Datasets:

Hendrik-a

/

arxiv

like 0

we describe the detonation mechanism comprising the `` pulsationally assisted '' gravitationally confined detonation ( gcd ) model of type ia supernovae ( sne ia ) . this model is analogous to the previous gcd model reported in ; however , the chosen initial conditions produce a substantively different detonation mechanism , resulting from a larger energy release during the deflagration phase . the resulting final kinetic energy andni yields conform better to observational values than is the case for the `` classical '' gcd models . in the present class of models , the ignition of a deflagration phase leads to a rising , burning plume of ash . the ash breaks out of the surface of the white dwarf , flows laterally around the star , and converges on the collision region at the antipodal point from where it broke out . the amount of energy released during the deflagration phase is enough to cause the star to rapidly expand , so that when the ash reaches the antipodal point , the surface density is too low to initiate a detonation . instead , as the ash flows into the collision region ( while mixing with surface fuel ) , the star reaches its maximally expanded state and then contracts . the stellar contraction acts to increase the density of the star , including the density in the collision region . this both raises the temperature and density of the fuel - ash mixture in the collision region and ultimately leads to thermodynamic conditions that are necessary for the zeldovich gradient mechanism to produce a detonation . we demonstrate feasibility of this scenario with three 3-dimensional ( 3d ) , full star simulations of this model using the flash code . we characterized the simulations by the energy released during the deflagration phase , which ranged from 38% to 78% of the white dwarf s binding energy . we show that the necessary conditions for detonation are achieved in all three of the models .

1202.3997

type ia supernovae ( sne ia ) are among the most energetic explosions in the known universe , releasing @xmath1 ergs of kinetic energy in their ejecta , and synthesizing @xmath2 @xmath3 of radioactive @xmath0ni . the discovery of the phillips relation @xcite has enabled the use of sne ia as standardizable cosmological candles which greatly enhances the accurate determination of their distance . the discovery of the accelerated expansion of the universe using sne ia @xcite has stimulated a tremendous amount of interest in the use of sne ia events as standard cosmological candles , allowing them to serve as probes of the equation of state of dark energy , as parameterized by the eos parameter @xmath4 . the main challenge to the enterprise of measuring @xmath5 using sne iais reducing the systematic errors in the accuracy with which such supernovae can be used as standard candles @xcite . the accuracy must be improved from the current level of about 15% to about 1% in order for large surveys to determine the values of @xmath6 and its rate of change with @xmath7 to better than 10% @xcite . the best hope for improvements in distance modulus accuracy is more accurate modeling of sne ia explosions . evidence suggests that sne ia are the results of the thermonuclear explosion of a carbon - oxygen ( c - o ) white dwarf ( wd ) . the leading scenario for sne ia explosions is the single - degenerate model in which a progenitor wd accretes material from a non - degenerate companion star until the mass of the wd grows nearly equal to the chandrasekhar limit @xcite . the wd then manages to release enough nuclear energy by fusing c and o into radioactive ni and other lighter @xmath8-elements in the time span of a few seconds or less @xcite , such that it deposits approximately @xmath9 ergs of energy , unbinding the star and accelerating the stellar material to speeds of thousands of kilometers per second @xcite . this rapid fusion process must proceed in two phases @xcite . the first phase begins with the initiation of a subsonic nuclear burning front ( referred to as a deflagration or flame ) . as the wd accretes material from its companion , convective carbon burn begins in its core . when carbon burning becomes too vigorous for convective cooling to be effective , a thermonuclear flame ( or deflagration ) is born in local hot spots that form in the convective region . the deflagration burns and rises due to buoyancy as it makes its way from the core to the outer layers of the star . the second phase consists of a supersonic burning front a detonation that consumes the remainder of the wd . the transition from the deflagration phase to the detonation phase is poorly understood , and has been the subject of much modeling research . a variety of models incorporates the scenario of a deflagration followed by a detonation , such as the deflagration - to - detonation transition ( ddt ) , the pulsating reverse detonation ( prd ) @xcite , or the gravitationally confined detonation gcd @xcite . these models differ primarily in the method by which the deflagration leads to a detonation . an alternative scenario to the single - degenerate model is the double - degenerate model in which two wds forming a binary system undergo a merger process that leads to the detonation of one of the wds . the double - degenerate channel @xcite has recently received revived attention from both observation ( e.g. , @xcite ) and theory ( e.g. , @xcite ) . however , while computational models have explored prompt detonations in near - equal mass super - chandrasekhar mergers and head - on collisions of binary wd systems @xcite it remains unclear whether more commonplace mergers between two typical c - o wds of masses @xmath10 will result in a type ia explosion . previous one - dimensional theoretical models suggest that such mergers will result in deflagration wave that sets off an accretion - induced collapse to a neutron star @xcite . in this work we focus on the gcd scenario of the single - degenerate model of sne ia . we use the flash code @xcite to extend the set of 3d whole - star gcd models in @xcite ( hereafter j08 ) to include multiple ignition points as initial conditions . these initial conditions provide more burning , and hence more energy release during the deflagration phase . the purpose of this work is to examine the conditions produced in the wd resulting from a scenario where more energy is released during the deflagration phase than in the previous gcd models and to demonstrate that the scenario manifests the necessary conditions for a detonation . to distinguish between the two versions of gcd models we refer to the previous body of gcd models ( @xcite , j08 , @xcite , and references therein ) as `` classical '' gcd models and refer to models in this paper as `` pulsationally - assisted '' gcd models or simply `` pulsational '' gcd ( pgcd ) models . the origin of the name stems from the fact that the wds in the models undergo a pulsation where they first expand due to energy input from the deflagration phase and then contract from the pull of gravity . these models require the contraction of the wd to create the thermodynamic conditions necessary to initiate a detonation , hence the phrase `` pulsationally - assisted '' . in fact , the paradigm we put forth has characteristics of both the classical gcd models and the prd models . in section [ sec : gcdrev ] we review the classical gcd model , introduce the pgcd model , and discuss the ignition of the deflagration . in section [ sec : det ] we discuss the zeldovich gradient mechanism as a detonation trigger and its implications for the pgcd . we additionally discuss the numerical detonation trigger used in our simulations . we give an overview of the flash code and the relevant physics modules used for our sne ia simulations in section [ sec : methods ] . in section [ sec : sims ] we describe the results of our simulations . we compare the detonation mechanisms between the classical gcd and the pgcd models in section [ sec : compare ] . finally , in section [ sec : discussion ] we discuss properties of the simulations as well as possible observational features of the models . the classical gcd model of sne ia falls under the general category of single - degenerate models . in the gcd scenario , an off - centered deflagration ignites and begins burning its way through the star . the deflagration burns and rises , forming a plume of ash whose volume is bounded by the flame . when the plume reaches the stellar surface , the ash breaks through and spreads around the star . the ash then collides with itself at a location on the star opposite that of breakout . during this collision process , cold , low - density fuel is pushed ahead of the ash flows . the ash flows compress and heat the fuel in the collision region . by squeezing the fuel in the collision region , a jet is formed . the jet pushes the hot , smoldering fuel towards the high density layers of the wd which leads to conditions necessary to trigger a detonation ( j08 , @xcite ) . the detonation occurs between 1.5 s and 3 s after ignition , which is about the time it takes for the ash to flow around the stellar surface . in the classical gcd scenario , the deflagration does not drive the wd to expand energetically due to the fact that the deflagration is so weak . typically , deflagration phase only release @xmath11 of the binding energy to the wd . the star has only mildly expanded by the time the ash flows have triggered a detonation , and in fact the star is still expanding when the detonation is triggered . additionally , as reported in j08 classical gcds produce more than 1 @xmath3 of @xmath0ni which correspond to over - luminous sne ia . in this work we introduce the pgcd scenario . the pgcd proceeds similarly to the classical gcd . an off - centered deflagration ignites and burns its way to the surface . the ash breaks through the surface of the star , some of which flows laterally over the surface towards the collision region . the pgcd differs from the classical gcd in the amount of energy released during the deflagration phase . in the classical gcd scenario , a relatively small amount of energy is released during the deflagration phase ; instead , substantially more is released during the deflagration phase in the pgcd scenario . the wd in the pgcd scenario responds to the deflagration by expanding rapidly . by the time the ash flows reach the collision region , the density has significantly dropped and the flows can not compress the fuel to the high temperatures and densities as in the classical gcd . thus , a detonation does not immediately occur as it did in the classical gcd . however , as the ash flowed towards the collision region , it mixed with fuel on the wd surface , forming a mixing layer at the interface between the ash and surface fuel . after the collision , the mixture pushes its way into the surface layers of the star . meanwhile , the wd has reached maximum expansion and begins to contract . as the wd contracts , it squeezes the fuel - ash mixture to high temperatures and densities . the conditions obtained are above the threshold necessary to trigger a detonation . thus , it is the stellar contraction , and not just the kinetic energy in the ash flows themselves ( as in the classical gcd ) , that gives rise to conditions that make it possible for the mixture to detonate . the simulations of the classical gcd model in j08 were initiated with a single ignition point that was offset from the center of the wd . these initial conditions lead to a weak deflagration which gives rise to the events described in section [ subsec : cgcdrev ] . we extended our set of initial classical gcd models by choosing initial conditions that would lead to more burning and thus releases more energy than we had previously obtained . to accomplish this we chose to initialize our simulations with a single cluster of multiple ignition points that was offset from the center of the wd . these initial conditions serve to represent two physical situations that may occur in the wd . first , it may be possible that ignition occurs at multiple points . recent work by @xcite and @xcite of the last moments of the smoldering phase before ignition conclude that a single ignition point is the most probable scenario ; however , ignition is an inherently stochastic process and many realizations are required to make definitive statements about all of the possible outcomes . currently only a handful of simulations exist . furthermore , the process of forming hot spots in the simulations that eventually lead to an ignition may be resolution dependent as stated in @xcite , and leaves open the door for a multiple ignition point scenario . finally , all but one of the simulations in @xcite and @xcite were stopped once a hot spot formed that lead to ignition . therefore , it may have been possible for more ignition points to form after the initial runaway , even though in the specific simulation that they continued past the formation of initial ignition point no secondary ignition points were observed . second , though we ignore the convective turbulence , most likely convective plumes exist with a velocity @xmath12 km / s @xcite . this is comparable to the flame speed and can potentially grow the flame surface through the kelvin - helmholtz instability . we essentially mock - up this effect by initializing our simulations with multiple ignition points in close proximity to one another . these ignition points in turn quickly merge and form a single asymmetric burning region enclosed by a distorted flame surface . in the pgcd model , a plume of ash , with a layer mixed with fuel , plunges into the star during its contraction phase and initiates a detonation . initially , as the plume advances around the star , a mixing layer of cold fuel and hot ash forms at the star - ash interface . compression due to the stellar contraction and to the flow pushing to higher density layers of the star acts to heat the mixture . the composition in the mixing layer transitions from pure , hot ash in the plume to pure , cold fuel in the star , resulting in a compositional and temperature gradient in the mixing layer . a spontaneous detonation is triggered by the zeldovich gradient mechanism @xcite when the mixture reaches the critical conditions . the zeldovich mechanism requires an induction time gradient resulting from a gradient in temperature or gradients in temperature and composition . rapid combustion in the high temperature region results in a shock wave as the burning front progresses to regions of low temperature . a detonation is formed if the gradients are shallow enough and the combustion becomes associated with the shock ; otherwise , the shock wave runs ahead of the burning front and the plasma does not detonate . for the purposes of this work , we wish to characterize the conditions under which a detonation would occur in the mixing layer that is formed when the ash plume flowed from its breakout point until it reaches the collision region . in the envisaged scenario , the mixing layer consists of a compositional gradient between pure ash in the plume and pure fuel on the surface of the wd , as well as a temperature gradient produced by combining the varying amounts of hot fuel and cold ash . in the following we discuss the critical conditions that are necessary to trigger a detonation in such a scenario . in particular , we concentrate on the possibility that the gradient mechanism would trigger a detonation on length scales that are under resolved or unresolved in our simulations , which corresponds to length scales @xmath13 km . much work has been done to determine the properties of the gradients required to successfully initiate a detonation by the zeldovich gradient mechanism @xcite . both @xcite and @xcite studied systems initialized with a temperature gradient that included a constant density and composition ( though the pgcd would also have a compositional gradient ) . @xcite examined the results of a linear temperature gradient . @xcite performed a more comprehensive study with several functional forms for the temperature gradient and a wider range of conditions . the goal of the work was to determine the minimum size of the region containing the induction gradient for a self sustaining shock - reaction complex to form . this minimum size ( or critical length scale , @xmath14 ) was determined to be a function of the input parameters of the system such as the ambient temperature , ambient density , peak temperature , steepness and functional form of the temperature profile , and composition of the fuel . in general , they find that for @xmath15 km , a composition of 50%-50% @xmath16c and @xmath17o with densities @xmath18 g/@xmath19 and peak temperatures @xmath20 k would successfully trigger a detonation . @xcite performed a resolution study on one of their models and examined how @xmath14 changed when they increased their resolution . in this study , the @xmath14 grew from @xmath21 m in the lowest resolution case to @xmath22 m in the highest resolution case ( see their table 3 ) . the grid resolution with which they ran their simulations may increase @xmath14 for a given parameter set , or said another way , the grid resolution will increase the critical conditions required to trigger a detonation for a particular @xmath14 . because they only performed the resolution study on one set of parameters it is difficult to asses how much the critical conditions would increase with adequate resolution for a given @xmath14 . @xcite also examined the conditions for the formation of a detonation for systems comprised of less than 50% @xmath16c by mass . in their most extreme case of carbon - poor fuel ( 30/70 c / o ratio ) the critical radius is roughly a factor of 10 larger than fuel composed of equal amounts of carbon and oxygen . in general , the reduction of @xmath16c increased the critical length scale ( their tables 8 - 11 ) . they also tested the effect of adding @xmath23he ( 14% by mass ) to the composition . the addition of helium dramatically decreased the critical length scale ( their table 12 ) . as with the resolution study , they only examined one set of parameters for their composition study , and thus did not determine how changing the composition with a fixed @xmath14 would change the values of peak temperature , density , etc . , required for a successful detonation . we note that by lowering the @xmath16c mass , the critical conditions would most likely would increase , and for extremely carbon - poor environments a detonation may not be triggered at all . for the remainder of this work we assume that extremely carbon - poor environments are not encountered and that small to moderate deviations from a 50/50 c / o ratio would not substantially increase the critical temperature and density at least to within the context of our approximation . @xcite examined the conditions under which the zeldovich gradient mechanism would produce a detonation given a gradient in both temperature and composition . they study two cases : in the first case turbulence tears apart an active flame and mixes cold fuel with hot ash to achieve the necessary conditions to initiate a detonation . in the second case , stellar expansion extinguishes the active flame allowing the hot ash and cold fuel to mix . as the star contracts , it squeezes the mixture - creating the necessary conditions for the mixture to detonate . @xcite argue that mixing is important in order for fuel to attain a high enough temperature to ignite at relatively low densities . they point out that cold fuel that obeys a wd equation of state would have to be compressed to @xmath24 g/@xmath19 in order to obtain a high enough temperature to ignite ; however , mixing hot ash at a level as low as 10% raises the entropy enough that the mixture would ignite at densities @xmath25 g/@xmath19 . their figure 2 shows the critical density for ignition as a function of fuel mass fraction in the mixture . in @xcite , a temperature gradient is set up in the mixing layer between cold fuel and hot ash . the temperature of the mixture is low in regions dominated by fuel and is high in regions dominated by ash . it is thus a matter of compressing the mixture to the critical density ( and thus the critical peak temperature through compressional heating ) for a detonation to initiate in the mixture . their figure 6 shows @xmath14 as a function of critical density for a gas initialized with a linear compositional gradient and a temperature gradient determined by the amount of ash in the mixture at a particular location . they conclude that in the wd environment , a detonation is possible for densities between @xmath26 g/@xmath19 and @xmath27 g/@xmath19 . for a mixing layer with a critical length scale on the order of 10 km , the critical density of the fuel is @xmath28 g/@xmath19 . interestingly , this is commensurate with the results of @xcite with a 50/50 c - o composition , even though the two approaches were distinct . we therefore conclude that for @xmath29 km , the critical density , @xmath30 g/@xmath19 and with a corresponding critical peak temperature , t @xmath20 k are the necessary criteria to successfully trigger a detonation on that length scale . we remind the reader that we have assumed that the environment is not carbon - poor and that a decrease in the fuel mass will not dramatically increase the critical conditions . we have further assumed that though the method used to determine the critical values was unresolved , the results of a fully resolved study would not dramatically increase the critical values from the unresolved case . though both the composition effects and uncertainties with the resolution of the method will affect the values critical for detonation , the values of temperature and density that we have selected are acceptable for our level of approximation . the pgcd detonates in the same manner as the second case described in @xcite , which depends on the formation of a fuel - ash mixture . in the pgcd , ash ejected from the star flows over the surface to the collision region . as the ash flows , it turbulently mixes with fuel on the surface forming a mixing layer of fuel and ash . in the following we demonstrate the feasibility of our mixing assumption by examining the growth of kelvin - helmholtz modes at the fuel - ash interface and showing that these modes can grow quickly enough and on the relevant length scales to provide the appropriate mixing . however , mixing is a complicated process and our assumptions below of constant physical conditions are meant to provide an order - of - magnitude estimates . we are particularly interested on the growth of the mixing layer on unresolved scales . as above , our length scale of interest is @xmath29 km . using typical values from the conditions present at the surface - ash interface , we can approximate the largest unstable kelvin - helmholtz length scale , @xmath31 , such that all @xmath32 would grow due to the instability . using linear stability analysis , the largest unstable length scale is expressed in terms of the physical conditions at the interface between the flowing ash and the surface of the wd as @xmath33 where @xmath34 , @xmath35 , @xmath36 is the fuel density ( i.e. the surface density of the wd ) , @xmath37 is the ash density , @xmath38 cm / s@xmath39 is the acceleration due to gravity , and @xmath40 is the velocity of the ash @xcite . the ash flows at a speed @xmath41 cm / s and the interface is located at @xmath42 cm from the center of the star . values for the densities are @xmath43 g/@xmath19 and @xmath44 g/@xmath19 . inserting these values into equation . ( [ eq : kh ] ) gives @xmath45 cm - comparable to the radius of the star and hence @xmath46 @xmath14 . therefore a region the size of @xmath14 would develop kelvin - helmholtz instabilities which would drive turbulent mixing . the ash must flow approximately half way around the star to reach the collision region . using the values of @xmath40 and @xmath47 we can approximate the time it would take for the ash to reach the collision region , @xmath48 , as @xmath49 therefore @xmath50s . note that this is consistent with the results of our simulations discussed in section [ subsec : simnumbers ] and shown in figure [ fig : thetaashvstime ] . the timescale for the growth of the unstable kelvin - helmholtz modes with length scale @xmath51 @xcite in the linear regime is @xmath52^{-1 / 2}. \label{eq : tkh}\ ] ] the growth time scale for a perturbation of length scale @xmath53 km is @xmath54 s. since @xmath55 , perturbations of length scale @xmath14 will grow for several ( linear analysis ) e - folding times before the ash reaches the collision region . we again note that the conditions at the ash - star interface are not constant , especially since the star is expanding as the ash flows over the surface . however , given the short growth timescale ( @xmath56 s ) for @xmath57 km , if the above conditions exist for only a small fraction of the time that the ash is flowing over the surface then the requisite modes would grow and mixing would develop by the time the ash reaches the collision region . we conclude that it is possible for the zeldovich mechanism to trigger a detonation in the mixed fuel - ash layer on length scales not adequately resolved in our simulation if the critical conditions are reached . the ash flows turbulently mix with fuel on the surface driven by kelvin - helmholtz instabilities . the mixing process produces the necessary thermal and compositional gradients and the increase in density of the mixture as it pushes into the star both accelerates the combustion and shrinks the critical length scale , @xmath14 , required for the gradients . to treat the detonation physics we have incorporated the numerical scheme used by @xcite and @xcite which we briefly describe in section [ sec : methods ] and which was not present in j08 . to correctly capture the zeldovich mechanism in the simulation we would need to resolve the required gradients as well as the carbon burning length scales within those gradients . this is prohibitively expensive for 3d simulations and we do not accomplish it here . thus the initiation of a detonation in our simulations is a process that is severely under resolved . the finest resolution in our simulation does not allow us to follow the simulation on length scales of @xmath13 km . in section [ subsec : theory ] we estimate that if a gradient is characterized by a peak temperature , t @xmath20 k and density , @xmath30 g/@xmath19 on length scales of @xmath58 km then the necessary conditions exist to produce a gradient - triggered detonation . we further show that mixing could occur on length scales on the order of @xmath58 km as the ash flows around the star . thus we adopt the criteria that if a computational cell s temperature and density exceed those specified above , then the necessary conditions exist for a gradient - triggered detonation and we allow the cell to detonate . numerically , we trigger the detonation by allowing the @xmath16c - @xmath16c reaction to run away . when a computational cell reaches a high enough temperature and density such that the carbon burning reaction rate becomes extremely high , all of the fuel in the cell burns in a single computational time step . if there is enough fuel in the cell , this releases a tremendous amount of energy , which dramatically increases the temperature and pressure in the cell , and creates a shock with the neighboring cells . the shock propagates outwards from the cell and burns material in its wake , forming a detonation . this is akin to the direct initiation of a detonation by a blast wave . since the initiation of the detonation is purely numerical in nature , we stress that it simply indicates that a computational cell has reached the necessary thermodynamic conditions for rapid carbon burning . however , the conditions for rapid carbon burning also coincide with the necessary ( but not sufficient ) conditions that we expect would enable the zeldovich mechanism to produce a detonation as outlined in section [ subsec : theory ] . we refer the reader to @xcite and @xcite for a discussion of the initiation of detonations in the context of numerical simulations . in our description of the detonation mechanism , we appeal to the fact that the ash ejected from the wd will mix with fuel on the surface of the star as it flows ; however , we do not treat mixing on unresolved scales . mixing on small scales occurs by numerical diffusion . though we do not capture the details of that mixing process , we fully expect the material to be well mixed as it flows around the star as postulated in section [ subsec : theory ] . finally , due to the above mentioned physical processes that are not included in our simulations , the physical quantities that make up the trends discuss in section [ sec : sims ] and listed in table [ tab : sims ] that are the results of the detonation should be taken with a grain of salt . these properties ( such as the radius at which the detonation occurs for example ) are the result of the time at which the numerical detonation is triggered and the state of the contracting wd at that time . though the conditions for which the numerical detonation is triggered roughly coincides with those that are sufficient to trigger a gradient - induced detonation , by approximating the relevant physics the exact time and location of the detonation are uncertain . altering the detonation time and location could change the physical properties resulting from the detonation . nevertheless because of the correspondence between the conditions for the numerical detonation and that of the physical detonation , we expect the general trends presented in section [ sec : sims ] to hold while we acknowledge that there may be large uncertainties . we use the adaptive mesh refinement ( amr ) flash application framework @xcite to perform the simulations presented in this paper . flash is a modular , component - based application code framework created to simulate compressible , reactive astrophysical flows . the framework supports multiple methods for managing the discretized simulation mesh , including the paramesh ( parallel adaptive mesh refinement ) library @xcite , which implements a block - structured adaptive grid . flash includes a directionally split piecewise - parabolic method ( ppm ) solver @xcite descended from the prometheus code @xcite . it has been successfully applied to a wide variety of large - scale terrestrial and astrophysical flow problems , ranging from simulations of homogeneous , isotropic turbulence @xcite , raleigh - taylor instability @xcite , shock - cylinder interaction , and laser - driven high energy density laboratory experiments @xcite , to buoyancy - driven turbulent combustion @xcite , wind - driven instabilities in neutron star atmospheres @xcite , contact binary stellar evolution @xcite , type ia supernovae @xcite , galaxy collisions @xcite , and cosmological simulations of large - scale structure formation @xcite . the energetics scheme employed to treat flames and detonation waves in our simulations uses three progress variables to track carbon burning , qse relaxation , and nse relaxation . details concerning the nuclear physics and the numerical implementation are presented in @xcite and @xcite . subsonic burning fronts ( deflagrations ) are advanced using an advection - diffusion - reaction ( adr ) equation . details concerning the implementation , calibration and noise properties of the flame treatment can be found in @xcite and references therein . detonations are handled naturally by the reactive hydrodynamics solver in flash without the need for a front tracker . this approach is possible because unresolved chapman - jouguet ( cj ) detonations retain the correct jump conditions and propagation speeds . cellular structure smaller than the grid scale will be suppressed in our simulations but is free to form on resolved scales . the impact of cellular structure on the global evolution of the model is still uncertain ; however , since cellular structure alters the detonation wave speed by only a few percent for the conditions being modeled @xcite , the effect is likely to be small . additional details related to the treatment of detonation waves are discussed in @xcite . self gravity is calculated by expanding the mass density field in multipole moments , which are used to approximate the scalar gravitational potential . the gravitational acceleration is calculated by approximating the derivative of the scalar potential at each location in the domain . the helmholtz equation of state of @xcite is used to describe the thermodynamic properties of the stellar plasma , including contributions from blackbody radiation , ions , and electrons of an arbitrary degree of degeneracy . we extended the study of the gcd model described in j08with two primary differences . first , we included detonation physics in the simulations ( as we described in sections [ subsec : numreal ] and [ sec : methods ] ) and followed the models from ignition , through the detonation phase , and to the free expansion phase . we terminated the simulations when the temperature decreases to the point that nuclear reactions ceased . second , we initiated these simulations with multiple ignition points ( as described in section [ subsec : ics ] ) instead of a single ignition point with the method used in j08 . for completeness we give the basic points here but refer the reader to j08 and references therein for more details . our simulations used flash s amr capabilities with a finest resolution of 8 km . each simulation contained a 1.365 @xmath3 wd in hydrostatic equilibrium composed of equal parts carbon and oxygen . the wd had a central density of @xmath59 g/@xmath60 , an ambient temperature of @xmath61 k , a radius of approximately @xmath62 km , and a binding energy of @xmath63 ergs . although the core of the star is most likely convective with a turbulent convective rms velocity @xmath64 km / s @xcite , the turbulent convective rms velocity is much less than the laminar flame speed , and so we have ignored the background convective turbulence , and initialized the star with zero velocity . we initiated this series of simulations with multiple ignition `` points '' , which are @xmath65 km spheres comprised of hot ash placed in the star at rest . we placed the ignition points in a spherical region whose center coincided with the z - axis . we parameterized the spherical region by its radius and z - offset ( the distance from the origin of the center of the spherical region along the z - axis ) . the radius of the spherical region was @xmath66 km and the z - offsets were @xmath67 km , @xmath68 km , and @xmath69 km - one z - offset for each of our three simulations . we randomly populated a @xmath66 km spherical region with @xmath70 ignition points . we placed ignition points so that they did not overlap with one another , and so that they were contained entirely within the spherical region . we used the same random distribution of ignition points for all three simulations ; they only differed by their relative location along the z - axis . table [ tab : sims ] contains labels for the simulations referred to in this work along with parameters for the initial conditions . in this section we describe the evolution of the pgcd model as demonstrated by the @xmath71 simulation . figures [ fig : gcdpbrads1 ] and [ fig : gcdpbrads2 ] show snapshots of the @xmath71 simulation starting with the initial conditions and ending with the passage of the detonation wave . the green contour in the figures approximates the location of the wd s surface . the orange regions are high temperature regions . these regions are primarily hot ash . the temperature ranges from @xmath72 k ( dark orange ) to @xmath73 k ( bright orange ) . each simulation begins with one of the above multiple ignition point configurations . panel ( @xmath74 ) of figure [ fig : gcdpbrads1 ] shows the initial conditions of the @xmath71 simulation . in the first few tenths of a second , each ignition point quickly burns radially outward from its center and merges with the other ignition points to form a large asymmetric bubble of ash . the ash in the bubble is less dense than the surrounding stellar material and is therefore buoyant . after 0.3 - 0.4 seconds , the bubble , which continues to grow in size , begins to quickly rise towards the wd s surface . during the rise , rayleigh - taylor instabilities grow on the flame surface , which further increases the surface s complexity and enhances the burning rate @xcite . panel ( @xmath75 ) of figure [ fig : gcdpbrads1 ] shows the simulation at 0.6s . this panel shows the ignition points after they have risen and merged to form a complex volume whose surface has been enhanced by fluid instabilities . after approximately 1.5 s , the rising ash bubble breaks through the stellar surface and begins spreading laterally across the star in all directions , converging on a region antipodal to the bubble breakout region . since the density of the surface layers of the star are too low to maintain thermonuclear combustion , the flame quenches and the deflagration subsides . as the ash flows over the surface of the star , cold fuel ( c and o ) mixes with the ash at the interface between the ash and the stellar surface , creating a fuel - ash mixture . panels ( @xmath76 ) , ( @xmath77 ) , and ( @xmath78 ) of figure [ fig : gcdpbrads1 ] are at 1.1 s , 1.4 s , and 1.9 s , respectively . these panels show the ash after it has broken out of the surface and has started to spread . in panel ( @xmath76 ) the ash has spread approximately half way around the star . by 1.4 s ( panel @xmath77 ) , the ash has made it three - fourths the way around the star , and at 1.9 s ( panel @xmath78 ) the ash has almost completely engulfed the wd . a significant amount of nuclear energy , @xmath79 ergs ( comparable to the binding energy of the wd ) , is released during the deflagration . this provides a kick to the star which causes it to expand . during the first few seconds , while the ash flows over the surface , the star expands to several times its original size . the expanding star slows the ash before it reaches the collision region . this shown in panels ( @xmath76 ) - ( @xmath80 ) of figure [ fig : gcdpbrads1 ] and panel ( @xmath74 ) of figure [ fig : gcdpbrads2 ] . panel ( @xmath80 ) of figure [ fig : gcdpbrads1 ] is at 2.2 s and panel ( @xmath74 ) of figure [ fig : gcdpbrads2 ] is at 2.5 s. the ash seems to disappear from the panels because the star has expanded and the ash has cooled to the point that it has fallen out of the color range . panel ( @xmath75 ) of figure [ fig : gcdpbrads2 ] is at 3.1 s. the wd is maximally expanded in this image . several seconds after the ash is ejected from the surface of the star , it converges at the opposite pole from where is broke out . the converging flow compresses the material in the collision region until the pressure is sufficient to stop the mixture from flowing laterally and to drive a plume of fuel and ash towards the interior of the star . additionally , after the star has reached its maximally expanded state , it begins to contract . the stellar contraction increases the global density structure of the star , and thus the local density of at the collision region , eventually resulting in explosive c and o burning . once the conditions for detonation are met ( as discussed in section [ subsec : theory ] ) and the detonation is triggered ( as discussed in section [ subsec : numreal ] ) the combustion immediately sweeps across the star in a few tenths of a second . panels ( @xmath76 ) and ( @xmath77 ) of figure [ fig : gcdpbrads2 ] , at 3.7 s and 4.2 s respectively , show the star as it contracts . a hot region can be seen forming at the `` bottom '' of the star . this hot region is the result of the compressional heating of the mixture in the collision region from the work done by the contracting star as well as the material flows themselves . panel ( @xmath78 ) of figure [ fig : gcdpbrads2 ] is at 4.3 s and shows the simulation just after the initiation of the detonation wave . panel ( @xmath80 ) of figure [ fig : gcdpbrads2 ] shows the simulation at 4.4 s. the detonation wave has consumed about half of the star in just over 0.1 s. the detonation wave transforms the wd into a super - heated remnant composed of material that has reached nuclear statistical equilibrium ( nse ) in the core ( which is dominated by iron - group elements , most of which is @xmath0ni ) surrounded by a layer of intermediate mass elements and topped of with the ash of partially burned c and o. this whole structure is shrouded by the products of the deflagration which consisted of iron - group elements , intermediate mass elements , carbon burning products , and unburned c and o. the super - heated structure quickly expands and cools , and is transitioning into a homologous structure when the simulations are stopped . in each of our simulations , we found the necessary conditions to trigger a detonation as discussed in section [ sec : det ] . the main properties of the three simulations are summarized in table [ tab : sims ] . the simulations differ in the location of the center of their sphere of ignition points , and are labeled @xmath81 ( ignition sphere centered 68 km from the center of the star ) , @xmath82 ( 88 km from the center of the star ) , and @xmath71 ( 168 km from the center of the star ) . the deflagration phase of the sn ia provides a kick to the wd which causes the wd to rapidly expand and then contract . it is the expansion and subsequent detonation on contraction that characterizes the pulsation gcd model . figure [ fig : enucvstime ] shows the amount of nuclear energy release as a fraction of the binding energy of the star as a function of time for the three simulations . the simulations trend together for the first 0.8s and then diverged . the deflagration continued to release nuclear energy until approximately 1.5s in each simulation , at which time the flame stopped burning . in general , the simulations whose ignition points were placed closer to the core of the wd burned more during the deflagration and thus imparted more energy to the star . the amount of energy released ranged from @xmath83 ergs to @xmath84 ergs , or between 38% and 78% of the binding energy of the star . by comparison , the single bubble initial conditions in j08 released between @xmath85 ergs gand @xmath86 ergs , or between 6% and 21% of the binding energy of the wd . the energy released during the deflagration phase was sufficient to cause the star to undergo an energetic pulsation . figure [ fig : rhocvstime ] shows the evolution of the maximum density found in the computational domain , which is equivalent to the central density of the wd ( prior to detonation ) for the three simulations and gives a measure of the strength and the period of the pulsation . in each simulation , the wd expanded to a maximum amplitude and then contracted before detonating . the simulation that expanded the least was @xmath71 which reached a minimum central density , @xmath87 , of @xmath88 g/@xmath19 . the star expanded for 3.1 s before it began to contract . by contrast , @xmath81 achieved @xmath89 g/@xmath19 , and the wd did not begin to contract until 4.71 s. thus the more energy released during the deflagration phase , the more the wd expands and the longer its pulsational period . the vertical lines in figure [ fig : rhocvstime ] mark the time at which the star detonates . as the wd expanded , ash was ejected from the surface and flowed laterally over the star , mixing with surface fuel as it flowed . figure [ fig : thetaashvstime ] plots the polar angle of the leading edge of the flow as a function of time . initially a small region of ash crossed the origin and was responsible for the large values of @xmath90 seen in the figure during the first second . after one second , the ash reached the surface and started to spread around the star . the ash , which mixes with fuel on the surface as it flows , quickly reached a polar angle of approximately @xmath91 and then stalled in each of the simulations . this was in part because the mixture pushed some material in front of it which compressed material in the collision region and increased the pressure there . also , the expansion of the star robbed kinetic energy from the flow which also contributed to the mixture stalling . once the star contracted , the mixture slowly pushed its way further around the star , as well as into the high density regions towards the core . as before , the lines on the graph highlight the time at which the star detonated . the mixture only stalled for a short period of time in the @xmath71 simulation whose wd had the shortest pulsational period . the opposite is true for the @xmath81 simulation in which the mixture stalled the longest and the wd had the longest pulsational period . even though the mixture encroached on the collision region within a few seconds from when the flow started , it was the contraction of the star that assisted the mixture in making the final move to higher densities . after the star began to contract , the fuel - ash mixture made its way into the high - density layers of the wd . figure [ fig : maxtempvstime ] shows the evolution of the thermodynamic conditions in the fuel - ash mixture as it pushed into the star . the top plot in figure [ fig : maxtempvstime ] shows the temperature of the computational cell with the maximum temperature in the mixture . the middle plot shows the density and the bottom plot shows the radius of the computational cell with the maximum temperature . both the temperature and density follow the trend of the central density , as the wd expanded and contracted . as the star expanded , the temperature and density of the mixture decreased . as the star contracted , the temperature and density of the mixture increased . the location of the hot spot followed the same pattern ; hence , as the star contracted , the radius of the hot spot moves closer to the core . this was due both to the stellar contraction and the continued flow of the mixture . once the temperature of the hot spot exceeded @xmath92 k and the density exceeded @xmath93 g/@xmath19 rapid combustion ensued , which transitioned into a detonation . we stress that the detonation is the result of the numerical scheme discussed in section [ subsec : numreal ] and is not the result of the zeldovich gradient mechanism . it merely indicates that the conditions in the computational cell surpass those that are necessary to produce a gradient - induced detonation . in the @xmath71 simulation , the deflagration phase released the least amount of energy of all the simulations and thus the wd expanded the least , it had a shortest pulsational period , it began contracting the earliest . as a result , the detonation occurred the soonest . the opposite is true for the @xmath81 simulation whose deflagration phase released the most amount of energy and whose detonation was triggered last . an important set of observables of an sn ia are the nucleosynthetic yields produced in the event . we have approximated the yields from the three models using the reaction progress variables from flash s burning module @xcite . table [ tab : det ] lists the post - explosion nucleosynthetic yields from our simulations in terms of the quasi - static equilibrium groups that are represented by the progress variables . note that the material that burned to nse is predominately iron - group elements , most of which is @xmath0ni . thus , the amount of nse material can be considered an upper limit on the amount of @xmath0ni produced during the explosion . the upper limit on the @xmath0ni produced during these simulations ranges from @xmath94 @xmath3 to @xmath2 @xmath3 . though yields of @xmath94 @xmath3 of @xmath0ni are associated with luminous sne ia , yields of @xmath2 @xmath3 of @xmath0ni are associated with sne ia of normal luminosity . the combined mass of the intermediate mass elements and nse material is @xmath95 @xmath3 and is consistent with observed sne ia @xcite . the deflagration phase of the pgcd models burns approximately 0.1 - 0.25 @xmath3 of material . this material is ejected into the outer regions of the system , surrounding the wd when it detonates . table [ tab : def ] lists the nucleosynthetic yields produced during the deflagration phase of the three simulations . the primary product of the deflagration phase is nse material with about a third of the material composed of carbon burning products and intermediate mass elements . these simulations show that through the course of an off - centered ignition , if a detonation is not triggered as the flame breaks down when it moves through the low density layers of the star on its way to the surface ( i.e. the first scenario in @xcite ) and if the energy release during the deflagration is sufficient to significantly disrupt the star and cause it to rapidly expand , ash will flow around the surface ( mixing with cold fuel as it flows ) , and stall in the collision region while the star is expanding . as the wd contracts , the conditions in the fuel - ash mixture are pushed to higher temperature and density until the values exceed that which are necessary to produce a gradient - triggered detonation . once the detonation is triggered in the mixture , it quickly consumes the star . the classical gcd model of sne ia is postulated to be possible if the energy released during the deflagration phase does not significantly expand the star . with little stellar expansion , the fuel density in the collision region is high enough for compression by the in - flowing ash to increase the temperature in the fuel and ultimately initiate a detonation . for the purposes of demonstrating this , we restarted the @xmath96 simulation from j08 just prior to detonation with the detonation physics included in simulation . figure [ fig : gcdcdet ] shows the evolution of the @xmath96 classical gcd model leading up to the detonation . the top left panel of the figure is at t=1.8s . it shows the ash as it approached the collision region . fuel pushed ahead of the ash increased in temperature due to compressional heating . this hot region formed a jet that flowed both towards and away from the surface of the star . the top right panel is at t=1.9s . the ash converged further into the collision region . the hot fuel continued to increase in temperature and had begun to slowly burn carbon as a result . this smoldering also increased the fuel temperature . the bottom left panel is at t=2.22 s. at this time the ash flows have collided . the hot fuel has further increased in temperature due to further compressional heating and due to combustion . the head of the hot jet has pushed the high temperature region and has reached the higher density layers of the white dwarf . the sufficient thermodynamic conditions to produce a detonation have been met and a detonation was triggered in the hot smoldering material . the formation of the detonation can be seen at the head of the hot jet . finally , the bottom right panel at t=2.3s shows the simulation after the detonation occurred . the detonation wave is propagating outward from its initiation point . the high temperature region behind the detonation as well as the smooth detonation front is visible in this image . the pgcd model of sne ia is possible if the deflagration phase releases enough energy to cause the wd to expand significantly , but not so much energy that the star becomes unbound . figure [ fig : gcdpdet ] shows the evolution of the pgcd model , @xmath71 , leading up to the detonation . the top left panel is at t=3.5s where the mixture was beginning to approach the collision region . in contrast to the classical gcd , the ash flows and collision region were at low densities and no significant compressional heating of the fuel in the collision region occurred . the top right panel is at t=3.8s . note that this panel is zoomed with respect to the previous panel . the fuel - ash mixture continued to push into the collision region . the mixture began to heat since the wd was contracting and the ash was pushing to higher densities towards the core of the wd . the bottom left panel is at t=4.1s . note that this panel is again zoomed in from the previous panel . the wd has continued to contract and the mixture continued to move to higher densities . the mixture heated significantly due to 1 ) compressional heating from the mixture pushing to higher densities and from the wd contracting and 2 ) the mixture reaching conditions that caused it to smolder . the mixture then approached the critical density above which a detonation would be initiated . the bottom left panel is at 4.22s note that , again , this panel is zoomed in from the previous panel . the mixture reached the necessary thermodynamic conditions for a gradient - triggered detonation . the detonation was triggered in the simulation as described in section [ subsec : numreal ] . two detonation waves can be seen emerging from the mixture layer in the image . the bottom right panel is at t=4.3s . the detonation is propagating outward from its initiation point through the star . the high temperature region behind the detonation wave as well as the smooth detonation front are clearly visible . in summary , the classical gcd model detonates as the wd is expanding . the thermodynamic conditions for detonation are reached when the ash flows compress and heat fuel in the collision region and force this hot fuel into the high density layers of the star . in contrast , the pgcd model detonates as the wd contracts . the thermodynamic conditions for detonation are achieved by the combination of the contraction of the wd which increases the temperature and density of the fuel - ash mixture , as the mixture pushes its way towards the high density core of the wd . in the classical gcd model , flowing ash must compress fuel in the collision region to achieve the thermodynamic conditions necessary to launch a detonation . this would seem to require the ash flows to converge symmetrically so that they focus on a localized region . thus far , the only successful classical gcd simulations have been performed in 2d cylindrical geometry ( @xcite for example ) which imposes azimuthal symmetry , and in 3d with single bubble ignition points as initial conditions which , by their nature , evolve in an azimuthally symmetric manner . our pgcd models were initiated with multiple ignition points . these ignition points evolved into an asymmetric volume of ash which produced asymmetric ash flows on the surface . however , the pgcd depends less stringently on the details of the ash flows . as long as the mixture is redirected towards the core of the star , the contraction of the wd will do the rest , even if the flow lacks symmetry . as a result , the pgcd model , in principle , is a more robust detonation mechanism compare to the classical gcd scenario in that it only depends on getting the fuel - ash mixture to a high enough density as the star contracts . for any sn ia model , the luminosity of the model is strongly related to the amount of @xmath0ni produced in the explosion . this depends in turn on the density of the plasma before it is burned . a fair rule of thumb is that if @xmath97 g/@xmath19 then that material will burn to nse , which is predominantly iron - group elements comprised mostly of @xmath0ni ( see e.g. , the comparison made in figure 12 of @xcite and the discussion in their section 5 ) . the wd in the pgcd model is contracting from an expanded state when it detonates . as a result , there is a reduced amount of high density material . it is therefore in principle possible for the model to produce range of abundances of @xmath0ni - and thus of luminosities - depending on the expanded state of the wd when it detonates . the classical gcd models described in j08 detonate as the wd is expanding after a weak deflagration phase . the star is still fairly compact and there is ample high density material when the star detonates . the amount of @xmath0ni produced in each of those models is greater that 1 @xmath3 which corresponds to overly luminous sne ia . by contrast , the pgcd model is more easily able to access regions in model space that correspond to the 0.7 @xmath3 values of @xmath0ni production typical of normal sne ia . furthermore , the production of intermediate mass elements between the two scenarios is comparable . the classical gcd models in j08 produced between 0.1 @xmath3 to 0.36 @xmath3 of intermediate mass elements whereas the pgcd produces between 0.22 @xmath3 and 0.33 @xmath3 of intermediate mass elements when adding the contributions of @xmath98 mg and si - group elements . the pgcd model has a post detonation geometry similar to the classical gcd model . the explosion is the result of a single off - centered detonation . the location of the detonation is at the antipodal point from where the star ejected the ash from its interior . this confers approximate azimuthal symmetry upon the system . the degree of asymmetry is determined in part by the radial offset of the detonation trigger . the larger the offset , the higher degree of asymmetry . we list the radial offset of the detonation location in our simulations in table [ tab : sims ] . in the pgcd model , the star detonates after the cessation of the deflagration phase . there is no active flame in the interior of the star at the time of detonation and thus no compositional inhomogeneities from deflagration ash . when the detonation occurs , it produces a smooth interior of nse material surrounded by a layer of intermediate mass elements . the intermediate mass elements are surrounded by a shroud of ash produced in the deflagration , which is a mixture of intermediate mass elements and nse material . this would suggest that elements , such as fe , would be present in the high - velocity , outer layers of the ejects and appear as such in the spectra . the flow of this ash over the surface of the wd produces a clumpy , choppy outer boundary to the intermediate mass element layer of the remnant . this is similar to structure suggested by the spectropolarimetry measurements of @xcite . furthermore , very early spectra taken at @xmath99 day after the explosion from the sn ia 2011fe show that there is o , mg , si , s , ca , and fe - which are products of the deflagration in our models - in the outer most layers of the remnant @xcite . finally , table [ tab : sims ] lists the post explosion kinetic energy for the three simulations . these energies are @xmath100 ergs and are consistent with observations of sne ia . it is interesting to compare our results to the results from the 3d simulations in @xcite . they initialize their simulations with a single spherical bubble , a region of small bubbles emulating a single bubble with surface perturbations , and a configuration of bubbles confined to a tear - drop - shaped envelope . of the six 3d simulations they performed , two of the wds in the simulation became unbound due to the energy released during the deflagration phase . of the four simulations in which the wd remained bound , the deflagration energy release was in the range of @xmath101 @xmath102 ergs , or roughly @xmath103 20% 60% of the binding energy of the wd . furthermore , one of their single - bubble ignition model released @xmath104 ergs comparable to our @xmath82 simulation ( their other single bubble ignition model disrupted the star ) . these values of the energy released during the deflagration phase are similar to those presented in our work . they found that the conditions for detonation are not reached in any of their simulations ; however , they stopped their simulations when the conditions for detonation are not met in the scenario presented in j08 . the further evolution of their simulations would have been interesting in light of the results presented here . we have conducted a series of numerical experiments simulating the evolution of a wd star in which we initiate a deflagration at off - center ignition points . the amount of energy released during the deflagration phase is enough to cause the star to undergo an energetic pulsation . as in the classical gcd model , the off - center ignition leads to a plume of material that is ejected from the star , flows laterally over the stellar surface , and converges on a collision region at the antipodal point from the ash breakout point . as the wd contracts , it creates thermodynamic conditions in the collision region such that it is possible for the zeldovich gradient mechanism to trigger a detonation . we find that these necessary ( but not sufficient ) conditions for detonation are reached in all three of our models . the energy deposition from the deflagration phase in these models ranges from 38% to 78% of the binding energy of the wd . we further comment that the detonation mechanism in the pgcd depends only on the bulk fluid motion of the system after the deflagration is ignited as opposed to a specific flow pattern , such as the focusing of the ash flows in the collision region in the classical gcd . this property confers a measure of robustness to the detonation mechanism . finally , the features of the pgcd qualitatively agree with the observations of sne ia , insofar as such comparisons can be made without subjecting the remnant to a radiation transfer treatment in order to compute actual light curves . the three models produced upper limits on the yields of @xmath0ni ranging from 0.72 @xmath3 to 0.98 @xmath3 . these @xmath0ni yields are less than those produced in the classical gcd models of j08 , primarily because the wd in our models is in a more expanded state at the time of detonation and contains a lower abundance of high - density , nse - producing matter . this class of models allows sne ia to evolve and detonate with characteristics similar to the classical gcd while producing supernovae of lower luminosity . the authors thank the flash code group , especially chris daley and anshu dubey for help with the development of and troubleshooting the code . we thank brad gallagher for creating figures [ fig : gcdpbrads1 ] and [ fig : gcdpbrads2 ] . we also thank katherine riley , mike papka , and the staff at the argonne leadership computing facility at argonne national laboratory for help running our large - scale simulations on intrepid at anl . this work was supported in part at the university of chicago by the u.s department of energy ( doe ) under contract b523820 to the asc alliances center for astrophysical nuclear flashes , and in part by the national science foundation under grant no . ast - 0909132 for the `` petascale computing of thermonuclear supernova explosions '' . this research used computational resources awarded under the incite program at alcf at anl , which is supported by the office of science of the us department of energy under contract no . de - ac02 - 06ch11357 . , a. , benzi , r. , berg , j. , biferale , l. , bodenschatz , e. , busse , a. , calzavarini , e. , castaing , b. , cencini , m. , chevillard , l. , fisher , r. t. , grauer , r. , homann , h. , lamb , d. , lanotte , a. s. , lvque , e. , lthi , b. , mann , j. , mordant , n. , mller , w. , ott , s. , ouellette , n. t. , pinton , j. , pope , s. b. , roux , s. g. , toschi , f. , xu , h. , & yeung , p. k. 2008 , physical review letters , 100 , 254504 , a. c. , fryxell , b. , plewa , t. , rosner , r. , dursi , l. j. , weirs , v. g. , dupont , t. , robey , h. f. , kane , j. o. , remington , b. a. , drake , r. p. , dimonte , g. , zingale , m. , timmes , f. x. , olson , k. , ricker , p. , macneice , p. , & tufo , h. m. 2002 , , 143 , 201 , g. c. , meakin , c. a. , hearn , n. , fisher , r. t. , townsley , d. m. , lamb , d. q. , & truran , j. w. 2009 , in astronomical society of the pacific conference series , vol . 406 , astronomical society of the pacific conference series , ed . n. v. pogorelov , e. audit , p. colella , & g. p. zank , 92+ , p. e. , sullivan , m. , cenko , s. b. , thomas , r. c. , kasen , d. , howell , d. a. , bersier , d. , bloom , j. s. , kulkarni , s. r. , kandrashoff , m. t. , filippenko , a. v. , silverman , j. m. , marcy , g. w. , howard , a. w. , isaacson , h. t. , maguire , k. , suzuki , n. , tarlton , j. e. , pan , y .- c . , bildsten , l. , fulton , b. j. , parrent , j. t. , sand , d. , podsiadlowski , p. , bianco , f. b. , dilday , b. , graham , m. l. , lyman , j. , james , p. , kasliwal , m. m. , law , n. m. , quimby , r. m. , hook , i. m. , walker , e. s. , mazzali , p. , pian , e. , ofek , e. o. , gal - yam , a. , & poznanski , d. 2011 , , 480 , 344 , s. , aldering , g. , goldhaber , g. , knop , r. a. , nugent , p. , castro , p. g. , deustua , s. , fabbro , s. , goobar , a. , groom , d. e. , hook , i. m. , kim , a. g. , kim , m. y. , lee , j. c. , nunes , n. j. , pain , r. , pennypacker , c. r. , quimby , r. , lidman , c. , ellis , r. s. , irwin , m. , mcmahon , r. g. , ruiz - lapuente , p. , walton , n. , schaefer , b. , boyle , b. j. , filippenko , a. v. , matheson , t. , fruchter , a. s. , panagia , n. , newberg , h. j. m. , couch , w. j. , & the supernova cosmology project . 1999 , , 517 , 565 , a. g. , filippenko , a. v. , challis , p. , clocchiatti , a. , diercks , a. , garnavich , p. m. , gilliland , r. l. , hogan , c. j. , jha , s. , kirshner , r. p. , leibundgut , b. , phillips , m. m. , reiss , d. , schmidt , b. p. , schommer , r. a. , smith , r. c. , spyromilio , j. , stubbs , c. , suntzeff , n. b. , & tonry , j. 1998 , , 116 , 1009 lccccccccccc @xmath71 & 8 & 63 & 128.0 & 168.0 & 18.9 & 0.38 & 3.51 & 3.10 & 4.20 & 1,630 & 1.23 + @xmath82 & 8 & 63 & 128.0 & 88.0 & 29.1 & 0.59 & 1.13 & 3.83 & 6.21 & 2,124 & 1.19 + @xmath81 & 8 & 63 & 128.0 & 68.0 & 38.6 & 0.78 & 0.43 & 4.71 & 8.67 & 2,660 & 1.05 +

this paper studies a specific metric on plane curves that has the property of being isometric to classical manifold ( sphere , complex projective , stiefel , grassmann ) modulo change of parametrization , each of these classical manifolds being associated to specific qualifications of the space of curves ( closed - open , modulo rotation etc ) using these isometries , we are able to explicitely describe the geodesics , first in the parametric case , then by modding out the paremetrization and considering horizontal vectors . we also compute the sectional curvature for these spaces , and show , in particular , that the space of closed curves modulo rotation and change of parameter has positive curvature . experimental results that explicitly compute minimizing geodesics between two closed curves are finally provided

0706.4299

the definition and study of spaces of plane shapes has recently met a large amount of interest @xcite , and has important applications , in object recognition , for the analysis of shape databases or in medical imaging . the theoretical background involves the construction of infinite dimensional manifolds of shapes @xcite . the riemannian framework , in particular , is appealing , because it provides shape spaces with a rich structure which is also useful for applications . a general discussion of several classes of metrics that can be introduced for this purpose can be found in @xcite . the present paper focuses on a particular riemannian metric that has very specific properties . this metric , which will be described in the next section , can be seen as a limit case of one of the classes studied in @xcite , and would receive the label @xmath0 in the nomenclature introduced therein . one of its surprising properties is that it can be characterized as a image of a grassmann manifold by a suitably chosen riemannian submersion . a consequence of this is the possibility to derive explicit geodesics in this shape space . a precursor of the @xmath0 metric has been introduced in @xcite and studied in the context of open plane curves . it has also recently been used in @xcite . because the metric is placed on curves modulo changes of parametrization , the computation of geodesics naturally provides an elastic matching algorithm . the paper is organized as follows . we first provide the definitions and notation that we will use for spaces of curves , the @xmath1 metric and the classical manifolds that will be shown to be isometric to it . we then study some local properties of the obtained manifold , discussing in particular its geodesics and sectional curvature . we finally provide experimental results for the numerical computation of geodesics and the solution of the related elastic matching problem . throughout this paper , we will assume our plane curves are curves in the _ complex _ plane @xmath2 . then real inner products and @xmath3 determinants of real 2-vectors are given by @xmath4 and @xmath5 . we first recall the notations for various spaces of plane curves which we will need , some of which were introduced in the previous paper @xcite . for all questions about infinite dimensional analysis and differential geometry we refer to @xcite . by @xmath6,\mathbb c)\ ] ] we denote the space of @xmath7-immersions @xmath8 \rightarrow \mathbb c$ ] . here ` @xmath9 ' stands for _ open curve_. @xmath10 is the quotient of @xmath11 by the group @xmath12)$ ] of @xmath7 increasing diffeomorphisms of @xmath13 $ ] . next @xmath14 are the spaces of @xmath7-immersions @xmath15 of _ even _ , respectively _ odd _ rotation degree . here , @xmath16 is the unit circle in @xmath2 , which will be identified in this paper to @xmath17 . then @xmath18 and @xmath19 are the quotients of @xmath20 , respectively @xmath21 by the group @xmath22 of @xmath7 orientation preserving diffeomorphisms of @xmath16 . for example , @xmath19 contains the simple closed plane curves , since they have index + 1 or @xmath23 ( depending on how they are oriented ) . these are the main focus of this study . to save us from enumerating special cases , we will often consider open curves as defined on @xmath16 but with a possible discontinuity at 0 . we will also consider the quotients of these spaces by the group of translations , by the group of translations and rotations and the group of translations , rotations and scalings . using the notation of @xcite , we can introduce the basic metric studied in this paper on these three spaces of immersions , but _ modulo translations _ , as follows . identify @xmath24 with the set of vector fields @xmath25 along @xmath26 modulo constant vector fields . then we consider the limiting case of the scale invariant metric of sobolev order 1 from @xcite , 4.8 : @xmath27 where , as in @xcite , @xmath28 is arclength measure , @xmath29 is the derivative with respect to arc length , and @xmath30 is the length of @xmath26 . we also recall for later use the notation @xmath31 for the unit tangent vector , and , as multiplication by @xmath32 is rotation by 90 degrees , @xmath33 for the unit normal . note that this metric is invariant with respect to reparametrizations of the curve @xmath26 , hence it induces a metric which we also call @xmath34 on the quotient spaces @xmath35 and @xmath19 also modulo translations . the geodesic equation in all these metrics is a simple limiting case of those worked out in @xcite . suppose @xmath36 is a geodesic . then : @xmath37 here the bar indicates the _ average _ of the quantity over the curve @xmath26 , i.e. @xmath38 . unfortunately , this case was not worked out in @xcite , hence we give the details of its derivation in appendix i. the local existence and uniqueness of solutions to this equations can be proved easily , essentially because of the regularizing influence of the term @xmath39 . this will also follow from the explicit representation of these geodesics to be given below , but because of its more general applicability , we give a direct proof in appendix i. it is convenient to introduce the _ momentum _ @xmath40 associated to a geodesic . using the momentum , the geodesic equation is easily rewritten in the more compact form : @xmath41 are nothing more than _ horizontal _ geodesics in @xmath42 , that is geodesics which are perpendicular at one hence all points to the orbit of the group of reparametrizations . as is shown in @xcite , horizontality is equivalent to the condition @xmath43 for some scalar function @xmath44 . substituting @xmath43 and taking the @xmath45-component of the last equation , we find that horizontal geodesics are given by : @xmath46 there are several conserved momenta along each geodesic @xmath47 of this metric ( see @xcite , 4.8 ) : the ` reparametrization ' momentum is @xmath48 which vanishes along all horizontal geodesics . the translation momentum vanishes because the metric does not feel translation ( constant vector fields along @xmath26 ) . the angular momentum is @xmath49 since the metric invariant under scalings , we also have the scaling momentum @xmath50 so we may equivalently consider either the quotient space @xmath51 or consider the section of the translation action @xmath52 . in the same way , we may either pass modulo scalings or consider the section by fixing @xmath53 , since the scaling momentum vanishes here . finally , in some cases , we will pass modulo rotations . we could consider the section where angular momentum vanishes : but this latter is not especially simple we introduce the three function spaces : @xmath54 \rightarrow \mathbb r \\ v_{\ev } & = \text{vector space of all } c^\infty \text { mappings } f : s^1 \rightarrow \mathbb r \text { such that } f(\th+2\pi ) \equiv f(\th)\\ v_{\od } & = \text{vector space of all } c^\infty \text { mappings } f : s^1 \rightarrow \mathbb r \text { such that } f(\th+2\pi ) \equiv -f(\th)\end{aligned}\ ] ] all three spaces have the weak inner product : @xmath55 given @xmath56 from any of these spaces , the basic map is : @xmath57 the map @xmath26 so defined carries @xmath13 $ ] or @xmath16 to @xmath2 . it need not be an immersion , however , because @xmath58 and @xmath59 might vanish simultaneously . define @xmath60 then we get three maps : @xmath61 looking separately at the three cases , define first the sphere @xmath62 to be the set of @xmath63 such that @xmath64 . @xmath65 is defined as the subset where @xmath66 . then the magic of the map @xmath67 is shown by the following key fact @xcite : @xmath67 defines a map : @xmath68 which is an isometric 2-fold covering , using the natural metric on @xmath69 and the metric @xmath70 on @xmath71 . proof the mapping @xmath67 is surjective : given @xmath72 with @xmath73 and @xmath53 , we write @xmath74 . then we may choose @xmath75 and @xmath76 . since @xmath77 we see that @xmath78 . the only choice here is the sign of the square root , i.e. @xmath79 , thus @xmath67 is 2:1 . to see that @xmath67 is an isometry , let @xmath80 and @xmath81 . then the differential @xmath82 is given by @xmath83 we have @xmath84 . this implies first that @xmath85 as required . then : @xmath86 the dictionary between pairs @xmath87 and immersions @xmath26 connects many properties of each with those of the other . curvature @xmath88 works out especially nicely . we list here some of the connections : @xmath89 and if @xmath90 is the wronskian , then : @xmath91 since geodesics on a sphere are always given by great circles , this theorem gives us the first case of explicit geodesics on spaces of curves in the metric of this paper . however , note that great circles in the open part @xmath92 are susceptible to crossing the ` bad ' part @xmath93 somewhere . this occurs if and only if there exists @xmath94 such that @xmath95 and @xmath96 have identical complex arguments modulo @xmath97 . so we find that our metric on @xmath42 is _ incomplete_. we can form a commutative diagram : @xmath98 ^ -{\text{2-fold } } \ar@{^{(}->}[d ] & \on{imm}_{\op}\big/ \left ( \text { transl , scalings } \right ) \ar@{^{(}->}[d ] \qquad \\ \widetilde{\ph } : \onb{s}(v_\op ) \ ; \ar[r ] & c^{\infty}([0,2\pi],\mathbb c ) \big/ \left ( \text{transl , scalings}\right)}\ ] ] where we have denoted the extended @xmath67 by @xmath99 . for rather technical reasons @xmath99 is not surjective : there are pathological non - negative @xmath7 functions which have no @xmath7 square root , see @xcite , e.g. but what this diagram does do is give some space of maps to hold the extended geodesics . the example : @xmath100 is shown in figure [ fig : genbifurc ] . this is a geodesic in which all curves are immersions for @xmath101 , but @xmath102 has a double zero at @xmath103 . the generic way in which a family of open immersions crosses the hypersurface where @xmath104 . the parametrized straight line in the middle of the family has velocity with a double zero at the black dot , hence is not an immersion . see text.,title="fig:",scaledwidth=70.0% ] + next , consider the periodic cases . here we need the _ stiefel manifolds _ : @xmath105 and @xmath106 the subset defined by the constraint @xmath66 . for later use , it is also convenient to note that @xmath107 where @xmath108 is the space of linear maps from @xmath2 to @xmath109 and @xmath110 . for @xmath111 , when is @xmath112 periodic ? if and only if we have : @xmath113 condition says that @xmath114 ( since the sum is 2 ) and @xmath115 , so that @xmath116 or @xmath117 . recall that the index @xmath45 of an immersed curve @xmath26 is defined by considering @xmath118 . the log must satisfy @xmath119 for some @xmath45 and this is the index . so this index is even or odd depending on whether the square root of @xmath120 is periodic or anti - periodic , that whether @xmath121 or @xmath122 . so if @xmath67 is restricted to @xmath123 or @xmath124 ( and is still denoted @xmath67 ) , it provides isometric 2-fold coverings @xmath125 all three of these maps @xmath67 can be modified so as to divide out by rotations . the mapping @xmath126 produces a rotation of the immersed curve @xmath127 through an angle @xmath128 . the complex projective space @xmath129 is @xmath130 divided by the action of rotations , and we denote by @xmath131 the subset gotten by dividing @xmath132 by rotations . the group generated by translations , rotations and scalings will be called the group of _ similitudes _ , abreviated as ` sim ' . then we get the variant : @xmath133 similarly , let @xmath134 be the grassmannian of unoriented 2-dimensional subspaces of @xmath109 and let @xmath135 be the subset of those @xmath136 with @xmath137 for @xmath138 or @xmath139 . then we have maps : @xmath140 for later use , we describe the tangent spaces of these spaces . the tangent space @xmath141 to @xmath134 at @xmath142 is naturally identified with @xmath143 and has the following norm , induced from that on @xmath109 : @xmath144 for @xmath145 and @xmath146 an orthonormal basis of @xmath136 . similarly , @xmath147 can be naturally identified with pairs @xmath148 in @xmath109 such that @xmath149 with norm : @xmath150 the same definition holds for @xmath151 , this time with the constraint @xmath152 . let @xmath153)$ ] be the group of increasing diffeomorphisms @xmath154 of @xmath13 $ ] ( so @xmath155 ) and let @xmath156 be the group of increasing @xmath7 diffeomorphisms @xmath157 such that @xmath158 for all @xmath159 . modulo the central subgroup of translations @xmath160 , the second group is just @xmath161 . for @xmath162 let @xmath163 be the group of unitary maps on @xmath109 given by : @xmath164 ) \text { if } v = v_{\op } \\ \ph \in c^{\infty,+}(\mathbb r ) \text { if } v = v_{\ev},v_{\od } \end{array } \right.\ ] ] these are the reparametrization groups for our various spaces . the infinitesimal action of a vector field @xmath165 on @xmath13 $ ] or a periodic vector field @xmath165 on @xmath16 is then @xmath166 for all three sets of isometries @xmath67 , we can now divide each side by the reparametrization group @xmath163 . for open curves , we get a diagram @xmath167 ^ -{\text{2-fold } } \ar[d ] & b_{i,\op}\big/ \left ( \text { transl , scalings } \right ) \ar[d ] \quad\\ \overline{\ph } : \onb{cp}^0(v^2_\op)/u(v_\op ) \ ; \ar[r]^-{\approx } & b_{i,\op } \big/ \left(\text{sim}\right)}\ ] ] and a similar one for closed curves of even and odd index where @xmath168 and @xmath169 : @xmath170 ^ -{\text{2-fold } } \ar[d ] & b \big/ \left ( \text{transl , scalings}\right ) \ar[d]\qquad\\ \overline{\ph } : \onb{gr}^0(2,v)/u(v ) \ ; \ar[r]^-\approx & b \big/ \left(\text{sim}\right)}\ ] ] here we have divided by isometries on both the left and right : by @xmath163 or @xmath171 on the left ( where @xmath16 rotates the basis @xmath146 ) and by reparametrizations and rotations on the right . thus @xmath172 is again an isometry if we make both quotients into riemannian submersions . this means we must identify the tangent spaces to the quotients with the _ horizontal _ subspaces of the tangent spaces in the larger space , i.e. those perpendicular to the orbits of the isometric group actions . for @xmath173 , this means : [ horizontalinst ] the tangent vector @xmath148 to @xmath173 satisfies : @xmath174 it is horizontal for the rotation action if and only if : @xmath175 it is horizontal for the reparametrization group if : @xmath176 where @xmath177 is the wronskian with respect to the parameter @xmath94 . proof consider the action of rotations , which is one - dimensional , with orbits @xmath178 ; the direction at @xmath87 is chosen as @xmath179 . so @xmath180 being horizontal at @xmath87 for this action means that @xmath181 . this proves the first assertion . for the action of @xmath163 , one has to note that horizontal vectors must satisfy @xmath182 for any periodic vector field @xmath165 on @xmath183 , which yields the horizontality condition after integration by parts of the terms in @xmath184 . horizontality on the shape space side means ( see @xcite ) : @xmath185 is horizontal for the action of @xmath186 if and only if @xmath187 is normal to the curve , i.e. @xmath188 . for any smooth path @xmath26 in @xmath189 there exists a smooth path @xmath154 in @xmath186 with @xmath190 depending smoothly on @xmath26 such that the path @xmath58 given by @xmath191 is horizontal : @xmath192 . this is a variant of @xcite . proof writing @xmath193 instead of @xmath194 we note that @xmath195 for @xmath196 . so we have @xmath197 . let us write @xmath198 for @xmath191 , etc . we look for @xmath154 as the integral curve of a time dependent vector field @xmath199 on @xmath16 , given by @xmath200 . we want the following expression to vanish : @xmath201 using the time dependent vector field @xmath202 and its flow @xmath154 achieves this . as we will see below , we can describe geodesics in the ` classical ' spaces @xmath203 quite explicitly . by the above isometries , this gives us the geodesics in the various spaces @xmath204 . but , as we mentioned above for the space @xmath69 , geodesics in the ` good ' parts @xmath205 do not stay there , but they cross the ` bad ' part where @xmath206 . now the basic mapping is still defined on the full sphere , projective space , stiefel manifold or grassmannian , giving us some smooth mappings of @xmath13 $ ] or @xmath16 to @xmath2 , possibly modulo translations , rotations and/or scalings . but when we divide by @xmath207 , a major problem arises . the orbits of @xmath207 acting on @xmath208,\mathbb c)$ ] are not closed , hence the topological quotient of the space @xmath208,\mathbb c)$ ] by @xmath207 is _ not hausdorff_. this is shown by the following construction : 1 . start with a @xmath7 non - decreasing map @xmath209 from @xmath13 $ ] to itself such that @xmath210 for @xmath94 in some interval @xmath211 . 2 . let @xmath212 . the sequence @xmath213 of diffeomorphisms of @xmath13 $ ] converges to @xmath209 . 3 . then for any @xmath214 , the maps @xmath215 are all in the orbit of @xmath26 . but they converge to @xmath216 which is constant on the whole interval @xmath211 , hence is not in the orbit . thus , if we want some hausdorff space of curves which a ) have singularities more complex than those of immersed curves and b ) can hold the extensions of geodesics in some space @xmath217 which come from the map @xmath67 , we must divide @xmath208 , \mathbb c)$ ] by some equivalence relation larger than the group action by @xmath163 . the simplest seems to be : first define a _ monotone relation _ @xmath218\times [ 0,2\pi]$ ] to be any closed subset such that @xmath219 $ ] ( @xmath220 and @xmath221 being the projections on the axes ) and for every pair of points @xmath222 and @xmath223 , either @xmath224 and @xmath225 or vice versa . then @xmath226 \rightarrow \mathbb c$ ] are _ frchet _ equivalent if there is a monotone relation @xmath227 such that @xmath228 . this is a good equivalence relation because if @xmath229 \rightarrow \mathbb c$ ] are two sequences and @xmath230 and @xmath231 are frchet equivalent for all @xmath45 , then @xmath232 are frchet equivalent . the essential point is that the set of non - empty closed subsets of a compact metric space @xmath165 is compact in the hausdorff topology ( see @xcite ) . thus if @xmath233 are the monotone relations instantiating the equivalence of @xmath234 and @xmath235 , a subsequence @xmath236 hausdorff converges to some @xmath218 \times [ 0,2\pi]$ ] and it is immediate that @xmath227 is a monotone relation making @xmath59 and @xmath237 frchet equivalent . define @xmath238,\mathbb c)/ \text{fr\'echet equivalence , translations , scalings}.\ ] ] then we have a commutative diagram : @xmath98 \ar@{^{(}->}[d ] & b_{i,\op}\big/ \left ( \text{transl , scalings } \right ) \ar@{^{(}->}[d ] \qquad \\ \widetilde{\ph } : \onb{s}(v^2_\op ) \ ; \ar[r ] & b_{big,\op}}\ ] ] thus the whole of a geodesic which enters the ` bad ' part of @xmath239 creates a path in @xmath240 . of course , the same construction works for closed curves also . we will see several examples in the next section . the space @xmath241 being the sphere of radius @xmath242 on @xmath243 , its geodesics are the great circles . thus , the geodesic distance between @xmath244 and @xmath245 is given by @xmath246 with @xmath247 and the geodesic is given by @xmath248 the corresponding geodesic on @xmath71 modulo translation and scaling is the time - indexed family of curves @xmath249 with @xmath250 the following notation will be used throughout this section : @xmath251 so that @xmath252 and @xmath253 for @xmath254 . thus the distance @xmath255 is @xmath256 the metric on @xmath71 modulo rotations is @xmath257 the distance on @xmath10 is the infimum of this expression over all changes of coordinate for @xmath258 . assuming that @xmath258 and @xmath259 are originally parametrized with @xmath260 times arc - length so that @xmath261 , this is @xmath262 and modulo rotations @xmath263 the supremum in both expressions is taken over all increasing bijections @xmath264 , [ 0 , 2\pi])$ ] . to shorten these formulae , we will use the following notation . define : @xmath265 then we have : @xmath266 these expressions give very explicit descriptions of distance and geodesics . we have already noted , however , that , even if both @xmath244 and @xmath245 belong to @xmath92 , the same property is not guaranteed at each point of the geodesic . @xmath267 happens for some @xmath268 whenever @xmath269 and @xmath270 are collinear with opposite orientations . this is not likely to happen for geodesics joining nearby points . when it does happen , it is usually a stable phenomenon : for example , if the geodesic crosses @xmath93 transversally , as illustrated in figure [ fig : genbifurc ] , then this happens for all nearby geodesics too . note that this means that the geodesic spray on @xmath11 is _ not _ surjective . in fact , any geodesic on @xmath11 comes from a great circle on @xmath271 and if it crosses @xmath93 , it leaves @xmath11 . when we pass to the quotient by reparametrizations , another question arises : does the inf over reparametrizations exist ? or equivalently is there is a _ horizontal _ geodesic joining any two open curves ? in fact , there need not be any such geodesic even if you allow it to cross @xmath93 . in general , to obtain a geodesic minimizing distance between 2 open curves , the curves themselves must be given parametrizations with zero velocity somewhere , i.e.they may need to be lifted to points in @xmath93 . this is best illustrated by the special case in which @xmath259 is the line segment from @xmath272 to @xmath273 , namely @xmath274 , @xmath275 . the curve @xmath258 can be arbitrary . then the reparametrization @xmath276 which minimizes distance is the one which maximizes : @xmath277 this variational problem is easy to solve : the optimal @xmath276 is given by : @xmath278 note that @xmath276 is _ not _ in general a diffeomorphism : it is constant on intervals where @xmath279 . its graph is a monotone relation in the sense of section [ sec : bigger ] . in fact , it s easy to see that monotone relations enjoy a certain compactness , so that the inf over reparametrizations is always achieved by a monotone relation . assuming @xmath280 is represented by a continuous function for which @xmath281 , the result is that the places on the curve @xmath258 where @xmath282 get squashed to points on the line segment . the result is that this limit geodesic is not actually a path in the space @xmath283 of smooth curves . figure [ fig : badinf ] illustrates this effect . the general problem of maximizing the functional @xmath284 with respect to increasing functions @xmath276 has been addressed in @xcite . existence of solutions can be shown in the class of monotone relations , or , equivalently functions @xmath276 that take the form @xmath285 for some positive measure @xmath286 on @xmath287 $ ] with total mass less or equal to @xmath288 ( @xmath289 being replaced by the radon - nykodim derivative of @xmath286 in the definition of @xmath290 ) . the optimal @xmath276 is a diffeomorphism as soon as @xmath291 whenever @xmath292 is smaller than a constant ( which depends on @xmath280 and @xmath293 ) . more details can be found in @xcite . it is easy to show that maximizing @xmath290 is equivalent to maximizing @xmath294 because one can always modify @xmath276 on intervals on which @xmath295 to ensure that @xmath296 . in @xcite , it is proposed to maximize @xmath297 this corresponds to replacing the lift @xmath298 by @xmath299 where @xmath300 for all @xmath301 , but this is beyond the scope of this paper . this is a geodesic of open curves running from the curve with the kink at the top left to the straight line on the bottom right . a blow up of the next to last curve is shown to reveal that the kink never goes away it merely shrinks . thus this geodesic is not continuous in the @xmath302-topology on @xmath303 . the straight line is parametrized so that it stops for a whole interval of time when it hits the middle point and thus it is @xmath302-continuous in @xmath11.,title="fig : " ] + the integrated path - length distance and explicit geodesics can be found in any grassmannian using _ jordan angles _ @xcite as follows : if @xmath304 are two 2-dimensional subspaces the singular value decomposition of the orthogonal projection @xmath305 of @xmath306 to @xmath307 gives orthonormal bases @xmath308 of @xmath306 and @xmath309 of @xmath307 such that @xmath310 , @xmath311 , @xmath312 and @xmath313 , where @xmath314 . write @xmath315 then @xmath316 , @xmath317 are the jordan angles , @xmath318 . the global metric is given by : @xmath319 and the geodesic by : @xmath320 we apply this now in order to compute the distance between the curves in the two spaces @xmath321 and @xmath322 , as well as in the unparametrized quotients @xmath323 and @xmath324 . we write as above @xmath325 and @xmath326 . we put @xmath327 thus lifting these curves to 2-planes in the grassmannian . the @xmath328 matrix of the orthogonal projection from the space @xmath329 to @xmath330 in these bases is : @xmath331 it will be convenient to use the notations : @xmath332 we have to diagonalize this matrix by rotating the curve @xmath258 by a constant angle @xmath333 , i.e. , the basis @xmath334 by the angle @xmath335 ; and similarly the curve @xmath259 by a constant angle @xmath336 . so we have to replace @xmath280 by @xmath337 and @xmath293 by @xmath338 in such a way that @xmath339 thus @xmath340 in the newly aligned bases , the diagonal elements of @xmath341 will be the cosines of the jordan angles . but even without preliminary diagonalization , the following lemma gives you a formula for them : if @xmath342 and @xmath343 , then the singular values of @xmath344 are : @xmath345 the proof is straightforward . this gives the formula @xmath346 this is the distance in the space @xmath347/ ( ) . if we want the distance in the quotient space @xmath348 by the group @xmath349 we have to take the infimum of over all reparametrizations . to simplify the formulas that follow , we can assume that the initial curves @xmath350 are parametrized by arc length so that @xmath351 . then consider a reparametrization @xmath352 of one of the two curves , say @xmath353 : @xmath354 where now @xmath355 to describe the inf , we can use the fact that geodesics on the space of curves are the horizontal geodesics in the space of immersions . consider the geodesic @xmath356 in @xmath134 described in , for @xmath357 where the rotations @xmath333 and @xmath336 must be computed from @xmath353 and @xmath259 . note that @xmath358 if the jordan angles are @xmath316 and @xmath317 , then the tangent vector to the geodesic @xmath359 at @xmath360 is described by @xmath361 by [ horizontalinst ] the geodesic is perpendicular to all @xmath186-orbits if and only if the sum of wronskians vanishes : @xmath362 this is an ordinary differential equation for @xmath276 which is coupled to the ( integral ) equations for calculating the @xmath363 s as functions of @xmath276 . if it is non - singular ( i.e. , the coefficient function of @xmath364 does not vanish for any @xmath94 ) then there is a solution @xmath276 , at least locally . but the non - existence of the inf described for open curves above will also affect closed curves and global solutions may actually not exist . however , for closed curves that do not double back on themselves too much , as we will see , geodesics do seem to usually exist . geodesics in the sphere are great circles , which go all the way around the sphere and are always closed geodesics . in the case of the grassmannian , using the jordan angle basis , the geodesic can be continued indefinitely using formula [ eq : neretingeod ] above . in fact , it will be a _ closed _ geodesic if the jordan angles @xmath365 are commensurable . it is interesting to look at an example to see what sort of immersed curves arise , for example , at the antipodes to the point representing the unit circle . to do this , we take @xmath366 to be the circle of unit length , giving the orthonormal basis @xmath367 . we want @xmath368 to lie in a direction horizontal with respect to these and the simplest choice satisfying the wronskian condition is : @xmath369 the result is shown in figure [ fig : grtcircle ] . a great circle geodesic on @xmath370 . the geodesic begins at the circle at the top left , runs from left to right , then to the second row and finally the third . it leaves @xmath370 twice : at the top right and bottom left , in both of which the singularity of figure [ fig : genbifurc ] occurs in 2 places . the index of the curve changes from @xmath371 to @xmath372 in the middle row . see text.,title="fig : " ] + we compute , in this section , the sectional curvature of @xmath324 ( i.e. , translations , rotations , scaling ) . we first compute the sectional curvature on the grassmannian which is non - negative ( but vanishes on many planes ) and conclude from oneill s formula @xcite that the sectional curvature on @xmath373 is non - negative but since the oneill correction term is difficult to compute in this setting we also do it in a more explicit way , computing first the curvature on the stiefel manifold by gauss equation , then carrying it over to @xmath374 . since this is an open subset in a frchet space , the oneill correction term can be computed more easily on @xmath374 and so we finally get a more explicit formula for the sectional curvature on @xmath375 . let @xmath376 be a fixed 2-plane which we identify again with @xmath377 . let @xmath378 be the isomorphism which equals @xmath23 on @xmath136 and @xmath273 on @xmath379 satisfying @xmath380 . then the grassmanian is the symmetric space @xmath381 with the involutive automorphism @xmath382 given by @xmath383 . for the lie algebra in the @xmath384-decomposition we have @xmath385 here @xmath386 . the fixed point group is @xmath387 . the reductive decomposition @xmath388 is given by @xmath389 let @xmath390 be the quotient projection . then @xmath391 is an isomorphism , and the @xmath392-invariant riemannian metric on @xmath134 is given by @xmath393 for @xmath394 , where @xmath395 is an orthonormal base of @xmath136 . by the general theory of symmetric spaces @xcite , the curvature is given by @xmath396,y_1 ] \\ & \left[\begin{pmatrix } 0 & -y_1^t \\ y_1 & 0 \end{pmatrix } , \begin{pmatrix } 0 & -y_2^t \\ y_2 & 0 \end{pmatrix } \right ] = \begin{pmatrix } -y_1^ty_2 + y_2^ty_1 \\ 0 & -y_1y_2^t + y_2y_1^t\end{pmatrix } \\ & \left[\left[\begin{pmatrix } 0 & -y_1^t \\ y_1 & 0 \end{pmatrix } , \begin{pmatrix } 0 & -y_2^t \\ y_2 & 0 \end{pmatrix } \right ] , \begin{pmatrix } 0 & -y_1^t \\ y_1 & 0 \end{pmatrix}\right ] = \\&\qquad\qquad = \left[\begin{pmatrix } -y_1^ty_2 + y_2^ty_1 & 0 \\ 0 & -y_1y_2^t + y_2y_1^t\end{pmatrix } , \begin{pmatrix } 0 & -y_1^t \\ y_1 & 0 \end{pmatrix}\right ] \\&\qquad\qquad = \begin{pmatrix } 0 & 2y_1^ty_2y_1^t - y_2^ty_1y_1^t - y_1^ty_1y_2^t \\ -2y_1y_2^ty_1+y_2y_1^ty_1+y_1y_1^ty_2 & 0 \end{pmatrix}\end{aligned}\ ] ] for the sectional curvature we have ( where we assume that @xmath397 is orthonormal ) : @xmath398,y_1 ] ) = \on{tr}_w(y_2^ty_2y_1^ty_1+y_2^ty_1y_1^ty_2 - 2y_2^ty_1y_2^ty_1 ) \\ & = \tfrac12\on{tr}_w\big((y_2^ty_1-y_1^ty_2)^t(y_2^ty_1-y_1^ty_2)\big ) \\&\qquad + \tfrac12\on{tr}_{w^\bot}\big((y_2y_1^t - y_1y_2^t)^t(y_2y_1^t - y_1y_2^t)\big ) \\ & = \tfrac12\|y_2^ty_1-y_1^ty_2\|^2_{l^2(w , w ) } + \tfrac12\|y_2y_1^t - y_1y_2^t\|^2_{l^2(w^\bot , w^\bot ) } \ge 0.\end{aligned}\ ] ] where @xmath399 stands for the space of hilbert - schmidt operators . note that there are many orthonormal pairs @xmath397 on which sectional curvature vanishes and that its maximum value 2 is attained when @xmath400 are isometries and @xmath401 where @xmath402 is rotation through angle @xmath403 in the image plane of @xmath404 . the curvature formula can be rewritten by ` lowering the indices ' which will make it much easier to express it in terms of the immersion @xmath26 . fix an orthonormal basis @xmath56 of @xmath136 and let @xmath406 . for @xmath407 , we use the notation @xmath408 then : @xmath409 to check this , note that @xmath410 is given by a skew - symmetric @xmath328 matrix whose off diagonal entry is just @xmath411 and this identifies the first terms in the two formulas for @xmath412 . on the other hand , @xmath413 is given by a matrix of rank 2 on the infinite dimensional space @xmath379 . in view of @xmath414 it is the 2-tensor @xmath415 . skew - symmetrizing , we identify the second terms in the two expressions for @xmath412 . going over to the immersion @xmath26 , the tangent vector @xmath416 to @xmath417 becomes the tangent vector @xmath418 to @xmath405 . to express the first term in the curvature , we have : @xmath419 @xmath420 hence @xmath421 hence @xmath422 the second term is not quite so compact : because it is a norm on @xmath423 , it requires double integrals over @xmath424 , not just a simple integral over @xmath425 . we use the notation as above @xmath426 . then we have : @xmath427 using @xmath428 and @xmath429 , we have : @xmath430 hence : @xmath431 skew - symmetrizing in the 2 vectors @xmath432 , we get : @xmath433 squaring and integrating over @xmath434 , the right hand becomes @xmath435 . on the left , first write @xmath436 as the sum of @xmath437 and @xmath438 . then when we square and integrate , the cross term drops out because it is odd when @xmath439 are reversed . we therefore obtain the expression of the curvature in @xmath405 : @xmath440 a major consequence of the calculation for the curvature on the grassmannian is : the sectional curvature on @xmath441 is @xmath442 . we apply oneill s formula @xcite to the riemannian submersion @xmath443^{\text{ver}}|_{w}\|^2 \ge 0\end{gathered}\ ] ] where @xmath444 is a horizontal vector field projecting to a vector field @xmath165 at @xmath445 ; similarly for @xmath446 . the horizontal and vertical projections exist and are pseudo differential operators , see [ curvb ] . the stiefel manifold is not a symmetric space ( as the grassmannian ) ; it is a homogeneous riemannian manifold . this can be used to compute its sectional curvature . but the following procedure is simpler : for @xmath448 we consider the functions @xmath449 then @xmath447 is the codimension 3 submanifold of @xmath243 defined by the equations @xmath450 , @xmath451 . the metric on @xmath447 is induced by the metric on @xmath452 . if @xmath453 and @xmath454 are tangent vectors at a point in @xmath452 , we have @xmath455 . for a function @xmath154 on @xmath452 its gradient @xmath456 of @xmath154 ( if it exists ) is given by @xmath457 . the following are the gradients of @xmath458 : @xmath459 and these form an orthonormal basis of the normal bundle @xmath460 of @xmath447 . let @xmath461 be two normal unit vectors tangent to @xmath447 at a point @xmath87 . since @xmath243 is flat the sectional curvature of @xmath447 is given by the gauss formula @xcite : @xmath462 where @xmath463 denotes the second fundamental form of @xmath447 in @xmath452 . moreover , when a manifold is given as the zeros of functions @xmath464 in a flat ambient space whose gradients are orthonormal , the second fundamental form is given by : @xmath465 where @xmath466 is the hessian of second derivatives . given @xmath467 with @xmath468 , we have : @xmath469 so that @xmath470 finally , _ the sectional curvature of @xmath471 for a normal pair of unit vectors @xmath472 in @xmath473 is given by : _ @xmath474 comparing this with the curvature for the grasmannian , we see that the oneill factor in this case is @xmath475 . moreover , we can write for the curvature of the isometric @xmath476 @xmath477 using the basic mapping @xmath478 , the manifold @xmath374 can be identified with the unscaled stiefel manifold which we view as the following submanifold of @xmath243 . we do not introduce a systematic notation for it . @xmath479 equipped with the metric @xmath480 consider the diffeomorphism @xmath481 defined by @xmath482 for @xmath483 we have @xmath484 thus , @xmath485 is an isometry if @xmath486 is equipped with the metric @xmath487 so that @xmath344 is isometric to the riemannian product of @xmath488 and @xmath447 , taking @xmath489 for the metric on @xmath488 . this implies that the curvature tensor on @xmath344 is the sum of the tensors on @xmath488 ( which vanishes ) and @xmath447 . thus , if @xmath490 , @xmath491 with @xmath492 orthonormal , @xmath493 note that we have the relations : @xmath494 for riemannian submersions , oneill formula @xcite states that the sectional curvature , in the plane generated by two horizontal vectors , is given by the curvature computed on the space `` above '' plus a positive correction term given by @xmath495 times the squared norm of the vertical projection of the lie bracket of any horizontal extensions of the two vectors . we now proceed to the computation of this correction for the submersion from @xmath496 to @xmath497 . because of the simplicity of local charts there , it will be easier to start from @xmath498 . let @xmath499 with @xmath500 . we first compute the vertical projection of a vector @xmath501 for the submersion @xmath502 . vectors in the vertical space at @xmath26 take the form @xmath503 each generator corresponding ( in this order ) to the action of diffeomorphisms , rotation and scaling ( @xmath504 is a function and @xmath505 are constants ) . denoting @xmath506 the vertical projection of @xmath507 , and using the fact that @xmath508 for any vertical @xmath509 , we easily obtain the fact that @xmath510 with @xmath511 where we have used the following notation : @xmath512 , @xmath513 and , as before @xmath514 from this , we obtain the fact that @xmath504 must satisfy @xmath515 the operator @xmath516 is of order two , unbounded , selfadjoint , and positive on @xmath517 thus it is invertible on @xmath518 by an index argument as given in lemma @xcite . the operator @xmath519 in the left - hand side of is also invertible under the condition that @xmath26 is not a circle , with an inverse given by @xmath520 this is well defined unless @xmath521 . indeed , letting @xmath522 , we have @xmath523 which implies @xmath524 . by schwartz inequality we have @xmath525 which ensures @xmath526 equality requires @xmath527 or @xmath528 , which in turn implies that @xmath529 and that @xmath26 is a circle . we note for future use that @xmath530 . we hereafter assume that @xmath26 has length 1 , is parametrized with its arc - length divided by @xmath288 , and that it is different from the unit circle ( which is a singular point in @xmath531 ) . we can therefore write @xmath532 with @xmath533 . the right - hand term in is the sum of three orthogonal terms , the last two forming the vertical projection for the submersion @xmath534 . applying oneill s formula two times to this submersion and to @xmath535 , we see that the correcting term for the sectional curvature on @xmath531 relative to the curvature on @xmath536 , in the direction of the horizontal vectors @xmath537 and @xmath538 is @xmath539_c ) v - i{\overline{\left\langle { \ka ( \tilde l^\top)^{-1}\psi([\bar h_1 , \bar h_2]_c ) } \right\rangle } } c\right\|^2,\ ] ] @xmath540 being horizontal extensions of @xmath537 and @xmath538 . from the identity @xmath541 we can write @xmath542_c ) ( \tilde l^\top)^{-1}\psi([\bar h_1 , \bar h_2]_c ) ds.\ ] ] we now proceed to the computation of the lie bracket : @xmath543_c ) = w_s(d_sh_1 \cdot n , d_sh_2\cdot n ) - { \overline{\left\langle { \mathrm{det}(d_sh_1 , d_sh_2 ) } \right\rangle}}\ka$ ] where @xmath544 is the wronskian with respect to the arc - length parameter . proof we take @xmath545 which are horizontal at @xmath26 , consider them as constant vector fields on @xmath374 and take , as horizontal extensions , their horizontal projections @xmath546 . then we compute the lie bracket evaluated at @xmath547 : @xmath548\big|_{\ga } & = & d_{c , h_2}h_1^\bot(\ga ) - d_{c , h_1}h_2^\bot(\ga ) \\ & = & - d_{c , h_2}h_1^\top(\ga ) + d_{c , h_1}h_2^\top(\ga ) \end{aligned}\ ] ] since @xmath549 is constant , for @xmath491 . we have @xmath550 with @xmath551 . we have added subscripts @xmath547 to quantities that depend on the curve , with @xmath552 holding for the derivative with respect to the @xmath547 arc - length ( we still use no subscript for @xmath553 ) . note that @xmath554 and @xmath555 which is a first simplification . also , since we assume that @xmath537 is horizontal at @xmath26 , we have @xmath556 and @xmath557 , which imply @xmath558 . we therefore have ( to simplify , we temporarily use the notation @xmath559 ) @xmath560 since we have @xmath561 and @xmath562 we immediately obtain the expression of the last two terms in , which are @xmath563 we now focus on the variation of @xmath564 . we need to compute @xmath565 if @xmath507 is a constant vector field , we have @xmath566 and @xmath567 this implies @xmath568 therefore @xmath569 using @xmath570 and the fact that @xmath571 , we can write @xmath572 which yields @xmath573 by symmetry @xmath574 where @xmath575 combining this with , we get @xmath576_c & = & ( \tilde l^\top)^{-1 } ( w_s(h'_1 \cdot n , h'_2\cdot n ) - { \overline{\left\langle { \mathrm{det}(h'_1 , h'_2 ) } \right\rangle}}\ka ) v\\ & - & i{\overline{\left\langle { \ka ( \tilde l^\top)^{-1 } ( w_s(h'_1 \cdot n , h'_2\cdot n ) - { \overline{\left\langle { \mathrm{det}(h'_1 , h'_2 ) } \right\rangle}}\ka ) } \right\rangle } } c - i{\overline{\left\langle { \mathrm{det}(h'_1 , h'_2 ) } \right\rangle } } c\end{aligned}\ ] ] so that @xmath577_c ) & = & w_s(h'_1 \cdot n , k'_2\cdot n ) - { \overline{\left\langle { \mathrm{det}(h'_1 , h'_2 ) } \right\rangle}}\ka . \end{aligned}\ ] ] ( we have used the fact that @xmath578 . ) we therefore obtain the formula @xmath579 with @xmath580 given by . finally , assuming that @xmath537 and @xmath538 are orthogonal , @xmath581 where @xmath582 is given in . a similar ( and simpler ) computation provides the correcting term for the space @xmath583 . in this case , the rotation part of the vertical space disappears , and the two remaining components ( parametrization and scale ) are orthogonal . the result is @xmath584 here we derive an explicit upper bound for @xmath586 at a fixed curve @xmath587 and a fixed tangent vector @xmath538 . this will show that geodesics ( such as the one in the @xmath537 direction ) have at least a small interval before they meet another geodesic . the size of this interval can be controlled , as we will see , by an upper bound that involves the supremum norm of the first two derivatives of @xmath537 . we assume that @xmath26 has length @xmath588 . since @xmath589 is isometric to @xmath134 , its sectional curvature is not larger than 2 as already remarked . we estimate the terms in @xmath590 where @xmath591 . for a fixed @xmath538 , @xmath592 is function of @xmath537 belonging to @xmath593 . we estimate @xmath594 and then @xmath595 by estimating the norm of the operator @xmath596 which maps @xmath593to @xmath597 . if @xmath598 , @xmath599 and @xmath600 . therefore , @xmath601 since @xmath537 has norm 1 , @xmath602 and @xmath603 are @xmath604 . @xmath605 this results in @xmath606 now @xmath607 since @xmath608 . if @xmath609 then @xmath610 proof let @xmath611 . let @xmath612 . then , @xmath613 and @xmath614 . let @xmath615 so that @xmath616 . the eigenvalues of @xmath617 are positive and bounded from below by @xmath618 , see @xcite . therefore , @xmath619 where @xmath620 . we also have @xmath621 hence , @xmath622 therefore , @xmath623 and @xmath624 . finally , @xmath625 putting all the estimates together we get , for orthonormal @xmath432 as always , @xmath626 the distance @xmath627 given in [ greatcircles ] can be computed in a very short time by dynamic programming , using a slightly modified procedure from the one described in @xcite . here is a sketch of how it works . let @xmath628 , and assume that the curves are discretized over intervals @xmath629 , @xmath630 , @xmath631 , so that the angles have constant values , @xmath632 on these intervals . the problem is then equivalent to maximizing @xmath633 with @xmath634 . because the integral of @xmath635 is maximal for linear @xmath276 , we must in fact maximize @xmath636 with the notation @xmath637 and @xmath638 . the method now essentially implements a coupled linear programming procedure over the values of @xmath639 and @xmath640 . see @xcite for more details . this procedure is very fast , and one still obtains an efficient procedure by combining it with an exhaustive search for an optimal rotation . for closed curves , we can furthermore optimize the result with respect to the offset @xmath641 , for the diffeomorphism . doing so provides the value of @xmath642 where the notation @xmath643 is here to remember that the minimization is over @xmath644 and not @xmath645)$ ] . this combination of the almost instantaneous dynamic programming method and of an exhaustive search over two parameters provides a feasible elastic matching method for closed curves . but this does not provide the geodesic distance over @xmath646 , since we worked with great circles instead of the neretin geodesics . there are two consequences for this : first , the obtained distance is only a lower bound of the distance on @xmath646 , and second , since the closedness constraint is not included , the curves generally become open during the evolution ( as shown in the experiments ) . however , the optimal diffeomorphism which has been obtained by this approach can be used to reparametrize the curve @xmath258 , and we can compute the geodesic between @xmath647 and @xmath259 in @xmath405 using neretin geodesics , which forms , this time , an evolution of closed curves . its geodesic length now obviously provides an upper - bound for the geodesic distance on @xmath648 . the numerical results that are presented in figures [ fig : bird02.arb2 ] to [ fig : hand90.heart02 ] compare the great circles and neretin geodesics obtained using this method . quite surprisingly , the differences between the lower and upper bounds in these examples are quite small . curve evolution with and without the closedness constraint . lower and upper bounds for the geodesic distance : 0.443 and 0.444,title="fig:",scaledwidth=80.0% ] + curve evolution with and without the closedness constraint . lower and upper bounds for the geodesic distance : 0.443 and 0.444,title="fig:",scaledwidth=80.0% ] curve evolution with and without the closedness constraint . lower and upper bounds for the geodesic distance : 0.462 and 0.464,title="fig:",scaledwidth=80.0% ] + curve evolution with and without the closedness constraint . lower and upper bounds for the geodesic distance : 0.462 and 0.464,title="fig:",scaledwidth=80.0% ] curve evolution with and without the closedness constraint . lower and upper bounds for the geodesic distance : 0.433 and 0.439,title="fig:",scaledwidth=80.0% ] + curve evolution with and without the closedness constraint . lower and upper bounds for the geodesic distance : 0.433 and 0.439,title="fig:",scaledwidth=80.0% ] curve evolution with and without the closedness constraint . lower and upper bounds for the geodesic distance : 0.498 and 0.532,title="fig:",scaledwidth=80.0% ] + curve evolution with and without the closedness constraint . lower and upper bounds for the geodesic distance : 0.498 and 0.532,title="fig:",scaledwidth=80.0% ] curve evolution with and without the closedness constraint . lower and upper bounds for the geodesic distance : 0.513 and 0.528,title="fig:",scaledwidth=80.0% ] + curve evolution with and without the closedness constraint . lower and upper bounds for the geodesic distance : 0.513 and 0.528,title="fig:",scaledwidth=80.0% ] we use the method of @xcite for the space @xmath650 which is an open subset in a frchet space , with tangent space @xmath651 . we shall use the following conventions and results from @xcite : @xmath652 then the derivative of the metric at @xmath26 in direction @xmath653 is : @xmath654 according to @xcite we should rewrite this as @xmath655 and thus we find the two versions @xmath656 and @xmath466 of the @xmath34-gradient of @xmath657 : @xmath658 which gives us the geodesic equation by @xcite : @xmath659 [ existenceg ] for each @xmath660 the geodesic equation derived in [ geodesicimm ] has unique local solutions in the sobolev space of @xmath661-immersions . the solutions depend @xmath7 on @xmath268 and on the initial conditions @xmath662 and @xmath663 . the domain of existence ( in @xmath268 ) is uniform in @xmath412 and thus this also holds in @xmath664 . proof the proof is very similar to the one of @xcite . we denote by @xmath665 any space of based loops @xmath666 . we consider the geodesic equation as the flow equation of a smooth ( @xmath7 ) vector field on the @xmath667-open set @xmath668 in the sobolev space @xmath669 where @xmath670 is @xmath667-open . to see that this works we will use the following facts : by the sobolev inequality we have a bounded linear embedding @xmath671 if @xmath672 . the sobolev space @xmath673 is a banach algebra under pointwise multiplication if @xmath674 . for any fixed smooth mapping @xmath59 the mapping @xmath675 is smooth @xmath676 if @xmath677 . we write @xmath678 just for the remainder of this proof to stress the dependence on @xmath26 . the mapping @xmath679 is smooth @xmath680 and is a bibounded linear isomorphism @xmath681 for fixed @xmath26 . this can be seen as follows ( compare with @xcite ) : it is true if @xmath26 is parametrized by arclength ( look at it in the space of fourier coefficients ) . the index is invariant under continuous deformations of elliptic operators of fixed degree , so the index of @xmath682 is zero in general . but @xmath682 is self - adjoint positive , so it is injective with vanishing index , thus surjective . by the open mapping theorem it is then bibounded . moreover @xmath683 is smooth @xmath684 ( by the inverse function theorem on banach spaces ) . the mapping @xmath685 is smooth @xmath686 for @xmath687 , and is linear in @xmath59 . we have @xmath688 and @xmath689 . the mapping @xmath690 is smooth on the @xmath667-open set @xmath691 into @xmath692 . keeping all this in mind we now write the geodesic equation as follows : here we used that along any geodesic the norm @xmath694 and the scaling momentum @xmath695 are both constant in @xmath268 . now a term by term investigation shows that the expression in the brackets is smooth @xmath696 since @xmath697 . the operator @xmath698 then takes it smoothly back to @xmath699 . so the vector field @xmath700 is smooth on @xmath701 . thus the flow @xmath702 exists on @xmath699 and is smooth in @xmath268 and the initial conditions for fixed @xmath412 . now we consider smooth initial conditions @xmath703 and @xmath704 in @xmath705 . suppose the trajectory @xmath706 of @xmath165 through these intial conditions in @xmath699 maximally exists for @xmath707 , and the trajectory @xmath708 in @xmath709 maximally exists for @xmath710 with @xmath711 . by uniqueness we have @xmath712 for @xmath713 . we now apply @xmath714 to the equation @xmath715 , note that the commutator @xmath716 $ ] is a pseudo differential operator of order @xmath717 again , and write @xmath718 . we obtain @xmath719(\,\dots\ , ) + \text{const}.w$ ] . in the term @xmath720 we consider now only the terms @xmath721 and rename them @xmath722 . then we get an equation @xmath723 which is inhomogeneous bounded linear in @xmath724 with coefficients bounded linear operators on @xmath699 which are @xmath7 functions of @xmath725 . these we already know on the intervall @xmath726 . this equation therefore has a solution @xmath727 for all @xmath268 for which the coefficients exists , thus for all @xmath728 . the limit @xmath729 exists in @xmath699 and by continuity it equals @xmath730 in @xmath699 at @xmath731 . thus the @xmath709-flow was not maximal and can be continued . so @xmath732 . we can iterate this and conclude that the flow of @xmath165 exists in @xmath733 .

we consider a minimally coupled scalar field with a monomial potential and a perfect fluid in flat flrw cosmology . we apply local and global dynamical systems techniques to a new three - dimensional dynamical systems reformulation of the field equations on a compact state space . this leads to a visual global description of the solution space and asymptotic behavior . at late times we employ averaging techniques to prove statements about how the relationship between the equation of state of the fluid and the monomial exponent of the scalar field affects asymptotic source dominance and asymptotic manifest self - similarity breaking . we also situate the ` attractor ' solution in the three - dimensional state space and show that it corresponds to the one - dimensional unstable center manifold of a de sitter fixed point , located on an unphysical boundary associated with the dynamics at early times . by deriving a center manifold expansion we obtain approximate expressions for the attractor solution . we subsequently improve the accuracy and range of the approximation by means of pad approximants and compare with the slow - roll approximation .

1503.06994

the present paper investigates general relativistic flat friedmann - lematre - robertson - walker ( flrw ) models with a minimally coupled scalar field with a monomial potential , @xmath0 ( @xmath1 , @xmath2 ) , and a perfect fluid . the perfect fluid is assumed to obey a linear equation of state , @xmath3 , where @xmath4 and @xmath5 are the pressure and the energy density , respectively . the adiabatic index is assumed to satisfy @xmath6 , where @xmath7 corresponds to dust and @xmath8 to radiation . when @xmath9 the matter term @xmath10 can be reinterpreted as @xmath11 , where @xmath12 is the spatial 3-curvature of the open flrw model , i.e. , @xmath9 leads to equations that are the same as those for a scalar field in open flrw cosmology . the case @xmath13 corresponds to a matter content described by a cosmological constant , i.e. , @xmath14 , while @xmath15 describes a stiff perfect fluid ; both cases are associated with significant bifurcations , and we therefore refrain from discussing them . the einstein and matter field equations for these models are given by [ hphieq ] @xmath16 here an overdot signifies the derivative with respect to synchronous proper time , @xmath17 ; @xmath18 is the hubble variable , which is given by @xmath19 , where @xmath20 is the cosmological scale factor , and throughout we assume an expanding universe , i.e. @xmath21 , where @xmath18 is related to the expansion @xmath22 according to @xmath23 . we use ( reduced planck ) units such that @xmath24 , where @xmath25 is the speed of light and @xmath26 is the gravitational constant ( in the inflationary literature the gravitational constant @xmath26 is often replaced by the planck mass , @xmath27 ) . heuristically eq . can be viewed as an equation for an anharmonic oscillator with a friction term @xmath28 . this suggests that @xmath29 toward the future in an oscillatory manner , which is indeed correct . this qualitative picture , however , does not show how this comes about in a quantitative way , nor how the fluid affects the situation via its influence on @xmath18 . running the time backwards allows one to heuristically interpret @xmath28 as an energy input , which suggests that the scalar field oscillates with increasing amplitude toward the past , but this picture breaks down in the limit @xmath30 . even though this is beyond the planck regime , this limit is also needed in order to describe the classical behavior at early times after the planck epoch . furthermore , eq . yields that @xmath31 where @xmath32 an @xmath33 are constants , and hence @xmath34 at late times while @xmath35 at early times . note that the above qualitative considerations say nothing about how e.g. @xmath36 behaves asymptotically , i.e. , if the model is fluid or scalar field dominated , or neither , asymptotically . nor does the above say anything about the role of the so - called attractor solution in a global solution space setting . this exemplifies that there is a need for a more careful examination , which is illustrated by some previous heuristic considerations for a scalar field with a monomial potential by e.g. turner @xcite and mukhanov @xcite p. 242 , which in turn inspired the rigorous work by rendall @xcite ; in addition de la macorra and piccinelli introduced a new heuristic approach to study dynamics at late times for a scalar field with a monomial potential and a perfect fluid @xcite ; rigorous work in this context was also obtained for the special case @xmath37 by giambo and miritzis @xcite . nevertheless , this still leaves room for improvements and extensions , and , as will be shown in this paper , it is possible to shed light on interesting previously neglected physical and mathematical aspects . the main purpose of this paper in , primarily , mathematical cosmology is two - fold : firstly , to obtain a global visual picture of the solutions space , thus , e.g. , situating the so - called attractor solution in a global solution space context . secondly , to prove issues concerning asymptotical behavior at late and early times . this includes introducing averaging techniques to determine late time behavior , generalizing and simplifying earlier proofs in the literature , and using center manifold theory to rigorously derive approximations for the attractor solution at early times , as well as clarifying the physically important issue of asymptotic self - similarity . the outline of the paper is as follows . in the next section we introduce our new three - dimensional dynamical systems reformulation of the field equations on a relatively compact state space . we also present two other complementary dynamical systems formulations of the field equations , which allow us to effectively obtain approximations for the attractor solution . in section [ sec : globaldynsys ] we apply global and local dynamical systems techniques to obtain a complete and illustrative picture of the solution space and its properties , including asymptotics . in particular , we introduce averaging techniques in our global dynamical systems setting , which allows us to prove the following theorem : [ theorem ] it should be pointed out that similar conclusions have been reached heuristically with quite different arguments in e.g. @xcite . furthermore , giambo and miritzis gave a proof for @xmath37 for the cases ( i ) and ( ii ) in @xcite ( in the case of general relativity ) . however , apart from that our proof rigorously generalizes previous results , our method can , in principle , be modified to treat even more general situations . moreover , we tie our results to the global dynamical systems picture and discuss their physical implications , e.g. , situating them in the context of future manifest asymptotic self - similarity breaking . in section [ sec : attrappr ] we focus on the attractor solution , where we introduce and compare several approximation schemes , such as center manifold and slow - roll based expansions and pad approximants , in order to describe it quantitatively . finally , section [ sec : concl ] contains some general remarks , e.g. about the de sitter solution on the unphysical boundary of the state space . our main global ( i.e. compact ) dynamical systems formulation is based on the dependent variables @xmath38 , which are defined as follows : [ vardef1 ] @xmath39 where @xmath40 in addition it is of interest to define @xmath41 to introduce a new suitable time variable we note the following : at early times it is natural to use a hubble - normalized time variable @xmath42 defined by @xmath43 , due to that the expansion @xmath44 provides a natural variable scale when @xmath45 via the raychaudhuri equation , as further discussed in e.g. @xcite , and references therein ( in an inflationary context @xmath42 is often interpreted as the number of @xmath46-folds @xmath47 ) . at late times the square root of the second derivative of the potential , @xmath48 ( for simplicity we here incorporate @xmath49 into @xmath50 ) , provides a natural variable ( mass ) scale . due to the gauss constraint , which relates a scale given in @xmath50 to one given in @xmath18 according to @xmath51 , this scale can be expressed in terms @xmath18 according to @xmath52 which leads to a dimensionless time variable @xmath53 defined by @xmath54 , where the constant have the same dimension as @xmath55 . to incorporate these features in a global dynamical systems setting we introduce a new time variable @xmath56 that interpolates between these two regimes , @xmath57 where @xmath58 when @xmath59 and @xmath60 , respectively . the above leads to the following three - dimensional dynamical system for @xmath61 : has been used ubiquitously in the scalar field literature ( often denoted by @xmath62 ) , while @xmath63 was used in @xcite where it was denoted by @xmath64 , however , as far as we know , the variable @xmath65 and the independent variable @xmath56 are new . the reason for the name @xmath66 is that mathematically this variable plays a role that is reminiscent to that of hubble - normalized shear , which is usually denominated by @xmath67 in anisotropic cosmology , for a number of situations ( the subscript @xmath68 follows the notation in @xcite ) . thus the present nomenclature is designed to pave the way for eventually situating the present problem in a broader context than isotropic scalar field cosmology . ] [ dynsys ] @xmath69 where the deceleration parameter , @xmath70 , defined via @xmath71 , is given by @xmath72 where @xmath73 where the last equation follows from the gauss constraint , while the inequality is due to @xmath5 . above we have also introduced an effective equation of state parameter @xmath74 for the scalar field which is defined according to @xmath75 from the above relations it follows that @xmath76 . in addition it is of interest to give the following auxiliary evolution equation for @xmath77 : @xmath78 where @xmath79 the state space @xmath80 associated with eq . is given by a finite ( when @xmath81 deformed ) cylinder described by the invariant pure _ scalar field boundary subset _ , @xmath82 ( i.e. @xmath83 ) , and thus @xmath84 , which we denote by @xmath85 , and @xmath86 . from now on , we analytically extend @xmath80 to the state space @xmath87 by including the unphysical invariant submanifold boundaries @xmath88 and @xmath89 . although these boundaries are unphysical , we stress that it is essential to include them since they describe the past and future asymptotic states , respectively , of all physical solutions . note that @xmath90 ( i.e. @xmath91 ) and hence @xmath92 is an interior invariant subset , @xmath93 , which is just a straight line in the center of the cylinder , describing the flat flrw perfect fluid model without a scalar field ( this solution appears as a straight line in the present state space due to that it is a self - similar solution , where @xmath65 describes the temporal change in the dimensional variable @xmath18 ) . note also that the dynamical system is invariant under the discrete symmetry @xmath94 , leading to a double representation of the physical solutions when @xmath95 , which is a consequence of that the potential is invariant when @xmath96 . when @xmath97 , where the differentiability depends on @xmath98 , describing problems with potentials @xmath99 , where @xmath63 is to be replaced with @xmath100 in . ] to describe the dynamics on the scalar field boundary @xmath101 , where @xmath84 , it is useful to introduce a complementary global formulation , which is based on the following transformation of @xmath66 and @xmath63 : @xmath102 this leads to the following regular _ unconstrained _ two - dimensional dynamical system : [ dynsysb ] @xmath103 in this case the deceleration parameter @xmath70 is given by @xmath104 the system constitutes a generalization of the system used in @xcite . note that for @xmath97 the present @xmath22 variable is not the same @xmath22 as that in @xcite , which in turn was based on the variables used in @xcite . we here introduce two complementary dynamical systems on unbounded state spaces that are useful for describing the dynamics at early times . the first system is based on the dependent variables @xmath105 and the independent variable @xmath42 , where we recall that @xmath106 and @xmath42 are defined by @xmath107 where @xmath42 can be viewed as the number of @xmath46-folds @xmath47 , i.e. , @xmath108 . this leads to the dynamical system : [ dynsysearly ] @xmath109 where @xmath70 is still given by , . and @xmath110 this system coincides with eq . ( 16 ) in @xcite ; incidentally , this model was also the example discussed by linde in his paper `` chaotic inflation '' @xcite . ] it is also useful to consider auxiliary equations for @xmath77 and @xmath111 : [ omeqrtau ] @xmath112 the second complementary dynamical system concerns the dynamics on the scalar field boundary @xmath82 . expressed in terms of @xmath106 and @xmath42 the unconstrained system takes the form : [ dynsysbearly ] @xmath113 note that the above systems share the same equations as the previous ones with bounded state spaces on the invariant boundary subset @xmath114 . it follows from that @xmath115 on @xmath80 ( note that since we have assumed that @xmath116 it follows from eq . that @xmath117 only when @xmath82 and @xmath118 ) . due to that @xmath119 , it follows from and that @xmath65 is a monotonically increasing function on @xmath80 ( although eq . shows that solutions on the scalar field boundary have inflection points when @xmath120 ) and hence @xmath65 can be viewed as a time variable if one is so inclined . as a consequence all orbits ( i.e. solution trajectories ) in @xmath80 originate from the invariant subset boundary @xmath88 , which is associated with the asymptotic ( classical ) initial state , and end at the invariant subset boundary @xmath89 , which corresponds to the asymptotic future , and therefore all fixed points are located at @xmath88 and @xmath89 . in the state space of a dynamical system @xmath121 for which @xmath122 . ] it also follows from the monotonicity of @xmath65 that the past ( future ) attractor resides on @xmath88 ( @xmath89 ) . the equations on the subset @xmath88 ( or , equivalently @xmath123 ) are given by [ dynsyst0 ] @xmath124 as follows from , or , equivalently ( with @xmath56 replaced with @xmath42 ) , where @xmath70 is given by and . it follows that the state space on @xmath88 is divided into four sectors defined by the invariant subsets @xmath125 and @xmath126 . the intersection of these subsets with @xmath82 and with each other yield five fixed points on @xmath88 : [ dynsyst0 ] @xmath127 where @xmath128 are two equivalent fixed points for which @xmath82 and @xmath129 ( and @xmath130 ) , i.e. , they are associated with a massless scalar field state , while the two equivalent fixed points @xmath131 , for which @xmath82 and @xmath132 ( and @xmath118 ) correspond to a de sitter state . for some constant finite value of @xmath50 for which both @xmath133 and @xmath18 have constant , bounded , and positive values . in contrast the present de sitter states correspond to the limits @xmath134 , @xmath135 , and therefore reside on the unphysical boundary @xmath88 . ] the fixed point @xmath136 gives @xmath92 and @xmath137 and corresponds to the flat perfect fluid friedman model . as shown below , the two fixed points @xmath128 are sources on @xmath138 ; the fixed points @xmath131 are sinks _ on _ @xmath88 , but they also have one zero eigenvalue that corresponds to a one - dimensional unstable center submanifold on @xmath85 , i.e. , one solution , called an attractor solution , originates from each fixed point @xmath131 into @xmath80 on the pure scalar field boundary subset @xmath82 . the fixed point @xmath136 is a saddle that gives rise to a 1-parameter set of solutions entering @xmath80 ( the associated unstable tangent space is given by @xmath125 ) , one being the perfect fluid solution given by @xmath139 . the system admits the following conserved quantity when @xmath140 : @xmath141 which determines the solution trajectories on the @xmath88 subset , see figure [ fig : t0 ] . the equations on the subset @xmath89 are given by [ dynsyst1 ] @xmath142 this system has a non - hyperbolic fixed point , @xmath143 with three zero eigenvalues . fortunately @xmath144 resides at the intersection of two invariant subsets : the invariant subset @xmath89 and the invariant perfect fluid subset @xmath139 , and thus , since @xmath65 is monotone in @xmath80 , @xmath144 attracts at least this orbit . _ on _ @xmath89 @xmath144 is conveniently analyzed by considering eq . on @xmath89 , which yields that @xmath145 which is an integral of , i.e. , the subset @xmath89 is foliated with periodic orbits surrounding the fixed point @xmath144 , see figure [ fig : t1 ] . to make further progress as regards the global properties of the solution space we need to consider the asymptotic dynamics at early and late times . the linearization of the system at the fixed points @xmath136 and @xmath128 is conveniently described as follows : 2 . |__0 & = _ m , & . | _ _ & = , + . |__0 & = -(2 - _ m ) , & . _ & = 3(2 - _ m ) , + . |__0 & = _ m , & . | _ _ & = , where the right hand sides constitute the eigenvalues of the fixed points . hence @xmath136 is a hyperbolic saddle , with an unstable manifold tangential to @xmath125 , i.e. , there is a 1-parameter set of solutions that originate from @xmath136 entering the state space @xmath146 tangentially to @xmath125 . the fixed points @xmath128 are hyperbolic sources ; it follows from the invariant submanifold structures that there exists a 2-parameter set of solutions entering the interior of the cylinder with @xmath147 from each fixed point @xmath128 , while a 1-parameter set of solutions originate from each fixed point @xmath128 into the boundary subset @xmath85 . _ on _ the subset @xmath88 linearization of the system gives @xmath148 and hence @xmath149 are hyperbolic sinks on @xmath88 , as illustrated in figure [ fig : t0 ] . in the full state space , however , each equivalent fixed point @xmath149 have an additional zero eigenvalue associated with a one - dimensional so - called center manifold . fortunately , the center direction lies on the @xmath85 subset , and hence we can investigate the center manifold by means of the unconstrained system , which will be done in section [ sec : attrappr ] . there we show that the center manifold of each ( equivalent ) fixed point @xmath149 corresponds to a single solution that enters the state space @xmath85 ( we will even obtain approximate expressions for this solution ) , and this solution , which hence resides on @xmath150 , is what is often referred to as the ` attractor ' solution . in the full state space @xmath138 the fixed points @xmath131 are thus center - saddles . from these considerations , in combination with the monotonicity of @xmath65 , it follows that all solutions are _ past asymptotically self - similar _ in the sense that all physical geometrical scale - invariant observables , such as the deceleration parameter @xmath70 , are asymptotically constant . however , there is a twist to this . the geometry of a flat flrw pure perfect fluid cosmology with a linear equation of state and the geometry of a pure massless scalar field are geometries that admit a proper spacetime transitive homothety group , and such spacetimes are invariant under scalings of the spacetime coordinates . this is the underlying reason why they can be represented by fixed points , but not all fixed points are associated with geometries admitting proper spacetime transitive homothety groups , as exemplified by the de sitter spacetime . for these spacetimes homothetic scale - invariance is broken by the dimensional cosmological constant , but the 1-parameter set of de sitter spacetimes ( parameterized by @xmath151 ) admits a scaling ( self - similar ) property that scales @xmath151 ( i.e. , a scaling that maps one de sitter spacetime to another with a different @xmath151 ) , and it is due to this scaling property de sitter spacetimes can appear as fixed points . in the present case the fixed points are not in the interior physical state space , but on the unphysical boundary , but they are nevertheless characterized by e.g. the same value of @xmath70 as the associated physical spacetime . the fact that they in the present context appear on the unphysical boundary has consequences , which we will come back to in a discussion about the de sitter fixed points in section [ sec : concl ] . finally we point out that , apart from a set of measure zero , all solutions originate from a massless scalar field state and hence the present models are _ past generically massless scalar field dominated_. the equations on the @xmath89 subset , i.e. equation , are equivalent to that of when setting @xmath152 , i.e. this problem is _ exactly _ that of an anharmonic oscillator ( when @xmath97 ; for @xmath37 the problem is that of a harmonic oscillator ) . this can be seen from eq . , which yields @xmath153 we will now apply the approximate ideas in mukhanov @xcite to the present exact problem of an anharmonic oscillator . we therefore first multiply the above equation with @xmath63 and rewrite it as @xmath154 each periodic orbit is characterized by a constant value of @xmath77 and has an associated time period @xmath155 . the time average of a function @xmath156 over a period for a periodic orbit characterized by @xmath77 is given by @xmath157 taking the time average of eq . for a periodic orbit gives @xmath158 where we for notational convenience from now on drop the subscript @xmath77 . again , note that in contrast to the result in @xcite , this is an exact relation _ on _ the subset @xmath89 . using this result on @xmath89 for a periodic orbit in the definition of @xmath74 yields @xmath159 which together with leads to @xmath160 on the subset @xmath89 , i.e. , @xmath161 is independent of @xmath77 . it therefore follows that on average , in the above sense , e.g. @xmath37 and @xmath110 on @xmath89 correspond to dust and radiation , respectively . note that the result coincides with the approximate heuristic results using proper time given in @xcite and @xcite ; see also @xcite for a quite different precise definition of @xmath162 , which still yields . before continuing it is instructive to consider a model that consists of two perfect fluids with constant equation of state parameters @xmath163 and @xmath164 . then @xmath165 ( this expression follows from that @xmath166 , but it can also be obtained from the equation @xmath167 ) . since it is not difficult to show that @xmath168 toward the future it follows that @xmath169 if @xmath170 ; @xmath171 if @xmath172 ; @xmath173 if @xmath174 , i.e. , the fluid with the softest equation of state dominates toward the future . _ assuming _ that asymptotically @xmath74 can be replaced with the asymptotic averaged result @xmath175 in eq . results in @xmath176 which suggests theorem [ theorem ] , given in the introduction , but this is of course no proof . next we introduce averaging techniques that are subsequently used to provide the proof of this theorem . standard averaging techniques and theorems can be found in chapter 4 in @xcite ( the periodic case ) and in @xcite ( the general case ) . in standard averaging theory , a perturbation parameter @xmath177 plays the key role : roughly speaking , a differential equation of the form @xmath178 for @xmath179 is approximated by the averaged equation at @xmath180 , i.e. , @xmath181 , where the average @xmath182 is defined in eq . . furthermore , the error @xmath183 has to be controlled . in the problem at hand , we consider the differential equations in the variables @xmath184 and @xmath65 , where the role of the parameter @xmath177 is played by @xmath185 . therefore , after setting @xmath186 we have to prove an averaging theorem for the case where @xmath177 is not a parameter , but a variable that slowly goes to zero . the evolution equation of @xmath184 is given in eq . , which in terms of @xmath177 takes the form @xmath187 where @xmath188 . this formulation is problematic due to that @xmath189 is not well - defined when @xmath77 is zero . we therefore use the following formulation : [ ophieps ] @xmath190 where @xmath191 solves the system , @xmath192 , and @xmath193 the general idea of averaging is to express @xmath184 as @xmath194 and prove that the evolution of the variable @xmath64 is approximated at first order by the solution @xmath195 of the averaged equation . for that , consider the average as defined in eq . of the right hand side of eq . . more precisely , considering an equation of the form @xmath196 with @xmath56-periodic @xmath197 of period @xmath198 , the averaged equation is given by @xmath199 , where @xmath200 . according to , we have @xmath201 , where @xmath202 is a constant that does not depend on @xmath77 . hence the averaged equation reads @xmath203 while @xmath204 in eq . will be chosen appropriately in the proof . to prove theorem [ theorem ] we first re - express the theorem in terms of @xmath77 : * if @xmath205 , initial conditions with positive @xmath206 and @xmath177 converge for @xmath207 to the outer periodic orbit with @xmath84 tangentially to the slice @xmath208 . * if @xmath209 , initial conditions with positive @xmath210 and @xmath177 converge for @xmath207 to the center with @xmath90 tangentially to the slice @xmath208 . * if @xmath211 , each periodic orbit on the slice @xmath212 attracts a 1-parameter set of trajectories with positive initial @xmath177 . let us first derive a differential equation for @xmath64 by taking the time derivative of eq . : @xmath213 on the other hand , @xmath214 let us now set @xmath215 note that for large times @xmath216 is well approximated by periodic functions with an average @xmath217 . the right hand side of is for large times almost periodic and has an average that is zero so that the variable @xmath204 is bounded . as a consequence , the differential equation for the variable @xmath64 takes the form @xmath218 using the fact that @xmath219 and that @xmath220 results in the following : @xmath221 next we have to prove that the solution @xmath64 of this equation and the solution @xmath195 of the averaged equation have the same asymptotics when @xmath222 . since the averaged equation is expected to govern the dynamics , we first study the late time behavior of the system [ sysaverage ] @xmath223 after euler multiplication by @xmath224 ( or equivalently , a singular change of time variable @xmath225 ) , this system reads @xmath226 in cases ( i ) and ( ii ) , for which @xmath227 , the two fixed points of this system are located at @xmath228 and @xmath229 . the line @xmath230 is a heteroclinic orbit between these two fixed points , whose direction depends on the sign of @xmath231 . undoing the euler multiplication does not affect the trajectories with positive @xmath177 . on the other hand , in the original averaged system , the line @xmath212 is a line of fixed points . solutions with positive initial @xmath177 will approach the line of fixed points at @xmath180 , but slowly slide along this line as @xmath207 in the direction prescribed by the sign of @xmath231 , and go to the left or right fixed point accordingly . next we prove that the variables @xmath64 and @xmath77 follow this evolution . first note that the sequences @xmath232 , @xmath233 , defined as follows , @xmath234 have limits @xmath235 since @xmath236 goes to zero when @xmath56 goes to infinity . for a sufficiently small @xmath179 , eq . guaranties that @xmath64 is monotone ( in- or decreasing , depending on the sign of the quantity @xmath231 ) and bounded . hence @xmath237 must have a limit when @xmath238 . next we estimate @xmath239 , where @xmath64 and @xmath195 are trajectories with identical ` initial ' conditions at time @xmath240 : @xmath241 for @xmath242 , where we used the fact that @xmath243 with @xmath244 constant . applying gronwall s lemma ( see lemma 4.1.2 in @xcite , or p. 37 in @xcite ) , results in @xmath245 hence for @xmath246 $ ] , i.e. @xmath247 $ ] , the inequality @xmath248 holds for a positive constant @xmath249 . letting @xmath98 go to infinity implies that @xmath250 tends to zero when @xmath56 goes to infinity . therefore @xmath64 and @xmath195 have the same limit as @xmath238 , i.e. 0 or 1 depending on the sign of the quantity @xmath251 . finally , recall that @xmath252 ; from the triangle inequality , and @xmath253 when @xmath207 , it follows that @xmath184 also converges to 0 or 1 according to the sign of @xmath251 . this completes the proof of the theorem for the non - critical cases ( i ) and ( ii ) for which @xmath254 . let us now consider the critical case ( iii ) , for which @xmath255 . in this case the right hand side of the averaged equation vanishes . furthermore , the evolution equation for @xmath64 becomes @xmath256 let us first consider the average of the right hand side . we therefore define @xmath257 and compute its average at @xmath180 , where @xmath258 is a periodic function and @xmath259 , @xmath260 note that in the critical case , @xmath261 . we therefore obtain the following averaged system , expected to give the leading order approximation : [ secondav ] @xmath262 again we study the dynamics on a euler multiplied version of this system , i.e. in the time variable @xmath263 : [ eulersecondav ] @xmath264 the linearization at the line of fixed points @xmath208 is @xmath265 this linearization admits one eigenvalue that is zero with a corresponding eigenvector that is parallel to the line of fixed points , and a stable ( i.e. negative ) eigenvalue @xmath266 , with a corresponding eigenvector @xmath267 pointing toward the inside of the cylinder ( i.e. the line of fixed points of the system is transversally hyperbolic ) . hence each fixed point @xmath268 has a one - dimensional stable manifold , i.e. , there exists a trajectory @xmath269 that solves the averaged system with a positive initial @xmath177 and converges to @xmath268 for each @xmath270 as @xmath271 . this asymptotic behavior is not affected by transforming the equations back to the time variable @xmath56 . a straightforward estimation of the term @xmath272 provides bootstrapping sequences @xmath233 and @xmath232 , defined similarly as in the non - critical case . in other words , we obtain a pseudo - trajectory @xmath273 of the original system with @xmath274 , \qquad \vert \omega_\phi^n(\bar{\tau } ) - \bar{z}(\bar{\tau})\vert < k \varepsilon_n,\ ] ] for a constant @xmath249 . by regularity of the flow and compactness of the cylinder , there is an initial data whose trajectory @xmath275 under the flow of the original equation shadows the above pseudo - trajectory in the following sense : @xmath276 , \qquad \vert \omega_\phi^n(\bar{\tau } ) -\omega_\phi ( \bar{\tau})\vert < k \varepsilon_n.\ ] ] by the triangle inequality , @xmath277 as @xmath238 ; therefore , for each @xmath278 $ ] , there exists a trajectory that is limiting to the periodic trajectory at @xmath180 , characterized by @xmath279 . translating these results into the state space of concludes the proof of case ( iii ) . expressing these results for orbits with @xmath280 and @xmath86 in terms of formal global future attractors @xmath281 of the global dynamical system leads to : * @xmath282 if @xmath283 , * @xmath284 if @xmath285 , * @xmath286 if @xmath287 . the above results are illustrated by the numerical examples that are depicted in figure [ fig : latetime ] . the present results have physical consequences . the solutions approach @xmath288 toward the future , which is the minkowski solution of the system . the minkowski solution admits an 11-dimensional homothety group and is therefore an example of a self - similar spacetime . however , the asymptotic future state for solutions in case ( i ) , with @xmath283 and @xmath289 , is the periodic orbit at @xmath290 , and hence , since the future asymptotic behavior is described by a limit cycle , it follows that the deceleration parameter @xmath70 oscillates toward the future ( this is also obviously true for the orbits on the subset @xmath84 ) . the solutions thus approach the minkowski spacetime in a foliation that is induced by the present models in a manner where the self - similar nature of the minkowski spacetime is not manifest and thus we say that these models exhibit _ future asymptotic manifest self - similarity breaking_. take asymptotic constant values , but since the future attractor is a limit cycle it follows that @xmath70 is asymptotically oscillating , i.e. , asymptotic manifest self - similarity is broken . for another example of future asymptotic self - similarity breaking in cosmology , see @xcite . note that asymptotic manifest self - similarity is a complicated issue in the present asymptotic minkowski case due that the only metrics with no conformal scalars ( and hence the only admitting a local conformal group not conformally isometric ) are either conformal to the plane wave metric with parallel rays or conformally minkowski , see @xcite . ] physically , case ( ii ) , for which @xmath285 , has the simplest asymptotic regime . the future state for this case when @xmath147 is the fixed point @xmath144 for which @xmath291 . this state describes the flat perfect fluid model with equation of state parameter @xmath292 . since this model is self - similar , the solutions asymptotically approach the minkowski spacetime in a manifest self - similar manner and we therefore say that the present class of models are _ future asymptotically _ ( manifestly ) _ self - similar_. case ( iii ) , where @xmath287 , implies that ( with the exception of the perfect fluid solution with @xmath90 ) all solutions oscillate asymptotically toward the future and thus this case is also characterized by future asymptotic manifest self - similarity breaking . since asymptotic manifest self - similarity , and breaking thereof , plays an important role in physics as a whole , it is of considerable interest to note the above features . furthermore , recall that the case @xmath293 is equivalent to a pure scalar field case in an open flrw model . as a consequence _ all _ monomial potentials ( with @xmath294 ) for these models lead to that the future end state is given by the fixed point @xmath144 , which in this context is associated with a milne state . we are now finally in a position to describe the global solution space and its features . there is a 1-parameter set of solutions that originate from the fixed point @xmath136 , corresponding to an initial perfect fluid dominated state , into the state space @xmath80 ; two equivalent 2-parameter sets of solutions that originate from each of the equivalent fixed points @xmath128 , corresponding to a self - similar initial massless scalar field dominated state , where two equivalent 1-parameter subsets , belonging to each of these sets of solution , reside on the scalar field boundary @xmath295 ; finally , there are two equivalent attractor solutions residing on @xmath295 that originate from each of the fixed points @xmath131 , thus exhibiting an initial de sitter state in the sense previously discussed . the future state of all orbits in @xmath80 resides on the subset @xmath89 , where the precise asymptotic location on @xmath89 depends on the relation between the monomial potential exponent @xmath296 and the equation of state parameter @xmath292 according to theorem [ theorem ] . in all three cases ( i ) , ( ii ) , and ( iii ) there exists an open set of solutions that is close to the attractor solutions at some intermediate stage of their evolution , but there also exists an open set of solutions that is not close to the attractor solutions furthermore , although there exists an open set of solutions that is close to the attractor solutions at late times in case ( i ) this is not true for case ( ii ) where each ( equivalent ) attractor solution only acts as a kind of ` saddle . ' in addition , note that the variables @xmath297 can be locally expressed in terms of the physical quantities @xmath298 . by imposing a euclidian measure on the space described by @xmath298 one finds that ` most ' of the evolution of ` most ' solutions is not governed by the ` attractor ' solution ; the addition of a degree of freedom , in this case associated with a perfect fluid , has aggravated the situation for arguments that attempt to establish that attractor solutions in some sense are attractors . nevertheless , attractor solutions are likely to continue to generate considerable attention , and thus there is a need to describe them accurately , which is the topic of the next section . in this section we will introduce and compare several analytical approximation schemes that describe the attractor solution . since the two equivalent attractor solutions are just the center manifolds of @xmath131 , we begin with a direct approach of approximately obtaining these center manifolds by means of center manifold theory . for presentations of center manifold analysis , see e.g. , @xcite , and for applications in cosmology , e.g. , @xcite . the equivalent center manifolds of @xmath131 reside on the scalar field subset @xmath82 and therefore it suffices to study the equations on this subset . in @xcite it was shown that for quadratic potentials the center manifold expansion using the bounded system for @xmath37 resulted in a larger range than the expansion based on the unbounded system . however , it was also shown that by using these expansions to produce so - called pad approximants resulted in much better approximations , both as regards accuracy and range . moreover , the pad approximants for the system gave the same expressions as the pad approximants obtained from ( or , more accurately , they produced the converging subset of pad approximants associated with ) , but in a simpler form . thus the most useful results from center manifold analysis are obtained by performing such an analysis for the unbounded system , followed by an introduction of pad approximants , instead of using . due to the discrete symmetry , we can , without loss of generality , restrict the center manifold analysis to the fixed point @xmath299 at @xmath300 . a linearization of the system at this fixed point yields [ linearanalysis ] @xmath301 where @xmath302 and @xmath303 denote the tangential stable and center subspaces , respectively . since the tangential center subspace is given by @xmath304 , we introduce @xmath305 as a new variable that replaces @xmath22 in order to study the center manifold @xmath306 , with the tangent space @xmath303 at @xmath307 . as follows from , this leads to the transformed system [ dynsysv ] @xmath308 where @xmath309 . the linearization of eq . yields @xmath310 , while eq . only has higher order terms . the center manifold @xmath306 can be obtained as the graph @xmath311 near @xmath307 ( i.e. , use @xmath106 as an independent variable ) , where @xmath312 ( fixed point condition ) and @xmath313 ( tangency condition ) . inserting this relationship into eq . and using @xmath106 as the independent variable leads to @xmath314 where @xmath315 . solving the above nonlinear ordinary differential equation amounts to finding the attractor solution , which for the present class of problems does not seem likely to be possible . instead the equation can be solved approximately by representing @xmath316 as a formal truncated power series and by taylor expanding the expressions involving @xmath22 ( and subsequently replace @xmath22 with its power series expression in @xmath106 ) , which makes it possible to algebraically solve for the coefficients in the formal series . before doing this , it is useful to note that all coefficients in the above equation are odd in terms of @xmath106 and @xmath22 , and as a consequence the power series for @xmath317 consists of odd powers of @xmath106 , since it is only then @xmath318 results in even powers ( odd powers for @xmath318 , and hence even powers for @xmath317 , are clearly zero due to the above properties ) . furthermore , it follows from the tangency condition that the series for @xmath317 in odd powers of @xmath106 begins with a cubic term . hence @xmath319 where the series for @xmath320 is truncated at some chosen order and where we have introduced @xmath321 . inserting this into a taylor expansion of eq . and algebraically solving for the coefficients leads to @xmath322 , \end{split}\ ] ] where we have chosen to truncate the series for @xmath22 at 5th order in @xmath106 . to improve the range and accuracy of the above approximation for the attractor solution we construct the @xmath323_{f}(x)$ ] pad approximant , which leads to the following approximate expression for @xmath22 : of a function @xmath324 , denoted by @xmath325_f(x)$ ] , is associated with a truncated taylor series @xmath326 and given by the polynomials @xmath327 and @xmath328 according to @xmath325_f(x ) = \frac{p_m(x)}{q_n(x)}$ ] , such that @xmath329 , where coefficients with the same powers of @xmath62 are equated up through @xmath330 . for details and examples , see @xcite and references therein . ] @xmath331_{f}(x ) , \qquad [ 1/1]_{f}(x ) = \frac{1-\frac{n}{36(3n-7 ) } \left(\frac{7}{2}n^2 - 43n+\frac{2753}{30}\right)x}{1-\frac{n}{36(3n-7 ) } \left(\frac{13}{2}n^2 - 57n+\frac{1081}{10}\right)x } .\ ] ] when this is subsequently expressed in @xmath65 , i.e. , @xmath332 and @xmath333 , this yields a curve @xmath334 which approximates the attractor solution in the state space @xmath80 on the boundary @xmath82 . the case @xmath37 was dealt with in figures 5 and 14 in @xcite , where these approximations curves for the attractor solution , as well as those obtained by considering higher order expansions , were compared with the numerically computed attractor solution . to obtain explicit curves in our state space picture we need to specify @xmath98 . to avoid details , we in this paper only compare the approximate solution curves with the numerically computed attractor solution for some representative values of @xmath98 for the above @xmath323_{f}(x)$ ] pad approximant in subsection [ sec : comp ] below ; although it should be pointed out that higher order pad approximants give better , although more complicated , results . ref . @xcite also illustrates that nonlinear transformations might have two important features . first , they affect analytical or numerical computations and therefore a suitable choice of variables can make a problem more tractable . second , they affect approximations and a suitable choice of variables can lead to better approximations for solutions like the attractor solution . we will therefore next consider an approximation scheme based on the variables @xmath66 and @xmath63 instead of @xmath22 . as in the previous case we use a system adapted to the dynamics at early times and we therefore use @xmath106 , but @xmath22 is replaced with @xmath66 and @xmath63 , i.e. , we consider the system . we are interested in finding new approximations for the attractor solution , and inspired by the result in eq . we assume that @xmath66 and @xmath63 can be written as formal truncated series in @xmath106 . these series are subsequently inserted in , but since we are interested in the attractor solution , which resides on the @xmath82 boundary , we also require that the constraint @xmath335 is satisfied . the analysis is simplified by noting that it follows from the system that @xmath66 must have only odd terms while @xmath63 have only even terms in their series expansions in @xmath106 . moreover , since we choose , without loss of generality , to consider the solution that originates from the @xmath299 fixed point it follows that to lowest order @xmath336 . algebraically solving for the coefficients , and writing @xmath337 as before , leads to [ sigmaxtildet ] @xmath338 where we have chosen to truncate the series for @xmath324 and @xmath339 at 2nd order in @xmath62 . to improve the range and accuracy we calculate the @xmath323_{f}(x)$ ] and @xmath323_{x}(x)$ ] pad approximants , which yield the following approximate expressions [ sigmaxpade ] @xmath340,\label{seriessigma}\\ x & \approx \frac{1+\frac{n}{36}\left(3 - 2n\right)x}{1+\frac{n}{36}\left(5 - 2n\right)x}.\label{xpade}\end{aligned}\ ] ] note that for small values of @xmath106 in the neighborhood of @xmath299 we obtain @xmath341 , which gives @xmath342 , which is just the tangency condition for the center manifold of @xmath299 . replacing @xmath106 and @xmath62 with @xmath65 in yield two curves in the state space @xmath80 on the scalar field subset @xmath82 , one for the approximation of @xmath63 and one for the approximation of @xmath66 , both approximating the attractor solution . to obtain accurate numerical results for the attractor solution it is preferable to use the unconstrained system rather than subjected to the constraint @xmath343 , especially when @xmath285 since the constraint surface then becomes unstable for sufficiently large @xmath65 . the above results are translated into the corresponding @xmath334 curves via @xmath344 ( the minus sign is due to that @xmath345 and taking the default branch ) and implicitly via @xmath346 , where @xmath347 and @xmath348 are obtained by replacing @xmath62 with @xmath65 in and , which lead to approximations that subsequently can be compared with the numerically computed attractor solution . below , in subsection [ sec : comp ] we for brevity only discuss the pad approximant given in eq . for some representative values of @xmath98 . as a final remark , note that the lowest order expansion just gives @xmath349 , which is equivalent to @xmath350 , since @xmath351 on the scalar field boundary . the condition @xmath350 and @xmath349 describes a straight vertical line in @xmath87 , corresponding to @xmath132 , originating from @xmath299 , which apart from @xmath299 is not a good approximation for the attractor solution . this is in contrast with the previous center manifold expansion , which is given by a unique curve @xmath334 , which to all orders in the limit of small @xmath65 is tangent to the center manifold . next we turn to approximations based on the slow - roll approximation and its extensions . we here extend the work in @xcite and @xcite to general @xmath98 and to higher order by approximating @xmath352 with so - called slow - roll hubble expansions , and illustrate this class of approximations in our global state space picture . to facilitate comparison with the inflationary literature , we initially keep the coupling constant @xmath353 in the slow - roll hubble expansion formulas below . the approach developed in @xcite is based upon a hierarchy of slow - roll parameters , the first being @xmath354 which are assumed to be small . these parameters allow one to produce a truncated expansion for the hubble variable in terms of @xmath50 , which for the present monomial scalar field potential is given by ( following the prescription given in @xcite , which we refer to for details ) : @xmath355 where we have found it convenient to define a quantity @xmath64 according to @xmath356 note that the 2nd ( 3rd ) order term is zero in when @xmath110 ( @xmath357 ) . the above series expansion contains no parameters and thus it can only describe a single solution , moreover , for the series expansion to make sense requires that @xmath64 is small and hence that @xmath50 is large , which leads to that @xmath18 is large , i.e. , for the present models the hubble slow - roll expansion attempts to approximately describe the dynamics at early times for a certain solution , but which one ? in terms of @xmath65 and @xmath63 , and therefore in @xmath65 and @xmath22 , the definitions in eq . lead to that the hubble expansion can be written on the form ( from now on we set @xmath358 ) @xmath359 ( as usual , @xmath360 ) where each order yields an implicit relation for a curve on the scalar field boundary , although it can be used to provide a series expansion for @xmath63 in terms of @xmath106 for small @xmath106 and hence for small @xmath65 . note that just as in the previous approximation scheme , the zeroth order hubble expansion just gives @xmath361 , which , as stated earlier , is not a good approximation for the attractor solution . nevertheless , this approximation is what is sometimes referred to as the slow - roll hubble approximation , see e.g. @xcite . at first order , when @xmath362 , eq . results in @xmath363 without loss of generality , consider the neighborhood of @xmath299 ; then the above expression yields the limit ( beyond the @xmath299 point itself ) @xmath364 which is the tangency condition for the center submanifold of @xmath299 . thus all curves associated with eq . originate from @xmath131 ; furthermore , all curves associated with orders larger than zero are tangential to the attractor solutions in the small @xmath65-limit toward @xmath131 . to improve the range and rate of convergence one can compute pad approximants for the hubble expansion for @xmath365 , i.e. , @xmath366_f(y ) = v(\phi)[l / m]_f(y),\ ] ] where [ hsrpades ] @xmath367_f(y)&=\frac{1+\frac{2}{n}y}{1-\left(1-\frac{2}{n}\right)y } , \\ [ 2/2]_f(y)&=\frac{1-\frac{5}{2n}- \left(1-\frac{11}{n}+\frac{22}{n^2}\right)y+\frac{8}{n^2 } \left(1-\frac{43}{16n}\right)y^2}{1-\frac{5}{2n}-2\left(1-\frac{27}{4n}+ \frac{11}{n^2}\right)y+\left(1-\frac{9}{n}+\frac{25}{n^2}-\frac{43}{2n^3}\right)y^2}.\end{aligned}\ ] ] in ref . @xcite , these approximations have been compared with the numerically computed attractor solutions in detail for @xmath37 , see fig . 7 and table 3 of @xcite . let us now turn to what is referred to as _ the slow - roll approximation _ in e.g. @xcite . this approximation is obtained by inserting the lowest order hubble expansion approximant @xmath368 into @xmath369 which yields @xmath370 in the present case with a monomial potential . expressed in terms of the variables @xmath66 and @xmath63 , this gives @xmath371 which when squared leads to @xmath372 note that this expression coincides with the first order hubble expansion approximation given in eq . , and thus the slow - roll approximation is an approximation to the attractor solution , since it yields a curve that is tangential to the center manifold in the limit of small @xmath65 . apart from the zeroth order approximation , hubble expansion slow - roll approximations , as well as all slow - roll hubble expansion based approximants , are approximations for the attractor solution_. the slow - roll approximation can be generalized by inserting higher order hubble approximants into @xmath365 . a complication arises since different approximations can be obtained by multiplying expressions by various combinations of @xmath373 ( they are all tangential to the center manifolds in the limit of small @xmath65 ) ; we here choose a combination that yields the above slow - roll approximation ( which was chosen to coincide with the first order expression in ) when setting @xmath374 below : @xmath375 inserting the rational approximants into leads to the corresponding slow - roll approximants based on the pad approximants constructed from the hubble expansion . in ref . @xcite the corresponding curves in the state space @xmath80 on the scalar field subset @xmath82 are given for @xmath37 in fig . 10 and the associated relative errors at the end of inflation at @xmath376 are given in table 4 . next we will compare some of the simplest , and thereby the most easily applicable , approximations for each of the above schemes . as seen from the above discussion , series expansion approximants and slow - roll approximation schemes lead to ambiguities , i.e. , different possible types of approximations . this is to be contrasted with the center manifold analysis , which yield a systematic scheme for obtaining a unique sequence of converging approximations for the attractor solution . on the other hand , the strength of slow - roll approximation schemes is that they can be applied to any potential and problem that satisfies the slow - roll conditions ; it remains to be seen to what extent this can be accomplished with other methods like center manifold analysis , an issue we will return to elsewhere . to keep the discussion reasonably short , we here restrict it to comparing only some of the fairly simple approximations , namely : ( i ) the slow - roll approximation given by @xmath377 , ( ii ) the center manifold pad approximant @xmath378_{f(x)}$ ] given in eq . , which we will refer to as @xmath323_\theta$ ] , and ( iii ) the series expansion pad approximant for @xmath66 given in eq . , which we will refer to as @xmath323_{\sigma_\dagger}$ ] . the accuracy of the different approximations for the attractor solution , which is computed numerically , is shown in figure [ fig : errorscomparisons ] , while the relative errors at @xmath376 are given in table [ tab : comparisons ] . by @xmath379 , where the subscript @xmath380 stands for the numerically computed attractor solution while the subscript @xmath381 stands for the approximant . ] .relative hubble errors @xmath382 at @xmath376 for the slow - roll @xmath377 , and the @xmath323_\theta$ ] and @xmath323_{\sigma_\dagger}$ ] pad approximants . [ cols="^,^,^,^,^,^,^,^,^,^,^,^,^ " , ] all the presently discussed approximations , including the slow - roll approximation @xmath377 , have quite small errors when @xmath383 , which is a reflection of that they are all tangential to the attractor solution when @xmath384 . for larger @xmath70 the slow - roll approximation give fairly good results for small @xmath98 while it becomes less competitive for steep potentials with large @xmath98 ( recall that @xmath385 in the de sitter fixed point limit ) , as can be expected however , for small @xmath98 the slow - roll approximation lead to fluctuations in accuracy when @xmath70 is no longer close to @xmath386 , indicating that it sometimes might be rather hard to predict how good the slow - roll approximation actually is going to be in a given situation . the above suggests that the slow - roll approximation , as well as its more accurate but more complicated higher expansion slow - roll based approximants , should be used with care when it comes to precision cosmology . it should also be pointed out that all approximations get worse for large @xmath98 , hinting at that steep potentials might pose a challenge for approximative methods ( although the physical viability of such models might be questioned in view of recent observational data ) . however , the @xmath323_{\sigma_\dagger}$ ] pad approximant does not get worse so fast as the other approximations and for @xmath387 it is the best approximation of the three . this suggests that a good analytic description , pertinent in the context of precision cosmology , might require the development of a variety of complementary approximation methods that can be applied to a range of different problems ; we emphasize that the present problem is just an illustrative stepping stone for such developments . in this paper we have used the inherent physical scales associated with the problem at hand to produce new variables . in particular we obtained a regular dynamical system on a compact state space @xmath138 . this allowed us to give a pictorial description of the entire solution space and moreover give proofs concerning the solutions asymptotic properties , which in turn were tied to physically important features such as asymptotic manifest self - similarity and breaking thereof . in addition we showed that the so - called attractor solution is just the unstable center manifold of a de sitter state on the unphysical past boundary subset on @xmath138 . we also found it convenient to introduce several auxiliary dynamical systems , adapted to the physical properties associated with asymptotic past dynamics , since this allowed us to efficiently compute approximations for the attractor solution by means of several complementary approximation schemes . the common feature of these schemes is that to lowest order toward the de sitter point ( apart from those approximations like the lowest order hubble expansion slow - roll approximation , which just give vertical lines at @xmath132 ) , from which the attractor solution originate , the approximations yield the tangent space of the center manifold . as a consequence they all lead to that the variable @xmath106 along the attractor solution in the limit of small @xmath106 is governed by the approximation @xmath388 this , in combination with the definition @xmath389 and that @xmath42 can be identified as @xmath390 , yields that when the scale factor @xmath391 then @xmath392^{\frac{n}{2 } } \rightarrow \infty,\ ] ] which is to be contrasted with the exact de sitter solution for which @xmath393 . finally , we would like to point out that in principle the methods we have introduced have a broader range of applicability than the presently studied models . for example , the use of averaging techniques should be applicable to the dynamics at late times for solutions that approaches a local minimum of a scalar field potential of the form @xmath394 where we for simplicity have translated the minimum of the potential so that it occurs at @xmath395 , and where @xmath396 obeys suitable differentiability conditions when @xmath397 . one might question the physical relevance of these potentials at very late times , but it should be pointed out that asymptotic dynamics can be used as pieces to provide approximations for global dynamics , as described in @xcite . aa is supported by the projects cern / fp/123609/2011 , excl / mat - geo/0222/2012 , and camgsd , instituto superior tcnico by fct / portugal through uid / mat/04459/2013 , and the fct grant sfrh / bpd/85194/2012 . furthermore , aa also thanks the department of physics at karlstad university , sweden , for kind hospitality . jh is supported by the dfg collaborative research center sfb647 space , time , matter . cu would like to thank the center for mathematical analysis , geometry and dynamical systems at the technical university of lisbon and the institut fr mathematik at freie universitt in berlin for kind hospitality . finally , it is a pleasure to thank prof . b. fiedler and dr . s. liebscher for useful suggestions . l. defrise - carter . conformal groups and conformally equivalent isometry groups . commun . math . * 40 * 273 ( 1975 ) . d. m. eardley . self - similar spacetimes : geometery and dynamics . phys . * 37 * 287 ( 1974 ) . g. n. remmen and s. m. carroll . attractor solutions in scalar - field cosmology . rev . * d88 * ( 2013).doi : 10.1103/physrevd.88.083518 j. j. sinou , f. thouverez and l. jezequel . extension of the center manifold approach , using rational fractional approximants , applied to non - linear stability analysis . nonlinear dynamics * 33 * 267 ( 2003).doi : 10.1023/a:1026060404109

turbulence is essential for understanding the structure and dynamics of molecular clouds and star - forming regions . there is a need for adequate tools to describe and characterize the properties of turbulent flows . one - point probability distribution functions ( pdf s ) of dynamical variables have been suggested as appropriate statistical measures and applied to several observed molecular clouds . however , the interpretation of these data requires comparison with numerical simulations . to address this issue , sph simulations of driven and decaying , supersonic , turbulent flows with and without self - gravity are presented . in addition , random gaussian velocity fields are analyzed to estimate the influence of variance effects . to characterize the flow properties , the pdf s of the density , of the line - of - sight velocity centroids , and of the line centroid increments are studied . this is supplemented by a discussion of the dispersion and the kurtosis of the increment pdf s , as well as the spatial distribution of velocity increments for small spatial lags . from the comparison between different models of interstellar turbulence , it follows that the inclusion of self - gravity leads to better agreement with the observed pdf s in molecular clouds . the increment pdf s for small spatial lags become exponential for all considered velocities . however , all the processes considered here lead to non - gaussian signatures , differences are only gradual , and the analyzed pdf s are in addition projection dependent . it appears therefore very difficult to distinguish between different physical processes on the basis of pdf s only , which limits their applicability for adequately characterizing interstellar turbulence . _ ( accepted for publication in the astrophysical journal ) _

astro-ph0001379

turbulence is an important ingredient for understanding the properties and characteristics of molecular clouds and star - forming regions . turbulent gas motions are highly supersonic as indicated by the superthermal line widths ubiquitously observed throughout molecular clouds ( williams , blitz & mckee 2000 ) . these motions carry enough energy to halt global collapse and act as stabilizing agent for the entire cloud . however , it can be shown that interstellar turbulence decays quite rapidly on time scales of the order of the free - fall time of the system ( mac low et al . 1998 , stone , ostriker & gammie 1998 , padoan & nordlund 1999 ) . to explain the observed long life times , turbulence in molecular clouds must be constantly driven ( gammie & ostriker 1996 , mac low 1999 ) . the interplay between self - gravity on the one hand ( leading to local collapse and star formation ) and turbulent gas motion on the other hand ( trying to prevent this process ) plays a key role in determining the structure of molecular clouds . altogether , understanding the characteristics of compressible , supersonic , and constantly replenished turbulence in self - gravitating media is an important ingredient for an adequate description of molecular clouds dynamics . and vice versa , from analyzing the spatial and dynamical structure of molecular clouds we can gain insight into the phenomenon of turbulence ( for an overview over interstellar turbulence see franco & carraminana 1999 ) . unfortunately , a complete and comprehensive theory of turbulence does not exist . due to the enormous complexity of the problem , progress has been slow since kolmogorov s pioneering work in 1941 , where he derived simple scaling laws for incompressible , stationary , and homogeneous turbulence by postulating a self - similar energy cascade downwards from the driving scale to the dissipation range . most effort has since been put in finding an adequate closure procedure , i.e. in finding a way to express the highest - order correlation in the hierarchy of equations governing turbulent motion ( for an excellent overview see lesieur 1997 ; also boratav , eden & erzan 1997 ) . however , a satisfying description of turbulence has yet to be found . correlation and distribution functions of dynamical variables are frequently deployed for characterizing the kinematical properties of turbulent molecular clouds . besides using 2-point statistics ( e.g. scalo 1984 , kleiner & dickman 1987 , kitamura et al . 1993 , miesch & bally 1994 , larosa , shore & magnani 1999 ) , many studies have hereby concentrated on 1-point statistics , namely on analyzing the probability distribution function ( pdf ) of the ( column ) density and of dynamical observables , e.g. of the centroid velocities of molecular lines and their increments . the density pdf has been used to characterize numerical simulations of the interstellar medium by vzquez - semadeni ( 1994 ) , padoan , nordlund , & jones ( 1997 ) , passot , & vzquez - semadeni ( 1998 ) and scalo et al . velocity pdf s for several star - forming molecular clouds have been determined by miesch & scalo ( 1995 ) and miesch , scalo & bally ( 1998 ) . lis et al . ( 1996 , 1998 ) analyzed snapshots of a numerical simulation of mildly supersonic , decaying turbulence ( without self - gravity ) by porter , pouquet , & woodward ( 1994 ) and applied the method to observations of the @xmath0-ophiuchus cloud . altogether , the observed pdf s exhibit strong non - gaussian features , they are often nearly exponential with possible evidence for power - law tails in the outer parts . this disagrees with the nearly gaussian behavior typically found in experimental measurements and numerical models of incompressible turbulence . the observed centroid velocity _ increment _ pdf s are more strongly peaked and show stronger deviations from gaussianity than numerical models of incompressible turbulence predict . furthermore , the spatial distribution of the largest centroid velocity differences ( determining the tail of the distribution ) appears ` spotty ' across the face of the clouds ; there is no convincing evidence for filamentary structure . miesch et al . ( 1998 ) conclude that turbulence in molecular clouds involves physical processes that are not adequately described by incompressible turbulence or mildly supersonic decay simulations ( see also mac low & ossenkopf 2000 ) . it is the principal goal of this paper to extend previous determinations of pdf s from numerical models into a regime more applicable for interstellar turbulence by ( 1 ) by calculating fully supersonic flows , ( 2 ) by including self - gravity , and ( 3 ) by incorporating a ( simple analytic ) description of turbulent energy input . for comparison with molecular cloud observations , i discuss the dynamical properties of decaying and stationary ( i.e. driven ) , supersonic , isotropic turbulence in self - gravitating isothermal gaseous media . the pdf s for the density , for the line centroid velocity and for their increments are derived as function of time and evolutionary state of the turbulent model . the structure of this paper is as follows : section [ sec : remarks ] introduces and defines the statistical tools applied in the current study . it is followed in sec . [ sec : model ] by a description of the numerical scheme used to compute the time evolution of the turbulent flows . [ sec : init ] shows that already simple variance effects in random gaussian fields are able to introduce strong _ non_-gaussian distortions to the pdf s which makes a clear - cut interpretation difficult . section [ sec : decay - non - grav ] contains the analysis of decaying , initially highly supersonic turbulence without self - gravity . this effect is then added to the simulations presented in sec . [ sec : decay - with - gravity ] . the model most relevant for molecular cloud dynamics is discussed in sec . [ sec : driven - with - gravity ] . it includes a simple driving term to replenish the turbulent cascade . finally , in sec . [ sec : summary ] all results are summarized . the kolmogorov ( 1941 ) approach to incompressible turbulence is a purely phenomenological one and assumes the existence of a stationary turbulent cascade . energy is injected into the system at large scales and cascades down in a self - similar way . at the smallest scales it gets converted into heat by molecular viscosity . the flow at large scales is essentially inviscid , hence for small wave numbers the equation of motion is dominated by the advection term . if the stationary state of fully developed turbulence results from random external forcing then one navely expects the velocity distribution in the fluid to be gaussian on time scales larger than the correlation time of the forcing , irrespectively of the statistics of the forcing term which follows from the central limit theorem . however , the situation is more complex ( e.g. frisch 1995 , lesieur 1997 ) . one of the most striking ( and least understood ) features of turbulence is its intermittent spatial and temporal behavior . the structures that arise in a turbulent flow manifest themselves as high peaks at random places and at random times . this is reflected in the pdf s of dynamical variables or passively advected scalars . they are sensitive measures of deviations from gaussian statistics . rare strong fluctuations are responsible for extended tails , whereas the much larger regions of low intensity contribute to the peak of the pdf near zero ( for an analytical approach see e.g. forster , nelson & stephens 1977 , falkovich & lebedev 1997 , chertkov , kolokolov & vergassola 1997 , balkovsky et al . 1997 , balkovsky & falkovich 1998 ) . for incompressible turbulence the theory predicts velocity pdf s which are mainly gaussian with only minor enhancement at the far ends of the tails . the distribution of velocity _ differences _ ( between locations in the system separated by a given shift vector @xmath1 ) is expected to deviate considerably from being normal and is likely to resemble an exponential . this finding is supported by a variety of experimental and numerical determinations ( e.g. kida & murakami 1989 , vincent & meneguzzi 1991 , jayesh & warhaft 1991 , she 1991 , she , jackson & orszag 1991 , cao , chen , & she 1996 , vainshtein 1997 , lamballais , lesieur , & mtais 1997 , machiels & deville 1998 ) . compressible turbulence has remained to be too complex for a satisfying mathematical analysis . it is not clear how to relate the analytical work on incompressible turbulence to molecular clouds . in addition to the fact that interstellar turbulence is highly supersonic and self - gravitating , there are also observational limitations . unlike the analytical approach or numerical simulations , molecular cloud observations allow access only to dimensionally reduced information . velocity measurements are possible only along the line - of - sight , and the spatial structure of a cloud is only seen in projection onto the plane of the sky , i.e. as variations of the column density . although some methods can yield information about the 3-dimensional spatial structure of the cloud ( see stutzki & gsten 1990 , williams , de geus , & blitz 1994 ) , the result is always model dependent and equivocal ( see also ballesteros - paredes , vzquez - semadeni , & scalo 1999 ) . a common way of obtaining knowledge about the velocity structure of molecular clouds is to study individual line profiles at a large number of various positions across the cloud . in the optical thin case line shapes are in fact histograms of the radial velocities of gas sampled along the telescope beam . falgarone & phillips ( 1990 ) and falgarone et al . ( 1994 ) showed that line profiles constructed from high - sensitivity co maps exhibit non - gaussian wings and attributed this to turbulent intermittency ( see also falgarone et al . 1998 on results from the iram - key project ) . dubinski , narayan , & phillips ( 1995 ) demonstrated that non - gaussian line profiles can be produced from _ any _ gaussian random velocity field if variance effects become important ( which is always the case for very steep or truncated power spectra ) . they concluded that non - gaussian line profiles do not provide clear evidence for intermittency . another method of inferring properties of the velocity distribution in molecular clouds is to analyze the pdf of line centroid velocities obtained from a large number of individual measurements scanning the entire projected surface area of a cloud ( miesch & scalo 1995 , lis et al . 1998 , miesch et al . 1998 ) . each line profile ( i.e. the pdf _ along _ the line - of - sight ) is collapsed into one single number , the centroid velocity , and then sampled _ perpendicular _ to the line - of - sight . hence , the two functions differ in the direction of the sampling and in the quantity that is considered . a related statistical measure is the pdf of centroid velocity increments , it samples the velocity differences between the centroids for line measurements which are offset by a given separation . the observational advantage of using centroid and increment pdf s is , that the line measurements can typically be taken with lower sensitivity as only the centroid has to be determined instead of the detailed line shape . these measures are also less dependent on large - scale systematic motions of the cloud and they are less effected by line broadening due to the possible presence of warm dilute gas . however , to allow for a meaningful analysis of the pdf s especially in the tails , the number of measurements needs to be very large and should not be less than about 1000 . in order to sample the entire volume of interstellar clouds , the molecular lines used to obtain the pdf s are optically thin . i follow this approach in the present investigation and use a mass - weighted velocity sampling along the line - of - sight to determine the line centroid . this zero - opacity approximation does not require any explicit treatment of the radiation transfer process . the observed pdf s are obtained from _ averaged _ quantities ( from column densities or line centroids ) . to relate these observational measures to quantities relevant for turbulence theory , i.e. to the full 3-dimensional pdf , numerical simulations are necessary as only they allow unlimited access to all variables in phase space . a first attempt to do this was presented by lis et al . ( 1996 , 1998 ) who analyzed a simulation of mildly supersonic decaying hydrodynamic turbulence by porter et al . since their model did neither include self - gravity nor consider flows at high mach number or mechanisms to replenish turbulence , the applicability to the interstellar medium remained limited . this fact prompts the current investigation which extends the previous ones by calculating _ highly supersonic flows _ , and by including _ self - gravity _ and a _ turbulent driving scheme_. the current study does not consider magnetic fields . their influence on the pdf s needs to be addressed separately . however , the overall importance of magnetic fields and mhd waves on the dynamical structure of molecular clouds may not be large . the energy associated with the observed fields is of the order of the ( turbulent ) kinetic energy content of molecular clouds ( crutcher 1999 ) . magnetic fields can not prevent the decay of turbulence ( e.g. mac low et al . 1998 ) which implies the presence of external driving mechanisms . these energy sources replenish the turbulent cascade and may excite mhd waves explaining the inferred equipartition between turbulent and magnetic energies . the one - point probability distribution function @xmath2 of a variable @xmath3 is defined such that @xmath4 measures the probability for the variable to be found in the interval @xmath5 $ ] . density pdf _ ( @xmath0-pdf ) discussed in this paper is obtained from the local density associated with each sph particle . it is basically the normalized histogram summed over all particles in the simulation , i.e. a mass - weighted sampling procedure is applied . the _ line - of - sight velocity centroid pdf _ ( @xmath6-pdf ) is more complicated to compute . the face of the simulated cube is divided into @xmath7 equal - sized cells . for each cell , the line profile is computed by sampling the normal ( line - of - sight ) velocity component of all gas particles that are projected into that cell . the line centroid is determined as the abscissa value of the peak of the distribution . this procedure corresponds to the formation of optically thin lines in molecular clouds , where all molecules within a certain column through the clouds contribute equally to the shape and intensity of the line . to reduce the sampling uncertainties , this procedure is repeated with the location of the cells shifted by half a cell size in each direction . altogether about 20@xmath8000 lines contribute to the pdf . this is procedure is repeated for line - of - sights along all three system axes to identify projection effects . the _ line centroid increment pdf _ ( @xmath9-pdf ) is obtained in a similar fashion . however , the sampled quantity is now the velocity _ difference _ between line centroids obtained at two distinct locations separated across the face of the cloud by a fixed shift vector @xmath1 . the @xmath9-pdf for a spatial lag @xmath10 is obtained as azimuthal average , i.e. as superposition of all individual pdf s with shift vectors of length @xmath10 . also statistical moments of the distribution can be used to quantify the spread and shape of pdf s . for the current analysis i use the first four moments . mean value @xmath11 and standard deviation @xmath12 ( the 1 . and 2 . moments ) quantify the location and the width of the pdf and are given in units of the measured quantity . the third and fourth moments , skewness @xmath13 and kurtosis @xmath14 , are dimensionless quantities characterizing the shape of the distribution . the skewness @xmath13 describes the degree of asymmetry of a distribution around its mean . the kurtosis @xmath14 measures the relative peakedness or flatness of the distribution . i use a definition where @xmath15 corresponds to a normal distribution . smaller values indicate existence of a flat peak compared to a gaussian , larger values point towards a stronger peak or equivalently towards the existence of prominent tails in the distribution . a pure exponential results in @xmath16 . gaussian random fields are statistically fully determined by their mean value and the 2-point correlation function , i.e. by their first two moments , @xmath11 and @xmath12 . all higher moments can be derived from those . the 2-point correlation function is equivalent to the power spectrum in fourier space ( e.g. bronstein & semendjajew 1979 ) . besides using moments there are other possibilities of characterizing a distribution . van den marel & franx ( 1993 ) and dubinski et al . ( 1995 ) applied gauss - hermite expansion series to quantify non - normal contributions in line profiles . a more general approach has been suggested by vio et al . ( 1994 ) , who discuss alternatives to the histogram representation of pdf s . however , as astrophysical data sets typically _ are _ histograms of various types and as histograms are the most commonly used method to describe pdf s , this approach is also adopted here . sph ( _ smoothed particle hydrodynamics _ ) is a particle - based scheme to solve the equations of hydrodynamics . the fluid is represented by an ensemble of particles , each carrying mass , momentum , and hydrodynamic properties . the time evolution of the fluid is represented by the time evolution of the particles , governed by the equations of motion which are supplemented by a prescription to modify the hydrodynamic properties . at any location these properties are obtained by averaging over an appropriate set of neighboring particles . excellent overviews over the method provide the reviews by benz ( 1990 ) and monaghan ( 1992 ) . for the current study i use sph because it is intrinsically lagrangian and because it is able to resolve very high density contrasts . another reason for choosing sph is the possibility to use it in combination with the special - purpose hardware device grape ( sugimoto et al . 1990 , ebisuzaki et al . 1993 ; and also umemura et al . 1993 , steinmetz 1996 ) . this allows calculations at supercomputer level on a normal workstation . the code is based on a version originally developed by benz ( 1990 ) , and is used with a standard description of a von neumann - type artificial viscosity ( monaghan & gingold 1983 ) with the parameters @xmath17 and @xmath18 for the linear and quadratic terms . the system is subject to periodic boundary conditions ( klessen 1997 ) and is integrated in time using a second - order runge - kutta - fehlberg scheme , allowing individual time steps for each particle . furthermore , the smoothing volume over which hydrodynamic quantities are averaged in the code is freely adjustable in space and time such that the number of neighbors for each particle remains approximately fifty . when including self - gravity , regions with masses exceeding the jeans limit become unstable and collapse . once a highly - condensed core has formed in the center of a collapsing gas clump , that core is substituted by a ` sink ' particle ( bate , bonnell , & price 1995 ) which inherits the combined masses , linear and ` spin ' angular momenta of the particles it replaces . it also has the ability to accrete further sph particles from its infalling gaseous envelope . for simulations of turbulent flows one also has to take into account that an explicit viscosity term is introduced in the sph method . this fact demands attention when studying dissipative processes , especially in the subsonic regime . the current study focuses on the properties of highly supersonic turbulent flows . in this regime , direct comparison between sph and grid - based methods has proven the close correspondence of both methods ( mac low et al . 1998 , klessen , heitsch & mac low 2000 ) . if one bears the above caveats in mind , the sph method calculates the time evolution of gaseous systems very reliably and accurately , and offers large spatial and dynamical flexibility . the numerical models discussed here describe isothermal gas . the hydrodynamic equations are extended to include self - gravity ( in sec.s [ sec : decay - with - gravity ] and [ sec : driven - with - gravity ] ) and to incorporate a random turbulent driving mechanism ( in sec.[sec : driven - with - gravity ] ) . all physical constants are set to unity . the same applies to mass and length scales , i.e. the total mass is @xmath19 and the simulated volume is the cube @xmath20 ^ 3 $ ] . the mean density is thus @xmath21 . the initial configuration of all dynamical systems discussed in this paper is a homogeneous gas distribution with a gaussian velocity field . without turbulence , the time evolution depends on one parameter , the _ ratio _ between internal and gravitational energy , @xmath22 . this quantity can be interpreted as dimensionless temperature and determines the number of thermal jeans masses contained in the system . molecular clouds are characterized by line widths which largely exceed the thermal broadening . the evolution away from the homogeneous initial state is thus strongly influenced by the adopted initial velocity distribution and depends on whether turbulence is decaying or driven . large turbulent kinetic energy can considerable slow down or even prevent the collapse of thermally jeans unstable gas . the situation is very complex and depends on the shape and strength of the turbulent velocity spectrum ( klessen et al . 2000 ; see also see bonazzola et al . 1992 and vzquez - semadeni & gazol 1995 for an analytical approach ) . to generate and maintain turbulent flows gaussian velocity fields are introduced . the spatial variations of each component of the velocity vector @xmath23 are described as superpositions of plane waves with wave numbers @xmath24 , where the phase of each wave is random and sampled from a uniform distribution in the interval @xmath25 . also the amplitude is random , but selected from a gaussian distribution centered on zero and with a width determined by the power spectrum @xmath26 . gaussian fields are isotropic and only depend on the absolute value of the wave vector @xmath27 . only waves in the range @xmath28 are considered . for large cut - off wave numbers @xmath29 the gaussian statistics is very well sampled . if only very few modes are used to generate the field , variance effects become strong and individual realizations of the field can deviate significant from the ensemble average ( see sec . [ sec : init ] ) . the field is then transformed back into real space and the resulting velocities are assigned to individual sph particles using the ` cloud - in - cell ' scheme ( hockney & eastwood 1988 ) . for the initial field , all velocities are multiplied by the appropriate factor to reach the desired rms mach number of the flow . in case of driven turbulence , this velocity field is also used to ` kick ' the sph particles at every time step such that a constant level of kinetic energy is maintained ( see mac low 1999 ) variance effects in poorly sampled gaussian velocity fields can lead to considerable _ non_-normal contributions to the @xmath6- and @xmath30-pdf s . if a random process is the result of sequence of independent events ( or variables ) , then in the limit of large numbers , its distribution function will be a gaussian around some mean value . however , only the properties of a large _ ensemble _ of gaussian fields are determined in a statistical sense . individual realizations may exhibit considerable deviations from the mean . the effect is strongest when only few ( spatial ) modes contribute to the field or , almost equivalently , when the power spectrum falls off very steeply . in this case , most kinetic energy is in large - scale motions . this is visualized in fig . [ fig : v - pdf - init ] , it shows @xmath6-pdf s for homogeneous gas ( sampled by @xmath31 sph particles placed on a regular grid ) with gaussian velocity fields with power spectra @xmath32 which are truncated at different wave numbers @xmath33 ranging from ( a ) @xmath34 to ( d ) @xmath35 . each realization is scaled such that the rms velocity dispersion is @xmath36 . the figure displays the pdf s for the @xmath3- , @xmath37- , and @xmath38-component of the velocity . the pdf s of the strongly truncated spectrum ( fig . [ fig : v - pdf - init]a ) do not at all resemble normal distributions . the gaussian statistics of the field is very badly sampled with only very few modes . note that the pdf s of the same field may vary considerably for different velocity _ components _ , i.e. for different _ projections_. with the inclusion of larger number of fourier modes this situation improves , and in fig . [ fig : v - pdf - init]d the pdf s of all projections sample the expected gaussian distribution very well . a similar conclusion can be derived for @xmath9-pdf . this measure is even more sensitive to deviations from gaussian statistics . figure [ fig : dv - pdf - init ] plots the @xmath9-pdf s for the same sequence of velocity fields . for brevity , only the line - of - sight component parallel to the @xmath3-axis is considered . furthermore , from the sequence of possible @xmath9-pdf s ( defined by the spatial lag @xmath10 ) only three are shown , at small ( @xmath39 , upper curve ) , medium ( @xmath40 , middle curve ) , and large spatial lags ( @xmath41 , upper curve ) . sampling the gaussian field with only two modes ( fig . [ fig : dv - pdf - init]a ) is again insufficient to yield increment pdf s of normal shape . the velocity field is very smooth , and the line centroid velocity difference between neighboring cells is very small . hence , for @xmath39 the pdf is dominated by a distinct central peak at @xmath42 . the tails of the distribution are quite irregularly shaped . the situation becomes ` better ' when sampling increasing distances , as regions of the fluid separated by larger @xmath10 are less strongly correlated in velocity . for @xmath40 and @xmath41 the pdf s follow the gaussian distribution more closely although irregularities in the shapes are still present . in fig.s [ fig : dv - pdf - init]b and c the @xmath9-pdf s for medium to large lags are very well fit by gaussians . deviations occur only at small @xmath10 , the pdf s are exponential ( and the distribution for @xmath43 is still a bit cuspy ) . finally , fig . [ fig : dv - pdf - init]d shows the three @xmath30-pdf s for the case where all available spatial modes contribute to the velocity field ( @xmath44 ) . the pdf s follow a gaussian for all spatial lags this behavior is also seen in the variation of the moments of the distribution as function of the spatial lag @xmath10 . applied to the above sequence of gaussian velocity fields , fig . [ fig : mom - init ] displays the dispersion @xmath12 and the kurtosis @xmath14 of the distribution . the corresponding models are indicated at the right hand side of each plot . the width of the distribution , as indicated by the dispersion @xmath12 ( fig . [ fig : mom - init]a ) , typically grows with increasing @xmath45 , reflecting the relative peakedness of the distribution at small lags . for example , the distribution ( a ) yields a slope of 0.3 in the range @xmath46 , and ( b ) leads to a value of 0.2 in relatively large interval @xmath47 . the effect disappears for the better sampled fields . typical values for that slope in observed molecular clouds are @xmath48 to @xmath49 ( miesch et al . 1998 ) . versus the spatial lag @xmath10 . for a comparison with the present study , their numbers have to be divided by a factor of two . furthermore , they use a relatively narrow range of @xmath10-values to compute the slope of the function ; larger intervals would on average tend to decrease these values ( see their fig . in addition , miesch et al . ( 1998 ) applied spatial filtering to remove large - scale velocity gradients in the clouds . these would lead to steeper slopes . the fact that in the present study the functions @xmath12 and @xmath14 level out for large spatial lags @xmath10 is a consequence of the periodic boundary conditions which do not allow for large - scale gradients.[comment ] ] a direct measure of the peakedness of the distribution is its fourth moment , the kurtosis @xmath14 ( fig . [ fig : mom - init]b ) . at small lags @xmath10 , clearly the pdf s of model ( a ) are more strongly peaked than exponential ( @xmath50 ) . comparing the entire sequence reveals again the tendency of the pdf s to become gaussian at decreasing @xmath10 with increasing number of modes considered in the construction of the velocity field . taking all together , it is advisable to consider conclusions about interstellar turbulence derived from solely analyzing one - point probability distribution functions from molecular clouds with caution . similar to what has been shown by dubinski et al . ( 1995 ) for molecular line profiles , deviations from the regular gaussian shape found in @xmath6- and @xmath9-pdf s need not be the signpost of turbulent intermittency . gaussian velocity fields which are dominated by only a small number of modes ( either because the power spectrum falls off steeply towards larger wave numbers , or because small wave length distortions are cut away completely ) will lead to very similar distortions . in addition , the properties of the pdf may vary considerably between different projections . the same velocity field may lead to smooth and gaussian pdf s for one velocity component , whereas another projection may result in strong non - gaussian wings ( see also fig . [ fig : v - pdf - xyz - decay - with - gravity ] ) . in this section the pdf s of freely decaying initially highly supersonic turbulence without self - gravity are discussed . they are calculated from an sph simulation with @xmath51 particles ( mac low et al . 1998 , model g ) . initially the system is homogeneous with a gaussian velocity distribution with @xmath52 in the interval @xmath53 . the rms mach number of the flow is @xmath54 . after the onset of the hydrodynamic evolution the flow quickly becomes fully turbulent resulting in rapid dissipation of kinetic energy . the energy decay is found to follow a power law @xmath55 with exponent @xmath56 . the overall evolution can be subdivided into several phases . the first phase is very short and is defined by the transition of the initially gaussian velocity field into fully developed supersonic turbulence . it is determined by the formation of the first shocks which begin to interact with each other and build up a complex network of intersecting shock fronts . energy gets transfered from large to small scales and the turbulent cascade builds up . the second phase is given by the subsequent self - similar evolution of the network of shocks . even though individual features are transient , the overall properties of this network change only slowly . in this phase of highly supersonic turbulence the loss of kinetic energy is dominated by dissipation in shocked regions . in the transsonic regime , i.e. the transition from highly supersonic to fully subsonic flow , energy dissipation in vortices generated by shock interactions becomes more and more important . only the strongest shocks remain in this phase . surprisingly , the energy decay law does not change during this transition . it continues to follow a power law with exponent @xmath57 . in the subsonic phase the flow closely resembles incompressible turbulence . its properties are similar to those reported from numerous experiments and simulations ( e.g. porter et al . 1994 , lesieur 1997 , boratav et al . the simulation is stopped at @xmath58 when the flow has decayed to a rms mach number of @xmath59 . since the energy loss rate follows a power law , the duration of each successive phase grows . this sequence of evolutionary stages is seen in the pdf s of the system . one noticeable effect is the decreasing width of the distribution functions as time progresses . as the kinetic energy decays the available range of velocities shrinks . this not only leads to ` smaller ' @xmath6- and @xmath9-pdf s , but also to a smaller @xmath0-pdf since compressible motions lose influence and the system becomes more homogeneous . this is indicated in fig . [ fig : rho - v - pdf - decay - no - gravity ] , it displays ( a ) the @xmath60-pdf and ( b ) @xmath6-pdf at the following stages of the dynamical evolution ( from top to bottom ) : shortly after the start , at @xmath61 when the first shocks occur , then at @xmath62 when the network of interacting shocks is established and supersonic turbulence is fully developed , during the transsonic transition at @xmath63 , and finally at @xmath58 when the flow has progressed into the subsonic regime . the rms mach numbers at these stages are @xmath64 , @xmath65 , @xmath66 , and @xmath59 , respectively . the density pdf always closely follows a log - normal distribution , i.e. it is gaussian in the _ logarithm _ of the density . also the distribution of line centroids at the four different evolutionary stages of the system is best described by a gaussian with only minor deviations at the far ends of the velocity spectrum . for the same points in time , fig . [ fig : dv - pdf - decay - no - gravity ] shows the @xmath9-pdf s for @xmath3-component of the velocity . the displayed spatial lags are selected in analogy to fig . [ fig : dv - pdf - init ] . note the different velocity scaling in each plot reflecting the decay of turbulent energy as the system evolves in time . throughout the entire sequence , spatial lags larger than about 10% of the system size always lead to @xmath9-pdf s very close to gaussian shape ( the middle and lower curves ) . considerable deviations occur only at small spatial lags ( the upper curves ) . for those , the increment pdf s exhibit exponential wings during all stages of the evolution . when scaling the pdf s to the same width , the distribution in the subsonic regime ( d ) appears to be more strongly peaked than during the supersonic or transsonic phase ( a c ) . there , the central parts of the pdf s are still reasonably well described by the gaussian obtained from the first two moments , whereas in ( d ) the peak is considerably narrower , or vice versa , the tails of the distribution are more pronounced . these results can be compared with the findings by lis et al . they report increment pdf s for three snapshots of a high - resolution hydrodynamic simulation of decaying mildly super - sonic turbulence performed by porter et al . they analyze the system at three different times corresponding to rms mach numbers of @xmath67 , @xmath68 , and @xmath69 . their first two data sets thus trace the transition from supersonic to subsonic flow and are comparable to phase ( c ) of the current model ; their last data set corresponds to to phase ( d ) . in the transsonic regime both studies agree : lis et al . ( 1998 ) report enhanced tails in the increment pdf s for the smallest spatial lags which they considered and near gaussian distributions for larger lags ( however , the largest separation they study is about 6% of the linear extent of the system ) . in the subsonic regime , lis et al . ( 1998 ) find near gaussian pdf s for very small spatial lags ( @xmath70% ) , but extended wings in the pdf s for lags of 3% and 6% of the system size . they associate this with the ` disappearance ' of large - scale structure . indeed , their fig . 7 exhibits a high degree of fluctuations on small scales which they argue become averaged away when considering small spatial lags in the @xmath9-pdf . comparing the pdf with spatial lags of 3% ( upper curves in fig . [ fig : dv - pdf - decay - no - gravity ] , compared to the pdf s labeled with @xmath71 in lis et al . 1998 ) both studies come to the same result . at these scales the @xmath9-pdf s tend to exhibit more pronounced wings in the subsonic regime as in the supersonic regime . the sph calculations reported here do not allow for a meaningful construction of @xmath72-pdf s for @xmath73% . the gaussian behavior of pdf s for very small spatial lags reported by lis et al . ( 1998 ) therefore can not be examined . however , neither of the purely hydrodynamic simulations lead to pdf s that are in good agreement with the observations . observed pdf s typically are much more centrally peaked at small spatial separation ( see e.g.fig . 4 in lis et al 1998 and miesch et al . 1998 ) . figure [ fig : dv - array - decay - no - gravity ] shows the spatial distribution of centroid velocity differences between cells separated by a vector lag of @xmath74 ( i.e. between neighboring cells along the diagonal ) . data are obtained at the same times as above . each figure displays the array of the absolute values of the velocity increments @xmath75 in linear scaling as indicated at the top . note the decreasing velocity range reflecting the decay of turbulent energy . the distribution of @xmath75 appears random , there is no clear indication for coherent structures . this is corresponds to most observations . miesch et al . ( 1998 ) find for their sample of molecular clouds that high - amplitude velocity differences for very small spatial lags typically are well distributed resulting in a ` spotty ' appearance . note , however , that using azimuthal averaging lis et al . 1998 report the finding of filamentary structures for the @xmath0-ophiuchus cloud . altogether , filamentary structure is difficult to define and a mathematical thorough analysis is seldomly performed ( for an astrophysical approach see adams & wiseman 1994 , for a discussion of the filamentary vortex structure in incompressible turbulence consult frisch 1995 or lesieur 1997 ) . the visual inspection of maps is often misleading and influenced by the parameters used to display the image . larger velocity bins for instance tend to produce a more ` filamentary ' structure than very fine sampling of the velocity structure . further uncertainty may be introduced by the fact that molecular clouds are only seen in one projection as the signatures of the dynamical state of the system can strongly depend on the viewing angle . in this section , i discuss the properties of decaying , initially supersonic turbulence in a self - gravitating medium . figure [ fig:3d - plot - decay - with - gravity ] displays an sph simulation with @xmath76 particles at six different times of its dynamical evolution . since the model is subject to periodic boundary conditions , every figure has to be considered infinitely replicated in each direction . analog to the previous model , the system is initially homogeneous and its velocity field is generated with @xmath77 using modes with wave numbers @xmath78 . from the choice @xmath79 it follows that the system contains 120 _ thermal _ jeans masses . the initial rms velocity dispersion is @xmath36 and with the sound speed @xmath80 the rms mach number follows as @xmath81 . these values imply that the initial turbulent velocity field contains sufficient energy to globally _ stabilize _ the system against gravitational collapse . scaled to physical units using a density @xmath82@xmath83 , which is typical for massively star - forming regions ( e.g. williams et al.2000 ) , the system corresponds to a volume of @xmath84 ^ 3 $ ] and contains a gas mass of 200@xmath8m@xmath85 . as the simulation starts , the system quickly becomes fully turbulent and loses kinetic energy . like in the case without self - gravity a network of intersecting shocks develops leading to density fluctuations on all scales . if the mass of a fluctuation exceeds the ( local ) jeans limit it begins to contract due to self - gravity . during the early evolution , there is enough kinetic energy to prevent this collapse process on all scales ( fig . [ fig:3d - plot - decay - with - gravity]b @xmath86 ) and the properties of the system are similar to those of pure hydrodynamic turbulence . however , as time progresses and turbulent energy decays the effective jeans mass decreases . _ local _ collapse of shock generated density fluctuations sets in despite the fact that the system is still globally stabilized by turbulence ( see also klessen et al . the central high - density cores of collapsing clumps are indicated by black dots . the cores form mainly at the intersection of filaments , where the density is highest and local collapse is most likely to set in . when turbulence is decayed sufficiently also large - scale collapse becomes possible . gas clumps follow the global flow pattern towards a common center of gravity where they may merge or sub - fragment . gradually a cluster of dense cores is built up . in the isothermal model this process continues until all available gas is accreted onto the ` protostellar ' cluster ( for more details klessen & burkert 2000 ) . the pdf s of ( a ) the density and of ( b ) the @xmath3-component of the line centroid velocities for the above six model snapshots are displayed in fig . [ fig : rho - v - pdf - decay - with - gravity ] . the corresponding time is indicated by the letters at the right side of each panel . during the dynamical evolution of the system the density distribution develops a high density tail this is the imprint of local collapse . the densities of compact cores are indicated by solid dots ( at @xmath87 and @xmath88 ) . virtually all particles in the high density tails at earlier times ( at @xmath89 and more so at @xmath90 ) are accreted onto these cores . the bulk of matter roughly follows a log - normal density distribution as indicated by the dotted parabola . the @xmath6-pdf s are nearly gaussian as long as the dynamical state of the system is dominated by turbulence . also the width of the pdf remains roughly constant during this phase . this implies that the decay of turbulent kinetic energy is in balance with the gain of kinetic energy due to gravitational ( ` quasi - static ' ) contraction on large scales . the time scale for this process is determined by the energy dissipation in shocks and turbulent eddies . however , once _ localized _ collapse is able to set in , accelerations on small scales increase dramatically and the evolution ` speeds up ' . for times @xmath91 the centroid pdf s become wider and exhibit significant deviations from the original gaussian shape . the properties of the pdf s are similar to those observed in star - forming regions ( miesch & scalo 1995 , lis et al . 1998 , miesch et al . 1998 ) . this is expected since gravitational collapse is a necessary ingredient for forming stars . gravity creates _ non_-isotropic density and velocity structure structures . when analyzing @xmath6- and @xmath9-pdf s , their appearance and properties will strongly depend on the viewing angle . this is a serious point of caution when interpreting observational data , as molecular cloud are seen only in _ one _ projection . as illustration , [ fig : v - pdf - xyz - decay - with - gravity ] plots the centroid pdf at the time @xmath87 for the line - of - sight projection along all three axes of the system . whereas the pdf s for the @xmath3- and the @xmath37-component of the velocity centroid are highly structured ( upper and middle curve the latter one is even double peaked ) , the distribution of the @xmath38-component ( lowest curve ) is smooth and much smaller in width , comparable to the ` average ' pdf at _ earlier _ stages of the evolution . as the variations between different viewing angles or equivalently different velocity components can be very large , statements about the 3-dimensional velocity structure from only observing one projection can be misleading . gravity effects the @xmath9-pdf . figure [ fig : dv - pdf - decay - with - gravity ] displays the increment pdf s at small , intermediate and large spatial lags , analog to fig.s [ fig : dv - pdf - init ] and [ fig : dv - pdf - decay - no - gravity ] . time ranges from ( a ) @xmath89 to ( d ) @xmath88 corresponding to fig.s [ fig:3d - plot - decay - with - gravity]c f . the pdf s for @xmath92 and @xmath86 are not shown since at these stages supersonic turbulence dominates the dynamic of the system and the pdf s are comparable to the ones without gravity ( fig . [ fig : dv - pdf - decay - no - gravity ] ) . this still holds for @xmath89 . the increment pdf s for medium to large spatial lags appear gaussian , however , the pdf for the smallest lag follows a perfect exponential all the way inwards to @xmath42 . unlike in the case without gravity , the peak of the distribution is not ` round ' , i.e. is not gaussian in the innermost parts ( when scaled to the same width ) . _ it is a sign of self - gravitating systems that the increment pdf at smallest lags is very strongly peaked and remains exponential over the entire range of measured velocity increments_. this behavior is also seen fig.s [ fig : dv - pdf - decay - with - gravity]b d . at these later stages of the evolution in addition non - gaussian behavior is also found at medium lags . this results from the existence of large - scale filaments and streaming motions . the same behavior is found for the increment pdf s from observed molecular clouds ( for @xmath0-ophiuchus see lis et al 1998 ; for orion , mon r2 , l1228 , l1551 , and hh83 see miesch et al . 1998 ) . in each case , the distribution for the smallest lag ( one pixel size ) is very strongly peaked at @xmath42 , in some cases even more than exponential . the deviations from the gaussian shape remain for larger lags but are not so pronounced . the inclusion of self - gravity into models of interstellar turbulence leads to good agreement with the observed increment pdf s . however , this result may _ not _ be unique as in molecular clouds additional processes are likely to be present that could also lead to strong deviations from gaussianity . the time evolution of the statistical moments of the @xmath9-pdf s for various spatial lags is presented in fig . [ fig : mom - decay - with - gravity ] . it plots ( a ) the dispersion @xmath12 , and ( b ) the kurtosis @xmath14 . the letters on the right - hand side indicate the corresponding time in fig . [ fig:3d - plot - decay - with - gravity ] . at @xmath92 the width @xmath12 of the pdf is approximately constant for all @xmath10 and the kurtosis @xmath14 is close to normal value of three . both indicate that gaussian statistics very well describes the initial velocity field . as turbulent energy decays , gravitational collapse sets in . because of the gravitational acceleration , the amplitudes of centroid velocity differences between separate regions in the cloud grow larger , the width @xmath12 of the @xmath9-pdf s increases . this becomes more important when sampling velocity differences on larger spatial scales , hence @xmath12 also increases with @xmath10 . the slope is @xmath93 . for @xmath94 it levels out , which is a result of the adopted periodic boundary conditions . they do not allow for large - scale velocity gradients . the increasing ` peakedness ' of @xmath30-pdf is reflected in the large values of the kurtosis @xmath14 at the later stages of the evolution . for small spatial lags the pdf s are more centrally concentrated than exponential ( i.e. @xmath95 ) , and even at large spatial separations they are still more strongly peaked than gaussian ( @xmath96 ) . the slope at @xmath88 is @xmath97 which is indeed comparable to what is found in observed star - forming regions ( miesch et al . 1998 ) . for the above simulation of self - gravitating , decaying , supersonic turbulence , fig.[fig : dv - array - decay - with - gravity ] plots the 2-dimensional distribution of centroid increments for a vector lag @xmath74 . the velocity profiles are sampled along the @xmath3-axis of the system . the magnitude of the velocity increment @xmath98 is indicated at the top of each plot . the spatial distribution of velocity increments during the initial phases appears random . later on , gravity gains influence over the flow and creates a network of intersecting filaments where gas streams onto and flows along towards local potential minima . at that stage , the velocity increments with the highest amplitudes tend to trace the large - scale filamentary structure . this is the sign of the anisotropic nature of gravitational collapse motions . figure [ fig:3d - plot - driven - with - gravity ] displays the gas distribution at different evolutionary stages of a simulation of driven , supersonic , self - gravitating turbulence . the number of sph particles is @xmath99 . again , the system is initially homogeneous in space and has a random gaussian velocity field with flat power spectrum in the wave number interval @xmath100 . it contains 64 _ thermal _ jeans masses and turbulence is continuously driven as described in sec . [ subsec : models ] . the initial evolution into equilibrium between the energy input by the driving force and the decay of turbulent kinetic energy is computed without self - gravity , then it is turned on . this phase is displayed in fig . [ fig:3d - plot - driven - with - gravity]a . in this state the _ turbulent _ jeans mass ( on scales larger than the maximum driving wave length ) exceeds the total mass in the system by a factor of two , the cloud is therefore stabilized by turbulence against gravitational collapse on global scales . however , local collapse ( on scales at or below the driving scale ) is still possible and does occur . as in the previous case without driving , the dynamical evolution of the system leads to the formation of a cluster of dense collapsed cores . this is shown in fig.s [ fig:3d - plot - driven - with - gravity]b d , which display the system when 20% , 40% , and 60% of the gas mass has accumulated in dense collapsed cores ( at time @xmath101 , @xmath102 , and @xmath103 , respectively ) . however , in the presence of the driving source the time scales for accretion are longer and the cluster is less dense . the pdf s of ( a ) the density and ( b ) the @xmath3-component of the line centroid velocities corresponding to the above four snapshots are displayed in fig . [ fig : rho - v - pdf - driven - with - gravity ] . as in the previous model , the bulk of gas particles that are not accreted onto cores build up an approximately log - normal @xmath0-pdf ( indicated by the dotted lines ) . also the @xmath6-pdf remains close to the gaussian value . this is different from the case of purely decaying self - gravitating turbulence , where at some stage global collapse motions set in and lead to very wide and distorted centroid pdf s . this is not possible in the simulation of driven turbulence , as it is stabilized on the largest scales by turbulence . collapse occurs only locally which leaves the width of the pdf s relatively unaffected and only mildly alters their shape . also the @xmath9-pdf s show no obvious sign of evolution . for the @xmath3-component of the velocity these functions are displayed in fig . [ fig : dv - pdf - driven - with - gravity ] , again for three different spatial lags . the chosen times correspond ( a ) to the equilibrium state at @xmath92 , and ( b ) to @xmath103 which is the final state of the simulation . the pdf s only marginally grow in width . at every evolutionary stage , the pdf for the smallest spatial lag is exponential , whereas the pdf s for medium and large shift vectors closely follow the gaussian curve defined by the first two moments of the distribution ( dotted lines ) . the functions are similar to the ones in the previous model before the large scale collapse motions set in ( fig . [ fig : dv - pdf - decay - with - gravity]a , b ) . only overall contraction will affect @xmath9-pdf at medium to large lags . this behavior also follows from comparing the statistical moments . figure [ fig : mom - driven - with - gravity ] plots ( a ) the dispersion @xmath12 and ( b ) the kurtosis @xmath14 as function of the spatial lag @xmath45 . figures [ fig : mom - decay - with - gravity]a and [ fig : mom - driven - with - gravity]a are very similar , as soon as turbulence is established the width @xmath12 of the pdf increases with @xmath10 with a slope of @xmath93 for small to medium lags and levels out for larger ones . however , when comparing the ` peakedness ' of the pdf as indicated by @xmath14 ( fig . s [ fig : mom - decay - with - gravity]b and [ fig : mom - driven - with - gravity]b ) the model of decaying self - gravitating turbulence yields much higher values since the pdf s are more strongly peaked due to the presence of large - scale collapse motions . figure [ fig : dv - array - driven - with - gravity ] finally shows the spatial distribution of the @xmath3-component of the line centroid increments for a vector lag @xmath104 . since the increment maps at different evolutionary times are statistically indistinguishable , only times ( a ) @xmath92 and ( b ) @xmath103 are displayed in the figure . as in the case of supersonic , purely hydrodynamic turbulence the spatial distribution of velocity increments appears random and uncorrelated . the adopted driving mechanism prevents global collapse . the bulk properties of the system therefore resemble hydrodynamic , _ non_-self - gravitating turbulence . however , local collapse motions do exist and are responsible for noticeable distortions away from the gaussian statistics . as the non - local driving scheme adopted here introduces a bias towards gaussian velocity fields , these distortions are not very large . there is a need to introduce other , more realistic driving agents into this analysis . these could lead to much stronger non - gaussian signatures in the pdf s . sph simulations of driven and decaying , supersonic , turbulent flows with and without self - gravity have been analyzed in this study . it extends previous investigations of mildly supersonic , decaying , _ non_-self - gravitating turbulence ( lis et al . 1996 , 1998 ) into a regime more relevant molecular clouds , by ( a ) considering highly supersonic flows and by including ( b ) self - gravity and ( c ) a driving source for turbulence . the flow properties are characterized by using the probability distribution functions of the density , of the line - of - sight velocity centroids , and of their increments . furthermore the dispersion and the kurtosis of the increment pdf s are discussed , as well as the spatial distribution of the velocity increments for the smallest spatial lags . \(1 ) to asses the influence of variance effects , simple gaussian velocity fluctuations are studied . the insufficient sampling of random gaussian ensembles leads to distorted pdf s similar to the observed ones . for line profiles this has been shown by dubinski et al . ( 1995 ) . \(2 ) decaying , initially highly supersonic turbulence without self - gravity leads to pdf s which also exhibit deviations from gaussianity . for the trans- and subsonic regime this has been reported by lis et al . ( 1996 , 1998 ) . however , neglecting gravity and thus not allowing for the occurance of collapse motions , these distortions are not very pronounced and can not account well for the observational data ( lis et al . 1998 , miesch et al . 1998 ) . \(3 ) when including gravity into the models of decaying initially supersonic turbulence , the pdf s get into better agreement with the observations . during the early dynamical evolution of the system turbulence carries enough kinetic energy to prevent collapse on all scales . in this phase the properties of the system are similar to those of non - gravitating hydrodynamic supersonic turbulence . however , as turbulent energy decays gravitational collapse sets in . first localized and on small scales , but as the turbulent support continues to diminish collapse motions include increasingly larger spatial scales . the evolution leads to the formation of an embedded cluster of dense protostellar cores ( see also klessen & burkert 2000 ) . as the collapse scale grows , the @xmath0- , @xmath6- , and @xmath9-pdf s get increasingly distorted . in particular , the @xmath9-pdf s for small spatial lags are strongly peaked and exponential over the entire range of measured velocities . this is very similar to what is observed in molecular clouds ( for @xmath0-ophiuchus see lis et al 1998 ; for orion , mon r2 , l1228 , l1551 , and hh83 see miesch et al . 1998 ) . \(4 ) the most realistic model for interstellar turbulence considered here includes a simple ( non - local ) driving scheme . it is used to stabilize the system against collapse on large scales . again non - gaussian pdf s are observed . despite global stability , local collapse is possible and the system again evolves towards the formation of an embedded cluster of accreting protostellar cores . as the adopted driving scheme introduces a bias towards maintaining a gaussian velocity distribution , the properties of the pdf s fall in between the ones of pure hydrodynamic supersonic turbulence and the ones observed in systems where self - gravity dominates after sufficient turbulent decay . this situation may change when considering more realistic driving schemes . \(5 ) a point of caution : the use of @xmath6- and @xmath9-pdf s to unambiguously characterize interstellar turbulence and to identify possible physical driving mechanisms may be limited . _ all _ models considered in the current analysis lead to non - gaussian signatures in the pdf s , differences are only gradual . in molecular clouds the number of physical processes that are expected to give rise to deviations from gaussian statistics is large . simple statistical sampling effects ( sec . [ sec : init ] ) and turbulent intermittency caused by vortex motion ( lis et al . 1996 , 1998 ) , as well as the effect self - gravity ( sec . [ sec : decay - with - gravity ] ) and of shock interaction in highly supersonic flows ( mac low & ossenkopf 2000 ) , _ all _ will lead to non - gaussian signatures in the observed pdf s . also stellar feedback processes , galactic shear and the presence of magnetic fields will influence the interstellar medium and create distortions in the velocity field . this needs to be studied in further detail . in addition , the full 3-dimensional spatial and kinematical information is not accessible in molecular clouds , measured quantities are always projections along the line - of - sight . as the structure of molecular clouds is extremely complex , the properties of the pdf s may vary considerably with the viewing angle . attempts to disentangle the different physical processes influencing interstellar turbulence therefore should no rely on analyzing velocity pdf s alone , they require additional statistical information . i thank a. burkert , f. heitsch , and m .- mac low for many fruitful and stimulating discussions , and the editor s. shore for his comments on the paper and his help with an extremely slow and non - responsive ( anonymous ) referee . adams , f.c . , wiseman , j.j . 1994 , , 435 , 693 balkovsky , e. , falkovich , g. 1998 , , 57 , r1231 balkovsky , e. , falkovich , g. , kolokolov , i. , lebedev , v. 1997 , , 78 , 1452 ballesteros - paredes , j. , vzquez - semadeni , e. , scalo , j. 1999 , , 515 , 286 bate , m.r . , bonnell , i.a , price , n.m . 1995 , , 277 , 362 bate , m.r . , burkert , a. 1997 , , 288 , 1060 benz , w. 1990 , in the numerical modeling of nonlinear stellar pulsations , ed . buchler , kluwer academic publisher , dordrecht bonazzola , s. , perault , m. , puget , j.l , heyvaerts , j. , falgarone , e. , panis , j.f . 1992 , j. fluid . , 245 , 1 boratav , o. , eden , a. , erzan , a. ( eds . ) 1997 , turbulence modeling and vortex dynamics , springer verlag , heidelberg bronstein , i.n . , semendjajew , k.a . 1979 , taschenbuch der mathematik , teubner verlagsgesellschaft , leibzig burkert , a. , klessen , r.s . , bodenheimer , p. 1998 , in the orion complex revisited , eds . m. mccaughrean & a. burkert , asp conference series , in press cao , n. , chen , s. , she , z .- s . 1996 , , 76 , 3711 chertkov , m. , kolokolov , i. , vergassola , m. 1997 , , 56 , 5483 crutcher , r. m. 1999 , , 520 , 706 dubinski , j. , narayan , r. , phillips , t.g . 1995 , , 448 , 226 ebisuzaki , t. , makino , j. , fukushige , t. , taiji , m. , sugimoto , d. , ito , t. , okumura , s.k . 1993 , , 45 , 269 falgarone , e. , lis , d.c . , phillips , t.g . , pouquet , a. , porter , d.h . , woodward , p.r . 1994 , , 436,728 falgarone , e. , panis , j .- f . , heithausen , a. , prault , m. , stutzki , j. , puget , j .- l . , bensch , f. 1998 , , 331 , 669 falgarone , e. , phillips , t.g . 1990 , , 359 , 344 falkovich , g. , lebedev , v. 1997 , , 79 , 4159 forster , d. , nelson , d.r . , stephen , m.j . 1977 , , 16 , 732 franco , j. , carraminana , a. ( eds . ) 1999 , interstellar turbulence , cambridge university press frisch , u. 1995 , turbulence the legacy of a. n.kolmogorov , cambridge university press gammie , c.f . , ostriker , e. e.c . 1996 , , 466 , 814 hockney , r.w . , eastwood , j.w . 1988 , computer simulation using particles , iop publishing ltd . , bristol and philadelphia jahyesh , warhaft , z. 1991 , , 67 , 3503 kleiner , s.c . , dickman , r.l . 1987 , , 312 , 837 klessen , r.s . 1997 , , 292 , 11 klessen , r.s . , burkert , a. , bate , m.r . 1998 , , 501 , l205 klessen , r.s . , burkert , a. 2000 , , 128 in press ( astro - ph/9904090 ) klessen , r.s . , heitsch , f. , mac low , m .- 2000 , , accepted ( astro - ph/9911068 ) kida , s. , murakami , y. 1989 , fluid dyn . res . , 4 , 347 kitamura , y. , sunada , k. , hayashi , m. , hasegawa , t. 1993 , , 413 , 221 kolmogorov , a.n . 1941 , dokl . nauk sssr , 30 , 301 ( reprinted in proc . r. soc . lond . a , 434 , 9 ( 1991 ) ) lamballais , e. , lesieur , m. , mtais , o. 1997 , , 56 , 761 larosa , t.n . , shore , s.n . , magnani , l. 1999 , , 512 , 761 lesieur , m. 1997 , turbulence in fluids , 3 . , kluwer academic publishers , dordrecht lis , d.c . , keene , j. , phillips , t.g . , pety , j. 1998 , , 504 , 889 lis , d.c . , pety , j. , phillips , t.g . , falgarone , e. 1996 , , 463 , 623 machiels , l. , deville , m.o . 1998 , j. comp . phys . , 145 , 256 mac low , m .- 1999 , , 524 , 169 mac low , m .- m . , klessen , r.s . , burkert , a. , smith , m.d . 1998 , , 80 , 2754 mac low , m .- , ossenkopf , v. 2000 , , 353 , 339 miesch , m.s . , bally , j. 1994 , , 429 , 645 miesch , m.s . , scalo , j.m . 1995 , , 450 , l27 miesch , m.s . , scalo , j.m . , bally , j. 1998 , , submitted ( astro - ph/9810427 ) monaghan , j.j . 1992 , , 30 , 543 monaghan , j.j . , gingold , r.a . 1983 , j. comp . phys . , 52 , 135 padoan , p. , nordlund , , jones , b.j.t . 1997 , , 288 , 145 padoan , p. , nordlund , . 1999 , , 526 , 279 passot , t. , vzquez - semadeni , e. 1998 , , 58 , 4501 porter , d.h . , pouquet , a. , woodward , p.r . 1994 , phys . fluids , 6 , 2133 scalo , j.m . 1984 , , 7 , 556 scalo , j.m . , vzquez - semadeni , e. , chappell , d. , passot , t. 1998 , , 504 , 835 she , z .- s . 1991 , fluid dyn . res . , 8 , 143 she , z .- s . , jackson , e. , orszag , s.a . 1991 , proc . lond . a , 434 , 101 steinmetz , m. 1996 , , 278 , 1005 stone , j. m. , ostriker , e. c. , gammie , c. f. : 1998 , , 508 , l99 stutzki , j. , gsten , r. 1990 , , 412 , 233 sugimoto , d. , chikada , y. , makino , j. , ito , t. , ebisuzaki , t. , umemura , m. 1990 , nature , 345 , 33 umemura , m. , fukushige , t. , makino , j. , ebisuzaki , t. , sugimoto , d. , turner , e.l . , loeb , a. 1993 , , 45 , 311 vainshtein , s.i . 1997 , , 56 , 6787 van der marel , r. p. , franx , m. 1993 , , 407 , 525 vzquez - semadeni , e. 1994 , , 423 , 681 vzquez - semadeni , e. , gazol , a. 1995 , , 303 , 204 vincent , a. , meneguzzi , m. 1991 , j. fluid mech . , 225 , 1 vio , r. , fasano , g. , lazzarin , m. , lessi , o. 1994 , , 289 , 640 williams , j.p . , de geus , e.j . , blitz , l. 1994 , , 428 , 693 williams , j. p. , blitz , l. , mckee , c. f. 2000 , in protostars and planets iv , eds . v. mannings , a. boss & s. russell , in press

in 2010 , tom halverson and georgia benkart introduced the motzkin algebra , a generalization of the temperley - lieb algebra , whose elements are diagrams that can be multiplied by stacking one on top of the other . halverson and benkart gave a diagrammatic algorithm for decomposing any motzkin diagram into diagrams of three subalgebras : the right planar rook algebra , the temperley - lieb algebra , and the left planar rook algebra . we first explore the right and left planar rook monoids , by finding presentations for these monoids by generators and relations , using a counting argument to prove that our relations suffice . we then turn to the newly - developed motzkin monoid , where we describe halverson s decomposition algorithm algebraically , find a presentation by generators and relations , and use a counting argument but with a much more sophisticated algorithm .

1301.4518

in this paper , the diagram monoids @xmath0 and one of its submonoids @xmath1 are first introduced in order to get an understanding of the diagram monoids @xmath2 and @xmath3,which are submonoids of @xmath1 . presentations of the latter two are provided in this paper , as well as references to papers in which presentations for the former two are provided . we then introduce the well known temperley - lieb algebra , @xmath4 . with these diagram monoids and the temperley - lieb algebra , we can give a presentation of the motzkin monoid , @xmath5 . in particular , we show that any diagram in @xmath5 can be decomposed into a product of the form @xmath6 , where @xmath7 , @xmath8 , and @xmath9 . ( note there is a particular form called standard " form mainly involving the placement of empty vertices . ) this is particularly important because of the inductive nature of our proof . + we start with a word in @xmath6 form , append a generator , @xmath10 , to the end of the word , and then try to get it back into @xmath6 form . we apply the relations of @xmath5 to first get @xmath11 into the form @xmath12 ( which we call @xmath13 form ) , where @xmath14 , and @xmath8 . from this @xmath13 form , we then get to minimal @xmath6 form , which is done by taking a diagram in @xmath13 form , and moving all the empty vertices to the right . this is done using five lemmas ( hop , burrow , slide , wallslide , and fuse wire ) , and also putting an order on @xmath1 . from here , we concern ourselves with the case that the edges in @xmath15 in minimal @xmath6 form have endpoints that are empty vertices of @xmath16 or @xmath17 . we show that we can put words of this form into standard @xmath6 form ( where standard @xmath6 form is as defined in section 4.3.1 ) . + the preceding inductive proof tells us that the number of distinct words is equal to the number of standard words , and thus equal to the number of diagrams in @xmath5 , giving us a presentation of @xmath5 . we define the _ rook monoid _ , denoted @xmath0 , as the set of one - to - one functions from a subset of @xmath18 to a subset of @xmath18 . these functions can be written as diagrams : a graph on two rows of @xmath19 vertices , each labeled from @xmath20 to @xmath19 from left to right . on the graph we connect vertices on the top line to vertices on the bottom line . take for example the following diagram in @xmath21 : + @xmath22 the rook monoid is a _ monoid _ under the operation of function composition , where the functions go from the top row of vertices to the bottom . to perform function composition with two diagrams @xmath23 , @xmath24 , we place @xmath23 above @xmath24 and identify the vertices in the bottom row of @xmath23 with the corresponding vertices in the top row of @xmath24 . in terms of functions , if we consider @xmath23 as the function @xmath25 , and @xmath24 as the function @xmath26 , then the product @xmath27 is thought of as the composition @xmath28 , where if @xmath29 is in the range of @xmath25 , and in the domain of @xmath26 , we get that the pre - image of @xmath29 , say @xmath30 , in @xmath25 is in the domain of @xmath28 , and the image of @xmath29 , say @xmath31 , in @xmath26 is in the range of @xmath28 . in particular , @xmath30 is in the domain of @xmath28 if and only if there exists @xmath29 in the domain of @xmath26 such that @xmath32 . this @xmath30 then corresponds to the value @xmath31 in the range of @xmath28 such that @xmath33 . note that in the diagram form we then draw an edge from @xmath30 on top to @xmath31 on bottom . for example , + if @xmath23= units < 0.3cm,0.15 cm > x from 1 to 6 , y from 0 to 3 = 0.3pt at 1 -1 at 1 2 at 2 -1 at 2 2 at 3 -1 at 3 2 at 4 -1 at 4 2 at 5 -1 at 5 2 1 -1 1 2 / 3 -1 2 2 / 5 -1 3 2 / 2 -1 4 2 / 4 -1 5 2 / and @xmath24= units < 0.3cm,0.15 cm > x from 1 to 6 , y from 0 to 3 = 0.3pt at 1 -1 at 1 2 at 2 -1 at 2 2 at 3 -1 at 3 2 at 4 -1 at 4 2 at 5 -1 at 5 2 1 -1 1 2 / 2 -1 3 2 / 3 -1 5 2 / 4 -1 4 2 / units < 0.3cm,0.15 cm > x from 1 to 12 , y from 0 to 3 = 0.3pt at 1 -3.5 at 1 -.5 at 1 1 at 1 4 at 2 -3.5 at 2 -.5 at 2 1 at 2 4 at 3 -3.5 at 3 -.5 at 3 1 at 3 4 at 4 -3.5 at 4 -.5 at 4 1 at 4 4 at 5 -3.5 at 5 -.5 at 5 1 at 5 4 at -4 .5 at 8 2 at 8 -1 at 9 2 at 9 -1 at 10 2 at 10 -1 at 11 2 at 11 -1 at 12 2 at 12 -1 at 6.5 .5 1 1 1 4 / 2 1 4 4 / 2 4 3 1 / 3 4 5 1 / 4 4 2 1 / 5 4 4 1 / 1 -.5 1 -3.5 / 2 -3.5 3 -.5 / 3 -3.5 5 -.5 / 4 -3.5 4 -.5 / 8 2 8 -1 / 9 2 9 -1 / 10 2 10 -1 / 12 2 11 -1 / another way to think about the rook monoid , @xmath0 , is as the set of @xmath34 matrices which have entries in the set @xmath35 , with the property that there is at most one @xmath20 in each row and each column . take for example , @xmath36 , which consists of the matrices : @xmath37 note that the elements in @xmath0 are in a one - to - one correspondence with the possible ways in which one can place non - attacking rooks on a @xmath34 chess board . we define the _ rank _ of a diagram as the number of edges it has , or the number of @xmath20s in the corresponding matrix . + if we desire a rook matrix of rank @xmath38 , we choose @xmath38 columns and @xmath38 rows in a total of @xmath39 ways . we then can choose to place the @xmath20s in the @xmath38 chosen rows / columns in a total of @xmath40 ways ( such that no row or column has two or more nonzero entries ) . summing over all the possible ranks up to and including @xmath19 gives rise to the order of @xmath0 : @xmath41 the relationship between diagrams and matrices is given in a very natural way . we connect the vertex in the @xmath42 position in the top row of a diagram to the @xmath43 position in the bottom row if and only if the corresponding matrix has a 1 in the ( @xmath44,@xmath45)-position . take for example the following matrix - diagram correspondence in @xmath21 : @xmath46 note that matrix multiplication is equivalent to diagram multiplication by stacking : @xmath47 is equivalent to @xmath48 a presentation of the rook monoid is provided on page 339 in @xcite . the _ planar rook monoid _ , @xmath1 , is the set of order - preserving one - to - one functions from a subset of @xmath49 to a subset of @xmath49 . these order - preserving functions correspond to those diagrams that can be drawn with edges that do not cross . for example , the set @xmath50 consists of the following diagrams : @xmath50 = @xmath51 @xmath52 @xmath53 @xmath54 @xmath55 @xmath56 for a diagram @xmath57 , we define @xmath58 and @xmath59 to be the sets containing the indices of the vertices of @xmath57 which are incident to an edge on top and on bottom respectively . for example , if @xmath57 = @xmath60 , then @xmath61(@xmath57 ) = \{1,2,3,5 } and @xmath62(@xmath57 ) = \{@xmath63,@xmath64,@xmath65,@xmath66 } + where we label the top vertices from 1 to @xmath19 and the bottom vertices from @xmath63 to @xmath67 . for a planar rook diagram @xmath57 , there is only one way to connect the vertices by edges , thus these sets @xmath61(@xmath57 ) and @xmath62(@xmath57 ) uniquely determine @xmath57 . + notice that the product of two planar rook diagrams is planar ( seen easily through diagram multiplication ) , thus @xmath1 is a submonoid of @xmath0 . to obtain the order of @xmath1 , as in @xmath0 , we choose @xmath38 columns , and @xmath38 rows in @xmath39 , however as stated above this choosing determines a unique diagram , thus the total number of planar rook diagrams is : @xmath68 let @xmath69 be the element of @xmath1 such that @xmath70\setminus\{i+1\}$ ] and @xmath71\setminus\{i\}=\{1,2,\ldots , i-1,i+i,\ldots , n\}$ ] , as shown below : @xmath72 let @xmath73 be the element of @xmath1 such that @xmath74\setminus\{i\}$ ] and @xmath75\setminus\{i+1\}$ ] as shown below : @xmath76 as proven in @xcite , every planar rook diagram can be written as a product of @xmath69 s and @xmath73 s . the following relations hold for all @xmath44 such that the terms in the relation are defined : 1 . @xmath77 2 . \a ) @xmath78 + b ) @xmath79 3 . \a ) @xmath80 + b ) @xmath81 4 . \a ) @xmath82 + b ) @xmath83 + c ) @xmath84 + d ) @xmath85 5 . @xmath86 6 . if @xmath87 , then @xmath88 these relations can easily be verified by drawing the diagram products they refer to . furthermore , as proven in @xcite , these relations suffice to completely characterize @xmath1 . we define the _ right planar rook monoid _ to be the submonoid of the planar rook monoid which contains all the diagrams with the property that the top vertex of each edge is directly above or above and to the right of the bottom vertex . similarly , the _ left planar rook monoid _ is the submonoid where the top vertex of each edge is above or above and to the left of the bottom vertex . we denote these monoids @xmath2 and @xmath3 respectively . for example , @xmath89 and @xmath90 are in @xmath91 and @xmath92 respectively . + we will now derive some facts about @xmath2 . first , we prove a closed form for the size of @xmath2 . through numerical experimentation we find that for the first few values of @xmath19 , we get : @xmath93 @xmath94 has diagrams : + @xmath95 @xmath96 + @xmath97 has diagrams : + @xmath51 @xmath98 @xmath99 @xmath100 @xmath101 + @xmath102 has diagrams : + @xmath103 @xmath104 @xmath105 @xmath106 @xmath107 @xmath108 @xmath109 @xmath110 @xmath111 @xmath112 @xmath113 @xmath114 @xmath115 @xmath116 + we notice that these cardinalities are catalan numbers , and indeed we can prove that this pattern continues : @xmath117 we begin by defining a new way to encode any diagram @xmath118 . to do this we create ordered pairs of the form @xmath119 for each edge in @xmath57 connecting the @xmath44th vertex on top to the @xmath45th vertex on bottom . we then gather all such pairs into a set along with the pair @xmath120 . we impose a well ordering on this set given by @xmath121 . for example , the diagram given by : @xmath122 yields the poset @xmath123 . now we begin by exhibiting a bijection from @xmath2 to the set of all @xmath124-sequences of @xmath125s , with exactly @xmath126 of each , such that no partial sum is negative . the set of such sequences is known to have cardinality @xmath127 . + define a function from @xmath2 to the set is as follows : given any diagram @xmath128 , start our sequence with @xmath129 @xmath20s followed by @xmath130 @xmath131s . inductively , append to this sequence @xmath132 @xmath20s and @xmath133 @xmath131s for @xmath134 . to see that the resulting sequence is indeed in our set , note that diagrams in @xmath2 must consist of ordered pairs with their first component greater than or equal to their second . this , in combination with the fact that the first coordinate of every pair is appended first as @xmath20s , makes it impossible to have a negative partial sum . + to see that this function is a bijection , note that there is a natural inverse . given any sequence , we create the first ordered pair by counting off all of the @xmath20s until the first @xmath131 , as well as the number of @xmath131s up to the next @xmath20 . these numbers will be the first and second coordinates of the first ordered pair , @xmath135 . inductively we count the @xmath44th set of @xmath20s and the @xmath44th set of @xmath131s . these numbers would then be added to @xmath136 and @xmath137 to give an ordered pair @xmath138 . therefore the cardinality of @xmath2 is equal to @xmath127 . let @xmath73 be defined as in subsection [ copying kathy ] . let @xmath139 be the element of @xmath2 such that @xmath140\setminus\{i\}$ ] as shown : @xmath141 this yields the following theorem . [ generators]every diagram of @xmath2 can be written as a product of @xmath142 and @xmath143 . suppose @xmath57 is a diagram in @xmath2 with rank @xmath38 . let @xmath144 with @xmath145 and @xmath146 with @xmath147 . define @xmath148 to be the identity diagram . for @xmath149 , define @xmath150 to be @xmath151 then @xmath152 is the diagram consisting of an edge from vertex @xmath153 to @xmath154 and vertical edges connecting all vertices to the right of @xmath154 or to the left of @xmath153 . let , @xmath155 all that remains to be shown is that @xmath156 . consider a node @xmath157 . all diagrams in @xmath158 have a vertical edge at @xmath159 . so do the diagrams @xmath160 for @xmath161 . then @xmath162 connects nodes @xmath159 and @xmath163 . finally , @xmath160 has a vertical edge at @xmath163 for @xmath164 . thus @xmath165 connects nodes @xmath159 and @xmath163 . if @xmath166 then the product of @xmath139 s ensures @xmath167 . thus @xmath168 is a word in our generators . note @xmath139 is not considered a generator in @xmath1 , since @xmath169 , for all @xmath170 and that @xmath171 for all @xmath172 . [ rpn relations ] the right planar rook monoid is generated by @xmath73 and @xmath139 subject to the following relations : 1 . @xmath173 2 . @xmath174 3 . @xmath175 4 . @xmath176 5 . @xmath177 6 . @xmath178 7 . if @xmath179 then @xmath180 theorem [ generators ] shows that @xmath73 and @xmath139 generate all of @xmath2 , so every diagram can be written as a formal word . we must show that there are at most as many distinct equivalence classes of words as there are diagrams . to this end we define the standard word as follows : a word @xmath181 is said to be in standard form if there exist subsets of @xmath182 $ ] , @xmath183 and @xmath15 with the following three properties : + @xmath184 the sets @xmath183 and @xmath15 have equal cardinality , say @xmath38 . + let @xmath185 and @xmath186 be the @xmath44th elements @xmath183 and @xmath15 respectively . for example , if @xmath187 , then @xmath188 . + @xmath184 for all @xmath189 . + let @xmath190 be defined as is theorem [ generators ] . + @xmath184 @xmath191 , where @xmath192 and @xmath193 since every diagram corresponds to exactly one standard word , we must show that each formal word is equal to a standard word . given a standard word @xmath194 and a generator @xmath195 we show that we can standardize the product @xmath196 . we will assume @xmath181 is not the empty word , since if @xmath181 is the empty word , after appending either of the generators the product is obviously equivalent to a standard word . first consider the product @xmath197 . we have two cases : * case 1 : @xmath198 . then necessarily @xmath139 appears in the product @xmath181 . since @xmath178 , @xmath181 can be written in the form : @xmath199 where @xmath200 does not contain @xmath139 . then using the relation @xmath201 we have that @xmath202 . * case 2 : @xmath203 . this means that @xmath204 . so if @xmath205 then @xmath206 . if @xmath207 then there exists @xmath45 such that @xmath208 for @xmath209 . we commute the @xmath139 next to the @xmath73 and use the relation @xmath210 to remove the @xmath211 s in @xmath212 . then we commute the @xmath213 s past the @xmath211 s . thus , @xmath214 now consider @xmath215 . let @xmath216 be the rightmost letter of @xmath217 there are four cases to consider : * case 1 : @xmath87 . since any of the generators commute when the indices are at least two apart , @xmath73 can be moved to the left . * case 2 : @xmath218 . since @xmath219 , we can eliminate @xmath73 . * case 3 : @xmath220 . since @xmath221 we can move @xmath73 to the left . * case 4 : @xmath222 . using the relation @xmath223 , we can eliminate @xmath224 . these relations can then be applied repeatedly until the @xmath211 s have been eliminated or commuted past @xmath225 . if we eliminate the @xmath211 s then we have that @xmath226 . if the @xmath211 s are commuted past @xmath225 then we must show that words of the form @xmath227 can be standardized . let @xmath224 be the letter immediately to the left of @xmath228 . we again have four cases depending on the difference between @xmath229 and @xmath45 : * case 1 : @xmath230 . since any of the generators commute when the indices are at least two apart , @xmath228 can be moved to the left . * case 2 : @xmath231 . we have that @xmath232 , which is already in standard form . * case 3 : @xmath233 . this implies that @xmath234 , which is already in standard form . * case 4 : @xmath235 . note that @xmath236 . moving the @xmath237 to the right , through @xmath211 s , is equivalent to moving @xmath211 s to the left through @xmath213 s , which is doable , as proven above . this process can be repeated as long as the word is not in standard form . so we have proven that @xmath196 can be standardized . by induction , this implies that when a string of arbitrary length is appended to a word in standard form the product can be standardized . thus any word can be put into standard form , implying that our relations are sufficient . for every diagram @xmath57 in @xmath2 , let @xmath238 be the diagram obtained by interchanging the vertices in the top and bottom rows of @xmath57 while maintaining edge connections . notice that @xmath239 for all @xmath44 such that the terms are defined . thus the function @xmath240 is an antiisomorphism and an involution , meaning that @xmath240 is an isomorphism from @xmath2 to the opposite of @xmath2 ( which is @xmath3 ) , and @xmath240 is its own inverse . therefore @xmath3 is antiisomorphic to @xmath2 , and from every theorem about @xmath2 one can easily derive a corresponding theorem regarding @xmath3 . in particular , it can easily be shown that @xmath3 is generated by @xmath69 and @xmath139 and that it is completely characterized by relations analogous to those in theorem [ rpn relations ] . a _ temperley - lieb diagram _ has two rows of @xmath19 vertices , connected by non - crossing edges . unlike planar rook diagrams , temperley - lieb diagrams do not have any empty vertices and they may have _ horizontal edges_. a horizontal edge is an edge connecting two distinct vertices in the same row . the collection of linear combinations of temperley - lieb diagrams forms the temperley lieb algebra , denoted @xmath241 . to multiply two temperley - lieb diagrams @xmath23 and @xmath24 , we place @xmath23 on top of @xmath24 , and identify the bottom vertices of @xmath23 with the top vertices of @xmath24 . additionally , horizontal edges between vertices in the bottom set of @xmath23 and top set of @xmath24 that form a closed loop produce a _ bubble_. we eliminate the bubbles and multiply the product diagram by a factor of @xmath29 for each bubble that is removed . as discussed in @xcite , every temperley - lieb diagram can be written as the product of diagrams @xmath242 , where @xmath243 for @xmath244 . the algebra @xmath241 is completely characterized by the following relations @xmath245 these relations are sufficient to show that any string of temperley - lieb generators is equivalent to a string of the form @xmath246 for some integers @xmath247 with the properties that @xmath248 the above form is called the _ standard form _ of a temperley - lieb word . + we will need to use some non - standard forms of words in the temperley - lieb algebra . for any temperley - lieb diagram , @xmath57 , in which vertex @xmath249 is connected to @xmath45 or @xmath250 , we can express @xmath57 as a word in one of the following five forms : 1 . if @xmath249 is connected to @xmath250 with @xmath251 . then @xmath57 can be written in the form + @xmath252 , where @xmath253 . note that if @xmath254 then the word is of the form @xmath255 . 2 . if @xmath249 is connected to @xmath250 with @xmath256 . then @xmath57 can be written in the form + @xmath257 , where @xmath258 . note that if @xmath259 then the word is of the form @xmath260 . 3 . if @xmath249 is connected to @xmath45 with @xmath256 . then @xmath57 can be written in the form + @xmath261 , where @xmath262 4 . if @xmath249 is connected to @xmath45 with @xmath263 . then @xmath57 can be written in the form + @xmath264 where @xmath265 . 5 . if @xmath249 is connected to @xmath45 with @xmath266 . then @xmath57 can be written in the form @xmath267 , where @xmath268 , @xmath269 . given that these are the only five ways the vertex @xmath249 is connected to @xmath45 or @xmath250 , by drawing diagrams , one can see that these are indeed the ways in which temperley - lieb diagrams can be drawn . for example , in _ 1 . _ , simply take the diagram @xmath270 and perform multiplication by stacking it above the diagram @xmath271 , along with any chosen diagram @xmath272 above this product , and @xmath273 below , and one can verify that indeed this gives the desired connection from @xmath249 to @xmath250 with @xmath274 . in this paper we will consider multiplication of diagrams , but we will not consider linear combinations of diagrams . for simplicity , we will therefore take @xmath29 to be @xmath20 and refer exclusively to the monoid @xmath4 , although all results can be generalized to the algebra @xmath241 , by multiplying by @xmath29 where appropriate . a motzkin diagram is similar to a temperley - lieb diagram , however motzkin diagrams may have _ empty _ vertices ( vertices not incident to an edge ) . for example : + @xmath57 = units < 0.5cm,0.3 cm > x from 0 to 8 , y from -1 to 2 at 1 -1 at 1 2 at 2 -1 at 2 2 at 3 -1 at 3 2 at 4 -1 at 4 2 at 5 -1 at 5 2 at 6 -1 at 6 2 at 7 -1 at 7 2 4 2 1 -1 / 5 2 2 -1 / 6 2 7 -1 / 1 2 2 1 3 2 / 3 -1 4.5 .25 6 -1 / 4 -1 4.5 -.25 5 -1 / at 1 3.0 at 2 3.0 at 3 3.0 at 4 3.0 at 5 3.0 at 6 3.0 at 7 3.0 at 1 -2 at 2 -2 at 3 -2 at 4 -2 at 5 -2 at 6 -2 at 7 -2 + we define the _ motzkin monoid _ , @xmath5 , as the set of motzkin diagrams with @xmath19 vertices . for example @xmath275 consists of the following diagrams : + @xmath276 units < 0.55cm,0.55 cm > x from 1 to 2 , y from 0 to 1 at 1 0 at 1 1 at 2 0 at 2 1 1 0 1 1 / to multiply two motzkin diagrams , @xmath23 and @xmath24 , we place @xmath23 on top of @xmath24 , and identify the bottom vertices of @xmath23 with the top vertices of @xmath24 . as with temperley - lieb daigrams , closed loops are discarded . for example , + @xmath277 then @xmath278 for any motzkin diagram @xmath57 , the symbols @xmath58 and @xmath59 refer to the sets of non - empty top and bottom vertices respectively , as in the planar rook monoid . in @xcite , it is proved that any diagram in the motzkin monoid can be written in the form @xmath57=@xmath279 where @xmath280 , @xmath281 , and @xmath282 . we now describe the algorithm to compute @xmath211 , @xmath283 , and @xmath229 for any given @xmath57 . as in @xcite , we let @xmath284 be the diagram with top set @xmath285 and bottom set , @xmath286 . let @xmath287 be the diagram with bottom set , @xmath288 and top set , @xmath289 \{1, ,@xmath290}. to obtain @xmath283 , first shift the isolated vertices of @xmath57 to the right of the diagram , while preserving connections to the other vertices , to produce the diagram @xmath165 . for example , if @xmath57 is the diagram below , then @xmath165 would be the diagram below it . @xmath57 = units < 0.5cm,0.3 cm > x from 0 to 8 , y from -1 to 2 at 1 -1 at 1 2 at 2 -1 at 2 2 at 3 -1 at 3 2 at 4 -1 at 4 2 at 5 -1 at 5 2 at 6 -1 at 6 2 at 7 -1 at 7 2 4 2 1 -1 / 5 2 2 -1 / 6 2 7 -1 / 1 2 2 1 3 2 / 3 -1 4.5 .25 6 -1 / 4 -1 4.5 -.25 5 -1 / at 1 3.0 at 2 3.0 at 3 3.0 at 4 3.0 at 5 3.0 at 6 3.0 at 7 3.0 at 1 -2 at 2 -2 at 3 -2 at 4 -2 at 5 -2 at 6 -2 at 7 -2 + @xmath165= units < 0.5cm,0.3 cm > x from 0 to 8 , y from -1 to 2 at 1 -1 at 1 2 at 2 -1 at 2 2 at 3 -1 at 3 2 at 4 -1 at 4 2 at 5 -1 at 5 2 at 6 -1 at 6 2 at 7 -1 at 7 2 3 2 1 -1 / 4 2 2 -1 / 5 2 7 -1 / 1 2 1.5 1 2 2 / 3 -1 4.5 .25 6 -1 / 4 -1 4.5 -.25 5 -1 / at 1 3.0 at 2 3.0 at 3 3.0 at 4 3.0 at 5 3.0 at 6 3.0 at 7 3.0 at 1 -2 at 2 -2 at 3 -2 at 4 -2 at 5 -2 at 6 -2 at 7 -2 next we deal with the issue of filling in the empty vertices to complete the temperley - lieb diagram . let @xmath291 and @xmath292 denote the number of empty vertices on the top and bottom of @xmath165 respectively . for example , if @xmath165 is as above , then @xmath293 and @xmath294 . note that the difference between @xmath291 and @xmath292 must even . we then have two cases for filling in the empty vertices : * case 1 : @xmath295 . to turn @xmath165 into a temperley - lieb diagram , first add horizontal edges on top from @xmath45 to @xmath296 for @xmath297 . now , add vertical edges from @xmath45 to @xmath250 for @xmath298 * case 2 : @xmath299 . to turn @xmath165 into a temperley - lieb diagram , first add horizontal edges on bottom from @xmath250 to @xmath300 for @xmath301 . now , add vertical edges from @xmath45 to @xmath250 for @xmath302 this gives a well defined way to decompose each diagram in the motzkin monoid . thus @xmath303 . [ mek ] the motzkin monoid is generated by @xmath242 , @xmath73 and @xmath69 subject to the following relations . @xmath304 2 . @xmath305 3 . @xmath306 @xmath307 4 . @xmath308 @xmath309 @xmath310 @xmath311 5 . @xmath312 6 . @xmath313 7 . @xmath314 8 . @xmath315 9 . @xmath316 10 . @xmath317 @xmath318 11 . a ) @xmath319 12 . @xmath320 each of these relations can be easily verified by drawing the diagrams in question . for example the picture below justifies the relation @xmath321 . @xmath322 in this section we prove theorem [ mek ] . to start we will define a set of formal words which are subject exactly to the relations in theorem [ mek ] . next we want to show that there is an isomorphism between this set and @xmath5 . we define @xmath323 to be the monoid generated by @xmath324 , @xmath325 , and @xmath326 subject to the same relations as above , and we let @xmath327 be the map given by @xmath328 @xmath329 @xmath330 since the diagrams @xmath331 and @xmath69 satisfy the relations on @xmath332 , the operations are preserved ; thus @xmath333 is a homomorphism . in order to prove that @xmath333 is also an isomorphism we want to show that @xmath323 has at most as many distinct elements as there are diagrams in @xmath5 . we begin by defining a standard word in @xmath323 . given a word @xmath57 in @xmath334 , let @xmath335 and @xmath336 . we say @xmath57 is in _ standard form _ if @xmath337 where @xmath338 + we then have three possibilities for @xmath15 : * if @xmath339 then , @xmath340 . * if @xmath341 then , @xmath342 . * if @xmath343 then , @xmath344 . in all the three cases , @xmath15 is in standard form for temperley - lieb words . this definition of a standard word corresponds to the diagram decomposition described by halverson in @xcite . by the definition of @xmath333 , @xmath323 maps to @xmath5 . so if we show that we can standardize any formal word in @xmath323 then we will have proven theorem [ mek ] . the following relations can be derived from the relations on @xmath323 and will be useful : * @xmath345 * @xmath346 * @xmath347 * @xmath348 * @xmath349 * @xmath350 in order to prove that every element of @xmath323 is equivalent to a standard word , note that the identity is a standard word . next , given a standard word @xmath181 and a generator @xmath351 , @xmath352 , @xmath353 we show that our relations are sufficient standardize the product @xmath196 . in order to do this we proceed in three steps : first we get the word @xmath196 into the form @xmath12 , known as @xmath13 form , where @xmath354 and @xmath8 . next we manipulate @xmath13 form in order to obtain minimal @xmath6 form where @xmath355 . after this , we manipulate the temperley - lieb diagram to obtain standard form . + [ rtlx to ptp ] if @xmath6 is a word in standard form and @xmath356 , then @xmath357 is equal to a word in @xmath13 form . notice that appending an @xmath358 or an @xmath352 yields a word already in @xmath13 form , so it suffices to show that @xmath359 is equivalent to a word in @xmath13 form . we will proceed using reverse induction on @xmath44 . + for the base case we append @xmath360 . there are several cases to consider depending on what the bottom set of @xmath17 is : + case 1 : @xmath361 . + because @xmath57 is in standard form , there exist integers @xmath45 and @xmath38 , with @xmath362 and @xmath363 such that @xmath364 . for simplicity , let @xmath365 . let @xmath366 thus @xmath367 by the relation @xmath368 for @xmath369 , we rearrange the terms to obtain @xmath370 . the indices of all the components of @xmath371 are strictly less than @xmath372 , which is in turn strictly less than @xmath44 , implying that the highest index of any of @xmath371 s factors is less than or equal to @xmath373 . thus , @xmath242 can commute past all these elements . thus , @xmath374 by repeatedly applying the relation @xmath375 we obtain @xmath376 all the factors of @xmath371 have index greater than or equal to @xmath377 , implying that @xmath378 commutes with @xmath371 . therefore @xmath379 . thus @xmath380 . + case 2 : @xmath381 . since the @xmath382 and @xmath67 vertices are empty , we have that @xmath383 where @xmath384 . we have then , @xmath385 now to get to @xmath13 form we need to move @xmath386 . note @xmath387 , so @xmath388 . since @xmath6 is in standard form , this tells us something about @xmath15 . there are three subcases depending on @xmath389 . subcase 1 : @xmath390 . then @xmath391 in this case every element of @xmath15 commutes with @xmath392 , and so @xmath393 subcase 2 : @xmath394 . then @xmath15 can be written in the form @xmath395 , where @xmath396 . so @xmath397 then @xmath398 subcase 3 : @xmath399 . then @xmath15 can be written as @xmath400 , where @xmath396 . by our relations we get , @xmath401 in all three subcases , @xmath402 is equal to a word in @xmath13 form . + case 3 : @xmath403 . since the @xmath19th vertex in @xmath404 is empty , @xmath405 . using our relations , this implies that @xmath406 now we are reduced to case 2 . + case 4 : @xmath407 . this case can be reduced to case 2 using analogous reasoning . + in conclusion , we have shown that if @xmath408 is any word in standard form , then @xmath402 is equivalent to a word in @xmath13 form . we now move on to the inductive step to show that this holds for any @xmath242 . assume that for @xmath409 @xmath410 can be put into @xmath13 form . we now show that this assumption implies that @xmath411 can be put into @xmath13 form . consider the following three cases : + case 1 : @xmath412 since @xmath413 we can write @xmath17 as @xmath414 . thus , @xmath415 the factor @xmath416 right can be put into @xmath13 form , by the inductive hypothesis . thus we have @xmath417 , which is in @xmath13 form . + case 2 : @xmath418 we know that @xmath17 can we written as @xmath419 where @xmath420 . so , @xmath421 case 3 : @xmath422 since @xmath17 is in standard form , and @xmath423 , we know that we can not have a vertical line from @xmath424 to @xmath425 . furthermore , @xmath423 allows us to write @xmath17 equivalently as @xmath426 ( because necessarily @xmath427 are in our product @xmath428 ) , where @xmath429 is defined by @xmath430 and@xmath431 . so we have , @xmath432 the factor in parenthesis can be put into @xmath12 form , by the inductive hypothesis . thus @xmath433 . note that @xmath434 contains @xmath435 . so @xmath436 can be put into standard form , by case 2 . + case 4 : @xmath437 note that in this case @xmath438 . using our relations we have , @xmath439 . note that @xmath44 and @xmath440 are not in @xmath441 , implying that this word can be standardized by the previous case . + case 5 : @xmath442 this case is analogous to case 4 . + rewriting a word in @xmath13 form as a word in standard form is analogous to taking a motzkin diagram drawn as the product of a temperley - lieb diagram and two planar rook diagrams and redrawing it so that the empty vertices in the bottom of the top diagram and the empty vertices in the top of the bottom diagram are all on the far right . this motivates the following lemmas , each of which concerns words in @xmath334 but corresponds to shifting empty vertices rightward in diagrams in @xmath5 . [ hopping ] let @xmath44 and @xmath38 be natural numbers . if @xmath443 and @xmath444 is even then , @xmath445 intuitively we are hopping " a dead end in the @xmath44th vertex over a series of horizontal edges to the @xmath38th vertex as shown below : @xmath446{$t_0$}$ } at 4 0.45 % \put{$\cdots$ } at 4 2 % \put{$\cdots$ } at 4 -3 % \put{$\cdots$ } at 4 -7 \plot 1 1 1 -3 / \plot 2 -3 2 -7 / \plot 3 -3 3 -7 / \plot 4 -3 4 -7 / \plot 5 -3 5 -7 / \setquadratic \plot 2 -3 2.5 -2 3 -3 / \plot 4 -3 4.5 -2 5 -3 / \plot 2 1 2.5 0 3 1 / \plot 4 1 4.5 0 5 1 / \put{$\scriptstyle{i\,}$ } at 1 -8 \put{$\scriptstyle{k\,}$ } at 5 -8 \endpicture } { \beginpicture \setcoordinatesystem units < 0.5cm,0.3 cm > \setplotarea x from 1 to 5 , y from 0 to 2 \put{$\bullet$ } at 1 1 % ( 1 ) \put{$\bullet$ } at 2 1 \put{$\bullet$ } at 3 1 \put{$\bullet$ } at 4 1 \put{$\bullet$ } at 5 1 \put{$\bullet$ } at 1 -3 \put{$\bullet$ } at 2 -3 \put{$\bullet$ } at 3 -3 \put{$\bullet$ } at 4 -3 \put{$\bullet$ } at 5 -3 % \put{$\framebox[0.9 in]{$t_0$}$ } at 4 0.45 % \put{$\cdots$ } at 4 2 % \put{$\cdots$ } at 4 -3 % \put{$\cdots$ } at 5 -7 \put{$\scriptstyle{i\,}$ } at 1 -8 \put{$\scriptstyle{k\,}$ } at 5 -8 \plot 1 -3 2 -7 / \plot 2 -3 3 -7 / \plot 3 -3 4 -7 / \plot 4 -3 5 -7 / \setquadratic \plot 1 1 2 -.5 3 -1 / \plot 3 -1 4 -1.5 5 -3 / \plot 1 -3 1.5 -2 2 -3 / \plot 3 -3 3.5 -2 4 -3 / \plot 2 1 2.5 0 3 1 / \plot 4 1 4.5 0 5 1 / \put{$\bullet$ } at 1 -7 \put{$\bullet$ } at 2 -7 \put{$\bullet$ } at 3 -7 \put{$\bullet$ } at 4 -7 \put{$\bullet$ } at 5 -7 \endpicture}\ ] ] we prove lemma [ hopping ] by induction on @xmath38 . + the case @xmath447 is simply the relation @xmath448 . + assume the result holds for @xmath449 : @xmath450 then @xmath451 [ slide ] if @xmath452 , then @xmath453 for some @xmath454 diagrammatically we are sliding " the product of @xmath229 s under the product of @xmath283 s , as shown below : @xmath455 we will prove lemma [ slide ] using induction on the length of the word @xmath15 . + base case : for our base case we want to show that @xmath456 for any @xmath457 . since @xmath458 $ ] then there exists an @xmath459 in our product of @xmath229 s . we will commute @xmath378 up to this point and then use our relations as follows : @xmath460 inductive step : now , assume @xmath461 where @xmath15 and @xmath462 are some products of @xmath283 s as described above . if we append some @xmath463 , we have that : @xmath464 where @xmath465 so by induction we have that @xmath466 as desired . [ burrow ] let i be a natural number . then , @xmath467 intuitively , we will use this to allow a horizontal edge to burrow under " a neighboring edge . @xmath468 the proof of this lemma follows easily from the relations @xmath469 and @xmath321 . thus , @xmath470 [ wallslide1 ] let @xmath44 be a natural number . let @xmath471 and @xmath472 be arbitrary . then @xmath473 . this is the case where we have a punctured " vertical line ( an edge from vertex @xmath44 on top to vertex @xmath249 on bottom ) , and want to move it over a non - punctured vertical linea wall" as in the diagram : @xmath474 since all of @xmath15 s factors have index less than @xmath44 and all of @xmath462 s factors have index greater than @xmath440 , we can commute @xmath73 to commute through @xmath15 and @xmath462 to get @xmath475 if vertex @xmath44 is connected to vertex @xmath45 , then @xmath44 being punctured is equivalent to @xmath45 being punctured . this fact can be seen from the diagram and is stated formally and proven below . @xmath476 [ fuse ] ( i ) if @xmath256 , @xmath477 is even , and @xmath478 then @xmath479 . + ( ii ) if @xmath480 , @xmath481 is even , and @xmath482 then @xmath479 . + ( iii ) if @xmath251 , @xmath481 is odd , and @xmath483 , then @xmath484 . + ( iv ) if @xmath256 , @xmath481 is odd , and @xmath485 , then @xmath486 . ( i ) by the commutativity relations , we can rearrange @xmath29 to be of the form @xmath487 this implies that @xmath488 this process can be repeated until @xmath489 has been transformed to @xmath490 . the equation @xmath491 can be derived by first noting that @xmath492 and using the above result to move one of the factors of @xmath216 to the far left . + ( ii ) the proof is analogous to that of the previous case . + ( iii ) @xmath493 $ ] . notice that the factor in brackets is of the form discussed in case ( ii ) . thus we obtain @xmath494 $ ] this is in turn equal to @xmath495 $ ] , by the relation @xmath496 . we can then commute the @xmath139 to the far right , yielding @xmath497 as desired . as discussed at the end of ( i ) , one can easily prove that @xmath498 . + ( iv ) the proof is analogous to that of the previous case . by putting an order on @xmath1 we can measure how close a given word is to being in @xmath6 form and thereby better describe the actions necessary to perform our decomposition and get the dead ends " , or non - incident vertices , on the right and incident vertices on the left . + let @xmath499 be any order on the power set of @xmath500 with the properties : + @xmath501if@xmath502 then @xmath503 + @xmath501if@xmath504 then @xmath505 + a word @xmath506 is said to be in _ minimal form _ if it is in the form @xmath6 for some @xmath507 , and @xmath508 with the property that for all @xmath509 , and @xmath510 such that @xmath511 @xmath512 in this case , @xmath16 and @xmath17 are said to be _ minimal_. [ ordering ] if @xmath513 and @xmath514 then @xmath515 with equality if and only if @xmath516 . notice that if @xmath517 , left multiplication by @xmath73 will send the top of @xmath225 to the @xmath518 index , leaving @xmath519 empty in the index @xmath44 on top , thus @xmath520 , hence @xmath521 . now if @xmath522 , then @xmath523 . furthermore , if @xmath524 , then left multiplication by @xmath224 leaves the @xmath525 vertex on top of @xmath526 empty , hence @xmath523 . left multiplication by @xmath224 only affects these two top vertices , thus in all other cases @xmath225 is unaffected . hence if @xmath527 , then @xmath528 , giving us the desired result . [ algorithm ] if @xmath529 is minimal , then the following three assertions hold : + ( 1 ) for all @xmath530 + if vertex @xmath249 is connected to @xmath250 in @xmath15 , then @xmath531 if and only if @xmath532 . + if vertex @xmath44 is connected to @xmath250 in @xmath15 , then @xmath533 if and only if @xmath532 . if vertex @xmath44 is connected to @xmath45 in @xmath15 , then @xmath533 if and only if @xmath534 . + ( 2 ) there exists a natural number @xmath38 such that @xmath535 + ( 3 ) there exists a natural number @xmath536 such that @xmath537 + \(1 ) if any of the implications are false , then the hypothesis for at least one of the cases of lemma [ fuse ] are satisfied , which implies that @xmath57 is not in minimal form . \(2 ) if @xmath538 , for any k , there exists an @xmath44 such that @xmath539 and @xmath540 there are five ways in which this can happen . in each case we can prove that @xmath541 , where @xmath542 and @xmath543 , implying that @xmath12 was not minimal . @xmath249 is connected to @xmath250 with @xmath251 . + then @xmath544 where @xmath545 . consider @xmath546 . we commute every temperley - lieb generator in @xmath547 with @xmath548 . with the @xmath548 in place , notice that the hypothesis for theorem [ hopping ] are satisfied where @xmath38 in theorem [ hopping ] is equal to @xmath549 . we hop the dead end at @xmath44 to @xmath549 . this yields : @xmath550 subsequently , we slide ( lemma [ slide ] ) to get : @xmath551 where @xmath552 . in conclusion , @xmath553 and by theorem [ ordering ] @xmath554 since @xmath555 . @xmath249 is connected to @xmath250 with @xmath556 . + in this case we have @xmath557 where @xmath558 and @xmath559 . since @xmath560 we have @xmath561 ( by lemma [ fuse ] ) . we can now obtain @xmath562 such that @xmath563 by referring to the previous case . @xmath249 is connected to @xmath45 with @xmath256 + in the case , @xmath15 is of the form @xmath564 where @xmath565 and @xmath566 . consider @xmath567 . as in case 1 we can commute @xmath139 through @xmath547 . notice that @xmath568 is even , and we have a divisor @xmath569 . thus by lemma [ fuse ] , we get : @xmath570 observe that taking the antiisomorphism of theorem [ hopping ] ( hop ) we obtain : @xmath571^ * & = & [ ( t_{k-1 } t_{k-3 } \ldots t_{i+1 } ) ( t_{k-2 } t_{k-4 } \ldots t_{i})(l_{k-1}l_{k-2 } \ldots l_i)]^*\\ p_i ( t_{i+1 } t_{i+3 } \ldots t_{k-1 } ) & = & ( r_{i}r_{i+1 } \ldots r_{k-1})(t_{i } t_{i+2 } \ldots t_{k-2})(t_{i+1 } t_{i+3 } \ldots t_{k-1})\end{aligned}\ ] ] hence by applying theorem [ hopping ] , @xmath572 as in case 1 we can use the relation @xmath573 if @xmath574 to simplify the product of @xmath283 s . now we use the relation @xmath575 and apply lemma [ wallslide1 ] to produce all the @xmath211 s between @xmath45 and @xmath576 . so , @xmath577 now the hypothesis for slide ( theorem [ slide ] ) are satisfied , so we obtain : @xmath578 where @xmath579 . in conclusion , @xmath580 notice that @xmath581 and @xmath582 by theorem [ ordering ] , since @xmath583 . case 4 . @xmath249 is connected to @xmath45 with @xmath263 + in this case @xmath15 can be written in the form @xmath584 where @xmath585 . consider @xmath567 . every element in @xmath547 commutes with @xmath139 . notice that the hypothesis in theorem [ hopping ] are satisfied for @xmath586 . thus , @xmath587 using the relations @xmath588 if @xmath589 and @xmath590 if @xmath179 as in case 1 , we can simplify the product of @xmath283 s to @xmath591 . now we use the relation @xmath592 and apply wallslide ( lemma [ wallslide1 ] ) to produce all the @xmath229 s between @xmath45 and @xmath576 , so that we will be able to slide the product of @xmath229 s ( theorem [ slide ] ) through @xmath547 . so we have : @xmath593 we can continue this process until we arrive at : + @xmath594 . + now the hypotheses for slide ( lemma [ slide ] ) are satisfied , we obtain : @xmath595 where @xmath596 . in conclusion , @xmath597 notice that @xmath598 and that @xmath599 , implying that the from @xmath12 was not minimal . case 5 : @xmath249 is connected to @xmath45 on top with @xmath266 + @xmath15 can be written as @xmath600 where @xmath601 , @xmath602 + the proof for this case is nearly identical to that of case 4 . we apply the hop relation to @xmath139 to get it through @xmath603 if it appears in the product ( otherwise we perform a wallslide ) , then produce @xmath73 and @xmath69 and commute @xmath73 through @xmath604 , and furthermore , obtain some product of @xmath229 s and apply theorem [ slide ] to slide them through @xmath547 . _ case 2 _ : if @xmath547 does not contain any factors of @xmath603 , we apply wallslide ( lemma [ wallslide1 ] ) to get @xmath605 . note that we can commute the @xmath73 through @xmath547 since @xmath606 , and @xmath607 thus @xmath608 and @xmath609 and @xmath610 , by theorem [ ordering ] since @xmath611 if ( 3 ) does not hold , we can prove that @xmath57 is not minimal using analogous reasoning . in this section we prove theorem [ std ] , namely , given a word @xmath506 written in the form @xmath6 , where @xmath16 and @xmath17 are minimal , @xmath57 can be written as a word in standard form . note that the difference between @xmath6 and standard form concerns edges in @xmath15 whose endpoints are empty vertices of @xmath16 or @xmath17 . the following lemma covers an extreme case . [ t death ] let @xmath15 be an arbitrary element of @xmath4 composed of generators whose indices are greater than or equal to some natural number @xmath38 , with @xmath612 . let @xmath613 . then @xmath614 we proceed by induction on the number of letters in @xmath15 . first consider the base case , when @xmath615 for some @xmath44 between @xmath38 and @xmath576 . then we obtain @xmath616 by the relations @xmath617 and @xmath618 ( from relation 9 ) . + assume that @xmath619 whenever @xmath15 is the product of @xmath536 letters . we wish to show that @xmath620 for any @xmath44 greater than or equal to @xmath38 , thereby proving that @xmath619 whenever @xmath15 is the product of @xmath621 letters . consider the following cases for what the @xmath249th vertex of @xmath15 is connected to to . case 1 : @xmath249 is connected to @xmath250 with @xmath622 . then @xmath15 is of the form @xmath623 where @xmath545 . @xmath624 case 2 : @xmath249 is connected to @xmath250 with @xmath625 . + then @xmath15 is of the form @xmath626 , where @xmath627 . @xmath628 case 3 : @xmath249 is connected to @xmath45 , with @xmath629 . + then @xmath15 is of the form @xmath630 where @xmath631 @xmath632 case 4 : @xmath249 is connected to @xmath45 , with @xmath256 . the proof of this case is analogous to case 3 . @xmath249 is connected to @xmath44 . this implies that @xmath15 does not contain @xmath242 or @xmath633 . thus we obtain @xmath634 , since @xmath635 . case 6 : @xmath249 is connected to @xmath636 . this implies by section [ okay ] that @xmath637 , for some @xmath462 . thus we obtain @xmath638 [ std ] let @xmath57 be a word in @xmath334 in the form @xmath6 where @xmath16 and @xmath17 are minimal . then @xmath57 can be put into standard form . we consider two cases . case 1 : @xmath639 assume without loss of generality that @xmath640 . this means that no vertex on bottom with index greater than @xmath641 can be connected to a vertex on bottom whose index is less than or equal to @xmath641 or any vertex on top with index less than or equal to @xmath642 . similarly , no vertex on top with index greater than @xmath642 can be connected to a vertex on top whose index is less than or equal to @xmath642 or any vertex on bottom with index less than or equal to @xmath641 . ( if this were not the case , @xmath17 or @xmath16 would not be minimal , as proven in theorem [ algorithm ] . ) the word @xmath57 can be written in the form @xmath643 for some @xmath644 and some @xmath510 . @xmath645 which is in standard form . case 2 : @xmath646 using similar reasoning as above , no vertex whose index is greater than @xmath38 can be connected to any vertex with index less than or equal to @xmath38 . this implies that @xmath15 can be written without any factors of @xmath647 . by the commutativity relations , @xmath57 can be written in the from @xmath648 , where @xmath649 , and @xmath650 , which is equal to @xmath651 by lemma [ t death ] , which in standard form . [ induction ] every word in @xmath334 is equivalent to a standard word in @xmath334 . we proceed by induction . obviously every single - letter word is equivalent to a standard word . assume that every word of @xmath38 letters is equivalent to a word in standard form . let @xmath57 be an arbitrary word of @xmath652 letters . let @xmath653 be the product of the first @xmath38 letters of @xmath57 , and let @xmath29 be the final letter of @xmath57 . by the inductive hypothesis , @xmath653 is equivalent to a standard word . if @xmath654 , then @xmath655 is already in ptp form . if @xmath656 , then , by lemma [ rtlx to ptp ] , @xmath655 can be put into @xmath13 form . thus , @xmath57 is equivalent to a word in @xmath13 from . by theorems [ algorithm ] and [ std ] , this implies that @xmath57 is equivalent to a word in standard from . thus every string of @xmath652 letters is equivalent to a word in standard form . thus every word in @xmath334 is equivalent to a word in standard from . the monoid @xmath323 is isomorphic to @xmath5 . by theorem [ induction ] , we know that the number of distinct words in @xmath323 is equal to the number of standard words . clearly there is a one - to - one correspondence between words in standard form and diagrams in @xmath5 which are in standard form . every diagram in @xmath5 is equivalent to a diagram in standard from . thus @xmath657 . this implies that the homomorphism @xmath333 is an isomorphism . in this paper we have introduced the rook , planar rook , left and right planar rook monoids , the temperley - lieb algebra , and the motzkin monoid . we have given a complete presentation of the motzkin monoid through a set of moves ( hop , burrow , wallslide ) , whose proof at times can get very involved . monoid presentations are very important in order for us to understand a monoid , and in order for us to define morphisms to other monoids . the motzkin monoid is used extensively in @xcite . another example of a presentation is completed in @xcite . in this paper , david penneys gives a presentation of the annular temperley - lieb category . the mathematical content described in this paper is derived from a collaboration over the summer of 2011 at the reu program at the university of california , santa barbara . the group consisted of megan ly ( loyola marymount university ) , eliezer posner ( city college of new york ) , and kristofer hatch ( university of california , santa barbara ) . + the reu was supported by the national science foundation research experience for undergraduates program ( nsf - reu grant dms-0852065 ) , and was conducted under the direction of our advisor , stephen bigelow .

this short review paper devoted to 90th anniversary of s. t. belyaev birthday . belyaev s ideas associated with the condensate state in bose interacting systems have stimulated intensive studies of the possible manifestation of such a condensation in fermi systems . in many fermi systems and compounds at zero temperature a phase transition happens that leads to a quite specific state called fermion condensation . as a signal of such a fermion condensation quantum phase transition ( fcqpt ) serves unlimited increase of the effective mass of quasiparticles that determines the excitation spectrum and creates flat bands . we show that the class of fermi liquids with the fermion condensate forms a new state of matter . we discuss the phase diagrams and the physical properties of systems located near that phase transition . a common and essential feature of such systems is quasiparticles different from those suggested by l. d. landau , by crucial dependence of their effective mass on temperature , external magnetic field , pressure etc . it is demonstrated that a huge amount of experimental data collected on different compounds suggests that they , starting from some temperature and down , form the new state of matter , and are governed by the fermion condensation . our discussion shows that the theory of fermion condensation develops completely good description of the nfl behavior of strongly correlated fermi systems . moreover , the fermion condensation can be considered as the universal reason for the nfl behavior observed in various hf metals , liquids , compounds with quantum spin liquids , and quasicrystals . we show that these systems exhibit universal scaling behavior of their thermodynamic properties . therefore , the quantum critical physics of different strongly correlated compounds is universal , and emerges regardless of the underlying microscopic details of the compounds . this uniform behavior , governed by the universal quantum critical physics , allows us to view it as the main characteristic of the new state of matter .

1311.0629

s. t. belyaev s contribution to theoretical physics is very impressive . in 1958 s. t. belyaev has published his classical works on the theory of nonideal bose - gas . in these works , s. t. belyaev demonstrates that peculiarity of bose systems comes from a microscopic number of particles in the condensed state with the momentum @xmath0 @xcite . to deal with such a system , he suggests splitting the system into two parts or subsystems , the condensate with @xmath0 and the rest with @xmath1 . it turns out that it is the condensate that creates the `` flavor '' of bose liquid , generating its vivid properties . one may try to figure out if physics like that of bose systems could be represented in fermi systems . belyaev s daring ideas of the two macroscopic parts in bose systems is adopted by a theory of fermion condensate that permits to construct the new class of strongly correlated fermi liquids with the fermion condensate ( fc ) @xcite , which quasiparticle system is also composed of two parts : one of them is represented by fc located at the chemical potential @xmath2 , and giving rise to the spiky density of states ( dos ) like that with @xmath0 of bose system . figure [ fig0 ] , panel * a * , shows dos of a fermi liquid with fc located at the momentum @xmath3 and energy @xmath4 . in contrast to the condensate of a bose system occupying the @xmath0 state , quasiparticles of fc with the energy @xmath4 must be spread out over the interval @xmath3 . [ ! ht ] at the momentum @xmath5 of a fermi liquid with fc . panel * b * , schematic plot of two - component fermi liquid at @xmath6 with fc . the system is separated into two parts shown by the arrows : the first part is a landau fermi liquid with the quasiparticle distribution function @xmath7 , and @xmath8 ; the second one is fc with @xmath9 and the single - particle spectrum @xmath10 . the fermi momentum @xmath11 satisfies the condition @xmath12.,title="fig:",scaledwidth=40.0% ] at the momentum @xmath5 of a fermi liquid with fc . panel * b * , schematic plot of two - component fermi liquid at @xmath6 with fc . the system is separated into two parts shown by the arrows : the first part is a landau fermi liquid with the quasiparticle distribution function @xmath7 , and @xmath8 ; the second one is fc with @xmath9 and the single - particle spectrum @xmath10 . the fermi momentum @xmath11 satisfies the condition @xmath12.,title="fig:",scaledwidth=37.0% ] the quasiparticle distribution function @xmath13 of fermi system with fc is determined by the ordinary equation for a minimum of the landau functional @xmath14 @xcite . in contrast to common functionals of the number density @xmath15 @xcite , the landau functional of the ground state energy @xmath14 becomes the exact functional of the occupation numbers @xmath16 . in case of homogeneous system a common functional becomes a function of @xmath17 , while e remains a functional , @xmath18 @xcite , @xmath19 equation represents an ordinary one to search the minimum of functional @xmath14 . in the case of bose system the equation @xmath20 describes a common instance . in the case of fermi systems such an equation , generally speaking , were not correct . thus , it is the binding constraint @xmath21 , taking place over some region @xmath3 , that makes eq . applicable for fermi systems . because of the binding constraint , fermi quasiparticles in the region @xmath3 can behave as bose one , occupying the same energy level @xmath4 , and eq . yields the quasiparticle distribution function @xmath22 that minimizes the ground - state energy @xmath14 . a possible solution @xmath22 of eq . ( [ fcm ] ) and the corresponding single - particle spectrum @xmath23 are depicted in fig . [ fig0 ] , panel * b*. as seen from the panel * b * , @xmath22 differs from the step function in the interval @xmath3 , where @xmath24 , and coincides with the step function outside this interval . thus , the fermi surface at @xmath25 transforms into the fermi volume at @xmath26 suggesting that the band is absolutely `` flat '' within this interval , giving rise to the spiky dos . the existence of such flat bands formed by inter - particle interaction has been predicted for the first time in ref . quasiparticles with momenta within the interval @xmath27 have the same single - particle energies equal to the chemical potential @xmath2 and form fc , while the distribution @xmath22 describes the new state of the fermi liquid with fc , and the fermi system is split up into to parts : a landau fermi liquid ( lfl ) and the fc part , as it is shown in fig . [ fig0 ] , panel * b * @xcite . in contrast to the landau , marginal , or luttinger fermi liquids , which exhibit the same topological structure of the green s function , in systems with fc , where the fermi surface spreads into the fermi volume , the green s function belongs to a different topological class . the topological class of the fermi liquid is characterized by the invariant @xcite @xmath28 where `` tr '' denotes the trace over the spin indices of the green s function and the integral is taken along an arbitrary contour @xmath29 encircling the singularity of the green s function . the invariant @xmath30 in takes integer values even when the singularity is not of the pole type , can not vary continuously , and is conserved in a transition from the landau fermi liquid to marginal liquids and under small perturbations of the green s function . as shown by volovik @xcite , the situation is quite different for systems with fc , where the invariant @xmath30 becomes a half - integer and the system with fc transforms into an entirely new class of fermi liquids with its own topological structure . in contrast to bose liquid , which entropy @xmath31 at temperature @xmath32 , a fermi liquid with fc possesses finite entropy @xmath33 at zero temperature @xcite . indeed , as it is seen from fig . [ fig0 ] , panel * b * , at @xmath6 , the ground state of a system with a flat band is degenerate , and the occupation numbers @xmath22 of single - particle states belonging to the flat band are continuous functions of momentum @xmath34 , in contrast to discrete standard lfl values 0 and 1 . such behavior of the occupation numbers leads to a @xmath35-independent entropy term @xmath36 with the entropy given by @xmath37.\label{s0}\ ] ] since the state of a system with fc is highly degenerate , fc serves as a stimulator of phase transitions that could lift the degeneracy of the spectrum and make @xmath33 vanish in accordance with the nernst theorem . for instance , fc can excite the formation of spin density waves , antiferromagnetic state and ferromagnetic state etc . , thus strongly stimulating the competition between phase transitions eliminating the degeneracy . the presence of fc facilitates a transition to the superconducting state , because the both phases have the same order parameter @xcite . thus , in contrast to bose systems with bose condensate , entropy of which at lowering temperatures @xmath31 , the @xmath33 peculiarity of fermi systems with fc incites to the emerging of great diversity of states . being generated by the same driving motive - @xmath33 - these , as we shall see , exhibit a universal behavior , and form a new state of matter demonstrated by many compounds . in this paper we briefly review the theory of fc that permits to describe a tremendously broad variety of experimental results in different systems . we assume that these systems are located near the fermion condensation quantum phase transition ( fcqpt ) , leading to the emergence of fc @xcite . the rest of the paper is organized as follows . in section 2 , we examine a scaling behavior of the effective mass and heavy fermion ( hf ) metals based on the extended quasiparticle paradigm that is employed to renovate the landau quasiparticle paradigm . in section 3 , we construct phase diagrams of fermi systems with fc , and compare these with the experimental ones , and show that fc leads to a new state of matter . in section 4 , we apply the fc theory to describe the thermodynamic properties of strongly correlated fermi systems represented by compounds holding a quantum spin liquid . section 5 is devoted to quasicrystals . section 6 summaries the main results . upon using the landau functional , one can obtain the well - known landau equation for the effective mass @xcite @xmath38 here , @xmath39 is the landau interaction . for simplicity , we omit the spin dependencies . to calculate @xmath40 as a function of @xmath35 , we construct the free energy @xmath41 , where @xmath42 is given by . minimizing @xmath39 with respect to @xmath43 , we arrive at the fermi - dirac distribution @xmath44\right\}^{-1}.\label{fl4}\ ] ] here @xmath2 is the chemical potential , @xmath45 is an external magnetic field , and @xmath46 is the bohr magneton . the term @xmath47 entering the right hand side of eq . describes zeeman splitting . is exact , and allows us to calculate the behavior of @xmath40 which now becomes a function of temperature @xmath35 , external magnetic field @xmath45 , number density @xmath15 and pressure @xmath48 , rather than a constant . it is this feature of @xmath40 that forms both the scaling and the non - fermi liquid ( nfl ) behavior observed in measurements on strongly correlated fermi systems @xcite . in case of finite @xmath40 and at @xmath6 the distribution function @xmath49 becomes the theta - function @xmath50 , as it follows from , and eq . ( [ fll ] ) yields the well - known result @xmath51 where @xmath52 , @xmath53 is the @xmath54-wave component of the landau interaction , @xmath55 is the bare mass of particles of the liquid in question , and @xmath56 is the dos of a free fermi gas . because @xmath57 in the landau fermi - liquid theory , the landau interaction can be written as @xmath58 . at a certain critical point @xmath59 the denominator @xmath60 tends to zero , @xmath61 , and we find that @xmath62 where @xmath63 and @xmath64 are constants and @xmath65 . as a result , at @xmath32 and @xmath66 the system undergoes a quantum phase transition @xcite , represented by fcqpt . we note that the divergence of the effective mass given by eq . does preserve the pomeranchuk stability conditions , for @xmath67 is positive @xcite . the behavior of @xmath68 given by ( [ fl7 ] ) is in good agreement with the results of experiments on @xmath69he @xcite . the divergence of the effective mass @xmath68 observed in measurements on two - dimensional @xmath69he @xcite is illustrated in fig . it is seen that calculations based on are in good agreement with the experimental data . [ ! ht ] in two - dimensional @xmath69he as a function of the number density @xmath15 of the liquid , obtained from heat capacity and magnetization measurements . the experimental data are marked by black squares @xcite , and the solid line represents the function given by eq . .,scaledwidth=47.0% ] it is instructive to briefly explore the behavior of @xmath40 in order to capture its universal behavior at fcqpt . let us write the quasiparticle distribution function as @xmath70 , with @xmath43 being the step function , and eq . then becomes @xmath71 at fcqpt , that is at @xmath72 , the effective mass @xmath40 diverges and eq . becomes homogeneous , determining @xmath73 as a universal function of temperature and magnetic field . the only role of @xmath39 is to drive the system to fcqpt and the solutions of eq . are represented by some universal function of variables @xmath35 , @xmath45 , @xmath15 . in that case @xmath40 strongly depends on the same variables . in contrast to the landau quasiparticle paradigm assuming the constancy of the effective mass , the extended quasiparticle paradigm is to be introduced @xcite . the main point here is that the well - defined quasiparticles determine as before the thermodynamic , relaxation and transport properties of strongly correlated fermi - systems , while @xmath40 becomes a function of @xmath35 , @xmath45 , @xmath15 , etc . a deeper insight into the behavior of @xmath73 can be achieved using some `` internal '' scales . namely , near fcqpt the solutions of eq . exhibit a behavior so that @xmath73 reaches its maximum value @xmath74 at some temperature @xmath75 @xcite . it is convenient to introduce the internal scales @xmath74 and @xmath76 to measure the effective mass and temperature . thus , we divide the effective mass @xmath40 and the temperature @xmath35 by the values , @xmath74 and @xmath76 , respectively . this generates the normalized effective mass @xmath77 and the normalized temperature @xmath78 . [ ! ht ] , @xmath79 , versus temperature @xmath35 as a function of magnetic field @xmath45 @xcite shown in the legend . the illustrative values of @xmath80 and @xmath76 at @xmath81 t are shown.,scaledwidth=47.0% ] as illustration to the above consideration , we analyze the specific heat @xmath82 of the hf metal @xmath83 @xcite . under the application of magnetic field @xmath45 , the specific heat exhibits a behavior that is described by a function of both @xmath35 and @xmath45 . [ ! ht ] versus normalized temperature @xmath84 . @xmath85 is extracted from the measurements of the specific heat @xmath79 on @xmath86 in magnetic fields @xmath45 @xcite listed in the legend . constant effective mass @xmath87 inherent in a normal landau fermi liquid ( lfl ) is depicted by the solid line . the transition region is shown by the arrow and the nfl behavior is depicted by two arrows . our calculation based on eq . is displayed by the solid curve.,scaledwidth=47.0% ] as seen from fig . [ fig2 ] , a maximum structure @xmath88 in @xmath89 at temperature @xmath76 appears under the application of magnetic field @xmath45 . @xmath76 shifts to higher @xmath35 and @xmath89 diminishes as @xmath45 is increased . the value @xmath79 is saturated towards lower temperatures , decreasing at elevated magnetic field . to obtain the normalized effective mass @xmath85 , we use @xmath88 and @xmath76 as `` internal '' scales : the maximum structure @xmath88 was used to normalize @xmath79 , and @xmath35 was normalized by @xmath76 . in fig . [ fig3 ] , the obtained @xmath90 , as a function of normalized temperature @xmath78 , is shown by geometrical figures . it is seen that the landau fermi liquid ( lfl ) state and nfl one are separated by the transition regime at which @xmath85 reaches its maximum value . figure [ fig3 ] reveals the scaling behavior of the normalized experimental curves - the curves at different magnetic fields @xmath45 merge into a single one in terms of the normalized variable @xmath91 . our calculations of the normalized effective mass @xmath92 , shown by the solid line and based on eq . , are in good agreement with the experimental facts @xcite . near fcqpt the normalized solution of eq . @xmath92 can be well approximated by a simple universal interpolating function @xcite . the interpolation occurs between the lfl and nfl regimes and represents the universal scaling behavior of @xmath85 @xcite @xmath93 here , @xmath94 , @xmath95 , @xmath96 are fitting parameters ; and the exponent @xmath97 if the landau interaction is an analytical function otherwise @xmath98 @xcite . it follows from eq . that @xmath99 where @xmath63 is a dimensionless factor . several remarks concerning the applicability of eqs . and to systems with violated translational invariance are in order . we study the universal behavior of hf metals , quantum spin liquids , and quasicrystals at low temperatures using the model of a homogeneous hf liquid @xcite . the model is applicable because we consider the scaling behavior exhibited by the thermodynamic properties of these materials at low temperatures , a behavior related to the scaling of quantities such as the effective mass @xmath40 , the heat capacity @xmath82 , the magnetic susceptibility @xmath100 , etc . the behavior of @xmath92 that characterizes them are determined by momentum transfers that are small compared to momenta of the order of the reciprocal lattice length . the high momentum contributions can therefore be ignored by substituting the lattice for the jelly model . while the values of the scales like the maximum @xmath101 of the effective mass , measured at some field @xmath102 , and @xmath103 at which @xmath74 takes place are determined by a wide range of momenta . thus , these scales are controlled by the specific properties of the system under consideration . it follows from our consideration that the scaled thermodynamic properties of different strongly correlated fermi systems can be described by universal function determining @xmath92 . it is seen from fig . [ fig3 ] , and demonstrated by a huge amount of experimental data collected on strongly correlated fermi systems that this observation is in good agreement with experimental facts @xcite . at @xmath6 , a quantum phase transition is driven by a nonthermal control parameter , e.g. the number density @xmath15 , magnetic field @xmath45 , pressure @xmath48 . at the quantum critical point ( qcp ) related to fcqpt and taking place at @xmath59 , the effective mass @xmath40 diverges . we note that there are different kinds of instabilities of normal fermi liquids connected with several perturbations of initial quasiparticle spectrum @xmath104 and occupation numbers @xmath105 , associated with the emergence of a multi - connected fermi surface , see e.g. @xcite . depending on the parameters and analytical properties of the landau interaction , such instabilities lead to several possible types of restructuring of the initial fermi liquid ground state . this restructuring generates topologically distinct phases . one of them is the fc , another one belongs to a class of topological phase transitions , where the sequence of rectangles @xmath106 and @xmath107 is realized at @xmath6 . in fact , at elevated temperatures the systems located at these transitions exhibit behavior typical to those located at fcqpt @xcite . therefore , we do not consider the specific properties of these topological transitions , and focus on the behavior of systems located near fcqpt . beyond fcqpt , the system forms fc that leads to the emergence of a flat band protected by topological invariants @xcite . the @xmath108 schematic phase diagram of the system which is driven to the fc state by variation of the number density @xmath15 is reported in fig . [ fig03 ] . upon approaching the critical density @xmath109 the system remains in the lfl region at sufficiently low temperatures , as it is shown by the shadow area . the temperature range of the shadow area shrinks as the system approaches qcp , and @xmath110 diverges as it follows from eq . . at this qcp shown by the arrow in fig . [ fig03 ] , the system demonstrates the nfl behavior down to the lowest temperatures . beyond the critical point at finite temperatures the behavior remains the nfl , and is determined by the temperature - independent entropy @xmath33 @xcite . in that case at @xmath111 , the system again demonstrates the nfl behavior , and approaches a quantum critical line ( qcl ) ( shown by the vertical arrow and the dashed line in fig . [ fig03 ] ) rather than a quantum critical point . upon reaching the quantum critical line from the above at @xmath32 the system undergoes the first order quantum phase transition , making @xmath33 vanish . as it is seen from fig . [ fig03 ] , at rising temperatures the system located before qcp does not undergo a phase transition , and transits from the nfl behavior to the lfl one . at finite temperatures there is no boundary ( or phase transition ) between the states of the system located before or behind qcp , shown by the arrows . therefore , at elevated temperatures the properties of systems with @xmath112 or with @xmath113 become indistinguishable . [ ! ht ] phase diagram of system with fc . the number density @xmath15 is taken as the control parameter and depicted as @xmath114 . at @xmath113 and sufficiently low temperatures , the system is in the lfl state as shown by the shadow area . this location of the system is depicted by both the solid square and the arrow . the vertical arrow illustrates the system moving in the lfl - nfl direction along @xmath35 at fixed control parameter . at @xmath115 the system is shifted beyond the qcp , and is at the quantum critical line depicted by the dashed line and shown by the vertical arrow . this location of the system is shown by the short dash arrow . at any finite low temperatures @xmath116 the system possesses finite entropy @xmath33 and exhibits the nfl behavior.,scaledwidth=47.0% ] as seen from fig . [ fig03 ] , the location of the system is controlled by the number density @xmath15 . at rising @xmath15 , @xmath113 , the system is located before fcqpt , and demonstrates the lfl behavior at low temperatures . we speculate that such a state can be induced by the application of positive pressure , including positive chemical pressure . on the other hand , at diminishing @xmath15 , @xmath115 , the system is shifted beyond fcqpt , and is at the quantum critical line depicted by the dashed line . in that case the system demonstrates the nfl behavior at any finite temperatures . we assume that such a state can be induced by the application of negative pressure , including negative chemical pressure . at low temperatures and above the critical line , the system has the finite entropy @xmath33 and its nfl state is strongly degenerate . the degeneracy stimulates the emergence of different phase transitions , lifting it and removing the entropy @xmath33 . the nfl state can be captured by other states such as superconducting , for example , by the superconducting state ( sc ) in @xmath117 , or by antiferromagnetic ( af ) state , e.g. the af in @xmath83 @xcite . the diversity of phase transitions occurring at low temperatures is one of the most spectacular features of the physics of many hf metals and strongly correlated compounds . within the scenario of ordinary quantum phase transitions , it is hard to understand why these transitions are so different from one another and their critical temperatures are so extremely small . however , such diversity is endemic to systems with a fc , since the fc state must be altered as @xmath111 , so that the excess entropy @xmath33 is shed before zero temperature is reached . at finite temperatures this takes place by means of some phase transitions which can compete , shedding the excess entropy @xcite . phase diagram of hf liquid with magnetic field as the control parameter . the vertical and horizontal arrows show lfl - nfl and nfl - lfl transitions at fixed @xmath45 and @xmath35 , respectively . at @xmath118 the system is in its nfl state , having a flat band , and demonstrates nfl behavior down to @xmath32 . the hatched area separates the nfl phase and the weakly polarized lfl phase and represents the transition region . the dashed line in the hatched area represents the function @xmath119 given by eq . . the functions @xmath120 and @xmath121 shown by two - headed arrows define the width of the nfl state and the transition area , respectively . the qcp located at the origin and indicated by the arrow denotes the critical point at which the effective mass @xmath40 diverges and both @xmath122 and @xmath123 tend to zero.,scaledwidth=48.0% ] the schematic @xmath124 phase diagram of a hf liquid is depicted in fig . [ fig1 ] , with the magnetic field @xmath45 serving as the control parameter . at @xmath118 , the hf liquid acquires a flat band corresponding to a strongly degenerate state . the nfl regime reigns at elevated temperatures and fixed magnetic field . with increasing @xmath45 , the system is driven from the nfl region to the lfl domain . as shown in fig . [ fig1 ] , the system moves from the nfl regime to the lfl regime along a horizontal arrow , and from the lfl to nfl along a vertical arrow . the magnetic - field - tuned qcp is indicated by the arrow and located at the origin of the phase diagram , since application of any magnetic field destroys the flat band and shifts the system into the lfl state @xcite . the hatched area denoting the transition region separates the nfl state from the weakly polarized lfl state and contains the dashed line tracing the transition region , @xmath119 . referring to eq . , this line is defined by the function @xmath125 , and the width @xmath122 of the nfl state is seen to be proportional @xmath35 . in the same way , it can be shown that the width @xmath123 of the transition region is also proportional to @xmath35 . phase diagram of a hf metal with magnetic field as the control parameter . the af phase boundary line is shown by the arrow and depicted by the solid curve , representing the nel temperature @xmath126 . the vertical and horizontal arrows show lfl - nfl and nfl - lfl transitions at fixed @xmath45 and @xmath35 , respectively . the hatched area separates the nfl phase and the weakly polarized lfl phase and represents the transition area . the solid curve in the hatched area represents the transition temperature @xmath127 . the solid short dash line @xmath128 depicts the transition temperature provided that the af state were absent . the functions @xmath120 shown by two - headed arrows defines the total width of both the nfl state and the transition area . the inset shows a schematic plot of the normalized effective mass versus the normalized temperature . the transition regime , where @xmath85 reaches its maximum value at @xmath129 , is shown as the hatched area in both the main panel and the inset . arrows indicate the transition region and the inflection point @xmath130 in the @xmath85 plot.,scaledwidth=48.0% ] [ ! ht ] phase diagram for @xmath83 . solid circles represent the boundary between af and nfl states . the curve is represented by the function @xmath131 @xcite . the diamonds along the short - dash curve denote the transition area between the nfl and lfl regime @xcite . the short - dash line is approximated by @xmath132 , see fig . [ fig04 ] . diamonds marking @xmath127 along this dash line signify the maxima @xmath103 of @xmath79 @xcite shown in fig . the transition temperature @xmath127 as determined from magnetostriction ( solid squares ) , longitudinal magnetoresistivity ( triangles ) , and susceptibility ( solid circles ) @xcite . the solid line depicts the linear fit to @xmath127.,scaledwidth=47.0% ] now we construct the @xmath124 schematic phase diagram of a hf metal like @xmath83 shown in fig . [ fig04 ] @xcite . we start with observing that eq . reveals the scaling behavior of the normalized effective mass @xmath92 : values of the effective mass @xmath73 at different magnetic fields @xmath45 merge into a single mass value @xmath85 in terms of the normalized variable @xmath133 . the inset in fig . [ fig04 ] demonstrates that scaling behavior of the normalized effective mass @xmath85 versus the normalized temperature @xmath84 . the lfl phase prevails at @xmath134 , followed by the @xmath135 regime at @xmath136 , with @xmath137 or @xmath138 as it follows from eq . . the latter phase is designated as nfl due to the strong dependence of the effective mass on temperature . the temperature region @xmath139 encompasses the transition between the lfl regime with almost constant effective mass and the nfl behavior . thus @xmath140 identifies the transition region featuring a crossover between lfl and nfl regimes . the transition ( crossover ) temperature @xmath141 is not actually the temperature of a phase transition . its specification is necessarily ambiguous , depending as it does on the criteria invoked for determination of the crossover point . as usual , the temperature @xmath141 is extracted from the field dependence of charge transport , for example from the resistivity @xmath142 given by @xmath143 where @xmath144 is the residual resistivity and @xmath145 is a @xmath35-independent coefficient . the term @xmath144 is ordinarily attributed to impurity scattering . the lfl state is characterized by the @xmath146 dependence of the resistivity with index @xmath147 . the crossover ( through the transition regime shown as the hatched area in both fig . [ fig04 ] and its inset ) takes place at temperatures where the resistance starts to deviate from lfl behavior , with the exponent @xmath148 shifting from 2 into the range @xmath149 . when constructing the phase diagram depicted in fig . [ fig04 ] , we assume that af order wins the competition , destroying the @xmath33 term at low temperatures . at @xmath150 , the hf liquid acquires a flat band corresponding to a strongly degenerate state . here , @xmath151 is a critical magnetic field , such that at @xmath32 the application of magnetic field @xmath152 destroys the af state restoring the paramagnetic state with the lfl behavior . in some cases as in the hf metal @xmath153 , @xmath154 , see e.g. @xcite , while in @xmath83 , @xmath155 t @xcite . obviously , @xmath151 is defined by the specific system properties , therefore we consider it as a parameter . the nfl regime reigns at elevated temperatures and fixed magnetic field . with increasing @xmath45 , the system is driven from the nfl region to the lfl domain . as shown in figs . [ fig1 ] and [ fig04 ] , the system moves from the nfl regime to the lfl regime along the corresponding arrows . the magnetic - field - tuned qcp is indicated by the arrow and located at @xmath150 . the hatched area denotes the transition region , and separates the nfl state from the weakly polarized lfl state . it contains both the dashed line tracing @xmath156 and the solid curve @xmath141 . referring to eq . , the latter is defined by the function @xmath157 and merges with @xmath156 at relatively high temperatures , and @xmath158 at lower @xmath159 , with @xmath160 being the nel temperature . as seen from eq . , both the width @xmath122 of the nfl state and the width of the transition region are proportional @xmath35 @xcite . the af phase boundary line is shown by the arrow and depicted by the solid curve . as it was mentioned above , the dashed line @xmath161 represents the transition temperature provided that the af state were absent . in that case the fc state is destroyed by any weak magnetic field @xmath162 at @xmath32 and the dashed line @xmath163 crosses the origin of coordinates , as it is displayed in figs . [ fig1 ] and [ fig04 ] . at @xmath164 the transition temperature @xmath156 coincides with @xmath141 shown by the solid curve , since the properties of the system are given by its local free energy , describing the paramagnetic state of the system . one might say that the system `` does not remember '' the af state , emerging at lower temperatures . this observation is in good agreement with experimental facts collected on the hf metal @xmath83 . these experimental facts are summarized in the phase diagrams of fig . [ f02 ] : at relatively high temperatures @xmath165 the transition temperature @xmath127 , obtained in measurements on @xmath83 @xcite , is well approximated by the straight lines representing @xmath163 . it is seen from fig . [ f02 ] , that the slope of the short dash line ( representing the maxima of the specific heat @xmath79 ) is different from the slope of the solid line ( representing maxima of the susceptibility @xmath166 ) . such a behavior is determined by the fact that the maxima of @xmath79 and @xmath166 are given by the different relations , determining the inflection points of the entropy : @xmath167 and @xmath168 , correspondingly . the theory of fc shows that the inflection points exist , provided that the system is located near fcqpt @xcite . [ ! ht ] phase diagrams for the three hf metals : @xmath83 ( a ) ; @xmath169 ( b ) ; @xmath170 ( c ) . in panels a and b , the af phase boundaries @xcite are shown by the solid lines . in panel b , the diamonds depict the measurements @xcite . in panel c , the phase boundary of possible phase transition is shown by short dot curve . pentagons correspond to the measurements of @xmath141 on the hf metals extracted from the analysis of @xmath171 as a function @xmath45 @xcite . the solid straight lines depict the transition temperature @xmath156 . dash lines schematically denote the boundary of the nfl and lfl regime.,title="fig:",scaledwidth=42.0% ] phase diagrams for the three hf metals : @xmath83 ( a ) ; @xmath169 ( b ) ; @xmath170 ( c ) . in panels a and b , the af phase boundaries @xcite are shown by the solid lines . in panel b , the diamonds depict the measurements @xcite . in panel c , the phase boundary of possible phase transition is shown by short dot curve . pentagons correspond to the measurements of @xmath141 on the hf metals extracted from the analysis of @xmath171 as a function @xmath45 @xcite . the solid straight lines depict the transition temperature @xmath156 . dash lines schematically denote the boundary of the nfl and lfl regime.,title="fig:",scaledwidth=42.0% ] phase diagrams for the three hf metals : @xmath83 ( a ) ; @xmath169 ( b ) ; @xmath170 ( c ) . in panels a and b , the af phase boundaries @xcite are shown by the solid lines . in panel b , the diamonds depict the measurements @xcite . in panel c , the phase boundary of possible phase transition is shown by short dot curve . pentagons correspond to the measurements of @xmath141 on the hf metals extracted from the analysis of @xmath171 as a function @xmath45 @xcite . the solid straight lines depict the transition temperature @xmath156 . dash lines schematically denote the boundary of the nfl and lfl regime.,title="fig:",scaledwidth=42.0% ] panels ( a , b , c ) of fig . [ ff4 ] are focused on the behavior of the transition temperature @xmath141 extracted from measurements of kinks in @xmath172 as a function of @xmath45 @xcite , where @xmath173 is the magnetization . positions of the kinks is shown by pentagons in fig . it is seen from fig . [ ff4 ] , panel a , that at @xmath164 , the transition temperature @xmath127 of @xmath86 is well approximated by the line @xmath163 . as mentioned above , upon using nonthermal tuning parameters like the number density @xmath15 , the nfl behavior can be destroyed and the lfl one will be restored . in our simple model , the application of positive pressure @xmath48 makes @xmath15 grow shifting the system from fcqpt to the lfl state , as it is shown in fig . [ fig03 ] , so that the electronic system of @xmath83 moves into the shadow area characterized by the lfl behavior at low temperatures . the new location of the system , represented by @xmath174 , is shown by the arrow pointing at the solid square . we note that the positive chemical pressure in the considered case is induced by @xmath175 substitution @xcite . as a result , the application of magnetic field @xmath176 does not drive the system to its fcqpt with the divergent effective mass because the qcp is already destroyed by the positive pressure , as it is shown in fig . [ ff4 ] , panel b. here @xmath151 is the critical magnetic field that eliminates the corresponding af order . at @xmath177 and raising temperatures , the system , moving along the vertical arrow , transits from the lfl regime to the nfl one . at relatively high temperatures both @xmath83 and @xmath174 are in their paramagnetic state . as a result , in that case @xmath127 is well approximated by the straight line @xmath163 . this observation is in accordance with experimental facts @xcite displayed in fig . [ ff4 ] , panels a , b . the system located above qcl exhibits the nfl behavior down to lowest temperatures unless it is captured by a phase transition . the behavior exhibiting by the system located above qcl is in accordance with the experimental observations that study the evolution of qcp in @xmath83 under the application of negative chemical pressure induced with @xmath178 substitution @xcite . we assume a simple model that the application of negative pressure reduces @xmath15 and the electronic system of @xmath83 moves from qcp to a new position over qcl shown by the dash arrow in fig . [ fig03 ] . thus , the electronic system of @xmath170 is located at qcl and possesses a flat band , while the entropy includes @xmath33 . we predict that at lowering temperatures , the electronic system of @xmath170 is captured by a phase transition , since the nfl state above qcl is strongly degenerate and the term @xmath33 should be eliminated . at diminishing temperatures , this degeneracy is to be removed by some phase transition which likely can be detected by the lfl state accompanying it . the tentative boundary line of that transition is shown by the short dot line in fig . [ ff4 ] , panel c. it is also seen from fig . [ ff4 ] , panel c , that at elevated temperatures @xmath127 is well approximated by the function @xmath163 . thus , at relatively high temperatures the transition temperature @xmath156 , shown in fig . [ ff4 ] , panels a , b , and c by the solid lines , coincides with @xmath141 depicted by the pentagons . for , as it was discussed above , the local properties of the systems in question are given by their local free energy , formed by the nfl region related with fc , as it is displayed in fig . [ fig03 ] . to conclude this section , we have for the first time theoretically carried out a systematic study of the phase diagrams of strongly correlated fermi systems , including hf metals like @xmath86 , and considered the evolution of these diagrams in case of the application of negative / positive pressure . we have observed that at sufficiently high temperatures outside the af phase the transition temperature @xmath141 follows an almost linear b - dependence , and coincides with @xmath156 , induced by the presence of fc . the obtained results are in good agreement with experimental facts @xcite . as we shall show in sections 4 and 5 , the phase diagram [ fig1 ] of hf liquid describes the corresponding phase diagrams of quantum spin liquids and quasicrystals as well , while the observed scaling behavior is described by eq . thus , all these strongly correlated compounds exhibit the uniform behavior , and allow us to view that behavior as representing the main characteristic of the new state of matter . the first experimental observation of quantum spin liquid ( qsl ) supporting exotic spin excitations - spinons - and carrying fractional quantum numbers in the herbertsmithite @xmath179 is reported in ref . qsl can be viewed as an exotic quantum state composed of hypothetic particles such as fermionic spinons which carry spin @xmath180 and no charge . the herbertsmithite @xmath181 has been exposed as a @xmath182 heisenberg antiferromagnet on a perfect kagome lattice , see ref . @xcite for a recent review . in @xmath181 , the @xmath183 ions with @xmath182 form the triangular kagome lattice , and are separated by nonmagnetic intermediate layers of @xmath184 and @xmath185 atoms . the planes of the @xmath183 ions can be considered as two - dimensional ( 2d ) layers with negligible magnetic interactions along the third dimension . a simple kagome lattice has a dispersionless topologically protected branch of the spectrum with zero excitation energy that is the flat band @xcite . in that case fcqpt can form a strongly correlated quantum spin liquid ( scqsl ) composed of fermions with zero charge , @xmath182 , and the effective mass @xmath186 , occupying the corresponding fermi sphere with the fermi momentum @xmath11 . a comparison of the qsl specific heat @xmath187 extracted from measurements on @xmath181 with @xmath79 of @xmath86 @xcite is shown in fig . the striking feature of the specific heat behavior is the strong dependence on the magnetic field seen from the figure . it is seen that both @xmath188 and @xmath79 exhibit the same qualitative behavior that allows us to view the herbertsmithite as insulator with qsl , and qsl itself as scqsl @xcite . [ ! ht ] of qsl is extracted from measurements of the specific heat on @xmath181 at different magnetic fields shown in the legend @xcite . right panel reports the @xmath35-dependence of the electronic specific heat @xmath79 of @xmath86 at different magnetic fields @xcite as shown in the legend.,scaledwidth=47.0% ] the schematic @xmath124 phase diagram of @xmath181 is reported in fig . [ fig1 ] . at @xmath6 and @xmath118 the system is near fcqpt without tuning . it can also be shifted from fcqpt by the application of magnetic field @xmath45 . magnetic field @xmath45 and temperature @xmath35 play the role of the control parameters , driving it from the nfl to lfl regions as shown by the vertical and horizontal arrows . at fixed @xmath45 and increasing @xmath35 the system transits along the vertical arrow from the lfl region to nfl one crossing the transition region . on the contrary , at fixed @xmath35 increasing @xmath45 drives the system along the horizontal arrow from the nfl region to lfl one . the inset to the phase diagram shown in fig . [ fig04 ] demonstrates the universal behavior of the normalized effective mass @xmath85 versus normalized temperature @xmath84 as given by eq . . it follows from eq . , and is seen from fig . [ fig1 ] , that the both width @xmath189 of the nfl and the width @xmath190 of the transition region , shown by the arrows in fig . [ fig1 ] , tend to zero at diminishing @xmath35 and @xmath45 since @xmath191 . [ ! ht ] extracted from measurements of susceptibility @xmath192 @xcite on @xmath181 and @xmath193 on @xmath83 @xcite . our calculations of @xmath85 are shown by solid curve.,scaledwidth=47.0% ] from fig . [ fig5 ] it is clearly seen that the data collected on both @xmath181 @xcite and @xmath83 @xcite merge into the same curve , obeying the scaling behavior . this demonstrates that the spin liquid of the herbertsmithite is close to fcqpt and behaves like hf liquid of @xmath83 in magnetic fields . figure [ t1 ] displays the normalized spin - lattice relaxation rates @xmath194 at fixed temperature versus normalized magnetic field @xmath195 . it is seen from fig . [ t1 ] , that the magnetic field @xmath45 progressively reduces @xmath196 , and the spin - lattice relaxation rate as a function of @xmath45 possesses an inflection point at some @xmath197 shown by the arrow . to clarify the universal scaling behavior of qsl in the herbertsmithite and in hf metals , we normalize both the function @xmath196 and the magnetic field . namely , we normalize @xmath198 by its value at the inflection point , and magnetic field is normalized by @xmath199 , @xmath200 . since @xmath201 @xcite , we expect that different strongly correlated fermi systems located near fcqpt exhibit the same behavior of the normalized spin - lattice relaxation rate . it is seen from fig . [ t1 ] , that both the herbertsmithite @xmath181 @xcite and hf metal @xmath202 @xcite demonstrate similar behavior of the normalized spin - lattice relaxation rate . as seen from fig . [ t1 ] , at @xmath203 ( or @xmath204 ) , when the system in its nfl region , the normalized relaxation rate @xmath194 depends weakly on the magnetic field , while at higher fields , as soon as the system enters the lfl region , @xmath194 diminishes in agreement with both eq . and phase diagram [ fig1 ] , @xmath205 thus , in accordance with the phase diagram shown in fig . [ fig1 ] and eq . , we conclude that the application of magnetic field @xmath45 leads to crossover from the nfl to lfl behavior and to the significant reduction in the relaxation rate . [ ! ht ] at fixed temperature as a function of magnetic field : solid squares correspond to data on @xmath194 extracted from measurements on @xmath181 @xcite , while the solid triangles correspond to those extracted from measurements on @xmath202 with @xmath206 @xcite . the inflection point , representing the transition region , where the normalization is taken is shown by the arrow . our calculations based on eqs . and are depicted by the solid curve , tracing the scaling behavior of @xmath207 and representing the @xmath45-dependence of the thermal resistivity @xmath208 , see main text and eq . .,scaledwidth=47.0% ] as it was mentioned above , qsl plays a role of hf liquid framed into the insulating compound . thus , we expect that qsl in the herbertsmithite behaves like the electronic liquid in hf metals if the charge of an electron were zero . in that case , the thermal resistivity @xmath208 of qsl is customarily related to the thermal conductivity @xmath209 @xmath210 the resistivity @xmath208 behaves like the electrical resistivity @xmath211 of the electronic liquid , since @xmath212 represents the contribution of spinon - spinon scattering to the thermal transport , being analogous to the contribution @xmath213 to the charge transport by electron - electron scattering . here , @xmath214 is the lorenz number , @xmath144 and @xmath215 are residual resistivity of electronic liquid and qsl , respectively , and the coefficients @xmath216 and @xmath217 @xcite . thus , in the lfl region the coefficient @xmath212 of the thermal resistivity of qsl under the application of magnetic fields at fixed temperature behaves like the spin - lattice relaxation rate shown in fig . [ t1 ] , @xmath218 , while in the lfl region at fixed magnetic fields the thermal conductivity @xmath209 is a linear function of temperature , @xmath219 . study of the thermal resistivity @xmath208 given by eq . allows one to reveal spinons as itinerant excitations . it is important that @xmath208 is not contaminated by contributions coming from localized excitations by impurity effects . the temperature dependence of thermal resistivity @xmath208 represented by the finite term @xmath215 directly shows that the behavior of qsl is similar to that of metals , and there is a finite residual term @xmath220 in the zero - temperature limit of @xmath209 . the presence of this term immediately proves that there are gapless excitation associated with the property of normal metals , in which gapless electrons govern the heat transport . the finite @xmath215 means that in qsl both @xmath221 and @xmath222 remain nonzero at @xmath32 . therefore , gapless spinons , forming the fermi surface , govern the specific heat and the transport . key information on the nature of spinons is further provided by the @xmath45-dependence of the coefficient @xmath212 . the specific @xmath45-dependence of the resistivity @xmath223 , shown in fig . [ t1 ] and given by eq . , would establish the behavior of qsl as scqsl . we note that the heat transport is polluted by the phonon contribution . on the other hand , the phonon contribution is hardly influenced by the magnetic field @xmath45 . therefore , we expect the @xmath45-dependence of the heat conductivity to be governed by @xmath224 . consider the approximate relation , @xmath225 where the coefficient @xmath226 is @xmath45-independent . to derive , we employ eq . , and obtain @xmath227 here , the term @xmath228 describes the phonon contribution to the heat transport . upon carrying out simple algebra and assuming that @xmath229\ll 1 $ ] , we arrive at eq . . it is seen from fig . [ t1 ] , that the effective mass @xmath230 is a diminishing function of magnetic field @xmath45 . then , it follows from eqs . and that the function @xmath231/\kappa(0,t)$ ] increases at elevated field @xmath45 . ht ] dependence of the thermal conductivity @xmath232 measured on the organic insulator @xmath233_{2}$ ] and standardized by the zero field value @xmath209 , @xmath234/\kappa(b=0,t)$ ] at temperatures shown in the legend @xcite . our calculations are based on eq . and shown by pentagons and stars.,scaledwidth=47.0% ] [ ! ht ] dependence of the thermal conductivity @xmath232 measured on the organic insulator @xmath235 and standardized by the zero field value @xmath209 , @xmath234/\kappa(b=0,t)$ ] at temperatures shown in the legend @xcite . our calculations are based on eq . and shown by geometrical figures as it is displayed in the legend . panel * b * : normalized thermal conductivity @xmath236 as function of the normalized magnetic field @xmath195 . the inflection point is shown by the arrow . our calculations based on eq . are depicted by the solid curve.,title="fig:",scaledwidth=47.0% ] dependence of the thermal conductivity @xmath232 measured on the organic insulator @xmath235 and standardized by the zero field value @xmath209 , @xmath234/\kappa(b=0,t)$ ] at temperatures shown in the legend @xcite . our calculations are based on eq . and shown by geometrical figures as it is displayed in the legend . panel * b * : normalized thermal conductivity @xmath236 as function of the normalized magnetic field @xmath195 . the inflection point is shown by the arrow . our calculations based on eq . are depicted by the solid curve.,title="fig:",scaledwidth=47.0% ] recent measurements of @xmath237 on the organic insulators @xmath238_{2}$ ] and @xmath235 @xcite are displayed in figs . [ kappa ] and [ kappa1 ] . the measurements show that the heat is carried by phonons and qsl , for the heat conductivity is well fitted by @xmath239 , where @xmath63 and @xmath64 are constants . the finite @xmath63 term implies that spinon excitations are gapless in @xmath233_{2}$ ] , while in @xmath240 gapless excitations are under debate @xcite . a simple estimation indicates that the ballistic propagation of spinons seems to be realized in the case of @xmath233_{2}$ ] @xcite . it is seen from figs . [ kappa ] and [ kappa1 ] , panel * a * , that the normalized data @xmath234/\kappa(b=0,t)$ ] demonstrate a strong @xmath45-dependence , namely the field dependence shows an increase of thermal conductivity for rising fields @xmath45 . such a behavior is in agreement with eq . and fig . [ t1 ] which demonstrate that @xmath241 is diminishing function of @xmath45 . as a result , it follows from eq . that @xmath232 is an increasing function of @xmath45 . our calculations based on eqs . and are depicted by geometrical figures in figs . [ kappa ] and [ kappa1 ] , panel * a*. since we can not calculate the parameter @xmath226 entering eq . we use it as a fitting parameter . temperature @xmath35 was also used to fit the data at temperatures shown in the legend in figs . [ kappa ] and [ kappa1 ] . it is seen from fig . [ kappa1 ] , panel * a * , that @xmath232 as a function of @xmath45 possesses an inflection point at some @xmath197 . to reveal the scaling behavior of qsl in @xmath235 , we normalize both the function @xmath232 and the magnetic field by their values at the inflection points , as it was done in the case of @xmath198 , see fig . [ t1 ] . in that case we get rid of the factor @xmath226 , entering eq . , and our calculations do not have any fitting parameters . it is seen from fig . [ kappa1 ] , panel * b * , that the normalized @xmath236 exhibits the scaling behavior and becomes a function of a single variable @xmath195 . our calculations show that @xmath236 extracted from measurements on @xmath238_{2}$ ] exhibits the scaling behavior as well . it is seen from both figs . [ kappa ] and [ kappa1 ] , that our calculations are in good overall agreement with the experimental facts and there is no need to suppose the existence of additional magnetic excitations activated by the application of magnetic field in order to explain the growth of @xmath232 at elevated @xmath45 @xcite . important signature of electron liquid in hf metals is excitations quasiparticles carrying electron quantum numbers and characterized by the effective mass @xmath186 . neutron scattering measurements of the dynamic spin susceptibility @xmath242 as a function of momentum @xmath243 , frequency @xmath244 and and temperature @xmath35 play important role when identifying the properties of quasiparticles . ht ] plotted against the dimensionless variable @xmath245 . the data are extracted from measurements on @xmath246 @xcite . panel * b * : @xmath247 versus @xmath245 . the data are extracted from measurements on @xmath181 @xcite . our calculations are shown by the solid curves.,title="fig:",scaledwidth=47.0% ] plotted against the dimensionless variable @xmath245 . the data are extracted from measurements on @xmath246 @xcite . panel * b * : @xmath247 versus @xmath245 . the data are extracted from measurements on @xmath181 @xcite . our calculations are shown by the solid curves.,title="fig:",scaledwidth=47.0% ] at low temperatures , such measurements allow one to reveal that quasiparticles excitations form a continuum , and populate an approximately flat band crossing the fermi level . in that case the normalized susceptibility @xmath248 exhibits a scaling behavior as a function of the dimensionless variable @xmath245 @xcite . the panel * a * of fig . [ fig6 ] reports @xmath248 extracted from measurements of the inelastic neutron scattering spectrum on the hf metal @xmath246 @xcite . the data @xmath248 exhibit the scaling behavior over three orders of the variation of both the function and the variable . the scaled data obtained in measurements on a such quite different strongly correlated system as @xmath181 @xcite are shown in the panel * b*. it is seen that our calculations shown by the solid curves are in good agreement with the experimental facts collected both on @xmath249 and @xmath179 over almost three orders of the scaled variables @xcite . thus , the spin excitations in @xmath181 exhibit the same behavior as electron excitations of the hf metal @xmath249 , and , therefore form a continuum . this observation of the continuum is of great importance since it clearly reveals the presence of scqsl in the herbertsmithite , as it was later confirmed by direct experimental observation @xcite . in this section , we have considered the non - fermi liquid behavior and the scaling one of such strongly correlated fermi systems as insulators @xmath181 , @xmath238_{2}$ ] , @xmath235 , and hf metals @xmath249 , @xmath250 , and @xmath83 . we have shown that these are described within the frame of the theory of fc . our calculations are in a good agreement with the experimental data , and allow us to identify the low - temperature behavior of @xmath179 , @xmath235 , and @xmath238_{2}$ ] as determined by scqsl . the same behavior is observed in the heavy fermion metals . thus , @xmath179 , @xmath240 , and @xmath238_{2}$ ] can be viewed as a new type of strongly correlated electrical insulator that possesses properties of heavy - fermion metals with one exception : it resists the flow of electric charge , supporting the spin current formed by spinons . new materials named quasicrystals ( qcs ) and characterized by noncrystallographic rotational symmetry and quasiperiodic translational properties have attracted scrutiny @xcite . study of quasicrystals may shed light on the most basic notions related to the quantum critical state observed in hf metals . experimental measurements on the gold - aluminium - ytterbium quasicrystal @xmath251 have revealed a quantum critical behavior with the unusual exponent @xmath252 defining the divergency of the magnetic susceptibility @xmath253 at @xmath32 @xcite . the measurements have also exposed that the observed nfl behavior transforms into the lfl one under the application of a tiny magnetic field @xmath45 . all these facts challenge theory to explain a quantum criticality of the gold - aluminum - ytterbium qc . in case of qcs electrons occupy a new class of states denoted as `` critical states '' , neither being extended nor localized . associated with these critical states , characterized by an extremely degenerate confined wave function , are the so - called `` spiky '' dos @xcite . these predicted dos are corroborated by experiments revealing that single spectra of the local dos demonstrate the spiky dos @xcite , which form flat bands @xcite . as a result , we assume that the electronic system of some quasicrystals is located at fcqpt without tuning . we now investigate the behavior of @xmath192 as a function of temperature at fixed magnetic fields . the effective mass @xmath73 can be measured in experiments , for @xmath254 , where @xmath192 is the ac or dc magnetic susceptibility . if the corresponding measurements are carried out at fixed magnetic field @xmath45 then , as it follows from eq . , @xmath192 reaches the maximum @xmath255 at some temperature @xmath76 . upon normalizing both @xmath192 and the specific heat @xmath79 by their peak values at each field @xmath45 and the corresponding temperatures by @xmath76 , we observe from eq . that all the curves are to merge into a single one , thus demonstrating a scaling behavior typical for hf metals @xcite . as seen from fig . [ fig7_1 ] , @xmath256 extracted from measurements on @xmath257 @xcite shows the scaling behavior given by eq . and agrees well with our calculations shown by the solid curve over four orders of magnitude in the normalized temperature . it is also seen , that @xmath256 agrees well with the normalized @xmath258 extracted from measurements in magnetic fields on @xmath83 @xcite . at @xmath259 the susceptibility @xmath256 exhibits the @xmath135 regime , @xmath260 , with @xmath138 . [ ! ht ] and the normalized magnetic susceptibility @xmath256 extracted from measurements in magnetic fields @xmath45 on @xmath86 @xcite and on @xmath257 @xcite , respectively . the corresponding magnetic fields are listed in the legends . the arrow shows the @xmath261 regime taking place at @xmath259 . our calculations are depicted by the solid curve tracing the scaling behavior of @xmath262 given by eq . .,scaledwidth=47.0% ] the schematic phase diagram of the gold - aluminum - ytterbium qc @xmath251 is reported in fig . the magnetic field @xmath45 plays the role of the control parameter , driving the system outwards fcqpt that occurs at @xmath118 and @xmath6 without tuning , since the qc critical state is formed by singular density of states @xcite . it follows from eq . and is seen from fig . [ fig1 ] , that at fixed temperatures the increase of @xmath45 drives the system along the horizontal arrow from nfl state to lfl one . on the contrary , at fixed magnetic field and increasing temperatures the system transits along the vertical arrow from lfl state to nfl one . the @xmath135 regime with @xmath138 is marked as nfl since contrary to the lfl case , where the effective mass is constant , the effective mass depends strongly on temperature . thus , the temperatures @xmath263 , shown by the arrow in fig . [ fig1 ] , can be regarded as the transition regime between lfl and nfl states . it is seen , that the common width of the lfl region @xmath189 and the transition one @xmath190 is proportional to @xmath35 . these theoretical results are in good agreement with the experimental facts @xcite . [ ! ht ] versus normalized temperature @xmath84 extracted from measurements of the magnetic susceptibility @xmath192 on the quasicrystal @xmath264 at different magnetic fields @xcite shown in the legend . the lfl region and nfl one are shown by the solid and dashed arrows , respectively . the crossover region is depickted by two arrows . the solid line marked by the nfl displays @xmath265 behavior.,scaledwidth=47.0% ] in order to validate the phase diagram displayed in fig . [ fig1 ] , we focus on the lfl , nfl and the transition lfl - nfl regions exhibited by the qc . to this end , we display in fig . [ fig7 ] the normalized @xmath256 on the double logarithm scale . as seen from fig . [ fig7 ] , @xmath256 extracted from the measurements is not a constant , as would be for a lfl . the two regions ( the lfl region and nfl one ) , separated by the transition region , as depicted by the hatched area in fig . [ fig1 ] , are clearly seen in fig . [ fig7 ] illuminating good agreement between the theory and measurements . the straight lines in fig . [ fig7 ] outline both the lfl and nfl behaviors of @xmath266 and @xmath267 . these straight lines in fig . [ fig7 ] are in good agreement with the behavior of @xmath85 given by eq . with @xmath98 . it is also seen from fig . [ fig7 ] that a tiny magnetic field of @xmath268 t destroys the nfl behavior hereby driving the system to the lfl region @xcite . it is seen that our calculations shown by the solid curve are in good agreement with the experimental facts over four orders of magnitude in the normalized temperature . as a result , we conclude that @xmath257 quasicrystal exhibits the typical scaling behavior of its thermodynamic properties , and belongs to the famous family of hf metals , while the quantum critical physics of the quasicrystal is universal , and emerges regardless of the underlying microscopic details of the quasicrystal . the condensate state in bose interacting systems , introduced by s. t. belyaev in his famous papers , forms the system s properties . it turns out that the fermion condensate takes place in many compounds , and generates the non - fermi liquid behavior by forming flat bands . we have for the first time theoretically carried out a systematic study of the phase diagrams of strongly correlated fermi systems , including hf metals , insulators with strongly correlated quantum spin liquid , and quasicrystals , and have demonstrated that these diagram have universal features . the obtained results are in good agreement with experimental facts . we have shown both analytically and using arguments based entirely on the experimental grounds that the data collected on very different heavy - fermion compounds , such as hf metals , compounds with quantum spin liquid and quasicrystals , have a universal scaling behavior , and materials with strongly correlated fermions can unexpectedly have a uniform behavior in spite of their microscopic diversity . thus , the quantum critical physics of different heavy - fermion compounds is universal , and emerges regardless of the underlying microscopic details of the compounds . this uniform behavior , induced by the universal quantum critical physics , allows us to view it as the main characteristic of the new state of matter . our analysis of strongly correlated systems is in the context of salient experimental results , and our calculations of the non - fermi liquid behavior are in good agreement with a broad variety of experimental facts . our theoretical analysis of numerous experimental facts shows that the theory of fermion condensation develops completely good description of the nfl behavior of strongly correlated fermi systems . moreover , the fermion condensate can be considered as the universal reason for the nfl behavior observed in various hf metals , liquids , insulators with quantum spin liquids , and quasicrystals . this paper is written as invited review dedicated to 90th anniversary of s. t. belyaev birthday . kgp acknowledges funding from the ural branch of the russian academy of sciences , basic research program no . 12-u-1 - 1010 , the presidium of the russian academy of sciences , program 12-p1 - 1014 , and rtp urb ras project p3 .

we determine the jet vertex for mueller - navelet jets and forward jets in the small - cone approximation for two particular choices of jet algoritms : the kt algorithm and the cone algorithm . these choices are motivated by the extensive use of such algorithms in the phenomenology of jets . the differences with the original calculations of the small - cone jet vertex by ivanov and papa , which is found to be equivalent to a formerly algorithm proposed by furman , are shown at both analytic and numerical level , and turn out to be sizeable . a detailed numerical study of the error introduced by the small - cone approximation is also presented , for various observables of phenomenological interest . for values of the jet `` radius '' , the use of the small - cone approximation amounts to an error of about 5% at the level of cross section , while it reduces to less than 2% for ratios of distributions such as those involved in the measure of the azimuthal decorrelation of dijets .

1501.07442

the identification of high - energy ( regge ) dynamics in qcd processes has been long since theoretically investigated . with the advent of high - energy colliders like hera , tevatron and lhc such studies have become possible at experimental level too . the processes which are expected to be more sensitive to this peculiar dynamical regime , where the center - of - mass energy @xmath1 is much larger than all hard scales involved in the scattering , are the so - called _ mueller - navelet _ ( mn ) _ jets _ at hadron - hadron colliders @xcite and _ forward jets _ at electron - hadron colliders @xcite . both of them are defined by the presence of qcd jets at large rapidity , accompanied by any hadronic activity which is inclusively collected in the central region . such processes can be theoretically described by factorization formulae which involve several ingredients : the partonic distribution functions ( pdfs ) of the incoming hadron(s ) , the gluon green s function ( ggf ) describing the high - energy dynamics of emitted and exchanged partons mostly ( reggeized ) gluons and finally the so - called jet vertices , describing the production of a forward jet from the interaction of one incoming parton and a reggeized gluon . in the case of an incoming electron , an additional quantity , the photon impact factor , has to be considered too . at present , all ingredients are known at next - to - leading level in the respective parameters : the pdfs which resum logarithms of collinear type , the ggf resumming logarithms of the energy , and the jet vertices ( and impact - factors ) which are computed at finite perturbative order . focusing for definiteness on mn jets in hadron - hadron collision , the process - dependent part of the cross section is represented by the jet vertex @xcite , which depends on the jet variables and the actual jet algorithm . by following the work of ivanov and papa @xcite , in this paper we reconsider the computation of the jet vertex in the small - cone approximation ( sca ) , namely for jets whose extension in the rapidity - azimuthal angle @xmath2-plane is small . in particular we apply their method to determine the analytic expressions of the jet vertex for two particular choices of jet algorithms : the _ kt algorithm _ @xcite and the _ cone algorithm _ @xcite . these are the mostly used algorithms in modern jet phenomenology , in particular the kt one . on the contrary , the algorithm used in @xcite can be traced back to the one considered by furman @xcite in early studies of qcd radiation , but not used for practical purposes anymore , being infra - red unsafe . the aim of our work is twofold : on one hand we want to give a precise estimate of the error introduced by the small - cone approximation in the description of qcd observables at high energies , i.e. , at large rapidities , so as to possibly justify its use in phenomenological analyses . on the other hand , we want to estimate the differences occurring by choosing different jet algorithms for the same process . the jet algorithm dependence has already been studied in the past , and a detailed analysis of the cone and kt algorithms in the sca was presented in ref . @xcite in the context of collinear factorization . here we carry out a similar analysis in the framework of high - energy ( kt - dependent ) factorization , the basic tool for the description of the regge regime in perturbative qcd . one should keep in mind that the small - cone expressions are fully analytic ( before their convolution with the pdfs ) and compact , and allow a simple implementation in numerical codes that run much faster than those with the exact jet vertices . for this reason the sca jet vertices have already been used@xcite in quantitative comparison with available data@xcite . however , the experimental results were extracted by clustering jets with the kt algorithm , while the sca jet vertices used for the theoretical calculation were those obtained with the furman algorithm by ivanov and papa ( fip ) . also the small - cone analysis performed in ref . @xcite compared calculations with the exact jet vertices in the kt algorithm versus the small - cone ones in the fip algorithm . starting from these premises , and after reviewing in sec . [ s : ts ] the theoretical setup for the description of mueller - navelet jets in terms of the collinear and high - energy factorization formulae , in sec . [ s : jetalg ] we discuss in detail the differences among the three ( kt , cone and fip ) jet algorithms in the relevant case of two near particles and we determine the kinematical configurations in the limit of small jet `` radius '' @xmath3 . we then derive the small - cone jet vertices for the kt and cone algorithms in sec . [ s : scjv ] , by computing their differences with respect to the fip algorithm induced by the different kinematical conditions . in sec . [ s : nc ] we perform a numerical study in order to asses the quantitative difference between the exact and small - cone jet vertex in the kt algorithm , by comparing the vertices themselves as well as a typical differential cross section for mn jets and some angular coefficients measuring the azimuthal decorrelation between the jets . in addition we also determine the discrepancies of the same quantities induced by a different choice of the jet algorithm . we discuss the results in sec . [ s : conc ] , where we conclude that the wrong choice of algorithm causes sizeable errors on the predictions , while the sca within the same algorithm provides a good approximation to the exact quantities and can therefore be used as a valuable tool for a quantitative description of mn and forward jets . the process we are considering was suggested long ago by mueller and navelet @xcite in order to study the high - energy behaviour of qcd . it is generated by the collision of two hadrons @xmath4 typically ( anti)protons and is characterized by the detection in the final state of two hard jets @xmath5 with large rapidity separation : @xmath6 where @xmath7 represents any additional emission . each jet @xmath8 represents a cluster of particles grouped together according to some given jet algorithm and is described by 3 variables : the rapidity @xmath9 , the _ transverse _ energy @xmath10 and the azimuthal angle @xmath11 , @xmath12 being the @xmath13-th jet transverse momentum . the kinematical region where one expects the high - energy qcd dynamics to play an important role is given by @xmath14 where the condition of hard jets ( @xmath15 ) is imposed for the applicability of perturbation theory . in the leading twist approximation , i.e. , up to power suppressed correction in the hard scale parameter @xmath16 , the hadronic cross section @xmath17 can be factorized in the ( longitudinal momentum fraction ) convolution of two partonic distribution functions ( pdfs ) @xmath18 and a partonic cross section @xmath19@xmath20 where @xmath21 denote the parton flavours ( quark or gluon ) and @xmath22 the partonic momentum fractions w.r.t . their parent hadrons . in turn , in the high - energy regge regime we are considering , the partonic cross section for jet production can be factorized in a ( transverse momentum ) convolution of ( process dependent ) jet vertices @xmath23 and a universal factor @xmath24 called gluon green s function ( ggf)@xmath25 where @xmath26 denotes the ( reggeized ) gluon transverse momentum flowing out from the ggf and entering the jet vertex , while @xmath27 is the center - of - mass energy squared of the partonic subsystem . the overall factorization structure is depicted in fig . [ f : fact ] . represents the parton densities , @xmath28 the jet vertices and @xmath24 the gluon green s function . _ _ ] the pdfs are non - perturbative objects that depend also on the renormalization and factorization scales @xmath29 , @xmath30 . while @xmath29 is introduced in the renormalization of uv divergencies , @xmath30 enters in the treatment of the ir collinear divergencies which are absorbed by the pdfs . the @xmath30 dependence of the pdfs is governed by the dglap equations @xcite , and their evolution kernels ( splitting functions ) are known at next - to - next - to - leading order ( nnlo ) . by means of global fits they have been determined in a wide range of the @xmath31 plane . the ggf is the central object in high - energy qcd , in that it resums the @xmath32 to all orders in perturbation theory . it obeys the bfkl equation @xcite @xmath33 where @xmath34 is the mellin transform of @xmath35 , defined in terms of an arbitrary energy scale @xmath36 . the bfkl kernel @xmath37 is known in next - to - leading - logarithmic ( nll ) approximation . the coefficient @xmath38 depends on the choice of both @xmath29 and @xmath36 . finally , the jet vertices are perturbative finite objects without energy ( @xmath1 ) dependence , and are known in nlo approximation : @xmath39 . they depend on the jet variables @xmath40 and also on the arbitrary scales @xmath41 , in such a way that the hadronic cross section be independent of those scales up to nll terms , i.e. , the scale dependence is present only in the terms of relative order @xmath42 . it is apparent that the determination of the mueller - navelet ( mn ) jet cross section involves quite a number of integrals , both in the factorization formulae ( [ collfact],[hefact ] ) , and in the determination of @xmath24 and @xmath23 , as can be checked from their explicit expressions @xcite . furthermore , when comparing the theoretical predictions with experiments , one needs integrated cross sections in some of the jet variables , in order to comply with the experimental binning . such integrations are mostly done numerically , and this would require a large amount of computing resources or time for reaching a precision at the level of 1% . in order to cope with such a problem , two techniques can be exploited so as to reduce the computing time and improve convergence : * to project the ggf and jet vertices on a complete set of functions which respects the symmetries of the process ( e.g. , azimuthal invariance ) ; * to use an approximated and simpler version of the jet vertices . these methods are often used in bfkl phenomenological analyses , and we shall illustrate them in the following subsections : the former in order to set up the theoretical framework and the main notations ; the latter in order to introduce the main subject of this paper . the first method to reduce computing time does not involve any approximation , at least at the nll level of accuracy we are working with . it is better illustrated in the ll approximation , where the strong coupling @xmath43 is fixed and thus the bfkl kernel and jet vertices , in addition of being invariant under azimuthal rotations , are also scale invariant . , @xmath44 . ] in this case , a fourier - mellin transform diagonalizes the transverse integrations and the ensuing expressions are considerably simpler to evaluate . one can proceed in this way : first of all , let s exploit the fact that , in the ll approximation , each partonic momentum fraction coincides with the corresponding jet s longitudinal momentum fraction ( because of a @xmath45 in @xmath46 ) : @xmath47 so that @xmath48 , where @xmath49 is the rapidity distance between the jets . if we adopt the convenient and natural choice of @xmath50 as energy scale , the green s function ( [ gomega ] ) is now independent of the partonic momentum fractions : @xmath51 secondly , we introduce the _ impact factor _ by integrating a jet vertex with the corresponding parton density : @xmath52 so that we can rewrite the factorization formula in the form @xmath53 at this point we project kernel and vertices onto the eigenfunctions of the ll bfkl kernel @xmath54 satisfying the completeness relation @xmath55 and providing the ll eigenvalue function @xmath56 @xmath57({\boldsymbol{k } } ) & \equiv \int{\mathrm{d}}^2{\boldsymbol{k}}'\ ; k^{(0)}({\boldsymbol{k}},{\boldsymbol{k } } ' ) { \mathcal{e}_{n\nu}}({\boldsymbol{k } } ' ) = \chi^{(0)}_{n\nu } { \mathcal{e}_{n\nu}}({\boldsymbol{k } } ) \label{k0e}\\ \chi^{(0)}_{n\nu } & = 2\psi(1)-\psi\left(\frac{1+n}{2}+{\mathrm{i}}\nu\right ) -\psi\left(\frac{1+n}{2}-{\mathrm{i}}\nu\right ) \;. \label{chi0}\end{aligned}\ ] ] finally , by inserting a completeness ( [ comp ] ) between each pair of factors in ( [ sigmafact ] ) , we arrive at the convenient expression for the differential cross section @xmath58 where in the last equality we have used the fourier - mellin transforms @xmath59 the delta functions on the r.h.s . of eq . ( [ gn ] ) are just a consequence of the azimuthal- and scale - invariance of the kernel , and allow us to trade two bidimensional integrals for a sum and a simple integral . the azimuthal correlation of the mn jets is usually measured by means of the fourier coefficients ( @xmath60 ) @xmath61 because of the azimuthal- and scale - invariance of the vertices , it is easy to show that @xmath62 where @xmath63 is dimensionless , as jet vertex . ] thus obtaining a factorization formula with just one integration : @xmath64 such a structure is preserved in the nll approximation too , provided the impact factors and ggf are suitably modified in order to take into account the loss of scale - invariance due to the renormalization procedure and to the factorization of collinear singularities , which translates in a dependence on @xmath65 , @xmath66 and @xmath67 of the ggf and impact factors . the expression of the ggf in mellin - space is very simply expressed in terms of the eigenvalue @xmath68 of the bfkl kernel : @xmath69 however , the computation of @xmath70 at nll level involves several integrations . the main advantage of this procedure is that such integrations can be done once and for all for each set of one - jet variables . nevertheless , such computations can still be rather lengthy , and the use of an approximate expression of the impact factors to be described in the next subsection turns out to be very convenient . let us conclude this section by noticing that the factorization formula ( [ cm1 ] ) is very useful also in the case of cross - section integrated in jet energies . in fact , a double integral in @xmath71 and @xmath72 factorizes into the product of simple integrals of the impact factors ( provided the integration domain @xmath73 can be factorized into the cartesian product of two one - dimensional sets ) @xmath74 \ ; \left[\int_{i_2 } { \mathrm{d}}e_2 \ ; ( e_2 ^ 2)^{{\mathrm{i}}\nu-1 } \psi_{b\,m\nu } \right]^ * \\ \nonumber & \equiv ( 2\pi)^2 \int{\mathrm{d}}\nu \ ; g_{m\nu}(y ) \ ; { \boldsymbol{\psi}}_{a\,m\nu}(y_1 ) \ ; { \boldsymbol{\psi}}_{b\,m\nu}^*(y_2 ) \;.\end{aligned}\ ] ] in this case , the integrated impact factors @xmath75 can be computed independently and stored in suitable grids , thus reducing a lot the computational effort of the phase - space integration . in the general case @xmath76 , however , the expression ( [ cm1 ] ) has to be numerically integrated in energy ( and possibly in rapidity ) , and a suitable approximation ( like the sca ) could be a very valuable tool to diminish the computing demand . in order to study the behaviour of the jet vertex for small values of the `` radius '' @xmath3 to be precisely defined later on and possibly to speed up the computation of the jet impact factor @xmath70 and of its energy - integrated version @xmath77 , one can use the small - cone approximation ( sca ) , as suggested and derived in @xcite . the dependence of the impact factor on the jet radius @xmath3 has the form @xcite ( @xmath78 is a shorthand for @xmath29 , @xmath29 and @xmath79 ) @xmath80 and the analytic expressions for the coefficients @xmath81 were explicitly computed @xcite for a particular jet algorithm ( fip ) . however , whereas the coefficient @xmath82 of @xmath83 depends only on the incoming hadron and is given in terms of the usual splitting functions as @xmath84 \sum_{a\in\{q,\bar{q}\ } } f_{a}\big(\frac{x_j}{\zeta}\big ) \\ & + \big[p_{gg}(\zeta)+2 n_f p_{qg}(\zeta)\big ] f_{g}\big(\frac{x_j}{\zeta}\big ) \bigg\ } \;,\end{aligned}\ ] ] the constant term @xmath85 depends also on the details of the jet algorithm . ivanov and papa @xcite computed such coefficient for an algorithm which was used in pioneering work on qcd jets by furman @xcite we shall refer to it as fip algorithm which , however , is no more used in present day phenomenology . the main purpose of our paper is to derive such coefficient for the two mostly used algorithms of qcd analysis , namely the cone - algorithm and the kt - algorithm . the computations can be repeated by following the procedure of @xcite . the expressions for the jet vertices at ll and nll level are extracted from the perturbative calculation of processes with two incoming partons producing 2-jet at lo and nlo respectively . at lo the amplitudes have just two partons in the final states , each of which is identified with a jet . the jets are emitted back - to - back in the azimuthal direction and have no substructure . therefore no dependence on the algorithm is found at lo . the same is true at nlo as far as the virtual corrections are concerned . on the other hand , the nlo real corrections involve 3 partons in the final state , therefore a jet can be constituted by either one or two of them . in the case of 1-parton ( simple ) jet , all algorithms are designed in such a way that no further emission is found within a region of radius @xmath3 in the @xmath2 plane around the position of that parton . therefore , the differences among the algorithms are to be found in the 2-parton ( composite ) jet configurations . in the following section we will carefully compare the definitions the _ cone _ , _ kt _ , and _ fip _ jet algorithms and , from their differences , we shall compute the small - cone impact factors for the cone and kt jets . according to ref . @xcite , when two partons @xmath86 and @xmath87 are combined into one jet of radius @xmath3 , the resulting jet variables @xmath88 are defined to be @xmath89 to determine whether the two partons are to be combined , we see if they fit in a cone of radius @xmath3 about the jet axis in the @xmath2 plane . in practice , by denoting with @xmath90 the @xmath2-distance @xmath91 one requires @xmath92 and @xmath93 for the composite jet , which amounts the condition @xmath94 a simple jet can then be defined only if the two partons can not be combined , i.e. , provided @xmath95 according to ref . @xcite , the kt clustering algoritm consists in an iterative procedure which is based on comparing a set of resolution variables of single - particle @xmath96 ( b for beam ) and pairs @xmath97 @xmath98 one then considers the smallest one : if it is a @xmath96 , particle @xmath13 is thrown in the beam basket and removed from the list ; if it is a @xmath97 then particles @xmath13 and @xmath99 are merged into a pseudoparticle @xmath100 e.g. , by using the recombination scheme ( [ conejet ] ) . the procedure is repeated from the beginning , until all resolution variables are greater than some ( hard ) stopping parameter @xmath101 . in the case of three partons in the final state there are these possibilities : * one of the @xmath96 , say @xmath102 is the smallest resolution variable . * * if @xmath103 then particle 1 belongs to the beam and we are left with two partons which , if not in the beam , form two simple jets ; * * if @xmath104 then the clustering stops and all three particles form simple jets ; * one of the @xmath97 , say @xmath105 is the smallest resolution variable . in this case 1 and 2 are merged into a pseudoparticle \{12 } and one considers three resolution variables : @xmath106 , @xmath107 and @xmath108 . * * if the smallest is larger than @xmath109 clustering stops and we have one simple jet \{3 } and one composite jet \{12 } ; * * if the smallest is less than @xmath109 , a ( pseudo)particle belongs to the beam and we can not have two jets in the final state ; to summarize , we have a composite jet , say \{12 } , only if @xmath110 in the other case @xmath111 only simple jets are present . in @xcite ivanov and papa define a jet as a set of particles within a cone of radius @xmath3 . to be more precise , they require that : * the jet s momentum is the sum of the particles momenta ; * all and only the particles of the jet belong to a circle of radius @xmath3 in the @xmath2 plane and centered at the jet s momentum . this is just the jet definition of furman @xcite used in the 80s for early phenomenology of nlo qcd jets we denote it `` fip algorithm '' . in the case under study , where at most two particles ( 1 and 2 ) can form a jet , the condition for a composite jet is nothing but the prescription adopted in the cone algorithm : two particles such that @xmath112 can be considered as a composite jet . on the other hand , particle 1 can form a simple jet if no other particle is found within a distance @xmath3 from it , namely @xmath113 this definition is somewhat pathological , because it may happen that a given configuration can give rise to both a composite jet and two simple jets . this fact can be easily understood in the case of two particles with the same transverse energy and whose distance satisfy @xmath114 . they can form simple jets because a cone of radius @xmath3 centered on either particle does not contain the other ; on the other hand , a cone of radius @xmath3 centered halfway the two particles contains both of them . for this reason , it is not possible to extend such jet definition into an ir - safe algorithm to all orders , i.e. , with an arbitrary number of particles , hence it has been abandoned in favour of better algorithms like the cone and especially the kt one . in any respect , this definition is different from both the cone algorithm ( in the case of simple jets ) and the kt algorithm ( in the case of composite jets ) , and yields different results when used to defined any observable . it is the main purpose of this paper to derive the correct expressions for the jet vertices in the sca for the kt algorithm and also for the cone one . in this section we shall compute the sca jet vertex in the cone and kt algoritm . the idea is to identify and calculate the contributions differing from those of the original paper @xcite where the fip algorithm was adopted . as explained in the previous section , for dijet production at nlo , the differences among jet algorithms occur only in the way that two partons can be combined to form a jet . in the bfkl approach , two partons ( 1 and 2 ) in the fragmentation region of an incoming hadron are produced by the interaction of a parton stemming from the hadron and a reggeized gluon , as depicted in fig . [ f : fragkinem ] . in the high - energy kinematics , the incoming parton has just a fraction @xmath115 of the pure longitudinal momentum of the hadron , while the reggeized gluon s momentum @xmath116 is essentially transverse . following ivanov - papa ( ip ) @xcite we work in dimensional regularization ( @xmath117 ) and we indicate with @xmath118 ( @xmath119 ) the ( @xmath120)-dimensional transverse momentum of the outgoing parton 1 ( 2 ) , and with @xmath121 ( @xmath122 ) its longitudinal momentum fraction with respect to the incoming parton . in terms of these variables , the relative rapidity and azimuthal angle between the two outgoing partons are @xmath123 we shall present our results by using the notations of ref . @xcite , namely in term of their `` @xmath124 '' quantities , which are related to our definition of jet impact factor ( [ phin ] ) by @xmath125 where @xmath13 ( @xmath126 ) are labels for the initial ( final ) state of the sub - processes contributing to the cross section , while the superscripts @xmath3 and @xmath23 denote real and virtual parts respectively . the condition for composite jet in the cone algorithm ( [ conecomp ] ) coincides with the one adopted by fip ( [ ipcomp ] ) , thus no difference is expected for the corresponding contributions . in contrast , the conditions for simple jets ( [ conesimple ] ) and ( [ ipsimple ] ) are different , causing different contributions to the jet vertices . by using the notations of ref . @xcite , @xmath127 is the transverse momentum of the parton forming the simple jet , @xmath128 is the transverse momentum of the parton outside the jet ( also called `` spectator '' ) , and one introduces the transverse vector @xmath129 @xmath130 which vanishes when the two partons are collinear . in fact , for small @xmath129 , we have @xmath131 and also @xmath132 therefore , the simple - jet condition ( [ conecomp ] ) becomes @xmath133 at variance with the fip condition ( [ ipcomp ] ) which , expressed in terms of @xmath129 , reads like eq . ( [ conedelta ] ) but without the denominator @xmath134 @xcite . in order to identify the contributions of the fip jet vertex that have to be modified in the cone algorithm , let us recall that the simple jet configurations were computed in two steps : _ 1 ) _ by allowing the spectator parton to span the whole phase space and then _ 2 ) _ by subtracting the contribution stemming from the spectator parton inside the jet cone . therefore , we have to modify only the subtractions by replacing @xmath135 . the * quark initiated * jet vertex ( quark+regge - gluon @xmath136 quark+gluon , see fig . [ f : fragkinem ] ) has two such subtractions , one for the quark - jet and one for the gluon - jet , whose results are reported in eqs . ( 5.36 ) and ( 5.38 ) of ref . @xcite respectively . in the gluon - jet term the substitution @xmath137 is straightforward : @xmath138 \to -\frac{{\alpha_s}}{2\pi}\frac{\gamma(1-{\epsilon})}{{\epsilon}(4\pi)^e}\frac{\gamma^2(1+{\epsilon})}{\gamma(1 + 2{\epsilon } ) } ( k^2)^{\gamma+{\epsilon}-\frac{n}{2 } } ( k\cdot l)^n \int_{x_j}^1\frac{{\mathrm{d}}\zeta}{\zeta}\;\zeta^{-2\gamma } \sum_{a = q,\bar{q } } f_a\left(\frac{x_j}{\zeta}\right ) \nonumber \\ & \qquad\qquad\times\left(\frac{r}{\max(\zeta,\bar\zeta)}\right)^{2{\epsilon } } \left[p_{gq}(\zeta ) \left(1 + 2{\epsilon}\log\frac{\bar\zeta}{\zeta}\right)+{\epsilon}c_f\zeta\right ] \nonumber \\ & = i_{q;g ,- q}^r - \frac{{\alpha_s}}{2\pi } ( k^2)^{\gamma-\frac{n}{2 } } ( k\cdot l)^n \int_{x_j}^1\frac{{\mathrm{d}}\zeta}{\zeta}\;\zeta^{-2\gamma } \sum_{a = q,\bar{q } } f_a\left(\frac{x_j}{\zeta}\right ) p_{gq}(\zeta ) 2\log\big(\max(\zeta,\bar\zeta)\big ) \ ; , \label{diffiqgq}\end{aligned}\ ] ] where @xmath139 and @xmath140 is a complex vector lying only in the first two of the @xmath120 transverse dimensions . in the quark - jet one has to proceed more carefully , because of the presence of a double pole in @xmath141 multiplying @xmath142 . this would generate , among other things , modified simple poles and also finite double logs . however , since the double pole is multiplied by a @xmath143 , for these terms @xmath144 and the outcome is identical to the fip algorithm . the only difference comes from the simple pole in front of the @xmath145 splitting function , and we obtain @xmath146 \nonumber \\ & \to i_{q;q ,- g}^r - \frac{{\alpha_s}}{2\pi } ( k^2)^{\gamma-\frac{n}{2 } } ( k\cdot l)^n \int_{x_j}^1\frac{{\mathrm{d}}\zeta}{\zeta}\;\zeta^{-2\gamma } \sum_{a = q,\bar{q } } f_a\left(\frac{x_j}{\zeta}\right ) p_{qq}(\zeta ) 2\log\big(\max(\zeta,\bar\zeta)\big ) \;. \label{diffiqqg}\end{aligned}\ ] ] the * gluon initiated * jet vertex has two subtractions too , one for the @xmath147 final state and one for the @xmath148 final state , whose results are reported in eqs . ( 5.49 ) should be @xmath149 instead of @xmath116 . also the last subscript in eq . ( 5.38 ) should be @xmath150 instead of @xmath151 . ] and ( 5.56 ) of ref . @xcite respectively . the situation is very similar to that of the quark initiated vertex : the subtraction in the ( anti)quark jet contains a simple pole times the @xmath152 splitting function in front of the @xmath153 factor , while the gluon jet contains both single and double poles , the former with the @xmath154 splitting function and the latter with the @xmath143 distribution . for both types of jets the substitution @xmath135 simply amounts to finite contributions proportional to @xmath155 : @xmath156 + \ { q \leftrightarrow \bar{q } \ } + ( 5.56)[1 ] \nonumber \\ & \to i_{g;q,-\bar{q}}^r + i_{g;\bar{q},-q}^r + i_{g;g ,- g}^r \label{diffig } \\ & - \frac{{\alpha_s}}{2\pi } ( k^2)^{\gamma-\frac{n}{2 } } ( k\cdot l)^n \int_{x_j}^1\frac{{\mathrm{d}}\zeta}{\zeta}\;\zeta^{-2\gamma } \sum_{a = q,\bar{q } } f_a\left(\frac{x_j}{\zeta}\right ) \left[2 n_f p_{qg}(\zeta ) + p_{gg}(\zeta ) \right ] 2\log\max(\zeta,\bar\zeta ) \;. \nonumber\end{aligned}\ ] ] to sum up , the jet vertex for the cone algorithm in the small - cone approximation is obtained by replacing @xmath135 in the final formulae ( 5.39 ) and ( 5.57 ) of @xcite . the complete expressions are written in eqs . ( [ finalq],[finalg ] ) . the condition for simple jet in the kt algorithm ( [ ktsimple ] ) coincides with the one adopted by fip ( [ ipsimple ] ) , while the conditions for composite jet ( [ ktcomp ] ) and ( [ ipcomp ] ) are different . in the composite jet configuration the jet s transverse momentum is @xmath157 and it is convenient to define the auxiliary transverse vector @xmath129 as @xmath158 thus obtaining @xmath159 the composite jet condition ( [ ktcomp ] ) becomes @xmath160 at variance with the fip condition ( [ ipcomp ] ) which reads @xcite @xmath161 . the corresponding contributions of @xcite , that have to be modified in order to recover the jet vertex in the kt algorithm , are found in sec . 5.1.1.c for the quark initiated vertex and secs . 5.2.1.b and 5.2.2.b for the gluon initiated one . in all such cases the modification amounts to replace @xmath162 and finally to substitute @xmath163 in the @xmath121-integral with the relevant splitting function . explicitly , in the * quark - initiated * case , such integrals for the fip and kt algoritms reads respectively @xmath164^{2{\epsilon } } \frac{1+\bar\zeta^2+{\epsilon}\zeta^2}{\zeta } = \frac1{{\epsilon } } -\frac32 + \left(\frac72 -\frac{\pi^2}{3}+3\log 2\right){\epsilon}\\ i_{q;q+g}^{r\;(\mathrm{kt } ) } & \propto \int_0 ^ 1{\mathrm{d}}\zeta\ ; \quad\;\ ; ( \zeta\bar\zeta)^{2{\epsilon } } \quad \frac{1+\bar\zeta^2+{\epsilon}\zeta^2}{\zeta } = \frac1{{\epsilon } } -\frac32 + \left(\frac{13}{2 } -\frac{2\pi^2}{3}\right){\epsilon}\;. \label{iqqgkt}\end{aligned}\ ] ] their difference @xmath165 , when multiplied by the overall @xmath166 pole , provides a finite contribution that has to be added to the ip result in order to obtain the proper expression for the kt algorithm : @xmath167 \to i_{q;q+g}^r + \frac{{\alpha_s}}{2\pi } ( k^2)^{\gamma-\frac{n}{2 } } ( k\cdot l)^n \sum_{a = q,\bar{q } } f_a(x_j ) c_f \left(3-\frac{\pi^2}{3}-3\log2 \right ) \;. \label{diffiqg}\end{aligned}\ ] ] in the * gluon - initiated * case the procedure is identical ; here we have two contributions : one from the @xmath147 jet @xmath168 \to i_{g;q+\bar{q}}^r + \frac{{\alpha_s}}{2\pi } ( k^2)^{\gamma-\frac{n}{2 } } ( k\cdot l)^n \frac{c_a}{c_f } f_g(x_j ) \ , 2 n_f t_r \left(\frac{2}{3}\log2 -\frac{23}{36 } \right ) \ ; , \label{diffiqqq}\end{aligned}\ ] ] and one from the @xmath148 jet @xmath169 \to i_{g;g+g}^r + \frac{{\alpha_s}}{2\pi } ( k^2)^{\gamma-\frac{n}{2 } } ( k\cdot l)^n \frac{c_a}{c_f } f_g(x_j ) c_a \left ( \frac{131}{36 } -\frac{\pi^2}{3 } -\frac{11}{3}\log2 \right ) \;. \label{diffiggg}\end{aligned}\ ] ] the integrals ( [ iqqgfip],[iqqgkt ] ) and the analogous ones for the gluon - initiated contributions , yielding eqs . ( [ diffiqg]-[diffiggg ] ) , were already considered and computed @xcite in the first study of the relation between the kt and the cone algorithm the latter sharing with fip the condition of composite jet . the result of the jet vertex for the cone and kt algoritms is reported below , by adding to the original expressions of ip @xcite the modifications computed in the previous subsections and here highlighted in boldface : @xmath170 for the cone and @xmath171 for the kt . the quark part is @xmath172\log\frac{k^2}{\mu_f^2 } + \nonumber \\ & \quad- 2 \zeta^{-2\gamma}[p_{qq}(\zeta)+p_{gq}(\zeta)]\log \frac{r}{\boldsymbol{\big\langle\max(\zeta,\bar\zeta)\big\rangle_c}}-\frac{\beta_0}{2}\log\frac{k^2}{\mu_r^2}\delta(1-\zeta ) \nonumber\\ & \quad+c_a\delta(1-\zeta)\left\{\chi^{(0)}_{n\nu}\log\frac{s_0}{k^2}+\frac{85}{18}+\frac{\pi^2}{2 } + \frac12\left[\psi'\left(1+\gamma+\frac{n}{2}\right)-\psi'\left(\frac{n}{2}-\gamma\right)-\chi^{(0)\,2}_{n\nu } \right]\right\ } \nonumber\\ & \quad+(1+\zeta^2)\bigg\{c_a\left[\frac{(1+\zeta^{-2\gamma})\chi^{(0)}_{n\nu}}{2(1-\zeta)_+ } -\zeta^{-2\gamma}\left(\frac{\log(1-\zeta)}{1-\zeta}\right)_+\right ] \nonumber \\ & \quad+\left(c_f-\frac{c_a}{2}\right ) \left[\frac{\bar\zeta}{\zeta^2}i_2-\frac{2\log \zeta}{\bar\zeta}+2\left(\frac{\log(1-\zeta)}{1-\zeta}\right)_+ \right]\bigg\ } \nonumber \\ & \quad + \delta(1-\zeta)\left [ c_f\left(3\log 2-\frac{\pi^2}{3}-\frac{9}{2 } + \boldsymbol{\big\langle 3-\frac{\pi^2}{3}-3\log 2 \big\rangle_k}\right ) -\frac{10}{9 } n_f t_r\right ] \nonumber \\ & \quad+c_a \zeta+c_f \bar\zeta + \frac{1+\bar\zeta^2}{\zeta}\left[c_a\frac{\bar\zeta}{\zeta}i_1 + 2c_a\log\frac{\bar\zeta}{\zeta } + c_f \zeta^{-2\gamma}(\chi^{(0)}_{n\nu}-2\log\bar\zeta)\right]\bigg\ } \ ; , \label{finalq}\end{aligned}\ ] ] where @xmath173 , @xmath174 , @xmath175 is the digamma function , while the splitting functions @xmath176 and the special functions @xmath177 are reported in app . [ a : def ] . the gluon part is @xmath178\log\frac{k^2}{\mu_f^2 } + \nonumber \\ & \quad- 2 \zeta^{-2\gamma}[p_{gg}(\zeta)+2 n_f p_{qg}(\zeta)]\log \frac{r}{\boldsymbol{\big\langle\max(\zeta,\bar\zeta)\big\rangle_c}}-\frac{\beta_0}{2}\log\frac{k^2}{4\mu_r^2}\delta(1-\zeta ) \nonumber\\ & \quad+c_a\delta(1-\zeta)\left\{\chi^{(0)}_{n\nu}\log\frac{s_0}{k^2 } + \frac12\left[\psi'\left(1+\gamma+\frac{n}{2}\right)-\psi'\left(\frac{n}{2}-\gamma\right)-\chi^{(0)\,2}_{n\nu } \right ] \right . \nonumber \\ & \qquad\qquad\qquad\qquad \left . + \frac1{12}+\frac{\pi^2}{6 } + \boldsymbol{\big\langle \frac{131}{36 } -\frac{\pi^2}{3 } -\frac{11}{3}\log 2 \big\rangle_k } \right\ } \nonumber\\ & \quad+2c_a(1-\zeta^{-2\gamma})\left[\left(\frac1{\zeta}-2+\zeta \bar\zeta\right)\log \bar\zeta + \frac{\log(1-\zeta)}{1-\zeta}\right ] \nonumber \\ & \quad+c_a\left[\frac1{\zeta}+\frac1{(1-\zeta)_+}-2+\zeta \bar\zeta\right ] \left[(1+\zeta^{-2\gamma})\chi^{(0)}_{n\nu}-2\log \zeta+\frac{\bar\zeta^2}{\zeta^2 } i_2\right ] \nonumber \\ & \quad+ 2 n_f t_r \left[2\frac{c_f}{c_a}\zeta \bar\zeta+(\zeta^2+\bar\zeta^2)\left(\frac{c_f}{c_a}\chi^{(0)}_{n\nu } + \frac{\bar\zeta}{\zeta}i_3\right ) \right.\nonumber \\ & \qquad\qquad\qquad \left . + \delta(1-\zeta)\left ( -\frac1{12 } + \boldsymbol{\big\langle \frac{2}{3}\log 2-\frac{23}{36 } \big\rangle_k } \right ) \right]\bigg\ } \;. \label{finalg}\end{aligned}\ ] ] it is apparent from the above expression that the @xmath83 coefficient @xmath82 of the jet vertex is independent of the jet algorithm , while the constant coefficient @xmath85 depends on it . in this section we assess the quantitative difference among the jet vertices in the three algorithms ( cone , kt , fip ) that we considered , and also the corresponding accuracy of their small - cone approximations ( sca ) . we shall use the term _ exact _ in the sense of `` without sca '' . we start by evaluating the `` exact '' jet vertices in the three algorithms we employ the numerical code used in ref . @xcite for various values of the jet radius @xmath3 , and we compare them with their sca . we expect the sca to be better , the smaller the values of @xmath3 , and increasing discrepancy with increasing @xmath3 . on the other hand , the differences among different algorithms should nt vanish with @xmath3 , according to our analysis . of the nlo jet vertices @xmath63 in the three jet algorithms discussed in the text : fip ( dashed red ) , cone ( dotted green ) and kt ( solid blue ) . the big dots correspond to the exact evaluation ( the small error bars showing montecarlo integration uncertainties ) , while the straight lines denote the small - cone approximation . here @xmath180 , @xmath181 , @xmath182 , @xmath183 . _ _ ] this is actually the case , as can be seen in fig . [ f : scavsr ] , where we plot the exact ( points ) and small - cone ( lines ) nlo part of the jet vertex versus @xmath3 , in the three algorithms mentioned before ( for a given choice of the parameters @xmath184 , @xmath185 , @xmath186 , @xmath187 ) . the common slope of the lines represents the coefficient @xmath82 of @xmath83 , which is the same for the three algorithms , while the intercepts at @xmath188 give the constant coefficients @xmath85 , which clearly depend on the algorithm . a detailed study , carried out with several values of the parameters , shows that the small - cone approximation works very well up to @xmath3 of few tenths , with an error below 1% for @xmath189 which increases up to 3 - 4% when @xmath0 . next , we specialize our analysis to realistic values of the jet radius and energy . since the typical phenomenological studies on mn jets use @xmath190 and jet transverse energies @xmath191 , in the following all quantities will be evaluated at @xmath0 and @xmath183 . in addition , nowadays the mostly used jet algorithm is the kt . therefore we adopt `` exact '' quantities , computed in the kt algorithm , as reference quantities , and we estimate the deviations to them introduced by the sca . to some extent , adopting the sca at fixed @xmath3 is a sort of choosing a jet algorithm . the natural question then arises : how does the discrepancy introduced by the small - cone approximation @xmath192 compare with the discrepancy caused by different choices of jet algorithm , i.e. , @xmath193 in order to answer this question , we compute the exact nlo part of jet vertex @xmath194 in the kt algorithm as function of @xmath185 , and compare it with the sca in the same algorithm and also with the sca in the fip algorithm . , @xmath195 , @xmath182 , @xmath183 . _ _ ] fig . [ f : scavsnu ] shows such a comparison for @xmath195 . we can see that the exact result in the kt algorithm is much better approximated by the sca in the same algorithm rather than by the fip choice , in particular at small values of @xmath185 , which are the most important in the @xmath185 integrals of eqs . ( [ cm1],[intcm ] ) since the ggf is peaked around @xmath181 . this conclusion is further supported by analysing the whole ( lo+nlo ) @xmath185-integrand of eq . ( [ intcm ] ) , again by comparing exact - kt , sca - kt and sca - fip , as in fig . [ f : integrandvsnu ] . the discrepancy introduced by the sca in the wrong fip algorithm is about three times larger than that introduced by the sca in the proper kt algorithm , the latter being of the order of 5% . actually , the relative error due to the sca is slightly larger for the full ( lo+nlo ) quantities than for the pure nlo ones . this is due to the fact that the nlo corrections usually have sign opposite to the common lo terms , giving rise to cancellations in their sum that amplify the relative differences . -integrand for kt algorithm ( points ) with the sca in the same algorithm ( solid blue ) and in the fip algorithm ( dashed red ) . on the left : @xmath180 ; on the right : @xmath195 . the upper plots display absolute values in arbitrary units , and below them we show the ratios w.r.t . exact integrand . the parameters are @xmath0 , @xmath196 , @xmath197 , @xmath198.__,title="fig : " ] -integrand for kt algorithm ( points ) with the sca in the same algorithm ( solid blue ) and in the fip algorithm ( dashed red ) . on the left : @xmath180 ; on the right : @xmath195 . the upper plots display absolute values in arbitrary units , and below them we show the ratios w.r.t . exact integrand . the parameters are @xmath0 , @xmath196 , @xmath197 , @xmath198.__,title="fig : " ] + -integrand for kt algorithm ( points ) with the sca in the same algorithm ( solid blue ) and in the fip algorithm ( dashed red ) . on the left : @xmath180 ; on the right : @xmath195 . the upper plots display absolute values in arbitrary units , and below them we show the ratios w.r.t . exact integrand . the parameters are @xmath0 , @xmath196 , @xmath197 , @xmath198.__,title="fig : " ] -integrand for kt algorithm ( points ) with the sca in the same algorithm ( solid blue ) and in the fip algorithm ( dashed red ) . on the left : @xmath180 ; on the right : @xmath195 . the upper plots display absolute values in arbitrary units , and below them we show the ratios w.r.t . exact integrand . the parameters are @xmath0 , @xmath196 , @xmath197 , @xmath198.__,title="fig : " ] finally , we present the results of the differential cross section @xmath199 and of few angular coefficients @xmath200 . ( @xmath180 ) for kt algorithm ( solid blue ) with the sca in the same algorithm ( dash - dotted blue ) and in the fip algorithm ( dashed red ) . top : absolute values in linear scale ; bottom : ratios of the scas w.r.t . the exact one . here @xmath0 , @xmath196.__,title="fig : " ] + ( @xmath180 ) for kt algorithm ( solid blue ) with the sca in the same algorithm ( dash - dotted blue ) and in the fip algorithm ( dashed red ) . top : absolute values in linear scale ; bottom : ratios of the scas w.r.t . the exact one . here @xmath0 , @xmath196.__,title="fig : " ] in fig . [ f : dsigmady ] we plot the differential cross section @xmath201 by comparing again the exact kt calculation with the small - cone approximations in the kt and fip algorithms . it is evident that the wrong choice of the algorithm yields a large error , especially at lower values of @xmath202 , while the sole sca with the proper algorithm introduces an error of 4 - 8% . the shape of the @xmath203 curves , which are not monothonically decreasing in @xmath202 as one could naively expect , is due to an additional cut in rapidity @xmath204 that we have imposed just for computational convenience , as will be shortly explained . due to this cut , the minimum value of @xmath202 that we allow is @xmath205 and in this limit the cross - section vanishes . it then quickly rises for @xmath206 before eventually decreasing at larger @xmath207 . the reason for imposing the @xmath208 cut is due to the fact that monochromatic ( fixed @xmath186 ) impact factors oscillate at large @xmath185 , causing the @xmath185 integration in eq . ( [ cm1 ] ) to be slowly convergent . actually , the convergence is provided by the ggf , whose modulus decreases at large @xmath185 , the decrease being faster at larger values of @xmath202 . on the other hand , at low values of @xmath202 , and in particular with very asymmetric rapidity configurations ( @xmath209 ) , the numeric @xmath185-integration has to be pushed to large @xmath185-values , thus demanding a large computational effort . since this problem is absent for realistic phenomenological studies ( where the impact factors are integrated in the @xmath186 variable and decrease themselves with @xmath185 ) , and because the goal of this analysis is just to compare the main features of the different algorithms , we solve the convergence issue by imposing the mentioned cut in rapidity @xmath210 . from the plots of figs . [ f : integrandvsnu],[f : dsigmady ] one can infer that the main effect of the sca is mostly an overall normalization change , and that this is also true even for the wrong choice of algorithm ( kt versus fip ) , though with a larger factor . if this were the case , by computing _ ratios _ of observables such effects should cancel out and reproduce more faithfully the exact quantities . this is partially true , as can be seen in fig . [ f : cmcn ] , where we plot some ratios of angular coefficients @xmath211 which are usually adopted in order to measure the azimuthal decorrelation of the mn jets . it is nevertheless evident that the cancellation of the systematic effects is more effective if the sca is made with the proper algorithm , leading to a discrepancy of about 2% or less for all the ratios considered . a different choice of algorithm yields definitely larger discrepancies , and therefore should be avoided . determined with the exact jet vertices in the kt algorithm ( solid blue ) with those obtained in the sca with the same algorithm ( dotted blue ) and with the fip algorithm ( dashed red ) . here @xmath0 , @xmath196.__,title="fig : " ] determined with the exact jet vertices in the kt algorithm ( solid blue ) with those obtained in the sca with the same algorithm ( dotted blue ) and with the fip algorithm ( dashed red ) . here @xmath0 , @xmath196.__,title="fig : " ] + determined with the exact jet vertices in the kt algorithm ( solid blue ) with those obtained in the sca with the same algorithm ( dotted blue ) and with the fip algorithm ( dashed red ) . here @xmath0 , @xmath196.__,title="fig : " ] in this paper we have reconsidered the high - energy factorization formula for mueller - navelet jet production , and we have computed the nlo jet vertices in the small - cone approximation for two different jet algorithms : the _ kt algorithm _ and the _ cone algorithm_. the small - cone approximation amounts to evaluate the generic jet vertex @xmath63 at small jet radius @xmath212 according to the expansion @xmath213 , where the coefficient @xmath82 is universal , depending only on the infra - red properties of qcd , while the constant term @xmath85 depends on the jet algorithm . it should be noted that the neglected term in the expansion is quadratic in @xmath3 , and that the linear term is missing . it turns out that the small - cone approximation is quite accurate for realistic values of @xmath214 and , because of its analytic and computational simplicity , it has been and will be used with benefit . the calculation of the mellin - fourier components of the jet vertices ( cfr.sec . [ s : rms ] ) was originally performed @xcite by adopting an algorithm which is equivalent to furman s one @xcite , but which is not used anymore in modern phenomenological studies , being infra - red unsafe . in order to use such vertices for the analysis of data from high - energy colliders , we have computed them in the two most popular algorithms : the kt and the cone ones . the ensuing analytical expressions have been taken out in sec . [ s : scjv ] , and the differences among the three algorithms have been highlighted . the quantitative difference between the small - cone approximations in the kt algorithm and in furman s one have been estimated in sec . [ s : nc ] , by plotting several quantities of interest : the jet vertices , the integrand of the high - energy factorization formula , the differential cross section w.r.t . the rapidity distance @xmath49 of the two mn jets , and some angular coefficients indicating the azimuthal decorrelations of the jets themselves . also the corresponding quantities with the exact jet vertices in the kt algorithm have been plotted . it turns out that the difference between the kt and furman s algorithms is sizeable , of the order of 20% at the level of cross section , and about 5% for the ratios @xmath215 and @xmath200 , the angular coefficients of azimuthal decorrelation . on the other hand , the discrepancies between the exact results and the small - cone approximated ones are much smaller , of the order of 5% at the level of cross section , and less than 2% for the angular ratios . we therefore conclude that the small - cone expansion , computed with the proper jet algorithm , provides a good approximation to the mueller - navelet jet vertices ; it can then be used as a very convenient tool to perform phenomenological studies , in that it requires much less computational resources than the exact computation . this aspect could be essential when analysing observables obtained by integrating the jet energies in a non - factorized domain , e.g. , by requiring @xmath216 . such condition is often used in dijet analysis since , at fixed nlo , it yields more stable results than the condition @xmath217 , while retaining the @xmath218 symmetry . on the other hand , a non - factorizable domain of integration in energy prevents the use of eq . ( [ intcm ] ) with independently integrated impact factors @xmath75 ; a numerical integration in energy ( and possibly in rapidity ) of the integrand ( [ cm1 ] ) is thus necessary , and the small - cone approximation represents a valuable tool to reduce the computational effort of such calculation . we wish to thank the _ galileo galilei institute for theoretical physics _ where part of this work was performed . the splitting functions found in the jet vertices are defined in the usual way : [ splitfun ] @xmath219 \\ p_{gq}(z ) & = c_f\frac{1+(1-z)^2}{z } \\ p_{qg}(z ) & = t_r\left[z^2+(1-z)^2\right ] \\ p_{gg}(z ) & = 2c_a\left[\frac{1}{(1-z)_+ } + \frac{1}{z } -2+z(1-z)\right ] + \left(\frac{11}{6}c_a-\frac{n_f}{3}\right)\delta(1-z)\;. \end{aligned}\ ] ] [ ij ] @xmath221 \label{i2 } \\ i_{1,3}(n,\gamma,\zeta ) & = \frac{\bar\zeta}{2\zeta } i_2(n,\gamma,\zeta ) \pm\frac{\zeta}{\bar\zeta } \left[\log\zeta+\frac{1-\zeta^{-2\gamma}}{2}\left(\chi^{(0)}_{n\nu}-2\log\bar\zeta\right)\right ] \label{i13 } \end{aligned}\ ] ] 99 a. h. mueller and h. navelet , `` an inclusive minijet cross - section and the bare pomeron in qcd , '' nucl . phys . b * 282 * ( 1987 ) 727 . a. h. mueller , `` parton distributions at very small x values , '' nucl . suppl . * 18c * ( 1991 ) 125 ; a. h. mueller , `` jets at lep and hera , '' j. phys . g * 17 * ( 1991 ) 1443 . j. bartels , d. colferai and g. p. vacca , `` the nlo jet vertex for mueller - navelet and forward jets : the quark part , '' eur . j. c * 24 * ( 2002 ) 83 [ hep - ph/0112283 ] ; j. bartels , d. colferai and g. p. vacca , `` the nlo jet vertex for mueller - navelet and forward jets : the gluon part , '' eur . j. c * 29 * ( 2003 ) 235 [ hep - ph/0206290 ] . d. y. .ivanov and a. papa , `` the next - to - leading order forward jet vertex in the small - cone approximation , '' jhep * 1205 * ( 2012 ) 086 [ arxiv:1202.1082 [ hep - ph ] ] . s. catani , y. l. dokshitzer , m. h. seymour and b. r. webber , `` longitudinally invariant @xmath222 clustering algorithms for hadron hadron collisions , '' nucl . b * 406 * ( 1993 ) 187 . s. d. ellis , z. kunszt and d. e. soper , `` the one jet inclusive cross - section at order alpha - s**3 . gluons only , '' phys . d * 40 * ( 1989 ) 2188 . m. furman , `` study of a nonleading qcd correction to hadron calorimeter reactions , '' nucl . b * 197 * ( 1982 ) 413 . b. jager , m. stratmann and w. vogelsang , phys . d * 70 * ( 2004 ) 034010 [ hep - ph/0404057 ] ; a. mukherjee and w. vogelsang , phys . d * 86 * ( 2012 ) 094009 [ arxiv:1209.1785 [ hep - ph ] ] . f. caporale , d. y. ivanov , b. murdaca and a. papa , `` mueller - navelet small - cone jets at lhc in next - to - leading bfkl , '' nucl . b * 877 * ( 2013 ) 73 [ arxiv:1211.7225 [ hep - ph ] ] . f. caporale , b. murdaca , a. sabio vera and c. salas , `` scale choice and collinear contributions to mueller - navelet jets at lhc energies , '' nucl . b * 875 * ( 2013 ) 134 [ arxiv:1305.4620 [ hep - ph ] ] . cms collaboration [ cms collaboration ] , `` azimuthal angle decorrelations of jets widely separated in rapidity in pp collisions at sqrts = 7 tev , '' cms - pas - fsq-12 - 002 . b. duclou , l. szymanowski and s. wallon , `` confronting mueller - navelet jets in nll bfkl with lhc experiments at 7 tev , '' jhep * 1305 * ( 2013 ) 096 [ arxiv:1302.7012 [ hep - ph ] ] . v. n. gribov and l. n. lipatov , `` deep inelastic e p scattering in perturbation theory , '' sov . j. nucl . * 15 * ( 1972 ) 438 [ yad . fiz . * 15 * ( 1972 ) 781 ] ; g. altarelli and g. parisi , `` asymptotic freedom in parton language , '' nucl . b * 126 * ( 1977 ) 298 ; y. l. dokshitzer , `` calculation of the structure functions for deep inelastic scattering and e+ e- annihilation by perturbation theory in quantum chromodynamics . , '' sov . jetp * 46 * ( 1977 ) 641 [ zh . * 73 * ( 1977 ) 1216 ] . v. s. fadin , e. a. kuraev and l. n. lipatov , `` on the pomeranchuk singularity in asymptotically free theories , '' phys . b * 60 * ( 1975 ) 50 ; e. a. kuraev , l. n. lipatov and v. s. fadin , `` multi - reggeon processes in the yang - mills theory , '' sov . jetp * 44 * ( 1976 ) 443 [ zh . eksp . * 71 * ( 1976 ) 840 ] [ erratum - ibid . * 45 * ( 1977 ) 199 ] ; e. a. kuraev , l. n. lipatov and v. s. fadin , `` the pomeranchuk singularity in nonabelian gauge theories , '' sov . jetp * 45 * ( 1977 ) 199 [ zh . * 72 * ( 1977 ) 377 ] ; i. i. balitsky and l. n. lipatov , `` the pomeranchuk singularity in quantum chromodynamics , '' sov . j. nucl . * 28 * ( 1978 ) 822 [ yad . * 28 * ( 1978 ) 1597 ] . d. colferai , f. schwennsen , l. szymanowski and s. wallon , `` mueller navelet jets at lhc - complete nll bfkl calculation , '' jhep * 1012 * ( 2010 ) 026 [ arxiv:1002.1365 [ hep - ph ] ] .

we investigate the condition for capture into first order mean motion resonances using numerical simulations with a wide range of various parameters . in particular , we focus on deriving the critical migration timescale for capture into the 2:1 resonance ; additional numerical experiments for closely spaced resonances ( e.g. , 3:2 ) are also performed . we find that the critical migration timescale is determined by the planet to stellar mass ratio , and its dependence exhibits power law behavior with index . this dependence is also supported by simple analytic arguments . we also find that the critical migration timescale for systems with equal mass bodies is shorter than that in the restricted problem ; for instance , for the 2:1 resonance between two equal mass bodies , the critical timescale decreases by a factor of 10 . in addition , using the obtained formula , the origin of observed systems that include first order commensurabilities is constrained . assuming that pairs of planets originally form well separated from each other and then undergo convergent migration and are captured in resonances , it is possible that a number of exoplanets experienced rapid orbital migration . for systems in closely spaced resonances , the differential migration timescale between the resonant pair can be constrained well ; it is further suggested that several exoplanets underwent migration that can equal or even exceed the type i migration rate predicted by the linear theory . this implies that some of them may have formed _ in situ_. future observations and the use of our model will allow us to statistically determine the typical migration speed in a protoplanetary disk .

1307.7776

when two bodies are in a mean motion resonance , their orbital periods are close to a ratio of two integers , which stabilizes the system in many cases . the laplace 4:2:1 resonance among the galilean satellites and the 3:2 resonances between neptune and trans neptunian objects ( tnos ) are well known examples in the solar system . more than 30 exoplanet systems also include confirmed planets exhibiting mean motion resonances . in addition , _ kepler _ has detected a large number of planet pairs near mean motion resonances ( e.g. , @xcite ; @xcite ) , although they are not necessarily in resonances . many exoplanets are not near mean motion resonances ( e.g. , @xcite ) . however , according to @xcite , such non resonant configurations can be established through orbit crossing among the planets after they are captured in mutual mean motion resonances . thus , mean motion resonances may have played an important role in the planet formation process . to date , mean motion resonances have been studied from many perspectives ( e.g. , @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ) . resonant configurations are thought to arise primarily from convergent migration ( e.g. , @xcite ) . several efforts have been made to derive a critical differential ( relative ) migration timescale @xmath1 above which mean motion resonances can be formed . @xcite analytically studied the restricted three body problem with adiabatic migrating bodies , where the relative migration timescale is much longer than the resonant libration timescale . in contrast , recent studies have adopted numerical methods . one approach is to perform direct _ n_body calculations ( e.g. , @xcite ; @xcite ) , which take a direct summation of the mutual interaction between the bodies . another approach uses the hamiltonian model ( e.g. , @xcite ; @xcite ) , where the canonical equations of the hamiltonian are solved numerically with some approximations . these studies examine the dependence of the critical migration timescale on the mass but consider the restricted three body problem with a massive planet and a massless test particle . @xcite performed direct _ n_body calculations with equal mass planets . however , because their goal was to specify the origin of individual systems , the planetary mass was not usually treated as a parameter . our main aim is to derive an empirical formula for the critical migration timescale of capture into first order @xmath2 mean motion resonances by performing direct _ n_body simulations . we handle the physical variables ( e.g. , mass and damping timescales ) as parameters and vary them over wide ranges . in this way , the dependences of the critical migration timescale on the parameters can be obtained . although the equal mass case is of primary importance for the study of observed exoplanets in resonances , previous general studies , which are not restricted to particular systems , have not considered this situation ; therefore , we examine the case of equal mass bodies . captures into 2:1 resonances , which are the outermost first order resonances , are extensively studied in this paper ; in addition , more closely spaced commensurabilities ( e.g. , 3:2 and 4:3 ) are also examined . from the empirical formula based on our numerical results , we try to constrain the origin of the orbital architecture of planetary systems exhibiting commensurabilities . the structure of this paper is as follows . in section [ sec : model ] , we describe the numerical methods ; in section [ sec : results ] , we present and summarize the results of _ n_body simulations . in section [ sec : discussion ] , we derive the mass dependence of the critical migration time , and in section [ sec : compare ] , we compare our results with those of previous studies . in section [ sec : constraints ] , we apply these results to systems with resonances and discuss their origin , and in section [ sec : conclusions ] , we offer our conclusion . figure [ fig : model ] presents the calculation model considered in this work . initially , two planets with masses of @xmath3 ( inner body ) and @xmath4 ( outer body ) are placed at semimajor axes @xmath5 and @xmath6 , respectively . the initial eccentricity of these bodies is set to @xmath7 , and their inclination is also set to about @xmath7 in radian . damping forces are applied to the bodies that damp the eccentricity and semimajor axis on timescales of @xmath8 and @xmath9 , respectively . we apply the @xmath10damping force only to one body ( mostly to the outer body ) ; therefore , @xmath9 is interpreted as the timescale of differential migration between the bodies . the initial locations are adopted to be separated from the resonance @xmath11^{2/3}$ ] ; namely , @xmath12 for 2:1 resonance , and for other closer resonances ( e.g. , @xmath13 ) the orbits are set just inside the @xmath14 resonance @xmath15^{2/3}$ ] . thus , when @xmath8 is much shorter than @xmath9 , the eccentricity is negligibly small at the resonant encounter . the equation of motion for planet 2 is @xmath16 where the first term on the right hand side is the gravitational force of the central star , the second term is the mutual gravity between the bodies , and the third and fourth are the indirect terms . @xmath17 and @xmath18 are the specific forces for @xmath19damping and @xmath10damping , respectively ( see , for example , @xcite ) . the orbits are directly integrated with the fourth order hermite scheme @xcite over about 0.2 times the migration timescale , which means that although the long term stability is not exmined in this study , a temporary capture with a capture duration of less than about 0.1 times the migration timescale can be excluded . in order to derive critical values of @xmath9 for resonance capture , numerical simuations are performed with different @xmath9 . the sampling interval of @xmath9 adopted in this study is usually 0.2 in a logarithmic scale . we treat the masses ( @xmath3 and @xmath4 ) , the @xmath19damping timescales ( @xmath20 and @xmath21 ) , and the initial eccentricity ( @xmath7 ) as parameters . for a fiducial model , the values are @xmath22 , @xmath23 , @xmath24 , @xmath25 , and @xmath26 , as shown in table [ tbl : parameters ] . here , @xmath27 indicates the orbital period of the inner body . we perform a series of calculations for a wide range of parameters to examine the dependences of resonant capture . cclcl parameters & fiducial value & + @xmath28 & 1 & @xmath29&&@xmath30 + @xmath31 & @xmath32 & @xmath33&&@xmath34 + @xmath35 & @xmath36 & @xmath37&&@xmath36 + @xmath38 & @xmath39 & @xmath37&&@xmath36 + @xmath7 & @xmath40 & @xmath41&&@xmath42 [ tbl : parameters ] the equation of motion can be normalized by the unit length @xmath43 , unit mass @xmath44 , and unit time @xmath45 as @xmath46 where values with tildes on top are those scaled by the unit values . therefore , by changing @xmath3 and @xmath4 , we can discuss the dependence on the stellar mass . hereafter , we usually use @xmath47 and present @xmath3 and @xmath4 in units of earth mass for simplicity . in this paper , we mostly present results assuming @xmath48 because we find that the condition for capture into resonance is approximately expressed by the mass of the larger body , which will be shown in section [ sec : internal_resonance ] . we estimate the critical migration timescale in a manner similar to that of @xcite . orbital integrations of two planets are performed with various parameters to determine whether the planets are captured into 2:1 mean motion resonances . figure [ fig : t_a ] shows the evolution of the semimajor axes and eccentricities of the two planets with a migration timescale @xmath9 of @xmath49 in the fiducial model . when a 2:1 commensurability is formed at @xmath50 , the eccentricities are excited . we also confirm that the two resonant angles and @xmath51 , where @xmath52 and @xmath53 are mean orbital longitudes and longitudes of the pericenter , respectively . ] librate about fixed values . we made 10 runs for each migration timescale with different initial orbital phase angles and derived the capture probability @xmath54 of 2:1 resonances . when the period ratio lies within an error of 1% , we call it a resonant capture . note that the long term stability of the resonant configuration is not considered . figure [ fig : ta_p_fiducial](a ) shows the capture probability for the fiducial model as a function of the migration time . the probability @xmath54 is assumed to have the form @xmath55^{-1},\end{aligned}\ ] ] where @xmath1 is defined as the migration time for which @xmath56 and @xmath57 is a constant . using a least squares fit to the data ( solid line ) , we specify @xmath1 and @xmath57 . the capture probability increases sharply at @xmath58 , which means that the critical migration time @xmath1 can be estimated without introducing large errors @xmath59 . figure [ fig : ta_p_fiducial](b ) presents the results for an initial eccentricity of @xmath60 . in this case , the eccentricity is 0.01 at the resonant encounter because the inner body does not undergo eccentricity damping . in contrast to low @xmath7 , high eccentricities reduce @xmath54 at @xmath61 @xmath62 , and @xmath54 gradually increases with @xmath9 @xmath63 . this broadening of the capture probability curve at higher eccentricity is also seen in previous studies ( e.g. , @xcite ; @xcite ) . we find that the critical migration timescale can be sharply defined except when the eccentricity is not small @xmath64 at the resonant encounter . in the following subsections , the value of @xmath57 is explicitly described only when the transition is not sharp enough . the eccentricity dependence of @xmath54 is summarized in section [ sec : eini ] . we performed @xmath65 runs for each set of parameters ( @xmath66 runs in total ) . we first examine the dependence of the critical migration timescale on the mass of the larger body @xmath3 . figure [ fig : m1_ta ] illustrates how @xmath1 varies with @xmath3 . open squares ( connected by a solid line ) represent the fiducial case , where @xmath67 , and @xmath68 with different values of @xmath3 . filled squares , open circles , filled circles , and open triangles show the results for @xmath69 and @xmath70 respectively , but the other parameters are the same as for the fiducial case . note that the open squares and open circles overlap at every @xmath3 . the dependence on @xmath71 is examined between @xmath72 and @xmath73 . in order to keep computational cost reasonable , we also put an upper limit of @xmath74 . for @xmath75 , the eccentricity of the inner body is @xmath76 at the resonant encounter and the capture probability curve is broadened ( @xmath77@xmath78 ) . this survey reveals two features . first , all of the @xmath1 values except that for @xmath79 are within about a factor of two of each other . this suggests that @xmath1 depends weakly on @xmath80 and @xmath7 , which can be seen in the following subsections in more detail . second , all the cases exhibit a similar power law dependence of @xmath1 on @xmath3 . the gradients are calculated for each case by least squares fits , and we obtain a typical power index approximately equal to @xmath0 . a physical interpretation for this dependence is discussed in section [ sec : discussion ] . next , we explore the effect of varying @xmath81 note that the dependence can be seen more clearly when @xmath31 is used instead of @xmath4 . figure [ fig : m2_ta ] shows @xmath1 as a function of @xmath81 open squares represent the fiducial case . filled squares , open circles , filled circles , and open triangles show the results for @xmath82 , and @xmath70 respectively . same as the previous subsection , the capture probability curve is not sharp for @xmath75 ( @xmath83@xmath84 ) . we find that if @xmath85 @xmath1 is independent of @xmath31 ; therefore , the restricted three body approximation is valid for this condition . the difference between @xmath1 for @xmath86 and the value for small @xmath31 is found to be a factor of about 10 . thus , although the restricted three body approach can not provide an accurate prediction of @xmath1 for bodies with comparable masses , it would be possible to roughly derive @xmath1 to an accuracy of a factor of 10 . see also discussions in sections [ sec : internal_resonance ] and [ sec : compare ] for the reason for the difference . note that we consider only the migration of the outer planet . because migration speed depends on planetary mass , the inner planet s migration is not negligible , especially for @xmath87 . however , if we apply the differential migration speed instead of the migration speed of the outer planet , the critical migration timescales we obtain are valid even considering the migration of both planets . figure [ fig : te1_ta ] shows the results for various eccentricity damping timescales for the inner ( larger ) planet , @xmath20 . again , the open squares represent the fiducial case . filled squares , open circles , filled circles , and open triangles show the results for @xmath88 , and @xmath70 respectively . the rightmost points on the horizontal axis are the cases without eccentricity damping . for @xmath75 and @xmath24 , the transition from a capture probability of zero to one is not very sharp ( @xmath77 ) . there is no systematic change in @xmath1 with @xmath89 even when @xmath20 is varied by more than four orders of magnitude , the differences in @xmath1 lie within a factor of two or three . figure [ fig : te2_ta ] shows the results for various eccentricity damping timescales for the outer ( smaller ) planet , @xmath21 . open squares represent the fiducial case . filled squares , open circles , filled circles , and open triangles are the results for @xmath90 , and @xmath70 respectively . for @xmath75 , the transition in the capture protability is not sharp ( @xmath77@xmath91 ) . we also do not see any clear trends in @xmath92 although the case with @xmath93 ( open triangles ) shows a slight variation , it still remains within a factor of a few . finally , figure [ fig : e_ta ] shows the results for various initial eccentricities @xmath7 . open squares represent the fiducial case . filled squares , open circles , filled circles , and open triangles are the results for @xmath90 , and @xmath94 respectively . when the eccentricity is not small @xmath64 at the resonant encounter , in other words , when @xmath95 and @xmath96 , the capture probability curve is not sharp the same as before . again , no clear trends in @xmath1 with @xmath7 are recognized . note that , for @xmath97 and @xmath79 ( open circles ) and for @xmath98 ( open triangles ) , @xmath1 can not be determined to have particular values . because the eccentricity of the smaller body @xmath4 at a resonant encounter is relatively large , adequate resonant capture does not occur even for long @xmath9 ; the capture probability does not exceed about 0.3 . except for large @xmath7 , the dependence is reasonably weak . several of these features are in agreement with previous studies , which will be discussed in section [ sec : compare ] . the critical migration timescales derived in the previous subsections are summarized as follows : @xmath99 where the dependence on the stellar mass @xmath44 is included . as shown in the results , @xmath1 can differ by a factor of two or three . although we assume that the inner body is more massive than the outer body @xmath100 , the results are almost the same when the outer one is more massive , which will be seen in section [ sec : internal_resonance ] . the formula is valid if the eccentricities of the smaller planets at a resonant encounter are much smaller than 0.1 . the capture condition for 2:1 mean motion resonances was examined above . we performed additional simulations to evaluate the critical migration timescale for closely spaced first order resonances ( e.g. , 3:2 and 4:3 ) because several exoplanet systems have such commensurabilities . figure [ fig : p_ta ] shows the critical migration timescales for @xmath2 resonances ( @xmath101 ) obtained by numerical simulations . here , we define @xmath102a_1,\ ] ] which roughly corresponds to the orbital separation from the inner planet to the outer planet in the mean motion resonance . open squares represent the results for @xmath103 , whereas filled squares represent those for equal mass bodies ( @xmath86 ) . in each case , the other parameters are set to the fiducial values @xmath104 . we see that the critical migration time decreases with decreasing separation . the critical migration timescale for equal mass bodies ( @xmath105 ) is always shorter than that with bodies that have a high mass ratio ( @xmath106 ) . the difference in @xmath1 is larger at 2:1 resonances than at closely spaced resonances . this is presumably because the strength of the 2:1 external resonance , where `` external '' means that the inner body is the dominant body , is significantly weakened by indirect terms in the disturbing function , which will also be discussed in section [ sec : internal_resonance ] . cccccc mass ratio & 2:1 & 3:2 & 4:3 & 5:4 & 6:5 + @xmath107 & @xmath108 & @xmath109 & @xmath110 & @xmath111 & @xmath111 + @xmath112 & @xmath113 & @xmath114 & @xmath109 & @xmath110 & @xmath110 [ tbl : fitting ] the combined results from the above surveys are plotted in figure [ fig : dis2_m1_ta ] . the solid lines indicate the critical migration timescale for capture into the 2:1 resonance as a function of @xmath3 , which is summarized in equation ( [ eq : ta_crit ] ) . squares , triangles , and inverted triangles are the numerical results for the 3:2 , 4:3 , and 5:4 resonances , respectively . from the results , the critical migration timescale is described as @xmath115 where @xmath116 depends on @xmath117 and the resonant commensurability . we derive the @xmath116 values from the fitting of the results and summarize @xmath116 in table [ tbl : fitting ] . note that although @xmath116 should depend on commensurability , the values are the same between the 5:4 and 6:5 resonances , which suggests that the difference is within the minimum interval of @xmath9 that we set for our investigation . the dashed lines in figure [ fig : dis2_m1_ta ] represent fitting results for equal mass cases @xmath118 . although the approximate fits for closer resonances are considered to be correct within a factor of a few , they should be treated with some caution . we observe that when @xmath119 and @xmath120 , the bodies undergo close encounters before being captured into resonances ; therefore , the bodies do not settle into stable resonant orbits , as indicated by the open symbols in figure [ fig : dis2_m1_ta ] . according to the analysis of the hill stability by @xcite , dynamical stability is almost guaranteed if @xmath121 , where @xmath122 is the mutual hill radius . note that if the orbital separation is within a few tens of percent beyond the critical hill separation , hill stable planetary systems may manifest lagrange instability when the outer planet escapes to infinity ( @xcite ; @xcite ) . there is no analytical criteria that describe lagrange stability , and additional numerical simulation is required to determine whether the system is lagrange stable or not . however , @xcite found that the hill stable condition is a good predictor of lagrange stability . the hill stable condition is rewritten as @xmath123 ^ 3 m_*. \label{eq : critical_mass}\ ] ] the critical masses for equal mass bodies assuming @xmath47 are plotted with crosses connected with by a solid line in figure [ fig : dis2_m1_ta ] . if the mass of the body is larger than the critical mass , the system becomes hill unstable , which is consistent with our results that exhibit close encounters ( open symbols ) . note that even if @xmath124 , the orbit in a resonance can become stable for specific orbital arguments ( some examples are provided in section [ sec : constraints ] ) . this criterion for the instability is a rough estimate , and long term orbital integration can reveal the instability time of planets in closely spaced mean motion resonances . so far , we have carried out simulations in the cases where the inner body is more massive or equal to the outer body . in order to evaluate the difference in @xmath1 between the internal and external resonances , additional simulations for outer massive body are performed . assuming @xmath125 , @xmath126 , @xmath127 , @xmath128 , and @xmath26 , the critical migration timescales are determined for 2:1 , 3:2 , 4:3 , 5:4 , and 6:5 resonances , which are plotted in figure [ fig : p_ta ] with crosses . we find that no significant difference between the internal and external resonances ; in fact , it lies within a factor of two . therefore , as stated in the last sentence of section [ sec : model ] , if the outer body is larger than the inner body , we apply the mass of the larger body to @xmath3 ; equation ( [ eq : ta_gen ] ) is then valid . although the difference is not significant , we see a decrease in @xmath1 at the 2:1 internal resonance , which means that the 2:1 internal resonance is stronger than the 2:1 external resonance . this tendency can be understood in terms of contributions of each term in the disturbing function . for the 2:1 external resonance , the contribution of the direct term is diminished by the indirect term , leading to weakening the strength of the resonance ( @xcite ; @xcite ) . one indicator of the strength of the resonance is @xmath129 in the work of @xcite ( see table 1 in their work for each value ) , where the strength of resonance decreases with increasing @xmath129 . we see that @xmath129 for the 2:1 external resonance is large . the fact that the 2:1 internal resonance is stronger than the 2:1 external resonance would also partially explain the decrease in @xmath130 for equal mass bodies in figures [ fig : m2_ta ] and [ fig : p_ta ] although this can not account for the entire change . we also see in figure [ fig : p_ta ] that internal resonances are slightly weaker than external resonances for closely spaced resonances ( e.g. , 3:2 ) , which is consistent with figure 11 in the work of @xcite . we discuss the numerical results using analytical arguments . through numerical investigations , we find that the critical migration timescale shows a power law behavior , @xmath131 , which was also seen in previous studies ( e.g. , @xcite ; @xcite ) . this tendency can be estimated using a simple pendulum model @xcite ; the following expressions describe the orbital properties of a massless particle that is in a mean motion resonance . note that these expressions are for the internal resonance ; however , those for the external resonance are almost the same . for the circular restricted problem , where the massless test particle is in a first order @xmath2 resonance with a body with mass of @xmath132 , the maximum width of libration is given by @xmath133^{1/2 } a , \label{eq : delta_res}\end{aligned}\ ] ] where @xmath134 is the excited eccentricity due to resonant perturbation , and @xmath135 and @xmath10 are the mean motion and semimajor axis of the test particle , respectively . the constant arising from the resonant term of the disturbing function is @xmath136 where @xmath137 , and @xmath138 are the ratio of semimajor axis of the inner body to that of the outer body , the laplace coefficient , and the derivative operator , respectively . for the first order resonance , @xmath139 is always negative . the libration timescale is derived as @xmath140 then the excited eccentricity during the resonant passage is approximated using an adiabatic invariant as @xmath141 as in @xcite . substituting this equation into equations ( [ eq : delta_res ] ) and ( [ eq : t_lib ] ) , we obtain @xmath142 @xmath143 @xmath144 comparing the libration timescale @xmath145 and the migration timescale through the resonant width @xmath146 , where @xmath147 we determine that the critical migration timescale can be roughly given by @xmath148 we confirm the power law dependence on the mass . in addition , given that @xmath149 , @xmath150 , and @xmath151 @xcite , @xmath1 for an earth mass body is @xmath152 . these estimates are roughly consistent with our numerical results ; note , however , that they are only order of magnitude estimates assuming the restricted problem . the contribution from the variation in the longitude of the pericenter @xmath153 is also neglected . we here discuss our results by comparing with previous studies . first , our results for two bodies that have a high mass ratio can be compared to the works that consider the restricted problem ( e.g. , @xcite ; @xcite ) . we find that several features in figure [ fig : e_ta ] , in which the dependence of @xmath1 on the eccentricity is investigated , are in agreement with the results of @xcite ( figure 2 in their paper ) as follows . @xcite showed that when the generalized momentum is small enough , which corresponds to small @xmath19 , the capture probability curve is steep and @xmath1 hardly depends on @xmath19 . in figure [ fig : e_ta ] , we observe that @xmath1 is well defined and almost independent of @xmath19 when @xmath19 at the resonant encounter is small ( @xmath154 for @xmath155 ) . in this case , the value of @xmath1 is also comparable to that derived by @xcite ; that is , their estimate for @xmath1 is @xmath156 for @xmath155 . we also see another feature of steepening dependence on @xmath19 at higher @xmath19 . according to @xcite , @xmath1 tends to sensitively depend on @xmath19 when @xmath19 is larger than @xmath157 and @xmath158 for @xmath155 and @xmath159 , respectively , which are seen in figure [ fig : e_ta ] in the models of @xmath128 and @xmath160 . the other feature is that the capture probability has a cutoff at high eccentricity and never exceeds about 0.3 in the case of @xmath128 . in our results , the cutoff eccentricity is @xmath161 , which quantitatively reproduces the results of @xcite . in addition , several similar features are observed even in the results for equal mass bodies ; namely , the weak dependence of @xmath1 on @xmath19 and the existence of the cutoff in the capture probability at high @xmath19 , which were expected from the previous study @xcite . note , however , that the values of @xmath1 differ by a factor of 10 between the restricted problem and the unrestricted problem . even if we consider the difference between the internal resonance and the unrestricted problem , there is at least a factor of a few difference in @xmath1 special care would have to be taken when one quantitatively discuss the critical migration timescale for equal mass bodies . several properties observed in our simulations were also seen in a dynamical study by @xcite . figure 2 of @xcite indicates the excluded resonant regions for the hd 200964 system ; figure 2(a ) shows their results for two planets of comparable mass , whereas figure 2(c ) shows the case where one body is a test particle . they found that the allowed region for the comparable mass case is wider than that for the zero mass particle , which suggests that planets with comparable masses are more easily captured into mean motion resonances . in our model , we find that the critical migration timescale for equal mass planets is shorter than that for systems with high mass ratios . in addition , we also see a weak dependence of the critical migration timescale on the @xmath19damping timescales . this tendency is also seen in figure 4 of @xcite . @xcite derived an analytical expression for the eccentricity of the outer planet captured into a resonance . the equilibrium eccentricity is attained by balancing pumping due to resonant effects with damping caused by migration , as described by equation ( 5 ) of @xcite . we confirmed that our numerical results are consistent with this analytical formula except for rapid migration @xmath162 and equal mass bodies . the results given in section [ sec : results ] can be used to constrain the history of systems in mean motion resonances and their formation scenarios . in this section , we first discuss exoplanet systems with resonances using the mass versus critical migration timescale diagram , and then move on to other systems . as of 2013 , more than 25 planetary systems that lie in or close to first order mean motion resonances have been detected ( e.g. , @xcite ) . table [ tbl : exoplanets ] lists the orbital properties of the confirmed planets that could be in first order resonances . the fourth and fifth columns show the masses and semimajor axes of the planets , respectively ; the sixth and seventh columns show the resonant properties . for example , the third row indicates that gliese 876 c could be in a 2:1 resonance with planet b , where the period ratio of these planets is 2.03 . the data are taken from the open exoplanet catalogue . note that many systems listed have ratios more distant from commensurability than the 1% used to classify resonant captures in our simulations . although the period ratios are close to commensurate values , not all of the planets necessarily lie in mean motion resonances . the solid and dashed lines in figure [ fig : dis_m1_ta ] , which indicate the critical migration speeds , are the same as in figure [ fig : dis2_m1_ta ] . for the resonant pair exoplanets listed in table [ tbl : exoplanets ] , the critical migration timescale , plotted as open circles , is obtained from their resonant commensurability and the mass ratio between the smaller and larger bodies ( @xmath163 or not ) as a function of the larger planet mass . all pairs except for the pair of gliese 876 b and e have roughly equal masses @xmath164 . the outer planet migration timescale must be longer than the critical migration time for capture into mean motion resonance to occur . we obtained the critical migration timescale in section [ sec : results ] while ignoring the migration of inner planets . if the inner planets also migrate inward , the differential ( or relative ) migration speed between the inner and outer planets should be applied to the critical migration speed . using figure [ fig : dis_m1_ta ] , we can place some constraints on the differential migration speeds that the exoplanets used to have . note that in the following discussions , we introduce several assumptions : ( i ) planets formed widely separated from the 2:1 resonance . ( ii ) the masses of planets do not evolve after the onset of migration . ( iii ) migration is smooth and not subjected to stochastic torques . ( iv ) the eccentricities of planets are low and hence the transition from certain capture to certain failure is well defined . ( v ) once captured in resonance , planets do not subsequently escape from the resonance . the 2:1 mean motion resonance is the outermost first order mean motion resonance . the capture probability for the outer , higher order mean motion resonance is very low . therefore , if a system has a 2:1 commensurability , the differential migration speed would have been slower than that estimated from the critical migration time . if a system has a 3:2 commensurability , the outer planet passed through the 2:1 mean motion resonance because of the high migration speed and was captured into the 3:2 mean motion resonance ; thus , the differential migration speeds is slower than the critical migration speed for the 3:2 mean motion resonance but faster than that for the 2:1 mean motion resonance . we can constrain the differential migration speed for capture in the other mean motion resonances in a similar manner . the pairs of exoplanets in the closely spaced mean motion resonances are expected to have migrated at significantly high speed . note that if two planets in a mean motion resonance formed with a small orbital separation , they may have been captured into closely spaced resonances under slow migration . in addition , if several planets simultaneously exhibit migration as a resonant convoy ( @xcite ; @xcite ) rather than the single outer planet migrating inward , the critical migration timescale can be somewhat longer than that derived in section [ sec : results ] . nonetheless , our approach provides useful constraints on formation models without the need for further dynamical analyses and calculations for individual systems . although the number of observed resonant exoplanets is too small to support a statistical discussion , we see in figure [ fig : dis_m1_ta ] that there is a general trend toward a decrease in the number of systems in closely spaced resonances with increasing mass . this trend can be understood in terms of the short critical migration timescale for capture into 2:1 resonances with high mass planets ; 1000earth mass planets can capture bodies into 2:1 resonances even if the migration timescale is quite short @xmath165 . as stated in section [ sec : closer_resonances ] , it is predicted that pairs with orbital separations smaller than @xmath166 hill radii can be hill unstable , and such configurations are rare . we note that the hill stability criterion , more properly bodies should be lagrange stable , gives only a sufficient condition for stability ; thus , if the planets are in mean motion resonances , the system can become stable . however , they should experience a configuration where planets are not near resonances at some point during migration phase ; therefore , it is possible that such planets undergo close encounters and are scattered away from the system . we find that almost all the pairs have separations larger than @xmath167 ; however , there are several exceptions . we discuss these systems in section [ sec : small_separation ] . on the other hand , several pairs have separations larger than @xmath168 hill radii , which is the typical orbital separation of planets formed via the oligarchic growth phase @xcite . the location of @xmath169 is also shown in figure [ fig : dis_m1_ta ] by crosses connected by a solid line , in the same way as for @xmath170 . thus , planets are believed to form in distant orbits , after which they migrate inward one by one , maintaining their orbital separations wider than @xmath171 @xcite . these systems include kepler-18 , kepler-48 , kepler-52 , kepler-53 , and kepler-57 , which all have 2:1 commensurabilities . figure [ fig : dis_m1_ta ] gives the lower limits on the migration timescale . for example , the kepler-18 system consists of two low density neptune mass planets ( c and d ) near a 2:1 resonance and an inner super earth ( b ) @xcite . the orbital separation of planets c and d is @xmath172 . the planets are believed to undergo migration with a differential migration speed slower than @xmath173 . next , we focus on super earth mass planets @xmath174 , which have a relatively large number of samples , and compare their histories with the migration theory . planets with a few tens of the earth mass or less might have experienced type i migration . through a linear calculation , the type i migration timescale is given by @xcite @xmath175 where @xmath176 denotes the surface density gradient . for optically thin disks , the temperature distribution is @xcite @xmath177 which determines the sound velocity @xmath178 . we scale the gas surface density as @xmath179 where @xmath180 is a scaling factor . if we introduce the type i migration efficiency @xmath181 to express the uncertainty in the type i migration theory , the type i migration timescale for @xmath182 and @xmath183 is written as @xmath184 the type i migration speed is still uncertain ) . ] . here , we consider a scenario in which the inner body is stationary and the outer body undergoes inward migration . then the type i migration timescale ( equation [ [ eq : typei ] ] ) , which is identical to the relative migration timescale between the two bodies , can be drawn as a function of @xmath185 ( dotted line in figure [ fig : dis_m1_ta ] ) . some of the exoplanets with masses of @xmath186 are in 2:1 resonances . as shown in figure [ fig : dis_m1_ta ] , they would not have undergone rapid migration , so @xmath187 . on the other hand , some planets are captured into closely spaced resonances . for them , a short migration time is required , and @xmath188 . in order to reconcile the low migration efficiency ( @xmath187 ) and the linear type i migration theory , the migration of inner planets should be considered . if the inner planets migrate when they are captured into mean motion resonances , the differential migration speeds become much slower than the migration rate estimated from equation ( [ eq : type1lin ] ) . this can explain the planets in the 2:1 mean motion resonances . on the other hand , the migration of inner planets is negligible if the inner planets are much smaller than the outer ones and/or if the inner planets have orbits around the inner edge of the disk . such pairs of planets can be in closely spaced resonance because of the high relative migration speed . in addition , the other possibility for a pileup of planets with masses of @xmath189 is that they are formed near the current resonance locations ( e.g. , 3:2 and 2:1 ) . the orbital separations for 10earth mass planets in 3:2 and 2:1 resonances are @xmath190 and @xmath191 , respectively . this is comparable to the typical separation after the oligarchic growth phase : planets are formed _ in situ _ and then migrate slightly and are captured in the resonances . as argued above , the formation of resonances between two planets can be discussed in terms of two cases of the orbital separation at the time migration begins ; namely , the planets are well separated from each other @xmath192 , or they have relatively close orbits @xmath193 . the orbital separation of the planets at the onset of migration is characterized by their migration and growth timescales . the growth timescale of a planet @xmath194 is determined by the accretion rate of surrounding planetesimals in the classical planet growth model , which is similar to or longer than the type i migration timescale at the time of migration ( e.g. , * ? ? ? * ; * ? ? ? * ) . in this case , the inner planet starts to migrate before the outer planet does , resulting in an expansion of the orbital separation @xmath192 . on the other hand , as planets grow , the surrounding planetesimals stirred by the planets are fragmented by mutual collisions . the resultant fragments effectively accrete onto planets @xcite ; @xmath194 depends on the initial planetesimal mass and radial gas density profile , and @xmath194 might be shorter than @xmath9 in some cases @xcite when migration begins . if @xmath195 , the two planets can start their migration almost simultaneously , in which case the orbital separation @xmath193 is maintained during migration . detailed calculations that include accretion , fragmentation , and migration are needed to clarify this behavior . below we discuss some individual systems . first , in this subsection the systems in closely spaced resonances ( e.g. , 4:3 and 5:4 ) are considered . as shown in figure [ fig : dis_m1_ta ] , the differences in @xmath1 between adjacent pairs are small ; for instance , the @xmath1 difference between the 4:3 and 5:4 resonances is only a factor of two . this means that the migration speeds for systems in closer resonances can be constrained well . the kepler-11 system , in which six planets have been confirmed , has a possible pair of planets in a 5:4 mean motion resonance ( planets b and c ) . the critical migration timescales for the 4:3 and 5:4 resonances are @xmath196 and @xmath197 , respectively . thus , if these planets are formed well separated from each other @xmath192 and then undergo convergent migration , the migration timescale would be a few thousand times @xmath27 . this migration speed is slightly higher than that of type i migration @xmath198 . because the other planets in the kepler-11 system are not considered in resonances , it may not be natural to suppose that the planets undergo significant migration . therefore , it seems likely that they are formed _ in situ _ and exhibit slight inward migration , which leads to capture into the 5:4 resonance . in addition , the required high type i migration speed is inconsistent with the typical migration speed which is predicted by the population synthesis model @xcite , which would also support the _ in situ _ formation . the orbital separation is slightly smaller than the typical orbital separation after oligarchic growth at 1 au @xmath199 ; however , it has been shown that planets can be formed with smaller orbital separations near the central star @xmath200 @xcite . it is also possible that the planets formed _ in situ _ and planet b exhibited outward migration due to the tidal torque from the central star , resulting in capture into the resonance . note that the difficulty of _ in situ _ accretion was pointed out in an investigation of the accretion and evolution of a hydrogen rich atmosphere for kepler-11 @xcite . another possibility is that the planets are formed in distant orbits and migrate inward as a resonant convoy , in which the planets can pass through 2:1 resonances as they are pushed inward by the outer bodies in the resonances @xcite . however , in such a case , it is likely that other planets are also in resonances in the final state . the kepler-60 system consists of three planets : the inner pair ( b and c ) has 5:4 commensurability , and the outer pair ( c and d ) seems to be in a 4:3 resonance . if we assume that these planets formed well separated and then migrated inward , the differential migration timescale would be approximately @xmath201 . using equation ( [ eq : typei ] ) , we find that the efficiency of type i migration multiplied by the scaling factor for the gas surface density is @xmath202 . this means that if the gas surface density is similar to that of the minimum mass solar nebula , the migration speed would be that predicted by the linear theory of type i migration @xcite . we also do not exclude the possibility of an _ in situ _ formation model for this system . next , we discuss resonant systems with small orbital separations @xmath203 , which can be hill unstable . in the hd 200964 system , two jovian mass planets lie in a 4:3 mean motion resonance ( e.g. , @xcite ; @xcite ) , and their orbital separation is @xmath204 . if we assume that these planets are stable despite their small separation , the differential migration timescale for passing through the 3:2 resonance would be extremely short @xmath205 . this migration speed is almost impossible to achieve because a large amount of angular momentum should be delivered to the disk . thus , it is not easy to provide a formation model that produces the orbital properties of the hd 200964 system . in fact , @xcite have also claimed that no formation scenarios are successful in reproducing 4:3 resonant planets similar to those in the hd 200964 system . because the pair of planets around hd 200964 have almost crossing orbits , they might be in temporary resonance following orbital instability . the hr 8799 system has inner and outer debris disks @xcite and four planets that were discovered by direct imaging @xcite . the masses have not been well constrained ; the planets have masses between a few and 13 jupiter masses ( e.g. , @xcite ) . because the orbital separations of the second innermost pair ( planets d and c ) and the outermost pair ( planets c and b ) are small @xmath206 , it is plausible that the two pairs of planets are in 2:1 mean motion resonances , which can stabilize the system over the estimated age of the star @xcite . in this case , our model gives a lower limit on the migration timescale of @xmath207 . this value is quite small , so it may not provide a useful constraint . in addition to this lower limit , the time of passage of the hill unstable region should be shorter than the orbital unstable time for a system with small orbital separation @xmath208 . according to @xcite , the crossing time , which is the timescale of the initiation of orbital instability , is @xmath209 ; therefore , the actual migration timescale would be between @xmath210 and @xmath196 . the type ii migration expected for such massive planets can not bring about the short migration timescale . @xcite pointed out that if the masses of the planets are small , the system can be hill stable . gliese 876 , an m dwarf star with a mass of @xmath211 , harbors four planets : planets c and b have comparable masses , and planets d and e are relatively small , with masses of @xmath186 . planets c b and b e are in or close to the 2:1 resonance . the orbital separations for the inner pair ( c b ) and the outer pair ( b e ) are @xmath212 and @xmath213 , respectively . these planets are believed to be in the 2:1 resonance to stabilize their orbits at such a small orbital separation ( e.g. , @xcite ) . the lower limits to the differential migration timescales are @xmath207 for the inner pair and @xmath214 for the outer pair , which is consistent with the migration timescale obtained from a hydrodynamical simulation of this system by @xcite . finally in this subsection , we add some comments on planets with small separations that are still hill stable . the hd 45364 system consists of two planets with masses comparable to those of saturn and jupiter @xcite , which seem to be in a 3:2 mean motion resonance with a separation of @xmath215 . this system should have undergone rapid migration to pass through the 2:1 resonance . the critical migration timescale for capture into the 2:1 resonance is given by @xmath216 ; thus , the differential migration timescale should be at least shorter than @xmath217 . @xcite also examined this system both numerically and analytically , and proposed that a relative migration timescale shorter than @xmath218 is needed to pass through the 2:1 resonance , which is consistent with our results . our model also provides a lower limit of @xmath219 on the migration timescale . the planets in this system would have undergone type iii migration . here we apply our results to resonant systems in the solar system . one good example is the galilean satellites around jupiter , where io europa and europa ganymede are in the 2:1 mean motion resonance . replacing the stellar mass @xmath220 with the jupiter mass , our model gives a lower limit of @xmath221 on the differential migration speed . this constraint is consistent with the result of _ n_body work on the formation of the galilean satellites ( equation [ 32 ] in @xcite ) . some trans neptunian objects are in mean motion resonance with neptune . the population of objects in the 3:2 resonance is much higher than that in the 2:1 resonance . the objects were captured into the mean motion resonances during outward migration of neptune . simulations by @xcite showed that the tendency could be explained by the migration speed of neptune . we apply our formula for the inward migration of the outer objects to resonant capture by an outwardly migrating neptune . the obtained migration timescale of @xmath222@xmath223myr is plausible for the high population of the 3:2 resonance , which is roughly consistent with @xcite . furthermore , this estimated migration timescale is also consistent with that obtained by @xcite of @xmath34@xmath223myr , which is derived from studying the proportion of tnos in the leading and trailing islands of the 2:1 resonance . neptune s outward migration , which is caused by interaction with the surrounding planetesimals , and other planets may not be as smooth as we assume @xcite . the application of our model to these objects should be done carefully . we can further discuss other planet formation models , which assume the establishment of mean motion resonances . in the grand tack model @xcite , saturn migrates faster than jupiter , which results in capture into a 3:2 mean motion resonance with jupiter . for capture into the 3:2 resonance , the differential migration timescale between the two planets should be @xmath224@xmath225 , which is smaller than the typical type i or ii migration timescales . @xcite considered high speed type iii migration , which is necessary to realize the grand tack scenario . we investigated capture into first order mean motion resonances in a system of two bodies undergoing damping of the eccentricity and semimajor axis using _ n_body integrations . in some of our calculations , we considered the case in which the mass of one body is negligible . in addition , we also studied systems with equal masses . in fact , orbital calculations were performed with a wide range of parameters ; we found that the critical migration timescale can be described using the mass ratio between the larger body and the central object , and depends weakly on the @xmath19damping timescale and initial eccentricity . the empirical formula is given by equation ( [ eq : ta_crit ] ) , where the critical migration timescale for equal mass bodies is about an order of magnitude shorter than that for systems with a massless particle . we also confirmed the power law dependence of the mass with index @xmath0 . this dependence is also supported by analytical arguments that compare the resonant libration timescale and the migration timescale . additional simulations of closely spaced resonances were run , and empirical fits to the results were derived . all the fitting formulae from our calculations are shown in figure [ fig : dis2_m1_ta ] and table [ tbl : fitting ] . the empirical formula we derived can constrain the relative migration speed in systems of two bodies undergoing convergent migration toward capture into mean motion resonances . this means that our model can be useful for understanding the origins of exoplanet systems in resonances . for systems in closely spaced mean motion resonances ( e.g. , kepler-11 , kepler-60 ) , the migration timescale can be well constrained . it is also possible that the planets formed _ in situ_. the systems in which the orbital separation is smaller than @xmath226 ( e.g. , hr 8799 , gliese 876 ) are believed to become stable owing to resonant effects . lower limits to the relative migration timescale were placed on several systems in 2:1 resonances . the origin of the hd 200964 system , which is in 4:3 resonance , remains unclear . our model also provides constraints on the migration timescale of systems other than exoplanets ( e.g. , the grand tack model of the solar system ) . furthermore , when the number of discovered exoplanets in resonances increases sufficiently in the future , the typical type i and ii migration timescales can be obtained using our results ; future observations will allow us to tackle this issue . note that if the eccentricity at a resonant encounter is large , the capture into resonance becomes probabilistic . in this case , our model can provide only a necessary condition . in this work , we considered first order mean motion resonances , which are certainly important for planets both inside and outside the solar system . in addition , higher order mean motion resonances ( e.g. , 3:1 and 5:2 ) can also be important for specific systems ( e.g. , @xcite ; @xcite ) ; therefore , it would be worth examining the conditions for capture into these resonances in future work . we thank to an anonymous referee for detailed helpful comments . we also thank s. inutsuka for helpful discussions . the numerical computations were conducted in part on the general purpose pc farm at the center for computational astrophysics , cfca , of national astronomical observatory of japan . this work was supported by a grant in aid for jsps fellows ( 23004841 ) . hk gratefully acknowledges the support of a grant in aid from mext ( 23103005 ) . barnes , r. , & greenberg , r. 2005 , , 647 , l163 baruteau , c. , & papaloizou , j. c. b. 2013 , , submitted cochran , w. d. et al . ( 2011 ) , 197 , 7 correia , a. c. m. , udry , s. , mayor , m. , benz , w. , bertaux , j .- l . , bouchy , f. , laskar , j. , lovis , c. , mordasini , c. , pepe , f. , & queloz , d. 2009 , , 496 , 521 fabrycky , d. c. , & murray - clay , r. a. 2010 , , 710 , 1408 friedland , l. 2001 , , 547 , l45 gladman , b. 1993 , icarus , 106 , 247 goldreich , p. 1965 , , 130 , 159 hayashi , c. 1981 , prog . suppl . , 70 , 35 henrard , j. 1982 , celest . , 27 , 3 ida , s. bryden , g. , lin , d. n. c. , & tanaka , h. 2000 , , 534 , 428 ida , s. & lin , d. n. c. 2008 , , 673 , 484 ikoma , m. , & hori , y. 2012 , , 753 , 66 johnson , j. a. , payne , m. , howard , a. w. , clubb , k. i. , ford , e. b. , bowler , b. p. , henry , g. w. , fischer , d. a. , marcy , g. w. , brewer , j. m. , schwab , c. , reffert , s. , & lowe , t. b. 2011 , , 141 , 16 kley , w. , peitz , j. , & bryden , g. 2004 , , 414 , 735 kobayashi , h. , ormel , c. w. , & ida , s. 2012 , , 756 , 70 kobayashi , h. , tanaka , h. , & krivov , a. v. 2011 , , 738 , 35 kobayashi , h. , tanaka , h. , krivov , a. v. , & inaba , s. 2010 , icarus , 209 , 836 kokubo , e. & ida , s. 1998 , icarus , 131 , 171 lissauer , j. j. et al . 2011 , , 197 , 8 levison , h. f. , morbidelli , a. , vanlaerhove , c. , gomes , r. , & tsiganis , k. 2008 , icarus , 196 , 258 makino , j. , & aarseth , s. j. 1992 , publ . jpn . , 44 , 141 marley , m. s. , saumon , d. , cushing , m. , ackerman , a. s. , fortney , j. j. , & freedman , r. 2012 , , 754 , 135 marois , c. , macintosh , b. , barman , t. , zuckerman , b. , song , i. , patience , j. , lafrenire , d. , & doyon , r. 2008 , science , 322 , 1348 marois , c. , zuckerman , b. , konopacky , q. m. , macintosh , b. , & barman , t. 2010 , , 468 , 1080 mcneil , d. , duncan , m. j. , & levison , h. f. 2005 , , 130 , 2884 marcy , g. w. , butler , r. p. , fischer , d. , vogt , s. s. , lissauer , j. j. , rivera , e. j. 2001 , , 556 , 296 mayor , m. , udry , s. , lovis , c. , pepe , f. , queloz , d. , benz , w. , bertaux , j. -l . , bouchy , f. , mordasini , c. , & segransan , d. 2009 , , 493 , 636 michtchenko , t. a. , & ferraz - mello , s. 2001 , icarus , 149 , 357 murray , c. d. , & dermott , s. f. 1999 , solar system dynamics ( cambridge : cambridge univ . press ) murray - clay , r. a. , & chiang , e. i. 2005 , , 619 , 623 mustill , a. j. , & wyatt , m. c. 2011 , , 413 , 554 nelson , r. p. , & papaloizou , j. c. b. 2002 , , 339 , 993 ogihara , m. , & ida , s. 2009 , , 699 , 824 ogihara , m. , & ida , s. 2012 , , 753 , 60 ogihara , m. , duncan , m. j. , & ida , s. 2010 , , 721 , 1184 paardekooper , s .- j . , baruteau , c. , & kley , w. 2011 , , 410 , 293 papaloizou , j. c. b. , & szuszkiewicz , e. 2005 , , 363 , 153 quillen , a. c. 2006 , , 365 , 1367 raymond , s. n. , barnes , r. , armitage , p. j. , & gorelick , n. 2008 , , 687 , l107 reidemeister , m. , krivov , a. v. , schmidt , t. o. b. , et al . 2009 , , 503 , 247 rein , h. 2012 , , 427 , l21 rein , h. , papaloizou , j. c. b. , & kley , w. 2010 , , 510 , a4 rein , h. , payne , m. j. , veras , d. , & ford , e. b. 2012 , , 426 , 187 snellgrove , m. d. , papaloizou , j. c. b. , & nelson , r. p. 2001 , , 374 , 1092 steffen , j. h. 2013 , , submitted steffen , j. h. et al . 2013 , , 428 , 1077 szuszkiewicz , e. , & podlewska - gaca , e. 2012 , orig . of life and evol . of the biosph . , 42 , 113 origins of life and evolution of the biosphere , 42 , 113 tanaka , h. , takeuchi , t. , & ward , w. r. 2002 , , 565 , 1257 terquem , c. , & papaloizou , j. c. b. 2007 , , 654 , 1110 veras , d. , & mustill , a. j. 2013 , , in press . walsh , k. j. , morbidelli , a. , raymond , s. n. , obrien , d. p. , mandell , a. m. 2011 , , 475 , 206 weidenschilling , s. j. , & davis , d. r. 1985 , icarus , 62 , 16 wisdom , j. 1980 , , 85 , 8 wittenmyer , r. a. , horner , j. , & tinney , c. g. 2012 , , 761 , 165 wyatt , m. c. 2003 , 598 , 1321 zhou , j .- l . , & lin , d. n. c , 2007 , , 666 , 447 lllllll star & @xmath227 & planet & @xmath228 & @xmath10 ( au ) & mmr [ pair ] & period ratio + gliese 876 & 0.334 & d & 6.68 & 0.0208 & & + & & c & 227 & 0.130 & 2:1 [ b ] & 2.03 + & & b & 723 & 0.208 & 2:1 [ e ] & 2.03 + & & e & 14.6 & 0.334 & & + hd 37124 & 0.83 & b & 215 & 0.534 & & + & & c & 207 & 1.71 & 2:1 [ d ] & 2.10 + & & d & 221 & 2.81 & & + hd 73526 & 1.08 & b & 922 & 0.66 & 2:1 [ c ] & 2.00 + & & c & 795 & 1.05 & & + hd 82943 & 1.18 & c & 639 & 0.746 & 2:1 [ b ] & 2.01 + & & b & 556 & 1.19 & & + hd 128311 & 0.84 & b & 693 & 1.10 & 2:1 [ c ] & 2.05 + & & c & 1020 & 1.76 & & + hr 8799 & 1.56 & e & 2861 & 15 & & + & & d & 3179 & 27 & 2:1 [ c ] & 2.00 + & & c & 3179 & 43 & 2:1 [ b ] & 1.99 + & & b & 2225 & 68 & & + kepler-9 & 1 & d & 6.99 & 0.0273 & & + & & b & 80.1 & 0.14 & 2:1 [ c ] & 2.02 + & & c & 54.4 & 0.225 & & + kepler-18 & 0.972 & b & 6.90 & 0.0447 & & + & & c & 17.2 & 0.0752 & 2:1 [ d ] & 1.94 + & & d & 16.5 & 0.117 & & + kepler-48 & 0.89 & b & 4.79 & & 2:1 [ c ] & 2.02 + & & c & 10.6 & & & + kepler-51 & 1.00 & b & 55.9 & & 2:1 [ c ] & 1.89 + & & c & 36.2 & & & + kepler-52 & 0.54 & b & 4.61 & & 2:1 [ c ] & 2.08 + & & c & 3.51 & & & + kepler-53 & 0.98 & b & 8.90 & & 2:1 [ c ] & 2.07 + & & c & 10.8 & & & + kepler-56 & 1.37 & b & 16.0 & & 2:1 [ c ] & 2.04 + & & c & 69.7 & & & + kepler-57 & 0.83 & b & 5.03 & & 2:1 [ c ] & 2.03 + & & c & 2.47 & & & + mu ara & 1.08 & c & 10.6 & 0.0909 & & + & & d & 166 & 0.921 & 2:1 [ b ] & 2.07 + & & b & 533 & 1.5 & & + & & e & 577 & 5.235 & & + 24 sex & 1.54 & b & 633 & 1.33 & 2:1 [ c ] & 1.95 + & & c & 273 & 2.08 & & + hd 45364 & 0.82 & b & 59.5 & 0.681 & 3:2 [ c ] & 1.51 + & & c & 209 & 0.897 & & + kepler-49 & 0.55 & b & 7.86 & & 3:2 [ c ] & 1.51 + & & c & 6.88 & & & + kepler-54 & 0.51 & b & 4.61 & & 3:2 [ c ] & 1.51 + & & c & 1.53 & & & + kepler-55 & 0.62 & b & 6.23 & & 3:2 [ c ] & 1.51 + & & c & 5.12 & & & + kepler-58 & 0.95 & b & 8.22 & & 3:2 [ c ] & 1.52 + & & c & 8.71 & & & + kepler-59 & 1.04 & b & 1.22 & & 3:2 [ c ] & 1.51 + & & c & 4.08 & & & + hd 200964 & 1.44 & b & 588 & 1.60 & 4:3 [ c ] & 1.34 + & & c & 286 & 1.95 & & + kepler-60 & 1.11 & b & 5.46 & & 5:4 [ c ] & 1.25 + & & c & 6.44 & & 4:3 [ d ] & 1.33 + & & d & 6.88 & & & + kepler-11 & 0.961 & b & 4.30 & 0.091 & 5:4 [ c ] & 1.26 + & & c & 13.5 & 0.106 & & + & & d & 6.10 & 0.159 & & + & & e & 8.40 & 0.194 & & + & & f & 2.30 & 0.25 & & + & & g & 302 & 0.462 & & + kepler-50 & 1.24 & b & 5.07 & 0.077 & 6:5 [ c ] & 1.20 + & & c & 8.28 & 0.087 & & [ tbl : exoplanets ]

a leading - power factorization formula for weight functions relating the photon spectrum to arbitrary partial decay rates in is derived . these weight functions are independent of the hadronic shape function and allow for the determination of in a model - independent way . we calculate the weight function in renormalization - group improved perturbation theory to complete next - to - next - to leading order at the jet scale and to next - to leading order at the hard scale . first - order power corrections are also included , where a model - dependence enters via the appearance of subleading hadronic shape functions . mit - ctp 3695 + hep - ph/0511098 + november 8 , 2005 * shape - function independent relations of charmless inclusive-decay spectra * bjrn o. lange +

hep-ph0511098

the determination of the cabibbo - kobayashi - maskawa ( ckm ) matrix element @xmath2 from inclusive semileptonic @xmath6 decays requires theoretical predictions for partial decay rates , which are then compared to their experimentally measured values . because of a dominating @xmath7 background in a large portion of phase space , this procedure is adopted for a variety of restricted regions in phase space as obtained by , e.g. , accepting only events with one or more of the following features : charged - lepton energy @xmath8 , hadronic invariant mass @xmath9 , leptonic invariant mass @xmath10 , hadronic @xmath11 . here , @xmath12 , where @xmath13 denotes the energy and @xmath14 the three - momentum of the final hadronic state in the @xmath5-meson rest frame . for many of these cuts a hierarchy of energy scales exists in the decay process , for example @xmath15 , and shape - function effects become important @xcite . the theoretical expressions for differential decay rates in this region of phase space factorize into hard functions at the scale @xmath16 , and the convolution of jet functions and shape functions at the scale @xmath17 @xcite . while the jet functions are perturbatively calculable , shape functions are non - perturbative objects that capture all strong - interaction effects below the scale @xmath18 . at leading power the jet function @xmath19 and shape function @xmath20 are universal , and enter the qcd - factorization theorems for both the triple differential decay rate in @xmath1 decays and in the @xmath0 normalized photon spectrum @xcite . one strategy for the inclusive determination of @xmath2 is to use the @xmath0 photon spectrum to extract the leading shape function @xmath20 , which then allows for the calculation of arbitrary semileptonic partial decay rates @xcite . it was shown in that reference that the factorization approach can be applied to all commonly used kinematic cuts . in practice this program is realized by adopting a parameterizable model for the shape function , and fitting the parameters using information of the measured photon spectrum . it was emphasized in @xcite that such a model is only acceptable if the values for parameters are stable when fitted to different aspects of the photon spectrum , such as moments of it , or its functional form ( provided that the resolution is coarse enough to smear out hadronic resonances ) . with improving data on the photon spectrum such an approach might require further and further refinement of the models used . a different strategy is to eliminate the necessity for the extraction of the leading shape function , and to use the experimental data directly . such ideas have been investigated previously in @xcite for some specific cuts , where partial rates in semileptonic decays are expressed as weighted integrals over the @xmath21 photon spectrum , @xmath22 in the most recent analysis of this type , @xcite , the cut was chosen as @xmath11 for simplicity , and the weight function @xmath23 was calculated at complete two - loop order at the intermediate scale @xmath18 . the second term on the right - hand side of this equation denotes a residual hadronic power correction ( rhc ) , which was absorbed into the weight function in that reference . in the above relation @xmath24 denotes the partial semileptonic decay rate and @xmath25 the normalized photon spectrum in radiative decays , where @xmath26 , @xmath27 is the photon energy in the @xmath5-meson rest frame , and the total decay rate @xmath28 is defined to include all events with @xmath29 . it is beneficial to use the normalized photon spectrum instead of the absolute spectrum , because the weight function as defined in ( [ eq : sffree ] ) is independent of @xmath30 , possesses a well - behaved perturbative expansion @xcite , and because the normalized photon spectrum can be determined with better accuracy than the absolute one @xcite . in order to determine @xmath2 from relation ( [ eq : sffree ] ) both @xmath24 and the normalized photon spectrum enter as experimental input , while the weight function @xmath31 and the residual hadronic correction @xmath32 are theoretical quantities . in this paper , we extend the technology developed in @xcite to derive the weight function for an _ arbitrary _ kinematic cut that includes events with @xmath33 and @xmath34 . in general , the weight function @xmath31 ( and the correction @xmath32 ) , as well as the integration limit @xmath35 in ( [ eq : sffree ] ) depend on the particular cut that is used . for example , if we consider a cut on the charged - lepton energy @xmath36 then @xmath37 , or if a cut on hadronic invariant mass @xmath9 is considered then one needs @xmath38 . a quick way to obtain some intuition on how @xmath31 and @xmath35 depend on the specific cut qualitatively is to consider the @xmath1 phase - space depicted in figure [ fig : pspace ] . note that @xmath39 in radiative decays , so that the phase - space of @xmath40 consists only of the vertical boundary on the right - hand side of the plot . this picture explains correctly how the maximal @xmath41 depends on the nature of the cut , and where to expect some kinks in @xmath31 , for example for a combined cut on @xmath42 and @xmath43 . in the next section we will derive an expression for the weight function at leading power , which follows from exact factorization theorems for differential rates in inclusive @xmath5 decays in the shape - function region . the formula for the weight function is thereby also valid to all orders in perturbation theory . we compute its explicit form to next - to - next - to leading order ( nnlo ) at the intermediate scale @xmath18 and to next - to leading order ( nlo ) at the hard scale @xmath44 in renormalization - group ( rg ) improved perturbation theory , including three - loop running effects . this approximation is sufficiently precise for phenomenological applications since effects at subleading power become as important as higher - order perturbative effects at leading power . in section [ sec : powcorr ] we discuss first - order power corrections . kinematical power corrections enter the weight function , while hadronic power corrections give rise the quantity @xmath32 in ( [ eq : sffree ] ) . we then apply our results to a few examples of kinematic cuts in section [ sec : examples ] and perform an analysis of theoretical uncertainties on @xmath2 by using a simple model for the experimental inputs . in this section we will adopt the leading - power approximation . the second term on the right - hand side of relation ( [ eq : sffree ] ) is then absent , because it collects contributions from subleading shape functions to both radiative and semileptonic decay rates , which start at order @xmath45 . we start the discussion by stating the exact factorization theorems for the fully differential leading - power decay rate in @xmath6 decays @xmath46 and for the normalized @xmath0 photon spectrum @xmath47 the superscript @xmath48 indicates that these equations are valid to leading power . in ( [ eq : xulnu ] ) @xmath49 denotes the hard function , which collects all matching corrections at the scale @xmath44 and depends on the kinematic variables @xmath50 for which the phase - space is @xmath51 . the jet function @xmath52 contains distributions that act on the shape function @xmath53 and depends on @xmath54 via @xmath55 . in formula ( [ eq : xsgamma ] ) for the photon spectrum , on the other hand , the hard function is denoted by @xmath56 , and the argument of the jet function is @xmath57 , independent of @xmath54 because @xmath39 in radiative decays . three powers of the @xmath58-quark mass appear due to phase - space integrations in the total rate @xmath28 . renormalization - group ( rg ) running effects between the scales @xmath44 and @xmath18 build up the functions @xmath59 and @xmath60 in the factorization formulas , which are such that @xmath61 and @xmath62 in the limit @xmath63 . expressions for them will be given below . we start the derivation of the weight function by considering the factorized expression for an arbitrary partial differential decay rate @xmath64}dy \int\limits_0^{\e_{\rm max}[p_+,y ] } d\e \ ; \frac{d^3 \gamma_u^{(0)}}{dp_+ \,dy \,d\e } \;.\ ] ] the integration limits @xmath65 $ ] and @xmath66 $ ] , as well as @xmath35 depend on the specifics of the cut . ( the use of squared brackets for these quantities is supposed to remind the reader that these functions are defined differently for different kinematic cuts . ) we are now going to rewrite this semileptonic decay rate such that it resembles a weighted integral over the normalized photon spectrum . clearly the biggest obstacle is that the argument of the jet function @xmath19 in ( [ eq : xulnu ] ) depends on the kinematic variable @xmath54 , while it does not in ( [ eq : xsgamma ] ) . the solution to this problem has been presented in @xcite , where it was shownthe quantity considered in that reference was @xmath67 . ] that a factorization of the integrated jet function in the sense that @xmath68 can be achieved for arbitrary @xmath69 , if @xmath70 is allowed to be a distribution in the variable @xmath71 . here , @xmath72 we will discuss the nature of @xmath70 and its perturbative expansion to two - loop order in the next section . the strategy for the derivation of the weight function is to interchange integrations in ( [ eq : partialgamma ] ) so that the @xmath73 integration acts only on the jet function @xmath19 , and we can make use of ( [ eq : ydef ] ) . before this can be done , however , it is necessary to transform the other @xmath73-dependent terms in the partial decay rate into @xmath73-independent ones . this can be achieved by inserting @xmath74 into ( [ eq : partialgamma ] ) , and by replacing @xmath75 in @xmath65 $ ] , @xmath76 $ ] and in the kinematic prefactor @xmath77 . it is now an easy exercise to interchange the integrations @xmath78 next , we apply ( [ eq : ydef ] ) , interchange the @xmath71 and @xmath73 integrations , and undo the steps in ( [ eq : intchanges ] ) . finally the integrations over @xmath79 and @xmath80 can be carried out , which identifies @xmath81 . as a result we arrive at an expression for the semileptonic decay rate , which is given as an integral over the product of the normalized photon spectrum in ( [ eq : xsgamma ] ) and @xmath82 where @xmath83}dy \int\limits_0^{\e_{\rm max}[p',y ] } d\e\ ; y^{-2a_\gamma(\mu_h,\mu_i ) } h_u(y,\e,\mu_h)\ , y(k,\ln y,\mu_i)\;.\ ] ] the above two formulas enable us to calculate the weight function in an automated fashion . the procedure is as follows : first , the integration limits @xmath35 , @xmath65 $ ] , and @xmath84 $ ] are specified from the kinematics of the cut . for the next step it is helpful to decompose @xmath85 , where @xmath86 are distributions in @xmath71 and independent of @xmath54 . ( for example , we will see below that there are only three different distributions in the perturbative expansion of @xmath70 to two - loop order . ) likewise , we decompose @xmath87 . it is then straight - forward to calculate the functions @xmath88 by integrating over @xmath89 and @xmath54 in ( [ eq : fdef ] ) . finally the integration over @xmath71 is performed in ( [ eq : wresult ] ) , where the distributions @xmath90 act on @xmath91 . equation ( [ eq : wresult ] ) together with ( [ eq : ydef ] ) is the main result of this paper . we stress that both formulas are exact factorization theorems ( in the convolution sense ) , valid to all orders in perturbation theory , and ( [ eq : wresult ] ) gives the exact leading - power weight function for arbitrary cuts . the computation of the kernel @xmath92 in ( [ eq : ydef ] ) requires the knowledge of the jet integral @xmath93 . the jet function @xmath52 has been calculated to one - loop order explicitly @xcite , and the non - constant part of @xmath93 at two - loop order can be extracted from rg evolution @xcite . because the jet integral is of central importance to the present work , we find it legitimate to re - derive its dependence on @xmath94 in detail . in particular , from the rg equations for the shape function and the leading - power current in soft - collinear effective theory ( scet , see also @xcite ) , @xmath95 \delta(\hat\omega-\hat\omega ' ) - 2 \gamma_c(\mu ) \left ( \frac{1}{\hat\omega-\hat\omega ' } \right)_*^{[\mu ] } \right\ } \hat s(\hat\omega,\mu ) \;,\\ 0 & = & \left[\frac{d}{d\ln\mu } + 2\gamma'(\mu ) + 2 \gamma_c(\mu ) \ln\frac{m_b}{\mu } \right ] \int_0^{p_+ } d\hat\omega\ ; m_bj(m_b(p_+-\hat\omega),\mu ) \ ; \hat s(\hat\omega,\mu)\;,\nonumber\end{aligned}\ ] ] we can derive the rg equation governing @xmath93 with @xmath96 and arbitrary @xmath97 . above , @xmath98 is the cusp anomalous dimension @xcite , which has been calculated to three - loop order @xcite , the remaining anomalous dimensions @xmath99 and @xmath100 are known to two - loop order @xcite . ( for a definition of the star distribution that acts on the shape function see ( [ stardistris ] ) below . ) when combining the two rg equations , one finds @xmath101 j(l,\mu ) - 2 \gamma_c \int_0 ^ 1 \frac{dz}{z } \big[j(l+\ln(1-z),\mu ) - j(l,\mu ) \big],\ ] ] with @xmath102 . this integro - differential equation can be solved perturbatively by choosing a polynomial ansatz for @xmath93 . as a result the integral in the above equation leads to the appearance of the riemann zeta - function @xmath103 . specifically , after expanding the qcd @xmath104-function and anomalous dimensions as @xmath105 and similarly @xmath106 , we find to two - loop accuracy @xmath107 \\ & & \mbox { } + \left ( \frac{\alpha_s(\mu)}{4\pi } \right)^2 \bigg [ b_0^{(2 ) } + \left ( b_0^{(1 ) } ( \gamma_0^j-\beta_0 ) + \gamma_1^j - \frac{\pi^2}{6}\gamma_0\gamma_0^j + \zeta_3 \gamma_0 ^ 2 \right ) l \nonumber \\ & & \mbox { } + \frac12 \left ( \gamma_0^j ( \gamma_0^j-\beta_0 ) + b_0^{(1)}\gamma_0+\gamma_1-\frac{\pi^2}{6 } \gamma_0 ^ 2 \right)l^2 - \left ( \frac16\beta_0-\frac12\gamma_0^j \right ) \gamma_0 l^3 + \frac18\gamma_0 ^ 2 l^4 \bigg ] \nonumber .\end{aligned}\ ] ] here , @xmath108 is the one - loop constant , and the two - loop constant @xmath109 is currently unknown . to determine it a multi - loop calculation will be necessary . however , this constant does not enter the two - loop result for the kernel @xmath70 . next , we need to find an ansatz for @xmath92 which satisfies ( [ eq : ydef ] ) . at tree - level @xmath110 and therefore @xmath111 is also independent of @xmath54 . it follows that the constant @xmath109 cancels in the relation ( [ eq : ydef ] ) to two - loop accuracy . beyond the tree approximation the integrated @xmath92 ( defined equivalently to ( [ eq : j ] ) ) must pick up a logarithmic dependence on @xmath69 , as can be seen by interchanging the integrations in ( [ eq : ydef ] ) ; but @xmath92 itself must not depend on @xmath69 . objects that accomplish that are already known from the jet function @xmath112 , and are called `` star - distributions '' @xcite ( see also @xcite ) . their definitions are such that , when integrated over an interval @xmath69 , they act on a function @xmath113 as @xmath114 } \phi(k ) & = & \int_0^{\omega } dk\,\frac{\phi(k ) - \phi(0)}{k}\ , + \phi(0 ) \,\ln \frac{m_b \omega}{\mu^2 } \ , , \\ \int_0^\omega dk\,\left(\frac{1}{k}\ln\frac{m_b k}{\mu^2 } \right)_*^{[\mu^2/m_b ] } \phi(k ) & = & \int_0^{\omega } dk\,\frac{\phi(k ) - \phi(0)}{k } \ln\frac{m_b k}{\mu_i^2 } \ , + \frac{\phi(0)}{2 } \,\ln^2 \frac{m_b \omega}{\mu^2 } \,.\nonumber\end{aligned}\ ] ] therefore , the ansatz for the jet kernel reads ( here with @xmath115 for brevity ) @xmath116 } _ * \right ] \\ & & \mbox { } + \left ( \frac{\alpha_s(\mu)}{4\pi } \right)^2 \left [ c_0^{(2)}(l)\ , \delta(k ) + c_1^{(2)}(l)\left ( \frac1k \right)^{[\mu^2/m_b ] } _ * + c_2^{(2)}(l)\left ( \frac1k \ln \frac{m_b k}{\mu^2 } \right)^{[\mu^2/m_b ] } _ * \right ] . \nonumber\end{aligned}\ ] ] in ( [ eq : wresult ] ) it acts on all @xmath71-dependent objects , i.e. , on the prefactor @xmath117 and on the integration limits @xmath118 $ ] and @xmath119 $ ] . to determine the coefficients @xmath120 , however , we consult ( [ eq : ydef ] ) , where the star distributions act on the @xmath71-dependent logarithms . we find , the connection is that the integrations over @xmath89 and @xmath54 were immediately performed , leading to the replacement of @xmath121 and subsequent devision by @xmath122 in the notation of @xcite . this was possible because for a pure cut on @xmath73 one has @xmath123 and @xmath124 independent of @xmath73 , unlike the general case considered here . ] @xmath125 l + \left [ \gamma_0 \gamma^j_0 - \beta_0 \frac{\gamma_0}{2 } \right ] l^2 + \frac{\gamma_0 ^ 2}{2 } l^3 \nonumber \ ; , \\ c^{(2)}_0(l ) & = & \left[\gamma^j_1 - b^{(1)}_0 \beta_0 \right ] l + \left [ \frac{\gamma^j_0}{2}(\gamma^j_0 - \beta_0 ) + \frac{\gamma_1}{2 } - \frac{\pi^2}{12 } \gamma_0 ^ 2 \right ] l^2 + \frac{\gamma_0}{2 } \left[\gamma^j_0 - \frac13 \beta_0 \right ] l^3 + \frac{\gamma_0 ^ 2}{8}\nonumber l^4 \;.\end{aligned}\ ] ] at this point we interrupt the discussion briefly and consider the result for the weight function in ( [ eq : wresult ] ) again . as mentioned earlier , the @xmath58-quark mass enters the normalization of the weight function since we also normalized the photon spectrum . in order to avoid the renormalon ambiguities of the pole scheme it is favorable to use a low - scale subtracted quark - mass definition , for which we adopt the shape - function scheme @xcite . the two mass definitions are connected via @xmath126 and we will refer to @xmath127 at @xmath128 gev as @xmath129 for brevity , for the rest of this paper . apart from the normalization to the weight function , @xmath129 also enters through radiative logarithms in the hard functions and through star distributions in the jet kernel . since the hard functions are only kept to one - loop order , the above redefinition of @xmath129 does not affect the result . @xmath70 , on the other hand , is given to two - loop accuracy in ( [ eq : yansatz ] ) . it follows that @xmath130 receives a contribution at this level , @xmath131 l + \frac{\gamma_0}{2 } l^2\;.\ ] ] in the remainder of this section we collect the other ingredients of ( [ eq : wresult ] ) for completeness . the hard function in semileptonic decays is known to 1-loop order in perturbation theory and given by @xmath132\bigg\ } - 6 ( y-\e ) \frac{c_f\alpha_s(\mu_h)}{\pi } \ln y\;. \end{aligned}\ ] ] the hard function for the normalized photon spectrum reads @xmath133 ^ 2}{[c^{\rm eff}_{7\gamma}(\mu_h)]^2}\ , \hat f_{11}(\delta _ * ) + \frac{c_1(\mu_h)}{c^{\rm eff}_{7\gamma}(\mu_h ) } \hat f_{17}(\delta _ * ) + \frac{c_1(\mu_h)\,c_{8g}^{\rm eff}(\mu_h ) } { [ c^{\rm eff}_{7\gamma}(\mu_h)]^2 } \hat f_{18}(\delta _ * ) \nonumber \\ & & \mbox { } + \frac{c_{8g}^{\rm eff}(\mu_h ) } { c^{\rm eff}_{7\gamma}(\mu_h ) } \hat f_{78}(\delta _ * ) + \frac{[c_{8g}^{\rm eff}(\mu_h)]^2 } { [ c^{\rm eff}_{7\gamma}(\mu_h)]^2 } \hat f_{88}(\delta _ * ) \bigg\}.\end{aligned}\ ] ] here , @xmath134 , and the functions @xmath135 capture effects from operator mixing @xcite and can be found in this notation in @xcite . note also the term proportional to @xmath136 , which ensures that the three powers of @xmath129 in ( [ eq : wresult ] ) are defined in the shape - function scheme . finally , the rg function @xmath137 is defined as the integrated cusp anomalous dimension from @xmath138 to @xmath139 , which yields @xmath140 \nonumber\\ & & \mbox{}+ \frac{\alpha_s^2(\mu_i ) - \alpha_s^2(\mu_h)}{32\pi^2 } \left [ \frac{\beta_1}{\beta_0 } \left ( \frac{\beta_1}{\beta_0 } - \frac{\gamma_1}{\gamma_0 } \right ) - \frac{\beta_2}{\beta_0 } + \frac{\gamma_2}{\gamma_0 } \right]+ \dots \bigg\}.\end{aligned}\ ] ] when combining the various quantities into ( [ eq : wresult ] ) the result should be re - expanded in @xmath141 to the order in which we are working . however , it is convenient to treat @xmath60 as a running `` physical '' quantity ( similar to @xmath142 ) , which is not expanded . this is the same approach as put forward in @xcite , and we will use it in the phenomenological applications in section [ sec : examples ] . [ [ expansion - coefficients - of - the - anomalous - dimensions . ] ] expansion coefficients of the anomalous dimensions . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + to three - loop order , the coefficients of the @xmath104 function read @xcite @xmath143 where @xmath144 is the number of light flavors , @xmath145 and @xmath146 . the cusp anomalous dimension to three - loop order is given by @xcite @xmath147 , \\ \gamma_2 & = & 16c_f \bigg [ \left(\frac{245}{24 } - \frac{67\pi^2}{54 } + \frac{11\pi^4}{180 } + \frac{11}{6}\,\zeta_3 \right ) c_a^2 + \left ( - \frac{209}{108 } + \frac{5\pi^2}{27 } - \frac73\,\zeta_3 \right ) c_a\,n_f \nonumber\\ & & \mbox { } \qquad + \left ( - \frac{55}{24 } + 2\zeta_3 \right ) c_f\,n_f - \frac{1}{27}\,n_f^2 \bigg ] . \nonumber \end{aligned}\ ] ] we also need the anomalous dimension coefficients for the integrated jet function @xmath93 , which are @xcite @xmath148 . \nonumber \end{aligned}\ ] ] factorization theorems exist at each level of power counting for differential decay rates in inclusive heavy - quark decays . we differentiate between two types of power corrections , `` kinematical '' and `` hadronic '' . the first class arises simply because we have restricted our discussion to a particular portion in phase - space , where @xmath149 . these corrections are power suppressed in the shape - function region , but are of leading power when integrated over a domain that is comparable to @xmath150 ( ope region ) . a different way of thinking about kinematical corrections is to view them as the equivalent of the factorization theorems ( [ eq : xulnu ] ) and ( [ eq : xsgamma ] ) with subleading hard and jet functions . however , no complete scale separation has been achieved for these power corrections yet , but the products of subleading hard and jet functions are known in fixed - order perturbation theory and to all powers from @xcite . since kinematical corrections start at @xmath151 and are numerically small for all prominent cuts , this approximate treatment suffices . the scale @xmath152 is typically near @xmath153 , but independent of @xmath18 and @xmath44 . the second class of corrections comes from subleading hadronic structure functions @xcite . already at first subleading order there are multiple independent shape functions entering the calculation of the differential decay rates . furthermore , they appear in different linear combinations in the semileptonic and radiative cases , so that at this stage a weight function can not relate semileptonic decay rates to radiative ones alone . in equation ( [ eq : sffree ] ) we have therefore added a second term labeled @xmath32 . note that this term is different from the subleading shape - function contribution to the semileptonic decay rate ( denoted @xmath154 in @xcite ) since contributions from subleading hadronic corrections to the photon spectrum are convoluted with the leading - power weight function and must be subtracted . as a result the residual hadronic power corrections are collected in @xmath32 . currently the hard and jet functions in the subleading shape - function contributions to the decay rates are only known at tree - level . since kinematical corrections come with the leading shape function , the corresponding contribution to the weight function @xmath155 can be calculated without the introduction of any hadronic uncertainty . these terms start at order @xmath151 and we only compute the correction to this order . it thus follows that @xmath156 , symbolically . ( here and below the abbreviation @xmath157 in statements within the main text denotes the _ normalized _ photon spectrum . ) the relevant expressions for the differential decay rates have been collected in @xcite and may be written as ( here and below @xmath158 for brevity ) @xmath159}dy \int\limits_0^{\e_{\rm max}[p_+,y]}d\e\;y^{-2a_\gamma } ( m_b - p_+)^4 \int\limits_0^{p_+}d\hat\omega\ ; k_u(x , y,\e)\ , \hat s(\hat\omega)\ ; , \nonumber \\ \frac{1}{\gamma_s}\frac{d\gamma_s^{\rm kin}}{dp_+ } & = & n_s \frac{(m_b - p_+)^2}{m_b^3 } \int\limits_0^{p_+}d\hat\omega\ ; k_s(x)\ , \hat s(\hat\omega)\ ; , \qquad \hbox{with}\;\ ; x = \frac{p_+-\hat\omega}{m_b - p_+}\;,\end{aligned}\ ] ] where we have collected all terms in the functions @xmath160 and @xmath161 , and abbreviated the different norms by @xmath162 and @xmath163 . the leading weight function at tree - level is taken from ( [ eq : wresult ] ) . we now transform the expression for @xmath164 in three steps : first , we interchange the order of the @xmath73 and @xmath165 integrations . next , the integration variable @xmath73 is substituted by @xmath166 . finally the variable @xmath165 is renamed by @xmath73 . after this has been done ( no manipulation is needed for the term @xmath167 ) , we find @xmath168}dy \int\limits_0^{\e_{\rm max}[p_++k , y]}d\e\;y^{-2a_\gamma } \nonumber \\ & & \mbox { } \times \left [ k_u\left(\frac{k}{m_b - p_+-k},y,\e\right ) - h_u^{(0)}(y,\e)\,k_s\left(\frac{k}{m_b - p_+-k}\right ) \right],\end{aligned}\ ] ] where @xmath169 denotes the tree - level part of the hard function in ( [ eq : hu ] ) . we are now going to restrict the calculation to the first - order power corrections because of mixing effects with hadronic power corrections at higher orders . for example , the second - order power correction to the weighted integral over the photon spectrum contains a term @xmath170 . ( as before , the superscript denotes the order in power counting . ) for such a term we would require a compensation in the residual correction @xmath32 at order @xmath151 , which goes beyond the scope of this paper . including only first - order power corrections is not a bad approximation ; the studies in @xcite have shown that the full kinematical corrections can be approximated very well by including the first term in the power expansion only . at this level the functions @xmath171 and @xmath161 involve only two different functional dependences on @xmath172 , a constant and a term proportional to @xmath173 . we find @xmath174 } dy \int\limits_0^{\e_{\rm max}[p_++k , y ] } d\e \;y^{-2a_\gamma } \left [ a(y,\e ) + b(y,\e ) \ln\frac{k}{m_b - p_+-k } \right ] \end{aligned}\ ] ] with @xmath175\ln y \bigg ] \nonumber \\ & & \mbox { } + 12 ( y-\e)(1-y+\e ) \bigg [ \left(\frac13 - \frac49 \ln \frac{m_b}{m_s } \right ) \frac{[c_{8g}^{\rm eff}(\mu_h)]^2}{[c_{7\gamma}^{\rm eff}(\mu_h)]^2 } - \frac{10}{3 } \frac{c_{8g}^{\rm eff}(\mu_h)}{c_{7\gamma}^{\rm eff}(\mu_h ) } \nonumber \\ & & \mbox { } + \frac83 \left ( \frac{c_1(\mu_h)}{c_{7\gamma}^{\rm eff}(\mu_h)}- \frac13 \frac{c_1(\mu_h ) c_{8g}^{\rm eff}(\mu_h)}{[c_{7\gamma}^{\rm eff } ( \mu_h)]^2 } \right ) \,g_1(z ) - \frac{16}{9}\ , \frac{[c_1(\mu_h)]^2}{[c_{7\gamma}^{\rm eff}(\mu_h)]^2}\,g_2(z ) \bigg ] , \nonumber \\ b(y,\e ) & = & 48 \ , \frac{y-\e}{y } ( 1-y)(-3 + 5y-5\e ) - \frac83 ( 1-y+\e)(y-\e ) \frac{[c_{8g}^{\rm eff}(\mu_h)]^2 } { [ c_{7\gamma}^{\rm eff}(\mu_h)]^2 } \;. \nonumber \end{aligned}\ ] ] in these expressions @xmath176 denote the ( effective ) wilson coefficients of the relevant operators in the effective weak hamiltonian for @xmath0 decay . charm - loop penguin contributions to the hard function of the photon spectrum are encoded in the functions @xmath177 and @xmath178 , which depend on the variable @xmath179 @xcite . they are @xmath180 , \nonumber\\\ ] ] @xmath181 2 \left ( \ln\big[(\sqrt{t}+\sqrt{t-4})/2\big ] - \displaystyle\frac{i\pi}{2 } \right)^2 & ~;~ t\ge 4 \ , . \end{array } \right . \nonumber\end{aligned}\ ] ] this concludes the calculation of the weight function . there are four different hadronic structures entering the first - order power corrections to the differential decay rates at tree level . following @xcite we denote them by @xmath182 , @xmath183 , @xmath184 , and @xmath185 . the first one in this list involves the leading shape function and the heavy - quark parameter @xmath186 . this parameter ensures that @xmath182 has a vanishing norm , as all subleading shape functions must have . it is possible to absorb the effect of this function into a weight - function contribution , which in turn only depends on @xmath187 ; however , because of the zero - norm constraint it is difficult to assign the correct numerical value of @xmath186 in that case . instead , we keep this contribution together with the subleading shape functions in @xmath32 . a straight - forward calculation yields @xmath188}dy \int\limits_0^{\e_{\rm max}[p_+,y]}d\e\;y^{-2a_\gamma}\ ; \frac{y-\e}{y}\nonumber \\ & & \mbox { } \times \bigg\ { 2(1-y)(y-\e)\;(\bar\lambda - p_+ ) \hat s(p_+ ) + 2(1-y ) \left ( y-\e + \frac{\e}{y } \right ) \hat t(p_+ ) \\ & & \mbox { } \quad + ( 1-y)(1-y+\e)\ , \hat u(p_+ ) + \left ( 1+y-\frac{y-\e}{y } ( 2-y+y^2)\right ) \hat v(p_+ ) \nonumber \\ & & \mbox { } \quad + y(1-y+\e)\,m_s^2 \hat s'(p_+ ) \bigg\ } , \nonumber \end{aligned}\ ] ] where we have kept the rg - evolution function @xmath189 , as well as @xmath59 @xcite . its structure is such that @xmath190 and we evaluate to leading - order the sudakov exponent ( with @xmath191 ) @xmath192 and the rg function @xmath193 analogous to ( [ eq : agamma ] ) . the use of the anomalous dimension @xmath194 of the leading scet current is a good approximation because most of the subleading operators in scet are build from the leading scet current and thus share the same anomalous dimension . ( a complete resummation requires knowledge of the anomalous dimension matrix of all subleading operators in scet . for some discussion , see e.g. @xcite . ) we have included a small correction from a finite @xmath195-quark mass , leading to the expression proportional to @xmath196 . the appearance of hadronic structure functions in ( [ eq : resssf ] ) introduces some irreducible uncertainties in phenomenological applications of our results . while in principle the leading shape function @xmath197 can be extracted from the photon spectrum , see e.g. @xcite , the forms of the shape functions @xmath183 , @xmath198 , and @xmath199 are unknown . what we know is that their norms vanish and their first moments are given in terms of the heavy - quark parameters @xmath200 and @xmath201 . in order to estimate effects from higher moments we define functions @xmath202 , @xmath203 , via @xmath204 as long as each of the @xmath202 have zero norm and first moment the constraints on the subleading shape functions are satisfied . let us now adopt a default model for the terms in ( [ eq : resssf ] ) . for simplicity we use for @xmath197 the familiar exponential - type functional form @xmath205 with parameters @xmath186 and @xmath200 . by construction the moment constraints on @xmath206 are respected . the default model for ( [ eq : ssfmodel ] ) is defined as @xmath207 in all three cases . in order to estimate the uncertainty introduced by adopting a specific model we closely follow the procedure of @xcite , where four different functions @xmath208 , @xmath209 , were constructed with vanishing norm and first moment . a variation of the functional form of the subleading shape functions was then achieved by setting @xmath210 , where @xmath211 and @xmath212 . combinatorially this means we have @xmath213 different models for the set of subleading shape functions . the estimator for the hadronic uncertainty is the maximal deviation from the default result when sampling over all models . let us demonstrate the phenomenological implications of the result ( [ eq : wresult ] ) , ( [ eq : wkinresult ] ) , and ( [ eq : resssf ] ) by applying the main relation ( [ eq : sffree ] ) for a few examples of typical kinematic cuts . to disentangle theoretical uncertainties from experimental ones we are going to pretend that both the normalized photon spectrum and the semileptonic partial decay rate were measured with no uncertainty . in particular we consider the photon spectrum given as @xmath214 with @xmath215 gev and @xmath216 . this model describes the experimental data by babar @xcite and belle @xcite reasonably well . furthermore we need as inputs the heavy - quark parameters @xmath217 gev@xmath218 , @xmath219 gev@xmath218 , and the quark masses @xmath220 gev @xcite , @xmath221 mev @xcite , and @xmath222 @xcite . here @xmath129 and @xmath200 are defined in the shape - function scheme at a scale @xmath128 gev , while @xmath223 is evaluated in the @xmath224 scheme at @xmath225 gev . the ratio @xmath226 is also evaluated in the @xmath224 scheme , where it is scale invariant . for the strong coupling @xmath142 we use three - loop running from @xmath227 down to @xmath228 gev , apply matching corrections onto a 4-flavor theory , and then run to @xmath44 , @xmath18 , or @xmath152 . default values for these scales are taken to be @xmath229 , @xmath225 gev , and @xmath225 gev , respectively . to assign a perturbative error , the scales are varied around these default settings by factors of @xmath230 and @xmath231 . in all cases considered below , the analysis of uncertainties closely follows @xcite , with similar outcome . first , let us consider a cut on @xmath11 with @xmath233 gev together with a cut on @xmath234 . in terms of the kinematic variables @xmath54 and @xmath89 this means that @xmath235 = 1\ ; , \qquad \e_{\rm max}[p_+,y ] = { \rm min}\left(1-\frac{2e_0}{m_b - p_+ } , y \right).\ ] ] using the central values of the input parameters and scales , the resulting weight functions are depicted on the left - hand side of figure [ fig : examples ] for a few examples of @xmath236 . there is an integrable singularity at the endpoint @xmath237 @xcite if the lepton cut @xmath236 is small . in the limit @xmath238 , corresponding to a pure cut on lepton energy , the weight function vanishes at the endpoint and the singularity disappears . we now study the case @xmath239 gev in more detail . this is a useful example because the partial branching fraction for this particular cut has already been measured @xcite . first , we investigate the perturbative uncertainty of the right - hand side of ( [ eq : sffree ] ) , which is obtained by studying the residual scale dependence entering via the weight function . ( it is clear that the relevant quantity is the entire integral , and not the values of the weight function for individual values of @xmath73 . ) we observe that the sum of the weighted integral over the photon spectrum @xmath240 ps@xmath241 and the residual hadronic corrections @xmath242 ps@xmath241 is very stable under scale variations , but not each of the two terms alone . the nnlo approximation for the weight function introduces roughly a @xmath243 uncertainty . the error from the lo approximation to the power suppressed corrections is numerically of equal magnitude , but tends to cancel a large portion of the scale sensitivity . this leads us to interpret the perturbative error on the convolution integral as also the perturbative uncertainty on the sum of both terms , to avoid counting this error twice . the hadronic uncertainty on the residual term @xmath32 is obtained by taking the maximal deviation from the central value when sampling over a large set of models for the subleading shape functions , as outlined in section [ sec : reshadr ] . finally we also vary the numerical value of @xmath129 and the remaining input parameters @xmath223 , @xmath244 , and @xmath200 within their stated errors . this yields @xmath245 } \ ; \pm 1.42 \,\hbox{\scriptsize [ hadr . ] } \ ; ^{+1.71}_{-1.67 } \,\hbox{\scriptsize [ $ m_b$ ] } \ ; ^{+0.46}_{-0.63 } \,\hbox{\scriptsize [ pars.]}\;.\ ] ] further uncertainty enters in practice , because the photon spectrum can not be measured over the entire range , but only over a certain window around the endpoint @xmath246 . the normalized photon spectrum is then obtained using theoretical information on the fraction of events that fall into this window . the current precision for these fractions is about @xmath247 @xcite , which impacts @xmath248 directly . if we further assume that the left - hand side of relation ( [ eq : sffree ] ) was given with no experimental uncertainty , we can extract @xmath2 with only a theoretical error . for example , let us take the central value @xmath249 @xcite , and dismiss the experimental error . using the average lifetime @xmath250 of the @xmath5 meson , and taking the normalization uncertainty on the photon spectrum into account , we find @xmath251 } \right ) \cdot 10^{-3}\;.\ ] ] when cutting on @xmath252 , @xmath10 , and @xmath253 , the maximal value of @xmath73 is given by @xmath254 . the phase space is such that @xmath255 & = & { \rm min}\left ( 1-\frac{q_0 ^ 2}{(m_b - p_+)^2 } , \frac{m_0 ^ 2 - p_+^2}{p_+(m_b - p_+ ) } \right ) , \\ \e_{\rm max}[p_+,y ] & = & { \rm min}\left ( 1-\frac{2e_0}{m_b - p_+ } , y \right ) . \nonumber\end{aligned}\ ] ] the right - hand side of figure [ fig : examples ] shows three different weight functions . in all cases @xmath256 gev , close to the optimal value @xmath257 , and @xmath239 gev as before . from top to bottom the three curves are for @xmath258 ( pure @xmath42 cut ) , @xmath259 gev@xmath218 ( mixed cut ) , and @xmath260 ( pure @xmath43 cut ) . note that the integrable singularity of the pure @xmath73 cut in the previous discussion is gone , because this point is no longer the endpoint . instead , the weight function has a kink , which is expected from considerations of the phase space depicted in figure [ fig : pspace ] . the endpoint @xmath35 is much larger in this case , and the weight function vanishes there , so that we do not encounter any singularity anymore . as a second example of a @xmath2 determination we consider the case @xmath256 gev , @xmath261 , @xmath239 gev , i.e. , the top curve in the plot on the right - hand side of figure [ fig : examples ] . because of phase - space restrictions the weight function has a kink at @xmath262 gev . for the central values and default models we find @xmath263 ps@xmath241 , and the analysis of uncertainties yields @xmath264 } \ ; \pm 0.47 \,\hbox{\scriptsize [ hadr . ] } \ ; ^{+1.93}_{-1.88 } \,\hbox{\scriptsize [ $ m_b$ ] } \ ; ^{+1.18}_{-0.99 } \,\hbox{\scriptsize [ pars.]}\;.\ ] ] again , we must also add a @xmath247 uncertainty to the norm of the photon spectrum . from the input @xmath265 @xcite follows @xmath266 } \right ) \cdot 10^{-3 } \;.\ ] ] the third and last example is the combined cut @xmath256 gev , @xmath267 gev@xmath218 , @xmath239 gev , whose weight function is shown as the curve in the middle of the right plot in figure [ fig : examples ] . in analogy to the previous cases we obtain @xmath268 ps@xmath241 and @xmath269 } \ ; \pm 0.43 \,\hbox{\scriptsize [ hadr . ] } \ ; ^{+1.10}_{-1.07 } \,\hbox{\scriptsize [ $ m_b$ ] } \ ; ^{+0.71}_{-0.70 } \,\hbox{\scriptsize [ pars.]}\;.\ ] ] taking @xmath270 , which is the mean of the central values of the measurements reported in @xcite , leads to a larger value , namely @xmath271 } \right ) \cdot 10^{-3 } \;.\ ] ] it is of course understood that these values change once an analysis with the full experimental data is conducted . we believe , however , that a theoretical error of about @xmath272 is realistic and consistent with the previous work in @xcite . in the last section we have given results for the weight function @xmath31 for several commonly employed kinematic cuts . in all cases the charged - lepton energy was bound to exceed 1 gev , which is typically used to identify semileptonic @xmath5 decays in practice . we stress that for an additional cut on @xmath73 the photon spectrum is required over only a small window , where already precise data exists . for a cut on the hadronic invariant mass , on the other hand , the photon spectrum is also needed in a regime where its measurement is very difficult . cutting away further events in the low @xmath73 region by virtue of an additional restriction on the leptonic invariant mass @xmath43 worsens the situation even more , because the relative importance of the high @xmath73 region is enhanced . it is therefore expected that the first example , the cut on @xmath73 , will ultimately lead to the most precise determination of @xmath2 . in summary , we have presented a formula in ( [ eq : wresult ] ) which is based on exact factorization theorems for the differential decay rates , and that allows for the calculation of weight functions for arbitrary kinematic cuts . quantities entering this formula were evaluated with one - loop precision at the hard scale , and complete two - loop precision at the intermediate scale , including three - loop running effects in renormalization - group improved perturbation theory . to achieve further precision we have also included first - order power corrections resulting from subleading kinematical and hadronic contributions . the details of the cut are encoded in three kinematic quantities , @xmath66 $ ] , @xmath65 $ ] , and @xmath35 . once they are specified , one only needs to carry out a few integrations that lead directly to the weight function . the use of relations such as ( [ eq : sffree ] ) circumvents the necessity for fitting models of the leading shape function to the @xmath0 photon spectrum and allows for the determination of @xmath2 in a model - independent way at leading power . the quest for a precision measurement of @xmath2 requires a variety of different approaches . the results of this paper represent an alternative route to direct theoretical predictions of partial decay rates . _ acknowledgments : _ i would like to thank matthias neubert for comments on the manuscript . i also thank gil paz for his collaboration in the early stages of this work and his comments on the manuscript . this work was supported in part by funds provided by the u.s . department of energy ( d.o.e . ) under cooperative research agreement de - fc02 - 94er40818 . m. neubert , _ `` qcd based interpretation of the lepton spectrum in inclusive @xmath273 lepton anti - neutrino decays , '' _ phys . rev . d * 49 * , 3392 ( 1994 ) [ hep - ph/9311325 ] . m. neubert , _ `` analysis of the photon spectrum in inclusive @xmath274 decays , '' _ phys . d * 49 * , 4623 ( 1994 ) [ hep - ph/9312311 ] . i. i. y. bigi , m. a. shifman , n. g. uraltsev and a. i. vainshtein , _ `` on the motion of heavy quarks inside hadrons : universal distributions and inclusive decays , '' _ int . j. mod . phys . a * 9 * , 2467 ( 1994 ) [ hep - ph/9312359 ] . c. w. bauer and a. v. manohar , _ `` shape function effects in @xmath275 and @xmath276 decays , '' _ phys . d * 70 * , 034024 ( 2004 ) [ hep - ph/0312109 ] . s. w. bosch , b. o. lange , m. neubert and g. paz , _ `` factorization and shape - function effects in inclusive b - meson decays , '' _ nucl . b * 699 * , 335 ( 2004 ) [ hep - ph/0402094 ] . m. neubert , _ `` renormalization - group improved calculation of the @xmath275 branching ratio , '' _ eur . j. c * 40 * , 165 ( 2005 ) [ hep - ph/0408179 ] . m. neubert , _ `` advanced predictions for moments of the @xmath277 photon spectrum , '' _ hep - ph/0506245 . b. o. lange , m. neubert and g. paz , _ `` theory of charmless inclusive b decays and the extraction of @xmath278 , '' _ hep - ph/0504071 . a. k. leibovich , i. low and i. z. rothstein , _ `` extracting @xmath278 without recourse to structure functions , '' _ phys . d * 61 * , 053006 ( 2000 ) [ hep - ph/9909404 ] . a. k. leibovich , i. low and i. z. rothstein , _ `` extracting @xmath2 from the hadronic mass spectrum of inclusive b decays , '' _ phys . b * 486 * , 86 ( 2000 ) [ hep - ph/0005124 ] . a. h. hoang , z. ligeti and m. luke , _ `` perturbative corrections to the determination of @xmath278 from the @xmath73 spectrum in @xmath279 , '' _ phys . d * 71 * , 093007 ( 2005 ) [ hep - ph/0502134 ] . b. o. lange , m. neubert and g. paz , _ `` a two - loop relation between inclusive radiative and semileptonic b decay spectra , '' _ hep - ph/0508178 . c. w. bauer , s. fleming , d. pirjol and i. w. stewart , _ `` an effective field theory for collinear and soft gluons : heavy to light decays , '' _ phys . d * 63 * , 114020 ( 2001 ) [ hep - ph/0011336 ] . g. p. korchemsky and a. v. radyushkin , _ `` renormalization of the wilson loops beyond the leading order , '' _ nucl . b * 283 * , 342 ( 1987 ) ; i. a. korchemskaya and g. p. korchemsky , _ `` on lightlike wilson loops , '' _ phys . b * 287 * , 169 ( 1992 ) . s. moch , j. a. m. vermaseren and a. vogt , _ `` the three - loop splitting functions in qcd : the non - singlet case , '' _ nucl . phys . b * 688 * , 101 ( 2004 ) [ hep - ph/0403192 ] . g. p. korchemsky and g. marchesini , _ `` structure function for large x and renormalization of wilson loop , '' _ nucl . b * 406 * , 225 ( 1993 ) [ hep - ph/9210281 ] . e. gardi , _ `` on the quark distribution in an on - shell heavy quark and its all - order relations with the perturbative fragmentation function , '' _ jhep * 0502 * , 053 ( 2005 ) [ hep - ph/0501257 ] . f. de fazio and m. neubert , _ `` @xmath280 decay distributions to order @xmath141 , '' _ jhep * 9906 * , 017 ( 1999 ) [ hep - ph/9905351 ] . a. l. kagan and m. neubert , _ `` qcd anatomy of @xmath275 decays , '' _ eur . j. c * 7 * , 5 ( 1999 ) [ hep - ph/9805303 ] . o. v. tarasov , a. a. vladimirov and a. y. zharkov , _ `` the gell - mann - low function of qcd in the three loop approximation , '' _ phys . b * 93 * , 429 ( 1980 ) . c. w. bauer , m. e. luke and t. mannel , _ `` light - cone distribution functions for b decays at subleading order in @xmath45 , '' _ phys . d * 68 * , 094001 ( 2003 ) [ hep - ph/0102089 ] . a. k. leibovich , z. ligeti and m. b. wise , _ `` enhanced subleading structure functions in semileptonic b decay , '' _ phys . b * 539 * , 242 ( 2002 ) [ hep - ph/0205148 ] . c. w. bauer , m. luke and t. mannel , _ `` subleading shape functions in @xmath281 and the determination of @xmath2 , '' _ phys . b * 543 * , 261 ( 2002 ) [ hep - ph/0205150 ] . m. neubert , _ `` subleading shape functions and the determination of @xmath2 , '' _ phys . b * 543 * , 269 ( 2002 ) [ hep - ph/0207002 ] . c. n. burrell , m. e. luke and a. r. williamson , _ `` subleading shape function contributions to the hadronic invariant mass spectrum in @xmath282 decay , '' _ phys . d * 69 * , 074015 ( 2004 ) [ hep - ph/0312366 ] . k. s. m. lee and i. w. stewart , _ `` factorization for power corrections to @xmath283 and @xmath281 , '' _ nucl . b * 721 * , 325 ( 2005 ) [ hep - ph/0409045 ] . s. w. bosch , m. neubert and g. paz , _ `` subleading shape functions in inclusive b decays , '' _ jhep * 0411 * , 073 ( 2004 ) [ hep - ph/0409115 ] . m. beneke , f. campanario , t. mannel and b. d. pecjak , _ `` power corrections to @xmath284 decay spectra in the shape - function region , '' _ jhep * 0506 * , 071 ( 2005 ) [ hep - ph/0411395 ] . r. j. hill , t. becher , s. j. lee and m. neubert , _ `` sudakov resummation for subleading scet currents and heavy - to - light form factors , '' _ jhep * 0407 * , 081 ( 2004 ) [ hep - ph/0404217 ] . c. m. arnesen , j. kundu and i. w. stewart , _ `` constraint equations for heavy - to - light currents in scet , '' _ hep - ph/0508214 . m. trott and a. r. williamson , _ `` towards the anomalous dimension to @xmath285 for phase space restricted @xmath281 and @xmath283 , '' _ hep - ph/0510203 . b. aubert _ et al . _ [ babar collaboration ] , _ `` results from the babar fully inclusive measurement of @xmath283 , '' _ hep - ex/0507001 . k. abe _ et al . _ [ belle collaboration ] , _ `` moments of the photon energy spectrum from @xmath283 decays measured by belle , '' _ hep - ex/0508005 . m. neubert , _ `` two - loop relations for heavy - quark parameters in the shape - function scheme , '' _ phys . b * 612 * , 13 ( 2005 ) [ hep - ph/0412241 ] . c. aubin _ et al . _ [ hpqcd collaboration ] , _ `` first determination of the strange and light quark masses from full lattice qcd , '' _ phys . d * 70 * , 031504 ( 2004 ) [ hep - lat/0405022 ] . e. gamiz , m. jamin , a. pich , j. prades and f. schwab , _ `` @xmath286 and @xmath223 from hadronic tau decays , '' _ phys . lett . * 94 * , 011803 ( 2005 ) [ hep - ph/0408044 ] . i. bizjak _ et al . _ [ belle collaboration ] , _ `` measurement of the inclusive charmless semileptonic partial branching fraction of b mesons and determination of @xmath2 using the full reconstruction tag , '' _ hep - ex/0505088 . b. aubert _ et al . _ [ babar collaboration ] , _ `` measurement of the partial branching fraction for inclusive charmless semileptonic b decays and extraction of @xmath278 , '' _ hep - ex/0507017 .

we have developed a numerical procedure to clarify the critical behavior near a quantum phase transition by analyzing a multi - point correlation function characterizing the ground state . this work presents a successful application of this procedure to the string order parameter of the chain with uniaxial single - ion anisotropy . the finite - size string correlation function is estimated by the density matrix renormalization group method . we focus on the gradient of the inversed - system - size dependence of the correlation function on a logarithmic plot . this quantity shows that the finite - size scaling sensitively changes at the critical point . the behavior of the gradient with increasing system size is divergent , stable at a finite value , or rapidly decreases to zero when the system is in the disordered phase , at the critical point , or in the ordered phase , respectively . the analysis of the finite - size string correlation functions allows precise determination of the boundary of the haldane phase and estimation of the critical exponent of the correlation length . our estimates of the transition point and the critical exponents , which are determined only by the ground - state quantities , are consistent with results obtained from the analysis of the energy - level structure . our analysis requires only the correlation functions of several finite sizes under the same condition as a candidate for the long - range order . the quantity is treated in the same manner irrespective of the kind of elements which destroy the order concerned . this work will assist in the development of a method to directly observe quantum phase transitions .

0811.1616

quantum phase transitions originating from quantum fluctuations have been extensively studied as a hot issue in condensed - matter physics . several interesting characteristics of the transitions appear in the low - energy behavior of the systems . two types of approach can capture the phase transitions and critical phenomena precisely when the transition is continuous . one is analyzing the energy - level structure . the other involves considering the ground - state behavior . in the former approach , a standard method is to analyze the structure of the energy levels of finite - size systems based on the finite - size scaling ( fss ) assumption . for example , the scaled energy gap @xcite is often used to estimate the boundary of the gapped phase as the transition point . this is called phenomenological renormalization group ( prg ) analysis . however , it is difficult to estimate the transition point when a logarithmic correction appears in the dependence of the energy difference . a typical example is the berezinskii kosterlitz thouless ( bkt ) transition @xcite . to resolve this difficulty , the level - spectroscopy method has been developed @xcite and precise determinations of phase transitions have been successfully made for various transitions in many models . unfortunately , this analysis is complicated in that appropriate adjustments of the procedure are required according to the type of phase transition , which must also be known in advance . in the latter approach , on the other hand , quantities that characterize the ground state are carefully observed . one of these quantities is the multi - point correlation function . the long - range behavior of correlation functions shows whether the system exhibits long - range order . if a correlation function survives to be nonzero in the long - range limit , it is an appropriate order parameter . however , it is not easy to capture a phase transition using this strategy because reliable and precise data on correlation functions are necessary for large systems . the system sizes that are treated in numerical - diagonalization calculations are insufficient . for this reason , the latter approach has been employed in only a few studies . therefore , no systematic procedure for analyzing ground - state quantities to capture quantum phase transitions has been established to date . in this paper , we develop a procedure to determine the transition point and critical exponents by analyzing correlation functions based only on the scaling assumption . a feature of this approach is that only the common quantities under the same condition are treated irrespective of the type of phase transition . we call the procedure ground - state phenomenological renormalization group ( gsprg ) analysis . to confirm its validity and usefulness in detecting phase transitions , we apply it to a nontrivial ground state in the af @xmath0 @xmath1 chain with uniaxial single - ion anisotropy by the density matrix renormalization group ( dmrg ) method @xcite . in the isotropic case of this system , there exists a nonzero energy gap between the unique ground state and the first excited state , called the haldane gap@xcite . it is known that when anisotropy is introduced , of the single - ion type or of the @xmath1-type exchange interaction , the haldane gap decreases and finally closes . the region where the nonzero haldane gap exists is called the haldane phase . the phase diagram of the af @xmath0 chain with anisotropy of the two types , including the haldane phase , has been extensively studied by analyzing the energy - level structure , assisted by the level - spectroscopy method @xcite . it is well known that in many af spin systems , the standard spin spin correlation function gives so - called n@xmath2el order . in the ground state in the haldane phase , however , the spin spin correlation function decays exponentially with a finite correlation length and the n@xmath2el order no longer exists . in this sense , the haldane phase is a disordered phase . however , the string order parameter is known to characterize the ground state in the haldane phase , in which the longitudinal and transverse string order parameters are nonzero @xcite . from the viewpoint of the string order , it is possible to treat the haldane phase as an ordered phase and capture the phase transition at the boundary . this approach has been applied to numerical - diagonalization data of the haldane phase in @xmath0 systems@xcite . unfortunately , only very small system sizes were available and hence it was quite difficult to capture precisely the critical behavior of the string order near the transition point . in this situation , we can obtain numerical data of the string order of this model for much larger system sizes by using dmrg for gsprg analysis . consequently , it is possible to examine the phase transition at the boundary of the haldane phase from the viewpoint of the string order . we compare our results with those from the analysis of the energy - level structure . this comparison provides a systematic and consistent understanding of the phase transition . this paper is organized as follows . the model hamiltonian and order parameters are defined in section [ sec : level2 ] . the analysis procedure which we have developed is introduced in section [ sec : level3 ] . the numerical results and discussions are given in section [ sec : level4 ] . section [ sec : level5 ] consists of a summary of this work and some remarks . we consider the following hamiltonian : @xmath3 \!\ ! + \ ! d \ ! \sum^n_{i=1 } \ ! { s^z_i}^2 , \label{eq.1}\end{aligned}\ ] ] where @xmath4 is the system size , @xmath5 are spin-1 operators , @xmath6 and @xmath7 represent the @xmath1-type anisotropic exchange interaction , and the parameter @xmath8 represents uniaxial single - ion anisotropy . in this paper , energies are measured in units of @xmath6 and hence we take @xmath9 hereafter . the boundary condition of the system is open . an antiferromagnetic chain is usually characterized by the n@xmath2el order parameter defined as @xmath10 where @xmath11 is the n@xmath2el correlation function given by @xmath12 here @xmath13 represents the ground - state expectation value of an arbitrary operator @xmath14 . in the haldane phase , the n@xmath2el order parameter vanishes . however , the string order introduced by den nijs and rommelse @xcite appears instead . the string order parameter is given by @xmath15 where the string correlation function @xmath16 is given by @xmath17 s^\alpha_j \right\rangle . \label{ep.3a}\ ] ] kennedy and tasaki @xcite extensively studied the string order and applied a nonlocal unitary transformation to the hamiltonian ( [ eq.1 ] ) . thereby they obtained @xmath18 where @xmath19 , and @xmath20 is obtained by applying the transformation to the original hamiltonian @xmath21 . in the transformed system , @xmath22 symmetry emerges . breaking of this symmetry is described by the behavior of the order parameters @xmath23 . when the system is in the haldane phase , this @xmath22 symmetry is fully broken . when the chain is in a phase other than the haldane phase , the full symmetry or a part of the symmetry survives @xcite . in this section we introduce our analysis procedure . the procedure consists of three steps . the first is to calculate the longitudinal and transverse order parameters by the dmrg method based on the finite - size algorithm @xcite . knowing the behavior of the order parameters enables us to observe the boundary of the haldane phase briefly and to confirm that the transition at the boundary is continuous . if we apply the fss analysis of the order parameters @xcite , we can roughly estimate the critical point and exponents . however , there remain finite - size effects in these estimates . in order to eliminate the effects , we carry out a finite - size extrapolation derived from the fss formula . as the second step , we introduce a finite - size quantity which we calculate from the string order parameters . this quantity reaches the critical exponent just at the phase boundary when we take the limit @xmath24 . by examining the behavior of this quantity , it is possible to obtain the phase boundary and the critical exponent of the order parameter at the transition point . at the third step , we estimate the critical exponent of the correlation length near the transition point by extrapolating a finite - size quantity at the above transition point . the present procedure gives consistent and precise estimates for the critical point and exponents . to calculate the order parameters , we use the finite dmrg algorithm with the acceleration algorithm introduced by white @xcite . a correlation function , such as eq . ( [ eq.2a ] ) or eq . ( [ ep.3a ] ) , is calculated as follows . first , we obtain a variational wave function of the ground state represented by the matrix product state ( mps ) : @xmath25 , s_2 } u_{\alpha_2 , \alpha_3}^{[3 ] , s_3 } \cdots u_{\alpha_{n/2 - 2 } , \alpha_{n/2 - 1}}^{[n/2 - 1 ] , s_{n/2 - 1 } } \right . \hspace{10 mm } \nonumber \\ a_{\alpha_{n/2 - 1 } , \alpha_{n/2 + 2}}^{s_{n/2 } , s_{n/2 + 1 } } % \nonumber \\ v_{\alpha_{n/2 + 2 } , \alpha_{n/2 + 3}}^{[n/2 + 2 ] , s_{n/2 + 2 } } \cdots \hspace{9 mm } \nonumber \\ \left . v_{\alpha_{n-2 } , \alpha_{n-1}}^{[n-2 ] , s_{n-2 } } v_{\alpha_{n-1 } , s_n}^{[n-1 ] , s_{n-1 } } \right ) \left| s_1 s_2 \cdots s_n\right\rangle , \nonumber \\ ( \{s\}=s_1 , s_2 , \cdots , s_n , \hspace{2 mm } \{\alpha\}=\alpha_2 , \alpha_3 , \cdots , \alpha_{n-1 } ) , \label{eq.5 - 5}\end{aligned}\ ] ] where @xmath26 and @xmath27 are matrices satisfying @xmath28\dagger } u^{[i ] } & = \mathbf{1 } , & ( i=2 , \cdots , n/2 - 1 ) , \label{eq.5 - 7a}\\ v^{[j ] } v^{[j ] \dagger } & = \mathbf{1 } , & ( j = n/2 + 2 , \cdots , n-1 ) , \label{eq.5 - 7b}\end{aligned}\ ] ] respectively , and @xmath29 is the ground state of the renormalized hamiltonian which is obtained by applying the density matrix renormalization transformation to the hamiltonian ( [ eq.1 ] ) . a truncation error @xmath30 due to the cut off @xmath31 is estimated by @xmath32 , where @xmath31 is the number of states preserved in the dmrg iterations @xcite . using this wave function , we estimate the expectation value @xmath33 as follows : @xmath34}_{\mathbf{1 } , \mathbf{1 } } e^{[3]}_{\mathbf{1 } } \cdots e^{[i-1]}_{\mathbf{1 } } e^{[i]}_{s^\alpha_{i } } e^{[i+1]}_{\tilde{s}^\alpha_{i+1 } } \right . \cdots \hspace{8 mm } \nonumber\\ e^{[n/2 - 1]}_{\tilde{s}^\alpha_{n/2 - 1 } } e^{[a]}_{\tilde{s}^\alpha_{n/2 } , \tilde{s}^\alpha_{n/2 + 1 } } % \nonumber \\ e^{[n/2 + 2]}_{\tilde{s}^\alpha_{n/2 + 2 } } \cdots \hspace{5 mm } \nonumber \\ \left . e^{[j-1]}_{\tilde{s}^\alpha_{j-1 } } e^{[j]}_{s^\alpha_{j } } e^{[j+1]}_{\mathbf{1 } } \cdots e^{[n-2]}_{\mathbf{1 } } e^{[n-1]}_{\mathbf{1 } , \mathbf{1 } } \right ) . \label{eq.5 - 8}\end{aligned}\ ] ] here , @xmath35 is the identity matrix in spin-1 space and @xmath36 is defined as @xmath37 . each @xmath38 is a matrix for an arbitrary local operator @xmath39 defined by @xmath40}_{o , o } \ ! & = & \sum_{s_1,s_1',s_2 , s_2 ' } \left\langle s_1 \right| o \left| s_1 ' \right\rangle \left\langle s_2 \right| o \left| s_2 ' \right\rangle \hspace{23 mm } \nonumber \\ & & u_{s_1}^{[2 ] , s_2 } \otimes \left ( u_{s_1'}^{[2 ] , s_2 ' } \right)^ * , \hspace{4 mm } \label{eq.5 - 9a}\\ e^{[i]}_{o } & = & \sum_{s_i , s_i ' } \left\langle s_i \right| o \left| s_i ' \right\rangle u_{}^{[i ] , s_i } \otimes \left ( u_{}^{[i ] , s_i ' } \right)^ * , \hspace{14 mm } \nonumber \\ & & ( 3 \leq i \leq n/2 - 1 ) , \hspace{5 mm } \label{eq.5 - 9b } \\ e^{[a]}_{o , o } & = & \sum_{s_k , s_k ' , s_{k+1},s_{k+1 } ' } \!\!\!\!\!\ ! \left\langle s_k \right| o \left| s_k ' \right\rangle \left\langle s_{k+1 } \right| o \left| s_{k+1 } ' \right\rangle \nonumber \hspace{13 mm } \\ & & a_{}^{s_k , s_{k+1 } } \otimes \left ( a_{}^{s_k ' , s_{k+1 } ' } \right)^ * , \nonumber \hspace{5 mm } \\ & & ( k = n/2 ) , \hspace{5 mm } \label{eq.5 - 9c}\\ e^{[j]}_{o } & = & \sum_{s_j , s_j ' } \left\langle s_j \right| o \left| s_j ' \right\rangle v_{}^{[j ] , s_j } \otimes \left ( v_{}^{[j ] , s_{j } ' } \right)^ * , \hspace{10 mm } \nonumber \\ & & ( n/2 + 2 \leq j \leq n-2 ) , \hspace{5 mm } \label{eq.5 - 9d}\\ e^{[n-1]}_{o , o } & = & \sum_{s_{n-1},s_{n-1 } ' , s_n , s_n ' } \hspace{-8 mm } \left\langle s_{n-1 } \right| o \left| s_{n-1 } ' \right\rangle \left\langle s_n \right| o \left| s_n ' \right\rangle \nonumber \hspace{11 mm } \\ & & v_{s_n}^{[n-1 ] , s_{n-1 } } \otimes \left ( v_{s_n'}^{[n-1 ] , s_{n-1 } ' } \right)^*. \hspace{5 mm } \label{eq.5 - 9e}\end{aligned}\ ] ] when calculating a two - point correlation function such as eq . ( [ eq.2a ] ) , we replace all the operators @xmath36 with @xmath35 . we also choose an appropriate correlation function as the longest - ranged component from among @xmath33 or @xmath41 for a fixed @xmath4 under the open boundary condition . three desirable conditions should be met , as follows . 1 . the measurement points @xmath42 and @xmath43 are as far as possible from the edges the correlation distance of @xmath44 is as long as possible . 3 . the distance @xmath44 should increase in proportion to the system size @xmath4 . in order to satisfy the above conditions , we take @xmath45 and @xmath46 . thus , we consider the order parameter @xmath47 where @xmath48 , @xmath49 represents @xmath50 or @xmath51 . note here that @xmath52 should be an integer . we emphasize again that eq . ( [ eq.5 ] ) is useful in the case of the open boundary condition @xcite . the error of the order parameters is estimated as follows . 1 . @xmath53 is calculated in the case of @xmath54 . 2 . @xmath55 and the truncation error @xmath30 are calculated in the case of @xmath56 . the error estimation of the order parameters @xmath57 is defined by the following formula : @xmath58 . \label{eq.6}\end{aligned}\ ] ] in this paper , all numerical data have a truncation error @xmath59 . in the general theory of phase transitions , the treatment of physical quantities depends on whether the transition is continuous or discontinuous . if the transition is continuous , the critical behavior of bulk quantities is extracted through the fss analysis of finite - size quantities @xcite . as we observe later , the dmrg data of correlation functions of the finite - size systems are continuous near the boundary of the haldane phase . the exact - diagonalization data of the string order parameters are also continuous . thus , it is possible to perform fss analysis of our dmrg data of the string order parameter . the present model includes two control parameters , @xmath8 and @xmath7 . when @xmath7 is fixed and @xmath8 is varied , we carry out the fss analysis based on the following equation : @xmath60 where @xmath61 or @xmath62 , @xmath63 is the fixed @xmath7 , and @xmath64 is the critical point . the same equation concerning the string - type order parameter was used in ref . the exponents @xmath65 and @xmath66 are defined as @xmath67 where @xmath68 represents the correlation length . when @xmath8 is fixed and @xmath7 is varied , on the other hand , the fss formula is given by @xmath69 where the critical exponents @xmath65 and @xmath66 are given by @xmath70 note that as the bkt transition point is approached , only eq . ( [ eq.8b ] ) or eq . ( [ eq.10b ] ) is realized . in this case , the correlation length grows exponentially @xcite ; the dependence is different from eq . ( [ eq.8a ] ) or eq . ( [ eq.10a ] ) , having a finite exponent . successfully obtaining a universal function @xmath71 irrespective of the system size @xmath4 in the critical region near the critical point , allows us to determine the critical point and critical exponents . however , the fss analysis still has a problem in that the width of the critical region is unknown . since the width depends on the values of the control parameters , it is difficult to determine or estimate an appropriate width . to avoid this difficulty , we perform the extrapolation explained below . in this paper , we perform a procedure to obtain the transition point and the critical exponents consistently considering the ground - state quantities . we call this procedure the ground - state phenomenological - renormalization - group ( gsprg ) analysis . the first step of the procedure is to examine a finite - size quantity defined as @xmath72 } { \log[n_1/n_2 ] } , \label{eq.11}\end{aligned}\ ] ] where @xmath73 , @xmath74 , and @xmath75 . here , we examine the direction @xmath76 such that @xmath51 shows critical behavior . the quantity indicates the gradient of the curve of the dependence of @xmath51 on @xmath77 in a plot with both the axes on the logarithmic scale . the gradient should be constant for large system sizes when the set of @xmath8 and @xmath7 corresponds to the boundary of the haldane phase . the quantity @xmath78 converges to the critical exponent @xmath79 defined by eq . ( [ eq.8b ] ) or eq . ( [ eq.10b ] ) for ( @xmath8 , @xmath7 ) on the boundary ; the @xmath80 dependence of @xmath78 shows a stable convergence to a finite value when the system size is increased . on the other hand , when the point ( @xmath8 , @xmath7 ) is not on the boundary , @xmath78 shows a different behavior . for ( @xmath8 , @xmath7 ) inside the haldane phase , @xmath51 tends towards a nonzero value as @xmath4 is increased . thus the gradient @xmath78 rapidly decreases . for ( @xmath8 , @xmath7 ) outside the haldane phase , @xmath51 decays rapidly with increasing @xmath4 . this decay is more rapid than that for eq . ( [ eq.8b ] ) or eq . ( [ eq.10b ] ) . in this case , the gradient @xmath78 rapidly increases . therefore , we can find the critical point from the difference in the @xmath80-dependence of @xmath78 . the difference is expected to be more apparent when @xmath80 increases sufficiently to diminish the edge effect . in order to estimate the critical point , we have investigated the behavior of @xmath78 for @xmath81 and found the characteristic behavior of @xmath78 in the region of large @xmath80 . the numerical procedure to determine the critical point by observing the behavior of @xmath78 is as follows . * when the differentiation of the finite - size quantity @xmath78 satisfies the following conditions for a large system size , + @xmath82 + we can consider that the system with @xmath83 is at a critical point . here @xmath84 is the numerical differentiation given by @xmath85 the differentiation is approximated by the difference because @xmath80 is integer or half integer . * if the differentiation reveals @xmath86 we can consider that @xmath78 will decrease rapidly with increasing system size . in this case , the system with @xmath83 is in the ordered phase with respect to the string order . * if the differentiation satisfies @xmath87 we can consider that @xmath78 will increase rapidly with increasing system size . in this case , the system with @xmath83 is in the disordered phase with respect to the string order . we summarize the difference in the behavior of @xmath78 in table [ tab.2 ] . now , the boundary for a finite - size system between the critical region and the string - ordered phase is given by @xmath88 defined in @xmath89 to find a boundary between eq . ( [ eq.12a ] , [ eq.12b ] ) and eq . ( [ eq.15 ] ) . we obtain @xmath90 as a real positive number because @xmath91 is an interpolated value of @xmath92 for @xmath93 . the boundary between the critical region and the string disordered phase is , on the other hand , given by @xmath88 defined in @xmath94 to find a boundary between eq . ( [ eq.12a ] , [ eq.12b ] ) and eq . ( [ eq.14 ] ) . when the critical behavior appears only at a point , the width between the critical - ordered boundary and the critical - disordered boundary shrinks as @xmath90 increases . such behavior will be presented in section [ sec : level4 - 2 ] . in this work , we consider the width of the critical region between the two boundaries to be an error in our analysis of the transition point if the width is very narrow . in a case of the bkt transition , the critical - disordered boundary does not appear . in this case , we must estimate the transition point carefully only from the critical - ordered boundary . details of this treatment will be given in section [ sec : level4 - 4 ] . .[tab.2 ] behavior of @xmath95 in the four phases and near the three transition lines . hn , hl , hx : haldane - n@xmath2el , haldane - large-@xmath8 , haldane-@xmath96 . rd , ri , sf : rapidly decreasing finite - size exponent ( [ eq.11 ] ) , rapidly increasing exponent , stably finite exponent , as the system size is increased . [ cols="^,^,^,^",options="header " , ] the estimation of @xmath79 of the string order parameter by prg analysis has been reported by hida @xcite based on finite - size data of the string orders by an exact diagonalization ( ed ) method . since the system size is limited to being very small , however , the finite - size effect becomes significant . to avoid this difficulty as much as possible , hida combined the critical point determined from the energy gap under the open boundary condition and the string correlation functions under the periodic boundary condition . for our purposes , we impose only the open boundary condition for our dmrg calculations and employ a definition of the string order eq . ( [ eq.5 ] ) as the longest - ranged component . in the final stage of the present analysis , we estimate the critical exponents @xmath97 and @xmath98 . we first consider the case where @xmath8 is controlled for a fixed @xmath7 . within the fss analysis based on eq . ( [ eq.7 ] ) , an appropriate set of @xmath64 , @xmath65 , and @xmath66 is expected to give a universal function @xmath71 near @xmath99 independent of @xmath4 . however , it is difficult to determine the width of the critical region , as we have mentioned . we instead focus our attention on the gradient of the universal function @xmath71 at @xmath100 . we note that the @xmath4-independence of the gradient is a necessary condition for the existence of the universal function @xmath71 near @xmath99 . therefore , we assume that the gradient for @xmath101 and that for @xmath102 agree with each other for the same @xmath64 , @xmath65 , and @xmath66 , to give @xmath103 } { \partial d } \right|_{d = dc } \hspace{10 mm } \nonumber \\ = \left . n_2^{\eta_\alpha-1/\nu_\alpha } \frac{\partial [ o^\alpha_{\textrm{str}}(n_2 , d , j^{\textrm{fix}}_z ) ] } { \partial d } \right|_{d = dc } \label{eq.15 - 5}\end{aligned}\ ] ] we input @xmath64 and @xmath104 determined above into @xmath8 and @xmath105 in this equation and solve with respect to @xmath106 . denoting the solution by @xmath107 , we obtain @xmath108}{\log [ n_2/n_1 ] } , \label{eq.16}\end{aligned}\ ] ] where @xmath109 represents @xmath110 we note that @xmath111 is a finite - size quantity and we examine the @xmath4-dependence of this quantity . an extrapolation to the limit @xmath112 provides the exponent @xmath113 . hereafter , we call @xmath78 and @xmath111 the finite - size exponents . we next consider the case where @xmath7 is controlled for a fixed @xmath8 . the same derivation as the above from eq . ( [ eq.9 ] ) leads to the finite - size exponent : @xmath114}{\log [ n_2/n_1 ] } , \label{eq.18}\end{aligned}\ ] ] where @xmath115 represents @xmath116 the extrapolation of @xmath117 gives the exponent @xmath98 . let us first review the behavior of the four order parameters under consideration in a finite - size system and summarize some important relations between them . in a moderately large system , we can see indications of asymptotic behavior in each order parameter , although slow convergence prevents a full description . some are characteristic for a given region of the parameter space , which is specified as one of the haldane , n@xmath2el , large - d , and @xmath96 phases . we illustrate @xmath118 and @xmath119 with @xmath120 or @xmath62 in fig . [ fig.3 ] . the @xmath8-dependences of the order parameters on the line @xmath121 and their @xmath7-dependences on the line @xmath122 are shown in fig . [ fig.3](a ) and [ fig.3](b ) , respectively . dependence of the order parameters for the hamiltonian ( [ eq.1 ] ) on ( a ) the @xmath121 line and ( b ) the @xmath123 line at @xmath124 . @xmath125 : @xmath126 , @xmath127 : @xmath128 , @xmath129 : @xmath130 , @xmath131 : @xmath132 . , width=283 ] we now compare the magnitudes of the four order parameters . ( i ) if @xmath133 , @xmath51 is larger than @xmath134 . this is a known relation found by kennedy and tasaki @xcite . ( ii ) when @xmath123 and @xmath135 , we have @xmath136 due to the isotropy of the system . ( iii ) when @xmath7 is decreased and when @xmath7 crosses a critical point at @xmath137 , @xmath132 is smaller than @xmath130 . this fact will be discussed and utilized in [ sec : level4 - 4 ] . although @xmath138 appears to be nonzero around @xmath139 with @xmath140 and around @xmath141 with @xmath123 , we can confirm that @xmath130 in this region vanishes for the long - ranged limit . on the other hand , @xmath128 around @xmath142 with @xmath123 looks very small but it survives as a nonzero quantity in an infinite system , as shown in [ sec : level4 ] . the phase boundaries of haldane n@xmath2el , haldane large-@xmath8 , and haldane@xmath96 are denoted by dotted lines , though they are given only as indicators as we will determine the boundaries in the following subsections . we can see that some or all of the order parameters vanish at the phase boundaries . also , the order parameters are continuous around the boundaries , which suggests that the phase transitions are continuous . therefore , the fss analysis and the gsprg procedure are feasible for capturing critical phenomena in this case , except for the berezinskii kosterlitz thouless ( bkt ) transition which does not satisfy the conditions of eq . ( [ eq.8a ] ) or ( [ eq.10a ] ) and thus requires extra consideration . for gsprg , however , it is possible for us to capture the transition by looking at the exponent @xmath65 as discussed in [ sec : level4 - 4 ] . we next observe the behavior of the four order parameters in each phase to determine their thermodynamic limits . in fig . [ fig.1 ] , we illustrate the behavior of the order parameters ( [ eq.5 ] ) as a function of the inverse of the system size at the representative points @xmath143 , @xmath144 , @xmath145 , and @xmath146 . these sets of parameters correspond to the n@xmath2el phase , large-@xmath8 phase , haldane phase , and @xmath96 phase , respectively . ) as a function of the inverse of the system size . each panel represents ( a ) the n@xmath2el phase , ( b ) the large-@xmath8 phase , ( c ) the haldane phase , and ( d ) the @xmath96 phase . @xmath125 : @xmath147 , @xmath127 : @xmath128 , @xmath129 : @xmath138 , @xmath131 : @xmath132 . , width=283 ] we observe that in the n@xmath2el phase , only the order parameters in the @xmath62 direction remain nonzero in the limit @xmath24 . all the four parameters vanish in the large-@xmath8 phase in the thermodynamic limit . we note that in the haldane phase , only the string order parameters in the two directions remain nonzero in the thermodynamic limit . it is difficult to judge in fig . [ fig.1](d ) whether both of the transverse order parameters in the @xmath96 phase vanish or remain nonzero in the thermodynamic limit . we plot the same data on a logarithmic scale in fig . [ fig.2 ] . ) in the @xmath96 phase as a function of @xmath77 . a fitting based on @xmath148 is carried out for @xmath149 . @xmath129 : @xmath138 , @xmath131 : @xmath132 . , width=283 ] for large @xmath4 , the data exhibits a linear behavior , which suggests that the transverse order parameters in the @xmath96 phase are critical , consistent with previous reports @xcite . consequently we can confirm that the order parameter @xmath150 vanishes in the @xmath96 phase . in the limit @xmath24 . in this subsection , we examine the transition from the haldane phase to the large-@xmath8 phase . this transition is known to be of gaussian type . as we observe in [ sec : level4 - 1 ] , the string order parameters in the haldane phase remain nonzero for both @xmath120 and @xmath62 while both of the n@xmath2el order parameters vanish along the directions @xmath151 and @xmath62 . we also observe critical behavior near the transition line in both @xmath132 and @xmath128 . to begin with , we consider difficulties in the fss analysis near the transition between the two phases . in this analysis , we have adjusted the critical point @xmath152 and exponents @xmath65 and @xmath66 so that the data for @xmath153 , 48 , and 96 follows a universal function . the results are depicted in fig . [ fig.4 ] . near the haldane large-@xmath8 critical point . @xmath131 , @xmath125 , and @xmath127 represent @xmath154 . , width=283 ] in fig . [ fig.4](a ) , we observe a deviation from the universal function @xmath71 at @xmath8 , not far from @xmath100 . the appearance of this deviation depends on the system size and the direction @xmath76 . thus , it is not easy to determine the critical region around @xmath100 with finite - size data less than 100 sites in this case . despite this difficulty , we can choose input parameters @xmath152 , @xmath155 , and @xmath79 such that a universal function @xmath71 appears near the transition point . in fig . [ fig.4](a ) , the string correlation functions in the direction @xmath120 provide us with @xmath156 and @xmath157 . on the other hand , in fig . [ fig.4](b ) , the string correlation functions in @xmath158 give @xmath156 and @xmath159 . the estimate of the transition point @xmath152 for @xmath120 and that for @xmath158 agree with each other . this fact strongly suggests that the string correlation functions for the transverse and longitudinal directions reveal a common phase transition . we should note that @xmath132 and @xmath128 are clearly different quantities near the transition point , because there are differences in their exponents , for example @xmath160 and @xmath161 . our fss analysis gives @xmath157 and @xmath162 . we recall that the growth of the correlation length determines the critical behavior near the transition point from a general argument of the renormalization group concerning critical phenomena . in this framework , only a single characteristic length in a system shows critical behavior . the characteristic length must be the correlation length of the system . thus , the exponent of the correlation length should be unique for the order parameters . in this case , the correlation functions of the string order parameters along both @xmath120 and @xmath158 show critical behavior as shown by the fss analysis . from this argument , @xmath163 and @xmath164 should exhibit a serious finite - size effect , which we will examine and solve by gsprg analysis . we consider the case of @xmath165 in order to observe the finite - size effect . in fig . [ fig.5 ] we illustrate our results for the exponents @xmath166 , @xmath167 , @xmath168 , and @xmath169 determined by gsprg analysis . , ( b ) @xmath170 , and ( c ) @xmath171 on @xmath172 . @xmath80 is given by eq . ( [ eq.11 ] ) . the dotted curves in ( a ) and ( b ) are guides for the eyes . the dotted lines in ( c ) are linear fitting lines . the string correlation functions @xmath173 and @xmath174 as a function of @xmath77 are shown in the logarithmic scale in the inset figures of ( a ) and ( b ) , respectively . , width=283 ] in fig.[fig.5](a ) , we observe the critical - disordered boundary at @xmath175 . the critical - ordered boundary is also observed at @xmath176 . on the other hand , we obtain no boundaries defined in eqs . ( [ co - boundary1])-([cd - boundary2 ] ) in the case of @xmath177 . this fact suggests that the critical region for finite - size systems in our study is realized around @xmath177 with a narrow width . in order to confirm whether the width shrinks or not as the system sizes increase , we examine the relationship between @xmath88 and @xmath8 so that the case is on the boundary . we obtain some of the critical - disordered boundaries at @xmath178 , @xmath179 , and @xmath180 . we also obtain some of the critical - ordered boundaries at @xmath181 , @xmath182 , and @xmath183 . these results indicate that the critical region for a given @xmath88 is gradually narrower when @xmath88 increases although the expression of the relationship between @xmath88 and @xmath8 on the boundary is unknown in the present stage . it is reasonable to conclude that the critical region shrinks and goes to the transition point for the infinite - size system . when one can confirm whether the critical region between the two boundaries is sufficiently narrow or not , the width of the region should be regarded as an error coming from the maximum system size and the interval of @xmath8 in the performed calculations . in this work , thus , we conclude @xmath184 . note here that we can obtain the same critical point from @xmath185 in fig.[fig.5](b ) in the same manner . hereafter , we determine critical points with an error in this way . in order to confirm whether the critical behavior ( [ eq.8b ] ) or ( [ eq.10b ] ) appears or not in the original correlation functions , each string correlation function as a function of @xmath77 is shown in the logarithmic scale in inset figures . the finite - size string correlations for each direction clearly reveal a power - law decay behavior at the critical point @xmath156 . on the other hand , a behavior deviating from power - law decay appears in the cases of @xmath186 and @xmath187 in the ordered and disordered phases , respectively , as we have mentioned in [ sec : level3 - 3 ] . note here that a comparison with these insets shows that the system size dependence of the finite - size quantity ( [ eq.11 ] ) sensitively change near the transition point . we next observe the @xmath80 dependence of the finite - size exponents of @xmath168 and @xmath169 for @xmath165 and @xmath188 in fig.[fig.5](c ) . these two finite - size exponents , @xmath163 and @xmath164 , get gradually closer with increasing @xmath80 . in the limit @xmath112 , @xmath163 and @xmath164 appear to approach a single value @xmath1891.2 . this is consistent with the above argument on the unique characteristic length . consequently , the problem of the disagreement of @xmath163 and @xmath164 in the fss analysis occurs due to the finite - size effect and is resolved by gsprg analysis . we now consider the transition point @xmath152 for a fixed @xmath135 . in this case , many studies have reported various estimates for the boundary of the haldane phase , @xmath152 : @xmath190 in ref . @xcite , @xmath191 in ref . @xcite , @xmath192 in ref . @xcite , @xmath193 in ref . @xcite , @xmath194 in ref . @xcite , @xmath195 in ref . @xcite , and @xmath196 in ref . @xcite . among these works , only a single study @xcite was based on the analysis of the string order , although data from the numerical - diagonalization calculations in this study for small clusters might not be sufficient to show the transition point . recently , tzeng and yang @xcite investigated the fidelity susceptibility @xcite of the ground state by the dmrg method to detect quantum phase transitions for the system . this work examines only the information of the ground state , a feature that is shared with our present analysis . other works analyzed the structure of low - energy levels . from the present analysis , our estimate is @xmath197 , which we have obtained irrespective of @xmath120 or @xmath158 . although the estimates are all very close to each other , there are small differences between them even taking errors into account . the reason for these differences is not clear at present and should be resolved as a future issue . next , we consider the transition point @xmath152 for a fixed @xmath121 . the estimation of this point is suitable for checking the availability of our analysis procedure , because a relatively large exponent @xmath198 which is reported 2.38 by analysis of the energy level structure appears @xcite . several previous studies presented numerical data of the transition point as follows : @xmath199 in ref . @xcite , @xmath200 in ref . @xcite , @xmath201 in ref . all of these works examined the free energy near the critical point to determine the critical point . in particular , the recent study @xcite develops rapidly converging methods by using the differentiations of a quantity , which is derivative of the ground state energy with respect to a controlled parameter , as a function of @xmath4 . from the viewpoint of using only information in the ground state for detecting a quantum phase transition , our analysis and their analysis have a common policy . our estimate for @xmath202 is @xmath203 , and this estimate is also consistent with all previous reports . in accordance with the above results , we apply the procedure to estimate the critical behavior for other @xmath7 , confirming the @xmath7 dependence of @xmath204 and @xmath205 . the error of @xmath205 is estimated by @xmath206 . we illustrate our results in fig . [ fig.6 ] together with those of previous reports @xcite . transition line and ( b ) critical exponent @xmath207 . @xmath131 : evaluated value of x - component , @xmath208 : evaluated value of z - component , + : ref . @xcite , @xmath129 : ref . @xcite , @xmath209 : ref . the inset of ( b ) magnifies the data at @xmath210 to allow a clear comparison to distinguish the data . , width=283 ] our estimates of @xmath204 and @xmath205 are common for @xmath120 and @xmath158 within errors . our transition line is almost consistent with those of previous reports @xcite , in which the energy - level structure is analyzed . our estimates of @xmath198 also agree well with previous reports within errors . consequently , our gsprg analysis successfully captures the transition between the haldane phase and the large-@xmath8 phase . the correlation length exponent @xmath198 is known to be related to other critical exponents . in the gaussian transition , okamoto obtained the following relationship from the argument by the bosonization method : @xmath211 where @xmath212 is the exponent defined by @xmath213 at the transition point . note here that @xmath214 holds . to confirm the consistency between our estimate of @xmath198 and the decay of the n@xmath2el correlation function , we plot our @xmath215 at @xmath216 and @xmath217 as a function of @xmath77 on a logarithmic scale in fig . [ sxsx_cor ] . we clearly observe a linear behavior for large @xmath4 . we have added the dotted line @xmath218 with @xmath219 . from eq . ( [ okamoto_relation_nu_eta ] ) , this value of @xmath220 gives @xmath221 , which is consistent with our estimate shown in the inset of fig . [ fig.6 ] . this consistency also supports the scaling hypothesis that the growth of the unique correlation length determines all the critical behavior around the transition point . dependence of @xmath215 at @xmath216 and @xmath217 . the dotted line shows @xmath218 with @xmath219 . , width=283 ] in this subsection , we examine the transition from the haldane phase to the n@xmath2el phase . this transition is considered to be of ising type . we recall that in a transition of ising type , the exponent of the correlation length is @xmath222 when the system approaches the transition point . we have mentioned in the above that the longitudinal string order is nonzero in both of the haldane phase and the n@xmath2el phase and that the order does not reveal the critical behavior at the transition point . this means that the longitudinal string order is not appropriate for studying the haldane n@xmath2el transition . therefore , to study this transition we examine only the transverse string order . by gsprg analysis of this order , we determine the transition point @xmath223 for a given @xmath8 or the transition point @xmath152 for a given @xmath7 and the critical exponent @xmath198 near the transition point . we consider the case of @xmath224 . we illustrate our result for finite - size exponents @xmath225 and @xmath198 in fig . [ fig.7](a ) and ( b ) , respectively . and ( b ) @xmath226 for @xmath227 from our dmrg results for @xmath228 . the dotted curves in ( a ) are guides for the eyes . the dotted line in ( b ) is a linear fitting line applied to the data for large @xmath80 . ( c ) extrapolation procedure of the finite - size critical point @xmath229 for @xmath227 by the numerical - diagonalization method for @xmath230 . the numerical - diagonalization data for @xmath231 are new in this paper . the dotted line is the extrapolation line in ref . the broken curve is a guide for the eyes . , width=283 ] our estimates are @xmath232 and @xmath233 . our estimate of the transition point is different from that of @xmath234 reported in ref . @xcite . to find the reason for the difference between the two estimates , we have made numerical - diagonalization calculations of finite - size clusters up to @xmath235 under the periodic boundary condition and obtained the eigenenergies of the low - energy states . we have performed the same analysis as that in ref . @xcite and determined the finite - size critical point @xmath236 as @xmath7 at which the scaled energy gap does not depend on the system size for @xmath230 . the results are depicted in fig . [ fig.7](c ) . from our numerical data for @xmath237 , we successfully reproduce the results of ref . @xcite . on the other hand , we can observe that @xmath236 of @xmath238 gradually departs from the fitting line of the extrapolation in ref . our new data points approach our estimates from the string order by the dmrg calculations , as shown by the guide for the eyes denoted by the broken curve in fig . [ fig.7](c ) . this agreement suggests that the results from the numerical - diagonalization and dmrg calculations are consistent with each other if we accept the interpretation suggested by the broken curve . hence , careful extrapolation with respect to system size is required . the @xmath80 dependence of our new data appears exponential rather than polynomial . a similar @xmath80 dependence of @xmath239 was reported in ref . @xcite , in which calculations up to @xmath240 based on the multi - target dmrg method with an infinite - system algorithm were carried out under the periodic boundary condition . our result and ref . @xcite suggest that the absence of polynomial components does not depend on the values of the parameters of the system . it is important to be careful when a system - size extrapolation of an ising transition point is carried out by the prg analysis of the energy - level structure . we now compare estimates of the transition point between ref . @xcite and the present analysis . reference @xcite gives @xmath241 . from the present analysis of our data up to @xmath242 , we obtain @xmath243 for the transition point . our estimate , with a very small error , agrees excellently with the estimate in ref . @xcite . we now discuss our estimate of @xmath198 . our estimate @xmath233 is in good agreement with @xmath222 of the ising - type transition . this agreement also suggests that our analysis successfully captures the haldane n@xmath2el transition as well as the haldane large-@xmath8 transition . we can now summarize our results for the transition points @xmath152 for a given @xmath7 and the critical exponents @xmath198 between the haldane and the n@xmath2el phases from our dmrg data . the results are depicted in fig . [ fig.8 ] . el transition points . @xmath131 : our work , + : ref . @xcite , @xmath208 : ref . ( b ) haldane n@xmath2el critical exponent @xmath198 . the inset figure compares our @xmath198 with that of ref . width=283 ] figure [ fig.8](a ) shows that our estimates for the transition points are in good agreement with the results in ref . @xcite of the multi - target dmrg method and the results in ref . @xcite of the numerical diagonalizations . in fig . [ fig.8](b ) , our estimates for the exponent agree well with @xmath222 irrespective of @xmath7 . note here that the center values of our estimates , namely the extrapolated results , are much closer to @xmath222 than the results in ref . @xcite , although our errors are estimated to be larger . note also that the error in @xmath244 is quite large . the reason for this is considered to be that the curve of the haldane n@xmath2el transition points and that of the haldane large-@xmath8 transition points approach each other . a similar phenomenon appears when the central charge @xmath245 on the curve of the haldane large-@xmath8 transition points was estimated in ref . @xcite , in which the estimate of @xmath245 gradually deviates from @xmath246 around @xmath247 . in the report of tzeng and yang @xcite , the transition point and the critical exponent are given as @xmath248 , @xmath249 , respectively , from fidelity susceptibility analysis . our estimated values at the same point are @xmath250 , @xmath251 , which are more precise than the values of tzeng and yang . in this subsection , we examine the transition from the haldane phase to the @xmath96 phase . this transition is considered to be a bkt - type transition . we recall that in a bkt - type transition , the exponents @xmath252 and @xmath253 appear at the transition point and the exponent @xmath198 can not be defined because the correlation length grows exponentially . we consider the case @xmath254 and examine the magnitudes of the string order and the n@xmath2el order . we refer back to the behavior of orders characterizing the haldane phase , in which we have @xmath255 under the condition @xmath256 this means that the region @xmath257 is not in the haldane phase because the inequality ( [ inequality_str ] ) and eq . ( [ vanishing_neel ] ) can not both be satisfied at the same time assuming the inequality ( [ inequality_orders ] ) . however , it is not as easy to make a direct comparison of these quantities in the limit @xmath24 as for the inequality ( [ inequality_orders ] ) . we can instead compare the finite - size quantities @xmath258 and @xmath259 . recall that for @xmath260 , @xmath258 is smaller than @xmath259 when @xmath254 , whereas @xmath258 is larger than @xmath259 when @xmath261 . we have studied the system size dependence of this behavior ; our results are depicted in fig . [ fig.9 ] . . @xmath131 and @xmath127 represent @xmath138 and @xmath132 . , width=283 ] the behavior is clearly independent of system size . we can also confirm this independence irrespective of @xmath8 for cases between the haldane phase and the @xmath96 phase . our present results suggest the inequality ( [ inequality_orders ] ) and indicate that the haldane@xmath96 transition point satisfies @xmath262 . the finding is entirely consistent with previous works . thus , it is sufficient to consider the case of @xmath263 hereafter in examining the haldane@xmath96 transition . we now estimate @xmath264 by our gsprg analysis . we consider the case @xmath265 for @xmath263 . for this purpose , we examine the finite - size exponent @xmath266 , and estimate the critical - ordered boundary point @xmath267 given by eq . ( [ cd - boundary1 ] ) or eq . ( [ cd - boundary2 ] ) . our results are depicted in fig . [ fig.10 ] . . the dotted lines are the rigorous exponent values of the bkt transition : @xmath268 , @xmath269 . @xmath131 , + , @xmath127 , @xmath129 , and @xmath208 represent @xmath270 , 0.1 , 0.15 , 0.18 , and 0.2 , respectively . , width=302 ] we find that the critical - ordered boundary given by eq . ( [ cd - boundary2 ] ) appears when @xmath7 is 0.1 , 0.15 , 0.18 , and 0.2 , but it does not appear when @xmath7 is 0 . concerning with @xmath151-component of the string order , we find the boundaries at @xmath271 , @xmath272 , and @xmath273 . concerning with @xmath62-component of the string order , on the other hand , we have @xmath274 , @xmath275 , and @xmath276 as the boundaries . in the cases of both of the components , one can observe that @xmath90 grows when @xmath7 approaches @xmath270 . these phenomena lead to our result that @xmath264 is between @xmath270 and @xmath277 . for estimating the transition point more accurately , the critical - ordered boundary point is extrapolated to the limit @xmath278 . the results are depicted in fig . [ fig.11 - 5 ] . at @xmath265 as a function of the system size @xmath90 . results are shown for @xmath279 . the error is estimated from the difference between the results for @xmath280 and for @xmath279 . , width=283 ] since the leading dependence of @xmath281 on @xmath282 is unknown , we here choose the power @xmath282 so that the dependence is almost linear . we can successfully determine an appropriate value of the power for each @xmath120 and @xmath158 , although the @xmath120 and @xmath158 values differ from each other . a linear extrapolation gives @xmath283 from the transverse component and @xmath284 from the longitudinal component . here we determine the error as being the difference between the values obtained by the extrapolation and the finite - size critical point @xmath281 for maximum @xmath88 . both results suggest @xmath285 irrespective of the direction of the string order parameter , which is consistent with a previous report @xcite . next , we examine what type of transition this is . our finite - size exponents in fig . [ fig.10 ] at @xmath137 indicate @xmath286 and @xmath287 . these values agree well with the exponents of the bkt transition @xmath288 and @xmath269 . our results are also consistent with many previous works @xcite . therefore , our gsprg analysis applied to the string correlation functions is useful in capturing bkt transitions . we have investigated critical behavior near the boundary of the haldane phase in the ground state of an anisotropic @xmath0 chain from the viewpoint of string correlation functions estimated precisely by standard finite - size dmrg under the open boundary condition . we have developed the ground - state phenomenological - renormalization - group analysis and used it to analyze the correlation functions . this analysis provides us with the transition point of the boundary of the haldane phase and the critical exponents at and near the transition point . our estimates for these quantities agree with those previously obtained from analysis of the energy - level structure . ) . the haldane large-@xmath8 , haldane n@xmath2el , and haldane@xmath96 transition points are denoted by @xmath289 , respectively . all the lines are guides for the eyes . x - haldane and z - haldane represent the @xmath290 region and @xmath291 regions , respectively . , width=283 ] we summarize the transition points as a ground - state phase diagram in fig . [ fig.12 ] . this figure presents the phase boundary of the haldane large-@xmath8 , haldane n@xmath2el , and haldane@xmath96 transitions . note additionally that the dominant order parameter is @xmath132 in most of the haldane phase . a feature of our approach , gsprg analysis , is that only common quantities under the same condition are treated in a unified manner irrespective of the type of phase transition . although we have employed the dmrg method in this paper to calculate the order parameters , we are not limited to the dmrg method if we can obtain precise estimates of the order parameters . the string order parameters of the haldane phase in the @xmath0 chain are examples . other multi - point correlation functions for finite - size clusters may also be applicable . if we precisely calculate an appropriate ground - state quantity that plays the role of an order parameter , the framework of the analysis would be widely applicable for capturing ground - state critical behavior irrespective of the method of calculation and the kind of order parameter . we hope that the procedure presented in this paper contributes to future studies of quantum phase transitions . we wish to thank prof . k. hida , dr . k. okamoto , dr . t. sakai , and dr . s. todo for fruitful discussions . we are grateful to prof . t. nishino , dr . k. okunishi , and dr . k. ueda for comments on the dmrg calculations . this work was partly supported by grants - in - aid from the ministry of education , culture , sports , science and technology ( nos . 19310094 , 17064006 , 20340096 ) , the 21st coe program supported by the japan society for promotion of science , and next generation integrated nanoscience simulation software . a part of the computations was performed using the facilities of the information initiative center , hokkaido university and the supercomputer center , institute for solid state physics , university of tokyo . 99 m. n. barber , _ in _ `` phase transitions and critical phenomena '' , vol . 8 , ( c. domb and m. s. green , eds . ) , academic press , london , p. 145 , ( 1983 ) . j. m. kosterlitz , j. phys . c , * 7 * , 1046 ( 1974 ) . k. nomura , j. phys . a * 28 * , 5451 ( 1995 ) . s. r. white , phys . lett . * 69 * , 2863 ( 1992 ) . s. r. white , phys . b * 48 * , 10345 ( 1993 ) . f. d. m. haldane , phys . lett . * 93a * , 464 ( 1983 ) . f. d. m. haldane , phys . * 50 * , 1153 ( 1983 ) . f. c. alcaraz and y. hatsugai , phys . b * 46 * , 13914 ( 1992 ) . w. chen , k. hida , and b. c. sanctuary , phys . b * 67 * , 104401 ( 2003 ) . c. d. e. boschi , e. ercolessi , f. ortolani , and m. roncaglia , eur . phys . j. b * 35 * , 465 ( 2003 ) . c. d. e. boschi and f. ortolani , eur . j. b * 41 * , 503 ( 2004 ) . m. den nijs and k. rommelse , phys . b * 40 * , 4709 ( 1989 ) . t. kennedy and h. tasaki , phys . b * 45 * , 304 ( 1992 ) . k. totsuka , y. nishiyama , n. hatano , and m. suzuki , j. phys : condens . matter * 7 * , 4895 ( 1995 ) . t. tonegawa , t. nakao , and m. kaburagi , j. phys . soc . jpn . * 65 * , 3317 ( 1996 ) . y. hatsugai and m. kohmoto , phys . b * 44 * , 11789 ( 1991 ) . s. tonooka , h. nakano , k. kusakabe , and n. suzuki , j. phys . . jpn . * 76 * , 084714 ( 2007 ) . s. r. white , phys . lett . * 77 * , 3633 ( 1996 ) . an estimation of string correlation functions with periodic boundary condition by the dmrg method has been reported @xcite . however , the result for correlation functions is translationally non - invariant in spite of the fact that the system is translationally invariant . under such circumstances , it is difficult to capture the phase transition precisely if we examine the correlation function of the longest distance between @xmath42 and @xmath292 . k. hida , j. phys . . jpn . * 62 * , 1466 ( 1993 ) . t. kennedy and h. tasaki , commun * 147 * , 431 ( 1992 ) . t. sakai and m. takahashi , phys . b * 42 * , 4537 ( 1990 ) . o. golinelli , t. jolicoeur , and r. lacaze , phys . b * 46 * , 10854 ( 1992 ) . a. koga , phys . lett . * 296 * , 243 ( 2002 ) . w. chen , k. hida , and b. c. sanctuary , j. phys . . jpn . * 69 * , 237 ( 2000 ) . y.c . tzeng and m.f . yang , phys . a * 77 * , 012311 ( 2008 ) . l. campos venuti and p. zanardi , phys . rev . lett . * 99 * , 095701 ( 2007 ) . m. roncaglia , l. campos venuti , and c. degli esposti boschi , phys . b * 77 * , 155413 ( 2008 ) . k. okamoto , j. phys . a * 29 * , 1639 ( 1996 ) . f. c. alcaraz and a. moreo , phys . b * 46 * , 2896 ( 1992 ) .

we analyze the influence of sterile neutrinos with the masses in the mev range on the primordial abundances of helium-4 and deuterium . we solve explicitly the boltzmann equations for all particle species , taking into account neutrino flavour oscillations and demonstrate that the abundances are sensitive _ mostly _ to the sterile neutrino lifetime and only weakly to the way the active - sterile mixing is distributed between flavours . the decay of these particles also perturbs the spectra of ( decoupled ) neutrinos and heats photons , changing the ratio of neutrino to photon energy density , that can be interpreted as extra neutrino species at the recombination epoch . we derive upper bounds on the lifetime of sterile neutrinos based on both astrophysical and cosmological measurements of helium-4 and deuterium . we also demonstrate that the recent results of izotov & thuan , who find higher than predicted by the standard primordial nucleosynthesis value of helium-4 abundance , are consistent with the presence in the plasma of sterile neutrinos with the lifetime 0.01 2 seconds .

1202.2841

the characteristic feature of the physical processes in the early universe is a peculiar interplay of gravity and microscopic physics . gravity introduces the _ hubble time parameter _ @xmath1 that indicates the timescale on which the global properties of the universe ( geometry , temperature , etc . ) change significantly . the hubble time is determined solely by the energy density of the matter filling the space the microscopic matter constituents , particles , are involved in the interaction processes , that are believed to be described fundamentally by three known forces electromagnetic , weak and strong . as long as the timescale @xmath2 of any given microscopic physical process is much smaller than @xmath1 , the expansion can be neglected on that timescale . if time @xmath2 is enough to establish thermal equilibrium between the particles , then the equilibrium it maintained in the course of the universe expansion , while @xmath3 holds . when this inequality ceases to hold , the state of equilibrium is lost . the main reason for that is that interparticle distances become larger , while the corresponding densities become lower , hence interactions are less likely to occur . in this paper we are consider the formation of light nuclei in the primordial environment big bang nucleosynthesis ( bbn ) . all three fundamental interactions are important for this phenomenon , all playing different roles . charged particles together with photons are subject to electromagnetic forces and the equilibration timescale of corresponding processes is tiny with respect to the expansion time . therefore the particles are kept in thermal equilibrium at the common temperature @xmath4 . due to expansion the temperature is _ decreasing _ with time . the equilibration time of the weak interactions changes abruptly so that at @xmath5 weakly interacting neutral particles ( neutrinos and neutrons ) stay in equilibrium , while at lower temperatures they fall out of it ( freeze out ) . at high temperatures processes like @xmath6 maintain _ chemical _ equilibrium , that is the neutron - to - proton conversion exhibits the same finite intensity as the opposite processes . chemical and thermal equilibria are interconnected , so they are lost simultaneously , when neutron - to - proton ratio freezes out . finally , the strong interactions are responsible for the production of nuclei comprising more than one nucleon . the most important fusion reaction for the formation of the first nucleus , deuteron , @xmath7 , _ releases _ energy of at least the binding energy of deuteron @xmath8 , and proceeds effectively in dense primordial medium . at temperatures of the order of @xmath9 , however , energetic photons collide with deuteron and lead to its destruction . as baryon density is much lower than the density of photons @xcite , there are _ many _ photons with energies much higher than @xmath9 that collide with deuterons and hence postpone the production of the significant deuteron density until the temperature when the photodissociation is not effective anymore , @xmath10 , much lower than the binding energy . the net abundance of deuterium is , however , non - zero at all times till this moment and is given by the equilibrium boltzmann distribution . deuterium that is created at lower temperatures , serves as a fuel for the formation of @xmath11he , @xmath12he and other nuclides . although the times of elements production and the moment of the departure from the chemical @xmath13 equilibrium are well - separated , the former process is very sensitive to the latter . firstly , the details of the freeze - out set the ratio of the neutron to proton densities , and secondly , the time elapsed between the two moments determines the fraction of neutrons that have decayed since then ( recalling that neutron is an unstable particle ) . the seminal ideas of the primordial synthesis of light elements were first outlined in the so - called @xmath14 paper , @xcite , published in the late 1940s . since then the theory of big bang nucleosynthesis has evolved and its main predictions were confirmed , making it a well - developed model from both theoretical and observational points of view . a lot of reviews of the standard bbn scenario and its implication for particle physics models exist ( see e.g. @xcite ) . the predictions of the primordial nucleosynthesis can change once one replaces the standard model of particle physics underlying the processes considered so far by some of its `` beyond the standard model '' ( bsm ) extensions . therefore the bbn plays the role of a benchmark for testing physical models . in this paper we investigate the influence of _ sterile neutrinos _ on primordial nucleosynthesis . sterile neutrinos are hypothetical massive _ super - weakly - interacting _ particles ( see e.g. @xcite for reviews ) , as opposed to their weakly - interacting counterparts ordinary standard model neutrinos @xmath15 , that are called `` _ _ active _ _ '' in this context . sterile neutrinos carry no charges with respect to the standard model gauge groups ( hence the name ) , but via their quadratic _ mixing _ to active neutrinos they effectively participate in weak reactions and at energies much below the mass of the @xmath16-boson their interaction can be described by the analog of the fermi theory with the fermi coupling constant @xmath17 replaced by @xmath18 , where the _ active - sterile mixing angle _ @xmath19 ( see fig . [ fig : fermi1 ] ) . here @xmath20 is a flavour index , @xmath21 , indicating that sterile neutrino can mix differently with neutrinos of different flavours . massive sterile neutrinos can decay , but due to their feeble interaction strength their lifetime can be of order seconds ( even for masses as large as mev ) . the decay products of the sterile neutrinos are injected into the primordial environment , increasing its temperature and shifting the chemical equilibrium . in this work we concentrate on sterile neutrinos with the masses in the mev range , motivated by the recent observations @xcite that particles with such masses can be responsible simultaneously for neutrino oscillations and generation of baryon and lepton asymmetry of the universe and can influence the subsequent generation of dark matter @xcite . the corresponding model has been dubbed @xmath22msm ( _ neutrino minimal standard model _ , see @xcite for review ) . several works had previously considered the influence of mev - scale particles on primordial nucleosynthesis . compared to the refs . @xcite this paper accounts for the neutrino flavour oscillations in the plasma and employs more accurate strategy of solving boltzmann equations , which results in the revision of the bounds of @xcite ( see section [ sec : discussion ] for detailed comparison ) . the authors of @xcite developed a new code that can perform treatment of active and sterile neutrinos with arbitrary distribution functions , non - zero lepton asymmetry , etc . however , as of time of writing this code has not been made publicly available and the ref . @xcite did not derive bounds on sterile neutrino parameters . the work @xcite concentrated on the bounds that cosmic microwave background measurements could provide on decaying sterile neutrinos with the masses @xmath23 mev , leaving bbn analysis for the future work . a number of other works ( @xcite ) analyzed the influence of decaying mev particles on bbn . we compare with them in the corresponding parts of the paper . [ [ the - paper - is - organized - as - follows . ] ] the paper is organized as follows . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + we explain the modifications of the standard bbn computations due to the presence of sterile neutrinos in the plasma and describe our numerical procedure in sec . [ sec : bbn - formalism ] . the results are summarized in sec . [ sec : results ] . we conclude in sec . [ sec : discussion ] . appendixes [ sec : numerical - tests][app : neutrino - oscillations ] provide the details of our numerical procedure . the section below summarizes our setup for the bbn analysis with decaying particles . the notations and conventions closely follow the series of works @xcite . we will be interested only in the tree - level fermi interactions of sterile neutrinos with the primordial plasma . in this case the interaction is fully determined by the squares of their mixing angles . we will consider one majorana particle with 4 degrees of freedom of additional chiral singlets ( i.e. `` neutrino - like '' species ) . actual number of degrees of freedom is of course twice larger : @xmath24 , because every chiral fermion has 2 different helicity states . ] and three _ active - sterile mixing angles _ @xmath25 . matrix elements of interactions of sterile neutrinos with the standard model particles are summarized in appendix [ app : matrix - elements ] ( tables [ tab : sterile - scatterings ] ) . we consider in this work only sterile neutrinos with the masses in the range @xmath26 . for heavier particles , two - particle decay channels appear ( e.g. @xmath27 ) and our procedure of solving boltzmann equations ( described below ) should be significantly modified . the lower bound was chosen to be around @xmath28 by the following considerations . the sterile neutrino lifetime @xmath29 is @xcite @xmath30 \\ & \approx 6.9~\mathrm{sec}^{-1}~\left(\frac{m_s}{10{\ensuremath{\:\mathrm{mev}}}}\right)^5\bigr [ 1.6\,\vartheta_e^2 + 1.13(\vartheta_\mu^2 + \vartheta_\tau^2)\bigr ] \end{aligned}\ ] ] where @xmath31 is the weinberg s angle and @xmath32 , @xmath33 , @xmath34 . ) is for majorana particle . for dirac particle the lifetime would be twice larger . ] from this expression one sees that sterile neutrinos lighter than about @xmath35 have lifetime of at least several hundred seconds even for very large mixing angles @xmath36 . therefore , such particles survive till the onset of the bbn , and freeze - out at temperatures @xmath37 . they would be relativistic at that time , i.e. their average momentum would be of the order of temperature , @xmath38 , and their contribution to the number of relativistic neutrino species would be significant , @xmath39 . in the course of the universe expansion @xmath40 would scale as temperature due to the gravitational redshift , and at some point would become smaller than the mass of sterile neutrino . at that moment the energy density of sterile neutrinos would start to change with expansion as @xmath41 rather than @xmath42 ( where @xmath43 is a scale - factor ) so that the contribution of these massive particles to the energy density would quickly become dominant , making @xmath44 ( or could even overclose the universe ) before the production of light elements starts . it contradicts the current bound that puts @xmath45 at @xmath0 @xcite . additionally , in the @xmath22msm the successful baryogenesis is possible only for the masses of sterile neutrinos above few mev @xcite . therefore we restrict the analysis to the region of masses higher than @xmath28 . we consider expansion of the homogeneous and isotropic universe with the flat friedmann walker metric in the form @xmath46 , where @xmath47 is a time - dependent scale factor , whose evolution is described by the friedmann equation @xmath48 with the quantity on the left - hand side being the hubble expansion rate , reciprocal to the expansion timescale @xmath1 discussed above . the total energy density @xmath49 is the sum of all the energy densities present in the medium , and @xmath50 is the newton s constant . the energy density together with the total pressure density @xmath51 satisfy the `` energy conservation '' law @xmath52 at the temperatures of interest the dominant components of the plasma are photons @xmath53 , electrons and positrons @xmath54 , three flavours of active neutrinos ( @xmath55 ) and sterile neutrinos .. see sec . [ sec : sterile - dynamics ] for details . ] working with the particle kinematics in the expanding universe it is convenient to use _ conformal momentum @xmath56 _ instead of the usual physical momentum @xmath51 . the two are related through @xmath57 . the quantitative description of the plasma population is provided by the distribution functions @xmath58 , that are the numbers of particles @xmath20 per `` unit cell '' of the phase space @xmath59 . mev temperatures the medium is homogeneous and the distribution functions are independent of spatial coordinates of particles , and due to isotropy they do not depend on the direction of the particle momentum . that simplifies the description of their evolution and therefore @xmath60 holds . the goal is to find the distribution functions of all relevant particles and to use them to compute the energy density and pressure as a function of time and scale - factor , closing the system of eqs . ( [ eq : friedmann - first])([eq : friedmann - second ] ) via @xmath61 here the summation goes over all plasma particles , @xmath62 is the number of degrees of freedom and mass of @xmath63-th particle respectively , @xmath64 . if interaction rate of the particles is much faster than the hubble expansion rate , their distribution functions are given by either the bose - einstein , or the fermi - dirac distributions . this is the case for photons , electrons and positrons that are kept in equilibrium due to intensive electromagnetic interactions @xmath65 their contribution to the energy and pressure in eqs . ( [ eq : friedmann - first ] ) , ( [ eq : friedmann - second ] ) is hence determined by the single parameter temperature . however , to describe the contributions of the other particles one has to solve _ kinetic _ equations involving them ( see secs . [ sec : active - dynamics][sec : sterile - dynamics ] below ) . the contribution of the baryonic matter to the evolution of the hot plasma of relativistic species is proportional to the so - called baryon - to - photon ratio @xmath66 . the measurements of relic radiation @xcite yield @xmath67 . one can see that baryons are present in negligible amount , and do not influence the dynamics of the remaining medium . this allows to analyze our problem in two steps . at step i we omit baryonic species and study how the temperature of the plasma , the expansion factor and neutrino distributions evolve in time from temperatures of the order of @xmath68 , when sterile neutrinos typically start to go out of equilibrium , down to @xmath69 when nuclear fusion reactions have ended . at step ii we use these results to determine the outcome of the nuclear reaction network against the background of evolving electromagnetic plasma ( sec . [ sec : nuclear - dynamics ] ) . weak interactions are not able to maintain the thermal equilibrium of active neutrinos with the plasma during all the expansion period we consider . a simple comparison of the weak collision rate @xmath70 and @xmath71 tells that neutrino maintain their equilibrium with the rest of the plasma down to temperatures @xmath72 . the process of neutrinos going out of equilibrium is usually referred to as _ neutrino decoupling_. throughout the paper we assume that _ no _ large lepton asymmetry is present so that the number of neutrinos is equal to the number of antineutrinos . at temperatures higher than @xmath73 the distribution is therefore given by the fermi - dirac one , while at lower temperatures we have to solve the set of three boltzmann equations @xmath74 the details of the interactions , such as particle collisions , are encoded in the so - called collision terms @xmath75 . the terms are explicitly @xcite @xmath76 ( 2\pi)^4 \delta^4 ( p_{\rm in}-p_{\rm out } ) \prod\limits_{i=2}^q \frac{d^3 p_i}{(2\pi)^3 2 e_i}\ ] ] the sum runs over all the possible initial states `` in '' involving @xmath77 ( represented by a particle set @xmath78 ) and the final states `` out '' ( @xmath79 ) . matrix element @xmath80 corresponds to the probability of the transition `` in''``out '' to occur and the delta - function ensures the conservation of 4-momentum @xmath81 . symmetrization factor @xmath82 is equal to @xmath83 , except of the transitions involving identical particles either in initial or in a final state . relevant matrix elements together with the symmetrization factors are listed in appendix [ app : matrix - elements ] . the interaction rates are dependent on the population of the medium , and the functional @xmath84 $ ] describes this . in case when all the incoming and outgoing particles are fermions , @xmath85=(1-f_{\nu_\alpha})\hdots ( 1-f_k ) f_{k+1 } \hdots f_q - f_{\nu_\alpha } \hdots f_k ( 1-f_{k+1 } ) \hdots ( 1-f_q).\ ] ] when some of particles are bosons , one has to replace @xmath86 by @xmath87 for every bosonic particle @xmath88 . a simple estimate ( see appendix [ app : neutrino - oscillations ] ) demonstrates that the rates of transitions between neutrinos of different flavours are much faster than weak reactions . we argue that this phenomenon can be approximately described by the following modification of the boltzmann equations @xmath89 summation is carried out over three active flavours and expressions for @xmath90 are listed in appendix [ app : neutrino - oscillations ] ( eqs . [ eq : oscillation - probabilities ] ) . as already mentioned , sterile neutrinos interact much more feebly than active neutrinos do . nevertheless , at some high temperature sterile neutrinos may enter thermal equilibrium . whether this happens or not depends on the thermal history of the universe before the onset of the synthesis.msmmodel at early times ( @xmath91 ) initial densities of sterile neutrinos are negligible @xcite . then the neutrinos come into equilibrium at temperature @xmath92 ( typically @xmath93 gev ) and freeze - out at temperatures @xmath94 @xcite . ] even if they were in thermal equilibrium at early times , sterile neutrinos then necessarily decouple at temperatures _ higher _ than those of active neutrino decoupling . if sterile neutrinos were light and stable ( or very long - lived ) , they would be relativistic and propagate freely in the medium , yielding @xmath95 together with active neutrinos ( @xmath96 is the number of sterile neutrinos ) . however , sterile neutrinos decay into active neutrinos and other particles . the energies of the decay products may be very different from the typical energies of plasma particles . for particles that equilibrate quickly ( such as electrons or photons ) , this `` injection '' results in the fast redistribution of the energy between all particles in equilibrium and effectively the process looks like a temperature increase ( more precisely , it just slows down the cooling of the universe ) . but for particles that either are not in equilibrium or are about to fall out of it , such as active neutrinos at few mev , the `` injection '' _ modifies _ the form of their spectra . the other mass - induced effect is that sterile neutrinos may switch from the relativistic regime ( when their average momentum is larger than mass ) , that is established at large temperatures , to the non - relativistic one , due to the gravitational redshift . for the quantitative description of sterile neutrino dynamics we utilize the boltzmann equation similar to ( [ eq : boltzmann - equation ] ) , replacing active neutrino everywhere therein by sterile neutrino @xmath97 @xmath98 reactions contributing to the right - hand side together with their probabilities are listed in tables [ tab : sterile - scatterings] of appendix [ app : matrix - elements ] . note that we neglect the processes involving baryonic particles . however , they become important for temperatures near the qcd crossover temperature @xmath99 , when their density is not negligible anymore . more scattering channels of sterile neutrino would appear and their proper account is involved . however it seems to be reasonable to assert that the only modification the account will bring is to _ lower _ the decoupling temperature of sterile neutrinos . oscillation phenomenon does not affect significantly sterile neutrinos and therefore boltzmann equation in its original form ( [ eq : boltzmann - equation - sterile ] ) is still valid , contrary to what we have found out for active neutrinos . an argument in favor of this statement is explained in appendix [ app : neutrino - oscillations ] . when sterile neutrino is heavier than muon , the former particle can appear in the decay @xmath100 . however , the branching fraction of this decay mode does not even reach a percent for masses of sterile neutrino we consider ( see e.g. @xcite ) . therefore we can neglect influence of both muons and other particles , appearing in the decay . as a result we have six equations ( [ eq : friedmann - first ] ) , ( [ eq : friedmann - second ] ) , ( [ eq : boltzmann - equation - oscillations ] ) , and ( [ eq : boltzmann - equation - sterile ] ) describing primordial plasma at temperatures of interest . these equations contain six unknowns scale factor @xmath101 , temperature @xmath102 and four neutrino distribution functions , @xmath103 and @xmath104 . the system of equations is therefore closed and we have solved it numerically at the step i. outcome of the nuclear reaction chains is found numerically . for the standard bbn model one of the earlier attempts was made with the code written by l. kawano @xcite . however , the program in its original form is inappropriate for the account of the bsm physics , and we modified it for this work . two technical remarks are in order here . first , we used the 1992 version of the program @xcite as a starting point , and not the 1988 one , @xcite . therefore , the integration time steps were taken small enough , so that the integration procedure did not introduce an error , that was compensated as a shift in the resulting value of the @xmath105 , the _ mass fraction _ of the @xmath12he , that is a fraction of the total baryon mass stored in the form of helium-4 ] the so - called `` kernan correction '' @xcite . second , the code did not take into account the coulomb and the nucleon finite - mass corrections to weak interaction rates , as well as radiative and finite - temperature effects . we do not calculate directly these effects , but assume their net result to be in the form of the additive correction , which we took to be @xmath106 @xcite . the tests described in appendix [ sec : stand - nucl - model ] demonstrate an agreement of thus modified `` kawano code '' with the results of the other code , parthenope @xcite , that takes a proper account of these effects . presence of sterile neutrinos alters the standard dynamics of the temperature and the expansion rate as well as the rates of weak interactions involving neutrons and protons . these quantities are known from the step i , so we have implemented the import of these data . together with the change of @xmath107 indicated above , it has lead to the code , that became an essential tool of step ii in our approach . the computations of nuclide evolution started from temperatures of several mev , when the chemical equilibrium ceases to hold , up to temperatures @xmath108 . the observables of the bbn are concentrations , or abundances , of light nuclides dispersed in the cosmos . the most relevant abundance in our problem is that of @xmath12he , as it is sensitive to the expansion rate of the universe at mev temperatures and neutrino distribution functions . the presence of sterile neutrinos in plasma typically increases the concentration of @xmath12he , described by @xmath105 . accurate calculations carried out in the standard model @xcite predict the values @xmath109 depending on the lifetime of neutron , @xmath110 , see below . there are two main methods of experimental determination of primordial helium abundance . the first one is related to the studies of low - metallicity astrophysical environments and extrapolating them to zero metallicity case . the @xmath105 measurements are known to be dominated by systematic uncertainties . therefore we adopt the @xmath105 values from the two most recent studies , refs . @xcite that have slightly different implications . for recent discussion of various systematic uncertainties in @xmath12he determination , see @xcite . in ref . @xcite the value @xmath111 was obtained . therefore , the @xmath0 intervals that we adopt in our studies are @xmath112 one notices that this result is more than @xmath0 away from the standard model bbn predicted value of @xmath105 , eq . ( [ eq:3 ] ) . using a subsample of the same data of @xcite , a different group had independently determined @xmath105 @xcite . from their studies we adopt ( eq . ( 8.2 ) of @xcite ) and leave the systematic error from @xcite . ] @xmath113 . as a result , @xmath114 ( this values of @xmath105 coincide with the standard bbn one , ( [ eq:3 ] ) , at about @xmath115 level ) .. its upper bound becomes very close to that of ( [ eq:7 ] ) if one employs an additional systematic uncertainty at the level @xmath116 ( twice the value of systematic uncertainty of @xcite ) . ] second method of determination of helium abundance is based on the cmb measurements . this method is believed to determine truly pristine value of @xmath105 , not prone to the systematics of astrophysical methods . however currently its uncertainties are still much larger than of the first method . the present measurements put it at @xmath117 again consistent with the standard model bbn at @xmath118 . here @xmath119 is the so - called effective number of neutrino species @xmath120 proportional to the ratio of the total energy , deposited into the active neutrino species to that of photons . notice , that the bound ( [ eq : yp - cmb - neff3 ] ) is based on assumption that before the onset of the recombination epoch the effective number of neutrino species is close to its sm value @xmath121 . as we will see later , sterile neutrinos can significantly distort @xmath119 . for the values of @xmath119 strongly deviating from @xmath122 the cmb bounds on @xmath105 gets modified . for example , the analysis carried out in @xcite reveals that @xmath123 the similar conclusion is reached if one employs the data of @xcite . the other element produced during the bbn is the deuterium , and recent observations determine its abundance to be @xmath124 this value is sensitive both to the baryon - to - photon ratio and to @xmath125 . in this work we adjust the value of baryon - to - photon ratio @xmath126 at the beginning of the computation so that by @xmath127 it is equal to the value given by cosmic microwave background measurements @xcite . finally , we mention another important uncertainty originating from the particle physics . there are two different measurements of neutron lifetime @xmath110 that are at tension with each other . particle data group @xcite provides @xmath128 , while measurements performed by serebrov et al . @xcite result in @xmath129 . we employ both results and explore the differences they lead to in what follows . in this section we present our main results : the bounds on sterile neutrino lifetime as a function of their masses and mixing patterns , as well as the bounds on the mixing angles . as discussed in the previous section , there are several systematic uncertainties in the determination of the @xmath12he abundance and therefore the results will depend on the adopted values of @xmath105 ( together with the neutron lifetime , @xmath110 ) . we summarize these systematic effects below . we start with comparing the upper bounds on sterile neutrino lifetime for different values of @xmath105 ( see section [ sec : abundances ] ) . the fig . [ fig : astro - bounds ] shows that the bounds from the two recent works @xcite are quite similar ( the difference is of the order of 30% ) . the bound , based on @xcite would give a result , similar to @xcite as discussed above . for the cmb bound in fig . [ fig : astro - bounds ] , we present only the results for masses @xmath130 where @xmath119 does not deviate significantly from @xmath122 . fig . [ fig : neff ] indicates that for smaller masses the number of effective neutrino species increases significantly . it in turn affects the cmb helium bounds ( c.f . ( [ eq : yp - cmb - neff3 ] ) and ( [ eq : neff - definition ] ) ) . the accurate account of this effect goes beyond the scope of this work and we choose instead to plot stronger deuterium - based bounds ( those of fig . [ fig : neff ] ) in fig . [ fig : astro - bounds ] for @xmath131 . the lower bound on @xmath105 from the recent work of @xcite is above the standard bbn value ( [ eq:3 ] ) at @xmath132 level ( see however @xcite ) . the presence of sterile neutrinos in plasma of course relaxes this tension and therefore at @xmath0 the adopted values of @xmath105 ( eq . [ eq:6 ] ) provide both upper and _ lower _ bounds on sterile neutrino lifetime . this is shown in fig . [ fig : izotov ] , right panel . at @xmath133 level the measurements of @xcite are consistent with standard bbn and the lower bound disappears . fig . [ fig : neff ] shows the changes in deuterium abundance and in the effective number of neutrino species , caused by sterile neutrinos ( with parameters corresponding to the upper bound based on @xcite ) . for these values of parameters the abundance lies within the @xmath133 boundaries ( [ eq : deuterium - observations ] ) . and for the highest effective number of neutrinos reached , @xmath134 , @xmath135 is close to the @xmath133 upper bound . notice that the same relation between @xmath125and @xmath135 is observed in the model without new particles but with the effective number of neutrinos different from @xmath122 . the effective number of neutrino species does not define the helium abundance though . otherwise the same @xmath105 bound @xcite would predict _ only one particular _ value of @xmath125 , which is not case , as the inspection of fig . [ fig : neff ] shows . range , based on @xcite . * right : * effective number of neutrino species ( the ratio of the effective neutrino temperature to the photon temperature at @xmath136 ) as a result of decay of sterile neutrino . the horizontal `` sm '' lines indicate @xmath119 that corresponds to the boundary of the @xmath133 range @xcite , in the sm with the number of relativistic species deviating from @xmath121 . in both panels , parameters of sterile neutrinos correspond to the upper bound on @xmath105 from @xcite ( see eq . ( [ eq:6 ] ) ) , except of the `` cmb '' line that corresponds to the upper bound from @xcite ( see eq . ( [ eq : yp - cmb - neff3 ] ) ) . , title="fig : " ] range , based on @xcite . * right : * effective number of neutrino species ( the ratio of the effective neutrino temperature to the photon temperature at @xmath136 ) as a result of decay of sterile neutrino . the horizontal `` sm '' lines indicate @xmath119 that corresponds to the boundary of the @xmath133 range @xcite , in the sm with the number of relativistic species deviating from @xmath121 . in both panels , parameters of sterile neutrinos correspond to the upper bound on @xmath105 from @xcite ( see eq . ( [ eq:6 ] ) ) , except of the `` cmb '' line that corresponds to the upper bound from @xcite ( see eq . ( [ eq : yp - cmb - neff3 ] ) ) . , title="fig : " ] the influence of another systematic uncertainty ( the lifetime of neutron , @xmath110 ) is negligible . indeed , the relative difference between sterile neutrino lifetimes were found to be of the order of @xmath137 for two choices of @xmath110 from @xcite and from @xcite ( taking the same @xmath105 bound from @xcite ) . next we investigate the dependence of the resulting bounds on the mixing patterns of sterile neutrinos . naively , one would expect that sterile neutrinos mixing `` only with @xmath138 '' and `` only with @xmath139 '' should have different effect of @xmath105 . however , it is the energy `` injection '' rate ( i.e. the overall decay rate of sterile neutrinos ) that is _ more important _ for the dynamics of plasma before the onset of nucleosynthesis . this quantity depends on the lifetime @xmath29 and the mass @xmath140 of the neutrino . mixing patterns affect mostly the concentration of particular decay products , but not the injection rate . in addition , the neutrino oscillations ( fast at the bbn epoch ) make the difference between flavours less pronounced ( see appendix [ app : neutrino - oscillations ] ) . as a result , mixing patterns give essentially the same results with the difference at the level of tens of per cent ( see figs . [ fig : mixing - patterns ] , [ fig : mixing - bounds ] ) . -only ( red dashed line ) , @xmath139-only ( green dashed - dotted line ) and equal mixing with @xmath138 and @xmath139 flavours ( black solid line ) . all bounds are derived for the lifetime of neutron @xmath110 adopted from @xcite . the effect of different mixing patterns is at the level @xmath141 and can only be seen in the right panel because of the different @xmath56 axis . in the right panel , only the masses @xmath130 are presented . for details , see sec . [ sec : results ] and fig . [ fig : neff ] . , title="fig : " ] -only ( red dashed line ) , @xmath139-only ( green dashed - dotted line ) and equal mixing with @xmath138 and @xmath139 flavours ( black solid line ) . both types of bounds are derived by assuming lifetime of the neutron @xmath110 from @xcite . in the right panel , only the masses @xmath130 are presented . for details , see sec . [ sec : results ] and fig . [ fig : neff].,title="fig : " ] in this work we considered the influence of decaying particles with the masses few mev 140 mev on the primordial abundance of light elements ( @xmath142 and @xmath12he ) . such particles appear in many cosmological scenarios @xcite . particularly , we concentrated on the properties of sterile neutrinos and derived constraints on their lifetime imposed by the present measurements of primordial helium abundance @xmath105 . sterile neutrinos are super - weakly interacting particles , quadratically mixed with the active flavours . we analyzed the case of one majorana sterile neutrino with 4 degrees of freedom ( if sterile neutrinos were kept in thermal equilibrium it would be equivalent to @xmath143 species of active neutrinos ) . since the plasma evolution is mostly affected by the overall decay rate of sterile neutrinos , the lifetime bounds that we obtained are essentially independent of the particular mixing patterns , as figs . [ fig : mixing - patterns],[fig : mixing - bounds ] demonstrate . in the paper @xcite a similar model was considered with one _ dirac _ sterile neutrino . dirac sterile neutrino has the same 4 degrees of freedom and influences primordial plasma in the same way ( if it has the same spectrum , lifetime and mixing pattern ) . however , in @xcite effect of active - neutrino oscillations was not taken into account , and some simplifying approximations like boltzmann statistics were employed . to provide corresponding analysis we wrote code that solves more accurate boltzmann equations describing kinetics of neutrino than what were used in @xcite . we compare the results of this work with the previous bounds @xcite in fig . [ fig : dolgov ] . we see that our results are broadly consistent with the previous works . the differences for a given mixing pattern of sterile neutrinos can be as large as a factor of 2.5 for some masses . the presence of sterile neutrinos in the plasma affects the effective number of neutrino degrees of freedom , @xmath125 . [ fig : neff ] , right panel shows that @xmath125between @xmath144 and @xmath145 are possible for different mixing angles and masses , which could explain a larger than 3 values of @xmath146 , reported recently in several cmb observations ( see e.g. @xcite , but also @xcite ) . decaying sterile neutrinos with the masses @xmath147 and lifetimes from seconds to minutes and their influence on @xmath148 and entropy production have been recently considered in @xcite ( see also @xcite ) where it was demonstrated that they can lead to @xmath149 and can therefore be probed with the cmb measurements . the results of the present work demonstrate that in the region @xmath150 where we overlap with the parameter space , studied in @xcite , the primordial nucleosynthesis restricts the lifetime of sterile neutrinos to be well below 1 sec ( see fig . [ fig : izotov ] , left panel ) . finally , it is interesting to compare the upper bound on sterile neutrino lifetime , derived in this paper with the _ lower bounds _ that come from direct experimental searches for sterile neutrinos ( see @xcite ) . these latter bounds are based on the assumption that sterile neutrinos with four degrees of freedom are solely responsible for the observed pattern of neutrino oscillations via the see - saw mechanism @xcite . the appropriate comparison , based on @xcite , is presented in fig . [ fig : lifetime - bbn ] . no allowed values of sterile neutrino lifetimes for @xmath151 exist for either type of neutrino mass hierarchy ( i.e. the upper bound is _ smaller _ than the lower bound , see the purple double - shaded region in fig . [ fig : lifetime - bbn ] ) . notice , that if the astrophysical bounds on helium @xcite were used for @xmath152 in fig . [ fig : lifetime - bbn ] , instead of the cmb bound , the resulting lifetime bounds would become stronger ( by as much as a factor of 4 ) in this mass range . _ we stress that for this conclusion it is essential _ that mev sterile neutrinos are responsible for neutrino oscillations . for example , a model in which sterile neutrinos couple to @xmath153 _ only _ ( and therefore do not contribute to the mixing between active neutrino flavours ) , _ is allowed _ even if one confronts the strongest bbn bounds ( based on the astrophysical helium measurements ) with the direct accelerator bounds , see fig . [ fig : lifetime - tau ] for details . _ lower _ bounds on the lifetime of sterile neutrinos @xcite ( solid line ) , combined with the _ upper _ bounds from this work ( dashed line ) , corresponding to the _ weakest _ bound in fig . [ fig : astro - bounds ] . the accelerator bounds are for two majorana sterile neutrinos solely responsible for neutrino oscillations . * left : * normal hierarchy , * right : * inverted hierarchy . combination of bbn bounds with direct experimental searches demonstrates that sterile neutrinos with the masses in 1 - 140 mev range , solely responsible for neutrino oscillations are ruled out . see secs . [ sec : results],[sec : discussion ] for details.,title="fig : " ] _ lower _ bounds on the lifetime of sterile neutrinos @xcite ( solid line ) , combined with the _ upper _ bounds from this work ( dashed line ) , corresponding to the _ weakest _ bound in fig . [ fig : astro - bounds ] . the accelerator bounds are for two majorana sterile neutrinos solely responsible for neutrino oscillations . * left : * normal hierarchy , * right : * inverted hierarchy . combination of bbn bounds with direct experimental searches demonstrates that sterile neutrinos with the masses in 1 - 140 mev range , solely responsible for neutrino oscillations are ruled out . see secs . [ sec : results],[sec : discussion ] for details.,title="fig : " ] _ only_. unlike the case , presented in fig . [ fig : lifetime - bbn ] there is an allowed region of parameter space for most of the masses below 140 mev . ] we would like to thank a. boyarsky , d. gorbunov , s. hansen , d. semikoz , m. shaposhnikov for valuable discussions and for help and encouragement during various stages of this project . we specially thank d. semikoz for sharing with us the original version of the code , used in @xcite and j. racle for writing an initial version of the bbn code as a part of his master s project @xcite at epfl . is also grateful to s. vilchynskiy , scientific and educational centre of the bogolyubov institute for theoretical physics in kiev , ukraine and to ukrainian virtual roentgen and gamma - ray observatory virgo.ua . the work of a.i . was supported in part from the swiss - eastern european cooperation project ( scopes ) no . iz73z0_128040 of swiss national science foundation . a.i . acknowledges support from the erc advanced grant 2008109304 . .values of helium abundance @xmath105 in the standard model bbn ( sbbn ) and their dependence on the neutron lifetime , @xmath110 . [ cols="<,^,^",options="header " , ] the active neutrinos of different flavours @xmath154 are related to the mass eigen - state basis @xmath155 via a non - diagonal pontecorvo - maki - nakagava - sakata ( pmns ) matrix @xmath156 @xmath157 ( see e.g. @xcite for reviews ) : @xmath158 here @xmath159 and @xmath160 are functions of the active - active neutrino mixing angles @xmath161 . exact treatment of active neutrino oscillation in the early universe is a difficult task ( see e.g. @xcite ) characteristic timescale of oscillation between @xmath63 and @xmath162 mass eigen - states for a neutrino with energy @xmath163 is @xcite @xmath164 average energy of relativistic fermi particles in equilibrium is @xmath165 @xcite . applying this relation to active neutrinos and using their measured mass differences @xcite @xmath166 , we obtain @xmath167 provided that influence of the surrounding environment on neutrino propagation is neglected . one sees therefore that about the moment active neutrino decouples @xmath168 typical oscillation timescales are much smaller than the hubble expansion time given by eq . ( [ eq : friedmann - first ] ) @xmath169 here @xmath170 ( at @xmath171 ) @xcite is the so - called number of relativistic species that enters energy - temperature relation @xmath172 . therefore , active neutrinos oscillate many times between the subsequent reactions involving them . in quantitative terms it means that probabilities @xmath173 to transform from flavour @xmath20 to flavour @xmath174 are oscillating functions of time . in realistic situation neutrinos do not have a definite momentum but are created in wave packets that are superpositions of states which have one . since oscillation periods are momentum - dependent according to eq . ( [ eq : oscillation - timescale ] ) , each state in the superposition will have his own period . therefore after sufficiently many periods initial phases characterizing superposition will change , and there is no reason for the phase changes to be correlated with each other . so the decoherence of states is what happens . this phenomenon can be described effectively by averaging transition probabilities @xmath173 over time . resulting expressions are @xcite to understand what happens with a neutrino , consider example of electron - neutrino created in electron - positron annihilation . at the production time this particle has probability 1 to oscillate into @xmath138 and zero for other final state . after long enough time for many oscillations to happen and before the time when a collision with other particle becomes quite probable , the decoherence comes into play . so now we may find the @xmath138 with probability @xmath176 , @xmath139 with probability @xmath177 and @xmath153 with @xmath178 . the production rate of the initial specimen per unit time is proportional to collision integral @xmath179 , according to the boltzmann equation ( [ eq : boltzmann - equation ] ) . but the actual number of produced electron neutrinos is actually reduced by factor @xmath176 . and even if ( consider this hypothetical situation ) muon neutrino does not interact with plasma , it will be anyway produced , at rate @xmath180 . generalization to other neutrino flavours leads us to conclusion that the modified boltzmann equation @xmath181 describes neutrino dynamics correctly ( that is not the case for the initial equation ( [ eq : boltzmann - equation ] ) ) . for the actual computations we use the following experimental best - fit values : @xmath182 from @xcite , and @xmath183 from the daya bay @xcite . the latter number is close to the result @xmath184 indicated by another recent experiment , reno @xcite . however , in dense medium oscillations proceed differently due to considerable effects of the plasma on properties of a single particle . still , the phenomenon can be described by the formalism of the pmns matrix . the difference is that mixing parameters together with masses now depend on properties of the environment . in case of plasma close to equilibrium with no non - trivial conserving charges present the parameter describing it is the temperature . so the parameters of the pmns become _ temperature - dependent_. in language of the effective hamiltonian approach the system of three neutrinos is described by the addition of medium potential @xmath185 to the hamiltonian @xmath186 of the system in vacuum @xcite the medium potential comprises effects of neutrino interactions . since neutrinos take part only in charged- and neutral - current interactions , matter potential has two terms @xmath189 and @xmath190 , respectively . all neutrinos couple to neutral currents identically , so @xmath190 is proportional to unit matrix . therefore this term just renormalizes energy , and does not affect oscillations . in contrast , the charged - current term is non - diagonal and is present only for @xmath138 . the reason is that due to abundance of electrons in plasma , @xmath138 couples effectively to charged currents , while at temperatures below the muon s mass there is no significant contribution of muons and tau - leptons to realize coupling of other neutrinos to w boson . so far we have dropped sterile neutrinos from consideration . but their mixing properties are also altered in hot plasma . using the approach of the effective hamiltonian for them , one finds that their effective mixing angles in medium @xmath193 differ from that in vacuum @xmath194 as @xcite @xmath195 for small mixing angles @xmath194 . therefore the mixing angle is not altered significantly for sterile neutrinos and matter effects are negligible for their dynamics . y. izotov and t. thuan , _ the primordial abundance of 4he : evidence for non - standard big bang nucleosynthesis _ , _ astrophys.j . _ * 710 * ( 2010 ) l67l71 [ http://arxiv.org/abs/1001.4440[1001.4440 ] ] . e. komatsu , k. m. smith , j. dunkley , c. l. bennett , b. gold , g. hinshaw , n. jarosik , d. larson , m. r. nolta , l. page , d. n. spergel , m. halpern , r. s. hill , a. kogut , m. limon , s. s. meyer , n. odegard , g. s. tucker , j. l. weiland , e. wollack and e. l. wright , _ seven - year wilkinson microwave anisotropy probe ( wmap ) observations : cosmological interpretation _ , _ _ * 192 * ( feb . , 2011 ) 18+ [ http://arxiv.org/abs/1001.4538[1001.4538 ] ] . r. alpher , h. bethe and g. gamow , _ the origin of chemical elements _ , _ phys.rev . _ * 73 * ( 1948 ) 803804 . f. iocco , g. mangano , g. miele , o. pisanti and p. d. serpico , _ primordial nucleosynthesis : from precision cosmology to fundamental physics _ , _ phys . * 472 * ( 2009 ) 176 [ http://arxiv.org/abs/0809.0631[0809.0631 ] ] . g. steigman , _ primordial nucleosynthesis in the precision cosmology era _ , _ ann . rev . nucl . part . * 57 * ( 2007 ) 463491 [ http://arxiv.org/abs/0712.1100[0712.1100 ] ] . m. pospelov and j. pradler , _ big bang nucleosynthesis as a probe of new physics _ , _ ann . rev . nucl . part . * 60 * ( 2010 ) 539568 [ http://arxiv.org/abs/1011.1054[1011.1054 ] ] . a. boyarsky , o. ruchayskiy and m. shaposhnikov , _ the role of sterile neutrinos in cosmology and astrophysics _ , _ ann . rev . nucl . part . _ * 59 * ( 2009 ) 191 [ http://arxiv.org/abs/0901.0011[0901.0011 ] ] . a. kusenko , _ sterile neutrinos : the dark side of the light fermions _ , * 481 * ( 2009 ) 128 [ http://arxiv.org/abs/0906.2968[0906.2968 ] ] . e. k. akhmedov , v. a. rubakov and a. y. smirnov , _ baryogenesis via neutrino oscillations _ , _ phys . _ * 81 * ( 1998 ) 13591362 [ http://arxiv.org/abs/hep-ph/9803255[hep-ph/9803255 ] ] . t. asaka , s. blanchet and m. shaposhnikov , _ the numsm , dark matter and neutrino masses _ , _ phys . * b631 * ( 2005 ) 151156 [ http://arxiv.org/abs/hep-ph/0503065[hep-ph/0503065 ] ] . t. asaka and m. shaposhnikov , _ the numsm , dark matter and baryon asymmetry of the universe _ , _ phys . lett . b _ * 620 * ( july , 2005 ) 1726 [ http://arxiv.org/abs/arxiv:hep-ph/0505013 [ arxiv : hep - ph/0505013 ] ] . m. shaposhnikov , _ the numsm , leptonic asymmetries , and properties of singlet fermions _ , _ jhep _ * 08 * ( 2008 ) 008 [ http://arxiv.org/abs/0804.4542[0804.4542 ] ] . l. canetti and m. shaposhnikov , _ baryon asymmetry of the universe in the numsm _ , _ jcap _ * 1009 * ( 2010 ) 001 [ http://arxiv.org/abs/1006.0133[1006.0133 ] ] . m. laine and m. shaposhnikov , _ sterile neutrino dark matter as a consequence of @xmath22msm - induced lepton asymmetry _ , _ jcap _ * 6 * ( june , 2008 ) 31+ [ http://arxiv.org/abs/arxiv:0804.4543 [ arxiv:0804.4543 ] ] . a. d. dolgov , s. h. hansen , g. raffelt and d. v. semikoz , _ cosmological and astrophysical bounds on a heavy sterile neutrino and the karmen anomaly _ , _ nucl * b580 * ( 2000 ) 331351 [ http://arxiv.org/abs/hep-ph/0002223[hep-ph/0002223 ] ] . a. d. dolgov , s. h. hansen , g. raffelt and d. v. semikoz , _ heavy sterile neutrinos : bounds from big - bang nucleosynthesis and sn 1987a _ , _ nucl . * b590 * ( 2000 ) 562574 [ http://arxiv.org/abs/hep-ph/0008138[hep-ph/0008138 ] ] . c. j. smith , g. m. fuller and m. s. smith , _ big bang nucleosynthesis with independent neutrino distribution functions _ , _ _ * d79 * ( 2009 ) 105001 [ http://arxiv.org/abs/0812.1253[0812.1253 ] ] . g. m. fuller , c. t. kishimoto and a. kusenko , _ heavy sterile neutrinos , entropy and relativistic energy production , and the relic neutrino background _ , http://arxiv.org/abs/1110.6479[1110.6479 ] . a. dolgov , s. hansen and d. semikoz , _ impact of massive tau neutrinos on primordial nucleosynthesis . exact calculations _ , _ nucl.phys . _ * b524 * ( 1998 ) 621638 [ http://arxiv.org/abs/hep-ph/9712284 [ hep - ph/9712284 ] ] . a. dolgov and d. kirilova , _ nonequilibrium decays of light particles and the primordial nucleosynthesis _ , _ int.j.mod.phys . _ * a3 * ( 1988 ) 267 . m. kawasaki , p. kernan , h .- s . kang , r. j. scherrer , g. steigman _ et . _ , _ big bang nucleosynthesis constraints on the tau - neutrino mass _ , _ nucl.phys . _ * b419 * ( 1994 ) 105128 . m. kawasaki , k. kohri and n. sugiyama , _ mev scale reheating temperature and thermalization of neutrino background _ , _ phys.rev . _ * d62 * ( 2000 ) 023506 [ http://arxiv.org/abs/astro-ph/0002127 [ astro - ph/0002127 ] ] . s. hannestad , _ what is the lowest possible reheating temperature ? _ , _ phys * d70 * ( 2004 ) 043506 [ http://arxiv.org/abs/astro-ph/0403291[astro-ph/0403291 ] ] . a. d. dolgov , s. h. hansen and d. v. semikoz , _ nonequilibrium corrections to the spectra of massless neutrinos in the early universe . ( addendum ) _ , _ nucl . * b543 * ( 1999 ) 269274 [ http://arxiv.org/abs/hep-ph/9805467[hep-ph/9805467 ] ] . a. d. dolgov , s. h. hansen and d. v. semikoz , _ non - equilibrium corrections to the spectra of massless neutrinos in the early universe _ , _ nucl . * b503 * ( 1997 ) 426444 [ http://arxiv.org/abs/hep-ph/9703315[hep-ph/9703315 ] ] . d. gorbunov and m. shaposhnikov , _ how to find neutral leptons of the numsm ? _ , _ jhep _ * 10 * ( 2007 ) 015 [ http://arxiv.org/abs/arxiv:0705.1729 [ hep - ph][arxiv:0705.1729 [ hep - ph ] ] ] . collaboration , k. nakamura _ et . _ , _ review of particle physics _ , _ j.phys.g _ * g37 * ( 2010 ) 075021 . a. serebrov and a. fomin , _ neutron lifetime from a new evaluation of ultracold neutron storage experiments _ , _ phys.rev . _ * c82 * ( 2010 ) 035501 [ http://arxiv.org/abs/1005.4312[1005.4312 ] ] . j. lesgourgues and s. pastor , _ cosmological implications of a relic neutrino asymmetry _ , _ _ * 60 * ( nov . , 1999 ) 103521+ [ http://arxiv.org/abs/hep-ph/9904411[hep-ph/9904411 ] ] . p. d. serpico and g. g. raffelt , _ lepton asymmetry and primordial nucleosynthesis in the era of precision cosmology _ , _ phys . rev . _ * d71 * ( 2005 ) 127301 [ http://arxiv.org/abs/astro-ph/0506162 [ astro - ph/0506162 ] ] . c. j. smith , g. m. fuller , c. t. kishimoto and k. n. abazajian , _ light element signatures of sterile neutrinos and cosmological lepton numbers _ , _ phys.rev . _ * d74 * ( 2006 ) 085008 [ http://arxiv.org/abs/astro-ph/0608377[astro-ph/0608377 ] ] . g. mangano , g. miele , s. pastor , o. pisanti and s. sarikas , _ constraining the cosmic radiation density due to lepton number with big bang nucleosynthesis _ , _ jcap _ * 1103 * ( 2011 ) 035 [ http://arxiv.org/abs/1011.0916[1011.0916 ] ] . e. kolb and m. turner , _ the early universe_. addison - wesley , reading , ma , usa , 1990 . prepared with latex . f. bezrukov , d. gorbunov and m. shaposhnikov , _ on initial conditions for the hot big bang _ , _ jcap _ * 0906 * ( 2009 ) 029 [ http://arxiv.org/abs/0812.3622[0812.3622 ] ] . l. kawano , _ let s go : early universe . guide to primordial nucleosynthesis programming _ , . l. kawano , _ let s go : early universe . primordial nucleosynthesis : the computer way _ , . j. kernan and l. m. krauss , _ refined big bang nucleosynthesis constraints on omega ( baryon ) and n ( neutrino ) _ , _ phys.rev.lett . _ * 72 * ( 1994 ) 33093312 [ http://arxiv.org/abs/astro-ph/9402010 [ astro - ph/9402010 ] ] . r. n. boyd , c. r. brune , g. m. fuller and c. j. smith , _ new nuclear physics for big bang nucleosynthesis _ , _ phys.rev . _ * d82 * ( 2010 ) 105005 [ http://arxiv.org/abs/1008.0848[1008.0848 ] ] . g. m. fuller and c. j. smith , _ nuclear weak interaction rates in primordial nucleosynthesis _ , _ phys.rev . _ * d82 * ( 2010 ) 125017 [ http://arxiv.org/abs/1009.0277[1009.0277 ] ] . a. coc , s. goriely , y. xu , m. saimpert and e. vangioni , _ standard big - bang nucleosynthesis up to cno with an improved extended nuclear network _ , _ astrophys.j . _ * 744 * ( 2012 ) 158 [ http://arxiv.org/abs/1107.1117[1107.1117 ] ] . s. sarkar , _ big bang nucleosynthesis and physics beyond the standard model _ , _ _ * 59 * ( 1996 ) 14931610 [ http://arxiv.org/abs/hep-ph/9602260[hep-ph/9602260 ] ] . dedicated to dennis sciama on his 67th birthday . o. pisanti , a. cirillo , s. esposito , f. iocco , g. mangano _ et . al . _ , _ parthenope : public algorithm evaluating the nucleosynthesis of primordial elements _ , _ comput.phys.commun . _ * 178 * ( 2008 ) 956971 [ http://arxiv.org/abs/0705.0290[0705.0290 ] ] . e. aver , k. a. olive and e. d. skillman , _ an mcmc determination of the primordial helium abundance _ , http://arxiv.org/abs/1112.3713 [ 1112.3713 ] . g. mangano and p. d. serpico , _ a robust upper limit on @xmath148 from bbn , circa 2011 _ , _ phys.lett . _ * b701 * ( 2011 ) 296299 [ http://arxiv.org/abs/1103.1261[1103.1261 ] ] . m. peimbert , v. luridiana and a. peimbert , _ revised primordial helium abundance based on new atomic data _ , _ astrophys.j . _ * 666 * ( 2007 ) 636646 [ http://arxiv.org/abs/astro-ph/0701580 [ astro - ph/0701580 ] ] . j. dunkley , r. hlozek , j. sievers , v. acquaviva , p. ade _ et . _ , _ the atacama cosmology telescope : cosmological parameters from the 2008 power spectra _ , _ astrophys.j . _ * 739 * ( 2011 ) 52 [ http://arxiv.org/abs/1009.0866[1009.0866 ] ] . r. keisler , c. reichardt , k. aird , b. benson , l. bleem _ et . _ , _ a measurement of the damping tail of the cosmic microwave background power spectrum with the south pole telescope _ , _ astrophys.j . _ * 743 * ( 2011 ) 28 [ http://arxiv.org/abs/1105.3182[1105.3182 ] ] . g. gelmini , s. palomares - ruiz and s. pascoli , _ low reheating temperature and the visible sterile neutrino _ , _ phys . * 93 * ( 2004 ) 081302 [ http://arxiv.org/abs/astro-ph/0403323 [ astro - ph/0403323 ] ] . g. b. gelmini , e. osoba and s. palomares - ruiz , _ inert - sterile neutrino : cold or warm dark matter candidate _ , http://arxiv.org/abs/0912.2478[0912.2478 ] . g. m. fuller , a. kusenko and k. petraki , _ heavy sterile neutrinos and supernova explosions _ , _ phys.lett . _ * b670 * ( 2009 ) 281284 [ http://arxiv.org/abs/0806.4273[0806.4273 ] ] . b. benson , t. de haan , j. dudley , c. reichardt , k. aird _ et . _ , _ cosmological constraints from sunyaev - zeldovich - selected clusters with x - ray observations in the first 178 square degrees of the south pole telescope survey _ , http://arxiv.org/abs/1112.5435[1112.5435 ] . m. moresco , l. verde , l. pozzetti , r. jimenez and a. cimatti , _ new constraints on cosmological parameters and neutrino properties using the expansion rate of the universe to z 1.75 _ , http://arxiv.org/abs/1201.6658[1201.6658 ] . t. asaka , m. shaposhnikov and a. kusenko , _ opening a new window for warm dark matter _ , _ phys . lett . _ * b638 * ( 2006 ) 401406 [ http://arxiv.org/abs/hep-ph/0602150[hep-ph/0602150 ] ] . t. asaka , s. eijima and h. ishida , _ mixing of active and sterile neutrinos _ , _ jhep _ * 1104 * ( 2011 ) 011 [ http://arxiv.org/abs/1101.1382[1101.1382 ] ] . o. ruchayskiy and a. ivashko , _ experimental bounds on sterile neutrino mixing angles _ , http://arxiv.org/abs/1112.3319[1112.3319 ] . j. racle , _ deriving bounds on interactions of numsm sterile neutrinos using primordial nucleosynthesis _ , master s thesis , epfl , 2008 . g. mangano , g. miele , s. pastor , t. pinto , o. pisanti _ et . _ , _ relic neutrino decoupling including flavor oscillations _ , _ nucl.phys . _ * b729 * ( 2005 ) 221234 [ http://arxiv.org/abs/hep-ph/0506164 [ hep - ph/0506164 ] ] . a. strumia and f. vissani , _ neutrino masses and mixings and . _ , http://arxiv.org/abs/hep-ph/0606054[hep-ph/0606054 ] . a. dolgov , s. hansen , s. pastor , s. petcov , g. raffelt _ et . _ , _ cosmological bounds on neutrino degeneracy improved by flavor oscillations _ , _ nucl.phys . _ * b632 * ( 2002 ) 363382 [ http://arxiv.org/abs/hep-ph/0201287[hep-ph/0201287 ] ] . a. dolgov and f. villante , _ bbn bounds on active sterile neutrino mixing _ , _ nucl.phys . _ * b679 * ( 2004 ) 261298 [ http://arxiv.org/abs/hep-ph/0308083[hep-ph/0308083 ] ] . d. kirilova , _ non - equilibrium neutrino in the early universe plasma _ , _ aip conf.proc._ * 1121 * ( 2009 ) 8389 . t. schwetz , m. tortola and j. valle , _ where we are on @xmath196 : addendum to global neutrino data and recent reactor fluxes : status of three - flavour oscillation parameters _ , _ new j.phys._ * 13 * ( 2011 ) 109401 [ http://arxiv.org/abs/1108.1376[1108.1376 ] ] . collaboration , f. an _ et . _ , _ observation of electron - antineutrino disappearance at daya bay _ , http://arxiv.org/abs/1203.1669[1203.1669 ] . collaboration , j. ahn _ et . _ , _ observation of reactor electron antineutrino disappearance in the reno experiment _ , http://arxiv.org/abs/1204.0626[1204.0626 ] . d. notzold and g. raffelt , _ neutrino dispersion at finite temperature and density _ , _ nucl . * b307 * ( 1988 ) 924 .

based on rxte / asm and exosat / me data we studied x - ray variability of persistent lmxbs in the hz frequency range , aiming to detect features in their power density spectra ( pds ) associated with the viscous time scale of the accretion disk . as this is the longest intrinsic time scale of the disk , the power density of its variations is expected to be independent on the frequency at . indeed , in the pds of 11 sources out of 12 we found very low frequency break , below which the spectra are nearly flat . at higher frequencies they approximately follow the law . the break frequency correlates very well with the binary orbital frequency in a broad range of binary periods , in accord with theoretical expectations for the viscous time scale of the disk . however , the value of is at least by an order of magnitude larger than predicted by the standard disk theory . this suggests that a significant fraction of the accretion occurs through the optically thin and hot coronal flow with the aspect ratio of . the predicted parameters of this flow , k and are in qualitative agreement with recent chandra and xmm - newton observations of complex absorption / emission features in the spectra of lmxbs with high inclination angle . we find a clear dichotomy in the value of between wide and compact systems , the compact systems having times shorter viscous time . the boundary lies at the mass ration , suggesting that this dichotomy is caused by the excitation of the 3:1 inner lindblad resonance in low- lmxbs . accretion , accretion disks x - rays : binaries

astro-ph0501215

x - ray binaries show x - ray flux variations in a broad range of time scales ( fig.[fig : pds_cygx2 ] ) . their power density spectra often reveal a number of periodic and quasi - periodic phenomena , corresponding to the orbital frequency of the binary system , the spin frequency of the neutron star and quasi - periodic oscillation , which nature is still poorly understood . in addition , aperiodic variability is observed , giving rise to the broad band continuum component in power density spectra , extending from @xmath15 msec to the longest time scales accessible for monitoring instruments . the aperiodic variability is often characterized by the fractional rms values of @xmath16 tens of per cent , indicating that variations of the mass accretion rate @xmath2 in the broad range of time scales are present in the innermost region of the accretion flow , where x - ray emission is produced . from the point of view of characteristic time scales , the high frequency variations could potentially be produced in the vicinity of the compact object . longer time scales , on the contrary , exceed by many orders of magnitude the characteristic time scales in the region of the main energy release . hence , the low frequency @xmath2 variations have to be generated in the outer parts of the accretion flow and to be propagated to the region of the main energy release . due to the diffusive nature of the standard shakura - sunyaev disk @xcite , it plays the role of the low - pass filter , at any given radius @xmath17 suppressing @xmath2 variations on the time scales shorter than the local viscous time @xmath18 . the viscous time is , on the other hand , the longest time scale of the accretion flow . hence , it is plausible to suppose that any given radius @xmath17 contributes to @xmath2 and x - ray flux variations predominantly at the frequency @xmath19 @xcite . @xcite demonstrated that in this picture @xmath20 power density spectra can naturally appear with slope @xmath21 , in good agreement with observations . owing to the finite size of the accretion disk , the longest time scale which can appear in the disk is restricted by the viscous time on its outer boundary @xmath22 . below this frequency the x - ray flux variations are uncorrelated , therefore the power density spectrum of the accretion disk should become flat at @xmath23 . due to the mechanism suggested by @xcite are discussed in section [ sec : qpo ] ] if there are several components in the accretion flow each having a different viscous time scale ( e.g. geometrically thin disk and diffuse corona above it ) , several breaks can appear on the power spectra at the frequencies , corresponding to the inverse viscous time scale of each component . yet another source of @xmath2 variations might be the variability of the mass transfer rate from the donor star . as these variations also have to be propagated through the accretion disk , only low frequency perturbations , with @xmath24 , will reach the region of the main energy release . this might give rise to an independent continuum component in the power density spectrum at @xmath24 . the shape of the final power density spectrum of x - ray binary will depend on the relative amplitudes of the @xmath2 variations due to different components of the accretion flow and the donor star . as these are independent from each other , a distinct features should appear in the power density spectrum of x - ray binary at frequencies @xmath25 . power law.,scaledwidth=51.0% ] the viscous time scale of the accretion disk is @xcite @xmath26 where @xmath27 is the disk half thickness and @xmath28 is the keplerian frequency at the outer disk boundary @xmath29 . for a geometrically thin disk , @xmath30 , the viscous time scale is significantly longer than the keplerian time at the outer disk boundary . if the disk fills a large fraction of the roche - lobe of the primary , as is the case for steady disks in persistent lmxbs , @xmath31 , ( section [ sec : tvisc_h2r ] ) , the viscous time at the outer boundary @xmath22 will exceed the orbital period of the binary system @xmath32 as well . therefore , one might expect to find features associated with the viscous time of the accretion disk at frequencies below the orbital frequency of the binary system . detection of such features for a sample of lmxbs would give an opportunity to probe the accretion disk parameters , such as viscosity parameter @xmath33 and the disk thickness at its outer boundary . in this paper we study a sample of persistent low mass x - ray binaries aiming to detect such features in their power density spectra . longevity of typical orbital periods in lmxbs requires long monitoring observations to achieve this goal . this has become possible thanks to the long operations of the all sky monitor aboard rossi x - ray timing explorer , which provided long term light curves of hundreds of galactic sources covering period of @xmath34 years . some sources with rather short orbital period , @xmath35 few hours may be studied based on the exosat data . lcccccccc source & @xmath36 & @xmath37 & @xmath38 & @xmath39 & q & @xmath40 & d & ref . + & @xmath41 & @xmath41 & hrs & @xmath42 cm & & @xmath43 & kpc & + & ( 1 ) & ( 2 ) & ( 3 ) & ( 4 ) & ( 5 ) & ( 6 ) & ( 7 ) & ( 8) + + grs 1915 + 105 & @xmath44 & @xmath45 & @xmath46 & @xmath47 & @xmath48 & @xmath49 & @xmath50 & 1,2 + gx 13 + 1 & 1.4 & @xmath51 @xmath52 & 592.8 & @xmath53 & @xmath54 & & @xmath55 & 3 + cir x-1 & 1.4 & @xmath56 & 398.4 & @xmath57 & @xmath58 & & @xmath59 & 4 + cyg x-2 & @xmath60 & @xmath61 & 236.3 & @xmath62 & @xmath63 & @xmath64 & @xmath65 & 5 + gx 349 + 2 & 1.4 & @xmath66 & 22.5@xmath670.1 & @xmath68 & @xmath69 & & @xmath70 & 6 + sco x-1 & 1.4 & 0.42 & 18.92 & @xmath71 & @xmath72 & @xmath73 & @xmath74 & 7 + + exo 0748 - 676 & 1.4 & @xmath75 & @xmath76 & @xmath77 & @xmath78 & @xmath79 & @xmath80 & 8 + 4u 1636 - 536 & @xmath81 & @xmath82 & @xmath83 & @xmath84 & @xmath85 & & 5.9 & 9 + 4u 1323 - 619 & 1.4 & 0.26 & 2.93 & @xmath86 & @xmath87 & @xmath88 & @xmath89 & 9 + 4u 1916 - 053 & 1.4 & @xmath90 & 0.83 & @xmath91 & @xmath92 & @xmath93 & 10.0 & 10 + 1h 1627 - 673 & @xmath81 & @xmath94 & @xmath95 & @xmath96 & @xmath97 & @xmath98 & 8.0 11,12 + 4u 1820 - 303 & @xmath81 & @xmath99 & @xmath100 & @xmath101 & @xmath102 & & 8.0 & 13 + + the paper is structured as follows . in section [ sec : data ] we describe the selection criteria and sample of lmxbs . power density spectra and their approximations are presented in the section [ sec : pds ] . the viscous time scale of the accretion disks as obtained from the power spectra analysis and constrains on parameters of the outer disk are discussed in the section [ sec : tvisc ] . in section [ sec : discussion ] we consider the vertical structure of the semi - thick accretion flow , derive parameters of the disk corona and compare these with other evidence of the existence of diffuse coronal flow in lmxbs . our results are summarized in section [ sec : summary ] . in the appendix we describe the method used to calculate power density spectra from the asm light curves . we used data of the asm instrument of rxte observatory ( swank , 1999 ) , covering the period from 19962004 ( mjd range from @xmath103 ) . the asm instrument operates in the 212 kev energy range and performs flux measurements for over @xmath104 x - ray sources from the asm source catalog once per satellite orbit , i.e. every @xmath105 min . each flux measurement ( dwell ) has duration of @xmath105 sec . due to navigational constrains and appearance of very bright transient sources , the asm light curves for individual sources sometimes have gaps of duration of up to few months . the dwell - by - dwell light curves were retrieved from the public rxte / asm archive at heasarc . for sources with short orbital periods , @xmath106 hours , we used the data of the medium energy ( me ) detector of exosat satellite @xcite . it provided @xmath16 several tens of ksec long light curves with typical time resolution of @xmath107 sec in the 0.98.9 kev energy range . the exosat data were also retrieved from heasarc . we have selected low mass x - ray binaries , satisfying the following criteria : 1 . _ persistent sources _ , which light curves do not show off - states or x - ray novae - like outbursts . this ensures that the accretion disk was in the same state during the analyzed period . 2 . the _ orbital period _ @xmath32 is known . orbital parameters of x - ray sources were taken from the catalog of @xcite . + the nyquist frequency in the power density spectra obtained from the asm light curves , @xmath108 hz , correspond to period of @xmath109 hours . in practice , due to specifics of the asm light curves , the high frequency part of the power spectra can sometimes be distorted by the aliasing effect . to minimize the contribution of this effect we selected source with sufficiently long orbital periods , @xmath110 hours . + although the aliasing is not an issue for exosat light curves , their duration , typically limited by @xmath111 ksec , imposes a different constrains on the @xmath38 . as we are looking for features at the time scales larger than @xmath38 , we required that the light curve covers at least several @xmath38 . this leads to the constrain for the sources which can be studied using exosat light curves , @xmath112 hours . source brightness . _ we selected sources with the average count rate exceeding @xmath113 cnts / s for asm and @xmath114 cnts / s for exosat . the _ noise level _ , calculated for asm power spectra , although approximately correct , is not accurate enough , due to existence of unaccounted systematic errors in the flux measurements ( e.g. * ? ? ? * ) . to avoid additional uncertainties in the power spectra we considered only sources with sufficiently high signal at the orbital frequency , @xmath115 at @xmath116 for exosat light curves the noise level is not an issue . there are 6 sources from the asm catalog and 6 sources observed by exosat , satisfying the selection criteria of the section [ sec : selection ] . the final asm sample includes the following sources : grs 1915 + 105 , gx 13 + 1 , cir x-1 , cyg x-2 , gx 349 + 2 , sco x-1 the exosat sample includes : 4u 1323619 , 4u 1636536 , 4u 1820303 , 4u 1916053 , 1h 1627673 , exo 0748676 the binary system parameters for selected sources are listed in table [ tab : general ] and their x - ray light curves obtained by asm are shown in fig . [ fig : lc_asm ] . power density spectra were computed in the 2 - 12 kev ( asm ) and 0.98.9 kev ( exosat me ) energy range . the pds of the sources from asm sample were obtained using the method based on the autocorrelation function calculation as described in the appendix [ sec : method ] . the exosat light curves were analyzed with the aid of the _ powspec _ task from ftools 5.1 . in analyzing the exosat data we averaged the power spectra obtained in ( nearly ) all individual observations with adequate time resolution available in the public archive at heasarc . all but one source in the exosat sample are x - ray bursters . the presence of x - ray bursts in their light curves can results in appearance of an additional component in the pds , having no relation to the power spectrum of @xmath2 variations due to the accretion disk . to avoid this contamination , we screened the exosat light curves to exclude the time intervals corresponding to x - ray bursts . in the case of exo 0748676 we also excluded the time intervals corresponding to the x - ray eclipses . no attempt to screen out the x - ray dips has been done , due to ambiguity of their identification . the power spectra based on the asm data are not subject to a significant contamination due to x - ray bursts and dips / eclipses . [ fig : pds_asm ] [ fig : pds_exosat ] the power spectra of selected sources are shown in figs.[fig : pds_asm ] and [ fig : pds_exosat ] . for reference , we plot in each panel a power law spectrum @xmath117 . flattening at frequencies @xmath118 hz can be seen in the pds of grs 1915 + 105 . this can potentially be related to the specifics of asm light curves , namely , it can be caused by the aliasing effect leading to the leakage of higher frequency power below the nyquist frequency . the assessment of the reality of this flattening is beyond the scope of the present paper . also seen for some sources ( most prominently for cir x-1 and 4u1820303 ) are the peaks due to the orbital modulation . in fig.[fig : pds_broadband ] we show combined power spectra based on the asm and exosat light curves . these spectra cover @xmath119 hz frequency range . for several faint sources from the exosat sample no meaningful power spectra can be obtained from the asm light curves due to too low count rate . for grs1915 + 105 , obviously , no exosat data exist . the power spectra of almost all sources show a clear low frequency break , below which they are nearly flat . for some sources the second very low frequency component is present at lower frequencies . the most clearly this component can be seen in the case of cir x-1 , sco x-1 and 4u1916536 . some sources also show broad qpo like features near or below the break frequency . at high frequencies , above the break , the power spectra follow the @xmath120 power law . remarkably , the slope of the power law appears to be similar in all 12 sources , @xmath121 . the same is true for the normalization . there is at least one exception from the above behavior . no evidence for break was found in the pds of 4u 1636 - 536 . although the power spectrum of this source might have several weak features , its overall behavior in an extremely broad frequency range from @xmath122 hz is well represented by the power law with slope @xmath123 and without any obvious break ( fig.[fig : pds_broadband ] ) . the second possible exception is 4u 1820 - 303 . in addition to clearly visible peak due to the orbital modulations , it appears to have a shoulder at @xmath124 hz followed by a power law at lower frequencies . the presence of this shoulder is more obvious in the data of individual exosat observations , shown in fig.[fig : pds_4u1820 ] . in the formal statistical sense the existence of the low frequency break is highly significant . for example , approximation of the aug . 1920 , 1985 data in the @xmath125 hz frequency range gave the following values : @xmath126 ( 11 d.o.f . ) and @xmath127 ( 12 d.o.f . ) for the power law model with and without low frequency break respectively , resulting in @xmath128 for one additional parameter . however , considering the overall shape of the power spectrum ( e.g. fig.[fig : pds_broadband ] ) , this source presents a less obvious case than the others in our sample . @xcite analyzing long term x - ray variability of 4u 1820 - 303 confirmed earlier suggestion that it is a hierarchical triple system , where ultra - close x - ray binary is orbited by companion with orbital period about 1.1 day . due to influence of the third star the well - known 176 day x - ray modulation of 4u 1820 - 303 is generated . influence of the third star can also give rise to very low frequency modulation of the mass transfer rate in this system , unrelated to the @xmath2 variations intrinsic to the accretion disk . this could , in principle , explain strong very low frequency power law component at @xmath129 hz . in the following we associate the shoulder at @xmath130 with the low frequency break similar to the ones observed in other lmxbs in our sample . we emphasize that this interpretation is not unique and 4u 1820 - 303 , like 4u 1636 - 536 , can show the behavior different from other sources in our sample . power law . the vertical dashed line marks the orbital frequency.,scaledwidth=50.0% ] the power density spectra were approximated by a model of a power law with low frequency break : @xmath131 an additional lorentzian component was added to the model for the sources with strong orbital modulation and/or broad low frequency qpo peaks . the frequency range was chosen for each source individually , depending on the duration of the available time series and the presence of the additional noise component below @xmath132 . the adopted frequency ranges and best fit values of the model parameters are listed in table 2 , along with the count rates and x - ray luminosities of the sources . the large width of the qpo component present in the power spectra of several sources close to the low frequency end makes the determination of the break frequency somewhat ambiguous . for these sources ( grs1915 + 105 , cyg x-2 and 4u 1323619 ) we list the values of the break frequency obtained with and without the lorentzian component in the model . in the following the average of the frequencies of the two models the statistical uncertainty of the break frequency in these sources are obtained from combined confidence intervals of the models with and without lorentzian component . at least three harmonics of the orbital frequency are present in the power spectrum of 4u1323619 . the corresponding bins of the power spectrum were excluded from the fitting procedure . [ cols="<,^,^,^,^,^,^,^ " , ] + for the standard accretion disk ( shakura & sunyaev , 1973 ) the viscous time scale at the outer edge of the disk @xmath29 can be estimated as : @xmath133 where @xmath29 is the disk outer radius , @xmath134 is the keplerian frequency , @xmath27 is the disk half - thickness at the outer edge , @xmath36 is the mass of the primary and @xmath33 is the dimensionless viscosity parameter . combining the eq.([eq : tvisc ] ) with the third kepler law : @xmath135^{-1/2 } , \label{eq : kepler3}\ ] ] and assuming that the binary has a circular orbit one finds : @xmath136 where @xmath137 is the mass ratio , @xmath39 is binary separation , @xmath138 and @xmath139 . from eq.([eq : fvisc2forb ] ) it follows that for fixed viscosity parameter and relative disk thickness and size , the viscous time scale is directly proportional to the orbital frequency of the binary system , as intuitively expected . to estimate the expected range of values of @xmath140 one needs to know the disk size and thickness . the thickness of the standard shakura sunyaev disk is @xmath141 where @xmath142 is the mass of the compact object in solar units , @xmath143 cm is defined so that @xmath144 , @xmath145 erg / s is x - ray luminosity , and @xmath146 cm is the outer disk radius . the disk thickness predicted by this relation varies from @xmath147 for the most compact and low luminosity systems to @xmath148 for sources in the upper part of the table [ tab : general ] . irradiation of the disk by the x - rays from the vicinity of the compact object , although does increase significantly the surface temperature of the outer disk , has a little effect on its mid - plane temperature , due to large optical depth of the shakura - sunyaev disk @xcite . correspondingly , account for irradiation effects does not change the above estimates significantly . with plausible values of two other parameters in eq.([eq : fvisc2forb ] ) , @xmath149 and @xmath150 , the standard theory predicts : @xmath151 power density spectra of all but one lmxbs from our sample have the low frequency break predicted in the simple qualitative picture outlined in the introduction . [ fig : fbreak_forb ] shows the dependence of the best fit value of the break frequency on the orbital frequency of the binary system . there is an obvious positive correlation between these two frequencies broadly consistent with eq.([eq : fvisc2forb ] ) . remarkably , this correlation holds over 3 order of magnitudes in the orbital frequency , from compact binaries 4u 1820 - 303 and 1h1627673 with @xmath152 and @xmath153 min to extremely wide system grs 1915 + 105 with @xmath154 days . the value of @xmath6 ranges from @xmath155 for long period binaries to @xmath156 for the more compact ones . the broad band spectra shown in fig.[fig : pds_broadband ] confirm that the breaks are indeed unique and well defined features in the power spectra . in the following we assume that the break frequency corresponds to the viscous time scale of the accretion disk , and they are related by a simple linear relation : @xmath157 note that this simplified picture does not take into account complexity of the power density distribution near the viscous time scale . with this definition the break frequency is related to the orbital frequency via : @xmath158 cir x-1 appears to deviate from the general trend in fig.[fig : fbreak_forb ] , having the break / viscous time scale by a factor @xmath12 shorter than expected given its orbital period . the significant orbital modulation of the x - ray activity suggests a highly eccentric orbit in this binary system @xcite . @xcite proposed a model , where the system consists of a neutron star on a highly eccentric , @xmath159 orbit around a @xmath160 @xmath161 sub - giant companion . in this model the donor star fills its roche lobe and the mass transfer occurs only during the periastron passage . outside the periastron the donor star is detached from its roche lobe surface and the mass transfer in the binary stops . in such a system the disk radius would be defined , to first approximation , by the minimal separation between the components , @xmath162 . as the eq.([eq : fvisc2forb ] ) was derived for a circular orbit , the following substitution should be made in interpreting the cir x-1 data : @xmath163 plotted in fig.[fig : fbreak_forb ] are two point for cir x-1 . the open circle corresponds to the observed orbital period of the source , the filled circle corresponds to the orbital period corrected for the eccentricity of the orbit assuming @xmath164 . with this substitution the consistency with other sources is restored . although the break frequency does increase linearly with the binary orbital frequency , the ratio @xmath165 is notably larger than predicted for the shakura - sunyaev disk , @xmath166 . the latter range of values is shown as the hatched area in the fig . [ fig : fbreak_forb ] . obviously , larger values of @xmath140 imply that the disk viscous time as traced by the position of the break on the pds is by a factor of @xmath167 shorter than predicted by the theory . as it follows from eq.([eq : fvisc2forb ] ) and ( [ eq : fbr2forb ] ) , there are several parameters affecting the disk viscous time , of which the strongest dependence is upon the disk size and thickness . the theoretical and observational constrains on the disk size are summarized in fig.[fig : rdisk ] . it has two well known theoretical limits . due to the angular momentum conservation , the disk radius can not be smaller than the circularization radius . the results of the numerical calculations of @xcite can be approximated by @xmath168 this approximation is accurate to @xmath169 in the range of the mass ratios @xmath170 . the upper limit on the disk size is given by the tidal truncation radius @xcite . in the case of small pressure and viscosity its value is close to the radius of the largest non - intersecting periodic orbit @xcite . the latter can be approximated by : @xmath171 accurate to @xmath169 in the range of @xmath172 . these two limits on the disk radius are shown as thick dash - dotted lines in fig.[fig : rdisk ] . on the observational side , there is a number of spectroscopic and photometric determinations of the disk radius in cvs @xcite and fewer in lmxbs @xcite . these data indicate ( fig.[fig : rdisk ] ) that for a steady disk the disk radius is close to the tidal truncation radius as defined by @xcite , with possible exception of the small - q systems ( section [ sec : tvisc_lindblad ] ) . with these values of the disk size and accepting @xmath173 as a plausible value of the viscosity parameter , the disk thickness of @xmath174 is required in order explain observed values of the break frequency . this conclusion relies on the association of the break in the power density spectra with the viscous time scale of the disk and as long as the assumption @xmath175 is valid , it is very robust . indeed , in order to describe the data with @xmath176 , as predicted by the standard theory , one would need to increase dramatically the @xmath33-parameter , up to implausibly high values of @xmath177 . alternatively , short viscous time of the disk could be achieved by reducing the roche - lobe filling factor of the disk , down to @xmath178 . this is significantly smaller than the circularization radius and contradicts to both theoretical expectations and observations of the disks in cvs and lmxbs ( fig.[fig : rdisk ] ) . the latter possibility can be also interpreted that only the very inner part of the accretion disk , @xmath179 , contributes to the observed variability of lmxbs in x - rays . this can not be excluded a priori . we note however , that in this case the sharpness of the breaks observed in the power spectra of at least some sources ( e.g. sco x-1 , exo 0748676 , etc ) implies that the transition from the ( outer ) region of negligibly small amplitude of the @xmath2 perturbation to the ( inner ) region of relatively large ones should occur in a rather narrow range of radii . as is evident from fig.[fig : fbreak_forb ] , there is a significant dichotomy in the value of @xmath6 ratio between long- and short - period binaries . this is further illustrated by fig.[fig : fbr2forb ] where the ratio @xmath6 is plotted against the binary orbital period and the mass ratio . the average values of @xmath6 are @xmath180 and @xmath181 for the wide and compact systems correspondingly . the sense of the difference is that the compact systems have systematically shorter , by a factor of @xmath182 , viscous time expressed in the orbital periods than the wide ones , the boundary lying at @xmath183 hours or @xmath184 . the fact that this bimodality occurs between compact and wide systems or , equivalently , ( with the exception of grs1915 + 106 ) between small- and large - q binaries suggests that it may be caused by the excitation of tidal resonances in the accretion disk . the phenomenon of tidal resonances in accretion disks is well studied in the context of dwarf novae @xcite . the resonance occurs if the angular frequency of the orbital motion of the particle in the disk is commensurate with the angular frequency of the orbital motion of the secondary . for a keplerian disk , the location of the resonant orbits is given by @xmath185 where @xmath186 are integers ( e.g. * ? ? ? an obvious condition for a resonance being excited is that this radius lies within the accretion disk . from analytical studies and numerical simulations it is known that the strongest resonance occurs at the lowest order commensurability , @xmath187 ( e.g. * ? ? ? * ) . however the 2:1 resonant radius lies within the accretion disk only at extreme values of the mass ratio , @xmath188 . the next strongest resonance , most important in the context of binaries with not too extreme values of the mass ratio lies at the 3:1 commensurability . the higher order resonances are probably not excited in the accretion disks . the dependence of the 3:1 resonance radius on the mass ratio @xmath14 is shown by the dashed line in fig.[fig : rdisk ] , confirming the well known fact that the @xmath189 resonance can be excited in the systems with the mass ration @xmath190 . note that the precise value of the threshold @xmath14 depends on the definition of the tidal truncation radius , due to small angle between the two curves in fig.[fig : rdisk ] . under the assumption that @xmath191 it is @xmath192 . with the definition of @xmath193 used in fig.[fig : rdisk ] the threshold value is slightly larger . numerical simulations of the non - linear stage of the instability have shown that excitation of tidal resonances results in a significant asymmetry of the accretion disk and causes its precession in the reference frame of the binary @xcite . additionally , the accretion disk is truncated near the resonant radius . this explains the values of the disk radii smaller than the tidal truncation radius observed in cvs and lmxbs with small mass ratios ( fig.[fig : rdisk ] ) . to conclude , with exception of grs1915 + 105 , discussed in section [ sec : grs1915 ] , the excitation of the 3:1 inner lindblad resonance provides a natural explanation of the dichotomy in the viscous time scale between wide and compact systems . the values of the viscous time scale of the disk inferred by the position of the low frequency break in the power spectra of lmxbs require rather large disk thickness , @xmath194 , by a factor of @xmath195 exceeding the prediction of the standard theory of the accretion disks . taken at the face value , @xmath7 appears to agree with the statistics of the eclipsing systems among lmxbs ( e.g. * ? ? ? * ; * ? ? ? . however , large values of @xmath196 result in small optical depth of the outer disk , which is inconsistent with numerous observations of the optical emission from accretion disks , as discussed below . we consider the structure of an accretion disk with some value of the vertical scale - height @xmath197 , without specifying the source of the additional energy dissipation . under the assumption of vertical hydrostatic equilibrium and stationarity , both the mid - plane disk temperature and the density are steep functions of the disk thickness : @xmath198 , @xmath199 . due to strong dependence of the rosseland mean opacity on the density and temperature , typically as @xmath200 in the parameters range of interest , the optical depth of the disk @xmath201 is also steeply decreasing function of the disk thickness , @xmath202 with the power law index @xmath203 changing from @xmath204 to @xmath109 as the @xmath196 increases ( fig.[fig : tau ] ) . therefore , even a moderate increase of the disk thickness would lead to dramatic decrease of its optical depth . as a result , for @xmath197 , the accretion disk in the majority of lmxbs from our sample would be optically thin . this would contradict to a number of observational facts , mostly from the optical band , proving existence of the optically thick outer disk in lmxbs . indeed , the free - free optical depth in the optical v - band equals @xmath205 where @xmath206 is the disk geometrical thickness normalized to 0.1 . the above approximation is accurate to @xmath207 in the @xmath208 k temperature range , the logarithmic term due to the gaunt factor varies from 2.9 to 7.5 for the temperature in the range @xmath209 k. for the parameter of binaries from our sample the optical depth in the v - band would be @xmath210 , with the exception of sources with the largest @xmath2 ( sco x-1 , gx349 + 2 and 4u1820 ) where @xmath211 . the low optical depth in the v - band would conflict with numerous spectroscopic observations of the optical emission from the lmxbs , suggesting existence of the continuum emission originating from the optically thick outer accretion disk ( e.g. * ? ? ? * ; * ? ? ? on this basis the possibility of the semi - thick , @xmath197 outer disk in lmxbs can be excluded . cm on the @xmath196 ratio for @xmath212 g / s , @xmath213 @xmath161 , @xmath214 . the solid line shows the optical depth based on the opal opacities @xcite , long dashed line kramer law ( @xmath215 @xmath216/g ) , short dashed free - free absorption in the optical v - band , dash - dotted thompson optical depth . ] instead , one can envisage a two phase accretion flow with a shakura - sunyaev - like disk in the mid - plane surrounded by the tenuous optically thin coronal flow with the aspect ratio @xmath217 and temperature of @xmath218 . in order to explain the prominence of the low frequency break in the power spectrum , a significant fraction of the accretion has to take place in this corona @xmath219 the temperature and density in the coronal flow are : @xmath220}\end{aligned}\ ] ] @xmath221}\end{aligned}\ ] ] where @xmath222 . as the surface mass density depends on the thickness of the disk as @xmath223 , such a corona will contain a small fraction of the mass of the accreting matter : @xmath224 assuming @xmath225 . the vertical and radial column density of the coronal flow are @xmath226 $ ] : @xmath227 the plausible mechanism of formation of the coronal flow is the disk evaporation process originally proposed by @xcite for dwarf novae . in x - ray binaries the physics of the disk evaporation is significantly affected by the illumination and compton heating and cooling of the disk - corona system by the x - ray emission produced in the vicinity of the relativistic object . although the the self - consistent treatment of the problem is yet to be done , the likely net effect of the x - ray irradiation is to increase the fraction of @xmath2 in the coronal flow ( * ? ? ? * ; * ? ? ? * f.meyer , private communication ) . an independent evidence of the existence of a diffuse ionized gas above the accretion disk plane is provided by observations of lmxbs with high inclination the adc sources and dippers . based on observations of partial x - ray eclipses in 4u1822371 , @xcite concluded that the compact x - ray source in this system is diffused by a large moderately compton thick highly ionized corona located above and below the accretion disk . detailed modeling of the eclipse light curves in the adc sources 4u1822371 and 4u2129 + 47 @xcite yielded estimates of the radial extent of the accretion disk corona in these sources , @xmath228 cm and @xmath229 . these numbers are comparable with the value of the tidal radius in the systems , @xmath230 cm . the partial nature of the eclipses in these sources , with the eclipsed flux at the level of @xmath231 of the uneclipsed value indicated that the corona had non - negligible radial optical depth , @xmath232 , in agreement with the estimate of eq.([eq : corona - nh ] ) . high resolution spectroscopic observations of lmxbs with chandra and xmm - newton gratings are revealing complex absorption / emission features in their x - ray spectra @xcite . these features are mostly pronounced in the systems with high binary inclination angle and suggest the presence of tenuous photo - ionized plasma orbiting the compact object above the orbital plane of the binary . from the analysis of the recombination emission lines of h- and he - like ions of o , ne and mg in exo0748676 @xcite constrained the density and radial extend of the photoionized plasma , @xmath233 @xmath10 , @xmath234 cm and the temperature @xmath235 k. this range of the parameters is consistent with the numbers computed from the above formulae for exo0748676 and assuming the aspect ratio of the coronal flow @xmath236 : the tidal truncation radius @xmath237 cm , density in the coronal flow @xmath238 @xmath10 , and temperature of @xmath239 k. based on the observations of the narrow absorption lines of the h- and he - like fe in the spectrum of 4u1916053 , @xcite estimated the ionization parameter and the column density of the absorbing gas : @xmath240 , @xmath241 @xmath242 . from these numbers one can estimate the radial extent and the density of the absorbing gas , @xmath243 cm , @xmath244 @xmath10 . the tidal radius in this system is @xmath245 cm , in agreement with the value derived from the xmm - newton data . the density in the coronal flow with @xmath236 is @xmath246 @xmath10 . the latter number is somewhat larger than derived from the xmm - newton data . this can be explained by smaller inclination angle in this source resulting in lower density along the line of sight . indeed , unlike exo0748676 showing deep x - ray eclipses , 4u1916053 shows only x - ray dips . in this picture , the @xmath2 variations in the geometrically thin shakura sunyaev disk should give rise to the second very low frequency power law component , revealing itself below the break frequency . such power law component is indeed observed in the power spectra of some of the sources . most obvious its presence is in cir x1 and 4u1916053 and , to less extent , in sco x1 ( fig.[fig : pds_broadband ] ) . remarkably , the slope of the power spectrum before and after the break are very similar , suggesting similar nature of the processes causing the variability furthermore , the second break should be expected at the low frequencies , corresponding to the ( larger ) viscous time of the geometrically thin disk . the frequency of the second break is described by eq.([eq : fbr2forb ] ) with the @xmath196 given by eq.([eq : h2r_ss73 ] ) . an evidence of the second break may be seen in the power spectra of sco x-1 and , possibly , of 4u1916053 . in the former case the break frequency equals @xmath247 hz and the ratio @xmath248 . this value is within the range predicted for standard shakura sunyaev disk with the aspect ratio of @xmath249 , eq.([eq : fvisc2vorb_std ] ) . the value of the break frequency in the case of 4u1916053 is probably @xmath250 hz , although this number is less secure than that for sco x-1 . if correct , this is significantly , by @xmath251 dex smaller that expected in the above scenario . in the conclusion of this section we note that due to power law dependence of the viscous time scale on the disk radius , @xmath252 , the radial drift velocity of the matter in the accretion flow should be increased throughout the most ( @xmath253 ) of the radial extent of the accretion disk . therefore the bulge at the outer edge of the disk , resulting from the disk stream interaction ( e.g. * ? ? ? * ) is insufficient as it is located at the outer edge of the disk . for the same reason this analysis based on the viscous time of the disk probes the outer disk and is insensitive to the conditions in its innermost parts , e.g. @xmath254 , which contribution to the total viscous time is negligibly small . the dichotomy in the viscous time between long- and short - period systems is naturally explained by excitation of the 3:1 inner lindblad resonance . this suggestion is supported by detection of superhumps in several lmxbs with small mass ratios ( e.g. * ? ? ? * ; * ? ? ? importantly , the superhumps are detected not only in transient sources but also in one or two persistent lmxbs , including one source from our sample , 4u1916053 @xcite . the specific mechanism by which the tidal instability is affecting the viscous time of the disk is , however , unclear . it is also unclear if the coronal flow in the small - q system has indeed @xmath255 and @xmath256 as it is formally required by the observed @xmath257 . several possibilities can be mentioned : 1 . mass transfer in tidal waves excited in the disk by the tidal forces . it has been shown theoretically and confirmed in numerical simulations , that in the non - linear regime strong tidal waves are excited in the accretion disk @xcite . these waves , propagating in the disk with the speed significantly exceeding the radial drift velocity of the matter , can decrease the effective viscous time scale . if correct , the aspect ratio and coronal temperature in the small - q systems may be not significantly different from those in the large - q ones . 2 . heating of the outer disk due to resonant tidal interaction @xcite can increase the temperature of the disk and/or coronal flow and , correspondingly , decrease the viscous time scale . as was mentioned above , in order to explain the observed viscous time scale the coronal flow temperature of @xmath256 is required . non - trivial definition of the viscous time for an eccentric and precessing disk . in the fully developed instability the particles in the disk move on non - circular trajectories with large eccentricity . therefore , the average velocity of the radial drift of the disk particles can be significantly different from the value predicted by the standard disk theory for a circular keplerian disk . we note , that the truncation of the accretion disk at the 3:1 resonance radius ( fig.[fig : rdisk ] ) can be excluded as the cause of the shorter viscous time scale in the small-@xmath14 systems , because the change of the outer disk radius by a @xmath258 is insufficient to explain the observed decrease of the viscous time by the factor of @xmath12 . grs1915 + 105 clearly stands out among other small - q systems ( fig.[fig : fbr2forb ] , right panel ) . although it has small mass ratio , @xmath259 @xcite , its ratio @xmath6 is close to the value found in the binaries with @xmath260 . the most plausible explanation of such behavior is related to the transient nature of the source . we outline two possibilities : 1 . the tidal instability is known to have a rather long growth time , of the order of @xmath261 binary orbital periods ( e.g. * ? ? ? * ) . for grs1915 + 105 @xmath154 days and @xmath262 corresponds to @xmath12 years . therefore the time passed since the onset of the outburst ( @xmath12 years ) was insufficient for the tidal instability to fully develop . in this scenario one would expect an increase of the break frequency in the course of the outburst , unless the instability growth time is significantly longer than @xmath12 years . however , no statistically significant difference in the break frequency has been found between the first and second halves of the data . this fact speaks against the above suggestion . 2 . due to the large size of the system ( @xmath263 cm ) , the irradiation effects are insufficient to ionize the outer disk in grs1915 + 105 which is in the cold , low - viscosity state ( h.ritter , private communication ) . therefore its outer boundary is located near the circularization radius rather than has expanded to the tidal radius , as is the case for the persistent sources . therefore there is no accretion disk at the 3:1 resonance radius ( fig.[fig : rdisk ] ) and the tidal instability is not excited . in this respect the present outburst of grs1915 + 105 might be similar to normal outbursts in dwarf novae . as well known , for the same reason the superhumps in dwarf novae are observed ( tidal instability is excited ) only in during the superoutbursts but not during the normal outbursts ( e.g. * ? ? ? * ) . for several sources broad qpo - like features are obvious in the power density spectra . most apparent these features are in the case of cyg x-2 and grs1915 + 105 ( fig.[fig : pds_asm ] , [ fig : pds_exosat ] ) . among plausible mechanisms of appearance of quasi periodic variability in x - ray binaries on these time scales are the radiation - driven warping and precession of the accretion disk ( e.g. * ? ? ? * ) and mass - flow oscillations caused by diffusion instability @xcite . both types of oscillations are caused by the irradiation of the accretion disk by the primary and their characteristic time scale is of the order of the viscous time scale of the disk . accordingly , these features should appear near the break frequency in the power density spectrum . another possible mechanism of appearance of broad low frequency qpos near the viscous frequency of the disk was suggested by @xcite . this mechanism is related to the anti - correlation of the @xmath2 perturbations in the accretion flow on the time scales @xmath264 . the anti - correlation appears in the presence of the steady mass supply through the outer boundary of the accretion flow and is a direct consequence of the energy conservation law . in this respect the broad qpo features observed near the viscous time scale of the outer disk might be similar to the broad qpos detected near the high frequency break of power density spectrum of cyg x-1 in the hard state , at frequency of @xmath265 hz @xcite . in the latter case the values of the break frequency and of the qpo centroid frequency are defined by the viscous time scale of the inner hot flow , @xmath266 r@xmath267 , responsible for the production of the hard comptonized emission in the low spectral state of black hole candidates . we studied the low frequency variability of low mass x - ray binaries in order to search for the signatures in their power density spectra related to the viscous time scale of the disk . our results can be summarized as follows : 1 . as the viscous time is the longest time scale of the accretion disk , the @xmath2 variations produced in the disk due to various instabilities should become independent of each other on the time scales longer than the viscous time , i.e. @xmath268 at @xmath269 correspondingly , the power density of @xmath2 perturbations produced in the disk and of the observed x - ray flux variations should become independent of the frequency at @xmath270 . 2 . using archival data of rxte / asm and exosat / me we studied x - ray variability of persistent lmxbs in the @xmath271 hz frequency range . in power density spectra of 11 sources out of 12 satisfying our selection criteria , we found a very low frequency break . the spectra approximately follow @xmath4 power law above the break and are nearly flat below the break ( figs . [ fig : pds_asm][fig : pds_broadband ] ) . in some cases ( sco x-1 , cir x-1 and 4u1916053 ) a second power law component with the same slope appears at @xmath272 . 3 . in a very broad range of binary periods , from @xmath273 min in ultracompact binary 4u1820 - 303 to @xmath274 days in grs1915 + 105 , the break frequency correlates with the orbital frequency @xmath138 of the binary system ( fig.[fig : fbreak_forb ] ) , in good agreement with the theoretical prediction for the viscous frequency @xmath275 ( eq.[eq : fvisc2forb ] ) 4 . assuming that the low frequency break is associated with the viscous time scale of the disk , @xmath276 , the measured values of the break frequency imply that the viscous time of the disk is by a factor of @xmath277 shorter than predicted by the standard theory ( section [ sec : tvisc_h2r ] , fig.[fig : fbreak_forb ] ) . taken at the face value , this requires the relative height of the outer disk @xmath7 . motivated by the low vertical optical depth of such semi - thick accretion flow we propose instead , that significant fraction of the accretion occurs through the coronal flow above the standard geometrically thin shakura - sunyaev disk ( section [ sec : corona ] , fig.[fig : tau ] ) . the aspect ratio of the coronal flow , @xmath7 , corresponds to the gas temperature of @xmath278 . the corona has moderate optical depth in the radial direction , @xmath232 , and contains @xmath279 of the total mass of the accreting matter . the estimates of the temperature and density of the corona are in quantitative agreement with the parameters inferred by the x - ray spectroscopic observations by chandra and xmm - newton of complex absorption / emission features in the lmxbs with large inclination angle . we found a clear dichotomy in the @xmath6 between wide and compact system , the accretion flow in the compact systems having @xmath12 times shorter viscous time expressed in the orbital periods of the binary , than in the wide ones ( fig.[fig : fbr2forb ] ) . the jump in the @xmath11 ratio occurs at the binary system mass ratio of @xmath184 . this strongly suggests that the dichotomy between small - q and large - q systems is caused by the excitation of the 3:1 lindblad resonance in the small - q systems ( section [ sec : tvisc_lindblad ] ) . we are grateful to friedrich meyer and hans ritter for numerous discussions of the physics of the accretion disks and superhumps phenomenon in dwarf novae . va would like to acknowledge the partial support from the president of rf grant ns-2083.2003.2 and from the rbrf grant 03 - 02 - 17286 . this research has made use of data obtained through the high energy astrophysics science archive research center online service , provided by the nasa / goddard space flight center . bandyopadhyay r. , shabaz t. , charles p. , naylor t. , 1999 , mnras , 306 , 417 boirin l. et al . , 2004 , a&a , 418 , 1061 callanan p.j . , grindlay j.e . & cool a.m. , 1995 , pasj , 47 , 153 chakrabarty d. , 1998 , apj , 492 , 342 chaty s. , haswell c.a . , malzac j. , hynes r.i . , shrader c.r . , cui w. , 2003 , mnras , 346 , 689 chou y. , & grindlay j. , 2001 , apj , 563 , 934 churazov e , gilfanov m. , revnivtsev m , 2001 , mnras , 321 , 759 cottam j. , kahn s.m . , brinkman a.c . , den herder j.w . , erd c. , a&a , 2001 , 365 , 277 edelson r.a . & krolik j.h . , 1988 , apj , 333 , 646 fender r. et al . , 1999 , mnras , 304 , 865 grimm h .- j . , gilfanov m. , & sunyaev r. , 2002 , a&a , 391 , 923 harlaftis e.t . & marsh t.r . , 1996 , a&a , 308 , 97 harlaftis e. & greiner j. , 2004 , a&a , 414 , l13 harrop - allin m.k . & warner b. , 1996 , mnras , 279 , 219 haswell c.a . , king a.r . , murray j.r . & charles p.a . , 2001 , mnras , 321 , 475 hellier c. & mason k.o . , mnras , 239 , 715 hessman f. 1988 , astron . suppl . , 72 , 515 hessman f. , & hopp u. , 1990 , astron . astrophys . , 228 , 387 hynes r.i . , haswell c.a . , chaty s. , shrader c.r . , cui , w. , 2002 , mnras , 331 , 169 ichikawa s. , & osaki y. , 1994 , pasj , 46 , 621 iglesias c.a . & rogers f.j . , 1996 , apj , 464 , 943 jimenez - garate m. a. , raymond j.c . & liedahl d.a . , 2002 , apj , 581 , 1297 jimenez - garate m. a. , schulz n.s . , marshall h. l. , 2003 , apj , 590 , 432 johnston h. , fender r. , wu k. , 1999 , mnras , 308 , 415 kallman t.r . et al . , 2003 , 583 , 861 leahy d.a . , 1983 , apj , 266 , 160 levine a. , ma c. p. , mcclintock j. , rappaport s. , van der klis m. , verbunt f. , 1988 , apj , 327 , 732 liu q. , van paradijs j. , van den heuvel e. , 2001 , a&a , 368 , 1021 lubow s.h . & shu f.h . , 1975 , apj , 198 , 383 lubow s.h . , 1991 , apj , 381 , 259 lyubarskii yu.e . , 1997 , mnras , 292 , 679 lyutyi v. & sunyaev r. , 1976 , sov . astronomy journal , 53 , 511 maloney p.r . , begelman m.c . , pringle j.e . , 1996 , apj , 472 , 582 milgrom m. , 1978 , a&a , 67 , l25 meyer f. & meyer - hofmister e. 1990 , a&a , 239 , 214 meyer f. & meyer - hofmister e. 1994 , a&a , 288 , 175 meyer f. , liu b.f . & meyer - hofmister e. 2000 , a&a , 361 , 175 murdin p. et al . , 1980 , a&a , 87 , 292 narayan r. & mcclintock j.e . , 2004 , astro - ph/0410556 nelson l. et al . , 1986 , apj , 304 , 231 orosz j. , & kuulkers e. , 1999 , mnras , 305 , 132 osaki y. & meyer f. , 2003 , a&a , 401 , 325 paczynski b. , 1977 , apj , 216 , 822 papaloizou j. & pringle j.e . , mnras , 181 , 441 parmar a. n. , white n. e. , giommi p. , gottwald , m. , 1986 , apj , 308 , 199 patterson j. , 1984 , apjs , 54 , 443 rappaport s. , nelson l. , ma c. , joss p. , 1987 , apj , 322 , 842 rutten r.g.m . , van paradijs j. & tinbergen j. , 1992 , a&a , 260 , 213 shakura n.i . & sunyaev r.a . , 1973 , astron . astrophys , 24 , 337 shahbaz t. et al . , 2004 , mnras , in press ( astro - ph/0408057 ) spruit h.c . , matsuda t. , inoue m. , sawada , k. , 1987 , mnras , 229 , 517 steeghs d. & casares j. , 2002 , apj , 568 , 273 swank j.h . , 1999 , nucl.phys . b - proc.suppl . , 69 , 12 torres m.a.p . , 2004 , apj in press ( astro - ph/0405509 ) truss m.r . , , murray j.r . , king , a. r. , 2002 , mnras , 337 , 1329 turner m. , smith a. , zimmermann h. , 1981 , space science rev . , 30 , 513 vikhlinin a. et al . , 1994 , apj , 424 , 395 vikhlinin a. , churazov e , & gilfanov m. , 1994 , a&a , 287 , 73 wachter s. & margon b. , 1996 , astron.j . , 112 , 2684 white n.e . et al , 1981 , apj , 247 , 994 white n.e . & holt s.s . , 1982 , apj , 257 , 318 whitehurst r. , 1988 , mnras , 232 , 35 whitehurst r. & king a. , 1991 , mnras , 249 , 25 zurita c. et al . , 2000 , mnras , 316 , 137 zurita c. et al . , 2002 , mnras , 333 , 791 to calculate power density spectra from asm light curves we used the method based on the auto - correlation function ( cf . we begin with consideration of the equally spaced data . the measured time series is @xmath280 and equals the number of counts detected in the @xmath281-th time bin corresponding to the time interval of @xmath282 , where index @xmath281 runs from 0 to @xmath283 , n is assumed to be even . the discrete fourier transform of the time series is defined as : @xmath284 where @xmath285 changes in the range @xmath286 . the power density in the @xmath285-th frequency bin , which center is @xmath287 hz , expressed in units of rms@xmath288/hz ( rms is fractional rms of variability ) equals : @xmath289 where @xmath290 is the total number of counts in the time series , @xmath291 is its total duration and @xmath292 is the average count rate . combining eqs.[eq : ft ] and [ eq : ft_power ] one finds @xmath293 where @xmath294 is discrete auto - correlation function : @xmath295 the eq.[eq : acf_power ] expresses the well known fact that power equals the cosine transform of the auto - correlation function . the noise level in the power density spectrum equals : @xmath296 where @xmath297 is the error of @xmath280 . for poisson errors in @xmath280 , @xmath298 and @xmath299 , in agreement with the conventional form ( e.g. * ? ? ? the asm time series consists of the count rate measurements @xmath300 , [ cnts / sec ] , at unevenly spaced moments @xmath301 , @xmath302 . the values of @xmath300 are averaged over the time interval ( @xmath105 sec ) much shorter than the distance between adjacent bins ( @xmath105 min ) . as before , the auto - correlation function @xmath303 is defined on the grid of @xmath304 time bins , with the @xmath281-th bin corresponding to the time interval @xmath305 , where index @xmath281 runs from 0 to @xmath283 . the @xmath294 is computed as @xmath306 where the averaging is performed over all pairs @xmath307 , @xmath308 , for which @xmath309 falls in the @xmath310-th time bin of the grid . the time interval @xmath311 and the duration of the time series @xmath291 are parameters of the transformation and define the frequency range covered by the output power density spectrum . their choice depends on the effective nyquist frequency of the time series and its duration . appropriate for the parameters of the asm light - curves are the following values : @xmath312 sec and @xmath313 sec . advantage of this method is that the correctness of the chosen values of @xmath311 and @xmath291 can be easily checked using the number of pairs @xmath307 used to compute the autocorrelation value at the given time scale @xmath314 and the uniformity of these numbers across the full range of the time scales , from @xmath311 to @xmath291 . with the auto - correlation computed according to eq.[eq : acf_asm ] , the power density spectrum can be easily computed using eq.[eq : acf_power ] . as usual in the time series analysis , the entire time series is divided into @xmath318 segments and the power spectrum is computed for each segment separately . their average and dispersion give estimates of the average power density spectrum and its uncertainty ( under the assumption of stationarity ) . if the errors for @xmath300 are known , the noise level in the output power spectrum can be computed as @xmath319 where @xmath297 is the error of @xmath300 ( in units of cnts / sec ) and @xmath320 is the number of asm measurements used to compute one power spectrum . in practice , the flux measurement errors given in the asm light - curves are slightly underestimated , therefore the noise level computed with eq.[eq : asm_noise ] is somewhat imprecise @xcite . the performance of the method described above was verified in simulations . the initial light curve with the time resolution of @xmath321 sec was computed assuming the power law spectrum with slope of 1.3 and low frequency break at @xmath322 hz and @xmath323 hz and random fourier phases . from this light curve the asm time series was simulated using the measurements times @xmath301 of the real asm light curve of cyg x-2 . the obtained light curve was randomized assuming gaussian errors with the standard deviation equal the real asm measurement error for the given measurement @xmath300 . the obtained time series was analyzed with the same code as used for the analysis of the real asm data with parameters @xmath324 sec and @xmath325 sec . the noise level estimated using eq.[eq : asm_noise ] was subtracted from the power density spectra . the results of simulations are shown in fig.[fig : pds_fake ] along with the model power density spectra used to generate the initial high resolution light curves .

the ising spin glass is a one - parameter exponential family model for binary data with quadratic sufficient statistic . in this paper , we show that given a single realization from this model , the maximum pseudolikelihood estimate ( mple ) of the natural parameter is-consistent at a point whenever the log - partition function has order in a neighborhood of that point . this gives consistency rates of the mple for ferromagnetic ising models on general weighted graphs in all regimes , extending the results of chatterjee where only-consistency of the mple was shown . it is also shown that consistent testing , and hence estimation , is impossible in the high temperature phase in ferromagnetic ising models on a converging sequence of weighted graphs , which include the curie - weiss model . in this regime , the sufficient statistic is distributed as a weighted sum of independent random variables , and the asymptotic power of the most powerful test is determined . we also illustrate applications of our results on synthetic and real - world network data . exponential family , graph limit theory , hypothesis testing , ising model , pseudolikelihood estimation , spin glass .

1507.07055

the ising spin glass is a discrete random field developed in statistical physics as a model for ferromagnetism @xcite , and is now widely used in statistics as a model for binary data with applications in spatial modeling , image processing , and neural networks ( cf . @xcite and the references therein ) . to describe the model , suppose that the data is a vector of dependent @xmath4 random variables @xmath5 , and the dependence among the coordinates of @xmath6 is modeled by a one - parameter exponential family where the sufficient statistic is a quadratic form : @xmath7 for any @xmath8 and an @xmath9 symmetric matrix @xmath10 with zeros on the diagonals . the elements of @xmath10 are denoted by @xmath11 , for @xmath12 . given any @xmath13 , the quadratic form ( [ q ] ) defines a parametric family of probability distributions on @xmath14 : @xmath15 where @xmath16 is the _ log - partition function _ which is determined by the condition @xmath17 , that is , @xmath18 where @xmath19 denotes the expectation over @xmath6 distributed as @xmath20 , the uniform measure on @xmath14 . the parameter @xmath21 is often referred to as the _ inverse temperature _ , so the high temperature regime corresponds to small values of @xmath22 . the family ( [ pmf ] ) includes many famous statistical physics models : the usual ferromagnetic ising model on generals graphs , the sherrington - kirkpatrick mean - field model @xcite , and the hopfield model for neural networks @xcite . estimating the parameter @xmath22 in , given one realization from the model , is extremely difficult using likelihood - based methods because of the presence of an intractable normalizing constant @xmath16 in the likelihood . a variety of numerical methods are known for approximately computing the likelihood @xcite , but they are computationally expensive and very little is known about the rate of convergence . one alternative to using likelihood - based methods is to consider the maximum pseudolikelihood estimator ( mple ) @xcite . chatterjee @xcite showed that given a spin single realization from the model , the mple @xmath23 is @xmath2-consistent at @xmath24 is said to be @xmath1-_consistent _ at @xmath24 if @xmath25 , that is , @xmath26 . ] , whenever @xmath27 . however , in many popular models such as regular graphs , random graphs , and weighted dense graphs , the log - partition function @xmath28 for certain ranges of @xmath22 , and chatterjee s result does not tell us anything about the consistency of the mple . in this paper we show that the mple is @xmath0-consistent at @xmath24 , if the log - partition function has order @xmath1 in a neighborhood of @xmath29 ( theorem [ gen ] ) , for a sequence @xmath30 . this gives the consistency rate of the mple for all values of @xmath31 away from the critical points , and shows that the rate of the mple undergoes phase transitions for ising models on various graphs ensembles ( corollary [ dense ] and [ erdosrenyirandom ] ) . we also show that _ no _ consistent test , and hence no estimator , exists if the log - partition function remains bounded ( theorem [ genfixed ] ) . as a consequence , consistent estimation is impossible in the high temperature regime in ferromagnetic ising models on a converging sequence ( in cut - metric as defined by lovaz and co - authors @xcite ) of weighted graphs ( theorem [ dense_testing ] ) . this strengthens previous results of comets and gidas @xcite and chatterjee @xcite where the mle and the mple was , respectively , shown to be inconsistent for @xmath32 in the curie - weiss model , which corresponds to taking @xmath33 , for all @xmath34 . finally , using the emerging theory of graph limits @xcite , the limiting distribution of the sufficient statistic @xmath35 , and the asymptotic power of the most powerful test are derived for weighted dense graphs in the high temperature regime ( theorem [ dense ] ) . + [ fig : fb1 ] while proving the consistency of the mple , we show that the asymptotic order of the sufficient statistic @xmath35 is same as the order of the log - partition function for general matrices @xmath10 ; a result which appears to be new and might be of independent interest . more precisely , the sequence of random variables @xmath36 is asymptotically tight under @xmath37 , and the limiting distribution ( if any ) is non - zero when the log - partition function has order @xmath1 in a neighborhood of @xmath29 ( lemma [ lem1 ] ) . moreover , simple bounds for matrices @xmath10 with non - negative entries provide the correct order of the log - partition function in the high temperature regime for a wide class of ising models ( lemma [ bound ] ) . finally , we illustrate the usefulness of the mple and the applicability of our results on two real datasets : in section [ sec : fbdataset ] we study the effect of gender among friends in two facebook friendship - networks from the stanford large network dataset ( snap ) collection ( see figure [ fig : fb1 ] ) . in section [ sec : statesdataset ] we model the spatial correlation among the different states in the us neighborhood graph using data from the last 26 presidential elections during 1912 - 2012 ( see figure [ fig : states2012 ] ) . + another active area of research is high - dimensional structure estimation in a sparse ising model , where the goal is to consistently estimate the underlying matrix @xmath10 , under certain structural constraints from i.i.d . samples from the model ( refer to @xcite and the references therein ) . this is in contrast with the present work , where the matrix @xmath10 is known and we estimate the natural parameter and its error rate given a _ single _ realization from the model . the rest of the paper is organized as follows : the consistency of the mple and general inconsistency results are described in section [ pl_estimate ] . applications of these results to various graph ensembles including regular graphs , random graphs , and general weighted graphs , are explained in section [ examples ] . theorem [ gen ] and theorem [ genfixed ] are proved in section [ pfmain ] and section [ pfgenfixed ] , respectively . the proofs of corollary [ dregular ] and [ erdosrenyirandom ] are given in section [ pfapplication ] . the results on converging sequence of weighted graphs are in section [ pfdense ] . the analysis of the real datasets are given in section [ sec : datasets ] . the maximum pseudolikelihood estimator ( mple ) , introduced by besag @xcite , can be conveniently used to approximate the joint distribution of @xmath38 that avoids calculations with the normalizing constant . given a random vector @xmath39 whose joint distribution is parametrized by a parameter @xmath40 , the mple of @xmath22 defined as @xmath41 where @xmath42 is the conditional probability density of @xmath43 given @xmath44 . given @xmath45 from the model ( [ pmf ] ) , the conditional density of @xmath46 , given @xmath47 can be easily computed . to this end , given @xmath48 , define the function @xmath49 as @xmath50 where @xmath51 note that @xmath52 does not depend on @xmath53 since the diagonal element @xmath54 . interpreting @xmath55 , the function @xmath56 can be extended to @xmath57 $ ] by defining @xmath58 . then it is easy to verify that ( see chatterjee ( * ? ? ? * section 1.2 ) ) @xmath59 . therefore , the mple for @xmath22 in the model ( [ pmf ] ) is @xmath60 where @xmath45 is a random element from ( [ pmf ] ) . hereafter , we suppress the dependence on @xmath6 and denote by @xmath61 the mple of @xmath22 . consistency results for the mple in ising models are known in the case of lattices @xcite , complete graphs @xcite , and spatial point processes @xcite . however , for general processes where the dependence is neither local nor mean - field , it is very difficult to prove consistency results for mple . in a major breakthrough , chatterjee @xcite developed a remarkable technique using exchangeable pairs and showed ( * ? ? ? * theorem 1.1 ) that the mple @xmath62 , given a single realization @xmath63 from ( [ pmf ] ) , is a @xmath2-consistent estimator at @xmath64 , whenever @xmath65 symmetric matrix @xmath66 , denote by @xmath67 the operator norm of @xmath66 . ] and @xmath68 + to the best of our knowledge , all results regarding mple @xmath62 are in the regime where it is @xmath2-consistent . however , in many examples such as the ising model on dense graphs , @xmath69-regular graphs with @xmath70 , and erds - rnyi graphs @xmath71 , with @xmath72 , the partition function @xmath28 for certain ranges for @xmath22 , and chatterjee s result is not applicable for deriving the consistency of the mple . the following theorem ( see section [ pfgen ] for proof ) shows that the consistency of the mple at a point is governed by the order of the log - partition function in a neighborhood of that point . this generalizes the result of chatterjee @xcite giving the rate of consistency of the mple for all values @xmath22 ( at all temperatures ) away from the critical points . [ gen ] let @xmath73 , and @xmath74 be fixed . suppose @xmath75 is a sequence of positive reals diverging to @xmath76 such that @xmath77 moreover , assume that the following conditions hold : * @xmath78 , where @xmath79 is as defined in ( [ mi ] ) . * @xmath80 . then the mple @xmath62 for the model ( [ pmf ] ) is a @xmath0-consistent sequence of estimators for @xmath24 . condition ( [ maincond ] ) in the above theorem demands the right order of the log - partition function in a small neighborhood around the point @xmath29 . this avoids the critical points , where the order of the log - partition function ( and its derivative ) undergoes a sharp transition . it follows from the proof of theorem [ gen ] that the following ( possibly slightly weaker ) condition works as well instead of : @xmath81 however , for most of the applications estimates of the log - partition are more readily available . thus , the sufficient conditions are stated in terms of the log - partition function instead of its derivative . conditions ( a ) and ( b ) are technical requirements arising out of the proof technique , which ensure that the main contributions come from @xmath79 that are small , and on average the entries in @xmath10 are not too large compared to @xmath1 . these conditions are satisfied in most commonly used models ( see section [ examples ] ) . moreover , the result of chatterjee ( * ? ? ? * theorem 1.1 ) is an immediate corollary of theorem [ gen ] ( refer to section [ pfcorchatterjee ] for the proof ) . ( ( * ? ? ? * theorem 1.1 ) ) [ sc ] let @xmath73 and @xmath74 be such that ( [ ptn ] ) holds . then the sequence of estimators @xmath62 is @xmath82 consistent for @xmath24 . note that theorem [ gen ] does not apply to the case @xmath83 . next , we show that if @xmath83 , then there is no sequence of estimators which consistently estimates @xmath29 . in fact , we show that even testing is impossible in this regime : given a single spin - configuration @xmath63 from ( [ pmf ] ) , there exists _ no _ sequence of consistent tests is said to be _ consistent _ for the testing problem ( [ testingbeta12 ] ) if @xmath84 and @xmath85 . ] for the hypothesis testing problem : @xmath86 this is summarized in the following theorem ( see section [ pfgenfixed ] for proof ) : [ genfixed ] let @xmath73 , and @xmath74 be fixed . suppose @xmath87 then for @xmath88 , there exists no consistent sequence of tests for the testing problem ( [ testingbeta12 ] ) . in particular , there exists no consistent sequence of estimators for @xmath22 in the interval @xmath89 $ ] . one of the main applications of above results is in deriving the rate of the mple for ising models on weighted graphs , that is , for matrices @xmath10 with non - negative entries . for such matrices , condition ( b ) in theorem [ gen ] can be directly verified , and we have the following simplified corollary : [ mplepositive ] consider the model ( [ pmf ] ) such that @xmath10 is a sequence of matrices with non - negative entries with @xmath90 . a. the sequence of estimators @xmath62 is @xmath91 consistent at @xmath24 for any @xmath92 , whenever condition ( a ) in theorem [ gen ] holds . b. if @xmath93 , then exists no consistent sequence of estimators for @xmath22 in the interval @xmath94 . the @xmath2-consistency of the mple in the sherrington - kirkpatrick ( sk ) model and the hopfield model , for all values of @xmath95 , follows from results of chatterjee @xcite . our results give the rate of consistency of the mple in the regime where it is not @xmath2-consistent . we begin with a simple example where the rate of the mple undergoes multiple phase transitions . [ ptblock ] consider the model with @xmath96 then the sequence of estimators @xmath62 is inconsistent for @xmath97 , @xmath98-consistent for @xmath99 , and @xmath82-consistent if @xmath100 . the proof of the above example is given in section [ ptblockpf ] . in fact , this example can be easily generalized to construct a @xmath101-block matrix @xmath10 such that the consistency rate of mple undergoes @xmath101 phase transitions . however , for most popular choices of @xmath10 the rate of the mple undergoes at most one phase transition from @xmath102-consistent to @xmath2-consistent . let @xmath103 be a sequence of @xmath104 regular graphs . consider the family of probability distributions ( [ pmf ] ) with the sufficient statistic @xmath105 where @xmath106 is the adjacency matrix of the graph @xmath103 . this includes ising models on lattices , complete graph , hypercube , and random regular graphs , among others , and have been extensively studied in probability and statistical physics . dembo et al . @xcite derived the limit of the log - partition function for random regular ( and other locally - tree like ) graphs . levin et al . @xcite showed that the mixing time of the glauber dynamics on the complete graph exhibits the cutoff phenomenon @xcite in the high temperature regime . the cutoff phenomenon for lattices was established by lubetzky and sly in a series of breakthrough papers ( refer to @xcite and the references therein ) . the next result gives the rate of consistency of the mple for general regular graphs . the proofs are deferred to section [ pfapplication ] . [ dregular ] fix @xmath74 and let @xmath103 be a sequence of @xmath104 regular graphs . suppose @xmath62 is the mple for the model ( [ pmf ] ) with sufficient statistic ( [ adjm ] ) . a. if @xmath107 , @xmath62 is a @xmath108-consistent sequence of estimators for @xmath29 . b. if @xmath109 , @xmath62 is a @xmath82-consistent sequence of estimators for @xmath29 . the above theorem shows that the rate of the mple undergoes a phase transition at @xmath110 for general regular graphs . in particular if @xmath111 remains bounded , then the above theorem shows that the mple is @xmath2 for all non - negative @xmath112 . however , in this case , it is easy to argue that @xmath113 , for all @xmath95 ( see proof of lower bound in corollary [ dregular ] ) . theorem [ gen ] then concludes that @xmath23 is @xmath82-consistent for all values of @xmath95 . in fact , using similar arguments as in the proof of corollary [ dregular ] , it follows that the mple @xmath62 is @xmath2-consistent for all @xmath31 in all bounded degree graphs with at least @xmath114 edges . this shows that mple is @xmath2-consistent for lattice graphs re - deriving classical results ( see @xcite and the references therein ) . for @xmath115 the behavior of the mple at @xmath110 remains unclear . it is believed that the mple might have a non - gaussian limiting distribution at the critical point @xmath110 @xcite . if @xmath116 , then theorem [ dregular ] shows that the mple is @xmath117 consistent for @xmath107 , suggesting that the mple might be inconsistent in this regime . chatterjee @xcite showed that this is indeed the case for the curie - weiss model ( where @xmath33 for all @xmath118 ) for @xmath119 . comets and gidas @xcite showed that even the mle of @xmath22 in the curie - weiss model is inconsistent for @xmath119 . later , in theorem [ dense ] we strengthen this result by showing that for ising models on arbitrary dense graphs , there exists no sequence of consistent estimators before the phase transition point . this extends the results in @xcite and justifies the @xmath117-rate of the mple in the dense case . + [ mplehist ] let @xmath120 be a sequence of erds - rnyi graphs . consider the family of probability distributions ( [ pmf ] ) with the sufficient statistic @xmath121 where @xmath106 is the adjacency matrix of the graph @xmath103 . [ erdosrenyirandom ] fix @xmath74 and consider a sequence @xmath120 of erds - rnyi graphs , with @xmath122 . let @xmath62 be the mple for the model ( [ pmf ] ) with sufficient statistic ( [ er ] ) . a. if @xmath107 , @xmath62 is a @xmath123-consistent sequence of estimators for @xmath29 . b. if @xmath109 , @xmath62 is a @xmath82-consistent sequence of estimators for @xmath29 . as in the regular case , the rate of the mple undergoes a phase transition at @xmath110 for erds - rnyi graphs . figure [ mpleerrorrate ] shows the error bars for the mple for the ising model on @xmath120 , with @xmath124 and @xmath125 , for a sequence of values of @xmath126 $ ] . figure [ mplehist ] shows the histograms of the mple for two values of @xmath22 ( before and after the phase transition point @xmath110 ) under various scalings . recall that the mple is inconsistent in the curie - weiss model in the high temperature regime , @xmath127 @xcite . in this section , using the emerging theory of graph limits and theorem [ genfixed ] above , we strengthen this result to show that consistent testing is impossible in the entire high temperature regime in ising models on a converging sequence of weighted dense graphs . we also calculate the distribution of the most powerful test and the asymptotic power in this regime . let @xmath128 be a weighted graph with vertices @xmath129 and @xmath130 a symmetric matrix of edge - weights in @xmath131 $ ] with zeros on the diagonal , that is , @xmath132,~ w(a , b)=w(b , a ) , \text { for } 1\leq a < b\leq n. \label{edgeweights}\ ] ] conversely , any symmetric @xmath9 matrix with entries in @xmath131 $ ] and zeros on the diagonal is the edge weight matrix of some weighted graph . a graph is said to be _ simple _ if @xmath133 and @xmath134 , for @xmath135 . in this case , @xmath136 denotes the set of edges in the graph @xmath103 . lovsz and co - authors @xcite developed a limit theory of weighted graphs , which connects various topics such as graph homomorphisms , szemerdi regularity lemma , and extremal graph theory . in the following , we summarize the basic results about converging graph sequences ( cf . lovsz @xcite for a detailed exposition ) . to this end , note that any weighted graph @xmath103 with edge - weights as in ( [ edgeweights ] ) can be represented as a function @xmath137 ^ 2\rightarrow [ 0 , 1]$ ] in a natural way : define @xmath138 , that is , partition @xmath131 ^ 2 $ ] into @xmath139 squares of side length @xmath140 , and define @xmath141 in the @xmath142-th square . let @xmath143 be the space of all measurable functions from @xmath131 ^ 2 $ ] into @xmath131 $ ] that satisfy @xmath144 for all @xmath145 $ ] . for every @xmath146 and any fixed simple graph @xmath147 define the _ homomorphism density _ @xmath148^{|v(h)|}}\prod_{(i , j)\in e(h)}w(x_i , x_j ) \mathrm d x_1\mathrm d x_2\cdots \mathrm d x_{|v(h)|}.\ ] ] a sequence of weighted graphs @xmath149 is said to converge to @xmath146 if for every finite simple graph @xmath150 , @xmath151 the limit objects , that is , the elements of @xmath143 , are called _ graph limits _ or _ graphons_. conversely , every such function arises as the limit of an appropriate graph sequence . it turns out that the above notion of convergence can be suitably metrized using the so - called _ cut - metric _ * chapter 8) for details ) . moreover , every function @xmath146 defines an operator @xmath152\rightarrow l_2[0 , 1]$ ] : @xmath153 @xmath154 is a hilbert - schmidt operator with operator norm denoted by @xmath155 , which is compact and has a discrete spectrum , that is , a countable multiset of non - zero real eigenvalues @xmath156 . in particular , every non - zero eigenvalue has finite multiplicity and @xmath157 ^ 2}w(x , y)^2\mathrm d x\mathrm d y:=||w||_2 ^ 2 . \label{lambda2}\end{aligned}\ ] ] recall that for a weighted graph @xmath103 , @xmath158 is the weighted adjacency matrix ( [ edgeweights ] ) . now , using graph limit theory we show the following result : [ dense_testing ] let @xmath149 be a sequence of weighted graphs as in ( [ edgeweights ] ) which converges in cut - metric to @xmath159 such that @xmath160 ^ 2}w(x , y)\mathrm d x\mathrm d y>0 $ ] . consider the testing problem ( [ testingbeta12 ] ) given a single realization @xmath63 from ( [ pmf ] ) with sufficient statistic @xmath161 . a. if @xmath162 , then there does not exist a sequence of consistent tests for ( [ testingbeta12 ] ) . b. if @xmath163 , then the mple @xmath164 is a sequence of @xmath82-consistent estimators for @xmath24 . the proof of the theorem is given in section [ pfdense ] . it involves showing that @xmath83 whenever @xmath165 , for any converging sequence of weighted graphs , which together with theorem [ genfixed ] proves ( a ) . to show ( b ) it suffices to show that @xmath166 , for @xmath163 ( by corollary [ sc ] ) . for ising models on a convergence sequence of graphs , @xmath167 is given by a variational problem ( [ denseopt ] ) ( cf . * theorem 2.14 ) ) . even though explicitly solving this variational problem for large values of @xmath22 is extremely difficult , a simple argument can be used to show that the value of the variational problem is positive for @xmath168 . + [ randomgraphpower ] by the neyman - pearson lemma , the most - powerful ( mp ) test for ( [ testingbeta12 ] ) is based on the sufficient statistic @xmath35 . by theorem [ dense_testing ] , the test based on @xmath35 is not consistent ( see figure [ randomgraphpower ] ) . however , the asymptotic power of the mp - test can be derived from the limiting distribution of @xmath35 , for any @xmath169 . [ dense ] let @xmath149 be a sequence of weighted graphs with entries bounded in @xmath131 $ ] which converges in cut - metric to @xmath159 , with @xmath160 ^ 2}w(x , y)\mathrm d x\mathrm d y>0 $ ] . if @xmath170 , then for @xmath171 @xmath172 where @xmath173 are i.i.d . @xmath174 random variables . hereafter , the random variable in the rhs of ( [ limitdense ] ) will be denoted by @xmath175 and can be used to compute the asymptotic power for the test based on @xmath35 for the testing problem ( [ testingbeta12 ] ) , when @xmath162 . to this end , we need the following definition : [ not ] let @xmath146 and @xmath176 . denote by @xmath177 the distribution function of the random variable @xmath175 defined in ( [ limitdense ] ) . also , let @xmath178 be the @xmath179-th quantile of @xmath177 , that is , @xmath180 . the following corollary is an immediate consequence of the neyman - pearson lemma and theorem [ dense ] . fix @xmath181 and @xmath162 . the most powerful level @xmath182 test for ( [ testingbeta12 ] ) rejects @xmath183 when @xmath184 , and has limiting power @xmath185 + in most of the relevant examples , the limiting graphon @xmath186 has finitely many non - zero eigenvalues , and the expression on the rhs of ( [ pb1 ] ) can be computed easily in terms of the quantiles of the chi - squared distribution . suppose @xmath187 be a erds - rnyi random graph with @xmath188 . then @xmath103 converges to the constant function @xmath189 on @xmath131 ^ 2 $ ] , which has only one non - zero eigenvalue @xmath190 . therefore , consistent testing is impossible for @xmath191 ( see figure [ randomgraphpower ] ) . moreover , for @xmath192 , ( [ pb1 ] ) simplifies to @xmath193 if @xmath194 denotes the @xmath179-th quantile of the @xmath3 distribution , then by ( [ pb1 ] ) , the limiting power of the test with rejection region @xmath195 for the testing problem @xmath196 versus @xmath197 is @xmath198 the limiting power of the mp - test for the curie - weiss model ( which corresponds to taking @xmath199 in ) is shown in figure [ limitingpower ] . note that it has a phase transition at @xmath110 , as stated in theorem [ dense_testing ] . ising models have been widely used to understand correlations among neighboring vertices in network data with binary node attributes . here , we use it to study the effect of gender in facebook friendship - networks , and spatial correlation of voting patterns among neighboring states based on the us presidential elections data during 1912 - 2012 . we use the facebook friendship - network data from the stanford large network dataset ( snap ) collection , available freely at http://snap.stanford.edu/data/egonets-facebook.html . the nodes are groups of users from facebook and there is an edge between two users if they are friends . the dataset also include several anonymized node features , such as hometown , gender , birthday , school , and university . we consider two networks ( referred to as fb1 and fb2 ) with gender as the binary node feature , encoding , without loss of generality , male by 1 and female by -1 . the nodes labelled 1 are colored blue and those labelled -1 are colored red . the fb1 network has 221 nodes and 3176 edges . among the 221 nodes , 170 are labelled 1 and 51 are labelled -1 . the fb2 network ( depicted in figure [ fig : fb1 ] ) has 333 nodes and 2519 edges , with 213 nodes labelled 1 and 120 labelled -1 . [ cols="^,^,^",options="header " , ] the dataset consists of political colors of the 48 states in the continental us ( excluding alaska , hawai , and washington d.c . ) in the last 26 presidential elections held during 1912 - 2012 . each state is assigned 1 for democratic ( colored blue ) and -1 for republican ( colored red ) . in a few cases , where a state was won by some third party , the color ( blue / red ) is assigned depending on which of the other two parties received more votes in that particular state . in order to understand the spatial correlation of political colors across the states , we consider an ising model on the neighborhood graph , where the vertices represent the states , and an edge between two nodes indicates that the two corresponding states share a border . the mple of @xmath22 for the 26 presidential elections are given in table [ table : elections ] . the mple - test column shows the @xmath200-values ( based on @xmath201 replications of the null distribution ) corresponding to these estimates for testing the null hypothesis of no spatial correlation . recall that the most - powerful test for this problem rejects for large values of @xmath202 , where @xmath203 is the scaled adjacency matrix of the network . + + + + the mp - test column shows the asymptotic @xmath200-values of the most powerful test calculated using the normal approximation @xmath204 . , whenever @xmath203 is the scaled adjacency matrix of a bounded degree graph @xmath103 . it is reasonable to use the normal approximation for the us neighborhood graph , as it has maximum degree 8 ( corresponding to the states missouri and tennessee ) . ] note that the outcomes of both tests in all 26 cases are the same at the 1% level , that is , they both reject or accept , demonstrating the use of asymptotic @xmath200-values and the significance of our theoretical results . there are several interesting values in table [ fig : states ] , which can often be explained by actual political events : a. the mple of @xmath22 for the 1924 election is the largest , which indicates very high spatial correlation , as illustrated from the perfect red - blue clustering in the graph ( figure [ fig : states](a ) ) . b. in contrast , large values of the mple for the 1936 , 1972 , and 1984 elections are not because of spatial correlation , but due to the landslide victories of fdr ( figure [ fig : states](b ) ) , nixon , and reagan , respectively . in this case , inferring existence of spatial correlation , as indicated by the large values of mple , would be incorrect . c. except for the 1960 and 1992 presidential elections , all other mple - tests are rejected at the @xmath205 level , indicating significant spatial correlation across the states ( except those in 1936 , 1972 , and 1984 for reasons discussed above ) . the strength of the spatial correlation is quantified by the values of the estimates . d. for the 1960 and 1992 presidential elections , the null hypothesis of independence is accepted at 5% . this can be explained by the apparent lack of clusters / patterns in the corresponding graphs ( figure [ fig : states](c ) and figure [ fig : states](d ) ) . interestingly , the 1960 election race was the closest in this period ( 1912 - 2012 ) , where jfk carried less states than the republican candidate , but won in the total percentage of votes received . the political colors of the states are governed by several attributes which are interacting with each other , and it is unlikely that such a complex dynamics can be modeled by a one - parameter ising model as above . nevertheless , the study above shows that even using such a simple model captures interesting aspects of the data , and can be used as a platform for further analysis . this section contains the proof of theorem [ gen ] . the technical lemmas required for the proof are listed in section [ pftechnicallemmas ] and proved later in appendix [ pftechnicallemmas ] . using this , we complete the proof of the theorem in section [ pfgen ] . corollary [ sc ] is proved in section [ pfcorchatterjee ] . the proof of theorem [ gen ] requires few technical lemmas . we begin by showing that in ising models satisfying ( [ maincond ] ) , the asymptotic order of the sufficient statistic @xmath35 is same as the order of the log - partition function , that is , ( a ) the sequence @xmath36 does not tend to @xmath206 in distribution , and ( b ) @xmath36 is @xmath207 . in fact , ( b ) is not required in the rest of the proof , however we include it because , together with ( a ) , it gives the correct order of @xmath35 , which appears to be new and might be of independent interest . the proof of the lemma is given in appendix [ pflem1 ] . [ lem1 ] under assumption , the following hold : a. @xmath208 , b. @xmath209 , the next lemma is similar to the ( * ? ? ? * lemma 1.2 ) , where it was shown that the second moment of the function @xmath210 is @xmath211 whenever the log - partition function scales like @xmath212 . here , by a finer analysis using part ( a ) of lemma [ lem1 ] we show that the @xmath213 , if the log - partition function has order @xmath1 . the proof of the lemma is given in appenidx [ pflem2moment ] . [ 2moment ] let @xmath214 be as defined in . then under the assumptions in theorem [ gen ] , for @xmath212 large enough , @xmath215 where @xmath216 is a constant depending only on @xmath29 . the above results can be combined to get a lower bound on the sum of squares of @xmath79 that are small " . the proof of the lemma is given in appendix [ pflemderivative ] . [ derivative ] under the assumptions in theorem [ gen ] , @xmath217 the above lemma replaces the application of paley - zygmund inequality of ( * ? ? ? * lemma 2.2 ) , and will be used to complete the proof of theorem [ gen ] . by chebyshev s inequality and lemma [ 2moment ] we have @xmath218 now , fix @xmath219 . therefore , it is possible to choose @xmath220 such that the rhs above is less than @xmath221 . also , by lemma [ derivative ] there exists @xmath222 and @xmath223 such that @xmath224 for @xmath212 large enough . thus , taking @xmath212 large enough and setting @xmath225 we have @xmath226 . for @xmath227 , @xmath228 therefore , @xmath229 let @xmath230 . this implies that @xmath231 and theorem [ gen ] follows . note that for @xmath48 and any @xmath232 , @xmath233 therefore , condition ( a ) in theorem [ gen ] holds with @xmath234 . moreover , @xmath235 that is condition ( b ) in theorem [ gen ] holds with @xmath234 . finally , to check note that @xmath236 , where @xmath237 therefore , @xmath238 by condition ( [ ptn ] ) . also , @xmath239 . this verifies ( [ maincond ] ) and by an application of theorem [ gen ] the result follows . in this section we give the proof of theorem [ genfixed ] , which shows that consistent testing and estimation is impossible whenever the partition function is @xmath117 . this is a consequence of a general result ( see proposition [ testing ] below ) which shows that distinguishing two probabiity measures @xmath240 versus @xmath241 is impossible whenever the kl divergence between the two measures @xmath240 and @xmath241 remains asymptotically bounded . for every @xmath242 , let @xmath243 be a measure space and @xmath240 and @xmath241 two distributions on this measure space . let @xmath244 be a dominating measure for both @xmath240 and @xmath241 , and @xmath245 and @xmath246 denote the respective densities with respect to this measure . also , denote the kullback - leibler ( kl ) divergence between @xmath241 and @xmath240 by @xmath247 consider the problem of testing @xmath240 versus @xmath241 . a sequence of tests @xmath248 is _ consistent _ for this testing problem if there exists a sequence of test functions @xmath249 such that @xmath250 , and @xmath251 . [ testing ] consider the problem of testing @xmath240 versus @xmath241 . if @xmath252 then there does not exist a consistent sequence of tests for this testing problem . the proof of the proposition is given in appendix [ pftesting ] . in the following , we use it to prove theorem [ genfixed ] . given proposition [ testing ] , it remains to verify that @xmath253 for @xmath254 ( where @xmath29 satisfies ) . by hypothesis there exists @xmath255 such that @xmath256 and @xmath257 , for @xmath212 large enough . moreover , by the monotonicity of @xmath258 , @xmath259 proving . in this section , we prove corollary [ mplepositive ] which will then be used to derive rates of consistency of the mple for ising models on different graph ensembles , using theorems [ gen ] and [ genfixed ] . to apply these results we need to determine the correct order of @xmath260 in a neighborhood of a point @xmath74 . however , the exact asymptotics @xmath260 is known only for specific choices of the matrix @xmath10 and for specific values of @xmath29 . nevertheless , the correct order of @xmath260 can be easily obtained in various examples , using , for instance , the following very useful lemma , which is of independent interest and may find other applications . [ bound ] consider the family of probability distributions on @xmath14 given by ( [ pmf ] ) . assume that the elements of the matrix @xmath10 are non - negative , and @xmath261 are the eigenvalues of the matrix @xmath10 . a. for @xmath262 , @xmath263 b. for any @xmath95 , @xmath264 let @xmath265 be a vector of i.i.d . @xmath266 random variables . note that for any @xmath267 and non - negative integers @xmath268 @xmath269 since the matrix @xmath10 has non - negative entries , by expanding the exponential function in power series every term can be bounded using the above inequality . this implies that @xmath270 the rhs of can be computed exactly as follows : let @xmath271 , be the spectral decomposition of @xmath10 , where @xmath272 are the normalized eigenvectors of @xmath10 . then setting @xmath273 for @xmath274 , we get @xmath275 note that @xmath276 is a vector of i.i.d . @xmath266 random variables . therefore , and ( [ expnorm ] ) implies @xmath277 using the mgf of the chi - squared distribution ( since @xmath278 , for all @xmath274 ) . the inequality follows by taking @xmath279 on both sides . to prove ( b ) let @xmath280 be i.i.d . with @xmath281 . then for any collection of non - negative integers @xmath282 , @xmath283 indeed , this follows on noting that both the lhs and rhs are @xmath284-valued , and the lhs is @xmath285 if and only if @xmath286 is even for all @xmath287 , which is when the rhs is @xmath285 as well . this implies , @xmath288 the inequality follows on taking @xmath279 on both sides . note that the upper bound is obtained by replacing the spin configuration @xmath5 with a vector of i.i.d . @xmath289 random variables . to get the lower bound , the collection @xmath290 is replaced by i.i.d rademacher random variables . surprisingly , the bounds obtained by these simple comparison techniques often give the correct asymptotic order of @xmath16 in the high temperature regime @xmath291 . to get the order of @xmath16 beyond the phase transition , the standard mean - field approximation can be used ( see section [ bound ] for details ) . for all @xmath31 , by the bound ( [ lower_bound ] ) in lemma [ bound ] we get @xmath292 where @xmath293 . to get the upper bound , we use ( [ upper_bound ] ) for @xmath294 @xmath295 where @xmath296 for any @xmath294 , and we use the fact that @xmath297 . the bounds ( [ lbpt ] ) and ( [ ubpt ] ) together implies ( [ maincond ] ) with @xmath298 for @xmath299 . therefore , if @xmath300 , part ( a ) follows by theorem [ gen ] . finally , if @xmath301 , then @xmath302 for @xmath294 and by theorem [ genfixed ] part ( b ) follows . it is well known that in the curie - weiss model for @xmath303 ( see ( * ? ? ? * example 3.9 ) ) @xmath304 note that @xmath305 where @xmath203 is a @xmath306 matrix with @xmath307 , for @xmath308 , and @xmath309 is a @xmath310 matrix with @xmath311 , for @xmath308 , and @xmath312 . therefore , @xmath313 note that @xmath314 and by @xmath302 for @xmath315 . thus , there exists no sequence of consistent estimators for @xmath316 by theorem [ genfixed ] . for @xmath317 , by @xmath318 since @xmath319 is the hamiltonian of a curie - weiss model on size @xmath2 . moreover , @xmath320 , for all @xmath321 and @xmath322 ; so taking @xmath323 , @xmath324 establishing condition ( a ) of theorem [ gen ] . therefore , the mple @xmath62 is @xmath98-consistent for @xmath99 by theorem [ gen ] . similarly , for @xmath100 @xmath325 and so the mple @xmath62 is @xmath2-consistent . note that when the sufficient statistic is of the form ( [ adjm ] ) , @xmath320 , for all @xmath322 . therefore , taking @xmath323 , @xmath326 which implies condition ( a ) of theorem [ gen ] . moreover , in this case , @xmath314 , and @xmath327 . therefore , part ( a ) follows by corollary [ mplepositive ] . by corollary [ sc ] , to show part ( b ) it suffices to verify that condition ( [ ptn ] ) holds for all @xmath109 . this is done using the mean field approximation of lemma [ meanfield ] . by plugging in the vector @xmath328 for the vector @xmath329 in the rhs of @xmath330}\left\{\frac{\beta_0 m^2}{2}-i(m)\right\ } , \label{znbeta0lb}\ ] ] where @xmath331 for @xmath332 $ ] . thus , it suffices to show that @xmath333}g(m)>0 $ ] , where @xmath334 . to this end , note that @xmath335 , that is , @xmath336 is not a local maximum of @xmath337 . this implies the rhs of ( [ znbeta0 lb ] ) is positive , thus verifying condition ( [ ptn ] ) . let @xmath338 be the degree of the vertex @xmath339 in @xmath103 , for @xmath274 . then @xmath340 , for all @xmath322 . in the regime @xmath122 , the maximum degree @xmath341 with high probability @xcite therefore , @xmath342 for all @xmath274 , and by taking @xmath343 it follows that @xmath344 , with high probability . this implies condition ( a ) of theorem [ gen ] . moreover , for @xmath122 , @xmath345 with high probability @xcite , and @xmath346 and part ( a ) follows from corollary [ mplepositive ] . to prove part ( b ) , we use the mean field approximation as in corollary [ dregular ] . by plugging in the vector @xmath328 for the vector @xmath329 in the rhs of we get @xmath330}\left\{\frac{\beta_0 m^2|e(g_n)|}{n^2p(n)}-i(m)\right\}.\ ] ] condition follows by arguments similar to those in corollary [ dregular ] and the fact @xmath347 . in this section we show the existence of a untestable / testable threshold in ising models on converging sequence of dense graphs , and compute the distribution and asymptotic power of the most powerful test , before the phase transition . if @xmath103 converges to @xmath186 , then @xmath348 converges to the operator norm of @xmath155 ( see ( [ eq : t ] ) ) . moreover , @xmath349 and part ( a ) follows by corollary [ mplepositive ] from ( * ? ? ? * theoem 2.14 ) , when @xmath103 converges to @xmath186 , then @xmath350 , where @xmath351\mapsto [ -1,1]}\left\{\frac{\beta}{2}\int_{[0,1]^2}m(x)m(y)w(x , y)\mathrm d x\mathrm d y-\int_0 ^ 1i(m(x))\mathrm d x\right\},\ ] ] and @xmath352 as in corollary [ dregular ] . by corollary [ sc ] , it enough to show that @xmath353 , for @xmath168 . to this end , let @xmath354 to be the eigenvector corresponding to the eigenvalue @xmath355 . then @xmath356 , and @xmath357}|v_1(x)|<\infty$ ] . thus , there exists @xmath219 such that for @xmath358 we have @xmath357}|zv_1(x)|\le 1 $ ] , and @xmath359 ^ 2}v_1(x)v_1(y)w(x , y)\mathrm d x\mathrm d y-\int_0 ^ 1i(z v_1(x))\mathrm d x\right\}\\ = & \sup_{|z|<\delta}\left\{\frac{\beta z^2\lambda}{2}\int_0 ^ 1 v_1(x)^2\mathrm d x-\int_0 ^ 1i(z v_1(x))\mathrm d x\right\}.\end{aligned}\ ] ] setting @xmath360 it suffices to show that @xmath361 is not a point of local maxima of the function @xmath362 . this follows on noting that @xmath363 . by lemma [ lem : exists ] ( see appendix [ pfpartition ] ) , the limiting distribution ( [ limitdense ] ) is well defined . the following proposition ( proved in appendix [ pfpartition ] ) gives the limit of the log - partition function , for a converging sequence of dense graphs , for @xmath169 . [ partition ] let @xmath149 be a sequence of weighted graphs with entries bounded in @xmath131 $ ] which converges in cut - metric to @xmath159 , such that @xmath160 ^ 2}w(x , y)^2\mathrm d x\mathrm d y>0 $ ] . then for any @xmath364 @xmath365 the above proposition can be used to complete the proof of theorem [ dense_testing ] as follows : fix @xmath219 such that @xmath366 . then for any @xmath367 , @xmath368 by proposition [ partition ] . by lemma [ lem : exists ] the rhs above is the mgf of the random variable @xmath369 defined in ( [ limitdense ] ) . 99 a. anandkumar , v. y. f. tan , f. huang , and a. s. willsky , high - dimensional structure estimation in ising models : local separation criterion , _ the annals of statistics _ , vol . 40 ( 3 ) , 13461375 vol . 2012 . s. banerjee , b. p. carlin , and a. e. gelfand , _ hierarchical modeling and analysis for spatial data _ , chapman & hall , boca raton , fl , 2004 . j. besag , spatial interaction and the statistical analysis of lattice systems , _ b _ , vol . 36 , 192236 , 1974 . j. besag , statistical analysis of non - lattice data , _ the statistician _ , vol . 24 ( 3 ) , 179195 , 1975 . b. b. bhattacharya , p. diaconis , and s. mukherjee , universal poisson and normal limit theorems in graph coloring problems with connections to extremal combinatorics , arxiv:1310.2336 , 2014 . g. bresler , efficiently learning ising models on arbitrary graphs , _ proc . symposium on theory of computing ( stoc ) _ , 771782 , 2015 . c. borgs , j.t . chayes , l. lovsz , v.t . ss , and k. vesztergombi , convergent sequences of dense graphs i : subgraph frequencies , metric properties and testing , _ advances in mathematics _ , vol . 219 , 18011851 , 2009 . c. borgs , j.t . chayes , l. lovsz , v.t . ss , and k. vesztergombi , convergent sequences of dense graphs ii . multiway cuts and statistical physics , _ annals of mathematics _ , vol . 176 , 151219 , 2012 . s. chatterjee , estimation in spin glasses : a first step , _ the annals of statistics _ 35 ( 5 ) , 19311946 , 2007 . f. comets , on consistency of a class of estimators for exponential families of markov random fields on the lattice , _ the annals of statistics _ , vol . 20 ( 1 ) , 455468 , 1992 . f. comets and b. gidas , asymptotics of maximum likelihood estimators for the curie - weiss model , _ the annals of statistics _ , vol . 19 ( 2 ) , 557578 , 1991 . a. dembo , a. montanari , and n. sun , factor models on locally tree - like graphs , _ ann . 41 , 41624213 , 2013 . a. dembo and a. montanari , gibbs measures and phase transitions on sparse random graphs , _ brazilian journal of probability and statistics _ , vol . 24 ( 2 ) , 137211 , 2010 . a. dembo and a. montanari , ising models on locally tree - like graphs . 20 , 565592 , 2010 . p. diaconis , the cutoff phenomenon in finite markov chains , _ proc . _ , vol . 93 ( 4 ) , 16591664 , 1996 . r. durrett , _ probability : theory and examples _ , 4th edition , cambridge u. press , 2010 . c. j. geyer and e. a. thompson , constrained monte carlo maximum likelihood for dependent data ( with discussion ) , _ j. roy . b _ , vol . 54 , 657699 , 1992 . b. gidas , consistency of maximum likelihood and pseudolikelihood estimators for gibbs distributions , _ in stochastic differential systems , stochastic control theory and applications ( w. fleming and p .- l . lions , eds . ) _ , 129145 , springer , new york , 1988 . p. j. green , and s. richardson , hidden markov models and disease mapping , _ journal of the american statistical association _ , 97 , 10551070 , 2002 . x. guyon , h. r. knsch , asymptotic comparison of estimators in the ising model , _ stochastic models , statistical methods , and algorithms in image analysis _ , lecture notes in statistics , vol , 74 , 177198 , 1992 . j. j. hopfield , neural networks and physical systems with emergent collective computational abilities , _ proc . 79 ( 8) , 25542558 , 1982 . e. ising , beitrag zur theorie der ferromagnetismus , _ zeitschrift fr physik _ , 31 , 253258 , 1925 . m. jerrum and a. sinclair , polynomial - time approximation algorithms for the ising model , _ siam journal of computing _ , vol . 22 , 10871116 , 1993 . m. krivelevich and b. sudakov , the largest eigenvalue of sparse random graphs , _ combinatorics , probability and computing _ , vol . 12 ( 1 ) , 6172 , 2003 . d. a. levin , m. luczak , and y. peres , glauber dynamics for the mean - field ising model : cut - off , critical power law , and metastability , _ probability theory and related fields _ , vol . 146 ( 1 - 2 ) , 223265 , 2010 . l. lovsz , _ large networks and graph limits _ , colloquium publications , vol . 60 , 2012 . e. lubetzky and a. sly , information percolation and cutoff for the stochastic ising model , arxiv:1401.6065 [ math.pr ] , 2015 , _ journal of the american mathematical society _ , to appear . e. lubetzky and a. sly , cutoff for the ising model on the lattice , _ inventiones mathematicae _ , vol . 191 ( 3 ) , 719755 , 2013 . j. l. jensen and j. mller , pseudolikelihood for exponential family models of spatial point processes , _ annals of applied probability _ , vol . 1 ( 3 ) , 445461 , 1991 . d. panchenko , _ the sherrington - kirkpatrick model _ , springer science & business media , 2013 . d. k. pickard , inference for discrete markov fields : the simplest nontrivial case , _ journal of the american statistical association _ , ( 397 ) , 9096 , 1987 . p. ravikumar , m. j. wainwright and j. lafferty , high - dimensional ising model selection using @xmath370-regularized logistic regression , _ the annals of statistics _ , vol . 38 ( 3 ) , 12871319 , 2010 . m. talagrand , _ spin glasses , a challenge to mathematicians _ , springer , 2003 . l. xue , h. zou , and t. cai , non - concave penalized composite conditional likelihood estimation of sparse ising models , _ annals of statistics _ , vol . 40 ( 3 ) , 14031429 , 2012 . in this appedix we prove the lemmas used in the proof of theorem [ gen ] . the rest of the section is organized as follows : appendix [ pflem1 ] contains the proof of lemma [ lem1 ] . the proofs of lemma [ 2moment ] and lemma [ derivative ] are given in appendix [ pflem2moment ] and appendix [ pflemderivative ] , respectively . by there exists @xmath371 such that @xmath372 . by the monotonicity of @xmath258 , @xmath373 it follows that @xmath374 . thus , for any @xmath375 @xmath376 which , on taking logarithms , implies that @xmath377 dividing both sides by @xmath1 and taking limits as @xmath378 followed by @xmath379 we have @xmath380 thus completing the proof of ( a ) . to show ( b ) , again invoking there exists @xmath219 such that @xmath381 . since @xmath382 it follows that @xmath383 thus , for any @xmath384 @xmath385 taking logarithm on both sides , @xmath386 from which dividing by @xmath1 and taking limits as @xmath387 followed by @xmath388 gives @xmath389 thus proving part ( b ) . we begin with a technical estimate which will be needed to bound the second moment of @xmath210 . [ lem0 ] under assumption and @xmath79 as defined in ( [ mi ] ) , @xmath390 by there exists @xmath219 such that @xmath391 . therefore , @xmath392 , and so @xmath393 now , observe that @xmath79 does not depend on @xmath46 , and @xmath394 . since @xmath395 the result follows from . the above lemma will be used to complete the proof of lemma [ 2moment ] . to this end , for @xmath396 and @xmath48 , let @xmath397 and @xmath398 from equation ( 10 ) of chatterjee @xcite it follows that @xmath399 setting @xmath400 , note that @xmath401 now , by a second order taylor expansion , @xmath402 where @xmath403 and @xmath404 for some @xmath405 in the interval @xmath406 $ ] . therefore , to prove the lemma , it suffices to control these three terms . to control @xmath407 , note that @xmath408 and @xmath409 for all @xmath410 for some @xmath411 , which gives @xmath412 which is finite as @xmath378 by an application of lemma [ lem0 ] . now , let us bound @xmath413 . by the cauchy - schwarz inequality , @xmath414 taking expectation on both sides above and using cauchy - schwarz inequality again @xmath415 now , since @xmath416 , for some constant @xmath417 , by lemma [ lem0 ] @xmath418 using this along with in ( [ t2cs ] ) gives @xmath419 . it remains to bound @xmath420 . since @xmath421 , we have @xmath422 where the last step uses @xmath423 . finally , taking expectations on both sides in ( [ t3bound ] ) , and using condition ( b ) on the first term , and ( [ t1 ] ) on the second term , gives @xmath424 . fixing @xmath219 by lemma [ lem1](a ) there exists @xmath425 such that @xmath426 for @xmath212 large enough . also , using lemma [ 2moment ] and chebyshev s inequality , for @xmath427 we have @xmath428 moreover , by condition ( a ) in theorem [ gen ] there exists @xmath429 such that for all @xmath212 large enough we have @xmath430 and so by markov s inequality @xmath431 defining @xmath432 we have @xmath433 , for @xmath212 large enough ( by combining ( [ c00 ] ) , ( [ c1 ] ) , and ( [ c0 ] ) ) . now , on the set @xmath434 using the bounds @xmath435 on @xmath436 , and @xmath437 on @xmath438 , @xmath439 thus , on the set @xmath440 , @xmath441 for all @xmath212 large , completing the proof . for every @xmath242 , let @xmath243 be a measure space and @xmath240 and @xmath241 two distributions on this measure space . recall the definition of kullback - leibler divergence @xmath442 from , and consider the problem of testing @xmath240 versus @xmath241 such that condition holds . since @xmath443 , by assumption ( [ divlemma ] ) @xmath444 for some @xmath445 and all large @xmath212 . also , there exists @xmath446 such that @xmath447 , for all @xmath212 . to see this , note that @xmath448 hence , by ( [ likelihood ] ) and ( [ ln ] ) , @xmath449 . therefore , by markov s inequality , for any @xmath375 @xmath450 now , suppose there exists a sequence of test functions @xmath248 such that @xmath451 . then @xmath452 taking limits on both sides gives , @xmath453 . since @xmath375 is arbitrary @xmath454 , that is , @xmath248 is not a consistent sequence of test functions . a standard technique to derive a lower bound on the log - partition function is the mean - field approximation ( refer to @xcite for details ) . here , we give a short proof for the sake of completeness . [ meanfield ] consider the family of probability distributions on @xmath14 given by ( [ pmf ] ) . then for any matrix @xmath455^n } \left\ { \frac{\beta}{2 } \it{\textbf{z}}'j_n\it{\textbf{z}}-\sum_{i=1}^ni(z_i)\right\ } , \label{mfieldb}\ ] ] where @xmath456 $ ] for @xmath332 $ ] . let @xmath457 be the kullback - leibler divergence between two probability measures . by a direction computation , for any probability mass function @xmath458 on @xmath459^n$ ] we have @xmath460 now , since @xmath461 we have @xmath462 one can obtain a lower bound on @xmath463 by taking supremum in lhs over product measures , that is @xmath464 . hence , setting @xmath465 $ ] , the bound in follows . we begin by deriving the mgf of the limiting distribution . the proof involves straightforward calculations using the mgf of the chi - squared distribution , similar to ( * ? ? ? * proposition 7.1 ) . to compute the moment generating function of @xmath480 , first note that @xmath481 . thus if the collection of random variables @xmath482 is uniformly integrable , then we have @xmath483 thus completing the proof of the lemma . it thus remains to prove uniform integrability , for which it suffices to show that for some @xmath219 we have @xmath484 since @xmath485 there exists @xmath219 such that @xmath486 . for this @xmath221 setting @xmath487 we have @xmath488 now setting @xmath489 we have @xmath490 for @xmath491 , and so the r.h.s . of can be bounded by @xmath492 which converges to @xmath493 therefore , @xmath494 is uniformly integrable , thus completing the proof of the lemma . the above lemma can be used to complete the proof of proposition [ partition ] . to this end , let @xmath495 . since the entries of @xmath496 are bounded in @xmath131 $ ] , the proof of ( * ? ? ? * theorem 1.4 ) can be generalized verbatim to get @xmath497 where @xmath498 are i.i.d . @xmath174 random variables . thus , @xmath499 if the lhs in is uniformly integrable , then @xmath500 where the last equality uses lemma [ lem : exists ] . the proof of part ( a ) then follows on taking log of both sides of the above equality . it remains to show that the lhs in is uniformly integrable , that is , @xmath501 for some @xmath219 . to this end , note that if @xmath502 , there exists @xmath219 such that @xmath503 . now , using and the fact @xmath504 , we have @xmath505 since @xmath506 in the cut metric , @xmath507 , and so there exists @xmath375 such that for all @xmath212 large enough @xmath508 . for @xmath509 there exists @xmath510 such that @xmath511 using this the rhs of can be bounded by @xmath512 which converges to @xmath513 , as @xmath378 . this proves ( [ unint ] ) and completes the proof of the proposition .

we present two examples of reductions from the evolution equations describing discrete schlesinger transformations of fuchsian systems to difference painlev equations : difference painlev equation d- with the symmetry group and difference painlev equation d- with the symmetry group . in both cases we describe in detail how to compute their okamoto space of the initial conditions and emphasize the role played by geometry in helping us to understand the structure of the reduction , a choice of a good coordinate system describing the equation , and how to compare it with other instances of equations of the same type .

1408.3778

it is well - known that many special functions , e.g. the airy , bessel , or legendre functions , originate as series solutions of _ linear _ ordinary differential equations ( odes ) . such functions play a crucial role in describing a wide range of important physical and mathematical phenomena . an essential point here is that for linear ordinary differential equations singularities of solutions can only occur at the points where the coefficients of the equation itself become singular . this makes it possible to talk about global properties of solutions ( and hence , of the corresponding special functions ) , and the asymptotic behavior of solutions near those fixed singular points . for nonlinear equations the situation is very different . although the cauchy existence theorem guarantees _ local _ existence of the solution to a given initial value problem at an ordinary ( regular ) point , in general the domain in which this solution is defined depends not just on the equation itself , but on the initial conditions as well solutions acquire _ movable _ ( i.e. , dependent on the initial values or , equivalently , on the constants of integration ) singularities . such singularities are called _ critical _ if a solution looses its single - valued character in a neighborhood of the singularity ( e.g. , when a singular point is a _ branch point _ ) . an ode is said to satisfy the _ painlev property _ if its general solution is free of _ movable _ critical singular points . otherwise , the riemann surface uniformizing such solution becomes dependent on the constants of integration , which prevents global analysis . thus , equivalently , the painlev property of an ode is the uniformizability of its general solution , see @xcite ( as well as the other excellent articles in the volume @xcite ) for a careful overview of these ideas . it is clear that liner equations satisfy painlev property . the importance of the painlev property for nonlinear equations is that , similar to the linear case , solutions of these equations give rise to new transcendental functions . in that sense , according to m.kruskal as quoted in @xcite , nonlinear equations satisfying this property are on the border between trivially integrable linear equations and nonlinear equations that are not integrable , and so the painlev property is essentially equivalent to ( and is a criterion of ) integrability . the search for new transcendental functions was the original motivation in the work of p.painlev who , together with his student b.gambier , had classified all of the rational second - order differential equations that have the painlev property , @xcite , @xcite , @xcite . among 50 classes of equations that they found , only _ six _ can not be reduced to linear equations or integrated by quadratures . these equations are now known as @xmath4 . solutions to these equations , the so - called _ painlev transcendents _ , are playing an increasingly important role in describing a wide range of nonlinear phenomena in mathematics and physics @xcite . almost simultaneously with the work of p.painlev and b.gambier , the most general painlev vi equation was obtained by r.fuchs @xcite in the theory of _ isomonodromic deformations _ of fuchsian systems . this theory , developed in the works of r.fuchs @xcite , l.schlesinger @xcite , r.garnier @xcite , and then extended to the non - fuchsian case by m.jimbo , t.miwa , and k.ueno @xcite , and also by h.flaschka and a.newell @xcite , as well as the related riemann - hilbert approach @xcite , @xcite , are now among the most powerful methods for studying the structure of the painlev transcendents . over the last thirty years a significant effort has been put towards understanding and generalizing results and methods of the classical theory of completely integrable systems to the _ discrete _ case . this is true for the theory of painlev equations as well . discrete painlev equations were originally defined as second order nonlinear difference equations that have usual painlev equations as continuous limits @xcite , @xcite . a systematic study of discrete painlev equations was started by b.grammaticos , j.hietarinta , f.nijhoff , v.papageorgiou and a.ramani , @xcite , @xcite , @xcite , and many different examples of such equations were obtained in a series of papers by grammaticos , ramani , and their collaborators by a systematic application of the singularity confinement criterion , see reviews @xcite , @xcite , and many references therein . discrete painlev equations also appear in a broad spectrum of important nonlinear problems in mathematics and physics , among which are the theory of orthogonal polynomials , quantum gravity , determinantal random point processes , reductions of integrable lattice equations , and , notably , as bcklund transformations of differential painlev equations . some of these problems are discrete anlogues or direct discretizations of the corresponding nonlinear problems , and some describe purely discrete phenomena . it turned out that classifying discrete painlev equations by their continuous limits , as well as the singularity confinement criterion , is not a very good approach , since such correspondence is far from being bijective . it is both possible for the same discrete equation to have different continuous painlev equations as continuous limits under different limiting procedures , and for different discrete equations to have the same continuous limit . in the seminal paper @xcite h.sakai showed that an effective way to understand and classify discrete painlev equations is through algebraic geometry . in this approach , to each equation , if we consider it as a two - dimensional first - order nonlinear system , we put in correspondence a family of algebraic surfaces @xmath5 , where @xmath6 is some collection of parameters that change , depending on the type of the equation , in an additive , multiplicative , or elliptic fashion as functions of a discrete `` time '' parameter . this family @xmath5 , by a slight abuse of terminology , is called the _ okamoto space of initial conditions _ ( that we often denote simply by @xmath7 omitting explicit dependence on parameters @xmath8 ) and it is obtained by resolving the indeterminacy points of the corresponding map @xmath9 via the blowing - up procedure . by sakai s theory , the complete resolution of indeterminacies is obtained by blowing up 9 ( possibly infinitely close ) points on @xmath10 ( or eight points on a birationally equivalent @xmath11 compactification of @xmath12 ) . the resulting surface @xmath7 has a special property that it admits a unique anti - canonical divisor @xmath13 of _ canonical type_. the orthogonal complement of @xmath14 in the picard lattice @xmath15 is described by the affine dynkin diagram @xmath16 that has two intersecting root subsystems of affine type : @xmath17 , that is generated by classes @xmath18 of irreducible components of the anticanonical divisor @xmath19 , and its orthogonal complement @xmath20 whose corresponding root lattice @xmath21 is called the _ symmetry sub - lattice_. then the _ type _ of the discrete painlev equation is the same as the _ type of its surface _ @xmath5 , which is just the type of an affine dynkin diagram describing the root system @xmath17 of irreducible components @xmath18 of @xmath19 ( essentially , the degeneration structure of the positions of the blowup points ) . moreover , nonlinear painlev dynamic now becomes a translation in the symmetry sub - lattice @xmath21 , and so it becomes `` _ _ linearized _ _ '' there , see @xcite for details . this is somewhat similar to the algebro - geometric integrability of classical integrable systems and soliton equations , where nonlinear dynamic is mapped to commuting linear flows on the jacobian of the spectral curve of the associated linear problem via the abel - jacobi map . one important observation from sakai s geometric approach is that in addition to _ additive _ ( difference ) and _ multiplicative _ ( @xmath22-difference ) _ discretizations _ of continuous painlev equations , there are some purely discrete painlev equations . it also led to the discovery of the master _ discrete painlev equation such that all of the other painlev equations can be obtained from it through degenerations ( which corresponds to more and more special configurations of the blow - up points ) . this degenerations can be described by the following scheme , where letters stand for the symmetry type of the equation ( which is the type of the affine dynkin diagram of the root subsystem @xmath20 ) , and the subscripts @xmath23 , @xmath22 , @xmath24 , and @xmath25 stand for elliptic , multiplicative , additive and differential painlev equations respectively , see figure [ fig : pain - diag ] . @xmath26 \matrix ( m ) [ matrix of math nodes , column sep={1pt},row sep={15pt } ] { \left(e^{(1)}_{8}\right)^{e } & & & & & & & \left(a_{1}^{(1)}\right)^{q } \\ \left(e^{(1)}_{8}\right)^{q } & \left(e^{(1)}_{7}\right)^{q } & \left(e^{(1)}_{6}\right)^{q } & \node{\left(d^{(1)}_{5}\right)^{q } } ; & \left(a^{(1)}_{4}\right)^{q } & \left((a_{1}\!\!+\!\ ! a_{2})^{(1)}\right)^{q } & \left((a_{1}\!\!+\!\!a_{1})^{(1)}\right)^{q } & \left(a_{1}^{(1)}\right)^{q } & \left(a_{0}^{(1)}\right)^{q } \\ \left(e^{(1)}_{8}\right)^{\delta } & \left(e^{(1)}_{7}\right)^{\delta } & \left(e^{(1)}_{6}\right)^{\delta } & \left(d^{(1)}_{4}\right)^{c,\delta } & \left(a^{(1)}_{3}\right)^{c,\delta } & \left((2a_{1})^{(1)}\right)^{c,\delta } & \left(a_{1}^{(1)}\right)^{c,\delta } & \left(a_{0}^{(1)}\right)^{c } \\ & & & & & \left(a_{2}^{(1)}\right)^{c,\delta } & \left(a_{1}^{(1)}\right)^{c,\delta } & \left(a_{0}^{(1)}\right)^{c } \\ } ; % \draw [ label = above ( m-1 - 1 ) : $ a$ ] \node [ blue ] at ( $ ( m-2-4.north ) + ( 0,0.1)$ ) { q-$p_{\text{vi}}$ } ; \node [ blue ] at ( $ ( m-3-4.south ) + ( 0,-0.1)$ ) { $ p_{\text{vi}}$,\ ! d-$p_{\text{v}}$ } ; \node [ blue ] at ( $ ( m-2-5.north ) + ( 0,0.1)$ ) { q-$p_{\text{v}}$ } ; \node [ blue ] at ( $ ( m-1-8.north ) + ( 0,0.1)$ ) { q-$p_{\text{i}}$ } ; \node [ blue ] at ( $ ( m-3-5.south ) + ( 0,-0.1)$ ) { $ p_{\text{v}}$,\ ! d-$p_{\text{iv}}$ } ; \node [ blue ] at ( $ ( m-3-5.south ) + ( 0,-0.4)$ ) { d-$p_{\text{iii}}$ } ; \node [ blue ] at ( $ ( m-2-6.north ) + ( 0,0.1)$ ) { q-$p_{\text{iv}}$,\ ! q-$p_{\text{iii}}$ } ; \node [ blue ] at ( $ ( m-3-6.south ) + ( 0,-0.1)$ ) { $ p_{\text{iii}}$ } ; \node [ blue ] at ( $ ( m-3-6.south ) + ( 0.1,-0.4)$ ) { alt.\ ! d-$p_{\text{ii}}$ } ; \node [ blue ] at ( $ ( m-4-6.south ) + ( 0,-0.1)$ ) { $ p_{\text{iv}}$,\ ! d-$p_{\text{ii}}$ } ; \node [ right , scale=0.8 ] at ( $ ( m-1-8.south east ) + ( -0.3,0.15)$ ) { $ { |\alpha|^{2}=8}$ } ; \node [ blue ] at ( $ ( m-4-7.south ) + ( 0,-0.1)$ ) { $ p_{\text{ii}}$,\ ! alt.d-$p_{\text{i}}$ } ; \node [ blue ] at ( $ ( m-4-8.south ) + ( 0,-0.1)$ ) { $ p_{\text{i}}$ } ; % \node [ right , scale=0.5 ] at ( $ ( m-1-8.south east ) + ( -0.5,0.05)$ ) { $ { |\alpha|^{2}=8}$ } ; \node [ right , scale=0.8 ] at ( $ ( m-2-7.south east ) + ( -0.95,0.02)$ ) { $ { |\alpha|^{2}=14}$ } ; \node [ right , scale=0.8 ] at ( $ ( m-2-8.south east ) + ( -0.65,0.02)$ ) { $ { |\alpha|^{2}=4}$ } ; \draw[- > ] ( m-1 - 1 ) edge ( m-2 - 1 ) ( m-2 - 1 ) edge ( m-3 - 1 ) ( m-2 - 2 ) edge ( m-3 - 2 ) ( m-2 - 3 ) edge ( m-3 - 3 ) ( m-2 - 4 ) edge ( m-3 - 4 ) ( m-2 - 5 ) edge ( m-3 - 5 ) ( m-2 - 6 ) edge ( m-3 - 6 ) ( m-2 - 7 ) edge ( m-3 - 7 ) ( m-2 - 8 ) edge ( m-3 - 8 ) ( m-3 - 5 ) edge ( m-4 - 6 ) ( m-3 - 6 ) edge ( m-4 - 7 ) ( m-3 - 8 ) edge ( m-4 - 8 ) ( m-2 - 7 ) edge ( m-1 - 8 ) ( m-1 - 8 ) edge ( m-2 - 9 ) ( m-2 - 9 ) edge ( m-4 - 8 ) ( m-2 - 8 ) edge ( m-4 - 7 ) % ( $ ( m-2-1.east ) + ( -0.15,0)$ ) edge ( $ ( m-2-2.west ) + ( 0.05,0)$ ) ( $ ( m-2-2.east ) + ( -0.15,0)$ ) edge ( $ ( m-2-3.west ) + ( 0.05,0)$ ) ( $ ( m-2-3.east ) + ( -0.15,0)$ ) edge ( $ ( m-2-4.west ) + ( 0.05,0)$ ) ( $ ( m-2-4.east ) + ( -0.15,0)$ ) edge ( $ ( m-2-5.west ) + ( 0.05,0)$ ) ( $ ( m-2-5.east ) + ( -0.15,0)$ ) edge ( $ ( m-2-6.west ) + ( 0.05,0)$ ) ( $ ( m-2-6.east ) + ( -0.15,0)$ ) edge ( $ ( m-2-7.west ) + ( 0.05,0)$ ) ( $ ( m-2-7.east ) + ( -0.15,0)$ ) edge ( $ ( m-2-8.west ) + ( 0.05,0)$ ) % ( $ ( m-2-8.east ) + ( -0.15,0)$ ) edge ( $ ( m-2-9.west ) + ( 0.05,0)$ ) % ( $ ( m-3-1.east ) + ( -0.15,0)$ ) edge ( $ ( m-3-2.west ) + ( 0.05,0)$ ) ( $ ( m-3-2.east ) + ( -0.15,0)$ ) edge ( $ ( m-3-3.west ) + ( 0.05,0)$ ) ( $ ( m-3-3.east ) + ( -0.15,0)$ ) edge ( $ ( m-3-4.west ) + ( 0.05,0)$ ) ( $ ( m-3-4.east ) + ( -0.15,0)$ ) edge ( $ ( m-3-5.west ) + ( 0.05,0)$ ) ( $ ( m-3-5.east ) + ( -0.15,0)$ ) edge ( $ ( m-3-6.west ) + ( 0.05,0)$ ) ( $ ( m-3-6.east ) + ( -0.15,0)$ ) edge ( $ ( m-3-7.west ) + ( 0.05,0)$ ) % ( $ ( m-4-6.east ) + ( -0.15,0)$ ) edge ( $ ( m-4-7.west ) + ( 0.05,0)$ ) ( $ ( m-4-7.east ) + ( -0.15,0)$ ) edge ( $ ( m-4-8.west ) + ( 0.05,0)$ ) % ( $ ( m-3-7.east ) + ( -0.15,0)$ ) - > ( $ ( m-3-8.west ) + ( 0.05,0)$ ) ; \draw [ - ] ( $ ( m-2-6.south ) + ( -0.48,0)$ ) to [ bend right = 50 ] ( $ ( m-4-6.north ) + ( -0.48,0.02)$ ) ; % \draw[->,gray , rounded corners ] ( kc ) -| node[auto , text = black , pos=.7 ] % { \(\partial\ ) } ( $ ( 01.east)+(.5,0)$ ) |- ( $ ( b)!.35!(b')$ ) -| % ( $ ( 02.west)+(-.5,0)$ ) |- ( ca ) ; \end{tikzpicture}\ ] ] the following question then becomes natural and important : _ how to represent these new purely discrete equations in the isomonodromic framework ? _ this question was posed by sakai in @xcite ( * problem a * for the difference case and * problems b , c * for the @xmath22-difference case ) . more precisely , in both continuous and discrete difference case we start with some fuchsian system and consider its isomonodromic deformations . in the continuous case , deformation parameters are locations of singular points of the system . the resulting isomonodromic flows on the space of coefficients of the fuchsian system are given by schlesinger _ equations_. in particular , for a @xmath27 fuchsian system with four poles , schlesinger equations reduce to the most general @xmath28 equation . in the discrete difference case , deformation parameters are the characteristic indices of the system and since the isomonodromy condition requires that the indices change by integral shifts , the resulting dynamic is indeed _ discrete_. it is expressed in the form of _ difference _ equations called schlesinger _ transformations_. it is also possible to get the isomonodromic description of difference and @xmath22-difference painlev equations by studying deformations of linear difference and @xmath22-difference analogues of fuchsian systems , see @xcite and @xcite for details . in @xcite we studied in detail a particular class of schlesinger transformations that are called _ these transformations change only two of the characteristic indices of the underlying fuchsian system ( and any other schlesinger transformation not involving characteristic indices with multiplicity can be represented as a composition of the elementary ones ) . in particular , we obtained explicit evolution equations governing the resulting discrete dynamic . our objective for the present paper is to carefully and very explicitly describe reductions of these _ discrete schlesinger evolution equations _ to the _ difference _ painlev equations . since the painlev equations are of second order , we focus on fuchsian systems that have two - dimensional moduli spaces ( coordinates on such moduli space are known as accessory parameters ) . it turns out that , modulo two natural transformations called _ katz s addition _ and _ middle convolution _ @xcite , that preserve the deformation equations @xcite , there are only four such systems , @xcite , that have the _ spectral type _ @xmath29 , @xmath30 , @xmath31 , and @xmath32 ( spectral type of a fuchsian system encodes the degeneracies of the characteristic indices , or eigenvalues of residue matrices at singular points , and it is carefully defined in the next section ) . isomonodromic deformations of a @xmath29 spectral type fuchsian system are well known continuous deformations reduce to painlev vi equation and schlesinger transformations reduce to the difference painlev d-@xmath33 equation , also known as d-@xmath34 , and in @xcite we showed that in this case our _ discrete schlesinger evolution equations _ indeed can be reduced to the standard form of d-@xmath33 . in @xcite p.boalch showed that for fuchsian systems with the spectral types @xmath30 , @xmath31 , and @xmath32 their schlesinger transformations reduce to difference painlev equations with the required symmetry groups @xmath1 , @xmath3 , and @xmath16 respectively , thus providing a theoretical answer to sakai s * problem a*. our goal for the present paper is to make this statement very concrete via explicit direct computation of reductions of discrete schlesinger evolution to difference painlev equations with symmetry groups @xmath1 and @xmath3 ( we plan to consider deformations of a fuchsian system of the spectral type @xmath32 with the symmetry group @xmath16 elsewhere ) . in addition to establishing that the resulting difference painlev equations have the required types d-@xmath35 and d-@xmath36 , we explicitly compare the resulting equations with the previously known instances of equations of the same type . we do so by finding an explicit identification between their okamoto spaces of initial conditions , which allows us to compute and compare the translation directions for different equations w.r.t . the same root basis , and also to match generic parameters to the characteristic indices of the fuchsian system , which in turn allows us to see and compare these different equations via their actions on the riemann scheme of our fuchsian system . we show that in both examples elementary schlesinger dynamic is indeed more elementary in the sense that standard examples of difference painlev equations can be realized as compositions of elementary schlesinger transformations . of particular interest here is the @xmath3 case which has two characteristic indices of multiplicity @xmath37 . we show that in that case the standard from of the equation can not be represented as a composition of elementary schlesinger transformations of rank one . thus we first generalize our discrete schlesinger evolution equations from @xcite to elementary schlesinger transformations of rank two , and then show how to represent the standard dynamic as a composition of two such rank two transformations . we also provide a lot of details on how to compute the okamoto space of initial conditions for our equations and how to identify two different instances of such spaces , hoping that this will be helpful for other researchers who are interested in the geometric approach to discrete painlev equations . the paper is organized as follows . in section [ sec : preliminaries ] we briefly describe our parameterization of a fuchsian system by its spectral and eigensystem data , define elementary schlesinger transformations , present evolution equations for elementary schlesinger transformations as a dynamic on the space of coefficient of our fuchsian system , and then show how to split them to get the dynamic on the space of eigenvectors of the coefficient matrices ( this is a brief overview of our paper @xcite ) . next we show how to generalize this to elementary schlesinger transformations of rank two , which is a new result . in section [ sec : reductions ] we consider two examples of reductions of the elementary schlesinger transformation dynamic . the first example of a difference painlev equation of type d-@xmath38 with the symmetry group @xmath39 was also briefly presented in @xcite , here we go into a lot more detail and show how the choice of good coordinates , which was essentially guessed in @xcite , is really forced on us by the geometric considerations . the next example of a @xmath40 fuchsian system of the _ spectral type _ @xmath41 ( i.e. , with three poles and two double eigenvalues at one pole ) is a first example that we have which has degenerate eigenvalues , and this is a completely new result . finally , we give a brief summary in section [ sec : conclusion ] . part of this work was done when a.d . was visiting t.t . at the tokyo university of marine science and technology and nalini joshi at the university of sydney , and a.d . would like to thank both universities for the stimulating working environment and , together with the university of northern colorado , for the generous travel support . the goal of this section is to write down evolution equations for elementary schlesinger transformations , as well as to introduce the necessary notation . our presentation here is very brief and we refer the interested reader to @xcite for details . the main new and important result of this section is the generalization of equations governing elementary schlesinger transformation dynamic on the decomposition space from rank - one to rank - two schlesinger transformations . consider a generic _ fuchsian system _ ( or a _ fuchsian equation _ ) written in the _ schlesinger normal form _ : @xmath42 where @xmath43 are constant @xmath44 matrices . in addition to simple poles at finite distinct points @xmath45 , this system also has a simple pole at @xmath46 if @xmath47 . the _ spectral data _ of system ( [ eq : fuchs-1 ] ) consists of locations of the simple poles @xmath48 and the eigenvalues ( also called the _ characteristic indices _ ) @xmath49 of @xmath50 and their multiplicities . these multiplicities are encoded by the _ spectral type _ of the system , @xmath51 where partitions @xmath52 , @xmath53 describe the multiplicities of the eigenvalues of @xmath50 . spectral type classifies fuchsian systems up to isomorphisms and the operations of addition and middle convolution . schlesinger transformations are discrete analogues of the usual schlesinger differential equations describing isomonodromic deformations of our fuchsian system . they are rational transformations preserving the singularity structure and the monodromy data of the system ( [ eq : fuchs-1 ] ) , except for the integral shifts in the characteristic indices @xmath49 , and so the coefficient matrix now depends on @xmath49 , @xmath54 . schlesinger transformations are given by the following _ differential difference lax pair _ : @xmath55 where @xmath56 is a specially chosen rational matrix function called the _ multiplier _ of the transformation . the compatibility condition of this lax pair is @xmath57 in @xcite we considered a special class of schlesinger transformations for which the multiplier matrix has the form @xmath58 it turns out that in this case it is possible to solve equation ( [ eq : sch - transf ] ) explicitly to obtain a discrete dynamic on the space of coefficient matrices . namely , after substituting @xmath56 of the form ( [ eq : mult - proj - gen ] ) in ( [ eq : sch - transf ] ) ( and its inverse ) we immediately see that @xmath59 , and if we put @xmath60 and @xmath61 , we get the following equations on the coefficient matrices : @xmath62 then either ( [ eq : q - cond ] ) or ( [ eq : p - cond ] ) imposes important constraints on the projector @xmath63 : @xmath64 and if this condition is satisfied , we get the following dynamic on the space of coefficient matrices : @xmath65 indeed , @xmath66 and the equation for @xmath67 is obtained in a similar fashion . we call these equations _ discrete schlesinger evolution equations_. it is sometimes more convenient to extend the dynamic from the space of coefficients of the fuchsian system to the space of eigenvectors of the coefficient matrices , we call this space the _ decomposition space_. in particular , this is the space on which both the continuous ( @xcite ) and the discrete ( @xcite ) hamiltonian equations for schlesinger deformations can be written . before defining this space it is convenient to reduce the number of parameters in our system by using scalar gauge transformations of the form @xmath68 , where @xmath69 is a solution of the _ scalar _ equation @xmath70 such transformations change the residue matrices by @xmath71 ( and consequently change the residue matrix at infinity by @xmath72 ) . hence we can always make one of the eigenvalues @xmath73 by choosing a good representative w.r.t . the action by the group of local _ scalar gauge transformations_. so we make the following assumption . [ assume : rank - reduce ] we always assume that at the finite point @xmath74 the eigenvalue @xmath75 of the highest multiplicity @xmath76 is zero . we also need the following important _ semi - simplicity _ assumption . [ assume : diagonal ] we assume that the coefficient matrices @xmath50 are _ diagonalizable _ ( even when we have multiple eigenvalues ) . in view of these assumptions , coefficient matrices @xmath50 are similar to diagonal matrices @xmath77 , where @xmath78 . omitting the zero eigenvalues , we put @xmath79 further , in view of assumption ( [ assume : diagonal ] ) there exist full sets of _ right _ eigenvectors @xmath80 , @xmath81 , and _ left _ eigenvectors @xmath82 , @xmath83 ( we use the @xmath84 symbol to indicate a _ row_-vector or a matrix of row vectors ) . in the matrix form , omitting vectors with indices @xmath85 that are in the kernel of @xmath50 , we can write @xmath86 with @xmath87 defined by ( [ eq : theta - i ] ) . then we have a decomposition @xmath88 , provided that @xmath89 . we call this last condition the _ orthogonality condition _ ( since @xmath87 is diagonal ) and we assume that it holds even when we have repeating eigenvalues . this condition is related to the normalization ambiguity of the eigenvectors . thus , given @xmath50 , we can construct ( in a non - unique way ) a corresponding decomposition pair @xmath90 . the space of all such pairs for all finite indices @xmath91 , without any additional conditions , is our _ decomposition space_. we denote it as @xmath92 and write an element @xmath93 of this space as a list of @xmath94 pairs @xmath95 . then , given a riemann scheme of a fuchsian system ( equivalently , a collection @xmath96 of the characteristic indices having the correct multiplicities and satisfying the fuchs relation ) , we denote by @xmath97 the corresponding fiber in the decomposition space ( since for schlesinger transformations locations of the poles are just fixed parameters of the dynamic , we occasionally omit them , as in the above notation ) . [ rem : transf ] there are two natural actions on the decomposition space @xmath98 . first , the group @xmath99 of global gauge transformations of the fuchsian system induces the following action . given @xmath100 , we have the action @xmath101 which translates into the action @xmath102 . we refer to such transformations as _ similarity transformations_. it is often necessary to restrict this action to the subgroup @xmath103 preserving the form of @xmath104 . second , for any pair @xmath90 the pair @xmath105 determines the same matrix @xmath50 for @xmath106 . the condition @xmath107 restricts @xmath108 to the stabilizer subgroup @xmath109 of @xmath110 . in particular , when all @xmath49 are distinct , @xmath108 has to be a diagonal matrix . we refer to such transformations as _ trivial transformations_. these two actions obviously commute with each other . the phase space for the schlesinger dynamic is the quotient space of @xmath111 by this action . in this section we explain how to lift the schlesinger evolution equations to the decomposition space . in @xcite we focused on the _ elementary _ schlesinger transformations @xmath112 that only change two of the characteristic indices by unit shifts , i.e. , @xmath113 and @xmath114 , @xmath115 . for such transformations the projector matrix @xmath63 has rank one and the multiplier matrix has the form ( [ eq : mult - proj - gen ] ) with @xmath116 in this case , under the semi - simplicity assumption ( [ assume : diagonal ] ) it is possible to decompose equations ( [ eq : ai][eq : abeta ] ) to get the dynamic on the space @xmath111 . [ thm : evolution ] an elementary schlesinger transformation @xmath117 defines the map @xmath118 given by the following evolution equations ( grouped for convenience ) , where @xmath119 are arbitrary non - zero constants . a. transformation vectors : @xmath120 b. generic indices : @xmath121 c. special indices : @xmath122 for the difference painlev equation d-@xmath123 we need to study schlesinger transformations of a fuchsian system that has the spectral type @xmath41 , and so we need to consider schlesinger transformations that change not one but two eigenvalues at each point @xmath124 and @xmath125 . in this section we show how to obtain the corresponding dynamic on the decomposition space . the resulting equations suggest what happens in the general case of a projector @xmath63 of arbitrary rank , but since the focus of the present paper is on examples , we plan to consider the general case elsewhere . naively , we want to consider schlesinger transformations of the form @xmath126 however , if one of the characteristic indices ( say , @xmath127 ) has multiplicity , applying a rank - one elementary schlesinger transformation will change the spectral type of the equation ( e.g. , in our example , a rank - one transformation @xmath128 maps the moduli space of fuchsian equations of spectral type @xmath31 to a smaller moduli space @xmath129 , and in fact our formulas in this case do not work , since some of the expressions become undefined ) . thus , we need to develop the rank - two version of the elementary schlesinger transformation separately . we start with a composition of two rank - one maps to get an insight on the structure of the multiplier matrix in the rank - two case , but then proceed to derive the dynamic equations independently . the resulting equations are then defined on moduli spaces of fuchsian systems that have multiplicity in the spectral type ( e.g. , in our example , the map @xmath130 is defined on both moduli spaces of fuchsian systems of spectral type @xmath131 and @xmath41 ) . so we start with the multiplier matrix that is a product ( and for simplicity we put @xmath132 and @xmath133 for this derivation ) @xmath134 let us now rewrite @xmath135 in a more symmetric form . first note that , since @xmath136 are rank - one projectors , @xmath137 also , note that for any rank - one projector @xmath138 and for any matrix @xmath139 we have an identity @xmath140 . therefore , @xmath141 where @xmath142 are two mutually orthogonal rank - one projectors , @xmath143 , @xmath144 . at the same time , since obviously @xmath145 , @xmath146 , where @xmath147 we also put @xmath148 , @xmath149 , and @xmath150 . in view of the orthogonality conditions @xmath151 , it is easy to describe the eigenvectors for each of those projectors ( we do it just for @xmath135s since for @xmath152s eigenvectors are the same but eigenvalues swap between @xmath153 and @xmath154 , below we use the notation @xmath155 , where @xmath156 is an eigenvalue ( which is either @xmath153 or @xmath154 for projectors ) , @xmath157 is a row ( or left ) eigenvector and @xmath158 is a column ( or right ) eigenvector ) : @xmath159 note that the sum of two rank - one projectors is not a projector . here @xmath135 is the `` correct '' way to add @xmath160 and @xmath161 so that the result is a rank - two projector that is a sum of two orthogonal rank - one projectors and that has the same row and column spaces as @xmath162 . also , note that there are many ways to choose bases in the row and column ranges of @xmath135 , the choices above reflect the splittings @xmath163 . we can now use these projectors to split the discrete schlesinger evolution equations to define dynamic on eigenvectors . the proof is very similar in spirit to the rank - one case proof in @xcite . [ thm : rank2 ] consider a multiplier matrix in the form @xmath164 and @xmath165 , @xmath166 are given by ( [ eq : p - projs][eq : pt - projs ] ) . then @xmath135 satisfies the constraints ( [ eq : p - constr ] ) and so defines a schlesinger transformation . this schlesinger transformation has the type @xmath167 and the corresponding map @xmath168 is given by the following evolution equations , where @xmath119 are again arbitrary non - zero constants . a. transformation vectors : @xmath169 b. generic indices : @xmath170 c. special indices ( here @xmath171 , @xmath172 , and @xmath173 ) : @xmath174 of course the statement that @xmath135 defines an elementary schlesinger transformation of the type @xmath175 follows from how we derived it , but it can also be seen directly . e.g. , conditions ( [ eq : p - projs][eq : pt - projs ] ) follow immediately from @xmath176 and the fact that @xmath177 can be seen , in particular , from our derivation of the evolution equations below . to establish ( i ) , we use ( [ eq : p - cond ] ) : @xmath178 since @xmath179 and @xmath180 , we see that @xmath181 and @xmath182 , where @xmath183 stands for ` proportional ' . then @xmath184 , where @xmath185 is some non - zero proportionality constant . the other equations in this part are proved similarly . note that the consequence of ( i ) is that we can write @xmath186 for the generic case @xmath187 in ( ii ) the proof is identical to the rank - one case . since it is also short , we opted to include it to make the paper more self - contained . from ( [ eq : ai ] ) we see that @xmath188 and so @xmath189 and @xmath190 , where @xmath191 is a diagonal matrix of non - zero proportionality constants . similarly , @xmath192 . the orthogonality condition @xmath193 implies that @xmath194 , which gives ( ii ) for generic indices . for @xmath195 , from ( [ eq : q - cond ] ) we see that @xmath196 , and so again @xmath197 . however , since @xmath198 , @xmath199 and we can not recover @xmath200 and @xmath201 . the case @xmath202 is similar . finally , let us consider special indices . to find @xmath203 for @xmath173 start with ( [ eq : aalpha ] ) and ( [ eq : cb - generic - rank2 ] ) : @xmath204 multiplying on the right by @xmath205 , using the orthogonality conditions and @xmath206 , @xmath207 , we get @xmath208 now left - multiply by @xmath200 and use expression ( [ eq : p - mixed ] ) for @xmath135 and orthogonality conditions again to get @xmath209 this gives @xmath210 repeating the same steps for @xmath211 , substituting the result into ( [ eq : baj - rank2 ] ) , solving for @xmath203 and simplifying gives ( [ eq : bb - ak - rank2 ] ) ; ( [ eq : cb - bk - rank2 ] ) is proved in a similar fashion . finally , to get expressions for @xmath212 and @xmath213 , use all of the previously obtained expressions to write @xmath214 multiplying on the right by @xmath215 we get @xmath216 solving for @xmath212 gives ( [ eq : bb - bn - rank2 ] ) for @xmath217 , and the expression for @xmath218 is obtained by right - multiplying by @xmath218 instead . equations ( [ eq : cb - am - rank2 ] ) are obtained along the same lines . in this section , which is the central section of the paper , we consider two examples of reductions from the schlesinger dynamic on the decomposition space to difference painlev equations . first we consider schlesinger transformations of a fuchsian system of spectral type @xmath219 . resulting difference painlev equation is of type d-@xmath220 and has the symmetry group @xmath1 . we have previously considered this example in @xcite , but the exposition there was very brief and it relied on a nontrivial observation on how to choose good coordinates parameterizing our fuchsian system . here we not only provide more details but also show how geometric considerations _ lead us _ to the appropriate coordinate choice . in the second example we consider schlesinger transformations of a fuchsian system of spectral type @xmath41 , which gives difference painlev equation of type d-@xmath123 with the symmetry group @xmath221 . this example is completely new and here , in addition to elementary schlesinger transformations of rank one , we also , for the first time , consider elementary schlesinger transformations of rank two we need such transformations to represent the standard example of a difference painlev equation of type d-@xmath123 , as written in @xcite , @xcite , as a composition of elementary schlesinger transformations . for our model example of type d-@xmath220 we take the equation that was first written by grammaticos , ramani , and ohta , @xcite , see also murata @xcite and sakai @xcite . following sakai s geometric approach , we view this equation as a birational map @xmath222 with parameters @xmath223 @xmath224 where @xmath225 , @xmath226 , @xmath227 , @xmath228 , @xmath229 , @xmath230 , @xmath231 , @xmath232 , @xmath233 , and @xmath234 and @xmath235 are given by the equation @xmath236 this map has the following eight indeterminate points : @xmath237 resolving which by the blow - up procedure then gives us a rational surface @xmath5 , known as the _ okamoto space of initial conditions _ for this difference painlev equation , that is described by the blow - up diagram on figure [ fig : dpa2-standard - blowup ] . for the model form of d-@xmath220 . ] the picard lattice of @xmath5 is generated by the total transforms @xmath238 and @xmath239 of the coordinate lines and the classes of the exceptional divisors @xmath240 , @xmath241 the anti - canonical divisor @xmath242 uniquely decomposes as a positive linear combination of @xmath243-curves @xmath18 , @xmath244 , where the irreducible components @xmath18 , in bold on figure [ fig : dpa2-standard - blowup ] , are given by @xmath245 the configuration of components @xmath18 is described by the dynkin diagram of type @xmath246 ( with nodes corresponding to classes of self - intersection @xmath243 and edges connecting classes of intersection index @xmath154 ) . to this diagram correspond two different types of surfaces , the generic one corresponds to the multiplicative system of type @xmath246 , and the degenerate configuration , where all three components @xmath18 intersect at one point , corresponds to the additive system denoted by @xmath247 , which is clearly our case , see figure [ fig : dpa2-configs ] . [ cols="^,^,^ " , ] the symmetry sub - lattice @xmath248 is of type @xmath3 , where the basis of @xmath249 is given on figure [ fig : dpa1-symm ] . @xmath250 to compute the action of @xmath251 on @xmath252 , we decompose @xmath253 , where @xmath254 is given by ( [ eq : dpa1-phi1f ] ) , and @xmath255 is given by ( [ eq : dpa1-phi1 g ] ) . then @xmath256 , it is straightforward to see that @xmath257 and that @xmath258 , @xmath259 , which gives @xmath260 , where @xmath261 . the situation with @xmath262 is completely symmetric , @xmath263 , @xmath264 , and @xmath265 . composing these two linear maps , we get the action of @xmath251 : @xmath266 and so the induced action @xmath251 on the sub - lattice @xmath20 is given by the following translation : @xmath267 we start with a @xmath40 fuchsian system of the spectral type @xmath268 . in this case it is convenient to have all singular points to be finite , since we need to consider two different kinds of elementary schlesinger transformations one between two points with non - repeating eigenvalues , and the other when at one point we have an eigenvalue of multiplicity two . as before , we can use scalar gauge transformations to make some of the eigenvalues to vanish , and so , putting @xmath269 we take our riemann scheme to be @xmath270 we first consider an elementary schlesinger transformation @xmath271 for which @xmath272 and @xmath273 . the multiplier matrix for this transformation is @xmath274 since this transformation does not involve the point @xmath275 with multiple eigenvalues , the dynamic is again given by equations ( [ eq : bb - cb - generators][eq : cb - am ] ) that now take the form @xmath276 where @xmath119 are again arbitrary non - zero constants . similarly to the previous example , we parameterize the matrices as @xmath277 using the condition that the eigenvalues of @xmath278 are @xmath279 we get a system of three linear equations on @xmath127 , @xmath280 , and @xmath281 with coefficients depending on @xmath282 and @xmath283 , which gives us rational functions @xmath284 , @xmath285 , and @xmath286 ( again , the resulting expressions are quite large and we omit them ) . thus , the space of accessory parameters for fuchsian systems of this type is two - dimensional and @xmath282 and @xmath283 are some coordinates on this space . the resulting mapping @xmath287 becomes very complicated ( and computing it requires a computer algebra system , in our work we have used * mathematica * ) and so we omit equations describing the map . nevertheless , it is possible to do a complete geometric analysis of the mapping . we find that the indeterminate points of @xmath288 are @xmath289 as well as the sequence of infinitely close points @xmath290 note also that the points @xmath291 all lie on a @xmath292-curve @xmath293 given by the equation @xmath294 resolving indeterminate points of this map using blow - ups gives us the okamoto surface @xmath295 pictured on figure [ fig : dpa1-schlesinger - blowup ] . note that the @xmath243 curves @xmath296 and @xmath297 touch at the point with coordinates @xmath298 . thus , we immediately see that this is indeed a surface of type @xmath299 . for the schlesinger transformations reduction to d-@xmath123 . ] we now proceed to match the surface @xmath300 described by the blow - up diagram on figure [ fig : dpa1-schlesinger - blowup ] with the surface @xmath5 described by diagram on figure [ fig : dpa1-standard - blowup ] . as in the previous example , we look for rational classes @xmath301 , @xmath302 , @xmath303 in @xmath304 such that @xmath305 and the resulting configuration matches diagram on figure [ fig : dpa1-standard - blowup ] . this time the ( virtual ) genus formula @xmath306 suggests we see that we should look for classes of rational curves of self - intersection zero among @xmath307 , @xmath308 , or @xmath309 and for classes of rational curves of self - intersection @xmath310 among @xmath311 , @xmath312 , @xmath313 , or @xmath314 . it is again convenient to start by comparing the @xmath243-curves on both diagrams , @xmath315 given the uniformity of the coordinates of @xmath316 , @xmath317 , we see that it makes sense to choose @xmath318 for @xmath317 , and also we can put @xmath319 . this results in @xmath320 , which suggests taking @xmath321 and @xmath322 ( and so @xmath323 and @xmath324 ) . then @xmath325 , so we take @xmath326 , @xmath327 , and @xmath328 . it is not very hard to show that such choice satisfies all of our requirements and moreover , it is essentially unique ( up to a permutation of the indices of exceptional divisors ) . to summarize , we get the following identification : @xmath329 let us now define the base coordinates @xmath330 and @xmath331 of the linear systems @xmath332 and @xmath333 that will map the exceptional fibers of the divisors @xmath334 to the points @xmath335 such that @xmath336 are on a line @xmath337 and @xmath338 are on the line @xmath339 . since the pencil @xmath332 consists of @xmath340 curves on @xmath341 passing through @xmath342 and @xmath343 , @xmath344 and @xmath345 . next we will do a series of affine change of variables to arrange that the points @xmath316 , @xmath317 , are on a line @xmath337 . we have ( below @xmath317 ) @xmath346 thus , our final change of coordinates is @xmath347 and for @xmath317 , @xmath348 , @xmath349 and so these points lie on the line @xmath350 . we also get the identification of some of the parameters , @xmath351 for @xmath317 , and @xmath352 . it remains to verify that this change of variables puts points @xmath338 on the line @xmath353 and identify the remaining parameters . the exceptional divisor @xmath354 in the @xmath355-coordinates corresponds to the line @xmath356 , and so @xmath357 . similarly , @xmath358 . the exceptional divisor @xmath359 corresponds to the line @xmath360 , and so @xmath361 . finally , @xmath362 is given by @xmath363 , and so @xmath364 . thus , we see that indeed @xmath338 lie on the line @xmath365 and the remaining identification between the parameters is @xmath366 , @xmath367 , @xmath368 , and @xmath369 . we are now in the position to compare the dynamic given by an elementary schlesinger transformation with the dynamic of our model example of d-@xmath370 . as in the previous example , there are two different ways to do so . first , we can compute the corresponding translation vector . it is not very difficult to show that the action of @xmath371 of an elementary schlesinger transformation @xmath271 on the classes @xmath301 , @xmath302 , and @xmath334 is @xmath372 comparing the action of @xmath371 with the action of the standard dynamic @xmath251 given by ( [ eq : dpa1-trans - std ] ) on the symmetry sub - lattice , we see that the translation vectors are different : @xmath373 to get more insight into the relationship between schlesinger transformations and the standard d-@xmath370 dynamic , it is better to compute the action of d-@xmath370 on the riemann scheme of our fuchsian system using the above identification of parameters : @xmath374 thus , we see that the standard d-@xmath370 dynamic changes the multiple eigenvalue and so it requires the use of rank - two elementary schlesinger transformations . in fact , the action on the riemann scheme suggests that @xmath375 thus , consider the transformation @xmath376 changing the characteristic indices by @xmath377 , @xmath377 , and @xmath378 . this transformation is given by equations ( [ eq : bb - cb - generators - rank2][eq : cb - am - rank2 ] ) that now take the form @xmath379 where @xmath380 the okamoto surface for this dynamic is the same as before ( it depends only on the fuchsian system rather than a particular transformation ) , and its action @xmath381 on @xmath304 is given by @xmath382 and the action on the symmetry sub - lattice is @xmath383 similarly , the action @xmath384 of the rank - two transformation @xmath385 on @xmath304 is given by @xmath386 and its action on the symmetry sub - lattice is @xmath387 thus , @xmath388 finally , using a computer algebra system we can verify by a direct calculation that @xmath389 holds on the level of equations as well . in this work we further develop a theory of discrete schlesinger evolution equations that correspond to elementary schlesinger transformations of ranks one and two of fuchsian systems . we showed how to obtain difference painlev equations of types d-@xmath0 and d-@xmath2 as reductions of elementary schlesinger transformations . we also tried to make our computations very detailed in order to illustrate general techniques on how to study discrete painlev equations geometrically . one interesting observation is that standard examples of difference painlev equations of these types in both cases can be represented as compositions of elementary schlesinger transformations . thus , schlesinger dynamic should in principle be simpler and it would be interesting to find a nice and simple form of equations giving this dynamic . anton dzhamay , hidetaka sakai , and tomoyuki takenawa , _ discrete schlesinger transformations , their hamiltonian formulation , and difference painlev equations _ , http://arxiv.org/abs/1302.2972[arxiv:1302.2972v2 [ math - ph ] ] , 2013 , pp . 129 . athanassios s. fokas , alexander r. its , andrei a. kapaev , and victor yu . novokshenov , _ painlev transcendents _ , mathematical surveys and monographs , vol . 128 , american mathematical society , providence , ri , 2006 , the riemann - hilbert approach . mr 2264522 ( 2010e:33030 ) b. grammaticos , a. ramani , and y. ohta , _ a unified description of the asymmetric @xmath390 and @xmath391 equations and their schlesinger transformations _ , j. nonlinear math . phys . * 10 * ( 2003 ) , no . 2 , 215228 . mr 1976382 ( 2004c:34273 ) katsunori iwasaki , hironobu kimura , shun shimomura , and masaaki yoshida , _ from gauss to painlev _ , aspects of mathematics , e16 , friedr . vieweg & sohn , braunschweig , 1991 , a modern theory of special functions . mr mr1118604 ( 92j:33001 ) alexander r. its and victor yu . novokshenov , _ the isomonodromic deformation method in the theory of painlev equations _ , lecture notes in mathematics , vol . 1191 , springer - verlag , berlin , 1986 . mr mr851569 ( 89b:34012 ) michio jimbo and tetsuji miwa , _ monodromy preserving deformation of linear ordinary differential equations with rational coefficients . iii _ , phys . d * 4 * ( 1981/82 ) , no . 1 , 2646 . mr mr636469 ( 83k:34010c ) michio jimbo , tetsuji miwa , and kimio ueno , _ monodromy preserving deformation of linear ordinary differential equations with rational coefficients . i. general theory and @xmath392-function _ , phys . d * 2 * ( 1981 ) , no . 2 , 306352 . mr mr630674 ( 83k:34010a ) vladimir petrov kostov , _ the deligne - simpson problem for zero index of rigidity _ , perspectives of complex analysis , differential geometry and mathematical physics ( st . konstantin , 2000 ) , world sci . publ . , river edge , nj , 2001 , pp . mr 1877365 ( 2002j:14013 ) f. w. nijhoff and v. g. papageorgiou , _ similarity reductions of integrable lattices and discrete analogues of the painlev @xmath393 equation _ , phys . lett . a * 153 * ( 1991 ) , no . 6 - 7 , 337344 . mr 1098879 ( 92b:58103 ) paul painlev , _ oeuvres de paul painlev . tome i , ii , iii . _ , ditions du centre national de la recherche scientifique , paris , 1973 , preface by ren garnier , compiled by raymond grard , georges reeb and antoinette sec . mr mr0532682 ( 58 # 27154a ) to3em , _ problem : discrete painlev equations and their lax forms _ , algebraic , analytic and geometric aspects of complex differential equations and their deformations . painlev hierarchies , rims kkyroku bessatsu , b2 , res . inst . math . ( rims ) , kyoto , 2007 , pp . 195208 . mr 2310030 ( 2008c:33020 )

in a-weighted version of the robinson - schensted algorithm was introduced . in this paper we show that this algorithm has a symmetry property analogous to the well known symmetry property of the normal robinson - schensted algorithm . the proof uses a generalisation of the growth diagram approach introduced by fomin . this approach , which uses `` growth graphs '' , can also be applied to a wider class of insertion algorithms which have a branching structure , including some of the other-weighted versions of the robinson - schensted algorithm which have recently been introduced by borodin - petrov .

1306.2208

in @xcite a @xmath0-weighted version of the robinson - schensted algorithm was introduced . in this paper we show that this algorithm enjoys a symmetry property analoguous to the well - known symmetry property of the robinson - schensted algorithm . the proof uses a generalisation of the growth diagram approach introduced by @xcite . the insertion algorithm we consider in this paper is based on column insertion , but the technique applies to any insertion algorithm belonging to a certain class of `` branching insertion algorithms '' , as described in section [ s : other ] below . for example , @xcite have recently introduced a @xmath0-weighted version of the row insertion algorithm , which is defined similarly to the column insertion version of @xcite ; they also consider a wider family of such algorithms ( and , more generally , dynamics on gelfand - tsetlin patterns ) , some of which fall into the framework considered in the present paper , and can similarly be shown to have the symmetry property . we discuss such extensions in section [ s : other ] below . the robinson - schensted ( rs ) algorithm is a combinatorial algorithm which was introduced by robinson @xcite and schensted @xcite . it has wide applications in representation theory and probability theory , e.g. last passage percolation , totally asymmetric simple exclusion process / corner growth model , random matrix theory @xcite , queues in tandem @xcite and more . there are two versions of rs algorithms , the row insertion and column insertion version . in most of this paper we deal with column insertion and its @xmath0-version . the rs algorithm transforms a word to a tableau pair of the same shape . a word can be treated as a path , hence a random word corresponds to a random walk . when taking such a random walk , the shape of the output tableaux is a markov chain , whose transition kernel is related to the schur symmetric functions @xcite . a geometric generalisation transforms a brownian motion with drift to a markov process whose generator is related to @xmath1-whittaker functions @xcite , which are eigenfunctions of the quantum toda chain @xcite . the @xmath0-whittaker functions on the one hand are a generalisation of the schur symmetric functions and a specialisation of macdonald @xmath2-symmetric functions when @xmath3 @xcite , and on the other hand are eigenfunctions of the @xmath0-deformed quantum toda chain @xcite . when @xmath4 they become the schur functions and when @xmath5 with a proper scaling @xcite they converge to whittaker functions . in the spirit of this connection , a @xmath0-weighted robinson - schensted algorithm was formulated in @xcite , which transforms the random walk to a markov chain that is related to @xmath0-whittaker functions . as is expected the algorithm degenerates to the normal rs algorithm when @xmath4 . part of the random @xmath6-tableau also has @xmath0-tasep dynamics @xcite , the latter introduced in @xcite , just as the same part of the random @xmath6-tableau of normal rs algorithm has tasep dynamics @xcite . recently , a @xmath0-version of the rs row insertion algorithm was also introduced in @xcite . it denegerates to the normal rs algorithm with row insertion when @xmath4 . in this paper we show that both algorithms enjoy a symmetry property analogous to the well - known symmetry property of the robinson - schensted algorithm . basically , the symmetry property for the robinson - schensted algorithms restricted to permutation inputs is the property that the output tableau pair is interchanged if the permutation is inversed . knuth @xcite generalised the normal row insertion algorithm to one which takes matrix input , which we refer to as robinson - schensted - knuth ( rsk ) algorithm . for this algorithm the symmetry property is that the output tableau pair is interchanged if the matrix is transposed . note that the matrix becomes the permutation matrix when the rsk algorithm is restricted to permutation , hence the transposition of the matrix corresponds to inversion of the permutation . burge gives a similar generalisation of the column insertion algorithm @xcite , which we refer to as burge s algorithm . the symmetry property for the normal rs algorithm is normally discussed in the literature for the row insertion . however , in the permutation case the output tableau pair for row insertion is simply the transposition of the pair for column insertion . therefore proofs of the symmetry property can be translated to column insertion instantly . in the matrix input case , the burge and rsk algorithms are also closely related so that proofs can be extended from one to the other naturally . for a proof of the symmetry property one can see e.g. @xcite . there are a few different approaches to deal with it . one is the viennot diagram @xcite which provides a nice geometric construction of the rs algorithms . in the matrix input case there are two approaches which reduce to the viennot diagram approach when restricted to permutations . one is the antichain or the inversion digraph construction of the rsk algorithm ( which should extend naturally to the burge algorithm ) due to @xcite ; the other ( for both rsk and burge ) is fomin s matrix - ball construction of @xcite . another method is the growth diagram technique due to fomin @xcite . it can be generalised to the rsk and burge algorithms . theorem @xcite gives a proof by showing the relation between the lengths of the longest subsequences of the input and the shape of the output . however , the growth diagram approach can be thought of as a fast construction to calculate these lengths . among all these techniques , the special structure of growth diagram can be extended to a class of algorithms what we will call branching algorithms , which include the @xmath0-weighted column and row insertion algorithms . therefore it is this approach which we use in this paper . for a simple description of the technique for normal row insertion see e.g. @xcite , whose column version will be shown in section [ s : symnormal ] . the rest of the paper is organised in the following way . in section [ s : classical ] we recall the insertion rule of a letter into a tableau for the normal robinson - schensted algorithm ( with column insertion ) . we describe it in a way that suits the growth diagram . in section [ s : symnormal ] we describe the insertion rule for a word and state the symmetry property for the rs algorithm applied to permutations . in section [ s : qrs ] and [ s : wordqrs ] we describe the @xmath0-weighted insertion algorithm for letters and words in a way which is different , but equivalent to the definition given in @xcite . in section [ s : symq ] we state and prove the symmetry property in the @xmath0-case . finally in section [ s : other ] we prove the symmetry property for the row insertion algorithm and more generally branching algorithms . _ acknowledgements_. the author would like to thank neil oconnell for guidance . research of the author is supported by epsrc grant number ep / h023364/1 . a partition @xmath7 is a vector of weakly decreasing non - negative integer entries . denote by @xmath8 the number of positive entries of @xmath9 and @xmath10 the size of the partition . we say @xmath9 is a partition of @xmath11 if @xmath12 , which we denote by @xmath13 . a young tableau @xmath6 with shape @xmath9 is a left aligned array of @xmath8 rows of positive integers such that the entries are strictly increasing along each column and weakly increasing along each row , and such that the length of the @xmath14th row is @xmath15 . for example below is a tableau with shape @xmath16 . @xmath17 we denote by @xmath18 the shape of tableau @xmath6 , and @xmath19 the set of all tableaux with entries no greater than @xmath20 . denote @xmath21=\{1,2,\cdots , n\}$ ] for a positive integer @xmath11 . a standard tableau @xmath22 is a tableau with distinct entries from @xmath23 $ ] . for example below is a standard tableau with shape @xmath16 . @xmath24 we denote by @xmath25 the set of standard tableaux with shape of size @xmath11 . for example the above tableau is an element of @xmath26 . for a tableau @xmath6 , we denote by @xmath27 its subtableau containing all entries no greater than @xmath28 and call it the @xmath28th subtableau of @xmath6 . for example the @xmath29th subtableau of the tableau shown in is @xmath30 a tableau @xmath6 can be identified by the shape of its subtableaux , which we usually denote by @xmath31 . also let @xmath32 to be the empty partition . evidently a tableau has only finitely many different @xmath33 s and there exists an @xmath20 such that @xmath34 for @xmath35 . we call @xmath33 the @xmath36th shape of @xmath6 . we can identify @xmath6 with these shapes and write @xmath37 , where `` @xmath38 '' is an interlacing relation : for @xmath39 and @xmath40 , @xmath41 means @xmath42 . for example the tableau @xmath6 in is identified as @xmath43 the basic operation of the robinson - schensted algorithm is to insert a letter @xmath44 into a tableau @xmath45 and produce a new tableau @xmath46 . to do this we first find the lowest row in @xmath47 such that appending a box at the end of that row would preserve the interlacement between the @xmath48th shape and the @xmath28th shape , and append the box to that row . suppose the row has index @xmath49 , then we find the lowest row in @xmath50 that is no lower than row @xmath49 such that appending a box at the end of that row would preserve the interlacement between the @xmath28th shape and the @xmath51th shape , and append the box to that row , and so on and so forth . more precisely , define @xmath52 then the new tableau @xmath53 is defined by @xmath54 where @xmath55 is the @xmath14th standard basis of @xmath56 . for example , if we insert a 6 into the tableau shown in , the insertion process is shown as follows : @xmath57 where each @xmath58 denotes a box in @xmath33 , and each entry denotes an appended box . the resultant tableau is thus @xmath59 one can visualise the insertion process by building the shapes up vertically . = [ fill = white ] ( 00 ) at ( 0,0 ) @xmath60 ; ( 10 ) at ( 3,0 ) @xmath60 ; iin 1, ... ,3 ( 0i ) at ( 0,i ) @xmath61 ; ( 1i ) at ( 3,i ) @xmath62 ; ( 04 ) at ( 0,4 ) @xmath63 ; ( 14 ) at ( 3,4 ) @xmath64 ; at ( 1.5,3.5 ) @xmath65 ; iin 0, ... ,4 ( 0i ) ( 1i ) ; i[count= ] in 0, ... ,2 ( 0i ) ( 0 ) ; ( 1i ) ( 1 ) ; ( 03 ) ( 04 ) ; ( 13 ) ( 14 ) ; ( 0,4.5 ) node[above]inserting @xmath28 ( 3,4.5 ) ; ( -.5,0 ) node[left]tableau @xmath6 ( -.5,4 ) ; ( 3.5,0 ) node[right]tableau @xmath53 ( 3.5,4 ) ; as we can see , this forms a one - column lattice diagram such that for each @xmath66 , the vertices labeled with @xmath67 and @xmath68 surround a box , which we call the @xmath36th box . we can put an @xmath69 into the @xmath28th box to indicate that the number inserted into @xmath6 is @xmath28 . the corresponding diagram of the previous example where we insert a 6 into tableau is : = [ fill = white ] ( 00 ) at ( 0,0 ) @xmath60 ; ( 01 ) at ( 0,1 ) @xmath70 ; ( 02 ) at ( 0,2 ) @xmath70 ; ( 03 ) at ( 0,3 ) @xmath71 ; ( 04 ) at ( 0,4 ) @xmath72 ; ( 05 ) at ( 0,5 ) @xmath73 ; ( 06 ) at ( 0,6 ) @xmath74 ; ( 07 ) at ( 0,7 ) @xmath75 ; ( 08 ) at ( 0,8 ) @xmath76 ; \(10 ) at ( 1.1,0 ) @xmath60 ; ( 11 ) at ( 1.1,1 ) @xmath70 ; ( 12 ) at ( 1.1,2 ) @xmath70 ; ( 13 ) at ( 1.1,3 ) @xmath71 ; ( 14 ) at ( 1.1,4 ) @xmath72 ; ( 15 ) at ( 1.1,5 ) @xmath73 ; ( 16 ) at ( 1.1,6 ) @xmath75 ; ( 17 ) at ( 1.1,7 ) @xmath77 ; ( 18 ) at ( 1.1,8 ) @xmath78 ; at ( .5,5.55 ) @xmath69 ; iin 0, ... ,8 ( 0i ) ( 1i ) ; i[count= ] in 0, ... ,7 ( 0i ) ( 0 ) ; ( 1i ) ( 1 ) ; for a word @xmath79^n$ ] , the robinson - schensted algorithm starts with the empty tableau @xmath80 . then @xmath81 is inserted into @xmath82 to obtain @xmath83 , then @xmath84 into @xmath83 to obtain @xmath85 and so on . the recording tableau @xmath22 is a standard tableau defined by : @xmath86 for example the following table shows the the process of inserting the word @xmath87 : [ cols="^,^,^,^,^,^",options="header " , ] ; this way , all the vertices of the diagram can be labelled recursively given @xmath60 as the boundary condition on the leftmost and bottom vertices . one could check that this is indeed equivalent to the definition in the previous section . moreover , the transposition of the lattice diagram preserves the algorithm . that is : if we put @xmath69 s in boxes @xmath88 rather than @xmath89 , and label each vertex @xmath90 with what was labelled on @xmath91 , we end up with the configuration that is the same as if we apply the rules 1 through 5 . this immediately finishes the proof . in a recent paper @xcite a @xmath0-weighted robinson - schensted algorithm was introduced . in this and the next section we describe the algorithm in a different way from the definition in @xcite . at the end of next section it is obvious to see that : the algorithm described in this section for inserting a letter to a tableau and in the next section for inserting a word to an empty tableau is an equivalent reformulation of the @xmath0-weighted robinson - schensted algorithm defined in @xcite . when inserting a letter to a tableau it outputs a weigted set of tableaux . to insert a @xmath28 into @xmath6 , we start with @xmath47 , append a box to different possible rows no lower than @xmath28 , record the index of rows , and for each one of these indices @xmath49 , we obtain a new @xmath28th shape @xmath92 with weight @xmath93 ; then we proceed to add a box to all possible rows in @xmath50 no lower than @xmath49 and obtain a weighted set of new @xmath51th shapes @xmath94 ; then for each @xmath95 we obtain the new @xmath96th shapes with weight @xmath97 by adding a box to @xmath98th row in @xmath99 and so on and so forth . we also prune all the 0-weighted branches . this way we obtain a forest of trees whose roots are @xmath100 s , leaves are @xmath101 s with edges labeled by the weights . if we prepend @xmath102 to this forest we obtain a tree with root @xmath60 , the first @xmath28 levels each having one branch with one edge , which we label with weight 1 . then for each leaf @xmath101 we obtain a unique tableau by reading its genealogy . we also associate this tableau with a weight which is the product of weights along the edges . by going through all leaves we obtain a set of weighted tableaux , which is the output of @xmath0-inserting a @xmath28 to @xmath6 . now we describe how we calculate the weights when inserting a box to @xmath36th shape @xmath33 with a @xmath103 ( let @xmath104 ) specified . first let us define functions @xmath105 and @xmath106 associated with a pair of nested partitions @xmath107 : @xmath108 when adding a box to the first affected partition @xmath47 , the weight @xmath93 associated with each @xmath109 is @xmath110 for @xmath111 , and for a fixed pair of @xmath112 , when adding a box to row @xmath113 of @xmath114 , the corresponding weight @xmath115 is @xmath116 for example , below is a path of the tree , i.e. genealogy of a possible output tableau when inserting a @xmath117 into tableau : @xmath118 where again each @xmath119 denotes a box in @xmath120 and the @xmath121 denotes the new box @xmath122 . we also have @xmath123 , @xmath124 in this example . the above output tableau is @xmath125 with weight @xmath126 . for any two tableaux @xmath6 and @xmath53 we denote the weight of obtaining @xmath53 after inserting @xmath28 into @xmath6 by @xmath127 . so in the previous example we have @xmath128 . as we can see , each branching to obtain a weighted set of new @xmath36th shapes @xmath129 is determined by @xmath33 and @xmath103 , the latter in turn is identified by the pair @xmath130 . therefore the branching is determined by the triplet @xmath131 . while each @xmath132 , together with @xmath33 and @xmath133 determines a next branching . hence the tree has the structure illustrated in figure [ fig : branch ] , ( see section [ s : other ] , where the @xmath0-weighted algorithm is an example ) . ] where we put the weights on the edges . this way we can write the weights : @xmath134 where @xmath135 , @xmath136 for @xmath137 , and @xmath138 is the indicator function . therefore , the @xmath0-insertion algorithm can be visualised in a `` one - column '' graph similar to the normal insertion . each node @xmath139 ( or @xmath140 ) representing the old @xmath36th shape ( or a new @xmath14th shape ) is associated with a vertex @xmath141 ( or @xmath142 ) as the northwest ( or northeast ) vertex of the @xmath36th ( or @xmath14th ) box . each triplet @xmath143 is represented as a connected triple nodes surrounding the @xmath36th box from the south and the west , for which each branch @xmath144 is represented as a node that connects to both @xmath33 and @xmath145 , where we put the weights on both the horizontal and vertical edges . @xmath146 as is in the normal insertion case , we put an @xmath69 in the @xmath28th box to indicate we are inserting a @xmath28 . figure [ fig : insertletter ] is such a visualisation corresponding to the tree in figure [ fig : branch ] . 6 = [ draw , rectangle ] ( 00 ) at ( 0,0 ) @xmath60 ; ( 10 ) at ( , 0 ) @xmath60 ; ( 01 ) at ( 0 , ) @xmath147 ; ( 11 ) at ( , ) @xmath148 ; ( 02 ) at ( 0,2 * ) @xmath149 ; ( 12 ) at ( , 2 * ) @xmath150 ; ( 03 ) at ( 0,3 * ) @xmath151 ; ( 13 ) at ( , 3 * ) @xmath152 ; ( 04 ) at ( 0 , ) @xmath153 ; ( 143 ) at ( c-,- ) @xmath154 ; ( 142 ) at ( , ) @xmath155 ; ( 141 ) at ( + , + ) @xmath156 ; ( 05 ) at ( 0 , ) @xmath50 ; ( 156 ) at ( c-2*,-2 * ) @xmath157 ; ( 155 ) at ( c-1*,-1 * ) @xmath158 ; ( 154 ) at ( , ) @xmath159 ; ( 153 ) at ( + , + ) @xmath160 ; ( 152 ) at ( + 2*,+2 * ) @xmath161 ; ( 151 ) at ( + 3*,+3 * ) @xmath162 ; = [ sloped , fill = white , draw = none , font= ] iin 0, ... ,3 ( 0i ) node@xmath163 ( 1i ) ; i[count= ] in 0, ... ,4 i[count= ] in 0, ... ,1 ( 1i ) node@xmath163 ( 1 ) ; ( 02 ) ( 03 ) ; ( 12 ) ( 13 ) ; ( 13 ) node@xmath164 ( 141 ) ; ( 13 ) node@xmath165 ( 142 ) ; ( 13 ) node@xmath166 ( 143 ) ; ( 141 ) node@xmath167 ( 151 ) ; ( 141 ) node@xmath168 ( 152 ) ; ( 141 ) node@xmath169 ( 153 ) ; ( 142 ) node@xmath170 ( 154 ) ; ( 143 ) node@xmath171 ( 155 ) ; ( 143 ) node@xmath172 ( 156 ) ; ( 04 ) node@xmath164 ( 141 ) ; ( 04 ) node@xmath165 ( 142 ) ; ( 04 ) node@xmath166 ( 143 ) ; ( 05 ) node@xmath167 ( 151 ) ; ( 05 ) node@xmath168 ( 152 ) ; ( 05 ) node@xmath169 ( 153 ) ; ( 05 ) node@xmath170 ( 154 ) ; ( 05 ) node@xmath171 ( 155 ) ; ( 05 ) node@xmath172 ( 156 ) ; at ( 3,6.5 ) [ font=]@xmath69 ; all the nodes associated with the right column @xmath173 form a tree , such that each @xmath53 in the output corresponds to a genealogy of a node associated with @xmath174 , whose weight @xmath127 can be obtained as the product of the weights along the genealogical line of the tree . we can also take a word as input for the @xmath0-weighted robinson - schensted algorithm and produce a set of weighted pairs of tableaux . we start with the set of only one pair of tableaux @xmath175 with weight 1 . suppose at time @xmath176 , we have obtained a set of weighted pairs of tableaux @xmath177 we use brackets in superscripts to indicate the different possibilities , in order not to confuse with subtableaux or shapes . we take each triplet + @xmath178 in the collection , insert @xmath179 into @xmath180 to produce a new set of @xmath6-tableaux paired with the weight generated during the insertion @xmath181 then for each @xmath182 , @xmath183 , we pair it with a @xmath184 by adding a box with entry @xmath185 to @xmath186 such that @xmath187 . we also attach a weight @xmath188 to this pair . this way , for each triplet @xmath178 we have obtained its branching set @xmath189 let @xmath36 run over @xmath190 , we obtained a collection of all possible branchings : @xmath191 note by _ collection _ we mean a vector of objects that allows repeats , as opposed to a _ set_. we merge triplets with the same tableau pair by adding up their weights . this way we have obtained the weighted set of pairs of tableaux at time @xmath185 . the output of inserting the word @xmath192 is defined as the weighted set at time @xmath11 . we denote by @xmath193 the weight of pair @xmath194 in this set . this is defined recursively by the following formula ( see @xcite ) . for @xmath6 and @xmath22 with the same shape of size @xmath11 and word @xmath195 of length @xmath196 , @xmath197 where the sum is over all tableau @xmath198 with the same shape as @xmath199 . this , like in the classical case , can be visualised as a graph whose nodes are associated with vertices on @xmath200 lattice diagram by concatenating the graphs of the one - column construction in figure [ fig : insertletter ] associated with the insertion of each letter . each possible @xmath201 is associated with the vertex @xmath202 . for example , we obtain the following graph if we apply the algorithm to @xmath203 ^ 4 $ ] , where the unlabeled edges have weight 1 , which will be the case hereafter . = [ draw ] iin 0, ... ,4 ( i0 ) at ( , ) @xmath60 ; iin 0, ... ,3 ( 0i ) at ( , ) @xmath60 ; ( 11 ) at ( , ) @xmath60 ; iin 2, ... ,4 ( i1 ) at ( , ) @xmath163 ; iin 2, ... ,3 ( 1i ) at ( , ) @xmath163 ; ( 2i ) at ( , ) @xmath70 ; ( 32 ) at ( , ) @xmath70 ; ( 331 ) at ( , ) @xmath204 ; ( 332 ) at ( -,- ) @xmath205 ; ( 421 ) at ( , ) @xmath204 ; ( 422 ) at ( -,- ) @xmath205 ; ( 431 ) at ( + , + ) @xmath206 ; ( 432 ) at ( -0.3*,-0.3 * ) @xmath71 ; ( 433 ) at ( -,- ) @xmath71 ; ( 434 ) at ( -2*,-2 * ) @xmath71 ; ( 435 ) at ( -3*,-3 * ) @xmath207 ; = [ ] at ( /2+/2,/2+/2 ) @xmath69 ; at ( /2+/2,/2+/2 ) @xmath69 ; at ( /2+/2,/2+/2 ) @xmath69 ; at ( /2+/2,/2+/2 ) @xmath69 ; = [ fill = white , sloped , font= ] i[count= ] in 0, ... ,3 ( i0 ) ( 0 ) ; ( i1 ) ( 1 ) ; i[count= ] in 0, ... ,2 ( 0i ) ( 0 ) ; ( 1i ) ( 1 ) ; ( 2i ) ( 2 ) ; i[count= ] in 0,1 ( i2 ) ( 2 ) ; ( i3 ) ( 3 ) ; ( 30 ) ( 31 ) ; ( 40 ) ( 41 ) ; ( 31 ) ( 32 ) ; ( 22 ) ( 32 ) ; ( 32 ) node@xmath208 ( 331 ) ; ( 32 ) node@xmath209 ( 332 ) ; ( 23 ) node@xmath208 ( 331 ) ; ( 23 ) node@xmath209 ( 332 ) ; ( 41 ) node@xmath210 ( 421 ) ; ( 41 ) node@xmath0 ( 422 ) ; ( 32 ) node@xmath210 ( 421 ) ; ( 32 ) node@xmath0 ( 422 ) ; ( 421 ) node@xmath211 ( 431 ) ; ( 421 ) node@xmath212 ( 432 ) ; ( 331 ) node@xmath211 ( 431 ) ; ( 331 ) node@xmath212 ( 432 ) ; ( 421 ) ( 434 ) ; ( 332 ) ( 434 ) ; ( 422 ) ( 433 ) ; ( 331 ) ( 433 ) ; ( 422 ) ( 435 ) ; ( 332 ) ( 435 ) ; from the definition of the algorithm , each node @xmath213 at @xmath214 corresponds to a @xmath194 tableau pair in the output set . indeed we can trace back the genealogy of @xmath213 and find a unique array of indices @xmath215 with @xmath216 such that nodes @xmath217 are connected and form a diagram that is isomorphic to the lattice diagram @xmath218 . clearly , @xmath219 are the corresponding @xmath194 tableau pair . moreover , we can identify a weight with the node @xmath220 by multiplying the weights of all the horizontal ( or all the vertical ) edges along its genealogical diagram and denote it by @xmath221 : @xmath222 then by definition , @xmath223 where the sum is over all nodes @xmath224 at vertex @xmath225 . when we take a permutation input @xmath226 which is identified as a word with distinct letters in the same way as in the normal rs algorithm , we also end up with a symmetry property , which is the main result of this paper : [ thm : main ] for any permutation @xmath227 and standard tableau pair @xmath194 , @xmath228 as in the normal rs algorithm case , permutation input are special in that we can restate the insertion rule in a symmetric fashion . for each box we pick a connected triplet on southwest , northwest and southeast corners . that is , suppose the box has index @xmath202 we fix arbitrary @xmath229 , @xmath230 and @xmath231 such that @xmath232 is a connected triplet . we denote by @xmath233 the set of partitions on the vertex @xmath202 that are connected to @xmath234 and @xmath235 . and for any @xmath236 , we write @xmath237 instead of @xmath238 or @xmath239 since they are equal , and for the sake of symmetry . define an operator @xmath240 by @xmath241 and a set @xmath242 by @xmath243 and denote @xmath244 . then @xmath233 and @xmath245 belongs to one of the following 5 cases . 1 . the box has an @xmath69 in it , and @xmath234 , @xmath235 are equal to @xmath9 . then @xmath233 consists of all possible partitions obtained by adding a box to @xmath9 : @xmath246 the weights are @xmath247 2 . the box does not have an @xmath69 in it and @xmath248 . then @xmath233 is a singleton which is the same as @xmath235 , and the weight is 1 : @xmath249 3 . ( the dual case of case 2)the box does not have an @xmath69 in it and @xmath250 . then @xmath233 is a singleton which is the same as @xmath234 , and the weight is 1 : @xmath251 4 . the box is empty . @xmath252 and @xmath253 for some @xmath254 . then @xmath233 again only contains one element @xmath255 , with weight 1 : @xmath256 5 . the box is empty and @xmath257 for some @xmath36 . then @xmath233 consists of all possible partitions that are obtained by adding a box to a row no lower than @xmath36th row of @xmath258 . that is @xmath259 the weights are @xmath260 we call this the _ growth graph rule _ as opposed to the _ insertion rule _ described in sections [ s : qrs ] and [ s : wordqrs ] . below is a group of illustrations of all 5 cases , where @xmath9 is the southwest partition . \(100 ) at ( 0,0 ) ( 0,0 ) ( 4,0 ) ( 4,-1 ) ( 2,-1 ) ( 2,-2 ) ( 0,-2 ) ( 0,0 ) ; ; ( 110 ) at ( 1,0 ) ( 0,0 ) ( 4,0 ) ( 4,-1 ) ( 2,-1 ) ( 2,-2 ) ( 0,-2 ) ( 0,0 ) ; ; ( 101 ) at ( 0,1 ) ( 0,0 ) ( 4,0 ) ( 4,-1 ) ( 2,-1 ) ( 2,-2 ) ( 0,-2 ) ( 0,0 ) ; ; ( 1111 ) at ( 1+,1 + ) ( 0,0 ) ( 4,0 ) ( 4,-1 ) ( 2,-1 ) ( 2,-2 ) ( 0,-2 ) ( 0,0 ) ; ( 1,-2 ) ( 1,-3 ) ( 0,-3 ) ( 0,-2 ) ; ; ( 1112 ) at ( 1,1 ) ( 0,0 ) ( 4,0 ) ( 4,-1 ) ( 2,-1 ) ( 2,-2 ) ( 0,-2 ) ( 0,0 ) ; ( 3,-1 ) ( 3,-2 ) ( 2,-2 ) ( 2,-1 ) ; ; ( 1113 ) at ( 1-,1- ) ( 0,0 ) ( 4,0 ) ( 4,-1 ) ( 2,-1 ) ( 2,-2 ) ( 0,-2 ) ( 0,0 ) ; ( 5,0 ) ( 5,-1 ) ( 4,-1 ) ( 4,0 ) ( 5,0 ) ; ; = [ sloped , fill = white ] ( 100 ) node@xmath261 ( 101 ) node@xmath262 ( 1111 ) node@xmath262 ( 110 ) node@xmath263 ( 100 ) ; ( 101 ) node@xmath264 ( 1112 ) node@xmath264 ( 110 ) ; ( 101 ) node@xmath265 ( 1113 ) node@xmath265 ( 110 ) ; at ( .5,.5 ) @xmath69 ; at ( .5,-.1 ) case 1 ; \(200 ) at ( 0,0 ) ( 0,0 ) ( 5,0 ) ( 0,-4 ) ( 0,0 ) ; ; ( 210 ) at ( 1,0 ) ( 0,0 ) ( 5,0 ) ( 5,-1 ) ( 3,-1 ) ( 3,-3 ) ( 2,-3 ) ( 2,-4 ) ( 0,-4 ) ( 0,0 ) ; ( 3,-3 ) ( 3,-4 ) ( 2,-4 ) ; ; ( 201 ) at ( 0,1 ) ( 0,0 ) ( 5,0 ) ( 5,-1 ) ( 3,-1 ) ( 3,-3 ) ( 2,-3 ) ( 2,-4 ) ( 0,-4 ) ( 0,0 ) ; ; ( 211 ) at ( 1,1 ) ( 0,0 ) ( 5,0 ) ( 5,-1 ) ( 3,-1 ) ( 3,-3 ) ( 2,-3 ) ( 2,-4 ) ( 0,-4 ) ( 0,0 ) ; ( 3,-3 ) ( 3,-4 ) ( 2,-4 ) ; ; = [ sloped , fill = white , font= ] ( 200 ) node@xmath261 ( 201 ) ( 211 ) ( 210 ) node@xmath263 ( 200 ) ; at ( .5,-.4 ) case 2 ; \(100 ) at ( 0,0 ) ( 0,0 ) ( 5,0 ) ( 5,-1 ) ( 0,-4 ) ( 0,0 ) ; ; ( 110 ) at ( 1,0 ) ( 0,0 ) ( 5,0 ) ( 5,-1 ) ( 3,-1 ) ( 3,-3 ) ( 2,-3 ) ( 2,-4 ) ( 0,-4 ) ( 0,0 ) ; ; ( 101 ) at ( 0,1 ) ( 0,0 ) ( 5,0 ) ( 5,-1 ) ( 3,-1 ) ( 3,-3 ) ( 2,-3 ) ( 2,-4 ) ( 0,-4 ) ( 0,0 ) ; ( 5,0 ) ( 6,0 ) ( 6,-1 ) ( 5,-1 ) ; ; ( 111 ) at ( 1,1 ) ( 0,0 ) ( 5,0 ) ( 5,-1 ) ( 3,-1 ) ( 3,-3 ) ( 2,-3 ) ( 2,-4 ) ( 0,-4 ) ( 0,0 ) ; ( 5,0 ) ( 6,0 ) ( 6,-1 ) ( 5,-1 ) ; ; = [ sloped , fill = white , font= ] ( 100 ) node@xmath261 ( 101 ) ( 111 ) ( 110 ) node@xmath263 ( 100 ) ; at ( .5,-.5/3 ) case 3 ; \(200 ) at ( 0,0 ) ( 0,0 ) ( 5,0 ) ( 5,-1 ) ( 3,-1 ) ( 3,-3 ) ( 2,-3 ) ( 2,-4 ) ( 0,-4 ) ( 0,0 ) ; ; ( 210 ) at ( 1,0 ) ( 0,0 ) ( 5,0 ) ( 5,-1 ) ( 3,-1 ) ( 3,-3 ) ( 2,-3 ) ( 2,-4 ) ( 0,-4 ) ( 0,0 ) ; ( 1,-4 ) ( 1,-5 ) ( 0,-5 ) ( 0,-4 ) ; ; ( 201 ) at ( 0,1 ) ( 0,0 ) ( 5,0 ) ( 5,-1 ) ( 3,-1 ) ( 3,-3 ) ( 2,-3 ) ( 2,-4 ) ( 0,-4 ) ( 0,0 ) ; ( 4,-1 ) ( 4,-2 ) ( 3,-2 ) ; ; ( 211 ) at ( 1,1 ) ( 0,0 ) ( 5,0 ) ( 5,-1 ) ( 3,-1 ) ( 3,-3 ) ( 2,-3 ) ( 2,-4 ) ( 0,-4 ) ( 0,0 ) ; ( 1,-4 ) ( 1,-5 ) ( 0,-5 ) ( 0,-4 ) ; ( 4,-1 ) ( 4,-2 ) ( 3,-2 ) ; ; = [ sloped , fill = white , font= ] ( 200 ) node@xmath261 ( 201 ) ( 211 ) ( 210 ) node@xmath263 ( 200 ) ; at ( .5,-.5/3 ) case 4 ; \(200 ) at ( 0,0 ) ( 0,0 ) ( 5,0 ) ( 5,-1 ) ( 3,-1 ) ( 3,-2 ) ( 1,-2 ) ( 1,-3 ) ( 0,-3 ) ( 0,0 ) ; ; ( 210 ) at ( 1,0 ) ( 0,0 ) ( 5,0 ) ( 5,-1 ) ( 3,-1 ) ( 1,-3 ) ( 0,-3 ) ( 0,0 ) ; ( 2,-2 ) ( 1,-2 ) ; ; ( 201 ) at ( 0,1 ) ( 0,0 ) ( 5,0 ) ( 5,-1 ) ( 3,-1 ) ( 3,-2 ) ( 1,-2 ) ( 1,-3 ) ( 0,-3 ) ( 0,0 ) ; ( 2,-2 ) ( 1,-2 ) ; ; ( 2111 ) at ( 1+,1 + ) ( 0,0 ) ( 1,-3 ) ( 0,-3 ) ( 0,0 ) ; ( 2,-2 ) ( 1,-2 ) ; ( 3,-2 ) ( 2,-2 ) ; ; ( 2112 ) at ( 1,1 ) ( 0,0 ) ( 0,-3 ) ( 0,0 ) ; ( 2,-2 ) ( 1,-3 ) ( 1,-2 ) ; ( 4,-1 ) ( 4,-2 ) ( 3,-1 ) ; ; ( 2113 ) at ( 1-+.05,1 - -.05 ) ( 0,0 ) ( 5,0 ) ( 5,-1 ) ( 3,-1 ) ( 3,-2 ) ( 1,-2 ) ( 1,-3 ) ( 0,-3 ) ( 0,0 ) ; ( 2,-2 ) ( 2,-3 ) ( 1,-3 ) ( 1,-2 ) ; ( 5,0 ) ( 6,0 ) ( 6,-1 ) ( 5,-1 ) ; ; at ( .5,-.125 ) case 5 ; = [ sloped , fill = white , font= ] ( 200 ) node@xmath261 ( 201 ) node@xmath266 ( 2111 ) node@xmath266 ( 210 ) node@xmath263 ( 200 ) ; ( 201 ) node@xmath267 ( 2112 ) node@xmath267 ( 210 ) ; ( 201 ) node@xmath268 ( 2113 ) node@xmath268 ( 210 ) ; it now suffices to show that the above constuction agrees with the definition of the algorithm in the preceeding section . that is , for any connected triplets @xmath269 surrounding the box @xmath202 from the southwest , its branching set with corresponding weights according to the insertion rule agree with @xmath233 and @xmath237 s according to the growth graph rule . to distinguish the context of insertion rule and growth graph rule we also write @xmath270 , @xmath271 , @xmath272 and @xmath273 in accordance with the definition of the insertion algorithm in section [ s : qrs ] . we show this by discussing the location of @xmath202 in the permutation matrix of @xmath226 , which corresponds to the 5 cases in the growth graph rule . 1 . the box @xmath202 has an @xmath69 in it . this means @xmath274 . so we are inserting a @xmath28 to the tableau at time @xmath176 , hence @xmath275 , i.e. @xmath276 . moreover since @xmath226 is a permutation , we have @xmath277 for all @xmath278 , hence @xmath279 , i.e. @xmath280 ( condition of case 1 satisfied ) . any branch of @xmath281 is one of the @xmath282 s such that the weight @xmath283 . @xmath284 so the weights agree . all the pruned branches @xmath282 with @xmath285 are exactly the @xmath282 s that are not partitions . therefore @xmath233 is exactly the set of all branches of the triplet . there s no @xmath69 in @xmath286 for @xmath287 . this means @xmath277 for @xmath288 . since @xmath277 for @xmath278 , @xmath279 , i.e. @xmath280 ( condition of case 2 satisfied ) . moreover since @xmath289 , the triplet produces only one branch which is equal to @xmath235 with weight 1 . this agrees with case 2 . there s no @xmath69 in @xmath290 for @xmath291 . this means @xmath292 . by the insertion rule , @xmath275 , i.e. @xmath276 ( condition of case 3 satisfied ) . again by the same rule the triplet only produces one branch that equals @xmath234 with weight 1 . there s one @xmath69 in each of @xmath293 and @xmath294 for some @xmath295 and @xmath296 . then on the one hand @xmath297 so @xmath298 for some @xmath36 , i.e. @xmath252 . on the other hand @xmath299 so @xmath300 for some @xmath14 and @xmath301 , i.e. @xmath253 ( condition of case 4 or case 5 satisfied ) . therefore the branches of the triplet @xmath302 are in the form of @xmath92 for @xmath303 with weights @xmath304 if @xmath305 ( condition of case 4 satisfied ) , then @xmath306 . therefore the branch has only one shape equal to @xmath307 with weight 1 . this agrees with case 4 . + if @xmath308 ( condition of case 5 satisfied ) , then by @xmath309 so the weights agree with case 5 . moreover , all the pruned branches are exactly the @xmath310 s that are not partitions . therefore @xmath233 is indeed the set of all branches . when inverting a permutation @xmath226 to @xmath311 , the @xmath69 marks are transposed , so is the weighted graph by the symmetry of the rule , thus we have arrived at the conclusion . note that in both rules although we consider an arbitrary box with index @xmath202 , the rules do not depend on either @xmath176 or @xmath28 . this is a key condition for the symmetry property to work and will be generalised in proposition [ p : branch ] in the following section . below is the growth graph of the permutation @xmath312 : 0,1,2.6,5.6,11.6 .3 ( s1 ) at ( .5,.5 ) @xmath69 ; ( s2 ) at ( 1.8,8.6 ) @xmath69 ; ( s3 ) at ( 4.1,1.8 ) @xmath69 ; ( s4 ) at ( 8.6,4.1 ) @xmath69 ; = [ font=,draw ] ; iin 0, ... ,4 ( i0 ) at ( [i],[0 ] ) @xmath60 ; ( 0i ) at ( [0],[i ] ) @xmath60 ; iin 1, ... ,4 ( i1 ) at ( [i],[1 ] ) @xmath163 ; ( 1i ) at ( [1],[i ] ) @xmath163 ; ( 22 ) at ( [2],[2 ] ) @xmath163 ; ( 23 ) at ( [2],[3 ] ) @xmath163 ; \(321 ) at ( [3 ] - c , [2 ] - c ) @xmath70 ; ( 322 ) at ( [3 ] + c , [2 ] + c ) @xmath313 ; \(331 ) at ( [3 ] - c , [3 ] - c ) @xmath70 ; ( 332 ) at ( [3 ] + c , [3 ] + c ) @xmath313 ; \(241 ) at ( [2 ] - c , [4 ] - c ) @xmath70 ; ( 242 ) at ( [2 ] + c , [4 ] + c ) @xmath313 ; \(421 ) at ( [4 ] - c , [2 ] - c ) @xmath70 ; ( 422 ) at ( [4 ] + c , [2 ] + c ) @xmath313 ; \(341 ) at ( [3 ] - c 3,[4 ] - c 3 ) @xmath205 ; ( 342 ) at ( [3 ] - c , [4 ] - c ) @xmath204 ; ( 343 ) at ( [3 ] + c , [4 ] + c ) @xmath204 ; ( 344 ) at ( [3 ] + c 3,[4 ] + c 3 ) @xmath204 ; \(431 ) at ( [4 ] - c 3.3,[3 ] - c 3.3 ) @xmath205 ; ( 432 ) at ( [4 ] - c , [3 ] - c ) @xmath204 ; ( 433 ) at ( [4 ] + c , [3 ] + c ) @xmath204 ; ( 434 ) at ( [4 ] + c 3,[3 ] + c 3 ) @xmath314 ; \(441 ) at ( [4 ] - c 11,[4 ] - c 11 ) @xmath207 ; ( 442 ) at ( [4 ] - c 9,[4 ] - c 9 ) @xmath71 ; ( 443 ) at ( [4 ] - c 7,[4 ] - c 7 ) @xmath71 ; ( 444 ) at ( [4 ] - c 5,[4 ] - c 5 ) @xmath315 ; ( 445 ) at ( [4 ] - c 3,[4 ] - c 3 ) @xmath71 ; ( 446 ) at ( [4 ] - c , [4 ] - c ) @xmath71 ; ( 447 ) at ( [4 ] + c , [4 ] + c ) @xmath206 ; ( 448 ) at ( [4 ] + c 3,[4 ] + c 3 ) @xmath71 ; ( 449 ) at ( [4 ] + c 5,[4 ] + c 5 ) @xmath315 ; = [ font=,fill = white , sloped ] i[count= ] in 0, ... ,3 ( i0 ) ( 0 ) ; ( 0i ) ( 0 ) ; ( i1 ) ( 1 ) ; ( 1i ) ( 1 ) ; iin 2, ... ,4 ( i0 ) ( i1 ) ; ( 0i ) ( 1i ) ; \(21 ) ( 22 ) ; ( 22 ) ( 23 ) ; \(12 ) ( 22 ) ; ( 13 ) ( 23 ) ; \(23 ) node@xmath0 ( 241 ) ; ( 23 ) node@xmath210 ( 242 ) ; ( 14 ) node@xmath0 ( 241 ) ; ( 14 ) node@xmath210 ( 242 ) ; \(31 ) node@xmath0 ( 321 ) ; ( 31 ) node@xmath210 ( 322 ) ; ( 22 ) node@xmath0 ( 321 ) ; ( 22 ) node@xmath210 ( 322 ) ; \(321 ) ( 331 ) ; ( 322 ) ( 332 ) ; ( 23 ) ( 331 ) ; ( 23 ) ( 332 ) ; \(321 ) ( 421 ) ; ( 322 ) ( 422 ) ; ( 41 ) ( 421 ) ; ( 41 ) ( 422 ) ; \(331 ) ( 341 ) ; ( 332 ) ( 342 ) ; ( 331 ) ( 343 ) ; ( 332 ) ( 344 ) ; ( 241 ) ( 341 ) ; ( 241 ) ( 342 ) ; ( 242 ) ( 343 ) ; ( 242 ) ( 344 ) ; \(421 ) node@xmath209 ( 431 ) ; ( 421 ) node@xmath208 ( 432 ) ; ( 422 ) node@xmath0 ( 433 ) ; ( 422 ) node@xmath210 ( 434 ) ; ( 331 ) node@xmath209 ( 431 ) ; ( 331 ) node@xmath208 ( 432 ) ; ( 332 ) node@xmath0 ( 433 ) ; ( 332 ) node@xmath210 ( 434 ) ; \(341 ) ( 441 ) ; ( 341 ) ( 442 ) ; ( 342 ) ( 443 ) ; ( 342 ) ( 444 ) ; ( 343 ) ( 445 ) ; ( 343 ) node@xmath316 ( 446 ) ; ( 343 ) node@xmath317 ( 447 ) ; ( 344 ) ( 448 ) ; ( 344 ) ( 449 ) ; \(431 ) ( 441 ) ; ( 432 ) ( 442 ) ; ( 433 ) ( 443 ) ; ( 434 ) ( 444 ) ; ( 431 ) ( 445 ) ; ( 432 ) node@xmath316 ( 446 ) ; ( 432 ) node@xmath317 ( 447 ) ; ( 433 ) ( 448 ) ; ( 434 ) ( 449 ) ; in @xcite the growth diagram technique was used to show the symmetry property for the rs algorithm with row insertion . to row insert a @xmath28 into a tableau @xmath6 , we again keep @xmath318 unchanged . then we append a box at the end of first row of @xmath319 , where @xmath320 is the smallest number in row 1 of @xmath6 that is larger than @xmath28 , then we append a box at the end of second row of @xmath321 , where @xmath322 is the smallest number in row 2 of @xmath6 that is larger than @xmath320 , and so on and so forth . more precisely , define : @xmath323 then the output tableau has : @xmath324 for example if we insert a 3 into the tableau , the one - column insertion diagram is as follows : = [ fill = white ] ( 00 ) at ( 0,0 ) @xmath60 ; ( 01 ) at ( 0,1 ) @xmath70 ; ( 02 ) at ( 0,2 ) @xmath70 ; ( 03 ) at ( 0,3 ) @xmath71 ; ( 04 ) at ( 0,4 ) @xmath72 ; ( 05 ) at ( 0,5 ) @xmath73 ; ( 06 ) at ( 0,6 ) @xmath74 ; ( 07 ) at ( 0,7 ) @xmath75 ; ( 08 ) at ( 0,8 ) @xmath76 ; \(10 ) at ( 1.1,0 ) @xmath60 ; ( 11 ) at ( 1.1,1 ) @xmath70 ; ( 12 ) at ( 1.1,2 ) @xmath70 ; ( 13 ) at ( 1.1,3 ) @xmath72 ; ( 14 ) at ( 1.1,4 ) @xmath73 ; ( 15 ) at ( 1.1,5 ) @xmath74 ; ( 16 ) at ( 1.1,6 ) @xmath325 ; ( 17 ) at ( 1.1,7 ) @xmath326 ; ( 18 ) at ( 1.1,8 ) @xmath327 ; the most elegant one is the duality . denote by @xmath329 and @xmath330 the tableau pairs when applying column and row insertions to word @xmath195 respectively . also denote by @xmath331 the inverse word of @xmath195 : @xmath332 . moreover , denote by @xmath333 the evacuation operation on @xmath22 , see e.g. @xcite . then another simple relation is between the original definitions . row insertion was initially defined as an algorithm of inserting and bumping based on the ordering of integers . this definition turns into column insertion if one replaces all occurrance of `` row '' by `` column '' and replaces the strong order ( greater than ) with the weak order ( greater than or equal to ) due to the asymmetry of the ordering in rows and columns in the definition of a tableau . for these definitions see e.g. @xcite . the third relation is a bit more complicated . in column insertion , we initialise @xmath301 . then in each step we find the largest row index @xmath339 for a letter @xmath28 such that @xmath340 , append a box to @xmath341 and increase @xmath28 by 1 . in row insertion , we initialise @xmath342 , and in each step one we find the smallest letter @xmath343 for a row index @xmath14 such that @xmath344 , append a box to @xmath345 and increase @xmath14 by 1 . in a recent paper @xcite , a @xmath0-weighted robinson - schensted algorithm with row insertion was proposed . it is defined in a similar way as the @xmath0-column insertion and has the third relation with the column insertion , which we detail below . in @xmath0-column insertion , we initialise @xmath104 . then in each step we run over all row indices @xmath339 for a letter @xmath28 , with some weight append a box to @xmath341 and increase @xmath28 by 1 . in @xmath0-row insertion , we initialise @xmath342 . in each step one we run over all letters @xmath343 for a row index @xmath14 such that , with some weight we append a box to @xmath345 and increase @xmath14 by 1 . our definition here is a reformulation equivalent to the one in @xcite . for @xmath107 define @xmath346 when @xmath0-row inserting a @xmath28 into a tableau @xmath6 , we keep the first @xmath48 shapes unchanged : @xmath347 , with weight @xmath163 , @xmath348 and for @xmath349 , we have a binary branching for each triplet @xmath350 : @xmath351 denote by @xmath352 the weight of obtaining a @xmath53 after row inserting a @xmath28 into @xmath6 . we define the @xmath0-row insertion for word input in the same way as @xmath0-column insertion of words : successively @xmath0-row inserting letters , multiplying the weights , keeping a tableau @xmath22 to record changes for each @xmath6 , and merging the same tableau pairs by adding up the weights . denote by @xmath353 the weight of obtaining tableau pair @xmath194 after @xmath0-row inserting a word @xmath195 . then we have the same recursion rule : for a pair of tableaux @xmath194 with the same shape of size @xmath11 and a word @xmath195 of length @xmath196 , @xmath354 where the sum is over all tableau @xmath198 with the same shape as @xmath199 . since the algorithm has a similar triplet branching structure as in figure [ fig : branch ] , we can build the one - column insertion construction like in figure [ fig : insertletter ] , and concatenate them into a growth graph . with the same approach we can show the symmetry property for this algorithm . again denote by @xmath356 nodes associated with vertices surrounding a box from the south and the west . the set @xmath357 of all partitions branched from the triplet with @xmath358 belongs to one of the following 5 cases . 1 . the box has an @xmath69 in it . and @xmath234 , @xmath235 are equal to @xmath9 . then @xmath357 consists of only one partition @xmath359 with weight 1 . 2 . the box does not have an @xmath69 in it and @xmath248 . then @xmath357 consists of only one partition @xmath235 with weight 1 . the box does not have an @xmath69 in it and @xmath250 . then @xmath357 consists of only one partition @xmath234 with weight 1 . the box is empty . @xmath252 and @xmath253 for some @xmath254 . then @xmath357 again only contains one element @xmath255 with weight 1 . the box is empty and @xmath257 for some @xmath36 . then @xmath360 with weight @xmath361 first we define a _ branching insertion algorithm _ as an algorithm that has the branching structure as in figure [ fig : branch ] when we insert a letter to a tableau . that is , there exists an _ initial branching weight function _ @xmath362 , a _ high level weight function _ @xmath81 and a _ low level weight function _ @xmath84 such that when inserting a letter @xmath28 , the weight of a new @xmath36th shape is : @xmath363 ( i ) : : the insertion of @xmath364 into a tableau results in increment of one coordinate by 1 in @xmath365 , while keep @xmath366 unchanged , that is , the support of @xmath367 and @xmath368 are in @xmath369 and @xmath370 , ( ii ) : : @xmath371 if @xmath372 ; ( iii ) : : @xmath373 does not depend on the inserted letter ; ( iv ) : : @xmath374 does not depend on the inserted letter , we use the same notations @xmath356 as in descriptions of growth graph rules such that they surround the box @xmath202 from the south and the west . the symmetry property is shown once we can construct a symmetric growth graph rule , where the relation of @xmath9 , @xmath234 and @xmath235 and whether there s an @xmath69 in the box , determines the location of the box in the permutation matrix , and vice versa . then the 5 cases of the growth diagram rule are satisfied given these conditions : for case 1 , the equivalence between @xmath375 and @xmath376 is given by ( i ) and ( ii ) , and the branched shapes and weights are given by ( iii ) ; for case 2 both the equivalence between @xmath377 and @xmath378 is given by ( ii ) , and the singleton branching with weight @xmath163 is also given by ( ii ) ; for case 3 both the equivalence between @xmath379 and @xmath380 is given by ( i ) , and the singleton shape with weight @xmath163 is also given by ( i ) ; for case 4 and 5 the equivalence between @xmath381 and @xmath382 is given by ( i)(ii ) , and the singleton shape with weight @xmath163 in case 4 is given by ( ii ) while the branched shapes and weights in case 5 is given by ( iv ) . for example , one column insertion algorithm proposed by @xcite satisfies the conditions and has a symmetry property . see dynamics 3 in section 6.5.3 and ( 8.5 ) in section 8.2.1 in that paper . here we restate the definition in the framework of branching insertion algorithms . denote for @xmath107 : @xmath383 the algorithm is defined by the following weights @xmath384 we can see from these weight function formulae that this algorithm is different from the @xmath0-column insertion algorithm in the previous sections . qualitatively , it has row - insertion like branching of inserting a bumped letter into a lower row , which never happens in normal column insertion where a bumped letter stays in the same row or is inserted to a higher row . as such , it is more natural to compare the @xmath0-column insertion discussed in the previous sections with the @xmath0-row insertion , as we have pointed out the similarity of the relation between the @xmath0-column insertion and the @xmath0-row insertion and the relation between the normal column and row insertions . in @xcite another pair of @xmath0-rs column and row insertion algorithms called `` @xmath0-whittaker - multivariate ` dynamics ' with deterministic long - range interactions '' were introduced . these are also branching insertion algorithms . we omit the definitions here and point interested readers to section 8.2.2 of that paper . these algorithms also satisfy the conditions in proposition [ p : branch ] and thus enjoy the symmetry property .

in this report we have made a systematic study of strangeness production in proton - proton(pp),proton - nucleus(pa ) and nucleus- nucleus(aa ) collisions at cern super proton synchroton energies , using + ( version ) . numerical results for mean multiplicities of neutral strange particles , as well as their ratios to negatives hadrons( ) for p - p , nucleon - nucleon(n - n),p - s , p - ag , p - au(min . bias)collisions and p - au , s - s , s - ag , s - au ( central)collisions are compared to experimental data available from cern experiments and also with recent theoretical estimations given by others models( + and ) . neutral strange particle abundances are quite well described for p - p , n - n and p - a interactions , but are underpredicted by a factor of two in a - a interactions for in symmetric collisions(s - s , pb - pb)and for in asymmetric ones(s - ag , s - au , s - w ) . the ratios antistrange / strange are well described for but the ratios for multistrange particles ( ) and their antiparticle , could not be predicted by the model.a detailed analysis in limited rapidity range is required in order to draw some definite conclusion . a qualitative prediction for rapidity , transverse kinetic energy and transverse momenta normalized distributions are performed at 200 gev / nucleon in p - s , s - s , s - ag and s - au collisions in comparison with recent experimental data . hijing model predictions for coming experiments at cern for s - au , s - w and pb - pb interactions are given . the theoretical calculations are estimated in a full phase space . the factor of two can not be explained by considering very central events ( ) and taking into account minijets . new ideas are necessary in order to improve the comparison with experiment for nucleus - nucleus interactions . = 16truecm = 24truecm 1 = = .5truecm = -2truecm 1truecm * dipartamento di fisica `` g.galilei '' + * * istituto di fisica nucleare + * * sezione di padova + * italy * * 2truecm * strangeness production in pp , pa , aa * + 0.3 cm * interactions * + 0.3 cm * at sps energies . hijing approach * + 1truecm * wa94 collaboration * + dipartimento di fisica``g.galilei'',infn sezione di padova , via marzolo 8 , 35131 padova,*italy + * , * romania + * and + dipartamento di fisica `` g.galilei'',infn sezione di padova , via marzolo 8 , 35131 padova,*italy + * * dfpd-94-np-42 8 june 1994 * _ work presented at the international conference `` physics of high energy heavy ion collisions '' , vuosaari(helsinki),finland,17 - 22 june 1994 _ 1truecm = -1truecm = -2truecm

hep-ph9407262

searching for possible exotic states of nuclear matter is one of the currently most interesting and stimulating subjects in high energy nuclear physics ( see references [ 110 ] ) to create exotic matter such as the quark - gluon plasma ( qgp ) [ 14,5]strange matter @xcite,@xcite , multi@xmath11-matter , pion condensate , kaon condensate @xcite,@xcite it is essential to generate extreme states of nuclear matter regarding density , temperature or energy density @xcite . such extreme states are considered to be produced most probably in high energy heavy ion reactions[1014 ] or in the annihilation of energetic antiprotons(@xmath12)on a nucleus @xcite the possibility suggested by witten @xcite that strange quark - matter might be the lowest energy state of large baryon number , would have many consequences in physics and astrophysics and also cosmology relevance [ 1620 ] . small drops of strange matter(strangelets ) may be natural results of the evaporation of a qgp formed in heavy ion collisions . enhanced strangeness production @xcite in ultrarelativistic heavy ion collisions ( urhic ) is one of the most widely discussed signature for creating qgp @xcite,@xcite . several experiments have been performed to investigate strangeness production in reactions involving relativistic nuclei , particularly at super proton synchrotron ( sps ) at cern : na35 [ 2232 ] , na36 [ 3337],wa85,wa94[3845],[128130],na44 [ 132133 ] , hellios collab.[4650 ] , and at the alternating gradient synchrotron(ags ) at brook- + haven e802 [ 5155 ] , e810 @xcite,@xcite and e814 @xcite,[134 ] . these experiments have measured also the yield of several particle species containing one or two strange quarks ( @xmath13 , @xmath14,@xmath15,@xmath16 and their antiparticles ) or three strange quarks(@xmath17 ) . in all the cases their production relative to pions or negative hadrons and some other ratios are larger in nucleus - nucleus collisions compared to proton - proton . now experiments with truly heavy ion projectile have just started(with a gold beam at bnl)@xcite,@xcite or are under way ( @xmath18 at cern)@xcite . the soft physics data have stimulated a great deal of theoretical activity and in the present introduction it is impossible to give more than a superficial overview of the main types of the monte carlo models used.many models for urhic have been developed : dual partons models(dpm ) [ 5966],@xcite quark gluon string models(qgsm)[6769 ] , venus models for very energetic nuclear scattering [ 7075 ] fritiof model @xcite,@xcite , attila @xcite model , relativistic quantum molecular dynamics ( rqmd ) [ 7983][135138],parton string model(psm)[139 ] , hijet model@xcite . an excelent review and detailed comparision of the models was done by werner @xcite . on the other hand , no unconventional explanation has so far put forward for the observed increase of antyhyperons ( @xmath19,@xmath16 ) . actually , it is the old pomeron exchange picture that is revitalized in some models with a new ingredient that is the pomeron is made of two colored strings.thus the problem of interaction of strings becomes important.this interaction may give rice to the fusion of strings[8588][137,138 ] making a new string with more color charge.these new objects could explain the strangeness enhancement @xcite , @xcite,[137,138]especially @xmath19 enhancement . such models can be seen as an interplay between independent string fragmentation and qgp formation.one version of venus generator also include such interaction @xcite , @xcite . we note also that an interpretation based on a particular dependence of the cross sections for urhic was taken into account @xcite . in this report we have performed a systematic study of strange particle production in proton proton(@xmath20),proton nucleus(@xmath21 ) and nucleus nucleus ( @xmath22 ) interactions at sps cern energies using hijing(heavy ion jet interacting generator ) monte carlo model [ 9197 ] developed at berkeley . the formulation of hijing is guided by the lund fritiof ( @xcite@xcite ) and dpm @xcite models phenomenology for soft nucleus - nucleus reactions at intermediate energies @xmath23 and the successfull implementation of perturbative qcd ( pqcd ) processes in the pythia model @xcite@xcite for hadronic interactions . hijing incorporate the pqcd to multiple jet processes and the nuclear effects such as parton shadowing and jet quenching . hijing provide a link between the physics at intermediate cern sps energies and the highest collider energies at relativistic heavy ion collider(rhic)(@xmath24 and the cern large hadron collider ( lhc ) ( @xmath25 ) . using hijing monte carlo model @xcite,@xcite a detailed discussion and comparison with a wide variety of data in @xmath20,@xmath21 and @xmath22 collisions was reported.however no study on the strangeness production at sps - cern energies in @xmath20,@xmath21 and @xmath22 data exist within this approach . in this paper we present some theoretical calculations for mean multiplicities of strange particle production in the full rapidity range as well as their ratios to negative hadron predicted by hijing model(hij.01 version -see section 2 and 3 for explanation of this version)for @xmath20 , nucleon - nucleon @xmath26,minimum biascollisions @xmath27 and + central collisions@xmath28 at the energy of @xmath29 . also a qualitative prediction for normalised distributions for rapidity , transverse kinetic energy and transverse momenta are done at 200 gev / nucleon in @xmath30 and sau collisions and the results are compared with recent experimental data from na 35 collaboration[140].we note that theoretical calculations are done for full phase space . for giving an idea of abundances and spectra in experiments planned in the near future at cern , we give some predictions for sw and pbpb interactions . we note that the version hij.01 of the model includes neither the formation of qgp nor some collective dynamics on either the hadronic or string level.although the existing trends of hadron distributions are quite satisfactorily approximated by the hijing model @xcite,@xciteand also neutral strange production in @xmath20 and @xmath21 interactions are quite well described , this version is getting less reliable for strange particle produced in heavy colliding systems . the same result is stressed out and in other models qgsm@xcite,@xcite , rqmd@xcite , fritiof@xcite , psm@xcite . a brief description of the hijing monte carlo model and theoretical background are given in section 2.detailed numerical results of this version hij.01 for @xmath31 and @xmath32 reactions especially at cern - sps energies@xmath33 for strangeness production are compared to experimental data and other models predictions like qgsm @xcite,@xcite , dpm @xcite,[141],142],venus @xcite , fritiof @xcite , and rqmd @xcite,[140 ] in section 3 . section 4 concludes with a summary and discussion of future applications . a detailed discussion of the hijing monte carlo model was reported in references @xcite,@xcite,@xcite . the formulation of hijing was guided by the lund - fritiof and dual parton model phenomenology for soft nucleus - nucleus reactions at intermediate energies ( @xmath34)and implementation of perturbative qcd(pqcd ) processes in the phythia model for hadronic interactions.comparison between some monte carlo generators are given in reference @xcite all pqcd models employ initial and final state radiation in an iterative fashion : a parton with a squared mass @xmath35 radiates a parton and leaves a remainder parton with reduced @xmath35,the latter one radiating again and so on .the hadronization finally is done by ussing string model(pythia).a link to the soft semihard models is provided by pythia , where it was realized that the perturbatively calculated cross section is an inclusive one and may well exceed the total cross section meaning multiple scattering.therefore the cross sections was eikonalized : for a given impact parameter multiple scattering occurs acording to a poissonian distribution , with the average number of scatterings , being @xmath36dependent , proportional to some profile function @xmath37.hijing is based on a particular model of high energy pp inelastic collisions . we give in this section brief review of the aspect relevants to hadronic interaction : 1 . the models includes multiple minijet production with initial and final state radiation along the lines of the pythia model @xcite,@xcite and with cross sections calculated within the eikonal formalism . soft beam jets are modeled by quark - diquark strings with gluon kinks along the lines of the dpm and fritiof models.multiple low @xmath38 exchanges among the end point constituents are included . exact diffuse nuclear geometry is used to calculate the impact parameter dependence of the number of inelastic processes @xcite . an impact parameter dependent parton structure function is introduced to study the sensitivity of observable to nuclear shadowing , especially of the gluon structure functions . 5 . a model for jet quenching is included to enable the study of the a - dependence of moderate and high @xmath38 observable on an assumed energy loss of partons traversing the produced dense matter . hijing does not incorporate the mechanism for final state interactions among low @xmath38 produced particles . the rate of multiple minijet production in hijing is constrained by the cross sections in nucleon - nucleon collision .within an eikonal formalism @xcite the total elastic cross sections @xmath39,total inelastic cross sections @xmath40 and total cross sections @xmath41 can be expressed as : 0.3 cm @xmath42 0.3 cm @xmath43 0.3 cm @xmath44 0.3 cm strong interactions involved in hadronic collisions can be generally divided into two categories depending on the scale of momentum transfer @xmath45 of the processes . if @xmath46 the collisions are nonperturbative and are considered _ soft_and if @xmath47 the subprocesses on the parton level are considered _ hard _ and can be calculated via pqcd @xcite . in the limit that the real part of the scattering amplitude is small and the eikonal function @xmath48is real , the factor 0.3 cm @xmath49 0.3 cm in terms of semiclassical probabilistic model can be interpreted as _ the probability for an inelastic event of nucleon nucleon collisions at impact parameter @xmath36 , which may be caused by hard , semihard or soft parton interactions_. to calculate the probability of multiple minijet a main dynamical assumption is that they are independent.this holds as long as their average number is not too large.the independence should apply up to lhc energies @xcite when shadowing can be neglected , the probability of no jets and @xmath50 independent jet production in an inelastic event at impact parameter @xmath36,can be written as : 0.3 cm @xmath51 0.3 cm @xmath52^{j}}{j!}\cdot exp(-2\,\chi_{h}(b , s))\,\,\,\,\,\ , j \geq 1 \label{e6}\ ] ] 0.3 cm where @xmath53 is the eikonal function for soft interaction , @xmath54 is the average number of hard parton interactions at a given impact parameter , @xmath55 is the probability for no soft interaction . summing eqs.(5 ) and ( 6 ) over all values of @xmath50 leads to : 0.3 cm @xmath56 comparing with eq.([e4 ] ) one has : 0.3 cm @xmath57 0.3 cm if we consider that the parton distribution function is factorizable in longitudinal and transverse directions and the shadowing can be neglected the average number of hard interaction @xmath58 at the impact parameter @xmath36 is given by : 0.3 cm @xmath59 0.3 cm where @xmath60 is the effective partonic overlap function of the nucleons at impact parameter @xmath36 . 0.3 cm @xmath61 0.3 cm with normalization 0.3 cm @xmath62 0.3 cm and@xmath63is the is the pqcd cross section of parton interaction or jet production @xcite,@xcite if we note @xmath64,where @xmath65provide a measure of the geometrical size of the nucleon @xmath66 we get assuming the same geometrical distribution for both soft and hard overlap functions 0.3 cm @xmath67 @xmath68 @xmath69\chi_{0}(\xi ) \label{e14}\ ] ] 0.3 cm we note that@xmath70 is a function not only of @xmath71 but also of @xmath72because of the @xmath73 dependence on the jet cross section @xmath74.geometrical scaling implies on the other hand , that @xmath75 is only a function of@xmath76 . thefore geometrical scaling is broken at high energies by the introduction of the nonvanishing @xmath77 of jet production . now we can rewrite the cross sections of nucleon - nucleon collisions as : 0.3 cm @xmath78 @xmath79 @xmath80 0.3 cm the calculation of these cross sections requires specifying @xmath81 with a coresponding value of cut - off momenta @xmath82 @xcite,@xcite in the energy range @xmath83,where only soft parton interactions are important cross sections @xmath81 is fixed by the data on total cross sections @xmath84 directly.in and the above @xmath85 energy range @xmath86 we fix @xmath81 at a value of 57 mb with @xmath87,in order to fit the data of the cross sections.between the two regions @xmath88,we simply use a smooth extrapolation for @xmath81@xcite . in hijing a nucleus - nucleus collisions is decomposed into binary collisions involving in general excited or wounded nucleons . wounded nucleon are assumed to be @xmath89 string like configurations that decay on a slow time scale compared to the collision time of the nuclei.in the fritiof scheme wounded nucleon interactions follow the same excitation law as the original hadrons . in the dpm scheme subsequent collisions essentially differ from the first since they are assumed to involve sea partons instead of valence ones.the hijing model adopt a hybrid scheme , iterating string - string collisions as in fritiof but utilizing dpm like distributions . by incorporating the successful multistring phenomenology for low @xmath38 interactions at intermediate energies , hijing provide a link between the dominant nonperturbative fragmentation physics at intermediate cern - sps energies and perturbative qcd physics at the highest collider energies(rhic;lhc ) strange particle production provide some information about the question of whether or not a quark - gluon plasma occurs in heavy ion collisions.this is major motivation to first study the production of strange particles in pp scattering.in addition , strange particle give some insight into dynamics already for pp collisions . we run the program hijing , with default parameters , mainly ihpr2(11)=1 , which means choise of baryon production model with diquark - antidiquark pair production allowed , initial diquark treated as unit ; ihpr2(12)=1 , decay of particle such as @xmath90,@xmath91,@xmath15,@xmath92,@xmath93 , @xmath94 are allowed;ihpr2(17)=1 - gaussian distribution of transverse momentum of the sea quarks ; ihpr2(8)=0 - jet production turned off for theoretical predictions in hijing model- hij.01 , and ihpr2(8)=10-the maximum number of jet production per nucleon - nucleon interaction for for theoretical predictions @xmath95 . the energy dependence of average multiplicities for different particle species , gives a first impression about the predictive power of the model and provide an important check of model approach . in table 1 we give average multiplicities of particle at @xmath96 in proton - proton interaction , @xmath97 being the laboratory energy . the theoretical values @xmath98 are obtained for @xmath99 generated events and in a full phase space . the values @xmath95 are for very central @xmath20 collisions(@xmath100 ) and with minijet production . the experimental data are taken from reference @xcite . the large difference between pions and kaons is due to the suppressed strangeness production from string fragmentation.pions(@xmath101 ) and kaons(@xmath101 ) are more frequent than the negative ones due to charge conservation.we note that the multiplicities for neutral strange particle @xmath102 and for antiproton @xmath103are quite well described in the limits of three standard error deviations for @xmath20 interactions at @xmath104 .however the values for @xmath105 and @xmath106 are slightly overpredicted by the model . in order to give some comparision at ultrahigh energy and to test the predictive power of generator we perform calculations at centre of mass energy @xmath107(@xmath85-energies ) , for @xmath108 interactions . by using many pieces of data from the different collider experiments ( charge particle @xmath109 and @xmath110 from * ua5 * , @xmath111 from * ua2 * ) , the * ua5 * collaboration has attempt to piece together a picture of the composition of a typical soft event at the @xmath85 @xcite.the measurements were made in various different kinematic regions and have been extrapolated in the full transverse momenta(@xmath38 ) range and rapidity range for comparison as described in reference @xcite . the experimental data are compared with @xmath112 results in table 2 . the number of generated events was @xmath113 . it was stressed out @xcite that the data show a substantial excess of photons compared to the mean @xmath114.it was suggested like explanation the possible emission of gluon cerenkov radiation in hadronic collision @xcite.at this energy the model @xmath95 seems to work fairlly well and the data are well described . our calculations ruled out such hypothesis.taken into account direct gamma production the agreement with the data is in the limit of experimental errors . the experimental ratio @xmath115 are predicted also by@xmath116 model ( 0.99 ) . we note that a study of the ratios of invariant cross sections of kaons to that of pions as a function of transverse momenta in the central region was done by wang and gyulassy @xcite,@xcite . the behaviour of average multiplicities for different particles species as a function of the energy can be reproduced by hijing model@xcite and also by venus model(see reference @xcite ) . appropriate variable to describe single particle properties are the transverse momentum @xmath38 and the rapidity @xmath117 0.3 cm @xmath118 0.3 cm with @xmath119,and @xmath120 being energy , longitudinal momentum and transverse mass 0.3 cm @xmath121 0.3 cm with @xmath122 being particle mass . in figures 1a , b and 2a , b we show normalized rapidity ( 1a,2a)and transverse momentum distributions(1b,2b ) for @xmath123 s ( fig.1a , b ) and for @xmath124 s for@xmath125 scattering at@xmath126 . the theoretical values hij.01 are compared with experimental data taken from jaeger et al.@xcite hijing results and data @xcite are very similar , although @xmath127 does not show some structures present in data . we see the two distinct lambda peaks , being due ( in zeroth order)the fragmentation of the forward baryonic string and the backward baryonic string into leading baryons.therefore lambda distributions are similar to proton distributions , just reduced by some factor and with the diffractive peak missing(see also reference @xcite ) . the kaons are peaked at mid rapidity and are similar to the pions , though again less in magnitude.the distributions are narrower compared to proton s and pion s . in the following we investigate strange particle production in hijing model approach , for proton - nucleus and nucleus- nucleus collisions . we have studied the production of neutral strange particle @xmath128 in comparison with recent experimental data from na35 cern - experiments , for @xmath129 interactions . we investigate the average multiplicities for negative hadrons @xmath2,negative pions(@xmath130 and neutral strange particles ( @xmath131,@xmath132 , @xmath106 ) in @xmath133 minimum bias collisionsand @xmath134 central collisions at @xmath135 . the default parameters were used during the simulation and also option jet no(ihpr2(8)=0 ) . the number of monte carlo generated events was @xmath99 for @xmath136 and @xmath137interactions , @xmath138 for@xmath139 and @xmath140 for @xmath141 and @xmath142 collisions . the results are given in tabel 3 in comparison with experimental data @xcite.the corresponding data for @xmath20 and @xmath143 at 200 gev @xcite are also shown in table 3 . we remark that hijing model describe quite well neutral strange particle multiplicities for @xmath136 and @xmath144 interactions , but underestimate by a factor of two the values for nucleus - nucleus interactions for @xmath132,@xmath106,@xmath131 in symmetric collisions @xmath145(expected values ) and for @xmath132 , @xmath106 in assymmetric ones ( @xmath146 ) . we note also that our theoretical calculations are done for @xmath144 minimum bias collisions and the experimental data are for the events with charged particle multiplicity greater than five , which contain a significant fraction ( about 90 % ) of the minimum bias events @xcite . in order to get some idea about the discrepancies between theoretical calculations and experimental data we try to run the program in the limiting condition for very central events ( @xmath147 ) and jet allowed . the results are given in table 4 and are compared also with recent theoretical approach in rqmd @xcite , qgsm @xcite,@xcite and dpm models ( @xmath148 are from reference @xcite , variant of dpm which include additionally @xmath149 production from the sea into the chain formation process , @xmath150 are from reference @xcite , variant of dpm which include chain fusion , good candidate to explain the anomalous antihyperon production ) . in order to study the total production of strangeness in these collisions in a model independent way and to use all available experimental information na35 collaboration @xcite has studied the ratio defined as : 0.3 cm @xmath151 0.3 cm we estimate this ratio in hijing approach for the above interactions and give the numerical results in table 5 together with the mean multiplicities for negative pions . we can conclude from this analysis that the factor of two for @xmath152 which was reclaimed from multiplicities analisis for symmetric interactions are ruled out.experimental error are to higher and new data with higher statistics are needed for draw a definite conclusions . taken into account a factor of two in neutral strange particle production we get a value @xmath153 independent of the target nucleus mass number for @xmath22 interaction from @xmath154 to @xmath142 . without factor of two the theoretical @xmath155 values for @xmath144 and @xmath32 are aproximatively equal to the values for nucleon - nucleon interactions . however the experimental values of @xmath155 are two times higher then the corresponding @xmath156 and @xmath144 values , a fact established from recent na35 data . therefore new ideas are required to modify hadronic models in order to reproduce strange particle production at high energies . a different approach of finding signals for qgp is simply to compare nucleus - nucleus ( aa ) or hadron - nucleus(ha ) data with nucleon - nucleon(nn ) data . however , in particular this assumption is certainly not correct , since at high energies aa scattering is not simply a superposition of nn collisions.the time scale between two nn interactions is too small for hadronization and therefore some intermediate object is involved . simple theoretical estimates are very unrealiable .comparing aa with nn data or pa and nn data does not help , because new results do not necessarily mean qgp.calculation based on qcd are not feasible , because we are in domaine where pqcd does not apply . soft interactions involving small transverse momenta transfer should be taken into account . this ratios for multistrange particles were analised in terms of thermal models@xcite,@xcite , + @xcite , @xcite . a constraint between thermal fireball parameters arises from the requirement that the balance of strangeness in a fireball is nearly zero.the impact of this constraint on ( multi-)strange ( anti-)baryon multiplicities comparing hadron gas and qgp has been analised.the data are compatible with the qgp hypothesis and appear to be inconsistent with the picture of an equilibrated hadron gas fireball @xcite,@xcite . however the two scenarios ( slow and rapid)for the expansion of a qgp were recently studied in detail .production ratios @xmath157 and @xmath158 for hadron gas at @xmath159 and plasma break - up have been compared to experimental data at baryonic chemical potential @xmath160.present data on strange particle production can not provide a distinction between the two scenarios @xcite . we calculate the ratios of the mean strange particles multiplicities to the mean multiplicities of negative hadrons and we show in table 6 a comparison with experimental na35 data @xcite for the same interactions like in table 3.the agreement can not be obtained without factor of two in neutral strange multiplicities . we note that negative hadrons as well as negative pions are well described even for very complex @xmath32 interactions . in tabel 7 we give the multiplicity ratios for negative hadrons as well as for neutral strange particle for pa / pp , aa / pp and aa / pa resuts from hijing ( @xmath1 ) and @xmath161 models are compared to na5 @xcite and na35 @xcite experimental data.the calculations on the models @xmath1 and @xmath161 reproduce the relative increase of the @xmath91 ratios in central @xmath162 collisions , but can not describe the data concerning @xmath15 and @xmath19 generation . comparing experimental data with theoretical prediction of the hijing model we see that these ratios can not be described witouth factor of two for @xmath163 and @xmath164 . data for charge kaon production have already been published by helios collaboration for @xmath165 and @xmath166 at @xmath167 in the limited target rapidity region @xmath168 and by na35 collab . for @xmath169and @xmath170 collisions at @xmath167 in a wider rapidity interval . both experiments have shown an increase of the ratio @xmath171 as compared with the elementary interaction , while no significant increase was observed for @xmath172 ratio . the experimental values @xcite,@xcite are compared with hijing results in tabel 8a and tabel 8b . in table 9 the results on average multiplicities of @xmath13 and @xmath173 and multiplicities ratios in @xmath174 interactions extrapolated to the full phase space are compared with model predictions : hijing ( hij.01),venus and fritiof . the experimental data are taken from reference @xcite . all models underestimate the @xmath175 ratio , but reproduce well the @xmath176 ratio . the models seems to overpredict the average @xmath177 multiplicities in nucleus - nucleus collisions.the secondary collisions induced by the pion and kaon mesons in nuclei may influence the ratios @xmath178 @xcite . experimental values for multistrange baryon ratios in nucleus nucleus interactions were recently reported @xcite , @xcite , @xcite.we check if the model could describe these ratios , but our calculations are for full phase space and experimental data are reported in limited rapidity range . we see from table 10 that only ratios @xmath8 are well described by the model.the model overestimate the productions of @xmath179 and @xmath180 . the ratios are well described for @xmath181 but the ratios for multistrange particles and their antiparticle could not be predicted by the model.a detailed analysis in limited rapidity range is required in order to draw some definite conclusions ( calculations are now in progress ) . we remark that the ratios for multistrange particle production are not predicted by this version of hijing model , nor by more sophisticated models like qgsm , rqmd , dpm , spm . a qualitative prediction for rapidity and transverse kinetic energy are performed at @xmath182 in @xmath183 and @xmath184 collisions and are represented in comparison with experimental data from t.alner et al.,@xcite . in fig.3a , b , c , d we give normalized rapidity distributions for @xmath123 produced in interactions quoted above : fig.3a for @xmath185 ; fig.3b for @xmath186 , fig.3c for @xmath187;and fig.3d for @xmath184 . in fig.4a , b , c , d and in fig.5a , b , c , d we give the rapidity distributions for @xmath188 and @xmath124 respectively . in fig.6a , b , c , d ; fig.7a , b , c , d and in fig.8a , b , c , d normalized distributions of transverse kinetic energy defined as @xmath189,are represented for the same interactions . theoretical hijing values ( hij.01 ) are multiplied by a factor of two ( hij.01 * 2 ) shown in the figures.the total number of events generated in full phase space are the same as for the results from table 3(see section 3.1 ) . transverse momentum distributions for @xmath123 , @xmath188 and @xmath124 particles produced in @xmath190 interactions are given in figures 9a , b , c . the ratio of @xmath188 to @xmath123 rapidity distributions for @xmath190 collisions are done in fig.10 . analysing the figures ( fig.3a , b , c , d fig.5a , b , c , d ) we see that similarly to @xmath191 collisions , the rapidity spectra of @xmath91 and @xmath19 particles in @xmath192 interactions have a sharp peak centered at mid rapidity and their width are close to the ones for the proton induced case . for @xmath15 particles the distribution is naturally symmetrical and its width is larger due to possible production of strange baryons in the target(projectile ) fragmentation region . a detailed discussion of @xmath174 collision system was done also by werner @xcite and amelin et al . @xcite , @xcite.the projectile and target mass asymmetry is reflected in the @xmath123 rapidity distributions which for @xmath185 , @xmath187and @xmath184 interactions reach a maximum at about 1.0 which is compatible with those remarked in experiment ( about 1 to 1.5 ) . the @xmath188 and @xmath124 rapidity distributions are less sensitive to the initial asymmetry of the system . as follows from figures , the shape of rapidity for kaons , lambdas and antilambdas are reproduced fairly well by hijing model , but the model uderestimate the production of @xmath15 in the central region of rapidity spectra , even for proton - sulphur interaction . new experimental data and also data on proton production in hadron - nucleus and nucleus - nucleus interactions are required in order to draw a definite conclusions . for transverse kinetic energy distributions ( taken into account the factor of two ) we obtain a good agreement with the data except the low @xmath120 values and in some cases the values at the tails . the normalised transverse momentum distributions for @xmath193 interactions are well described except some discrepacies seen for @xmath15 ( see fig.9a , b , c ) . the ratio of the @xmath19 and @xmath15 yields is suggested to reflect the net baryon density of the particle source . in fig.10 the ratio of @xmath19 to @xmath15 rapidity distributions for @xmath191 is presented in comparison with experimental data .the theoretical values are higher than experimental ones . it should be stressed out that in the most string models the produced strings do not interact and decay independently . introduction of some collective effects like string fusion @xcite,@xcite or firecreacker model @xcite can change the predictions essentially , especially for rapidity plateau heights and high multiplicity distribution tails . some analysis was done recently by sorge @xcite , @xcitein the framework of the rqmd .a doubling of lambda production in the target fragmentation region for collision on heavy targets were found due to nuclear cascading of the produced mesons.also in rqmd approach the resonances play an important role in creating strange quark pairs.werner @xcite has shown that rescattering should be considered at cern - sps energies with the effect of increasing neutral strange particles production especially in the central region.hijing does not incorporate these ingredients . no model seems to get a real description of absolute values for neutral strange particle production without additional mechanisms . since recently @xcite the data on multiple strange particle production ( @xmath194 ) was reported , in figures 11a - b we give rapidity distribution for full phase space @xmath16 ( fig.11a ) and @xmath195 ( fig.11b)for @xmath174 interaction . we can see from these figures that the ratio of multiplicities @xmath196 has strogly dependence on rapidity bins . improvement of the statistics and considerations of limited rapidity experimental intervals , are needed in order to due some more toward comparison with experimental data . for giving an idea about of the abundances of particles produced in the near planed experiments at cern we have studied @xmath197 interaction at @xmath198 and in addition to multiplicities given in tabel 3 and tabel 5 , normalised rapidity distributions are given in figures 12a b for ( @xmath199 ) . the number of generated events was@xmath200 and the results are represented for full phase space . the dependences rapidity - transvers momentum , which define theoretical values for acceptances as well as their bidimensional projections are also given in figures 13 a - b for @xmath15 , in figures 14 a - b for @xmath19 and for all negatives charges in @xmath186,@xmath201 and @xmath202 interactions in figures 15a , b;16a , b and 17a , b(respectively ) and for all positives charges in @xmath203 interactions in figures 18a - b . we hope that from such kind of correlations we will get more inside the dynamics of the systems and we will find some limits of the model concerning neutral strange particles production . in this report , we give a systematic study of strange particle production in pp , pa and aa collisions mainly at sps cern - energies , using hijing monte carlo model developed for high energy collisions . we analyse data from cern experiments ( na35,na36 , + wa85 ) and we give a detailed comparison for mean multiplicities of neutral strange particle for pp , ps , pag , pw , pau , ss , sw and pbpb interactions . one should note that hijing model includes neither hipotheses of qgp . strange particle abundances are quite well described in pp and pa collisions and are underpredicted by a factor of two in nucleus - nucleus interactions , by this version ( theoretical values for hij.01 ) .the ratios for multistrange particles could not be predicted by the model hij.01 .the abundances of @xmath16 and @xmath17 are underpredicted and the abundances of @xmath179 and @xmath180 are overpredicted . more carreful analisis are necessary in limited rapidity interval . we present also some calculations for the ratios @xmath204 and @xmath205 given by hijing model compared with other models predictions like venus and fritiof.all models seems to overpredict @xmath205 ratios and underpredict @xmath204 ratios . rescattering should be important even at sps -energies . a qualitative prediction for rapidity distributions of neutral strange particles in @xmath206 , @xmath207,@xmath208,@xmath209 @xmath210per nucleon are done compared also with recent experimental data[140 ] . the models describes correctly the shapes of the rapidity in central @xmath211 collisions for @xmath212 and @xmath188 but absolute numbers for multiplicities are underestimated by a factor of two for @xmath213 in symmetric collisions and for @xmath214 in asymmetric ones . some discrepancies are seen for @xmath15 rapidity distribution and transverse momentum distributions even for @xmath190 interactions.the factor of two can not be explained taken into account minijets and very central events(@xmath215 ) . the theoretical model predictions hij.01 presented above for central @xmath201,@xmath184 and @xmath202 collisions give an idea for particle abundances in experiments which are under the way or are planned in the near future at cern . the hijing model simulations are however still somewhat too low compared to the data.also the ratios for multistrange baryons could not be described . new ideas are necessary in order to improve the comparison of theoretical calculations with experimental data.the main goal is to increase the rate of neutral strange particle production by a factor of two in nucleus - nucleus interaction at @xmath216 . we remark that other models ( qgsm , venus , dpm and psm ) include now some aditional hypothesis in order to improve the discrepancies ( colective string - string interactions , double or multiple colour exchanges , string fusion ) . one of the authors ( vtp ) would like to thank to professor miklos gyulassy and dr xin - nian wang for providing the monte carlo program hijing and many helpful communications . + _ nucleus collisions _ , gattinburg , tennesse , usa , november 1115(1991 ) , eds . t. c. awes , f. e. obenshaim , f . plasil , m . r. strayer and c. y. wong(qm91 ) + nucl . a544*(1992 ) aussois , france , march 1620,1992,eds . m.bex , b.borderie,m.ploszajczak,j.richter,p.schuck and e.surand + nucl.phys.*a545*(1992 ) wiesbaden , germany + , july 26august 1,1992 , eds . r. bock , h . emling , e . grosse , u . grundinger + k. d. hildenbrand and j. knoll + nucl.phys.*a553*(1993 ) university of aarhus , denmark,2024 may 1991,eds.j.madsen and p.haensel + nucl.phys.*b25*(1992 ) eds .hans.h.gutbrod and johann rafelski , plenum press new york and london ( 1993 ) edited by e. stenlund , h . a. gustafsson , a . oskarsson and i. otterlund + nucl.phys.*a566*(1994 ) + perugia , italy 28 june2 july 1993,book of abstracts , courmayeur , aosta valley , italy 1992,eds.f.iazzi and a.zenoni + nucl.phys.*a558*(1993 ) s.a.bonometto,ornela pantano , phys.rep.*228 * ( 1993)177 p.koch.b.muller and j.rafelski phys.rep.*c142*(1986)167 s.nagamiya in _ perspective of meson science _ eds.t.yamazaki,k.nakai,k.nagamine(north holland , amsterdam - london - new york -tokyo,1992),cp 24 _ strange meson production in heavy ion collisions_,p735 m.sano and m.wakai(see reference [ 11]),cp25 , _ strangeness production in high energy collisions_,p763 j.schukraft,nucl.phys.*a553*(1993)31c h. r. schmidt and j. schukraft , cern - ppe-/9242 ( 1992 ) j.of phys.*g*(1993)1705 s.ahmad et al.nucl.phys.*a558*(1993)393c e.witten,phys.rev.*d30*(1984)272 e.farhi and r.l.jaffe,phys.rev.*d30*(1984)2379 k.kajantie and l.mc.lerran,ann.rev.and part.sci . * 37*(1987)293 h.stocker,c.greiner,j.schafner , nucl.phys.*b24*(1991)239c,246c h.bartz,nucl.phys.*b24*(1991)211c e.quercigh , see reference [ 5 ] p 499 j.bartke et al.,z.phys.*c48*(1990)191 r.stock et al.,nucl.phys.*a525*(1991)221c p.seyboth et al.,nucl.phys.*a544*(1992)293c m.kowalski et at.,nucl.phys.*a544*(1992)609 h.bialkowska et al.,z.phys.*c55*(1992)491 j.baechler et al.,z.phys.*c58*(1993)367 t.akeson et al.,nucl.phys.*b353*(1991)1 t.akeson et al.,z.phys.*c46*(1990)369 a.bamberger et al.,z.phys.*c43*(1989)25 i.derado et al.,z.phys.*c50*(1991)31 m.gazdzicki and o.hansen,nucl.phys . , * a528*(1991)754 d.e.greiner et al.,nucl.phys.*a544*(1992)309 e.anderson et al.,nucl.phys.*a553*(1993)817c e.andersen et al.,phys.lett.*b294*(1992)127 e.andersen et al.,nucl.instr.methods * a301 * ( 1991)69 e.andersen et al.,phys.lett.*b316*(1993)603 s.abatzis et al.,phys.lett.*b244*(1990)130 s.abatzis et al.,phys.lett.*b259*(1991)508 s.abatzis et al.,phys.lett.*b270*(1991)123 s.abatzis et al.,nucl.phys.*a525*(1991)445c s.abatzis et al.,nucl . phys . * a566*(1994)225c j.b.kinson et al.,nucl.phys.*a544*(1992)321c di bari , doctoral thesis ( january 1993 ) s.abatzis et al.,phys.lett.*b316*(1993)615 t.akesson et al.,preprint cern - ppe/87 - 176(1987 ) t.akesson et al.,preprint cern - ppe/89 - 97(1987 ) t.akesson et al.,preprint cern - ppe/87 - 170(1987 ) t.akesson et al.,preprint cern - ppe/89 - 111(1989 ) t.akesson et al.,preprint cern - ppe/92 - 130(1992 ) t.akesson et al.,z.phys.*c53*(1992)183 t.abbot et al.,phys.rev.lett.*64*(1990)847 t.abbot et al.,phys.rev.lett.*66*(1991)1567 t.abbot et al.phys.rev.*d45*(1992)3906 y.miake et al.,nucl.phys.*a525*(1991)231c t.abbot et al.,phys.rev.*c47*(1993)1351 k.j.foley et al.,nucl.phys.*a544*(1992)335c , w.a.love et al.,e810 collaboration , in _ proceedings of the workshop on heavy ion physics at ags , bnl_,(1990),edited by o.hansen r.a.salmeron,nucl.phys.*b389*(1993)301 a.capella et al.,phys.lett.*b81*(1979)68 a.capella and j.tran thanh , z.phys.*c10 * ( 1981)249 a.capella,j.kwiecinski,j.tran thanh , phys.lett.*b108*(1982)347 a.capella et al.,z.phys.*c33*(1987)343 j.ranft and s.ritter,z.phys.*c27 * ( 1985)413 h.j.mohring,j.ranft,s.ritter , z.phys.*c27*(1985)419 h.j.mohring,j.ranft,z.phys.*c52 * ( 1991)643 h.j.mohring et al.,nucl.phys.*a525 * ( 1991)493c a.b.kaidalov,sov.j.nucl.phys.*33*(1981)733 a.b.kaidalov and o.i.piskunova,z.phys.*c30 * ( 1986)145 a.b.kaidalov , nucl.phys.*a525*(1991)39c k.werner,phys.lett.*b197*(1987)225 k.werner,z.phys.*c38*(1988)193 k.werner,phys.rev.*d39*(1989)780 k.werner,z.phys.*c42*(1989)85 k.werner and p.koch,z.phys.*c47 * ( 1990)215 k.werner,nucl.phys.*a525*(1991)501c b.andersson,g.gustafson and nilsson - almqvist nucl.phys.*b281*(1987)289 b.nilsson-almqvistand e.stenlund , comp.phys.commun.*43*(1987)387 m. gyulassy , cern report th-4794/87(1987 ) j. aichelin et al.,phys.rev.*c37 * ( 1988)2451 r. mattielo , h . sorge , h . stocker and w. greiner , phys.rev.lett.*63*(1989)1459 h. sorge , h . stocker and w. greiner , nucl.phys.*a498*(1989)567c a. von keitz , h . sorge , h.stocker and w.greiner,nucl.phys.*a527*(1991)601 h. sorge et al.,see ref.[5 ] p 335 k. werner , preprint hd - tvp-93 - 1(1993 ) phys.rep.*232*(1993)87 m. braun and c.pajares,nucl.phys.*b390 * ( 1993)542 m. braun and c.pajares,nucl.phys.*b390 * ( 1993)559 m. braun and c.pajares,phys.lett.*b287 * ( 1992)154 m. braun and c.pajares,phys.rev.*d47 * ( 1993)114 c. merino , c.pajares and j.ranft,phys.lett . * b276*(1992)168 h. mohring , j.ranft , c.merino and c.pajares , phys.rev.*d47*(1993)4142 xin - nian wang , phys.rev.*d43*(1991)104 xin - nian wang and miklos gyulassy , phys.rev . * d44*(1991)3501 xin - nian wang and miclos gyulassy , phys.rev . * d45*(1992)844 xin - nian wang and miclos gyulassy , phys.rev.lett . * 68*(1992)1480 berndt muller and xin - nian wang , phys , rev.lett.*68*(1992)2437 xin - nian wang and miclos gyulassy , preprint duke - th-91 - 24(1991 ) xin - nian wang , preprint duke - th-92 - 33 ( 1993 ) j. ranft , phys.rev.*d37*(1988)1842 t. sjostrand , comput.phys.commun . * 39*(1986)347 t. sjostrand , comput.phys.commun . * 43*(1987)367 n. s. amelin et al.,phys.rev.*c44 * ( 1991)1541 n. s. amelin et al.,phys.rev.*c47 * ( 1993)2299 r. c. hwa , phys.rev.*d37*(1988)1830 d. r. ward _ properties of soft proton antiproton collisions _ in advances series on direction in high energy physics , vol.4 , eds.g.altarelli and l.di lella,1989 , p.85 i. m. dremin , sov.j.nucl.phys . * 33*(1981)726 g.j.alner et al . , phys.rep.*154*(1987)247 jaeger et al.,phys.rev.*d11 * ( 1975)2405 liu bo and chiang huan ching , j. phys . g.nucl . part . phys.*19*(1993)161 h.strobele et al . nucl.phys.*a525*(1991)59c h.sorge,l . a. winckelmann , h. stocker , w . greiner , z.phys.*c59*(1993)85 j. letessier , a . tounsi , u . heinz , j. sollfrank and j. rafelski , _ strangeness conservation in hot fireballs_preprint par / lpthe/92 - 27(1992 ) j. cleymans , k . redlich , h . satz and e. suhonen , _ the hadronisation of a quark - gluon plasma _ , preprint cern - th 6759/92(1992 ) helmut satz , nucl.phys.*a544*(1992)371c e. quercigh , nucl . phys.*a566*(1994)321c r. s. longacre et al . , nucl . phys.*a566*(1994)167c johana stachel et . nucl.phys.*a566*(1994)183 g. e. diebold et . al . , phys . * c48*(1993)2984 m. n. namboodiri et . al . , nucl . phys.*a566*(1994)443c r. folman , a . shor , nucl . phys.*a566*(1994)917 a.capella,u.sukhatme,c . i. tan and j. tran thanh van , phys . rep.*236*(1994)225 j. letessier , j . rafelski , a . tounsi , phys . * b323*(1994)393 j. sollfrank , m . gazdzichi , u . heinz , j . rafelski , z. phys . * c61*(1994)659 e.schnedermann,j . sollfrank and u.heinz , phys . * c48*(1993)2462 u.heinz , nucl . * a566*(1994)205 m. gazdzicki et al . , nucl . * a566*(1994)503c e. andersen et . * a566*(1994)217c e.andersen et . * a566*(1994)487c s.abatzis et . al . , nucl . phys . * a566*(1994)225c f. antinori et . al . , nucl . * a566*(1994)491c s.abatzis et . al . , nucl . * a566*(1994)499c j. p. blaizot , nucl . * a566*(1994)333c m. sarabura et al.,nucl . * a544*(1992)125c m. murray et al.,nucl . phys . * a566*(1994)515c j. barette et al.,z.phys.*c59*(1993)211 h. sorge , phys.rev.*c49*(1994)1253 h. sorge , m.berenguer,h.stocker,w.greiner , preprint * la - ur-92 - 1078*(1992 ) h. sorge , preprint * la - ur-93 - 1103*(1993 ) h. sorge , nucl.phys.*a566*(1994)633c n.s.amelin et al.,preprint * us - ft/1 - 94 * ( santiago university)(1994 ) t.alber et.al , preprint ikf - henpg/1 - 94(university of frankfurt)(1994 ) h. j. mohring et al.,phys.rev . * d47*(1993)4146 h. j. mohring et al.,phys.rev . * d47*(1993)4142 ben - hao sa and an tai , preprint lund university , may 1994(private communication ) * table 3 . * + _ average multiplicities for negative hadrons , neutral strange particle in pp , pa and aa interactions predicted by hijing model are compared with experimental data.the values for @xmath218 are theoretical values for @xmath199 and@xmath219 multiplied by a factor two and the values for@xmath220 are theoretical values for @xmath199 multiplied by a factor two . _ * table 5 . * + _ the mean multiplicities of negative pions and @xmath221 ratios(see the text for definition ) for nuclear collisions at 200 gev per nucleon .the experimental data are from na 35 collaboration(reference[140 ] ) and the nn data are from reference[32 ] . _ * table 6 . * + _ the ratios of the mean strange particle multiplicity to the mean multiplicity of negative hadrons . the values for @xmath218 and @xmath220 have the same meaning as in table 3.the experimental data are from t.alner et al . [ 140 ] _ * table 7 * + _ multiplicity ratios for negative hadrons as well as for several identified strange particles and diferent reactions at @xmath222;results from the hij.01 model are compared to dpm model[65 ] , na5[31 ] and na35[22 ] data_. * table 9 * + _ average multiplicities of @xmath13 and @xmath173 and multiplicity ratios in @xmath174 interactions extrapolated to the full phase space . the experimental data are from baechler et al.[27 ] . the results for @xmath224 interaction [ 32 ] are also included for comparison . _ 0.3 cm + 0.3 cm * fig.1 + * 0.3 cm rapidity distributions ( fig.1a ) and transverse momentum distributions(fig.1b ) for @xmath15 particles produced in @xmath136 interactions at @xmath225 . experimental data are taken from jaeger et al.[107 ] . 0.3 cm * fig.2 * + 0.3 cm rapidity distributions ( fig.2a)and transverse momentum distributions(fig.2b ) for @xmath91 particles produced in @xmath136 interactions at @xmath225 . experimental data are taken from jaeger et al.[107 ] . 0.3 cm * fig.3 * + 0.3 cm rapidity distributions for @xmath123 particles produced in minimum bias @xmath185 ( a ) interactions and central @xmath186(b),@xmath187(c ) and @xmath184 ( d ) collisions at @xmath226 per nucleon.the experimental data ( full circles ) are taken from t.alner et al.,[140 ] . the dashed histograms are theoretical values@xmath227 and the solid histograms are theoretical predictions @xmath1 multiplied by a factor of two @xmath228 . the open circles show the distributions for @xmath186 collisions reflected at @xmath229 0.3 cm * fig.4 * + 0.3 cm rapidity distributions for @xmath188 particles produced in minimum bias @xmath185 ( a ) interactions and central @xmath186(b),@xmath187(c ) and @xmath184 ( d ) collisions at @xmath226 per nucleon . the experimental data ( full circles ) are taken from t.alner et al.,[140].the dashed histograms are theoretical values@xmath227 and the solid histograms are theoretical predictions @xmath1 multiplied by a factor of two @xmath228 . the open circles show the distributions for @xmath186 collisions reflected at @xmath229 0.3 cm * fig.5 * + 0.3 cm rapidity distributions for @xmath230 particles produced in minimum bias @xmath185 ( a ) interactions and central @xmath186(b),@xmath187(c ) and @xmath184 ( d ) collisions at @xmath226 per nucleon . the experimental data ( full circles ) are taken from t.alner et al.,[140].the dashed histograms are theoretical values@xmath227 and the solid histograms are theoretical predictions @xmath1 multiplied by a factor of two @xmath228 . the open circles show the distributions for @xmath186 collisions reflected at @xmath229 0.3 cm * fig.6 * + 0.3 cm transverse kinetic energy distributions for @xmath123 particles produced in minimum bias @xmath185 ( a ) interactions and central @xmath186(b ) @xmath187(c ) and @xmath184 ( d ) collisions at @xmath226 per nucleon . the experimental data ( full circles ) are taken from t.alner et al.,[140 ] . the solid histograms are theoretical hijing values ( @xmath1 ) or @xmath231 shown in the figures.the vertical scale is given in @xmath232 . 0.3 cm * fig.7 * + 0.3 cm transverse kinetic energy distributions for @xmath188 particles produced in minimum bias @xmath185 ( a ) interactions and central @xmath186(b ) @xmath187(c ) and @xmath184 ( d ) collisions at @xmath226 per nucleon . the experimental data ( full circles ) are taken from t.alner et al.,[140 ] . the solid histograms are theoretical hijing values ( @xmath1 ) or @xmath231 shown in the figures.the vertical scale is given in @xmath232 . 0.3 cm * fig.8 * + 0.3 cm transverse kinetic energy distributions for @xmath124 particles produced in minimum bias @xmath185 ( a ) interactions and central @xmath186(b ) @xmath187(c ) and @xmath184 ( d ) collisions at @xmath226 per nucleon . the experimental data ( full circles ) are taken from t.alner et al.,[140 ] . the solid histograms are theoretical hijing values ( @xmath1 ) or @xmath231 shown in the figures.the vertical scale is given in @xmath232 . 0.3 cm * fig.9 * + 0.3 cm transverse momentum distributions for @xmath123 ( a ) , @xmath188 ( b ) and @xmath124 ( c ) particles produced in minimum bias @xmath185 interactions at @xmath226 per nucleon.the solid histograms are theoretical hijing values ( @xmath1).the vertical scale is given in @xmath233 . 0.3 cm * fig.10 * + 0.3 cm the ratio of @xmath188 to @xmath123 rapidity distributions for @xmath193 interactions at @xmath226 per nucleon . the corresponding ratio calculated using hijing model is shown by solid histogram . the experimental data [ 140 ] are represented by full circles . 0.3 cm * fig.11 * + 0.3 cm rapidity distributions for @xmath234 ( a ) and for @xmath235 ( b ) particles produced in @xmath236 interactions at @xmath226 per nucleon . histograms are hijing results . 0.3 cm * fig.12 * + 0.3 cm predicted rapidity distributions for @xmath123 ( a ) and for @xmath188 ( b ) particle produced in central @xmath202 collisions at @xmath237 0.3 cm * fig.13 * + 0.3 cm theoretical values for acceptances are given ( a ) for @xmath123 particles produced in @xmath202 collisions at @xmath237 . a bidimensional plot ( rapidity y - transverse momentum @xmath38 ) ( b ) for the same interaction . 0.3 cm * fig.14 * + 0.3 cm theoretical values for acceptances are given ( a ) for @xmath188 particles produced @xmath202 collisions at @xmath237 . a bidimensional plot ( rapidity y - transverse momentum @xmath38 ) ( b ) for the same interaction . 0.3 cm * fig.15 * + 0.3 cm theoretical values for acceptances are given ( a ) for all negatives charges produced in @xmath186 central collisions at @xmath238 . a bidimensional plot ( rapidity y - transverse momentum @xmath38 ) ( b ) for the same interaction . 0.3 cm * fig.16 * + 0.3 cm theoretical values for acceptances are given ( a ) for all negatives charges produced in @xmath201 central collisions at @xmath238 . a bidimensional plot ( rapidity y - transverse momentum @xmath38 ) ( b ) for the same interaction . 0.3 cm * fig.17 * + 0.3 cm theoretical values for acceptances are given ( a ) for all negatives charges produced in @xmath202 central collisions at @xmath237 . a bidimensional plot ( rapidity y - transverse momentum @xmath38 ) ( b ) for the same interaction . 0.3 cm * fig.18 * + 0.3 cm theoretical values for acceptances are given ( a ) for all positives charges produced in @xmath202 central collisions at @xmath237 . a bidimensional plot ( rapidity y - transverse momentum @xmath38 ) ( b ) for the same interaction .

it is a known result that given a extended modular functor where is a semisimple abelian category , we can find a structure of weakly rigid fusion category on . also if we have a structure of a weakly rigid fusion category on then from this we can define a extended modular functor . in this paper , we extend this notion of modular functor and fusion category to what we called equivariant modular functor and equivariant fusion category where is a finite group . then we establish a similar correspondence between equivariant modular functor and equivariant fusion category .

0807.0939

this paper is a continuation of @xcite and @xcite . its goal is to develop a formalism of @xmath1-equivariant modular functors , which would provide a suitable algebraic formalism for orbifold models in conformal field theory , much as usual modular functors can be used for describing various structures appearing in usual conformal field theory . we will also establish a relation of this approach to the theory of @xmath1-equivariant fusion categories as defined in @xcite ( following earlier work of turaev @xcite ) . the main result of this paper is that the notion of @xmath1-equivariant modular functor ( in genus zero ) and a structure of a @xmath1-equivariant fusion category are essentially equivalent ; precise statement is given in the main theorem of this paper . our approach only discusses topological setting , in which the main objects are oriented surfaces with boundary ( or , in @xmath1-equivariant case , @xmath1-covers of such curves ) . complex - analytic analog , which uses the language of flat connections on the moduli spaces of curves , will be discussed in a subsequent paper . it should be noted that some of the results here are parallel to the results of @xcite . however , unlike turaev , our approach is not based on 3d tqft , in which the main technical tool is presentation of 3-manifolds via surgery and using kirby moves . instead , we follow the approach suggested ( for non-@xmath1-equivariant case ) by moore and seiberg , presenting a surface as a result of gluing of `` standard '' spheres with holes , and then writing the generators and relations in the groupoid of all such presentations . throughout this paper , @xmath1 is a fixed finite group . [ geqcat ] a _ @xmath1-equvariant category _ is an abelian category @xmath0 with the following additional structure : @xmath1-*grading * : : decomposition @xmath2 where each @xmath3 is a full subcategory of @xmath0 . we will call objects @xmath4 `` @xmath5 twisted '' . in particular , objects @xmath6 will be called `` neutral '' . in physical literature , the subcategory @xmath7 is usually called the `` untwisted sector '' . * action of @xmath1 * : : for each @xmath8 , we are given a functor @xmath9 , and for each pair @xmath10 , a functorial isomorphism @xmath11 . these functorial isomorphisms must satisfy the following conditions : + 1 . @xmath12 2 . @xmath13 . this means the action of @xmath1 respects the grading . @xmath14 . here both sides are functorial isomorphism from @xmath15 ( see figure [ assg ] ) . this might be thought of as the associativity of @xmath1 action . + we will frequenlty use notation @xmath16 for @xmath17 . [ [ ht ] this definition was introduced by turaev @xcite under a differnt name . [ geqfucat ] a _ @xmath1-equivariant fusion category _ is a semisimple @xmath1 equivariant abelian category over the base field @xmath18 with the following additional structure : 1 . structure of a monoidal category such that + @xmath19 is a simple object + for any simple object @xmath20 , @xmath21 + @xmath22 is a tensor functor + for @xmath23 and @xmath24 , @xmath25 . 2 . structure of rigidity : each object @xmath26 has a right dual @xmath27 , with evaluation and coevaluation map , @xmath28 and @xmath29 which satisfy the following rigidity conditions : + @xmath30 is the identity map on @xmath26 + @xmath31 is the identity map on @xmath32 + for detailed discussion on this , see e.g. @xcite . functorial isomorphism @xmath33 , satisfying the same compatibility conditions as in the absence of @xmath1 . for detailed discussion see @xcite . here , for readers convenience , we list these conditions : + @xmath34 + @xmath35 + @xmath36 + in addition to the above reations , we also require + @xmath37 4 . a collection of functorial isomorphisms , @xmath38 for every @xmath4 and @xmath39 . this is similar to the braiding in the @xmath40 case but with the addtion of the twist . these functorial isomorphisms must satisfy an analog of two hexagon axioms . the first hexagon axiom is shown in figure [ hexagon_fusion ] ; the other one is similar but with @xmath41 replaced by @xmath42 . + @xmath43^{{\alpha } } & { } ^hv\otimes ( u\otimes w ) \ar[dr]^{\operatorname{id}\otimes r_{u , w}}\\ ( u\otimes v)\otimes w \ar[ur]^{r_{u , v}\otimes\operatorname{id}}\ar[dr]^{\alpha } & & & { } ^hv\otimes ( { } ^hw\otimes u)\\ & u\otimes ( v\otimes w ) \ar[r]^{r_{u , v\otimes w } } & { } ^h(v\otimes w)\otimes u \ar[ur]_{\alpha}}\ ] ] the definition immediately implies that @xmath44 and if @xmath45 then @xmath46 . also , since in a rigid monoidal category the unit object and dual is unique up to a unique isomorphism , we have canonical identification @xmath47 from now on we will refer to the isomorphisms described above : associativity , unit , braiding , rigidity , @xmath48 morphism , @xmath1-action , and their compositions as canonical morphisms . we will omit these canonical morphisms in the formulas , writing , e.g , @xmath49 rather than @xmath50 . thus all formulas and identities only make sense after the insertion of appropriate canonical morphisms . pedantic readers may complete all computations by inserting appropriate canonical morphisms . as in the @xmath51 case , existence of morphism @xmath52 is equivalent to existence of the twist @xmath53 . hence we have the following lemma . [ l : theta ] let @xmath0 be a @xmath1-equivariant fusion category . then one can define a collection of functorial morphisms @xmath54 for @xmath55 which satisfy the following conditions ( here @xmath56 and @xmath57 ) : @xmath58 conversely , if we have @xmath59 satisfying the above condition then we can recover @xmath48 from this @xmath59 , @xmath41 , and the monoidal structure . the proof is completely parallel to the @xmath51 case discussed in ( * ? ? ? * section 2.2 ) . details of @xmath1equivarint case can be found in @xcite . let @xmath0 be a @xmath1-equivariant category . an object @xmath60 is called a weak dual of @xmath26 if @xmath61 in other words , the functor @xmath62 is represented by the object @xmath60 . in this case , we will denote @xmath60 by @xmath32 . obviously , in the case of a @xmath1 equivariant fusion category the usual right dual of an object is also a weak dual . a @xmath1-equivariant category @xmath0 is called a @xmath1-equivariant weakly fusion category if it satisfies all the conditions of a @xmath1-equivariant fusion category except for the rigidity condition . instead of the condition that each object has a right dual we require that each object has a weak dual of course , a @xmath1-equivariant fusion category is also a @xmath1-equivariant weakly fusion category but the converse is not true . the main topological object of our study is the notion of a @xmath1-cover of a surface . detailed description of them is given in the @xcite . for readers convenience , we recall basic definitions here . [ d : exsurf ] an _ extended surface _ is a compact , smooth , oriented , closed surface @xmath63 ( not necessarily connected ) , possibly with boundary and with a choice of a distinguished ( marked ) point on each of its boundary components . we denote by @xmath64 the set of the boundary components of @xmath65 . so an extended surface will be denoted by @xmath66 where @xmath67 is the choice of marked point on the @xmath68 th boundary component . [ d : gcover ] a @xmath1 cover of @xmath69 is a pair @xmath70 where @xmath71 is a principal @xmath1-cover ( possibly not connected ) and @xmath72 are choice of points in the fiber of @xmath67 : @xmath73 for all @xmath74 . for brevity , we will usually denote a @xmath1-cover just by @xmath75 , suppressing all other data . note that are @xmath1-covers are not required to be connected but are required to be unbranched . one can easily define the notion of a morphism between two @xmath1-covers ; also , since each @xmath1-cover comes with a choice of a marked points , we can define , for each boundary component @xmath68 of @xmath65 , the monodromy @xmath76 ( see , e.g. , @xcite for details ) . [ l : glue ] let @xmath77 be a @xmath1-cover , and let @xmath78 be two boundary components . let @xmath79 be an orientation reversing homeomorphism of boundary circles such that @xmath80 it is well - known that such a homeomorphism is unique up to isotopy . then @xmath81 can be lifted to an orientation - reversing homeomorphism of boundary components of the @xmath1-cover @xmath82 such that @xmath83 iff the monodromy satisfies the following condition : @xmath84 if this condition is satisfied , then one can form a new @xmath1-cover by identifying `` gluing '' @xmath85 using @xmath86 . we will denote this new @xmath1-cover by @xmath87 as a special case , we can consider the situation where @xmath65 is disconnected : @xmath88 , and @xmath89 , @xmath90 . in this case we will use the notation @xmath91 for the result of identification , or `` gluing '' . for every @xmath92 , we define the standard sphere , @xmath93 , to be the riemann sphere @xmath94 with @xmath95 disks @xmath96 removed and with the marked points being @xmath97 , here @xmath98 . of course , we could replace these @xmath95 disks with any other @xmath95 non - overlapping disks with centers on the real line and with marked points in the lower half plane . any two such spheres are homeomorphic and the homeomorphism can be chosen canonically up to homotopy . note that the set of boundary components of the standard sphere is naturally indexed by numbers @xmath99 . we will now define the notion of standard blocks , or some `` distinguished '' @xmath1-covers of the standard sphere . let us start with the standard sphere with @xmath95 holes , @xmath93 and @xmath100 elements @xmath101 and @xmath102 in @xmath1 such that @xmath103 . make the cuts on @xmath93 as in figure [ cutsons11 ] . ] here the point @xmath104 in figure [ cutsons11 ] is the point at @xmath105 . in fact @xmath106 can be chosen to be any point on the upper hemisphere as long as it does not belong to the boundary circles . then one can easily sees that @xmath107cuts is simply connected . consider the trivial @xmath1-cover @xmath108 . now , define a @xmath1-cover of @xmath109 by gluing along @xmath110-th cut , identifying point @xmath111 on the left hand side of the @xmath110-th cut with the point @xmath112 on the right - hand side . one easily sees that this agrees with the action of @xmath1 on the fibers ( recall that @xmath1 acts by left multiplication ) ; condition @xmath113 ensures that this gluing defines a cover which is unbranched at point @xmath106 . finally , define a marked point @xmath114 on @xmath110-th boundray circle to be @xmath115 . this defines a @xmath1-cover of @xmath109 . [ d : stblock ] the @xmath1 cover of @xmath93 constructed above will be called the standard block and will be denoted by @xmath116 . note that it is only defined if @xmath117 . [ isolemma ] let @xmath118 and @xmath119 be two standard blocks . then the identity isomorphism @xmath120 can be lifted to an isomorphism @xmath121 iff there exists @xmath122 so that @xmath123 and @xmath124 for @xmath125 ; in this case , the isomorphism is unique . we denote the isomorphism @xmath121 by @xmath126 . see the paper @xcite . for the standard block @xmath127 , the monodromy @xmath128 around the @xmath110-th boundary circle is given by @xmath129 see the paper @xcite . @xmath110-th boundary circle of @xmath130 can be glued to the @xmath131-th boundary circle of @xmath132 iff @xmath133^{-1}$ ] . [ d : param ] let @xmath134 be a @xmath1 cover . a _ parameterization _ of @xmath75 an isomorphism of this @xmath1-cover with one or gluing of several standard blocks : @xmath135 it is easy to see that parameterization can be equivalenlty described by the following data : 1 . a finite set @xmath136 of closed non - intersecting curves ( `` cuts '' ) @xmath137 . 2 . a choice of marked points @xmath138 , one points for each cut @xmath139 , and a choice of lifting @xmath140 . 3 . for every connected component @xmath141 of @xmath142 , an isomorphism of @xmath1-covers @xmath143 . finally , note that if @xmath65 is an extended surface with the set of boundary components @xmath144 , then we have a natural action of the group @xmath145 on the category of @xmath1-covers of @xmath65 by changing the marked points @xmath146 : if @xmath147 , then we define @xmath148 in this section we introduce the @xmath1-equivariant analog of the notion of modular functor ; we will call such an analog a _ @xmath1-equivariant modular functor _ , or simply @xmath1-mf . let @xmath0 be a semisimple abelian @xmath1 equvariant category ; we assume that the set @xmath149 of equivalence classes of simple objects is finite . [ d : r ] let @xmath0 be a @xmath1-equivariant category . an object @xmath150 is called symmetric and @xmath1-invariant if 1 . @xmath151 2 . @xmath152 is symmetric , i.e. we have an isomorphism @xmath153 as in ( * ? ? ? * section 2.4 ) . 3 . for every @xmath1 we have an isomorphism @xmath154 ; these isomorphisms should be compatible with each other and with the symmetry isomorphism @xmath155 . as we will show later , a typical example of such an object is when @xmath0 is a @xmath1-fusion category and @xmath156 , where @xmath20 are simple objects . we will frequently use the following standard convention : if @xmath157 , then we will drop the index @xmath110 and summation from our formulas , writing , for example @xmath158 for @xmath159 . the following definition is the main definition of this paper ; it generalizes the well - known definition of the modular functor to @xmath1-equivariant case . [ d : mf ] let @xmath160 be as above . a @xmath0-extended @xmath1-equivariant modular functor ( @xmath1-mf for short ) is the following collection of data : 1 . to every @xmath1-cover @xmath77 is assigned a polylinear functor @xmath161 ( here @xmath162 is monodromy around @xmath68-th boundary component of the @xmath1-cover ) . + in other words , for every choice of objects @xmath163 attached to every boundary component of @xmath65 is assigned a finite - dimensional vector space @xmath164 , and this assignment is functorial in @xmath165 . [ d : mf-2 ] to every morphism of @xmath1-covers @xmath166 is assigned a functorial isomorphism @xmath167 , which depends only on the isotopy class of @xmath168 . d : mf-3 ] functorial isomorphisms @xmath169 , @xmath170 . [ d : mf-4 ] * gluing isomorphism : * let @xmath171 be a @xmath1-cover and @xmath172 be two boundary components such that condition holds . let @xmath173 be the surface obtained by gluing components @xmath174 of @xmath75 as in lemma [ l : glue ] . then we have a functorial isomorphism @xmath175 where @xmath176 means that we assign the symmetric object @xmath177 to boundary components @xmath174 . [ d : mf-5 ] for any @xmath1-cover @xmath178 and any @xmath179 , we have functorial isomorphisms @xmath180 the above data have to satisfy the following axioms : multiplicativity : : @xmath181 , @xmath182 . functoriality : : all isomorphisms in parts [ d : mf-3 ] , [ d : mf-4 ] , [ d : mf-5 ] above are functorial in @xmath75 . compatibility : : all isomorphisms in parts [ d : mf-3 ] , [ d : mf-4 ] , [ d : mf-5 ] above are compatible with each other . symmetry of gluing : : after the identification @xmath183 , we have @xmath184 . normalization : : @xmath185 . explicit statements of all functoriality and compatibility axioms are similar to the ones in @xmath186 case which can be found in @xcite . the only new compatibility relations are those involving @xmath187 . for the most part , they are quite obvious ; the only one which is not immediately obvious is the one involving compatibility of @xmath187 and gluing , which is given below . let @xmath75 be a @xmath1-cover and @xmath188 . assume that condition is satisfied so that we can glue @xmath189 and @xmath190 boundary components . let @xmath191 , be a collection of objects from @xmath0 and let @xmath192 be such that @xmath193 . then the following diagram commutes : @xmath194 \ar[r]^{g_{{\alpha } , { \beta } } } \ar[d]_{t_{{\mathbf{x } } } } & \tau [ \sqcup_{{\alpha},{\beta } } { \widetilde \sigma},\{{\tilde{p}_a}\ } ; \{w_a\ } ] \ar[d]_{t_{{\mathbf{x}}'}}\\ \tau[{\widetilde \sigma } , \{x_a{\tilde{p}_a}\ } ; \{{}^{x_a}w_a\ } , { } ^x{\mathcal{r}}^1 , { } ^x{\mathcal{r}}^2 ] \ar[r]_{g_{{\alpha},\beta } } & \tau [ \sqcup_{{\alpha},{\beta } } { \widetilde \sigma},\{x_a{\tilde{p}_a}\ } ; \{{}^{x_a}w_a\ } ] } \ ] ] here we assigned @xmath195 and @xmath196 to @xmath189 and @xmath197 boundary components respectively ; the bottom arrow also uses isomorphism @xmath198 ( see definition [ d : r ] ) . also note that condition @xmath193 ensures that if we can glue the @xmath189 and @xmath197 boundary components of @xmath199 then we can also glue the @xmath189 and @xmath197 boundary components of @xmath200 we leave all the other compatibility conditions , which are rather obvious , to the imagination of the readers . [ d : nondegenerate ] a @xmath1-mf is called non - degenerate if for every non - zero object @xmath201 , there exists a @xmath1-cover @xmath199 and a collection of objects @xmath202 such that @xmath203 . as in @xmath186 case , it is sometimes convenient to consider more restricted version of modular functor , in which we only allow genus 0 surfaces . [ d : mf0 ] a genus zero @xmath0 extended @xmath1-equivariant modular functor ( genus 0 @xmath1-mf for short ) consists of the same data as defined in definition [ d : mf ] except that @xmath204 is only defined for surfaces all connected components of which have genus zero , and the gluing is only defined if @xmath174 are in different connected components of @xmath65 . in this paper , we will only consider genus zero @xmath0 extended @xmath1-eqivariant modular functor ; case of positive genus will be addressed in forthcoming papers . the following is the main theorem of this paper . the rest of the paper is devoted to prove this theorem . let @xmath0 be a semisimple @xmath1-equivariant abelian category @xmath0 and let @xmath205 be a set of representative of isomorphism classes of simple objects . assume that @xmath206 . 1 . if we have a non - degenerate @xmath0-extended genus zero @xmath1-mf , then we can construct a structure of @xmath1-equivariant weakly fusion category on @xmath0 . conversely , if we have a structure of @xmath1 equivariant fusion ( or weakly fusion ) category on @xmath0 then we can define a non - degenerate @xmath0 extended genus zero @xmath1-mf . 3 . the above two constructions are inverse of each other . the above correspondence between modular functor and fusion category satisfies the following properties : 1 . the @xmath1-invariant symmetric object @xmath152 used in the definition of the mf is given by @xmath156 . 2 . let @xmath207 be the standard block as defined in definition [ d : stblock ] . then @xmath208 = \operatorname{hom}_{{{\mathcal c}}}({\mathbf{1}},{}^{h_1^{-1}}w_1 \otimes\dots \otimes { } ^{h_n^{-1}}w_n).\ ] ] 3 . let @xmath209 be the isomorphism between standard blocks given by the rotation ( see @xcite for details ) : @xmath210 then @xmath211 is given by : @xmath212 where @xmath213 . 4 . let @xmath214 be the braiding morphism between standard blocks see @xcite for details : @xmath215 then @xmath216 is given by @xmath217 where , as before , @xmath213 . note that since @xmath218 , @xmath219 . so by the definition of the braiding isomorphism of the fusion category , we need to twist @xmath220 by @xmath221 . let @xmath222 be the isomorphism between standard blocks described in lemma [ isolemma ] . then @xmath223 is given by the following formula : @xmath224 we use the fact that @xmath225 and the action of the group @xmath1 is a tensor functor . this is not the full list of all the properties enjoyed by the correspondence between modular functors and equivariant categories ; more properties will be seen from the construction . instead of proving a direct correspondence between @xmath1-equivariant modular functor and @xmath1-equivariant fusion category , we will introduce an intermediate object , which we will call @xmath1 equivariant moore - seiberg data and then established the equivalence between these three notions : @xmath226 in this section we wil introduce the intermediate object , the moore - seiberg data ( ms for short ) ; the goal of this is encoding the structure of a fusion category in terms of vector spaces @xmath227 and suitable isomorphisms between such spaces . for moore - seiberg data for @xmath51 case , see ( * ? ? ? * section 5.3 ) . here we need the extension of this concept to @xmath1-equivariant case . [ d : msdata ] let @xmath0 be a @xmath1-equivariant abelian category ( see definition [ geqcat ] ) . then @xmath1-equivariant moore - seiberg data ( @xmath228ms data for short ) is the following collection of data : conformal blocks : : [ ms1 ] for each @xmath229 and @xmath230 satisfying @xmath231 we have a functor : @xmath232 where @xmath233 denotes the category of finite - dimensional vector spaces . + * note * : we can trivially extend this functor to a functor @xmath234 by letting @xmath235 if @xmath236 , @xmath237 . @xmath1-invariance : : [ ms2 ] for each @xmath8 , we have functorial isomorphism @xmath238 satisfying @xmath239 . rotation isomorphism : : [ ms3 ] functorial isomorphism @xmath240 symmetric object : : [ ms4 ] a symmetric @xmath1-invariant object @xmath241 as in definition [ d : r ] . gluing isomorphism : : [ ms5 ] for each @xmath242 , there exist functorial isomorphism @xmath243 ( as before , @xmath244 should be understood as in ) . commutativity isomorphism : : [ ms6 ] for @xmath245 and @xmath246 , we have functorial isomorphism @xmath247 * note * : if @xmath248 then @xmath249 is zero unless @xmath250 . but then @xmath251 and the right hand side is zero unless we have @xmath252 ; thus , if one side is zero for grading reasons , then so is the other . these above data must satisfy the axioms formulated below : non - degeneracy , normalization , rotation axiom , associativity of @xmath1 , symmetry of @xmath1 , compatibility of @xmath253 , hexagon axiom , and dehn twist axiom . before formulating the axioms , it would be convenient to define certain compositions of elementary gluing , rotation , and commutativity isomorphisms as follows . generalized gluing : : for any @xmath254 , we define _ generalized gluing _ isomorphism @xmath255 as the following compoisition : @xmath256 generalized commutativity : : for any @xmath257 , we define the generalized commutativity isomorphisms @xmath258 where @xmath259 and @xmath260 , as the following composition @xmath261 ( here @xmath262 is the generalized gluing ) . + from now on we will denote by @xmath263 both the generalized commutativity isomorphisms and the usual commutativity isomorphisms . one can define even more general isomorphisms ; however , it wo nt be necessary for our purposes . now we are ready to formulate the axioms of the ms data . [ msax1 ] for each object @xmath264 , there exists an object @xmath265 so that @xmath266 . [ msax2 ] for @xmath267 @xmath268 is the identity functor ( as before , @xmath233 denotes the category of finite - dimensional vector spaces ) . [ msax3 ] let @xmath269 be two copies of @xmath152 . then the diagram in figure [ associativity_of_g ] is commutative . @xmath270_{{{\mathcal g}}\otimes\operatorname{id } } \ar[dr]^{\operatorname{id}\otimes { { \mathcal g}}}\\ \scriptstyle { { \langle}a_1,\dots , a_k , b_1,\dots , b_l , \tilde { \mathcal{r}}^2{\rangle}\otimes{\langle}\tilde { \mathcal{r}}^2,c_1,\dots , c_p{\rangle } } \ar[dr]_{{{\mathcal g } } } & & \scriptstyle{{\langle}a_1,\dots , a_k , { \mathcal{r}}^1{\rangle}\otimes{\langle}{\mathcal{r}}^2,b_1,\dots , b_l , c_1,\dots , c_p{\rangle } } \ar[dl]^{{{\mathcal g } } } \\ & \scriptstyle { { \langle}a_1,\dots , a_k , b_1,\dots , b_l , c_1,\dots , c_p{\rangle } } } \ ] ] [ msax4 ] the isomorphism of the vector spaces @xmath271 is equal to identity . [ msax5 ] again let @xmath272 and @xmath273 be the usual isomorphism between the vector spaces @xmath274 given by @xmath275 . then the diagram in figure [ ms_g_symmetry ] is commutative . @xmath276^{{\mathcal g}}\ar[d]_{z\otimes z^{-1 } } & { \langle}a_1,\dots , a_k , b_1,\dots , b_l{\rangle}\ar[dd]^{z^l}\\ { \langle}{\mathcal{r}}^1 , a_1,\dots , a_k { \rangle}\otimes { \langle}b_1,\dots , b_l , { \mathcal{r}}^2{\rangle}\ar[d ] \\ { \langle}b_1,\dots , b_l , { \mathcal{r}}^2{\rangle}\otimes { \langle}{\mathcal{r}}^1 , a_1,\dots , a_k{\rangle}\ar[r]^{{\mathcal g } } & { \langle}b_1,\dots , b_l , a_1,\dots , a_k{\rangle}}\ ] ] [ msax7 ] the functorial isomorphisms @xmath277 must be compatible with all the other isomorphisms . more precisely we have the following : 1 . @xmath253 must be compatible with rotation isomorphism , i.e. the following diagram is commutative : @xmath278^{z } \ar[d]^{\phi_g } & { \langle}w_n , w_1,\dots , w_{n-1}{\rangle}\ar[d]^{\phi_g}\\ { \langle}{}^gw_1,\dots , { } ^gw_n{\rangle}\ar[r]^{z } & { \langle}{}^gw_n , { } ^gw_1,\dots , { } ^gw_{n-1}{\rangle}}\ ] ] 2 . @xmath253 must be compatible with the gluing isomorphism , i.e. for each @xmath279 the following diagram must be commutative @xmath280^{{{\mathcal g}}}\ar[d]^{\phi_g\otimes\phi_g } & { \langle}a_1,\dots , a_k , b_1,\dots , b_l{\rangle}\ar[dd]^{\phi_g } \\ { \langle}{}^ga_1,\dots , { } ^ga_k,{}^g{\mathcal{r}}^1{\rangle}\otimes { \langle}{}^g{\mathcal{r}}^2,{}^gb_1,\dots , { } ^gb_l{\rangle}\ar[d ] \\ { \langle}{}^ga_1,\dots , { } ^ga_k,{\mathcal{r}}^1{\rangle}\otimes { \langle}{\mathcal{r}}^2,{}^gb_1,\dots , { } ^gb_l{\rangle}\ar[r]^{{{\mathcal g } } } & { \langle}{}^ga_1,\dots , { } ^ga_k , { } ^gb_1,\dots , { } ^gb_l{\rangle}}\ ] ] the diagram above uses the isomorphism @xmath281 . 3 . @xmath253 must be compatible with commutativity isomorphism , i.e. for any @xmath282 and @xmath283 the following diagram must be commutative . @xmath284^{\sigma}\ar[d]_{\phi_g } & { \langle}x , { } ^pb , a{\rangle}\ar[d]_{\phi_g}\\ { \langle}{}^gx , { } ^ga , { } ^gb{\rangle}\ar[r]^{\sigma } & { \langle}{}^gx , { } ^{gp}b , { } ^ga{\rangle}}\ ] ] [ msax8 ] for any @xmath285 and @xmath286 , the diagram below and a similar diagram with @xmath155 repalced by @xmath287 must be commutative @xmath288^{{\sigma}_{a , bc } } \ar[dr]_{{\sigma}_{a , b } } & & { \langle}x , { } ^pb,{}^p c , a{\rangle}\\ & { \langle}x , { } ^pb , a , c{\rangle}\ar[ur]_{{\sigma}_{a , c } } } \ ] ] where @xmath289 , @xmath290 are generaized braidings defined by and @xmath291 is defined as the following composition @xmath292 [ msax9 ] let @xmath282 and @xmath283 with @xmath293 . then the dehn twist axiom is the commutativity of the diagram in figure [ f : ms_dehn_twist ] . note that if @xmath294 then all the vector spaces in figure [ f : ms_dehn_twist ] will be zero and the diagram trivially commutes . @xmath295^{\sigma}\ar[dd]_z & { \langle}{}^pb , a{\rangle}\ar[dr]^z\\ & & { \langle}a , { } ^p b{\rangle}\\ { \langle}b , a{\rangle}\ar[r]^{\sigma}&{\langle}{}^qa , b{\rangle}\ar[ur]_{\varphi}}\ ] ] we will divide the proof of the main theorem into two steps , first relating the notion of @xmath1-ms data with @xmath1-equivariant fusion categories , and then relating @xmath1-ms data with the @xmath1-modular functor . in this section we do the first step , showing the equivalence of @xmath1-equivariant ( weakly ) fusion category and @xmath1-equivariant ms data . [ fums ] 1 . if we have a structure of @xmath1 equivariant ( weakly ) fusion category on @xmath0 then from this we can create a @xmath1 equivariant ms data on @xmath0 . conversely , given a @xmath1 equivariant ms data on @xmath0 , we can define a structure of @xmath1 equivariant weakly fusion category on @xmath0 . 3 . the above two constructions are inverse of each other . the proof of this theorem is given below . for the most part , it is parallel to the proof in @xmath186 case , given in ( * ? ? ? * section 5.3 ) ; thus we will only provide detailed explanations of the steps which are new to @xmath1-equivariant case . assume that @xmath0 has a structure of @xmath1 equivariant ( weakly ) fusion category . we define the @xmath1 equivariant ms data as follows conformal blocks : : let @xmath296 , where @xmath297 . then define @xmath298 @xmath253-axiom : : for @xmath8 , define @xmath299 by @xmath300 here @xmath22 is the action of the element @xmath8 . since @xmath22 is by definition a tensor functor , we have canonical isomorphisms @xmath301 and @xmath302 rotation isomorphism : : define the rotation isomorphism @xmath303 by @xmath304 + here we use the rigidity isomorphisms , balancing isomorphism @xmath305 , and the fact that in any weakly rigid fusion category we have an isomorphism @xmath306 ( see @xcite ) . symmetric object : : the symmetric object @xmath152 is defined by @xmath307 where @xmath308 are representatives of the isomorphism classes of simple objects . the fact that it is independent of the choice of representatives and symmetry of @xmath152 are shown in the same way as in @xcite . + to show @xmath1-invariance , note that since @xmath22 is a tensor functor , it must take a simple object to a simple object . thus , for every @xmath110 we can choose an isomorphism @xmath309 for some @xmath310 . now define isomorphism @xmath311 it is easy to see that this isomorphism does not depend on the choice of @xmath312 and is compatible with the symmetry @xmath313 . gluing isomorphism : : the gluing isomorphism @xmath314 is defined by @xmath315 commutativity isomorphism : : let @xmath282 and @xmath283 . define the commutativity isomorphism @xmath316 by @xmath317 then the above data satisfies the definition of @xmath1-ms data . indeed , it can easily be shown that the isomorphisms defined above satisfy all the axioms of ms - data ; the easiest way to do this is to use the technique of using appropriately marked ribbon graphs to represent morphisms in a @xmath1-equivariant fusion category ( see @xcite , @xcite ) . now we are assuming that we have a @xmath1-equivariant moore - seiberg data and from this we want to construct a structure of @xmath1 equivariant weakly fusion category on @xmath0 . this construction is parallel to the construction in the case of @xmath40 given in ( * ? ? ? * section 5.3 ) . we will construct the structure of weakly fusion category step by step as follows : duality : : define the functor @xmath318 by @xmath319 + every object @xmath26 of @xmath0 is completely determined by the functor @xmath320 . + the arguments in ( * ? ? ? * section 5.3 ) , without any changes , show that @xmath318 is an antiequivalence of categories , and that one has a canonical isomorphism @xmath321 . construction of tensor product : : as in @xcite , define tensor product functor by @xmath322 note that for @xmath323 , one has @xmath324 unless @xmath325 ; this implies that @xmath326 . + the same arguments as in ( * ? ? ? * section 5.3 ) show that this tensor product has a canonical associativity isomorphism ; moreover , we have canonical isomorphisms @xmath327 construction of unit object : : we define the unit object by @xmath328 again , the same argument as in ( * ? ? ? * section 5.3 ) , with no changes at all , shows that so defined @xmath19 is a unit object with respect to previously defined tensor product , and that we have a canonical isomorphism @xmath329 . construction of commutativity isomorphism : : let @xmath282 and @xmath330 . we define @xmath331 as the following composition : @xmath332 here @xmath263 is the commutativity isomorphism of @xmath1 equivariant ms data . + easy explicit computation shows that the hexagon axiom of fusion category is exactly equivalent to the hexagon axiom of @xmath1 equivariant ms data . construction of balancing isomorphims : : we know that having functorial isomorphism @xmath333 is equivalent to having functorial isomorpisms ( or twist ) @xmath334 for @xmath56 satisfying certain conditions ( see lemma [ l : theta ] ) . + to define @xmath335 , recall the generalized commutativity isomorphisms defined by . in particular , letting @xmath336 , we get a commutativity isomorphism @xmath337 now define the twist functor @xmath338 , @xmath339 as follows . for any @xmath340 , consider the composition + @xmath341 this gives the functorial isomorphism between the functor @xmath342 and @xmath343 which in turn gives the twist @xmath344 . + all the required properties of @xmath335 , listed in lemma [ l : theta ] , now easily follow from the properties of commutativity and associativity morphisms in teh definition of ms data . in this section we do the second step of the proof , showing the equivalence of @xmath1 equivariant genus zero modular functor and @xmath1 equivariant ms data . 1 . if we have a non - degenerate @xmath1 equivariant @xmath0 extended genus zero modular functor then we can define a @xmath1 equivariant ms data on @xmath0 2 . conversely , given a @xmath1 equivariant ms data on @xmath0 , we can define a non - degenerate @xmath1 equivariant @xmath0 extended genus zero modular functor . 3 . the above two constructions are inverse to each other . now we are assuming that we have a @xmath1 equivariant ms data on @xmath0 and from this we want to construct a @xmath1 equivariant @xmath0 extended genus zero modular functor . the construction is similar ot the one given i n @xcite in @xmath186 case : we first define the modular functor on standard blocks . since any @xmath1 cover is isomorphic to gluing of several standard blocks ( this identification is not unique ; in fact there are infinitely many parameterization of a given @xmath1 cover ) , this will give us the modular functor on parametrized @xmath1-covers . after this , we show that the modular functor spaces obtained from any two parameterizations of the same @xmath1 cover are canonically isomosrphic . note that monodromy around @xmath110th boundary component of + @xmath348 is @xmath349 , so we have @xmath350 , as required in the definition of the modular functor . note also that @xmath351 , so condition @xmath113 given by definition of standard block matches the condition required for @xmath352 to be non - zero . next , let @xmath75 be a @xmath1-cover and @xmath353 a collection of objects assigned to boundray components of @xmath65 . let @xmath354 be a parameterization of @xmath75 ( see definition [ d : param ] ) . for each connected component @xmath355 of @xmath356 , let us assign an object of @xmath0 to each boundary component of @xmath355 , by putting @xmath165 on the boundary of @xmath65 and a copy of @xmath152 on each cut ( i.e. , assigning @xmath195 to the component on one side of the cut and @xmath196 on the other side ; since @xmath152 is symmetric , choice of the side is not important ) . then we define the modular functor for parameterized @xmath1-covers by @xmath357 = \bigotimes_i \tau[s_{n_{i}}({\mathbf{g}}^i,{\mathbf{h}}^i)]\ ] ] with the choice of objects as explained above . this defines a modular functor for parameterized surfaces ; by definition , it satisfies the gluing axiom . now we need to identify the spaces @xmath358 $ ] for different parameterizations . to do so , we use the same strategy used in @xmath186 case in @xcite . namely , recall the complex @xmath359 from @xcite ; vertices of this complex are exactly different parameterizations of @xmath75 , edges are certain `` simple moves '' between parameterizations , and 2-cells describe relations . the main result of the paper @xcite is that under these basic moves and relations the complex @xmath359 is connected and simply connected . now , for any simple move @xmath360 connecting parameterizations @xmath361 , we will define a functorial isomorphism @xmath362\to \tau [ { \widetilde \sigma},\{w_a\},g]$ ] as follows . recall from @xcite that simple moves are @xmath363 ( rotation ) , @xmath364 ( braiding ) , @xmath365 ( fusion , or gluing ) , @xmath366 ( which is related to isomorphsim @xmath367 between standard blocks ) , @xmath368 ( change of marked points on a cut ) . @xmath363 move : : if @xmath168 is the parameterization @xmath369 then @xmath370 is the parameterization @xmath371 now the corresponding map between @xmath358 $ ] and @xmath372 $ ] is given by the rotation isomorphism of @xmath1-equivariant ms data which we also denoted by @xmath373 : @xmath374 @xmath364 move : : now let @xmath168 be the parameterization @xmath375 then @xmath376 is the parameterization @xmath377 then the corresponding map between @xmath358 $ ] and @xmath378 $ ] is given by the commutativity isomorphism of @xmath1 equivariant ms data which we denoted by @xmath263 : @xmath379 + since @xmath218 , we must have @xmath380 . thus from the definition of the commutativity isomorphism , we need to twist @xmath381 by @xmath221 . @xmath365 move : : let @xmath168 be the parameterization @xmath382 here we write @xmath383 , etc . for simplicity . assume additionally that @xmath384 . then @xmath385 is the parameterization given by @xmath386 where @xmath387 + then the corresponding map between @xmath358 $ ] and @xmath388 $ ] is given by the gluing isomorphism of @xmath1 equivariant ms data which we denoted by @xmath389 @xmath390 @xmath391 move : : let @xmath392 , and let @xmath168 be the following parameterization @xmath393 then @xmath394 is the parameterization @xmath395 then the corresponding map between @xmath358 $ ] and @xmath396 $ ] is given by the @xmath126 isomorphism as defined in @xmath253 axiom of @xmath1 equivariant ms data : @xmath397 @xmath368 move : : let @xmath168 be the parameterization @xmath398 where we avoid writing all @xmath399 and @xmath400 for simplicity and only write the labels for the boundary we want to glue . then @xmath368 move replaces the label @xmath401 by another label @xmath402 : @xmath403 ( geometrically , it means that we are changing the choice of marked point on the corresponding cut ) . then the corresponding map between @xmath358 $ ] and @xmath404 $ ] is given by the symmetry of @xmath152 under the action of @xmath1 , which is part of the definition of @xmath1-equivariant modular functor : @xmath405 is given by identifying both @xmath406 and @xmath407 with @xmath408 so far we have translated all our simple moves from the language of parameterizations of @xmath1-covers to the language of @xmath1-equivariant ms data . since any two parametrizations @xmath168 and @xmath5 can be connected by a sequence of simple moves ( the complex @xmath359 is connected ) , we can construct an isomorphism of the corresponding vector spaces , @xmath409 $ ] and @xmath410 $ ] . now we need to show that we get the same isomorphism of the vector spaces independent of the choice of path ( composed of simple moves ) . since the complex @xmath359 is simply connected , it is enough to show that all the basic relations or 2-cells of @xmath359 can be translated to the corresponding axioms of @xmath1-equivariant ms data . * @xmath391 relation * @xmath411 relation * @xmath363 relation * @xmath364 relation * @xmath368 relation * rotation axiom * commutativity of disjoint union * symmetry of @xmath365 move * associativity of cuts * cylinder axiom * braiding axiom * dehn twist axiom @xmath391 relation : : the relation @xmath412 corresponds to the compatibility of @xmath253 with rotation axiom . + the relation @xmath413 corresponds to the compatibility of @xmath253 with commutativity isomorphism . + the relation @xmath414 corresponds to the compatibility of @xmath253 with gluing isomorphism . + the relation @xmath415 corresponds to the @xmath253 relation of the definition of ms data . rotation axiom : : the rotation axiom , @xmath416 , of the complex @xmath359 corresponds to the rotations axiom @xmath417 of the ms data symmetry of @xmath365 move : : symmetry of the @xmath365 move in @xmath359 corresponds to the symmetry of gluing isomorphism , @xmath262 , in ms data associativity of cuts : : associativity of cuts of @xmath359 corresponds to the associativity of gluing isomorphism , @xmath262 , in ms data braiding axiom : : braiding axioms of @xmath359 corresponds to the hexagon axiom in ms data dehn twist axiom : : dehn twist axioms of @xmath359 corresponds to the dehn twist axiom in ms data . cylinder axiom : : cylinder axiom of @xmath359 follows from functoriality of all the isomorphisms in the definition of ms data . thus , we have defined , for any @xmath1-cover @xmath75 and a parametrization @xmath168 , a vector space @xmath418 $ ] and have shown that for any two prametrizations , there is a canonical isomorphism between the vector spaces @xmath358 $ ] and @xmath419 $ ] . now the same arguments as in @xmath186 case ( see @xcite ) show that this allows us to define a vector space @xmath420 $ ] , independent of parametrization , thus giving us the @xmath1-equivaraint genus zero modular functor . readers can easily show that the @xmath1 equvariant genus zero modular functor defined this way satisfies all the axioms of modular functor . conformal blocks : : given @xmath229 and @xmath421 satisfying @xmath422 , we define @xmath423 ) \right]\ ] ] where @xmath424 is the standard block defined in definition [ d : stblock ] . @xmath253 axiom : : for each @xmath8 we define a functorial isomorphism @xmath425 as the following composition : @xmath426 ) \right ] \\ & \qquad { \xrightarrow}{(\phi_g ) _ * } \tau\left [ s_{n}(gm_{1}g^{-1},\dots , gm_{n}g^{-1};g^{-1},\dots , g^{-1 } ; [ w_1,\dots , w_n ] ) \right]\\ & \qquad { \xrightarrow}{t_{g,\dots , g } } \tau \left [ s_{n}(gm_{1}g^{-1},\dots , gm_{n}g^{-1};1,\dots,1 ; [ { } ^gw_1,\dots , { } ^g w_n ] ) \right ] \end{aligned}\ ] ] here @xmath277 is the morphism between standard blocks described in lemma [ isolemma ] , and @xmath427 is as in the definition of the modular functor ( see eq . ) . rotation axiom : : the rotation isomorphism @xmath428 is given by @xmath429 , where @xmath430 is the rotation homeomorphism of the standard block ( see eq . ) . symmetric object : : the symmetric object @xmath152 directly comes from the definition of @xmath1 equivariant modular functor . gluing isomorphism : : let @xmath431 and @xmath432 , where @xmath433 and @xmath434 and they satisfy @xmath435 . then the gluing isomorphism @xmath436 is given by the @xmath437 , where @xmath438 is the homeomomorphism defined in paper @xcite . note that we must have @xmath439 . commutativity isomorphism : : the commutativity isomorphisms @xmath440 is defined as follows . recall the braiding homeomorphism ( see @xcite ) : @xmath441 then we define @xmath155 as the following composition : @xmath442 { \xrightarrow}{\tilde b _ * } \tau [ s_3(r , pqp^{-1},p ; 1 , p^{-1},1;x , b , a)]\\ & \qquad { \xrightarrow}{t_{1 , p,1 } } \tau [ s_3(r , pqp^{-1},p;1,1,1;x , { } ^pb , a ) ] \end{aligned}\ ] ] * non - degeneracy of ms data follows from the non - degeneracy of modular functor . * normalization axiom of ms data follows from the normalization axiom of modular functor . * associativity of @xmath389 of ms data follows from the associativity of cuts of the complex @xmath359 * rotation axiom of ms data follows from the rotation axiom of the complex @xmath443 . * symmetry of @xmath389 in ms data follows from the symmetry of @xmath365 move in the complex @xmath359 . * @xmath253-relation and the compatibility of @xmath253 follows from the @xmath391 relation of the complex @xmath359 and the associativity of the group multiplication in @xmath1 . * hexagon axiom of ms data follows from the braiding axiom of the complex @xmath359 . * dehn twist axiom of ms data follows from the dehn twist axiom of the complex @xmath359 . if we look at the dehn twist axiom of ms data and rewrite everything using our definition of conformal block , replacing @xmath444 by @xmath445)]$ ] etc , we get the following diagram ( as before , @xmath446 . @xmath447 ) ] \ar[r]^{\sigma}\ar[dd]_z & \tau[s_2(q , p;1,1;[{}^pb , a])]\ar[dr]_z\\ & & \tau[s_2(p , q;1,1;[a,{}^pb])]\\ \tau[s_2(q , p;1,1;[b , a])]\ar[r]^{\sigma}&\tau[s_2(p , q;1,1;[{}^qa , b])]\ar[ur]_{\varphi}}\ ] ] now it is easy to derive the dehn twist axiom of ms data from the axioms of modular functor . recall that in the complex @xmath359 , we had the following relation , which we also called `` dehn twist axiom '' : ( we assume @xmath448 is the boundary associated with @xmath449 and the label @xmath450 and @xmath190 is the boundary circle associated with @xmath451 and the label @xmath452 ; as before , @xmath293 ) @xmath453_z \ar[r]^{b_{\alpha , \beta } } & s_2(q , p;q,1)\ar[dr]^z\\ & & s_2(p , q;1,q)\\ s_2(q , p;1,1)\ar[r]^{b_{{\beta},{\alpha } } } & s_2(p , q;p,1)\ar[ur]^{{\varphi}_p } } \ ] ] the above relation is a certian relation between homeomorphsism of @xmath1-covers ; bu functoriality axiom of modular functor , this relation must also hold between the corresponding modualr functor spaces , which exactly gives us realtion of figure [ f : dehn_twist_explained ] , this proving the dehn twist axiom for ms data .

by adding a small , irrelevant four fermi interaction to the action of noncompact lattice quantum electrodynamics ( qed ) , the theory can be simulated with massless quarks in a vacuum free of lattice monopoles . simulations directly in the chiral limit of massless quarks are done with high statistics on , and lattices at a wide range of couplings with good control over finite size effects , systematic and statistical errors . the lattice theory possesses a second order chiral phase transition which we show is logarithmically trivial , with the same systematics as the nambu - jona lasinio model . the irrelevance of the four fermi coupling is established numerically . our fits have excellent numerical confidence levels . the widths of the scaling windows are examined in both the coupling constant and bare fermion mass directions in parameter space . for vanishing fermion mass we find a broad scaling window in coupling which is essential to the quality of our fits and conclusions . by adding a small bare fermion mass to the action we find that the width of the scaling window in the fermion mass direction is very narrow . only when a subdominant scaling term is added to the leading term of the equation of state are adequate fits to the data possible . the failure of past studies of lattice qed to produce equation of state fits with adequate confidence levels to seriously address the question of triviality is explained . the vacuum state of the lattice model is probed for topological excitations , such as lattice monopoles and dirac strings , and these objects are shown to be non - critical along the chiral transition line as long as the four fermi coupling is nonzero . our results support landau s contention that perturbative qed suffers from complete screening and would have a vanishing fine structure constant in the absence of a cutoff . -1 truecm -1 truecm

hep-lat0112009

simulation studies of nambu - jona lasinio models have proven to be much more quantitative than those of other field theories @xcite . in particular , the logarithmic triviality of these models has been demonstrated , although determining logarithmic singularities decorating mean field scaling laws is a daunting numerical challenge . the reason for this success lies in the fact that when one formulates these four fermi models in a fashion suitable for simulations , one introduces an auxiliary scalar field @xmath3 in order to write the fermion terms of the action as a quadratic form . in this formulation @xmath3 then acts as a chiral order parameter which receives a vacuum expectation value , proportional to the chiral condensate @xmath4 , in the chirally broken phase . most importantly , the auxiliary scalar field @xmath3 becomes the dynamical mass term in the quark propagator . the dirac operator is now not singular for quarks with vanishing bare mass and its inversion @xcite , @xcite is successful and very fast . the algorithm for nambu - jona lasinio models is `` smart '' it incorporates a potential feature of the solution of the field theory , chiral symmetry breaking and a dynamical fermion mass , into the field configuration generator . the good features of the simulation algorithm for the nambu - jona lasinio model can be generalized to lattice qcd @xcite and qed @xcite by incorporating a weak four fermi term in their actions . these generalized models now depend on two couplings , the familiar gauge coupling and a new four fermi coupling . by choosing the four fermi coupling small we can be confident that all the dynamics resides in the gauge and fermi fields and the four fermi term just provides the framework for an improved algorithm which allows us to simulate the chiral limit of massless quarks directly . we shall find a line of spontaneously broken chiral symmetry transition points in the two dimensional coupling constant parameter space of the u(1)-gauged nambu - jona lasinio model . by simulating the model at several regions along the transition line , we will see that the theory is logarithmically trivial and that the four fermi term is irrelevant in the continuum limit . our conclusions will be supported by fits with very high confidence levels . because of the irrelevance of the pure four fermi interaction , this model will make `` textbook '' qed accessible and this paper will address the classic problem of whether qed suffers from complete charge screening . our measurements will show that the theory is logarithmically trivial and the systematics of the logarithms of triviality follow those of the nambu - jona lasinio model rather than the scalar @xmath5 model as usually assumed . simulating the @xmath6 case directly has substantial advantages , both theoretical and practical . when @xmath7 is set to zero , the theory has the exact chiral symmetry of the interaction terms in the action and this forbids chiral symmetry breaking counterterms from appearing in its effective action . this simplicity can lead to a large scaling window in the direction of the gauge or four fermi coupling in the theory s parameter space . our simulation results will support this point . however , when @xmath7 is not zero , as in most past studies of lattice qed and qcd , the effective action has no protection from dangerous symmetry breaking counterterms . in fact we will find that the scaling window of the lattice theory in the @xmath7-direction is very small and this fact is responsible for the failure of past approaches to lattice qed to address the question of triviality in a straightforward , convincing fashion . in fact , @xcite claimed non - triviality for the theory while @xcite found triviality and backed up their claim further in @xcite by calculating the sign of the beta function , which is directly relevant to the question of triviality . in addition , we shall check that the algorithm used in this work generates gauge field configurations for couplings near the chiral transition line which are free of lattice artifacts , such as monopoles @xcite and dirac strings , etc . in this paper we will present data and analyses . preliminary results have already appeared in letter form @xcite , but this article will contain new data , analyses and discussions . other applications of the use of a four fermi term to speed lattice gauge theory simulations are also under development and are being applied to qcd @xcite . it is important to note that in these applications the strength of the four fermi term is weak , so it is not responsible for chiral symmetry breaking . it just acts as scaffolding which leads to an algorithm that converges efficiently in the limit of massless quarks . the dynamics resides in the gauge and fermion field interactions . this paper is organized as follows . in the next section we present the formulation of the lattice action and discuss its symmetries and general features . in the third section we test the algorithm and tune its parameters . in the next three sections we present data and analyses over a range of gauge couplings for three choices of the irrelevant four fermi coupling on @xmath1 lattices . the irrelevance of the four fermi coupling is demonstated explicitly and equation of state fits are presented which show that the theory is logarithmically trivial with the same systematics as the nambu - jona lasinio model . the confidence levels of these fits range from approximately @xmath8 to @xmath9 percent . analyses of the order parameter s susceptibility reinforce our conclusions . in the seventh section we consider simulations at nonzero bare fermion masses in order to make contact with past work on pure lattice qed . we find that subdominant scaling terms are needed to fit the data . in other words , the usual assumption that the scaling window is wide enough to address the issue of triviality by simulating the model at nonzero fermion masses and fitting to a logarithmically improved mean field form is shown to be incorrect . in section eight we present data on lattices ranging in size from @xmath0 to @xmath2 to check that our data for the chiral condensate is not influenced significantly by finite size effects for the range of couplings used in the fits . in section nine we consider measurements of lattice monopole observables to check that they are not critical at the chiral transition points as long as the bare four fermi coupling is nonzero . in section ten we discuss the possible role of lattice artifacts in simulations of pure lattice qed and address some concerns in the literature . finally , in section eleven we suggest additional work in this field . we considered the @xmath10gauged nambu jona lasinio model with four species of fermions . the lagrangian for the continuum nambu - jona lasinio model is , @xmath11 the symmetries and other properties of @xmath12 have been discussed in @xcite and we refer the reader to that and related references for details . we will be brief here and just review a few conceptually important points . the pure nambu - jona lasinio model ( eq . 1 . with @xmath13 set to zero ) has been solved at large @xmath14 by gap equation methods @xcite , and an accurate simulation study of it has been presented @xcite . the lattice action for eq . 1 reads : @xmath15 where @xmath16 where @xmath3 is an auxiliary scalar field defined on the sites of the dual lattice @xmath17 @xcite , and the symbol @xmath18 denotes the set of the 16 lattice sites surrounding the direct site @xmath19 . the factors @xmath20 are the gauge connections and @xmath21 are the staggered phases , the lattice analogs of the dirac matrices . @xmath22 is a staggered fermion field and @xmath7 is the bare fermion mass , which will be set to 0 . note that the lattice expression for @xmath23 is non - compact in the lattice field @xmath24 , while the gauge field couples to the fermion field through compact phase factors which guarantee local gauge invariance . this point will be discussed further in sec.10 below . interesting limiting cases of the above action are the pure @xmath25 nambu - jona lasinio model ( @xmath26 ) , which has a phase transition at @xmath27 @xcite and the pure lattice qed ( @xmath28 ) limit , whose chiral phase transition is near @xmath29 for four flavors @xcite , @xcite . the pure qed ( @xmath28 ) model also has a monopole percolation transition which is probably coincident with its chiral transition at @xmath30 @xcite . past simulations of this lattice model have led to contradictory results @xcite , @xcite . since the gauged nambu - jona lasinio model can be simulated at @xmath6 for all gauge couplings , the results reported here will be much more precise and decisive than those of the pure lattice qed ( @xmath28 ) limit . we scanned the 2 dimensional parameter space ( @xmath31,@xmath32 ) using the hybrid molecular dynamics algorithm tuned for four continuum fermion species @xcite and measured the chiral condensate and monopole susceptibility as a function of @xmath31 and @xmath32 . we found that as we increased @xmath32 and moved off the @xmath28 axis , the peak of the monopole susceptibility shifted from @xmath30 at @xmath28 to @xmath33 at large @xmath34 . by contrast the chiral transition point shifted to a larger @xmath31 than the monopole percolation transition for a given value of @xmath34 and became distinct from the monopole percolation point as soon as @xmath34 became nonzero , as shown in the phase diagram , fig.1 . before turning to physically interesting measurements , we should address some technical issues concerning the algorithm . unlike the hybrid monte carlo algorithm , the hybrid molecular dynamics algorithm is not exact . the molecular dynamics equations of motion can be found in the literature @xcite . in order to evolve the noisy equations of motion and generate an ensemble of field configurations , one must choose a monte carlo time step @xmath35 @xcite . the discretization errors have been exhaustively studied and it has been shown that systematic errors in observables behave as @xmath36 @xcite . therefore , we must choose @xmath35 small enough that these systematic errors are no larger than the statistical errors we will encounter . in table 1 and fig . 2 we show the order parameter @xmath3 evaluated on a @xmath0 lattice at gauge coupling @xmath37 and four fermi coupling @xmath38 . ( we write @xmath3 here as a shorthand for @xmath39 , the expectation value of the field . this is a standard notational shortcut which , hopefully , should nt lead to confusion . ) the table shows that as long as @xmath40 the systematic error in @xmath3 is negligible . the figure shows that the theoretically expected quadratic dependence of the systematic error on @xmath36 has been confirmed numerically . the error bars quoted in the table have been obtained using the usual binning techniques , so they reflect the correlations in the measurements . the last column of the table gives the number of trajectories in each data set . a trajectory here means an interval of one monte carlo time unit of the algorithm ( for @xmath41 a trajectory consists of one hundred sweeps ) . after each trajectory a single measurement of @xmath3 was made . most of our production runs were done using @xmath42 . particularly close to the critical point where these systematiic errors are most dangerous , we checked our results with runs having @xmath43 . no problems were encountered . as stated in the introduction , we made accurate measurements on the chiral critical line for many choices of couplings ( @xmath31 , @xmath32 ) and lattice sizes ranging from @xmath0 to @xmath2 . in this section we review our data collected varying @xmath44 at fixed @xmath38 on a @xmath1 lattice . a discussion and presentation of this data has appeared in @xcite , so we will be brief . the data is presented in table 2 . the columns list the average values of @xmath3 , @xmath45 which is the longitudinal susceptibility of the order parameter @xcite , @xmath46 which is the monopole percolation order parameter and @xmath47 which is the susceptibility of the monopole order parameter @xcite . the monopole observables will be discussed later . the data for the order parameter was fit to a form which could accomodate either @xmath5 or nambu - jona lasinio triviality : @xmath48 , where the parameter @xmath49 , the critical point @xmath50 , the amplitude @xmath51 and the scale @xmath52 are determined by the fitting routine . recall that @xmath5 triviality gives @xmath53 and nambu - jona lasinio triviality gives @xmath54 . for the scaling window of gauge couplings @xmath31 between @xmath55 and @xmath56 , we found the parameters @xmath57 , @xmath58 , @xmath59 and @xmath60 with a confidence level of 34 percent . the reader should consult the figures and discussison in @xcite for more detail and perspective . as will be discussed below in sec.9 , these simulations also measured topological observables for the system s vacuum and we confirmed that monopoles and related objects were critical near the chiral transition @xmath57 , @xmath38 . ( we shall see there that the monopole percolation transition is very narrow and occurs at @xmath61 for @xmath38 . ) are other fitting forms possible for this data ? this is certainly true , of course . the point we are making , however , is that log - improved mean field theory fits the data with very high confidence levels and there are compelling theoretical reasons for it . the data and the fits support the conventional wisdom that qed is a trivial field theory and that the logarithms of triviality follow the systematics of the nambu - jona lasinio model rather than the scalar @xmath5 model . this last point is different from that usually assumed . in retrospect , it is very plausible that the nambu - jona lasinio model represents the triviality of qed better than @xmath5 , but the differences between the two models have not been emphasized or appreciated in the past . let s end our discussion of @xmath3 with some examples of other fits . simple power laws are the first ones to try . for example , a fit of the form @xmath62 is expected to work rather well with @xmath63 slightly larger than @xmath64 since the nambu - jona lasinio fits have worked so well . if we choose the range of @xmath31 to extend from @xmath65 to @xmath56 , we find @xmath66 , @xmath67 , @xmath68 , but the confidence level is very poor , ( @xmath69 ) . if we accept only a smaller range of couplings closer to the critical point , @xmath31 extending from @xmath55 to @xmath56 , the quality of the fit improves ( @xmath70 , confidence level of @xmath71 percent ) while the critical index @xmath63 rises to @xmath72 . this is the trend we find in the data : powerlaw estimates of the critical index @xmath63 increase as the range of couplings is restricted closer and closer to the critical coupling . this systematic drift in the fitting results suggests that a simple power is not an adequate representation of the full data set , but is simply mocking up the logarithm of the nambu - jona lasinio fit , which has a higher confidence level and is stable as different ranges of @xmath31 are considered . in @xcite we also analyzed the susceptibility associated with @xmath3 . in mean field theory , the singular piece of the longitudinal susceptibility @xmath73 diverges at the critical point @xmath50 as @xmath74 , @xmath75 , as @xmath76 approaches zero from above in the broken phase , and as @xmath77 in the symmetric phase @xcite . the critical index @xmath78 is exactly unity in mean field theory . in logarithmically trivial models @xmath78 remains unity , but the amplitudes @xmath79 and @xmath80 develop weak logarithmic dependences @xcite . in the two component @xmath5 model , @xmath81 , while in the @xmath25 nambu - jona lasinio model , @xmath82 @xcite , where the scale @xmath52 comes from the order parameter fit . constrained linear fits to the data @xcite produced the amplitude ratio @xmath83 . since @xmath3 varies from @xmath84 to @xmath85 over the @xmath31 range @xmath55 - @xmath56 of the scaling window in the broken phase , the logarithm in the theoretical prediction of the nambu - jona lasinio model states that @xmath86 should range from @xmath87 to @xmath88 . again , the agreement between the simulation data and theory is very good . we find no support for the approximate analytic schemes discussed in @xcite which predicted that gauged @xmath89 nambu - jona lasinio models with a four fermi term with continuous chiral symmetry are nontrivial , have powerlaw critical singularities with indices that vary continuously with the couplings @xmath31 and @xmath32 . additional simulations in sec.5 and 6 below will give strong evidence for the irrelevance of the four fermi term contrary to the results of @xcite . the reader should recall that truncated @xmath89 nambu - jona lasinio models which account for only restricted sets of feynman diagrams produce nontrivial `` theories '' with critical indices that vary continuously as @xmath31 and @xmath32 are varied . for example , this occurs if only `` rainbow graphs '' of gauged nambu - jona lasinio models are summed @xcite . some of these exercises may be relevant to technicolor model building . on the basis of the work here , however , we suspect that when fermion vacuum polarization is accounted for , one would find complete charge screening and every gauged nambu - jona lasinio model based on continuum noncompact @xmath89 gauge dynamics would be trivial for all couplings . we suspect that nontriviality and lines of nontrivial field theories are aspects of truncation procedures only . we suspect , on the basis of the present work and past triviality investigations in scalar qed @xcite , that only models with dynamics beyond continuum noncompact @xmath89 gauge fields and fermions can be nontrivial and have a renormalization group fixed point at nonzero gauge coupling . an example might be afforded by @xmath89 theories with fundamental monopoles @xcite . in this section we consider new data collected varying @xmath91 at fixed @xmath90 on a @xmath1 lattice . the purpose of this series of simulations is to 1 . to verify that the four fermi coupling is irrelevant , and 2 . to accumulate more evidence that the theory is logarithmically trivial in the sense of the nambu - jona lasinio model . the data is presented in table 3 in the same format as table 2 . in fig . 3 we show the data for the chiral condensate @xmath3 , at fixed @xmath90 and variable @xmath31 . we use the same fitting procedures as used in sec.4 : @xmath48 , where the parameter @xmath49 , the critical point @xmath50 , the amplitude @xmath51 and the scale @xmath52 are determined by the fitting routine . for the scaling window of gauge couplings @xmath31 between @xmath92 and @xmath93 , we found the parameters @xmath94 , @xmath95 , @xmath96 and @xmath97 with a confidence level of 87 percent . this excellent fit is the one shown in the figure . note that eight data points for @xmath31 between @xmath98 and @xmath99 were used in the fit while the figure has two additional points at stronger coupling . those points lie slightly below the fit , are slightly outside the scaling window and show the extent of the scaling window . we plot the eight data points between @xmath31 of @xmath98 and @xmath99 and the fit as shown in fig.4 . , @xmath100 vs. @xmath101 to illustrate the importance and numerical significance of the logarithm . the dashed line is the previous fit redrawn in this format , where we have zoomed in on the scaling window for emphasis . clearly this fit is stable to further cuts on the data set since all the data points lie on the fit . we conclude that nambu - jona lasinio triviality accommodates the lattice data at @xmath90 with very good confidence levels . this success also shows the irrelevance of the four fermi term in the lattice action : the scaling law for the order parameter is the same as that at the larger @xmath32 value although the lattice parameters , such as the location of the critical point , have changed . next , in fig.5 we show the inverse of the longitudinal susceptibility of the auxiliary field @xmath3 at fixed @xmath90 and variable @xmath31 . we follow the same procedures as used in sec.4 to analyze and plot the data here . the plot picks out a critical point @xmath102 and is consistent with the mean field value of the critical index @xmath103 . the constrained linear fits determine the amplitude ratio , @xmath104 . since @xmath3 varies from @xmath105 to @xmath106 over the @xmath31 range @xmath92 - @xmath93 of the scaling window in the broken phase , the logarithms in the theoretical prediction of the nambu - jona lasinio model for the amplitude ratio predict that @xmath86 should range from @xmath87 to @xmath88 . again , the agreement between the simulation data and theory is good , but is not comparable in quality or decisiveness to our other fits . in this section we consider new data collected varying @xmath108 at fixed @xmath107 on a @xmath1 lattice . in this case the four fermi coupling is four times stronger than the data discussed in the previous section , but far too weak to cause chiral symmetry breaking in the absence of the gauge coupling . the analysis and plots here are identical to the previous discussions of @xmath38 and @xmath90 , so we will be brief . the data is presented in table 4 in the same format as table 1 . in fig.6 we show the data for the chiral condensate @xmath3 , at fixed @xmath107 and variable @xmath31 . we use the same fitting procedures as used in sec.3 : @xmath48 , where the parameter @xmath49 , the critical point @xmath50 , the amplitude @xmath51 and the scale @xmath52 are determined by the fitting routine . for the scaling window of gauge couplings @xmath31 between @xmath109 and @xmath110 , we found the parameters @xmath111 , @xmath112 , @xmath113 and @xmath114 with a confidence level of 99.9 percent . this impressive fit is the one shown in the figure . we plot the data and the fit as shown in fig.7 , @xmath100 vs. @xmath101 to illustrate the importance and numerical significance of the logartihm . the dashed line is the previous fit redrawn in this format . the success of this fit reiterates the irrelevance of the four fermi term : the scaling law for the order parameter is the same as that at the @xmath32 values of @xmath115 and @xmath116 although the lattice parameters , such as the location of the critical point , have changed . next , in fig.8 we show the inverse of the longitudinal susceptibility of the auxiliary field @xmath3 at fixed @xmath107 and variable @xmath31 . the plot picks out a critical point @xmath117 and the constrained linear fits to the data shown in the figure produced the amplitude ratio @xmath118 . which compares well to the theoretical prediction @xmath119 . again , the agreement between the simulation data and theory is good , but is not comparable in quality or decisiveness to our order parameter fits . past simulations of lattice qed had to be done at nonzero fermion mass @xcite . the standard algorithms fail to converge in the limit @xmath120 because the lattice dirac operator becomes singular in the chiral limit @xcite , @xcite . this algorithmic problem has led to indecisive results for lattice qed because of large statistical and systematic errors . it is interesting to use the algorithm of this paper to discover , assess and clarify the problems in past work in this field . we chose to do simulations at nonzero fermion mass at the critical coupling @xmath50 determined by our fits presented in the previous section . in this way we can look for the width of the scaling window in the @xmath7-direction in a particularly simple fashion . recall that at criticality the order parameter @xmath3 should scale with the fermion mass @xmath7 , an explicit symmetry breaking parameter , as @xmath121 where the critical index @xmath122 should be @xmath123 in a logarithmically trivial theory and @xmath124 , the power of the logarithm should be @xmath125 for a @xmath5 theory and should be @xmath126 for a nambu - jona lasinio model @xcite . by accumulating data over a range of small @xmath7 values , we can look for the region where eq . 6 might apply and determine the width of the scaling window . it is important to keep the number of variables and parameters to a minimum in this sort of investigation . this is the reason we work at criticality . the critical coupling has been determined to be @xmath127 in sec.4 . the data for the order parameter and its susceptibility are shown in table 5 for @xmath7 ranging from @xmath128 to @xmath129 . note that the statistics for this data set is particularly high as smaller and smaller @xmath7 values are considered and the critical point is approached . the error bars in @xmath3 recorded there account for critical slowing down which forced us to accumulate such high statistics . the statistics are at least an order of magnitude greater than those of past studies and produce @xmath3 values with errors ranging from @xmath64 percent to @xmath130 percent . we learned in past studies of the pure nambu - jona lasinio that small @xmath7 values , typically below @xmath131 , are needed to find a scaling window @xcite . however , in this case the dynamics is controlled by the gauge coupling which alone is driving chiral symmetry breaking . the four fermi coupling is tiny and is not affecting the dynamics in a numerically significant fashion . therefore , the width of the scaling window must be determined anew from the data in table 5 . in fig . 9 we plot @xmath132 vs. @xmath7 in order to assess visually the relevance of the leading logarithm result eq the data clearly pick out the value @xmath133 for the dominant powerlaw singularity of the scaling law for very small @xmath7 values , all less than @xmath131 . however , we also see that the deviations from the mean field result are numerically significant over the entire mass range shown . in fact , they are far too large to be accommodated by a weak logarithmic scaling violation as expected in eq . 6 . in fact , a fit of that form to the data ranging from @xmath134 to @xmath135 produces a huge value for the power of the logarithm , @xmath136 , and a very small confidence level of @xmath137 percent ( @xmath138 ) . therefore , the data rule out the applicability of logarithmic improved mean field scaling to describe the data at nonzero @xmath7 except for the very smallest values of @xmath7 , @xmath139 . unfortunately , most data used to study the potential triviality of qed using the conventional action employed @xmath7 values considerably larger than @xmath140 in order to run efficiently and generate sufficient statistics . typical ranges of @xmath7 have been between @xmath131 and @xmath141 @xcite and are very sensitive to data taken with @xmath142 , @xmath143 and @xmath144 . this criticism applies to all past studies of noncompact qed , for example , @xcite . it also means that the methods of analysis introduced in @xcite do not apply to this data set because those methods require data in a scaling window , controlled by a single asymptotic form . higher precision data taken at the smallest values of @xmath7 , @xmath139 , are required apparently and , in fact , larger lattices than @xmath1 might be necessary also because of the possibility of significant finite size effects . two possible explanations for the data come to mind : 1 . perhaps the real critical point is significantly different from @xmath145 as determined by our fits at @xmath146 , or 2 . perhaps subdominant singularities in the scaling law are numerically significant over this range of @xmath7 . it is easy to rule out option 1 . ignoring logarithms , the mean field equation of state reads @xmath147 . this implies that if @xmath31 were different from @xmath50 , then @xmath132 would behave as @xmath148 , and the correction term would be large for small @xmath3 , which is just the opposite of the behavior observed in fig . 9 . now consider option 2 . if a subdominant singularity contributes to the equation of state , then at criticality the relation @xmath149 should be replaced by , @xmath150 where @xmath122 should be @xmath123 and @xmath151 should be considerably larger @xcite . this hypothesis fits the data beautifully : the curved dashed line in fig . 9 shows the fit which has a confidence level of @xmath152 percent ( @xmath153 ) . in fact this fitting form can be well approximated in a fashion that is useful for practical purposes , @xmath154 eq . 8 approximates eq . 7 because the correction to the constancy of @xmath132 is less than @xmath155 percent over the range of @xmath7 values in the figure . the fit gives @xmath156 , @xmath157 , and @xmath158 . we learn several lessons from this exercise . previous simulations of pure qed at nonzero @xmath7 could not possibly have detected the logarithms of triviality decorating mean field singularities . for the present range of @xmath7 values and lattice sizes , data at nonzero @xmath7 have contributions from subdominant critical singularities which are larger numerically than logarithmic corrections to mean field theory . since we are using a new algorithm which works in the limit of massless quarks , we should be careful to monitor finite size effects . some of our data are taken very near to critical points in order to find critical indices that control continuum limits of the lattice models . at these points the model s correlation length diverges and there are potentially dangerous finite size effects which could mimic finite temperature effects , for example . we need to check that the lattice is large enough to contain correlations larger than the lattice spacing but smaller than the system s spatial extent in order to work within a scaling window where we can extract continuum features of the field theory . in table 6 we show data for @xmath3 taken for gauge couplings @xmath31 ranging from @xmath159 to @xmath160 at fixed four fermi coupling @xmath161 for @xmath0 , @xmath1 and @xmath2 lattices . the comparison of the three data sets shows coincidence everywhere except at @xmath162 between the smallest lattice @xmath0 and the other two . @xmath162 was our closest approach to the critical point in the symmetry broken phase and it appears that our @xmath1 lattice was sufficient , given our @xmath64 percent statistical errors . reliance on a @xmath0 lattice would have failed us . in the next table we show @xmath0 data for a simulation where the four fermi coupling is fixed at @xmath38 and @xmath31 ranges from @xmath159 to @xmath163 . the data consists of @xmath3 as well as monopole observables that will be discussed in a later section . comparing the @xmath3 data here to that in table 3 , we confirm the absence of finite size effects within our statistical errors . in summary , the @xmath1 data we have used to extract scaling laws from @xmath3 measurements appears free of significant finite size effects . the significance of finite size effects depends strongly on the observable being simulated . we also checked that the longitudinal susceptibility data that was used to extract the logarithmic violations of scaling in the amplitudes was not distorted by finite size effects . since these susceptibilities are determined with much larger statistical error bars , this test was less demanding . certainly the finite size effects in @xmath45 are much larger than those in @xmath3 itself . however , since @xmath3 was determined within a fraction of a percent while the statistical uncertainty in @xmath45 was typically several percent , a @xmath1 lattice was adequate for the range of couplings used in this study . noncompact lattice qed was first studied with the goal of simulating the dynamics of @xmath89 gauge fields without the monopoles that accompany compact lattice qed @xcite . it was found , however , that even the noncompact formulation has monopole - like dislocations in its lattice formulation because of the space - time cutoff itself @xcite . these dislocations can undergo a percolation transition where long range correlations develop @xcite . because of this transition , it is not obvious that simulation results in pure noncompact lattice qed reflect the physics of textook qed in which field configurations are smooth and have no topological excitations . the formulation of noncompact lattice qed with a four fermi term is free of the issues raised in @xcite . the point is , as discussed in the section 2 above , the monopole percolation transition does not coincide with the chiral transition as long as the four fermi coupling is nonzero . therefore the gauge field vacuum is free of critical dislocations at the gauge couplings of interest , so we know that we are studying a model free of topological excitations , as we wish . let s find the monopole percolation transition in the model with a fixed four fermi coupling @xmath38 . the data for the monopole concentration @xmath46 and the associated monopole percolation susceptibility @xmath47 , both defined exactly as in @xcite , are given in table 2 . in fig.10 we plot the monopole concentration against the gauge coupling and find a percolation transition at @xmath164 . we determined in sec.4 that the chiral transition occurs at considerably weaker coupling , @xmath165 , where the monopole concentration is insignificant , as we read off fig.10 . it is also informative to confirm this conclusion by considering the monopole percolation susceptility , @xmath47 . in fig.11 we plot this susceptibility against the gauge coupling and see that it appears to diverge in the vicinity of @xmath164 ( we also confirmed this impression with powerlaw fits ) . in addition , in fig.12 we plot the longitudinal susceptibility of the chiral transition and confirm that it diverges near @xmath165 , as already determined in sec.4 . the two susceptibility peaks are cleanly separated : @xmath164 vs. @xmath165 . we end this section with a minor remark about the finite size effects observed in the monopole observables . comparing table 2 and table 7 , we see that as the monopole percolation transition s critical coupling is approached , there are numerically significant differences between the @xmath0 and the @xmath1 data sets for both the concentration @xmath46 as well as its associated susceptibility @xmath47 . as expected , the percolation susceptibility @xmath47 is strongly suppressed by the lattice size near the transition . in fact , as we have discussed elsewhere @xcite , finite size scaling of the peak of the susceptibility is an effective and accurate means to measure the percolation critical indices . it would take simulations on a series of lattice sizes to carry out such a program for this model . the only point we wish to make here , however , is that the percolation and chiral transitions are well separated inside the phase diagram fig . it is interesting ( and fortunate for the success of this project ) that the finite size effects in the chiral order parameter @xmath3 are significantly smaller than those in the monopole concentration . although the major topic in this research is the behavior of the gauged nambu - jona lasinio model for @xmath167 , we will briefly discuss the present confusing state of theory and simulations at the edge of the phase diagram @xmath166 where past simulations have been carried out . as we have already emphasized , the real problem with studies at @xmath166 is that they must be done at nonzero fermion mass away from the chiral limit and this has caused several problems : 1 . the simulations become excessively slow for small @xmath7 values because the lattice dirac operator is singular in that limit . therefore , at low values of @xmath7 where the best statistics are required , the statistics of the data sets are typically the poorest . the scaling window in the @xmath7-direction is extremely narrow , so fitting forms which only account for the leading critical behavior are inadequate and misleading . attempting to go beyond leading order critical singularities in fits leads to a vast proliferation of parameters which undermines firm conclusions . another potential problem concerning the @xmath166 edge of the phase diagram concerns lattice monopoles . recall that one motivation for inventing and studying noncompact lattice qed @xcite was to make a model free of monopoles in order to understand the relation between chiral symmetry breaking and single gluon exchange . however , hands and wensley @xcite pointed out that even the the noncompact model has monopole - like lattice dislocations because of gauge invariance of the pure gauge field piece of the action and because of the lattice cutoff itself . these authors also pointed out that these lattice monopoles experience a percolation transition as the gauge coupling becomes strong and in the case of quenched simulations , the monopole percolation transition is very close to the chiral transition experienced by light fermions @xcite . this led these authors to speculate that noncompact lattice qed might not be a sound framework for studying `` textbook '' qed at strong coupling @xcite . what does this possibility mean for this paper ? since we work at @xmath167 where the critical line of monopole percolation is distinct from the chiral transition line , these lattice artifacts are not relevant to our conclusions . we believe that we have a firm theoretical and numerical grasp of gauged nambu - jona lasinio models everywhere within the phase diagram fig.1 but not along the edge @xmath166 . how could this be ? following hands and wensley , the gauge field piece of the action eq.2 is invariant under local gauge transformations defined by the group of real numbers @xmath168 , while the fermionic piece of the action , which describes the gauge invariant hopping of the fermion around the lattice , has a gauge symmetry based on phases , @xmath89 . the cutoff theory described by the pure gauge piece of the action has monopole excitations attached by dirac strings @xcite . these are singular field configurations whose actions diverge when the lattice spacing is taken to zero . they would be of no concern if it were nt for the fact that as the coupling increases they experience a percolation transition where monopole clusters develop macroscopic dimensions . since the fermions are sensitive to monopole clusters through their @xmath89 phase , hands and wensley speculated that they could affect the chiral transition in the quenched and unquenched model . this speculation could be wrong for several reasons : 1 . the underlying gauge action is just a quadratic form , so it is a perfectly solvable free field theory . a free field theory ca nt have a phase transition , as emphasized in @xcite and 2 . percolation transitions need not affect the bulk properties of the underlying field theory . many examples of this sort can be cited . these complaints can be answered in part : 1 . the phase transition of percolation is not in local observables constructed out of the gauge fields , but rather is in nonlocal matrix elements . it is not unusual in statistical mechanics to make models where non - local matrix elements experience phase transitions when the underlying local field theory has no transition itself . condensed matter physics provides many examples of enormous practical importance including , for example , the localization - delocalization transition of single electrons in background fields of varying degrees of disorder . the chiral transition is sensitive to loops of the @xmath89 phase and is of this type . 2 . since fermions flip their chirality in the presence of monopoles , it is plausible that a percolating network of monopole - like excitations can induce chiral symmetry breaking in the bulk system . there is a possibility that the @xmath169 pure qed model has qualitatively different physics than that found anywhere within the phase diagram in fig.1 . only at the edge of the diagram would the percolating monopole - like excitations be critical where chiral symmetry is broken . only there are new degrees of freedom , percolating monopoles , relevant so only there could there be a new universality class . it might be that on the edge of the phase diagram , the chiral condensate is driven by monopole percolation and the chiral transition inherits a correlation length critical index @xmath170 from the percolating network and becomes the basis for a nontrivial field theory @xcite . it has been noticed that as the number of fermions is varied , both the chiral and monopole percolation transitions move in unison @xcite . in addition , in unquenched models , such as the four flavor model on the edge of the phase diagram fig.1 , the fermions induce @xmath89 plaquette terms into the theory s action which can support conventional lattice monopoles @xcite . we have nothing to add to the pros and cons of these qualitative arguments . we hope that the physics issues brought up here could be answered by striking out in new directions and finding approaches or arguments which are more precise and quantitative . the monopole percolation picture may contain only half truths , but some of those ideas might be testable in the context of models with real monopoles , generalizations of compact @xmath89 lattice qed @xcite , perhaps . we presented numerical evidence for the triviality of textbook qed using a new algorithm which converges for massless quarks . past simulations using the action with massive quarks but no four fermi term produced controversial results . recall that @xcite claimed non - triviality for the theory while @xcite found triviality and backed up their claim further in @xcite by calculating the sign of the beta function , which is directly relevant to the question of triviality . one could calculate the theory s renormalized couplings and their rg trajectories in the chiral limit , extending the work of @xcite to a two parameter space . both the gauge and the four fermi couplings should vanish as the reciprocal of the logarithm of the ultra - violet cutoff . as discussed in @xcite this calculation has some technical challenges specific to lattices of finite extent which necessitate the extrapolation of raw lattice data to achieve physical results . it would be worthwhile to investigate improved strategies here to avoid crude , indecisive results . the high quality of the equation of state fits in sec.4 , 5 and 6 should lead to improved determinations of the renormalized couplings because the lattice critical couplings are determined with excellent precision . one could also simulate the model with the @xmath25 chiral group replaced by a continuous group so the model would have goldstone bosons even on a coarse lattice @xcite . it would then be possible to test the approach and results of @xcite more quantitatively . it would also be interesting to generalize the results of sec.7 , that a subdominant critical singularity is needed to describe the data at nonzero @xmath7 , away from the critical coupling . in other words , fit the finite @xmath7 data points of previous investigations such as @xcite to equations of state with both a dominant and subdominant singularity and check that improved confidence levels are achieved with simple hypotheses . unfortunately , there will be a proliferation of fitting parameters in such a program , so its numerical significance might be questioned . nonetheless , it would definitely be worth consideration . such a program would also influence the determination of renormalized couplings because these calculations use critical couplings inferred from equation of state fits @xcite . finally , it would be interesting to simulate compact qed with a small four fermi term and study the interplay of monopoles , charges and chiral symmetry breaking . since the @xmath171 limit of the compact model is known to have a first order transition @xcite , generalizations of the action will be needed to find a continuous transition where a continuum limit of the lattice theory might exist . since the parameter space of the generalized model would be at least three dimensional , this interesting problem would be quite challenging . this work was partially supported by nsf under grant nsf - phy96 - 05199 . s. k. is supported by the korea research foundation . l. wishes to thank the _ ect@xmath172 _ , trento , for hospitality during the final stages of this project . the simulations were done at npaci and nersc . s. kim , a. koci@xmath173 and j.b . kogut , nucl . phys . * b429 * , 407 ( 1994 ) . phys . * b220 * , 102 ( 1983 ) . s. duane , a.d . kennedy , b.j . pendleton and d. roweth , phys . lett . * b195 * , 216 ( 1987 ) . s. duane and j.b . kogut , phys . * 55 * , 2774 ( 1985 ) . s. gottlieb , w. liu , d. toussaint , r.l . renken and r.l . sugar , phys . rev . * d35*,2531 ( 1987 ) . kogut , j .- f . lagae , and d.k . sinclair , phys . rev . * d58 * , 34004 ( 1998 ) . kogut , and d.k . sinclair , hep - lat/0005007 . s. kim , j.b . kogut and m .- lombardo , phys . * b55 * , 2774 ( 2001 ) . s. hands and r. wensley , phys . . lett . * 63 * , 2169 ( 1989 ) . b. rosenstein , b. warr and s. park , phys . * 205 * , 497 ( 1991 ) . y. cohen , s. elitzur and e. rabinovici , nucl . phys . * b220 * , 102 ( 1983 ) . v. azcoiti , g. di carlo , a. galante , a.f . grillo , v. laliena , and c.e . piedrafita , phys . lett . * b353 * , 279 ( 1995 ) ; * b379 * , 179 ( 1996 ) . kogut and k.c . wang , phys.rev . * d53 * , 1513 ( 1996 ) . e. dagotto and a. koci@xmath173 and j.b . b317 * , 271 ( 1989 ) . s. duane and j.b . . phys . * b275 * [ fs17 ] , 398 ( 1986 ) . a. koci@xmath173 , j.b . kogut and m .- b398 * , 376(1993 ) ; a. koci@xmath173 and j.b . . phys . * b422 * [ fs ] , 593 ( 1994 ) . c. itzykson and j .- m . drouffe , statistical field theory ( cambridge university press , 1989 . ) v. azcoiti , g. di carlo , a. galante , a.f . grillo , v. laliena , and c.e . piedrafita , phys.lett . * b355 * , 270 ( 1995 ) . c.n leung , s.t . love and w.a . bardeen , nucl . phys.*b327 * , 649(1986 ) . m. baig , h. fort , s. kim , j.b . kogut and d.k . phys . rev . * d48 * , r2385 ( 1993 ) ; m. baig , h. fort , j.b . kogut and s. kim . phys . rev . * d51 * , 5216 ( 1995 ) . a. koci@xmath173 , j.b . kogut and s. hands . b357 * , 467 ( 1991 ) ; s. hands and j.b . kogut , nucl . phys . * b462 * , 291 ( 1996 ) . a. koci and j. kogut , nucl . b * 422 * , 593 ( 1994 ) . r. myerson , t. banks and j.b . * b129 * , 493 ( 1977 ) . s. hands , a. koci@xmath173 and j.b . kogut . 289b * , 400 ( 1992 ) . j. bartholomew , j.b . kogut , s. h. shenker , j. sloan , j. shigemitsu , d. k. sinclair and h. w. wyld . b230 * , [ fs10 ] , 222 ( 1984 ) . m. gckeler , r. horsley , p.e.l . rakow and g. schierholz , d * 53 * , 1508 ( 1996 ) . m. gockeler , r. horsley , v. linke , p. rakow , g. schierholz and h. stuben , phys . * 80 * , 4119 ( 1998 ) . e. dagotto and j.b . kogut , phys . 59 * , 617 ( 1987 ) . cdd @xmath35 & @xmath3 & @xmath174 + @xmath175 & 0.1279(6 ) & 1200 + @xmath176 & 0.1280(5 ) & 2000 + @xmath177 & 0.1284(4 ) & 2400 + @xmath178 & 0.1293(4 ) & 2800 + @xmath179 & 0.1301(3 ) & 3500 + @xmath180 & 0.1314(2 ) & 4800 + cddddd @xmath181 & @xmath3 & @xmath45 & @xmath46 & @xmath47 & @xmath174 + @xmath182 & 0.11980(7 ) & 0.1756(5 ) & 0.97676(1 ) & 0.1245(5 ) & 930 + @xmath183 & 0.11248(8 ) & 0.1843(20 ) & 0.9566(1 ) & 0.261(1 ) & 950 + @xmath184 & 0.10438(8 ) & 0.1993(50 ) & 0.9221(1 ) & 0.552(2 ) & 1030 + @xmath185 & 0.09531(9 ) & 0.2160(50 ) & 0.8644(2 ) & 1.220(5 ) & 1010 + @xmath186 & 0.08520(8 ) & 0.2621(30 ) & 0.7669(3 ) & 3.04(2 ) & 1500 + @xmath187 & 0.0738(1 ) & 0.3230(40 ) & 0.5976(7 ) & 10.3(1 ) & 1310 + @xmath93 & 0.0674(1 ) & 0.3671(30 ) & 0.463(1 ) & 25.3(6 ) & 1012 + @xmath188 & 0.0609(1 ) & 0.4301(30 ) & 0.250(2 ) & 124.0(9 ) & 1130 + @xmath189 & 0.0537(2 ) & 0.4849(40 ) & 0.0812(8 ) & 122.4(9 ) & 2701 + @xmath190 & 0.0456(2 ) & 0.6591(40 ) & 0.0338(5 ) & 69.7(5 ) & 1120 + @xmath56 & 0.0367(2 ) & 0.9356(40 ) & 0.0192(2 ) & 43.5(2 ) & 1670 + @xmath191 & & & 0.0130(2 ) & 30.2(1 ) & 810 + @xmath192 & -0.00008(9 ) & 3.360(90 ) & 0.00798(8 ) & 19.57(5 ) & 810 + @xmath193 & 0.0002(5 ) & 1.903(80 ) & 0.00681(4 ) & 16.90(2 ) & 2518 + @xmath194 & -0.00002(3 ) & 1.316(70 ) & 0.00601(5 ) & 14.87(3 ) & 960 + @xmath195 & 0.0003(4 ) & 0.919(50 ) & 0.00544(2 ) & 13.34(1 ) & 3350 + @xmath196 & -0.0007(3 ) & 0.786(40 ) & 0.00483(4 ) & 12.07(2 ) & 820 + @xmath197 & -0.0006(3 ) & 0.625(50 ) & 0.00407(4 ) & 10.28(1 ) & 1010 + @xmath198 & -0.0003(2 ) & 0.525(10 ) & 0.00303(6 ) & 9.051(9 ) & 1070 + @xmath199 & -0.0002(2 ) & 0.484(10 ) & 0.00166(6 ) & 8.185(7 ) & 1230 + @xmath200 & 0.0002(2 ) & 0.432(10 ) & 0.00074(5 ) & 7.521(6 ) & 1150 + cddddd @xmath181 & @xmath3 & @xmath45 & @xmath46 & @xmath47 & @xmath174 + @xmath183 & 0.05201(4 ) & 0.066(2 ) & 0.9479(1 ) & 0.327(1 ) & 730 + @xmath201 & 0.04932(4 ) & 0.068(1 ) & 0.9293(1 ) & 0.487(2 ) & 770 + @xmath184 & 0.04642(5 ) & 0.075(2 ) & 0.9047(2 ) & 0.729(3 ) & 660 + @xmath202 & 0.04332(5 ) & 0.074(2 ) & 0.8729(3 ) & 1.106(5 ) & 600 + @xmath185 & 0.03996(6 ) & 0.082(2 ) & 0.8314(4 ) & 1.72(1 ) & 500 + @xmath203 & 0.03644(6 ) & 0.089(3 ) & 0.7765(5 ) & 2.81(2 ) & 640 + @xmath186 & 0.03267(5 ) & 0.096(3 ) & 0.7054(6 ) & 4.82(3 ) & 960 + @xmath204 & 0.02848(7 ) & 0.110(4 ) & 0.6091(9 ) & 9.35(9 ) & 770 + @xmath187 & 0.02401(9 ) & 0.136(4 ) & 0.4751(16 ) & 24.2(9 ) & 640 + @xmath93 & 0.01872(9 ) & 0.190(5 ) & 0.2677(29 ) & 115(4 ) & 780 + @xmath188 & 0.01250(9 ) & 0.38(1 ) & 0.08801(19 ) & 127(2 ) & 960 + @xmath189 & -0.00013(64 ) & 1.62(9 ) & 0.03625(95 ) & 74.1(9 ) & 310 + @xmath190 & -0.00027(38 ) & 0.75(9 ) & 0.02333(52 ) & 51.3(7 ) & 320 + @xmath56 & 0.00009(15 ) & 0.33(4 ) & 0.01576(26 ) & 35.6(2 ) & 490 + @xmath191 & 0.00009(9 ) & 0.23(4 ) & 0.01172(15 ) & 27.9(1 ) & 660 + @xmath205 & 0.00009(9 ) & 0.19(4 ) & 0.00937(12 ) & 22.8(8 ) & 590 + @xmath192 & 0.00010(8 ) & 0.16(2 ) & 0.00779(9 ) & 19.22(6 ) & 640 + @xmath193 & 0.00010(6 ) & 0.145(9 ) & 0.00665(6 ) & 16.63(4 ) & 910 + @xmath194 & 0.00009(6 ) & 0.133(7 ) & 0.00586(5 ) & 14.67(3 ) & 790 + cddd @xmath181 & @xmath3 & @xmath45 & @xmath174 + @xmath182 & 0.2525(2 ) & 0.39(5 ) & 1500 + @xmath183 & 0.2434(2 ) & 0.43(2 ) & 1500 + @xmath184 & 0.2339(2 ) & 0.43(5 ) & 1500 + @xmath185 & 0.2237(2 ) & 0.44(5 ) & 1500 + @xmath186 & 0.2129(2 ) & 0.48(3 ) & 1500 + @xmath187 & 0.2012(2 ) & 0.51(4 ) & 1500 + @xmath188 & 0.1885(2 ) & 0.61(3 ) & 1500 + @xmath190 & 0.1751(3 ) & 0.68(3 ) & 1500 + @xmath191 & 0.1606(3 ) & 0.84(2 ) & 1500 + @xmath192 & 0.1450(3 ) & 1.04(5 ) & 1500 + @xmath194 & 0.1281(4 ) & 1.26(4 ) & 1500 + @xmath196 & 0.1095(4 ) & 1.68(4 ) & 1500 + @xmath197 & 0.0881(5 ) & 2.23(5 ) & 1500 + @xmath206 & 0.000(6 ) & 5.66(10 ) & 1500 + @xmath207 & 0.000(7 ) & 3.68(10 ) & 1500 + @xmath208 & 0.000(7 ) & 3.17(10 ) & 1500 + @xmath209 & 0.000(6 ) & 2.61(10 ) & 1500 + @xmath210 & 0.000(4 ) & 2.33(10 ) & 1500 + @xmath211 & 0.000(5 ) & 2.31(7 ) & 1500 + @xmath212 & 0.000(5 ) & 1.80(6 ) & 1500 + cddd @xmath7 & @xmath3 & @xmath45 & @xmath174 + @xmath213 & 0.0286(1 ) & 0.927(9 ) & 3400 + @xmath214 & 0.0315(1 ) & 0.779(2 ) & 3194 + @xmath215 & 0.0340(1 ) & 0.681(4 ) & 3040 + @xmath216 & 0.0361(1 ) & 0.639(5 ) & 3397 + @xmath217 & 0.0380(1 ) & 0.582(8 ) & 3207 + @xmath218 & 0.03944(9 ) & 0.538(5 ) & 2174 + @xmath219 & 0.04247(9 ) & 0.474(5 ) & 1837 + @xmath220 & 0.04890(8 ) & 0.415(5 ) & 2521 + @xmath221 & 0.05357(8 ) & 0.377(4 ) & 1740 + @xmath222 & 0.06110(9 ) & 0.325(8 ) & 1260 + @xmath223 & 0.06688(8 ) & 0.290(7 ) & 1380 + @xmath224 & 0.07160(7 ) & 0.269(7 ) & 1400 + @xmath225 & 0.07571(8 ) & 0.255(6 ) & 960 + @xmath226 & 0.07916(7 ) & 0.240(6 ) & 1020 + @xmath227 & 0.08214(8 ) & 0.231(7 ) & 1000 + @xmath228 & 0.08480(8 ) & 0.217(6 ) & 910 + @xmath229 & 0.08724(7 ) & 0.211(6 ) & 970 + @xmath182 & 0.09636(7 ) & 0.193(3 ) & 970 + @xmath187 & 0.10237(8 ) & 0.177(3 ) & 730 + cddd @xmath181 & @xmath230 & @xmath231 & @xmath232 + @xmath182 & & 0.2525(2 ) & + @xmath183 & & 0.2434(2 ) & + @xmath184 & 0.2341(4 ) & 0.2339(2 ) & + @xmath185 & 0.2239(4 ) & 0.2237(2 ) & + @xmath186 & 0.2130(4 ) & 0.2129(2 ) & + @xmath187 & 0.2013(5 ) & 0.2012(2 ) & + @xmath188 & 0.1885(5 ) & 0.1885(2 ) & + @xmath190 & 0.1747(6 ) & 0.1751(3 ) & 0.1748(4 ) + @xmath191 & 0.1606(6 ) & 0.1606(3 ) & 0.1617(5 ) + @xmath192 & 0.1451(7 ) & 0.1450(3 ) & 0.1454(3 ) + @xmath194 & 0.1281(6 ) & 0.1281(4 ) & 0.1283(4 ) + @xmath196 & 0.1089(8 ) & 0.1095(4 ) & 0.1093(4 ) + @xmath197 & 0.0866(8 ) & 0.0881(5 ) & 0.0885(4 ) + cdddd @xmath181 & @xmath3 & @xmath46 & @xmath47 & @xmath174 + @xmath182 & 0.1203(2 ) & 0.9770(1 ) & 0.123(1 ) & 1000 + @xmath183 & 0.1129(2 ) & 0.9568(2 ) & 0.261(2 ) & 1000 + @xmath184 & 0.1047(2 ) & 0.9223(4 ) & 0.551(5 ) & 1000 + @xmath185 & 0.0956(2 ) & 0.8641(6 ) & 1.21(1 ) & 1000 + @xmath186 & 0.0854(3 ) & 0.7660(10 ) & 3.10(5 ) & 1000 + @xmath187 & 0.0741(3 ) & 0.5954(22 ) & 10.6(3 ) & 1000 + @xmath188 & 0.0610(4 ) & 0.2669(49 ) & 71.5(9 ) & 1000 + @xmath190 & 0.0449(7 ) & 0.0651(16 ) & 51.0(7 ) & 1000 + @xmath191 & & 0.0291(5 ) & 26.6(2 ) & 1000 + @xmath192 & & 0.0203(6 ) & 18.4(2 ) & 1000 + @xmath194 & & 0.0152(4 ) & 14.1(1 ) & 1000 +

we implement the elliptical gauss - laguerre ( egl ) galaxy - shape measurement method proposed by bernstein & jarvis ( 2002 ) [ ] and quantify the shear recovery accuracy in weak lensing analysis . this method uses a deconvolution fitting scheme to remove the effects of the point - spread function ( psf ) . the test simulates noisy galaxy images convolved with anisotropic psfs , and attempts to recover an input shear . the tests are designed to be immune to shape noise , selection biases , and crowding . the systematic error in shear recovery is divided into two classes , calibration ( multiplicative ) and additive , with the latter arising from psf anisotropy . at s / n , the deconvolution method measures the galaxy shape and input shear to multiplicative accuracy , and suppresses of the psf anisotropy . these systematic errors increase to for the worst conditions , with poorly resolved galaxies at s / n . the egl weak lensing analysis has the best demonstrated accuracy to date , sufficient for the next generation of weak lensing surveys .

astro-ph0607062

weak gravitational lensing , the shearing of galaxy images by gravitational bending of light , is an effective tool to probe the large - scale matter distribution of the universe . it is also a means to measure the cosmological parameters by comparing observation to numerical simulations of large scale structure growth @xcite . there are many weak lensing ( wl ) surveys underway to obtain the cosmological parameters to higher precision , and in particular to probe the evolution of the dark energy by observing its effects on the evolution of matter distribution ( dls , cfhtls ) . the wl signal is very subtle , however ; it is necessary to measure these small distortions ( typical shear @xmath6 ) in the presence of optical distortions and the asymmetric point - spread - function ( psf ) of real - life imaging . the level of systematic error in the wl measurement methods are currently above the statistical accuracy expected from future wide and deep wl surveys ( pan - starrs , snap , lsst , ska ) . because there are no `` standard shear '' lenses on the sky , shear - measurement techniques are tested by applying them to artificial galaxy images and seeing if one can correctly extract a shear applied to the simulation . in most cases , the recovered shear can be written as @xmath7 . departures from the ideal @xmath8 we will term `` calibration '' or `` multiplicative '' errors and quote as percentages . deviations from the ideal @xmath9 can result from uncorrected asymmetries in the psf and optics , and will be termed `` additive errors '' or `` incomplete psf suppression . '' such tests of the most widely applied analysis method @xcite[ksb ] , find @xmath100.80.9 , but this coefficient is implementation dependent @xcite , and depends upon the characteristics of the simulated galaxies . hirata & seljak ( 2003 ) [ @xcite ] demonstrate that various psf - correction methods can produce shear measurements miscalibrated by a few % to 20% or more . heymans et al . ( 2005 ) [ shear testing programme , ( @xcite ) ] present testing of many existing shear - measurement pipelines using a common ensemble of sheared simulated images . these methods show a median calibration error of 7% , although some ( the bj02 rounding kernel method , an implementation of a ksb method , as well as the one described in this paper ) show no calibration error , to within the @xmath11 noise level of the first step tests . although the statistical accuracy in past surveys was comparable to the 7% systematics , it is expected to be well below 1% in future surveys . hence , understanding and eliminating the wl systematic errors require the most urgent attention today . in this paper , we implement the elliptical gauss - laguerre ( egl ) deconvolution method as described in bj02 , and subject it to a series of tests designed to be more stringent than any previous test of wl measurements . the deconvolution method is distinct from the @xcite method , also described bj02 , in which the anisotropic psf effects are removed using a `` rounding kernel '' instead . wl testing regimes are of two types : in end - to - end tests ( _ e.g. _ step ) , one produces simulated sky images with a full population of stars and galaxies , analyzes them with the same pipeline as one would real data , then checks the output shear for veracity . we perform here more of a dissection , in which we analyze the performance of the method one galaxy type at a time , and vary the parameters of the galaxy and psf images to determine which , if any , conditions cause the measurement to fail . while lacking the realism of an end - to - end test , this allows us to isolate and fix weaknesses . if we can demonstrate that the method succeeds under a set of conditions that will circumscribe those found on the real sky , then we can have confidence that our method is reliable , whereas end - to - end testing is reliable only to the extent that the simulated sky reproduces the characteristics of the real sky . we investigate here the performance of our egl method across the range of noise levels , degree of resolution by the psf , pixel sampling rates , galaxy ellipticity , and psf ellipticity , using both highly symmetric and asymmetric galaxy shapes . we test not only the accuracy of shear recovery , but also the accuracy of the shear uncertainty estimates . the egl method is further elaborated in 2 , while the implementation , , is detailed in 3 . the shear accuracy test procedure is described in 4 . the conditions under which the shape measurement succeeds , and the accuracy of its estimates of shear , are presented in 5 . previous dissection tests include @xcite and @xcite . the former studies the performance of several methodologies on varied galaxy and psf shapes / sizes in the absence of noise . the latter study verified its `` polar shapelet '' method to better than 1% calibration accuracy . in 6 and 7 we conclude with comparisons to other shape - measurement methodologies and tests , and draw inferences for future surveys . the task of this weak lensing methodology is to assign some _ shape _ @xmath12 to observed galaxy @xmath13 , then to derive from the ensemble @xmath14 an estimate of the applied lensing _ shear _ . more precisely , a shape analysis can only determine the _ reduced shear _ @xmath15 , where @xmath16 is the lens convergence . following bj02 , we use distortion to describe the shear , where @xmath17 ( @xmath18 for @xmath19 ) . in this paper , both the shear and the shapes are expressed as distortions ; while in other wl literatures , shear is usually expressed as @xmath20 . following bj02 , we will quantify the lensing by decomposing its magnification matrix @xmath21 into a diagonal dilation matrix @xmath22 and a unit - determinant symmetric shear matrix @xmath23 : @xmath24 where @xmath25 is the direction of the shear axis , and @xmath26 is a measure of shear . the `` conformal shear '' @xmath26 can be reparameterized as the distortion @xmath27 or the reduced shear @xmath28 . the shape must be assigned to an image of a galaxy with some surface - brightness distribution @xmath29 . initially we will ignore the effects of psf convolution on the observed image . the bj02 definition of shape is to specify roundness criteria or `` circularity tests , '' @xmath30 and @xmath31 , that operate on @xmath32 to yield one scalar for each component of the shear typically these are quadrupole moments . the object is deemed circular ( @xmath33 ) if @xmath34 . if the object is not circular , then we assign to the object the shape which yields the solutions @xmath35 = m_\times[i({\bf s}_{\boldeta}\boldtheta ) ] = 0,\ ] ] _ i.e. _ we find the shear that , when applied to the coordinate system , makes the image appear circular in that coordinate system , and declare the galaxy shape to be this shear . any circularity test will do , as long as has a unique solution ; in particular the matrix @xmath36 must be non - singular . the shape is defined by @xmath37 , keeping the position angle @xmath25 . defining shape in this way with a suitable circularity test has the virtue that the effect of a lensing distortion upon the galaxy shape is completely defined by the multiplication of shear matrices . in particular , the component - wise formulae for transformation of a shape under a shear must take the form given by @xcite : [ emap ] @xmath38 ( e_\times\delta_+ - e_+\delta_\times ) } \over { 1 + \bolde \cdot \bolddelta } } ; \ ] ] @xmath39 ( e_+\delta_\times - e_\times\delta_+ ) } \over { 1 + \bolde \cdot \bolddelta } } .\ ] ] we can take the limit of a weak shear @xmath40 , @xmath41 : [ weakmap ] @xmath42 bj02 describe ( 5 ) a scheme for optimally weighting and combining an ensemble of shapes to produce an accurate estimate of the distortion . this scheme is predicated on the assignment of shapes that transform under shear according to . hence to test the accuracy of our methodology in recovering weak lensing shear , _ we need only test that assigned shapes transform under shear as in equations ( [ weakmap ] ) . _ furthermore , the isotropy of the universe guarantees that the @xmath12 of an unlensed population will be uniformly distributed in @xmath25 , hence we need only verify that equations ( [ weakmap ] ) hold when averaged over an ensemble of galaxies with fixed unlensed @xmath43 but random orientations : [ avgweakmap ] @xmath44 here the brackets refer to averaging over the pre - lensing orientation @xmath25 . the second order term in @xmath45 vanishes , so these equations are valid to @xmath46 . we refer to this as the _ ring test _ [ fig : ringtest ] ) , since we construct an ensemble of test galaxies which form a ring in the plane , then shear them , measure their shapes , and take the mean . we also note that as a special case , we should obtain @xmath47 when there is _ no _ applied shear . when there is no psf or the psf is symmetric under 90 rotation , then this result holds for any measurement scheme that is symmetric under inversion or exchange of the @xmath48 and @xmath49 axes of images . but for an asymmetric psf , this is a stringent test of the ability of the shape - measurement technique to remove the effects of the psf from the galaxy shapes . the circularity test described in bj02 involves decomposing the pre - convolution surface brightness distribution of the galaxy , @xmath29 , into the gauss - laguerre set of orthonormal basis functions in the plane . we consider here a general set of two dimensional functions @xmath50 that are complete ( though not necessarily orthogonal ) over the plane . any set of complete functions can be transformed to a new complete set @xmath51 via @xmath52 here @xmath53 is any remapping of the sky plane that compounds a displacement @xmath54 , a shear @xmath55 , and a dilation @xmath56 . we refer to @xmath53 as the _ basis ellipse _ of the function set : if the unit circle is an isophote of @xmath57 , then @xmath58 describe the center , size , and shape of an ellipse that traces the same isophote for @xmath59 . we will also use @xmath53 to refer to this set of 5 parameters that defines an ellipse ; it will be clear from context whether we are referring to the parameter vector or the coordinate transform . for any * e * we must have @xmath60 for ( at least ) one vector of coefficients . for the function set to be complete , the vectors must be infinite - dimensional , but a real measurement will model @xmath32 with a finite subset of the basis functions . the model must be fit to the observed image plane which we denote with coordinates . the action of the atmosphere , optics , and detector will operate on @xmath29 to produce an observed surface - brightness distribution @xmath61 on the detector plane . this observation operator @xmath62 likely includes convolution by the psf and perhaps some distortion by the optics . we will assume that this operation is known and that it is linear over surface brightness , so that @xmath63 , where @xmath64 and @xmath65 are any two scalars . in this case , must also imply that @xmath66.\end{aligned}\ ] ] the observed - plane brightness is sampled at the centers of the pixels @xmath67 yielding measurements @xmath68 with uncertainties @xmath69 , assumed henceforth to be gaussian . the model which maximizes the likelihood of the decompositions ( [ ibpsi ] ) and ( [ ibphi ] ) , given the data is that which minimizes @xmath70 ^ 2 \over \sigma_p^2}\ ] ] we call this a `` forward fit '' to the galaxy image , because we are positing a distribution @xmath71 of flux on the sky , then propagating this to the detector plane , where we compare the model to the observations . our task will be to find the @xmath53 for which the @xmath72-minimizing vector satisfies the circularity constraints . if the is always constrained to be circular . we do not yet understand the significance , if any , of this distinction . ] for fixed @xmath53 , the @xmath72 minimization is a linear least - squares problem over with the usual analytic solution @xmath73 if the @xmath57 are orthogonal _ and _ the noise is stationary and white _ and _ the sampling approaches the continuum limit _ and _ the psf approaches a delta - function , then will be diagonal ( modulo some complex conjugation operations ) and the solution is very simple . these conditions are not , however , typically met by real data in particular the sampling and psf conditions so it is not safe to use orthogonality to decompose the observed data @xcite . for finite data , the solution must be done over some truncated basis set . we will assume that @xmath74 of the basis functions are being used , so that , , and have dimension @xmath74 . the @xmath72-minimizing is uniquely defined , regardless of whether the @xmath57 are complete or orthogonal , as long as @xmath75 is not singular . to follow the bj02 prescription we must define circularity tests and iterate the shear in @xmath53 until the two circularity components vanish . at the same time , we may wish to define null tests for the centroid @xmath54 and/or size @xmath56 of the galaxy , so there will be some number @xmath76 such statistics ( @xmath77 when there is a null test for each of the parameters of @xmath53 ) . any test which is linear over the true intensity @xmath29 may be expressed as a linear operation on . hence the task of the fitter is to adjust @xmath78 components of the @xmath53 parameters until the @xmath72-minimizing satisfies @xmath79 @xmath21 is a @xmath80 matrix that defines our tests for matching the basis ellipse to the center , size , and shape of the pre - convolution galaxy . once is satisfied for a chosen set of basis functions @xmath57 and any choice of the circularity - test matrix @xmath21 , we obtain a well - defined measurement of the shape ( or more generally the defining ellipse * e * ) of the pre - convolution galaxy image . since these shapes are defined by shear operations on the @xmath81 plane , they should transform according to and be amenable to all the weighting and responsivity schemes of bj02 . before describing how we solve for @xmath53 , we summarize the conceptual and practical advantages of combining a transformation - based definition of shape with forward fitting of a pre - convolution model to the observed pixels . note the fitting approach has been advocated by several authors @xcite . * the values determined in this way transform in a well - known way under lensing shear . the response of a galaxy population to lensing shear can be calculated without recourse to empirical calibration factors beyond the distribution of the values themselves . * the forward - fitting procedure can in principle work with any sort of psf , even those such as the airy function that have divergent second moments . * the forward - fitting procedure properly handles pixelization and sampling by the detector . furthermore any aliasing ambiguities will properly propagate into the covariance matrix for , and as described below can be propagated into uncertainties in @xmath53 . * pixels rendered useless by defects or cosmic rays are easily omitted from the measurement . * the method is easily extended to simultaneous fitting of multiple images with distinct psfs . the sum over pixels in is run over all pixels in all exposures of the galaxy . all exposures share the same , , and @xmath53 vectors , but have distinct observation operators @xmath62 and hence different @xmath82 . information available in the best - seeing images is properly exploited . * the method is easily adapted to the analysis of @xmath83-plane interferometric data rather than image - plane data . the desired ellipse parameters @xmath84 enter the least - squares fit non - linearly , hence some iterative scheme ( or markov chain , _ cf . _ @xcite ) is required to determine the values that meet the constraint ( [ mb0 ] ) . for a chosen @xmath53 , the determination of has the rapid solution ( [ sumpix ] ) . if the current estimate @xmath85 yields a @xmath86 that does not meet the circularity condition , the newton - raphson iteration would be @xmath87 the derivative @xmath88 follows from noting the effect of a small change @xmath89 to the basis of : @xmath90 here @xmath91 is the _ generator _ for the transformation indicated by @xmath92th parameter of @xmath53either translation , dilation , or shear . these matrices are fixed by the choice of basis functions @xmath93 . these alterations to the basis - function values can be propagated through the solution ( [ sumpix ] ) to give the perturbation to : @xmath94.\end{aligned}\ ] ] the matrix @xmath88 is apparent from this last equation . here we have taken the generator matrices @xmath91 to each be @xmath95 ; the vector @xmath96 now must be , in general , augmented to infinite dimension , and we also take @xmath97 to be @xmath98 . note that the parenthesized portion of the solution ( [ dbde1 ] ) would vanish if not for the distinction between the truncated and infinite - dimensional versions of and , since the first @xmath74 elements of @xmath99 are zero . likewise we could set the initial @xmath100 of the final term to identity if not for the truncation , in which case the transformation would become the very simple @xmath101 ( with some abuse of notation here ) . since we are using @xmath102 only to help us iterate the solution for @xmath53 , we could use this simple approximation , or extend to some order beyond @xmath74 using . we use the gauss - laguerre decomposition in our shape measurements . these are the eigenfunctions of the 2-dimensional quantum harmonic oscillator , and are most compactly expressed as complex functions indexed by two integers @xmath103 : @xmath104 the elliptical - basis versions are taken to be @xmath105 note we have tweaked the normalizations so that the flux is @xmath106 rather than making the functions orthonormal . the gauss - laguerre ( gl ) functions are still , however , a complete and orthogonal set over the plane . the functions are equivalent to the `` polar shapelets '' of @xcite . the advantages of the gl functions as our basis are : * the @xmath57 are rapidly calculable via recursion relations . * decompositions of typical ground - based psfs and exponential - disk galaxies are reasonably compact . * there are obvious choices for the circularity tests : the centroiding condition is @xmath107 ; the circularity condition is @xmath108 ; and a size - matching condition is @xmath109 . the matrix @xmath21 is very simple , _ i.e. _ it just restricts to five relevant ( real - valued ) elements . * the generator matrices @xmath91 are sparse and simple . in particular , if the truncated extends to @xmath110 , then @xmath111 vanishes for @xmath112 . this means that the augmented and in can be calculated exactly simply by extending the sums in to order @xmath113 ( in @xmath114 ) , just slightly beyond the order @xmath74 needed to calculate . * matrices for finite transformations are calculable with fast recursion relations . * convolution by a psf is also expressible as a matrix operation over the vector , if we also express the psf as a gl decomposition @xmath115 in some basis ellipse @xmath116 . this means that the @xmath82 corresponding to a chosen psf are rapidly calculable . bj02 contains details on many of these properties , and gives the generators ( 6.3.2 ) , the forms for the finite - transformation matrices ( appendix a ) , and the convolution matrices ( appendix b ) . we emphasize that other basis sets may be chosen which yield valid shape measurements , but we choose the gl decomposition for its efficiency and clarity . similarly , one need not execute the operation @xmath117 using the gl matrices ; this is just a computational convenience in our case . we follow bj02 in using @xmath108 as our definition of a circular object . it is possible that other gl coefficients carry shape information that could be used to produce lower - noise shape estimates @xcite . in fact with the formalism used below ( [ errors ] ) to estimate shape uncertainties , one can derive for each object a linear circularity test @xmath118 that is optimally sensitive to shear . we have implemented a scheme whereby such an optimally weighted combination of all the quadrupole coefficients @xmath119 is derived for each galaxy . our testing has not yet demonstrated any significant advantage to this scheme , however , so we have not pursued it further for fear that such custom estimators might produce subtle noise - rectification biases . we may produce uncertainties on the ellipse parameters @xmath120 by propagating the pixel flux uncertainties . the covariance matrix for the vector takes the usual form @xmath121 . the covariance matrix for the ellipse parameters then follows : @xmath122^t\ ] ] we may also assign a detection significance to the object . in the simplest case , where there is no psf to consider , the signal - to - noise ratio @xmath123 for detection by the optimally - sized elliptical gaussian filter is given by @xmath124 the maximization of this quantity with respect to the size , shape , and center of the elliptical - gaussian filter generates conditions that are identical to our default circularity tests . we define a significance in the more general case such as when the psf is non - trivial to be @xmath125 the variance on the flux can be obtained by contracting @xmath126 . this estimate of the significance corresponds to the signal - to - noise on the galaxy detection that would be obtained by applying a filter to the _ deconvolved _ image that is matched to the shape , location , and radial profile of the galaxy . the maximum order of the sum in ( [ nu2 ] ) determines the allowable complexity of the filter . is the c++ implementation of the methods described in the previous section . the code includes classes to represent the vectors , transformations , and covariance matrices over quantities indexed by the integer pairs @xmath127 . to fit to a single galaxy , the code requires as input : * image - format data for both the surface brightness @xmath68 and the weight @xmath128 ( _ cf . _ eq . ( [ chisq ] ) ) . the latter image should be set to zero at saturated , defective , or contaminated pixels . data / weight image pairs from multiple exposures may be submitted for simultaneous fitting of a single object . it is assumed that each image has been flat - fielded so that a source of uniform surface brightness has equal value in all pixels . * for each exposure , a map @xmath129 from the pixel coordinate system into the world coordinate system . this is typically either the distortion map into the local tangent plane to the celestial sphere , or an identity map . * for each exposure , the local sky brightness and a photometric gain factor . * if we are executing a _ deconvolution fit _ , whereby we measure the shape of the galaxy before psf convolution , then we require , for each exposure , a vector @xmath115 and basis ellipse @xmath116 describing the estimated psf at the location of the galaxy . if we are executing a _ native fit _ , whereby we are trying to measure the _ observed _ shape of the galaxy , then no psf information is required . * a choice for the initial order @xmath74 @xmath130 of the gl decomposition of the galaxy model . * lastly an initial estimate of the size , location , and shape of the galaxy , _ e.g. _ the values generated by detection software such as sextractor @xcite . if we are executing a deconvolution fit , we need to know both the size @xmath131 of the object as observed , and the size @xmath132 of the psf in which the object is observed . these are by default subtracted in quadrature to estimate the size @xmath133 of the intrinsic ( pre - seeing ) galaxy . the procedure for shape determination has the following steps : 1 . if we are executing a deconvolution fit , we choose the size of the `` source basis '' @xmath134 which is the basis ellipse for the @xmath59 that will be used to model the pre - convolution image . we select @xmath135 where @xmath136 and @xmath137 are the sizes of the pre - seeing galaxy image and the psf , respectively , and @xmath138 is a prefactor chosen by default to be 1.2 . if there are multiple exposures , we take @xmath132 to be the harmonic mean of all the exposures psf sizes . as discussed in bj02 6.2 , we expect a choice of gaussian basis @xmath139 that is larger than both the intrinsic galaxy and the psf to offer the most sensitive measure of pre - convolution quadrupole moments . during a deconvolution fit , @xmath140 is held fixed during all iterations and there is no constraint on size in the circularity test @xmath21 . the initial guess for @xmath134 is mapped from the world coordinate system into each exposure s pixel system . an elliptical mask is created for each exposure , defined by the @xmath141 contour of the elliptical gaussian . typically we select @xmath142 . pixels outside the mask are not included in the fit . larger masks reduce the possibility of degeneracy or non - positive models for high - order fits , but increase the chance of contamination by neighboring objects . if any of the exposures do not fully contain the fitting mask , the edge flag is set . if the initial centroid falls outside _ all _ the exposures , then the outofbounds flag is set and the fit is terminated . the basis ellipse @xmath116 of the psf is changed to have the same shape ( not size ) as the current estimate of the galaxy shape , by applying a finite shear transformation matrix to @xmath115 . the convolution by the psf is now expressed as a matrix operation , so we have an observed - plane image model @xmath143 c_{ij } b_j .\ ] ] the observed - plane basis ellipse @xmath144 has second moments that are the sum of the second moments of the chosen source - plane basis @xmath134 and the psf basis ellipse . the convolution matrix elements @xmath145 can be computed as in appendix b of bj02 . we now have a model for the observed data which is linear over the expansion coefficients . [ psfstep ] 5 . the linear solution for is executed by summing over all valid pixels within the masks on all exposures . if a singular value decomposition of the matrix indicates that there are poorly constrained combinations of elements , the order of the gl decomposition is reduced by one ( to a minimum of @xmath146 ) , the reducedorder flag is set , and the solution is attempted again . the matrix @xmath88 is calculated from the augmented and ( which in fact were calculated in the previous step ) as per . 7 . if the circularity test applied to is _ not _ zero within a specified tolerance , then an iteration increment @xmath147 is calculated according to . if a singular - value decomposition of @xmath88 indicates a nearly - degenerate matrix , we know that there is no sensible estimate for the next step . in other words , at least one of our circularity tests is currently ill - defined , for example no small change of basis makes the object appear more circular . this should , for example , be the case for a point source , but also arises for some very irregular or noisy galaxies . the behavior of the next step depends upon the fit type : in a native fit , we set the frozedilation flag , set @xmath56 back to its initial value , and restart the fitting process , this time holding @xmath56 fixed . if we are executing a deconvolution fit , we increment @xmath138 by 0.5 , set the raisedmu flag , and restart the fitting process , since fits to a larger basis tend to be more stable ( but less sensitive ) . if frozedilation has been set ( for a native fit ) or @xmath148 already ( for a deconvolution fit ) , and @xmath88 is still near - degenerate , then our last resort is to set the frozeshear flag , reset to its starting value , and restart the fit process while holding the shape fixed . the next iteration of the fitting process continues at step [ psfstep ] . if the circularity test is not satisfied within a chosen number of iterations , the didnotconverge flag is set and the fit is terminated . if any other matrix singularities are encountered , or if there are too few pixels to conduct the fit , the singularity flag is set , and the fit terminates . 9 . once the circularity tests are satisfied within a desired tolerance , the resultant @xmath53 and for the galaxy are available for the user . the matrix provides the covariance matrix for . the errors on the centroid , size , and ellipticity ( or whichever of these were free to vary ) are calculated using , and a detection significance can be reported using . in most cases the solution for @xmath53 is executed twice : first , a coarse fit at order @xmath146 is attempted . if this fails , the coarsefailure flag is set . if it succeeds , the coarse solution is used as the starting point for a fit to the full requested @xmath74 . another subtlety is that the pixel mask is not changed between iterations the mask maintains the shape and size specified by the initial guess ( or the result of the coarse fit ) . this should not bias the fit toward the initial guess , however , because we are fitting to the data within the mask rather than executing a sum of moments over pixels within the mask . if the size @xmath56 of the object changes dramatically during the coarse fit ( _ e.g. _ due to a very poor initial estimate from sextractor ) , then we do resize the masks and start over . if at any time the mask on an image grows to encompass too large a number of pixels , we set the toolarge flag and quit . this prevents the program from getting hung up calculating pixel sums for nonsensically large objects . the algorithm is considered to have converged to a valid shape measurement if it completes without setting of the flags frozeshear , didnotconverge , singularity , outofbounds , toolarge , or coarsefailure . in our tests , postage - stamp images are created of individual galaxies , as convolved with a known psf , pixelized , and given noise . we then extract shape and shear measurements from ensembles of such postage - stamp tests . in the real sky and in end - to - end tests which draw ( pre - lensing ) galaxy shapes at random , the uncertainties in the output shear are usually dominated by the finite variance of the mean pre - shearing shape ( shape noise ) . we construct our ensembles with zero mean shape to eliminate shape noise , leaving our testing precision dominated by the shot noise in the images . because we provide the shape - measurement algorithms with an approximate location for every simulated galaxy , our tests do not include the effects of possible selection biases @xcite . similarly the psf is always presumed to be known exactly , so we do not test for errors that would result from psf measurement or interpolation errors @xcite . effects of galaxy crowding or overlap , detector nonlinearities , intrinsic galaxy - shape correlations , or redshift determination errors @xcite are not examined here . instead we concentrate on the sources of systematic error that arise specifically from ( 1 ) the shape measurement , ( 2 ) removal of the psf effects , and ( 3 ) the shear estimation , in images with finite noise and sampling . in order to test our wl analysis methodology , we need to find the conditions under which works , and then test for any biasing in the shape measurement or shear estimation method . the following is the list of the test conditions or choices made : * native fit or deconvolution fit . all measurements are tested for cases without ( native fit ) and with ( deconvolution fit ) the presence of a psf . in a wl analysis of real sky data , native fits are used in ( 1 ) characterizing the psf from stellar images , and ( 2 ) measuring the shape of the psf - smeared galaxy , whose value is used to produce an initial estimate of the intrinsic galaxy shape . * galaxy size with respect to pixel or psf . as the pixel sampling rate decreases , it becomes harder to obtain shape information about the galaxy from the pixelized image , or for the to converge . for the native fits , the galaxy size is defined with respect to the pixel ; for the deconvolution fits , to the psf . * galaxy detection significance @xmath149 , or equivalently @xmath123 . if the background noise dominates the signal from the galaxy , is also unlikely to converge . * galaxy shape @xmath150 . the galaxies are sheared to various shapes , ranging from @xmath151 to @xmath152 ( @xmath153 axis ratio @xmath154 ) . in some cases , the major axis orientation ( parallel or diagonal to the pixel axis ) is also examined . * galaxy type . we choose a `` symmetric '' or `` asymmetric '' galaxy model , with the latter being a more stringent test of the methodology . the galaxy models are described further in [ imagemodel ] , and the shear accuracy is examined by the ring test , described in [ ringtest ] . for the deconvolution fit , we have the following additional choices for the psf : * psf type . we choose an airy disk psf as the worst possible case for a psf . an airy disk offers particular difficulties for shape - measurement algorithms , such as its non - trivial morphology and divergent second moments . * psf shape . the airy disk is either circular or elliptical ( sheared at @xmath155 ) . we look for any leakage of the psf shape into the galaxy shape measurement or the shear estimate . in this paper , the psf is always known ; we do not test for errors due to incomplete knowledge of the psf . two types of galaxies , symmetric and asymmetric , are used to test the shape measurement and shear recovery . the galaxy models , along with any shear , rotation , and dilation , are described as an analytic function over the pixel coordinates . the `` symmetric '' galaxies have either a gaussian or exponential circularly symmetric radial profile that is then sheared . all isophotes are similar ellipses , so these galaxies have unambiguous shapes , which allows direct comparison between the input and measured shape . the elliptical symmetry of these galaxies could be masking errors in the shape - measurement methodology by causing fortuitous cancellations of errors . in other words , all of their egl expansion coefficients with @xmath156 are exactly zero , so any biases that couple to these coefficients will not be tested by the symmetric galaxies . the `` asymmetric '' galaxies are designed so that they have no inversion symmetry , and their isophotal shape varies with radius ; i.e. , there is no unambiguously defined shape . the ring test is therefore needed to see if input shear is accurately recovered . these are constructed as a combination of exponential and devaucouleur ellipses , each with different flux , size , and shape , as well as a slight offset between the two centroids . the asymmetric galaxies offer a more stringent test of shear measurement , since we have broken any symmetries that could mask errors in the shape measurement . a pixelized postage stamp image is created from the analytic functions that describe the galaxy and psf models at the intended sampling rate . the centroid of the analytic functions is randomized within the pixel over multiple image realizations ( pixel phase randomization ) . the psf convolution is done either analytically if possible , or via fft of noiseless pixelized images . the poisson photon noise is then added to produce the final image to produce the desired @xmath123 for the object . we then run for shape measurement with starting parameters randomly displaced from the true values . the ring test evaluates the mean shape the weighted average of the galaxy shape components with a simplified galaxy shape distribution , as seen in figure [ fig : ringtest ] . when a shear is uniformly applied to the ensemble , the mean shape @xmath157 should equal @xmath158 , where @xmath159 ( eq . ( [ avgweakmap ] ) ) is called the responsivity . this relation then gives an estimate of the shear from the mean shape . the value of the ring test is two - fold : the expected signal is known , and the shape noise is absent . the shape noise in wl is the statistical noise due to the random distribution of shapes with finite magnitudes ; the uniform distribution of shapes in the ring test eliminates this noise . the lack of shape noise and the known expected value then allow systematic errors to be quantified . we run on objects with different size and detection significance to understand the conditions under which converges reliably , which we define to be @xmath3 rate of successful shape measurement . shear estimate accuracy tests are subsequently performed only in @xmath3 convergence conditions , to preclude any selection biases that could be generated by the non - convergence of . figure [ fig : native - contour ] shows the @xmath160 convergence contour for native fits on exponential - profile symmetric galaxies of various ellipticities . the upper right hand region delineated by the contour indicates where the fit has @xmath3 convergence . the horizontal axis indicates the sampling rate , and is expressed in terms of the `` minor axis width '' ( maw ) , defined below . the vertical axis incidates the detection significance @xmath149 of the object . we define the minor axis width for a gaussian object as its full width at half maximum ( fwhm ) along the minor axis direction . for non - gaussian objects , the `` gaussian size '' @xmath139 is defined from the size - matching condition @xmath161 ( [ gldecomposition ] ) ; an exponential profile @xmath162 has a gaussian size @xmath163 . the gaussian size along the minor axis is @xmath164 , where @xmath26 is the conformal shear ( [ shape - as - shear ] ) , so the minor axis width is @xmath165 . figure [ fig : native - contour](a ) clearly shows that for a native fit to converge reliably , @xmath123 must be @xmath166 and its minor axis width be @xmath167 pixels . the convergence contour for gaussian radial profiles is found to be identical to the exponential - profile contour . the fact that the minor axis dimension is the limiting factor for convergence implies a shear - dependent selection , since @xmath168 does not remain constant under a shear operation . the required sampling rate depends , however , on the orientation of the minor axis . figure [ fig : native - contour](a ) shows the convergence contour when the ellipse is elongated along the pixel axis , where figure [ fig : native - contour](b ) shows the case when the elongation is along the pixel diagonal . in the second case , the required sampling rate appears to decrease with increasing @xmath150 , until it settles at 2.0 pixel for @xmath169 . figure [ fig : diag - sampling ] illustrates that for objects elongated along the pixel diagonal , the effective sampling rate along the minor axis becomes @xmath170 of the pixel spacing . the gl decomposition of the psf is determined via a native fit . hence , the images must sample the psf minor axis width by more than 2.8 pixels for the to be applied successfully to individual stars . note that it is possible to increase the sampling rate by dithering for a ccd of any pixel size . for the deconvolution test , we convolve the model galaxies with a circular airy - function psf , whose characteristic width is @xmath171 . the model galaxies in this test are symmetric elliptical objects of exponential radial profile with minor axis gaussian width @xmath168 before psf convolution . the pixel size is irrelevant to the convergence rate , as long as @xmath172 pixels ( _ cf . _ [ nativeconvergence ] ) . we find that the convergence rate of is independent of the orientation of the major axis under these conditions . figure [ fig : deconv - contour ] shows the 99% convergence contour for deconvolution fits . the horizontal axis is the minor - axis resolution @xmath173 we find that high-@xmath150 galaxies are less likely to converge at low resolution than more circular galaxies . in analyses of the real sky , the dimmest and smallest galaxies are most numerous , so we draw our attention to the lower left corner of the contour in figure [ fig : deconv - contour ] . will converge reliably for objects with @xmath174 down to @xmath175 . to obtain a reliable shape measurement of marginally resolved objects ( @xmath176 ) at all shapes @xmath150 , the significance must be @xmath177 . having determined where is consistently successful in measuring a shape , we now examine the accuracy of the shape measurement and error estimate that result . when an elliptical object is used as an input , the input shape is well defined , and hence so is the measurement error . in this section we use symmetric elliptical galaxies with exponential radial profile that have an unambiguous shape @xmath178 . the shape measurement error is expressed as a conformal shear @xmath179 , where @xmath180 is the shear addition operator , equivalent to shear matrix multiplication : @xmath181 where @xmath182 is the unique rotation matrix that allows @xmath183 to be symmetric ( bj02 2.2 ) . the quantity @xmath184 is the amount of shear required to bring the input shape to the measured shape , hence describes a shear bias regardless of the intrinsic object shape . each data point is an average over @xmath185 measurements of independent realizations of pixel phase and photon noise . we then plot the averaged error components @xmath186 or @xmath187 as the measurement accuracy . the uncertainty in the mean @xmath188 component is then @xmath189 . produces error estimates @xmath190 , @xmath191 for every shape measurement from the covariance matrix @xmath192 ( [ errors ] ) . we test for the accuracy of this estimation by examining the averaged ratios @xmath193 and @xmath194 , where the average is over shape measurement trials . each panel in figure [ fig : nativeshapeerror ] show the average error @xmath195 ( @xmath196 ) as a function of the input shape @xmath150 of the symmetric exponential galaxies . we first examine the accuracy of native fits to galaxies with no psf convolution . the panels correspond the sampling rates and detection significances marked by the points inside the 99% convergence contour in figure [ fig : native - contour ] . those that are well into the @xmath3 convergence region ( the inner three panels ) show no systematic bias in the shape measurement within the uncertainty @xmath197 . for those on the @xmath160 convergence contour ( left - most and bottom panels ) , shape measurements show some systematic errors , but the worst systematic is still within @xmath198 . the shear error @xmath199 ( @xmath200 if @xmath9 ) that would result from such measurement error is @xmath201 assuming @xmath202 , where @xmath203 ( 1-\langle e^2\rangle)\ ] ] for a constant @xmath204 . the function @xmath205 is always @xmath206 , and rapidly decreases to zero for @xmath207 . hence a shape measurement systematic of @xmath198 approximately corresponds to a shear calibration error of @xmath208 . figure [ fig : nativeerrorestimate ] shows the root - mean - square ( rms ) of the actual to estimated error ratios , @xmath209 , where @xmath210 is the error estimate in the @xmath13th component as calculated by . overall , we see that the error estimate does fairly well . there is a slight tendency to overestimate the error as @xmath149 becomes low ; the @xmath210 bias has no dependence on the pixel sampling . the @xmath149 dependence is approximately @xmath211 @xmath212 . figure [ fig : dcvlshapeerror100 ] shows the accuracy of a deconvolution fit when a symmetric , exponential galaxy is convolved with an airy psf . we first discuss a high-@xmath123 case , @xmath213 . the measurements were done at various minor axis resolutions , where @xmath214 , 1 , 2 , and 5 each correspond to the columns from left to right . the panels in the top two rows show the shape error @xmath195 as a function of the input galaxy ellipticity . in the top row , the airy psf is circular ( @xmath215 ) , while in the middle row the airy psf is anisotropic , with an ellipticity of @xmath155 . the psf and galaxy ellipticity both are elongated along @xmath216 ; we find that the results do not change even if we add a non - zero @xmath217 component to either the psf or the galaxy . the major axis orientation with respect to the pixel axis is irrelevant for deconvolution fits ( _ cf . _ [ deconvconvergence ] ) . the bottom row plots the difference in average measured shape between @xmath155 and @xmath215 . if the psf anisotropy is fully suppressed in the deconvolved galaxy - shape measurement , this difference should be zero . in general , figure [ fig : dcvlshapeerror100 ] shows the following two systematic shape measurement errors : ( 1 ) regardless of the psf anisotropy , the measurement error exhibit a slight systematic when the resolution is low ( @xmath218 ) . there is a constant positive slope in @xmath186 between @xmath219 . ( 2 ) for @xmath220 , the slope turns around , eventually underestimating the shear at @xmath221 . in the region described by ( 1 ) , the shear measurement errors are approximately @xmath222 , corresponding to @xmath223 . the well - resolved galaxies ( @xmath224 ) exhibit biases in shape measurement , which are a consequence of a peculiarity of the airy function psf . we will explain this below , and describe a remedy which yields the improved performance of the points marked `` blur . '' despite these measurement systematics , however , the difference between the @xmath155 and @xmath225 measurements are less than a shear of 0.001 , equivalent to @xmath226 of the psf anisotropy , except for the poorly resolved ( @xmath214 ) and highly elliptical galaxies . hence our implementation has a psf anisotropy bias suppression of @xmath3 at @xmath213 , except for the poorly - resolved case . we now examine deconvolution fits with galaxies of lower @xmath123 , using the combinations of resolution and significance marked by the circles sampling the @xmath3 convergence region in figure [ fig : deconv - contour ] . figure [ fig : dcvlshapeerror ] shows the measurement error @xmath186 for a circular psf , and figure [ fig : dcvlshapebias ] shows the shape bias induced by a psf ellipticity of @xmath155 . the content of these figures are similar to those on the top and bottom row of figure [ fig : dcvlshapeerror100 ] , respectively . evaluated at input @xmath227 , the initial slope in @xmath186 and the value of the psf bias increase as @xmath228 as @xmath149 decreases . in both cases , there is a turn - around in the magnitude of the systematic , with a zero crossing in the region @xmath229 . a rough estimate for the shear error @xmath199 due to these shape errors ( assuming a flat @xmath150 distribution ) is 12% . figure [ fig : dcvlerrorestimate ] plots the rms error ratio of the actual to the estimated @xmath230 for deconvolution fits , and is analogous to figure [ fig : nativeerrorestimate ] . for the deconvolution , the error estimate @xmath190 does fairly well for @xmath231 , while @xmath230 @xmath232 when @xmath233 . the open - circle points in the rightmost column of panels in figure [ fig : dcvlshapeerror100 ] mark a tendency for the deconvolution procedure to underestimate the ellipticity of large objects . this can be traced to the fact that the airy function is rather poorly described by a gauss - laguerre expansion . in our default procedure , the size of the basis set of the egl expansion of the psf is made similar to the size of the psf itself ; in this case the airy function is poorly modelled at radii @xmath234 . for example the airy function has divergent second radial moment , while any egl expansion at finite order has a finite second radial moment , indicating that the egl expansion has failed to capture the large-@xmath235 behavior of the airy function . it takes gl order @xmath236 and @xmath237 to describe the first and second airy ring , respectively ; our simulations use gl order 12 to describe the psf . an equivalent statement is that , in the fourier domain , an egl expansion fails to properly describe the small-@xmath92 behavior of the airy function . this is not surprising since the airy function has a cusp at @xmath238 ( which is equivalent to having infinite second moment in real space ) . the cusp in turn results from the sharp edge of the illumination function of an unapodized circular mirror . the poor description of the airy function at large @xmath235 or small @xmath92 becomes important when trying to deconvolve images of well - resolved galaxies , because the shape information for large galaxies is carried at large @xmath235 and small @xmath92 relative to the psf . the finite egl expansion of the airy psf underestimates its circularizing effect on large galaxies , hence the deconvolved shapes are too round for large galaxies . marginally - resolved galaxies do nt have this difficulty because they carry their ellipticity information in the part of @xmath92-space where the egl expansion is a good match to the airy function . we see that this `` airy failure '' is not an intrinsic difficulty of the shape - measurement methodology , but rather stems from a poor model of the psf . this leads us to consider several possible solutions : * describe the psf using a more appropriate function set than the gauss - laguerre expansion , if one expects nearly diffraction - limited images . this would complicate numerical calculation of the operator @xmath62 from [ fits ] , but should fully restore the numerical accuracy for any psf . * when fitting well - resolved galaxies , choose a basis for the egl expansion of the psf which matches the size of the galaxy rather than the size of the psf . this puts the fitting freedom of the egl expansion in the range of @xmath92-space that is more important for the problem at hand ; the core of the airy function will be poorly fit but the wings will be better . * change the psf . in these tests we implement the last of these three options , by blurring the postage - stamp and psf images with a gaussian that is a @xmath239 of the size of the galaxy , and then apply deconvolution using the smeared image pair . the blurring decreases the size mismatch between the psf and galaxy gl bases , which improves the accuracy of the gl expansion at the size scale relevant to the large galaxy . the blurring very slightly increases the scatter of the shape measurement , but greatly reduces the bias , as seen in the figure . to summarize , the cusp at @xmath238 in the airy psf is not well fit by our default psf characterization , leading to biases in the shape measurements . this can be remedied in a number of ways without invalidating our general approach . the previous section discussed shape measurement accuracies with a well defined input shape and high degree of symmetry . however , real galaxy images do not have an unambiguous shape nor a perfect exponential radial profile . in this section , we measure the shapes of asymmetric galaxy images that have irregular isophotes ( [ imagemodel ] ) , for which there are no unambiguous definition of shape . however , in weak lensing , it is not the accuracy of the individual shape measurement , but rather the ability to recover the shear from an ensemble of shapes , that is well - defined and needs to be accurate . in this sense , the ring test provides a measure of accuracy for assigning a shape for these irregular galaxies . in the ring test , the shape @xmath150 is determined by measurements ; the same image , before pixelization , is rotated by a series of angles equally spaced between @xmath240 . each rotated image is then sheared by @xmath241 , @xmath242 or @xmath243 , then pixelized for the ring test . with @xmath244 different orientations and @xmath245 ( or more ) realizations each , the uncertainty in the mean shape is @xmath246 , where @xmath247 is the scatter in the @xmath216 measurements due to photon statistics . the ring test has no intrinsic - shape noise . we test both the native and deconvolution fit cases at a series of different significance @xmath149 , resolution @xmath248 , and galaxy ellipticity @xmath150 . the asymmetric galaxies of different shape @xmath150 are distinct combinations of exponential - disk and devaucouleur profiles ; they are not sheared versions of the same underlying shape . additionally , the shear calibration factor @xmath249 and additive bias @xmath250 can be measured with the ring test . the additive bias is the non - zero offset of the average shape when setting @xmath251 in the ring test , and is typically present when @xmath252 . the calibration factor @xmath249 is then obtained by repeating the ring test with @xmath253 . figure [ fig : nativeshear ] shows the mean shape @xmath254 as a function of the galaxy ellipticity @xmath150 for native fits . the mean shape is expected to follow the curve @xmath255 . the approximation @xmath256 , valid to @xmath257 , is more than adequate here . for the various significance ( @xmath258 , @xmath259 , or @xmath260 ) , the mean shape follows the expected curve very well . in particular , we see that there is no significant additive bias @xmath250 seen in the @xmath241 data points , to an uncertainty of @xmath261 . figure [ fig : nativeshearnorm ] plots the calibration factor @xmath249 by normalizing @xmath254 for the @xmath262 cases to the expected signal @xmath255 , @xmath263 . the calibration is well within 1% of unity at @xmath264 , but at lower detection significance , there is a systematic underestimation of the order @xmath265 . the ring test with asymmetric galaxies is similarly performed with deconvolution fits , with a circular airy psf of @xmath266 and an elliptical psf with @xmath155 . figure [ fig : dcvlshear ] shows the additive bias @xmath250 in the shape average @xmath267 when @xmath241 and @xmath155 , while figure [ fig : dcvlshearnorm ] show the calibration factor @xmath249 for @xmath268 . both figures show various cases of @xmath269 pairs , which traces the @xmath160 convergence contour from figure [ fig : deconv - contour ] . figure [ fig : dcvlshear ] ( a ) shows @xmath250 as a function of @xmath150 at various @xmath269 pairs . there is an @xmath150 dependence to the additive bias , which increases in magnitude with decreasing @xmath149 . comparing with figure [ fig : dcvlshapebias ] , we see that the additive bias is essentially identical to the psf bias found in symmetric galaxies , indicating that the psf bias is independent of the galaxy symmetry , or the orientation of the galaxy shape over the ensemble average . figure [ fig : dcvlshear ] ( b ) shows that for a given @xmath269 , the additive bias @xmath250 scales linearly with @xmath270 for reasonable @xmath270 values . figure [ fig : dcvlshearnorm ] shows the calibration factor @xmath249 for @xmath266 , @xmath268 . at @xmath213 and better , the measured shear is within @xmath271 of the input shear signal . as the significance decreases ( @xmath272 ) , there is a multiplicative bias for galaxy shapes over @xmath273 , whose magnitude is about @xmath274 at @xmath275 ( axis ratio @xmath276 ) . similar results are found for @xmath249 for the case @xmath252 , where @xmath277 , but with increased error bars due to the additional uncertainty in @xmath250 . there are limitations to with implications for wl surveys . if is to be used for shape analysis , the sampling rate must be such that the psf minor - axis fwhm is sampled by @xmath278 pixels ; otherwise , fails to converge to a unique shape ( [ nativeconvergence ] ) . similarly , the psf - smeared objects which produce useful shapes must have a minimum resolution that depends on the significance ( or vice - versa ) in order to converge to a measurement of the pre - seeing galaxy shape . if we require shape measurements of all values of @xmath150 to be successful , then @xmath279 is required at @xmath280 , or @xmath281 at @xmath282 ( [ deconvconvergence ] ) . this implies that improvements in wl statistical accuracy are rapidly limited by the resolution of the galaxies : although going deep in the exposure increases the @xmath123 of the detected objects and reduces the scatter and systematic in the shear estimate , the required @xmath123 grows very rapidly for poorly resolved images , and it becomes difficult to produce additional useful wl data for poorly resolved galaxies . we have also found that larger , well resolved galaxies ( @xmath283 ) exhibit a problem with the deconvolution method when the psf is an airy function , due to the poor approximation of the airy - functions wings by the gl expansion . a simple remedy is available : smooth the psf and galaxy images before measuring both . another solution is to choose an alternative to gauss - laguerre as the psf decomposition basis functions , one that is better suited to describe an airy function . however , a non - gl decomposition will make the deconvolution process excessively complex ; another solution is to apodize the telescope to suppress the airy wings . @xcite offers an excellent description of the difference between various wl techniques ; we refer the reader to this paper for the difference between , for example , the ksb method and kuijken s method , which is applicable to the difference bewteen ksb method and the egl method . our egl method is similar to kuijken s polar shapelet method @xcite . what the methods have in common are : the deconvolution of the psf , which in principle allows for any psf effects to be removed ; forward fitting , which allows error propagation , and hence an error estimate to the measured shape ; and the definition of shape as shear , which has a well - defined shear transformation . all of these features contribute to a better shear accuracy . the differences between egl and kuijken s methods are subtle . the first difference is that kuijken method works the deconvolution and shearing in shapelet - coefficient space . our egl method determines the shape - as - shear by iteratively fitting within the pixel space . secondly , kuijken s method obtains the shape using a shear transformation valid to first - order in @xmath150 , which can be off by up to 10% at @xmath284 ( @xmath285 ) . our method uses basis functions that are elliptical , _ i.e. _ the shear transformation is valid to all orders , which allows for the shape to be measured accurately for any @xmath150 . the third difference is that , in kuijken s method , only the @xmath286 terms are used to describe the galaxy . this allows for less coefficients needed , and hence is efficient . in comparison , our method obtains the full set of coefficients to the specified order , which theoretically is more accurate under shearing or deconvolution . another difference is in the choice of scale radius . kuijken s method uses 1.3 times the best - fit gaussian @xmath139 for the basis function scale radius , which apparently is optimal for the first - order method ; egl uses the best - fit elliptical gaussian size , which gives the optimal sensitivity to small shear ( bj02 3.1 ) . unlike the ksb method , our egl deconvolution method , the @xcite polar shapelet and the @xcite methods do not rely on the approximation that the psf is circular up to a small linear `` smear . '' these methods are expected to do well on theoretical grounds ; their differences ( at least with * ? ? ? * ) seem to matter at the 1% level . as mentioned earlier , the main difference between this paper and previous end - to - end tests @xcite is that we test the individual steps of the analysis . the difference is mainly in the distribution of galaxies : these end - to - end tests mimic the distribution to an actual wl field image , while we test at specific s / n and galaxy size ( resolution ) sets . by controlling the exact distribution of galaxy shapes , we also eliminate shape noise in the shear estimate , and are able to quantify the effect that noise or resolution / sampling has on the shear . the main differences from other `` dissection '' methods @xcite are the inclusion of and quantifying noise effects , testing with asymmetric galaxies , and determining the limits of shape - measurement convergence . @xcite demonstrates the calibration errors for a variety of psfs ( gaussian , seeing - limited , diffraction - limited ) , while we only test for airy psfs as the worst - case scenario . in @xcite , the galaxy types were varied as well ( gaussian , exponential , devaucouleurs ) , but they are symmetric , which can mask potential problems in shape measurements . their simulation focuses on the calibration accuracy , but does not quantify the additive error or include noise effects . @xcite includes an analysis with noise , but does not quantify them ; a variety of sersic indices ( for galaxies ) are tested , but are symmetric . their psfs do not test the diffraction - limited ( airy ) case , although asymmetric psfs , which are not considered in this paper , are tested . with the egl method , the systematic due to the shape measurement method is well under the 0.1% level under ideal native - fit conditions , where the galaxy is symmetric and psf is absent , as long as the minimum sampling and s / n criteria ( minor axis width @xmath287 pixels , @xmath288 ) are met . the deviation from this high accuracy comes from real - life complications . in native fits , when the galaxies are not elliptically symmetric , the shear calibration factor degrades by @xmath289 as s / n is decreased . the presence of a psf further complicates the matter ; deconvolution with psfs of truncated coefficients becomes a source of error for well - resolved objects with an airy psf ( up to 0.5% ) , but solutions to this poor approximation of the psf are straightforward . more generally , the circularly - symmetric airy psf itself introduces an @xmath150 dependent systematic in the calibration factor , at worst @xmath290 at @xmath291 . a non - circular psf adds a shear additive error that is proportional to the shape of the psf , which also increases with decreasing s / n ( up to @xmath292 of @xmath270 at @xmath280 ) . for the deconvolution fits , the sign and magnitude of the calibration and additive error is approximately the same for both the symmetric and asymmetric galaxies . the systematics are typically inversely proportional to the s / n . at high s / n ( @xmath293 ) , we have calibration error @xmath294 , and the psf anisotropy is @xmath3 suppressed . these errors become @xmath4 problems as @xmath295 . the systematics are a function of the galaxy shape itself as well , with the systematic at low @xmath150 having the reverse sign of that at high @xmath150 . to attain 0.1% accuracy that is desired in future wl surveys , it is obvious that this shape measurement method itself needs to be refined . in addition , the whole pipeline needs to be rid of any systematics that are unrelated to the shape measurement , such as crowding or selection effects . the elliptical gauss - laguerre decomposition is one of the most stringently tested methods to characterize shapes of galaxies . with the egl decomposition , shapes are measured without the need for empirical shear / smear polarizabilities , and psfs are removed by deconvolution . the shear , obtained from averaging an isotropic ensemble of galaxy shapes , is highly accurate due to the definition of shape as shear . we have demonstrated that the egl method allows shear recovery of unprecedented accuracy , and quantified its degradation due to psf , truncation of the egl decomposition , image noise , and sampling rate / resolution . however , the current work is limited to the extraction of shapes ; further work , including the full pipeline analysis , is required for attaining 0.1% accuracy in shear estimation . we would like to thank david rusin and mike jarvis for much assistance and consultation . this work is supported by grants ast-0236702 from the national science foundation , department of energy grant doe - de - fg02 - 95er40893 and nasa befs-04 - 0014 - 0018 . bridle , s. , kneib , j .- p . , bardeau , s. , & gull , s. 2002 , the shapes of galaxies and their dark halos , proceedings of the yale cosmology workshop `` the shapes of galaxies and their dark matter halos '' , new haven , connecticut , usa , 28 - 30 may 2001 . edited by priyamvada natarajan . singapore : world scientific , 2002 , isbn 9810248482 , p.38 , 38

five planets are presently believed to orbit the primary star of 55 cnc , but there exists a large 5 au gap in their distribution between the two outermost planets . this gap has attracted considerable interest because it may contain one or more lower mass planets whose existence is not contradicted by long - term orbit stability analyses , in fact it is expected according to the `` packed planetary systems '' hypothesis and an empirical titius bode relation recently proposed for 55 cnc . furthermore , the second largest planet is just the second farthest and it orbits very close to the star . its orbit , the most circular of all , appears to be nearly but not quite commensurable with the orbit of the third planet , casting doubt that any migration or resonant capture of the inner planets has ever occurred and lending support to the idea of `` in situ '' giant planet formation by the process of core accretion . all of the above ideas will be tested in the coming years in this natural laboratory as more observations will become available . this opportunity presents itself in conjunction with a physical model that relates the orbits of the observed planets to the structure of the original protoplanetary disk that harbored their formation at the early stages of protostellar collapse . using only the 5 observed planets of 55 cnc , this model predicts that the surface density profile of its protoplanetary disk varied with distance precisely as , as was also found for the minimum mass solar nebula . despite this similarity , the disk of 55 cnc was smaller , heavier , and less rotationally supported than the solar nebula , so this system represents a different mode of multi planet formation compared to our own solar system .

0811.0868

eighteen years of doppler shift measurements of the primary star in the nearby binary system 55 cancri ( hd 75732 ; hereafter 55 cnc ) have gradually raised the number of orbiting planets from 1 ( butler et al . 1997 ) , to 3 ( marcy et al . 2002 ) , to 4 ( mcarthur et al . 2004 ) , and recently to 5 ( fischer et al . 2008 ) . although the 4 inner planets are closely packed together ( their semimajor axes @xmath2 au ) , the orbits of all 5 planets are remarkably circular ( see table 1 ) and presumably coplanar . the system exhibits astonishing regularity , reminiscent only of the planetary order in our solar system and unmatched by any other currently known multiple planet extrasolar system ( for reviews of orbital and dynamical characteristics of exoplanets , see marcy et al . [ 2005 ] and butler et al . [ 2006 ] ) . the most massive planet , 55 cnc d with @xmath3 au and a minimum mass ( mcarthur et al . 2004 ) , which increases all masses by 25% . ] of 3.835 jupiter masses ( table 1 ) , is the most distant gaseous giant known among exoplanets with well defined orbits and , like jupiter , it orbits beyond 5 au from the star . furthermore , there are no giant planets in the area roughly between 1 and 5 au , just like in our solar system . in fact , no planets have been found to date in the 5 au gap between the two outermost planets , f and d , with semimajor axes @xmath4 au and @xmath5 au , respectively . cccccc + + index & planet & orbital & semimajor & minimum & orbital + @xmath6 & designation & period & axis & mass & eccentricity + & & @xmath7 ( days ) & @xmath8 ( au ) & @xmath9 ( @xmath10)@xmath11 & @xmath12 + 1 & e & 2.817 & 0.038 & 0.034 & 0.070 + 2 & b & 14.65 & 0.115 & 0.824 & 0.014 + 3 & c & 44.34 & 0.240 & 0.169 & 0.086 + 4 & f & 260.0 & 0.781 & 0.144 & . . . @xmath13 + 5 & d & 5218 & 5.77 & 3.835 & 0.025 + + + + despite the similarities in orbital characteristics , the planetary system of 55 cnc exhibits some conspicuous differences when compared to our solar system ( fischer et al . [ 2008 ] and table 1 ) : ( a ) three of the planets orbit too close to the star ( within 0.24 au ) , that is closer than the orbit of mercury around the sun . ( b ) the second planet , 55 cnc b ( @xmath14 au ) , is also the second largest with a mass of @xmath15 jupiter masses . this planet also shows the lowest orbital eccentricity ( @xmath16 ) among all 5 planets . ( c ) the 4 inner planets are all quite massive , with the innermost planet , 55 cnc e ( @xmath17 au ) , having the smallest mass ( @xmath18 earth masses ) , a value roughly comparable to the mass of uranus . most , if not all , of these planets are expected to be gaseous giants in contrast to the terrestrial planets in our solar system . the central star of 55 cnc and the dynamics of its system have kept observers occupied for other reasons as well : 1 . this star is the primary in a visual binary ( separation @xmath1910@xmath20 au ) and belongs to a group of @xmath1930 multiple stellar systems in which giant planets are seen orbiting very close to the primary stars ( eggenberger et al . 2004 ; raghavan et al . in fact , the most massive inner planets tend to be found in multiple stellar systems and those with the shorter periods ( @xmath21 days ) tend to also have very low eccentricities . the three inner planets of 55 cnc , with orbital periods @xmath22 days and eccentricities @xmath23 , all fit precisely this trend . 2 . the primary star of 55 cnc has very high metallicity compared to the sun ( @xmath24 ; gonzalez & vanture 1998 ; valenti & fischer 2005 ) , a property that has been associated with a higher probability of forming massive protoplanetary cores and gaseous giant planets in the surrounding , also metal rich , disk by the process of core accretion ( ida & lin 2004b ; bodenheimer & pollack 1986 ; pollack et al . it seemed for a time that an extended disk of dust was visible in the infrared , but this result was negated when additional infrared observations failed to detect a disk ( schneider et al . 2001 ; bryden et al . 2006 ) and submillimeter observations indicated that the excess flux could be attributed to background sources ( jayawardhana et al . 4 . the orbits of the two more massive of the inner planets , 55 cnc b and c , appear to be close to but not quite in mean motion resonance with a period ratio of nearly 3:1 ( marcy et al . 2002 ; fischer et al . this is the only known 3:1 apparent resonant pair , whereas in all the other known cases the planets appear to have been captured in 2:1 resonances ( butler et al . 2006 ) . the picture that emerges from the observations of 55 cnc is enigmatic . the most important questions can be summarized as follows ( fischer et al . 2008 ) : why did this star end up with so many massive planets when 90% of the observed stars do not have any giants at all ? the similarities to our solar system seem to indicate that multiple planet systems like 55 cnc ( and @xmath25 ara ara ( hd 160691 , gj 691 ) also has a gaseous giant orbiting beyond 5 au ( @xmath26 au ) and 3 additional giants orbiting close to the star ( santos et al . 2004 ; pepe et al . 2007 ; and table 5 below ) . their distribution shows a large gap between 1.5 and 5.2 au ; but the orbital eccentricities of the inner planets are comparatively larger ( @xmath27 ) than those in 55 cnc ( see also [ muara ] ) . ] ) should be common , but why have they not been observed in larger numbers ? and if such multi planet systems around metal rich stars represent the new norm of planet evolution ( ida & lin 2004b ) , where are the systems with central stars chemically similar to our sun ? one would certainly be reluctant to admit that our solar system is special because of the inherent bias associated with such a proposition . in addition , were the inner planets of 55 cnc ( and @xmath25 ara ) formed `` in situ '' on the observed orbits ( lissauer 1995 ; bodenheimer et al . 2000 ) , or are they the result of inward migration ( ward & hahn 2000 ; lin et al . 2000 ; armitage et al . 2002 ; alibert et al . 2004 , 2005a , 2006 ) that terminated conveniently ( marcy et al . 2000 ; trilling et al . 2002 ; eggenberger et al . 2004 ) to allow for the planets to survive and circularize their orbits ? and why are all the planetary orbits in 55 cnc so nearly circular when large eccentricities are quite common in exoplanets ( marcy et al . 2000 , 2005 ) and it appears that resonant tidal interactions did not take hold since an actual 3:1 commensurability was not established ? finally , could there be smaller planets or an asteroid belt in the 5 au gap and , in any case , why was matter acquisition so disparate as to produce a series of smaller planets surrounded by two very massive planets , 55 cnc b and d ? finding answers to even some of these questions can be a giant step toward understanding planet formation and evolution . we will revisit the underlying issues in [ questions ] below . the observations of 55 cnc have motivated several empirical modeling efforts that attempted to explore the various compelling issues that arise from the impressive regularity of this planetary system , the existence of the 5 au gap , and the unusual chemical properties of the parent star : 1 . _ metal rich environment for planet formation_. ida & lin ( 2004a , b , 2005 ) carried out monte carlo simulations of protoplanetary core growth , orbital migration , and gas acquisition . their model predicts that the probability of forming gas giants increases rapidly with increasing metal abundances in the host stars and the surrounding disks , which is consistent with the multitude of planets observed in 55 cnc and @xmath25 ara . on the other hand , the physical and orbital properties of these planets are in disagreement with the main features of this model , namely that there should not be planets with @xmath28 earth masses at @xmath29 au and that short period , neptune size planets should not be common around such g - type stars as opposed to m dwarfs . this led ida & lin ( 2005 ) to speculate that the planets seen in 55 cnc and @xmath25 ara have not migrated to their present locations but they were instead formed in situ with the help of sweeping mean motion or secular resonances . however , there is no evidence that gas - giant migration or resonances ever played a role in the formation of these planets , and this adds to the mystery surrounding such well ordered planetary systems . 2 . _ long term orbital stability_. the stability of the planetary orbits in 55 cnc has been investigated independently by several groups : raymond & barnes ( 2005 ) integrated the orbits of the 3 most massive planets ( b d ) for 100 myr and found that the system could afford to have more planets in stable orbits between planets c and d. fischer et al . ( 2008 ) integrated the orbits of all 5 planets for 1 myr and found that planets b and c are not locked into a 3:1 mean motion resonance . gayon et al . ( 2008 ) integrated the orbits of all 5 planets for 400 myr and showed that the system represents a case of stable chaos in which , however , close encounters between planets are avoided . all of these models indicate clearly that the system is dynamically stable over long time scales ( @xmath30 0.1 gyr ) . 3 . _ apparent 3:1 mean motion resonance_. the apparent 3:1 mean motion resonance between the orbits of planets b and c in 55 cnc was initially hailed as a new example of a low order commensurability in a `` resonant system '' ( barnes & quinn 2004 ) , with 2:1 resonances providing the only previous example in observed systems . although barnes & quinn ( 2004 ) did not study 55 cnc explicitly , their models showed that planets in resonant systems have very narrow zones of stability and raised the possibility that such systems are only marginally stable . modeling of 55 cnc was undertaken by zhou et al . ( 2008 ) who showed that planets b and c can be captured in a 3:1 resonance if they start with modest eccentricities ( @xmath31 ) and migrate slowly ( for @xmath190.1 myr ) . the above models would indicate that the arrangement of the inner planets in 55 cnc is the result of special circumstances , an uncomfortable notion . since we now know that planets b and c are not in resonance ( fischer et al . 2008 ; gayon et al . 2008 ) , a more palatable conjecture is that no special migration has ever occurred in 55 cnc and that its inner planets form a typical `` interacting system '' ( barnes & quinn 2004 ) in which the planets are able to perturb one another and have broad zones of stability available to them . modeling efforts and related conjectures concerning the apparent 3:1 resonance in 55 cnc demonstrate how difficult it is to simulate such complex systems because we do not have a handle on the initial conditions ; numerical models that start from arbitrary but convenient initial conditions are capable of producing `` easily '' just about all of the observed results . so it is not surprising that more recent modeling efforts ( although they did not address 55 cnc directly ) now seem to provide evidence against 3:1 resonant captures of planets in general : pierens & nelson ( 2008 ) found that massive planets captured by a dominant , jupiter size planet tend to end up exclusively in 2:1 or 3:2 mean motion resonances , depending on their mass . furthermore , adams et al . ( 2008 ) and lecoanet et al . ( 2008 ) found that turbulence in the host protoplanetary disk can upset mean motion resonances and thus resonant systems should be rarely observed . _ packed planetary systems hypothesis_. the orbital stability calculations of barnes & quinn ( 2004 ) and related simulations by barnes & raymond ( 2004 ) used massless test particles in models of several known extrasolar multiplanetary systems in order to identify broad regions full of dynamically stable orbits and to predict the possible existence of additional , yet unseen planets . these calculations were extended by raymond & barnes ( 2005 ) , raymond et al . ( 2006 ) , and barnes et al . ( 2008 ) to include various saturn mass and terrestrial mass particles in between the orbits of the known planets in each simulated system . although it was easy to find stable orbits for additional planets with terrestrial or smaller masses in some of these systems , the main result of these numerical experiments came as a surprise : many systems are on the edge of stability and small changes to the values of @xmath32 or @xmath33 push them to become unstable . following the ideas of laskar ( 1996 ) , barnes & raymond ( 2004 ) conjectured that the remaining apparently stable systems are also packed with enough planets to be marginally stable . this is the packed planetary systems ( pps ) hypothesis and it predicts that stable systems may contain unseen planets that bring them to the edge of stability . the pps models successfully predicted the existence of hd 74156 d ( barnes et al . 2008 ) before its discovery ( bean et al . 2008 ) and proved to be consistent with the presence of 55 cnc f that was found ( fischer et al . 2008 ) orbiting at the inner edge of the predicted stable zone ( @xmath34 au @xmath35 3.2 au , @xmath36 ; barnes & raymond 2004 ) . the stable zone in 55 cnc is too wide , as a result the 5planet system is still not packed and there could be more undetected planets between the planets f and d ( raymond et al . 2008 ) . specifically , the pps models suggest the following interesting possibilities : ( a ) a giant planet with up to saturn mass may exist near 2 au ( the center of the stable zone ) with @xmath37 ( raymond & barnes 2005 ) . ( b ) alternatively , several terrestrial planets with up to 0.63 earth masses may exist at 1.13.6 au with @xmath38 ( raymond et al . although the possibility of having earth like planets in the `` habitable zone '' around 1 au is intriguing , our own interest in this work is focused entirely on the possibility of having a gas giant orbiting at @xmath192 au for the reasons that we explain in [ newmodel ] below . the current observations of 55 cnc do not rule out such a planet : at a distance of @xmath192 au , a giant planet could avoid detection if its minimum mass is smaller than 100 earth masses ( fischer et al . 2008 ) , i.e. , approximately equal to or less than the mass of saturn . 5 . _ empirical titius bode relation_. poveda & lara ( 2008 ) presented an exponential fit of the semimajor axes of the 5 known planets in 55 cnc . this empirical relation is akin to the titius bode ( tb ) rule for the solar system , but it works only if there exists a hypothetical sixth planet at @xmath39 au . based on this result , poveda & lara ( 2008 ) proposed that an unseen planet with @xmath40 au should be present in 55 cnc . there is no physical basis for this prediction , yet it is quite interesting that the empirical tb relation of 55 cnc is in complete agreement with the prediction of the pps model for a gas giant at the same orbital distance . since the pps model and the tb relation converge to the same prediction , they can both be tested by additional doppler shift measurements of the primary star of 55 cnc . by extending their baseline , future observations should be able to search for an additional periodicity in the 5 au gap of 55 cnc and , irrespective of the outcome , the result can be pivotal for planet formation theories ( see also [ newmodel].5 ) . 6 . _ the protoplanetary disk of 55 cnc_. fischer et al . ( 2008 ) made an attempt to estimate some key parameters of the minimum mass protoplanetary disk of 55 cnc . using a typical surface density profile for the disk that varied with distance @xmath0 as @xmath1 and an estimate of the core masses of the 5 observed planets that took into account the excessive metallicity of the system , they found a total mass of 0.031 @xmath41 out to 6 au and a surface density for the dust of @xmath42 at 5 au . the dust - to - gas mass ratio was set to the nominal value of 1% and the mass estimate also took into account the intermediate inclination of the orbital plane of the system ( @xmath43 , as derived by mcarthur et al . [ 2004 ] for 55 cnc d ) . the assumed @xmath44 profile was first established for the minimum mass solar nebula ( mmsn ) by weidenschilling ( 1977 ) and hayashi ( 1981 ) and it has been commonly used ever since in calculations of this type . recently , this profile was investigated again in our solar system ( davis 2005 ) , in extrasolar systems ( kuchner 2004 ) , and in dusty circumstellar disks around t tauri stars ( kitamura et al . 2002 ; andrews & williams 2007 ) . due to large systematic uncertainties , none of these investigations has been able to directly confirm or reject the original @xmath44 relation , although all of them provide useful hints about possible improvements of the `` typical '' surface density profile of the solar nebula and extrasolar protoplanetary disks . for what follows , it is interesting to note that the the model of davis ( 2005 ) indicates that the original power law @xmath45 with @xmath46 tends to overestimate the surface density within the central 1 au of the solar nebula . also , kuchner ( 2004 ) finds a steeper profile with @xmath47 for the surface density of the typical minimum mass `` extrasolar nebula '' ( mmen ) constructed by weidenschilling s ( 1977 ) method from 26 exoplanets found in multi planet systems . the @xmath47 power law index results , again , in an overestimate of the surface density within the central 1 au of the mmen . the lack of a central core region , such as that suggested for the mmsn ( lissauer 1987 ; davis 2005 ) and for circumstellar disks ( garaud & lin 2007 ) , is the obvious reason for the excessive central densities in models that adopt a single power law density profile for the entire protoplanetary disk . we are interested in 55 cnc because this is the only extrasolar system with so many planets in well ordered , long lived , circular orbits . we believe that the high degree of regularity seen in this system makes it an ideal laboratory for exploring the above discussed issues and for testing the leading current hypotheses and models about planet formation in protostellar disks . our work starts with a new physical model that we formulated rigorously based on exact solutions of the isothermal lane emden equation with rotation and that relates the observed planets in a well ordered , multi planet system to key physical properties of the protoplanetary disk that hosted their formation . first we applied this model successfully to the 11 largest planets of our solar system ( christodoulou & kazanas 2008 , hereafter ck ) and we proceed here to apply it to the 5 known planets of 55 cnc . the results obtained from our modeling are tied to the theoretical models of [ th ] in the following ways : 1 . our model uses all the planets in regular orbits irrespective of size , mass , and metal abundances . as such , it works equally well for solar type systems like our solar system and for systems with different metallicities or with gas giants in short period orbits like 55 cnc . we do expect however that the modeling results will have to be interpreted consistently with the physical and the chemical properties of the detected planets and their parent stars . it is important that the planetary orbits of 55 cnc were found to be dynamically stable over long time scales because our model associates such long lived orbits with local minima of the gravitational potential well in the midplane of the original protoplanetary disk . this association does not hold for many odd systems in which one or two planets have migrated inward and are now seen very close to their parent stars . however , the regular spacing of the planets and the stability of 55 cnc argue against migration as a process that dominated its evolution . it is now clear that 55 cnc b and c are not caught in a 3:1 mean motion resonance . this result also suggests that no significant migration has taken place in the disk of 55 cnc and it supports our picture of a well - ordered system in which the planets formed in situ by core accretion . in turn , such regular planetary orbits readily provide information about the locations of density enhancements ( the local potential minima ) in the disk during the planet formation stage . the pps models predict that there must exist at least one more planet in the 5 au gap of 55 cnc and they will be tested by future observations for the first time in such a well ordered , multi planet , extrasolar system . our model is also not impervious to the possible existence of an additional planet , but it does not necessarily need a 6@xmath48 planet in order to work . as it predicts entirely different physical characteristics for the disk of 55 cnc depending on whether 5 or 6 planets are used ( see [ properties ] ) , our model too needs an observational resolution of this dilemma . 5 . based on the work of ck , we now understand that the tb relation can be accommodated within our approach , but the rule has no inherent physical significance for our solar system or for 55 cnc . since the tb relation of 55 cnc necessarily predicts a 6@xmath48 planet at @xmath40 au , the absence of such a planet from the 5 au gap will settle this issue without a doubt . such an observational result will however be detrimental to the pps hypothesis as well . alternatively , if an additional planet is found in the 5 au gap , then the doubts will persist , the pps hypothesis will persevere , and our model ( and presumably models such as those of weidenschilling [ 1977 ] and davis [ 2005 ] ) will then show a clear preference for a different radial density profile of the protoplanetary disk of 55 cnc ( see [ properties ] for details ) . the attempt of fischer et al . ( 2008 ) to estimate some of the properties of the minimum mass protoplanetary nebula ( mmpn ) of 55 cnc is noteworthy by insufficient . the main drawback of these estimates is that the authors assumed a @xmath49 profile for the disk without actually using weidenschilling s ( 1977 ) method or any of the competing methods cited in [ th].6 above . consequently , the particular density profile produced by our disk model of 55 cnc can not be compared to this previous work . some of the parameter values obtained by fischer et al . ( 2008 ) and by kuchner ( 2004 ) , who did use weidenschilling s method , can however be compared to the results of our modeling , and we do so in [ surface ] below . in [ model ] , we describe in detail the new physical model of the midplane of the protoplanetary disk of 55 cnc and we derive the fundamental dynamical parameters of this disk ( minimum mass , density profile , specific angular momentum , rotation frequency , and equation of state ) . we also examine the dynamical stability of the model and we describe the effects that a possible 6@xmath48 planet will have if it is discovered orbiting at @xmath40 au , as expected by the pps and tb models of the system ( [ th ] ) . in [ discussion ] , we interpret the results from our model and we discuss the issues and the questions mentioned above that pertain to the impressive regular structure of this planetary system and its intriguing 5 au gap . we also report briefly on our modeling effort of the related multi planet system @xmath25 ara . we are interested in the equilibrium structure of a rotating , self gravitating , isothermal disk of angular velocity @xmath50 , mass density @xmath51 , and sound speed @xmath52 . following ck , we adopt cylindrical coordinates ( @xmath53 ) , the assumption of axisymmetry ( @xmath54 ) , and the assumption of cylindrical symmetry ( @xmath55 ) that is valid on the midplane of any disk . we cast all quantities in dimensionless form as follows : we use the central density @xmath56 and the length scale of the disk @xmath57 where @xmath58 is the gravitational constant , and we define the normalized radius as @xmath59 and the normalized density as @xmath60 . in addition , we adopt a rotation profile of the form @xmath61 where @xmath62 and the dimensionless function @xmath63 is to be determined . finally , we use @xmath64 and the jeans frequency @xmath65 to define a dimensionless rotation parameter of the form @xmath66 the equilibrium structure of the midplane of the disk is then described by the isothermal lane emden equation with rotation that takes the dimensionless form @xmath67 in ck , we showed that any power law density profile is an exact analytic solution of this equation , provided that the rotation profile @xmath63 is also also determined self consistently from eq . ( [ main1 ] ) . such `` baseline '' power law profiles are particular solutions of the lane emden equation , but they are unable to satisfy the central boundary conditions , namely that @xmath68 they do however determine analytically the appropriate rotation profile @xmath63 and the mean density profile irrespective of the applicable ( physical ) boundary conditions . the exact solutions can then be obtained by a simple numerical integration that obeys the boundary conditions ( [ bc ] ) , and they are bound to oscillate permanently about the corresponding baseline solutions without ever settling on to them ( that would be a violation of the boundary conditions ) . it is important to note that , although any monotonic power law profile can be chosen as a baseline solution and it will produce a monotonic rotation profile , the corresponding exact solution subject to the boundary conditions ( [ bc ] ) will still have to be oscillatory , creating thus a series of local maxima and minima across the entire density profile . for our baseline equilibrium model of the midplane of the mmpn of 55 cnc , we adopt a composite analytic solution in which the mean density profile is a combination of a flat inner region ( a `` core '' ) followed by a declining power law : @xmath69 where @xmath70 is the radius of the constant density core region and @xmath71 . compared to the model of the solar nebula in ck , this model does not have a flat outer region because no outer equidistant planets are observed in 55 cnc . the rotation pofile is determined next by substituting eq . ( [ den ] ) into eq . ( [ main1 ] ) and by solving the resulting differential equation : @xmath72 } \ , \ \ \ { \rm if } \ \ \ x > x_1 \end{array } \right . . \label{den31}\ ] ] it is easy to show that this rotation law obeys the physical requirements that @xmath73 and @xmath74 everywhere for any choice of @xmath71 . the functions @xmath75 and @xmath63 are plotted in figure [ fig1 ] for @xmath76 and for various choices of the power law index @xmath71 . we see that , as @xmath77 is increased , the rotation profile is not affected as much as the density profile . the above composite equilibrium model is characterized by three free parameters : the core radius @xmath78 , the rotation parameter @xmath79 , and the power law index @xmath71 . the density profile ( eq . [ [ den ] ] ) of this baseline solution of the lane emden equation ( [ main1 ] ) is not capable of satisfying the physical boundary conditions ( [ bc ] ) at the center and it serves only as a mean approximation to the mean density of the corresponding physical model . the general form of the rotation law of the baseline ( eq . [ [ den31 ] ] ) can , however , be adopted for the differential rotation @xmath63 of the physical model as well . then eq . ( [ den ] ) provides a prescription for the rhs of the lane emden equation ( [ main1 ] ) which can thus be written as @xmath80 this differential equation can be integrated numerically subject to the physical boundary conditions ( [ bc ] ) . for all choices of the free parameters \{@xmath81 , @xmath79 , @xmath71 } , the numerical solutions oscillate about the baseline solutions ( [ den ] ) . the local maxima of the density profile correspond to minima of the gravitational potential well in the midplane of the disk and they are ideal locations for the formation of planetary cores from the dust grains that also have to aggregate inside these wells . we construct a model of the mmpn of 55 cnc by choosing the 3@xmath82 planet ( @xmath83 au ) from the star to set the physical scale of the model . ( similar results are obtained when any of the inner 4 planetary orbits is used for scaling . ) then the third density peak of the model ( not counting the central peak at @xmath84 ) is associated with a distance of 0.24 au from the center , and the exact solutions of the lane emden equation ( [ main2 ] ) are optimized by varying the free parameters until the remaining density enhancements match as closely as possible the observed semimajor axes of the remaining planetary orbits of 55 cnc . fit model is shown in figure [ fig2 ] and its density peaks are listed in table 2 along with the observed semimajor axes given by fischer et al . the third density peak occurs at @xmath85 in the best fit model of 55 cnc , implying that the length scale of the protoplanetary disk in the isothermal phase of its evolution was extremely small ( @xmath86 au ) . this value is @xmath1930 times smaller than the length scale of our solar system ( @xmath87 au ; ck ) and indicates that the inner disk of 55 cnc was denser than the solar nebula ( see [ properties ] below ) . furthermore , there is no other density peak out to 100 au , so this model predicts no more planets beyond the orbit of 55 cnc d. cccccc + + + index & planet & semimajor & peak & relative & peak + @xmath6 & designation & axis & location & deviation & density + & & @xmath8 ( au ) & @xmath88 ( au ) & ( % ) & @xmath89 + 1 & e & 0.038 & 0.042 & 10.5 & @xmath90 + 2 & b & 0.115 & 0.091 & @xmath91 & @xmath92 + 3 & c & 0.240 & 0.240 & . . . & @xmath93 + 4 & f & 0.781 & 0.862 & 10.4 & @xmath94 + 5 & d & 5.77 & 5.61 & @xmath95 & @xmath96 + the relative deviations between the locations of the model peaks and the planetary orbits are also listed in table 2 . the largest relative deviation ( @xmath1921% ) is observed for the second peak that corresponds to the orbit of planet b. this difference is still quite small ( @xmath190.02 au ) in absolute terms but it appears disproportionately large because our method of optimization is skewed toward the inner planets for which it produces larger relative deviations ( see ck for more details ) . the values listed in table 2 can not be improved individually because of the extreme nonlinear nature of the problem . only the value of @xmath97 can be optimized independently of the other parameters by minimizing the deviation of the innermost planet from the corresponding density peak . the strong coupling of the remaining two parameters then demonstrates that fitting 5 planetary orbits with 3 free parameters is not a simple matter , as nonlinearities do not generally allow for one - to - one relations between parameters and planetary positions . the parameters of the best fit model determined by the optimizing algorithm along with the scaling assumption that @xmath98 au ( see table 2 ) constitute a set of important dynamical parameters of the mmpn of 55 cnc : @xmath99 the value of the power law index @xmath100 indicates that the density profile of the differentially rotating region of the disk declined with radius , on average , as @xmath101 . this region extended outward beyond the core radius of @xmath102 au . the low value of @xmath97 indicates that the rotation of the isothermal nebula was sufficiently slow to avoid nonaxisymmetric instabilities ( see [ stability ] below ) . the unusually small value of @xmath103 , along with eq . ( [ length ] ) and a typical low gas temperature , indicates that the nebula was very dense , about @xmath1910@xmath20 times denser than the solar nebula ( see table 3 below ) . so the picture that emerges from the best fit model of 55 cnc is that of a relatively heavy and slowly rotating disk that is long lived and that can comfortably build massive planets over time by accumulating the solids and then accreting the gas contained within each local potential well . the optimization procedure also finds some models of high quality in which somewhat larger magnitudes of @xmath97 are compensated by smaller values of @xmath70 . the most different model gives us an idea about how shallow the area around the true minimum is in the three parameter space of the model . this model is characterized by the following values : @xmath104 , @xmath105 , @xmath106 au , @xmath107 , @xmath108 au , and the relative deviations are smaller than 22% . however , the power law index does not differ from @xmath109 by more than 1.2% in any of the high quality models . similarly , the variations of @xmath103 and @xmath70 always produce core radii to within 1% of @xmath1100.052 au in all of these models , so there is no doubt that the core region of the mmpn of 55 cnc was extremely small ( @xmath1910 solar radii ) . emden equation that we solved can serve as a model of a differentially rotating disk supported by thermal pressure in the @xmath111direction , so we expect that the scale height from pressure support will be @xmath112 down the radial density gradient . ] where @xmath113 is the rotation speed . in our models , @xmath113 becomes asymptotically flat when the density exhibits a power law profile with @xmath71 and @xmath52 is constant , leading to the approximate relation that @xmath112 . ] in this case , the value of @xmath100 implies that the corresponding power law index in the surface density profile ( @xmath114 ) of the nebular disk was @xmath115 . this value is virtually identical to that obtained for the solar nebula by weidenschilling ( 1977 ) who used an entirely different method and by ck who used the same optimization technique as in this work . the conformance of these results does not hold if a 6@xmath116 planet is assumed to exist in the 5 au gap of 55 cnc . we have also optimized models that included a hypothetical planet at @xmath117 au ( as predicted by the pps and tb models of 55 cnc ) , and the best fit model of the six planetary orbits has the following parameter values : @xmath118 , @xmath119 , @xmath120 au , @xmath121 , @xmath122 au , and the relative deviations are smaller than 19% . these values are comparable to those in the best fit model of the 5 observed planets , except for @xmath77 whose smaller value produces a softer density gradient to accommodate the hypothetical planet at 2 au . this softer gradient can also accommodate one additional planet farther out , but the outermost density peak occurs at 20.56 au which differs from @xmath123 au predicted for a hypothetical 7@xmath116 planet by the empirical tb relation of poveda & lara ( 2008 ) . this peak is also not consistent with the recent pps prediction ( raymond et al . 2008 ) that an unseen outer planet could exist close to 9 or 10 au , i.e. near the inner edge of a wide zone of stability found beyond 55 cnc d. thus , our model with hypothetical planets is still markedly different than the pps and tb models of 55 cnc . our modeling also delineates some of the fundamental physical characteristics behind the mean density profile of the midplane of the disk of 55 cnc . with the aid of our best fit model , we can deduce substantial information concerning the structure and the dynamics of the nebular disk . in addition to the structural and rotational parameters discussed above , we can use our analytic baseline model in order to probe the dynamical state of the protoplanetary disk of 55 cnc in its isothermal phase . following ck , we calculate the relevant physical quantities for the mmpn of 55 cnc and we summarize our results in table 3 . ccccc + + description & quantity & & 55 cnc value @xmath11 & solar system value @xmath11 + equation of state & @xmath124 & @xmath125 & @xmath126 & @xmath127 + & @xmath128 & @xmath125 & @xmath129 & @xmath130 + + central density @xmath13 & @xmath131 & @xmath125 & @xmath132 & @xmath133 + jeans frequency & @xmath134 & @xmath125 & @xmath135 & @xmath136 + rotation frequency & @xmath64 & @xmath125 & @xmath137 & @xmath138 + rotation period & @xmath139 & @xmath125 & 1.4 yr & 12 yr + + + + surface density & @xmath140 & @xmath125 & @xmath141 & @xmath142 + mass & @xmath143 & @xmath125 & @xmath144 & @xmath145 + specific angular momentum & @xmath146 & @xmath125 & @xmath147 & @xmath148 + + + + mass & m & @xmath125 & @xmath149 & @xmath150 + + + + + the corresponding results obtained by ck for the solar nebula are also listed in table 3 for comparison purposes . compared to the core of the solar nebula , the mmpn core of 55 cnc was smaller ( @xmath151 ratio @xmath1916 ) , denser ( @xmath131 ratio @xmath1910@xmath20 ) , heavier ( @xmath152 ratio @xmath1930 ) , and it was rotating @xmath1910 times faster . the disk contained @xmath19100 times less mass ) and @xmath103 is extremely small in both cases . the @xmath111integration needs to be extended for about 1 au away from the midplane in order to produce masses @xmath190.01 @xmath41 . ] within a 6-au radius and only 2/5 of the angular momentum per unit mass . there is no doubt that these two disks were different and their differences are reflected in the distributions and the physical properties of the various planets that they formed over time . using the @xmath153 value of 55 cnc from table 3 , the core radius @xmath154 au , and the power law index @xmath155 , we can write the surface density profile of the mmpn of 55 cnc for @xmath156 in the form @xmath157 where @xmath0 is measured in au . this equation shows that the surface density of the gas is @xmath158 at @xmath159 au and @xmath160 at @xmath161 au . these values are 48% of the corresponding values in the solar nebula . the 1-au value is in agreement with kuchner s ( 2004 ) result of @xmath162 , although he used just the 3 planets b d , the only ones known at the time . in particular , planet f was not included in the model and this is the reason that the 1-au value was underestimated and the density profile turned out to be too steep with @xmath163 . in general , undetected planets cause the main problem in this method , and if such planets are included , they tend to soften the density gradients and push kuchner s average mmen value of @xmath164 closer toward @xmath165 . the 5-au value corresponds to a typical surface density of solids of @xmath166 for a nominal value of 1% for the dust - to - gas mass ratio . this value is comparable to the low end values calculated by fischer et al . ( 2008 ) using only the planetary masses inferred from observations . some additional estimates by the same authors reach as high as @xmath42 . such estimates rely on uncertain assumptions about the masses of the solid planetary cores and they are obviously too high , as they predict much larger densities in the disk of 55 cnc compared to those in the solar nebula . but we believe that the opposite is true since our model and kuchner s ( 2004 ) model , following entirely different methodologies , both predict significantly lower densities at 5 au for the mmpn of 55 cnc than the mmsn . we note that such low densities in the outer disk of 55 cnc do not pose a problem for the formation of the massive planet d. as fig . [ fig2 ] shows , the half width of the corresponding density peak is about 3 au , and all the solids within this emormous potential well were bound to aggregate at the bottom and form the massive core of this planet . all composite power law models with @xmath71 described above are stable to axisymmetric perturbations because they satisfy the rayleigh criterion : the specific angular momentum @xmath167 is an increasing function of @xmath0 at all radii . this can be shown easily by using eq . ( [ den31 ] ) to provide the function @xmath168 and then by proving that @xmath169 for all @xmath170 and @xmath171 . the above models are also stable to nonaxisymmetric perturbations because they do not rotate too fast . for sufficiently slow rotation , the nonaxisymmetric modes are neutral and incapable of merging to produce instability . this behavior is captured in the @xmath172parameter stability criterion ( christodoulou et al . 1995 ) which for gaseous disk models can be written as @xmath173 in our notation , this is equivalent to requiring that @xmath174 for stability , thus our best fit model with @xmath175 is far from the threshold of nonaxisymmetric instability . a disk of dust would also be stable with this rotation law . for an unisotropic disk of particles , the corresponding stability threshold switches from 0.35 to 0.25 , and the requirement for stability then becomes @xmath176 . in [ intro ] , we placed in proper context all the available information about the multi planetary system of 55 cnc . our knowledge of this system comes from 18 years of doppler shift observations of the central star ( fischer et al . 2008 ; mcarthur et al . 2004 ; marcy et al . 2002 ; butler et al . 1997 ) and from theoretical modeling of the inferred planetary orbits ( fischer et al . 2008 ; raymond et al . 2008 ; poveda & lara 2008 ; gayon et al . 2008 ; zhou et al . 2008 ; raymond & barnes 2005 ; ida & lin 2005 ; kuchner 2004 ) . the metal rich primary star of 55 cnc has 5 massive planets in remarkably circular orbits and , outside of our sun , this is the only star with so many planets in well ordered , perfectly stable , nonresonant orbits ( table 1 ) . the system does however exhibit a large 5 au gap between its two outermost planets and this vast empty region has sparked speculation that one or more additional planets may still remain undetected between planets f ( @xmath4 au ) and d ( @xmath5 au ) . our contribution to the investigation of 55 cnc is a new physical model of the midplane of the protoplanetary disk in which the planets were formed . this model was introduced in [ newmodel ] and it was described in detail in [ model ] . it is based on exact solutions of the isothermal lane emden equation with rotation ( eq . [ [ main1 ] ] ) subject to the appropriate central boundary conditions ( eq . [ [ bc ] ] ) . these solutions describe radial density profiles that are oscillatory by nature , despite the fact that the corresponding rotation profiles are strictly monotonic and well - behaved . the oscillations occur about an average ( baseline ) density ( eq . [ [ den ] ] ) that is a combination of two power laws , a flat uniform core and a decreasing outer section ( [ baseline ] ) . such baseline density profiles are also exact intrinsic solutions of the lane - emden equation but they are incapable of satisfying the central boundary conditions ( see ck for more details and for an application of the model to our solar system ) . the baselines do however obey the same rotation law ( eq . [ [ den31 ] ] ) as the corresponding oscillatory solutions . some typical baselines and their rotation profiles are shown in fig . [ fig1 ] , while a typical oscillatory solution and its baseline are shown in fig . [ fig2 ] . the oscillatory density profile shown in fig . [ fig2 ] represents the best fit model for the midplane of the protoplanetary disk of 55 cnc . the fit was obtained by nonlinear unconstrained optimization that matches the orbital semimajor axes of the observed planets to consecutive density maxima ( or gravitational potential minima ) of the oscillatory profile ( see [ exact ] and table 2 ) and that uses the semimajor axis of the third planet from the star to set the physical scale of the model . the dynamical parameters of the best fit model are listed in eq . ( [ neb ] ) and its physical properties are summarized in [ properties][surface ] and in table 3 . the midplane of the disk of 55 cnc was composed of a very small ( @xmath190.05 au ) and very dense ( @xmath1910@xmath177 ) uniform core in slow rotation ( period @xmath191 yr ) , followed by a power law gradient such that the surface density varied with distance as @xmath178 ( eq . [ [ sigma ] ] ) out to @xmath196 au . this configuration is stable to both axisymmetric and nonaxisymmetric perturbations ( [ stability ] ) . the @xmath178 dependence of the surface density profile of 55 cnc is effectively identical to that found by ck for the solar nebula . despite this resemblance , the detailed comparison between the two nebular disks shown in table 3 indicates that the two models are generally dissimilar , and the differences seen in table 3 are also reflected in the differing characteristics of the planetary systems that were created around these two stars . specifically , the inner disk of 55 cnc was denser and heavier , and this explains why 4 massive planets were formed within the inner 0.8 au . these high densities are also responsible for the large growth of the second planet from the star , 55 cnc b , which is the most massive of the inner 4 planets . this planet was free to grow in a high density environment , while the area around the orbit of the first planet could have been depleted by the protostar ( just as mercury s growth was severely limited by the protosun ) . on the other hand , the same disk showed much smaller densities beyond 2 au , and this explains why only one massive planet managed to form at the bottom of the single , very wide potential well outside of 5 au . ccc + physical & large light disks & small heavy disks + property & ( solar system ) @xmath179 & ( 55 cancri a ) @xmath180 + & & + disk & & + & & + metallicity ( @xmath181 ) & @xmath182 & @xmath183 + radial scale ( @xmath103 ) & 0.02 au & 0.001 au + radial size & 4070 au & 610 au + core size ( @xmath151 ) & 1 au & 0.05 au + core density ( @xmath131 ) & @xmath184 & @xmath185 + & & + rotation & & + & & + @xmath97 & 0.4 & 0.1 + rotation period ( @xmath139 ) & 12 yr & 1 yr + jeans period ( @xmath186 ) & 5 yr & 0.1 yr + & & + gas surface density & & + & & + @xmath153 & 2500 & 75000 + @xmath187 & 2000 & 1000 + @xmath188 & 200 & 100 + @xmath189 , @xmath190 & @xmath165 & @xmath165 + + + + + the planetary system of 55 cnc and the solar system are both remarkably well ordered despite the differences in structure and dynamical characteristics and the differences between their host stars . this comparison seems to indicate that there exist two modes of planet formation that lead to stable , well organized , multi planet systems . based on our modeling and on the results of ck , we can describe these two extremes of multiple planet formation in the isothermal regime of protostellar collapse as follows ( see also table 4 ) : 1 . _ large light disks._solar type protoplanetary disks with @xmath191 au cores and moderate rotational support ( no more than @xmath1940%50% of maximum rotation ) that have core densities a few times above those required to form planets ( @xmath1910@xmath191 ; lissauer 1993 ) are capable of forming rocky planets in the first 2 au and , due to the comparatively high densities in their outer regions , several gas giants beyond 5 au as well . such disks can extend for at least 4070 au in the radial direction . because their core densities are barely supercritical , these disks can form only terrestrial planets in the vicinity of the core radius . 2 . _ small heavy disks._metal rich protoplanetary disks with @xmath190.05 au cores and little rotational support ( no more than @xmath1910%15% of maximum rotation ) that have core densities @xmath191000 times above the critical value are capable of forming giant planets in the first 2 au and , due to the substantial density drop in their outer regions , only one gas giant beyond 5 au . such disks can not extend for more than 610 au ( and if they do , their densities are too low to form giant planets ) . because their core densities are strongly supercritical , these disks can easily form gas giants around 1 au , so terrestrial planets are not expected to be present in the habitable zone in this mode . returning now to the questions posed in [ obs ] , we can outline some feasible answers suggested by the above results : most of the observed stars ( @xmath1990% ) do not show any giant planets at all . some of them , especially those of solar type , may have planetary systems formed in large light disks ( llds ) , in which case their planets are not observable yet with the current resolution because the inner planets have very low masses and the outer planets are too distant from their stars . this is probably the reason why the solar system is currently the only example of a system formed in an lld . on the other hand , the inner giant planets formed in small heavy disks ( shds ) can be presently detected . besides the standard example of 55 cnc , all other planets found within 23 au from their stars in nearly circular orbits are good candidates for the shd mode . this also includes @xmath25 ara , hd 37124 , and the massive , short period ( @xmath21 days ) planets with low eccentricities found in multiple stellar systems ( see [ obs ] ) . on the question of in situ formation versus migration , the slow rotation found in our models ( table 4 ) and the structure of the disks that show several local potential minima strongly support the in situ formation of planets by planitesimal aggregation and gas accretion inside these potential wells . this does not mean that planetary migration can not occur at all , it just indicates that the multiple planets of well ordered systems do not migrate . migration may still be the dominant mechanism in systems with planets in highly eccentric and/or irregularly spaced orbits . but even in such cases , our modeling adds a new dimension to the migration problem that has not been previously explored : the ability of a planet to migrate can no longer be determined solely by the disk torques that redistribute the angular momentum of a planetary orbit . that was done in the past ( e.g. , lin et al . [ 2000 ] , ward & hahn [ 2000 ] , and references therein ) while the density profile of the underlying disk was assumed to be monotonically decreasing without any localized potential wells . in the present setting in which the planets are trapped inside gravitational potential minima , the energy content of each planet needs to be considered as well . a planet orbiting inside a local gravitational potential well can migrate only if the interaction with the disk provides enough energy for it to leap out of its well otherwise the torqing of the planetary orbit will simply lead to stable oscillations and the development of some eccentricity insde this potential minimum . on the question of the existence of additional planets in 55 cnc , our model boldly predicts that no other planet will be found in the future . in particular , we do not believe that there is room in the 5 au gap even for a small rocky terrestrial planet or an asteroid belt because the potential minima of planets f and d reach into the gap ( see fig . [ fig2 ] ) , and all debris must have been swept by these two massive planets . this prediction is in contrast to the predictions of the pps models ( raymond et al . 2008 ) and the tb conjecture ( poveda & lara 2008 ) for the system , and the disagreement can be resolved by future observations of 55 cnc ( see also [ properties ] above ) . the question of the impressive growth of inner planet 55 cnc b ( @xmath14 au , mass @xmath192 jupiter masses ) was discussed briefly in [ modes ] above . the density of the peak where this planet was formed ( table 2 ) is quite high ( @xmath193 ) . only planet e was formed in a higher density environment in the shd of 55 cnc , but this planet resides deep inside the core of the nebula and its gaseous envelope could have easily been ablated by the protostar at a later time . the situation is similar in the inner solar system , although venus is slightly less massive than the earth . the emerging paradigm , valid for both llds and shds , is that of a depleted innermost planet followed by planets that are free to grow as much as the local conditions in the disk would allow . this picture appears to be borne out also in another shd candidate , @xmath25 ara , but with some caveats that we discuss below . recent observations ( pepe et al . [ 2007 ] and table 5 ) of the metal rich star @xmath25 ara ( @xmath24 ; santos et al . 2004 ) indicate 4 orbiting planets in a configuration very similar to that of 55 cnc : a single massive outer planet , @xmath25 ara e , is separated from the 3 inner planets ( c , d , and b ) by a large empty region between 1.5 and 5.2 au . the minimum mass of the innermost planet is comparable to that of the innermost planet in 55 cnc , so this planet is also a low mass ( depleted ) giant while its two neighbors have minimum masses comparable to the mass of jupiter . cccccc + + index & planet & orbital & semimajor & minimum & orbital + @xmath6 & designation & period & axis & mass & eccentricity + & & @xmath7 ( days ) & @xmath8 ( au ) & @xmath9 ( @xmath10)@xmath11 & @xmath12 + 1 & c & 9.6386 & 0.091 & 0.03321 & 0.172 + 2 & d & 310.55 & 0.921 & 0.5219 & 0.0666 + 3 & b & 643.25 & 1.497 & 1.676 & 0.128 + 4 & e & 4205.8 & 5.235 & 1.814 & 0.0985 + + + although the eccentricities of its planetary orbits are larger than those of 55 cnc ( see tables 1 and 5 ) , @xmath25 ara appears to be a good example of an shd system . the size of the system is also comparable to that of 55 cnc ( @xmath196 au ) , so we expected our best fit model to produce parameter values similar to those of 55 cnc . we also thought that , by analogy , a pps model of the system would find a stable zone in the 3.7 au gap between planets b and e , and a tb model would predict a planet in the same gap . such models are not yet available , but the tb relation of poveda & lara ( 2008 ) can be easily applied to the observed semimajor axes . it turns out that the tb models are not particularly good fits to the data with or without a hypothetical 5@xmath116 planet in the 3.7 au gap . the planetary orbits of @xmath25 ara shown in table 5 can not be matched to the density maxima in any of our models . the reason for this failure is that planet d is too close to planet b , and this leaves an inconspicuous 0.8 au gap in the distribution between planets c and d. a hypothetical planet is needed in this gap to restore the regular spacing of the inner planets . this , of course , does not mean that an undetected planet necessarily exists in the 0.8 au gap , it just implies that a density maximum is needed in that area of the model for the distribution to be typical of a well organized shd system . by the same token , some models allow for yet another density maximum interior to the orbit of planet c , at @xmath190.05 au . such experimental models are characterized by different parameter sets , so a `` best fit model '' can not be determined , unless additional planets are discovered in @xmath25 ara by future observations . a surface density profile of the form @xmath194 is applicable outside the central core region of the protoplanetary disk of 55 cnc and the solar nebula ( table 4 ) . the presence of a roughly uniform core is strongly suggested by both theory and observations ( see [ th].6 ) , and our models suggest that llds and shds can be distinguished by their core sizes ( @xmath191 au and @xmath190.05 au , respectively ) . on the other hand , we do not know why both of the nebular disks in table 3 show the @xmath178 trend , although this fact that we independently obtained the same result ( as opposed to two different power laws ) from two different planetary systems argues in favor of our modeling assumptions . at this point , we do not have any more evidence of the ubiquity of the @xmath178 profile among llds and shds , but we can certainly argue that this density power law is more relevant to planet formation than the profiles from steady state viscous and irradiated accretion disks ( kitamura et al . [ 2002 ] , andrews & williams [ 2007 ] , garaud & lin [ 2007 ] , and references therein ) . the isothermal phase of protostellar collapse that our models address is an extended intermediate phase of the whole process during which the magnetic fields have diffused away , the central core mass is a minute fraction of the solar mass , matter continues being deposited onto the disk from above and below , most of the gas in the disk is molecular , and all the energy produced is efficiently radiated away ( see tohline [ 2002 ] for a recent review ) . at this stage , the evolution of the gas is not driven by viscosity or by ionization because the gas is transparent to radiation ; the gas responds fully to the effects of self gravity and ( non keplerian ) rotation , and this is why it is very encouraging to know that our modeled disks are dynamically stable ( [ stability ] ) . theoretical calculations of viscous , magnetized , or irradiated accretion disks are not applicable in this regime of protostellar collapse , while observations of thermal radiation from t tauri stars probe the disks long after the isothermal phase is over , the gas has become opaque to its own radiation , and the grown protostars have imposed nearly keplerian rotation profiles onto the disks . therefore , there is no reason to expect that the surface density profiles should have power law indices of @xmath195 or softer ( seen in t tauri disks ; kitamura et al . 2002 ; andrews & williams 2007 ) during the isothermal phase of collapse ; and if the planetary results of kuchner ( 2004 ) give any indication about extrasolar systems , we should expect instead @xmath196indices between @xmath165 and @xmath197 . adams , f. c. , laughlin , g. , & bloch , a. m. 2008 , , 683 , 1117 alibert , y. , baraffe , i. , benz , w. , et al . 2006 , a&a , 455 , l25 alibert , y. , mordasini , c. , & benz , w. 2004 , a&a , 417 , l25 alibert , y. , mordasini , c. , benz , w. , & winisdoerffer , c. 2005a , a&a , 434 , 343 alibert , y. , mousis , o. , mordasini , c. , & benz , w. 2005b , apj , 626 , l57 andrews , s. m. , & williams , j. p. 2007 , , 659 , 705 armitage , p. j. 2003 , , 582 , l47 armitage , p. j. , livio , m. , lubow , s. h. , & pringle , j. e. 2002 , mnras , 334 , 248 barnes , r. , gozdiewski , k. , & raymond , s. n. 2008 , , 680 , l57 barnes , r. , & quinn , t. 2004 , , 611 , 494 barnes , r. , & raymond , s. n. 2004 , , 617 , 569 bean , j. l. , mcarthur , b. e. , benedict , g. f. & armstrong , a. 2008 , , 672 , 1202 bodenheimer , p. , hubickyj , o. , & lissauer , j. j. 2000 , icarus , 143 , 2 bodenheimer , p. , & pollack , j. p. 1986 , icarus , 67 , 391 bryden , g. , beichman , c. a. , trilling , d. e. , et al . 2006 , , 636 , 1098 butler , r. p. , marcy , g. w. , williams , e. , hauser , h. , & shirts , p. 1997 , , 474 , l115 butler , r. p. , wright , j. t. , marcy , g. w. , et al . 2006 , , 646 , 505 christodoulou , d. m. , & kazanas , d. 2008 , , submitted ( astro - ph/0706.3205 ) ( ck ) christodoulou , d. m. , shlosman , i. , & tohline , j. e. 1995 , , 443 , 551 davis , s. s. 2005 , , 627 , l153 desort , m. , lagrange , a .- m . , galland , f. , et al . 2008 , a&a , preprint ( astro - ph/0809.3862 ) eggenberger , a. , udry , s. , & mayor , m. 2004 , a&a , 417 , 353 fischer , d. a. , marcy , g. w. , butler , r. p. , et al . 2008 , , 675 , 790 garaud , p. , & lin , d. n. c. 2007 , , 654 , 606 gayon , j. , marzari , f. , & scholl , h. 2008 , , 389 , l1 gonzalez , g. , & vanture , a. d. 1998 , a&a , 339 , l29 hayashi , c. 1981 , prog . suppl . , 70 , 35 ida , s. , & lin , d. n. c. 2004a , apj , 604 , 388 ida , s. , & lin , d. n. c. 2004b , apj , 616 , 567 ida , s. , & lin , d. n. c. 2005 , apj , 626 , 1045 jayawardhana , r. , holland , w. s. , kalas , p. , et al . 2002 , , 570 , l93 kitamura , y. , momose , m. , yokogawa , s. , et al . 2002 , , 581 , 357 kuchner , m. j. 2004 , apj , 612 , 1147 laskar , j. 1996 , celest . astron . , 64 , 115 lecoanet , d. , adams , f. c. , & bloch , a. m. 2008 , , preprint ( astro - ph/0810.4076 ) lin , d. n. c. , papaloizou , j. c. b. , terquem , c. , bryden , g. , & ida , s. 2000 , protostars and planets iv , ed . v. mannings , a. p. boss , & s. s. russell ( univ . of arizona press : tucson ) , 1111 lissauer , j. j. 1987 , icarus , 69 , 249 lissauer , j. j. 1993 , , 31 , 129 lissauer , j. j. 1995 , icarus , 114 , 217 marcy , g. w. , butler , r. p. , fischer , d. a. , et al . 2002 , , 581 , 1375 marcy , g. , butler , r. p. , fischer , d. , et al . 2005 , prog . suppl . , 158 , 24 marcy , g. w. , cochran , w. d. , & mayor , m. 2000 , protostars and planets iv , ed . v. mannings , a. p. boss , & s. s. russell ( univ . of arizona press : tucson ) , 1285 mcarthur , b. e. , endl , m. , cochran , w. d. , et al . 2004 , , 614 , l81 pepe , f. , correia , a. c. m. , mayor , m. 2007 , a&a , 462 , 769 pierens , a. , & nelson , r. p. 2008 , a&a , 482 , 333 pollack , j. b. , hubickyj , o. , bodenheimer , p. , et al . 1996 , icarus , 124 , 62 poveda , a. , & lara , p. 2008 , rmxaa , 44 , 243 raghavan , d. , henry , t. j. , mason , b. d. , et al . 2006 , apj , 646 , 523 raymond , s. n. , & barnes , r. 2005 , , 619 , 549 raymond , s. n. , barnes , r. , & gorelick , n. 2008 , , , preprint ( astro - ph/0808.3295 ) raymond , s. n. , barnes , r. , & kaib , n. a. 2006 , , 644 , 1223 santos , n. c. , bouchy , f. , mayor , m. , et al . 2004 , a&a , 426 , l19 schneider , g. becklin , e. e. , smith , b. a. , et al . 2001 , , 121 , 525 tohline , j. e. 2002 , , 40 , 349 trilling , d. e. , lunine , j. i. , & benz , w. 2002 , a&a , 394 , 241 valenti , j. a. , & fischer , d. a. 2005 , apjs , 159 , 141 vogt , s. s. , butler , r. p. , marcy , g. w. , et al . 2005 , , 632 , 638 ward , w. r. , & hahn , j. m. 2000 , protostars and planets iv , ed . v. mannings , a. p. boss , & s. s. russell ( univ . of arizona press : tucson ) , 1135 weidenschilling , s. j. 1977 , a&ss , 51 , 153 zhou , l .- y . , ferraz - mello , s. , & sun , y .- s . 2008 , exoplanets : detection , formation and dynamics , proc . iau symposium no . sun , s. ferraz - mello , & j .- zhou ( cambridge univ . press : cambridge ) , 119 0.5truein

we review current ideas on entropy production during the different stages of a relativistic nuclear collision . this includes recent results on decoherence entropy and the entropy produced during the hydrodynamic phase by viscous effects . we start by a discussion of decoherence caused by gluon bremsstrahlung in the very first interactions of gluons from the colliding nuclei . we then present a general framework , based on the husimi distribution function , for the calculation of entropy growth in quantum field theories , which is applicable to the early ( `` glasma '' ) phase of the collision during which most of the entropy is generated . the entropy calculated from the husimi distribution exhibits linear growth when the quantum field contains unstable modes and the growth rate is asymptotically equal to the kolmogorov - sina ( ks ) entropy . we outline how the approach can be used to investigate the problem of entropy production in a relativistic heavy - ion reaction from first principles . we show that the same result can be obtained in the framework of a completely different approach called _ eigenstate thermalization hypothesis_. finally we discuss some recent results on entropy production in the strong coupling limit , as obtained from ads / cft duality .

1110.2378

high - energy heavy - ion collisions show a vary rapid transition from a quantum mechanical initial state to a thermalized state known as a quark - gluon plasma . this poses a fundamental , conceptual theory problem , namely how the necessary production of entropy can be reconciled with the t - invariance of quantum chromodynamics ( qcd ) . one could argue in analogy to the situation in the einstein - rosen - podolsky paradox that all produced particles stay in a highly entangled state which only collapses when a measurement is performed and that entropy production occurs only in that moment . nevertheless , a hydrodynamic description of the quark - gluon plasma seems to be possible long before any measurement takes place . the resolution of this apparent contradictionis of interest for a wide spectrum of physical phenomena , e. g. , reheating of the early universe after the end of the presumed period of inflationary growth , when the universe transitioned from an exponentially inflated vacuum bubble to the hot cosmos we observe today . another example is the gravitational collapse of matter into a black hole , which behaves as if it were a thermal object . in this review we will focus on the thermalization of strongly interacting matter after the collision of two relativistic heavy nuclei , thereby forming a quark - gluon plasma , a process which has the immense advantage that it can be studied in great detail in laboratory experiments . for example it is known that the quark gluon matter in a heavy - ion collision behaves like a nearly ideal thermal fluid already after @xmath0 fm/@xmath1 . justifying the use of thermodynamics in the context of quantum field theory is an extremely difficult task in any case , even for phenomena described by quantum electrodynamics ( qed ) , where the quantum field theory is under excellent control . for qcd the situation is even more difficult due to the strong coupling dynamics , which permits analytical calculations only in very much constrained situations . to illustrate the extent of the problem let us note that even in quantum mechanics the problem of thermalization is far from being settled . a nice example is a recent study@xcite , where a finite one - dimensional quantum system was shown to thermalize only when the so called `` eigenstate thermalization hypothesis '' ( eth@xcite ) is satisfied . rigol@xcite stresses that quantum and classical thermal states have very different natures . eth claims that each eigenstate of the hamiltonian is related to a thermal state which is reached by the `` de - phasing '' of the coherent components of the quantum state . while we do not want to discuss the merits of this specific concept in more detail , it serves as an example illustrating the fact that the de - phasing or decoherence of a quantum state constitutes one way ( if not _ the _ way ) to generate entropy . generated entropy generally is proportional to the number of lost bits of information . consequently phases carry an important part of the information contained in a quantum state , their loss generates a significant amount of entropy . a general mechanism for losing information about the state of a quantum system is the entanglement of its wave function with that of its environment . this mechanism obviously does not apply to an isolated system . the conservation of the von neumann entropy for an isolated system implies that any information encoded in its quantum state is never lost , at least in principle . however , the information may become encoded in observables of such complexity or requiring such a degree of experimental precision that its retrieval is impractical . a standard approach to describe this effective loss of information in an isolated system is _ coarse graining_. by averaging over finite regions of phase space , information is lost , and thus entropy is produced . often , coarse graining is discussed in connection with measurements performed on a system . however , in section 2 we will discuss how even in the absence of physical measurements the uncertainty principle implies a natural , measurement independent coarse graining . it thus implies effective entropy production for nonlinear , time - reversal invariant systems , in which individual trajectories approach all points in some region of phase space , while all ensembles of trajectories respect liouville s theorem , i.e. occupy a constant phase space volume . this argument also applies to the information residing in the phases between the different components of a quantum system , which are smeared by coarse graining . thus decoherence or de - phasing plays a crucial role for the discussion of thermalization and entropy production , which naturally leads to the question whether decoherence alone is sufficient to reach thermalization . this is not the case , because decoherence may lead to systems with statistical features , which have energy and momentum distributions that are far from equilibrium . heavy - ion collisions provide a good example for such systems . after decoherence of the initial parton distributions in the colliding nuclei one is left with some sort of quark - gluon matter that can , for most part , be treated statistically and semi - classically . the natural question is whether the equations of fluid dynamics can be applied to describe the subsequent evolution of this matter . fluid dynamics is an effective theory for the time evolution of the expectation value of the components of the energy - momentum tensor @xmath2 , which encodes the dynamics of the system in terms of macroscopic quantities : energy density , pressure , collective flow velocity , and the equation of state . its applicability requires that spatial and temporal gradients of these quantities are small , and that the deviations of the microscopic statistical state of the system from local thermal equilibrium are small and can be represented by viscous ( gradient ) corrections describing anisotropies of the stress tensor . the early - time energy momentum tensor for heavy - ion collisions ( colliding along the @xmath3 axis ) obtained by solving the classical yang - mills equations of motion has the approximate form@xcite @xmath4 with negative longitudinal pressure , which makes it unsuitable for hydrodynamical calculations . the negative longitudinal pressure is an expression of the fact that gauss law does not permit all components of the gauge field to decohere , because they are not independent of the field generating color charges . in this case , secondary gluon production depletes the remaining coherent longitudinal fields and eventually generates positive longitudinal pressure . only then can the system establish acceptable initial conditions for the following hydrodynamical evolution . the particle production process and the interaction among the produced particles therefore must be included in the description of the thermalization process . a natural question to ask in this situation is , how quickly the components of the stress tensor develop a sufficient degree of isotropy in the locally co - moving frame for a hydrodynamical description to be valid , and how this _ isotropization time _ , @xmath5 , compares to the _ decoherence time _ , note that isotropization is a less stringent requirement than thermal equilibration . in principle , an isotropic stress tensor could be generated by an isotropic , yet strongly non - thermal gluon distribution . indeed , if one considers a dilute ensemble of de - phased particles , which is described by the boltzmann equation , in a state near thermal equilibrium , its relaxation to equilibrium is governed by an infinite set of relaxation times and their associated modes . because of the local conservation of energy and momentum , the viscous modes are usually those which survive at late times , and thus entropy growth at late time is usually controlled by the viscosities and the thermal conductivity . but this does not have to be the case when the system is far off equilibrium , where unstable , exponentially growing modes may exist that do not permit a linearized description . such processes can somtimes be interpreted as run - away ( stimulated ) particle production . while particle production is usually not interpreted as decoherence , it may in fact be related to it . for example , bremsstrahlung can be understood as decoherence of the coherent , quasi - real weizscker - williams fields associated with any fast moving charge . while such radiation does not result in exponential growth in qed in the vacuum , it can do so in the presence of strong fields ( e.g. in a free electron laser ) or in the nonlinear dynamics of an electromagnetic plasma or a quark - gluon plasma far off equilibrium . the interpretation of how entropy is generated in such processes may then depend on the theoretical and conceptual framework in which the processes that lead to thermalization are described . the question of whether , how , and how fast a highly excited state of qcd matter equilibrates thermally is thus a multi - faceted one , and we certainly do not profess to fully understand all relevant points of view . below , we will outline and discuss those we are aware of , without claiming completeness . a general framework for the study of thermalization processes starts from the observation that thermal equilibrium corresponds to the state of highest entropy . our approach thus focuses on the rate at which entropy is produced in the system . as already mentioned , the von neumann entropy of an isolated quantum system never increases , because the time evolution of an isolated quantum state is unitary . however , when we speak of the thermalization of an isolated system , we mean that its state `` looks '' thermal to most practically feasible experiments that measure a select number of observables . owing to the lack of complete observation we replace the microscopic , possibly pure quantum state of the system by a density matrix that traces over all unmeasured observables . by this coarse graining procedure we assign an entropy to the system , which can grow as a function of time if the microscopic structure of the quantum state gets hidden in more and more complex , and hence practically unobservable , degrees of freedom . the coarse grained , apparent entropy is then a measure of the complexity of the microscopic state of the system . the generic time evolution of the coarse grained entropy for a dynamical system with ergodic properties is sketched in fig.[fig : as1 ] . at very early times , the behavior is extremely sensitive to details of the initial state and will depend on the initial phase correlations and the occupation probabilities of the various eigenmodes of the system . if all modes are randomly occupied the length of this period can be estimated@xcite to be of order @xmath7 , where @xmath8 is the number of independent modes . the period of approximately linear rise of the coarse grained entropy is governed by the sum of all positive lyapunov exponents of the system in the classical limit , the so - called kolmogorov - sina ( ks ) entropy growth rate . the rise starts when the amplitude of the unstable modes has outgrown the average mode occupation , and it terminates when the coarse grained entropy approaches the equilibrium entropy . while the rate of growth of the entropy is an intrinsic property of the system , the total duration of the quasi - linear period depends on the specific nature of the initial state . finally , the last phase is governed by the relaxation times of small deviations from equilibrium . the dynamical evolution of the quark - gluon plasma created in a relativistic heavy ion collision is commonly described by relativistic viscous hydrodynamics , which is well suited to describe the last , asymptotic phase sketched in fig.[fig : as1 ] . obviously , the question , when the transition from the microscopic description to the hydrodynamical one should be taken , has no unique answer but has to be decided on the basis of heuristic and practical considerations . in principle , the initial conditions for the hydrodynamic simulation will depend on which moment is chosen as the starting time . however , if hydrodynamics applies at all , the late time results of the simulation will depend only little on when exactly the transition from a microscopic description to hydrodynamics is made . according to our discussion , the only constraint is that this time should be chosen close to the end of the phase of linear entropy growth . in fact , the situation is even more involved . in practice , the information about the supposedly thermal character of the final state is obtained from single - particle observables , including emitted particle yields , spectra , and angular distributions , as well as of selected two - particle correlation functions , such as balance functions . these observables are all of hadronic nature ; thus , one also has to take into account the entropy production which occurs during hadronization , as well as that caused by interactions in the expanding hadron gas before freeze - out . these final one - particle or limited two - particle observables trace over a very large number of degrees of freedom and thus imply a highly coarse grained description in terms of the fundamental degrees of freedom . however , we are not so much interested in this late state as in the question when the transition to a hydrodynamical description is justified early in the reaction . for this question , a coarse grained entropy definition that is dictated by the exigencies of final - state measurements , long after the collision is over , makes little sense , because a thus defined entropy may have very little relevance for the dynamics governing the time evolution of the system . this suggests that it may be useful to study entropy growth under a _ minimal _ amount of coarse graining ( as imposed by the uncertainty principle , see below ) and to use this approach to estimate the rate at which the coarse grained entropy approaches its thermal equilibrium value . the different conceptual approaches to the thermalization problem are to some extent complementary . obviously , when the system truly equilibrates , all approaches will indicate thermalization . however , isotropy of its macroscopic properties alone is a poor measure of thermal equilibrium of a quantum system , nor does the growth of the coarse grained entropy ensure that the system approaches local isotropy . a fluid dynamical description with a thermal equation of state requires both . in this review , we focus on the progress of our understanding of how and when the apparent entropy in the final state of a relativistic heavy ion collision is generated by the internal dynamics of the highly excited qcd matter . however , we caution the reader that , while substantial progress has been made over the last years , we are still far from having a complete answer to the problem of thermalization . as already mentioned , much of our information on the behavior and properties of hot qcd matter is derived from measurements of the particle yields and spectra in the final state of relativistic heavy ion collisions and their interpretation in terms of thermodynamic and hydrodynamic concepts . at the energies available at the relativistic heavy ion collider ( rhic ) and the large hadron collider ( lhc ) one of the most relevant quantities is the azimuthal quadrupole anisotropy of the collective flow , usually called _ elliptic flow _ and denoted as @xmath9 . an excellent agreement of hydrodynamical calculations of the flow anisotropy of the matter produced in the nuclear collisions with the elliptic flow measurements requires the assumption of a rapid thermal equilibration of the matter on a time scale of the order of 1 fm / c . an important problem in the description of relativistic heavy ion reactions is thus to understand how the produced matter equilibrates so quickly , _ i.e. _ , when and how entropy is created in the reaction . in principle , one can distinguish five different stages of entropy production : 1 . decoherence of the initial nuclear wave functions ; 2 . thermalization of the partonic plasma ( `` glasma '' ) ; 3 . dissipation due to shear viscosity in the hydrodynamical expansion ; 4 . hadronization accompanied by large bulk viscosity ; 5 . viscous hadronic freeze - out . indicate the entropy per unit rapidity reached at the end of various collision stages based on experimental data and model estimates.,scaledwidth=85.0% ] our present incomplete knowledge about the contribution of these different stages to the final entropy is depicted in fig . [ fig : bm1 ] . before we discuss these stages in turn , we note that the final entropy per unit rapidity @xmath10 is one of the best known quantities in relativistic heavy ion physics . in the case of au+au collisions at rhic the entropy distribution @xmath10 at particle freeze - out can be determined from an analysis of the final hadron spectra in combination with the information on the source radius derived from identical particle ( hbt ) correlations.@xcite the slightly extrapolated result for the 6% most central au+au collisions at @xmath11 gev is @xmath12 at midrapidity.@xcite alternatively , the final entropy can be deduced from the measured hadron abundances , combined with the calculated entropy per particle for a hadron gas in chemical equilibrium at @xmath13 mev , @xmath14 , which yields the result@xcite @xmath15 . the 10 percent difference between these values can be attributed to the entropy production during the hadronic freeze - out and reflects the entropy production due to the decay of excited hadronic resonance states as well as the contribution from the substantial shear viscosity of a thermal hadron gas.@xcite entropy production by decoherence dominates the very first stage of a heavy ion collision . the loss of coherence is measured by the decay of off - diagonal elements of the density matrix @xmath16 describing the system . a practical way of investigating this process is to calculate the decay rate of the quantity @xmath17 ^ 2 $ ] . the time scale of the decoherence of the initial nuclear wave function has been studied by this method in the color glass condensate model ( cgc).@xcite in this model , the initial nuclear gluon distributions are coherent over a wide range of longitudinal momenta @xmath18 ( in the beam direction ) due to the high degree of lorentz contraction of the fast moving nuclei . the elementary process driving the decay of the off - diagonal elements of the density matrix is the fusion of one gluon from each nucleus , which removes gluons from the initial - state wave functions . the relevant time scale in this process is set by the saturation scale @xmath19 of the nuclear gluon distribution . @xmath19 is a measure of the average transverse momentum of gluons in a fast moving nucleus . gluons with transverse momentum @xmath20 gev can be thought of as components of a quasi - classical gauge field . a careful analysis yields the result that the characteristic decoherence time is @xmath21 , where @xmath1 denotes a calculable constant close to unity.@xcite while this result is expected on dimensional grounds , the fact that @xmath22 is important because it shows that decoherence occurs over a time of less than 0.2 fm / c . here we do not want to present the complete chain of arguments leading to this result , but only provide a few details showing how the cgc model enters into the calculation . we study perturbative gluon interactions between the gluons in a nucleus at rest ( nucleus 1 ) and a fast moving one ( nucleus 2 ) . the time evolution of the density matrix describing nucleus 1 reads @xcite @xmath23 the cgc model enters by specifying the correlator @xmath24 between gauge fields in nucleus 2 as @xmath25 where @xmath26 this form implies decoherence in transverse momentum but coherence in longitudinal momentum . for the gluon field correlator in nucleus 1 ( at rest ) one makes a simple gaussian ansatz @xmath27 it turns out that a finite result is only obtained if the following running of the qcd coupling constant is used@xcite @xmath28 where @xmath29 is the infrared cutoff of the cgc , which cancels in the ratio @xmath17 ^ 2 $ ] . this ratio then turns out to be inversely proportional to the life - time of the system , and one can define @xmath6 as the time after which this ratio has dropped by a factor @xmath30 . the amount of entropy created from the loss of coherence of the components of the nuclear gluon wave functions can be estimated as follows . the complete decoherence of a coherent quantum state with average occupancy @xmath31 results in a mixed state with entropy@xcite @xmath32 . the decoherence occurs in transverse domains of spatial size @xmath33 , which are causally disconnected during the decoherence process . the number of causally disconnected transverse domains in a central au+au collision is of order @xmath34 , where @xmath35 denotes the nuclear radius . accounting for a longitudinal coherence length @xmath36 and using @xmath37 one finds:@xcite @xmath38 at midrapidity for a central au+au collision this is roughly one quarter of the observed final entropy per unit rapidity . the other stage which is reasonably well understood is the hydrodynamic expansion stage . rhic data on the elliptic flow of hadrons in non - central collisions indicate that the shear viscosity during this phase is small . the bounds on the shear viscosity @xmath39 compatible with the rhic data lie characteristically in the range @xmath40.@xcite such a small viscosity can not increase the total entropy of the expanding fluid by much . systematic studies of the entropy growth in longitudinally boost - invariant , viscous fluid dynamics for different initial conditions were performed by fries _ et al_.@xcite the viscous stress - energy tensor in the locally co - moving frame has the form @xmath41 . the transverse and longitudinal pressures can be decomposed into the thermodynamic pressure @xmath42 and the components @xmath43 and @xmath44 of the shear and bulk stress : @xmath45 ; @xmath46 . at early times , @xmath43 and @xmath44 are given by the initial conditions for the stress tensor established during the thermalization process . in view of the dilution effect of the longitudinal expansion imprinted on the matter by the collision , the pressure components are expected to satisfy the ordering @xmath47 . the smallest physically meaningful value of the longitudinal pressure is @xmath48 , which arises when the matter is completely in the form of coherent longitudinal fields.@xcite however , the fluid dynamical approach does not apply to such a situation , because it assumes small deviations from local equilibrium . the minimal condition for fluid dynamics to be applicable is the mechanical stability condition : @xmath49 , but the regime where the fluid dynamical description can be trusted only begins when @xmath50 . the equations governing the longitudinal expansion of the medium are:@xcite @xmath51 @xmath39 and @xmath52 are the shear and bulk viscosity , respectively . the relaxation times for the shear and bulk stress , @xmath53 and @xmath54 , were assumed to be those derived from the boltzmann equation , @xmath55 was taken from the supersymmetric gauge theory ( see fries _ et al._@xcite for details ) . the entropy per unit rapidity and transverse area , @xmath56 obeys the equation:@xcite @xmath57 fries _ et al._@xcite started the hydrodynamical simulations after an initial time @xmath58 fm/@xmath1 , which is compatible with the expected decoherence time but considerably shorter than the equilibration times estimated from the elliptic flow analysis . the initial energy density was fixed at @xmath59 gev/@xmath60 . the early starting time was chosen to explore , in a schematic way , how long it would take to reach the domain of applicability of the hydrodynamical description defined by the conditions @xmath61 fries _ et al._@xcite considered three different initial conditions : * @xmath62 , corresponding to thermal equilibrium at @xmath63 ; * @xmath64 and @xmath65 , corresponding to the shear stress of the relativistic navier - stokes theory ; * @xmath64 as in ( ii ) , but @xmath66 for vanishing initial longitudinal pressure . typical results from these calculations are shown in fig [ fig : visc - hydro ] . the lower panel ( a ) indicates that the condition ( [ eq : hydro - cond ] ) is approximately satisfied for @xmath67 fm/@xmath1 independently of the choice of the initial condition for the stress tensor . on the other hand , the upper panel ( b ) shows that most of the viscous entropy production occurs before this time during a period when the validity of the hydrodynamical approach is questionable . the relative contribution to the final entropy from viscous forces after @xmath68 does not exceed 10 percent for the parameter choices explored in this calculation . ( dashed ) and @xmath69 ( solid ) , as functions of time @xmath70 for the initial conditions ( i ) ( black ) , ( ii ) ( blue ) , and ( iii ) ( red ) . ( b ) relative entropy production from bulk and shear stress , @xmath71 ( solid ) and @xmath72 ( dashed ) for the same scenarios . for further details , see fries _ et al_.@xcite ] the same conclusion was reached by song and heinz@xcite in a two - dimensional boost - invariant calculation ( see fig . [ fig:2d - hydro ] ) . we may thus conclude conservatively that the entropy at the moment of local equilibration must be @xmath73 . these considerations tell us that approximately half of the final entropy must be generated during the thermalization process , which can not be described by fluid dynamics . we next discuss various theoretical approaches to the entropy growth rate during the thermalization stage . , for two - dimensional boost invariant viscous hydrodynamical expansion of a quark - gluon plasma for three different equations of state . the entropy increases by approximately 10 percent , if the shear visscosity is equal to the kss bound @xmath74 . for further details , see song and heinz@xcite . ] if the quark - gluon plasma were a weakly coupled system , the growth of the entropy during the equilibration phase could be calculated from a partonic boltzmann equation . however , the small shear viscosity and other observations from the rhic experiments , such as measurements of jet quenching , indicate that the plasma is strongly coupled . in addition , it is thought that the state created by the decoherence of the nuclear gluon wave functions ( the _ glasma_@xcite ) is characterized by strong , still partially coherent color fields . such fields may even be regenerated by plasma instabilities in the pre - equilibrium quark - gluon plasma.@xcite a formalism that professes to describe entropy production during this early stage must , therefore , be able to describe the growing complexity of a quantum system as it makes the transition from a regime of gauge field dominance to the hydrodynamical regime characterized by complete randomness on thermal length scales . a general formalism that can describe this transition in terms of the gauge field dynamics has been proposed by kunihiro _ et al_.@xcite the idea underlying this approach is that the growing entropy measures the increasing intrinsic complexity of the quantum state of the system after appropriate coarse graining . as mentioned in the introduction , the problem is how to impose a minimal amount of coarse graining without assuming the answer . a general solution to this problem dates back to husimi.@xcite the husimi distribution is defined as a convolution of the wigner function @xmath75 of the system with a minimum - uncertainty gaussian wave packet : @xmath76 here @xmath77 and @xmath78 stand for all `` position '' and `` momentum '' variables characterizing the system , which may include particle positions as well as field amplitudes.@xcite the husimi distribution depends on the squeezing parameter @xmath79 , which measures the ratio of position and momentum uncertainty . its value can be chosen at liberty , but once fixed , will not change with time . the husimi distribution can be understood as a coarse - grained phase space distribution of the quantum system , where the coarse graining corresponds to the projection on a coherent state . we recall that coherent states are the closest quantum analogues of classical systems compatible with the uncertainty relation . because the husimi distribution can be shown to be positive ( semi-)definite , it permits the definition of a coarse - grained entropy , first introduced by wehrl:@xcite @xmath80 quantum systems containing unstable modes , _ i.e. _ modes with an exponentially growing amplitude , have a linearly growing husimi - wehrl entropy.@xcite the entropy growth rate is found to be independent of the squeezing parameter @xmath79 and thus independent of details of the coarse graining . it is given by the sum of the exponential growth rates of all unstable modes . in classical dynamical systems , this quantity is known as the kolmogorov - sina entropy , or ks entropy , and defined as the sum over all positive lyapunov exponents @xmath81 of the system : @xmath82 the ks entropy is understood to be a measure of the growth rate of the coarse grained entropy of a dynamical system starting from a configuration far away from equilibrium , after an initial start - up phase during which unstable fluctuations grow to dominance and before it gets too close to its micro - canonical equilibrium.@xcite and @xmath83 . the horizontal axis denotes the scaled position @xmath84 ; the vertical axis represents the momentum @xmath78.,title="fig : " ] and @xmath83 . the horizontal axis denotes the scaled position @xmath84 ; the vertical axis represents the momentum @xmath78.,title="fig : " ] and @xmath83 . note that the extent of the distribution in the off - diagonal direction ( @xmath85 ) does not shrink beyond the resolution limit set by the gaussian smearing introduced by the husimi transform.,title="fig : " ] and @xmath83 . note that the extent of the distribution in the off - diagonal direction ( @xmath85 ) does not shrink beyond the resolution limit set by the gaussian smearing introduced by the husimi transform.,title="fig : " ] as stated in the introduction , the eigenstate thermalization hypothesis@xcite ( eth ) states that an isolated quantum system thermalizes under its own internal dynamics if every energy eigenstate contains a thermal component . more precisely , the eth posits@xcite that the expectation value @xmath86 of a few - body observable @xmath87 in an energy eigenstate @xmath88 of the hamiltonian of a large , interacting many - body system equals the microcanonical average of @xmath87 at the mean energy e : @xmath89 for a system with very many degrees of freedom , the microcanonical average is generally agrees with the thermal average for an appropriate choice of the temperature . a corrollary of this hypothesis is that a state that is an eigenstate of a few - body observable @xmath87 will be a superposition of energy eigenstates @xmath90 with probabilities @xmath91 that thermally distributed : @xmath92 . as the state evolves with time , @xmath93 the phases of the different energy components increasingly diverge from each other . the contributions from different energy eigenstates thus effectively decohere at late times . any measurement of a few - body observable on the pure quantum state yields a result that is indistinguishable from the results of a measurement on a mixed state defined by the diagonal density matrix@xcite @xmath94 rigol _ et al._@xcite argued on the basis of this argument that the coarse grained entropy of a pure quantum state immediately after a quench can be calculated as the entropy associated with the diagonal energy density matrix in the hamiltonian after the quench : @xmath95 however , this argument does not take into account the time evolution of the phases of the different energy components , which only become quasi - random numbers at late times . at a time @xmath96 after the quench , the phases of energy eigenstates within an energy band @xmath97 differ by less than @xmath98 , and thus these components of the wave function remain approximately coherent . the time evolution of the coarse grained entropy should therefore be approximately given by the following expression : @xmath99 where @xmath100 denotes an energy band of width @xmath101 and @xmath102 as time progresses , @xmath101 continues to shrink , until it is eventually smaller than any spacing between energy eigenvalues of the system . thus @xmath103 defined in ( [ eq : st ] ) approaches the asymptotic value ( [ eq : sdiag ] ) at late times . however , immediately after the quench , @xmath104 , and the coarse grained entropy only increases gradually as more and more energy eigenstate components dephase from each other . we will show in the next subsection quantitatively how this mechanism operates in the case of an exactly solvable quantum quench . the ideas behind the husimi - wehrl entropy are best illustrated by a simple example . we choose the inverted harmonic oscillator because this case is so simple that we can analyze it also in a completely different approach built on the _ eigenstate thermalization hypothesis _ ( eth ) . for the inverted harmonic oscillator @xmath105 we start from a gaussian wave packet @xmath106 and calculate @xmath107 analytically . the result is shown in figs . [ fig : husimi_1 ] and [ fig : husimi_2 ] , which illustrate how the phase space volume stays constant for the wigner function but increases for the husimi function because the collapse in the shrinking directions is halted by the resolution @xmath79 . one finds that@xcite @xmath108 it is noteworthy that this results is independent of @xmath79 and thus independent of the details of the coarse graining . next we show that on gets the same result in the approach based on the eigenstate thermalization hypothesis . the basic concept underlying the eth is that every eigenstate of a hamiltonian contains a thermal state . the initial quantum state is a coherent superposition of such eigenstates with the property that this thermal component does not contribute to physical observables . if coherence is lost this state becomes visible and can dominate at late times @xcite . for the inverted harmonic oscillator we can use the wkb approximation @xmath109 for the gaussian initial state ( [ init ] ) one expands the initial state in energy eigenstates @xmath88 : @xmath110 the time evolution of a narrow energy band @xmath111 within the initial state is given by @xmath112 at time @xmath96 , the phases remain coherent within an energy band of width @xmath113 . the eth approach rests on the fact that this band shrinks with time @xmath96 . the coherence is controlled by the right - most sine function which for large times imposes the constraint @xmath114 and by the term @xmath115 in the exponent . the phase coherent bands in the variable @xmath116 are determined by the exponential , yielding the position uncertainty @xmath117 for the occupation probability of each coherent energy band this condition implies for large enough times @xmath96 : @xmath118 which implies the normalization @xmath119 . note the appearance of the thermal weight factor @xmath120 , as predicted by the eth . this result can be easily extended to include the block structure in @xmath77 : @xmath121 with the normalization @xmath122 from the block matrix of all coherent energy position bands @xmath123 it is then straightforward to calculate the coarse grained entropy for large times : @xmath124 \nonumber \\ & \approx & \int_0^{\infty } de~ \int_0^{\infty}dx~ [ \rho(e , x)\ln ( e^{\lambda t } ) ] ~=~ \lambda t ~~,\end{aligned}\ ] ] thus leading to a constant entropy growth rate @xmath125 in agreement with the results ( [ eq : dsdt ] ) obtained in the husimi formalism . apparently , dephasing of @xmath126 blocks has a similar effect as phase space smearing , at least in this toy example . this suggests that the phase of linear entropy growth may be a generic feature of entropy production in quantum field theories . the analysis of the previous sections suggests that entropy production is , in general , dominated by a phase of linear growth with time , where the slope is given by the kolmogorov - sina entropy growth rate ( in short , the ks entropy ) . we assume that this is also true for highly excited quark - gluon systems like those created in high - energy heavy - ion collisions . in this case , the ks entropy can be calculated from the dynamics of _ classical _ gauge fields . et al._@xcite studied the real - time dynamics of classical yang - mills fields numerically after replacing continuous space by a three - dimensional cubic lattice . on this lattice they investigated the hamiltonian @xmath127 where @xmath128 denotes the central difference operator in the @xmath129-direction . the classical equations of motion are then @xmath130 the spectrum of lyapunov exponents and the kolmogorov - sina entropy were determined with two different distance measures . as expected , the slope was found to be independent of the distance measure used , as shown in fig.[fig : ay1 ] . lattice for the two different distance measures . all scales are given in the lattice unit . ] in order to convert the results of the lattice simulations to dimensional , physical values , one needs to take into account that classical gauge theories have unphysical ultraviolet divergencies . this limitation of classical field theory is already encountered for point charges in classical electrodynamics . the physical reason is that classical point charges do not exist ; no particle can be localized better than within its de broglie wave length . thus the lattice spacing @xmath131 is not an unphysical parameter which should be sent to zero as in lattice simulations of the euclidean quantum field theory . rather it determines the length scale below which neglected quantum effects become relevant . there are various ways to determine @xmath131 . for su(@xmath132 ) gauge theories the inverse damping rate for infrared modes is e.g.@xcite @xmath133 for @xmath134 . alternatively one can compare the energy density on the lattice and in the continuum @xmath135 the fact that @xmath131 should scale like @xmath136 implies that the entropy slope in lattice units should scale like @xmath137 which it does , as shown in fig.[fig : ay6 ] . . ] kunihiro _ et al . _ also confirmed that the results for the lyapunov exponents are volume independent , see fig.[fig : ay2 ] . , and @xmath138 lattices with the same energy density . ] finally , in view of the discussion below it is worth mentioning that within classical yang - mills theory thermalization follows the bottom - up rather than the top - down scenario , i. e. soft modes grow stronger and thermalize faster than hard modes . more details can be found in kunihiro _ al._@xcite the conclusion from these investigations is that the thermalization time in a relativistic heavy - ion collision must be approximately 2 fm/@xmath1 and can hardly be shorter than 1 fm/@xmath1 . the success of the classical lattice simulation for the ks entropy of the non - abelian gauge field poses the question how generic this behaviour is . the lattice regularized yang - mills equations describing the dynamics of classical color fields are known to be strongly chaotic,@xcite and the ks - entropy of the classical yang - mills field was shown to be a thermodynamically extensive quantity.@xcite it could be that systems with less pronounced ergodic properties show a less clear - cut behaviour . even for quantum chromodynamics there are still many open issues . for example , one can extend the lattice approach to include gaussian fluctuations.@xcite it would also be interesting to evaluate the increase of the coarse grained entropy associated with the growth of the initial quantum fluctuations around classical gauge fields in the colliding nuclei.@xcite finally , let us comment on related work by berges _ et al._@xcite these authors analyzed the isotropization of the energy momentum tensor of classical yang - mills fields . in their fig . 8 , low momentum modes show a behaviour reminiscent of that observed by kunihoro _ _ with a similar isotropization time of @xmath0 fm/@xmath1 . on the other hand , the unexpected behaviour of high momentum modes may not be physically significant , because these modes are subject to lattice artefacts . the _ entanglement entropy _ is defined as follows : consider a quantum system @xmath139 composed of two complementary subsystems @xmath140 and @xmath141 such that @xmath142 , which is in a pure state @xmath143 . if the state of the subsystem @xmath140 is not a pure state , one says that the quantum states of @xmath140 and @xmath141 are entangled . the density matrix of the subsystem @xmath140 can be written as @xmath144 , and the entanglement entropy is defined as @xmath145 .\ ] ] this definition measures the amount of entanglement of the quantum state of @xmath140 with the quantum state of @xmath141 . obviously , for this definition to make sense @xmath146 must be equal to @xmath147 obtained by tracing the density matrix over @xmath140 . the relationship @xmath148 implies that the entropy can only depend on the area of the common surface separating @xmath140 from @xmath141 , which we call @xmath149 . if the entanglement entropy is proportional to the surface area , as explicit calculations in tractable quantum field theories show , this implies that @xmath150 , where @xmath151 is the proportionality constant . entanglement entropy thus contains by construction the holographic principle , making it one of the primary objects of interest for the thermodynamic properties of event horizons . the entanglement entropy was first quantitatively studied for a three - dimensional quantum field theory by srednicki , who found that the value of the entropy depends quadratically on the ultraviolet momentum cut - off of the field theory.@xcite the concept trivially generalizes to spatial dimensions other than @xmath152 . the one - dimensional case has been extensively studied for conformal field theories.@xcite in that case , the entanglement entropy of an interval of length @xmath153 in the vacuum state of the field theory is given , up to a possible constant , by the general formula@xcite @xmath154 where @xmath1 denotes the central charge of the conformal field theory and @xmath131 is the short - distance cut - off . for this simple case , the analogue to the apparent conflict in three dimensions between the volume dependence of thermal entropy and surface dependence of entanglement entropy can be resolved as follows . when the field is not in its vacuum state but in thermal equilibrium at @xmath155 , then @xmath146 measures not only the entanglement entropy with the surroundings of domain @xmath140 , but also the thermal entropy contained in @xmath140 . as expected , the thermal contribution is proportional to the volume of @xmath140 ( in one dimension simply the length @xmath153 of the interval ) . the thermal generalization of ( [ eq : s - entangled ] ) is@xcite @xmath156 which goes over into ( [ eq : s - entangled ] ) in the limit @xmath157 , and approaches the thermal entropy @xmath158 in the limit @xmath159 . when the original vacuum state is perturbed by a quantum quench , the entanglement entropy increases with time and eventually reaches the thermal value . in the limit of a near critical one - dimensional conformal field theory on an interval of length @xmath153 one finds@xcite @xmath160 where @xmath63 is the so - called extrapolation length . the final entropy corresponds to the temperature @xmath161 remarkably , the entanglement entropy reaches its equilibrium value after a finite time @xmath162 , which implies that the entangled information travels away from the center of the interval at the speed of light . the concept of entanglement entropy has an important application to black holes . in the theoretical investigation of the properties of black holes it was realized early on that thermodynamic consistency requires black holes to have some form of entropy , which was suggested by bekenstein@xcite to be proportional to its surface @xmath163 . the relation @xmath164 , which holds when some object falls into the black hole , requires black holes to have a nonzero temperature , e. g. @xmath165 for a schwarzschild black hole , which is exactly the hawking temperature . that the latter is of quantum origin ( notwithstanding the fact that some of its properties can be derived from thermodynamics ) underlines the fact that entropy can not be formulated consistently in a classical theory , as usually discussed in connection with the third law of thermodynamics . equating statistical entropy and thermodynamic entropy implies that the number of quantum states must be proportional to @xmath166 . this relation implies that all information contained in a black hole is encoded in the area of its horizon , which is a special case of the holographic principle.@xcite the latter states that the number of degrees of freedom @xmath8 , understood as the number of independent quantum states describing a region @xmath140 of space - time is bounded by the area @xmath167 of its boundary.@xcite many detailed investigations suggest that this definition may , indeed , avoid all apparent paradoxes of information flow across an event horizon that have been constructed over the years . for our discussion , two aspects are worth highlighting : * the surface entropy of a horizon can be identified , up to a constant factor , with the _ entanglement entropy_. * there exists a fundamental difference between a static black hole and a dynamically created and evaporating one , as the entanglement entropy ( which is often identified with the von neumann entropy ) can remain zero for the latter . thus the unitarity of the boundary gauge field theory might find its counterpart in in the unitarity of the black hole formation and evaporation process , see e. g. takayanagi and ugajin.@xcite there is a growing consensus that the formation and evaporation of a black hole is a unitary process , at least in asymptotically flat or ads space - times.@xcite this implies that the von neumann entropy stays constant throughout the process and information about the initial state can , in principle , be recovered from careful measurements of correlations within the hawking radiation.@xcite the apparent ( bekenstein ) entropy assigned to a black hole is an expression of the inability of a classical observer to distinguish between the many micro - states of the true quantum geometry . in the `` fuzzball '' picture@xcite of the black hole , the horizon encloses the region in which the average micro - state deviates strongly from the classical geometry . the finite size of the fuzzball is set by the uncertainty relation , which dictates that a micro - state carrying enough energy to represent the mass of the black hole will , in general , be spread out in space . one can easily imagine that a phase space smearing of a micro - state , similar to the husimi transform of the wigner function , would map the the pure state of the quantum geometry into the entropy carrying , quasi - thermal geometry of a classical black hole . the bekenstein entropy would then have to be understood as the coarse grained entropy for a transitory black hole created by gravitational collapse as well as the entanglement entropy of an eternal black hole . takayanagi and ugajin@xcite have argued that under certain circumstances the entanglement entropy of a transient black hole can remain zero . this could happen , e.g. , if the forming and decaying black hole is microscopic so that it decays into a small number of excitations whose detailed phase relationship can be observed . they construct a toy model in which `` black hole '' formation is a periodic process , which obviously does not entail entropy growth . this is a special case of the general observation that the concept of entropy for a finite system only makes sense on time scales much shorter than the poincar recurrence time . while this is practically true for most systems with many degrees of freedom , it does not apply to microscopic systems , and it may not apply to simplified systems that have been constructed to be mathematically tractable . we also note that the von neumann entropy of an eternal black hole may be non - zero . the fundamental difference between a stationary and a transitory black hole with respect to their entanglement entropy is illustrated in fig.[fig : entangle_1 ] . before the black hole is formed , @xmath140 and @xmath141 cover the full quantum system @xmath142 . thus entanglement entropy can be defined as the minimal surface between @xmath140 and @xmath141 . if a black hole develops inside of @xmath141 this minimal surface does not change at all ( case @xmath168 ) . as the process is unitary , @xmath141 undergoes some non - trivial internal reordering of states , but this does not affect the surface between @xmath140 and @xmath141 . however , in the presence of a stationary black hole , @xmath169 at early times and one has to define entanglement entropy from the minimal surfaces between the three regions @xmath140 , @xmath141 and @xmath170 ( case @xmath1 ) . ) containing a transitory black hole ; c ) one region ( b@xmath171 ) containing an eternal black hole , which can be considered as a third region of space ( b@xmath172 ) with its event horizon as natural boundary with respect to region b@xmath171 . ] the reason we discussed the question of entropy of a black hole at some length is that black holes play an important role in holographic gravity duals of interacting quantum field theories . the _ ads / cft duality_@xcite posits that superstring theory in ads@xmath173 is holographically dual to a conformal field theory in minkowski space , the @xmath174 supersymmetric @xmath175 gauge theory . in the limit of a large number of colors @xmath132 the string theory approaches the classical limit of the supergravity theory on ads space . when a black hole with schwarzschild radius much larger than the ads curvature radius ( a `` black brane '' ) is added to the ads@xmath176 space , the string theory becomes dual to the thermal super - yang - mills theory in minkowski space . the connection to relativistic heavy - ion collisions originates in the observation that for very high temperatures , i. e. deeply in the deconfined phase , qcd is approximately conformal and fermionic contributions to thermal properties should be much less important than gluonic ones for large @xmath132 . in addition , supersymmetry is broken by finite temperature , and the adjoint superpartners of the gauge field ( `` gauginos '' ) become effectively massive and decouple from the infrared sector of the theory . thus the absence of @xmath177-fold supersymmetry in thermal qcd should be less relevant than in the vacuum theory . finally , the question whether the thermodynamic properties of @xmath175 gauge theories already have reached the large-@xmath132 limit for @xmath134 has recently been answered in the affirmative by lattice - qcd simulations.@xcite for those readers who are not familiar with ads / cft duality , we now give a very brief description and refer to natsuume@xcite for more details and references . as already stated , ads / cft duality establishes a correspondence between superstring theory in the ads@xmath173 space - time and the @xmath174 supersymmetric @xmath175 gauge theory in minkowski space . for many purposes , the @xmath178 part of the ten - dimensional space - time can be ignored . the correspondence is thus between a five - dimensional theory of quantum gravity and a four - dimensional gauge theory . the most important aspect of the correspondence is that the strong t hooft coupling limit ( @xmath179 ) of the gauge theory is mapped onto the weak coupling limit of the string theory , which is well described by its classical limit , the supergravity theory on space - times with asymptotic ads@xmath176 geometry . the field theory can be considered as `` living '' on the asymptotic boundary of the ads@xmath176 space - time ; the ads@xmath176 space - time is therefore usually referred to as the `` bulk '' . there is an extensive dictionary of correspondences between expectation values of operators in the field theory and geometrically defined objects in the dual supergravity theory.@xcite for example , the expectation values of some local operators of the gauge theory , such as the energy - momentum tensor , are related to the metric or other local fields in the gravity theory . the expectation values of non - local operators in the gauge theory , such as two - point functions or wilson loops , correspond to the actions associated with geodesics , minimal surfaces , or other geometric objects in the dual supergravity theory.@xcite the thermal state of the field theory , i. e. the equilibrated gauge field plasma , is represented by anti - desitter space with an imbedded black hole of mass @xmath180 , whose hawking temperature @xmath181 equals the physical temperature @xmath182 of the field theory . empty anti - desitter space corresponds to the vacuum state of the field theory . thermalization of the theory is thus dual to the process of black hole formation in the bulk . the energy deposition occurring in a heavy ion collision corresponds to the injection of energy in the outer regions of the bulk , which then falls by gravitational attraction and eventually forms a black hole , providing a holographic description of the process of thermalization in the gauge theory . in recent years it has been understood that ads / cft duality makes it possible to obtain rigorous predictions for strongly coupled gauge theories at high temperature which seem to agree with phenomenological results deduced from high - energy heavy - ion collision data.@xcite however , it is important to keep in mind that qcd is not a strongly coupled theory in the high temperature limit , where it becomes approximately conformal . on the other hand , in the temperature range of practical interest ( @xmath183 500 mev ) where thermal qcd is a strongly coupled theory , it is far from being conformally invariant . lattice - qcd simulations show that the trace anomaly of the energy - momentum tensor , @xmath184 , is large and of the same order as @xmath185 itself ( see borsanyi _ et al._@xcite , fig . it is thus not obvious that the ads / cft correspondence in its simplest form is a good model for thermal qcd in the region of interest . this has motivated attempts to find improved holographic duals ( see , e. g. , alanen _ et al._@xcite , kajantie _ et al._@xcite ) . a volume of minkowski space @xmath186 filled with a quantum field in its vacuum state has a non - vanishing entanglement entropy associated with it , because the field modes of its complement necessarily leak across the boundary ( see fig . [ fig : ee ] ) . although the field is known to be in the vacuum state within @xmath186 , nothing is assumed to be known about the state of the field outside . as mentioned in section [ subsect : bc ] , this ultraviolet divergent entanglement entropy is proportional to the surface area of the volume.@xcite a holographic dual description of the entanglement entropy of a region of minkowski space - time was proposed by ryu and takayanagi@xcite ( see nishioka et al.@xcite for a review ) . they extended the surface of the volume into dimension of the bulk to form a minimal hypersurface @xmath187 , whose boundary coincides with the surface of the minkowski space volume @xmath186 . this is illustrated in fig . [ fig : ee ] . the entanglement entropy associated with this surface is defined as:@xcite @xmath188 where @xmath189 denotes the volume of @xmath187 , which has the same dimensionality as @xmath186 , and @xmath190 is newton s gravitational constant . when the field is in the vacuum state , @xmath187 extends into pure ads@xmath176 space , and one finds that its ultraviolet divergent volume is proportional to the surface area of @xmath186 . when the anti - desitter space contains a black hole and the region @xmath186 is large enough , the hypersurface @xmath187 hugs the event horizon of the black hole , and @xmath189 acquires an additional contribution proportional to the covered horizon area , which is proportinal to the volume of @xmath186 . by explicit calculation one finds that the additional contribution to the entanglement entropy is precisely the thermal entropy of the gauge field in the minkowski space region @xmath186 . ryu an takayanagi confirmed the equivalence between the entanglement entropy calculated in the field theory and that calculated holographically by means of ( [ eq : rt ] ) in several other tractable cases , but no explicit equivalence proof is known . in the following , we assume that the equivalence holds , in general . of space as the area of the minimal hypersurface @xmath187 in the bulk , which has the same boundary as @xmath186 . ] the relation ( [ eq : rt ] ) is remarkable in that it relates the bekenstein entropy of black holes to the usual thermal entropy of a field theory by holographic duality . we can imagine applying this concept to a relativistic heavy ion collision , in which a certain region of space is suddenly filled with high energy that quickly equilibrates into a hot quark - gluon plasma , which then expands , cools , and eventually dissolves . in the dual description , energy is injected asymptotically into an ads@xmath176 space , collapses to form a black hole ( or black brane ) , which eventually evaporates by hawking radiation . the formation of the black hole , i. e. the dual of the thermalization process , can be tracked by the growth of the entanglement entropy defined in ( [ eq : rt ] ) . however , in this context , the holographic duality of the entanglement entropy concept raises a number of questions : * how is the entropy of a black brane in ads@xmath176 related to the von neumann and wehrl entropies in the boundary gauge theory ? * can the duality be extended to the dynamical process of black brane formation in ads@xmath176 and thermalization in the boundary field theory ? * is it correct to apply the entanglement entropy correspondence , which has been verified to be dual in the case of thermal equilibrium , to non - equilibrium situations ? none of these questions can be conclusively answered at present . nevertheless , we want to review some of the ideas and insights which might contribute to find relevant answers in the future . also we will review results from one specific model study . a crucial element in the discussion of thermalization is the form the fluctuation - dissipation theorem takes in the bulk . following caron - huot _ et al._@xcite one can argue as follows : the large black brane in ads emits hawking radiation in the same way as black holes in minkowski space . this does not lead to a heating of the boundary , because perturbations along outgoing geodesics are reflected at the boundary and , therefore , end up again in the black brane . however , the quantum fluctuations associated with this hawking radiation induce fluctuations in the boundary , which by the fluctation - dissipation theorem are related to dissipation , and push the system towards equilibrium . the closer a fluctuation is located to the horizon , the longer it takes to reach the boundary and the stronger it will be red - shifted , while all higher momentum fluctuations get reabsorbed by the black brane on a time scale of order @xmath191 . in the boundary field theory , therefore , ultraviolet modes thermalize first , infrared modes last . to gain more insight into this complex situation one studies the reaction of ads / cft for various types of sudden violent perturbations , often called _ quantum quenches_. the entanglement entropy then traces the evolution from such a highly excited , but highly phase correlated initial state to the state of thermal equilibrium.@xcite to reduce the technical difficulty of such calculations , one often studies lower dimensional theories . in the next section we will review results in 2 , 3 , and 4 dimensions for a schematic model of black hole formation in the bulk.@xcite albash _ et al._@xcite and balasubramanian _ et al._@xcite considered the case where a finite energy density was deposited in the bulk near the boundary at a certain initial time , resulting in a shell of null dust falling in the fifth dimension until it forms an event horizon and a black brane is created . the resulting metric was then probed by strings and membranes with endpoints on the boundary . thermalization was studied as a function of the minkowski space separation between the endpoints , respectively the area of the wilson loop , as a function of time . the time required for the geodesic length or the membrane area to reach that of the black brane equilibrium configuration was interpreted as an estimate for the physical thermalization time , as measured by the entanglement entropy . the results of this study showed that complete thermalization occurs within a time that is equal to half of the spatial diameter of the probe in the boundary space . the ads space dual to the thermalizing field theory is probed by geodesics and surfaces with endpoints on the boundary , see fig.[fig : string_1 ] . the effect of the in - falling shell can be expressed geometrically in so - called poincar coordinates by the vaidya metric @xmath192 \ , , \label{eq : vaidya}\ ] ] where @xmath193 labels in - going null trajectories and the ads radius has been set equal to unity . the boundary space - time is located at @xmath194 and @xmath195 correspond to the spatial coordinates on the boundary . the mass function of the in - falling shell was taken to be : @xmath196 where @xmath197 parametrizes the thickness of the shell falling along @xmath198 , which is assumed to be very small.@xcite for a sufficiently large separation of the two endpoints on the boundary and before the shell has formed an event horizon , the geodesic curve in the bulk connecting the endpoints on the ads boundary will `` punch '' through the shell and explore the still un - thermalized ads geometry below the shell . the larger the portion of the curve extending below the shell , the larger is the difference between the geodesic length of the curve and that of a geodesic connecting the two endpoints in the ads - schwarzschild geometry , which represents full thermalization . ) with a separation @xmath199 . outside the shell , the geodesic propagates in a black brane geometry , while inside it propagates in an empty ads geometry . , scaledwidth=60.0% ] it is then possible to obtain analytic or numerical solutions for different dimensions and probes . some results were obtained by balasubramanian _ et al._@xcite for the entanglement entropy . in one spatial dimension ( @xmath200 ) the entanglement entropy is holographically given by the length of the bulk geodesic connecting the two endpoints of the linear region in the boundary space . in two spatial dimensions ( @xmath152 ) the entanglement entropy is holographically given by the minimal area of the bulk surface , whose perimeter coincides with the perimeter of the circular boundary area . in three spatial dimensions ( @xmath201 ) the entanglement entropy is holographically given by the minimal volume of the bulk hypersurface , whose surface coincides with the surface of the spherical boundary region . the results are shown in fig . [ fig : deltase ] . in each dimension , the difference between the minimal length ( or area or volume ) in the bulk for the geometry at a given time and minimal length ( area , volume ) in the fully thermalized ads - schwarzschild geometry is shown , normalized to the boundary volume ( @xmath202 , @xmath203 , @xmath204 ) . in every case one finds that thermalization as measured by the entanglement entropy is achieved within the time given by precisely half the diameter of the probe in the boundary divided by the speed of light . the main results of this investigation are:@xcite * thermalization is an extremely fast process in the strongly coupled gauge theory , which is only constrained by causality . the information loss from a finite region proceeds at the speed of light . this is especially remarkable because the strongly coupled gauge theory does not support quasi - particle excitations , with the exception of phonons , which do not play a role in this process . the mechanism by which the information is transported out of the thermalizing volume is unknown . however , the result suggests that the strongly coupled gauge theory behaves like a _ fast scrambler _ as defined by sekino and susskind.@xcite * other observables reach their equilibrium values faster than the entanglement entropy . this is not surprising , because the entropy is sensitive to all degrees of freedom of the gauge theory , while other observables are sensitive only to a subset . * short distances thermalize first , i.e. thermalization occurs top - down for a gauge theory in the limit of very strong coupling , while it occurs bottom - up for weakly coupled gauge theories.@xcite , minimal surface area @xmath205 , and minimal volume @xmath206 , respectively , to that of the thermalized state as a function of boundary time @xmath207 for dimensions @xmath200 ( left ) , @xmath152 ( center ) , and @xmath201 ( right ) . the results are for a thin in - falling shell ( @xmath208 ) . the boundary separations were taken to be @xmath209 and radii @xmath210 ( from top to bottom curve ) , respectively . all quantities are given in units of @xmath180 . complete thermalization occurs precisely at @xmath211 or @xmath212 , respectively . , title="fig:",height=151 ] , minimal surface area @xmath205 , and minimal volume @xmath206 , respectively , to that of the thermalized state as a function of boundary time @xmath207 for dimensions @xmath200 ( left ) , @xmath152 ( center ) , and @xmath201 ( right ) . the results are for a thin in - falling shell ( @xmath208 ) . the boundary separations were taken to be @xmath209 and radii @xmath210 ( from top to bottom curve ) , respectively . all quantities are given in units of @xmath180 . complete thermalization occurs precisely at @xmath211 or @xmath212 , respectively . , title="fig:",height=151 ] , minimal surface area @xmath205 , and minimal volume @xmath206 , respectively , to that of the thermalized state as a function of boundary time @xmath207 for dimensions @xmath200 ( left ) , @xmath152 ( center ) , and @xmath201 ( right ) . the results are for a thin in - falling shell ( @xmath208 ) . the boundary separations were taken to be @xmath209 and radii @xmath210 ( from top to bottom curve ) , respectively . all quantities are given in units of @xmath180 . complete thermalization occurs precisely at @xmath211 or @xmath212 , respectively . , title="fig:",height=151 ] the problem of thermalization of highly excited states in quantum chromodynamics and gravity has been intensively studied during the past decade . in qcd , the interest was driven by the experimental results from relativistic heavy ion collisions , which indicated that a thermal quark - gluon plasma is formed on an extremely short time scale of order 1 fm/@xmath1 . in gravity theories , the main motivation was the desire to resolve the information paradox associated with black hole formation . in both situations , the question can be phrased in terms of the growth rate of an appropriately defined entropy . while the von neumann entropy of a closed system does not change with time in any quantum theory due to the unitarity of the time evolution , other measures of entropy , which take into account the principal or practical inability of an observer to measure all details of a system , can grow with time and eventually approach the thermal equilibrium value . in this review , we covered several different definitions of such a coarse grained entropy with broad applicability : husimi s phase - space smearing of the wigner distribution , de - phasing of the quantum state underlying the eigenstate thermalization hypothesis , and entanglement entropy applied to holographic duals of the quantum system . we showed that de - phasing and husimi smearing result in the same entropy growth rate in the case of a single unstable mode , which was equal to the ks entropy describing the growth rate of the coarse grained entropy in the classical limit . this observation motivated a numerical study of the entropy growth rate of the non - abelian su(2 ) gauge theory , reviewed in section [ subsect : su2 ] , which confirmed the rapid thermalization observed in the experiments . the concept of entanglement entropy and its holographic equivalent enables rigorous calculations of the approach to thermalization in the strongly coupled conformal supersymmetric gauge theory . in this approach , the process of thermalization is dual to the formation of a black hole , linking the study of thermalization in the @xmath213dimensional gauge theory to the study of black hole formation in @xmath214dimensional supergravity . the concept of entanglement entropy for a bounded region of space parallels that of the husimi smeared entropy ( the wehrl entropy ) , because it accounts for the position uncertainty of field modes with a given momentum . the two concepts differ in that the entanglement entropy only accounts for the uncertainty relation smearing of field modes near the boundary , while the wehrl entropy implements the uncertainty smearing of all field modes . the holographic duality also provides compelling evidence for the view that the formation and subsequent evaporation of a black hole must be a unitary process in a quantum theory of gravity , which does not result in an increase of the von neumann entropy , but is accompanied by a maximal increase in an appropriately defined coarse grained entropy . notwithstanding these significant theoretical advances , there remain many unresolved questions . how exactly are the different notions of coarse grained entropy connected ? when does black hole formation and decay result in an increase in the coarse grained entropy , and when does it represent a practically reversible process as in the toy model constructed by takayanagi and ugajin?@xcite how do the same concepts apply to the cosmic big bang ? we hope that this review will motivate some of its readers to study and ultimately resolve such questions . this work was supported in part by grants from the u. s. department of energy ( de - fg-05er-41367 ) and the german bmbf .

the fermion four - point functions and condensates as the chiral symmetry order parameters are calculated analytically in u(1 ) gauge theory in the massless phase . it is shown that in leading order of the loop expansion of the effective action , there is a critical coupling for the nonperturbative parity - invariant chirality - changing four - fermion functions ; this critical coupling however is the same as that for dynamical mass generation . therefore there is no breaking of chiral symmetry through the condensation of four - fermion functions below the critical coupling for dynamical mass generation for the fermions .

hep-ph9511403

dynamical chiral symmetry breaking ( dcsb ) plays a very important role in the research of fundamental interactions . up to now , most investigations are focussed on dcsb with mass generation for the fermions . in such a scenario , the bilinear fermion condensate is taken as the order parameter of chiral symmetry breaking and it is known that it can receive a nonzero value when the coupling of the system exceeds some critical value . for a review on dynamical chiral symmetry breaking , see e.g. @xcite . but this is not the complete story for dcsb , since the order parameter may not be the bilinear fermion condensate : in principle , higher @xmath0-point functions which break chiral symmetry might become nonzero , without dynamical mass generation for the fermions @xcite . in a massless @xmath1 gauge theory for example , four - fermion green s functions which break chiral symmetry might become nonzero at values of the coupling below the critical value for dynamical mass generation . such a breaking of the chiral symmetry by the four - fermion condensates , without mass generation for the fermions , was discussed recently by holdom and triantaphyllou @xcite , who have calculated these four - point functions numerically and found a critical coupling constant which is almost the same as that responsible for dynamical mass generation . one could also think of theories where some other symmetry prevents the dynamical generation of a bilinear condensate and a dynamical fermion mass , but where a nonzero condensate of a higher @xmath0-point function is not restricted by this symmetry . in a chiral gauge theory , the gauge symmetry will be broken by the generation of a fermion mass , but a chirality - changing four - fermion condensate does not break the gauge symmetry , and might therefore be generated dynamically . in such a scenario , the chiral symmetry can be realized in more delicate way which may be useful for model building @xcite : it allows for a hierarchy of chiral symmetry breaking , first by the four - fermion condensate at relatively small values of the coupling , and eventually a second phase transition , at higher values of the coupling , leading to mass generation for the fermions , due to the four - fermion condensates . for these purposes one should study this phenomenon in a chiral gauge theory , as suggested in @xcite , but for simplicity and to see whether such a symmetry breaking patterns is possible , we restrict ourselves to a @xmath1 gauge theory . in our paper we confirm analytically the numerical results obtained by holdom and triantaphyllou @xcite , using a slightly different method to calculate the nonperturbative four - fermion functions . we find that there is a critical coupling for dynamical chiral symmetry breaking via the four - fermion condensates , below which there are no nontrivial solutions for the four - fermion functions which break chiral symmetry , but preserve parity , lorentz symmetry , and the flavor symmetry . this critical coupling is ( in our truncation scheme ) exactly the same as the critical coupling for fermion mass generation . this result is obtained by calculating the loop expansion of the effective action to leading order for the nonperturbative four - point functions . our result is also in agreement with an intuitive argument comparing the condensation of two fermions to the condensation of four fermions : for the four - fermion condensate we need at least the same binding force as for the bilinear fermion condensate . our paper is organized as follows . in the next section , we discuss some general features about dynamical symmetry breaking , and we show how in principle the chiral symmetry can be broken without mass generation for the fermions . in sec . [ seceffact ] , we derive a linear integral equation for the four - fermion function , using the loop expansion of the effective action . next , we analyze this equation for the nonperturbative chirality - changing four - fermion functions , using some expansions in order to perform the angular integrations analytically ; more details can be found in the appendix . we find that there is a critical coupling for these functions ; this critical coupling turns out to be the same as the critical coupling for dynamical mass generation . finally we give some concluding remarks in sec . [ secconc ] . the kind of symmetry breaking we are dealing with in this paper is the breaking of chiral symmetry in a @xmath1 gauge theory with @xmath2 massless fermion flavors . the global continuous chiral symmetry is @xmath3 , and this symmetry can be broken in different ways , characterized by the corresponding chiral order parameters . since the flavor symmetry is not broken in vector - like gauge theories @xcite , only flavor - symmetry invariant order parameters may receive nonzero contributions ; also parity @xcite and lorentz symmetry is believed not to be broken spontaneously , so in this paper we limit us to parity- , flavor- , and lorentz - invariant order parameters . the simplest and most wellknown way of breaking the chiral symmetry is characterized by the trace of the two - point function @xmath4 . this condensate , @xmath5 , is the only possible two - fermion condensate which keeps the flavor symmetry and is parity and lorentz invariant . it gives rise to a dynamically generated fermion mass , and the symmetry breaking pattern is @xmath6 an explicit mass for the fermions would also break the symmetry in this way . however , this is not the only way to break the chiral symmetry , an alternative way is characterized by condensates of four fermions . the possible four - point function which keep flavor and lorentz symmetry and which are parity invariant are @xmath7 \ , , \label{0.02}\\ { \cal g}_{0,t}(x_1,x_2,x_3,x_4 ) & = & \langle 0|{\bf t}{\bar\psi}(x_1)\sigma_{\mu\nu}\lambda^{0 } \psi ( x_2){\bar\psi}(x_3)\sigma^{\mu\nu}\lambda^{0}\psi ( x_4 ) |0\rangle \ , , \label{0.04}\\ { \cal g}_{t}(x_1,x_2,x_3,x_4 ) & = & \sum_{\alpha = 1}^{n_f^2 - 1 } \langle 0|{\bf t}{\bar\psi}(x_1)\sigma_{\mu\nu}\lambda^{\alpha } \psi(x_2){\bar\psi}(x_3)\sigma^{\mu\nu}\lambda^{\alpha}\psi ( x_4 ) |0\rangle \ , . \label{0.05}\end{aligned}\ ] ] in these expressions , @xmath8 is the flavor matrix with @xmath9 ; @xmath10 is proportional to the unit matrix . these four - fermion functions can be related to the axial charge , and the condensate of these four - fermion functions ( with all the coordinates @xmath11 in the same space - time point ) can be considered as the order parameters of dcsb . a nonzero condensate of ( one of ) these four - fermion functions gives the symmetry breaking pattern @xmath12 here @xmath13 is a nontrivial chiral subgroup which plays a role in preventing dynamical mass generation for the fermions . it corresponds to the discrete chiral transformation @xmath14 and therefore the fermions remain massless in such a breaking of the chiral symmetry . note that in principle there are a lot more four - fermion functions which break the chiral symmetry . in the case of one fermion flavor , @xmath15 , there are in total 54 independent functions which all lead to the same chiral symmetry breaking pattern , eq . ( [ symbrpat ] ) , but the ones above , which we will call @xmath16 and @xmath17 type four - fermion functions , are the only ones which preserve parity and lorentz symmetry . the other 52 independent functions are not all parity - invariant or lorentz scalars , which means they will not receive nonzero condensates , assuming that the lorentz symmetry and parity remains unbroken . however , it is not excluded that ( some of ) the nonlocal four - fermion functions become nonzero , in which case they might couple to the @xmath16 and @xmath17 type functions . we leave this problem to future investigations and for simplicity we only consider the nonperturbative four - point functions which are flavor- , lorentz- and parity - invariant . this scenario can easily be generalized to condensates of @xmath18 fermions , breaking the chiral symmetry into @xmath19 with the nontrivial chiral subgroup @xmath20 , corresponding to the discrete chiral transformation @xmath21 unbroken . it is wellknown that if a continuous symmetry of the lagrangian is broken dynamically , there will appear some massless bosons , the socalled goldstone or nambu goldstone bosons @xcite . these goldstone bosons couple both to the conserved current which is associated with the symmetry , and to some fields @xmath22 which characterize the breaking of the symmetry . note that these fields @xmath22 , which can be regarded as the order parameters of the phase transition , can be either composite or elementary . in the case of dynamical breaking of chiral symmetry , the goldstone bosons are related to the chiral - symmetry breaking @xmath0-point functions through the axial ( or chiral ) ward takahashi identity . the conserved chiral current is @xmath23 and we consider the following green s function in momentum space @xmath24 where @xmath22 is a ( composite ) field associated with the chiral order parameter we are considering which is related to goldstone pole in momentum space ; in general , @xmath22 can also depend on other space - time coordinates , which do not relate to the goldstone pole . ] . the breaking of chiral symmetry means that this order parameter becomes nonzero , and due to the chiral ward takahashi identity this leads to a pole at zero momentum in the above green s function ; in the limit @xmath25 this green s function behaves as @xmath26 to be more specific , we first consider conventional chiral symmetry breaking , with a dynamically generated mass for the fermions . in that case , the composite field @xmath27 is @xmath28 , and the chiral ward takahashi identity leads to @xmath29 where @xmath30 is the full fermion propagator . from this equation , we see immediately that a dynamically generated mass for the fermions will lead to a massless pole in the pseudoscalar green s function @xmath31 if @xmath32 ; this pole corresponds to the massless goldstone boson . in the case of dynamical chiral symmetry breaking through the four - point functions , as we are considering here , the situation is somewhat more complicated since we have to consider a composite field @xmath22 which consists of four fermion fields . using the chiral ward takahashi identity , one can show that the green s functions @xmath33 have to satisfy the relation @xmath34 where @xmath35 is the fourier transform of a four - fermion function , with momentum conservation @xmath36 . in the limit @xmath37 , this leads to massless poles , if either the @xmath16 type or the @xmath17 type four - fermion function @xmath38 is nonzero . this can be seen easily by decomposing the green s function @xmath39 into @xmath40 so eq . ( [ axward ] ) leads to @xmath41 thus massless goldstone bosons appear as soon as any of these nonperturbative four - point functions becomes nonzero . if we restrict ourselves to _ local _ composite fields @xmath22 we get a similar expression , but now involving the _ condensates _ of these four - point functions . defining the condensate as @xmath42 where @xmath43 denotes either @xmath16 or @xmath17 , the chiral ward takahashi identity relates this condensate to the goldstone bosons . in momentum space this finally leads to massless poles for @xmath44 @xmath45 thus the condensates of the four - point functions we are considering here are directly related to the existence of massless goldstone bosons . note that both a nonzero condensate and a nonzero four - point function would signal dynamical chiral symmetry breaking , and in principle it is conceivable that the condensate is zero , and the four - point function not ; the other way around ( no nonzero four - point function , but only a nonzero condensate ) is obviously impossible . in this paper , instead of four - fermion condensate , we directly calculate four - point function and take it as a generalized order parameter . the generalization to the case of @xmath46 fermion flavors is in principle straightforward ; also the generalization to higher @xmath0-point functions is a trivial exercise . in order to calculate the chirality - changing four - fermion functions , we need some nonperturbative calculation scheme . for this purpose , we use a loop expansion of the effective action @xcite . for simplicity we will restrict ourselves to just one fermion flavor , so the generating functional in minkowski space is @xmath47 \ ; = \ ; { \rm e}^{w[{\bar i},i , j ] } & = & { \int}{\cal d}\bar\psi { \cal d}\psi { \cal d}a_{\mu } \ ; { \rm e}^{i{\int}d^{4}x [ { \cal l}(a ) + { \bar\psi}(i\partial\!\!{\!}/-ea\!\!{\!}/)\psi + { \bar i}{\psi}+{\bar\psi}i + j_{\mu}a^{\mu } ] } \ , , \label{genfun}\end{aligned}\ ] ] where @xmath48 is the lagrangian for the photon field @xmath49 in a general covariant gauge . in order to calculate the effective action , we start by integrating out the photon field , which gives @xmath50 } \ ; = } \nonumber \\ & = & \exp\bigg[\frac{-i}{2}{\int}d^{4}xd^{4}y \left(-e\bar{\psi}(x)\gamma^{\mu}\psi ( x)+j^{\mu}(x)\right ) d_{\mu\nu}(x , y ) \left(-e\bar{\psi}(y)\gamma^{\nu}\psi ( y ) + j^{\nu}(y)\right)\bigg ] \,,\end{aligned}\ ] ] where @xmath51 is the bare photon propagator @xmath52 \delta^4(x - y ) \,,\end{aligned}\ ] ] in which the derivative @xmath53 is acting on the coordinate @xmath54 , and @xmath55 is the gauge - fixing parameter . next , we introduce a generalized green s function containing four space - time points , with spinor indices @xmath56 , and @xmath57 @xmath58 such that @xmath59 note that in the case with @xmath2 fermion flavors , we just have to add additional indices to label the different flavors . thus the generating functional can be written as @xmath60 & = & \int{\cal d}\psi{\cal d}\bar{\psi}\exp\,i \bigg [ { \int}d^{4}x \left(\bar{\psi}i\partial\!\!\!/{\psi}+{\bar i}\psi+\bar{\psi}i \right ) \nonumber\\ & & + \frac{e^2}{2 } { \int}d^{4}xd^{4}x'd^{4}yd^{4}y ' \overline{d}^{ab}_{cd}(x , x',y , y ' ) \bar{\psi}^{a}(x){\psi}^{c}(x ' ) \bar{\psi}^{b}(y){\psi}^{d}(y ' ) \nonumber\\ & & - \frac{1}{2 } { \int}d^{4}xd^{4}y d_{\mu\nu}(x , y ) \left(-2e\bar{\psi}(x)\gamma^{\mu}\psi ( x)j^{\nu}(y ) + j^{\mu}(x)j^{\nu}(y)\right ) \bigg ] \,.\end{aligned}\ ] ] inserting a gaussian integral of a bilocal auxiliary field @xmath61 @xmath62 \,,\end{aligned}\ ] ] gives for the generating functional @xmath63 \ ; = \ ; \int{\cal d}\chi{\cal d}\psi{\cal d}\bar{\psi}\exp\,i \bigg [ { \int}d^{4}x \big ( \bar{\psi}(x)i\partial\!\!\!/_{x } \psi(x ) + { \bar i}(x)\psi(x ) + \bar{\psi}(x)i(x)\big ) } \nonumber\\ & & + { \int}d^{4}xd^{4}x'd^{4}yd^{4}y ' \overline{d}^{ab}_{cd}(x , x',y , y ' ) \left ( e \bar{\psi}^{a}(x ) \chi^{bd}(y , y ' ) \psi^{c}(x ' ) - \frac{1}{2 } \chi^{ac}(x , x')\chi^{bd}(y , y ' ) \right ) \nonumber\\ & & + e{\int}d^{4}x d^{4}y \bar{\psi}(x ) \gamma^{\mu } d_{\mu\nu}(x , y ) j^{\nu}(y ) \psi(x ) - \frac{1}{2}{\int}d^{4}xd^{4}y j^{\mu}(x ) d_{\mu\nu}(x , y ) j^{\nu}(y ) \bigg ] \,.\end{aligned}\ ] ] we see that the fermion field appears quadratically in the generating functional , so that they can be easily be integrated out . thus we arrive at @xmath64 & = & \int{\cal d}\chi { \rm e}^{is_{\rm{eff}}[\chi,{\bar i},i , j ] } \nonumber \\ & = & \int{\cal d}\chi \exp\,i\bigg [ \frac{-1}{2}{\int}d^{4}xd^{4}x'd^{4}yd^{4}y ' \overline{d}^{ab}_{cd}(x , x',y , y')\chi^{ac}(x , x')\chi^{bd}(y , y ' ) \nonumber\\ & & + i{\rm tr}\,{\rm ln}\,s - { \int}d^{4}xd^{4}y \left(\frac{1}{2 } j^{\mu}(x ) d_{\mu\nu}(x , y)j^{\nu}(y ) + { \bar i}^{a}(x)s_{ab}(x , y)i^{b}(y)\right ) \bigg ] \,,\end{aligned}\ ] ] in which @xmath65 is defined as @xmath66 the effective action for the @xmath67-field , induced by eq . ( [ 506 ] ) and with external sources , is then given by @xmath68 \;= \ ; \frac{-1}{2 } { \int}d^{4}xd^{4}x ' d_{\mu\nu}(x , x')\gamma^{\mu}_{ad}\gamma^{\nu}_{bc } \chi^{ac}(x , x')\chi^{bd}(x',x ) } \nonumber\\ & & + i{\rm tr}\,{\rm ln}\,s - { \int}d^{4}xd^{4}y \left(\frac{1}{2}j^{\mu}(x)d_{\mu\nu}(x , y)j^{\nu}(y ) + { \bar i}^{a}(x)s_{ab}(x , y)i^{b}(y)\right ) \,,\end{aligned}\ ] ] where we have used eq . ( [ 502.1 ] ) . given the effective action @xmath69 $ ] in eq . ( [ 507 ] ) , the calculation of various physical quantities concerns the calculation of the following generating functional @xmath70 & = & { \rm e}^{i\tilde{w}[{\bar i},i , j,\tilde j ] } \nonumber\\ & = & \int{\cal d}{\chi}\exp\,i\left[s_{\rm{eff}}[\chi,{\bar i},i , j ] + { \int}d^{4}x { \;\;}\tilde j^{b}(x)\chi_{b}(x ) \right ] \,,\end{aligned}\ ] ] where @xmath71 is an external source coupled to the @xmath72 field , and we use a shorthand notation @xmath73 and @xmath74 for the bilocal variables @xmath75 and indices @xmath76 . the generating functional @xmath77 $ ] in eq . ( [ 504 ] ) is related to @xmath78 $ ] by @xmath77=\tilde{z}[{\bar i},i , j,\tilde j=0]$ ] . now we define the classical field @xmath79 by @xmath80}{{\delta } \tilde j^{b}(x ) } \,,\end{aligned}\ ] ] and the partial legendre transformation @xmath81 & = & w[{\bar i},i , j,\tilde j ] - { \int}d^{4}x{\;\;}\tilde j^{b}(x ) { \hat{\chi}}_{b}(x ) \,.\end{aligned}\ ] ] it is easy to see that @xmath82 } { \delta{\hat{\chi}}_{b}(x ) } & = & -\tilde j^{b}(x ) \,,\end{aligned}\ ] ] and hence @xmath83 \;=\ ; \tilde{w}[{\bar i},i , j,0 ] & = & \tilde{\gamma}[{\bar i},i , j,{\hat{\chi } } ] \bigg|_{\frac{\delta\tilde{\gamma}}{\delta{\hat{\chi}}_{b}}=0 } \;.\end{aligned}\ ] ] therefore what we need to do is to calculate @xmath84 $ ] . when we turn off the source for the bilocal field , @xmath71 , eq . ( [ 515 ] ) becomes @xmath85 } { \delta{\hat{\chi}}_{b } } & = & 0 \,,\end{aligned}\ ] ] which reflects the stability of the physical vacuum . the standard loop expansion formula for @xmath86 $ ] is given in @xcite , and reduces in the present case to @xmath87 & = & s_{\rm{eff}}[{\hat\chi},{\bar i},i , j ] + \frac{i}{2}{\rm tr}\,{\rm ln}\left [ \frac{\delta^{2}s_{\rm{eff}}[{\hat\chi},{\bar i},i , j ] } { \delta{\hat\chi}_{b_{1}}(x_{1})\delta{\hat\chi}_{b_{2}}(x_{2 } ) } \right ] + \tilde{\gamma}_{2 } \,,\end{aligned}\ ] ] where the first term is the tree level contribution , the second term is the one - loop contribution , and @xmath88 represent higher - loop contributions , containing the sum over all higher loop 1pi vacuum diagrams with the propagator @xmath89 } { \delta{\hat{\chi}}_{b_{1}}(x_{1 } ) \delta{\hat{\chi}}_{b_{2}}(x_{2 } ) } \,,\end{aligned}\ ] ] and vertices @xmath90 } { \delta{\hat{\chi}}_{b_{1}}(x_{1})\ldots \delta{\hat{\chi}}_{b_{n}}(x_{n } ) } \hspace{2 cm } ( n\;{\geq}\;3 ) \,.\end{aligned}\ ] ] this expansion , eq . ( [ 518 ] ) , gives systematic rules to calculate the effective action , and we base our further discussion on this expansion . for simplicity , we consider the effective potential up to tree - level order only in our present paper , but _ in principle _ it is straightforward to go beyond this leading order calculation . at tree level , when the external source is switched off , we find @xmath91 & = & s_{\rm{eff}}[{\hat\chi},{\bar i},i , j ] \,,\end{aligned}\ ] ] at the stationary point of the effective action @xmath92 } { \delta { \hat{\chi}^{ac}(x , y ) } } & = & 0 \,.\end{aligned}\ ] ] using the definitions eqs . ( [ 507 ] ) and ( [ 506.1 ] ) , the stationary condition for the effective action , eq . ( [ 520 ] ) , leads to @xmath93 combined with eq . ( [ 506.1 ] ) , we find @xmath94 \,.\end{aligned}\ ] ] note from eq . ( [ 519 ] ) and eqs . ( [ 507 ] ) and ( [ 520 ] ) , that the function @xmath95 is related to the full fermion propagator ( two - point function ) @xmath96}{\delta{\bar i}_a(x)\delta i_b(y ) } & = & s_{ab}(x , y ) - \int d^{4}z{\bar i}_{c}(z ) \frac{\delta s_{cb}(z , y)}{\delta{\bar i}_{a}(x ) } \\ \label{522b } & = & s_{ab}(x , y ) + \int d^{4}z\frac{\delta s_{ac}(x , z)}{\delta i_{b}(y)}i_{c}(z)\,,\end{aligned}\ ] ] where the functional derivatives @xmath97 and @xmath98 can be calculated using eq . ( [ 521.1 ] ) . combined with eq . ( [ 521 ] ) , we find that the @xmath67-field is the full fermion propagator if we switch off the external sources . on the vacuum , all external sources should be taken to zero , and therefore eq . ( [ 521.1 ] ) gives @xmath99 which is just the quenched ladder approximation for the schwinger dyson equation for the full fermion propagator . by differentiating the generating functional twice with respect to the source @xmath100 and imposing the stationary condition for the effective action , we can get the schwinger dyson for the photon field . for the four - point function , we use eq . ( [ 522a ] ) and ( [ 522b ] ) to relate the four - point function to derivatives of the function @xmath95 . after putting the sources equal to zero , and defining @xmath101 we find for the full four - point function @xmath102 } { \delta{\bar i}_{a}(x)\delta i_{c}(x ' ) \delta{\bar i}_{b}(y)\delta i_{d}(y')}\bigg|_{j=0,i={\bar i}=0 } \nonumber\\ & = & g^{acbd}(x , x',y , y')-g^{bcad}(y , x',x , y')\nonumber\\ & = & g^{acbd}(x , x',y , y')-g^{adbc}(x , y',y , x'),\end{aligned}\ ] ] we have written down all possible expressions , which we can use to solve the four - point functions . note that the four - point function @xmath103 is written as the sum of the @xmath104- and @xmath105-channel components @xmath106 . using eq . ( [ 521.1 ] ) we find for the @xmath104-channel @xmath107 e^{2}d_{\mu\nu}(x , x ' ) \gamma^{\nu}_{bc } \,.\end{aligned}\ ] ] this is nothing but the ( inhomogeneous ) bethe salpeter equation . for the @xmath16 and @xmath17 type four - point functions in the massless - fermion phase ( or more general : for all nonperturbative four - point functions ) , the second term on the right hand side of eq . ( [ 528 ] ) vanishes , so we arrive at following equation , namely @xmath108 this equation is the @xmath104-channel equation for the nonperturbative four - fermion functions in the quenched ladder approximation . for the @xmath105-channel equation , we can get a similar equation in exactly the same way . this @xmath105-channel equation however is completely equivalent to the @xmath104-channel equation , and it can also be obtained by applying a fierz transformation to the @xmath104-channel equation . next we make a fourier transformation to momentum space . the equation for the @xmath104-channel four - point function becomes @xmath109 where @xmath110 is the bare photon propagator in momentum space @xmath111 \,.\end{aligned}\ ] ] as mentioned before , the @xmath105-channel equation is completely equivalent to this @xmath104-channel equation . for the complete nonperturbative four - point functions in momentum space , @xmath112 , we thus find @xmath113 so finally we have the leading order equations for the nonperturbative four - fermion functions , which is shown in diagrams fig . [ figsteqs ] . for the perturbative four - fermion functions ones we have to add the tree contribution coming from the second term in the right hand side of eq.([528 ] ) . one can show that the perturbative four - fermion functions do not contribute at all to the nonperturbative ones , so the nonperturbative four - point functions form a closed set of equations , at least below the critical coupling for dynamical mass generation . furthermore , since we assume that all the other types of nonperturbative four - point functions ( e.g. the ones which break parity or lorentz invariance ) remain zero , and we only have to deal with the @xmath16 and @xmath17 type four - point functions @xmath114 using the general nonperturbative equation for the four - fermion function derived in the previous section , we can project out the equations for these @xmath16 and @xmath17 type four - point functions , which form closed set of linear integral equations . the four - fermion functions @xmath115 and @xmath116 are related to the functions @xmath117 and @xmath118 by @xmath119 after the projection and reduction of the @xmath120-algebra , we get for the @xmath104-channel @xmath16 and @xmath17 type four - fermion functions @xmath121 where @xmath122 is the bare photon propagator . remember that in @xmath123 and @xmath124 , momentum conservation condition @xmath125 should be satisfied . since the @xmath105-channel equations follow from these @xmath104-channel equations , it is enough to analyze these two @xmath104-channel equations , at least for the question whether or not there is a critical coupling . we also have to choose gauge . if we take the feynman gauge , then the equation eq . ( [ ts ] ) gives immediately @xmath126 , whereas the equation for @xmath117 , eq . ( [ sps ] ) , leads automatically to @xmath127 in the yennie gauge . these results are obviously artifacts of the specific gauges , and one should choose the `` correct '' gauge in order to get reasonable results . it is known that the quenched ladder approximation for the fermion schwinger dyson equation , which also follows from our effective action , is most reliable in the landau gauge , @xmath128 . therefore , we analyze the equations for these four - fermion functions in the landau gauge , and use the photon propagator @xmath129 we start by looking at the equation for @xmath117 , eq . ( [ sps ] ) , which becomes in euclidean space , after a wick rotation , @xmath130 we redefine @xmath131 which allows us to rewrite this equation into @xmath132 in order to solve this equation , we first expand @xmath133 in terms of @xmath134 @xmath135 where we use the notation @xmath136 for the angular dependence , and @xmath137 for the absolute value of the momenta . note also that @xmath138 is symmetric in the indices @xmath139 . inserting this in eq . ( [ 306.1 ] ) shows immediately that the coefficients @xmath138 have to satisfy @xmath140 \ , , \label{309.1}\end{aligned}\ ] ] and @xmath141 \nonumber\\ & = & 3e^{2}\int\frac{d^{4}q}{(2\pi ) ^{4}}\frac{1}{(q - p_{1})^{2 } } \bigg[-p_{1}^{\mu_{1}}{\tilde g}_{n-1}^{\mu_{2}\ldots \mu_{n } } ( q , p_{2r},p_{3},p_{4 } ) + p_{1}^{2 } { \tilde g}_{n}^{\mu_{1}\ldots \mu_{n}}(q , p_{2r},p_{3},p_{4 } ) \nonumber\\ & & + \hbox { all symmetric permutations in } ( \mu_{1}\ldots\mu_{n } ) \bigg ] \ , , \label{310.1}\end{aligned}\ ] ] with @xmath142 . the homogeneous part of above equation satisfies @xmath143 in order to prove that the only solution of the equations eqs . ( [ 308.1])([310.1 ] ) is the trivial solution , it is enough to prove that the only solution of the homogeneous equation eq . ( [ 311.1 ] ) is the trivial solution @xmath144 , for all @xmath145 . note that only a solution which is symmetric in the lorentz indices @xmath146 will contribute to the four - point function @xmath147 ; from eq . ( [ 307.1 ] ) it follows immediately that any nonsymmetric part does not contribute . the angular integration in eq . ( [ 311.1 ] ) can be performed analytically , using an expansion in a kind of generalized chebyshev polynomials @xmath148\ , , \label{408.1}\\ { \hat u}_{2n+1}^{\mu_{1}\ldots\mu_{2n+1}}({\hat q } ) & = & \sum_{k=0}^{n } \frac{(-1)^{k}2^{2n+1 - 2k}(2n+1-k)!}{k!(2n+1 - 2k)!(2n+1 ) ! } \big[{\hat q}^{\mu_{1}}\ldots{\hat q}^{\mu_{2n+1 - 2k } } \delta^{\mu_{2n+2 - 2k}\mu_{2n+3 - 2k}}\ldots \delta^{\mu_{2n}\mu_{2n+1 } } \nonumber\\ & & + { \hbox { all symmetric permutations in } } ( \mu_{1}\ldots\mu_{2n+1 } ) \big ] \ , , \label{409.1}\end{aligned}\ ] ] where @xmath149 . these polynomials satisfy the identity @xmath150^{n+1 } \ , , \label{313.1}\end{aligned}\ ] ] which is shown in appendix [ angint ] , where we have given more details on these polynomials . assuming that our four - point function is smooth enough ( or more precisely , has a taylor - expansion ) , we expand first , followed by an expansion of @xmath151 in terms of @xmath152 using eqs . ( [ 417 ] ) and ( [ 418 ] ) , see the appendix . ] the unknown functions @xmath153 in terms of these generalized chebyshev polynomials @xmath154 after this decomposition , which allows us to perform the angular integration , eq . ( [ 311.1 ] ) has to hold for each function @xmath155 separately , which leads to an equation for each coefficient @xmath156 for convenience we will drop all irrelevant indices and the spectator momenta in the discussion of this equation , defining @xmath157 as a scalar function of the absolute value of @xmath158 @xmath159 so the equation we have to consider becomes @xmath160 this integral equation is exactly the same as the equation for the dynamical mass function in quenched ladder qed , which has been studied extensively @xcite . it can be transformed into the following differential equation @xmath161 there are two independent solutions for the above differential equation , say @xmath162 with @xmath163 satisfying @xmath164 \,.\end{aligned}\ ] ] from this equation , it follows immediately that there is a critical coupling @xmath165 , below which both @xmath163 are real , and above which they are complex . so the general solution is @xmath166 in terms of the two independent solutions of the differential equation , with @xmath167 some arbitrary functions , depending on the spectator momenta @xmath168 , @xmath169 , and @xmath170 in the original equation eq . ( [ 314.1conv ] ) , which needs to be determined . this general solution is constrained by the integral equation , and therefore , we substitute this solution back in the original integral equation , which gives us @xmath171 \ , . & = & 0\end{aligned}\ ] ] this boundary condition can only be satisfied above the critical coupling , if the exponents @xmath163 are complex ; below the critical coupling , both @xmath163 are real numbers , and it is impossible to satisfy this boundary condition . assuming there is a nontrivial solution , eq . ( [ 63.1 ] ) is equivalent to the set of equations @xmath172 below the critical coupling , this set of equations only has a solution if @xmath173 , which means the infrared cutoff bigger than the ultraviolet cutoff . this is obviously not a physical solution , and therefore the only solution of eq . ( [ 56.1 ] ) is the trivial solution . that means that the only solution of the integral equation eq ( [ 314.1 ] ) is the trivial solution @xmath174 now going back to the original equation for the @xmath16 type four - fermion function , using the decomposition eqs . ( [ 312.1 ] ) and ( [ 307.1 ] ) , and definition eq . ( [ 305.1 ] ) , we find that only the trivial solution @xmath175 is allowed . above the critical coupling , @xmath176 , both @xmath163 have a nonzero imaginary part , which implies oscillating solutions . instead of using the complex exponents @xmath163 , we write the general solution of the differential equation eq . ( [ 56.1 ] ) as @xmath177 where we have used the ultraviolet cut - off @xmath178 to set the mass scale . substituting this solution back in original integral equation , we finally find a condition for @xmath179 @xmath180 now one can always construct nontrivial solutions , depending on this ratio @xmath181 . we assume that @xmath182 , the infrared cutoff , is related to the energy scale of the dynamically generated condensate through nonlinear effects , just as in the linearized quenched ladder approximation for the fermion schwinger dyson equation , where the infrared cut - off sets the scale for the infrared mass function , @xmath183 . thus we allow the infrared cut - off to be a function of the ultraviolet cut - off , and in general , the lowest value of @xmath0 for which @xmath184 will correspond to the ground state solution ; higher values of @xmath0 correspond to oscillating solutions , i.e. excited states with higher energy . so there is a critical coupling @xmath185 , above which one can find a nontrivial solution for the homogeneous equation , which is also the zeroth - order solution for @xmath186 if one expand it in terms of @xmath187 . the only special point is that the ratio @xmath181 must satisfy some constraint . in the limit that the coupling goes towards the critical coupling , @xmath188 , this ratio goes to infinity , and we can remove the ultraviolet and infrared cut - offs . note that this critical coupling coincides with the critical coupling for dynamical mass generation for the fermions in the quenched ladder approximation . this is no surprise , since we noted already that we arrived at the same integral equation as that for the mass function , eq ( [ 314.1conv ] ) . looking back at the original equation for @xmath117 , eq . ( [ spseucl ] ) , we can understand this , because for @xmath189 this equation is exactly the same as the linearized quenched ladder approximation for the fermion propagator . now we calculate @xmath190 , eq . ( [ ts ] ) , which becomes in euclidean space @xmath191 we can solve this equation in the same way as the @xmath16 type equation , redefining @xmath192 which allows us to rewrite this equation into @xmath193 similarly , we expand @xmath194 in terms of @xmath134 @xmath195 inserting this in eq . ( [ 306 ] ) leads to a set of equations for the functions @xmath196 , and in order to show that the only solution for eq . ( [ 306 ] ) is the trivial solution , it is enough to prove that the only solution of the equation @xmath197 is the trivial solution @xmath198 , for all @xmath145 , just as before . the generalized chebyshev polynomials satisfy the identity @xmath199 \frac{1}{p_{1r}q_r } \left[\frac{min(p_{1r},q_r)}{max(p_{1r},q_r)}\right]^{n+1 } \ , , \label{313}\end{aligned}\ ] ] so if we expand the unknown functions @xmath200 in terms of these polynomials @xmath201 we can perform the angular integrals in eq . ( [ 311 ] ) analytically , see appendix [ angint ] . this gives an equation for each of the coefficients separately @xmath202 \nonumber\\ & & + \int^{\lambda^{2}}_{p_{1}^{2}}dq^{2 } \widetilde{g_l^n}^{\mu_{1}\ldots\mu_{n}}_{\nu_{1}\ldots\nu_{l } } ( q_{r},p_{2r},p_{3},p_{4})q^{2 } \left[\delta_{l0 } + \left(-\frac{l}{l+1}\frac{p_{1}^{2}}{q^{2 } } + \frac{l-1}{l+1}\right)\left(\frac{p_{1r}}{q_{r}}\right)^{l } \right ] \bigg\ } \ , . \label{314}\end{aligned}\ ] ] again , for convenience we will drop all irrelevant indices and the spectator momenta in the discussion of this equation , defining @xmath157 as a scalar function of the absolute value of @xmath158 @xmath203 so the equation we have to consider becomes @xmath204 \nonumber\\ & & + \int^{\lambda^{2}}_{p^{2}}dq^{2 } q^2 g(q ) \left[\delta_{l0 } + \left(-\frac{l}{l+1}\frac{p^{2}}{q^{2 } } + \frac{l-1}{l+1}\right)\left(\frac{p}{q}\right)^{l}\right ] \bigg\ } \ , . \label{314conv}\end{aligned}\ ] ] since the @xmath205 and @xmath206 cases are different , we discuss them separately . for @xmath205,the integral equation can be transformed into the following differential equation @xmath207 there are two independent solutions for the above differential equation , say @xmath208 with @xmath209 satisfying @xmath210 \ , . \label{57}\end{aligned}\ ] ] this means that the general solution of the differential equation can be written as @xmath211 with @xmath167 some arbitrary functions , depending on the spectator momenta @xmath168 , @xmath169 , and @xmath170 in the original equation . we substitute this solution back in the original integral equation in order to get the correct boundary conditions , which gives us @xmath212 & = & 0 \,.\end{aligned}\ ] ] the critical coupling for the nontrivial solution is @xmath213 , as can be seen from eq . ( [ 57 ] ) , and just as for the @xmath16 type equation , below the critical coupling this constraint allows only the trivial solution @xmath214 above this critical coupling , were the roots @xmath163 become complex , we can construct a nontrivial solution similar as before . for @xmath206 , the integral equation can be transformed into the following differential equation @xmath215 \ , . \label{59}\end{aligned}\ ] ] now there are four independent solutions for the above differential equation , say @xmath216 with @xmath217 satisfying the equation @xmath218 & = & 0 \,.\end{aligned}\ ] ] we can calculate some special values @xmath219 , to see where the roots are located , and find for small values of the coupling @xmath220<0 \nonumber\\ f^{\alpha}_{l}(l)&=&\frac{\alpha}{12\pi}l(l-1)\geq 0 \nonumber\\ f^{\alpha}_{l}(0)&=&\frac{\alpha}{12\pi}l(l+2)>0 \label{61}\\ f^{\alpha}_{l}(-\frac{1}{2})&= & - \frac{1}{4}\left[l^{2}+2l+\frac{3}{4 } - \frac{\alpha}{3\pi}\left(l^{2 } + \frac{7l}{2}+\frac{9}{4}\right)\right]<0 \nonumber\\ f^{\alpha}_{l}(-1)&=&\frac{\alpha}{12\pi}(l+2)(l+3)>0 \nonumber\end{aligned}\ ] ] from this it follows immediately that at small values of the coupling all four roots of the equation are real : two of them are positive and the other two are negative . in principle , there might be a critical coupling @xmath221 , where two roots become degenerate ; such a critical coupling satisfies the equations @xmath222 this critical coupling will in general depend on the parameter @xmath223 , and from eq . ( [ 61 ] ) we can see immediately that a lower bound for these critical coupling is @xmath224 a more detailed calculations shows that the critical coupling for @xmath225 is indeed @xmath226 . again , the general solution can be written as @xmath227 with @xmath167 depending on the spectator momenta @xmath168 , @xmath169 , and @xmath170 in the original equation eq . ( [ 314 ] ) , and constrained by the integral equation . therefore , we substitute this solution back in the integral equation , which gives us in eq . ( [ lastconstr ] ) is to keep the equations finite in the case of @xmath225 , in which case @xmath228 is a solution . we find in this case that in fact @xmath229 . ] @xmath230 & = & 0 \,,\end{aligned}\ ] ] which is equivalent to @xmath231 we also have the additional constraint that the infrared cut - off is smaller than the ultraviolet cut - off . this set of equations has no nontrivial solution below the critical coupling , with all the roots @xmath232 real , but above the critical coupling , some of @xmath233 develop an imaginary part which means that the solution starts oscillating . just as in the case of @xmath117 , these oscillations allow for nontrivial solutions of the integral equation , with @xmath234 . note , that the critical coupling for @xmath118 depends on @xmath223 , and all of them are bigger than that for @xmath117 which is @xmath235 ; the smallest one , corresponding to @xmath205 is @xmath236 . this means that one needs a stronger coupling to form a nontrivial @xmath17 type function . if the coupling is in the region @xmath237 , only @xmath117 will have a nontrivial solution , while @xmath118 still remains zero ; at @xmath238 this @xmath118 might become nonzero as well . however , one should keep in mind that these @xmath117and @xmath118 type function are not identical to the four - fermion functions @xmath239 and @xmath240 itself , see eqs . ( [ sym1])([sym4 ] ) . so as soon as either @xmath117 ( or @xmath118 ) becomes nonzero , both the @xmath16 and the @xmath17 type nonperturbative four - point functions @xmath239 and @xmath241 will become nonzero . we have shown that , in leading order , both the @xmath16 and the @xmath17 type four - fermion functions and its condensates are zero below the critical coupling @xmath242 in the massless fermion phase of a @xmath1 gauge theory . above this critical coupling , there can be a nontrivial solution for these nonperturbative chiral - symmetry breaking four - point functions . this result is found using an expansion of the effective action , assuming that the other nonperturbative four - fermion functions ( which are not parity or lorentz invariant ) are zero . it agrees with the result found by holdom and triantaphyllou @xcite , who used a truncation of the schwinger dyson equation for the four - fermion functions . our result shows that the critical coupling for chiral symmetry breaking by four - fermion condensates is the same as that for dynamical mass generation . therefore , in a strongly coupled abelian gauge theory , beyond the critical coupling , the chiral symmetry will be broken through the two - fermion condensate since this is energetically favored . this means that in this theory , there is no hierarchy of chiral symmetry breaking , first by the four - fermion condensates , while the fermions remain massless , and another phase transition , at a lower energy , where the fermions acquire a mass . the generalization to the @xmath243 flavor case is straightforward , and does not change our detailed calculations . the only thing we need to do is to label every @xmath244 function with flavor indices . there are no flavor mixing effects in the equations , and the flavor symmetry is not believed to be broken dynamically , so the green s functions for the different flavor indices decouple and the form of the equations is completely same as the ones we discussed . this means that in a @xmath245 gauge theory it seems to be impossible to use four - fermion functions or its condensates to substitute the two - fermion condensate as the chiral order parameter . in a more complicated theory like a chiral gauge theory , as suggested in @xcite , there exists in principle the possibility that chiral symmetry is broken through the forming of four - fermion condensates while the fermions remain massless . for such a scenario to be true , there should be some additional symmetry to prevent the fermions from becoming massive . whether such a hierarchy of chiral symmetry breaking can indeed be realized in a chiral gauge theory , remains an open question ; we only showed that in a @xmath1 gauge theory the critical coupling for the four - fermion condensates is the same as that for the two - fermion condensates . a similar result has been obtained using a truncation of the schwinger dyson equation for the four - fermion functions @xcite . however , the problem with the set of schwinger dyson equations is that it forms an infinite set of coupled equations , and it is not at all obvious how to truncate this set of equations in a consistent way . using an approximation which is similar to the quenched ladder approximation for the fermion propagator schwinger dyson equation , holdom and triantaphyllou have shown numerically that the critical coupling for chiral symmetry breaking by four - fermion condensates is almost the same as that for dynamical mass generation . it is remarkable that both their calculation and our analytical calculation give the same result ( the same critical coupling ) , although the set of feynman diagrams included in their and our truncation is not the same . this indicates that the essential diagrams are included in both schemes , and that the exact details of the truncation are less important . of course , this should be confirmed by a study of the next - to - leading order effective potential , or a more accurate truncation scheme for the schwinger dyson equation . we should also consider alternative formulations for the usual effective action , since there are different ways to introduce the auxiliary fields , which might lead to a different result . we would like to thank koichi yamawaki , yoonbai kim , and bob holdom for useful and stimulating discussions . this work has been supported in part by the nishina memorial fund of japan , the natural science fund of china , and the japanese society for the promotion of science . in this appendix we discuss the angular integration . consider the case with three independent momenta , and choose the axes ( in euclidean space ) to be @xmath246 where we use the notation @xmath247 for a four - dimensional vector and @xmath248 for its absolute value . the angular integration we are interested in is @xmath249 where the last integration is performed along the unit circle in the complex @xmath105 plane . there are two poles inside the circle , one is zero and the other is @xmath250 , which is defined as @xmath251}{\max[p_{1},q ] } , \frac{\max[p_{1},q]}{\min[p_{1},q]}\bigg\}.\ ] ] their contributions to the integration are denoted by @xmath252 and @xmath253 respectively . we further find @xmath254}c_{n+1}^{k } \sin[(n+1 - 2k)\theta_{p } ] \bigg[u_{n-1 - 2k}\left(\frac{p_{1}^{2}+q^{2}}{2p_{1}q}\right ) \nonumber\\ & & -u_{n+1 - 2k}\left(\frac{p_{1}^{2}+q^{2}}{2p_{1}q}\right)\bigg ] - \sum_{k=[\frac{n}{2}]}^{[\frac{n+1}{2 } ] } c_{n+1}^{k } \sin[(n+1 - 2k)\theta_{p } ] u_{n+1 - 2k } \left(\frac{p_{1}^{2}+q^{2}}{2p_{1}q}\right ) \bigg\ } \ , , \label{203}\\ i_{n,1}&=&-\frac{q^{n-1}p_{2}^{n}}{2^{n+1}(n+1)p_{1}\sin\theta_{p } } \sum_{k=0}^{n+1 } c_{n+1}^{k}t_{-}^{-n-1 + 2k } \sin[(n+1 - 2k)\theta_{p } ] \,,\end{aligned}\ ] ] where we have used the standard chebyshev expansion @xcite @xmath255 using another property of the chebyshev functions @xmath256=u_{n}(\cos\theta ) \sin\theta \,,\label{205}\end{aligned}\ ] ] the result of eq . ( [ 202 ] ) can be further expressed as @xmath257 where @xmath258 to calculate @xmath259 , we consider the quantity @xmath260 \nonumber\\ & = & \frac{1}{t}\left[-1 + \frac{1-t^{2}-2t\sqrt{x^{2}-1}}{1 - 2xt+t^{2}}\right ] \nonumber\\ & = & 2\frac{x-\sqrt{x^{2}-1}}{1-t(x-\sqrt{x^{2}-1 } ) } \nonumber\\ & = & 2 \sum_{n=0}^{\infty}(x-\sqrt{x^{2}-1})^{n+1}t^{n } \,,\end{aligned}\ ] ] where eq . ( [ 204 ] ) is used . comparing both sides of the above equation , we find @xmath261 now use the definition of the chebyshev polynomials @xcite @xmath262}(-1)^{k}\frac{(n - k ) ! } { k!(n-2k)!}(2x)^{n-2k } \ , , \label{208}\end{aligned}\ ] ] to obtain @xmath263 in which @xmath264 and @xmath265 are defined as @xmath266 finally , to calculate the coefficients @xmath267 , we introduce two functions @xmath268 so we get @xmath269 for @xmath270 we find @xmath271 whereas for @xmath272 @xmath273 \ , , \label{218}\\ { \bar f}_{n , k}(x)&= & \frac{1}{n!}x^{-2k-2}\frac{d^{n-1}}{dx^{n-1}}[x^{2k+n+1}(1-x)^{n}]\ , , \label{219}\end{aligned}\ ] ] so the total result can be written as @xmath274 substituting this result into eqs . ( [ 209 ] ) and ( [ 210 ] ) , we finally get the result @xmath275}{\max[p_{1},q ] } \bigg]^{n+1}u_{n}({\hat p}_{1}\cdot{\hat p}_{2 } ) \ , , \label{405}\end{aligned}\ ] ] where @xmath276 denotes the angular dependence . next , we introduce a `` generalized chebyshev function '' with lorentz indices @xmath277 by extracting the @xmath278 dependence in @xmath279 @xmath280 which defines the generalized chebyshev functions @xmath281 as @xmath282 \ , , \\ { \hat u}_{2n+1}^{\mu_{1}\ldots\mu_{2n+1}}({\hat q } ) & = & \sum_{k=0}^{n } \frac{(-1)^{k}2^{2n+1 - 2k}(2n+1-k)!}{k!(2n+1 - 2k)!(2n+1 ) ! } \big[{\hat q}^{\mu_{1}}\ldots{\hat q}^{\mu_{2n+1 - 2k } } \delta^{\mu_{2n+2 - 2k}\mu_{2n+3 - 2k}}\ldots \delta^{\mu_{2n}\mu_{2n+1 } } \nonumber\\ & & + { \hbox { all symmetric permutations in } } ( \mu_{1}\ldots\mu_{2n+1 } ) \big ] \,.\end{aligned}\ ] ] since in eq . ( [ 405 ] ) , @xmath283 is an arbitrary momentum variable , we can use the expansion eq . ( [ 407 ] ) to find @xmath284}{\max[p_{1},q ] } \bigg]^{n+1 } { \hat u}_{n}^{\mu_{1}\ldots\mu_{n}}({\hat p}_{1 } ) \,.\end{aligned}\ ] ] this formula allows us to perform all of the angular integration which are needed , rest us to show that we can indeed expand our four - point functions in terms of these generalized chebyshev functions . the chebyshev polynomials @xmath285 form a complete set of independent functions , so from eq . ( [ 407 ] ) , it follows immediately that our generalized chebyshev functions are also a complete set for functions which are totally symmetric in their lorentz indices . the explicit expression of this statement can be seen from following identities , noting that @xmath286 which follows directly from eq . ( [ 407 ] ) and the fact the chebyshev polynomials form a complete set . now due to the fact that @xmath278 is arbitrary , the eqs ( [ 415 ] ) and ( [ 416 ] ) are equivalent to @xmath287 \label{417}\\ { \hat q}^{\mu_{1}}\ldots{\hat q}^{\mu_{2n+1}}&= & \sum_{k=0}^{n } \frac{c'_{n , k}}{(2n+1 ) ! } \big[{\hat u}_{2k+1}^{\mu_{1}\ldots\mu_{2k+1}}({\hat q } ) \delta^{\mu_{2k+2}\mu_{2k+3}}\ldots \delta^{\mu_{2n}\mu_{2n+1 } } \nonumber \\ & & + { \hbox { all symmetric permutations in } } ( \mu_{1}\ldots\mu_{2n+1})\big ] \ , . \label{418}\end{aligned}\ ] ] so all @xmath288 can be expanded in terms of our generalized chebyshev polynomials . since any smooth function of @xmath289 can always be expanded in terms of @xmath290 times a function which depends only on the absolute value @xmath291 , this means that we can always expand our four - point functions in terms of these chebyshev functions . b. holdom , _ four fermion dynamics and fermion masses _ , utpt9522 , hep - ph/9510249 , to appear in the proceedings of the yukawa international seminar 95 : from the standard model to grand unified theories , kyoto , japan , 21 - 25 august 1995 .

in this paper we present recent results toward the computation of the homflypt skein module of the lens spaces , , via braids . our starting point is the knot theory of the solid torus st and the lambropoulou invariant , , for knots and links in st , the universal analogue of the homflypt polynomial in st . the relation between and is established in and it is shown that in order to compute , it suffices to solve an infinite system of equations obtained by performing all possible braid band moves on elements in the basis of , , presented in . the solution of this infinite system of equations is very technical and is the subject of a sequel paper .

1702.06290

skein modules were independently introduced in 1987 by przytycki @xcite and turaev @xcite . they generalize knot polynomials in @xmath5 to knot polynomials in arbitrary 3-manifolds . the essence is that skein modules are quotients of free modules over ambient isotopy classes of links in 3-manifolds by properly chosen local ( skein ) relations . let @xmath6 be an oriented @xmath7-manifold , @xmath8 $ ] , @xmath9 the set of all oriented links in @xmath6 up to ambient isotopy in @xmath6 and let @xmath10 be the submodule of @xmath11 generated by the skein expressions @xmath12 , where @xmath13 , @xmath14 and @xmath15 comprise a conway triple represented schematically by the illustrations in figure [ skein ] . locally.,width=163 ] for convenience we allow the empty knot , @xmath16 , and add the relation @xmath17 , where @xmath18 denotes the trivial knot . then the _ homflypt skein module _ of @xmath6 is defined to be : @xmath19,u^{-1 } l_{+ } -ul_{- } -zl{}_{0 } \right)={\raise0.7ex\hbox{$ r\mathcal{l } $ } \!\mathord{\left/ { \vphantom { r\mathcal{l } s } } \right . \kern-\nulldelimiterspace}\!\lower0.7ex\hbox{$ s $ } } .\ ] ] skein modules of @xmath7-manifolds have become very important algebraic tools in the study of @xmath7-manifolds , since their properties renders topological information about the @xmath7-manifolds . unlike the kauffman bracket skein module , the homflypt skein module of a @xmath7-manifold , also known as _ conway skein module _ and as _ third skein module _ , is very hard to compute and very little is known so far . more precisely , @xmath20 $ ] , where the empty link is a generator of the module ( @xcite , @xcite ) . also , for the solid torus st , @xmath3 is a free , infinitely generated @xmath21$]-module isomorphic to the symmetric tensor algebra @xmath22 , where @xmath23 denotes the conjugacy classes of non trivial elements of @xmath24 ( @xcite , @xcite , @xcite , @xcite ) . further , let @xmath25 denote a surface . then @xmath26 is an algebra which , as an @xmath27 module , is a free module isomorphic to the symmetric tensor algebra , @xmath28 , where @xmath29 denotes the conjugacy classes of nontrivial elements of @xmath30 ( @xcite ) . moreover , @xmath31 is freely generated by standard oriented unlinks as presented in figure [ mroc ] ( @xcite ) and @xmath32 is freely generated by the empty link over a properly chosen ring ( @xcite ) . finally , @xmath33 is isomorphic to @xmath34 modulo torsion , where @xmath35 are oriented 3-manifolds and @xmath36 their connected sum ( @xcite ) . in @xcite the homflypt skein module of the lens spaces @xmath0 is computed using diagrammatic method . the diagrammatic method could in theory be generalized to the case of @xmath37 , but the diagrams become even more complex to analyze and several induction arguments fail . .,width=124 ] in @xcite the most generic analogue of the homflypt polynomial , @xmath2 , for links in the solid torus @xmath38 has been derived from the generalized hecke algebras of type @xmath39 , @xmath40 , via a unique markov trace constructed on them . this algebra was defined by lambropoulou in @xcite and is related to the knot theory of the solid torus , the artin group of coxeter group of type b , @xmath41 , and to the affine hecke algebra of type a. the lambropoulou invariant @xmath2 recovers the homflypt skein module of st , @xmath3 , and is appropriate for extending the results to the lens spaces @xmath42 , since the combinatorial setting is the same as for @xmath38 , only the braid equivalence includes the braid band moves ( shorthanded to _ bbm _ ) , which reflect the surgery description of @xmath42 . for the case of @xmath0 , in order to extend @xmath2 to an invariant of links in @xmath0 we need to solve an infinite system of equations resulting from the braid band moves . namely we force : @xmath43 for all @xmath44 and for all possible slidings of @xmath45 . the above equations have particularly simple formulations with the use of a new basis , @xmath4 , for the homflypt skein module of @xmath38 , that we give in @xcite . this basis was predicted by przytycki and is crucial in this paper , since _ bbm _ s are naturally described by elements in this basis . in order to show that the set @xmath4 is a basis for @xmath46 , we started in @xcite with the well - known basis of @xmath46 , @xmath47 , discovered independently in @xcite and @xcite with diagrammatic methods , and a basis @xmath48 of the algebra @xmath40 and we followed the steps below : * an ordering relation in @xmath47 is defined and it is shown that the set is totally ordered . * elements in @xmath47 are converted to linear combinations of elements in the new set @xmath4 as follows : * elements in @xmath47 are first converted to elements in the linear basis of @xmath49 , @xmath48 . * using conjugation , the gaps appearing in the indices of the looping generators in the monomials in @xmath50 are managed . * using conjugation , the exponents of the looping generators are ordered . * using conjugation and stabilization moves , the ` braiding tails ' are removed from the above monomials and thus , the initial elements in @xmath50 are converted to linear combination of elements in @xmath4 . * finally , the sets @xmath47 and @xmath4 are related via a block diagonal matrix , where each block is an infinite lower triangular matrix . * the diagonal elements in the above matrix are invertible , making the matrix invertible and thus , the set @xmath4 is a basis for @xmath46 . the new basis is appropriate for computing the homflypt skein module of the lens spaces @xmath42 in general . note that @xmath46 plays an important role in the study of homflypt skein modules of arbitrary c.c.o . @xmath7-manifolds , since every c.c.o . @xmath7-manifold can be obtained by surgery along a framed link in @xmath5 with unknotted components . the family of the lens spaces , @xmath42 , comprises the simplest example , since they are obtained by rational surgery on the unknot . equations ( [ eqbbm ] ) are very controlled in the algebraic setting , because , as shown in @xcite , they can be performed only on elements in @xmath4 . this is shown by following the technique developed in @xcite . the difference lies in the fact that here we deal with elements in @xmath4 and at the same time with their result after the performance of a _ bbm _ and we keep track of how _ bbm _ s affect the steps described above . more precisely , in @xcite we followed the steps below : * equations ( [ eqbbm ] ) boil down by linearity to considering only words in the canonical basis @xmath51 of the algebra @xmath52 . * for words in @xmath53 equations of the form @xmath54 are obtained , where @xmath55 is the result of the performance of a braid band move on the _ first _ moving strand of the closed braid @xmath56 , and @xmath57 . * then , elements in @xmath53 are expressed to elements in the linear basis @xmath48 of @xmath52 and it is shown that the equations for words in @xmath53 are equivalent to equations of the form @xmath58 , where @xmath59 . * a set @xmath60 is then introduced , consisting of monomials in the looping generators @xmath61 s with no gaps in the indices ( as in @xmath4 ) but not necessarily ordered exponents . * equations for words in @xmath50 are now reduced to equations obtained from elements in the @xmath52-module @xmath60 , where the braid band moves are performed on _ any moving strand_. * equations of the form @xmath62 are now obtained , where @xmath63 is the result of the performance of a braid band move on the @xmath64 moving strand of the closed braid @xmath65 , and @xmath66 an element in the augmented set @xmath60 followed by a ` braiding tail ' . * using conjugation , the exponents then become in decreasing order and equations obtained from elements in the @xmath52-module @xmath60 by performing _ s on all moving strands are reduced to equations for words in @xmath52-module @xmath4 by performing _ s on all moving strands . * the ` braiding tails ' from elements in the @xmath52-module @xmath4 are now eliminated and it is shown that equations for words in the @xmath52-module @xmath4 by performing braid band moves on any strand , are now reduced to equations obtained from elements in the basis @xmath4 of @xmath3 by performing braid band moves on every moving strand : @xmath67 in @xcite we elaborate on the infinite system . the importance of our approach is that it can shed light on the problem of computing skein modules of arbitrary c.c.o . @xmath7-manifolds , since any @xmath7-manifold can be obtained by surgery on @xmath5 along unknotted closed curves . indeed , one can use our results in order to apply a braid approach to the skein module of an arbitrary c.c.o . the main difficulty of the problem lies in selecting from the infinitum of band moves ( or handle slide moves ) some basic ones , solving the infinite system of equations and proving that there are no dependencies in the solutions . note that the computation of @xmath1 is equivalent to constructing all possible analogues of the homflypt or 2-variable jones polynomial for knots and links in @xmath0 , since the linear dimension of @xmath1 means the number of independent homflypt - type invariants defined on knots and links in @xmath0 . the paper is organized as follows : in [ basics ] we recall the setting and the essential techniques and results from @xcite . more precisely , we describe braid equivalence for knots and links in @xmath0 and we present a sharpened version of the reidemeister theorem for links in @xmath0 . we also provide geometric formulations of the braid equivalence via mixed braids in @xmath5 using the @xmath68-moves and the braid band moves and give algebraic formulations in terms of the mixed braid groups @xmath69 . in [ solidtorus ] we present results from @xcite and @xcite . more precisely , we recover the homflypt skein module of the solid torus st , @xmath3 , via algebraic techniques and we present a new basis for @xmath3 , @xmath4 , from which the braid band moves are naturally described . the aim of this section is to set a homogeneous ground in computing skein modules of c.c.o . @xmath7-manifolds in general via algebraic means . in 3 we derive the relation between @xmath3 and @xmath1 and show that in order to compute @xmath1 we only need to consider elements in the basis @xmath4 of @xmath3 and impose on the lambropoulou invariant @xmath2 relations coming by performing all possible braid band moves on elements in @xmath4 . we consider st to be the complement of a solid torus in @xmath5 . then , an oriented link @xmath68 in st can be represented by an oriented _ mixed link _ in @xmath70 , that is , a link in @xmath70 consisting of the unknotted _ fixed part _ @xmath71 representing the complementary solid torus in @xmath5 and the _ moving part _ @xmath68 that links with @xmath71 . mixed link diagram _ is a diagram @xmath72 of @xmath73 on the plane of @xmath71 , where this plane is equipped with the top - to - bottom direction of @xmath74 ( see @xcite ) . .,width=105 ] it is well known that the lens spaces @xmath0 can be obtained from @xmath5 by surgery on the unknot with surgery coefficient @xmath75 . surgery along the unknot can be realized by considering first the complementary solid torus and then attaching to it a solid torus according to some homeomorphism on the boundary . thus , isotopy in @xmath0 can be viewed as isotopy in st together with the _ band moves _ in @xmath5 , which reflect the surgery description of the manifold , see figure [ bmov ] ( see @xcite ) . in @xcite we show that in order to describe isotopy for knots and links in a c.c.o . @xmath7-manifold , it suffices to consider only the type @xmath45 band moves ( see figure [ unknsurg ] ) and thus , isotopy between oriented links in @xmath0 is reflected in @xmath5 by means of the following result ( cf . * theorem 5.8 ) , ( * ? ? ? * theorem 6 ) ) : _ two oriented links in @xmath0 are isotopic if and only if two corresponding mixed link diagrams of theirs differ by isotopy in st together with a finite sequence of the type @xmath45 band moves . _ band move follows from a type-@xmath45 band move in the case of integral surgery coefficient . , width=480 ] by the alexander theorem for knots in solid torus ( ( * ? ? ? * theorem 1 ) ) , a mixed link diagram @xmath72 of @xmath73 may be turned into a _ mixed braid _ @xmath76 with isotopic closure . this is a braid in @xmath70 where , without loss of generality , its first strand represents @xmath71 , the fixed part , and the other strands , @xmath66 , represent the moving part @xmath68 . the subbraid @xmath66 is called the _ moving part _ of @xmath76 ( see fig . [ mbtoml ] ) . the sets of mixed braids related to st form groups , which are in fact the artin braid groups of type b. the mixed braid group on @xmath77 moving strands is denoted as @xmath69 and it admits the following presentation : @xmath78 where the ` braiding ' generators @xmath79 and the ` looping ' generator @xmath80 are illustrated in figure [ gen ] ( @xcite ) . .,width=220 ] in @xmath69 we also define the elements : @xmath81 for @xmath82 , where @xmath83 . for illustrations see figure [ genh ] . [ tprt]in ( * ? ? ? 1 ) it is shown using the artin combing , that every element in @xmath41 can be written in the form @xmath84 , where @xmath85 is a word in the looping elements @xmath86 and @xmath87 is the ` braiding tail ' . furthermore , the looping elements @xmath86 are conjugates . on the other hand , the elements @xmath61 commute . moreover , the @xmath61 s are consistent with the braid band move used in the link isotopy in @xmath0 , in the sense that a braid band move can be described naturally with the use of the @xmath61 s . [ loopttpr ] and @xmath88.,width=307 ] in @xcite the authors give a sharpened version of the classical markov theorem for braid equivalence in @xmath5 based only on one type of moves , the @xmath68-moves : an _ @xmath68-move _ on a mixed braid @xmath89 , consists in cutting an arc of the moving sub - braid open and pulling the upper cut point downward and the lower upward , so as to create a new pair of braid strands with corresponding endpoints , and such that both strands cross entirely _ over _ or _ under _ with the rest of the braid . stretching the new strands over will give rise to an _ @xmath90-move _ and under to an _ @xmath91-move_. for an illustration see figure [ lmove ] . an algebraic @xmath68-move has the following algebraic expression for an @xmath90-move and an @xmath91-move respectively : @xmath92 where @xmath93 , @xmath94 are elements of @xmath69 and @xmath95 , @xmath96 are obtained from @xmath93 , @xmath94 by replacing each @xmath97 by @xmath98 for @xmath99 . -moves , width=470 ] in order to translate isotopy for links in @xmath0 into mixed braid equivalence , we first define a _ braid band move _ to be a move between mixed braids , which is a type @xmath45 band move between their closures . it starts with a little band oriented downward , which , before sliding along a surgery strand , gets one twist _ positive _ or _ negative _ ( see figure [ bbm12 ] ) . then , isotopy in @xmath0 is translated on the level of mixed braids by means of the following theorem : ( * ? ? ? * theorem 5 ) [ markov ] let @xmath100 be two oriented links in @xmath0 and let @xmath101 be two corresponding mixed braids in @xmath70 . then @xmath102 is isotopic to @xmath103 in @xmath0 if and only if @xmath104 is equivalent to @xmath105 in @xmath106 by the following moves : @xmath107 where @xmath108 is the word @xmath45 with all indices shifted by 1 ( see figure [ bbm12 ] ) . equivalently , @xmath102 is isotopic to @xmath103 if and only if @xmath104 and @xmath105 differ by moves ( iii ) and ( iv ) as above , while moves ( i ) and ( ii ) are replaced by the two types of @xmath68-moves , see eq . ( [ l - m ] ) . we denote a braid band move by _ bbm _ and , specifically , the result of a positive or negative braid band move performed on the @xmath64-moving strand of a mixed braid @xmath66 by @xmath109 . note that in the statement of theorem [ markov ] in @xcite the braid band moves take place on the last strand of a mixed braid . clearly , this is equivalent to performing the braid band moves on the first moving strand or , in fact , on any specified moving strand of the mixed braid . indeed , in a mixed braid @xmath110 consider the last moving strand of @xmath66 approaching the surgery strand @xmath74 from the right . before performing a _ bbm _ we apply conjugation ( isotopy in st ) and we obtain an equivalent mixed braid where the first strand is now approaching @xmath74 ( see figure [ bbmconj ] ) . in terms of diagrams we have the following ( ( * ? ? ? * lemma 1 ) ) : @xmath111 it is well - known that @xmath69 is the artin group of the coxeter group of type b , which is related to the hecke algebra of type b , @xmath112 and to the cyclotomic hecke algebras of type b. in @xcite it has been established that all these algebras are related to the knot theory of st . the basic one is @xmath112 , a presentation of which is obtained from the presentation ( [ presb ] ) of the artin group @xmath69 by corresponding the braiding generator @xmath113 to @xmath114 and by adding the quadratic relations : @xmath115 and the relation @xmath116 . the cyclotomic hecke algebras of type b are denoted @xmath117 , and they admit presentations that are obtained by the quadratic relation ( [ quad ] ) and the relation @xmath118 . in @xcite also the _ generalized iwahori hecke algebra of type b _ , @xmath119 , is introduced , as the quotient of @xmath120b_{1,n}$ ] over the quadratic relations ( [ quad ] ) . namely : @xmath121b_{1,n } } { \langle \sigma_i^2 -\left(q-1\right)\sigma_i - q \rangle}.\ ] ] note that in @xmath119 the generator @xmath80 satisfies no polynomial relation , making the algebra @xmath119 infinite dimensional , and as observed by t. tom dieck , @xmath122 is closely related to the affine hecke algebra of type a , @xmath123 . in @xcite the algebra @xmath119 is denoted as @xmath124 . in @xcite v.f.r . jones gives the following linear basis for the iwahori - hecke algebra of type a , @xmath125 : @xmath126 for @xmath127 . the basis @xmath10 yields directly an inductive basis for @xmath125 , which is used in the construction of the ocneanu trace , leading to the homflypt or @xmath128-variable jones polynomial . we also introduce in @xmath49 the ` looping elements ' @xmath129 that correspond to the elements of eq . [ lgen ] of @xmath69 . from @xcite we have : ( * ? ? ? * proposition 1 & theorem 1 ) [ basesh ] the following sets form linear bases for @xmath52 : @xmath130 where @xmath131 and @xmath132 a basic element in @xmath133 . for an illustration of the @xmath61 s and the @xmath86 s recall figure [ genh ] . [ conind]note that the form of the elements in the set @xmath134 is consistent with the artin combing in the mixed braid group @xmath41 ( recall remark [ tprt ] ) . however the looping part of a mixed braid after the combing contains elements of a free group on @xmath77 generators , so the indices as well as the exponents have no restrictions . in theorem [ basesh ] though , the indices of the @xmath61 s and the @xmath135 s in the above sets are ordered but are not necessarily consecutive , neither do they need to start from @xmath80 . also , the exponents are arbitrary . [ sspr ] we shall denote @xmath136 in @xcite , @xcite the homflypt skein module of the solid torus st has been computed using diagrammatic methods by means of the following theorem : [ turaev ] the skein module @xmath3 is a free , infinitely generated @xmath21$]-module isomorphic to the symmetric tensor algebra @xmath22 , where @xmath23 denotes the conjugacy classes of non trivial elements of @xmath24 . in the diagrammatic setting of @xcite and @xcite , st is considered as @xmath137 . a basic element of @xmath3 in the context of @xcite , is illustrated in figure [ tur ] . the homflypt skein module of st is particularly important , because any closed , connected , oriented ( c.c.o . ) @xmath7-manifold can be obtained by surgery along a framed link in @xmath5 with unknotted components . such component can be viewed as a solid torus complementing st in @xmath5 . .,width=124 ] in @xcite the bases @xmath51 are used for constructing a markov trace on @xmath139 . * theorem 6 ) [ tr ] given @xmath140 , with @xmath141 specified elements in @xmath142 $ ] , there exists a unique linear markov trace function @xmath143 determined by the rules : @xmath144 note that the use of the looping elements @xmath86 enable the trace @xmath145 to be defined by just extending by rule ( 4 ) the three rules of the ocneanu trace on the algebras @xmath146 @xcite , recall remark [ tprt ] . using @xmath147 the second author constructed a universal homflypt - type invariant for oriented links in st . namely , let @xmath9 denote the set of oriented links in st . then : ( * ? ? * definition 1 ) [ inv ] the function @xmath148 @xmath149 @xmath150 where @xmath151 , @xmath152 , @xmath153 is a word in the @xmath79 s and @xmath154 s , @xmath155 is the closure of @xmath45 , @xmath156 is the exponent sum of the @xmath79 s in @xmath157 , and @xmath158 the canonical map of @xmath69 on @xmath52 , such that @xmath159 and @xmath160 , is an invariant of oriented links in st . in the braid setting of @xcite , the elements of @xmath3 correspond bijectively to the elements of the following set @xmath47 : @xmath161 so , we have that @xmath47 is a basis of @xmath3 in terms of braids . note that @xmath47 is a subset of @xmath162 and , in particular , @xmath47 is a subset of @xmath163 . note also that in contrast to elements in @xmath163 , the elements in @xmath47 have no gaps in the indices , the exponents are ordered and there are no ` braiding tails ' . the lambropoulou invariant @xmath2 recovers @xmath3 , because it gives distinct values to distinct elements of @xmath47 , since @xmath164 . note also that the invariant @xmath2 is defined by the skein relation : @xmath165 and an infinitum of initial conditions , one for each element in @xmath3 as shown in figure 12 . namely , if @xmath166 denotes the unknot , then @xmath167 and @xmath168 . in @xcite we give a different basis @xmath4 for @xmath3 , which was predicted by j. przytycki and which is described in eq . [ basis ] in open braid form . * theorem 2)[newbasis ] the following set is a @xmath169$]-basis for @xmath3 : @xmath170 the importance of the new basis @xmath4 of @xmath3 lies in the simplicity of the algebraic expression of a braid band move , which extends the link isotopy in st to link isotopy in @xmath0 and this fact was our motivation for establishing this new basis @xmath4 ( recall theorem 1(iv ) , remark [ tprt ] and figure [ bbm12 ] ) . .,width=153 ] .,width=124 ] notice that comparing the set @xmath4 with the set @xmath171 , we observe that there are no gaps in the indices of the @xmath61 s and the exponents are in decreasing order . also , there are no ` braiding tails ' in the words in @xmath4 . our method for proving theorem [ newbasis ] is the following : @xmath172 we define total orderings in the sets @xmath47 and @xmath4 , @xmath172 we show that the two ordered sets are related via a lower triangular infinite matrix with invertible elements on the diagonal , and @xmath172 using this matrix , we show that the set @xmath4 is linearly independent . in order to relate the two sets via a lower triangular infinite matrix we start with elements in the basic set @xmath47 and we first convert them into sums of elements in @xmath171 , containing the linear bases of the algebras @xmath49 . these elements consist of two parts : arbitrary monomials in the @xmath61 s followed by ` braiding tails ' in the bases of the algebras @xmath146 . then , these elements are converted into elements in the set @xmath4 by : * managing the gaps in the indices , * by ordering the exponents of the @xmath61 s and * by eliminating the ` braiding tails ' . it is worth mentioning that these procedures are not independent in the sense that when , for example one manages the gaps in the indices of the looping generators @xmath61 s , ` braiding tails ' may occur and also the exponents of the @xmath61 s may alter . similarly , when the ` braiding tails ' are eliminated , gaps in the indices of the @xmath61 s might occur . this is a long procedure that eventually stops and only elements in the set @xmath4 remain . for defining orderings in the sets @xmath174 and @xmath47 we need the notion of the _ index _ of a word @xmath175 in any of these sets , denoted @xmath176 . in @xmath47 or @xmath4 @xmath176 is defined to be the highest index of the @xmath86 s , resp . of the @xmath61 s in @xmath175 . similarly , in @xmath163 or @xmath171 , @xmath176 is defined as above by ignoring possible gaps in the indices of the looping generators and by ignoring the braiding parts in the algebras @xmath125 . moreover , the index of a monomial in @xmath125 is equal to @xmath177 . @xmath178 we now proceed with defining an ordering relation in the sets @xmath171 and @xmath163 , which passes to their respective subsets @xmath4 and @xmath47 : ( * ? ? ? * definition 2 ) [ order ] let @xmath179 and @xmath180 in @xmath163 , where @xmath181 for all @xmath182 and @xmath183 . then , we define the following ordering in @xmath163 : * if @xmath184 , then @xmath185 . * if @xmath186 , then : + \(i ) if @xmath187 , then @xmath185 , + \(ii ) if @xmath188 , then : + ( @xmath45 ) if @xmath189 , then @xmath190 , + ( @xmath66 ) if @xmath191 for all @xmath80 and @xmath192 , then @xmath185 , + ( @xmath193 ) if @xmath191 for all @xmath80 and @xmath194 and @xmath195 , then @xmath185 , + ( @xmath196 ) if @xmath197 and @xmath198 , @xmath199 , then @xmath200 . the ordering in the set @xmath171 is defined as in @xmath163 , where @xmath86 s are replaced by @xmath61 s . @xmath201 in order to eventually get to the infinite matrix relating the two basic sets @xmath47 and @xmath4 we need to define the _ subsets of level @xmath202 _ , @xmath203 and @xmath204 , of @xmath4 and @xmath47 respectively ( ( * ? ? ? * definition 3 ) ) , to be the sets : @xmath205 in @xcite it was shown that the sets @xmath203 and @xmath204 are totally ordered and well ordered for all @xmath202 ( ( * ? ? ? * propositions 1 & 2 ) ) . note that the sets @xmath4 and @xmath47 admit a natural grading : @xmath206 and @xmath207 . in the rest of the section we will be using the ordering in the transitions from @xmath163 to @xmath171 and from @xmath171 to @xmath4 . in this subsection we convert monomials in the @xmath135 s to expressions containing the @xmath61 s . full details and related technical lemmas are provided in @xcite . in order to simplify the expressions in this step we first introduce the following notation . [ nt ] we set @xmath208 and @xmath209 , for @xmath210 , @xmath211 for all @xmath212 . we now introduce the notion of _ homologous words _ , which is crucial for relating the sets @xmath47 and @xmath4 via a triangular matrix . * definition 4 ) we say that two words @xmath213 and @xmath214 are _ homologous _ , denoted @xmath215 , if @xmath175 is obtained from @xmath216 by changing @xmath135 into @xmath61 for all @xmath217 . the words @xmath218 and @xmath219 are homologous . note also that @xmath220 . in ( * ? ? ? * lemma 11 ) it is shown that the following relations hold in @xmath52 for @xmath221 : @xmath222 where @xmath223 , @xmath224 and @xmath225 , if @xmath226 and @xmath227 , if @xmath228 . using now these relations , we have that every element in @xmath47 can be expressed as a linear sum of each homologous word , the homologous word with ` braiding tails ' and elements in @xmath171 of lower order , as illustrated abstractly in figure [ fthm7 ] . more precisely : ( * ? ? ? * theorem 7)[convert ] the following relations hold in @xmath52 for @xmath229 : @xmath230 where @xmath231 , for all @xmath232 , @xmath233 a monomial of the @xmath61 s such that @xmath234 for all @xmath212 and @xmath235 , for all @xmath232 . .,width=384 ] [ eg1]we shall now give examples of monomials in @xmath47 converted into sums of elements in @xmath171 using technical lemmas from @xcite ( ( * ? ? ? * lemmas 3 , 4 , 5 , 6 , 9 & 11 ) ) . the examples provide at the same time the motivation for the subsections that follow . * consider the monomial @xmath236 . we have that : + @xmath237\ + \\ & & \\ & + & tt_1^{-1}\cdot \left [ ( q-1)(q^2-q+1)\cdot g_2^{-1}- ( q-1)^2\cdot g_1g_2g_1^{-1}g_2^{-1 } \right]+\\ & & \\ & + & tt_2^{-1}\cdot \left[q^2(q-1)\cdot g_2^{-1}+q(q-1)^3\cdot g_2^{-1}-q(q-1)^2\cdot g_2g_1^{-1}g_2^{-1 } \right]+\\ & & \\ & + & t_1t_2^{-1}\cdot \left [ q(q-1)\cdot g_2g_1^{-1}g_2^{-1}-q(q-1)^2\cdot g_1^{-1}g_2^{-1 } \right]+\\ & & \\ & + & t^{-1}t_1 \cdot \left[-(q-1)\cdot g_2g_1^{-1}g_2^{-1 } - q^{-1}(q-1)^2\cdot g_1^{-1}g_2^{-1 } \right ] \end{array}\ ] ] + so , we obtain the homologous word @xmath238 , the word @xmath175 again followed by the braiding element @xmath239 and terms in @xmath171 of order less than @xmath175 : since either their index is less that @xmath176 ( the terms @xmath240 , @xmath241 and @xmath242 ) , or they contain gaps in the indices ( the terms @xmath243 and @xmath244 ) . * a simpler example is the following : @xmath245 + we obtained the homologous word @xmath246 , the element @xmath247 in @xmath171 comprising the monomial @xmath248 with gaps in the indices , followed by ` braiding tail ' @xmath239 and also the lower order term @xmath249 in the @xmath133-module @xmath4 . in theorem [ convert ] note that in the right hand side there are terms which do not belong to the set @xmath4 . the point now is that these terms are elements in the set of bases @xmath171 on the hecke algebras @xmath49 , but , when we are working in @xmath3 , which is the knots and links level , such elements must be considered up to conjugation by any generator of the alegbra and up to stabilization moves ( recall theorem [ markov ] ) . topologically , conjugation corresponds to closing a mixed braid . in this subsection we show how to deal with monomials in @xmath171 where the looping elements do not have consecutive indices . we call _ gaps _ in monomials of the @xmath61 s any gaps occurring in the indices . after managing , that is eliminating , the gaps we pass to the augmented @xmath146-module @xmath60 , which consists of monomials in the @xmath61 s with consecutive indices but not necessarily ordered exponents . * definition 3)[expsetl ] we define the sets : @xmath250 and the _ subset of level @xmath202 _ of @xmath60 @xmath251 : @xmath252 note that in @xcite the set @xmath60 is denoted by @xmath68 . obviously the set @xmath4 ( eq . 10 ) is a subset of @xmath60 . whenever we talk about a module with coefficients in @xmath253 we shall be denoting it by @xmath146-module . in what follows for the expressions that we obtain after appropriate conjugations we shall use the notation @xmath254 . * theorem 8)[gp ] let @xmath255 be a monomial in the @xmath61 s with gaps in the indices . then we have that : @xmath256 where @xmath257 and @xmath258 for some @xmath259 and for all @xmath217 . theorem [ gp ] is best demonstrated in the following example on a word with two gaps . note that we underline expressions which are crucial for the next step . for the 2-gap word @xmath260 we have : @xmath261 \ t_6^{-1}g_3 ^ 2g_5g_4 & = \\ & & & & & \\ & = & q^2t^{k_0}t_1^{k_1}t_2t_3 ^ 2\underline{g_4g_5t_6^{-1}}g_3 ^ 2g_5g_4 & + & q(q-1)t^{k_0}t_1^{k_1}t_2t_3t_4\underline{g_5t_6^{-1}}g_3 ^ 2g_5g_4 & + \\ & & & & & \\ & & & + & ( q-1)t^{k_0}t_1^{k_1}t_2t_3t_5\underline{g_4t_6^{-1}}g_3 ^ 2g_5g_4 & = \\ & & & & & \\ & = & q^2t^{k_0}t_1^{k_1}t_2t_3 ^ 2\underline{t_6^{-1}}g_4g_5g_3 ^ 2g_5g_4 & + & ( q-1)t^{k_0}t_1^{k_1}t_2t_3\underline{t_5}t_6^{-1}g_4g_3 ^ 2g_5g_4 & + \\ & & & & & \\ & & & + & q(q-1)t^{k_0}t_1^{k_1}t_2t_3t_4\underline{t_6^{-1}}g_5g_3 ^ 2g_5g_4 & \widehat{=}\\ \end{array}\ ] ] @xmath262 by theorem [ gp ] we have to deal with elements in @xmath60 , where the looping generators have consecutive indices but their exponents are not in decreasing order , followed by ` braiding tails ' . we show that these elements are conjugate equivalent to sums of elements in the @xmath263-module @xmath4 , namely , elements in @xmath4 followed by ` braiding tails ' . * theorem 9)[exp ] for an element in the @xmath146-module @xmath60 we have that : @xmath264 where @xmath265 and @xmath266 for all @xmath212 . consider the element @xmath267 and apply theorem [ exp ] on the first ` bad ' exponent occurring in the word , starting from right to left . in that way we obtain a word with one less ` bad ' exponent , so applying theorem [ exp ] again we obtain elements in the set @xmath4 followed by braiding tails . more precisely : @xmath268 where @xmath269 , for all @xmath217 and @xmath270 $ ] . so far we have seen how to convert elements in the basis @xmath47 to sums of elements in @xmath171 and then , using conjugation , how these elements are expressed as sums of elements in the @xmath263-module @xmath4 . we now present results on how using conjugation and stabilization moves all these elements in the @xmath263-module @xmath4 are expressed as sums of elements in the set @xmath4 with scalars in the field @xmath271 . we will use the symbol @xmath272 when a stabilization move is performed and @xmath273 when both stabilization moves and conjugation are used . more precisely , in @xcite we prove the following : ( * ? ? ? * theorem 10 ) [ tails ] for a word in the @xmath146-module , @xmath4 we have : @xmath274 such that @xmath275 , @xmath276 and @xmath277 for all @xmath212 . in this example we demonstrate how to eliminate the ` braiding tail ' in a word . @xmath278 we have that : @xmath279 and so @xmath280 [ rm4]this is a long procedure , since eliminating the tails will give rise to gaps in the indices again . recall however that when managing gaps in the indices and ordering the exponents of the @xmath61 s we also obtain ` braiding tails ' . this is a long procedure which , as shown in @xcite using complex and technical inductions , this procedure eventually terminates and only elements in @xmath4 remain . this means that @xmath4 is a generating set for @xmath3 . this procedure is abstractly demonstrated in figure [ alg ] . in the figure , we start with a word @xmath85 in the old basis of @xmath3 , @xmath47 , and applying theorems 810 we end up with a linear combination of the homologous word @xmath281 and of elements @xmath282 smaller than @xmath255 . to @xmath4.,width=499 ] with the orderings given in definition [ order ] , in @xcite we showed that the infinite matrix converting elements of the basis @xmath47 of @xmath3 to elements of the set @xmath4 is a block diagonal matrix , where each block corresponds to a subset of @xmath47 of level @xmath202 and it is an infinite lower triangular matrix with invertible elements in the diagonal . this constitutes our strategy for proving theorem [ newbasis ] . more precisely , fixing the level @xmath202 of a subset of @xmath47 , the proof of theorem [ newbasis ] is based on the following : * a monomial @xmath283 can be expressed as linear combinations of elements in @xmath284 , @xmath285 , followed by monomials in @xmath263 , with scalars in @xmath271 such that there exists @xmath286 . * applying conjugation and stabilization moves on all @xmath285 s results in elements @xmath287 in @xmath203 , such that @xmath288 for all @xmath217 . * the coefficient of @xmath175 is an invertible element in @xmath271 . * @xmath289 . * using this infinite diagonal matrix , in ( * ? ? ? * theorem 11 ) we showed that the set @xmath4 is linearly independent . hence , using the above and remark [ rm4 ] , @xmath4 forms a basis for the homflypt skein module of st . we now return to our initial aim , that is , the computation of the homflypt skein module of a lens space @xmath0 . as explained in the introduction , in order to compute @xmath291 we need to normalize the invariant @xmath2 ( recall theorem 5 ) by forcing it to satisfy all possible braid band moves ( _ bbm _ ) , recall eq . 1 . at this point the reader should recall the discussion in the introduction culminating to equation 2 . in order to simplify this system of equations , in @xcite we first show that performing a _ bbm _ on a mixed braid in @xmath41 reduces to performing _ s on elements in the canonical basis , @xmath134 , of the algebra @xmath52 and , in fact , on their first moving strand . we then reduce the equations obtained from elements in @xmath163 to equations obtained from elements in @xmath171 . in order now to reduce further the computation to elements in the basis @xmath4 of @xmath3 , we first recall that elements in @xmath171 consist in two parts : a monomial in @xmath61 s with possible gaps in the indices and unordered exponents , followed by a ` braiding tail ' in the basis of @xmath146 . so , we first manage the gaps in the indices of the looping generators of elements in @xmath171 , obtaining elements in the augmented @xmath133-module @xmath60 ( followed by ` braiding tails ' ) . note that the performance of _ bbm _ s is now considered to take place on any moving strand . we then show that the equations obtained from elements in the @xmath133-module @xmath60 by performing _ _ s on any strand are equivalent to equations obtained from elements in the @xmath133-module @xmath4 by performing _ bbm _ s on any strand ( ordering the exponenets in the @xmath61 s ) . we finally eliminate the ` braiding tails ' from elements in the @xmath133-module @xmath4 and reduce the computations to the set @xmath4 , where the _ bbm _ s are performed on any moving strand ( see @xcite ) . thus , in order to compute @xmath291 , it suffices to solve the infinite system of equations obtained by performing _ bbm _ s on any moving strand of elements in the set @xmath4 . in this section we present the above steps in more details . the procedure is similar to the one described in @xcite ( see 2 in this paper ) , but now we do it simultaneously before and after the performance of a braid band move . we now show that it suffices to perform _ bbm _ s on elements in the linear basis of @xmath52 , @xmath163 . as already mentioned , this is the first step in order to restrict the performance of _ bbm _ s only on elements in the basis @xmath4 . we first recall that by the artin combing we can write words in @xmath41 in the form @xmath84 , where @xmath85 is a monomial in the @xmath86 s and @xmath87 ( remark 1 ) . we then note the following : [ artcomb ] braid band moves are interchangeable with the artin combing . let @xmath292 . then , the proof is clear from the following diagram : @xmath293 ( * ? ? ? * lemma 2)[lbbmskein ] braid band moves and the quadratic relation ( skein relation ) are interchangeable . by lemma [ artcomb ] , a word in @xmath41 can be assumed in the form @xmath294 , where @xmath295 is a monomial in @xmath86 s and @xmath296 . seen as a monomial in @xmath146 and applying the quadratic relation , the element @xmath175 can be written as a sum : @xmath297 , where the @xmath298 s are words in @xmath146 in canonical form and the @xmath299 are expressions in @xmath271 for all @xmath217 . we perform a braid band move on the element @xmath300 and we obtain : @xmath301{(\pm)(p,1 ) } \ t^p\tau_2^{\prime}\cdot w_+\sigma_1^{\pm1},\ ] ] where @xmath302 is the same word as @xmath175 but with all indices shifted by one . hence , on the algebra level we have @xmath303 and @xmath304 . so : @xmath305{(\pm)(p,1 ) } \ t^p\tau_2^{\prime}\cdot \sum_{i=1}^{n}{f_{i}(q)w_{i_+ } } g_1^{\pm1}.\ ] ] on the other hand , for each mixed braid @xmath306 we also have : @xmath307{(\pm)(p,1)}\ t^p\tau_2^{\prime}\cdot w_{i_+ } \sigma_1^{\pm1}\ \forall\ i,\ ] ] and thus , on the algebra level we finally obtain : @xmath308 that is , the following diagram commutes : @xmath309 see also figure [ bbmskein ] . so the proof is concluded . .,width=432 ] furthermore we have : [ sprbbm ] the procedure of bringing a looping monomial in the @xmath86 s in the form of elements in the sets @xmath134 of theorem 2 is consistent with the braid band moves . let @xmath294 be an element in @xmath41 and @xmath310 the result of a performance of a _ bbm _ on @xmath294 . we now follow ( * ? ? ? * proposition 2 & theorem 1 ) so as to order the indices of the monomials @xmath295 and @xmath311 in the @xmath86 s ( before and after the performance of the _ bbm _ ) . cabling the first moving strand coming from the performance of the _ bbm _ with the fixed strand of the mixed braid , and viewing this cable as one thickened strand , we have that the procedure we follow to order the indices of the @xmath86 s is identical before and after the performance of the _ bbm _ and this concludes the proof . using lemmas [ artcomb ] , [ lbbmskein ] and [ sprbbm ] we now have the following : ( * ? ? ? * proposition 1)[bbms ] it suffices to consider the performance of braid band moves on the first strand of only elements in the set @xmath163 . by the artin combing , any @xmath312 can be written in the form @xmath84 , where @xmath85 is a monomial in @xmath86 s and @xmath313 . by lemma [ artcomb ] we have that : @xmath314 where @xmath298 are words in reduced form in @xmath315 and @xmath316 . then , we order the indices of the monomials in the @xmath86 s before and after the performance of the _ bbm_. the result follows from lemma [ sprbbm ] . we shall now show that it suffices to perform _ bbm _ s on elements in the linear bases @xmath171 of the algebras @xmath52 , which includes as a proper subset the basis @xmath4 of @xmath3 , described in [ lamb ] . indeed , let @xmath317 as above . then , by ( 8) we have : @xmath318 see top row of fig . we then perform a _ bbm _ on the first moving strand of both @xmath84 and @xmath319 ( see bottom row of fig . 18 ) and we cable the new parallel strand together with the surgery strand . denote the result of cabling the new strand appearing after the performance of the _ bbm _ with the fixed strand as @xmath320 . then : @xmath321 so : @xmath322 . but since @xmath323 for some @xmath259 , we can express @xmath324 as a sum of elements in the linear basis @xmath48 of @xmath52 , that is , @xmath325 , where @xmath326 and @xmath327 , for all @xmath217 and @xmath328 is a monomial in the @xmath329 s with possible gaps in the indices and unordered exponents . then , by theorem 5 : @xmath330 we conclude that : @xmath331 the main ponts in the above procedure is that when performing a _ bbm _ , the looping generators @xmath61 ( in some @xmath122 ) remain formally the same ( in some @xmath332 ) . this implies that after the _ bbm _ and the cabling operation , the words @xmath84 and @xmath324 remain formally the same . furthermore , when converting the words @xmath319 and @xmath333 into linear sums of elements in the bases @xmath48 and @xmath334 respectively , the coefficients @xmath335 in @xmath336 remain the same . the above are summarized in the following proposition : ( * ? ? ? * proposition 2)[picprop2 ] the equations @xmath337 result from equations of the form @xmath338 where @xmath339 and @xmath340 . . , width=211 ] as mentioned in 2 , a word in @xmath171 is a monomial in the @xmath61 s followed by a ` braiding tail ' , a monomial in the @xmath114 s . this ` braiding tail ' is a word in the algebra @xmath146 and the monomial in the @xmath61 s may have gaps in the indices . using the ordering relation given in definition [ order ] and conjugation these gaps are managed by showing that a monomial in the @xmath61 s can be expressed as a sum of monomials in the @xmath61 s with consecutive indices , which are of less order than the initial word and which are followed by ` braiding tails ' ( theorem 8) . note that the exponents of the @xmath61 s are in general not ordered , so these end monomials do not necessarily belong to the basis @xmath4 of @xmath3 . in order to restrict the _ bbm _ s only on elements in @xmath4 , we need first to augment the set @xmath4 . so , as a first step we consider the augmented set @xmath60 in @xmath3 that contains monomials in the @xmath61 s with consecutive indices but arbitrary exponents . note that , due to the presence of ` braiding tails ' , the set @xmath60 is considered as an @xmath146-module . we now proceed with showing that eqs ( [ eq2 ] ) ( proposition 2 ) reduce to equations of the same type , but with elements in the set @xmath60 . for that , we need the following lemma about the monomial @xmath341 , which serves as the basis of the induction applied for proving the main result of this section , proposition [ imp ] . * lemma 3)[gap ] the equations @xmath342 are equivalent to the equations @xmath343 indeed , in @xcite we prove that : @xmath344 which captures the main idea of the proof of lemma [ gap ] . let now @xmath345 denote a word containing gaps in the ( ordered ) indices but not starting with a gap . when managing the gaps , the first part of the word ( before the first gap ) remains intact after managing the gaps and the same carries through after the performance of a _ bbm _ on the first moving strand . that is , the following diagram commutes : @xmath346 where @xmath347 and @xmath257 , for all @xmath217 . in the case where the word @xmath347 starts with a gap , in @xcite it is shown that equations obtained from @xmath324 are equivalent to equations obtained from elements @xmath348 , where @xmath349 are monomials in the @xmath61 s _ not _ starting with a gap , _ but with the _ bbm _ performed on any strand _ ( see fig . [ exgaps ] ) . the above are summarized in the following proposition : ( * ? ? ? * proposition 3)[imp ] in order to obtain an equivalent infinite system to the one obtained from elements in @xmath171 by performing braid band moves on the first moving strand , it suffices to consider monomials in @xmath60 followed by braiding tails in @xmath146 and perform braid band moves on any moving strand . the monomials in the @xmath61 s that we obtain after managing the gaps are not elements in the set @xmath4 , since the exponents of the loop generators are not necessarily ordered . we now order the exponents of the @xmath61 s and we show that equations obtained from elements in the @xmath146-module @xmath60 reduce to equations obtained from elements in the @xmath146-module @xmath4 . the procedure we follow is similar to the one described in @xcite , but , as we mentioned earlier , in this case we do it simultaneously before and after the performance of a _ bbm_. ( * ? ? ? * proposition 4 ) equations of the infinite system obtained from elements in @xmath60 followed by braiding tails in @xmath133 are equivalent to equations obtained from elements in @xmath4 followed by braiding tails , where a braid band move can be performed on any moving strand . it follows from theorem 9 , since all steps followed so as to order the exponents in a monomial in the @xmath61 s remain the same after the performance of a _ bbm _ , ignoring the @xmath350 appearing after the _ bbm_. we now deal with the ` braiding tails ' . applying the same technique as in theorem 10 before and after the performance of a _ bbm _ , we first prove that equations obtained by performing _ s on any moving strand on elements in @xmath4 followed by words in @xmath146 , reduce to equations obtained by performing _ s on any moving strand from elements in @xmath60 ( with no ` braiding tails ' ) . * proposition 5)[tailbbm ] equations obtained from _ bbm _ s on elements in @xmath4 followed by words in @xmath146 are equivalent to equations obtained by performing a braid band move on any moving strand on elements in @xmath60 . we perform a _ bbm _ on an element @xmath351 in the @xmath146-module @xmath4 and we cable the parallel strand with the surgery strand , see figure [ tailfig ] . we then apply theorem 10 before and after the performance of the _ bbm _ and uncable the parallel strand . the proof is illustrated in figure [ tailfig ] . . , width=518 ] in this example we demonstrate proposition [ tailbbm ] . @xmath352\cdot t^3 & \underset{bbm}{\overset{1^{st } str.}{\longrightarrow } } & \left[q^2(q-1)(q^2-q+1)z^2\right]\cdot t^pt_1 ^ 3g_1^{\pm1}\\ + & & + \\ a\cdot t^2t_1 & \underset{bbm}{\overset{1^{st } str.}{\longrightarrow } } & a\cdot t^pt_1 ^ 2t_2g_1^{\pm1}\\ \end{matrix}\ ] ] where @xmath353 . in the process of eliminating the ` braiding tails ' we have so far reached _ bbm _ s in the set @xmath60 ( with no ` braiding tails ' ) . following the same procedure and applying the same techniques as in @xcite , we finally reach the set @xmath4 by means of the following : ( * ? ? ? * theorem 8) in order to obtain the _ bbm _ equations needed for computing the homflypt skein module of the lens spaces @xmath0 , it suffices to perform braid band moves on any strand on elements in the basis @xmath4 of @xmath3 . the proof is based on theorems 8 , 9 & 10 and the fact that the _ bbm _ s commute with the stabilization moves ( lemma 1 ) and the skein ( quadratic ) relation ( lemma 2 ) . the fact that the _ bbm _ s and conjugation do not commute , results in the need of performing braid band moves on all moving strands of elements in @xmath4 . and on the @xmath354 strand of an element in @xmath4.,width=240 ] in this paper we present recent results toward the computation of the homflypt skein module of the lens spaces @xmath0 , @xmath290 , via braids . we first presented a new basis @xmath4 of @xmath3 in braid form , which is crucial for the braid approach to @xmath290 and we then related @xmath290 to @xmath3 by means of equations resulting from _ bbm _ s . in particular , we showed that by considering elements in the basis @xmath4 and imposing on values on them of the lambropoulou invariant @xmath2 for knots and links in st relations coming from the performance of _ bbm _ s on all their moving strands , we arrive at an infinite system of equations , the solution of which coincides with the computation of @xmath290 . our results are summarized to the following : , skein module of links in a handlebody , _ topology 90 , proc . of the research semester in low dimensional topology at osu _ , editors : b.apanasov , w.d.neumann , a.w.reid , l.siebenmann , de gruyter verlag ( 1992 ) , 315 - 342 .

at wide separations , planetary - mass and brown dwarf companions to solar type stars occupy a curious region of parameters space not obviously linked to binary star formation or solar - system scale planet formation . these companions provide insight into the extreme case of companion formation ( either binary or planetary ) , and due to their relative ease of observation when compared to close companions , they offer a useful template for our expectations of more typical planets . we present the results from an adaptive optics imaging survey for wide ( - 500au ) companions to solar type stars in upper scorpius . we report one new discovery of am companion around gsc 06214 - 00210 , and confirm that the candidate planetary mass companion 1rxs j160929.1 - 210524 detected by is in fact co - moving with its primary star . in our survey , these two detections correspond to% of solar type stars having companions in the 6 - 20 mass and - 500au separation range . this figure is higher than would be expected if brown dwarfs and planetary mass companions were drawn from an extrapolation of the binary mass function . finally , we discuss implications for the formation of these objects . [ firstpage ]

1011.2201

over the past five years , direct imaging surveys for extrasolar planets have discovered a small but significant number of ultra - low mass companions ( henceforth ulmcs , masses @xmath320@xmath4 ) at @xmath550 au separations from their primaries . these objects have estimated masses that are in the same range as radial velocity or transiting planets with @xmath65au separations , which have a continuous mass distribution up to @xmath020@xmath4 , then a gap in mass until arguably star - like objects are found at @xmath760m@xmath8 @xcite . for this reason , ulmcs are often called `` planetary mass '' companions . the prototypical wide ulmc , 2m1207 - 3933 , consists of a 48 @xmath9 companion located @xmath050 au away from a 10 myr old brown dwarf @xcite . since its discovery , half a dozen other ulmcs have also been reported , most of which orbit much higher - mass primaries ( @xmath00.52.0 @xmath10;e.g . @xcite ; @xcite ; @xcite ) . some of these systems appear to be genuinely scaled - up versions of our own solar system , with planets that are consistent with formation in a disk . one example is the companion to fomalhaut , which is coplanar with its debris disk . another is the hr 8799 , which has multiple planetary companions of similar mass . other cases like chxr 73 and 1rxs j160929.1 - 210524 are more ambiguous since their orbital radii are even wider , and it is unclear whether they lie in the original plane of planet formation . such companions could very well form like planets within a circumstellar disk or like binaries from the collapse of a molecular cloud . these ulmcs pose a significant challenge to existing models of planet and binary formation . their orbital radii are so large that the core accretion timescale ( @xmath7100 myr at 100 au ; pollack et al . 1996 ) should be much longer than the typical protoplanetary disk dissipation timescale ( @xmath03 - 5 myr ; @xcite ; @xcite ; @xcite ) . some of the closer companions could potentially form on the gravitational instability timescale ( e.g. * ? ? ? * ) , which can be very short at intermediate radii ( 10100au ) , but gravitational instability in disks has not been extensively modeled at @xmath5100 au , much less at @xmath0300 au . these companions could represent the extreme end of the binary mass function , which appears to be linearly - flat ( with all companion masses being equally probable , @xcite ) well into the substellar regime . however , it is unclear whether this trend could extend to planetary masses ( by which we mean @xmath620m@xmath1 in this paper ) . these companions have masses near the opacity - limited minimum mass @xcite , and unless their formation occurred exactly as the circumstellar envelope was exhausted , then they should have quickly accreted enough mass to become high - mass brown dwarfs or stars . in this paper , we report the discovery of two ulmcs in the upper scorpius ob association , one of which was independently discovered by @xcite . the survey consists of a subset of the aperture - masking interferometry sample reported by @xcite . we describe our observations and data analysis techniques in section [ sectobs ] , and in section [ sectfollowup ] , we report the detections and detection limits from our survey . in section [ sectfreq ] , we report the first measurement of a frequency for ulmcs around young stars . finally , in section [ sectdiscuss ] , we discuss the implications for possible formation mechanisms for ulmcs . nearby ( @xmath3200pc ) young ( @xmath320myr ) stars have been the object of numerous high - resolution imaging campaigns over the past several decades . these observations have included lunar occultation ( e.g. @xcite ) , speckle interferometry ( e.g. @xcite ) , adaptive optics ( ao ) imaging ( e.g. @xcite ) , and most recently non - redundant masking ( nrm ) inteferometry @xcite . the earlier techniques of lunar occultaion and speckle were only sensitive to the presence of bright , stellar - mass , companions , and only recently did ao imaging and nrm interferometry managed to probe within the brown dwarf regime , going as far as sampling the top of the planetary mass regime . in k08 , we reported the results of one such survey of young stars in the upper sco ob association that used a combination of conventional ao imaging and nrm - interferometry . in addition to the results for stellar and brown dwarf companions , that paper also reported detection limits for ulmcs at small separations ( within 50 au ) where the probability of background star contamination was negligible . in this work , we report the corresponding analysis for candidate companions at wide separations , including multi - epoch follow - up imaging for three candidates , of which two appear to be associated and of approximately planetary mass . table 4 of k08 lists the ao imaging observations conducted with the pharo camera at the palomar 200 " telescope and the nirc2 camera at the keck - ii 10 m telescope . we found that 10 out of these 62 targets had stellar binary companions at separations of @xmath00.255.0 , which should mean that the majority of additional faint companions are not dynamically stable in this range of separation ; these were omitted from our sample . in addition , we omitted the few targets with spectral types of later than m2 ( @xmath11 ) so as to work with a single mass range of approximately `` solar - type '' stars , leaving 49 targets in consideration . all observations used the smallest pixel scales ( 10 mas / pix with nirc2 and 25 mas / pix with pharo ) in order to achieve the best psf sampling . we used a @xmath12 filter at palomar and the @xmath13 filter at keck , which yielded diffraction - limited resolutions of 100mas and 50mas , respectively . the keck observations used the narrowband filter despite the penalty in sensitivity , because the brighter primary stars in our sample would have saturated the detector within the minimum exposure time . much of this sensitivity was regained by using more fowler samples per frame , which significantly reduces the high read noise of nirc2 s detector . these observations used relatively short total integration time , of the order of a minute , in comparison with other ao imaging surveys ( e.g. @xcite ) , resulting in limiting magnitudes of @xmath141517 for the companions . given the young age of these upper sco targets ( @xmath05 myr ) , even these shallow observations were able to reach the planetary mass regime ( at 2 arcsec separations , @xmath314 @xmath9 for 48 targets and @xmath37 @xmath9 for 42 targets ) . the detections and detection limits were derived using methods we previously described in @xcite and kraus & hillenbrand ( 2010 , in prep ) . since the detection limits at more than a few @xmath15 are driven by speckle noise , detections and detection limits were determined from the coadded image stacks by placing a large number of photometric apertures ( with diameter @xmath16@xmath17 ) around each target , then measuring the mean and standard deviation of the brightness distribution for all apertures in concentric rings around the primary . for each ring , we identified all candidate detections with significance @xmath55@xmath18 above the mean brightness , then compared those detections to other stars taken on the same night to identify and reject the quasistatic speckles and diffraction spikes that are seen in common for many targets . all candidate detections that could not be identified as psf artifacts were then adopted as genuine sources and hence as candidate bound companions , while the 5@xmath18 limits were adopted as our formal detection limits . at large separations , the detection limit from this algorithm converges to the sky background limit , and at these separation ( @xmath72 arcsec ) , we used a 10@xmath18 limit , because of the large number of pixels in this regime . we found no candidate companions near the detection limits in the speckle - limited regime , but there are many candidate companions in the sky - limited regime . we determined the photometry and astrometry for these sources using the methods described in k08 and @xcite . to briefly summarize , we measured astrometry and aperture photometry for each source with respect to the known usco member using the iraf task daophot @xcite ; all measurements were conducted using apertures of 0.5 , 1.0 , and 2.0 @xmath15 , and then the optimal aperture was chosen to maximize the significance of the detection ( given the competing uncertainties from the sky background and the poisson noise for the source itself ) . in order to estimate the uncertainties from the data , we analyzed the measurements in individual frames and then combined those measurements to estimate the mean and standard deviation . we then accounted for the systematic uncertainty in the plate scales and distortion solutions of pharo ( 0.3% ; section 2.3 ) and nirc2 ( 0.05% ; ghez et al . 2008 , @xcite ) by adding those terms in quadrature with the observed scatter . most of the new candidate companions were identified with pharo ; as we describe below , its astrometric calibration is not yet well - understood and might be further improved with more calibration observations , but our systematic uncertainties should account for this effect . many of the wider ( @xmath55 ) companions are also affected by anisoplanatism , yielding systematically low brightness estimates for aperture photometry . a correction of the photometry would require a detailed knowledge of the atmospheric turbulence profile and isoplanatic patch size , so it can not be accomplished for our data . however , as we describe below , all of these sources have well - calibrated @xmath19 magnitudes available from ukidss . we choose instead to defer to those measurements in determining colors ( section 3.2 ) . finally , all candidate companions with separations @xmath54.2 have spatially resolved counterparts in the ukidss galactic cluster survey @xcite , and many of the widest companions also have optical counterparts in the usno - b1.0 digitization of the palomar observatory sky survey @xcite . as we describe further in section 3.2 , we have used those observations to identify most of these candidate companions as background stars . the rest require multi - epoch astrometric monitoring to determine whether they are associated . around three targets in the k08 survey we found faint visual candidate companions with projected separations of 23 and brightnesses of @xmath201516 : gsc 06214 - 00210 , 1rxs j160929.1 - 210524 , and 1rxs j160703.4 - 203634 , whose basic properties are listed in table [ tabprops ] . we re - observed these objects with the telescope - instrument - filter combinations as detailed in table 2 . the expected surface density of unassociated background stars at these brightnesses is relatively low , but since upper scorpius is at moderate galactic latitude and is projected over the background milky way bulge , this probability was not sufficiently low that we could assume they were bound ulmc companions . follow - up observations were therefore required to confirm common proper - motion . the density of background stars at @xmath21 is approximately 0.002 per square arcsecond ( kraus & hillenbrand 2010 , in prep . ) , which gives a 86% chance of a chance alignment within 2.5for at least one star in our sample , but only a 30% chance of all 3 candidates being chance alignments . it is impossible to use proper motions to rule out chance alignments with other association members because the internal velocity dispersion of upper sco is only @xmath01kms@xmath22 ( @xmath01.5masyr@xmath22 , @xcite ) , meaning that they would be comoving even if they were only seen in chance alignment . however , the surface density of young stars in upper sco is no more than 100deg@xmath23 @xcite or @xmath24arcsec@xmath23 , meaning that the probability of a chance alignment is negligible . the proper motion of the upper scorpius association is ( -11.5,-23.5)masyr@xmath22 in equatorial coordinates , determined from the members listed in @xcite , and using updated proper motions from @xcite . this is relatively small when compared to nearby moving groups , so an astrometric accuracy of order 2mas or 0.1% at 2is required in order to clearly determine whether a companion is co - moving or not on a 1 year time baseline . the nirc2 camera has been shown to have a stability better than this in the precise galactic center astrometric work of @xcite . @xcite showed that the palmao adaptive optics system @xcite and the pharo near infrared ( nir ) camera are stable in distortion to @xmath0100@xmath25as over several months . however , palmao underwent several upgrades between these common proper motion confirmation observations , including preliminary work for a reconfiguration of the output beam path to accommodate new science instruments . we searched for palmao distortion solution changes during this period using observations of the core of the m5 globular cluster from four nights ( 28/5/2007 ; 29/5/2007 ; 17/07/2008 and 17/08/2008 ) . the observations consisted of 100 co - added 1.4 second exposures at each epoch , were taken in the 25 mas pharo plate scale , used the br-@xmath26 narrow - band filter to avoid differential chromatic refraction , and were timed to be observed at very similar airmasses and hour angles . the same ao guide star was used in each case , and care was taken to align each epoch to the same pointing within an arcsecond and to keep all other ao and camera parameters consistent . we fitted 2d gaussians to derive the positions of 34 stars covering the 25``@xmath2725 '' field in each dataset ; we also checked the fitted positions using sextractor @xcite and found very similar results . we first used a simple model to match the stellar positions between epochs , allowing for a change in pointing position , an arbitrary rotation , and a separate scaling in the x and y axes . there was no detectable change in rotation or scale within 2007 or within 2008 , but between those years the field rotation changed by 0.18 degrees , while the x and y plate scales changed by -0.5 % and + 0.2% respectively . however , the simple rotation and plate scale change model leaves 3 - 5mas residuals in matching the 2007 and 2008 measured positions . the residuals are reduced to @xmath01mas using a general 4-parameter linear transformation , where the additional parameter represents the plate scale changes being in arbitrary orthogonal axes , or alternatively the addition of a shear term . however , some statistically significant higher - order distortion is still apparent after application of this solution . it appears that long - term precision astrometry using adaptive optics requires careful attention to changes in the detailed distortion solutions ( beyond simple rotation and scale ) ; this is particularly important as , like palmao , the optics of many adaptive optics systems are regularly upgraded . therefore , wherever at least two nirc2 epochs are available , we only consider the astrometry with nirc2 for our proper motion measurements . all followup nirc2 data were acquired using the narrow camera with a @xmath010mas / pixel image scale and the @xmath28 filter , which we found did not quite saturate for these targets using the shortest exposure time with a minimum number of fowler samples . the follow - up nirc2 data were reduced using a custom pipeline written in idl . after standard image - processing tasks ( background subtraction , flat - fielding and bad pixel removal ) were completed , the distortion in the image plane was then removed by the program nirc2dewarp available from the nirc2 camera home page , using the updated distortion solution of b.cameron ( personal communication , 2007 ) . finally , aperture - photometry and centroids were computed with the imexam function of atv , version 2.0b4 . selected data sets were also analysed with the daophot function of iraf , giving results consistent well within the error bars . for centroiding , a centering box size of 5 pixels was used , with the exception of one single observation in l - band , where a box size of 9 pixels was used . the aperture - radius for photometry was 5 pixels for the kp filter , but 7 pixels for the j and h filters due to low strehls and dispersion , and 10 pixels in l - band due to the larger diffraction - limited core . the sky annulus was always set at 10 to 20 pixels . note that we did not attempt to calibrate absolute photometry , and have only computed the relative photometry between primary and secondary . the following subsections describe the multi - epoch astrometry obtained on each of the three targets in table [ tabprops ] , in order to confirm proper motion . two clearly associated objects are discussed , as well as two objects identified as background stars . lccccrr gsc 06214 - 00210 & 16 21 54.67&-20 43 09.1&m1&0.60&11.6&9.15 + 1rxs j160929.1 - 210524&16 09 30.30&-21 04 58.9&k7/m0&0.68 - 0.77&12.1&8.92 + 1rxs j160703.4 - 203634&16 07 03.56&-20 36 26.5&m0+?&0.68 + 0.59&11.3&8.10 + [ tabprops ] gsc 06214 - 00210 is a young m1 star near the eastern edge of upper sco . it was originally identified as a candidate young star by preibisch et al . ( 1998 ) based on its x - ray emission , then confirmed to have strong lithium absorption ( @xmath29 ) and h@xmath30 emission ( @xmath31 ) as compared to stars of equivalent spectral type in several young clusters . it was also found to have a proper motion @xmath32 mas yr@xmath22 @xcite , consistent with the motion of upper sco . given its m1 spectral type , the 5 myr mass - temperature relations of @xcite predict a mass of 0.60@xmath33 . note that this mass must be taken with caution , as an uncertainty of 1 subclass in spectral type should correspond to an uncertainty of @xmath00.1 @xmath33 in mass . one should keep in mind that the models themselves carry an unknown uncertainty since the mass - luminosity and mass - temperature relations of young stars are almost completely uncalibrated for @xmath31 @xmath33 ( e.g. @xcite ) . an example image of gsc 06214 - 00210b used for astrometry is shown in figure [ figimage ] . our photometric and astrometric observations of this companion are given in table [ tabp98 - 70 ] and figure [ figp70 ] . the candidate companion showed a relative motion of 7@xmath344 mas with respect to gsc 06214 - 00210 over 2.9 years . this is much less than the @xmath070 mas motion expected if gsc 06214 - 00210b were a background star , and demonstrates that it is most likely physically associated with gsc 06214 - 00210 . the apparent relative proper motion of 2.5@xmath341.3masyr@xmath22 corresponds to 1.7@xmath340.9 kms@xmath22 assuming a 7 mas parallax for upper scorpius , which is roughly equal to the circular orbital velocity of 1.7kms@xmath22 expected for a @xmath0300 au orbit . as can be seen from the larger contrast in bluer filters ( cf . tabp98 - 70 ] ) , gsc 06214 - 00210b is quite red compared to its primary star . the primary has an observed 2mass color of @xmath35 , and most m1 stars have a typical @xmath36 color of @xmath00.15@xmath340.05 @xcite , so the inferred colors for the companion are @xmath37 and @xmath38 . the mean distance for upper sco is 145 pc ( de zeeuw et al . 1999 ) , and a @xmath39 degree spread on the sky of the densest region ( where all our candidates belong ) corresponds to a @xmath014pc error on the distance of any individual star . this gives a a distance modulus of @xmath40 , so given the 2mass magnitudes of the primary , the absolute magnitudes of the companion are @xmath41 , @xmath42 , @xmath43 , and @xmath44 . the interstellar extinction toward upper sco is negligible ( @xmath45 or @xmath46 ) , so no extinction corrections should be required . the observed @xmath47 color of 1.3 is consistent with the m8-l4 spectral type range of field dwarfs from @xcite . the k - l color of 1.05 , however , would clearly place the star at the red end of this range , at l3-l4 . the absolute k - magnitude of 9.1 is then about 2 magnitudes brighter than for corresponding field objects , providing clear evidence of a larger radius and young age ; this height above the main sequence is similar to that observed for other late - type members of upper sco @xcite and tw hya @xcite . the colors and spectral type of gsc 06214 - 00210b are not expected to match field dwarfs due to the low surface gravity . @xcite find that j - k and k - l colors are significantly redder for low gravity objects at a given spectral type . although gsc 06214 - 00210b is a low - mass object , @xcite focusses on redder object at j - k colors of @xmath02.0 , and it is nt clear that their result should hold for bluer objects like gsc 06214 - 00210b , with j - k=1.3 . in figure [ figcolorcolor ] we show the nir the colors of gsc 06214 - 00210b with respect to field dwarfs and giants ( which bracket gsc 06214 - 00210b in gravity ) , where it appears to have a small but significant k - l excess . the red k - l color could be due to a disk , as @xmath51/3 of young brown dwarfs retain a disk for @xmath55 myr ( e.g. @xcite ) , with a trend of increasing disk lifetime with decreasing mass . spatially resolved spectroscopy of the system would be required in order to determine the spectral type more accurately , and to search for signs of accretion onto the secondary . given the observed brightness in the @xmath48 filters , a comparison to the 5 myr dusty models @xcite et al . 2000 ) suggests that the mass of the companion is @xmath01215 @xmath9 , with bluer filters suggesting slightly lower masses than redder filters . the cond models @xcite never predict sufficiently red @xmath47 colors to match our observations , and indeed are not appropriate for objects with @xmath49k . this trend qualitatively matches the observed trend for young low - mass objects to be redder than older field counterparts of similar spectral type . if the comparison was based only on observed j - k color , then the dusty models would predict masses of 1012 @xmath9 . as we discussed above , the @xmath50 photometry could have an excess from a circumstellar disk , so we suggest that the @xmath50 magnitude should not be used directly in the mass estimate . llrrr + 2454258.0 & kp & 2203.3@xmath341.5 & 176.04@xmath340.06 & 5.74@xmath340.05 + 2454634.8 & kp & 2204.7@xmath340.9 & 175.99@xmath340.03 & 5.78@xmath340.03 + 2454634.8 & j & 2205.2@xmath340.9 & 176.00@xmath340.09 & 6.30@xmath340.03 + 2454982.9 & kp & 2204.1@xmath340.9 & 175.91@xmath340.03 & 5.88@xmath340.03 + 2455313.1 & kp & 2205.6@xmath341.1 & 175.93@xmath340.03 & 5.73@xmath340.03 + 2455313.1 & h & 2202.8@xmath342.2 & 175.91@xmath340.04 & 6.21@xmath340.03 + 2455313.1 & l & - & - & 4.75@xmath340.05 + + 2454634.8 & kp & 2210.1@xmath341.0 & 27.62@xmath340.04 & 7.27@xmath340.02 + 2454982.9 & kp & 2211.3@xmath340.9 & 27.61@xmath340.05 & 7.23@xmath340.03 + + 2454250.8 & ks & 4261@xmath3414 & 219.5@xmath340.2 & 8.63@xmath340.04 + 2454982.9 & kp & 4215@xmath345 & 220.03@xmath340.08 & 8.6@xmath340.1 + + 2454634.8 & kp & 2143.2@xmath341.5 & 234.01@xmath340.07 & 8.20@xmath340.08 + 2454634.8 & j & 2138.2@xmath341.0 & 234.05@xmath340.07 & 8.06@xmath340.05 + 2454982.9 & kp & 2123.1@xmath340.5 & 234.45@xmath340.01 & - + [ tabp98 - 70 ] 1rxs j160929.1 - 210524 is a young m0 star located near the center of upper sco . like gsc 06214 - 00210 , it was first identified as a likely upper sco member by preibisch et al . ( 1998 ) and exhibits both strong lithium absorption ( @xmath51 ) and h@xmath30 emission ( @xmath52 ) . the proper motion reported by ucac3 @xcite for this object is ( -11.2,-21.9 ) @xmath341.5 mas yr@xmath22 , which is also consistent with the value for upper sco . the spectral type reported by preibisch et al . ( 1998 ) was m0 , but , as was described by @xcite , newer measurements by d. c. nguyen suggest a more likely spectral type of k7 . the inferred mass would be 0.68 - 0.77 @xmath33 for the two estimates , with uncertainties similar to that for gsc 06214 - 00210 . our observations for 1rxs j160929.1 - 210524b are listed in table [ tabp98 - 70 ] and plotted in figure [ figp50 ] . the companion was first detected in palomar images as early as may 2007 , but as the astrometric performance of the pharo camera was unverified on large timescales ( cf . section [ sectfollowup ] ) , we were not able to convincingly demonstrate common proper motion until 2009 . this object was independently detected by @xcite and further characterized by @xcite . our astrometric results are consistent with those recently reported by @xcite at 2-@xmath18 in separation and at 1-@xmath18 in position angle , but have higher precision due to the accurate astrometric characterization of nirc2 . based on a spectrum indicating a late spectral type and low gravity , these authors argued that this faint companion was a young object and therefore likely a physical companion . since that time , we have continued monitoring this object ( cf . table [ tabp98 - 70 ] , and can now confirm that 1rxs j160929.1 - 210524b is a physical companion of 1rxs j160929.1 - 210524 . the apparent proper motion of 1.2@xmath341.3masyr@xmath22 , corresponding to 0.8@xmath340.9kms@xmath22 . this is consistent with the @xmath01.7kms@xmath22 orbital motion expected , especially considering possible projection effects . since @xcite already reported nir colors , we only took @xmath19-band observations . these measurements are consistent with theirs , and indicate an absolute magnitude of @xmath53 . the predicted mass from the dusty and cond models is @xmath08 @xmath9 from the @xmath19 band photometry or @xmath07 @xmath9 from @xmath54 band photometry , though again , these estimates are completely uncalibrated by observations . given the assumptions listed above for the distance of upper sco , the projected separation is @xmath0320 au , similar to the value for gsc 06214 - 00210 system . unassociated background stars typically are identified based on multi - epoch astrometric monitoring ( which indicates that they are not comoving ) or multi - wavelength observations to measure colors ( which indicate that they do not fall along the same color - magnitude sequence ) . as we describe in the next several subsections , several of the closer companions will require astrometric monitoring with high - resolution imaging to confirm or disprove their association , but the wider companions can be identified and rejected based on archival photometry from seeing - limited all - sky surveys . in addition to the companion listed as companion b in table [ tabp98 - 70 ] , an additional wider companion candidate , listed as companion c in table [ tabp98 - 70 ] was found around 1rxs j160929.1 - 210524 in both the 2007 palomar images ( see section [ sectwider ] ) and the 2009 keck images . this companion candidate was also reported by @xcite . although the palomar astrometry had significant errors , the time baseline of 2 years was sufficient to clearly show that this object was a background star , as shown in figure [ figp50c ] . 1rxs j160703.4 - 203634 is a young , close binary system which is also located near the center of upper sco . it was originally identified in a survey for new sco - cen members by @xcite , using methods very similar to the survey by preibisch et al . ( 1998 ) : targets were identified as potential x - ray emitters , then confirmed to have strong lithium for their age . however , the survey by kunkel was never published in the refereed literature , so the equivalent widths of relevant spectral lines are not available . we discovered its multiplicity during our direct imaging and aperture masking survey ( k08 ) , finding that it had a companion with projected separation of 184 mas ( @xmath027 au ) and flux ratio @xmath55 ( mass ratio @xmath56 ) . the same observations also revealed a faint candidate companion at a projected separation of @xmath02.1 . as we summarize in table [ tabp98 - 70 ] and figure [ figk99 ] , multi - epoch astrometry for the candidate companion shows that it is not comoving with 1rxs j160703.4 - 203634 , but instead appears to be nearly stationary , as would be expected for a background star . the relative motion of the primary with respect to the companion is ( 7@xmath342,-26@xmath342)masyr@xmath22 , consistent with the proper motion of upper scorpius . the contrast with respect to the brighter member of the close binary pair is @xmath57 and @xmath58 , indicating that the companion has a color of @xmath59 . given that the total extinction along this line of sight is only @xmath601.2 or @xmath610.2 ( schlegel et al . 1998 ) , the companion must be intrinsically cool ( @xmath620.7 or spectral type @xmath5k5 ; kraus & hillenbrand 2007 ) . combined with its nonmotion ( which indicates it is likely to be quite distant ) , we therefore conclude that the candidate companion is likely a background k or early m giant , perhaps located in the milky way bulge . multi - epoch observing campaigns can be observationally expensive , so where possible , it is best to use archival data to rule out possible companions . this is sometimes impossible since many candidate companions of interest can only be distinguished from their candidate primary using high - resolution imaging techniques . however , wider companions can often be resolved in seeing - limited data , especially for new surveys that have very good spatial resolution ( i.e. , @xmath30.5 for ukidss images ) . of the 25 companions with separations of @xmath53.5 , 21 were detected in both the @xmath63 and @xmath19 filters by ukidss , so we can use their @xmath64 colors to determine whether they might be associated . ukidss observations of known low - mass members of upper sco by lodieu et al . ( 2007 ) show that most members fainter than @xmath2013 ( i.e. with spectral type @xmath5m7 ) have colors of @xmath65 ; this limit is consistent with the typical @xmath64 colors of @xmath5m7 field dwarfs as compiled in kraus & hillenbrand ( 2007 ) . as we show in table 3 , none of the candidate companions with both @xmath63 and @xmath19 magnitudes meet this criterion , so we identify all of them to be unassociated field stars , most likely in the distant background behind upper sco . in addition , 17 of the wider candidate companions have counterparts visible in the usno - b1.0 digitization of the palomar observatory sky survey @xcite . most are not present in the usno - b1.0 source catalog since its source identification algorithm was extremely conservative in identifying faint neighbors to bright stars . however , these sources can be manually identified by visual inspection . in table 3 , we list the bluest plate ( b or r ) at which each of the companions was visible . as we noted above , any true companions should have spectral types of @xmath5m7 , so our compilation of field dwarf colors ( kraus & hillenbrand 2007 ; kraus & hillenbrand , in prep ) suggests that the expected colors for true companions are @xmath66 and @xmath67 . the detection limits of the poss survey were @xmath6821 and @xmath6920 , so all 17 sources with counterparts in the b or r plates must be background stars that are bluer than these limits . in all cases , these identifications agree with the ukidss identification . three wider companions in table [ tblwider ] can not be eliminated as upper scorpius members due to insufficient photometry or astrometry . these are the companions to gsc 06793 - 00994 , gsc 06794 - 00156 and scopms 015 . the closest of these systems is scopms 015 , with a companion just outside of 500au . this small incompleteness defines the outer limit of our survey until additional follow - up observations can be obtained , though some azimuthal uncertainty remains at separations of @xmath6500 au due to image boundaries ( section [ sectdectlimits ] ) . lrrrcrrr rxj1603.6 - 2245&10959@xmath3434&206@xmath340.2&7.52@xmath340.03&r2&15.339@xmath340.011&15.074@xmath340.013&0.265@xmath340.017 + rxj1603.9 - 2031a&9484@xmath3429&280@xmath340.2&7.49@xmath340.04&b2&15.668@xmath340.012&15.547@xmath340.018&0.121@xmath340.022 + rxj1606.2 - 2036&6820@xmath3422&62.6@xmath340.2&7.07@xmath340.07& .. &16.449@xmath340.022&16.306@xmath340.034&0.143@xmath340.040 + rxj1607.0 - 2036&11939@xmath3436&47.7@xmath340.2&6.50@xmath340.03&b2&15.573@xmath340.010&15.433@xmath340.015&0.140@xmath340.018 + usco-160517.9 - 202420&7638@xmath344&153.998@xmath340.013&6.23@xmath340.02&b2&15.299@xmath340.009&15.173@xmath340.013&0.126@xmath340.016 + usco-160801.4 - 202741&7175@xmath3424&323.8@xmath340.2&6.88@xmath340.11& .. &16.344@xmath340.029&16.145@xmath340.039&0.199@xmath340.049 + usco-160801.4 - 202741&10666@xmath3432&69.1@xmath340.2&6.67@xmath340.04&b2&16.053@xmath340.022&15.959@xmath340.033&0.094@xmath340.040 + usco-160900.7 - 190852&6850@xmath3427&274.2@xmath340.2&7.06@xmath340.05&b2&15.571@xmath340.012&15.406@xmath340.017&0.165@xmath340.021 + gsc 06205 - 00954&15398@xmath3447&183.7@xmath340.2&5.03@xmath340.02&b2&13.887@xmath340.003&13.723@xmath340.004&0.164@xmath340.005 + gsc 06209 - 01501&8994@xmath3427&216.6@xmath340.2&8.17@xmath340.04&b2&16.513@xmath340.033&16.51@xmath340.055&0.003@xmath340.064 + gsc 06213 - 01358&4261@xmath3414&219.5@xmath340.2&8.63@xmath340.04 & .. & .. & .. & .. + gsc 06793 - 00797&11086@xmath3433&116.4@xmath340.2&6.46@xmath340.02&b2&14.426@xmath340.005&14.339@xmath340.008&0.087@xmath340.009 + gsc 06793 - 00797&12330@xmath3438&223.3@xmath340.2&4.46@xmath340.03&b2&12.265@xmath340.001&12.121@xmath340.001&0.144@xmath340.001 + gsc 06793 - 00994&5462@xmath3417&357.4@xmath340.2&7.79@xmath340.04& .. & .. &15.400@xmath340.018 & .. + gsc 06794 - 00480&11820@xmath3436&313.0@xmath340.2&6.41@xmath340.03&b2&15.072@xmath340.008&14.726@xmath340.010&0.346@xmath340.013 + gsc 06214 - 00210&12941@xmath3439&30.4@xmath340.2&5.88@xmath340.02&b2&14.507@xmath340.006&14.415@xmath340.008&0.092@xmath340.010 + gsc 06794 - 00537&15942@xmath3449&81.5@xmath340.2&8.29@xmath340.06& .. &16.310@xmath340.023&15.951@xmath340.030&0.359@xmath340.038 + gsc 06794 - 00537&5214@xmath3416&73.6@xmath340.2&7.99@xmath340.04& .. &15.548@xmath340.012&15.433@xmath340.019&0.115@xmath340.022 + gsc 06794 - 00156&5973@xmath3418&338.7@xmath340.2&9.43@xmath340.04& .. & .. &15.029@xmath340.013 & .. + scopms 015&16954@xmath3451&180.8@xmath340.2&8.02@xmath340.07&b2&16.664@xmath340.033&16.491@xmath340.044&0.173@xmath340.055 + scopms 015&3538@xmath3411&94.9@xmath340.2&7.19@xmath340.02 & .. & .. & .. & .. + scopms 015&7485@xmath3423&25.7@xmath340.2&7.20@xmath340.03&b2&16.085@xmath340.020&16.054@xmath340.030&0.031@xmath340.036 + scopms 015&12482@xmath3438&47.0@xmath340.2&7.06@xmath340.04&b2&16.773@xmath340.036&16.618@xmath340.049&0.155@xmath340.061 + scopms 045&7063@xmath3421&190.3@xmath340.2&8.44@xmath340.04&b2&16.408@xmath340.029&16.134@xmath340.029&0.274@xmath340.041 + scopms 045&11286@xmath3434&173.5@xmath340.2&7.76@xmath340.03&b2&16.216@xmath340.024&15.970@xmath340.025&0.246@xmath340.035 + [ tblwider ] in order to infer the properties of the distribution of wide ulmcs , it is essential not only to establish the existance of a small number of physical companions , but also to determine the magnitude limit as a function of separation for possible companions not confidently identified in the observations . this has not been general practice in previously reported wide ulmcs , in particular not for several of the @xmath340m@xmath8 companions in our separation and primary mass range : gq lup b , ct cha b and 1rxs j160929.1 - 210524 b. in this section , we will describe the detection limits for all stars in our sample . as we described in section [ sectdisco ] , we measured our detection limits in a method that accounts for both spurious detections from speckle noise ( at small separations ) and the sky background limit ( at large separations ) . these limits could in principle be used as input for monte carlo or bayesian techniques to study the underlying population , though as we discuss in the section [ sectfreq ] , using the results from our survey alone could yield a biased measurement . table [ tablimits ] lists and figure [ figlimits ] shows the companion detection limits for each star in the sample . the last column of table [ tablimits ] also gives the maximum separation in arcseconds where we surveyed all position angles for companions . the few small ( @xmath33 ) values of @xmath70 in this column are due to quick image sets taken in the camera sub - array mode we used for aperture - masking interferometry . the limits at separations smaller than 1 are set by the separation - dependent speckle noise and extended psf halo of the primary star , while the wider limits at separation greater than 1.5are constant and result from the sky background ( for broadband @xmath19 observations ) or read noise ( for narrowband @xmath13 observations ) . for each star , we list the primary @xmath19 magnitude ( from 2mass , with the flux from any close binary companions subtracted ) , the detection limit in @xmath71 at a range of angular separation , and the corresponding detection limit in @xmath9 at the corresponding projected orbital distances . the mass detection limits were derived from the k magnitudes of the 5 myr dusty isochrone @xcite . lrrrrrrrcccccccc gsc 06205 - 00954&12.9&13.4&13.4&14.1&15.0&16.4&17.8&42 & 28 & 28 & 21 & 14&6.9&4.5&10.4 + gsc 06208 - 00834&12.6&13.5&13.6&14.0&14.8&16.4&17.5&57 & 27 & 26 & 22 & 16&6.8&4.7&10.7 + gsc 06209 - 01501&12.3&13.3&13.1&13.7&14.8&16.1&17.7&74 & 30 & 33 & 25 & 16&7.6&4.5&10.4 + gsc 06213 - 00194&12.1&12.7&13.3&13.6&14.0&15.7&17.1&78 & 52 & 30 & 26 & 22&11&5.4&10.2 + gsc 06213 - 01358&12.9&13.6&13.8&14.2&15.2&16.5&17.9&39 & 26 & 23 & 20 & 14&6.5&4.4&10.4 + gsc 06214 - 00210&12.5&13.3&13.8&14.6&14.8&16.0&17.4&62 & 29 & 24 & 17 & 16&8.4&4.8&9.4 + gsc 06214 - 02384&12.3&12.9&13.5&13.9&14.4&15.8&17.5&76 & 41 & 27 & 23 & 18&9.9&4.8&10.2 + gsc 06764 - 01305&12.4&13.1&13.6&14.1&14.8&16.2&16.9&71 & 33 & 27 & 21 & 16&7.4&5.7&11.1 + gsc 06793 - 00797&11.7&12.9&13.2&13.8&14.4&16.0&17.3&97 & 42 & 31 & 24 & 19&8.3&5.0&10.2 + gsc 06793 - 00994&12.1&12.8&13.4&13.5&14.1&15.7&17.2&78 & 43 & 29 & 28 & 21&10&5.2&10.2 + gsc 06794 - 00156 & 9.8&10.7&11.4&12.1&12.9&14.6&17.2&413&212&130&78 & 41&17&5.1&10.2 + gsc 06794 - 00480&12.0&12.7&13.1&13.8&13.9&16.0&17.0&81 & 53 & 33 & 24 & 23&8&5.5&10.2 + gsc 06794 - 00537&11.9&12.7&13.1&13.6&14.0&15.7&17.7&89 & 54 & 34 & 26 & 22&10&4.6&10.2 + rxj1550.0 - 2312 & 13.7&14.4&15.1&16.8&17.4&17.2&17.1&25 & 19 & 14 & 5.8&4.9&5.1&5.4&2.1 + rxj1550.9 - 2534 & 11.4&12.8&13.8&15.6&16.8&17.1&17.3&129&44 & 24 & 11 & 5.9&5.4&4.9&1.3 + rxj1551.1 - 2402 & 14.3&15.5&15.8&17.0&17.1&17.1&17.2&19 & 12 & 10 & 5.5&5.3&5.3&5.2&1.5 + rxj1557.8 - 2305 & 13.7&15.0&15.5&16.9&17.3&17.2&17.1&25 & 15 & 12 & 5.7&4.9&5.1&5.4&3.7 + rxj1558.1 - 2405 & 11.4&12.5&13.2&14.1&14.6&16.1&17.0&126&64 & 32 & 21 & 17&7.7&5.6&8.9 + rxj1558.2 - 2328 & 11.4&12.1&12.7&13.3&13.9&15.4&17.1&128&78 & 54 & 29 & 23&12&5.2&10.2 + rxj1600.7 - 2127 & 12.5&13.6&14.1&14.2&15.0&16.3&17.6&61 & 26 & 21 & 20 & 15&7.1&4.6&10.6 + rxj1601.1 - 2113 & 12.7&13.4&13.5&13.7&14.8&16.5&17.4&56 & 28 & 27 & 25 & 16&6.6&4.8&9.6 + rxj1601.9 - 2008 & 11.7&12.2&12.3&12.9&13.5&15.6&17.6&101&76 & 72 & 39 & 28&11&4.6&11 + rxj1602.0 - 2221 & 13.1&14.4&15.1&16.8&16.9&17.0&17.1&35 & 19 & 14 & 5.8&5.7&5.5&5.3&3.7 + rxj1602.8 - 2401a&11.1&11.8&12.4&14.1&15.0&16.0&17.0&156&95 & 71 & 20 & 15&8.4&5.6&3.6 + rxj1602.8 - 2401b&10.1&10.6&11.3&12.4&13.3&14.7&15.1&332&226&131&67 & 30&16&14.0&10.2 + rxj1603.6 - 2245 & 11.9&12.7&13.4&13.8&14.4&15.8&17.4&89 & 54 & 28 & 24 & 19&9.3&4.8&10.2 + rxj1603.9 - 2031a&11.9&12.7&13.4&13.6&14.5&16.0&17.6&88 & 54 & 29 & 27 & 18&8.4&4.6&10.1 + rxj1604.3 - 2130 & 11.9&12.3&13.0&13.7&14.3&15.9&17.1&83 & 73 & 37 & 25 & 19&8.7&5.2&8.9 + rxj1606.2 - 2036 & 10.4&11.1&11.7&12.8&13.5&15.1&15.4&271&164&96 & 45 & 27&14&12.0&11.1 + rxj1607.0 - 2036 & 9.1&12.4&13.3&13.9&14.5&15.6&17.9&648&69&30&23&18&11&4.3&8.7 + scopms015&11.3 & 13.4&13.8&14.1&15.1&16.9&17.9&140&28&24&20&14&5.7&4.3&10.4 + scopms017&13.9 & 14.7&15.0&16.9&17.6&17.4&17.2&23&17&15&5.7&4.6&4.9&5.2&1.5 + scopms019&11.3 & 12.1&12.5&13.5&13.9&15.5&16.6&133&79&65&27&23&11&6.2&8.8 + scopms022&14.3 & 15.5&15.5&17.4&17.4&17.3&17.3&19&12&12&4.9&4.9&4.9&5.0&1.4 + scopms027&11.8 & 12.6&12.7&12.9&13.5&15.3&16.4&89&57&49&42&28&13&6.9&3.0 + scopms028&13.1 & 15.0&15.9&17.1&17.2&17.0&16.9&34&15&9.3&5.3&5.1&5.4&5.7&3.7 + scopms044&11.8 & 12.3&12.7&13.0&13.4&15.5&17.7&91&72&53&36&28&12&4.5&10.3 + scopms045&12.8 & 13.4&13.8&14.2&14.6&15.8&18.0&48&28&24&20&17&9.6&4.2&10.7 + usco-160341.8 - 200557&14.0&15.4&15.7&17.3&17.4&17.4&17.3&22 & 12 & 10 & 5 & 4.8&4.9&5.0&2.2 + usco-160643.8 - 190805&10.1&10.8&11.6&13.2&13.8&15.6&15.8&336&203&110&30&24&11&9.4&8.5 + usco-160707.7 - 192715&14.6&16.1&16.6&17.9&17.9&17.7&17.6&17 & 7.6&6.2&4.4&4.3&4.5&4.7&1.5 + usco-160801.4 - 202741&11.1&11.9&12.5&13.7&14.1&15.8&16.1&158&86 & 64 & 25&20&10&7.8&8.5 + usco-160823.2 - 193001&14.4&15.5&15.7&17.3&17.4&17.3&17.3&18 & 11 & 11 & 5&4.9&5.0&5.0&3.7 + usco-160825.1 - 201224&14.3&15.8&16.3&17.3&17.5&17.5&17.5&19 & 10 & 7 & 4.9&4.8&4.8&4.8&1.5 + usco-160900.7 - 190852&10.8&11.5&12.2&13.6&14.1&15.6&16.0&196&115&76 & 26&21&11&8.3&6.0 + usco-160916.8 - 183522&14.3&15.8&16.0&17.4&17.7&17.6&17.4&19 & 9.6&8.1&4.9&4.5&4.7&4.9&2.1 + usco-160954.4 - 190654&14.5&15.5&16.0&17.4&17.4&17.5&17.5&18 & 12 & 8.4&4.9&4.9&4.8&4.7&2.2 + usco-161031.9 - 191305&10.8&11.5&12.2&13.3&13.9&15.5&16.1&200&114&77 & 29&23&12&7.7&8.5 + usco-161347.5 - 183459&13.4&13.5&13.5&13.4&13.3&13.2&13.1&28 & 27 & 27 & 29&29&31&34&2.1 + [ tablimits ] the most basic step in assessing the ability of current planet formation theories ( see section 1 ) to explain the widely separated and relatively massive candidate planetary companions that are observed is to measure their frequency . most of the 49 stars in our sample have relatively uniform detection limits ( @xmath046 @xmath9 at @xmath5300au ) , so a naive estimate of the frequency is approximately 2/49=4.1@xmath72% . here we have quoted the most likely frequency , and the bayesian 68% confidence interval on the frequency with a prior distribution where all frequencies are equally likely . the detection limits are not completely uniform , so the exact frequency will depend on more sophisticated analysis using monte carlo ( k08 ) or bayesian ( @xcite ; kraus and hillenbrand , in prep . ) techniques , and ultimately should depend on the separation and mass distributions of ulmcs . there is also a more fundamental issue that must be considered : this survey is not the first to be sensitive to the presence of ulmcs , and a full treatment should consider all surveys that have properly reported null detections and detection limits ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? @xcite reported 1 detection ( ab pic ) from 30 - 40 young solar - type targets surveyed ( depending on the definitions of `` young '' and `` solar - type '' ) . indeed , the full sample of past null detections is larger than our observed sample , suggesting that the true frequency could be lower by up to a factor of @xmath52 , and that our survey had good fortune to discover two new companions . alternatively , including the companions to ct cha and possibly gq lup ( likely more massive than 20@xmath2 ) may increase the true frequency of wide ulmcs , if only the details of the survey samples in which these companions were discovered were known . a large census of the literature is beyond the scope of a discovery paper , and the results of this analysis will be reported in a companion paper ( kraus et al . , in prep . ) the population of directly imaged wide ( @xmath540au ) ulmcs poses a significant challenge to planet formation models . existing models of solar - system scale planet formation ( i.e. , via core accretion or gravitational instability in a class ii disk ; @xcite ) have not been succesful at forming planets at the orbital radii they are observed here ( @xmath5200 au ) . it is also difficult to form wide ulmcs like a binary ( i.e. via fragmentation of the freefalling protostellar core or fragmentation in the massive protostellar disk ) without subsequently accreting sufficient mass to become a stellar or brown dwarf companion . however , one of these mechanisms must occur . complicating this picture is the possibility , and indeed likelihood , of multiple planet scattering in some formation scenarios . therefore , we can not answer the question of formation mechanism without examining the population of ulmcs in separation and mass space alongside more well studied classes of companions . the core - accretion mechanism for planetary formation only operates close to the host star ( @xmath05au ) , where protoplanetary disk densities are high enough to enable dust and ice to coagulate into protoplanets . therefore , a wide ulmc could only originate from core - accretion if it came from a scattering event . in any scattering event , low - mass companions are preferentially scattered outwards , so a core - accretion and scattering origin for wide ulmcs would require an even greater population of close , higher - mass objects . as brown dwarfs are well known to be rare less than @xmath05au from their host stars ( the brown - dwarf desert * ) , this scenario is very unlikely . another plausible formation scenario for wide ulmcs appears to be formation like stars from the fragmentation of protostellar clouds and massive circumstellar disks ( e.g. @xcite ) . in existing models , it is difficult to form ulmcs because fragments subsequently accrete mass , and the only fragments that remain near the opacity limit for fragmentation are those that are ejected @xcite . in an extreme case , the fragmentation of each core might be expected to be independent and companions would be expected to follow an initial - mass function ( i.e. random pairing ) . this is now well known to be incorrect , with an approximately linearly flat distribution of companion masses for solar type primaries and @xmath010@xmath73au separations ( @xcite ; @xcite ; kraus et al 2010 in prep . ) . as young solar - type stars have stellar companion fractions in the range 1222% ( @xcite ; k08 ; kraus et al , 2010 , submitted ) per decade of separation , this would mean that 0.2 - 0.3% of solar - type stars should have @xmath050 - 500au , @xmath740.006 to 0.02 companions ( i.e. 620@xmath2 for solar - type stars . we see a clear surplus to this model , strongly suggesting that wide ulmcs follow a different formation path to stars . if wide ulmcs were to form via fragmentation of a circumstellar disk when the primary is at the class ii stage ( disk masses 0.001 - 0.1m@xmath75 , e.g. @xcite ) it would be much easier to form wide companions than with core accretion . although most observed disks have masses of @xmath05m@xmath8 , a few disks around solar - type stars ( e.g. dl tau ) have large enough linear dimensions and mass ( up to @xmath01000au and 0.1@xmath76 , @xcite ) to fragment into observed ulmcs . relatively little work has been done examining in detail how these large disks might fragment , with the bulk of disk - fragmentation literature discussing the possible formation of solar - system scale planets via fragmentation . where models of large disks have been computed ( e.g. @xcite ) , it is clear that it is easier for them to become toomre - unstable and fragment than @xmath020au disks . the remaining questions to be answered about this fragmentation include how many fragments are expected , and if the fragments can accrete enough of the disk mass to become objects like gsc 06214 - 00210 b and 1rxs j160929.1 - 210524 b. the question of where each formation mechanism operates is certainly not answered yet , but could be in the next few years . radial velocity techniques will provide real constraints on the @xmath05 - 10au giant planet frequency as their time baselines increase , while direct imaging surveys will finish probing the wide ulmc frequency around young stars . finally , high angular resolution techniques such as aperture - masking and high - efficiency coronography will fill much of the yet - unprobed @xmath05 - 30au regime of orbital separations which likely forms the boundary between the dominance of core - accretion and fragmentation processes . m.i . would like to acknowledge support from the australian research council through an australian postdoctoral fellowship . we would like to express thanks to undergraduate students alison hammond and matthew hill from the university of sydney who made a first - pass astrometric analysis of the data . alk was suported by a sim science study and by nasa through hubble fellowship grant 51257.01 awarded by the space telescope science institute , which is operated by the association of universities for research in astronomy , inc . , for nasa , under contract nas 5 - 26555 . some of these observations were obtained at the hale telescope at palomar observatory , as part of a collaborative agreement between the california institute of technology , jpl and cornell university . 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 . the authors wish to recognize and acknowledge the very significant cultural role and reverence that the summit of mauna kea has always had within the indigenous hawaiian community . we are most fortunate to have the opportunity to conduct observations from this mountain .

we present the first results from the chang - es survey , a new survey of 35 edge - on galaxies to search for both in - disk as well as extra - planar radio continuum emission . chang - es is exploiting the new wide - band , multi - channel capabilities of the karl g. jansky very large array ( i.e. the expanded very large array , or evla ) with observations in two bands centered at 1.5 and 6 ghz in a variety of array configurations with full polarization . the motivation and science case for the survey are presented in a companion paper ( paper i ) . these first results are based on c - array test observations in both observing bands of the well - known radio halo galaxy , ngc 4631 . in this paper , we outline the observations and the data reduction steps that are required for wide - band calibration and mapping of evla data , including polarization . with modest on - source observing times ( 30 minutes at 1.5 ghz and 75 minutes at 6 ghz for the test data ) we have achieved best rms noise levels of 22 and 3.5jy beam at 1.5 ghz and 6 ghz , respectively . new disk - halo features have been detected , among them two at 1.5 ghz that appear as loops in projection . we present the first 1.5 ghz spectral index map of ngc 4631 to be formed from a single wide - band observation in a single array configuration . this map represents tangent slopes to the intensities within the band centered at 1.5 ghz , rather than fits across widely separated frequencies as has been done in the past and is also the highest spatial resolution spectral index map yet presented for this galaxy . the average spectral index in the disk is indicating that the emission is largely non - thermal , but a small global thermal contribution is sufficient to explain a positive curvature term in the spectral index over the band . two specific star forming regions have spectral indices that are consistent with thermal emission . polarization results ( uncorrected for internal faraday rotation ) are consistent with previous observations and also reveal some new features . on broad scales , we find strong support for the notion that magnetic fields constrain the x - ray emitting hot gas .

1205.5771

we present results of radio continuum observations of the edge - on galaxy , ngc 4631 , using the karl g. jansky very large array ( hereafter , the expanded very large array , or evla ) in its c configuration . these are the first results from a new survey , continuum halos in nearby galaxies an evla survey ( chang - es ) , whose motivation , science goals , galaxy sample criteria , and expectations for improvements over previous surveys are described in a companion paper ( * ? ? ? * hereafter , paper i ) . the sample consists of 35 nearby edge - on galaxies and observations are being carried out in two frequency bands ( 1.5 ghz and 6 ghz , i.e. in l - band and c - band , respectively ) in all polarization products . the survey includes observations in three evla array configurations ( b , c , and d ) but the test data presented here were obtained at both frequencies in the c configuration only . in this second paper , our goals are to outline the observations and data reduction procedures used for the test data ( sect . [ sec : n4631 ] ) and to present the initial scientific results for ngc 4631 ( sect . [ sec : results ] ) . we pay particular attention to the data reduction procedures , especially with respect to differences that are introduced by using the wide evla frequency bands that are now available . the galaxy , ngc 4631 , was chosen because of its extensive radio continuum halo which has been known for some time @xcite . for a recent summary of previous observations of this galaxy , see @xcite . in this paper , we refer to the region , @xmath3 kpc , as the disk - halo interface and use ` halo ' for emission on larger scales ( @xmath4 kpc ) . ` high - latitude ' or ` extraplanar ' are also used to describe either of these components . it is worth keeping in mind that halos are not necessarily smooth since substructure is generally observed , depending on the spatial scales that are probed . observations of ngc 4631 were carried out on 11 nov . 2010 with the pointing center set to the center of the galaxy at ra = 12@xmath5 42@xmath6 08@xmath701 , dec = 32@xmath8 32@xmath9 29@xmath104 the observing set - up is summarized in table [ table : obs - n4631 ] . all evla observations ( including continuum ) are now done in spectral line mode which facilitates the excision of radio frequency interference ( rfi ) from the data , effectively enables broader uv coverage via multi - frequency synthesis techniques ( sect . [ subsec : imaging ] ) and permits the extraction of spectral index information ( sect . [ sec : spectral_index ] ) . at the time of the observations , the widar correlator was restricted to a maximum of 8 spectral windows ( spws ) per base - band , i.e. per ac or bd intermediate frequency ( if ) . each spectral window has its own bandpass response , and contains 64 channels . with ac and bd ifs contiguous in frequency , the result was 1024 spectral channels centered at 1.5 ghz ( l - band ) for a total bandpass of 512 mhz , and 1024 channels centered at 6.0 ghz ( c - band ) for a total bandpass of 2.048 ghz . clearly , at the lower frequency , the achievable bandwidth can not be higher than the frequency itself , but the l - band bandwidth ( more so than c - band ) also suffers from rfi , restricting the bandwidth further . in fact , although we avoided known rfi at the ends of the band , we found that the l - band rfi , especially in the 1.52 to 1.64 ghz region within our band , to be persistent throughout our observations . together with other flagging , this reduced the effective l - band bandwidth to approximately 300 mhz and has prompted a modification of the correlator set - up for subsequent observations . the strength of the rfi in l - band achieved a maximum value which was 300 times that of 3c286 ( the flux calibrator ) so that even the sidelobes of the rfi swamped that of our astronomical sources at some frequencies . experimentation with flagged and unflagged data showed that it was not , in general , possible to salvage the long - spacing uv data while flagging only the short - spacings ; once rfi was detected , all uv data had to be removed over the time and frequency ranges that were affected . rfi at c - band was also present , but much less severe than in l - band . observations were carried out in the standard fashion , including scans of a flux calibrator ( which was also used as a bandpass calibrator as well as for the determination of absolute polarization position angle ) , a ` zero - polarization ' calibrator ( to determine the instrumental polarization ) and a phase calibrator . the flux calibrator was observed in 3 independent scans at each frequency for these test observations . only a single scan is necessary , in principle , but these data were intentionally ` over - calibrated ' for consistency checks . the zero - polarization calibrator was observed once at each frequency . the l - band observation of ngc 4631 was carried out in a single time block containing one on - source scan flanked by phase calibrator scans ( plus the flux and zero - polarization calibrator ) . the c - band observation of ngc 4631 with its calibrators was similarly observed in a single time block , but separated into 3 individual scans interspersed with phase calibrator scans . observations at each frequency were carried out within single blocks of total observing time in order to minimize time spent on overheads . for the regular data , however , attempts are being made , where possible , to split the observing time for given galaxy at each frequency into two observing blocks that are widely separated in hour angle ; more time is then spent on overheads but the result should be improved uv coverage with a better - behaved beam , as well as lower off - axis instrumental polarization ( sect . [ subsec : pol - cal ] ) . the data were reduced using the common astronomy software applications ( casa ) package , unless otherwise indicated . as improvements to casa are frequently being made during evla commissioning , a number of versions of casa were employed in the data reductions , in particular versions that were extant over the period may 2011 to july 2011 . the following steps were carried out for both the l - band and c - band data separately . on - line hanning smoothing is no longer an option at the evla , so all data were first hanning smoothed to minimize gibbs ringing in frequency , a step which significantly ameliorated rfi detection throughout the band . as a result , the frequency resolution is twice the channel width , the latter listed in table [ table : obs - n4631 ] . a significant amount of time was then spent flagging rfi . experimentation with automatic flagging routines had not yet provided results that were as dependable as manual flagging , requiring that flagging be done by hand ; however , such routines are improving and should shorten the data reduction process in the future . the flux density , i@xmath11 , ( stokes i ) of the flux calibrator , 3c 286 , was first set for the frequency , @xmath12 , of each spectral window using a known model and the ` perley - taylor-99 ' calibration scale . since each spectral window of each antenna has its own bandpass response , bandpass calibrations were required , again using 3c 286 . however , since the data are vector - averaged when producing an average bandpass for each spectral window , it is necessary to first carry out an initial gain and phase calibration with time ; applying the solution ensures that phase variations that may occur over the different scans of 3c 286 when it is observed through different elevations do not result in decorrelation when vector - averaging occurs . with the initial gain / phase solutions applied , bandpasses were then determined , one for each spectral window of each antenna , and the initial gain and phase solution is no longer used . the gain and phase were then calibrated with time for each of the flux , zero - polarization , and phase calibrators , where an antenna - based solution is found for each spectral window and each separate scan , followed by bootstrapping of the flux from the flux calibrator to both the zero - polarization ( oq 208 ) and phase ( j1221 + 2813 ) calibrators for each spectral window . at l - band , the flux densities of oq 208 and j1221 + 2813 varied systematically over the band from 0.72 to 1.16 jy ( positive spectral index . ] with an error of about 0.3% on each spectral window value ) and from 0.55 to 0.53 jy ( negative spectral index with an error of about 0.7% on each spectral window value ) , respectively . at c - band , the flux densities of the two calibrators ( in the same order ) varied from 2.4 to 1.9 jy ( negative spectral index with an error of about 0.4% on each spectral window value ) and from 0.46 to 0.44 jy ( negative spectral index with an error of about 0.5% on each spectral window value ) , respectively . the changing spectral index of oq 208 ( positive in l - band and negative in c - band ) is in good agreement with the known spectrum which peaks around 4 ghz ( * ? ? ? * or the nasa extragalactic database ) . the bandpass and gain / phase solutions as a function of time were then applied to all sources , antenna by antenna , and spectral window by spectral window , with the gain / phase solutions from the phase calibrator interpolated in time when applied to ngc 4631 . polarized intensity , p , and the polarization angle on the sky , @xmath13 , require calibration of stokes q and u. to initially set stokes q and u for 3c 286 , the absolute position angle on the sky is assumed to be constant at @xmath14 deg , but the percentage polarization varies with frequency . we obtained detailed evla data on this calibrator ( r. perley , private communication ) and fit a curve through the l - band data of the form , @xmath15 where @xmath16 is the percent polarization at frequency , @xmath12 ( mhz ) . in c - band , only a linear fit was necessary , @xmath17 ( @xmath12 in mhz ) . the estimated error , including measurement errors and fitted curves is of order 0.3% . q and u were then calculated at the frequency of each spectral window via @xmath18 where the total intensity , i@xmath11 , has previously been set via a known model ( sect . [ subsec : n4631-calibration ] ) . there is leakage between the right circularly polarized ( r ) and left circularly polarized ( l ) feeds at the evla and this leakage must be determined in order to calibrate the cross - terms ( rl , lr ) from which q and u are calculated . the leakage terms ( referred to as the ` d ' terms ) were determined from the ` zero - polarization ' source , oq 208 after first applying the previously determined gain / phase and bandpass solutions . finally , the absolute polarization angle on the sky ( the r - l phase difference ) was calibrated from the known position angle of 3c 286 , after first applying solutions for gain / phase , bandpass , and d terms . all corrections were then applied to the calibrators and the source ( gain / phase , bandpass , d terms , and absolute position angle corrections ) . the polarization calibration described in this section corrects for ionospheric faraday rotation , provided that this rotation is constant with time and position in the sky . as neither may be the case , it is important to determine the magnitude of such variations . at present , there is no casa routine to do this , but there is ( the task , tecor ) in the astronomical image processing system ( aips ) . we therefore copied a small amount of l - band data from each calibrator , ensuring that all observing times were represented in the copied data set , into aips for this purpose . ( corrections for the source are assumed to be the same as those for the phase calibrator , given their close proximity in the sky . ) tecor derives corrections for ionospheric faraday rotation from maps of total electron content obtained from nasa s crustal dynamics data interchange system ( ccdis ) which cover the time range of the observations . these measurements provide a crude correction since the spatial resolution is typically coarser than the source size . therefore , if significant fluctuations exist on smaller scales , we can not correct for them . in l - band , we find that the differential ionospheric faraday rotation ranges from a minimum of 2.365 to a maximum of 2.805 rad m@xmath19 over all sources , positions , and times . this result corresponds to a variation of 0.44 rad m@xmath19 , or 0.74 deg at the high frequency end of the band and 1.45 deg at the low frequency end . thus a typical error introduced by ignoring time variable or position variable ionospheric faraday rotation is about 1 deg and we therefore have made no additional correction for faraday rotation at l - band beyond the non - differential term that has already been taken into account in the normal data reduction . moreover , since faraday rotation @xmath20 and since the l - band and c - band observations were close together in time , no additional correction was necessary in c - band either . finally , the accuracy of wide field linear polarization imaging is limited by variations in instrumental polarization as a function of angle from the pointing center ; that is , the corrections that have now been applied are accurate only for the on - axis position . the off - axis instrumental polarization for a ` snapshot ' observation ] is estimated to be less than a few percent within the half - power point of the primary beam and increases at larger angles . in this paper , our polarization results ( sect . [ sec : pol_maps ] ) are shown only over regions that are within the primary beam full width at half maximum ( fwhm ) . in addition , when sources are tracked , the source parallactic angle rotates through the primary beam , minimizing the off - axis polarization errors . hence for these data , this effect is negligible . nevertheless , for future wide - field polarization mapping , a correction will likely be required ; a casa algorithm is currently being developed to aid in this process . our goal is to form single i , q , and u maps of ngc 4631 from each of the l - band and c - band data ; maps of linear polarization intensity , p , and magnetic field position angle , @xmath13 , can then be formed from q and u. as has always been done , the uv data are first fourier transformed ( fted ) into the map plane which results in an image ( the ` dirty image ' ) which is the convolution of the ft of the antenna distribution in the uv plane ( the ` dirty beam ' ) with the source sky brightness distribution . the dirty beam is then deconvolved from the dirty image ( i.e. it is ` cleaned ' ) to find an image of the source . the clark method @xcite of separating the cleaning process into major and minor cycles , treating each stokes plane separately , was used . the broad bandwidths of the evla widar correlator , however , introduce some new imaging requirements and considerations that were , for the most part , unnecessary with the vla . these are described briefly here:_a ) _ _ bandwidth smearing : _ if the observations had not been in spectral line mode , then the broad frequency band would have introduced severe chromatic aberration into the image ; this aberration worsens with distance from the map center . however , since spectral line mode is being universally employed at the evla , bandwidth smearing is instead restricted to the spectral resolution of the data ( by using the multifrequency synthesis technique , see next section ) . for example , for our data at the center of either band and at a distance from the map center that is twice the fwhm of the primary beam , the loss in amplitude due to bandwidth smearing is only about 1.5% , or essentially negligible where most of the emission occurs._b ) _ _ multi - frequency synthesis ( mfs ) : _ the positions of the antennas in the uv plane are measured in wavelengths and those positions will vary as the frequency changes across a given band . consequently , if each spectral channel is gridded separately , there is an improvement in uv coverage over a case in which channels are averaged together . the clean algorithm employed in casa utilizes this multi - frequency synthesis technique ( * ? ? ? * also called bandwidth synthesis)._c ) _ _ widefield imaging : _ the sensitivity of the array to emission that is far from the map center and even outside of the telescope s primary beam means that the array itself is non - coplanar to such emission . thus , the ` w term ' in the ft ( the coordinate in the line - of - sight direction towards the source ) , which can be ignored near the map center , becomes important . casa clean employs the ` w - projection algorithm ' @xcite to correct for this effect._d ) _ _ spectral index fitting : _ with wide bands , the spectral index of the source itself , which is spatially variable , needs to be fitted during imaging @xcite . in casa s clean algorithm , this is accomplished by assuming that the emission at any position can be described by , @xmath21 where @xmath22 is a reference frequency within the band ( here taken to be the band center ) , i@xmath23 is the specific intensity at the reference frequency , @xmath24 is the spectral index and @xmath25 describes the spectral curvature . the function is expanded in a taylor series about @xmath22 , resulting in a polynomial fit to the spectrum . the fit includes all unflagged channels across any given band and is not done on a ` per spectral window ' basis . in addition , the fitting is done using all data so that the resulting parameters correspond to a single spatial resolution ( for example , the @xmath24 map is not formed such that the spatial resolution would vary across the band ) . @xcite outline the implementation of spectral fitting and provide examples as to how the image can improve with increasing numbers of taylor terms , @xmath26 . _ e ) _ _ multi - scale clean : _ improvements have been made over the traditional clean which modeled all emission as a collection of point sources . the clean implementation in casa now allows for a multi - scale clean @xcite which assumes that the emission can be modeled as a collection of components over a variety of spatial scales . these scales are typically chosen to span the range from an angular size of ` zero ' ( corresponding to a point source , or the classic clean analogy ) to a scale of order the maximum spatial scale expected to be present in the emission . since the latter is not always known in advance , in practice , the maximum scale size is increased systematically until no more flux is actually cleaned out in that largest scale . for details of the casa implementation of the combined multi - scale , multi - frequency synthesis ( ms - mfs ) clean , see @xcite . _ f ) _ _ primary beam : _ the primary beam ( pb ) varies from one end of the band to the other and , as a result , nowhere in the field of view is there a ` null ' as there would be for a monochromatic primary beam @xcite . consequently , it is the frequency - averaged pb ( pb@xmath27 ) that the final cleaned image must be corrected for . moreover , since the frequency dependence of the pb imposes its own spectral index across the field of view , all spectral index maps ( sect . [ sec : spectral_index ] ) must be corrected for this effect as well . ngc 4631 was mapped taking into account all of the above points . at both bands , a very wide field was first mapped ( 3 degrees for l - band , or 6 times the primary beam fwhm ; 34.1 arcmin for c - band or 4.6 times the primary beam ) to determine how large an image needed to be mapped to include all sources of significant flux in the clean . smaller fields were then mapped completely , as required . for the multi - scale clean , we used typically 4 to 6 equally - spaced spatial scales , the first one always corresponding to point sources as in the classic clean . for example , at l - band , scales up to 4.2 arcmin were employed , where the largest scale was chosen according to the considerations of point _ e ) _ above . we also self - calibrated each data set once , finding that a single amplitude + phase self - calibration resulted in a small improvement in the rms map noise . after extensive experimentation , we adopted 3 sets of parameters representing 3 different spatial resolutions for each frequency and display them , along with some ancillary data , in figs . [ fig : l - band ] and [ fig : c - band ] ; the corresponding map parameters are given in table [ table : map - n4631 ] . the theoretical noise values for these observations ( including confusion estimates ) are 18 and 3.7 @xmath0jy beam@xmath1 at l - band and c - band , respectively . our lowest rms noise values are only 1.6 and 1.3 times higher for the total intensity images . there are , however , several residual linear cleaning artifacts in the l - band images that are visible immediately to the east and west of the galaxy s disk ( see fig . [ fig : l - band]c in particular ) . the conclusions that we present in this paper are not affected by these artifacts . the disconnected emission seen approximately 2 arcmin above and below the plane in fig . [ fig : c - band]c is likely real and part of the halo ; missing low - order spacings contribute to the well - known appearance of the galaxy s disk emission sitting on a negative bowl . note that , as is usually done , all displayed total intensity images are not corrected for the primary beam in order to show images with uniform noise . when numerical results are given , however , the results have been corrected for the frequency - averaged primary beam ( pb@xmath27 ) . note also , that we carried out some subsequent analysis using aips . stokes q and u maps were formed with the same sets of parameters as the total intensity images , except that fewer spatial scales were required during cleaning , given the weaker emission in these maps . the relevant parameters for these images are given in table [ table : map - n4631 ] . in principle , the rms noise on the q and u maps should be the same as the i map ; however , the lower signal in the cross - hands means that residual errors associated with cleaning the high intensity emission are no longer present . at l - band , our lowest rms noise is 1.2 times the theoretical value and at c - band , our lowest noise matches the theoretical value . note that , in order to obtain sufficient signal - to - noise ( s / n ) in the polarized emission , all spectral windows were used . we formed polarized intensity images , p@xmath28 , for each of the uv weightings of figs . [ fig : l - band ] and [ fig : c - band ] , correcting for the bias introduced by the fact that p images do not obey gaussian statistics @xcite . maps of the polarization angle of the electric field vector , @xmath29 , were also formed for points for which the linear polarization was @xmath30 4@xmath31 was relaxed slightly from 5@xmath31 in order to show the trends in vector orientation with position . ] . when displayed , we have rotated @xmath13 by 90 degrees to illustrate the observed ( uncorrected for internal faraday rotation ) direction of the magnetic field . we show the results for only a single representative data set in the two frequency bands ( robust = 2 for l - band and robust = 0 + uv taper for c - band ) in fig . [ fig : pol_maps ] . consistent with the total intensity images ( sect . [ subsec : imaging ] ) the displayed maps are shown without primary beam corrections . we then carried out a pb correction using the same frequency - averaged primary beam , pb@xmath27 , as was used for the total intensity images . this correction is good to zeroth order , but does not take into account first order effects which would result from differences between the right and left circularly polarized pbs . first order corrections are not yet implemented in casa but we expect them to be small for this data set . for example , if the difference between the right and left circularly polarized pb voltage response is of order 10% at a distance from the map center corresponding to 30% of the primary beam ( 17.4 arcmin and 4.6 arcmin at l - band and c - band , respectively ) , then the error in q or u would be 3% of the q or u intensity at that point . pb corrections of the linearly polarized intensity maps shown in fig . [ fig : pol_maps ] of all emission greater than 5@xmath31 resulted in increases in flux of 5% and 40% at l - band and c - band , respectively . maps of @xmath13 ( shown in fig . [ fig : pol_maps ] ) do not require pb correction . we then formed maps of percentage polarization for the two bands ( these also do not require pb correction ) . the maps were formed for emission which was greater than 5@xmath31 in both linear polarization intensity as well as total intensity , ensuring that the spatial resolutions are matched . we do not display the l - band maps since the result had sufficient s / n in only 20 beams and these positions were non - contiguous spatially . the c - band percentage polarization map is shown in fig . [ fig : percentpol ] . we calculated the average percentage polarization over the region shown by measuring the total polarized flux ( rather than per pixel values ) and dividing by the total flux , so that high values around the periphery of the emission do not skew the result ; we find an average of 7.0% for the result corresponding to the region displayed . this region is mainly in the disk where some faraday depolarization may be occurring ( sect . [ sec : pol_maps ] ) . as indicated in sect . [ subsec : imaging ] point _ d ) _ , the wide bands that are now available at the evla make it possible , in principle , to create spatially resolved spectral index maps , @xmath24 , ( and curvature maps , @xmath25 , if there were sufficient signal - to - noise ) over a single band with a single observation and common spatial resolution . one could form a map of @xmath24 for l - band and for c - band individually , and then form such a map between the two bands , assuming a constant spectral index between the two bands . for observations at two frequencies with a single evla array , as for our c - array test observations , however , there are some limitations . first , as specified in paper i , we do not detect spatial scales at l - band and c - band that are greater than 16.2 and 4 arcmin , respectively one minute duration ( snapshots ) these values could reduce to as little as 8 and 2 arcmin , respectively ( see _ observational status summary _ at http://evlaguides.nrao.edu ) . our observations ( see table [ table : obs - n4631 ] ) are intermediate between a snapshot and full synthesis . consequently , the spatial scales quoted in the text are likely upper limits . ] . the halo of ngc 4631 at both bands has previously been observed to extend to approximately 12 arcmin ( see , for example , fig . 9 of @xcite and fig . 3 of @xcite ) . it is therefore clear that not all flux has been detected at c - band and , as fig . [ fig : l - band ] illustrates , the broadest scale halo emission has not been detected at l - band either . future d - array observations , with additional shorter spacings , should remedy this situation at l - band . using previously published total flux estimates , we can normally quantify the amount of flux that has been missed in our observations . however , at c - band , a wide range of 4.85 ghz fluxes have been reported , likely reflecting similar problems with spatial scales , for example , @xmath32 mjy , 243 mjy ( see ned ) and 480 mjy @xcite . our primary - beam - corrected flux in c - band is @xmath33 mjy . using the average global spectral index of @xcite , @xmath34 , our frequency - adjusted flux would be @xmath35 mjy , indicating that there is significant flux missing from the c - band maps . we thus do not show our c - band spectral index maps in this paper . at l - band , we do form spectral index maps . the primary - beam - corrected flux from these c - array observations is @xmath36 mjy , or 77% of the vla d - array vla fluxes of 1.22 jy at @xmath37 ghz @xcite or 1.29 jy from the westerbork synthesis radio telescope at @xmath38 ghz @xcite . since the missing flux is on the largest scales , we form spectral index maps at l - band beginning with the highest resolution image ( fig . [ fig : l - band])a , and then consider how this and other errors affect the result . since the l - band 500 mhz bandwidth is much narrower than the frequency spacing which has been used in the past to calculate @xmath24 ( e.g. 610 mhz to 1412 mhz , or 4.75 ghz to 10.7 ghz , see * ? ? ? * ) , the values of @xmath24 presented here can be thought of as tangents to the frequency spectrum at each spatial position in l - band . spectral index and curvature maps were formed from ratios of the taylor expansion coefficient maps which describe eqn . [ eqn : spectralfit ] and @xmath39 , where @xmath40 , @xmath41 , and @xmath42 are maps of the first 3 taylor terms . ] , after first correcting these coefficient maps for pb@xmath27 and then introducing a 5@xmath31 total intensity cut - off . it has been shown that increasing the number of terms to the polynomial fit will generally improve the result , subject to s / n considerations @xcite . for example , fitting both @xmath24 and @xmath25 allows @xmath24 to be well - determined at the reference frequency even though @xmath25 may not have sufficient s / n in each synthesized beam to be presented as a map . in fig . [ fig : alpha ] we show spectral index maps corresponding to the high resolution l - band data ( fig . [ fig : l - band]a ) for a straight spectral index assumption ( a ) and then allowing for curvature within the band ( b ) . the `` mottled '' appearance of the two maps is due in part to the fact that point - to - point variations are significant . for example , after subtracting the two spectral index maps , the rms of the result is @xmath43 ( omitting approximately a beam width around the perimeter of the emission ) suggesting that the per - pixel error from the fitting is of this order . there are , however , variations in the map that exceed this uncertainty . in addition , there are some global trends in @xmath24 which are very similar between the two maps and the average spectral indices ( @xmath44 , averaged over the maps ) are in very good agreement , differing by only 0.01 . clearly , the error bar on the spectral index decreases with the number of points that are averaged , i.e. with increasing s / n , as one would expect ; we will return to this point in sect . [ results : spectral_index ] . regarding the effect of missing broad scale flux , our measured @xmath24 will not be affected provided the broad scale emission has the same spectral index as the disk and contributes a constant intensity at each position . it is known , however , that the halo spectral index , which we associate with the broad scale emission , is steeper than that of the disk . a variety of measurements have been made over different frequency ranges , but typically , off - major - axis values fall in the range , @xmath45 ( see discussion in * ? ? ? * ) . taking the extreme case of @xmath46 , we can determine how the addition of a ` plateau ' of emission of this spectral index to our data would affect our measured values of @xmath24 . if the halo emission occupies an emission ellipse of major @xmath47 minor axis extent , approximately 15 arcmin @xmath47 12 arcmin @xcite , then the emission region of fig . [ fig : alpha ] occupies approximately 20% of this area . scaling the missing flux of fig . [ fig : l - band]a to this area , assuming that the broad scale flux uniformly fills the emission ellipse , then the missing emission corresponds to a plateau of 0.12 mjy beam@xmath1 , on average , over the disk at the central frequency . using the measured average intensity from fig . [ fig : l - band]a over the same region as the spectral index ( taking @xmath48 ) the addition of such a steep spectral index plateau would adjust our measurement of @xmath24 from -0.84 to -0.88 ( @xmath49 ) . this is a worst - case scenario unless there are missing spatial scales that are _ smaller _ than 11 arcmin ( the major axis size ) . spectral index , @xmath24 , and curvature , @xmath25 , maps for l - band were also formed for the remaining mid and low resolution maps corresponding to fig . [ fig : l - band]b and c , respectively and are shown in fig . [ fig : alpha_lowres ] . for these resolutions , we find @xmath50 and @xmath51 , respectively . the curvature , @xmath52 , averaged over the emission regions for these two resolutions is @xmath53 in both cases , agreeing to within @xmath54 with the result at high resolution . if we consider fitting errors , possible errors from missing flux , and variations between the different uv weightings , our average spectral index is @xmath55 . as mentioned earlier , it is worth stressing that the wide bands used here allow for considerable flexibility in ` tuning ' the spatial resolution of the array by appropriate use of uv weighting . this means that the spatial resolutions of l - band and c - band data can actually be matched exactly , even with observations in only a single array configuration . therefore spectral index maps can also be made between l and c - bands from observations in a single array configuration . although we formed such maps for ngc 4631 , we defer the results until all missing c - band flux is recovered through additional observations ( paper i ) . these new evla observations , even though they were obtained during a limited test run in a single array configuration ( table [ table : obs - n4631 ] ) , have revealed a number of new features not before seen in ngc 4631 . in this section , our intention is not to make an exhaustive list of all features , but rather summarize the more obvious or new results and to place them in the context of previous knowledge of this galaxy . the total intensity images displayed in figs . [ fig : l - band ] and [ fig : c - band ] show numerous extensions away from the plane at both frequencies . at l - band , in particular , two of the more prominent features appear in projection as loops or partial loops ( although they could represent shells ) that have been outlined in red in fig . [ fig : l - band]b . although we have noted a few cleaning artifacts in this image to the east and west of the major axis of the galaxy ( sect . [ subsec : imaging ] ) , the loops do not fall within the affected regions and they are either of sufficient signal - to - noise and/or persistent over various uv weightings ( some not displayed here but formed during experimentation with uv weightings as described in sect . [ subsec : imaging ] ) that we consider them to be real features . the diameter of the larger northern loop , which is only partial in fig . [ fig : l - band]b , is 6.3 kpc . it is also visible in the lower resolution fig . [ fig : l - band]c where it appears to ` frame ' the companion dwarf elliptical galaxy , ngc4627 which is 2.6 arcmin to the nw of the galaxy s center . this is likely a coincidence , since we will show below that the loop is related to the hot x - ray gas . the diameter of the smaller more easterly loop is 2.6 kpc and this loop becomes ` filled in ' in the lower resolution fig . [ fig : l - band]c . the eastern side of this loop and the weaker emission that extends upwards from it corresponds to the hi ` worm ' observed by @xcite . the latter feature sits above a significant hi supershell found by those authors whose diameter ( 3 kpc ) is approximately the same as our eastern loop . it is likely that these features are related and are associated with star forming regions that are present in the underlying disk . the closest prominent star formation complex is located at ra @xmath56 12@xmath5 42@xmath6 22@xmath57 , dec @xmath56 32@xmath8 32@xmath9 42@xmath58 ; it is marked with a cross in the h@xmath24 map of fig . [ fig : c - band]c . a search of previous radio continuum images of ngc 4631 @xcite does not reveal these loops , although the continuum image in @xcite shows hints of the larger central loop . since we see extraplanar structure on all scales in this galaxy , detecting specific features requires appropriately weighted uv data over scales that match the feature of interest . the spatial scales highlighted by our new data have now revealed these loops and the other arcs and filaments that we see extending from the plane . it is possible that some of the other broad extensions ( for example , at ra @xmath56 12@xmath5 42@xmath6 12@xmath57 , dec @xmath56 32@xmath8 31@xmath9 30@xmath58 ) would resolve into loops if viewed at higher resolution and sensitivity . the close connection between x - ray emission and the radio halo , first pointed out for this galaxy by @xcite and @xcite , finds stronger support in these observations . in fig . [ fig : l - band_xray ] , we show the soft x - ray emission from fig . [ fig : l - band]a , enhanced to show extra - planar emission , together with the low resolution l - band image from fig . [ fig : l - band]c . in this figure , we see that the x - ray emission forms a loop interior to the northern radio continuum loop , suggesting that hot gas may be confined by magnetic pressure . it is interesting that there is even a smaller vertical x - ray protrusion ( ra @xmath56 12@xmath5 42@xmath6 07@xmath57 , dec @xmath56 32@xmath8 36@xmath59 ) above the gap on the north - eastern side of the radio continuum loop . a close relationship has been found between radio continuum and x - ray emission in the outflow of ngc 253 as well @xcite . we can determine whether such confinement is feasible , assuming a simple case of unity filling factors , by estimating the minimum energy magnetic field strength following @xcite . assuming a proton to electron number density ratio of 100 , a line of sight distance equal to the width of the radio loop as shown in fig . [ fig : l - band_xray ] , an isotropic field direction , and a spectral index equal to the global average disk value , we find b@xmath60 @xmath0 g . the magnetic pressure is then @xmath61 erg @xmath62 . by comparison , the thermal pressure from the hot , x - ray - emitting gas is @xmath63 @xcite , implying that magnetic pressure is indeed sufficient to confine the thermal gas . at c - band , where we have higher spatial resolution ( fig . [ fig : c - band ] ) , we continue to see many vertical extensions . again , we do not discuss every feature , but point out the brightest extension which is on the north of the disk at ra = 12@xmath5 42@xmath6 09@xmath57 , dec = 32@xmath8 33@xmath59 10@xmath58 , and best seen in fig . [ fig : c - band]b . this extension is directly below the west side of the large northern loop seen in fig . [ fig : l - band]b . this feature protrudes 1 kpc above the main emission below it , or about 1.6 kpc from mid - plane . the detail and numerous extensions observed on this map reveal the complexity of the disk - halo interface in this galaxy . the disconnected ` patches ' of emission visible in fig . [ fig : c - band]c above and below the plane appear to represent brighter parts of the broader scale halo emission which has not been detected in its entirety in our data due to missing short spacings at c - band ( sect . [ subsec : imaging ] ) . the feature at ra @xmath56 12@xmath5 42@xmath6 23@xmath57 , dec @xmath56 32@xmath8 34@xmath59 30@xmath58 for example , corresponds to the hi worm mentioned above . the well - known close relationship between radio continuum , h@xmath64 , and co emission is also evident from fig . [ fig : c - band ] , to the extent that ` kinks ' , widenings and narrowings along the major axis are correlated between these components . radio continuum and h@xmath24 peaks are nicely coincident in the wider ` bulge ' region @xcite about 3 arcmin to the east of the galaxy s center within which the bright star formation complex mentioned above is located . the linear polarization maps with observed magnetic field vectors are shown in fig . [ fig : pol_maps ] and the percentage polarization with observed magnetic field vectors for c - band is shown in fig . [ fig : percentpol ] . recall that the q and u images have lower rms than the total intensity images ( sect . [ sec : polarization ] ) and hence some polarized emission can be seen outside of regions of total intensity . all maps have been corrected for foreground ionospheric faraday rotation , where necessary ( sect . [ subsec : pol - cal ] ) , but no faraday rotation correction has been done for the galaxy itself ; therefore , especially in the disk at l - band , we expect significant faraday depolarization , as indicated in sect . [ sec : polarization ] . this is evident in fig . [ fig : pol_maps]a which shows little polarized intensity in the disk at l - band . previous l - band images ( see especially * ? ? ? * ; * ? ? ? * ) similarly show little or no polarization right in the disk . the halo , however , should not be as strongly affected . the effects of faraday rotation should also be lower at c - band ( fig . [ fig : pol_maps]b ) but will not be negligible in the disk . for example , rotation measures of up to @xmath65 300 rad / m@xmath66 have been found for ngc 4631 in a larger 85 arcsec beam @xcite , corresponding to 43 degrees of rotation at 6 ghz ; the value could be higher still in our data set , given our smaller beam . again , away from the plane , faraday rotation should be minor , as has been demonstrated previously for the halo region of the edge - on galaxy , ngc 5775 ( * ? ? ? * cf . figs . 3 and 5 of that paper ) nevertheless , we prefer to be cautious about our comments regarding the magnetic field orientation in the following discussion . at l - band , the most prominent halo feature is a highly linearly polarized spur located at ra @xmath56 12@xmath5 42@xmath6 20@xmath57 , dec @xmath56 32@xmath8 34@xmath9 30@xmath58 . this spur is visible in the linearly polarized map of @xcite and also in the lower resolution 5 ghz map of @xcite . the spur is directly above the eastern radio continuum loop and may be related to the disturbance that has caused that loop as well as the hi supershell below it ( sect . [ sec : results_total ] ) . however , the galaxy also displays an v - shaped field structure on the northern side ( or ` x - shaped ' globally , see above references ) and this spur may be associated with the eastern side of that larger feature . at 6 ghz ( fig . [ fig : pol_maps]b , fig . [ fig : percentpol ] ) , it is interesting that the observed magnetic field orientation within the disk is parallel to the disk . previously published maps have shown the magnetic fields near the center of ngc 4631 to be perpendicular to the disk ( for example , see the 8 ghz image of * ? ? ? * arcsec resolution ) , although at larger radii the fields are parallel to the disk again @xcite . our 12 arcsec observations and omission of the largest spatial scales are targetting different scales and may now be revealing a disk - parallel component near the galaxy s center , although faraday - corrected maps are still required to confirm this . away from the disk , the magnetic field tends to become perpendicular to the plane , in agreement with previous observations . extra - planar polarized emission both to the north - east and north - west of the galaxy s center agrees in general with the image of @xcite at 5 ghz . a final point of interest is the sharpness with which the _ apparent _ magnetic field orientation changes ( 90 degrees within a single beam ) at ra @xmath56 12@xmath5 42@xmath6 12@xmath57 , dec @xmath56 32@xmath8 32@xmath9 30@xmath58 ( see fig . [ fig : percentpol ] ) . this change occurs at the location of a bright hii region complex visible in fig . [ fig : c - band]c which has been labelled cm67 by @xcite . if such a change is still visible after having corrected for faraday rotation , then it lends support for the notion that the magnetic field is swept outwards with the outflow from specific star forming regions . note that there are radio continuum spurs above and below this location ( fig . [ fig : c - band]c ) . if the change is not intrinsic but is due to faraday rotation alone , then either the ordered magnetic field in the line of sight has changed sharply at this position or the electron density has changed abruptly , or both . in either case , cm67 is an important region for follow - up studies with applied rotation measure corrections . our l - band spectral index maps ( figs . [ fig : alpha ] and [ fig : alpha_lowres ] ) are the first that have been formed from _ within _ an observing band for an edge - on galaxy and represent a tangent to the spectrum at 1.5 ghz , rather than an average between broadly separated frequencies as has been done in the past . the average spectral index is @xmath67 = - 0.84 @xmath65 0.05 ( sect . [ sec : spectral_index ] ) and therefore the l - band maps are clearly dominated by non - thermal emission . this result is typical of the values found in other galaxies @xcite and the milky way @xcite . previous values for the disk of ngc 4631 cover a rather wide range ( see * ? ? ? * ) , for example , -0.68 ( from 610 mhz and 10.7 ghz measurements ) , -0.45 ( from 327 mhz and 1.49 ghz ) , -0.69 ( from 327 mhz , 610 mhz , 1412 mhz , and 10.7 ghz data ) , and -0.9 ( from 610 mhz and 1412 mhz ) . also , although the global spectral index of ngc 4631 has been studied extensively @xcite the only previous spectral index _ maps _ that have been published can be found in @xcite whose highest resolution was 58 arcsec . our maps are the highest resolution spectral index maps yet presented for this galaxy . the average , global spectral curvature is positive ( @xmath68 , sect . [ sec : spectral_index ] ) . the most likely explanation is the contribution of thermal emission from the disk . to see whether this is feasible , we use the thermal flux density estimate of @xcite of @xmath69 mjy who makes use of the h@xmath24 luminosity of ngc 4631 and adjusts upwards to account ( roughly ) for dust obscuration . extrapolated to 1.5 ghz with a @xmath70 frequency dependence gives @xmath71 mjy with a variation of 2 mjy across l - band . the curvature that we have found from eqn . [ eqn : spectralfit ] corresponds to an increase of 2% at the upper frequency band edge in comparison to a non - curving spectral index . with a total flux density of 935 mjy at the band center ( sect . [ sec : spectral_index ] ) , a curved spectral index results in an increase of 16 mjy at the high frequency end of the observing band , over the value that would be observed if the spectral index were constant . this increase is less than the estimated thermal flux density at the upper end of the band , so it is clear that a thermal contribution is capable of accounting for the observed curvature . however , thermal emission contributes , on average , less than 7% of the flux at l - band , assuming that the thermal estimate of @xcite can be applied . an interesting result is the detail with which we have mapped the spectral index distribution in a single observation in a single frequency band . as indicated in sect . [ sec : spectral_index ] , the uncertainty at any point is approximately @xmath720.4 , though it becomes lower depending on the size of the region over which averages are taken . for the average spectral index over all maps , the uncertainty is smaller , i.e. @xmath720.05 . consequently , in the following , we point out only values of @xmath24 that are significant in comparison to the uncertainties which have been appropriately averaged over the region of interest and for which there are similar trends for all maps shown in figs . [ fig : alpha ] and [ fig : alpha_lowres ] . for example , contrast in the spectral index is observed in the region of the eastern loop ( see dashed arc in fig . [ fig : l - band]b ) . the contrast is most clearly seen in figs . [ fig : alpha_lowres]a and b. the average spectral index along the sides of the loop itself ( all maps ) is @xmath73 and this value does not differ significantly from the underlying disk at the loop ` footprints ' . in the regions interior to the loop and just exterior to it , on the other hand , the values are much steeper ( darker regions in fig . [ fig : alpha_lowres]a and b ) . the average spectral index interior to the loop is @xmath74 . both values indicate that a thermal contribution will be negligible at these locations and , as we did in sect . [ sec : results_total ] , we estimate a minimum energy magnetic field strength using the same assumptions as in that section but with these measured spectral indices . the result is b@xmath75 @xmath0 g along the loop to @xmath76 hz , results in b@xmath77 @xmath0 g . ] . this is not the first time that spectral index contrasts or steep spectral indices have been observed in galaxies that experience outflows . for example , @xcite show that the spectral index of radio continuum features that are located along hi shells in ngc 5775 is flatter than in the interior and that this flattening can not be due to a thermal contribution . as in ngc 4631 , the spectral index along the loop itself is not significantly different from that of the underlying disk at its location . moreover , although such steep spectral indices are not generally seen in our own galaxy @xcite , @xcite also find regions of very steep spectral index in the galaxy , ngc 253 , i.e. @xmath78@xmath79 is seen within the outflow cone originating from the nuclear region of that galaxy . ngc 4631 is known to be experiencing widely distributed disk - related activity rather than a localized nuclear starburst ( e.g. see * ? ? ? * ) , but the physics may be similar . for ngc 253 , strong electron energy losses are implied , for example . we defer the point - by - point decomposition of the map into thermal and non - thermal components to a future paper ; nevertheless , it is possible to see evidence for the dominance of thermal emission in at least two specific star forming regions , best seen at the highest resolution spectral index maps ( fig . [ fig : alpha ] ) . one has been pointed out earlier ( sect . [ sec : results_total ] ) and is marked with a cross on the h@xmath24 map of fig . [ fig : c - band]c as well as the high resolution spectral index map of fig . [ fig : alpha]b . the spectral index at this position is @xmath24 = -0.16 @xmath65 0.09 ( averaged from fig . [ fig : alpha]b over the star forming region ) which is entirely consistent with a thermal spectrum , within uncertainties . another star forming region located 1 arcmin to the ssw of this location and visible as a discrete peak in fig . [ fig : c - band]c also shows a similarly flat index , consistent with thermal emission . in this paper , we have presented the first results from a new survey , called chang - es , to observe radio continuum halos in 35 edge - on , normal spiral galaxies . the galaxies are being observed in all polarization products in two different bands , 1.5 ghz ( l - band ) and 6 ghz ( c - band ) , and over 3 different evla array configurations . this is the first comprehensive radio continuum survey of halos to include all polarization products . the motivation and science goals for the survey have been presented in paper i. our initial chang - es test observations of ngc 4631 have been carried out at c array alone and therefore are sensitive to the disk , the disk - halo interface , and inner halo emission ; we do not detect the faint large scale halo emission because of the lack of large scale sensitivity in this array configuration . our new results demonstrate that , even with modest integration times , new details of the emission in this galaxy have emerged . one advantage of the wide - band , multi - channel evla capabilities is the ability to form spectral index maps at a common spatial resolution _ within _ a single observing band in a single array configuration . we have formed such maps at l - band . our l - band spectral index map has the highest resolution yet obtained for this galaxy and is the first in - band spectral index map published for an edge - on galaxy . in addition , with appropriate uv weighting , the evla wide bands have allowed us to match spatial resolution between l - band and c - band in a single array configuration . @xmath80 at 1.5 ghz , two extra - planar features appear to form loops in projection . the larger loop is 6.3 kpc in diameter and located on the north side of the disk slightly to the west of center . the smaller loop , 2.6 kpc in diameter , is also on the north side of the disk but towards the east ( see fig . [ fig : l - band]b ) . @xmath80 the larger loop is exterior to the soft x - ray feature observed by @xcite ( fig . [ fig : l - band_xray ] ) . the minimum energy magnetic field is 8 @xmath0 g and the magnetic pressure in this loop ( @xmath61 erg @xmath62 ) is sufficient to constrain the hot x - ray gas ( @xmath63 erg @xmath62 ) . @xmath80 the smaller loop is approximately at the location of the hi worm and supershell identified by @xcite . the spectral index within and adjacent to the loop is much steeper ( @xmath81 ) than along the loop itself ( @xmath82 ) . @xmath80 at 6 ghz , the polarization is higher and the average percentage polarization is 7% over regions within which the s / n was high enough to be measured , i.e. predominantly in the disk . the _ apparent _ magnetic field orientation in the plane of the disk ( uncorrected for faraday rotation ) appears to be parallel to the disk . @xmath80 there is one position in the disk , corresponding to the hii region complex , cm67 , at which the apparent field orientation changes sharply ( fig . [ fig : percentpol ] ) ; this is a target region for future in - depth analysis including faraday rotation . @xmath80 at 1.5 ghz , the average spectral index is @xmath55 , indicating that the emission is predominantly non - thermal , on average , throughout the disk . however , a small thermal contribution is sufficient to explain the observed positive spectral curvature in this band . @xmath80 at specific discrete locations in the disk , it is clear that the thermal contribution is not negligible . an example is the location of a specific star forming region identified with a cross in fig . [ fig : c - band]c whose flat 1.5 ghz spectral index is consistent with thermal emission . jai and djs would like to thank the staff at the evla for their warm welcome and assistance during their sojourn in socorro . research at ruhr - universitt , bochum , is supported by deutsche forschungsgemeinschaft through grants , for1048 and for1254 . the digitized sky surveys were produced at the space telescope science institute under u.s . government grant nag w-2166 . the second palomar observatory sky survey ( poss - ii ) was made by the california institute of technology with funds from the national science foundation , the national geographic society , the sloan foundation , the samuel oschin foundation , and the eastman kodak corporation . the national radio astronomy observatory is a facility of the national science foundation operated under cooperative agreement by associated universities , inc . beck , r. , & krause , m. 2005 , astron . 6 , 326 , 414 bhatnagar , s. , rau , u. , green , d. a. , & rupen , m. p. 2011 , , ( arxiv1106.2796b ) braun , r. , oosterloo , t. a. , morganti , r. , klein , u. , & beck , r. 2007 , , 461 , 455 briggs , d. s. 1995 , high fidelity deconvolution of moderately resolved sources , phd thesis , the new mexico institute of mining and technology , socorro , nm brisken , w. 2003 , evla memo 58 , using grasp8 to study the vla beam cenacchi , e. , dallacasa , d. , & orfei , a. 2006 , mem . suppl . , 10 , 77 clark , b. g. 1980 , , 89 , 377 condon , j. j. 1992 , , 30 , 575 condon , j. j. , cotton , w. d. , & broderick , j. j. 2002 , , 124 , 675 conway , j. e. , cornwell , t. j. , & wilkinson , p. n. 1990 , , 246 , 490 cornwell , t. j. 2008 , ieee j. of selected topics in signal proc . 2 , no . 5 , 793 cornwell , t. j. , golap , k. , & bhatnagar , s. 2008 , ieee j. of selected topics in signal proc . 2 , no . 5 , 647 dumke , m. , krause , m. , & wielebinski , r. , & klein , u. 1995 , , 302 , 691 ekers , r. d. , & sancisi , r. 1977 , , 54 , 973 ghosh , t. , et al . 2011 , arxiv:1112.0509v1 golla , g. 1999 , , 345 , 778 golla , g. , & hummel , e. 1994 , , 284 , 777 heald , g. h. , et al . 2009 , , 503 , 409 heesen , v. , beck , r. , krause , m. , & dettmar , r .- j . 2011 , , 535 , a79 hoopes , c. g. , walterbos , r. a. m. , & rand , r. j. 1999 , , 522 , 659 hummel , e. , & dettmar , r .- j . 1990 , , 236 , 33 hummel , e. , lesch , h. , wielebinski , r. , & schlickeiser , r. 1988 , , 197 , l29 hummel , e. , beck , r. , & dettmar , r .- j . 1991 , , 87 , 309 irwin , j. a. et al . 2011 , , 410 , 1423 irwin , j. a. et al . 2012 , , paper i klein , u. , wielebinski , r. , & beck , r. 1984 , , 133 , 19 krause , m. 2004 , in the magnetized interstellar medium , eds . b. uyaniker , w. reich , and r. wielebinski , copernicus gmbh , katlenburg - lindau . , 173 krause , m. 2009 , in magnetic fields in the universe ii : from laboratory and stars to the primordial universe , revista mexicana de astronoma y astrofsica ( serie de conferencias ) , eds . a. esquivel , j. franco , g. garcia - segura , e. m. de gouvela dal pino , a. lazarian , s. lizano , & a. raga , vol . 36 , 25 lee , s .- w . , irwin , j. a. , dettmar , r .- j . , cunningham , c. t. , golla , g. , & wang , q. d. 2001 , , 377 , 759 niklas , s. , klein , u. , & wielebinski , r. 1997 , , 322 , 19 pacholczyk , a. b. 1970 , radio astrophysics , freeman and co. , san francisco pohl , m. , & schlickeiser , r. 1990 , , 239 , 424 rand , r. j. , & van der hulst , j. m. 1993 , , 105 , 2098 rau , u. , & cornwell , t. j. 2011 , , 532 , a71 reich , p. , reich , w. , & testori , j. c. 2004 , in the magnetized insterstellar medium , ed . b. uyanikar , w. reich , & r. wielebinski , 63 - 68 . sault , r. j. , & wieringa , m. h. 1994 , , 108 , 585 , 448 , 123 simmons , j. f. l. , & stewart , b. g. , , 142 , 100 soida , m. , krause , m. , dettmar , r .- j . , & urbanik , m. 2011 , , 531 , 127 strong , a. w. , orlando , e. , & jaffe , t. r. 2011 , , 534 , a54 sukumar , s. , & velusamy , t. 1985 , , 212 , 367 vaillancourt , j. e. 2006 , , 118 , 1340 wang , q. d. , et al . 1995 , , 439 , 176 wang , q. d. , immler , s. , walterbos , r. , lauroesch , j. r. , & breitschwerdt , d. 2001 , , 555 , l99 werner , w. 1988 , , 201 , 1 array & c & c + no . antennas & 25 & 22 + ac central frequency ( ghz ) & 1.375 & 5.5 + bd central frequency ( ghz ) & 1.625 & 6.5 + spw bandwidth ( mhz ) & 32 & 128 + no . spws ( ac plus bd ) & 16 & 16 + total bandwidth ( ac plus bd)(ghz ) & 0.512 & 2.048 + no . channels / spw & 64 & 64 + total no . channels & 1024 & 1024 + channel separation ( mhz ) & 0.500 & 2.00 + integration time ( s ) & 10 & 10 + obs . time ( min ) & 30 & 75 + flux calibrator & 3c286 & 3c286 + zero - polarization calibrator & oq208 & oq208 + phase calibrator & j1221 + 2813 & j1221 + 2813 + lcccccc uv weighting & briggs rob = 0 & briggs rob = 2 & briggs rob = 2 & briggs rob = 0 & briggs rob = 2 & briggs rob = 0 + uv taper ( k@xmath83 ) & none & none & 5 & none & none & 10 + nterms & 3 & 3 & 3 & 3 & 3 & 3 + no . self - cals & 1 a&p & 1 a&p & 1 a&p & 1 a&p & 1 a&p & 1 a&p + & + figure label & fig . [ fig : l - band](a ) & fig . [ fig : l - band](b ) & fig . [ fig : l - band](c ) & fig . [ fig : c - band](a ) & fig . [ fig : c - band](b ) & fig . [ fig : c - band](c ) + synth . beam + ( @xmath58 , @xmath58 , @xmath8 ) & 11.16 , 10.28 , -175.0 & 16.69 , 14.77 , 34.5 & 29.90 , 26.08 , 48.9 & 2.71 , 2.63 , -50.9 & 4.07 , 3.80 , 27.5 & 8.60 , 8.52 , 41.5 + rms ( @xmath0jy beam@xmath1 ) & 28 & 45 & 100 & 4.8 & 5.5 & 8.0 + & + synth . beam + ( @xmath58 , @xmath58 , @xmath8 ) & 11.01 , 10.13 , -176.8 & 12.28 , 14.41 , 33.7 & 29.35 , 25.51 , at 48.3 & 2.71 , 2.63 , -50.8 & 4.07 x 3.80 at 27.5 & 13.15 , 11.45 , 27.2 + rms ( @xmath0jy beam@xmath1 ) & 27 & 22 & 23 & 4.7 & 3.5 & 7.0 +

we consider one - dimensional reaction - diffusion equations for a large class of spatially periodic nonlinearities including multistable ones and study the asymptotic behavior of solutions with heaviside type initial data . our analysis reveals some new dynamics where the profile of the propagation is not characterized by a single front , but by a layer of several fronts which we call a terrace . existence and convergence to such a terrace is proven by using an intersection number argument , without much relying on standard linear analysis . hence , on top of the peculiar phenomenon of propagation that our work highlights , several corollaries will follow on the existence and convergence to pulsating traveling fronts even for highly degenerate nonlinearities that have not been treated before . * key words*. multistable reaction - diffusion equation , periodic environment , long time behavior , propagating terrace , zero - number argument . * 2010 mathematical subject classification*.35k55 , 35c07 , 35b08 , 35b40

1203.6206

we consider in this work a cauchy problem for the following reaction - diffusion equation in one space dimension : @xmath0 supplemented with the initial condition @xmath1 here the function @xmath2 satisfies the periodicity condition @xmath3 for some @xmath4 . we will assume , throughout this paper , that there exists a positive and @xmath5-periodic stationary solution @xmath6 of : @xmath7 the function @xmath8 is also a stationary solution of the following auxiliary equation , the @xmath5-periodic counterpart of : @xmath9 it is obvious that any solution of is also a solution of . equation will later play an important auxiliary role in the analysis of . + our aim is to investigate the profile of solutions of connecting the two stationary states @xmath10 and @xmath8 . in particular , we will study the long time behavior of solutions with heaviside - type initial data . roughly speaking , our main results state that the solution will converge to what we call a propagating terrace , the meaning of which will be specified later . this result , in particular , implies the existence of pulsating traveling waves ( or a set of traveling waves ) under rather mild assumptions . + we now state our main assumptions . the first one is concerned with the attractiveness of @xmath8 with respect to at least one compactly supported initial data . our main theorems ( theorems [ heavip2 ] and [ heavip2-cv ] ) will only need this assumption : [ assumption - p1 ] there exists a solution @xmath11 of - with compactly supported initial data @xmath12 that converges locally uniformly to @xmath8 as @xmath13 . this assumption covers a wide variety of nonlinearities that include not only such standard ones as monostable , bistable or combustion nonlinearities , but also much more general and complex ones . for instance , it even allows an infinite number of stationary solutions between 0 and @xmath8 . in this paper , we will show that this rather weak condition is in fact sufficient for deriving our main results on the convergence to a propagating terrace . the next assumption guarantees that our propagating terrace consists of a single ( pulsating ) traveling wave . thus , under this additional assumption , our main results imply the existence of a pulsating traveling wave , as well as the convergence of solutions to this traveling wave ( theorem [ heavip1 ] ) : [ assumption - p2 ] there exists no @xmath5-periodic stationary solution @xmath14 with @xmath15 that is both isolated from below and stable from below with respect to . let us clarify the notions introduced in this assumption . a stationary solution @xmath14 of is said to be * isolated from below * ( resp . * above * ) if there exists no sequence of other stationary solutions converging to @xmath14 from below ( resp . above ) . a stationary solution @xmath14 is said to be * stable from below * ( resp . * above * ) with respect to equation if it is stable in the @xmath16 topology under nonpositive ( resp . nonnegative ) perturbations . otherwise , @xmath14 is called * unstable from below * ( resp . * above * ) . it is known that , if @xmath14 is isolated from below , then it is stable from below if and only if there exists a solution @xmath17 converging to @xmath14 as @xmath18 , and unstable from below if and only if there exists an ancient solution ( that is , a solution defined for all sufficiently negative @xmath19 ) @xmath17 converging to @xmath14 as @xmath20 ( see theorem 8 in @xcite ) . note that this additional assumption holds for a large class of standard nonlinearities including the following : there exists no @xmath5-periodic stationary solution @xmath14 satisfying @xmath21 for all @xmath22 . furthermore , @xmath10 is unstable from above . the stationary solution @xmath10 is stable from above with respect to , and @xmath8 is stable from below with respect to . furthermore , all other stationary solutions between 0 and @xmath8 are unstable . there exists a family of @xmath5-periodic stationary solutions @xmath23}$ ] that forms a continuum in @xmath24 and satisfies @xmath25 . furthermore , there exists no stationary solution @xmath14 satisfying @xmath26 for all @xmath27 . a classical example of the bistable nonlinearity is the allen - cahn nonlinearity @xmath28 , where @xmath29 , @xmath30 . an important subclass of the monostable nonlinearity is the kpp type nonlinearity , in which @xmath10 is assumed to be linearly unstable and @xmath31 is sublinear with respect to @xmath11 ; a typical example being @xmath32 , with @xmath33 . kpp type equations have been widely studied , even in the periodic setting , by numerous authors including @xcite . while most of those studies rely heavily on the linear instability of @xmath10 , our approach in the present paper largely avoid the need for linear analysis , allowing our results to be applicable even to strongly degenerate situations that have not been treated before . we now introduce some notions which will play a fundamental role in this paper . we begin with the following : [ def : steep ] let @xmath34 be two entire solutions of . we say that @xmath35 is * steeper than * @xmath36 if for any @xmath37 , @xmath38 and @xmath39 in @xmath40 such that @xmath41 , we have either @xmath42 here , by an entire solution " we mean a solution that is defined for all @xmath43 . the above property implies that the graph of the solution @xmath35 ( at any chosen time moment @xmath37 ) and that of the solution @xmath36 ( at any chosen time moment @xmath38 ) can intersect at most once unless they are identical , and that if they intersect at a single point , then @xmath44 is positive on the left - hand side of the intersection point , while negative on the right - hand side . note that , according to this definition , if the ranges of @xmath35 and @xmath36 are disjoint , then @xmath35 and @xmath36 are steeper than each other , since their graphs never intersect . [ def : puls ] given two distinct periodic stationary states @xmath45 and @xmath46 , by a * pulsating traveling wave solution * ( or * pulsating traveling front * ) of connecting @xmath45 to @xmath46 , we mean any entire solution @xmath11 satisfying , for some @xmath47 , @xmath48 for any @xmath22 and @xmath49 , along with the asymptotics @xmath50 where the convergence is understood to hold locally uniformly in the space variable . the ratio @xmath51 is called the * average speed * ( or simply the * speed * ) of this pulsating traveling wave . one can easily check that , for any @xmath52 , @xmath53 is a pulsating traveling wave connecting @xmath45 to @xmath46 with speed @xmath54 if and only if it can be written in the form @xmath55 , where @xmath56 satisfies @xmath57 along with the following equation that is equivalent to : @xmath58 let us recall some known results on traveling waves from the literature . in the case of spatially homogeneous problems , existence of traveling waves is well studied ( see for instance @xcite for a review of the area ) . more precisely , in the kpp case , there exists a continuum of admissible speeds @xmath59 , while in the bistable or combustion cases , the admissible speed is unique . stability and convergence to those traveling waves are also studied extensively . among other things , in the one - dimensional kpp case ( or , more generally , the monostable case ) , uchiyama @xcite , bramson @xcite and lau @xcite proved that solutions of the cauchy problem with compactly supported initial data converge to the traveling front with minimal speed as @xmath60 . in this case , the solution does not converge to the traveling wave with an asymptotic phase , but a phase drift of order @xmath61 occurs @xcite . similar results hold for multi - dimensional problems as long as the parameters of the equation are invariant in the direction of propagation @xcite . in the case of spatially periodic problems , the state of research is slightly behind , for obvious technical difficulties . nonetheless , in the kpp case , the existence of a continuum of admissible speeds is well established , as in the case of spatially homogeneous problems . it is also known that there is a close relation between the speed of a traveling wave @xmath62 and its decay rate as @xmath63 , at least under some assumptions on the linearized problem around @xmath10 ; the smaller the speed @xmath54 , the faster the decay ; hence steeper the front profile . convergence to those traveling waves was studied in @xcite in a periodic framework . more precisely , it has been shown that if the initial data has the same exponential decay as a given traveling wave as @xmath64 , then the solution of the cauchy problem converges to this traveling wave as @xmath18 . however , the case of very fast decaying initial data ( for instance , a heaviside or a compactly supported function ) has been left open up to now in the periodic framework , although the appearance of some phase drift of order @xmath61 has also been highlighted in @xcite . let us now come back to the main theme of the present paper a propagating terrace . as we mentioned earlier , a traveling wave is a special case of a propagating terrace , but the latter is a more suitable notion for describing typical frontal behaviors in equations of multistable nature . the aim of the present paper is to study properties of propagating terraces in a spatially periodic setting , thereby generalizing ( and improving ) some of the aforementioned results on pulsating traveling waves . [ def : terrace ] a * propagating terrace * connecting @xmath10 to @xmath8 is a pair of finite sequences @xmath65 and @xmath66 such that : * each @xmath67 is an @xmath5-periodic stationary solution of satisfying @xmath68 * for each @xmath69 , @xmath70 is a pulsating traveling wave solution of connecting @xmath67 to @xmath71 . * the speed @xmath72 of each @xmath70 satisfies @xmath73 . furthermore , a propagating terrace @xmath74 connecting @xmath10 to @xmath8 is said to be * minimal * if it also satisfies the following : * for any propagating terrace @xmath75 connecting @xmath10 to @xmath8 , one has that @xmath76 * for each @xmath77 , the traveling wave @xmath70 is steeper than any other traveling wave connecting @xmath67 to @xmath71 . roughly speaking , a propagating terrace can be pictured as a layer of several traveling fronts going at various speeds , the lower the faster ( figure [ fig : terrace ] ) . the aim of the present paper is to show that the solution of with heaviside - type initial data will converge to a minimal propagating terrace , as illustrated in figure [ fig : terrace - like ] . in some standard problems such as the kpp and the bistable equations , the terrace actually consists of a single front ( that is , @xmath78 ) , which means that the solution will eventually look like a single traveling wave ; see theorem [ heavip1 ] . however , in more general equations , one can not expect such simple dynamics , and this is where the notion of terrace plays a fundamental role . the existence of a multi - step terrace has been known in the spatially homogeneous case ( where @xmath79 ) . let us give a simple example . consider @xmath31 as in figure [ fig : f_terrace ] ( left ) , which is kpp on @xmath80 $ ] , and bistable on @xmath81 $ ] . the speed of the upper part of the solution is bounded from above by the speed , say @xmath54 , of the traveling wave for the bistable nonlinearity @xmath82}$ ] . on the other hand , the lower part of the solution is pushed from behind by a spreading front for the kpp nonlinearity @xmath83}$ ] , whose speed is known to approach @xmath84 . therefore , if @xmath85 , the upper and lower parts of the solution necessarily move at two distinct speeds . [ fig : fterrace ] that produce a multi - step terrace , title="fig:",scaledwidth=90.0% ] another example was exhibited by fife and mcleod in @xcite , where they considered a specific case of a non - degenerate tristable nonlinearity , that is , when @xmath86}$ ] and @xmath87}$ ] are both bistable for some @xmath88 , and that @xmath89 , @xmath90 and @xmath91 are all strictly negative ; see figure [ fig : f_terrace ] ( right ) . they showed that if the speed of the upper bistable part is smaller than the speed of the lower bistable part , then there does not exist any single front connecting @xmath10 to @xmath92 . furthermore , some solutions of the cauchy problem , in particular for heaviside type initial data , converge to a combination of those two fronts . this may be seen as an early study of a propagating terrace for some very specific examples . although the method in @xcite was expected to hold for homogeneous nonlinearities composed of a finite number of bistable parts , it relied strongly on the particular shape of @xmath31 , and on the non - degeneracy of the equilibria . this means that they needed some important a priori knowledge on the shape of the nonlinearity , which we do not need in the present paper . more importantly , what makes our work different from those early observations is that we are not simply giving examples of propagating terraces but are establishing the ubiquity of such terraces for large classes of reaction nonlinearities , thus showing that the notion of propagating terrace is fundamental for studying the dynamics of fronts in general reaction - diffusion equations . we consider solutions of - whose initial data are given in the form @xmath93 where @xmath94 is any constant , and @xmath95 denotes the heaviside function , which is defined by @xmath96 hereafter , for each @xmath97 , we denote by @xmath98 such solutions . we will prove that @xmath99 converges in some sense to a minimal propagating terrace as @xmath60 . [ heavip2 ] let assumption [ assumption - p1 ] hold . then there exists a propagating terrace @xmath100 that is minimal in the sense of definition [ def : terrace ] . such a minimal propagating terrace is unique , in the sense that any minimal propagating terrace shares the same @xmath101 and that @xmath70 is unique up to time - shift for each @xmath102 . moreover , it satisfies : a. for any @xmath103 , the @xmath5-periodic stationary solution @xmath67 is isolated and stable from below with respect to . b. all the @xmath67 and @xmath70 are steeper than any other entire solution of . the existence of a minimal terrace as stated in the above theorem gives various useful information about the qualitative properties of the equation . for example , statement @xmath104 implies , in particular , that there exists no traveling wave that intersects any of the @xmath67 . note that , in the spatially homogeneous case ( namely , @xmath79 ) , the stability of @xmath67 in statement @xmath105 implies that each @xmath67 is a constant ; hence the terrace consists of flat steps . we now state our convergence result : [ heavip2-cv ] let assumption [ assumption - p1 ] hold . then for any @xmath106 , the solution @xmath107 converges as @xmath18 to the minimal propagating terrace @xmath100 in the following sense : a. there exist functions @xmath108 with @xmath109 as @xmath18 such that @xmath110 locally uniformly on @xmath40 , @xmath72 being the speed of @xmath70 . b. for any @xmath111 , there exists @xmath112 such that , for any @xmath113 , @xmath114 ) } \leq \delta \ \ \hbox{as}\ t\to + \infty , \ ] ] together with @xmath115 ) } \leq \delta \ \ \hbox{as}\ t\to + \infty , \ ] ] @xmath116 roughly speaking , statements @xmath105 and @xmath104 of the above theorem describe , respectively , the ascending part and the stationary part of the terrace , the latter being flat if @xmath79 , as mentioned above . it should be noted that , under assumption [ assumption - p1 ] , there may exist an infinite number of isolated stationary solutions between @xmath10 and @xmath8 , but our theorem [ heavip2 ] states that only a finite number of layers appear in the limiting terrace . the solution seems to ignore excessive complexity of such nonlinearities . the proofs of theorems [ heavip2 ] and [ heavip2-cv ] are , in a sense , one and the same . in fact , by first showing that the steepness of the heaviside type initial data implies that the limiting profile of the solution is steeper than any other entire solution , we will then use this fact to prove the convergence of the solution to a minimal terrace without assuming the existence of a terrace . hence it automatically implies the existence of a minimal terrace . in the special case where assumption [ assumption - p2 ] also holds , the above two theorems reduce to the following result on pulsating traveling waves . thus it gives a new and highly original proof for the existence of pulsating traveling waves . [ heavip1 ] let assumptions [ assumption - p1 ] and [ assumption - p2 ] hold . then there exists a pulsating traveling wave @xmath117 connecting @xmath10 to @xmath8 with speed @xmath118 that is steeper than any other entire solution between @xmath10 and @xmath8 . furthermore , for any @xmath106 , there exists a function @xmath119 with @xmath120 as @xmath121 such that @xmath122 let us make some comments on theorem [ heavip1 ] , which is a special case of the previous two theorems . as regards the existence part , there have been earlier studies of the existence of pulsating traveling waves for monostable ( possibly degenerate ) and combustion cases @xcite , as well as for some periodic bistable case @xcite . in contrast to those earlier results , which are derived by various different methods depending on the type of nonlinearities , our theorem relies on a new , unified and rather straightforward proof , thus avoiding to deal directly with the particular features and difficulties of each case . as mentioned earlier , the convergence result for fast decaying initial data such as our heaviside type ones was previously known only in the homogeneous setting . indeed , even in the standard kpp case with spatially periodic coefficients , almost all the convergence results in the literature are concerned with solutions whose initial data have roughly the same decay rate as one of the traveling waves near @xmath123 . only very recently , there are some works in progress that are trying to deal with some fast decaying initial data @xcite . note that , though our results cover a large class of equations , they are concerned only with a specific type of inital data . however , by analogy with the homogeneous case ( see for instance the proofs in @xcite ) , we expect that similar convergence results hold for more general initial data , such as compactly supported ones . this is a topic of particular relevance from an applied point of view , and we will give a partial answer to this question in a forthcoming paper @xcite . [ [ plan - of - the - paper ] ] plan of the paper + + + + + + + + + + + + + + + + + our paper is organized as follows . in section [ sec : preliminaries ] , we will present some preliminaries . as our proofs largely rely on the so - called intersection number ( or the zero number ) argument , we will first give its precise definition and basic properties . we will then use this method to prove our fundamental lemma ( lemma [ lem : omega - steep ] ) , which roughly states the following : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ * fundamental lemma : * any function that appears in the @xmath124-limit set of @xmath99 is steeper than any other entire solution , where @xmath125 is the solution of for the initial data _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ one immediate consequence of the above fundamental lemma is that any two elements of the @xmath124-limit set are steeper than each other . this means that they are either identical ( up to time shift ) or strictly ordered , that is , one is above or below the other . this observation is important both for establishing convergence results and for the construction of a multi - step terrace . here we note that the definition of the @xmath124-limit set in this paper is slightly different from the standard one , in that we consider arbitrary spatial translations while taking the limit as @xmath126 ; see definition [ def : omegalimit ] . the reason for adopting this slightly non - standard defintion is that , since each step of the terrace moves at a different speed , we can not capture the asymptotic profile of the solution in a single frame . a multi - speed observation is unavoidable in the case of a multi - step terrace . in the last part of section [ sec : preliminaries ] , we will prove a lemma on the spreading speed of @xmath99 , which will be used repeatedly in later sections but is also of independent interest in its own right . in section [ sec : lemmas ] , we will use our fundamental lemma to prove the convergence of solutions with heaviside type initial data to a unique limit around any level set . the entire solution thereby constructed possesses , in some sense , some qualitative properties of traveling fronts , such as monotonicity in time . the same result could in fact be shown with a similar argument in a more general setting without the periodicity assumption . we will then show in section [ sec : periodic ] that this limit is a pulsating traveling wave connecting some pair of @xmath5-periodic stationary solutions @xmath127 that lie between @xmath10 and @xmath8 . once again the above - mentioned fundamental lemma plays a key role in deriving this result . this leads to construction of a multi - step minimal terrace inductively , as described in theorems [ heavip2 ] and [ heavip2-cv ] . in the special case where assumtion [ assumption - p2 ] holds , we have @xmath128 , thus the terrace is a single traveling wave . as the existence of any traveling wave is not a priori assumed , this leads to both the existence and the convergence results in theorem [ heavip1 ] . our proof of the main results relies strongly on a zero - number argument . the application of this argument or the sturmian principle " to the convergence proof in semilinear parabolic equations first appeared in @xcite . but what makes the present paper different from earlier work is that we employ the zero - number argument to prove not only the convergence but also the existence of the target objects , namely the terrace and pulsating traveling waves . in this paper , besides the standard zero - number @xmath129 $ ] , we introduce a related notion @xmath130 $ ] , which turns out to be exceedingly useful in establishing our fundamental lemma . [ def : zeros ] for any real - valued function @xmath131 on @xmath40 , we define : * @xmath132 $ ] is the number of sign changes of @xmath131 , namely the supremum over all @xmath133 such that there exist real numbers @xmath134 with @xmath135 we set @xmath136=-1 $ ] if @xmath137 . * @xmath138 $ ] , which is defined when @xmath139<\infty$ ] , is the word consisting of @xmath140 and @xmath141 that describes the signs of @xmath142 , where @xmath143 is the sequence that appears in the definition of @xmath136 $ ] with maximal @xmath102 . we set @xmath144 = [ \ ] $ ] , the empty word . if @xmath131 is a smooth function having only simple zeros on @xmath40 , then @xmath145 $ ] coincides with the number of zeros of @xmath131 . for example , @xmath146=2,\ \ sgn[x^2 - 1]=[+ - + ] .\ ] ] by definition , the length of the word @xmath147 $ ] is equal to @xmath136 + 1 $ ] . it is also clear that @xmath136=0 $ ] if and only if either @xmath148 or @xmath149 . the former implies @xmath147=[+]$ ] , and the latter @xmath147=[-]$ ] . if @xmath150 are two words consisting of @xmath140 and @xmath141 , we write @xmath151 ( or , equivalently , @xmath152 ) if @xmath153 is a subword of @xmath154 . for example , @xmath155 \rhd b\ \ \hbox{for}\ b=[+\,-],\,[+],\,[-],\,[\ ] \quad\ \hbox{but not}\ \ [ + \,-]\rhd [ -\,+].\ ] ] let us recall some properties of @xmath156 and @xmath157 : [ lem : z1 ] let @xmath158 be a bounded solution of a parabolic equation of the form @xmath159 where @xmath54 is bounded . then , for each @xmath160 , the zeros of @xmath161 do not accumulate in @xmath40 . furthermore , 1 . @xmath162 $ ] and @xmath163 $ ] are nonincreasing in @xmath19 , that is , for any @xmath164 , @xmath165\geq z\left[w(t',\cdot)\right],\quad\ sgn\left[w(t,\cdot)\right ] \rhd sgn\left[w(t',\cdot)\right];\ ] ] here the assertion remains true even for @xmath166 if @xmath131 can be extended to a continuous function on @xmath167 ; 2 . if @xmath168 for some @xmath169 and @xmath170 , then @xmath171 - 2 \geq z\left [ w(s,\cdot ) \right]\geq 0 \ \ \ \hbox{for any}\ t \in ( t_1 , t ' ) \ \hbox{and}\ s \in ( t ' , t_2)\ ] ] whenever @xmath172 < \infty$ ] . the second inequality of statement @xmath104 above implies that , for any @xmath173 , the function @xmath174 does not vanish entirely on @xmath40 unless @xmath137 on @xmath175 . statement @xmath104 is due to @xcite , where this result is proved by using similarity variables and expansion by hermitian polynomials . though @xcite deals with only equations on bounded intervals , the result can easily be extended to @xmath40 by applying the maximun principle near @xmath176 ; see @xcite . the statement @xmath105 for @xmath136 $ ] follows from @xmath104 , at least when the domain is a bounded interval , but it can be shown more directly by a combination of the maximum principle and a topological argument similar to the jordan curve theorem ( which is a more standard way to prove this statement ) . in fact , this direct proof proves the assertion for @xmath147 $ ] , from which the assertion for @xmath136 $ ] follows automatically ; see , for example , @xcite for a similar argument . one can also check that @xmath156 is semi - continuous with respect to the pointwise convergence , that is : [ lem : z - lim ] let @xmath177 be a sequence of functions converging to @xmath131 pointwise on @xmath40 . then @xmath178 \leq \liminf_{n \rightarrow \infty } z\left [ w_n \right],\ ] ] @xmath179 \lhd \liminf_{n \rightarrow \infty } sgn [ w_n].\ ] ] combining the above two lemmas , we obtain the following lemma : [ lem : z(u1-u2 ) ] let @xmath35 and @xmath36 be solutions of such that the initial data @xmath180 is a piecewise continuous bouded function on @xmath40 , while @xmath181 is bounded and continuous on @xmath40 . assume also that @xmath182 changes sign at most finitely many times on @xmath40 . then 1 . for any @xmath183 , @xmath184 & \geq z \left [ u_1 ( t',\cdot ) - u_2 ( t',\cdot)\right],\\ sgn \left [ u_1 ( t,\cdot ) - u_2 ( t,\cdot)\right ] & \rhd sgn \left [ u_1 ( t',\cdot ) - u_2 ( t',\cdot)\right ] ; \end{split}\ ] ] 2 . if , for some @xmath185 , the graph of @xmath186 and that of @xmath187 are tangential at some point in @xmath40 , and if @xmath188 , then for any @xmath189 with @xmath190 , @xmath191 -2 \geq z \left [ u_1 ( s,\cdot ) - u_2 ( s,\cdot)\right]\geq 0.\ ] ] the same conclusion holds if @xmath192 are entire solutions of , in which case @xmath193 is arbitrary and @xmath194 . the function @xmath195 satisfies an equation of the form on @xmath196 with @xmath197 being bounded . thus the conclusion of the lemma follows from lemma [ lem : z1 ] except for with @xmath198 . moreover statement with @xmath198 also follows from lemma [ lem : z1 ] if @xmath180 and @xmath181 are both continuous . in the general case where @xmath180 is only piecewise continuous , we approximate @xmath35 by a sequence of solutions of , say @xmath199 , whose initial data @xmath200 are continuous and satisfy * @xmath201 and @xmath202 pointwise on @xmath40 ; * @xmath203=sgn[u_1(0,\cdot ) -u_2(0,\cdot)]$ ] for @xmath204 . then we have , for each @xmath185 , @xmath205 \rhd sgn \left [ u_{1,n } ( t',\cdot ) - u_2 ( t',\cdot)\right].\ ] ] letting @xmath206 and applying lemma [ lem : z - lim ] , we obtain the desired conclusion . + the following corollary will be used repeatedly later : [ cor : steeper ] let @xmath207 be two entire solutions of , and assume that @xmath208\lhd [ + \;- ] \quad \hbox{for any}\ \ t_1 , t_2 \in{\mathbb{r}}.\ ] ] then @xmath209 is steeper than @xmath210 in the sense of definition [ def : steep ] . fix @xmath211 arbitrarily . from the assumption we see that @xmath212 \leq 1 \quad \ \hbox{for all}\ \ t\in { \mathbb{r}}.\ ] ] if @xmath213 , then by lemma [ lem : z(u1-u2 ) ] ( ii ) , the function @xmath214 has at most one zero on @xmath40 , and that this zero is simple . let @xmath39 be such a zero ; that is , @xmath215 . then the simplicity of this zero and the sign property @xmath216\lhd [ + \;-]$ ] imply that @xmath217 . this proves that @xmath209 is steeper than @xmath210 . the following definition of the @xmath124-limit set of a solution @xmath11 is slightly different from the standard one , as we add arbitrary spatial translations while taking the long - time limit . the reason for adopting this definition is that , since each step of the terrace moves at a different speed , we need multi - speed observations in order to fully capture the asymptotic profile of the solution . [ def : omegalimit ] let @xmath53 be any bounded solution of cauchy problem - . we call @xmath218 an * @xmath219-limit orbit * of @xmath11 if there exist two sequences @xmath220 and @xmath221 such that @xmath222 by parabolic estimates , the above convergence takes place in @xmath223 in @xmath224 and @xmath225 in @xmath19 . hence one can easily check that any @xmath124-limit orbit of @xmath11 is an entire solution of . moreover , if @xmath218 is an @xmath124-limit orbit of @xmath11 , then so is @xmath226 for any @xmath227 and @xmath228 . let us now state a fundamental lemma that will be used repeatedly throughout our paper : [ lem : omega - steep ] let @xmath94 and let @xmath209 be any @xmath124-limit orbit of @xmath107 . then @xmath209 is steeper than any entire solution of in the sense of definition [ def : steep ] , provided that this entire solution lies between @xmath10 and @xmath8 . fix @xmath94 , and let the sequences @xmath229 and @xmath221 be such that @xmath230 locally uniformly as @xmath231 . by standard parabolic estimates , the convergence in fact holds in @xmath232 . let @xmath233 be any entire solution lying between 0 and @xmath8 . since @xmath234 , we have @xmath235 for @xmath236 and @xmath237 for @xmath238 . consequently , for any @xmath239 and @xmath240 , @xmath241=1 \ \ sgn\left [ \widehat{u } ( 0,\cdot;a ) - v(\tau - t_j,\cdot ) \right]=[+ \ ; -].\ ] ] it follows from lemma [ lem : z(u1-u2 ) ] that , for all @xmath242 and @xmath243 , @xmath244 \leq 1,\ ] ] @xmath245 \lhd [ + \ ; -].\ ] ] passing to the limit as @xmath246 , we get @xmath247 \lhd [ + \ ; - ] \quad\ \hbox{for any}\ \ t,\tau\in{\mathbb{r}}.\ ] ] hence , by corollary [ cor : steeper ] , @xmath209 is steeper than @xmath233 in the sense of definition [ def : steep ] . before going to the proof of our main results , we investigate some spreading property of solutions of . the result below , which can be of independent interest , will be used repeatedly later . recall that @xmath107 is the solution of with initial data @xmath248 . [ spread ] let assumption [ assumption - p1 ] be satisfied . then there exist constants @xmath249 that do not depend on @xmath250 , such that * for each @xmath251 , one has : @xmath252 ; * for each @xmath253 one has @xmath254 note that from the @xmath225-regularity and the periodicity of @xmath31 , there exists @xmath255 such that for any @xmath22 and @xmath256 , we have that @xmath257 . now define @xmath258 which is a solution of the linear homogeneous equation @xmath259 then @xmath260 is a supersolution of in the range @xmath256 . since @xmath261 for all @xmath262 , while @xmath263 for all @xmath264 , it follows from the comparison principle that for all @xmath265 and @xmath27 , @xmath266 therefore , for any speed @xmath267 , @xmath268 uniformly with respect to @xmath269 as @xmath121 . let us now find a positive lower bound for the spreading speed . let @xmath270 be the compactly supported function given in assumption [ assumption - p1 ] . this means that the solution @xmath271 of the cauchy problem - with initial data @xmath272 converges locally uniformly to @xmath8 as @xmath13 . thanks to the periodicity of , one can assume without loss of generality that @xmath273\ ] ] for some @xmath112 . since @xmath274 as @xmath60 , there exists @xmath47 such that @xmath275 by the comparison principle , it follows that @xmath276 by induction , we obtain that for all @xmath133 , @xmath277 since @xmath278 for all @xmath279 $ ] , we have that @xmath280 applying the comparison principle , one gets for all @xmath281 and @xmath22 that @xmath282 hence , for all @xmath27 , @xmath283 therefore , from assumption [ assumption - p1 ] , we have that for any @xmath133 , @xmath284 where the convergence holds as @xmath285 , uniformly with respect to @xmath286 and @xmath287 $ ] . let us now define @xmath288 and choose any @xmath54 with @xmath289 . denote by @xmath290 the ceiling function of @xmath291 , that is , the least integer not smaller than @xmath291 . then for any @xmath265 , let @xmath292 as @xmath293 , one can easily check that @xmath294 as @xmath18 . thus , @xmath295 and , since @xmath296 and @xmath297 for all @xmath265 , @xmath298 which concludes the proof of lemma [ spread ] . in this section , we aim to prove an important lemma on the convergence of the solutions of the cauchy problem with some shifted heaviside type initial data around a given level set . [ heavi1lemma ] let assumption [ assumption - p1 ] be satisfied . let @xmath299 be given . for any @xmath300 and @xmath301 , let us define @xmath302 then the following limit exists for the topology of @xmath303 @xmath304 function @xmath305 is an entire solution of that is steeper than any other entire solution . furthermore , the following alternative holds true : either it is a stationary solution , or @xmath306 for all @xmath307 . the former assertion is impossible for each @xmath308 close enough to @xmath309 . this lemma states that if we look at some well chosen level set for shifted initial data , the profile of the solution locally converges to a monotonically increasing entire solution @xmath310 of , which connects two stationary solutions . in the sequel we will show that @xmath310 is a pulsating traveling wave . a similar result holds in a nonperiodic heterogeneous framework , where one could show that @xmath311 converges as @xmath312 to the steepest entire solution of taking the value @xmath308 at the point @xmath313 . however , as the notion of traveling wave is not clear in general , we chose to restrict ourselves to the more standard periodic setting . + the proof of this result is split into three parts . before we begin the proof of this lemma , let us make the following remark , explaining the choice of such shifted initial data . [ rem : period ] notice that one has @xmath314 this implies that for any @xmath315 and @xmath133 : @xmath316 now since the initial data @xmath317 is increasing with respect to @xmath250 , the comparison principle provides that for each given @xmath318 , the maps @xmath319 and @xmath320 are nondecreasing . let @xmath299 and @xmath321 be given . then function @xmath310 defined in lemma [ heavi1lemma ] rewrites as @xmath322 as it is clear that @xmath323 as @xmath324 , the above computations explain the choice of the shifts of the initial data in order to study the large time behavior of the solution of the cauchy problem . let us first recall that assumption [ assumption - p1 ] holds true . let @xmath299 and @xmath300 be given . let us first note that for any @xmath301 , the following quantity exists and is finite @xmath326 indeed , let @xmath102 be some large enough given integer such that @xmath327 $ ] , wherein @xmath270 is the compactly supported function arising in assumption [ assumption - p1 ] . recall that the corresponding solution of the cauchy problem converges locally uniformly to @xmath8 . since @xmath328 for all @xmath262 , one has @xmath329 therefore , @xmath107 converges locally uniformly with respect to @xmath224 to @xmath8 as @xmath18 . in particular , this leads us to @xmath330 as @xmath18 . on the other hand , since @xmath331 , we have @xmath332 and the existence of @xmath333 immediately follows . + we now aim to prove that the following limit exists for all @xmath334 : @xmath335 to do so , let us first notice that from parabolic estimates , the family of functions @xmath336 is uniformly bounded along with their derivatives . therefore it is relatively compact for the topology of @xmath337 with respect to @xmath338 . let @xmath339 be a given sequence such that @xmath340 as @xmath324 , and such that the following limit holds true : @xmath341 as @xmath324 , wherein @xmath310 is some function and where the convergence holds in @xmath232 . up to a subsequence , one may assume that @xmath342 in @xmath343 . then , from remark [ rem : period ] , one can check that @xmath310 is an @xmath124-limit orbit of @xmath344 , and it therefore follows from lemma [ lem : omega - steep ] that it is steeper than any other entire solution in the sense of definition [ def : steep ] . recalling definition [ def : steep ] , there is a unique entire solution @xmath310 of that is steeper than any other entire solution , and such that @xmath345 . it follows that @xmath310 does not depend on the choice of the sequence @xmath346 . finally the relative compactness of the family of functions @xmath336 completes the proof of the existence of @xmath347 together with the convergence for the topology of @xmath348 . in order to complete the proof of lemma [ heavi1lemma ] , it remains to prove the alternative part . to do so , we will show that function @xmath310 is nondecreasing with respect to time . we will more precisely prove that for any @xmath49 , @xmath349 does not change sign . to prove this statement , we will argue by contradiction by assuming that for some given @xmath350 , there exist @xmath351 and @xmath352 such that @xmath353 it is then clear that for any @xmath354 small enough , one has @xmath355 \geq 1.\ ] ] besides , recall that @xmath310 is an @xmath124-limit orbit of @xmath99 , and so is @xmath356 for any @xmath227 . therefore , they are steeper than each other and it immediately follows that @xmath357 for each @xmath354 small enough . hence , @xmath358 a contradiction together with . this implies that for any @xmath43 , one has @xmath359 = [ \ ; ] \mbox { or } [ + ] \mbox { or } [ -].\ ] ] let us denote by @xmath360 . it is an entire solution of the linear parabolic equation @xmath361 we infer from and the strong maximum principle that either @xmath362 , either @xmath363 or @xmath364 . next due to the definition of @xmath325 , one has @xmath365 . this completes the proof of the alternative part of lemma [ heavi1lemma ] . to conclude the proof of lemma [ heavi1lemma ] , let us show that when @xmath308 is chosen close enough to @xmath309 then @xmath310 can not be a stationary solution of . to show that let us first notice that due to assumption [ assumption - p1 ] , the stationary solution @xmath8 is isolated with respect to the other stationary solutions . therefore , one can choose @xmath308 close enough to @xmath309 so that there is no stationary solution @xmath14 with @xmath366 . then due to assumption [ assumption - p1 ] , @xmath310 is not a stationary solution and it converges to @xmath8 as @xmath18 . this completes the proof of the lemma . the aim of this section is to prove the convergence of the solutions to a propagating terrace . since only small differences will arise depending on whether assumption [ assumption - p2 ] holds or not , we give in this section a common proof for both theorem [ heavip1 ] and theorem [ heavip2 ] . we will explicitely write down whenever we use assumption [ assumption - p2 ] . in this section , we will first show that the functions @xmath367 , constructed in the previous sections , are either traveling waves or stationary solutions . using some well chosen values of @xmath308 , we will then be able to construct , by using iterative arguments , the minimal propagating terrace describing the long time behavior of the solution @xmath99 of @xmath368 with an heaviside - type initial data , as stated in theorem [ heavip2 ] . lastly , we will prove that it satisfies all the required statements . the proof of this result relies on some properties of the sequence @xmath372 . it is split into two parts . let us first assume that there exists some subsequence @xmath373 converging to some @xmath374 . then we obtain that @xmath375 moreover , lemma [ heavi1lemma ] provides that @xmath376 and therefore it converges as @xmath377 to two periodic stationary solutions @xmath378 ( the periodicity follows from the above equality ) . if the two functions @xmath379 and @xmath380 are distinct , then @xmath310 is a pulsating traveling wave . if they are identically equal , then @xmath310 is a periodic stationary solution . furthermore , it is positive since @xmath381 and the strong maximum principle . + let us now consider the case when no subsequence of @xmath382 converges to some positive constant and let us show that @xmath310 is stationary . due to remark [ rem : period ] , it is clear that for all @xmath133 , @xmath383 . on the other hand , it follows from the spreading speed property provided by lemma [ spread ] that @xmath384 therefore , since no subsequence of @xmath382 converge to some positive constant , implies that one can find two subsequences converging respectively to @xmath10 and @xmath385 . by considering a subsequence converging to @xmath10 , the same computations as above with @xmath386 lead us to @xmath387 and function @xmath310 is @xmath5-periodic with respect to the space variable for all time . in order to show that @xmath310 is stationary , let us argue by contradiction by assuming that @xmath310 is not stationary . then using lemma [ heavi1lemma ] , one has that @xmath388 thanks to the @xmath389 convergence of @xmath390 to @xmath310 as @xmath312 , there exists some @xmath111 such that for any @xmath391 and @xmath250 large enough , one has @xmath392 on the other hand , using , for any @xmath393 , the following holds true for any @xmath250 large enough @xmath394 then one gets @xmath395 by choosing @xmath396 small enough , we conclude that @xmath397 , thus @xmath398 for any @xmath250 large enough . in particular , one gets that the sequence @xmath372 is bounded , which contradicts the existence of a subsequence going to @xmath385 . this completes the proof of the result . + in this subsection , we have proven that the limit @xmath310 is either a periodic positive stationary solution or a pulsating traveling wave . in the monostable case , there is no periodic positive stationary solution between @xmath10 and @xmath8 , so that @xmath310 is always a pulsating traveling wave connecting @xmath10 to @xmath8 , which already gives part @xmath105 of theorem [ heavip1 ] . together with assumption [ assumption - p1 ] , it is clear that @xmath8 is isolated , so that one can choose @xmath308 close enough to @xmath309 , so that function @xmath310 is a pulsating traveling wave connecting some periodic stationary solution @xmath45 to @xmath8 . [ cvtaukspeed ] note that it follows from the above proof as well as the uniqueness of the speed , that in the case where @xmath310 is a pulsating traveling wave , then the whole sequence @xmath399 converges to @xmath400 where @xmath54 is the speed of @xmath310 . this will be used later in the paper . we now aim to construct a terrace composed of pulsating fronts . we will proceed by iteration to construct such a terrace . let us first notice that , as mentioned above , by choosing @xmath308 close enough to @xmath309 , one can find a wave @xmath401 connecting some periodic stationary solution @xmath402 to @xmath8 . this gives us the first step of our iterative argument which is related to the following claim : then @xmath67 is isolated from below and stable from below with respect to . furthermore , there exists some @xmath407 such that @xmath408 is a pulsating traveling wave connecting some stationary periodic solution @xmath409 to @xmath67 . let @xmath233 be an entire solution of such that @xmath412 and that @xmath413 for some @xmath414 . we know from lemma [ lem : omega - steep ] that @xmath415 is steeper than any other entire solution between @xmath10 and @xmath8 , thus for any @xmath416 and @xmath19 in @xmath40 , @xmath417 \leq 1,\ ] ] @xmath418 \lhd [ + \ ; -].\ ] ] passing to the limit as @xmath419 , one gets , for all @xmath49 , @xmath420 \leq 1,\ ] ] @xmath421 \lhd [ + \ ; -].\ ] ] this implies , by corollary [ cor : steeper ] , that @xmath67 is steeper than @xmath233 . since @xmath233 is arbitrary , in particular , @xmath67 is steeper than @xmath422 . on the other hand , we know by lemma [ lem : omega - steep ] that @xmath422 is steeper than @xmath67 . thus these two functions are steeper than each other . furthermore , neither lies strictly above or below the other since @xmath423 . thus we conclude that @xmath424 , which completes the proof of claim [ walphapk ] . [ supersolpk ] let @xmath425 be a given function satisfying @xmath426 and let @xmath427 be a nondecreasing function moving with average speed @xmath428 ( where @xmath429 the minimal speed of spreading of @xmath99 provided by lemma [ spread ] ) . assume that @xmath430 @xmath431 then there exists a sequence @xmath220 such that for any @xmath269 , @xmath432 let us look at the intersection of @xmath433 and @xmath434 for any @xmath242 . note that @xmath435 is a super - solution for on the domain @xmath436 and , moreover , that @xmath437 on the half - space @xmath438 for each @xmath242 large enough . since @xmath427 moves with the average speed @xmath54 smaller than the minimal spreading speed @xmath429 of @xmath99 , one has that for any @xmath242 , @xmath439 as @xmath18 . thus , for any @xmath440 large enough , there exists some minimal time @xmath441 such that @xmath442 since @xmath435 is a super - solution on the domain @xmath443 , one can check that @xmath444 one can easily check that @xmath220 as @xmath445 . therefore , by standard parabolic estimates and possibly up a subsequence , one may assume that @xmath446 in @xmath447 and that @xmath448 converges as @xmath246 to some @xmath124-limit @xmath449 that satisfies @xmath450 since @xmath449 is steeper than @xmath67 , and conversely from claim [ walphapk ] , it follows that @xmath451 . therefore , we get that @xmath452 for any @xmath269 and the result follows . assume by contradiction that there exists some sequence @xmath453 of periodic stationary solutions such that @xmath454 as @xmath246 and @xmath455 for any @xmath242 . using standard elliptic estimates , one can easily show that the convergence holds uniformly in @xmath456 . the first statement @xmath105 follows from the following formula taken from @xcite ( see proposition 7.1 ) , which is adapted from nadin in @xcite : @xmath461 where @xmath462 is the functional defined by @xmath463 next @xmath104 follows from the fact that @xmath67 is an accumulation point of periodic stationary solutions . let us now construct some super - solution of . consider the function @xmath233 defined by @xmath464 wherein @xmath465 ( @xmath429 being the minimal speed of spreading of @xmath99 , provided by lemma [ spread ] ) , @xmath466 while @xmath467 is a solution of . it is clear that there exists some increasing map @xmath468 moving with the average speed @xmath54 and such that @xmath469 @xmath470 let us now define @xmath471 and compute on this set the following quantity : @xmath472 where the last inequality holds for any @xmath440 large enough and any @xmath473 small enough , on the domain @xmath474 . next claim [ supersolpk ] applies and provides the existence of a sequence @xmath220 such that @xmath475 as @xmath246 and for any @xmath269 . on the other hand , from the definition of @xmath233 , there exists some constant @xmath476 such that for any @xmath49 : @xmath477 this contradicts and we conclude that @xmath67 is isolated from below with respect to . assume first that @xmath67 is linearly unstable , that is @xmath478 where @xmath479 is defined as in . then , proceeding as in the well - known monostable case ( see for instance @xcite ) , one can find a stationary super - solution of the form @xmath480 , with @xmath481 is small enough and wherein @xmath482 is a principal eigenfunction of the following problem : @xmath483 one can check , using the regularity of @xmath31 and the fact that @xmath484 as @xmath485 , that @xmath233 is a super - solution of . as before , one can then apply claim [ supersolpk ] to reach a contradiction in this case . assume now that @xmath67 is not linearly unstable , namely @xmath460 . since @xmath67 is unstable from below with respect to , we also know that there exists some entire solution @xmath486 , decreasing in time and periodic with respect to the space variable , that belongs to the unstable set of @xmath67 in the downward direction ( we refer to @xcite ) , that is , such that @xmath487 @xmath488 let @xmath489 be a sequence converging to @xmath490 as @xmath246 , and such that the @xmath5-periodic function @xmath491 satisfies @xmath492 similarly as above , we construct a super - solution crossing @xmath67 and use claim [ supersolpk ] to reach a contradiction . let us consider the function @xmath493 wherein @xmath428 and @xmath429 is the minimal speed of spreading of @xmath99 ) , @xmath466 , while @xmath467 is a solution of . as before , one can check that function @xmath131 satisfies the hypotheses of claim [ supersolpk ] and stays away from below to @xmath67 as @xmath18 . we again reach a contradiction . [ [ step-3-convergence - to - a - pulsating - traveling - wave - for - alpha - p_k - x_0 ] ] step 3 : convergence to a pulsating traveling wave for @xmath494 + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + this last step is rather easier . we already know that for any @xmath495 , we have that @xmath367 is either a pulsating traveling wave or a positive and periodic stationary solution . but we have just shown that @xmath67 is isolated from below , therefore similarly as we have done before in the case @xmath496 to begin our iteration , for any @xmath308 close enough to @xmath497 , @xmath367 is a pulsating traveling wave connecting some stationary solution @xmath498 to @xmath67 . let us argue by contradiction and assume that it is not . since the sequence @xmath499 is monotonically decreasing , it converges uniformly to some @xmath500 , a periodic stationary solution of . as in the proof of lemma [ iteration - terrace ] , we use a super - solution crossing some @xmath67 to get a contradiction . let us now introduce the following function @xmath507 wherein @xmath508 , @xmath466 and @xmath509 is a solution of . using the same computations as in the proof of lemma [ iteration - terrace ] , one gets that @xmath510 for all @xmath473 small enough , on some domain of the form @xmath511 , wherein @xmath427 moves with the average speed @xmath54 and satisfies for some @xmath102 large enough and any @xmath49 : @xmath512 as @xmath513 as @xmath63 uniformly with respect to @xmath49 , one can proceed as before to reach a contradiction together with claim [ supersolpk ] . in fact , we have not yet proven that the sequence @xmath516 of the speeds of the pulsating traveling waves @xmath70 is nondecreasing . however , this directly follows from lemma [ iteration - terrace ] . indeed lemma [ iteration - terrace ] also states that we have a decreasing sequence @xmath517 such that @xmath518 then , from remark [ cvtaukspeed ] , the speed @xmath72 of @xmath70 can be obtained as @xmath519 and since @xmath520 is increasing for any @xmath521 and @xmath522 ( see remark [ rem : period ] ) , one obtains that the sequence @xmath72 is nondecreasing . to conclude the proof of theorem [ heavip2 ] , it remains to check that the propagating terrace @xmath515 satisfies all the required statements . this is in fact straightforward from all the above . part @xmath105 indeed immediately follows from the construction of the terrace and lemma [ iteration - terrace ] . part @xmath104 follows from claim [ walphapk ] and the construction of the @xmath70 as some @xmath124-limit orbits of @xmath99 . moreover , one can easily check that @xmath515 is minimal . indeed , let us argue by contradiction by assuming that it is not . then , one can easily check that there exists some @xmath102 and some traveling wave @xmath523 crossing @xmath67 . this contradicts the fact that it is steeper than any other entire solution . lastly , let us check that any other minimal propagating terrace @xmath524 is equal to @xmath515 . let @xmath525 be a given other minimal propagating terrace . then it immediately follows from the definition that the two sequences @xmath526 and @xmath101 are identically equal . then , for any @xmath102 , @xmath527 and @xmath70 are steeper than each other ( from definition [ def : terrace ] and part @xmath104 of theorem [ heavip2 ] ) and intersect , hence they are identically equal up to some time shift . let us first show the locally uniform convergence to the pulsating traveling waves @xmath528 along the moving frames with speed @xmath72 and some sublinear drifts . let us fix some @xmath69 . for any large enough @xmath19 , let us define @xmath529 such that @xmath530 and let us introduce @xmath531 wherein @xmath532 is defined as in section [ cvpulswave ] , with @xmath533 chosen so that @xmath534 . let us now consider @xmath535 , the piecewized affine function , defined by @xmath536 recall that the sequence @xmath537 converges to @xmath538 , so that @xmath539 as @xmath13 . furthermore , since @xmath540 where the above convergence is understood to hold locally uniformly with respect to @xmath541 , and since @xmath542 and @xmath543 stay bounded , one can check that @xmath544 thus , we obtain @xmath545 wherein both of the above convergences hold locally uniformly with respect to @xmath27 . this completes , in the general case , the convergence result stated in theorem [ heavip2-cv ] . + it now remains to consider what happens `` outside '' of the moving frames with speed @xmath546 . this will follow from the following monotonicity property : + let us first look on the left of the terrace , that is , when @xmath549 . in that case , we will use the fact that @xmath550 let @xmath111 be a given small enough number . from the asymptotics of @xmath551 , there exists @xmath552 such that for all @xmath19 : @xmath553 next for each time large enough one has for all @xmath554 $ ] : @xmath555 then , using claim [ claim : period - mon ] , one gets for all large enough @xmath19 : @xmath556 one can proceed similarly to get that for any @xmath111 , there exists @xmath557 such that for any @xmath558 , @xmath559 note that under assumption [ assumption - p2 ] , since @xmath78 , the uniform convergence immediately follows from the above computations . + lastly , let @xmath69 be a given integer . then one has @xmath560 as above , one can use claim [ claim : period - mon ] to show that there exists some constant @xmath112 such that for each large time @xmath19 : @xmath561 @xmath562 this ends the proof of theorem [ heavip2-cv ] .

user - generated content can be distributed at a low cost using peer - to - peer ( p2p ) networks , but the free - rider problem hinders the utilization of p2p networks . in order to achieve an efficient use of p2p networks , we investigate fundamental issues on incentives in content production and sharing using game theory . we build a basic model to analyze non - cooperative outcomes without an incentive scheme and then use different game formulations derived from the basic model to examine five incentive schemes : cooperative , payment , repeated interaction , intervention , and enforced full sharing . the results of this paper show that 1 ) cooperative peers share all produced content while non - cooperative peers do not share at all without an incentive scheme ; 2 ) a cooperative scheme allows peers to consume more content than non - cooperative outcomes do ; 3 ) a cooperative outcome can be achieved among non - cooperative peers by introducing an incentive scheme based on payment , repeated interaction , or intervention ; and 4 ) enforced full sharing has ambiguous welfare effects on peers . in addition to describing the solutions of different formulations , we discuss enforcement and informational requirements to implement each solution , aiming to offer a guideline for protocol designers when designing incentive schemes for p2p networks . game theory , incentives , network economics , peer - to - peer networks , pricing schemes .

0910.4618

recent developments in technology have significantly reduced the cost of producing and distributing content in various forms such as images , sounds , videos , and text . once produced only by companies with a large capital , content can now be produced by end - users . in today s internet - based social communities , peer - to - peer ( p2p ) networks offer a cost effective and easily deployable framework for sharing user - generated content @xcite . while p2p networks have many advantages such as scalability , resilience , and effectiveness in coping with dynamics and heterogeneity @xcite , they are vulnerable to intrinsic incentive problems in that the transfer of content incurs costs to uploaders as well as to downloaders but benefits only downloaders . since the social cost of transfer ( the sum of upload and download costs ) exceeds the private cost of transfer ( download costs ) , peers tend to download excessibly as in the tragedy of the commons problem . on the other hand , since upload incurs costs to uploaders without giving them direct benefit , peers tend to upload too little . the incentive problem stating that peers desire to benefit from p2p networks while not contributing to them is referred to as the free - rider ( or freeloader ) problem . various incentive schemes to mitigate the free - rider problem have been proposed and analyzed in the literature . cooperative schemes ( e.g. , @xcite , @xcite ) utilize helpers that download files on behalf of a peer in the same collaborative group . helpers can improve the download performance of p2p networks by sharing their spare upload capacities . however , forming and sustaining collaborative groups in a distributed system poses a main challenge to cooperative schemes . payment schemes ( e.g. , @xcite , @xcite ) use virtual currency or micropayment to reward upload and charge download . payment schemes have a solid theoretical foundation as they are based on economic models . however , they are regarded impractical because they require an accounting infrastructure to track the transactions of peers @xcite . differential service schemes ( e.g. , @xcite , @xcite ) allow peers to make upload decisions based on the rating of a peer that requests content from them . since a peer with a good reputation is treated preferentially by other peers , differential service schemes provide incentives for peers to contribute in order to build and maintain a good reputation . however , differential service schemes require large communication overheads to determine and announce the ratings of peers . the rating of a peer is determined by its past actions , which are observed by different peers , and peers need to know the rating of every other peer that they interact with . game theory @xcite offers a useful framework to model multi - user interaction and has been applied to analyze the behavior of peers in p2p networks . incentive schemes such as payment schemes and differential service schemes have been investigated using non - cooperative game theory . payment schemes can be easily incorporated in static game models as in @xcite , while differential service schemes have been studied in the context of different game models . @xcite uses a static game model to analyze the steady - state outcome of learning dynamics under a differential service scheme . @xcite simulates an evolutionary game model to examine the performance of a differential service scheme based on peer reciprocation . @xcite uses a repeated game model to construct a differential service scheme based on the idea of social norms @xcite . @xcite and @xcite apply the mechanism design approach to build optimal incentive - compatible differential service schemes . @xcite uses both repeated game and mechanism design approaches to propose cheat - proof and attack - resistant differential service schemes . cooperative game theory has also been used to investigate coalition formation among peers @xcite , @xcite . in this paper , we investigate fundamental issues on incentives in content production and sharing over p2p networks using game theory . unlike existing game - theoretic works on p2p networks , which focus on a particular game model to construct incentive schemes , we build a basic model and use it as a unified framework based on which different incentive schemes are examined applying various game theoretic models . specifically , we analyze the basic model as a non - cooperative game and examine five incentive schemes cooperative , payment , repeated interaction , intervention , and enforced full sharing using different game formulations derived from the basic model , as summarized in table i. hence , instead of arguing for a particular incentive scheme and a modeling approach , we show that alternative incentive schemes can provide incentives for sharing in p2p networks from a neutral perspective . as can be seen from table i , different incentive schemes and the corresponding game models have different requirements for implementation . since the characteristics of p2p networks vary depending on their architecture , the effectiveness of an incentive scheme will depend on the network environment . thus , our analysis in this paper can serve as a guideline for a protocol designer when modeling , comparing , and selecting incentive schemes . .comparisons of the approaches discussed in the paper . @xmath0 represents the utility functions of peers . @xmath1 represents the pareto efficient sharing levels of peers desired by the protocol designer ( pd ) . @xmath2 and @xmath3 represent optimal payment and intervention schemes , respectively , that implements @xmath4 . rationality of peers are assumed throughout . [ cols="^,^,^,^,^",options="header " , ] another distinctive feature of our framework is that we allow peers to make production decisions whereas most of existing works assume that peers are endowed with a certain amount of content ( see , for example , @xcite , @xcite ) . when produced content and downloaded content are substitutable in consumption , the amount of content a peer produces is affected by the amount of content available in a p2p network . by endogenizing the amount of content that peers produce , we can capture the strategic link between producing and downloading content . in addition , we consider scenarios where the sharing decisions of peers can be enforced while production and download decisions are made in a non - cooperative manner . these scenarios can be formulated as games with partial cooperation in which the strategies of players can be enforced only in some stages . the concepts developed in the discussion of enforced sharing decisions can be applied to other scenarios that can be modeled as multi - stage games . the rest of this paper is organized as follows . in section ii , we formulate the basic model that describes a scenario of content production and sharing . in section iii , we analyze the basic model as a non - cooperative game and identify the free - rider problem . in section iv , we investigate cooperative schemes by deriving a coalitional game based on the basic model . in section v , we augment the basic model with a payment scheme to achieve cooperative outcomes among non - cooperative peers . in section vi , we study differential service schemes applying repeated game and intervention approaches to the basic model . in section vii , we analyze a partially cooperative scenario where full sharing is enforced . in section viii , we provide numerical illustration . in section ix , we conclude and discuss future directions . proofs of propositions are provided either following propositions or in appendix b. we consider a completely connected p2p network of @xmath5 peers as in @xcite , @xcite . peers produce content ( e.g. , photos , videos , news , and customer reviews ) and use the p2p network to distribute produced content . following @xcite , we model the content production and sharing scenario as a sequential game consisting of three stages , which is called the content production and sharing ( cps ) game . * _ stage one ( production ) : _ each peer determines its level of production . @xmath6 represents the amount of content produced by peer @xmath7 and is known only to peer @xmath7 . * _ stage two ( sharing ) : _ each peer specifies its level of sharing . @xmath8 $ ] represents the amount of content that peer @xmath7 makes available to other peers . @xmath9 is known to all peers at the end of stage two . * _ stage three ( transfer ) : _ each peer determines the amounts of content that it downloads from other peers . peer @xmath7 serves all the requests it receives from any other peer up to @xmath10 . @xmath11 $ ] represents the amount of content that peer @xmath7 downloads from peer @xmath12 , or equivalently peer @xmath13 uploads to peer @xmath7 . let @xmath14 be the set of peers in the p2p network . for notations , we define @xmath15 , @xmath16 , and @xmath17_{i , j \in \mathcal{n}}$ ] , an @xmath5-by-@xmath5 matrix whose @xmath18-entry is given by @xmath19 , where we set @xmath20 for all @xmath21 . the download profile of peer @xmath7 is given by the @xmath7th row of @xmath22 , denoted by @xmath23 . similarly , the upload profile of peer @xmath7 is given by the @xmath7th column of @xmath22 , denoted by @xmath24 . given the elements of @xmath22 , we can compute the download volume of peer @xmath7 by @xmath25 and its upload volume by @xmath26 . for notational convenience , we suppress the dependence of @xmath27 and @xmath28 on @xmath22 and write @xmath27 and @xmath28 instead of @xmath29 and @xmath30 , respectively . also , we define @xmath31 to be the total transfer volume of the p2p network given @xmath22 , i.e. , @xmath32 , which can be considered as a measure of the utilization of the p2p network . we assume that peers produce nonidentical content of homogeneous quality , which allows us to focus on the quantity of content . the total amount of content that peer @xmath7 has at the end of the cps game , which we call the consumption of peer @xmath7 , is given by the sum of the amounts it produces and downloads , @xmath33 . the utility of peer @xmath7 is given by the benefit of consumption minus the costs of production , download , and upload : @xmath34 we analyze the case of homogeneous peers in that @xmath35 , @xmath36 , @xmath37 , and @xmath38 are the same for all peers . the benefit of consumption is measured by a concave function @xmath39 as in @xcite . we assume that @xmath35 is twice continuously differentiable and satisfies @xmath40 , @xmath41 and @xmath42 on @xmath43 . we also assume that @xmath44 is finite , to represent the right derivative of @xmath35 at 0 . ] @xmath45 , and @xmath46 so that for every @xmath47 $ ] there exists a unique @xmath48 that satisfies @xmath49 . we use linear cost functions as widely adopted in the literature ( see , for example , @xcite , @xcite ) . the cost of producing the amount of content @xmath50 is given by @xmath51 , where @xmath52 is the marginal cost of production . download and upload create costs in terms of bandwidth usage , and transferring the amount of content @xmath19 from peer @xmath7 to peer @xmath13 induces a cost of @xmath53 to peer @xmath7 ( the downloader ) and @xmath54 to peer @xmath13 ( the uploader ) , where @xmath55 and @xmath56 are the marginal costs of download and upload , respectively . the p2p network has a positive social value only if obtaining a unit of content through the p2p network costs less to peers than producing it privately . hence , we assume that @xmath57 to ensure that the p2p network is socially valuable . we illustrate the considered scenario with an example of financial data . suppose that peers need financial data ( e.g. , earnings of companies , gross domestic products , and interest rates ) in order to make forecasts based on which they make investment decisions ( e.g. , trade stocks and bonds ) . to obtain financial data , peers can either collect data by themselves or download data shared by other peers . a peer can make a more informed decision with a larger amount of data . hence , benefits that peers receive from data can be represented by an increasing benefit function . the benefit function is concave when the marginal returns of information are diminishing in the total amount . alternatively , we can obtain a concave benefit function by assuming that there is possible duplication in collected data , peers can not identify the contents of data before downloading them , and the benefit is proportional to the amount of distinct data . appendix a presents a formal proof of this statement . we first study the non - cooperative outcome of the cps game , without any incentive or enforcement device . non - cooperative peers choose their strategies to maximize their own utilities given others strategies . thus , peers strategies should be self - enforcing at non - cooperative equilibrium in that no peer can gain by choosing a different strategy unilaterally . a strategy for peer @xmath7 in the cps game is its complete contingent plan over the three stages and is denoted by @xmath58 . a stage - one strategy for peer @xmath7 is represented by @xmath6 , a stage - two strategy by a function @xmath59 such that @xmath60 for all @xmath6 , and a stage - three strategy by a function @xmath61 such that @xmath62 for all @xmath12 and @xmath63 for @xmath64 , where @xmath65 , y_j \in \mathbb{r}_+ , \forall j \neq i \}$ ] is the set of possible information sets at the beginning of stage three . nash equilibrium ( ne ) of the cps game is defined as a strategy profile such that no peer can improve its utility by a unilateral deviation . the play on the equilibrium path at an ne is called an ne outcome of the cps game . a refinement of ne for sequential games is subgame - perfect equilibrium ( spe ) , which requires that players choose ne strategies in any subgame , thereby eliminating incredible threats . subgame perfection provides robustness in equilibrium strategies in that deviation is unprofitable not only at the beginning of the game but also at any stage of the game . however , formally there is no subgame of the cps game starting from stage two or three because the stage - one choice of a peer is not revealed to other peers . hence , spe fails to provide a refinement of ne in the cps game . in order to extend the spirit of subgame perfection to non - singleton information sets , we can use sequential rationality , which postulates that players behave optimally in each information set for a given belief system @xcite . sequential rationality is required by the solution concepts of perfect bayesian equilibrium ( pbe ) and sequential equilibrium ( se ) . the difference between these two solution concepts disappears in the cps game because the consistent belief of peer @xmath7 on @xmath66 should be the correct @xmath67 in both solution concepts . hence , we use se to refer to a solution concept requiring sequential rationality and specify only the strategy profile to describe se suppressing the belief system with an implicit premise that peers hold correct beliefs . an se strategy profile of the cps game can be found applying a backward induction argument , which is described in detail in @xcite . as with ne , the play on the equilibrium path at an se is called an se outcome of the cps game . at the unique se outcome of the cps game , we have @xmath68 , @xmath69 , @xmath70 for all @xmath21 . thus , @xmath71 at se . a formal proof can be found in ( * ? ? ? . 1 ) . since sharing can incur the cost of upload while it gives no benefit to the sharing peer , it is never optimal for a peer to share a positive amount . expecting no sharing , each peer produces the autarkic optimal amount of content , @xmath72 , which maximizes @xmath73 . _ suppose that the cost of production is given by a general function @xmath74 instead of @xmath51 . there is still no sharing at se , but each peer produces an amount that maximizes @xmath75 , assuming that a maximum exists . proposition 1 shows that when peers behave non - cooperatively , even a small cost of upload makes the socially valuable p2p network never utilized because peers are not compensated for their upload . this result explains the free - riding behavior of peers in file sharing p2p networks such as napster and gnutella , as reported in @xcite , @xcite . using a similar argument as in the formal proof of proposition 1 , we can show that the ne outcome of the cps game is the same as the se outcome . ne may prescribe suboptimal strategies off the equilibrium path , but there can not be a positive amount shared on the equilibrium path . individual utility and total utility at non - cooperative equilibrium are @xmath76 and @xmath77 , respectively , where we define @xmath78 for @xmath79 as the conjugate of @xmath35 @xcite . is a concave function . ] we consider cooperative schemes in the cps game , which allow peers to form collaborative groups and to maximize their joint welfare . in order to prevent peers from behaving non - cooperatively , cooperative schemes need to enforce the actions of peers by a contract or a protocol . the protocol designer can implement a cooperative scheme if he knows the utility functions of all peers in order to determine a desired operating point and can enforce the operating point . an example of p2p networks to which a cooperative scheme can be applied is a camera network , where cameras in different locations capture the images of an object from various angles . a property that a desired operating point should possess is pareto efficiency ( pe ) , which is satisfied when there is no other operating point that makes some peers better off without making other peers worse off . we define social welfare by the sum of the utilities of peers , i.e. , @xmath80 . then an allocation is pareto efficient ( pe ) if it maximizes social welfare among feasible allocations . is feasible if @xmath81 , and @xmath82 for all @xmath12 , for all @xmath21 . ] let @xmath83 . at pe , we have @xmath84 and @xmath85 for all @xmath12 , for all @xmath21 . thus , @xmath86 at pe . a formal proof can be found in ( * ? ? ? 2 ) . pe can occur only when peers share all produced content and download all shared content . then the social welfare maximization problem can be written as @xmath87 \sum_{i=1}^n x_i.\end{aligned}\ ] ] the first - order optimality condition for @xmath88 is @xmath89 . note that @xmath90 is the per capita marginal cost of obtaining one unit of content when @xmath5 peers share all produced content . thus , at pe , the level of total production is chosen to equate the marginal benefit and the marginal cost of supplying content to every peer in the p2p network . _ proposition 2 determines production up to the aggregate level , leaving the individual levels unspecified . this is a by - product of the linear cost function of production ( i.e. , constant returns to scale ) . if we assume a strictly convex cost function ( i.e. , decreasing returns to scale ) , @xmath74 , instead of @xmath51 , then pe requires that every peer produce the same amount @xmath91 that maximizes @xmath92 , eliminating indeterminacy in the allocation of total production to peers . at pe , peers jointly produce @xmath93 and share all produced content so that each peer consumes the total amount produced . the utility of peer @xmath7 producing @xmath94 at a pe allocation @xmath95 is given by @xmath96 x_i^o.\end{aligned}\ ] ] note that the utility of a peer is decreasing in its production level given that the total amount of production is fixed and that all produced content is shared . total utility at pe is given by @xmath97 . since @xmath98 for @xmath99 , the consumption of a peer and total utility are smaller at non - cooperative equilibrium than at pe , i.e. , @xmath100 and @xmath101 . in order to derive a coalitional game @xcite based on the cps game , we need to compute the maximum total utility that a subset of peers can achieve . define @xmath102 by @xmath103 for @xmath104 . note that @xmath105 , @xmath106 , and @xmath107 as @xmath108 . @xmath102 can be interpreted as the per capita marginal cost of obtaining one unit of content when @xmath109 peers share all produced content . the maximum total utility achievable with @xmath109 peers is given by @xmath110 , and the maximum average individual utility achievable with @xmath109 peers by @xmath111 . marginal product ( mp ) measures an increment in the maximum total utility when the @xmath109th peer joins the p2p network , i.e. , @xmath112 and @xmath113 for @xmath114 . the following proposition gives some properties of the functions @xmath115 and @xmath116 . \(i ) @xmath117 is increasing in @xmath109 , and @xmath118 . + ( ii ) @xmath119 is increasing in @xmath109 , @xmath120 for all @xmath114 , and @xmath121 = 0 $ ] . since @xmath117 is increasing in @xmath109 , there are increasing returns to scale when inputs and outputs are taken to be peers and total utility , respectively . as there are more peers in the p2p network , the cost of production can be shared by more peers and the socially valuable p2p network can be utilized more extensively , which results in an increase in the maximum average individual utility . proposition 3(i ) in addition states that the maximum average individual utility is bounded above . proposition 3(ii ) shows that the no - surplus condition in the sense of @xcite is satisfied only in the limiting case with infinitely many peers . this implies that the distribution of total utility to peers according to their mp , which is proposed by the marginal productivity theory of distribution of neoclassical economics , is not feasible unless there are infinitely many peers . thus , we rely on cooperative game theory as an alternative theory of distribution . let @xmath122 with @xmath123 and @xmath124 be a coalition of peers . the characteristic function @xmath125 , which assigns each coalition the maximum total utility it can create , is given by @xmath126 where @xmath127 denotes the number of peers in coalition @xmath122 . we set @xmath128 . a coalitional game is described by the characteristic function @xmath125 , and we consider two solution concepts for coalitional games , the core and the shapley value . an allocation @xmath129 has the core property in the coalitional game @xmath125 if @xmath130 the first condition states that the maximum total utility with the grand coalition @xmath131 is distributed to peers ( i.e. , a pe allocation is chosen ) while the second condition states that no coalition can improve the utilities of its members from the current allocation . hence , the core describes the stable distributions of total utility in that no coalition of peers can improve their utilities by separating from the grand coalition . this implies that , when the protocol designer enforces an allocation with the core property , no coalition of peers can object the allocation credibly by threatening to leave the p2p network . the shapley value , whose expression can be found in @xcite , is a distribution of total utility , @xmath132 , that satisfies a certain set of axioms . the shapley value can be considered as a fair distribution of utility as it takes into account the mp of peers in all possible orders of arrival in the p2p network . \(i ) the core of the coalitional game @xmath125 is a nonempty convex set whose vertices are given by @xmath133 and all of its permutations . at the core , we have @xmath84 , @xmath85 for all @xmath12 , for all @xmath21 , and @xmath134 for all @xmath135 . + ( ii ) the shapley value of the coalitional game @xmath125 is @xmath136 for all @xmath21 , which is attained at the symmetric pe allocation , @xmath137 for all @xmath12 , for all @xmath21 . [ fig : solcon ] illustrates the results in propositions 1 , 2 , and 4 with two peers . since @xmath101 , pe allocations achieve a higher total utility than se allocations . also , since the core imposes additional constraints on pe , the core is a subset of pe utility profiles . proposition 3(ii ) implies that the coalitional game @xmath125 is convex @xcite , and thus the results in proposition 4 can be considered as the corollaries of theorems in @xcite . in particular , the core is nonempty and coincides with the unique stable set in the sense of @xcite . also , the shapley value is the center of gravity of the core , which is consistent with the illustration in fig . [ fig : solcon ] . hence , by prescribing the allocation that yields the shapley value , the protocol designer can obtain the stability property of the core and the fairness property of the shapley value at the same time . the maximum utility that a peer can obtain by itself is @xmath76 , which can be considered as a reservation utility . an allocation @xmath138 satisfies the participation ( or individual rationality ) constraint for peer @xmath7 if @xmath139 . an allocation @xmath140 is participation - efficient if it is pe and satisfies the participation constraint for every peer . among pe allocations , condition for a singleton coalition @xmath141 , which can be written as @xmath142 is required for the participation constraint for peer @xmath7 . since the utility of a peer decreases in its production level among pe allocations as shown in , the participation constraint puts an upper bound on the individual production level to prevent a peer from leaving the p2p network . the core is a stronger concept than participation - efficiency in that the core prevents not only a single peer from leaving the p2p network but also a subset of peers from forming their own p2p network . peers choose actions @xmath143 , @xmath144 , and @xmath22 over the three stages of the cps game . suppose that the protocol designer can enforce the sharing levels of peers in stage two while he can not enforce the choices in stages one and three . then the cps game is reduced to the cps game with enforced sharing levels @xmath145 , where the stage - two choice of peers is fixed at some @xmath145 . suppose that @xmath146 . at the se outcome of the cps game with enforced sharing levels @xmath145 , we have @xmath147 for all @xmath12 , for all @xmath21 . proposition 5 shows that when peers are required to share @xmath145 that satisfies @xmath146 , they produce exactly the enforced sharing levels and download all shared content in their self - interest . since @xmath148 , the protocol designer can implement a pe allocation by enforcing only the sharing levels @xmath145 such that @xmath149 , leaving peers to choose the production and download levels on their own . suppose that the protocol designer is unable to enforce the sharing levels of peers . the non - cooperative analysis in section iii suggests that , in order to avoid the collapse of the p2p network , peers need to be incentivized to share and upload their produced content . pricing is an extensively studied form of incentives to achieve an efficient use of network resources @xcite . pricing schemes have been used in p2p - based web services such as mojonation in the forms of tokens and credits . we say that a pricing scheme is optimal if it induces a pe non - cooperative equilibrium . in order to determine an optimal pricing scheme , the protocol designer needs to know the utility functions of peers . an optimal pricing scheme can be implemented when the protocol designer can enforce payments and each peer knows the pricing rule applied to it . in this section , we examine a class of pricing schemes under which the payment to a peer is increasing in its upload volume and decreasing in its download volume at the same rate . we call such a pricing scheme a linear pricing scheme , which can be expressed formally as @xmath150 for some price @xmath151 . note that a linear pricing scheme with any price satisfies budget balance since it simply transfers payments from downloaders to uploaders , i.e. , @xmath152 for all @xmath22 and @xmath153 . the payoff to peer @xmath7 in the cps game with the linear pricing scheme with price @xmath153 is given by @xmath154 in effect , the linear pricing scheme with price @xmath153 increases the cost of download from @xmath37 to @xmath155 and decreases the cost of upload from @xmath38 to @xmath156 . if the reward for upload exceeds the cost of upload , i.e. , @xmath157 , then peers receive a net benefit from uploading , which provides them with an incentive for sharing . the following proposition shows that there exists an optimal pricing scheme in the class of linear pricing schemes . moreover , the optimal linear pricing scheme transfers the utilities of peers so that the equilibrium payoff profile coincides with the shapley value for any pe allocation chosen by peers . let @xmath158/n$ ] . at the se outcome of the cps game with the linear pricing scheme with price @xmath159 , we have @xmath84 and @xmath160 for all @xmath12 , for all @xmath21 . the payoff of each peer at se is given by @xmath161 . the result follows from lemma 1 and proposition 7 of @xcite . _ as is the case with pe , indeterminacy in the allocation of total production to peers can be eliminated by assuming a strictly convex cost function of production , @xmath74 . with a strictly convex cost function of production , the optimal price @xmath159 can be found as in proposition 6 replacing @xmath36 with @xmath162 , where @xmath91 is the individual production level at pe as discussed in the remark following proposition 2 . a notable feature of the optimal price @xmath159 is that peers are indifferent between the two alternative methods of obtaining data , production and download , when they face the optimal price . at pe without pricing schemes , peers prefer download to production because the marginal cost of download , @xmath37 , is smaller than that of production and upload , @xmath163 ( see the coefficient of the term @xmath94 in ) . the optimal price is chosen such that it equates the effective marginal cost of download , @xmath164 , with that of production and upload , @xmath165 . as a result , at se with the optimal linear pricing scheme , peers obtain the same payoff regardless of their production levels . with linear cost functions and linear prices , this is a necessary property of any nondiscriminatory pricing scheme that induces non - cooperative homogeneous peers to produce a positive bounded amount of content in aggregate . if the payoff to a peer is increasing in its production level , then the peer would be willing to produce and upload as much as it can ( i.e. , the peer is overcompensated for its production and the supply of content is unbounded ) . on the other hand , if the payoff to a peer is decreasing in its production level , then the peer would not produce at all ( i.e. , the peer is undercompensated for its production and the supply of content is zero ) . the optimal price can be considered as an equilibrium price of content in that the compensation is enough to provide incentives for production and upload but prevents overproduction so that supply equals demand . imposing that @xmath166 and @xmath167 for all @xmath21 , the social welfare maximization problem can be written as @xmath168 x_i - \delta d_i \right\ } \nonumber \\ & \textrm{subject to } x_i \geq 0 , 0 \leq d_i \leq \sum_{j \neq i } x_j , \textrm{for all $ i = 1 , \ldots , n$}.\end{aligned}\ ] ] let @xmath169 be a lagrange multiplier on the constraint @xmath170 for each @xmath21 , which can be interpreted as the price that peer @xmath7 pays for its download . then the lagrangian function can be written as @xmath171 x_i - \delta d_i + p_i(\sum_{j \neq i } x_j - d_i ) \right\},\end{aligned}\ ] ] and the first - order optimality conditions for @xmath50 and @xmath27 are given by @xmath172 and @xmath173 at pe , we have @xmath174 , which yields the lagrange multiplier @xmath175 for all @xmath21 . the dual decomposition of can be written as @xmath176 x_i - ( p_i + \delta ) d_i\end{aligned}\ ] ] for each @xmath7 . the solution to is given by @xmath177 } & \textrm{if $ \sum_{j=1}^n p_j = \kappa + ( n-1 ) \sigma - \delta$}&,\nonumber\\ & x_i = \hat{x}_{[\kappa + ( n-1 ) \sigma - \sum_{j \neq i } p_j ] } \textrm { and } d_i = 0 & \textrm{if $ \sum_{j=1}^n p_j > \kappa + ( n-1 ) \sigma - \delta$}&.\footnotemark\end{aligned}\]]thus , the maximum value of is given by @xmath178 , and @xmath175 for all @xmath7 is the solution of the dual problem , @xmath179 . thus , a uniform linear pricing scheme suffices to obtain pe allocations . the problem is more general than the resource allocation problem in @xcite in the following aspect . in our problem , peers can choose to become either a seller or a buyer ( or both ) , and the amount of resources ( i.e. , content ) supplied in the p2p network is chosen by peers . on the contrary , in @xcite , buyers and sellers are predetermined , and sellers hold a fixed supply of resources . if the protocol designer knows the utility functions of peers , he can compute the optimal price @xmath159 using the expression in proposition 6 . at the optimal price , equilibrium requires that peers produce @xmath93 in aggregate . peers can coordinate to achieve total production @xmath93 using the following quantity adjustment process . initially , each peer @xmath7 chooses arbitrary optimal production and download levels @xmath180 , which satisfy @xmath181 . peers share their production fully throughout the process , and thus they can observe the production levels of other peers indirectly . if @xmath182 ( respectively , @xmath183 ) , then @xmath184 ( respectively , @xmath185 ) for all @xmath7 . hence , if each peer @xmath7 adjusts its production and download levels by @xmath186 for some constant @xmath187 , can not be negative , we assume that peer @xmath7 stops adjusting its quantity when @xmath188 and @xmath189 . ] then the allocation will converge to an equilibrium allocation , which satisfies @xmath84 . suppose instead that the protocol designer does not know the utility functions of peers . in this case , the protocol designer can still find the optimal price by using a price adjustment process similar to that in @xcite . in other words , a price adjustment process can substitute knowledge about the utility functions of peers . in the proposed price adjustment process , the protocol designer announces a price @xmath153 , which applies to every peer . given the price , each peer @xmath7 chooses @xmath190 by solving and reports its choice to the protocol designer . we assume that peers can coordinate their choices to satisfy @xmath191 for all @xmath21 whenever possible , for example , by using the quantity adjustment process . the total demand for content at price @xmath153 is denoted by @xmath192 and can be computed as @xmath193 . similarly , the total supply of content at price @xmath153 is denoted by @xmath194 and can be computed as @xmath195 , as a peer can upload its production up to @xmath196 times . using , we obtain @xmath197 } & \textrm{if $ p > p^*$ } \end{array } \right.,\end{aligned}\ ] ] as depicted in fig . [ fig : linprice ] . we define the excess demand at price @xmath153 by @xmath198 . the protocol designer adjusts the price following the process @xmath199 for some constant @xmath200 . since @xmath201 for @xmath202 and @xmath203 for @xmath204 , the price will converge to the optimal price @xmath159 starting from any initial price . we compare the above price adjustment process with that in @xcite . in @xcite , there are multiple resources with fixed supply , and each resource manager adjusts the price of his resource so that aggregate demand for the resource equals the supply of the resource . in our formulation , by focusing on a uniform linear pricing scheme , we treat resources provided by different peers as a single resource . hence , the protocol designer needs to aggregate demand and supply by all peers and adjust the price of content to eliminate excess demand or supply . . @xmath192 represents the total demand for download and @xmath194 the total supply of upload at price @xmath153 . the equilibrium price equates demand and supply , i.e. , @xmath205 . the protocol designer can reach the equilibrium price by adjusting the price depending on the excess demand , @xmath206.,scaledwidth=60.0% ] we have assumed that peers report their demand and supply truthfully in the price adjustment process . suppose instead that peers know the price adjustment process used by the protocol designer and can engage in strategic misrepresentation as in @xcite . we find that , unlike in @xcite , no peer can gain from influencing the equilibrium price by misreporting its demand or supply provided that other peers report truthfully . in @xcite , a user , acting as a buyer , can benefit from a lower price of a resource by underreporting its demand . on the contrary , in our model , a peer is both a buyer and a seller , and thus it can lower the price only by increasing its supply , which hurts it as a seller . peer @xmath7 can make the price adjustment process stop at @xmath207 by reporting @xmath208 and @xmath209 . since @xmath210 , the payoff of peer @xmath7 is @xmath211 \hat{x}_{(p'+\delta ) } < f(\hat{x}_{(p'+\delta ) } ) - \beta \hat{x}_{(p'+\delta ) } < f^*(\beta),\end{aligned}\ ] ] and thus it obtains a lower payoff by manipulating the equilibrium price at @xmath212 . for @xmath204 , the optimal production and download levels for peer @xmath13 are @xmath213}$ ] and @xmath214 . since @xmath215 , peer @xmath7 alone can not induce @xmath216 , or @xmath217 , for some @xmath204 , if @xmath218 . when @xmath219 , peer @xmath7 can make the price adjustment process stop at @xmath220 by reporting @xmath188 and @xmath221}$ ] . since @xmath222 , the payoff of peer @xmath7 is @xmath223}\right ) - ( p''+\delta ) \hat{x}_{[\kappa - ( n-1)(p''-\sigma)]}\\ & < f\left(\hat{x}_{[\kappa - ( n-1)(p''-\sigma)]}\right ) - \beta \hat{x}_{[\kappa - ( n-1)(p''-\sigma ) ] } < f^*(\beta).\end{aligned}\ ] ] again , peer @xmath7 can not gain from misreporting . another form of incentives that encourage sharing by non - cooperative peers is differential service , in which peers obtain different qualities of service depending on their contribution levels . differential service schemes are widely adopted in file sharing p2p networks such as bittorrent @xcite and kazaa , in the forms of tit - for - tat and reputation . in this section , we capture the differential service in the cps game using two modeling approaches based on repeated games and intervention . in a repeated game model , peers can reciprocate service to each other based on private or public history . in an intervention model , the system treats peers differentially based on their contribution to the system . suppose that peers interact repeatedly over time in the p2p network . the repeated game model can support a cooperative outcome among non - cooperative peers by providing rewards and punishments depending on the past behavior of peers . the repeated cps game is a supergame in which the cps game is played repeatedly . we use the limit of means criterion @xcite to evaluate the utility of a peer in the repeated cps game to obtain the following result . any participation - efficient allocation can be supported as a non - cooperative equilibrium of the repeated cps game . by proposition 5 , any deviation that is profitable in the current cps game involves a deviation in sharing levels , which can be publicly observed by peers . hence , the protocol designer can deter peers from free - riding in the p2p network by making peers play the se of the one - shot cps game in all subsequent cps games whenever a peer does not share its required amount of content . we have assumed that peers serve all the download requests they receive in stage three . suppose instead that a peer can choose whether to upload or not to another peer that requests its content . then the punishment following a deviation in sharing levels can be asymmetric by prescribing peers not to upload to a peer that has ever deviated , which effectively excludes the deviating peer from the p2p network . similarly , refusing a download request from a peer that has not deviated can also be deterred by using private retaliation ( i.e. , a non - deviating peer whose request was refused does the same thing in return to the peer that has refused its request ) in all subsequent cps games . intervention @xcite refers to the system directly influencing the usage of users depending on their behavior . we consider a particular form of intervention applicable to p2p networks . suppose that the p2p network can reduce the download rate of a peer depending on its rating , where the rating of peer @xmath7 is defined by its upload to download ratio , i.e. , @xmath224 . then a differential service scheme based on intervention can be described by an intervention function @xmath225 , which represents an increase in the marginal cost of download . that is , when peer @xmath7 has rating @xmath226 , its marginal cost of download after intervention is given by @xmath227 . note that the range of @xmath228 is constrained to be nonnegative since we assume that the system can only decrease the download rates . this imposes a restriction in incentive design compared to a payment scheme , where it is usually assumed that a payment function can take any positive or negative real number . however , incentive schemes based on intervention have advantages in implementation over those based on payment and repeated games . unlike a payment scheme , there is no need for transactions in an intervention scheme since intervention affects peers directly through the system . also , intervention can be considered as a substitute of the punishment strategy in repeated games , but it requires neither repeated interaction among peers nor the maintenance of history since punishment is executed by the system architecture rather than by peers . we say that an intervention scheme is optimal if it achieves a pe allocation with zero intervention level at non - cooperative equilibrium . since any positive level of intervention results in performance degradation , it is desirable to have intervention only as a threat , which is called for when misbehavior occurs . define an intervention function @xmath3 by @xmath229^+$ ] , where @xmath158/n$ ] and @xmath230^+ = \max \{r,0\}$ ] . at the se outcome of the cps game with the intervention scheme @xmath3 , we have @xmath137 for all @xmath12 , for all @xmath21 . moreover , @xmath231 at se . as long as @xmath232 , or @xmath233 , for all @xmath7 , the intervention scheme @xmath3 is equivalent to the optimal linear pricing scheme @xmath159 . since increasing @xmath28 beyond @xmath27 can only increase the cost of upload without affecting the level of intervention , we must have @xmath233 for all @xmath7 at se . among se with the optimal pricing scheme @xmath159 given in proposition 6 , @xmath233 is satisfied for all @xmath7 only when the amount of total production @xmath93 is split equally to all peers . at this allocation , @xmath234 , and thus @xmath235 and @xmath231 at se . proposition 8 shows that an optimal intervention scheme can be constructed to achieve the symmetric pe allocation without intervening at equilibrium . under the optimal intervention scheme @xmath3 , a peer experiences a reduced download rate whenever it downloads more than it uploads . since every peer downloads and uploads the same amount at the symmetric pe allocation , reduced download rates act only as a threat at equilibrium , deterring peers from deviation . the model of @xcite can be considered as using another form of intervention , where the system determines the proportion of shared content that a peer is allowed to download as a function of the contribution of the peer . the model of @xcite can also be interpreted as using an intervention scheme , where the system no longer serves a peer when its cumulative average rating falls below a threshold level . @xcite applies an intervention scheme to a multi - user access network , where the system can jam packets randomly with a probability that depends on the transmission probabilities of users . in @xcite , intervention affects all users in the system to the same degree , thus represented by a function that depends on the actions of all users . on the contrary , intervention considered in the cps game , @xcite , and @xcite influences a peer depending only on its own action , thus allowing the differential service to peers . we have seen from proposition 5 that the protocol designer can achieve a pe allocation by enforcing the sharing levels of peers . as an alternative scenario , suppose that the protocol designer can enforce full sharing among peers , but not sharing levels . the resulting cps game is called the cps game with enforced full sharing . formally , the cps game with enforced full sharing is a restricted version of the cps game where the stage - two choice of each peer @xmath7 is fixed as @xmath166 . note that enforced sharing levels constrain the production decisions of peers in that peers need to produce at least the required sharing levels . on the contrary , under enforced full sharing , peers can choose any levels of production in stage one . the following proposition characterizes allocations at the se of the cps game with enforced full sharing . let @xmath236 . define @xmath237 by @xmath238 if @xmath239 and @xmath240 otherwise ( i.e. , @xmath241 ) . at the se outcome of the cps game with enforced full sharing , we have @xmath242 and @xmath160 for all @xmath12 , for all @xmath21 . thus , @xmath243 at se with enforced full sharing . a formal proof can be found in ( * ? ? ? since peers download all shared content at se , enforced full sharing increases the effective marginal cost of production from @xmath36 to @xmath244 , which includes the marginal cost of upload to @xmath196 peers . the stage - one problem for peer @xmath7 can be written as @xmath245 x_i - \delta \sum_{j \neq i } x_j\end{aligned}\ ] ] given @xmath67 , and the result follows . as peers face effectively a higher cost of production with enforced full sharing , non - cooperative peers reduce their production when full sharing is enforced , i.e. , @xmath246 . total utility at se with enforced full sharing is given by @xmath247 $ ] . to make welfare comparisons , we first consider a scenario in which the number of peers in the p2p network is fixed as @xmath5 . the price of anarchy ( poa ) is defined to be the ratio of social welfare at the worst non - cooperative equilibrium to that at pe , i.e. , @xmath248 the price of no sharing ( pons ) compares social welfare at se with and without enforced full sharing , i.e. , @xmath249 finally , the price of underproduction ( pou ) compares social welfare at se with enforced full sharing and at pe , i.e. , @xmath250 when @xmath251 , the poa can be decomposed as the product of the pons and the pou , i.e. , @xmath252 . the poa is a widely used measure of the inefficiency of non - cooperative equilibria . the pons measures the welfare implication of enforced full sharing on selfish peers , and thus it can be used to analyze the value of a technology that enables enforced full sharing . the pou measures inefficiency due to underproduction caused by the selfish behavior of peers assuming that full sharing is enforced . the following proposition examines the range of values that each measure of inefficiency can take when we vary the utility specification of the model , @xmath35 , @xmath36 , @xmath37 , and @xmath38 . for a fixed size @xmath99 of the p2p network , @xmath253 , @xmath254 $ ] , and @xmath255 . these bounds are tight . a formal proof can be found in ( * ? ? ? 4 ) . since @xmath256 for @xmath99 , it follows immediately from and that @xmath257 , which shows that selfish behavior results in efficiency losses regardless of whether full sharing is enforced or not . the relative size of @xmath258 and @xmath259 is ambiguous , which implies that the enforcement of full sharing may make peers worse off . this is because enforced full sharing has two offsetting effects on social welfare . on one hand , full sharing has a positive effect on welfare by reducing the cost of obtaining one unit of content to @xmath90 , compared to @xmath36 in the case of no sharing . on the other hand , full sharing has a negative effect by increasing the effective cost of producing one unit of content from @xmath36 to @xmath260 . therefore , the overall welfare implication of enforced full sharing is determined by the stronger of the two effects . next we consider a scenario in which the number of peers in a p2p network is endogenously determined by peers . there are total @xmath5 peers that are connected to each other , and they can form groups to share their content within a group . the maximum average individual utility increases with the number of peers in a group as shown in proposition 3(i ) . thus , in a cooperative scenario , peers will form a p2p network with all the @xmath5 peers if they accept a new peer as long as the inclusion of an additional peer benefits existing peers assuming that peers split total utility equally . in a non - cooperative scenario , peers do not share content at all , and thus their utilities do not depend on the number of peers . hence , the previous results that @xmath97 and @xmath77 are still valid with endogenous network formation . to analyze a scenario with enforced full sharing , define @xmath261 by @xmath262 for @xmath263 . then the average individual utility of a peer in a p2p network of size @xmath109 is given by @xmath264 increasing the size of a p2p network has two opposing effects on average individual utility . on one hand , increasing the number of peers benefits peers by reducing the effective marginal cost of obtaining content as represented by @xmath265 , which decreases with @xmath109 . on the other hand , increasing the number of peers does harm to peers by increasing the effective marginal cost of producing content as represented by @xmath266 , which increases with @xmath109 . hence , we can expect that there exists an optimal size of a p2p network that balances these positive and negative effects . since @xmath267 and @xmath268 for all @xmath269 , there must exist a maximizer of @xmath270 among @xmath271 , denoted by @xmath272 , where @xmath273 is the largest integer smaller than or equal to @xmath274 . we assume that @xmath272 is unique , which will hold for a generic specification of the utility function . when @xmath5 peers form p2p networks endogenously to maximize their individual utilities , they will form @xmath275 networks of size @xmath272 and one network of residual peers . hence , total utility that a coalition @xmath122 can create is given by @xmath276 in order to examine the stability property of endogenous network formation , we characterize the core of the coalitional game @xmath277 . suppose that @xmath278 . if @xmath5 is a multiple of @xmath272 , then the core of the coalitional game @xmath277 consists of a unique element @xmath279 for all @xmath21 . otherwise , the core is empty . note that we necessarily have @xmath278 when @xmath280 . when @xmath5 is not a multiple of @xmath272 , there is a residual network , whose size is smaller than @xmath272 . a peer in the residual network can bid a utility smaller than @xmath281 to form a network of size @xmath272 including itself , yielding instability for the networks of size @xmath272 . suppose that @xmath5 is a multiple of @xmath272 so that the core is nonempty . the utility profile in the core is achieved by peers forming @xmath282 networks , producing @xmath283 for all @xmath7 , and sharing all produced content within a network . social welfare at the allocation with the core property is @xmath284 . since @xmath285 , we have @xmath286 when peers can form p2p networks of the optimal size . that is , with endogenous network formation , enforced full sharing can only improve the welfare of peers because peers are given the option of operating in an autarkic manner . in this section , we provide illustrative results using a particular utility specification and varying the number of peers . for the utility function of peers , we use @xmath287 , @xmath288 , @xmath289 , and @xmath290 . we consider the ( exogenous ) number of peers in the p2p network , @xmath5 , from 1 to 100 . [ fig : num ] shows average individual utility in the three scenarios : @xmath161 in the cooperative case , @xmath76 in the non - cooperative case , and @xmath291 in the partially cooperative case ( i.e. , enforced full sharing ) . it can be seen that @xmath161 is increasing in @xmath5 , verifying proposition 3(i ) , that @xmath76 is independent of @xmath5 , and that @xmath291 reaches a peak at @xmath292 and is zero for all @xmath293 . [ fig : num ] plots total utility in the three scenarios : @xmath294 in the cooperative case , @xmath258 in the non - cooperative case , and @xmath259 in the partially cooperative case . [ fig : num ] compares the mp of the @xmath109th peer , @xmath119 , with the maximum average individual utility that @xmath109 peers can achieve , @xmath117 , verifying proposition 3(ii ) . [ fig : num ] plots the three inefficiency measures defined in section vii . since @xmath76 is independent of @xmath5 , we can see that the poa and the pons change with @xmath5 in the opposite way that @xmath161 and @xmath291 change , respectively . since @xmath161 converges to @xmath295 , the poa converges to @xmath296 as @xmath5 goes to infinity . fig . [ fig : num ] shows the utilization of the p2p network in the three scenarios : @xmath297 in the cooperative case , @xmath298 in the non - cooperative case , and @xmath299 in the partially cooperative case . we can see no utilization in the non - cooperative case and underutilization ( and no utilization for @xmath293 ) in the partially cooperative case compared to the cooperative case , which exhibits a high utilization of the p2p network . finally , fig . [ fig : num ] plots the optimal linear price @xmath159 as a function of @xmath5 . as can be seen its expression in proposition 6 , @xmath159 decreases with @xmath5 and converges to @xmath290 as @xmath5 goes to infinity . in this paper , we have provided a unified framework to investigate incentive issues in content production and sharing using various game theoretic approaches . we have characterized the non - cooperative and cooperative outcomes of the cps game and have shown that incentive schemes such as payment schemes and differential service schemes can yield a cooperative outcome among non - cooperative peers . throughout the paper , we have discussed enforcement and information requirements to implement the solutions of different approaches with a protocol . our analysis allows protocol designers to compare the performance and the overheads of different approaches and eventually helps them select the best approach given a network environment they face . we have maintained the homogeneity assumption in order to keep our model tractable so that we can better illustrate different approaches by providing analytic results . however , all the concepts in this paper can be straightforwardly applied to the case of peers with heterogeneous utility functions . the convexity of the coalitional game in the cooperative approach will still hold , but computing cooperative solutions will become more complicated with heterogeneous peers . non - cooperative solutions will remain the same with a minor change that individually optimal production levels will differ across peers . also , heterogeneous peers in a distributed system offer a natural scenario to which a mechanism design approach can be applied . lastly , the protocol designer may need to discriminate heterogeneous peers in order to achieve a cooperative outcome using a linear pricing scheme . investigating how the results in this paper extend to and change in p2p networks with heterogeneous peers will provide interesting and challenging future research directions . for simplicity , suppose that peers choose the number of files from the set of nonnegative integers . there are total @xmath300 files that can be potentially produced by a peer , where @xmath300 is a large positive integer . when peer @xmath7 produces @xmath50 files , it draws @xmath50 files with replacement from the @xmath300 files . ( now @xmath36 can be considered as a constant cost of a draw . ) each file is drawn with equal probability of @xmath301 . let @xmath302 be the number of files that peer @xmath7 consumes , i.e. , @xmath303 . since peers can not identify the content of files produced by others before download , @xmath302 files that peer @xmath7 consumes can be considered as @xmath302 independent draws from the @xmath300 files . the probability that a given file is not one of the @xmath302 files is @xmath304 . hence , the expected number of distinct files in the @xmath302 files is @xmath305.\end{aligned}\ ] ] since @xmath306 , we have @xmath305 \leq c_i,\end{aligned}\ ] ] and @xmath307 $ ] is the expected number of redundant files in the @xmath302 files . if the benefit of consumption is proportional to the number of distinct files , the expected benefit function is given by @xmath308\end{aligned}\ ] ] for some constant @xmath309 . note that @xmath35 satisfies all the assumptions for a benefit function given in section ii when @xmath310 if we take @xmath302 as a nonnegative real number rather than a nonnegative integer . in particular , @xmath311 and @xmath35 is concave . * proof of proposition 3 . * ( i ) note that @xmath312 . also , @xmath313 , where @xmath49 , for @xmath47 $ ] . that is , @xmath314 is the unique maximizer of @xmath315 on @xmath316 . choose @xmath317 $ ] such that @xmath318 . then @xmath319 . hence , @xmath320 is decreasing on @xmath321 $ ] . since @xmath322 @xmath265 is decreasing in @xmath109 and its range lies in @xmath323 \subset ( 0,f'(0)]$ ] . since @xmath115 is a composite function of two decreasing functions , it is increasing . \(ii ) to prove that @xmath116 is increasing in @xmath109 , it suffices to show the strict convexity of @xmath326 , taking the domain of @xmath265 and @xmath326 as @xmath43 instead of @xmath327 . since @xmath328 and @xmath320 and @xmath265 are differentiable on @xmath43 , by the chain rule @xmath326 is differentiable and @xmath329 note that @xmath330 on @xmath331 , @xmath332 is continuously differentiable on @xmath43 , and @xmath333 for all @xmath334 . by the inverse function theorem , @xmath320 is twice continuously differentiable on @xmath331 , and we have @xmath335 + n ( f^*)''(\tilde{\beta}(n ) ) ( \tilde{\beta}'(n))^2.\end{aligned}\ ] ] since @xmath336 and @xmath320 is strictly convex on @xmath331 , we have @xmath337 for all @xmath338 . thus , @xmath119 is increasing in @xmath109 . since @xmath320 is convex and differentiable , we have @xmath341\\ & = \hat{x}_{\tilde{\beta}(n ) } [ \tilde{\beta}(n-1 ) - \tilde{\beta}(n)]\\ & < \hat{x}_{(\delta + \sigma ) } [ \tilde{\beta}(n-1 ) - \tilde{\beta}(n)].\end{aligned}\ ] ] note that @xmath342 = \frac{1}{n } ( \kappa - \delta - \sigma).\end{aligned}\ ] ] hence , @xmath343 < \hat{x}_{(\delta + \sigma ) } \frac{1}{n } ( \kappa - \delta - \sigma),\end{aligned}\ ] ] and taking limits as @xmath108 yields the desired result.@xmath344 * proof of proposition 4 . * ( i ) proposition 3(ii ) implies that the coalitional game @xmath125 is convex . hence , the first sentence follows from theorems 3 and 4 of @xcite . the first condition for the core property , @xmath345 , is equivalent to pe , which requires @xmath346 , @xmath160 for all @xmath12 , for all @xmath21 as shown in proposition 2 . choose an arbitrary coalition @xmath122 . for a pe allocation @xmath129 , we have @xmath347 \sum_{i \in \mathcal{s } } x_i.\end{aligned}\ ] ] hence , using , we can show that is equivalent to the second condition for the core property , @xmath348 . \(ii ) let @xmath349 be the shapley value of the coalitional game @xmath125 . by the efficiency property of the shapley value , we have @xmath350 . also , by the symmetry axiom , we have @xmath351 for all @xmath352 . combining these two yields @xmath353 for all @xmath21 . using , we can see that @xmath354 is necessary to obtain the shapley value.@xmath344 * proof of proposition 5 . * suppose that @xmath355 for some @xmath21 at se . then it must be the case that @xmath356 . since @xmath357 , we have @xmath358 . then we obtain a contradiction to se because peer @xmath7 can improve its utility by reducing @xmath50 and increasing @xmath27 by the same amount . thus , @xmath359 for all @xmath12 , for all @xmath21 . the requirement for peer @xmath7 that @xmath360 in stage two restricts its stage - one choice with @xmath361 . suppose that @xmath362 for some @xmath21 at se . since @xmath363 at se , the first - order effect of increasing @xmath50 on @xmath364 is given by @xmath365 . since @xmath362 implies @xmath366 , we have @xmath367 for @xmath362 , and thus peer @xmath7 becomes worse off by choosing @xmath362 , contradicting se.@xmath344 * proof of proposition 7 . * let @xmath95 be a participation - efficient allocation . consider the following repeated game strategy for peer @xmath7 : start with a cooperative strategy in the cps game @xmath368 , @xmath369 , and @xmath370 , where @xmath371 is the optimal download profile of peer @xmath7 given @xmath372 , play the cooperative strategy if @xmath373 in all the previous cps games , and play the se strategy of the one - shot cps game , i.e. , @xmath68 , @xmath374 , and @xmath370 , if @xmath375 in at least one of the previous cps games . proposition 5 implies that peer @xmath7 can not gain in the current cps game by deviating to @xmath376 or @xmath377 for some @xmath12 . hence , a profitable deviation involves either @xmath378 in stage one or @xmath379 in stage two ( or both ) . either case results in a reduction in the sharing level from @xmath380 . since sharing levels are publicly observed , any profitable deviation is detectable and punishment will be triggered . since the gain from deviation in the current cps game is bounded above , it will be erased by the punishment in the long run . in other words , peer @xmath7 receives @xmath381 on average if it follows the described repeated game strategy and @xmath76 if it deviates in a way that the deviation increases the current utility . since @xmath95 is participation - efficient , we have @xmath382 for all @xmath21 . hence , the described repeated game strategy , which realizes the allocation @xmath95 in every cps game , is a non - cooperative equilibrium of the repeated cps game.@xmath344 * proof of proposition 11 . * assume that the core is nonempty and choose a utility profile @xmath383 in the core . suppose that there exists a peer @xmath7 with @xmath384 . consider a coalition @xmath122 of size @xmath272 that do not include peer @xmath7 , which must exist since @xmath385 . then @xmath386 , and thus @xmath387 , where peer @xmath388 is the one that receives the highest utility among peers in @xmath122 . then @xmath389 , and thus peer @xmath7 and peers in @xmath390 can block the utility profile @xmath391 . hence , we need to have @xmath392 for all @xmath21 . this is possible , with equality , only if @xmath5 is a multiple of @xmath272 . we can confirm that the core is nonempty since the utility profile @xmath279 for all @xmath21 satisfies the definition of the core.@xmath344 j. wang , c. yeo , v. prabhakaran , and k. ramchandran , `` on the role of helpers in peer - to - peer file download systems : design , analysis and simulation , '' presented at the 6th int . workshop peer - to - peer syst . ( iptps07 ) , 2007 . t. b. ma , s. c. m. lee , j. c. s. lui , and d. k. y. yau , `` incentive and service differentiation in p2p networks : a game theoretic approach , '' _ ieee / acm trans . 5 , pp . 978991 , oct . 2006 . w. s. lin , h. v. zhao , and k. j. r. liu , `` incentive cooperation strategies for peer - to - peer live multimedia streaming social networks , '' _ ieee trans . multimedia _ , vol . 3 , pp . 396412 , apr . m. k. h. yeung and y .- k . kwok , `` game theoretic peer selection for resilient peer - to - peer media streaming systems , '' in _ proc . conf . distributed computing syst . ( icdcs 08 ) _ , 2008 , pp . 817824 . h. park and m. van der schaar , `` coalition based resource negotiation for multimedia applications in informationally decentralized networks , '' _ ieee trans . multimedia _ , vol . 4 , pp . 765779 , jun . 2009 . kelly , a. k. maulloo , and d. k. h. tan , `` rate control for communication networks : shadow prices , proportional fairness and stability , '' _ j. operational research soc . 49 , no.3 , pp . 237252 , mar .

we study the elasto - plastic behaviour of materials made of individual ( discrete ) objects , such as a liquid foam made of bubbles . the evolution of positions and mutual arrangements of individual objects is taken into account through statistical quantities , such as the elastic strain of the structure , the yield strain and the yield function . the past history of the sample plays no explicit role , except through its effect on these statistical quantities . they suffice to relate the discrete scale with the collective , global scale . at this global scale , the material behaves as a continuous medium ; it is described with tensors such as elastic strain , stress and velocity gradient . we write the differential equations which predict their elastic and plastic behaviour in both the general case and the case of simple shear . an overshoot in the shear strain or shear stress is interpreted as a rotation of the deformed structure , which is a purely tensorial effect that exists only if the yield strain is at least of order 0.3 . we suggest practical applications , including : when to choose a scalar formalism rather than a tensorial one ; how to relax trapped stresses ; and how to model materials with a low , or a high , yield strain .

1002.0732

discrete rearranging patterns include cellular patterns , for instance liquid foams , biological tissues and grains in polycrystals ; assemblies of particles such as beads , granular materials , colloids , molecules and atoms ; and interconnected networks @xcite . many of these disordered materials display elastic and plastic properties , so that the stress tensor can rotate and is not necessarily aligned with the strain rate tensor ; in models this effect is included in objective derivatives @xcite . use of simplified geometries , _ e.g. _ in a rheometer , allows a first characterization of the material through measurements of shear stress . an overshoot in the shear stress is seen during the first loading in materials such as polymers @xcite , granular materials @xcite , and emulsions @xcite . for liquid foams this effect has been observed in a plate - plate rheometer @xcite and in simulations @xcite . it is unclear whether this is due to a change in the material s structure , or a tensorial effect of shear ; but nevertheless the overshoot is an essential ingredient in a recent model @xcite of the strain - rate discontinuity in the cylindrical couette foam flow experiments of ref . @xcite . such an overshoot results in mechanical bistability : two different values of strain correspond to the same value of stress between the plateau and the maximum , and can thus coexist . here , we investigate the elastic regime and elasto - plastic transition in a fully tensorial model . to describe the mechanical behaviour we use a formalism adapted for discrete rearranging disordered patterns which enables us to quantify rotational effects and to test the relevant parameters @xcite . we use as an example a sheared liquid foam @xcite . although a liquid foam consists only of gas bubbles surrounded by liquid walls , it exhibits a complex mechanical behaviour . it is elastic for small strains , plastic for large strains and flows at large strain rates @xcite . this behavior is useful for numerous applications such as ore separation , oil extraction , foods and cosmetics . the individual objects , namely the bubbles , are easily identified , which makes a liquid foam ( or alternatively an emulsion , made of droplets ) a model for the study of other complex fluids . this paper is organized as follows . in section [ sec : simulations ] , we simulate the quasistatic 2d flow of a foam in a couette shear geometry ; we explain how we perform and represent the measurements . in sec . [ sec : model ] , we present our equations , and discuss the specific effects due to the use of tensors , such as the overshoot . sec . [ sec : comparisonsimulationmodel ] compares the model and the simulation , and extracts the relevant information . [ sec : practical ] presents applications to practical situations , _ i.e. _ how and when to use the model . sec . [ sec : conclusions ] summarizes our findings . an appendix explains the notation and provides the detailed equations . we simulate numerically a 2d foam flowing in a linear couette shear geometry . simulations of dry foams offer several advantages : ( i ) the parameters are homogeneous ( liquid fraction , bubble area ) and controlled ( no diffusion - driven coarsening or film rupture ) ; ( ii ) the yield strain is of order of 0.3 , which is large enough to observe a full tensorial elastic regime while small enough that plastic effects can be easily observed ; ( iii ) all physical quantities can be easily measured . . characteristics of simulated foams . the different columns correspond to the symbols used in figs . [ fig : plaplasticlimit ] and [ fig : plaovershootus ] , effective liquid fraction @xcite , area dispersity , boundary conditions and maximal amplitude of the cycles . [ cols="^,^,^,^,^",options="header " , ] ( 8,12)(0.0,0.0 ) ( 0,8.1 ) and [ fig : simulationrepresentation ] . bubbles with 6 neighbours are displayed in white , otherwise in gray.,title="fig:",width=151 ] ( 4.5,8.1 ) and [ fig : simulationrepresentation ] . bubbles with 6 neighbours are displayed in white , otherwise in gray.,title="fig:",width=151 ] ( 0,4 ) and [ fig : simulationrepresentation ] . bubbles with 6 neighbours are displayed in white , otherwise in gray.,title="fig:",width=151 ] ( 4.5,4 ) and [ fig : simulationrepresentation ] . bubbles with 6 neighbours are displayed in white , otherwise in gray.,title="fig:",width=151 ] ( 2,-0.1 ) and [ fig : simulationrepresentation ] . bubbles with 6 neighbours are displayed in white , otherwise in gray.,title="fig:",width=151 ] ( -0.5,8.1 ) ( 4,8.1 ) ( -0.5,4 ) ( 4,4 ) ( 1.5,-0.1 ) several ideal , two - dimensional , dry foams @xcite are simulated ( fig . [ fig : plastructurereffp ] and table [ tab : defsimul ] ) . we use the surface evolver @xcite in a mode in which each film is represented as a circular arc . the value of surface tension is taken equal to 1 throughout , without loss of generality . a realistic foam structure is found by minimizing the total film length subject to the constraint of fixed bubble areas , prescribed at the beginning of the simulation . the simulations are quasistatic , which means that the system has time to relax between successive time steps ( increments in applied strain ) . relaxation effects are thus neglected and viscosity does not need to be included . the behaviour is expected to be elasto - plastic . the simulation procedure is as follows . a voronoi construction of randomly distributed points @xcite ( not shown ) is first used to generate a fully periodic tessellation of the plane . to create a confined foam , bubbles at the top and bottom are sequentially deleted until the required number of bubbles remains . in each case , the structure is imported into the surface evolver and target bubble areas prescribed , either all the same ( monodisperse , @xmath0 ) , a small random variation of up to 20% about monodisperse ( @xmath1 ) , or equal to the areas given by the voronoi construction ( @xmath2 ) . the initial foam configuration for each simulation ( _ e.g. _ label ( 1 ) in fig . [ fig : plastructurereffp ] ) is found by reducing the total film length to a local minimum . during this minimization , neighbour swappings ( so - called t1s " @xcite ) are triggered by deleting each film that shrinks below a certain critical length @xmath3 and allowing a new film to form to complete the process . the critical length @xmath3 defines and measures an effective liquid fraction , @xmath4 @xcite , here chosen to be very dry ( table [ tab : defsimul ] ) . one geometry consists of a unit cell of 400 bubbles with fully periodic boundary conditions to eliminate any artefacts due to small sample sizes . the second geometry mimics more closely a real experiment , and consists of 296 bubbles with two parallel bars ( about 15 bubble diameters apart ) confining the foam and with periodicity in one direction only . to shear the foams , two different procedures are required . for the periodic foams , one off - diagonal component of the matrix describing the periodicity of the unit cell is adjusted by a small amount @xcite . for the confined foams , a small step in strain is applied by moving one of the confining walls a small distance and then moving all vertices affinely . in each case this is followed by reduction of the film length to a minimum correct to 16 d.p . using conjugate gradient ( without biasing the search by introducing any large - scale perturbations of the structure ) . in tab . [ tab : defsimul ] ) . horizontal axis : time is in arbitrary units , equivalent to the cumulated strain " @xmath5 , where @xmath6 is the time and @xmath7 is defined up to an arbitrary prefactor ; here 2.25 cycles are represented . vertical axis : all curves represent @xmath8 ( left scale ) except for the saw - tooth which is the applied @xmath9 ( right scale ) . numbers correspond to the pictures in fig . [ fig : plastructurereffp ] . the first step , plotted with a thick solid line , starts at the @xmath10 ( indicated also by a number 1 ) and its end is labelled by number 2 : @xmath11 . the second step , plotted with a thin solid line , is from number 2 to 3 . the third step , plotted with a middle solid line , starts at the @xmath12 ( indicated also by a number 3 ) , and extends to number 5 : @xmath13 . four predictions of the model are plotted as dashed lines ( see fig . [ fig : pla - h - n ] for explanation of the legend ) ; for clarity they are plotted only from 1 to 2 and from 3 to 5 : note that from 3 to 4 all predictions are indiscernable from the simulation . , width=302 ] each foam is subjected to at least two saw - tooth " shear cycles of amplitude @xmath14 . positive and negative steps correspond respectively to shear toward increasing or decreasing imposed strain @xmath9 ( fig . [ fig : integral_strain ] ) are tested during the shear of the reference simulation ( @xmath15 in tab . [ tab : defsimul ] ) . a ) strain eigenvalues @xmath16 _ vs _ @xmath17 ; black solid line : initial shear ( @xmath18)which anneals the disorder ; blue dots : shear cycles ( @xmath19 ) ; dashed blue line : straight line of slope -1 passing through the origin . b ) deviatoric stress - strain relation : red , @xmath20 _ vs _ @xmath8 ; blue , @xmath21 _ vs _ @xmath22 ; red and blue data almost perfectly overlap ( see zoom in inset ) ; the slope determines @xmath23 , where @xmath24 is the shear modulus . , title="fig:",width=302 ] + + are tested during the shear of the reference simulation ( @xmath15 in tab . [ tab : defsimul ] ) . a ) strain eigenvalues @xmath16 _ vs _ @xmath17 ; black solid line : initial shear ( @xmath18)which anneals the disorder ; blue dots : shear cycles ( @xmath19 ) ; dashed blue line : straight line of slope -1 passing through the origin . b ) deviatoric stress - strain relation : red , @xmath20 _ vs _ @xmath8 ; blue , @xmath21 _ vs _ @xmath22 ; red and blue data almost perfectly overlap ( see zoom in inset ) ; the slope determines @xmath23 , where @xmath24 is the shear modulus . , title="fig:",width=302 ] + at each step , the positions of the bubble centres and films are recorded . tensorial quantities are measured by averaging over all bubbles , as follows @xcite . the texture tensor @xmath25 is computed statistically as an average over vector links @xmath26 between centres of neighbouring bubbles . we assume here that the reference texture at rest , @xmath27 , is isotropic . we thus define it by measuring the average of the determinant of @xmath28 over the duration of the whole simulation , @xmath29 . the elastic strain of bubbles expresses the deviation from the reference state , @xmath30 ( eq . [ eq : fundament_4_ann ] ) . this tensor is symmetric by construction ( @xmath31 ) ; it can be diagonalized and has two eigenvalues ( @xmath32 , @xmath33 ) in two orthogonal eigendirections . the simulations ( fig . [ fig : verifutraceless]a ) verify that we can reasonably assume its trace to be always close to zero , @xmath34 , as is roughly expected for an incompressible material @xcite . thus , due to its symmetry and vanishing trace , @xmath35 has only two independent components : @xmath36 fig . [ fig : verifutraceless]a shows that our measurement of the elastic strain makes evident the effect of shear - induced shuffling : the annealed foam ( blue dots ) differs significantly from the initial one ( black solid line ) . the contribution to the stress of the network of bubble walls is obtained by integrating over all films @xcite ; it yields the deviatoric ( _ i.e_. traceless ) part of the elastic stress tensor @xmath37 . the trace of the stress , namely the pressure , is unimportant here . the simulations are quasistatic and the viscous part of the stress is not relevant . we check ( fig . [ fig : verifutraceless]b ) that the stress and strain are strongly correlated ; that their correlation is linear ; and that it is isotropic ( the same for @xmath38 and @xmath39 components ) @xcite . half the slope thus defines and measures the elastic shear modulus @xmath24 . ( 7,4)(0.0,0.0 ) ( -1,0 ) ( 3.5,0 ) ( -0.5,0 ) ( 3.5,0 ) to summarize , @xmath28 , @xmath35 and @xmath37 characterize the current state of the foam . these three tensors carry here the same information , since @xmath24 appears constant . in what follows , texture , elastic strain and stress tensors are always aligned . we choose to display @xmath35 only , because it is dimensionless , and thus more general : it makes the comparison of different materials easy . one possibility @xcite is to represent the traceless tensor @xmath35 as a circle of radius @xmath40 , with a straight line to indicate the direction @xmath41 of its positive eigenvalue : see thick lines ( circle and straight line ) in fig . [ fig : representationphysicalstress]a . we do not use it here , except in the inset of fig . [ fig : plaplasticlimit ] . in fact , it is easier to represent @xmath35 at a given time by a point , enabling us to plot trajectories . its two independent components can be represented in two different but equally useful ways , as follows . both representations are equally appropriate in the problem considered here because of the circular symmetry of the yield criterion ( see eq . [ eq : ucriterion ] ) . first , in the case of a traceless tensor , the absolute value of the two eigenvalues is the same and equal to the amplitude @xmath40 of the tensor @xmath35 defined as @xmath42 or equivalently @xmath43 , where @xmath44 is the euclidian norm of @xmath35 . we call @xmath41 the direction of the greatest eigenvalue . we call _ physical space _ the representation of the parameters ( @xmath40,@xmath41 ) . it is useful because it shows the evolution of the structure ( elongation , orientation ) , in particular , we plot the trajectory of the point ( @xmath45 ) ( fig . [ fig : representationphysicalstress]a ) . the other representation , which has already been used for foams @xcite , is called _ component space_. it plots the trajectory of the point ( @xmath46 ) ( fig . [ fig : representationphysicalstress]b ) . it is more suitable for comparison with experimental data , since rheometers measure the tangential stress ( @xmath38 ) , and sometimes the normal stress difference ( @xmath39 ) . these two possible choices are related by [ eq : uprincipalcomponent ] & u_xy & = u + & & = u. complete data for one simulation are plotted in fig . [ fig : simulationrepresentation ] and are discussed in the next section . ( 14,13)(0.0,0.0 ) ( -1.7,0 ) ( -1.2,0 ) ( -1.7,7.1 ) ( -1.2,7.1 ) ( 8.4,0 ) ( 8.5,0 ) ( 8.1,7 ) ( 8.5,7.1 ) as already mentioned , we consider a quasistatic limit in which flow is slow enough that we may neglect viscous stresses . the stress is then related to the elastic strain ( fig [ fig : verifutraceless]b ) . we do nt consider here the effect of external forces , such as gravity or friction on the boundary if the system is confined between glass plates @xcite . in the present 2d case , classical plasticity @xcite suggests that the material begins to yield when the difference between the stress eigenvalues becomes too large : @xmath47 . the yield stress @xmath48 separates a domain of pure elasticity from a domain in which the material flows plastically . a complete set of continuous equations ( reuss equations @xcite ) can then be derived ; @xmath49 is assumed to be constant ( no strain hardening ) . the effect of pressure ( trace of the stress ) is neglected , which is usually a good first approximation for metals for instance @xcite . it must be even more appropriate for soft materials , like a foam , for which the shear modulus is several orders of magnitude smaller than the bulk modulus @xcite . in what follows we prefer to use the component yield criterion @xcite : @xmath50 equivalently , eq . [ eq : componentyieldcriterion ] can be written for @xmath35 , since the deviatoric parts of @xmath37 and @xmath35 are proportional ( fig . [ fig : verifutraceless]b ) . from eqs . [ eq : uprincipalcomponent_3 ] and [ eq : componentyieldcriterion ] , we can write the yield criterion as @xmath51 this is represented by a circle in both physical and component spaces ( fig . [ fig : simulationrepresentation]bd ) for the example of foams and highly - concentrated emulsions , marmottant and graner @xcite suggested that the transition between elastic and plastic regimes is not sharp , but can be described by a yield function @xmath52 . this function is 0 , respectively 1 , in the pure elastic , respectively plastic , domain . between these two limits , both effects are present and the proportion seems to depend mostly on the elastic strain amplitude @xmath40 ( fig . [ fig : pla - h - n]a ) . this assumption was successfully tested on different flow geometries of a 2d foam @xcite . ( 8,5)(0.0,0.0 ) ( -0.4,-0.1 ) used as examples in the present paper are power laws @xmath53 with @xmath54 ( green , dash - dots ) , @xmath55 ( red , dashes ) , @xmath56 ( blue , dots ) and @xmath57 ( black , thick dashes , equal to 0 everywhere except at @xmath58 where it is equal to 1 ) . b ) corresponding limit cycles predicted by the model , plotted in _ component space _ ( same legend ) . , title="fig:",height=188 ] ( 5.8,0 ) used as examples in the present paper are power laws @xmath53 with @xmath54 ( green , dash - dots ) , @xmath55 ( red , dashes ) , @xmath56 ( blue , dots ) and @xmath57 ( black , thick dashes , equal to 0 everywhere except at @xmath58 where it is equal to 1 ) . b ) corresponding limit cycles predicted by the model , plotted in _ component space _ ( same legend ) . , title="fig:",height=188 ] ( 0,0 ) ( 5.8,0 ) by assuming that the deformation is affine @xcite and according to the prediction of plasticity ( eq . 22 in @xcite ) , we can then write a tensorial equation of evolution of the texture ( eq . 9 in @xcite ) . this dictates the evolution of the texture due to imposed strain ( deformation , rotation ) and due to relaxation ( rearrangements ) , for details see eqs . [ eq : fundament_3_ann],[eq : plasticmarmottant ] : @xmath59 here @xmath60 is the lagrangian derivative in time ( including advection ) ; @xmath61 is the variation of @xmath28 due to convection by the velocity gradient @xmath62 ; as explained in appendix [ sec : notationsfortensors ] and [ sec : tensorcomponents ] , the notation @xmath63 is the scalar product of the elastic strain tensor with the symmetrized velocity gradient tensor @xmath64 @xcite ; conversely , @xmath65 is the usual product of tensors ; here @xmath66 is the heaviside function , which is equal to 1 if @xmath63 is positive and 0 otherwise . [ eq : evolutionequation ] links the evolution of the foam texture with the elastic strain . it is quasistatic in the sense that the strain is relevant , not the strain - rate . it can be generalized to evolutions quicker than the relaxation times of the structure @xcite . plasticity occurs only when the elastic strain is oriented in the direction of shear , as expressed by the heaviside function @xmath67 ( appendix [ sec : elastoplasticcomponentequations ] ) . the model is continuous and analytic , without fluctuations . the information regarding disorder is recorded in @xmath52 . trapped stresses @xcite are recorded in the initial value @xmath68 ( or equivalently @xmath69 ) . the material s yielding criterion is encoded in @xmath70 . the history of the material only plays a role in determining @xmath52 , @xmath68 ( or @xmath69 ) , and @xmath70 , which together fully describe the material . according to the expression of @xmath52 , eq . [ eq : evolutionequation ] can be integrated analytically or numerically . to study the structure - evolution equation , _ i.e. _ the competition between elasticity and plasticity , and predict the rheological behaviour , we impose a strain rate @xmath71 on the material . we take @xmath72 as the direction of the shear , which gives the following velocity field : @xmath73 and hence @xmath74 this factor @xmath75 appears when comparing the scalar and tensorial descriptions ( appendix [ sec : annscalarlimit ] ) . in this geometry , the advection term is taken equal to zero and the resulting system of equations is given in appendix [ sec : elastoplasticcomponentequations ] . the reference state @xmath27 is considered isotropic and constant throughout the evolution . we recall that this evolution is quasistatic : @xmath71 appears as a prefactor in the time evolution , eqs . [ eq : ann2equationquasistatique ] . we thus follow the evolution with the strain @xmath76 , instead of the time . tensor operations and the time evolution of @xmath28 are implemented by a finite difference procedure . between two time steps , @xmath35 is recalculated with eq . [ eq : fundament_4_ann ] . we now address the resolution of the full elasto - plastic set of equations [ eq : ann2equationquasistatique ] . our representation underlines the specifically tensorial effects . ( 8,4)(0.0,0.0 ) ( -0.5,0 ) ( center , red ) and @xmath77 ( initial points scattered around the circle , black ) . here @xmath78 : all the trajectories evolve elastically . for @xmath79 , @xmath8 increases , that is , time evolve upwards on ( b ) . if @xmath80 , these graphs are unchanged , due to their symmetry with respect to the horizontal axis , showing that the purely elastic trajectories are reversible . , title="fig:",width=173 ] ( 3.8,0 ) ( center , red ) and @xmath77 ( initial points scattered around the circle , black ) . here @xmath78 : all the trajectories evolve elastically . for @xmath79 , @xmath8 increases , that is , time evolve upwards on ( b ) . if @xmath80 , these graphs are unchanged , due to their symmetry with respect to the horizontal axis , showing that the purely elastic trajectories are reversible . , title="fig:",width=185 ] ( 0,0 ) ( 4.3,0 ) as a first example of our representation , we consider here the pure elastic regime . this means that we allow the structure to deform elastically ( stretching , contraction ) , but not to relax plastically ( no rearrangements ) . our formalism allows us to describe pure elasticity by computing the elastic strain and its evolution when the material is deformed . our formalism extends to large strains , even those of order one ( for strains much larger than one , without plasticity , the formalism of large amplitude strain @xcite might be preferable ) . [ fig : modelcurvesuyelast]ab shows trajectories for different initial elastic strains . for an initially isotropic material , @xmath81 , we recover the classical results in the small strain limit : @xmath82 and @xmath83 . the poynting relation @xcite thus takes the form of a parabola ; we even extend it to an initially anisotropic material , @xmath84 ( see fig . [ fig : test ] and [ eq : approximationcomplete ] ) . for higher strains , @xmath8 is less linear with respect to @xmath9 , due to the rotation of the elastic strain . in all cases , @xmath8 increases monotonically . note that this is not the case for @xmath85 nor for @xmath40 . when the structure is aligned perpendicularly to the shearing direction ( @xmath86 ) , it contracts ( @xmath40 decreases ) under shear , until it aligns with the shear . when the structure is aligned parallel to the shearing direction ( @xmath87 ) , it stretches ( @xmath40 increases ) under shear . since @xmath35 is a tensor , it can continuously decrease , change direction and increase again without ever vanishing ( as opposed to a scalar , which can change sign only when it is equal to zero ) . for instance , a trajectory which starts with a direction opposed to that of shear has first a decreasing @xmath40 ( contraction , with @xmath8 negative and increasing ) , then an increasing @xmath40 ( stretching , with @xmath8 positive and increasing ) , then a constant @xmath40 ( yielding , with a rotation of @xmath35 towards the plastic limit ) . ( 7,5)(0.0,0.0 ) ( 0,-0.2 ) ( 1.4,0.8 ) the yield strain is the amplitude of the strain when the material yields , that is , a scalar number . the _ plastic limit _ is defined as the elastic strain tensor @xmath35 obtained after an infinitely long shearing ( @xmath88 ) . the amplitude of this tensor is that of the yield strain . its direction is obtained by solving eq . [ eq : evolutionequation ] when its left - hand side equals 0 , @xmath52 equals 1 , and @xmath67 equals 1 : [ eq : plasolutionplastique ] [ eq : plasolutionplastique_1 ] u & = & u_y , + [ eq : plasolutionplastique_2 ] & = & , + [ eq : plasolutionplastique_3 ] & = & ( ) . this plastic limit is represented on fig . [ fig : plaplasticlimit ] . it shows that the larger @xmath70 , the less aligned @xmath35 is with respect to @xmath89 . this tensorial effect is strong because @xmath90 decreases quickly with @xmath70 . the scalar limit corresponds to @xmath91 , as discussed in sec . [ sec : comparisonscalartensorial ] . ( 8,12)(0.0,0.0 ) ( 0,8 ) in _ physical _ ( a , c , e ) and _ component _ ( b , d , f ) spaces . if @xmath80 , the vertical axis of all these graphs should be inverted . ( a - b ) for @xmath78 , all the trajectories evolve elastically ( @xmath8 increases , that is , time evolves in the direction of the arrows ) up to the yield strain , here taken as @xmath92 , then evolve plastically ( @xmath40 constant ) towards the plastic limit ( blue point ) which corresponds to a specific angle , see eq . [ eq : plasolutionplastique ] . if the trajectory reaches the yield strain on the left of the plastic limit , @xmath8 passes through a maximum ( overshoot ) . ( c - d ) for @xmath93 , plasticity appears more progressively , thus smoothening the transition between elastic and plastic regimes , and decreasing ( or even suppressing ) the overshoot . ( e - f ) summary of the mechanical behavior : stretching " and contraction " according to the direction of @xmath35 with respect to shear . here plasticity on " or plasticity off " refers to the heaviside function in the last term of eq . [ eq : evolutionequation ] , when the plasticity is progressive ( @xmath94 finite ) ; when @xmath94 increases the plasticity on " zone narrows , and for @xmath94 infinite it is reduced to the limit circle . , title="fig:",width=151 ] ( 4,8 ) in _ physical _ ( a , c , e ) and _ component _ ( b , d , f ) spaces . if @xmath80 , the vertical axis of all these graphs should be inverted . ( a - b ) for @xmath78 , all the trajectories evolve elastically ( @xmath8 increases , that is , time evolves in the direction of the arrows ) up to the yield strain , here taken as @xmath92 , then evolve plastically ( @xmath40 constant ) towards the plastic limit ( blue point ) which corresponds to a specific angle , see eq . [ eq : plasolutionplastique ] . if the trajectory reaches the yield strain on the left of the plastic limit , @xmath8 passes through a maximum ( overshoot ) . ( c - d ) for @xmath93 , plasticity appears more progressively , thus smoothening the transition between elastic and plastic regimes , and decreasing ( or even suppressing ) the overshoot . ( e - f ) summary of the mechanical behavior : stretching " and contraction " according to the direction of @xmath35 with respect to shear . here plasticity on " or plasticity off " refers to the heaviside function in the last term of eq . [ eq : evolutionequation ] , when the plasticity is progressive ( @xmath94 finite ) ; when @xmath94 increases the plasticity on " zone narrows , and for @xmath94 infinite it is reduced to the limit circle . , title="fig:",width=151 ] ( 0,4 ) in _ physical _ ( a , c , e ) and _ component _ ( b , d , f ) spaces . if @xmath80 , the vertical axis of all these graphs should be inverted . ( a - b ) for @xmath78 , all the trajectories evolve elastically ( @xmath8 increases , that is , time evolves in the direction of the arrows ) up to the yield strain , here taken as @xmath92 , then evolve plastically ( @xmath40 constant ) towards the plastic limit ( blue point ) which corresponds to a specific angle , see eq . [ eq : plasolutionplastique ] . if the trajectory reaches the yield strain on the left of the plastic limit , @xmath8 passes through a maximum ( overshoot ) . ( c - d ) for @xmath93 , plasticity appears more progressively , thus smoothening the transition between elastic and plastic regimes , and decreasing ( or even suppressing ) the overshoot . ( e - f ) summary of the mechanical behavior : stretching " and contraction " according to the direction of @xmath35 with respect to shear . here plasticity on " or plasticity off " refers to the heaviside function in the last term of eq . [ eq : evolutionequation ] , when the plasticity is progressive ( @xmath94 finite ) ; when @xmath94 increases the plasticity on " zone narrows , and for @xmath94 infinite it is reduced to the limit circle . , title="fig:",width=151 ] ( 4,4 ) in _ physical _ ( a , c , e ) and _ component _ ( b , d , f ) spaces . if @xmath80 , the vertical axis of all these graphs should be inverted . ( a - b ) for @xmath78 , all the trajectories evolve elastically ( @xmath8 increases , that is , time evolves in the direction of the arrows ) up to the yield strain , here taken as @xmath92 , then evolve plastically ( @xmath40 constant ) towards the plastic limit ( blue point ) which corresponds to a specific angle , see eq . [ eq : plasolutionplastique ] . if the trajectory reaches the yield strain on the left of the plastic limit , @xmath8 passes through a maximum ( overshoot ) . ( c - d ) for @xmath93 , plasticity appears more progressively , thus smoothening the transition between elastic and plastic regimes , and decreasing ( or even suppressing ) the overshoot . ( e - f ) summary of the mechanical behavior : stretching " and contraction " according to the direction of @xmath35 with respect to shear . here plasticity on " or plasticity off " refers to the heaviside function in the last term of eq . [ eq : evolutionequation ] , when the plasticity is progressive ( @xmath94 finite ) ; when @xmath94 increases the plasticity on " zone narrows , and for @xmath94 infinite it is reduced to the limit circle . , title="fig:",width=151 ] ( 0,0 ) in _ physical _ ( a , c , e ) and _ component _ ( b , d , f ) spaces . if @xmath80 , the vertical axis of all these graphs should be inverted . ( a - b ) for @xmath78 , all the trajectories evolve elastically ( @xmath8 increases , that is , time evolves in the direction of the arrows ) up to the yield strain , here taken as @xmath92 , then evolve plastically ( @xmath40 constant ) towards the plastic limit ( blue point ) which corresponds to a specific angle , see eq . [ eq : plasolutionplastique ] . if the trajectory reaches the yield strain on the left of the plastic limit , @xmath8 passes through a maximum ( overshoot ) . ( c - d ) for @xmath93 , plasticity appears more progressively , thus smoothening the transition between elastic and plastic regimes , and decreasing ( or even suppressing ) the overshoot . ( e - f ) summary of the mechanical behavior : stretching " and contraction " according to the direction of @xmath35 with respect to shear . here plasticity on " or plasticity off " refers to the heaviside function in the last term of eq . [ eq : evolutionequation ] , when the plasticity is progressive ( @xmath94 finite ) ; when @xmath94 increases the plasticity on " zone narrows , and for @xmath94 infinite it is reduced to the limit circle . , title="fig:",width=151 ] ( 4,0 ) in _ physical _ ( a , c , e ) and _ component _ ( b , d , f ) spaces . if @xmath80 , the vertical axis of all these graphs should be inverted . ( a - b ) for @xmath78 , all the trajectories evolve elastically ( @xmath8 increases , that is , time evolves in the direction of the arrows ) up to the yield strain , here taken as @xmath92 , then evolve plastically ( @xmath40 constant ) towards the plastic limit ( blue point ) which corresponds to a specific angle , see eq . [ eq : plasolutionplastique ] . if the trajectory reaches the yield strain on the left of the plastic limit , @xmath8 passes through a maximum ( overshoot ) . ( c - d ) for @xmath93 , plasticity appears more progressively , thus smoothening the transition between elastic and plastic regimes , and decreasing ( or even suppressing ) the overshoot . ( e - f ) summary of the mechanical behavior : stretching " and contraction " according to the direction of @xmath35 with respect to shear . here plasticity on " or plasticity off " refers to the heaviside function in the last term of eq . [ eq : evolutionequation ] , when the plasticity is progressive ( @xmath94 finite ) ; when @xmath94 increases the plasticity on " zone narrows , and for @xmath94 infinite it is reduced to the limit circle . , title="fig:",width=151 ] ( 0,8 ) ( 4,8 ) ( 0,4 ) ( 4,4 ) ( 0,0 ) ( 4,0 ) we now consider the shearing of an initially anisotropic elastic strain , @xmath95 . in _ physical _ or _ component space _ , the state of the material is initially situated on an elastic trajectory and must arrive at the plastic limit point ( fig . [ fig : modelcurvesuyplast]a - d ) . at this limit point , an increase of elastic strain is immediately transformed into plastic strain . plasticity may occur only if @xmath87 . graphically , both in _ physical space _ and in _ component space _ [ fig : modelcurvesuyplast ] ) , the plastic limit is represented by the point where a trajectory reaches perpendicularly the circle at @xmath96 ( the elastic strain increases along the tangent of the trajectory , the plastic strain relaxes towards the centre of the circle : to balance each other , they must be parallel ) . the shape of the yield function @xmath52 then determines how the material reaches the plastic limit . for simplicity , we take @xmath52 as a power law function : @xmath97 ( fig . [ fig : pla - h - n]a ) . two examples of the resulting behaviour are plotted on fig . [ fig : modelcurvesuyplast ] . the limit @xmath98 is intuitive : the material follows the elastic trajectory up to @xmath96 ; then @xmath40 is fixed and the plastic limit is reached by describing an arc of a circle in _ physical _ and _ component spaces _ [ fig : modelcurvesuyplast]ab ) . for other cases ( finite @xmath94 ) plasticity occurs earlier ( fig . [ fig : modelcurvesuyplast]ef ) and trajectories converge to the plastic limit ( fig . [ fig : modelcurvesuyplast]cd ) . the behaviour changes qualitatively if the sign of @xmath71 is abruptly reversed . unlike the elastic term , the plasticity term is irreversible due to the heaviside function @xmath99 in eq . [ eq : evolutionequation ] . this leads to an inversion of the plastic domain in _ physical _ and _ component spaces _ ( fig . [ fig : modelcurvesuyplast]ef ) and to a new plastic limit position ( @xmath100 , eq . [ eq : plasolutionplastique ] ) . this hysteretic effect is shown on fig . [ fig : simulationrepresentation ] by reversing @xmath71 once the plastic limit is reached . for high @xmath94 , the new plastic limit is quickly reached , since the two plastic limits are on the same elastic trajectory . if we perform alternate sign changes of @xmath71 , we observe that the material is stuck in a limit trajectory ( fig . [ fig : pla - h - n]b ) . this trajectory is almost insensitive to @xmath52 and therefore close to the elastic trajectory joining the two plastic limits . this has an important consequence : once in the plastic regime , the elastic strain ( and thus the stress ) can not be totally relaxed if we only reverse the shearing direction . this is examined in more detail below ( fig . [ fig : relaxationcycles ] ) . \a ) versus @xmath9 curve , and the value of the plateau which follows this maximum . calculations are performed in the case of an initially isotropic structure ( @xmath81 ) . ( a ) zoom over the maxima of the curves of fig . [ fig : simulationrepresentation]a . ( b ) maxima for different @xmath52 functions ( same legend as fig . [ fig : pla - h - n ] ) , and plateau value ( starred line , which is the same for all @xmath52 functions ) , plotted _ vs _ @xmath70 . simulation points are plotted for comparison , with the same symbols as in table [ tab : defsimul ] ; closed symbols and @xmath101 correspond to the maximal value ( averaged over a few successive points ) during the first shear step ; open symbols and @xmath15 correspond to the averaged value along the plateau of the last shear step ; @xmath70 is measured as the plateau value of @xmath40 . ( c ) zoom of ( b ) . , title="fig:",width=264 ] + b ) versus @xmath9 curve , and the value of the plateau which follows this maximum . calculations are performed in the case of an initially isotropic structure ( @xmath81 ) . ( a ) zoom over the maxima of the curves of fig . [ fig : simulationrepresentation]a . ( b ) maxima for different @xmath52 functions ( same legend as fig . [ fig : pla - h - n ] ) , and plateau value ( starred line , which is the same for all @xmath52 functions ) , plotted _ vs _ @xmath70 . simulation points are plotted for comparison , with the same symbols as in table [ tab : defsimul ] ; closed symbols and @xmath101 correspond to the maximal value ( averaged over a few successive points ) during the first shear step ; open symbols and @xmath15 correspond to the averaged value along the plateau of the last shear step ; @xmath70 is measured as the plateau value of @xmath40 . ( c ) zoom of ( b ) . , title="fig:",width=264 ] + c ) versus @xmath9 curve , and the value of the plateau which follows this maximum . calculations are performed in the case of an initially isotropic structure ( @xmath81 ) . ( a ) zoom over the maxima of the curves of fig . [ fig : simulationrepresentation]a . ( b ) maxima for different @xmath52 functions ( same legend as fig . [ fig : pla - h - n ] ) , and plateau value ( starred line , which is the same for all @xmath52 functions ) , plotted _ vs _ @xmath70 . simulation points are plotted for comparison , with the same symbols as in table [ tab : defsimul ] ; closed symbols and @xmath101 correspond to the maximal value ( averaged over a few successive points ) during the first shear step ; open symbols and @xmath15 correspond to the averaged value along the plateau of the last shear step ; @xmath70 is measured as the plateau value of @xmath40 . ( c ) zoom of ( b ) . , title="fig:",width=264 ] + as observed for @xmath78 ( fig . [ fig : modelcurvesuyplast]b ) , the overshoot is due to the transition from an elastic trajectory to the plastic limit . the structure itself has no overshoot : @xmath40 increases monotonically . the overshoot appears in the tangential strain @xmath8 : it is a purely tensorial effect due to a rotation of the structure . in fact , @xmath8 increases in all elastic trajectories . upon reaching @xmath58 there is a sudden transition to the plastic regime . for trajectories to the left of the plastic limit ( fig . [ fig : modelcurvesuyplast]b ) , we see that @xmath8 decreases towards the plastic limit . the overshoot corresponds to the difference between the maximum value of @xmath8 ( where the trajectory meets the circle ) and the plateau value ( plastic limit ) . > from fig . [ fig : modelcurvesuyplast]b , we observe a tiny overshoot for the normal stress difference if the trajectories reach the @xmath58 circle to the right of the plastic limit . in that case , the trajectories move towards the right in the elastic regime , then towards the left in the plastic regime . such trajectories correspond to the right of fig . [ fig : modelcurvesuyplast]b , that is , a structure with a large trapped normal stress difference @xmath102 . for smaller @xmath94 , the elastic - plastic transition is smoother and the overshoot is reduced . the overshoot amplitude for different @xmath52 and @xmath70 is plotted on fig . [ fig : plaovershootus ] , for the case of an initially isotropic structure ( @xmath81 ) . the overshoot increases with @xmath70 because the plastic limit moves away from the initial elastic trajectory . the model ( section [ sec : transientregim ] ) predicts that after a few cycles a limit trajectory is reached . this trajectory is also displayed in simulations by kabla and debregeas @xcite . the plastic limit can thus be evaluated in simulations : @xmath70 is estimated by averaging @xmath40 on the last plateau ( using any of fig . [ fig : simulationrepresentation]a - d ) ; similarly , @xmath90 is estimated by averaging @xmath41 on the last plateau of fig . [ fig : simulationrepresentation]b or d. on fig . [ fig : plaplasticlimit ] , results from simulations are compared with the model . agreement is good , and the model captures tensorial effects , especially because the measured @xmath90 deviate much from the @xmath103 scalar limit . taking larger @xmath3 in the simulations favours neighbour swappings ( t1s " ) , thus it corresponds to an increased effective liquid fraction . as expected , we see a decrease in @xmath70 . however , since the effective liquid fraction we are simulating remains in a very dry range ( @xmath104 ) , it does not influence much @xmath70 , which thus varies over a narrow range ( @xmath105 ) . reaching higher @xmath70 is possible with other materials , but not with disordered 2d foams . reaching lower @xmath70 is possible ( and usual ) in experiments on disordered wet foams , but not in the present simulations where the algorithm would require adaptation at high @xmath3 . simulation results fluctuate , due to the limited number of bubbles ( discrete description ) , while model curves are smooth , corresponding to the limit of a large number of bubbles ( continuous material description ) . there is a qualitative agreement , which is good enough to deduce @xmath70 and @xmath52 approximately . for instance , fig . [ fig : simulationrepresentation ] compares a simulation with models using various @xmath52 functions . we observe that @xmath106 ( red dashes on fig . [ fig : simulationrepresentation ] ) describes well the simulation during the first positive shearing step , and @xmath107 ( blue dotted lines on fig . [ fig : simulationrepresentation ] ) during the second one . similarly , @xmath70 is deduced from the plateau value of @xmath40 ( figs . [ fig : plaplasticlimit ] and [ fig : plaovershootus ] ) . in practice , in a first approximation , it is enough to consider @xmath70 and @xmath52 as constant . their variations are small and thus have a small effect on the foam rheology . however , these variations do exist . for instance , in this example of fig . [ fig : simulationrepresentation ] , @xmath94 ( and thus @xmath52 ) evolves throughout the simulation , revealing that the structure evolves too ; @xmath52 seems to be sensitive to the ( topological ) disorder of the foam . here @xmath70 is constant , but there are other cases ( data not shown , see @xcite ) where , due to the decrease of the topological disorder during the shearing , @xmath70 decreases . more generally , a real foam is constantly evolving under the effect of drainage , coarsening @xcite , or shuffling . these effects should probably have to be considered in future models , which would try to predict @xmath70 and @xmath52 , based on the average and fluctuations of the structure , respectively . we can now identify two distinct physical mechanisms which can cause a stress overshoot in shear experiments of elasto - plastic materials . the first one is an _ orientation _ effect , suggested in sec . [ sec : overshoot ] . @xmath40 increases monotonically , but if @xmath70 is large enough then the rotation of @xmath35 under shear implies that the tangential shear strain @xmath8 passes through a maximum . this purely tensorial effect is absent from scalar models . under certain additional conditions on the initial elastic strain @xmath108 , which are also described correctly only when taking into account the tensorial aspects , the normal strain difference @xmath102 too passes through a maximum . [ fig : plaovershootus ] shows a comparison between the model and the simulations . given that in the range of simulated @xmath70 the overshoot is tiny and difficult to extract from the fluctuations , the agreement is surprisingly good . in most foam experiments , where @xmath70 is even lower , this effect should be too small to be measurable . the second one is outside of the scope of the present paper . it is due to an evolution of the _ structure _ itself during the first shear step ( see section [ sec : evolutionofthestructure ] ) . this might be invoked to explain the larger overshoot of the data corresponding to the confined simulations ( red and pink ) , as well as most experimental observations ( such as that of ref . as long as the applied shear keeps a constant direction , and the elastic strain remains much smaller than 1 , its eigenvectors correspond to that of @xmath89 . that is , they are at @xmath103 to the direction of shear . this is called the _ scalar approximation _ , and it considerably simplifies the study of the mechanical behaviour . in that case , a single ( scalar ) number is enough to fully describe the elastic strain . this scalar number can equally well be chosen as the amplitude @xmath40 , or the eigenvalue @xmath16 , or the tangential shear strain @xmath8 , among others . to switch from one choice to the other requires care regarding the prefactors @xcite : this is often a source of confusion in the literature , especially regarding the definition and value of the yield strain . the link between the simplified ( scalar ) and complete ( tensorial ) equations is detailed in appendix [ sec : annscalarlimit ] , using @xmath109 as a scalar . if @xmath110 , which is the case for wet foams and emulsions , then @xmath40 remains always much smaller than 1 , and @xmath111 , so that the scalar approximation holds , see fig . 19 in ref . @xcite ( except if the direction of the shear changes , in 2d or in 3d ) . in that limit tensorial effects such as normal differences or stress overshoot are negligible . quantitatively , @xmath8 is linked to @xmath112 ( eq . [ eq : uprincipalcomponent_1 ] ) . this implies that a difference of 10% between the scalar and tensorial equations is reached when @xmath113 , which corresponds to @xmath114 ( eq . [ eq : plasolutionplastique ] ) . very dry foams , such as those simulated here , are slightly above this limit : a tensorial model is therefore useful . \a ) for the first step and @xmath115 , @xmath78 . between two steps : the direction of shearing is turned @xmath116 clockwise ( blue dots ) ; the amplitude is decreased from @xmath117 to 0 in 15 steps ( black dashes ) ; simultaneously , the direction of shear is turned and its amplitude is decreased ( red solid line ) . a ) component space . b ) @xmath118 versus time ( @xmath119s@xmath120).,title="fig:",width=264 ] + b ) for the first step and @xmath115 , @xmath78 . between two steps : the direction of shearing is turned @xmath116 clockwise ( blue dots ) ; the amplitude is decreased from @xmath117 to 0 in 15 steps ( black dashes ) ; simultaneously , the direction of shear is turned and its amplitude is decreased ( red solid line ) . a ) component space . b ) @xmath118 versus time ( @xmath119s@xmath120).,title="fig:",width=264 ] + a dry foam is a material with sufficiently high @xmath70 that normal stresses may exist even when the material is at rest @xcite . to relax such residual ( or trapped " ) stresses , we should first shear the foam enough to reach the plastic stage , so that plastic rearrangements anneal the disorder . we then must perform cycles of shear . if the direction of shear is kept constant , and the shear simply reversed , the foam asymptotically reaches a limit trajectory , and the stress is not relaxed . decreasing the amplitude of the shear cycle does not enable to leave this limit trajectory ( black dashes in fig . [ fig : relaxationcycles]ab ) . simulated dry 3d foam and applied shearing cycles ( actually uniaxial contractions ) of amplitude @xmath121 in different directions , rotated by @xmath116 ; this procedure decreases the trapped stress by a factor of around 2 , which does not improve with more cycles ( fig . 7 of @xcite ) . here we propose a reproducible procedure based on section [ sec : transientregim ] , which couples shearing cycles in different directions and decreasing amplitudes , as follows : * the amplitude @xmath122 of the first step is large enough to completely reach the plastic stage : @xmath123 . * at each step , the shearing direction is rotated by @xmath116 and the shearing amplitude is decreased . * the decrease in amplitude between successive steps is smaller than @xmath124 , ensuring there are at least 5 steps between @xmath125 and 0 ( _ i.e. _ the total number of steps is at least @xmath126 ) . the red solid line in fig . [ fig : relaxationcycles]ab shows that the normal stress difference decreases more at each cycle ; for instance , 6 cycles yield a decrease by a factor of 10 , apparently without saturating . in 3d the procedure is the same , rotating the shearing direction successively along the @xmath72 , @xmath127 and @xmath128 axes @xcite . this procedure is easy to apply to simulations , especially of fully periodic foams . in experiments , a special set - up should be built : in 2d , it can be a rubber frame in the spirit of refs . @xcite , if the four corners can be independently displaced . ( 7,4.5)(0.0,0.0 ) ( -0.7,-0.1 ) ( 3.5,0 ) ( -0.5,0 ) ( 4,0 ) ( 8,4.5)(0.0,0.0 ) ( 0,-0.5 ) for practical purposes we plot the reference curves for two limiting types of materials : those with @xmath70 much higher or much lower than 0.23 . fig . [ fig : bigovershootu1 ] shows the example of @xmath129 . the plastic limit corresponds to a small angle @xmath90 , resulting in a strong overshoot . [ fig : test ] shows that for small strains in the elastic regime , all curves can be expressed using a single parameter ; for instance , as here , @xmath130 , which is the normal elastic strain @xmath131 at zero tangential shear ( @xmath132 ) . a rough parabolic approximation , and a refined one ( eq . [ eq : approximationcomplete ] ) are plotted here for @xmath92 . for smaller @xmath70 , this approximation is good over its whole range of validity ( namely the elastic regime ) , but this range is smaller . we propose a continuous model of the elasticity and plasticity of disordered , discrete materials such as cellular patterns ( for instance liquid foams or emulsions ) and assemblies of particles ( for instance colloids ) . it is based on statistical quantities including ( i ) the elastic strain @xmath35 , a dimensionless quantity measurable on images , which facilitates the comparison between different experiments or models , and makes apparent the effect of shear on the material s structure ; ( ii ) the yield strain @xmath40 , a classical criterion for the transition between reversible , elastic and irreversible , plastic regimes ; ( iii ) and the yield function @xmath133 , which describes how progressive this transition is , by measuring the relative proportion of elastic and plastic deformation . they suffice to relate the discrete scale with the collective , global scale . at this global scale , the material behaves as a continuous medium ; it is described with tensors such as strain , stress and velocity gradient . we give the differential equations which predict the elastic and plastic behaviour . the model is fully tensorial and thus general , in 2d or in 3d . we study in detail the case of simple shear . an original representation , suitable for 2d incompressible materials , is introduced to follow the evolution of the material during shear . since @xmath35 is a tensor , it has an orientation and an amplitude , which both evolve under shear . it can continuously decrease its amplitude , change direction and increase again its amplitude without ever vanishing ( as opposed to a scalar , which can change sign only when it is equal to zero ) . predictions of the model regarding orientation and stretching are plotted . they include a rotation of the structure , which can induce an overshoot of the shear strain or shear stress ( and a smaller , rarer overshoot in normal stress differences ) even without overshoot in the elastic strain amplitude . this purely tensorial effect exists if @xmath70 is at least of order of 0.3 . independently , the shear can also induce a change in the material s structure , sometimes resulting in a ( purely scalar ) overshoot in the modulus of the elastic strain . the model extends a classical plasticity criterion to disordered media . it can be solved numerically and yields testable predictions . we successfully compare them with carefully converged quasistatic simulations of shear cycles in 2d foams : the elastic strain increases , saturates and reverses . from this comparison between model and simulation we determine @xmath70 and estimate @xmath52 . this method is similar to that which we have used in experiments to extract @xmath70 @xcite , and a rough estimate of @xmath52 . we still lack a model to predict @xmath70 and @xmath52 . both quantities evolve throughout the simulation , probably due to the evolution of the foam s internal structure , as well as the disorder and fluctuations . in short , the material obeys a continuous description determined by its average properties , while @xmath70 and @xmath52 account for the effect at large scale of its fluctuations . all quantities involved in the model are directly measurable , as tensors , in the current state of the material ; this includes trapped stresses which we discuss ( we also explain how to relax them ) : the history of the sample which led to this current state plays no other direct , explicit role . we explain how and when to use the model in practice , and provide a set of curves and analytical approximations , including a discussion and an extension of the poynting relation . at low strain , typically below 0.2 , tensorial effects vanish and an approximate scalar simplification holds . we thank c. quilliet , b. dollet , s. attai talebi and other participants in the grenoble foam mechanics workshop 2008 for stimulation and useful discussions . we thank k. brakke for his development and maintenance of the surface evolver code . fg thanks alexandre kabla for critical reading of the manuscript and recalling the link between overshoot and bistability . sjc thanks the british council alliance programme , cnrs and epsrc ( ep / d048397/1 , ep / d071127/1 ) for financial support and ujf for hospitality during the period in which this work was conceived . cr thanks the alliance programme for having supported one visit to aberystwyth university , project 15154xb _ foam rheology in two dimensions_. we collect here a list of our notation , since the definitions are scattered throughout the text . for a symmetric tensor @xmath134 , we denote by @xmath135 and @xmath136 its eigenvalues , by @xmath137 its trace , by @xmath138 its off - diagonal term , by @xmath139 half its normal difference , by @xmath140 its amplitude , and by @xmath141 its euclidian norm defined as @xmath142 . the scalar product of two tensors @xmath134 and @xmath143 is defined as @xmath144 . for a traceless tensor , if @xmath41 is the angle corresponding to @xmath135 , then the point with coordinates @xmath146 is a representation of the actual direction of the tensor ( _ physical space _ ) . conversely , the point with coordinates @xmath147 and @xmath148 directly represents the components of the tensor ( _ component space _ ) . for a tensor , @xmath41 is defined modulo @xmath149 ( and not @xmath150 as for vectors ) , so that @xmath151 has usually more relevance than @xmath41 . similarly , for traceless tensors with eigendirections making a relative angle @xmath152 , their scalar product is proportional to @xmath153 . this scalar product is maximal when the two eigenvectors of the positive eigenvalues coincide , and minimized when they are perpendicular . [ eq : fundament_1_ann ] is the equation of dynamics , equivalent to navier - stokes , except that here the viscous stress is assumed to be negligible compared to the elastic stress . [ eq : fundament_2_ann ] assumes that the flow is incompressible ; this assumption is often valid for foams at small deformation but can be relaxed if needed . [ eq : fundament_4_ann ] defines the elastic strain from the texture @xcite , that is , it assumes that each bubble s internal degrees of freedom depend on its shape . [ eq : fundament_3_ann ] is the evolution of the texture , see eq . [ eq : evolutionequation ] for the definitions of its terms ( transport and source ) . here the plasticity rate @xmath154 is predicted according to eq . 22 in ref . @xcite : @xmath155 the meaning of each term is the following . the direction of @xmath156 is set by that of @xmath35 , indicating that the plasticity is opposed to the increase of @xmath35 . the amplitude of @xmath156 , that is the rate of plastic rearrangements , is the inverse of a time . it is determined by the total strain rate @xmath89 ; more precisely , by one component of @xmath89 , determined by the scalar product with @xmath35 ( and only if this scalar product is positive , as expressed by the heaviside function @xmath67 ) . finally , the amplitude of @xmath156 depends on the yield criterion , as expressed by @xmath157 : the plasticity appears ( progressively or abruptly ) when @xmath40 approaches then exceeds the yield strain . [ eq : plasticmarmottant ] is written here by assuming that @xmath28 and @xmath35 commute ( see eq . 20 of ref . @xcite ) , which is always the case if @xmath158 is isotropic . like [ eq : fundament_2_ann ] , it assumes that the flow is incompressible , but can be extended to more general cases . it also assumes that the flow is slow : see ref . @xcite for a discussion of quasistatic " flow , and @xcite for the extension to higher velocity . the next appendices examine more restrictive cases , that is , additional approximations : simple shear ( [ sec : tensorcomponents ] ) , small strain ( [ sec : annscalarlimit ] ) , and the purely elastic regime ( [ sec : analyticalapproximation ] ) . here , due to our conventions , the angle between both tensors is @xmath160 , hence the term @xmath161 this scalar product is maximal when the two eigenvectors of the positive eigenvalues coincide ( which happens for @xmath162 ) , and minimized when they are perpendicular ( @xmath163 ) . the elastic regime can be studied by taking the last term of these equations equal to 0 ( limit of high @xmath70 ) . the plastic limit is calculated by taking the left hand sides of these equations equal to 0 , @xmath164 , and @xmath165 . in the limit of small strain , @xmath35 can be linearized : @xmath166 where @xmath167 is the isotropic eigenvalue of @xmath27 . in that limit eq . [ eq : ann2equationquasistatique_3 ] becomes @xmath168 assuming that @xmath41 remains close to @xmath103 leads to in a purely elastic regime , the evolution equation for the texture is @xmath170 to express all curves analytically , we choose a single parameter , for instance the elastic strain in a non - sheared state ( @xmath171 ) : @xmath172 @xmath173 which can be rewritten by eliminating @xmath9 : @xmath174 there are still two constants left , @xmath175 and @xmath176 . to eliminate one of them , we use the fact that the trace of @xmath35 is almost zero , and thus the determinant of @xmath28 is almost constant : @xmath177 or equivalently , using eq . [ eq : step1 ] : @xmath178 solving eq . [ determin ] yields @xmath179 or equivalently , eliminating @xmath176 using eq . [ determin ] : @xmath180 coming back to @xmath35 using eq . [ eq : linearized ] : @xmath181 eq . [ m_n ] yields a parabolic approximation : @xmath182 the parameter which determines each elasticity curve is the normal strain difference at zero shear ( which is thus equal to the amplitude of elastic strain at zero shear ) [ eq : approximationcomplete ] is tested on fig . [ fig : test ] for @xmath40 up to 0.3 . the prefactor of the parabola , _ i.e. _ the bracket in eq . [ eq : approximationcomplete ] , is exactly 1 if @xmath183 : this is the poynting relation @xcite ( black curve on fig . [ fig : test ] , starting from the point @xmath184 ) . in fact , even for @xmath185 , the bracket in eq . [ eq : approximationcomplete ] remains close to 1 : as shown in fig . [ fig : test ] , the poynting relation extends even to an initially anisotropic material .

turbo codes are a very efficient method for communicating reliably through a noisy channel . there is no theoretical understanding of their effectiveness . in ref . they are mapped onto a class of disordered spin models . the analytical calculations concerning these models are reported here . we prove the existence of a no - error phase and compute its local stability threshold . as a byproduct , we gain some insight into the dynamics of the decoding algorithm . lptens 00/13

cond-mat0003218

communication through a noisy channel is a central problem in information theory @xcite . error correcting codes are a widespread method for compensating the information corruption due to the noise , by cleverly increasing the redundancy of the message . turbo codes @xcite are a recently invented class of error correcting codes with nearly optimal performances . they allows reliable communication ( i.e. very low error per bit probability ) with practical communication rates . it is known , since the work of sourlas @xcite , that there exists a close relationship between the statistical behavior of error correcting codes and the physics of some disordered spin models . recently the tools developed in statistical physics have been employed in studying gallager - type codes @xcite . in ref . @xcite the equivalence discovered by sourlas is extended to turbo codes , and the basic features of the corresponding spin models are outlined . a remarkable property of a large family of turbo codes , presented in ref . @xcite , is the existence of a no - error phase . in other words the error probability per bit vanishes beyond some critical ( finite ) signal to noise ratio . in ref . @xcite some intuitive arguments supporting this thesis are given . some analytical results concerning the critical value of the signal to noise ratio are announced without giving any derivation . these results are compared with numerical simulations . in this paper we present the analytical results in their full generality , and explain their derivation . we prove the existence of the no - error phase and find the condition for its local stability . this condition is derived in two different approaches . in the first one we study the asymptotic dynamics of the decoding algorithm . in the second approach we use replicas and establish the condition for stability in the full replica space . local stability is a necessary but not sufficient condition for the stability of the no - error phase . the critical signal to noise ratio obtained from local stability is the correct one only if the phase transition is a second order one : in the general case it is only a lower bound . the spin models which are equivalent to turbo codes have the following statistical weight @xcite : @xmath0 the dynamical variables of the model are the spins @xmath1 with @xmath2 . we shall choose them to be ising spins , that is @xmath3 . the spins enters in the hamiltonians @xmath4 through the local interaction terms @xmath5 and @xmath6 which are products of @xmath7 s . their exact form can be encoded in two set of numbers @xmath8 and @xmath9 as follows : @xmath10 and @xmath11 . in order to fix completely our notation we set @xmath12 . the quenched variables are : * the couplings @xmath13 , whose distribution @xmath14 satisfies the conditions @xmath15 and @xmath16 ; * the permutation @xmath17 , which has uniform distribution . it is convenient to impose a fixed boundary condition at one end of the chain ( i.e. @xmath18 for @xmath19 ) and a free boundary condition at the other end . the model is composed by two one dimensional substructures ( chains ) , which interact through the kronecker delta functions in eq . ( [ models1 ] ) . when the average over permutations is taken into account this interaction turns into a mean field one . this interplay between the two subsystems , each one possessing a one dimensional structure , and the mean field interaction which couples them is clearly displayed by the analytical calculations . for further explanations on eqs . ( [ models1]-[models2 ] ) and their motivation we refer to @xcite . the paper is organised as follows . in sections [ withoutreplicas1 ] and [ withoutreplicas2 ] we present a first derivation of the stability condition . we write a `` mean field '' equation which describes the dynamics of the decoding algorithm ( sec . [ withoutreplicas1 ] ) , we show that it possesses a no - error fixed point and then study its behavior in a neighbourhood of this fixed point ( sec . [ withoutreplicas2 ] ) . thanks to this derivation we will understand how this fixed point is reached . in section [ withreplicas ] replicas are introduced in order to compute the average over the permutations . we exhibit the no - error saddle point . in section [ stability ] the stability of the no - error saddle point is studied by diagonalizing the second derivative of the free energy . finally in section [ conclusion ] the validity of our calculations is discussed . appendix [ appendicealgebra ] collects some useful ( although simple ) facts of algebra . in appendix [ appendiceprobabilita ] the type of integral equations which appear in section [ withoutreplicas2 ] is studied in detail . [ section ] some properties concerning the models defined by eqs . ( [ models1]-[models2 ] ) can be obtained by considering the `` turbo decoding '' algorithm and making some factorization hypothesis . these hypothesis enable us to obtain a recursive integral equation for the probability distribution of a local field . they can be justified on heuristic grounds and arguments of this kind will be given later in this section . moreover the replica calculation presented in the section [ withreplicas ] does support our arguments . in particular this approach allows us to derive the critical noise below which `` perfect '' decoding is possible . turbo decoding is an iterative algorithm . the iteration variables are the fields @xmath20 with @xmath2 . the step @xmath21 of the turbo decoding algorithm is defined as follows @xcite : @xmath22 -\gamma_{\rho^{-1}(i)}^{(2)}(t)\label{turbodecoding1}\\ \gamma_i^{(2)}(t+1 ) & = & \frac{1}{\beta}\arctanh\left [ \<\epsilon_{\rho(i)}(\sg)\>^{(1)}_{\gamma^{(1)}(t)}\right ] -\gamma_{\rho(i)}^{(1)}(t)\label{turbodecoding2}\end{aligned}\ ] ] the expectation value @xmath23 is intended to be taken with respect to the boltzmann weight with the modified hamiltonian @xmath24 . the iteration variables @xmath25 should be interpreted as external fields conjugate to the operators @xmath26 . they describe , in an approximate way , the action of each of the two chains on the other one . in order to lighten the notation , let us write eqs . ( [ turbodecoding1]-[turbodecoding2 ] ) in the form : @xmath27 with @xmath28 if @xmath29 , and @xmath30 if @xmath31 . due to the randomness in the couplings @xmath32 , the fields @xmath33 are random variables . equation ( [ turbodecodingshort ] ) implies an integral equation for the probability distribution of @xmath33 : @xmath34 \label{probabiltyiteration}\end{aligned}\ ] ] let us state a few approximations which allow us to reduce eq . ( [ probabiltyiteration ] ) to a much simpler one . * we make the substitution @xmath35 in eq . ( [ probabiltyiteration ] ) . this yields a closed integral equation describing the evolution of the distribution @xmath36 . * we neglect correlations between the fields at different sites : @xmath37 these two hypothesis imply that eq . ( [ probabiltyiteration ] ) is equivalent to : @xmath38\int\!d{\cal p}[\jg]\;\delta\left(y- \frac{1}{\beta}\arctanh\left(\<\epsilon_{\hat{\rho}(i)}(\sg)\>_{\jg,\xg } \right)+x_{\hat{\rho}(i)}\right ) \label{pdependent}\\ d\pi^{(k')}_t[\xg ] & \equiv&\prod_{i=1}^n dx_i \;\pi^{(k')}_t(x_i)\end{aligned}\ ] ] where @xmath39 is the appropriate permutation of @xmath40 , i.e. @xmath41 if @xmath42 and @xmath43 if @xmath44 . the expectation value @xmath45 on the right hand side of eq . ( [ pdependent ] ) has to be taken with respect to the hamiltonian @xmath46 . let us now define a field distribution averaged over the permutations and the sites : @xmath47 where we made explicit the dependence of @xmath48 upon the specific permutation @xmath49 which defines the code . we can now state our last approximation . * we make the substitution @xmath50 on the right hand side of eq . ( [ pdependent ] ) . this yields a recursive equation for @xmath51 : @xmath52\int d{\cal p}[\jg]\ \delta\left(y- \frac{1}{\beta}\arctanh\left(\<\epsilon_{i}(\sg)\>_{\jg,\xg}\right ) + x_{i}\right ) \label{pindependent}\end{aligned}\ ] ] the indices @xmath53 and @xmath54 have been dropped since we can define @xmath55 for @xmath21 odd , and @xmath56 for @xmath21 even , or vice - versa . a byproduct of this heuristic derivation is the expression for the probability distribution of the expectation values @xmath57 after @xmath21 iterations of the turbo decoding algorithm : @xmath58\int d{\cal p}[\jg]\ \delta\left(\epsilon- \<\epsilon_{i}(\sg)\>_{\jg,\xg}\right)$ ] . let us discuss the validity of the approximations made in deriving eq . ( [ pindependent ] ) . * these approximations should be accurate in the thermodynamic limit for a generic random permutation @xmath49 . the reason is that the correlations produced by eqs . ( [ turbodecoding1]-[turbodecoding2 ] ) have short range : @xmath59 and @xmath60 have a significant correlation only if @xmath61 is less than some characteristic length . the random permutation @xmath49 reshuffles the sites so that the correlation between two fields @xmath25 and @xmath62 is vanishing with high probability if @xmath61 is required to be `` small '' . the correlations which `` survive '' ( non vanishing only between `` distant '' sites ) are irrelevant when computing the expectation values of local operators . in order to make this last assertion plausible , let us suppose that , for each site @xmath63 , we can find a `` large '' . ] interval @xmath64 $ ] of the chain , such that the correlations between the couplings inside the interval are negligible . the expectation value @xmath65 will not depend ( as @xmath66 ) upon the couplings outside @xmath64 $ ] ( this is always true in one dimension at non zero temperature ) and can be then safely computed without taking into account the correlations . it is easy to find a similar argument concerning the correlations between @xmath67 and @xmath68 in eq . ( [ probabiltyiteration ] ) . * this is the probabilistic analogue of the replica symmetric approximation . let us consider the fixed point equation @xmath69 corresponding to the dynamics defined by eq . ( [ pindependent ] ) . it is remarkable that this fixed point equation coincides with the saddle point equation obtained by the standard replica method in the replica symmetric approximation ( see section [ withreplicas ] ) . this fact confirms our conclusions about the relevance of the various approximations . equation ( [ pindependent ] ) is the final outcome of our heuristic derivation . we want to study its behavior when the distribution @xmath70 is concentrated on large values of the field @xmath71 , that is when the error probability is very small . in this regime the most relevant spin configuration satisfies @xmath72 for each @xmath73 . the lowest excitations are such that @xmath74 only on a few sites . the first crucial point will be to understand that , for a class of hamiltonians of the type ( [ models2 ] ) ( which will be defined as `` recursive ' ' ) , the energy to be paid for flipping a single @xmath75 variable diverges in the thermodynamic limit . the second point will be to evaluate the energy to be paid for flipping two @xmath75 variables . in order to treat both these passages in full generality it is convenient to use an algebraic bookkeeping technique which we shall soon explain . the results concerning these two points will be useful again in section [ stability ] . a preliminary step consists in making the change of variables @xmath76 and introducing the corresponding distribution function @xmath77 . low @xmath78 s correspond then to large local fields , i.e. to low error probability . the result is @xmath79\int d{\cal p}[\jg]\ \delta\left(y- \frac{1}{x_i}\,\frac{z(\epsilon_i(\sg)=-1;\jg,\xxg ) } { z(\epsilon_i(\sg)=+1;\jg,\xxg)}\right ) \label{riscritta}\end{aligned}\ ] ] where @xmath80 with @xmath81 . let us introduce some notations in order to write down the small @xmath78 expansion of @xmath82 : @xmath83 is an @xmath84-uple ( not ordered ) of integers in @xmath85 ; @xmath86 is the configuration such that @xmath72 for all the sites @xmath63 ; @xmath87 is the configuration such that @xmath88 if @xmath89 and @xmath90 otherwise ( there is at most one such configuration once the boundary conditions have been specified ) ; @xmath91 is the energy of the ordered configuration ; finally @xmath92 . the following expressions are straightforward : @xmath93 the `` bookkeeping technique '' which we shall adopt in treating the above expansions consists in using the algebra of `` generating polynomials '' @xcite . this approach allows us to consider a general hamiltonian of the type ( [ models2 ] ) . let us define the following polynomials on @xmath94 : @xmath95 , with @xmath96 ; @xmath97 ; @xmath98 . notice that the boundary condition on @xmath99 can be translated as follows : @xmath100 is a series of strictly positive powers of @xmath71 . it is necessary to distinguish two types of models : in the first case @xmath101 divides @xmath102 , i.e. @xmath103 is a polynomial ( these are the `` non recursive '' models , a particular case being @xmath104 ) ; in the second one @xmath101 does not divide @xmath102 , i.e. @xmath103 is a series ( `` recursive '' models ) . we shall treat the `` recursive '' models first . in this case the first order terms in the expansions ( [ lowx1 ] ) and ( [ lowx2 ] ) are exponentially small in the size . in order to prove this assertion , let us consider the configuration @xmath105 . the relevant generating polynomials are @xmath106 and @xmath107 . the form of @xmath108 is given by the following result of algebra [ eccitazioni1 ] let @xmath109 and @xmath110 be two polynomials on @xmath94 such that @xmath111 , @xmath112 , and their greatest common divisor @xmath113 is equal to 1 . then there exists an integer @xmath114 such that @xmath115 with @xmath116<\omega$ ] and @xmath117 if @xmath118 is large enough . hereafter we shall call @xmath119 the smallest of such integers . an explicit expression for @xmath119 is given in the appendix [ appendicealgebra ] . the lemma [ eccitazioni1 ] applies to our case if we divide both @xmath101 and @xmath102 by their greater common divisor : @xmath120 , so that @xmath121 . it implies that if we write down the numbers @xmath122 we get an antiperiod followed by a non trivial periodic sequence with period @xmath123 . let us consider a site `` in the bulk '' : @xmath124 with @xmath125 a ( small ) positive number . then , using the convention @xmath126 for @xmath127 , we get : @xmath128 which diverges almost surely in the thermodynamic limit if @xmath129 ( see the introduction on this point ) . in eq . ( [ riscritta ] ) we must sum also terms which are `` near '' the boundaries , i.e. @xmath130 or @xmath131 . these give however a negligible contribution . let us now consider the second order terms of the expansions ( [ lowx1 ] ) and ( [ lowx2 ] ) . they involve configurations @xmath132 with two flipped @xmath133 s . the only configurations which give a non negligible contribution are the ones which involve a finite ( in the @xmath66 limit ) number of flipped @xmath134 s . this corresponds to choosing @xmath135 and @xmath84 such that @xmath136 is a polynomial ( and not an infinite series ) . the following useful result is proved in the appendix [ appendicealgebra ] . [ eccitazioni2possibili ] let @xmath110 be a polynomial on @xmath94 such that @xmath137 and @xmath135 an integer . then there exists an integer @xmath119 such that @xmath110 divides @xmath138 if and only if @xmath135 is a strictly positive multiple of @xmath119 . as suggested by the notation the @xmath119 s cited in lemmas [ eccitazioni1 ] and [ eccitazioni2possibili ] are indeed equal . the terms which give a non vanishing contribution at order @xmath139 in the expansions ( [ lowx1]-[lowx2 ] ) are the ones corresponding to configurations @xmath132 such that @xmath140 is a multiple of @xmath123 . in order to evaluate these terms we must count the number of flipped @xmath134 s . this number is nothing but the number of non zero coefficients in the polynomial @xmath136 . let us define the weight of a polynomial @xmath141 over @xmath94 as the number of its non zero coefficients : @xmath142 . the weight of @xmath136 is given , for a large class of hamiltonians , by the following lemma . [ eccitazioni2energia ] let @xmath110 and @xmath109 be two polynomial on @xmath94 such that @xmath143 , @xmath112 , and @xmath144 . if @xmath145 then the weight of @xmath146 is given by @xmath147 for each @xmath148 . the coefficients @xmath149 and @xmath150 are positive integers whose explicit expressions are given by eqs . ( [ peso1]-[peso2 ] ) . appendix [ appendicealgebra ] contains also an illustration of what could happen in the more general case . by using lemmas [ eccitazioni2possibili ] and [ eccitazioni2energia ] we can linearize with respect to @xmath78 the expression on the r.h.s . of eq . ( [ riscritta ] ) : @xmath151 and defining @xmath152 we get @xmath153 if @xmath154 . clearly eq . ( [ sviluppo ] ) holds only for @xmath63 in the `` bulk '' ( i.e. @xmath155 ) up to terms which are exponentially small in the size @xmath156 . our first important observation is that the right hand side of eq . ( [ sviluppo ] ) vanishes if @xmath157 for @xmath158 . this means that @xmath159 is a fixed point of eq . ( [ riscritta ] ) for `` recursive '' models . recall that the change of variables which yields eq . ( [ riscritta ] ) is @xmath160 and that @xmath71 has the meaning of an effective field acting on @xmath161 . the solution @xmath162 corresponds then to a phase with completely frozen spins : @xmath163 . we would like to understand if this phase is stable for some temperature @xmath164 and some distribution of the couplings . a possible approach is to study the turbo decoding dynamics ( as described by eq . ( [ riscritta ] ) ) when starting from a distribution `` near '' @xmath162 . let us suppose that , for @xmath165 near enough to @xmath162 , we can safely neglect @xmath166 terms on the r.h.s . of eq . ( [ sviluppo ] ) : @xmath167\int\ ! d{\cal p}[\jg]\ \delta\left(y- \sum _ { m\neq 0 } x_{i+m\omega(f_1)}e^{-\beta\delta(i , i+m\omega(f_1))}\right ) \label{linearizzata}\end{aligned}\ ] ] this equation is very similar to a class of recursive equations which appear in a completely different context : polymers on disordered trees @xcite . these are of the type @xmath168 the only non trivial difference is that the linear function of @xmath169 appearing inside the delta function on the r.h.s . of eq . ( [ linearizzata ] ) depends upon a macroscopic ( indeed linear in @xmath156 ) number of @xmath78 s . in eq . ( [ alberi ] ) , instead , only a finite number of variables appears : @xmath170 is the coordination number of the tree minus one . notice however that , for @xmath171 large , @xmath172 . we can thus truncate the sum in eq . ( [ linearizzata ] ) to @xmath173 by making an error of order @xmath174 and we guess that the limit @xmath175 can be taken at the end without problems . for more convincing arguments . ] let us summarize some results of @xcite which are useful in our discussion . it turns out that eq . ( [ alberi ] ) is equivalent to a discretization of the kolmogorov - petrovsky - piscounov ( kpp ) equation @xcite ( a well studied partial differential equation ) . using this equivalence the large time limit of eq . ( [ alberi ] ) is obtained : @xmath176 corresponding to a front wave solution of the kpp equation with front velocity @xmath177 . if we define the function @xmath178 then the front velocity is given by the following construction : @xmath179 with @xmath180 given by @xmath181 at the critical temperature @xmath180 a freezing phenomenon takes place with the front velocity sticking to its minimal value . let us apply these results to our case , i.e. to eq . ( [ linearizzata ] ) . the large time solution @xmath182 gives the correct behavior for @xmath183 only if @xmath184 . in this case @xmath185 and it is then correct to linearize eq . ( [ riscritta ] ) : the frozen phase is stable . if , on the other hand , @xmath186 then we must take into account higher order terms in the low @xmath78 expansion and the asymptotic form is no longer of the type defined by eq . ( [ larget ] ) : the frozen phase is unstable . in the thermodynamic limit we get @xmath187 e^{-\beta\delta(i , i+m\omega(f_1))}= \label{prevelocity1}\\ & = & 2\left(\int\ ! dj \;p(j)\ ; e^{-2\beta j}\right)^2 \sum_{m=1}^{\infty}\left(\int\ ! dh\ ; p(h)\ ; e^{-2\beta h}\right ) ^{{\rm weight}(s_m ) } \nonumber\end{aligned}\ ] ] the front velocity @xmath177 is obtained by applying the construction given in eqs . ( [ frontvelocity]-[criticaltemperature ] ) to eq . ( [ prevelocity1 ] ) . if the hypothesis of lemma [ eccitazioni2energia ] are satisfied we can easily sum the series : @xmath188 we discuss now eq . ( [ prevelocity2 ] ) , the more general case being completely analogous . the series converges only if @xmath189 . if @xmath190 , as we supposed sinc the beginning , then convergence is assured for @xmath191 with @xmath192 . it is easy to see that @xmath193 is strictly convex for @xmath191 and thus @xmath194 has either one global minimum or is strictly monotonic for @xmath191 . since @xmath195 the first possibility is excluded and we conclude that @xmath196 . the important point is that the right hand side of eq . ( [ prevelocity1 ] ) is well defined every time we need of it , i.e. for @xmath197 . in applications to turbo codes a simplification occurs : we are interested in a particular temperature , @xmath198 , and we are left with a unique parameter : the signal to noise ratio @xmath199 . moreover the probability distributions of the couplings are fixed by the characteristics of the communication channel @xcite . if we introduce the auxiliary variables @xmath200 and @xmath201 , which correspond to the output of the channel , the probability distributions are obtained as follows @xmath202 where @xmath203 is the probability distribution of the output of the channel conditional to the input @xmath204 . a similar expression holds for @xmath205 . if the channel is symmetric ( i.e. if @xmath206 ) one easily obtains @xmath207 and then @xmath208 . we can distinguish the two cases defined below . * if @xmath209 for some @xmath210 then we are in the no - error phase and the turbo decoding algorithm converges to the message with `` velocity '' @xmath211 . we expect the condition @xmath212 to be verified in the `` low noise '' region @xmath213 . * if @xmath214 in the interval @xmath210 then @xmath215 and the linearization in eq . ( [ linearizzata ] ) is no longer reliable . in this case @xmath216 is expected to converge for @xmath183 to some distribution supported on finite fields @xmath71 . the decoded message will be plagued by a finite error probability per bit , no matter how many times do we iterate the turbo decoding algorithm . let us now study some examples . we consider a gaussian channel with : @xmath217 this choice of the variances is justified since it corresponds to a code with rate @xmath218 ( see ref . it is useful to define the function @xmath219\end{aligned}\ ] ] the three cases below have been already considered in ref . we refer to the appendix [ appendicealgebra ] for the calculation of the constants @xmath220 and @xmath221 to be used in eq . ( [ prevelocity2 ] ) . 1 . a model with nearest neighbours interaction is : @xmath222 and @xmath223 ( which corresponds to the generating polynomials @xmath224 and @xmath225 ) . using eq . ( [ prevelocity2 ] ) and the fact that @xmath226 and @xmath227 we get @xmath228 it is easy to see that @xmath214 for each @xmath210 if @xmath229 . [ range1 ] 2 . for @xmath230 and @xmath231 ( generating polynomials : @xmath232 and @xmath233 ) we obtain @xmath234 and @xmath235 and then @xmath236 finally @xmath237 where @xmath238 is the only real solution of the equation @xmath239 . [ range3 ] 3 . if we consider the model given by @xmath240 and @xmath241 ( generating polynomials : @xmath242 and @xmath243 ) we obtain @xmath234 and @xmath235 as in the previous example . both @xmath244 and @xmath245 coincide with the ones obtained above . let us make a few observations about the validity of our calculation . the threshold @xmath245 has been obtained by starting from a distribution @xmath246 very near to the `` frozen '' one @xmath162 and linearizing eq . ( [ riscritta ] ) in @xmath78 . it must then be interpreted as a threshold for local stability of the `` frozen '' solution . moreover , if we take seriously the heuristic derivation of eq . ( [ pindependent ] ) , we can deduce something about the dynamics of the turbo decoding algorithm in the error - free phase : the probability distribution of the auxiliary fields @xmath247 moves towards infinitely large fields with an average velocity @xmath248 . this conclusion is compared with numerical data in fig . ( [ speedfig ] ) : the agreement seems to be quite good . an interesting outcome of the previous calculation is that the approach to the perfect decoding becomes slower near to the critical signal to noise ratio . let us now discuss the `` non recursive '' models , that is models such that @xmath101 divides @xmath102 . in this case the energy @xmath249 to be paid for flipping @xmath250 remains finite in the thermodynamic limit . the low @xmath78 expansions in eqs . ( [ lowx1]-[lowx2 ] ) have a non vanishing term of order @xmath251 . this implies that @xmath252 is no longer a fixed point of eq . ( [ riscritta ] ) . let us compute @xmath249 . for `` non recursive '' models we can define the polynomial @xmath253 . it is easy to show that @xmath254 . a simple approximation of the fixed point of eq . ( [ riscritta ] ) is : @xmath255 with @xmath256 . this approximation is supposed to be good in the low noise region where we expect the distributions @xmath165 to be concentrated on small @xmath78 s . the replica method @xcite starts with the computation of the ( integer ) moments of the partition function . this can be done by introducing an appropriate order parameter ( the choice is a matter of convenience ) and by recurring to standard tricks . here we choose to use the ( multi)-overlaps @xmath257 and their complex conjugates @xmath258 : @xmath259}\\ s[q,\hat{q } ] & = & -1+q_0\hat{q}_0+\sum_a q_a\hat{q}_a + \sum_{(a , b ) } q_{ab}\hat{q}_{ab}+\dots+n\log2+\label{action}\\ & & + \beta { \cal f}_{1d , n}[q ] + \beta { \cal f}_{1d , n}[\hat{q}]\nonumber\\ { \cal f}_{1d , n}[q]&\equiv & -\lim_{n\to\infty}\frac{1}{n\beta } \log z_{1d , n}[q]\\ z_{1d , n}[q ] & \equiv&\sum_{\{\sigma_i^a\ } } \prod_{i=1}^n[q_0 + \sum_a q_a\epsilon_i(\sg^a)+ \sum_{(a , b)}q_{ab}\epsilon_i(\sg^a)\epsilon_i(\sg^b)+\dots]\cdot \\ & & \phantom{\sum_{\sigma_i^a } } \cdot \int d{\cal p}[\jg ] \exp\left\{-\beta\sum_a h(\sg^a;\jg)\right\}\nonumber\end{aligned}\ ] ] where @xmath260 . and the replica indices @xmath261 run from @xmath262 to @xmath118 . the usual mean field models have no geometrical structure at all . in those cases the introduction of the order parameters leads to a ( replicated ) partition function which factorizes over the sites . in our case we are left with the problem of computing the one dimensional partition functions @xmath263 $ ] . these correspond to the one dimensional sub - structures which are not destroyed by the randomness of the model . the saddle point equations are easily written @xmath264 } \right\>_q \label{saddle}\end{aligned}\ ] ] where the expectation values @xmath265 , @xmath266 are defined as follows @xmath267}\int d{\cal p}[\jg ] \sum_{\{\sigma_i^a\ } } ( \;\cdot \;)\ ; \prod_{i=1}^n[q_0+\sum_a q_a\epsilon_i(\sg^a)+\dots]\ ; e^{-\beta\sum_a h[\mbox{\boldmath $ { \scriptstyle\sigma}$}{}^a ; \mbox{\boldmath $ { \scriptstyle j}$}]}\end{aligned}\ ] ] in the recursive case eq . ( [ saddle ] ) admits the following solution ) : @xmath268=s[e^{i\theta}q , e^{-i\theta}\hat{q}]$ ] . however the integration over the parameter @xmath269 does not pose any problem . we shall fix this freedom by imposing @xmath270 to be real.[notasimmetria ] ] corresponding to a no - error phase : @xmath271 . the free energy of this phase is @xmath272 . if we parametrize the replica symmetric ansatz as in ref . @xcite @xmath273 and analogously for @xmath274 ( with a different distribution @xmath275 ) , the following free energy functional can is obtained in the limit @xmath276 : @xmath277 & = & \frac{1}{\beta}\int \!dx\;dy\;\pi(x)\hat{\pi}(y ) \log\left[2\cosh(\beta x+\beta y)\right]+ { \cal f}_{1d}^{rs}[\pi]+{\cal f}_{1d}^{rs}[\hat{\pi}]\\ { \cal f}_{1d}^{rs}[\pi ] & \equiv & -\lim_{n\to\infty } \frac{1}{\beta n}\int \!d{\cal p}[\jg ] \int \!d\pi[\xg ] \log z_{1d}^{rs}[\jg,\xg]\\ z_{1d}^{rs}[\jg,\xg]&\equiv & \sum_{\sg}\exp\left [ \beta\sum_{i=1}^n ( j_i+x_i ) \epsilon_{i}(\sg ) + \beta\sum_{i=1}^n h_i \eta_{i}(\sg)\right]\end{aligned}\ ] ] the distributions @xmath278 and @xmath279 are normalized ( @xmath280 ) and satisfy the saddle point equation below : @xmath281\int d{\cal p}[\jg]\ \delta\left(y-\frac{1}{\beta } \arctanh\left(\<\epsilon_{i}(\sg)\>_{\jg,\xg}\right)+x_{i}\right ) \label{saddlers}\end{aligned}\ ] ] which is identical to the fixed point equation corresponding to eq . ( [ pindependent ] ) , if we suppose the order parameters to be real at the saddle point . equation ( [ saddlers ] ) is unpractical since it involves the unknown distributions @xmath70 and @xmath275 infinitely many times . however due to the short range structure of the hamiltonians defined in eq . ( [ models2 ] ) , it can be rewritten as a simple integral equation . obviously the precise form of this equation depends upon the form of the hamiltonian ( [ models2 ] ) . in particular it becomes simpler as the range of the interaction becomes shorter . let us illustrate this point by considering the model [ range1 ] of the previous section : @xmath282 , @xmath223 . we start by defining the following ( right and left ) partition functions : @xmath283\\ z^{(l)}_{i , m}(\sigma_i ) & \equiv & \sum_{\sigma_{i - m}\dots\sigma_{i-1 } } \exp\left[\beta\sum_{k = i - m+1}^i(j_k+x_k)\sigma_k\sigma_{k-1}+ \beta\sum_{k = i - m}^i h_k\sigma_k\right]\end{aligned}\ ] ] and the ( right and left ) fields : @xmath284 we define now a new couple of order parameters , the probability distributions @xmath285 and @xmath286 of the right ( or left ) fields : @xmath287\right)\end{aligned}\ ] ] it easy to show that , at the saddle point , @xmath285 and @xmath286 satisfy the following integral equation : @xmath288\right\}\label{omegaricorsivo}\\ \theta_{\beta}(x;y)&\equiv&\frac{1}{\beta } \arctanh[\tanh(\beta x)\tanh(\beta y)]\end{aligned}\ ] ] and that the solution of eq . ( [ saddlers ] ) is related to the solution of the previous equation as follows : @xmath289\end{aligned}\ ] ] equation ( [ omegaricorsivo ] ) reduces to the dyson schmidt equation @xcite for a one dimensional model with nearest neighbour interaction if we keep the distribution @xmath286 fixed . the interaction between the two one - dimensional subsystems turns it into a nonlinear equation . moreover eq . ( [ omegaricorsivo ] ) can be treated numerically more easily than eq . ( [ saddlers ] ) . a possible approach consists in representing the unknown distribution as @xmath290 and iterating eq . ( [ omegaricorsivo ] ) until a fixed point is reached . an example of this kind of computations is shown in fig . ( [ rsafig ] ) . it is simple to obtain the analogous of eq . ( [ omegaricorsivo ] ) for the simplest non recursive model , defined by : @xmath291 , @xmath292 . the final result is @xmath293\right)\label{omeganonricorsivo}\\ \pi(x ) & = & \int\!dj\;p(j)\int\!dx_l\;\hat{\omega}(x_l)\int\!dx_r\ ; \hat{\omega}(x_r)\;\delta(x - j - x_l - x_r)\end{aligned}\ ] ] a simple approximation to the solution of eq . ( [ omeganonricorsivo ] ) can be obtained by starting from a distribution @xmath285 supported on very large fields @xmath71 and iterating eq . ( [ omeganonricorsivo ] ) one time . the result is @xmath294 , which coincides with the more general eq . ( [ apprnonric ] ) after the change of variables @xmath160 . no such approximation is possible for eq . ( [ omegaricorsivo ] ) . expressions equivalent to eqs . ( [ omegaricorsivo]-[omeganonricorsivo ] ) can be derived for more complicated types of interaction . in general the distribution @xmath285 , which is defined on the real line , will be replaced by a distribution defined on @xmath295 , @xmath296 being the range of the hamiltonian . we would like to study local stability of the no - error phase in the context of the replica method . this can be done by computing the eigenvalues of the matrices : @xmath297 = \delta_{a_1\dots a_l , b_1\dots b_m } \pm \frac{\partial^2 \beta { \cal f}_{1d , n}[q ] } { \partial q_{a_1\dots a_l}\partial q_{b_1\dots b_m}}\end{aligned}\ ] ] @xmath298 $ ] are the mass matrices for purely real ( @xmath299 $ ] ) , or purely immaginary ( @xmath300 $ ] ) , fluctuations of the order parameter around the value @xmath301 . we are interested in the saddle point @xmath302 . in order to write down all the @xmath303 eigenvectors of @xmath304 $ ] it is convenient to change slightly our notation for the overlaps . let us denote by @xmath305 the set of @xmath306 different indices @xmath307 . we can use the @xmath308 s as indices for the overlaps with the natural identification @xmath309 . it is not difficult to show that @xmath310 } { \partial q_{a_1\dots a_l}\partial q_{b_1\dots b_m}}\right|_{q = q^*}=\\ & = & \frac{2^{1-n}}{n}\left\{\sum_{(i , j ) } \int\!d{\cal p}[\jg]\;e^{nn\beta f - n\beta e_0}\ ; \left(1+e^{-\beta\delta(i , j)}\right)^n \left [ \tanh\left(\frac{\beta\delta(i , j)}{2}\right ) \right]^{u}-\frac{n^2}{2}\right\}\nonumber\end{aligned}\ ] ] where @xmath311 is defined in section [ withoutreplicas2 ] , @xmath312\;e^{-n\beta h(\sg_0)}$ ] and @xmath313 counts the indices which are either in the set in @xmath314 or in the set @xmath315 but not in both . if @xmath301 is an eigenvector of @xmath316 with eigenvalue @xmath317 , then it is an eigenvector of @xmath304 $ ] with eigenvalue @xmath318 . notice that @xmath316 is an hermitian matrix with respect to to the scalar product : @xmath319 we shall use another subset of @xmath320 ( let us call it @xmath321 ) to label the different eigenvectors of @xmath316 , which we now exhibit : @xmath322 the vectors @xmath323 form an orthonormal set with respect to the scalar product defined in eq . ( [ scalarproduct ] ) . this is easily proven by induction on @xmath118 . the vector @xmath324 is nothing but the constant one . the corresponding eigenvalue is @xmath325 , whence @xmath326 and @xmath327 . the eigenvalue @xmath327 is a remnant of the invariance of the action under the symmetry cited in the footnote [ notasimmetria ] of the previous section . in order to compute the eigenvalues in the subspace orthogonal to @xmath324 , the following formula turns out to be useful : @xmath328 where @xmath329 denotes the symmetric difference of @xmath308 and @xmath330 ( i.e. @xmath331 ) . using eq . ( [ evalue1 ] ) and the results of algebra outlined in section [ withoutreplicas2 ] we get ( for @xmath332 ) : @xmath333 where @xmath334 is defined in section [ withoutreplicas2 ] and @xmath335 for @xmath336 or @xmath337 . when the one dimensional hamiltonians ( [ models2 ] ) satisfy the hypothesis of lemma [ eccitazioni2energia ] , the sum in eq . ( [ evalue2 ] ) can be explicitly computed yielding : @xmath338 if @xmath339 then @xmath340 is positive and decreasing with @xmath164 . moreover @xmath341 and @xmath342 . we can thus define the critical temperatures @xmath343 for @xmath344 , by requiring ) can have more than one solution for @xmath345 . the `` physical '' critical point is obtained by taking the limit @xmath276 of the solution of eq . ( [ critical ] ) which exists for any @xmath118 . ] @xmath346 if @xmath347 the `` frozen '' saddle point is stable with respect to the direction @xmath348 . if @xmath349 it becomes unstable : @xmath350 while @xmath351 ( it could be guessed that the `` imaginary '' directions would be stable because of the physical interpretation of the overlaps ) . in the limit @xmath276 , @xmath352 : the critical directions are the ones corresponding to @xmath353 . it is easy to see that the critical temperature @xmath180 coincides with the one obtained in section [ withoutreplicas2 ] . we have presented two derivations of the local stability condition for the no - error phase . both will be object of the criticism of the skeptical reader . in the first one we obtained the `` mean field '' equation describing the dynamics of the decoding algorithm , eq . ( [ pindependent ] ) , by making use of heuristic arguments . indeed we argued eq . ( [ pindependent ] ) to be valid only in the replica symmetric approximation . in the second derivation we made use of the replica method , which has not ( yet ) well founded mathematical basis . we think that the two derivations compensate each other for their defects . moreover they yield the same replica symmetric saddle point equation ( [ saddlers ] ) and give the same picture of the instability which destroys the no - error ( frozen ) phase . this corresponds to couples of flipped @xmath133 s . finally thanks to the first derivation we get some insight on the behavior of the decoding algorithm . in particular we have seen that , in the frozen phase , it approaches a no - error fixed point . this approach becomes slower near to the boundary of the frozen phase . in ref . @xcite the local stability threshold computed here has been compared with numerical simulations for two types of code , respectively the models [ range1 ] and [ range3 ] presented in section [ withoutreplicas2 ] . good agreement was found only for model [ range1 ] . we propose two possible explanations of the disagreement for model [ range3 ] : * the phase transition is a first order one ; * the turbo decoding algorithm used in ref . @xcite gets sticked in some local minimum of the free energy , characterized by a finite error probability per bit . we have not yet enough informations for choosing between these two scenarios . [ section ] [ section ] in this appendix we remind to the reader some known facts in the theory of finite fields and we prove the propositions stated in section [ withoutreplicas2 ] . these are nothing but simple exercises and we work out them in detail only for greater convenience of the reader . finally we illustrate a few applications of the results obtained . the reader interested in a more complete treatment can consult refs . @xcite . let us begin with some elementary definitions . the basic object is @xmath94 i.e. the field of integer numbers modulo @xmath354 . a polynomial over @xmath94 , @xmath355 $ ] is simply a polynomial whose coefficients are in @xmath94 . we say @xmath355 $ ] to be irreducible if there do not exist two noncostant polynomials @xmath356 $ ] such that @xmath357 . any @xmath355 $ ] possess an unique factorization , i.e. a decomposition of the form @xmath358 where @xmath359 $ ] are irreducible and @xmath360 are integer numbers . given two polynomials @xmath361 $ ] we say that @xmath110 divides @xmath109 ( in symbols @xmath362 ) if there exists @xmath363 $ ] such that @xmath364 . , we say that @xmath365 divides @xmath301 ( and write @xmath366 ) if there exists @xmath367 , such that @xmath368 . ] for an irreducible polynomial @xmath355 $ ] it does make sense to define the order @xmath369 : @xmath369 is the smallest positive integer @xmath135 such that @xmath370 . the basic result which we shall employ in this appendix is the following : [ irriducibili ] let @xmath110 be an irreducible polynomial over @xmath94 . then @xmath370 if and only if @xmath371 . it is useful to know how to compute the order of an irreducible polynomial . the main tool is the theorem below : [ ordine ] let @xmath110 be an irreducible polynomial of degree @xmath372 over @xmath94 . then @xmath372 is the smallest positive integer for which @xmath373 . moreover it is obvious from the definition that @xmath374 our first step will be the proof of lemma [ eccitazioni2possibili ] which we restate here as follows [ eccitazioni2possibilibis ] let @xmath110 be a polynomial on @xmath94 with the following factorization @xmath375 where the polynomials @xmath376 are irreducible over @xmath94 . let @xmath377 be the smallest integer such that @xmath378 . then @xmath379 if and only if @xmath380 and @xmath381 for @xmath382 . _ proof of lemma [ eccitazioni2possibilibis ] . _ let us begin by noticing that , since the @xmath376 are irreducible , @xmath379 if and only if @xmath383 for @xmath382 . we can then limit ourselves to the case @xmath384 with @xmath385 irreducible . it is convenient to work in an extension of @xmath94 , i.e. in a field containing @xmath94 as a subfield . we choose an extension ( let us call it @xmath386 ) of @xmath94 such that both @xmath385 and @xmath387 can be decomposed in linear factors . the existence of such an extension is a basic fact of field theory . we are then looking for the @xmath135 such that all the root of @xmath385 ( in @xmath386 ) are roots of @xmath387 with multiplicity at least @xmath296 . it is then necessary to study the multiplicity of the roots of @xmath387 . the first observation is that , if @xmath135 is odd , all the roots are simple . in fact @xmath388 has no roots in common with @xmath387 . the second observation consists in noticing that @xmath389 . we deduce that @xmath390 with @xmath135 odd has @xmath135 distinct roots ( the same as @xmath387 ) , each one with multiplicity @xmath391 . the final outcome is that @xmath392 if and only if @xmath393 and @xmath394 @xmath395 from lemma [ eccitazioni2possibilibis ] the explicit form of the period @xmath119 used in sec . [ withoutreplicas2 ] is easily obtained : @xmath396 the lemmas [ eccitazioni1 ] and [ eccitazioni2energia ] are easy consequences of lemma [ eccitazioni2possibili ] . _ proof of lemma [ eccitazioni1 ] . _ let us begin by considering the series @xmath397 . we can always define the polynomials @xmath398 with @xmath399 such that @xmath400 . since @xmath110 divides @xmath401 for all @xmath148 , we conclude that @xmath402 for all @xmath403 and @xmath404 . with the following definition @xmath405<\omega(f ) \label{prodotto}\end{aligned}\ ] ] we get @xmath406 notice that , for @xmath407 , @xmath408 . an upper bound on @xmath409 is easily obtained from eq . ( [ prodotto ] ) yielding . ] @xmath410 for @xmath411 - 1)/\omega(f)\rceil\ge l$ ] . clearly it can not be @xmath412 otherwise we would conclude that @xmath110 divides @xmath109 in contradiction with the hypothesis . in order to complete the proof of let us suppose the following equation to hold @xmath413 with @xmath414 , @xmath415 and @xmath416 for @xmath118 large enough . this implies that @xmath110 divides @xmath417 but , since @xmath418 , we would conclude that @xmath110 divides @xmath419 contradicting lemma [ eccitazioni2possibili ] @xmath395 _ proof of lemma [ eccitazioni2energia ] . _ it suffices to specialize the content of the previous paragraph to the case @xmath420\le\omega(f)$ ] : @xmath421 whence @xmath422 & = & w_0 + w_1\cdot m \label{peso1}\\ w_0&\equiv & \mbox{weight}[g_0(x)]+\mbox{weight}[g_1(x ) ] -\mbox{weight}[g_0(x)+g_1(x ) ] \label{peso2}\\ w_1&\equiv & \mbox{weight}[g_0(x)+g_1(x ) ] \label{peso3}\end{aligned}\ ] ] @xmath395 what does it happen when the hypothesis of lemma [ eccitazioni2energia ] are not satisfied ? it is easy to guess the answer . there exists a positive integer @xmath423 such that , for @xmath424 , @xmath334 grows linearly with @xmath171 : @xmath425 with @xmath426 . thanks to this fact we can always sum the series in eq . ( [ prevelocity1 ] ) in the interval @xmath191 . the discussion of the behavior of eq . ( [ linearizzata ] ) presented in section [ withoutreplicas2 ] is then completely general . let us return down to the earth and make a few examples . we shall consider the codes presented in ref . @xcite : 1 . the simplest non trivial case : @xmath427 , @xmath428 . clearly both the polynomials are irreducible . the degree of @xmath110 is @xmath429=1 $ ] . because of theorem [ ordine ] @xmath430 whence @xmath431 . theorem [ irriducibili ] implies that @xmath432 for each @xmath433 . this conclusion is easily confirmed by the well known formula @xmath434 . lemma [ eccitazioni1 ] tells us that @xmath435 with @xmath436 for @xmath437 and that @xmath438 ( @xmath262 is the unique non zero polynomial of degree zero ) . we have thus rediscovered the simple fact that @xmath439 . finally we observe the hypothesis of lemma [ eccitazioni2energia ] are satisfied and that ( with the notation of eq . ( [ gg ] ) ) , @xmath440 and @xmath441 . from eqs . ( [ peso1]-[peso2]-[peso3 ] ) it follows that @xmath442 = m$ ] which is easily confirmed by observing that @xmath443 . [ codicesemplice ] 2 . a less elementary example is : @xmath444 , @xmath445 . it is easy to see that @xmath110 is irreducible and that @xmath446 whence @xmath418 . from @xmath447 and @xmath448 we deduce that @xmath449 . in fact @xmath450 thus by lemma [ eccitazioni2possibili ] @xmath379 if and only if @xmath135 is a multiple of @xmath451 . we can use lemma [ eccitazioni2energia ] in order to compute the weight of @xmath452 . we see that @xmath453 and @xmath454 whence @xmath455=2 + 2m$ ] . with some book - keeping one can confirm this result : @xmath456=2 + 2m\end{aligned}\ ] ] 3 . finally the generating polynomials used in ref . @xcite to build the first example of turbo code : @xmath457 , @xmath458 . also in this case @xmath110 is irreducible and @xmath459 yielding @xmath418 . since @xmath460 and @xmath461 , we deduce that either @xmath462 or @xmath463 . however we know that @xmath464 and we conclude that @xmath465 . in fact @xmath466 using the fact that @xmath467 and @xmath454 we get @xmath455=2m+2 $ ] . in this appendix we study eq . ( [ linearizzata ] ) in order to extend to this case the results concerning eq . ( [ alberi ] ) used in section [ withoutreplicas2 ] . we shall examine both the approach of ref . @xcite , which is based on the analogy with the kpp equation and is a non rigorous one , and the approach of ref . @xcite , which employs probability theory and is entirely satisfactory from the mathematical point of view . we would like to deal with this type of equation : @xmath468 with the requirement that @xmath469 and the initial condition @xmath470 . following ref . @xcite we introduce the function : @xmath471 which satisfy this recurrence equation @xmath472 let us make a few elementary observations concerning eq . ( [ recurrencegen ] ) : if @xmath473 for some @xmath171 and all @xmath71 then @xmath474 for all @xmath71 and @xmath475 ; if @xmath476 then @xmath477 ; if @xmath478 is increasing and @xmath479 for some @xmath71 ( both these hypothesis are implied by eq . ( [ definizionegen ] ) ) then @xmath480 . the stationary uniform solutions of eq . ( [ recurrencegen ] ) are @xmath481 and @xmath482 . the first one is obviously stable . if we consider a small fluctuation around @xmath483 , @xmath484 we get : @xmath485 which can be diagonalized in fourier space : @xmath486 notice that @xmath487 for @xmath135 small enough and that @xmath488 diverges at @xmath489 in agreement with the previous observation that if @xmath490 than @xmath491 . the preceding observations lead us to the hypothesis that the @xmath492 behavior of our problem is controlled by front - like solutions @xmath493 interpolating between the stable state @xmath481 at @xmath494 and @xmath482 at @xmath495 . this scenario is easily confirmed in the case without disorder . if @xmath496 and @xmath497 one obtains @xmath498 , @xmath499 with @xmath500 in the general case we assume the existence of front - like solutions with the large @xmath71 behavior @xmath501 . the front velocity is obtained through the construction given in eqs . ( [ frontvelocity]-[criticaltemperature ] ) with @xmath502 notice that @xmath129 implies that @xmath503 in some interval @xmath191 and that @xmath504 . this remark allows us to sum the series @xmath505 in the range @xmath197 , thus obtaining eq . ( [ treespeed ] ) . the same remark will be useful in the following . let us consider now the more rigorous approach used in ref . we start by defining the polymer model which corresponds to eq . ( [ type ] ) . we have to use a tree with a numerable set of branches stemming from each node . a node of the @xmath118-th generation is identified by @xmath118 integer numbers @xmath506 ; its generation is denoted by @xmath507 . we denote by @xmath508 the root node ( i.e. the only node of the zeroth generation ) . we say that the node @xmath509 belonging to the @xmath171-th generation is a descendant of the node @xmath510 of the @xmath118-th generation ( and write @xmath511 if @xmath512 or @xmath513 if @xmath514 ) if @xmath515 . the node @xmath509 is said to be an older brother of the node @xmath510 with @xmath516 if @xmath517 and @xmath518 . a pair of random variables @xmath519 and @xmath520 is attached at each node . all these variables are statistically independent and have marginal distributions @xmath521 ( the @xmath520 s ) and @xmath522 ( the @xmath519 s ) . a directed polymer is given by a pair of nodes @xmath523 . to each polymer we assign an energy as follows : @xmath524 moreover we use the shorthand @xmath525 and define the following partition functions : @xmath526 the velocity of the front wave studied in the previous paragraphs corresponds in this language to the random variable : @xmath527 the model has two phases . in the high temperature phase ( @xmath528 ) the fluctuations of @xmath529 are small and @xmath530 in the low temperature phase ( @xmath531 ) the fluctuations become large and @xmath177 is fixed by simple convexity and monotonicity arguments . the key point of the approach used in ref . @xcite is to estimate these fluctuations by proving that , for @xmath532 : @xmath533 for some @xmath534 uniformly in @xmath118 . this is enough for obtaining eq . ( [ hightfreeenergy ] ) . let us define the normalized variables @xmath535 . in ref . @xcite the bound in eq . ( [ bound ] ) is obtained starting with the second moment of @xmath536 , and then refining the inequality for the fractional moments of order @xmath534 . notice that looking at the @xmath171-th moment of the partition function is a well known method @xcite for obtaining an upper estimate on the critical temperature ( the estimate becomes worser as @xmath171 gets larger ) . let us have a look at the first two integer moments : @xmath537+\nonumber\\ & & + \<e^{-2\beta v}\>\sum_{k=0}^{\infty}\<e^{-2\beta h}\>^k \left(1 + 2\sum_{l=1}^{\infty}\<e^{-\beta h}\>^l\right)\ ; \<z_n(2\beta)\>\end{aligned}\ ] ] in general the @xmath171-th moment is finite ( but not necessarily uniformly bounded ) only if @xmath538 i.e. if @xmath539 . there is no integer moment of order greater than one which remains finite in the interval @xmath540 . this fact forces us to a slight modification of the proof presented in ref . we use the trivial identity : @xmath541 and estimate the @xmath542-th moment ( with @xmath534 ) as follows : @xmath543}\ ; z_n(\beta|\underline{\omega}^1)\;z_n(\beta|\underline{\omega}^2 ) \right\}^{\alpha/2}\le\nonumber\\ & \le & \sum_{\underline{\omega}^1:\atop |\underline{\omega}^1|=1 } \sum_{\underline{\omega}^2:\atop |\underline{\omega}^2|=1 } e^{-\frac{\alpha\beta}{2 } [ e(\underline{\omega}^1)+e(\underline{\omega}^2)]}\ ; z_n^{\alpha/2}(\beta|\underline{\omega}^1)\ ; z_n^{\alpha/2}(\beta|\underline{\omega}^2)\end{aligned}\ ] ] for a temperature such that @xmath544 we can take the averages and sum up the series : @xmath545}\>\ ; \<z_n^{\alpha/2}(\beta)\>^2\le\nonumber\\ & \le & \phi(\alpha\beta)\<z_n^{\alpha}(\beta)\>+ 2 \phi(\alpha\beta)\sum_{l=1}^{\infty}\<e^{-\frac{\alpha\beta}{2}h}\>^l \<z_n(\beta)\>^{\alpha}\end{aligned}\ ] ] rewriting this formula for the normalized variables we get @xmath546 \ < m_n^{\alpha}(\beta)\>+ 2\left[\frac{\phi(\alpha\beta)}{\phi(\beta)^{\alpha}}\right ] \sum_{l=1}^{\infty}\<e^{-\frac{\alpha\beta}{2}h}\>\equiv\nonumber\\ & \equiv & \psi(\alpha,\beta)\ < m_n^{\alpha}(\beta)\ > + \chi(\alpha,\beta)\end{aligned}\ ] ] at this point we observe , following ref . @xcite , that , if @xmath547 ( i.e. @xmath532 ) then we can choose @xmath548 such that @xmath549 . the condition to be imposed on @xmath542 for obtaining this inequality is @xmath550 ( notice that this inequality implies the previous one @xmath551 ) . the desired bound is obtained by using gronwall lemma together with the fact that @xmath552 : @xmath553

we update the indirect constraints on the georgi - machacek model from-physics and electroweak precision observables , including new constraints from and . we illustrate the effect of these constraints on the couplings of the standard model - like higgs boson by performing scans using the most general scalar potential , subject to vacuum stability and perturbativity constraints . we find that simultaneous enhancements of all the higgs production cross sections by up to 39% are still allowed after imposing these constraints . lhc rate measurements on the higgs pole could be blind to these enhancements if unobserved non - standard higgs decays are present .

1410.5538

since the 2012 discovery of a standard model ( sm)-like higgs boson at the cern large hadron collider ( lhc ) @xcite , there has been considerable interest in models with extended higgs sectors to be used as benchmarks for lhc searches for physics beyond the sm . one such model is the georgi - machacek ( gm ) model @xcite , which adds isospin - triplet scalar fields to the sm in a way that preserves custodial su(2 ) symmetry . its phenomenology has been extensively studied @xcite . the gm model has also been incorporated into the scalar sectors of little higgs @xcite and supersymmetric @xcite models , and an extension with an additional isospin doublet @xcite has also been considered . the gm model has the interesting feature that the coupling strengths of the sm - like higgs boson @xmath3 to @xmath4 or @xmath5 boson pairs can be larger than in the sm . and @xmath6 couplings above their sm strength while preserving custodial su(2 ) symmetry can also be obtained in higher - isospin generalizations of the gm model @xcite or in an extension of the higgs sector by an isospin septet with appropriately - chosen hypercharge @xcite . ] such an enhancement is not possible in higgs - sector extensions that contain only isospin doublets or singlets . in light of the upcoming lhc data - taking period during which higher - precision measurements of the sm - like higgs boson couplings and searches for additional higgs states will be pursued , it is timely to revisit the indirect constraints on the gm model from @xmath0-physics and electroweak precision data . constraints from the oblique parameter @xmath7 have been studied in refs . @xcite and constraints from the nonoblique @xmath5-pole observable @xmath8 have been studied in refs . @xcite . in this paper we point out that the dominant one - loop contributions of the additional gm higgs bosons to nonoblique @xmath5-pole observables and to @xmath0-physics observables can be taken over directly from calculations in the type - i two - higgs - doublet model ( 2hdm ) @xcite . we use this fact to determine for the first time the constraints on the gm model from @xmath1 , @xmath9@xmath10 mixing , and @xmath11 . for @xmath1 , we adapt the numerical implementation for the 2hdm in the public code superiso v3.3 @xcite . for @xmath12 , we make use of a new calculation of @xmath13 @xcite in the aligned 2hdm @xcite . of these observables , we find that @xmath1 provides the strongest constraint , though it may be surpassed in the near future as the precision on the lhc measurement of the @xmath11 branching fraction improves . we also provide an analytic formula for the @xmath7 parameter in the gm model in the approximation that the new scalars are heavy compared to the @xmath5 mass . we then examine the effect of these indirect constraints on the accessible ranges of the sm - like higgs boson couplings . we find that simultaneous enhancements of the @xmath14 , @xmath6 , and @xmath15 couplings above their sm value are still allowed , and could simultaneously enhance the sm - like higgs boson production cross sections in all production modes by up to 39% . because the lhc measures higgs production rates only in particular higgs - decay final states , it could be blinded to such an enhancement by the presence of new unobserved higgs decay modes that would suppress the higgs branching ratios into detectable final states . disentangling these effects will be a major phenomenological and experimental challenge at the lhc . can be made with no reference to the higgs decay branching ratios by using the recoil mass method ; see , e.g. , ref . @xcite . ] this paper is organized as follows . in sec . [ sec : model ] we briefly review the gm model and set our notation . in sec . [ sec : st ] we discuss the constraints from the oblique parameters and give our analytic formula for @xmath7 . in sec . [ sec : nonoblique ] we discuss @xmath8 and the @xmath0-physics observables , and compare their constraints on the gm model parameter space . in sec . [ sec : numerical ] we illustrate the effects of these indirect constraints through numerical scans over the gm model parameter space , imposing all relevant theoretical constraints . we conclude in sec . [ sec : conclusions ] . the scalar sector of the gm model @xcite consists of the usual complex doublet @xmath16 with hypercharge . ] @xmath17 , a real triplet @xmath18 with @xmath19 , and a complex triplet @xmath20 with @xmath21 . the doublet is responsible for the fermion masses as in the sm . custodial symmetry is preserved at tree level by imposing a global su(2)@xmath22su(2)@xmath23 symmetry on the scalar potential . in order to make this symmetry explicit , we write the doublet in the form of a bidoublet @xmath24 and combine the triplets to form a bitriplet @xmath25 : @xmath26 the vacuum expectation values ( vevs ) are defined by @xmath27 and @xmath28 , where the @xmath4 and @xmath5 boson masses constrain @xmath29 the most general gauge - invariant scalar potential involving these fields that conserves custodial su(2 ) is given , in the conventions of ref . @xcite , by @xmath30 ^ 2 + \lambda_2 \text{tr}(\phi^\dagger \phi ) \text{tr}(x^\dagger x ) \nonumber \\ & & + \lambda_3 \text{tr}(x^\dagger x x^\dagger x ) + \lambda_4 [ \text{tr}(x^\dagger x)]^2 - \lambda_5 \text{tr } ( \phi^\dagger \tau^a \phi \tau^b ) \text{tr } ( x^\dagger t^a x t^b ) \nonumber \\ & & - m_1 \text{tr}(\phi^\dagger \tau^a \phi \tau^b)(u x u^\dagger)_{ab } - m_2 \text{tr}(x^\dagger t^a x t^b)(u x u^\dagger)_{ab}. \label{eq : potential}\end{aligned}\ ] ] here the su(2 ) generators for the doublet representation are @xmath31 with @xmath32 being the pauli matrices , the generators for the triplet representation are @xmath33 and the matrix @xmath34 , which rotates @xmath25 into the cartesian basis , is given by @xcite @xmath35 the physical fields can be organized by their transformation properties under the custodial su(2 ) symmetry into a fiveplet , a triplet , and two singlets . the fiveplet and triplet states are given by @xmath36 where the vevs are parameterized by @xmath37 and we have decomposed the neutral fields into real and imaginary parts according to @xmath38 the masses within each custodial multiplet are degenerate at tree level and can be written ( after eliminating @xmath39 and @xmath40 in favor of the vevs ) as can be written using the minimization condition @xmath41 as @xmath42,\ ] ] which is finite in the limit @xmath43 . ] @xmath44 the two custodial - singlet mass eigenstates are given by @xmath45 where @xmath46 the mixing angle and masses are given by @xmath47 , \label{eq : hmass}\end{aligned}\ ] ] where we choose @xmath48 , and @xmath49 , \nonumber \\ \mathcal{m}_{22}^2 & = & \frac{m_1 v_{\phi}^2}{4 v_{\chi } } - 6 m_2 v_{\chi } + 8 \left ( \lambda_3 + 3 \lambda_4 \right ) v_{\chi}^2.\end{aligned}\ ] ] the @xmath7 parameter @xcite is given in terms of the @xmath5 boson and photon self - energies as @xmath50 \nonumber \\ & \simeq & \frac{4 s_w^2 c_w^2}{\alpha_{\rm em } } \left[\pi^{\prime}_{zz}(0 ) -\frac{c_w^2-s_w^2}{s_w c_w}\pi^{\prime}_{z\gamma}(0)-\pi^{\prime}_{\gamma\gamma}(0)\right ] , \end{aligned}\ ] ] where @xmath51 and @xmath52 stand for the sine and cosine of the weak mixing angle , @xmath53 is the electromagnetic fine structure constant , @xmath54 is the @xmath5 boson mass , and @xmath55 denotes the derivative of the self - energy with respect to its argument @xmath56 . here the second expression holds when the new physics scale is large compared to @xmath54 . this second expression can be written in an analytical form , and we use it in what follows . the new contributions to the @xmath7 parameter in the gm model are given by @xmath57 -|g_{zzh}^{\rm sm}|^2\left[\frac{f_1(m_z , m_h^{\rm sm})}{2 m_z^2}-f_3(m_z , m_h^{\rm sm})\right]\right . \nonumber\\ & & \left.+|g_{zzh}|^2\left[\frac{f_1(m_z , m_h)}{2 m_z^2}-f_3(m_z , m_h)\right ] + |g_{zzh_5 ^ 0}|^2\left[\frac{f_1(m_z , m_5)}{2 m_z^2}-f_3(m_z , m_5)\right]\right . \nonumber\\ & & \left.+2|g_{zw^+h_5^{+*}}|^2\left[\frac{f_1(m_w , m_5)}{2 m_w^2}-f_3(m_w , m_5)\right]\right\ } , \label{eq : sgm}\end{aligned}\ ] ] where @xmath58 is the unit of electric charge , @xmath59 is the reference sm higgs mass for which the oblique parameters are extracted , and the loop functions are given when the new physics scale is large compared to @xmath54 by @xmath60\end{aligned}\ ] ] and @xmath61 when their arguments are equal , @xmath62 and @xmath63 are still finite ; taking @xmath64 , where @xmath65 , @xmath62 can be expanded as @xmath66 and @xmath63 can be expanded as @xmath67 the couplings that appear in eq . ( [ eq : sgm ] ) are given by @xcite @xmath68 where we abbreviate @xmath69 , @xmath70 . the sm coupling @xmath71 is given by @xmath72 setting @xmath73 , the experimental values for the oblique parameters @xmath7 and @xmath74 are extracted for a reference sm higgs mass @xmath75 gev as @xmath76 and @xmath77 with a correlation coefficient of @xmath78 @xcite . we implement the constraint using a @xmath79 variable involving @xmath7 and @xmath74 , @xmath80^{-1}_{ij},\ ] ] where @xmath81 is the @xmath82th observable and @xmath83^{-1}_{ij}$ ] is the inverse of the matrix of uncertainties , @xmath84_{ij } = \delta \mathcal{o}_i \ , \delta \mathcal{o}_j \ , \rho_{ij},\ ] ] where @xmath85 are the relative correlations ( note @xmath86 ) . for the two - observable case of interest , we can invert the matrix @xmath87 explicitly and write @xmath88.\ ] ] here @xmath89 and @xmath90 are the experimental central values , @xmath91 and @xmath92 are their @xmath93 experimental uncertainties , @xmath94 is the relative correlation between the two oblique parameters , and @xmath7 and @xmath74 are the new - physics contributions from the gm model . it is well known that , in the gm model , hypercharge interactions break the su(2)@xmath23 global symmetry at one - loop level , yielding a divergent value for the @xmath74 parameter @xcite . this would be corrected in a more complete theory by the counterterm of an su(2)@xmath23-breaking quartic coupling in the scalar potential @xcite , the finite part of which could in turn be adjusted to compensate the one - loop contributions to the @xmath74 parameter . in our analysis we thus take a conservative approach and marginalize over the value of @xmath74 in the @xmath79 , ] which yields @xmath95 ] resulting in a constraint on @xmath7 alone . our constraint on @xmath7 agrees numerically with that shown in fig . 1 of ref . @xcite . extended higgs sectors are typically also constrained by nonoblique corrections to @xmath5-pole observables , as well as @xmath0-physics observables . these constraints come from one - loop diagrams involving higgs boson couplings to fermions and to sm gauge bosons . the analysis of these constraints in the gm model is greatly simplified by the observation that the relevant diagrams are completely analogous to those of the type - i two - higgs - doublet model . in the gm model , fermion masses are generated in the same way as in the sm through yukawa couplings involving the _ single _ su(2)@xmath96 doublet . the resulting feynman rules for vertices involving a scalar and two fermions , with all particles incoming , are given by @xcite @xmath97 here @xmath98 is any charged fermion , @xmath99 is the appropriate element of the cabibbo - kobayashi - maskawa ( ckm ) matrix , and the projection operators are defined as @xmath100 . the @xmath101 couplings are the same as the @xmath102 couplings with @xmath103 . the custodial fiveplet states do not couple to fermions as they have no su(2)@xmath96-doublet component . in particular , the scalar couplings to fermions in the gm model have exactly the same structure as those in the type - i two - higgs - doublet model @xcite with the replacement @xmath104 . in this situation , large enhancements of scalar couplings to light fermions ( in particular to the bottom quark or to charged leptons ) are not possible due to perturbativity constraints on the top quark yukawa coupling . the dominant new - physics contributions to nonoblique @xmath5-pole and @xmath0-physics observables are then due solely to diagrams involving scalar couplings to the top quark ; in particular , diagrams involving the @xmath105 coupling . in other words , @xmath106 is the only new scalar in the gm model that contributes significantly to @xmath5-pole and @xmath0-physics observables . since custodial symmetry requires that the @xmath107 coupling be identical to the @xmath108 coupling in the 2hdm , all of the relevant @xmath106 couplings have the same form as those of @xmath109 in the 2hdm . this implies that all of the nonoblique @xmath5-pole and @xmath0-physics constraints on the gm model can be obtained by making the replacements @xmath104 and @xmath110 in the corresponding calculations for the type - i 2hdm . in what follows we use this correspondence to consider the constraints on the gm model from @xmath8 , @xmath9@xmath10 mixing , @xmath111 , and @xmath112 . these observables each put an upper bound on @xmath113 ( equivalently @xmath114 ) as a function of @xmath115 . in each case we combine the experimental and gm theoretical uncertainties in quadrature and constrain the gm model prediction for the observable in question to lie within 2@xmath116 of the experimental central value . we will refer to these as `` tight '' constraints . however , in the gm model the @xmath106 contributions to @xmath8 , @xmath111 , and @xmath112 worsen the agreement with experiment compared to the sm limit ( i.e. , compared to taking @xmath43 or @xmath117 ) . as we will see , the sm limit is already 0.8@xmath116 , 1.0@xmath116 , and 1.3@xmath116 away from the experimental central values of these three observables , respectively . for this reason , we also consider a second , more conservative approach to constraining the parameter space for these three observables : we require that the gm model prediction lie within 2@xmath116 of the best - fit value obtainable in the gm model ( i.e. , the sm limit ) , again combining the experimental and gm theoretical uncertainties in quadrature . we will refer to these more conservative constraints as `` loose '' constraints . these `` loose '' and `` tight '' constraints are respectively shown in the right- and left - hand panels of fig . [ fig : expconst ] . , we choose the input parameters for all our numerical results from the 2014 review of particle physics @xcite . for @xmath118 , we use the first combination of tevatron and lhc measurements of the top quark mass @xcite . in particular , we set @xmath119 , @xmath120 , @xmath121 , @xmath122 , @xmath123 , @xmath124 and @xmath125 . in addition , we obtain the dependent parameters @xmath126 and @xmath127 at tree level . we thus edit the input files of superiso v3.3 which by default uses inputs from the 2011 review of particle physics @xcite . [ note : parameters ] ] details on each process follow . the @xmath5-pole observable @xmath8 , defined as @xmath128 has been calculated in the sm including two - loop electroweak @xcite and three - loop qcd corrections . the correction to @xmath8 due to one - loop diagrams involving additional higgs bosons has been calculated in the 2hdm @xcite . in the type - i 2hdm , the contribution of the neutral scalars can be neglected @xcite as it is suppressed by a relative factor of @xmath129 compared to the charged higgs contribution . the results for the type - i 2hdm can easily be adapted to the gm model @xcite . following ref . @xcite , the one - loop charged higgs correction to @xmath130 can be written as depends on the @xmath131 couplings and the bottom quark mass . updated values of these quantities have a very small effect on the coefficient . for example , using more recent values from ref . @xcite the change in the coefficient is 0.1% . ] @xmath132 where @xmath133 , @xmath134 , and we neglect @xmath54 in the loop calculation . dependence have been given in refs . because the constraint from @xmath8 is weaker than the other constraints , we use here only the approximation given in eq . ( [ eq : drb ] ) . ] here @xmath135 is the @xmath136 running mass and is evaluated at @xmath137 . the approximation in eq . ( [ eq : drb ] ) can be made because @xmath138 is suppressed by a factor of @xmath129 compared to @xmath139 @xcite . the correction is always negative and interferes destructively with the sm contribution . the measured value of @xmath8 is @xcite @xmath140 while the sm prediction is @xmath141 @xcite . therefore the 2@xmath116 upper bound relative to the experimental central value yields the `` tight '' constraint @xmath142 , where we have combined the experimental and sm theoretical uncertainties in quadrature . ) depend on the sm @xmath8 prediction in a complex way , the @xmath8 observable can not straightforwardly be calculated using a ratio of the sm and gm constributions ( as we will do with the other observables ) . for this reason , in the @xmath8 case only we take the theory uncertainty on the gm prediction to be the same as that of the sm prediction . ] the sm prediction is 0.8@xmath116 below the measured value ; as the gm correction interferes destructively with the sm contribution , it increases the discrepancy between theory and experiment . as a result , the best agreement with the experimental measurement of @xmath8 in the gm model occurs in the sm limit ( @xmath143 , @xmath144 ) . requiring that the gm prediction for @xmath8 lie within 2@xmath116 of the sm value yields a `` loose '' constraint of @xmath145 . as we see in fig . [ fig : expconst ] , @xmath8 provides the weakest non - oblique tight " and loose " constraints . furthermore , substantial improvement in the @xmath8 experimental measurement is unlikely in the near future , as this would require better z - pole measurements using a next - generation @xmath146 collider like the international linear collider ( ilc ) with the gigaz option @xcite . the effect of charged scalars on @xmath148@xmath149 mixing in a type - i 2hdm has been studied @xcite by adapting results of similar processes on @xmath150@xmath151 mixing @xcite . these can be extended to the @xmath9@xmath147 system , which is more constraining than the @xmath148@xmath149 system due to lower errors in both the experimental measurement and the sm prediction @xcite . to leading order , the oscillation frequency of a @xmath9 meson in the gm model is determined by the mass splitting @xcite @xmath152 here @xmath153 is a scaling factor , @xmath154 is the weak decay constant , @xmath155 is the bag parameter , @xmath156 is the meson mass , and @xmath157 here @xmath158 , @xmath159 , and @xmath160 . we set the top mass renormalization scale @xmath161 , where @xmath118 is the top quark pole mass . the inami - lim functions @xcite @xmath162 , @xmath163 , and @xmath164 are given by @xcite @xmath165 , \nonumber \\ i_{wh}(x , y , z ) & = & y\left[\frac{(2 z - 8) \log y}{(1-y)^2(1-z ) } + \frac{6 z \log x}{(1-x)^2(1-z ) } - \frac{8 - 2 x}{(1-y)(1-x)}\right].\end{aligned}\ ] ] under the assumption that the overall coefficients do not vary substantially due to new scalar contributions , a prediction for @xmath166 in the gm model may be obtained using the ratio @xmath167 since @xmath162 , @xmath164 and @xmath163 are all positive , the gm model contribution interferes constructively with the sm contribution . because the theoretical uncertainty on the mass splitting is due almost entirely to uncertainties in the coefficients of @xmath168 in eq . ( [ eq : dmbs ] ) , we scale the sm theoretical uncertainty @xmath169 by @xmath170 to obtain the theoretical uncertainty in the gm model , i.e. , @xmath171 . the measured value for the @xmath172@xmath173 mass difference is given by @xcite @xmath174 the largest uncertainty in the sm prediction comes from the lattice qcd calculation of @xmath175 . using a ckmfitter @xcite average of several lattice results based on separate extractions of @xmath176 and @xmath177 , ref . @xcite obtains the sm prediction @xmath178 . however , a preliminary lattice calculation of the product @xmath175 from the fermilab / milc collaboration @xcite yields a larger uncertainty and considerably different central value , leading to @xmath179 @xcite . for our numerical results we use the ckmfitter central value but take the more conservative uncertainty as advocated in ref . @xcite , @xmath180 the above results can be translated into an experimental measurement of @xmath181 and combined experimental and sm theoretical uncertainty of @xmath182 the sm prediction @xmath183 is thus only 0.13@xmath116 below the measured value . in the case of @xmath9@xmath147 mixing , the charged higgs contributions in the gm model increase the predicted value of @xmath184 , so that the best - fit value and the experimental central value are the same . thus our `` tight '' and `` loose '' constraints from this observable are the same . the 2@xmath116 constraint is @xmath185 , where we have combined the experimental and gm theoretical uncertainties in quadrature . the resulting constraint on the @xmath115@xmath113 plane is shown in the left and right panels of fig . [ fig : expconst ] . it is slightly more constraining than the bounds from @xmath8 , and is about the same as the `` loose '' bound from @xmath186 . , the uncertainty on @xmath187 becomes @xmath188 and the bound would tighten to match the `` tight '' bound from @xmath186 in the left - hand panel . if we were instead to use the central value @xmath189 , the best - fit reference point would become the sm prediction , @xmath190 , and the `` loose '' and `` tight '' bounds would each be slightly stronger than the corresponding bounds from @xmath191 . this variability illustrates the very large remaining theoretical uncertainty in this observable . ] in both cases the bound is weaker than that from @xmath191 . an improvement in the constraint from @xmath172@xmath192 mixing relies on an improved lattice determination of @xmath193 . a full leading - order computation of the average time - integrated branching ratio @xmath194 in the aligned 2hdm @xcite was recently performed in ref . the calculation can be easily specialized to the type - i 2hdm and hence to the gm model ; the result is conveniently expressed in terms of a ratio to the sm prediction , @xmath195 where the wilson coefficients @xmath196 and @xmath197 are given by @xcite @xmath198^{1.53 } \left[\frac{\alpha_s(m_z)}{0.1184}\right]^{-0.09 } , \nonumber \\ c_{10}^{\rm gm } & = & c_{10}^{\rm sm } + \tan^2\theta_h \frac{x_{tw}}{8 } \left [ \frac{x_{t3}}{1-x_{t3 } } + \frac{x_{t3 } \log x_{t3}}{(1-x_{t3})^2 } \right ] , \label{eq : c10gm}\end{aligned}\ ] ] with @xmath159 and @xmath199 as before . , the expression in square brackets can be expanded in powers of @xmath200 and reads @xmath201 \simeq - \frac{1}{2 } - \frac{\delta}{6 } + \mathcal{o}(\delta^2).\ ] ] ] here @xmath161 is the top quark pole mass and @xmath202 is the @xmath203 running mass . the theoretical uncertainty on the resulting gm branching ratio is taken to be @xmath204 . the approximation made in eq . ( [ eq : rsmu ] ) is to neglect contributions from the wilson coefficients @xmath205 and @xmath206 , which arise from scalar and pseudoscalar penguins and box diagrams and are suppressed by an extra factor of @xmath129 compared to @xmath207 . the expression for @xmath196 includes next - to - leading order ( nlo ) electroweak and qed corrections , as well as nlo and next - to - next - to - leading order ( nnlo ) qcd corrections . we note that the expression for the @xmath106 contribution to @xmath197 has the same dependence on @xmath115 and @xmath114 as the expression for the correction to @xmath8 in the @xmath208 limit given in eq . ( [ eq : drb ] ) . this is because the charged higgs contribution to @xmath197 comes from the same @xmath5 penguin diagrams as in @xmath8 , but with a generation - changing @xmath209 vertex in place of the generation - conserving @xmath210 vertex and @xmath211 . the current experimental measurement of @xmath194 from a combination of cms and lhcb results is @xcite @xmath212 the sm prediction is given in ref . @xcite as @xmath213 this number differs slightly from the result in ref . @xcite , upon which it is based , due to the use of a slightly different central value and more conservative uncertainty on the top quark mass . these yield an experimental measurement of @xmath214 and combined experimental and sm theoretical uncertainty of @xmath215 in particular , the sm prediction , @xmath216 , is 1.0@xmath116 above the current experimental value . the 2@xmath116 constraint on the gm model relative to the experimental central value yields a bound of @xmath217 , where we have combined the experimental and gm model theoretical uncertainties in quadrature . tight " constraint is shown in the left - hand panel of fig . [ fig : expconst ] ; it is stronger than the corresponding constraints from @xmath8 and @xmath9@xmath147 mixing , as previously discussed , but remains weaker than the tight " @xmath191 constraint . however , the gm model contribution to @xmath207 in eq . ( [ eq : c10gm ] ) is always negative , leading to constructive interference with the sm contribution and increasing the prediction for @xmath218 compared to its value in the sm . as the sm value is already larger than the experimental value ( @xmath219 ) , the best agreement with the experimental measurement of @xmath194 in the gm model occurs in the limit @xmath43 or @xmath117 ( i.e. , the sm limit ) . the best - fit 2@xmath116 bound taken relative to the sm prediction yields a loose " constraint of @xmath220 , which is shown in the right - hand panel of fig . [ fig : expconst ] . the `` loose '' constraint from @xmath194 is also weaker than that from @xmath112 . the current uncertainty on @xmath214 is dominated by the experimental statistical uncertainty . this has the potential to be significantly reduced in the near future as more data is collected at the lhc . in particular , the upgraded lhcb experiment is expected to measure @xmath194 with an ultimate experimental uncertainty of better than 10% with 50 fb@xmath221 of data @xcite , which corresponds to about ten years of lhc running . assuming an experimental rate consistent with the sm prediction and no change in the theoretical uncertainty , this would correspond to a combined uncertainty on @xmath222 of 0.12 . this measurement thus has the potential to become the most stringent constraint on the gm model parameter space in the near future . the @xmath223 branching ratio has been measured at several different experiments , including cleo , babar , belle , and aleph . the current experimental average from the heavy flavour averaging group is @xcite @xcite . ] @xmath224 for a photon energy @xmath225 gev . @xmath112 is known up to nnlo in qcd in the sm @xcite . , resulting in a contribution to the theory uncertainty from the interpolation ambiguity . ] the two current sm predictions are @xmath226 @xcite and @xmath227 @xcite . these predictions differ due to different approaches to handling higher - order contributions to the photon energy cutoff corrections ; however , their difference is within the @xmath228 theoretical uncertainty due to uncalculated higher orders @xcite . the charged higgs contributions in the type - i 2hdm , first calculated in ref . @xcite , are themselves now known up to nlo in qcd @xcite . because @xmath112 will provide the most stringent constraint on the gm model parameter space , we will use the full implementation of the sm and 2hdm contributions in the public code superiso v3.3 @xcite , which is based on the calculations in refs . superiso calls the code 2hdmc v1.6.4 @xcite for spectrum calculations within the type - i 2hdm . in the limit @xmath43 or @xmath117 , the calculation of @xmath229 by superiso v3.3 , using the input parameters given in footnote [ note : parameters ] , yields a prediction @xmath230 the difference compared to the sm predictions quoted above is primarily due to differences in the input parameters , particularly @xmath231 and @xmath232 @xcite . however , the difference is still within the theoretical uncertainty due to parametric uncertainties of @xmath228 @xcite . we take the total theoretical uncertainty on this sm prediction to be @xmath233 from ref . @xcite . combining this in quadrature with the experimental uncertainty yields a total uncertainty of @xmath234 . in particular , the value of @xmath229 in the sm limit is 1.3@xmath116 below the experimental value . the charged higgs contribution to @xmath229 in the gm model interferes destructively with the sm contribution , leading to a smaller predicted value for @xmath229 than in the sm . because the sm prediction is already below the experimental central value , the best agreement with the experimental measurement in the gm model occurs in the limit @xmath43 or @xmath117 ( i.e. , the sm limit ) . since even the sm limit yields a prediction that is only 0.7@xmath116 from the experimental bound , the 2@xmath116 experimental constraint on the gm @xmath115@xmath113 plane is quite strong , as can be seen in the left panel of fig . [ fig : expconst ] . this bound corresponds to @xmath235 ( `` tight '' constraint ) , where we have combined the experimental and gm theoretical uncertainties in quadrature ; again , here we estimate the gm theory uncertainty to be that of the sm prediction scaled by a ratio of the gm and sm predictions . in comparison , the @xmath236 constraint with respect to the best - fit point , the sm limit , yields @xmath237 ( `` loose '' constraint ) . this is shown in the right panel of fig . [ fig : expconst ] together with the `` loose '' constraints from the other observables . in either case , @xmath229 is the strongest constraint on these parameters . because of the large theoretical uncertainty on @xmath229 and the sensitivity of the resulting constraint to the particular choice of input parameters and the handling of partial higher - order corrections , we consider it safer to take the more conservative approach and apply the `` loose '' constraint from @xmath229 as our primary constraint on the @xmath115@xmath113 plane . we will nevertheless also show the effect of applying the `` tight '' @xmath1 constraint in our numerical scans . the current theoretical and experimental uncertainties on @xmath229 are comparable in size . the experimental uncertainty is expected to be reduced with measurements at the super @xmath0 factory experiment belle ii currently under construction at kek . a conservative treatment of systematics yields an estimated future experimental precision on @xmath229 of 7% ( i.e. , about @xmath238 ) with 5 ab@xmath221 of data , or 6% ( i.e. , about @xmath239 ) with 50 ab@xmath221 of data @xcite . with the current theoretical uncertainties , these would reduce the combined uncertainty only to about @xmath240 or @xmath241 , respectively . a more significant improvement in the constraining power of @xmath229 would require a simultaneous reduction in the theoretical uncertainty . we now illustrate the effects of the indirect experimental constraints from @xmath112 ( computed using superiso v3.3 @xcite , which calls 2hdmc v1.6.4 @xcite ) and the @xmath7 parameter on the parameter space of the gm model . we scan over the full range of gm model parameters allowed after imposing the theoretical requirements of perturbative unitarity , bounded - from - belowness of the potential , and the absence of alternative custodial - symmetry breaking minima @xcite . we require that either @xmath3 or @xmath242 has mass 125 gev and set the sm higgs vev @xmath243 using @xmath244 . we take @xmath245 , which fully populates the mass ranges shown in figs . [ fig : mvchi][fig : kfkvmass ] below . in fig . [ fig : kfkvmass ] we will include additional points generated by a dedicated scan with @xmath246 in order to better populate the low - mass region . in all cases , we show the effects of the following constraints : * the prediction for @xmath7 yields @xmath247 after marginalizing over the @xmath74 parameter . points eliminated by this constraint are shown by red ( medium gray ) @xmath248 shapes . * the prediction for @xmath112 lies within @xmath236 of the model point that gives the best agreement with the experimental measurement ( `` loose '' constraint ) . we combine theoretical and experimental uncertainties in quadrature . points eliminated by this constraint are shown by light green ( light gray ) @xmath249 shapes . * the prediction for @xmath112 lies within @xmath236 of the experimental measurement ( `` tight '' constraint ) . we combine theoretical and experimental uncertainties in quadrature . points eliminated by this constraint are shown by dark green ( dark gray ) @xmath249 shapes . points depicted in black are allowed by all constraints . we start by showing the effect of the @xmath1 measurement on the @xmath115@xmath113 plane in the left panel of fig . [ fig : mvchi ] . the prediction for @xmath112 in the gm model depends only on these two parameters . we see that , due to its interplay with the decoupling effect of falling @xmath113 with increasing triplet masses @xcite , the `` loose '' @xmath1 constraint eliminates all model points with @xmath250 gev and the `` tight '' @xmath1 constraint eliminates all model points with @xmath251 gev . obtained in the spirit of ref . @xcite by requiring @xmath252 to avoid parameter regions in which the top quark yukawa coupling becomes too large ( @xmath253 plays the same role as is played by @xmath254 in the type - i 2hdm ) ; this requirement yields @xmath255 gev . ] this is reflected in the right panel of fig . [ fig : mvchi ] , where we plot @xmath113 as a function of @xmath256 . because @xmath257 in general , values of @xmath113 up to the limit of @xmath258 gev are allowed even for @xmath256 masses as low as 100 gev under the `` loose '' @xmath1 constraint . this indirect constraint on @xmath113 as a function of @xmath256 is especially interesting in light of the recent recasting @xcite of an atlas measurement @xcite of the like - sign @xmath259 cross section in 8 tev data in the context of the gm model . the like - sign @xmath259 cross section receives contributions especially from the @xmath260-channel production of @xmath261 in @xmath262 fusion , followed by decays back to @xmath262 . the analysis of ref . @xcite excludes a triangular region of parameter space in the @xmath256@xmath113 plane extending from @xmath263 gev to 610 gev at @xmath264 gev , down to @xmath265 gev at @xmath266 gev . in the right panel of fig . [ fig : mvchi ] we also see the effect of the @xmath7 parameter constraint , which eliminates a few model points at very low @xmath256 , as well as moderate to high @xmath256 and high @xmath113 . this is illustrated in more detail in fig . [ fig : m5 ] . in particular , points with very low values of @xmath256 tend to have a large @xmath256@xmath115 splitting , which leads to large positive values of the @xmath7 parameter . finally we show the effect of the constraints from @xmath112 and the @xmath7 parameter on the allowed ranges of the couplings of the 125 gev higgs boson @xmath3 to @xmath4 and @xmath5 boson pairs and to fermion pairs . we parameterize these couplings in terms of scaling factors @xmath267 and @xmath268 @xcite , which represent the @xmath269 ( @xmath270 ) and @xmath271 couplings , respectively , normalized to their sm values . in the gm model , these couplings are given in terms of the triplet vev @xmath113 and the custodial singlet mixing angle @xmath272 by @xmath273 where @xmath274 corresponds to the sm higgs vev . , and the lighter custodial singlet @xmath3 has a mass below 125 gev . in these cases , we plot the coupling scaling factors @xmath267 and @xmath268 that represent the @xmath275 ( @xmath270 ) and @xmath276 couplings , respectively , normalized to their sm values . these couplings are given in this case by @xmath277 ] one of the most interesting features of the gm model is the possibility that @xmath278 , which is not possible at tree level in extended higgs sectors that contain only su(2)@xmath96 doublets and/or singlets . in particular , for maximal @xmath113 and @xmath279 ( corresponding to @xmath3 being entirely composed of triplet ) , @xmath267 can be as large as 1.6 . this maximal value is reduced to @xmath280 by the upper bound on @xmath113 imposed by the `` loose '' @xmath1 constraint , as shown in the left panel of fig . [ fig : kappas ] . it would be reduced even further to @xmath281 by the `` tight '' @xmath1 constraint . similarly , @xmath268 can be significantly enhanced in the gm model if @xmath282 is small . by limiting the maximum size of @xmath113 , the `` loose '' @xmath1 constraint puts a lower bound on @xmath282 , and thereby imposes an upper bound on @xmath268 of @xmath283 , as shown in the right panel of fig . [ fig : kappas ] . the `` tight '' constraint on @xmath1 would reduce this further to @xmath284 . the @xmath7 parameter measurement does not further constrain the allowed ranges of either @xmath267 or @xmath268 in a significant way once either of the @xmath1 constraints has been applied . finally we show the allowed range of correlations between @xmath267 and @xmath268 in fig . [ fig : kvkfcor ] . we note in particular that the gm model can accommodate _ simultaneous _ enhancements of both @xmath267 and @xmath268 . such enhancements are constrained by the @xmath1 measurement to lie below @xmath285 ( `` loose '' constraint ) . the `` tight '' @xmath286 constraint would reduce this to about 1.09 . this is interesting primarily because higgs coupling fits from lhc data suffer from a flat direction @xcite if unobserved decay modes are allowed , corresponding to a simultaneous increase in the unobserved decay branching ratio and in all the higgs couplings to sm particles . this flat direction can be cut off by imposing additional theory assumptions , such as the absence of new , unobserved higgs decay modes @xcite or the imposition of @xmath287 valid when the higgs sector contains only isospin doublets and/or singlets @xcite . the gm model provides a concrete example of a model that violates the second assumption while being consistent with other experimental constraints . the flat direction could also be tamed by constraining the total higgs width through measurements of off - shell @xmath288 @xcite ; the interpretation of this measurement in terms of a higgs width constraint , however , is itself model - dependent @xcite and it is not yet clear what effect the presence of additional higgs states will have . crucially , however , the simultaneous enhancement of @xmath267 and @xmath268 occurs only when the new scalars are relatively light . this is illustrated in fig . [ fig : kfkvmass ] , where we plot @xmath267 as a function of the mass of the lightest new scalar , for @xmath268 within 5% ( red ) or 10% ( blue ) of @xmath267 . the remaining points are shown in green . under the `` loose '' @xmath1 constraint , for @xmath268 within 5% of @xmath267 , an 18% enhancement of these couplings is possible only when at least one of the new scalars has mass below about 375 gev . this provides a complementary ( albeit model dependent ) way to constrain the flat direction by directly searching for the new scalars . we leave a full consideration of the direct - search constraints on these additional scalars to future work . in this paper we updated the indirect experimental constraints on the gm model coming from electroweak and @xmath0-physics observables in particular the @xmath7 parameter , @xmath8 , @xmath1 , @xmath9@xmath10 mixing , and @xmath12 . except for the @xmath7 parameter , all of these constrain only two of the gm model parameters : the isospin - triplet vev @xmath113 and the custodial - triplet mass @xmath115 . we gave the analytic expressions for the one - loop contributions from the additional higgs bosons for the @xmath7 parameter , @xmath8 , @xmath9@xmath10 mixing , and @xmath12 ; in the case of the @xmath7 parameter and @xmath8 these are in the approximation that the new scalars are heavy compared to @xmath54 . for @xmath1 we adapted the 2hdm calculation implemented in the code superiso . the constraints from @xmath1 , @xmath9@xmath10 mixing , and @xmath11 have not been studied in the gm model before . we found that @xmath1 is currently the strongest of the @xmath0-physics constraints on the gm model . however , this may be surpassed in the next few years by the constraint from @xmath12 , which will become more important as its statistical uncertainty is reduced with further lhc data - taking . combined with the theoretical requirements of vacuum stability and perturbativity , the @xmath289 constraint puts a conservative upper bound of about 65 gev on the isospin - triplet vev @xmath113 , which leads to upper bounds on the @xmath14 , @xmath6 , and @xmath290 couplings . in particular , a _ simultaneous _ enhancement of the @xmath14 , @xmath6 , and @xmath290 couplings of up to 18% compared to their sm values is still allowed by the indirect constraints , leading to a simultaneous enhancement of all the higgs production cross sections by up to 39% . such an enhancement could mask ( and be masked by ) the presence of undetected new decay modes of the sm - like higgs boson at the lhc . we thank f. mahmoudi for helpful discussions about superiso . also thanks the organizers of the lisbon workshop on multi - higgs models 2014 for partial support and a stimulating environment while part of this work was performed . this work was supported by the natural sciences and engineering research council of canada . was also supported by the government of ontario through an ontario graduate scholarship . g. aad _ et al . _ [ atlas collaboration ] , phys . b * 716 * , 1 ( 2012 ) [ arxiv:1207.7214 [ hep - ex ] ] ; s. chatrchyan _ et al . _ [ cms collaboration ] , phys . b * 716 * , 30 ( 2012 ) [ arxiv:1207.7235 [ hep - ex ] ] . h. georgi and m. machacek , nucl . b * 262 * , 463 ( 1985 ) . m. s. chanowitz and m. golden , phys . b * 165 * , 105 ( 1985 ) . j. f. gunion , r. vega and j. wudka , phys . d * 42 * , 1673 ( 1990 ) . j. f. gunion , r. vega and j. wudka , phys . d * 43 * , 2322 ( 1991 ) . j. f. gunion , h. e. haber , g. l. kane , and s. dawson , _ the higgs hunter s guide _ ( westview , boulder , colorado , 2000 ) . h. e. haber and h. e. logan , phys . d * 62 * , 015011 ( 2000 ) [ hep - ph/9909335 ] . m. aoki and s. kanemura , phys . d * 77 * , 095009 ( 2008 ) [ arxiv:0712.4053 [ hep - ph ] ] ; erratum phys . d * 89 * , 059902 ( 2014 ) . s. godfrey and k. moats , phys . rev . d * 81 * , 075026 ( 2010 ) [ arxiv:1003.3033 [ hep - ph ] ] . i. low and j. lykken , jhep * 1010 * , 053 ( 2010 ) [ arxiv:1005.0872 [ hep - ph ] ] ; i. low , j. lykken and g. shaughnessy , phys . rev . d * 86 * , 093012 ( 2012 ) [ arxiv:1207.1093 [ hep - ph ] ] . h. e. logan and m .- a . roy , phys . d * 82 * , 115011 ( 2010 ) [ arxiv:1008.4869 [ hep - ph ] ] . a. falkowski , s. rychkov and a. urbano , jhep * 1204 * , 073 ( 2012 ) [ arxiv:1202.1532 [ hep - ph ] ] . s. chang , c. a. newby , n. raj and c. wanotayaroj , phys . d * 86 * , 095015 ( 2012 ) [ arxiv:1207.0493 [ hep - ph ] ] . d. carmi , a. falkowski , e. kuflik , t. volansky and j. zupan , jhep * 1210 * , 196 ( 2012 ) [ arxiv:1207.1718 [ hep - ph ] ] . chiang and k. yagyu , jhep * 1301 * , 026 ( 2013 ) [ arxiv:1211.2658 [ hep - ph ] ] . chiang , a .- l . kuo and k. yagyu , * 1310 * , 072 ( 2013 ) [ arxiv:1307.7526 [ hep - ph ] ] . s. kanemura , m. kikuchi and k. yagyu , phys . d * 88 * , 015020 ( 2013 ) [ arxiv:1301.7303 [ hep - ph ] ] . c. englert , e. re and m. spannowsky , phys . d * 87 * , 095014 ( 2013 ) [ arxiv:1302.6505 [ hep - ph ] ] . r. killick , k. kumar and h. e. logan , phys . d * 88 * , 033015 ( 2013 ) [ arxiv:1305.7236 [ hep - ph ] ] . g. belanger , b. dumont , u. ellwanger , j. f. gunion and s. kraml , phys . d * 88 * , 075008 ( 2013 ) [ arxiv:1306.2941 [ hep - ph ] ] . c. englert , e. re and m. spannowsky , phys . d * 88 * , 035024 ( 2013 ) [ arxiv:1306.6228 [ hep - ph ] ] . a. efrati and y. nir , arxiv:1401.0935 [ hep - ph ] . k. hartling , k. kumar and h. e. logan , phys . d * 90 * , 015007 ( 2014 ) [ arxiv:1404.2640 [ hep - ph ] ] . chiang and t. yamada , arxiv:1404.5182 [ hep - ph ] . chiang , s. kanemura and k. yagyu , arxiv:1407.5053 [ hep - ph ] . s. i. godunov , m. i. vysotsky and e. v. zhemchugov , arxiv:1408.0184 [ hep - ph ] . s. chang and j. g. wacker , phys . d * 69 * , 035002 ( 2004 ) [ hep - ph/0303001 ] . s. chang , jhep * 0312 * , 057 ( 2003 ) [ hep - ph/0306034 ] . l. cort , m. garcia and m. quiros , phys . d * 88 * , 075010 ( 2013 ) [ arxiv:1308.4025 [ hep - ph ] ] . m. garcia - pepin , s. gori , m. quiros , r. vega , r. vega - morales and t. t. yu , arxiv:1409.5737 [ hep - ph ] . s. el hedri , p. j. fox and j. g. wacker , arxiv:1311.6488 [ hep - ph ] . p. galison , nucl . b * 232 * , 26 ( 1984 ) . r. w. robinett , phys . d * 32 * , 1780 ( 1985 ) . h. e. logan , hep - ph/9906332 . j. hisano and k. tsumura , phys . d * 87 * , 053004 ( 2013 ) [ arxiv:1301.6455 [ hep - ph ] ] . c. alvarado , l. lehman and b. ostdiek , jhep * 1405 * , 150 ( 2014 ) [ arxiv:1404.3208 [ hep - ph ] ] . for a recent review , see g. c. branco , p. m. ferreira , l. lavoura , m. n. rebelo , m. sher and j. p. silva , phys . * 516 * , 1 ( 2012 ) [ arxiv:1106.0034 [ hep - ph ] ] . f. mahmoudi , comput . commun . * 178 * , 745 ( 2008 ) [ arxiv:0710.2067 [ hep - ph ] ] ; comput . commun . * 180 * , 1579 ( 2009 ) [ arxiv:0808.3144 [ hep - ph ] ] ; comput . commun . * 180 * , 1718 ( 2009 ) . x .- q . li , j. lu and a. pich , arxiv:1404.5865 [ hep - ph ] . a. pich and p. tuzon , phys . d * 80 * , 091702 ( 2009 ) [ arxiv:0908.1554 [ hep - ph ] ] . h. baer , t. barklow , k. fujii , y. gao , a. hoang , s. kanemura , j. list and h. e. logan _ et al . _ , arxiv:1306.6352 [ hep - ph ] . m. e. peskin and t. takeuchi , phys . d * 46 * , 381 ( 1992 ) . m. baak , j. cuth , j. haller , a. hoecker , r. kogler , k. moenig , m. schott and j. stelzer , arxiv:1407.3792 [ hep - ph ] . k. a. olive _ et al . _ [ particle data group collaboration ] , chin . phys . c * 38 * , 090001 ( 2014 ) . atlas , cdf , cms , and d0 collaborations , arxiv:1403.4427 [ hep - ex ] . k. nakamura _ et al . _ [ particle data group collaboration ] , j. phys . g * 37 * , 075021 ( 2010 ) . a. freitas and y .- c . huang , jhep * 1208 * , 050 ( 2012 ) [ erratum - ibid . * 1305 * , 074 ( 2013 ) ] [ erratum - ibid . * 1310 * , 044 ( 2013 ) ] [ arxiv:1205.0299 [ hep - ph ] ] . a. denner , r. j. guth , w. hollik and j. h. kuhn , z. phys . c * 51 * , 695 ( 1991 ) . a. k. grant , phys . d * 51 * , 207 ( 1995 ) [ hep - ph/9410267 ] . u. haisch and s. westhoff , jhep * 1108 * , 088 ( 2011 ) [ arxiv:1106.0529 [ hep - ph ] ] . g. g. athanasiu , p. j. franzini and f. j. gilman , phys . d * 32 * , 3010 ( 1985 ) . l. f. abbott , p. sikivie and m. b. wise , phys . rev . d * 21 * , 1393 ( 1980 ) . a. lenz , arxiv:1205.1444 [ hep - ph ] . a. wahab el kaffas , p. osland and o. m. ogreid , phys . 095001 ( 2007 ) [ arxiv:0706.2997 [ hep - ph ] ] . j. urban , f. krauss , u. jentschura and g. soff , nucl . b * 523 * , 40 ( 1998 ) [ hep - ph/9710245 ] . t. inami and c. s. lim , prog . phys . * 65 * , 297 ( 1981 ) [ erratum - ibid . * 65 * , 1772 ( 1981 ) ] . f. mahmoudi and o. stal , phys . d * 81 * , 035016 ( 2010 ) [ arxiv:0907.1791 [ hep - ph ] ] . j. beringer _ [ particle data group collaboration ] , phys . d * 86 * , 010001 ( 2012 ) . a. lenz , u. nierste , j. charles , s. descotes - genon , h. lacker , s. monteil , v. niess and s. tjampens , phys . d * 86 * , 033008 ( 2012 ) [ arxiv:1203.0238 [ hep - ph ] ] . ckmfitter results are summarized at http://ckmfitter.in2p3.fr . the rfit method used by ckmfitter is described in j. charles _ et al . _ [ ckmfitter group collaboration ] , eur . j. c * 41 * , 1 ( 2005 ) [ hep - ph/0406184 ] . u. nierste , arxiv:1212.5805 [ hep - ph ] . c. m. bouchard , e. d. freeland , c. bernard , a. x. el - khadra , e. gamiz , a. s. kronfeld , j. laiho and r. s. van de water , pos lattice * 2011 * , 274 ( 2011 ) [ arxiv:1112.5642 [ hep - lat ] ] . a. lenz and u. nierste , arxiv:1102.4274 [ hep - ph ] . cms and lhcb collaborations [ cms and lhcb collaboration ] , cms - pas - bph-13 - 007 . c. bobeth , m. gorbahn , t. hermann , m. misiak , e. stamou and m. steinhauser , phys . lett . * 112 * , 101801 ( 2014 ) [ arxiv:1311.0903 [ hep - ph ] ] . r. aaij _ et al . _ [ lhcb collaboration ] , eur . j. c * 73 * , 2373 ( 2013 ) [ arxiv:1208.3355 [ hep - ex ] ] . d. asner _ [ heavy flavor averaging group collaboration ] , arxiv:1010.1589 [ hep - ex ] . j. p. lees _ et al . _ [ babar collaboration ] , phys . d * 86 * , 112008 ( 2012 ) [ arxiv:1207.5772 [ hep - ex ] ] . m. misiak , h. m. asatrian , k. bieri , m. czakon , a. czarnecki , t. ewerth , a. ferroglia and p. gambino _ et al . _ , phys . rev . lett . * 98 * , 022002 ( 2007 ) [ hep - ph/0609232 ] . t. becher and m. neubert , phys . lett . * 98 * , 022003 ( 2007 ) [ hep - ph/0610067 ] . v. d. barger , j. l. hewett and r. j. n. phillips , phys . d * 41 * , 3421 ( 1990 ) . m. ciuchini , g. degrassi , p. gambino and g. f. giudice , nucl . b * 527 * , 21 ( 1998 ) [ hep - ph/9710335 ] . m. misiak and m. steinhauser , nucl . b * 764 * , 62 ( 2007 ) [ hep - ph/0609241 ] . d. eriksson , j. rathsman and o. stal , comput . commun . * 181 * , 189 ( 2010 ) [ arxiv:0902.0851 [ hep - ph ] ] ; comput . commun . * 181 * , 833 ( 2010 ) . t. aushev , w. bartel , a. bondar , j. brodzicka , t. e. browder , p. chang , y. chao and k. f. chen _ et al . _ , arxiv:1002.5012 [ hep - ex ] . atlas collaboration , atlas - conf-2014 - 013 , available from ` http://cds.cern.ch ` . a. david _ et al . _ [ lhc higgs cross section working group collaboration ] , arxiv:1209.0040 [ hep - ph ] . d. zeppenfeld , r. kinnunen , a. nikitenko and e. richter - was , phys . d * 62 * , 013009 ( 2000 ) [ hep - ph/0002036 ] ; a. djouadi , r. kinnunen , e. richter - was , h. u. martyn , k. a. assamagan , c. balazs , g. belanger and e. boos _ et al . _ , hep - ph/0002258 . m. duhrssen , s. heinemeyer , h. logan , d. rainwater , g. weiglein and d. zeppenfeld , phys . d * 70 * , 113009 ( 2004 ) [ hep - ph/0406323 ] ; k. a. assamagan _ et al . _ [ higgs working group collaboration ] , hep - ph/0406152 . f. caola and k. melnikov , phys . d * 88 * , 054024 ( 2013 ) [ arxiv:1307.4935 [ hep - ph ] ] . cms collaboration , cms - pas - hig-14 - 002 ; the atlas collaboration , atlas - conf-2014 - 042 ; available from http://cds.cern.ch . c. englert and m. spannowsky , arxiv:1405.0285 [ hep - ph ] ; m. ghezzi , g. passarino and s. uccirati , arxiv:1405.1925 [ hep - ph ] ; g. cacciapaglia , a. deandrea , g. d. la rochelle and j. b. flament , arxiv:1406.1757 [ hep - ph ] ; a. azatov , c. grojean , a. paul and e. salvioni , arxiv:1406.6338 [ hep - ph ]

we study the large - coalition asymptotics of fingerprinting and group testing , and derive explicit decoders that provably achieve capacity for many of the considered models . we do this both for simple decoders ( fast but suboptimal ) and for joint decoders ( slow but optimal ) , and both for informed and uninformed settings . for fingerprinting , we show that if the pirate strategy is known , the neyman - pearson - based log - likelihood decoders provably achieve capacity , regardless of the strategy . the decoder built against the interleaving attack is further shown to be a universal decoder , able to deal with arbitrary attacks and achieving the uninformed capacity . this universal decoder is shown to be closely related to the lagrange - optimized decoder of oosterwijk et al . and the empirical mutual information decoder of moulin . joint decoders are also proposed , and we conjecture that these also achieve the corresponding joint capacities . for group testing , the simple decoder for the classical model is shown to be more efficient than the one of chan et al . and it provably achieves the simple group testing capacity . for generalizations of this model such as noisy group testing , the resulting simple decoders also achieve the corresponding simple capacities . fingerprinting , traitor tracing , group testing , log - likelihood ratios , hypothesis testing .

1404.2825

protect copyrighted content against unauthorized redistribution , distributors commonly embed watermarks or fingerprints in the content , uniquely linking copies to individual users . if the distributor finds an illegal copy of the content online , he can then extract the watermark from this copy and compare it to the database of watermarks , to determine which user was responsible . to combat this solution , a group of @xmath0 pirates may try to form a coalition and perform a collusion attack . by comparing their unique versions of the content , they will detect differences in their copies which must be part of the watermark . they can then try to create a mixed pirate copy , where the resulting watermark matches the watermark of different pirates in different segments of the content , making it hard for the distributor to find the responsible users . the goal of the distributor of the content is to assign the watermarks to the users in such a way that , even if many pirates collude , the pirate copy can still be traced back to the responsible users . a different area of research that has received considerable attention in the last few decades is group testing , introduced by dorfman @xcite in the 1940s . suppose a large population contains a small number @xmath0 of infected ( or defective ) items . to identify these items , it is possible to perform group tests : testing a subset of the population will lead to a positive test result if this subset contains at least one defective item , and a negative result otherwise . since the time to run a single test may be very long , the subsets to test need to be chosen in advance , after which all group tests are performed simultaneously . then , when the test results come back , the subset of defective items needs to be identified . the goal of the game is to identify these defectives using as few group tests as possible , and with a probability of error as small as possible . the above problems of fingerprinting and group testing can be jointly modeled by the following two - person game between ( in terms of fingerprinting ) the distributor @xmath1 and the adversary @xmath2 ( the set of colluders , or the set of defectives ) . throughout the paper we will mostly use terminology from fingerprinting ( i.e. users instead of items , colluders instead of defective items ) , unless we are specifically dealing with group testing results . first , there is a universe @xmath3 of @xmath4 users , and the adversary is assigned a random subset of users @xmath5 of size @xmath6 . this subset @xmath2 is unknown to the distributor ( but we assume that the distributor does know the size @xmath0 of @xmath2 ) , and the aim of the game for the distributor is ultimately to discover @xmath2 . the two - person game consists of three phases : ( 1 ) the distributor uses an _ encoder _ to generate a fingerprinting code , used for assigning versions to users ; ( 2 ) the colluders employ a _ collusion channel _ to generate the pirate output from their given code words ; and ( 3 ) the distributor uses a _ decoder _ to map the pirate output to a set @xmath7 . the distributor generates a fingerprinting code @xmath8 of @xmath4 binary code words of length @xmath9 . , but in this paper we restrict our attention to the binary case @xmath10 . ] the parameter @xmath9 is referred to as the code length , and the distributor would like @xmath9 to be as small as possible . for the eventual embedded watermark , we assume that for each segment of the content there are two differently watermarked versions , so the watermark of user @xmath11 is determined by the @xmath9 entries in the @xmath11th code word of @xmath8 . a common restriction on the encoding process is to assume that @xmath8 is created by first generating a bias vector @xmath12 ( by choosing each entry @xmath13 independently from a certain distribution @xmath14 ) , and then generating code words @xmath15 according to @xmath16 . this guarantees that watermarks of different users @xmath11 are independent , and that watermarks in different positions @xmath17 are independent . fingerprinting schemes that satisfy this assumption are sometimes called bias - based schemes , and the encoders in this paper ( both for group testing and fingerprinting ) are also assumed to belong to this category . after generating @xmath8 , the code words are used to select and embed watermarks in the content , and the content is sent out to all users . the colluders then get together , compare their copies , and use a certain collusion channel or pirate attack @xmath18 to determine the pirate output @xmath19 . if the pirate attack behaves symmetrically both in the colluders and in the positions @xmath17 , then the collusion channel can be modeled by a vector @xmath20^{c+1}$ ] , consisting of entries @xmath21 ( for @xmath22 ) indicating the probability of outputting a @xmath23 when the pirates received @xmath24 ones and @xmath25 zeroes . a further restriction on @xmath26 in fingerprinting is the marking assumption , introduced by boneh and shaw @xcite , which says that @xmath27 and @xmath28 , i.e. , if the pirates receive only zeros or ones they have to output this symbol . finally , after the pirate output has been generated and distributed , we assume the distributor intercepts it and applies a decoding algorithm to @xmath29 , @xmath8 and @xmath30 to compute a set @xmath7 of accused users . the distributor wins the game if @xmath31 and loses if this is not the case . while the above model is described in fingerprinting terminology , it also covers many common group testing models . the users then correspond to items , the colluders translate to defectives , the code @xmath8 corresponds to the group testing matrix @xmath32 ( where @xmath33 if item @xmath11 is included in the @xmath17th test ) , and the pirate output corresponds to positive / negative test results . the collusion channel is exactly what separates group testing from fingerprinting : while in fingerprinting it is commonly assumed that this channel is not known or only weakly known to the distributor , in group testing this channel is usually assumed known in advance . this means that there is no malicious adversary in group testing , but only a randomization procedure that determines @xmath29 . note also that in ( noisy ) group testing , the boneh - shaw marking assumption may not always hold . work on the fingerprinting game described above started in the late 90s , and lower bounds on the code length were established of the order @xmath34 @xcite , until in 2003 tardos @xcite proved a lower bound of the order @xmath35 and described a scheme with @xmath36 , showing this bound is tight . since the leading constants of the upper and lower bounds did not match , later work on fingerprinting focused on finding the optimal leading constant . based on channel capacities , amiri and tardos @xcite and huang and moulin @xcite independently derived the optimal leading constant to be @xmath37 ( i.e. an asymptotic code length of @xmath38 ) and many improvements to tardos s original scheme were made @xcite to reduce the leading constant from @xmath39 to @xmath40 . recently it was shown that with tardos original ` score function ' one can not achieve capacity @xcite , which lead to the study of different score functions . based on a result of abbe and zheng @xcite , meerwald and furon @xcite noted that a score function designed against the worst - case attack achieves capacity against arbitrary attacks . this also lead to a proposal for a capacity - achieving score function in @xcite , which achieves the lower bound on the leading constant of @xmath37 . most of the work on fingerprinting focused on the setting of arbitrary , unknown attacks , but some work was also done on the informed setting , where the decoder knows or tries to estimate the pirate strategy @xcite . it is well known that for suboptimal pirate attacks the required code length may be significantly smaller than @xmath38 , but explicit schemes provably achieving an optimal scaling in @xmath9 are not known . research on the group testing problem started much longer ago , and already in 1985 exact asymptotics on the code length for probabilistic schemes were derived as @xmath41 @xcite , whereas deterministic schemes require a code length of @xmath35 @xcite . later work focused on slight variations of the classical model such as noisy group testing , where a positive result may not always correspond to the presence of a defective item due to ` noise ' in the test output @xcite . for noisy group testing exact asymptotics on the capacities ( with leading constants ) are yet unknown , and so it is not known whether existing constructions are optimal . in this paper we show how to build schemes using log - likelihood decoders that provably satisfy given bounds on the error probabilities , and have a code length with the optimal asymptotic scaling . we do this both for the informed setting ( where @xmath26 is known to the decoder ) and the universal setting , where @xmath26 is only known to satisfy the marking assumption . the results for the informed setting for fingerprinting are summarized in table [ tab : tab1 ] . simple informed and universal decoders are discussed in sections [ sec : dec - simple ] and [ sec : dec - simple - universal ] , and joint informed and universal decoders are discussed in sections [ sec : dec - joint ] and [ sec : dec - joint - universal ] respectively . in - between , section [ sec : dec - simple - universal - emi ] discusses a completely different approach to obtain a universal simple decoder ( based on the empirical mutual information decoder of moulin @xcite ) and shows how in the end the result is again quite similar . .an overview of the provable asymptotic code lengths of the informed fingerprinting decoders discussed in this paper . [ tab : tab1 ] [ cols="<,<,<",options="header " , ] in this section we will discuss simple decoders with explicit scheme parameters ( code lengths , accusation thresholds ) that provably satisfy given bounds on the error probabilities . the asymptotics of the resulting code lengths further show that these schemes are capacity - achieving ; asymptotically , the code lengths achieve the lower bounds that follow from the simple capacities , as derived in @xcite . we will follow the bias - based and score - based framework introduced by tardos @xcite , which was later generalized to joint decoders by moulin @xcite . for simple decoding , this means that a user @xmath11 receives a score @xmath42 of the form @xmath43 and he is accused iff @xmath44 for some fixed threshold @xmath45 . the function @xmath46 is sometimes called the score function . note that since @xmath46 only depends on @xmath8 through @xmath47 , any decoder that follows this framework is a simple decoder . several different score functions @xmath46 have been considered before @xcite , but in this work we will restrict our attention to log - likelihood scores , which are known to perform well and which turn out to be quite easy to analyze . first , when building a decoder we naturally want to be able to distinguish between two cases : user @xmath11 is guilty or user @xmath11 is not guilty . to do this , we assign scores to users based on the available data , and we try to obtain an optimal trade - off between the false positive error ( accusing an innocent user ) and the false negative error ( not accusing a guilty user ) . this problem is well known in statistics as a hypothesis testing problem , where in this case we want to distinguish between the following two hypotheses @xmath48 and @xmath49 : @xmath50 the neyman - pearson lemma @xcite tells us that the most powerful test to distinguish between @xmath48 and @xmath49 is to test whether the following likelihood ratio exceeds an appropriately chosen threshold @xmath45 : @xmath51 taking logarithms , and noting that different positions @xmath17 are i.i.d . , it is clear that testing whether a user s likelihood ratio exceeds @xmath52 is equivalent to testing whether his score @xmath42 exceeds @xmath53 for @xmath46 defined by @xmath54 thus , the score function @xmath46 from corresponds to using a neyman - pearson score over the entire code word @xmath47 , and therefore @xmath46 is in a sense optimal for minimizing the false positive error for a fixed false negative error . score functions of this form were previously considered in the context of fingerprinting in e.g. @xcite , but these papers did not show how to choose @xmath45 and @xmath9 to provably satisfy certain bounds on the error probabilities . let us first see how we can choose @xmath9 and @xmath45 such that we can prove that the false positive and false negative error probabilities are bounded from above by certain values @xmath55 and @xmath56 . for the analysis below , we will make use of the following function @xmath57 , which is closely related to the moment - generating function of scores in one position @xmath17 for both innocent and guilty users . for fixed @xmath58 , this function @xmath57 is defined on @xmath59 $ ] by @xmath60 and it satisfies @xmath61 . [ thm : dec - simple - informed ] let @xmath58 and @xmath26 be fixed and known to the decoder . let @xmath62 , and let the code length @xmath9 and threshold @xmath45 be chosen as @xmath63 then with probability at least @xmath64 no innocent users are accused , and with probability at least @xmath65 at least one colluder is caught . for innocent users @xmath11 , we would like to prove that @xmath66 , where @xmath42 is the user s total score over all segments . if this can be proved , then since innocent users have independent scores , it follows that with probability at least @xmath67 no innocent users are accused . to get somewhat tight bounds , we start by applying the markov inequality to @xmath68 for some @xmath69 : @xmath70 let us now first switch to guilty users . we will prove that for some guilty user @xmath11 , @xmath71 . again using markov s inequality with some fixed constant @xmath72 , we get @xmath73 for both guilty and innocent users , we can now obtain bounds by choosing appropriate values for @xmath74 and @xmath75 . investigating the resulting expressions , it seems that good choices for @xmath76 leading to sharp bounds are @xmath77 and @xmath78 . substituting these choices for @xmath74 and @xmath75 , and setting the bounds equal to the desired upper bounds @xmath79 and @xmath56 , we get @xmath80 combining these equations we obtain the given expression for @xmath45 , and solving for @xmath9 leads to the expression for @xmath9 in . compared to previous papers analyzing provable bounds on the error probabilities @xcite , the proof of theorem [ thm : dec - simple - informed ] is remarkably short and simple . note however that the proof that a colluder is caught assumes that the attack used by the colluders is the same as the one the decoder is built against , and that the actual value of @xmath9 is still somewhat mysterious due to the term @xmath81 . in section [ sec : dec - simple - codelength ] we will show how to get some insight into this expression for @xmath9 . before going into details how the code lengths of theorem [ thm : dec - simple - informed ] scale , note that theorem [ thm : dec - simple - informed ] only shows that with high probability we provably catch _ at least one _ colluder with this decoder . although this is commonly the best you can hope for when dealing with arbitrary attacks in fingerprinting , if the attack is colluder - symmetric it is actually possible to catch _ all _ colluders with high probability . so instead we would like to be able to claim that with high probability , the set of accused users @xmath82 _ equals _ the set of colluders @xmath2 . similar to the proof for innocent users , we could simply replace @xmath56 by @xmath83 , and argue that the probability of finding all pirates is the product of their individual probabilities of getting caught , leading to a lower bound on the success probability of @xmath84 . this leads to the following heuristic estimate for the code length required to catch _ all _ pirates . let @xmath85 in theorem [ thm : dec - simple - informed ] be replaced by @xmath86 . then with probability at least @xmath64 no innocent users are accused , and with probability at least @xmath65 _ all colluders are caught_. the problem with this claim is that the pirate scores are related through @xmath29 , so they are not independent . as a result , we can not simply take the product of the individual probabilities @xmath87 to get a lower bound on the success probability of @xmath65 . on the other hand , especially when the code length @xmath9 is large and @xmath56 is small , we do not expect the event @xmath88 to tell us much about the probability of e.g. @xmath89 occurring . one might thus expect that @xmath89 does not become much less likely when @xmath88 occurs . but since it is not so simple to prove a rigorous upper bound on the _ catch - all _ error probability without assuming independence , we leave this problem for future work . let us now see how the code lengths @xmath9 from scale in terms of @xmath0 and @xmath4 . in general this expression is not so pretty , but if we focus on the regime of large @xmath4 ( and fixed @xmath55 and @xmath56 ) , it turns out that the code length always has the optimal asymptotic scaling , regardless of @xmath58 and @xmath26 . [ thm : dec - simple - asymptotics ] for large @xmath4 and fixed @xmath55 and @xmath56 , the code length @xmath9 of theorem [ thm : dec - simple - informed ] scales as @xmath90 , \label{eq : dec - simple - asymptotics}\end{aligned}\ ] ] where @xmath91 is the mutual information between a pirate symbol @xmath92 and the pirate output @xmath93 . as a result , @xmath9 has the optimal asymptotic scaling . first , note that if @xmath94 and @xmath95 are fixed , then @xmath96 . let us first study the behavior of @xmath81 for small @xmath85 , by computing the first order taylor expansion of @xmath81 around @xmath97 . for convenience , below we will abbreviate @xmath98 by @xmath99 . @xmath100 here the second equality follows from the fact that if @xmath101 , then the factor @xmath102 in front of the exponentiation would already cause this term to be @xmath103 , while if @xmath104 , then also @xmath105 and thus their ratio is bounded and does not depend on @xmath85 . now , recognizing the remaining summation as the mutual information ( in natural units ) between a colluder symbol @xmath92 and the pirate output @xmath93 , we finally obtain : @xmath106 substituting this result in the original equation for @xmath9 , and noting that the factor @xmath55 inside the logarithm is negligible for large @xmath4 , we finally obtain the result of . note that in the discussion above , we did not make any assumptions on @xmath58 . in fact , both theorems [ thm : dec - simple - informed ] and [ thm : dec - simple - asymptotics ] hold for _ arbitrary _ values of @xmath58 ; the decoder always achieves the capacity associated to that value of @xmath58 . as a result , if we optimize and fix @xmath58 based on @xmath26 ( using results from ( * ? ? ? * section ii ) ) , we automatically end up with a decoder that provably achieves capacity for this attack . let us now consider specific pirate attacks which are often considered in the fingerprinting literature , and investigate the resulting code lengths . we will again consider the following five attacks of @xcite , which were also considered in e.g. @xcite . recall that @xmath107 is the probability of the pirates outputting a @xmath23 when they received @xmath24 ones . * * interleaving attack * : the coalition randomly selects a pirate and outputs his symbol . this corresponds to @xmath108 * * all-@xmath23 attack * : the pirates output a @xmath23 whenever they can , i.e. whenever they have at least one @xmath23 . this translates to @xmath109 * * majority voting * : the colluders output the most common received symbol . this corresponds to @xmath110 * * minority voting * : the colluders output the least common received symbol . this corresponds to @xmath111 * * coin - flip attack * : if the pirates receive both symbols , they flip a fair coin to decide which symbol to output : @xmath112 using theorem [ thm : dec - simple - informed ] we can now obtain exact , provable expressions for the code length @xmath9 in terms of @xmath113 . in general these expressions are quite ugly , but performing a taylor series expansion around @xmath114 for the optimal values of @xmath58 from ( * ? ? ? * section ii.a ) we obtain the following expressions for @xmath9 . note that @xmath115 . @xmath116 , \label{eq : dec - simple - int } \\ \ell({{\bm}{\theta}_{\text{all1 } } } ) & = \frac{c \ln\left(\frac{n}{{\varepsilon}_1}\right)}{\ln(2)^2 } \left[\frac{\sqrt{\gamma}\left(1 + \sqrt{\gamma } - \gamma\right)\ln 2}{1 - 2^{-\sqrt{\gamma } } } + o\left(\frac{1}{c}\right)\right ] , \\ \ell({{\bm}{\theta}_{\text{maj } } } ) & = \pi c \ln\left(\frac{n}{{\varepsilon}_1}\right ) \left[\frac{1 + \sqrt{\gamma } - \gamma}{1 - \sqrt{\gamma } } + o\left(\frac{1}{c}\right)\right ] , \\ \ell({{\bm}{\theta}_{\text{coin } } } ) & = \frac{4c}{\ln(2)^2 } \ln\left(\frac{n}{{\varepsilon}_1}\right ) \\ & \times \left[\frac{\sqrt{\gamma } \left(1 + \sqrt{\gamma } - \gamma\right ) \ln 2 } { 2 - \left(1+\frac{1}{\sqrt{2}}\right)^{\sqrt{\gamma } } - \left(1-\frac{1}{\sqrt{2}}\right)^{\sqrt{\gamma } } } + o\left(\frac{1}{c}\right)\right].\end{aligned}\ ] ] if we assume that both @xmath117 and @xmath96 , then we can further simplify the above expressions for the code lengths . the first terms between brackets all scale as @xmath118 , so the code lengths scale as the terms before the square brackets . these code lengths match the capacities of @xcite . as for the score functions @xmath46 , let us highlight one attack in particular , the interleaving attack . the all-@xmath23 decoder will be discussed in section [ sec : dec - simple - gt ] , while the score functions for other attacks can be computed in a similar fashion . for the interleaving attack , working out the probabilities in , we obtain the following score function : @xmath119 for group testing , we will consider three models : the classical ( noiseless ) model and the models with additive noise and dilution noise . other models where the probability of a positive test result only depends on the tally @xmath120 ( such as the threshold group testing models considered in @xcite ) may be analyzed in a similar fashion . note that the classical model is equivalent to the all-@xmath23 attack in fingerprinting , as was previously noted in e.g. @xcite . * * classical model * : the test output is positive iff the tested pool contains at least one defective : @xmath109 * * additive noise model * : just like the classical model , but if no defectives are tested the result may still be positive : @xmath121 * * dilution noise model * : similar to the classical model , but the probability of a positive result increases with @xmath24 : @xmath122 we can again expand the expressions of theorem [ thm : dec - simple - informed ] around @xmath114 for the optimal values of @xmath58 from ( * ? ? ? * section ii.b ) , but with the added parameter @xmath123 the resulting formulas are quite a mess . if we also let @xmath96 then we can use theorem [ thm : dec - simple - asymptotics ] to obtain the following simpler expressions : @xmath124 , \\ \ell({{\bm}{\theta}_{\text{add } } } ) & = \frac{c \ln n}{\ln(2)^2 - r \ln 2 + o(r^2 ) } \left[1 + o\left(\sqrt{\gamma } + \frac{1}{c}\right)\right ] , \\ \ell({{\bm}{\theta}_{\text{dil } } } ) & = \frac{c \ln n}{\ln(2)^2 - o(r \ln r ) } \left[1 + o\left(\sqrt{\gamma } + \frac{1}{c}\right)\right].\end{aligned}\ ] ] for more detailed expressions for @xmath9 , one may combine theorems [ thm : dec - simple - informed ] and [ thm : dec - simple - asymptotics ] with ( * ? ? ? * section ii.b ) . for the classical model , working out the details we obtain the following result . [ thm : dec - simple - cla ] for the classical group testing model , the simple decoder for the optimal value @xmath125 is given by by a factor @xmath126 , and so also @xmath45 should be scaled by a factor @xmath126 . ] @xmath127 using this decoder in combination with the parameters @xmath45 and @xmath9 of theorem [ thm : dec - simple - informed ] , we obtain a simple group testing algorithm with an optimal asymptotic number of group tests of @xmath128 this asymptotically improves upon results of e.g. chan et al . @xcite who proposed an algorithm with an asymptotic code length of @xmath129 . their algorithm does have a guarantee of never falsely identifying a non - defective item as defective ( whereas our proposed decoder does not have this guarantee ) , but the price they pay is a higher asymptotic number of tests to find the defectives . while in the previous section we discussed simple decoders for the setting where @xmath26 is completely known to the decoder , let us now consider the setting which is more common in fingerprinting , where the attack strategy is not assumed known to the decoder . this setting may partially apply to group testing as well ( where there may be some unpredictable noise on the test outputs ) but the main focus of this section is the uninformed fingerprinting game . let us now try to investigate how to build a decoder that works against arbitrary attacks in fingerprinting . to build such a decoder , meerwald and furon @xcite previously noted that abbe and zheng @xcite proved in a more general context that under certain conditions on the set of allowed pirate strategies @xmath18 , a decoder that works against the worst - case attack @xmath130 also works against arbitrary other attacks @xmath131 . in this case the set of allowed pirate strategies we consider is the set of all attacks satisfying the marking assumption @xmath132 : @xmath133^{c + 1 } \mid \theta_0 = 0 , \theta_c = 1\}.\end{aligned}\ ] ] for finite @xmath0 the worst - case attack in fingerprinting ( from an information - theoretic perspective ) has been studied in e.g. @xcite , but in general this attack is quite messy and unstructured . since this attack is not so easy to analyze , let us therefore focus on the asymptotics of large @xmath0 and @xmath4 . huang and moulin @xcite previously proved that for large coalitions , the optimal pirate attack is the interleaving attack . so combining this knowledge with the result of abbe and zheng , perhaps a good choice for a universal decoder is the interleaving decoder , which we recall is given by : @xmath134 let us take a closer look at this decoder . for fixed @xmath135 , if we look at values @xmath136 $ ] and focus on the regime of large @xmath0 , we can perform a taylor series expansion around @xmath114 to get @xmath137 . the resulting expressions then turn out to be closely related to oosterwijk et al.s @xcite decoder @xmath138 : @xmath139 this implies that @xmath46 and @xmath138 are asymptotically equivalent for @xmath58 sufficiently far away from @xmath103 and @xmath23 . since for oosterwijk et al.s score function one generally uses _ cut - offs _ on @xmath14 ( i.e. only using values @xmath136 $ ] for fixed @xmath135 ) to guarantee that @xmath140 ( cf . @xcite ) , and since the decoder of oosterwijk et al . is known to achieve capacity using these cut - offs , we immediately get the following result . [ prop : dec - simple - uninformed ] the score function @xmath46 of together with the bias density function ( encoder ) @xmath141 on @xmath142 $ ] of the form @xmath143 asymptotically achieve the simple capacity for the uninformed fingerprinting game when the same cut - offs @xmath144 as those of @xcite are used . so combining the log - likelihood decoder tuned against the asymptotic worst - case attack ( the interleaving attack ) with the arcsine distribution with cut - offs , we obtain a universal decoder that works against arbitrary attacks . although proposition [ prop : dec - simple - uninformed ] is already a nice result , the cut - offs @xmath144 have been a nagging inconvenience ever since tardos introduced them in 2003 @xcite . in previous settings it was well - known that this cut - off @xmath144 had to be large enough to guarantee that innocent users are not falsely accused , and small enough to guarantee that large coalitions can still be caught . for instance , when using tardos original score function , it was impossible to do without cut - offs , and the same seems to hold for oosterwijk et al.s decoder @xmath138 , since the scores blow up for @xmath145 . looking at the universal log - likelihood decoder , one thing to notice is that the logarithm has a kind of mitigating effect on the tails of the score distributions . for @xmath146 the resulting scores are roughly a factor @xmath0 smaller than those obtained with @xmath138 , but where the blow - up effect of @xmath138 for small @xmath58 is proportional to @xmath147 , the function @xmath46 only scales as @xmath148 in the region of small @xmath58 . this motivates the following claim , showing that with this decoder @xmath46 we finally do not need any cut - offs anymore ! [ thm : dec - simple - universal ] the decoder @xmath46 of and the encoder @xmath149 , defined on @xmath59 $ ] by @xmath150 together asymptotically achieve the simple capacity for the uninformed fingerprinting game . we will argue that using this new universal decoder @xmath46 , the difference in performance between using and not using cut - offs on @xmath14 is negligible for large @xmath0 . since the encoder with cut - offs asymptotically achieves capacity , it then follows that without cut - offs this scheme also achieves capacity . let us first prove that all moments of user scores are finite , even if no cut - offs are used . we will show that @xmath151 < \infty$ ] for any @xmath152 and @xmath153 , so that after taking weighted combinations we also get @xmath154 . let us consider the case where @xmath155 ; other cases can be analyzed in the same way . using the density function @xmath156 of , we have @xmath157 \\ & = \int_{0}^1 \frac{dp}{\pi \sqrt{p(1-p ) } } \log^k\left(1 + \frac{1 - p}{c p}\right).\end{aligned}\ ] ] splitting the interval @xmath59 $ ] into two parts @xmath158 $ ] and @xmath159 $ ] ( where @xmath144 depends on @xmath160 but not on @xmath0 ) we obtain @xmath161 let us denote the two terms by @xmath162 and @xmath163 . for the first term , we can perform a taylor series expansion to obtain : @xmath164 here @xmath165 follows from considering sufficiently large @xmath0 while @xmath144 remains fixed . ( note that for large @xmath0 we even have @xmath166 . ) for the other term we do not expand the logarithm : @xmath167 the last step @xmath168 follows from the fact that the integration is done over an interval of width @xmath144 , while the integrand scales as @xmath169 times some less important logarithmic terms . for arbitrary @xmath160 , we can thus let @xmath170 as a function of @xmath160 to see that this is always bounded . similar arguments can be used to show that for other values of @xmath171 we also have @xmath172 < 0 $ ] . as a result , all innocent and guilty user score moments are finite , and so for large @xmath0 from the central limit theorem it follows that the distributions of user scores will converge to gaussians . if the scores of innocent and guilty users are indeed gaussian for large @xmath0 , then as discussed in e.g. @xcite all that matters for assessing the performance of the scheme are the mean and variance of both curves . similar to , the effects of small cut - offs on the distribution function @xmath14 are negligible as both means and variances stay the same up to small order terms . so indeed , in both cases the ` performance indicator ' @xcite asymptotically stays the same , leading to equivalent code lengths . note that the same result does _ not _ apply to the score function @xmath138 of oosterwijk et al . @xcite , for which the effects of values @xmath145 are not negligible . the main difference is that for small @xmath58 , the score function @xmath138 scales as @xmath147 ( which explodes when @xmath58 is really small ) , while the log - likelihood decoder @xmath46 only scales as @xmath173 . figure [ fig : pdf ] illustrates the difference in the convergence of normalized innocent user scores to the standard normal distribution , when using the score functions @xmath46 and @xmath138 . these are experimental results for @xmath174 and @xmath175 based on @xmath176 simulated scores for each curve , and for both score functions we did not use any cut - offs . as we can see , using @xmath46 the normalized scores @xmath177 are close to gaussian , while using @xmath138 the curves especially do not look very gaussian for @xmath178 ; in most cases the distribution tails are much too large . for the minority voting attack the resulting curve does not even seem close to a standard normal gaussian distribution . with the above result in mind , let us now briefly discuss how to actually build a universal scheme with the interleaving decoder @xmath46 . from theorem [ thm : dec - simple - universal ] it is clear that for generating biases we should use the arcsine distribution @xmath156 , and our decoder will be the interleaving decoder @xmath46 of . what remains is figuring out how to choose @xmath9 and @xmath45 for arbitrary attacks . first , it is important to note that the expected innocent and guilty scores per segment ( @xmath179 and @xmath180 ) and the variance of the innocent and guilty scores ( @xmath181 and @xmath182 ) heavily depend on the collusion channel @xmath26 . this was not the case for tardos original decoder @xcite and the symmetrized decoder @xcite , for which @xmath45 could be fixed in advance regardless of the collusion strategy . this means that we will either have to delay fixing @xmath45 until the decoding stage , or scale / translate scores per segment accordingly at each position @xmath17 . for choosing the code length @xmath9 and threshold @xmath45 let us focus on the regime of reasonably large @xmath0 . in that case , as argued above the total innocent and guilty scores will behave like gaussians , with parameters @xmath183 and @xmath184 . to distinguish between these two distributions , using e.g. sanov s theorem the code rate @xmath185 should be proportional to the kullbeck - leibler divergence between the two distributions : @xmath186 a similar expression appears in @xcite , where it was noted that @xmath187 , so that the first term is the most important term . in @xcite the ratio @xmath188 was coined the ` performance indicator ' , and it was argued that this ratio should be maximized . in @xcite it was further shown that when using their decoder @xmath138 , this ratio is minimized by the pirates when they choose the interleaving attack @xmath189 . in other words : assuming scores are gaussian for large @xmath0 , the best attack the pirates can use is the interleaving attack when using @xmath138 as the decoder . since our new decoder @xmath46 is very similar to oosterwijk et al.s decoder @xmath138 ( by @xmath190 ) , a natural conjecture would be that also for this new score function , asymptotically the best pirate attack maximizing the decoder s error probabilities is the interleaving attack . experiments with @xmath46 and previous experiments of @xcite with @xmath138 indeed show that other pirate attacks ( such as those considered in section [ sec : dec - simple - fp ] ) generally perform worse than the interleaving attack . as a result , a natural choice for selecting @xmath9 would be to base @xmath9 on the code length needed to deal with the ( asymptotic ) worst - case attack for this decoder , which we conjecture is the interleaving attack . and for the interleaving attack we know how to choose @xmath9 by theorem [ thm : dec - simple - informed ] , equation [ eq : dec - simple - int ] and theorem [ thm : dec - simple - asymptotics ] : @xmath191 \label{eq : ell - int - uni2 } \\ & = 2 c^2 \ln n \left[1 + o\left(\sqrt{\gamma } + \frac{1}{c^2}\right)\right ] . \label{eq : ell - int - uni3}\end{aligned}\ ] ] these choices for @xmath9 thus seem reasonable estimates for the code lengths required to deal with arbitrary attacks . finally , for choosing @xmath45 , as argued before this parameter depends on the pirate strategy @xmath26 , which may lead to different scalings and translations of the curves of innocent and guilty user scores . what we could do is compute the parameters @xmath192 based on the pirate output @xmath193 and normalize the scores accordingly . this means that after computing user scores @xmath42 , we apply the following transformation : @xmath194 this guarantees that the scores of innocent users will roughly be distributed as @xmath195 , and for guilty users this results in a distribution of the form @xmath196 . to guarantee that the probability that an innocent user has a score below @xmath45 is at least @xmath197 , it then suffices to let @xmath198 where @xmath199 denotes the distribution function of the standard normal distribution @xmath195 this means that after transforming the scores , the threshold can be fixed independent of the pirate strategy . besides oosterwijk et al.s lagrangian approach and our neyman - pearson - based approach to obtaining efficient decoders , let us now mention a third way to obtain a similar capacity - achieving universal decoder . to construct this decoder , we start with the empirical mutual information decoder proposed by moulin @xcite , and for now let us assume @xmath200 is fixed is not fixed and is drawn from a continuous distribution function @xmath14 , the empirical probabilities considered below do not make much sense , as each value of @xmath201 only occurs once . in that case one could e.g. build a histogram for the values of @xmath58 , and compute empirical probabilities for each bin , or discretize the distribution function for @xmath58 @xcite . ] . with this decoder , a user is assigned a score of the form @xmath202 and again the decision to accuse depends on whether this score exceeds some fixed threshold @xmath45 . here @xmath203 is the empirical estimate of the actual probability @xmath204 , i.e. , @xmath205 . writing out the empirical probability outside the logarithm , and replacing the summation over @xmath171 by a summation over the positions @xmath17 , this is equivalent to @xmath206 now , this almost fits the score - based simple decoder framework , except for that the terms inside the logarithm are not independent for different positions @xmath17 . to overcome this problem , we could try to replace the empirical probabilities @xmath203 by the actual probabilities @xmath204 , but to compute @xmath207 we need to know whether user @xmath11 is guilty or not . solving this final problem using bayesian inference , we get the following result . approximating the empirical probabilities in the empirical mutual information decoder with actual probabilities using bayesian inference ( with an a priori probability of guilt of @xmath208 ) , this decoder corresponds to using the following score function @xmath209 : @xmath210\right ) . \label{eq : dec - simple - moulin1}\end{aligned}\ ] ] the value of @xmath211 can be computed without any problems , so let us focus on the term @xmath212 . using bayesian inference , we have : @xmath213 assuming an a priori probability of guilt of @xmath208 and dividing by @xmath214 we get @xmath215 taking logarithms , this leads to the expression of . although this score function looks very similar to the log - likelihood decoder , there are some essential differences . for instance , for the all-@xmath23 attack we have @xmath216 while @xmath217 . for the interleaving attack , for which we may again hope to obtain a universal decoder using this approach , we do get a familiar result . [ thm : dec - simple - moulin ] for the interleaving attack , the bayesian approximation of the empirical mutual information decoder of satisfies @xmath218 for values of @xmath136 $ ] with @xmath135 and large @xmath0 , this decoder is again equivalent to both the log - likelihood score function @xmath46 and oosterwijk et al.s score function @xmath138 : @xmath219 for @xmath145 the logarithm again guarantees that scores do not blow up so much , but due to the factor @xmath4 in the denominator ( rather than a factor @xmath0 , as in @xmath46 ) the scores relatively increase more when @xmath58 approaches @xmath103 than for the score function @xmath46 . in this section we will discuss informed joint decoders which we conjecture are able to find pirates with shorter code lengths than simple decoders . the asymptotics of the resulting code lengths further motivate that these schemes may be optimal , but some open problems remain for proving that they are indeed optimal . following the score - based framework for joint decoders of moulin @xcite , we assign tuples @xmath220 of size @xmath0 a score of the form @xmath221 note that if @xmath26 is colluder - symmetric , then this is equivalent to @xmath222 where @xmath223 is the tally of the number of ones received by the tuple @xmath220 in position @xmath17 . for the accusation phase , we now accuse all users in @xmath220 iff @xmath224 for some fixed threshold @xmath45 . note that this accusation algorithm is not exactly well - defined , since it is possible that a user appears both in a tuple that is accused and in a tuple that is not accused . for the analysis we will assume that the scheme is only successful if the single tuple consisting of all colluders has a score exceeding @xmath45 and no other tuples have a score exceeding @xmath45 , in which case all users in that guilty tuple are accused . this may be too pessimistic for evaluating the performance of the scheme . for building a joint decoder we would like to be able to distinguish between the all - guilty tuple and other tuples , so a natural generalization of the hypotheses @xmath48 and @xmath49 for simple decoding would be to let @xmath225 and @xmath226 . however , with this choice of @xmath49 , computing probabilities @xmath227 is complicated : the event @xmath49 does not completely determine @xmath228 , since that depends on exactly how many colluders are present in @xmath220 . to be able to compute the likelihood ratios , we therefore use the following two hypotheses , which were also used in e.g. @xcite : @xmath229 we again use the corresponding log - likelihood ratio per position as our score function @xmath46 , which is again the logarithm of the likelihood ratio over all positions @xmath17 : @xmath230 using this joint score function @xmath46 corresponds to a most powerful test according to the neyman - pearson lemma @xcite , so @xmath46 is in a sense optimal for distinguishing between @xmath48 and @xmath49 . joint decoders of this form were previously considered in e.g. @xcite . let us again study how to choose @xmath9 and @xmath45 such that we can prove that the false positive and false negative error probabilities are bounded from above by certain values @xmath55 and @xmath56 . below we will again make use of the function @xmath57 of where the simple hypotheses have been replaced by our new joint hypotheses @xmath48 and @xmath49 . [ thm : dec - joint - informed ] let @xmath58 and @xmath26 be fixed and known to the decoder . let @xmath231 , and let the code length @xmath9 and the threshold @xmath45 be defined as @xmath232 then with probability at least @xmath64 all all - innocent tuples are not accused , and with probability at least @xmath65 the single all - guilty tuple is accused . the proof is very similar to the proof of theorem [ thm : dec - simple - informed ] . instead of @xmath4 innocent and @xmath0 guilty users we now have @xmath233 all - innocent tuples and just @xmath23 all - guilty tuple , which changes some of the numbers in @xmath85 , @xmath45 and @xmath9 above . we then again apply the markov inequality with @xmath77 and @xmath78 to obtain the given expressions for @xmath45 and @xmath9 . for deterministic strategies @xmath234 , choosing the scheme parameters is much simpler . similar to ( * ? ? ? * lemma 1 ) , where it was shown that for deterministic attacks the capacity is exactly @xmath235 , in this case we get a code length of roughly @xmath236 . [ thm : dec - joint - deterministic ] let @xmath26 be a deterministic attack , and let @xmath58 be chosen such that @xmath237 . let @xmath9 and @xmath45 be chosen as : @xmath238 then with probability @xmath64 all all - innocent tuples will not be accused , and the single all - guilty tuple will _ always _ be accused . for deterministic attacks , we have @xmath239 as a result , the score function @xmath46 satisfies @xmath240 with the capacity - achieving choice of @xmath58 of ( * ? ? ? * lemma 1 ) , we have @xmath241 for @xmath242 leading to a score of @xmath243 for a match , and @xmath244 for cases where @xmath245 does not match the output that follows from @xmath26 and the assumption that @xmath220 is the all - guilty tuple . for @xmath246 , clearly we will always have a match , so this tuple s score will always be @xmath247 , showing that this tuple is always accused . for innocent tuples , since @xmath248 it follows that in each position @xmath17 , with probability @xmath249 this tuple s score will not be @xmath244 . so with probability @xmath250 , the tuple s score after @xmath9 segments will not be @xmath244 , in which case it equals @xmath247 . to make sure that this probability is at most @xmath251 so that the total error probability is at most @xmath55 , we set @xmath252 , leading to the resulting expression for @xmath9 . note that for deterministic attacks , any choice of @xmath253 works just as well as choosing @xmath45 ; after @xmath9 segments all tuples will either have a score of @xmath244 or @xmath45 . theorem [ thm : dec - joint - informed ] does not prove that we can actually find the set of colluders with high probability , since mixed tuples ( consisting of some innocent and some guilty users ) also exist and these may or may not have a score exceeding @xmath45 . theorem [ thm : dec - joint - informed ] only proves that with high probability we can find a set @xmath82 of @xmath0 users which contains at least one colluder . basic experiments show that in many cases the only tuple with a score exceeding @xmath45 ( and thus the tuple with the highest score ) is the all - guilty tuple , and so all mixed tuples have a score below @xmath45 . proving that mixed tuples indeed get a score below @xmath45 is left as an open problem . to further motivate why using this joint decoder may be the right choice , the following proposition shows that at least the resulting code lengths are optimal . the proof is analogous to the proof of theorem [ thm : dec - simple - asymptotics ] . [ thm : dec - joint - asymptotics ] if @xmath254 then the code length @xmath9 of theorem [ thm : dec - joint - informed ] scales as @xmath255,\end{aligned}\ ] ] thus asymptotically achieving an optimal scaling of the code length for arbitrary values of @xmath58 . since the asymptotic code length is optimal regardless of @xmath58 , these asymptotics are also optimal when @xmath58 is optimized to maximize the mutual information in the fully informed setting . finally , although it is hard to estimate the scores of mixed tuples with this decoder , just like in @xcite we expect that the joint decoder score for a tuple is roughly equal to the sum of the @xmath0 individual simple decoder scores . so a tuple of @xmath0 users consisting of @xmath160 colluders and @xmath256 innocent users is expected to have a score roughly a factor @xmath257 smaller than the expected score for the all - guilty tuple . so after computing the scores for all tuples of size @xmath0 , we can get rough estimates of how many guilty users are contained in each tuple , and for instance try to find the set @xmath82 of @xmath0 users that best matches these estimates . there are several options for post - processing that may improve the accuracy of using this joint decoder , which are left for future work . using theorem [ thm : dec - joint - informed ] we can obtain exact expressions for @xmath9 in terms of @xmath113 . for the optimal values of @xmath58 of ( * ? ? ? * section iii.a ) we can use theorem [ thm : dec - joint - asymptotics ] to obtain the following expressions . note again that @xmath115 . @xmath258 , \label{eq : dec - joint - int } \\ \ell({{\bm}{\theta}_{\text{all1 } } } ) & = c \log_2 n \left[1 + o\left(\sqrt{\gamma } + \frac{1}{c}\right)\right ] , \\ \ell({{\bm}{\theta}_{\text{maj } } } ) & = c \log_2 n \left[1 + o\left(\sqrt{\gamma } + \frac{1}{c}\right)\right ] , \\ \ell({{\bm}{\theta}_{\text{coin } } } ) & = c \log_{5/4 } n \left[1 + o\left(\sqrt{\gamma } + \frac{1}{c}\right)\right].\end{aligned}\ ] ] let us again highlight one resulting decoder in particular ; the one for the interleaving attack . in general we can rewrite the ratio inside the logarithm as follows : @xmath259 for the interleaving attack we further have @xmath260 and @xmath261 . this leads to the following joint decoder @xmath46 . this means that the joint scores are purely based on the similarities between the tuple tally @xmath24 and the expected tuple tally @xmath263 for each position @xmath17 . if a tuple s tally @xmath24 is larger than the expected tally @xmath263 , then the score is positive if @xmath264 and negative otherwise , while if @xmath24 is smaller than @xmath263 , then the score is positive if @xmath265 and negative otherwise . for innocent tuples this leads to an expected score of roughly @xmath103 , while for the guilty tuple this leads to a high ( positive ) expected score . similar to the above , for group testing models we can also use theorem [ thm : dec - joint - informed ] to obtain exact expressions for @xmath9 in terms of @xmath113 with provable error bounds . for the optimal values of @xmath58 of ( * section iii.b ) we can use theorem [ thm : dec - joint - asymptotics ] to obtain the following refined expressions . @xmath266 , \\ \ell({{\bm}{\theta}_{\text{add } } } ) & = \frac{c \log_2 n}{1 - \frac{1}{2 } h(r ) + o(r^2 ) } \left[1 + o\left(\sqrt{\gamma } + \frac{1}{c}\right)\right ] , \\ \ell({{\bm}{\theta}_{\text{dil } } } ) & = \frac{c \log_2 n}{1 - \frac{\ln 2}{2 } h(r ) + o(r^2 ) } \left[1 + o\left(\sqrt{\gamma } + \frac{1}{c}\right)\right].\end{aligned}\ ] ] note that as discussed in theorem [ thm : dec - joint - deterministic ] the score function for the classical model is equivalent to simply checking whether some subset of @xmath0 items matches the test results , i.e. whether these would indeed have been the test results , had this subset been the set of defectives . with high probability , only the correct set of defectives passes this test . let us now again consider the more common setting in fingerprinting where the attack strategy @xmath26 is assumed unknown to the distributor . with the results for simple decoding in mind , and knowing that the interleaving attack is also the asymptotically optimal pirate attack in the joint fingerprinting game , we again turn our attention to the decoder designed against the interleaving attack . similar to the setting of simple decoding , explicitly proving that this decoder achieves the uninformed capacity is not so easy . however , through a series of reductions we can prove that this decoder is asymptotically capacity - achieving , for certain parameters @xmath9 and @xmath45 . [ thm : dec - joint - universal ] the joint log - likelihood decoder designed against the interleaving attack , using the score function @xmath46 defined by @xmath267 and the arcsine distribution encoder @xmath156 of together asymptotically achieve the joint capacity for the uninformed fingerprinting game . first , the simple uninformed capacity is asymptotically equivalent to the joint uninformed capacity , which follows from results of huang and moulin @xcite and oosterwijk et al . @xcite ( and section [ sec : dec - simple ] ) . this means that the simple universal decoder of theorem [ thm : dec - simple - universal ] already asymptotically achieves the joint capacity . we will prove that asymptotically , the proposed universal joint decoder is equivalent to the universal simple decoder of section [ sec : dec - simple - universal ] , thus also achieving the joint uninformed capacity . suppose we have a tuple @xmath220 of size @xmath0 , and suppose in some segment @xmath17 there are @xmath24 users who received a @xmath23 and @xmath25 users who received a @xmath103 . for now also assume that @xmath136 $ ] for some @xmath135 that does not depend on @xmath0 . in case @xmath265 , the combined simple decoder score of this tuple @xmath220 ( using the simple universal decoder @xmath46 of section [ sec : dec - simple - universal ] ) would be : @xmath268 on the other hand , if we look at this tuple s joint score with the joint universal decoder @xmath46 of , we have @xmath269 note that the last step follows from the fact that for large @xmath0 , with overwhelming probability we have @xmath270 ( since @xmath120 is binomially distributed with mean @xmath263 and variance @xmath271 ) , in which case @xmath272 . combining the above , we have that @xmath273 . so the joint universal decoder score for a tuple @xmath220 is asymptotically equivalent to the sum of the simple universal decoder scores for the members in this tuple , if @xmath136 $ ] . since as argued before the distribution tails @xmath274 $ ] and @xmath275 $ ] are negligible for the performance of the scheme for sufficiently small @xmath144 , and since the same result holds for @xmath264 , the simple and joint decoders are asymptotically equivalent . note that for the uninformed setting , the simple and joint capacities are asymptotically equivalent , which allowed us to prove the result . for finite @xmath0 the joint capacity may be slightly higher than the simple capacity , but the fact that they are asymptotically the same does show that there is not as much to gain with joint decoders as there is with e.g. joint group testing decoders , where the joint capacity is asymptotically a factor @xmath276 higher than the simple capacity . and since assigning joint scores to all tuples of size @xmath0 for each position @xmath17 is computationally very involved ( and since the resulting joint scores are very similar to the sum of simple universal scores anyway ) a more practical choice seems to be to use the simple universal decoder of section [ sec : dec - simple - universal ] instead . let us now briefly discuss the main results in this paper and their consequences . for the setting of simple decoders with known collusion channels @xmath26 , we have shown how to choose the score functions @xmath46 in the score - based framework , as well as how to choose the threshold @xmath45 and code length @xmath9 to guarantee that certain bounds on the error probabilities are met . with log - likelihood decoders , we showed that these decoders achieve capacity _ regardless of @xmath58 and regardless of @xmath26_. this means that no matter which pirate strategy @xmath26 and bias @xmath58 you plug in , this construction will always achieve the capacity corresponding to those choices for @xmath26 and @xmath58 . a trivial consequence is that for the optimal values of @xmath58 derived in @xcite , one also achieves the optimal code length for arbitrary @xmath58 . since in fingerprinting it is usually more common to assume that @xmath26 is unknown , we then turned our attention to the uninformed setting . we showed that the decoder designed against the interleaving attack , with the score function @xmath46 given by @xmath277 is actually a universal decoder , and achieves the uninformed fingerprinting capacity . we also showed how the proposed universal decoder is very similar to both oosterwijk et al.s decoder @xmath138 and an approximation of moulin s empirical mutual information decoder @xmath209 for @xmath146 by @xmath278 and we highlighted the differences between these decoders for @xmath145 . we argued that the proposed decoder @xmath46 is the most natural choice for a universal decoder ( motivated from the neyman - pearson lemma ) , and that it has some practical advantages compared to oosterwijk et al.s decoder @xmath138 , such as finally being able to get rid of the cut - offs @xmath144 on the density function @xmath14 . in sections [ sec : dec - joint ] and [ sec : dec - joint - universal ] we then turned our attention to joint decoders , which have the potential to significantly decrease the required code length @xmath9 at the cost of a higher decoding complexity . we considered a natural generalization of the simple decoders to joint decoders , and argued that our choice for the joint score functions @xmath46 seems to be optimal . there are still some gaps to fill here , since we were not able to prove how scores of mixed tuples ( tuples containing some innocent and some guilty users ) behave , and whether their scores also stay below @xmath45 with high probability . on the other hand , for deterministic attacks it is quite easy to analyze the behavior of these decoders , and for arbitrary attacks we did show that the code lengths have the optimal asymptotic scaling . finally , for the uninformed setting with joint decoders , we proved that the joint interleaving decoder achieves the joint uninformed capacity . since the joint uninformed capacity is asymptotically the same as the simple uninformed capacity , and since joint decoding generally has a much higher computational complexity than simple decoding , this decoder may not be as practical as the proposed simple universal decoder . finally , let us finish by mentioning some open problems which are left for future work . while in section [ sec : dec - simple - universal ] we showed that the new simple universal decoder achieves capacity and how one should roughly choose the code length @xmath9 and threshold @xmath45 , we did not provide any provable bounds on the error probabilities for the uninformed setting . for earlier versions of tardos scheme various papers analyzed such provable bounds @xcite and a similar analysis could be done for the log - likelihood decoder designed against the interleaving attack . perhaps such an analysis may once and for all establish the best way to choose the scheme parameters in universal fingerprinting . although this paper focused on applications to the ` static ' fingerprinting game , in some settings the feedback @xmath93 may be obtained in real - time . for instance , in pay - tv pirates may try to duplicate a fingerprinted broadcast in real time , while in group testing it may sometimes be possible to do several rounds of group testing sequentially . the construction of @xcite can trivially be applied to the decoders in this paper as well to build efficient dynamic fingerprinting schemes with the same asymptotics for the code lengths , but where ( i ) the order terms in @xmath9 are significantly smaller ; ( ii ) one can provably catch _ all _ pirates regardless of the ( asymmetric ) pirate strategy ; and ( iii ) one does not necessarily need to know ( a good estimate of ) @xmath0 in advance ( * ? ? ? * section v ) . an important open problem remains to determine the dynamic uninformed fingerprinting capacity , which may prove or disprove that the construction of @xcite combined with the universal decoder @xmath46 of section [ sec : dec - simple - universal ] is asymptotically optimal . in this work we focused on the binary case of @xmath10 different symbols for generating the code @xmath8 , but in ( universal ) fingerprinting it may be advantageous to work with larger alphabet sizes @xmath279 , since the code length decreases linearly with @xmath280 @xcite . this generalization to @xmath280-ary alphabets was considered in e.g. @xcite . for the results in this paper we did not really use that we were working with a binary alphabet , so it seems a straightforward exercise to prove that the @xmath280-ary versions of the log - likelihood decoders also achieve the @xmath280-ary capacities . a harder problem seems to be to actually compute these capacities in the various informed settings , since the maximization problem involved in computing these capacities then transforms from a one - dimensional optimization problem to a @xmath281-dimensional optimization problem . the author is grateful to pierre moulin for his comments during the author s visit to urbana - champaign that lead to the study of the empirical mutual information decoder of theorem [ thm : dec - simple - moulin ] and that eventually inspired work on this paper . the author further thanks jeroen doumen , teddy furon , jan - jaap oosterwijk , boris kori , and benne de weger for valuable discussions and comments regarding earlier versions of this manuscript . chan , p. h. che , s. jaggi , and v. saligrama , non - adaptive probabilistic group testing with noisy measurements : near - optimal bounds with efficient algorithms , " _ 49th allerton conference on communication , control , and computing _ , pp . 18321839 , 2011 . chan , s. jaggi , v. saligrama , and s. agnihotri , non - adaptive group testing : explicit bounds and novel algorithms , " _ ieee international symposium on information theory ( isit ) _ , pp . 18371841 , 2012 . m. cheraghchi , a. hormati , a. karbasi , and m. vetterli , group testing with probabilistic tests : theory , design and application , " _ ieee transactions on information theory _ , vol . 57 , no . 10 , pp . 70577067 , 2011 . y .- w . huang and p. moulin , on the saddle - point solution and the large - coalition asymptotics of fingerprinting games , " _ ieee transactions on information forensics and security _ , vol . 7 , no . 1 , pp . 160175 , 2012 . t. laarhoven , capacities and capacity - achieving decoders for various fingerprinting games , " _ acm workshop on information hiding and multimedia security ( ih&mmsec ) _ , 2014 . this is a preliminary version of the present paper . j. neyman and e. s. pearson , on the problem of the most efficient tests of statistical hypotheses , " _ philosophical transactions of the royal society a : mathematical , physical and engineering sciences _ , vol . 231 , pp . 289337 , 1933 . k. nuida , s. fujitsu , m. hagiwara , t. kitagawa , h. watanabe , k. ogawa , and h. imai , an improvement of discrete tardos fingerprinting codes , " _ designs , codes and cryptography _ , vol . 52 , no . 3 , pp . 339362 , 2009 . d. sejdinovic and o. johnson , note on noisy group testing : asymptotic bounds and belief propagation reconstruction , " _ 48th allerton conference on communication , control , and computing ( allerton ) _ , pp . 9981003 , 2010 . d. r. stinson , t. van trung , and r. wei , secure frameproof codes , key distribution patterns , group testing algorithms and related structures , " _ journal of statistical planning and inference _ , vol . 86 , no . 2 , pp . 595617 , 2000 .

the present panorama of hpc architectures is extremely heterogeneous , ranging from traditional multi - core cpu processors , supporting a wide class of applications but delivering moderate computing performance , to many - core gpus , exploiting aggressive data - parallelism and delivering higher performances for streaming computing applications . in this scenario , code portability ( and performance portability ) become necessary for easy maintainability of applications ; this is very relevant in scientific computing where code changes are very frequent , making it tedious and prone to error to keep different code versions aligned . in this work we present the design and optimization of a state - of - the - art production - level lqcd monte carlo application , using the directive - based openacc programming model . openacc abstracts parallel programming to a descriptive level , relieving programmers from specifying how codes should be mapped onto the target architecture . we describe the implementation of a code fully written in openacc , and show that we are able to target several different architectures , including state - of - the - art traditional cpus and gpus , with the same code . we also measure performance , evaluating the computing efficiency of our openacc code on several architectures , comparing with gpu - specific implementations and showing that a good level of performance - portability can be reached .

1701.00426

the use of processors based on multi- and many - core architectures is a common option in high performance computing ( hpc ) . several variants of these processors exist , differing mainly in the number and architecture of the cores integrated in a single silicon die . conventional cpus integrate tens of fat cores sharing a large on - chip cache . fat cores include several levels of caches and complex control structures , able to perform hardware optimization techniques ( branch - speculation , instruction scheduling , register renaming , etc ) . vector instructions are also supported by these cores , with a moderate level of data parallelism : 2 to 4 vector elements are processed by one vector instruction . this architecture is reasonably efficient for many type of regular and non - regular applications and delivers a level of performance of the order of hundreds of gigaflops per processor . on the other side of the spectrum we have graphics processor units ( gpu ) , available as accelerator boards attached to conventional cpus . gpus integrate thousands of slim cores able to efficiently support regular streams of computation , and deliver performances of the order of several teraflops . gpus are extremely aggressive in terms of data - parallelism , implementing vector units with large vector sizes ( 16 and 32 words are presently available options ) . midway between these two architectures , we have the intel _ many integrated cores _ ( mic ) architecture based on several tens of slim cores . in this case , cores are similar to their fat counterparts , but their design has been simplified removing many hardware control structures ( instruction scheduler , register renaming , etc ) and adopting wider vector units , able to process up to 4 or 8 vector elements in parallel . large scale computing centers today have not reached a common consensus on the `` best '' processor option for hpc systems , also because system choices are driven not only by application performances , but also by cost of ownership and energy aspects which are becoming increasingly critical parameters@xcite . several computing centers do adopt machines based on gpus , but other ones prefer to stay on more traditional cpus , offering a lower peak performance , but better computing efficiency for a wider range of applications . in this scenario , the development of applications would greatly benefit from the availability of a unique code version , written in an appropriate programming framework , able to offer portability , in terms of code and performance , across several present and possibly future state - of - the - art processor architectures . a single code version , portable across several architectures , is of great convenience in particular for scientific applications , where code changes and development iterations are very frequent , so keeping several architecture - specific code versions up - to - date is a tedious and error prone effort@xcite . directives based programming models are going exactly in this direction , abstracting parallel programming to a descriptive level as opposite to a prescriptive level , where programmers must specify how the code should be mapped onto the target machine . openmp@xcite and openacc@xcite are among the most common such programming models , already used by a wide scientific community . both are based on directives : openmp was introduced to manage parallelism on traditional multi - core cpus , while openacc is mainly used to target gpus ( although designed to be architecture agnostic)@xcite . these two frameworks are in fact converging and extending their scope to cover a large subset of hpc applications and architectures : openmp version 4 has been designed to support also accelerators , while compilers supporting openacc ( such as the pgi@xcite ) are starting to use directives also to target multi - core cpus . in this work we describe the implementation of a lattice qcd ( lqcd ) monte carlo code designed to be portable and efficient across several architectures . lqcd simulations represent a typical and well known hpc grand challenge , with physics results strongly limited by available computational resources@xcite ; over the years , several generations of parallel machines , optimized for lqcd , have been developed@xcite , while the development of lqcd codes running on many core architectures , in particular gpus , has seen large efforts in the last 10 years@xcite . our goal is to have just one code able to run on several processors without any major code changes , and possibly to have roughly the same level of efficiency , looking for an acceptable trade - off between portability and efficiency@xcite . as a programming model we have selected openacc , as it currently has a wider compiler support , in particular targeting nvidia gpus , which are widely used in hpc clusters and commonly used for scientific computations . openacc has been successfully used to port and run other scientific codes , such as lattice boltzmann applications@xcite in computational fluid - dynamics , showing a good level of code and performance portability on several architectures . the migration of our code to openmp4 , if needed , as soon as compiler support becomes more mature , is expected to be a simple additional effort . we have developed a code with all key features for a state - of - the - art simulations of qcd with dynamical fermions . using this code as a user test case , we assess : i ) if it is possible to write the code in such a way that the most computationally critical kernels can be executed on accelerators , as in previous cuda implementations@xcite ; ii ) how many of the presently available multi and many - core architectures can be really used ; iii ) how efficient are these codes , and in particular what is the price to pay in terms of performance with respect to a code written and optimized for a specific architecture ( e.g. , using cuda for gpus ) . we believe that our work is a non trivial step forward in the development of a fully portable production - grade lqcd monte carlo code , using the openacc programming model . an earlier paper@xcite presented tests of selected portions of an openacc lqcd implementation on fermi and k20 nvidia gpus , comparing performances with an openmp implementation for cpus . similarly , in a preliminary study@xcite , we compared the performance of selected kernels of a full simulation , written in openacc , with an equivalent cuda implementation , on a k20 nvidia gpu . in this work , we extend the use of openacc in several new directions : i ) we show the portability of a complete implementation across several architectures ; ii ) we show performance figures for the same openacc code on a variety of multi and many - core processors , including the most recent gpus like the k80 and the recently released p100 ; iii ) we compare results with a previous implementation of the same full application written in cuda@xcite . the remainder of the paper is organized as follows : in section [ simalg ] we give a brief introduction to lqcd and to the main computational aspects of our application ; in section [ hpctrend ] we highlight recent developments in hpc hardware and programming tools ; in section [ implementation ] we describe the openacc implementation of our code ; in section [ results ] we analyze our results ; finally , section [ conclusions ] , contains our concluding remarks . quantum chromodynamics ( qcd ) is the quantum field theory that describes strong interactions in the standard model of particle physics . it is a non - abelian gauge theory , based on the @xmath0 group ( the `` color '' group ) , describing the interactions of six different species ( `` flavors '' ) of quarks , mediated by @xmath1 vector bosons , the `` gluons '' . in principle qcd is not different from the theory that describes other sectors of the standard model ( i.e. the electroweak interaction ) ; however , strong interactions are indeed strong , i.e. the coupling constant of qcd is generically not small . asymptotic freedom ensures that the coupling constant gets smaller and smaller as the energy scale increases ( a summary of experimental results is available in 9.4 of the particle data group review@xcite ) , but a wealth of interesting phenomena take place for energies well below the perturbative regime ; a systematically improvable computational scheme , that does not rely on the smallness of the coupling constant , is needed to study this phenomenology from first principles . lattice qcd provides such a scheme . lqcd uses the feynman path - integral quantization and approximates the infinite dimensional path - integral by a finite dimensional integral : continuous space - time is replaced by a finite lattice of sizes @xmath2 , @xmath3 , @xmath4 , @xmath5 and lattice spacing @xmath6 . in order to maintain gauge invariance , the variables @xmath7 associated with the gauge fields are elements of the @xmath0 group and live on the links of the lattice ; the quark fields @xmath8 live on the lattice sites and transform under the gauge group as @xmath9dimensional complex vectors@xcite . the fundamental problem of lqcd is the evaluation of expectation values of given functions of the fields , @xmath10 $ ] , that is integrals of the form @xmath11\det(m[u])e^{-s_g[u]}\ , \quad z = \int \mathscr{d}u \det(m[u])e^{-s_g[u]}\ ; \end{aligned}\ ] ] the exponent @xmath12 is the discretization of the action of the gauge fields ( usually written as a sum of traces of products of @xmath7 along closed loops ) and @xmath13 describes the gluon - quark interaction . here , @xmath14 $ ] is a large and sparse structured matrix ( i.e. containing both space - time and color indexes ) which is the discretization of the continuum fermion operator @xmath15 where @xmath16 is the fermion mass , multiplying the identity operator , and @xmath17 is the dirac operator , which is constructed in terms of covariant derivatives . the integral in @xmath18 extends over all the @xmath7 variables on the lattice using the haar measure of @xmath0 . ( [ eq : pathint ] ) refers to a single quark species ( flavor ) ; in the realistic case of multiple flavors , @xmath19 , @xmath20 , @xmath21 , @xmath22 , @xmath23 and ordered by increasing quark mass . in a realistic simulation , one usually takes into account the first 3 ( or 4 , at most ) flavors , since the heaviest species give a negligible contribution to the low - energy dynamics of the theory . ] , one has to introduce a separate determinant for each flavor . this formulation makes contact with a standard problem in statistical mechanics : importance sampling of the distribution @xmath24)e^{-s_g[u]}$ ] . what is non - standard is the form of this distribution and in particular the presence of the determinant . the best strategy devised so far to cope with this problem is to introduce the so called pseudofermion fields@xcite @xmath25 and rewrite the integral as follows : @xmath26\det(m[u ] ) e^{-s_g[u]}\propto \int \mathscr{d}u \mathscr{d}\phi \ , o[u]\exp\left(-s_g[u]-\phi^{\dag } m[u]^{-1}\phi\right)\ ; \end{aligned}\ ] ] the action is still a non - local function of the field variables , but the computational burden required for the solution of a large sparse linear system is much lower than the one needed for the computation of its determinant . the explicit form of @xmath27 $ ] and @xmath14 $ ] is not fully determined , as these functions only have the constraint to go over to the correct continuum limit as the lattice spacing goes to zero . much in the same way as several discretization schemes exist for the numerical solution of a partial differential equation , several discretization schemes of the qcd action exist . in this paper we consider a specific state - of - the - art discretization , the tree - level symanzik improved action@xcite for the gauge part and the stout - improved@xcite `` staggered '' action for the fermion part . staggered actions have a residual degeneracy , that has to be removed by taking the @xmath28th root of the determinant . so , becomes in the staggered case @xmath29\exp\big(-s_g[u]-\phi^{\dag } m[u]^{-1/4}\phi\big)\ .\end{aligned}\ ] ] the physical system that one would like to simulate by the lattice box has a characteristic physical length @xmath30 , which is of the order of @xmath31 m. in order to reduce systematic effects related to discretization and to the finite box size , one would like that , at the same time , the lattice spacing @xmath6 be much smaller , and the box size @xmath32 much larger than @xmath30 , i.e. @xmath33 . making the reasonable approximation that @xmath34 translates into one order of magnitude means that the number of sites in each direction should be @xmath35 ; the corresponding fermion matrix , considering also internal ( e.g. , color ) indexes , has a dimension slightly exceeding @xmath36 ; note that it is a sparse matrix , since the discretization of the dirac operator @xmath17 connects only neighbor lattice sites . in finite temperature simulations the size of the lattice is typically smaller , since in that case the temporal direction is shortened and equal to the inverse of the temperature , @xmath37 . the most computationally demanding task in the typical lqcd algorithm is the solution of a linear system involving the fermion matrix @xmath38 . the numerical difficulty of this problem is fixed by the condition number of @xmath38 , hence , since the highest eigenvalue is typically @xmath39 , by the smallest eigenvalue of @xmath38 . here the physical properties of qcd play a significant role : the eigenvalues of the dirac operator are dense around zero , a property related to the so - called _ spontaneous breaking of chiral symmetry _ , so the smallest eigenvalue is set by @xmath40 where @xmath16 is quark mass . since nature provides us with two quark flavors ( @xmath41 and @xmath19 quarks ) whose mass is significantly lower ( by two orders of magnitude ) than other energy scales of the theory , typical values of @xmath40 are typically very small , resulting in a bad condition number ( @xmath42 being a typical value ) . also regarding this aspect , the situation becomes better when one is interested in the regime of very high temperatures , since in that case the spontaneous breaking of chiral symmetry disappears , the minimum eigenvalue of @xmath17 is non - zero , and the condition number significantly improves . in lqcd , the usual local updates adopted in statistical mechanics scale badly with the volume , as the action of is non - local . this problem is partly solved by the hybrid monte carlo ( hmc ) algorithm@xcite ; in hmc we associate fake conjugate momenta entering quadratically in the action to each degree of freedom of the system . for an @xmath0 gauge theory , momenta conjugate to the link variable are again @xmath43 matrices @xmath44 associated to each link of the lattice , this time living in the group algebra ( hence hermitian and traceless ) . ( [ eq : rooting ] ) is rewritten as @xmath45\exp\left(-\frac{1}{2 } h^2 -s_g[u]-\phi^{\dag } m[u]^{-1/4}\phi\right)\ , \end{aligned}\ ] ] where the momenta term is a shorthand to indicate the sum of @xmath46 over the whole lattice . the update then proceeds as follows : random gaussian initial momenta @xmath47 and pseudofermions @xmath25 are generated ; 2 . starting from the initial configuration and momenta @xmath48 , a new state @xmath49 is generated by integrating the equations of motion ; 3 . the new state @xmath49 is accepted with probability @xmath50 , where @xmath51 is the change of the total ( i.e. included the momenta ) action . step 2 is an unphysical evolution in a fictitious time and , under mild conditions on the numerical integration of the equations of motion , it can be shown to satisfy the detailed balance principle@xcite , so it provides a stochastically exact way to estimate the integral in . the more time consuming steps of the update are the ones that involve the non - local term in the exponent of . in particular , the most time consuming single step of the whole algorithm is the solution of a linear system @xmath52\varphi = b \ , .\ ] ] this calculation is needed to compute the forces appearing in the equations of motion and also to evaluate @xmath51 , and one usually resorts to krylov solvers . in the case of staggered fermions , corresponding to , it is customary to use the so - called rational hmc ( rhmc ) algorithm@xcite , in which the algebraic matrix function appearing in is approximated to machine precision by a rational function . in this case one replaces by @xmath53 equations ( @xmath53 is the order of the approximation adopted ) @xmath54+\sigma_i)\varphi_i = b\ , \quad i\in\{1,\ldots , r\}\ , \ ] ] where the real numbers @xmath55 are the poles of the rational approximations . these equations can again be solved by using krylov methods : by exploiting the shift - invariance of the krylov subspace it is possible to write efficient algorithms that solve all the equations appearing in ( [ eq : shlineq ] ) at the same time , using at each iteration only one matrix - vector product@xcite . for most of the discretizations adopted in qcd ( and in particular for the one we use ) , the matrix @xmath14 $ ] can be written in block form @xmath56 matrices @xmath57 and @xmath58 connect only even and odd sites . it is thus convenient to use an even / odd preconditioning@xcite ; in this case , is replaced by : @xmath59 @xmath60 is defined only on even sites and the matrix is positive definite ( because of ) , so we can use the simplest of the krylov solvers : the conjugate gradient ( or its shifted counterpart ) . over the years , many improvements of this basic scheme have been developed ; these are instrumental in reducing the computational cost of actual simulations but their implementation is straightforward , once the basic steps of the `` naive '' code are ready . for this reason we will not discuss in the following the details of multi - step integrators@xcite , improved integrators@xcite , multiple pseudofermions@xcite or the use of different rational approximations and stopping residuals in different parts of the hmc@xcite , even if our code uses all these improvements . our most important data structures are the collection of all gauge variables @xmath7 ( elements of the group of @xmath0 matrices , one for each link of the four - dimensional lattice ) and of the pseudofermion fields @xmath61 ( @xmath9dimensional complex vectors , one for each even site of the lattice when using the even / odd preconditioning ) . we also need many derived and temporary data structures , such as : 1 . the configurations corresponding to different stout levels ( @xmath62 , again @xmath0 matrices ) , used in the computation of the force ( typically less than five stout levels are used ) and the momenta configuration ( which are @xmath43 hermitian traceless matrices ) ; 2 . some auxiliary structures needed to compute the force acting on the gauge variables , like the so called `` staples '' @xmath63 and the @xmath64 and @xmath65 matrices@xcite ; @xmath63 and @xmath64 are generic @xmath66 complex matrices and @xmath65 are @xmath66 hermitian traceless matrices ; 3 . the solutions @xmath67 of and some auxiliary pseudofermion - like structure needed in the krylov solver . at the lowest level , almost all functions repeatedly multiply two @xmath66 complex matrices ( e.g. , in the update of the gauge part ) , or a @xmath66 complex matrix and a @xmath9dimensional complex vector ( e.g. , in the krylov solver ) or compute dot products and linear combinations of complex @xmath9vectors . all these operations have low computational intensity , so it is convenient to compress as much as possible all basic structures by exploiting their algebraic properties . the prototypical example is @xmath7 : one only stores the first two rows of the matrix and recovers the third one on the fly as the complex conjugate of the wedge product of the first two rows@xcite . this overhead is negligible with respect to the gain induced , at least for gpus , by the reduction of the memory transfer@xcite real numbers , but in this case the reconstruction algorithm presents some instabilities@xcite . ] . at a higher level the single most time consuming function is the krylov solver , which may take @xmath68 of the total execution time of a realistic simulation ( depending e.g. on the value of the temperature ) and consists basically of repeated applications iterations are needed to reach convergence , depending on the temperature . ] of the @xmath57 and @xmath58 matrices defined in , together with some linear algebra on the pseudofermion vectors ( basically _ zaxpy_-like functions ) . an efficient implementation of @xmath58 and @xmath57 multiplies is then of paramount importance , the effectiveness of this operation being often taken as a key figure of merit in the lqcd community . there is a clear trend in high - performance computing ( hpc ) to adopt multi - core processors and accelerator - based platforms . typical hpc systems today are clusters of computing nodes interconnected by fast low - latency communication networks , e.g. infiniband . each node typically has two standard multi - core cpus , each attached to one or more accelerators , either graphic processing unit ( gpu ) or many - core systems . recent development trends see a common path to performance for cpus and accelerators , based on an increasing number of independent cores and on wider vector processing facilities within each core . in this common landscape , accelerators offer additional computing performance and better energy efficiency by further pushing the granularity of their data paths and using a larger fraction of their transistors for computational data paths , as opposed to control and memory structures . as a consequence , even if cpus are more tolerant for intrinsically unstructured and irregular codes , in both class of processors computing efficiency goes through careful exploitation of the parallelism available in the target applications combined with a regular and ( almost ) branch - free scheduling of operations . this remark supports our attempt to write just one lqcd code which is not only portable , but also efficiency - portable across a large number of state - of - the - art cpus and accelerators . in this paper we consider intel multi - core cpus , and nvidia and amd gpus , commonly used today by many scientific hpc communities . [ tab : architecture ] summarizes some key features of the systems we have used@xcite , that we describe very briefly in the following . intel xeon - e5 architectures are conventional x86 multi - core architectures . we have used two generations of these processors , differing for the number of cores and for the amount of integrated last - level cache . performances in both cases rely on the ability of the application to run on all cores and to use 256-bit vector instructions . nvidia gpus are also multi - core processors . a gpu hosts several streaming multiprocessors ( sm ) , which in turn include several ( depending on the specific architecture ) compute units called cuda - cores . at each clock - cycle sms execute multiple warps , i.e. groups of 32 instructions , belonging to different cuda - threads , which are executed in _ single instructions multiple threads _ ( simt ) fashion . simt is similar to simd execution but more flexible , e.g. different cuda - threads of a simt - group are allowed to take different branches of the code , although at a performance penalty . each cuda - thread has access to its copy of registers and context switches are almost at zero cost . this structure has remained stable across several generations with minor improvements . the nvidia k80 has two gk210 gpus ; each gpu has 13 _ next generation _ streaming multiprocessor , ( smx ) running at a base frequency of @xmath69 mhz that can be increased to @xmath70 mhz under specific condition of work - load and power . the corresponding aggregate peak performance of the two gk210 units is then @xmath71 and @xmath72 tflops in double precision . the peak memory bandwidth is @xmath73 gb / s considerably higher compared to that of e5-xeon cpus . the gp100 gpu , based on the pascal architecture , has recently become available . it has 56 streaming processors running at base - frequency of @xmath74 that can be increased to @xmath75 ghz , delivering a peak double - precision performance of @xmath76 and @xmath77 tflops . peak memory bandwidth has been increased to @xmath78 gb / s . the amd gpus are conceptually similar to nvidia gpu . the amd firepro w9100 has 44 processing units , each one with 64 compute units ( stream processors ) , running at @xmath79 mhz . this board delivers a peak double - precision performance of @xmath80 tflops , and has a peak memory - bandwidth of @xmath81 gb / s . native programming models , commonly used for the systems shown in [ tab : architecture ] , differ in several aspects . for xeon - e5 cpus , the most common models are openmp and openmpi . both models support core - parallelism , running one thread or one mpi process per logical core . moreover openmp is a directive based programming model , and allows to exploit vector - parallelism properly annotating for - loops that can be parallelized@xcite . on gpus , the native programming model is strongly based on data - parallel models , with one thread typically processing one element of the application data domain . this helps exploit all available parallelism of the algorithm and hide latencies by switching among threads waiting for data coming from memory and threads ready to run . the native language is cuda - c for nvidia gpus and opencl for amd systems . both languages have a very similar programming model but use a slight different terminology ; for instance , on opencl the cuda - thread is called work - item , the cuda - block work - group , and the cuda - kernel is a device program . a cuda - c or opencl program consists of one or more functions that run either on the host , a standard cpu , or on a gpu . functions that exhibits no ( or limited ) parallelism run on the host , while those exhibiting a large degree of data parallelism can go onto the gpu . the program is a modified c ( or c++ , fortran ) program including keyword extensions defining data parallel functions , called _ kernels _ or _ device programs_. kernel functions typically translate into a large number of threads , i.e. a large number of independent operations processing independent data items . threads are grouped into blocks which in turn form the execution _ grid_. when all threads of a kernel complete their execution , the corresponding grid terminates . since threads run in parallel with host cpu threads , it is possible to overlap in time processing on the host and the accelerator . new programming approaches are now emerging , mainly based on directives , moving the coding abstraction layer at an higher lever , over the hardware details . these approaches should make code development easier on heterogeneous computing systems@xcite , simplifying the porting of existing codes on different architectures . openacc is one such programming models , increasingly used by several scientific communities . openacc is based on _ pragma _ directives that help the compiler to identify those parts of the code that can be implemented as _ parallel functions _ and offloaded on the accelerator or divided among cpu cores . the actual construction of the parallel code is left to the compiler making , at least in principle , the same code portable without modifications across different architectures and possibly offering more opportunities for performance portability . this make openacc more descriptive compared to cuda and opencl which are more prescriptive oriented . .... # pragma acc data copyin(x ) , copy(y ) { # pragma acc kernels present(x ) present(y ) async(1 ) # pragma acc loop vector(256 ) for ( int i = 0 ; i < n ; + + i ) y[i ] = a*x[i ] + y[i ] ; # pragma wait(1 ) ; } .... listing [ lst : saxpy ] shows an example of the _ saxpy _ operation of the _ basic linear algebra subprogram _ ( blas ) set coded in openacc . pragma acc kernels _ clause identifies the code fragment running on the accelerator , while _ pragma acc loop ... _ specifies that the iterations of the for - loop can execute in parallel . the standard defines several directives , allowing a fine tuning of applications . as an example , the number of threads launched by each device function and their grouping can be tuned by the _ vector _ , _ worker _ and _ gang _ directives , in a similar fashion as setting the number of _ work - items _ and _ work - groups _ in cuda . data transfers between host and device memories are automatically generated , and occur on entering and exiting the annotated code regions . several data directives are available to allow the programmer to optimize data transfers , e.g. overlapping transfers and computation . for example , in listing [ lst : saxpy ] the clause _ copyin(ptr ) _ copies the array pointed by _ ptr _ from the host memory into the accelerator memory before entering the following code region ; while _ copy(ptr ) _ perform the additional operation of copying it also back to the host memory after leaving the code region . an asynchronous directive _ async _ is also available , instructing the compiler to generate asynchronous data transfers or device function executions ; a corresponding clause ( i.e. _ # pragma wait(queue ) _ ) allows to wait for completion . openacc is similar to the openmp ( open multi - processing ) framework widely used to manage parallel codes on multi - core cpus in several ways@xcite ; both frameworks are directive based , but openacc targets accelerators in general , while at this stage openmp targets mainly multi - core cpus ; the latest release of openmp4 standard has introduced directives to manage also accelerators , but currently , compilers support is still limited . regular c / c++ or fortran code , already developed and tested on traditional cpu architectures , can be annotated with openacc pragma directives ( e.g. _ parallel _ or _ kernels _ clauses ) to instruct the compiler to transform loop iterations into distinct threads , belonging to one or more functions to run on an accelerator . ultimately , openacc is particularly well suited for developing scientific hpc codes for several reasons : * it is highly hardware agnostic , allowing to target several architectures , gpus and cpus , allowing to develop and maintain one single code version ; * the programming overhead to offload code regions to accelerators is limited to few _ pragma _ lines , in contrast to cuda and in particular opencl verbosity ; * the code annotated with openacc _ pragmas _ can be still compiled and run as plain c code , ignoring the _ pragma _ directives . in this section we describe the openacc implementation of our lqcd code . we first describe the data structures used , then we highlight the most important openacc - related details of our implementation . in writing the openacc version , we started from our previous code implementations@xcite : a c++/cuda@xcite developed for nvidia gpus aggressively optimized with cuda - specific features , and a c++ one , developed using openmp and mpi directives , targeting large cpu clusters@xcite . data structures have a strong impact on performance@xcite and can hardly be changed on an existing implementation : their design is in fact a critical step in the implementation of a new code . we have analyzed in depth the impact of data - structures for lqcd on different architectures ( i.e. a gpu and a couple of cpus ) , confirming that the _ structure of arrays _ ( soa ) memory data layout is preferred when using gpus , but also when using modern cpus@xcite . this is due to the fact that the soa format allows vector units to process many sites of the application domain ( the lattice , in our case ) in parallel , favoring architectures with long vector units ( e.g. with wide simd instructions ) . modern cpus tend indeed to have longer vector units than older ones and we expect this trend to continue in the future . for this reason , all data structures related to lattice sites in our code follow the soa paradigm . of each array c0 , c1 and c2 is a c99 complex value . see sections [ memalloc_section ] for details.,scaledwidth=80.0% ] matrices ; this structure contains 3 vectors . to mitigate memory - bandwidth requirements , one can avoid reading and writing the r2 member and recompute it on the fly , exploiting the unitarity constraint.,scaledwidth=60.0% ] in our implementation , we use the c99 double complex as basic data - type which allows to use built - in complex operators of the c library making coding easier and more readable without loss of performance . the algorithm is based on even / odd preconditioning , so the pseudo - fermion variables ( implemented as vec3_soa data - types ) live only on the even sites of the lattice . this comes at the price of requiring that all sides of the lattice must be even ; in the following we call lnh_sizeh half the number of lattice sites . the pseudofermion field has three complex values for each even lattice site , corresponding to the three qcd `` colors '' that we label c0 , c1 , c2 . a schematic representation of the vec3_soa structure is shown in fig . [ fig : mem - vec ] and a lexicographical ordering was used for the even lattice sites : @xmath82}{2 } \qquad \mathrm{s.t.}\ \sum_{i=0}^3 x_i \% 2 = 0\ , \ ] ] where lnh_n0 , lnh_n1 and lnh_n2 are the lattice sizes ; we allow for full freedom in the mapping of the physical directions @xmath83 , @xmath84 , @xmath85 and @xmath23 onto the logical directions @xmath86 , @xmath87 , @xmath88 and @xmath89 , as this option will be important for future versions of the code able to run on many processors and accelerators . the data structure used for the generic @xmath66 complex matrices is the su3_soa data - type , matrices , while actual @xmath0 matrices require in principle less memory . ] used e.g. for the `` staples '' @xmath90 and the @xmath91 matrices needed in the stouting procedure@xcite . structure su3_soa is a collection of 3 vec3_soa structures ( r0 , r1 , r2 , see fig . [ fig : mem - su3 ] ) , and data that has to be stored in this structure typically involve a number of matrices equal to the number of links present in the lattice , i.e. 8 lnh_sizeh ; this means that an array of 8 su3_soa elements is required . gauge configurations , i.e. the set of the gauge links @xmath7 and their stouted counterparts , are stored in memory as an array of 8 su3_soa structures . as previously explained the algorithm is typically bandwidth limited and for @xmath0 matrices it is convenient to read and write just the first two rows , computing the third one on the fly as @xmath92 . note that the soa memory layout avoids the prefetching problems discussed in similar cases@xcite . other data structures are needed to store in memory @xmath66 traceless hermitian matrices or @xmath66 traceless anti - hermitian matrices . in these cases , only 8 real parameters per matrix are needed : 3 complex numbers for the upper triangular part and the first two elements of the diagonal , which are real ( imaginary ) numbers for ( anti-)hermitian traceless matrices . these data structures have been implemented according to the soa scheme as follows : thmat_soa and tamat_soa contain 3 vectors of c99 double complex numbers and 2 vectors of double numbers , in a form that closely resemble the one of vec3_soa . data movements between device and host are negligible , with significant transfers happening only at the beginning and at the end of each monte carlo update , and managed mainly with the update device and update host openacc directives . to initially assess the performance level achievable using openacc , we have developed a mini - application benchmark of the dirac operator@xcite . as previously underlined this is the fundamental building block of the krylov solver , commonly accounting for not less than @xmath93 of the running time , and reaching up to @xmath94 in low temperature simulations . this compute intensive part of an lqcd simulation is where most of the optimization efforts are usually concentrated@xcite . the dirac operator code uses three functions : deo , doe ( corresponding respectively to the application of functions @xmath58 and @xmath57 defined in ) and a _ zaxpy_-like function which is negligible in terms of execution time . a direct comparison indicated that the performance of the openacc versions of the double precision deo and doe functions were comparable with the cuda ones@xcite . this promising start was a strong indication that also for lqcd the higher portability of the openacc implementation is not associated with a serious loss of performance , and motivated us to proceed to an openacc implementation of the full rhmc code . as a side benefit , the use of the openacc programming model significantly simplified the implementation of algorithmic improvements . the implementation of these new features started with the coding and testing of the improvements on a single thread version . after the algorithm is validated , the acceleration is switched on by annotating the code with # pragma directives . in order to have a more readable code , the most complex kernels have been split in several functions . while small functions can be used in kernels if declared as static inline , for larger ones we had to use the routine seq openacc directive as large functions can not be inlined . kernels have been parallelized following two different approaches . those using data belonging to nearest ( and/or next - to - nearest ) neighbors have been parallelized via the # pragma acc loop directive on @xmath95 nested loops , one for each dimension . this allows to use 3d thread blocks , which should improve data reuse between threads thus reducing bandwidth requirements , which is our major performance concern . the other kernels , i.e. the ones performing only single - site operations , have been parallelized using a single cycle running on the lattice sites . after the implementation of a first full working openacc simulation , various optimization iterations took place , in particular for the performance critical steps . these include the dirac operator in the first place , but also the gauge part of the molecular dynamics steps , since their relative impact on the overall execution time is very large , as shown in [ tab : functions ] for a few representative examples . during the full development phase , every time a new openacc feature has been introduced , extensive checks have been performed to ensure the correctness of the improved code , against possible semantic misunderstanding of openacc clauses or compiler bugs . this section describes the overall structure of our code , and focuses on the openacc implementation of selected performance - critical parts . read gauge configuration @xmath96 create momenta @xmath97 generate pseudofermions by heatbath [ pseudoferm_generation_algorithm ] calculation of initial action molecular dynamics [ possibly in single precision ] [ moldyn_step_algorithm ] calculate action variation @xmath51 [ final_action_calculation_algorithm ] montecarlo step accepted with probability @xmath98 take measurements algorithm [ big_scheme_of_things_algorithm ] is a top - level description of the full code , showing the main computational tasks . for performances , the most critical steps are molecular dynamics ( step [ moldyn_step_algorithm ] ) followed by the heatbath generation of the pseudofermions ( step [ pseudoferm_generation_algorithm ] ) , and the calculation of the final action ( step [ final_action_calculation_algorithm ] ) . steps [ pseudoferm_generation_algorithm ] and [ final_action_calculation_algorithm ] consist basically in function calls to the multishift inverter routine , with a high target accuracy . the outer level of the multistep integrator for molecular dynamic evolution ( step [ moldyn_step_algorithm ] ) in algorithm [ big_scheme_of_things_algorithm ] is expanded in algorithm [ outer_cycle_algorithm ] . as explained in sec.([grand_challenge_section ] ) , in zero temperature simulations or for small quark masses usually the heaviest computational parts are the calculations of the fermion force , while in high temperature simulations the load is shifted inside the gauge cycles , as already shown in [ tab : functions ] . the fermion force calculation step is implemented following@xcite ; for this step a large fraction of the execution time is is spent in computation of deo and doe functions implementing the dirac operator . .... void acc_deo ( _ _ restrict const su3_soa * const u , _ _ restrict vec3_soa * const out , _ _ restrict const vec3_soa * const in , _ _ restrict const double_soa * const backfield ) { int hd0 , d1 , d2 , d3 ; # pragma acc kernels present(in ) present(out ) present(u ) present(backfield ) async(1 ) # pragma acc loop independent gang(gang ) for(d3=0 ; d3<nd3;d3++ ) { # pragma acc loop independent vector tile(tile0,tile1,tile2 ) for(d2=0 ; d2<nd2 ; d2++ ) { for(d1=0 ; d1<nd1 ; d1++ ) { for(hd0=0 ; hd0 < nd0h ; hd0++ ) { ... } } } } } .... the deo openacc implementation is shown in listing [ lst : dirac ] , showing the 4 dimension nested loops and the corresponding pragma directives . in this listing openacc directives are used : i ) to identify the data structures already present in the accelerator memory , when targeting accelerators ( present ( ) clause ) ; ii ) to make the compiler aware of the data independence of loops iterations ( independent clause ) ; iii ) to request to group iterations in order to execute them in the same ( or close ) compute units ( tile clause ) . in particular , the tile openacc clause asks the compiler to split or strip - mine each loop in the nest into two loops , an outer tile loop and an inner element loop . where possible ( e.g. in deo and doe ) , performing computations of adjacent lattice sites in close hardware compute units may increase data reuse ( i.e. matrices shared between sites)@xcite for all the architectures where data caches are present , which means almost every modern processing architecture . the tile sizes offering the best performance depend , for each kernel , on several features of each specific architecture , e.g. vector units size , register numbers , cache levels and sizes . we keep a door open for limited architecture - specific optimization , by allowing to specify the tile0 , tile1 , tile2 variables at compile time , telling the compiler how to group together iterations involving adjacent lattice sites . fermion force calculation evolve momenta for @xmath99 @xcite ( @xmath100 ) fermion force calculation evolve momenta for @xmath101 ( @xmath100 ) fermion force calculation evolve momenta for @xmath102 ( @xmath100 ) fermion force calculation evolve momenta for @xmath101 ( @xmath100 ) fermion force calculation evolve momenta for @xmath103 the actual evolution of the gauge configuration happens inside the inner gauge cycles , where the gauge contribution to the momenta evolution is also calculated . among the tasks performed in the gauge cycles , the computation of staples in the gauge force calculation is the most time consuming . it consists of calculating 6 products of 3 and 5 @xmath0 matrices representing links on c - shaped paths on the lattice . the implementation of one of these functions is sketched in listing [ lst : rect_staples ] : also in this case the parallelization has been done using the tile directive over the 3 innermost nested cycles . this allows us also in this case to use 3d thread blocks , which should improve data reuse between threads , reducing the bandwidth needs . we shall also remark that in this case , since second - nearest - neighbor - site addressing is needed , for the sake of simplicity we use indirect addressing . notice that the function staple_type1 ( as well as similar ones ) has to be declared with # pragma acc routine seq to be used inside a kernel . .... # pragma acc routine seq void staple_type1 ( ... ) { ... } void calc_staples_type1 ( _ _ restrict const su3_soa * const u , _ _ restrict su3_soa * const loc_stap ) { int d0 , d1 , d2 , d3 , mu , iter ; # pragma acc kernels present(u ) present(loc_stap ) present(nnp_openacc ) present(nnm_openacc ) # pragma acc loop independent gang(impstapgang3 ) for(d3=0 ; d3<nd3 ; d3++ ) { # pragma acc loop independent vector tile(impstaptile0 , impstaptile1 , impstaptile2 ) for(d2=0 ; d2<nd2 ; d2++ ) { for(d1=0 ; d1<nd1 ; d1++ ) { for(d0=0 ; d0 < nd0 ; d0++ ) { # pragma acc loop seq for(mu=0 ; mu<4 ; mu++ ) { ... const int idx_pmu = nnp_openacc[idxh][mu][parity ] ; ... staple_type1(&u[dir_nu_1r ] , idx_pmu , ... ) .... in order to improve performance , we also implemented a single precision version of the code for the molecular dynamics evolution . due to the low arithmetic density of the lqcd algorithms , on gpus at least , all kernels are memory - bound ; this means that , when precision is not an issue , it is preferable to have single precision versions of selected functions and structures , as a plain @xmath104 increase in performance is expected with respect to the double precision implementation . to compare the performance of our code on different architectures we consider two different benchmarks taking into account the most computational intensive parts of the code . the first benchmark evaluates the performance of the dirac operator , both single and double precision version , and the latter evaluates the performance of the gauge part of the molecular dynamics step . depending on input configuration parameters either the former or the latter kernels make up most of the execution time of a typical simulation , as shown in [ tab : functions ] . we present the execution time per site of the dirac operator for different lattice sizes in [ tab : dirac - norm ] . exactly the same code has been run on all platforms without requiring any change ; we have just re - compiled it with different flags instructing the pgi 16.10 compiler to target the corresponding architectures and using the best tile dimensions for each of them . we tested two different nvidia gpus , the k80 based on the kepler architecture and the recently released p100 board based on the pascal architecture . for the k80 the single precision version takes @xmath105 per lattice site , while the double precision version requires @xmath106 . running on the p100 we measure @xmath107 for single and @xmath108 for double precision , improving approximately by a factor @xmath109 over the k80 . this results perfectly scales with architecture performance of p100 that has @xmath110 more cores and @xmath111 more memory bandwidth , see [ tab : architecture ] . concerning intel cpus , we have compared two different processors , the 8-core e5 - 2630v3 cpu based on haswell architecture , and the 18-core e5 - 2697v4 cpu based on broadwell . since computing resource of the cpus are roughly @xmath109 lower than on gpus , see [ tab : architecture ] , a performance drop is expected . however , the actual performance drop measured on both cpus is much larger than this expected theoretical figure ; indeed time per site is approximately @xmath112 or larger on the haswell than on one k80 gpu . the broadwell performs approximately a factor @xmath113 better compared to haswell , at least for some lattice sizes . we have identified two main reasons for this non - optimal behavior , and both of them point to some still immature features of the pgi compiler when targeting x86 architectures , that we expect should be soon resolved : * * parallelization * - the compiler is only able to split outer - loops across different threads , while inner loops are executed serially or vectorized within each thread . this explains why on the broadwell cpu running on a lattice @xmath114 we have a performance @xmath113 better than for a @xmath115 lattice , which has the same volume but allows to split the outer loop only on @xmath1 threads . * * vectorization * - as reported by the compilation logs , the compiler fails to vectorize the deo and doe functions computing the dirac operator ( see listing [ lst : dirac ] ) reporting to be unable to vectorize due to the use of `` mixed data - types '' . to verify if this is related to how we have coded these functions , we have translated the openacc pragmas into the corresponding openmp ones without changing the c code and compiled using the intel compiler ( version 17.0.1 ) . in this case the compiler succeeds in vectorizing the two functions , running a factor @xmath116 faster compared to the openacc version compiled by pgi compiler . [ tab : md - norm ] shows the execution time of the gauge part of the molecular dynamics step . as already remarked this is one of the two most time - consuming steps together with the application of the dirac operator . as we see the update time per site is quite stable for all lattice sizes we have tried and for all architectures . going from the nvidia k80 to the p100 the time improves by a factor @xmath117 , while between haswell and broadwell we have roughly a factor @xmath118 / @xmath119 . we finally mention that we have also been able to compile and run our code on an amd firepro w9100 gpu and on the latest version of the intel xeon phi processor , the knights landing ( knl ) . however , in these cases , results are still preliminary . in more details , the compiler itself crashes when compiling the code for the amd gpu for some specific lattice sizes ; for the knl , specific compiler support is still missing , but this processor is able to run the code compiled for the haswell architecture , implying however that 512-bit vectorization is not used . these problems do not allow us to perform a systematic comparison of performance for these architectures . once again , we believe that this is due to some immaturity of the compiler , and we expect that these issues will be resolved in future versions . [ tab : cuda - openacc ] addresses the question of the efficiency costs ( if any ) of our architecture - portable code ; the table compares the execution time for a _ full _ monte carlo step ( in double precision ) of the openacc code and a previously developed cuda implementation@xcite , optimized for nvidia gpus . although the two codes are not exactly in a one to one correspondence , the implementations are similar enough to make such a test quantitatively meaningful . one immediately sees that the performances of the two implementations are comparable and the use of openacc does not imply a dramatic performance loss , the differences between the execution times of the two versions being of the order of @xmath120 . the worst case is the one of the @xmath121 lattice , in which openacc is about @xmath122 slower than cuda . since we are comparing an high - level version of the code with one specifically developed for nvidia gpus , this would not be a dramatic loss , however in this case the comparison is also not completely fair . indeed for this high temperature simulation the gauge part of the molecular dynamic step starts to be the computationally heaviest task and , in the cuda implementation , part of it had been explicitly hard coded in single precision . for the low - temperature test cases the differences between the cuda and the openacc implementation are much smaller and , in fact , in one case the openacc version is the fastest one . a possible explanation of this is the following : in the cuda version unidimensional blocks are adopted to parallelize the dirac operator , while in the openacc implementation three - dimensional block structures are used , that fit better the larger cache of recent gpus and , especially on larger lattices , improves data reuse . in this work we have developed a full state - of - the - art production - grade code for lattice qcd simulations with staggered fermions , using the openacc directive - based programming model . our implementation includes all steps of a complete simulation , and most of them run on accelerators , minimizing the transfer of lattice data to and from the host . we have used the pgi compiler , which supports the openacc standard and is able to target almost all current architectures relevant for hpc computing , even if with widely different levels of maturity and reliability . exactly the same code runs successfully on nvidia many - core gpus and intel multi - core cpus , and for both architectures we have measured roughly comparable levels of efficiency . also , the performance of the complete code is roughly the same as that of an equivalent code , specifically optimized for nvidia gpus and written in the cuda language . our code also runs on amd gpus and on the knl intel phi processor , even if the compilation and run - time environment for these processors is still unable to deliver production - grade codes ; in these cases , we have strong indications that these problems come from a residual immaturity of the compilation chain and we expect that they will be soon resolved . all in all , our final result is a lqcd monte carlo code portable on a large subset of hpc relevant processor architectures and with consistent performances . some further comments are in order : i ) using a directive - based programming model , we are able to target different computing platforms presently used for hpc , avoiding to rewrite the code when moving from one platform to another ; ii ) the openacc standard provides a good level of hardware abstraction requiring the programmer to only specify the function to be parallelized and executed on the accelerator ; the compiler is then able to exploit the parallelism according to the target processor , hiding from the programmer most hardware optimizations ; iii ) the openacc code has roughly the same level of performance of that implemented using a native language such as cuda for nvidia gpus , allowing to efficiently exploit the computing resources of the target processor . in the near future we plan to carefully assess performances on amd and knl systems , in order to enlarge the platform portfolio of our code . we also plan to assess whether openmp4 provides the same level of portability as openacc , as soon as compilers supporting this programming standard become available . this is important to have a unique directive - based programming model which is widely used by several scientific communities and supported by several compilers ( gcc , icc , ) . finally , we are already working on a massively parallel version of our code , able to run concurrently on a large clusters of cpus and accelerators . we warmly thank francesco sanfilippo ( the developer of the nissa code@xcite ) for his advice and support . we thank the infn computing center in pisa for providing us with the development framework , and universit degli studi di ferrara and infn - ferrara for the access to the coka gpu cluster . this work has been developed in the framework of the suma , coka and cosa projects of infn . fn acknowledges financial support from the infn suma project . o. villa , d. r. johnson , m. oconnor , e. bolotin , d. nellans , j. luitjens , n. sakharnykh , p. wang , p. micikevicius , a. scudiero , s. w. keckler and w. j. dally , scaling the power wall : a path to exascale , in _ proceedings of the international conference for high performance computing , networking , storage and analysis _ , sc 14 ( ieee press , piscataway , nj , usa , 2014 ) . doi : 10.1109/sc.2014.73[doi : 10.1109/sc.2014.73 ] . c. bonati , e. calore , s. coscetti , m. delia , m. mesiti , f. negro , s. f. schifano and r. tripiccione , development of scientific software for hpc architectures using openacc : the case of lqcd , in _ the 2015 international workshop on software engineering for high performance computing in science ( se4hpcs ) _ , icse companion proceedings2015 . doi : 10.1109/se4hpcs.2015.9[doi : 10.1109/se4hpcs.2015.9 ] . d. pflger , m. mehl , j. valentin , f. lindner , d. pfander , s. wagner , d. graziotin and y. wang , the scalability - efficiency / maintainability - portability trade - off in simulation software engineering : examples and a preliminary systematic literature review , in _ proceedings of the fourth international workshop on software engineering for hpc in computational science and engineering _ , se - hpccse 162016 . doi : 10.1109/se - hpccse.2016.8[doi : 10.1109/se - hpccse.2016.8 ] . s. wienke , c. terboven , j. beyer and m. mller , _ lecture notes in computer science ( including subseries lecture notes in artificial intelligence and lecture notes in bioinformatics ) _ * 8632 * , 812 ( 2014 ) . m. g. lopez , v. v. larrea , w. joubert , o. hernandez , a. haidar , s. tomov and j. dongarra , towards achieving performance portability using directives for accelerators , in _ proceedings of the third international workshop on accelerator programming using directives _ , waccpd 162016 . doi : 10.1109/waccpd.2016.9[doi : 10.1109/waccpd.2016.9 ] . c. bernard _ et al . _ , nucl . phys . proc . suppl . * 106 * , 199 ( 2002 ) . g. bilardi , a. pietracaprina , g. pucci , f. schifano and r. tripiccione , _ lecture notes in computer science _ * 3769 * , 386 ( 2005 ) , doi : 10.1007/11602569_41[doi : 10.1007/11602569_41 ] . f. belletti , s. f. schifano , r. tripiccione , f. bodin , p. boucaud , j. micheli , o. pene , n. cabibbo , s. de luca , a. lonardo , d. rossetti , p. vicini , m. lukyanov , l. morin , n. paschedag , h. simma , v. morenas , d. pleiter and f. rapuano , _ computing in science and engineering _ * 8 * , 50 ( 2006 ) , doi : 10.1109/mcse.2006.4[doi : 10.1109/mcse.2006.4 ] . p. a. boyle , d. chen , n. h. christ , m. clark , s. cohen , z. dong , a. gara , b. joo , c. jung , l. levkova , x. liao , g. liu , r. d. mawhinney , s. ohta , k. petrov , t. wettig , a. yamaguchi and c. cristian , qcdoc : a 10 teraflops computer for tightly - coupled calculations , in _ supercomputing , 2004 . proceedings of the acm / ieee sc2004 conference _ , nov 2004 . doi : 10.1109/sc.2004.46[doi : 10.1109/sc.2004.46 ] . h. baier , h. boettiger , m. drochner , n. eicker , u. fischer , z. fodor , a. frommer , c. gomez , g. goldrian , s. heybrock , d. hierl , m. hsken , t. huth , b. krill , j. lauritsen , t. lippert , t. maurer , b. mendl , n. meyer , a. nobile , i. ouda , m. pivanti , d. pleiter , m. ries , a. schfer , h. schick , f. schifano , h. simma , s. solbrig , t. streuer , k .- h . sulanke , r. tripiccione , j .- s . vogt , t. wettig and f. winter , _ computer science - research and development _ * 25 * , 149 ( 2010 ) , doi : 10.1007/s00450 - 010 - 0122 - 4[doi : 10.1007/s00450 - 010 - 0122 - 4 ] . j. doi , peta - scale lattice quantum chromodynamics on a blue gene / q supercomputer , in _ proceedings of the international conference on high performance computing , networking , storage and analysis _ , sc 12 ( ieee computer society press , los alamitos , ca , usa , 2012 ) . r. babich , m. a. clark and b. joo , parallelizing the quda library for multi - gpu calculations in lattice quantum chromodynamics , in _ sc 10 ( supercomputing 2010 ) new orleans , louisiana , november 13 - 19 , 2010 _ , 2010 . o. philipsen , c. pinke , a. sciarra and m. bach , cl@xmath123qcd - lattice qcd based on opencl , in _ proceedings , gpu computing in high - energy physics ( gpuhep2014 ) : pisa , italy , september 10 - 12 , 2014 _ , 2015 . doi : 10.3204/desy - proc-2014 - 05/30[doi : 10.3204/desy - proc-2014 - 05/30 ] . s. blair , c. albing , a. grund and a. jocksch , accelerating an mpi lattice boltzmann code using openacc , in _ proceedings of the second workshop on accelerator programming using directives _ , waccpd 15 ( acm , new york , ny , usa , 2015 ) . doi : 10.1145/2832105.2832111[doi : 10.1145/2832105.2832111 ] . j. kraus , m. schlottke , a. adinetz and d. pleiter , accelerating a c++ cfd code with openacc , in _ accelerator programming using directives ( waccpd ) , 2014 first workshop on _ doi : 10.1109/waccpd.2014.11[doi : 10.1109/waccpd.2014.11 ] . b. jo , m. smelyanskiy , d. d. kalamkar and k. vaidyanathan , wilson dslash kernel from lattice qcd optimization , in _ high performance parallelism pearls volume 2 : multicore and many - core programming approaches _ , eds . j. reinders and j. jeffers ( morgan kaufmann , boston , 2015 ) , pp . doi : 10.1016/b978 - 0 - 12 - 803819 - 2.00023 - 9[doi : 10.1016/b978 - 0 - 12 - 803819 - 2.00023 - 9 ] . e. calore , n. demo , s. f. schifano and r. tripiccione , experience on vectorizing lattice boltzmann kernels for multi- and many - core architectures , in _ parallel processing and applied mathematics : 11th international conference , ppam 2015 , krakow , poland , september 6 - 9 , 2015 . revised selected papers , part i _ , eds . r. wyrzykowski , e. deelman , j. dongarra , k. karczewski , j. kitowski and k. wiatrlecture notes in computer science ( springer international publishing , cham , 2016 ) , pp . doi : 10.1007/978 - 3 - 319 - 32149 - 3_6[doi : 10.1007/978 - 3 - 319 - 32149 - 3_6 ] . b. jo , m. smelyanskiy , d. d. kalamkar and k. vaidyanathan , chapter 9 - wilson dslash kernel from lattice qcd optimization , in _ high performance parallelism pearls _ , eds . j. reinders and j. jeffers ( morgan kaufmann , 2015 ) , pp . 139 170 . doi : 10.1016/b978 - 0 - 12 - 803819 - 2.00023 - 9[doi : 10.1016/b978 - 0 - 12 - 803819 - 2.00023 - 9 ] . b. jo , d. d. kalamkar , t. kurth , k. vaidyanathan and a. walden , _ optimizing wilson - dirac operator and linear solvers for intel knl _ , in _ high performance computing : isc high performance 2016 international workshops _ , eds . m. taufer , b. mohr and j. m. kunkel ( springer international publishing , june 2016 ) , pp . 415427 . revised selected papers .

for any group , there is a natural ( pseudo-)norm on the vector space of real homogenized ( group )-boundaries , called the _ stable commutator length _ norm . this norm is closely related to , and can be thought of as a relative version of , the gromov ( pseudo)-norm on ( ordinary ) homology . we show that for a free group , the unit ball of this pseudo - norm is a rational polyhedron . it follows that stable commutator length in free groups takes on only rational values . moreover every element of the commutator subgroup of a free group rationally bounds an injective map of a surface group . the proof of these facts yields an algorithm to compute stable commutator length in free groups . using this algorithm , we answer a well - known question of bavard in the negative , constructing explicit examples of elements in free groups whose stable commutator length is not a half - integer . [ section ] [ theorem]lemma [ theorem]proposition [ theorem]corollary [ theorem]conjecture [ theorem]question [ theorem]problem [ theorem]definition [ theorem]construction [ theorem]notation [ theorem]remark [ theorem]example p

0802.1352

stable commutator length is a numerical invariant of elements in the commutator subgroup of a group . it is intimately related to ( two - dimensional ) bounded cohomology , and appears in many areas of low - dimensional topology and dynamics , from the milnor - wood inequality , to the @xmath2 conjecture . however , although a great deal of work has gone into estimating or bounding stable commutator length in many contexts , there are very few known nontrivial examples of finitely presented groups in which it can be calculated exactly , and virtually no cases where the range of stable commutator length on a given group can be understood arithmetically . the most significant results of this paper are as follows : 1 . we show that stable commutator length in free groups takes on only rational values , and give an explicit algorithm to compute the value on any given element . this is the first example of a group with infinite dimensional second bounded cohomology group @xmath3 in which stable commutator length can be calculated exactly . we show how to extend stable commutator length to a ( pseudo)-norm on the vector space @xmath0 of homogenized real ( group ) @xmath1-cochains that are boundaries of @xmath4-cochains . in the case of a free group , this is a genuine norm . we show that the intersection of the unit ball in this norm with any finite dimensional rational subspace of @xmath0 of a free group is a finite - sided rational polyhedron . this invites comparison with the thurston norm on the @xmath4-dimensional homology of a @xmath5-manifold @xcite , although the relationship between the two cases is subtle and deserves further investigation . we give examples of explicit elements in the commutator subgroup of @xmath6 ( the free group of rank @xmath4 ) for which the stable commutator length is not in @xmath7 . this answers in the negative a well - known question of bavard @xcite . we now elaborate on these points in turn . let @xmath8 be a group . for @xmath9 $ ] , the _ commutator length _ of @xmath10 , denoted @xmath11 , is the smallest number of commutators in @xmath8 whose product is equal to @xmath10 . the _ stable commutator length _ of @xmath10 , denoted @xmath12 , is the limit of @xmath13 as @xmath14 . in geometric language , ( see e.g. gromov @xcite ) @xmath15 is sometimes called `` filling genus '' , and @xmath16 is called `` stable filling genus '' . this quantity is intimately related , by bavard duality and an exact sequence ( see theorem [ duality_theorem ] and proposition [ exact_sequence_prop ] ) to the second bounded cohomology @xmath3 of @xmath8 , with its banach norm . despite a considerable amount of research , there are very few known examples of groups @xmath8 in which @xmath16 can be calculated exactly ( except when it vanishes identically ) . this is partly because the groups @xmath3 , when nontrivial , tend to be very large in general : when @xmath8 is word - hyperbolic , @xmath3 is not merely infinite dimensional , but is not even separable as a banach space . calculating @xmath16 is tantamount to solving an extremal problem in @xmath3 . ( technically , one solves the extremal problem in the space of _ homogeneous quasimorphisms _ @xmath17 , see definition [ quasimorphism_definition ] . the spaces @xmath17 and @xmath3 are both banach spaces , and are related by the coboundary @xmath18 , which is fredholm when @xmath8 is finitely presented . ) gromov ( @xcite 6.@xmath19 ) asked whether @xmath16 is always rational in finitely presented groups . the answer to gromov s question is known to be _ no _ : dongping zhuang gave the first examples in @xcite . these examples occur in generalized stein - thompson groups of pl homeomorphisms of the circle , where one can show that @xmath3 is actually finite dimensional , and everything can be calculated explicitly . in this paper we show that _ @xmath16 is rational _ in free groups ( and some closely related groups ) , and moreover we give an explicit algorithm to compute the value of @xmath16 on any element . if @xmath20 are elements in @xmath8 , define @xmath21 to be the smallest number of commutators in @xmath8 whose product is equal to the product of conjugates of the @xmath22 . let @xmath23 denote the limit of @xmath24 as @xmath14 . this function can be extended by linearity and continuity in a unique way to a pseudo - norm on @xmath25 , the vector space of real group @xmath1-chains on @xmath8 that are in the image of the boundary map @xmath26 . this function vanishes identically on the subspace @xmath27 of @xmath25 spanned by terms of the form @xmath28 and @xmath29 for @xmath30 and @xmath31 , and descends to a pseudo - norm on the quotient @xmath32 , or @xmath0 for short . when @xmath8 is hyperbolic , @xmath16 is a genuine norm on @xmath0 . we show that in a free group , this @xmath16 norm is _ piecewise rational linear _ ( denoted pql ) on finite dimensional rational subspaces of @xmath0 . so for any finite set of elements @xmath33 , there is a uniform upper bound on the denominators of the values of @xmath16 on integral linear chains @xmath34 in @xmath0 . one should compare the @xmath16 norm with the gromov - thurston norm @xcite , which is a norm on @xmath35 of an irreducible , atoroidal @xmath5-manifold , and whose most significant feature is that it is a piecewise rational linear function . in thurston s definition ( in which one restricts to embedded surfaces ) this is straightforward to show . in gromov s definition ( in terms of chains , or immersed surfaces ) this is a very deep theorem , whose proof depends on the full power of gabai s theory of sutured hierarchies @xcite , and taut foliations . in fact , it is reasonable to think of the @xmath16 norm as a relative gromov - thurston norm , with gromov s definition . our proof of rationality is conceptually close in some ways to an argument due to oertel @xcite insofar as both proofs reduce the problem of calculating the norm to a linear programming problem in the vector space of weights carried by a finite constructible branched surface . however , there are crucial differences between the two cases . in oertel s case , the branched surface might have complicated branch locus , but it comes with an embedding in a @xmath5-manifold . in our case , the branch locus is simple , but the branched surface is merely _ immersed _ in a @xmath5-manifold . it is intriguing to try to find a natural generalization of both theories . in his seminal paper @xcite on stable commutator length , bavard asked whether stable commutator length in free groups takes values in @xmath7 . there were several pieces of direct and indirect evidence for this conjecture . firstly , where certain ( geometric or homological ) methods for estimating stable commutator length in free groups give exact answers , these answers in every case confirm bavard s guess . secondly , in the ( analogous ) context of @xmath5-manifold topology , one knows that the gromov norm of an integral @xmath4-dimensional homology class is in @xmath36 ( the factor of @xmath37 arises because gromov norm counts triangles , whereas stable commutator norm counts handles ) . it was generally felt that bavard s conjecture was eminently plausible , and it is therefore surprising that our algorithm produces many elements whose stable commutator length is not in @xmath7 . in fact experiments suggest that arbitrarily large denominators occur , with arbitrary prime factors . in view of these examples , the fact that stable commutator length is rational in free groups is seen to be a more delicate and subtle fact than one might have imagined , and stable commutator length to be a richer invariant than previous work has suggested . the organization of this paper is as follows . in [ background_section ] we state definitions and sketch proofs of background results which pertain to stable commutator length in groups in general . in [ free_group_section ] we specialize to the case of free groups . the purpose of this section is to state and prove the `` rationality theorem '' , whose precise statement is the following : let @xmath38 be a free group . 1 . @xmath39 for all @xmath40 $ ] . every @xmath41 $ ] rationally bounds an extremal surface ( in fact , every rational chain @xmath42 in @xmath0 rationally bounds an extremal surface ) 3 . the function @xmath16 is piecewise rational linear on @xmath0 . 4 . there is an algorithm to calculate @xmath16 on any finite dimensional rational subspace of @xmath0 . similar rationality results hold for stable commutator length in virtually free groups , and fundamental groups of noncompact seifert - fibered @xmath5-manifolds . finally , in [ algorithm_section ] we explicitly describe an algorithm for computing the stable commutator length in free groups , and discuss a simple example that answers bavard s question in the negative . for the convenience of the reader , we collect here some basic definitions and properties that will be used in subsequent sections . as general background , see @xcite , @xcite , @xcite and @xcite . note that the reference @xcite , although the most detailed , complete and relevant to the material in this paper , is an unfinished manuscript ( which is readily available online ) and therefore we have tried to refer to this manuscript by section number ( which one can expect to be reasonably stable ) rather than by page number . in this section we give the definitions and basic properties of stable commutator length in groups . this is a numerical invariant of elements in the commutator subgroup of a given group which is _ universal _ for certain kinds of extremal problems . for background or proofs , see @xcite or @xcite . [ scl_definition ] let @xmath8 be a group . for @xmath9 $ ] the _ commutator length _ of @xmath10 , denoted @xmath11 , is the smallest number of commutators in @xmath8 whose product is equal to @xmath10 . the _ stable commutator length _ , denoted @xmath12 , is the following limit @xmath43 note that the function @xmath13 is subadditive , so the limit in definition [ scl_definition ] exists . notice further that @xmath15 and @xmath16 are class functions , and that they are monotone non - increasing under homomorphisms between groups . if we need to emphasize that @xmath15 or @xmath16 is being calculated in a fixed group , we will use subscripts ; hence @xmath44 and @xmath45 . we sometimes extend @xmath15 and @xmath16 to all of @xmath8 by defining @xmath46 if @xmath10 is not in @xmath47 $ ] , and replacing @xmath48 by @xmath49 in the definition of @xmath16 . notice that @xmath50 if and only if some power of @xmath10 is in @xmath47 $ ] . the functions @xmath15 and @xmath16 can be interpreted geometrically . let @xmath51 be a connected cw complex with @xmath52 , and let @xmath53 be a loop in @xmath51 whose free homotopy class represents the conjugacy class of @xmath10 . then @xmath54 if and only if there exists an orientable surface @xmath55 of genus @xmath56 with one boundary component , and a map @xmath57 taking @xmath58 to the free homotopy class of @xmath53 . in order to be able to speak interchangeably about loops @xmath53 in spaces @xmath51 as above and their images , we assume in the sequel that all spaces @xmath51 are such that every free homotopy class of loop can be realized by an embedded circle . this can be achieved , for an arbitrary homotopy type of cw complex @xmath51 , by multiplying by a sufficiently high dimensional cube . genus is not multiplicative under finite covers , but euler characteristic is . so when we stabilize @xmath15 , the relevant geometric quantity to keep track of is derived from euler characteristic . let @xmath55 be a compact , connected , oriented surface . then set @xmath59 extend @xmath60 additively to compact , oriented ( but not necessarily connected ) surfaces @xmath55 , so that @xmath61 where @xmath62 ranges over the components of @xmath55 ( cf . @xcite ) . let @xmath55 be a compact , connected , oriented surface . let @xmath51 be a topological space , and @xmath63 a continuous loop . let @xmath57 be such that there is a commutative diagram @xmath64 where @xmath65 is the inclusion map , and define @xmath66 by the identity @xmath67 = n(s)[s^1]$ ] in @xmath68 . if the ( oriented ) components of @xmath58 are denoted @xmath69 , then @xmath66 is the sum of the degrees of the maps @xmath70 . informally , @xmath66 is the degree with which @xmath58 wraps around the loop @xmath53 . if @xmath66 is nonzero , one says that the surface @xmath55 _ rationally bounds @xmath53_. strictly speaking , @xmath66 depends on @xmath71 and not just on @xmath55 , but we suppress this in our notation . with these definitions , one can give a geometric interpretation of @xmath16 . [ euler_lemma ] let @xmath51 be a connected cw complex with @xmath52 and let @xmath53 be a loop in @xmath51 in the free homotopy class corresponding to the conjugacy class of @xmath10 . then @xmath72 where the infimum is taken over all maps @xmath57 wrapping @xmath58 around @xmath53 with any degree @xmath66 . an inequality in one direction can be obtained by restricting the class of admissible @xmath55 to those that are connected with exactly one boundary component . to obtain the inequality in the other direction , first observe that components without boundary can be thrown away without increasing @xmath73 . passing to a cover multiplies both @xmath73 and @xmath56 by the same factor . moreover an orientable surface with @xmath74 boundary components admits a cover ( in fact , a cyclic cover ) of degree @xmath75 which also has @xmath74 boundary components , providing @xmath75 and @xmath76 are coprime . after passing to such a cover with @xmath75 very large , multiple boundary components can be tubed together with @xmath1-handles ( whose image in @xmath51 can be taken to be a point ) , increasing @xmath73 by a term which is arbitrarily small compared to @xmath56 , thereby proving the theorem . by changing the orientation on @xmath55 if necessary , we may always take @xmath66 to be positive . in the sequel we therefore adhere to the convention that @xmath66 is positive unless we explicitly say otherwise . on the other hand , even if @xmath66 is positive , if @xmath55 has more than one boundary component , some components might map to @xmath53 with positive degree , and others with negative degree . say further that a surface @xmath55 is _ monotone _ if the degree of every component @xmath77 is positive . the following lemma shows that for the purposes of computing @xmath16 , one can restrict attention to monotone surfaces . [ monotone_surface_lemma ] let @xmath57 be a connected surface with @xmath78 that rationally bounds @xmath53 . then there is another surface @xmath79 which is monotone , and satisfies @xmath80 . each component @xmath69 of @xmath58 maps to @xmath53 with some degree @xmath81 , where @xmath82 . if some @xmath83 is zero , the image @xmath84 is homotopically trivial in @xmath51 , so we may reduce @xmath85 without affecting @xmath66 by compressing @xmath69 . so without loss of generality , assume that no @xmath83 is zero . since @xmath86 is negative , there is a finite cover of @xmath55 with positive genus . if @xmath55 is a connected surface with positive genus and negative euler characteristic , there is a connected degree @xmath4 cover of @xmath55 such that each boundary component in @xmath55 has exactly two preimages in the cover . hence , after passing to a finite cover if necessary ( which does not affect the ratio of @xmath73 to @xmath87 ) we can assume that the boundary components @xmath69 of @xmath55 come in _ pairs _ with equal degrees @xmath83 . let @xmath88 be the least common multiple of the @xmath89 . define a function @xmath90 from the set of boundary components of @xmath55 to @xmath91 as follows . divide the set of components into pairs @xmath92 for which @xmath93 , and define @xmath94 and @xmath95 . then @xmath96 , so @xmath90 extends to a surjective homomorphism from @xmath97 to @xmath91 . if @xmath98 is the cover associated to the kernel , then each component of @xmath99 maps to @xmath53 with degree @xmath100 . pairs of components whose degrees have opposite sign can be glued up ( which does not affect @xmath73 or @xmath87 ) until all remaining components have degrees with the same ( positive ) sign . the resulting surface @xmath101 satisfies the conclusion of the lemma . for more details , see @xcite , 2.1 . a map @xmath57 rationally bounding @xmath53 is _ extremal _ if @xmath55 has no disks or closed components , and there is an equality @xmath102 . notice that @xmath12 must be rational for an extremal surface to exist . [ extremal_lemma ] an extremal surface is @xmath103-injective . let @xmath57 be extremal . suppose there is some essential immersed loop @xmath104 in @xmath55 for which @xmath105 is null - homotopic . since surface groups are lerf ( see @xcite ) there is a finite cover @xmath106 of @xmath55 to which @xmath104 lifts as an embedded loop . let @xmath107 lift the map @xmath71 ( i.e. @xmath108 is the composition of @xmath71 with the covering projection @xmath109 ) . note that @xmath110 . let @xmath111 denote an embedded preimage of @xmath112 in @xmath106 . since @xmath111 is nullhomotopic under @xmath108 , we can surger @xmath106 along @xmath111 to produce a surface @xmath101 with @xmath113 but with @xmath114 . but this contradicts the hypothesis that @xmath55 is extremal . lemma [ monotone_surface_lemma ] shows that if there is an extremal surface for @xmath53 , there is a monotone extremal surface . a brief discussion of quasimorphisms , though not strictly logically necessary for the results of this paper , nevertheless provides some useful context and explains an important connection with the theory of bounded cohomology . [ quasimorphism_definition ] let @xmath8 be a group . a _ quasimorphism _ on @xmath8 is a function @xmath115 for which there exists some least non - negative constant @xmath116 called the _ defect _ , so that the following inequality holds @xmath117 for all @xmath30 . a quasimorphism is _ homogeneous _ if @xmath118 for all @xmath119 and all @xmath31 . in words , a quasimorphism on a group is a homomorphism up to a bounded error . a quasimorphism is a genuine homomorphism if and only if the defect is zero . the set of quasimorphisms ( resp . homogeneous quasimorphisms ) on @xmath8 admits the structure of a real vector space . denote the vector space of quasimorphisms on @xmath8 by @xmath120 , and the vector space of homogeneous quasimorphisms by @xmath121 . [ exact_sequence_prop ] let @xmath8 be a group . there is an exact sequence @xmath122 where @xmath123 denotes _ bounded cohomology _ ( with real coefficients ) , and @xmath18 denotes the coboundary on group @xmath1-cochains . see @xcite for an introduction to bounded cohomology . note that when @xmath124 and @xmath125 are finite dimensional ( as is the case when @xmath8 is finitely presented ) then @xmath18 is fredholm ( with respect to natural banach norms on @xmath126 and @xmath3 ) . there is a kind of duality , called _ bavard duality _ , relating commutator length and quasimorphisms . the most concise statement of this duality is the following : [ duality_theorem ] let @xmath8 be a group . for any @xmath9 $ ] there is an equality @xmath127 note that one should restrict attention to @xmath128 since if @xmath90 is a homomorphism , then both @xmath129 for @xmath9 $ ] , and @xmath130 . the functions @xmath15 and @xmath16 can be extended to finite sums as follows . [ general_scl_definition ] let @xmath8 be a group . let @xmath131 be elements in @xmath8 ( not necessarily distinct ) . define @xmath132 and define @xmath133 note that @xmath15 and @xmath16 depend only on the individual conjugacy classes of the summands , and are commutative in their arguments . geometrically , if @xmath51 is a cw complex with @xmath52 and @xmath134 are loops representing the conjugacy classes of @xmath135 respectively , then @xmath136 is the smallest genus surface @xmath55 with @xmath75 boundary components @xmath69 for which there is a map @xmath57 wrapping each @xmath69 around @xmath137 . it is worth remarking that the function @xmath138 is not subadditive , but that the `` corrected '' function @xmath139 is subadditive , and therefore the limit exists in definition [ general_scl_definition ] , providing @xmath15 is not infinite . let @xmath55 be a compact , connected , oriented surface . let @xmath51 be a topological space , and @xmath140 for @xmath141 continuous loops . let @xmath57 be such that there is a commutative diagram @xmath142 where @xmath65 is the inclusion map . suppose there is an integer @xmath66 so that @xmath67 = n(s)[\coprod_i s^1]$ ] in @xmath68 . then say @xmath57 is _ admissible_. informally , @xmath66 is the common degree with which @xmath58 wraps around each loop @xmath137 . with this notation , the generalization of lemma [ euler_lemma ] to arbitrary sums is as follows : [ general_euler_lemma ] let @xmath51 be a connected cw complex with @xmath143 . further , let @xmath134 be loops in @xmath51 in free homotopy classes corresponding to conjugacy classes @xmath135 . then @xmath144 where the infimum is taken over all admissible maps @xmath57 wrapping @xmath58 around each @xmath137 with degree @xmath66 . the proof is almost identical to that of lemma [ euler_lemma ] . moreover , one may restrict attention to monotone admissible maps , by the argument of lemma [ monotone_surface_lemma ] . for details , see @xcite , 2.6.1 . the function @xmath16 as above can be extended to integral group @xmath1-chains . from lemma [ general_euler_lemma ] follow equalities @xmath145 and @xmath146 valid for any @xmath147 and any non - negative integer @xmath56 . hence one may define @xmath148 for any integers @xmath83 ( not necessarily non - negative ) and elements @xmath149 and observe that the result is well - defined on integral group @xmath1-chains , and is subadditive under addition of chains . consequently , @xmath16 can be extended to rational chains by linearity , and to real chains by continuity . see @xcite , 2.6.1 . denote the vector space of real ( group ) @xmath1-chains on @xmath8 by @xmath150 and @xmath1-boundaries by @xmath151 ( or just @xmath152 and @xmath25 if @xmath8 is understood ) . the expression @xmath153 is finite if and only if @xmath154 . bavard duality holds in the broader context of arbitrary real @xmath1-boundaries , and with essentially the same proof . the statement is : [ general_duality_theorem ] let @xmath8 be a group . for any finite set of elements @xmath149 and numbers @xmath155 for which @xmath154 there is an equality @xmath156 any homogeneous quasimorphism vanishes identically on any chain of the form @xmath28 or @xmath29 for @xmath157 and @xmath31 . define @xmath27 to be the subspace of @xmath25 spanned by such chains . then @xmath16 descends to a pseudo - norm on @xmath158 ( hereafter denoted @xmath0 ) . when @xmath8 is hyperbolic , @xmath16 is a _ norm _ on @xmath0 ; this is corollary 3.57 from @xcite , restating work of @xcite ( this fact is logically superfluous for the results of this paper ) . in fact , the reader can take lemma [ euler_lemma ] and lemma [ general_euler_lemma ] as the _ definition _ of @xmath16 . this is the point of view we shall take in the sequel . this section contains the proof of the rationality theorem for free groups . the main result is that @xmath16 is a piecewise rational linear function on @xmath159 , for @xmath38 a free group . a similar statement holds for some groups derived in simple ways from free groups . throughout this section , we use lemma [ euler_lemma ] and lemma [ general_euler_lemma ] as an operational definition of @xmath16 . in this section , for convenience , we use some language and basic facts from elementary @xmath5-manifold topology ; for a reference , see @xcite . in the sequel , let @xmath38 denote a free group of some fixed rank and let @xmath27 denote a handlebody of genus equal to @xmath160 . as in figure [ handledec ] ( illustrating the case of @xmath161 ) , we consider a system of compressing disks @xmath162 which decompose @xmath27 into @xmath163 components , each of which retracts down to one of @xmath163 compressing disks @xmath164 . denote the union of the @xmath162 by @xmath165 and the union of the @xmath164 by @xmath166 . @xmath167 at 0 370 @xmath168 at -20 260 @xmath169 at 0 150 @xmath170 at 180 520 @xmath171 at -15 315 @xmath172 at -15 200 @xmath173 at 375 315 @xmath174 at 375 200 @xmath175 at 180 -15 and @xmath176 for @xmath177,title="fig : " ] given a conjugacy class in @xmath38 , we construct a representative loop in the corresponding free homotopy class in @xmath27 of a simple kind . such a representative will be made up of certain kinds of arcs , which we call _ horizontal _ and _ vertical _ , and which are defined as follows . a _ horizontal _ arc is an embedded arc @xmath178 whose image is contained in some @xmath179 . vertical _ arc is an arc @xmath178 which is properly embedded in the complement of @xmath166 , and which intersects some @xmath176 transversely in one point . note that any two horizontal arcs with the same endpoints are homotopic rel . endpoints through horizontal arcs . moreover , any two vertical arcs whose endpoints are contained in the same @xmath179 are properly homotopic through vertical arcs ; call arcs which differ from each other by such homotopies _ equivalent_. [ branch_remark ] if one does not want the psychological convenience of working in a manifold , one can substitute in place of @xmath27 a union of @xmath160 solid tori @xmath180 , each with a marked disk @xmath179 in their boundary , and glue the tori up by identifying the @xmath179 with a single disk @xmath181 by homeomorphisms . the resulting space is a manifold away from the disk @xmath181 . in the case @xmath182 , the two approaches are equivalent . dual to the system @xmath165 of compressing disks there is a graph @xmath183 with one vertex for each component of @xmath184 and one edge for each @xmath162 in @xmath165 . there is an isomorphism @xmath185 . the universal cover @xmath186 of @xmath183 is a tree ( for an introduction to trees in geometric group theory , see @xcite , especially chapter 1 ) . every element in @xmath38 acts on the tree @xmath186 with a unique axis ; this axis covers a closed loop in @xmath183 . each arc in @xmath183 corresponds to a unique equivalence class of vertical arc in @xmath27 . so to each conjugacy class of element @xmath187 is associated a ( cyclically ordered ) sequence of ( equivalence classes of ) vertical arcs in @xmath27 . if two consecutive vertical arcs are on opposite sides of some @xmath179 in @xmath166 , they can be homotoped ( rel . endpoints ) until they have a common endpoint in @xmath179 , and their union is transverse to @xmath179 at that point . if two consecutive vertical arcs are on the same side of some @xmath179 in @xmath166 , their endpoints in @xmath179 can be joined by a horizontal arc in @xmath179 . in this way , a conjugacy class in @xmath38 determines a loop @xmath53 in @xmath27 , unique up to equivalence , made up of vertical and horizontal arcs . say that such a @xmath53 is in _ bridge position_. in @xmath6 . there are eight vertical arcs ( one for each letter ) and two horizontal arcs ( one for each `` double letter '' ) . this loop happens to be embedded in @xmath27 , but the isotopy class of the loop is not significant , just its homotopy class . ] if @xmath188 is a family of elements in @xmath38 , then a family of loops @xmath134 in @xmath27 representing the conjugacy classes of the @xmath22 is in bridge position if each loop individually is in bridge position . see figure [ bridgeloop ] for an example of a loop in bridge position . since each horizontal or vertical arc has distinct endpoints , a circle in bridge position decomposes into at least two arcs . we may ( and do ) assume without loss of generality that any family of circles in bridge position is actually embedded in @xmath27 . however , it is very important to note that the _ isotopy class _ of the loop is _ not _ important ; all that matters is its _ homotopy class_. a system @xmath165 and @xmath166 of compressing disks for a handlebody determines a generating set for @xmath38 as a groupoid , whose generators are equivalence classes of vertical arcs . if we use in place of a handlebody the space described in remark [ branch_remark ] , vertical arcs correspond to generators for @xmath38 as a group . now , let @xmath187 be a conjugacy class in @xmath189 $ ] , and let @xmath53 be a loop in bridge position representing @xmath10 . let @xmath190 be the union of the endpoints of the horizontal and vertical arcs in @xmath53 ( remember that for convenience we have assumed that @xmath53 is embedded ) . note that @xmath191 is a finite set . let @xmath192 be a map of a surface whose ( possibly multiple ) boundary components wrap some number of times around @xmath53 . by lemma [ monotone_surface_lemma ] , we may assume that @xmath71 is monotone , so that each component of @xmath58 maps over @xmath53 with positive degree . this is not strictly necessary for what follows , but it simplifies some arguments . we will gradually adjust @xmath71 and @xmath55 , never changing @xmath66 or increasing @xmath73 , until the end result is built up from a finite number of simple `` pieces '' . since by hypothesis @xmath71 is monotone , first adjust @xmath71 by a homotopy so that the restriction of @xmath71 to each boundary component is an orientation - preserving covering map @xmath77 . next put @xmath71 into general position ( rel . @xmath58 ) with respect to the disks @xmath165 . since @xmath71 is in general position , and since @xmath53 is transverse to @xmath165 , the preimage @xmath193 is a union of disjoint embedded loops and proper arcs . since @xmath71 restricted to @xmath58 is an immersion , every arc of @xmath193 is essential . let @xmath104 be a component of @xmath193 in @xmath55 . if @xmath104 is an ( innermost ) inessential loop , then @xmath104 can be pushed off @xmath165 by a homotopy , reducing the number of components of @xmath193 . if @xmath104 is an essential loop , then @xmath55 can be compressed along @xmath104 , and the compressing disks can be mapped to the component of @xmath165 containing @xmath105 . this does not change @xmath66 but reduces @xmath73 . after finitely many such compressions and homotopies , we can assume that @xmath193 is a union of disjoint embedded proper essential arcs . in short , we have proved the following `` preparation lemma '' : [ tight_lemma ] let @xmath192 monotone be given . then after possibly replacing @xmath194 with @xmath195 satisfying @xmath196 and @xmath197 , we can assume that @xmath193 is a union of disjoint embedded proper essential arcs . note that distinct components of @xmath193 might be ( and typically will be ) parallel in @xmath55 , especially for complicated @xmath53 . let @xmath198 be a regular neighborhood of @xmath193 in @xmath55 , so that @xmath198 consists of a union of disjoint embedded proper essential rectangles . by general position we can take @xmath198 to be equal to the inverse image ( under @xmath199 ) of an open tubular neighborhood @xmath200 of @xmath165 . now , there is a deformation retraction of pairs @xmath201 to @xmath202 . this deformation retraction can be extended to a map @xmath203 which restricts to a homotopy equivalence of pairs from @xmath204 to @xmath205 . so after composing @xmath71 with such a retraction @xmath206 we get a new map , which by abuse of notation we also denote by @xmath71 , homotopic to the old map @xmath71 , such that @xmath207 . each component @xmath208 of @xmath209 either maps by @xmath71 to a point in @xmath191 , or to a horizontal arc in @xmath53 . in the first case , we collapse @xmath208 to a point in @xmath55 by a homotopy equivalence , so we assume without loss of generality that every arc of @xmath209 is a horizontal arc . see figure [ retract ] . to @xmath181 , the preimage @xmath210 consists of a union @xmath198 of rectangles . ] let @xmath211 be a component of @xmath212 . then @xmath71 maps @xmath211 to some component @xmath179 of @xmath181 . if @xmath211 is not a disk , then it contains an essential simple loop @xmath53 . since @xmath179 is a disk , @xmath71 maps @xmath53 to a homotopically trivial loop in @xmath179 , so we can compress @xmath55 along @xmath53 , mapping the compressing disks to @xmath179 , to get a new surface @xmath101 with @xmath196 and @xmath197 . so without loss of generality we can assume that each component @xmath211 is a disk . in fact , @xmath211 inherits the structure of a polygon , whose edges are arcs of the boundary of components of @xmath198 , and horizontal arcs . the vertices of @xmath211 map by @xmath71 to @xmath191 . [ embedded_vertices ] with notation as above , after possibly replacing @xmath194 with @xmath195 satisfying @xmath196 and @xmath197 we can assume that for each polygon @xmath211 of @xmath212 the image of the vertices of @xmath211 under @xmath71 are _ distinct _ elements of @xmath191 . suppose that @xmath211 is a polygon mapping under @xmath71 to the disk @xmath181 , and let @xmath213 be distinct vertices of @xmath211 mapping to the same point @xmath214 . let @xmath208 be an embedded proper arc in @xmath211 from @xmath215 to @xmath216 . let @xmath98 be obtained from @xmath55 by identifying @xmath215 to @xmath216 , then cut @xmath98 along the loop @xmath208 , and glue in two disks . the restriction of @xmath71 to @xmath217 extends to these disks , since @xmath181 is contractible . the result is a new surface @xmath101 as in the statement of the lemma . the argument of lemma [ embedded_vertices ] may be summarized by saying that @xmath101 is obtained from @xmath55 by first adding an oriented @xmath1-handle from @xmath215 to @xmath216 which maps trivially under @xmath71 , and then compressing the trivial embedded loop that runs over the core of this @xmath1-handle . adding a @xmath1-handle increases @xmath73 by @xmath1 , but doing a compression reduces it by @xmath4 . these operations are uniquely defined up to homotopy , and keep the surface oriented and the map monotone . notice that @xmath55 and @xmath101 might have different numbers of boundary components . call the result of this whole operation a _ boundary compression_. the core of each component of @xmath198 runs between two vertical arcs of @xmath53 , after mapping by @xmath71 . there are only finitely many combinatorial types of such rectangles ; an upper bound is the number of pairs of vertical arcs in @xmath53 which intersect the same component of @xmath165 . in fact , since @xmath55 is monotone , the two ( oriented ) arcs of @xmath53 in the boundary of each rectangle must run over a handle of @xmath27 in opposite directions . similarly , there are only finitely many polygon types @xmath211 , since each is determined by a cyclically ordered subset of @xmath218 for some @xmath179 . it is convenient to introduce the language of branched surfaces in what follows . we give cursory definitions below to standardize terminology , but the reader who is not familiar with branched surface should consult e.g. @xcite chapter 6 , 6.3 , or @xcite 1 for precise definitions . a _ branched surface _ is a finite , smooth @xmath4-complex obtained from a finite collection of smooth surfaces by gluing compact subsurfaces . the set of non - manifold points of a branched surface @xmath219 is called the _ branch locus _ , and denoted @xmath220 . a branched surface in which the branch locus is a @xmath1-manifold is called a _ simple branched surface_. in this paper we are only interested in simple branched surfaces . one can also consider branched surfaces with boundary ; the boundary of a branched surface is a _ train - track _ ( see e.g. @xcite ) . the components of @xmath221 are called the _ sectors _ of the branched surface . a branched surface is oriented if the sectors can be oriented in such a way that the orientations are compatible along @xmath220 . in a simple branched surface , several distinct sectors locally bound a component of the branch locus from either side ; see figure [ local_models ] for an example of the local model . a _ weight _ is a linear function @xmath222 from the sectors of @xmath219 to @xmath223 satisfying the _ gluing conditions _ : along each component of @xmath220 , the weights of the sectors on each side sum to the same value . a branched surface @xmath219 has a well - defined tangent bundle , even along the branch locus , so it makes sense to say that a map from a surface into @xmath219 is an immersion . a proper immersion @xmath224 from an oriented surface to an oriented branched surface @xmath219 is said to be a _ carrying map _ , and @xmath55 is said to be _ carried _ by @xmath219 . a carrying map determines a weight @xmath222 , where @xmath225 is the cardinality of @xmath226 for @xmath74 a point in the interior of a sector @xmath227 . given a non - negative integral weight @xmath222 , we say that a carrying map @xmath224 _ realizes _ @xmath222 if the weight associated to @xmath55 is @xmath222 . note that @xmath55 is not uniquely determined by its weight in general . for an abstract branched surface , not every non - negative integral weight is realized , but for a _ simple _ branched surface , every non - negative integral weight is realized . this is a somewhat subtle point , so we make a few clarifying remarks . let @xmath222 be a weight . an edge @xmath228 of @xmath220 has sectors @xmath229 on one side , and @xmath230 on the other . take @xmath231 copies of @xmath232 for each @xmath233 . the gluing condition along @xmath228 is the equality @xmath234 hence there is a bijection between the set of copies of the @xmath229 and the set of copies of the @xmath230 . after choosing such a bijection , these copies can be glued up along copies of @xmath228 to build the surface @xmath55 near @xmath228 . for a general branched surface , there is a holonomy problem at the triple points of @xmath220 : around each triple point , the holonomy must be trivial , or else the surface constructed will map to @xmath219 with branch points . but if @xmath219 is simple , there are no such triple points , and the holonomy problem goes away . this concludes our summary of the theory of branched surfaces . if @xmath219 is an _ embedded _ branched surface ( possibly with triple points ) in a @xmath5-manifold @xmath235 , the local transverse order structure canonically solves the holonomy problem : associated to each weight there is a unique _ embedded _ surface contained in a tubular neighborhood of @xmath219 . the requirement that the surface be embedded makes the bijections along edges canonical , and therefore the holonomy around a triple point is necessarily trivial . a similar phenomenon occurs in normal surface theory : a vector of weights satisfying the gluing equalities and inequalities determines a unique embedded normal surface . when one tries to do immersed normal surface theory in @xmath5-manifolds , the holonomy problem reasserts itself and the situation becomes very tricky ; see @xcite for a discussion of some of the phenomena which arise in the theory of immersed normal surfaces . we construct a branched surface @xmath219 as follows . @xmath219 is obtained by taking a disjoint copy of each combinatorial type of ( oriented ) marked polygon @xmath211 , and a copy of each combinatorial type of ( oriented ) rectangle @xmath236 as above , and gluing them along their common oriented edges . there is a unique way to choose the smooth structure along @xmath220 compatibly with the orientations . notice that each marked polygon is glued to exactly one rectangle along each non - horizontal edge , but each rectangle is typically glued to several polygons along each of two edges . hence a weight is determined by its values on polygon sectors ( cf . quadrilateral co - ordinates in normal surface theory ) . the resulting branched surface is special in a few ways , of which we take note : 1 . it is _ oriented _ , since it is obtained by gluing oriented pieces in a manner compatible with their orientations . the branch locus consists of a finite union of embedded arcs along which a rectangle is glued to several possible polygons . in particular , the branch locus contains no triple points and is _ simple_. 3 . the branched surface has boundary which is an oriented train - track . every surface @xmath55 carried by @xmath219 has boundary @xmath58 which is carried by @xmath237 . 4 . the branched surface admits a tautological immersion @xmath238 , by choosing a map for each rectangle and polygon type and gluing these maps together along the edges . the restriction of @xmath239 to each train - track component of @xmath237 is an oriented immersion to @xmath53 for each polygon @xmath211 let @xmath240 denote the number of edges , and @xmath241 the number of horizontal edges . note that @xmath242 and @xmath243 , since a pair of adjacent edges of a polygon can not both be horizontal . furthermore , if @xmath244 then @xmath245 . let @xmath246 be decomposed into polygons and rectangles . each rectangle contributes @xmath247 to @xmath86 , and each polygon @xmath211 contributes @xmath248 this can be seen by giving each rectangle and polygon a singular foliation tangent to @xmath237 and transverse to @xmath220 , and using the hopf - poincar formula . in particular , the contribution of each polygon to @xmath86 is _ nonpositive _ , and therefore @xmath85 is a _ linear function _ of the number and kinds of rectangles and polygons that make it up . in summary , the results of the previous sections show that for any monotone @xmath192 , after possibly replacing @xmath55 by a new surface with smaller @xmath73 and the same @xmath66 , we can homotope @xmath71 so that it factors through a carrying map to @xmath219 , and determines a non - negative integral weight . conversely , since @xmath219 is simple , every non - negative integral weight on @xmath219 corresponds to a surface @xmath55 as above ( possibly not unique ) , and @xmath73 is a linear function of the values of @xmath222 on the branches of @xmath219 . notice that @xmath73 depends only on @xmath222 , though the topology of @xmath55 might not . let @xmath249 denote the ( finite dimensional ) real vector space of weights on @xmath219 , and @xmath250 the subspace of nonnegative weights . moreover , let @xmath251 ( resp . @xmath252 ) denote the subset of weights with rational ( resp . non - negative rational ) coefficients . then @xmath73 is a linear function on @xmath249 which is non - negative on @xmath250 , and takes rational values on @xmath251 . we abbreviate this by saying that @xmath73 is a _ rational linear function_. let @xmath253 be the subspace of @xmath249 spanned by @xmath250 ( note that this subspace is not necessarily equal to @xmath249 ) . then @xmath253 is a rational subspace of @xmath249 , since it is spanned by finitely many rational vectors . there is a rational linear function @xmath254 defined as follows . given a positive integral weight @xmath222 , let @xmath55 be a surface carried by @xmath219 associated to @xmath222 . then define @xmath255 and extend by linearity to @xmath253 . notice that @xmath256 does not depend on the choice of @xmath55 , but only on the induced weight on the train - track @xmath237 . the inverse @xmath257 is a rational polyhedron . moreover , by construction , there is an equality @xmath258 since @xmath73 is non - negative on @xmath250 , this infimum is _ realized _ , and the set of points which realize the infimum is itself a rational polyhedron . if @xmath222 is an integral weight in the projective class of an element of this polyhedron and @xmath55 is a surface carried by @xmath219 with weight @xmath222 , then @xmath55 is an extremal surface for @xmath10 . from this discussion we can conclude the following : 1 . @xmath39 for every @xmath40 $ ] 2 . an extremal surface exists for every @xmath10 3 . there is an algorithm to calculate @xmath16 and to construct all monotone extremal surfaces for every @xmath40 $ ] if @xmath192 is extremal , and there is an essential embedded loop @xmath104 in @xmath55 such that @xmath105 is freely homotopic to a power of @xmath53 , we may cut open @xmath55 along @xmath104 to produce a new extremal ( but not necessarily monotone ) surface . in fact , very little is required to extend the results of the last few sections to finite dimensional vector spaces of @xmath259 . let @xmath260 be represented by loops @xmath261 in bridge position in @xmath27 . denote @xmath262 . fix an orientation on each @xmath137 . a monotone surface @xmath192 whose boundary wraps some positive number of times around the various @xmath137 can be compressed , boundary compressed and homotoped until it is composed of a union of rectangles and polygons carried by a fixed simple branched surface @xmath219 as above . there is a rational linear map @xmath263 . let @xmath264 be the kernel of the inclusion map @xmath265 . let @xmath266 be the intersection of @xmath264 with the orthant spanned by non - negative combinations of the @xmath267 $ ] in @xmath268 . given @xmath269 corresponding to a collection of non - negative weights on the @xmath22 , we have an equality @xmath270 and therefore @xmath16 is a piecewise rational linear function on @xmath266 . there are finitely many orthants of this kind , corresponding to choices of orientation on each @xmath137 , so @xmath16 is piecewise rational linear on @xmath264 . putting this together proves our main result : let @xmath38 be a free group . 1 . @xmath39 for all @xmath40$].[rational_bullet ] 2 . every @xmath41 $ ] rationally bounds an extremal surface ( in fact , every rational chain @xmath42 in @xmath0 rationally bounds an extremal surface)[extremal_bullet ] 3 . the function @xmath16 is piecewise rational linear on @xmath0.[polyhedron_bullet ] 4 . there is an algorithm to calculate @xmath16 on any finite dimensional rational subspace of @xmath0 . note that bullet ( [ rational_bullet ] ) is a special case of bullet ( [ polyhedron_bullet ] ) . say that a group @xmath8 is pql ( pronounced `` pickle '' ) if @xmath16 is piecewise rational linear on @xmath271 . the pql property is inherited by supergroups of finite index . this follows in a straightforward way from the following lemma , which relates @xmath16 in groups and in finite index subgroups . [ finite_index_formula ] let @xmath8 be a group , and @xmath27 a subgroup of @xmath8 of finite index . let @xmath51 be a cw complex with @xmath52 , and let @xmath272 be a covering space with @xmath273 . let @xmath135 be elements in @xmath8 , and for each @xmath233 , let @xmath137 be a loop in @xmath51 representing the conjugacy class of @xmath22 . let @xmath274 be the preimages of the @xmath137 in @xmath272 , and @xmath275 the corresponding conjugacy classes in @xmath27 . then @xmath276 if @xmath57 is a map of a surface whose boundary maps to the @xmath137 , then @xmath55 admits a finite index cover @xmath106 such that @xmath71 lifts to @xmath277 . conversely , given @xmath278 with boundary mapping to the @xmath279 , the composition of @xmath71 with the covering projection @xmath280 takes @xmath58 to @xmath53 . now apply lemma [ euler_lemma ] and lemma [ general_euler_lemma ] . let @xmath235 be a non - compact seifert - fibered @xmath5-manifold . then @xmath281 has the pql property . since @xmath235 is noncompact , there is a finite index subgroup @xmath27 of @xmath281 of the form @xmath282 where @xmath38 is free . since @xmath283 is amenable , a theorem of bouarich ( see @xcite or @xcite , 2.4 ) implies that @xmath284 is an isomorphism , and @xmath285 . hence @xmath27 is pql by theorem [ general_duality_theorem ] . but then @xmath281 is pql by lemma [ finite_index_formula ] . the braid group @xmath286 is isomorphic to @xmath287 where @xmath264 is the trefoil knot ( of either handedness ) . this group is a central @xmath283 extension of @xmath288 , which admits a free subgroup of finite index . hence @xmath286 has the pql property . does every @xmath5-manifold group have the pql property ? in this section we discuss the implementation of the algorithm implicit in [ free_group_section ] , and study an explicit example . in order to describe the algorithm in a uniform way for free groups of any rank , it is convenient to use the formalism described in remark [ branch_remark ] , in which vertical arcs in @xmath183 correspond to elements in a free generating set for @xmath38 . if one is more comfortable working with systems of compressing disks in a handlebody , one must work with _ groupoid _ generators for @xmath38 associated to the system of vertical arcs in a splitting . in the case of a free group of rank @xmath4 , both formalisms agree ; the cautious reader may prefer to stick to this case in what follows . fix a free symmetric generating set @xmath55 for a free group @xmath38 , and let @xmath289 be an integral chain in @xmath159 , expressed as a formal sum of cyclically reduced words in the generating set . we denote the generators of @xmath38 by @xmath290 and their inverses by @xmath291 . [ arcs_and_polygons ] a _ letter _ is a specific character in a specific word ; its _ value _ is the element of the generating set that it represents . an _ arc _ is an ordered pair @xmath292 of letters whose values are inverse in @xmath38 . a _ polygon _ is a cyclically ordered list of distinct arcs @xmath293 so that for each @xmath233 , the last letter of @xmath294 immediately precedes the first letter of @xmath295 in some ( cyclic ) word @xmath296 ( indices @xmath233 and @xmath297 are taken mod @xmath75 ) . the _ length _ of a polygon is the number of arcs appearing in the list . in the language of [ free_group_section ] , each oriented rectangle has four edges , two of which ( the `` free edges '' ) correspond to vertical edges in @xmath183 , and two of which ( the `` glued edges '' ) are glued up to some polygon . each vertical edge in @xmath183 corresponds to a letter in some @xmath298 . the two glued edges of a rectangle determine two arcs @xmath104 and @xmath299 consisting of the same pair of letters in opposite orders ; i.e. if @xmath300 then @xmath301 . a polygon ( in the sense of [ free_group_section ] ) has edges that are glued up to edges of rectangles , and edges on horizontal edges of @xmath183 . horizontal edges are of purely psychological value ; they do not contribute to @xmath302 . so to each such polygon there is associated a polygon ( in the sense of definition [ arcs_and_polygons ] ) consisting of a cyclically ordered list of arcs . the condition that the arcs making up a polygon are distinct is the analog of the condition that each polygon ( in the sense of [ free_group_section ] ) has vertices mapping to distinct points of @xmath191 . let @xmath211 denote the real vector space spanned by the set of polygons . let @xmath303 denote the vector space spanned by the set of arcs modulo the relation that @xmath304 . there is a linear map @xmath305 sending a polygon to the formal sum of the arcs that make it up . let @xmath249 denote the kernel of @xmath306 , and @xmath250 the cone of non - negative vectors in @xmath249 . the relations @xmath307 and the condition @xmath308 express the gluing equations for @xmath219 in terms of weights on polygon sectors . the spaces @xmath249 and @xmath250 are naturally isomorphic to the spaces with the same names described in [ free_group_section ] . a polygon of length @xmath75 contributes @xmath309 to @xmath310 . the condition that a formal sum of polygons has boundary equal to the chain @xmath42 is a further list of linear conditions , one for each word in @xmath42 . minimizing the objective function @xmath310 is a linear programming problem , which can be solved in a number of ways . let @xmath311 ( remember that @xmath312 and @xmath313 ) . there are @xmath314 arcs , one for each ordered pair of letters in @xmath222 with opposite values . labeling the letters of @xmath222 by integers from @xmath247 to @xmath315 , and using the shorthand @xmath316 and @xmath317 , these arcs are @xmath318 @xmath319 there are 625 polygons , 43 of length 4 or less ( since the letters of @xmath222 alternate between one of @xmath320 and one of @xmath321 , every polygon has even length ) , including @xmath322 let @xmath323 denote the vector space of formal linear combinations of polygons , and let @xmath324 for @xmath325 denote the components of a vector in @xmath211 . for each @xmath233 , let @xmath326 denote the length of polygon @xmath233 . compatibility of gluing along rectangles ( i.e. @xmath327 above ) imposes one equation of the form @xmath328 for each pair of arcs @xmath329 where the polygons of type @xmath330 are those that contain the arc @xmath331 , and polygons of type @xmath332 are those that contain the arc @xmath333 ( note that a polygon type might contain both @xmath331 and @xmath333 or neither ) . there are half as many equations of this kind as arcs , hence @xmath334 equations . restricting to geometrically sensible answers imposes the conditions @xmath335 . the condition that the boundary of a ( formal ) surface corresponding to a weight represents @xmath336 $ ] in homology is the equation @xmath337 subject to this list of constraints , which determine a compact convex polyhedron in @xmath338 , we minimize the objective function @xmath339 this linear programming problem can be solved using exact arithmetic by the gnu package glpsol @xcite or masashi kiyomi s program exlp @xcite using dantzig s simplex method ( see @xcite ) , and gives the answer @xmath340 . an extremal solution found by the simplex method is always a vertex , which can be projectively represented by an extremal surface . one such extremal surface found by this method is determined by the identity @xmath341 \cdot [ ababa , babababababba ] \\ \cdot [ bababababa , aabaab ] = a(babababababa)^3a\end{gathered}\ ] ] exhibiting the cube of @xmath342 ( i.e. the inverse of a cyclic conjugate of @xmath343 ) as a product of three commutators . since extremal surfaces are @xmath103-injective , the group generated by @xmath344 is a free subgroup of @xmath6 of rank @xmath345 , equal to the image of @xmath103 of a genus @xmath5 surface with one puncture under an injective homomorphism . this fact may be verified independently e.g. by using stallings folding , and gives an independent check of the validity of the calculation . as was discussed in the introduction , this example answers bavard s question in the negative . other simple examples are @xmath346 and @xmath347 several developments building on this material have taken place between the time this paper was submitted and was accepted for publication . it would be inappropriate to go into too much detail , but for the convenience of the reader , we summarize some of the most interesting points . firstly , the algorithm described above has been improved to run in _ polynomial time_. an account of this improvement is described in @xcite , 4.1.78 . the program scallop ( source code available at @xcite ) implements this algorithm , and can be used to compute @xmath16 in @xmath6 on words of length @xmath348 60 . experiments using scallop reveal additional structure in the @xmath16 spectrum of a free group , partially explained in a forthcoming paper @xcite . in particular , it is possible to prove rigorously that in any nonabelian free group , the image of @xmath16 contains nontrivial accumulation points , and takes on values with any denominator . secondly , the polyhedral structure on the @xmath16 norm unit ball is studied in @xcite , and it is shown that every realization of a free group @xmath38 as @xmath97 where @xmath55 is an oriented surface with boundary is associated to a codimension one face of the boundary of the @xmath16 norm unit ball in @xmath0 . thirdly , the extremal surfaces guaranteed in bullet ( [ extremal_bullet ] ) of the rationality theorem are exploited to construct closed surface subgroups in certain graphs of free groups amalgamated along cyclic subgroups . this is studied in @xcite , and further generalized in @xcite . while writing this paper i was partially funded by nsf grant dms 0707130 . i would like to thank roger alperin , nathan dunfield , dieter kotschick , lars louder , jason manning , bill thurston , dongping zhuang , the anonymous referees , and the members of the dsemi at the tokyo institute of technology , especially shigenori matsumoto . i presented an incorrect argument , purporting to prove the main result of this paper , at the dsemi in april 2007 . matsumoto asked a simple - sounding question about branch points , which turned out to be a crucial detail that i had overlooked . understanding this detail was the key to obtaining the results in this paper . i would also like to explicitly thank jason manning for a number of conversations that substantially increased my confidence in the experimental results discussed in [ algorithm_section ] . 99 c. bavard , _ longueur stable des commutateurs _ , enseign . math . ( 2 ) , * 37 * , 1 - 2 , ( 1991 ) , 109150 a. bouarich , _ suites exactes en cohomologie borne relle des groupes discrets _ , c. r. acad . * 320 * ( 1995 ) , no . 11 , 13551359 r. brooks , _ some remarks on bounded cohomology _ , riemann surfaces and related topics : proceedings of the 1978 stony brook conference ( suny stony brook ny 1978 ) ann . math . stud . * 97 * , princeton univ . press 1981 , 5363 d. calegari , _ foliations and the geometry of @xmath5-manifolds _ , oxford mathematical monographs , oxford univ . press , oxford , 2007 d. calegari , _ surface subgroups from homology _ , geom . * 12 * ( 2008 ) , 19952007 d. calegari , _ faces of the scl norm ball _ , geom . top . * 13 * ( 2009 ) , 13131336 d. calegari , _ scl , sails and surgery _ , preprint , in preparation d. calegari , scallop , computer program , available from d. calegari and k. fujiwara , _ stable commutator length in word hyperbolic groups _ , groups , geom . dyn . to appear g. dantzig , _ linear programming and extensions _ , princeton univ . press , princeton , 1963 d. gabai , _ foliations and the topology of @xmath5-manifolds _ , j. diff . geom . * 18 * ( 1983 ) , no . 3 , 445503 c. gordon and h. wilton , _ on surface subgroups of doubles of free groups _ , preprint , arxiv:0902.3693 m. gromov , _ volume and bounded cohomology _ , inst . hautes tudes sci . 56 ( 1982 ) , 599 m. gromov , _ asymptotic invariants of infinite groups _ , london math . lecture note ser . * 182 * cambridge univ . press , cambridge , 1993 j. hempel , _ @xmath5-manifolds _ , ann . math . stud . * 86 * , princeton univ . press , princeton , 1976 m. kiyomi , exlp , computer program , available from a. makhorin , glpsol , computer program , available from http://www.gnu.org l. mosher and u. oertel , _ two - dimensional measured laminations of positive euler characteristic _ , quart . j. math . * 52 * ( 2001 ) , 195216 r. penner with j. harer , _ combinatorics of train tracks _ , ann . math . stud . * 125 * , princeton univ . press , princeton , 1992 u. oertel , _ homology branched surfaces : thurston s norm on @xmath349 _ , low - dimensional topology and kleinian groups ( coventry / durham , 1984 ) , 253272 london math . soc . lecture notes ser . * 112 * cambridge univ . press , cambridge , 1986 r. rannard , _ computing immersed normal surfaces in the figure - eight knot complement _ , experiment . * 8 * ( 1999 ) , no . 1 , 7384 p. scott , _ subgroups of surface groups are almost geometric _ , j. london math . ( 2 ) * 17 * ( 1978 ) , no . 3 , 555565 j .- serre , _ trees _ , springer monographs in mathematics ( corrected 2nd printing , trans . j. stillwell ) , springer - verlag , berlin , 2003 w. thurston , _ a norm for the homology of @xmath5-manifolds _ , mem . soc . * 59 * ( 1986 ) , no . 339 , i - vi and 99 - 130 d. zhuang , _ irrational stable commutator length in finitely presented groups _ , dyn . 2 ( 2008 ) no . 3 , 499507

a multiple - antenna amplify - and - forward two - hop interference network with multiple links and multiple relays is considered . we optimize transmit precoders , receive decoders and relay af matrices to maximize the achievable sum rate . under per user and total relay sum power constraints , we propose an efficient algorithm to maximize the total signal to total interference plus noise ratio ( tstinr ) . computational complexity analysis shows that our proposed algorithm for tstinr has lower complexity than the existing weighted minimum mean square error ( wmmse ) algorithm . we analyze and confirm by simulations that the tstinr , wmmse and the total leakage interference plus noise ( tlin ) minimization models with per user and total relay sum power constraints can only transmit a single data stream for each user . thus we propose a novel multiple stream tstinr model with requirement of orthogonal columns for precoders , in order to support multiple data streams and thus utilize higher degrees of freedom . multiple data streams and larger multiplexing gains are guaranteed . simulation results show that for single stream models , our tstinr algorithm outperforms the tlin algorithm generally and outperforms wmmse in medium to high signal - to - noise - ratio scenarios ; the system sum rate significantly benefits from multiple data streams in medium to high snr scenarios . mimo af relay network , sum rate maximization , total signal to total interference plus noise ratio , alternating iteration

1211.5884

relays are often used to aid communications , not only to improve the quality of service ( qos ) of the user pairs , which have weak direct links due to poor channel conditions , but also to increase the multiplexing gain of the network @xcite . among various relay transmit schemes , the most effective ones are amplify - and - forward ( af ) , compute - and - forward ( cf ) and decode - and - forward ( df ) . especially af protocol is standardized as layer 1 relaying @xcite , and thus in popular research , because of its simplicity and low complexity . in this paper we consider the multiple link multiple relay network with non - regenerative relaying . there has been many works discussing the optimization of the relay beamforming weights . for single antenna case , @xcite and @xcite study models to solve the optimal relay af weights , where total relay transmit power is minimized under guaranteed signal - to - interference - plus - noise - ratio ( sinr ) requirements . there is also extension to multiple antenna case . in @xcite the authors explore the network with one multiple - antenna relay , and according to various relay af matrix schemes , proposes irc flexcobf " algorithm for transmit beamforming matrices . for the networks with one user pair and parallel relays , @xcite discusses the joint optimization of source and relay beamforming with different receiver filters . with the similar mimo relay network as @xcite , @xcite investigates the optimal joint source and relay power allocation to maximize the end - to - end achievable rate . extended to one transmitter , multiple receiver and multiple relay network , @xcite proposes a weighted mean square error minimization ( wmmse ) model to solve the source and relay beamforming matrices with mmse receiving filter . recent work of @xcite is based on general mimo af relay networks with multiple links and multiple relays . the authors provide algorithms to jointly optimize users precoders , decoders and the relay af matrices . total leakage interference plus noise ( tlin ) minimization and wmmse models are proposed , both with per user and total relay transmit power constraints . the idea to construct the wmmse models in @xcite and @xcite are similar . the wmmse model in @xcite is also extended to that with individual user and individual relay power constraints . the precoders , decoders and the relay af matrices are solved alternatively , where each subproblem can achieve its optimal solution . however the algorithm has quite high computational complexity . here we propose different approaches to approximate the system sum rate and derive lower complexity algorithm to solve the corresponding optimization problem . the degrees of freedom ( dofs ) of one network is closely related to its channel capacity . in high signal - to - noise - ratio ( snr ) scenarios , the capacity increases linearly with the number of dofs . the authors in @xcite propose a new technique of interference alignment ( ia ) , to maximize the achievable dofs for mimo networks . such technique optimizes the precoders and decoders , in order to eliminate the network interference and approach the capacity of mimo network . in mimo networks , the ia technique has been deeply investigated @xcite . it is shown in @xcite that , with relays the achievable dofs of the mimo interference network are increased , and the capacity as well as the reliability are improved . in two - hop networks , @xcite studies the feasibility conditions and the algorithms for relay aided ia , restricted on single antenna case . aiming to achieve the maximum dofs of the @xmath0 mimo relay network , @xcite and @xcite study similar technique of aligned interference neutralization to explore the optimal transmission scheme for single antenna and multiple antennas cases , respectively @xcite investigates the ergodic capacity of a class of fading 2-user 2-hop networks with interference neutralization technique . in @xcite the maximum achievable dofs for different kinds of mimo interference channels and mimo multiple hop networks are listed and concluded . for the general @xmath1 mimo relay network , the maximum dofs are only analyzed with restriction to the number of relays . interestingly , we observe by simulations that the algorithms proposed in @xcite all lead to precoders with linearly dependent columns , which result in single transmit data stream corresponding to one dof for each user , regardless of the number of antennas at relay and user nodes . this is an impetus for us to propose multiple data stream models . in our paper , we propose several models for the general mimo relay network , according to different purposes and situations . the general transmit process and system model are introduced in section . in section we set up a total signal to total interference plus noise ratio ( tstinr ) maximization model to approximate the system sum rate , with per user and total relay transmit power constraints . then this tstinr model as well as the tlin and wmmse model in @xcite are extended to those with individual user and individual relay power constraints . also , the computational complexity of our algorithm is analyzed and compared with the wmmse algorithm in @xcite . our proposed algorithm is shown to have lower complexity . furthermore , to achieve more than one data streams for each user , we propose a multiple stream tstinr model in section . compared to the tstinr model in section , additional orthogonal constraints for precoders are added . in all the models , the precoding matrices , decoding matrices and relay beamforming matrices are iterated alternatively . each subproblem is efficiently solved , with sufficient reduction of the objective function in each iteration guaranteed . we provide simulation results in section . since the network model in @xcite is more general than @xcite , we compare our proposed algorithms with those in @xcite . the results indicate that tstinr outperforms tlin generally , and achieves higher sum rate than wmmse in medium to high snr scenarios , for the single stream cases . the system sum rate benefits much from the multiple stream model in medium and high snr scenarios . parts of our work are reported in @xcite and @xcite . compared to them , we add more details of the proposed algorithms and provide detailed proof for all the mentioned theorems . furthermore , we analyze and compare the detailed computational complexity of our proposed algorithm and the wmmse algorithm from @xcite . _ notation _ : lowercase and uppercase boldface represent vectors and matrices , respectively . @xmath2 represents the complex domain . @xmath3 means the real part of scalar @xmath4 . @xmath5 and @xmath6 are the trace and the frobenius norm of matrix @xmath7 , respectively . @xmath8 represents the @xmath9 identity matrix . @xmath10 and @xmath11 represent the set of the user indices @xmath12 and that of relay indices @xmath13 , respectively . and we use @xmath14 to denote the statistical expectation . @xmath15 means the same order amount of @xmath16 . @xmath17 is composed of the eigenvectors of @xmath7 corresponding to its @xmath18 smallest eigenvalues . consider a two - hop interference channel consisting of @xmath19 user pairs and @xmath20 relays as in fig . transmitter @xmath21 , receiver @xmath21 and relay @xmath22 are equipped with @xmath23 , @xmath24 and @xmath25 antennas , respectively , for any @xmath26 , @xmath27 . user @xmath21 wishes to transmit @xmath28 parallel data streams . @xmath29 denotes the transmit signal vector of user @xmath21 , where @xmath30 . due to the poor channel conditions between user pairs , there is no direct links among users . low - complex relays aid to communicate and the af transmit protocol is used . here we assume perfect channel state information ( csi ) is available at a central controller . transmission process includes two time slots . in the first time slot , all sources transmit signals to all relays . relay @xmath22 receives @xmath31 , for all @xmath32 , where @xmath33 is the precoding matrix of user @xmath21 , @xmath34 is the channel coefficient between the transmitter @xmath21 and relay @xmath22 , and @xmath35 with zero mean and variance matrix @xmath36 is the noise at relay @xmath22 . in the second time slot , by the af protocol all relays broadcast to all destinations @xmath37 , for all @xmath27 , where @xmath38 is the beamforming matrix of relay @xmath22 . receiver @xmath21 observes : @xmath39 for all @xmath26 , where @xmath40 is the channel coefficient between relay @xmath22 and receiver @xmath21 , and @xmath41 with zero mean and variance matrix @xmath42 is the noise at receiver @xmath21 . multiplying the decoding matrix @xmath43 , receiver @xmath21 obtains : @xmath44 the right hand side of ( [ eq : eq5 ] ) contains three terms : the desired signal , the interference from other users and the noise including relay enhanced noise and the local noise . the effective channel from transmitter @xmath21 to receiver @xmath45 is given by @xmath46 . suppose all the transmit signals and noise in the system are independent of each other . the transmit powers at each user and each relay are , respectively : @xmath47 then the total relay transmit power is @xmath48 . in the following two sections we propose different models with corresponding algorithms to obtain efficient system precoders , decoders and relay beamforming matrices . for the sake of expression simplicity , we predefine some symbols here : precoded and decoded effective channel from transmitter @xmath21 to relay @xmath22 as @xmath49 and @xmath50 , respectively ; precoded and decoded effective channel from relay @xmath22 to receiver @xmath21 @xmath51 and @xmath52 , respectively , @xmath53 . in this section , we develop a new model to approximate sum rate maximization . a low complexity algorithm to optimize the users precoders , decoders and relay beamforming matrices is proposed . and its computational complexity is analyzed . first , we set up the new model of maximizing tstinr with per user and total relay transmit power constraints . define @xmath54 , where @xmath55 are the desired signal power , the leakage interference and the noise power at receiver @xmath21 , respectively . we wish to maximize the system sum rate @xmath56 with @xmath57 . the direct optimization of the system sum rate is complicated . therefore , we approximate it by the tstinr and maximize the tstinr instead . as the tstinr remains invariant with @xmath58 replaced by @xmath59 , where @xmath60 is any @xmath18-dimensional unitary matrix , we require the decoders @xmath61 to be orthogonal , as the bases of the @xmath18-dimensional solution subspaces . also , the following theorem holds : [ th : th1 ] for any precoder @xmath62 , relay beamforming matrix @xmath63 and any decoder @xmath58 satisfying @xmath64 , we have @xmath65\leq r_{\textrm{sum}}(\{\mathbf{u}\},\{\mathbf{w}\})$ ] . the detailed proof is shown in appendix - a . this states that the result from maximizing tstinr provides a guaranteed system throughput . besides the orthogonality constraints of decoders , we add fixed transmit power constraints for per user and total relay . then the corresponding optimization problem is : [ eq : eq4 ] @xmath66 \hspace{-0.8cm}&\textrm{s.t . } & \mathbf{v}_{k}^{h}\mathbf{v}_{k}=\mathbf{i}_{d_{k}},\\[-0.2 cm ] \hspace{-0.8cm}&&\|\mathbf{u}_{k}\|_{f}^{2}=p_{0}^{t } , k\in\mathcal { k},\label{eq : eq4.2}\\[-0.2 cm ] \hspace{-0.8cm}&&\sum_{r\in\mathcal { r}}\!(\sum_{k\in\mathcal { k}}\!\|\mathbf{w}_{r}\mathbf{g}_{rk}\mathbf{u}_{k}\|_{f}^{2 } + \sigma_{1}^{2}\|\mathbf{w}_{r}\|_{f}^{2})=p_{\max}^{r}.\label{eq : eq4.1}\end{aligned}\ ] ] there is a lack of efficient methods to deal with ( [ eq : eq4.3 ] ) because it is a fraction . this makes problem ( [ eq : eq4 ] ) difficult to solve . stimulated by dinkelbach s work @xcite for nonlinear fraction optimization problem on convex sets , we use a parameter @xmath67 to combine the denominator and the numerator as the new objective function , whereas the conclusions in @xcite can not be extended to the problem ( [ eq : eq4 ] ) with nonconvex feasible set . reformulate ( [ eq : eq4 ] ) as follows : [ eq : eq6 ] @xmath68\label{eq : eq6.1}\\ \hspace{-0.8cm}&\textrm{s.t . } & \mathbf{v}_{k}^{h}\mathbf{v}_{k}=\mathbf{i}_{d_{k}},\label{eq : eq6.2}\\ \hspace{-0.8cm}&&\textrm{tr}(\mathbf{u}_{k}^{h}\mathbf{u}_{k})=p_{0}^{t } , k\in\mathcal { k},\label{eq : eq6.3}\\ \hspace{-0.9cm}&&\sum_{r\in\mathcal { r}}(\sum_{k\in\mathcal { k}}\|\mathbf{w}_{r}\mathbf{g}_{rk}\mathbf{u}_{k}\|_{f}^{2 } + \sigma_{1}^{2}\|\mathbf{w}_{r}\|_{f}^{2})=p_{\max}^{r}.\label{eq : eq6.4}\end{aligned}\ ] ] thus in each iteration we solve ( [ eq : eq6 ] ) , and then update the parameter @xmath67 as follows : initially @xmath67 is set as a small positive scalar ( for example @xmath69 ) , then after each iteration it is updated as @xmath70 with such updating strategy of @xmath67 , we have the following theorem , which is proved in appendix - b : [ th : th2 ] if the objective function of ( [ eq : eq6 ] ) has sufficient reduction in each iteration and @xmath67 is updated as ( [ eq : eq33 ] ) , then the objective function of ( [ eq : eq4 ] ) , tstinr , is monotonically increasing . any stationary point of ( [ eq : eq6 ] ) is also a stationary point of ( [ eq : eq4 ] ) . the programming ( [ eq : eq6 ] ) itself is a nonconvex nonlinear matrix optimization problem , which is difficult to handle jointly . thus we solve precoders @xmath62 , decoders @xmath71 and relay beamforming matrices @xmath63 alternatively . efficient algorithms are developed for each subproblem . firstly , we fix @xmath62 and @xmath63 , then all @xmath71 are independent of each other . the subproblem for @xmath58 becomes : @xmath72 where @xmath73 represents variable @xmath58 , and @xmath74 . since @xmath7 is hermitian , we obtain the closed form solution of ( [ eq : eq7 ] ) as @xmath75 . next , we solve the subproblem for @xmath76 . given a certain index @xmath27 , we fix @xmath77 , @xmath71 and @xmath78 . thus the optimization subproblem for @xmath76 is : @xmath79 + 2\textrm{re}\big[\sum_{k\in\mathcal { k}}\!\sum_{\substack{l\neq r,\\l\in\mathcal { r}}}\!\textrm{tr}(\mathbf{x}\mathbf{p}_{rl}^{k}\mathbf{w}_{l}^{h } \bar{\mathbf{v}}_{kl}^{h}\bar{\mathbf{v}}_{kr})\big]\nonumber\\ \!&\textrm{s.t.}&\!\ ! \textrm{tr}\big[\mathbf{x}(\sum_{k\in\mathcal { k}}\bar{\mathbf{g}}_{rk}\bar{\mathbf{g}}_{rk}^{h } + \sigma_{1}^{2}\mathbf{i}_{l_{r}})\mathbf{x}^{h}\big]=\eta_{1},\end{aligned}\ ] ] where @xmath80 , @xmath81 and @xmath82 . problem ( [ eq : eq9 ] ) is equivalent to a specific quadratic constrained quadratic programming ( qcqp ) with @xmath83 : [ eq : eq10 ] @xmath84 here @xmath85 , @xmath86 and @xmath87 . from the expressions , we know that @xmath88 and generally @xmath89 is indefinite . here we also discuss the case that @xmath89 is positive semi - definite . first we show that the following theorem holds : given @xmath90 , @xmath91 , @xmath92 , @xmath93 and @xmath94 , ( [ eq : eq10 ] ) is equivalent to @xmath95 further if @xmath89 is indefinite , ( [ eq : eq10 ] ) is equivalent to : @xmath96 for the case of positive semi - definite @xmath89 : if the optimal solution of ( [ eq : eq17 ] ) @xmath97 is not that of ( [ eq : eq24 ] ) , then @xmath98 . it is trivial to prove that ( [ eq : eq10 ] ) is equivalent to ( [ eq : eq24 ] ) and thus the detailed proof is omitted . the global optimality conditions of ( [ eq : eq17 ] ) are as follows : there exists @xmath99 , such that @xmath100 , @xmath101 , @xmath102 and @xmath103 . for the case that @xmath89 is indefinite , it follows that @xmath104 is also indefinite . thus it must hold that @xmath105 . then from the complementary optimality condition we must have @xmath106 . such @xmath107 and @xmath108 satisfy the global optimal condition of ( [ eq : eq24 ] ) . therefore to solve ( [ eq : eq10 ] ) is equivalent to solve ( [ eq : eq17 ] ) with indefinite @xmath89 . if the optimal solution of ( [ eq : eq17 ] ) @xmath97 is not that of ( [ eq : eq24 ] ) , then @xmath109 . from the complementary optimality condition we have @xmath110 . then @xmath98 . problem ( [ eq : eq17 ] ) is a typical trust region ( tr ) subproblem in trust region optimization method . * chapter 6.1.1 ) provides an efficient algorithm to achieve its optimal solution . it first checks whether @xmath111 with @xmath110 . if so , @xmath112 is the optimal solution of ( [ eq : eq17 ] ) ; if not , the optimality conditions are used directly , and the optimal lagrange multiplier @xmath107 is calculated by newton s root - finding method from @xmath106 is solved instead @xcite . ] . when @xmath89 is indefinite , we solve ( [ eq : eq17 ] ) with the corresponding algorithm . when @xmath89 is positive semi - definite , we modify the algorithm to solve ( [ eq : eq24 ] ) : check whether @xmath113 , and if it is not the case , calculate @xmath107 by newton s root - finding method . with this tr method we are able to solve ( [ eq : eq10 ] ) efficiently , and construct @xmath76 from @xmath114 . for a precoder @xmath77 , we fix @xmath71 , @xmath63 and @xmath115 to get the following subproblem . @xmath116 here @xmath73 represents @xmath62 , and @xmath117 let @xmath83 , @xmath118 and @xmath119 . then ( [ eq : eq11 ] ) is turned into a nonconvex qcqp : @xmath120 as all constraints are equalities , we use the sequential quadratic programming ( sqp ) algorithm in @xcite to solve it . we set the initial point in the sqp algorithm as the precoder calculated in the previous iteration . as sqp converges to a local optimal solution from the initial point , we are able to guarantee the sufficient reduction of the objective function . with the above analysis , we conclude the framework of the algorithm to solve ( [ eq : eq4 ] ) : as we guarantee sufficient reduction of the objective function in each subproblem , the objective function value in our algorithm will converge . however as we have separated the variables into more than two parts , there is no theoretical guarantee that the algorithm converges to a stationary point of ( [ eq : eq4 ] ) . [ re : re1 ] sharing similar expression of objective function , the algorithm here for tstinr is also applicable to the tlin model in @xcite . the objective function of ( [ eq : eq6 ] ) is the linear combination of the total leakage interference plus noise @xmath121 and the desired signal power @xmath122 , and the parameter @xmath67 balances their weights , while the model tlin in @xcite only minimizes @xmath121 . from the sum rate point of view , our tstinr model is better motivated . this is verified by simulation results , where significant improvement of system sum rate by tstinr compared to tlin is shown . a similar objective function has been discussed in @xcite , where the desired signal power and the leakage interference are combined and optimized . in that case the leakage interference is aimed to be aligned perfectly , thus the parameter @xmath67 in @xcite approaches to infinity to satisfy the interference alignment constraint . in our paper , @xmath121 might not be reduced to zero , and consequently @xmath67 might not grow to infinity . in @xcite , @xmath67 is enlarged when the interference does not have sufficient reduction , which is different from the update strategy here . in this subsection , we extend our new tstinr model , as well as the tlin and wmmse model in @xcite , to the ones with individual user and individual relay fixed transmit power constraints . with individual user and individual relay fixed transmit power constraints , the tstinr model becomes : @xmath123 assume there is a preprocess to carefully select active relays in the communication stage . here we require all the users and relays to transmit signals with fixed power . the difference from ( [ eq : eq4 ] ) is that , the relay sum power constraint is replaced by @xmath20 individual power constraints for each relay . we use the objective function @xmath124 from ( [ eq : eq6 ] ) in each iteration and preserve the same update strategy of parameter @xmath67 as ( [ eq : eq33 ] ) . when applying alternating iterations , the subproblems for decoders @xmath71 are the same as ( [ eq : eq7 ] ) . the objective functions in the subproblems for relay beamforming matrices @xmath63 and precoders @xmath62 remain the same as in ( [ eq : eq9 ] ) and ( [ eq : eq11 ] ) , respectively . the differences are the constraints . with @xmath73 represents @xmath76 , for any @xmath27 , its constraint is : @xmath125=p_{0}^{r}.\ ] ] the transformed problem has the same structure as ( [ eq : eq10 ] ) , and we solve it with the same method . for a precoder @xmath77 , @xmath26 , the constraints are : @xmath126 where @xmath73 here represents the variable @xmath77 , @xmath127 . then the subproblem is reformulated with @xmath83 : @xmath128 where @xmath129 . similar to solving ( [ eq : eq12 ] ) , we achieve a local optimal solution of ( [ eq : eq14 ] ) and sufficient reduction of the objective function by sqp algorithm . as the constraints in ( [ eq : eq14 ] ) are nonlinear equations , to ensure feasibility we normally require that the number of variables is no less than the number of equations . because we turn the variables from complex domain into real domain to solve them , the number increases to @xmath130 . so we have @xmath131 , for any @xmath26 , as a requirement for such problem . this limits our algorithm . the extension of our low complexity algorithm to constraints with power control is ongoing and future work . the algorithm in @xcite can not be extended to solve problems with individual relay fixed transmit power constraints , because it uses semi - definite programming ( sdp ) relaxation to solve the subproblems for precoders . if @xmath132 , it may get a suboptimal solution by relaxation technique . thus the objective function is not guaranteed to have sufficient reduction . with the adjustment to the subproblems , the basic framework of the algorithm is the same as that in the last subsection . as the objective function of tlin in @xcite is similar to that of the reformulated problem of tstinr , we extend the individual relay power constraints case to tlin with the corresponding replacement of the objective functions of each subproblem . now we extend the individual fixed power constraint model to wmmse model in @xcite . the corresponding optimization problem is as follows : @xmath133 -\textrm{log}_{2}\textrm{det}(\mathbf{s}_{k})\big\}\nonumber\\ \hspace{-0.8cm}&\textrm{s.t . } & \mathbf{s}_{k}\succeq0 , \|\mathbf{u}_{k}\|_{f}^{2}=p_{0}^{t } , k\in\mathcal { k},\nonumber\\ \hspace{-0.8cm}&&\sum_{k\in\mathcal { k}}\|\mathbf{w}_{r}\mathbf{g}_{rk}\mathbf{u}_{k}\|_{f}^{2 } + \sigma_{1}^{2}\|\mathbf{w}_{r}\|_{f}^{2}=p_{0}^{r},r\in\mathcal { r},\end{aligned}\ ] ] with @xmath134 . and @xmath135 are the weight matrices . in this approach , we try to minimize the mean square error . @xcite shows that , if we use the linear mmse receiver filter , ( [ eq : eq15 ] ) shares the same stationary points with the sum rate maximization problem . we also apply the alternating minimization algorithm to solve ( [ eq : eq15 ] ) . from the above analysis , @xmath58 , for all @xmath26 , are set as mmse filter . with fixed @xmath77 , for all @xmath26 and @xmath76 , for all @xmath27 , we have @xmath136 take the partial derivative of the objective function of ( [ eq : eq15 ] ) with respect to @xmath137 and set the expression be zero . then we obtain ( [ eq : eq20 ] ) : @xmath138^{-1}=\mathbf{i}_{d_{k}}+\mathbf{t}_{kk}^{h}\mathbf{f}_{k}^{-1}\mathbf{t}_{kk}.\end{aligned}\ ] ] fixing all other variables , the subproblem for relay beamforming matrix @xmath76 , for any @xmath27 , represented by @xmath73 is expressed as follows : @xmath139 - 2\textrm{re}\big[\sum_{k\in\mathcal { k}}\!\sum_{q\in\mathcal { k}}\!\sum_{\substack{l\neq r,\\ l\in\mathcal{r } } } \textrm{tr}(\mathbf{x}\bar{\mathbf{g}}_{rq}\bar{\mathbf{g}}_{lq}^{h}\mathbf{w}_{l}^{h}\bar{\mathbf{v}}_{kl}^{h}\mathbf{s}_{k } \bar{\mathbf{v}}_{kr})\big]\nonumber\\ \hspace{-0.8cm}&\textrm{s.t.}&\textrm{tr}\big[\mathbf{x}(\sum_{k\in\mathcal { k}}\bar{\mathbf{g}}_{rk}\bar{\mathbf{g}}_{rk}^{h } + \sigma_{1}^{2}\mathbf{i}_{l_{r}})\mathbf{x}^{h}\big]=p_{0}^{r}.\end{aligned}\ ] ] by applying @xmath83 , ( [ eq : eq16 ] ) is transformed into a qcqp similar to ( [ eq : eq10 ] ) . we use the same method to solve it . for a precoder @xmath77 , for any @xmath26 , while fixing all other variables , the subproblem becomes : @xmath140\mathbf{x}\right\ } -2\textrm{re}\left[\textrm{tr}(\sum_{r\in\mathcal { r}}\mathbf{s}_{k}\bar{\mathbf{v}}_{kr}\bar{\mathbf{w}}_{rk}\mathbf{x})\right]\nonumber\\ & \textrm{s.t.}&\|\mathbf{x}\|_{f}^{2}=p_{0}^{t},\nonumber\\ & & \textrm{tr}(\mathbf{x}^{h}\bar{\mathbf{w}}_{rk}^{h}\bar{\mathbf{w}}_{rk}\mathbf{x})=\eta_{3}^{r } , r\in\mathcal { r},\end{aligned}\ ] ] where @xmath73 represents @xmath77 . with @xmath83 , we transform it into a qcqp similar to ( [ eq : eq14 ] ) and solve it efficiently by sqp method . the algorithm to solve the wmmse model with individual user and individual relay fixed power constraints is as follows : in this subsection , we compare the computational complexity of the algorithm for our new model tstinr with that of the algorithm in @xcite for the wmmse model , both with per user and total relay transmit power as representation . as both algorithms consist of three main parts of subproblems , we analyze them individually . first , we consider the complexity for solving decoder @xmath58 , as well as the weight matrix @xmath137 in wmmse , @xmath26 . the construction for the matrix @xmath7 in ( [ eq : eq7 ] ) of tstinr has the same computations as that for @xmath141 and @xmath142 to solve @xmath58 and @xmath137 of wmmse , similar to ( [ eq : eq19 ] ) and ( [ eq : eq20 ] ) . besides , tstinr requires @xmath143 operations for eigenvalue decomposition of @xmath7 ; wmmse requires @xmath144 operations for eigenvalue decompositions of @xmath141 and @xmath142 . thus in the first part , tstinr requires less complexity than wmmse . second , we focus on the part for the relay beamforming matrix @xmath76 , with any @xmath27 . both algorithms require similar complexity to construct the corresponding subproblem . to solve subproblem ( [ eq : eq10 ] ) , our tstinr algorithm applies the tr method , which is mainly the newton s root finding method for @xmath107 . it requires only a few inner iterations to find the optimal @xmath107 and in each inner iteration the main calculation is the qr factorization , with complexity of @xmath145 . to the contrary , the wmmse algorithm in @xcite applies sdp method , whose complexity is @xmath146 . the complexity in the second part of tstinr is much less than that of wmmse . third is the part for the precoder @xmath77 , with any @xmath26 . tstinr solves ( [ eq : eq12 ] ) by sqp method , which complexity is @xmath147 . wmmse solves a qcqp with the same structure but with inequality constraints . and it applies sdp relaxation method , with complexity of @xmath148 . with similar computations to construct the corresponding subproblem , tstinr has much lower compelxity than wmmse in the third part . .computational complexity analysis for each subproblem in one iteration [ cols="<,^,^",options="header " , ] the complexity differences of each subproblem in one iteration between the tstinr and the wmmse model are listed in table 1 . from the comparison of the three main parts of the two algorithms , we conclude that our proposed new algorithm for tstinr enjoys lower complexity than the algorithm for wmmse in @xcite . similarly , it is analyzed that our new algorithm for tstinr has lower complexity than the algorithm for tlin in @xcite . with numerical evidence , it turns out that the compared algorithms have similar number of iterations to solve the problem . because the sqp method , which solves ( [ eq : eq12 ] ) , ( [ eq : eq14 ] ) and ( [ eq : eq21 ] ) , can only deal with equality constraints , we restrict our models to fixed power constraints . although the models with power control constraints are more general and practical , the application of sqp method saves computational complexity . even with fixed power constraints , our tstinr model performs better than the wmmse model with power control constraints in medium to high snr scenarios , as shown in section [ sec : sec3 ] . in this section , we study the multiple data stream model , with the purpose to maximize the system sum rate . it is pointed out in @xcite that multiple data streams , corresponding to multiple dofs , help to increase the capacity of single - hop network in medium to high snr . this conclusion can be extended to the two - hop case , by treating the network as an equivalent single - hop network between users and assuming the same power at the relays and the transmitters . first , we analyze the achievable number of data streams of the models from section . then our proposed tstinr model and the corresponding algorithm are modified to support multiple data streams for each user pair . here all the models include per user and total relay transmit power constraints . the dimension @xmath28 of the transmit signal @xmath149 , is expected as the achieved number of data streams at user @xmath21 . however , in simulations when we apply our tstinr algorithm to the system with @xmath150 , we always observe that the system precoder @xmath77 has rank one . this implies that with linearly dependent columns of each precoder we can only achieve one data stream for each user pair , regardless of @xmath28 . similar phenomena are observed for the tlin and wmmse algorithm in @xcite . the following theorem provides theoretical evidence for the phenomena of the tstinr and tlin models : [ th : th3 ] in our proposed tstinr model and in the tlin model from @xcite , the subproblem for precoder @xmath77 always has a rank one optimal solution , regardless of @xmath28 . define @xmath151 and drop the rank constraint of @xmath152 , the subproblem ( [ eq : eq11 ] ) in tstinr is relaxed to the following semi - definite programming : @xmath153 in ( * ? ? ? * theorem 4.1 ) , it is shown that ( [ eq : eq29 ] ) always has a rank one optimal solution . suppose @xmath154 is the optimal solution of ( [ eq : eq29 ] ) . let @xmath155 $ ] with @xmath156 . that is , the columns of @xmath157 consist of @xmath158 . from the fact that @xmath159=\sum_{i=1}^{d_{k}}a_{i}^{2}(\mathbf{y}^{*})^{h}\mathbf{q}_{k}\mathbf{y}^ { * } = \textrm{tr}(\mathbf{y}^{*}\mathbf{q}_{k})$ ] , for any feasible @xmath73 of ( [ eq : eq11 ] ) , we have : @xmath160.\ ] ] thus we conclude @xmath157 is an optimal solution of ( [ eq : eq11 ] ) . because the subproblem for precoder in tlin from @xcite has the same structure as ( [ eq : eq11 ] ) , we conclude the same result for tlin . theorem [ th : th3 ] shows that based on the structure of the subproblem , the optimization always has rank one precoders in tstinr and tlin as solutions . simulations verify this behavior in all cases . the same phenomenon is observed for the wmmse model of @xcite whereas the conclusion of theorem [ th : th3 ] can not be extended to wmmse , due to the extra linear term in the objective function of the precoder subproblem . therefore , with the existing models we can only achieve a single data stream for each user pair . thus a new model should be proposed to achieve multiple data streams . in this subsection , we propose the new model based on the tstinr model in section to support multiple data streams . sufficient motivation for the construction of the new model is also provided . to achieve the required number of parallel data streams , we should have independent columns of precoder @xmath77 for all @xmath26 . without loss of generality we require the columns of @xmath77 to be orthogonal . whereas there is a transmit power constraint ( [ eq : eq4.2 ] ) for each user in ( [ eq : eq4 ] ) , we have the power allocation among @xmath28 parallel data streams for user @xmath21 . first , we modify our tstinr model as follows : @xmath161 here @xmath162 is a @xmath163 diagonal positive semi - definite matrix , which contains the data stream power allocation variable of user @xmath21 . from the optimization point of view , the feasible set of precoder @xmath77 is restricted to have orthogonal columns , comparing with that of ( [ eq : eq4 ] ) . this avoids the phenomenon observed in the solution of ( [ eq : eq4 ] ) that all columns of the rank one precoders @xmath77 are nonzero but linearly dependent . different from what has been mentioned in theorem [ th : th3 ] , here the rank one case of @xmath77 only happens when one diagonal element of @xmath162 is nonzero , which result in all columns of @xmath77 but one are all zeros . hence , we focus on the analysis of the subproblem to solve @xmath77 as well as @xmath162 . the reformulation of the objective function of ( [ eq : eq39 ] ) and the update strategy of the parameter @xmath67 are similar to the algorithm to solve ( [ eq : eq4 ] ) . given @xmath26 , fixing all variables other than @xmath77 and @xmath162 , the precoder subproblem becomes : [ eq : eq40 ] @xmath164 where @xmath73 represents @xmath77 . [ th : th4 ] the optimal @xmath162 of ( [ eq : eq40 ] ) is of rank one , i.e. , there is only one positive element on the diagonal of @xmath162 . the detailed proof is shown in appendix - c . as described in theorem [ th : th4 ] , at the optimal solution of ( [ eq : eq40 ] ) the complete transmit power should be assigned to one data stream . this leads to @xmath165 and thus only one data stream can be transmitted for each user . in our new model for multiple stream case , without transmit power optimization , we assume each user has fixed transmit power @xmath166 , and require equal power allocation among parallel data streams for each user . this choice accords with the optimal power allocation scheme to maximize the system sum rate in the high snr scenario @xcite . suppose user @xmath21 has @xmath28 parallel data streams , and the corresponding optimization problem of the new model becomes : @xmath167 similar to ( [ eq : eq4 ] ) , we reformulate the objective function of ( [ eq : eq32 ] ) with parameter @xmath67 , which adopts the update strategy ( [ eq : eq33 ] ) , and becomes @xmath168 . we apply the alternating minimization method to solve precoders , decoders and relay beamforming matrices in the reformulated problem . the subproblems for decoders @xmath61 are the same as ( [ eq : eq7 ] ) in section [ sec : sec2 ] . for any @xmath27 , the subproblem for @xmath76 while fixing all other variables is reformulated as ( [ eq : eq10 ] ) with inequality constraint . then it is equivalent to the typical trust region subproblem ( [ eq : eq17 ] ) , and solved by tr method in @xcite . for the precoder @xmath77 , the corresponding subproblem becomes following , while fixing @xmath169 , @xmath170 and @xmath115 : [ eq : eq35 ] @xmath171 where @xmath73 represents the variable @xmath77 , and @xmath172 , @xmath173 and @xmath174 are mentioned just after the equation of ( [ eq : eq11 ] ) . before presenting the algorithm to solve ( [ eq : eq35 ] ) , we first show its global optimality conditions : [ th : th5 ] the global optimality conditions for ( [ eq : eq35 ] ) are stated as follows . there exists @xmath175 as the lagrange multiplier of ( [ eq : eq35.1 ] ) , such that : + oc1 : : @xmath176 is the optimal solution for : @xmath177.\end{aligned}\ ] ] oc2 : : complimentary condition holds : @xmath178-\eta_{2}\}=0 $ ] . oc3 : : @xmath179 as the function of @xmath180 satisfies ( [ eq : eq35.1 ] ) : + @xmath181^{h}\mathbf{l}_{k}\mathbf{x}^{*}(\mu^{*})\}\leq \eta_{2}$ ] . suppose there exists @xmath182 as a feasible point of ( [ eq : eq35 ] ) , which satisfies : @xmath183 $ ] . + if @xmath184 , then from oc1 , @xmath157 is the global optimal solution of the relaxed problem of ( [ eq : eq35 ] ) by dropping the constraint ( [ eq : eq35.1 ] ) . thus for @xmath73 as any feasible point of ( [ eq : eq35 ] ) , it holds that @xmath185\leq\textrm{tr}(\mathbf{x}^{h}\mathbf{q}_{k}\mathbf{x})$ ] , which contradicts the assumption . + if @xmath186 , then @xmath187=\eta_{2}$ ] holds from oc2 . thus we deduce @xmath188 $ ] . then we have @xmath189 < \textrm{tr}[(\mathbf{x}^{*})^{h}(\mathbf{q}_{k}+\mu^{*}\mathbf{l}_{k})\mathbf{x}^{*}]$ ] , which contradicts the fact that @xmath157 is the global optimal solution of ( [ eq : eq38 ] ) . for both cases we have proved that the assumption for @xmath182 does not hold . thus @xmath157 is the global optimal solution of ( [ eq : eq35 ] ) . with eigenvalue decomposition , we obtain the optimal solution of ( [ eq : eq38 ] ) as @xmath190 . we want to obtain @xmath175 , to satisfy oc2 and oc3 . if @xmath191 , then @xmath192 is the optimal lagrange multiplier . otherwise @xmath192 can not be optimal and @xmath193 should be strictly greater than @xmath194 . thus we should always have @xmath195 = \eta_{2}$ ] from oc2 . also from the constraint ( [ eq : eq35.1 ] ) we should have @xmath196 for a feasible problem . then with @xmath179 as a continuous function , there exists @xmath197 such that @xmath198 . thus we use newton s root finding method @xcite to search for @xmath193 . the algorithm to solve subproblem ( [ eq : eq35 ] ) is summarized as follows : with the methods for all the three subproblems , the algorithm for ( [ eq : eq32 ] ) is presented as algorithm 3 : by enforcing the orthogonality constraints to the columns of each precoder , it is guaranteed that @xmath199 . thus each user pair has @xmath28 parallel data streams as expected . in this section , we evaluate the performances of our proposed algorithms . simulations include two parts , where the single stream and the multiple streams models are analyzed , respectively . in both parts , each element of @xmath200 and @xmath201 are generated as i.i.d complex gaussian distribution with zero mean and unit variance . the noise variances are set as @xmath202 . initial values of @xmath62 and @xmath63 are randomly generated , and scaled to be feasible . initially , the parameter in tstinr model is set as @xmath69 . for each plotted point , @xmath203 random realization of different channelcoefficients are generated to evaluate the average performance . here we define snr as snr@xmath204 , and @xmath205 . we use system sum rate @xmath206 as the measure of qos . we use matlab simulation server r2010a based on an intel core i7 - 875k cpu with 8 gb ram with the operating system 64 bits debian linux to run all the simulations . in this subsection , we consider a @xmath207 mimo relay system , which means @xmath208 , for all @xmath26 and @xmath27 , and @xmath209 . first , we analyze models with per user and total relay power constraints . our algorithm for tstinr model is compared with those of the tlin model and the wmmse model with power control in @xcite . here we use the sedumi toolbox to solve the subproblems with sdp algorithm in wmmse . for the tlin model , we apply similar algorithm as proposed for tstinr to speed up . fig . 2 and fig . 3 depict the achieved sum rate and the computing time with respect to different snr values for the three algorithms , respectively . from the aspect of the achieved sum rate , tstinr outperforms tlin for general snr values as expected , and outperforms wmmse in medium and high snr scenarios . furthermore , the computing time of tstinr and tlin are much less than that of wmmse , which accords with the computational complexity analysis . by numerical evidence , the computing time of wmmse increases with increasing snr value , mainly because the rank - one solution is less frequently observed for sdp , and the relaxation technique @xcite has to be used . then we compare the algorithms for models with individual user and individual relay power constraints . 4 depicts the average achievable sum rate by our proposed algorithms for tlin , tstinr and wmmse model with fixed transmit power , as well as the algorithm in @xcite for the wmmse model with power control . it is shown that wmmse with power control enjoys higher sum rate than wmmse with fixed power constraints generally . thus system sum rate benefits substantially from power control . even so , our proposed tstinr with fixed power constraints outperforms wmmse with power control for medium to high snr scenarios . also , tstinr outperforms tlin in general . here the complexity as well as the computing time of the tlin , tstinr and wmmse with power control behave similar to the algorithms in section [ sec : sec2 ] . the comparison with both total relay and individual relay constraints cases are similar . in both fig . 2 and fig . 4 the curves representing tstinr and wmmse with power control have intersection point at about snr@xmath210db . with snr value higher than @xmath211db , our proposed tstinr algorithm outperforms the wmmse algorithm with power control . the performance confirms the analysis : in low snr scenarios mmse receiver filter are almost optimal considering linear filter , while in high snr scenarios the receiver filter should be close to zero - forcing solution , which can be achieved by tstinr . in this subsection we consider multiple stream models of the mimo relay networks . we investigate three kinds of @xmath0 networks with different number of antennas , that is , @xmath212 .. here for the scheme @xmath213 we choose the maximum system sum rate results between our tstinr model and the wmmse model with power control in @xcite . first we consider a network with @xmath214 antennas for each user and @xmath215 antennas for each relay . the number of data streams for user @xmath21 , @xmath28 , varies from @xmath216 to @xmath214 for both @xmath217 . for different choices of @xmath218 the average sum rate results corresponding to different snr values are shown in fig . as expected , in the low snr scenario the single stream scheme with @xmath213 outperforms other schemes ; in medium to high snr scenarios , the scheme @xmath219 becomes dominant and the scheme @xmath213 performs worse than all others , in terms of sum rate . in the second example , the considered network has the same parameters as the previous one , except that each relay owns @xmath214 antennas . similar to fig . 5 , the average achieved sum rate results corresponding to different snr values for different requirements of data streams are shown in fig . 6 . the curves are quite different from the previous example . here the schemes @xmath220 and @xmath221 outperform the other two schemes in medium to high snr scenarios . and generally the scheme @xmath219 performs very bad . the performances shown in fig . 5 and fig . 6 indicate similar behavior as the recent theoretical result on dof of mimo relay networks . from the cut - set bound theory , the maximum dofs of the first network is no greater than @xmath214 for each user ( * ? ? ? * theorem 15.1 ) . and simulations verify the benefit to transmit @xmath214 data streams for each user over other schemes in fig . 5 . however in the second example , there is no extra relay antenna to align interference besides transmitting the desired signal . without symbol extension or time division , @xmath214 dofs for each user is not achievable . this accords with the performances in fig . 6 . in @xcite the authors show that @xmath222 dofs for each user are achievable in @xmath0 network with @xmath214 antennas for each user and each relay . and the essential idea of the transmission scheme is to sacrifice one data stream for interference and make full use of all other streams . correspondingly , in fig . 6 for the medium to high snr scenarios the schemes @xmath220 and @xmath221 perform the best . this indicates substantial benefit of system sum rate from such schemes . this is also verified in the third example with @xmath223 antennas for each user and each relay . the comparison of different data stream schemes are depicted in fig . 7 , where the scheme @xmath224 achieves the highest sum rate among all the schemes in medium to high snr . in general , multiple stream schemes improve the system sum rate in medium to high snr scenarios . this paper considered the general @xmath1 mimo relay network . with per user and total relay power constraints , an algorithm for the tstinr model was proposed . for the individual user and individual relay fixed power constraints , the tstinr algorithm was extended , and the algorithms for the tlin and wmmse model in @xcite were also modified to solve the problem with such constraints . computational complexity analysis showed that our proposed algorithm for tstinr has much lower complexity than the wmmse algorithm from @xcite . focusing on per user and total relay power constraints , we proposed a multiple stream tstinr model , to overcome the disadvantages of the previous models that only single data stream can be transmitted for each user . simulations showed that for single stream case tstinr algorithm performs better than wmmse in medium and high snr scenarios for both constraint cases , and outperforms tlin generally , in terms of achievable sum rate ; the system sum rate significantly benefits from the multiple data stream transmission in medium to high snr scenarios . \1 . we prove that for any @xmath26 , @xmath240.\end{aligned}\ ] ] for simplicity we omit the subscript @xmath21 . let @xmath241 be the bases of the complementary subspace of the subspace spanned by the columns of @xmath242 . that is , @xmath243\in\mathbb{c}^{n\times n}$ ] is a unitary matrix . then we deduce that @xmath244 here the equality is deduced from the property of unitary matrix @xmath60 . and the inequality comes from lemma 1 , @xmath245 and that @xmath246 \2 . @xmath247 induces @xmath248 . lemma 2 shows that @xmath249 from ( [ eq : eq23 ] ) and ( [ eq : eq25 ] ) , it is concluded that @xmath250\geq \sum_{k\in\mathcal{k}}\textrm{log}_{2}\frac{\textrm{tr}(\mathbf{v}_{k}^{h}\mathbf{c}_{k}\mathbf{v}_{k } ) } { \textrm{tr}(\mathbf{v}_{k}^{h}\mathbf{b}_{k}\mathbf{v}_{k})}.\end{aligned}\ ] ] \3 . finally we prove that : @xmath251 with any scalar @xmath252 and the fact that @xmath253 , it is deduced that : @xmath254 let @xmath255 , divide @xmath256 and take logarithm for both sides , and thus we have ( [ eq : eq22 ] ) . combining ( [ eq : eq26 ] ) and ( [ eq : eq22 ] ) we prove theorem [ th : th1 ] . let @xmath257 represent the set of the iterative points @xmath258 . suppose @xmath259 are the feasible points achieved from the @xmath260th iteration . define @xmath261 . then the expression of parameter @xmath67 used in the @xmath260th iteration as well as the tstinr achieved in the @xmath262th iteration is : @xmath263 suppose @xmath266 are the stationary points of ( [ eq : eq6 ] ) and @xmath267 is the lagrange multiplier of ( [ eq : eq6.4 ] ) : @xmath268 . then the first order optimality conditions of the problem ( [ eq : eq6 ] ) with respect to @xmath76 are : @xmath269 when the iterative points converges to @xmath270 , we have @xmath271 . taking it into ( [ eq : eq27 ] ) , and let @xmath272 . then we have @xmath273 ^ 2}\left[p^{i+n}(\{\mathbf{x}^{*}\})\frac{\partial p^{s}(\{\mathbf{x}^{*}\})}{\partial\mathbf{w}_{r } } -p^{s}(\{\mathbf{x}^{*}\})\frac{\partial p^{i+n}(\{\mathbf{x}^{*}\})}{\partial\mathbf{w}_{r}}\right ] -\tilde{\lambda}\frac{\partial h(\{\mathbf{x}^{*}\})}{\partial\mathbf{w}_{r}}=\mathbf{0}.\end{aligned}\ ] ] with @xmath274 as the lagrange multiplier of ( [ eq : eq4.1 ] ) , ( [ eq : eq37 ] ) and ( [ eq : eq36 ] ) consist of the first order optimality conditions of ( [ eq : eq4 ] ) with respect to @xmath76 . similarly , we are able to achieve the first order optimality conditions of ( [ eq : eq4 ] ) with respect to other variables . thus @xmath270 are also the stationary points of ( [ eq : eq4 ] ) . similar to theorem [ th : th5 ] for problem ( [ eq : eq35 ] ) , the optimality conditions for ( [ eq : eq40 ] ) are as follows : + 1 . @xmath275 are the optimal solutions for the problem below : @xmath276\nonumber\\ & \textrm{s.t . } & \mathbf{x}^{h}\mathbf{x}=\boldsymbol{\phi}_{k},\nonumber\\ & & \textrm{tr}(\boldsymbol{\phi}_{k})\leq p_{0}^{t } , \boldsymbol{\phi}_{k}\hspace{0.1cm}\textrm{is diagonal } , \boldsymbol{\phi}_{k}\succeq0.\end{aligned}\ ] ] 2 . complimentary condition holds : @xmath277-\eta_{2}\}=0 $ ] . @xmath278 satisfies ( [ eq : eq40.1 ] ) : @xmath279^{h}\mathbf{l}_{k}\mathbf{x}^{*}\}\leq \eta_{2}$ ] . here @xmath280 is the optimal lagrange multiplier of ( [ eq : eq40.1 ] ) . with @xmath281 we are able to obtain the optimal @xmath157 and @xmath282 of ( [ eq : eq40 ] ) . if the @xmath260th diagonal element of @xmath162 is zero , then the @xmath260th column of @xmath73 should be @xmath283 . in this situation we can delete this column and optimize the remaining ones . without loss of generality we assume each element of the diagonal of @xmath162 is strictly positive , that is , @xmath284 . let @xmath285 , and @xmath286 . omit the index @xmath21 for simplicity . problem ( [ eq : eq8 ] ) is equivalent to the following problem : in the following we analyze the optimal solutions of ( [ eq : eq30 ] ) to show the property of @xmath288 and @xmath289 . here we treat the columns of @xmath290 as the linear combination of the eigenvectors of @xmath291 . let @xmath292 be the eigenvalues of @xmath291 . suppose @xmath293 , the @xmath260th column of @xmath290 is the linear combination of the eigenvectors corresponding to the eigenvalues @xmath294 , where @xmath295 . then according to ( [ eq : eq30.1 ] ) , different columns correspond to different eigenvalues of @xmath291 , that is , @xmath296 . then the feasible set is divided into @xmath18 independent parts with @xmath293 as variables , for @xmath297 , respectively . define @xmath298 as the @xmath260th diagonal element of @xmath299 . the objective function of ( [ eq : eq30 ] ) is rewritten as @xmath300 . for any fixed feasible @xmath301 , the optimal solution of ( [ eq : eq30 ] ) , @xmath302 , is the eigenvector of matrix @xmath291 corresponding to the eigenvalue @xmath303 , as long as @xmath304 , for any @xmath297 . then taking these solutions back into ( [ eq : eq30 ] ) , the problem becomes : @xmath305 as the objective function of ( [ eq : eq41 ] ) is linear in @xmath298 for any @xmath306 , the optimal solutions of ( [ eq : eq41 ] ) should be @xmath307 and @xmath308 . thus , the optimal solution @xmath282 of ( [ eq : eq40 ] ) is of rank one . c. sun and y. yuan , a fast algorithm for beamforming problems in distributed communication of relay networks , " in _ proc . speech signal process . ( icassp ) _ , prague , czech , may . 2011 , pp . 3304 - 3307 . j. li , f. roemer and m. haardt , spectrum and infrastructure sharing in the mimo interference channels , " in _ proc . 19-th european signal processing conference ( eusipco 2011 ) _ , barcelona , spain , aug . 2011 , pp . 181 - 185 . a .- toding , m .- ra . khandaker and y. rong , joint source and relay optimization for parallel mimo relay networks , " in _ eurasip journal on advances in signal processing _ , doi:10.1186/1687 - 6180 - 2012 - 174 , aug . 2012 . j. liu , n. b. shroff , and h. d. sherali , optimal power allocation in multi - relay mimo cooperative networks : theory and algorithms , " _ ieee journal on selected areas in communications _ , vol . 331 - 340 , feb . 2012 . choi , k - j . lee , c. song , h. song and i. lee , weighted sum rate maximization for multi - user multi - relay mimo systems , " _ accepted by ieee trans . vehicular technology _ , doi : 10.1109/tvt.2012.2226068 , 2012 . k. t. truong , p. sartori and r. w. heath , jr . , cooperative algorithms for mimo amplify - and - forward relay networks , " _ online available _ , arxiv:1112.4553v3 , _ submitted to ieee trans . signal process . _ , 2012 . m. razaviyayn , g. lyubeznik and z. q. luo , on the degrees of freedom achievable through interference alignment in a mimo interference channel , " _ ieee trans . signal process . 812 - 821 , feb . 2012 . k. gomadam , v. r. cadambe and s. a. jafar , a distributed numerical approach to interference alignment and applications to wireless interference networks , " _ ieee trans . inf . theory _ 57 , no . 6 , pp . 3309 - 3322 , jun , 2011 . c. sun y. yang and y. yuan , low complexity interference alignment algorithms for desired signal power maximization problem of mimo channels , " in _ eurasip journal on advances in signal processing _ , doi : 10.1186/1687 - 6180 - 2012 - 137 , jul . 2012 . t. gou , s. a. jafar , c. wang ; s .- w . jeon and s .- y . chung , aligned interference neutralization and the degrees of freedom of the @xmath0 interference channel , " _ ieee trans . inf . theory _ , vol . 4381 - 4395 , 2012 . c. s. vaze and m. k. varanasi , beamforming and aligned interference neutralization achieve the degrees of freedom region of the @xmath0 mimo interference network , " in _ information theory and applications workshop _ , ca , usa , 2012 , pp . 199 - 203 . q. shi , m. razaviyayn , z. luo and c. he , an iterative weighted mmse approach to distributed sum - utility maximization for a mimo interfering broadcast channel , " _ ieee trans . signal process . _ , vol.59 , no.9 , pp . 4331 - 4340 , sept . 2011 .

in this paper , we study the hlder regularity of set - indexed stochastic processes defined in the framework of ivanoff - merzbach . the first key result is a kolmogorov - like hlder - continuity theorem derived from the approximation of the indexing collection by a nested sequence of finite subcollections . increments for set - indexed processes are usually not simply written as , hence we considered different notions of hlder - continuity . then , the localization of these properties leads to various definitions of hlder exponents . in the case of gaussian processes , almost sure values are proved for these exponents , uniformly along the sample paths . as an application , the local regularity of the set - indexed fractional brownian motion and the set - indexed ornstein - uhlenbeck process are proved to be constant , with probability one . _ ams classification _ : 60g60 , 60g17 , 60g15 , 60g22 , 60j65 . _ key words _ : fractional brownian motion , gaussian processes , hlder exponents , kolmogorov criterion , local regularity , random fields , sample path properties , multiparameter and set - indexed processes .

1203.0750

sample path properties of stochastic processes have been deeply studied for a long time , starting with the works of kolmogorov , lvy and others on the modulus of continuity and laws of the iterated logarithm of the brownian motion . since the late 1960s , these results were extended to general gaussian processes , while a finer study of the local properties of these sample paths was carried out ( we refer to berman @xcite , dudley @xcite , orey and pruit @xcite , orey and taylor @xcite and strassen @xcite , for the early study of gaussian paths and their rare events ) . among the large literature dealing with fine analysis of regularity , hlder exponents continue to be widely used as a local measure of oscillations ( see @xcite for examples of recent works in this area ) . two different definitions , called local and pointwise hlder exponents , are usually considered for a stochastic process @xmath1 , depending whether the increment @xmath2 is compared with a power @xmath3 or @xmath4 inside a ball @xmath5 when @xmath6 . as an example , with probability one , the local regularity of fractional brownian motion @xmath7 is constant along the path : the pointwise and local hlder exponents at any @xmath8 are equal to the self - similarity index @xmath9 ( e.g. see @xcite ) . this field of research is also very active in the multiparameter context and a non - exhaustive list of authors and recent works in this area includes ayache @xcite , dalang @xcite , khoshnevisan @xcite , lvy - vhel @xcite , xiao @xcite . as an extension to the multiparameter one , the set - indexed context appeared to be the natural framework to describe invariance principles studying convergence of empirical processes ( e.g. see @xcite ) . the understanding of set - indexed processes and particularly their regularity is a more complex issue than on points of @xmath10 . the simple continuity property is closely related to the nature of the indexing collection . as an example , brownian motion indexed by the lower layers of @xmath11 ^ 2 $ ] ( i.e. the subsets @xmath12 ^ 2 $ ] such that @xmath13\subseteq a$ ] for all @xmath14 ) is discontinuous with probability one ( we refer to @xcite or @xcite for the detailed proof ) . as a matter of fact , necessary and sufficient conditions for the sample path continuity property were investigated , starting with dudley @xcite who introduced a sufficient condition on the metric entropy of the indexing set , followed by fernique @xcite who gave a necessary conditions in the specific case of stationary processes on @xmath10 . talagrand gave a definitive answer in terms of majorizing measures @xcite ( see @xcite or @xcite for a complete survey and also @xcite for a lil and lvy s continuity moduli for set - indexed brownian motion ) . the question was left open so far concerning the exact hlder regularity of set - indexed processes . a formal set - indexed setting has been introduced by ivanoff and merzbach in order to study standard issues of stochastic processes , such as martingale and markov properties ( see @xcite ) . in this framework , an _ indexing collection _ @xmath15 is a collection of subsets of a measure space @xmath16 , which is assumed to satisfy certains properties such as stability by intersection of its elements . section [ sec : ho - cont ] of the present paper uses these properties , instead of conditions on the metric entropy , to derive a kolmogorov - like criterion for hlder - continuity of a set - indexed process . the collection of sets @xmath15 is endowed with a metric @xmath17 and a nested sequence @xmath18 of finite subcollections of @xmath15 such that each element of @xmath15 can be approximated as the decreasing limit ( for the inclusion ) of its projections on the @xmath19 s . we consider a supplementary assumption ( @xmath20 ) on @xmath21 and @xmath17 which impose that : 1 ) the distance from any @xmath22 to @xmath19 can be related to the cardinal @xmath23 , roughly by @xmath24 , where @xmath25 is called the _ discretization exponent _ of @xmath26 ; and 2 ) a minimality condition on the class @xmath27 that is verified in most cases . this is discussed in section [ sec : ho - cont ] , together with the links between our assumption and entropic conditions of previous works . we prove in theorem [ kolmth ] : if @xmath28 is a set - indexed process and @xmath29 are positive constants such that @xmath30\leq k\ { d_{\mathcal{a}}}(u , v)^{{q_{\underline{\mathcal{a}}}}+\beta } $ ] for all @xmath31 , then for all @xmath32 , there exist a random variable @xmath33 and a constant @xmath34 such that almost surely @xmath35 , the increments of a set - indexed process @xmath28 are defined on the class @xmath36 of sets @xmath37 where @xmath38 by the _ inclusion - exclusion formula _ @xmath39 this definition extends the notion of rectangular increments for multiparameter processes . for instance , quantities like @xmath40 } \mathbf{b } = \mathbf{b}_{\mathbf{v } } - \mathbf{b}_{(u_1,v_2 ) } - \mathbf{b}_{(v_1,u_2 ) } + \mathbf{b}_{\mathbf{u}}$ ] , where @xmath41 and @xmath42 is the brownian sheet , were proved to be useful to derive geometric sample path properties of the process ( see e.g. some of the works of dalang and walsh @xcite ) . let us notice that some processes can satisfy an increment stationarity property with respect to these increments while they do not for quantities @xmath0 . moreover , this inclusion - exclusion principle is very useful when it comes to martingale and markov properties . according to this definition , another way to express the hlder - continuity of @xmath43 is @xmath44 , for @xmath45 . this question is clarified in section [ sec : kolm ] . the purpose of hlder exponents is the ( optimal ) localization of the hlder - continuity concept . following the previous discussion , the first definition for local and pointwise hlder exponents is based on the comparison between @xmath46 and a power @xmath47 or @xmath4 in a ball @xmath48 around @xmath49 when @xmath6 . another definition compares @xmath50 for @xmath51 in @xmath36 with @xmath52 and @xmath53 for each @xmath54 , to a power @xmath55 when @xmath6 . as in the real - parameter setting , these two kinds of exponents , precisely defined in sections [ sec : hoexp ] and [ sec : hocl ] , provide a fine knowledge of the local behaviour of the sample paths . in section [ sec : flow ] , the different hlder exponents are linked to the hlder regularity of projections of the set - indexed process on increasing paths . the _ pointwise continuity _ has been introduced in the multiparameter setting in @xcite and in the set - indexed setting in @xcite as a weak form of continuity . in this definition , the _ point mass jumps _ are the only kind of discontinuity considered . without any supplementary condition on the indexing collection , the set - indexed brownian motion satisfies this property , even on lower layers where it is not continuous . in section [ sec : hopc ] , we define the pointwise continuity hlder exponent of a pointwise continuous process @xmath43 by a comparison between @xmath56 with a power @xmath57 when @xmath58 , where @xmath59 is a decreasing sequence of elements in @xmath60 which converges to @xmath61 . in the gaussian case , we prove in section [ sec : as - exp ] that the different aforementioned hlder exponents admit almost sure values . assumption ( @xmath20 ) is the key to extend this result from the multiparameter to the set - indexed setting . moreover these almost sure values can be obtained uniformly on @xmath15 for the local exponent . however , this a.s . result can not be obtained for the pointwise exponent ( even for multiparameter processes ) . nevertheless , we proved that it holds for the set - indexed fractional brownian motion ( defined in @xcite ) in section [ sec : appl ] , thus improving on a result in the multiparameter case @xcite . as this requires some specific extra work , we believe that the uniform almost sure result might not be true for the pointwise exponent of any gaussian process . finally , we also applied our results to the set - indexed ornstein - uhlenbeck ( siou ) process @xcite , for which all exponents are almost surely equal to @xmath62 at any set @xmath22 . in the classical case of real - parameter ( or multiparameter ) stochastic processes , kolmogorov s continuity criterion is a useful tool to study sample path hlder - continuity ( e.g. see @xcite ) . in this section , we focus on the definition of a suitable assumption on the indexing collection , that allows to prove an extension of this result to the set - indexed ( possibly non - gaussian ) setting . a general framework was introduced by ivanoff and merzbach to study martingale and markov properties of set - indexed processes ( we refer to @xcite for the details of the theory ) . the structure of these indexing collections allowed the study of the set - indexed extension of fractional brownian motion @xcite , its increment stationarity property @xcite and a complete characterization of the class of set - indexed lvy processes @xcite . let @xmath63 be a locally compact complete separable metric and measure space , with metric @xmath64 and radon measure @xmath65 defined on the borel sets of @xmath63 . all stochastic processes will be indexed by a class @xmath66 of compact connected subsets of @xmath63 . in the whole paper , the class of finite unions of sets in any collection @xmath67 will be denoted by @xmath68 . in the terminology of @xcite , we assume that the _ indexing collection _ @xmath66 satisfies stability and separability conditions in the sense of ivanoff and merzbach : [ basic ] a nonempty class @xmath66 of compact , connected subsets of @xmath63 is called an _ indexing collection _ if it satisfies the following : 1 . @xmath69 and for all @xmath70 , @xmath71 if @xmath72 . @xmath66 is closed under arbitrary intersections and if @xmath73 are nonempty , then @xmath74 is nonempty . if @xmath75 is an increasing sequence in @xmath66 then @xmath76 . the @xmath77-algebra @xmath78 generated by @xmath66 is the collection @xmath79 of all borel sets of @xmath63 . separability from above _ : there exists an increasing sequence of finite subclasses @xmath80 ( @xmath81 ) of @xmath66 closed under intersections and a sequence of functions @xmath82 defined by @xmath83 and such that for each @xmath84 , @xmath85 . ( note : ` @xmath86 ' and`@xmath87 ' denote respectively the closure and the interior of a set . ) standard examples of indexing collections can be mentionned , such as rectangles @xmath13 $ ] of @xmath10 , arcs of the circle @xmath88 or lower layers . some of them are detailed in examples [ ex : index ] and [ ex : layers ] below . * distances on sets . * in order to study the hlder - continuity of set - indexed processes , we consider a distance on the indexing collection . along this paper , we may sometimes specify the distance on @xmath15 that we are using . among them , the following distances are of special interest : * the classical hausdorff metric @xmath89 defined on @xmath90 , the nonempty compact subsets of @xmath63 , by @xmath91 where @xmath92 ; * and the pseudo - distance @xmath93 defined by @xmath94 where @xmath65 is the measure on @xmath63 and @xmath95 denotes the symetric difference of sets . in the case of @xmath96;\;t\in{\mathbf{r}}^n_{+ } \right\}$ ] , @xmath97,[0,t])$ ] induces a distance on @xmath98 . this distance can be compared to the classical distances of @xmath10 , @xmath99 if @xmath65 is the lebesgue measure of @xmath10 , the distance @xmath100 is equivalent to @xmath101 , @xmath102 and @xmath103 on any compact of @xmath104 . + more precisely , for all @xmath105 in @xmath104 , there exist two positive constants @xmath106 and @xmath107 such that @xmath108;\quad m_{a , b}\ d_1(s , t)\leq m([0,s]\bigtriangleup [ 0,t ] ) \leq m_{a , b}\ d_{\infty}(s , t).\ ] ] we refer to @xcite for a proof of these assertions . * total boundedness of indexing collections . * as discussed in the introduction , the study of continuity of stochastic processes is closely related to the control of the metric entropy of the indexing collection . following the conditions of definition [ basic ] , some additional assumptions on the collection @xmath66 are required to guarantee that @xmath109 is totally bounded ( or at least locally totally bounded ) . we recall that a metric space @xmath110 is _ totally bounded _ if for any @xmath111 , @xmath63 can be covered by a finite number of balls of radius smaller than @xmath112 . the minimal number of such balls is called the _ metric entropy _ and is denoted by @xmath113 . before getting to the main assumption on the metric @xmath17 and the finite subclasses @xmath27 that approximate @xmath15 , we notice that the sequence @xmath114 is an increasing sequence that tends to @xmath115 , as @xmath58 . this property comes from condition _ ( 4 ) _ in definition [ basic ] . we will say that @xmath116 is _ admissible _ if : @xmath117 this should not appear as a restriction anyhow , because : if @xmath118 was going to @xmath115 too slowly , it would suffice to extract a subsequence ; and in the opposite situation , the gap between one scale to the other is too large and can then be filled with additional subclasses . * ( @xmath20).*[assumptdiscr2 ] let @xmath17 be a ( pseudo-)distance on @xmath66 . let us suppose that for @xmath119 , there exist positive real numbers @xmath25 and @xmath120 such that : 1 . for all @xmath121 , @xmath122 2 . and the collection @xmath26 is minimal in the sense that : setting for all @xmath121 and all @xmath123 , @xmath124 the sequence @xmath125 defined by @xmath126 for all @xmath127 satisfies @xmath128 the real @xmath25 is not unique and it depends _ a priori _ on the distance @xmath17 and the sub - semilattices @xmath18 . such a real @xmath25 is called _ discretization exponent _ of @xmath129 . note that if @xmath130 can be bounded independently of @xmath131 , then the last assumption is satisfied by admissibility of @xmath132 . without loss of generality , the distance @xmath17 can be normalized such that @xmath133 . the summability condition of assumption ( @xmath20 ) is close to the notion of entropy with inclusion developped by dudley @xcite in the context of empirical processes . on the contrary to the present work , @xcite focused exclusively on the sample path boundedness and continuity in the brownian case . the following example shows that assumption ( @xmath20 ) is satisfied in simple situations . [ ex : index ] * in the case of @xmath96;\;t\in [ 0,1]\subset{\mathbf{r}}_{+ } \right\}$ ] , the subclasses @xmath134 ( @xmath121 ) are commonly @xmath135;\;k=0,\dots,2^n \right\}$ ] . the two ( pseudo-)distances @xmath89 and @xmath136 ( where @xmath137 is the lebesgue measure on @xmath138 ) on @xmath66 are equal to @xmath139,[0,t])\mapsto |t - s|,\ ] ] and we have @xmath140,[0,(k+1).2^{-n}])=2^{-n}.\ ] ] then the two conditions of assumption ( @xmath20 ) both are satisfied for @xmath141 . * in the case of @xmath96;\;t\in [ 0,1]\subset{\mathbf{r}}^n_{+ } \right\}$ ] , the subclasses @xmath134 ( @xmath121 ) can be chosen as @xmath142;\;0\leq l_1,\dots , l_n\leq 2^n \right\}.\ ] ] let @xmath143 be a set in @xmath15 . the distance ( induced by the lebesgue measure @xmath137 ) between @xmath143 and @xmath144 is the volume difference between the two sets . it can be easily bounded from above by the sum of the volumes of the outer faces , minus a negligible residue @xmath145 since @xmath146 , this leads to @xmath147 with @xmath148 and the other condition of assumption ( @xmath20 ) are satisfied . on the contrary to the rectangles case , the following example shows that the collection of _ lower layers _ of @xmath10 does not satisfy assumption ( @xmath20 ) . we will see later that this result is not surprising in the view of theorem [ kolmth ] , since brownian motion indexed by lower layers of @xmath11 ^ 2 $ ] does not have a continuous modification , as can be seen for instance in @xcite . [ ex : layers ] let @xmath66 be the collection of lower layers of @xmath11 ^ 2 $ ] , i.e. the subsets @xmath149 of @xmath11 ^ 2 $ ] such that @xmath150 , @xmath13 \subseteq a$ ] . for all @xmath121 , let @xmath134 be the collection of finite unions of sets in the dissecting collection of the diadic rectangles of @xmath11 ^ 2 $ ] , i.e. @xmath151 : 2^n x \in \mathbf{z}^2 \cap ( 0,2^n]^2 \right\ } \cup \{0\ } \cup \{\emptyset \}.\ ] ] then , it can be shown that the cardinal @xmath132 of @xmath134 satisfies @xmath152 for all @xmath121 . for all @xmath153 , we can see that @xmath154 , hence there does not exist any @xmath25 such that @xmath155 and @xmath156 are of the same order . consequently the subclasses @xmath26 can not verify assumption ( @xmath20 ) . to conclude this section , we emphasize the fact that assumption ( @xmath20 ) implies the total boundedness of @xmath157 . since @xmath158 @xmath134 constitutes a @xmath156-net for all @xmath121 , and thus @xmath159 is totally bounded . as @xmath160 is not generally totally bounded , for any deterministic function @xmath161 , we consider the _ modulus of continuity _ on any totally bounded @xmath162 @xmath163 recall that the function @xmath164 is said _ hlder continuous of order @xmath165 _ if for all totally bounded @xmath162 one of the following equivalent conditions holds ( e.g. see @xcite , chapter @xmath166 ) a. @xmath167 b. there exists @xmath168 and @xmath169 such that for all @xmath170 with @xmath171 , @xmath172 . for any general set - indexed gaussian process , dudley s corollary 2.3 in @xcite allows to compute a modulus of continuity ( giving the same kind of result than following corollary [ gausskolm ] ) . this result holds under certain entropic conditions on the indexing collection , which are different from these of our setting . assumption ( @xmath20 ) and more precisely its second condition allows to prove a continuity criterion in the non - gaussian case . although adler and taylor @xcite emphasize that the gaussian property is only used through the exponential decay of the tail probability of the process in the proof of the previous results , they do not suggest any kolmogorov criterion for non - gaussian processes . the following theorem [ kolmth ] do so in the general set - indexed framework of ivanoff and merzbach , thanks to the discretization exponent . [ def : contract ] a ( pseudo-)distance @xmath17 on @xmath15 is said : a. [ cond : oc ] _ outer - continuous _ if for any non - increasing sequence @xmath173 in @xmath15 converging to @xmath174 , @xmath175 tends to @xmath176 as @xmath131 goes to @xmath115 ; b. [ cond : contract ] _ contractive _ if it is outer - continuous and if for any @xmath177 , @xmath178 the most important metrics in the context of set - indexed processes , @xmath100 and @xmath89 , are contractive . assumption ( @xmath20 ) on the subcollections @xmath129 and the contractivity of the metric @xmath17 allow to state the first important result of the paper : [ kolmth ] let @xmath17 be a contractive ( pseudo-)distance on the indexing collection @xmath66 , whose subclasses @xmath18 satisfy assumption ( @xmath20 ) with a discretization exponent @xmath179 . let @xmath180 be a set - indexed process such that @xmath181 \leq k\ { d_{\mathcal{a}}}(u , v)^{{q_{\underline{\mathcal{a}}}}+\beta}\ ] ] where @xmath182 , @xmath183 and @xmath184 are positive constants . + then , the sample paths of @xmath43 are almost surely locally @xmath185-hlder continuous for all @xmath186 , i.e. there exist a random variable @xmath33 and a constant @xmath34 such that almost surely latexmath:[\[\forall u , v\in\mathcal{a},\qquad { d_{\mathcal{a}}}(u , v)<h^ * \rightarrow let us fix @xmath186 and denote @xmath188 a countable dense subset of @xmath15 . first , let @xmath189 be any sequence of positive real numbers such that @xmath190 , and for @xmath191 such that @xmath192 , we have @xmath193 now applying successively tchebyshev s inequality , ( [ eq : hypoprocess ] ) and equation ( [ eq : assump1 ] ) of assumption ( @xmath20 ) , @xmath194 \\ & \leq k\ k_{n+1}^{\alpha \gamma/{q_{\underline{\mathcal{a } } } } } \sum_{j = n}^{\infty } a_j^{-\alpha } k_{j+1 } \sup_{v\in { \mathcal{a}}_{j+1 } } { d_{\mathcal{a}}}(v , g_j(v))^{{q_{\underline{\mathcal{a}}}}+ \beta } \\ & \leq k\ k_{n+1}^{\alpha \gamma/{q_{\underline{\mathcal{a } } } } } \ \sum_{j = n}^{\infty } a_j^{-\alpha } \frac{k_{j+1}}{k_j } k_{j}^{-\beta/{q_{\underline{\mathcal{a}}}}}.\end{aligned}\ ] ] the admissibility of @xmath118 implies that for @xmath195 , and for @xmath131 large enough ( depending on @xmath196 ) , @xmath197 , so that : @xmath198 since @xmath199 , @xmath200 can be chosen equal to @xmath201 ( which is indeed summable because @xmath132 is admissible ) , and then : @xmath202 which finally leads to , for @xmath203 , @xmath204 thus , this probability is summable and borel - cantelli s theorem implies the existence of @xmath205 with @xmath206 such that @xmath207 , @xmath208 where we used @xmath196 as in the previous paragraph . this is summable by ( [ eq : assump2 ] ) , hence there exists @xmath209 a measurable subset of @xmath210 of probability @xmath211 and @xmath212 a integer - valued finite random variable such that on @xmath209 : @xmath213 for any @xmath214 , there is a unique @xmath191 such that @xmath215 . let @xmath216 . without any restriction , we assume that @xmath217 . indeed , if this not the case , we shall consider @xmath218 , where @xmath219 by contractivity . since this implies that @xmath220 , we will write , on @xmath221 , for any @xmath222 : @xmath223 as a consequence of equations ( [ eq : bc1 ] ) and ( [ eq : bc2 ] ) . since @xmath221 is of probability @xmath211 , we have proved that there exist a constant @xmath34 and a random variable @xmath33 such that @xmath224 ) to the whole class @xmath66 . from the outer - continuity of @xmath17 , we can claim : + on @xmath225 , for all @xmath226 , for all @xmath143 and @xmath227 in @xmath66 with @xmath228 , there exists @xmath229 such that @xmath230 for all @xmath231 and @xmath232 . thus by ( [ eq : resultd ] ) , @xmath233 by * @xmath234 , @xmath235 , @xmath236 , * @xmath207 , * * @xmath237 , @xmath238 * * @xmath239 , @xmath240 . + applying ( [ resultg_n ] ) with @xmath241 , the outer - continuity property of @xmath17 implies that @xmath242 is a cauchy sequence and then converges in @xmath138 . the process @xmath243 satisfies almost surely latexmath:[\[\forall u , v\in\mathcal{a};\quad { d_{\mathcal{a}}}(u , v)<h^*\rightarrow moreover , * @xmath237 , @xmath245 almost surely . * @xmath239 , by construction , @xmath246 as @xmath58 . + since @xmath247 $ ] converges to @xmath176 when @xmath58 , the sequence @xmath248 converges in probability to @xmath249 . then , there exists a subsequence converging almost surely . + from these two facts , we get @xmath245 a.s . as in the multiparameter s case , a simpler statement holds for gaussian processes ( see @xcite for a detailed study of the kolmogorov criterion in the multiparameter frame ) . [ gausskolm ] let @xmath17 be a ( pseudo-)distance on the indexing collection @xmath66 , whose subclasses @xmath18 satisfy assumption ( @xmath20 ) . let @xmath180 be a centered gaussian set - indexed process such that @xmath250 \leq k \ { d_{\mathcal{a}}}(u , v)^{2\beta}\ ] ] where @xmath251 and @xmath252 . + then , the sample paths of @xmath43 are almost surely locally @xmath185-hlder continuous for all @xmath253 . for any @xmath254 , there exists a constant @xmath255 such that for all centered gaussian random variable @xmath256 , we have @xmath257=m_p\;({\mathbf{e}}\left[y^2\right])^p$ ] . then , @xmath258\leq k\ m_p\ { d_{\mathcal{a}}}(u , v)^{2p\beta}.\ ] ] for all @xmath253 , there exists @xmath254 such that @xmath259 , where @xmath25 is the discretization exponent of @xmath129 . by theorem [ kolmth ] the result follows . [ rem : hypgauss ] the proof of theorem [ kolmth ] shows that when condition ( [ eq : assump2 ] ) is removed from assumption ( @xmath20 ) , the conclusion remains true when the hypothesis ( [ eq : hypoprocess ] ) is strengthened in @xmath260 \leq k\ { d_{\mathcal{a}}}(u , v)^{2{q_{\underline{\mathcal{a}}}}+\beta}.\ ] ] the result follows from the simple estimation @xmath261 in equation ( [ eq : pairs ] ) . in that case , the validity of corollary [ gausskolm ] persists , since the integer @xmath262 can be chosen such that @xmath263 ( instead of @xmath259 ) . as previously mentioned , the brownian motion indexed by the lower layers of @xmath11 ^ 2 $ ] is discontinuous with probability one ( e.g. see theorem 1.4.5 in @xcite or @xcite ) . the previous theorem [ kolmth ] and corollary [ gausskolm ] do not contradict this fact , since the collection of lower layers of @xmath11 ^ 2 $ ] do not satisfy assumption ( @xmath20 ) according to example [ ex : layers ] in the specific case of the separating subclasses @xmath26 mentioned there . this latter result is improved by the following corollary of theorem [ kolmth ] . any subclasses @xmath26 satisfying condition ( 4 ) of definition [ basic ] for the indexing collection of lower layers of @xmath11 ^ 2 $ ] do not satisfy assumption ( @xmath20 ) . following the early work of dudley , the restriction of the set - indexed brownian motion to an indexing collection satisfying certain conditions can admit a continuous modification . we refer to @xcite for a modern survey of these results . in particular , the set - indexed brownian motion is a.s . continuous over any vapnik - ervonenkis class of sets ( see corollary 1.4.10 in @xcite ) , as the collection of rectangles of @xmath10 is an example . according to dudley s theorem ( see theorem 2.7.1 of chapter 5 in @xcite and also theorems 1.3.5 and 1.5.4 in @xcite ) , the existence of a continuous modification of a centered gaussian @xmath15-indexed process can be proved if @xmath264 is totally bounded and @xmath265 , where @xmath266 denotes the entropy function ( relative to the distance @xmath17 ) . following the continuity on processes indexed by vapnik - ervonenkis classes of sets and the role of assumption ( @xmath20 ) in theorem [ kolmth ] , we emphasize the fact that upper bounds for the entropy function can be obtained in the two cases . let us define @xmath267 let us also define , for @xmath268 $ ] , @xmath269 . + from condition ( [ eq : assump1 ] ) of assumption ( @xmath20 ) , for all @xmath22 , @xmath270 which implies @xmath271 . + we can see easily that : @xmath272 which allows to get a bound for the entropy function ( relative to the distance @xmath17 ) , @xmath273 in the case of a vapnik - ervonenkis class @xmath67 of sets in a measure space @xmath274 , the entropy function ( relative to the distance @xmath275 ) is bounded as : @xmath276 where @xmath182 and @xmath277 are positive constants ( e.g. see @xcite , theorem 1.4.9 ) . so far , we only considered simple _ increments _ of @xmath43 of the form @xmath278 for @xmath279 not necessarily ordered . however these quantities do not constitute the natural extension of the one - parameter @xmath2 ( @xmath280 ) to multiparameter ( e.g. @xcite ) and set - indexed ( e.g. @xcite ) settings , particularly when increment stationarity property is concerned . this section is devoted to usual increments of set - indexed processes , which extend the rectangular increments of multiparameter processes . let us define , for any given indexing collection @xmath15 , the collection @xmath36 of subsets of @xmath63 , defined as @xmath281 this collection is used to index the process @xmath282 , defined by @xmath283 for @xmath284 , where @xmath285 is given by the _ inclusion - exclusion _ formula @xmath286 the existence of the increment process @xmath282 indexed by @xmath36 requires that for any @xmath287 , the value @xmath288 does not depend on the representation of @xmath289 . [ corcl ] under the hypotheses of theorem [ kolmth ] and if the distance @xmath17 on the class @xmath15 is assumed to be contractive , for each fixed integer @xmath290 , for all @xmath32 , there exist a random variable @xmath291 and a constant @xmath34 such that , with probability one , @xmath292 . corollary [ corcl ] , as a result on the class @xmath293 where @xmath294 , does _ not _ extend to the whole @xmath295 , as the following example shows . the next result is an adaptation of an example in @xcite to the set - indexed setting . it states that the brownian motion can be unbounded on @xmath36 when @xmath15 is the collection of rectangles of @xmath11 ^ 2 $ ] . [ prop : exbm ] let @xmath296 be a brownian motion indexed by the borelian sets of @xmath11 ^ 2 $ ] , i.e. a centered gaussian process with covariance structure @xmath297 = \lambda ( c \cap c'),\quad\forall c , c ' \in \mathcal{b}([0,1]^2)\ ] ] where @xmath137 denotes the lebesgue measure . + let @xmath15 be the collection of rectangles of @xmath11 ^ 2 $ ] . in the sequel , we consider the restriction on the class @xmath36 , related to @xmath15 , of the brownian motion defined above . + then for all @xmath298 , all @xmath168 , and for almost all @xmath299 , there exist sequences of sets @xmath300 , @xmath301 in @xmath36 such that @xmath302 and for @xmath131 big enough , @xmath303 without any stronger condition than assumption ( @xmath20 ) on the sub - semilattices @xmath129 , the previous example of set - indexed brownian motion dismisses a possible definition of the hlder continuity for stochastic processes of the form : @xmath304 back to the beginning of section [ sec : kolm ] , localizing the two expressions ( i ) and ( ii ) for hlder - continuity leads to two different notions . indeed , for the distance @xmath17 on @xmath66 , if @xmath48 ( or simply @xmath305 if the context is clear ) denotes the open ball centered in @xmath306 and whose radius is @xmath307 , we get 1 . @xmath308 2 . there exist @xmath168 and @xmath169 such that latexmath:[\[\forall u , v\in b_{{d_{\mathcal{a}}}}(u_0,\delta_0),\quad although the conditions ( i ) and ( ii ) are equivalent , localizing around @xmath306 only gives ( ii)@xmath310(i)@xmath311 . this leads usually to consider two kinds of hlder exponent at @xmath306 : * the pointwise hlder exponent @xmath312 * and the local hlder exponent @xmath313 each one allows to measure the regularity of the function @xmath164 . in general , we have @xmath314 but the inequality can be strict . we can see that condition ( i)@xmath311 is equivalent to @xmath315 , and condition ( ii)@xmath311 is equivalent to @xmath316 . then @xmath317 is another statement for ( ii)@xmath310(i)@xmath311 . note that the discussion of this whole paragraph is not specific to indexing collection , but can be adapted to any totally bounded metric space . consider the case of the metric space @xmath318 . fix @xmath319 and @xmath195 . let @xmath164 be a chirp function defined by @xmath320 . the two hlder exponents at @xmath176 can be computed and @xmath321 . this example shows that the sole pointwise exponent is not sufficient to describe the irregularity of the function . the local exponent _ can see _ the oscillations around @xmath176 , while the pointwise exponent can not . these two notions can be applied to study sample path regularity of a stochastic process . in the case of gaussian processes ( see @xcite ) , we define the _ deterministic pointwise hlder exponent _ @xmath322 ^ 2}{\rho^{2\sigma}}<\infty \right\}\ ] ] and the _ deterministic local hlder exponent _ @xmath323 ^ 2}{{d_{\mathcal{a}}}(u , v)^{2\sigma}}<\infty \right\}.\ ] ] on the space @xmath324 , it is shown in @xcite that for all @xmath325 , the pointwise and local hlder exponents of @xmath43 at @xmath326 satisfy almost surely @xmath327 in the following sections , several other definitions are studied for hlder regularity of set - indexed processes . they are connected to the various ways to study the local behaviour of the sample paths of @xmath43 around a given set @xmath306 . following expression ( [ eq : inc ] ) for the definition of the increments of a set - indexed process , we consider alternative definitions for hlder exponents , where the quantities @xmath0 are substituted with @xmath329 . as stated in section [ sec : cinc - kolm ] , it is not wise to consider @xmath329 when @xmath84 and @xmath330 are close to a given @xmath306 . indeed , proposition [ prop : exbm ] shows that the quantity @xmath331 can stay far away from @xmath176 when @xmath332 is small , even in the simple case of a brownian motion indexed by @xmath11 ^ 2 $ ] . however , when @xmath84 and @xmath227 is restricted to sets of the form @xmath333 where @xmath334 is fixed and @xmath335 , the hlder regularity can be defined from the study of @xmath329 . fix any integer @xmath290 and set for all @xmath22 and @xmath307 , @xmath336 the pointwise and local hlder @xmath337-exponents at @xmath49 are respectively defined as @xmath338 and @xmath339 the following result shows that the @xmath337-exponents do not depend on @xmath334 and , consequently , they provide a definition of hlder exponents on the class @xmath60 . moreover , these exponents can be compared to the exponents defined by ( [ eq : pointholder ] ) and ( [ eq : localholder ] ) . [ propcomparaisonalphas ] if @xmath17 is a contractive distance , for any @xmath306 , the exponents @xmath340 and @xmath341 do not depend on the integer @xmath290 . they are denoted by @xmath342 and @xmath343 respectively . moreover , for all @xmath306 and all @xmath299 , @xmath344 we only detail the case of the pointwise exponent . the proof for the local exponent is totally similar . + from the definition of the @xmath337-exponents , since @xmath345 implies @xmath346 , it is clear that @xmath347 for the sake of readability , we prove the converse inequality for @xmath348 ( the other cases are very similar ) . for any @xmath307 , let @xmath349 , and @xmath350 with @xmath351 . from the inclusion - exclusion formula , @xmath352 we have @xmath353 , @xmath354 and @xmath355 using @xmath356 from the contracting property of @xmath17 . + then , for all @xmath357 , @xmath358 which says that @xmath359 . thus , @xmath360 . + this inequality achieves to prove that @xmath361 does not depend on the integer @xmath290 . to prove the second part of the proposition , it suffices then to prove the inequality for @xmath362 . this is straightforward , since for a fixed @xmath349 , @xmath363 hence @xmath364 . the inequality for the local exponent can be obtained identically , or one can notice that it is a direct consequence of corollary [ corcl ] . the converse inequality does not hold in general since quantities @xmath365 can not be obtained from the increment process @xmath282 when @xmath366 are not ordered . the previous definition of the pointwise hlder exponent on @xmath328 is not equivalent to the quantity @xmath367 as the following example shows . + in the particular case of the indexing collection @xmath15 equal to the rectangles of @xmath368 and the distance @xmath369 induced by the lebesgue measure @xmath137 of @xmath370 , we show that the assertion @xmath371 is not equivalent to @xmath372 . consider @xmath373 , v_2 = [ 0;(n^2+\frac{1}{n},n^2)]$ ] and @xmath374 $ ] . we have @xmath375 then , @xmath376 for small @xmath377 and it is not possible to control the quantity @xmath378 using @xmath379 , @xmath379 and @xmath380 as was done in the previous proofs . the notation @xmath381 must be considered with care : proposition [ prop : exbm ] shows that the hlder exponents can not be defined directly by taking the supremum on @xmath84 and @xmath330 with @xmath382 and @xmath383 ( and then , on the class @xmath60 ) . this is the reason why the set @xmath227 is restricted to be in @xmath384 . the arguments of the proof of proposition [ propcomparaisonalphas ] in the particular case of @xmath362 leads to : for all @xmath385 , @xmath386 and @xmath387 the converse inequalities follow from the fact that the set of @xmath388 with @xmath389 is included in the set of @xmath390 and @xmath391 . then , we can state : [ cor : expc ] if @xmath17 is a contractive distance , the pointwise and local hlder @xmath60-exponents at @xmath49 are respectively given by @xmath392 and @xmath393 as previously mentioned , the set - indexed brownian motion can be not continuous , when the indexing collection is not a vapnik - ervonenkis class ( see @xcite for the detailed study ) . in @xcite , a weak form of continuity is considered in the study of set - indexed poisson process , set - indexed brownian motion and more generally set - indexed lvy processes . in particular , the sample paths of the set - indexed brownian motion are proved to be pointwise continuous as a set - indexed lvy process with gaussian increments . notice that such a property does not require assumption ( @xmath20 ) on @xmath15 . we recall the following definitions : the _ point mass jump _ of a set - indexed function @xmath394 at @xmath395 is defined by @xmath396 and for each @xmath127 , @xmath397 denotes the collection of subsets @xmath398 with @xmath399 and @xmath400 . a set - indexed function @xmath394 is said _ pointwise continuous _ at @xmath395 if @xmath401 . let us recall that a subset @xmath402 of @xmath66 which is closed under arbitrary intersections is called a _ lower sub - semilattice _ of @xmath66 . the ordering of a lower sub - semilattice @xmath403 is said to be _ consistent _ if @xmath404 . proceeding inductively , we can show that any lower sub - semilattice admits a consistent ordering , which is not unique in general ( see @xcite ) . if @xmath405 is a consistent ordering of a finite lower sub - semilattice @xmath402 , the set @xmath406 is called _ the left neighbourhood _ of @xmath407 in @xmath402 . since @xmath408 , the definition of the left neighbourhood does not depend on the ordering . as in the classical kolmogorov criterion of continuity , the pointwise continuity of a set - indexed process @xmath43 can be proved from the study of @xmath409 $ ] when @xmath131 goes to infinity . [ prop : crit - pc - unif ] let @xmath180 be a set - indexed process and let @xmath410 be a subset in @xmath15 such that @xmath411 and assume that there exist @xmath412 , @xmath413 , @xmath414 and @xmath251 such that for all @xmath415 and all @xmath416 , @xmath417 \leq k\ m(c_n(t))^{q } \ .\ ] ] then , for any @xmath418 , there exists an increasing function @xmath419 and a random variable @xmath420 satisfying , with probability one , @xmath421 , we assume in this proof that the indexing collection @xmath15 is included in @xmath410 . + for all @xmath422 , we consider @xmath423 , where @xmath424 is defined in ( [ eq : jtx ] ) . when @xmath425 ranges @xmath410 , the subset @xmath424 ranges @xmath426 , the collection of the disjoint left - neighbourhoods of @xmath19 . consequently we can write @xmath427 . for any integer @xmath428 , we have @xmath429}{m(c)^{\gamma p } } \leq k\ \sum_{c\in { \mathcal{c}}^l({\mathcal{a}}_n ) } m(c)^{q-\gamma p } \ .\end{aligned}\ ] ] since @xmath430 , we have @xmath431 where the fact that the @xmath432 are disjoint is used . up to choosing an extraction @xmath433 for the sequence @xmath434 , we can assume that @xmath435 is summable . hence the borel - cantelli lemma implies that for @xmath436 , @xmath437 happens infinitely often , which gives the result . proposition [ prop : crit - pc - unif ] does not require assumption ( @xmath20 ) for the collection @xmath438 and the distance @xmath100 . from proposition [ prop : crit - pc - unif ] , it is natural to define local hlder regularity of a set - indexed process by a comparison of @xmath56 to quantities @xmath57 with @xmath165 , when @xmath131 is large . the _ pointwise continuity hlder exponent _ at any @xmath395 is defined by @xmath439 according to proposition [ prop : crit - pc - unif ] , if @xmath43 is a @xmath15-indexed process satisfying hypothesis ( [ eq : hypoprocess3 ] ) , then with probability one , @xmath440 for all @xmath441 . as in the continuity criterion ( theorem [ kolmth ] and corollary [ gausskolm ] ) , the proof of proposition [ prop : crit - pc - unif ] can be improved for @xmath442 for any @xmath443 , when the process is gaussian . in that specific case , the upper bound for admissible values of @xmath185 is @xmath444 ( instead of @xmath445 ) . in this section , we consider the concept of _ flow _ , which is a useful tool to reduce characterization or convergence problems to a one - dimensional issue . _ flows _ have been used to characterize : strong martingales @xcite , set - markov processes @xcite , set - indexed fractional brownian motion @xcite and set - indexed lvy processes @xcite . [ flowdef ] an _ elementary flow _ is defined to be a continuous increasing function @xmath446\subset\mathbf{r}_+\rightarrow\mathcal{a}$ ] , i.e. such that @xmath447;\quad & s < t \rightarrow f(s)\subseteq f(t)\\ \forall s\in [ a , b);\quad & f(s)=\bigcap_{v > s}f(v)\\ \forall s\in ( a , b);\quad & f(s)=\overline{\bigcup_{u < s}f(u)}.\end{aligned}\ ] ] a _ simple flow _ is a continuous function @xmath446\rightarrow\mathcal{a}(u)$ ] such that there exists a finite sequence @xmath448 with @xmath449 and elementary flows @xmath450\rightarrow\mathcal{a}$ ] ( @xmath451 ) such that @xmath452;\quad f(s)=f_i(s)\cup \bigcup_{j=1}^{i-1}f_j(t_j).\ ] ] the set of all simple ( resp . elementary ) flows is denoted @xmath453 ( resp . @xmath454 ) . according to @xcite , we use the parametrization of flows which allows to preserve the increment stationarity property under projection on flows ( it avoids the appearance of a time - change ) . for any set - indexed process @xmath455 on the space @xmath456 and any simple flow @xmath446\rightarrow\mathcal{a}(u)$ ] , the _ @xmath65-standard projection of @xmath43 on @xmath164 _ is defined as the process @xmath457)\right\},\ ] ] where @xmath458 is the function @xmath459 $ ] and @xmath460 its right inverse . the importance of flows in the study of set - indexed processes follows the fact that the finite dimensional distributions of an additive @xmath461-indexed process @xmath43 determine and are determined by the finite dimensional distributions of the class @xmath462 ( @xcite , lemma 6 ) . as the projection of a set - indexed process on any flow is a real - parameter process , its classical hlder exponents can be considered and compared to the exponents of the set - indexed process . in the sequel , we study how regularity of flows connects the exponents @xmath463 ( resp . @xmath464 ) and @xmath465 ( resp . @xmath466 ) , when @xmath49 and @xmath467 . for any @xmath468 , let us denote by @xmath469 the subset of @xmath470 containing all the simple flows @xmath471 such that there exists @xmath472 satisfying @xmath473 , and where @xmath474 is a closed interval of @xmath475 containing a ball centered in @xmath326 . such a @xmath326 does not depend on the flow @xmath164 , since @xmath476 . in the same way , we define @xmath477 for elementary flows . [ lem : flow ] let @xmath478 and @xmath479 such that @xmath480 . for all @xmath481 , @xmath482 . as @xmath458 is increasing , @xmath460 is increasing as well . we assume without loss of generality that @xmath484 . then , @xmath485 using lemma [ lem : flow ] , we can compare the hlder regularity of @xmath43 and the hlder regularity of its projections on flows . let @xmath486 be a set - indexed process on @xmath456 , with finite hlder exponents at @xmath487 . then , @xmath488 where the metric considered on @xmath15 is @xmath100 . the proof is only given for the pointwise hlder exponent . the case of the local hlder exponent is totally similar . + from proposition [ propcomparaisonalphas ] , the inequality @xmath489 for all @xmath299 is already known . the equality @xmath490 follows from corollary [ cor : expc ] and lemma [ lem : flow ] . the natural question is then to wonder if the previous inequality could be improved in an equality . the answer is generally no , as the following example shows . in this example , we only consider deterministic functions , instead of random processes . let @xmath491 be a set - indexed function on @xmath15 , the usual collection of rectangles of @xmath11 ^ 2 $ ] . let @xmath487 and assume that @xmath491 is @xmath183-hlder continuous in @xmath492 , for some @xmath493 . we assume without loss of generality that @xmath494 . + let us divide @xmath15 into four quadrants around @xmath495 $ ] in the following manner : @xmath496 \in { \mathcal{a } } : x\leq x_0 \ and \ y < y_0\ } , \\ \mathcal{q}_2 & = \{[0,(x , y ) ] \in { \mathcal{a } } : x\leq x_0 \ and \ y\geq y_0\ } , \\ \mathcal{q}_3 & = \{[0,(x , y ) ] \in { \mathcal{a } } : x > x_0 \ and \ y\geq y_0\ } , \\ \mathcal{q}_4 & = \{[0,(x , y ) ] \in{\mathcal{a } } : x > x_0 \ and \ y < y_0\}.\end{aligned}\ ] ] [ fig : functiong ] let us fix @xmath497 . as @xmath491 is @xmath183-hlder continuous at @xmath492 , for all @xmath251 , there exists a sequence of sets in @xmath15 converging to @xmath492 and such that @xmath498 . up to a rotation , assume @xmath499 is this quadrant . we now assume ( without restriction ) that a subsequence of @xmath500 belongs to a closed subset @xmath501 ( see figure [ fig : functiong ] ) . let @xmath502 be a smooth function except maybe at @xmath492 , taking its values in @xmath11 $ ] and such that @xmath503 and @xmath504 for all @xmath505 ( see figure [ fig : functiong ] above ) , and @xmath506 for all @xmath507 . we denote by @xmath508 the product of @xmath491 and @xmath502 . + up to an extraction that we detailed previously , the sequence @xmath173 belongs to @xmath509 . then , latexmath:[\[\forall n\geq 0,\quad thus , if @xmath508 is @xmath184-hlder continuous at @xmath492 , then the inequality @xmath511 holds necessarily . for @xmath512 , there exist @xmath513 and @xmath251 such that latexmath:[\ ] ] then , the same development as the previous proof gives @xmath704 for all @xmath705^n$ ] . consequently , we can state : @xmath688^n , \quad { \mathbf{p}}\left({\texttt{\large $ \boldsymbol{\alpha}$}}_{{\mathbf{b}}^h}^{pc}(t_0 ) = \mathbb{\bbalpha}_{{\mathbf{b}}^h}^{pc}(t_0)= h \right)=1,\ ] ] and , @xmath689^n , \quad { \texttt{\large $ \boldsymbol{\alpha}$}}_{{\mathbf{b}}^h}^{pc}(t_0 ) \geq \mathbb{\bbalpha}_{{\mathbf{b}}^h}^{pc}(t_0)= h \right ) = 1.\ ] ] in order to prove corollary [ corcl ] , we need the following lemma : [ lem : cl ] if the distance @xmath17 on the class @xmath66 is contracting , then for @xmath706 , @xmath707 moreover , for any integer @xmath290 and for all @xmath708 , @xmath709 for some constant @xmath710 which only depends on @xmath334 . the proof relies on the triangular inequality and the contracting property of @xmath17 . assuming that @xmath711 can be extended to @xmath712 in the following way : @xmath713 the following inequality holds : @xmath714 since for all @xmath715 , @xmath716 one can express @xmath717 now assume that @xmath718 . when @xmath719 , the number of terms in the right side of inequality ( [ eq : delta - gn ] ) is bounded by a constant , independent of the set @xmath720 . thus , there exists a positive constant @xmath721 such that @xmath722 using the same sequence @xmath189 as in the proof of theorem [ kolmth ] , and the above equation ( [ eqchain4 ] ) in the third inequality below : @xmath723 we obtain the same expression ( [ eq : proofkolm1 ] ) that we had in the proof of theorem [ kolmth ] , thus the same conclusion holds : if @xmath724 , then almost surely , @xmath725 implies that : @xmath726 in the same way , the second term of the upper bound ( [ eq : maj - deltaxc ] ) is proved to be bounded by @xmath727 , where @xmath728 only depends on @xmath185 and @xmath25 . + the first term of ( [ eq : maj - deltaxc ] ) can be bounded by a finite sum ( whose number of terms only depends on @xmath334 ) of the form @xmath729 , where @xmath730 for @xmath731 : latexmath:[\[\label{eqc } finally , if @xmath724 , condition ( [ eq : assump1 ] ) of assumption ( @xmath20 ) and lemma [ lem : cl ] imply @xmath733 hence , theorem [ kolmth ] implies that when @xmath734 , each term of equation ( [ eqc ] ) is bounded by a quantity proportional to @xmath735 . then , the random variable @xmath291 of the statement can be chosen to be @xmath736 and the result follows . a lower bound for the local hlder exponent is directly given by corollary [ gausskolm ] . + for all @xmath306 and all @xmath737 , there exists @xmath738 and @xmath251 such that @xmath739 \leq k\ { d_{\mathcal{a}}}(u , v)^{2\alpha}.\ ] ] therefore , the sample paths of @xmath43 are almost surely @xmath740-hlder continuous in @xmath741 for all @xmath742 , which leads to @xmath743 almost surely . then we get @xmath744 by inequality ( [ ineqholder ] ) , any lower bound for the local hlder exponent is also a lower bound for the pointwise exponent . moreover it can be improved in the case of strict inequality @xmath745 . for any @xmath111 , there exist @xmath746 and @xmath168 such that @xmath747 \leq m\ \rho^{\epsilon}.\ ] ] then setting @xmath748 , the exponential inequality for the centered gaussian variable @xmath0 implies @xmath749}\right ) \leq \exp\left ( -\frac{1}{2 } m \rho^{\epsilon } \right).\ ] ] we consider the particular case @xmath750 for @xmath121 large enough . using the above estimate in the proof of theorem [ kolmth ] still leads to equation ( [ eq : finchain ] ) , where we had that on @xmath225 , for all @xmath751 : @xmath752 [ lem_uppoint ] let @xmath759 be a centered gaussian process . assume that for @xmath306 , there exists @xmath760 such that for all @xmath111 , there exist a sequence @xmath761 of @xmath66 converging to @xmath492 , and a constant @xmath551 such that @xmath762 \geq c\ ; { d_{\mathcal{a}}}(u_n , u_0)^{2\mu+\epsilon}.\ ] ] then , we have almost surely @xmath763 [ lem_upbound ] let @xmath759 be a centered gaussian process . assume that for @xmath306 , there exists @xmath760 such that for all @xmath111 , there exist two sequences @xmath761 and @xmath766 of @xmath66 converging to @xmath492 , and a constant @xmath551 such that @xmath767 \geq c\ ; { d_{\mathcal{a}}}(u_n , v_n)^{2\mu+\epsilon}.\ ] ] then , we have almost surely @xmath768 this section is devoted to the proof of theorem [ th_unifalmostsure ] . we only consider the local hlder exponent . the uniform almost sure lower bound for the pointwise exponent is proved in a similar way . starting with the lower bound , from theorem [ kolmth ] , for all @xmath487 and all @xmath111 , there is a modification @xmath771 of @xmath43 which is @xmath183-hlder continuous for all @xmath772 on @xmath773 . * in the first step , @xmath774 is assumed to be constant over @xmath15 . hence the local hlder exponent of @xmath771 satisfies almost surely @xmath775 + the collection @xmath15 is totally bounded , so it can be covered by a countable number of balls of radius at most @xmath776 , for all @xmath574 . let @xmath777 be one of these balls . for all @xmath487 , we consider @xmath738 such that ( [ eq : uniflocholder ] ) holds . we have obviously @xmath778 for each open ball , there exists an integer @xmath131 such that @xmath779 so that for @xmath780 , there exists an integer @xmath781 such that @xmath782 . thus @xmath783 where the union is countable . each of these balls satisfies @xmath784 and since @xmath66 is a countable union of balls @xmath785 , we get @xmath786 taking @xmath787 , we conclude that @xmath788 * in the general case of a not constant exponent @xmath774 , for any ball @xmath777 of radius @xmath776 previously introduced , we set @xmath789 , @xmath111 . then , there exists a constant @xmath635 such that @xmath790 \leq c\ { d_{\mathcal{a}}}(u , v)^{2\beta}.\ ] ] in a similar way as we proved ( [ eq_111 ] ) , we deduce the existence of an event @xmath791 of probability one such that for all @xmath792 : @xmath793 by letting @xmath58 , the previous equation leads to @xmath794 using the fact that @xmath795 is countable , lemma [ lem_upbound ] gives @xmath796 let @xmath797 be the set of @xmath385 , such that @xmath798 for all @xmath799 . let @xmath800 . let @xmath801 be a sequence in @xmath67 converging to @xmath492 . on @xmath802 , @xmath803 , for all @xmath804 . for each fixed @xmath804 , there exist two sequences @xmath805 and @xmath806 in @xmath15 converging to @xmath807 as @xmath58 , and for all @xmath121 , @xmath808 as in @xcite , we build two other sequences @xmath809 and @xmath810 so that @xmath811 and @xmath812 and @xmath813 this implies the expected inequality for all @xmath487 . + the general case for @xmath774 not constant is proved in the same way as for the lower bound . y. xiao . sample path properties of anisotropic gaussian random fields . in : _ a minicourse on stochastic partial differential equations , ( d. khoshnevisan and f. rassoul - agha , editors ) _ , lecture notes in math . 1962 , pp . 145212 , springer , new york , 2009 .

we present a comprehensive view of the w51b hii region complex and the w51c supernova remnant using new radio observations from the vla , vlba , merlin , jcmt , and cso along with archival data from _ spitzer _ , _ rosat _ , _ asca _ , and _ chandra_. our vla data include the first cm ( 74 mhz ) continuum image of w51 at high resolution ( ) . the 400 cm image shows non - thermal emission surrounding the g49.2 - 0.3 hii region , and a compact source of non - thermal emission ( w51b_nt ) coincident with the previously - identified oh ( 1720 mhz ) maser spots , non - thermal 21 and 90 cm emission , and a hard x - ray source . w51b_nt falls within the region of high likelihood for the position of tev-ray emission . using the vlba three oh ( 1720 mhz ) maser spots are detected in the vicinity of w51b_nt with sizes of 60 to 300 au and zeeman effect magnetic field strengths of 1.5 to 2.2 mg . the multiwavelength data demonstrate that the northern end of the w51b hii region complex has been partly enveloped by the advancing w51c snr and this interaction explains the presence of w51b_nt and the oh masers . this interaction also appears in the thermal molecular gas which partially encircles w51b_nt and exhibits narrow pre - shock ( km s ) and broad post - shock ( km s ) velocity components . radex radiative transfer modeling of these two components yield physical conditions consistent with the passage of a non - dissociative c - type shock . confirmation of the w51b / w51c interaction provides additional evidence in favor of this region being one of the best candidates for hadronic particle acceleration known thus far .

1305.2793

star formation triggered by the interaction of supernova remnants ( snrs ) with their parent molecular clouds has long been thought to play an important role in the production of new generations of stars . indeed , it is thought that the enrichment of our own solar system with heavy elements might have resulted from such an interaction . in practice such interactions are very difficult to observe due to the extreme kinematic complexity present toward the inner galactic plane where most massive stars and snrs are found . over the last two decades it has been recognized that when oh ( 1720 mhz ) masers are found toward snrs they act as signposts for snr / molecular cloud interactions ( see for example * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? to date , oh ( 1720 mhz ) masers have been found in @xmath6 24 snrs , or 10% of the known snrs in our galaxy @xcite . indeed , all 24 oh ( 1720 mhz ) maser snrs can be found in the @xcite catalogue of 34 snrs thought to be interacting with associated molecular clouds . the observational properties of these snr masers are quite different from those of hii region oh ( 1720 mhz ) masers . snr oh ( 1720 mhz ) masers have : ( 1 ) larger maser spot sizes ; ( 2 ) narrow and simple line profiles ; ( 3 ) low levels of circular polarization ( typically @xmath7 ) ; ( 4 ) low magnetic field strengths ( @xmath8 mg ) ; ( 5 ) relatively non - variable flux densities ; ( 6 ) a notable lack of any other maser species ; and ( 7 ) low luminosity @xcite . these observational facts can be explained if oh ( 1720 mhz ) snr masers originate in the post - shock molecular gas behind c - type shocks , and are collisionally pumped in contrast to their radiatively pumped hii region counterparts @xcite . current theories further suggest that the collisional pump is most efficient for post - shock densities of @xmath9 @xmath10 and temperatures in the range 50 k @xmath11 125 k , and that the abundance of oh must be enhanced by the dissociation of water due to x - rays and/or @xmath2-rays emanating from the snr @xcite . observations of the physical conditions in the pre- and post - shock gas available to date are in good agreement with these theoretical expectations ( see e.g. * ? ? ? * ; * ? ? ? prior to this work , however , there have been few in - depth studies employing multi - transition , multi - species analyses . interestingly , a number of recent publications have also noted the growing number of @xmath2-ray detected snrs that are interacting with nearby molecular clouds . * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * to name a few ) . indeed , many of the @xmath2-ray snrs are associated with well known oh ( 1720 mhz ) sources : e.g. g349.7 + 0.2,ctb 37a , 3c391 , w44 , ic443 and w51c . hence , as a powerful probe of snr / molecular cloud interactions , observations of this oh maser line can potentially pinpoint sites of particle acceleration . of the known snr oh ( 1720 mhz ) masers , the origin of those observed toward the w51 complex remain among the most poorly understood @xcite . w51 is composed of two large hii region complexes w51a ( also known as w51 main ) and w51b , as well as the snr w51c ( see the 90 and 400 cm images in figure [ fig1]a , b ) . w51 is located near the sagittarius arm tangent point at @xmath12 and @xmath13 and thus velocity crowding in this direction has made distance determinations ( and relative orientations along the line of sight ) for the three w51 components ( a , b , and c ) uncertain . in the past , their distances were considered to be in the range of 5 - 7 kpc @xcite . a maser parallax distance has recently been measured for w51a of 5.4 ( @xmath14 ) kpc @xcite . additionally , the soft x - ray absorption seen toward the w51b region compared to the rest of the w51c snr suggests that w51b lies in front of w51c @xcite . however , it remains uncertain whether w51a is in front of , co - distant with , or behind w51b . the oh ( 1720 mhz ) masers reported by @xcite are located @xmath15 west of the w51b hii region g49.2 - 0.3 . subsequent observations of these masers by @xcite demonstrated that these masers exhibit all the hallmarks of snr oh ( 1720 mhz ) masers . like green et al . , these authors conclude that these masers are most likely excited by the interaction between the w51c snr ( the only known snr in the region ) and the molecular gas associated with the w51b hii region . however , the exact nature of this interaction remains unclear especially since snr oh ( 1720 mhz ) maser theory suggests that a sufficient column of oh with the necessary velocity coherence can only be achieved in shocks that move ( more or less ) perpendicular to the line of sight ( see for example * ? ? ? it is difficult to reconcile the need for an edge - on shock with the fact that w51b lies in front of the w51c snr ( i.e. it seems that any shock would propagate with a substantial component along the line of sight ) . this paper describes a multiwavelength investigation of the w51b / c oh ( 1720 mhz ) masers and their environment spanning ten decades in wavelength ( 400 cm to x - rays ) and six decades of angular resolution ( 10 mas to @xmath16 ) . the goal of this study is to better understand the properties of the maser emission through high resolution merlin and very long baseline array ( vlba ) observations , as well as to understand the physical conditions that have led to the maser emission using very large array ( vla ) 400 cm , 90 cm , and 20 cm data , james clerk maxwell telescope ( jcmt ) co(32 ) and @xmath17co(21 ) data , and caltech submillimeter observatory ( cso ) hco@xmath18(32 ) and hcn(32 ) data . for comparison , we also make use of archival _ spitzer _ mid - ir 8 @xmath19 m data , as well as _ , _ rosat _ , and _ asca _ x - ray data . the observing parameters for all of the observations described in this paper are listed in table [ observing ] . additional details are provided below . the oh ( 1720 mhz ) masers were observed for approximately 18 hours ( 15 hrs on source ) over the course of three days using the six antennas of the merlin array ( lovell was not used ) . the doppler tracking corrections , in the lsr convention , were applied in the correlator . in addition to right circular and left circular polarization data , the rl and lr cross correlation data were also recorded . absolute flux and bandpass calibration were carried out using observations of 3c 84 . initial phase calibration was obtained from periodic observations of j1922 + 1530 . instrumental polarization leakage corrections were derived from the unpolarized source 3c84 , and 3c286 was used to set the absolute polarization angle . the data were processed using the aips software package . after initial amplitude , delay , and phase calibration , several iterations of self - calibration on the strongest ( w51_2 ) maser channel were carried out and subsequently applied to the full spectral line dataset . the oh ( 1720 mhz ) masers were observed for approximately 12 hours ( 6 hrs on source ) over the course of two days using the ten antennas of the vlba plus one vla ( y1 ) antenna . both right circular and left circular polarization data were recorded . all calibration and imaging for these data were carried out using the aips software . absolute flux calibration was determined from periodic system temperature measurements coupled with existing antenna gain tables . bandpass and delay calibration were performed using observations of the strong continuum source j2253 + 1608 ( 3c 454.3 ) . in order to obtain accurate absolute positions and sensitivity to potentially weak maser emission , these observations were phase referenced . since no calibration source suitable for vlba observations was known at that time , a short vlba survey of near - by vla calibrators was carried out prior to the target observations . from this survey , the jvas source j1911 + 1611 ( @xmath20 , @xmath21 with an uncertainty of @xmath22 mas ) had the strongest compact emission and was used to phase reference the maser data with a cycle time of 2 minutes ( this source is now available in the vlba calibrator catalog ) . corrections for doppler tracking of the target data were applied offline in aips . after initial amplitude , delay , and phase calibration , several iterations of self - calibration on the strongest ( w51_2a ) maser channel were carried out and subsequently applied to the full line dataset . we have obtained vla observations from the b+c configurations at 400 cm and a+b+c+d configurations at 90 cm ( see table 1 for details ) . the fwhp primary beam size of the 90 cm and 400 cm data are @xmath23 and @xmath24 , respectively . the data were reduced and imaged using the low frequency wide - field imaging techniques described in ( * ? ? ? * and references therein ) . gain , bandpass , and absolute flux calibration for the 90 cm data were carried out using j1924 + 334 , cygnus a , and 3c 48 respectively . in order to best match the 400 cm data a 90 cm image was constructed using only the b+c+d configurations and the multiscale clean algorithm in aips , with scales of @xmath25 ( i.e. the intrinsic beam given the combination of data and weighting ) , @xmath26 , and @xmath27 . to achieve the highest angular resolution , a 90 cm image was also constructed from the a+b+c+d configuration data and multiscale clean with scales of @xmath28 ( i.e. the intrinsic beam given the combination of data and weighting ) , @xmath29 , and @xmath30 . absolute flux , bandpass , and coarse phase calibration at 400 cm were all obtained from observations of cygnus a ( located @xmath31 distant ) . of particular note is the lack of accurate absolute position information for the 400 cm data due to the large distance to the 400 cm phase calibrator . typically this problem is solved by using a 90 cm image as an initial model for the 400 cm self - calibration . however , in the case of the w51 region , much of the total 90 cm flux is dominated by hii regions making it unsuitable for a 400 cm model ( 400 cm emission is absorbed by ionized thermal emission along the line - of - sight creating distinct differences in their morphology , see fig . [ fig1]a , b ) . therefore , the 400 cm images have been shifted by @xmath32 in r.a . and @xmath33 in dec . so that the background point sources detected at both 400 cm and 90 cm ( throughout the @xmath34 90 cm primary beam ) are coincident . we have also obtained new vla c - configuration 21 cm hi spectral line data toward the w51b region ( primary beam @xmath35 ) . the continuum from the line free - channels of these data have been combined with the continuum from line free - channels of archival vla d - configuration 21 cm hi data @xcite to form a 21 cm continuum image with a resolution matching that of the 90 cm b+c+d configuration data . it should be noted , however , that owing to spatial filtering the 21 cm data are missing flux on scales larger than @xmath36 compared to @xmath37 for the 90 cm data . in order to investigate the nature of the shocked molecular gas in the vicinity of the oh ( 1720 mhz ) masers we observed the co(32 ) line using the 345 ghz b3 receiver at the jcmt . the zenith opacity at 225 ghz measured by the tipping radiometer at the cso ranged from 0.10 - 0.12 . data were taken in raster mode with 4-s integrations per spectrum with samples separated by @xmath38 in right ascension and @xmath39 in declination . the final raster is @xmath40 in extent . typical system temperatures were between 400 - 500 k. the data were reduced using the specx software package . a linear baseline was subtracted from each spectrum using channels outside of the line emission region . in order to minimize baseline ripples we used a nearby off - position that contained low level co(32 ) emission . to account for this emission we also observed the reference position for 10 minutes in frequency switching mode , and added the resulting emission back to each spectrum . in this paper , the co(32 ) line intensity is presented in units of @xmath41 ( k ) , where @xmath42 . the final image cube was smoothed to @xmath43 resolution . a smaller region ( @xmath44 ) around the oh ( 1720 mhz ) masers was also mapped in @xmath17co(21 ) using receiver a3 when the zenith opacity was 0.13 . the samples were taken every @xmath39 in right ascension and every @xmath45 in declination with 4-s integrations . the system temperatures ranged from 250 - 300 k. the angular resolution of these data is @xmath43 . for @xmath17co(21 ) , we used @xmath46 . we observed the maser position in hcn ( 32 ) at 265.886 ghz and hco@xmath18 ( 32 ) at 267.558 ghz simultaneously using the caltech submillimeter observatory ( cso ) in good conditions ( @xmath47 ) . at these frequencies , the beam size is 26 and the main beam efficiency is 0.7 . using position - switched mode , we acquired 11 minutes of on source integration using the 230 ghz receiver and the facility ffts autocorrelators . the reference position was @xmath48 ( j2000 ) @xcite . the data were processed using the class software package . the final velocity resolution was smoothed to 0.60 km s@xmath4 for hco@xmath18 and 1.2 km s@xmath4 for hcn to improve the signal - to - noise . the results are presented in @xmath49 . figure [ mermom]a shows the merlin integrated intensity of the two oh ( 1720 mhz ) masers spots detected toward w51b . the locations of the maser spots with respect to the w51 region is shown in fig . the positions and central velocities are in good agreement with those reported by @xcite from vla observations ( see table [ lines ] ) . the results from fitting a gaussian profile to the maser spectra at the emission peaks are given in table [ lines ] , along with the results from earlier vla observations for reference @xcite . at the resolution of the merlin data ( @xmath50 mas , p.a.=@xmath51 ) the two maser spots remain unresolved . however , maps of the velocity field of the two masers ( see figs . [ mermom]b , and c ) are notably different , with the w51_1 spot showing a moire pattern indicative of no resolved velocity structure ( note the very small width of the displayed velocities ) , while the w51_2 spot shows a clear se - nw velocity gradient with higher velocities to the nw . the fitted merlin peak velocity and fwhm line width is @xmath52 km s@xmath4 and @xmath53 km s@xmath4 for w51_1 , while for w51_2 the fitted values are @xmath54 and @xmath55 km s@xmath4 ( also see table [ lines ] ) . for comparison , at assumed kinetic temperatures of 50 to 125 k , the thermal line width of oh ranges from 0.3 to 0.6 km s@xmath4 . thus , roughly 1/2 to 1/3 of the observed line width can be attributed to thermal motion , with the remaining line width is due to bulk or non - thermal motions . the integrated intensities and velocities ( moment 1 ) of this region as observed with the vlba with a resolution of @xmath56 mas ( @xmath57 ) are shown in figures [ vlbamom ] a , b , c , and d. these results are in excellent agreement with the merlin results w51_1 remains a single , though elongated maser feature with little velocity structure while w51_2 is composed of two distinct maser spots w51_2a , and w51_2b with velocities differing by only @xmath58 km s@xmath4 but with considerable velocity structure . the fitted line parameters from the vlba data are given in table [ lines ] . the fitted fwhm line width for w51_1 is in excellent agreement with that observed with merlin . the combined line width of the two components of w51_2 resolved by the vlba is consistent with the merlin line width of this maser , although the individual vlba line widths of w51_2a and w51_2b are slightly narrower . we also fitted 2-d gaussian components to the merlin and vlba integrated intensity images in order to measure the maser locations , angular sizes , and flux densities . the results are given in table 2 . the masers were unresolved by the vla and so the peak intensity @xmath59 is equivalent to the integrated flux density @xmath60 . interestingly , only about half of the vla a - configuration integrated flux density is recovered by the merlin observations , while the vlba and merlin recover a similar integrated flux density . if these masers are non - time variable , this suggests that merlin and the vlba resolved out about half the total flux density . this suggests that these masers have a `` core - halo '' morphology as seen previously for oh ( 1720 mhz ) , ch@xmath61oh , and h@xmath62o masers ( see for example * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) . from the vlba integrated intensity images , the devonvolved fitted sizes of the maser spots are w51_1 : @xmath63 mas ( p.a.@xmath64 ) , w51_2a : @xmath65 mas ( p.a.@xmath66 ) , and w51_2b : @xmath67 mas ( p.a.@xmath68 ) ; with statistical uncertainties of about @xmath69 . for reference , at 6 kpc , 10 mas@xmath70 cm or @xmath71 au using the peak intensities given in table [ lines ] these sizes correspond to peak brightness temperatures of @xmath72 , @xmath73 , and @xmath74 k for w51_1 , w51_2a , and w51_2b , respectively . from the merlin full stokes polarization data , the maser spot w51_1 has a polarized intensity ( @xmath75=sqrt(@xmath76 ) of @xmath77 and a position angle of @xmath78 , while the polarized intensity of w51_2 is @xmath79 at @xmath80 . the uncertainty in the position angles ( @xmath81 ) are dominated by the position angle calibration uncertainty which we estimate to be about @xmath34 . the merlin linear polarization position angles are shown in fig . @xcite reported a polarized intensity of @xmath82 and @xmath83 for w51_2 , but the polarization position angle calibration available for those data were very uncertain . following @xcite , we fit the zeeman magnetic field strength using the thermal zeeman equation @xmath84 @xmath85@xmath86 where @xmath87 is stokes @xmath88 , @xmath89=0.6536 hz @xmath19g@xmath4 , and @xmath86 is the derivative of stokes @xmath90 . the parameter @xmath85@xmath91@xmath92 , is the magnitude of the magnetic field strength times a constant @xmath93 that may depend on the angle @xmath94 between the magnetic field @xmath95 and the line - of - sight . for example , for thermal lines @xmath96 ; the meaning of @xmath93 for the oh ( 1720 mhz ) maser case will be discussed further in 4.3 . figures [ b_mer ] and [ b_vlba ] show the zeeman fits for the merlin and vlba data , respectively . the fitted values of @xmath85 are also listed in table 2 . the signal - to - noise of these fits are outstanding with s / n up to 30 ( merlin for w51_1 ) and s / n=7 ( vlba for w51_1 ) . the observed values of @xmath85 from 1.5 to 2.2 mg are in good agreement with the vla observations of @xcite . for the maser with the simplest velocity structure , w51_1 , the merlin and vlba magnetic field results are in excellent agreement . for w51_2 , which the vlba data reveal is composed of two spots , the merlin value is essentially an intensity weighted average of the two individual vlba results . these data are of sufficient quality that by following @xcite we also independently derived @xmath85 by fitting gaussian components to the left and right circularly polarized profiles independently and determining the line splitting directly , i.e. @xmath85=@xmath97 . for example , for w51_2a the line splitting is @xmath98 km s@xmath4 . this method yielded consistent results with the thermal zeeman equation values listed in table 2 . the 90 cm ( 330 mhz ) images presented in this work ( for example fig [ fig1]a ) are qualitatively similar to the @xmath99 resolution 90 cm image presented by @xcite , but the resolution and sensitivity are significantly improved . figure [ fig1]a shows the vla b+c+d configuration 90 cm image of the w51 region ( resolution @xmath100 , see table 1 ) with the three major components w51a ( hii region complex ) , w51b ( hii region complex ) , and w51c ( snr ) identified . the oh ( 1720 mhz ) masers lie about @xmath101 west of the w51b hii region g49.2@xmath1020.3 , and are coincident with an unresolved 90 cm source with an angular size of @xmath103 at a position of ( j2000 ) @xmath104 ( see fig . [ fig1]a ) . the w51 region is shown for the first time at a wavelength longer than 2 m in figure [ fig1]b . the vla 400 cm b+c configuration image with a resolution of @xmath105 ( see table 1 ) reveals diffuse , extended non - thermal emission concentrated mostly to the eastern side of the w51c snr , with some diffuse emission also appearing toward the northwestern boundary of the w51c snr . most interesting for the purpose of the current study is the non - thermal 400 cm emission that appears to partially encircle the g49.2@xmath1020.2 hii region and the discovery of an unresolved 400 cm source coincident with 90 cm emission and the oh ( 1720 mhz ) masers . hereafter , we will call this source of non - thermal emission w51b_nt . it is notable that the area toward the southern portion of the w51b string of hii regions is invisible at 400 cm , consistent with these regions being in the foreground of the w51c snr , and free - free absorbing the 400 cm emission from w51c along this line - of - sight ( see for example * ? ? ? * ; * ? ? ? to aid comparison of the thermal vs non - thermal gas , figures [ big]a , b , shows three - color images constructed from the spitzer glimpse 4.5 and 8 @xmath19 m bands and the vla 90 cm and 400 cm data , respectively . the mid - infrared emission traces thermal ionized gas , as well as emission from dust and pahs @xcite . its morphology matches closely that of the molecular cloud associated with w51b ( see for example * ? ? ? * ; * ? ? ? * ) . the 90 cm emission ( fig . [ big]a ) traces both bright synchrotron ( snr ) emission _ and _ optically thick free - free emission from hii regions . in contrast , the 400 cm emission ( fig . [ big]b ) traces _ only _ synchrotron emission except where it is absorbed by foreground free - free emission . [ big]c and d show the large scale _ rosat _ soft and _ asca _ hard x - ray emission , respectively , first presented by @xcite . soft x - rays trace hot ( @xmath106 kev ) thermal gas from the snr , except where it is absorbed by the intervening total column of atomic and molecular gas . in contrast , the hard x - rays are mostly unaffected by the column of gas along the line of sight and originates from even hotter thermal gas , including a pulsar wind nebula candidate and young stars with fast winds . we detect diffuse 90 cm emission at the location of the pulsar wind nebula candidate cxo j192318.5 + 1403035 from ( * ? ? ? * the location is indicated on figs . [ big]a - d ) , but it is not clear if it arises from a distinct source compared to the overall diffuse snr emission . similar to the 400 cm emission , there is a lack of strong soft x - ray emission toward the southern portions of the w51b hii region complex . in fig . [ big]d we also indicate the region in which strong tev emission was detected by the magic telescope @xcite . a zoomed in view of the 21 cm continuum emission in the vicinity of g42.9@xmath1020.3 and w51b_nt is shown in figure [ zoom]a at @xmath100 resolution . for comparison , the 400 cm emission and archival soft and hard x - ray emission from _ chandra _ ( see * ? ? ? * ) of the same field are shown in figures [ zoom]b , c , and d. these images show the coincidence of 21 , 90 , and 400 cm emission toward w51b_nt , as well as the presence of a faint hard x - ray source at this position . unfortunately , this emission is too weak to extract a useful spectrum . the nature of the stronger hard x - ray source to the nw of w51b_nt is unknown . after accounting for the background contribution we estimate integrated flux densities for w51b_nt ( defined by the 90 cm 0.15 jy beam@xmath4 contour shown in fig . [ big]a , though it is partially obscured by the maser symbol ) of @xmath6 0.5 , 1.3 , and 0.6 jy at 400 cm , 90 cm , and 21 cm , respectively . the uncertainty of these flux densities are dominated by the background estimate , and are not accurate to better than @xmath107 jy . the low 400 cm flux density is almost certainly a result of free - free absorption along this complex line - of - sight ( see for example * ? ? ? the 21 and 90 cm flux densities imply a non - thermal radio spectral index of @xmath108 ( @xmath109 ) . other groups have searched for possible non - thermal sources in the w51 complex . in particular , @xcite used the bonn 11 cm galactic plane survey ( resolution @xmath110 ) along with iras 60 @xmath19 m images to disentangle the thermal and non - thermal components of w51 . though uncertain with respect to absolute flux densities , their resulting images show the well - known thermal sources in w51a and w51b as well as the non - thermal w51c snr . interestingly , they also see evidence for a non - thermal source west of the g49.2 - 0.3 hii region . more qualitative evidence for the existence of non - thermal emission in the vicinity of w51b_nt was presented by @xcite at 151 mhz with a resolution of @xmath111 with a double source detected near g49.2 - 0.3 with a separation of a few arcminutes ( note that the published declination scale of the 151 mhz image must be displaced to the south by 5 arc min * ? ? ? though suggestive , these early data have too poor an angular resolution to be useful in the current analysis . the jcmt co(32 ) integrated intensity from 40 to 95 km s@xmath4 with a resolution of @xmath112 is shown in figure [ co]a . representative spectra are shown in figs . [ co]b - g ) , and where observed , the @xmath17co(21 ) spectra are also shown ( figs . [ co]d - f ) . from these data we find that co emission partially encircles w51b_nt with the strongest emission along the eastern and southern boundaries . this enhanced co(32 ) emission is coincident with the locations of the oh ( 1720 mhz ) masers which are located towards the se corner of the w51b_nt source ( fig . [ co]e , profile `` nt_oh '' is taken from the maser location ) . the spectra in the w51b_nt region ( figs . [ co]d , e , f ) show at least three components : ( i ) a narrow ( @xmath113 km s@xmath4 ) , but strong feature @xmath114 k at @xmath115 km s@xmath4 ; ( ii ) a broad ( @xmath116 km s@xmath4 wide ) feature with a strength of @xmath117 k also centered around 70 km s@xmath4 ; and ( iii ) a narrow ( @xmath118 km s@xmath4 ) , weak blue - shifted feature at 60 km s@xmath4 . the appearance of the dramatically broadened component ( ii ) in the vicinity of w51b_nt is strongly indicative of a shock . the co throughout the mapped region has an average peak velocity of about 70 km s@xmath4 ( @xmath119 km s@xmath4 ) , in agreement with components ( i ) and ( ii ) toward the masers . thus , while the broad velocity width observed toward the w51b_nt region is exceptional , it does not appear that the w51b_nt region arises from a distinct cloud along the line of site , but is co - distant with the nearby hii regions like g49.2@xmath1020.3 . additionally , the velocities of the narrow and broad co components at 70 and 71 km s@xmath4 , respectively , are in excellent agreement with the range of oh ( 1720 mhz ) maser velocities ( 69 - 72 km s@xmath4 ; see table [ lines ] ) . other features of note in the full co(32 ) raster ( see fig . [ co]b - g ) : ( i ) the emission associated with the g49.2@xmath1020.3 hii region is comparatively weak and is composed of two main components that bracket the radio recombination line velocity from @xcite of 67.2 km s@xmath4 . ( ii ) along the sw rim of the hii region is a compact `` clump '' with the strongest co(32 ) emission in the mapped region ( 40 k ) , but with narrow ( 4 km s@xmath4 ) line widths . ( iii ) the southernmost co(32 ) clump labelled `` yso '' shows evidence of bipolar outflow emission predominantly in an east - west direction , but with a third component to the north . the `` yso '' source is also associated with weak 90 cm and 20 cm emission , though its appearance is unremarkable in the mid - ir . from these clues , it seems likely this source is an intermediate to massive protostar with the radio continuum arising from optically thick free - free emission . the appearance of the co(32 ) and @xmath17co(21 ) spectra toward w51b_nt , with a broad ( @xmath116 km s@xmath4 ) component superposed on a narrow ( @xmath113 km s@xmath4 ) feature at the lsr velocity is in good agreement with other co(32 ) studies of snr oh ( 1720 mhz ) maser regions ( see for example * ? ? ? * ; * ? ? ? * ) . to facilitate comparison of all four observed thermal molecular line species , we convolved the jcmt co(32 ) and @xmath17co(21 ) cubes to @xmath120 resolution to match the cso hcn(32 ) and hco@xmath18(32 ) data . spectra of all four species at this resolution toward the centroid of the oh maser location are shown in figure [ densitytemperature]a , and the parameters from gaussian fits using the casa task specfit are given in table [ cofits ] . using the @xmath120 resolution fitted parameters at the oh maser position ( table [ cofits ] ) , we ran the radex radiative transfer code @xcite with the lamda molecular data files @xcite to simultaneously model the co(32 ) and @xmath17co(21 ) line strengths and thereby estimate the physical parameters of the pre - shock and post - shock gas . because these two lines trace somewhat different excitation conditions , combining observations of both lines helps to constrain the possible densities and temperatures present . we chose the slab option in radex as it is the most appropriate geometry for both a shock and a molecular cloud filament . we ran grids of radex models over density and temperature for a range of plausible column densities , while assuming n(co)/n(h@xmath62)@xmath121 and an isotopic abundance of @xmath122c/@xmath17c=60 @xcite . the h@xmath62 column densities and corresponding densities and kinetic temperatures that simultaneously produce co(32 ) and @xmath17co(21 ) antenna temperatures that match the observed spectra are plotted in fig.[densitytemperature]b . next we used radex to determine what abundance of hcn(32 ) and hco@xmath18(32 ) would be required to produce the observed intensities of these high density tracers at the densities and kinetic temperatures derived from the co observations . figure [ densitytemperature]b summarizes the results for the four species modeled . in the lower left corner of fig . [ densitytemperature]b , the sequence of points of varying h@xmath62 column density ( from @xmath123 to @xmath124 @xmath125 ) indicate the conditions that reproduce the observed line profiles in the pre - shock gas , with kinetic temperatures ranging from 26 to 32 k and h@xmath62 densities of 850 to 4000 @xmath10 . the corresponding hco@xmath18 and hcn line strengths are consistent with abundances of @xmath126 , which are typical of the values seen in massive star formation regions ( * ? ? ? * and references therein ) . we can compare the pre - shock column densities derived from the radex analysis with those estimated from previous millimeter continuum and x - ray observations toward this region . w51b_nt is not detected at 1.1 mm in the @xmath127 resolution bolocam galactic plane survey ( the hii region g49.2 - 0.3 is detected ; see * ? ? ? * for survey details ) . the @xmath128 upper limit measured in the vicinity of w51b_nt is about 0.36 mjy beam@xmath4 after multiplying by the 1.5 flux density correction factor recommended by @xcite . assuming the dust temperature is about @xmath129 k in the pre - shock gas , a dust to gas ratio of 100 , and a dust opacity of 0.0114 @xmath130 g@xmath4 @xcite , we find a dust - derived @xmath131 column density upper limit of @xmath132 @xmath125 . we can also use the results of previous x - ray observations toward this region to derive a lower limit . using the _ suzaku _ x - ray satellite , @xcite find an average total proton column density @xmath133 toward their region 3 , which encompasses most of w51b including w51b_nt , of @xmath134 @xmath125 . using a combination of arecibo and vla hi emission and absorption data , @xcite find @xmath135 @xmath125 along the line - of - sight to the nearby hii region g49.2@xmath1020.3 , with @xmath136 of the total column originating close to the w51b region . since @xmath137 k is an upper limit , @xmath138 represents an upper limit to the true n(hi ) . when compared to the x - ray derived @xmath139 , this suggests @xmath140 @xmath125 . thus , the allowed range estimated from previous observations is @xmath141 @xmath125 , compared to @xmath142 @xmath125 from the radax molecular line analysis . the post - shock gas is represented in the upper portion of figure [ densitytemperature]b , with the gray shaded rectangular region indicating the theoretical range of parameters sufficient to pump the 1720 mhz oh maser . a beam filling factor of significantly less than 1 is required to explain the low observed line strengths in the broad co component . this result is not surprising given that the @xmath143 beam corresponds to a linear diameter of @xmath144 cm , because the width of the region of elevated temperature in a typical interstellar c - type shock is @xmath145 cm , depending on the ionization rate @xcite . the hco@xmath18 abundance is significantly lower ( factors of 3 - 20 ) throughout the shaded region than its value in the pre - shock gas , suggesting a difference in chemistry . indeed , pdr models indicate that the hco@xmath18/co abundance ratio drops precipitously when the ionization rate exceeds 10@xmath146 s@xmath4 @xcite . as mentioned previously , the w51b / w51c interface region is a strong emitter of @xmath2-rays with properties that are consistent with a hadronic emission mechanism ( see * ? ? ? * and references therein ) . @xcite estimate @xmath2-ray and x - ray induced ionization rates of @xmath147 s@xmath4 and @xmath148 s@xmath4 , respectively , for this region . though dependent on many assumptions , together , the @xmath2-ray plus x - ray induced ionization rate is about 30 times the local interstellar rate , and is in excellent accord with both that needed to produce an adequate abundance of oh for the masers @xcite , and that required to lower the hco@xmath18/co abundance as predicted by @xcite . similar to hco@xmath18 , the inferred abundance of hcn is also reduced in much of the post - shock parameter space , with the exception of the low - temperature , low - density end of the range ( see fig . [ densitytemperature]b ) . while this suggests that the physical conditions in the post - shock gas may lie in this triangular region , it may be that the abundance of hcn is also altered by the shock chemistry . in any case , a range of physical conditions almost certainly exists across the shock , such that there is no single unique answer . still these observations demonstrate that the range of expected densities and temperatures from the maser models are supported by the thermal molecular line data . first , to summarize the key multi - wavelength results : a non - thermal source of low frequency radio emission ( w51b_nt ) has been detected in the vicinity of the w51b oh ( 1720 mhz ) masers . this source is partially encircled by shocked molecular gas that is coincident in space and velocity with the masers . moreover , detailed modeling of the molecular emission indicates that the physical conditions in the pre- and post - shock gas are in good agreement with that expected from collisional pump models . it is also notable that the molecular gas associated with w51b_nt shares a similar lsr velocity with nearby hii regions ( for example g49.2@xmath1020.3 ) , suggesting that the masers are co - distant with the northern members of the w51b hii region complex . the non - thermal radio source is also coincident with a hard x - ray source and falls within the region of high likelihood for the position of tev @xmath2-ray emission , both of which have been suggested as necessary to produce enough oh abundance ( from water ) in c - type shocks for snr oh ( 1720 mhz ) maser emission . thus , the evidence indicates that these oh ( 1720 mhz ) masers are of the snr / molecular cloud interaction variety . the multi - wavelength data presented in this work suggests two possible scenarios for the nature of w51b_nt : ( 1 ) the non - thermal radio continuum emission , co shock , oh ( 1720 mhz ) masers , and hard x - ray emission are due to the w51c snr , which lies close enough behind w51b for its shock wave to have hit the hii region complex ; or ( 2 ) along this crowded , tangent point line of sight there is a previously unknown small diameter ( @xmath149 ) supernova remnant co - distant with the w51b hii regions . as will be described below , option ( 1 ) seems the most consistent with the full range of data available for this region . it is clear from both 4 m and soft x - ray absorption that the w51b hii region complex lies at least partially _ in front _ of the w51c snr . the key uncertainty is whether w51b is close enough to w51c for the snr s shock front to have reached its back side . in addition to the presence of w51b_nt and the oh ( 1720 mhz ) masers , there is further evidence that the w51c snr is interacting with much of the northern end of w51b . for example , @xcite report a filamentary arc of high velocity ( hv ) , shocked hi emission at velocities between @xmath150 km s@xmath4 toward the northern section of w51b . this arc of emission begins just west of w51b_nt , curves around to the south , and ends just north of the hii region g49.1@xmath1020.4 . figure [ hvhi]a shows the morphology of the hv hi compared to the 90 cm continuum ; the co(32 ) emission from the smaller field of view imaged with the jcmt is also shown . the arc of the filament matches very closely with the morphology of the 90 cm continuum emission in this region . @xcite also find hv co(10 ) emission roughly co - spatial with the hv hi ( most of the hv co(10 ) is outside the field of view of the co(32 ) jcmt data , see fig . [ hvhi]a ) . these authors suggest that the hv gas is tracing a shock interaction between the w51c snr and the molecular gas associated with the ( foreground ) w51b complex of hii regions . indeed , the hv hi does appear to lie along the transition between the unabsorbed 400 cm snr emission and the large scale filamentary molecular cloud associated with the w51b complex the co filament is roughly coincident with the _ spitzer _ 8 @xmath19 m emission shown in fig . [ big]a , b ( see for example * ? ? ? * ; * ? ? ? @xcite finds that the properties of the hv hi shock are consistent with a dissociative j - type shock , and further postulate that the post - shock gas is mostly atomic since the hv hi column density is quite high . therefore , the non - detection of hv hi toward w51b_nt is not surprising since it is spatially coincident with the oh ( 1720 mhz ) masers that are thought to arise in non - dissociative c - type shocks @xcite . another piece of evidence indicating a more extensive interaction is the morphology of the 400 cm emission . in particular , the non - thermal 400 cm emission appears to both encircle and partially overlap with the g49.2@xmath1020.3 hii region emission as defined by the shorter wavelength 21 cm image ( see figs.[zoom]a , b ) . as described , for example , in @xcite and @xcite , relatively little thermal ionized gas is required to absorb 400 cm emission so it would be surprising to detect such low frequency emission in the direction of a foreground hii region . this expectation can be quantified by comparing an estimate of the 400 cm free - free continuum opacity from the hii region with the observed 400 cm emission . the 400 cm opacity was calculated by first convolving the c+d 21 cm continuum image to the 400 cm image resolution of @xmath151 , and then converting this image to brightness temperature . next the @xmath152 was calculated from @xmath153 assuming an electron temperature of 10,000 k , and then extrapolated to @xmath154 using @xmath155 . using this method , we find a peak 400 cm opacity of 22.5 with @xmath156 extending well into the observed 400 cm emission , see figure [ hvhi]b . of course , in reality the @xmath157 within the hii region is unlikely to be constant , but the relatively high assumed value yields a lower limit to @xmath152 . the strongest observed 400 cm emission is on the eastern side of w51c at 1.35 jy beam@xmath4 ( see fig . [ fig1]b ) . since it seems unlikely that the unabsorbed 400 cm emission underlying the hii region is significantly brighter than this intensity , we take it to be an upper limit to the unabsorbed 400 cm non - thermal emission toward g49.2@xmath1020.3 . when this assumption is coupled with the @xmath158 rms noise of the 400 cm image of 0.3 jy beam@xmath4 , we find that lines of sight with @xmath159 would be undetectable _ if the hii region lies entirely in front of the 400 cm non - thermal emission_. however , as demonstrated in fig . [ hvhi]b detectable 400 cm emission is observed toward regions with @xmath154 up to @xmath160 ! this analysis suggests , as does the morphology , that in fact the w51c snr has partially enveloped the g49.2@xmath1020.3 hii region . interestingly , from analysis of the optical extinction toward the w51b region , @xcite found that the northern half of w51b is more heavily extincted than the south / southwestern end . since the total molecular column density is actually larger to the sw , these authors ( also see * ? ? ? * ) suggest that the northern end of w51b is tilted deeper into the cloud than the south / southwestern end , placing it closer to w51c . from all of these comparisons it seems clear that the w51c snr is interacting with the northern part of the w51b hii region complex . as a shock propagates , the gas is compressed in the direction the shock is moving and the component of the magnetic field parallel to the shock front will be amplified . for collisionally pumped maser emission it is further expected that the shock must be moving more or less perpendicular to our line - of - sight in order to assure adequate velocity coherence , and column density along the line - of - sight . thus , collisionally pumped masers like oh ( 1720 mhz ) masers associated with snrs and class i ch@xmath61oh masers are always found near the systemic velocity . these two facts together imply that the magnetic field vector is oriented somewhere in the plane perpendicular to the direction the shock front is moving , and we are viewing that shock front edge - on . using this as the starting geometry , one would like to use the polarization measurements to constrain the direction of the magnetic field vector and the value of @xmath93 in the @xmath85@xmath91@xmath92equation . maser propagation theory ( see e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) suggests that the observed percentage of linear polarization @xmath75 is a strong function of the angle between the line - of - sight and the magnetic field vector @xmath161 . indeed , for @xmath162 the observed @xmath75 should be zero ( it is also zero for @xmath163 ) . away from the critical angles of @xmath164 and @xmath165 , @xmath75 is non - zero , though its magnitude is a strong function not only of @xmath161 , but also the saturation level of the maser and to some extent to the angular momentum of the molecular state that is masing . the brightness temperatures of oh ( 1720 mhz ) masers @xmath166 k , along with the lack of significant variability are consistent with at least a moderate degree of saturation ( also see additional arguments in * ? ? ? * ; * ? ? ? * ; * ? ? ? assuming moderate saturation , according to the model of @xcite , for a j=21 state , @xmath75 is expected to be within about @xmath167 to @xmath168 , with positive values corresponding to @xmath169 and negative values corresponding to @xmath170 . if @xmath161 is known , this model can also be used to constrain the @xmath93 parameter ( which may depend on @xmath94 ) in the stokes v conversion to magnetic field strength . for small values of @xmath75 and moderate saturation , @xmath171 ( z = the zeeman coefficient , and @xmath172 is the total field strength ) with no @xmath94 dependence , i.e. @xmath173 and hence @xmath174 . currently , no detailed analysis is available for the j=3/2 oh ( 1720 mhz ) transition , though it would likely be qualitatively similar to the j=21 case ( private communication , william watson 2005 ) . this suggests that given the small measured @xmath75 values of @xmath175 and @xmath176 ( i.e. stokes u is consistent with zero intensity and stokes q is positive ) for w51_1 and w51_2 , @xmath177 . unfortunately , unless a maser polarization model is created for the j=3/2 oh ( 1720 mhz ) transition a more quantitative analysis is not be possible . a maser s linear polarization position angle @xmath81 can be _ either _ parallel or perpendicular to the magnetic field in the plane - of - the - sky @xmath178 . however , maser theories also predict that for @xmath169 , @xmath179 ( see e.g. * ? ? ? * ; * ? ? ? as mentioned above , we expect that @xmath92 lies parallel to the plane of the shock front ( perpendicular to its direction of motion ) , and in principle the orientation of the shock front itself can be discerned from the morphology of the shocked co(32 ) emission . although there is significant confusion from quiescent gas at the lsr velocity of the w51b molecular cloud ( @xmath180 km s@xmath4 ) , the co(32 ) emission from the wings of the broad component ( see fig . [ co]e , f ) appears to be oriented east - west . this is also apparent from the co(32 ) integrated intensity image ( fig . [ co]a ) which is dominated by the broad component . an east - west orientation for the molecular shock front is approximately perpendicular to that found for the linear polarization position angle @xmath81 ( see 3.2.1 ; fig . [ mermom]a ) . this result is in good agreement with the expectation described above based on @xmath75 that @xmath169 and hence we have @xmath179 . the situation where @xmath81 is perpendicular to the shock front is also seen in w28 but the opposite orientation is observed for w44 @xcite . the fact that some sources fall into the @xmath181 case and some the @xmath182 breaks the uncomfortable coincidence pointed out by @xcite that since all snr oh ( 1720 mhz ) masers have linear polarization percentages of @xmath183 they could all have essentially the same @xmath184 which is extremely unlikely . a final consideration is whether the observed @xmath75 represents the true polarized intensity . the observed polarized intensity could be a lower limit to the true intensity , owing to the factors mentioned by @xcite , including including faraday depolarization and tangling of the magnetic field lines within the masing region . those authors found that the faraday depolarization length is only about three times larger than the estimated maser gain length of about @xmath185 cm @xcite . the similarity of the gain and faraday depolarization lengths suggests that faraday depolarization is possible , but the new data do not allow further insight on this issue . luckily even in the presence of faraday depolarization the net polarization p.a . is not expected to change @xcite . we can now exclude the second possibility down to sizescales of @xmath70 cm ( i.e. 10 mas at 6 kpc ) based on the fact that across 4 orders of magnitude in beam area from vla to vlba resolution the observed zeeman line splitting did not increase ( except as consistent with spectral blending ) as would be expected if the magnetic field were tangled on small size scales . we have presented a wide range of new data toward the w51c snr and the w51b hii region complex . using merlin and the vlba we have spatially resolved the oh ( 1720 mhz ) masers in this region and explored their magnetic field properties , establishing that ( 1 ) the magnetic field strengths range from 1.5 to 2.2 mg ( using the thermal zeeman equation ) ; ( 2 ) the field strengths do not increase with angular resolution ( except where features were spatially unresolved and spectrally blended by previous , lower - resolution observations ) suggesting the field is relatively smooth on these scales ; ( 3 ) the maser spots sizes are relatively large compared to hii region masers , with linear dimensions of @xmath186 cm or @xmath187 au at 6 kpc and brightness temperatures of @xmath188 k ; ( 4 ) the linear polarized intensities are a few percent and the position angle of the linear polarization is nearly zero ; and ( 5 ) the linear polarization properties suggest that the angle between the magnetic field vector and the line - of - sight is @xmath189 , and that the difference between the magnetic field strengths measured using the thermal equation and the total field strength is likely to be small . more quantitative analysis requires a detailed polarization model for the j=3/2 case . we have presented the most sensitive , highest angular resolution long wavelength images of this region to date at 90 cm and 400 cm observed with the vla showing the non - thermal radio continuum emission from the w51c snr with unprecedented detail . these data reveal ( 1 ) the presence of non - thermal radio continuum emission in the vicinity of the oh ( 1720 mhz ) masers , which we denote w51b_nt ; ( 2 ) that the w51b hii region complex must lie in front of the w51c snr in agreement with previous soft x - ray observations ; and ( 3 ) that the nearby hii region g49.2@xmath1020.3 ( @xmath190 east of the masers ) has been at least partially enveloped by the w51c snr by comparing the expected versus observed 400 cm absorption . through comparison with previous x - ray data , we also find that a source of hard x - rays is coincident with w51b_nt , though the signal is too faint to model the spectrum . using observations of co(32 ) , @xmath17co(21 ) , hco@xmath18(32 ) , and hcn(32 ) in the vicinity around the oh ( 1720 mhz ) masers , we have discovered a ring of shocked gas partially encircling the non - thermal emission of w51b_nt , coincident spatially and kinematically with the oh ( 1720 mhz ) masers . radiative transfer modeling of the physical conditions in the narrow velocity , unshocked gas yields a column density of @xmath191 @xmath125 , a density of @xmath192 @xmath10 , and a temperature of 26 - 32 k. the broad velocity , shocked gas is significantly smaller than the beam leading to a range of possible column densities ( @xmath193 @xmath125 ) and hence physical conditions , but the most likely density and temperatures are @xmath194 @xmath10 and 50 - 100 k , consistent with the passage of a c - shock . clb thanks the jcmt and the university of hawaii for fellowship support during the course of this research . imh thanks nrao for pre - doctoral fellowship support during the course of this project . we would also like to thank professors william ( bill ) watson and moshe eliztur for helpful discussions about maser polarization properties . we are also grateful for the help provided by darek lis and simon radford in obtaining the cso spectra . this research has made use of nasa s astrophysics data system . we acknowledge the use of nasa s _ skyview _ facility ( http://skyview.gsfc.nasa.gov ) located at nasa goddard space flight center . ll date & 2002 jan 14 , 15 , 18 ( mmo1b06 ) + bandwidth & 0.25 mhz + spectral channels & 256 + channel separation & 0.17 km s@xmath4 + velocity resolution & 0.24 km s@xmath4 + spectral line rms noise@xmath195 & 5 mjy beam@xmath4 + synthesized beam & 221 mas @xmath196 125 mas p.a.@xmath197 + dates & 2000 dec 02,03 , & 04 ( bb129 ) + bandwidth & 0.25 mhz + spectral channels & 256 + channel separation & 0.17 km s@xmath4 + velocity resolution & 0.24 km s@xmath4 + spectral line rms noise@xmath195 & 9 mjy beam@xmath4 + synthesized beam & 12.5 mas@xmath198 mas p.a.@xmath199 + dates b - array & 2002 jun 06 & 22 ( ab1031 ) + date c - array & 2006 oct 24 ( ab1219 ) + bandwidth & 1.5 mhz + spectral channels & 64 + continuum rms noise & 100 mjy beam@xmath4 + synthesized beam ( b+c ) & @xmath200 p.a . @xmath201 + date a - array & 2003 aug 23 ( ab1089 ) + dates b - array & 2002 jun 06 & 22(ab1031 ) + dates c - array & 2002 nov 02 ; 2002 dec 13 & 30(ab1031 ) + date d - array & 2003 feb 08 ( ab1077 ) + bandwidth & 3.0 mhz + spectral channels & 32 + continuum rms noise & 12 mjy beam@xmath4 + synthesized beam ( b+c+d ) & @xmath202 p.a.@xmath203 + synthesized beam ( a+b+c+d ) & @xmath204 p.a.@xmath205 + date c - array & 2006 nov 13 ( ab1219 ) + date d - array ( used for continuum only ) & 1992 jul ( ak301 ) + bandwidth ( c ) & 3.0 mhz + spectral channels ( c ) & 32 + continuum rms noise ( c+d ) & 5 mjy beam@xmath4 + spectral line rms noise@xmath195 ( c ) & 3 mjy beam@xmath4 + synthesized beam ( c+d continuum ) & @xmath202 p.a.@xmath203 + synthesized beam ( c hi line ) & @xmath206 p.a.@xmath207 + dates & 2004 mar 29 ; apr 20 - 25 ; jun 11 & 12 ( m04ah45a2 ) + rest frequency & 345.79599 ghz + observing mode & single sideband + velocity resolution & 1.08 km s@xmath4 + primary beam & @xmath208 + spectral line rms noise@xmath195 & 0.1 k + dates & 2005 sep 22 ( m05bh49c ) + rest frequency & 220.39868 ghz + observing mode & double sideband + velocity resolution & 1.1 km s@xmath4 + primary beam & @xmath43 + spectral line rms noise@xmath195 & 0.15 k + dates & 2012 nov 08 + rest frequencies & 265.886 & 267.558 ghz + observing mode & double sideband + velocity resolution & 0.6 & 1.2 km s@xmath4 + primary beam & @xmath143 + spectral line rms noise@xmath195 & 0.02 & 0.014 k + [ observing ] lcccccccc w51_1 & vla & & & 72.0 ( 0.02 ) & 0.9 ( 0.05 ) & 2.6 ( 0.05 ) & 2.6 ( 0.05 ) & @xmath209 + & merlin & & & 72.018 ( 0.002 ) & 0.902 ( 0.004 ) & 1.234 ( 0.004 ) & 1.38 ( 0.02 ) & @xmath209 + & vlba & 19 22 53.8210 & + 14 15 43.462 & 72.06 ( 0.01 ) & 0.901 ( 0.03 ) & 0.211 ( 0.007 ) & 1.05 ( 0.05 ) & @xmath210 + w51_2 & vla & & & 69.1 ( 0.02 ) & 1.2 ( 0.05 ) & 6.1 ( 0.05 ) & 6.1 ( 0.05 ) & @xmath211 + & merlin & & & 69.031 ( 0.001 ) & 1.183 ( 0.002 ) & 2.939 ( 0.004 ) & 3.12 ( 0.02 ) & @xmath212 + & vlba_a & 19 22 54.3632 & + 14 15 40.219 & 68.965 ( 0.002 ) & 1.064 ( 0.004 ) & 1.029 ( 0.004 ) & 2.07 ( 0.02 ) & @xmath213 + & vlba_b & 19 22 54.3608 & + 14 15 40.247 & 69.050 ( 0.004 ) & 0.957 ( 0.009 ) & 0.700 ( 0.005 ) & 1.20 ( 0.04 ) & @xmath214 + [ lines ] lcccc @xmath122co 32 & narrow & 70.65 @xmath215 0.03 & 18.57 @xmath215 0.28 & 4.99 @xmath215 0.09 + & broad & 70.12 @xmath215 0.18 & 6.72 @xmath215 0.26 & 18.64 @xmath215 0.52 + & blueshifted & 59.40 @xmath215 0.15 & 2.30 @xmath215 0.26 & 3.00 @xmath215 0.43 + @xmath17co 21 & narrow & 71.05 @xmath215 0.07 & 4.31 @xmath215 0.22 & 3.20 @xmath215 0.20 + & broad & 70.1 & 0.49 @xmath215 0.16 & 15.47 @xmath215 4.43 + & blueshifted & 59.4 & 0.43 @xmath215 0.18 & 3.0 + hco+ 32 & narrow & 71.03 @xmath215 0.16 & 0.20 @xmath215 0.02 & 4.4 @xmath215 0.5 + & broad & 69.6 @xmath215 0.3 & 0.21 @xmath215 0.02 & 18.0 @xmath215 1.0 + & blueshifted & 59.3 @xmath215 0.3 & 0.07 @xmath215 0.01 & 3.0 + hcn 32 & narrow & 69.6 @xmath215 0.2 & 0.07 @xmath215 0.01 & 1.7 @xmath215 0.4 + & broad & 69.7 @xmath215 0.3 & 0.16 @xmath215 0.02 & 14.9 @xmath215 1.0 + & blueshifted & 57.4 @xmath215 0.4 & 0.06 @xmath215 0.01 & 3.0

the distribution of bar strengths in disk galaxies is a fundamental property of the galaxy population that has only begun to be explored . we have applied the bar / spiral separation method of buta , block , and knapen to derive the distribution of maximum relative gravitational bar torques , , for 147 spiral galaxies in the statistically well - defined ohio state university bright galaxy survey ( osubgs ) sample . our goal is to examine the properties of bars as independently as possible of their associated spirals . we find that the distribution of bar strength declines smoothly with increasing , with more than 40% of the sample having 0.1 . in the context of recurrent bar formation , this suggests that strongly - barred states are relatively short - lived compared to weakly - barred or non - barred states . we do not find compelling evidence for a bimodal distribution of bar strengths . instead , the distribution is fairly smooth in the range 0.0 0.8 . our analysis also provides a first look at spiral strengths in the osu sample , based on the same torque indicator . we are able to verify a possible weak correlation between and , in the sense that galaxies with the strongest bars tend also to have strong spirals .

astro-ph0505079

bars and spirals are an important part of the morphology of disk galaxies . these showy disk morphological features which characterize the ( hubble ) tuning fork " ( firmani & avila - reese 2003 ) play a role in general classification schemes ( e.g. , hubble 1926 ; sandage 1961 ; de vaucouleurs 1959 ; sandage & bedke 1994 ) , and also can be tied to disk galaxy evolution ( e.g. , kormendy & kennicutt 2004 ) . over the past two decades , there has been a great deal of interest in the properties of bars , including quantification of bar strength ( e.g. , elmegreen & elmegreen 1985 ; martin 1995 ; wozniak et al . 1995 ; martinet & friedli 1997 ; rozas et al . 1998 ; aguerri et al . 1998 ; seigar & james 1998 ; aguerri 1999 ; chapelon et al . 1999 ; abraham & merrifield 2000 ; shlosman et al . 2000 ; buta & block 2001 ; laurikainen & salo 2002 ; knapen et al . 2002 ) , bar pattern speeds ( elmegreen et al . 1996 ; corsini et al . 2003 , 2004 ; debattista and williams 2004 ; aguerri et al . 2003 ; debattista et al . 2002 ; gerssen et al . 1999 ; merrifield and kuijken 1995 ) , mass inflow rates ( quillen et al . 1995 ) , and studies of the distribution of bar strengths ( block et al . 2002 ; whyte et al . 2002 ; buta , laurikainen , & salo 2004 ) . the most recent studies have indicated , on one hand , that bar and spiral strength can be quantified in a reasonable manner from near - infrared images and , on another hand , that such quantifications are useful for probing both bar and spiral evolution . the distribution of bar strengths is a particularly important issue . it is well known that as much as 70% of normal bright galaxies are barred at some level ( e.g. , eskridge et al . 2002 ) , which suggests that bars might be long - lived features . however , in the presence of gas , bars are not expected to be permanent features of galaxies , but should dissolve in much less than a hubble time owing to mass inflow into the nuclear region which can build up a central mass concentration and destroy a bar ( pfenniger and norman 1990 ) . the high frequency of bars has thus led to the idea that bars dissolve and reform many times during a hubble time ( combes 2004 ) . if this is the case , the distribution of bar strengths will tell us the relative amount of time a galaxy stays in a given bar state ( strong , weak , or non - barred ; bournaud & combes 2002 ; block et al . 2002 ) . block et al . ( 2002 ) and buta , laurikainen , & salo ( 2004=bls04 ) used the gravitational torque method ( gtm ; buta & block 2001 ; laurikainen & salo 2002 ) to derive maximum relative nonaxisymmetric torque strengths @xmath4 for the ohio state university bright galaxy survey ( osubgs , eskridge et al . 2002 ) , a statistically well - defined sample of nearby bright galaxies . block et al . ( 2001 , 2004 ) , bls04 , and laurikainen , salo , & rautiainen ( 2002 ) showed that @xmath4 correlates with deprojected bar ellipticity , a popular parameter suggested by athanassoula ( 1992 ) to be a useful ( although incomplete ) measure of bar strength ( e.g. , martin 1995 ; whyte et al . the correlation was found by laurikainen , salo , & rautiainen ( 2002 ) to be much better when objectively - measured near - ir ellipticities are used as opposed to the optical ellipticities estimated by martin ( 1995 ) from blue light photographs . the good correlation is very important , because the shape of the bar relates to the shape of the orbits which build up the bar , which should depend on the global force field . also , bls04 found that @xmath4 correlates well with the bar ellipticity parameter @xmath5 measured by whyte et al . @xmath4 is a bar strength indicator that is sensitive to the mass of the bar , and as such should be a better measure of bar strength than bar ellipticity . however , @xmath4 is affected also by spiral arm torques which can dominate over the torques due to weak bars . thus , @xmath4 alone can not tell us the actual distribution of bar strengths , but only the distribution for stronger bars . one way to derive the distribution of real bar strengths is to remove the spiral contribution to @xmath4 . buta , block , and knapen ( 2003=bbk03 ) developed a fourier - based method of separating bars from spirals that utilizes a symmetry assumption ( section 3 ) . block et al . ( 2004 ) applied this method to deep near - ir images of 17 bright galaxies to derive true bar strengths @xmath0 and spiral strengths @xmath3 . this analysis detected a possible correlation between @xmath0 and @xmath3 in the sense that among bars having @xmath0 @xmath6 0.3 , spiral strength increases with increasing bar strength . block et al . suggested that the apparent correlation implies that for stronger bars , the bar and the spiral grow together and have the same pattern speed . our goal with the present paper is to apply the bbk03 method to nearly 150 spiral galaxies in the osubgs , a database of @xmath7-band ( 1.65@xmath8 m ) images that have enough depth of exposure to allow reliable fourier analyses . in the @xmath7-band , the extinction is only 19% that in the @xmath9-band ( cardelli , clayton , & mathis 1989 ) , and such images are suitable for the derivation of gravitational potentials using fast - fourier transform techniques ( quillen , frogel , and gonzlez 1994 ; salo et al . 1999 ; laurikainen & salo 2002 ) . from the separated images , we derive the distributions of bar and spiral strengths and investigate what these tell us about disk galaxies . we also further investigate the correlation between @xmath0 and @xmath3 . our sample consists of 147 bright galaxies drawn from the same sample as used by bls04 , laurikainen , salo , and buta ( 2004a = lsb04 ) , and laurikainen et al . ( 2004b = lsbv04 ) . these previous studies used 180 galaxies , including 158 osubgs galaxies having total magnitudes @xmath10 @xmath2 12.0 , @xmath11 @xmath2 6@xmath125 , 0 @xmath1 @xmath13 @xmath1 9 , inclination @xmath2 65@xmath14 , and @xmath1580@xmath14 @xmath2 @xmath16 @xmath2 + 50@xmath14 . in addition , this sample included 22 galaxies from the two micron all - sky survey ( 2mass , skrutskie et al . 1997 ) which satisfy similar criteria as the osubgs but are larger than the 6@xmath125 diameter limit . however , the 2mass images are sufficiently underexposed that they proved inadequate for bar / spiral separation . whereas bars are detected fairly well in such images , the spirals and background disks are often too faint to characterize reliably , and we do not use these images further in this paper . figure [ histtypes ] shows that our subset of 147 osubgs galaxies is dominated mainly by sbc and sc galaxies . the base osu sample is typical of the bright galaxy population , as shown by eskridge et al . ( 2002 ) and whyte et al . bls04 and lsb04 showed that their osubgs/2mass sample is biased mainly against inclusion of very late - type , low luminosity barred spirals and low surface brightness galaxies . our subset of 147 galaxies has a similar bias . the bar / spiral separation method of bbk03 depends on a simple assumption concerning the behavior of the relative fourier intensity amplitudes as a function of radius in a bar : the relative intensities decline past a maximum in the same or a similar manner as they rose to that maximum . this is known as the _ symmetry assumption_. in a complicated bar and spiral system , only the rising portion of the symmetric curve is seen , as in bbk03 s example of ngc 6951 , and the symmetry assumption allows the extrapolation of the bar into the spiral region . the assumption is justified from studies of barred galaxies lacking strong spiral structure . bbk03 used the study of 6 barred galaxies from ohta et al . ( 1990 ) and the case of ngc 4394 from the osubgs to justify the assumption . the assumption has found further support in studies of sb0 and sb0/a galaxies from the near - infrared s0 survey ( buta et al . 2005 ; see buta 2004 for a preliminary summary of these results ) . in these cases , the bars in the near - infrared are observed against only bulge and disk components , so that the bar is the only significant nonaxisymmetric contribution . since we can not know _ apriori _ the form of any particular bar , we judge the effectiveness of a given bar / spiral separation by examining bar plus disk and spiral plus disk intensity maps ( see bbk03 ) . if the bar length is underestimated or overestimated , we can detect the failure as positive or negative residuals in the spiral plus disk image . the application of the bbk03 technique to the osubgs sample required a number of modifications . first , the method was developed using deep @xmath17 images with pixel sizes of 0@xmath1824 ( block et al . in contrast , the osubgs @xmath7-band images have pixel sizes ranging from 1@xmath1811 to 1@xmath1850 and are noisier at large radii than the @xmath17 images used by block et al . ( these two factors complicate separation , but the pixel size problem could be handled effectively by resampling the images into pixels one - quarter as large using iraf routine imlintran . in our analysis of the osubgs sample , we encountered a greater range of complications in the relative fourier intensity curves , such as multiple bars and the effects of decomposition errors on central isophotes . thus , it was necessary to adapt the bbk03 method to deal with the new complications . the effects of decomposition errors were particularly serious . for all separations , we used deprojected images based on bulge / disk / bar decompositions from lsbv04 . in each case , the bulge was assumed to be spherical , but in those cases where the assumption may be wrong , the decomposition led to symmetric regions of lower intensity on each side of the center where too much bulge light was subtracted . in about a dozen cases , the problem was sufficiently serious that bar / spiral separation could not be carried out . in most cases , the problem could be treated in a two - step separation process , which also proved very effective for cases with multiple bars or ovals . figure [ extraps ] shows the lower - order fourier representations used to separate the bars and spirals in 24 osubgs galaxies . these objects illustrate well the types of extrapolations needed to deal with the wide range of bar types found in the sample . while for many ( e.g. , ngc 289 , 864 , 1637 , 3261 , 3686 , 4027 , 4254 , 4303 , 4548 , 4995 , 5085 , 5483 , 5921 , and 6300 ) the symmetry assumption appears to work well , for others ( e.g. , ngc 613 , 1087 , 4457 , 4579 , 4593 , and the higher order terms for ngc 3261 and 4548 ) we found it effective to fit a single gaussian to the rising relative intensities ( or even a double gaussian , as in ngc 1087 ) . we also followed the procedure of bbk03 to extrapolate the bars as little as possible , so that if the observed relative fourier amplitudes due to the bar could be detected beyond the maximum , as much as possible of the decline is used as observed . two cases shown in figure [ extraps ] are ngc 1559 and ngc 2964 . it is likely that some bar profiles are indeed gaussian in nature , although the physical implication of such a representation is not explored here . some profiles are symmetric but not necessarily gaussian ( e.g. , ngc 1637 ) or are clearly asymmetric ( as in ngc 1087 , 1559 , and 2964 ) . the effectiveness of the separations of the 24 galaxies are shown in figure [ separations ] . in general , very good separations are possible by the bbk03 procedure . the partly gaussian - fitted bar representation for ngc 4548 has cleanly removed the bar with little residual bar light remaining . in ngc 4579 , gaussian fits to all the main bar fourier terms allows the inner part of the spiral to be more clearly seen . the complex bar in ngc 1087 , represented by two non - coincident gaussians , is well - separated from the complex spiral which itself appears to be affected by considerable star formation . this bar does not follow the symmetry assumption , except for the individual gaussian components . in many of the bar images , ring - like structures , sometimes slightly oval and sometimes weakly spiral , are seen . these rings in large part represent the axisymmetric component of the associated spirals . also , some failure of the bar extrapolations can cause these weak structures . in the spiral images also , one often sees a filling in of the inner part of the pattern . this is due to the axisymmetric part of the bar . maximum relative gravitational torques must be calculated against the total axisymmetric background , including whatever contributions the spirals and bars themselves make to this background . bbk03 noted that in some bars where the spiral is weak or absent , the maxima of the higher order bar terms shift towards larger radii ( e.g. , ohta et al . 1990 ) . these shifts are sometimes seen in our bar representations , but in many cases there is little or no shift detectable . also , figure 1 shows that our bar representation for ngc 4593 has fitted peaks in @xmath19=4 and 6 at smaller radii than for @xmath19=2 . the symmetry assumption leads to double - humped bar profiles in the strong bars of ngc 1300 and ngc 5921 . in both of these cases , the bar image includes a weak elongated ring pattern that contributes little to the torques . in both cases also , the spiral plus disk image looks reasonably bar - free , but a slight asymmetry in the bar leaves a small residual bar spot on the lower right end of the bar in ngc 5921 . separation was especially effective for inner ovals . small ovals in ngc 4254 , 5085 , 5247 , 5248 , and 5483 were easily mapped and removed with just a few terms . in the case of ngc 5085 , we used the symmetry assumption twice , one for the inner oval and one for an outer oval ( figure [ extraps]s ) . the bar mapping for ngc 5248 shows a double - humped profile that could in principle be represented as a double - gaussian . in a few cases ( e.g. , ngc 613 , 4457 , 4579 , 4593 ) , the separation successfully removed the primary bar but left a small oval in the center . these ovals tend to be weak compared to the primary bar , and were sometimes left in the spiral plus disk image . in other cases , a two - step process could be used to remove them from the spiral plus disk image , if necessary . we were able to carry out reasonably successful bar / spiral separations for 147 osubgs galaxies . the 33 missing objects from the lsbv04 sample include the original 22 2mass galaxies in their sample and 11 other cases where the decompositions left complex residual structure in the bulge region that prevented a reliable fourier extrapolation of the bar light . for the 147 separated galaxies , we computed bar and spiral strengths following lsbv04 . gravitational potentials are inferred from the bar and spiral images assuming a constant mass - to - light ratio . the potentials are derived from poisson s equation using a fast - fourier transform technique . a polar grid approach is used to minimize the effects of noise at large radii ( e.g. , laurikainen and salo 2002 ) . vertical thickness is taken into account using an exponential density function having a scaleheight @xmath20 scaled from the radial scalelength @xmath21 using a type - dependent formulation ( de grijs 1998 ) . for each image , the radial variation , @xmath22 , of the maximum tangential force relative to the mean background radial force is computed . then the maximum ratio from the bar image defines @xmath0 and the radius @xmath23 and the maximum ratio from the spiral image defines @xmath3 and @xmath24 . figure [ schematic ] shows a schematic of these definitions , based on ngc 6951 ( from bbk03 ) . since bar / spiral separation uses mainly even fourier terms for the bar , the procedure leaves the odd fourier terms and much of the image noise in the spiral plus disk image . thus , it was necessary to inspect the plots for the spirals to eliminate spurious maxima due to noise at large radii . the bbk03 definition of @xmath4 , shown in figure [ schematic ] , differs slightly from that actually used in bls04 , lsb04 , and lsbv04 , where @xmath4 was taken to be the maximum @xmath22 in the bar / oval region when such features were present , and the general maximum @xmath22 for mostly non - barred spirals . in general , the differences between the formal @xmath4 defined by bbk03 and that used in the previously cited papers will not be large , but were necessitated by the higher noise level in the osubgs images compared to those used by bbk03 and block et al . ( 2004 ) . table 1 lists the derived parameters for the 147 galaxies . the typical uncertainties in maximum relative gravitational torques are discussed by bbk03 , bls04 , and lsbv04 . in the present paper , we note that the uncertainty in the constant mass - to - light ratio assumption , as well as the effects of disk truncation , will likely affect @xmath0 and @xmath3 differently , because @xmath24 can significantly exceed @xmath23 , as shown for many cases in table 1 . in bls04 , we showed that the typical dark halo correction to @xmath4 is about 5% for the galaxies that define the osubgs sample , based on a universal rotation curve " analysis ( persic , salucci , & stel 1996 ) . since @xmath25 will generally be intermediate between @xmath23 and @xmath24 , we expect the dark halo contribution affects @xmath3 more than @xmath0 . to minimize the effects of disk truncation , we chose a maximum radius ( radmax ) for all calculations of 127 pixels , which is the maximum circle contained in each image . radmax is the radius of the zone for which and from which forces are calculated . laurikainen & salo ( 2002 ) showed that as long as radmax/@xmath25 @xmath6 2 , disk truncation should not significantly affect the force ratio . in almost all cases , @xmath23 satisfies this condition , but @xmath24 may or may not satisfy it . thus , our derived @xmath3 values can not be taken as definitive , but as indicative of the approximate spiral strength . some @xmath3 values are also affected by star formation , and in general @xmath3 is probably overestimated on average in our analysis . we examine correlations between our measured @xmath0 and @xmath3 values and other parameters . figure [ byfam ] shows plots of @xmath26 and @xmath27 versus rc3 family classification . figure [ byfam]a shows a virtually linear correlation between @xmath26 and rc3 family over all types , verifying that the intermediate de vaucouleurs family class sab is justified ( see also block et al . 2001 ) . this is the case even when the sample is divided into types @xmath28 ( figure [ byfam]c ) and @xmath29 ( figure [ byfam]e ) . the right panels of figure [ byfam ] show only weaker correlations of @xmath27 with bar family . in all three panels , @xmath27 is higher for sb galaxies than for sa galaxies . table 2 summarizes the mean values displayed . table 1 also lists the `` osu @xmath30 '' and osu @xmath7 " family classifications from eskridge et al . ( 2002 ) , based on visual inspection of the osubgs @xmath30-band and @xmath7-band images . eskridge et al . noted that many galaxies classified as non - barred or weakly barred in rc3 appear more strongly barred in the near - ir . we have computed a @xmath0 family " to compare with these estimates ( see table 3 ) . we define an sa galaxy to be one which has @xmath0 @xmath2 0.05 , while an sb galaxy is defined to have @xmath0 @xmath31 0.25 . between these extremes we define classes s@xmath32b , sab , and sa@xmath33 using the notation of de vaucouleurs ( 1963 ) , designed to illustrate the continuity aspect of bar strength . the different @xmath0 families do not involve equal steps in @xmath0 : the sb category involves a much broader range of bar strengths than does sa , and we give a broader range to sab compared to s@xmath32b and sa@xmath33 . comparison between the @xmath0 family and the osu @xmath7 family shows the two often disagree . many osu @xmath7 sb galaxies end up as @xmath0 family sab because the bars are really not that strong . table 4 summarizes 6 galaxies from eskridge et al . ( 2000 ) noted to have changed classification from sa to sb in going from the @xmath30 to the @xmath7 band . however , the @xmath0 family indicates the bars are still weak , even in the near - ir . some disagreements occur for cases where the spiral comes off the ends of the bar with a large pitch angle , an example being ngc 1042 . in this case , there is an oval that we have interpreted as all or some of the bar for bar / spiral separation . in other cases , the @xmath0 family gives an sab or sa@xmath33 family for cases which are clearly sb in blue light . some notable examples are ngc 4643 , 5101 , and 5701 , all early - type spirals . in these cases , the bars are simply not as strong as they appear to be because of the presence of a strong bulge component , which contributes significantly to the axisymmetric background . table 5 summarizes a general comparison between the @xmath0 family and the other sources of bar family classification , including the classification of lsbv04 whereby a fourier bar " is defined to be one where the phases of the main @xmath19=2 and @xmath19=4 fourier components are maintained nearly constant in the bar region . the most striking aspect of the rc3 comparison is the number of objects classified as sa in rc3 which have a @xmath0 family of s@xmath32b , meaning some bar or oval was detected in the near - ir . a similar comparison is found for the osu @xmath30 classifications , which is not surprising since rc3 families are also based on @xmath30-band images . the comparison with the osu @xmath7 classifications shows that , as highlighted before , many osu @xmath7 sb galaxies have bars that are not that strong , and come out with a @xmath0 family of sab . the fourier bar comparison gives very similar results but requires mainly that the low order fourier phases be constant , not that the bar should be especially strong . figure [ byparams ] shows plots of @xmath26 and @xmath27 versus : ( a , b ) the extinction- and tilt - corrected absolute blue magnitude @xmath34 based on parameters from rc3 and distances from tully ( 1988 ) ; ( c , d ) the extinction and tilt - corrected mean effective blue light surface brightness @xmath35 ( mag arcmin@xmath36 ; see equation 71 of rc3 ) ; and ( e , f ) de vaucouleurs revised hubble - sandage type , coded on the rc3 numerical scale . little correlation with absolute magnitude is found , although this is partly due to the fact that the sample is biased against late - type , low luminosity barred spirals ( bls04 ) . except for the lowest surface brightness bin , there is little correlation between @xmath26 and @xmath35 , while some correlation between @xmath27 and @xmath35 is found . the mean spiral strength appears to increase with increasingly fainter surface brightness , changing by more than a factor of two , an effect that may be partly due to star formation and partly due to increased image noise for the lower surface brightness objects . figure [ byparams]e shows the same kind of correlation between @xmath26 and type discussed by bls04 and lsb04 for @xmath37 , in the sense that maximum relative gravitational torques are larger for later types . figure [ byparams]f shows the same may be true for spiral strengths as well . bls04 and lsb04 attributed the type dependence in @xmath4 as due to the increased prominence of the bulge in early - type spirals . this rather counterintuitive result , that significant - looking bars in early - types actually have weaker relative torques than those in later types , is due to the fact that what the eye recognizes in photographs as a bar strength" is the local surface density contrast , which is different from the tangential to radial force ratio or its maximum value @xmath4 . the latter is a global quantity , measuring forces from all parts of the galaxy , and should be a more reliable indicator of actual bar strength . this highlights the advantage of the gtm in quantifying bar strength beyond the visual appearance of bars ( lsb04 ) . early - type bars may in fact be more massive and intrinsically stronger than those in later types , but relative to the axisymmetric disks they are imbedded within , late - type bars can be stronger . figure [ bymoreparams ] shows correlations between @xmath26 and @xmath27 and @xmath38 , the rc3 logarithmic standard isophotal axis ratio ( used as an indicator of inclination ) , and visually estimated arm class ( elmegreen and elmegreen 1987 ) . arm classes ( acs ) emphasize the symmetry and continuity of spiral arms , and it is worth investigating whether these might correlate with @xmath3 . no significant trend of either nonaxisymmetric strength parameter is found with @xmath38 , confirming that there is no systematic bias introduced to the torques due to deprojection corrections . however , we detect a weak correlation between @xmath27 and ac in the sense that @xmath27 is higher for acs of 9 and 12 ( there are no acs of types 10 and 11 ) , which include the most symmetric , longest arms , compared to acs of 13 , which include the chaotic , fragmented arms seen in flocculent spirals . in spite of the apparent correlation , @xmath3 is not necessarily a suitable replacement for the ac because there is considerable overlap among the classes and the two parameters measure different aspects of spiral structure . finally , figures [ bymoreparams]e , f show @xmath26 and @xmath27 for sb galaxies as functions of rc3 variety classification : ringed ( r ) , pseuodringed ( rs ) , and spiral - shaped ( s ) . with our direct estimates of bar strength , we can investigate the claim made by kormendy & kennicutt ( 2004 ) that sb(r ) galaxies have stronger bars than sb(s ) galaxies , based on the hydrodynamic simulations of sanders & tubbs ( 1980 ) . table 2 summarizes @xmath26 and @xmath27 for the three varieties separated by family . figure [ bymoreparams]e shows that , on average , sb(r ) galaxies have weaker bars than sb(s ) galaxies , contrary to the conclusion of kormendy & kennicutt . in our sample also , sb(rs ) galaxies have bars as strong on average as those in sb galaxies . the differences are not that significant owing to the large scatter at each variety . table 2 shows that the statistics are more uncertain for sa galaxies , which strongly emphasize the ( s)-variety , and sab galaxies , which strongly emphasize the ( rs ) variety . figures [ histos]a and b show both differential and cumulative distributions of @xmath0 , while figures [ histos]c and d show the same for @xmath3 , for 147 osubgs galaxies . for comparison , the distributions of @xmath4 values ( from lsbv04 ) for the same galaxies are shown in figures [ histos]e and f. figures [ histos]a and b show that when bars are isolated from spirals in galaxy images , the lowest bar strength bins , 0.0 - 0.05 and 0.05 - 0.10 , fill up considerably over the @xmath4 bins . more than 40% of the galaxies have bars with @xmath0 @xmath1 0.10 , while only 22% have @xmath4 @xmath1 0.10 . it is clear that weak bars or ovals are often masked by spirals and not detected via the @xmath4 parameter ; these bars are visible in @xmath22 profiles but have force maxima much lower than those induced by spiral arms in the outer parts of the disks . thus , @xmath4 does not give a reliable indication of the relative frequency of weak bars . for the spirals , the lowest @xmath3 bin is deficient in galaxies compared to the next highest @xmath3 bin , which is not unexpected given that the sample excludes s0 galaxies . most spirals are nevertheless fairly weak , with more than 75% having @xmath39 0.20 . these parameters allow us to assign all the sample galaxies to bar and spiral strength classes ( see table 1 ) . we follow buta & block ( 2001 ) to make these assignments . for bar class 0 we include any galaxy having @xmath0 @xmath2 0.05 , while for spiral class 0 we include any galaxy having @xmath3 @xmath2 0.05 . for bar class 1 we include galaxies having 0.05@xmath1 @xmath0 @xmath20.15 , while for spiral class 1 we include galaxies having 0.05@xmath1 @xmath3 @xmath20.15 , etc . thus , the 0 class for bars and spirals involves a narrower range since @xmath0 and @xmath3 can not be negative as defined . these spiral and bar classes define a quantitative near - infrared classification of bars and spirals and can be incorporated into the dust - penetrated classification scheme of block & puerari ( 1999 ; see buta & block 2001 ) . while bar class may represent a suitable replacement for de vaucouleurs family classifications , spiral class only distinguishes early and late - type spirals and does not discriminate well between individual @xmath13-types ( figure [ byparams]f ) . elmegreen & elmegreen ( 1985 ) used bar / interbar and arm / interarm contrasts to show that strong spirals are associated with strong bars . although the bulk of their correlation is based on only a few galaxies , the implication was that the bars might be driving the spirals . however , sellwood & sparke ( 1988 ) used numerical simulations to show that bars and spirals might be independent features , with different pattern speeds . block et al . ( 2004 ) applied the bbk03 technique to 17 intermediate to late - type spirals and found some correlation between bar and spiral arm torques , but only for the strongest bars . these authors suggested that in strongly - barred galaxies , the bar and the spiral may be growing together and have the same pattern speed . figure [ qsqb ] shows the correlation between @xmath3 and @xmath0 ( crosses ) for the osubgs sample . for comparison , the values from block et al . ( 2004 ) for 17 bright spirals are plotted also . the solid curve shows the medians in @xmath3 for successive bins of 0.1 in @xmath0 . the plot shows that the median @xmath3 increases from 0.1 to 0.30 as @xmath0 increases from 0.05 to 0.75 . the rise agrees with that found by block et al . ( 2004 ) within the uncertainties , and again suggests that at lower bar strengths , spiral and bar strengths are largely uncorrelated , while at stronger bar strengths some correlation may be present . the result is difficult to interpret because the numbers of galaxies decrease significantly with increasing @xmath0 . also , @xmath0 and @xmath3 have correlated uncertainties , in the sense that if @xmath0 is overestimated by the separation procedure , then @xmath3 will be underestimated and vice versa . these uncertainties could be reduced with better quality images as used by block et al . our results largely support the idea of sellwood & sparke ( 1988 ) that spirals and bars are independent features , with likely different pattern speeds , at least for @xmath40 0.3 . this is not definitive , however , because as noted by block et al . ( 2004 ) , the frequent alignment of bars and rings , which are often parts of the spiral pattern , implies similar pattern speeds in some cases . for higher bar strengths , some correlation between @xmath0 and @xmath3 may be present that can only be confirmed with a larger sample of strongly barred spirals . bar strength in isolated disk galaxies is thought to be determined largely by the effectiveness with which a bar can transfer angular momentum to other galactic components , such as spiral structure , resonances , live halos , and outer bulge stars ( athanassoula 2003 ) . a bar can get very strong if there is nothing to negate this effect . however , a bar can affect its own evolution by driving gas into the center . this builds up the central mass concentration and can lead to an inner lindblad resonance ( ilr ) , which will feed angular momentum to the bar . when this happens , the bar s orbital structure can be destroyed and the bar itself fades away ( norman , sellwood , & hasan 1996 ) . bar strength in non - isolated galaxies can be affected by tidal interactions ( noguchi 1996 ; miwa & noguchi 1998 and also accretion of gas - rich dwarfs or infalling external gas ( sellwood & moore 1999 ; bournaud & combes 2002 ; combes 2004 ) . miwa and noguchi ( 1998 ) have argued that which bar - forming mechanism ( spontaneous or tidal ) dominates depends on the relative importance of the disk and halo . they suggest that spontaneous bars will be important if disks are massive relative to their halo , while tidally - induced bars will dominate if the disks are stable against spontaneous bar formation . noguchi ( 1987 ) suggested that the exponential " and flat " bars of elmegreen & elmegreen ( 1985 ) are distinguished by these same two mechanisms , with the former being spontaneous and the latter being tidally triggered . if gas flow helps to dissolve a bar , an interaction may regenerate a bar if there is little disk gas remaining ( berentzen et al . if a galaxy accretes substantial external gas , it may be susceptible to multiple or recurring bar episodes ( bournaud & combes 2002 , 2004 ) . several simulation studies ( athanassoula 2003 ; athanassoula & misiriotis 2002 ; miwa & noguchi 1998 ; berentzen et al . 2004 ) have found a correlation between bar strength and bar pattern speed , in the sense that stronger bars have lower pattern speeds . these results suggest that bar strength is not a permanent feature of galaxies , but can be highly variable over a hubble time . evidence in support of this idea comes from the inverse correlation between central mass concentration and bar ellipticity in a sample of spiral galaxies ( das et al . thus , the distribution of bar strengths in galaxies may be influenced by a complex variety of effects : environment , mass distribution , the interstellar medium , and the properties of dark matter halos . we have shown in this paper that a straightforward fourier technique can be used to separate bars from spirals , allowing us to examine the distribution of bar strengths in galaxies unaffected by the torques due to spirals . we find a preponderance of low bar strengths that was masked in previous @xmath4 studies partly because of the effects of spiral arm torques . @xmath4 is only reliable as a bar strength indicator if the bar is the dominant nonaxisymmetric feature in a galactic disk . in cases where the spiral dominates , or where the bar and spiral have comparable strengths , @xmath4 will be an overestimate of bar strength . the reason for wanting to look at the distribution of bar strengths alone is sellwood s ( 2000 ) assertion that most real bars are not made by the bar instability . " this global dynamical instability was first identified in @xmath41-body models that showed that a disk - shaped galaxy having sufficient kinetic energy in ordered rotational motion would be unstable to the formation of a bar ( sellwood 1996 and references therein ) . the way to avoid the linear instability would be a high central concentration , guaranteeing the existence of an ilr inside the bar . sellwood ( 2000 ) noted that many strong bars , such as those found in galaxies like ngc 1300 and 1433 , include small circumnuclear rings whose presence has been tied to the existence of an ilr region ( although the exact locations of the rings may not be coincident with ilr ; regan and teuben 2003 ) . sellwood noted that enough barred galaxies showed these features to cast considerable doubt on the bar instability as being the explanation of most bars . other features of the strong bars that suggest the influence of ilrs are the shapes of offset dust lanes ( athanassoula 1992 ) and observed gas velocity fields . sellwood also brought attention to the results of early high redshift studies ( e.g. , abraham et al . 1999 ) that indicated bars are less frequent for @xmath42 @xmath6 0.5 , suggesting that bars develop long after the disk formed . however , this conclusion has been refuted by more recent studies ( sheth et al . 2003 ; elmegreen , elmegreen , & hirst 2004 ; jogee et al . 2004 ) , which indicate no significant drop in the bar fraction out to @xmath42@xmath431 . jogee et al . ( 2004 ) present the most comprehensive study of bar fraction as a function of redshift , and found that this fraction is virtually constant at 30%@xmath446% to @xmath42=1 . the implications of their result are thought to be that cold , unstable disks were already in place by @xmath42=1 , and that bars must survive at least 2 gyr . in looking for an alternative to the bar instability , sellwood ( 2000 ) suggested bar growth occurs through an episodic process , where the interaction between a bar and a spiral can add particles to the bar and make it longer , while at the same time reducing the bar s pattern speed . he suggested that it would be useful to be able to predict the distribution of bar strengths for various bar formation scenarios . of course , it is also useful to know the observed distribution of bar strengths . the bbk03 method and the osubgs have allowed us to consider this for the first time . the only theoretical predictions of an expected distribution of bar strengths have been made for recurrent bar formation models ( bournaud & combes 2002 ) . block et al . ( 2002 ) used a preliminary @xmath4 analysis of the osubgs sample to derive an observed distribution , and then used the bournaud & combes simulation database to derive a theoretical distribution using as much as possible the same assumptions : constant mass - to - light ratio ; exponential vertical density distribution having @xmath45 ; inclusion of spiral torques ; bulges assumed as flat as the disks ; and dark matter ignored . these authors noted that the observed @xmath4 distribution shows a deficiency of low @xmath4 galaxies and an extended tail " of high @xmath4 galaxies . the comparison showed that both characteristics were best explained if galaxies are open systems , accreting enough external gas to double their mass in a hubble time . the distribution of bar strengths would then mainly tell us the relative amount of time galaxies spend in different bar states ( strong , weak , or non - barred ) . the deficiency of low @xmath4 galaxies was interpreted as due to the duty cycle " between bar episodes . that is , accretion prevented most galaxies from spending much time in a perfectly axisymmetric state . some of the nonaxisymmetric torques could be due to spirals which would also be maintained by accretion . the refined gtm analysis carried out by bls04 , lsb04 , and lsbv04 provided a more reliable distribution of @xmath4 . bls04 showed that even with refinements that account properly for bulge shapes , and using improved estimates of @xmath20 that allow for the type dependence of @xmath46 as well as values of @xmath21 derived from two - dimensional bar / bulge / disk decompositions , the observed distribution of @xmath4 still shows a deficiency of objects having @xmath4 @xmath2 0.05 and an extended tail of high @xmath4 objects . however , the refined distribution shows more low @xmath4 values than did the block et al . ( 2002 ) analysis , due to a variety of effects discussed by bls04 . we find that when spiral torques are removed , the distribution of bar strengths is a relatively smoothly declining function with increasing @xmath0 . it appears that galaxies spend more time in a relatively weakly - barred or non - barred state than they do in a strongly - barred state . even in these weakly - barred states , they can have significant spiral torques . the question now is whether gas accretion models can account for the actual distribution of bar strengths rather than simply the distribution of total nonaxisymmetry strengths . in principle , a separation analysis could be made for simulations as for images . whyte et al . ( 2002 ) analyzed the blue and near - infrared images in the osubgs , and derived a quantitative bar strength parameter , @xmath5 , which is a rescaled measure of bar ellipticity ( abraham & merrifield 2000 ) . for the large and well - defined osubgs sample , they derived a distribution of @xmath5 which they claim shows evidence for bimodality , and argue that the bimodality is likely due to rapid evolution from the sb phase to sa and sab phases . however , the distribution of @xmath0 suggests a continuous distribution of bar strengths , with no evidence of bimodality . the two results are not really in disagreement because the evidence for bimodality in @xmath5 is very weak , especially in the plot of @xmath5 versus concentration shown by whyte et al . the original evidence was found in this same kind of plot by abraham & merrifield ( 2000 ) . in agreement with our results , whyte et al . ( 2002 ) found that sab galaxies have values of @xmath5 intermediate between sa and sb galaxies . using a simple fourier technique , we have separated the bars and spirals in 147 osubgs galaxies , and for the first time derived the distribution of actual bar strengths in disk galaxies . we find that the relative frequency of bars is a declining function of bar strength , with more than 40% of the sample being very weakly - barred or non - barred with @xmath0 @xmath2 0.1 . the higher frequency of weak bars compared to strong ones suggests that strong bars are either very transient or may require more special conditions , such as an interaction . if , in fact , bars are long - lived , as suggested by the results of high redshift studies ( e.g. , jogee et al . 2004 ) , then the observed distribution of bar strengths is telling us that cold , unstable disks preferentially form weak bars . an important piece of the whole picture of barred galaxies is still missing : sb0 galaxies . block et al . ( 2002 ) suggested that in the absence of gas , bars are very robust and can last a hubble time . what is the distribution of bar strengths in such galaxies ? our sb0 survey ( buta et al . 2005 ; buta 2004 ) should be able to answer this question . we thank an anonymous referee for helpful comments on the manuscript . rb and sv acknowledge the support of nsf grant ast-0205143 to the university of alabama . el and hs acknowledge the support of the academy of finland , and el also from the magnus ehrnrooth foundation . sv acknowledges the support of the academy of finland during two summer visits to oulu in 2002 and 2003 . funding for the osu bright galaxy survey was provided by grants from the national science foundation ( grants ast-9217716 and ast-9617006 ) , with additional funding from the ohio state university . buta , r. 2004 , in penetrating bars through masks of cosmic dust , the hubble tuning fork strikes a new note , d. l. block , i. puerari , k. c. freeman , r. groess , & e. k. block , eds . , springer , dordrecht , 101 combes , f. 2004 , in penetrating bars through masks of cosmic dust , the hubble tuning fork strikes a new note , d. l. block , i. puerari , k. c. freeman , r. groess , & e. k. block , eds . , springer , dordrecht , 57 llllccrrccl ngc 150 & sb & sab & sb & 0.475 & 0.254 & 23 & 33 & 5 & 3 & sb + ngc 157 & sab & sb & sa & 0.024 & 0.323 & 3 & 21 & 0 & 3 & sa + ngc 210 & sab & sa & sb & 0.052 & 0.037 & 29 & 63 & 1 & 0 & s@xmath32b + ngc 278 & sab & sa & sa & 0.046 & 0.064 & 5 & 19 & 0 & 1 & sa + ngc 289 & sb & sb & sb & 0.212 & 0.089 & 11 & 49 & 2 & 1 & sa@xmath33 + ngc 428 & sab & sab & sb & 0.254 & 0.100 & 19 & 50 & 3 & 1 & sb + ngc 488 & sa & sa & sa & 0.028 & 0.020 & 11 & 40 & 0 & 0 & sa + ngc 578 & sab & sab & sb & 0.180 & 0.168 & 9 & 37 & 2 & 2 & sab + ngc 613 & sb & sb & sb & 0.298 & 0.319 & 39 & 65 & 3 & 3 & sb + ngc 685 & sab & sb & sb & 0.389 & 0.157 & 9 & 23 & 4 & 2 & sb + ngc 864 & sab & sab & sb & 0.321 & 0.134 & 13 & 19 & 3 & 1 & sb + ngc 1042 & sab & sab & sab & 0.044 & 0.530 & 5 & 21 & 0 & 5 & sa + ngc 1058 & sa & sa & sa & 0.129 & 0.097 & 15 & ... & 1 & 1 & sab + ngc 1073 & sb & sb & sb & 0.561 & 0.264 & 15 & 29 & 6 & 3 & sb + ngc 1084 & sa & sa & sa & 0.038 & 0.197 & 5 & 23 & 0 & 2 & sa + ngc 1087 & sab & sb & sb & 0.428 & 0.265 & 5 & 25 & 4 & 3 & sb + ngc 1187 & sb & sb & sb & 0.117 & 0.183 & 17 & 31 & 1 & 2 & sab + ngc 1241 & sb & sab & sb & 0.181 & 0.153 & 11 & 19 & 2 & 2 & sab + ngc 1300 & sb & sb & sb & 0.524 & 0.184 & 57 & 111 & 5 & 2 & sb + ngc 1302 & sb & sab & sb & 0.061 & 0.033 & 17 & 89 & 1 & 0 & s@xmath32b + ngc 1309 & sa & sa & sab & 0.091 & 0.132 & 9 & 15 & 1 & 1 & s@xmath32b + ngc 1317 & sab & sb & sb & 0.085 & 0.031 & 35 & 83 & 1 & 0 & s@xmath32b + ngc 1371 & sab & sab & sab & 0.049 & 0.109 & 7 & 19 & 0 & 1 & sa + ngc 1385 & sb & sb & sb & 0.269 & 0.262 & 3 & 31 & 3 & 3 & sb + ngc 1493 & sb & sab & sb & 0.319 & 0.159 & 9 & 19 & 3 & 2 & sb + ngc 1559 & sb & sb & sb & 0.328 & 0.185 & 5 & 45 & 3 & 2 & sb + ngc 1617 & sb & sa & sab & 0.034 & 0.078 & 7 & 35 & 0 & 1 & sa + ngc 1637 & sab & sa & sb & 0.193 & 0.066 & 11 & 17 & 2 & 1 & sab + ngc 1703 & sb & sa & sab & 0.073 & 0.097 & 9 & 23 & 1 & 1 & s@xmath32b + ngc 1792 & sa & sa & sa & 0.060 & 0.150 & 5 & 31 & 1 & 2 & s@xmath32b + ngc 1832 & sb & sab & sb & 0.176 & 0.131 & 11 & 39 & 2 & 1 & sab + ngc 2090 & sa & sab & sa & 0.087 & 0.090 & 9 & 17 & 1 & 1 & s@xmath32b + ngc 2139 & sab & sb & sb & 0.356 & 0.198 & 3 & 21 & 4 & 2 & sb + ngc 2196 & sa & sa & sa & 0.069 & 0.094 & 7 & 107 & 1 & 1 & s@xmath32b + ngc 2442 & sab & sb & sb & 0.412 & 0.600 & 45 & 71 & 4 & 6 & sb + ngc 2559 & sb & sab & sb & 0.334 & 0.169 & 25 & 43 & 3 & 2 & sb + ngc 2566 & sb & sb & sb & 0.270 & 0.220 & 45 & 79 & 3 & 2 & sb + ngc 2775 & sa & sa & sa & 0.037 & 0.043 & 11 & 33 & 0 & 0 & sa + ngc 2964 & sab & sa & sab & 0.270 & 0.110 & 13 & 24 & 3 & 1 & sb + ngc 3059 & sb & sb & sb & 0.533 & 0.305 & 8 & 113 & 5 & 3 & sb + ngc 3166 & sab & sa & sb & 0.108 & 0.073 & 21 & 77 & 1 & 1 & sab + ngc 3169 & sa & sa & sa & 0.089 & 0.036 & 11 & 113 & 1 & 0 & s@xmath32b + ngc 3223 & sa & sa & sa & 0.025 & 0.047 & 7 & 81 & 0 & 0 & sa + ngc 3227 & sab & sab & sb & 0.151 & 0.078 & 21 & 43 & 2 & 1 & sab + ngc 3261 & sb & sab & sb & 0.166 & 0.100 & 15 & 30 & 2 & 1 & sab + ngc 3275 & sb & sb & sb & 0.183 & 0.166 & 19 & 102 & 2 & 2 & sab + ngc 3319 & sb & sb & sb & 0.537 & 0.309 & 9 & 75 & 5 & 3 & sb + ngc 3338 & sa & sab & sab & 0.049 & 0.076 & 5 & 33 & 0 & 1 & sa + ngc 3423 & sa & sa & sa & 0.037 & 0.163 & 7 & 63 & 0 & 2 & sa + ngc 3504 & sab & sb & sb & 0.286 & 0.069 & 19 & 33 & 3 & 1 & sb + ngc 3507 & sb & sb & sb & 0.188 & 0.098 & 13 & 19 & 2 & 1 & sab + ngc 3513 & sb & sb & sb & 0.521 & 0.293 & 13 & 47 & 5 & 3 & sb + ngc 3583 & sb & sab & sb & 0.170 & 0.189 & 9 & 17 & 2 & 2 & sab + ngc 3593 & sa & sa & sa & 0.151 & 0.010 & 7 & 60 & 2 & 0 & sab + ngc 3596 & sab & sa & sab & 0.080 & 0.200 & 7 & 30 & 1 & 2 & s@xmath32b + ngc 3646 & si & sa & sab & 0.081 & 0.260 & 5 & 42 & 1 & 3 & s@xmath32b + ngc 3675 & sa & sab & sb & 0.078 & 0.083 & 11 & 49 & 1 & 1 & s@xmath32b + ngc 3681 & sab & sab & sb & 0.187 & 0.070 & 5 & 19 & 2 & 1 & sab + ngc 3684 & sa & sab & sab & 0.086 & 0.163 & 3 & 113 & 1 & 2 & s@xmath32b + ngc 3686 & sb & sab & sb & 0.225 & 0.082 & 7 & 15 & 2 & 1 & sa@xmath33 + ngc 3726 & sab & sab & sb & 0.212 & 0.174 & 17 & 41 & 2 & 2 & sa@xmath33 + ngc 3810 & sa & sa & sab & 0.049 & 0.110 & 7 & 11 & 0 & 1 & sa + ngc 3887 & sb & sab & sb & 0.093 & 0.175 & 9 & 23 & 1 & 2 & s@xmath32b + ngc 3893 & sab & sa & sab & 0.122 & 0.132 & 9 & 47 & 1 & 1 & sab + ngc 3938 & sa & sa & sa & 0.022 & 0.052 & 11 & 37 & 0 & 1 & sa + ngc 3949 & sa & sab & sab & 0.171 & 0.269 & 3 & 17 & 2 & 3 & sab + ngc 4027 & sb & sb & sb & 0.569 & 0.316 & 3 & 19 & 6 & 3 & sb + ngc 4030 & sa & sa & sa & 0.020 & 0.059 & 5 & 53 & 0 & 1 & sa + ngc 4051 & sab & sb & sb & 0.097 & 0.257 & 23 & 45 & 1 & 3 & s@xmath32b + ngc 4123 & sb & sb & sb & 0.331 & 0.195 & 21 & 31 & 3 & 2 & sb + ngc 4136 & sab & sab & sb & 0.150 & 0.114 & 7 & 17 & 2 & 1 & sab + ngc 4138 & sa & s & s & 0.039 & 0.035 & 5 & 17 & 0 & 0 & sa + ngc 4145 & sab & sab & sb & 0.427 & 0.124 & 3 & 25 & 4 & 1 & sb + ngc 4151 & sab & sb & sb & 0.114 & 0.039 & 43 & 87 & 1 & 0 & sab + ngc 4212 & sa & sa & sab & 0.060 & 0.210 & 5 & 19 & 1 & 2 & s@xmath32b + ngc 4242 & sab & sb & sb & 0.225 & 0.050 & 29 & 60 & 2 & 1 & sa@xmath33 + ngc 4254 & sa & sa & sab & 0.098 & 0.101 & 9 & 51 & 1 & 1 & s@xmath32b + ngc 4303 & sab & sb & sb & 0.075 & 0.243 & 13 & 27 & 1 & 2 & s@xmath32b + ngc 4314 & sb & sb & sb & 0.439 & 0.084 & 35 & 61 & 4 & 1 & sb + ngc 4394 & sb & sb & sb & 0.259 & 0.070 & 21 & 41 & 3 & 1 & sb + ngc 4414 & sa & sa & sa & 0.088 & 0.143 & 7 & 21 & 1 & 1 & s@xmath32b + ngc 4450 & sa & sa & sb & 0.116 & 0.085 & 25 & 63 & 1 & 1 & sab + ngc 4457 & sab & sa & sb & 0.078 & 0.050 & 19 & 41 & 1 & 1 & s@xmath32b + ngc 4487 & sab & sab & sb & 0.178 & 0.070 & 7 & 34 & 2 & 1 & sab + ngc 4504 & sa & sa & sb & 0.075 & 0.138 & 7 & 23 & 1 & 1 & s@xmath32b + ngc 4548 & sb & sb & sb & 0.285 & 0.155 & 33 & 51 & 3 & 2 & sb + ngc 4571 & sa & sa & sa & 0.022 & 0.080 & 3 & 30 & 0 & 1 & sa + ngc 4579 & sab & sb & sb & 0.188 & 0.050 & 21 & 49 & 2 & 1 & sab + ngc 4580 & sab & sa & sa & 0.077 & 0.088 & 7 & 13 & 1 & 1 & s@xmath32b + ngc 4593 & sb & sb & sb & 0.263 & 0.104 & 37 & 53 & 3 & 1 & sb + ngc 4618 & sb & sb & sb & 0.354 & 0.197 & 7 & 67 & 4 & 2 & sb + ngc 4643 & sb & sb & sb & 0.245 & 0.039 & 27 & 45 & 2 & 0 & sa@xmath33 + ngc 4647 & sab & sb & sb & 0.108 & 0.112 & 7 & 57 & 1 & 1 & sab + ngc 4651 & sa & sa & sab & 0.061 & 0.095 & 7 & 13 & 1 & 1 & s@xmath32b + ngc 4654 & sab & sab & sb & 0.136 & 0.144 & 5 & 45 & 1 & 1 & sab + ngc 4665 & sb & sb & sb & 0.257 & 0.037 & 25 & 73 & 3 & 0 & sb + ngc 4689 & sa & sa & sa & 0.050 & 0.067 & 13 & 39 & 1 & 1 & s@xmath32b + ngc 4691 & sb & sb & sb & 0.499 & 0.063 & 9 & 87 & 5 & 1 & sb + ngc 4698 & sa & sa & sa & 0.088 & 0.059 & 45 & 105 & 1 & 1 & s@xmath32b + ngc 4699 & sab & sb & sb & 0.138 & 0.030 & 9 & 19 & 1 & 0 & sab + ngc 4772 & sa & sa & sb & 0.042 & 0.030 & 45 & 63 & 0 & 0 & sa + ngc 4775 & sa & sa & sa & 0.105 & 0.125 & 3 & 27 & 1 & 1 & sab + ngc 4781 & sb & sab & sb & 0.205 & 0.312 & 7 & 17 & 2 & 3 & sa@xmath33 + ngc 4900 & sb & sb & sb & 0.372 & 0.167 & 5 & 19 & 4 & 2 & sb + ngc 4902 & sb & sb & sb & 0.272 & 0.060 & 15 & 67 & 3 & 1 & sb + ngc 4930 & sb & sb & sb & 0.210 & 0.110 & 31 & 109 & 2 & 1 & sa@xmath33 + ngc 4939 & sa & sab & sab & 0.119 & 0.084 & 11 & 97 & 1 & 1 & sab + ngc 4995 & sab & sab & sb & 0.203 & 0.207 & 11 & 19 & 2 & 2 & sa@xmath33 + ngc 5054 & sa & sa & sab & 0.065 & 0.088 & 13 & 69 & 1 & 1 & s@xmath32b + ngc 5085 & sa & sa & sab & 0.155 & 0.109 & 19 & 43 & 2 & 1 & sab + ngc 5101 & sb & sb & sb & 0.186 & 0.033 & 39 & 109 & 2 & 0 & sab + ngc 5121 & sa & sa & sa & 0.024 & 0.030 & 25 & 57 & 0 & 0 & sa + ngc 5248 & sab & sa & sa & 0.061 & 0.270 & 7 & 51 & 1 & 3 & s@xmath32b + ngc 5247 & sa & sa & sa & 0.020 & 0.327 & 3 & 65 & 0 & 3 & sa + ngc 5334 & sb & sb & sb & 0.322 & 0.145 & 5 & 11 & 3 & 1 & sb + ngc 5427 & sa & sa & sa & 0.083 & 0.235 & 7 & 33 & 1 & 2 & s@xmath32b + ngc 5483 & sa & sab & sb & 0.174 & 0.109 & 7 & 19 & 2 & 1 & sab + ngc 5643 & sab & sab & sb & 0.321 & 0.236 & 27 & 45 & 3 & 2 & sb + ngc 5676 & sa & sa & sab & 0.087 & 0.080 & 11 & 23 & 1 & 1 & s@xmath32b + ngc 5701 & sb & sb & sb & 0.139 & 0.053 & 27 & 105 & 1 & 1 & sab + ngc 5713 & sab & sab & sb & 0.335 & 0.111 & 7 & 15 & 3 & 1 & sb + ngc 5850 & sb & sb & sb & 0.311 & 0.053 & 39 & 65 & 3 & 1 & sb + ngc 5921 & sb & sb & sb & 0.255 & 0.349 & 21 & 37 & 3 & 3 & sb + ngc 5962 & sa & sab & sb & 0.141 & 0.055 & 9 & 15 & 1 & 1 & sab + ngc 6215 & sa & sa & sab & 0.079 & 0.230 & 3 & 24 & 1 & 2 & s@xmath32b + ngc 6221 & sb & sab & sb & 0.430 & 0.207 & 25 & 43 & 4 & 2 & sb + ngc 6300 & sb & sab & sb & 0.222 & 0.175 & 29 & 63 & 2 & 2 & sa@xmath33 + ngc 6384 & sab & sb & sb & 0.135 & 0.050 & 11 & 35 & 1 & 1 & sab + ngc 6753 & sa & sa & sa & 0.029 & 0.032 & 5 & 15 & 0 & 0 & sa + ngc 6782 & sab & sab & sb & 0.163 & 0.030 & 21 & 44 & 2 & 0 & sab + ngc 6902 & sa & sa & sb & 0.034 & 0.080 & 11 & 30 & 0 & 1 & sa + ngc 6907 & sb & sb & sb & 0.071 & 0.329 & 3 & 25 & 1 & 3 & s@xmath32b + ngc 7083 & sa & sa & sa & 0.033 & 0.071 & 5 & 23 & 0 & 1 & sa + ngc 7217 & sa & sa & sa & 0.033 & 0.036 & 9 & 109 & 0 & 0 & sa + ngc 7205 & sa & sa & sab & 0.048 & 0.061 & 7 & 55 & 0 & 1 & sa + ngc 7213 & sa & sa & sa & 0.004 & 0.024 & 11 & 93 & 0 & 0 & sa + ngc 7412 & sb & sab & sab & 0.060 & 0.434 & 11 & 45 & 1 & 4 & s@xmath32b + ngc 7418 & sab & sab & sb & 0.158 & 0.153 & 11 & 35 & 2 & 2 & sab + ngc 7479 & sb & sb & sb & 0.702 & 0.260 & 27 & 41 & 7 & 3 & sb + ngc 7552 & sb & sb & sb & 0.393 & 0.055 & 39 & 65 & 4 & 1 & sb + ngc 7713 & sb & sa & sa & 0.040 & 0.097 & 5 & 23 & 0 & 1 & sa + ngc 7723 & sb & sb & sb & 0.319 & 0.120 & 11 & 22 & 3 & 1 & sb + ngc 7727 & sab & sab & sa & 0.087 & 0.145 & 7 & 99 & 1 & 1 & s@xmath32b + ngc 7741 & sb & sb & sb & 0.736 & 0.324 & 11 & 27 & 7 & 3 & sb + ic 4444 & sab & sa & sb & 0.254 & 0.140 & 5 & 16 & 3 & 1 & sb + ic 5325 & sab & sa & sab & 0.030 & 0.213 & 5 & 11 & 0 & 2 & sa + eso 138- 10 & sa & sa & sa & 0.038 & 0.134 & 7 & 67 & 0 & 1 & sa + lccccccr + sa & 0.069 & 0.043 & 0.006 & 0.101 & 0.069 & 0.010 & 48 + sab & 0.177 & 0.114 & 0.017 & 0.147 & 0.118 & 0.018 & 45 + sb & 0.294 & 0.166 & 0.023 & 0.171 & 0.100 & 0.014 & 53 + + sa & 0.062 & 0.041 & 0.008 & 0.082 & 0.072 & 0.014 & 28 + sab & 0.143 & 0.100 & 0.020 & 0.134 & 0.127 & 0.025 & 26 + sb & 0.247 & 0.131 & 0.022 & 0.143 & 0.101 & 0.017 & 36 + + sa & 0.080 & 0.044 & 0.010 & 0.127 & 0.054 & 0.012 & 20 + sab & 0.225 & 0.117 & 0.027 & 0.165 & 0.105 & 0.024 & 19 + sb & 0.395 & 0.191 & 0.046 & 0.228 & 0.069 & 0.017 & 17 + + r & 0.047 & 0.037 & 0.012 & 0.066 & 0.058 & 0.019 & 9 + rs & 0.077 & 0.025 & 0.008 & 0.111 & 0.034 & 0.011 & 9 + s & 0.074 & 0.047 & 0.009 & 0.109 & 0.077 & 0.014 & 30 + + r & 0.199 & 0.094 & 0.033 & 0.092 & 0.056 & 0.020 & 8 + rs & 0.167 & 0.119 & 0.022 & 0.164 & 0.105 & 0.020 & 29 + s & 0.193 & 0.124 & 0.044 & 0.141 & 0.189 & 0.067 & 8 + + r & 0.201 & 0.107 & 0.029 & 0.135 & 0.081 & 0.022 & 14 + rs & 0.327 & 0.139 & 0.029 & 0.177 & 0.093 & 0.019 & 23 + s & 0.329 & 0.215 & 0.054 & 0.193 & 0.119 & 0.030 & 16 + lrrrrr + rc3 sa & 21 & 18 & 9 & 0 & 0 + rc3 sab & 5 & 9 & 16 & 3 & 12 + rc3 sb & 2 & 6 & 9 & 6 & 31 + + osu @xmath30 sa & 23 & 23 & 8 & 0 & 2 + osu @xmath30 sab & 3 & 6 & 16 & 5 & 9 + osu @xmath30 sb & 1 & 4 & 10 & 4 & 32 + + osu @xmath7 sa & 20 & 12 & 3 & 0 & 1 + osu @xmath7 sab & 5 & 12 & 4 & 0 & 1 + osu @xmath7 sb & 2 & 9 & 27 & 9 & 41 + + without fourier bar & 26 & 23 & 6 & 1 & 2 + with fourier bar & 2 & 10 & 28 & 8 & 41 +

motivated by the work of yang et al . , _ mod . phys . lett . a _ , * 26 * , 191 ( 2011 ) , we report a study on the new holographic dark energy ( nhde ) model with energy density given by in the framework of chameleon brans - dicke cosmology . we have studied a correspondence between the quintessence , the dbi - essence and the tachyon scalar field models with the nhde model in the framework of chameleon brans - dicke cosmology . deriving an expression of the hubble parameter and , accordingly , in the context of chameleon brans - dicke chameleon cosmology , we have reconstructed the potentials and dynamics for these scalar field models . furthermore , we have examined the stability for the obtained solutions of the crossing of the phantom divide under a quantum correction of massless conformally - invariant fields and we have seen that quantum correction could be small when the phantom crossing occurs and the obtained solutions of the phantom crossing could be stable under the quantum correction . it has also been noted that the potential increases as the matter - chameleon coupling gets stronger with the evolution of the universe .

1401.8208

approaches to account for the late time cosmic acceleration , which is suggested by the two independent observational signals on distant type ia supernovae ( sneia ) @xcite , the cosmic microwave background ( cmb ) temperature anisotropies measured by the wmap and planck satellites @xcite and baryon acoustic oscillations ( bao ) @xcite fall into two representative categories : in the first , the concept of dark energy " is introduced in the right - hand side of the einstein equation in the framework of general relativity ( for good reviews see @xcite ) while in the second one the left - hand side of the einstein equation is modified , leading to a modified gravitational theory ( which is well reviewed in @xcite ) . in a recent review , bamba et al . @xcite demonstrated that both dark energy models and modified gravity theories seem to be in agreement with data and hence , unless higher precision probes of the expansion rate and the growth of structure will be available , these two rival approaches could not be discriminated . physical origin of dark energy ( de ) is one of the largest mysteries not only in cosmology but also in fundamental physics @xcite . the cosmological constant @xmath3 represents the earliest and the simplest theoretical candidate proposed in order to explain the observational evidence of accelerated expansion . some tentative deviations from @xmath3cdm model may eventually rule out an exact cosmological constant @xcite . a considerable number of models for de have been proposed till date to explain the late - time cosmic acceleration without the cosmological constant . such models include a canonical scalar field , the so - called quintessence , a non - canonical scalar field such as phantom , tachyon scalar field motivated by string theories and a fluid with a special equation of state ( eos ) called as chaplygin gas . other well studied candidates for de are the k - essence , the quintom and the agegraphic dark energy ( ade ) models . studies on the models previously mentioned include @xcite . there also exists a proposal known as holographic dark energy ( hde ) proposed by li @xcite , following the idea that the short distance cut - off is related to the infrared cut - off and it was assumed in @xcite that the infrared cut - off relevant to the dark energy is the size of the event horizon . some notable works on hde include @xcite . furthermore , there exist plethora of literatures on hde in theoretical aspects as well as observational constraints e.g. @xcite . the equation of state ( eos ) parameter , defined as @xmath4 ( where @xmath5 and @xmath6 denote the pressure and density of de , respectively ) , is one of the most important quantity used to describe the features of de models . if we restrict ourselves in four - dimensional einstein s gravity , almost all de models can be classified according to the behavior of eos parameter as follow @xcite : ( i ) cosmological constant : @xmath7 ; ( ii ) quintessence : @xmath8 ; ( iii ) phantom : @xmath9 and ( iv ) quintom : its eos is able to evolve across the cosmological constant boundary . scalar field models of dark energy are among the most promising and best elaborated ones to match observations of the accelerated expansion of the universe . the phantom - like behavior of @xmath10 may appear from brans - dicke ( bd ) scalar tensor gravity , from non - standard ( negative ) potentials , from the non - minimal coupling of scalar lagrangian with gravity or even usual matter may appear in phantomlike form @xcite . studies devoted to phantom cosmology include @xcite and to quintessence include @xcite . in this contribution , we are working on new holographic reconstruction of scalar field dark energy models in the framework of chameleon brans - dicke cosmology . in the context of cosmological reconstruction problem , some notable contributions are @xcite . the current work is primarily motivated by @xcite and also got inspiration from @xcite . it is already stated that hde model is based on the holographic principle , according to which , the number of degrees of freedom of a physical system scales with the area of its boundary @xcite . although hde gives the observational value of de in the universe and can drive the universe to an accelerated expansion phase , an obvious drawback concerning causality appears in this proposal @xcite . in view of this limitation granda and oliveros @xcite proposed a new infrared cut - off for hde , which is proportional to the square of the hubble parameter squared @xmath11 and to the time derivative of the hubble parameter @xmath12 , and dubbed this model as new hde ( nhde ) model . the energy density of the nhde model is given by @xcite : @xmath13 where @xmath14 and @xmath15 are two positive constants . this model could avoid the problem of causality and could solve the coincidence problem @xcite . in a more recent work , @xcite confirmed through action principle that the nhde model overcomes the causality and circular problems in the original hde model and putting constraints on the model from the union@xmath16+bao+cmb+h@xmath17 data @xcite got the goodness of - fit @xmath18 , which they found comparable with the results of the original hde model @xmath19 and the concordant @xmath3cdm model @xmath20 and this lead them to conclude that nhde fit well to the data . viewing scalar field dark energy models as an effective description of the underlying theory of dark energy , and considering the holographic vacuum energy scenario as pointing in the same direction , granda and oliveros @xcite demonstrated how the scalar field models can be used to describe the holographic energy density as effective theories , for this purpose they studied correspondence between the quintessence , tachyon , k - essence and dilaton energy densities with this nhde in the flat frw universe . connecting these scalar field models with nhde , they found the explicit forms of the scalar fields and of the potentials in this reconstruction approach . karami and fehri @xcite extended the work of @xcite to non - flat frw universe i.e. they studied a correspondence between nhde and quintessence , tachyon , k - essence and dilaton scalar field models in the presence of a spatial curvature and reconstructed scalar field and potential . sharif and jawad @xcite established a correspondence between the nhde model and the quintessence , the tachyon , the k - essence and the dilaton scalar field models . in a recent work , jawad et al . @xcite explored holographic reconstruction of modified @xmath21 horava - lifshitz gravity via power - law scale factor and discussed eos parameter as well as stability of the reconstructed model and remarked about quintessence era in near future with instability . dependency of the evolution of equation of state , deceleration parameter and cosmological evolution of hubble parameter on the parameters of nhde model were studied in malekjiani et al @xcite . considering interaction between dark matter and nhde , debnath and chattopadhyay @xcite investigated the statefinder and the @xmath22 diagnostics and also checked the validity of the gsl of thermodynamics with apparent horizon as the enveloping horizon of the universe . before demonstrating the current contribution in view of the stated works , let us have a brief overview of the theories of modified gravity as the current contribution is going to explore a cosmological reconstruction in the framework of a modified gravity theory . modified gravity has become an essential part of theoretical cosmology nowadays @xcite . it is proposed as generalization of general relativity with the purpose to understand the qualitative change of gravitational interaction in the very early and/or very late universe . in particular , it is accepted nowadays that modified gravity may not only describe the early - time inflation and late - time acceleration but also may propose the unified consistent description of the universe evolution epochs sequence : inflation , radiation / matter dominance and dark energy @xcite . nojiri and odintsov @xcite summarized the usefulness of modified gravity as follow : 1 . it provides natural gravitational alternative for dark energy , 2 . it presents very natural unification of the early - time inflation and late - time acceleration thanks to different role of gravitational terms relevant at small and at large curvature , 3 . it may serve as the basis for unified explanation of dark energy and dark matter . modified gravity models include @xmath21 gravity ( where @xmath23 is the ricci scalar curvature ) @xcite , @xmath24 gravity ( where @xmath25 represents the torsion scalar ) @xcite , scalar - tensor theories @xcite , braneworld models @xcite , galileon gravity @xcite , gauss - bonnet gravity @xcite and so on . bamba et al . @xcite investigated the future evolution of the dark energy universe in modified gravities including @xmath21 gravity , string - inspired scalar - gauss - bonnet and modified gauss - bonnet ones and an ideal fluid with inhomogeneous equation of state and constructed several examples of the modified gravity that produces accelerating cosmologies ending at the finite - time future singularity by applying the reconstruction program . cosmological evolution of the equation of state for dark energy @xmath10 in the exponential and logarithmic as well as their combination in the framework of @xmath24 theories was studied in bamba et al . @xcite . a reconstruction scheme for modified gravity realizing a crossing of the phantom divide was proposed in bamba et al . appearance of finite - time future singularities in @xmath24 gravity was demonstrated in bamba et el . @xcite . in another work , bamba et al @xcite explored the cosmological evolution in a modified gravity @xmath26 and demonstrated that the late - time cosmic acceleration following the matter - dominated stage can be realized n that model . bamba @xcite showed that the crossing of the phantom divide can be realized in the combined @xmath24 theory constructed with the exponential and logarithmic terms . recently , various scalar tensor theories have been considered extensively and one important example of the scalar tensor theories is the brans - dicke ( bd ) theory of gravity which was introduced by brans and dicke @xcite to incorporate the mach principle in the einstein s theory of gravity . bd theory is proposed as the natural extension of kaluza - klein idea of unification @xcite . bd parameter has some interesting properties as a candidate of de when it has been studied in the non minimally coupled regime @xcite . the popularity of bd modified gravity @xcite lies in the fact that it naturally arises as the low energy limit of many other quantum gravity theories , like the kaluza - klein one or the superstring theory . in this paper , we decided to consider the new holographic dark energy ( nhde ) model in the framework of the chameleon brans - dicke modified gravity theory , in which there is a non - minimal coupling between the matter field and the scalar field @xmath27 which is usually known in literature with the name of chameleon field @xcite since its main physical properties strongly depend on the environment . waterhouse @xcite derived that the deviations from newtonian gravity due to the chameleon field of the earth are suppressed by nine orders of magnitude by the thin - shell effect . in a recent work , chattopadhyay @xcite studied the matter - chameleon coupling considering extended holographic ricci dark energy model in chameleon brans - dicke cosmology . instead , bisbar @xcite considered a generalized brans - dicke model in which the scalar field has a potential function and it can couple non - minimally with the matter sector . late - time dynamics of a chameleonic generalized brans - dicke cosmology with a power law chameleonic function has been well studied in a recent paper of el - nabulsi @xcite . other important works on chameleon gravity have been done in @xcite . the present contribution is organized as follow : in section ii , we study the main cosmological properties of the nhde model in the framework of brans - dicke chameleon cosmology . in section iii , we make a correspondence between the reconstructed nhde model and three differen scalar field models , i.e. the quintessence , the dbi - essence and the tahcyon scalar field models . finally , in section iv , we write the conclusions of this paper . we begin this section with the description of the main cosmological properties of the chameleon brans - dicke ( bd ) theory . + according to bd theory , the scalar field is coupled non - minimally to the matter field via the action @xmath28 given by @xcite : @xmath29 where @xmath23 indicates the ricci scalar curvature , @xmath27 represents the brans - dicke scalar field with an associate potential @xmath30 , @xmath31 indicates the dimensionless brans - dicke parameter , @xmath32 represents the metric tensor with determinant given by @xmath33 , @xmath34 represents the matter lagrangian and , finally , @xmath35 represents an arbitrary function of the scalar field @xmath27 . the last term in the action @xmath28 given in eq . ( [ action ] ) represents the term which gives us information about the interaction between the matter lagrangian and the arbitrary function @xmath36 . we must also emphasize here that , in the limiting case corresponding to @xmath37 , we obtain the standard bd cosmology . + varying the action @xmath28 given in eq . ( [ action ] ) with respect to the metric tensor @xmath38 and @xmath27 , we obtain the following field equations : @xmath39 where @xmath40 is the einstein tensor , @xmath41 ( with @xmath42 representing the covariant derivative ) @xmath43 and the prime denotes a differentiation with respect to @xmath27 . in eq . ( [ field1 ] ) , we have that : @xmath44 and @xmath45 because of the explicit coupling between matter system and @xmath27 , the stress tensor @xmath46 is not divergence free . we now apply the above framework to a homogeneous and isotropic universe described by the friedman - robertson - walker metric given by : @xmath47 the universe is open , closed or flat according as @xmath48 . moreover , we have that @xmath49 represents the scale factor ( which gives information about the expansion of the universe ) , @xmath50 gives the radial component of the metric , @xmath51 indicates the cosmic time and @xmath52 denotes the solid angle element ( squared ) . @xmath53 and @xmath54 are the usual azimuthal and polar angles , with @xmath55 and @xmath56 . the coordinates @xmath57 are known as comoving coordinates . + in a spatially flat universe ( i.e. for @xmath58 , eqs . ( [ field1 ] ) and ( [ field12 ] ) yield @xcite : @xmath59f\right\ } \nonumber \\ & & -\omega\frac{\dot{\phi}^2}{\phi^2}+3h \frac{\dot{\phi}}{\phi}+\frac{1}{2\omega+3}\left[3v'+(2\omega-3)\frac{v}{\phi}\right ] , \\ \label{fried3 } ( 2\omega+3)(\ddot{\phi}+3h\dot{\phi})&-&2(2v-\phi v')=\rho\left[(1 - 3\gamma)f+2\gamma\phi f'\right].\end{aligned}\ ] ] we must underline here that a dot indicates a time derivative while the prime indicates a derivative with respect to @xmath27 . + in this paper , our purpose is to generalize the work of karami and fehri @xcite to the nhde model with energy density @xmath2 given by @xcite : @xmath60 where @xmath61 and @xmath62 are two constant parameters and the overdot represents the first time derivative . we consider the following ansatz for @xmath27 , @xmath63 and @xmath64 @xcite : @xmath65 where @xmath66 and @xmath67 are constant parameters and @xmath68 and @xmath69 are positive quantities representing the present day values of the corresponding quantities . one attribute of taking this kind of ansatz is scale invariance of power laws . given a relation in power law form , scaling the argument by a constant factor @xmath70 causes only a proportionate scaling of the function itself . issues related to power - law choice of potential are discussed in @xcite . + using the aforesaid ansatz in eq . ( [ fried1 ] ) we get the following differential equation for @xmath11 : @xmath71 h^2=-\frac{8 \left(e^{\alpha x } \phi_0\right)^{-2+\beta -\nu } v_0 \omega } { 3 f_0 \nu } .\ ] ] solving eq . ( [ diffeqn ] ) , we get the expression of reconstructed @xmath72 as a function of the scale factor @xmath73 as follow : @xmath74 , \label{hsqr } \end{array}\ ] ] where : @xmath75 moreover , we have that in eq . ( [ hsqr ] ) @xmath76 represents the gamma function . + subsequently , using the relation @xmath77 , we obtain the following relation for @xmath78 : @xmath79\right ) \label{hdot } \end{array}\ ] ] we can use eq . ( [ hsqr ] ) and ( [ hdot ] ) in eq . ( [ nhde ] ) to reconstruct the density of the nhde in chameleon bd cosmology , obtaining : @xmath80\right\}\right ] \end{array}\ ] ] since we are assuming that the universe is filled with nhde , the conservation equation for @xmath2 is given by the following relation : @xmath81 where @xmath82 represents the eos parameter of de . + from eqs . ( [ density ] ) and ( [ conserve ] ) , we can derive the reconstructed equation of state ( eos ) parameter @xmath82 as a function of the scale factor @xmath73 as follow : @xmath83\}\times\\ \left\{3 \left(s \left(a^s \eta_5\right)^{\eta_2 } \nu -e^{a^s \eta_5 } \left(2 \mu -\nu \left(-s \eta_4+a^s s \eta_6+\alpha ( 2-\beta + \nu ) \right)\right ) \gamma\left[\eta_2,a^s \eta_5\right]\right)\right\}^{-1 } \end{array}\ ] ] ( null energy condition test ) . ( [ density ] ) and ( [ eos ] ) . ] ( null energy condition test ) . ( [ density ] ) and ( [ eos ] ) . ] ( null energy condition test ) . ( [ density ] ) and ( [ eos ] ) . ] ( null energy condition test ) . ( [ density ] ) and ( [ eos ] ) . ] we now plot the reconstructed cosmological parameters against the redshift @xmath84 . in all figures , we have that red , green and blue lines correspond to @xmath85 , @xmath86 and @xmath87 , respectively . for all other figures , the other parameters present in the equations we derived are set as @xmath88 . the choice of the value of the bd parameter @xmath31 is based on ref . observational results coming from sneia data suggest a range of possible values for the eos parameter of @xmath89 @xcite . using a set of variations in the values of @xmath61 and @xmath62 in eq . ( [ eos ] ) we find that the results are in good agreement with @xcite . in table i , we are showing a set of values of the reconstructed eos based on the chosen values of the parameters . it is apparent from this table that the eos parameter @xmath89 . .values of reconstructed @xmath82 ( eq . ( [ eos ] ) ) for different choices of parameter values . [ cols="<,<,<,<",options="header " , ] in fig . [ fig1 ] , we observe the evolution of the reconstructed hubble parameter @xmath90 with redshift @xmath91 . we observe a decaying pattern of @xmath90 with evolution of the universe . this in consistent with the accelerated expansion of the universe . in fig . [ fig2 ] , the reconstructed nhde density is plotted and indicates its dominance with the evolution of the universe . this is consistent with the current dark - energy dominated era . + we now want to have a deeper look into the eq . ( [ eos ] ) derived through the reconstructed @xmath90 . in eq . ( [ eos ] ) , the expression @xmath92\right)~\neq0 $ ] . considering the forms of @xmath93 it can be verified that the above expression is surely positive if @xmath94 or @xmath95 . since we have taken @xmath96 , we have @xmath97 . if @xmath98 , then the scalar field decays with the evolution of the universe . since the behavior of the remaining part of the expression is too complicated to make any inference on its quintessence " or phantom "- like behavior , we only depend on the _ strong _ and _ null _ energy conditions to put some light into the behavior of the equation of state parameter . null energy condition is satisfied if @xmath99 while the strong energy condition condition is satisfied if @xmath100 . figs . [ fig3 ] and [ fig4 ] indicate that both of the energy conditions are violated . this indicates phantom "- like behavior of the equation of state parameter . in this section , we reconstructed the nhde in the framework of chameleon bd cosmology . in the following section , our objective is to study the correspondence between this reconstructed nhde model and the scalar field models , namely ( i ) quintessence dark energy , ( ii ) dbi - essence dark energy and ( iii ) tachyon dark energy . following @xcite we examine the stability for the obtained solutions of the crossing of the phantom divide under a quantum correction of massless conformally - invariant fields . quantum effects produce the conformal anomaly @xcite : @xmath101 where @xmath102 in the frw universe , we have that : @xmath103 for @xmath104 real scalar , @xmath105 dirac spinor , @xmath106 vector fields , @xmath107 gravitons and @xmath108 higher derivative conformal scalars , we have the following expressions for @xmath109 and @xmath110 : @xmath111 with @xmath112 which can be arbitrary . if we assume that @xmath113 can be given by the effective energy density @xmath114 and pressure @xmath115 from the conformal anomaly as : @xmath116 the following expressions for @xmath114 and @xmath115 can be found : @xmath117 at phantom crossing , we must have @xmath118 . if we assume that the magnitude of the hubble rate @xmath1 could be the order of the present hubble constant @xmath119 , then for phantom crossing we have : @xmath120 where @xmath121 $ ] . the results obtained in eqs . ( [ ta ] ) and ( [ ta2 ] ) tell that we may assume @xmath122 . then , we find @xcite : @xmath123 where @xmath124 represents a dimensionless constant of the order of @xmath125 . thus , for the reconstructed model , we obtain : @xmath126 hence we can conclude that @xmath127 new holographic reconstruction of scalar field models in bd cosmology --------------------------------------------------------------------- sahni and starobinsky @xcite discussed various aspects of reconstructing the expansion history of the universe and to probe the nature of dark energy . below , we will study the correspondence between nhde model and the quintessence , the dbi - essence and the tachyon scalar field models in the framework of flat chameleon brans - dicke universe . we will also reconstruct the potentials and the dynamics for these scalar field models . we can give the related results of scalar fields and potentials for the nhde model in the flat chameleon brans - dicke universe . in order to establish this correspondence , we compare the energy density of the nhde model given in eq . ( [ density ] ) with the corresponding energy density of the scalar field model , and we also equate the eos for these scalar models with the eos for the nhde model given in eq . ( [ eos ] ) . we must also emphasize here that we indicate the scalar field with @xmath54 in order to differ it from the scalar field @xmath27 in brans - dicke theory . quintessence is a dynamical , evolving , spatially inhomogeneous component with negative pressure . unlike a cosmological constant , the quintessential pressure and energy density evolve with the time and the eos parameter may also do so . a common model of quintessence is the energy density associated with a scalar field @xmath128 slowly rolling down a potential @xmath129 . a detailed discussion on quintessence dark energy is available in the review @xcite . the energy density @xmath130 and pressure @xmath131 of the quintessence scalar field @xmath54 are given , respectively , by @xcite : @xmath132 moreover , the eos parameter can be written as follow : @xmath133 as we are reconstructing the quintessence model based on nhde in the framework of chameleon bd cosmology , we shall consider @xmath134 and @xmath135 . hence , we have : @xmath136 where @xmath2 and @xmath82 are given in eqs . ( [ density ] ) and ( [ eos ] ) , respectively . based on the reconstructed hubble parameter , we express @xmath137 and @xmath138 as functions of the scale factor @xmath73 as follow : @xmath139\right ) , % \end{array}\end{aligned}\ ] ] @xmath140\right ) \left(1-\left(s \left(a^s \eta_5\right)^{\eta_2 } \left(-2 \mu + \nu \left(-3-s \left(\eta_2 + 2 \eta_4-a^s ( \eta_5 + 2 \eta_6)\right)+2 \alpha ( 1-\beta + \nu ) \right)\right)+\right.\right . \nonumber\\ & & \left.\left.e^{a^s \eta_5 } \left(a^{2 s } s^2 \eta_6 ^ 2 \nu + ( 3+s \eta_4+\alpha ( \beta -\nu ) ) ( 2 \mu + ( s \eta_4+\alpha ( -2+\beta -\nu ) ) \nu ) -\right.\right.\right . \nonumber\\ & & \left.\left.\left.a^s s \eta_6 ( 2 \mu + ( 3+s+2 s \eta_4 + 2 \alpha ( -1+\beta -\nu ) ) \nu ) \right ) \gamma\left[\eta_2,a^s \eta_5\right]\right)\right.\nonumber \\ & & \left./ \left(3 \left(s \left(a^s \eta_5\right)^{\eta_2 } \nu -e^{a^s \eta_5 } \left(2 \mu -\nu \left(-s \eta_4+a^s s \eta_6+\alpha ( 2-\beta + \nu ) \right)\right ) \gamma\left[\eta_2,a^s \eta_5\right]\right)\right)\right ) . % \end{array}\end{aligned}\ ] ] ( eq . ( [ qpotential ] ) ) with reconstructed scalar field @xmath54 ( eq . ( [ qfield ] ) ) of the quintessence dark energy model . ] ( [ qpotential ] ) ) with reconstructed scalar field @xmath54 ( eq . ( [ qfield ] ) ) of the quintessence dark energy model . ] during the last few years , there have been many works aiming at connecting string theory with inflation which is also a phase of accelerated expansion . martin and yamaguchi @xcite introduced a scalar - field model of dark energy with a non - standard dirac - born - infeld ( dbi ) kinetic term . this model is dubbed as dbi - essence dark energy " and the energy density @xmath141 and the pressure @xmath142 of the dbi - essence model are given , respectively , by @xcite : @xmath143 where : @xmath144 the eos parameter for the dbi - essence scalar field model can be written as follow : @xmath145 in the present work , we shall assume that @xmath146 , where @xmath147 . since we are considering a correspondence between dbi - essence dark energy and the reconstructed nhde model , we consider @xmath148 and @xmath149 . the based on eq . ( [ dbi ] ) we get the reconstructed scalar field @xmath54 as function of the scale factor @xmath73 as follow : @xmath150\right ) , % \end{array}\end{aligned}\ ] ] @xmath151\right)+\right.\nonumber \\ & & \left.\left(-1+\sqrt{\frac{n}{n-1}}\right ) \sqrt{\frac{n-1}{n } } n \left(s \left(a^s \eta_5\right)^{\eta_2 } \left(-2 \mu + \nu \left(-s \left(\eta_2 + 2 \eta_4-a^s ( \eta_5 + 2 \eta_6)\right)+\right.\right.\right.\right . \nonumber\\ & & \left.\left.\left.\left.\alpha ( 2 - 2 \beta + 2 \nu ) \right)\right)+e^{a^s \eta_5 } \left(\alpha ^2 \nu \left(\beta ^2 - 2 \beta ( 1+\nu ) + \nu ( 2+\nu ) \right)+s \left(s \eta_4 ^ 2 \nu + 2 \eta_4 \left(\mu -a^s s \eta_6 \nu \right)+\right.\right.\right.\right . \nonumber \\ & & \left.\left.\left.\left.a^s \eta_6 \left(-2 \mu + s \left(-1+a^s \eta_6\right ) \nu \right)\right)+2 \alpha \left(\beta \left(\mu + s \left(\eta_4-a^s \eta_6\right ) \nu \right ) \right . \nonumber \\ & & \left . \left . -\nu \left(\mu + s \left(\eta_4-a^s \eta_6\right ) ( 1+\nu ) \right)\right)\right ) \gamma\left[\eta_2,a^s \eta_5\right]\right)\right ) . % \end{array}\end{aligned}\ ] ] ( eq . ( [ dbiv ] ) ) with reconstructed scalar field @xmath54 ( eq . ( [ dbiphi ] ) ) of the dbi - essence dark energy model.we have chosen @xmath152 . ] ( [ dbiv ] ) ) with reconstructed scalar field @xmath54 ( eq . ( [ dbiphi ] ) ) of the dbi - essence dark energy model.we have chosen @xmath152 . ] tachyonic condensate in a class of string theories can be described by an effective scalar field with a lagrangian of the form @xmath153 . since this lagrangian has also a potential function @xmath154 , any form of cosmological evolution ( that is , any @xmath155 ) can be obtained with the tachyonic field as the source by choosing @xmath154 suitably @xcite . cosmological effects of homogeneous tachyon matter coexisting with non - relativistic matter and radiation have been studied by @xcite . the energy density @xmath156 and the pressure @xmath157 of the tachyon scalar field model are given , respectively , by @xcite : @xmath158 while the eos parameter can be written as follow : @xmath159 for the correspondence under consideration , we have @xmath160 and @xmath161 . using the same procedure used before , we reconstruct the scalar field @xmath54 and the potential @xmath138 as follow : @xmath162\right)\times \nonumber\\ & & \left(3 \left(s \left(a^s \eta_5\right)^{\eta_2 } \nu -e^{a^s \eta_5 } \left(2 \mu -\nu \left(-s \eta_4+a^s s \eta_6+\alpha ( 2-\beta + \nu ) \right)\right ) \gamma\left[\eta_2,a^s \eta_5\right]\right)\right)^{-1 } , % \end{array}\end{aligned}\ ] ] @xmath163\right ) \left(s \left(a^s \eta_5\right)^{\eta_2 } \nu -e^{a^s \eta_5 } \left(2 \mu -\nu \left(-s \eta_4+a^s s \eta_6+\right.\right.\right.\right.\nonumber\\ & & \left.\left.\left.\left.\alpha ( 2-\beta + \nu ) \right)\right ) \gamma\left[\eta_2,a^s \eta_5\right]\right)\right]^{1/2 } % \end{array}\end{aligned}\ ] ] ( eq . ( [ tachv ] ) ) with reconstructed scalar field @xmath54 ( eq . ( [ tachphi ] ) ) of the tachyon dark energy model . ] ( [ tachv ] ) ) with reconstructed scalar field @xmath54 ( eq . ( [ tachphi ] ) ) of the tachyon dark energy model . ] in this paper , we studied the main cosmological properties of the new holographic dark energy ( hnde ) model in the framework of brans - dicke chameleon cosmology . we considered a particular ansatz for the parameters @xmath27 , @xmath63 and @xmath64 in which their expressions are given in the power law form . we decided to consider different aspects to study . first of all , we reconstructed the expression of the hubble parameter @xmath1 and , accordingly , the expression of the density @xmath2 of the nhde in the context of chameleon brans - dicke chameleon cosmology . we also tested the weak energy condition ( wec ) and the strong energy condition ( sec ) for the reconstructed model we obtained . considering three cases , namely @xmath85 , @xmath86 and @xmath87 setting the other parameters as @xmath88 and bd parameter @xmath31 following @xcite we have computed the reconstructed eos parameter . observational results coming from sneia data suggest a limit of the eos parameter as @xmath89 @xcite . using a set values of @xmath61 and @xmath62 in eq . ( [ eos ] ) we found that the results are in good agreement with observations of @xcite . finally , we reconstructed three scalar field models of dark energy ( namely , the quintessence , the dbi - essence and the tachyon ones ) based on the nhde model in the framework of bd cosmology . for the three scalar field models we considered , we have reconstructed the corresponding potentials and scalar fields . to further elucidate our reconstructions , we have plotted the reconstructed potential @xmath138 against @xmath84 and made parametric plots between @xmath54 and @xmath164 in figs . [ fig6 ] , [ fig7 ] , [ fig9 ] , [ fig10 ] , [ fig11 ] and [ fig12 ] . it is apparent from the plots that the potential @xmath138 is increasing up to redshifts of the order of @xmath165 , afterwards it starts to decay . in the plots of @xmath166 , it appears that the potential has a decreasing behavior with the scalar field @xmath54 . with @xmath167 for tachyon dark energy model . ] with @xmath167 for tachyon dark energy model . ] with @xmath167 for tachyon dark energy model . ] in order to have a look into the behavior of the reconstructed potential against coupling function @xmath64 we discuss the figs . [ quint ] , [ dbi ] and [ tachyon ] , where we observe that the reconstructed potentials are increasing with @xmath64 . this indicates that the potential increasing as the matter - chameleon coupling is getting stronger . it is further noted that the rate of increase in the potential is much higher in the case of tachyon than the cases of quintessence and dbi . in the present work we have used a reconstruction scheme for new holographic dark energy model with energy density given by @xmath168 in the framework of brans - dicke cosmology taking the ansatz @xmath169 . outcomes of the study are : * considering @xmath170 in the first modified field equation of bd theory leads to a linear differential equation which could be solved analytically to have a solution for the reconstructed hubble parameter in terms of scale factor @xmath73 ; when plotted against redshift @xmath91 , it exhibited decaying pattern with the evolution of the universe ( i.e. , decreases in @xmath84 ) and this is consistent with the accelerated expansion of the universe . * the nhde energy density , as reconstructed through hubble parameter , when plotted against @xmath84 , is found to increase with evolution of the universe and it is consistent with the evolution of the universe from matter to dark energy domination . * violation of strong energy condition , as expected in the framework of einstein gravity , has also been found for the reconstructed nhde model in the framework of bd gravity . * the reconstructed equation of state ( eos ) parameter has been found to exhibit phantom "- like behavior , i.e. @xmath171 . * considering three different combinations of the parameters @xmath61 and @xmath62 , namely @xmath85 , @xmath86 and @xmath87 and setting the other parameters as @xmath88 and the bd parameter @xmath31 following @xcite , we have computed the reconstructed eos parameter for the reconstructed nhde . observational results coming from sneia data suggest a limit of the eos parameter as @xmath89 @xcite . using a set of variations in the values of @xmath61 and @xmath62 in eq . ( [ eos ] ) we found that the results are in good agreement with observations of @xcite ( see table i ) . in the following phase of the study , we considered the correspondence between the reconstructed new holographic dark energy in the framework of bd gravity and some scalar field dark energy models in a manner under which the two scenarios can be simultaneously valid . this type of approach is available in cosmological literature ( e.g. , @xcite ) . we have constructed the potentials and the scalar fields of these models . we observed that @xmath172 for all of the reconstructed scalar field models upt o @xmath165 and , at very late stage , ( i.e. , @xmath173 ) , we have @xmath174 . moreover , @xmath175 for all of the models . in summary , by generalizing the previous works @xcite to the nhde model with @xmath0 in the framework of chameleon brans - dicke cosmology , we have obtained the evolution of eos . following @xcite we have considered @xmath63 , @xmath27 and @xmath64 in power - law form and accordingly reconstructed hubble parameter . this approach differs from @xcite in the sense that instead of considering brans - dicke cosmology , we have considered chameleon brans - dicke with coupling function @xmath64 . we have tested sec and wec conditions and interpreted evolution of eos from them . with some choice of the model parameters we computed eos and found the the computed values of eos are consistent with the observational results coming from sneia data that suggest a limit of the eos parameter as @xmath89 @xcite . subsequently we examined the stability for the obtained solutions of the crossing of the phantom divide under a quantum correction of massless conformally - invariant fields and we have seen that quantum correction could be small when the phantom crossing occurs and the obtained solutions of the phantom crossing could be stable under the quantum correction . in the subsequent phase , we have established a correspondence between the nhde model and the quintessence , the dbi - essence and the tachyon scalar field models in the framework of chameleon brans - dicke cosmology . we reconstruct the potentials and the dynamics for these three scalar field models we have considered . the reconstructed potentials are found to increase with evolution of the universe and in a very late stage they are observed to decay . it is also observed through @xmath176 plot that the potential is increasing with @xmath64 , which indicates that the potential increases as the matter - chameleon coupling gets stronger with evolution of the universe . sincere thanks are due to the anonymous reviewer for constructive suggestions . project grant of dst , govt . of india sr / ftp / ps-167/2011 is duly acknowledged by the first author . also , the first authors acknowledges the facilities provided by iucaa , pune , india , where a major portion of the work was carried out during a scientific visit in december , 2013-january , 2014 . a. g. riess et al . j. * 116 * , 1009 ( 1998 ) [ doi : 10.1086/300499 ] . s. perlmutter , astrophys . j. * 517 * , 565 ( 1999 ) [ doi : 10.1086/307221 ] . r. a. knop et al . , astrophys . j. * 598 * , 102 ( 2003 ) [ doi : 10.1086/378560 ] . d. n. spergel et al . , astrophys . * 148 * , 175 ( 2003 ) [ doi : 10.1086/377226 ] . e. komatsu et al . , astrophys . j. suppl . * 192 * , 18 ( 2011)[doi : 10.1088/0067 - 0049/192/2/18 ] . planck collaboration , ade , p. a. r. , aghanim , n. , et al . 2013 , arxiv:1303.5076 . d. j. eisenstein et al . [ sdss collaboration ] , astrophys . j. * 633 * , 560 ( 2005 ) [ doi : 10.1086/466512 ] . w. j. percival et al . , mon . not . * 401 * , 2148 ( 2010 ) [ doi : 10.1111/j.1365 - 2966.2009.15812.x ] . e. j. copeland , m. sami , s. tsujikawa , int . d , * 15 * , 1753 ( 2006 ) [ doi : 10.1142/s021827180600942x ] . k. bamba , s. capozziello , s. nojiri , s. d. odintsov , astrophys . space sci . , * 342 * , 155 ( 2012 ) [ doi : 10.1007/s10509 - 012 - 1181 - 8 ] . r. r. caldwell , m. kamionkowski , ann . nucl . part . sci . * 59 * , 397 ( 2009 ) [ doi : 10.1146/annurev - nucl-010709 - 151330 ] . s. nojiri , s. d. odintsov , phys . , * 505 * , 59 ( 2011 ) [ doi : 10.1016/j.physrep.2011.04.001 ] . t. clifton , p. g. ferreira , a. padilla , c. skordis , phys . * 513 * , 1 ( 2012 ) [ doi : 10.1016/j.physrep.2012.01.001 ] . s. capozziello , m. de laurentis , s. d. odintsov , eur . j. c * 72 * , 2068 ( 2012 ) [ doi : 10.1140/epjc / s10052 - 012 - 2068 - 0 ] . s. tsujikawa , lect . notes phys . * 800 * , 99 ( 2010 ) [ doi : 10.1007/978 - 3 - 642 - 10598 - 23 ] . v. sahni , a. a. starobinsky , int . j. mod d * 9 * , 373 ( 2000 ) [ doi : 10.1142/s0218271800000542 ] . v. sahni , a. a. starobinsky , int . j. mod d * 15 * , 2105 ( 2006 ) [ doi : 10.1142/s0218271806009704 ] . h. motohashi , a. a. starobinsky , j. yokoyama , prog . . phys . * 123 * , 887 ( 2010 ) [ doi : 10.1143/ptp.124.541 ] . p. j. e. peebles , b. ratra , rev . phys . * 75 * , 559 ( 2003 ) [ doi : 10.1103/revmodphys.75.559 ] . a. shafieloo , v. sahni , a. a. starobinsky , phys . d * 80 * , 101301 ( 2009 ) [ doi : 10.1103/physrevd.80.101301 ] . a. a. starobinsky , phys . b * 91 * , 99 ( 1980 ) [ doi : 10.1016/0370 - 2693(80)90670-x ] . s. nojiri , s. d. odintsov , o. g. gorbunova , j. phys . a : math * 39 * , 6627 ( 2006 ) [ doi : 10.1088/0305 - 4470/39/21/s62 ] . a. v. astashenok , s. nojiri , s. d. odintsov , r. j. scherrer , phys . b * 713 * , 145 ( 2012 ) [ doi : 10.1016/j.physletb.2012.06.017 ] . b. gumjudpai , t. naskar , m. sami , s. tsujikawa , j. cosmol . astropart . phys . * 06 * , 007 ( 2005 ) [ doi : 10.1088/1475 - 7516/2005/06/007 ] . e. elizalde , s. nojiri , s. d. odintsov , phys . d * 70 * , 043539 ( 2004 ) [ doi : 10.1103/physrevd.70.043539 ] . s. nojiri , s. d. odintsov , s. tsujikawa , phys . d * 71 * , 063004 ( 2005 ) [ doi : 10.1103/physrevd.71.063004 ] . h. zhang , z - h . zhu , phys . d * 73 * , 043518 ( 2006 ) [ doi : 10.1103/physrevd.73.043518 ] . k. bamba , j. matsumoto , s. nojiri , phys . d * 85 * , 084026 ( 2012 ) [ doi : 10.1103/physrevd.85.084026 ] . m. forte , phys . rev . d * 90 * , 027302 ( 2014 ) [ doi : 10.1103/physrevd.90.027302 ] . m. li , phys . b , * 603 * , 1 ( 2004 ) [ doi : 10.1016/j.physletb.2004.10.014 ] . e. elizalde , s. nojiri , s. d. odintsov , p. wang , phys . d * 71 * , 103504 ( 2005 ) [ doi : 10.1103/physrevd.71.103504 ] . s. nojiri , s. d. odintsov , gen . . grav . * 38 * , 1285 ( 2006)[doi : 10.1007/s10714 - 006 - 0301 - 6 ] . s. del campo , j. c. fabris , r. herrera , w. zimdahl , phys . d * 83 * , 123006 ( 2011 ) [ doi : 10.1103/physrevd.83.123006 ] . cui , j - f . zhang , eur . j. c * 74 * , 2849 ( 2014 ) [ doi : 10.1140/epjc / s10052 - 014 - 2849 - 8 ] . huang , y. gong , jcap * 0408 * , 006 ( 2004 ) [ doi : 10.1088/1475 - 7516/2004/08/006 ] . huang , m. li , jcap * 0503 * , 001 ( 2005 ) [ doi : 10.1088/1475 - 7516/2005/03/001 ] . x. zhang , f - q . wu , phys . d * 76 * , 023502 ( 2007 ) [ doi : 10.1103/physrevd.76.023502 ] . cai , e. n. saridakis , m. r. setare , j - q . xia , phys . 493 * , 1 ( 2010 ) [ doi : 10.1016/j.physrep.2010.04.001 ] . s. nojiri , s. d. odintsov , s. tsujikawa , phys . rev . d * 71 * , 063004 ( 2005 ) [ doi : 10.1103/physrevd.71.063004 ] . a. a. starobinsky , phys . lett . * 85 * , 2236 ( 2000)[doi : 10.1103/physrevlett.85.2236 ] . s. nojiri and s. d. odintsov , phys . b * 562 * , 147 ( 2003 ) [ doi : 10.1016/s0370 - 2693(03)00594-x ] s. nojiri and s. d. odintsov , phys . b * 571 * , 1 ( 2003 ) [ doi : 10.1016/j.physletb.2003.08.013 ] s. nojiri and s. d. odintsov , phys . b * 565 * , 1 ( 2003 ) [ doi : 10.1016/s0370 - 2693(03)00753 - 6 ] i. brevik , s. nojiri , s . d. odintsov and l. vanzo , phys . d * 70 * , 043520 ( 2004 ) [ doi : 10.1103/physrevd.70.043520 ] s. nojiri , s. d. odintsov , proc . wc * 2004 * , 024 ( 2004 ) [ arxiv : hep - th/0412030 ] . b. novosyadlyj , o. sergijenko , r. durrer , v. pelykh , jcap ( 2013 ) [ doi : 10.1088/1475 - 7516/2013/06/042 ] . m. r. setare and e. n. saridakis , phys . d * 79 * , 043005 ( 2009 ) [ doi : 10.1103/physrevd.79.043005 ] . h. wei , r - g . cai , d - f . zeng , class . quantum grav . * 22 * , 3189 ( 2005 ) [ doi : 10.1088/0264 - 9381/22/16/005 ] . m. sharif , a. jawad , eur . j. c * 72 * , 2097 ( 2012 ) [ doi : 10.1140/epjc / s10052 - 012 - 2097 - 8 ] . a. jawad , s. chattopadhyay , a. pasqua , astrophys . space sci . * 346 * , 273 ( 2013)[doi : 10.1007/s10509 - 014 - 2010-z ] . m. malekjani , a. khodam - mohammadi , n. nazari - pooya , astrophys . space sci . * 332 * , 515 ( 2011 ) [ doi : 10.1007/s10509 - 010 - 0550 - 4 ] . u. debnath , s. chattopadhyay , int . j. theor . phys . * 52 * , 1250 ( 2013 ) [ doi : 10.1007/s10773 - 012 - 1440-z ] . s. nojiri , s. d. odintsov , phys . b * 716 * , 377 ( 2012 ) [ doi : 10.1016/j.physletb.2012.08.049 ] . s. nojiri , s. d. odintsov , n. shirai , j. cosmol . . phys . * 1305 * , 020 ( 2013 ) [ doi : 10.1088/1475 - 7516/2013/05/020 ] . k. bamba , s. nojiri , s. d. odintsov , report ocha - pp-323 ( 2014 ) [ arxiv:1406.2417 [ hep - th ] ] . w. yang , y. wu , l. song , y. su , j. li , d. zhang , x. wang , mod . a * 26 * , 191 ( 2011 ) [ doi : 10.1142/s0217732311034682 ] k. karami , j. fehri , phys . b * 684 * , 61 ( 2010 ) [ doi : 10.1016/j.physletb.2009.12.060 ] . a. sheykhi , phys . b * 681 * , 205 ( 2009 ) [ doi : 10.1016/j.physletb.2009.10.011 ] . l. n. granda , a. oliveros , phys . b * 669 * , 275 ( 2008 ) [ doi : 10.1016/j.physletb.2008.10.017 ] . m. li , x - d . li , j. meng , z. zhang , phys . d * 88 * , 023503 ( 2013 ) [ doi : 10.1103/physrevd.88.023503 ] . l. n. granda , a. oliveros , phys . b * 671 * , 199 ( 2009 ) [ doi : 10.1016/j.physletb.2008.12.025 ] . s. nojiri , s. d. odintsov , j. phys . ser . * 66 * , 012005 ( 2007 ) [ doi : 10.1088/1742 - 6596/66/1/012005 ] . s. nojiri , s. d. odintsov , int . phys . , * 4 * , 115 ( 2007 ) [ doi : 10.1142/s0219887807001928 ] , econf c0602061 ( 2006 ) 06 . s. nojiri , s. d. odintsov , int . 11 * , 1460006 ( 2014 ) [ doi : 10.1142/s0219887814600068 ] . s. capozziello , v. f. cardone , v. salzano , phys . d * 78 * , 063504 ( 2008 ) [ doi : 10.1103/physrevd.78.063504 ] . k. bamba , c - q . geng , c - c . lee , int . j. mod . d * 20 * , 1339 ( 2011 ) [ doi : 10.1142/s0218271811019517 ] . k. bamba , c - q . geng , prog . . phys . * 122 * , 1267 ( 2009)[doi : 10.1143/ptp.122.1267 ] . r. myrzakulov , eur . j. c * 71 * , 1752 ( 2011 ) [ doi : 10.1140/epjc / s10052 - 011 - 1752 - 9 ] . k. bamba , c - q . geng , c - c . lee , l - w . luo , jcap * 1101 * , 021 ( 2011 ) [ doi : 10.1088/1475 - 7516/2011/01/021 ] . k. bamba , y. kokusho , s. nojiri , n. shirai , class . quantum grav . * 31 * , 075016 ( 2014 ) [ doi:10.1088/0264 - 9381/31/7/075016 ] . y. ito , s. nojiri , s. d. odintsov , entropy * 14 * , 1578 ( 2012 ) [ doi : 10.3390/e14081578 ] . k. nozari , a. behboodi , s. akhshabi , phys . b * 723 * , 201 ( 2013 ) [ doi : 10.1016/j.physletb.2013.04.058 ] . a. ali , r. gannouji , m. sami , phys . d * 82 * , 103015 ( 2010 ) [ doi : 10.1103/physrevd.82.103015 ] . k. bamba , a. n. makarenko , a. n. myagky , s. d. odintsov , phys . b * 732 * , 349 ( 2014 ) [ doi : 10.1016/j.physletb.2014.04.004 ] . k. bamba , s. nojiri , s. d. odintsov , jcap * 0810 * , 045 ( 2008 ) [ doi : 10.1088/1475 - 7516/2008/10/045 ] . k. bamba , c - q . geng , s. nojiri , s. d. odintsov , phys . d * 79 * , 083014 ( 2009 ) [ doi : 10.1103/physrevd.79.083014 ] . k. bamba , r. myrzakulov , s. nojiri , s. d. odintsov , phys . d * 85 * , 104036 ( 2012 ) [ doi : 10.1103/physrevd.85.104036 ] . k. bamba , c - q . geng , c - c . lee , jcap * 1008 * , 021 ( 2010 ) [ doi : 10.1088/1475 - 7516/2010/08/021 ] . k. bamba , proceedings of the kmi inauguration conference , nagoya university , nagoya , japan , 24 26 october 2011 ( 2011 ) [ arxiv : 1202.4317 [ gr - qc ] ] . k. bamba , d. momeni , r. myrzakulov , ( 2014 ) arxiv:1404.4255 [ hep - th ] . c. brans , r. h. dicke , phys . rev . * 124 * , 925 ( 1961 ) [ doi : 10.1103/physrev.124.925 ] . m. jamil , d. momeni , chinese phys . * 28 * , 099801 ( 2011 ) [ doi:10.1088/0256 - 307x/28/9/099801 ] . m. jamil , d. momeni , m. a. rashid , eur . j. c * 71 * , 1711 ( 2011 ) [ doi : 10.1140/epjc / s10052 - 011 - 1711 - 5 ] . m. jamil , i. hussain , d. momeni , eur . j. plus * 126 * , 80 ( 2011 ) [ doi : 10.1140/epjp / i2011 - 11080 - 2 ] . m. jamil , d. momeni , m. raza , r. myrzakulov , eur . j. c * 72 * , 1999 ( 2012 ) [ doi : 10.1140/epjc / s10052 - 012 - 1999 - 9 ] . d. momeni , m. r. setare , mod . a * 26 * , 2889 ( 2011 ) [ doi : 10.1142/s0217732311037169 ] . s. sen , s. , t. r. seshadri , int . j. mod d * 12 * , 445 ( 2003 ) [ doi : 10.1142/s0218271803003049 ] . h. alavirad , a. sheykhi , phys . b. * 734 * , 148 ( 2014 ) [ doi : 10.1016/j.physletb.2014.05.023 ] . j. khoury , a. weltman , phys . d * 69 * , 044026 ( 2004 ) [ doi : 10.1103/physrevd.69.044026 ] . j. khoury , a. weltman , phys . * 93 * , 171104 ( 2004 ) [ doi : 10.1103/physrevlett.93.171104 ] . t. p. waterhouse , ( 2006 ) [ arxiv : astro - ph/0611816 ] . s. chattopadhyay , isrn high energy physics * 2013 * , 414615 ( 2013 ) [ doi : 10.1155/2013/414615 ] . y. bisbar , phys . d * 86 * , 127503 ( 2012 ) [ doi : 10.1103/physrevd.86.127503 ] . r. a. el - nabulsi , eur . j. plus * 127 * , 23 ( 2012 ) [ doi : 10.1140/epjp / i2013 - 13055 - 7 ] . a. upadhye , s. s. gubser , j. khoury , phys . d , * 74 * , 104024 ( 2006 ) [ doi : 10.1103/physrevd.74.104024 ] . s. s. gubser , j. khoury , phys . d , * 70 * , 104001 ( 2004 ) [ doi : 10.1103/physrevd.70.104001 ] . s. chattopadhyay , u. debnath , int . j. mod d * 20 * , 1135 ( 2011 ) [ doi : 10.1142/s0218271811019293 ] . h. farajollahi , a. salehi , phys . d * 85 * , 083514 ( 2012 ) [ doi : 10.1103/physrevd.85.083514 ] . saaidi , a. mohammadi , t. golanbari , h. sheikhahmadi , b. ratra , phys . d * 86 * , 045007 ( 2012 ) [ doi : 10.1103/physrevd.86.045007 ] . a. pasqua , s. chattopadhyay , astrophys . space sci . * 348 * , 283 ( 2013 ) [ doi : 10.1007/s10509 - 013 - 1557 - 4 ] . e. n. saridakis , nuclear phys . b * 819 * , 116 ( 2009 ) [ doi : 10.1016/j.nuclphysb.2009.04.011 ] . m. yashar , b. bozek , a. abrahamse , a. albrecht , m. barnard , phys . d * 79 * , 103004 ( 2009 ) [ doi : 10.1103/physrevd.79.103004 ] . o. hrycyna , m. szydowski , phys . d * 88 * , 064018 ( 2013 ) [ doi : 10.1103/physrevd.88.064018 ] . a. a. usmani , p. p. ghosh , u. mukhopadhyay , p. c. ray , s. ray , mon . not . . soc . * 386 * , l92 ( 2008 ) [ doi : 10.1111/j.1745 - 3933.2008.00468.x ] . k. bamba , g. cognola , s. d. odintsov , s. zerbini , phys . d * 90 * , 023525 ( 2014 ) [ doi : 10.1103/physrevd.90.023525 ] . p. j. steinhardt , phil . a * 361 * , 2497 ( 2003 ) 361 [ doi : 10.1098/rsta.2003.1288 ] . j. martin , m. yamaguchi , phys . d * 77 * , 123508 ( 2008 ) [ doi : 10.1103/physrevd.77.123508 ] . t. padmanabhan , phys . d * 66 * , 021301 ( 2002 ) [ doi : 10.1103/physrevd.66.021301 ] . j. s. bagla , h. k. jassal , t. padmanabhan , phys . d * 67 * , 063504 ( 2003 ) [ doi : 10.1103/physrevd.67.063504 ] .

we study the nonhomogeneous boundary value problem for navier stokes equations of steady motion of a viscous incompressible fluid in a three dimensional bounded domain with multiply connected boundary . we prove that this problem has a solution in some axially symmetric cases , in particular , when all components of the boundary intersect the axis of symmetry .

1110.6301

let @xmath0 be a bounded domain in @xmath1 with with multiply connected lipschitz boundary @xmath2 consisting of @xmath3 disjoint components @xmath4 : @xmath5 , and @xmath6 . consider in @xmath0 the stationary navier stokes system with nonhomogeneous boundary conditions @xmath7 { \mathop{\rm div}}\,{\bf u } & = & 0 \qquad \hbox{\rm in } \;\;\omega , \\[4pt ] { \bf u } & = & { \bf a } \qquad \hbox{\rm on } \;\;\partial\omega . \end{array}\right.\ ] ] the continuity equation @xmath8 implies the necessary compatibility condition for the solvability of problem ( [ ns ] ) : @xmath9 where @xmath10 is a unit vector of the outward ( with respect to @xmath0 ) normal to @xmath2 and @xmath11 . starting from the famous paper of j. leray @xcite published in 1933 , problem ( [ ns ] ) was a subject of investigation in many papers ( see , e.g. , @xcite , @xcite , @xcite@xcite , @xcite@xcite , @xcite@xcite , etc . ) . however , for a long time the existence of a weak solution @xmath12 to problem ( [ ns ] ) was proved only under the condition @xmath13 or for sufficiently small fluxes ( see @xcite , @xcite@xcite , @xcite , @xcite , @xcite , etc . ) . condition ( [ flux0 ] ) requires the net flux @xmath14 of the boundary value @xmath15 to be zero separately across each component @xmath4 of the boundary @xmath2 , while the compatibility condition ( [ flux ] ) means only that the total flux is zero . thus , ( [ flux0 ] ) is stronger than ( [ flux ] ) ( condition ( [ flux0 ] ) does not allow the presence of sinks and sources ) . for a detailed survey of previous results one can see the recent papers @xcite or @xcite@xcite . in particular , in the last papers v.v . pukhnachev has established the existence of a solution to problem ( [ ns ] ) in the three dimensional case when the domain @xmath0 and the boundary value @xmath15 have an axis of symmetry and a plane of symmetry which is perpendicular to this axis , moreover , this plane intersects each component of the boundary . in this paper we study the problem in the axial symmetric case . let @xmath16 be coordinate axis in @xmath1 and @xmath17 , @xmath18 , @xmath19 be cylindrical coordinates . denote by @xmath20 the projections of the vector @xmath21 on the axes @xmath22 . a function @xmath23 is said to be _ axially symmetric _ if it does not depend on @xmath24 . a vector - valued function @xmath25 is called _ axially symmetric _ if @xmath26 , @xmath27 and @xmath28 do not depend on @xmath24 . a vector - valued function @xmath29 is called _ axially symmetric without rotation _ if @xmath30 while @xmath27 and @xmath28 do not depend on @xmath24 . we will use the following symmetry assumptions . ( so ) @xmath31 is a bounded domain with lipschitz boundary and @xmath32 is the axis of symmetry of the domain @xmath0 . ( as ) the assumptions ( so ) are fulfilled and the boundary value @xmath33 is axially symmetric . ( aswr ) the assumptions ( so ) are fulfilled and the boundary value @xmath33 is axially symmetric without rotation . denote by @xmath34 the bounded simply connected domain with @xmath35 , @xmath36 . let @xmath37 be the largest domain , i.e. , @xmath38 here and henceforth we denote by @xmath39 the closure of the set @xmath40 . let @xmath41 @xmath42 we shall prove the existence theorem if one of the following two additional conditions is fulfilled : @xmath43 or @xmath44 is sufficiently small ( @xmath45 is specified below in section [ poet ] ) . in particular , ( [ flux^2 ] ) includes the case @xmath46 , i.e , when each component of the boundary intersects the axis of symmetry . notice that in ( [ flux^1 ] ) , ( [ flux^2 ] ) the fluxes @xmath47 , could be arbitrary large . \(a ) m = n=2,title="fig : " ] ( b ) m=2 , n=3,title="fig : " ] ( c ) m=1 , n=3,title="fig : " ] on fig.1 we show several possible domains @xmath0 . in the case ( a ) all fluxes @xmath48 and @xmath49 are arbitrary ; in the case ( b ) fluxes @xmath50 are arbitrary , while the flux @xmath51 has to be nonnegative , but there are no restriction on its size ; in the case ( c ) fluxes @xmath48 are arbitrary , while @xmath49 and @xmath51 has to be `` sufficiently small '' . + the main result of the paper reads as follows . [ kmpth4.2 ] _ let the conditions ( as ) , ( [ flux ] ) be fulfilled . suppose that one of the conditions ( [ flux^1 ] ) or ( [ flux^2 ] ) holds . then the problem @xmath52 admits at least one weak axially symmetric solution @xmath53 . _ if , in addition , the conditions ( aswr ) are fulfilled , then the problem @xmath52 admits at least one weak axially symmetric solution without rotation . ( for the definition of a weak solution , see section 2.1 . ) the analogous results for the plane case were established in @xcite . the proof of theorem [ kmpth4.2 ] uses the bernoulli law for a weak solution of the euler equations and the one - side maximum principle for the total head pressure corresponding to this solution ( see section [ eueq ] ) . these results were obtained in @xcite for plane case ( see @xcite for more detailed proofs ) . the proof of the bernoulli law for solutions from sobolev spaces is based on recent results obtained in @xcite ( see also section [ sms ] ) . the short version of this paper was published in @xcite . the preprint version of this paper see in @xcite . by _ a domain _ we mean an open connected set . let @xmath54 be a bounded domain with lipschitz boundary @xmath2 . we use standard notations for function spaces : @xmath55 , @xmath56 , @xmath57 , @xmath58 , @xmath59 , where @xmath60 $ ] . in our notation we do not distinguish function spaces for scalar and vector valued functions ; it is clear from the context whether we use scalar or vector ( or tensor ) valued function spaces . @xmath61 is subspace of all solenoidal vector fields ( @xmath62 ) from @xmath63 with the norm @xmath64 . note that for functions @xmath65 the norm @xmath66 is equivalent to @xmath67 . working with sobolev functions we always assume that the `` best representatives '' are chosen . if @xmath68 , then the best representative @xmath69 is defined by @xmath70 0 \qquad\qquad\quad & \ ; { \rm otherwise } , \end{array}\right.\ ] ] where @xmath71 , @xmath72 is a ball of radius @xmath73 centered at @xmath74 . further ( see theorem [ kmpth2.1 ] ) we will discuss some properties of the best representatives of sobolev functions . the next lemmas concern the existence of a solenoidal extensions of boundary values and the integral representation of the bounded linear functionals vanishing on solenoidal functions . [ see corollary 2.3 in @xcite ] [ kmplem14.1 ] _ let @xmath75 be a bounded domain with lipschitz boundary . if @xmath76 and the equality ( [ flux ] ) is fulfilled , then there exists a solenoidal extension @xmath77 of @xmath15 such that @xmath78 _ from this lemma we can deduce some assertions for the symmetric case . [ kmplem14.1 ] _ let the conditions ( as ) , ( [ flux ] ) be fulfilled . then there exists an axially symmetric solenoidal extension @xmath79 of @xmath15 such that the estimate ( [ is2 ] ) holds . _ * proof . * let @xmath80 be solenoidal extension of @xmath15 from lemma [ kmplem14.1 ] . put @xmath81 clearly , each @xmath82 is also solenoidal extension of @xmath15 and the estimate ( [ is2 ] ) holds for @xmath82 with the same @xmath83 ( not depending on @xmath84 ) . by construction @xmath85 take a weakly convergence sequence @xmath86 in @xmath87 . then by construction @xmath88 , @xmath89 , and the estimate ( [ is2 ] ) holds . from ( [ is2.1 ] ) it follows that @xmath90 for all @xmath91 . hence @xmath92 is axially symmetric . @xmath93 [ kmplem14.1]_let the conditions ( aswr ) , ( [ flux ] ) be fulfilled . then there exists a solenoidal extension @xmath79 of @xmath15 such that @xmath94 is axially symmetric without rotation and the estimate ( [ is2 ] ) holds . _ * let @xmath95 be solenoidal extension of @xmath15 from the previous lemma [ kmplem14.1 ] . then by classical formula @xmath96 = \frac{1}{r } \frac{\partial}{\partial r}(\tilde a_r r ) + \frac{\partial}{\partial z}(\tilde a_z)=0 . \end{array}\ ] ] here @xmath97 because of axial symmetry . define the vector field @xmath98 by the formulas @xmath99 then by construction @xmath94 is axially symmetric without rotation , @xmath89 , and the estimate ( [ is2 ] ) holds . from ( [ is2.2 ] ) it follows that @xmath100 . @xmath93 [ see @xcite ] [ kmplem14.2 ] _ let @xmath75 be a bounded domain with lipschitz boundary and @xmath101 be a continuous linear functional defined on @xmath63 . if @xmath102 then there exists a unique function @xmath103 with @xmath104 such that @xmath105 moreover , @xmath106 is equivalent to @xmath107 . _ [ kmplem14.2]_if , in addition to conditions of lemma [ kmplem14.2 ] , the domain @xmath0 satisfies the assumption ( so ) and @xmath108 for all @xmath109 , @xmath110 $ ] , where @xmath111 , then the function @xmath112 is axially symmetric . _ * proof . * take the function @xmath112 from the assertion of lemma [ kmplem14.2 ] . for @xmath110 $ ] define the function @xmath113 by the formula @xmath114 . by construction , @xmath115 @xmath116 since @xmath112 is unique , we obtain the identity @xmath117 . @xmath93 [ see @xcite ] [ kmplem14.3 ] _ let @xmath75 be a bounded domain with lipschitz boundary and let @xmath77 be divergence free . then there exists a unique weak solution @xmath118 of the stokes problem satisfying the boundary condition @xmath119 , i.e. , @xmath120 and @xmath121 moreover , @xmath122 _ [ kmplem14.3 ] _ if , in addition to conditions of lemma [ kmplem14.3 ] , the domain @xmath0 satisfies the assumptions ( so ) and also @xmath92 is axially symmetric , then @xmath123 is axially symmetric too . _ * let @xmath123 be a solution of the stokes problem from lemma [ kmplem14.3 ] . for @xmath110 $ ] define the function @xmath124 by the formula @xmath125 . by construction , @xmath126 . moreover , @xmath127 where @xmath111 . because of the uniqueness , we obtain the identity @xmath128 . @xmath93 [ kmplem14.3]_if , in addition to conditions of lemma [ kmplem14.3 ] , the vector field @xmath92 is axially symmetric without rotation , then @xmath123 is axially symmetric without rotation too . _ * proof . * take the function @xmath129 from the assertion of lemma [ kmplem14.3 ] and define @xmath130 by the formulas @xmath131 then from lemma [ kmplem14.3 ] it follows the inclusion @xmath109 ( see also the formula ( [ is2.2 ] ) ) . consequently , from ( [ 4.1 ] ) we obtain @xmath132 but by the direct calculation @xmath133 formulas ( [ 4.3ax ] ) , ( [ 4.4ax ] ) imply the required equality @xmath134 . @xmath93 + for a function @xmath135 , consider a continuous linear functional @xmath136 . because of riesz representation theorem there exists a unique function @xmath137 such that @xmath138 denote @xmath139 . evidently , @xmath140 is a continuous linear operator from @xmath141 to @xmath61 . denote by @xmath142 the space of all axially symmetric vector - function from @xmath141 . analogously define the spaces @xmath143 , @xmath144 , @xmath145 , @xmath146 , @xmath147 , etc . [ kmplem14.4 ] _ the operator @xmath148 has the following symmetry properties : @xmath149 @xmath150 _ * proof . * the property ( [ 4.10ax ] ) can be proved in the same way as lemma [ kmplem14.3 ] and the property ( [ 4.11ax ] ) as lemma [ kmplem14.3 ] . @xmath93 [ kmplem14.5 ] _ the following inclusions are fulfilled : @xmath151 @xmath152 _ * proof . * by direct calculation . @xmath93 + assume that @xmath33 and let conditions @xmath153 , ( as ) ( or ( aswr ) ) be fulfilled . take the corresponding axially symmetric functions @xmath92 , @xmath123 from the above lemmas . denote @xmath154 . then the problem ( [ ns ] ) is equivalent to the following one @xmath155 = - \nabla p -\big({\bf u}\cdot \nabla\big){\bf u}\qquad \hbox{\rm in } \;\;\omega,\\[4pt ] { \mathop{\rm div}}\,{\bf w } = 0 \qquad \hbox{\rm in } \;\;\omega , \\[4pt ] { \bf w } = 0 \qquad \hbox{\rm on } \;\;\partial\omega . \end{array}\right.\ ] ] by a _ weak solution _ of problem ( [ ns ] ) we understand a function @xmath156 such that @xmath157 and @xmath158 @xmath159 because of riesz representation theorem for any @xmath160 there exists a unique function @xmath161 such that the right - hand side of the equality ( [ 4.3 ] ) is equivalent to @xmath162 for all @xmath109 . obviously , @xmath163 is a nonlinear operator from @xmath61 to @xmath61 . [ kmplem14.6 ] _ the operator @xmath164 is a compact operator . moreover , @xmath163 has the following symmetry properties : @xmath165 @xmath166 _ * proof . * the first statement is well known ( see @xcite ) . the statements about symmetry follow from the above lemmas . @xmath93 + obviously , the identity ( [ 4.3 ] ) is equivalent to the operator equation in the space @xmath61 : @xmath167 thus , we can apply the leray schauder fixed point theorem to the compact operators @xmath168 and @xmath169 . there hold the following statements . [ kmplem14.7]_let the conditions ( as ) , @xmath153 be fulfilled . suppose that all possible solutions of the equation @xmath170 , @xmath171 $ ] , @xmath172 are uniformly bounded in @xmath144 . then the problem @xmath52 admits at least one weak axially symmetric solution . _ [ kmplem14.9 ] _ let the conditions ( aswr ) , @xmath153 be fulfilled . suppose that all possible solutions of the equation @xmath173 , @xmath171 $ ] , @xmath174 are uniformly bounded in @xmath145 . then the problem @xmath52 admits at least one weak axially symmetric solution without rotation . _ first we recall some classical differentiability properties of sobolev functions . [ kmpthdor ] _ let @xmath176 . then the function @xmath177 is continuous and there exists a set @xmath178 such that @xmath179 , and the function @xmath177 is differentiable ( in the classical sense ) at each @xmath180 . furthermore , the classical derivative at such points @xmath74 coincides with @xmath181 , where @xmath182 . _ here and henceforth we denote by @xmath183 the one - dimensional hausdorff measure , i.e. , @xmath184 , where @xmath185 . the next theorems have been proved recently by j. bourgain , m. korobkov and j. kristensen @xcite . [ kmpth1.1 ] _ let @xmath186 be a bounded domain with lipschitz boundary and @xmath187 . then _ \(i ) @xmath188 ; \(ii ) for every @xmath189 there exists @xmath190 such that for any set @xmath191 with @xmath192 the inequality @xmath193 holds . \(iii ) for @xmath183almost all @xmath194 the preimage @xmath195 is a finite disjoint family of @xmath196curves @xmath197 , @xmath198 . each @xmath197 is either a cycle in @xmath199 @xmath200i.e . , @xmath201 is homeomorphic to the unit circle @xmath202 or it is a simple arc with endpoints on @xmath203 @xmath200 in this case @xmath197 is transversal to @xmath204 . [ kmpth1.2 ] _ let @xmath205 be a bounded domain with lipschitz boundary and @xmath206 . then for every @xmath189 there exists an open set @xmath207 and a function @xmath208 such that @xmath209 , and for each @xmath210 if @xmath211 then @xmath212 , the function @xmath177 is differentiable at the point @xmath74 , and @xmath213 , @xmath214 . _ we shall say that a value @xmath215 is _ regular _ if it satisfies the condition ( iii ) of theorem [ kmpth1.1 ] and @xmath211 for some @xmath216 from theorem [ kmpth1.2 ] . note , that by above theorems almost all values @xmath215 are regular . we will study the euler equation under the following assumptions . \(e ) _ let the conditions ( so ) be fulfilled . suppose that some axially symmetric functions @xmath217 and @xmath218 satisfy the euler system @xmath219 { \mathop{\rm div}}{\mathbf v } & = & 0 \end{array}\right.\ ] ] for almost all @xmath220 . moreover , suppose that @xmath221 _ denote @xmath222 , @xmath223 , @xmath224 . of course , on @xmath225 the coordinates @xmath226 coincides with coordinates @xmath227 . from the conditions ( so ) one can easily see that ( s@xmath228 ) @xmath229 is a bounded plane domain with lipschitz boundary . moreover , @xmath230 is a connected set for each @xmath36 . in other words , the family @xmath231 coincides with the family of all connected components of the set @xmath232 . then @xmath233 and @xmath112 satisfy the following system of equations in the plane domain @xmath199 : @xmath234 \frac{\partial p}{\partial r}-\lambda_0\frac{(v_\theta)^2}r+\lambda_0v_r\frac{\partial v_r}{\partial r}+ \lambda_0v_z\frac{\partial v_r}{\partial z}=0,\\[6pt ] \frac{v_\theta v_r}r+v_r\frac{\partial v_\theta}{\partial r}+ v_z\frac{\partial v_\theta}{\partial z}=0,\\[6pt ] \frac{\partial ( rv_r)}{\partial r}+\frac{\partial ( rv_z)}{\partial z}=0 \end{array}\right.\ ] ] ( these equations are fulfilled for almost all @xmath235 ) . the next statement was proved in ( * ? ? ? * lemma 4 ) and in ( * ? ? ? * theorem 2.2 ) . [ kmpth2.3 ] _ let the conditions ( e ) be fulfilled . then @xmath236 in particular , by axial symmetry , @xmath237 _ [ l_est]_under conditions of theorem [ kmpth2.3 ] , the following estimate @xmath238 holds , where the constant @xmath239 depends on @xmath0 only . _ one of the main purposes of this section is to prove the following fact . [ kmpth2.3 ] _ under conditions of theorem [ kmpth2.3 ] , the equalities @xmath240 are fulfilled . _ to prove the last theorem , we need some preparation , especially , a version of bernoulli law for sobolev case ( see below theorem [ kmpth2.2 ] ) . from the last equality in ( [ 2.1 ] ) and from ( [ ax1 ] ) it follows that there exists a stream function @xmath241 such that @xmath242 we have the following integral estimates : @xmath243 , @xmath244 by identities ( [ ax7 ] ) , we can rewrite the last formula in the following way : @xmath245 fix a point @xmath246 . for @xmath189 denote by @xmath247 the connected component of @xmath248 containing @xmath249 . since @xmath250 by sobolev embedding theorem @xmath251 . hence @xmath177 is continuous at points from @xmath252 . denote by @xmath253 the total head pressure corresponding to the solution @xmath254 . obviously , @xmath255 by direct calculations one easily gets the identity @xmath256 for almost all @xmath257 . [ kmpth2.2 ] _ let the conditions ( e ) be valid ( see the beginning of this section ) . then there exists a set @xmath258 such that @xmath259 and for any compact connected set @xmath260 , if @xmath261 then the identities @xmath262 hold . _ theorem [ kmpth2.2 ] was obtained for plane case in ( * ? ? ? * theorem 1 ) ( see also @xcite for detailed proof ) . to prove theorem [ kmpth2.2 ] , we need some preliminaries . [ pp ] _ let the conditions ( e ) be fulfilled . then the inclusion @xmath263 holds . _ * proof . * clearly , @xmath112 is the ( unique ) weak solution to the poisson equation @xmath264 p & = & \tilde p,\quad\hbox{\rm in } \partial\omega,\end{array}\right.\ ] ] with @xmath265 . let @xmath266 by the results of @xcite @xmath267 belongs to the hardy space @xmath268 so that by caldern zygmund theorem for hardy s spaces @xcite @xmath269 . let @xmath270 be the trace of @xmath271 on @xmath2 and let @xmath272 be the solution to the problem @xmath273 p _ * & = & \tilde p-\bar g \quad\hbox{\rm in } \partial\omega . \end{array}\right.\ ] ] by the uniqueness theorem @xmath274 @xmath93 from inclusion ( [ co1 ] ) it follows that @xmath275 for almost all @xmath276 . denote @xmath277 . equations ( [ 2.1 ] ) yield the equality @xmath278 and it is easy to deduce that @xmath279 consider the stream function @xmath177 . from ( [ ax1 ] ) , ( [ ax7 ] ) we have @xmath280 for @xmath183-almost all @xmath281 . then from the morse - sard property ( see theorem [ kmpth1.1 ] ) it follows that @xmath282 then by ( s@xmath228 ) ( see the beginning of section [ eueq ] ) @xmath283 [ rem_ext]since @xmath284 on @xmath285 ( in the sense of traces ) , we can extend the function @xmath177 to the whole half - plane @xmath225 : @xmath286 the functions @xmath287 can be extended to @xmath225 as follows : @xmath288 @xmath289 then the extended functions inherit the properties of the previous ones . namely , formulas ( [ 2.1 ] ) , ( [ ax7])([2.2 ] ) , ( [ 2.2a ] ) are fulfilled with @xmath290 , @xmath247 replaced by @xmath225 and @xmath291,\ z\in[-\frac1\varepsilon,\frac1\varepsilon]\}\ ] ] respectively . for @xmath292 denote by @xmath293 the straight line parallel to the @xmath294-axis : @xmath295 . working with sobolev functions we always assume that the `` best representatives '' are chosen . the basic properties of these `` best representatives '' are collected in the following [ kmpth2.1 ] _ there exists a set @xmath296 such that : _ ( i)@xmath297 ; \(ii ) for all @xmath298 @xmath299 @xmath300 moreover , the function @xmath177 is differentiable at @xmath74 and @xmath301 ; \(iii ) for all @xmath302 there exists an open set @xmath303 such that @xmath304 , @xmath305 , and the functions @xmath306 are continuous in @xmath307 ; \(iv ) for each @xmath308 and for any @xmath189 the convergence @xmath309 holds , where @xmath310 @xmath311}|\phi(r , z_0+t)-\phi(x_0)|+ \sup\limits_{z\in[z_0-\rho , z_0+\rho]}|\phi(r_0+t , z)-\phi(x_0)|<\varepsilon\}.\ ] ] \(v ) take any function @xmath312 and a closed set @xmath313 such that @xmath314 on @xmath315 . then for almost all @xmath316 and for all the connected components @xmath317 of the set @xmath318 the equality @xmath319 holds , the restriction @xmath320 is an absolutely continuous function , and formulas ( [ 2.1]),([2.2 ] ) are fulfilled @xmath321-almost everywhere on @xmath317 . most of these properties are from @xcite . for the detailed proof of theorem [ kmpth2.1 ] see , e.g. , @xcite . the last property ( v ) follows ( by coordinate transformation , cf . @xcite ) from the well - known fact that any function @xmath322 is absolutely continuous along almost all coordinate lines . the same fact together with ( [ 2.2a ] ) , ( [ 2.2 ] ) imply [ kmplem2]_for almost all @xmath292 the equality @xmath323 holds , moreover , @xmath324 , @xmath325 are absolutely continuous functions ( locally ) and @xmath326 _ below we prove that for the set @xmath327 from theorem [ kmpth2.1 ] the assertion of bernoulli law ( theorem [ kmpth2.2 ] ) holds . before we need some lemmas . [ lemth2.2 ] _ for almost all @xmath328 the equality @xmath329 holds , and for each continuum @xmath330 the identities @xmath331 are fulfilled . _ * proof . * fix any @xmath189 and consider a function @xmath312 and an open set @xmath332 with @xmath209 from theorem [ kmpth1.2 ] applied to the function @xmath333 , where the rectangle @xmath334 was defined by formula ( [ t2 ] ) . . then @xmath336 and @xmath337 for any @xmath338 . thus , by theorem [ kmpth2.1 ] ( v ) for almost all @xmath339 and for any connected component @xmath317 of the set @xmath340 the equality @xmath341 holds and the restriction @xmath320 is absolutely continuous , moreover , for any @xmath196smooth parametrization @xmath342\to k$ ] the identity ( [ 2.2 ] ) gives @xmath343'=\nabla \phi(\gamma(t))\cdot\gamma'(t)=0\quad\mbox{for } \ha^1\mbox{-almost all } t\in[0,1]\ ] ] ( the last equality is valid because @xmath344 on @xmath317 and , hence , @xmath345 ) . so , we have @xmath346 on @xmath317 . in view of arbitrariness of @xmath189 we have proved the assertion of the lemma . we need also some technical facts about continuity properties of @xmath347 at `` good '' points @xmath348 . [ lemth4 ] _ let @xmath349 . suppose that there exist a constant @xmath350 and a sequence of continuums @xmath351 such that @xmath352 , @xmath353 , @xmath354 as @xmath355 , and @xmath356 . then @xmath357 as @xmath355 . _ * proof . * without loss of generality we may assume that the projection of each @xmath358 on the @xmath359-axis is a segment @xmath360 $ ] of the length @xmath361 ( otherwise the corresponding fact is valid for projection of @xmath358 on the @xmath362-axis , etc . ) so by theorem [ kmpth2.1 ] ( iv ) for any @xmath189 we have @xmath363 for sufficiently large @xmath364 . thus @xmath365 for sufficiently large @xmath364 . @xmath93 [ lemth5 ] _ suppose for @xmath292 the assertion of lemma [ kmplem2 ] is fulfilled , i.e. , the equality @xmath323 holds , @xmath324 , @xmath325 are absolutely continuous functions , and formula ( [ z1 ] ) is valid . let @xmath366 be a compact set such that @xmath367 and @xmath368 ( recall that @xmath369 is called an interval adjoining @xmath315 if @xmath370 and @xmath371 ) . then @xmath372 _ * proof . * take any pair @xmath373 , @xmath374 . on the interval @xmath375 $ ] define a function @xmath376 by the rule @xmath377 . by construction , @xmath378 is an absolute continuous functions , and from ( [ t1.2 ] ) it follows that @xmath379 for any interval @xmath380 $ ] adjoining @xmath315 . since by definition the absolutely continuous function @xmath376 is differentiable almost everywhere and it coincides with the lebesgue integral of its derivative , we obtain @xmath381 hence , @xmath382 if @xmath383 $ ] and the interval @xmath384 contains only a finite number of points from @xmath315 . consider now the closed set @xmath385 : \mbox{\sl in\ any\ neighborhood\ of\ the\ point}\ z \ \mbox{\sl there\ exist}\\ \\ \qquad\qquad\qquad \qquad \mbox{\sl infinitely\ many\ points\ from}\ f \}. \end{array}\ ] ] it follows from ( [ 2.7 ] ) that @xmath386\setminus f_\infty } g^\prime(z)\,dz=0.\ ] ] according to the properties ( ii ) in theorem [ kmpth2.1 ] , the function @xmath177 is differentiable at any point @xmath387 , @xmath388 . from this fact and identity ( [ t1.1 ] ) we obtain @xmath389 for all @xmath390 . using ( [ ax7 ] ) , we can rewrite the last fact in the form @xmath391 for all @xmath390 . then , in view of formula ( [ z1 ] ) , we immediately derive @xmath392 summing formulas ( [ 2.8 ] ) and ( [ 2.9 ] ) , we get @xmath393 the last relation is equivalent to the target equality @xmath394 . the lemma is proved . * proof of theorem [ kmpth2.2]*. step 1 . because of remark [ rem_ext ] we can assume without loss of generality that a continuum @xmath317 is a connected component of the set @xmath395 , where @xmath396 , @xmath397 is a rectangle @xmath398,\ z\in[z_1,z_2]\}$ ] , @xmath399 , and @xmath400 , @xmath401 for each @xmath402 , where we denote @xmath403 put @xmath404 . for @xmath189 denote by @xmath405 the connected component of the compact set @xmath406\}$ ] containing @xmath317 . clearly , @xmath407 in hausdorff metric as @xmath408 . by theorem [ kmpth1.1 ] and lemma [ lemth2.2 ] for almost all @xmath189 the set @xmath409 is a finite disjoint union of @xmath196-curves and functions @xmath410 are constant on each of these curves . from the last two sentences by topological obviousness it follows that for each component @xmath411 of the open set @xmath412 there exists a sequence of continuums @xmath413 such that each @xmath414 is a @xmath196-curve homeomorphic to the segment @xmath415 $ ] or to the circle @xmath416 , @xmath414 is a connected component of the set @xmath417 , @xmath418 , @xmath419 in hausdorff metric as @xmath355 , and for any @xmath420 there exists an index @xmath421 such that @xmath74 and @xmath317 lie in the different connected components of the set @xmath422 for @xmath423 . from these facts using lemma [ lemth4 ] it is easy to deduce that for any @xmath411 there exists a limit @xmath424 such that @xmath425 step 2 . we claim that for almost all @xmath426 the identities @xmath427 hold . indeed , let @xmath426 satisfying the assertion of lemma [ kmplem2 ] and @xmath428 . put @xmath429:(r_0,z)\in k\}$ ] . from ( [ t4 ] ) it follows that @xmath430 for any interval @xmath431 $ ] adjoining @xmath315 . thus the target identity ( [ t5 ] ) follows directly from lemma [ lemth5 ] . we claim that there exists @xmath432 such that @xmath433 for each component @xmath411 ( see formula ( [ t4 ] ) ) . the proof of this claim splits in two cases . 3a ) let @xmath434 . then by construction ( see the beginning of step 1 ) @xmath435 , @xmath436 , @xmath437 , and from ( [ t4])([t5 ] ) it is easy to deduce that @xmath438 . 3b ) now suppose @xmath439 . let @xmath440 be a component such that @xmath441 . then for each horizontal line @xmath293 if @xmath442 then @xmath443 . from the last fact and from ( [ t4])([t5 ] ) it is easy to deduce that @xmath444 . formula ( [ t6 ] ) is proved completely . now we can rewrite ( [ t4])([t5 ] ) as follows : @xmath445 @xmath446 ( where @xmath447 is equal either to @xmath448 or to @xmath449 ) . we claim that @xmath450 for each @xmath451 . indeed , fix @xmath452 . the proof of the claim splits in two cases . 4a ) let there exists @xmath190 such that for any @xmath453 the inequality @xmath454 holds . then the equality ( [ t8 ] ) follows from ( [ t5 ] ) and the assertion ( iv ) of theorem [ kmpth2.1 ] . namely , fix @xmath189 and take @xmath455 ( this intersection is nonempty for sufficiently small @xmath456 ) such that @xmath457 and the identity ( [ t5 ] ) is fulfilled for @xmath458 , i.e. , @xmath459 by construction , @xmath460 . by our assumption 4a ) there exists a point @xmath461 such that @xmath462 . from theorem [ kmpth2.1 ] ( iv ) it follows that @xmath463 . using ( [ t5 ] ) , we finally obtain @xmath464 . 4b ) let the assumption 4a ) be false . then there exists a sequence @xmath465 such that @xmath466 we can assume without loss of generality that each segment @xmath467 is contained in some @xmath468 . denote by @xmath469 the open squares @xmath470 . then it is easy to deduce that for sufficiently large @xmath471 each set @xmath472 contains a continuum @xmath473 such that @xmath474 . indeed , by construction there exists @xmath475 such that @xmath476 ( the existence of such @xmath477 follows from above inclusions @xmath478 , @xmath479 ) . let @xmath473 be the closure of the connected component of the set @xmath480 containing the point @xmath481 . then @xmath482 ( otherwise there would be a contradiction with connectedness of @xmath317 ) . but by assumption ( [ tp5 ] ) @xmath473 does not intersect the segment @xmath467 . hence @xmath473 intersects at least one of other three sides of @xmath483 . in each case thus the equality ( [ t8 ] ) follows from ( [ t4 ] ) and lemma [ lemth4 ] . now equality ( [ t8 ] ) is proved for each @xmath484 . thus the proof of theorem [ kmpth2.2 ] is finished . @xmath93 . * proof of theorem [ kmpth2.3 ] . * to prove the equalities ( [ bp3 ] ) , we shall use the bernoulli law and the fact that the axis @xmath362 is `` almost '' a stream line . more precisely , @xmath362 is a singularity line for @xmath21 , @xmath177 , @xmath112 , but it can be accurately approximated by usual stream lines ( on which @xmath485 ) . first of all , let us simplify the geometrical setting . put @xmath486 and consider an extension of @xmath410 to @xmath487 by formulas of remark [ rem_ext ] . then the extended functions @xmath410 inherit the properties of the previous ones . namely , the bernoulli law ( see the assertion of theorem [ kmpth2.2 ] ) and ( [ ax3])([axc2 ] ) hold with @xmath290 , @xmath247 replaced by @xmath487 , @xmath488 , respectively . below we will use these facts only . so , we may assume , without loss of generality , that @xmath46 , i.e. , that @xmath489 is a _ simply connected _ plane domain . from ( [ ax3 ] ) it follows that there exists a sequence @xmath490 such that the convergence @xmath491 holds for lines @xmath492 . fix a point @xmath493 and denote by @xmath494 the connected component of the open set @xmath495 containing @xmath496 . obviously , for sufficiently large @xmath84 the open set @xmath494 is a simply connected plane domain with a lipschitz boundary , @xmath497 . we also have @xmath498 @xmath499 where @xmath500 , @xmath501 . then from ( [ ax8 ] ) and ( [ ax10 ] ) we conclude that @xmath502 in particular , @xmath503 , i.e. , @xmath504 our plan for the rest part of the proof is as follows . first , we prove that for any @xmath505 there exists a set @xmath506 such that @xmath507 @xmath508 @xmath509 notice that @xmath510 does not depend on @xmath74 , while @xmath511 can a priory depend on @xmath74 . however , finally we prove that @xmath512 for all @xmath513 . this fact will easily imply the target equalities ( [ bp3 ] ) . on @xmath514 define an equivalence relation by the rule @xmath515 a continuum @xmath516 such that @xmath517 and both @xmath518 do not belong to the unbounded connected component of the open set @xmath519 . by @xmath520 denote the corresponding class of equivalence . illustrate this definition by some examples . ( @xmath521 ) if @xmath516 is a continuum and @xmath522 , then @xmath523 . ( @xmath524 ) if @xmath516 is homeomorphic to the circle and @xmath525 , then @xmath526 , where @xmath527 is a bounded domain such that @xmath528 . for each @xmath529 the following properties of the relation @xmath530 hold ( for the proof of them , see appendix ) . ( @xmath531 ) @xmath532 and each @xmath520 is a compact set . ( @xmath533 ) the set @xmath520 is connected . ( @xmath534 ) @xmath535 . ( @xmath536 ) the set @xmath537 is connected . ( @xmath538 ) the set @xmath539 is connected . ( @xmath540 ) the formula @xmath541 holds . for @xmath542 put @xmath543 . because of topological obviousness @xmath544 then from ( @xmath534 ) , ( [ ax11])([ax11 ] ) and ( [ ax13 ] ) we conclude that the identity ( [ ax14 ] ) holds . from the bernoulli law ( see theorem [ kmpth2.2 ] ) it follows that @xmath545 fix any point @xmath546 and @xmath364 such that @xmath547 . by construction ( see the properties ( @xmath538)(@xmath540 ) , ( [ ax15- ] ) ) there exists a sequences of continuums @xmath548 and points @xmath549 such that @xmath550 for all sufficiently large @xmath84 , @xmath551 , and @xmath552 converges to some set @xmath317 with respect to the hausdorff metric as @xmath553 . hence @xmath554 , @xmath317 is a compact connected set , @xmath555 , and @xmath556 . consequently , @xmath557 again by the bernoulli law it follows that @xmath558 then from ( [ ax15])([ax15 ] ) , the connectedness of @xmath559 , and from the continuity properties of @xmath347 ( see theorem [ kmpth2.1](iii ) ) we conclude that the convergence @xmath560 holds . in particular , @xmath561 because the right - hand side of the last equality does not depend on the choice of @xmath562 , we have proved the identities ( [ ax16 ] ) with @xmath563 . now let @xmath292 satisfying the assertion of lemma [ kmplem2 ] and the conditions @xmath564 , @xmath565 . to finish the proof of the theorem , we need to show that @xmath566 put @xmath567:\ ( r_0,z)\in \partial u((r_0,z))\}.\ ] ] then by construction @xmath373 and the set @xmath315 is compact . indeed , denote @xmath568 , @xmath569 . since @xmath570 , we have @xmath571 , consequently , @xmath572 . analogously , @xmath573 , i.e. , @xmath373 . further , let @xmath574 . denote @xmath575 . then @xmath576 , @xmath577 . of course , @xmath578 ( otherwise @xmath579 for large @xmath471 ) . therefore @xmath580 , i.e. , @xmath581 . so , we prove that @xmath373 and that the set @xmath315 is compact . now from ( [ ax14])([ax16 ] ) the identities ( [ t1.1])([t1.2 ] ) hold . thus by lemma [ lemth5 ] we have the target equality ( [ 2.6 ] ) . @xmath93 in particular , during the last proof we established the following assertion . [ kmp_b]_assume that the conditions ( e ) be fulfilled . let @xmath552 be a sequence of compact sets with the following properties : @xmath582 , @xmath583 , and let there exist @xmath584 such that @xmath585 , @xmath586 . then there exist @xmath587 such that @xmath588 @xmath589 and @xmath590 as @xmath553 . _ let @xmath591 be a domain with lipschitz boundary . we say that the function @xmath592 satisfies a _ weak one - side maximum principle locally _ in @xmath593 , if @xmath594 holds for any strictly interior subdomain @xmath595 ( @xmath596 with the boundary @xmath597 not containing singleton connected components . ( in ( [ 2.13 ] ) negligible sets are the sets of 2dimensional lebesgue measure zero in the left _ esssup _ , and the sets of 1dimensional hausdorff measure zero in the right _ esssup_. ) if ( [ 2.13 ] ) holds for any @xmath598 ( not necessary strictly interior ) with the boundary @xmath597 not containing singleton connected components , then we say that @xmath592 satisfies a _ weak one - side maximum principle globally _ in @xmath593 ( in particular , we can take @xmath599 in ( [ 2.13 ] ) ) . [ kmpth2.4 ] _ let the conditions ( e ) be fulfilled . assume that there exists a sequence of functions @xmath600 such that @xmath601 and @xmath602 weakly in @xmath603 for some @xmath604 . if all @xmath605 satisfy the weak one side maximum principle locally in @xmath199 , then @xmath606 _ * proof . * let the conditions of theorem [ kmpth2.4 ] be fulfilled . then from ( * ? ? ? * theorem 2 ) ( see also @xcite for more detailed proof ) it follows that @xmath607 to prove the estimate ( [ bp4 ] ) in the whole domain @xmath290 we will use the same methods as in the proof of theorem [ kmpth2.3 ] . first of all , we simplify the situation : as above define the domain @xmath487 by equality ( [ axcdom ] ) and extend the functions @xmath410 to @xmath487 by formulas ( [ axc10])([axc11 ] ) . the extended functions @xmath410 inherit the properties of the previous ones . namely , ( [ ax3])([axc2 ] ) and the bernoulli law ( see theorem [ kmpth2.2 ] ) hold with @xmath290 , @xmath247 replaced by @xmath487 , @xmath488 , respectively . moreover , the maximum property ( * ) holds with @xmath290 replaced by @xmath487 . since in the proof below we will use only these facts , we may assume without loss of generality that @xmath46 , i.e. , @xmath489 is _ a simply connected _ plane domain . suppose the assertion of the theorem is false . then there exists a point @xmath608 such that @xmath609 take a sequences of numbers @xmath610 , the corresponding lines @xmath611 and domains @xmath612 from the proof of theorem [ kmpth2.3 ] ( in particular , the formula ( [ ax10 ] ) holds ) . denote by @xmath613 the connected component of the level set @xmath614 containing @xmath249 . by bernoulli law @xmath615 we have two possibilities : \(i ) @xmath616 for all @xmath84 . then , by lemma [ kmp_b ] , @xmath617 , and we obtain a contradiction with the assumption ( [ bp5 ] ) . \(ii ) there exists @xmath618 such that @xmath619 . then the sequence @xmath613 stabilizes after @xmath620 , i.e. , @xmath621 denote @xmath622 . then by construction @xmath623 now consider the family of sets @xmath624 defined during the proof of theorem [ kmpth2.3 ] . from ( [ ax15 ] ) it follows that @xmath625 where @xmath626 ( the last convergence follows from lemma [ kmp_b ] ) . take @xmath627 sufficiently large such that @xmath628 denote @xmath629 . by construction , @xmath630 but the last inequalities contradict the assertion ( * ) . the proof is complete . consider first the axially symmetric case with possible rotation . according to lemma [ kmplem14.7 ] , in order to prove the existence of the solution to problem ( [ ns ] ) it is enough to show that all possible solutions to the operator equation @xmath631 , \ \w\in h_{as}(\omega)\ ] ] are uniformly bounded in @xmath632 . we shall prove this estimate by contradiction , following the well - known argument of j. leray @xcite ( this argument was used also by many other authors , e.g. @xcite , @xcite , @xcite , @xcite , see also @xcite ) . suppose that the solutions to ( [ opereq ] ) are not uniformly bounded in @xmath632 . then there exists a sequence of functions @xmath633 such that @xmath634 with @xmath635 $ ] and @xmath636 . note that @xmath637 and the corresponding axially symmetric pressures @xmath638 satisfy the following integral identity @xmath639 @xmath640 for any @xmath641 . here @xmath642 is an axially symmetric solution to the stokes problem ( see lemmas [ kmplem14.3][kmplem14.3 ] ) . denote @xmath643 , @xmath644 , @xmath645 , @xmath646 . then @xmath647 and the following estimates @xmath648 hold for any @xmath649 ( the detailed proof of the above estimates see , for example , in @xcite ) . extracting a subsequences , we can assume without loss of generality that @xmath650,\ ] ] @xmath651 @xmath652 + multiplying the integral identity ( [ identity ] ) an arbitrary fixed @xmath653 by @xmath654 and passing to a limit as @xmath655 , yields that the limit functions @xmath233 and @xmath112 satisfy the euler equations @xmath656 { \mathop{\rm div}}{\mathbf v } & = & 0 \\[4pt ] { \mathbf v}|_{\partial\omega } & = & 0 \end{array}\right.\ ] ] ( the details of the proof see , for example , in @xcite ) . from equations ( [ euler ] ) and from inclusions ( [ exc3 ] ) , ( [ exc3 ] ) it follows that @xmath657 . thus the assumptions ( e ) from the beginning of the section [ eueq ] are fulfilled . moreover , @xmath658 . now , taking in ( [ identity ] ) @xmath659 we get @xmath660 using the compact embedding @xmath661 @xmath662 , we can pass to a limit as @xmath655 in equality ( [ 4.7 ] ) . as a result we obtain @xmath663 from the last formula and euler equation ( [ euler ] ) , we derive @xmath664 because of ( [ bp2 ] ) the last equality could be rewritten in the following equivalent form @xmath665 now using ( [ flux ] ) and ( [ bp3 ] ) from ( [ exc5 ] ) we derive @xmath666 consider , first , the case ( [ flux^2 ] ) . if the condition ( [ flux^2 ] ) is fulfilled with @xmath667 , where @xmath239 is a constant from lemma [ l_est ] , then from ( [ exc6 ] ) and ( [ bp3 ] ) it follows a contradiction ( recall that @xmath668 $ ] ) . thus , the proof the case ( [ flux^2 ] ) is complete . consider now the case when condition ( [ flux^1 ] ) is fulfilled . then the equality ( [ exc6 ] ) takes the following form : @xmath669 from ( [ flux^1 ] ) , ( [ exc7 ] ) it follows that @xmath670 consider the identity @xmath671 @xmath672 integrating the above identity by parts in @xmath0 , we get @xmath673 @xmath674 @xmath675 @xmath676 hence , @xmath677 the total head pressures @xmath678 for the navier stokes system ( [ ns ] ) satisfy the equations @xmath679 satisfy the one - side maximum principle locally in @xmath0 . denote @xmath680 . from ( [ exc3])([exc3 ] ) and from the symmetry assumptions it follows that the sequence @xmath681 weakly converges to @xmath682 in the space @xmath683 . therefore , by theorem [ kmpth2.4 ] , @xmath684 ( the last equality follows from the conditions @xmath685 and ( [ exc8 ] ) ) . then it follows from ( [ exc22 ] ) that @xmath686 and we obtain the contradiction with ( [ exc8 ] ) , which proves theorem in the case of condition ( [ flux^1 ] ) . if the boundary value @xmath687 is axially symmetric without rotation , the proof of theorem [ kmpth4.2 ] is just the same as in the first part ; we need only to use lemma [ kmplem14.9 ] instead of lemma [ kmplem14.7 ] . let us prove the topological properties ( @xmath531)(@xmath540 ) of the equivalence class @xmath520 , @xmath529 , which were used in the proof of theorem [ kmpth2.3 ] . ( @xmath531 ) indeed , if @xmath688 , then by definition there exists a sequence of continuums @xmath358 such that @xmath689 and @xmath690 do not belong to the unbounded connected component of the set @xmath691 . without loss of generality we may assume that @xmath358 converge with respect to the hausdorff metric to the set @xmath317 . then @xmath317 is a continuum , @xmath522 , and it is easy to see that neither @xmath74 nor @xmath692 belongs to the unbounded connected component of the open set @xmath519 . ( @xmath533 ) fix any @xmath693 . take the corresponding set @xmath317 from the definition of @xmath694 . then @xmath516 is a compact connected set such that @xmath517 and both @xmath518 do not belong to the unbounded connected component of the open set @xmath519 . denote by @xmath695 the family of connected components of the open set @xmath519 . let @xmath696 be an unbounded component . since the domain @xmath494 is simply connected , we have @xmath697 for each @xmath698 . hence by definition of @xmath530 we obtain @xmath699 for each @xmath698 . by construction , each set @xmath317 , @xmath700 is connected and @xmath701 . from these facts we conclude that the set @xmath702 is connected and the inclusions @xmath703 hold . the last assertion and arbitrariness of @xmath693 imply the connectedness of @xmath520 . ( @xmath534 ) to prove the property @xmath704 , we may assume , without loss of generality , that @xmath705 . fix any @xmath706 . take the corresponding set @xmath317 from the definition of @xmath694 and the sets @xmath695 from the proof of property ( @xmath533 ) . then it is easy to see that @xmath707 indeed , if for example @xmath708 , then @xmath709 for some @xmath698 . but by construction @xmath695 is an open set and @xmath710 . these facts contradict the assumption @xmath711 . this proves the inclusion ( [ appp1 ] ) . from ( [ appp1 ] ) and the assumption @xmath517 we obtain the required equality @xmath712 . using similar elementary arguments , it is easy to prove the next two properties ( @xmath536)(@xmath538 ) . therefore , we shall prove in detail only the last property ( @xmath540 ) . ( @xmath540 ) suppose the formula ( [ ax13 ] ) is not true , i.e. , @xmath713 hence , @xmath714 let @xmath715 for all @xmath706 . fix @xmath716 . from properties ( @xmath521 ) , ( @xmath534 ) , ( @xmath538 ) it follows that @xmath717 where we denote by @xmath718 the connected component of the level set @xmath719 containing the point @xmath720 . by construction , the closure of each connected component @xmath721 of the set @xmath722 intersects the line @xmath611 and @xmath723 ( see the formulas ( [ axc13])([axc14 ] ) , ( [ ax11 ] ) ) . hence , the conditions ( [ ax19])([ax19 ] ) imply the assertion @xmath724 take a sequence @xmath725 such that each value @xmath726 is regular from the viewpoint of morse - sard theorem ( see theorem [ kmpth1.1 ] ( iii ) ) . denote by @xmath727 the connected component of the level set @xmath728\}$ ] containing @xmath718 . then for sufficiently large @xmath364 the boundary @xmath729 consists of finite disjoint family of @xmath196cycles in @xmath730 ( it follows from the formula ( [ ax19 ] ) and from the evident convergence @xmath731 ) . denote by @xmath732 the cycle separating the set @xmath727 from infinity , and denote by @xmath733 the bounded domain such that @xmath734 . then by construction @xmath735 , @xmath736 , and @xmath737 . consequently , @xmath738 on the other hand , by property ( @xmath524 ) all points of @xmath733 are @xmath530 equivalent . the last assertion contradicts the formula ( [ ax20 ] ) and the definition of @xmath520 . the property ( [ ax13 ] ) is proved . @xmath739the authors are deeply indebted to v.v . pukhnachev for valuable discussions . : on stationary solutions to navier stokes equation in symmetric plane domain under general outflow condition , _ pitman research notes in mathematics , proceedings of international conference on navier stokes equations . theory and numerical methods . june 1997 . varenna , italy _ ( 1997 ) * 388 * , 1630 . : on spaces of solenoidal vector fields and boundary value problems for the navier stokes equations in domains with noncompact boundaries , _ trudy mat . inst . steklov _ * 159 * ( 1983 ) , 536 . english transl . : _ proc . steklov _ * 159 * ( 1984 ) , 334 . : on some problems of vector analysis and generalized formulations of boundary value problems for the navier stokes equations , _ zapiski nauchn . * 59 * ( 1976 ) , 81116 ( in russian ) ; english translation in _ journal of soviet mathematics _ * 10 * ( 1978 ) , no.2 , 257286 . : a new approach to the existence of weak solutions of the steady navier stokes system with inhomoheneous boundary data in domains with noncompact boundaries , _ arch . rational mech . anal _ * 198 * , no . 1 ( 2010 ) , 331348 . : viscous flows in domains with a multiply connected boundary , _ new directions in mathematical fluid mechanics . the alexander v. kazhikhov memorial volume . fursikov a.v . , galdi g.p . and pukhnachev v.v . berlin : birkhauser _ ( 2009 ) 333348 .

in this paper , we study the graphical structure of elementary trapping sets ( ets ) of variable - regular low - density parity - check ( ldpc ) codes . etss are known to be the main cause of error floor in ldpc coding schemes . for the set of ldpc codes with a given variable node degree and girth , we identify all the non - isomorphic structures of an arbitrary class of etss , where is the number of variable nodes and is the number of odd - degree check nodes in the induced subgraph of the ets . our study leads to a simple characterization of dominant classes of etss ( those with relatively small values of and ) based on short cycles in the tanner graph of the code . for such classes of etss , we prove that any set in the class is a layered superset ( lss ) of a short cycle , where the term `` layered '' is used to indicate that there is a nested sequence of etss that starts from the cycle and grows , one variable node at a time , to generate . this characterization corresponds to a simple search algorithm that starts from the short cycles of the graph and finds all the etss with lss property in a guaranteed fashion . specific results on the structure of etss are presented for , and in this paper . the results of this paper can be used for the error floor analysis and for the design of ldpc codes with low error floors .

1308.1259

performance of low - density parity - check ( ldpc ) codes under iterative decoding algorithms in the error floor region is closely related to the problematic structures of the code s tanner graph @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite . following the nomenclature of @xcite , here , we collectively refer to such structures as _ trapping sets_. the most common approach for classifying the trapping sets is by a pair @xmath2 , where @xmath3 is the size of the trapping set and @xmath4 is the number of odd - degree ( unsatisfied ) check nodes in the subgraph induced by the set in the tanner graph of the code . among the trapping sets , the so - called _ elementary trapping sets ( ets ) _ are known to be the main culprits @xcite , @xcite , @xcite , @xcite , @xcite , @xcite . these are trapping sets whose induced subgraph only contains check nodes of degree one or two . for a given ldpc code , the knowledge of dominant trapping sets , i.e. , those that are most harmful , is important . such knowledge can be used to estimate the error floor @xcite , to modify the decoder to lower the error floor @xcite , @xcite , @xcite , or to design codes with low error floor @xcite , @xcite . while the knowledge of dominant trapping sets is most helpful in the design and analysis of ldpc codes , attaining such knowledge is generally a hard problem @xcite . much research has been devoted to devising efficient search algorithms for finding small ( dominant ) trapping sets , see , e.g. , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , and to the ( partial ) characterization of such sets @xcite , @xcite @xcite , @xcite , @xcite , @xcite . asymptotic analysis of trapping sets has also been carried out in @xcite , @xcite , @xcite , @xcite , @xcite , @xcite . @xcite studied the characterization of small @xmath2 trapping sets of size up to 8 ( @xmath9 ) and @xmath10 in ldpc codes from steiner triple systems ( sts ) . sts ldpc codes are a special category of regular ldpc codes with variable node degree 3 . @xcite showed that for a regular ldpc code with variable node degree @xmath11 and check node degree @xmath12 , and girth @xmath13 , no @xmath14 trapping set with @xmath15 and @xmath16 can exist . they also studied the trapping sets of euclidean geometry ( eg ) ldpc codes and provided some bounds on the size and the number of unsatisfied check nodes of the trapping sets of such codes @xcite . an eg - ldpc code with parameter @xmath17 is a regular ldpc code of length @xmath18 , with variable node degree @xmath19 and check node degree @xmath17 . a consequence of the bounds derived in @xcite is that for the case where @xmath20 , there is no trapping set of size smaller than the minimum distance of the code , i.e. , @xmath21 , with less than @xmath22 unsatisfied check nodes . a subset of trapping sets , called _ absorbing sets _ , for array - based ldpc codes with variable node degrees 2 , 3 and 4 , were studied in @xcite , @xcite . absorbing sets are trapping sets in which each variable node is connected to more satisfied than unsatisfied check nodes in the induced subgraph of the set @xcite , @xcite . array - based ldpc codes are a subclass of ( regular ) protograph ldpc codes which are constructed by lifting a fully - connected base graph using cyclic permutations . the analysis in @xcite and @xcite was focused on minimal absorbing sets , i.e. , the ones with the smallest size and with the smallest number of unsatisfied check nodes for a given size . @xcite studied the topological structure of trapping sets of size up to @xmath23 in regular ldpc codes with variable node degree 3 , and proposed a hierarchical search method to find them . the study of the graphical structure of trapping sets so far has been mainly limited to structured codes , codes with certain variable node degrees , and to relatively small trapping sets . in this work , for the category of variable - regular ldpc codes with a certain variable node degree and a given girth , we study the topological structure of @xmath2 etss for given values of @xmath3 and @xmath4 , and find all the non - isomorphic structures of such etss . a careful examination of these structures , which are independent of the check node degree distribution of the code , reveals that for relatively small values of @xmath3 and @xmath4 , the structures are all layered supersets ( lss ) of small cycles , i.e. , they can be characterized by a nested sequence of etss which starts from a short cycle and grows to the ets one node at a time . the lss property lends itself to a simple search algorithm that starts from short cycles of the code s tanner graph as input and can find all the etss with lss property in a guaranteed fashion . although the general approach discussed here can be applied to any category of variable - regular ldpc codes with arbitrary variable node degree @xmath0 and girth @xmath1 and to any class of etss with arbitrary values of @xmath3 and @xmath4 , the results presented here are for @xmath6 , @xmath7 , and @xmath8 . one of the main advantages of the results presented here is that they are applicable to specific codes , rather than just to an ensemble or a category of codes . in particular , the search algorithm based on lss property can be used to efficiently find the dominant etss of a code in a guaranteed fashion . this , for example , would imply having a faster and more accurate estimation of error floor for the code under consideration using techniques such as importance sampling . moreover , the results presented here can be used in the design of ldpc codes with low error floor . this can be achieved by avoiding certain dominant etss in the tanner graph of the code . in such a context , this work can help in identifying the dominant etss . it has been known that dominant trapping sets of ldpc codes have a close relationship with short cycles in the code s tanner graph @xcite , @xcite , @xcite . this work takes a rigorous step in establishing such a relationship . in general , in comparison with existing results on characterization of trapping sets such as @xcite , @xcite , @xcite , the results presented here are more general in terms of being applicable to both structured and random codes , and to cover a wider range of variable node degrees and trapping set classes . the remainder of this paper is organized as follows . basic definitions and notations are provided in section ii . in section iii , we present and discuss the lss property . in section iv , we develop the algorithm which guarantees to find all the etss with the lss property starting from the short cycles of the tanner graph . sections v , vi , vii , and viii present the results for variable - regular ldpc codes with variable node degrees 3 , 4 , 5 and 6 , respectively . as part of the material presented in these sections , we provide the lengths of short cycles that are required for the proposed algorithm to find all the @xmath2 etss with lss property in a guaranteed fashion , for different values of @xmath3 and @xmath4 . section ix is devoted to discussions and conclusions . let @xmath24 be the bipartite graph , or tanner graph , corresponding to the ldpc code @xmath25 , where @xmath26 is the set of variable nodes , @xmath27 is the set of check nodes and @xmath28 is the set of edges . the notations @xmath26 and @xmath27 refer to left " and right , " respectively , pointing to the side of the bipartite graph where variable nodes and check nodes are located , respectively . a _ cycle _ of length @xmath29 in a graph @xmath30 is a non - empty alternating sequence @xmath31 of nodes and edges in @xmath30 such that @xmath32 , and @xmath33 for all @xmath34 , @xmath35 , and all the other nodes are distinct . the length of the shortest cycle in a graph @xmath30 is denoted by @xmath1 and is called the _ girth _ of @xmath30 . the degree of a node @xmath36 is denoted by @xmath37 . a bipartite graph is called _ variable - regular _ or _ left - regular _ with left degree @xmath0 if @xmath38 . the graphs @xmath39 and @xmath40 are _ isomorphic _ if there is a bijection @xmath41 such that nodes @xmath42 , @xmath43 are joined by an edge if and only if @xmath44 and @xmath45 are joined by an edge . for a subset @xmath46 , @xmath47 denotes the set of neighbors of @xmath48 in @xmath49 . the _ induced subgraph _ of @xmath50 , represented by @xmath51 , is the graph containing nodes @xmath52 with edges @xmath53 . the set of check nodes in @xmath47 with odd degree in @xmath51 is denoted by @xmath54 . similarly , @xmath55 represents the set of check nodes in @xmath47 with even degree in @xmath51 . in this paper , we use the terms _ satisfied check nodes _ and _ unsatisfied check nodes _ to refer to the check nodes in @xmath55 and @xmath54 , respectively . given a tanner graph @xmath56 , a set @xmath57 is called an @xmath2 _ trapping set _ if @xmath58 and @xmath59 . the integer @xmath3 is referred to as the _ size _ of the trapping set @xmath50 . we also refer to all the trapping sets with the same parameters @xmath3 and @xmath4 as a _ class _ of trapping sets . an @xmath2 trapping set @xmath50 is called _ elementary _ if all the check nodes in @xmath51 have degree one or two . a set @xmath57 is called an @xmath2 _ absorbing set _ if @xmath50 is an @xmath2 trapping set and if all the nodes in @xmath50 are connected to more nodes in @xmath55 than to nodes in @xmath54 . trapping sets with smaller values of @xmath3 and @xmath4 are generally more harmful to iterative decoding . loosely speaking , such trapping sets are called _ dominant_. without loss of generality , we assume that the induced subgraph of a trapping set is connected . disconnected trapping sets can be considered as the union of connected ones . moreover , to the best of our knowledge , almost all the structures reported as dominant trapping sets ( of regular ldpc codes ) in the literature have the property that every variable node is connected to at least two satisfied check nodes in the induced subgraph . we thus focus on trapping sets with this property . in the rest of the paper , we use the notation @xmath60 to denote the set of all trapping sets @xmath48 in a graph @xmath30 whose induced subgraph @xmath51 is connected and for which every node @xmath61 is connected to at least two nodes in @xmath55 . in the following , we also assume that the tanner graph @xmath30 has no parallel edges . elementary trapping sets ( ets ) in left - regular tanner graphs are the main focus of this paper . to investigate the structure of etss of a certain @xmath2 class in left - regular tanner graphs with left - degree @xmath0 and girth @xmath1 , we need to obtain all the non - isomorphic graphical structures of such trapping sets . to simplify the representation of the subgraph induced by an ets in a left - regular graph , we often use an alternate graphical representation called _ normal _ graphs @xcite . the normal graph of an ets is obtained from the induced subgraph of the set by removing all the degree-1 check nodes and their edges from the subgraph , and by replacing each degree-2 check node with an edge . figures [ normal]@xmath62 and [ normal]@xmath63 represent the induced subgraph and the normal graph of a @xmath64 ets in a left - regular graph with @xmath65 and @xmath66 , respectively . in figure [ normal]@xmath62 , and the rest of the paper , variable nodes are represented by circles , and satisfied and unsatisfied check nodes are shown by empty and full squares , respectively . elementary trapping set and its normal graph.,width=288 ] it is easy to see that there is a one - to - one correspondence between an ets of a left - regular graph and its normal graph . given @xmath0 , one can construct the subgraph of the ets from a given normal graph by replacing each edge @xmath67 of the normal graph with two edges @xmath68 and @xmath69 , where @xmath70 is a degree-2 check node which is also added to the graph , and by connecting @xmath71 check nodes of degree one to every variable node @xmath72 with @xmath73 . in the following , we provide an example to demonstrate how all the non - isomorphic structures of a class of etss can be found for rather small values of @xmath0 , @xmath3 and @xmath4 . any @xmath74 ets of a left - regular ldpc code with @xmath65 and @xmath66 has one of the structures presented in figure [ 6_2_g ] . [ prop1 ] etss in left - regular ldpc codes with @xmath75 and @xmath76.,width=499 ] we use the normal graph representation , and prove that any @xmath74 ets in a left - regular ldpc code with @xmath75 and @xmath76 has one of the normal graph representations given in figure [ 6_2_ng ] . etss in left - regular ldpc codes with @xmath75 and @xmath66.,width=480 ] there are only two possibilities for @xmath74 etss : ( i ) both unsatisfied check nodes are connected to the same variable node , and ( ii ) the two unsatisfied check nodes are connected to two different variable nodes . we show that there is only one structure for the first case and only 2 non - isomorphic structures for the second case . for case ( i ) , let @xmath42 be the only variable node connected to two unsatisfied check nodes . this means that in the normal graph representation , @xmath42 is connected to only two other variable nodes . suppose that those variable nodes are @xmath77 and @xmath78 . since the degree of all the other variable nodes @xmath79 , @xmath80 and @xmath81 in the normal graph is 4 , each of them must have 4 distinct neighbors . the only possibility is for each of these three nodes to be connected to the other two nodes in addition to nodes @xmath77 and @xmath78 . this also satisfies the degree requirements for @xmath77 and @xmath78 ( each of them has degree 4 and is connected to all the nodes @xmath79 , @xmath80 , @xmath81 and @xmath42 ) . this topology is shown in figure [ 6_2_ng]@xmath62 . for case ( ii ) , let @xmath42 and @xmath77 be the two variable nodes , each connected to one unsatisfied check node . since all the other 4 nodes have degree 4 and must be connected to 4 distinct nodes , each of them must be connected to at least one of the nodes @xmath42 and @xmath77 . there are two possibilities : a ) @xmath42 and @xmath77 are connected together , and b ) @xmath42 and @xmath77 are not connected together . one can show that the only possible topology for case ( a ) is the one shown in figure [ 6_2_ng]@xmath63 , and the only possible topology for case ( b ) is the one shown in figure [ 6_2_ng]@xmath82 . for case ( a ) , there always exists a node @xmath78 which is connected to one of the nodes @xmath42 and @xmath77 , but not to both . otherwise , if all the other 4 nodes are connected to both @xmath42 and @xmath77 , both @xmath42 and @xmath77 will have degree 5 which contradicts the assumption of the proposition that @xmath65 . without loss of generality , we assume @xmath78 to be connected to @xmath42 , and we grow the normal graph from @xmath78 as the root . figure [ ab_grow]@xmath62 shows such a graph growth , in which all the nodes except @xmath77 are located in the first layer and @xmath77 is in the second layer . node @xmath77 must be connected to two nodes other than @xmath42 . without loss of generality , we assume those nodes to be @xmath79 and @xmath80 ( figure [ ab_grow]@xmath63 ) . from this point on , there is no option in connecting the nodes . node @xmath81 must be connected to 4 distinct nodes , and the only possibility is to have it connected to all the nodes @xmath79 , @xmath80 and @xmath42 . this will satisfy the degree of @xmath42 , which must be 3 . to satisfy the degrees of @xmath79 and @xmath80 ( @xmath83 ) , the only option is to have a connection between them , which results in the topology shown in figure [ 6_2_ng]@xmath63 . the proof that case ( b ) , i.e. , the case where @xmath42 and @xmath77 are not connected , will result in the topology of figure [ 6_2_ng]@xmath82 is similar and is thus omitted . .,width=288 ] the structures in figures [ 6_2_g]@xmath63 and [ 6_2_g]@xmath82 are absorbing sets while the one in figure [ 6_2_g]@xmath62 is not . moreover , proposition [ prop1 ] implies that there are no @xmath74 etss in left - regular graphs with @xmath65 and @xmath84 . finding the non - isomorphic structures for etss with rather large values of @xmath0 , @xmath3 or @xmath4 can be a formidable task . we thus resort to software programs to find such structures . one of the well - known software programs related to graph isomorphism is the _ nauty _ program @xcite . this program can be used to efficiently generate all the non - isomorphic graphs with a given number of nodes ( up to 32 ) and a given number of edges . the program has many input options to determine the minimum and maximum values of the node degrees , to select the girth , and to generate only bipartite graphs or all the possible graphs . for the case of bipartite graphs , however , the program does not have the option of taking the degree distribution of each part of the graph as an input . in the case of finding the non - isomorphic structures ( for induced subgraphs ) of etss , this limitation results in having a large number of undesired structures at the output of the program . one is thus required to check all the output structures to find the ones that satisfy the particular degree distributions of the class of etss under consideration . this difficulty can be circumvented by using the normal graph representation of etss as explained in the following example . consider the class of @xmath85 etss in a left - regular graph with @xmath65 and @xmath66 . figure [ 66_normal]@xmath62 shows one such ets . elementary trapping set and its normal graph.,width=268 ] it is easy to see that the induced subgraph of any ets in this class has 6 variable nodes of degree 4 , 6 check nodes of degree 1 , and @xmath86 check nodes of degree 2 . this is a total of 21 nodes ( check and variable nodes ) and 24 edges . to generate the non - isomorphic structures of @xmath85 etss using _ nauty _ package , one way is to set the package parameters to generate all the connected bipartite graphs of girth at least 6 with 21 nodes and 24 edges , and with the minimum and maximum degrees of 1 and 4 , respectively . this results in @xmath87 graphs , from which an overwhelming majority are not @xmath85 etss . alternatively , we can use the normal graph representation of @xmath85 etss . figure [ 66_normal]@xmath63 shows the normal graph of the structure shown in figure [ 66_normal]@xmath62 . similar to figure [ 66_normal]@xmath63 , any normal graph of a @xmath88 ets in a left - regular graph with @xmath65 has 6 nodes and 9 edges . to generate all the non - isomorphic normal structures using the _ nauty _ program , one needs to generate all the bipartite and non - bipartite graphs with 6 nodes and 9 edges , and with the minimum and maximum node degrees of 2 and 4 , respectively . this reduces the number of possible graphs from @xmath87 to only 11 . all the 11 graphs correspond to valid normal structures for the class of etss under consideration . these structures are shown in figures [ 66_4_c ] , [ 66_4_a ] , [ 66_4_b ] , and [ 66_4_d ] . among the 11 structures , only 2 structures are absorbing sets . these are shown in figure [ 66_4_c ] . the other structures are grouped together based on the number of variable nodes with two unsatisfied check nodes . the structures with 1 , 2 , and 3 variable nodes connected to two unsatisfied check nodes are shown in figures [ 66_4_a ] , [ 66_4_b ] , and [ 66_4_d ] , respectively . ets in a left - regular graph with @xmath65 and @xmath76 : the only two possible absorbing set topologies.,width=355 ] ets in a left - regular graph with @xmath65 and @xmath76 : topologies with only one variable node connected to two unsatisfied check nodes.,width=422 ] ets in a left - regular graph with @xmath65 and @xmath76 : topologies with only two variable nodes connected to two unsatisfied check nodes . , width=422 ] ets in a left - regular graph with @xmath65 and @xmath76 : the only possible topology with three variable nodes connected to two unsatisfied check nodes.,width=163 ] consider an @xmath14 ets @xmath50 in @xmath60 . let @xmath89 be an ets in @xmath60 of size @xmath90 . we say that @xmath50 is a _ layered superset ( lss ) _ of @xmath91 if there exists a nested sequence of etss : @xmath92 , such that @xmath93 has size @xmath94 for @xmath95 . when there is no risk of confusion , we also refer to @xmath5 as having the lss property . consider the ets @xmath96 of a left - regular ldpc code with @xmath65 shown in figure [ lss_ex]@xmath62 , and one of its subsets @xmath97 . ets in a left - regular graph with @xmath65 , ( b ) a @xmath98 ets in a left - regular graph with @xmath99.,width=508 ] it is easy to see that @xmath50 and @xmath100 are @xmath85 and @xmath101 etss . careful inspection of figure [ lss_ex]@xmath62 reveals that @xmath5 is not an lss of @xmath100 . set @xmath50 however , is an lss of @xmath102 with the following nested sequence of etss : @xmath103 . figure [ lss_ex]@xmath63 shows a @xmath98 ets @xmath5 in a left - regular code with @xmath99 . this set has lss property with respect to @xmath104 with the following nested sequence of etss : @xmath105 . one should note that any cycle in a tanner graph is an ets . in the sequel , we are particularly interested in the lss property of more complex etss with respect to short cycles of the graph . the following proposition is an example . all the @xmath74 etss of a left - regular ldpc code with @xmath65 and @xmath66 are layered supersets of any one of their 6-cycle subsets . [ prop2 ] based on proposition [ prop1 ] , there are only 3 non - isomorphic structures for @xmath74 etss in left - regular ldpc codes with @xmath65 and @xmath66 . these structures are shown in figure [ 6_2_g ] . it is not difficult to see that all three structures are lss of any of their 6-cycle subsets . for example , the structure @xmath50 in figure [ 6_2_g]@xmath63 is an lss of its 6-cycle subset @xmath106 with the following nested sequence of etss : @xmath107 . one advantage of lss property is that it corresponds to a simple algorithm for finding larger etss with lss property starting from one of their subsets in the nested sequence of etss . the basic step is explained in the following lemma . consider an ets @xmath108 of size @xmath109 . suppose that @xmath5 has an elementary trapping subset @xmath110 of size @xmath3 . then , the variable node @xmath111 is only connected to unsatisfied check nodes of @xmath112 ( at least two of them ) , i.e. , there is no connection between @xmath72 and the satisfied check nodes of @xmath112 . [ lemma1 ] the proof is simple and is based on the definitions of an ets and the set @xmath113 . the pseudo code of an algorithm corresponding to lemma [ lemma1 ] is given in routine 1 . ( 1,0)500 + * routine 1 : expansion of an ets @xmath50 of size @xmath3 to etss of size @xmath114 + routine ex = oneexpansion@xmath115 * + ( 1,0)500 * initialization : * @xmath116 . let @xmath54 and @xmath55 be the set of unsatisfied check nodes and satisfied check nodes of @xmath50 , respectively . let @xmath117 be the set of variable nodes which have at least two connections with the check nodes in @xmath54 and have no connection with the check nodes in @xmath55 . * output : * ex ( 1,0)500 based on lemma [ lemma1 ] , the set @xmath117 contains _ all _ the nodes that can be part of the expansion of the ets . complexity of routine 1 for expanding an @xmath2 ets to etss of size @xmath114 in a @xmath120 regular graph , where @xmath0 and @xmath121 are left and right degrees , respectively , is of order @xmath122 . this comes from the fact that for finding the larger trapping sets , the algorithm needs to check at most @xmath123 variable nodes as possible candidates . the memory requirement in this case is of order @xmath124 . it should be however noted that , imposing the condition of line 3 removes a large portion of neighboring variable nodes of @xmath50 from the set of possible candidates for expansion . this highly reduces the complexity of the algorithm . consider the case where an ets @xmath50 of size @xmath3 is an lss of an ets @xmath91 of size @xmath125 . clearly , starting from @xmath91 , the successive application of routine 1 will result in finding all the etss which are layered supersets of @xmath91 . in particular , ets @xmath50 will be among the outputs after @xmath126 applications of routine 1 . algorithm 1 contains the pseudo code of an algorithm that starts from a set of etss and finds _ all _ the etss of size up to @xmath29 that are layered supersets of the initial set of etss . ( 1,0)250 + * algorithm 1 * : expansion of input etss to etss of size up to @xmath29 in @xmath127 . + ( @xmath128 and @xmath129 are the lists of input and output trapping sets , respectively . ) + ( 1,0)250 * inputs : * @xmath30 , @xmath130 , @xmath29.*initialization : * @xmath131 } for @xmath132 . ex= oneexpansion@xmath115 @xmath133 ex . * output : * @xmath134 . it is easy to see that any trapping set in @xmath113 , including etss , contains at least one cycle . it can thus be argued that cycles are the most basic structure for etss in @xmath113 . this motivates the study of the relationship between cycles , as the most basic etss , and the more complex etss of @xmath113 . short cycles have long been known to be problematic for iterative decoding in general @xcite , and for the performance in the error floor region , in particular @xcite , @xcite , @xcite . short cycles are also easy to enumerate @xcite . in this section , we establish a simple relationship between short cycles and the more complex etss . in particular , we prove that an overwhelming majority of dominant etss of left - regular ldpc codes are layered supersets of short cycles . one important implication of this result is that , starting from short cycles of the graph , algorithm 1 , presented in the previous section , can be used to find all such etss in a _ guaranteed _ fashion . the following examples demonstrate the relationship between short cycles and the more complex etss of left - regular graphs . figure [ 51_3 ] shows a possible structure @xmath5 of a @xmath135 ets in a left - regular graph with @xmath99 and @xmath66 . the @xmath136 ets @xmath137 forms a 6-cycle and is a subset of @xmath5 . set @xmath5 is an lss of @xmath91 with the following nested sequence of etss : @xmath138 . starting from @xmath91 , in the first round of expansion by algorithm 1 , variable node @xmath79 will be added and the @xmath139 ets @xmath140 will be found . in the second round of expansion , with set @xmath140 as the input , variable node @xmath80 will be added and @xmath50 will be found . trapping set in a left - regular graph with @xmath99.,width=153 ] note that although @xmath80 is not part of any 6-cycle , this node will be added at the second round of expansion . in other words , for a trapping set to be found by the algorithm , it is not necessary that all the nodes in the set participate in short(est ) cycles . starting from a short cycle , each round of expansion by algorithm 1 , may result in several new etss . these trapping sets all have the same size , but may have different number of unsatisfied check nodes , and thus belong to different classes of trapping sets . this is demonstrated in the following example . consider the structure in figure [ 6_2_g]@xmath63 and one of its subsets @xmath141 which is a @xmath101 ets . starting from @xmath91 , as the input to algorithm 1 , there are two variable nodes @xmath42 and @xmath80 as candidates for the first expansion . the resultant etss are @xmath142 and @xmath143 , which are @xmath144 and @xmath145 trapping sets , respectively . [ 2b ] it is important to note that if an element of a class of etss is an lss of a short cycle , this does not necessarily mean that all the elements of that class are also lsss of short cycles of the same length . the reason is that there may be other non - isomorphic structures in that class which are not lsss of any of the short cycles under consideration . figure [ 66_4 ] shows two possible structures for a @xmath85 ets in a left - regular graph with @xmath65 and @xmath66 . the structure of figure [ 66_4]@xmath62 is an lss of any of its 6-cycles and thus can be obtained by the recursive expansion of any of them through algorithm 1 . the structure of figure [ 66_4]@xmath63 , on the other hand , is not an lss of any of its 6-cycles , and is thus out of the reach of algorithm 1 , if the algorithm starts from any 6-cycle . ets in a left - regular graph with @xmath65 and @xmath66.,width=268 ] it is however , easy to see that the structure in figure [ 66_4]@xmath63 is an lss of the set @xmath146 , which itself forms an 8-cycle . the trapping set can thus be found by algorithm 1 if cycles of length 8 are included in the input set . in general , adding short cycles longer than the girth to the input of algorithm 1 can improve the coverage of this algorithm . nonetheless , there are structures which do not satisfy the lss property with respect to any of their cycles . this means that even arbitrarily enlarging the size of the cycles in the initial input set will not result in finding such structures by the algorithm . figure [ 82_3 ] shows an @xmath147 ets @xmath50 in a left - regular graph with @xmath99 and @xmath66 . set @xmath50 contains cycles of length 6 , 8 and 10 . it is easy to check that @xmath50 does not satisfy the lss property with respect to any of these cycles , and thus can not be found using algorithm 1 starting from any of them . ets in a left - regular graph with @xmath99 and @xmath66 which is not an lss of any of its cycles.,width=268 ] in this part of the paper , we investigate the structure of etss for left - regular ldpc codes with left degrees 3 , 4 , 5 and 6 . for each category of codes , we consider girth values 6 and 8 , and study all the non - isomorphic structures of different classes of @xmath2 etss with values of @xmath3 and @xmath4 up to 10 . for each class of etss with given values of @xmath3 and @xmath4 , we first find all the non - isomorphic structures using the approach described in section iii ( with constraints on @xmath0 and @xmath1 ) . we then examine each of these structures to find out whether the structure is an lss of any of its cycles . this can be performed using algorithm 1 . let @xmath5 be the structure under consideration . we can start with the set of shortest cycles , say of length @xmath148 , in @xmath5 , and apply algorithm 1 to recursively expand them to larger subsets of @xmath5 . if this process ends with finding @xmath5 , then we report @xmath5 as being an lss of a cycle of length @xmath148 . in the case that this process will not result in finding @xmath5 , we use the set of cycles of next larger size as the input and repeat the process . this will continue until @xmath5 is identified as an lss of one of its cycles or until all the cycles are exhausted and @xmath5 is not an lss of any of them . in the former case , if the cycle length is @xmath149 , we refer to @xmath5 as an lss@xmath150 structure . the results for different values of @xmath0 and @xmath1 are reported in the following subsections . for each value of @xmath3 , we mostly consider the values of @xmath4 which satisfy @xmath151 . for a given value of @xmath3 , these values of @xmath4 are believed to correspond to dominant trapping sets @xcite . it is easy to see that for a tanner graph with girth @xmath152 , it is impossible to have any @xmath2 trapping set in @xmath60 with @xmath153 . for a tanner graph with @xmath76 and variable node degree @xmath0 , short cycles of length 6 are trivial @xmath154 trapping sets . for this reason , we only consider @xmath2 trapping sets with @xmath155 . for left - regular graphs with @xmath99 and @xmath66 , the multiplicity of non - isomorphic lss@xmath150 structures for different values of @xmath149 are listed in table [ d3g6 ] for different classes of etss in @xmath60 . [ thm1 ] -1 cm & & & & & & + & & & & & & + & & & & & & + & & & & & & + & & & & & & + & & & & & & + & & & & & & + & & & & & & + & & & & & & + [ d3g6 ] ( each row in table [ d3g6 ] corresponds to a specific size @xmath3 of an ets , and each column corresponds to a specific number of unsatisfied check nodes , @xmath4 . for each pair @xmath2 , table [ d3g6 ] lists the multiplicity of lss@xmath150 structures for different values of @xmath149 . for example , having @xmath157 for a specific class of etss means that there are 4 different non - isomorphic structures for that class of trapping sets : three of them are lss@xmath158 and one of them is an lss@xmath159 structure . starting with cycles of length @xmath160 and @xmath161 , algorithm 1 is thus guaranteed to find all such etss . having the symbol - " for a class of @xmath2 trapping sets means that for the underlying conditions , i.e. , @xmath99 and @xmath66 , it is impossible to have such a class of trapping sets . for the structures which do not satisfy the lss property with respect to any of their cycles , we use the notation na " ( stands for not applicable ) . starting from any set of cycles of the graph , algorithm 1 can not find such structures . ) in general , the results reported in table [ d3g6 ] for any @xmath2 ets can be proved by first obtaining all the non - isomorphic structures using the nauty program , as described in section iii , and then examining each such structure for the lss property , using algorithm 1 , as described earlier in this subsection . we however , provide a formal proof for the results pertaining to the ets class @xmath162 as well as all the classes that can not exist in left - regular graphs with @xmath99 and @xmath66 ( i.e. , those with designation - " in the table ) . the general approach for the formal proof of the rest of the results is similar and not provided . : we first prove that the two structures presented in figure [ 6_0_ng ] are the only possible non - isomorphic structures for @xmath163 etss in left - regular graphs with @xmath99 and @xmath66 . we use the normal graph representation of these structures , also shown in figure [ 6_0_ng ] , for the proof . etss in left - regular graphs with @xmath99 and @xmath66 , and their normal graph representations.,width=355 ] since the number of unsatisfied check nodes is zero , every node in the normal graph must be connected to three other nodes . starting from an arbitrary node , say @xmath78 , as the root , we grow the normal graph , and will have three nodes in the first layer . we arbitrarily denote these nodes by @xmath42 , @xmath79 , and @xmath80 . consider the case where the nodes in the first layer do not have any edges in common ( see figure [ ab_grow1]@xmath62 ) . in this case , they must have 6 edges connected to the two remaining nodes ( @xmath77 and @xmath81 ) . according to the girth constraint ( i.e. , @xmath76 ) , each pair of nodes can have at most one edge in common . this implies that the only possible scenario is the case where each of the three nodes in the first layer is connected to both @xmath77 and @xmath81 . this results in the structure of figure [ 6_0_ng]@xmath63 . note that in this case , the length of the shortest cycles in the structure is 8 . now , consider the case where the nodes in the first layer have some edge(s ) in common . based on the girth constraint ( @xmath66 ) , the two remaining nodes ( @xmath77 and @xmath81 ) must have at least 4 edges in common with the nodes in the first layer . ( otherwise , for @xmath77 and @xmath81 to have degree 3 , they need to have more than one edge in common , which contradicts the girth constraint . ) this implies that the nodes in the first layer can have only one edge in common . without loss of generality , we assume @xmath42 and @xmath79 are connected ( see figure [ ab_grow1]@xmath63 ) . from this point on , there is no choice in connecting the nodes . node @xmath80 must be connected to both @xmath77 and @xmath81 , and each of the nodes @xmath42 and @xmath79 must have one connection to @xmath77 or @xmath81 . ( switching the connections will result in isomorphic structures . ) this results in the structure of figure [ 6_0_ng]@xmath62 . etss of a left - regular graph with @xmath99 and @xmath66.,width=288 ] there are two cycles of length 6 in the structure @xmath164 of figure [ 6_0_ng]@xmath62 . however , @xmath164 is not an lss of any of them . there are also three cycles of length 8 in @xmath164 , and @xmath164 is an lss of all of them . for example , @xmath164 is an lss of @xmath165 with the following nested sequence of etss : @xmath166 . the structure @xmath167 of figure [ 6_0_ng]@xmath63 does not have any 6-cycles . it has four 8-cycles and has lss property with respect to all of them . : all such cases in table [ d3g6 ] can be proved using two facts : ( i ) an @xmath2 ets in a left - regular graph with @xmath99 has @xmath168 edges and @xmath169 satisfied check nodes . for the latter to be integer , there can not be any @xmath2 ets with an odd @xmath3 and an even @xmath4 , or vice versa . ( ii ) by definition , each variable node of an ets @xmath108 of a left - regular graph with @xmath99 can be connected to at most one unsatisfied check node . this means that it is not possible to have an @xmath2 ets with @xmath170 . the results of table [ d3g6 ] indicate that the majority of ets structures satisfy the lss property with respect to short cycles . this is particularly the case for smaller values of @xmath3 and @xmath4 which correspond to more problematic trapping sets . a cycle of length @xmath171 in a left - regular graph with @xmath99 is an @xmath172 ets . for example , figure [ 55_3_n ] shows the structure of a @xmath173 ets in a left - regular graph with @xmath99 , which is a cycle of length 10 . in fact , as the results of table [ d3g6 ] show cycles of length @xmath171 are the only @xmath172 etss for the graphs under consideration . such cycles trivially satisfy the lss property . ets in a left - regular graph with @xmath99 and @xmath66.,width=144 ] in figure [ n64_73_91_3_1 ] , the three structures corresponding to the `` na '' cases in table [ d3g6 ] for @xmath174 , @xmath175 and @xmath176 etss are presented . all three structures consist of two etss connected by a check node . for @xmath174 , @xmath175 and @xmath176 etss.,width=432 ] in the rest of the paper , we omit the proofs as they are similar in nature to that of theorem [ thm1 ] . for left - regular graphs with @xmath99 and @xmath178 , the multiplicity of non - isomorphic lss@xmath150 structures for different values of @xmath149 are listed in table [ d3g8 ] for different classes of etss in @xmath60 . & & & & & & + & & & & & & + & & & & & & + & & & & & & + & & & & & & + & & & & & & + & & & & & & + & & & & & & + & & & & & & + [ d3g8 ] the results of table [ d3g8 ] also indicate that the majority of ets structures satisfy the lss property with respect to short cycles of the graph . compared to the results of table [ d3g6 ] , a larger portion of ets classes have the property that all their non - isomorphic structures are lsss of short cycles . in a left - regular graph with @xmath179 , any ets in @xmath60 is also an absorbing set . so the results presented in tables [ d3g6 ] and [ d3g8 ] are also applicable to the absorbing sets of these graphs . for left - regular graphs with @xmath65 and @xmath66 , the multiplicity of non - isomorphic lss@xmath150 structures for different values of @xmath149 are listed in tables [ d4g6_1 ] and [ d4g6_2 ] for different classes of etss in @xmath60 . & & & & + & & & & + & & & & + & & & & + & & & & + & & & & + & & & & + & & & & + & & & & + & & & & + & & & & + & & & & + & & & & + & & & & + [ d4g6_1 ] & & + & & + & & + & & + & & + & & + & & + & & + & & + & & + & & + & & + & & + & & + [ d4g6_2 ] for graphs with @xmath180 , not every ets is an absorbing set . to identify the structures that are absorbing sets , we use the notation `` as '' in the tables . for each class of ets , we thus have two sets of entries : the ones for absorbing sets denoted by `` as '' , and those that correspond to all the etss denoted by `` ts '' to stand for `` trapping sets '' . for each entry , the number of possible non - absorbing set structures can be obtained by subtracting the corresponding results in the two sets . for example , the results for the ets class @xmath64 in table [ d4g6_1 ] indicate that there are two non - isomorphic structures for this class that are both lsss of cycles of length @xmath1 . only one of the two structures however , is an absorbing set . tables [ d4g6_1 ] and [ d4g6_2 ] indicate that an overwhelming majority of ets structures satisfy the lss property with respect to short cycles . in particular , all the etss of size less than 7 with less than 6 unsatisfied check nodes are lsss of 6-cycles . for left - regular graphs with @xmath65 and @xmath178 , the multiplicity of non - isomorphic lss@xmath150 structures for different values of @xmath149 are listed in table [ d4g8 ] for different classes of etss in @xmath60 . & & & & & & + & & & & & & + & & & & & & + & & & & & & + & & & & & & + & & & & & & + & & & & & & + & & & & & & + & & & & & & + & & & & & & + & & & & & & + & & & & & & + & & & & & & + & & & & & & + [ d4g8 ] the results of table [ d4g8 ] indicate that all the etss of size less than 10 with less than 9 unsatisfied check nodes satisfy the lss property . comparison with the results of tables [ d4g6_1 ] and [ d4g6_2 ] reveals that by increasing the girth of the graph from 6 to 8 , the number of classes whose structures all satisfy the lss property increases . for left - regular graphs with @xmath181 and @xmath66 , the multiplicity of non - isomorphic lss@xmath150 structures for different values of @xmath149 are listed in table [ d5g6 ] for different classes of etss in @xmath60 . -.6 cm & & & & & & + & & & & & & + & & & & & & + & & & & & & + & & & & & & + & & & & & & + & & & & & & + & & & & & & + & & & & & & + & & & & & & + & & & & & & + & & & & & & + & & & & & & + & & & & & & + & & & & & & + & & & & & & + & & & & & & + & & & & & & + & & & & & & + & & & & & & + [ d5g6 ] table [ d5g6 ] shows that except for a small fraction of ets structures , all the rest of the structures of ets classes with size less than 10 and with less than 10 unsatisfied check nodes satisfy the lss property . in particular , all the structures of size less than 10 with less than 5 unsatisfied check nodes are lsss of 6-cycles . for left - regular graphs with @xmath181 and @xmath178 , the multiplicity of non - isomorphic lss@xmath150 structures for different values of @xmath149 are listed in table [ d5g8 ] for different classes of etss in @xmath60 . & & & + & & & + & & & + & & & + & & & + & & & + & & & + & & & + & & & + & & & + & & & + & & & + & & & + & & & + [ d5g8 ] tables [ d4g8 ] and [ d5g8 ] , in comparison with their counterparts for @xmath66 , show that in graphs with larger girth , small trapping sets with small number of unsatisfied check nodes can not exist . in particular , table [ d5g8 ] indicates that in left - regular graphs with @xmath181 and @xmath178 no @xmath2 ets with @xmath182 and @xmath183 can exist . the inspection of table [ d5g6 ] however , reveals that many of these ets classes can exist in left - regular graphs with @xmath181 and @xmath66 . the results of table [ d5g8 ] also demonstrate that , for a left - regular graph with @xmath181 and @xmath178 , all the @xmath2 etss with @xmath184 and @xmath183 are lsss of 8-cycles . it can be proved that it is not possible to have any @xmath2 ets with @xmath184 and @xmath183 in a left - regular graph with @xmath181 and @xmath185 . for left - regular graphs with @xmath186 and @xmath66 , the multiplicity of non - isomorphic lss@xmath150 structures for different values of @xmath149 are listed in table [ d6g6 ] for different classes of etss in @xmath60 . & & & & & & + & & & & & & + & & & & & & + & & & & & & + & & & & & & + & & & & & & + & & & & & & + & & & & & & + & & & & & & + & & & & & & + & & & & & & + & & & & & & + & & & & & & + & & & & & & + & & & & & & + & & & & & & + & & & & & & + [ d6g6 ] the results of table [ d6g6 ] indicate that for left - regular graphs with @xmath186 and @xmath66 , all the possible @xmath2 absorbing sets with @xmath184 and @xmath187 satisfy the lss property with respect to 6-cycles . moreover , except for one structure ( out of 5411 structures ) of the ets class @xmath188 , all the structures of @xmath2 ets classes with @xmath184 and @xmath189 are lsss of short cycles . ( the structure that does not satisfy the lss property is shown in figure [ 910_6 ] . ) among these classes ( excluding the @xmath188 class ) , only one structure out of the 2274 possible structures for the @xmath190 ets class is an lss of 10-cycles . all the other possible structures for all the ets classes are lsss of 6-cycles . ets in a left - regular graph with @xmath186 and @xmath66 that does not satisfy the lss property . , width=240 ] for left - regular graphs with @xmath186 and @xmath84 , there does not exist any @xmath2 ets in @xmath113 with @xmath184 and @xmath187 . in this paper , we investigated the structure of elementary trapping sets ( ets ) of left - regular ldpc codes . we developed an approach to find _ all _ the non - isomorphic structures of a given @xmath2 class of etss , where @xmath3 is the size and @xmath4 is the number of unsatisfied check nodes of the ets . for left - regular ldpc codes with left degrees @xmath6 , and girths @xmath7 , we studied such structures and demonstrated that an overwhelming majority of them are layered supersets ( lss ) of short cycles in the tanner graph of the code . in particular , we proved that for any category of left - regular ldpc codes with given @xmath0 and @xmath1 , there exist integers @xmath125 and @xmath191 such that all the classes of @xmath2 etss with @xmath192 and @xmath193 , are lsss of short cycles . this implies that for any category of left - regular ldpc codes , the dominant etss are all lsss of short cycles . the lss characterization of dominant etss is particularly important as it corresponds to a simple algorithm that can find _ all _ such etss in a _ guaranteed _ fashion starting from the short cycles of the graph . for any class of @xmath2 etss , the lengths of the required short cycles were provided in this paper . one important contribution of this paper is the approach developed to exhaustively find all the non - isomorphic structures of a given class of @xmath2 etss for arbitrary values of @xmath3 and @xmath4 and for left - regular ldpc codes of arbitrary left degree and girth . in a more general context , the database of such structures can be very helpful in the analysis and the design of ldpc codes with low error floor . in particular , one can use this information to find all the etss of a certain class in a guaranteed fashion regardless of whether those etss satisfy the lss property or not . to the best of our knowledge , the results presented in tables [ d3g6 ] - [ d6g6 ] are the most comprehensive results available so far on the structure of etss of regular ldpc codes . s. abu - surra , d. declercq , d. divsalar , and w. ryan , trapping set enumerators for specific ldpc codes , " proc . theory and applications workshop , _ san diego , ca , jan . 31- feb . 5 , 2010 , pp . 1 - 5 . s. abu - surra , w. e. ryan , d. divsalar , and w. ryan , ensemble trapping set enumerators for protograph - based generalized ldpc codes , " proc . theory and applications workshop , _ san diego , ca , jan . 27feb . 1 2008 , pp . 6365 . e. cavus and b. daneshrad , a performance improvement and error floor avoidance technique for belief propagation decoding of ldpc codes , " proc . _ 16th ieee international symposium pers . indoor mobile radio communications , _ los angeles , ca , sept . 2005 , vol . 2386 - 2390 . qiuju diao , ying yu tai , shu lin , and khaled a. s. abdel - ghaffar , trapping set structure of finite geometry ldpc codes , " proc . _ ieee international symposium on information theory ( isit12 ) _ , 2012 , pp . 3088 - 3092 . l. dolecek , z. zhang , v. anantharam , m.wainwright , and b. nikolic , analysis of absorbing sets for array - based ldpc codes , " proc . _ interntional conf . communications , _ glasgow , scotland , jun . 24 - 28 , 2007 , pp . 6261 - 6268 . l. dolecek , z. zhang , m. wainwright , v. anantharam , and b. nikolic , evaluation of the low frame error rate performance of ldpc codes using importance sampling , " proc . _ ieee inform . theory workshop _ , lake tahoe , ca , sept . 2 - 6 , 2007 , pp . 202 - 207 . l. dolecek , p. lee , z. zhang , v. anantharam , b. nikolic , and m. wainwright , predicting error floors of ldpc codes : deterministic bounds and estimates , " _ ieee journal on selected areas in communications _ , vol . 27 , no . 6 , pp . 908 - 917 , aug . 2009 . l. dolecek , z. zhang , v. anantharam , m. j. wainwright , and b. nikolic , analysis of absorbing sets and fully absorbing sets of array - based ldpc codes , " _ ieee trans . inform . theory _ 181 - 201 , jan . 2010 . c. di , d. proietti , i. e. telatar , t. j. richardson , and r. l. urbanke , finite - length analysis of low - density parity - check codes on the binary erasure channel , " _ ieee trans . inform . theory _ 48 , no . 6 , pp . 1570 - 1579 , june 2002 . m. ivkovic , s. k. chilappagari , and b. vasic , eliminating trapping sets in low - density parity - check codes by using tanner graph covers , " _ ieee trans . inform . theory _ 3763 - 3768 , aug . 2008 . s. laendner and o. milenkovic , algorithmic and combinatorial analysis of trapping sets in structured ldpc codes , " proc . _ ieee international conference on wireless networks , communications and mobile computing _ , hawaii , usa , june 13 - 16 , 2005 , pp . 630 - 635 . s. laendner , t. hehn , o. milenkovic and j.b . huber , characterization of small trapping sets in ldpc codes from steiner triple systems , " proc . _ 6th international symposium on turbo codes & iterative information processing , _ brest , france , sept . 6 - 10 , 2010 , pp . 93 - 97 . c. koller , a. graell i amat , j. kliewer , and d. j. costello , on trapping sets for repeat accumulate accumulate codes , " proc . _ theory and applications workshop _ , san diego , la jolla , ca , feb . 8 - 13 , 2009 . c. koller , a. graell i amat , j. kliewer , and d. j. costello , trapping set enumerators for repeat multiple accumulate code ensembles , " proc . _ ieee international symposium on information theory ( isit09 ) _ , seoul , korea , jun . - jul . 2009 . g. b. kyung and c .- c . wang , finding the exhaustive list of small fully absorbing sets and designing the corresponding low error - floor decoder , " _ ieee trans . communications _ , vol . 60 , pp . 1487 - 1498 , june . 2012 . y. mao and a.h . banihashemi , a heuristic search for good low - density parity - check codes at short block lengths , " proc . _ ieee international conference on communications _ , helsinki , finland , june 2001 , pp . 41 - 44 . e. pusane , d. j. costello , and d. g.m . mitchell , trapping set analysis of protograph - based ldpc convolutional codes , " proc . _ ieee international symposium on information theory ( isit09 ) _ , seoul , korea , jun . - jul . 2009 , pp . 561565 . c .- c . wang , on the exhaustion and elimination of trapping sets : algorithms & the suppressing effect , " proc . _ ieee international symposium on inform . nice , france , june 24 - 29 , 2007 , pp . 2271 - 2275 . h. xiao and a. h. banihashemi , estimation of bit and frame error rates of low - density parity - check codes on binary symmetric channels , " _ ieee trans . 2234 - 2239 , dec . 2007 . h. xiao and a. h. banihashemi , error rate estimation of low - density parity - check codes on binary symmetric channels using cycle enumeration , " _ ieee trans . communications _ , vol . 1550 - 1555 , june 2009 . h. xiao , a. h. banihashemi and m. karimi , error rate estimation of low - density parity - check codes decoded by quantized soft - decision iterative algorithms , " _ ieee trans . communications _ , vol . 474 - 483 , feb . 2013

this paper presents a careful and detailed light curve analysis of bright detached eclipsing binaries ( deb ) in the small magellanic cloud , discovered by ogle collaboration , on the basis of recently available difference image analysis ( dia ) photometry . the 19 binaries brighter than 16.4 mag in band and with the depth of primary and secondary eclipse greater than 0.25 mag were investigated . the solutions were obtained by a modified version of the wilson - devinney program . the quality of dia light curves a good phase coverage and relatively small scatter is enough to calculate realistic estimates for the third light and the argument of periastron . it was found that solutions of detached , eccentric systems with flat light curve between eclipses usually may suffer from indetermination of in contrast to those of similar systems having some proximity effects . the physical properties of the stars were estimated on the basis of their photometric elements and indices assuming the distance modulus to smc and consistency between an empirical mass - luminosity relation and the flux scaling . the method was tested on three lmc stars of known absolute dimensions and a good agreement was found for . such an approach may give fast and accurate estimates of absolute dimensions for large and homogeneous samples of eclipsing binaries in the magellanic clouds and other close galaxies . moreover , this method allows also for independent estimation of in the direction to a particular binary . the subset of six bright deb s worth future intensive investigations as likely distance indicators to smc , was chosen . they are sc3 139376 , sc4 53898 , sc5 129441 , sc6 67221 , sc6 215965 and sc9 175336 . [ firstpage ] binaries : eclipsing magellanic clouds

astro-ph0303550

accurate determination of absolute dimensions of eclipsing binaries based , among others , on detailed light curve analysis is still a very important task of modern astronomy . it gives an opportunity for testing advanced evolutionary models of stars ( e.g. pols et al . 1997 ) but also allows for very precise distance determination . the method , at least in principle , is quite straightforward ( see paczyski 1997 for an outline and clausen 2000 for a review ) . three distance determination for eclipsing binaries in the large magellanic cloud were quite recently presented : hv 2274 ( guinan et al . 1998 ) , hv 982 ( fitzpatrick et al . 2001 ) and eros 1044 ( ribas et al . 2002 ) giving the distance modulus @xmath5 . light curve solutions together with estimation of physical parameters and the distance to the small magellanic cloud were obtained also for two eclipsing binaries : hv 2226 ( bell et al . 1991 ) and hv 1620 ( pritchard et al . 1998 ) giving for the distance modulus @xmath6 . very recently ten eclipsing binaries have been analysed in the central part of the smc giving mean distance modulus @xmath7 ( harries , hilditch & howarth 2003 ) . the light curve analysis in their paper was based on dophot ogle photometry ( udalski et al . 1998 ) . in recent years , extensive ccd photometry have been undertaken during the projects of microlensing searching toward the magellanic clouds and galactic bulge . as a byproduct , thousands of eclipsing binaries were discovered and placed in catalogs ( grison et al . 1995 , alcock et al . 1997 , udalski et al . preliminary solutions for macho lmc catalog were found by alcock et al . they fitted 611 light curves of eclipsing binaries using the ebop code ( etzel 1993 ) . that code uses a tri - axial ellipsoid approximation what makes this code very fast . but the code can not properly account for the proximity effects present in most of the eclipsing binaries detected by macho . for smc eclipsing binaries ogle collaboration ( udalski et al . 1998 ) extracted from their catalog a sample of 153 detached binaries with well defined , narrow eclipses of similar depth and relatively good photometry . it was argued that the sample contains the best systems for distance determination . wyithe & wilson ( 2001 , 2002 hereafter ww1 , ww2 ) presented an excellent analysis of the whole ogle data containing 1459 eclipsing binaries using an automated version of the wilson - devinney ( 1971 , wilson 1992 ) code . this program can provide very detailed treatment of proximity effects . they used public domain dophot photometry available at ogle homepage _ http://sirius.astrouw.edu.pl/ogle/ogle2/var@xmath8stars/smc/ecl _ in the analysis . they extracted in an objective way two subsets of detached binaries with very narrow eclipses and complete eclipses as the most likely good distance indicators . subset of semi - detached eclipsing binaries was also chosen as distance indicators on the basis of similar surface brightness between components and complete eclipses . the authors intend , also , to extract such subsets for contact binaries . in general it is believed that the close - to - ideal system chosen as a distance indicator , should have two stars of similar temperature and luminosity and the separation of the components should be significantly larger than their radii . but well - detached , eccentric systems may suffer from the aliasing of photometric solutions . ww2 argued that semi - detached and contact binaries , and also detached eclipsing binaries showing larger proximity effects , have their photometric solutions better defined and may serve as potential best distance indicators as in the case of recently analyzed 15-th magnitude binary eros 1044 ( ribas et al . 2002 ) . in this paper i focused my attention only on detached eclipsing binaries . the release of new dia photometry by ogle collaboration motivated me to reanalyse the light curves of eclipsing binaries in smc for the purpose of extracting a sample of the best distance indicators . this sample was chosen using additional important criteria which were not used in the selections made by wyithe & wilson . the analysis was restricted only for bright systems . it is obvious that those systems have , in general , the largest probability of obtaining high - quality light curves and radial velocity curves . thus the searching for distance indicators should start from the analysis of the brightest candidates . moreover using dia photometry we can achieve more realistic determination of the light curve parameters than in the case of using the older dophot photometry . ccd differential photometric observations of smc eclipsing binaries were obtained by ogle between 1997 and 2000 . the data were acquired using 1.3 m warsaw telescope at las campanas observatory . the photometric @xmath0 indices were calculated using the technique of difference image analysis ( dia ) as was reported in zebru , soszyski & woniak ( 2001a ) . the new photometry was quite recently facilitated on ogle homepage _ http://sirius.astrouw.edu.pl/ogle/ogle2/dia _ ( zebru et al . the ogle database contains all variables detected in the lmc and smc ( 68 thousand ) during the course of the ogle - ii project . the dia is argued as a superior method over classical dophot method because of its better accuracy , especially in crowded fields ( zebru et al . the light curves based on ogle - dia photometry are characterized by considerably smaller scatter of observations and higher smoothness of the light curves . also , in most cases the phase coverage is better because the dia photometry includes ogle observations which have extended almost two years after the moment when original catalog of smc eclipsing binaries ( udalski et al . 1998 ) was published . the quality of ogle - dia photometry can be demonstrated by a comparison of the @xmath0 light curves obtained with both methods . as an example i present a 17-th magnitude eccentric eclipsing binary smc sc1 25589 fig . [ fig:25589 ] together with preliminary light curve solutions . note that according to the analysis of dia light curve the system has apparently total eclipses and much larger separation between components than in the case of the analysis of dophot data . the candidates for the analysis were chosen after the visual inspection of the light curves of eclipsing binaries in smc released by udalski et al . a preliminary set of about 25 binaries were selected brighter than @xmath9 mag at maximum brightness and with the depth of the primary and secondary eclipses greater than 0.25 mag . the candidates fall into three separate groups . the first group , hereafter called a , consist of well - detached , eccentric systems with , usually , deep narrow eclipses . the second group , hereafter called b , comprise the short - period systems with apparently circular orbits and the small separation between components . as a result , the systems from group b show quite large proximity effects in their light curves . the third group c comprises well separated long period systems with circular orbits . as we expect long period systems to be , in general , eccentric this is most probably a chance selection . @|l|c|c|c|c|c|c|c|@ object & @xmath10 & @xmath11 & period[period ] . ] & epoch 3.1 ] & @xmath0 & @xmath12 & @xmath13 + & ( h : mm : ss ) & ( deg : mm : ss ) & ( days ) & ( @xmath14 ) & & & + + sc3 139376 & 0:44:08.69 & @xmath1573:14:17.9 & 6.05213 & 629.07158 & 14.482 & @xmath150.165 & @xmath150.126 + sc5 129441 & 0:49:40.56 & @xmath1573:00:22.5 & 8.05104 & 474.20881 & 16.012 & @xmath150.105 & @xmath150.072 + sc5 311566 & 0:51:34.84 & @xmath1572:45:45.9 & 3.29132 & 468.56556 & 16.032 & @xmath150.120 & @xmath150.062 + sc6 11143 & 0:51:39.69 & @xmath1573:18:44.8 & 5.72606 & 471.95035 & 16.251 & @xmath150.138 & @xmath150.092 + sc6 67221 & 0:52:06.22 & @xmath1572:45:14.3 & @xmath16 & 466.88188 & 15.327 & @xmath150.101 & @xmath150.095 + sc6 221543 & 0:53:40.39 & @xmath1572:52:22.0 & 3.41680 & 468.59872 & 16.089 & @xmath150.218 & @xmath150.109 + sc6 272453 & 0:54:33.22 & @xmath1573:10:39.6 & 5.70970 & 469.30090 & 15.805 & @xmath150.172 & @xmath150.057 + sc9 163575 & 1:02:46.27 & @xmath1572:24:44.2 & 1.97109 & 629.49776 & 15.181 & @xmath150.248 & @xmath150.221 + sc10 37223 & 1:03:41.42 & @xmath1572:03:06.9 & 2.58124 & 630.63675 & 16.336 & @xmath150.197 & @xmath150.164 + + sc4 53898 & 0:46:32.79 & @xmath1573:26:39.4 & 1.74107 & 624.86290 & 16.201 & @xmath150.128 & @xmath150.129 + sc4 103706 & 0:47:25.50 & @xmath1573:27:16.7 & 1.35585 & 623.94692 & 15.416 & @xmath150.195 & @xmath150.208 + sc4 163552 & 0:47:53.24 & @xmath1573:15:56.5 & 1.54581 & 625.35629 & 15.774 & @xmath150.046 & @xmath150.006 + sc5 38089 & 0:49:01.85 & @xmath1573:06:06.9 & 2.38941 & 469.77239 & 15.261 & @xmath150.238 & @xmath150.111 + sc6 215965 & 0:53:33.35 & @xmath1572:56:24.1 & 3.94603 & 473.26734 & 14.133 & @xmath150.234 & @xmath150.190 + sc6 319966 & 0:54:23.49 & @xmath1572:37:23.0 & 1.06466 & 466.68981 & 16.363 & @xmath150.151 & @xmath150.105 + sc8 104222 & 0:58:25.08 & @xmath1572:19:09.8 & 1.56972 & 629.30952 & 16.042 & @xmath150.144 & @xmath150.146 + sc9 175336 & 1:02:53.40 & @xmath1572:06:43.2 & 2.98654 & 628.36738 & 14.867 & @xmath150.222 & @xmath150.192 + + sc4 192903 & 0:48:22.69 & @xmath1572:48:48.6 & @xmath17 & 761.00300 & 16.242 & @xmath180.771 & @xmath180.956 + sc8 201484 & 1:00:18.05 & @xmath1572:24:07.5 & @xmath19 & 641.56700 & 14.230 & @xmath180.883 & @xmath180.977 + afterwards , the photometric data for each selected star were extracted , using the coordinates as identifier . some stars were not found in ogle - dia database because they probably passed undetected through the variability checking filters ( i. soszyski 2002 , private communication ) . the table [ tab : candidates ] gives the names and identification of 19 selected systems together with a summary of photometric parameters . the observations were folded with trial periods following the method given by kaluzny et al . the searches were usually restricted to a period close to the period given by udalski et al . the final period was found by tuning to the shape of the minima . it was found that in most cases the resulting periods agreed very well with those given by udalski et al . however , for sc6 67221 , i have found a new period resulting from observations of both eclipses . this binary turned out to be a very eccentric system with the orbital period of 10.977 days . the period is close to a whole number of 11 days and this is the reason for spurious secondary eclipse detection in the original ogle catalog . [ fig:67221com ] presents the light curve based on the proper period and , for comparison , the light curve from the original ogle catalog . two eccentric binaries sc3 139376 and sc5 311566 show small changes in the light curve most probably due to the apsidal motion . in both cases the tuning to the shape of the primary and the secondary minimum gives significantly different periods see as an example sc5 311566 ( fig.[fig:311566com ] ) . smooth light curves were obtained assuming that the change of the shape of minima is small and the secondary minimum was progressively shifting in respect to the primary minimum during the period of observations . the linear correction was applied to all phases corresponding to the secondary eclipse ( phases : 0.35 - 0.55 for sc3 139376 and 0.4 - 0.6 for sc5 311566 ) in the form : @xmath20 where @xmath21 and @xmath22 are the phase and the epoch number of the @xmath23-th observation . denoting the period resulting from the tuning to the primary eclipse by @xmath24 and to the secondary one by @xmath25 , the factor @xmath26 was calculated by formula : @xmath27 and was @xmath28 and @xmath29 for sc3 139376 and sc5 311566 , respectively . the old reference epochs @xmath30 given by udalski et al . ( 1998 ) were used to fold up the observations . new reference epochs were determined later during the light curve analysis see @xmath31 3.1 . for each star i calculated out - of - eclipse @xmath0 magnitude at the maximum of the brightness . the dia data does not include the @xmath32 and @xmath33 photometry . for the purpose of obtaining the most accurate @xmath12 and @xmath13 colours i recalculated the out - of - eclipse @xmath32 , @xmath33 and @xmath0 magnitudes using the old dophot photometry . note , that the magnitudes which i have found refer to the mean magnitudes at the maximum brightness rather than to the maximal magnitudes given in the original ogle catalog . resulting out - of - eclipse @xmath0 magnitudes were then used to normalize the light curves . in two cases ( sc6 11143 and sc9 163575 ) the light curve shows considerable amount of scatter , much larger than the average error of the observations . the light curves of both systems are presented on fig . [ fig : blends ] ( upper panel ) . i have removed observational points when the eclipses occurred and rerun the period finding program to find some periodicity in the scatter . it turns out that in the case of sc6 11143 the out - of - eclipse observations assemble with the period of 2.618 days and form a light curve with two well - defined , narrow minima of different depth ( middle panel of fig . [ fig : blends ] ) . it is likely that the binary sc6 11143 blends with another short period and slightly eccentric eclipsing binary . the similar analysis done for the system sc9 163575 shows that the out - of - eclipse observations assemble with the period of 1.985 days and form two minima separated by a half of the period . it seems that the system sc9 163575 blends also with a short period eclipsing binary . their periods are very close to each other and in the original light curve the points , corresponding to the minima of the blend , form two clumps situated around the minima of sc9 163575 . the light curves of both systems were cleaned by removing from the light curve all points corresponding to the moments when the eclipses occurred in the blended binary . the resulting light curves are presented in the lower panel of fig . [ fig : blends ] . also the system sc5 38089 in spite of its relatively high brightness ( @xmath34 ) has quite large scatter of observations . the analysis done for this system shows that the changes of the brightness are randomly distributed and thus , it is likely that the reason of the scatter is some kind of an intrinsic variability . + + + + all fits to the light curves were obtained using a modified version of the wilson - devinney(1971 , wilson 1979 , 1992 , hereafter wd ) program with a model atmosphere procedure based on the atlas9 code and developed by milone , stagg & kurucz ( 1992 ) . an automated iterations scheme of wd program was employed . for a subset a detached configuration ( mode 2 ) was chosen for all solutions and simple reflection treatment was used ( mref=1 , nref=1 ) . in the case of b subset both detached and semi - detached ( mode 5 ) configurations were used to obtain solutions . for b and c subsets the detailed reflection treatment with a double reflection was taken into account ( mref=2 , nref=2 ) . for all stars @xmath35 was fixed except for stars from c subset for which @xmath36 was assumed . i have found the initial guess via trial - and - error fitting of the light curves . the convergence is defined to have been achieved if the parameters corrections given by wd program are smaller than 0.75 times the standard errors ( 1.0 in the case of eccentricity and argument of periastron ) on two consecutive iterations . the adjusted parameters were the surface potentials ( @xmath37 ) , the effective temperature of the secondary ( @xmath38 ) , the luminosity of the primary ( @xmath39 ) , orbital inclination ( @xmath23 ) and the zero epoch offset ( @xmath40 ) . for a subset i have adjusted two more parameters : eccentricity ( @xmath41 ) and the argument of periastron ( @xmath2 ) . if for an eccentric system the convergence was not achieved i suppressed the adjustment of @xmath41 , @xmath42 and @xmath40 provided the correction were smaller than the standard error for these parameters on three consecutive iterations . it turned out that such procedure allowed for better convergence and the solution was achieved in all cases . afterwards i have run additional iterations , allowing all parameters to vary , to find error estimates of the final solution . similar procedure was employed for the systems from b and c subsets , just fixing the circular orbit and the suppressing included only the zero phase offset . among the systems in b subset the third light ( @xmath1 ) adjustment was needed to get the convergence in most cases . for all systems correction @xmath43 to the epoch given by udalski et al . ( 1998 ) was applied in the form : @xmath44 , where @xmath40 is the zero epoch offset of the final solution and @xmath45 is the period of the system . ww1 introduced a quantity @xmath46 which is greater than unity for systems with complete eclipses : @xmath47 where @xmath48 is the fractional the radius of larger star and @xmath49 is the fractional radius of the smaller star , both expressed in units of semi - major axis @xmath50 . @xmath46 is valid , in the form presented above , for circular orbits or perpendicular orbits ( @xmath23=90@xmath51 ) and also for eccentric orbits provided the fractional radii of components are expressed in units of the distance between stars _ at the moment _ of the mid - eclipse . in general case of eccentric orbits , if one eclipse occurs in the neighbourhood of periastron and is complete , the other may be partial and in the extreme cases , there may be an eclipse near periastron and none near apastron . the instantaneous distance between components @xmath52 , in units of semi - major axis @xmath50 , is given by the equation : @xmath53 where @xmath54 is the true anomaly . the primary and secondary minima occur when the true anomalies @xmath55 and @xmath56 are ( neglecting higher powers of @xmath41 , see kopal 1959 , chapter 6 , equation ( 9 - 22 ) ) : @xmath57 respectively . substituting @xmath55 and @xmath56 into eq . [ distance ] the distances were calculated at the moments of both mid - eclipses . then appropriate quantities @xmath58 and @xmath59 referring to the primary and the secondary eclipse , respectively , were computed . semidetached configuration ( mode 5 ) were investigated for all stars from the b subset in order to check which systems are definitively separated . when the semidetached configuration was investigated the mass ratio ( @xmath60 ) of the components was adjusted and secondary surface potential ( @xmath61 ) was set by @xmath60 . however , it turned out that no solution could be found in this mode for any star from b subset in spite of many tests for different sets of adjustable parameters . accordingly , in subsequent part of the paper i will refer only to detached solutions found in mode 2 . @|l|c|c|c|c|c|c|c|c|c|c|@ & radius & radius & temperature & inclination & eccentricity & argument of & third & luminos . & + object & @xmath62 & @xmath63 & ratio @xmath64 & @xmath23 & @xmath41 & periastron @xmath42&light @xmath1 & ratio @xmath65 & @xmath58 & @xmath59 + + sc3 139376@xmath66 & 0.177@xmath670.004 & 0.194@xmath670.004 & 1.007@xmath670.006 & 83.3@xmath670.2 & 0.064@xmath670.003 & 133.7@xmath675.7 & 0.0&1.191 & 0.734 & 0.704 + sc5 129441@xmath66@xmath68 & 0.158@xmath670.001 & 0.162@xmath670.002&0.999@xmath670.001 & 85.2@xmath671.0&0.375@xmath670.001&172.0@xmath670.4 & @xmath690.0@xmath70 & 1.079 & 0.797 & 0.773 + sc5 311566 & 0.157@xmath670.004 & 0.147@xmath670.002 & 0.962@xmath670.005 & 85.7@xmath670.6 & 0.251@xmath670.004 & 070.0@xmath670.1 & 0.0 & 0.843 & 0.841 & 0.721 + sc6 11143@xmath68 & 0.133@xmath670.004 & 0.112@xmath670.002 & 0.972@xmath670.020 & 88.8@xmath670.1 & 0.263@xmath670.001 & 234.3@xmath670.6 & 0.30@xmath670.05 & 0.704 & 0.983 & 1.022 + sc6 67221@xmath66 & 0.131@xmath670.002 & 0.119@xmath670.002&1.154@xmath670.011 & 83.9@xmath670.3&0.401@xmath670.002&043.5@xmath670.9 & @xmath690.0@xmath70 & 1.058 & 0.757 & 0.533 + sc6 221543 & 0.202@xmath670.006 & 0.182@xmath670.005 & 0.960@xmath670.007 & 84.2@xmath670.7 & 0.099@xmath670.002 & 146.1@xmath672.7 & 0.0 & 0.745 & 0.794 & 0.764 + sc6 272453 & 0.149@xmath670.004 & 0.180@xmath670.004 & 0.961@xmath670.006 & 85.2@xmath670.6 & 0.035@xmath670.004 & 236.0@xmath674.5 & 0.0 & 1.392 & 0.815 & 0.831 + sc9 163575 & 0.234@xmath670.004 & 0.225@xmath670.004 & 0.891@xmath670.007 & 80.6@xmath670.4 & 0.043@xmath670.002 & 010.7@xmath677.4 & 0.10@xmath670.05 & 0.762 & 0.661 & 0.655 + sc10 37223 & 0.202@xmath670.004 & 0.195@xmath670.004 & 0.951@xmath670.006 & 83.8@xmath670.3 & 0.087@xmath670.003 & 142.3@xmath672.4 & 0.0 & 0.874 & 0.757 & 0.728 + + sc4 53898@xmath66 & 0.189@xmath670.001 & 0.356@xmath670.002 & 0.910@xmath670.006 & 85.7@xmath671.2 & 0.0 & 090.0 & 0.187@xmath670.012 & 3.035 & + sc4 103706@xmath68 & 0.350@xmath670.002 & 0.284@xmath670.006 & 0.964@xmath670.004 & 86.8@xmath672.0 & 0.0 & 090.0 & 0.382@xmath670.014 & 0.612 & + sc4 163552@xmath68 & 0.338@xmath670.006 & 0.326@xmath670.008 & 1.004@xmath670.008 & 85.4@xmath670.9 & 0.0 & 090.0 & 0.263@xmath670.007 & 0.951 & + sc5 38089 & 0.246@xmath670.009 & 0.255@xmath670.007 & 0.988@xmath670.003 & 76.8@xmath671.9 & 0.0 & 090.0 & 0.0 & 1.053 & + sc6 215965@xmath66@xmath68 & 0.277@xmath670.003 & 0.335@xmath670.002&0.922@xmath670.001 & 75.6@xmath670.3&0.0&090.0 & @xmath690.0@xmath70 & 1.356 & + sc6 319966@xmath68 & 0.339@xmath670.006 & 0.324@xmath670.008 & 1.007@xmath670.007 & 88.3@xmath673.0 & 0.0 & 090.0 & 0.462@xmath670.005 & 0.861 & + sc8 104222 & 0.311@xmath670.011 & 0.331@xmath670.009 & 0.993@xmath670.008 & 86.8@xmath671.4 & 0.0 & 090.0 & 0.087@xmath670.013 & 1.129 & + sc9 175336@xmath66 & 0.194@xmath670.002 & 0.358@xmath670.002 & 0.965@xmath670.005 & 88.7@xmath670.8 & 0.0 & 090.0 & 0.174@xmath670.010 & 3.228 & + + sc4 192903 & 0.111@xmath670.004 & 0.145@xmath670.003 & 0.927@xmath670.004 & 87.2@xmath670.7 & 0.0 & 090.0 & 0.0 & 1.165 & + sc8 201484 & 0.171@xmath670.003 & 0.224@xmath670.002 & 0.888@xmath670.002 & 78.6@xmath670.2 & 0.0 & 090.0 & 0.0 & 1.107 & + the stars chosen in this paper as likely distance indicators are marked by an asterisk . the stars indicated by wyithe & wilson ( 2001 , 2002 ) are marked by a dagger . a double dagger means that the binary has most probably no third light contribution . in the analysis presented here i have included the temperature of binary s components resulting from an intrinsic colour @xmath71 of the system . all @xmath72 colours from table [ tab : candidates ] were dereddened using the average reddening toward smc @xmath73 ( massey et al . 1995 , udalski 2000 ) . i set the temperature of the primary ( @xmath74 ) respectively to the @xmath71 colour of the system using flower s ( 1996 ) empirical calibration . almost all the selected binaries have minima of similar depth what indicates that the difference between the temperatures of the components is small . thus such a procedure is justifiable . for those stars for which @xmath75 colour suggested a very high temperature ( @xmath76 k for sc5 38089 , sc6 215965 , sc9 163575 , sc9 175336 ) i set up @xmath77 k. i used a logarithmic law for the limb darkening ( e.g. van hamme 1993 ) . the @xmath0 limb darkenings @xmath78 and bolometric limb darkenings @xmath79 were calculated from van hamme s ( 1993 ) tables of limb darkenings for appropriate @xmath80 and @xmath81 . during the iterations the secondary s limb darkening coefficients were calculated according to its temperature @xmath38 . the bolometric albedo and gravity brightening exponent were set to the canonical value of 1.0 for stars with radiative envelopes ( von zeipel 1924 ) . however , for two red long - period systems sc4 192903 and sc8 201484 values typical for convective envelopes were assumed bolometric albedo was set to 0.5 and gravity brightening exponent to 0.32 . the synchronous rotation of both components was assumed ( @xmath82 ) . no level depending weighting of observations was applied parameter noise was set to 0 . the mass ratio was assumed to be @xmath83 in all calculations . ww2 pointed out that in the case of detached eclipsing binary the light curve is almost insensitive to the mass ratio parameter . only if one of the components is highly evolved ( when its radius @xmath84 increase to about a half of the total separation @xmath50 ) or if the system has a very low mass ratio , the light curve becomes sensitive for this parameter . the preliminary tests show that no component from the set of binaries has fractional radius @xmath85 larger than 0.4 . however , for two systems from b subset it was necessary to adjust the mass ratio in order to get a good solution ( see 3.4 ) . for each system from a group the automated fitting scheme described in 3.1 was employed . only in three cases : sc6 67221 , sc6 221543 and sc6 272453 was the convergence achieved by adjusting simultaneously all free parameters . the figure [ fig : fita ] shows the best - fitting model light curves , together with @xmath86 residuals , whose photometric parameters are given in table [ tab : sola ] . all arguments of the periastron refer to the epoch 1999.0 . for all systems from the a subset reasonable solutions were found assuming @xmath87 . however , in the case of sc6 11143 and sc9 163575 we expect that some amount of the third light is present in the system because of the blending with another eclipsing variable . in order to estimate third light component an additional grid of solutions for different values of @xmath1 was calculated in the range of 0.05 - 0.40 with a step of 0.05 . for sc6 11143 the best solution was found at @xmath88 , while for sc9 163575 at @xmath89 . it is worth noting that the solutions found with @xmath90 for these systems have almost no systematic trends in o - c s residuals , thus the intrinsic variability of the blend is the only sign of the third light component . the solutions for sc6 67221 converged very rapidly and had the smallest residuals . only slight systematic residuals are visible near phase 0.04 ( just after the egress from primary minimum ) , most probably due to the strong mutual reflection effect . similar but even stronger systematic effect is visible near first quadrature ( @xmath91 ) in sc5 129441 . because of the presence of the proximity effects both systems were additionally investigated in order to estimate possibly the third light contribution by adjusting @xmath1 as a free parameter . in each case synthetic light curves produce a tiny improvement only for unphysical values @xmath1@xmath920 , but such solutions were invariably unstable . the conclusion is that these binaries show virtually insignificant third light contribution what is consistent with @xmath90 . for a brief discussion of both systems see 5.1.2 . small residuals and relatively quick convergence characterize also sc3 139376 and sc9 163575 , whereas , the largest residuals are visible in sc5 221543 mostly due to a big intrinsic scatter in observations . having calculated @xmath41 and @xmath42 we could estimate the rate of apsidal motion for two binaries from the sample . the change of displacement of minima @xmath93 of an eccentric binary with central eclipses ( @xmath23=90 ) is given by ( see e.g. kopal 1959 , chapter 6 , equations ( 9 - 8 ) , ( 9 - 9 ) ) : @xmath94 however for our purpose we can still use this formula also for inclination close to 90provided the timescale of a displacement change is small with respect to an apsidal period @xmath95 . the change of displacement per one orbital revolution ( @xmath45 ) is simply @xmath26 ( eq . [ shift ] ) so the rate of advance of periastron is approximated by : @xmath96 . \label{rate}\ ] ] substituting numerical values we get 58/yr and 18/yr for sc3 139376 and sc5 311566 , respectively . + + + + + + + + + + + + the photometric parameters for the stars from b and c subsets are presented in table [ tab : sola ] . appropriate light curve solutions are shown in figs . [ fig : fitb ] and [ fig : fitc ] . in contrast to the a subset , almost all solutions from b subset ( beside sc6 215956 ) were found only by adjusting the third light @xmath1 in the system . there is a simple explanation of this effect - see a discussion in section [ l3 ] . sc6 215956 was also investigated for possible third light contribution but values of @xmath1 suggested by wd program were invariably negative . sc6 215965 has very small residuals and the solution for this system converged rapidly . also it seemed to be most stable of all in b subset . the solution for sc5 38089 was obtained using the spectroscopic mass ratio @xmath97 and more appropriate temperatures ( @xmath98 k ) determined for this binary by harries et al . third light contribution in this system is spurious although some improvements to the fits can be obtained for @xmath99 . in spite of the small residuals in sc9 175336 its solution was extremely unstable and exhibited large systematic @xmath86 deviations , especially near external contacts . the tests showed that the reason was the sensitivity of the light curve to the mass ratio @xmath60 as the larger and cooler component seemed to be noticeably more evolved than the hotter one . the same problems arose for sc4 53898 which seems to be similar in many respects to the former , although its light curve is less sensitive for @xmath60 . both systems show complete eclipses which allow for more accurate determination of photometric parameters including the photometric mass ratio . thus i allowed for adjusting of @xmath60 for both systems in detached configuration . the convergence was achieved and a stable solution obtained : in the case of sc9 175336 for @xmath100 and in the case of sc4 53898 for @xmath101 . the solutions for two systems from c subset converged very rapidly . no third light @xmath1 was adjusted in both cases because of the lack of any systematic residuals in the solutions ( fig . [ fig : fitc ] ) . moreover the solution for sc8 201484 has the smallest residuals from all of the subsets . we can not directly calculate absolute dimensions and masses without knowing spectroscopic orbit . however it is possible to estimate stellar parameters using the inversion of the method of parallaxes of eclipsing binaries . the method of parallaxes was elaborated by dworak ( 1974 ) who based it on gaposchkin ( 1938 ) algorithm . we assume that we _ know _ distance to smc and we use it to calculate absolute dimensions by scaling the system to obtain the observed flux at earth . the calculations are quite straightforward . for early b or o type stars the rayleigh - jeans approximation may be used in visible and near - infrared region of the spectrum so total @xmath33 or @xmath0 luminosity of each star is a linear function of temperature @xmath102 . using this approximation we can disentangle the temperatures @xmath74 and @xmath38 of both components : @xmath103 where @xmath104 is the luminosity ratio of the components known from the light curve solution ( table [ tab : sola ] ) and @xmath105 is the `` mean '' temperature corresponding to the total @xmath75 colour of the system . the temperature @xmath38 is then directly calculated from the temperature ratio . initially @xmath75 was found for each binary assuming total , foreground and internal , mean reddening to smc @xmath73 ( massey et al . the foreground reddening of @xmath106 was assumed ( schlegel , finkebeiner & davis 1998 ) with the ratio of the selective to the total extinction @xmath107 , while for the internal reddening i assumed @xmath108 ( bouchet et al . the resulting temperatures @xmath109 were used to calculate approximate intrinsic colours @xmath75 and bolometric corrections @xmath110 of both components via flower s ( 1996 ) calibration for main sequence stars . the `` standard '' mean value @xmath111 of the distance modulus ( @xmath112 ) for smc was assumed . the semi - major axis @xmath50 can be then computed by the following formula : @xmath113 where @xmath114 is the @xmath33 absolute magnitude of the sun ( assumed + 4.75 ) , @xmath115 is the out - of - eclipse @xmath33 magnitude of the primary ( calculated assuming @xmath116 and that @xmath1 contribution in @xmath33 band is comparable to that in @xmath0 band ) . the total mass of the system results simply from application of kepler s third law . the absolute scale of the system gives immediately the bolometric luminosity @xmath117 of each component . now we proceed to the second step of the method . we can represent the luminosity of the star as a function of its mass , which is well - known as eddington s relation . this relation works only for main sequence stars , but it should be kept in mind that for an early spectral type star ( massive and luminous at the same time ) its luminosity is not a strong function of the age . the dependence is rather weak and such star tends to evolve from zams along almost horizontal line on hr diagram . moreover , no binary from the sample ( beside two red systems from c subset ) seems to contain considerably evolved components . [ fig : law ] presents the mass - luminosity relation . the data for massive ( @xmath118 ) galactic stars were taken from compilation of detached , double lined eclipsing binaries done by pols et al . ( 1997 and references therein ) and for three lmc eclipsing systems from ribas et al . ( 2000 ) , fitzpatrick et al . ( 2001 ) and ribas et al . ( the inclusion of lmc stars gives an opportunity to check how this relation depends on the metallicity . indeed the galactic stars seems to have a little smaller luminosity then lmc stars ( having [ fe / h]@xmath119 ) at a given mass . for a galactic star the mass - luminosity @xmath120 relation may be approximated by a simple linear function in wide range of masses ( @xmath121 ) : @xmath122 i.e. @xmath123 . the sample of lmc stars do not cover such wide range of masses and i assumed that the slope of the @xmath120 relation is the same like for the galactic stars . then the data can be fitted by : @xmath124 above equation ( labeled 2 in fig . [ fig : law ] ) gives the approximate @xmath120 relation for lmc stars . this relation can be compared with a mass determination for two detached eclipsing systems in smc done by harries et al . ( 2003 ) see fig . [ fig : law ] . components of sc6 215965 lie very close to the predicted m - l relation , but there are some problems with components of sc5 38089 : especcially a more massive component of 38089 seems to be a less luminous one . as the physical parameters of 215965 are more precisely determined than in the case of 38089 i decided to adopt the mass - luminosity relation given by eq . [ mas : lmc ] for smc stars . relying on this assumption we can check the consistency between parameters found from the flux scaling ( eq . [ axis ] ) and from @xmath120 relation ( eq . [ mas : lmc ] ) . @|l|c|c|c|c|c|c|c|c|c|c|c|c|c|@ & & & mass & & & & & & + object&spectrum&@xmath125&@xmath126&ratio @xmath60&@xmath127&@xmath128&@xmath129&@xmath130 & @xmath38 & @xmath74 & @xmath131 & @xmath132&@xmath133@xmath134 + + sc3 139376@xmath66&b0v+b0v & 15.4@xmath672.0 & 16.3@xmath672.1 & 1.06@xmath670.11 & 7.8@xmath670.9 & 8.6@xmath671.0 & 3.84 & 3.79 & @xmath13531500@xmath671100 & @xmath13531700@xmath671100 & 4.731 & 4.823 & 0.128 + sc5 129441@xmath66@xmath68 & b3iii+b3iii & @xmath1357.1@xmath670.6 & @xmath1357.2@xmath670.6 & 1.01@xmath670.07 & 6.5@xmath670.4 & 6.7@xmath670.4 & 3.67 & 3.65 & 16900@xmath67400 & 16900@xmath67400 & 3.486 & 3.506 & 0.094 + sc5 311566 & o9v+o9v & 11.9@xmath671.6 & 11.0@xmath671.5 & 0.92@xmath670.10 & 4.2@xmath670.5 & 3.9@xmath670.5 & 4.28 & 4.30 & @xmath13534200@xmath671300 & @xmath13532900@xmath671200 & 4.324 & 4.199 & 0.180 + sc6 11143@xmath68&b2v+b2v & @xmath1357.5@xmath670.8 & @xmath1356.6@xmath670.7 & 0.88@xmath670.08 & 4.3@xmath670.5&3.6@xmath670.4 & 4.04 & 4.14 & 21600@xmath67600 & 21000@xmath67600 & 3.565 & 3.367 & 0.103 + sc6 67221@xmath66 & b2iv+b1iv & 10.0@xmath671.1&11.1@xmath671.2 & 1.11@xmath670.09 & 7.5@xmath670.6 & 6.8@xmath670.6 & 3.69 & 3.82 & 21700@xmath67600 & 25100@xmath67700 & 4.052 & 4.218 & 0.153 + sc6 221543 & b2v+b2v & @xmath1358.2@xmath671.0 & @xmath1357.4@xmath671.0 & 0.90@xmath670.09 & 4.8@xmath670.5 & 4.3@xmath670.5 & 3.99 & 4.03 & 22400@xmath67600 & 21500@xmath67600 & 3.715 & 3.554 & 0.027 + sc6 272453 & b2v+b2iv & @xmath1357.9@xmath670.9 & @xmath1358.4@xmath671.0 & 1.06@xmath670.09 & 5.1@xmath670.6 & 6.1@xmath670.6 & 3.93 & 3.79 & 21400@xmath67600 & 20500@xmath67500 & 3.680 & 3.775 & 0.065 + sc9 163575 & o5+o7 & 16.9@xmath672.4 & 14.6@xmath672.0 & 0.86@xmath670.10 & 4.9@xmath670.6 & 4.7@xmath670.6 & 4.29 & 4.26 & @xmath13543300@xmath671800 & @xmath13538600@xmath671500 & 4.875 & 4.640 & 0.073 + sc10 37223 & b1v+b1v & @xmath1358.4@xmath671.0 & @xmath1357.8@xmath670.9 & 0.93@xmath670.08 & 4.0@xmath670.4 & 3.9@xmath670.4 & 4.15 & 4.15 & 25100@xmath67700 & 23900@xmath67650 & 3.766 & 3.648 & 0.064 + + sc4 53898@xmath66 & b1v+b2iv & @xmath1356.5@xmath670.6 & @xmath1358.9@xmath670.9 & @xmath1351.36@xmath670.12@xmath70 & 2.9@xmath670.2 & 5.4@xmath670.4 & 4.34 & 3.93 & 24400@xmath67700 & 22200@xmath67600 & 3.415 & 3.801 & 0.122 + sc4 103706@xmath68 & b0v+b0v & 10.7@xmath671.3 & @xmath1359.1@xmath671.1 & 0.86@xmath670.08 & 4.9@xmath670.4 & 4.0@xmath670.4 & 4.09 & 4.21 & @xmath13528700@xmath671000 & 27700@xmath67800 & 4.162 & 3.917 & 0.085 + sc4 163552@xmath68 & b1v+b1v & @xmath1359.3@xmath671.0 & @xmath1359.2@xmath671.0 & 0.98@xmath670.09 & 5.0@xmath670.5 & 4.9@xmath670.5 & 4.01 & 4.03 & 24700@xmath67700 & 24800@xmath67700 & 3.922 & 3.897 & 0.216 + sc5 38089 & b0v+b0v & 11.7@xmath672.0 & 11.8@xmath671.9 & 1.01@xmath670.12 & 5.3@xmath670.9 & 5.5@xmath670.8 & 4.10 & 4.07 & @xmath13529700@xmath671000 & @xmath13529400@xmath671000 & 4.291 & 4.302 & 0.047 + sc6 215965@xmath66@xmath68&b0v+b1iv & 14.6@xmath671.1 & 14.8@xmath671.1 & 1.02@xmath670.05 & 9.0@xmath670.7&@xmath13610.9@xmath670.8 & 3.70 & 3.54 & 28100@xmath67900 & 25900@xmath67800 & 4.653 & 4.677 & 0.039 + sc6 319966@xmath68 & b2v+b2v & @xmath1356.2@xmath670.6 & @xmath1356.1@xmath670.6 & 0.98@xmath670.07 & 3.4@xmath670.3 & 3.3@xmath670.3 & 4.16 & 4.19 & 20700@xmath67500 & 20800@xmath67500 & 3.286 & 3.259 & 0.085 + sc8 104222 & b1v+b1v & @xmath1358.4@xmath671.0 & @xmath1358.7@xmath671.1 & 1.03@xmath670.10 & 4.6@xmath670.6 & 4.9@xmath670.5 & 4.05 & 4.01 & 23900@xmath67650 & 23800@xmath67650 & 3.784 & 3.826 & 0.113 + sc9 175336@xmath66 & b1v+b1iv & 10.0@xmath671.0 & 11.4@xmath671.1 & @xmath1351.15@xmath670.05@xmath70 & 4.7@xmath670.4 & 8.7@xmath670.6 & 4.09 & 3.62 & 25200@xmath67700 & 24300@xmath67700 & 3.901 & 4.371 & 0.039 + + sc4 192903 & f9iv+g2iv & @xmath1351.5@xmath670.3 & @xmath1351.6@xmath670.3 & 1.1@xmath670.2 & 15@xmath672 & 19@xmath672 & 2.26 & 2.06 & @xmath1355900@xmath67100 & @xmath1355400@xmath67100 & 2.372 & 2.472 & 0.087 + sc8 201484 & g0iii+g7iii & @xmath1351.9@xmath670.3 & @xmath1351.9@xmath670.3 & 1.0@xmath670.2 & 36@xmath672 & 48@xmath673 & 1.59 & 1.36 & @xmath1355700@xmath67100 & @xmath1355000@xmath67100 & 3.088 & 3.116 & 0.087 + + eros 1044 & b2v+b2iv & @xmath1357.1@xmath670.8 & @xmath1358.7@xmath670.9 & 1.22@xmath670.10 & 4.2@xmath670.4 & 6.7@xmath670.6 & 4.04 & 3.73 & 21400@xmath67600 & 20500@xmath67500 & 3.524 & 3.844 & 0.065 + hv 982 & b1v+b1v & 10.4@xmath671.1 & 10.6@xmath671.1 & 1.02@xmath670.08 & 7.0@xmath670.5 & 7.6@xmath670.5 & 3.77 & 3.70 & 23300@xmath67650 & 22800@xmath67650 & 4.111 & 4.143 & 0.081 + hv 2274 & b2iv+b2iv & 12.5@xmath671.3 & 12.0@xmath671.2 & 0.96@xmath670.07&@xmath13610.0@xmath670.7 & 9.2@xmath670.7 & 3.54 & 3.60 & 23100@xmath67650 & 23200@xmath67650 & 4.407 & 4.341 & 0.121 + note : all parameters were estimated assuming @xmath111 for smc stars and @xmath137 for lmc stars . the spectrum was estimated from @xmath81 and @xmath80 and , in the case of lmc stars , adopted from the literature . the stars chosen in the present paper as likely distance indicators are marked by an asterisk . the stars indicated by wyithe & wilson ( 2001 , 2002 ) are marked by a dagger . a double dagger indicates a photometric mass ratio derived from the light curve analysis . such treatment allows for independent determination of the unknown reddening in the direction to the particular binary and thus a reasonable @xmath75 intrinsic colour can be obtained . although the mean foreground reddening toward smc is quite low some large random internal reddenings ( up to 0.2 ) were reported for individual b type stars ( larsen , clausen & storm 2000 ) . it is obvious ( e.g. flower s 1996 calibration ) that for early type stars the temperature is a very strong function of the @xmath75 colour and a small change of this colour causes a large change of the temperature . as the temperature enters eq . [ axis ] at a relative high power even a small error in determination of reddening @xmath138 produces large errors in luminosities and masses . let us return to the problem of the determination of physical parameters . we can compute the expected mass ratio @xmath60 as follows : @xmath139 where @xmath140 are the absolute luminosities computed by using eq . then the masses of individual components @xmath141 are calculated from eq . [ mas : lmc ] . if there is a consistency between the flux scaling and @xmath120 relation a quantity @xmath142 : @xmath143 should be equal or close to the free term in eq . [ mas : lmc ] i.e. 0.38 . i have allowed @xmath142 parameter to vary in the range of 0.36 - 0.40 . if @xmath142 were in allowed range the consistency was achieved , the @xmath73 was approved as the final reddening and calculations stopped . if not , the iterative determination of the reddening were done . for @xmath142 greater than 0.40 the correction of @xmath144 were added to @xmath138 , while for @xmath142 smaller than 0.36 the correction of @xmath145 was applied and all calculations were repeated for this new value of reddening , until the @xmath142 fell in the allowed range . the above method was tested on three lmc eclipsing binaries which have their physical parameters known with high accuracy . as a free parameter i assumed distance modulus to the lmc . the best `` fit '' to the original data were obtained for @xmath137 what is a value still accepted as close to the true one ( e.g. gibson 1999 ) and marginally consistent with @xmath112 obtained using eclipsing binaries ( ribas et al . the comparison between original data and parameters given in table [ tab : par ] ( lowest panel ) show that the `` fit '' is excellent the method allows for consistent and accurate estimation of masses , sizes , temperatures , luminosities and reddenings of all three systems . it may serve as a proof that such an approach of estimation of physical parameters works well and may give fast and proper information about the absolute dimensions for large samples of eclipsing binaries see [ fast ] for a discussion . it is worth to note that the initial value of reddening is irrelevant to the method and the same results we obtain adopting another initial @xmath138 . table [ tab : par ] gives the final estimation of the physical parameters of each component of binaries in the sample . the spectral type and luminosity class of each star was estimated from @xmath81 and @xmath80 . sc4 53898 and sc9 175336 were treated in a slightly different way . the mass ratio for both systems was not calculated from eq . [ ratio ] but was assumed to be equal to the photometric mass ratio determined from the light curve solution ( see 3.4 ) . the further calculations were done in a similar way like for the rest of the sample . for two binaries from c subset which host evolved and cool components the reddening was assumed to be equal to the initial value and just the simple flux scaling was applied . the temperatures @xmath74 from table [ tab : par ] in a few cases ( sc4 163552 , sc5 311566 , sc9 175336 ) significantly differ from initial @xmath74 for which light curve solutions were found . additional tests show , however , that photometric parameters recalculated for the temperatures resulted from the flux scaling were consistent with those from table [ tab : sola ] within quoted errors . below the discussion of individual eclipsing binaries as distance indicators to smc is presented . six candidates were chosen according to the following criteria : the presence of complete eclipses , small residuals , small and well determined third light contribution and the spectral type . two binaries from this subset were also advocated to be possible distance indicators by ww1 . another four systems which were selected by ww1 and ww2 were also briefly discussed in a separate subsection , where reasons against their selection were given and additionaly sc5 38089 used recently for distance determination to smc is also discussed . these two binaries may serve as the best candidates because of the presence of complete eclipses , high brightness ( especially sc9 175336 ) , modest proximity effects allowing for reasonable determination of the third light contribution to the system and at last , small residuals . in both cases the primary is slightly cooler and considerably larger than the secondary component . as a result there is a quite high luminosity ratio between components @xmath146 ( see table [ tab : sola ] ) which may by regarded as the only disadvantage of using these systems . the expected velocity semiamplitudes should be @xmath147 km / s and @xmath148 km / s for sc4 53898 , while @xmath149 km / s and @xmath150 km / s for sc9 175336 . the latter lies near the large h@xmath18ii emission region and ob association ngc 371 . both stars were not included by ww1 or ww2 as likely distance indicators . very high eccentricity and clear proximity effects visible near the periastron passage are features of both systems . sc5 129441 consists of two somewhat evolved b3 stars of equal temperatures . in the second system sc6 67221 during the primary eclipse ( phase 1.0 ) the cooler and probably larger component is eclipsed and the strong mutual reflection causes the brightening of the system visible as a hump around phase 1.0 . both systems may serve as ideal testers of mutual reflection treatment used in modern light curve synthesis programs . indeed , the presence of small residuals visible near phase 1.1 in sc5 129441 and near the primary eclipse in sc6 67221 may ( fig . [ fig : fita ] ) suggest that simple reflection used in the analysis ( see 3.1 ) might be too crude a simplification . unfortunately the use of the detailed reflection ( which should be more appropriate for these binaries ) together with the eccentric orbits makes wd program practically unsuitable for analysis due to an enormously large computational time ( wilson 1992 ) . both systems show no substantial third light contribution . the expected velocity semiamplitudes should be @xmath151 km / s for sc5 129441 and @xmath152 km / s and @xmath153 km / s for sc6 67221 . sc5 129441 was suggested by ww1 to have complete eclipses , but the present light curve analysis did not support this thesis . sc6 67221 was not included by ww1 . this binary seems to be the most massive and the most luminous in the sample ( beside sc9 163575 ) , consisting of two b0 v - iv stars of similar temperature : @xmath154 k. it is the only `` classical '' well - detached system included in the subset of likely distance indicators the fractional radii of both components are below 0.2 . the residuals of solution are very small and only some larger scatter can be seen during eclipses due to very fast apsidal motion . the rate of the motion is @xmath155 58/year , so the apsidal period @xmath95 is only @xmath156 years . thus the analysis of this system can give parallel information about the components and an independent check of stellar evolution models . the expected velocity semiamplitudes should be @xmath157 km / s and @xmath158 km / s . the brightest binary in the sample ( @xmath159 ) with the spectrum b0 v - iv + b1 iv . table [ tab : par ] suggests that the cooler component is more massive one and it is slightly evolved . the light curve solution showed that the contribution of the third light to this system might be neglected and photometric parameters such as the temperature ratio ( surface brightness ratio ) could be very accurately determined . also very small residuals suggested that the system had no essential intrinsic variability . this system was yet used by harries et al . ( 2003 ) for a distance determination to smc giving @xmath160 . the physical parameters of both components are reasonably close to that presented in table [ tab : par ] . it can be considered as an independent check of the method given in section [ physics ] and an additional argument for use this star for an accurate individual distance determination . ww1 included this binary in their catalog on the basis of having complete eclipses but the present analysis ( and results of harries et al . ) exludes this suggestion . harries et al . ( 2003 ) determined absolute dimensions and a distance to this b0v+b0v system : @xmath161 . their light curve solution is essentially the same as the solution reported in this paper . however , there is an apparent discrepancy between the observational data on sc5 38089 and the mass - luminosity relationship shown on fig [ fig : law ] . thus i would not recommended this star for individual distance determination , although it is very interested target for spectroscopic investigations in the future . table [ tab : par ] gives the estimate of physical parameters assuming that both components are normal main sequence stars . the expected @xmath32 luminosity ratio is very close to unity what indeed was reported by harries et al . formally this binary also seems to have complete eclipses ( it was included by ww1 in their subset of binaries showing totality ) but the very large third light contribution ( @xmath162 ) may cause serious problems during modelling of the spectral energy distribution and thus i decided not to include this star into the most suitable candidates . ww2 included this binary according to the presence of complete eclipses , but as in the case of sc6 215965 , the light curve analysis contradicted this thesis . the third light contribution to this system is sunstantial , at about 26% . this binary has small residuals , a very smooth light curve and can be considered as a `` reserve '' candidate . the mass ratio and temperature ratio are very close to unity . the expected velocity amplitude is @xmath163 km / s for both components . ww1 classified this system as a well - detached . indeed , my analysis showed that the components had the smallest fractional radii of all the sample ( beside sc6 67221 ) and are normal b2 v stars . however , this binary happens to blend with another eclipsing binary ( see 2.2 ) which contributes about 30% to the total light of system . thus any detailed analysis of sc6 11143 should at the same time take into account a careful analysis of the blending binary . probably this includes simultaneous solution of multicolour light curves of both binaries what is , in principle , possible , but may give ambiguous results . this binary was included by ww1 according to the presence of complete eclipses which seems to be incorrect in the view of my analysis . this binary had the highest @xmath1 contribution from all the sample nearly 50 % which is most probably too large for a detailed modelling of spectrophotometry . the observed difference of the @xmath1 contribution to the total light between binaries from a and b subsets may be understood as follows . the components of binaries from b subset are relatively much closer to each other than those from a subset and the proximity effects , depending on the gravitational distortion and overall geometry of the system , are much more distinct . thus proximity effects give us additional constraints to the solution . when we consider the third light @xmath1 we see that its influence on the light curve of well - detached systems is of little importance : equally good solutions can be obtained for wide range of @xmath1 values . the situation is different for closer systems small changes of @xmath1 produce noticeably systematic o - c residuals , especially near quadratures . this is an important reason for preferring such systems over classical well - detached systems in determination of the distance using individual eclipsing binaries , especially in the case of the extragalactic binaries observed usually on a rich stellar background of a host galaxy . the probability of blending with another star(s ) in such environment is high ( crowding effect ) . also the blend may be a member of the system itself as at least 30% of binaries ( batten 1973 ) are found to form hierarchical multiple systems of stars . indeed , neglecting of the third light contribution in the case of hv 2274 and hv 982 may be one of the prime sources of small differences on the level 1.5@xmath164 between distances derived for three lmc systems as was announced by ribas et al . ( 2002 ) . the idea of deriving the absolute properties of binary stars from knowledge of their photometric properties alone has long tradition ( e.g. gaposchkin 1938 , kopal 1959 - chapter 7 ) . the methods based on this idea use somewhat statistical approach : one or both components of a binary conform an empirical mass - luminosity relation and the temperature is imposed by observed spectral type . there are two important advantages of such statistical method 1 ) it may be used for large samples of binaries and 2 ) it gives fast , though approximate physical parameters of binaries . afterward we can choose from the sample most interesting systems and investigate them more accurately using spectroscopy , spectrophotometry or multicolor photometry . however , past methods neglected the interstellar reddening ( which influences both the brightness and the colours ) and , what is more important , the distance to the binary . as a matter of fact the distance is irrelevant to these methods and thus it can be derived simply by comparison with the apparent magnitude @xmath165 , even without any radial velocity measurements ( e.g. gaposchkin 1968 , dworak 1974 ) . moreover , they based their reasoning on quite crude spectral type - temperature calibrations and , of course , as the temperature come in high powers into equations involved in the method , the obtained parameters were usually very inaccurate . in present paper this `` old - fashioned '' method was innovated to account for the interstellar reddening @xmath138 and recent @xmath12 colour - temperature calibration ( flower 1996 ) . it turns out that such approach causes a big improvement in the temperature determination and overall physical parameters as was demonstrated by an accurate `` reproducing '' of absolute dimensions of three lmc eclipsing binaries . however , it was done for a little bit larger distance to lmc than the distance derived , on average , from these lmc stars . a shift of the distance modulus @xmath166 is quite small and comes probably from some inaccuracies of flower s temperature or / and bolometric corrections calibration for hot stars . anyway we should expect that this kind of method may give fast approximate determination of absolute dimensions for many binaries in the magellanic clouds , especially as we have large archives of binaries discovered by ogle , macho and eros projects at our disposal , and may be used also for eclipsing binaries in other close galaxies in the near future . i am grateful to the anonymous referee for his / her very valuable comments and suggestions . i would like to thank professor andrzej udalski for his comments on ogle database and dr tnu viik for reading the manuscript .

we have obtained oh spectra of four transitions in the ground state , at 1612 , 1665 , 1667 , and 1720 mhz , toward 51 sightlines that were observed in the herschel project galactic observations of terahertz c+ . the observations cover the longitude range of ( 32 , 64 ) and ( 189 , 207 ) in the northern galactic plane . all of the diffuse oh emissions conform to the so - called ` sum rule ' of the four brightness temperatures , indicating optically thin emission condition for oh from diffuse clouds in the galactic plane . the column densities of the hi ` halos ' surrounding molecular clouds increase monotonically with oh column density , , until saturating when 2 and 2 , indicating the presence of molecular gas that can not be traced by hi . such a linear correlation , albeit weak , is suggestive of halos contribution to the uv shielding required for molecular formation . about 18% of oh clouds have no associated co emission ( co - dark ) at a sensitivity of 0.07 k but are associated with emission . a weak correlation exists between intensity and oh column density for co - dark molecular clouds . these results imply that oh seems to be a better tracer of molecular gas than co in diffuse molecular regions .

1703.05864

the hydroxyl radical ( oh ) is a relatively abundant , simple hydride , and thus a potentially important probe of interstellar medium ( ism ) structure . it was first detected in absorption against continuum sources @xcite and then in emission toward interstellar dust clouds ( heiles 1968 ) . a large number of studies have revealed the widespread existence of oh throughout dense , dusty clouds @xcite , high - latitude translucent clouds @xcite , and diffuse regions outside the co - bright molecular clouds @xcite . four 18 cm ground - state transitions of oh at 1612 , 1665 , 1667 , and 1720 mhz can be readily observed in l band . local thermodynamic equilibrium ( lte ) was initially considered to be valid for these four transitions ( e.g. , * ? ? ? * ) . under optically thin assumption , lte implies the ratios @xmath5 . subsequent observations revealed anomalies in satellite ( 1612 , 1720 mhz ) and main ( 1665 , 1667 mhz ) lines ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) . on - source absorption and off - source emission observations toward continuum sources have been used to obtain the optical depth and excitation temperature of each transition independently . the aforementioned surveys have found non - lte gas with a typical excitation temperature difference of @xmath6 k for the oh main lines ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? recent observations have shown that co , the widely used tracer of 2 , does not trace molecular gas well in regions with intermediate extinctions @xmath7 mag ( e.g. , * ? ? ? we refer to such regions as diffuse , or translucent clouds , and when appropriate , @xmath8 transition regions , throughout this work . oh and c@xmath9 are key initiators of the chemistry that leads to co in diffuse and translucent regions through the reactions @xcite @xmath10 , @xmath11 , @xmath12 . oh has been detected toward the outer shells , also referred to as ` halos ' , of molecular clouds with low co abundances ( e.g. , * ? ? ? * ; * ? ? ? * ) , and clouds toward continuum sources @xcite . in order to improve the understanding of the distribution of oh and of the ism traced by oh , large and sensitive surveys of oh in diffuse gas are necessary . the first blind " survey of diffuse oh taken by @xcite was unsuccessful . @xcite carried out oh survey near galactic plane with a sensitivity of 0.18 k. the surveys with high sensitivity ( a few mk rms ) by @xcite covered regions ( @xmath13)=(108.0@xmath1 , 5.0@xmath1 ) and ( 105.0@xmath1 , 1.0@xmath1 ) with the 25 m radio telescope of the onsala space observatory and green bank telescope , respectively . the southern parkes large - area survey in hydroxyl ( splash ) covered ( @xmath13)=(334@xmath1 - 344@xmath1 , -2@xmath1 - 2@xmath1 ) in a pilot region , and will cover @xmath14 ranges of ( 332@xmath1 , 10@xmath1 ) , @xmath15 including some additional coverage of higher altitude around the galactic centre ( dawson , in prep ) . in these regions , observations of c@xmath9 are not available . in this work , we adopted a more limited and focused approach by following up on galactic observations of terahertz c+ ( gotc+ ) survey @xcite with oh observations . with a data set of the three important tracers of molecular gas , c@xmath9 , oh , and co , we here examine their correlation and their relative efficiency in tracing molecular gas . this paper is organized as follows : in sections [ sec : observations ] and [ sec : reduction_detection ] , we describe the observations and data reduction of oh and associated spectral data . in section [ sec : analysis ] we show procedures for gaussian decomposition and analysis of oh , @xmath16 , and co column density . the results are presented in section [ sec : results ] . in section [ sec : discussion ] we provide a discussion of oh column density and atomic / molecular transition . in section [ sec : conclusion ] we provide the conclusions from our study . there are 92 sightlines of the gotc+ project that are covered by the arecibo telescope . we chose observed sightlines based on the following three criteria : ( 1 ) if the sightline can be observed for at least half an hour , ( 2 ) if there exist `` co - dark '' candidates toward the sightline , and ( 3 ) if there exist abundant self - absorption features , c@xmath9 , and co emission toward the sightline . criteria ( 1 ) and ( 2 ) have higher priority . all sightlines satisfying criteria ( 1 ) and ( 2 ) have been observed . in this survey , oh spectra toward 51 sightlines were obtained . the positions of these sightlines covering 43 points in the galactic longitude range of ( 32@xmath1 , 64@xmath1 ) ( range a ) and 8 points in the galactic longitude range of ( 189@xmath1 , 207@xmath1 ) ( range b ) are shown in figure [ fig : oh_count_plot ] . the oh observations were carried out with the arecibo telescope in two periods , september 15th to november 7th , 2014 and february 26th to march 3rd , 2015 . the observations were made with the interim correlator backend with bandwidth of 3.125 mhz , providing a velocity resolution of 0.28 km s@xmath17 at 1.66 ghz . the integration time for each sightline was half an hour . to reduce the effect of radio frequency interference ( rfi ) and the instability of the receiver and to avoid difficulty in choosing a clean `` off '' position , we developed an observation script that changes the central reference velocity by 200 km s@xmath17 every 15 min . this is equivalent to frequency switching , which is not supported in arecibo . corresponding @xmath18co(1 - 0 ) and @xmath19co(1 - 0 ) observations were made with the delingha 13.7 m telescope between may 4th and 10th , 2016 . the 13.7 m telescope , located in northwestern china , has an angular resolution of 1 arcmin ( fwhp , full width at half power ) at 115 ghz . the system temperature varied from 250 k to 360 k with a typical value of 300 k during the observations . the observations were taken using position switching . the total observation time per target was 30 min or 45 min depending on the system temperature . the backend has a 1 ghz bandwidth and 61 khz spectral resolution , corresponding to a velocity resolution of 0.16 km / s at 115.271 ghz . c@xmath9 data were obtained from the got c+ project @xcite . the data have already been smoothed into a channel width of 0.8 km s@xmath17 with an average rms noise of 0.1 k. the hi data representing brightness temperature were taken from the galactic arecibo l - band feed array hi with a noise level of 0.33 k in a 0.18 km s@xmath17 channel . the oh data were reduced with our idl procedures . scans with obvious rfi were firstly removed by checking the correlation map of the data . rfi was further checked by comparing averaged spectra in two separate 15 min observations . this is especially important for the 1612 mhz spectra , which are significantly affected by rfi . after deriving the bandpass spectrum , we ignored the edges of the spectrum where gain of the bandpass varies and only fitted middle part of the spectrum . spectral channels with obvious oh lines were marked to avoid being included in the bandpass fit . most of the bandpass spectra are flat and can be fitted with a first - order polynomial . the other spectra were fitted with higher - order polynomials . weak oh emission / absorption lines with wide velocity widths ( full width at half maximum @xmath20 km s@xmath17 ) may be missed during this step . the final noise level is 35 mk in a 0.28 km s@xmath17 channel . a main beam efficiency of 0.52 was used to transform co antenna temperatures to main - beam brightness temperatures . the gildas software was used for baseline fitting and spectral smooth of co data . the co spectra were smoothed to 0.32 km / s to reach a velocity resolution comparable to that of oh data . the final noise level of main - beam brightness temperature is @xmath21 mk for @xmath18co(1 - 0 ) in 0.32 km s@xmath17 and @xmath22 mk for @xmath19co(1 - 0 ) in 0.33 km s@xmath17 channel width . the hi data were smoothed to a velocity resolution of 0.36 km s@xmath17 that is comparable to the velocity resolution of oh and co data . the rms noise level after smoothing is 0.23 k. with a rms of @xmath23 mk , the detection statistics of the 4 oh lines are displayed in figure [ fig : oh_count_plot ] . oh emission / absorption is detected in 44 of 51 sightlines . oh main lines appear in 9 of 44 sightlines alone while oh satellite lines appear in 2 of 44 sightlines alone . the detection rate of oh main and satellite lines varies depending on their locations in the galaxy . no oh satellite lines were detected in the outer galaxy . figure [ fig : oh_count_plot ] indicates that the detection rate of oh lines ( including both the main and satellite lines ) in the outer galaxy is 62.5% , much smaller than that of 93.0% in the inner galaxy . this is consistent with the fact that the amount of co - bright molecular gas in the outer galactic plane is smaller than that in the inner galactic plane ( e.g. , * ? ? ? absorption features are commonly present in oh main lines in the inner galaxy even though there is no h ii region in the beam . but absorption features are absent in oh main lines in the outer galaxy , indicating a lower level of continuum background in the outer galaxy . , 210@xmath1 ) and ( 30@xmath1 , 65@xmath1 ) , respectively . circle , plus , and cross signs represent detection of oh main lines , detection of oh satellite lines , and no detection , respectively.,scaledwidth=47.0% ] we developed an idl script to decompose oh , and co spectra . this script uses the classical nonlinear least squares technique , which utilizes analytically - calculated derivatives , to iteratively solve for the least squares coefficients . for each spectrum , the number of gaussian components was fixed . initial guesses of each gaussian component were required . the decomposition results were then checked by eye . we fit the oh profiles first . in general , central velocities of the four oh lines should be the same for a cloud . a switch from emission to absorption as a function of velocity in oh satellite lines exists in some clouds . in these clouds , the central velocities of the main lines are the same as the cross points of the satellite lines . an example is shown in figure [ fig : sate_anomaly ] and discussed in section [ sec : oh_column_density_dis ] . this always occurs in the clouds near h ii regions and can be explained by infrared pumping of the @xmath24 level @xcite . we treat this kind of feature as a single component . the central velocities of derived oh components were adopted as the initial guess for decompositions of c@xmath9 and co data . finally , 151 cloud components with oh emission or absorption lines were identified . an example is shown in figure [ fig : g036.4 + 0.0_spec ] . . this corresponds to the velocity where the 1612 and 1720 mhz flips occur . , scaledwidth=47.0% ] the brightness temperature ratio between the 4 oh lines ( t@xmath25:t@xmath26:t@xmath27:t@xmath28 ) is 1:5:9:1 under assumptions of lte and optically thin emission ( e.g. , * ? ? ? . an anomalous ratio of oh lines that deviates from the 1:5:9:1 ratio can not be explained by optical depth effects . an oh anomaly implies non - lte conditions leading to differential excitation of 4 oh lines . satellite line anomaly is seen more often than main line anomaly . the main line transitions occur between levels with the same total angular momentum quantum number ( f ) . for satellite lines , transitions occur between energy levels with different f , which are easily affected by non - thermal excitation @xcite . inversion of satellite lines is commonly seen without inversion of main lines ( see figures [ fig : sate_anomaly ] and [ fig : g036.4 + 0.0_spec ] for examples ) , making it difficult to calculate the oh column density with satellite lines . we thus calculated oh column densities only for clouds with main line emission . the radiative transfer of the main lines in lte can be written as @xmath29 @xmath30 where @xmath31 and @xmath32 are the brightness temperatures of 1665 and 1667 mhz lines , respectively , @xmath33 is the beam filling factor , @xmath34 is the excitation temperature , and @xmath35 is the background continuum temperature at 1.6 - 1.7 ghz . in high latitude regions , @xmath36 k at 1.6 ghz , which is the sum of cosmic microwave background ( cmb ) of 2.73 k @xcite and galactic synchrotron emission of @xmath37 k extrapolating from 408 mhz survey @xcite with a spectral index of 2.7 @xcite . the contribution from continuum sources ( e.g. , h ii regions ) becomes important at low latitudes , especially the galactic plane in this survey . the hipass 1.4 ghz continuum survey @xcite was used to estimate continuum emission at 1.6 - 1.7 ghz toward each sightline . we subtracted 3.3 k ( the sum of 2.73 k cmb and @xmath38 k from galactic synchrotron emission at 1.4 ghz ; e.g. , @xcite ) from hipass data and then estimated values at 1612 , 1666 , and 1720 mhz with a spectral index of @xmath39 found in the splash survey . a fraction factor @xmath40 ( @xmath41 ) was utilized to derive continuum contribution behind oh cloud . with the assumption that the continuum contribution is uniformly distributed along the sightline across the milky way , @xmath40 is represented as @xmath42 , in which @xmath43 is distance to oh cloud and @xmath44 is the sightline length across the milky way . during the calculations , we applied the milky way rotation curve in @xcite and a maximum galactocentric radius of 16 kpc . the values of @xmath40 vary from 0.48 to 0.98 with a median of 0.90 . finally , a correction of 3.1 k was added back to derive @xmath45 at 1.6 - 1.7 ghz . the uncertainties are discussed in the end of this section . the oh column density @xmath46 can be calculated with the following two general equations @xcite @xmath47 @xmath48 where @xmath49 and @xmath50 are the main beam brightness temperatures of the 1665 mhz and 1667 mhz lines , respectively . @xmath51 is the correction factor for the optical depth @xmath52 of the oh transitions . the correction factor for @xmath34 , @xmath53 , approaches 1 when @xmath54 k. in lte , the ratio between brightness temperature of main lines ( @xmath55 ) varies between 1.8 for optically thin conditions and 1.0 for infinite optical depth @xcite . when @xmath56 was in the range of [ 1.0 , 1.8 ] , the combination of equation [ eq:65ta ] and equation [ eq:67ta ] can solve for @xmath57 and @xmath58 simultaneously . then @xmath57 and @xmath58 can be inserted into equation [ eq:65_col ] or [ eq:67_col ] to solve for n(oh ) . previous oh observations have revealed ubiquitous anomalies between excitation temperatures of the main lines . non - lte excitation can lead to ratios mimicking lte range @xcite . beside this , lte calculations are limited by satisfaction of sum rule , which implies small optical depth as described in section [ sec : sumrule ] . the values of @xmath56 in 29 oh clouds are in the lte range . as shown in figure [ fig : lte_nonlte ] , the values of @xmath58 in 4 clouds are smaller than 0.5 with the lte assumption . with consideration of satisfying sum rule implying optically thin as described in section [ sec : sumrule ] , we adopted lte calculation results for these 4 clouds . the method for non - lte oh clouds in case 3 described below was adopted to calculate oh column densities of the remaining 25 clouds . as shown in figure [ fig : lte_nonlte ] , lte assumption generally leads to higher optical depth , @xmath59 and larger oh column density , @xmath60(oh)@xmath61 @xmath62 than non - lte assumption . ( oh)@xmath63 and @xmath64 represent total oh column density and optical depth of 1667 line using the lte assumption , respectively . n(oh ) represents total oh column density using non - lte assumption . the vertical dashed line represents n(oh)@xmath63/n(oh ) ratio of 1 . the horizontal dashed line represents @xmath65 . , scaledwidth=47.0% ] we now consider the non - lte cases . as shown in equation [ eq:65_col ] and [ eq:67_col ] , oh column density and its uncertainty are very sensitive to @xmath57 through the function , @xmath66 . it would be ten times lower for @xmath67 than that of @xmath68 . but there exists a constraint on @xmath69 in order that oh be detected with our sensitivity as shown in figure [ fig : oh_detect_paraspace ] . it requires a larger deviation of @xmath70 from 1 for small @xmath60(oh ) to be detected . moreover , we are able to apply reasonable assumptions to different non - lte cases . the following cases are clearly non - lte when we consider the main lines ( masers are ignored here ) . 1 . the existence of 1665 mhz line alone . 2 . the existence of 1667 mhz line alone . 3 . both 1665 and 1667 mhz lines are present , but @xmath56 is out of the lte range . in case 1 , the existence of the 1665 mhz line alone with the absence of the 1667 mhz line implies the equality between excitation and background temperature of 1667 line , @xmath71 . previous emission / absorption observations toward continuum sources revealed @xmath72@xmath73 @xmath74 k ( e.g. , * ? ? ? * ) . we adopted @xmath75 1.0 k , in which plus and minus are for emission and absorption of the 1665 mhz line , respectively . this adoption leads to @xmath76= 6.0 or 8.0 when @xmath77 k , where we have adopted @xmath78 due to minor difference between them . a similar strategy for calculations in case 2 was adopted . we can not exclude the possibility of a detection limit that leads to absence of 1665 mhz detection in case 2 , since 1665 mhz line is generally weaker than 1667 mhz line . but the expected 1665 intensities ( @xmath79 ) are greater than 3 @xmath80 rms in 63% of case 2 clouds and are greater than 2 @xmath80 rms in all clouds of case 2 . thus the assumption of case 2 is reasonable . uncertainties in case 1 and 2 are given with @xmath72@xmath73 in the range of [ 0.5 , 2.0 ] k. the value of @xmath81 for case 1 and case 2 ranges from 4.3 to 11.5 with a median 7.04 . we applied this median value for all calculations in case 3 . the uncertainty in case 3 is given with @xmath81 ranges of [ 4.3,11.5 ] . /(@xmath57-@xmath45 ) . positive excitation temperature @xmath57 was considered , leading to temperature factor range of ( -@xmath82,0 ) and ( 1,@xmath82 ) . the two red vertical lines represent values of 0 and 1 for the temperature factor . a typical background continuum temperature of 8 k and fwhm of 1.5 km s@xmath17 in this survey were adopted , and optically thin condition was assumed . the grey shaded region covers the parameter space within the 3@xmath80 detection limit of 0.21 k. oh lines within that region can not be detected in our survey . when we zoom to the temperature factor in the range of [ -0.06 , 0.01 ] for @xmath83 @xmath62 , @xmath84 approaches a constant value when the temperature factor approaches 0 as shown in the small plot . , scaledwidth=47.0% ] the optically thin assumption was applied to clouds under non - lte conditions . this assumption is reasonable because there is no deviation from the ` sum rule ' as presented in section [ sec : sumrule ] . during the calculation of @xmath60(oh ) of case 1 , equation [ eq:65_col ] was employed . equation [ eq:67_col ] was employed for cases 2 and 3 . oh column densities of 117 clouds with main line emission were calculated . @xmath2 ranges from @xmath85 2 to @xmath86 2 with a median of @xmath87 2 . compared to oh column densities in clouds previously observed , this median value is about one order of magnitude larger than that determined explicitly through on / off observations toward 3c 133 and is more than 3 times the value in the w44 molecular cloud @xcite . two main uncertainties exist in the above assumptions of @xmath40 . the first originates from the distance ambiguity for directions toward the inner galaxy . for oh clouds associated with hi self absorption , near distance is preferred @xcite as we have adopted . for other oh clouds , the distance ambiguity leads to a maximum difference of @xmath40 between near and far distance of 0.57 . only 17 oh clouds are affected . the deviation factor of @xmath60(oh ) caused by the distance ambiguity ranges from 0.049 to 2.0 with a median of 1.6 . the second uncertainty is the difference between three - dimensional distribution of radio continuum emission over the entire galaxy and the uniform distribution we assumed . @xcite reproduced a three - dimensional model of the galactic radio emission from 408 mhz continuum map @xcite , and found exponentially decreasing distribution of emissivities along galactic radius ( 4 kpc @xmath88 r @xmath88 16 kpc ) in the galactic plane . we adopted the detailed radial distribution in fig . 6a of @xcite . the differences of @xmath40 varies from -0.01 to 0.24 with a median value of 0.028 . the deviation factor of @xmath2 caused by three - dimensional model of radio emission ranges from @xmath89 to 0.1 with a median value of 0.02 . thus the uncertainty from three - dimensional model is much smaller than that from intrinsic excitation temperature . permeates in the milky way . the spectrum includes all contributions along a sightline and toward the galactic plane is broad in velocity . it is difficult to distinguish a single cloud without the help of special spectral features , e.g. , narrow self absorption ( hinsa ) against a warmer background ( e.g. , * ? ? ? the excitation temperature ( t@xmath90 ) and optical depth ( @xmath52 ) of the hinsa cloud is essential for deriving the column density for a cloud with a hinsa feature . @xcite introduced a method of fitting the second derivative of the spectrum to derive the background spectrum and fitted @xmath91 of hinsa cloud . we combine the radiation transfer equations in @xcite and the analysis method in @xcite for calculation of n(@xmath92 ) . we assume a simple three - body radiative transfer configuration with background warm gas , cold cloud , and foreground warm gas . the background spectrum without absorption of cold cloud , @xmath93 , is related to the observed spectrum in which continuum has been removed , @xmath94 through the following equation ( see details of equation 8 in @xcite ) , @xmath95 where @xmath96 represents the background continuum temperature contributed by the cosmic background and the galactic continuum emission , @xmath97 is the excitation temperature of the atomic hydrogen in the cold cloud , which is equal to the kinetic temperature , @xmath52 is the optical depth of the cold cloud . @xmath98 and @xmath99 are the optical depths of warm gas in front and behind the hinsa cloud . the total optical depth of warm gas along the line of sight , @xmath100 . @xmath101 is defined as the fraction of background warm hi , @xmath102 . the value of @xmath101 is calculated through @xmath103 where @xmath104 and @xmath105 are the integrated surface densities behind the hinsa cloud and along the all line of sight . the surface density distribution in @xcite and the milky way rotation curve in @xcite were used for this calculation . we try to recover the background spectrum with equation [ eq : hibackspec ] to fit the second derivative as that in @xcite . information on the kinetic temperature is needed . hinsa features are pervasive in the taurus molecular cloud . analysis of pixels with both @xmath18co and @xmath19co emission in this region reveals a kinetic temperature in the range of [ 3,21 ] k , but concentrated in range of [ 6 , 12 ] k. in most cases , we choose a fixed kinetic temperature of 12 k for @xmath19co that is widely used in molecular clouds @xcite and an initial hinsa optical depth of 0.1 . the fitting result with a comparable thermal temperature of gas to 12 k was chosen , otherwise we modify the initial parameter , e.g. , relax the kinetic temperature in the range of [ 6,15 ] k as a free parameter . an example is shown in figure [ fig : g032.6 + 0.5_hinsa ] . the hinsa column density is given by the fitted @xmath52 and fwhm of hinsa cloud , @xmath106 by @xmath107 where @xmath108 is the kinetic temperature of the hinsa cloud . the hinsa column density depends on the value of the kinetic temperature , thus uncertainties are given from kinetic temperature in the range of [ 6,15 ] k. : hinsa spectrum at @xmath109 km s@xmath17 along g032.6 + 0.5 . the observed spectrum ( @xmath110 ) and derived background spectrum ( @xmath111 ) are represented by the black and red solid lines , respectively . the dotted line shows the residual spectrum . during the fitting , the kinetic temperature of co was fixed at 12 k. the original and fitted optical depths are 0.1 and 0.31 . @xmath112 : second derivatives of and background spectra . @xmath113 : corresponding @xmath19co spectrum of hinsa . the green vertical line marks the fitted central velocity of the hinsa cloud from @xmath19co.,scaledwidth=48.0% ] hinsa traces the cold component of neutral hydrogen in a molecular cloud which may have a warm halo @xcite . we were able to determine column density of the halo of molecular clouds with and without hinsa features through the spectra . due to the omnipresence of in the galactic plane , we did not apply gaussian decomposition to profiles without hinsa feature . we derived the column density of gas through the integrated intensity . the integrated intensity of recovered background spectrum was used for clouds with hinsa features . with the assumption of low optical depth , the column density @xmath60(hi ) is given by @xmath114 where the intensity is obtained through integrating the velocity channels determined by oh lines . the effect of adopting different velocity widths of is discussed in section [ sec : oh_vs_hi ] . @xmath60(hi ) derived using this method is limited by the optically thin assumption , the intensity contribution from clouds in neighboring velocities , and absorption features corresponding to oh emission lines ( e.g. , * ? ? ? the hinsa column density , @xmath60(hinsa ) derived in 52 clouds ranges from @xmath115 @xmath62 to @xmath116 @xmath62 with a median value of @xmath117 @xmath62 , which is 1/36 of the median of the halos of these clouds . the median @xmath60(hinsa ) is consistent with that derived in hinsa survey outside the taurus molecular cloud complex , log@xmath118=18.8@xmath119 0.53 @xcite . six masks are defined for different detection cases . they are shown in table [ table : detections ] . when @xmath120,@xmath121 were detected simultaneously ( mask 3 and 6 in table [ table : detections ] ) , the clouds should be dense molecular gas . in this case , @xmath120 is assumed to be optically thick with @xmath122 . @xmath123 , the excitation temperature of @xmath120 , is given by @xmath124\}^{-1 } , \label{eq : exciation_12co}\ ] ] where @xmath125 is the brightness temperature of @xmath120 . the total column density of @xmath19co , @xmath126 , is given by @xcite @xmath127 where @xmath128 is the brightness temperature of @xmath121 . @xmath129 is the correction factor of @xmath130 , the optical depth of @xmath131 . @xmath130 is given by @xmath132 in which @xmath122 and @xmath133 are adopted . these are reasonable when the excitation is dominated by collisions . the galactic distribution of @xmath134 ratio derived from synthesized observations of co , cn , and h@xmath135co in @xcite was adopted to convert @xmath136 to @xmath137 , @xmath138 , where @xmath139 is distance to the galactic center . we adopted the milky way rotation curve of @xcite to derive @xmath139 . velocity dispersions and non - circular motions @xcite are expected to affect the calculations of @xmath139 with an uncertainty of @xmath88 10% , leading to an uncertainty of @xmath140 1% in @xmath134 . for clouds with detection of only @xmath120 ( masks 2 and 5 in table [ table : detections ] ) , it is difficult to determine @xmath141 without observations of higher lines , e.g. , co(2 - 1 ) . with the assumption that @xmath120 is optically thin , we derive a lower limit . adoption of a 3@xmath80 detection of the @xmath121 will give an upper limit to the column density . combining these two facts , we adopted the average value of upper limit and lower limit for @xmath142 . under the optically thin assumption , the total column density of @xmath120 can be expressed as @xmath143 where @xmath144 is the brightness temperature , @xmath145 is the level correction factor , @xmath146 is the correction factor of opacity and @xmath147 was adopted under optically thin condition , @xmath148 $ ] is the correction for the background , @xmath149 is the beam filling factor of the cloud ( assumed to be 1.0 ) . a common excitation temperature of @xmath150 k in molecular regions ( e.g. , taurus molecular cloud ; goldsmith et al . 2008 ) was adopted during the calculation . @xmath151 k is the partition function . @xmath152 represents the degeneracy of the upper transition level and equals 3 for co(1 - 0 ) transition . @xmath153 is the opacity of the @xmath18co(1 - 0 ) transition . @xmath154 was assumed , indicating an underestimation of @xmath137 by a factor of @xmath140 5 when @xmath155 . @xmath45 is the background brightness temperature , adopted to be 2.73 k. a @xmath156 upper limit on @xmath157 of 0.12 k was adopted in equations [ eq : col_13co ] and [ eq : tau_13co ] for calculation of the upper limit to @xmath137 . the column density of @xmath18co ranges from @xmath158 @xmath62 to @xmath159 @xmath62 with a median value of @xmath160 @xmath62 . the median value of @xmath137 for clouds with both @xmath18co and @xmath19co emission is @xmath161 2 , which is @xmath162 times that in clouds with @xmath18co emission alone . @xcite presented a brightness temperature ` sum rule ' , which relates the intensities of the four oh ground - state transitions under the assumptions of small optical depths , a flat background continuum spectrum , and @xmath163 k. the ` sum rule ' is , @xmath164 diffuse oh emission and absorption in the pilot region of splash survey followed this relation , ( where diffuse " oh is defined as signal from the extended molecular ism , in which maser action is either absent or very weak ) . similar to @xcite , we found no deviation from the ` sum rule ' by more than 3 @xmath80 for all diffuse oh emission and absorption . an example is shown in figure [ fig : g035.1 + 0.5_oh ] . according to appendix a , the ` sum rule ' is valid for optically thin condition despite the existence of strong differences between the excitation temperatures of four oh lines . the deviation from the ` sum rule ' is dominated by the opacity of oh lines . no deviation from the ` sum rule ' confirms the validity of the optically thin assumption that has been used for calculation in section [ sec : oh_col_den ] . maser amplification , which indicates strong non - lte behavior and large optical depth , leads to deviation from the ` sum rule ' as shown in figure [ fig : g035.1 + 0.5_oh ] . based on this fact , the ` sum rule ' can be used as a filter for finding maser candidates . . the 1 @xmath80 and 3 @xmath80 levels of the spectrum are indicated by in dashed and dash - dotted lines , respectively . significant deviations are present for the peaks around 64 km s@xmath17 and 90 km s@xmath17 , which represent an evolved stellar maser associated with the infrared source iras 18510 + 0203 . bottom panel : oh spectra for g035.1 + 0.5 are displayed as solid lines with different colors . , scaledwidth=47.0% ] is the tracer of atomic gas while co is a tracer of molecular gas . c@xmath9 emission traces both atomic and molecular gas . all oh clouds have associated emission . spectra of these lines toward g036.4 + 0.0 are shown in figure [ fig : g036.4 + 0.0_spec ] . the statistics of clouds with c@xmath9 and co emission corresponding to oh are listed in table [ table : detections ] . c@xmath9 and co are present in 45% and 80% of all the oh clouds , respectively . we present a detailed comparison between column density of hi , co line and intensity of line with @xmath60(oh ) in the following sections . ccccccc 1 & @xmath165 & x & x & x & 17 & 1 + 2 & @xmath165 & x & @xmath165 & x & 17 & 5 + 3 & @xmath165 & x & @xmath165 & @xmath165 & 50 & 24 + 4 & @xmath165 & @xmath165 & x & x & 10 & 1 + 5 & @xmath165 & @xmath165 & @xmath165 & x & 9 & 2 + 6 & @xmath165 & @xmath165 & @xmath165 & @xmath165 & 48 & 16 there exist uncertainties in both the oh and h data . thus we adopt the idl procedure @xmath166 for fitting , which considers uncertainties in both x and y directions ( corresponding to @xmath167 and @xmath168(h ) , respectively ) during linear least - squares fitting . the value of the fitted slope is larger than that when x error is not considered during fitting ( see table [ table : fittingpara ] for comparison ) . as shown in table [ table : fittingpara ] , the linear fit for clouds with hinsa in figure [ fig : hi_vs_oh ] is expressed as , @xmath169(h)=@xmath170+@xmath171 . the value of the slope is 0.20 with an uncertainty of 0.20 , indicating a weak correlation . the linear fit for h halo is @xmath169(h)=@xmath172 . the value of the slope is 1.0 with an uncertainty of 0.028 , indicating a strong correlation . these results show that the correlation between oh and warm is better than that between oh and hinsa . this seems to conflict with the fact that cold rather than warm is mixed with molecular gas ( e.g. , * ? ? ? . the explanation may be in part the following . 1 . @xcite studied local dark clouds not in directions toward the galactic plane . there was thus little velocity ambiguity for these observations . in the present study of sightlines along the galactic plane , the oh velocity width was used for calculating @xmath60(h ) for halo gas . this may produce a bias toward apparent correlation . + we note that the velocity width for the halo is always larger than that of molecular tracers , e.g. , co. but no strong correlation between them is found @xcite . @xcite compared correlation between derived @xmath60(hi ) with different widths and 2mass extinction , finding the best correlation between emission and extinction at 20 around the co velocities . the width is much larger than linewidth found for co , oh , and self - absorption , making such a correlation suspicious . the logic here almost runs in a circle if one tries to study the behavior of associated 2 by only looking at the velocity range best associated with 2 . the analysis toward clouds in the galactic plane is more complicated . firstly , the extinction along a sightline represents sum of all clouds in this sightline . secondly , the brightness temperature in velocity range of a molecular cloud would be diluted by extended emission of other clouds in this sightline . thus , we adopted the oh velocity range to calculate column density of to examine the possible halo around our targets . two correlations with contrary behavior exist between hinsa and oh . firstly , hinsa content has a positive correlation with increasing molecular cloud size , which can be represented by n(oh ) . secondly , is depleted to form 2 , leading decreasing n(hinsa ) as the proportion of oh increases . if these two factors are comparable , the absence of correlation between hinsa column density and oh column density is expected . a feature in figure [ fig : hi_vs_oh ] is that @xmath60(h ) trends to saturate at @xmath173 @xmath62 when @xmath174 @xmath62 . a similar feature is seen in spider and ursa major cirrus clouds , where the asymptotic value of @xmath60(h ) is @xmath175 2 when @xmath176 2 @xcite . the asymptotic values of @xmath60(h ) between this study and @xcite are consistent but the critical value of @xmath46 in this paper is larger by two orders of magnitude . this asymptotic behavior implies that the mass of halo will be same for different clouds when the molecular core is large enough . this behavior also implies that a portion of molecular gas may be not well traced by in the halo . llllll 1 & n(@xmath16 ) vs n(oh ) & halo & 1.05 @xmath119 0.03 & 4.57 @xmath119 0.43 & 0.58@xmath119 0.07 + 2 & n(@xmath16 ) vs n(oh ) & hinsa & 0.20 @xmath119 0.20 & 15.9 @xmath119 3.1 & 0.14 @xmath119 0.18 + 3 & i(c@xmath9 ) vs n(oh ) & co - dark & 0.63 @xmath119 0.46 & -9.44 @xmath119 7.12 & 0.51 @xmath119 0.20 in hinsa and halo versus oh column density , @xmath60(oh ) . integrated emission spectra were used to derive @xmath177 in halo . the two red solid lines show linear fit for two category of samples . , scaledwidth=48.0% ] with @xmath46 on a log - log scale for 100 clouds . the green filled circles represent 24 oh clouds with @xmath18co emission alone . the blue filled squares represent 76 oh clouds with both @xmath18co and @xmath19co emission . solid , dotted , dashed , and dash dotted lines represent @xmath60(co)/@xmath60(oh ) values of 1 , 15 , 225 , and 3375 , respectively . the error bars of @xmath178 for clouds with @xmath120 detections represent upper and lower limits . for clouds with detections of both @xmath120 and @xmath121 , statistical uncertainties of the co spectrum are given . the uncertainty of @xmath2 is the same as that described in figure [ fig : hi_vs_oh].,scaledwidth=47.0% ] the oh column densities are compared with @xmath18co column densities in figure [ fig : co_vs_oh ] . there is no obvious correlation between @xmath60(@xmath18co ) and @xmath60(oh ) . a possible reason for this is that oh may reveal larger fraction of molecular gas than co , e.g. , the co - dark " gas component , the fraction of which can reach 0.3 even in co emission clouds @xcite . the ratio between co and oh column density , @xmath60(co)/@xmath60(oh ) varies from 3.7 to @xmath179 with a median value of 59 for clouds with both @xmath18co and @xmath19co emission . it varies from 0.81 to 52 with a median value of 7.1 for clouds with only @xmath18co emission . these results confirm that oh will be depleted to form co , resulting in larger @xmath60(co)/@xmath60(oh ) ratios in more massive molecular clouds . the c@xmath9 158 @xmath180 m fine - structure transition is sensitive to column density , volume density , and kinetic temperature of and 2 , making it difficult to determine column density based on present data . we adopted c@xmath9 intensity rather than c@xmath9 column density as a parameter for comparison . c@xmath9 emission can be produced by both photon - dominated regions ( pdrs ) and the ionized gas in h regions . the average ratio of c@xmath9 emission from h and pdrs in ic 342 is 70:30 @xcite . this fact is included in estimating the uncertainty of c@xmath9 intensity . we compared the relation between i(c@xmath9 ) and n(oh ) in co - dark clouds ( mask 4 ) and molecular clouds ( mask 5 and 6 ) . the clouds were divided into two categories , co - bright and co - dark . as seen in figure [ fig : cp_vs_oh ] , no correlation was found between @xmath181 and @xmath60(oh ) for co - bright category . but this comparison is limited by the large uncertainty in the c@xmath9 data and the fact that c@xmath9 traces both atomic and molecular components . for the co - dark category , the fitted slope is @xmath182 ( table [ table : fittingpara ] ) with fitted chi - squre @xmath183 , indicating a linear correlation . based on the fact that is a good tracer of 2 in not well - shielded gas ( e.g. , * ? ? ? * ) , the correlation is consistent with the suggestion that oh is a better trace of 2 than co in diffuse clouds , though the sample size is small . @xcite found that oh column density is proportional to the extinction following @xmath184 @xmath62 mag@xmath17 in @xmath185 within the range of 0.47 mag , which implies oh / h @xmath186 . the minimum value of @xmath60(oh ) found in this study is @xmath187 @xmath62 , which corresponds to an extinction of 2.3 mag . if the @xmath188 relation extends to higher extinction , the maximum and median @xmath60(oh ) values would correspond to 138 mag and 24 mag . the value of 24 mag is comparable to the largest extinction in taurus cloud @xcite while the value of 138 mag requires more dense gas . one possible reason is that the value of @xmath189 is larger than @xmath190 @xmath62 mag@xmath17 when @xmath191 mag . the other reason is that we may have overestimated @xmath60(oh ) . some satellite lines of oh show ` flip ' feature inverting from emission to absorption at a velocity . an example is shown in figure [ fig : sate_anomaly ] . this feature can be interpreted with overlap of infrared transition of oh and implies a transition column density of @xmath192 @xmath62 km@xmath17 s , where @xmath193 is the full width at half - maximum of oh line ( e.g. , * ? ? ? * ; * ? ? ? these ` flip ' features were found in three clouds of this survey . the values of n(oh ) for ` flip ' feaure in these clouds can be derived . to compare the results that derived with non - lte assumption in section [ sec : oh_col_den ] , we listed calculated @xmath60(oh ) with two different methods in table [ table : ohcolden_discuss ] . the results are consistent within a factor of 2.5 , confirming the validity of our calculation of @xmath60(oh ) in section [ sec : oh_col_den ] . lllll g033.8 - 0.5 & 11.0 & 1.3 & @xmath194 & 1.3 + g033.8 + 0.0 & 55.2 & 2.7 & @xmath195 & 2.7 + g035.1 + 1.0 & 13.2 & 1.6 & @xmath196 & 1.6 a large fraction of molecular gas is expected to exist in the transition region between the fully molecular co region and the purely atomic region . the molecular gas in this region , which is called the co - dark molecular gas " ( dmg ) , can not be traced by co , but is associated with ions or molecules that are precursors of co formation . c@xmath9 and oh are two of them . one example of a dmg cloud is shown around a velocity of 41.9 in figure [ fig : g036.4 + 0.0_spec ] . there exist oh and c@xmath9 emission without corresponding @xmath18co detections with a sensitivity of 0.07 k. twenty seven dmg clouds , which comprise 18% of all oh clouds , are identified as shown in table [ table : detections ] . this fraction is smaller than that of @xmath197 found in pilot oh survey toward outer galaxy @xcite . the co sensitivity of 0.05 k in @xcite is comparable to that in this study . but the oh sensitivity of @xmath198 mk in @xcite is ten times lower than that in this study . this comparison indicates that the dmg fraction detected depends on oh sensitivity . higher dmg fraction is expected with higher oh sensitivity . the dmg can be a significant fraction even in clouds with co emission @xcite and in observations ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? the fact that the dmg can be traced by oh may explain the absence of correlation between oh and co in figure [ fig : co_vs_oh ] . is the main reservoir of carbon in diffuse gas . it converts to co quickly through -oh chemical reactions once oh is formed ( e.g. , * ? ? ? * ) . thus and oh are expected to have a tight correlation . as shown in figure [ fig : cp_vs_oh ] , a @xmath199-@xmath60(oh ) correlation may exist for dmg clouds but not for all clouds . the atomic to molecular transition occurs in the dmg region . though hinsa other than warm gas is associated with molecular formation , the halo outside the dmg region provides shielding from uv radiation . the asymptotic value of , @xmath200 2 in figure [ fig : hi_vs_oh ] , corresponding visual extinction @xmath185 of 0.5 mag , approaching extinction @xmath201 mag ( e.g. , * ? ? ? * ) required to provide effective shielding for co formation . this extinction of 0.5 mag is much larger than that required for forming abundant 2 , which has a large self - shielding coefficient and can be the dominant form of hydrogen even when @xmath202 mag @xcite . thus a large fraction of dmg will exist before abundant co formation . we have obtained oh spectra of four 18 cm lines toward 51 got c+ sightlines with the arecibo telescope . using gaussian decomposition , we identified 151 oh components . a combined analysis of oh , co , , and hinsa reveals the following results . 1 . oh emission is detected in both main and satellite lines in the inner galactic plane but is only detected in the main lines in the outer galaxy . a large fraction of detected main lines show absorption features in the inner galaxy but no oh absorption feature was found in the outer galaxy . this is in agreement with more molecular gas and a higher level of continuum background emission being present in the inner galaxy than in the outer galaxy . 2 . there is no deviation from the ` sum rule ' by more than 3@xmath80 for all of the detected diffuse oh emission , suggesting small opacities of oh lines for clouds in the galactic plane . the column density in the cloud halos has an obvious correlation with the oh column density @xmath60(oh ) following @xmath169(h)=@xmath172 . reaches an asymptotic value of @xmath200 2 when @xmath203 2 . no correlation was found between the cold column density @xmath60(hinsa ) from narrow self - absorption feature and @xmath60(oh ) . @xmath2/@xmath204 ratios are ten times lower in translucent clouds with only @xmath18co detection than in dense clouds with both @xmath18co and @xmath19co detections . this confirms that oh is depleted to form co. no correlation between @xmath2 and @xmath204 was found . a weak correlation was found between c@xmath9 intensity @xmath205 and @xmath60(oh ) for co - dark molecular clouds . this is consistent with oh being better tracer of 2 in diffuse molecular clouds as traces 2 well in not well - shielded gas . no correlation was found for @xmath205 and @xmath60(oh ) for co - bright molecular clouds . we thank anonymous referee for significantly improving this paper by pointing out uncertainties that had been missed . this work is supported by international partnership program of chinese academy of sciences no.114a11kysb20160008 , the strategic priority research program the emergence of cosmological structures " of the chinese academy of sciences , grant no . xdb09000000 , national natural science foundation of china no . 11373038 , and national key basic research program of china ( 973 program ) 2015cb857100 , the china ministry of science . this work was carried out in part at the jet propulsion laboratory , which is operated for nasa by the california institute of technology . m. k. acknowledges the support of special funding for advanced users , budgeted and administrated by center for astronomical mega - science , chinese academy of sciences ( cams ) . n. m. m .- g . acknowledges the support of the australian research council through grant ft150100024 . the arecibo observatory is operated by sri international under a cooperative agreement with the national science foundation ( ast-1100968 ) , and in alliance with ana g. mndez - universidad metropolitana , and the universities space research association . co data were observed with the delingha 13.7 m telescope of the qinghai station of purple mountain observatory . we appreciate all the staff members of the delingha observatory and zhichen pan for their help during the observations . allen , r. j. , hogg , d. e. , & engelke , p. d. 2015 , , 149 , 123 allen , r. j. , ivette rodrguez , m. , black , j. h. , & booth , r. s. 2012 , , 143 , 97 andersson , b .- g . , wannier , p. g. , & morris , m. 1991 , , 366 , 464 barriault , l. , joncas , g. , lockman , f. j. , & martin , p. g. 2010 , , 407 , 2645 beuermann , k. , kanbach , g. , & berkhuijsen , e. m. 1985 , , 153 , 17 brand , j. , & blitz , l. 1993 , , 275 , 67 brooks , k. j. , & whiteoak , j. b. 2001 , , 320 , 465 calabretta , m. r. , staveley - smith , l. , & barnes , d. g. 2014 , , 31 , e007 clemens , d. p. 1985 , , 295 , 422 cotten , d. l. , magnani , l. , wennerstrom , e. a. , douglas , k. a. , & onello , j. s. 2012 , , 144 , 163 crutcher , r. m. 1979 , , 234 , 881 crutcher , r. m. 1977 , , 216 , 308 dame , t. m. , hartmann , d. , & thaddeus , p. 2001 , , 547 , 792 dawson , j. r. , walsh , a. j. , jones , p. a. , et al . 2014 , , 439 , 1596 dickey , j. m. , crovisier , j. , & kazes , i. 1981 , , 98 , 271 giardino , g. , banday , a. j. , grski , k. m. , et al . 2002 , , 387 , 82 gibson , s. j. , taylor , a. r. , higgs , l. a. , & dewdney , p. e. 2000 , , 540 , 851 goldsmith , p. f. , & li , d. 2005 , , 622 , 938 goldsmith , p. f. , heyer , m. , narayanan , g. , et al . 2008 , , 680 , 428 - 445 grenier , i. a. , casandjian , j .- m . , & terrier , r. 2005 , science , 307 , 1292 grossmann , v. , heithausen , a. , meyerdierks , h. , & mebold , u. 1990 , , 240 , 400 guibert , j. , rieu , n. q. , & elitzur , m. 1978 , , 66 , 395 haslam , c. g. t. , salter , c. j. , stoffel , h. , & wilson , w. e. 1982 , , 47 , 1 heiles , c. 1969 , , 157 , 123 heiles , c. e. 1968 , , 151 , 919 jackson , j. m. , bania , t. m. , simon , r. , et al . 2002 , , 566 , l81 kro , m. , goldsmith , p. f. , brown , r. l. , & li , d. 2008 , , 689 , 276 - 289 kro , m. , & goldsmith , p. f. 2010 , , 724 , 1402 langer , w. d. , velusamy , t. , pineda , j. l. , et al . 2010 , , 521 , l17 langer , w. d. , velusamy , t. , pineda , j. l. , willacy , k. , & goldsmith , p. f. 2014 , , 561 , a122 lee , m .- y . , stanimirovi , s. , douglas , k. a. , et al . 2012 , , 748 , 75 li , d. , & goldsmith , p. f. 2003 , , 585 , 823 li , d. , xu , d. , heiles , c. , pan , z. , & tang , n. 2015 , publication of korean astronomical society , 30 , 75 liszt , h. , & lucas , r. 1996 , , 314 , 917 liszt , h. s. , & pety , j. 2012 , , 541 , a58 mather , j. c. , cheng , e. s. , cottingham , d. a. , et al . 1994 , , 420 , 439 milam , s. n. , savage , c. , brewster , m. a. , ziurys , l. m. , & wyckoff , s. 2005 , , 634 , 1126 myers , p. c. 1975 , , 198 , 331 nakanishi , h. , & sofue , y. 2003 , , 55 , 191 nguyen - q - rieu , winnberg , a. , guibert , j. , et al . 1976 , , 46 , 413 peek , j. e. g. , heiles , c. , douglas , k. a. , et al . 2011 , , 194 , 20 penzias , a. a. 1964 , , 69 , 146 pineda , j. l. , goldsmith , p. f. , chapman , n. , et al . 2010 , , 721 , 686 pineda , j. l. , langer , w. d. , velusamy , t. , & goldsmith , p. f. 2013 , , 554 , a103 planck collaboration , ade , p. a. r. , aghanim , n. , et al . 2011 , , 536 , a19 qian , l. , li , d. , & goldsmith , p. f. 2012 , , 760 , 147 rllig , m. , simon , r. , gsten , r. , et al . 2016 , , 591 , a33 reich , p. , & reich , w. 1986 , , 63 , 205 robinson , b. j. , & mcgee , r. x. 1967 , , 5 , 183 roman - duval , j. , jackson , j. m. , heyer , m. , et al . 2009 , , 699 , 1153 tang , n. , li , d. , heiles , c. , et al . 2016 , , 593 , a42 turner , b. e. 1973 , , 186 , 357 turner , b. e. , & heiles , c. 1971 , , 170 , 453 turner , b. e. 1979 , , 37 , 1 van dishoeck , e. f. , & black , j. h. 1988 , , 334 , 771 wannier , p. g. , andersson , b .- g . , federman , s. r. , et al . 1993 , , 407 , 163 weinreb , s. , barrett , a. h. , meeks , m. l. , & henry , j. c. 1963 , , 200 , 829 wolfire , m. g. , hollenbach , d. , & mckee , c. f. 2010 , , 716 , 1191 the column densities of upper and lower levels of each transition , @xmath206 and @xmath207 , are related by where @xmath209 is the excitation temperature of oh transition line , @xmath206 and @xmath207 are partition weight the oh transition , and @xmath210 and @xmath211 are the statistical weights of i and j levels , respectively . for oh lines , @xmath212 , where f is the total angular momentum quantum number . the transition of 1612 mhz gives @xmath213 @xmath214 and @xmath215 are upper and lower level of 1612 mhz line as shown in figure [ fig : rp - pth ] . with @xmath225 for oh lines in general , we derive @xmath226 when most oh molecules are in the ground state of @xmath227 , the total oh column density , @xmath228 , is sum of molecules in four energy levels , @xmath228=@xmath229=@xmath230=@xmath231 . @xmath232 , where @xmath233 is peak optical depth and @xmath106 is full width at half maximum of transition line . combining equations [ eq : nu_tex ] , equation [ eq : nu_tij ] , and values of four oh transition coefficients , we derive where @xmath235,@xmath236,@xmath237,and @xmath58 are peak optical depth of four oh transitions . the only requirement for equation [ eq : sum_rule ] is @xmath225 . thus equation [ eq : sum_rule ] is valid even for non - lte conditions . the brightness of emission line , @xmath144 , is calculated through @xmath238 , where @xmath45 is background continuum temperature , @xmath52 is optical depth of transition line . when @xmath239 is the same for the four oh lines , which is valid under lte conditions , we derive the ` sum rule ' for brightness temperature under optically thin conditions to estimate the contribution of non - lte and optical depth leading to deviation from equation [ eq : sum_rule2 ] , we plot deviation fraction and maximum optical depth of oh as function of @xmath241 and @xmath60(oh ) in figure [ fig : sum_rule_contour ] . the deviation from the sum rule ( dsr ) is significant when @xmath60(oh ) @xmath242 2 . in this parameter space with significant deviation , the optical depth of oh lines @xmath243(oh ) is greater than 0.5 . we conclude that the condition of large optical depth results in significant dsr . = [ @xmath244+@xmath245-@xmath246-@xmath247/@xmath248 , as a function of @xmath241 and total oh column density @xmath249(oh ) . @xmath250 mk is the rms of the summed spectra . @xmath251 k , @xmath252 k , and continuum brightness temperature at 1667 mhz of 5.0 k are assumed . @xmath96 at 1612 mhz is calculated through @xmath253=3.1+@xmath254 . the continuum brightness at 1720 mhz is calculated using similar method . _ bottom : _ contour of the maximum optical depth of oh lines . , title="fig:",scaledwidth=47.0% ] = [ @xmath244+@xmath245-@xmath246-@xmath247/@xmath248 , as a function of @xmath241 and total oh column density @xmath249(oh ) . @xmath250 mk is the rms of the summed spectra . @xmath251 k , @xmath252 k , and continuum brightness temperature at 1667 mhz of 5.0 k are assumed . @xmath96 at 1612 mhz is calculated through @xmath253=3.1+@xmath254 . the continuum brightness at 1720 mhz is calculated using similar method . _ bottom : _ contour of the maximum optical depth of oh lines . , title="fig:",scaledwidth=47.0% ]

we present the results of a time - series analysis of 130 chelle spectra of ori ( b0 ia ) , acquired over seven observing seasons between 1998 and 2006 at ritter observatory . the equivalent widths of h ( net ) and he i were measured and radial velocities were obtained from the central absorption of he i . temporal variance spectra ( tvs ) revealed significant wind variability in both h and he i . the he i tvs have a double - peaked profile consistent with radial velocity oscillations . a periodicity search was carried out on the equivalent width and radial velocity data , as well as on wavelength - binned spectra . this analysis has revealed several periods in the variability with time scales of . many of these periods exhibit sinusoidal modulation in the associated phase diagrams . several of these periods were present in both h and he i , indicating a possible connection between the wind and the photosphere . due to the harmonic nature of these periods , stellar pulsations may be the origin of some of the observed variability . periods on the order of the rotational period were also detected in the he i line in the 98 - 99 season and in both lines during the 04 - 05 season . these periods may indicate rotational modulation due to structure in the wind .

1301.5368

early - type supergiant stars are the most luminous stars in the universe , and their intense radiation is responsible for driving their powerful stellar winds . absorption of uv photons by metal ions ( e.g. n iv , o iv , fe - group elements ) transfers momentum to the ions . winds driven by this mechanism are known as line - driven winds . in the 1970 s it was realized that , if the momentum gained by the metal ions could be shared with the hydrogen and helium ions , which are more abundant , then this process could lead to significant mass loss ( lucy & solomon , 1970 ; castor , abbott & klein , 1975 ) . the global properties of line - driven winds are successfully described by a spherically symmetric , homogeneous model wind with a velocity that increases radially outward . this model leads to the classic p cygni profiles indicative of mass loss . an example of a p cygni profile for @xmath0 ori can be seen in the spectrum of 2002 february 6 shown in figure [ fig1 ] . however , some observations can not be explained by this simple model . the observed x - ray emission from hot stars is believed to be due to shocks within the wind , which may be due to the inherent line - driven instability of these winds , first described by lucy & solomon ( 1970 ) . the observed electron - scattering wings of recombination lines in wolf - rayet stars are too weak to be matched by synthetic line profiles unless wind clumping is included in the models ( hillier ( 1991 ) . finally , line profile variability is observed in the spectra of early - type stars . this variability , typically observed in uv and optical lines , may be related to structure in the winds , stellar pulsations , or magnetic activity . line profile variability in ob stars was firmly established with observations from the iue satellite in unsaturated p cygni profiles in the ultraviolet . the emission component is fairly constant while absorption features are superimposed on the broad , blue absorption troughs . commonly referred to as discrete absorption components ( dacs ) , these features typically move blueward across the profile and often recur at regular intervals on the order of the estimated rotational period . their cyclical nature is thought to be due to co - rotating interaction regions ( cirs ) ( mullan , 1984 ; cranmer & owocki , 1996 ; kaper et al . the cirs may form when a density enhancement at the base of the wind propagates outward and forms a spiral density pattern in the wind as the star rotates . the mechanism responsible for the formation of the cirs in hot stars is not known but is likely due to disturbances at the photosphere . cantiello et al . ( 2009 ) propose that the cause may be tied to magnetic fields arising in sub - surface convection zones and that micro - turbulence due to these convection zones may be responsible for inducing wind clumping near the photosphere . the h@xmath1 line profiles of hot massive stars are also known to be highly variable . the wind contribution of the h@xmath1 line , being mainly formed by recombination and therefore proportional to the square of the electron density , is formed in the dense inner wind region . in the case of weak wind emission , the observed h@xmath1 profiles often alternate between p cygni , inverse p cygni , pure emission , and double absorption . variability in optical photospheric lines such as he i @xmath25876 is also observed . fullerton et al . ( 1996 ) found that nearly 23 out of 30 o - stars in their sample had variable photospheric lines . the time scales of the photospheric and wind variability are often comparable , suggesting a connection between phenomena in the two regions . one mechanism for variability in the photospheric lines is radial or non - radial pulsation . in addition , correlations between uv and h@xmath1 variability ( kaper et al . 1997 ) and between x - ray and h@xmath1 variability ( berghoefer et al . 1996 ) have been reported in o - stars . this implies the propagation of structures or disturbances throughout the wind . the search for a photosphere - wind connection in massive stars is of current interest . stellar pulsations may affect the mass loss by coupling to the wind ( townsend 2007 ) . in addition , townsend ( 2000a , 200b ) found that non - radial pulsations can `` leak '' into the base of the wind , possibly inducing the formation of wind structure that could lead to a cir . kaufer et al . ( 2006 ) studied the variability of optical lines in the b0 ib supergiant hd 64760 , which is similar in spectral type to @xmath0 ori ( b0 ia ) . they identified three closely spaced frequencies around in photospheric lines and demonstrated that these frequencies led to a beating pattern with a period of . this period was observed in photospheric lines as well as h@xmath1 . multi - periodic , non - radial pulsation modeling showed that three closely spaced , high - order modes could explain the beat period . non - radial @xmath3- and @xmath4-modes have been detected in early b - type supergiants ( lefever et al . 2007 ; saio et al . 2006 ) . using the @xmath5 satellite , saio et al . detected @xmath3- and @xmath4-mode pulsations in the b2 ib / ii supergiant hd 163899 with periods ranging from @xmath60.35 to . stellar modeling by saio et al . of supergiants between 12 and 20 m@xmath7 show that both @xmath3- and @xmath4-modes are excited in these stars . also known as alnilam , @xmath0 ori ( hd 37128 ) is the middle star of orion s belt and is a fairly typical early b supergiant . some basic properties of the star are given in table 1 . though it is not known to be a binary , morrell and levato ( 1991 ) and jarad et al . ( 1989 ) give conflicting reports of radial velocity variability for the system . it has been reported to be moderately deficient in nitrogen , along with the other two belt stars , @xmath8 and @xmath9 ori , by walborn ( 1976 ) . @xmath0 ori is one of the few early - type supergiants to have a measured radio flux density of the wind ( scuderi et al . 1998 ; blomme et al . 2002 ) . there have been two recent quantitative studies of early b supergiants based on uv and optical spectra and nlte model atmosphere codes ( crowther et al . 2006 ; searle et al . 2008 ) that included @xmath0 ori . the latter study included the effects of wind clumping . some nlte derived parameters for this star reported in those two papers are listed in table 2 . it is interesting to note the discrepancy between the @xmath10 and radius values found in these two studies . searle et al . found a larger gravity as well as a larger radius , so a major discrepancy between the masses is implied . llllc b0 ia & 1.69 & @xmath11 & @xmath12 & 91 lcllcccc 27000 & 2.9 & 5.44 & -6.3@xmath130.5 & 24.0 & 1910 & 12.5 & 1 + 27500 @xmath13 1000 & 3.13 & 5.73 @xmath13 0.11 & -6.89 @xmath13 0.05 & 32.4 @xmath13 0.75 & 1600 & 15 & 2 variability in @xmath0 ori has been studied at multiple wavelengths including the x - ray , uv , optical , and radio regions of the spectrum . cassinelli et al . ( 1983 ) made simultaneous x - ray and uv observations of @xmath0 and @xmath14 ori in order to study variability in the x - ray flux and uv line profiles . neither star exhibited variability on time scales of 100 - 5000 s. ultraviolet observations from the iue satellite showed variability on the time scales of @xmath610 hours in the form of dacs and narrow absorption components ( prinja et al . 2002 ; sapar & sapar 1998 ) . the iue time series were not extensive enough to allow for a study over longer time scales . blomme et al . ( 2002 ) carried out an extensive radio study of @xmath0 ori . they found no evidence of radio variability and found an average 6 cm flux of @xmath15 mjy . blomme et al . reported flux measurements at millimeter wavelengths that are larger than what is expected for a smooth wind model . the most extensive studies of @xmath0 ori concern the variability of optical line profiles , particularly h@xmath1 . the h@xmath1 emission in @xmath0 ori was first noted by cherrington ( 1937 ) in which he described the line as having a `` peculiar structure . '' the variability of h@xmath1 in @xmath0 ori was studied in some detail by ebbets ( 1982 ) as part of a survey of variability in early - type stars . ebbets noted that the broad emission in h@xmath1 typically extended from -800 km s@xmath16 to 800 km s@xmath16 and the average equivalent width of h@xmath1 for the 6 observations of @xmath0 ori was -0.56 with a range of -0.01 to -0.96 . the time scale for the observed variability in h@xmath1 and the photospheric lines was 1 - 10 d. morel et al . ( 2004 ) conducted a time - series analysis of h@xmath1 line profile variability for 22 ob supergiants . their study included 21 observations of @xmath0 ori obtained on 17 nights between nov . 18 , 2001 and may 7 , 2002 . they conducted a periodicity search on the measured net equivalent widths of h@xmath1 and the pixel - by - pixel h@xmath1 time series . significant variability was observed in h@xmath1 and he i @xmath26678 , as quantified using temporal variance spectra . on several nights multiple observations were obtained for @xmath0 ori and significant changes in the h@xmath1 profile were observed on time scales of a few hours . their period analysis found periods of and in the h@xmath1 data . the period is on the order of the rotational period of @xmath0 ori , for which morel et al . quote a maximum of . prinja et al . ( 2004 ) quote a maximum rotational period of . a more extensive time - series analysis of h@xmath1 and photospheric lines in @xmath0 ori was conducted by prinja et al . their study was primarily based on 61 spectra obtained over 17 nights between nov . 26 and dec . 13 , 1998 , supplemented by archival spectra taken over several months in 1996 . in addition to h@xmath1 , their data included the absorption lines h@xmath17 , h@xmath18 , he i @xmath24713 , he i @xmath25876 , he i @xmath26678 , c ii @xmath24267 , and si iii @xmath24553 . radial velocity variations were observed in all of the absorption lines . the period analysis involved constructing 2d ( pixel - by - pixel ) periodograms using the clean algorithm . they report periods of 1.9 , 6.6 and , each identified in h@xmath1 , h@xmath17 , and he i @xmath26678 . finally , they carried out a non - radial pulsation study of the period to see if non - radial pulsations could replicate the observations . the authors concluded that non - radial pulsations are not sufficient to explain the observed variability for the period . this paper examines the line profile variability of the h@xmath1 and he i @xmath25876 lines of @xmath0 ori . the remainder of this paper is organized in the following manner . section [ sec3 ] describes the instrumentation , observations , and data reduction . section [ sec4 ] discusses the temporal variance spectra and the equivalent width and radial velocity measurements . details of the line profile variability are described in section [ sec5 ] . a discussion of the techniques used in the period search is presented in section [ sec6 ] . the results of the period search are presented in section [ sec7 ] . finally , a summary of the findings and implications for future work are given in section [ sec8 ] . lccc 1998 - 1999 & 15 & 8.2 & 115 + 2000 - 2001 & 14 & 12.5 & 163 + 2001 - 2002 & 26 & 6.0 & 145 + 2002 - 2003 & 21 & 6.4 & 128 + 2003 - 2004 & 10 & 16.5 & 149 + 2004 - 2005 & 24 & 8.9 & 205 + 2005 - 2006 & 22 & 9.8 & 196 one hundred and thirty spectra of @xmath0 ori were obtained at ritter observatory using the 1-meter ritchey - chrtien telescope and fiber - fed chelle spectrograph . target acquisition and manual guiding were done with a sbig st-9 ccd camera . the detector for the spectrograph was a front - illuminated eev ccd in a wright instruments camera system . the ccd had 1200 @xmath19 800 pixels with physical dimensions of 22.5 @xmath19 22.5 @xmath20 m . the detector was cooled with liquid nitrogen to an operating temperature of 140 k. the entrance slit of the spectrograph projected to a width of 4.3 pixels and corresponded to a resolving power r = 26,000 . in its usual configuration , nine partial chelle orders were included on the chip with h@xmath1 falling in the center of the lowest - numbered order on the ccd . the included chelle orders , each spanning @xmath670 , covered the spectral range from 5285 to 6594 . in this paper , spectra of @xmath0 ori from seven observing `` seasons '' between 1998 and 2006 are analyzed . there were no observations during the 1999 - 2000 season . in toledo , ohio , an observing season for orion spans roughly from september to april . a typical season from this data set includes @xmath620 spectra , each with a signal - to - noise ratio ( s / n ) of @xmath6100 or better per pixel in the continuum . the average interval between observations is @xmath610 nights . prominent features in the ritter spectra of @xmath0 ori include the h@xmath1 line , the he i 5876 d3 multiplet , as well as interstellar na i d lines . weak c ii lines at 6578 and 6583 are visible in a few spectra . this work examines the line profile variability of h@xmath1 and he i @xmath25876 . the time series are summarized in table 3 . the raw chelle spectra were reduced in iraf with the observatory s pipeline script . the standard reductions included ( in this order ) bias subtraction , chelle order extraction , flat fielding , and wavelength calibration . the wavelengths were calibrated using comparison spectra from a th - ar discharge lamp . the rms of the calibration fit is typically between 0.001 - 0.005 ( mulliss , 1996 ) . further operations on spectra produced by the pipeline included the removal of telluric lines , corrections for the motion of the earth about the sun , and normalization to the continuum . beginning with data taken in the spring of 2002 , there was a problem with the wavelength calibration coming out of the standard pipeline . it is suspected that the problem began after the observatory replaced the th - ar lamp . as a result , beginning with the 2002 - 2003 season , the spectra had to be re - calibrated . the re - calibration used known telluric water lines that appear in the spectra in same way that th - ar lines were used in the original calibration . the wavelengths of the water lines used were measured in the telluric spectrum provided in fits format by hinkle et al . the water lines were identified in the spectra using the iraf task _ identify _ and the known wavelengths were fitted to the observed spectrum . the dispersion solution was then applied to the spectrum to correct the wavelength scale . this procedure was carried out separately for the two apertures containing h@xmath1 and he i @xmath25876 . the he i @xmath25876 and h@xmath1 spectra from the 2001 - 2002 season are shown in figures [ fig1 ] and [ fig1 ] , respectively . spectra from all seven observing seasons are available in the online version of the journal . cccc 1149.803 & -0.3801 & 1.139 & 11.8 + 1163.736 & -0.1594 & 1.080 & 8.53 + 1208.596 & -0.0179 & 1.125 & 11.7 + 1214.670 & -0.0530 & 1.215 & 7.20 + 1220.648 & -0.5581 & 1.010 & 0.52 in order to quantify the line profile variability , temporal variance spectra were calculated for h@xmath1 and he i line for each season . the temporal variance spectrum was first outlined in fullerton ( 1990 ) and fullerton , gies and bolton ( 1996 ) . the advantage of the technique is that it allows one to determine the level of statistical significance of detected line profile variability . the tvs method accounts for spectrum - to - spectrum and pixel - to - pixel differences in data quality . the tvs of a time series of @xmath21 spectra has the same statistical distribution as @xmath22 , where @xmath23 is the chi - squared distribution with @xmath24 degrees of freedom . thus , it is possible to calculate the significance level for a given value of the tvs from the appropriate @xmath22 distribution . it is then easy to determine the wavelength or velocity range over which significant variability occurs . the he i and h@xmath1 temporal variance spectra for all seasons are shown in figures [ fig2 ] and [ fig3 ] . the tvs@xmath25 is plotted because it scales linearly with the spectral standard deviations . the region of significant variability at the 99% confidence level was identified from the tvs for each season for both h@xmath1 and the he i 5876 triplet . the range of significant variability was then used in defining the wavelength region over which the equivalent width , @xmath26 , was measured . the goal was to define a consistent wavelength range for each line that would be used for all @xmath26 measurements without excluding any significant variability . the bluest extent of significant variability , from any of the seasons , defined the blue edge , @xmath27 , of the region used to measure the equivalent widths . similarly , the reddest extent of significant variability , from any season , defined the red edge , @xmath28 , of the region used to measure the equivalent widths . this was done for both the h@xmath1 and the he i @xmath25876 lines . the equivalent width is then defined as @xmath29 the equivalent widths of each line were measured over the ranges of 6554.25 to 6571.75 and 5871.7 to 5879.8 for h@xmath1 and he i , respectively . for h@xmath1 , the net equivalent width was measured which included both emission and absorption components . for this reason , the @xmath26 values for h@xmath1 are small , often close to zero . the equivalent width data for both h@xmath1 and he i may be found in table 4 . the errors in the equivalent width measurements were computed using the method given in chalabaev & maillard ( 1983 ) . for he i , the errors were @xmath60.035 or @xmath63% . for h@xmath1 , the errors were @xmath60.089 . since he i @xmath25876 is primarily a photospheric absorption line the radial velocity of the absorption core was measured to search for variability . absorption from the wind is present in the wings of the line , particularly at the blue edge . however , the line is symmetrical at normalized flux levels below @xmath60.9 , suggesting a negligible wind contribution there . only this central region of the line , from the flux minimum up to 0.9 in normalized flux , was used in the radial velocity determinations . the centroid of the line core was calculated using iraf and the wavelength was converted to a radial velocity using a rest wavelength of 5875.704 . this method attempts to minimize any wind contribution to the radial velocity variability . however , in cases of particularly strong wind absorption or emission , it is possible that the wind may contribute to variability in the line core . the radial velocity data may be found in table 4 . the errors for the radial velocity measurements were calculated using the method of brown ( 1990 ) . a typical value for @xmath30 is @xmath60.25 km s@xmath16 . previous radial velocity studies using the same instrument at ritter observatory have found an instrumental error of 0.25 km s@xmath16 due to motions of the ccd relative to the spectrum . the instrumental errors and the measurement errors were added in quadrature to produce the total error of the radial velocity data . the spectra shown in figures [ fig1 ] and [ fig1 ] demonstrate the variability in the h@xmath1 and he i @xmath25876 lines of @xmath0 ori . the h@xmath1 profile is highly variable , often appearing as p cygni , nearly pure emission , or double absorption . in contrast to h@xmath1 , the he i line is predominantly an absorption line with weak wind variability present in the wings . enhanced absorption is often observed in the blue wing of the line and a slight emission bump is sometimes seen at the red edge of the line . the strength of the absorption line is fairly constant but the profile often oscillates back and forth . prinja et al . ( 2004 ) noted this `` swaying '' in the profiles of several absorption lines , including he i @xmath25876 . since it is the variable wind emission that dominates the measured h@xmath1 equivalent widths , it is worth while to examine the distribution of the equivalent width values to gain insight into the typical wind emission . the left frame of fig . [ fig4 ] shows a histogram of all 130 h@xmath1 @xmath26 measurements , where the values have been put into 0.1 bins . the distribution is approximately gaussian , with the exception of a slight tail extending beyond -0.6 . the peak of the distribution is near -0.1 , indicating a slight preference toward net emission in the profile . the right side of fig . [ fig4 ] shows the cumulative distribution function of the h@xmath1 @xmath26 values . the cumulative distribution function ( cdf ) indicates that net emission is observed approximately 65% of the time . the temporal variance spectra , shown in figures [ fig2 ] and [ fig3 ] , provide more insight into the observed variability . the horizontal line indicates the level of significant variability at the 99% confidence level of the @xmath22 distribution . the vertical dashed lines mark the @xmath31 of @xmath1391 km s@xmath16 ( howarth et al . the velocity scale has been converted to the stellar rest frame . significant variability extends beyond the @xmath31 threshold in both lines . the range of significant variability for h@xmath1 extends approximately -350 to 300 km s@xmath16 while for he i , the range is roughly -200 to 175 km s@xmath16 . the variability in both lines is confined to velocities much less than the terminal wind speed of 1600 km s@xmath16 ( searle et al . 2008 ) . the h@xmath1 tvs are fairly symmetric , with the exception of the 2003 - 2004 season , and the peaks are shifted toward the blue . this indicates that the bulk of the observed variability occurs in the column of wind material approaching the observer . the he i tvs are characterized by double peaks with the central minimum occurring near rest velocity as one would expect for a line undergoing radial velocity variations . the double peaks are nicely contained within the @xmath1391 km s@xmath16 threshold , consistent with a photospheric origin for the radial velocity variations . it should be noted that the projected rotational velocity is an upper limit , since the line broadening is likely due to macroturbulence in addition to rotation ( ryans et al . 2002 ) . the he i tvs tend to be slightly asymmetric with broader wings on the blue side , likely due to the wind absorption often seen on the blue edge . the h@xmath1 photospheric absorption is heavily filled in by wind emission and photospheric variability is masked by the wind . to better understand the underlying photospheric absorption , a tlusty photospheric model with a superimposed @xmath0 ori h@xmath1 spectrum is shown in fig . the spectrum , from 2001 dec 21 , is an example of a minimum emission profile . the synthetic profile is from the bstar2006 grid of nlte line - blanketed model atmospheres ( lanz & hubeny 2007 ) . the model has parameters very similar to those of @xmath0 ori with @xmath32 k , @xmath33 , @xmath34 km s@xmath16 , and solar metallicity ( see table 2 and references therein ) . the model profile has not been rotationally broadened . even when the emission is weak , it is clear that the absorption is nearly completely filled in by wind emission . variability in the strength of wind emission gives rise to the many observed profile types . for example , strong emission gives rise to the emission or p cygni profiles seen in the nov 22 and feb 6 spectra of fig . weaker wind emission leads to the double absorption - type profile in the dec 21 spectrum shown in fig . since wind emission is the dominant factor in shaping the observed profile of h@xmath1 , most of the variability in the measured net equivalent widths is likely attributable to the wind . however , in the case of constant wind emission , variability in the measured equivalent widths could be due to a variable photospheric component . in contrast to h@xmath1 , the morphology of the he i @xmath25876 line is dominated by photospheric absorption . however , weak wind features are observed as absorption in the red and blue wings and emission in the red wing . it is important to note that the range used in measuring the he i equivalent widths ( 5871.7 - 5879.8 ) included a portion of these wind signatures . the absorption in the blue wing seen in most of the spectra is fairly broad and does lie within the region of significant variability . wind absorption is also observed on the red edge as well , as seen in the dec 22 spectrum of fig . this absorption in the red occurs over a narrower region but also lies within the region of significant variability . weak emission in the red wing , as seen in the nov 6 spectrum of fig . [ fig1 ] , is observed much less frequently and occurs right on the edge of the region of significant variability , near 5880 . this emission also occurs at the edge of the region over which the equivalent widths are measured ; therefore , any contribution would be very minor and may not be significant . boyajian et al . ( 2007 ) observed a correlation between h@xmath1 emission and radial velocity changes in the he i @xmath26678 line in the b3 ia supergiant hd 61827 . they attribute the radial velocity variations to weak wind emission in the red wing of the he i line causing an apparent shift of line center toward the blue . therefore , although the core of the he i @xmath25876 line in @xmath0 ori appears to be photospheric , the variability in the measured radial velocities could have a wind origin . the equivalent widths of h@xmath1 and he i , and the radial velocities were plotted against one another to look for any correlations between them . these plots are shown in fig . . there does appear to be a weak , positive correlation between the equivalent widths of h@xmath1 and he i , with a pearson s correlation coefficient @xmath35 = 0.29 . if a connection exists between activity in the photosphere and the wind , one would expect to find such a correlation . a variable wind component would have a smaller effect on he i than on h@xmath1 and the small slope of 0.15 seems to support that . there is at best a very weak correlation between either @xmath36 and radial velocity or @xmath37 and radial velocity , with correlation coefficients of 0.0062 and 0.014 , respectively . this suggests that the he i radial velocity measurements are not contaminated by wind emission and do originate in the photosphere . vs. @xmath37 , @xmath36 vs. @xmath38 , and @xmath37 vs. @xmath38 . the radial velocities are from the he i line . ] in the present work , the most important tool for finding periodicities is the periodogram . two standard techniques for handling unevenly sampled data such as these are the lomb - scargle periodogram ( lomb , 1976 ; scargle , 1982 ) and the clean algorithm ( roberts et al . another approach , outlined by bretthorst ( 2001a , 2001b ) , combines bayesian inference with the lomb - scargle ( ls ) periodogram . all three methods were employed in the analysis presented here . the period search was done on the data from each season except for 2003 - 2004 . that season only had 10 observations with an average time of between observations and it was felt that this was insufficient sampling for a proper analysis . a global analysis was also done on the combined data from all seasons , including 2003 - 2004 , but no long - term trends were found . bretthorst ( 2001a , 2000b ) showed that the lomb - scargle ( ls ) periodogram could be derived in a general way by applying bayesian probability theory to the lomb model ( lomb , 1976 ; scargle , 1982 ) . for astronomical time series , assuming a sinusoidal signal , bretthorst s algorithm yields the ls periodogram as well as the bayesian probability of the peaks in the ls periodogram . the algorithm assumes a jeffreys prior which represents equal probability per decade in @xmath39 ( gregory , 2005 ) . a plot of the resulting bayesian probability density ( pdf ) as a function of frequency resembles a typical periodogram but many of the smaller peaks are suppressed . for simplicity , the plot of the bayesian probability of the ls periodogram will be referred to as the _ bayesian periodogram_. the lomb - scargle and bayesian periodograms were calculated for the equivalent width data of both h@xmath1 and he i @xmath25876 as well as the radial velocities from the he i line . the lomb - scargle and bayesian periodograms were calculated using _ mathematica _ code adapted from files found in gregory ( 2005 ) . prior to calculating the periodograms , the average of each time series was subtracted . since the smallest interval between successive observations was , the frequency range was limited to @xmath40 d@xmath16 , because 0.5 d@xmath16 is the nyquist frequency . the bayesian analysis covered the frequency range @xmath41 d@xmath16 , with steps of @xmath42 . the probability density was normalized so that @xmath43 where the frequency units are d@xmath16 . the strongest peak in the bayesian periodogram was used to locate the first frequency identified in a time series . a gaussian was fit to the peak to determine the central frequency of the peak . once the central frequency was identified , a least - squares sine fit at that frequency was made to the data . this fit was then used to subtract that signal from the data . the average of the resulting secondary time series was subtracted and the analysis was repeated . this procedure was continued through three cycles so there is a first , second , and third frequency for each time series . with each signal subtracted the residual data set becomes more noisy . extending the analysis beyond three cycles was not productive . as an example of the results , figure [ fig7 ] shows the lomb - scargle and bayesian periodograms for the h@xmath1 equivalent width data from the 2001 - 2002 season . this bayesian algorithm was also used to calculate two - dimensional bayesian periodograms . before computing the periodograms , the average spectrum for each season was subtracted . the residual spectra were put into 1 bins creating a time series for each wavelength bin . the bayesian periodogram was computed for each of these time series . the bins covered the wavelength ranges of 6550 to 6575 and 5868 to 5883 for h@xmath1 and he i , respectively . these regions extend beyond the range of significant variability defined by the tvs for both lines . however , only periods detected within the regions of significant variability are reported . as an aid in identifying the strongest peaks in the 2d periodograms , the probability densities at each frequency were summed over wavelength . for brevity , we refer to this sum as the summed pdf power as a function of frequency . the summed pdf power was then plotted as a function of frequency to create a pdf power spectrum . figure [ fig8 ] shows the 2d periodogram of the he i spectra from the 2000 - 2001 season along with the corresponding summed pdf power spectrum . the 2d periodograms often contained peaks at numerous frequencies . it was desirable to limit the analysis to the strongest of these peaks . we wanted to identify the frequencies where the most variability occurred and the wavelength bin in which that variability was the strongest . it was found that a pdf power of 300 or more was a sufficient criterion for selecting the frequencies of strongest variability . at this level , we were able to identify at least one wavelength bin containing a peak whose bayesian probability was 0.2 or more for the selected frequency . since the power was often spread across many wavelength bins , the bin with the highest peak was chosen . further , this wavelength bin had to lie within the region of significant variability . a period search using clean was performed on the h@xmath1 and he i @xmath25876 equivalent width data . as with the bayesian analysis , the clean algorithm was run on mean subtracted data and the search was limited to frequencies less than the nyquist frequency . the clean analysis used the range @xmath40 d@xmath16 in frequency steps of @xmath42 . the same frequency step was used to maintain consistency between the bayesian and clean approaches . the clean gain used was 0.5 and the number of iterations was set to 1000 . the central frequency of each peak was found by fitting a gaussian to the peak . for each frequency found , a least - squares sine fit at that frequency was made . the fit was then used to remove that signal from the data and clean was run on the new data set . again , this cycle was repeated three times . determining the significance of peaks in a periodogram is a well - known and difficult problem . a standard metric for determining the significance of periodogram peaks is the false alarm probability ( fap ) . the false alarm probability as defined by scargle ( 1982 ) depends on a correct determination of the number of independent frequencies at which the periodogram is evaluated . horne & baliunas ( 1986 ) determined an empirical formula for calculating the number of independent frequencies , based on their monte carlo simulations . using the results of horne & baliunas ( 1986 ) , the 5% false alarm probability level ( or 95% confidence ) was calculated for each lomb - scargle periodogram . the calculated fap values should be viewed only as an estimate of the confidence level for each peak . every period reported from the bayesian analysis had a corresponding peak in the ls periodogram that exceeded the 95% confidence level . a simple way of determining whether a detected period is persistent throughout a time series is to plot the data phase - folded at the period of interest . the phase was defined as @xmath44\ ] ] where @xmath45 is the date of the observation , @xmath46 is the date representing zero phase , and @xmath47 is the period of interest . if a detected signal with period @xmath47 is harmonic , then a sinusoidal modulation should be evident in the phase - folded data . phase plots were created for all detected periods . for the equivalent width and radial velocity data , the mean - subtracted data were pre - whitened to remove the detected signals not being used in the current phase plot . for phase plots for periods found in the 2d periodograms , the mean - subtracted flux from the bins was phase folded . all of the frequencies detected in the analysis were converted to periods and are listed in tables 5 and 6 . the periods from each season are listed according to data type and periodogram method . the periods from the equivalent width and radial velocity data are listed in order of their detection ( i.e. first , second , or third run through ) . periods from the 2d periodograms are simply listed in ascending order of the wavelength bin . tables 5 and 6 also give the bayesian probability for the periods that were detected in bayesian periodograms . these probabilities were determined by simple numerical integration over the peak . in order to provide an objective assessment of whether a sinusoidal trend is present in a phase diagram , a function of the form @xmath48 was fit to the data phased with a given period . the parameters @xmath49 , @xmath50 , and @xmath51 were freely varied . the correlation coefficient , @xmath35 , from the fit is also given in tables 5 and 6 for each period . cccccccccccccccccccccccc bp & 73.4 & 0.16 & 0.60 & & 3.06 & 0.15 & 0.76 & & 4.38 & 0.40 & 0.40 & & 11.2 & 0.16 & 0.64 & & 19.3 & 0.86 & 0.63 & & 4.32 & 0.29 & 0.56 + & 4.48 & 0.12 & 0.72 & & 7.80 & 0.13 & 0.59 & & 41.6 & 0.26 & 0.28 & & 4.28 & 0.50 & 0.55 & & 4.07 & 0.13 & 0.54 & & 2.60 & 0.12 & 0.45 + & 13.1 & 0.17 & 0.52 & & 40.6 & 0.18 & 0.72 & & 10.7 & 0.28 & 0.23 & & 4.66 & 0.14 & 0.38 & & 8.65 & 0.21 & 0.35 & & 4.39 & 0.44 & 0.45 + cl & 13.4 & & 0.50 & & 3.07 & & 0.42 & & 38.4 & & 0.20 & & 4.26 & & 0.59 & & 19.1 & & 0.63 & & 2.24 & & 0.17 + & 3.51 & & 0.50 & & 4.12 & & 0.32 & & 4.43 & & 0.18 & & 11.2 & & 0.71 & & 4.70 & & 0.33 & & 4.33 & & 0.51 + & 4.30 & & 0.36 & & 9.31 & & 0.53 & & 11.0 & & 0.15 & & 9.54 & & 0.36 & & 2.77 & & 0.27 & & 2.61 & & 0.18 + 2d & 13.7 & 0.74 & 0.75 & & 2.42 & 0.25 & 0.66 & & 4.63 & 0.40 & 0.04 & & 21.7 & 0.74 & 0.51 & & 31.2 & 0.46 & 0.47 & & 5.46 & 0.41 & 0.50 + & 3.62 & 0.88 & 0.69 & & 4.55 & 0.26 & 0.66 & & 70.9 & 0.68 & 0.48 & & 11.2 & 0.96 & 0.59 & & 9.70 & 0.45 & 0.47 & & 93.0 & 0.64 & 0.58 + & 2.08 & 0.89 & 0.77 & & 3.72 & 0.22 & 0.64 & & 5.32 & 0.55 & 0.49 & & 2.86 & 0.64 & 0.47 & & 6.29 & 0.49 & 0.52 & & 6.62 & 0.81 & 0.60 + & & & & & 3.26 & 0.33 & 0.65 & & 36.9 & 0.64 & 0.48 & & 3.73 & 0.87 & 0.60 & & 2.22 & 1.00 & 0.71 & & 4.33 & 0.41 & 0.52 + & & & & & 3.06 & 0.54 & 0.71 & & 22.7 & 0.61 & 0.45 & & & & & & 2.48 & 0.43 & 0.50 & & & & + & & & & & 25.0 & 0.31 & 0.66 & & & & & & & & & & & & & & & & cccccccccccccccccccccccc bp & 2.47 & 0.32 & 0.56 & & 6.39 & 0.28 & 0.88 & & 6.01 & 0.27 & 0.58 & & 4.64 & 0.17 & 0.00 & & 19.4 & 0.15 & 0.67 & & 4.23 & 0.49 & 0.61 + & 18.1 & 0.94 & 0.59 & & 5.12 & 0.12 & 0.82 & & 4.39 & 0.19 & 0.48 & & 4.99 & 0.85 & 0.01 & & 4.47 & 0.21 & 0.56 & & 4.40 & 0.41 & 0.62 + & 6.89 & 0.06 & 0.53 & & 3.15 & 0.29 & 0.71 & & 5.28 & 0.26 & 0.40 & & 2.10 & 0.37 & 0.02 & & 2.67 & 0.09 & 0.37 & & 2.65 & 0.35 & 0.46 + cl & 2.48 & & 0.91 & & 6.39 & & 0.83 & & 6.51 & & 0.31 & & 5.81 & & 0.18 & & 6.26 & & 0.46 & & 4.23 & & 0.56 + & 18.4 & & 0.78 & & 2.48 & & 0.58 & & 3.92 & & 0.33 & & 4.24 & & 0.11 & & 12.0 & & 0.48 & & 4.43 & & 0.58 + & 3.05 & & 0.73 & & 3.51 & & 0.51 & & 5.66 & & 0.36 & & 5.52 & & 0.06 & & 5.71 & & 0.28 & & 2.64 & & 0.39 + 2d & 3.57 & 0.86 & 0.74 & & 3.25 & 0.38 & 0.68 & & 39.4 & 0.43 & 0.42 & & 2.43 & 0.76 & 0.60 & & 4.78 & 0.35 & 0.51 & & 3.83 & 0.56 & 0.56 + & 3.10 & 0.35 & 0.74 & & 2.75 & 0.47 & 0.69 & & & & & & 10.2 & 0.65 & 0.55 & & 19.5 & 0.37 & 0.48 & & 4.40 & 0.39 & 0.51 + & & & & & 3.40 & 0.72 & 0.75 & & & & & & 5.91 & 0.56 & 0.54 & & 9.70 & 0.40 & 0.49 & & 2.69 & 0.53 & 0.51 + & & & & & & & & & & & & & & & & & & & & & & & + @xmath38 & 63.3 & 0.17 & 0.63 & & 2.75 & 0.39 & 0.01 & & 121 & 0.52 & 0.74 & & 2.92 & 0.07 & 0.33 & & 7.60 & 0.17 & 0.27 & & 2.23 & 0.34 & 0.61 + & 2.17 & 0.66 & 0.84 & & 2.58 & 0.60 & 0.70 & & 3.54 & 0.90 & 0.69 & & 8.70 & 0.08 & 0.47 & & 2.23 & 0.50 & 0.70 & & 5.00 & 0.52 & 0.60 + & 16.2 & 0.69 & 0.69 & & 2.89 & 0.28 & 0.64 & & 3.85 & 0.21 & 0.38 & & 9.86 & 0.10 & 0.36 & & 3.33 & 0.69 & 0.50 & & 5.90 & 0.73 & 0.58 as examples of the phase diagrams , figures [ fig9 ] - [ fig11 ] show the phase diagrams for periods found in the 2000 - 2001 h@xmath1 and 1998 - 1999 he i equivalent width data and the radial velocities from the 2005 - 2006 data . for clarity , the phase diagrams are plotted over two cycles with zero phase arbitrarily set to the first observation of the season . in fig . [ fig9 ] , the phase diagram of the period found in the h@xmath1 equivalent width data is reasonably well fit by the sinusoid with an @xmath35 value of 0.76 . in fig . [ fig10 ] , the period found in the he i equivalent width data is fit very well , with an @xmath35 value of 0.91 while the 18.4 d period had an @xmath35 value of @xmath60.78 . we found that an @xmath35 value of @xmath60.7 or more was generally a good indication that sinusoidal modulation was present . however , there were a few cases where modulation was judged by eye to be present even with an @xmath35 value as low as 0.48 . from these figures , it is also clear that the bayesian and clean methods often find the same periods , though not always in the same order . but with he i @xmath25876 radial velocity data from the 2005 - 2006 season . ] phase diagrams were also made for periods detected in the 2d periodograms . here , the wavelength bin that contained the peak was phase - folded at the particular period . in many cases a frequency was detected over several wavelength bins but only the bin with the highest peak was used for the phase plot . the error bars on the data points represent @xmath52 , where @xmath53 is the reciprocal of the s / n in the continuum of each spectrum . the phase plots from the 2004 - 2005 season are shown in fig . [ fig12 ] . in this case , the 19.5 and 9.7 d periods appear to show sinusoidal modulation even though their @xmath35 values are @xmath60.5 . in assessing the many periods found in the period search , we used several criteria . first , periods identified using more than one technique were considered more robust detections . second , periods detected in both h@xmath1 and he i were of interest because of the possible wind - photosphere connection . third , we looked for sinusoidal modulation in the phase diagrams . finally , periods that were on the order of the rotational period of the star ( 13 - 18 d ; see sec . 7.2 below ) were of interest because they could indicate rotational modulation . some key results are listed in table 7 . lcccccclc [ ] 1998@xmath541999 & & 2.47 , 2.48 & & @xmath55 & & & & yes + & & 3.51@xmath543.62 & @xmath55 & & & @xmath55 & 5876 , 6566 & yes + & & 13.1@xmath5413.7 & @xmath55 & & & @xmath55 & 6562 & yes + & & 18.1 , 18.4 & & @xmath55 & & & & yes + 2000@xmath542001 & & 2.75 & & & @xmath55 & @xmath55 & 5877 & yes + & & 3.06 , 3.07 & @xmath55 & & & @xmath55 & 6564 & yes + & & 3.25 , 3.26 & & & & @xmath55 & 5876 , 6561 & yes + & & 6.39 & & @xmath55 & & & & yes + 2001@xmath542002 & & 4.38@xmath544.43 & @xmath55 & @xmath55 & & & & no + 2002@xmath542003 & & 4.24@xmath544.28 & @xmath55 & @xmath55 & & & & yes + & & 11.2 & @xmath55 & & & @xmath55 & 6563 & yes + 2004@xmath542005 & & 2.22 , 2.23 & & & @xmath55 & @xmath55 & 6567 & yes + & & 9.7 & & & & @xmath55 & 5876 , 6563 & yes + & & 19.1@xmath5419.5 & @xmath55 & @xmath55 & & @xmath55 & 5873 & yes + 2005@xmath542006 & & 2.23 , 2.24 & @xmath55 & & @xmath55 & & & yes + & & 2.60@xmath542.69 & @xmath55 & @xmath55 & & @xmath55 & 5878 & no + & & 4.32@xmath544.43 & @xmath55 & @xmath55 & & @xmath55 & 5876 , 6568 & yes many of the periods detected in our search were in the range with no obvious connection to rotational modulation . these periods may be associated with stellar pulsations or some type of evanescent events . in the he i @xmath26 data from 2000 - 2001 , a period of was identified in by both bayesian and clean periodograms , with sinusoidal modulation quite strong in both phase diagrams . a period was found in the he i @xmath26 data of the 1998 - 1999 season and periods of 6.01 and were detected in the he i @xmath26 data of the 2001 - 2002 season . in addition , a period was found in the 6565 bin of the 2005 - 2006 spectra . these periods are all similar to the period reported in spectra from 1998 by prinja et al . ( 2004 ) and may be due to a persistent or recurring phenomenon . many of the identified periods , especially those in the h@xmath1 equivalent width , showed harmonic modulation in their phase diagrams . this modulation is particularly associated with periods found in the h@xmath1 @xmath26 . given the scaling relation found by puls et al . ( 1996 ) between h@xmath1 @xmath26 and the mass - loss rate , it is tempting to conclude that the star s mass - loss rate is undergoing harmonic modulation , possibly due to stellar pulsations . however , caution is warranted in making any such connection . if the wind is translucent , the variable flux from pulsations at the surface may be modulating the ionization fraction of hydrogen , leading to changes in the h@xmath1 equivalent width . in the 2004 - 2005 season , a period of was found in the radial velocity data as well as in the 6567 bin of the 2d periodogram . this period was also detected in the h@xmath1 @xmath26 and radial velocity data in the 2005 - 2006 season . harmonic modulation was present in the phase diagrams for these periods . the fact that this period was found in back - to - back seasons implies that this may be a persistent or recurrent signal . indeed , in the 1998 - 1999 season a period of was detected in the radial velocity data and a period was found in the 6572 bin of the 2d periodogram . inspection of tables 4 - 7 reveals that periods of were detected in every season . sinusoidal modulation was evident in some of the phase diagrams for these periods . these periods were found in the equivalent width data as well as the 2d periodograms but not in the radial velocity data . it is interesting to note that the period may be an overtone of the periods . if these periods are separate detections of the same signal , it indicates a persistent or recurrent trend . from the 2d periodograms , we find that much of the variability in he i is detected near the central wavelength bin suggesting a photospheric origin . meanwhile , this same period may be detected in h@xmath1 but in a bin that implies an origin in the wind . for example , in 1998 - 1999 spectra a period of was found in the 5876 bin and a period of was found in the 6566 bin . in the 2005 - 2006 data , a period of was identified 5876 bin and a period of was found in the 6568 bin . the time scales of the periods can provide a hint as to the origin of the variability . a period due to rotational modulation of wind structure , such as a co - rotating interaction region ( cir ) , would be expected to be on the same time scale as the rotational period of the star . rotational modulation could also have time scales that are some fraction of the rotation period , if there were more than one wind structure present and they were approximately equally spaced . the dependence of the rotational period on the radius of the star is a major source of uncertainty . an accurate angular diameter measurement could help to constrain the stellar radius and the rotational period . the projected rotational velocity is another source of uncertainty since it is an upper limit , and a fraction of it may be due to macroturbuence rather than rotation . adopting a @xmath31 of 91 km s@xmath16 and a stellar radius of 32 r@xmath7 gives a maximum rotational period of @xmath5618 d. assuming a radius of 24 r@xmath7 with the same @xmath31 yields a maximum rotational period of @xmath5613 d. several of the detected periods fall within the range of estimated for the period of rotation . in the he i @xmath26 data from the 1998 - 1999 season periods of 18.1 and were found by the bayesian and clean periodograms , respectively . both phase diagrams exhibited sinusoidal modulation , particularly the period . a period of was detected in the radial velocity data from the same season . it is possible that these two periods have the same origin but with the period rooted in the photosphere while the period arises farther out in the wind . periods ranging from 19.1 to were identified in both the h@xmath1 and he i @xmath26 data for the 2004 - 2005 season . the phase diagrams for these periods all exhibit sinusoidal modulation . a period of was also found in the 5873 bin of the 2d periodogram the same season . the 5873 bin is in the wavelength region where we begin to see wind contamination in the he i line . also in this season , a period of was found in the central wavelength bins of 5876 and 6563 in the 2d periodograms . the period could be an overtone of the periods and may indicate a connection between the photosphere and wind emission . a period was also reported by prinja et al . ( 2004 ) in h@xmath1 , h@xmath17 , and he i @xmath26678 observations from 1998 . the line profile variability of the h@xmath1 and he i @xmath25876 lines in the supergiant @xmath0 ori has been examined through time - series analysis of spectra from seven seasons of observations . this star has a strong and variable stellar wind , as demonstrated by its highly variable h@xmath1 line which is observed to have several profile morphologies including p cygni , inverse - p cygni , double absorption , and pure emission . profiles of the h@xmath1 line show net emission approximately 65% of the time . the he i line is primarily an absorption line with weak wind features in the wings of the line . the he i profile is seen to sway back and forth , possibly indicating pulsational radial velocity oscillations . temporal variance spectra were calculated for each season for both spectral lines . the tvs clearly show that both lines undergo significant variability extending beyond @xmath13@xmath57 region indicating the presence of wind variability . in both cases , the variability is greatest on the blue side of line center . the tvs for the he i line has a double - peaked profile indicative of radial velocity oscillations due to pulsations . the double peaks are contained within the @xmath57 region consistent with a photospheric or near photospheric origin for the radial velocity variations . the region of significant variability defined by the 99% confidence level of the tvs was used to set the wavelength regions over which equivalent width measurements were made for the h@xmath1 and he i spectra . radial velocity measurements were also made from the central absorption of the he i profile . a periodicity search was carried out on data from six of the seven seasons individually . a technique that applies bayesian statistics to the lomb - scargle periodogram was employed on the equivalent width , radial velocity , and binned - spectra time series . the clean algorithm was also used on the equivalent width time series . numerous periods were identified in the variability as a result of these analyses . all periods identified in the bayesian analysis had corresponding ls periodogram peaks that exceeded the 5% false alarm probability . phase diagrams were created for each identified period and a sinusoidal function was fit to the phased data . the resulting phase plots indicate that several of the periods are harmonic . periods on the order of the rotational period of the star ( 13 - 18 d ) were found in data from two seasons indicating possible rotational modulation . sinusoidal modulation was evident in the phase diagrams in both cases . in the 1998 - 1999 observing season , periods were detected in the he i equivalent width data . in the 2004 - 2005 h@xmath1 and he i equivalent width data , periods of 19.1 to were found . a period was also present in the 5876 and 6563 bins in the 2d periodograms from that season . this is a possible overtone of the period . most of the periods identified were in the range of . many of these periods were present in both h@xmath1 and he i data and exhibited sinusoidal patterns in their phase diagrams . stellar pulsations are a possible origin of this variability and provide evidence for a connection between the photospheric activity and the wind . particularly intriguing is the fact that periods of are present in every season and were found in both lines . this may represent a persistent mode of variability in this star . the spectral variability of @xmath0 ori is clearly complex and in need of further study . stellar pulsation modeling of this star is needed in order to determine if the periods reported here can be produced by radial or non - radial pulsations . observationally , a campaign to acquire simultaneous spectroscopic and photometric data could provide conclusive evidence for a connection between stellar pulsation and wind variability . a more complete time series at optical and uv wavelengths , probing the inner and outer wind regions , would provide a more complete picture of the wind variability . we wish to thank jon bjorkman , karen bjorkman , douglas gies , and thomas kvale , for their support and many excellent suggestions . this work was partially supported by an nsf - prest grant ( ast-0440784 ) awarded to nancy morrison and karen bjorkman . support for the acquisition of these spectra at ritter observatory came from : the university of toledo , the fund for astrophysical research , inc . , and the american astronomical society small research grants program . financial support for the publication of this work was received from the scott e. smith fund for research at ritter observatory and from adrian college . finally , thanks go to the following members of the ritter observing team for obtaining spectra of @xmath0 ori : nancy morrison , chris mulliss , david knauth , howard ritter , will fischer , karen bjorkman , anatoly miroshnichenko , john wisniewski , josh thomas , amanda gault , doug long , noel richardson , erica hesselbach , nick sperling , dan kittell , and ian mcginness . berghoefer , t.w . , baade , d. , schmitt , j. h. m. m. , et al . 1996 , , 306 , 899 blomme , r. , prinja , r. k. , runacres , m. c. , & colley , s. 2002 , , 382 , 921 boyajian , t.s . , gies , d.r . , barai , p. , et al . 2007 , , 119 , 742 bretthorst , g.l . , 2001a , in aip conf . 567 , bayesian inference and maximum entropy methods in science and engineering : 19th international workshop , ed . rychert , j. , erickson , g. & smith , c.r . ( melville , ny : aip ) , 1 bretthorst , g.l . 2001b , in aip conf . 568 , bayesian inference and maximum entropy methods in science and engineering : 20th international workshop , ed . ali mohammad - djafari , ( melville , ny : aip ) , 241 brown , t.m . 1990 , in asp conf . proc . 8 , ccds in astronomy , ed . jacoby , g.h . , ( san francisco , ca : asp ) , 335 cantiello , m. , langer , n. , brott , i. , et al . 2009 , , 499 , 279 cassinelli , j. p. , myers , r. v. , hartmann , l. , et al . 1983 , , 268 , 205 castor , j. i. , abbott , d. c. , & klein , r. i. 1975 , , 195 , 157 chalabaev , a. , & maillard , j.p 1983 , , 127 , 279 cherrington , e. 1937 , , 85 , 139 conti , p. s. , leep , e. m. , and lorre , j. j. 1977 , , 214 , 759 cranmer , s.r . , & owocki , s.p . 1996 , , 462 , 469 crowther , p.a . , lennon , d.j . , & walborn , n.r . 2006 , , 446 , 279 ebbets , d. 1982 , , 48 , 399 evans , d.s . 1967 , in iau symposium 30 , determination of radial velocities and their applications , ed . batten , a.h . , & heard , j.f . , ( london : academic press ) , 57 fullerton , a.w . 1990 , ph.d . thesis , university of toronto fullerton , a.w . , gies , d.r . , & bolton , c.t . 1996 , , 103 , 475 gregory , p.c . 2005 , bayesian logical data analysis for the physical sciences , ( cambridge : cambridge university press ) hillier , d.j . 1991 , , 247 , 455 hinkle , k. , wallace , l. , valenti , j. , & harmer , d. 2000 , visible and near infrared atlas of the arcturus spectrum , 37279300 , ( san francisco , ca : asp ) horne , j.d . , & baliunas , s.l 1986 , , 302 , 757 howarth , i.d . , siebert , k.w . , hussain , g.a.j . , & prinja , r.k.1997 , , 284 , 265 jarad , m.m . , hilditch , r.w . , & skillen , i. 1989 , , 238 , 1085 johnson , h.l . , & morgan , w.w . 1953 , , 117 , 313 kaper , l. , henrichs , h. f. , fullerton , a. w. , et al . 1997 , , 327 , 281 kaufer , a. , stahl , o. , prinja , r.k . , & witherick , d. 2006 , , 447 , 325 lanz , t. , & hubeny , i. 2007 , , 169 , 83 lee , t.a . 1968 , 152 , 913l lefever , k. , puls , j. , & aerts , c. 2007 , , 463 , 1093 lomb , n.r . 1976 , , 39 , 447 lucy , l.b . & solomon , p.m. 1970 , , 159 , 879 morel , t. , marchenko , s. v. , pati , a. k. , et al . 2004 , , 351 , 552 morrell , n. , & levato , h. 1991 , , 75 , 965 mullan , d.j . 1984 , , 283 , 303 mulliss , c.l . 1996 , m.s . thesis , university of toledo prinja , r.k . , massa , d. , & fullerton , w. 2002 , , 388 , 587 prinja , r.k . , rivinius , th . , stahl , o. , et al . 2004 , , 418 , 727 puls , j. , kudritzki , r.p . , herrero , a. , et al . 1996 , , 305 , 171 roberts , d.h . , lehr , j. , & dreher , j.w . 1987 , , 93 , 968 ryans , r.s.i . , dufton , p.l . , rolleston , et al . 2002 , , 336 , 577 saio , h. , kuschnig , r. , gautschy , a. , et al . 2006 , , 650 , 1111 sapar , l. , & sapar , a. 1998 , baltic astronomy , 7 , 451 scargle , j.d . 1982 , , 263 , 835 scuderi , s. , panagia , n. , stanghellini , c. , trigilio , c. , & umana , g. 1998 , , 332 , 251 searle , s.c . , prinja , r.k . , mass , d. , & ryans , r. 2008 , , 481 , 777 townsend , r.h.d . 2000a , , 318 , 1 townsend , r.h.d . 2000b , , 319 , 289 townsend , r.h.d . 2007 , in aip conf . 948 , unsolved problems in stellar physics : a conference in honor of douglas gough , ed . stancliffe , r.j . , houdek , g. , martin , r.g . , & tout , c.a . , ( melville , ny : aip ) , 345 van leeuwen , f. ( ed . ) 2007 , hipparcos , the new reduction of the raw data , astrophysics and space science library , ( berlin : springer ) , 350 walborn , n.r . 1976 , , 205 , 419

we report here on the photometric and kinematic properties of a well defined group of nearly 200 low - mass pre - main sequence stars , concentrated within of the early - b star 25 ori , in the orion ob1a sub - association . we refer to this stellar aggregate as the 25 orionis group . the group also harbors the herbig ae / be star v346 ori and a dozen other early type stars with photometry , parallaxes , and some with ir excess emission , consistent with group membership . the number of high and low - mass stars is in agreement with expectations from a standard initial mass function . the velocity distribution for the young stars in 25 ori shows a narrow peak centered at , very close to the velocity of the star 25 ori . this velocity peak is offset from the velocity characterizing the younger stars of the ori ob1b sub - association , and from the velocity of more widely spread young stars of the ori ob1a population near the ori cluster ; this result provides new and compelling evidence that the 25 ori group is a distinct kinematic entity , and that considerable space and velocity structure is present in the ori ob1a sub - association . the low - mass members follow a well defined band in the color - magnitude diagram , consistent with an isochronal age of myr , depending on the assumed evolutionary model . the drop in the overall li i equivalent widths and accretion fraction between the younger ori ob1b and the 25 ori group , is consistent with the later being significantly older , independent of the absolute age calibration . + the highest density of members is located near the star 25 ori , but the actual extent of the cluster can not be well constrained with our present data . in a simple - minded kinematic evolution scenario , the 25 ori group may represent the evolved counterpart of a younger aggregate like the ori cluster . + the 25 ori stellar aggregate is the most populous myr sample yet known within 500 pc , setting it as an excellent laboratory to study the evolution of solar - like stars and protoplanetary disks .

astro-ph0701710

samples of @xmath7 myr old , low - mass stars are important for understanding protoplanetary disk and pre - main sequence stellar evolution . many estimates ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) indicate that by @xmath8 myr the majority of inner disks have dissipated and planet formation has been mostly completed @xcite . however , finding large samples of stars at this age has been a significant observational challenge . the traditional means of identifying low - mass pre - main sequence ( pms ) stars ( or t tauri stars - tts , as they are usually known ) has been to look toward molecular clouds . if cloud complexes had long lifetimes , of @xmath9 myr , ( e.g. * ? ? ? * ) , and star formation had proceeded in a more or less continuous fashion , one would expect to find roughly 10 times as many 10 myr - old stars as 1 myr old objects toward dark clouds with crossing times of order 10 myr . the lack of such older pms stars in regions like the taurus clouds was called by @xcite the post - t tauri problem . recently , accumulating observational and theoretical evidence suggest that molecular clouds lifetimes are much shorter than previously thought ( e.g. * ? ? ? * ; * ? ? ? * ) ; by ages @xmath8 myr the parent molecular clouds have dissipated and no longer serve as markers of the young population ( c.f . this is consistent with the absence of @xmath10 emission in regions like the orion ob1a sub - association , in which the estimated ages of the massive oba stars are of order 10 myr @xcite . one approach to searching for somewhat older pms stars over wide areas , extending beyond the confines of molecular clouds , has been to select x - ray active g and k type stars in large scale x - ray surveys like the rosat all - sky survey ( rass ) . however , in regions like orion the rass only reached down to @xmath11 stars @xcite , missing the more numerous late k and m type stars ; in addition , it suffered from significant contamination by foreground x - ray active zero age main sequence stars @xcite . nevertheless , a few @xmath7 myr old stellar groups have been discovered in the solar neighborhood in recent years . the nearby tw hya association is one of the most important such groups , but the coevality or even membership is somewhat uncertain for some objects @xcite . another important nearby group of similar age is the @xmath12 cha cluster @xcite but , like tw hya , it contains no more than @xmath13 stars . @xcite found a 10 myr old sample ( @xmath14 members ) in ngc 7160 in cepheus , but this is a much more distant cluster ( @xmath15 pc ) . the orion ob1a sub - association @xcite , at an estimated of @xmath16 myr @xcite is a promising region for searching for numerous samples of low - mass stars at this important age range . in @xcite we presented the first sample of the low - mass stellar population in this region , and identified a concentration of pms stars around the b1v star 25 ori . here we investigate the kinematics of members of this stellar aggregate , which we refer to as the 25 orionis group . we show that the stars in this region have a relatively narrow velocity dispersion that is kinematically distinct from the stars in the ori ob1b sub - association . the location of these stars in the color - magnitude diagram indicates a small dispersion in age , and provides indications of systematic discrepancies in pre - main sequence isochrones . the stars in the 25 ori group provide a well defined , populous , sample for studies of disk evolution . in section [ obs ] we describe the photometric selection of candidate pms stars , and the spectroscopic follow - up for membership confirmation , and measurement of radial velocities . in [ results ] we present our results , describing the overall 25 ori stellar population ( [ population ] ) , kinematics ( [ kinematics ] , age ( [ ages ] ) , the behavior of li i as a youth indicator ( [ lithium ] ) , and the fraction of stars showing evidence for disk accretion ( [ accretion ] ) . in order to study the star - forming history and characterize the stellar population of the orion ob1 association we are carrying out a large scale optical variability survey and spectroscopic follow - up . in briceo ( 2005 ) we presented results for a region spanning @xmath17 , between @xmath18 and @xmath19 ( in figure [ fig1 ] we show a part of that larger area ) , which includes the ori ob1b sub - association , and large fraction of ori ob1a . the multi - band , multi - epoch photometric survey , has been done with the @xmath20 quest ccd mosaic camera @xcite installed on the stock 1 m schmidt - type telescope at the venezuela llano del hato national astronomical observatory . we use variability and optical colors as a means to select candidate pms low - mass stars in color - magnitude diagrams , down to a limiting magnitude of v@xmath21 . we refer the reader to @xcite for details of the instrument , photometry and candidate selection criteria . the photometric sample considered here is composed of candidate pms stars located in the same area of the @xcite study , for which spectroscopic confirmation is presented here for the first time , plus stars already confirmed spectroscopically as orion ob1 members in that work . low - resolution spectra provide confirmation of the tts nature of a star . such objects can be reliably identified through optical spectroscopy by the presence of : ( 1 ) h@xmath22 emission , strong in stars actively accreting from their surrounding disks ( the so called classical t tauri stars - ctts , which exhibit w(h@xmath22)@xmath23 at m0 , with larger values at later spectral types , see @xcite ) , and weak ( a few ) in non - accreting stars , the weak - lined t tauri stars - wtts ; ( 2 ) the li i @xmath24 line strongly in absorption ( e.g. * ? ? ? spectra in the range @xmath25 also include several strong molecular features like tio bands , characteristic of cooler stars ( spectral types later than @xmath26 ) , which are sensitive t@xmath27 indicators . we obtained low - resolution optical spectra of our candidate pms stars with a variety of telescopes and instruments . the brighter stars ( v@xmath28 ) were observed at the 1.5 m telescope of the whipple observatory equipped with the fast spectrograph @xcite , with a spectral coverage of 3400 centered at 5500 , and a resolution of 6 . though spectra for a large fraction of the bright candidates were obtained in @xcite , here we include a number of stars for which spectra were not available at the time . with these additions we consider our follow - up spectroscopy in this magnitude range complete . for the fainter ( v@xmath29 ) candidates in our photometric survey the bulk of the spectroscopic follow - up was done with the hectospec multi - fiber spectrograph on the 6.5 m mmt telescope at mount hopkins @xcite . hectospec is a low - resolution optical spectrograph with a total of 300 fibers that can be placed within a @xmath30 diameter field . we used the 270 groove @xmath31 grating , obtaining spectra in the range @xmath32 , with a resolution of 6.2 . with a s / n ratio @xmath33 in our spectra we could detect li @xmath24 absorption down to @xmath34 to identify the low - mass , young stars . we measured h@xmath22 and li i equivalent widths in all our low resolution spectra , using the _ splot _ routine in iraf . from november 2004 through april 2005 we observed 2216 candidates in eight fields distributed throughout the orion 1a and 1b sub - associations ; a complete analysis of this entire sample will be presented in briceo et al . ( 2007 ) . in this contribution we concentrate on newly identified tts in a smaller area surrounding the early type star 25 orionis , in the orion ob1a region , and near the belt star @xmath35 ori , located in the orion ob1b sub - association ; for this sample we present high - resolution spectroscopy ( [ hectochelle ] ) . the data reduction of both fast and hectospec data was performed by s. tokarz through the cfa telescope data center , using iraf tasks and other customized reduction scripts . spectral types were derived using the sptclass code developed at cida , which includes spectral indices in the hectospec wavelength range . in order to study the kinematics of the region , we obtained high resolution spectra of a subset of orion ob1 candidate members , using the hectochelle fiber - fed multi - object echelle spectrograph , mounted on the 6.5 m mmt @xcite at the mmto , arizona , usa . hectochelle uses 240 fibers in a @xmath30 circular field . each fiber has a diameter of @xmath36 m , subtending 1.5 arc sec on the sky . we observed in the spectral order centered near h@xmath22 , from 6465 to 6650 ( 185 wide ) , with a resolution r@xmath37 ( equivalent to @xmath38 fwhm ) . two observations were obtained on the night of december 2 , 2004 , in two fields containing a total of 243 t tauri stars confirmed through fast and hectospec spectra , out of which we were able to obtain high - resolution spectra for 147 objects ( 60% ) . the first field observed ( field 1 ) is located in the 25 orionis group ( figure [ fig1 ] ; @xcite ) ; it contains 124 tts , of which 65% were assigned fibers . the second field ( field 2 ) is located within the ori ob1b sub - association , next to the belt star @xmath35 ori ( figure [ fig1 ] ) ; it includes 119 tts , of which 57% were assigned fibers . for each field we obtained three exposures , that were later combined to aid in cosmic ray removal . in table [ hcobs ] we summarize the characteristics of the observations . we allocated fibers according to the following priorities : first , to members already confirmed from low - resolution spectroscopy , mostly bright ( v @xmath39 ) members observed with fast ; we then added some fainter ( v@xmath40 ) , spectroscopically confirmed members . finally we allocated the remaining fibers to other objects in the field brighter than v=16 , which though not being confirmed members , had optical or 2mass colors suggestive of them possibly being candidate young objects . in each field we observed @xmath13 objects with dubious status from our low - resolution spectroscopy , that could be either wtts or dme stars . to establish the zero point for the radial velocities derived from our spectra , during the same night we obtained template spectra of 14 relatively bright stars with spectral types similar to our target stars , in a field located in selected area 57 , near the north galactic pole . the radial velocities of these stars have been monitored for many years using the cfa digital speedometers @xcite without showing any signs of variation , and their absolute velocities have all been established to an accuracy better than @xmath41 @xcite . for wavelength calibration we obtained several thorium - argon lamp exposures throughout the night ; the thermal shift was monitored to be of a few hundreds of a pixel ( @xmath42 m / s ) . spectra were reduced and extracted using standard iraf routines . no sky subtraction was applied . to derive the dispersion correction we used 47 lines fitted with a cubic spline . the rms fit for the dispersion correction itself was roughly @xmath43 m / s , however , because of fiber - to - fiber variations , which have not been properly calibrated out yet , the true scatter is closer to @xmath44 . typical signal - to - noise ( snr ) ratios ranged from @xmath45 for the fainter stars ( v@xmath46 ) , to snr @xmath43 for the brighter objects ( v@xmath47 ) . radial velocities were derived using the cross - correlation routine _ xcsao _ in the iraf saotdc package . this routine obtains the radial velocity from the shift of the cross - correlation peak . the r value from the cross - correlation process is a measurement of the ratio of the peak of the cross - correlation to the estimated noise , and is therefore an indicator of the accuracy of the results @xcite . in our measurements we blocked out h@xmath22 as well as other emission lines . for each star in our sample we ran _ xcsao _ using all the template spectra , we selected the result with the largest r ( the best matching template ) . in our final list we used only @xmath48 results , which yield velocity errors @xmath49 ( for slowly - rotating stars ; ( see * ? ? ? * ) ) . in this way we avoided spectra contaminated by scattered sunlight , which mostly correspond to small r values ( low snr ) that tend to cluster around the velocity value of the heliocentric correction . in order to eliminate multiple systems in our sample , we selected only objects which showed a clear single peak in the cross - correlation plot . the derived radial velocities were corrected for the motion of the earth around the sun . the final sample includes 78 `` single '' tts with @xmath48 ( 47 in field 1 and 31 in field 2 ) . the photometric and spectroscopic measurements for these stars are presented in table [ ttsdata ] . the first column provides the cida variability survey of orion ( cvso ) running number , continuing the numbering sequence from @xcite ; stars from that article have cvso numbers @xmath50 . the newly identified young members of 25 ori and orion ob1b presented here start with object cvso-198 . we provide two values of the equivalent width of the h@xmath22 emission line ( @xmath51 $ ] ) , the first one estimated in our low - resolution spectra , and the second measured in the hectochelle spectra . both values agree for the majority of stars , within the measurement uncertainties and allowing for variability of the h@xmath22 line . all equivalent widths were obtained using the _ splot _ routine in iraf , by measuring the line respect to the local pseudo - continuum , estimated to be roughly at the base of the emission feature . the 25 orionis group stands out as a concentration of tts stars in the ori ob1a sub - association ( figure [ fig1 ] ) , roughly surrounding the b1vpe star 25 ori ( @xmath52 ) . this star has been classified as a classical be object @xcite ; more recently @xcite determined a spectral type b2 . in @xcite we already pointed out the existence of this stellar aggregate as a distinct feature in the spatial distribution of the brighter ( v@xmath29 ) members , albeit with lower number statistics . @xcite classify the region around 25 ori as a new cluster ( ascc 16 ) , based on parallaxes , proper motions and v , b - v data for _ hipparcos _ and tycho-2 stars . they place the cluster center coordinates at @xmath53 , @xmath54 south - west of 25 ori , and derive a cluster radius of @xmath55 ( figure [ fig1 ] ) . the parallaxes of the _ hipparcos _ stars yield a mean distance of 323 pc ; we assume here a value of 330 pc ( see * ? ? ? we are aware we are missing the northernmost region of the 25 ori cluster , as can be seen in figure [ fig1 ] . however , because our continuing survey ultimately will span orion up to @xmath56 we will be in a position to study whether the cluster extends further north . for the present analysis we defined a circle of @xmath30 radius around 25 ori as the 25 orionis group . though this is larger than the cluster radius derived by @xcite , it allows us to improve our member statistics , and at the same time is not large enough to expect contamination from the younger stars in the orion ob1b region . the @xcite catalog lists 143 stars within this circle and south of our @xmath57 declination limit ( figure [ fig1 ] ) : 12 are b - type stars , including 25 ori , 18 are a - type stars , one of which is the herbig ae / be star v346 ori ( spectral type a8 ; @xcite ) , and 21 are f - type stars . a number of these early type stars show ir emission properties consistent with membership in this young stellar group . @xcite used spitzer data to look for circumstellar disks in a subset of the @xcite stars . they studied 7 of the 12 _ hipparcos _ b - type stars , and found one with 24@xmath58 m excess emission and lack of near - ir excesses indicative of a debris disk . their sample also included 13 a - type stars , out of which 7 had debris disks , and 1 ( v346 ori ) had an optically thick disk , for a disk fraction among a - stars of 62% ; of the 5 f - type stars in their sample 3 ( 60% ) had debris disks . in our subsequent discussion we adopt these sources as early type members of the 25 ori group . if a standard initial mass function ( imf ; e.g. @xcite ) is assumed , one would expect that a cluster formed with 12 b - type stars will produce roughly 12 a - stars , 12 f - stars , 18-g stars , 48-k stars and 336-m stars ( down to @xmath59m6 ) . within this area we identify 197 tts from our low - resolution follow up spectroscopy ( 124 inside our 25 ori hectochelle field - field 1 ) , all of them k and m - type ; 21 objects were reported in @xcite and 176 are new identifications , whose detailed properties will be discussed in @xcite . the magnitude limit of our spectroscopic follow - up effectively places our completeness at a spectral type of roughly m4 ( we discuss this with better number statistics in @xcite ) ; therefore , our sample is missing a number of the later type m - stars . still , the number of low - mass young stars we have identified within this region is in agreement , within a factor of @xmath60 , to expectations from a typical imf . figure [ fig2 ] shows the velocity distribution of stars in table [ ttsdata ] . because in our sample @xmath61 the typical velocity errors are @xmath62 . the two regions show clearly distinct kinematics , with well defined peaks located at @xmath63 for the 25 ori field and @xmath64 for the ori ob1b field . the narrow cores @xmath65 of both distributions are not much larger than the estimated errors , and consistent with velocity dispersions in typical regions of star formation . the peak of the velocity distribution in ori ob1a is consistent with the majority of radial velocity measurements for 25 ori ( @xmath66 , @xcite ; @xmath67 , @xcite ; @xmath68 , @xcite ; @xmath69 , @xcite ) , suggesting that the bulk of the tts are indeed related to the star 25 ori . the presence of a @xmath70 shift between the cores of the velocity distributions in ob1a/25 ori region and ob1b qualitatively supports the findings that these two sub - associations are at very different distances , though nearly along the line of sight @xcite . in a recent study , @xcite found two distinct velocity components when analyzing velocity data toward four fields around @xmath4 ori , located in the sub - association ori ob1b . their kinematic _ group 1 _ has velocities @xmath71 . taking the average value of @xmath72 determined by @xcite for ori ob1a , derived for each ori ob1 sub - association are the error of the mean , @xmath73 . from their ob1a and ob1b measurements , the actual mean internal error is @xmath74 and their mean external error is @xmath75 . ] they relate stars in _ group 1 _ with this sub - association ; this is consistent with suggestions that the closer ori ob1a region in fact extends in front of ori ob1b @xcite . the velocities of _ group 1 _ roughly overlap our distribution for 25 ori . however , it is interesting to highlight that the peak of the distribution for the _ group 2 _ of @xcite is located at @xmath76 , @xmath77 off the @xmath78 peak of our 25 ori @xmath79 distribution . with errors in both studies in the range @xmath80 , this is a statistically robust difference . if the kinematics of the `` field '' population of ori ob1a is indeed characterized by the velocity distribution of _ group 1 _ in the @xcite sample , then our results strongly argue in favor of 25 ori cluster being a a distinct entity in velocity space , different from the general population of ob1a , which is widely spread with a low spatial density ( c.f . figs . 5 and 6 of * ? ? ? the distribution of velocities in 25 ori ( field 1 ) looks somewhat skewed to higher velocities ( @xmath81 ; figure [ fig2 ] ) , though with few objects . this is what would be expected from the presence of a few interloper pms stars from the widespread population of ob1a . if this degree of contamination ( 5/47 ) is taken as representative of the field pms population of ori ob1a in the 25 ori area , we would expect that of the 124 tts within our hectochelle field , @xmath82 could be `` contaminants '' from ob1a . the ob1b stars in our sample lie to the west of the main distribution of molecular gas , corresponding to the respective ends of the orion a and b clouds ( figure [ fig1 ] ) . the gas nearest the stars has a heliocentric radial velocity @xmath83 ( e.g. , figure 5 of * ? ? ? * ) , velocity as the stars in our hectochelle field . there is also a degree of asymmetry in the velocity distribution for ori ob1b ( right panel in figure [ fig2 ] ) , this time extending to lower @xmath79 , toward the same velocity range expected of ob1a stars and of 25 ori . this is not surprising if we recall that ori ob1a is closer than ob1b , and much more spatially extended , such that along any line of sight in the direction of ob1b there probably will be some pms members of the general population of ori ob1a , that will contribute at these velocities . another fact worth pointing out is that the gas is present over a range of several @xmath84 , which could also account for this asymmetry . the orion belt star nearest to our hectochelle field 2 , in ori ob1b , is @xmath35 ori , with estimated velocities that range from @xmath85 @xcite to @xmath86 @xcite , though most measurements cluster around @xmath87 . @xcite classify @xmath35 ori as a member of ob1b . the _ hipparcos _ parallax is @xmath88 @xcite , which corresponds to a distance of 412 pc , consistent with our assumed distance of 440 pc . the tts in our southern hectochelle field share the same velocity as the nearby gas and @xmath35 ori , indicating that these objects indeed belong to what is called the ob1b sub - association . in their study @xcite found for their _ group 2 _ velocities in the range @xmath89 , peaking at @xmath90 . they identify this group as the `` @xmath4 ori cluster '' , with an age of @xmath45 myr and at a distance of 440 pc , similar to ori ob1b ; however , they argue it is kinematically distinct to ob1b , because they assume the value of @xmath91 from @xcite for this sub - association . our findings do not support their claim . first , the @xmath92 peak velocity we find for ob1b , and the small width of the distribution shown in figure [ fig2 ] , strongly suggest that ob1b and the @xmath4 ori cluster share similar kinematics . ori appears to be younger than the general population of ori ob1b , and with a higher disk fraction ( their figure 13 ) . ] second , if we recompute the average group velocity for the ob1b stars in @xcite , but considering only objects with reliable membership , that are not spectroscopic binaries , and flagged as having constant @xmath79 , we find @xmath93 with a @xmath94 dispersion of @xmath95 . therefore , the peak velocity of the @xcite _ group 2 _ is consistent with the radial velocity of both our sample of spectroscopically confirmed low - mass members , and samples of early type stars in ori ob1b . in contrast to the strongly peaked velocity distributions for the young stars , field stars ( objects showing strong h@xmath22 absorption in our hectochelle spectra ) exhibit an almost flat distribution , with stars at all velocity bins ( lower panels in figure [ fig2 ] ) . other objects recognized as part of the field population are dme stars , which look very similar to wtts in that they have weak h@xmath22 emission , but lack the strong li i 6707 absorption feature characteristic of m stars younger than @xmath96 myr . however , the velocity distribution of dme stars in both fields shows a small excess of objects at the velocity peak for each region . a few of these objects could actually be wtts , but the snr of their low resolution spectra does not allow detection of the li i line . but even if this were the case , we expect to be missing no more than 1 - 2 objects in 25 ori and 2 - 4 in orion ob1b . as mentioned in [ lowres ] and above , the presence of the li i 6707 line strongly in absorption is a clear indicator of youth in k and m - type stars . in figure [ fig3 ] , we show the distribution of li i equivalent widths , @xmath97 , measured in our low - resolution spectra , plotted as a function of the effective temperature for each star , derived from the adopted spectral type and the temperature scale in table a5 of @xcite . in order to compare with our sample , we show the lithium equivalent width measurements by @xcite and @xcite of taurus tts , representative of @xmath98 myr old , low - mass stars . the dotted line indicates the lower boundary of w(li i ) for the majority of taurus tts . we also plot the upper envelopes for the pleiades cluster ( age @xmath99 myr ; @xcite ) using data from @xcite and @xcite , and for the ic 2602 cluster ( age @xmath100 myr ; @xcite ) , from the dataset of @xcite . all the 25 ori and ori ob1b members fall above the upper cluster envelopes ; however , an important fraction of the m - stars in each field fall below the lower boundary of the taurus tts . also , in both orion fields the observed @xmath101 is lower than in taurus . the large spread toward lower values of @xmath97 for the m - type stars has also been observed by @xcite in @xmath4 ori . comparison of the @xmath97 in 25 ori and ob1b members shows that while 40% of all the ori ob1b members in table [ ttsdata ] fall within the taurus tts locus only 22% of the tts in the 25 ori group share this region of the diagram ( see figure [ fig3 ] ) . this analysis includes all stars with velocities within @xmath102 of the peak @xmath79 in each field . in figure [ fig3 ] we have also plotted the line separating wtts from post - t tauri stars ( ptts ) . @xcite used the term ptts for pms stars which have lost their t tauri properties ; @xcite identify as ptts those pms stars with essentially wtts properties , except they fall below the the lithium isoabundance line plotted in figure [ fig3 ] . in general , ptts are expected to be slightly more evolved tts . in the ori ob1b field , the wtts / ptts fraction is 0.48 ; in the 25 ori field it is 0.25 . in contrast , the wtts / ptts fraction for taurus members is 5.57 . we argue that the lower @xmath103 , the increasing fraction of tts with @xmath97 values below the taurus lower boundary as a function of age , and the decreasing wtts / ptts fraction , suggest we are looking at the effects of li depletion over the age range spanned by these three regions . in figure [ fig4 ] we show the color - magnitude diagrams ( cmds ) , of _ all _ tts within our two hectochelle fields ( open and solid circles ) , and for the subset of members for which high - resolution spectra were obtained ( hc - only , listed in table [ ttsdata ] ; open and solid diamonds ) . as in @xcite we adopted distances of 330 pc for the 25 ori group and 440 pc for the ori ob1b field . we also show two sets of isochrones for 1 , 3 , 30 , and 100 myr from @xcite and @xcite . the tts in the 25 ori field follow a relatively well defined band in the cmd . the spread observed is largely consistent with the upper limit of 0.75 magnitudes expected from unresolved binaries ( the dashed - lined isochrone in figure [ fig4 ] ) , plus a small spread introduced if we assume the stars follow some distance distribution along the line of sight ( @xmath104 at 400 pc corresponds to @xmath105 pc , or 0.04 magnitudes ) . variability should not affect in an important way our cmds because we use for each object the mean magnitudes derived from our multi - epoch optical survey ( see * ? ? ? * ) . from our photometry and spectral types for the tts members , assuming the intrinsic colors in table a5 of @xcite , and the extinction law of @xcite with @xmath106 , we compute an overall mean @xmath107 mags for the 25 ori cluster , in good agreement with the e(b - v)= 0.09 obtained by @xcite . for the 25 ori group we also plot the v and @xmath108 values for _ hipparcos _ stars ( including 25 ori ; * ? ? ? * ) , v346 ori and the excess ir emission early type stars from @xcite ; the v , b photometry for these later objects is from @xcite , transformed to @xmath108 assuming a mean @xmath107 and the appropriate colors for their spectral types from @xcite . similarly , we show the location of the _ hipparcos _ b - type stars and the belt star @xmath35 ori in the cmd diagram of ori ob1b . all these early type stars fall on or close to the the zams , consistent with them being the higher mass members of these stellar populations . the band of tts in the 25 ori field is roughly encompassed by the 3 - 10 myr isochrones . at the bright end of our tts sample ( v@xmath109 ) the stars seem to cross over to the left , towards the 30 myr isochrone ; this could be due to the photometry for some of these stars being affected by saturation ( the bright limit for our photometry varies with each ccd in the quest camera , but is roughly between v@xmath110 ) . alternatively , this could be due to uncertainties in the location of the birthline at these range of masses @xcite . at the faint end we note a deviation from the theoretical models that we attribute to limitations in the currently available pms isochrones . at v@xmath111 , @xmath112 the bulk of our stars in the cmd seems to deviate slightly downwards from the @xcite isochrones . with the baraffe tracks , which are considered to be more appropriate for low mass stars ( e.g. * * ; * ? ? ? * ) , the behavior is opposite , the lower mass stars seem systematically younger ( or the tracks bend down too steeply ) . a similar discrepancy is found in the ori ob1b field . overall , the baraffe models produce mean ages with are @xmath113 younger than those derived from the the siess isochrones . for the larger sample we derive an average age of @xmath7 myr from the @xcite isochrones , and of @xmath105 myr from the @xcite isochrones , in agreement with our determination in @xcite with smaller number of objects . for the hc - only sample in 25 ori the resulting average ages are similar . for both tts samples within the orion ob1b field , the resulting average ages are @xmath114 myr from the siess et al . isochrones and @xmath115 myr from the baraffe et al . models . in the ob1b field we find an average @xmath116 mags . within the uncertainties , the age derived here for 25 ori is the same as for the general ori ob1a sub - association in @xcite , though for this larger region the spread in ages for the individual stars is larger , probably caused in some measure by the increase in distance depth along the line of sight as one encompasses much wider areas . with a well defined , spatially confined , numerous sample in ori ob1b , resulting in a smaller spread in the cmd , the age difference between this younger region and the older ori ob1a is now firmly established . this age difference of a factor @xmath117 is consistent with the indications of li i depletion we discuss in [ lithium ] , and with the decay in accretion indicators shown by @xcite and considered here in the following section . + our finding of slightly different kinematics between the 25 ori cluster and what can be interpreted as the widely spread pms low - mass population of ori ob1a , suggest that although both stellar populations probably share a common origin , a previously unrecognized degree of substructure is present in this region . the fraction of ctts , or stars thought to be accreting from their circumstellar disks , can be derived by comparing the value of w(h@xmath22 ) for each spectral type in our low resolution spectroscopy , with the criteria in @xcite : a star is considered to be a ctts if @xmath118 in the range k0-k5 , @xmath119 for k7-m2.5 , @xmath120 for m3-m5.5 , and @xmath121 for m6-m7.5 . inside each 0.5 deg radius hectochelle field , and considering _ all _ tts , the ctts fraction in the 25 ori field is 5.6% ( 7 out of 124 stars ) , while in the ori ob1b field the ctts fraction is 12.6% ( 15 out of 119 stars , @xmath6 higher ) . these values are almost a factor of @xmath122 lower than those reported in @xcite . however , this apparent discrepancy can be easily explained by considering the following : first , in @xcite our sample for ori ob1b included regions like the very young ngc 2024 cluster ( @xmath123 myr ) which has a large number of ctts , and is nominally located within ori ob1b if one uses the @xcite boundaries . second , as one increases the census of pms stars in older regions like ori ob1a , the most frequent type of members are wtts , which tends to lower the accretor fraction . if we assume that our present ctts fraction in 25 ori is more representative of this region , then we find it is a factor of @xmath124 larger than what @xcite obtained in the @xmath7 myr old ngc 7160 ( 1 accretor in their sample of 55 members ) , but much smaller than the @xmath125% they report for the @xmath115 myr old cluster tr 37 . the higher ctts fraction in 25 ori when compared to ngc 7160 could be because this region is indeed younger , or due to the small number statistics in the @xcite sample . in order to provide a first look at the spatial distribution of young stars in this aggregate , we computed a spatial density map by counting the number of tts in squares @xmath126 in side ; because the spectroscopic followup is most spatially uniform for the brighter stars , we only considered tts with v@xmath127 ( the non - overlapping hectospec circular fields produce circular `` imprints '' on the spatial distribution for the fainter members , as can be seen in figure [ fig1 ] ) . the result is shown in figure [ fig5 ] . the largest starred symbol marks the location of 25 ori , and the belt stars @xmath128 and @xmath35 ori ; the small solid starred symbols correspond to b - type stars included as probable members of the 25 ori cluster in the list of @xcite , and _ hipparcos _ b - stars classified as members of ori ob1b by @xcite . there is a peak in the spatial distribution of low - mass young stars in the 25 ori region , containing 8 tts ( an implied density of 128 @xmath129 ) , at @xmath130 , @xmath131 south - east of 25 ori . @xcite place their cluster center at @xmath132 , @xmath133 , which given our bin size agrees with our determination . the density of stars falls off to @xmath123/bin at a radius of @xmath134 , which corresponds to 7 pc at the assumed distance of 330 pc . this value is slightly larger than what @xcite found for the younger @xmath4 ori cluster ( @xmath135 pc ) . at a velocity dispersion of @xmath62 , an unbound stellar aggregate would expand roughly 1 pc every 1 myr . if the 25 ori group resembled the @xmath4 ori cluster at an age of @xmath115 myr , a naive dynamical picture would have it evolve to a cluster radius of @xmath136 pc at @xmath137 myr . however , we must caution that the detailed spatial structure can not yet be fully investigated from our present data . first , because of gaps between the ccd detectors in the quest ccd mosaic camera , artifacts appear such as the e - w gap apparent at @xmath138 ( figure [ fig1 ] ) . second , we have not yet completed follow - up spectroscopy of the photometric candidates located north of @xmath57 ( figures [ fig1 ] and [ fig2 ] ) . therefore , if more young stars turn up northward of that line , the actual membership and spatial extent of the 25 ori group may be larger than shown here , and the point with the highest density of members could well shift somewhat . to illustrate this point we have plotted as small magenta stars the tts with @xmath79 within 1@xmath4 of the peak velocity for the 25 ori cluster ( 34 objects ) . it can be seen that these stars concentrate in the northern half of the hectochelle field , with 74% located in this area . the highest density is in the ne quadrant of the field , mostly within the highest contour of the spatial density map . we conclude that the spatial distribution of the kinematical members within the hectochelle field suggests the cluster may extend further north ( the apparent dearth of members close at the northernmost part of the field could well be an incompleteness effect ) . @xcite used sloan digital sky survey data ( sdss ) to look for tts in a @xmath139 wide equatorial strip between @xmath140 and @xmath141 . in his fig.8 he shows an excess of candidate pms stars at @xmath142 and @xmath143 , which he identifies as the south tip of the 25 ori cluster . because of this he derives a larger cluster radius of 8 - 11 pc . though his sample does not overlap the 25 ori nor the ori ob1b fields discussed here , we can compare with the larger scale distribution of stars in our orion ob1 survey . the feature identified by mcgehee as part of 25 ori can be seen as an elongated density enhancement in our spatial density map , roughly @xmath144 south of 25 ori ; additionally , we find a number of faint tts at this location ( figure [ fig1 ] ) . even taking into account this group of young objects , there is an apparent paucity of stars beyond @xmath134 south from 25 ori ; radial velocities need to be obtained before these objects can be interpreted as related to the 25 ori cluster . at present we are obtaining additional hectochelle fields north and south of the 25 ori field discussed here , and also in nearby regions of ori ob1a . these observations will not only help us constrain the spatial extent and refine the kinematics of this unique sample of young , low - mass stars , but will also provide for the first time important information on the actual velocity distribution of the sparse pms population of the orion ob1a sub - association . we have used the hectochelle spectrograph on the 6.5 m mmt to obtain high resolution spectra of 149 low - mass pms stars distributed in two fields located in the orion ob1 association , one field placed on the 25 ori group originally identified by @xcite , the other near the belt star @xmath35 ori , in the ori ob1b sub - association . we determined radial velocities with errors @xmath145 for 78 stars ( 47 in the 25 ori field and 31 in the ori ob1b field ) . the pms members of the 25 ori group have radial velocities in a narrow range , peaking at @xmath1 ; this value is very close to that of 25 ori itself , suggesting this may be indeed the most massive star in this stellar aggregate . the stars belonging to the ori ob1b sub - association peak at @xmath146 , and also show a very narrow distribution . these two differing values clearly show that these are two separate populations , and give strength to findings of a significant distance difference between these two groups . however , the 25 ori group not only is distinct in velocity space from the ori ob1b sub - association . the cluster peak @xmath79 is @xmath77 off from the peak velocity found by @xcite for stars they identify as belonging to the orion ob1a sub - association . therefore the 25 ori group constitutes a distinct kinematic group , clearly differentiated in velocity space from both the younger and more distant ob1b , and from what seems to be the general , widely spread population of ori ob1a . the color - magnitude diagrams yield ages of 7 - 10 myr for the 25 ori group , and of 4 - 5 myr for ori ob1 , depending on the assumed set of isochrones . the observed spread in each sample is roughly consistent with the upper limit expected for a population of unresolved binaries , combined with the distance depth estimated for each region . our finding of a similar age but slightly different kinematics between the 25 ori cluster and what can be interpreted as the widely spread pms low - mass population of ori ob1a , suggest that both stellar populations probably share a common origin , but also that a previously unrecognized degree of spatial and kinematic substructure is present in this region . the overall behavior of the li i equivalent widths from one region to the other , and the decline of a factor @xmath122 in the fractions of ctts are consistent with the 25 ori group being significantly older than ori ob1b , independent of the absolute age calibration . a first approach to the spatial structure of the 25 orionis group , with the caveat of an incomplete census , suggests that the low - mass pms stars peak at @xmath130 , @xmath131 south - east of the star 25 ori , with a maximum surface density of @xmath147 . however , the pms stars with @xmath79 values within @xmath94 of the cluster peak velocity concentrate slightly north of the density maximum . we argue that this may indicate that the cluster does indeed extend further north . + we derive a cluster radius of @xmath105 pc . in a simple - minded kinematic evolution scenario , this value is consistent with the 25 ori group representing the evolved counterpart of a @xmath115 myr old aggregate like the @xmath4 ori cluster . c. briceo acknowledges support from grant s1 - 2001001144 of fonacit , venezuela . this work has been supported by nsf grant ast-9987367 and nasa grant nag5 - 10545 . we are grateful to susan tokarz at cfa , who is in charge of the reduction and processing of fast , hectospec and hectochelle spectra . we thank the invaluable assistance of the observers and night assistants , in particular freddy moreno , orlando contreras , ubaldo sanchez and gregory rojas , at the venezuela schmidt telescope that made possible obtaining the data over these past years . we also acknowledge the support from the cida technical staff , and in particular of gerardo snchez . briceo , c. , preibisch , t. , sherry , w. h. , mamajek , e. e. , mathieu , r. d. , walter , f. m. , zinnecker , h. 2006 , in protostars & planets v. , b. reipurth , d. jewitt , and k. keil ( eds ) , university of arizona press , tucson . fabricant , d.g . , hertz , e.n . , szentgyorgyi , a.h . , fata , r.g . , roll , j.b . , & zajac , j.m . 1998 , construction of the hectospec : 300 optical fiber - fed spectrograph for the converted mmt , proc . spie , 3355 , 285 szentgyorgyi , a.h . , cheimets , p. , eng , r. , fabricant , d.g . , geary , j.c . , hartmann , l. , pieri , m.r . , roll , j.b . 1998 , proc . spie , vol . 3355 , p. 242 - 252 , optical astronomical instrumentation , sandro dodorico , ed . rcccclrrrrrrcc 19 & 05:23:30.93 & + 01:45:49.3 & 15.87 & 2.44 & m3 & -5.4 & 0.4 & -4.70 & 17.6 & 1.5 & 5.9 & w & 25ori + 198 & 05:23:46.57 & + 01:45:15.8 & 13.97 & 1.27 & k4 & -0.1 & 0.5 & -0.08 & 19.1 & 0.5 & 17.8 & w & 25ori + 199 & 05:23:55.36 & + 01:34:18.7 & 16.92 & 2.40 & m3 & -4.4 & 0.3 & -5.21 & 20.1 & 1.4 & 5.8 & w & 25ori + 200 & 05:24:02.30 & + 01:53:21.5 & 16.54 & 2.36 & & -5.1 & & -5.12 & 19.1 & 1.2 & 5.7 & w & 25ori + 24 & 05:24:04.26 & + 01:48:30.5 & 16.34 & 2.22 & m2 & -6.1 & 0.3 & -5.67 & 19.3 & 0.9 & 9.1 & w & 25ori + 201 & 05:24:07.56 & + 01:38:41.7 & 16.28 & 2.17 & m3 & -3.2 & 0.2 & -6.07 & 19.9 & 1.7 & 3.4 & w & 25ori + 25 & 05:24:10.36 & + 01:44:07.6 & 15.42 & 1.88 & m0 & -2.6 & 0.3 & -1.42 & 19.6 & 0.7 & 13.4 & w & 25ori + 202 & 05:24:18.14 & + 01:39:22.9 & 17.41 & 2.61 & m3 & -4.7 & 0.5 & -4.44 & 18.8 & 1.8 & 3.1 & w & 25ori + 203 & 05:24:20.27 & + 01:40:54.8 & 16.66 & 2.37 & m3 & -5.6 & 0.4 & -3.37 & 19.0 & 1.4 & 5.9 & w & 25ori + 204 & 05:24:25.45 & + 01:48:14.3 & 17.45 & 2.72 & m4 & -6.6 & 0.6 & -6.37 & 18.0 & 1.7 & 3.2 & w & 25ori + 205 & 05:24:26.64 & + 01:18:13.9 & 14.77 & 1.99 & m2 & -2.1 & 0.3 & -2.08 & 25.1 & 0.7 & 12.0 & w & 25ori + 206 & 05:24:41.04 & + 01:54:38.6 & 14.93 & 1.54 & k5 & -53.5 & 0.5 & -43.42 & 18.8 & 0.9 & 11.4 & c & 25ori + 207 & 05:24:52.27 & + 01:38:43.9 & 13.63 & 1.16 & k4 & -0.8 & 0.5 & -0.61 & 24.2 & 0.8 & 15.0 & w & 25ori + 208 & 05:24:54.18 & + 01:34:02.1 & 18.08 & 3.22 & m5 & -10.0 & 0.6 & -3.51 & 19.9 & 1.8 & 3.1 & w & 25ori + 209 & 05:25:01.92 & + 01:34:56.4 & 16.87 & 2.56 & m3 & -5.8 & 0.5 & -4.74 & 18.5 & 1.2 & 5.6 & w & 25ori + 29 & 05:25:02.07 & + 01:37:21.2 & 16.19 & 2.49 & m3 & -7.6 & 0.5 & -7.13 & 22.1 & 1.1 & 7.1 & w & 25ori + 210 & 05:25:03.61 & + 01:44:12.1 & 16.92 & 2.46 & m3 & -5.9 & 0.3 & -4.15 & 19.6 & 1.6 & 5.2 & w & 25ori + 211 & 05:25:08.76 & + 01:45:54.3 & 13.98 & 1.24 & k5 & -0.1 & 0.5 & -0.13 & 19.4 & 0.5 & 17.6 & w & 25ori + 212 & 05:25:13.55 & + 01:16:20.6 & 17.55 & 2.72 & m3 & -6.9 & 0.5 & -3.64 & 21.4 & 2.1 & 3.5 & w & 25ori + 213 & 05:25:16.72 & + 01:16:15.9 & 16.08 & 2.53 & m3 & -4.4 & 0.2 & -4.07 & -0.1 & 2.1 & 5.2 & w & 25ori + 214 & 05:25:17.04 & + 01:48:22.4 & 13.42 & 1.04 & k2 & -0.2 & 0.5 & -0.06 & 19.1 & 0.5 & 17.6 & w & 25ori + 215 & 05:25:27.70 & + 01:41:46.9 & 17.54 & 2.63 & m3.5 & -8.0 & 0.2 & -4.28 & 17.4 & 2.0 & 4.0 & w & 25ori + 216 & 05:25:34.25 & + 01:34:05.7 & 18.45 & 2.95 & m4.5 & -7.1 & 0.3 & -7.02 & 17.5 & 1.6 & 4.1 & w & 25ori + 217 & 05:25:34.39 & + 01:52:19.8 & 15.29 & 1.92 & m1 & -2.1 & 0.3 & -1.70 & 19.2 & 0.7 & 13.2 & w & 25ori + 218 & 05:25:35.18 & + 01:43:40.6 & 17.29 & 2.62 & m4 & -7.7 & 0.3 & -4.31 & 19.0 & 1.7 & 4.2 & w & 25ori + 219 & 05:25:35.65 & + 01:28:23.2 & 15.05 & 1.82 & m0 & -1.3 & 0.5 & -0.13 & 21.0 & 0.6 & 13.4 & w & 25ori + 220 & 05:25:38.89 & + 02:01:29.9 & 14.01 & 1.19 & k4 & -0.7 & 0.4 & -0.40 & 19.0 & 0.7 & 15.8 & w & 25ori + 221 & 05:25:39.57 & + 01:42:53.7 & 15.10 & 1.94 & m0 & -1.8 & 0.5 & -1.53 & 20.6 & 0.7 & 12.4 & w & 25ori + 222 & 05:25:39.72 & + 01:25:46.9 & 16.03 & 2.49 & m3 & -9.9 & 0.3 & -7.47 & 15.5 & 1.9 & 6.1 & w & 25ori + 223 & 05:25:39.94 & + 01:40:21.1 & 16.50 & 2.41 & m3 & -5.3 & 0.4 & -4.35 & 21.2 & 1.1 & 6.7 & w & 25ori + 35 & 05:25:45.91 & + 01:45:50.2 & 14.72 & 1.91 & k7 & -10.3 & 0.4 & -15.72 & 18.8 & 1.0 & 11.0 & c & 25ori + 224 & 05:25:46.74 & + 01:43:30.4 & 17.21 & 2.60 & m3 & -20.3 & 0.3 & -13.67 & 18.4 & 1.7 & 3.1 & w & 25ori + 36 & 05:25:50.37 & + 01:49:37.3 & 16.00 & 2.22 & m3 & -8.9 & 0.5 & -3.92 & 17.9 & 0.8 & 10.7 & w & 25ori + 225 & 05:25:51.82 & + 01:47:36.1 & 16.17 & 2.58 & m4 & -4.9 & 0.2 & -3.45 & 18.5 & 1.3 & 5.8 & w & 25ori + 226 & 05:25:52.61 & + 01:34:44.2 & 17.75 & 2.73 & m4 & -4.9 & 0.3 & -3.98 & 20.4 & 1.3 & 4.5 & w & 25 ori + 227 & 05:25:59.05 & + 01:23:34.7 & 16.30 & 2.60 & m4 & -4.8 & 0.2 & -4.99 & 20.2 & 1.2 & 6.0 & w & 25ori + 228 & 05:26:06.36 & + 01:45:46.7 & 16.30 & 2.26 & m2 & -3.4 & 0.2 & -3.84 & 20.6 & 0.9 & 8.8 & w & 25ori + 38 & 05:26:06.41 & + 01:37:11.8 & 16.18 & 2.26 & m2 & -7.5 & 0.5 & -4.04 & 20.8 & 1.1 & 9.1 & w & 25ori + 229 & 05:26:07.84 & + 01:23:08.3 & 15.95 & 2.43 & m3 & -4.3 & 0.2 & -3.71 & 24.1 & 1.6 & 5.5 & w & 25ori + 230 & 05:26:08.02 & + 01:41:15.2 & 15.44 & 1.94 & m2 & -4.9 & 0.5 & -2.90 & 19.9 & 1.1 & 11.6 & w & 25ori + 231 & 05:26:10.10 & + 01:18:44.3 & 16.42 & 2.69 & m4 & -9.0 & 0.2 & -7.28 & 21.4 & 1.7 & 3.9 & w & 25ori + 232 & 05:26:11.18 & + 01:54:52.3 & 14.70 & 1.94 & m2 & -3.7 & 0.3 & -3.22 & 19.1 & 1.2 & 10.5 & w & 25ori + 233 & 05:26:21.98 & + 01:47:57.6 & 16.30 & 2.46 & m3 & -5.0 & 0.3 & -4.46 & 18.3 & 1.2 & 6.8 & w & 25ori + 234 & 05:26:38.04 & + 01:21:45.5 & 15.35 & 2.08 & m2 & -4.0 & 0.4 & -3.77 & 24.0 & 0.8 & 10.4 & w & 25ori + 235 & 05:26:42.65 & + 01:39:47.3 & 16.65 & 2.33 & m2.5 & -5.7 & 0.3 & -7.69 & 20.3 & 1.3 & 5.5 & w & 25ori + 42 & 05:26:50.43 & + 01:55:03.6 & 15.99 & 2.16 & m2 & -3.8 & 0.3 & -2.15 & 20.7 & 0.9 & 8.7 & w & 25ori + 236 & 05:27:00.52 & + 01:49:27.2 & 16.51 & 2.36 & m3.5 & -7.7 & 0.3 & -5.45 & 29.6 & 1.1 & 8.2 & w & 25ori + 89 & 05:31:19.99 & -01:46:53.3 & 15.47 & 1.57 & k7 & -1.6 & 0.4 & -1.18 & 30.5 & 0.7 & 13.3 & w & ob1b + 92 & 05:31:31.75 & -01:47:05.6 & 16.90 & 2.21 & m2 & -3.6 & 0.2 & -5.53 & 23.6 & 1.9 & 3.7 & w & ob1b + 97 & 05:31:42.80 & -01:39:48.3 & 16.79 & 2.41 & m2 & -5.4 & 0.2 & -6.46 & 30.4 & 1.8 & 4.8 & w & ob1b + 237 & 05:31:48.29 & -01:47:59.2 & 16.36 & 2.36 & m3 & -4.3 & 0.4 & -3.78 & 30.8 & 0.8 & 10.5 & w & ob1b + 99 & 05:31:50.79 & -01:55:17.5 & 17.37 & 2.44 & m2 & -5.5 & 0.2 & -5.06 & 30.9 & 1.1 & 8.2 & w & ob1b + 238 & 05:32:00.42 & -01:40:11.0 & 15.85 & 1.88 & & -47.5 & & -47.46 & 32.4 & 0.8 & 13.1 & c & ob1b + 239 & 05:32:03.47 & -01:56:32.0 & 14.91 & 1.56 & k5 & -1.2 & 0.3 & -1.31 & 4.6 & 0.6 & 16.3 & w & ob1b + 240 & 05:32:05.41 & -01:46:31.8 & 17.42 & 2.47 & & -5.7 & 0.2 & -4.16 & 29.9 & 1.7 & 4.0 & w & ob1b + 106 & 05:32:21.51 & -01:43:45.6 & 16.80 & 2.44 & m2 & -7.0 & 0.3 & -8.39 & 31.0 & 2.1 & 3.4 & w & ob1b + 241 & 05:32:25.69 & -01:33:11.8 & 14.92 & 1.67 & k6 & -3.8 & 0.6 & -3.23 & 29.9 & 0.8 & 13.0 & c & ob1b + 242 & 05:32:28.93 & -01:27:34.6 & 17.85 & 2.06 & m1 & -0.2 & 0.2 & -2.03 & 55.6 & 2.1 & 3.2 & w & ob1b + 243 & 05:32:29.90 & -01:52:40.0 & 15.37 & 1.69 & k6 & -0.9 & 0.5 & -0.61 & 31.5 & 0.6 & 16.7 & w & ob1b + 110 & 05:32:34.35 & -01:50:46.3 & 15.41 & 1.60 & m0 & -2.0 & 0.6 & -1.72 & 31.7 & 0.7 & 12.5 & w & ob1b + 244 & 05:32:59.21 & -02:00:38.1 & 16.00 & 2.47 & m1 & -4.8 & 0.5 & -4.45 & 20.5 & 0.8 & 9.8 & w & ob1b + 245 & 05:33:06.62 & -01:47:03.2 & 17.60 & 2.79 & m3 & -6.6 & 0.3 & -5.01 & 33.0 & 1.7 & 5.1 & w & ob1b + 246 & 05:33:10.69 & -01:31:35.8 & 15.06 & 2.17 & & -2.5 & & -2.54 & 25.5 & 0.9 & 11.6 & w & ob1b + 247 & 05:33:19.87 & -01:56:28.9 & 17.57 & 2.37 & m2 & -2.8 & 0.2 & -2.63 & 32.3 & 1.1 & 7.7 & w & ob1b + 248 & 05:33:22.72 & -01:43:42.2 & 16.04 & 1.91 & k7 & -3.3 & 0.6 & -2.12 & 27.6 & 0.8 & 11.8 & w & ob1b + 118 & 05:33:25.76 & -01:32:18.7 & 14.70 & 1.26 & k5 & -1.0 & 0.5 & -0.72 & 31.6 & 0.6 & 15.9 & w & ob1b + 249 & 05:33:29.30 & -01:30:28.2 & 17.00 & 2.69 & m3 & -5.5 & 0.3 & -6.05 & 32.9 & 1.9 & 4.2 & w & ob1b + 250 & 05:33:36.11 & -01:28:52.6 & 15.00 & 1.69 & k6 & -1.3 & 0.5 & -1.30 & 29.8 & 0.7 & 15.8 & w & ob1b + 251 & 05:33:41.87 & -01:47:40.7 & 16.14 & 1.94 & m0 & -3.3 & 0.5 & -8.35 & 29.7 & 0.8 & 11.5 & w & ob1b + 252 & 05:33:43.63 & -01:13:36.2 & 17.29 & 2.48 & m3 & -3.2 & 0.3 & -3.11 & 27.9 & 1.8 & 4.0 & w & ob1b + 253 & 05:33:46.84 & -01:56:55.1 & 15.01 & 1.49 & k3 & -0.5 & 0.4 & -0.63 & 31.5 & 0.6 & 17.2 & w & ob1b + 122 & 05:33:47.29 & -01:20:27.3 & 15.85 & 1.85 & m1 & -1.9 & 0.5 & -1.66 & 29.8 & 0.8 & 11.6 & w & ob1b + 254 & 05:34:00.70 & -01:50:08.4 & 15.70 & 1.78 & k6 & -2.4 & 0.5 & -2.06 & 31.2 & 0.8 & 12.4 & w & ob1b + 255 & 05:34:06.17 & -01:58:55.9 & 16.75 & 2.39 & m3 & -4.4 & 0.2 & -2.98 & 26.7 & 2.2 & 5.1 & w & ob1b + 256 & 05:34:15.31 & -01:15:19.6 & 16.57 & 2.46 & m3 & -3.8 & 0.4 & -3.82 & 30.2 & 1.7 & 5.6 & w & ob1b + 257 & 05:34:15.46 & -01:34:29.8 & 17.77 & 2.80 & m4 & -7.3 & 0.5 & -7.35 & 27.6 & 2.1 & 3.3 & w & ob1b + 258 & 05:34:17.87 & -01:41:08.0 & 14.57 & 1.50 & k4 & -1.6 & 0.5 & -0.56 & 27.2 & 0.7 & 14.1 & w & ob1b + 134 & 05:34:38.98 & -01:46:56.4 & 15.25 & 1.71 & m0 & -4.6 & 0.5 & -2.84 & 29.0 & 0.6 & 14.9 & w & ob1b + ) wtts , and small stars to bright ctts . plus symbols indicate the faint ( v@xmath154 ) wtts , and plus symbols enclosed in open squares the faint ctts . the belt stars , 25 ori and @xmath4 ori are indicated with large starred symbols . the solid circles indicate the location of the hectochelle fields ( field 1 and field 2 ) . the smaller dashed - lined circle around 25 ori corresponds to the kharchenko et al . ( 2005 ) cluster radius . the grey scale image shows the dust extinction map from schlegel et al . ( 1998 ) ; the countours represent @xmath155 values of 0.1 to 1.5 magnitudes . the thick dashed line outlines the boundaries between ori ob1a and ob1b from @xcite . the horizontal dashed lines mark the limits of the survey region where most of our spectroscopic follow - up has been carried out so far.[fig1 ] ] measurements . these values have later been rounded off to one decimal place for their presentations in table [ ttsdata ] , more representative of our actual error . data for taurus pms stars ( age @xmath98 myr ) are shown as @xmath156 s @xcite . the dotted line traces the lower boundary to the majority of taurus tts . the long dash - dot line corresponds to the lithium isoabundance line from @xcite , which separates wtts from ptts . we show as a short - dash line the upper envelope for the pleiades cluster ( age @xmath99 myr ; @xcite . data from @xcite and @xcite ) , and in the long - dash line the upper envelope of the ic 2602 cluster ( age @xmath100 myr ; @xcite . data from @xcite ) . our typical error bar is also indicated in each panel.[fig3 ] ] ) t tauri stars in the orion ob1a and ob1b sub - associations . the contours correspond to stellar densities of 2 , 4 , 6 , 8 , and 10 stars per 0.25 deg square bin . the largest starred symbol marks the location of 25 ori ( @xmath52 ) . the smaller stars around 25 ori indicate the b - type stars from the list of kharchenko et al . ( 2005 ) in their ascc 16 cluster . the orion belt stars @xmath128 and @xmath35 , in the ori ob1b region , are also indicated with large starred symbols . the short dashed line around the belt stars indicates the ori ob1b boundaries of @xcite . the two light circles show the 1 deg wide hectochelle fields . the small open stars in the 25 ori hectochelle field indicate tts with velocities within 1@xmath4 of the @xmath79 peak for the cluster ( @xmath157).[fig5 ] ]

we present results of a high - resolution , near - infrared survey of 41 nearby , young ( ) m0m5.0 dwarfs using the altair natural guide star adaptive optics system at the gemini north telescope . twelve of the objects appear to be binaries , 7 of which are reported here for the first time . one triple system was discovered . statistical properties are studied and compared with earlier ( f to k ) and later ( m6 very low - mass , vlm ) populations . we find that the separation distribution of the binaries in this sample peaks at , which is consistent with previous measurements of early - m binaries . hence , early - m binaries seem to occur in on average tighter systems than g binaries . at the same time they are significantly wider than field vlm binary stars . the distribution of mass ratios of primary and secondary stars was found to show an intermediate distribution between the strongly peaked distribution of field vlm systems and the almost flat distribution of earlier - type stars . consequently , we show evidence for relatively young , early - m binaries representing a transition between the well known earlier star distributions and the recently examined field vlm population characteristics . despite the fact that this survey was dedicated to the search for faint brown dwarf and planetary mass companions , all planetary mass candidates were background objects . we exclude the existence of physical companions with masses greater than 10 jupiter masses ( ) at separations of and masses greater than for separations around 37 of the 41 observed objects .

astro-ph0609356

binary and multiple stars have long provided a highly effective method of testing stellar evolution theory . as coeval systems , the intrinsic properties deduced for individual components ( luminosity , effective temperature , color ) should be consistent with models drawn from the same isochrone . alternatively , parameters that are observationally well determined for one component ( such as distance or metallicity ) can be associated with the other component(s ) . this is particularly useful in characterizing the intrinsic properties of faint companions , including both low - mass stars ( e.g. gould & chanam 2004 ) and brown dwarfs ( e.g. kirkpatrick et al . 2001)of f , g , and early - k main sequence stars or subdwarfs . there is a long astronomical tradition of inverting the latter technique , and searching the environs of nearby stars for low luminosity companions . effectively , the main - sequence stars act as cosmic lamp posts , each illuminating a small corner of the solar neighborhood . the efficiency of this search process is well illustrated by its success in identifying the archetypical ultracool m dwarf , vb10 @xcite , the first l dwarf , gd165b @xcite , and the first t dwarf , gl229b @xcite . subsequent studies have built on this foundation , using a variety of methods , including coronagraphy ( e.g. oppenheimer et al . 2001 ) , wide - field ground - based imaging ( e.g. simons et al . 1996 ) , and high - resolution imaging with both the _ hubble space telescope _ ( _ hst _ ; e.g. golimowski et al . 2004 ) and ground - based adaptive optics ( ao ) systems ( e.g. close et al . 2003 ; siegler et al . 2005 ; burgasser et al . 2006 and references within ) . this paper describes our use of ao techniques on the gemini telescope to search for low - mass brown dwarf and planetary mass companions around young , early - m dwarfs in the vicinity of the sun . section [ sampsel ] describes the sample selection and section [ obs ] describes our observations . sections [ red ] and [ ana ] outline reduction and derived system parameters and section [ dis ] discusses their implications . the primary goal of our observing program is to identify very low - mass brown dwarfs as companions to nearby stars . that goal constrains our choice of targets . the probability of detecting a companion depends on the contrast with respect to the underlying flux distribution of the primary . the contrast , in turn , depends on the relative luminosity of primary and secondary , the angular separation , and the point - spread function ( psf ) . lacking a long - lived central energy source , brown dwarfs cool and fade over relatively rapid timescales ; moreover , low - mass brown dwarfs cool faster than high - mass brown dwarfs ( @xmath9 ; burrows et al . 2001 ) . thus , all other factors being equal , targeting young stars maximizes the probability of detecting low - mass brown dwarf companions . the nearest young star forming regions , however , lie at distances of more than @xmath10 parsecs , while even the 10 to 20 myr - old members of field associations , such as tw hya and tucana , are at distances of 50 to 100 parsecs from the sun @xcite . thus , the advantages offered by youth need to be balanced against the concomitant disadvantages of loss of linear resolution and flux . optimizing the contrast between primary and secondary presents a similar dilemma . in principle , very low - luminosity primaries offer the best prospects of detecting even lower luminosity secondaries ; hence , ( young ) ultracool dwarfs ( spectral types later than m7 ) might appear to be the best targets for our program . however , recent surveys have shown that the binary frequency decreases towards later spectral types , primarily because the maximum separation is a strong function of the total system mass . few ultracool binaries are found in the field with separations exceeding 10 au ( reid et al . 2001 ; close et al . 2003 ; burgasser et al . 2003 , 2006 ; bouy et al . 2003 ; gizis et al . 2003 ; martn et al . 2003 ; siegler et al . 2005 ) , and it is helpful for any survey for very low - mass companions to focus on systems within the immediate solar neighborhood . the space density of ultracool dwarfs is relatively low , with only @xmath11 l dwarfs within 20 parsecs of the sun @xcite . most of those systems are likely to be older than @xmath12myr @xcite . given these competing factors , we have chosen to target early - m dwarfs in the present survey . while these systems are more luminous than l dwarfs ( @xmath13 as compared with @xmath14 ) , they are much more numerous ( @xmath15 systems within 20 parsecs ) and support binary systems with separations as wide as 400au . moreover , the relative level of coronal activity provides a mechanism for selecting the youngest systems for observation in our program . m dwarfs have active chromospheres and coronae , characterized by optical emission ( ca ii h&k and balmer lines ) and x - ray emission , respectively . figure [ fig1 ] shows x - ray activity ( plotting @xmath16 , where @xmath17 is the flux measured by rosat ) as a function of spectral type and @xmath18 color for stars from the tw hydrae association ( @xmath19 myr ; reid 2003 ) , the pleiades ( @xmath20 myr ; micela et al . 1999 ) , the hyades ( @xmath21 myr ; reid et al . 1995b ) and local field stars @xcite . there is a clear decrease in activity with age , with most of the field stars lying well below the cluster members . ] using the data plotted in figure [ fig1 ] as a guide , we have defined selection criteria in the [ @xmath22 , @xmath18 ] plane that are designed to identify m dwarfs with activity levels comparable to or exceeding those of pleiades members : @xmath24 given the dispersion in activity levels at a given age , most single stars in this sample are likely to have ages younger than @xmath25 myr . we have applied these criteria to two data sets : @xmath26 proper motion stars from the 2mass - based nstars survey described by @xcite ; and @xmath27 previously - uncataloged late - type field dwarfs that show significant motion between the possi and 2mass surveys ( the moving m sample , reid , cruz & allen , in prep . ) . most of the former stars are drawn from the nltt proper motion catalogue @xcite , but the sample also includes lower motion dwarfs from the third catalogue of nearby stars ( cns3 ; gliese & jahreiss 1991 ) . the effective proper motion limits of both samples correspond to @xmath28 arcsec / yr , or a tangential velocity of 19 km / sec at 20pc . it is important to note that this threshold does not bias against young stars . systems like the pleiades with a relative velocity of @xmath29 with respect to the sun @xcite or tw hydrae with a relative motion of @xmath30 with respect to the sun @xcite offer exemplary proof that young stellar objects do not need to have velocities within a few km / sec of the local standard of rest . distance estimates are based primarily on photometric or spectroscopic parallaxes . both samples are drawn from the area covered by the 2mass second incremental release ( @xmath31 of the sky ) , and both were cross - referenced against the _ rosat _ bright and faint source catalogues ( voges et al . 1999 ; 2000 ) , searching for counterparts within 15 arcseconds . 196 stars from the first sample and 81 from the second are matched with rosat sources . figure [ fig2 ] compares the coronal activity of the stars in these two samples against our selection criteria . a total of 92 stars exceed the pleiades - like ( 100 myr ) activity threshold . however , 36 of these stars are known to be spectroscopic binaries , and it is likely that the high activity reflects rapid rotation in tidally - locked systems . we have excluded those stars from consideration in the current program , limiting observations to systems that are currently identified as single stars . nonetheless , it is probable that some of the stars targeted in this program will eventually prove to be unrecognized spectroscopic binaries , rather than pleiades - age single stars . as a function of ( @xmath18 ) color for the @xmath32 rosat - detected sources from the nstars @xmath33 pc nltt sample @xcite . the overall distribution is consistent with the larger @xcite nearby star sample . we have superposed x - ray activity data for members of the 100-myr old pleiades cluster , and the dotted lines mark the selection criteria that we have used to segregate candidate young stars in the nstars nltt sample . the lower panel plots similar data for the newly - identified nearby star candidates in the `` moving m '' sample ( reid , cruz , & allen , in prep . ) . [ fig2],title="fig : " ] as a function of ( @xmath18 ) color for the @xmath32 rosat - detected sources from the nstars @xmath33 pc nltt sample @xcite . the overall distribution is consistent with the larger @xcite nearby star sample . we have superposed x - ray activity data for members of the 100-myr old pleiades cluster , and the dotted lines mark the selection criteria that we have used to segregate candidate young stars in the nstars nltt sample . the lower panel plots similar data for the newly - identified nearby star candidates in the `` moving m '' sample ( reid , cruz , & allen , in prep . ) . [ fig2],title="fig : " ] the current paper presents observations of a subset of the x - ray active stars from the nltt / cns3 and moving m samples . all targeted objects are estimated to lie within 20 parsecs of the sun . all observations were taken in queue mode with the gemini north telescope using the facility near - infrared imager ( niri ; hodapp et al . 2003 ) in conjunction with the altair ao system . the detector is a 1024@xmath341024 insb aladdin infrared hybrid array with 27@xmath35 pixels . the f/32 camera was used providing a @xmath36/pixel platescale and 22@xmath37 field of view . the observations were conducted with the cassegrain rotator purposefully disabled , leading to a fixed image pupil relative to camera and detector . for all observed targets , both short ( @xmath38 ) and deep ( 30s ) exposures in the h and k@xmath39 filter bands were obtained . each of the targets were imaged in a 5-position , @xmath40-offset dithering pattern ; 3 images were taken per dither position . the short frames were designed to primarily obtain unsaturated target images so as to reveal any near - equal mass companions at separations @xmath41 . they also served to provide astrometry and psf sources . in a few cases even the minimum integration time of @xmath42 allowed by the niri ir camera led to saturated short frames due to the brightness of the target star . in these cases , other unsaturated sources were used for psf fitting . all of the discovered physical binary systems were detected in the short frames . the deep frames were designed to saturate the target star to gain sensitivity to potential very faint companions ( @xmath43h @xmath44mag ) at radial distances @xmath45 . integrated counts were typically @xmath46@xmath47 in excess of full well depth at both h and k@xmath39 in the primary s core . through the use of coadds , 30s integration times were obtained per image ; 3 images per dither position gave 7.5min of total integration time per target . forty - one of the 56 objects from our survey sample were observed with the gemini north telescope during semesters 2005a , 2005b , and 2006a . table [ tab1 ] lists the 23 objects observed with no likely physical companions according to our sensitivity limits ( examination of the detection limits is given in [ sen ] ) . four of the observed objects showed faint companions of currently undetermined nature . follow - up observations are necessary to reveal whether the detections are consistent with proper motion objects and therefore physical companions . one additional object is believed to be a tight equal - mass binary system with an estimated separation comparable to the fwhm of @xmath48 of these observations . poor image quality probably due to bad weather , however , makes follow - up observations necessary to confirm its nature . none of the other binary images suffered from similar aberrations . table [ tab2 ] shows the observable properties of the 12 binary systems observed , 7 of which are newly discovered binaries , as well as the parameters of the newly discovered triple system . lllcc j00115302 + 2259047 & lp348 - 40 & 8.00 & m3.5 & 1 + j013513930712517 & & 8.08 & m4.0 & 2 + j015311332105433 & lp828 - 89 & 7.14 & m1.0 & 1 + j02085359 + 4926565 & g173 - 39 & 7.58 & m3.5 & 3 + j034723330158195 & hip17695 & 6.93 & m3.0 & 2 + j03494325 + 2419046 & iitau & 8.94 & m4.0 & 1 + j042442600647313 & & 8.63 & m4.5 & 2 + j043737460229282 & steph497 & 6.41 & m0.5 & 1 + j08081359 + 2106094 & lhs 5134 & 6.52 & m2.5 & 1 + j10193634 + 1952122 & gl388 & 4.59 & m3.0 & 3 + j11032125 + 1337571 & lp491 - 51 & 7.91 & m3.5 & 1 + j11115176 + 3332111 & g119 - 62 & 7.50 & m3.5 & 3 + j141304921201262 & gl540.2 & 8.16 & m4.5 & 3 + j14200478 + 3903014 & steph1145 & 7.71 & m2.5 & 1 + j15215291 + 2058394 & wo9520 & 5.76 & m1.5 & 1 + j15530484 + 4457446 & & 9.25 & m3.5 & 4 + j18021660 + 6415445 & lp71 - 82 & 7.65 & m4.5 & 5 + j18130657 + 2601519 & lp390 - 16 & 8.07 & m4.0 & 1 + j19311257 + 3607300 & g125 - 15 & 8.83 & m4.5 & 1 + j204643601148132 & lp756 - 3 & 8.44 & m4.0 & 4 + j221438352141535 & bd-225866 & 6.72 & m0.5 & 1 + j22464980 + 4420030 & gl873 & 5.30 & m3.5 & 3 + j22515345 + 3145153 & gl875.1 & 6.87 & m3.0 & 3 + llcr@@xmath49lr@@xmath49lr@@xmath49lr@@xmath49lcc j00285391 + 5022330 & g172 - 1 & m4.0 & @xmath50 & @xmath51 & @xmath52 & @xmath53 & @xmath54 & @xmath55 & @xmath56 & @xmath57 & 05aug05 & tp + j012427673355086 & g274 - 24 & m5.0 & @xmath58 & @xmath51 & @xmath59 & @xmath53 & @xmath60 & @xmath55 & @xmath61 & @xmath57 & 24aug05 & tp + & & & & & @xmath62 & @xmath57 & @xmath63 & @xmath64 & jul99 & 5 + j04381255 + 2813001 & g39 - 29 & m4.0 & @xmath65 & @xmath51 & @xmath66 & @xmath51 & @xmath67 & @xmath55 & @xmath68 & @xmath69 & 15oct05 & tp + & & & & & & & 13sep02 & 6 + j052541660909123 & lp717 - 36 & m3.5 & @xmath70 & @xmath53 & @xmath71 & @xmath51 & @xmath72 & @xmath55 & @xmath73 & @xmath74 & 14oct05 & tp + j072851373014490 & gj2060 & m0.5 & @xmath75 & @xmath76 & @xmath75 & @xmath71 & @xmath77 & @xmath78 & @xmath79 & @xmath80 & 30oct02 & tp + & & & & & & & ? & 7 + j10364483 + 1521394 & & m4.0 & @xmath81 & @xmath53 & @xmath82 & @xmath83 & @xmath84 & @xmath55 & @xmath85 & @xmath57 & 18may06 & tp + & & & @xmath59 & @xmath53 & @xmath51 & @xmath53 & @xmath86 & @xmath55 & @xmath87 & @xmath88 & & + j132719663110394 & lhs2739 & m3.5 & @xmath88 & @xmath89 & @xmath90 & @xmath59 & @xmath91 & @xmath92 & @xmath93 & @xmath69 & 12jun05 & tp + j13345147 + 3746195 & & m3.5 & @xmath94 & @xmath51 & @xmath95 & @xmath51 & @xmath96 & @xmath97 & @xmath98 & @xmath99 & 26mar06 & tp + j15553178 + 3512028 & g180 - 11 & m4.5 & @xmath100 & @xmath53 & @xmath101 & @xmath83 & @xmath102 & @xmath55 & @xmath103 & @xmath57 & 27may05 & tp + & & & & & & & mar98 & 8 + j17035283 + 3211456 & lp331 - 57 & m2.5 & @xmath104 & @xmath105 & @xmath104 & @xmath89 & @xmath106 & @xmath55 & @xmath107 & @xmath74 & 22apr05 & tp + j223322640936537 & steph2018 & m3.0 & @xmath69 & @xmath89 & @xmath51 & @xmath51 & @xmath102 & @xmath97 & @xmath108 & @xmath88 & 09jun05 & tp + & & & & & & & jul97 & 8 + j232057660147373 & lp642 - 48 & m4.0 & @xmath109 & @xmath53 & @xmath99 & @xmath51 & @xmath110 & @xmath55 & @xmath111 & @xmath69 & 12jun05 & tp + j235813661724338 & lp764 - 40 & m2.0 & @xmath53 & @xmath53 & @xmath53 & @xmath53 & @xmath112 & @xmath55 & @xmath113 & @xmath114 & 16jul05 & tp + the data reduction used an ao data reduction pipeline written in the iraf language as described in @xcite . the pipeline outputs final unsaturated exposures in @xmath115 and @xmath116 band ( @xmath117 ) , and deep @xmath118 ( @xmath119 ) exposures for each observed object . the dithering produces a central @xmath120 high signal - to - noise - ratio ( s / n ) region in a @xmath121 final image centered on the object . since the cassegrain rotator was disabled , the residual psf aberrations including telescope primary and secondary structures remained static and fixed in the final images , allowing for comparative identification of faint companions without confusion from rotating psf artifacts . the pipeline compensates for this by rotating each image by the parallactic angle before combining as described in @xcite . in order to detect faint companions within @xmath122 distance to the central star , we filtered out the low spatial frequency components of the deep images leaving behind only the high frequency residuals in the psf ( unsharp masking ) . figures [ fig3 ] and [ fig4 ] show images of the discovered binaries ; figure [ fig5 ] shows the triple system . . [ fig5 ] ] photometry was performed using the allstar psf - fitting task in iraf . psf stars were selected from similar magnitude targets observed either the same night ( if available ) or as close as possible to the observation date , at similar airmass . uncertainties in @xmath123 and separations listed in table [ tab2 ] were determined by comparing the results employing two distinct psf stars for each binary . in the case of gj2060 , however , saturated raw images ruled out the use of allstar . instead , we used an artificial ghost that appeared in every bright image in the h filter at fixed separation and orientation relative to the detector . this ghost , probably due to secondary reflection in the h filter , remained unsaturated in the long exposures and responded linearly to the flux of the generating object . by averaging the ratio of ghost flux difference to binary flux difference of two similar binaries ( lp 717 - 36 and g 39 - 29 ) we were able to estimate @xmath124 for gj2060 . since there is no ghost in the @xmath116 images , @xmath125 was assumed which is reasonable given that on average @xmath126 for the other binaries is consistent with 0 . to account for uncertainties in this case , a larger error was assumed . the binary data are listed in table [ tab2 ] . we do not have spatially resolved spectra of any of the new multiple systems . moreover , intrinsic scatter in the @xmath127 colors of m dwarfs do not allow us to derive accurate spectral types from our observed @xmath128 and @xmath129 colors . however , we can estimate component spectral types and absolute magnitudes using averaged absolute @xmath130 magnitudes for each spectral type [ see figure [ fig6 ] for the function @xmath131 . component spectral types were estimated by assuming a flux - weighted relation between component and integrated spectral types of @xmath132 , where @xmath133 and @xmath134 are flux values derived from @xmath116 for primary and secondary component respectively . we have evidence that a linear relation can be used to estimate spt@xmath135 and spt@xmath136 from spt@xmath137 since the tio and vo band strengths from which the spectral types are derived are quite linear for m0m7 @xcite . the above relation defines possible pairs of spectral types @xmath138 for each binary . another restriction on spectral types is given by @xmath139 in combination with our measured values for @xmath140 : @xmath141 defines another relation between spt@xmath135 and spt@xmath136 . intersecting these two relations we find an unambiguous combination of @xmath142 and @xmath143 for each binary . typically @xmath43k is small for example g172 - 1 has @xmath144 and spt@xmath145and so we find spt@xmath146 and spt@xmath147 . however , in the case of g180 - 11 @xmath148 ; therefore , we estimate spt@xmath149 and spt@xmath150 from spt@xmath151 . equivalent considerations were made to derive parameter values for the a , b , and c components of the triple system 2m1036 , exploiting the fact that the b and c component are of equal magnitude . for each spectral type . the line shows the spline interpolation of the averages . [ fig6 ] ] absolute @xmath152 magnitudes were derived from applying @xmath139 on interpolated spectral type values . we account for the conversion from k to k@xmath39 band for reasons of completeness , although the difference turns out to be negligible for most of the binaries . using a sample of stars with published values for k and k@xmath39 magnitudes , differences between magnitudes in these two filter bands have been found to scatter around 0 , with an amplitude of @xmath153 . the uncertainties of the individual m@xmath154 values account for this incertitude in conversion by adding in an extra uncertainty of 0.035 . from apparent and absolute magnitudes distances were calculated . derived component spectral types , absolute magnitudes , and distance estimates are listed in table [ tab3 ] . lr@@xmath49lr@@xmath49lr@@xmath49lr@@xmath49lr@@xmath49lr@@xmath155lr@@xmath49lr@@xmath155l g172 - 1a & @xmath156 & @xmath51 & @xmath157 & @xmath53 & m@xmath158 & @xmath159 & @xmath160 & @xmath161 & @xmath162 & @xmath163 & @xmath164 & @xmath165 & @xmath166 & @xmath167 & @xmath168 & @xmath169 + g172 - 1b & @xmath170 & @xmath51 & @xmath171 & @xmath53 & m@xmath172 & @xmath159 & @xmath173 & @xmath174 & & @xmath175 & @xmath176 & & + g274 - 24a & @xmath177 & @xmath59 & @xmath178 & @xmath51 & m@xmath179 & @xmath159 & @xmath180 & @xmath181 & @xmath182 & @xmath183 & @xmath88 & @xmath184 & @xmath185 & @xmath186 & @xmath187 & @xmath188 + g274 - 24c & @xmath189 & @xmath59 & @xmath171 & @xmath51 & m@xmath179 & @xmath159 & @xmath157 & @xmath181 & & @xmath190 & @xmath184 & & + g39 - 29a & @xmath191 & @xmath53 & @xmath192 & @xmath53 & m@xmath158 & @xmath159 & @xmath193 & @xmath161 & @xmath194 & @xmath195 & @xmath196 & @xmath197 & @xmath198 & @xmath199 & @xmath200 & @xmath201 + g39 - 29b & @xmath202 & @xmath51 & @xmath203 & @xmath53 & m@xmath172 & @xmath159 & @xmath204 & @xmath174 & & @xmath58 & @xmath176 & & + lp717 - 36a & @xmath205 & @xmath51 & @xmath191 & @xmath51 & m@xmath206 & @xmath159 & @xmath207 & @xmath50 & @xmath208 & @xmath209 & @xmath210 & @xmath211 & @xmath212 & @xmath213 & @xmath214 & @xmath215 + lp717 - 36b & @xmath216 & @xmath51 & @xmath217 & @xmath51 & m@xmath158 & @xmath159 & @xmath218 & @xmath219 & & @xmath220 & @xmath221 & & + gj2060a & @xmath222 & @xmath190 & @xmath223 & @xmath90 & m@xmath224 & @xmath159 & @xmath225 & @xmath99 & @xmath226 & @xmath183 & @xmath227 & @xmath228 & @xmath229 & @xmath230 & @xmath231 & @xmath232 + gj2060b & @xmath233 & @xmath234 & @xmath235 & @xmath236 & m@xmath237 & @xmath238 & @xmath239 & @xmath240 & & @xmath241 & @xmath242 & & + 2m1036a & @xmath243 & @xmath51 & @xmath244 & @xmath51 & m@xmath206 & @xmath159 & @xmath245 & @xmath246 & @xmath247 & @xmath248 & @xmath249 & @xmath221 & @xmath250 & @xmath251 & @xmath252 & @xmath253 + 2m1036b & @xmath254 & @xmath51 & @xmath255 & @xmath51 & m@xmath172 & @xmath159 & @xmath256 & @xmath50 & & @xmath257 & @xmath258 & @xmath259 & @xmath260 & @xmath261 & @xmath262 + 2m1036c & @xmath263 & @xmath51 & @xmath255 & @xmath51 & m@xmath172 & @xmath159 & @xmath256 & @xmath50 & & @xmath257 & @xmath258 & & + lhs2739a & @xmath264 & @xmath105 & @xmath265 & @xmath51 & m@xmath206 & @xmath159 & @xmath266 & @xmath267 & @xmath268 & @xmath269 & @xmath270 & @xmath271 & @xmath272 & @xmath273 & @xmath274 & @xmath275 + lhs2739b & @xmath276 & @xmath105 & @xmath277 & @xmath89 & m@xmath206 & @xmath159 & @xmath233 & @xmath246 & & @xmath76 & @xmath278 & & + 2m1334a & @xmath279 & @xmath51 & @xmath189 & @xmath53 & m@xmath206 & @xmath159 & @xmath280 & @xmath161 & @xmath281 & @xmath282 & @xmath283 & @xmath284 & @xmath285 & @xmath286 & @xmath287 & @xmath288 + 2m1334b & @xmath289 & @xmath51 & @xmath290 & @xmath53 & m@xmath206 & @xmath159 & @xmath291 & @xmath174 & & @xmath76 & @xmath292 & & + g180 - 11a & @xmath293 & @xmath51 & @xmath192 & @xmath53 & m@xmath158 & @xmath159 & @xmath294 & @xmath246 & @xmath182 & @xmath295 & @xmath58 & @xmath296 & @xmath297 & @xmath298 & @xmath299 & @xmath300 + g180 - 11b & @xmath301 & @xmath89 & @xmath302 & @xmath51 & m@xmath303 & @xmath304 & @xmath305 & @xmath306 & & @xmath307 & @xmath308 & & + lp331 - 57a & @xmath309 & @xmath53 & @xmath310 & @xmath53 & m@xmath311 & @xmath159 & @xmath312 & @xmath88 & @xmath313 & @xmath314 & @xmath70 & @xmath315 & @xmath316 & @xmath317 & @xmath318 & @xmath319 + lp331 - 57b & @xmath320 & @xmath59 & @xmath321 & @xmath89 & m@xmath158 & @xmath159 & @xmath322 & @xmath219 & & @xmath323 & @xmath324 & & + steph2018a & @xmath325 & @xmath51 & @xmath293 & @xmath51 & m@xmath326 & @xmath159 & @xmath327 & @xmath270 & @xmath328 & @xmath329 & @xmath71 & @xmath330 & @xmath331 & @xmath332 & @xmath333 & @xmath334 + steph2018b & @xmath335 & @xmath51 & @xmath336 & @xmath51 & m@xmath326 & @xmath159 & @xmath337 & @xmath270 & & @xmath71 & @xmath330 & & + lp642 - 48a & @xmath338 & @xmath51 & @xmath339 & @xmath53 & m@xmath158 & @xmath159 & @xmath340 & @xmath219 & @xmath341 & @xmath342 & @xmath95 & @xmath343 & @xmath167 & @xmath344 & @xmath345 & @xmath346 + lp642 - 48b & @xmath347 & @xmath51 & @xmath338 & @xmath51 & m@xmath158 & @xmath159 & @xmath348 & @xmath219 & & @xmath349 & @xmath350 & & + lp746 - 40a & @xmath351 & @xmath51 & @xmath352 & @xmath53 & m@xmath311 & @xmath159 & @xmath353 & @xmath257 & @xmath354 & @xmath304 & @xmath355 & @xmath228 & @xmath356 & @xmath357 & @xmath358 & @xmath359 + lp746 - 40b & @xmath180 & @xmath51 & @xmath192 & @xmath51 & m@xmath311 & @xmath159 & @xmath360 & @xmath257 & & @xmath355 & @xmath228 & & + @xmath139 is derived from main - sequence stars ( 5gyr ) . young ( @xmath361myr ) , low - mass objects like those in this sample have not yet reached the main sequence and are therefore slightly brighter than estimated . according to the models of @xcite , this difference is negligible ( @xmath362 in k@xmath39 ) for objects brighter than @xmath363 . the companion to g180 - 11 , however , is estimated to have a magnitude of @xmath364 . its mass estimate ( see [ mass ] ) therefore suffers from an additional uncertainty due to this effect . we estimate masses for each of the binary components from their m@xmath154 magnitudes using 300myr evolutionary models from @xcite . @xcite compared 5gyr baraffe models to observational data finding very good agreement in k band . although the objects in this paper are believed to be much younger and there exists only little observational data constraining a mass luminosity relation for these low mass objects , we expect a sufficiently high accuracy for the 300myr models . the inaccuracies of the models ( see close et al . 2006 ) are assumed to be small compared to our uncertainties in spectral type . from this the binary periods are estimated by applying kepler s third law , @xmath365 , using separations scaled by a factor of 1.26 to account for random inclinations and eccentricities , as derived in @xcite . the calculation of the period of the triple system is based on the assumption that components b and c orbit each other in a tight system and they in turn orbit the primary component a in the same plane . this is a more or less likely assumption , given that gravitationally bound systems consisting of three objects of comparable mass are normally stable in a hierarchical configuration like that assumed and a few pathological situations . thus , we list two periods , one for the bc system and one for a(bc ) . however , we can not exclude a non - relaxed system in a more complicated configuration or a different hierarchy [ e.g. the proximity of b and c being due to projection and therefore c revolving around ( ab ) in a wide orbit ] . table [ tab3 ] lists estimated masses and periods for the binaries and the triple system . an instrument sensitivity estimation using modeled faint companions was made . unsaturated images of lp756 - 3 were scaled down and inserted into the fully reduced deep @xmath116 image of the same star , which was considered to be a representative medium brightness star from our sample ( 2mass @xmath366 , @xmath367 ; scholz et al . these `` fake planets '' were adjusted in brightness to simulate a 5@xmath368 detection in the unsharp masked image ( an example is shown in figure [ fig7 ] ) . to determine the detection limit with sufficient accuracy , 5 fake planets were distributed randomly on a circle drawn around the primary . for each of these positions , 4 to 6 noise measurements from a @xmath369 pixel region close to the companions positions were taken and averaged . the final s / n was calculated by averaging the individual ratios of peak and noise count . figure [ fig8 ] shows the sensitivity of our observation to the detection of faint close - in companions . plotted are the maximum magnitude difference in @xmath116 band vs. distance to the central star in order to detect faint companions at the 5@xmath368 level . @xmath116 exposure of lp756 - 3 , unsharp masked ( low spatial frequencies removed ) . a fake planet at the @xmath370 detection limit was inserted by scaling down the brightness of an unsaturated image of lp756 - 3 and adding it to the deep exposure . the detectable magnitude difference in @xmath116-band is as high as @xmath371 at a separation of @xmath372 . this corresponds to a @xmath373 planet 23au from the central star from the models of @xcite at 500myr . the companion can be distinguished from speckle noise by `` blinking '' the image with an image of another star in the same wave band . because the cassegrain rotator was purposefully disabled , speckles would remain fixed in both images , while the companion reveals itself by appearing in only one of the images . [ fig7 ] ] @xmath140 detection vs. distance from central star . the @xmath374 deep @xmath116 images of lp756 - 3 were used to insert unsaturated scaled - down images of the same star ( example shown in figure [ fig7 ] ) and the s / n was measured at several locations in the image . [ fig8 ] ] our instrumental detection limits predict that objects with @xmath375 can be detected at separations @xmath376 to the central star and objects with @xmath377 at separations of @xmath378 . with lp756 - 3 as a reference , the absolute magnitudes of these hypothetical companions are m@xmath154 @xmath379 and @xmath380 , respectively . all of our single stars are believed to have distances less than 20pc to the sun . accordingly , conservatively assuming an age of 500myr , the models of @xcite translate the detection limit to : @xmath381 at separations @xmath382 and @xmath373 at separations @xmath383 ( see figure [ fig9 ] ) . hence , we can exclude the existence of objects above these limits around the objects in table [ tab1 ] . we are using these evolutionary models as a convenient guide , although , there are likely uncertainties associated with these models at the lowest temperatures and hence our sensitivities ( in terms of the lowest masses ) should be viewed as estimates only . furthermore , the mass - luminosity relation of young planetary mass objects depends strongly on the exact age . the assumed age of 500myr serves as a conservative limit estimate , since our objects are probably less than 300myr old . the models of @xcite , calculated for ages 100 and 500myr ( fig . [ fig9 ] ) , show a maximum difference of @xmath384 9mag in the range of the lowest masses important for these observations . hence , objects comparable to @xmath385myr or even younger might be substantially brighter than the 500 myr models predict and therefore the mass limit would be lower than our estimated values above . level derived from reference star lp756 - 3 at separations of @xmath386 , @xmath122 , and @xmath387 . in our search for planetary companions we come to the same result as @xcite of no faint companion candidates around g172 - 1 . however , we were easily able to resolve this @xmath388 binary for the first time into two almost equal - magnitude components . this binary was first observed by @xcite in 1999 as part of a triple system with a very wide b component with @xmath389 separation to the primary . this corresponds to a separation of @xmath390 when using our distance estimate of @xmath391 . we were not able to image the wide component due to our limited field of view of @xmath121 . g39 - 29 was discovered to be a binary by @xcite . interestingly , there seems to be significant observable dynamics to the system since their measurement of @xmath392 in september 2002 differs strongly from our @xmath393 in october 2005 , which corresponds to a tangential motion in the plane of the sky of @xmath394 relative to the primary . at the same time the position angle stayed approximately the same ( @xmath395 in 2002 , @xmath396 in 2005 ) . we therefore conclude that the system is probably highly inclined . further observations will be needed to set constraints on orbital elements . we are the first to resolve this 0.53 arcsecond binary . a recent distance estimate by @xcite using low resolution spectroscopy resulted in a value of @xmath397 for the unresolved object . our value of @xmath398 agrees with that value on the @xmath399-level , taking the newly discovered binarity into account by multiplying the @xcite - result by a factor of @xmath317 . the binarity of this highly interesting variable object with the alternative identifier v372pup was discovered by @xcite . unfortunately , separation and position angle are not published , which would have allowed further analysis of this close @xmath400 projected separation ( this paper ) binary . a trigonometric parallax was measured by the _ hipparcos _ satellite @xcite . its value of @xmath401 corresponding to a distance of @xmath402 is in excellent agreement with our value of @xmath403 . this object ( subsequently 2m1036 ) is found to be a triple system . it shows a @xmath404 , k@xmath405 primary component accompanied by a pair of equal magnitude ( @xmath406 , k@xmath407 ) objects at @xmath408 , which themselves are separated by @xmath409 ( see figure [ fig5 ] ) . with a distance of @xmath410 , this high proper motion system was calculated to have the second - largest distance of all objects in our sample . its binary nature is reported here for the first time . this binary ( hereafter 2m1334 ) was discovered with a separation of @xmath411 , just at the resolution limit of this survey of @xmath412 . after its binarity was first published by @xcite , the primary was discussed by @xcite as hosting either an 11,000k white dwarf or an m6 companion . from our observed @xmath413 and @xmath414 for the companion we can exclude the possibility of a white dwarf companion , since we would expect @xmath127 to be less than zero for a predicted 11,000k white dwarf according to models of @xcite . we therefore conclude the companion to be a late type main sequence star . our spectral type estimations yield a value of m@xmath415 for the secondary , which is therefore the latest type companion in this sample . the inferred mass of the secondary is possibly below the hydrogen burning limit ( @xmath416 ) , which makes this object a brown dwarf candidate . this is also the lowest mass companion found in this survey . @xcite found a photometric distance of @xmath417 for the unresolved binary . this is in good agreement with our value of @xmath418 when multiplying with a factor of @xmath419 to account for the binarity discovered by our survey . the binary nature of this @xmath420 separated slow - moving system was first found by @xcite . we were able to clearly resolve the components of this newly discovered tight @xmath421 binary . our observations are the first to identify this system as a binary , with separation 1.9 arcseconds . the optical spectrum @xcite shows strong balmer emission , consistent with the young age inferred from the strong x - ray flux . our observations reveal a number of faint sources near the targeted stars that we have rejected as potential companions . here , we outline the rational behind those decisions for a selection of objects . a very faint @xmath422 and @xmath423 point - like object was found at a position angle of @xmath424 at a distance of @xmath425 to the primary ( see figure [ fig10 ] ) . follow - up observations with the gemini north telescope were conducted 2 months after first detection to test for proper motion correlation of the primary and companion candidate . from the two epoch data a relative motion of the central object and companion was derived and compared with the known proper motion of @xmath426 of the primary . we find that the data are consistent with a distant background object . and @xmath427 companion to g173 - 39 was detected with a separation of @xmath428 to the central star . assuming that this object resides in the same distance as g173 - 39 ( 14.3au ; gliese & jahrei 1991 ) , this corresponds to a @xmath429 companion in 53.3au separation to the central object at an age of 500myr . follow - up observations revealed its background nature through proper motion arguments . [ fig10 ] ] we detected a companion to lp390 - 16 at only @xmath430 separation with a delta magnitude of more than 7mag in h and k band . combining proper motion with point source locations in old exposures from possi , however , we find the object to be consistent with a non - moving object , and therefore conclude it to be a background star . follow - up observations helped to reject 7 companion candidates at separations between @xmath431 and @xmath432 to g125 - 15 . after a first observation in june 2005 the central object moved @xmath433 due to proper motion until a second observation in april 2006 . this corresponds to @xmath434 on the detector , which exposes the fixed companion candidates to be distant background stars . in @xmath435 separation to the central object we detected two faint objects with a separation of @xmath436 . the proper motion of lp756 - 3 ( 0.353 arcsec / yr ) and second epoch data 10 months after our first observation were used to reveal their background nature . four of the 13 observed multiples already have published separations and position angles ; proper motions are known for all . comparing our data taken in 2005 with the earlier epoch data of the other groups we find that all four binaries are consistent with physically related companions . evidence for the physical nature of the other multiples is given by proper motion arguments . our sample consists of relatively nearby m stars that show significant proper motions . looking at multiple epoch data , a background star mimicking a bound companion would remain fixed while the foreground object would move by . as the @xmath123 of all observed binaries range from @xmath437 to @xmath438 only ( implying apparent magnitudes of @xmath439 ) , these bright companions would show up in possi data taken at much earlier epoch . no bright background objects were found in the vicinity of any of the estimated secondary locations leading to the conclusion that these cool companions are physically bound to their primaries . observations of binaries and their inherent parameters are important for setting constraints on the formation mechanism of stars in general . this sample of young , early - m dwarfs allows insight into a population of stars in between the well - known g and k star population and the distinctly different vlm population @xcite , which became analyzable only a few years ago with the development and employment of highly sensitive ao and the _ hst _ ( close et al . 2003 ; gizis et al . 2003 ; bouy et al . 2003 ) . we found 12 binaries and 1 triple system in our sample of 41 objects . in order to derive a multiple star fraction free of malmquist bias , we consider only the systems in our sample whose distances are within 20pc . eight of the observed multiples are likely within 20pc of the sun , as are estimatedly all of the observed single stars . however , the distances of two of the 8 objects ( g172 - 1 , 2m1036 ) are within @xmath440 of the 20pc limit , leaving a rather large probability that they are not part of the defined volume . likewise , two binaries have estimated distances @xmath441 greater than 20pc ( lp717 - 36 and lp642 - 48 ) , so they may be part of the limited sample . this therefore leaves @xmath442 confirmed binaries within 20pc . to eliminate uncertainties caused by the distance dependence of the sensitivity , we conservatively cut the sample at a mass ratio of @xmath443 . there are 28 objects in this sub - sample other than the confirmed binaries . four of these have candidate companions with @xmath444 . hence , we include them in this binary fraction as single stars . in addition , we include one possible binary system in the upper limit estimate . this results in a total of @xmath445 multiples in a sample of @xmath446 objects ( upper and lower limits correspond to each other , respectively ) , resulting in a nominal binary fraction of @xmath447 for @xmath448 and @xmath449au ( we use binomial statistics to estimate the uncertainty as described in @xcite ) . the cut in separation is determined by the minimum resolution of @xmath450 at the distance limit of 20pc . we note , however , that this above examination of the binary fraction does not account for selection effects in the sample selection . figure [ fig11 ] shows the mass ratio @xmath451 distribution derived from the 13 observed multiples using the masses listed in table [ tab3 ] . the triple system 2m1036 is accounted for as two binaries , the binary bc and the a(bc ) system . mass ratios for these two objects are therefore @xmath452 and @xmath453 . in the following paragraphs we sometimes refer to 14 `` binaries '' , but one should keep in mind that this means 12 binaries and 1 triple system . twelve out of the 14 have mass ratios of @xmath454 ; we regard this as evidence that equal - mass systems are preferred for young , early - m dwarf primaries . our observations are sensitive to binaries with mass ratios @xmath444 for separations greater than @xmath455 ( see figure [ fig12 ] ) . observations have shown that there is a significant lack of very low - mass objects around main sequence stars for separations smaller than 5 au ( the `` brown dwarf desert '' ; marcy & butler 2000 ) . we therefore do not expect to have missed a large quantity of objects . however , we have to assume that for @xmath444 the sample is likely incomplete . the continuous line shows the distribution derived from the 14 early - m binaries in this paper ( triple 2m1036 included as two binaries , see [ massratio ] ) . the dashed line shows the distribution of binaries included in an early f to k star sample by @xcite that have separations @xmath456 . the dotted line shows the @xmath3 distribution for vlm stars from @xcite . while the f to k binary distribution is described by a rather flat distribution , slightly more populated at higher @xmath3 , the vlm distribution shows a significant peak for equal mass binaries . the early - m binary distribution with its broad peak at @xmath457 might represent an intermediate distribution between the earlier and later types . due to the high sensitivity of our measurements , only the last , @xmath444 bin is significantly affected by incompleteness . [ fig11 ] ] as a function of projected binary separation in astronomical units for the 12 discovered binaries ( filled circles ) . the triple 2m1036 is included as the two binaries a(bc ) and bc ( squares ; for further explanation see [ massratio ] ) . the dashed line represents the sensitivity limit of our survey , derived from the sensitivity relation shown in figure [ fig3 ] and 500myr models from @xcite . as a demonstration of detection capabilities , the open circle at @xmath458 and @xmath459 is the companion seen in figure [ fig10 ] if it were a physical companion to g173 - 39 . [ fig12 ] ] this emphasis on @xmath460 is not seen in the distributions of earlier type stars . @xcite show that for late f to k binaries with separations of less than 100au the distribution of mass ratios is rather flat with only marginal tendency towards equal mass binaries . for field vlm binaries , however , the opposite effect is observed ; recent statistical evaluations of all currently known vlm binaries @xcite show a significant peak at @xmath461 . it is therefore possible that the mass ratio distribution of our early - m binaries sample has an intermediate distribution between the flat f - k distribution and the strongly peaked field vlm distribution . however , our sample of binaries is rather small . further observations of a larger number of young m stars will be needed to show if this relation holds . mass ratios @xmath462 as a function of binary separation are shown in figure [ fig12 ] . while at separations @xmath463au systems are distributed over a broad range of mass ratios , all three binaries with separations @xmath464au show mass ratios of @xmath465 . the sensitivity of the observations would have allowed for the detection of much lower mass ratios at these separations ( see figure [ fig10 ] for example ) , indicating a real lack of objects with low mass ratios and high separations . however , a sample of 14 binaries can not fill the accessible space in the @xmath466 plane uniformly , limiting the significance of this result due to low number statistics . figure [ fig13 ] shows the separation distribution of our set of 14 binaries . the physical projected separations range from 2.2 to 49.7au and possess a distribution that shows no significant peak . however , since the observations cover separations up to 15 arcseconds , which corresponds to 300au at a distance of 20pc , we conclude that the decreasing number of detected objects for separations higher than @xmath467au is a real effect . the underlying distribution therefore probably peaks within 0 to 50au . a least-@xmath468 fit of a poisson distribution to the data suggests a peak at @xmath2 . this is consistent with earlier results , for instance by @xcite , who report a broad peak at @xmath469@xmath470 for m2m4.5 binaries . bin is probably incomplete due to the finite resolution of @xmath48 corresponding to 1.6au in 20pc . [ fig13 ] ] the separation distribution , again , shows evidence that young , early - m binaries possess transitional characteristics between earlier and later type populations . with a distribution peaking at @xmath471 @xcite , g stars seem to have on average greater separations than early - m stars . field vlm binaries , however , are preferably found in significantly tighter systems . the distribution of 70 binaries with spectral types later than m5.5 shows a significant peak at @xmath472@xmath473 @xcite with a steep quantitative decrease of systems with higher separations . g k binaries are described by a mass ratio distribution and a separation distribution that are very different from the vlm binaries characteristics ; young , early - m binaries appear to show an intermediate behavior . the higher and the lower mass distributions are apparently connected through a smooth transition embodied by early - m binaries . we have conducted a survey of 41 nearby , early - m stars of ages comparable to the pleiades cluster , searching for companions down to planetary masses using the gemini north telescope with its ao system altair . thirteen stars were observed to host fairly bright companions , 8 of these multiples ( 7 binaries , 1 triple ) are new discoveries . we derived physical parameters for each of the multiple components allowing for further analysis of the early - m star population confirming existing statistical results and contributing additional empirical constraints on star and binary formation . the binary fraction of this sample of young , early - m stars was found to be @xmath447 for mass ratios @xmath448 and @xmath449au . however , a full analysis of systematic errors and biases could not be made , this value might therefore differ significantly from the actual value for this group of stars . the mass ratio distribution was analyzed and compared with distributions drawn from f k stars and from vlm stars . the young , early - m binary data was shown to be best described as an intermediate distribution between the almost flat early type star distribution and the strongly @xmath474 peaked field vlm distribution , however , our sample size is relatively small . the most likely separation was found to be @xmath2 , which is less than the median value of the g binary distribution . field vlm binaries , however , tend to be bound in significantly tighter systems . we conclude that early - m binaries possess characteristics filling the gap between the very distinct field vlm population and stars earlier than m. a sensitivity estimate to the detection of planetary mass companions was made using planetary evolution models . since we found no faint physical companions around 37 of the observed sources we can exclude conservatively assuming an age of 500myr companions more massive than 10m@xmath475 at separations of @xmath383 and heavier than @xmath476 at separations of @xmath382 around all these objects . we thank the anonymous referee for many valuable comments leading to an improved paper . the authors are grateful to jean - ren roy , deputy director & head of science at gemini north , for kindly allocating special ddt time for follow - up observations . this work is based 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 particle physics and astronomy research council ( united kingdom ) , the national research council ( canada ) , conicyt ( chile ) , the australian research council ( australia ) , cnpq ( brazil ) and conicet ( argentina ) . program ids of the gemini observations are gn-2005a - q-28 , gn-2005b - q-14 , and gn-2006a - q-83 . this publication makes use of data products from the two micron all sky survey , which is a joint project of the university of massachusetts and the infrared processing and analysis center / california institute of technology , funded by the national aeronautics and space administration and the national science foundation . this research has made use of the simbad database , operated at cds , strasbourg , france . the national geographic society - palomar observatory sky atlas ( poss - i ) was made by the california institute of technology with grants from the national geographic society . lmc is supported by an nsf career award and the nasa origins of solar systems program . allen , p. r. , koerner , d. w. , reid , i. n. , & trilling , d. e. 2005 , , 625 , 385 baraffe , i. , chabrier , g. , allard , f. , & hauschildt , p. h. 1998 , , 337 , 403 baraffe , i. , chabrier , g. , barman , t. s. , allard , f. , & hauschildt , p. h. 2003 , , 402 , 701 becklin , e. e. , & zuckerman , b. 1988 , , 336 , 656 beers , t. c. , rossi , s. , ulrich , d. , & wilhelm , r. 1996 , , 112 , 1188 bergeron , p. , wesemael , f. , & beauchamp , a. 1995 , , 107 , 1047 beuzit , j .- l . , et al . 2004 , , 425 , 997 bouy , h. , brandner , w. , martn , e. l. , delfosse , x. , allard , f. , & basri , g. 2003 , , 126 , 1526 burgasser , a. j. , kirkpatrick , j. d. , reid , i. n. , brown , m. e. , miskey , c. l. , gizis , j. e. 2003 , , 586 , 512 burgasser , a. j. , reid , i. n. , siegler , n. , close , l. , allen , p. a. , lowrance , p. , gizis , j. e. 2006 , in _ protostars & planets v _ b. reipurth , u. arizona press burrows , a. , hubbard , w.b . , lunine , j.i . , liebert , j. 2001 , rev . phys . , 73 , 719 close , l. m. , potter , d. , brandner , w. , lloyd - hart , m. , liebert , j. , burrows , a. , & siegler , n. 2002 , , 566 , 1095 close , l. m. , siegler , n. , freed , m. , & biller , b. 2003 , , 587 , 407 close , l. m. et al . 2006 , , submitted cruz , k. l. , & reid , i. n. 2002 , , 123 , 2828 cruz , k. l. , reid , i. n. , liebert , j. , kirkpatrick , j. d. , & lowrance , p. j. 2003 , , 126 , 2421 delfosse , x. , forveille , t. , sgransan , d. , beuzit , j .- l . , udry , s. , perrier , c. , & mayor , m. 2000 , , 364 , 217 duquennoy , a. , & mayor , m. 1991 , , 248 , 485 fischer , d.a . & marcy , g. w. 1992 , , 396 , 178 gizis , j. e. , reid , i. n. , knapp , g. r. , liebert , j. , kirkpatrick , j. d. , koerner , d. w. , & burgasser , a. j. 2003 , , 125 , 3302 gliese , w. , & jahrei , h. 1991 , on : the astronomical data center cd - rom : selected astronomical catalogs , vol . i ; l.e . brotzmann , s.e . gesser ( eds . ) , nasa / astronomical data center , goddard space flight center , greenbelt , md golimowski , d. a. , et al . 2004 , , 128 , 1733 gould , a. , & chanam , j. 2004 , , 150 , 455 hawley , s. l. , gizis , j. e. , & reid , i. n. 1996 , , 112 , 2799 hodapp , k. w. , et al . 2003 , , 115 , 1388 hnsch , m. , schmitt , j. h. m. m. , sterzik , m. f. , & voges , w. 1999 , , 135 , 319 jao , w .- c . , henry , t. j. , subasavage , j. p. , bean , j. l. , costa , e. , ianna , p. a. , & mndez , r. a. 2003 , , 125 , 332 jones , b. f. , fischer , d. a. , & stauffer , j. r. 1996 , , 112 , 1562 kirkpatrick , j. 1996 , hst proposal , 2795 kirkpatrick , j. d. , dahn , c. c. , monet , d. g. , reid , i. n. , gizis , j. e. , liebert , j. , & burgasser , a. j. 2001 , , 121 , 3235 luyten , w. 1980 , nltt catalogue ( minneapolis , university of minnesota ) marcy , g. w. , & butler , r. p. 2000 , , 112 , 137 mccarthy , c. , zuckerman , b. , & becklin , e. e. 2001 , , 121 , 3259 martn , e. l. , barrado y navascus , d. , baraffe , i. , bouy , h. , & dahm , s. 2003 , , 594 , 525 micela , g. , sciortino , s. , harnden , f. r. , kashyap , v. , prosser , c. f. , damiani , f. , stauffer , j. , caillault , j .- p . 1999 , , 341 , 751 nakajima , t. , oppenheimer , b. r. , kulkarni , s. r. , golimowski , d. a. , matthews , k. , & durrance , s. t. 1995 , , 378 , 463 oppenheimer , b. r. , golimowski , d. a. , kulkarni , s. r. , matthews , k. , nakajima , t. , creech - eakman , m. , & durrance , s. t. 2001 , , 121 , 2189 perryman , m. a. c. , & esa 1997 , the hipparcos and tycho catalogues . astrometric and photometric star catalogues derived from the esa hipparcos space astrometry mission , publisher : noordwijk , netherlands : esa publications division , 1997 , series : esa sp series vol no : 1200 , isbn : 9290923997 ( set ) reid , n. 2003 , , 342 , 837 reid , i. n. , gizis , j. e. , & hawley , s. l. 2002 , , 124 , 2721 reid , i. n. , gizis , j. e. , kirkpatrick , j. d. , & koerner , d. w. 2001 , , 121 , 489 reid , i. n. , hawley , s. l. , & gizis , j. e. 1995a , , 110 , 1838 reid , i. n. , hawley , s. l. , mateo , m. 1995b , , 272 , 828 reid , i. n. , et al . 2003 , , 126 , 3007 reid , i. n. , et al . 2004 , , 128 , 463 riaz , b. , gizis , j. e. , & harvin , j. 2006 , , 132 , 866 siegler , n. , close , l. m. , cruz , k. l. , martn , e. l. , & reid , i. n. 2005 , , 621 , 1023 scholz , r .- d . , meusinger , h. , & jahreiss , h. 2005 , vizier online data catalog , 344 , 20211 simons , d. a. , henry , t. j. , & kirkpatrick , j. d. 1996 , , 112 , 2238 van biesbroeck , g. 1944 , , 51 , 61 voges , w. , et al . 1999 , , 349 , 389 voges , w. , et al . 2000 , , 7432 , 1 zuckerman , b. , & song , i. 2004 , , 42 , 685 zuckerman , b. , song , i. , & bessell , m. s. 2004 , , 613 , l65

the interlace polynomials introduced by arratia , bollobs and sorkin extend to invariants of graphs with vertex weights , and these weighted interlace polynomials have several novel properties . one novel property is a version of the fundamental three - term formula that lacks the last term . it follows that interlace polynomial computations can be represented by binary trees rather than mixed binary - ternary trees . binary computation trees provide a description of that is analogous to the activities description of the tutte polynomial . if is a tree or forest then these `` algorithmic activities '' are associated with a certain kind of independent set in . three other novel properties are weighted pendant - twin reductions , which involve removing certain kinds of vertices from a graph and adjusting the weights of the remaining vertices in such a way that the interlace polynomials are unchanged . these reductions allow for smaller computation trees as they eliminate some branches . if a graph can be completely analyzed using pendant - twin reductions then its interlace polynomial can be calculated in polynomial time . an intuitively pleasing property is that graphs which can be constructed through graph substitutions have vertex - weighted interlace polynomials which can be obtained through algebraic substitutions . keywords . interlace polynomial , vertex weight , pendant vertex , twin vertex , series , parallel , graph composition , graph substitution , join , computational complexity , tree , tutte polynomial , jones polynomial mathematics subject classification .

0808.1888

motivated by problems arising from dna sequencing and the properties of circle graphs of 2-in , 2-out digraphs , arratia , bollobs and sorkin introduced a new family of graph invariants , the interlace polynomials , in @xcite . these invariants may be defined either through recursive elimination of vertices or as sums indexed by subsets of vertices @xcite , much as the tutte polynomial may be defined either through recursive elimination of edges or as a sum indexed by subsets of edges . ellis - monaghan and sarmiento @xcite have shown ( among other results ) that one of the one - variable interlace polynomials , @xmath3 , can be computed in polynomial time for bipartite distance hereditary graphs . their proof depends on the corresponding result for the tutte polynomials of series - parallel graphs , first proved in @xcite , and the fact that bipartite distance hereditary graphs are circle graphs of euler circuits in medial graphs of series - parallel graphs . in this paper we discuss several useful features of interlace polynomials that have been modified to incorporate vertex weights . after defining the weighted interlace polynomials in section 2 , we observe that a simple adjustment of weights makes it unnecessary to have the third term in the fundamental recursion @xmath4 of @xcite . in section 3 we present reduction formulas that can be used to eliminate a vertex that is a twin of another , or pendant on another . these pendant - twin reductions are analogous to the series - parallel reductions of electrical circuit theory , in which two resistors wired in series ( resp . parallel ) are replaced by one resistor with @xmath5 ( resp . @xmath6 ) . the pendant - twin reductions are used to extend the result of ellis - monaghan and sarmiento mentioned above to the two - variable interlace polynomials of looped , non - bipartite distance hereditary graphs . in section 4 we show that if @xmath7 is a looped graph obtained using the composition construction of cunningham @xcite then @xmath1 is equal to the interlace polynomial of a suitably re - weighted version of @xmath8 ; this generalizes results of @xcite that describe the @xmath3 polynomials of simple graphs constructed through substitution . composition has proven useful in the study of circle graphs ( see for instance @xcite ) , so it is not surprising to see it appear in the theory of the interlace polynomials . in section 5 we discuss some elementary properties of the unweighted @xmath3 polynomial , focusing on simple ( unlooped ) graphs . in section 6 we sketch a combinatorial description of the interlace polynomials of trees and forests introduced by anderson , cutler , radcliffe and the present author in @xcite . this combinatorial description bears a striking resemblance to the activities description of the tutte polynomial , and in section 7 we extend it to arbitrary graphs using activities defined with respect to recursive interlace polynomial calculations . we do not know whether or not these activities have a convenient combinatorial description in general . we should observe that the idea of using vertex weights for interlace polynomials has appeared before , though our implementation of the idea is different from those we have seen elsewhere . in @xcite , courcelle introduced a multivariate interlace polynomial that is more complicated than the polynomials we consider here , and involves assigning indeterminates to the vertices of a graph . he used monadic second - order logic to show that it is possible to compute bounded portions of this polynomial ( and the entire unweighted interlace polynomial @xmath9 ) in polynomial time for graphs of bounded clique - width . this technique is quite general but involves large built - in constants , so for pendant - twin reductions and compositions the formulas presented here are considerably more practical . also , blser and hoffmann @xcite use the idea of assigning indeterminates to vertices , along with the adjunction of two types of pendant - twin vertices , to show that evaluating interlace polynomials is generally @xmath10-hard for almost all values of the variables . in @xcite arratia , bollobs and sorkin observe that there is a natural ( so natural it is `` practically a tautology '' ) correspondence between the kauffman bracket of an alternating link diagram and an interlace polynomial of an associated 2-in , 2-out digraph . the situation is clear enough that we do not discuss it in detail , but it is worth mentioning that this correspondence may be extended to arbitrary link diagrams using vertex weights . the well - known relationship between the jones and tutte polynomials is similar , in that an edge - weighted or -signed version of the tutte polynomial conveniently incorporates crossing information when dealing with non - alternating links @xcite . we recall terminology and notation of @xcite . _ graphs _ may have loops but not multiple edges or multiple loops . the _ rank _ @xmath11 and _ nullity _ @xmath12 of a graph @xmath2 are those of its adjacency matrix considered over @xmath13 . if @xmath14 then @xmath15 $ ] is the subgraph of @xmath2 induced by @xmath16 . in addition , we say a graph is ( _ vertex- _ ) _ weighted _ by functions @xmath17 and @xmath18 mapping @xmath19 into some commutative ring with unity @xmath20 . for ease of notation we prefer to denote a weighted graph @xmath2 rather than using the triple @xmath21 ; even when two graphs differ only in their weights we will denote them @xmath2 and @xmath22 rather than @xmath21 and @xmath23 . also , if a graph is modified then unless otherwise stated , we presume that the weight functions @xmath17 and @xmath18 are modified in the most natural way . for instance , if @xmath24 then the other vertices of @xmath2 have the same vertex weights in @xmath25 as they have in @xmath2 . if @xmath2 is a weighted graph then the _ unweighted version _ of @xmath2 is denoted @xmath26 ; it has the same underlying graph and the trivial weights @xmath27 . [ 2varq]if @xmath2 is a vertex - weighted graph then the _ weighted interlace polynomial _ of @xmath2 is @xmath28)}(y-1)^{n(g[s])}.\ ] ] if @xmath2 is a vertex - weighted graph then the _ weighted vertex - nullity interlace polynomial _ of @xmath2 is @xmath29)}.\ ] ] if @xmath2 is a vertex - weighted graph then the _ weighted vertex - rank interlace polynomial _ of @xmath2 is @xmath30)}.\ ] ] the original , unweighted interlace polynomials of @xmath2 are recovered by using @xmath31 the three weighted interlace polynomials of a particular weighted graph @xmath2 may be substantially different from each other . considered as functions defined on the class of all weighted graphs , however , the three polynomials are essentially equivalent : if @xmath32 is obtained from @xmath2 by re - weighting @xmath33 using the indeterminates in the polynomial ring @xmath34 $ ] , then the value of any one weighted interlace polynomial on @xmath35 determines the values of all three polynomials on all weighted versions of @xmath2 . for instance , @xmath1 can be obtained from @xmath36 by substituting @xmath37 for @xmath38 , @xmath39 for @xmath40 and @xmath41 for @xmath42 . ( in contrast , the functions defined by the three unweighted interlace polynomials are rather different from one another , as there are many pairs of graphs distinguished by @xmath43 but not by @xmath3 @xcite . ) we mention all three weighted polynomials simply because one of them may be more convenient for some purposes than the others . most of our results are stated for @xmath9 because it specializes to the others most readily . in fact any one of @xmath44 , @xmath36 , @xmath45 determines @xmath2 up to isomorphism , because the vertex weights identify the contribution of each @xmath14 . the looped vertices of @xmath2 appear in the 1-element subsets @xmath16 of rank 1 ( nullity 0 ) , some pairs of adjacent vertices of @xmath2 appear in the 2-element subsets @xmath16 that contain at least one unlooped vertex and have rank 2 ( nullity 0 ) , and the other pairs of adjacent vertices of @xmath2 appear in the 2-element subsets @xmath16 that contain two looped vertices and have rank 1 ( nullity 1 ) . moreover the same comment applies even if only one of @xmath46 is nontrivial , i.e. , if @xmath47 or @xmath48 . this might make it seem possible to simplify our discussion by considering only @xmath17 or only @xmath18 , but we prefer to use both weights because they produce especially easy - to - read formulas , as noted in the discussions of theorems [ theorem1 ] and [ theorem3 ] below . setting one weight identically equal to 1 would not simply leave us with the other weight ; in essence , the remaining one would replace the ratio of the original two . consequently , using only @xmath17 or only @xmath18 would entail unnecessary algebraic complications and losses of generality , because many formulas would require division . for instance , with @xmath47 proposition [ prop1 ] would state that replacing @xmath49 with @xmath50 results in a graph @xmath22 with @xmath51 . the special case @xmath52 would require a separate statement , and of course @xmath53 would be ruled out . our first proposition follows immediately from definition [ 2varq ] . [ prop1]suppose @xmath24 and @xmath54 . let @xmath22 be obtained from @xmath2 by changing the weights of @xmath55 to @xmath56 and @xmath57 . then @xmath58 . a fundamental property of the interlace polynomials is that they can be calculated recursively with the local complementation and pivot operations used by kotzig @xcite , bouchet @xcite and arratia , bollobs and sorkin @xcite . ( local complementation ) if @xmath55 is a vertex of @xmath2 then @xmath59 is obtained from @xmath2 by toggling adjacencies @xmath60 involving neighbors of @xmath55 that are distinct from @xmath55 . ( pivot ) if @xmath55 and @xmath61 are distinct vertices of @xmath2 then the graph @xmath62 is obtained from @xmath2 by toggling adjacencies @xmath60 such that @xmath63 , @xmath64 is adjacent to @xmath55 in @xmath2 , @xmath42 is adjacent to @xmath61 in @xmath2 , and either @xmath64 is not adjacent to @xmath61 or @xmath42 is not adjacent to @xmath55 . note that local complementation includes loop - toggling ( when @xmath65 is a neighbor of @xmath55 distinct from @xmath55 ) but pivoting does not . [ theorem1]if @xmath2 is a weighted graph then @xmath1 can be calculated recursively using the following properties . \(a ) if @xmath55 is a looped vertex then @xmath66 \(b ) if @xmath55 and @xmath61 are loopless neighbors then @xmath67 \(c ) if @xmath2 has no non - loop edges then @xmath1 is @xmath68 the proofs of parts ( a ) and ( b ) of theorem [ theorem1 ] are essentially the same as the proofs of the corresponding formulas for the unweighted two - variable interlace polynomial @xcite . indeed , if we read @xmath17 as includes and @xmath18 as excludes then the weighted formulas serve as mnemonic devices to recall the proofs . for instance part ( a ) is proven as follows . the term @xmath69 reflects the fact that if @xmath70 then @xmath15=(g - a)[s]$ ] , so the contributions of @xmath16 to @xmath1 and @xmath71 differ only by a factor @xmath72 . the term @xmath73 reflects the fact that if @xmath74 then @xmath75)=r \begin{pmatrix } 1 & \mathbf{1 } & \mathbf{0}\\ \mathbf{1 } & m_{11 } & m_{12}\\ \mathbf{0 } & m_{21 } & m_{22}% \end{pmatrix } = 1+r \begin{pmatrix } m_{11}^{c } & m_{12}\\ m_{21 } & m_{22}% \end{pmatrix } = 1+r((g^{a}-a)[s - a]),\ ] ] where bold numerals represent rows and columns and @xmath76 differs from @xmath77 in every entry . consequently the contributions of @xmath16 to @xmath1 and @xmath78 to @xmath79 differ by a factor @xmath80 . the more complicated formula of part ( b ) reflects , among other things , the fact that each subset @xmath14 with @xmath81 contributes to all three terms ; the last two contributions cancel each other . part ( c ) follows directly from definition [ 2varq ] . it is not technically necessary to mention graphs with loops in ( c ) , as loops can always be removed using part ( a ) . however applying part ( a ) to completely disconnected graphs is obviously inefficient . other properties of the unweighted interlace polynomials also extend naturally to the weighted polynomials . for instance , the second and third parts of theorem [ theorem2 ] below extend two properties discussed in section 3 of @xcite . the first part will be useful in section 4 , where we discuss substituted graphs . [ theorem2](a ) if @xmath24 is loopless then @xmath82 \(b ) if @xmath83 are loopless neighbors then @xmath84 \(c ) if @xmath2 is the union of disjoint subgraphs @xmath85 and @xmath86 then @xmath87 . to prove ( a ) , note that if @xmath88 then the adjacency matrix of @xmath15 $ ] may be represented by @xmath89 adding the first row to each of those in the second group results in @xmath90 the adjacency matrix of @xmath91 $ ] . definition [ 2varq ] then tells us that @xmath92)}(y-1)^{n(g[s])}\\ & ~\\ & = \sum_{a\in s\subseteq v(g)}(% % tcimacro{\dprod \limits_{s\in s}}% % beginexpansion { \displaystyle\prod\limits_{s\in s } } % endexpansion \alpha(s))(% % tcimacro{\dprod \limits_{v\not \in s}}% % beginexpansion { \displaystyle\prod\limits_{v\not \in s } } % endexpansion \beta(v))(x-1)^{r(g^{a}[s])}(y-1)^{n(g^{a}[s])}\\ & ~\\ & = q(g^{a})-\beta(a)q(g^{a}-a).\end{aligned}\ ] ] the proofs of ( b ) and ( c ) are essentially the same as the proofs of the corresponding results in @xcite . many properties of the unweighted interlace polynomials extend directly to the weighted polynomials , as we see in theorems [ theorem1 ] and [ theorem2 ] . it may be a surprise that some properties of the unweighted interlace polynomials can be significantly simplified using vertex weights . for instance , consider a computation tree representing a recursive implementation of theorem [ theorem1 ] . the tree has two branches for each application of part ( a ) , three branches for each application of part ( b ) , and a leaf for each application of part ( c ) . as noted in the proof of theorem [ theorem1 ] , the three terms of part ( b ) all incorporate contributions from the same subsets @xmath14 . the recursive computation calculates these same contributions three times , on separate branches . this inefficiency can be eliminated by rephrasing part ( b ) of theorem [ theorem1 ] so that no three - fold branches are necessary . [ cor1]if @xmath55 and @xmath61 are loopless neighbors in @xmath2 then @xmath93 , where @xmath94 is obtained from @xmath95 by changing the weights of @xmath55 to @xmath96 and @xmath97 . let @xmath98 be the graph obtained from @xmath95 by changing the weights of @xmath55 as in proposition [ prop1 ] , with @xmath99 and @xmath100 @xmath101 . then @xmath102 and @xmath103 , so @xmath104 = @xmath105 . on the other hand , proposition [ prop1 ] tells us that @xmath106 if @xmath107 and @xmath108 are used instead of the weights given in corollary [ cor1 ] , then the resulting formula @xmath109 still has only two branches , and bears an interesting resemblance to part ( a ) of theorem [ theorem1 ] . however using division to define @xmath110 may occasionally cause some algebraic difficulties , e.g. , it complicates the evaluation at @xmath111 , and it prohibits the use of rings in which @xmath112 is a divisor of zero . corollary [ cor1 ] is one of several results in which vertex weights allow us to extend properties of the unweighted version of @xmath3 , seemingly the simplest kind of interlace polynomial , to the other interlace polynomials . in this instance the extended property is that of possessing a recursive description represented by a binary computation tree . another result of this type is corollary [ cor2 ] , which extends remark 18 of @xcite : if @xmath55 and @xmath61 are loopless neighbors in @xmath2 then the unweighted vertex - nullity interlace polynomials of @xmath2 and @xmath62 are the same . [ cor2]let @xmath55 and @xmath61 be unlooped neighbors in @xmath2 , and let @xmath113 be the weighted graph obtained from @xmath62 by changing the weights of @xmath55 and @xmath61 to @xmath114 , @xmath115 , @xmath116 and @xmath117 . then @xmath118 . applying corollary [ cor1 ] to @xmath113 , with the roles of @xmath55 and @xmath61 reversed , tells us that @xmath119 = @xmath120 . observe that @xmath121 has the underlying graph @xmath122 , and differs from @xmath25 only in the weights of @xmath61 , which are given by @xmath123 and @xmath124 . consequently @xmath125 = @xmath126 = @xmath127 . as @xmath128 = @xmath129 , the result follows directly from corollary [ cor1 ] . if we are confident that division by @xmath112 will cause no trouble , corollary [ cor2 ] may be restated with a simpler conclusion : using @xmath130 , @xmath108 , @xmath131 and @xmath132 yields @xmath133 . other novel properties of the weighted interlace polynomials arise from a general observation . suppose @xmath83 happen to have the property that for @xmath14 , the rank and nullity of @xmath16 are determined by the rank and nullity of @xmath134 , perhaps in different ways according to which of @xmath135 is contained in @xmath16 . then it may be possible to adjust @xmath49 and @xmath136 in such a way that the weighted interlace polynomial of the weight - adjusted version of @xmath137 incorporates all the information in @xmath1 . the simplest instance of this observation occurs when @xmath55 and @xmath61 give rise to identical rows and columns in the adjacency matrix of @xmath2 . two vertices @xmath135 of @xmath2 are _ _ identical twins _ _ if ( i ) either they are looped and adjacent or they are unlooped and not adjacent , and ( ii ) they have the same neighbors outside @xmath138 . identical twins are called _ clones _ in @xcite , and unlooped identical twins are called _ duplicates _ in @xcite and _ false twins _ in @xcite . we prefer the present terminology because the adjective _ false _ seems inappropriate , and because we do not know what we would call non - identical duplicates or clones in definition 7 below . the following result extends proposition 40 of @xcite , proposition 4.14 of @xcite and section 3.1 of @xcite to vertex - weighted graphs . [ theorem3]suppose @xmath55 and @xmath61 are identical twins in @xmath2 . let @xmath22 be the graph obtained from @xmath137 by changing the weights of @xmath55 : @xmath139 and @xmath140 . then @xmath141 . if @xmath142 then @xmath143)$ ] = @xmath144)$ ] , because the adjacency matrices are the same . if @xmath145 then @xmath143)$ ] = @xmath144)$ ] = @xmath146)$ ] = @xmath147)$ ] , because the only difference among the adjacency matrices is that the matrix of @xmath148 $ ] has two identical rows and columns , corresponding to @xmath55 and @xmath61 . as mentioned above , it is helpful to read @xmath17 as includes and @xmath18 as excludes so that ( for instance ) the appearance of @xmath149 in @xmath150 indicates that if @xmath151 then the summand of @xmath152 corresponding to @xmath16 includes the summand of @xmath1 corresponding to @xmath153 . theorem [ theorem3 ] has the following inductive generalization . suppose @xmath154 and @xmath155 are identical twins in @xmath2 . if @xmath22 is the graph obtained from @xmath156 by changing the weights of @xmath55 to @xmath157 then @xmath158 . we refer to the process of combining several identical twins into a single vertex as an _ identical twin reduction _ no matter how many identical twins are combined . [ theorem4]suppose @xmath61 is an unlooped degree - one vertex pendant on @xmath55 . let @xmath22 be the graph obtained from @xmath137 by changing the weights of @xmath55 : @xmath159 and @xmath160 . then @xmath158 . if @xmath151 then @xmath143)$ ] = @xmath144)$ ] , and if @xmath142 then @xmath143)$ ] = @xmath161)-2 $ ] = @xmath146)$ ] = @xmath144)$ ] . if @xmath162 are unlooped and pendant on @xmath55 then @xmath163 are identical twins , so we can combine them into a single re - weighted vertex @xmath61 using theorem [ theorem3 ] and then remove @xmath164 using theorem [ theorem4 ] . we refer to the removal of any number of unlooped vertices pendant on the same vertex as an _ unlooped pendant vertex reduction_. theorem [ theorem4 ] extends proposition 4.12 of @xcite and section 3.2 of @xcite , where collections of pendant vertices are called _ combs_. a recursive calculation that depends solely on theorem [ theorem1 ] and corollary [ cor1 ] is represented by a mixed binary - ternary computation tree ; if corollary [ cor1 ] is always used in place of part ( b ) of theorem [ theorem1 ] then the tree will be binary . using identical twin and unlooped pendant vertex reductions makes the computation tree smaller , because each time one of these reductions is used , we avoid splitting the resulting portion of the calculation into branches . similarly , a tutte polynomial computation that incorporates series - parallel reductions and deletion - contraction operations will generally result in a smaller formula than a computation that involves only deletion - contraction operations can provide . this latter observation was made precise in @xcite : computing the tutte polynomial of a matroid @xmath165 using series - parallel reductions and deletion - contraction operations will result in an expression with at least @xmath166 terms , where @xmath166 is crapo s @xmath18 invariant @xcite ; moreover if @xmath167 the lower bound is attainable . ( this result is merely the extension to the tutte polynomial of the important theory of reliability domination ; see @xcite for expositions . ) for interlace polynomials , analogous lower bounds are derived from the unweighted vertex - nullity polynomial @xmath168 . the coefficient of @xmath42 in @xmath169 is denoted @xmath170 as in @xcite , and the evaluation @xmath171 is denoted @xmath172 . [ theorem5]if part ( c ) of theorem [ theorem1 ] is applied only to loopless graphs then a computation of @xmath1 using corollary [ cor1 ] and theorems [ theorem1 ] , [ theorem3 ] and [ theorem4 ] is represented by a computation tree with no fewer than @xmath173 leaves . if @xmath2 is a simple graph then the computation tree has no fewer than @xmath174 leaves . we quickly review some elementary properties of @xmath169 from @xcite . this polynomial is described recursively as follows . if @xmath2 has a loop at @xmath55 then @xmath169 = @xmath175 + @xmath176 , if @xmath55 and @xmath61 are loopless neighbors then @xmath169 = @xmath176 + @xmath177 , and if @xmath178 is the edgeless @xmath179-vertex graph then @xmath180 . the latter includes the empty graph @xmath181 with @xmath182 . it follows by induction on the number of vertices that no graph has any negative coefficient in @xmath169 , and hence every graph has @xmath183 and @xmath184 . moreover , every nonempty simple graph has @xmath185 and every disconnected simple graph has @xmath186 . theorem [ theorem5 ] is certainly true for @xmath178 , as @xmath187 . proceeding inductively , observe that if corollary [ cor1 ] or part ( a ) or ( b ) of theorem [ theorem1 ] is applied to a graph @xmath188 represented by a certain node of the computation tree , and @xmath189 and @xmath190 are the graphs represented by the ( first two ) resulting branch nodes , then @xmath191 = @xmath192 + @xmath193 ; certainly then @xmath194 = @xmath195 + @xmath196 . ( by the way , we call vertices of the computation tree _ nodes _ in order to distinguish them from vertices of @xmath2 . ) if theorem [ theorem3 ] or [ theorem4 ] is applied to remove a vertex @xmath61 from a graph @xmath197 then @xmath198 = @xmath199 , because the formulas of theorems [ theorem3 ] and [ theorem4 ] yield @xmath200 when @xmath49 = @xmath136 = @xmath201 = @xmath202 = @xmath203 , @xmath204 and @xmath205 . it remains to consider the special case involving simple graphs . we actually prove a slightly different result , namely : if @xmath2 is simple then the portion of the computation tree involving only nodes corresponding to connected graphs has no fewer than @xmath206 leaves . if @xmath2 is disconnected then @xmath186 so @xmath2 satisfies the result trivially . if @xmath207 is a connected , simple graph with @xmath208 then @xmath2 satisfies the result because @xmath209 . proceeding inductively , observe that if corollary [ cor1 ] or theorem [ theorem1 ] ( b ) is applied to a connected graph @xmath188 represented by a certain node of the computation tree , and @xmath189 and @xmath190 are the graphs represented by the ( first two ) resulting branch nodes , then @xmath191 = @xmath192 + @xmath193 . it follows that @xmath210 = @xmath211 + @xmath212 . if theorem [ theorem3 ] or theorem [ theorem4 ] is used to remove a vertex @xmath61 from a connected , simple graph @xmath188 with 3 or more vertices then @xmath213 , by corollary 4.17 of @xcite . theorem [ theorem5 ] is of limited value because the computations discussed are not optimal . the restriction that part ( c ) of theorem [ theorem1 ] is only applied to loopless graphs is an obvious inefficiency . in addition , if some ( combinations of ) weights are 0 then it would be natural to simply ignore the corresponding parts of the computation . there are also other useful twin reductions that do not fall under theorem [ theorem5 ] . two vertices @xmath135 of @xmath2 are _ _ fraternal twins _ _ if ( i ) either they are looped and nonadjacent or they are unlooped and adjacent , and ( ii ) they have the same neighbors outside @xmath138 . unlooped fraternal twins are called _ true twins _ in @xcite , but we prefer the present terminology because the rows and columns of the adjacency matrix corresponding to fraternal twins are not quite the same . here is an extension of proposition 4.15 of @xcite to weighted graphs . [ theorem6]suppose @xmath55 and @xmath61 are fraternal twins in @xmath2 . let @xmath22 be the graph obtained from @xmath137 by changing the weights of @xmath55 : @xmath150 = @xmath214 and @xmath215 = @xmath216 . then @xmath158 . if @xmath151 then @xmath143)$ ] = @xmath144)$ ] = @xmath217)$ ] , because the adjacency matrices are the same . if @xmath142 then @xmath143)$ ] = @xmath144)$ ] = @xmath161)-2 $ ] . theorem [ theorem6 ] has the following inductive generalization . suppose @xmath154 and @xmath155 are fraternal twins in @xmath2 . if @xmath22 is the graph obtained from @xmath156 by changing the weights of @xmath55 to @xmath218 then @xmath158 . we refer to the process of combining several fraternal twins into a single vertex as a _ fraternal twin reduction _ no matter how many fraternal twins are combined . in general , fraternal twin reductions are just as useful as identical twin reductions . however , as noted in proposition 38 of @xcite and corollary 4.16 of @xcite they have the effect of multiplying the unweighted vertex - nullity polynomial by powers of 2 . ( observe that if @xmath49 = @xmath136 = @xmath201 = @xmath202 = @xmath203 and @xmath204 then theorem [ theorem6 ] gives @xmath150 = @xmath215 = @xmath219 . ) consequently the lower bounds of theorem [ theorem5 ] are not valid for computations that utilize theorem [ theorem6 ] along with corollary [ cor1 ] and theorems [ theorem1 ] , [ theorem3 ] and [ theorem4 ] . if we are given a reduction of a graph @xmath2 to disconnected vertices using twin reductions and unlooped pendant vertex reductions , then theorems [ theorem3 ] , [ theorem4 ] and [ theorem6 ] describe @xmath1 in linear time simply update the vertex weights at each step , and at the end refer to part ( c ) of theorem [ theorem1 ] . if we are not given such a reduction then determining whether or not any such reduction exists , and finding one if possible , can be accomplished in polynomial time : as in corollary 5.3 of @xcite , simply search @xmath19 repeatedly for unlooped degree - one vertices and pairs of vertices @xmath135 with the same neighbors outside @xmath138 . hence if @xmath2 has such a reduction then in polynomial time , theorems [ theorem3 ] , [ theorem4 ] and [ theorem6 ] provide a description of @xmath1 that completely avoids the branching formulas of corollary [ cor1 ] and parts ( a ) and ( b ) of theorem [ theorem1 ] . this observation extends theorem 6.4 of @xcite from @xmath3 to @xmath9 and from simple graphs that can be analyzed without fraternal twin reductions to looped graphs that can be analyzed using all three types of reductions : if a graph @xmath2 can be reduced to a collection of disconnected vertices using unlooped pendant vertex reductions and the two types of twin reductions then theorems [ theorem3 ] , [ theorem4 ] and [ theorem6 ] provide a polynomial - time description of @xmath1 . the theorem refers to a _ description _ rather than a _ computation _ because we have ignored the cost of arithmetic operations in the ring @xmath20 . in case @xmath220 arithmetic operations have low cost , and describing a weighted polynomial @xmath1 is the same as computing an evaluation of the unweighted polynomial @xmath1 . the full unweighted polynomial may be recovered from several evaluations by interpolation , so the theorem provides a polynomial - time computation of the unweighted polynomial . in more complicated rings like @xmath221 $ ] arithmetic operations may be so expensive as to prohibit polynomial - time computation of entire weighted polynomials . a thorough discussion of these matters is given by courcelle @xcite . in this section we reformulate and extend some results of arratia , bollobs and sorkin @xcite regarding substituted graphs , using the following version of a construction introduced by cunningham @xcite . [ comp]a vertex - weighted graph @xmath2 is the _ composition _ of vertex - weighted graphs @xmath188 and @xmath8 , @xmath7 , if the following conditions hold . \(a ) @xmath222 consists of a single unlooped vertex @xmath55 . \(b ) the vertex @xmath55 is unweighted in both @xmath188 and @xmath8 , i.e. , @xmath223 in @xmath188 and @xmath8 . \(c ) @xmath224 , and the vertices of @xmath2 inherit their weights from @xmath188 and @xmath8 . \(d ) @xmath225 and @xmath226 . requiring @xmath55 to be unlooped and unweighted in both @xmath188 and @xmath8 guarantees that no information is lost when we remove @xmath55 in constructing @xmath2 . [ h ] weighif.ps definition [ comp ] includes several other familiar notions . if @xmath55 is isolated in @xmath188 or @xmath8 then @xmath227 is the disjoint union of @xmath228 and @xmath229 . if @xmath55 is adjacent to every other vertex of @xmath188 and @xmath8 then @xmath227 is also denoted @xmath230 . this is traditionally called a join but that term has recently been used for general compositions @xcite . if @xmath55 is adjacent to every other vertex of @xmath188 then @xmath227 is the graph obtained by substituting @xmath228 for @xmath55 in @xmath8 . if @xmath55 is adjacent to every other vertex of @xmath188 and @xmath228 is edgeless or complete then the ( un)looped vertices of @xmath228 are twins in @xmath227 . the following observation will be useful . [ prop2 ] suppose two graphs @xmath231 and @xmath232 are identical except for the weights of a single vertex @xmath55 , and let @xmath233 be the graph that is identical to both @xmath231 and @xmath232 except for @xmath234 and @xmath235 . then @xmath236 . let @xmath55 be an unweighted vertex of a simple vertex - weighted graph @xmath188 , and let @xmath189 and @xmath190 be the full subgraphs of @xmath228 induced by the neighbors ( resp . non - neighbors ) of @xmath55 . any weighted interlace polynomial @xmath237 may be analyzed in the following way . eliminate all edges between vertices of @xmath228 using pivots and weight - changes as in corollary [ cor1 ] . ( in general there will be many different sequences of pivots that may be used ; this lack of uniqueness is not important in the analysis . ) the assignments of individual vertices of @xmath188 to @xmath189 and @xmath190 may change during this process , and vertex weights may also change ; but these reassignments and weight changes will be the same for all graphs @xmath8 . as in the proof of proposition 39 of @xcite these pivots will not affect the internal structure of @xmath8 , because no two vertices of @xmath8 have distinct , nonempty sets of neighbors in @xmath188 . here the phrase `` internal structure '' refers to vertex weights , the positions of loops , and adjacencies , including adjacencies between @xmath55 and other vertices of @xmath8 . when step 1 is complete , @xmath237 is expressed as a sum in which each summand is the product of an initial multiplying factor and the weighted interlace polynomial @xmath238 of a graph in which every edge is incident on a vertex of @xmath239 . if @xmath240 then @xmath241 is isolated , so the only effect of @xmath241 is to multiply that summand by @xmath242 . this same effect is realized by removing @xmath241 and incorporating @xmath243 into the initial multiplying factor , so we may assume that every summand has @xmath244 . in a summand with @xmath245 the vertices of @xmath246 are all identical twins , and may be consolidated into a single vertex @xmath55 using theorem [ theorem3 ] . in a summand with @xmath247 a single vertex @xmath248 may be introduced with @xmath249 and @xmath250 ; this will not affect the value of the corresponding summand . as in step 1 , these manipulations are the same for all @xmath8 . step 3 . the weighted interlace polynomial @xmath237 is now expressed as a sum in which each summand is the product of an initial multiplying factor and a weighted interlace polynomial @xmath251 in which @xmath252 differs from @xmath8 only in the weights of @xmath55 . in each summand we multiply the @xmath17 and @xmath18 weights of @xmath55 by the initial multiplying factor . this has the effect of multiplying @xmath251 by that factor , so the summand is now simply @xmath251 . proposition [ prop2 ] tells us that the sum is equal to a single weighted interlace polynomial @xmath251 , where @xmath49 and @xmath136 are obtained by adding together the @xmath17 and @xmath18 weights of @xmath55 in the various summands . we deduce that there are weights @xmath49 and @xmath136 that depend only on @xmath188 and @xmath55 , and have the following property : in every instance of definition [ comp ] involving @xmath188 , the interlace polynomial @xmath237 equals @xmath251 , where @xmath252 is obtained from @xmath8 by using @xmath49 and @xmath136 as weights for @xmath55 . finding explicit formulas for these weights is not difficult . [ theorem8]let @xmath188 be a vertex - weighted simple graph with an unweighted vertex @xmath55 . then every composition @xmath227 has @xmath253 , where @xmath252 is obtained from @xmath8 by using the following weights for @xmath55 . @xmath254 with @xmath255 , we have @xmath256 = @xmath237 = @xmath251 = @xmath257 . with @xmath8 consisting of two adjacent , unlooped , unweighted vertices @xmath55 and @xmath241 we have @xmath258 = @xmath237 = @xmath259 = @xmath260 . the stated formulas for @xmath49 and @xmath136 follow . in case @xmath261 might be a zero divisor in the ring @xmath20 , one can avoid any difficulty with the formulas of theorem [ theorem8 ] by first evaluating them in the polynomial ring @xmath262 $ ] , and then evaluating the resulting division - free formulas in @xmath20 . theorem [ theorem8 ] concerns compositions @xmath227 in which @xmath188 is simple . if @xmath188 has looped vertices , a similar analysis requires two new steps . begin by removing all loops in @xmath228 using local complementation as in part ( a ) of theorem [ theorem1 ] . the result is a description of @xmath263 as a sum in which each summand is the product of an initial factor and an interlace polynomial @xmath238 or @xmath264 , with no loops in @xmath265 . step 4 . after applying steps 1 - 3 to each of these summands , collect terms to obtain a formula @xmath266 in order to distinguish the two terms on the right - hand side we denote by @xmath267 the copy of @xmath55 in @xmath268 . it might seem that we now have to determine four unknowns , namely the vertex weights @xmath49 , @xmath269 , @xmath136 and @xmath270 it turns out though that these unknowns are not independent . there is an obvious isomorphism between @xmath271 and @xmath268 , and consequently @xmath272 may be obtained from @xmath273 simply by replacing @xmath274 with @xmath49 . theorem [ theorem2 ] tells us that @xmath275 so @xmath276 may be obtained from @xmath273 by replacing @xmath274 with @xmath49 . that is , the coefficient of @xmath49 in @xmath251 is precisely the same as the coefficient of @xmath274 in @xmath277 . it follows that the sum @xmath251 + @xmath277 is unchanged if we replace @xmath49 by @xmath49 + @xmath274 and replace @xmath274 by @xmath278 . [ theorem9]let @xmath188 be a vertex - weighted graph with an unweighted , unlooped vertex @xmath55 . then @xmath188 and @xmath55 determine weights @xmath49 , @xmath136 and @xmath270 such that every composition @xmath227 has @xmath279 , where @xmath252 is obtained from @xmath8 by using @xmath49 and @xmath136 as weights for @xmath55 and @xmath280 is obtained from @xmath268 by using @xmath281 and @xmath270 as weights for @xmath267 , the copy of @xmath55 in @xmath280 . formulas for the three weights mentioned in theorem [ theorem9 ] may be derived from three instances of the theorem . we use @xmath228 and @xmath188 as in theorem [ theorem8 ] , and also @xmath282 , the graph obtained from @xmath188 by attaching a loop at @xmath55 . these correspond respectively to compositions of @xmath188 with graphs @xmath283 such that @xmath284 ; @xmath285 with @xmath241 an unweighted , unlooped neighbor of @xmath55 ; and @xmath286 with @xmath287 an unweighted , looped neighbor of @xmath55 . definition [ 2varq ] gives the following values . @xmath288 @xmath289 @xmath290 we deduce these formulas . @xmath291 @xmath292 @xmath293 separate formulas for @xmath136 and @xmath270 are derived in the obvious ways by adding and subtracting the last two . as before , possible problems with denominators may be avoided by first evaluating the formulas in @xmath294 $ ] . we close this section with different formulas for the weights @xmath49 , @xmath136 and @xmath270 that appear in theorems [ theorem8 ] and [ theorem9 ] . suppose @xmath188 has an unweighted , unlooped vertex @xmath55 . let @xmath295 denote the open neighborhood of @xmath55 , i.e. , the set containing the vertices @xmath296 that neighbor @xmath55 . given a subset @xmath297 let @xmath298 ( resp . @xmath299 ) be the row ( resp . column ) vector with entries indexed by @xmath300 whose @xmath301 entry is 1 or 0 according to whether @xmath302 or @xmath303 . also let @xmath304 be the adjacency matrix of @xmath305 $ ] . note that @xmath306 and @xmath307{cc}% m & \kappa\\ \rho & 0 \end{array } \right ) , \;\;r\left ( \begin{array } [ c]{cc}% m & \kappa\\ \rho & 1 \end{array } \right ) \leq r(m)+2\ ] ] because adjoining a single row or column to a matrix raises the rank by 0 or 1 . [ def9]the _ type _ of @xmath16 with respect to @xmath55 is defined as follows . @xmath16 is of type 1 if @xmath308{cc}% m & \kappa\\ \rho & 0 \end{array } \right ) = r\left ( \begin{array } [ c]{cc}% m & \kappa\\ \rho & 1 \end{array } \right ) -1.$ ] @xmath16 is of type 2 if @xmath309{cc}% m & \kappa\\ \rho & 0 \end{array } \right ) = r\left ( \begin{array } [ c]{cc}% m & \kappa\\ \rho & 1 \end{array } \right ) .$ ] @xmath16 is of type 3 if @xmath308{cc}% m & \kappa\\ \rho & 1 \end{array } \right ) = r\left ( \begin{array } [ c]{cc}% m & \kappa\\ \rho & 0 \end{array } \right ) -1.$ ] every @xmath297 is of one of these types . moreover , if @xmath55 has no looped neighbor then there is no @xmath297 of type 3 . the fact that every @xmath297 is of type 1 , 2 or 3 appears in lemma 2 of @xcite . suppose @xmath16 is of type 3 , so @xmath310{cc}% m & \kappa\\ \rho & 0 \end{array } \right ) .\ ] ] the row vector @xmath311 can not be a sum of rows of @xmath312 . for if it were then @xmath313 would be the corresponding sum of rows of @xmath165 , and by symmetry @xmath314 would be the corresponding sum of columns of @xmath165 . consequently it would follow that @xmath315{cc}% m & \kappa \end{array } \right ) = r\left ( \begin{array } [ c]{cc}% m & \kappa\\ \rho & 0 \end{array } \right ) .\ ] ] on the other hand , @xmath313 must be the sum of the rows of @xmath165 corresponding to the elements of some subset @xmath316 , for if it were not then it would follow that @xmath310{c}% m\\ \rho \end{array } \right ) = r\left ( \begin{array } [ c]{cc}% m & \kappa \end{array } \right ) = r\left ( \begin{array } [ c]{cc}% m & \kappa\\ \rho & 0 \end{array } \right ) -1.\ ] ] every such @xmath317 must contain an odd number of neighbors of @xmath55 , to avoid giving a sum of rows of @xmath312 equal to @xmath311 . choose such a @xmath317 , and consider the induced subgraph @xmath318 $ ] . as the sum of the rows of @xmath165 corresponding to elements of @xmath317 is @xmath313 , the sum of the rows of the adjacency matrix of @xmath318 $ ] is the row vector @xmath319 . @xmath320 is odd , so the adjacency matrix of @xmath318 $ ] has an odd number of entries equal to 1 . the matrix is symmetric , so at least one of these entries must appear on the diagonal . that is , at least one vertex of @xmath318 $ ] is looped . for @xmath297 let @xmath321 be the matrix obtained from @xmath165 by toggling every entry @xmath322 that has @xmath323 . then @xmath324 = @xmath325 because the first matrix is transformed into the second by adding the last column to every column corresponding to a neighbor of @xmath55 . similarly , adding the last column of the first matrix below to every column corresponding to a neighbor of @xmath55 tells us that @xmath326{cc}% m & \kappa\\ \rho & 1 \end{array } \right ) = r\left ( \begin{array } [ c]{cc}% m^{a } & \kappa\\ 0 & 1 \end{array } \right ) = 1+r(m^{a}).\ ] ] consequently , definition [ def9 ] may be restated using the relationship between @xmath327 and @xmath328 : if @xmath16 is of type 1 then @xmath329 , if @xmath16 is of type 2 then @xmath330 , and if @xmath16 is of type 3 then @xmath331 . [ prop3]for @xmath332 let @xmath333)}(y-1)^{n((h - a)[s])}.\ ] ] then the following equations hold . @xmath334 also , @xmath335 the first equality is obvious . for the second , note that each @xmath336 gives rise to two subsets of @xmath337 , namely @xmath16 and @xmath338 ; these correspond to the adjacency matrices @xmath165 and @xmath339{cc}% m & \kappa\\ \rho & 0 \end{array } \right ) $ ] . similarly , the third equality is derived by considering the contributions of two adjacency matrices for each @xmath297 , namely @xmath165 and @xmath339{cc}% m & \kappa\\ \rho & 1 \end{array } \right ) $ ] . the last equality follows from the discussion preceding the proposition . the weights mentioned in theorems [ theorem8 ] and [ theorem9 ] are @xmath340 , @xmath341 and @xmath342 . in particular , @xmath343 if @xmath55 has no looped neighbor . the corollary follows from proposition [ prop3 ] and the formulas given in theorem [ theorem8 ] and immediately after theorem [ theorem9 ] . in this section we focus our attention on the unweighted vertex - nullity polynomial . [ prop4]if @xmath2 is a connected , unweighted graph with at least one looped vertex then @xmath344 . the proposition is certainly true if @xmath2 has @xmath208 vertices , as all three such graphs have @xmath345 . the argument proceeds by induction on @xmath346 . recall that @xmath184 for every graph @xmath2 , and let @xmath55 be a looped vertex of @xmath2 . then @xmath347 , so @xmath348 , @xmath349 . suppose @xmath55 is not a cutpoint of @xmath2 . if @xmath2 has some looped vertex other than @xmath55 then the inductive hypothesis implies that @xmath350 . if @xmath2 has no looped vertex other than @xmath55 then every neighbor of @xmath55 is looped in @xmath351 . every component of @xmath351 contains at least one neighbor of @xmath55 , so the inductive hypothesis implies that @xmath352 for every component @xmath353 of @xmath351 . hence @xmath354 . suppose now that @xmath2 has a looped cutpoint @xmath55 . for each component @xmath353 of @xmath25 and each vertex @xmath355 that lies in some other component of @xmath25 , no edge connecting @xmath241 to an element of @xmath356 appears in @xmath25 . consequently every such edge appears in @xmath351 , so all of @xmath357 is contained in a single component of @xmath351 . as @xmath2 is connected , this implies that @xmath351 is also connected . hence if any neighbor of @xmath55 is unlooped in @xmath2 , @xmath351 is a connected graph with a looped vertex and the inductive hypothesis implies that @xmath358 . if instead every neighbor of @xmath55 is looped in @xmath2 , then every component @xmath353 of @xmath25 has a looped vertex and the inductive hypothesis implies @xmath359 . [ cor4]an unweighted graph @xmath2 has @xmath360 if and only if @xmath207 has no nonempty simple component . as noted in remark 20 of @xcite , every nonempty simple graph has @xmath185 . it follows that every graph with a nonempty simple component also has @xmath185 . on the other hand , proposition [ prop4 ] and the fact that @xmath361 together imply that every graph with no nonempty simple component has @xmath360 . let @xmath2 be an unweighted graph , and let @xmath362 be its _ complement _ , i.e. , the graph with @xmath363 whose edges ( including loops ) are precisely the edges absent in @xmath2 . then the following are equivalent : @xmath2 is simple , @xmath364 and @xmath365 . suppose first that @xmath2 is simple , and let @xmath366 with @xmath55 the vertex of @xmath367 . let @xmath8 be the two - vertex graph with @xmath368 in which @xmath287 is an unweighted , looped neighbor of @xmath55 . theorem [ theorem8 ] tells us that @xmath369 on the other hand , the recursive description of @xmath9 @xcite tells us that @xmath370 recalling that @xmath371 = @xmath257 , we see that @xmath372 and consequently @xmath373 . recalling that @xmath374 = @xmath258 = @xmath260 , we see that latexmath:[\[q_{n}(g+e_{1}^{u})=q(g+e_{1}^{u})|_{x=2}=y\left ( \alpha(a)+\beta(a)\right ) now suppose that @xmath2 is not simple . then @xmath2 has a looped vertex , and so does @xmath376 . as @xmath376 is connected , corollary [ cor4 ] states that @xmath377 ; consequently @xmath378 . a combinatorial description of the interlace polynomials of trees and forests is given in @xcite . in order to motivate the next section we sketch this description briefly here , omitting details and proofs . recall that a tree @xmath317 is _ rooted _ by specifying a root vertex @xmath379 . each non - root vertex @xmath380 then has a unique _ parent _ @xmath381 , a neighbor whose distance from @xmath382 is less than the distance from @xmath382 to @xmath241 . the elements of @xmath383 are the _ children _ of @xmath381 , and the children of @xmath381 other than @xmath241 itself are _ siblings _ of @xmath241 . an _ ordered _ tree is a rooted tree given with an order on the set of children of each parent vertex ; non - root vertices may then have _ earlier siblings _ and _ later siblings_. a set of vertices that contains no adjacent pair is _ independent _ , and a set of vertices _ dominates _ a vertex @xmath241 if it contains @xmath241 or contains some neighbor of @xmath241 . an _ earlier sibling cover _ ( or _ es - cover _ ) in an ordered tree @xmath317 is an independent set @xmath384 that dominates @xmath382 and has the property that for every non - root vertex @xmath385 , every earlier sibling of @xmath241 is dominated by @xmath384 . for integers @xmath386 and @xmath387 the _ es - number _ @xmath388 is the number of @xmath386-element es - covers in @xmath317 whose non - root elements have @xmath387 different parents . if @xmath389 is a subtree of @xmath317 and @xmath390 then we presume that the children of each parent vertex @xmath241 in @xmath389 are ordered by restricting the order of the children of @xmath241 in @xmath317 . with this convention , it is easy to verify that the earlier sibling covers in large trees arise from earlier sibling covers in subtrees . [ recur ] let @xmath317 be an ordered tree with a leaf @xmath287 such that @xmath391 , all the siblings of @xmath287 are leaves , and @xmath287 has no later siblings . then @xmath392 and @xmath393 lemma [ recur ] is the key to an inductive proof that the terms in the definition of @xmath394 can be collected into sub - totals corresponding to earlier sibling covers . let @xmath317 be an ordered tree with vertex weights , and let @xmath384 be an es - cover in @xmath317 . let @xmath395 for each vertex @xmath241 define the @xmath384__-weight _ _ @xmath396 as follows : @xmath397 the product @xmath398 is the _ total weight _ of @xmath384 in @xmath317 , denoted @xmath399 . [ treepoly ] if @xmath317 is an ordered tree with vertex weights then @xmath400 if @xmath317 has no more than one vertex of degree @xmath401 , it is not difficult to verify the theorem directly . if @xmath317 has more than one vertex of degree @xmath401 then the theorem follows inductively from theorem [ theorem1 ] and corollary [ cor1 ] using lemma [ recur ] , as detailed in @xcite . setting @xmath48 and @xmath47 we deduce that the unweighted interlace polynomial of a tree is determined by a simple formula involving earlier sibling covers . [ treeint]if @xmath317 is a tree then the unweighted interlace polynomial of @xmath317 is @xmath402 as the interlace polynomials are multiplicative on disjoint unions , these results extend directly to disconnected forests . the tutte polynomial is a very useful invariant of graphs and matroids , which incorporates a great deal of information and can be defined in several different ways ; see @xcite , @xcite and @xcite for more detailed discussions than we provide here . one of its definitions is given in definitions [ oldact ] and [ act ] . [ oldact ] suppose @xmath2 is an unweighted graph with @xmath403 , and @xmath317 is a maximal spanning forest of @xmath2 . an element @xmath404 is _ externally active _ with respect to @xmath317 if @xmath405 is the least index of an element of the unique circuit contained in @xmath406 . an element @xmath407 is _ internally active _ with respect to @xmath317 if @xmath405 is the least index of an element of the unique cutset contained in @xmath408 . the numbers of edges that are externally and internally active with respect to @xmath317 are denoted @xmath409 and @xmath410 , respectively . theorem [ treepoly ] and corollary [ treeint ] bear a strong resemblance to this formula , with earlier sibling covers in rooted trees replacing maximal spanning forests in arbitrary graphs . we do not know whether there is a combinatorial analogue of definition [ act ] that describes the interlace polynomials of an arbitrary graph , but there is an algorithmic analogue . before presenting it , we recall another definition of the tutte polynomial . to recursively calculate @xmath423 one simply chooses an arbitrary edge of @xmath2 , and applies the appropriate part of definition [ delcont ] ; this process is repeated as many times as necessary . such a computation is represented by a computation tree in which a node that represents an instance of part ( a ) has two children and a node that represents an instance of part ( b ) or ( c ) has only one . [ actact ] let @xmath2 be a graph with @xmath411 , and consider the recursive implementation of definition [ delcont ] in which @xmath424 is removed first , then @xmath425 is removed in all branches , then @xmath426 is removed in all branches , and so on . the leaves of the computation tree representing this implementation correspond to the maximal spanning forests of @xmath2 , with the leaf corresponding to @xmath317 resulting from the portion of the computation in which edges of @xmath317 are contracted and elements of @xmath427 are deleted . a node of this portion of the computation tree represents the removal of an active edge if and only if it has precisely one child . the proposition is implicit in the fact that definition [ delcont ] and definition [ act ] both yield @xmath423 , so it appears implicitly in just about every presentation of the tutte polynomial . see theorem ix . 65 of @xcite or theorem x.10 of @xcite , for instance . explicit discussions of the connection between activities and computation are less common in the literature , though there are some @xcite . the proof is a direct induction on @xmath428 . if @xmath424 is a loop then the maximal spanning forests of @xmath2 and @xmath429 coincide , and the computation tree for @xmath423 is obtained from the computation tree for @xmath430 by attaching a new root node of degree 1 , representing the removal of @xmath424 . if @xmath424 is an isthmus then the maximal spanning forests of @xmath2 and @xmath431 correspond , and the computation tree for @xmath423 is obtained from the computation tree for @xmath432 by attaching a new root node of degree 1 . otherwise , the maximal spanning forests of @xmath2 that contain @xmath424 correspond to the maximal spanning forests of @xmath431 , and the maximal spanning forests of @xmath2 that do not contain @xmath424 are the maximal spanning forests of @xmath429 . the computation tree for @xmath423 consists of the root and two disjoint subtrees that are the computation trees for @xmath432 and @xmath430 . definition [ oldact ] defines active edges using the structure of @xmath2 , and this leads to definition [ act ] s description of @xmath423 as a generating function for maximal spanning forests . proposition [ actact ] shows that we may also see activity from an algorithmic viewpoint : the external activity of a particular @xmath433 is revealed in the fact that one step of a calculation of @xmath423 involves removing @xmath434 using part ( b ) of definition [ delcont ] rather than part ( a ) . this distinction affects the result of the computation , so it would be important even if activity could not be conveniently described using the structure of @xmath2 , or did not contribute to a convenient closed form for @xmath423 . [ algact ] a node of a computation tree representing a recursive implementation of definition [ alg ] is _ active _ if it has precisely one child , i.e. , if it represents an application of part ( c ) or part ( d ) . an activities formula for @xmath1 arises directly from a computation tree representing an implementation of definition [ alg ] . the formula has one summand for each leaf of the computation tree ( each call to part ( e ) of definition [ alg ] ) , representing the product of the coefficients contributed by the nodes in the portion of the computation tree that gives rise to that leaf . active and non - active nodes contribute different coefficients . let @xmath317 be a rooted tree with root @xmath382 , and consider a recursive implementation of definition [ alg ] structured as follows . if possible , apply part ( d ) of definition [ alg ] ; if not and there is a parent vertex other than @xmath382 then apply part ( b ) with a leaf of the type denoted @xmath287 in lemma [ recur ] as @xmath55 ; otherwise apply part ( b ) with the last child of @xmath382 as @xmath55 . the leaves of the computation tree representing this implementation correspond to the earlier sibling covers of @xmath317 , with the es - cover @xmath384 corresponding to a given leaf constructed from the portion of the computation that gives rise to that leaf as follows : an occurrence of part ( d ) of definition [ alg ] contributes its @xmath55 to @xmath384 , and an occurrence of the @xmath441 branch of part ( b ) contributes its @xmath55 to @xmath384 . if @xmath317 has no vertex other than @xmath382 then @xmath442 is the only es - cover in @xmath317 , and the computation consists simply of a single call to part ( e ) of definition [ alg ] . if @xmath317 contains no parent vertex other than @xmath382 and @xmath55 is the last child of @xmath382 , then the es - covers in @xmath317 include @xmath443 and the es - covers of @xmath444 . the first step of the computation is an application of part ( b ) of definition [ alg ] with @xmath445 . the inductive hypothesis applies to @xmath444 , and the branch of the computation corresponding to @xmath446 consists solely of calls to part ( d ) because every vertex of @xmath447 is isolated ; consequently the latter part of the computation tree contains only one leaf , corresponding to @xmath443 . if @xmath317 contains a parent vertex other than @xmath382 , then the first step of the computation is an application of part ( b ) of definition [ alg ] with @xmath448 as in lemma [ recur ] , and @xmath449 . the computation tree contains a single node representing this first step and also two subtrees , one corresponding to @xmath450 and the other corresponding to @xmath451 . the proposition follows inductively from lemma [ recur ] . we do not know whether or not it is possible to reformulate definition [ algact ] so that it always refers to @xmath2 instead of a computation tree . such a reformulation would certainly be valuable , as the resulting activities formulas would shed light on the combinatorial significance of the interlace polynomials . definition [ algact ] extends directly to computation trees representing implementations of other recursions . for instance , if definition [ alg ] is augmented by incorporating pendant - twin reductions then the resulting computation trees will have active nodes representing these reductions , in addition to active nodes representing parts ( c ) and ( d ) of definition [ alg ] . arratia , r. , bollobs , b. , and sorkin , g. b. ( 2000 ) the interlace polynomial : a new graph polynomial . in _ proceedings of the eleventh annual acm - siam symposium on discrete algorithms _ ( san francisco , 2000 ) , association for computing machinery , new york , ps . 237 - 245 . blser , m. and hoffmann , c. ( 2008 ) on the complexity of the interlace polynomial , in : _ stacs 2008 : 25th international symposium on theoretical aspects of computer science _ ( bordeaux , 2008 ) , ps . 97 - 108 , available at http://www.stacs - conf.org/.

we develop various complementary concepts and techniques for handling quantum fluctuations of goldstone bosons.we emphasise that one of the consequences of the masslessness of goldstone bosons is that the longitudinal fluctuations also have a diverging susceptibility characterised by an anomalous dimension in space - time dimensions.in these fluctuations diverge logarithmically in the infrared region.we show the generality of this phenomenon on the basis of i ) . renormalization group flows , ii ) . ward identities , and iii ) . schwinger - dyson equations.we also obtain an explicit form for the generating functional of one - particle irreducible vertices of the o(n ) ( non)linearmodels in the leading approximation.we show that this incorporates all infrared behaviour correctly both in linear and non - linear models . these techniques provide an alternative to chiral perturbation theory.some consequences are discussed briefly . * infrared behaviour of systems with goldstone bosons * + ramesh anishetty , rahul basu , n.d . hari dass and h.s.sharatchandra + _ the institute of mathematical sciences , madras-600 113 , india _

hep-th9502003

the concept of spontaneous breaking of a continuous global symmetry with the attendant goldstone phenomenon of massless scalar excitations has deeply influenced many developments in field theory and statistical physics . the existence of massless modes can already be demonstrated in landau - ginzberg mean field analysis ( or tree level , in the field theory jargon ) . however , it is not sufficient to stop at this level . precisely because they are massless , the quantum / statistical fluctuations of the goldstone bosons may significantly alter the conclusions drawn from mean field theory.proofs of the goldstone theorem @xcite in quantum field theory have traditionally been used only to establish the existence of massless poles coupling to the currents of broken symmetry . they have not usually focussed on the correlation functions . most of the applications in quantum field theory have been based on the understanding at tree level . notable exceptions are the soft pion theorems @xcite combining pcac with current algebra.these are mostly applied to extract the leading infrared behaviour . a general analysis of these issues based essentially only on symmetry arguments was given some time ago by weinberg @xcite . in that article he also proposed methods to go beyond the leading order results . he also applied renormalization group methods to understand some universal features of these results . quite recently , some of weinberg s ideas have been extensively used in the context of the so called `` chiral perturbation theory''@xcite . in contrast , in statistical physics , powerful techniques have been developed for handling the fluctuations in this context . these are of importance for quantum field theory also.some examples are : i)the finite temperature chiral symmetry restoring transition in qcd , where goldstone phenomenon in three eucilidean dimensions is of relevance @xcite ; ii)qcd at temperatures below the chiral transition temperature as well as zero temperature qcd , where goldstone phenomenon in 4 euclidean dimensions is of relevance , and iii)the problems that are handled by chiral perturbation theory at present , i.e. effects of pion loops on various processes @xcite . in this paper we point out various complimentary concepts and techniques for handling the effects due to quantum fluctuations of goldstone bosons . we develop techniques for computing correlation functions away from the infrared regime.we show that there are certain universal features , valid even in four space - time dimensions.our approach provides an alternate procedure for chiral perturbation theory which is closer in its spirit to what has been proposed by weinberg on the basis of rg equations . this aspect is elaborated upon in section 7 . often the paradigm in quantum field theory is what is seen from a tree level analysis of the o(n ) linear and non - linear @xmath3-models . the linear @xmath3-model is given by the action @xcite , @xmath5 @xmath6 transforms as the fundamental representation of o(n ) . if @xmath7 , the minimum of the classical potential is at @xmath8 . therefore the vev @xmath9 is no longer zero in a tree level analysis.choosing the vacuum @xmath10 , among the equivalent vacua , we get , @xmath11 where @xmath12 . therefore at tree level , we have @xmath13 massless pions and a massive @xmath3 with a mass @xmath14 . precisely because the goldstone bosons are massless , loop corrections may drastically alter the above picture . now it is well known that infrared divergences due to massless fluctuations drastically alter the conclusions of a semi - classical analysis , especially if the spacetime dimension d is less than four @xcite.already , way back in 1940 , holstein and primakoff @xcite argued that not only the transverse mode ( i.e. @xmath15 ) but also the longitudinal mode ( @xmath3 ) is soft in a quantum ferromagnet in three dimensions . their arguments can be understood , as follows , in terms of the non - linear @xmath3 -model . the non - linear @xmath3- model is relevant to the infrared dynamics of goldstone bosons @xcite .a simple heuristic argument is that if @xmath3 is massive , then for the study of the infrared dynamics of the pions , we may use an effective lagrangian obtained by integrating over the heavy @xmath3 degree of freedom . also for leading infrared behaviour we may retain only the lowest ( two ) derivative terms in the effective action.the result is simply the heisenberg model , @xmath16 with the constraint , @xmath17 . statistical physics in d space dimensions is simply the euclidean quantum field theory in @xmath18 space - time dimensions . in statistical physics @xmath19 is proportional to the temperature , whereas in quantum field theory it is related to the pion decay constant , @xmath20(in what follows , we set @xmath21 ) . if the spins are ordered in the direction of @xmath22 , then eliminating this field using the constraint , the action can be written entirely in terms of @xmath23 . they describe the spin waves or the transverse fluctuations.at the tree level for the @xmath15 propagator we get , @xmath24 we may also consider the correlations for longitudinal fluctuations i.e. of the field @xmath25 ( : refers to any ordering prescription that makes products of fields at the same point meaningful ) . this means , @xmath26 . we may expect the leading infrared behaviour to come from the first term . this would mean that so far as the leading infrared behaviour is concerned , @xcite , @xcite @xmath27 . if we ignore self - interactions of @xmath28 , and use the above propagator for @xmath28 we get , for the @xmath3 propagator , @xmath29 this gives , atleast for @xmath30,in momentum space , @xmath31 ignoring the self interactions of @xmath15 s will later be seen to be justified because these are soft in the infrared region a consequence of the soft pion theorems.the result is that in @xmath32 the @xmath3propagator is @xmath33 , to be contrasted with the @xmath15-propagator.this means that @xmath3 also has long range correlations . this appears to negate the naive deduction of the non - linear @xmath3 model obtained by integrating over the heavy @xmath3 field in the linear @xmath3 model @xcite . one could dismiss these results as artifacts of a 1-loop calculation . in fact the infrared divergences grow with the loop order . therefore it is imperative to tackle this issue in a framework that transcends loop expansion.there is an exactly solvable model , the berlin - kac model @xcite which goes beyond loop - expansion and again gives precisely the same diverging longitudinal susceptibility and also a massless @xmath3 in @xmath30 . again it is not clear whether this is an artifact of the model . in order to get a handle on the infrared dynamics of the @xmath3 field , pokrovsky and patashinky @xcite proposed a `` conserved modulus principle '' ( cmp).this principle holds that the fluctuations are such as to maintain the modulus of the n- component order parameter . in mathematical terms cmp states that + @xmath34 now the expectation value of @xmath35 is given by the ornstein- zernike form : @xmath36 it is convenient to use susceptibility to describe the infrared behaviour.the longitudinal and the transverse susceptibility in presence of an external magnetic field @xmath37 are given by , @xmath38 using eqns . [ pp1 ] and [ pp2 ] , we get , for @xmath30 , @xmath39 thus the longitudinal suceptibility also diverges . patashinsky and pokrovski have offered a proof for cmp in @xcite . they express the change in the thermodynamic potential u computed to quartic order in the fluctuations as + @xmath40\ ] ] and interpret this as implying cmp . though to quartic order the correct form of this variation , which differs from the one given above ( and which can be verified using the ward identities discussed later ) , is @xmath41 + { 1\over 2 \chi_{l}}[{(\delta \vec\phi^2)^2\over 4 \bar \phi^2}]\ ] ] the fact that cmp is equivalent to the correct statement that the extreme ir behaviour is governed by the non - linear @xmath3-model indicates the possibility of a better proof of cmp . + the new era for understanding the role of fluctuations of goldstone bosons started with the renormalization group techniques . brezin , wallace and wilson @xcite computed the equation of state for o(n ) @xmath3model using the @xmath42expansion.they found important differences from the corresponding calculation for the n=1 case ( i.e. the ising model).this could be traced to the infrared singularities due to the goldstone bosons that are present everywhere in the coexistence region for @xmath43 . in particular it meant that the conclusions of the tree level analysis could not be right ( atleast for @xmath30 ) . at the critical point ( @xmath44 ) , @xmath3 and @xmath15 are on the same footing . both their susceptibilities diverge with vanishing magnetic field as , @xmath45 governed by the critical exponent @xmath46 . the question of interest is the behaviour for @xmath47 as @xmath48.the assumption that the singularities due to goldstone bosons exponentiate in the @xmath42expansion suggested that , @xmath49 holstein -primakoff @xcite , berlin - kac @xcite , and patashinski - pokrovski @xcite models have all made the same prediction . such a result has also been proved rigorously for o(2 ) case using correlation inequalities in ref . @xcite . a deeper understanding can be obtained by considering the renormalization group flows shown in the figure below : at the critical point , @xmath44 , the infrared dynamics is governed by the o(n ) non - gaussian fixed point @xmath50 when @xmath1.everywhere on the critical surface the renormalization group ( rg ) trajectories flow from the the o(n ) gaussian fixed point @xmath51 ( which is infrared unstable ) to the o(n ) non - gaussian fixed point @xmath50 . for @xmath52 we have the symmetric phase with massive ( and degenerate ) @xmath3 and @xmath53.the infrared behaviour is governed by the trivial fixed point @xmath54 which corresponds to zero correlation length.@xmath47 presents a new feature compared to n=1 case ( i.e.the ising model ) . the theory has long range correlations for all @xmath55 as a consequence of the goldstone phenomenon.hence the infrared behaviour must be determined by a fixed point of infinite correlation length.what is this fixed point ? @xmath56 , the deviation from the critical temperature in statistical mechanics , is to be identified with the square of the renormalized mass @xcite in the linear @xmath3-model . in particular they are parameters with positive mass dimensionality . under rg flow they necessarily scale.thus , once @xmath57 , the flow is towards @xmath58.this means that we are interested in the zero temperature fixed point . an indication as to the nature of this fixed point came from a study of @xmath59expansion of o(n ) non linear @xmath3model by brezin and zinn - justin @xcite . the o(n ) non - linear @xmath3- model in d=2 is renormalizable . even though the renormalized perturbation is in terms of the spin wave modes , mermin - wagner theorem @xcite requires that the system is never in a phase with spontaneously broken o(n ) symmetry . a hint of this is contained in the renormalized perturbation theory : the theory is asymptotically free.the coupling constant grows with the distance so that the long distance structure of the theory could be drastically different from the perturbation theory spectrum.indeed the spectrum is believed be that of massive and degenerate @xmath3 and @xmath15 s in the fundamental representation of o(n ) . a consequence of the asymptotic freedom in @xmath60 is that in @xmath61 , the @xmath62-function @xmath63 has a non - trivial zero ( say at @xmath64 ) in addition to the trivial zero , @xmath65 , corresponding to the theory of ( n-1 ) free scalars.at @xmath64 the theory is scale invariant and corresponds to the critical theory with n massless scalars and with anomalous dimensions ( denoted by @xmath66 ) .it should be noted in fig . 1 that the fixed point @xmath50 of the linear @xmath67-model is identified with the fixed point @xmath66 of the non - linear model . this follows essentially from a remark made in @xcite that at @xmath50 the correlation functions of the linear model are identical to those of the non - linear model where in the latter one includes correlation functions of composite operators also . the coincidence of these two fixed points can also be established in the @xmath4-expansion scheme @xcite . for @xmath68 , the coupling constant grows with the distance , and as in @xmath60 , the theory is in the unbroken phase . @xmath69 is of interest to us.in this case the infrared flow is towards @xmath65 . the infrared behaviour is of o(n-1 ) gaussian theory ( this fixed point is denoted by @xmath70 in fig.1 . ) and the ultraviolet behaviour is of the critical theory with anomalous dimensions . the fact that the infrared behaviour for @xmath69 is governed by the o(n-1 ) gaussian fixed point gives exhaustive information on the effects of quantum fluctuations of the goldstone bosons.it means that the pions decouple from each other at low momenta , and their propagators behave as the free propagators for small momenta.it also allows us to obtain the non - leading ir - behaviour as well as the effects of an external magnetic field.in particular we now understand why various heuristic considerations all gave the correct answer : it was the fortuitious behaviour of a free massless theory.anything more complicated would have been much more difficult to handle . upto what dimension d is the analysis of @xmath61 calculation correct?in particular is it applicable to d=4?the case d=4 is somewhat different because of the absence of the non - gaussian fixed point @xmath50.at the critical temperature the infrared behaviour is believed to be given by free massless theory of n scalars as the beta function is positive for small couplings ; there is no evidence for a non - gaussian fixed point.this appears to exclude the validity of the @xmath59 analysis to this case.we need an alternate technique to handle this situation.this is provided , for example , by the @xmath4 expansion . the @xmath4 expansion provides results which agree with the above analysis for all d.(the berlin - kac model @xcite is a variant of this technique . ) it yields the non - gaussian fixed point for @xmath1 and shows its absence for @xmath2.in either case it shows that the leading infrared behaviour anywhere in the broken phase is given by n-1 free pions.the only handicap of this technique is that it does not provide non - anomalous dimension for the non - gaussian fixed point @xmath50 in the leading n approximation.however even this is overcome once one takes higher order ( in @xmath4 ) corrections . there are also indications that the technique gives qualitatively correct and quantitatively reasonable results for all @xmath71 . the fact that even in d=4 the pion dynamics is governed by the o(n-1 ) gaussian fixed point @xmath70 in the infrared has important implications.in particular the propagator of @xmath3 ( to be interpreted as the o(n ) partner of the pions ) has an universal logarithmic singularity in the infrared region , @xmath72 where @xmath73 is the renormalisation scale . this follows from the same arguments as for @xmath30 , in particular because @xmath27 in the infrared.this feature is present whether @xmath3-propagator has a pole at some @xmath74 or not . the precise location of the pole is a non - universal feature depending on the microscopic dynamics . it could have interesting implications for nuclear forces at low energies in the sigma-channel ( i.e. , spin j=0,isospin i=0 ) as will be discussed elsewhere . in sec . 3 and 4 we develop a new formalism for @xmath75 which will allow for explicit calculation of all correlation functions even away from the infrared regime.we will explicitly demonstrate all the properties mentioned above.our formalism gives the same results in the leading infrared limit whether one starts with the linear or the non - linear @xmath3- model . that the pions behave like free massless bosons in the infrared limit , in any @xmath76 , is such a general phenomenon that there should be an easier way of understanding it . a careful analysis of the proof of the goldstone theorem along with current algebra and pcac provide one such non - perturbative argument.another is the treatment based on ward identities given in sec 5.low energy theorems based on pcac and current algebra do imply that pions decouple from each other and with other matter at low momenta , but they do not directly imply that the pion has a canonical dimension in the infrared.current algebra requires that the current coupling to the pion has canonical dimension @xmath77 . however , the canonical dimension ( for the @xmath78 behaviour of the pion ) is @xmath79 . this only means that the pcac relating the two needs a dimensionful parameter , but does not directly relate the two dimensions . consider the time ordered correlation function of a current in a broken direction and the corresponding @xmath15 field : @xmath80 we get , @xmath81 \mid 0 > \not = 0\ ] ] since @xmath82 = \sigma(0 ) $ ] has a non - zero expectation value . this means for the fourier transform of this correlation function , @xmath83 this requires that in the spectral representation of the correlation function there is necessarily a @xmath84 contribution implying a _ simple pole _ at zero mass.any other singularity , such as a branch point will not satisfy the above equation . the fact that the @xmath3 channel also has an infrared singularity is equally general and should be understandable in a general way . in sec . 5 and 6 we provide two proofs which rely essentially on symmetry arguments . the first uses ward identities , and the second , the schwinger- dyson equations . it was argued in sec . 2 that the leading term in @xmath4 expansion retains all important effects of the goldstone boson fluctuations . analytic calculations are possible in this limit and therefore we can analyse such effects in detail.there is an extensive literature on this technique @xcite.in this section we provide a new arsenal to this technique.we define a generalized generating functional of one - particle irreducible ( 1pi ) vertices and obtain an explicit expression for it in the large n limit . this is made possible by involving an auxiliary field . our method provides the easiest way of extracting all correlation functions without doing a separate calculation for each @xcite . we present results for both linear and non - linear @xmath3-models . @xmath87 $ ] , the generating functional of connected green s functions , is given by , @xmath88 } = \int { \cal d } { \vec \phi } e^ { i(s[{\vec \phi } ] + { \vec h}.{\vec \phi})}\ ] ] where , @xmath89 connected green s functions of @xmath90 are obtained from various functional derivatives of @xmath87 $ ] . @xmath91 $ ] , the generating functional of 1pi vertices , is obtained from @xmath92 $ ] by a legendre transformation . @xmath93 = w[\vec h ] - \vec h.\vec \phi\ ] ] where , @xmath94}{\delta h_i(x)}\ ] ] the vacuum expectation value ( vev ) of @xmath90 is obtained as the value at which @xmath85 is minimized.the second functional derivative about this minimum gives the inverse propagator . further , various functional derivatives of @xmath85 about this minimum give the 1pi vertices . the connected green s functions can be extracted by building _ only _ tree diagrams using these propagators and vertices . in the linear @xmath3-model , introduce an auxiliary field @xmath95 to make the exponent bilinear in @xmath96 : @xmath88 } = \int { \cal d } \lambda { \cal d } { \vec \phi } e^{i \int d^dx ( \frac{1}{2 } ( \partial_{\mu } { \vec \phi}(x))^{2 } -\frac{1}{2 } \lambda(x)({\vec \phi}^2(x)-c^2 ) + \frac{1}{4u } \lambda^{2}(x)+{\vec h}(x ) \cdot { \vec \phi}(x)}\ ] ] now an integration over each @xmath97 , gives a @xmath98 factor . therefore , @xmath88 } = \int { \cal d } \lambda e^{i \int d^dx ( \frac{1}{4u } \lambda^{2}(x)+ \frac{c^2}{2 } \lambda ( x ) ) -\frac{n}{2 } tr ln \nabla -i \frac{1}{2 } h.\nabla^{-1}.h}\ ] ] where , @xmath99 on defining , @xmath100 = n w[{\vec h } ] \label{redefs}\ ] ] we get , @xmath101 } = \int { \cal d } \lambda e^{i n [ ( \int d^dx(\frac{1}{4u } \lambda^{2 } + \frac{1}{2}c^2 \lambda ) ) + \frac{i}{2}tr ln \nabla - \frac{1}{2 } h. \nabla^{-1}.h]}\ ] ] we are interested in the limit @xmath102 , @xmath103 holding @xmath104 fixed . with the variables redefined as in eqn . [ redefs ] , it is clear that this limit is dominated by the saddle point . the fluctuations about the saddle point give contributions which are non - leading in n. thus the leading contribution is , @xmath105 = \int d^dx(\frac{1}{4u } \lambda^{2}(x)+\frac{1}{2}c^2 \lambda(x ) ) \\ + \frac{i}{2 } tr ln \nabla - \frac{1}{2 } h. \nabla^{-1}.h \label{ggf2}\ ] ] where @xmath106 , the saddle point , is a functional of @xmath107 as given by , @xmath108 we may obtain @xmath91 $ ] from @xmath109 $ ] by a legendre transform as described earlier . but this form is not easily amenable to further analysis . instead we note that the saddle point equation may be rewritten as @xmath110 where , @xmath111 notice that these equations can be obtained from the following functional of @xmath112 and @xmath95 : @xmath113 & = & \int d^dx(\frac{1}{4u } \lambda^{2}(x ) + \frac{1}{2}c^2 \lambda(x ) + \frac{1}{2}(\partial_{\mu } \vec \phi)^2 ( x ) \nonumber \\ & & - \frac{1}{2 } \lambda(x ) \vec \phi^2(x ) ) + \frac{i}{2}tr ln \nabla \label{gammainf}\end{aligned}\ ] ] by taking functional derivatives : @xmath114 } { \delta \phi_i(x ) } , \quad - \kappa(x)=\frac{\delta \gamma_{\infty}[\vec \phi,\lambda ] } { \delta \lambda(x ) } \label{ggf1}\ ] ] here we have temporarily added a source @xmath115 for @xmath95.when restricted to correlation functions with only @xmath116 on external legs , @xmath115 can be set to zero , yielding eqn 30 . these equations imply that @xmath117 $ ] and @xmath118 $ ] are related by a functional legendre transformation in both arguments . this means that the connected green s functions of @xmath112 and in fact also of @xmath119 ( i.e. functional derivatives of @xmath117 $ ] wrt both @xmath120 and @xmath115 ) can be obtained from tree diagrams built from the action @xmath121$].this connection is simply an algebraic consequence of the legendre transformation . we have thus the option of constructing green s functions involving the composite operator @xmath122 also by having @xmath95 as external legs . by introducing an auxiliary field @xmath95 we have obtained an explicit form of the generating functional of 1pi vertices which is far more amenable to analysis than if we had obtained a @xmath123 by substituting the saddle point solution for @xmath95 .moreover we find that the properties of the phase with spontaneously broken symmetry can be understood in a simple way in terms of this auxiliary field . ofcourse , now @xmath95 also appears in propagators and vertices . the usual generating function @xmath124 $ ] can however be recovered by using @xmath125 from eq.[ggf1 ] to replace @xmath95 as a functional of @xmath126 in @xmath127 $ ] . there is hardly any change in our calculations ( and results ) when passing over to the non - linear @xmath3-model . now we have , @xmath128 } = \int { \cal d } \vec \phi \pi_x \delta^d(\vec \phi^2(x)-c^2 ) e^{i \int d^dx ( \frac{1}{2 } ( \partial_{\mu } \vec \phi(x)^{2 } + \vec h(x ) \cdot \vec \phi ( x)})\ ] ] using the fourier representation for the functional @xmath46 function , @xmath129 } = \int { \cal d } \lambda { \cal d } \vec \phi e^ { i\int d^dx ( \frac{1}{2 } ( \partial_{\mu } \vec \phi(x)^{2 } ) -\frac{1}{2 } \lambda ( \vec \phi^2 ( x)-c^2 ) + \vec h(x ) \cdot \vec \phi(x))}\ ] ] in this form the only difference with the linear @xmath3-model is that a quadratic term in @xmath95 is missing in the exponent.equivalently , the non - linear @xmath3-model is simply the special case @xmath130 of the linear @xmath3-model . thus all results for the non - linear @xmath3-model can be extracted by taking this limit in our formulae.we find that it is straightforward to take this limit and the results are not very different . we have been cavalier in our manipulations with the functional integrals.for instance , the @xmath131 term has to be carefully defined.we may choose a pauli - villars or a lattice cut - off for defining this term.if we consider a functional taylor expansion of the tr ln term about a constant value of @xmath95 , for @xmath1 , only the term linear in @xmath95 is sensitive to the cut-off.we may therefore obtain a renormalized @xmath132 $ ] by making a specific cut - off independent choice for this term . in @xmath2 , in addition , the term quadratic in @xmath95 is also divergent and requires a renormalization.for the sake of uniformity , we shall introduce local counterterms @xmath133 for both @xmath32 and @xmath2 . in the @xmath32 case , this amounts to a finite renormalisation of u. because of the presence of massless modes , renormalisation should be so performed that it does not introduce spurious infra - red divergences . hence we choose to define renormalised quantities at @xmath134(say ) . then the renormalised effective action is @xmath113 & = & \int d^dx(\frac{1}{4u } ( \lambda(x)-\mu)^{2 } + \frac{1}{2}c^2 ( \lambda(x)-\mu ) + \frac{1}{2}(\partial_{\mu } \vec \phi)^2 ( x ) \nonumber \\ & & - \frac{1}{2 } \lambda(x ) \vec \phi^2(x ) ) + \frac{i}{2}tr^\prime_{\mu } ln \nabla \label{gammaren}\end{aligned}\ ] ] @xmath135 means that the * local * part of the first two terms in the functional expansion around @xmath134 are to be ignored.now u and c are renormalised quantities . thus the model is renormalizable in the usual sense.it is also seen that the wave function renormalization for @xmath95 , equivalently the composite operator @xmath119 , is finite in the @xmath102 limit.(for a discussion of renormalizability of the @xmath4 series see ref . @xcite ) . the discussion of renormalizability is equally valid for the the non - linear @xmath3-model . the only difference is that the bare @xmath136 term is not present in the action.the quantum effects introduce such a term which is finite in @xmath1 and diverges logarithmically with the cut - off in @xmath2.thus the non - linear @xmath3-model is not renormalizable in this sense in @xmath2 even in the @xmath137 limit . in this section , we extract the infrared behaviour of the correlation functions , in particular , in the phase with sbs . first we need to obtain the vev s @xmath138 and @xmath139 of the fields @xmath116 and @xmath95.for uniform ( in particular zero ) fields @xmath140 ( and @xmath115 ) , the vev s are space- time independent.therefore we get @xmath141 @xmath142 \label{mini}\ ] ] the analysis below closely follows ref . we first consider the situation in the absence of the external magnetic field , @xmath143.now @xmath144 and so there are three cases : 1 . @xmath145 and @xmath146 corresponds to the critical point.this is reached for the critical value @xmath147 where , + @xmath148 2 . @xmath149 and @xmath150 is the symmetric phase . @xmath139 is given by , + @xmath151 + where we have used eqn.[casei ] 3 . @xmath152 , @xmath153 is the phase with sbs.the vev @xmath154 is given by , + @xmath155 + as seen by using eqn.[casei ] for any @xmath156 we have a solution for a real @xmath154 and this corresponds to a phase with sbs . on the other hand for any @xmath157 we have a solution for @xmath139 in eqn.[caseii ] , corresponding to the symmetric phase . in terms of the fluctuations @xmath158 we have for the symmetric phase , @xmath113 & = & \int d^dx ( \frac{1}{2 } ( \partial_{\mu } \vec \phi(x)^{2 } ) -\frac{1}{2}\lambda_0 \vec \phi^2 ( x ) -\frac{1}{2}\lambda'(x ) \vec \phi^2 ( x ) ) \nonumber \\ & & + \frac{1}{4 } \int \int d^dx d^dy \lambda'(x)i(x - y,\lambda_0 ) \lambda'(y ) \nonumber \\ & & + \frac{i}{2}tr'_\mu ln ( -\partial^2- \lambda_0-\lambda')\end{aligned}\ ] ] where @xmath159 is the fourier transform of @xmath160+\frac{1}{u}\ ] ] all terms linear in the fluctuations drop out because we are expanding about the extremum value.@xmath139 is seen to be the @xmath161 of the @xmath126 field . for the phase with sbs we choose @xmath162 . in terms of the fluctuations , @xmath163 we get , @xmath164 & = & \int d^dx ( \frac{1}{2 } ( \partial_{\mu } \vec \phi(x))^{2 } -v \lambda(x)\sigma(x ) -\frac{1}{2 } \lambda(x ) ( \sigma^{2}(x)+\vec \pi^2(x ) ) \nonumber \\ & & + \frac{1}{4u } \int \int d^dx d^dy \lambda ( x)i(x - y,0 ) \lambda ( y ) + i \frac{1}{2}tr'_\mu ln ( -\partial^2-\lambda)\nonumber \\\end{aligned}\ ] ] @xmath165 is massless as is to be expected . the mixing of @xmath3 with the auxilary field @xmath95 is a crucial feature of this phase . this mixing governs the infrared behaviour of the correlation functions as seen below . the terms quadratic in @xmath3 and @xmath95 may be written in momentum space as , @xmath166 now , @xmath167 we get the propagators by inverting the matrix of the quadratic form in eqn.[quad ] . @xmath168 we are now in a position to extract the infrared behavior of various correlation functions . the propagators have the following infrared behaviour : @xmath169 thus the leading n approximation reproduces the correct infrared behaviour for the @xmath15 and the @xmath3 propagators , viz . , @xmath15 has canonical @xmath170 behaviour , whereas @xmath3 has an anomalous dimension @xmath0 , precisely as if @xmath171.we now demonstrate explicitly that the other correlation functions also have the infrared behaviour required by the soft pion theorems . the interactions of @xmath15 and @xmath3 are only mediated through the auxiliary field @xmath95 in this formalism , and the @xmath95 propagator has an unusual @xmath172 infrared behaviour.this is the essential reason why @xmath15 s and @xmath3 s decouple from each other in the infrared , and the results of soft pion theorems are valid . for the leading ir - behaviour , the multi-@xmath95 vertices are irrelevant . the effective action is then quadratic in @xmath95 and consequently , the @xmath95-field can be exactly integrated out.this leads to the following effective action to describe the leading ir behaviour : @xmath173\ ] ] this summarises all the essential results of this paper . thus the large n approximation provides a valuable non perturbative tool for obtaining effects due to quantum fluctuations of goldstone bosons.it incorporates the infrared effects correctly.it also incorporates the ultraviolet behavior correctly : for @xmath1 , the ultraviolet behaviour is of the non - gaussian fixed point , and for d=4 , it is of the o(n ) gaussian fixed point . it has an explicit @xmath3 field , so that effects due to its infrared behaviour on other matter can be investigated . also the technique provides a way of going beyond the leading infrared region.it provides a detailed structure for the propagators of both @xmath15 and @xmath3 with correct asymptotic behaviours . we now consider the situation with a uniform external magnetic field @xmath174 . now , @xmath175 implies that neither @xmath139 nor @xmath176 is zero.this is the choice that removes terms linear in the shifted field @xmath3 one gets essentially the same @xmath177 as the one we got for the symmetric phase but with a crucial additional term @xmath178 and with the proviso that elsewhere @xmath119 now stands for @xmath179 . thus @xmath15 has a mass @xmath180 . @xmath3 has a mixing with @xmath95 as in the sbs case.the @xmath181 propagators may be computed as before . the equation governing @xmath139 is now , @xmath182 with @xmath183 . by an analysis of these equations one obtains a mass that scales as @xmath184 for the pions and as @xmath185 for the sigma . for applications to statistical mechanics , one should use the expressions derived here continued to the relevant number of euclidean dimensions . in this section we demonstrate that the @xmath186 diverges for @xmath187 , that @xmath188 in the infrared regime and the softness of multipion scattering amplitudes using ward identities . chiral ward identities merely reflect the underlying symmetry structure of the theory and are useful both in the symmetric phase where they relate correlation functions of the same type ( n - point functions with fixed n ) as well as in the ordered phase where they relate different n - point functions in a non - trivial way . note that in the generating function both the measure @xmath189 , and the action @xmath190 , are invariant under @xmath67 rotations generated by @xmath191;the source term is also invariant under a simultaneous rotation of both the source @xmath192 and the @xmath112 field.considering an infinitesimal transformation , this implies for @xmath85 , @xmath193 = 0 \label{b}\ ] ] when an external field @xmath192 is present.this means that @xmath85 has only o(n ) invariant combinations such as @xmath194 etc . , except for one additional term @xmath195 , depending on the external field . by successive differentiations of this identity w.r.t . the fields @xmath196 , one can generate various forms of ward identities.in particular , an identity that will prove most useful later on is @xmath197 where @xmath198 is the expectation value of the order parameter . + to explore the infrared properties , we consider the thermodynamic potential , which is called the the exact effective potential in quantum field theory . we include an external magnetic field to serve as an infrared regulator and consider the singular behaviour as this field goes to zero . the various zero momentum n - point functions are obtained by functionally differentiating this potential w.r.t . constant fields and evaluating the derivatives at the minima of the potential . in the ordered phase this minima occurs at a non - vanishing value @xmath199 as the magnetic field is reduced to zero . + as a consequence of the ward identity , eqn.[b ] , the effective potential has the form , @xmath200 since @xmath201 the vev @xmath154 is given by , @xmath202 where h is the magnetic field along the @xmath3-direction . we also have @xmath203 @xmath204 @xmath205 etc . here @xmath206 etc . are the derivatives of @xmath207 w.r.t . its argument . since @xmath208 we get , @xmath209 also , @xmath210 it can be seen that eqns . [ d ] , [ c ] and [ e ] together are equivalent to the @xmath78 limit of the identity in eq.[a ] . > from eq.[f],the 1pi pion four - point vertex at zero momentum is given by @xmath211 and from eq.[e ] , the @xmath212 1pi vertex is given by @xmath213 hence the @xmath214-scattering amplitude at zero momentum @xcite is , @xmath215\ ] ] in other words @xmath216 irrespective of @xmath217 , @xmath218 is at least as soft as h. this is equivalent to the soft pion theorems of current algebra @xcite . note that this result has been proven here without the detailed assumptions of current algebra and pcac . + on introducing the shorthand notation @xmath219\ ] ] and @xmath220 one can write the general result @xmath221 now let us consider the zero momentum 1pi vertex function @xmath222 ( with @xmath223 ) corresponding to n - longitudinal and m - transverse external legs.then all tensors @xmath224 will vanish when @xmath225 as then at least some of the @xmath126 s in these tensors will have to be transverse.furthermore the leading ir behaviour is equivalent to treating @xmath154 as being very large ( compared to the momenta and equivalently to h ) . it is clear that due to the exchange of the massless goldstone particles in the intermediate state , the n - point functions should have infrared singularities for sufficiently large @xmath226 . from very general considerations , it follows that the nature of these singularities becomes more severe with increasing @xmath226 . thus , for sufficiently large @xmath227 , the derivative @xmath228 has the structure @xmath229 where @xmath230 therefore @xmath231 now three situations can arise : ( a ) @xmath232 is finite or goes to zero slower than @xmath233 as @xmath234 . then @xmath235 is more singular than @xmath236 . also @xmath237 from ( 67 ) : ( b)@xmath232 and @xmath233 have the same leading infrared behaviour . further , ( 63 ) already implies that @xmath238 and @xmath239 which is a rather strong result . then all @xmath240 for @xmath241 have the same infrared behaviour . we shall show shortly that this is inconsistent . ( c ) @xmath242 vanishes faster than @xmath233 as @xmath243 . this has the consequence that @xmath244 has softer ir singularity than @xmath245 for @xmath246 . by virtue of ( 63 ) this choice also implies that @xmath247 . this choice also implies that @xmath218 vanishes faster than @xmath248 from ( 67 ) . as these features are already invalidated in perturbation theory , ( c ) will not be considered any more . now , for both the situations a ) and b ) it follows that the dominant contribution to @xmath222 will come from @xmath249 in ( 70 ) ; this is the maximum possible value for @xmath250 . hence @xmath251 this further implies that @xmath252 this means that as far as the infrared behaviour of all the vertex functions is concerned @xmath253 this and the fact that pions are non - interacting in the ir leads to the stated result on @xmath186 _ viz . _ @xmath254 for @xmath1 and @xmath255 for @xmath2 . but these behaviours for @xmath186 are inconsistent with the situation ( b ) which demand @xmath256 . thus only ( a ) need be considered . in order to get an appreciation for the nature of the effective potential , we analyse it in the large n limit.from the generating functional @xmath257 $ ] of sec.3 . one can recover the usual generating functional @xmath258 $ ] by eliminating @xmath95 using its equation of motion . since we are interested in constant fields @xmath259 , it is sufficient to have the @xmath95 field also constant . subtracting eqn . [ mini ] from the analog of eqn . [ saddle ] that results from the renormalised @xmath177(eqn [ gammaren ] ) , we get , @xmath260 = ( \vec\phi^2 - v^2)\ ] ] which can be solved to yield the implicit solution @xmath261 for given @xmath262.therefore the effective potential has the form , @xmath263 where @xmath95 is to be substituted as in eqn . further we know that the minimum of the effective potential is at , @xmath264 such that @xmath265 , as @xmath266.this means that @xmath95 and also @xmath267 when expanded about the minimum @xmath268 has taylor coefficients which are increasingly singular in @xmath139 and hence in @xmath269 . in this section we present an argument based on schwinger - dyson equations that shows the diverging @xmath186 as a consequence of the infrared divergence of the goldstone bosons . the heirarchy of schwinger - dyson equations are obtained by exploiting the following property of functional integrals @xmath270 to be specific let us consider @xmath271\ ] ] the above identity implies the following equation for @xmath85 , the generating functional of 1pi diagrams : @xmath272\end{aligned}\ ] ] where @xmath273 and @xmath274 it should be noted that @xmath275 and @xmath276 are the propagator and 3-point vertex respectively only when the above derivatives are evaluated at the minimum of @xmath85 namely , at @xmath277 such that @xmath278 hence @xmath279 and @xmath280 in eqs([dinv ] ) and ( [ vinv ] ) are @xmath126-dependent . + it will prove useful to display explicitly the behaviour of the 1pi graphs in the large for example , the four - point vertex has a @xmath4 dependence for large n. likewise , the expectation value of the order parameter has a @xmath281 behaviour , and hence the 3-point functions in the ordered phase will have a @xmath282 factor . now consider the sd equations for @xmath283 . this equation is best represented diagrammatically as in fig . 2 . in these figures solid lines represent exact @xmath3-propagators and dashed lines exact @xmath15-propagators ( except the lines with subscript ` 0 ' denoting tree level propagators ) , and all blobs represent exact vertices . + we assume some infrared regulator like the external field h. then @xmath284 , the goldstone boson mass is not zero and we can safely study the zero momentum behaviour of various graphs . the degree of infrared divergence is reflected by the leading @xmath284-dependence . our input is that the exact pion propagator has the behaviour @xmath285 for @xmath78 , i.e. we put in the information that the pion has a canonical behaviour in the infrared and study the consequences for the @xmath3-propagator.diagram 2(a ) contributing to @xmath283 would behave like @xmath286 if @xmath287 were o(1 ) in the ir ( and consequently the total @xmath283 would also be o(1 ) due to the ward identity ( eqn [ a ] ) ) . this contribution is of order o(1 ) in the large - n count . thus , either the ir - singular behaviour of 2(a ) is cancelled by other diagrams in fig ( 2 ) or * @xmath287 must be at least as soft as @xmath288*. diagrams with potential ir - divergences are 2(b),2(c ) and 2(d ) ; apart from all of them being of o(1/n ) , none of them is as ir - singular as 2(a ) . hence cancellation is ruled out and @xmath287 must behave at least like @xmath289 with @xmath290-positive semidefinite . consequently @xmath283 must vanish as either @xmath291 or @xmath289 whichever is dominanat . + while this proves conclusively that @xmath186 must diverge , we still need to restrict @xmath290 further . it should be noted that @xmath283 has terms of order @xmath292 and @xmath293 in the large - n count . on the other hand @xmath283 must vanish as some power of @xmath284 . this is only possible if the @xmath292 terms are cancelled by such terms from the other diagrams . as all the diagrams except 2(a ) are @xmath294 , generically this can only happen from the ultraviolet contributions to 2(a ) . within @xmath4 perturbation theory at least , such cancellations are not possible in which case the @xmath292 terms must be generated in the infrared by 2(a ) and this is possible only if @xmath295 . this completes the proof based on sd - equations . we have thus established that the divergence of the longitudinal susceptibility is very general depending only on the goldstone phenomena . in fact , we have shown on very general grounds that the leading infrared behaviour of the longitudinal two - point function is completely universal . now we show that the same considerations lead to certain universal features of multipion scattering amplitudes . such universal features were alluded to in weinberg s article @xcite . consider , for example , the expression ( 67 ) for the @xmath214 scattering amplitude at zero momentum . eqn ( 74 ) can be recast as @xmath296 hence @xmath297 as @xmath298 and one has @xmath299 on recalling that in the absence of the infrared cut - off h , the scale that replaces @xmath248 is @xmath300 , we recognise that the leading universal infrared behaviour @xmath301 is indeed the well known result for soft pion scattering ( usually derived in @xmath2 ) . this universal behaviour does not depend on the space time dimensionality . however ( [ new])demonstrates that even the next to leading term is universal . the above considerations can be applied to multi - pion scatering amplitudes and one again finds the same universality of the first two terms . it is also clear from this paper that these results are non - perturbative in nature . the presence of these non - analytic ( in energy - momentum variables ) terms in @xmath2 had been noted by li and pagels @xcite , the coefficients of which in the case of @xmath214 scattering had been calculated by lehmann , and lehmann and trute @xcite using unitarity . weinberg showed that the application of renormalisation group can be used to show the universality of these logarithmic terms . he implicitly uses chiral symmetry . though superficially it appears that the universality of the logarithmic terms is a one - loop result , careful reexamination of weinberg s demonstartion shows that it is an all - loop result . in our approach , the origin of the universality of these subdominant logarithmic terms is directly related to universal divergence of the longitudinal susceptibility . our large - n derivation , which links the divergence of the @xmath186 to the special properties of the auxiliary field also demonstrates the universal features in a very straightforward manner . the multipion scattering amplitudes in this case can be classified into three categories : a ) those involving only @xmath95 and @xmath15 propagators , and the @xmath302 vertices . for the scattering amplitude involving 2n pions ( including initial and final states ) , one needs @xmath303 and @xmath304 propagators . hence the ir - behaviour of all multipion amplitudes of this class is @xmath305 . b ) amplitudes involving one @xmath306 vertex and n @xmath302 vertices . the @xmath306-vertex scales like @xmath307 . hence this contribution to the amplitude scales like @xmath308 , which is less dominant than the behaviour a ) . finally , there is the class c ) which consists of amplitudes involving combinations of multi-@xmath95 vertices with @xmath309.their contribution can be seen to be even less dominant than b ) . because of the @xmath310 factors in the denominators of the @xmath95-propagator , one gets a universal correction factor @xmath311 $ ] and the leading ir behaviour of these correction factors is the same in the linear and non - linear @xmath3-models . this is precisely the same quantity that governs the universal divergence of @xmath186 . we now make a few geenral remarks about conventional chiral perturbation theory ( chpt ) and compare it with our approach . as we have already stated many times earlier , the leading ir behaviour of systems with goldstone bosons is completely universal , and in the sector of goldstone bosons only , is represented by the lagrangian @xmath312 this , following weinberg , can be cast in the form @xmath313 the validity of ( [ new-1 ] ) is , however , strictly in the extreme infrared . one of the procedures adopted to extend ( [ new-1 ] to regions beyond the extreme ir is the so - called chiral perturbation theory . in the form that is mostly practised , chpt amounts to starting , for example , from a lagrangian of the form @xmath314 contrary to the attitude of the early days of effective lagrangians , one now carries out loop calculations with lagrangians of the type ( [ new-2 ] ) and ( [ new-3 ] ) . owing to perturbatie non - renormalisability of these lagrangians , one needs infinitely many counter terms to absorb the various resulting infinities .. however , to any any desired accuracy in the momenta involved , one needs only a finite number of counter terms . the structure of the resulting renormalised lagrangian whose domain of validity is somewhat larger that that of ( [ new-1 ] ) is that it is charaterised by a number of unknown parameters corresponding to the number of counteer terms used as well as certain universal parts independent of these unknown coefficients ( @xmath315 is always treated as known ) . this is indeed the general structure argued by weinberg . our analysis that the leading ir behaviour is governed by the gaussian fixed point @xmath70 points to a more efficient way of realising the results of chpt . this suggestion was implicit in @xcite also . this consists of perturbing the ir fixed point directly with the help of irrelevant operators . their influence on all correlation functions can be directly computed by the use of the rg techniques . the ultraviolet degrees of freedom have no role to play in this scheme . this method will be applied to practical calculations of chpt elsewhere . in this paper we have emphasised that the quantum fluctuations of the goldstone bosons have some drastic effects which are not encountered either in a tree level or loopwise analysis of the linear or the non - linear @xmath3- models.nevertheless,the leading infrared behaviour of the correlation functions again become identical in the linear and the non - linear versions.it is remarkable that the longitudinal fluctuations also become ` massless ' as a consequence of the masslessness of the goldstone bosons . we have exhibited the generality of this phenomenon giving three separate arguments based on i ) . renormalization group flows , ii ) . ward identities , iii ) . schwinger - dyson equations . heuristic derivations of the soft - pion theorems within the linear @xmath3-model are based on the fact that @xmath3-mass ( @xmath316 ) being non zero , for all processes involving momenta much smaller than this mass the @xmath3-propagator can be replaced by @xmath317 . one may wonder whether the results of this paper would vitiate this and consequently the soft - pion theorems too . but as demonstrated in the sections on the large - n analysis as well as the one on ward identities , the soft pion theorems are still valid . the deeper reason for this is , of course , the renormalisation group point of view according to which the leading ir - behaviour is controlled by the gaussian fixed point . + as stressed sufficiently in this article , both the qualitative and quantitative aspects of the result , as far as the susceptibilities are concerned , were known in the literature from somewhat differing perspectives which were not general enough . what we have done here is to demonstrate the full generality of this result as well as extend the analysis to include the nonperturbative ir behaviour of correlation functions also . despite the fact that the result concerning @xmath186 had appeared in the literature in different guises , its significance does not appear to have been well appreciated . + we emphasise that in the chiral limit the ` sigma particle ' has a singular infrared behaviour ( i.e. logarithmic divergence ) even in four dimensions . independently of the space - time dimensions , the goldstone bosons have canonical dimensions and decouple from each other and their logitudinal partners in the infrared.we pointed out that this can be understood from the goldstone theorem , current algebra and pcac on the one hand and the flow towards @xmath318 gaussian fixed point on the other hand . this latter point of view is very valuable because one can easily study the non - leading corrections to the infrared behaviour by standard renormalization group methods @xcite . all effects of the microscopic theory could be parametrised in a few irrelevant and marginal operators around the @xmath318 gaussian fixed point and their effects computed . this approach provides a technically more efficient and conceptually simpler alternative to chiral perturbation theory , which focusses on the infrared divergence of the goldstone bosons instead of the non - renormalizable ultraviolet divergences . we showed that the leading @xmath319 approximation takes into account all the infrared behaviours accuratly , including that of the longitudinal fluctuations , in contrast to perturbation theory based on the linear or the non - linear @xmath3- models.we developed a technique which allows explicit computation of the correlation functions even far away from the infrared region.this provides a useful non - perturbative technique for obtaining a qualitative understanding of the effects of fluctuations . a convenient tool for understanding longitudinal susceptibility is the equation of state @xcite.in terms of the reduced variables @xmath320,where m is the magnetisation , t the deviation from the critical temperature , and @xmath321 the standard exponents , the equation of state takes the universal form + @xmath322 as h tends to zero , @xmath323 must tend to one of the roots of @xmath324 . with suitable normalisation this root ( corresponding to spontaneous magnetization ) can be arranged to be at @xmath325.now , the question of whether @xmath186 diverges or not can be answered by knowing the manner in which @xmath324 approaches zero as @xmath323 approaches @xmath326.if this approach is parametrised as + @xmath327 where p is a positive semidefinite number , the behaviour of @xmath186 as h tends to zero is given by + @xmath328 in the large - n expansion for arbitrary d ( except d=4 where there are logarthmic modifications),the equation of state is @xmath329 as can be proven using the euclidean version of eqn([state ] ) of sec 4(see also @xcite ) . as the exponent of @xmath330 is greater than 1 one concludes that the longitudinal susceptibility diverges for all @xmath331 . modulo the wavefunction renormalisation constant , @xmath186 is just @xmath332 . there are interesting consequences of this phenomena both for finite temperature qcd as well as qcd at zero temperature . in the case of the former , under the well argued scenario that this is a second order transition , wilczek and rajagopal @xcite conclude that the transition must belong to the universality class of @xmath333 magnet models . then the general results of this paper become applicable . + in numerical simulations of the qcd chiral phase transition at finite temperature , the quantity that has been used to characterise the nature of the phase transition is the so - called @xmath334-cumulant defined by @xcite @xmath335 where @xmath336 is the chiral condensate and @xmath337 the current - quark mass.thus @xmath338 below @xmath339,but close to it , one should expect , on the basis of the considerations of this paper @xmath340 the same considerations apply to the spontaneously broken phase of zero temperature qcd . here one should find @xmath341 characterstic of goldstone phenomena in four dimensions . + of course , the arguments of wilczek and rajagopal that maps the problem to a classical stat mech problem in @xmath32 is valid exactly at @xmath339 . but continuity would demand that at temperatures in the vicinity of @xmath339 too the effective dimensionality should be close to 3 . thus , the expectations based on eqs ( [ cum1],[cum2 ] and [ cum3 ] ) are the correct ones . this should yield the experimentally interesting signal that for qcd at finite temperatures @xmath331 , @xmath186 should smoothly go over from eqn ( [ cum2 ] ) to eqn([cum3 ] ) as t is lowered . + indeed , wherever there are goldstone bosons , one should expect to see the singular behaviour of @xmath186 . it is therefore somewhat disappointing that clear experimental signatures of this effect have not yet been found even for the classic ferromagnets.the lack of isotropy in most real life ferromagnets would make the experimental establishment of this effect a challenging one . an attempt to see this for ferromagnets has recently been made by @xcite . they should have been looking for a @xmath342 behaviour rather than @xmath343 . the fact that the behaviour of @xmath186 is super - universal in the sense that it depends only on dimension and the existence of goldstone modes should be established in as many diverse experimental circumstances as possible . as patashinsky and pokrovsky @xcite have pointed out , the effect should also be observable in the superfluid phase of he.numerical simulations of the @xmath67- models in three dimensions , performed in the ordered phase , should also establish this effect . + kogut _ et al . _ @xcite report a completely different behaviour and different exponents for the finite temperature chiral transition in qcd in comparison to @xcite . in particular , kogut _ et al . _ do not find the crucial softening of the longitudinal mode . more specifically , they argue that pcac restricts the equation of state to be of the form + @xmath344 there is a fallacy in their argument ; pcac implies the above form of the equation of state if and only if the longitudinal susceptibility is assumed to be finite . only then a taylor expansion around zero applied magnetic field is permissible.but our results about the divergence of @xmath186 imply that the assumption of a finite @xmath186 is incorrect . once allowance is made for a diverging @xmath186 , pcac no longer places any restriction on the form of the equation of state . also , kogut _ et al . _ carry out their analysis in @xmath2 even at the critical temperature . + * acknowledgements * + the main inspiration for this work came from the paper by wilczek and rajagopal . we have benefitted from discussions with many of our colleagues , in particular , v.soni , g.baskaran , deepak kumar , p.k . mitter j.b . zuber and e. seiler . 99 for a review see , _ aspects of symmetry _ , s.coleman ( cambridge univ . press(1985 ) ) . for a review see , _ current algebra _ s.adler and r.dashen , a. benjamin inc . ( 1968 ) f.wilczek , int.j.mod.phys.*a7*(1992 ) 3911 ; k.rajagopal and f.wilczek , nucl.phys . * b399 * ( 1993 ) 395 . s. weinberg , physica * 96a * ( 1979 ) 341 . for reviews see , h.leutwyler,_recent aspects of quantum fields_,ed . h.mitter and m.gausterer , lecture notes in physics * 396*(1991)1;_principles of chiral perturbation theory _ , preprint no butp-94/13 , also published in hadron physics 94 , 1 - 46 ; hep - ph/9406283 . h. lehmann , phys . * b41 * , ( 1972 ) 529 ; h. lehmann and h. trute , nucl . phys . * b52 * ( 1973 ) 280 ; l .- f . li and h. pagels , phys . * 26 * ( 1971 ) 1204 ; _ ibid . _ * 27 * ( 1971 ) 1089 ; p. langacker and h. pagels phys . rev . * d8 * ( 1973 ) 4595 . see for instance , j.zinn-justin , _ quantum field theory and critical phenomena _ , clarendon press , oxford , 1989 . t. holstein and h. primakoff , phys . rev.*58 * , 1098(1940 ) . wallace and r.k.p . zia , phys . rev . * b12*,5340(1975 ) . w.a.bardeen , b.w.lee and r.e.schrock * d14*(1976)985 . s.weinberg , phys . rev . lett * 18 * 188(1967 ) t.h berlin and m. kac , phys . rev * 86 * ( 1952 ) 521 ; h.e . stanley , phys . rev * 176*(1968)718 . a.z . patashinsky , v.l . pokrovsky , sov phys jetp * 37*,733(1973 ) . see also , v.g . vaks , a.i . larkin and s.i . pikin , sov . 26*,647(1968 ) ; m.e . fisher , m.n . barber and d. jasnow , phys . rev * a8*,1111(1973 ) . patashinsky , v.l . pokrovsky,_fluctuation theory of phase transition _ , moscow , nauka 2075:pergamon , london . e. brezin , d.j . wallace and k.g . wilson , phys . rev . * b7 * , 232(1973 ) . f.dunlop and c.m.newman , commun . phys.*44 * ( 1975)223 . + j. lebowitz and o. penrose ( as communicated to dunlop and newman ) e. brezin and j.zinn-justin , phys . rev . lett.*36 * ( 1976 ) 691 . private communication by j.zinn-justin . amit , _ field theory , the renormalisation group and critical phenomena _ , world scientific , 1980 . i.ya.arefeva,e.r.nissimov and s.j.pacheva commun . * 71 * ( 1980 ) 213 ; i.ya.arefeva,s.i.azakov , nucl.phys . * b162 * ( 1980 ) 298 . f. karsch , phys . * d49 * ( 1994 ) 3791 . + f. karsch and e. laerman , phys * d50*(1994 ) 6954 . + c. de tar , nucl . * 42 * ( 1995 ) 457 . + koetzler and muschke , phys rev * b34*(1986)3543 ; a.kocic , j.b.kogut and m - p lombardo , nuc phys * b413*(1994 ) 503 .

understanding the interaction between cavity photons and electronic nanocircuits is crucial for the development of mesoscopic quantum electrodynamics ( qed ) . one has to combine ingredients from atomic cavity qed , like orbital degrees of freedom , with tunneling physics and strong cavity field inhomogeneities , specific to superconducting circuit qed . it is therefore necessary to introduce a formalism which bridges between these two domains . we develop a general method based on a photonic pseudo - potential to describe the electric coupling between electrons in a nanocircuit and cavity photons . in this picture , photons can induce simultaneously orbital energy shifts , tunneling , and local orbital transitions . we study in details the elementary example of a single quantum dot with a single normal metal reservoir , coupled to a cavity . photon - induced tunneling terms lead to a non - universal relation between the cavity frequency pull and the damping pull . our formalism can also be applied to multi quantum dot circuits , molecular circuits , quantum point contacts , metallic tunnel junctions , and superconducting nanostructures enclosing andreev bound states or majorana bound states , for instance .

1501.00803

cavity quantum electrodynamics ( cavity qed ) enables the study of the interaction between light and matter at the most elementary level , thanks to the achievement of a strong coupling between a single atom and a single photon trapped in a microwave or optical cavity@xcite . this paradigm has been recently brought into superconducting circuits : artificial atoms consisting of two level superconducting circuits have been coupled to superconducting cavities@xcite , in the context of circuit qed . these experiments provide an ideal playground to test the basic laws of quantum mechanics because they can be described in terms of simple models like the jaynes - cummings hamiltonian . however , such models conceal essential physical differences between cavity and circuit qed . on the one hand , the coupling between isolated atoms and cavity photons mainly occurs due to the sensitivity of the atom electric dipole to the cavity electric field . this coupling depends on microscopic details since the atomic dipole is set by the structure of the atom electronic orbitals . furthermore , one can generally perform the `` electric - dipole approximation '' which assumes that the cavity field varies little on the scale of the atomic system@xcite . on the other hand , the behavior of submicronic superconducting circuits is essentially insensitive to microscopic details due to the rigidity of the superconducting phase@xcite . for instance , the behavior of a superconducting charge qubit can be described with one macroscopic variable , i.e. the total charge of a superconducting island@xcite . this charge can vary due to the josephson coupling between the island and an external superconducting reservoir . the coupling between the superconducting charge qubit and the cavity is usually described in terms of a capacitive coupling between the superconducting island and the cavity central conductor . as a result , the chemical potential of the superconducting island is shifted proportionally to the cavity electric potential@xcite . this picture implies strong inhomogeneities of the photonic electric field on the scale of the superconducting qubit , in contrast to what is generally considered in atomic cavity qed for a single atom . recent technological progress is enabling the development of a new type of experiments where nanocircuits based on carbon nanotubes , semiconducting nanowires , two - dimensional electron gases or graphene , are coupled to coplanar microwave cavities@xcite . this paves the way for the development of `` mesoscopic qed '' , a denomination introduced in a pioneering theory work@xcite . mesoscopic qed opens many possibilities because nanoconductors can be tunnel - coupled to various types of fermionic reservoirs such as normal metals , ferromagnets@xcite or superconductors@xcite , in a large variety of geometries . so far , theoretical studies on mesoscopic qed have mainly focused on quantum dot circuits@xcite . several configurations have been suggested to reach the strong coupling regime between an electronic spin and cavity photons@xcite , or more generally , to develop quantum computing schemes@xcite . mesoscopic qed also tackles problems which go beyond the mechanics of closed two level systems coupled to cavities , usually considered in cavity or circuit qed . the interaction between electronic transport and the light - matter interaction leads to a rich phenomenology@xcite . besides , coplanar cavities could be used as a powerful probe to reveal some exotic properties of hybrid nanocircuits , like for instance the existence of topological superconducting phases@xcite , majorana quasiparticle modes@xcite , or spin - entanglement in a cooper pair beam splitter@xcite . on the experimental side , pioneering works have focused on mesoscopic rings@xcite and metallic tunnel junctions@xcite . more recently , experiments have been performed with single quantum dots@xcite and double quantum dots ( dqds ) @xcite with normal metal reservoirs . reaching the strong coupling regime between the charge states of a dqd and a cavity is still a challenge@xcite . nevertheless , interesting resonance phenomena have already been observed@xcite . several experiments have also provided evidence for a modification of the cavity behavior by finite bias transport through a dqd@xcite , including a maser effect@xcite . these recent developments call for a full description of the coupling between a hybrid nanocircuit and cavity photons . one question naturally arises : is mesoscopic qed closer to atomic cavity qed or superconducting circuit qed ? what are the specificities of the coupling between a nanocircuit and a cavity ? so far , most theory works have considered a capacitive coupling between the nanocircuit and the cavity central conductor , by analogy with circuit qed@xcite . this approach implies a coarse grained electric description of the nanocircuit , and a concentration of the non - homogeneous photonic electric field inside some capacitive elements . a few works have considered a direct coupling between the motion of electrons trapped in the nanoconductors and the bare cavity electric field , which is assumed to be constant on the scale of the nanocircuit . this is reminiscent of the descriptions used in cavity qed@xcite . in this article , we introduce a description of mesoscopic qed which bridges between these two approaches . we use a model which focuses on conduction electrons tunneling between the different elements of a nanocircuit . the tunneling electrons occupy quasi - localized orbitals in each nanocircuit element , which recalls the atomic orbital degree of freedom of cavity qed . however , there also exists collective plasmonic modes in the nanocircuit , which can screen at least partially the cavity fields . we use a gauge - invariant mesoscopic qed hamiltonian which accounts for the non - uniform screening of the cavity fields inside the nanocircuit , and for the electromagnetic boundary conditions provided by cavity conductors and voltage - biased nanocircuit dc gates . in the limit where photon - induced magnetic effects are negligible , we can reexpress the mesoscopic qed hamiltonian in terms of a scalar photonic pseudo - potential . this picture unifies the different approaches used so far to describe mesoscopic qed devices , since the photonic pseudo - potential can vary linearly with space in the case of a locally uniform photonic electric field ( dipolar coupling limit ) , as well as remain constant inside a given circuit element in the limit of a coarse grained circuit model . in the framework of a tunneling model , the photonic pseudo - potential leads to various types of linear electron / photon coupling terms : cavity photons shift the orbital energies of the different nanocircuit elements , but also induce simultaneously tunneling and local orbital transitions . this general description can be used to study the behavior of many different types of mesoscopic qed devices . for instance , it can be applied to quantum dot circuits , molecular circuits , quantum point contacts , metallic tunnel junctions , and superconducting nanostructures enclosing andreev bound states or majorana bound states . to illustrate the richness of our approach , we study in details the elementary example of a cavity coupled to a `` quantum rc circuit '' , i.e. a single quantum dot coupled to a single normal metal reservoir . the photon - induced tunneling terms between the quantum dot and the reservoir induce a non - universal relation between the cavity frequency pull and the cavity damping pull , contrarily to what is expected with purely capacitive coupling schemes at low temperatures@xcite . this paper is organized as follows . in section ii a , we discuss the gauge - invariant mesoscopic qed hamiltonian , which involves a photonic vector potential . in section ii b , we perform a unitary transformation to obtain a new hamiltonian where the electron / photon coupling is due to the scalar photonic pseudo - potential . in section ii c , we reexpress the photonic pseudo - potential scheme in the framework of a tunnelling model . in section ii d , we discuss the application of our formalism to the case of nanocircuits with superconducting elements . in particular , we give an explicit hamiltonian for a nanostructure with majorana bound states coupled to a cavity . in section iii , we work out in details the case of the quantum rc circuit coupled to a cavity . section iv concludes . appendix a gives a detailed mathematical justification for the form of the hamiltonian of section ii a , on the basis of an effective model which separates physically the tunneling electrons occupying individual orbital states in the different elements of a nanocircuit , from the plasmonic collective modes of this nanocircuit . appendix b discusses the advantages of the photonic pseudo - potential scheme . a nanocircuit encloses a large variety of degrees of freedom . our main interest in this paper is the interaction of the cavity with the ensemble @xmath0 of the `` tunneling '' conduction electrons , which can occupy orbital states in the different elements of a nanocircuit and tunnel between them . however , there also exists plasmonic electronic modes which are responsive for the screening of external fields from a massive metallic electrode . plasmons can also convey screening currents in the reservoir electrical lines , which reequilibrate locally charge distributions on tunnel junctions after a tunneling event . the question of plasmonic modes in a nanocircuit is complicated since nanoconductors have generally a low electronic density , which allows only a partial screening of the cavity field@xcite . in principle , even the metallic contacts of the nanocircuit are not able to screen totally the cavity field due to their thinness . in the limit where the nanocircuit contacts are connected to outside electrical lines with a low impedance , it is reasonable to assume that the dynamics of plasmonic modes in the nanocircuit is very fast compared to the dynamics of the cavity modes and of the tunneling electrons@xcite . in this case , one can assume that the plasmonic modes in the nanocircuit generate a distribution of screening charges which is enslaved to the position of the tunneling charges and to the value of the cavity field . this produces a renormalization of the cavity field properties , and in particular its spatial dependence nearby and inside the nanocircuit . it also produces a renormalization of the coulomb interactions between the tunneling charges . in general , the coulomb interactions between the nanocircuit charges is also renormalized by the screening charges on the cavity conductors ( see appendix a ) . the positive ions and valence electrons constituting the crystalline structure of the nanocircuit can be treated in a mean field approach , leading to an effective potential @xmath1 which confines tunneling electrons inside the nanocircuit . this confinement potential can delimit quantum dots , or induce disorder effects in nanoconductors , for instance . this approach is sufficient provided the background charges producing @xmath1 and the microwave cavity are far off resonant . in this framework , gauge invariance imposes the hamiltonian @xmath2 with@xmath3@xmath4@xmath5 and @xmath6 the elementary charge . the field operator @xmath7 includes all the tunneling charges of the nanocircuit . above , @xmath8 is a single electron hamiltonian , @xmath9 describes the coulomb interaction between the tunneling electrons ( see appendix a for details ) , and @xmath10 is the hamiltonian of a ( renormalized ) cavity electromagnetic mode . the potential @xmath11 is due to dc voltages applied on the nanocircuit electrostatic gates and to constant charges on floating conductors like the cavity central conductor in a coplanar geometry . the coupling between the cavity and the tunneling electrons arises from the photonic vector potential term of eq.([6 ] ) . for simplicity , we have considered a single cavity mode with creation operator @xmath12 and vector potential profile @xmath13 . this description can be generalized straightforwardly to the multimode case by introducing a sum on a cavity mode index in eqs . ( [ 1 ] ) and ( [ 6 ] ) . note that @xmath13 can strongly vary on the scale of the nanocircuit . the formal description of the electromagnetic field and the light / matter interaction in mesoscopic qed requires some care . in atomic cavity qed , the cavity field is generally assumed to vary slowly on the scale of an atom@xcite . in contrast , we have to take into account strong spatial variations of the photonic field on the scale of a nanocircuit . in particular , a protrusion of a cavity conductor can be used to increase the photonic field locally , close to a given quantum dot @xcite ( see fig . the dc voltage - biased electrostatic gates , used to control the nanocircuit spectrum , provide supplementary boundary conditions on the electromagnetic field . besides , we have already mentioned that the plasmonic screening charges on the nanocircuit conductors can modify the photonic vector potential profile , especially inside and around fermionic reservoirs . in order to describe this complex reality and justify mathematically the shape of eqs.([1]-[6 ] ) , we introduce in appendix a an effective model which separates physically the tunneling electrons occupying individual orbital states in the nanocircuit from the collective plasmonic modes of the nanocircuit . in this framework , one can introduce a hodge decomposition of the electromagnetic field@xcite , which can be obtained thanks to the use of an auxiliary electrostatic green s function . this leads to the terms in @xmath14 , @xmath9 and @xmath11 in eqs . ( [ 1]-[6 ] ) . in this section , we disregard the spin of charge carriers because we want to focus on regimes where charge effects prevail . the introduction of the spin degree of freedom in hamiltonian ( [ 1 ] ) would be straightforward by invoking gauge invariance ( see for instance section [ ad ] ) . the direct coupling between the particles spin and the photonic magnetic field can usually be disregarded because this coupling is expected to be very weak unless collective excitations are considered in a large ensemble of spins@xcite . from recent predictions , real@xcite or effective@xcite spin - orbit coupling provide an alternative way to obtain a spin - photon coupling . in eq.([202 ] ) , the coupling of tunneling electrons to cavity photons occurs through two terms , one in @xmath15 , corresponding to single photon transitions , and the other one in @xmath16 , corresponding to two photon transitions . the second has no reason to be disregarded in the general case , even if it has no structure in the electronic sector . for instance , this term is crucial for determining the existence of a dicke quantum phase transition when many two - level systems are coupled transversely to a cavity@xcite . the @xmath16 term can also bring a non - negligible contribution to the cavity frequency pull caused by one nanocircuit , which is a central quantity in mesoscopic qed experiments ( see appendix b ) . hence , it is convenient to introduce a new representation in which the @xmath16 term is eliminated . this task is completed in the next section . in atomic cavity qed , the power - zienau - wooley ( pzw ) transformation formally enables the elimination of a space - dependent @xmath17 term from the system hamiltonian@xcite . this generalizes the widely known `` electric - dipole '' transformation used in the case where @xmath18 can be considered as constant on the scale of an atom . however , the natural variables of the pzw hamiltonian are polarization and magnetization densities associated to the charge distribution of an atomic system , which are not directly useful in our case . to reexpress the pzw hamiltonian in terms of particles coordinates and momenta , it is necessary to perform a multipolar development of the light / matter interaction@xcite . in practice , this development is performed to a limited order , which means that only a moderate space dependence of the photonic electric and magnetic fields is taken into account . in this section , we perform a unitary transformation of hamiltonian ( [ 1 ] ) in the same spirit as the pzw transformation , but with specificities required for mesoscopic qed . upon disregarding magnetic effects , we can take into account a strong space dependence of the photonic electric field , thanks to the use of a scalar photonic pseudo - potential . this procedure avoids the development of the light / matter interaction in terms of multipoles . this is necessary to describe e.g. multi - quantum dot circuits with strongly asymmetric capacitive couplings to the cavity . furthermore , we use a second - quantized description of charges , which is more convenient to describe central phenomena of mesoscopic physics like tunneling but also superconductivity ( see next section ) . to account for the possibly strong spatial variations of @xmath13 on the scale of a nanocircuit , we define inside the volume of the nanocircuit a reference point @xmath19 and a continuous functional path @xmath20 relating any points @xmath21 and @xmath22 . we also define a photonic pseudo - potential @xmath23 with@xmath24 importantly , one can use @xmath25 inside the nanocircuit volume , provided the flux of @xmath26 through the nanocircuit can be disregarded . this criterion is generally satisfied , considering the typical amplitude of the cavity magnetic field and the size of nanoconductors . in this case , the arbitrariness on the choice of the contours @xmath20 leads only to marginal effects . one exception is nanocircuits with large loops leading to magnetic aharonov - bohm effects . this case is beyond the scope of the rest of this article , which focuses on photon - induced electric effects . we perform the transformation @xmath27 with @xmath28 this gives @xmath29 with@xmath30 and @xmath31 . in this new representation , the single electron potential and the coulomb interactions are renormalized by the term in @xmath32 , but the electron / photon coupling takes a simpler linear form . the motion of tunneling electrons is modified by the pseudo - potential @xmath33 . note that the arbitrariness on the choice of @xmath19 is not an issue , since a change in @xmath19 shifts @xmath34 by a global constant and leads to a unitary transformation of @xmath35 . in the limit where the spatial dependence of the photonic vector potential can be disregarded i.e. @xmath36 , one finds @xmath37 with @xmath38 . this corresponds to the electric - dipole approximation frequently used in atomic cavity qed . in the opposite limit of a small enough conductor or a perfect screening of @xmath39 inside this conductor , one can use a @xmath40 constant inside this conductor . this is reminiscent of the capacitive network models frequently used to describe superconducting charge qubits in circuit qed . note that we do not assume that the ensemble @xmath0 of the tunneling charges is neutral . in this section , we consider a nanocircuit gathering quantum dots and reservoirs connected by tunnel junctions . the effect of time - dependent classical fields on tunneling through quantum dots has been studied since the 1990s@xcite . mesoscopic circuit qed offers the opportunity to study the interaction between microwave photons and quantum dot circuits from a different perspective . quantum dot circuits are often described in terms of a tunneling model in which the dots and reservoirs contain quasi - localized orbitals coupled by tunnel elements@xcite . this decomposition is instrumental in order to particularize the treatment of the reservoirs and account for the irreversibility of transport processes or other reservoir - induced damping effects . importantly , the tunnel couplings arise from an overlap between neighboring orbitals . this means that these orbitals can not be considered as orthogonal , except in the limit of weak tunnel couplings ( see for instance @xcite ) . however , in practice , it may be necessary to take into account a strong inter - dot tunneling . for example , in the case of a dqd , there can exist a strong splitting between bonding and antibonding states , which can be resonant with the cavity@xcite . depending on the absolute energy scale of the dqd confinement potential , it might be necessary to go to the limit of a non - perturbative inter - dot tunnel coupling to obtain such a splitting . in order to circumvent the difficulty related to the non - orthogonality of the nanocircuit orbitals , we divide the nanocircuit into the ensemble @xmath41 of tunnel - coupled quantum dots from one side , and the different reservoirs with an index @xmath42 with @xmath43 , from the other side . we use an index @xmath44 ... @xmath45 to denote these different elements , and an index @xmath46 to denote the different orbitals in a given element @xmath47 . we assume that the low energy spectrum for the whole ensemble @xmath41 of the quantum dots can be determined in the absence of the reservoirs as the discrete spectrum of a potential profile with one or several potential wells . this procedure can be performed for an arbitrary inter - dot tunneling strength and leads to exactly orthogonal orbitals , so that @xmath48 . one can also define exactly orthogonal orbitals in the isolated reservoir @xmath42 , with @xmath43 , so that @xmath49 . in a second step , one can evaluate the tunnel coupling @xmath50 between the orbitals @xmath46 and @xmath51 of @xmath41 and @xmath42 from the overlap between their wavefunctions . this gives @xmath52 only at lowest order in tunneling@xcite . however , using a weak tunneling between @xmath41 and @xmath42 is not a severe restriction since the large density of states in the reservoirs can compensate for the smallness of the tunnel coupling elements , and lead to a large tunneling rate compatible with the kondo effect , for instance . to summarize , in the absence of the cavity , the tunnel hamiltonian writes@xmath53 where the creation operator @xmath54 corresponds to an orbital @xmath46 with energy @xmath55 and wavefunction @xmath56 mainly localized inside element @xmath44 ... @xmath45 . one can use @xmath57 at lowest order in the dot / reservoir tunnel couplings . we now reexpress the photonic pseudo - potential scheme of the previous section in the framework of the tunneling model . for this purpose , one needs to decompose the field operator @xmath58 associated to tunneling charges on the nanocircuit orbital states . at lowest order in the dot / reservoir tunneling , one can use@xcite @xmath59 hence , hamiltonian ( [ htottild ] ) directly gives@xmath60 with@xmath61 note that in principle , the term in @xmath32 from eq.([htottild ] ) renormalizes @xmath55 , @xmath62 and the coulomb interaction term @xmath63 , but this is not essential for the physics we discuss below . since the photonic pseudo - potential modifies the potential landscape seen by the tunneling charges , it can naturally affect all the parameters in the tunneling model . first , cavity photons shift the orbital energy @xmath64 , with a coupling coefficient@xmath65 in general , @xmath66 strongly depends on the indexes @xmath47 and @xmath46 due to the space dependence of @xmath67 . this makes the cavity - induced orbital energy shifts particularly relevant experimentally . for a standard metallic reservoir , it is reasonable to disregard the dependence of @xmath66 on the orbital index @xmath46 , i.e. @xmath68 , because the properties of the electronic wavefunctions in the reservoir can be considered as constant near the fermi energy@xcite . in this framework , a behavior similar to the capacitive coupling of the cavity central conductor to the reservoirs is recovered . this type of reservoir / cavity coupling enables one to interpret data obtained for a quantum dot coupled to two normal metal leads@xcite . in contrast , @xmath66 may strongly depend on @xmath46 for @xmath69 , as illustrated for instance by the example of a dqd with an asymmetric @xmath67@xcite . the hamiltonian @xmath70 also contains coupling terms in @xmath71 with @xmath72 . these terms include photon - induced tunneling terms in @xmath73 , with @xmath74 , and photon - induced transitions internal to @xmath47 , in @xmath75 . in principle , all the above cavity / nanocircuit coupling terms can coexist . in practice , we expect the terms in @xmath66 to be dominant over the tunnel terms in @xmath73 with @xmath74 , which involve a weak overlap between wavefunctions . nevertheless , as shown in the next section , the signatures of the photon - induced tunneling terms can be boosted by the large number of states they affect in the reservoirs . if @xmath67 varies slowly inside @xmath41 , the photon - induced transitions internal to @xmath41 are negligible , i.e. @xmath76 , since @xmath77 and @xmath78 are orthogonal by definition . however , these transitions become possible if the cavity field is weakly screened ( @xmath79 ) , or if a protrusion of a cavity conductor is placed close to one of the quantum dots in order to reinforce the spatial variations of @xmath80@xcite . transitions in @xmath81 can be particularly interesting in case of real or artificial spin orbit coupling which mixes spin and orbital states . in this case , the @xmath82 terms can correspond to spin - transitions inside a single quantum dot@xcite or a dqd@xcite . in the reservoirs , the @xmath83 terms are probably always non - resonant due to strong energy relaxation . hence , we will assume that these terms do not need to be treated explicitly because they can be included in a renormalization of the states @xmath84 . in practice , the potential @xmath67 can be evaluated numerically by removing all nanoconductors from the device and simulating the microwave electromagnetic field in the cavity in the presence of the metallic gates , sources and drains of the nanocircuit . this can be done using standard microwave simulation tools , which can take into account imperfect metals . it enables a realistic evaluation of the elements @xmath66 and @xmath73 . even if @xmath67 is not calculated numerically but replaced by a phenomenological expression , the expressions ( [ jj1 ] ) and ( [ jj2 ] ) remain interesting because they set constraints between the different @xmath66 and @xmath85 , which all depend on the same @xmath67 profile . we do not give details on the coulomb interaction term @xmath63 which stems directly from eq.([htottild ] ) . if coulomb blockade in the nanocircuit is already strong in the absence of the cavity , the effect of the cavity on interactions may be disregarded or treated perturbatively . in the next section , we discuss another situation where we assume that the effects of @xmath63 are negligible due to the large tunnel rate between a quantum dot and a normal metal reservoir . in the general case , coulomb interactions between the tunneling charges can lead to a large variety of effects , which we will not discuss in this work . to conclude , we obtain a decomposition of the system hamiltonian in terms of nanocircuit elements connected by tunnel couplings . while this is reminiscent of the coarse graining description of superconducting microcircuits frequently used in circuit qed , a full analogy is not possible due to the presence of the orbital degree of freedom in the mesoscopic qed case . due to this orbital degree of freedom , we obtain a large variety of electron / photon coupling terms . in this section , we have discussed the case of quantum dot circuits which raises most experimental efforts so far . however , the above approach can be used for any other type of system in which a decomposition in terms of a tunneling model is relevant . this includes quantum point contacts , molecular circuits , metallic tunnel junctions , and hybrid superconducting systems enclosing andreev bound states or majorana bound states , for instance . this last case is discussed in more details in section [ supra ] . many theoretical works on mesoscopic qed use as a starting basis a tunnel hamiltonian where all parameters ( orbital energies , tunnel rates , ... ) are perturbed by corrections proportional to the cavity electric field . these empirical approaches lack of a formal justification and can lead to unphysical predictions . our work provides a rigorous framework for the use of tunnelling models with a linear light / matter interaction ( the superconducting case will be discussed explicitly in section [ supra ] ) . as already mentioned above , our approach proceeds along a spirit similar to the pzw transformation , but with modifications necessary to take into account the specificities of mesoscopic qed , among which a strong space dependence of the cavity electric field on the scale of a nanocircuit , and boundary conditions provided by cavity conductors and voltage - biased dc electrostatic gates . interestingly , the effect of boundary conditions provided by grounded conductors on the pzw transformation has been discussed recently@xcite . however , this ref . considers only neutral atomic systems affected by locally constant cavity fields ( dipolar coupling limit ) . in the above sections , superconductivity was not explicitly taken into account . it is important to discuss the generalization of our approach to the case of superconducting hybrid nanocircuits , considering the present interest for majorana fermions@xcite or andreev bound states for instance@xcite . for simplicity , we consider below standard bcs superconducting correlations characterized in the mean field approach by a superconducting gap @xmath86 coupling electrons and holes with opposite spins . the gauge invariant minimal coupling hamiltonian of eq.([1 ] ) can be generalized as@xmath87 with @xmath88 the field operator for tunneling electrons with spin @xmath89 and@xmath90 the term @xmath91 is the straightforward generalization of eq.([1 ] ) to the spin dependent case , i.e. the single electron term @xmath92 can have a structure in spin - space , due to magnetic fields and/or spin - orbit coupling , for instance . hence , @xmath7 must be replaced by @xmath93 $ ] in the first term of eq.([1 ] ) and by @xmath94 in eq.([coul ] ) . due to gauge invariance , @xmath92 can involve the electron momentum through @xmath95 only . importantly , we define @xmath86 as a gauge invariant quantity , so that the phase parameter @xmath96 ensures the gauge invariance the pairing term of @xmath97 . the form used above for @xmath96 is valid provided photon - induced magnetic effects can be disregarded , in agreement with the approach of section [ pseudop ] . using eqs.([unit ] ) and ( [ unit2 ] ) with @xmath98 replaced by @xmath99 , hamiltonian ( [ hsupr ] ) is transformed into@xmath100 where @xmath101 is again the generalization of eq.([htottild ] ) to the spin dependent case . importantly , eq.([bim ] ) shows that in the photonic pseudo - potential scheme , the phase of the gap term is not affected by photons anymore . this represents one more advantage of this scheme . this result remains valid in the case of d - wave or p - wave superconducting correlations@xcite . in the case where electron - electron interactions can be disregarded , bogoliubov - de gennes equations are a widely used approach@xcite , which enables a diagonalization of the circuit hamiltonian in terms of quasiparticles with creation operator @xmath102 . one can define quasiparticles modes in the absence of the cavity , by using@xmath103.w_{n}(\overrightarrow{r})\ ] ] with a spinorial wavefunction @xmath104\ ] ] fulfilling @xmath105 with @xmath106 $ ] and @xmath107{cc}\widetilde{h}_{\mathcal{t}}(\vec{r } ) & \delta(\vec{r})\sigma_{0}\\ \delta^{\ast}(\vec{r})\sigma_{0 } & -\sigma_{y}\widetilde{h}_{\mathcal{t}}^{\ast}(\vec{r})\sigma_{y}\end{array } \right ] ~\text{.}\ ] ] above , @xmath108 and @xmath109 are the identity and the second pauli matrices in spin space . from the definition of @xmath102 , one gets @xmath110 with@xmath111@xmath112@xmath113 and @xmath114 , @xmath115 , @xmath116 pauli matrices in nambu space . in this section , we discuss the case of a nanocircuit enclosing majorana bound states , coupled to a cavity**. * * in principle , these self - adjoint bound states can appear at the interface between the topological and non - topological superconducting phases of a nanostructure@xcite . although an indirect majorana bound state / cavity coupling mediated by a superconducting quantum bit or a josephson junction has been considered in many works@xcite , a direct coupling can also exist@xcite . using our formalism , we can provide a general hamiltonian describing such a situation . the direct majorana bound state / cavity coupling occurs due an overlap between neighboring majorana bound states , caused by the finite size of the nanocircuit . this overlap is naturally dependent on the photonic pseudo - potential . proposals to obtain majorana bound states in condensed matter systems rely on the use of superconducting elements , which calls for the use of the results of section [ supra ] . for simplicity , we disregard coulomb interactions and use the framework of the bogoliubov - de gennes equations of section [ bdgsec ] . to remain general , we do not specify the details of the system . we consider an ensemble of majorana bound states with creation operators @xmath117 and wavefunction @xmath118 , localized inside a nanoconductor . in general , the nanoconductor also encloses a continuum of ordinary fermionic states @xmath119 with wavefunction @xmath120 located above an energy gap . we assume that the nanoconductor is tunnel coupled to a fermionic reservoir with orbital states @xmath121 with wavefunctions @xmath122 . such a reservoir can be used to measure the conductance through the majorana system@xcite . in the limit of weak tunneling , one can reexpress eq.([bdg ] ) by using the ensemble of the operators @xmath123 with @xmath124 , following similar considerations as for the tunneling model of section [ tunnel ] . this gives:@xmath125 above , the coefficients @xmath126 have expressions similar to the @xmath127 coefficients of eqs . ( [ htot]-[ho ] ) with @xmath128 replaced by @xmath129 . we use energies @xmath130 for the orbitals in the reservoir ( @xmath131 ) and the nanoconductor continuum of states ( @xmath132 ) , and a tunnel coupling parameter @xmath133 between overlapping majorana bound states . the majorana bound states and the reservoir are connected by tunnel coupling terms in @xmath134 . for simplicity , we have omitted the dc tunnel coupling between the reservoir and the nanoconductor continuum of states . photons can shift all the above mentioned parameters , due to the two last lines of eq.([tt ] ) . in particular , photons can modify the tunnel coupling between neighboring majorana bound states , due to the terms in @xmath135 , as discussed in ref.@xcite . note that in hamiltonian ( [ tt ] ) , there is no dc coupling between the majorana bound states and the nanoconductor continuum of states , which are assumed to be orthogonal by construction of @xmath136 and @xmath137 . however , there can be photon - induced transitions between the majorana bound states and the nanocircuit continuum of states due to terms in @xmath138 and @xmath139 . effects related to these last transitions have been discussed in @xcite . our approach provides a general justification for such terms . we now give an example of a specific prediction given by hamiltonian ( [ ht ] ) . we consider a cavity coupled to a `` quantum rc circuit '' made out of a single quantum dot tunnel contacted to a single normal metal reservoir . we assume that the dot has a large intrinsic level spacing so that a single dot orbital with energy @xmath140 needs to be considered . the dot is capacitively coupled to a dc gate electrode , which enables one to tune @xmath140 . we study the effects of the dot on the photonic mode with frequency @xmath141 . following the previous section , the total hamiltonian of the system is:@xmath142 the hamiltonian @xmath143 of the dot and its reservoir involves the lead density of states @xmath144 . for later use we have introduced a cutoff @xmath145 . the terms in @xmath146 describes tunneling between the dot and the reservoir , whose orbitals correspond to fermionic creation operators @xmath147 and @xmath148 , respectively , with @xmath149 and @xmath150 . one can use extra bosonic modes @xmath151 with @xmath152=\delta(\epsilon-\epsilon^{\prime})/n_{0}$ ] to describe the intrinsic damping of the cavity , i.e.@xmath153 the coupling between the quantum dot circuit and the cavity is described by the term @xmath154 with@xmath155 it is not necessary to write explicitly the coupling between the chemical potential of the reservoir and the cavity because since there is a single reservoir , it is only the difference between the potentials of the dot and the reservoir which matters . note that the terminology `` quantum rc circuit '' mainly refers to the fact that no dc current can flow through the device and dissipation is due to transport through a single tunnel barrier . however , in the general case , the concept of a local capacitor is not sufficient to describe the coupling of the electromagnetic field to the device , since the photonic pseudo potential modifies @xmath140 and @xmath146 simultaneously . in practice , one can measure the response of the cavity to a microwave drive with frequency @xmath156 , which can be described by@xmath157 above , we have disregarded coulomb interactions between the tunneling charges ( including the contribution from the @xmath158 term arising from the unitary transformation of section ii.b ) , because we assume that these interactions are smaller than the tunnel rate @xmath159 between the dot and the reservoir . such a simplification is frequently performed for interpreting standard current measurements in open quantum dot circuits@xcite . in practice , one can check the relevance of this hypothesis by measuring for instance the ac current through the dot versus @xmath140 . . the wavy lines correspond to photonic propagators , the simple dashed lines to bare electronic propagators in the normal metal reservoir , and the double full lines to electronic propagators in the quantum dot dressed by the dot / reservoir tunneling processes.,title="fig:"][diagrams ] our purpose is to identify signatures of the terms in @xmath160 , which could become visible in the cavity response , considering the recent progress on the mesoscopic qed technology@xcite . in the framework of the linear response theory , the cavity electromagnetic response to @xmath161 is determined by the green s function@xcite @xmath162 with @xmath163\right\rangle $ ] and @xmath164 . one can use the exact relation @xmath165 with @xmath166 the free cavity green s function . in the simplest model where @xmath167 and @xmath168 are independent of energy , the constant @xmath169 corresponds to pure cavity damping . in the following , we consider the limit @xmath170 @xmath171 which corresponds to standard cavities . in the limit @xmath172 , one can use a random phase approximation which leads to @xmath173 with @xmath174 the electronic response function calculated in the absence of the cavity . in the framework of a non - interacting diagrammatic calculation , @xmath175 can be decomposed into the eleven contributions represented in fig . our calculation is not perturbative in @xmath146 since we use an exact expression for the dressed propagator of electrons on the dot , corresponding to the double full lines in fig.2 . there are nine types of diagrams corresponding to the possible ordered pairs formed with the interaction constants @xmath176 , @xmath160 , and @xmath177 . the diagrams of fig.2a correspond to pure contributions from photon - induced orbital shifts in @xmath178 . the diagrams of figs . 2b and c correspond to contributions from photon - induced tunneling in @xmath179 , @xmath180 or @xmath181 . note that the contributions in @xmath179 have been separated into the diagrams of fig . 2b which diverge logarithmically for @xmath182 large and the bottom diagrams in fig.2c which are regular . there also exist interferences between photon - induced orbital shifts and photon - induced tunneling , as shown by the diagrams of fig . 2d which depend on @xmath183 or @xmath184 . the constants @xmath185 and @xmath182 set the scale of variations of @xmath186 . therefore , in the limit @xmath187 , one can replace @xmath188 by @xmath189 in eq.([rep ] ) . this gives a cavity frequency pull @xmath190 and a cavity damping pull @xmath191 . in the limit @xmath192 and at zero temperature , one obtains @xmath193 where @xmath194^{2}\gamma + 4\pi\alpha\operatorname{re}[t\gamma^{\ast}]n_{0}\varepsilon_{d}-\alpha ^{2}\gamma\right ) \nonumber\end{aligned}\ ] ] stems from the diagrams of fig.2,a , c and d , and@xmath195+\left\vert \varepsilon_{d}\right\vert ( 2\arctan[\frac{\gamma}{\left\vert \varepsilon _ { d}\right\vert } ] -\pi)\right ) \nonumber\end{aligned}\ ] ] stems from the diagrams of fig.2,b . note that the cutoff @xmath196 on the reservoir density of states is necessary to avoid a divergence of @xmath197 . the cavity damping pull is given by@xmath198n_{0}\varepsilon_{d}\right ) ^{2}~\text{. } \label{pos}\ ] ] interestingly , @xmath199 and @xmath200 vanish for @xmath201 , whereas @xmath197 vanishes only for @xmath202 . above , we have implicitly included the spin degree of freedom in the model . in the spinless case , @xmath203 and @xmath200 should be divided by a factor 2 . from eqs.([a1]-[pos ] ) , at low temperatures , when the cavity has a frequency smaller than the dot / reservoir tunnel rate @xmath185 and the reservoir bandwidth @xmath182 , and when @xmath204 , the cavity frequency pull and the damping pull fulfill the relation @xmath205 with@xmath206 this property is a manifestation of the korringa - shiba relation@xcite , whose universality is still actively discussed for quantum rc circuits @xcite . remarkably , we find that this universality is broken for @xmath160 finite . hence , one can identify the existence of the @xmath160 coupling by comparing the @xmath207 and @xmath208 curves ( fig . 3 ) or studying @xmath209 ( fig . 4 ) . here , we present results for @xmath210 because from eqs . ( [ jj1]-[jj2 ] ) , this is the most probable regime of parameters . when @xmath204 , @xmath203 and @xmath211 show identical variations with @xmath140 , and are both even functions of @xmath140 , while @xmath205 . when @xmath212 , @xmath211 presents a resonance with @xmath140 which is wider than that of @xmath203 ( fig.3 , middle panel ) . a good resolution on the tails of the @xmath203 and @xmath211 resonances could also reveal that @xmath213 and @xmath200 are not even with @xmath140 in the general case , except if @xmath214=\pi/2 $ ] ( fig.3 , bottom panel ) . in principle , even a relatively small @xmath160 can be detected . indeed , for @xmath215 , @xmath209 already significantly deviates from @xmath216 at @xmath217 where the signals @xmath203 and @xmath211 are maximal ( fig.4 , pink curves ) . this is because the effect of the photon - induced tunneling term is boosted by the large number of states involved in the reservoirs . to conclude , in order to reveal the photon - induced tunneling terms , one promising possibility is to compare the cavity frequency pull and damping pull caused by a quantum dot with a single normal metal reservoir . even a relatively weak photon - induced tunneling can affect these quantities because the large number of reservoir states reinforces the tunnel effect . and @xmath211 versus @xmath140 for different values of @xmath160 . we have used @xmath218 , @xmath219 , and pulsation scales @xmath220 and @xmath221 . ] versus @xmath140 for different values of @xmath160 . we have used the same values of @xmath146 , @xmath182 and @xmath222 as in fig.3 ] in principle , circuit - qed should enable an accurate determination of @xmath209 thanks to the use of resonant techniques . the fabrication of a quantum dot with a single normal metal reservoir inside a coplanar cavity is fully accessible with present techniques . however , so far , the closest case studied experimentally is a quantum dot with two normal metal reservoirs , which leads to more complicated physics due to an asymmetric coupling of the two reservoirs chemical potentials to the cavity@xcite . different types of quantum rc circuits based on 2-degs but also carbon nanotubes or semiconducting nanowires deserve an investigation . the ratio @xmath223 should be strongly dependent on the type of nanoconductor and the contacts configuration used . the physics discussed in this section is related to the problem of a quantum rc circuit which is not coupled to a cavity but directly excited by a classical microwave field with amplitude @xmath224 and frequency @xmath225 , which corresponds to a term in @xmath226 . at low frequencies and @xmath227 , it has been predicted@xcite that a spinless quantum rc circuit excited classically should have the same admittance as an rc circuit , with a universal resistance @xmath228 . however , this result was obtained in a purely capacitive coupling scheme i.e. @xmath204 . assuming that the admittance of the circuit is determined from the current @xmath229 with @xmath230 , this value of @xmath231 can be recovered by calculating the charge susceptibility @xmath232 , i.e. the diagram of fig . 2.a , which leads to @xmath233 because @xmath204 . however , in the general case , one should calculate the current response of the quantum rc circuit from @xmath234 because the oscillating drive can also modify the tunnel barrier transparency . only the diagram of fig . 2.a and half of the diagrams of fig . 2d should contribute to @xmath231 , whereas the diagrams of fig.2.b leading to the logarithmic term of eq.([a2 ] ) should not contribute . therefore , @xmath231 should show a behavior qualitatively different from @xmath209 . we will not discuss the detailed behavior of @xmath231 here . we will simply point out that for @xmath212 , @xmath231 is not universal , and furthermore , @xmath231 and @xmath209 are not trivially related . interestingly , the universal value @xmath228 expected for a spinless system with @xmath204 has been checked experimentally in the limit @xmath235 for a 2-deg based circuit@xcite . however , this measurement is not incompatible with a finite @xmath160 , due to its uncertainty of the order of @xmath236@xcite . note that since @xmath231 and @xmath209 are qualitatively different signals , it can be interesting to measure both in the same experiment in order to obtain more information on the system . more generally , the study of tunneling effects can benefit from a simultaneous measurement of the cavity response and the current through the nanocircuit . such a joint measurement has already been realized in the case of a dqd . the cavity dispersive shift is directly sensitive to the population imbalance between the dqd bounding and antibounding states whereas the current though the dqd corresponds to a more complex combination of state populations and tunnel rates . the simultaneous measurement of the two signals gives stronger constraints to determine the system parameters@xcite . another possibility to gain more information on mesoscopic qed systems might be to measure cross - correlations between the electronic current and the cavity output field . importantly , the example of the quantum rc circuit shows that dissipation of the cavity photons has to be considered carefully , since an open quantum dot circuit with a single contact can already induce photon dissipation . on a more general level , photonic dissipation in mesoscopic qed is a very rich problem . another interesting possibility would be to prepare non - trivial quantum photonic states by combining the effects of photonic and electronic dissipation with the non - linearity of quantum dot circuits , by analogy with the methods developped for two level systems@xcite . this so - called `` reservoir engineering '' could benefit from the specificities of nanocircuits like for instance the use of reservoirs with specific electronic orders or finite bias electronic transport . mesoscopic qed has common ingredients with atomic cavity qed , like the relevance of an orbital degree of freedom , which does not exist for superconducting qubits . it has also common ingredients with standard circuit qed , like the tunneling physics and the strong inhomogeneities of the photonic modes . it is therefore necessary to develop a specific description which combines all these ingredients . in the case where photon induced magnetic effects can be disregarded , one can express the coupling between tunneling quasiparticles and photons in terms of a scalar photonic pseudo - potential . in the framework of a tunneling model , this leads to photon - induced orbital energy shifts , which coexist with photon - induced tunneling terms and photon - induced local orbital transitions . * * to illustrate the richness of our approach , we have discussed the example of a cavity coupled to a quantum rc circuit , i.e. a single quantum dot coupled to a single normal metal reservoir . the photon - induced tunneling terms between the dot and the reservoir induce a non - universal relation between the cavity frequency pull and the cavity damping pull , contrarily to what is expected with purely capacitive coupling schemes at low temperatures@xcite . this case represents only one example of use for our formalism . we have given explicit hamiltonians for the cases of a multi - quantum dot circuit in a cavity , and a superconducting nanostructure enclosing majorana bound states coupled to a cavity . one could also consider cavities coupled to quantum point contacts , molecular circuits , and metallic tunnel junctions , for instance . _ we acknowledge useful discussions with m. baillergeau , l. e. bruhat , c. cohen - tannoudji , m. c. dartiailh , m. r. delbecq , m. p. desjardins , r. b. saptsov , m. trif , and j. j. viennot . this work was financed by the erc starting grant cirqys , the eu fp7 project se2nd[271554 ] and the anr - nanoquartet [ anr12bs1000701 ] ( france ) . _ _ _ of the perfect conductors considered for the generation of the photonic modes . the photonic field in yellow is inhomogeneous , as represented by darker yellow areas near the cavity protusions , and by the white ( screening ) areas between the small purple conductors ( c ) effective orbital nanocircuit @xmath237 . ] in the main text , eqs ( [ 1])-([6 ] ) have been introduced on the basis of a physical discussion and gauge invariance considerations . it is also instructive to use an effective model which separates spatially the tunneling and plasmonic parts of the nanocircuit . note that this effective model does not aim at calculating quantitatively the nanocircuit / cavity couplings . it rather aims at providing a full mathematical justification for the form of eqs . ( [ 1])-([6 ] ) and a deeper insight on the underlying physics . for simplicity , we assume that the nanoconductors host only individual electronic orbital modes since their electronic density is generally very low . we assume that the plasmonic modes in the fermionic reservoirs ( blue elements in fig . 1 ) are ultra fast due to the absence of dynamical coulomb blockade@xcite . in this case , one can decompose each fermionic reservoir into a purely orbital conductor ( dark grey elements in fig.5.a ) which contains no plasmons and is tunnel coupled to the nanocircuit , and perfect conductors ( small purple elements in fig.5.a ) which can host plasmonic modes , but are only capacitively coupled to the nanoconductors . these perfect conductors , or `` effective plasmonic reservoirs '' , can be placed such that the partial or full decrease of the photonic fields around the nanocircuit orbital states is reproduced at least qualitatively . we denote with @xmath238 the ensemble of the perfect conductors , which includes the cavity conductors , gate electrodes , and the effective plasmonic reservoirs ( see fig . some of these conductors are voltage biased and some others are left floating , like for instance the central conductor in a coplanar cavity . the rest of the circuit represents an effective orbital nanocircuit @xmath237 where tunneling physics prevails ( see fig . note that in principle , to grant current conservation , the `` orbital '' reservoirs have to be connected to voltage sources through wirings which host plasmonic modes , but one can assume that they are too far from the nanoconductors to have any significant influence on the nanocircuit / cavity coupling . for later use , we attribute to each conductor of the ensemble @xmath238 an index @xmath239 and we note @xmath240 and @xmath241 the corresponding surface and volume . in the following treatment we make a distinction between the conductors @xmath242 which are left floating with a constant total charge @xmath243 , and the conductors @xmath244 which are voltage - biased with a constant voltage @xmath245 or grounded ( @xmath246 ) . the generators imposing the voltages @xmath245 can be omitted from the description and included in electromagnetic boundary conditions since they are far from the cavity and nanocircuits . the volume @xmath247 complementary to @xmath248 hosts a charge distribution @xmath249 of particles @xmath176 with charges @xmath250 at positions @xmath251 , which belong to the nanocircuit @xmath237 . we also define the corresponding current distribution @xmath252 . in cavity qed , the treatment of the boundary conditions provided by the cavity conductors is usually omitted , because the atoms are very far from the cavity mirrors . therefore , a procedure based on a spatial fourier transformation can be used to separate the transverse electric field @xmath253 , which has no divergence , from the longitudinal electric field @xmath254 , which has no rotational@xcite . it is found that the value of @xmath254 is instantaneously imposed by the charge distribution @xmath255 associated to the atomic system . in contrast , the transverse fields @xmath253 and @xmath256 correspond to propagating modes . in mesoscopic qed , it is necessary generalize this procedure to account for the presence of the perfect conductors of the ensemble @xmath238 , including voltage - biased electrostatic gates . as already mentioned in the previous section , we use a charge distribution @xmath255 which includes the charges from the @xmath237 subsystem only . we will take into account the screening charges on the surface of the conductors of @xmath238 through electromagnetic boundary conditions . inside @xmath257 , @xmath255 , @xmath258 , and the total electric and magnetic fields @xmath259 and @xmath260 are related by maxwell s equations . to understand how these relations are affected by the screening charges lying on @xmath238 , we introduce the hodge decomposition of a generic field @xmath261 as @xmath262 , where @xmath263 has a finite divergence but no rotational , @xmath264 has a finite rotational but no divergence , and @xmath265 has none@xcite . a similar decomposition was invoked very recently in the context of cavity qed @xcite . however , in this ref . , the effects of the dc voltage biases @xmath245 or the floating charges @xmath243 are not considered ( i.e. @xmath246 is used for any @xmath239 ) . besides , the volume @xmath257 outside the cavity conductors is assumed to be bounded . this hypothesis is not possible in our case due to openings in the cavity planes , which are necessary to connect the nanocircuits to their outside voltage bias circuitry . hence , we need to perform the hodge decomposition in the fully general case . below , we explain how to perform this task . it is convenient to introduce a potential @xmath11 and a green s function @xmath266 which fulfill the static laplace equations:@xmath267 and @xmath268 for @xmath269 , with boundary conditions@xmath270@xmath271 and @xmath272@xmath273 above , @xmath274 is the outward pointing unit vector perpendicular to @xmath240 . one can check that each of these two sets of eqs . has a unique solution provided @xmath275 is not empty@xcite . * * from eqs . ( [ bbc0]-[bc2 ] ) and the maxwell eqs . , one can check that:@xmath276@xmath277 and@xmath278 while @xmath279 is still valid . we also find @xmath280 with@xmath281 and@xmath282 the harmonic component @xmath283 represents a static contribution to the electric field , set by the charges @xmath243 or potentials @xmath245 imposed on the conductors @xmath239 . thanks to the use of @xmath266 , the definitions ( [ d1]-[d4 ] ) take into account the effect of the screening charges lying on @xmath238 . using the maxwell equations and the expressions ( [ d0]-[d4 ] ) , one can check that the longitudinal components @xmath284 and @xmath285 are set instantaneously by @xmath255 since @xmath286 for @xmath287 . furthermore , @xmath288 , @xmath253 and @xmath289 follow an independent system of propagation equations , i.e. @xmath290@xmath291 for @xmath287 . the above equations have to be supplemented with the boundary conditions@xmath292@xmath293 with @xmath294 any trajectory connecting conductors @xmath239 and @xmath46 . at this stage , it is convenient to introduce scalar and vector potentials @xmath295 and @xmath296 in the coulomb gauge i.e. @xmath297 . we also define the decomposition @xmath298 , such that@xmath299@xmath300 and@xmath301 a comparison between eqs . ( [ bc0]-[bc2 ] ) , and ( [ b1],[bb1],[bb2 ] ) leads to the identification @xmath302 up to a global constant which one can disregard . hence , @xmath284 but also @xmath303 are instantaneously set by the charge distribution @xmath255 . in contrast , the transverse fields @xmath253 and @xmath304 are determined by @xmath296 which follows the propagation equation @xmath305 given by eq.([b4 ] ) , with @xmath306 , and boundary conditions similar to eqs.([bb1])-([bb2 ] ) . in this framework , a proper classical lagrangian is @xmath307 with @xmath308 in the sums . the lagrangian equations deriving from @xmath309 correspond to eq.([field ] ) and to the standard newton equations for the motion of particles @xmath176 . from eq . ( [ ll ] ) , one can check that @xmath310 is the conjugate of @xmath311 and @xmath312 is the conjugate of @xmath313 . therefore one can express the hamiltonian of the system as @xmath314 . this hamiltonian can be quantized straightforwardly by replacing @xmath315 , @xmath311 , @xmath316 and @xmath317 by operators @xmath318 , @xmath319 , @xmath320 and @xmath321 . this gives@xmath322 in principle , the ensemble of the charges @xmath308 includes tunneling electrons ( @xmath323 ) but also ions and valence electrons from an underlying crystalline structure . in the main text , the two latter are assumed to be decoupled from the cavity and are thus treated in a mean field approach . note that the harmonic potential term in the above hamiltonian is omitted by ref . @xcite . in our case , this term is crucial to account for the effect of dc electrostatic gates . assuming that the coupling between @xmath238 and @xmath237 is perturbative , the fields @xmath321 and @xmath324 can be expressed in terms of photonic modes calculated in the absence of @xmath325 , using for instance eq.([field ] ) with @xmath326 we treat explicitly only one of these modes in eq.([6 ] ) . screening charges on the nanocircuits reservoirs can strongly modify @xmath321 . this is why , in the general case , it is not adequate to make a perturbative treatment between the photonic field generated by the empty cavity and the whole nanocircuit . instead , the present work considers a perturbative treatment between the photonic field generated by the plasmonic modes in the whole mesoscopic qed device , and the effective orbital nanocircuit @xmath237 . similarly , in superconducting circuit qed , the cavity modes can be strongly renormalized by the presence of a superconducting quantum bit , and perturbation schemes must therefore be defined carefully@xcite . in this appendix , the renormalization of the bare cavity modes by the nanocircuit plasmons is described formally by supplementary boundary conditions imposed by effective plasmonic reservoirs . this enables a simple mathematical justification for the form of eqs ( [ 1])-([6 ] ) . in order to get realistic estimates of the fields @xmath321 and @xmath320 , one can make a numerical microwave simulation using the real device geometry , with the nanoconductors omitted ( see main text ) . the advantages of a formalism without the @xmath16 terms have already been discussed thoroughly for cavity qed@xcite . however , since the physical ingredients and measured quantities are different in cavity and mesoscopic qed , it is useful to discuss more specifically the advantages of the photonic pseudo - potential scheme of section [ ppp ] . the first quantity which is generally measured in mesoscopic qed experiments is the cavity frequency pull for a given cavity mode , which is modified by the nanocircuits coupled to the cavity . since the minimal coupling scheme and the photonic pseudo - potential scheme are related by a unitary transformation , they must of course predict the same cavity pull . it is nevertheless instructive to check this equivalence for a simple example . here , we use a simpler approach than in section iv , because we do not need to particularize the effect of the nanocircuit reservoirs and we will not discuss the cavity damping pull . we consider a cavity coupled to a nanocircuit with unperturbed many electron states @xmath327 which satisfy the general equation @xmath328 with@xmath329 and @xmath330 . we discuss the limit where the nanocircuit and the cavity are not resonant i.e. @xmath331 for any @xmath46 and @xmath51 . at second order in the cavity / nanocircuit coupling ( i.e. @xmath332 or @xmath333 ) , an elementary perturbation theory gives the cavity pull @xmath334 with@xmath335 note that @xmath336 depends on the state @xmath337 occupied by the nanocircuit . this can be used , in principle , for a noninvasive readout of the nanocircuit state in the non - resonant regime@xcite . equation ( [ pull ] ) can be obtained by using hamiltonian ( [ htottild ] ) or equivalently hamiltonian ( [ 1 ] ) . in order to check the equivalence between the two approaches , it is necessary to invoke the completeness relation @xmath338 for the nanocircuit states . interestingly , one can check that the @xmath16 terms in eq.([1 ] ) give a contribution @xmath339 to @xmath336 , which is not negligible for @xmath340 . we conclude that in the framework of mesoscopic qed where the cavity frequency pull is a central quantity , it is particularly important to eliminate in a rigorous way the @xmath16 terms if one wants to use a linear light / matter coupling . the photonic pseudo - potential scheme completes such a task . the cavity frequency pull is not the only means to characterize the interactions between a nanocircuit and a cavity . a more general workout of mesoscopic qed devices requires the knowledge of the matrix elements between the different nanocircuit eigenstates , generated by the cavity photons . another argument in strong favor of the use of the photonic pseudo - potential scheme is the general relation @xmath341 it indicates that the single - photon coupling elements decay more quickly with @xmath342 in the photonic pseudo - potential scheme . in practice one often uses a truncated space for the nanocircuit electronic states which is difficult to handle or determine globally . this is what we do for instance in section iv , where we consider a single orbital quantum dot . equation ( [ el ] ) shows that in this case it is more accurate to use eq.([htottild ] ) to predict the cavity behavior . finally , the photonic pseudo - potential scheme involves only photon - independent superconducting gap terms ( see section [ supra ] ) . this can be another significant advantage considering the already rich structure of the nanocircuit / light coupling in the absence of superconductivity . a. blais , r .- s . huang , a. wallraff , s. m. girvin , and r. j. schoelkopf , _ cavity quantum electrodynamics for superconducting electrical circuits : an architecture for quantum computation _ , physical review a * 69 * , 062320 ( 2004 ) . j. j. viennot , m. r. delbecq , m. c. dartiailh , a. cottet , and t. kontos , _ out - of - equilibrium charge dynamics in a hybrid circuit quantum electrodynamics architecture _ , phys . b * 89 * , 165404 ( 2014 ) . strictly speaking , this assumption is not valid for non ergodic reservoirs such as integrable billiards or anderson localized systems . our theory can be applied to these cases provided one keeps the @xmath46 dependence of the coupling parameter @xmath66 . the results of section [ supra ] can be generalized straighforwardly to d - wave or p - wave superconductivity by using a non - local gap term in @xmath343 in eq.([hsupr ] ) , with a different structure in spin - space in case of p - wave superconductivity .

we present the results of time - resolved spectroscopy of 13 o - type stars in the cas ob6 stellar association . we conducted a survey for radial velocity variability in search of binary systems , which are expected to be plentiful in young ob associations . here we report the discovery of two new single - lined binaries , and we present new orbital elements for three double - lined binaries ( including one in the multiple star system hd 17505 ) . one of the double - lined systems is the eclipsing binary system dn cas , and we present a preliminary light curve analysis that yields the system inclination , masses , and radii . we compare the spectra of the single stars and the individual components of the binary stars with model synthetic spectra to estimate the stellar effective temperatures , gravities , and projected rotational velocities . we also make fits of the spectral energy distributions to derive , , and angular diameter . a distance of 1.9 kpc yields radii that are consistent with evolutionary models . we find that 7 of 14 systems with spectroscopic data are probable binaries , consistent with the high binary frequency found for other massive stars in clusters and associations .

astro-ph0512407

the frequency of spectroscopic binaries ( sbs ) in star forming regions has been a topic of interest for many years . @xcite review the fraction of binary systems found among o stars in clusters / associations , the field , and runaway systems . the results from their large survey agree with previous studies ( e.g. * ? ? ? * ; * ? ? ? * ) and indicate that a large fraction of o stars in clusters and associations are binaries while fewer field stars show multiplicity and very few runaway stars have companions . the values from @xcite are broken down into spectroscopic and resolved binaries . for o stars in clusters and associations they give a total of 75% as binaries ; 61% of the binaries are spectroscopic and 41% are resolved ( with some obvious overlap ) . the spectroscopic binary group includes , however , a significant number of cases labeled as sb1 ? or sb2 ? ( systems suspected of radial velocity variability or showing double lines ) , so the fraction of confirmed spectroscopic binaries is then 34% among o stars in clusters and associations . we have conducted a radial velocity survey of o - type stars in the cas ob6 stellar association in search of spectroscopic binary systems . the cas ob6 region is one of the largest complexes in the perseus arm of the galaxy and is located at a distance of about 2.3 kpc @xcite . the warm dust and ionized gas in this region extends over 150 pc along the arm and is dominated by the w3/w4/w5 chain of regions and the hb3 supernova remnant @xcite . the open cluster ic 1805 ( ocl 352 = melotte 15 ) is situated at the western end in the w4 region and contains 8 o - type stars @xcite . the three brightest of these , hd 15558 , 15570 , and 15629 , are probably the ionization source that powers the blowout of gas in the galactic chimney @xcite and the emission arc above w4 @xcite . the open cluster ic 1848 ( ocl 364 ) is found in the eastern part of the association in the w5 region . its primary source of ionizing flux is the multiple o - star system hd 17505 @xcite . in addition to these two clusters , @xcite find evidence of 19 other young clusters that are still embedded in their primordial clouds . our knowledge of the binary star population among the young , massive stars of cas ob6 is quite limited at present . there are a number of radial velocity investigations of the sample but the overall small number of measurements has made the detection of binary stars difficult @xcite . spectroscopic orbits are available for only three systems , hd 15558 @xcite , hd 16429 @xcite , and bd@xmath2 @xcite , while a fourth star , bd@xmath3 = dn cas , is a known eclipsing binary @xcite . however , results for o - stars in other clusters and associations suggest that many more binaries may exist @xcite . it is important to determine the extent of the binary star population in cas ob6 in order to estimate accurately the ionizing flux of the key o - type stars and to assess the role of gravitational encounters with binary stars that might lead to ejection of runaway stars @xcite . here we present a reconnaissance of the radial velocity variations observed in 13 bright stars in cas ob6 that we obtained over a 5 night run with the kpno 4 m mayall telescope in 2003 october . we also present the first radial velocity study of the eclipsing binary system dn cas , which we combine with published photometry to determine the component masses . we discuss our observations and reduction techniques in 2 . the radial velocity measurements and the derivation of orbital elements are summarized in 3 . we use a tomography algorithm to separate the spectra of the binary star components , and we describe a comparison of the stellar spectra with model spectra in 4 that we use to estimate the basic stellar parameters . we examine in 5 the spectral energy distributions to determine reddening , ratio of total - to - selective extinction , and angular size for each target , and we discuss the probable distance to cas ob6 . details about the individual objects follow in 6 . finally we summarize in 7 our derived fraction of multiple and single systems for cas ob6 in the context of past results . the blue spectra were obtained with the rc spectrograph at the mayall 4 m telescope at kitt peak national observatory from 2003 september 30 through 2003 october 4 . the spectra were made with the bl 380 grating ( 1200 grooves mm@xmath4 , blazed at 4500 in second order ) using a bg39 order sorting filter . the detector was the t2 kb ccd , a @xmath5 device with 24 @xmath6 m square pixels . the spectra recorded wavelengths between 4180 and 4940 , although internal vignetting occurs for wavelengths red - ward of about 4800 . the reciprocal dispersion of the spectra is approximately 0.37 pixel@xmath4 , and the resolving power is @xmath7 . exposure times were usually a few minutes in length , and each stellar spectrum was preceded and followed by a he - ne - ar lamp exposure for careful wavelength calibration . most of the resulting spectra have a signal - to - noise ratio in excess of 100 pixel@xmath4 . we usually obtained at least two spectra of each target on each of five consecutive nights . we also made numerous bias and flat field frames each night . a list of the names of targets appears in table 1 . we also obtained an additional 32 spectra of hd 17505a with the kpno 0.9 m coude feed telescope in runs in 2000 october and 2000 december @xcite . the spectra were made with the long collimator , grating b ( in second order with order sorting filter og550 ) , camera 5 , and a ford @xmath8 ccd ( f3 kb ) detector . this arrangement produced a resolving power , @xmath9 , and covered a range of 829 around h@xmath10 . exposure times were 10 15 minutes , and the resulting spectra have a s / n = 200 in the continuum . we also observed the rapidly rotating a - type star , @xmath11 aql , which we used for removal of atmospheric water vapor lines . each set of observations was accompanied by numerous bias , flat field , and th ar comparison lamp calibration frames . the spectra were reduced using standard routines in iraf to create continuum rectified versions of each spectrum . then all the spectra of a given target were transformed to a common heliocentric frame in @xmath12 form and collected in a spectrum matrix of wavelength and time for subsequent analysis . the rectification of the vignetted portion of the spectrum was generally problem - free , so we used the entire spectral range in the following analysis . we found that the spectra of 10 of the 13 targets were composed of one set of spectral lines , and for these targets we obtained radial velocities by finding the extremum point of the cross - correlation of each spectrum with a spectrum template ( omitting those spectral regions containing interstellar features ) . we initially obtained velocities using the first spectrum as the template , and then we shifted each spectrum to the reference frame of the first spectrum and co - added all the spectra to form a high s / n template . the final relative velocities were then found by cross - correlation with this high s / n template spectrum . these relative velocities were then transformed to absolute velocity by adding the velocity measured in the template spectrum by fitting parabolae to the lower portions of selected spectral lines . rest wavelengths were taken from the list of @xcite . the final velocities for these 10 stars are listed in table 2 ( which appears only in the electronic version ) , and the mean velocity @xmath13 for each star is given in column 3 of table 1 . note that the adopted absolute velocity of the template spectrum depends critically on what subset of lines is measured because of the presence of significant line - to - line velocity differences . these differences mainly result from line formation at differing heights in an expanding stellar atmosphere @xcite . we generally selected high excitation lines that form deep in the atmosphere where the outflow velocity is smallest ( see discussion of the individual targets in 6 ) . we list in column 4 of table 1 @xmath14 , the standard deviation of the measured velocity among different lines in the template spectrum . we used a version of the statistical @xmath15-test to determine if a star was a velocity variable object . since we usually obtained at least two spectra every night of each object and since we do not expect large variations in velocity over the time span of a few hours , we estimated the average measurement error from the mean of the absolute value of the differences within a night divided by @xmath16 . this spectrum - to - spectrum error estimate @xmath17 is given in column 5 of table 1 . we then compared this to the overall standard deviation of the velocities @xmath18 ( column 6 of table 1 ) , and set the criterion for variability as @xmath19 , which corresponds to the @xmath20 probability that both variances are drawn from the same population ( for 9 degrees of freedom = 10 observations minus 1 mean value ) . we found that three of the ten single - lined targets were velocity variable according to this criterion ( see the final column of table 1 ) . the stars hd 17520 and bd@xmath21 are probably long period binary systems while the very luminous star hd 15570 exhibits variations related to atmospheric structure fluctuations ( see the discussion in 6 ) . the spectra of the targets dn cas and bd@xmath2 both showed evidence of line - doubling in some observations , and the spectra of hd 17505a displayed three components in the best separated cases . we were able to use a tomography algorithm to isolate the stationary line component in hd 17505a ( 6.5 ) , and after removal of the stationary component the spectra were treated in the same way as for the other double - lined systems . we measured velocities for these double - lined binaries in a two - step process @xcite . preliminary radial velocities were found using the multiple gaussian fitting technique within the iraf splot routine . then preliminary orbital elements were determined that provided starting estimates of the velocities for both components in each spectrum . we then formed template spectra for both components by interpolating in the o - star synthetic spectrum grid of @xcite for a temperature and gravity appropriate for each star , and these were convolved with broadening functions to account for rotational and instrumental broadening ( 4 ) . then we obtained velocities for specific spectral lines by a non - linear least squares fit of the observed profile with the two template spectra profiles co - added with a pre - defined monochromatic flux ratio ( details are given for each case in 6 ) . we took care to form an average velocity for each spectrum based on the same set of selected lines in order to avoid spurious variations introduced by line - to - line velocity differences . the resulting velocities are given below in 6 ( tables 4 , 6 , and 8) . we determined orbital elements for these three double - lined binaries using the non - linear , least - squares fitting routine of @xcite . the starting value for the orbital period was obtained with the period search technique of @xcite as modified by @xcite ( or from photometry in the case of dn cas ) . we first solved for the best fit period and then we formed separate elliptical solutions for both the primary and secondary components . the derived eccentricity in the case of dn cas was not significantly different from zero @xcite , but in the other two targets the eccentricity was significant and the same within errors in the solutions for the primary and secondary components . we formed average values of the eccentricity @xmath22 , longitude of periastron @xmath23 , and time of periastron @xmath24 that were then fixed in solutions to obtain the semiamplitude @xmath25 and systemic velocity @xmath26 for the primary and secondary components individually ( since differing atmospheric expansion rates may lead to different systemic velocities ) . the final orbital elements are given in 6 ( tables 5 , 7 , and 9 ) . we co - added our spectra to form high s / n versions for use in spectral classification , temperature and gravity estimation , and measurement of the projected rotational velocity @xmath27 . we used a simple shift - and - add algorithm to form an average spectrum in the case of the single - lined targets . however , we relied upon the doppler tomography algorithm described by @xcite to reconstruct the individual component spectra of the multiple component systems . the tomography algorithm is an iterative scheme that uses the derived orbital velocity shifts and monochromatic flux ratio to estimate the spectra of each component . we ran the algorithm for 50 iterations with a gain of 0.9 , although the results are insensitive to both parameters . the final , continuum rectified spectra of all our targets ( except hd 17505 ; see 6.5 ) and all their components are shown in figures 1 and 2 in order of decreasing effective temperature . we estimated the spectral classification of each target ( table 1 ) by comparing their spectra with standard star spectra in the atlas of @xcite . the corresponding stellar effective temperature @xmath28 was determined by comparing the equivalent widths of various / line ratios with those measured in the synthetic spectral library for o - stars by @xcite . in most cases , we simply assumed a surface gravity of @xmath29 as a suitable value for main sequence stars , but this approach was inadequate in several cases of evolved stars where the h balmer line wings are narrower and indicative of a lower gravity . we experimented with a number of model synthetic spectra to best estimate the gravity in these cases . this was accomplished through comparing models by eye with the data in order to find an appropriate match for surface gravity . the most difficult spectrum to match was that of the very luminous star hd 15570 , which displays clear evidence of the atmospheric expansion into a stellar wind that is not included in the static atmosphere formulation adopted by @xcite . we estimate that our fitting errors are approximately @xmath30 k in temperature and @xmath31 dex in @xmath32 , and the results are listed in table 1 . we estimated the projected rotational velocity @xmath27 by comparing the observed and line widths with those found by convolving the model profiles from @xcite with instrumental and rotational broadening functions . the instrumental broadening of our spectra was represented by a gaussian with @xmath33 km s@xmath4 . the rotational broadening function @xcite was computed assuming a spherical star with a linear limb - darkening coefficient determined from @xcite . a grid of theoretical profiles was computed for a range of @xmath27 , and a @xmath34 minimization was used to determine the best fit to the data . the results and errors are listed in table 1 . note that this approach does not account for variations in the intensity profile across the disk , nor the non - spherical shape and gravity darkening found in rapid rotators @xcite . these shortcomings are most important for the two rapid rotators in our sample , bd@xmath35 and bd@xmath21 , and we have probably underestimated the true rotational broadening in these cases . all of our targets are considered to be members of the cas ob6 association based upon their position in the sky and placement in the hertzsprung - russell diagram ( hrd ) @xcite . the proper motion study of ic 1805 made by @xcite confirms membership for five targets but curiously indicates a somewhat discrepant proper motion for hd 15558 . we will assume that all the targets share a common distance , so their observed fluxes are directly related to their relative stellar radii . since we have determined temperature and multiplicity for each target , we have all parameters needed to estimate their radii as a function of distance by comparing their observed and model flux distributions , and we can then check the distance by comparing the derived and model radii ( as a function of temperature ) . the angular diameter of the limb darkened disk @xmath36 ( in units of radians ) is found by the inverse - square law : @xmath37 where the ratio of the observed and emitted fluxes ( reduced by the effects of interstellar extinction @xmath38 ) depends on the square of the ratio of stellar radius @xmath39 to distance @xmath40 . all our targets have a large reddening , so we need to fit the relation above over the entire observed spectrum in order to account reliably for the wavelength dependent shape of the extinction curve . we adopted the extinction curve law from @xcite that is a function of the reddening @xmath0 and the ratio of total - to - selective extinction @xmath1 . we then fit the observed flux distribution of each target to determine @xmath0 , @xmath41 , and angular diameter @xmath36 . we compiled data on the observed fluxes from the uv to the near - ir . there are low dispersion _ international ultraviolet explorer ( iue ) _ spectra available ( swp and lwr cameras ) for all of the targets except dn cas , hd 17520 , hd 237019 , and bd@xmath42 , and we formed binned versions of the uv spectral fluxes from these . we found johnson @xmath43 photometry for all the targets ( plus cousins @xmath44 in a few cases ) , and these magnitudes were transformed to fluxes using the calibration of @xcite . we also included strmgren photometry for 7 of the 13 targets using the calibration of @xcite . all of the stars are included in the _ 2mass all - sky catalog of point sources _ @xcite , and we converted the @xmath45 magnitudes to fluxes using the calibration of @xcite . the model fluxes were selected from the grid of @xcite based upon the values of @xmath28 and @xmath32 in table 1 . these model fluxes are based upon plane - parallel , non - lte , line blanketed atmospheres that do not include the effects of stellar wind mass loss . however , the inclusion of winds generally alters the spectral energy distributions only at wavelengths much larger and much smaller than those considered here @xcite , so wind effects are insignificant in this context . we co - added model fluxes using the observed flux ratios for the spectroscopic binaries and using the observed magnitude difference for close visual binaries ( details are given in next section for individual stars ) . note that in the two single - lined binary cases , hd 17520 and bd@xmath21 , we simply assumed that the unseen companion contributes @xmath46 of the total flux ( a compromise between the expectation that the mass ratio is not too extreme and that brighter companions would have visible spectral features in our observations ) . we have no spectral information for some of the companions , and in these cases we assumed that the companion was a main sequence star with a temperature corresponding to the observed flux ratio ( from the calibration in table 4 of @xcite ) . we set a lower limit of @xmath47 k and @xmath48 for the faint companions , which corresponds to the cool boundary in the grid of @xcite . this simplification is acceptable since any detected companion is probably hot enough that the optical and near - ir regions correspond to the common rayleigh - jeans portion of the flux distribution . we compared the observed and model flux distributions and made a grid search fit to find the values of @xmath0 and @xmath41 that produced the lowest @xmath49 residuals ( by weighting the observations approximately equally between the uv , optical , and near - ir bands ) . the normalization term in the final fit yields the angular diameter . our results are gathered in table [ sedfits ] which lists the star name , home cluster , angular diameter ( in micro - arcsec ) , derived radius ( see below ) , @xmath0 , and @xmath41 . the acccompaning error estimates for @xmath0 and @xmath41 are derived from the point in grid search where @xmath49 attains a value higher by 1 than normalized minimum value , and the errors in diameter are based upon the scatter in flux distribution fit and the errors propagated from uncertainty in @xmath28 , @xmath0 , and @xmath41 . the model fits were generally excellent , but they were less satisfactory for the spectra of the three hottest stars , where it was difficult to find a consistent match of the far- and near - uv fluxes . we also found the @xmath45 fluxes were too large relative to the optical fluxes for hd 17520 , which we suspect is due to an ir excess associated with the be disk emission of the companion @xcite , so the fit was based on the optical fluxes alone in this case . the stellar radius is found by @xmath50 where the distance @xmath40 is given in kpc . the distance estimates for ic 1805 and ic 1848 range from 1.9 kpc @xcite to 2.4 kpc @xcite ( with a scatter of @xmath51 among recent estimates ) . we selected a distance of @xmath52 kpc from the compendium of open cluster data by @xcite because this low end value appears to give the best match between the observed and theoretical stellar radii and provides good agreement with radii of similar stars in eclipsing binaries @xcite . the resulting radii are listed in table [ sedfits ] and are plotted against stellar temperature in figure [ figradt ] . there is a clear and consistent trend for the hotter main sequence stars to be larger , and the observations match the predictions for main sequence stars from the atmosphere models of @xcite ( shown as a dotted line in fig . [ figradt ] ) . we also show in figure [ figradt ] the ( @xmath53 ) loci for stars with ages of 1.00 and 3.16 myr from the evolutionary models of @xcite , which correspond to the age range for ic 1805 given by @xcite . we see that most of the stars fall within the expected region in this diagram for this age range . however , the positions of several stars indicate that there is a real dispersion in the ages among this sample . for example , bd@xmath54 appears to have a relatively small radius for its temperature , indicating it is the youngest object in our sample . the derived stellar radii are directly proportional to the assumed distance , and , for example , if we adopted the larger distance @xmath55 kpc advocated by @xcite then the stars would be approximately @xmath56 larger , a size that would be difficult to reconcile with the evolutionary models of @xcite . we show below that our derived radii for dn cas agree well with those determined independently from an analysis of the eclipsing light curve . bd@xmath54. the two balmer lines h@xmath26 and h@xmath57 have the lowest measured velocities in the template spectrum , so they were omitted from the average ( i.e. , based only on the and lines ) . our observed mean velocity of @xmath58 km s@xmath4 is reasonably consistent with the mean from the work of @xcite of @xmath59 km s@xmath4 who also find the star to be a constant velocity object . hd 15558. there is evidence here for a velocity progression ( lower velocity among low excitation lines ) , so we used only the lines @xmath60 to set the radial velocity of the template spectrum . there are no obvious variations over the 5 night run , but the star may be a long period , single - lined binary @xcite . according to the recent orbital elements from @xcite , the orbital period is @xmath61 d and the predicted velocity at the mean time of our observations is @xmath62 km s@xmath4 ( near the peak of the radial velocity curve ) , which is 27.7 km s@xmath4 higher than our observed mean of @xmath63 km s@xmath4 . this discrepancy could be caused by differences in measurement techniques or errors in the preliminary orbital solution of @xcite . a long - term study by m. de becker and g. rauw ( 2005 , private communication ) has recently confirmed the binary status of this star and will provide an improved orbital solution . therefore , we list this star as `` constant ( sb1 ) '' in table [ summary ] denoting both a lack of short term line variability in our observations as well as the recent confirmation of binarity . there is a visual companion at a separation of @xmath64 that is 2.6 mag fainter @xcite . hd 15570. there is a clear velocity progression between spectral lines with the low excitation lines ( h , ) appearing at a much lower velocity than the high excitation lines ( , ) . there is also a correlation with line width in the sense that stronger , broader lines have lower velocity . we adopted a radial velocity for the template spectrum from the lines @xmath65 and @xmath66 . the night - to - night scatter between velocities is small except on the last night when the measured velocity is @xmath67 km s@xmath4 more positive than the mean from the other nights . the of emission lines do not show any shift in velocity on the final night . we suspect that the change was related to an atmospheric fluctuation rather than orbital motion . @xcite also consider this star to be a velocity variable . @xcite find no evidence of orbital motion and discuss rotational modulation as a possible source of line variability . hd 15629. the low excitation balmer and lines have systematically low velocities , so the template mean was adopted from the lines @xmath68 . we find no evidence of velocity variability . @xcite , on the other hand , flag this star as possibly variable , but given the presence of line - to - line velocity differences , some of the velocity differences found by various authors are probably due to the line sample adopted ( and not to orbital motion ) . @xcite also find no evidence of orbital motion for this star . bd@xmath69. this star has a broad - lined spectrum , and we selected only the strongest lines to measure the template mean velocity ( h@xmath26 , h@xmath57 , @xmath70 , @xmath71 ) . we confirm the lack of short term velocity variability reported by @xcite , and our mean velocity of @xmath72 km s@xmath4 is probably consistent with their mean value , @xmath73 km s@xmath4 . bd@xmath74. the h@xmath57 line had the lowest velocity measured , so we omitted the h balmer lines and weak lines in forming the template mean velocity . the star appears to be radial velocity constant over the run , and our mean velocity , @xmath75 km s@xmath4 , is probably consistent with that from @xcite of @xmath76 km s@xmath4 . hd 237019. all the strong lines were used to form the template mean velocity . the star has a constant radial velocity over the run . our mean velocity of @xmath77 km s@xmath4 is higher than the only other value in the literature of @xmath78 km s@xmath4 from @xcite . bd@xmath42. we omitted h@xmath57 and @xmath79 from the template mean calculation , since both lines had significantly lower velocities . the star was radial velocity constant during the run . our mean velocity , @xmath80 km s@xmath4 , agrees with that found by @xcite , @xmath81 km s@xmath4 , but is lower than the value reported by @xcite , @xmath82 km s@xmath4 . the star is a member of a visual multiple system ( ads 2194 ) with companions at separations of @xmath83 ( @xmath84 ) and @xmath85 ( @xmath86 ) . hd 17520. the h@xmath57 line has a blue - shifted emission component which also appears weakly in h@xmath26 . the balmer lines and weak lines were omitted in the calculation of the template mean velocity . the velocities show a progressive increase over the duration of the run , reaching values similar to those observed by @xcite and @xcite . @xcite showed how h@xmath10 emission developed between 1985 and 1991 leading to a spectral classification as a be - type star ( probably its status in 2003 as well since the emission is so apparent in h@xmath57 ) . this star is also the central object in a visual multiple system ( ads 2165 ) with a close companion at a separation of @xmath87 and with a magnitude difference of 0.52 mag @xcite . our spectra show evidence of only one set of lines ( for the o9 v component ) , and unfortunately , we do not know if the be emission originates in the o - star or its nearby companion ( which could have broad shallow lines as observed in many be stars ) . we note that over the course of the run the h@xmath57 emission appeared to be almost stationary while the absorption lines shifted red - ward by @xmath88 km s@xmath4 . the simplest interpretation is that the o - star component is a long - period , single - lined binary , while the be emission originates in the fainter component of the close visual pair . there is another star at a separation of @xmath89 that is 3.2 mag fainter in @xmath25 @xcite . bd@xmath21. we used the strong lines to set the template mean velocity for the broad - lined spectrum of this star . the first measured velocity is close to the value obtained by @xcite , @xmath90 km s@xmath4 , but the subsequent velocities decline monotonically by @xmath91 km s@xmath4 over the next 4 days . this suggests that the star is a single - lined binary with a period of @xmath92 d and a semiamplitude of @xmath93 km s@xmath4 . dn cas has been known as an eclipsing binary for some time and was classified as an algol type binary with a period of 1.155479 d by @xcite . @xcite classified the spectrum as o8 v. a later photometric study by @xcite used @xmath43 photoelectric photometry to refine the orbital period to 2.310955 d , twice that of the original value . unfortunately , the comparison star used by @xcite turned out to be itself an eclipsing binary , compromising their final photometry . @xcite made another photometric study , and based upon the curvature near the maxima of the light curve he reclassified dn cas as a @xmath57 lyrae type system . interestingly no spectroscopic observations of dn cas have appeared in the literature , though @xcite claimed that such observations were underway . additionally , the only photometry that has been published , other than in phase - folded figures , is the compromised data of @xcite . this overlooked opportunity to obtain masses for this massive binary is rectified here , where we present the first full orbital solution based on our spectroscopy and photometry from the literature . we determined velocities for both components using non - linear least - squares fits of two templates with temperatures and rotational velocities given in table [ summary ] and with a monochromatic flux ratio in the blue of @xmath94 . our final radial velocities are given in table [ dncasrv ] and the spectroscopic orbital elements are given in table [ dncasorb ] . the period was adopted from the eclipse timing observations collected by @xcite . here we adopt the time of minimum light ( primary superior conjunction ) to set the zero - point of orbital phase and epoch @xmath95 , and our derived epoch agrees with the current photometric epoch of minimum light from @xcite of hjd 2,452,501.928 @xmath96 . eccentric fits offered no significant improvement over circular fits @xcite , so circular elements are given in table 5 . the radial velocities and orbital solutions are shown in figure 4 . the only published photometric data of dn cas are those from @xcite who made @xmath43 photoelectric observations between 1971 october and 1972 december . unfortunately , their comparison star , hd 14817 , is itself an eclipsing binary as mentioned above . however , the light curve of hd 14817 from _ hipparcos _ @xcite is flat outside of eclipse , so we can retain the results from @xcite that are based on outside - of - eclipse times for hd 14817 . these measurements are plotted against orbital phase in figure 5 , and because there remain significant phase gaps in this restricted set , our light curve fitting results should be considered preliminary . we calculated a model light curve using the wilson - devinney code @xcite . most of the model parameters are set by the spectroscopic orbital solution ( table [ dncasorb ] ) , the adopted temperatures and gravities from the spectra ( table [ summary ] ) , the associated limb darkening parameters @xcite , and the monochromatic flux ratio ( which sets the ratio of the mean stellar radii ) . in addition , we set the bolometric albedo and gravity darkening exponents to 1.0 , which are valid approximations for stars with fully radiative envelopes . we then fit the observed @xmath97-band observations by varying the inclination @xmath98 and the stellar radii . the ratio of stellar radii was set using the observed monochromatic flux ratio and model surface flux ratio ( set according to the temperatures and gravities in table 1 ) @xmath99^{0.5 } = 0.72 \pm 0.09.\ ] ] if the stars are rotating synchronously , then the ratio of projected rotational velocities also yields the ratio of radii as @xmath100 and the good agreement between these two estimates suggests that synchronous rotation has been attained in this close binary . we found that the radii obtained from a fit of the spectral energy distribution and distance of 1.9 kpc ( table [ sedfits ] ) led to an excellent fit of the light curve for an orbital inclination of @xmath101 ( fig . 5 ) , and the derived stellar masses are @xmath102 for the o8 v primary and @xmath103 for the b0.2 v secondary ( table 5 ) . the mass of the primary is consistent with its spectral type for the calibration given by @xcite , and the mass of the secondary agrees well with the mass for its spectral type according to the relations given by @xcite . both stars are well within their respective roche lobes . there is a companion to dn cas in the 2mass catalogue at a separation of @xmath104 and at a position angle of @xmath105 . however , inspection of the 2mass images at this position suggests that this object is probably not a real star since its point spread function looks very sharp and unlike those of nearby stars . furthermore , we see no evidence of such a tertiary in our observations . since the near east - west alignment of the putative object and dn cas is the same as the slit orientation for our kpno 4 m spectra , the tertiary should appear in our spectra with a separation from dn cas of 7.7 pixels in the spatial dimension . the average 2mass magnitude difference is 2.13 mag ( for @xmath106 ) , so that the flux ratio of tertiary to binary is 0.14 in the near - ir . however , our blue spectra show no sign of a companion brighter than @xmath107 , so if this companion is real , it is probably a very red foreground star that can be ignored in our analysis . we found that the introduction of a third light contribution of @xmath108 in the @xmath97-band light curve results in an increase in the inclination by @xmath109 , but since the actual flux contribution of a tertiary is probably much smaller than this , any changes in our results due to the presence of a tertiary are probably comparable to or less than the errors quoted in table 5 . the spectra of bd+@xmath111497 show clear evidence that this star is a double - lined spectroscopic binary . in their recent paper , @xcite give the first orbital solution for bd+@xmath111497 , and they find an orbital period of 3.96 days . we measured velocities using the scheme outlined in 3 by matching template synthetic profiles for models with temperatures and gravities given in table 1 , together with an estimated flux ratio of @xmath112 . we note , however , that in some of the lines ( for example , @xmath113 ) we found that the approaching , blue - shifted component ( primary and secondary ) often appeared to be weaker than expected based upon this fixed flux ratio ( @xcite also observed a similar variation ) . we doubt that this variation has a significant effect on our measurements , which appear in table 6 . we combined our velocities with those of @xcite in making the orbital solution . a number of alias periods remain viable , but we obtained the best fit with a period close to that obtained by @xcite ( see table 7 ) . however , unlike @xcite , we found that a significant eccentricity is present , and we made full elliptical fits for both components . otherwise our results are in reasonable agreement with their solution ( compared in table 7 ) . the velocities and orbital fit are shown folded on the derived ephemeris in figure 6 . assuming masses typical of their spectral classes @xcite , the system probably has an inclination of @xmath114 , so we do not expect eclipses to occur . the star system hd 17505 dominates the center of the cluster ic 1848 , and there are seven visual companions noted by @xcite at distances ranging from @xmath115 to @xmath116 . multiplicity in the central object hd 17505a has been discussed by @xcite , @xcite , @xcite , and @xcite and it is known to contain at least three stars since the single _ international ultraviolet explorer _ spectrum shows evidence of three doppler shifted components @xcite . @xcite pointed out that the multiplicity of the star is the reason it appears so bright in the hrd . we obtained 10 spectra of hd 17505a over the five night duration of the kpno 4 m run , and we found that the lines were strongly blended in all but a few spectra corresponding to the maximum doppler separation of the binary motion . however , we also obtained an additional 32 red spectra of the target in 2000 with the kpno 0.9 m coude feed telescope ( 3 ) , and these turned out to be particularly helpful in finding the orbital elements . initial inspection of these higher dispersion red spectra showed that all three components were cleanly resolved in the profiles observed near orbital quadrature phases and indicated an orbital period of approximately 8 days . we measured radial velocities by first making unconstrained three - gaussian fits of the least blended feature , @xmath117 , in the quadrature phase spectra . we obtained a preliminary orbital solution from this subset of spectra for the two moving components , and then we used this solution to estimate the doppler shifts for all the coude feed observations . then we made three - component gaussian fits of @xmath117 in all the red spectra using these starting estimates for the doppler shifts and constraining the widths to be the mean values found from the fits of the quadrature phase spectra . a new orbital solution was obtained , and the solution was used to make a three component reconstruction of the individual component spectra using the doppler tomography algorithm @xcite . we subtracted the reconstructed spectrum of the stationary component from each spectrum and renormalized them to end up with a set corresponding to the spectra of the double - lined binary alone . we then measured the radial velocities using the non - linear , least - squares method by matching template spectra to the profiles of h@xmath10 , @xmath118/ @xmath119 , and @xmath117 . the template spectra for the two components were taken from their respective tomographic reconstructions ( @xmath120 ) rather than using the model profiles from @xcite , because we were concerned that some of these red features are not well represented by the synthetic models . this approach worked well except in one conjunction phase observation ( from hjd 2,451,892.743 ) where the lines were too badly blended for reliable measurement . our final radial velocities from the red spectra for the close binary are collected in table [ 17505rv ] . the estimate for the orbital period from the red spectra alone was @xmath121 d , and the error was sufficiently small that we could determine the cycle number and approximate orbital phase corresponding to the times of the kpno 4 m observations in 2003 october . we measured the line width of the @xmath122 feature in these blue spectra to find that maximum velocity separation occurred at hjd 2,452,915.469 @xmath123 , and we compared this to the time of the same orbital phase for the 2000 coude feed observations to arrive at a revised period estimate of @xmath124 d. the final orbital solution is given in table 9 and the radial velocity curve appears in figure 7 . the predictions from this fit for the time of the _ iue _ observation are @xmath125 and @xmath126 km s@xmath4 , in reasonable agreement with the observed velocities of @xmath127 and @xmath128 km s@xmath4 @xcite , but we did not include the _ iue _ measurements in the final solutions because of possible systematic differences between velocities measured in the uv and optical . we made tomographic reconstructions of both the blue and red spectra using the orbital solution in table [ 17505orb ] and assuming flux contributions of 30% , 30% , and 40% for aa1 , aa2 , and ab , respectively ( based upon the relative equivalent widths of the @xmath118 @xmath129 @xmath119 feature ) . both the blue and red spectral reconstructions are plotted in figure [ 17505spec ] . we determined radial velocities of the stationary component ab of @xmath130 and @xmath131 from the blue and red reconstructed spectra , respectively . once again , the optical velocities are systematically more negative than the uv determined value for ab of @xmath132 km s@xmath4 @xcite . our results are summarized in a schematic representation of the system hierarchy in figure [ 17505mob ] . note that the coude feed spectra were made with a @xmath133 wide slit in an approximately north - south orientation , so the b component ( at a separation of @xmath134 and a position angle of @xmath135 with a magnitude difference of @xmath136 mag ; @xcite ) may be slightly mixed in with the light from the ab component for the red reconstructed spectrum . however , the 4 m spectra were made with an east - west slit orientation , so it was possible to separate the a and b components spatially along the slit . a single extracted spectrum of the b component is shown in figure [ 17505spec ] along with the tomographic reconstructions of the aa1 , aa2 , and ab components . the spectrum of the b component is clearly different and indicates a cooler effective temperature than the stars of component a , and , thus , the spectroscopic triple we detect is clearly related to a alone . the hottest and brightest of all the components is ab , and its spectral type agrees with that obtained by @xcite for the entire system . the stars of the aa1-aa2 central binary are far enough apart that no eclipses should be seen , and indeed none are observed @xcite . for example , if we adopt the masses and radii associated with o - type main sequence stars from @xcite , then the expected system inclination should be @xmath137 ( based upon the @xmath138 values in table [ 17505orb ] ) while eclipses would be observed only if @xmath139 ( based upon the semimajor axis in table [ 17505orb ] ) . the orbital period of b around the triple a can be estimated as @xmath140 y if we take the projected angular separation as the semimajor axis and the distance is 1.9 kpc . the orbital period of the aa - ab central triple is probably much less . the system was unresolved in the speckle survey of @xcite , and if we take their resolution limit of 35 mas as the projected semimajor axis then the orbital period should be @xmath141 y. the target is clearly worthy of continued spectroscopic and interferometric observation to search for evidence of such orbital motion . we have monitored thirteen o stars in the cas ob6 association in search of multiple star systems . five of the thirteen were found to have radial velocity variations related to orbital motion . of these five , we have determined orbital solutions for three double - lined systems : the eclipsing binary dn cas , bd@xmath2 , and the spectroscopic triple hd 17505a . the remaining two binaries , bd@xmath21 and hd 17520 , are single - lined binaries with periods longer than 5 days . in addition to these five , one of our targets , hd 15558 , is probably a long period binary @xcite . one other bright o - star in cas ob6 is hd 16429 , which was found by @xcite to be a spectroscopic triple . if we include all of these objects in the known sample of observed o - stars in the cas ob6 association , then we find that 7 of the 14 , or @xmath142 , are spectroscopic binaries . this value is well within the range found by @xcite for spectroscopic binaries among stars in clusters and associations . we note that two systems , hd 17505 and dn cas , are of special interest and would benefit from follow - up observations . we are planning to obtain new three - color photometry of the eclipsing binary dn cas in order to make a better fit of the light curve and to determine an accurate mass of the o8 v primary star . the multiple system hd 17505 is a fascinating system with at least four o stars ( with a total mass close to @xmath143 ) which are apparently gravitationally bound . high resolution studies via interferometry to separate the aa and ab components would produce a more accurate kinematic model for the system and would help us better understand stellar interactions in these systems and their role in the formation and ejection of massive stars . we are grateful to the director and staff of kpno for their help in making these observations possible . we also thank gregor rauw and michael de becker for sharing their results with us in advance of publication . financial support was provided by the national science foundation through grant ast@xmath1440205297 ( drg ) . institutional support has been provided from the gsu college of arts and sciences and from the research program enhancement fund of the board of regents of the university system of georgia , administered through the gsu office of the vice president for research . mvm is supported by an nsf astronomy and astrophysics postdoctoral fellowship under award ast@xmath1440401460 . this publication makes use of data products from the two micron all sky survey , which is a joint project of the university of massachusetts and the infrared processing and analysis center / california institute of technology , funded by the national aeronautics and space administration and the national science foundation . this research has also made use of the washington double star catalog maintained at the u.s . naval observatory . lccccccccc dn cas a & o8 v & @xmath146 & & 4.9 & 149.3 & 34.2 & 4.0 & @xmath147 & sb2 + dn cas b & b0.2 v & @xmath148 & & 5.3 & 207.3 & 29.7 & 4.0 & @xmath149 & sb2 + bd@xmath2 a & o6 v & @xmath150 & & 2.2 & 126.8 & 37.5 & 4.0 & @xmath151 & sb2 + bd@xmath2 b & o8 v & @xmath152 & & 3.6 & 166.7 & 33.0 & 4.0 & @xmath153 & sb2 + bd@xmath54 & o6.5 v & @xmath154 & 9.4 & 0.5 & 1.1 & 36.8 & 4.0 & @xmath155 & constant + hd 15558 & o5 iii(f ) & @xmath156 & 5.3 & 1.8 & 2.9 & 39.5 & 3.7 & @xmath157 & constant ( sb1 ) + hd 15570 & o4 if+ & @xmath158 & 0.9 & 0.8 & 4.2 & 42.9 & 3.5 & @xmath159 & atmospheric var . + hd 15629 & o5 v((f ) ) & @xmath160 & 8.1 & 1.2 & 2.2 & 41.5 & 4.0 & @xmath161 & constant + bd@xmath69 & o7.5 vn & @xmath162 & 6.7 & 1.5 & 2.8 & 34.8 & 4.0 & @xmath163 & constant + bd@xmath74 & o6.5 v((f ) ) & @xmath164 & 3.7 & 0.8 & 1.7 & 36.2 & 3.8 & @xmath165 & constant + hd 17505 aa1 & o7.5 v((f ) ) & @xmath166 & & 8.6 & 115.3 & 36.3 & 4.0 & @xmath167 & sb2 + hd 17505 aa2 & o7.5 v((f ) ) & @xmath168 & & 7.2 & 118.2 & 36.5 & 4.0 & @xmath169 & sb2 + hd 17505 ab & o6.5 iii((f ) ) & @xmath148 & 8.2 & & & 37.5 & 3.8 & @xmath170 & constant + hd 17520 & o9 v & @xmath171 & 2.7 & 0.7 & 6.0 & 33.2 & 4.0 & @xmath172 & sb1 + be + hd 237019 & o8 v & @xmath150 & 8.1 & 0.7 & 1.1 & 34.0 & 4.0 & @xmath173 & constant + bd@xmath42 & o7.5 v & @xmath174 & 4.7 & 1.0 & 1.0 & 35.4 & 4.0 & @xmath175 & constant + bd@xmath21 & o8.5 vn & @xmath176 & 8.3 & 1.0 & 17.7 & 33.5 & 4.0 & @xmath177 & sb1 + lccccc dn cas a & ic 1805 & 36 ( 1 ) & 7.4 ( 7 ) & 0.97 ( 1 ) & 3.13 ( 1 ) + dn cas b & ic 1805 & 26 ( 3 ) & 5.3 ( 8) & 0.97 ( 1 ) & 3.13 ( 1 ) + bd@xmath2 a & ic 1805 & 49 ( 4 ) & 10.0 ( 13 ) & 0.89 ( 2 ) & 2.94 ( 6 ) + bd@xmath2 b & ic 1805 & 32 ( 5 ) & 6.5 ( 11 ) & 0.89 ( 2 ) & 2.94 ( 6 ) + bd@xmath54 & ic 1805 & 36 ( 3 ) & 7.2 ( 9 ) & 0.76 ( 2 ) & 3.12 ( 5 ) + hd 15558 a & ic 1805 & 81 ( 10 ) & 16.4 ( 26 ) & 0.84 ( 3 ) & 3.07 ( 9 ) + hd 15558 b & ic 1805 & 28 ( 4 ) & 5.6 ( 9 ) & 0.84 ( 3 ) & 3.07 ( 9 ) + hd 15570 & ic 1805 & 95 ( 16 ) & 19.3 ( 37 ) & 1.01 ( 2 ) & 3.13 ( 7 ) + hd 15629 & ic 1805 & 57 ( 7 ) & 11.6 ( 19 ) & 0.76 ( 3 ) & 3.09 ( 12 ) + bd@xmath69 & ic 1805 & 43 ( 4 ) & 8.7 ( 12 ) & 0.83 ( 3 ) & 3.00 ( 5 ) + bd@xmath74 & ic 1805 & 49 ( 3 ) & 9.9 ( 11 ) & 0.75 ( 2 ) & 3.02 ( 4 ) + hd 17505 ab & ic 1848 & 56 ( 6 ) & 11.4 ( 16 ) & 0.72 ( 3 ) & 2.83 ( 9 ) + hd 17505 aa1 & ic 1848 & 49 ( 8) & 10.0 ( 18 ) & 0.72 ( 3 ) & 2.83 ( 9 ) + hd 17505 aa2 & ic 1848 & 49 ( 8) & 10.0 ( 18 ) & 0.72 ( 3 ) & 2.83 ( 9 ) + hd 17505 b & ic 1848 & 44 ( 6 ) & 8.9 ( 15 ) & 0.72 ( 3 ) & 2.83 ( 9 ) + hd 17520 aa & ic 1848 & 38 ( 11 ) & 7.7 ( 24 ) & 0.57 ( 5 ) & 3.10 ( 120 ) + hd 17520 ab & ic 1848 & 14 ( 4 ) & 2.8 ( 9 ) & 0.57 ( 5 ) & 3.10 ( 120 ) + hd 17520 b & ic 1848 & 34 ( 10 ) & 7.0 ( 22 ) & 0.57 ( 5 ) & 3.10 ( 120 ) + hd 17520 c & ic 1848 & 14 ( 4 ) & 2.8 ( 9 ) & 0.57 ( 5 ) & 3.10 ( 120 ) + hd 237019 & ic 1848 & 37 ( 1 ) & 7.4 ( 7 ) & 0.77 ( 1 ) & 3.20 ( 6 ) + bd@xmath42 a & ic 1848 & 45 ( 3 ) & 9.1 ( 11 ) & 0.58 ( 3 ) & 3.27 ( 32 ) + bd@xmath42 b & ic 1848 & 25 ( 2 ) & 5.1 ( 7 ) & 0.58 ( 3 ) & 3.27 ( 32 ) + bd@xmath21 a & ic 1848 & 35 ( 4 ) & 7.2 ( 11 ) & 0.68 ( 2 ) & 2.88 ( 10 ) + bd@xmath21 b & ic 1848 & 10 ( 2 ) & 2.0 ( 5 ) & 0.68 ( 2 ) & 2.88 ( 10 ) + cccccc 2912.849 & 0.814 & @xmath183 & @xmath184 & @xmath185 & @xmath186 + 2912.932 & 0.850 & @xmath187 & @xmath188 & @xmath189 & @xmath188 + 2913.003 & 0.881 & @xmath190 & @xmath191 & @xmath192 & @xmath193 + 2913.825 & 0.237 & @xmath194 & @xmath195 & @xmath196 & @xmath197 + 2913.834 & 0.241 & @xmath198 & @xmath199 & @xmath200 & @xmath201 + 2913.839 & 0.243 & @xmath202 & @xmath203 & @xmath204 & @xmath205 + 2913.844 & 0.245 & @xmath206 & @xmath207 & @xmath208 & @xmath209 + 2913.927 & 0.281 & @xmath210 & @xmath211 & @xmath212 & @xmath213 + 2914.004 & 0.314 & @xmath214 & @xmath215 & @xmath216 & @xmath217 + 2914.814 & 0.665 & @xmath218 & @xmath219 & @xmath220 & @xmath221 + 2914.943 & 0.721 & @xmath222 & @xmath223 & @xmath224 & @xmath225 + 2915.025 & 0.756 & @xmath226 & @xmath227 & @xmath228 & @xmath229 + 2915.656 & 0.029 & @xmath230 & @xmath231 & @xmath232 & @xmath233 + 2915.883 & 0.127 & @xmath234 & @xmath235 & @xmath236 & @xmath237 + 2915.931 & 0.148 & @xmath238 & @xmath239 & @xmath240 & @xmath241 + 2916.015 & 0.184 & @xmath242 & @xmath243 & @xmath244 & @xmath245 + 2916.747 & 0.501 & @xmath246 & @xmath247 & @xmath248 & @xmath249 + 2916.756 & 0.505 & @xmath250 & @xmath251 & @xmath252 & @xmath253 + 2916.792 & 0.521 & @xmath254 & @xmath255 & @xmath256 & @xmath257 + 2916.797 & 0.523 & @xmath258 & @xmath259 & @xmath260 & @xmath261 + 2916.805 & 0.526 & @xmath262 & @xmath263 & @xmath264 & @xmath265 + 2916.809 & 0.528 & @xmath266 & @xmath267 & @xmath268 & @xmath269 + 2916.818 & 0.532 & @xmath270 & @xmath271 & @xmath272 & @xmath205 + 2916.822 & 0.534 & @xmath273 & @xmath274 & @xmath275 & @xmath276 + 2916.846 & 0.544 & @xmath277 & @xmath278 & @xmath279 & @xmath203 + 2916.851 & 0.546 & @xmath280 & @xmath281 & @xmath282 & @xmath283 + 2916.943 & 0.586 & @xmath284 & @xmath285 & @xmath286 & @xmath287 + 2917.018 & 0.618 & @xmath288 & @xmath289 & @xmath290 & @xmath291 + lr @xmath292 ( days ) & 2.310950 ( 2 ) + @xmath95 ( hjd-2,450,000 ) & 2915.579 ( 2 ) + @xmath22 & 0 + @xmath23 ( @xmath110 ) & + @xmath293 ( km s@xmath4 ) & 210.5 ( 20 ) + @xmath294 ( km s@xmath4 ) & 292.0 ( 38 ) + @xmath295 & 0.721 ( 12 ) + @xmath26 ( km s@xmath4 ) & @xmath296 ( 14 ) + @xmath26 ( km s@xmath4 ) & @xmath297 ( 27 ) + @xmath98 ( @xmath110 ) & 79.2 ( 11 ) + @xmath298 ( @xmath299 ) & 19.2 ( 8) + @xmath300 ( @xmath299 ) & 13.9 ( 5 ) + @xmath301 ( @xmath302 ) & 7.4 ( 6 ) + @xmath303 ( @xmath302 ) & 5.4 ( 8) + @xmath304 & 0.78 ( 6 ) + @xmath305 & 0.66 ( 10 ) + @xmath306 ( @xmath302 ) & 23.3 ( 2 ) + r.m.s.@xmath307 ( km s@xmath4 ) & 7.4 + r.m.s.@xmath308 ( km s@xmath4 ) & 14.1 + cccccc 2912.871 & 0.151 & @xmath309 & 0.6 & 136.3 & @xmath310 + 2912.953 & 0.171 & @xmath311 & @xmath312 & 145.4 & @xmath313 + 2913.874 & 0.404 & @xmath314 & 2.7 & 30.7 & 34.7 + 2913.946 & 0.422 & @xmath315 & 22.4 & @xmath316 & 2.7 + 2914.823 & 0.644 & 70.7 & 8.0 & @xmath317 & 2.7 + 2914.826 & 0.644 & 65.0 & 2.0 & @xmath318 & 1.7 + 2914.832 & 0.646 & 63.6 & @xmath319 & @xmath320 & 2.9 + 2914.836 & 0.647 & 60.1 & @xmath321 & @xmath322 & @xmath323 + 2914.844 & 0.649 & 64.5 & @xmath324 & @xmath325 & 1.4 + 2914.848 & 0.650 & 68.6 & 2.6 & @xmath326 & 0.2 + 2914.857 & 0.652 & 69.8 & 2.6 & @xmath327 & 5.1 + 2914.862 & 0.654 & 63.9 & @xmath328 & @xmath329 & @xmath330 + 2914.965 & 0.680 & 74.6 & @xmath331 & @xmath332 & 3.1 + 2915.825 & 0.897 & 41.3 & @xmath333 & @xmath334 & 1.9 + 2915.829 & 0.898 & 39.8 & @xmath335 & @xmath336 & @xmath337 + 2915.833 & 0.899 & 36.6 & @xmath338 & @xmath339 & @xmath340 + 2915.842 & 0.901 & 38.9 & @xmath341 & @xmath342 & @xmath343 + 2915.846 & 0.902 & 38.0 & @xmath341 & @xmath344 & @xmath340 + 2915.952 & 0.929 & 16.9 & 6.6 & @xmath345 & @xmath346 + 2916.831 & 0.151 & @xmath347 & 4.6 & 142.0 & @xmath348 + 2916.835 & 0.152 & @xmath349 & 9.3 & 148.4 & 5.7 + 2916.859 & 0.158 & @xmath350 & 2.5 & 139.0 & @xmath340 + 2916.863 & 0.159 & @xmath351 & 0.6 & 145.6 & 1.4 + 2916.932 & 0.176 & @xmath352 & 5.0 & 139.0 & @xmath353 + 2917.011 & 0.196 & @xmath354 & @xmath337 & 132.5 & @xmath355 + lcc @xmath292 ( days ) & 3.96 ( 9 ) & 3.95863 ( 21 ) + @xmath24 ( hjd-2,450,000 ) & & 2916.233 ( 15 ) + @xmath356 ( hjd-2,450,000 ) & 2935.98 ( 15 ) & + @xmath22 & 0 & 0.156 ( 19 ) + @xmath23 ( @xmath110 ) & & 100 ( 11 ) + @xmath293 ( km s@xmath4 ) & 172 ( 12 ) & 159.9 ( 22 ) + @xmath294 ( km s@xmath4 ) & 221 ( 15 ) & 207.7 ( 29 ) + @xmath295 & 0.78 ( 8) & 0.770 ( 15 ) + @xmath357 ( km s@xmath4 ) & @xmath171 ( 9 ) & @xmath358 ( 17 ) + @xmath359 ( km s@xmath4 ) & @xmath360 ( 12 ) & @xmath361 ( 23 ) + @xmath298 sin@xmath362 ( @xmath299 ) & 13.9 ( 25 ) & 11.1 ( 5 ) + @xmath300 sin@xmath362 ( @xmath299 ) & 10.9 ( 20 ) & 8.6 ( 4 ) + @xmath306 sin @xmath98 ( @xmath302 ) & 30.7 ( 19 ) & 28.39 ( 28 ) + r.m.s.@xmath307 ( km s@xmath4 ) & & 10.9 + r.m.s.@xmath308 ( km s@xmath4 ) & & 14.4 + lccccc 1817.887 & 0.772 & @xmath144181.1 & 5.3 & 143.7 & 5.2 + 1818.890 & 0.889 & @xmath144184.6 & @xmath1442.8 & 136.3 & 2.6 + 1819.873 & 0.004 & @xmath14477.9 & @xmath1441.3 & 18.9 & @xmath1446.9 + 1820.887 & 0.122 & 63.4 & @xmath1440.2 & @xmath144124.7 & @xmath1446.7 + 1821.864 & 0.236 & 148.2 & 15.8 & @xmath144190.4 & @xmath1441.9 + 1822.865 & 0.353 & 120.9 & 3.7 & @xmath144171.9 & 1.1 + 1823.814 & 0.464 & 49.2 & 0.0 & @xmath144105.9 & @xmath1442.9 + 1823.924 & 0.476 & 35.7 & @xmath1443.4 & @xmath14495.7 & @xmath1442.8 + 1824.825 & 0.582 & @xmath14445.5 & 4.8 & 1.0 & 2.0 + 1824.950 & 0.596 & @xmath14454.8 & 8.4 & 16.6 & 4.5 + 1830.842 & 0.284 & 142.9 & 7.5 & @xmath144187.7 & 3.8 + 1830.946 & 0.296 & 136.5 & 2.5 & @xmath144180.0 & 10.1 + 1888.797 & 0.045 & @xmath14429.1 & @xmath1444.5 & @xmath14434.0 & @xmath1446.5 + 1889.746 & 0.156 & 91.2 & @xmath1442.1 & @xmath144148.5 & @xmath1440.0 + 1892.743 & 0.506 & & & & + 1893.774 & 0.626 & @xmath14477.2 & 11.8 & 25.5 & @xmath14413.0 + 1894.771 & 0.742 & @xmath144168.5 & 4.5 & 131.3 & 6.6 + 1894.845 & 0.751 & @xmath144183.3 & @xmath1446.0 & 136.3 & 7.2 + 1895.721 & 0.853 & @xmath144196.3 & @xmath1441.8 & 144.7 & @xmath1442.1 + 1895.811 & 0.864 & @xmath144184.3 & 7.5 & 144.7 & 0.6 + 1896.651 & 0.962 & @xmath144132.7 & @xmath1448.0 & 76.7 & 1.4 + 1896.784 & 0.977 & @xmath144103.8 & 4.1 & 45.3 & @xmath14412.6 + 1897.652 & 0.078 & 6.0 & @xmath14410.1 & @xmath14471.1 & @xmath1441.8 + 1897.784 & 0.094 & 21.0 & @xmath14412.9 & @xmath14487.4 & 0.1 + 1898.656 & 0.196 & 114.5 & @xmath1443.8 & @xmath144177.2 & @xmath1443.2 + 1898.793 & 0.212 & 126.0 & 0.8 & @xmath144173.9 & 7.2 + 1899.658 & 0.312 & 133.2 & 2.5 & @xmath144182.4 & 4.3 + 1899.791 & 0.328 & 124.9 & @xmath1441.5 & @xmath144182.2 & 0.2 + 1900.653 & 0.429 & 69.7 & @xmath1444.9 & @xmath144136.9 & @xmath1447.7 + 1900.784 & 0.444 & 60.6 & @xmath1443.2 & @xmath144124.9 & @xmath1446.7 + 1901.637 & 0.543 & @xmath14432.7 & @xmath14415.8 & @xmath14414.1 & 21.2 + 1901.770 & 0.559 & @xmath14427.6 & 3.0 & @xmath14424.4 & @xmath1443.0 + lr @xmath292 ( days ) & 8.5710 ( 8) + @xmath24 ( hjd-2,450,000 ) & 1862.696 ( 16 ) + @xmath22 & 0.095 ( 11 ) + @xmath23 ( @xmath110 ) & 252 ( 6 ) + @xmath293 ( km s@xmath4 ) & 166.5 ( 18 ) + @xmath294 ( km s@xmath4 ) & 170.8 ( 18 ) + @xmath295 & 0.975 ( 15 ) + @xmath357 ( km s@xmath4 ) & @xmath363 ( 12 ) + @xmath359 ( km s@xmath4 ) & @xmath364 ( 12 ) + @xmath298 sin@xmath362 ( @xmath299 ) & 17.1 ( 6 ) + @xmath300 sin@xmath362 ( @xmath299 ) & 16.6 ( 6 ) + @xmath306 sin @xmath98 ( @xmath302 ) & 56.8 ( 4 ) + r.m.s.@xmath307 ( km s@xmath4 ) & 7.0 + r.m.s.@xmath308 ( km s@xmath4 ) & 7.0 +

both the naturalness of the electroweak symmetry breaking and the resolution of the strong cp problem may require a small higgsino mass generated by a realization of the dfsz axion model . assuming the axino is the lightest supersymmetric particle , we study its implications on and the axion scale . copiously produced light higgsinos at collider ( effectively only neutral nlsp pairs ) eventually decay to axinos leaving prompt multi - leptons or displaced vertices which are being looked for at the lhc . we use latest lhc7 + 8 results to derive current limits on and the axion scale . various higgsino - axino phenomenology is illustrated by comparing with a standard case without lightest axinos as well as with a more general case with additional light gauginos in the spectrum .

1407.1218

the strong cp problem is elegantly resolved by introducing a peccei - quinn ( pq ) symmetry @xcite and its spontaneous breaking resulting in a dynamical field called the axion @xcite . in this mechanism , the cp - violating qcd @xmath1 term is determined by a vacuum expectation value of the axion which dynamically cancels out the non - zero qcd @xmath1-term . the pq symmetry can be realized either by introducing heavy quarks ( ksvz ) @xcite or by extending the higgs sector ( dfsz ) @xcite and its breaking breaking scale @xmath2 is related with the axion coupling constant as @xmath3 with @xmath4 being the domain wall number counting the qcd anomaly , but a certain variations can allow @xmath5 to avoid the domain wall problem ] . the conventionally allowed window of the axion coupling constant is @xmath6 ( for a review , see @xcite ) . the upper bound comes from the axion cold dark matter contribution which is cosmological model dependent . a recent simulation of axionic topological defect contributions provides a stringent upper bound @xmath7 if pq - symmetry were broken after inflation @xcite . the window can be widen if pq - symmetry were broken before or during inflation in certain class of pq symmetry breaking models avoiding too large axionic isocurvature perturbations @xcite . the existence of such a high scale causes quadratic divergences to the higgs boson mass and thus requires a huge fine - tuning to keep stable two scales , the electroweak scale and the pq scale ( or a generic uv scale ) . supersymmetry ( susy ) would be the best - known framework to avoid such a hierarchy problem . however , the electroweak symmetry breaking in susy suffers from a certain degree of fine - tuning to maintain a desirable potential minimization condition : @xmath8 where @xmath9 are the soft masses of the two higgs doublets , @xmath10 is the ratio of their vacuum expectation values , and @xmath0 is the higgs bilinear parameter in the superpotential . as lhc finds no hint of susy , it pushes up the soft mass scale above tev range , the minimization condition ( [ mzmin ] ) requires a fine cancellation among different terms . barring too huge cancellation , one may arrange @xmath9 and @xmath0 not too larger than @xmath11 . this has been advocated as `` natural susy '' @xcite implying stops / sbottoms at sub - tev and light higgsinos with @xmath12 such a spectrum can also be obtained radiatively with multi - tev soft masses at a uv scale @xcite . an electroweak @xmath0 may be related to the pq symmetry in the manner of dfsz @xcite , which introduces a non - renormalizable superpotential in the higgs sector : @xmath13 where @xmath14 and thus @xmath15 carries a non - trivial pq charge and @xmath16 is the reduced planck mass . upon the pq symmetry breaking @xmath17 , a @xmath0 term is generated by @xmath18 . once pq - symmetry is broken , there appear the axion @xmath19 , its scalar partner , the saxion @xmath20 , and the fermion super - partner , the axino @xmath21 . forming an axion superfield @xmath22 , one can schematically write down the effective @xmath0-term superpotential ; @xmath23 where @xmath24 is a parameter depending on the pq symmetry breaking sector ; we use @xmath25 in this paper . in the context of the natural susy having a small @xmath0 parameter , a neutral higgsino tends to be the lightest supersymmetric particle ( lsp ) and thus is a dark matter candidate assuming r - parity . in this case , a heavy axino decay to the lsp can change the standard thermal higgsino dark matter density resulting in different mixtures of the axion and higgsino dark matter components depending on the pq scale @xcite . in this paper , we investigate implications of the axino lsp in the framework of `` the natural susy dfsz model '' . naively speaking , the axino mass is expected to be of order of the soft susy breaking scale , but it is in general model dependent @xcite . as a dark matter , the abundance of axino depends on the history of the universe involving either the condensation of saxion or the reheating temperature of the primordial inflation . axinos can be produced abundantly either by saxion decay @xcite or by interactions with thermal particles @xcite . to avoid axino over - production , we assume the axino is very light or the reheat temperature is low enough to suppress the thermal production in this paper . since higgsinos are predicted to be light in the natural susy scenario , they can be copiously produced at the lhc and decay to axino plus the higgs boson @xmath26 or @xmath27 boson through the coupling in eq . ( [ weff ] ) . this leads to interesting signatures of multi - leptons / jets and missing transverse energy(met ) which can be prompt or displaced depending on the pq scale . notice that the standard higgsino lsp scenario is hard to probe as heavier higgsino decays produce unobservably soft leptons or pions due to a small mass gap between a heavier higgsino and the higgsino lsp . currently , the atlas and cms collaborations look for prompt multi - lepton plus met and displaced di - jet / lepton signatures . applying the current search results to the higgsino - axino system , we obtain various limits on the @xmath0 parameter as well as the pq scale . we assume that sleptons , squarks and gluinos are heavy , but see refs . @xcite for earlier collider studies in the presence of light sleptons . in sec . [ sec : noaxino ] , we first translate the current multi - lepton + met search results to the higgsino - bino system where the higgsino and bino are taken to be the next - to - lsp ( nlsp ) and the lsp , respectively , and thus the nlsp decay to the lsp plus @xmath26 or @xmath27 can lead to prompt multi - lepton signatures . in sec . [ sec : axinolsp ] , we turn into a case of the higgsino nlsp and the axino lsp which can lead to displaced vertices from the nlsp decay . then , we extend our analysis to the case of the higgsino nnlsp and the bino nlsp with the axino lsp in sec . [ sec : higgsinonnlsp ] . lhc14 projections of displaced vertex searches are estimated in sec . [ sec : lhc14 ] to see how far the axion scale can be probed . finally , we conclude in sec . [ sec : conclusion ] . and @xmath27 bosons ) from the @xmath28+met search ( 20.3/fb ) is shown as the solid line for reference @xcite . assuming the higgsino nlsp with the bino lsp , we re - interprete the @xmath28+met search(dashed ) . the relevant br is taken into account with @xmath29 , and the @xmath28 search is not sensitive to the sign of @xmath0 as depicted in fig . [ fig : br - higgsinopair ] . we assume @xmath302 tev and @xmath31 . more on sec . [ sec : noaxino].,scaledwidth=60.0% ] + met search . both @xmath29(solid ) or @xmath32(dashed ) are shown . the relevant br shown does not vary much in most of the parameter space with both signs of @xmath0 @xcite . this result is used in fig . [ fig : higgsino3l].,scaledwidth=49.0% ] before considering the axino lsp , let us first deduce and summarize the current exclusion bounds in the case of ( 1 ) the standard higgsino - like nlsp and bino - like lsp as well as in the case of ( 2 ) higgsino - like lsp . the results will be later compared with those with axino lsps . consider first the case ( 1 ) with higgsino nlsp and bino lsp . being relatively light , a sizable number of charged and neutral higgisinos , @xmath33 and @xmath34 , can be produced electroweakly and decay to the lsp @xmath35 through @xmath36 and @xmath37 . the neutral higgsino decays to the @xmath27 boson are relevant to the multi - lepton searches although higgs decay products can certainly be useful when higgsinos are heavy @xcite ; see also @xcite . considering light higgsinos , we ignore higgs contributions in this work . ] , and its branching ratio ( br ) is a function of @xmath38 and the sign of @xmath0 . in fig . [ fig : higgsino3l ] , we show current bounds on the higgsino nlsp overlapping the officially reported bound on the wino nlsps from the @xmath28+met search @xcite for reference . the associate production of charged and neutral higgsinos is the largest and is constrained from the @xmath28+met search : @xmath39 . here , the dependence on the underlying model parameters such as @xmath38 and the sign of @xmath0 is weak as demonstrated in the left panel of fig . [ fig : br - higgsinopair ] the relevant br of higgsino pairs is in general close to a half @xcite . so we use a positive @xmath0 to draw the bound in fig . [ fig : higgsino3l ] . the bound on higgsinos is weaker than the official bound on winos due to two modifications : i ) the total production cross - section of higgsino pair @xmath40 is smaller than that of wino pairs @xmath41 ( by about a factor 2 for @xmath42gev nlsps ) , and ii ) the br for @xmath43 is smaller than 1 . the actual bound on winos will also be weaker than the officially reported one according to a smaller br . on the other hand , other multi - lepton searches contributed mainly from other pair productions of higgsinos currently lead to weaker or null bounds ; for example , the associate production of two neutral higgsinos is only weakly constrained from the @xmath44+met search @xcite via @xmath45 . in all , the nlsp higgsino mass exclusion currently reaches up to about 250gev while the lhc sensitivity drops quickly as the mass - gap between the nlsp and lsp becomes smaller . there exist other experimental results on @xmath28+met @xcite and @xmath44 + met @xcite . but they are not essentially different from the ones used above . the @xmath46 @xcite and the same - sign dilepton @xcite searches do not give a much stronger bound for such a light @xmath0 . the @xmath47 search for the @xmath48 is also potentially useful @xcite . in any case , our interpretation of a few standard searches in fig . [ fig : higgsino3l ] ( and similar figures throughout in this paper ) is a reasonable and useful estimation of current higgsino exclusion limits . see appendix [ app : simulation ] for more details on how we obtain the bounds . the higgsino can also be the lsp ( the case ( 2 ) above ) . if other gauginos are far away in mass , all three higgsino states one charged and two neutral are nearly degenerate . even though light higgsinos are abundantly produced , visible decay products of decays between them are generally too soft to be observable at collider and two lsps are produced in back - to - back directions giving a small met . it is why the search of nearly degenerate spectrum is difficult . the squeezed spectrum is typically searched by triggering hard initial state radiations(isr ) which subsequently boost the visible and invisible decay products . no dedicated lhc search is reported yet , but several theoretical studies of lhc prospects have been carried out in refs . it is expected that the monojet+met alone at lhc14 would be sensitive to nearly degenerate @xmath49gev higgsinos only with @xmath50/ab of data , but somewhat more optimistic approach would be to utilize soft leptons from heavier higgsino decays when the ( model - dependent ) mass - splitting is @xmath51gev or larger . decays between higgsinos are rather prompt @xcite ( even when the splitting is dominated by small loop - induced contributions ) , so the disappearing track searches @xcite that are sensitive to the degenerate wino lpss are not so useful for higgsino lsps ; see appendix [ app : inodecays ] . gev are assumed here . in the lower panel , we show tree - level mass splittings between higgsino states . the loop - induced mass splitting of the higgsinos , @xmath52mev @xcite , is marked as a horizontal dotted line . , title="fig:",scaledwidth=49.0% ] gev are assumed here . in the lower panel , we show tree - level mass splittings between higgsino states . the loop - induced mass splitting of the higgsinos , @xmath52mev @xcite , is marked as a horizontal dotted line . , title="fig:",scaledwidth=49.0% ] + gev are assumed here . in the lower panel , we show tree - level mass splittings between higgsino states . the loop - induced mass splitting of the higgsinos , @xmath52mev @xcite , is marked as a horizontal dotted line . , title="fig:",scaledwidth=49.0% ] in this section , we consider the situation of light ( nlsp ) higgsinos and heavy gauginos with the axino lsp . in the decoupling limit of gauginos , there occurs an interesting and rich situation for the decays of heavier higgsinos . since the axino lsp is weakly interacting , higgsinos can dominantly decay either to the lightest higgsino or to the axino lsp , depending on the gaugino masses and the pq symmetry breaking scale , @xmath2 . in fig . [ fig : h2hpwidth ] , we show relative decay widths of the heavier higgsinos for massless axinos and @xmath53 gev . heavier axinos ( for a fixed @xmath0 ) and a higher @xmath2 scale only make the decays to the axino smaller . for @xmath54 a few tev , both charged and neutral higgsinos decay dominantly to the lightest higgsinos even with massless axinos and @xmath53gev . for larger @xmath55 , the mass splitting between higgsino states are too small to have quick enough decays between them . in this paper , we simply assume @xmath56tev for which all heavier higgsinos decay to the lightest higgsinos ; also as long as @xmath57 and @xmath58 are tev scales , the mass splitting between higgsinos are @xmath50gev(see fig . [ fig : h2hpwidth ] ) and soft leptons from decays between higgsinos are too soft to be reliably measurable . , in the presence of the axino lsp . @xmath53gev here , and the lifetime scales with @xmath59 . we mark @xmath60 m with a blue line as a convenient reference for the displaced decay , and we mark @xmath61 m with a red line for decaying outside detector . the rapid increase of the lifetime below @xmath62gev is due to the closing of any two - body decay modes . , scaledwidth=49.0% ] whether or not the decays of the higgsino nlsp to the axino lsp can leave observable displaced vertices depends on the values of @xmath0 , @xmath2 and the mass gap between the nlsp and lsp . the proper decay length of the higgsino nlsp , @xmath63 , is shown in fig . [ fig : width - higgsino ] . the distinction between the prompt and displaced decays ( also whether decaying inside or outside detector ) is not determined solely by the @xmath64 but also by kinematics of decay products and the probabilistic distributions of decay lengths . but by conveniently referring to the contours of @xmath65m(blue ) and 10m(red ) standard tight leptons are required to satisfy @xmath66 m at lhc @xcite and the size of atlas detector , for example , is @xmath6710 m @xcite , we find that the decay is most likely be inside detector ( and to be displaced ) at collider for the favored region of parameter space with a smaller @xmath0 and @xmath68 gev ( unless the mass - gap between higgsino nlsps and axino lsps is very small ) . see ref . @xcite for earlier studies of displaced decays of singlinos in a most related context , refs . @xcite for displaced decays of standard neutralinos and refs . @xcite for lightest higgsino phenomenology with gravitino lsps . , @xmath69 is shown as blue . for the higgsino - axino case in sec . [ sec : axinolsp ] , @xmath70 is shown as red . for the higgsino - bino - axino case in sec . [ sec : higgsinonnlsp ] , it is the @xmath71 shown as black ; @xmath72gev is assumed . all prompt pair productions of inos effectively leading to the aforementioned production are added ; see text for more discussions.,scaledwidth=49.0% ] for @xmath29(upper ) and @xmath73(lower ) . the @xmath44+met search(blue ) @xcite and the cms dijet dv search(red ) @xcite are most relevant . we assume two extreme values of the dv reconstruction efficiencies : @xmath74(left ) and 0.01(right ) . @xmath75gev is used , but see fig . [ fig : higgsino - axino2 ] for results on other values of @xmath76 . @xmath55=2tev and @xmath31 . more on sec . [ sec : axinolsp ] . , title="fig:",scaledwidth=49.0% ] for @xmath29(upper ) and @xmath73(lower ) . the @xmath44+met search(blue ) @xcite and the cms dijet dv search(red ) @xcite are most relevant . we assume two extreme values of the dv reconstruction efficiencies : @xmath74(left ) and 0.01(right ) . @xmath75gev is used , but see fig . [ fig : higgsino - axino2 ] for results on other values of @xmath76 . @xmath55=2tev and @xmath31 . more on sec . [ sec : axinolsp ] . , title="fig:",scaledwidth=49.0% ] + for @xmath29(upper ) and @xmath73(lower ) . the @xmath44+met search(blue ) @xcite and the cms dijet dv search(red ) @xcite are most relevant . we assume two extreme values of the dv reconstruction efficiencies : @xmath74(left ) and 0.01(right ) . @xmath75gev is used , but see fig . [ fig : higgsino - axino2 ] for results on other values of @xmath76 . @xmath55=2tev and @xmath31 . more on sec . [ sec : axinolsp ] . , title="fig:",scaledwidth=49.0% ] for @xmath29(upper ) and @xmath73(lower ) . the @xmath44+met search(blue ) @xcite and the cms dijet dv search(red ) @xcite are most relevant . we assume two extreme values of the dv reconstruction efficiencies : @xmath74(left ) and 0.01(right ) . @xmath75gev is used , but see fig . [ fig : higgsino - axino2 ] for results on other values of @xmath76 . @xmath55=2tev and more on sec . [ sec : axinolsp ] . , title="fig:",scaledwidth=49.0% ] based on the higgsino decay patterns discussed above , we have a simple scenario where any higgsino pair productions would essentially be the same as the @xmath77 pair production and relevant collider signals come only from @xmath78 . it is useful to summarize several differences between the current situation and the standard higgsino nlsp and bino lsp case discussed in sec . [ sec : noaxino ] : 1 . the @xmath77 production is sizable . any pair productions of higgsinos essentially lead to the @xmath77 and resulting total production rate ( adding all ) is about 8 times larger than that of the usual @xmath79 pair production as shown in fig . [ fig : neutralhiggsinopair ] . note that pair productions of neutral winos or binos are highly suppressed . the enhanced neutralino pair production can also be resulted in the case with the weakly interacting gravitino lsp @xcite . 2 . among standard multi - lepton searches , the @xmath44+met search is most relevant through @xmath80 . the @xmath77 can now contribute to the stringent @xmath28+met ( @xmath81+met as well ) searches only by accidentally losing one or more leptons . thus , such multi - lepton searches are weakened . higgsino phenomenology depends only on the decay pattern of @xmath63 . decays of a single neutral higgsino , @xmath63 , depends sensitively on @xmath38 and the sign of @xmath0 ( as can be seen , e.g. in fig . [ fig : h1pairbr ] ) . on the other hand , in the standard case without axino lsps , decays of all higgsino states are indistinguishable at collider and are equally important , and summing all indistinguishable decays make some standard higgsino phenomenology less sensitive to those parameters @xcite ; see one example in fig . [ fig : br - higgsinopair ] . as discussed , the decay of @xmath63 is likely displaced . the displaced decay further weakens the standard multi - lepton susy searches . however , dedicated displaced vertex(dv ) searches are now relevant . for the given parameter space of higgsino nlsp with axino lsp ; for example , see fig . [ fig : higgsino - axino1 ] that the value would be @xmath82 for the 250 - 0 case with @xmath83 . although there can be a smaller @xmath2 not excluded , we conveniently choose this highest excluded value to show in these plots . cms dijet dv and @xmath44+met are used . numbers without(with ) parentheses are results with @xmath840.1(0.01 ) . the `` '' implies no existing bounds . the parameter space without anything written is not simulated by ourselves . the light - gray - dashed diagonal lines imply @xmath85gev below which the decays to the off - shell @xmath27 boson begins to be phase - space suppressed and @xmath86gev is already high enough to make all higgsinos decay far outer region or outside the detector thus , no collider bounds in general.,title="fig:",scaledwidth=49.0% ] for the given parameter space of higgsino nlsp with axino lsp ; for example , see fig . [ fig : higgsino - axino1 ] that the value would be @xmath82 for the 250 - 0 case with @xmath83 . although there can be a smaller @xmath2 not excluded , we conveniently choose this highest excluded value to show in these plots . cms dijet dv and @xmath44+met are used . numbers without(with ) parentheses are results with @xmath840.1(0.01 ) . the `` '' implies no existing bounds . the parameter space without anything written is not simulated by ourselves . the light - gray - dashed diagonal lines imply @xmath85gev below which the decays to the off - shell @xmath27 boson begins to be phase - space suppressed and @xmath86gev is already high enough to make all higgsinos decay far outer region or outside the detector thus , no collider bounds in general.,title="fig:",scaledwidth=49.0% ] channel relevant to the @xmath44+met search . ( left ) : the higgsino - bino case in sec . [ sec : noaxino ] . ( right ) : the higgsino - axino case in sec . [ sec : axinolsp ] . both @xmath29(solid ) and @xmath32(dashed ) are shown . the dotted diagonal lines are at @xmath87 above which the decays to the ( on- or off - shell ) @xmath27 boson is almost 100% . the br can be larger than a half in the right panel with @xmath29 . these results are used in obtaining the @xmath44+met bounds . , title="fig:",scaledwidth=49.0% ] channel relevant to the @xmath44+met search . ( left ) : the higgsino - bino case in sec . [ sec : noaxino ] . ( right ) : the higgsino - axino case in sec . [ sec : axinolsp ] . both @xmath29(solid ) and @xmath32(dashed ) are shown . the dotted diagonal lines are at @xmath87 above which the decays to the ( on- or off - shell ) @xmath27 boson is almost 100% . the br can be larger than a half in the right panel with @xmath29 . these results are used in obtaining the @xmath44+met bounds . , title="fig:",scaledwidth=49.0% ] in fig . [ fig : higgsino - axino1 ] , we analyze the exclusion bounds on the @xmath0-@xmath2 parameter space with @xmath88 gev . both the @xmath44+met search @xcite ( constraining too much prompt decays ) and the cms dijet dv search @xcite ( constraining a certain range of displaced decay ) are relevant . fig . [ fig : higgsino - axino2 ] shows results in the more general parameter space . for high enough @xmath2 scales , no bound exists ; either the higgsino decays still dominantly inside detector but its dv is not searched efficiently or the higgsino dominantly decays outside detector and its phenomenology is essentially the same as that of the higgsino - lsp case whose current null bounds are discussed in sec . [ sec : noaxino ] . the bound from the dv search is sensitive to the dv reconstruction efficiency , @xmath89 , which is an experimental factor capturing how much fraction of dvs are really reconstructed . for the low extreme value of @xmath90 ( see ref . @xcite that @xmath91 is a reasonable range to consider ) , the bound almost disappears . the bound from the @xmath44+met search is stronger for @xmath29 than @xmath32 because the relevant br is larger as depicted in the right panel of fig . [ fig : h1pairbr ] . the total decay width of the higgsino depends slightly on the sign of @xmath0 , thus so does fig . [ fig : higgsino - axino2 ] . it is useful to understand why the @xmath28+met search is now significantly weaker than the @xmath44+met search here as opposed to the results of sec . [ sec : noaxino ] . the main reason why the @xmath44+met search is now sensitive to this model while it is not sensitive to the standard higgsino - bino case in sec . [ sec : noaxino ] is the enhanced neutral higgsino pair production in this model as discussed in regard of fig . [ fig : neutralhiggsinopair ] . another minor reason is that the relevant br ( right panel of fig . [ fig : h1pairbr ] ) can be somewhat larger than a half while it is typically not in the standard case ( left panel of fig . [ fig : h1pairbr ] ) . , is pair produced here . the decays of @xmath63 and @xmath92 are typically opposite @xcite . ] on the other hand , compared to the dominant @xmath93 production in the higgsino - bino case leading to the @xmath28+met signal , the @xmath77 here is not much larger , thus a small selection efficiency to the @xmath28+met here ( needing to accidentally lose one lepton ) has a big impact to decrease the exclusion reach of the @xmath28+met in this model . the cms dilepton dv search @xcite can also give a relevant bound , but this search looks for a similar range of decay length @xmath94 cm ; so we conservatively use the dijet dv results to obtain bounds . other dedicated dv searches @xcite are less relevant and less stringent . in all , by having the axino lsp , some ranges of @xmath0 and @xmath2 can be probed at the lhc since the currently allowed range of @xmath2 falls in the right range to allow nlsp higgsinos to decay inside detector either promptly or with dvs . on the other hand , a higher @xmath95gev with @xmath96 100 - 400gev can avoid all the current lhc searches . when the mass gap between the higgsino and the axino is smaller than about @xmath11 , the higgsino generally decays far outer region or outside the detector and no current collider searches constrain the model . the higgsino nnlsp is assumed to be nearby with @xmath97 . for a positive @xmath0 assumed here , the mass eigenvalue of the bino nlsp is related as @xmath98gev due to the mild mixing between binos and higgsinos . more on sec . [ sec : higgsinonnlsp ] . , scaledwidth=49.0% ] having another gauginos in the light spectrum is another interesting possibility . in this section , we consider the case with the higgsino nnlsp , bino nlsp and axino lsp . as direct bino pair production is very small , the collider phenomenology relies on all possible pair productions of nnlsp higgsinos and nlsp binos . we assume that @xmath99gev so that these productions are big enough for collider analysis . due to this close - by masses and resulting mild mixing between binos and higgsinos , the mass eigenvalue of the bino - like lsp is 20gev lighter than the @xmath57 : @xmath100gev . similarly to the case of the higgsino nlsp with axino lsp discussed in sec . [ sec : axinolsp ] , the produced higgsinos dominantly and promptly decay to the bino nlsp . it is more obviously true here because the decays between higgsinos and binos are not small - gap - suppressed . then , again all the higgsino productions essentially lead to abundant bino pair productions in the collider physics point of view . binos can also decay to the axino lsp with substantial lifetime . in fig . [ fig : width - bino ] , we show the proper decay length of bino nlsps . in the majority of relevant parameter space , binos likely decay inside detector either promptly or with dvs . compared to the higgsino nlsp s decay in fig . [ fig : width - higgsino ] , binos have a somewhat longer lifetime because binos couple to axinos via higgsino mixtures in the dfsz model . numerically , it turns out that the bino typically has a 35 times longer lifetime ( with the same other parameters ) which implies that about 2 times lower @xmath2 scale is needed for a similar lifetime . if higgsinos are much heavier , the bino decays are much slower with about 1020 times longer lifetime due to a smaller bino - higgsino mixing . for the given parameter space of the higgsino nnlsp , bino nlsp and axino lsp discussed in sec . [ sec : higgsinonnlsp ] with @xmath29 . figure details are as in fig . [ fig : higgsino - axino1 ] and [ fig : higgsino - axino2 ] . no bound is derived from multi - lepton searches in this case . for @xmath32 , a similar bound is obtained from the dv search . @xmath74 here , and no dv bound exists for @xmath90.,title="fig:",scaledwidth=49.0% ] for the given parameter space of the higgsino nnlsp , bino nlsp and axino lsp discussed in sec . [ sec : higgsinonnlsp ] with @xmath29 . figure details are as in fig . [ fig : higgsino - axino1 ] and [ fig : higgsino - axino2 ] . no bound is derived from multi - lepton searches in this case . for @xmath32 , a similar bound is obtained from the dv search . @xmath74 here , and no dv bound exists for @xmath90.,title="fig:",scaledwidth=49.0% ] it is useful to note several differences between this scenario and the higgsino - axino case in sec . [ sec : axinolsp ] . ( i ) for the given nlsp mass , the effective total production of nlsp pairs is smaller here as shown in fig . [ fig : neutralhiggsinopair ] because the model here relies on the ( associate ) productions of heavier higgsinos . ( ii ) decays of nnlsp higgsinos to nlsp binos can produce observable particles as we assume about 50gev mass - gap . we will explain later how we treat these visible particles in our analysis . ( iii ) now the decay pattern of the bino nlsp is relevant to collider searches instead of that of the higgsino . in fig . [ fig : binodv ] , we analyze the exclusion bounds . again , both the @xmath44+met search ( constraining too much prompt decays ) and the cms dijet dv search ( constraining a certain range of displaced decay ) are relevant . for high enough @xmath2 scales , no bound exists ; either the bino decays still dominantly inside detector but its dv is not searched efficiently or the bino dominantly decays outside detector . when the bino decays outside detector , the visible decay products of nnlsp higgsinos can be important in collider searches the collider physics will then be essentially the same as that of the higgsino nlsp and bino lsp considered in sec . [ sec : noaxino ] as if axinos were absent . however , the mass - gap between the higgsino and bino is only 50gev in this work , and by referring to fig . [ fig : higgsino3l ] showing the current bounds on the higgsino - bino model , we find that the visible decay products of higgsino nnlsp with such small - gap is weakly constrained . we conservatively assume that we can ignore all ( soft ) leptons from higgsino decays in our multi - lepton analysis , but we will include all and only leptons from bino decays to axinos in our analysis ( when binos decay promptly inside detector ) the more accurate analysis will not give a much stronger bound anyway . the fig . [ fig : binodv ] , compared with fig . [ fig : higgsino - axino1 ] and [ fig : higgsino - axino2 ] , shows that the bound on this model is somewhat weaker than that of the higgsino - axino case in sec . [ sec : axinolsp ] . for @xmath90 , no bounds from the dv search is derived . for @xmath32 , no bounds from the multi - lepton search is derived . these weaker bounds are mainly because the effective total production of bino pairs is smaller for the given bino mass as discussed above and as shown in fig . [ fig : neutralhiggsinopair ] . the results depend on the choice of @xmath101gev . the heavier higgsinos , the smaller signal productions and the weaker collider constraints it is thus a less interesting scenario . the lighter higgsino closer to the bino can induce a larger mixing making binos decay more promptly ( but not faster than pure higgsinos discussed in previous section ) and the excluded parameter space change slightly . if we still assume that lepton from decays between those states are soft enough , not much qualitative change in the collider physics would arise . but again , in all , by having the axino lsp as well as light gauginos , some ranges of @xmath0 and @xmath2 can be probed at the lhc since the currently allowed range of @xmath2 falls in the right range to allow nlsp binos to decay inside detector either promptly or with dvs . on the other hand , a higher @xmath95gev with @xmath96100 - 400gev can avoid all the current lhc searches . if the nnlsp higgsino is much heavier , the model has a looser connection with the naturalness ; in any case , no any sizable production modes are available then and the search will rely on heavier particle productions . and @xmath31 . the lower horizontal dashed line is at the current cms data 18.6/fb while the upper horizontal line at 1/ab is just shown for easy reading . * right : * the highest excluded @xmath102 value with 100/fb as in the left panel of fig . [ fig : higgsino - axino2 ] with @xmath74 . increasing to 3/ab of data roughly enhances the reach by 0.7 of @xmath102 . , title="fig:",scaledwidth=49.0% ] and @xmath31 . the lower horizontal dashed line is at the current cms data 18.6/fb while the upper horizontal line at 1/ab is just shown for easy reading . * right : * the highest excluded @xmath102 value with 100/fb as in the left panel of fig . [ fig : higgsino - axino2 ] with @xmath74 . increasing to 3/ab of data roughly enhances the reach by 0.7 of @xmath102 . , title="fig:",scaledwidth=49.0% ] as the lhc 14tev will start in a year , it is interesting to estimate the prospect of it . we project the current cms dijet dv search results to study how high @xmath2 scale can be probed at 14tev . it is a techincally difficult task because future detectors are different and pile - up backgrounds at higher energy collisions are larger . we , however , parameterize the dv reconstruction efficiency which will be most dependent on detector performance by an unknown @xmath89 , and relatively hard cuts on jet @xmath103 used in this analysis ( @xmath104gev and @xmath105gev which shall be scaled up at 14tev ) will make the soft pile - up effects less influential . if we assume that cut / reconstuction efficiencies and the signal - to - background ratio after optimal cuts stay relatively constant between 8tev and 14tev analyses , the following simple scaling rule of the statistical significance is obtained ( significance)_i = = ( ) , = = , [ eq : scaling ] where each factor in the parenthesis , the signal - to - background ratio @xmath106 , signal cut efficiency @xmath107 and the assumed @xmath74 , stay constants as discussed . @xmath108 are production rates of signal and background , and the probability for displaced decays to be selected by the search , @xmath109 , depends on the @xmath2 and mass spectrum . in all , the significance simply scales with the square root of signal event counts . 8tev cms dijet dv search bounds can be extrapolated to the 14tev bounds by finding proper @xmath2 and mass spectrum giving the same signal event counts as the upper bound of 8tev results . we show 14tev projected results in fig . [ fig : higgsino14tev ] obtained in this way . the higgsino - axino model in sec . [ sec : axinolsp ] is used . for the given mass spectrum , lhc14 100(3000)/fb can probe higher @xmath2 scale by 0.60.7(1.31.4 ) of @xmath102 as shown in the left panel for one choice of parameters . similar size of improvement is expected for the most of light higgsino parameter space shown in the right panel . with 3000/fb , @xmath2 as high as @xmath110gev which is a general upper bound is expected to be probed with light higgsinos . a more dedicated search will be useful in the near future . the electroweak - scale axino and higgsino are perhaps predicted altogether by a naturalness philosophy of particle physics . the implications and the consistency of having both light axinos and higgsinos are studied in the context of a few benchmark models of supersymmetry . interestingly , for the typical range of the pq scale , @xmath111gev , the electroweak - scale nlsp can still decay to the axino lsp inside detector both promptly and by leaving a dv . the @xmath44+met signature from the prompt decay of the nlsp is enhanced among standard susy searches as all heavier neutralinos and charginos decay promptly first to nlsp neutralinos so that nlsp neutralino pair productions which are relevant to the collider physics are effectively enhanced . the displaced decay of the nlsp is constrained by dedicated dv searches for a certain range of @xmath2 typically of @xmath112 depending on the mass spectrum searches for a wider range of decay lengths maybe possible @xcite . a higher pq scale of @xmath113 with the electroweak - scale @xmath0 or the mass spectrum with small mass - gap between the nlsp and lsp is generally safe from all current collider searches . lhc14 , however , is expected to probe the large part of interesting parameter space with light higgsinos according to our naive estimation , thus a more dedicated search is motivated . we hope that we provided a basic collider physics of the natural supersymmetry with the axino lsp and light higgsino which can also be complementary to the widely studied axino sector cosmology . thanks kiwoon choi , hyung do kim and chang sub shin for useful conversations . s.j . thanks kias center for advanced computation for providing computing resources . g.b . acknowledges support from mec grant fpa2011 - 23596 , the gv grant prometeoii/2013/017 and eu fp7 itn invisibles ( marie curie actions , pitn - ga-2011 - 289442 ) . was supported in part by the national science foundation under grant no . nsf phy11 - 25915 . is supported in part by national research foundation ( nrf ) of korea under grant 2013r1a1a2058449 . is supported in part by nrf research grant 2012r1a2a1a01006053 . all the relevant two- and three - body decay widths of inos are calculated and collected in ref . @xcite ( see also refs . @xcite for earlier results ) . in this appendix , we further summarize how we calculate the two - body decays to pions which is relevant when the mass gap is very small @xmath114gev . the two - body decay @xmath115 is calculated as @xcite ( ^+ ^0 ^+ ) = ( ^+ ) ( 1- ) ^1/2 ( 1- ) ^-2 , where the total decay width of a charged pion is @xmath116 m or @xmath117ns or @xmath118gev @xcite . the mass splitting between the chargino and the neutralino is denoted by @xmath119 . we use @xmath120mev , @xmath121mev @xcite . for @xmath122(355)mev which is the one - loop asymptotic wino(higgsino ) mass splitting @xcite , the proper decay length is @xmath123(0.34)cm ( equivalently , @xmath1240.20(0.011)ns ) . the current disappearing track search @xcite is sensitive to @xmath125ns , thus is currently not so sensitive to the nearly degenerate higgsinos . we list methods and numerical results that we used to obtain various exclusion bounds in this paper . for the @xmath28+met result , we use the reported upper limits on the number of events in various ` sr0\taua ` bins of ref . @xcite . the ` sr0\taua - bin16 ` is usually strongest for heavy nlsps . for the @xmath44+met result , we interpret the result in the bin of ` 2ossf + 0\tau_h ` with met@xmath126100gev of ref . @xcite to the upper limit of number of events @xmath127 at 1.96@xmath128cl . interestingly , a very similar analysis has been carried out by atlas in ref . @xcite , but their weaker cut on met@xmath12675gev leads to a much weaker bound . thus , the optimization of the @xmath44+met cuts in each parameter space as roughly done for the @xmath28+met above will be useful . for the dijet dv result in ref . @xcite , we conservatively use the result for @xmath129 cm ( combined with 2 observed events ) to obtain the upper limit on the new physics contribution @xmath130 at 1.96@xmath128cl . for all results , we generate ` madgraph ` @xcite events with up to one additional parton and showered them by interfacing with ` pythia ` @xcite using the ` mlm ` @xcite matching . we use ` fastjet ` @xcite for particle reconstruction . r. d. peccei and h. r. quinn , phys . * 38 * ( 1977 ) 1440 . s. weinberg , phys . lett . * 40 * ( 1978 ) 223 ; f. wilczek , phys . rev . * 40 * ( 1978 ) 279 . j. e. kim , phys . * 43 * ( 1979 ) 103 ; m. a. shifman , a. vainstein and v. i. zakharov , nucl . phys . * b166 * ( 1980 ) 493 . m. dine , w. fischler and m. srednicki , phys . lett . * b104 * ( 1981 ) 199 ; a. p. zhitnitskii , sov . * 31 * ( 1980 ) 260 . j. e. kim and g. carosi , `` axions and the strong cp problem , '' rev . * 82 * ( 2010 ) 557 [ arxiv:0807.3125 [ hep - ph ] ] . t. hiramatsu , m. kawasaki , k. i . saikawa and t. sekiguchi , `` production of dark matter axions from collapse of string - wall systems , '' phys . d * 85 * , 105020 ( 2012 ) [ erratum - ibid . d * 86 * , 089902 ( 2012 ) ] [ arxiv:1202.5851 [ hep - ph ] ] . t. higaki , k. s. jeong and f. takahashi , `` solving the tension between high - scale inflation and axion isocurvature perturbations , '' arxiv:1403.4186 [ hep - ph ] ; k. choi , k. s. jeong and m. -s . seo , `` string theoretic qcd axions in the light of planck and bicep2 , '' arxiv:1404.3880 [ hep - th ] ; e. j. chun , `` axion dark matter with high - scale inflation , '' arxiv:1404.4284 [ hep - ph ] . r. kitano and y. nomura , `` supersymmetry , naturalness , and signatures at the lhc , '' phys . d * 73 * ( 2006 ) 095004 [ hep - ph/0602096 ] ; c. brust , a. katz , s. lawrence and r. sundrum , `` susy , the third generation and the lhc , '' jhep * 1203 * ( 2012 ) 103 [ arxiv:1110.6670 [ hep - ph ] ] . m. papucci , j. t. ruderman and a. weiler , `` natural susy endures , '' jhep * 1209 * ( 2012 ) 035 [ arxiv:1110.6926 [ hep - ph ] ] ; h. baer , v. barger , p. huang , a. mustafayev and x. tata , `` radiative natural susy with a 125 gev higgs boson , '' phys . rev . * 109 * ( 2012 ) 161802 [ arxiv:1207.3343 [ hep - ph ] ] . j. e. kim and h. p. nilles , `` the mu problem and the strong cp problem , '' phys . b * 138 * ( 1984 ) 150 ; e. j. chun , j. e. kim and h. p. nilles , `` a natural solution of the mu problem with a composite axion in the hidden sector , '' nucl . b * 370 * ( 1992 ) 105 . k. j. bae , h. baer and e. j. chun , `` mainly axion cold dark matter from natural supersymmetry , '' phys . d * 89 * ( 2014 ) 031701 [ arxiv:1309.0519 [ hep - ph ] ] . k. j. bae , h. baer and e. j. chun , `` mixed axion / neutralino dark matter in the susy dfsz axion model , '' jcap * 1312 * ( 2013 ) 028 [ arxiv:1309.5365 [ hep - ph ] ] . t. goto and m. yamaguchi , `` is axino dark matter possible in supergravity ? , '' phys . b * 276 * , 103 ( 1992 ) ; e. j. chun , j. e. kim and h. p. nilles , `` axino mass , '' phys . b * 287 * ( 1992 ) 123 [ hep - ph/9205229 ] . e. j. chun and a. lukas , `` axino mass in supergravity models , '' phys . b * 357 * ( 1995 ) 43 [ hep - ph/9503233 ] ; e. j. chun , d. comelli and d. h. lyth , `` the abundance of relativistic axions in a flaton model of peccei - quinn symmetry , '' phys . d * 62 * ( 2000 ) 095013 [ hep - ph/0008133 ] ; e. j. chun , h. b. kim and d. h. lyth , `` cosmological constraints on a peccei - quinn flatino as the lightest supersymmetric particle , '' phys . d * 62 * ( 2000 ) 125001 [ hep - ph/0008139 ] . s. kim , w. -i . park and e. d. stewart , `` thermal inflation , baryogenesis and axions , '' jhep * 0901 * , 015 ( 2009 ) [ arxiv:0807.3607 [ hep - ph ] ] . e. j. chun , `` dark matter in the kim - nilles mechanism , '' phys . d * 84 * ( 2011 ) 043509 [ arxiv:1104.2219 [ hep - ph ] ] . k. j. bae , e. j. chun and s. h. i m , `` cosmology of the dfsz axino , '' jcap * 1203 * ( 2012 ) 013 [ arxiv:1111.5962 [ hep - ph ] ] . k. j. bae , k. choi and s. h. i m , `` effective interactions of axion supermultiplet and thermal production of axino dark matter , '' jhep * 1108 * ( 2011 ) 065 [ arxiv:1106.2452 [ hep - ph ] ] . k. j. bae , h. baer , a. lessa and h. serce , `` coupled boltzmann computation of mixed axion neutralino dark matter in the susy dfsz axion model , '' arxiv:1406.4138 [ hep - ph ] . for a review on the ksvz axino , see , k. -y . choi , j. e. kim and l. roszkowski , `` review of axino dark matter , '' j. korean phys . * 63 * ( 2013 ) 1685 [ arxiv:1307.3330 [ astro-ph.co ] ] . a. brandenburg , l. covi , k. hamaguchi , l. roszkowski and f. d. steffen , `` signatures of axinos and gravitinos at colliders , '' phys . b * 617 * , 99 ( 2005 ) [ hep - ph/0501287 ] . k. hamaguchi , m. m. nojiri and a. de roeck , `` prospects to study a long - lived charged next lightest supersymmetric particle at the lhc , '' jhep * 0703 * , 046 ( 2007 ) [ hep - ph/0612060 ] . a. freitas , f. d. steffen , n. tajuddin and d. wyler , `` axinos in cosmology and at colliders , '' jhep * 1106 * , 036 ( 2011 ) [ arxiv:1105.1113 [ hep - ph ] ] . h. baer , v. barger , a. lessa , w. sreethawong and x. tata , `` wh plus missing - et signature from gaugino pair production at the lhc , '' phys . d * 85 * , 055022 ( 2012 ) [ arxiv:1201.2949 [ hep - ph ] ] . k. howe and p. saraswat , `` excess higgs production in neutralino decays , '' jhep * 1210 * , 065 ( 2012 ) [ arxiv:1208.1542 [ hep - ph ] ] . a. arbey , m. battaglia and f. mahmoudi , `` higgs production in neutralino decays in the mssm - the lhc and a future e+e- collider , '' arxiv:1212.6865 [ hep - ph ] . t. han , s. padhi and s. su , `` electroweakinos in the light of the higgs boson , '' phys . d * 88 * , 115010 ( 2013 ) [ arxiv:1309.5966 [ hep - ph ] ] . s. jung , `` resolving the existence of higgsinos in the lhc inverse problem , '' jhep * 1406 * , 111 ( 2014 ) [ arxiv:1404.2691 [ hep - ph ] ] . d. ghosh , m. guchait and d. sengupta , `` higgs signal in chargino - neutralino production at the lhc , '' eur . j. c * 72 * , 2141 ( 2012 ) [ arxiv:1202.4937 [ hep - ph ] ] . a. bharucha , s. heinemeyer and f. von der pahlen , `` direct chargino - neutralino production at the lhc : interpreting the exclusion limits in the complex mssm , '' eur . j. c * 73 * , 2629 ( 2013 ) [ arxiv:1307.4237 ] . f. yu , `` anatomizing exotic production of the higgs boson , '' arxiv:1404.2924 [ hep - ph ] . g. aad _ et al . _ [ atlas collaboration ] , `` search for direct production of charginos and neutralinos in events with three leptons and missing transverse momentum in @xmath131 = 8 tev @xmath132 collisions with the atlas detector , '' arxiv:1402.7029 [ hep - ex ] . v. khachatryan _ et al . _ [ cms collaboration ] , `` searches for electroweak production of charginos , neutralinos , and sleptons decaying to leptons and w , z , and higgs bosons in pp collisions at 8 tev , '' arxiv:1405.7570 [ hep - ex ] . [ atlas collaboration ] , `` search for supersymmetry in events with four or more leptons in 21@xmath133fb@xmath134 of pp collisions at @xmath135tev with the atlas detector , '' atlas - conf-2013 - 036 . g. aad _ et al . _ [ atlas collaboration ] , `` search for direct production of charginos , neutralinos and sleptons in final states with two leptons and missing transverse momentum in @xmath132 collisions at @xmath136 8 tev with the atlas detector , '' jhep * 1405 * , 071 ( 2014 ) [ arxiv:1403.5294 [ hep - ex ] ] . s. gori , s. jung and l. -t . wang , `` cornering electroweakinos at the lhc , '' jhep * 1310 * , 191 ( 2013 ) [ arxiv:1307.5952 [ hep - ph ] ] . c. han , a. kobakhidze , n. liu , a. saavedra , l. wu and j. m. yang , `` probing light higgsinos in natural susy from monojet signals at the lhc , '' jhep * 1402 * , 049 ( 2014 ) [ arxiv:1310.4274 [ hep - ph ] ] . p. schwaller and j. zurita , `` compressed electroweakino spectra at the lhc , '' jhep * 1403 * , 060 ( 2014 ) [ arxiv:1312.7350 [ hep - ph ] ] . h. baer , a. mustafayev and x. tata , `` monojets and mono - photons from light higgsino pair production at lhc14 , '' phys . d * 89 * , 055007 ( 2014 ) [ arxiv:1401.1162 [ hep - ph ] ] . z. han , g. d. kribs , a. martin and a. menon , `` hunting quasidegenerate higgsinos , '' phys . d * 89 * , 075007 ( 2014 ) [ arxiv:1401.1235 [ hep - ph ] ] . m. low and l. -t . wang , `` neutralino dark matter at 100 tev , '' arxiv:1404.0682 [ hep - ph ] . s. d. thomas and j. d. wells , `` phenomenology of massive vectorlike doublet leptons , '' phys . lett . * 81 * , 34 ( 1998 ) [ hep - ph/9804359 ] . the atlas collaboration , `` search for charginos nearly mass - degenerate with the lightest neutralino based on a disappearing - track signature in @xmath132 collisions at @xmath137 with the atlas detector , '' atlas - conf-2013 - 069 . [ atlas collaboration ] , `` expected performance of the atlas experiment - detector , trigger and physics , '' arxiv:0901.0512 [ hep - ex ] . s. p. martin , `` collider signals from slow decays in supersymmetric models with an intermediate scale solution to the mu problem , '' phys . d * 62 * , 095008 ( 2000 ) [ hep - ph/0005116 ] . p. meade , m. reece and d. shih , `` long - lived neutralino nlsps , '' jhep * 1010 * , 067 ( 2010 ) [ arxiv:1006.4575 [ hep - ph ] ] . p. w. graham , d. e. kaplan , s. rajendran and p. saraswat , `` displaced supersymmetry , '' jhep * 1207 * , 149 ( 2012 ) [ arxiv:1204.6038 [ hep - ph ] ] . s. bobrovskyi , j. hajer and s. rydbeck , `` long - lived higgsinos as probes of gravitino dark matter at the lhc , '' jhep * 1302 * , 133 ( 2013 ) [ arxiv:1211.5584 [ hep - ph ] ] . p. meade , m. reece and d. shih , `` prompt decays of general neutralino nlsps at the tevatron , '' jhep * 1005 * , 105 ( 2010 ) [ arxiv:0911.4130 [ hep - ph ] ] . k. t. matchev and s. d. thomas , `` higgs and @xmath27 boson signatures of supersymmetry , '' phys . d * 62 * ( 2000 ) 077702 [ hep - ph/9908482 ] . h. baer , p. g. mercadante , x. tata and y. -l . wang , `` the reach of tevatron upgrades in gauge mediated supersymmetry breaking models , '' phys . d * 60 * , 055001 ( 1999 ) [ hep - ph/9903333 ] . s. dimopoulos , m. dine , s. raby and s. d. thomas , `` experimental signatures of low - energy gauge mediated supersymmetry breaking , '' phys . rev . lett . * 76 * , 3494 ( 1996 ) [ hep - ph/9601367 ] . s. dimopoulos , m. dine , s. raby , s. d. thomas and j. d. wells , `` phenomenological implications of low - energy supersymmetry breaking , '' nucl . phys . proc . suppl . * 52a * , 38 ( 1997 ) [ hep - ph/9607450 ] . h. baer , p. g. mercadante , x. tata and y. -l . wang , `` the reach of the cern large hadron collider for gauge mediated supersymmetry breaking models , '' phys . d * 62 * , 095007 ( 2000 ) [ hep - ph/0004001 ] . j. t. ruderman and d. shih , `` general neutralino nlsps at the early lhc , '' jhep * 1208 * , 159 ( 2012 ) [ arxiv:1103.6083 [ hep - ph ] ] . cms collaboration [ cms collaboration ] , `` search for long - lived neutral particles decaying to dijets , '' cms - pas - exo-12 - 038 . cms collaboration [ cms collaboration ] , `` search for long - lived particles decaying to final states that include dileptons , '' cms - pas - exo-12 - 037 . v. m. abazov _ et al . _ [ d0 collaboration ] , phys . lett . * 103 * , 071801 ( 2009 ) [ arxiv:0906.1787 [ hep - ex ] ] . t. aaltonen _ et al . _ [ cdf collaboration ] , phys . d * 85 * , 012007 ( 2012 ) [ arxiv:1109.3136 [ hep - ex ] ] . g. aad _ et al . _ [ atlas collaboration ] , phys . lett . * 108 * , 251801 ( 2012 ) [ arxiv:1203.1303 [ hep - ex ] ] . g. aad _ et al . _ [ atlas collaboration ] , phys . b * 721 * , 32 ( 2013 ) [ arxiv:1210.0435 [ hep - ex ] ] . cms collaboration [ cms collaboration ] , cms - pas - b2g-12 - 024 . s. jung and j. d. wells , `` gaugino physics of split supersymmetry spectrum at the lhc and future proton colliders , '' phys . rev . d * 89 * , 075004 ( 2014 ) [ arxiv:1312.1802 [ hep - ph ] ] . g. aad _ et al . _ [ atlas collaboration ] , `` triggers for displaced decays of long - lived neutral particles in the atlas detector , '' jinst * 8 * , p07015 ( 2013 ) [ arxiv:1305.2284 [ hep - ex ] ] . j. f. gunion and h. e. haber , `` two - body decays of neutralinos and charginos , '' phys . d * 37 * , 2515 ( 1988 ) . m. ibe , s. matsumoto and r. sato , `` mass splitting between charged and neutral winos at two - loop level , '' phys . b * 721 * , 252 ( 2013 ) [ arxiv:1212.5989 [ hep - ph ] ] . j. beringer _ [ particle data group collaboration ] , `` review of particle physics ( rpp ) , '' phys . d * 86 * , 010001 ( 2012 ) . j. alwall , m. herquet , f. maltoni , o. mattelaer and t. stelzer , `` madgraph 5 : going beyond , '' jhep * 1106 * , 128 ( 2011 ) [ arxiv:1106.0522 [ hep - ph ] ] . t. sjostrand , s. mrenna and p. z. skands , `` pythia 6.4 physics and manual , '' jhep * 0605 * , 026 ( 2006 ) [ hep - ph/0603175 ] . m. l. mangano , m. moretti , f. piccinini and m. treccani , `` matching matrix elements and shower evolution for top - quark production in hadronic collisions , '' jhep * 0701 * , 013 ( 2007 ) [ hep - ph/0611129 ] . m. cacciari , g. p. salam and g. soyez , `` fastjet user manual , '' eur . j. c * 72 * , 1896 ( 2012 ) [ arxiv:1111.6097 [ hep - ph ] ] .

it is known that the spectral energy distribution ( sed ) of the nuclear radiation of narrow - line seyfert 1 galaxies ( nls1s ) has different shapes with respect to that of ordinary broad - line seyfert 1 galaxies ( bls1s ) , particularly in wavelengths of x - ray . this may cause some differences in the ionization degree and the temperature of gas in narrow - line regions ( nlrs ) between nls1s and bls1s . this paper aims to examine whether or not there are such differences in the physical conditions of nlr gas between them . for this purpose , we have compiled the emission - line ratios of 36 nls1s and 83 bls1s from the literature . comparing these two samples , we have found that the line ratios of [ o i]/[o iii] and [ o iii]/[o iii] , which represent the ionization degree and the gas temperature respectively , are statistically indistinguishable between nls1s and bls1s . based on new photoionization model calculations , we show that these results are not inconsistent with the difference of the sed between them . the influence of the difference of seds on the highly ionized emission lines is also briefly discussed .

astro-ph0008006

seyfert nuclei are typical active galactic nuclei ( agns ) in the nearby universe . they have been broadly classified into two types based on presence or absence of broad emission lines in their optical spectra ( khachikian & weedman 1974 ) ; seyferts with broad lines are type 1 ( hereafter s1 ) while those without broad lines are type 2 ( s2 ) . these two types of seyfert nuclei are now unified by introducing the viewing angle dependence toward the central engine surrounded by the geometrically and optically thick dusty torus ( antonucci & miller 1985 ; see for a review antonucci 1993 ) . in addition to these typical types , narrow - line seyfert 1 galaxies ( nls1s ) have also been recognized as a distinct type of seyfert nuclei . the optical emission - line properties of nls1s are summarized as follows ( e.g. , osterbrock & pogge 1985 ) . ( 1 ) the balmer lines are only slightly broader than the forbidden lines such as [ o iii]@xmath05007 ( typically less than 2000 km s@xmath1 ) . this property makes nls1s a distinct type of ordinary broad - line s1s ( bls1s ) . ( 2 ) the [ o iii]@xmath05007/h@xmath2 intensity ratio is smaller than 3 . this criterion was introduced to discriminate s1s from s2s by shuder & osterbrock ( 1981 ) . and , ( 3 ) they present strong fe ii emission lines which are often seen in s1s but generally not in s2s . moreover , the x - ray spectra of nls1s are very steep ( puchnarewicz et al . 1992 ; boller , brandt , & fink 1996 ; wang , brinkmann , & bergeron 1996 ; vaughan et al . 1999 ; leighly 1999b ) and highly variable ( boller et al . 1996 ; turner et al . 1999 ; leighly 1999a ) . because of these complex properties , it is still not understood what nls1s are in the context of the current agn unified model . in order to understand what nls1s are , it is important to investigate the narrow - line regions ( nlrs ) of nls1s because of the following two reasons . first , the intrinsic spectral energy distribution ( sed ) of the nuclear radiation of nls1s is rather different from that of bls1s ; i.e. , the soft and hard x - ray spectra of nls1s are steeper than those of bls1s ( boller et al . 1996 ; brandt , mathur , & elvis 1997 ; vaughan et al . 1999 ; leighly 1999b ) . it is often considered that the nlrs are photoionized by the nonthermal continuum radiation from central engines ( yee 1980 ; shuder 1981 ; cohen 1983 ; cruz - gonzlez et al . 1991 ; osterbrock 1993 ; evans et al . 1999 ) though shock ionization may play an important role of ionization of the nlr ( e.g. , contini & aldrovandi 1983 ; viegas - aldrovandi & contini 1989 ; dopita & sutherland 1995 ) . if the dominant mechanism of ionization is the photoionization , the degree or the structure of ionization of nlrs in nls1s may be different from that of bls1s . such difference can be probed by forbidden emission - line ratios . second , it is known that there are some differences between the nlr properties of s1s and s2s , for example , the gas temperature in the [ o iii ] zone ( e.g. , heckman & balick 1979 ; shuder & osterbrock 1981 ) . although the reason of such differences has not yet been understood fully , it is meaningful to investigate how the nlrs in nls1s share the properties with those in s1s or those in s2s . analyzing optical spectra of 7 nls1s and 16 bls1s , recently , rodrguez - ardila , pastoriza , & donzelli ( 2000b ) and rodrguez - ardila et al . ( 2000a ) reported that the nlrs of nls1s are less excited than those of the bls1s . they suggested that this is due to the difference in the shape of the seds of nuclear radiation between nls1s and bls1s . in their analysis they used the intensities of the forbidden lines normalized by the narrow components of balmer lines . however , it is not clear whether or not the `` narrow components '' of the balmer lines of nls1s are radiated from only the nlrs . for example , line widths of the balmer lines radiated from broad - line regions ( blrs ) may be narrow like nlr emission if we see nls1s from a nearly pole - on viewing angle ( taniguchi , murayama , & nagao 1999 and reference therein ) . therefore it seems better to use some combinations among forbidden emission lines . in this paper , we present the comparisons of some emission - line flux ratios among nls1s and bls1s ( see nagao , taniguchi , & murayama 2000c for highly ionized emission lines ) using the data compiled from the literature . in order to investigate the properties of the nlrs in nls1s and in bls1s , we compiled the following emission lines from the literature ; [ o i]@xmath06300 , [ o ii]@xmath33726,3729 , [ o iii]@xmath05007 , [ o iii]@xmath04363 , [ n ii]@xmath06583 , and [ s ii]@xmath36717,6731 . these lines are respectively referred as [ o i ] , [ o ii ] , [ o iii]@xmath05007 , [ o iii]@xmath04363 , [ n ii ] , and [ s ii ] below . as mentioned in section 1 , we do not use the flux of the balmer lines in order to avoid any ambiguity . the number of compiled objects is 119 ; 36 nls1s and 83 bls1s . the so - called seyfert 1.2 galaxies ( see osterbrock 1977 and winkler 1992 ) are also included in the bls1 sample . all the seyfert galaxies are listed up in table 1 together with their redshifts and 60@xmath4 m luminosities = 50 km s@xmath1 mpc@xmath1 and a deceleration parameter _ q_@xmath5 = 0 . ] . the 60 @xmath4 m luminosities are taken from the _ iras _ faint source catalogue ( moshir et al . 1992 ) . the emission - line flux ratios for each object are given in table 2 . each ratio is the averaged value among the references given in table 1 . since it is often difficult to measure the narrow balmer component for s1s accurately , there might be the systematic error if we make reddening corrections using the balmer decrement method ( e.g. , osterbrock 1989 ) for both types of seyferts . therefore , we do not make the reddening correction . the effect of dust extinction on our results is discussed in section 3.3 . some galaxies do not have all the emission - line ratios . the lack of the data in table 2 is attributed to the following five reasons ; ( 1 ) the observation did not cover the wavelengths where the emission lines exist , ( 2 ) the emission - line was not detected and the upper limit is not given in the reference , ( 3 ) the only upper limit is given in the reference , ( 4 ) the flux of the emission lines were not given in the reference because the author(s ) of the reference were not interested in those emission lines , and ( 5 ) the de - blending of the [ o iii]@xmath04363 emission from the h@xmath6 emission and the [ n ii]@xmath06583 emission from the h@xmath7 emission were not performed in the reference . since the number of the upper - limit data is quite small and those values are too large to be used for any scientific discussion , we do not use these upper - limit data in later analyses . because we do not impose any selection criteria upon our samples , it is necessary to check whether or not the two samples are appropriate for our comparative study . if there are some systematic differences in the redshift distribution and in the intrinsic agn power distribution between the two samples , there would be possible biases . first we investigate the redshift distribution . we show the histograms of the redshift in figure 1 . the average redshifts and 1@xmath8 deviations are 0.0568 @xmath9 0.0502 for the nls1s and 0.1141 @xmath9 0.1455 for the bls1s . it is noted that the average redshift of the bls1s is a little higher than that of the nls1s . in order to investigate whether or not the frequency distributions of the redshift are statistically different between two samples , we apply the kolmogorov - smirnov ( ks ) statistical test ( press et al . 1988 ) . the null hypothesis is that the redshift distributions of the nls1s and the bls1s come from the same underlying population . the resultant ks probability is 4.650 @xmath10 10 @xmath1 , which means that the two distributions are statistically indistinguishable . hence we conclude that there is no redshift bias . second , we investigate whether or not the intrinsic agn power is systematically different between the two samples using the _ iras _ 60@xmath4 m luminosity , which is regarded as a rather isotropic emission ( pier & krolik 1992 ; efstathiou & rowan - robinson 1995 ; fadda et al . 1998 ) . the 60@xmath4 m luminosity is thought to scale the nuclear continuum radiation which is absorbed and re - radiated by the dusty torus ( see storchi - bergmann , mulchaey , & wilson 1992 ) . the histograms of the 60@xmath4 m luminosity are shown in figure 2 . the average 60@xmath4 m luminosities and 1@xmath8 deviations in logarithm ( in units of solar luminosity ) are 11.646 @xmath9 0.646 for the nls1s and 11.627 @xmath9 0.487 for the bls1s . we apply the ks test where the null hypothesis is that the distribution of the 60@xmath4 m luminosity of the two samples come from the same underlying population . the resultant ks probability is 3.884 @xmath11 , which means that there is no systematic difference of the 60@xmath4 m luminosity between two samples . although the 60@xmath4 m luminosity might be contaminated with the influence of circumnuclear star formation and have the weak unisotropic tendency , this test supports the validity of the statistical comparisons in our study . in figure 3 we also show that the line ratios which we compiled do not correlate with the redshift and the 60@xmath4 m luminosity . to investigate whether or not the ionization degree of the nlrs is different between nls1s and bls1s , we compare some emission - line ratios between the two samples . although emission - line ratios of agns have been traditionally discussed in the form normalized by a narrow component of balmer lines , for example [ o iii]@xmath05007/h@xmath2 , we do not use such line ratios because of the difficulty in deblending of the narrow component of balmer lines from the broad component . in this study , we investigate the ionization degree of nlrs using [ o i ] , [ o ii ] , and [ o iii]@xmath05007 . the ionization potentials of the lower stage of ionization ( to produce the relevant ions ) are 0.0 ev , 13.6 ev , and 35.1 ev , respectively . this set is free from the chemical abundance effect . here we use [ o i]/[o iii]@xmath05007 and [ o ii]/[o iii]@xmath05007 . it is noted that the critical density of [ o iii]@xmath05007 is similar to that of [ o i ] ( 7.0 @xmath10 10@xmath12 @xmath13 and 1.8 @xmath10 10@xmath14 @xmath13 , respectively ) and far larger than that of [ o ii ] ( 4.5 @xmath10 10@xmath15 @xmath13 ) . this may mean that [ o i ] and [ o iii]@xmath05007 are radiated from the similar region in nlrs while the [ o ii ] emission comes from relatively lower - density clouds . for example , it may be such the case that matter - bounded parts of a relatively high - density cloud radiate the [ o iii]@xmath05007 emission while ionization - bounded parts of the same cloud radiate the [ o i ] emission ( see figure 4b of binette , wilson , & storchi - bergmann 1996 ) . therefore [ o i]/[o iii]@xmath05007 seems better to investigate the ionization degree of gas clouds in nlrs than [ o ii]/[o iii]@xmath05007 . we show the histograms of these emission - line ratios for the nls1s and the bls1s in figure 4 . though the nls1s appear to have larger [ o ii]/[o iii]@xmath05007 than the bls1s , it seems that there is no systematic difference in [ o i]/[o iii]@xmath05007 between the two samples . in order to investigate whether or not these distributions of the emission - line ratios are statistically different between the two samples , we apply the ks test . the null hypothesis is that the distributions of the relevant ratio of the samples come from the same underlying population . the ks probabilities are 7.089 @xmath16 for [ o i]/[o iii]@xmath05007 and 6.318 @xmath17 for [ o ii]/[o iii]@xmath05007 , which lead to the following results . 1 ) there is no statistical difference in [ o i]/[o iii]@xmath05007 between two samples . 2 ) it is , however , not clear whether or not there is statistical difference in [ o ii]/[o iii]@xmath05007 between them . these results seem to suggest that there is little difference of the ionization degree of the nlrs between the two types of seyferts . if this is true , it is contradictory to the result of rodrguez - ardila et al . ( 2000b ) , who reported that nls1s are less excited objects than bls1s . to make this issue clear , we have carried out the model calculations , which is presented in section 4 . we investigate the [ o iii]@xmath04363/[o iii]@xmath05007 ratio , which is sensitive to the gas temperature ( e.g. , osterbrock 1989 ) . in figure 5 , we show the histograms of [ o iii]@xmath04363/[o iii]@xmath05007 for the nls1s and the bls1s . in order to investigate whether or not these distributions of both samples are statistically different , we also apply the ks test where the null hypothesis is that the distributions of the emission - line ratio between two samples come from the same underlying population . the ks probability is 7.877 @xmath11 , which means that there is no statistical difference in this line ratio between the nls1s and the bls1s . as mentioned in section 2.1 , no reddening correction has been made for all the collected emission - line ratios analyzed here . however , it is known that dust grains are present in the nlr of seyferts ( e.g. , dahari & de robertis 1988a , 1988b ; netzer & laor 1993 ) . hence we check how the extinction affects the emission - line ratios discussed in previous sections . because the difference of the average amounts of the extinction between s1s and s2s is about 1 magnitude ( dahari & de robertis 1988a ; see also de zotti & gaskell 1985 ) , we investigate the extinction effect in the case of @xmath18 = 1.0 mag using the cardelli s extinction curve ( cardelli , clayton , & mathis 1989 ) . correction factors for the observed values of [ o i]/[o iii]@xmath05007 , [ o ii]/[o iii]@xmath05007 , and [ o iii]@xmath04363/[o iii]@xmath05007 for the extinction ( @xmath18 = 1.0 mag ) are 0.786 , 1.471 , and 1.222 , respectively . these values correspond to about a half bin in figures 3 , 4 , and 5 . this suggests that the extinction might affect the results in sections 3.1 and 3.2 @xmath19 there is a systematic difference in the amounts of the extinction much more than 1 magnitude between nls1s and bls1s . however , the sample of rodrguez - ardila et al . ( 2000b ) showed little difference in the amounts of the extinction between nls1s and bls1s : @xmath18 = 0.457 @xmath9 0.137 for 7 nls1s and 0.663 @xmath9 0.345 for 16 bls1s . though the number of objects is small , this suggests that the difference in the amounts of the extinction is so small that the extinction does not affect the results presented in previous sections . now we must consider the following problem . it has been known that the shape of seds of nuclear radiation is different between nls1s and bls1s particularly in x - ray band . since uv to x - ray photons are closely connected with the photoionization process , such difference in the sed may cause some distinctions in physical properties of the ionized gas in nlrs , such as the ionization degree and the temperature . on the other hand , our comparative study described in section 3 suggests that there is little difference in the ionization degree and in the temperature of the gas in nlrs between nls1s and bls1s . is this result plausible in terms of photoionization models ? in order to investigate this issue , we carry out photoionization model calculations and compare the model results with the compiled emission - line ratios . up to now , many efforts have been made to reveal the difference of the seds between nls1s and bls1s . we summarize such studies and construct template seds for the nls1s and the bls1s which will be used in the following model calculations . first , we mention the infrared properties of nls1s and bls1s . rodrguez - pascual , mas - hesse , & santos - lle ( 1997 ) pointed out that the fir properties of the nls1s and the bls1s are very similar to each other . murayama , nagao , & taniguchi ( 1999 ) have reported that the mid - infrared properties of the nls1s are also similar to those of the bls1s . therefore we assume that the infrared properties of nls1s are nearly the same as those of bls1s . second , we mention the x - ray properties of nls1s and bls1s . boller et al . ( 1996 ) revealed that nls1s have generally steeper soft x - ray spectra observed by _ rosat _ than bls1s . the weighted mean soft x - ray photon index for their sample of nls1s is 3.13 with an uncertainty in the mean of less than 0.03 . this is statistically larger than that of bls1s : the weighted mean soft x - ray photon index for the 51 bls1s in the sample of walter & fink ( 1993 ) is 2.34 and the uncertainty in this mean is 0.03 ( see boller et al . moreover , it is known that the hard x - ray spectra of nls1s are also steeper than those of bls1s . brandt et al . ( 1997 ) gave the average photon indices of the hard x - ray spectra observed by _ asca _ for 15 nls1s and 19 bls1s : the mean hard x - ray photon index of the nls1s is 2.15 where the variance of this value is 0.036 and the standard error is 0.049 , and the mean hard x - ray photon index of the bls1s is 1.87 where the variance of this value is 0.025 and the standard error is 0.036 . third , we mention optical to x - ray properties of nls1s and bls1s . the ratios of optical ( i.e. , 2500 @xmath20 ) to x - ray flux at 2 kev are parameterized using @xmath21 , which is defined as @xmath22}{\log[\nu({\rm 2 kev})/\nu(2500 { \rm\aa})]}\ ] ] ( tananbaum et al . 1979 ) . the average @xmath21 for optically - selected radio - quiet agns is 1.46@xmath23 ( zamorani et al . the mean value of @xmath21 derived by puchnarewicz et al . ( 1996 ) , whose sample is x - ray selected one , is harder than the others : 1.14 @xmath9 0.18 . in order to investigate whether or not this value is systematically different between nls1s and bls1s , we compare @xmath21 of 10 nls1s and 28 bls1s taken from the sample of walter & fink ( 1993 ) . it is noted that the values of @xmath21 for the sample of walter & fink ( 1993 ) have slightly different from those described as equation ( 1 ) because they measured the optical continuum flux at 2675 @xmath20 , not at 2500 @xmath20 : this leads to a difference of 0.02 in @xmath21 ( see puchnarewicz et al . the average spectral indices and 1@xmath8 deviations for the nls1s and the bls1s are 1.31 @xmath9 0.16 and 1.36 @xmath9 0.24 , respectively . the ks probability that the underlying distribution of these two distributions are the same is 5.984 @xmath10 10@xmath1 . therefore there is little or no difference in @xmath21 between nls1s and bls1s . this seems to be rather contradictory to some previous works ( walter & fink 1993 ; laor et al . 1994 ; puchnarewicz et al . 1996 ) , which claimed the existence of the correlation between @xmath21 and the x - ray spectral index , because the large x - ray spectral index is one of the characteristic properties of nls1s . the reason why such a complex situation is caused may be that some of nls1s in the sample of walter & fink ( 1993 ) is out of the correlation ( see figure 8 of walter & fink 1993 ) although it is not clear whether or not this property is a general one in nls1s . here we construct the template seds of the nls1 and the bls1 taking the above observational properties into account . we adopt the following function for the templates : @xmath24 ( see ferland 1996 ) . we adopt the following parameter values . ( i ) @xmath25 is the infrared cutoff of the so - called big blue bump component , and we fix @xmath25 = 0.01 ryd following to ferland ( 1996 ) . ( ii ) @xmath26 is the slope of the low - energy side of the big blue bump component . we adopt @xmath26 = 0.5 , which is the typical value for agns ( ferland 1996 ; see also francis 1993 ) . note that the photoionization process is not sensitive to this parameter . ( iii ) @xmath21 is the uv to x - ray spectral slope mentioned above , which determines the parameter @xmath27 in the equation ( 2 ) . we adopt @xmath21 = 1.35 , which is the average value for the sample of walter & fink ( 1993 ) mentioned above , for both sample . however , there are some claims that this parameter correlates to the x - ray spectral index ( walter & fink 1993 ; laor et al . 1994 puchnarewicz et al . 1996 ) , that is , @xmath21 may be different between nls1s and bls1s . hence we check the dependence of the calculation outputs on @xmath21 in section 4.3.3 . ( iv ) @xmath28 is the slope of the x - ray component . we adopt @xmath28 = 1.15 for nls1s and 0.85 for bls1s corresponding to the observational results by @xmath29 described in section 4.1.1 . this power - law component is not extrapolated below 1.36 ev or above 100 kev . below 1.36 ev , this term is set to zero , while above 100 kev , the continuum is assumed to fall off as @xmath30 . and finally , ( v ) @xmath31 is the parameter which characterizes the shape of the big blue bump . we choose this parameter to reproduce the soft x - ray spectral index measured by _ rosat _ described in section 4.1.1 . it results in 1,180,000 k for nls1s and 490,000 k for bls1s . they correspond to @xmath32 = 3.13 and 2.35 , respectively . the template seds constructed in such a way are shown in figure 6 ; hereafter we refer the nls1 sed and the bls1 sed , respectively . it is notable that these template seds are not theoretically predicted ones , but the empirical ones . although it has not been understood whether or not the soft excess component is well described by a blackbody , pounds et al . ( 1994 ) mentioned that the soft excess can be characterized by a blackbody of @xmath33 = 70 @xmath9 10 ev [ or @xmath34 k ] . this suggests that the temperature of our adopted seds is not too high . puchnarewicz et al . ( 1996 ) also mentioned that the soft excess may be represented by thermal bremsstrahlung with @xmath35 = 10@xmath14 k. mineshige et al . ( 2000 ) proposed a slim disk model whose maximum blackbody temperature is @xmath36 kev [ or @xmath37 for the soft excess of nls1s where @xmath38 is the mass of a supermassive black hole . these studies are almost consistent with our empirical template seds . we perform photoionization model calculations using the spectral synthesis code _ cloudy _ version 90.04 ( ferland 1996 ) , which solves the equations of statistical and thermal equilibrium and produces a self - consistent model of the run of temperature as a function of depth into the nebula . here we assume an uniform density gas cloud with a plane - parallel geometry . the parameters for the calculations are ( 1 ) the hydrogen density of the cloud ( @xmath39 ) , ( 2 ) the ionization parameter ( @xmath40 ) , which is defined as the ratio of the ionizing photon density to the electron density , ( 3 ) the chemical compositions of the gas , and ( 4 ) the shape of the input sed . we perform several model runs covering the following ranges of parameters : 10@xmath15 @xmath41 @xmath13 ( and 10@xmath42 @xmath13 in section 4.3.4 ) and 10@xmath43 . we set the gas - phase elemental abundance to be either solar or subsolar . the adopted solar abundances relative to hydrogen are taken from grevesse & anders ( 1989 ) with extensions by grevesse & noels ( 1993 ) . the subsolar abundances are assumed that 90% of mg , si , and fe , 50% of c and o , and 25% of n and s are locked into dust grains , as estimated for the orion h ii region ( e.g. , baldwin et al . 1991 , 1996 ) . for the input seds , we use the two types of sed : the nls1 sed and the bls1 sed , mentioned in the last section . the calculations are stopped when the temperature fall to 3000 k , below which gas does not contribute significantly to the optical emission lines . we show the results of the model calculations in the case of the solar abundances and compare them to the observations in figure 7 , which is a diagram of [ o i]/[o iii]@xmath05007 versus [ o ii]/[o iii]@xmath05007 . this diagram has been used to discuss the physical properties of ionized gas traditionally ( e.g. , heckman 1980 ; baldwin , phillips , & terlevich 1981 ; ferland & netzer 1983 ; evans et al . it is shown that there is a slight difference of [ o i]/[o iii]@xmath05007 between the two model grids . the reason for this difference is thought as follows . the relative intensity of soft x - ray of nls1s is stronger than that of bls1s . this results in a larger partially ionized zone in nlrs of nls1s . since [ o i ] is selectively radiated from such partially ionized zone because its ionization potential is close to the ionization potential of hydrogen , nls1s tend to exhibit stronger [ o i ] . however , the dispersion of the compiled data is larger than this difference . this means that _ the line ratios used in this diagram are insensitive to the difference of the shape of the template seds_. this result is consistent with the previous work of rodrguez - ardila et al . ( 2000a ) . they presented their photoionization model calculations assuming two types of seds ; i.e. , the nls1-like sed and the bls1-like sed . as shown in figure 7 of rodrguez - ardila et al . ( 2000a ) , the ratio of [ o i]/[o iii]@xmath05007 is not so different between the model nls1 and the model bls1 ( less than factor 3 ) . the comparison between the models and the observations shown in figure 7 suggests @xmath44 and @xmath45 for the nlrs of both samples . the estimated @xmath40 values seem to be rather lower than those calculated in some previous works ( e.g. , ferland & netzer 1983 ; ho , shields , & filippenko 1993 ) . in order to make it clear that this is not due to any selection effect of the samples , we show model grids on the diagram of [ o iii]@xmath05007/h@xmath2 versus [ n ii]/h@xmath7 , which is a familiar diagnostic diagram proposed by veilleux & osterbrock ( 1987 ) , in figure 8 . comparing this with figure 4 of veilleux & osterbrock ( 1987 ) , typical seyferts are reproduced by the models with @xmath45 . this discrepancy between the result by us and by previous literature is thought to be partly because the following reason . the energy peak of the template seds in our models is at rather high energy than those in previous studies . accordingly the relative amounts of photons whose energy is near the ionization potential of hydrogen increase . this leads to the lower @xmath40 because this parameter is defined using _ all _ photons which exceed the ionization potential of hydrogen although the photoionization is effective in the energy of near the ionization potential of hydrogen . we investigate the gas properties with another diagnostic diagram : [ s ii]/[o iii]@xmath05007 versus [ n ii]/[o iii]@xmath05007 ( figure 9 ) . it results in that the derived ranges in both @xmath39 and @xmath40 are consistent with those obtained in figure 7 . there is very little difference between the model grids for nls1s and those for bls1s . it is notable that the scatter of the plotted data in this diagram is larger than that in figure 6 . this may be due to that the deblending h@xmath7 from [ n ii ] is not be well done in some case if the spectral resolution is not so high . if this is the case , the flux measurement of h@xmath7 may not also be well done . it means that it is dangerous to use traditional emission - line ratios such as [ n ii]/h@xmath7 , [ s ii]/h@xmath7 , and [ o i]/h@xmath7 for s1s . alternatively , the scatter may reflect the variety of the nitrogen abundance because some previous works reported that some of seyferts show evidences in favor of a nitrogen overabundance ( storchi - bergmann & pastoriza 1990 ; storchi - bergmann 1991 ; storchi - bergmann et al . 1998 ) . in any case , the diagram of [ s ii]/[o iii]@xmath05007 versus [ n ii]/[o iii]@xmath05007 is less suitable to discuss the properties of gas in nlrs than that of [ o i]/[o iii]@xmath05007 versus [ o ii]/[o iii]@xmath05007 . in figures 10 and 11 , we show the result of the model calculations for the case of the subsolar abundances and compare them with the observations on the diagrams of [ o i]/[o iii]@xmath05007 versus [ o ii]/[o iii]@xmath05007 and [ s ii]/[o iii]@xmath05007 versus [ n ii]/[o iii]@xmath05007 , respectively . the loci of model grids in figure 10 slightly shift to be larger in [ o i ] than those in figure 7 . this may be attributed to the fact that the partially ionized region become thicker due to the decrease of the heavy elements . however , the estimated parameters , @xmath39 and @xmath40 , are almost the same as those in the case of the solar abundances . as mentioned in section 3.2 , the temperature - sensitive emission - line ratio , [ o iii]@xmath04363/[o iii]@xmath05007 , is scarcely different between the nls1s and the bls1s though the intrinsic seds are clearly different between them . in order to investigate whether or not this difference in seds causes a detectable difference in the temperature of nlr gas through photoionization processes , we carry out the model calculations concerning the [ o iii]@xmath04363/[o iii]@xmath05007 ratio . in figure 12 , we show the diagram of [ o iii]@xmath04363/[o iii]@xmath05007 versus [ o i]/[o iii]@xmath05007 for the case of the solar abundances . it is shown that the very high density condition ( 10@xmath46 or higher ) is needed to explain the observed [ o iii]@xmath04363/[o iii]@xmath05007 ratios for both nls1s and bls1s . this derived @xmath39 is far higher than the value obtained using the diagnostic diagrams presented in section 4.3.1 . the reason for this is partly because of the oversimplification of the photoionization models : the [ o iii]@xmath04363/[o iii]@xmath05007 ratio is difficult to be reproduced by one - zone photoionization models as mentioned in other works ( e.g. , filippenko & halpern 1984 ; tadhunter , robinson , & morganti 1989 ; wilson , binette , & storchi - bergmann 1997 ; nagao , murayama , & taniguchi 2000a ) . we are not going to make further discussion for this problem because this is out of the purpose of this paper . in this diagram , the loci of the model grid for nls1s slightly shift to be larger [ o iii]@xmath04363/[o iii]@xmath05007 with respect to those for bls1s , which means that the temperature of gas in nlrs of the model nls1 is higher than that of the model bls1 . this is because the number of high energy photon is larger in nls1s than in bls1s ( see figure 6 ) . however , it is evident that this difference of the model loci is much smaller than the dispersion of the observed data points . therefore , we conclude that the difference of seds between the nls1s and the bls1s is not important when one investigates the [ o iii]@xmath04363/[o iii]@xmath05007 ratio . as shown in figure 13 , almost the same results are obtained when the subsolar abundances are assumed on the model calculations . in order to investigate the difference of gas temperature between nls1s and bls1s in more detail , we show the temperature structure in a cloud as a function of the hydrogen column density from the inner surface in figure 14 . here we adopt @xmath47 @xmath13 and @xmath48 for the four model calculations . it is shown that the gas temperature in the case of the subsolar abundances is higher than the other , which is due to a decrease of coolant elements . generally the temperature decreases as the column density increases . however , there is a small turn - up just before the ionization front where the temperature drops off . this is attributed to the fact that the most effective coolant , o@xmath49 , is exhausted at this region . it is important that the abundances more affect the temperature than the input sed . therefore , the difference of seds between the nls1s and bls1s does not affect significantly the gas temperature proved by the [ o iii ] emission lines . this result is also consistent with rodrguez - ardila et al . ( 2000a ) , in which the ratio of [ o iii]@xmath04363/[o iii]@xmath05007 is not so different between the model nls1 and the model bls1 ( less than factor 3 ) though they suggested that the calculated ratio of [ o iii]@xmath04363/[o iii]@xmath05007 is higher in the bls1 than in the nls1 . as mentioned in section 4.1 , we have assumed that nls1s and bls1s have similar @xmath21 . however , there are some previous studies ( walter & fink 1993 ; laor et al . 1994 ; puchnarewicz et al . 1996 ) in which it is claimed that the soft x - ray spectral index correlates with @xmath21 . since the large x - ray spectral index is one of the characteristic properties of nls1s , their claim means that nls1s have the softer @xmath21 than bls1s , systematically . therefore , we investigate the dependence of the calculations on @xmath21 . when various values of @xmath21 are adopted , @xmath50 must be correspondingly adjusted to reproduce the observed soft x - ray photon index , @xmath32 = 3.13 . we adopt @xmath50 = 980,000 k , 840,000 k , 730,000 k , 650,000 k , and 590,000 k for the cases of @xmath21 = 1.40 , 1.45 , 1.50 , 1.55 , and 1.60 , respectively . in figure 15 , we show a diagram of calculated line ratios versus @xmath21 , adopting @xmath47 @xmath13 , @xmath48 , solar abundances , and the sed template of nls1s . it is clearly shown that the [ o ii]/[o iii]@xmath05007 ratio is almost independent of @xmath21 . on the other hand , the [ o i]/[o iii]@xmath05007 and the [ o iii]@xmath04363/[o iii]@xmath05007 ratios become smaller and to be close to the value of bls1s as @xmath21 becomes softer . thus , we conclude that the difference of intrinsic seds between nls1s and bls1s scarcely affects the nlr emission even if @xmath21 of nls1s is systematically softer than that of bls1s . in figure 16 , we show the temperature structure in a nlr cloud for various values of @xmath21 , adopting @xmath47 @xmath13 , @xmath48 , solar abundances , and the sed template of nls1s . the gas temperature also becomes to be close to that of bls1s as @xmath21 becomes softer . seyfert galaxies often present highly ionized emission lines such as [ fe vii]@xmath06087 , [ fe x]@xmath06374 , [ fe xi]@xmath07892 , and [ fe xiv]@xmath05303 ( see nagao et al . 2000c and references therein ) . these emission lines are useful to investigate the viewing angle toward dusty tori of seyfert nuclei ( murayama & taniguchi 1998a ; nagao et al . therefore it is important to investigate how the feature of intrinsic seds affects such highly ionized emission - line intensities . we show the results of the model calculations for the case of the solar abundances and compare them with the observations in a diagram of [ fe vii]@xmath06087/[o iii]@xmath05007 versus [ o i]/[o iii]@xmath05007 ( figure 17 ) . the data of observations are taken from nagao et al . ( 2000c ) . being different from the results described in section 4.3.1 and 4.3.2 , there are evident differences in the behavior of the calculated line ratios between two models as follows . ( 1 ) the calculated [ o i]/[o iii]@xmath05007 ratio for nls1s is smaller than that for bls1s when @xmath51 although the opposite trend is seen when @xmath52 . this is because the volume of the fully ionized region becomes larger with increasing ionization parameter , and thus the [ o iii]@xmath05007 emission becomes more prominent relative to the [ o i ] emission . ( 2 ) the calculated [ fe vii]@xmath06087/[o iii]@xmath05007 ratio for nls1s is several times larger than that for bls1s when @xmath53 . this is because the number of the high - energy ionizing photons is 99.1 ev and 3.6 @xmath54 @xmath13 , respectively . ] producing fe@xmath55 in the model for nls1s is much larger than that in the model for bls1s when we adopt the same ionization parameter and the gas density for both cases . however , clearly shown in figure 17 , the calculated [ fe vii]@xmath06087/[o iii]@xmath05007 is much smaller than the observed one in both models . this means that another component which radiates highly ionized emission lines is needed to explain the observations , which is consistent with previous studies ( stasiska 1984 ; ferland & osterbrock 1986 ; binette et al . 1996 ; murayama & taniguchi 1998a , 1998b ) . therefore this result _ does not _ suggest that the observed [ fe vii]@xmath06087/[o iii]@xmath05007 of nls1s should be larger than that of bls1s . in figure 18 , we show the same diagram adopting the subsolar abundances . similar to the case of the solar abundances , the calculated [ fe vii]@xmath06087/[o iii]@xmath05007 for bls1s is smaller than that for nls1s . in the case of the subsolar abundances , the large fraction of iron is depleted ( see section 4.2 ) . therefore , the calculated [ fe vii]@xmath06087/[o iii]@xmath05007 is much smaller than that calculated adopting the solar abundances . this paper has presented the comparisons of emission - line ratios which represent the ionization degree and the gas temperature of nlr clouds between the nls1s and the bls1s . the emission - line ratio of [ o i]/[o iii]@xmath05007 , which probes the ionization degree of nlrs , and that of [ o iii]@xmath04363/[o iii]@xmath05007 , which probes the gas temperature of nlrs , are indistinguishable between the two samples . this means that there is little difference in the physical properties of nlrs between nls1s and bls1s . using photoionization models , we have confirmed that these results are consistent with the presence of differences in seds between nls1s and bls1s . in both cases , using the template seds of nls1s and bls1s , we have shown that the observed emission line ratios are well reproduced when we adopt @xmath44 and @xmath45 for either solar or subsolar abundances . this study tells us that we need not consider the effects of difference of intrinsic seds between nls1s and bls1s when we discuss ionized gas properties using diagnostic diagrams as used by , e.g. , ferland & netzer ( 1983 ) and ho et al . ( 1993 ) , unless the high ionization nuclear emission - line region ( binette 1985 ; murayama , taniguchi , & iwasawa 1998 ; murayama & taniguchi 1998a , 1998b ; nagao et al . 2000b , 2000c ) is concerned . we would like to thank the anonymous referee for useful comments and gary ferland for making his code @xmath56 available to the public . this research has made use of the ned ( nasa extra galactic database ) which is operated by the jet propulsion laboratory , california institute of technology , under construct with the national aeronautics and space administration . tm is supported by a research fellowship from the japan society for the promotion of science for young scientists . this work was financially supported in part by grant - in - aids for the scientific research ( nos . 10044052 and 10304013 ) of the japanese ministry of education , culture , sports , and science . lllcl ngc 4051 & & 0.0024 & 2.348@xmath1010@xmath57 & m95 , v88 ngc 4748 & mcg -2 - 33 - 34 , cts r12.02 & 0.0146 & 1.419@xmath1010@xmath58 & w92 , rpd00 mrk 42 & & 0.0240 & 1.053@xmath1010@xmath58 & c91 , k78 , op85 mrk 291 & & 0.0352 & 2.428@xmath1010@xmath58 & o77 mrk 335 & pg 0003 + 199 & 0.0258 & 1.316@xmath1010@xmath58 & c94 , o77 , p78 mrk 359 & ugc 1032 & 0.0174 & 1.956@xmath1010@xmath58 & c94 , dk78 , op85 , v88 mrk 478 & pg 1440 + 356 & 0.0791 & 2.161@xmath1010@xmath59 & o77 , p78 mrk 493 & ugc 10120 & 0.0319 & 4.099@xmath1010@xmath58 & c91 , op85 mrk 504 & pg 1659 + 294 & 0.0359 & & o77 mrk 507 & & 0.0559 & 1.010@xmath1010@xmath59 & k78 mrk 739 & ngc 3758 & 0.0299 & 6.499@xmath1010@xmath58 & c94 , so81 mrk 766 & ngc 4253 & 0.0129 & 3.830@xmath1010@xmath58 & c94 , op85 , v88 mrk 783 & & 0.0672 & 8.392@xmath1010@xmath58 & op85 mrk 896 & & 0.0264 & 2.068@xmath1010@xmath58 & c94 , mw88 mrk 957 & 5c 3.100 & 0.0711 & 6.390@xmath1010@xmath59 & k78 mrk 1126 & ngc 7450 & 0.0106 & & op85 mrk 1239 & & 0.0199 & 3.038@xmath1010@xmath58 & c91 , c94 , op85 , rpd00 1e 1031 + 5822 & & 0.248 & & s89 1e 1205 + 4657 & & 0.102 & & s89 1e 12287 + 123 & & 0.116 & & s89 2e 1226 + 1336 & & 0.150 & & s89 2e 1557 + 2712 & & 0.0646 & & s89 i zw 1 & ugc 545 & 0.0611 & 5.004@xmath1010@xmath59 & c94 , o77 , p78 akn 564 & ugc 12163 & 0.0247 & 2.901@xmath1010@xmath58 & c94 , v88 cts h34.06 & iras f06083 - 5606 & 0.0318 & 1.396@xmath1010@xmath58 & rpd00 h 1934 - 063 & iras 19348 - 0619 & 0.0106 & 1.787@xmath1010@xmath58 & rpd00he 1029 - 1831 & cts j04.08 , iras 10295 - 1831 & 0.0403 & 2.413@xmath1010@xmath59 & rpd00 iras 15091 - 2107 & & 0.0446 & 1.777@xmath1010@xmath59 & w92 cts j03.19 & cts 90 & 0.0532 & & rpd00 cts j13.12 & cts 103 & 0.0120 & & rpd00 kaz 163 & vii zw 742 & 0.0630 & & s89 kaz 320 & & 0.0345 & & z92 ms 01119 - 0132 & & 0.12 & & s89 ms 01442 - 0055 & & 0.08 & & s89 ms 12235 + 2522 & & 0.067 & & s89 q 0919 + 515 & & 0.161 & & s89 ngc 4235 & & 0.0080 & 1.158@xmath1010@xmath57 & mw88 , m95 ngc 4593 & & 0.0090 & 1.401@xmath1010@xmath58 & mw88 ngc 5940 & & 0.0337 & 4.915@xmath1010@xmath58 & mw88 mrk 10 & ugc 4013 & 0.0293 & 4.025@xmath1010@xmath58 & o77 mrk 40 & arp 151 & 0.0211 & & o77 mrk 50 & & 0.0234 & & mw88 mrk 69 & & 0.0760 & & o77 mrk 79 & ugc 3973 & 0.0222 & 4.250@xmath1010@xmath58 & c83 , o77 , v88 mrk 106 & & 0.1235 & & o77 mrk 124 & & 0.0563 & 1.286@xmath1010@xmath59 & o77 , p78 mrk 141 & & 0.0417 & 7.533@xmath1010@xmath58 & o77 mrk 142 & pg 1022 + 519 & 0.0449 & & o77 , p78 mrk 236 & & 0.0520 & 2.772@xmath1010@xmath58 & o77 mrk 279 & ugc 8823 & 0.0294 & 6.275@xmath1010@xmath58 & c83 , o77 mrk 304 & ii zw 175 , pg 2214 + 139 & 0.0658 & & o77 mrk 358 & & 0.0452 & 3.816@xmath1010@xmath58 & o77 mrk 374 & & 0.0435 & 2.952@xmath1010@xmath58 & o77 , p78 mrk 382 & & 0.0338 & 1.431@xmath1010@xmath58 & o77 mrk 486 & pg 1535 + 547 & 0.0389 & & o77 , p78 mrk 541 & & 0.0394 & 3.182@xmath1010@xmath58 & o77 , p78 mrk 590 & ngc 863 , um 412 & 0.0264 & 1.965@xmath1010@xmath58 & o77 mrk 618 & & 0.0356 & 1.990@xmath1010@xmath59 & o77 , p78 mrk 704 & & 0.0292 & 1.797@xmath1010@xmath58 & c83 , c94 , uw83 mrk 705 & ugc 5025 & 0.0292 & 2.900@xmath1010@xmath58 & c94 mrk 876 & pg 1613 + 658 & 0.1290 & 6.328@xmath1010@xmath59 & s89 mrk 975 & ugc 774 & 0.0496 & 1.163@xmath1010@xmath59 & c83 , c94 , v88 mrk 1040 & ngc 931 & 0.0167 & 4.047@xmath1010@xmath58 & c94 , v88 mrk 1044 & pg 0157 + 001 & 0.0165 & 6.656@xmath1010@xmath57 & t91 mrk 1347 & & 0.0497 & 1.337@xmath1010@xmath59 & mw88 1e 0514 - 0030 & & 0.292 & & s89 1h 1836 - 786 & iras f18389 - 7834 & 0.0743 & 1.246@xmath1010@xmath59 & mw88 1h 1927 - 516 & cts g03.04 & 0.0403 & & rpd00 1h 2107 - 097 & & 0.0268 & & w92 , rpd00 2e 0150 - 1015 & & 0.361 & & s89 2e 0237 + 3953 & & 0.528 & & s89 2e 1401 + 0952 & & 0.441 & & s89 2e 1530 + 1511 & & 0.090 & & s89 2e 1556 + 2725 & pgc 56527 & 0.0904 & & s89 2e 1847 + 3329 & & 0.509 & & s89 3c 263 & & 0.646 & & p78 3c 382 & cgcg 173 - 014 & 0.0579 & & o77 iii zw 002 & & 0.0898 & & o77 vii zw 118 & & 0.0797 & & c94 , ks79 cts a08.12 & cts 109 & 0.0293 & & rpd00 ab 125 & & 0.281 & & s89 akn 120 & ugc 3271 & 0.0323 & 3.891@xmath1010@xmath58 & p78 , v88 arp 102b & & 0.0242 & & ssk83 cts b25.02 & tololo 0343 - 397 & 0.0432 & 2.622@xmath1010@xmath58 & t91 b2 1425 + 26 & ton 202 , pg 1425 + 267 & 0.366 & & p78 b2 1512 + 37 & 4c 37.43 , pg 1512 + 370 & 0.3707 & & p78 cts c16.16 & & 0.0795 & & rpd00 eso 141-g55 & & 0.0360 & 4.335@xmath1010@xmath58 & mw88 , w92 eso 198-g24 & & 0.0455 & & w92 eso 323-g77 & & 0.0150 & 7.275@xmath1010@xmath58 & w92 eso 438-g09 & & 0.0245 & 1.088@xmath1010@xmath59 & kf83 eso 578-g09 & cts j14.05 & 0.0349 & & rpd00 cts f10.01 & cts 114 & 0.0784 & & rpd00 fairall 9 & eso 113-ig45 & 0.0470 & & w92 fairall 265 & & 0.0295 & 3.388@xmath1010@xmath58 & w92 fairall 1116 & tololo 0349 - 406 & 0.0582 & 3.175@xmath1010@xmath58 & w92 fairall 1146 & & 0.0316 & & rpd00 cts h34.03 & iras f05561 - 5357 & 0.0967 & 2.769@xmath1010@xmath59 & rpd00 h 0557 - 385 & cts b31.01 , iras f05563 - 3820 & 0.0344 & & w92 , rpd00 ic 4218 & & 0.0194 & & mw88 ic 4329a & & 0.0161 & 2.987@xmath1010@xmath58 & mw88 , w92 , wp79 cts j10.09 & iras f12312 - 2047 & 0.0230 & 1.164@xmath1010@xmath58 & rpd00 cts m02.30 & iras f10306 - 2651 & 0.0688 & 7.266@xmath1010@xmath58 & rpd00 mc 1104 + 167 & 4c 16.30 & 0.632 & & p78 ms 02255 + 3121 & & 0.058 & & s89 ms 07451 + 5545 & & 0.174 & & s89 ms 08451 + 3751 & & 0.307 & & s89 ms 08495 + 0805 & & 0.062 & & s89 ms 10590 + 7302 & & 0.089 & & s89 ms 11397 + 1040 & & 0.150 & & s89 ms 13396 + 0519 & & 0.266 & & s89 ms 13575 - 0227 & & 0.416 & & s89 ms 15251 + 1551 & & 0.230 & & s89 ms 22152 - 0347 & & 0.242 & & s89 pg 1352 + 183 & & 0.152 & & s89 pks 1417 - 19 & cts j15.22 , cts 105 & 0.120 & & rpd00 tololo 20 & & 0.030 & & mw88 ton 1542 & pg 1229 + 204 & 0.0640 & & p78 zw 0033 + 45 & cgcg 535 - 012 & 0.0476 & & c94 lccccc ngc 4051 & 0.1300 & & & 1.0190 & 0.3160 ngc 4748 & 0.0400 & 0.1883 & 0.0492 & 0.3000 & 0.2100 mrk 42 & 0.1435 & 0.3696 & 0.1339 & 1.8248 & 0.5680 mrk 291 & 0.1031 & 0.4742 & 0.1031 & 1.1031 & 0.4845 mrk 335 & 0.0667 & 0.4316 & 0.0699 & @xmath60 1.4783 & 0.0571 mrk 359 & 0.0577 & 0.1790 & 0.0735 & 0.2120 & 0.1297 mrk 478 & 0.0740 & 0.4280 & & & 0.2040 mrk 493 & & 0.5600 & 0.4568 & 2.0400 & 0.8240 mrk 504 & & & 0.0308 & & 0.1192 mrk 507 & 0.2099 & 0.5679 & 0.0593 & 4.3951 & 1.1481 mrk 739 & 0.3700 & 0.8894 & @xmath60 0.0740 & 4.2800 & 2.1500 mrk 766 & 0.0241 & 0.0763 & 0.0294 & 0.2166 & 0.0695 mrk 783 & 0.0584 & 0.2957 & 0.0895 & 0.1479 & 0.1479 mrk 896 & & 0.5276 & & & 0.1626 mrk 957 & 0.3016 & 0.8730 & 0.0952 & 3.4444 & 1.1111 mrk 1126 & 0.0791 & 0.1279 & 0.1163 & 5.8140 & 0.3233 mrk 1239 & 0.0200 & 0.0852 & 0.0892 & 0.3011 & 0.1128 1e 1031 + 5822 & 1.1280 & 0.7082 & 0.0683 & & 1e 1205 + 4657 & 0.0555 & & 0.3646 & & 0.7376 1e 12287 + 123 & & & 0.5395 & & 0.6974 2e 1226 + 1336 & & 0.3493 & 0.0312 & & 0.4521 2e 1557 + 2712 & & & & & 0.6269 i zw 1 & 0.0636 & 0.4756 & 0.0761 & & 0.0295 akn 564 & & 0.4068 & & & cts h34.06 & & 0.3421 & 0.1023 & & h 1934 - 063 & & 0.1257 & 0.0437 & & he 1029 - 1831 & & 0.1961 & 0.0980 & & iras 15091 - 2107 & & 0.1400 & 0.0500 & 0.3200 & 0.3600 cts j03.19 & & 0.1626 & & & cts j13.12 & & & 0.3101 & & kaz 163 & & & & & 0.1418 kaz 320 & 0.0629 & 0.3452 & 0.0595 & 0.2151 & 0.2738 ms 01119 - 0132 & & & 0.1188 & & ms 01442 - 0055 & & & 0.0207 & & 0.2434 ms 12235 + 2522 & & & & & 0.2120 q 0919 + 515 & & 0.2191 & & & ngc 4235 & 0.1967 & 0.6389 & 0.2837 & 0.6610 & 0.6252 ngc 4593 & 0.0470 & 0.1696 & & & 0.1357 ngc 5940 & 0.0772 & & & & 0.1680 mrk 10 & 0.0172 & 0.0871 & 0.0247 & 0.1183 & 0.0763 mrk 40 & 0.1111 & 0.5000 & 0.1667 & 0.2639 & 0.1444 mrk 50 & 0.1218 & 0.3261 & & & 0.1851 mrk 69 & & 0.4186 & 0.0837 & 1.0698 & 0.3512 mrk 79 & 0.0424 & 0.1763 & 0.0457 & 0.2591 & 0.1741 mrk 106 & 0.0414 & 0.2000 & 0.0310 & 0.2241 & mrk 124 & 0.0628 & 0.3577 & 0.0730 & & 0.2394 mrk 141 & & 0.1632 & 0.0816 & 0.9474 & 0.3763 mrk 142 & @xmath60 0.0857 & 0.3760 & 0.0680 & 1.3200 & 0.5669 mrk 236 & 0.0717 & 0.1848 & 0.0804 & 0.1587 & 0.2565 mrk 279 & 0.1312 & 0.4635 & 0.0670 & 0.6483 & 0.3204 mrk 304 & @xmath60 0.0455 & 0.1212 & 0.1061 & 0.3788 & 0.1818 mrk 358 & @xmath60 0.0977 & 0.2326 & @xmath60 0.0605 & 0.3256 & 0.3953 mrk 374 & 0.0686 & 0.0914 & 0.0738 & 0.0628 & 0.1188 mrk 382 & & 0.0805 & 0.0322 & 0.3333 & 0.7805 mrk 486 & 0.0917 & @xmath60 0.1500 & & & 0.0773 mrk 541 & @xmath60 0.2050 & 0.1739 & 0.2217 & 1.3755 & 0.5600 mrk 590 & 0.2000 & 0.2182 & 0.2000 & 0.5455 & 0.2345 mrk 618 & 0.0549 & 0.1356 & 0.0410 & 0.8667 & 0.2320 mrk 704 & 0.0287 & 0.2016 & 0.0726 & 0.1710 & 0.1507 mrk 705 & & 0.1532 & & & mrk 876 & 0.1659 & & 0.0272 & & 0.7093 mrk 975 & 0.0248 & 0.1998 & 0.0303 & 0.4421 & 0.0634 mrk 1040 & & 0.0850 & & & mrk 1044 & & 0.1238 & & & mrk 1347 & & 0.6032 & & & 0.2063 1e 0514 - 0030 & & & 0.4949 & & 1h 1836 - 786 & 0.0349 & 0.1623 & & & 0.1623 1h 1927 - 516 & & & 0.0792 & & 1h 2107 - 097 & 0.1300 & 0.1300 & 0.1696 & & 2e 0150 - 1015 & & 0.1291 & 0.0341 & & 2e 0237 + 3953 & & 0.3364 & 0.0757 & & 2e 1401 + 0952 & & & 0.0565 & & 2e 1530 + 1511 & & & & & 0.2091 2e 1556 + 2725 & 0.0430 & & 0.0884 & & 0.2033 2e 1847 + 3329 & & 0.1800 & 0.0649 & & 3c 263 & & & 0.0515 & & 3c 382 & 0.0700 & 0.1300 & 0.1600 & 0.0400 & 0.1600 iii zw 002 & 0.0576 & 0.1394 & 0.0273 & 0.1333 & 0.1576 vii zw 118 & & 0.4444 & & & cts a08.12 & & 0.1280 & 0.0759 & & ab 125 & & 0.2922 & & & akn 120 & 0.0917 & 0.2658 & & & 0.3108 arp 102b & 0.8030 & 1.1970 & 0.0571 & 1.9343 & 1.2475 cts b25.02 & & 0.3207 & & & b2 1425 + 26 & & & 0.0417 & & b2 1512 + 37 & & & 0.0393 & & cts c16.16 & & 0.1793 & 0.0588 & & eso 141-g55 & & 0.0924 & & & 0.2800 eso 198-g24 & & 0.2700 & & & 0.1300 eso 323-g77 & & 0.1300 & & & 0.2400 eso 438-g09 & 0.0811 & 0.4324 & 0.0541 & 1.0000 & 0.4595 eso 578-g09 & & 0.1719 & 0.0898 & & cts f10.01 & & 0.1048 & 0.0391 & & fairall 9 & & 0.0800 & 0.1200 & 0.2000 & fairall 265 & & 0.7600 & & 1.3500 & fairall 1116 & & 0.2000 & 0.0900 & & 0.5400 fairall 1146 & & 0.1570 & 0.0436 & & cts h34.03 & & & 0.1287 & & h 0557 - 385 & & 0.0816 & & & 0.1200 ic 4218 & & 0.2318 & & & 0.1706 ic 4329a & 0.0964 & 0.0429 & & 0.3091 & 0.3156 cts j10.09 & & 0.1034 & 0.1565 & & cts m02.30 & & 0.1597 & 0.1875 & & mc 1104 + 167 & & & 0.0941 & & ms 02255 + 3121 & & & 0.1270 & & ms 07451 + 5545 & & 0.2066 & 0.0337 & & ms 08451 + 3751 & & & 0.1198 & & ms 08495 + 0805 & 0.0313 & & 0.0395 & & 0.1261 ms 10590 + 7302 & & & & & 0.1080 ms 11397 + 1040 & & & 0.0215 & & 0.1304 ms 13396 + 0519 & & 0.5666 & & & ms 13575 - 0227 & & 0.1029 & & & ms 15251 + 1551 & & 0.0675 & 0.0405 & & ms 22152 - 0347 & & 0.1071 & 0.0329 & & pg 1352 + 183 & & & & & 0.2090 pks 1417 - 19 & & & 0.1039 & & tololo 20 & 0.0425 & 0.0800 & 0.0924 & & 0.0337 ton 1542 & & 0.1944 & 0.1167 & & 0.1333 zw 0033 + 45 & 0.2727 & & & & +

we model the assembly of planets from planetary embryos under the conditions suggested by various scenarios for the formation of the planetary system around the millisecond pulsar b1257 + 12 . we find that the most likely models fall at the low angular momentum end of the proposed range . models that invoke supernova fallback produce such disks , although we find that a solar composition disk produces a more likely evolution than one composed primarily of heavy elements . furthermore , we find that dust sedimentation must occur rapidly as the disk cools , in order that the solid material be confined to a sufficiently narrow range of radii . a quantitative comparison between the observations and the best - fit models shows that the simulations can reproduce the observed eccentricities and masses , but have difficulty reproducing the compactness of the pulsar planet system . finally , we examine the results of similar studies of solar system terrestrial planet accumulation and discuss what can be learned from the comparison .

0908.0736

although the discovery of the planet 51 pegasi b ( mayor & queloz 1995 ) is often hailed as the start of the extrasolar planet revolution , the field of extrasolar planet studies began in 1991 , with the discovery of the two planet - sized bodies orbiting the pulsar psr 1257 + 12 ( wolszczan & frail 1991 ) . subsequent study has confirmed this identification by identifying the effect of mutual gravitational perturbations between the two planets in the timing residuals , thereby also constraining ( wolszczan 1994 ; wolszczan et al . 2000a ; konacki & wolszczan 2003 ) the inclinations and planetary masses , the current values of which are given in table [ params ] . yet , despite their early discovery , the origin of the pulsar planets have received little attention after the initial burst of post - discovery papers ( summarised in podsiadlowski 1992 ) . part of this neglect is no doubt due to the unusual nature of the host star and the relative uniqueness of the planets natal circumstances . however , as has been noted before ( phinney & hansen 1993 ) , the formation scenarios all produce a qualitatively similar outcome namely a gaseous disk on a compact scale around a 1.4@xmath0 pulsar . thus , the exact details of the formation are relevant primarily in the manner in which they determine the size of the original mass and angular momentum budget , and also the disk composition . the subsequent evolution of the gaseous and planetesimal disks are expected to follow a path very similar to the one they would follow if orbiting a normal main sequence star of similar mass . as such , studying the formation of the pulsar planets can offer us insights into the formation of terrestrial planets as a whole . this will be of particular interest in the next few years as the searches for planetary companions to main sequence stars continues to push down into the ` super - earth ' regime ( e.g. rivera et al . 2005 ; udry et al . 2007 ; mayor et al . 2008 ) . in that spirit , we wish to examine the quantitative evolution of pulsar protoplanetary disks that arise from different formation scenarios . in currie & hansen ( 2007 ) , hereafter paper i , we examined the expansion and evolution of gaseous disks for a range of mass and angular momentum , and identified the manner in which they laid down solid material that might eventually coalesce to form planets . in this paper we wish to extend that analysis to the question of how planetesimals , formed from such a distribution , would assemble into a final planetary configuration , and how the resulting planetary systems compare to that observed . in [ setup ] we review the outcomes of paper i and how we construct the initial conditions for the simulations described in this paper . in [ sims ] we describe the qualitative properties of the final planetary systems in each of the simulated scenarios . in [ comp ] we compare these results to the observed system and discuss the issues of broader relevance in [ discuss ] . essentially all of the proposed pulsar planet provenances invoke the capture of material from an external source , resulting in the formation of a gaseous disk of material in orbit around the pulsar . the nature of the interaction that generates the material will determine the mass and angular momentum of the initial disk , but all scenarios result in a disk that is initially quite compact , with characteristic scales @xmath1 . thus the initial state is one of super - eddington accretion onto the pulsar from a hot disk . as mass is accreted onto the star some material moves outwards to conserve angular momentum in the traditional accretion disk manner . although the physics that describes the evolution of the outer edge of this expanding disk is similar to that of normal protoplanetary disks , it offers an interesting variation on the traditional situation , in which the scale of the disk is set by the infall of material from larger radii . as we explain below , this leads to important differences that may help to explain the observed configuration . in paper i we describe the evolution of such disks for a range of total angular momentum . we motivate this by making reference to two specific scenarios , although they would apply equally well to different situations that produce disks with the same global properties . the first scenario is the formation of a disk that results from supernova fallback material . models of this kind lie at the lower end of the range of the disk angular momentum distribution . the second scenario results in the formation of a disk by the tidal disruption of a companion star with which a newly kicked neutron star collides . this yields angular momentum @xmath2@xmath3 ergs s , which lies at the upper end of the theoretical distribution . in paper i we incorporate a model to describe the thermal and viscous evolution of the disk , assuming a traditional metallicity for the gaseous material ( which is consistent with most , although not all , formation scenarios ) . we also consider two variations for the viscous evolution . in the first case , we simply assume a traditional parameterised @xmath4-viscosity ( shakura & sunyaev 1973 ) . in the second case , motivated by recent considerations regarding the ultimate physical origin of the @xmath4 viscosity , we consider a two - viscosity model in the spirit of the ` layered accretion ' model of gammie ( 1996 ) . in this model , we assume that the viscosity goes to zero when the local temperature of the disk material at a given radius drops to a level where the material is no longer expected to be ionized . this has the effect of regulating the outwards spread of the disk ( hansen 2000 ; paper i ) and imposing a final mass profile that depends on the total angular momentum of the disk . into each of these evolutionary models we also incorporate a prescription for the deposition of solid material , which is assumed to decouple from the gas and sediment to the midplane when the temperature drops below a certain threshold . for the purposes of this paper , we will assume that the local deposition of solid material determines the initial distribution of planetesimal mass . there are a variety of proposed physical effects , such as radial drift due to gas drag or ` type i migration ' , that may affect this distribution , but the degree to which such processes operate in reality is poorly understood and a matter of discussion . we have opted to calculate the simplest model , and will discuss possible refinements in [ otherstuff ] . using this model , we can predict the form of the solid material profile at the end of the gaseous evolution and adopt this as the initial conditions for the planetesimal assembly simulations . the models in paper i provide a surface density profile of solids that provide the initial condition for the next stage of the evolution , namely the assembly of the rocky material into the observed planets . in the standard model of planetary assembly , the initial stage from planetesimals to planetary embryos is essentially a local phenomenon , and proceeds until most of the material is concentrated in a series of ` oligarchs ' ( e.g. kokubo & ida 1998 ) , which contain the majority of the mass within an annulus whose width is roughly several times the hill sphere of the oligarch in question . the final stage of planet formation is then the subsequent collisional assembly of this population of planetary embryos into the final planets . we wish to simulate this final stage evolution of the planetesimal disks in each case discussed above , using the planetary dynamics code mercury6 ( chambers 1999 ) . this code incorporates a hybrid scheme which treats the longer - range interactions using a symplectic integrator but treats close encounters with a direct bulirsch - stoer integration . this makes it particularly useful for our purposes , because the transition from collisional accumulation to a dynamically quiescent configuration is presumably what will determine the final planetary configuration . we will assume an initial distribution composed of equal mass planetary embryos , whose spatial distribution is chosen to mimic the solid material surface density profile left at the end of gas disk evolution as described in paper i , for each individual scenario . we choose to start with these masses rather than the masses estimated for the final stage of oligarchic growth because the surface densities we use below are often significantly larger than those normally used in the solar system context , and sometimes lead to estimated oligarchic masses @xmath5 . in such cases , it is probably not a good approximation to assume the assembly proceeds to full oligarchic completion without interacting with neighbouring annuli . in effect , the distinction between the local runaway accretion into oligarchs and the final completion of assembly is blurred . we therefore choose to start with smaller masses and more bodies in order to be sure not to miss some of the essential dynamics . this amounts to essentially starting the simulation close to the end of the oligarchic growth phase ( we begin at the point where is still a significant population of small bodies ) but before the individual annuli are fully cleared . let us now consider each of the cases in turn . we consider bodies with densities @xmath6 , initially on perfectly circular , coplanar orbits , and assume an integration accuracy of @xmath7 per timestep ( which is taken to be 4 days , so that all relevant orbits are well resolved ) . we assume all physical collisions result in mergers , although for some of the more compact disks , encounter velocities are high enough that fragmentation is a real possibility . disks with total angular momentum @xmath8 ergs s lie at the upper end of the distribution produced in the various proposed scenarios . one scenario that gives rise to such a disk is the tidal disruption of a close stellar companion to the progenitor star , which is disrupted by the passage of the post - supernova kicked neutron star ( e.g. phinney & hansen 1993 ) . the large angular momentum means that the disk expands to radii @xmath9au before becoming cool enough to form dust and allow the deposition of solid material . if the disk remains fully viscous throughout , the resulting solid material is deposited in a relatively flat surface density profile and extending out as far as 7 au . following the results of paper i , we assume the initial surface density of solids is constant , with a value @xmath10 , out to 7 au . the mass is spread initially between 100 identical bodies , each of mass 0.58 @xmath11 . this results in a planetesimal disk of total mass @xmath12 . figure [ snap1 ] shows the evolution of the resulting planetary embryo swarm . as some of the bodies grow to masses @xmath5 , the eccentricities and inclinations are amplified by the mutual scattering . on timescales @xmath13 years , the outer edge of the swarm is pushed out to @xmath14 au , and there are five bodies with mass @xmath15 . after @xmath16 years , there is a dominant body , with mass @xmath17 , semi - major axis 0.91 au and an eccentricity of 0.263 . there is also still a significant amount of mass at larger radii . figure [ massive1 ] shows the radial evolution of the body that grew to be the most massive one . it initially began at @xmath18 au , and underwent a brownian random walk during the initial accumulation , migrating out as far as 8 au before undergoing an inward migration as its mass began to grow . once the mass is large enough to scatter smaller bodies outwards instead of accreting then , it begins to preferentially lose binding energy , much in the same manner as happens in our solar system ( e.g. fernandez & ip 1984 ) . the fact that the available reservoir of material is comparable in total mass to the scattering body means that this system suffers from the same kind of instability as described by murray et al . ( 1998 ) , albeit in a slightly different context . the rapidity of the migration also explains why some of the material is left in partially scattered orbits ( hansen 2000 ) , even though the dominant body is now capable of ejecting such material from the system ( see [ dynamics ] ) . in fact , only 1.3 @xmath11 of material is ejected during the evolution . the size of the dominant remaining body also invites speculation as to whether it could form a jupiter - class gas giant planet , since the mass is larger than the @xmath19 threshold for nebular accretion in the context of the core accretion scenario for giant planet formation ( e.g. mizuno et al . 1978 ; pollack et al . 1996 ) . in figure [ massive1 ] the dotted lines show the points at which the body crosses the mass thresholds of @xmath20 and @xmath21 . these occur on timescales @xmath22myr & @xmath23myr respectively , suggesting that giant planet formation is unlikely , since the gas density drops rapidly on myr timescales ( paper i ) . thus , the final configuration is likely to be accurately reflected in figure [ snap1 ] . the final configuration from this model is clearly wildly discrepant with regards to the observed system psr b1257 + 12 , as the observed planets are smaller and much closer to the pulsar than anything produced in this model . this is the case in every model realisation of this scenario , as can be seen in table [ comptab ] . indeed , the dynamical signatures that result from this type of system would produce time - of - arrival fluctuations of order 10 ms or more , and are thus potentially detectable around even normal pulsars . the larger orbits also imply orbital periods that are frequently several years , so that it is an interesting question as to whether any objects of this ilk are lurking in the timing noise of presently known systems . the model of the previous section clearly produces too much mass on too large a scale to produce anything resembling the observed pulsar planets . however , a fully viscous disk is unlikely to be an accurate model for any protoplanetary disk on these scales . in paper i , we also considered an evolutionary sequence for a large disk within the context of the layered accretion model of gammie ( 1996 ) . in this case , the final solid surface density is still approximately constant with radius , but with a higher value ( @xmath24 ) , and the disk is more compact ( with an outer edge of 4 au ) . once again , we begin with 100 bodies , this time each with mass @xmath25 . the total mass in this case is thus @xmath26 . figure [ snap2 ] shows the corresponding system configurations at 1 , 10 and 100 myr . the behaviour is similar to that observed in figure [ snap1 ] , except that the evolution proceeds a little faster because of the more compact configuration . once again , the final state is a moderately massive body ( in this case 12 @xmath11 ) , with semi - major axis 1.2 au and eccentricity 0.206 . in this case , at 100 myr , this is only the second innermost body , with another , 2.3 @xmath11 , body at smaller radii , pushed inwards as the larger body migrated . in both this case and the former , the dynamical evolution of this system of planetary embryos results in a single body which rapidly accretes a substantial fraction of the total mass , but probably on a long enough timescale that the gas inventory was reduced before it had a chance to form a gas giant planet through core accretion . this is a consequence of the excretion disk nature of the protoplanetary disk in this context , which results in a smaller and more rapidly evolving gas budget than in a traditional protoplanetary disk . at the lower end of the proposed angular momentum range is a disk with total angular momentum @xmath27 ergs s. a simple model which produces such is disk is the supernova fallback disk model , in which some fraction of the material in the progenitor core remains bound even after the passage of the supernova shock wave , and falls back to form a disk around the newly born neutron star . disks that result from such a scenario have masses @xmath28 and total angular momentum @xmath27 ergs s ( chevalier et al . 1989 ; hashimoto et al . 1989 ; lin et al . 1991 ; menou et al . the amount of material that actually expands to the scales of interest is smaller , @xmath29 . this also depends on the composition of the material . if the fallback material comes from the outer layers of the star , it may contain significant hydrogen and helium , and may evolve in a manner similar to a solar composition disk . on the other hand , if the material originates from the interior , onion - skin layers of the star , the evolution may be quite different because of the heavy element - dominated composition . in this section we will discuss the solar composition disk , leaving the heavy element disk to [ kristen ] . once again , the rate of expansion is also affected by what one assumes for the viscous evolution of the disk . in the case of a fully viscous disk , with @xmath30 , the final distribution of solid material can be characterised by a surface density profile @xmath31 with a cutoff at 1.5 au . the resulting mass is lower than in the previous models , only 10.6@xmath11 . once again , we spread this between 100 initial bodies , each of mass 0.106 @xmath11 . the dynamical evolution is shown in figure [ snap3 ] . the comparison with observations is much more encouraging in this case , primarily because the mass budget is much closer to that found in the observed system . the agreement is not perfect however . different realisations of this scenario ( table [ comptab ] ) all produce several planets with @xmath32 , but spread over a range that extends beyond 1 au , which is markedly more extended than the observed system . in the case shown here , we have four planets , with masses @xmath33 , @xmath34 , @xmath34 and @xmath35 , with semi - major axes extending from 0.18 au to 1.2 au . there is also potential for smaller , dynamically decoupled bodies at @xmath36 au . however , as we noted before , a fully viscous evolution is unlikely and a more physically motivated gaseous evolution is provided by the layered disk model . in the case of layered accretion , the remaining solid surface density is even steeper , with @xmath37 with a cutoff at 1 au . once again , the simulations are begun with 100 bodies , with individual masses 0.082 @xmath11 . the resulting total mass is 8.2@xmath11 . the results of this scenario are shown in figure [ snap4 ] and table [ comptab ] . again , there are some qualitative agreements with the observations . planets of approximately the right mass are formed in the region 0.30.5 au . however , as in the previous case , our final configuration contains several more planets than are observed . in this case we are left with five planets , all of whose masses are @xmath38 and are thus readily observable . we have also verified that this system is dynamically stable by letting it run out to @xmath39 years . although we used the supernova fallback scenario to motivate our choice of total angular momentum above , we still described the evolution of the gaseous disk in terms appropriate to cosmic composition . in many scenarios , the remaining disk is hydrogen poor and dominated by heavier elements , such as carbon and oxygen . in such cases , the higher ionization potentials of the dominant disk material means that the disk is likely to become neutral at much higher temperatures and thus remain quite compact ( chevalier 1989 , hansen 2000 ) . menou , perna & hernquist ( 2001 ) discuss the evolution of such gaseous disks in some detail . they find that the expanding disk is likely to stall ( because heavy elements become neutral at much higher temperatures ) while still very compact , indeed at radii smaller than the tidal disruption radius ( @xmath40 cm ) for a planet - size rocky body . in this event , the starting point for a planetary embryo calculation should thus be a very compact disk , which could potentially scatter bodies out to larger radii , where they might form pulsar planets as a residual , scattered population . to model this situation , we start with an annulus of width 0.05 au , located at 0.1 au . if we imagine that the planets move outwards to conserve angular momentum while most of the mass goes inwards , in the traditional fashion of accretion disks , then we need to assume an initial disk mass that is larger , @xmath41 . however , the end result of the simulation does not contain any planets at radii significantly larger than the original annulus , but rather the original inventory is spread amongst three larger planets , spread between 0.049 au and 0.112 au . the problem is essentially that the planets involved are not big enough to drive an outwards migration in the manner of our outer solar system ( e.g. fernandez & ip 1984 ; hahn & malhotra 1999 ) , because the repulsive force is mediated through multiple scatterings of smaller bodies . when the scatterers are @xmath42 , this leads to collision and accretion more frequently , short - circuiting any tendency towards outwards migration ( see [ dynamics ] ) . we repeated this calculation with several realisations , and found no significant transport of mass outwards in any of them . for completeness , we also performed the same simulation using an initial disk inventory of @xmath21 , initially spread amongst 100 bodies , which resulted in two planets , of mass 4.9@xmath11 and 5.1@xmath11 , located at 0.043 au and 0.088 au . thus , it appears as though an initially compact planetesimal disk can not produce pulsar planets on the scales observed . we note that the accumulation of planets in this scenario is extremely rapid , with the final system essentially in place within @xmath43 years , although the accumulation may be lengthened if collisional erosion is considered , as the encounter velocities in such a compact disk are likely to be higher than in the other cases . a qualitative comparison between the observations and the data suggests that the layered disk evolution of the smaller disk represents the most likely match to the observed pulsar planets . the larger disks simply produce too much mass on too large a scale , resulting in planets that are far too massive . in the case of the smaller disk , the final masses are closer to those observed , although spread over too large a range in radius if the disk is assumed to evolve via the @xmath4 disk model . in the case of the heavy element disks , one can reproduce the number and masses of the planets , but they are invariably found too close to the central star . however , even in the case of our favoured scenario , the comparison is not perfect . multiple realisations of the small , layered disk all result in more planets than those observed essentially , nature incorporates the same amount of mass into fewer bodies than our model predicts . in order to make the comparison more quantitative , table [ comptab ] shows the values of a variety of statistics proposed by chambers ( 2001 ) for the quantification of terrestrial planet systems , as applied to simulations of the above scenarios . these include @xmath44 , the fraction of the total mass stored in the largest object , @xmath45 , a statistic that corresponds roughly to the mean spacing between planets in terms of their hill radii , @xmath46 , the normalised angular momentum deficit ( a measure of how circular the orbits are ) and @xmath47 , a mass - concentration statistic , which measures how localised in radius the planets are . we also include the mass - weighted semi - major axis , @xmath48 , of chambers & cassen ( 2002 ) . we restrict our attention to those particles in the simulation which would yield fluctuations in the timing residuals @xmath49 and with periods less than 20 years . this is in order to approximately match the observational limits . finally , we also show the value of the most massive planet ( @xmath50 ) and it s location ( @xmath51 ) . the numbers bear out the qualitative observations above . particularly striking is the disparity between the observed and simulated values of @xmath47 , the localisation statistic . the observed pulsar planets have a value of @xmath47 an order of magnitude larger than the results from all the simulated cases , indicating that most proposals produce systems that are simply too spread out in radius . note that the pulsar planets are even more severely localised ( by a factor of 4 ) than our own terrestrial planet system . the simulation planets are also , in general , more eccentric than the observed planets , as indicated by the @xmath46 statistic . figure [ ss2 ] shows this result in graphical terms . figure [ ma0 ] shows the comparison of models and observations in terms of @xmath50 and @xmath48 , which reflects whether the final planets are of the right size and in the right general location . we see here the basis of our earlier assertion that the smaller disk models provide the best qualitative match . in performing the above comparisons , we have assumed that the current census ( konacki & wolszczan 2003 ) of the pulsar planet system ( which stands now at three objects ) is complete . there have been reports in the past ( wolszczan et al . 2000b ) of additional timing residuals in the b1257 + 12 data that might be indicative of the presence of additional small planets at larger radii . however , subsequent observations suggest these are due to effects related to the propagation of radio waves through the interstellar plasma , and not gravitational in origin ( alex wolszczan , private communication ) . we have tested the sensitivity of the above measures to the presence of additional small bodies whose contribution to timing residuals would have amplitudes @xmath52 , and found negligible influence . one statistic that is affected by observational uncertainties is the @xmath46 statistic the angular momentum deficit . the relative inclination of the two large planets reported by konacki & wolszczan is @xmath53 . this has a significant effect on @xmath46 because the statistic is sensitive to any deviations from circular , coplanar orbits . thus , in figure [ ss2 ] we show the error in @xmath46 due to the @xmath54 error in relative inclinations . the results of the preceding sections can be understood within the context of the theory of planetary assembly as developed for our own solar system . goldreich , lithwick & sari ( 2004 ) recently reviewed the understanding of the final stages of planetary assembly , with specific focus on the origin of uranus & neptune . within the terminology of goldreich et al . , our simulations apply to the period referred to as ` completion ' , when all small bodies have been consumed and the dynamics is governed by the interaction and merging of bodies of planetary embryo size . there has been a long - standing discussion as to whether there is a need for a dynamically cooled population of small bodies ( not present in the simulations shown here ) during these late stages of planetary accretion . the argument in favour of this population is that , without it , the accumulation of material to form uranus and neptune is too slow the result of the fact that the rate of accretion is reduced as the velocity dispersion of the accreting bodies is increased by scattering off the growing planet . these considerations do not , in fact , impact the simulations shown here because the systems we simulate are more compact , and the planets can accumulate within a finite amount of time without any additional dynamical cooling . as an illustration of this , we can recast equation ( 56 ) of goldreich et al . ( 2004 ) , which describes the characteristic accretion time for a large planet , in the form @xmath55 where @xmath56 and @xmath57 are is the characteristic random velocity of the accreting bodies and @xmath57 is the escape velocity from the surface of the growing planet . our simulations , without any dynamical cooling , automatically produce @xmath58 . the estimated formation time of @xmath59 is in good agreement with simulations shown in figures [ snap3 ] and [ snap4 ] , for example . application of this formula to the simulations shown in figures [ snap1 ] and [ snap2 ] also explains why it takes somewhat longer to form the largest body in those systems . the accumulation time is longer because the bodies start forming at 23 au , and so @xmath60 can be an order of magnitude larger . one aspect of these simulations not described in the goldreich et al . treatment is the inwards migration of the dominant body . for this , we turn to the framework laid out by tremaine ( 1993 ) . as the principal body grows in size , it scatters the smaller bodies in the vicinity . the initial evolution can be characterised as a growth of the velocity dispersion , as described by goldreich et al . , but as the growth continues , a more accurate characterisation is as a diffusion process in energy space , with fixed periastron . it is this process that is thought to drive the evolution of comets towards the oort cloud ( e.g. fernandez & ip 1984 ; duncan , quinn & tremaine 1987 ) . using a diffusion coefficient based on numerical simulations , tremaine derives a criterion for a mass threshold above which a planet can drive orbital evolution of smaller bodies on a characteristic timescale @xmath61 @xmath62 where we have assumed a pulsar mass of 1.4@xmath0 and where @xmath63 is the semi - major axis of the scattering planet . thus , combining equations ( [ gls ] ) & ( [ t93 ] ) , we may derive a criterion for a planet to migrate during it s accumulation process @xmath64 which corresponds quite well to the observed evolution . once a planet has begun to scatter other bodies in the system , the question is whether the cumulative effect of encounters and scatterings is to accrete material or eject it from the system . once again , tremaine estimates the threshold mass for this process to be @xmath65 where @xmath66 indicates the average range of inclinations of the population of small bodies being scattered . this number is taken from the simulations directly . if the vertical velocity dispersion @xmath67 is generated by scattering off the same body ( mass @xmath68 ) as is driving the collisional evolution , then @xmath69 where @xmath70 is the radius of the planet . assuming a mean planet density of @xmath71 , this results in a modified collision / ejection transition mass @xmath72 figure [ ma3 ] shows these criteria , along with the final states of two simulations , those shown in figures [ snap2 ] and [ snap4 ] . bodies that lie above the dotted line labelled ` migrate ' will migrate inwards faster than they can accrete . bodies that lie below the dotted line labelled ` accrete ' will preferentially accrete the bodies they encounter , while those above the line will eventually eject bodies through repeated scatterings . we see that the small disk simulation ( solid points ) is firmly within the regime where all mass is eventually accreted into the largest bodies , while the large disk simulations ( open points ) are capable of losing material to ejection , and large bodies are able to migrate a significant amount . the largest body in this latest simulation is located near the intersection of the two criteria , which is not a coincidence . the body grows to the point where it migrates inwards by scattering companions , but then eventually stalls as a result of reaching the threshold between accretion and ejection . thus , there is a qualitative difference in the two cases small disks accumulate material together quasi - locally , while the final configuration of a larger disk is regulated by global considerations of energy and angular momentum . this is , to a large extent , also the difference between the inner and outer parts of our own solar system . for completeness , we note that the presence of a population of smaller and dynamically colder bodies ( neglected in these simulations ) in [ tdfv ] or [ tfla ] could potentially speed up the accretion so that a large core is formed before the gas disk dissipates ( in a similar spirit to the models of our own solar system ) . in that event , one could possibly form a true gas giant planet in those scenarios . however , in the absence of any observational evidence , we have not considered that model in detail . the models presented in [ sims ] produce an encouraging , but not entirely satisfactory result . a disk of total angular momentum @xmath27 ergs s , made of approximately solar composition , and whose gaseous evolution proceeds via the layered disk model , can produce planets of approximately the correct mass , in approximately the correct location , to be a plausible progenitor of the psr b1257 + 12 system . however , the models generically produce planets that are too numerous , too spread out and more eccentric than the observations . this suggests that the model , as described , does not yet tell the full story . in examining the results of paper i , we can identify a probable reason why the mass distribution is more extended than the observations indicate . the initial deposition of mass is limited to the region 0.40.6 au , as the outer edge of the expanding disk drops below the temperature threshold for dust particles to condense out . as the disk continues to evolve , more mass expands outwards causing the surface density at the outer edge to increase , which , in turn , results in an increase in the local temperature . if one assumes , as we did in paper i , that the sedimented material is reabsorbed into the gas when the local temperature increases above condensation temperature , then the deposition profile will be influenced by the continued evolution of the gaseous disk and the final profile will only be layed down at late times , when the disk has ceased to expand significantly . the lower angular momentum disk scenarios are more sensitive to this issue than larger , more extended disks , such as a normal protoplanetary disk or the disks that result from tidal disruption . however , if the dust that sediments out forms larger and more robust agglomerates on a timescale short compared to the disk evolution time , then it will not be reabsorbed into the reheated gas , and then the correct distribution of solids is more accurately represented by the initial deposition profile , such as seen in figure 4 of paper i. thus , we have also performed simulations in which the mass is distributed in a uniform surface density annulus between 0.4 and 0.6 au . for a disk mass of @xmath73 , this results in a surface density @xmath74 . we simulate this as before , starting with 300 bodies , each of mass @xmath75 . the evolution of a planetary embryo swarm drawn from such a starting disk is shown in figure [ snap5 ] . the evolution is particularly rapid in this case , being largely complete within @xmath76 years . the more rapid pace is simply because the material is concentrated in a limited range of radii . we also note that there is relatively little orbital evolution of the largest bodies . the eventual outcome is also potentially the most encouraging of all the calculations , as the number , masses and locations of the bodies are now qualitatively similar to those observed . figure [ ss3 ] and table [ comptab2 ] show the quantitative comparison . this is much better than that shown in table [ comptab ] or figure [ ss2 ] . the values of @xmath46 and @xmath47 are closer to the observed system than before ( and comparable to the terrestrial system , although @xmath47 is still not quite large enough to match the pulsar planets ) . other measures are also in approximate agreement the masses of the planets are similar , and the mass - weighted semi - major axis @xmath48 is closer to the observed value , although still approximately 25% too large . although the values of @xmath48 are too large , we note that there is little dispersion from one realisation to the next , with @xmath48 almost always lying close to the midpoint of the annulus . thus , given the encouraging agreement in other parameters , we might improve the fit by simply postulating an annulus at a given radius effectively allowing for the possibility that our gaseous evolution models might be slightly overestimating the outward expansion of the protoplanetary disk . thus , we have calculated similar models but centered at 0.4 au rather than 0.5 au . we have also narrowed the annulus to investigate whether this can help with the mismatch in @xmath47 . the second set of results in table [ comptab2 ] thus show the results of an annulus that starts with material between 0.350.45 au . the results are indeed qualitatively similar as before , but now with @xmath48 matching the observed value , as expected . the simulations can easily match the observed @xmath46 and get within a factor of two of @xmath47 , in extreme cases . the values of @xmath48 and @xmath44 also match well , and are shown in figure [ ss4 ] . thus , viewed overall , the simulations now provide plausible matches to the observations in all the measures proposed by chambers , with the possible exception of @xmath47 . the ultimate goal of this paper is to evaluate the various proposed models for pulsar planet formation within the context of a physically motivated theory of protoplanetary disk evolution and planetary assembly . a secondary goal is to use our understanding of the successful model(s ) to illuminate our understanding of terrestrial planet formation in general , including in our own solar system and other extrasolar planetary systems . we have essentially considered formation scenarios to fall into one of two classes either solar composition protoplanetary disks , which evolve in a fashion similar to those around normal stars , or disks composed primarily of heavy metals , whose evolution we assume to be truncated at small radii according to the models of menou et al . ( 2001 ) . in the former case , these models form a one - parameter family , whose evolution is dictated by the overall angular momentum . we have examined the consequences of planet formation in this family of models and find that those with large values ( @xmath77@xmath3 ergs s ) produce too much solid material over too large a range of scales . the collisional assembly of the solid material from these models yields planets that are too large and with a significant number of secondary planets , whose presence would be easy to detect , possibly even around normal pulsars . the models which produce planetary systems that qualitatively resemble the observed system have total angular momenta at the low end of the proposed range ( @xmath27 ergs s ) . one proposed scenario that produces such values is the supernova fallback scenario . however , in most versions of this scenario the composition of the material is dominated by heavier elements , and the above evolution was calculated with an opacity and equation of state appropriate to a solar composition . to examine the consequences of heavy element disks , we also calculated models that begin with planetary embryo swarms localised at @xmath78au , to see whether sufficient mass could be pushed outwards to the locations of the pulsar planets , in the same manner as uranus & neptune could have been forced outwards in our own solar system . however , in this case , the masses are smaller and the material is deeper in the potential well , with the result that there is very little expansion and the resulting planetary systems are far too compact . the fact that our models favour the supernova fallback model is consistent with another analysis by miller & hamilton ( 2001 ) . they conclude that the b1257 + 12 pulsar is actually a relatively young pulsar , more closely associated ( at least in terms of origin ) with normal , longer period pulsars than with the dynamically older population of millisecond pulsars into whose company one would nominally place it based on the observed spin period . in part , their analysis is based on the survival of the protoplanetary disk in the face of irradiation from the central object . this is an effect that we have not considered in detail , based on the fact that the initial accretion rates from the disks we postulate are super - eddington , so that the expanding disk is already ionized , and furthermore , the inner parts of the disk are expected to shield the outer disk from much of the central object irradiation . nevertheless , our models are in complete accord , as the disk masses we use fall well above the threshold they require , and our models produce recognisable planetary systems on timescales @xmath13 years , appropriate for the dynamical age of a newly born neutron star . there is also independent observational evidence in favour of the existence of supernova fallback disks . wang , chakrabarty & kaplan ( 2006 ) report the presence of an infra - red excess associated with the isolated young neutron star 4u 0142 + 61 . this is interpreted as the reprocessing of x - rays by a source of dust in orbit around the neutron star . the wang et al . spitzer measurements place a lower limit on the disk outer edge @xmath78au , consistent with our models but not significantly constraining . however , one should note that this source is a slowly rotating neutron star , rather than a rapidly rotating one like b1257 + 12 , so that this system can not be considered an exact analogue at an earlier stage but only an example that such phenomena do occur . within the context of the rapid sedimentation models in [ dump ] , planet assembly occurs quite rapidly . figure [ assemble ] shows the assembly and orbital histories of the two massive surviving bodies in figure [ snap5 ] . we see that the bulk of the mass growth occurs between 0.1 and 1 myr , with a significant fraction resulting from mergers with other objects of significant mass . on timescales @xmath791 myr , the growth is slower , consisting essentially of the accretion of the remaining small bodies in the vicinity . we see also the orbital evolution , with the repulsive interaction of planetary scattering driving the two bodies to opposite ends of the original annulus . note that , despite the fact that our final model represents the outcome of a narrowly confined initial distribution , we never end up with a single planet . this is presumably because scattering of planetesimals can mediate a repulsive interaction between two accreting bodies , driving them apart . formation of multiple final bodies also allows for simultaneous conservation of total energy and angular momentum in this largely conservative system ( e.g. wetherill 1990 ) . the most common outcome is two or three planets of comparable mass . this is frequently the end result , but sometimes the process of accumulation does not extend fully to completion . this can be seen from the final panel in figure [ snap5 ] , as the final system does contain two scattered bodies on orbits with semimajor axes between 1 and 2 au . these appear dynamically decoupled from the larger bodies and can thus potentially survive long enough to be observed . the total mass in this scattered population is @xmath80 of the total system mass , i.e. of order a moon s mass or less , at an age of 100 myr . this should yield timing residuals @xmath81 , which is potentially observable . at an age of 10 myr ( probably the youngest plausible age for such a pulsar ) , the mass in this population is larger , @xmath82 , spread amongst several objects . figure [ ast232 ] shows how such remnant objects become dynamically decoupled from their parent population . beginning at @xmath83 myr , the periastron of the ( eventual ) outermost object in figure [ snap5 ] starts to move out slowly , so that the body no longer re - enters the region ( delineated by dashed lines ) where it is scattered by the large bodies . this drift is driven by the interaction with another scattered body whose semi - major axis is larger but whose periastron is smaller . the gravitational interaction between these two small scattered objects continues until 13.5 myr ( shown as the dotted line ) , at which point the other body is ejected from the system entirely ( recall the argument in [ dynamics ] is only statistical a few bodies are ejected rather than accreted ) , leaving a remnant body which is now dynamically decoupled from the main planetary region even though it was formed there originally . one of the most interesting aspects of the pulsar planet system is that it represents an opportunity to test the theory of terrestrial planet formation on an independent system . so , how does our preceding discussion fit in with current notions about the assembly of the solar system terrestrial planets ? the solar system planets are usually assumed to originate from a planetesimal distribution drawn originally from some variation on the minimum mass solar nebula ( e.g. weidenshilling 1977 ) or from a viscous accretion disk model of the protoplanetary nebula ( e.g. bell et al . the lack of remaining material in the region that encloses mars and the asteroid belt is usually associated with the disruptive influence of resonant perturbations associated with the giant planets ( wetherill 1992 ; lecar & franklin 1997 ; nagasawa et al . simulations of the final ( completion ) stages of accumulation from embryos into planets ( chambers & wetherill 1998 ; agnor et al . 1999 ; chambers 2001 ) , based on such models have both successes and failures , often in common with our results in [ sims ] . for reasonable choices of parameters they produce plausible analogues of earth and venus , both in terms of mass and position , although they consistently fail to reproduce mercury and mars accurately ( when they produce planets in the correct location , the planets are usually too massive ) . they also tend to produce systems that are too eccentric and not sufficiently localised ( chambers 2001 ) , in much the same manner as shown in figure [ ss2 ] . a proposed solution for this latter problem requires a dissipation mechanism , in order to reduce the epicyclic component of the surviving planets velocities . one possibility is that the final stages of accumulation occur while there is still a significant mass contained in smaller bodies , which damp the eccentricity growth through dynamical friction ( goldreich et al . 2004 ; obrien , morbidelli & levison 2006 ) . whether such a significant small body population can last for long enough is a matter of some debate ( thommes , nagasawa & lin 2008 ) , depending on the modelling of collisional accretion / fragmentation process . another possible source of dissipation is the interaction with a remaining gas disk ( papaloizou & larwood 2000 ; agnor & ward 2002 ; kominami & ida 2002 ) , although this scenario too , has been criticised . the effects of gas drag have to be tuned somewhat , otherwise the drag can be either too strong ( preventing orbit crossing to begin with ) or irrelevant ( kominami & ida 2002 ) . such tuning may , however , emerge naturally in a model where the migration of jupiter in a gaseous disk excites the eccentricities of the terrestrial embryos by resonance sweeping ( nagasawa , lin & thommes 2005 ; thommes , nagasawa & lin 2008 ) , thereby guaranteeing that the planet assembly occurs while the gas density is still high enough to be significant . the first point to note is the empirical fact that the psr b1257 + 12 system occupies an even more extreme position in the parameter space than the terrestrial system , with a similar value of @xmath46 but a significantly higher @xmath47 . this poses an interesting problem for the aforementioned resonant sweeping scenario , since there are no jovian mass planets in the system to trigger the planet assembly . while a population of smaller bodies or gas drag may indeed operate in this scenario too , it can not be triggered by the mechanism suggested by nagasawa et al . however , a second point to note is that our simulations do not overpredict the value of @xmath46 , and can match both the pulsar planet and terrestrial system quite easily . thus it does not appear necessary to invoke an additional dissipation mechanism to explain the eccentricities , although it may be required to explain the remaining mismatch in @xmath47 . the reason our simulations match @xmath46 is that the material is initially concentrated in a relatively narrow range of radii , so that the planetary eccentricities do not need to grow to the levels required to induce orbit crossing and collisions in traditional terrestrial planet accumulation studies . whether a similar concentration of material is a plausible starting point for our own solar system is a subject for future calculation . although the above results are very encouraging , there are still some remaining discrepancies . the simulations can match the observations in most of the statistics proposed by chambers , but the one remaining mismatch , in @xmath47 , suggests that there may still be some physics missing . in this case , there is still possibly some additional form of dissipation required from a population of small bodies or gas drag , although it can not be tuned by jovian planet migration as has been proposed for our solar system . we have also ignored any effects due to possible radial migration due to torques with any remnant gas disk . this could potentially explain both the more compact configuration and the fact that the best - fit annulus is slightly closer to the star than suggested by the gaseous disk simulations . tidal effects were also briefly considered but ruled out as irrelevant in this context . by systematically modelling the various formation scenarios proposed for the psr b1257 + 12 planets , we have arrived at the conclusion that the most likely origin for the observed planet properties is formation from an expanding disk of roughly solar composition , with total angular momentum @xmath27 ergs s. this is towards the low end of the angular momentum range of proposed scenarios , and consistent with models that postulate an origin for the disk in material captured during supernova fallback ( although significantly super solar metallicity does not provide an outcome consistent with the observations ) . we furthermore find that the models that provide the best fit are those in which the gaseous disk expands in the popular ` layered ' disk formalism , and that the material that condenses out of the nebula forms planetesimals rapidly and becomes decoupled from the gas on a timescale short compared to the viscous evolution time . this results in the majority of the solid material being deposited in a narrow annulus . the subsequent collisional assembly of the final planets produces planets of the right mass and approximate location , although some additional dissipation mechanism may be necessary to explain the compactness of the observed system . bh acknowledges support by the sloan foundation , the nasa astrobiology society and the space interferometry mission . in particular , the computing cluster purchased with sloan funds made the calculations presented in this paper considerably less painful than they would otherwise have been . agnor , c. b. , canup , r. m. & levison , h. f. , 1999 , icarus , 142 , 219 agnor , c. b. & ward , w. , 2002 , apj , 567 , 529 bell , k. r. , cassen , p. , klahr , h. h. & henning , t. , 1997 , apj , 486 , 372 cassen , p. , 2000 , lun . planet . , 31 , # 1141 chambers , j. e. , 1999 , mnras , 304 , 793 chambers , j. e. , 2001 , icarus , 152 , 205 chambers , j. e. & cassen , p. , 2002 , met . & plan . , 37 , 1523 chambers , j. e. & wetherill , g. w. , 1998 , icarus , 136 , 304 chevalier , r. , 1989 , apj , 346 , 847 currie , t. & hansen , b. m. s. , 2007 , apj , 666 , 1232 duncan , m. , quinn , t. & tremaine , s. , 1987 , aj , 94 , 1330 fernandez , j. a. & ip , w .- h . , 1984 , icarus , 58 , 109 gammie , c. f. , 1996 , apj , 457 , 355 hahn , j. m. & malhotra , r. , 1999 , aj , 117 , 3041 hansen , b. , 2000a , in ` stellar collisions , mergers and their consequences ' , asp conf . ser . m. m. shara , 221 , astro - ph/0008226 hansen , b. , 2000b , astro - ph/0004058 hashimoto , m. , nomoto , k. & shigeyama , t. , 1989 , a&a , 210 , l5 kokubo , e. & ida , s. , icarus , 131 , 171 kominami , j. & ida , s. , icarus , 157 , 43 konacki , m. & wolszczan , a. , 2003 , apj , 591 , l147 lecar , m. & franklin , f. , 1997 , icarus , 129 , 134 lin , d. n. c. , woosley , s. e. & bodenheimer , p. h. , 1991 , nature , 353 , 827 mayor , m. & queloz , d. , 1995 , nature , 378 , 355 mayor , m. , et al . , 2008 , arxiv:0806.4587 menou , k. , perna , r. & hernquist , l. , 2001 , apj , 559 , 1032 mizuno , h. , nakazawa , k. & hayashi , c. , 1978 , prog . phys . , 60 , 699 murray , n. , hansen , b. , holman , m. & tremaine , s. , 1998 , science , 279 , 69 nagasawa , m. , lin , d. n. c. & thommes , e. , 2005 , apj , 635 , 578 obrien , d. p. , morbidelli , a. & levison , h. f. , 2006 , icarus , 184 , 39 papaloizou , j. c. b. & larwood , j. d. , 2000 , mnras , 315 , 823 phinney , e. s. & hansen , b. m. s. , 1993 , in ` planets around pulsars ' , asp conf . ser . 36 , ed . j. a. phillips , s. e. thorsett & s. r. kulkarni , 371 podsiadlowski , p. , 1993 , in ` planets around pulsars ' , asp conf . j. a. phillips , s. e. thorsett & s. r. kulkarni , 197 pollack , j. b. , hubickyj , o. , bodenheimer , p. , lissauer , j. j. , podolak , j. & greenzweig , y. , 1996 , icarus , 124 , 62 rivera , e. j. et al . , 2005 , apj , 634 , 625 shakura , n. i. & sunyaev , r. a. , 1973 , a&a , 24 , 337 thommes , e. w. , duncan , m. j. & levison , h. f. , 1999 , nature , 402 , 635 thommes , e. w. , nagasawa , m. & lin , d. n. c. , 2008 , arxiv:0802.0541 tremaine , s. , 1993 , in ` planets around pulsars ' , asp conf . j. a. phillips , s. e. thorsett & s. r. kulkarni , 335 udry , s. et al . 2007 , a&a , 469 , l43 wang , z. , chakrabarty , d. & kaplan , d. l. , 2006 , nature , 440 , 772 weidenschilling , s. j. , 1977 , ast . , 51 , 153 wetherill , g. w. , 1990 , ann . rev . earth . , 18 , 205 wetherill , g. w. , 1992 , icarus , 100 , 307 wolszczan , a. , 1994 , science , 264 , 538 wolszczan , a. , hoffman , i. m. , konacki , m. , anderson , s. b. & wilouris , k. m. , 2000a , apj , 540 , l41 wolszczan , a. , et al . , 2000b , apj , 528 , 907 wolszczan , a. , & frail , d. a. , 1992 , nature , 355 , 145 lccc planet semi - major axis ( au ) & 0.19 & 0.36 & 0.46 + orbital period ( days ) & 25.262(3 ) & 66.5419(1 ) & 98.2114(2 ) + planet mass ( @xmath11 ) & 0.020(2 ) & 4.3(2 ) & 3.9(2 ) + eccentricity & 0.0 & 0.0186(2 ) & 0.0252(2 ) + inclination ( degrees ) & @xmath84 & 53(4 ) & 47(3 ) + & @xmath84 & 127(4 ) & 133(3 ) + lcccccccc ldv1 & 6 & 0.49 & 2.30 & 17.1 & 0.139 & 6.5 & 16.3 & 0.91 + ldv2 & 4 & 0.37 & 2.22 & 29.0 & 0.035 & 9.1 & 8.5 & 1.99 + ldv3 & 4 & 0.40 & 1.84 & 32.2 & 0.105 & 7.6 & 6.8 & 0.52 + ldv4 & 6 & 0.38 & 1.67 & 21.7 & 0.057 & 7.2 & 6.8 & 1.96 + ldv5 & 7 & 0.74 & 1.91 & 14.8 & 0.093 & 17.6 & 11.3 & 1.16 + ldl1 & 7 & 0.55 & 1.90 & 18.2 & 0.086 & 8.0 & 12.0 & 1.21 + ldl2 & 7 & 0.49 & 1.18 & 17.9 & 0.048 & 8.5 & 12.0 & 1.13 + ldl3 & 4 & 0.56 & 1.38 & 32.7 & 0.025 & 8.6 & 10.8 & 1.90 + ldl4 & 6 & 0.52 & 1.35 & 22.8 & 0.044 & 7.7 & 10.0 & 0.87 + ldl5 & 5 & 0.36 & 1.86 & 20.6 & 0.063 & 5.4 & 9.6 & 0.38 + sdv1 & 6 & 0.35 & 0.68 & 26.5 & 0.0058 & 10.7 & 3.4 & 1.25 + sdv2 & 7 & 0.29 & 0.65 & 24.7 & 0.0111 & 7.1 & 3.0 & 1.06 + sdv3 & 9 & 0.33 & 0.63 & 19.5 & 0.0125 & 9.3 & 3.4 & 0.24 + sdv4 & 9 & 0.36 & 0.63 & 25.9 & 0.0120 & 8.5 & 3.4 & 0.72 + sdv5 & 8 & 0.38 & 0.57 & 21.1 & 0.0230 & 5.0 & 3.8 & 0.80 + sdl1 & 5 & 0.37 & 0.41 & 25.4 & 0.0005 & 21.5 & 3.0 & 0.30 + sdl2 & 5 & 0.27 & 0.38 & 29.4 & 0.0027 & 24.7 & 2.2 & 0.69 + sdl3 & 5 & 0.29 & 0.40 & 25.2 & 0.0021 & 19.3 & 2.3 & 0.22 + sdl4 & 5 & 0.31 & 0.42 & 25.2 & 0.0024 & 22.1 & 2.5 & 0.29 + sdl5 & 6 & 0.33 & 0.44 & 22.3 & 0.0018 & 25.9 & 2.6 & 0.36 + hd1 & 3 & 0.42 & 0.08 & 20.9 & 0.0026 & 50.5 & 8.4 & 0.09 + hd2 & 2 & 0.64 & 0.06 & 41.3 & 0.0007 & 42.8 & 5.1 & 0.09 + hd3 & 2 & 0.56 & 0.07 & 27.5 & 0.0002 & 92.6 & 5.6 & 0.09 + hd4 & 3 & 0.42 & 0.07 & 24.3 & 0.0011 & 55.6 & 4.2 & 0.07 + hd5 & 3 & 0.63 & 0.07 & 26.2 & 0.0017 & 75.1 & 6.3 & 0.07 + b1257 + 12 & 3 & 0.52 & 0.41 & 28.0 & 0.0037(28 ) & 320 & 4.3(2 ) & 0.36 + mvem & 4 & 0.51 & 0.90 & 37.7 & 0.0018 & 90 & 1.0 & 1.0 + lcccccccc flr1 & 3 & 0.69 & 0.50 & 29.0 & 0.0039 & 77.1 & 5.7 & 0.53 + flr2 & 3 & 0.73 & 0.51 & 42.6 & 0.0057 & 69.5 & 6.1 & 0.42 + flr3 & 4 & 0.55 & 0.52 & 28.6 & 0.0040 & 54.9 & 4.6 & 0.38 + flr4 & 3 & 0.46 & 0.51 & 25.0 & 0.0036 & 65.4 & 3.7 & 0.50 + flr5 & 4 & 0.60 & 0.53 & 28.5 & 0.0050 & 73.7 & 4.8 & 0.53 + flr6 & 3 & 0.46 & 0.52 & 25.8 & 0.0050 & 60.3 & 3.7 & 0.38 + flr7 & 4 & 0.65 & 0.55 & 27.1 & 0.0020 & 84.2 & 5.6 & 0.57 + flr8 & 6 & 0.52 & 0.64 & 31.7 & 0.0275 & 35.7 & 4.5 & 0.44 + flr9 & 4 & 0.54 & 0.57 & 25.1 & 0.0017 & 71.2 & 4.6 & 0.42 + flr10 & 4 & 0.59 & 0.51 & 32.9 & 0.0026 & 71.4 & 4.6 & 0.61 + flr11 & 4 & 0.38 & 0.52 & 22.1 & 0.0010 & 78.6 & 2.9 & 0.38 + flr21 & 3 & 0.52 & 0.52 & 30.5 & 0.0031 & 65.3 & 4.1 & 0.39 + flr22 & 3 & 0.64 & 0.52 & 31.9 & 0.0044 & 62.9 & 5.0 & 0.50 + flr23 & 4 & 0.39 & 0.53 & 25.7 & 0.0030 & 59.8 & 3.0 & 0.37 + flr24 & 3 & 0.59 & 0.52 & 33.9 & 0.0015 & 76.4 & 4.6 & 0.40 + flr25 & 6 & 0.52 & 0.54 & 26.4 & 0.0098 & 48.0 & 4.0 & 0.39 + flr26 & 4 & 0.46 & 0.52 & 26.7 & 0.0034 & 66.8 & 3.6 & 0.39 + flr27 & 3 & 0.39 & 0.52 & 25.4 & 0.0021 & 57.2 & 3.0 & 0.36 + flr28 & 6 & 0.42 & 0.52 & 25.8 & 0.0061 & 49.0 & 3.3 & 0.61 + flr15 & 2 & 0.77 & 0.401 & 25.4 & 0.0003 & 166 & 6.4 & 0.36 + flr16 & 5 & 0.86 & 0.433 & 19.8 & 0.0116 & 56.3 & 7.2 & 0.36 + flr17 & 3 & 0.56 & 0.411 & 24.8 & 0.0004 & 88.4 & 4.7 & 0.38 + flr18 & 2 & 0.66 & 0.408 & 30.0 & 0.0012 & 92.8 & 5.5 & 0.35 + flr19 & 3 & 0.92 & 0.412 & 28.9 & 0.0037 & 113 & 7.7 & 0.37 + flr20 & 3 & 0.68 & 0.411 & 26.5 & 0.0012 & 111 & 5.7 & 0.43 + flr29 & 3 & 0.64 & 0.408 & 31.5 & 0.0005 & 88 & 4.9 & 0.33 + flr30 & 3 & 0.60 & 0.404 & 22.8 & 0.0004 & 106 & 4.7 & 0.43 + flr31 & 3 & 0.58 & 0.411 & 23.8 & 0.0013 & 96.7 & 4.5 & 0.42 + flr32 & 4 & 0.55 & 0.410 & 47.9 & 0.0007 & 99.6 & 4.3 & 0.48 + flr33 & 4 & 0.54 & 0.406 & 28.9 & 0.0015 & 99.3 & 4.2 & 0.32 + flr34 & 3 & 0.38 & 0.408 & 18.7 & 0.0012 & 107 & 3.0 & 0.42 + flr35 & 3 & 0.47 & 0.406 & 22.3 & 0.0004 & 77.8 & 3.7 & 0.30 + flr36 & 3 & 0.61 & 0.409 & 27.6 & 0.0021 & 92.9 & 4.7 & 0.43 + flr37 & 3 & 0.51 & 0.411 & 37.2 & 0.0013 & 105 & 4.0 & 0.48 + flr38 & 4 & 0.64 & 0.403 & 31.9 & 0.0025 & 94.8 & 5.0 & 0.36 + flr39 & 3 & 0.72 & 0.410 & 35.1 & 0.0031 & 89.5 & 5.6 & 0.45 + flr40 & 3 & 0.57 & 0.411 & 32.0 & 0.0014 & 83.7 & 4.4 & 0.32 + flr41 & 3 & 0.61 & 0.407 & 23.0 & 0.0010 & 96.9 & 4.7 & 0.33 + flr42 & 3 & 0.88 & 0.402 & 35.1 & 0.0018 & 143.3 & 6.9 & 0.37 + flr43 & 2 & 0.73 & 0.406 & 21.3 & 0.0005 & 222.4 & 5.7 & 0.37 + flr44 & 3 & 0.55 & 0.403 & 27.5 & 0.0006 & 103.6 & 4.2 & 0.33 + flr45 & 4 & 0.60 & 0.416 & 26.3 & 0.0023 & 67.7 & 4.7 & 0.32 + flr46 & 2 & 0.70 & 0.399 & 28.4 & 0.0002 & 115.8 & 5.4 & 0.34 + flr47 & 2 & 0.80 & 0.407 & 30.8 & 0.0005 & 127.1 & 6.2 & 0.36 + flr48 & 3 & 0.40 & 0.412 & 22.3 & 0.0042 & 78.6 & 3.1 & 0.31 + flr49 & 4 & 0.60 & 0.408 & 22.9 & 0.0010 & 121.1 & 4.6 & 0.45 + flr50 & 3 & 0.43 & 0.410 & 21.4 & 0.0012 & 96.0 & 3.4 & 0.51 + flr51 & 3 & 0.47 & 0.409 & 23.2 & 0.0012 & 80.0 & 3.6 & 0.41 + flr52 & 3 & 0.72 & 0.408 & 34.8 & 0.0083 & 64.6 & 5.6 & 0.40 + flr53 & 3 & 0.54 & 0.409 & 41.4 & 0.0007 & 77.4 & 4.2 & 0.31 + flr54 & 3 & 0.50 & 0.410 & 23.3 & 0.0022 & 86.1 & 3.9 & 0.39 + flr55 & 3 & 0.59 & 0.417 & 37.3 & 0.0040 & 78.6 & 4.6 & 0.33 + b1257 + 12 & 3 & 0.52 & 0.408 & 28.0 & 0.0037(28 ) & 320 & 4.3(2 ) & 0.36 + mvem & 4 & 0.51 & 0.898 & 37.7 & 0.0018 & 90 & 1.0 & 1.0 +

the inverse problem which consists of determining the flow at the earth s core mantle boundary according to an outer core magnetic field and secular variation model , has been investigated through a bayesian formalism . to circumvent the issue arising from the truncated nature of the available fields , we combined two modelization methods . in the first step , we applied a filter on the magnetic field to isolate its large scales by reducing the energy contained in its small scales , we then derived the dynamical equation , referred as filtered frozen flux equation , describing the spatio - temporal evolution of the filtered part of the field . in the second step , we proposed a statistical parametrization of the filtered magnetic field in order to account for both its remaining unresolved scales and its large scale uncertainties . these two modelization techniques were then included in the bayesian formulation of the inverse problem . to explore the complex posterior distribution of the velocity field resulting from this development , we numerically implemented an algorithm based on markov chain monte carlo methods . after evaluating our approach on synthetic data and comparing it to previously introduced methods , we applied it on real data for the single epoch .

1310.6848

the core magnetic field ( mf ) of the earth is sustained by the dynamo action taking place in its outer core . here , the variations in chemical composition and in temperature of the liquid metal allows convection to develop . since the fluid is electrically conducting , it interacts non - linearly with the magnetic field . while the flow is advecting the mf , this latter is constraining the fluid motions through the lorentz forces . the evolution of the fluid velocity field ( vf ) and the magnetic field are therefore entirely connected to one another through energy exchanges between them . studying such a system is difficult in many aspects . from a numerical point of view , simulating directly the dynamic of the outer core is a challenging task . because of the strong regime of turbulence that the magnetohydrodynamic ( mhd ) flow exhibits , the separation between the smallest and the largest scales of the system is extremely broad . yet , to properly describe the evolution of both the vf and the mf , all these scales should be considered in simulations since they interact non - linearly together . but this is , with the actual computation power available , impossible . from an experimental point of view , observing directly the evolution of the outer core is also impossible . nevertheless , measurements of the mf from ground observatories or satellites may allow to indirectly infer some dynamical properties of the flow and the mf in the core . in particular , because of the low conductivity of the mantle ( @xcite ) , a knowledge of the core mf at the earth s surface is sufficient to evaluate it at the level of the core - mantle boundary ( cmb ) . at this very specific location , the mf is coupled to the outer core vf through the frozen flux equation ( a simplified version of the induction equation introduced by @xcite , and in which the diffusion effects are neglected ) . by inverting this equation it is therefore possible to evaluate the vf at the cmb . unfortunately , the problem is ill - posed for different reasons . first , the two components of the velocity field are connected to the radial component of the secular variation ( sv ) through a unique equation . then , the available secular variation given by mf models such as grimm 2 of @xcite is only resolved at large scales whereas any scale of the velocity field can contribute to this resolved sv . finally , as it is the case for the sv , only the large scale core mf can be determined at the earth surface , yet interactions between the unknown small scale mf with the vf can generate large scale sv ( see @xcite ) . as shown by @xcite , to resolve the non - uniqueness of the velocity field in this inverse problem , constraints have to be imposed to the flow behavior . different formulations have been proposed over the past decades . this includes tangential - geostrophy , tangential - magnetostrophy , columnar flow , helical flow , or purely toroidal flow ( see @xcite and @xcite for an exhaustive review of the different constraints usually applied and their physical implications ) . nevertheless , these physical constraints are not sufficient to provide a unique flow solution ( see @xcite and references therein ) , and additional regularization assumptions have to be introduced in the problem ( see @xcite ) . although identified for more than twenty years ( @xcite ) , it is only recently that the effects of the unknown small scale mf on the large scale sv have been modeled in the inversion of the frozen flux approximation . in particular two methods have given promising results , namely the ensemble approach of @xcite , and the iterative method of @xcite . the philosophy of these two approaches is quite distinct from one another . whereas in the method of @xcite an ensemble of magnetic field containing small scales is generated and directly used to evaluate an ensemble of velocity field , in the method of @xcite the effects of the unknown magnetic field is transposed into a modeling error which is iteratively estimated . in this study , we propose a development of these approaches in the context of bayesian modeling . the article is therefore constructed in the following manner . in section [ governingequations ] , after describing the principle of derivation of the frozen flux equation , we recall the issue raised in the inversion of this equation by the truncated nature of the available mf . we then describe how to isolate the large scales of the mf and present the approximated dynamical equation , referred as filtered frozen flux ( fff ) equation , which determines the evolution of these large scales . at the end of the section we present how to formulate the inverse problem in a bayesian framework when variations of the mf around the prescribed one are allowed to occur . in section [ numericalmethod ] tests on synthetic data are performed . in the first one , we evaluate the improvement brought by using the fff equation in the inverse problem . in the second test , the bayesian formalism developed in section [ governingequations ] is considered to recover the velocity field from a set of artificially generated data , and the results are compared to the flow obtained with three other approaches . the methodology we developed is then used to determine the velocity field and its underlying uncertainties for the epoch @xmath0 . finally we present our conclusions in section [ conclusion ] . in this section we introduce briefly the hypothesis necessary to derive the frozen flux approximation . for a more detailed description see @xcite . outside the core , the geodynamo s mf @xmath1 is irrotational , it can therefore be expressed through a potential @xmath2 according to the relation : @xmath3 as mentioned previously , the low conductivity of the mantle ( @xcite ) allows to evaluate the mf and therefore its radial component @xmath4 at the level of the core mantle boundary . there its evolution is prescribed by the induction equation : @xmath5 with @xmath6 the horizontal divergence operator , @xmath7 the two - dimensional velocity field , @xmath8 the magnetic diffusivity , and @xmath9 a radial unitary vector . because for the earth , on short period of time the dissipation effects are dominated by advection effects ( see @xcite ) and since our study is limited to single epoch inversion , the fluid can be considered as a perfect conductor . @xcite showed that under this assumption , known as the frozen flux ( ff ) approximation , the induction equation can be simplified as : @xmath10 to perform a consistent inversion of the frozen - flux equation at the cmb , in addition to imposing a certain behavior to the flow and to regularizing it ( see @xcite ) , every scale composing the secular variation and the magnetic field have to be known . unfortunately , in the actual models describing the spatio - temporal evolution of the earth s magnetic field derived from satellite and observatory data , the resolution of both the sv and the mf is limited . for the grimm 2 model of @xcite , for example , the fields do not excess the degree @xmath11 when expanded in spherical harmonics . so if @xmath12 corresponds to the spherical harmonics ( sh ) coefficient at degree @xmath13 and order @xmath14 associated with the scalar potential @xmath2 , such as : @xmath15 where @xmath16 is the core radius , and @xmath17 is the schmidt semi - normalized sh , then according to equation ( [ potential ] ) , the available mf @xmath18 , and its unknown part @xmath19 are respectively given by : @xmath20 with @xmath21 the cut - off scale ( which is equal to @xmath11 for the mf provided by the grimm 2 model ) . note that the total radial component of the mf corresponds to the sum of these two quantities . to account for the truncated nature of the available mf and sv , the frozen flux approximation has to be rewritten as : @xmath22 where the advection term , on the right hand side of equation ( [ truncatedff1 ] ) , is decomposed into two parts , one depending on the resolved magnetic field and the other function of the undetermined field . @xcite were the first to highlight the issue raised by the unknown part of the mf in the inversion of the ff equation . nevertheless , because at this epoch the uncertainties on sv measurements were large , the contribution of the term @xmath23 could be neglected in the inverse problem . the recent increase in quality of both the measurements and models describing the evolution of the core mf ( see @xcite ) invalidate this latter statement , and @xcite showed that the unresolved part of the mf could not be neglected anymore . to model the effects of the unresolved part of the magnetic field on the large scale secular variation , assumptions on this unknown field behavior have to be made . in particular one can prescribe to it a certain energy spectrum . this operation can be performed , for instance , by extrapolating the spectrum associated with the resolved scales which is defined by : @xmath24 the resulting spectrum @xmath25 can then be used to statistically model the unknown mf @xmath19 . following the development of @xcite where the field is assumed to be isotropically distributed , the covariance of the coefficients @xmath26 is directly proportional to the extrapolated spectrum through the relation : @xmath27 = \frac { e_{b^>}(l)}{(l+1)(2l+1)}\delta(l - l^\prime ) \delta(m - m^\prime ) \ , \label{variances}\ ] ] where @xmath28 $ ] corresponds to the mathematical expectation . the issue with such a modelization is that no universal spectrum of the geodynamo s mf at the core mantle boundary is available . nevertheless , different formulations have been proposed over the past decades , but unfortunately most of them strongly differ from one another . an illustration of this statement is given in figure [ filteredextendedspectra ] where the spectrum of the mf prescribed by the grimm 2 model for the epoch @xmath0 ( between sh degree @xmath29 ) together with three different extrapolations ( thin lines ) are plotted . in this example , the laws derived by @xcite , @xcite and @xcite were used to extrapolate the resolved scale spectrum . they respectively read : @xmath30 with @xmath31 and @xmath32 . the constants @xmath33 , @xmath34 and @xmath35 were determined by fitting the resolved magnetic field spectrum between the degree @xmath36 . ( between sh degree @xmath29 ) and its filtered ( thick lines ) and non filtered ( thin lines ) extrapolations . the laws used to extrapolate the large scale spectrum were taken from @xcite ( plus symbols ) , @xcite ( circles ) and @xcite ( triangles ) . ] as it is confirmed in figure [ filteredextendedspectra ] the three extrapolated spectra are presenting distinct behaviors . furthermore , the law proposed by @xcite predicts a strong concentration of energy at small scales . if this latter formulation was to be the closest one from reality , the unresolved mf would have to be modeled at very high degree ( up to @xmath37 ) , which is nowadays numerically impossible we propose therefore to reduce the impact of the small scale magnetic field on the large scale secular variation by applying a filter on the magnetic field and by determining the dynamical equation governing the evolution of the resulting filtered field . as it is the case for the earth s dynamo , the degree of turbulence in astrophysical or geophysical systems is most of the time extremely high . this implies that the fields are populated by a great variety of scales . capturing all these scales in observations or in numerical simulations is usually impossible . therefore , focusing on the large scales , and their dynamics , is probably the best solution to study such systems . this approach has been widely followed in the context of hydrodynamic ( hd ) and magnetohydrodynamic ( mhd ) turbulence ( see @xcite ) . to extract the large scales from a given field , one has to filter it . in hd or mhd turbulence three different filters are generally used , the top - hat or the gaussian filter for studies in cartesian space , or the sharp cut - off filter in spectral space . while the top - hat or gaussian filter are continuous filters , the cut - off filter truncate the field above a certain predefined scale . in spherical coordinates the top - hat filter is usually preferred ( @xcite ) since until recently no equivalent of the gaussian filter existed for such geometries ( see however @xcite ) . in our study , we decided to consider the isotropic filter developed by @xcite . the principle of derivation of this filter is detailed in appendix [ appa ] . applying this filter to a scalar field @xmath38 leads to : @xmath39 where @xmath40 is the filtered field , the @xmath41 symbol denotes the convolution product over the sphere between convolution kernel @xmath42 and the scalar field @xmath43 . according to the convolution theorem ( see @xcite ) , in spectral space equation ( [ convolution ] ) becomes : @xmath44 where @xmath45 and @xmath46 are respectively the filtered and the total sh coefficients associated with @xmath43 , and @xmath47 is convolution kernel expressed in spectral space at degree @xmath13 . since for the filter developed by @xcite , @xmath42 is zonal , only its coefficients at order @xmath48 are different from zero . as shown in appendix [ appa ] equation ( [ spectralfilter ] ) , equation ( [ spectralconvolution ] ) can be expressed as : @xmath49 where @xmath50 corresponds to the filter width . with this formulation one can directly connect the spectrum of the scalar field @xmath51 to the spectrum of its filtered part @xmath52 as : @xmath53 we then applied this filter to the three different extrapolated spectra presented in the previous section ( see equations ( [ buffett])-([voohries ] ) ) . in figure [ filteredextendedspectra ] the non filtered ( thin lines ) , and the filtered ( thick lines ) spectra are plotted . the width of the filter has been set to @xmath54 km in order to preserve the large scale fields and to suppress the small scale ones . one can observe that most of the energy that was contained in the small scales of the initial fields has now vanished in the filtered fields . furthermore , the variations between the different extrapolated filtered spectra are now much less pronounced than in the non filtered case . applying the spatial filter presented in the preceding section to the velocity and the magnetic field allows to decompose them into an averaged part @xmath55 and @xmath56 , and a fluctuating part @xmath57 and @xmath58 such as : @xmath59 since the spatial filter considered here is homogeneous and time invariant , it commutes with all the differential operators encountered in the induction equation . therefore applying this filter to the frozen flux approximation ( equation ( [ frozenflux ] ) ) leads to : @xmath60 following the decomposition introduced by @xcite , the part of the subgrid stress tensor which allows interactions between the tangential components of the velocity field and the radial component of the mf reads : @xmath61 now equation ( [ filteredfrozenflux ] ) can be rewritten as : @xmath62 the subgrid stress tensor @xmath63 , through the averaged product @xmath64 , incorporates interactions between averaged and fluctuating part of the fields , thus equation ( [ fff ] ) is not closed in the sense that it does not only contains large scale quantities . to close this equation , expressions of the fluctuating quantities depending on the averaged ones have to be derived . as mentioned in appendix [ appa ] , the filtered velocity and magnetic field are both solutions of the diffusion equations : @xmath65 where the filtered and total fields are respectively @xmath66 and @xmath67 , and @xmath68 and @xmath69 , and the product between @xmath70 and the two diffusion coefficients @xmath71 and @xmath72 determines the width of the filter . note that since @xmath71 is the diffusion coefficient applied to each term of the frozen flux equation it has to be applied to the mf but not necessarily to the vf . to link fluctuating to filtered fields , taylor expansions at the first order in @xmath70 of @xmath7 , @xmath73 , and @xmath74 , using relations ( [ diffusionu ] ) to ( [ diffusionub ] ) are performed , leading to : @xmath75 by analogy to the usual gaussian filters used in large eddy simulations , one can define the following characteristic lengths : @xmath76 letting @xmath77 , injecting equations ( [ approxu ] ) and ( [ approxb ] ) into equation ( [ approxub ] ) , and keeping only the first order terms , one gets : @xmath78 which can be reduced to : @xmath79 so the subgrid stress tensor @xmath63 in its approximated closed form reads : @xmath80 the filtered frozen flux equation can now be written for the filtered secular variation @xmath81 truncated at degree @xmath82 as : @xmath83 with @xmath84 . in our study , the extrapolated filtered magnetic field @xmath85 is extended up to the spherical harmonic degree @xmath86 . above this scale , the filtered magnetic field ( for @xmath87 km ) is extremely weak ( see figure [ filteredextendedspectra ] ) and its influence on the large scale secular variation is neglected . the problem of determining the velocity field at the cmb knowing the exact mf and the sv together with its uncertainties is an ill - posed inverse problem : in a discrete approximation , the number of unknown is twice as large as the number of equations . one of the most common methods to tackle this problem consists in minimizing an energy functional composed of two main terms ; a quantity measuring the discrepancies between the model and the data , balanced , through a regularization parameter , with a quantity expressing a prior knowledge on the expected solution . this method has been widely used to evaluate the flow at the cmb and for a review of the different parametrization employed see @xcite . another option consists in formulating the problem in a bayesian framework . the solution becomes then the full posterior distribution of the velocity field given the secular variation . this method allows the estimation of a model for the flow together with the quantification of its uncertainties . when variations around the prescribed mf are allowed to occur , the inverse problem becomes more complicated . for models of the earth s core mf derived from satellite or observatory data , the nature of these variations is diverse . at large scales they can be due to a leakage of the external and lithospheric field into the core field , whereas at small scales , the entire mf is undetermined because of the dominance of the lithospheric field at the earth s surface . recent models such as grimm 2 ( @xcite ) are able to separate at large scales ( between sh degree @xmath29 ) the external from the core field , but not the lithospheric field from the core field . as a consequence , the large scale lithospheric field becomes a source of uncertainty on the geodynamo s mf . to parametrize the unresolved part of the mf in the inverse problem different approaches have been recently developed . this includes the ensemble method of @xcite and the iterative algorithm of @xcite . in this study we propose to extend these methods to the context of bayesian modeling , following the development of @xcite . furthermore , in addition to parametrizing the unresolved mf , we also consider the uncertainties on the large scale mf due to the lithospheric field . from now on , in order to simplify the notations , the filtered mf and vf as well as the filtered and truncated sv will be written as : @xmath88 in this section , the distribution we want to characterize is the posterior distribution of the velocity field given the secular variation @xmath89 . but since we want to account for the unknown small scale magnetic field together with the uncertainties on the large scale field , this distribution can not be expressed directly . nevertheless , it can be obtained by marginalizing the joint posterior distribution of the velocity field and the magnetic field as following : @xmath90 according to @xcite the distribution on the right hand side of relation ( [ posterior1 ] ) can be decomposed into : @xmath91 with @xmath92 the joint prior distribution of the vf and the mf , @xmath93 the distribution of the sv , which is constant with respect to both @xmath94 and @xmath95 , and finally @xmath96 the likelihood distribution . because @xmath94 and @xmath95 are a priori assumed to be independent random variables their joint distribution can be split into two distributions such as : @xmath97 the prior distribution of the vf @xmath98 is assumed to be gaussian with the following general form : @xmath99}{(2\pi)^{\frac{d}{2 } } vf vector , which in our case is twice as large as the dimension of the mf and sv , and @xmath100 the velocity covariance matrix chosen to enforce the spatial smoothness of the flow . for this purpose we impose that : @xmath101 the right hand - side of equation ( [ norm ] ) is composed of two different norms on @xmath94 , namely the bloxham s `` strong norm '' ( @xcite ) for the first one and the standard @xmath102-norm for the second one . the domain of integration @xmath103 is the surface of the sphere describing the cmb , and the balance factor @xmath104 is chosen to be small enough not to modify the correlation length induced by the bloxham norm . furthermore , the covariance is rescaled such as the averaged standard deviation of the velocity field intensity is @xmath105km / yr . this implies that at any location of the core - mantle boundary , the probability for the flow intensity to excess the value of @xmath106km / yr , an upper limit calculated by @xcite , is of the order of @xmath107 . as for the velocity field , the magnetic field @xmath95 is also assumed to be normally distributed , but with a mean @xmath108 corresponding to the resolved mf , and a covariance @xmath109 . its prior distribution can therefore be expressed as : @xmath110}{(2\pi)^{\frac{d}{4}}|\sigma_{b}|^{\frac{1}{2 } } } \label{bprior } \ . \ ] ] in this equation the two parts of the mf @xmath111 have different behaviors . the truncated field @xmath112 is taken from the grimm 2 model with @xmath113 = b_0 $ ] whereas the unknown mf @xmath114 is characterized by the averaged value @xmath115 = 0 $ ] . since in the grimm 2 model , the core field and the litospheric field are overlapping at large scales , this latter field can be viewed as a source of uncertainties on @xmath112 . we propose therefore to use the theoretical spectrum of the litospheric field given by @xcite to build the covariance matrix @xmath116 for the resolved scales magnetic field . this spectrum reads : @xmath117 with @xmath118 the averaged crust magnetization , @xmath119 km the equivalent magnetized layer thickness , and the constants @xmath120 , @xmath121 , and @xmath122 km . the two functions @xmath123 and @xmath124 are given by : @xmath125 letting the truncated mf @xmath112 being linked to its spectral counterpart @xmath126 through the relation @xmath127 , where the operator @xmath128 projects the coefficients in physical space , and the operator @xmath129 filters the field , and assuming that the litospheric field is isotropically distributed , the covariance of the truncated mf is : @xmath130 \\ & \!\ ! = \!\ ! & \!\ ! e \left[pf\left ( g_{l , m}^ < - g_{l , m}^0\right ) \left ( g_{l , m}^ < - g_{l , m}^0\right)^tf^tp^t\right ] \\ & \!\!= \!\!&\!\ ! pf \frac{e_b^l(l)}{(l+1)(2l+1)}f^tp^t \label{lsmfcovariance}\ , \end{aligned}\ ] ] where @xmath131 are the spherical harmonics coefficients associated with @xmath108 . the extrapolated sh coefficients of the mf @xmath26 are assumed to individually have a @xmath132 mean and a variance depending on the extrapolated spectrum @xmath133 from @xcite as shown in equation ( [ variances ] ) . therefore , in physical space , the mf also has a @xmath132 mean at any spatial location and a covariance given by : @xmath134 by combining @xmath116 with @xmath135 , one gets the total covariance @xmath109 for the mf . the last distribution to characterize is the likelihood distribution which measures the discrepancies between the model ( the filtered frozen flux equation ) and the data ( the secular variation ) . it reads : @xmath136 } { ( 2\pi)^{\frac{d}{4}}|\sigma_{\gamma}|^{\frac{1}{2 } } } \ , \label{likelyhood}\ ] ] where the operator @xmath137 , when applied to @xmath95 , allows to calculate the non linear term of the filtered frozen flux equation @xmath138 , and the covariance matrix @xmath139 is the sv posterior covariance of the grimm 2 model . all the distributions entering the velocity field posterior distribution being detailed , this latter can be evaluated . the integral given in equation ( [ posterior1 ] ) has already been calculated by @xcite , so we only present the result which reads : @xmath140 \label{likelihood1}\end{aligned}\ ] ] with : @xmath141 by multiplying the density ( [ likelihood1 ] ) with the prior distribution ( [ prioru ] ) one gets the posterior distribution : @xmath142 \label{posteriortotj}\nonumber\end{aligned}\ ] ] since this expression does not allow to easily apprehend the effects arising from the modelization of the mf variations on the posterior distribution of the velocity field , we decided to rewrite it into a more intuitive form . it reads : @xmath143 \nonumber \\ & \times & \frac{1 } { |\sigma_u|^{\frac{1}{2}}}\exp\left[-\frac{1}{2 } u^t \sigma_u^{-1 } u \right ] \times \frac{1}{p(\gamma ) } \label{posteriortot1}\\ \sigma_{\tilde{\gamma } } & = & \sigma_{\gamma } + a_{u } \sigma_b a_{u}^t \ .\end{aligned}\ ] ] this formulation is very similar to the posterior distribution of the velocity field in the case where the magnetic field is exactly know . indeed this latter distribution can be obtained by simply replacing the covariance matrix @xmath144 by @xmath139 . one can therefore observe that accounting for the small scale magnetic field and the lithospheric field when formulating the inverse problem in a bayesian framework leads to an increase of the secular variation uncertainties through the quantity @xmath145 . because of the dependency of this latter term on the velocity field @xmath94 , the maximum of the posterior distribution can not be analytically calculated as already mentioned by @xcite . nevertheless , it is numerically possible to extract the main statistical characteristics of this posterior distribution using a markov chain monte carlo method ( see @xcite ) . the algorithm we chose to explore the posterior distribution and the results we obtained are presented in the next section . the sv , mf and vf are expressed in physical space , therefore the surface of the cmb has been discretized by recursively dividing an initial icosahedron ( see figure [ mesh ] ) . the grid construction and its properties , as well as the approximation of the differential operators are explicited in appendix [ appb ] . given that the grid is composed of @xmath146 nodes , the vectors @xmath95 , @xmath147 , and @xmath148 are given by : @xmath149 to evaluate the fff model , an inversion of this equation was performed using grimm 2 as well as artificially generated data . the large scale magnetic field was taken from the grimm 2 model at the epoch @xmath150 , whereas the small scales were randomly generated for sh degree lying between @xmath151 and @xmath152 according to the exponential law ( [ buffett ] ) . the mf was then filtered ( with @xmath153 km ) such as its smallest scales , which can not be properly represented on the grid , exhibited a low energy level . this mf is referred as @xmath108 . to create a sv associated with this mf , we drew randomly a velocity field and used it to advect the mf with the ff equation ( [ frozenflux ] ) . the velocity field was decomposed into a poloidal and a toroidal part such as : @xmath154 in spectral space , the spherical harmonics coefficients for the poloidal and toroidal field respectively read @xmath155 and @xmath156 . in order to promote interactions between small and large scales , @xmath155 and @xmath156 were extended up to spherical harmonic degree @xmath157 with the following statistical properties : @xmath158 = e[\psi_{l , m } ] = 0 \quad \forall l , m\ ] ] @xmath159 & = & c l^{-14/3 } |m|^{-11/3}\delta_{l l^{\prime } } \delta_{m m^{\prime } } \\ e[\psi_{l , m } \psi_{l^\prime , m^\prime } ] & = & c l^{-14/3 } |m|^{-11/3}\delta_{l l^{\prime } } \delta_{m m^{\prime } } \\ e[\phi_{l , m } \psi_{l^\prime , m^\prime } ] & = & 0 \quad \forall l , l^{\prime},m , m^{\prime}\ , \end{aligned}\ ] ] where @xmath160 is a normalization constant . the choice of a @xmath161 power law imposed onto the poloidal and toroidal field correlation allowed to generate a flow with similar statistical properties than a two - dimensional turbulent flow ( see @xcite ) . to directly simulate the advection of the mf , a fourth order runge - kutta scheme had been implemented . the integration time was taken to be @xmath162 year , and the computation was performed on a grid refined @xmath163 times ( with @xmath164 nodes ) . with an artificial velocity field . ] in figure [ dtbadvected ] , the spectrum of the resulting secular variation is plotted . as it can be observed , the energy of the sv is maximum at scales lying between @xmath165 and @xmath166 indicating that the mf is strongly affected by the velocity field at these scales . to mimic the situation encountered when using models fitting satellite data where both the mf and sv can not be entirely taken into account , the artificial fields were truncated at degree 40 . the resulting mf and sv are respectively referred as @xmath167 and @xmath168 . these fields were then used as an input for the inversion of the fff equation . two different filter widths were tested , @xmath54 km , and @xmath169 km . in the latter case the equation reduces then to the usual ff approximation . since the statistical properties of the velocity field were exactly known , they were directly injected in the prior information @xmath170 . no measurement errors on the secular variation and the magnetic field had been generated , therefore the likelihood and the mf prior distributions respectively read : @xmath171 multiplying together these two distributions and marginalizing the result with respect to @xmath95 leads to : @xmath172 with @xmath173 the subgrid stress tensor evaluated with @xmath168 . the posterior distribution of the velocity field is then proportional to : @xmath174 the bayesian formulation of the problem being described , the discrete velocity field that maximizes the posterior distribution could be determined for the two cases . to do this , a particular solution of the equation @xmath175 was calculated . in addition , the null space of the operator @xmath176 defined such as @xmath177 was parametrized . the final solution corresponded then to the sum of the particular solution with the null - space one which minimized the prior information on the vf . the grid used to realize the computation was composed of @xmath178 nodes ( approximately @xmath179 times less than the one taken to advect the mf ) . for the results to be comparable , the three different velocity fields were truncated at sh degree @xmath180 . furthermore the artificially generated field as well as the field obtained by inverting the ff approximation ( @xmath169 ) , were filtered with @xmath54 km . on the left side of figure [ artificialspectra ] the poloidal and toroidal spectra respectively defined by : @xmath181 are plotted for the three velocity fields . the behavior of the spectra associated with the artificial field ( full line ) is correctly reproduced at large scale ( @xmath182 ) by both the ff ( crosses ) and the fff ( circles ) flow models , whereas at sh degree close to the cutoff @xmath183 , the discrepancies between the exact spectra and the spectra of the inverted fields become larger . nevertheless , the spectra of the difference between exact and inverted fields , displayed on the right side of figure [ artificialspectra ] , show that the use of the fff equation allows to reproduce more accurately the artificial field at almost any scale . to quantify the spatial error of the two inverted fields , the following quantities were computed : @xmath184 where the integration domain is the surface of the cmb , @xmath94 is the exact filtered velocity field , @xmath185 and @xmath186 are respectively the filtered velocity field obtained by inverting the ff and the fff equations . the result of these computations shows that although the energy associated to the error fields is weak in comparison to the total energy of the exact flow , performing an inversion of fff approximation reduces the global error on the vf . in this part we present a method to sample the posterior distribution @xmath187 given in equation ( [ posteriortot1 ] ) . since this distribution exhibits a complex form with respect to the velocity field @xmath94 , directly drawing sample from it is impossible . nevertheless , by building an appropriate markov chain on the vf one can map the distribution ( for a complete description of markov chain monte carlo methods see @xcite ) . for this study we chose an algorithm of the metropolis - hastings type to construct the chain . the principle of the method is the following : a. in the initial step , a vf @xmath188 with @xmath189 is generated . no particular property have to be imposed on this field , but choosing a field which is as close as possible to the one maximizing the target distribution will allow the chain to converge faster . b. from @xmath188 a field @xmath190 is constructed according to some arbitrary transition kernel @xmath191 . c. the next step consists in accepting or rejecting the move from @xmath188 to @xmath190 . therefore , an acceptance probability @xmath192 is defined . in the case of the metropolis - hastings algorithm this probability is expressed as : @xmath193 d. the process returns then to step ( b ) with @xmath190 if the move from @xmath188 to @xmath190 is accepted , and with @xmath188 otherwise . the ensemble is then assumed to be representative of the posterior distribution , once its averaged field as converged towards a fixed vector . note that this algorithm has already been employed by @xcite for sampling the posterior distribution given in equation ( [ posteriortotj ] ) , but with a different parametrization of the magnetic field and the velocity field . in particular , only the uncertainties on the large scale mf were considered in these two studies . as already mentioned previously , analytically calculating the maximum of the posterior distribution @xmath187 is not feasible . we therefore decided to approximate it by taking the averaged vf of the ensemble generated by the markov chain , such as : @xmath194 to evaluate our method , we created an artificial set of data , and performed the inversion of the fff equation using these data . we also compared our results to the ones obtained with alternative approaches . the construction of the synthetic fields was performed as following : * * artificial magnetic field * + the large scale mf ( between spherical harmonic degree @xmath29 ) was taken from the grimm 2 model at the epoch @xmath195 . its spectrum ( the thick black line in the top of figure [ artificialspectra2004 ] ) was then extrapolated up to degree @xmath86 according to the formulation ( [ buffett ] ) of @xcite . from this small scale spectrum , and under the assumption of isotropy and @xmath132 mean of the field , a mf was randomly generated and added to the grimm 2 large scale field . in figure [ artificialspectra2004 ] ( top ) the spectrum of the small scale mf is represented by the plus symbols . * * artificial velocity field * + the coefficients , in spectral space , of the poloidal ( @xmath196 ) and toroidal ( @xmath197 ) fields were assumed to be isotropically distributed with a @xmath132 mean and a covariance @xmath198 derived from the following power law spectrum : @xmath199 where the value of the amplitude @xmath200 was chosen such as the averaged velocity intensity at the cmb was of @xmath201 km.yr@xmath202 . a velocity field extending up to degree @xmath203 was then randomly drawn accordingly to these statistical properties . * * artificial secular variation * + to generate an artificial large scale sv ( extending up to degree @xmath204 ) , the mf was advected by the vf through the non - linear term of the frozen flux approximation ( [ frozenflux ] ) . the energy spectrum of the resulting sv is presented with the thick line in figure [ artificialspectra2004 ] ( bottom ) . the next step of this evaluation was to recover the velocity field according to the magnetic field and the secular variation . but for this inverse problem to be more realistic , uncertainties were added to the data . a large scale lithospheric field ( @xmath29 ) was randomly generated accordingly to the theoretical spectrum @xmath205 of @xcite ( see equation [ litosphericspectrum ] ) and added to the artificial mf . this contaminated field was then truncated at degree @xmath204 . it is referred as @xmath108 and its energy spectrum is plotted on top of figure [ artificialspectra2004 ] with circles . the uncertainties on the secular variation were built by randomly drawing a field from a gaussian distribution with a @xmath132 mean and a covariance given by the posterior covariance matrix of the grimm 2 secular variation @xmath206 . the resulting field was then superimposed on the artificial field . the spectrum associated with the total sv is shown with circles on the bottom of figure [ artificialspectra2004 ] . ( thick line ) , extrapolated mf ( plus symbols ) and grimm 2 mf contaminated with a randomly generated litospheric field ( circles ) . bottom : secular variation energy spectra . artificially generated sv ( thick line ) , error field ( plus symbols ) and combination of the two fields ( circles).,title="fig : " ] ( thick line ) , extrapolated mf ( plus symbols ) and grimm 2 mf contaminated with a randomly generated litospheric field ( circles ) . bottom : secular variation energy spectra . artificially generated sv ( thick line ) , error field ( plus symbols ) and combination of the two fields ( circles).,title="fig : " ] except for the prior distribution of the velocity field @xmath98 , all the distributions derived in section [ baysian ] are consistent to describe the posterior distribution of the velocity field in this test . we recall that they read : @xmath207 where @xmath208 corresponds to the normal distribution centered in @xmath209 and with covariance @xmath210 , and @xmath211 is the covariance of the mf in which are included the uncertainties due to the lithospheric field and the modelization of the unresolved mf . for this evaluation phase , the prior distribution of the vf was given by : @xmath212 with @xmath198 the covariance of the vf derived from the power laws ( [ powerlawartificial ] ) . the estimation of the flow with respect to the artificial mf and sv was then realized with the four algorithms presented below . * * mcmc method * + in this approach the full posterior distribution of the velocity field @xmath187 is explored with the metropolis - hastings algorithm described in the beginning of the section . the transition kernel @xmath191 entering the algorithm ( step b ) is derived from the prior distribution of the vf as following : @xmath213}{(2\pi)^{\frac{d}{2 } } |\tilde{\sigma}_u|^\frac{1}{2 } } \label{transition}\ , \ ] ] where the factor @xmath214 allows to rescale the covariance @xmath198 in order to limit the distance of the move from one field to the other . since this kernel is symmetric with respect to @xmath190 and @xmath188 , the acceptance probability @xmath192 of equation ( [ acceptance ] ) reduces to : @xmath215 the initial field was set to @xmath132 and an ensemble of @xmath216 vf @xmath217 was generated . we then approximated the flow maximizing the posterior distribution through the averaging operation : @xmath218 * * least square method * + in this method , the magnetic field @xmath95 is assumed to be exactly known such as : @xmath219 as a consequence , the posterior distribution of the velocity field becomes proportional to : @xmath220 \nonumber \\ & \times & \exp\left[-\frac{1}{2 } u^t \tilde{\sigma}_u^{-1 } u \right ] \label{leastsquare}\ , \end{aligned}\ ] ] where @xmath221 is the operator allowing to evaluate the non linear term of the fff equation when it is applied to @xmath94 . the maximum of this distribution can be calculated analytically and corresponds to the usual least square solution : @xmath222 * * ensemble method * + this approach was developed by @xcite and consists in generating an ensemble of magnetic field and calculating their associated velocity field . in this test , the ensemble of mf was randomly drawn from the distribution @xmath223 given in equation ( [ priorbtest ] ) . in total we generated @xmath224 magnetic fields @xmath225 ( with @xmath226 ) extending up to degree @xmath86 . each velocity field @xmath217 were then determined by the relation : @xmath227 where the operator @xmath228 now depends on each realization @xmath229 of the magnetic field . the flow solution of the complete inverse problem is given by : @xmath230 * * iterative method * + in this approach , @xcite proposed to determine the cmb velocity field through the following iterative process : @xmath231 where @xmath232 is the index of the iteration , and the covariance @xmath233 is given by : @xmath234 with @xmath235 the operator depending on the velocity field @xmath188 , and which allows to evaluate the non linear term of the fff equation when it is applied to @xmath95 . in the numerical implementation of this method , the initial field @xmath236 was taken as the one solution of the least square approach . + note that this algorithm provides an estimation of the maximum of the posterior distribution given in equation ( [ posteriortot1 ] ) if the quantity @xmath237 is assumed to vary slowly with respect to @xmath94 . in the previous section we mentioned the algorithm of @xcite which also allows to take into account the variations of the mf in the inverse problem . we decided not to implement it in this evaluation since it is an approximation of the method proposed by @xcite . to compare the different approaches , the artificial velocity field @xmath94 is decomposed into a toroidal ( @xmath197 ) and a poloidal ( @xmath196 ) field , with @xmath156 and @xmath155 their respective spectral counterpart . the same operation is performed on the velocity fields @xmath238 given by the four inversion methods . their toroidal and poloidal parts are then referred as @xmath239 and @xmath240 in physical space , and @xmath241 and @xmath242 in spectral space . to measure the accuracy , scale by scale , of the various fields , we define the poloidal and toroidal error fields as : @xmath243 and their associated energy spectra : @xmath244 in figure [ errorspectra2004 ] the spectra of the error fields ( symbols ) are plotted together with the poloidal and toroidal spectra of the artificial velocity field ( solid lines ) . the first observation one can make is that above the sh degree @xmath245 , the energy associated with the poloidal and toroidal error is of the same order than the energy of the artificial fields . this implies that above this degree , the estimation of the flow is not reliable whatever the method employed to determine it . at large scale , the performance of the different approaches varies strongly . whereas the flow obtained with the least square method ( triangles ) is the one which deviates the most from the artificial flow , the velocity fields evaluated with the iterative method ( circles ) and the mcmc algorithm ( plus symbols ) are the ones presenting the lowest error intensities . in between , in terms of accuracy , lies the ensemble method ( crosses ) . it can be noted that at degree @xmath246 and between @xmath246 and @xmath247 for respectively the ensemble and the least square method , the energy associated with the toroidal error field is larger than the energy of the artificial field itself . this is not the case anymore when using the iterative or mcmc approaches . to evaluate the different models in physical space , the coefficients associated with the poloidal and toroidal error fields ( equations [ poloerror2004]-[toroerror2004 ] ) were truncated at degree @xmath245 , since at smallest scales the uncertainties are maximal , and projected in real space . figure [ errorfield2004 ] shows the intensity of these fields at the level of america for the four approaches . although the locations of the errors are similar , their intensities differs strongly from one flow to the other . a computation of the averaged energy associated with the poloidal and toroidal error fields ( see table [ tableerror ] ) shows that globally the best approximation of the artificial velocity field is provided by the mcmc algorithm , followed in order by the iterative , ensemble and least square methods . note that the differences between the mcmc and iterative approach are very low , and since the computation time required to sample the full posterior distribution is much larger than the one to approximate its maximum using the iterative method , if one wants to determine the flow without its underlying uncertainties , using the algorithm of @xcite is certainly more appropriate . .averaged energy of the error in km@xmath248yr@xmath249 associated with the different inverted velocity fields . these values have to be compared with the averaged energies of the artificial poloidal and toroidal fields which are respectively of @xmath250 km@xmath248yr@xmath249 and @xmath251 km@xmath248yr@xmath249 . [ cols="<,^,^,^,^",options="header " , ] , of the difference between the artificial velocity field and the velocity fields evaluated with the following approaches : mcmc ( top left ) , iterative from @xcite ( top right ) ensemble from @xcite ( bottom left ) and least square ( bottom right ) . ] since the mcmc algorithm provides an information on the flow uncertainties , we extracted the standard deviation of the velocity intensity @xmath252 , from the variance @xmath253 as following : @xmath254 p(u|\gamma ) \mathrm{d}u\ ] ] where the term @xmath255 means that only the diagonal elements of the matrix lying within the brackets are kept . at each node of the discrete cmb , the vf @xmath94 is composed of a polar and an azimuthal component , so to get the variance associated with the velocity intensity @xmath256 , the variance of each component has to be summed up . in figure [ sdevaluation ] the quantity @xmath257 , corresponding to the @xmath258 confidence interval on the flow intensity , is displayed at the level of @xmath259 . this picture presents a pessimistic view of the uncertainties to be expected when evaluating the flow maximizing the posterior distribution . because the probability for the real flow to lie within the tails of the posterior distribution is very low , the predicted error will globally be larger than the effective one . nevertheless , locations where the differences between the solution flow and the real one are important corresponds to areas of high posterior variance . confidence interval on the velocity intensity in km.yr@xmath202 according to the mcmc flow ensemble . ] finally , we wanted to measure the impact of the velocity field prior information on the solution of the inverse problem . we therefore realized a simulation with the iterative algorithm of @xcite , in which only the covariance matrix of the velocity field @xmath198 of equation ( [ iterativelesur ] ) has been modified . instead of being the covariance imposed to generate the artificial velocity field , this latter was replaced by the covariance @xmath260 defined in equation ( [ norm ] ) and derived from the bloxham s `` strong norm '' . we then calculated , as previously , the averaged energy of the error field using the velocity field we obtained . it is of @xmath261 km@xmath248yr@xmath249 and @xmath262 km@xmath248yr@xmath249 for respectively the poloidal and toroidal part of the error . this shows us that the choice of the prior information on the velocity field influences in a significant manner the results of the inverse problem . nevertheless , it can be noticed that the intensity of the error in this case remains lower than the intensity of the error for the flow resulting from the least square approach , showing the importance of modeling the possible variations of the mf . in the simulation we realized , the mf and the sv as well as the covariance for the sv , are given up to sh degree @xmath204 by the grimm 2 model of @xcite for the epoch @xmath0 . for the magnetic field , the covariance associated with its large scales ( @xmath263 ) is derived from the theoretical spectrum of the litospheric field of @xcite as shown in equations ( [ litosphericspectrum])-([lsmfcovariance ] ) , whereas the extrapolation of the mf spectrum proposed by @xcite is used to build the covariance matrix of the small scale mf ( @xmath264 ) as presented in equation ( [ ssmfcovariance ] ) . the velocity field is extended up to sh degree @xmath203 , and its prior distribution is given by the relation ( [ prioru ] ) . the width of the filter has been set to @xmath54 km , and the initial field of the markov chain @xmath238 , is the velocity field solution of the least square approach : @xmath265 with @xmath266 . as for the synthetic test of section [ mcmcevaluation ] , the transition kernel @xmath191 was derived from the prior distribution of the vf as shown in equation ( [ transition ] ) . the metropolis - hastings algorithm was then numerically simulated on a discrete cmb refined 4 times . an ensemble of @xmath267 vf @xmath94 mapping the posterior distribution have been generated . the velocity field @xmath268 maximizing the posterior , is approximated by taking the average field of the ensemble we created . and its intensity for the epoch @xmath0 . ] in figure [ vf2005 ] , the vector field @xmath268 and its intensity are displayed in different locations of the core mantle boundary . many features of the flow we obtained have already been reported in previous studies . in particular , the eccentric and planetary scale anticyclonic gyre observed by @xcite and @xcite is also present in the flow we calculated . this observation reinforces the hypothesis that the fluid motions in the outer core can be well - described by the compressible quasi - geostrophic assumption , a constraint a priori applied in both studies . nevertheless , according to our results , deviations from quasi - geostrophy ( qg ) have also to be expected . indeed , under this hypothesis , the flow is forced to be symmetric with respect to the equator outside the tangential cylinder , a condition which is not fulfilled everywhere in our case . although the symmetric part of the velocity field is dominant in our simulation ( @xmath269 of the energy of the total vf is concentrated in its symmetric components ) certain patterns , such as the flow crossing the equator below india or the larger intensity of the westward drift in the southern hemisphere are violating this property . the flow we obtained also exhibits a much smoother spatial behavior than the ones presented by @xcite . this is certainly due to the choice we made to characterize a priori the velocity field . through the bloxham s `` strong norm '' , we imposed a very steep spectrum ( in @xmath270 ) to both the poloidal and toroidal field , as a consequence the intermediate scales of the vf were probably over - damped in our simulation . for the epoch @xmath0 and its associated uncertainties on the intensity @xmath271 in km.yr@xmath202 ( left ) and orientation @xmath272 in degree ( right ) . ] possessing the full posterior distribution of the vf allows to extract many useful information on the flow . in particular the uncertainties on the velocity field intensity and orientation ( with respect of course to the prescribed modelization ) can be evaluated at any spatial location on the grid . to compute the standard deviation of the velocity field intensity @xmath271 we followed the protocol given in section [ mcmcevaluation ] , whereas the standard deviation of the velocity orientation @xmath272 , is derived from the formula : @xmath273 p(u|\gamma ) \mathrm{d}u \ , \ ] ] where @xmath274 is the angle , in degree , between the velocity field @xmath94 and @xmath268 . the quantities @xmath271 and @xmath272 , are displayed in figure [ uncertainties2005 ] ( on the left for the @xmath271 and on the right for the @xmath272 ) . this figure first shows that a strong ( resp . weak ) uncertainty on the intensity coincides with a strong ( resp . weak ) uncertainty on the orientation of the flow . secondly one can notice that the uncertainties are not homogeneously distributed on the surface of the cmb . while the planetary scale eccentric gyre seems to be very robust , the vf at the level of the pacific ocean is much more uncertain . it is known that in this latter area the magnetic field activity is moderate ( @xcite ) . as a consequence the secular variation is low , and its associated uncertainties , due to the inaccuracy of the measurements and to the interactions between the unresolved mf and the vf , may be larger than the signal itself . it is therefore very difficult to evaluate the velocity field in this region . nevertheless , this maps shows that there is a large part east of australia and around the longitude of new - zealand where the vf can be accurately estimated . another particular feature can be noticed at the level of the north - atlantic ocean , where the uncertainties on both the flow intensity and orientation are very large . as already questioned by @xcite , the robustness of the clockwise gyre usually observed in this area by models assuming tangential - geostrophy seems to be very weak according to our results . ( thick line ) , and with the spectral uncertainties ( circles).,title="fig : " ] ( thick line ) , and with the spectral uncertainties ( circles).,title="fig : " ] since we are using truncated mf and sv it may be interesting to investigate the spectral properties of the posterior vf . the poloidal @xmath196 and toroidal @xmath197 part of the vf @xmath94 are expanded in spherical harmonics . the resulting fields are respectively referred as @xmath155 and @xmath156 . the same operation is performed on @xmath275 , with @xmath276 and @xmath277 its poloidal and toroidal sh coefficients . in figure [ spectraluncertainty ] are plotted the spectra associated with @xmath276 and @xmath277 as well as the variance on these coefficient summed up over the order @xmath14 and rescaled by the factor @xmath278 . we can observe that the largest scales of the flow are dominated by the toroidal field . a computation of the poloidal and toroidal energy shows that more than @xmath279 of the total energy is of toroidal nature . the other information which can be extracted from figure [ spectraluncertainty ] is that , as for the synthetic test we realized previously , above the sh degree @xmath245 , the intensity of the flow uncertainties becomes larger than the intensity of the flow itself . as a consequence the evaluation of the small scale velocity field can not be considered as reliable . it has to be emphasized that all the results we obtained are conditioned by the choice of the prior information imposed to the flow , and should not be considered as absolute . in particular the globally low level of uncertainties on the flow solution is certainly a consequence of the strong regularization we employed . in this study we have presented a new method to determine the velocity field at the earth s core mantle boundary according to an outer core magnetic field and secular variation model . we showed that using an appropriate dynamical equation to prescribe the large scale magnetic field evolution in the inverse problem , permitted to reduce the modeling errors arising from the truncated nature of the available fields . we also demonstrated that the bayesian formalism we developed to account for the large scale uncertainties on the magnetic field and to model the unresolved small scale mf , allowed to properly describe the inverse problem as soon as the information introduced a priori were accurate . through the evaluation of our method and the comparison with other approaches , we could indirectly confirm that the unresolved part of the magnetic field contributed significantly to the observed secular variation , and that its modelization was necessary to obtain a more accurate description of the flow at the core mantle boundary . when we applied our method to real data , provided by the grimm 2 model for the epoch @xmath0 , we could recover many features of the flow already observed in previous studies where a different prior information on the velocity field had been considered . in particular we could retrieve the planetary - scale eccentric gyre characteristic of flow evaluated under the compressible quasi - geostrophy assumption ( see @xcite ) . nevertheless , according to our simulation , the flow is crossing the equator below india and the intensity of the westward drift is the larger in the southern hemisphere , indicating that the equatorial symmetry imposed by the quasi - geostrophy hypothesis is broken . through this observation one can conclude that deviations from quasi - geostrophy should be allowed to occur when this latter constraint is imposed in the inverse problem . another specificity of the velocity field we obtained is its very smooth spatial behavior . this property is certainly induced by the prior distribution of the velocity field we chose , since this latter imposes a very steep spectrum to both the poloidal and toroidal part of the velocity field . finally , thanks to the ensemble of velocity field we generated to map the posterior distribution , we could evaluate the uncertainties associated with the flow solution of the inverse problem . according to our results , whereas on the one hand , the robustness of the flow is questioned in many area , and particularly in almost the entire pacific ocean , and in the northern part of the atlantic ocean , on the other hand , the planetary - scale eccentric gyre seems to be a very robust structure . from the evaluation of the uncertainties in spectral space , we concluded that the flow at the cmb could only be accurately estimated at large scales ( between spherical harmonics degree @xmath280 and @xmath281 ) . in cartesian space , applying a gaussian filter to a scalar or a vector field , or letting the field evolve through a diffusion process can be interpreted as being similar operations . indeed , the kernel of the gaussian filter is : @xmath282 where @xmath283 corresponds to the spatial dimension , @xmath50 is the filter width , and @xmath167 is a constant usually set to @xmath284 . the solution of the diffusion equation : @xmath285 where @xmath286 is a diffusion coefficient and @xmath287 the time , is the convolution between the scalar field @xmath288 , and the green function : @xmath289 so if one sets @xmath290 , diffusing a scalar field through equation ( [ diffusion ] ) is equivalent to filtering that field with the convolution kernel expressed in ( [ cartesiankernel ] ) . based on this observation , @xcite derived a gaussian - like filter on the surface of a sphere of radius @xmath16 , by determining the convolution kernel of the spherical diffusion equation : @xmath291 which reads : @xmath292 where @xmath293 is the spherical harmonic of degree @xmath13 and order @xmath48 . so in spectral space , the filtering operation of a scalar field @xmath43 expanded in spherical harmonics : @xmath294 simply reduces to the operation : @xmath295 the grid describing the core mantle boundary is obtained by recursively subdividing an initial icosahedron as explained in @xcite . for each grid refinement procedure , a node is added in the middle of the geodesic arc linking every two neighboring points . the refinement degree @xmath296 of the grid corresponds to the number of time this procedure has been applied . therefore @xmath297 corresponds to the icosahedron itself which possess @xmath298 nodes , and @xmath299 spherical triangle cells . as the refinement degree increases , the number of grid points and cells increases as : @xmath300 to approximate differential operators , a voronoi - based finite volume method is chosen . this approach has been widely used in advection - diffusion problems such as in @xcite and has proven to be efficient to tackle these kind of problems . since in finite volume methods , differential operators are converted to surface integrals , control volumes ( or voronoi cells ) surrounding each grid points have to be defined . of the different spherical triangles @xmath302 and @xmath303 denote respectively the unit vectors normal and tangential to the voronoi cell contour . b ) spherical triangle formed by the three nodes @xmath304 , @xmath305 and @xmath306 . @xmath307 corresponds to the gravity center of the triangle , and @xmath308 , @xmath309 , and @xmath310 are the area of the three sub - triangles . ] as shown in figure [ operators ] a ) , each grid point is surrounded by @xmath284 ( or @xmath311 when the point corresponds to a generator of the initial icosahedron ) nodes in its direct neighborhood . from this cluster of node , one can build an ensemble of spherical triangle cells @xmath312 all connected together by the common vertex @xmath304 . taking the gravity center @xmath307 of each cell allow then to draw a hexagonal ( or pentagonal ) volume control around the central node @xmath304 . @xcite have shown that moving the grid points in a manner that they coincide with the gravity center of their control volume , allows to improve the accuracy of the differential operators to the second - order . the procedure employed here is the constrained centroidal voronoi tessellations(ccvt ) developed by @xcite . the principle of this procedure as we implemented it , known as the lloyd s method , is the following : 1 . starting with an initial distribution of nodes on the sphere , taken here as the different subdivision of the icosahedron , as explained previously in this part . 2 . building the voronoi cells associated with each grid point . moving each node to the gravity center of the cell it is belonging to . 4 . returning to step 2 until some convergence criteria is reached . in our case , we imposed that the averaged geodesic distance between the node and the gravity center of the voronoi cell has to be smaller than @xmath313 . the horizontal divergence and curl applied to a vector field @xmath314 , and the gradient applied to a scalar field @xmath196 can be expressed in their integral form as : @xmath315 where the control surface as an area @xmath103 , delimited by the contour @xmath316 , and the unit tangential and normal vector to the contour are respectively @xmath317 and @xmath318 . on the discrete sphere , vector and scalar fields are known on the nodes , therefore , to differentiate them , one need to approximate them on the contour of the voronoi cells . first the different quantities are evaluated on the corners of the cells : @xmath319 where the areas @xmath308 , @xmath309 , @xmath310 are shown in figure [ operators ] b ) . then , following the notation of figure [ operators ] a ) the discrete approximation of the different differential operators becomes : @xmath320 where @xmath321 corresponds to the area of the control volume , @xmath322 is the geodesic length between the points @xmath323 and @xmath324 , and @xmath325 is the number of vertices of the control volume . note that when the subscript @xmath326 then @xmath327 .

1.2 potential based no - regret dynamics are shown to be related to fictitious play . roughly , these are-best reply dynamics where is the maximal regret , which vanishes with time . this allows for alternative and sometimes much shorter proofs of known results on convergence of no - regret dynamics to the set of nash equilibria . _ keywords : _ regret minimization , no - regret strategy , fictitious play , best reply dynamics , nash equilibrium , hannan set , curb set _ jel classification numbers : _ c73 , d81 ,

1207.0660

_ no - regret strategies _ are simple adaptive learning rules that recently received a lot of attention in the literature . in a repeated game , a player has a _ regret _ for an action if , loosely speaking , she could have obtained a greater average payoff had she played that action more often in the past . in the course of the game , the player reinforces actions that she regrets not having played enough , for instance , by choosing next action with probability proportional to the regret for that action , as in hart and mas - colell s @xcite _ regret matching _ rule . existence of _ no - regret strategies _ ( i.e. , strategies that guarantee no regrets almost surely in the long run ) is known since @xcite ; wide classes of no - regret strategies are identified by @xcite and @xcite . a _ no - regret dynamics _ is a stochastic process that describes trajectories of the average correlated play of players and that emerges when every player follows a no - regret strategy ( different players may play different strategies ) . by definition , it converges to the hannan set ( the set of all correlated actions that satisfy the no - regret condition first stated by @xcite).the hannan set of a game is also known as the set of _ weak correlated equilibria _ @xcite or _ coarse correlated equilibria _ * ch.3 ) . ] this set is typically large . it contains the set of correlated equilibria of the game and we show that it may even contain correlated actions that put positive weight _ only _ on strictly dominated actions . thus convergence of the average play to the hannan set often provides very little information about what the players will actually play , as it does not even imply exclusion of strictly dominated actions . in this paper we show that no - regret dynamics are intimately linked to the classical fictitious play process @xcite . drawing on @xcite , we first show that contrary to the standard , discrete - time version , continuous fictitious play leads to no regret . we then show that , for a large class of no - regret dynamics , if a player s maximal regret is @xmath1 , then she plays an @xmath0-best reply to the average correlated play of the others . since in this class the maximal regret vanishes ( see corollary [ cor:1 ] below ) , it follows that , for a good choice of behavior when all regrets are negative , the dynamics is a vanishingly perturbed version of fictitious play . for two - player finite games , this observation and the theory of perturbed differential inclusions @xcite allow us to relate formally the asymptotic behavior of no - regret dynamics and of continuous fictitious play ( or its time - rescaled version , the best - reply dynamics @xcite ) . in classes of games in which the behavior of continuous fictitious play is well known , this provides substantial information on the asymptotic behavior of no - regret dynamics . in particular , we recover most known convergence properties of no - regret dynamics . our results do not just allow us to find new and sometimes much shorter proofs of convergence of no - regret dynamics towards the set of nash equilibria in some classes of games , such as dominance solvable game or potential games . they also allow us to relate the asymptotic behavior of no - regret dynamics and continuous fictitious play in case of divergence , as in the famous shapley game @xcite . these results extend only partially to @xmath2-player games ( though they fully extend to @xmath2-player games with linear incentives @xcite ) . the issue is that in @xmath2-player games no - regret dynamics turn out to be related to the correlated version of continuous fictitious play , in which the players play a best - reply to the _ correlated _ past play of the others . this version of fictitious play is defined through a correspondence which is not convex valued . this creates technical difficulties , because the theory of perturbed differential inclusions is not developed for non convex valued correspondences . a different way to analyze no - regret dynamics is to show that some sets attract nearby solution trajectories . we show that strict nash equilibria and , more generally , the intersection of the hannan set and the sets that are _ _ closed under rational behavior ( curb ) _ _ are attracting for no - regret dynamics , in a sense to be defined in section [ sec : curb ] . the remainder of the note is organized as follows . the next section introduces no - regret dynamics . section [ sec : main ] studies the links between no - regret dynamics and fictitious play . section [ sec : curb ] shows that the intersection of the hannan set and curb sets is attracting for no - regret dynamics . section [ sec : cont ] studies the continuous - time version and the expected version of no - regret dynamics . finally , the appendix contains the proofs of the main results , as well as counterexamples illustrating the complexity of the relationship between ict and limit sets . consider a bimatrix game @xmath3 , where @xmath4 is the set of actions of player @xmath5 and @xmath6 is her payoff function , with @xmath7 . for any finite set @xmath8 , denote by @xmath9 the set of probability distributions over @xmath8 . a mixed action of player @xmath5 is an element of @xmath10 . a correlated action @xmath11 is a probability distribution over the set of pure action profiles , i.e. , @xmath12 given such a @xmath11 , let @xmath13 and @xmath14 denote its marginals for player @xmath5 and her opponent , respectively . thus , @xmath15 . throughout , @xmath16 refers to @xmath5 s opponent . as usual , let @xmath17 and @xmath18 for @xmath19 . depending on the context , @xmath20 may refer to a pure action an element of @xmath21 or to a vertex of @xmath10 , i.e. , a dirac measure on a pure action . the game is played repeatedly in discrete time periods @xmath22 . in every period @xmath23 each player @xmath5 chooses an action @xmath24 and receives payoff @xmath25 where @xmath26 . denote by @xmath27 the history of play up to @xmath23 , and let @xmath28 be the set of all finite histories ( including the empty history ) . a strategy of player @xmath5 is a function @xmath29 that stipulates to play in every period @xmath30 a mixed action @xmath31 as a function of the history before @xmath23 . the weight that this mixed action puts on action @xmath19 is denoted by @xmath32 the _ average correlated play _ up to period @xmath23 is @xmath33 , where we identify @xmath34 with the corresponding vertex of @xmath35 . since @xmath36 $ ] , it follows that for all @xmath37 , @xmath38 for a correlated action @xmath11 , the _ regret _ of player @xmath5 for action @xmath39 is defined as @xmath40 , and her maximal regret as @xmath41 . typically we deal with the regret based on the average correlated play , @xmath42 , up to some period @xmath23 . in this case the regret of player @xmath5 for action @xmath19 is equal to the difference between the average payoff she would have obtained by always playing @xmath39 ( assuming that her opponent s play remains the same ) and her average realized payoff : @xmath43.\ ] ] to simplify notations , we will often write @xmath44 for @xmath45 and @xmath46 for @xmath47 . player @xmath5 has no asymptotic regret if her average realized payoff is asymptotically no less than her best - reply payoff against the empirical distribution of her opponent : @xmath48 a strategy of player @xmath5 is a _ no - regret strategy _ if for any strategy of the other player , inequality ( [ eq : obj ] ) holds almost surely . this property is also called _ hannan consistency _ @xcite or _ universal consistency _ @xcite . it is well known in the literature since @xcite that there exist simple no - regret strategies . @xcite describe a wide class of _ potential based _ no - regret strategies . a twice differentiable , convex function @xmath49 is called a _ potential _ if it satisfies the following conditions : ( r1 ) : : @xmath50 , and @xmath51 for all @xmath52 ; ( r2 ) : : @xmath53 , and @xmath54 for all @xmath55 ; ( r3 ) : : if @xmath55 and @xmath56 , then @xmath57 , where @xmath58 denotes the partial derivative with respect to @xmath59 . the potential @xmath60 can be viewed as a generalized distance function between a vector @xmath61 and the nonpositive orthant @xmath62 . let @xmath63 denote player @xmath5 s regret vector . [ thm : hm ] let @xmath60 satisfy ( r1)(r3 ) and let strategy @xmath64 satisfy @xmath65 whenever @xmath66 . then @xmath64 is a no - regret strategy . this holds by theorem 3.3 of @xcite , whose conditions ( r1 ) and ( r2 ) are satisfied by our conditions ( r1) and ( r2 ) , respectively , and whose proof is based on the blackwell s approachability theorem @xcite . a standard example of no - regret strategy satisfying the above conditions is obtained by letting @xmath67 be the @xmath68-norm on @xmath69 , i.e. @xmath70_{+}^{\mathbf{p}})^{1/\mathbf{p}}$ ] with @xmath71 , where @xmath72_{+}=\max(0 , x_k)$ ] . the resulting strategy @xmath64 is called the @xmath68__-norm _ _ strategy @xcite . it is defined by @xmath73_{+}^{\mathbf{p}-1}}{\sum_{s\in a_{i}}[r_{i , s}(t)]_{+}^{\mathbf{p}-1 } } , \text { \ } \forall k\in a_{i},\ ] ] whenever @xmath66 . the @xmath74-norm strategy is the _ regret - matching strategy _ @xcite , that stipulates to play an action in the next period with probability proportional to the regret for that action . for large @xmath75 , the @xmath68-norm strategies approximate fictitious play . we say that the average correlated play @xmath42 follows a _ no - regret dynamics _ if both players use ( possibly different ) no - regret strategies . a trajectory @xmath76 of a no - regret dynamics is thus a solution of ( [ eq : dt ] ) where @xmath77 is a realization of @xmath78 and @xmath79 are no - regret strategies . we focus on the class @xmath80 of no - regret dynamics such that : \(i ) the no - regret strategies @xmath81 , @xmath82 of the players are potential - based : they satisfy ( q1 ) for some potentials @xmath83 , @xmath84 satisfying ( r1)-(r3 ) ; \(ii ) if a player has no - regret then he takes some constant pure action : for each @xmath85 , there exists @xmath86 such that @xmath87 our results are valid for a somewhat wider class of no - regret dynamics . what we really need , beside a no - regret dynamics , is that from some period @xmath88 on : ( i@xmath89 ) if a player has positive regret for some actions , then she plays one of these actions . ( ii@xmath89 ) if a player never has any positive regret , then she plays an @xmath90-best reply to the empirical distribution of her opponent , where @xmath91 almost surely . [ rem : q1 ] property ( i@xmath89 ) follows from ( r3 ) and ( [ eq : pb ] ) . this is a _ better reply property _ that stipulates to assign a positive probability only on better reply actions to the opponent s empirical distribution of play ( better " with respect to the realized payoff ) . also it implies that if @xmath92 in some period @xmath23 , then @xmath93 for all @xmath94 . indeed , when an action @xmath39 with positive regret is played , the sign of @xmath44 does not change , hence the maximal regret remains positive ( * ? ? ? * proposition 4.3 ) . [ rem : q2 ] assumption ( q2 ) is a simple way of ensuring ( ii@xmath89 ) , and in addition , that if @xmath95 for all @xmath23 , then @xmath96 as @xmath97 . below , but for our main results ( ii@xmath89 ) suffices . ] indeed , if @xmath95 for all @xmath98 then by ( q2 ) , for all @xmath98 , @xmath99 , hence @xmath100 . it follows that @xmath96 and that for all @xmath101 , player @xmath5 plays an @xmath90-best reply with @xmath102 . for a discussion of other possible assumptions , see @xcite , appendix a. note that there are no - regret dynamics that do not satisfy ( i@xmath89 ) . for instance , stochastic fictitious play with a noise parameter that declines with time at an appropriate rate ( see , e.g. , @xcite ) . this process is not potential based in our sense due to the time inhomogeneity , but this is not the crucial point , since ( i@xmath89)-(ii@xmath89 ) would suffice . define the _ hannan set _ @xmath103 of the stage game @xmath104 as the set of all correlated actions of the players where each player has no regret : @xmath105 the _ reduced hannan set _ @xmath106 is the subset of @xmath103 in which at least one regret is exactly zero for each player : @xmath107 the next property of no - regret dynamics is straightforward by the definition of no - regret strategies and remark [ rem : q2 ] ( see , e.g. , ( * ? ? ? * corollary 3.2 ) ) . [ cor:1 ] for every no - regret dynamics in class @xmath80 , the trajectories converge almost surely to the reduced hannan set . convergence of the average play @xmath42 to set @xmath106 does not imply its convergence to any particular point in @xmath106 . moreover , even if @xmath42 converges to a point , this point need not be a nash equilibrium . in _ discrete fictitious play _ , in every period @xmath23 after the initial one , player @xmath5 plays a pure best reply @xmath108 to the average past play of her opponent @xmath109 ( here @xmath110 is a vertex of @xmath111 ) . the latter is called the _ belief _ of player @xmath5 on her opponent s next move . formally , for any @xmath112 in @xmath113 , denote by @xmath114 player @xmath5 s set of best replies to @xmath115 : @xmath116 let @xmath117 . a discrete - time trajectory @xmath118 on @xmath119 is a solution of _ discrete fictitious play _ ( dfp ) if for every @xmath120 @xmath121 where @xmath122 and @xmath123 is a vertex of @xmath10 associated with some pure best reply action , @xmath85 . analogously , an absolutely continuous function @xmath124 is a solution of _ continuous fictitious play _ ( cfp ) if for almost all @xmath125 , @xmath126 is differentiable and @xmath127 where @xmath128 is now a profile of _ mixed _ actions . this may be written as the differential inclusion : @xmath129 the average correlated play satisfies @xmath130 for some initial condition @xmath131 such that @xmath132 , @xmath85 . thus , for almost all @xmath23 , @xmath42 is differentiable and @xmath133 where @xmath134 is the product distribution corresponding to the mixed strategy profile @xmath135 , and @xmath136 is a best - reply to @xmath137.this definition of cfp guarantees that solutions exist in all games and for all initial conditions , and that by the change of time scale @xmath138 , cfp corresponds to the _ best - reply dynamics _ @xcite defined by @xmath139 . another definition of cfp ( e.g. , @xcite and @xcite ) consider only trajectories that are piecewise linear , such that @xmath140 is always a pure action ( technically , a vertex of @xmath10 ) , and that the times at which @xmath141 changes have no finite accumulation point . this restricted definition is easier to handle , but in many games there do not exist such trajectories from every initial condition . ] in discrete or continuous fictitious play , the marginals @xmath142 , @xmath143 of the average past play are equal to the beliefs @xmath144 , @xmath145 . by analogy , if @xmath42 is the average past play generated by a no - regret dynamics , it is convenient to call @xmath146 the belief of player @xmath5 about her opponent s next move . this illuminates a crucial difference between fictitious play and no - regret dynamics in class @xmath80 : under fictitious play , a player chooses a _ best reply _ to her belief , whereas under no - regret dynamics , she chooses a _ better reply _ ( better " with respect to her average realized payoff ) . it is well known that discrete fictitious play does not lead to _ no regret _ @xcite . consider the following example : @xmath147 because @xmath148 is irrational , @xmath149 and @xmath150 can not both be best - replies to the empirical past play of the other player . thus , any dfp process is entirely determined by its first move . assume that the first move is off the diagonal , say @xmath151 . due to the symmetry of the game and the absence of ties , both players always switch to another action simultaneously . therefore the play is locked off the diagonal and the maximal regret is at least @xmath152 at any stage . this holds in the mixed extension of the game , since at any stage the players have a unique , pure best reply . since the continuous fictitious play process is a continuous - time version of dfp , intuitively , it should not lead to _ no regret _ either . the following result a generalization of theorem d of @xcite shows that this intuition is misleading . [ p : cfp ] under any solution of continuous fictitious play , the average correlated play converges to the reduced hannan set . this discrepancy between dfp and cfp may be explained as follows . playing an action with positive regret decreases the regret for this action . in cfp , roughly , when an action is played it remains a best reply , hence it is associated with maximal regret for a small time increment . precisely , the derivative of the regret for the action played is equal to the derivative of the maximal regret . since the regret for this action decreases , so does the maximal regret . in contrast , in dfp , an action played at stage @xmath23 has maximal regret at stage @xmath23 , but not necessarily at stage @xmath153 . thus the fact that the regret for this action decreases does not entail that the maximal regret does . for comparison with ( * theorem 3.1 ) , rescale time ( let @xmath154 ) so that ( [ eq : cont ] ) becomes @xmath155 . for any mixed action @xmath156 , let @xmath157 let @xmath158 . note that @xmath159 is lipschitz continuous for all @xmath39 in @xmath21 . thus it follows from theorem a.4 of @xcite that , for almost all @xmath23 , @xmath160 and @xmath159 are differentiable , and for all @xmath39 such that @xmath161 , we have @xmath162 . it follows that @xmath163 . furthermore : @xmath164 = -r_{i , q_i}= - v_i.\label{eq : a-1b}\ ] ] thus , @xmath165 . therefore , @xmath166 converges to zero for all @xmath85 , hence @xmath167 . [ rem : uni ] in the proof , we did not use that @xmath168 is a best - reply to @xmath169 . this shows that the fact that cfp leads to no - regret is a unilateral property . that is , if a player s behavior evolves according to cfp , then she has no asymptotic regret , independently of her opponent s behavior ( see also @xcite ) . cfp and the best - reply dynamics converge to the set of nash equilibria in finite zero - sum games @xcite . the usual proof is to show that the duality gap " @xmath170 converges to zero . this follows from the above proof , since in a two - player zero sum game @xmath171 , where @xmath172 is a solution of cfp and @xmath11 the associated correlated play . in the previous subsection we showed that cfp leads to no regret conversely , we now show that any no - regret dynamics in class @xmath173 ( as defined in section [ sec : prelim ] ) is closely related to cfp . we first explain the intuition . denote by @xmath174 the set of @xmath0-best replies of player @xmath5 to the mixed action @xmath115 of her opponent : @xmath175 the crucial observation is the following . [ lm0 ] assume that the maximal regret is less than @xmath0 . then any action with positive regret is an @xmath0-best reply to the average play of the opponent . if player @xmath5 has positive regret for action @xmath20 at some @xmath12 , then @xmath176 . but by assumption @xmath177 . therefore , @xmath178 @xmath179 , and @xmath20 is an @xmath0-best reply to @xmath137 . since no - regret dynamics in class @xmath80 only pick actions with positive regret , they only pick @xmath0-best replies to the average play of the others , where @xmath0 is the maximal regret . since this maximal regret approaches zero almost surely , eventually only almost - exact best replies are picked . this provides the intuition why no - regret dynamics and fictitious play may exhibit similar asymptotic behavior . finding a precise link , however , is not obvious . for instance , there could exist actions that are @xmath180-best replies in each period @xmath23 , with @xmath181 , but never exact best replies . thus a limit play of no - regret dynamics may include such actions , but this can not happen under fictitious play . @xmath182 consider the example shown on fig . [ f : emp ] . let @xmath183 . it is easy to verify that action @xmath184 is player 1 s best reply to player 2 s mixed action @xmath185 if and only if @xmath186 . let us first consider dfp . since @xmath187 is an irrational number , after every finite history of play , @xmath188 ; consequently dfp never picks @xmath184 ( except , possibly , at the initial period ) . would not help since @xmath184 is a best - reply only when @xmath189 , which happens at most once . ] however , it may be shown that under any dfp trajectory , the average play @xmath145 of player @xmath190 converges to @xmath191 , to which @xmath184 is a best - reply . it follows that @xmath184 is an @xmath180-best reply to @xmath145 for some sequence @xmath192 . thus a no - regret dynamics with the same trajectory of the marginal play of player 2 _ might _ choose action @xmath184 a positive fraction of time in the long run . this example does not apply to cfp , as in this case @xmath145 need not be a rational number ; and as we show below , the asymptotic behavior of no - regret dynamics and cfp can be formally related using the theory of perturbed differential inclusions @xcite . before stating a precise result , we need some definitions . a set @xmath193 is _ invariant _ under cfp if for every initial point @xmath194 there exists a solution @xmath195 of cfp , defined for all @xmath196 ( not only @xmath125 ) and such that @xmath197 and @xmath198 for all @xmath199 . a nonempty compact invariant set is an _ attractor _ if it attracts uniformly all trajectories starting in its neighborhood . an invariant set @xmath149 is _ attractor - free _ if no proper subset of @xmath149 is an attractor for the dynamics restricted to @xmath149 . a nonempty compact set @xmath149 is _ internally chain transitive _ ( ict ) for continuous fictitious play if every pair of points in @xmath149 can be connected by finitely many arbitrarily long pieces of orbits of cfp lying completely within @xmath149 with arbitrarily small jumps between them . ) . this explains that their definition considers solutions defined for all @xmath200 while ours considers solutions defined for all @xmath199 . ] every ict set is invariant and attractor free ( * ? ? ? * property 2 ) . limit set of the beliefs _ of a trajectory @xmath42 on @xmath201 is the set of all accumulation points of its marginals @xmath202 as @xmath203 . [ thm : main ] for every no - regret dynamics in class @xmath173 , the limit set of the beliefs is almost surely internally chain transitive for continuous fictitious play . , cfp can be replaced by the best - reply dynamics since they clearly have the same ict sets ( see footnote [ ft : brd ] ) . ] we give here a sketch of the proof . the details are given in appendix a.1 . a discrete - time trajectory @xmath204 on @xmath119 is a _ payoff perturbed _ dfp trajectory if there exists a positive sequence @xmath205 converging to zero such that ( [ eq : fp ] ) holds and @xmath206 is a vertex of @xmath10 associated with a pure @xmath180-best reply to @xmath207 , for all @xmath85 and all @xmath120 . a no - regret dynamics in class @xmath80 generates a trajectory @xmath208 on @xmath35 and an associated _ sequence of beliefs _ @xmath209 on @xmath113 . building on lemma [ lm0 ] , we show that this sequence of beliefs is almost surely a payoff perturbed dfp trajectory . by an auxiliary lemma , this implies that this is almost surely a _ graph - perturbed _ dfp trajectory : a notion similar to payoff - perturbed trajectory , but for another definition of perturbed best - reply , the one used in the theory of perturbed differential inclusions @xcite . it follows that the continuous - time interpolation of this sequence of beliefs is almost surely a perturbed solution of cfp , in the sense of @xcite . theorem [ thm : main ] then follows from theorem 3.6 of @xcite . since ict sets are invariant , a consequence of theorem [ thm : main ] is the following : [ cor : main ] let @xmath210 be the global attractor of cfp ( i.e. , its maximal invariant set , see @xcite ) . for any no - regret dynamics in class @xmath173 , the limit set of the beliefs is almost surely a subset of @xmath210 . note the similarity with propositions 5.1 and 5.2 of @xcite , who study the links between the time - average of the replicator dynamics and cfp . theorem [ thm : main ] allows for alternative and sometimes much shorter proofs of most known convergence properties of no - regret dynamics . below , we write that no - regret dynamics converge to some set @xmath211 if the limit set of the beliefs is almost surely a subset of @xmath211 . ( points ( a ) , ( b ) and ( c ) below ) lead to the same conclusions about no - regret dynamics as those about the time average of the replicator dynamics described in ( * ? ? ? * , points ( 2 ) , ( 3 ) and ( 4 ) ) . ] * ( a ) * for any game which is best - reply equivalent to a two - person zero sum game , the global attractor of cfp is the set of nash equilibria @xcite . hence all no - regret dynamics in class @xmath173 converge to the set of nash equilibria . actually , in zero - sum games , if the correlated action @xmath11 is in the hannan set ( recall that this is the set of correlated actions that satisfy _ no - regret _ for all players ) , then @xmath212 is a nash equilibrium . consequently , in zero - sum games all dynamics that lead to no regret ( not only those in class @xmath173 ) converge to the set of nash equilibria . this holds more generally for _ stable bimatrix games _ @xcite , because these are rescaled zero - sum games in the sense of @xcite , as is easily shown and was known to josef hofbauer ( private communication ) . * ( b ) * for games with strictly dominated strategies , the global attractor of cfp is contained in the face of the simplex with no weight on these strategies . hence all no - regret dynamics in class @xmath173 converge to this face . similarly , these dynamics converge to the unique nash equilibrium in strictly dominance solvable games . [ cols="^,^ " , ] contrary to ( a ) , this need not be true for all dynamics that lead to no regret . indeed , convergence to the hannan set or even to the reduced hannan set does not guarantee elimination of strictly dominated strategies . consider , for instance , the games shown on fig . [ f : h ] . both games are symmetric , so we indicate only the payoffs of the row player . game ( i ) is an identical interest game which is strictly dominance solvable ; yet the correlated action putting probabilities @xmath213 on each diagonal square is in the reduced hannan set . for @xmath214 , game ( ii ) is a coordination game with duplicate strategies . for @xmath1 , the duplicates @xmath215 , @xmath216 are penalized and become strictly dominated . thus , the correlated action putting probability @xmath217 on @xmath218 and @xmath217 on @xmath219 puts only weight on strictly dominated actions . yet , for @xmath220 , it belongs to the hannan set . * ( c ) * in weighted potential games , all internally chain transitive sets of cfp are ( subsets of ) connected components of nash equilibria on which the payoffs are constant ( see * ? ? ? * theorem 5.5 and remark 5.6 ) . hence by theorem [ thm : main ] , all no - regret dynamics in class @xmath173 converge to such components . note that the original proof is much longer ( * ? ? ? * appendix a ) . * ( d ) * if the beliefs @xmath221 of a no - regret dynamics converge to the set of nash equilibria , then the average realized payoff converges to the set of nash equilibrium payoffs . to see why this is true , let @xmath222 be a limit point of @xmath223 and let the marginals @xmath224 constitute a nash equilibrium . by corollary [ cor:1 ] the maximal regret converges to zero , so for every @xmath85 @xmath225 this result illuminates an important difference between no - regret dynamics and discrete fictitious play . it is well known that under dfp , if the beliefs of the players converge to a nash equilibrium , their average realized payoffs _ need not _ approach the set of nash equilibrium payoffs , whereas under no - regret dynamics it is always the case . @xmath226 * ( e ) * consider the @xmath227 game of fig . [ f : s ] due to @xcite , the historical counterexample to the convergence of fictitious play . this game has a unique equilibrium , in which both players randomize uniformly . though this equilibrium attracts some solutions of continuous fictitious play ( e.g. all those that start and remain symmetric ) , almost all solutions converge to a hexagon , the so - called shapley polygon @xcite . it may be shown that the only ict sets are the nash equilibrium and the shapley polygon . consequently , the limit set of any no - regret dynamics in class @xmath173 is almost surely one of these two sets . * ( f ) * in a number of classes of games , convergence of discrete fictitious play to the set of nash equilibria has been established , but analogous results for continuous fictitious play are lacking . thus we can not use theorem [ thm : main ] . these classes of games include generic @xmath228 games @xcite , generic ordinal potential games , quasi - supermodular games with diminishing returns @xcite , and some other special classes ( see , e.g. , @xcite ) . for ordinal potential games and quasi - supermodular games with diminishing returns , @xcite proves convergence to the set of nash equilibria of _ some _ solutions of continuous fictitious play as defined by ( [ eq : cfp2 ] ) ( see our footnote [ ft : brd ] ) . this is not enough to apply the results of @xcite . the same problem arises in @xcite . actually , as explained below , convergence of cfp to the set of nash equilibria would not suffice to use theorem [ thm : main ] : we would need some additional structure , such as a lyapunov function , to get more information on the ict sets . * ( g ) * consider a bimatrix game in which all solutions of cfp converge to the set of nash equilibria . because the definition of attractor requires uniform attraction , this does not imply that the set of nash equilibria is an attractor . neither does it imply that all ict sets are contained in the set of nash equilibria , as shown in appendix a.2 . therefore , we can not deduce from theorem [ thm : main ] that no - regret dynamics in class @xmath80 converge to the set of nash equilibria ; whether this is always the case remains an open question . * ( h ) * we show in section [ sec : cont ] that theorem [ thm : main ] also applies , and under weaker assumptions , to the continuous - time version and to the expected version of no - regret dynamics in class @xmath80 . as apparent from the proof , the existence of a potential is not essential : for a good choice of behavior when there are no regrets , theorem [ thm : main ] holds for any no - regret dynamics such that a player always chooses an action with positive regret whenever he has one . it also applies to certain no - regret dynamics that do not have this property , such as the exponential weight algorithm ( see remark [ rem : ew ] at the end of appendix a.1 ) . * ( i ) * let us now comment on extensions of our results to @xmath2-player games . the definition of no - regret dynamics , as well as proposition [ thm : hm ] , extend to the @xmath2-player setting straightforwardly ( e.g. , @xcite ) . the appropriate extension of cfp is _ correlated cfp _ where at each time @xmath23 every player chooses a best reply action to the _ correlated _ past average play of the others . specifically , an absolutely continuous function @xmath229 is a solution of correlated cfp if it is almost everywhere differentiable and satisfies @xmath230 where the correlated best - reply correspondence @xmath231 is defined by @xmath232 where @xmath233 is the set of mixed best replies of player @xmath5 to the correlated action @xmath234 . in @xmath2-player games with linear incentives @xcite , also known as polymatrix games @xcite , the correlated and independent best - reply correspondences coincide ; that is , for any correlated action @xmath12 , @xmath235 where @xmath236 is the vector of marginals of @xmath11 , and @xmath237 the standard ( independent ) best - reply correspondence . for such games , theorem [ thm : main ] extends easily . however , this is not the case in general . the main problem is that the correlated best reply correspondence is not convex valued ; that is , @xmath238 is not in general a convex subset of @xmath35 . are independent distributions and that the average of two independent distributions need not be an independent distribution . ] this creates two issues : \(i ) existence of solutions of correlated cfp is not guaranteed by the classical results on differential inclusions we are aware of ( e.g. , @xcite ) . \(ii ) the theory of perturbed differential inclusions @xcite does not apply to non - convex valued correspondences . the first issue can be solved by building piecewise linear solutions of correlated cfp following the same ideas as for two - player games ( see @xcite ) . is well defined , call @xmath239 the game in which the players are reduced to their best - replies to @xmath42 . start with some initial condition @xmath240 . then point to a nash equilibrium of @xmath241 ( i.e. fix @xmath242 and choose @xmath243 ) till the first time , @xmath244 , when , for some player @xmath5 , a strategy which was not a best - reply to @xmath240 is a best - reply to @xmath245 . then iterate . if the times @xmath246 accumulate towards some time @xmath247 , then use the fact that @xmath42 must have a limit when @xmath248 ( because @xmath42 is lipschitz ) . call it @xmath249 and restart from @xmath249 . note that there might in principle be a countable infinity of such accumulation points @xmath247 , and that they might themselves accumulate in some point @xmath250 , but then define @xmath251 as before and restart from there , etc . the largest ( forward time ) interval on which such a solution can be built is both open and closed in @xmath252 and is thus equal to @xmath252 . ] moreover , due to remark [ rem : uni ] , proposition [ p : cfp ] extends to the @xmath2-player setting . it then asserts that correlated cfp leads to no regrets . lemma [ lm0 ] also extends : it asserts that if the maximal regret of player @xmath5 is less than @xmath0 , then she plays only @xmath0-best reply actions to the _ correlated _ average play of the opponents . it follows that , analogously to two - player games , interpolated trajectories of no - regret dynamics are almost surely perturbed solutions of correlated cfp . however , we can not proceed to an analog of theorem [ thm : main ] because of the second issue . thus , whether there is a formal relation between no - regret dynamics and correlated cfp in @xmath2-player games remains an open question . similarly , the results of @xcite on the links between the time - average of the replicator dynamics and cfp are restricted to bimatrix games ( or games with linear incentives ) . theorem [ thm : main ] does not answer whether _ attracting sets _ of cfp have an analogous property under no - regret dynamics . a set @xmath253 is _ eventually attracting _ under a no - regret dynamic process if with any given probability it captures all no - regret trajectories originating from a small enough neighborhood of @xmath254 at all distant enough periods . formally , @xmath254 is eventually attracting if for every @xmath255 there exists @xmath1 and a period @xmath256 such that : for every @xmath257 , if @xmath240 is in an @xmath0-neighborhood of @xmath254 , then @xmath42 converges to set @xmath254 with probability at least @xmath258 . converges to @xmath254 if @xmath259 as @xmath260 . ] for this section it is convenient to replace assumption ( q2 ) by the following one : @xmath261 this is not essential , since the interesting histories are those where both players have positive regrets , in which case ( q2 ) plays no role . , if a player has positive maximal regret , then it remains positive forever . so we can consider histories from a distant enough period @xmath88 where both players have positive regrets and ( q2 ) plays no role . if @xmath88 does not exist , i.e. , some player _ always _ has nonpositive maximal regret , then proposition [ thm : hm ] and ( q2 ) imply that her play is constant , whereas her opponent s play must approach a best reply to it , leading to nash equilibrium . by replacing ( q2 ) by ( q2@xmath89 ) we avoid dealing with this issue . ] a strict nash equilibrium is eventually attracting . indeed , if @xmath240 is close enough to a vertex of @xmath35 corresponding to a strict nash equilibrium @xmath262 , then for each player @xmath5 , action @xmath20 is the unique best reply and there is a negative regret for any action other than @xmath20 . since by ( r3 ) only actions with positive regret can be chosen , and by ( q2@xmath89 ) only best - reply actions can be chosen if _ all _ regrets are nonpositive , action @xmath20 will be played by each player @xmath5 in the following period , and so on . let us now consider a standard generalization of strict nash equilibria . for each @xmath85 , let @xmath263 . with a slight abuse of notation , denote by @xmath264 the set of probability measures on @xmath21 with support on @xmath265 only . the product set @xmath266 is _ closed under rational behavior ( curb ) _ ( @xcite ) if @xmath267 that is , the set @xmath8 is curb if the players pure best reply profiles are contained in @xmath8 whenever they believe that no actions outside of @xmath8 should be played . curb sets are known to be attracting under cfp ( e.g. , ( * ? ? ? * lemma 7 ) ) . however , they need not be attracting under no - regret dynamics in class @xmath80 . indeed , even if the support of @xmath240 is contained in some curb set @xmath8 , there may be positive regrets for actions outside of @xmath8 , since @xmath8 need not be closed under _ better _ replies . however , we show that the intersection of the hannan set and the set of correlated actions with support on a curb set is eventually attracting . formally , let @xmath266 be a curb set . let @xmath268 denote the set of correlated actions with support on @xmath8 only . let @xmath269 . [ prop : curb ] for every curb set @xmath8 , the set @xmath270 is eventually attracting under every no - regret dynamics in @xmath173 . the proof is based on the following observations . for every curb set @xmath8 , if the average play is close enough to @xmath270 , then regrets for all actions outside of @xmath8 are negative ( since @xmath8 is curb ) . hence , by condition ( r3 ) , only actions in @xmath8 will be played in the immediate future . on the other hand , almost sure convergence of maximum regret to zero suggests that , so long as the players choose only actions in @xmath8 , the average play will approach @xmath270 , thus reinforcing the former observation . to prove the result , however , we need to establish bounds on the maximal future regret _ conditional on certain histories _ ( namely , conditional on being close to @xmath270 ) that @xcite do not provide . the complete proof is relegated to appendix a.3 . we now prove an analog of theorem [ thm : main ] for continuous - time dynamics @xcite and the expected version of discrete - time dynamics . both describe trajectories of average _ intended ( mixed ) _ play , rather than average realized ( pure ) play . for this reason , condition ( r3 ) is not needed . indeed , the interest of ( r3 ) is that , together with ( q1 ) , it requires every realized action to be a better reply to the opponents empirical distribution of play ( whenever such actions exist ) . but now we only need every mixed ( expected ) action to be a better reply , and this follows already from conditions ( r1)-(r2 ) and ( q1 ) . besides , these dynamics are deterministic , hence the results we obtain hold _ surely _ ( not just _ almost surely _ ) . the proofs are based on appendix a.1 and are best understood after reading it . consider a continuous - time dynamics @xmath271 where @xmath272 is the ( independent ) joint play at time @xmath23 and @xmath42 the average correlated play . there are two differences with ( [ eq : dt ] ) : time is now continuous , and , more importantly , realized play @xmath77 has been replaced by intended mixed play @xmath273 . as in cfp , start at time @xmath274 with some initial condition @xmath275 . assume that whenever @xmath92 @xmath276 where @xmath67 is a @xmath277 potential function satisfying ( r1 ) , ( r2 ) and the technical condition : + * ( p4@xmath89 ) * there exists @xmath278 such that @xmath279 for all @xmath280 . + this is a part of condition ( p4 ) in @xcite . [ prop : cont ] let @xmath42 be a solution of ( [ eq : contz ] ) and ( [ eq : contq ] ) with @xmath67 satisfying conditions _ ( r1 ) _ , _ ( r2 ) _ and _ ( p4@xmath89 ) _ for all @xmath85 . assume that the initial condition @xmath131 is such that both players have some positive regrets : @xmath281 for all @xmath85 . then the limit set of the beliefs is internally chain transitive for continuous fictitious play . let @xmath282 . * theorem 3.1 and lemma 3.3 ) is used in the proof of lemma 3.3 in @xcite . ] ) show that if @xmath283 , then @xmath284 for all @xmath23 , and @xmath285 as @xmath97 . moreover , by ( r2 ) applied to @xmath286 and definition of @xmath136 , we have : @xmath287 ( this is equation ( 3.3 ) in @xcite ) . thus by lemma [ lm0 ] , @xmath288 . together with lemma [ lm2 ] in appendix a.1 , this implies that @xmath289 is a perturbed solution of cfp in the sense of @xcite . the result then follows from theorem 3.6 of @xcite . assume that if all initial regrets of a player are nonpositive then the dynamics is defined as in @xcite , equation ( 4.9 ) . then it is easily seen that the result of proposition [ prop : cont ] holds for any initial condition @xmath131 . * expected discrete - time dynamics . * the expected motion in ( [ eq : dt ] ) is described by @xmath290 where @xmath291 is the expectation of @xmath77 . assume that @xmath136 is derived by ( q1 ) from a potential function satisfying ( r1)(r2 ) . let @xmath282 . it is easily seen that , as for continuous - time dynamics , @xmath285 as @xmath97 , and if @xmath283 , then for all @xmath23 , @xmath284 and @xmath288 . due to lemmata [ lm2 ] to [ lm4 ] of appendix a.1 and to theorem 3.6 of @xcite , it follows that for a good choice of behavior when all regrets are initially nonpositive , the limit set of the beliefs is internally chain transitive for cfp . denote by @xmath292 the correspondence whose graph is the @xmath0-neighborhood of the graph of @xmath293 : @xmath294 let @xmath295 . in words , action @xmath296 is an _ @xmath0-graph perturbed best reply to @xmath115 _ if there is an action @xmath0-close to @xmath296 which is an exact best - reply to an action @xmath0-close to @xmath115 . this is the notion of perturbation used in the theory of perturbed differential inclusions ( @xcite ) . as illustrated by the example below , it is different from the notion of perturbation of payoffs in the @xmath0-best reply correspondence , i.e. @xmath297 with @xmath175 consider a game where the payoffs of player 1 are given by fig [ f : emp2 ] . let @xmath299 and let @xmath300 . the pure action @xmath184 is a @xmath301-best reply to @xmath302 . using the sup norm , it is at distance 1 from pure action @xmath256 , the unique exact best reply to @xmath302 . nevertheless , @xmath184 is an @xmath0-graph perturbed best reply , because it is an exact best reply to @xmath303 , which is @xmath0-close ( in sup norm ) to @xmath302 . by contrast , for all @xmath304 , action @xmath8 is an @xmath305-best reply , but only a @xmath274-graph perturbed best reply to @xmath302 . a discrete - time trajectory @xmath204 on @xmath119 is a _ payoff perturbed _ fictitious play trajectory if there exists a positive sequence @xmath205 converging to zero such that @xmath306 with @xmath307 and @xmath308 for all @xmath85 and all @xmath120 . it is a _ graph perturbed _ fictitious play trajectory if the same holds but replacing @xmath309 with @xmath310 . a trajectory @xmath208 on @xmath35 generates a _ sequence of beliefs _ @xmath209 in @xmath113 . the proof goes as follows . lemma [ lm1 ] shows that the sequence of beliefs generated by a no - regret dynamics is a payoff perturbed fp trajectory . together with lemma [ lm2 ] , this implies that it is a graph - perturbed fp trajectory ( lemma [ lm3 ] ) . it follows that the interpolated process of a no - regret dynamics trajectory is a perturbed solution of cfp ( lemma [ lm4 ] ) . the result then follows from @xcite . if @xmath95 for all @xmath23 , then by remark [ rem : q2 ] , player @xmath5 plays an @xmath90-best reply for some @xmath90 converging to zero . otherwise , @xmath311 for some @xmath312 . then for all times @xmath313 , @xmath314 ( by remark [ rem : q1 ] ) and player @xmath5 plays an @xmath46-best reply by lemma [ lm0 ] and conditions ( r3 ) and ( q1 ) . since @xmath96 almost surely , the result follows . let @xmath315 be a compact subset of @xmath316 and @xmath317 a correspondence from @xmath315 to itself . for any @xmath318 , let @xmath319 denote the correspondence whose graph is the @xmath320-neighborhood of the graph of @xmath317 : @xmath321 for any @xmath322 , let @xmath323 be an u.s.c . correspondence from @xmath315 to itself . assume that for each @xmath172 in @xmath315 : by contradiction , assume that there exists @xmath327 , a decreasing sequence @xmath329 converging to zero , and sequences @xmath330 and @xmath331 of points in @xmath315 such that @xmath332 for all @xmath2 . by compactness of @xmath315 , we can assume that @xmath330 and @xmath331 converge respectively to @xmath333 and @xmath334 . fix @xmath335 . for all @xmath336 , @xmath337 by ( i ) . since @xmath338 is u.s.c . , it follows that @xmath339 . therefore , by ( i ) and ( ii ) @xmath340 but for @xmath2 large enough , @xmath341 , hence @xmath342 , a contradiction . applied to the best - reply correspondence , lemma [ lm2 ] implies that for any @xmath327 , an @xmath0-perturbed best - reply is a @xmath320-graph perturbed best - reply , provided @xmath0 is small enough . thus we have the next result . given a discrete - time trajectory @xmath348 on @xmath113 , with @xmath349 , define its interpolated process @xmath350 as follows . for all @xmath351 let @xmath352 , where @xmath353 , @xmath85 . this is equivalent to @xmath354 hence for all @xmath355 we have @xmath356 and @xmath357 an absolutely continuous function @xmath358 is a perturbed solution of cfp if there exists a vanishing function @xmath359 such that for almost all @xmath23 , @xmath360 consider a discrete time trajectory @xmath361 such that @xmath362 with @xmath363 and @xmath364 . for all @xmath2 and all @xmath351 , let @xmath365 . obviously , @xmath366 as @xmath367 . moreover , for all @xmath355 , the interpolated process satisfies @xmath368 , so @xmath369 . therefore ( [ eq : int ] ) implies ( [ eq : ptbcfp ] ) ( see also ( * ? ? ? * proposition 2.2 ) ) . we can now prove theorem [ thm : main ] . by lemmata [ lm1 ] and [ lm3 ] , the sequence of beliefs of a solution of a no - regret dynamics in class @xmath80 is almost surely a graph perturbed dfp trajectory . hence , by lemma [ lm4 ] , its interpolated process @xmath126 is a perturbed solution of cfp . this implies that @xmath370 is almost surely a perturbed solution of the best - reply dynamics , in the sense of ( * ? ? ? * definition ii ) . theorem [ thm : main ] now follows from theorem 3.6 of @xcite . [ rem : ew ] assume that at stage @xmath23 , for each @xmath85 , player @xmath5 chooses a pure action according to a mixed action @xmath140 that depends on the previous history @xmath371 . do not assume conditions ( r1)(r3 ) and ( q1 ) , but assume that there exists a vanishing sequence @xmath205 such that for for all @xmath120 and any previous history @xmath371 , @xmath372 , @xmath85 . then it follows from lemma [ lm2 ] , the above proof and ( * ? ? ? * proposition 1.4 and a variant of proposition 1.3 ) that theorem [ thm : main ] applies . as is well known , this is the case for the exponential weights algorithm @xcite that corresponds to @xmath373 with @xmath374 , @xmath375 as @xmath367 , and @xmath376 for some @xmath377 to ensure that this is a no - regret dynamics ( see , e.g. , @xcite ) . the above assumptions are not ( or not trivially ) satisfied by no - regret dynamics in class @xmath80 . indeed , the rate at which the maximal regret vanishes , hence the value @xmath180 such that @xmath378 , may depend on the trajectory . denote a mixed action by @xmath381 . then @xmath382 is a nash equilibrium if and only if @xmath383 or @xmath384 . it is easily seen that all solutions of the best - reply dynamics converge to the set of symmetric nash equilibria . however , the whole state space is ict . indeed , any mixed action @xmath172 can be connected to any other mixed action @xmath385 as follows : starting from @xmath172 , follow a solution pointing towards the edge @xmath384 , then jump on this edge and follow a solution pointing towards the pure strategy @xmath8 ; once close to @xmath8 , jump on the edge @xmath383 , and follow a solution pointing towards @xmath184 ; once close to @xmath184 , make a small jump to reach a point from which a solution points toward @xmath385 ; follow this solution and if needed ( i.e. if @xmath386 ) , make one more jump to reach @xmath385 . this example is also valid for the replicator dynamics and any payoff monotone dynamics in the sense of , e.g. , @xcite . the only difference is that traveling from @xmath387 to @xmath8 and from @xmath8 to @xmath184 can not be done by following solutions of the dynamics but only through long sequences of jumps . note also that in an inward cycling rock - paper - scissors game ( see e.g. , @xcite , or @xcite ) , all solutions of the replicator dynamics converge to one of the four rest points but the whole boundary of the state space is ict ( for the replicator dynamics ) . [ @xmath2-population dynamics ] [ ict : ex2 ] similarly , in the bimatrix version of example [ ict : ex1 ] , all solutions of the two - population best - reply dynamics converge to the set of nash equilibria but the whole state space is ict . again , this is true for all payoff monotone dynamics . similar examples can be given for @xmath2-population dynamics for any @xmath388 . at least for the best - reply dynamics , @xmath389 examples can also be given . consider , for instance , the @xmath389 game : @xmath390 denote mixed actions of players @xmath274 and @xmath190 by @xmath391 and @xmath392 , respectively . the set of nash equilibria is the union of the edges @xmath393 and @xmath394 and all solutions of the two - population best - reply dynamics converge to this set . however , the whole triangle @xmath395 is ict . in these examples , a direct analysis shows that all solutions of no - regret dynamics in class @xmath80 converge to the set of nash equilibria . thus we do not know whether , in general , convergence of all solutions of cfp to the set of nash equilibria entails convergence of no - regret dynamics . the point is that this is not guaranteed by theorem [ thm : main ] . we need some notation . for @xmath12 and @xmath396 , let @xmath397 denote the probability of @xmath398 under the correlated action @xmath11 . let @xmath399 be the neighborhood of @xmath270 in which the total weight on action profiles outside of @xmath8 and the potential of each player are below @xmath400 : @xmath401 where @xmath402 is the regret vector of player @xmath5 : @xmath403 . let @xmath266 be a curb set . let @xmath404 and note that @xmath405 , since @xmath8 is curb . now , consider a no - regret dynamics in @xmath173 defined by potentials @xmath67 , @xmath85 , with trajectory @xmath406 . let @xmath407 be the smallest number such that for all @xmath12 @xmath408 and let @xmath409 . let @xmath410 be the solution of @xmath411 where @xmath412 is a payoff bound . since @xmath413 is weakly increasing in @xmath400 and @xmath414 , there exists a unique solution @xmath410 of and @xmath415 . let @xmath12 . if @xmath11 is close enough to @xmath8 , then @xmath427 for all @xmath85 . if @xmath11 is close enough to @xmath103 , then @xmath428 . thus in the neighborhood of @xmath270 , @xmath429 hence @xmath430 for all @xmath431 . in particular , this holds if @xmath432 ( we omit the proof : easy but lengthy ) . it follows that if @xmath433 , then by conditions ( r3 ) and ( q2@xmath89 ) , @xmath434 for each @xmath85 . suppose that @xmath419 holds and let @xmath435 . then @xmath426 by lemma [ lem : u ] . we proceed by induction . assume @xmath436 for some @xmath437 . since @xmath435 , @xmath438 together with @xmath419 , this implies that @xmath439 . consequently , by lemma [ lem : u ] , @xmath440 . the proof of claim 2 builds up on the proof of theorem 2.1 of @xcite . it is different though , since we need to find the convergence rate of of the maximal regret _ conditional _ on a given initial history ( in particular , on those where the past average play is close to a curb set ) , which @xcite do not provide . so our result can not be directly derived from their proof . hjek - rnyi inequality ( e.g. , @xcite ) implies @xmath446\le\frac{1}{{\varepsilon}^2}\sum\nolimits_{k = m+1}^n c_k^2 var[x_k].\ ] ] using @xmath447 and @xmath448\le \bar \sigma^2 $ ] , the right - hand side can be bounded as follows , @xmath449\le \bar \sigma^2 \sum\nolimits_{k=1}^{n - m } \frac{1}{(m+k)^2}\le \bar \sigma^2 \frac{1}{m}.\ ] ] taking the limit @xmath450 yields @xmath451\le\frac{\bar \sigma^2 } { m{\varepsilon}^2}.\ ] ] the result is immediate by substitution @xmath452 . let @xmath458 and @xmath286 . note that @xmath459 , where @xmath460_{k\in a_i}$ ] . hence @xmath461 the regret for an action is bounded by @xmath462 and the difference between two regret terms by @xmath463 . thus , in sup norm , @xmath464 and @xmath465 . we now show that @xmath457\le c / t$ ] for @xmath474 . by definition of @xmath469 and taylor - lagrange theorem , @xmath475 using ( [ eq : xx0 ] ) we get : @xmath476 since @xmath67 is convex and @xmath477 , we have : @xmath478 therefore @xmath479 so that @xmath480 to prove that @xmath457\le c / t$ ] , it suffices to show that @xmath481=0 $ ] . to see this , note that @xmath482= u_i(k , a_{-i}(t ) ) - u_i(q_i(t ) , a_{-i}(t))$ ] hence @xmath483= q_i(t ) \cdot e[r_{i } | h(t-1 ) ] = \sum_{k \in a_i } q_{i , k } [ u_i(k , a_{-i } ) - u_i(q_i , a_{-i})]= 0.\ ] ] since @xmath140 is proportional to @xmath484 , the result follows . by we have @xmath485+\frac{t}{t+n}p_i(r_i(t)),\end{aligned}\ ] ] where @xmath486 $ ] . as we have assumed @xmath487 , we have @xmath488 next , by lemma [ lem : l ] , @xmath489\le \frac{1}{t+n}\sum_{s = t+1}^{t+n}\frac{c}{s}\le c\frac{\ln ( t+n)-\ln t}{t+n}.\ ] ] maximizing @xmath490 among all @xmath491 yields @xmath492 , hence @xmath489\le \frac{c}{te}.\ ] ] let @xmath493 be a bound on @xmath494 $ ] for all @xmath495 ( this bound exists , since by lemma [ lem : l ] variables @xmath496 are uniformly bounded ) . applying lemma [ lem : k ] , we obtain that for every @xmath255 and every @xmath23 , @xmath497\le \frac{\pi}{2}.\ ] ] hence with probability at least @xmath498 the following holds for all @xmath421 , @xmath499 and it holds for both @xmath85 simultaneously with probability at least @xmath500 . choosing @xmath501 so large that @xmath502 , we obtain that event @xmath419 : @xmath503 occurs with probability at least @xmath500 .

we present a formulation of the interaction between a given source field and a scalar huygens metasurface ( hms ) , a recently introduced promising concept for wavefront manipulation based on a sheet of orthogonal electric and magnetic dipoles . utilizing the equivalent surface impedance representation of these metasurfaces , we establish that an arbitrary source field can be converted into directive radiation via a passive lossless hms if two physical conditions are met : local power conservation and local impedance . expressing the fields via their plane - wave spectrum and harnessing the slowly - varying approximation we obtain semi - analytical formulae for the scattered fields , and prescribe the surface reactance required for the metasurface implementation . the resultant design procedure indicates that the local impedance induces a fresnel - like reflection , while local power conservation forms a radiating virtual aperture which follows the total excitation field magnitude . the semi - analytical predictions are verified by finite - element simulations of hmss designed for different source configurations . besides serving as a flexible design procedure for hms radiators , the proposed formulation also provides a mechanism to incorporate a variety of source configurations into general hms models , as well as physical insight on the conditions enabling purely reactive implementation of this novel type of metasurfaces . epstein and eleftheriades : passive lossless huygens metasurfaces for directive radiation metasurfaces , huygens sources , wavefront manipulation , plane - wave spectrum .

1409.1076

thin sheets with repetitive metallic inclusions or exclusions have been used extensively in the past in antenna applications to control the properties of reflected or transmitted power , e.g. its direction , phase , or polarization @xcite . such surfaces have received increasing attention lately , as part of the intensive research in the field of optical and microwave metamaterials , in an attempt to harness ideas from bulk metamterial explorations to design low - profile components with extraordinary wavefront manipulation capabilities @xcite . in contrast to bulk metamaterials , where subwavelength elements are combined to form a volumetric entity with prescribed local response to electromagnetic fields , in metamaterial sheets , or metasurfaces , these subwavelength atomic units are confined to a region with subwavelength thickness . this geometrical difference should decrease significantly fabrication complexity of metasurfaces and also loss - related problems ; however , it requires development of new design methodologies , as the interaction of electromagnetic fields with metasurfaces is naturally described via effective boundary conditions @xcite , as opposed to effective permeabilities and permittivities ( or effective wave equations ) , more suitable for modelling volumetric metamaterials @xcite . in particular , it was recently recognized that as metasurfaces act as sources of tangential field discontinuities , they can be modelled by a distribution of electric and magnetic surface currents , prescribed by the equivalence principle [ , pp . 575 - 579 ] . hence , in principle , for a given incident field , a desirable electromagnetic field distribution in space can be achieved by engineering the surface to induce currents that would produce the required tangential fields on both of its facets . approximating the required continuous surface currents by a dense distribution of electric and magnetic dipoles , passive surfaces implementing plane - wave refraction were demonstrated @xcite . the elementary sources were formed by subwavelength inductive and capacitive elements that produced the suitable magnitudes of the current in response to the exciting incident plane - wave . simultaneously , it was demonstrated that also active elementary sources may be utilized to introduce desirable field discontinuities , e.g. , to implement a cloaking device based on the same equivalence principle @xcite . as these surfaces were composed of orthogonal electric and magnetic dipoles engineered to induce unidirectional radiation , i.e. acting as huygens sources [ , pp . 653 - 660],@xcite , they were named huygens metasurfaces ( hms ) @xcite . in addition to plane - wave refraction and cloaking , recent reports proposed designs of huygens metasurfaces which implement beam shaping , transmission or reflection coefficient engineering and polarization manipulation ( using tensor huygens metasurfaces ) @xcite . although the design methodologies differ between the various authors , they all rely on the fact that if the dimensions of the unit cells and their spatial arrangement obey certain conditions , the metasurface can be modelled by effective electric and magnetic polarizability distributions , which are translated to position - dependent sheet boundary conditions @xcite . these , in turn , can be equivalently described as surface impedance and surface admittance matrices relating the electric and magnetic field components at the two facets of the metasurface @xcite . following this approach greatly simplifies device design , as it facilitates the development of simple circuit models to huygens metasurfaces @xcite . moreover , as the effective surface impedance and admittance matrices are directly related to the _ locally averaged _ polarizabilities of the elementary scatterers , we may assess the local equivalent surface impedance of a unit cell by simulating or measuring the effective impedance of an infinite periodic array of such identical unit cells ( local periodicity approximation ) @xcite . indeed , this modelling approach was used in recent demonstrations of hmss : given the incident field and the desirable transmitted ( or reflected ) field , the required boundary conditions to support the field discontinuities can be formulated , resulting in the required surface impedance and admittance matrices ( or electric and magnetic polarizability distributions ) . however , for realizing the huygens metasurfaces , it is desirable for the elements implementing the required polarizability distributions to be passive and lossless , i.e. the surface impedance and admittance to possess pure imaginary values . implementing impedance or admittance sheets with nonvanishing real parts requires engineering of gain or loss elements , thereby complicating greatly the design and realization . nonetheless , following the simplistic methodology in which the discontinuities between the desirable and incident fields are directly translated into surface impedances and admittances by no means guarantees the passivity of the resultant metasurface ( see , e.g. , @xcite ) . in fact , in @xcite the design procedure did not consider the passivity limitation , however the resultant leading to a well - functioning , prototype . the reasons for that encouraging outcome were not analyzed therein , though . in addition , almost all huygens metasurfaces presented in the literature to date were designed to be excited by a plane - wave or a beam propagating towards the surface in homogeneous medium @xcite . nevertheless , to facilitate the development of realizable antenna devices based on huygens metasurfaces it would be necessary to extend the current design techniques to enable excitation of the metasurface by localized ( impulsive ) sources , or waveguided modes . a step in that direction was made by holloway _ @xcite , where a line source excitation was considered , but only for metasurfaces with _ constant _ polarizability density ; as demonstrated by @xcite , allowing the polarizability densities to vary along the metasurface could be beneficial , providing more degrees of freedom for the design . in this work , we derive simple rules for designing passive lossless huygens metasurfaces producing directive radiation to a prescribed angle when excited by a given ( arbitrary ) source field . decomposing the fields to their plane - wave spectrum and generalizing the approach presented in @xcite for plane - wave excitation , we show that satisfying two physical conditions is sufficient to guarantee that the desirable functionality can be achieved by purely reactive surfaces : local power conservation across the surface , and local impedance the fields on both sides of the metasurface . enforcing these conditions locally , i.e. at each point on the surface , leads to a complementary set of simple expressions for the surface impedance and surface admittance , as well as facilitates the semi - analytical evaluation of the reflected and transmitted fields . these results enable design of directive hms radiators with a wide range of source excitations , thus extending significantly the possible applications . moreover , the derivation provides clear physical interpretation of the conditions required to implement passive lossless hmss , which may also be indicative to equivalent requirements in more generalized scenarios ( e.g. hmss which perform other functionalities ) . altogether , this forms an efficient and powerful tool for hms engineering , promoting design of novel antenna devices . we consider a 2d configuration @xmath0 in which a huygens metasurface situated at @xmath1 is excited by an arbitrary current distribution limited to the half - space @xmath2 ( fig . [ fig : physicalconfiguration ] ) . the surrounding media is assumed to be homogeneous , with permittivity @xmath3 and permeability @xmath4 , defining the wave impedance @xmath5 . harmonic time dependency of @xmath6 is assumed ( and suppressed ) , defining the wavenumber @xmath7 . . the formalism applies without modification also to scenarios in which the region @xmath8 is occupied by an inhomogeneous medium , as long as its cross - section remains uniform with respect to the @xmath9 axis , i.e the permittivity , permeability and conductivity are a function of @xmath9 coordinate only ( e.g. , plane - stratified media ) [ , pp . 183 - 202].,width=321 ] the is characterized by its surface impedance @xmath10 and surface admittance @xmath11 , inducing discontinuities in tangential magnetic and electric field components , respectively , given by the generalized sheet transition conditions ( gstc ) as formulated by kuester _ @xcite @xmath12 = \\ \bigskip = \dfrac{1}{2}\left[\left.\vec{e}\right|_{x\rightarrow0^{+}}+\left.\vec{e}\right|_{x\rightarrow0^{-}}\right ] \\ y_{sm}\left(z\right)\vec{m}_s = -y_{sm}\left(z\right)\hat{x}\times\left[\left.\vec{e}\right|_{x\rightarrow0^{+}}-\left.\vec{e}\right|_{x\rightarrow0^{-}}\right ] = \\ \bigskip = \dfrac{1}{2}\left[\left.\vec{h}\right|_{x\rightarrow0^{+}}+\left.\vec{h}\right|_{x\rightarrow0^{-}}\right ] , \end{array } \right.\!\!\!\!\!\!\ ! \label{equ : general_impedance_admittance_discontinuity}\end{aligned}\ ] ] where @xmath13 and @xmath14 are the surface currents induced by the tangential electric and magnetic field components , respectively , and we assumed the impedance and admittance matrices can be described by scalar quantities . the half - spaces below and above the metasurface are referred to as region 1 and region 2 , respectively , and we require the sources not to be infinitesimally close to the hms such that a source - free region @xmath15 can be defined just below it ( fig . [ fig : physicalconfiguration ] ) . in region 1 , we distinguish between the incident field and reflected field , while region 2 is populated only by the transmitted fields ( denoted by @xmath16 , @xmath17 , and @xmath18 superscripts , respectively ) @xmath19 and the incident fields are defined as the fields produced by the sources in the absence of the hms includes scattering elements ( e.g. , as we allow in subsection [ subsec : results : line_source_pec ] ) will be given in - . ] . in 2d configurations maxwell s equations can be separated to two decoupled sets of equations related to the transverse electric ( te ) or transverse magnetic ( tm ) polarized field components . in source - free regions , the te - polarized ( @xmath20 ) nonvanishing field components are @xmath21 , and maxwell s equations can be reduced to a wave equation for @xmath22 and curl equations relating the other components to it , @xmath23 analogously , in source - free regions , the tm - polarized ( @xmath24 ) nonvanishing field components are @xmath25 , and maxwell s equations can be reduced to @xmath26 as this work does not deal with polarization manipulation of the source fields , we may design an hms directive radiator for each polarization independently , namely a te - hms and a tm - hms . the te - hms would interact only with @xmath22 and @xmath27 , therefore the required surface impedance @xmath10 would be implemented using scatterers sensitive to electric field in the @xmath28 direction ( e.g. loaded wires parallel to the @xmath28-axis ) and the surface admittance @xmath11 would be implemented using scatterers sensitive to magnetic field in the @xmath29 direction ( e.g. loaded loops whose axis is parallel to the @xmath29-axis ) @xcite . on the other hand , the tm - hms should interact only with @xmath30 and @xmath31 , therefore @xmath10 and @xmath11 would be composed by scatterers sensitive to electric field in the @xmath29 direction and magnetic field in the @xmath28 direction , respectively . as the elements implementing the te - hms and the tm - hms are orthogonal , there should be no coupling between them ; thus , ideally , the two hmss may be combined to a single metasurface without any changes to the independent designs . in view of this observation , we formulate the procedure to design an hms assuming the incident field is te - polarized ; the design rules for tm - polarized excitation can be readily derived by duality ( appendix [ app : theory_tm_excitation ] ) . extension of this work to polarization manipulating hmss @xcite will be addressed in a separate report . as denoted , when the excitation field is te - polarized , the nonvanishing field components are @xmath32 , @xmath33 , and @xmath34 ; the scalar surface impedance only induces _ electric _ currents in the @xmath28 direction ; and the scalar surface admittance only induces _ magnetic _ currents in the @xmath29 direction . in view of a general solution for the fields in the _ source - free region _ @xmath35 can be formulated in the spectral domain as @xcite @xmath36 where @xmath37 is the transverse wavenumber ( associated with the propagation along @xmath29 ) , and @xmath38 is the longitudinal wavenumber ( associated with the propagation along @xmath9 ) ; to satisfy the radiation condition we demand that @xmath39 . the @xmath40 left superscript denotes te - hms related parameters throughout the paper , and @xmath41 is a unit current magnitude . it can be readily verified that all three integrals in satisfy the wave equation of , where @xmath42 is the source - related plane - wave spectrum , @xmath43 is the reflection coefficient in the spectral domain , and @xmath44 corresponds to the spectral content of the transmitted fields . as the sources reside at @xmath8 in region 1 , and the scattering metasurface is situated at @xmath1 , the general solution in @xmath15 must consider both upwards ( incident ) and downwards ( reflected ) propagating waves . in contrast , as neither sources nor scatterers exist in region 2 , only upwards ( transmitted ) propagating waves are allowed for @xmath45 , as to satisfy the radiation condition . we emphasize that the formal solution presented in is valid in general only for @xmath35 and that the presence of the sources at @xmath8 introduces a discontinuity in the fields , which should be accounted for in that region . nonetheless , for designing the hms only the tangential fields at @xmath46 as formulated in are required @xcite , and this formulation is valid _ regardless _ of the nature of the source distribution @xmath8 . in fact , applies without modification also to scenarios in which the region @xmath8 is occupied by an inhomogeneous medium , as long as its cross - section remains uniform with respect to the @xmath9 axis , i.e the permittivity , permeability and conductivity are a function of @xmath9 coordinate only ( e.g. , plane - stratified media ) [ , pp . 183 - 202 ] . as shall be demonstrated in subsection [ subsec : results : line_source_pec ] , this allows for even a wider range of excitation schemes to be investigated using our model . from and the curl equations of we derive the corresponding tangential component of the magnetic field at @xmath35 , @xmath47 and can be rewritten as @xcite @xmath48}{h_z^{\mathrm{trans}}\left(0,z\right)-\left[h_z^{\mathrm{inc}}\left(0,z\right)+h_z^{\mathrm{ref}}\left(0,z\right)\right ] } \\ \!\!\!\!\!{{\vphantom{y}}^{e}_{}{y}}_{sm}\left(z\right)=-\dfrac{1}{2 } \dfrac{h_z^{\mathrm{trans}}\left(0,z\right)+\left[h_z^{\mathrm{inc}}\left(0,z\right)+h_z^{\mathrm{ref}}\left(0,z\right)\right]}{e_y^{\mathrm{trans}}\left(0,z\right)-\left[e_y^{\mathrm{inc}}\left(0,z\right)+e_y^{\mathrm{ref}}\left(0,z\right)\right ] } \end{array}\!\!\!\!\!\!\!\!\ ! \right . \label{equ : te_impedance_admittance_discontinuity}\end{aligned}\ ] ] given an incident field , we would like to determine the required variation of the surface impedance and admittance along the metasurface such that * both the surface impedance and surface admittance are purely reactive ( passive and lossless ) , namely @xmath49 . for that we are willing to allow some reflections from the metasurface . * the transmitted field @xmath50 , @xmath51 , @xmath52 will form a directional radiation to a specified angle @xmath53 with respect to the @xmath9 axis ( fig . [ fig : physicalconfiguration ] ) . to this end we would like to form a virtual aperture on the surface with surface current having the suitable linear phase variation and as uniform as possible magnitude . in order to satisfy the second demand , we require that the spectral content of the transmitted fields @xmath44 will be localized around @xmath54 . @xmath55 and demand that the hms would be designed such that the resulting virtual aperture window would form a slowly - varying as @xmath56 , namely , @xmath57 applying this constraint on enables us to approximate the tangential magnetic field at @xmath56 as @xmath58 substituting and into yields @xmath59 where we have defined dimensionless quantities , proportional to the fields on the lower ( minus - sign superscript ) and upper ( plus - sign superscript ) facets of the metasurface , as follows @xmath60 \\ \vspace{5pt } { { \vphantom{f}}^{e}_{}{\!f}}_{\!\!h}^{-}\left(z\right ) \triangleq \dfrac{1}{i_0k\cos\theta_0 } \left[h_z^{\mathrm{inc}}\left(0,z\right)+h_z^{\mathrm{ref}}\left(0,z\right)\right ] \\ { { \vphantom{f}}^{e}_{}{\!f}}^{+}\left(z\right ) \triangleq { { \vphantom{w\left(0,z\right)}}^{e}_{}{w\left(0,z\right)}}e^{-jkz\sin\theta_0}. \end{array}\!\!\!\!\!\ ! \right . \label{equ : te_dimensionless_quantities}\end{aligned}\ ] ] as we require the hms to be purely reactive , the real power across the metasurface must be conserved . however , to obtain sufficient conditions for designing passive lossless hmss we may require a stricter condition to be met , namely , that the real power impinging the metasurface from region 1 is equal _ locally _ to the real power transmitted to region 2 , at _ each point _ @xmath29 on the surface . to assess the consequences of this local power conservation condition we utilize to express the local power densities along the metasurface in region 1 and 2 using the dimensionless field quantities . these are given , respectively , by the projection of the poynting vector on the @xmath9 axis as @xmath61 @xmath62_{x\rightarrow 0 ^ - } \nonumber \\ & \bigskip = \frac{1}{2}k\eta\left|i_0\right|^2{{\vphantom{f}}^{e}_{}{\!f}}_{\!\!e}^{-}\left(z\right ) { { \vphantom{f}}^{e}_{}{\!f}}_{\!\!h}^{-*}\left(z\right)k\cos\theta_0 , \label{equ : te_poynting_vector_region1}\end{aligned}\ ] ] and as @xmath63 @xmath64_{x\rightarrow 0^+ } \nonumber \\ & \bigskip = \frac{1}{2}k\eta\left|i_0\right|^2\left|{{\vphantom{f}}^{e}_{}{\!f}}^{+}\left(z\right)\right|^2k\cos\theta_0 , \label{equ : te_poynting_vector_region2}\end{aligned}\ ] ] where the asterisk indicates the complex - conjugate operation . the condition for local power conservation is @xmath65 which reads , using - , @xmath66 multiplying the rational functions in by the complex conjugate of their respective denominators and plugging in the local power conservation requirement yields @xmath67 in view of , a _ sufficient _ condition for the real part of both @xmath68 and @xmath69 to vanish is given by @xmath70 resulting in a purely imaginary numerator for both fractions . the physical meaning of the latter is revealed by rewriting using : the condition locally the wave impedance of the fields on the two facets of the metasurface . indeed , substituting into leads to the local impedance condition @xmath71 when is satisfied , the local power conservation condition takes a simpler form , namely , @xmath72 subsequently , we may define the dimensionless field quantities above and below the metasurface as @xmath73 where @xmath74 . substituting these definitions into leads to a complementary set of compact expressions for the desirable surface impedance and surface admittance @xmath75 \\ \!\!\!\!{{\vphantom{y}}^{e}_{}{y}}_{sm}\left(z\right)=-j\dfrac{{{\vphantom{y}}^{e}_{}{y}}_0}{2 } \cot\left[\dfrac{\varphi_{-}\left(z\right)-\varphi_{+}\left(z\right)}{2}\right ] , \end{array}\!\!\!\!\ ! \right . \label{equ : te_impedance_admittance_final}\end{aligned}\ ] ] where @xmath76 is the wave impedance of a te - polarized plane - wave propagating in region 2 at an angle of @xmath53 with respect to the @xmath9 axis , given generally by [ ] @xmath77 , \label{equ : transmitted_wave_impedance}\end{aligned}\ ] ] @xmath78 and @xmath79 being the tangential components of the fields . as @xmath80 , @xmath81 , @xmath82 are all real , the hms defined via is indeed purely reactive , as required . equation prescribes the required variation of the surface reactance and susceptance to implement the desirable hms . however , in order to use this formula , we should instruct how the phases of the fields at @xmath46 , defined as @xmath82 , are to be evaluated . to that end , we first indicate how to find the reflected and transmitted fields from the given source field and the requirement that the power is locally conserved and the impedance is locally . local impedance requires @xmath83 , \label{equ : te_local_impedance_matching_explicit_2}\end{aligned}\ ] ] from which , using the spectral representation of the fields - , the reflection coefficient can be evaluated as @xmath84 _ regardless _ of the source field . equation is merely a private case of the fresnel reflection formula for an incident plane - wave at an angle @xmath85 and a transmitted plane - wave at an angle of @xmath53 , travelling in media with the same permittivity and permeability @xcite . actually , it is a generalization of the reflection coefficient derived in @xcite for passive lossless hmss excited by a plane - wave . however , in the general case considered herein , the source field consists of an infinite number of plane - waves , propagating in different directions . therefore , the reflection coefficient of actually ensures _ individual _ impedance of each of the plane - waves in the source field spectrum - to that of a plane - wave propagating in region 2 with an angle of @xmath53 with respect to the @xmath9 axis . c |l| l| l| l no . & step & input & output & relevant equations + 1 & l local impedance + & [ cols= " < " , ] & @xmath86 & @xmath87 & @xmath88 & @xmath89 & @xmath90 & @xmath91 & @xmath92 & @xmath93 & @xmath94 + to verify our formulation , we have designed and simulated three huygens metasurfaces according to the procedure described in this subsection ( with @xmath95 ) , designated to convert the fields produced by an electric line source at @xmath96 to directive radiation towards @xmath97 , @xmath98 , and @xmath99 ( fig . [ fig : linesourceresults ] ) . for an hms stretching from @xmath100 to @xmath101 ( total length @xmath102 ) the required variation of the surface reactance @xmath103 and surface susceptance @xmath104 with @xmath29 is presented in fig . [ fig : linesourceresults](a)-(c ) . for clarity , only values of @xmath105 and @xmath106 are presented . recalling the physical meaning of the equivalent angle of incidence we can indeed observe a change of signs in the cotangent argument of when the angles formed between the source and the evaluation point approaches @xmath53 , i.e. in the proximity of @xmath107 . this would happen around @xmath108 ( fig . [ fig : linesourceresults](a ) ) , @xmath109 ( fig . [ fig : linesourceresults](b ) ) , and @xmath110 ( fig . [ fig : linesourceresults](c ) ) , for @xmath97 , @xmath98 , and @xmath99 , respectively . moreover , as the difference between the equivalent angles and the transmission angle becomes larger , the effective period of the cotangent should become shorter ; indeed , if we focus on the @xmath111 region of the metasurface , where the equivalent angles of incidence are mostly negative , we observe that the effective period for @xmath97 ( fig . [ fig : linesourceresults](a ) ) , is longer than that of @xmath98 ( fig . [ fig : linesourceresults](b ) ) , which is , in turn , longer than the one corresponding to @xmath99 ( fig . [ fig : linesourceresults](c ) ) . the practical implication of this observation is that if we desire to harness those rays incident upon the metasurface at angles which differ significantly from the desirable transmission angle , we should anticipate a fast variation of the corresponding surface impedance and admittance , which , in turn , requires smaller distances between the elements implementing the metasurface . we have implemented the designed hmss in a commercially available finite - element solver ( ansys hfss ) using one hundred @xmath112-long unit cells comprised of loaded wires and loops @xcite , as described in appendix [ app : hfss_implementation ] . the hms was excited by an electric line source carrying @xmath113 current oscillating at a frequency of @xmath114 ; the simulated electric field variation as a function of position , @xmath115 , is presented in fig . [ fig : linesourceresults](g)-(i ) for the three hmss considered . to compare these results with the theoretical predictions , we have calculated the spectral integrals - in conjunction with , , , and - , to evaluate the fields in the region @xmath116 , as presented in fig . [ fig : linesourceresults](j)-(l ) . the theoretical plots rely on semi - analytical approximations ( appendix [ app : asymptotic_evaluation ] ) , which assume the hms is of infinite extent to calculate the fields in region 1 ( @xmath2 ) and on the virtual aperture @xmath63 . then , to account for the finite hms length when evaluating the fields in region 2 , the virtual aperture window function @xmath117 is truncated at @xmath118 before utilizing . these approximations yield accurate results when most of the excitation power interacts with the hms on its finite extent ( @xmath119 ) . in addition , the steepest - descent - path method is employed to evaluate the radiated power in the far - field regions @xmath120 using the spectral integrals - and based on the same assumptions @xcite ( appendix [ app : asymptotic_evaluation ] ) . this allows us to compare in fig . [ fig : linesourceresults](d)-(f ) the theoretically predicted far - field radiation patterns ( dashed blue line ) with the ones calculated by the hfss simulation ( solid red line ) ; all radiation patterns are normalized to their maximum . although some discrepancies between the simulation results and the theoretical predictions are observed in fig . [ fig : linesourceresults](d)-(l ) , it is clear that the designed hms successfully convert the line source fields to directive radiation toward the desirable angle . both the beam - width and the immediate side lobe levels are in a good agreement with the semi - analytical theory . in all three cases , the simulated directivity values outside the main beam ( in region 2 ) are at least @xmath121 below the peak directivity . importantly , the theoretical calculations indicate that the slowly - varying condition is indeed satisfied for all transmission angles considered , except at some points towards the edge of the metasurface , where the exclusion of diffraction effects introduce a discontinuity in the fields along the @xmath29 axis ( not shown ) . another support for the validity of the theoretical results is provided by the fact that the absolute values of the fields as presented in fig . [ fig : linesourceresults](g)-(i ) and fig . [ fig : linesourceresults](j)-(l ) in @xmath122 units , are to scale . table [ tab : line_source_hms_performance ] concentrates performance parameters calculated from the theoretical and simulated radiation patterns of fig . [ fig : linesourceresults](d)-(f ) . these include the transmission efficiency , i.e. the ratio between the power transmitted to region 2 and the total power radiated by the source ; the half - power beamwidth ( hpbw ) , i.e. the angular difference between the half - power points ; the aperture efficiency , i.e. the ratio between the hpbw of the hms radiation and the hpbw of a uniformly excited aperture with the same length @xmath123 @xcite ; and the ( 2d ) peak directivity @xcite . for each parameter , table [ tab : line_source_hms_performance ] indicates its value as calculated from the simulation results and the theoretical predictions , and the respective relative performance , defined as the ratio between the two . in consistency with the results presented in fig . [ fig : linesourceresults](d)-(l ) , the performance parameters also indicate that the power radiated into region 2 is successfully funnelled into a directive beam , with performance comparable with the theoretical predictions , peak directivity excepted . in view of the theoretically predicted values themselves , two comments are in place . first , we refer to the predicted transmission efficiencies , which are calculated to be around @xmath124 . due to the finite length @xmath123 of the hms , only of the line source power interacts with the hms . if we assume minor reflections from the metasurface , and consider @xmath125 as in our case , this rough estimation leads to a transmission efficiency of @xmath126 , very close to the values in table [ tab : line_source_hms_performance ] hence , these values indicate that if the hms implementation would perfectly match its design , most of the power interacting with the metasurface is expected to be transmitted to region 2 . in other words , the reflection coefficient resulting from the enforcement of local impedance , of which we have little control , should not deteriorate significantly the hms performance . the second comment refers to the aperture efficiencies presented in table [ tab : line_source_hms_performance ] , which do not exceed @xmath127 for all transmission angles considered , even in the ideal theoretical scenario . the reason for these values is the utilization of local power conservation in the hms design , that coerces the virtual aperture window function to follow the profile of the total ( incident+reflected ) fields at @xmath128 . as the line source is situated only @xmath129 away from the metasurface in our configuration , most of the incident power is concentrated in a small region near @xmath108 . thus , effectively , the full length of the hms can not be utilized for radiation , and , in turn , the theoretical limit for the aperture efficiency is below @xmath130 . to conclude this subsection , we note that both the results presented in table [ tab : line_source_hms_performance ] and in fig . [ fig : linesourceresults](d)-(l ) indicate that the main deviations from the theoretical predictions are the increased reflections from the metasurface in region 1 and the incomplete elimination of the incident field in region 2 ; both discrepancies contribute to a significant difference in peak directivity ( table [ tab : line_source_hms_performance ] ) . we believe that these differences originate in the fact that the unit cells implementing the hms have not yet been optimized to exhibit the prescribed surface admittance and impedance accurately over the entire required dynamical range . however , the optimization of the metasurface unit cells requires specialized treatment , including scattering element selection @xcite , and is outside the scope of this paper . to further illustrate the versatility of our formulation , we consider a third excitation configuration , that of an electric line source positioned in front of a pec , the latter is separated from the hms by a distance @xmath131 ( fig . [ fig : physicalconfigurationthreesources](c ) ) ; as in subsection [ subsec : results : line_source ] , the line source current is given by @xmath132 . we present this configuration herein for two reasons : first , to demonstrate how scenarios including multiple reflections ( more generally , plane - stratified configurations ) can be treated using the design procedure presented in section [ sec : theory ] ; second , to provide an example for virtual aperture engineering via careful selection of the source excitation . when region 1 contains not only sources but also scatterers ( e.g. , abrupt interfaces ) the field expressions must take into account the boundary conditions induced by these scatterers , on top of the source condition already encountered in subsection [ subsec : results : line_source ] . the enforcement of these boundary conditions gives rise to multiple - reflection terms in the spectral response of the fields @xcite , which in turn form a dependency between the source - related spectrum @xmath42 of - and the reflection coefficient of the hms . in the configuration considered herein ( fig . [ fig : physicalconfigurationthreesources](c ) ) the boundary conditions introduced by the scatterers at @xmath133 require that the tangential electric field vanishes on the pec , i.e. @xmath134 for each @xmath29 . combining this requirement with the source condition at @xmath135 yields the following expressions for the incident and reflected fields - @xcite @xmath136 where @xmath137 , @xmath138 , and the upper and lower signs in should be used when @xmath35 or @xmath139 , respectively . as in , this change of signs for the regions below and above the source provides the required discontinuity in the magnetic field at @xmath135 due to the electric current ( source condition ) . we emphasize once more that the formulation of - is valid for any reflection coefficient dependency @xmath43 , retaining this degree of freedom required to employ our design procedure . the fraction in the braces of the integrands accounts for the reflection from the pec , and its numerator vanishes when @xmath140 as required . its denominator corresponds to the multiple reflections taking place between the hms and the pec@xcite ; consequently , the poles of the integrands correspond to guided or leaky modes of this structure @xcite . moreover , although the source - related spectrum is now dependent on the reflection coefficient , it can be readily verified that when @xmath141 the incident field in - is reduced to the field produced by a line source and its image , positioned symmetrically @xmath142 below the pec . as in subsubsection [ subsubsec : results : line_source : design ] , we compare - with - in the _ source - free region _ @xmath15 ( i.e. , where @xmath143 and @xmath144 ) to extract the source - related spectrum . this results in @xmath145 a careful examination of the derivation in section [ sec : theory ] reveals that the dependency of the source - related spectrum in the reflection coefficient does not affect the conditions for local impedance ( table [ tab : design_rules ] / step 1 ) , as in the source - free region , this dependency is the same for all fields ( e.g. , - ) . this means that enforcing local impedance on the fields - in that region will result in the same expression for the reflection coefficient . utilizing this , the incident and reflected fields can be completely evaluated in region 1 ( @xmath146 ) , and the dimensionless total field at @xmath61 will thus be given by , reading @xmath147 } { j\beta\cos\left(\beta d\right)-k\cos\theta_0\sin\left(\beta d\right)}e^{jk_tz}. \label{equ : line_source_pec_spectral_domain_f_minus}\end{aligned}\ ] ] as the poles of @xmath148 and the integrand of @xmath149 coincide , the pole contributions to the integral indicate the guided and leaky modal fields in region 1 @xcite . however , as these poles are complex ( @xmath150 ) , their presence do not , in general , introduce significant difficulties to numerical evaluation of the integral . hence , step 1 of table [ tab : design_rules ] may be completed by evaluating @xmath151 and @xmath152 from . as in subsubsection [ subsubsec : results : line_source : design ] we follow [ table [ tab : design_rules ] / step 2 ] and [ table [ tab : design_rules ] / step 3 ] to establish the shape of the virtual aperture and the phase compensation of the hms , respectively , leading to the formulation of the hms surface impedance and surface admittance @xmath153 \\ \!\!\!\!{{\vphantom{y}}^{e}_{}{y}}_{sm}\left(z\right)=-j\dfrac{{{\vphantom{y}}^{e}_{}{y}}_0}{2 } \cot\left[\dfrac{kz\left(\sin\theta_0-\sin\theta_i\left(z\right)\right)+\xi_0}{2}\right ] , \end{array}\!\!\!\!\ ! \right . \label{equ : line_source_pec_impedance_admittance_final}\end{aligned}\ ] ] where we have used again the definition of the equivalent angle of incidence . executing these steps enable the assessment of the transmitted fields as well . as the total field at @xmath61 is now an outcome of multiply - reflected field interference , it seems that controlling the shape of the virtual aperture window function becomes an even more difficult task . however , the introduction of the pec reflector actually enhances our ability to control this function . as discussed in subsubsection [ subsubsec : results : line_source : aperture ] , in many cases the total field on the lower facet of the hms can be approximated by the incident field . when the pec is present , the incident field is created by an interference between the source and its image . therefore , by controlling the relative position of the source and the pec , we may affect this interference pattern , and subsequently the variation of the total field magnitude at @xmath61 . as an example , we may utilize the fact that in the absence of the hms , the distance between the source and the pec , @xmath154 , is related to the internal angle @xmath155 in which constructive interference occurs in region 1 ( fig . [ fig : physicalconfigurationthreesources](c ) ) via @xcite @xmath156 formation of two lobes travelling towards @xmath157 within region 1 should , in general , broaden the effective interaction length of the source field with the hms on its lower facet , which , in turn , may enhance its aperture efficiency . as shall be demonstrated in the following subsubsection , can be used as an initial aperture engineering step , by which the suitable hms - pec distance @xmath131 is determined for given source position @xmath158 and desirable constructive interference direction @xmath155 . in front of a pec at @xmath159 to directive radiation towards @xmath97 . ( a ) required surface reactance @xmath160 ( blue dashed line ) and surface susceptance @xmath161 ( red solid line ) calculated from . ( b ) theoretically predicted ( blue dashed line ) and hfss - simulated ( red solid lines ) normalized radiation pattern ( db scale ) . ( c ) real part of the electric field phasor @xmath115 as simulated by hfss . ( d ) theoretical ( semi - analytical ) prediction of @xmath115.,width=207 ] to verify the design procedure for the configuration considered in this subsection , we have designed and simulated a huygens metasurface according to the prescribed procedure ( with @xmath95 ) , designated to convert the fields produced by an electric line source at @xmath162 , positioned in front of a pec at @xmath163 , to directive radiation towards @xmath97 ( fig . [ fig : linesourcepecresults ] ) . following the discussion in subsubsection [ subsubsec : results : line_source_pec : aperture ] , we aimed at harnessing the pec reflector to improve the theoretical limit for aperture efficiency for a line source @xmath129 below the hms , calculated in subsubsection [ subsubsec : results : line_source : results ] to be @xmath164 for @xmath97 . to that end we positioned the pec such that constructive interference would take place at @xmath165 in region 1 , yielding , utilizing with @xmath166 , an hms - pec distance of @xmath167 . although other values of @xmath155 also broaden the effective interaction length of the incident field and the hms , we have found using the semi - analytically estimated fields ( appendix [ app : asymptotic_evaluation ] ) that @xmath165 yields an optimal result for the @xmath168-long hms under consideration . [ fig : linesourcepecresults](a ) presents the surface reactance and surface susceptance required to implement the desirable hms . as in subsubsection [ subsubsec : results : line_source : results ] , the design procedure output was used to implement the hms in ansys hfss ( following the same procedure and unit cell structure , _ cf . _ appendix [ app : hfss_implementation ] ) , as well as to assess the predicted reflected and transmitted fields ( following the semi - analytical approach of appendix [ app : asymptotic_evaluation ] ) . [ fig : linesourcepecresults](b ) compares between the normalized radiation patterns evaluated using the hfss simulation ( red solid line ) and the semi - analytical theoeretical calculations ( blue dashed line ) , showing a very good agreement between the theory and simulation for the angular range @xmath169 . as implied by the simulated ( fig . [ fig : linesourcepecresults](c ) ) and theoretical ( fig . [ fig : linesourcepecresults](d ) ) field plots , most of the discrepancy at large angles ( @xmath170 ) originate from the fact that the semi - analytical approximations neglect the contribution of fields incident at the plane @xmath1 outside the metasurface @xmath171 to the radiation in region 2 ( appendix [ app : asymptotic_evaluation ] ) . [ fig : linesourcepecresults](c)-(d ) highlight two additional properties of the line - source / pec configuration . first , these two subfigures indicate in a clear manner that the profile of the virtual aperture @xmath63 indeed follows the total field magnitude at @xmath61 . the two spots formed on the lower facet of the hms due to the interference between the line source and its image ( subsubsection [ subsubsec : results : line_source_pec : aperture ] ) are clearly translated into two dominant beams originating from the same positions on the upper facet . second , as discussed briefly in subsubsection [ subsubsec : results : line_source_pec : design ] , the expression for the total field in region 1 contains complex poles , and indeed the field plots indicate that the examined structure supports leaky modes . this is particularly pronounced in the theoretical prediction fig . [ fig : linesourcepecresults](d ) , as the semi - analytical formulation used assumes the hms is of infinite extent over the @xmath29 axis when evaluating the fields in region 1 ( appendix [ app : asymptotic_evaluation ] ) , thus facilitating the observed long - range guidance of moderately leaking modes . l|c c c & hfss & theory & relative performance + + half - power beam width & @xmath172 & @xmath172 & @xmath130 + aperture efficiency & @xmath173 & @xmath173 & @xmath130 + peak directivity & @xmath174 & @xmath175 & @xmath164 + table [ tab : line_source_pec_hms_performance ] summarizes the performance of the els / pec excited hms as a directive radiator . in consistency with fig . [ fig : linesourcepecresults](b ) , the theoretically predicted and numerically simulated hpbw perfectly agree . as anticipated in subsubsection [ subsubsec : results : line_source_pec : aperture ] , the aperture efficiency has indeed increased from @xmath164 for the els excited hms ( table [ tab : line_source_hms_performance ] ) to @xmath173 after introduction of the pec , due to the utilization of the interference between the source and its image to broaden the incident field profile . an enhanced directivity is also recorded , where we note that when the pec is present no power is lost due to radiation to region 1 . it should be also noted that the theoretically predicted directivity is clearly overestimated , as is does not account for the fraction of the source power reaching region 2 without interacting with the hms ( i.e. , via @xmath171 , _ cf . _ appendix [ app : asymptotic_evaluation ] ) , which has non - negligible contribution , as the numerical simulations reveal ( fig . [ fig : linesourcepecresults](c ) ) . we have presented a formulation of a design procedure for scalar huygens metasurface directive radiators , applicable for an arbitrary 2d source excitation . our derivation reveals that satisfaction of two physical conditions is sufficient to guarantee that the designed hms is passive and lossless : local power conservation and local impedance . by expressing the incident , reflected and transmitted fields via their plane - wave spectrum , and utilizing the slowly - varying approximation , we have shown that enforcing local impedance results in a fresnel - like reflection for the various spectral components . furthermore , enforcing local power conservation dictates that the profile of the virtual aperture forming the transmitted directive radiation follows the magnitude of the total ( incident and reflected ) excitation fields . at the end of the design procedure , the fields at all regions may be assessed semi - analytically , and the required variation of the surface impedance and surface admittance along the hms is prescribed . we have verified our formulation using three different source configurations , showing good agreement between the semi - analytical predictions and finite - element simulation results executed by implementing the hms as consecutive unit cells in ansys hfss , yet to be optimized in a dedicated future work . means to control the virtual aperture via modification of the source excitation were discussed and demonstrated as well . the proposed design procedure establishes a flexible and robust foundation for the design and verification of novel antennas , allowing exploration of a vast variety of excitation forms via a single uniform formalism . moreover , the derivation provides insight regarding the physical requirements to achieve passive lossless huygens metasurfaces with desirable functionality , shedding light on previously investigated hmss as well as indicating possible directions for future hms applications . for completeness , we provide here the final results of the derivation of the surface reactance required to implement the dual hms , converting an arbitrary tm - polarized source field to directive radiation towards @xmath53 ( derivable by duality @xcite ) . as denoted in subsection [ subsec : theory : formulation ] , when the excitation field is tm - polarized , the nonvanishing field components are @xmath176 , @xmath177 , and @xmath178 ; the scalar surface impedance only induces _ electric _ currents in the @xmath29 direction ; and the scalar surface admittance only induces _ magnetic _ currents in the @xmath28 direction . analogously to - , the tangential magnetic fields in the _ source - free region _ @xmath35 can be generally formulated in the spectral domain as @xmath179 and the respective tangential electric fields can be derived via ; the @xmath180 left superscript denotes tm - hms related quantities . the scalar surface impedance and admittance may be written in analogy to as @xmath181}{h_y^{\mathrm{trans}}\left(0,z\right)-\left[h_y^{\mathrm{inc}}\left(0,z\right)+h_y^{\mathrm{ref}}\left(0,z\right)\right ] } \\ \!\!\!\!\!{{\vphantom{y}}^{m}_{}{y}}_{sm}\left(z\right)\!=\!\dfrac{1}{2 } \dfrac{h_y^{\mathrm{trans}}\left(0,z\right)+\left[h_y^{\mathrm{inc}}\left(0,z\right)+h_y^{\mathrm{ref}}\left(0,z\right)\right]}{e_z^{\mathrm{trans}}\left(0,z\right)-\left[e_z^{\mathrm{inc}}\left(0,z\right)+e_z^{\mathrm{ref}}\left(0,z\right)\right]}. \end{array}\!\!\!\!\!\!\ ! \right . \label{equ : tm_impedance_admittance_discontinuity}\end{aligned}\ ] ] the transmitted tangential fields are defined using a virtual aperture window function @xmath182 , similarly to , @xmath183 which we assume to satisfy the slowly - varying condition as @xmath56 ( with @xmath40 superscript replaced by @xmath180 ) . the dimensionless field quantities are defined for tm - polarized source field as @xmath184 \\ \vspace{5pt } { { \vphantom{f}}^{m}_{}{\!f}}_{\!\!h}^{-}\left(z\right ) \triangleq -\dfrac{1}{i_0k } \left[h_y^{\mathrm{inc}}\left(0,z\right)+h_y^{\mathrm{ref}}\left(0,z\right)\right ] \\ { { \vphantom{f}}^{m}_{}{\!f}}^{+}\left(z\right ) \triangleq { { \vphantom{w\left(0,z\right)}}^{m}_{}{w\left(0,z\right)}}e^{-jkz\sin\theta_0 } \end{array}\!\!\!\!\!\ ! \label{equ : tm_dimensionless_quantities}\end{aligned}\ ] ] using which local impedance and local power conservation conditions retain the same formulae as in and , with the @xmath40 superscript replaced by @xmath180 , likewise defining the analogue of . finally , the surface reactance for the tm case is given by @xmath185 \\ \!\!\!\!{{\vphantom{y}}^{m}_{}{y}}_{sm}\left(z\right)=-j\dfrac{{{\vphantom{y}}^{m}_{}{y}}_0}{2 } \cot\left[\dfrac{\varphi_{-}\left(z\right)-\varphi_{+}\left(z\right)}{2}\right ] , \end{array}\!\!\!\!\ ! \label{equ : tm_impedance_admittance_final}\end{aligned}\ ] ] where @xmath186 is the wave impedance of a tm - polarized plane - wave propagating in region 2 at an angle of @xmath53 with respect to the @xmath9 axis ( see ) . despite the different wave impedance for te and tm polarizations , the reflection coefficient arising from local impedance remains the same . consequently , explicit evaluation of the scattered fields and the phases @xmath82 may be obtained by using - with the @xmath40 left superscript replaced by @xmath180 . the huygens metasurface designs investigated in subsubsections [ subsubsec : results : line_source : results ] and [ subsubsec : results : line_source_pec : results ] were simulated using a commerically available finite - element solver ( ansys hfss ) to verify the theoretical predictions ( calculated using the continuous equivalent surface impedance and admittance ) via comparison to a more realistic implementation of the hms . -axis while the electric dipole formed by the loaded wires is oriented parallel to the @xmath28-axis.,width=245 ] the simulated @xmath168-long hmss were implemented using one - hundred @xmath112-long unit cells , where the frequency of the time - harmonic excitations was @xmath114 . [ fig : unitcell ] presents the physical configuration of a unit cell , along with its dimensions . each unit cell is composed of two square pec loops , loaded by identical lumped capacitors , and two pec wires , loaded either by identical lumped capacitors or by identical lumped inductors , positioned symmetrically with respect to the center of the cell . the magnetic dipole formed by the loaded loops is oriented parallel to the @xmath29-axis while the electric dipole formed by the loaded wires is oriented parallel to the @xmath28-axis , indicating that the unit cell is only sensitive to te - polarized fields ( see subsubsection [ subsubsec : theory : formulation : spectral_decomposition ] ) . we simulate a 2d environment by placing two pec surfaces at @xmath187 . the surface impedance and admittance of a unit cell for a given lumped loading is evaluated by simulating the response of an infinite periodic array of identical unit cells to a normally incident ( te - polarized ) plane - wave excitation . utilizing the impedance matrix calculated by hfss for the scattering problem , and the circuit model introduced in @xcite , we extract the equivalent surface impedance and admittance of the cell . variation of the lumped loading of the wires and loops induces , respectively , variation of the equivalent surface impedance and admittance of the cell @xcite . this facilitates the generation of a lookup table , matching pairs of surface reactance and susceptance to sets of lumped capacitors or inductors . invoking the principle of local periodicity , a required polarizability variation such as the ones in fig . [ fig : linesourceresults](a)-(c ) and fig . [ fig : linesourcepecresults](a ) can be sampled at a suitable unit - cell - length period ( @xmath112 in our case ) and translated using the lookup table into a set of unit cells with prescribed lumped loadings @xcite . although the electric and magnetic dipoles are orthogonal , and ideally should exhibit decoupled responses to orthogonal electric and magnetic fields , we discovered that due to the finite size of the structures , some coupling does exist . this is especially pronounced when the lumped loading is such that the element is near resonance at the working frequency @xmath188 . thus , to enhance the accuracy of the lookup table , we have chosen to simulate the unit cell as a whole and to relate pairs of reactance / susceptance to pairs of lumped elements . to establish an effective engineering methodology based on the design procedure introduced in section [ sec : theory ] , it is desirable to develop computationally efficient tools to estimate the hms performance before resorting to numerical simulation tools for final optimization . in subsubsection [ subsubsec : results : line_source : results ] and subsubsection [ subsubsec : results : line_source_pec : results ] we have employed such tools , in the form of semi - analytical formulae , to assess the agreement between simulated results of detailed design and the predictions of the more idealized theory . in this appendix we describe the details regarding the evaluation of these semi - analytical formulae , as well as the key approximations used . formally , the theoretical derivation of section [ sec : theory ] is valid only for infinitely long hmss . if the hms is finite , the problem is no longer uniform ( the configuration is not separable ) , and the spectral analysis employed herein can not be carried out . nevertheless , in practice , if an efficient conversion of the source power to directive radiation is to be achieved , the hms length @xmath123 should be judiciously chosen such that most of the interaction between the source power and an infinitely long hms takes place over a finite region @xmath119 of the @xmath189 plane . assuming this is indeed the case ( _ assumption 1 _ ) , we may treat the hms as infinite over the @xmath29 axis without introducing significant errors with respect to the realistic finite - length implementation . this , in turn , facilitates the utilization of to evaluate the fields in region 1 , disregarding any effects arising from the edges of the hms , e.g. diffraction or a discontinuity in the reflection coefficient . a second assumption essential for the validity of our derivation is that the magnitude of the total field impinging upon the hms at @xmath61 , including the reflection induced by local impedance , varies moderately such that the slowly - varying condition is satisfied . utilizing this assumption ( _ assumption 2 _ ) , we may use - to evaluate the fields on the upper facet of the hms ( @xmath63 ) . however , when evaluating the fields in region 2 , we would like to account for the fact that the hms is of finite length @xmath123 , which may have significant effects on the radiated fields . hence , we truncate the virtual aperture window function @xmath117 at the edges of the implemented hms ; practically , the integral of is executed with the infinite limits replaced by @xmath190 . this is equivalent to assuming that the fields at the hms plane @xmath1 vanish in the region where the hms is absent @xmath171 . _ assumption 1 _ above facilitates this approximation . these assumptions are used to evaluate the spectral content of the reflected and transmitted fields , and consequently , via the spectral integrals - , the fields everywhere above and below the hms ( fig . [ fig : linesourceresults](j)-(l ) and fig . [ fig : linesourceresults](d ) ) . nonetheless , when far - field evaluation is required , such as for the assessment of the radiation pattern , straightforward spectral integration is problematic due to the highly oscillating nature of the radiation integrals as @xmath191 . hence , for the evaluation of far - field radiated power we apply the steepest - descent - path method , yielding closed - form asymptotic approximations for the spectral integrals . for the typical spectral integrals - @xmath192 the asymptotic evaluation at a point @xmath193 where @xmath194 and @xmath195 $ ] is given by @xcite @xmath196 where @xmath197 for @xmath45 and @xmath198 for @xmath2 , assuming the sign attached to @xmath199 in the exponent of satisfies the radiation condition at the respective regions . a.e . gratefully acknowledges the support of the lyon sachs postodoctoral fellowship foundation as well as the andrew and erna finci viterbi fellowship foundation of the technion - israel institute of technology , haifa , israel . d. sievenpiper , l. zhang , r. f. j. broas , n. alexopolous , and e. yablonovitch , `` high - impedance electromagnetic surfaces with a forbidden frequency band , '' _ ieee trans . theory techn . _ , vol . 47 , no . 11 , pp . 20592074 , nov . 1999 . c. l. holloway , e. f. kuester , j. a. gordon , j. ohara , j. booth , and d. r. smith , `` an overview of the theory and applications of metasurfaces : the two - dimensional equivalents of metamaterials , '' _ ieee antennas propag . mag . _ , vol . 54 , no . 2 , pp . 1035 , apr . 2012 . e. kuester , m. mohamed , m. piket - may , and c. holloway , `` averaged transition conditions for electromagnetic fields at a metafilm , '' _ ieee trans . antennas propag . _ , vol . 51 , no . 26412651 , oct . 2003 . e. martini , g. sardi , and s. maci , `` homogenization processes and retrieval of equivalent constitutive parameters for multisurface - metamaterials , '' _ ieee trans . antennas propag . _ , vol . 62 , no . 4 , pp . 20812092 , apr . 2014 . m. selvanayagam and g. v. eleftheriades , `` discontinuous electromagnetic fields using orthogonal electric and magnetic currentsfor wavefront manipulation , '' _ opt . express _ 21 , no . 12 , pp . 1440914429 , june 2013 . y. g. rapoport , s. tretyakov , and s. maslovski , `` nonlinear active huygens metasurfaces for reflectionless phase conjugation of electromagnetic waves in electrically thin layers , '' _ j. electromagnet . wave . _ , vol . 27 , no . 11 , pp . 13091328 , 2013 . c. holloway , m. mohamed , e. kuester , and a. dienstfrey , `` reflection and transmission properties of a metafilm : with an application to a controllable surface composed of resonant particles , '' _ ieee trans . _ , vol . 47 , no . 4 , pp . 853865 , 2005 . c. l. holloway , d. c. love , e. f. kuester , j. a. gordon , and d. a. hill , `` use of generalized sheet transition conditions to model guided waves on metasurfaces / metafilms , '' _ ieee trans . antennas propag . _ , vol . 60 , no . 11 , pp . 51735186 , 2012 . a. epstein , `` rigorous electromagnetic analysis of optical emission of organic light - emitting diodes , '' ph.d . dissertation , dept . , technion - israel insititute of technology , haifa , israel , 2013 . [ online ] . available : http://arielepstein.webs.com/research/thesis.pdf g. lovat , p. burghignoli , f. capolino , d. jackson , and d. wilton , `` analysis of directive radiation from a line source in a metamaterial slab with low permittivity , '' _ ieee trans . antennas propag . _ , vol . 54 , no . 3 , pp . 10171030 , ariel epstein ( s12-m14 ) received the b.a . degree in computer science from the open university of israel , raanana , the b.a . degree in physics , and the b.sc . and ph.d . degrees in electrical engineering from the technion - israel institute of technology , haifa , israel , in 2000 , 2003 , and 2013 respectively . he is currently a lyon sachs postdoctoral fellow at the department of electrical and computer engineering in the university of toronto , ontario , canada . his current research interests include utilization of electromagnetic theory , with emphasis on analytical techniques , for novel applications , e.g. electromagnetic metamaterials for antennas and microwave components . epstein was the recipient of the young scientist best paper award in the ursi commission b international symposium on electromagnetic theory ( emts2013 ) , held in hiroshima , japan on may 2013 , as well as the best poster award at the 11th international symposium on functional @xmath200-electron systems ( f@xmath200 - 11 ) , held in arcachon , france on june 2013 . george v. eleftheriades ( s86-m88-sm02-f06 ) earned the m.s.e.e . and ph.d . degrees in electrical engineering from the university of michigan , ann arbor , mi , usa , in 1989 and 1993 , respectively , where he was involved in the development of sub - millimeter - wave and thz technology sponsored by nasa . from 1994 to 1997 , he was with the swiss federal institute of technology , lausanne , switzerland , where he was engaged in the design of millimeter and sub - millimeter - wave receivers for the european space agency . currently , he is a professor in the department of electrical and computer engineering at the university of toronto , on , canada , where he holds the canada research chair in nano- and micro - structured electromagnetic materials and applications and the velma m. rogers graham chair in engineering . he is a recognized international authority and pioneer in the area of negative - refractive - index metamaterials . these are man - made materials which have electromagnetic properties not found in nature . he introduced a method for synthesizing metamaterials using loaded transmission lines . together with his graduate students , he provided the first experimental evidence of imaging beyond the diffraction limit and pioneered several novel microwave components and antennas using these transmission - line based metamaterials . his research has impacted the field by demonstrating the unique electromagnetic properties of metamaterials ; used in lenses , antennas , and other microwave components to drive innovation in fields such as defence , medical imaging , microscopy , automotive radar , and wireless telecommunications . presently , he is leading a group of graduate students and researchers in the areas of electromagnetic negative - refraction metamaterials , transformation optics , metasurfaces , small antennas and components for broadband wireless communications , novel antenna beam - steering techniques , plasmonic and nanoscale optical components , and fundamental electromagnetic theory . eleftheriades served as an associate editor for the ieee transactions on antennas and propagation ( ap ) . he also served as a member of the ieee ap - society administrative committee ( adcom ) from 2007 to 2012 and was an ieee ap - s distinguished lecturer from 2004 to 2009 . he served as the general chair of the 2010 ieee international symposium on antennas and propagation held in toronto , on , canada . papers that he coauthored have received numerous awards such as the 2009 best paper award from the ieee microwave and wireless propagation letters , twice the r. w. p. king best paper award from the ieee transactions on antennas and propagation ( 2008 and 2012 ) , and the 2014 best paper award from the ieee antennas and wireless propagation letters . his work has been cited more than 10,000 times , and his h - index is 47 . he received the ontario premier s research excellence award and the university of toronto s gordon slemon award , both in 2001 . in 2004 he received an e.w.r . steacie fellowship from the natural sciences and engineering research council of canada . he received the 2008 ieee kiyo tomiyasu technical field award `` for pioneering contributions to the science and technological applications of negative - refraction electromagnetic materials . '' in 2009 , he was elected a fellow of the royal society of canada .

we use computer simulations to study the relaxation dynamics of a model for oil - in - water microemulsion droplets linked with telechelic polymers . this system exhibits both gel and glass phases and we show that the competition between these two arrest mechanisms can result in a complex , three - step decay of the time correlation functions , controlled by two different localization lengthscales . for certain combinations of the parameters , this competition gives rise to an anomalous logarithmic decay of the correlation functions and a subdiffusive particle motion , which can be understood as a simple crossover effect between the two relaxation processes . we establish a simple criterion for this logarithmic decay to be observed . we also find a further logarithmically slow relaxation related to the relaxation of floppy clusters of particles in a crowded environment , in agreement with recent findings in other models for dense chemical gels . finally , we characterize how the competition of gel and glass arrest mechanisms affects the dynamical heterogeneities and show that for certain combination of parameters these heterogeneities can be unusually large . by measuring the four - point dynamical susceptibility , we probe the cooperativity of the motion and find that with increasing coupling this cooperativity shows a maximum before it decreases again , indicating the change in the nature of the relaxation dynamics . our results suggest that compressing gels to large densities produces novel arrested phases that have a new and complex dynamics .

1502.00249

in nature and in our daily life , many soft materials are formed due to the dynamical arrest of the constituent particles @xcite . usually they are labelled as gels if the particle density is low and as glasses if the density is large . however , the difference between these two states is at present not very well understood and therefore it is not always easy to distinguish them . despite this difficulty quite a few features in this glass - gel cross - cover regime have been studied extensively . in dense glass forming liquids the slowing down of dynamics is related to the mutual steric - hindrance in the motion of the constituent particles . the dynamical properties of these glass formers are characterized by a stretched - exponential shape of relaxation functions @xcite , or similarly by the anomalous , exponential tails in the van hove distributions of particle displacements @xcite . many of these dynamical features are described well by mode - coupling theory @xcite . at even lower temperatures , the relaxation dynamics can be understood by means of the random first order transition theory @xcite . in these glass formers , the structure is usually close - packed for hard sphere or van der waals type interactions and is accompanied by a super - arrhenius increase of viscosity . or , if there are covalent bondings , they form network - like structures and exhibit an arrhenius increase in viscosity . on the other hand , chemical gels or rubbers are soft solids having random network structures @xcite . the cross linking of the permanent bonds between the constituent monomers happens during the synthesis process inducing a vulcanization transition once the density of the links exceeds the percolation threshold @xcite . different static and dynamical properties in the vicinity of this transition have been studied , both using simulations and theoretical models ( e.g. , see @xcite ) . physical gels are on the other hand low - density network structures with bonds that can be broken / realigned by thermal fluctuations within finite timescales @xcite . one possible nonequilibrium path to physical gelation is via a thermal quench across the liquid - gas spinodal leading to dynamically arrested states @xcite , that show complex aging phenomena @xcite . in general , these paths lead to spatially heterogeneous structures . however , in recent times , considerable effort has been made to devise ways by which spatially homogeneous physical gels can be formed @xcite . for such gel - forming systems , a wide variety of relaxation functions have been reported : logarithmic @xcite , stretched @xcite or compressed exponentials @xcite . while theoretical models @xcite have been proposed to account for such dynamical properties , they are at this time certainly not yet comprehensive . of particular interest is the interplay between these different arrest mechanisms , viz . gel and glass , since their competition can be used to engineer materials with novel functionalities @xcite . similar studies have been carried out in systems with competing lengthscales @xcite or interactions @xcite . yet , only few models do allow to study the low density gel phase and high density glass at the same time . accessing this regime is , however , necessary for investigating the structure and dynamics at those intermediate densities where the gel transforms into a glass and vice versa . here , we study a simple model with direct experimental relevance @xcite , and which permits us to traverse the density regime of interest and hence to study the interplay of different processes which lead to either gelation or glassiness . furthermore , our model also allows us to tune the lifetime of bonds , a feature that is usually not present in other models ( for example , see the recent work @xcite ) . on one hand , this facilitates a wider exploration of the relaxation dynamics of such model physical gels , but also allows to disentangle the origin of the apparently anomalous relaxation observed in these systems . in fact , mode - coupling theory predicts , e.g. , logarithmic relaxation whenever two different arrest lines meet , one gel - like and another glass - like , and relates it to an underlying higher - order singularity in the theory @xcite . the versatility of our model , and in particular the possibility of tuning at will the bonds lifetime , allows us to explore the different mechanisms at play and hence to elucidate the interplay of various lengths and time - scales @xcite . the paper is structured as follows . in section ii we explain the details of our model transient - network fluid , together with the numerical schemes used to simulate its dynamics . the phase diagram of the model system and its structural properties are discussed in sections iii and iv , respectively . in section v , we analyze in details the fluid s dynamics , quantified by the mean squared displacement and the incoherent scattering function . for that , we follow different routes across the phase diagram , which allows us to clearly understand the interplay between the gel and the glass regimes . finally , the dynamical heterogeneities which characterize the slow dynamics in the gel and glass phases are studied in section vi , followed by a summary of our results and a broader perspective in section vii . our model system is a coarse - grained representation @xcite of a transient gel which has been studied in experiments @xcite . in this system , an equilibrium low - density gel is obtained by adding telechelic polymers to an oil - in - water microemulsion . since the polymer end - groups are hydrophobic , the polymers effectively act as ( attractive ) bridges between the oil droplets they connect . the strength , lengthscale and typical lifetime of these bridging polymers can be controlled at will . denoting by @xmath0 the number of polymers connecting droplets @xmath1 and @xmath2 , we have established in refs . @xcite , that the following interaction is a reasonable coarse - grained representation of this ternary system : @xmath3 the first term is a soft repulsion acting between bare oil droplets , where @xmath4 , @xmath5 is the diameter of droplet @xmath1 , and @xmath6 is the distance between the droplet centers . the second term describes the entropic attraction induced by the telechelic polymers , which has the standard `` fene '' ( finitely - extensible nonlinear elastic ) form known from polymer physics @xcite , @xmath7 , and accounts for the maximal extension @xmath8 of the polymers . the last term introduces the energy penalty @xmath9 for polymers that have both end - groups in the same droplet . the most drastic approximation of the model ( [ model ] ) is the description of the polymers as effective bonds between the droplets , which is justified whenever the typical lifetime of the bonds is much larger than the timescale for polymer dynamics in the solvent @xcite . thus , for exploring the different properties of such a model , the relevant variables are the droplet volume fraction @xmath10 and the number of polymer heads per droplet @xmath11 @xcite . in order to describe the dynamics of the system , we use a combination of molecular dynamics to propagate the droplets with the interaction ( [ model ] ) , and local monte carlo moves with metropolis acceptance rates @xmath12 $ ] to update the polymer connectivity matrix @xmath0 , where @xmath13 is the difference in potential energy of the system for the two bond configurations @xcite . thus @xmath14 is the timescale governing the renewal of the polymer network topology . in order to prevent crystallization at high volume fractions ( which would be the case for the monodisperse model studied earlier @xcite ) , we use a polydisperse emulsion with a flat distribution of particle sizes in the range @xmath15 $ ] ( having a mean diameter @xmath16 ) . the units of length , energy and time are respectively @xmath17 , @xmath18 and @xmath19 where @xmath20 is the mass of the particles . the space of control parameters is quite large . therefore we set @xmath21 as measured in experiments @xcite , @xmath22 , and @xmath23 and @xmath24 , and vary the remaining parameters @xmath25 . these choices for the parameter values leads to a phase diagram which is similar to the one obtained in experiments . our numerical simulations are done for a three dimensional system of @xmath26 particles . the equations of motions of these particles are integrated using a velocity verlet scheme with a time step of @xmath27 . here , most of the results are reported for @xmath28 , although we also explore other values of @xmath14 : @xmath29 to illustrate some of the dynamical features of the system . at each volume fraction @xmath10 , we first equilibrate the system of particles without any links ( @xmath30 ) . once equilibrated , bonds are introduced corresponding to the required value of @xmath11 and then the system is again equilibrated to obtain the proper distribution of bonds per particle . since the structure of the network is independent of the choice of @xmath14 @xcite , we use a small value of @xmath31 to expedite the equilibration process . subsequently , data is generated by continuing the simulations with different @xmath14 values when required and the averages are typically calculated over 100 different time origins . we also do simulations for the case when the bonds between two particles are completely frozen . in order to do a proper sampling of the network configurations for this situation , we use 6 initial configurations ( positions , connectivities ) from the simulations with a finite @xmath14 as initial inputs for subsequent evolution of the particle positions using molecular dynamics with the connections now permanently fixed . and @xmath10 . the diamonds correspond to the coexistence region between gas and liquid , the circles to the sol phase and the squares to the gel phase . the horizontal dashed line corresponds to onset of percolation , the dot - dashed line indicates the phase - coexistence boundary and the thick line marks the predicted glass line . snapshots of typical configurations in each phase have been published elsewhere , see fig . 3 in @xcite.,width=302 ] we begin by summarizing our earlier findings for the phase diagram ( shown in fig . [ phasediagram ] ) for this model . if the number of polymers is small ( i.e @xmath32 ) , the system is in a simple liquid phase ( the sol ) at small values of @xmath10 . in this regime , the distribution of connectivities per particle is just an exponential @xcite . with increasing @xmath10 , the dynamics in the sol regime becomes slow and one eventually enters a glassy phase at large @xmath10 , characterized by a very strong increase of the timescales for structural relaxation . if the number of bonds is large and @xmath10 is small , phase separation is observed due to strong attractions between the droplets . here , the distribution of connectivities becomes bimodal , with one of the two peaks corresponding to a well - connected liquid and the other to free particles . for intermediate values of @xmath11 the system is in a gel phase . in this region of the phase diagram the particles are connected together and form a percolating cluster and the spatial density is homogeneous . here , the connectivity distribution is peaked around the average value and has an exponential tail . note that this gel is an equilibrium phase since the polymer network is constantly rewired on the timescale @xmath14 . however , if we go to large enough volume fractions , we observe again a glassy system for all connectivities , with the corresponding divergence of relaxation timescales . at low @xmath10 , in the gel phase , the main slow relaxation process is related to the connections between the droplets by means of the polymers and the timescale associated with its reconfiguration . at low @xmath11 , as the system becomes glassy at large @xmath10 , the origin of the slow dynamics is the steric hindrance caused by the caging of each particle by its neighbors . in the region where both @xmath11 and @xmath10 act as a source for slow dynamics , we have shown that the generic relaxation process has three steps but with proper tuning of the two relaxation timescales , one can also obtain logarithmic decays of the relaxation function @xcite . in the following we will discuss in details the interplay between these two relaxation processes and show the consequences on the nature of different dynamical quantities as we move around in the phase diagram . computed for the particles at a volume fraction of @xmath33 for different connectivities @xmath34 and 12 . inset : @xmath35 at @xmath36 for @xmath37 and 50 . the dashed line corresponds to @xmath38 . , width=302 ] before we discus the different dynamical properties of the system , we briefly look at its structure . in fig . [ sq70 ] we plot the static structure factor , @xmath35 , for a system that is dense , @xmath33 , varying the connectivity @xmath11 . the general shape of @xmath35 is very similar to the one of a simple liquid and hence we can conclude that the system is homogeneous . also , we see no significant dependence of @xmath35 on @xmath11 , thus showing that at high densities the structure is mainly governed by steric hindrance . for intermediate densities , however , we do note the weak dependence on @xmath11 in the regime of small wave - vectors , illustrated in the inset of fig . [ sq70 ] where we show @xmath35 for @xmath36 . this is at a volume fraction at which for large @xmath11 the system approaches the co - existence region , and hence one starts to see the emergence of a power - law behavior at small values of @xmath39 with increasing @xmath11 ; the data for @xmath40 can be approximated by @xmath38 ( which is expected in proximity to phase co - existence @xcite ) . with changing @xmath11 for ( a ) @xmath41 , ( b ) @xmath42 , and ( c ) @xmath33 , using a bond lifetime of @xmath28 . , title="fig:",width=302 ] with changing @xmath11 for ( a ) @xmath41 , ( b ) @xmath42 , and ( c ) @xmath33 , using a bond lifetime of @xmath28 . , title="fig:",width=302 ] with changing @xmath11 for ( a ) @xmath41 , ( b ) @xmath42 , and ( c ) @xmath33 , using a bond lifetime of @xmath28 . , title="fig:",width=302 ] . ( a ) @xmath41 , ( b ) @xmath42 , and ( c ) @xmath33 , using @xmath28 . note the change of time span in different panels , from a maximum time of @xmath43 in ( a ) to @xmath44 in ( c ) . the inset in panel ( c ) shows the collapse of correlation functions for @xmath33 after rescaling by the relaxation time , similar to time - temperature superposition principle . , title="fig:",width=302 ] . ( a ) @xmath41 , ( b ) @xmath42 , and ( c ) @xmath33 , using @xmath28 . note the change of time span in different panels , from a maximum time of @xmath43 in ( a ) to @xmath44 in ( c ) . the inset in panel ( c ) shows the collapse of correlation functions for @xmath33 after rescaling by the relaxation time , similar to time - temperature superposition principle . , title="fig:",width=302 ] . ( a ) @xmath41 , ( b ) @xmath42 , and ( c ) @xmath33 , using @xmath28 . note the change of time span in different panels , from a maximum time of @xmath43 in ( a ) to @xmath44 in ( c ) . the inset in panel ( c ) shows the collapse of correlation functions for @xmath33 after rescaling by the relaxation time , similar to time - temperature superposition principle . , title="fig:",width=302 ] with r , for different @xmath10 ( shown in fig . [ msdt100 ] ) . the dotted line , corresponding to an exponential function , is drawn as a guide to the eye.,width=302 ] we now characterize the dynamical properties of the system by focusing on two quantities : ( i ) the mean squared displacement , defined as @xmath45 \right)\rangle$ ] . here @xmath46 is the position of particle @xmath1 at time @xmath47 , @xmath39 is the wave - vector , and @xmath48 corresponds to the ensemble average . for increasing volume fraction @xmath49 , we discuss simultaneously the data for @xmath50 , shown in fig . [ msdt100 ] , and @xmath51 , shown in fig . [ fsq6t100 ] , computed at a wave - vector value @xmath52 . thus , the measured @xmath51 probes the relaxation dynamics on length scales that are slightly larger than the average particle diameter ( the peak in the structure factor @xmath35 occurs at @xmath53 ) . we start in the pure gel phase ( @xmath36 ) . the mean squared displacement , fig . [ msdt100]a , shows that the increasing number of bonds restricts the motion of the particles in that for @xmath54 we see the emergence of an intermediate regime which develops into a well - defined plateau at @xmath55 . the height of this plateau depends significantly on @xmath11 , showing that this cage motion is directly related to the transient bonds between the particles . the self intermediate scattering function , fig . [ fsq6t100]a , shows for small @xmath11 a very rapid decay . this changes in that for @xmath11 around 4 - 6 a plateau develops at intermediate and long times , the height of which depends strongly on @xmath11 . the presence of this increasing plateau height , which is reminiscent to the so called type - a transition of mode - coupling theory @xcite , indicates that the relaxation mechanism is changing : for small @xmath11 the motion of the particles is only weakly slowed down by the presence of the bonds , which typically break on the time scale of @xmath14 . however , for larger @xmath11 breaking a few bonds is not enough to allow the particles to move since the remaining bond still allow to maintain the particle inside its cage . hence this makes that at large @xmath11 the relaxation dynamics does not depend very strongly on @xmath11 anymore . this effect is seen in fig . [ difft100 ] where we show the diffusion constant of the particles , @xmath56 , ( as obtained from the mean squared displacement at long times ) as a function of @xmath11 . for small @xmath11 , @xmath56 shows a rather strong @xmath57dependence , whereas for @xmath58 this dependence becomes weaker . below we will discuss the @xmath57dependence of the relaxation time in more detail . next , we look at the data for an intermediate density , viz . @xmath42 , and in fig . [ msdt100]b we show the corresponding @xmath50 . like for @xmath36 , the longtime diffusion decreases with increasing @xmath11 , and the @xmath57dependence of the diffusion constant shows again a break at around @xmath59 ( see fig . [ difft100 ] ) . for this value of @xmath10 we observe , however , for @xmath60 , at intermediate times a shoulder in @xmath50 . this shoulder , clearly visible for @xmath61 , is related to the presence of the bonds that lead to a caging of the particles on the length scale related to @xmath8 , the maximum extension of a bond . thus the hint of the short - time plateau ( when @xmath62 ) and again one at later time ( when @xmath63 ) reflects the presence of the two different mechanisms for constraining particle motion , viz . local steric hindrance and the network bonds . since each type of caging leads to a plateau in the intermediate scattering function @xcite , the existence of the two competing mechanisms makes that at intermediate times @xmath51 , shown in fig . [ fsq6t100]b , has a very slow , almost logarithmic , decay , if @xmath11 is around 6 . more details on this particular case are give in the context of fig . [ decon ] . if one compares the data for @xmath55 at the two volume fractions @xmath36 and @xmath42 , we recognize that the height of the plateaus in @xmath51 increases with @xmath10 from which one can conclude that the proximity of the particles leads to increased tightening of the cage . if the density is increased further to @xmath33 , the relaxation dynamics becomes strongly dominated by the steric hindrance mechanism . already for @xmath30 one sees a weak plateau in @xmath50 and its length grows rapidly with increasing @xmath11 without changing much its height ( see fig.[msdt100]c ) . this is the typical behavior of simple glass - forming liquids @xcite . at the same time the self intermediate scattering function shows the growth of a shoulder with _ finite _ height and this height depends again only weakly on @xmath11 , fig . [ fsq6t100]c . in contrast to the case at lower densities , here the shape of the correlator is basically independent of @xmath11 . this is demonstrated in the inset of fig . [ fsq6t100]c where we plot @xmath51 as a function of @xmath64 , with the relaxation time @xmath65 defined by @xmath66 . the fact that this presentation of the curves leads to a nice master curve shows that we have for this system a time-@xmath11 superposition , in analogy to the time - temperature superposition found in simple glass - forming systems @xcite . despite the qualitative changes seen in @xmath50 and @xmath51 if @xmath10 is increased , the @xmath57dependence of the diffusion constant for @xmath33 is very similar to the one seen at lower densities , see fig . [ difft100 ] . also for this high value of @xmath10 we see that this dependence is relatively strong at small @xmath11 and becomes weaker if @xmath67 . hence the fact that at low @xmath11 just few bonds have to be broken in order to allow a particle to move whereas at high @xmath11 this is not a sufficient condition , is reflected in the @xmath57dependence of @xmath56 at _ all _ @xmath10 . for @xmath68 , a ) @xmath69 b ) @xmath28 and c ) @xmath70.,title="fig:",width=302 ] for @xmath68 , a ) @xmath69 b ) @xmath28 and c ) @xmath70.,title="fig:",width=302 ] for @xmath68 , a ) @xmath69 b ) @xmath28 and c ) @xmath70.,title="fig:",width=302 ] we will now explore further the interplay between the two processes leading to slow relaxation , i.e. the nearest - neighbor caging and the constrained motions due to the polymer bonds . since the relative importance of these two processes depends on the lifetime of the polymer bonds , we will in the following vary this lifetime and consider values of @xmath71 , and @xmath72 , and at fixed state - point of @xmath73 , where logarithmic decay in the time - correlation function is observed . we study how the shape of @xmath51 changes with varying @xmath14 and will relate this to the interplay between the two processes . this is done for different values of wave - vector in order to see the how relaxation timescales vary over different lengthscales . we begin by looking at the case of the large @xmath69 , i.e. when the polymer bonds hinder the motion on a time scale longer than the steric hindrance effect , see fig . [ r63pv61fsqall]a . the correlation function reflects three different relaxation processes which can be seen for all value of @xmath39 . initially , the particles rattle inside the cage , resulting in partial relaxation of the correlation function on a timescale @xmath74 . later on , the particles escapes from the cage of neighboring particles , ( which for this value of @xmath10 not very pronounced ) , but the relaxation process is then held up by the polymer bonds . eventually , the polymer network rewires on a timescale which is proportional to @xmath69 and the particles start the final relaxation process . the height of the plateau in @xmath51 increases with decreasing @xmath39 , which is the typical behavior for a glassy system @xcite . however , for the range of wave - vectors explored , the final timescale for decay of @xmath51 depends only weakly on @xmath39 . thus , over these length scales , the relaxation process is determined by the reconfiguration of the network . however , at larger length scales , one can expect that hydrodynamic effects will eventually dominate and this will then determine the relaxation timescales . for intermediate values of the bond lifetime , @xmath28 in fig . [ r63pv61fsqall]b , the final decay has moved to shorter times and makes that now the interplay between the two processes results in a logarithmic decay of @xmath51 as discussed elsewhere @xcite . this logarithmic dependence is seen for a range of @xmath39 values ( see fig . [ r63pv61fsqall]b ) , with the time - window over which it exists decreasing with decreasing @xmath39 . this implies that this form for the correlation function occurs only for specific combination of the relaxation timescales of the two processes of steric hindrance and eventual network relaxation . at sufficiently small @xmath39 , hydrodynamics makes that the relaxation becomes so slow that the second relaxation step is no longer visible out and hence the logarithmic @xmath75dependence is no longer observed . finally we mention that the logarithmic shape in the relaxation function as discussed here is not related to any underlying higher order mode - coupling transition , in contrast to the case of certain colloidal systems for which similar relaxation functions have been observed @xcite . if we set @xmath14 to a small value , this three - step relaxation can no longer be observed , as is seen for the case of @xmath70 in the bottom panel of fig . [ r63pv61fsqall ] : for all @xmath39 s the curve show a ( seemingly ) simple two step relaxation , since the third step ( related to the bonds ) starts already when the second step ( related to steric hindrance ) is not yet completed and hence the two processes become completely mixed in time . below we will briefly come back to this effect . a useful way to check the influence of the polymer bonds on the relaxation dynamics is by comparing the relaxation functions for the case when the bonds are permanent , i.e. @xmath76 , to those when the lifetime is finite . in fig . [ fsq6t100tinf ] , we do this comparison for different connectivities ( @xmath77 and 8) and different values of @xmath10 . in fig . [ fsq6t100tinf]a , we show @xmath78 for the case of @xmath61 with the two different bond lifetimes @xmath28 and @xmath76 . for @xmath79 0.60 , and 0.65 , the timescales for overcoming the steric hindrance are the same for both lifetimes . however , for all @xmath10 , we find that the correlation function for @xmath76 shows a plateau at long times ( not visible in this plot ) , which is due to the fact that the frozen bonds in the percolating gel - network prevent the complete relaxation of the system . in contrast the curves for the finite @xmath14 vanish at long times . for small and intermediate @xmath10 the two sets of curves are very similar , indicating that the presence of a few bonds does not change the dynamics significantly . only for @xmath80 one sees a substantial difference in that the correlator for the permanent links decays slower than the one with @xmath28 . it is reasonable that these differences are noticeable for times somewhat longer than @xmath81 , i.e. the time scale of @xmath14 . for different @xmath10 and @xmath77 and 8 when the bond lifetime is @xmath82 ( lines with symbols ) and when @xmath76 ( lines ) . the wave - vector is @xmath83,title="fig:",width=302 ] for different @xmath10 and @xmath77 and 8 when the bond lifetime is @xmath82 ( lines with symbols ) and when @xmath76 ( lines ) . the wave - vector is @xmath83,title="fig:",width=302 ] for different @xmath10 and @xmath77 and 8 when the bond lifetime is @xmath82 ( lines with symbols ) and when @xmath76 ( lines ) . the wave - vector is @xmath83,title="fig:",width=302 ] if we increase the connectivity to @xmath84 ( fig . [ fsq6t100tinf]b ) , we see that the behavior is qualitatively similar to @xmath61 in that for all values of @xmath10 the two curves track each other up to times around @xmath81 , i.e. the time of the finite @xmath14 . for larger times the correlators for @xmath28 decay to zero whereas the ones for @xmath76 show at long times a marked plateau . the height of this plateau depends now more strongly on @xmath10 than it was the case for @xmath61 , showing that if @xmath11 is increased the life time of the bonds becomes more influential . this is reasonable since it is related to the general observation that in glass - forming system small changes influence the relaxation dynamics increasingly more the slower the dynamics is . we also note that for @xmath33 the correlator for the permanent bonds becomes very stretched . this sluggish relaxation might be related to the fact that for this value of @xmath11 there are , in addition to the percolating cluster , clusters of different sizes ( see ref . @xcite for typical distributions ) , thus giving rise to relaxation dynamics that spans many orders of magnitude in time and hence to a very stretched average correlation function . the stretching of the correlator for the frozen bonds could , however , also be due to the fact that these different clusters hinder each other resulting in the overall slowdown of the dynamics @xcite . next , we increase the number of bonds even more , viz . @xmath85 , as shown in fig . [ fsq6t100tinf]c . we see that for the case of permanent bonds the height of the asymptotic plateau has increased strongly in comparison to the case of @xmath84 . as a result the correlator for @xmath33 shows only a negligible decay of the correlation function if the bonds are permanent . the motion is so constrained by these bonds that the height of the asymptotic plateau , caused by the permanent bonds , becomes comparable to the one related to the steric hindrance . also for @xmath86 the height of the second plateau has increased so much that the relaxation from caging is now completely masked . however , for the intermediate values of @xmath10 , one does notice a difference between the two plateau heights and the correlations functions decay in a very stretched fashion from one to the other . in fact , the decay is so slow that the time - dependence is seen to be logarithmic ( nearly for five decades in the case of @xmath87 ) . note that this logarithmic decay is due to the heterogeneous relaxation of the floppy clusters of frozen bonds , which we will discuss later in further detail . thus this mechanism is different from the one leading to the logarithmic relaxation seen in fig . [ fsq6t100 ] which was due to a unique combination of the two _ finite _ relaxation timescales . with wave - vector @xmath39 for @xmath88 and 4 at @xmath89 with @xmath76 . ( c ) : height of plateau , @xmath90 , of @xmath51 at long times as a function of @xmath39 for @xmath88 , 4 , and 6 . the dashed lines are fits of the form @xmath91 with the corresponding @xmath92 , 0.99 , and 2.27 . ( d),(e ) : @xmath90 vs. @xmath39 for different values of @xmath10 for @xmath61 and 6 . , title="fig:",width=264 ] with wave - vector @xmath39 for @xmath88 and 4 at @xmath89 with @xmath76 . ( c ) : height of plateau , @xmath90 , of @xmath51 at long times as a function of @xmath39 for @xmath88 , 4 , and 6 . the dashed lines are fits of the form @xmath91 with the corresponding @xmath92 , 0.99 , and 2.27 . ( d),(e ) : @xmath90 vs. @xmath39 for different values of @xmath10 for @xmath61 and 6 . , title="fig:",width=264 ] with wave - vector @xmath39 for @xmath88 and 4 at @xmath89 with @xmath76 . ( c ) : height of plateau , @xmath90 , of @xmath51 at long times as a function of @xmath39 for @xmath88 , 4 , and 6 . the dashed lines are fits of the form @xmath91 with the corresponding @xmath92 , 0.99 , and 2.27 . ( d),(e ) : @xmath90 vs. @xmath39 for different values of @xmath10 for @xmath61 and 6 . , title="fig:",width=302 ] we now study the floppiness of this network of particles connected by the permanent bonds by probing the wave - vector dependence of the relaxation functions @xmath51 . in figs . [ ap1]a and b , we show for @xmath89 the variation of @xmath51 for @xmath88 and 4 , i.e. in the region of the phase diagram where gelation sets in . the height of the plateau at long times , also called non - ergodicity parameter , is a measure for the stiffness of the network on the length scale @xmath39 considered . comparing the two panels we recognize that , for a given @xmath39 , the height of the plateau increases with increasing @xmath11 . denoting this height by @xmath93 , we show in fig . [ ap1]c that @xmath90 shows basically an exponential decrease in @xmath39 with a slope that decreases rapidly with increasing @xmath11 . that @xmath90 decreases with increasing @xmath39 is of course reasonable since on small length scales the particles have more leeway to flop around than on large length scales . note , however , that this exponential dependence is in contrast to the one found for the height of the plateau due to the steric hindrance , the latter being basically a gaussian function @xcite . since in the representation of fig . [ ap1]c such a gaussian dependence is given by a parabola , we see that such a curve will intersect the one for @xmath90 at a certain value of @xmath94 . for @xmath95 the plateau due to the steric hindrance is above @xmath90 , thus making that one observes two plateaus in the correlator . however , for @xmath96 the plateau at long times is higher than the steric one , thus making that the latter one will be completely masked by the former and thus the correlator will show only one plateau . we have also studied how the @xmath97dependence of @xmath90 changes with the volume fraction and in figs . [ ap1]d and e , we show @xmath90 for @xmath61 and 6 , respectively . for both cases we see that the rigidity of the network at large scales , i.e. small @xmath39 , is not affected by the volume fraction which is not surprising . for @xmath61 we see that this is also true at small length scales whereas for @xmath84 we note a significant @xmath98dependence if @xmath39 is large . this difference is likely related to the fact that the system with @xmath84 is much more sluggish than the one for @xmath61 , see fig.[fsq6t100tinf ] , and hence small changes ( here in @xmath10 ) will have a stronger impact on the dynamics . finally we disentangle the dynamics of the particles that belong to the percolating cluster from the one that are not attached to it . in the following discussion these are referred as clustered and non - clustered particles , respectively . the objective is to clarify the respective contributions to the different dynamical quantities that we have discussed above . we do this comparison for an increasing number of connections @xmath11 at a fixed ( large ) volume fraction of @xmath87 . in fig . [ figclustfree ] , we show the data for @xmath50 and @xmath51 for these two families of particles . ) , at @xmath61 , 6 , and 8 for volume fraction @xmath87 . ( a ) : mean squared displacement @xmath50 . ( b ) self intermediate scattering function @xmath51 for @xmath52.,title="fig:",width=302 ] ) , at @xmath61 , 6 , and 8 for volume fraction @xmath87 . ( a ) : mean squared displacement @xmath50 . ( b ) self intermediate scattering function @xmath51 for @xmath52.,title="fig:",width=302 ] the mean squared displacement of the clustered particles shows after the ballistic regime at short times a shoulder that is related to the cage of the steric hindrance ( fig . [ figclustfree]a ) . this localization is , however , only temporary and is followed by a further increase of @xmath50 . only at longer times @xmath50 saturates at a height that depends on @xmath11 . we note that the approach to this asymptotic height becomes increasingly slow with increasing @xmath11 and in fact for @xmath85 the time dependence is close to logarithmic and does not end within the time window of our simulation . this behavior is also seen in the self intermediate scattering function @xmath51 ( fig . [ figclustfree]b ) . at short times the correlator decays quickly onto a plateau ( not very pronounced ) before the relaxation of the steric hindrance starts . for @xmath61 and 6 , this process ends in that the correlator reaches the final plateau ( which is given by @xmath90 discussed above ) . however , for @xmath85 the final decay is so slow , again compatible with a logarithmic time dependence , that we do not see the asymptotic behavior . finally we look at the motion of the non - clustered particles and compare it with the one for @xmath30 . from fig . [ figclustfree]a we recognize that at this volume fraction also these particles are slowed down by the cage effect in that one sees for all values of @xmath11 a shoulder in @xmath50 at time @xmath99 . for @xmath30 the @xmath50 shows then immediately the diffusive behavior , i.e. it is proportional to @xmath47 . however , if @xmath11 is increased , the @xmath75dependence of @xmath50 for the un - clustered particles follows first the one of the clustered particles . only once the latter starts to reach the plateau discussed above , do the former cross over to the diffusive behavior . hence we can conclude that before this crossover the relaxation dynamics of the two population of particles are strongly coupled . this result is reasonable because in order to move , the un - clustered particles have to explore the holes within the percolating cluster formed by the particles which are permanently linked . also the self intermediate scattering function of the non - clustered particles tracks the one of the clustered particles at short times ( fig . [ figclustfree]b ) . however , once the latter starts to show at long times a plateau that has a significant height , the two correlators differ strongly since the one for the non - clustered particles decays to zero at long times . from the graph we also see that for @xmath85 the correlator is extremely stretched and shows almost a logarithmic @xmath75dependence . this very slow decay indicates that the mobile clusters can move around the percolating cluster only with great difficulty , a behavior that is similar to the relaxation dynamics of particles moving in random porous media @xcite . it is also interesting that for the highest @xmath11 the mean squared displacement shows for the last two decades in time a nice diffusive behavior , whereas @xmath51 is far from having decayed to zero . this apparent contradiction is related to the fact that @xmath50 is dominated by the particles that move relatively fast ( i.e. they are in the small clusters ) whereas @xmath51 is dominated by the slowly moving particles ( i.e. the large clusters ) . this cluster - size dependent dynamics leads to a so - called dynamical heterogeneity and in section [ sec_vi ] we will discuss this phenomenon in more detail . although we show in fig . [ figclustfree ] the comparison between the dynamics of clustered and non - clustered particles for the case that @xmath100 is infinitely large , it is evident that for a very large but _ finite _ value of @xmath14 the relaxation dynamics will be very similar . hence if , e.g. , @xmath14 is on the order of @xmath101 , basically none of the shown curves will change significantly and thus the conclusions drawn from fig . [ figclustfree ] will be apply also for such value of @xmath14 . we now investigate how the two different relaxation timescales , one due to breaking of local cages and the other due to the reconfiguration of the network - bonds , vary with the volume fraction @xmath10 and the connectivity @xmath11 . to start we consider the case of structural relaxation related to the steric hindrance . in order to avoid that this relaxation process is influenced by the one of the network , we consider the case in which the latter is completely suppressed , which can be achieved by choosing @xmath76 . in the following we will study the relaxation times associated with the intermediate scattering function for wave - vector @xmath52 . as discussed above , this correlator shows at long times an asymptotic plateau the height of which , @xmath90 , depends on @xmath11 and @xmath10 . to take this into account we define the relaxation time @xmath102 , to be the time at which @xmath103 . the evolution of @xmath104 with @xmath10 is shown in fig . [ figrelaxtime]a , for different value of @xmath11 . we see that in the absence of any bonds , i.e. for @xmath30 , we have the usual slowing down of dynamics with increasing @xmath10 . the @xmath98dependence of the relaxation time can be fitted by a vogel - fulcher - tammann - law of the form @xmath105}$ ] , with @xmath106 , from which we can estimate the volume fraction @xmath107 at which the relaxation times would diverge . if we increase the number of bonds among the particles , we see that @xmath104 increases . that this increase is not just a constant ( @xmath57dependent ) factor but depends also on @xmath10 is demonstrated in the inset of the figure where we have normalized the relaxation times to its value at @xmath86 . using the vogel - fulcher - tammann - law we can thus extract from these data the @xmath57dependence of @xmath107 , which can be considered as a proxy for the @xmath57dependence of the glass transition temperature . this @xmath108 line is included in fig . [ phasediagram ] as well and we see that it has a weak negative slope with @xmath109 and @xmath110 . thus we can conclude that the glass transition as induced by the steric hindrance mechanism does not depend strongly on the value of @xmath11 . however , since the prefactor of the vogel - fulcher - law does depend strongly on @xmath11 , we can conclude that a line of iso - relaxation time will bend significantly more . vs. @xmath10 for different values of @xmath11 . inset : same data scaled by @xmath111 . ( b ) main panel : @xmath112 vs. @xmath11 , for different values of @xmath10 . inset : same data as in the main panel but normalized by the relaxation time for @xmath30 , leading to a master curve . note the double logarithmic scale in this plot.,title="fig:",width=264 ] vs. @xmath10 for different values of @xmath11 . inset : same data scaled by @xmath111 . ( b ) main panel : @xmath112 vs. @xmath11 , for different values of @xmath10 . inset : same data as in the main panel but normalized by the relaxation time for @xmath30 , leading to a master curve . note the double logarithmic scale in this plot.,title="fig:",width=272 ] next , we investigate how the timescale for the full relaxation of the network of particles depends on the average number of bonds , @xmath11 . for this we define a relaxation time using the @xmath75dependence of @xmath78 with a very short lifetime for the bonds ( @xmath113 ) . @xmath112 is then defined via @xmath114 and the @xmath57dependence of this relaxation time is shown in fig . [ figrelaxtime]b . the figure shows that for small and intermediate values of @xmath11 the @xmath11-dependence is independent of the volume fraction in that the curves for the different values of @xmath10 seem to be just shifted vertically . that this is indeed the case is demonstrated in the inset where we show that a plot of the same data , but now normalized to the relaxation time for @xmath30 , gives a master curve . the main panel shows that at small concentration of bonds the @xmath57dependence is close to an exponential , a result that is likely related to the fact that increasing @xmath11 leads to a tighter cage for the steric hindrance and hence a slower dynamics . however , for intermediate values of @xmath11 the curve @xmath112 start to bend over towards a weaker @xmath57dependence . in order to investigate this effect better we have carried out simulations for @xmath36 at very high values of @xmath11 : 16 , 30 , and 50 . we find that in this regime the relaxation time follows closely an exponential ( see main fig . [ figrelaxtime]b ) , but with an exponential scale that is smaller than the one seen at small @xmath11 . note that this dependence implies that there is no singularity in the relaxation dynamics at any finite value of @xmath11 , at least for this volume fraction . instead the dynamics shows a behavior that is similar to an arrhenius law in that the barrier for the relaxation depends only on the number of bonds between the particles . this observation is in agreement with earlier simulations for equilibrium gels @xcite . @xmath51 vs. @xmath47 for @xmath115 , and different @xmath14 . the inset shows @xmath50 for the same parameters . the dotted lines in the main figure mark the plateaus in @xmath51 : @xmath116 and @xmath117 , which respectively correspond to the localization lengths @xmath118 and @xmath119 indicated in the inset.,width=302 ] we conclude this discussion by using the different relaxation timescales to develop a criterion that tells whether or not the correlator @xmath51 will show the logarithmic @xmath75dependence discussed in the context of fig . [ fsq6t100]b . for this we define @xmath120 as the height of the plateau associated to the @xmath121-process , with @xmath122 , and recall that @xmath123 , @xmath104 and @xmath124 are the timescales for the three different relaxation processes described above , namely the rattling inside the cage , the escape from the local cage , and the network renewal process . we can now define a simple criterion for the anomalous logarithmic relaxation to be observed by requiring that the slope of the two segments defined by pairs @xmath125 $ ] and @xmath126 $ ] is the same , see fig . [ decon ] . it is known that for glass - forming systems the plateau related to steric hindrance is a gaussian functions of the wave - vector @xmath39 , @xmath127 @xcite . we showed earlier , see fig . [ ap1]c , that for the pn - process , @xmath90 has an exponential shape at large @xmath39 ( the regime corresponding to anomalous logarithmic behavior ) . putting these elements together we thus obtain @xmath128 \ , . \label{eq1}\ ] ] this equation relates a purely structural observable which depends on the competition between two localization lengthscales with dynamical information encoded in the different relaxation times , predicting a precise connection between structure and dynamics whenever anomalous logarithmic relaxation is to be expected . typical glass formers show a heterogeneous dynamics of the particles when the system is increasingly supercooled . it reflects the broad distribution in the timescales for local structural relaxations @xcite . on the other hand , we have reported earlier @xcite that also the dynamics in the gel phase can be heterogeneous : the particles in the percolating cluster and those that are unattached have different nobilities till the timescales at which the bonds in the network are reconfigured . in the following we will thus explore how these two different sources of heterogeneity interact as we increase the density of particles as well as the number of connectivities in the system . we use two different measures of dynamical heterogeneity . the first one is the non - gaussian parameter @xmath129 , defined as @xmath130 , where @xmath131 and @xmath132 are the second and fourth moments of the distribution function of single particle displacements @xmath133 , i.e. of the self part of the van hove function . a non - zero value of @xmath129 quantifies the extent of deviation from a gaussian shape for @xmath134 @xcite . note that @xmath134 is gaussian at short times , i.e. in the ballistic regime , and again at long times when the particles are diffusive . for different values of @xmath11 ( as marked in the graphs ) for @xmath28 . ( a ) : @xmath36 in double - logarithmic representation and ( b ) : @xmath33.,title="fig:",width=302 ] for different values of @xmath11 ( as marked in the graphs ) for @xmath28 . ( a ) : @xmath36 in double - logarithmic representation and ( b ) : @xmath33.,title="fig:",width=302 ] the second measure is the dynamic susceptibility @xmath135 , computed via the fluctuations of time - correlation function : @xmath136 $ ] . it is designed to capture the spatiotemporal correlations of particle nobilities and provides , from its peak value , an estimate of the dynamic correlations @xcite . such functions have also been analyzed for gels with permanent bonds @xcite or low density gels @xcite . we begin by looking at how the non - gaussian parameter varies with changing connectivities @xmath11 , either when we go from the liquid to the gel phase , at @xmath36 , or when we are in a strongly glassy region , @xmath33 . for the case @xmath28 the data is shown in fig . [ alpha2phi ] . for @xmath36 , fig . [ alpha2phi]a , @xmath129 shows at @xmath137 a peak if @xmath30 and a shoulder for @xmath54 . this feature is thus related to the dynamical heterogeneity due to the steric hindrance mechanism , as in usual glass - forming systems @xcite . for @xmath54 we find in addition a very prominent peak the location of which is basically independent of @xmath11 , which shows that for this packing fraction the time at which the system is maximally heterogeneous does not depend on @xmath11 . this observation is in tune with our earlier discussion that the timescale for structural relaxation at small values of @xmath10 does not change with @xmath11 if the number of bonds is not very large ( see fig . [ fsq6t100]a ) . the maximum occurs around @xmath138 , a time which corresponds to a timescale for which a significant number of particles have broken the bonds with their neighbors and exit the constraints of the network . the location of the peak is thus somewhat larger than @xmath14 . as has been documented in ref . @xcite , at small packing fractions one has on this timescale two families of particles , one for the mobile particles and the other related to those that are immobile and as a consequence the shape of @xmath134 deviates strongly from a gaussian . figure [ alpha2phi]a indicates thus that the same behavior persists to the larger volume fraction of @xmath139 . one interesting feature is that the peak height is non - monotonous in @xmath11 in that it increases till @xmath84 and then decreases again for larger @xmath11 values . the reason for the growth is that an increasing @xmath11 allows for more diverse values of the connectivity for the particles , and hence to a stronger variety in the dynamical behavior . on the other hand , if @xmath11 is very large most of the particles are strongly connected all the times and hence show a much smaller variation in their relaxation dynamics , i.e. they behave like in a mean - field like regime . next we study the behavior for @xmath33 , fig . [ alpha2phi]b . when there are no bonds , @xmath30 , we see that @xmath129 is peaked at around @xmath140 . this peak , which is related to the steric hindrance , corresponds thus to the one seen for @xmath36 at @xmath99 and which , due to the higher density has shifted to larger times . for @xmath88 ( when the percolating cluster of the bonds develops ) , the location of this peak slightly shifts to larger times and one sees the appearance of a second peak at @xmath141 ( corresponding to the timescales for renewal of the network connectivities ) . however , the dominant heterogeneity is still due to the local steric hindrances . as @xmath11 is increased , the location of the first peak continues to shift to longer timescales , in track with the increasing structural relaxation timescales ( see fig . [ fsq6t100]c ) , and also its height increases , in qualitative agreement with the behavior found in simple glass - formers if the coupling is increased @xcite . for @xmath84 , the two peaks have merged , since the timescales for the two sources of heterogeneity are nearly the same , and thus @xmath129 has a single peak . if @xmath11 is increased even more , the position of the peak moves to larger times , but its height starts to decrease . the reason for this decrease is likely the same as the one we indicated when we discussed the data for @xmath36 , i.e. that for large @xmath11 the system starts to become mean - field like and hence heterogeneities are suppressed . we also note that the height of the peak is significantly smaller than the one for @xmath36 . thus , at large density , steric hindrance dominates and even the faster particles have less space to move around resulting in significantly less non - gaussian shapes for @xmath134 . as a consequence , the dynamics is more homogeneous in this regime compared to the gel at lower @xmath10 . and @xmath33 , parameters for which @xmath129 shows two peaks ( see fig . [ alpha2phi]b).,width=302 ] to elucidate the origin of the two peaks in the non - gaussian parameter at @xmath142 , we determine how the distribution of particle displacements @xmath134 evolves with time . this is shown in fig . [ vanhover4pv70 ] . at short times , particle motion is restricted by the local cage of neighbors and thus @xmath134 is a gaussian at small @xmath143 . at larger distances the distribution has an exponential tail which is due to a few particles that have escaped the steric hindrance cage , in agreement with the usual heterogeneous glassy dynamics in supercooled systems @xcite . the presence of these two processes gives rise to the maximum in @xmath144 at short times . at around @xmath145 , most of the particles have escaped from this steric hindrance cage and thus the dynamics of the system becomes more homogeneous with @xmath134 assuming a gaussian form and hence @xmath144 decreases again . but with time , the motion of the particles is again restricted , this time by the bonds which have not yet relaxed . thus again @xmath134 develops a gaussian shape ( with a larger width , determined by the length of the connecting bonds ) and an exponential tail , implying an increase in @xmath144 . later , the network eventually relaxes , the particles are diffusive and thus @xmath144 decreases again . for different values of @xmath14 at @xmath146 . inset : for the case of fixed bonds ( @xmath76 ) , @xmath144 for the full system ( maroon ) along with those for the clustered ( green open symbols ) and non - clustered particles ( orange filled symbols ) . , width=302 ] the presence of these two different contributions to the shape of @xmath129 can be further clarified by varying the lifetime of the network , i.e. @xmath14 , and in fig . [ alpha2taulink ] we show this for the case @xmath147 . since we scan a large span of timescales and @xmath144 changes strongly , the data is shown in logarithmic scales . for @xmath31 , we see only a small peak at @xmath148 , which corresponds to the one due to the steric hindrance . for @xmath28 , we find that this peak has increased a bit in height and its location has also shifted to @xmath149 . this change is a consequence of the increased effective coupling between the particles due to the increased lifetime of the bonds . in addition , we see the development of a new ( weak ) peak at @xmath138 , which is caused by the network renewal process . now , if we increase @xmath100 further , the location of the first peak remains at the same place , since the local crowding effects are unaffected by the network dynamics . in contrast to this , the location of the second peak shifts to longer timescales and its height increases strongly . since the position of this peak scales with @xmath14 , we recognize that this peak is related to the renewal process . finally , if we take the case of permanent bonds between the particles , the curve is nearly identical to that for @xmath150 , except that we see no signature of it eventually decreasing with time . the reason for this runaway effects is the fact that with increasing lifetime of the network , the unattached particles diffuse away and travel long distances before the network is again renewed . this results in very long extended tails in @xmath134 which shows up as very large values for the non gaussian parameter . for the case that @xmath14 diverges the dynamics never becomes gaussian and @xmath129 diverges at long times . it is interesting that if one computes , for @xmath76 , @xmath129 separately for the clustered and non - clustered particles , only the peak at short times is observed , i.e. the one which is due to the crowding effects . this is shown in the inset of fig . [ alpha2taulink ] for the case of the system of particles where there is a permanently frozen percolating cluster and few unattached particles . in this case , the @xmath129 for the clustered particles follows the curve for the full system , shows the bump at short times and becomes at long times a constant ( see the corresponding data for @xmath50 in fig . [ figclustfree ] ) . the result that at long times @xmath129 does not go to zero , as it would be the case for vibrations in a typical amorphous solid with @xmath30 , is related to the fact that the spanning cluster is a disordered network of particles that have different local environments . thus the large variety of slow floppy motions , related to the slow relaxation in @xmath51 over long times ( see fig . [ figclustfree ] ) , gives a non - gaussian shape for @xmath134 even at very long times . this non - gaussianity is also the reason for the exponential shape of @xmath151 , as observed earlier , see fig . [ ap1 ] . for the non - clustered particles , @xmath129 does show a slight decrease beyond the short - time maximum which shows that the relaxation dynamics starts to become a bit more homogeneous . however , it can not be expected that the non - gaussian parameter will go to zero even at very long times , since the mobile clusters do have a spread in size and hence a different diffusion constant . for @xmath82 . ( a ) @xmath30 and different values of @xmath10 . ( b ) @xmath33 and different values of @xmath11 ( as marked on plot ) ( c ) @xmath152 for different values of @xmath14 ( as marked on plot).,title="fig:",width=302 ] for @xmath82 . ( a ) @xmath30 and different values of @xmath10 . ( b ) @xmath33 and different values of @xmath11 ( as marked on plot ) ( c ) @xmath152 for different values of @xmath14 ( as marked on plot).,title="fig:",width=302 ] for @xmath82 . ( a ) @xmath30 and different values of @xmath10 . ( b ) @xmath33 and different values of @xmath11 ( as marked on plot ) ( c ) @xmath152 for different values of @xmath14 ( as marked on plot).,title="fig:",width=302 ] we conclude by measuring to what extent the observed heterogeneous dynamics is related to a correlated motion of the particles . this can be quantified by @xmath135 , defined above , measured for @xmath153 , i.e. we look for relaxations over distances which are slightly larger than the average particle diameter . to have a reference we consider first the case @xmath30 , i.e. when there are no polymers and the system is just a standard glass - forming system . the time dependence of @xmath135 is shown in fig . [ chi4]a for different packing fractions . in agreement with earlier studies on similar systems , ref . @xcite , we find that @xmath135 shows a maximum at a time @xmath47 that increases with @xmath10 and which tracks the increasing relaxation time @xmath104 . the fact that the height of the peak increases with increasing @xmath10 indicates that the relaxation dynamics become more cooperative , also this in agreement with previous studies for other glass - forming systems @xcite . next , we contrast this behavior with the case of increasing gelation , i.e. we fix a volume fraction and increase the number of polymer bonds . for this , we choose @xmath33 , i.e. the same packing fraction for which we have studied the correlation functions and the non - gaussian parameter , since for this @xmath10 the effects due to the steric hindrance as well as the network constraints are clearly observed . the corresponding @xmath135 is shown in fig . [ chi4]b for @xmath28 . for @xmath154 , the location of the peak in @xmath135 shifts to larger times and the height of the maximum increases . thus this trend follows the behavior observed in the data for @xmath51 as well as @xmath129 , indicating that steric hindrance dominates the relaxation process and that the dynamics becomes more and more collective . however , for @xmath67 , the peak height decreases again and its location becomes independent of @xmath11 . we can thus conclude that in this regime the relaxation dynamics becomes less cooperative , a result that makes sense since the rewiring of the network is not really a collective process . this result is also in qualitative agreement with the decrease of the maximum in @xmath129 , see fig . [ alpha2phi]b , which showed that the dynamics becomes more homogeneous . we conclude by investigating how the lifetime of the network - bonds influences the function @xmath135 . for this we have calculated this observable for the case @xmath87 and @xmath84 , i.e. at intermediate density and in the gel phase for which one sees a significant @xmath14-dependence in the relaxation behavior ( see fig . [ fsq6t100tinf]b ) . we see , fig . [ chi4]c , that at small and intermediate @xmath14 the location of the peak in @xmath135 moves to larger time and that its height increases somewhat , i.e. a behavior that is directly related to the steric hindrance mechanism in which the effective cage becomes stiffer due to the increased lifetime of the bonds . however , once @xmath14 exceeds @xmath43 , we see that @xmath135 no longer depends on this lifetime , i.e. the rewiring of the network no longer affects the cooperativity of the dynamics and the latter is solely dependent on the steric hindrance . in this paper , we have studied a coarse - grained model for a transient network fluid , a system that can be realized in experiments by a oil - in - water emulsion . by tuning the volume fraction @xmath10 of the constituent particles , the number of bonds @xmath11 between the mesoparticle , and the lifetime @xmath14 of these connections , we have scanned across the phase diagram to investigate the relaxation dynamics of this system , in particular the interplay between the gel - transition and the glass transition induced by steric hindrance effects . by analyzing the mean - squared displacement and self intermediate scattering function we find that the nature of the slowing down depends on the packing fraction : at low @xmath10 an increase of @xmath11 leads to a two step relaxation with a plateau that increases continuously with @xmath11 and an @xmath155relaxation time that is independent of @xmath11 and that is related to the lifetime of the bonds . in contrast to this we find at large @xmath10 a two step relaxation with a plateau height that is basically independent of @xmath11 whereas the relaxation time does depend on the density of bonds . thus these two different behaviors show that the system can show a glassy dynamics that is related on one hand to a gel transition and on the other hand to a glass transition associated with the steric hindrance mechanism . for intermediate densities and a certain range of @xmath11 and @xmath14 the interplay between these two mechanisms leads to a decay of the time correlation function that is logarithmic over several decades in time , whereas for other combinations it gives rise to a three step relaxation . we note here that a similar multistep relaxation pattern has been recently reported for mixtures of _ multiarm _ telechelic polymers and oil - in - water microemulsions @xcite , suggesting the existence of a competition of different arrest length- and time - scales in these systems . we have also studied how the height of the plateau at long times , i.e. the debye - waller factor , depends on the wave - vector and found that it decays in an exponential manner in @xmath39 , i.e. very different from the gaussian decay observed in more standard glass - forming systems . furthermore we have determined the relaxation times of the system and find that these can approximately be factorized into a function that depends strongly on @xmath11 and a vogel - fulcher type dependence on @xmath10 . the @xmath57dependent factor shows for @xmath156 a strong exponential dependence that is related to the escape of the particles from the local cage , whereas for larger values of @xmath11 the dependence is weaker and linked to an arrhenius - like process for bond - breaking . by studying the non - gaussian parameter we have probed to what extent the relaxation dynamics of the system is heterogeneous . at low packing fraction this dynamics becomes extremely heterogeneous , if @xmath11 is not too large , since some of the particles are very strongly connected to their neighbors ( and hence are immobile ) whereas others can move almost freely . however , if @xmath11 becomes larger than 6 , the relaxation dynamics becomes again quite homogeneous since at any time all particles are well connected to their neighbors . at large packing fractions the non - gaussian parameter shows a double peak structure and , by monitoring the van hove function , we can show that this feature is directly related to the two relaxation processes , i.e. the steric hindrance and the connectivity of the network . whether or not the relaxation dynamics is cooperative can be characterized by the four - point correlation function @xmath135 . we find that for large packing fractions the height of the peak in @xmath135 increases rapidly with @xmath11 , showing that the strengthened coupling leads to an enhanced cooperative motion . however , if @xmath11 is increased beyond 6 , this cooperativity decreases again , since the relaxation dynamics is strongly dominated by the rewiring of the network , i.e. a non - cooperative process . summarizing we see that the interplay between the two different mechanisms giving rise to glassy behavior can lead a rather complex and unusual relaxation dynamics . in the present study we have focused on a system in which the particles are connected by polymers having a fixed extension length @xmath157 . for much smaller extension lengths , one can expect to recover the re - entrant scenario observed in colloidal gels @xcite . for longer extension lengths , the particles will instead have more space to explore and it will likely result in a even more floppy network . in the future , it would certainly be worthwhile to explore also systems that have polymers with different value of @xmath8 , since this implies multiple localization lengths and thus provide novel relaxation scenarios that can subsequently also studied in experiments . and finally we recall that our model is motivated by an experimental system which has highly nontrivial rheological properties , e.g. this material flows like a liquid but eventually breaks as a brittle solid @xcite , being capable of self - healing through thermal fluctuations . thus , in the future we plan to study the rheological properties of our model in order to understand the microscopic mechanisms that lead to the experimental observations . financial support from anr s tsanet is acknowledged . pih acknowledges financial support from spanish micinn project fis2009 - 08451 , fis2013 - 43201-p , university of granada , junta de andaluca projects p06-fqm1505 , p09-fqm4682 and genil pyr-2014 - 13 project . wk is member of the institut universitaire de france .

we analyze the long time behavior of initial value problems that model a process where particles of type and diffuse in some substratum and react according to the case has been studied before ; it presents nontrivial behavior on the reactive scale only . in this paper we discuss in detail the cases and prove that they show nontrivial behavior on the reactive and the diffusive length scale .

math-ph9904013

we consider , for arbitrary but fixed @xmath5 @xmath6 the reaction diffusion problem @xmath7 for @xmath8 @xmath9 with initial conditions @xmath10 @xmath11 satisfying @xmath12 the choice of the initial time @xmath13 and a class of initial conditions @xmath14 @xmath15 will be described later on , but for the purpose of this introduction it is useful to have in mind the `` natural '' case : @xmath16 @xmath17 for @xmath18 @xmath19 for @xmath20 and @xmath21 for @xmath20 @xmath22 for @xmath23 this initial value problem models the time evolution of a chemical system of two ( initially separated ) substances @xmath0 and @xmath24 that diffuse in some substratum and react according to @xmath25 with a substance @xmath26 that is supposed not to participate in the reaction anymore . the model is a mean field description of such a situation where the functions @xmath27 and @xmath28 represent the densities of the substances @xmath0 and @xmath29 for more details see @xcite . equations ( [ system1 ] ) and ( [ system2 ] ) are best studied in terms of the sum @xmath30 and the difference @xmath31 which satisfy the equations @xmath32 with initial conditions @xmath33 and @xmath34 ( at time @xmath35 that satisfy @xmath36 and @xmath37 for initial conditions @xmath38 @xmath15 with @xmath39 the functions @xmath33 and @xmath34 are even and odd , respectively , and the equations ( [ equ ] ) , ( [ eqv ] ) preserve this symmetry . furthermore , for the special initial condition @xmath40 with @xmath41 defined by the equation @xmath42 equation ( [ equ ] ) has the explicit solution @xmath43 we note that the initial condition ( [ initsym ] ) for @xmath44 ( at time @xmath45 ) is simply the solution of equation ( [ equ ] ) with the `` natural '' initial condition , @xmath46 for @xmath18 @xmath47 for @xmath20 evaluated at @xmath48 to keep this paper as simple as possible we now restrict the discussion to this case , i.e. , we consider from now on equation ( [ eqv ] ) with initial conditions satisfying ( [ limitv ] ) , and @xmath44 given by ( [ uxt ] ) . we note , however , that more general ( asymmetric ) initial conditions for @xmath44 could be treated as well . this would lead to corrections to @xmath44 of the order @xmath49 and such corrections do not change in any way the discussion of the equation for @xmath50 that follows . the reaction diffusion problems considered here develop , in addition to the built in diffusive length scale @xmath51 an additional shorter length scale , on which the reaction takes place . the function @xmath52@xmath53 is called the reaction term or reaction front , and we are interested in describing the asymptotic behavior of the function @xmath54 for large times . the knowledge of this behavior is useful , since it appears to be universal , in the sense that it is largely independent of the choice of the initial conditions and of the details of the model under consideration . as mentioned above , if @xmath33 is an even function , then @xmath50 and as a consequence @xmath54 are even functions of @xmath55 we will see that the critical point of @xmath54 at @xmath56 is a maximum , and that @xmath54 decays ( rapidly ) for large @xmath55 before proceeding any further we note that the factor of @xmath57 in ( [ system1 ] ) , ( [ system2 ] ) and ( [ reaction001 ] ) is just a normalization , and has been chosen for convenience to make the equation ( [ eqv ] ) for @xmath50 look simple . in fact , any system of the form @xmath58 with positive @xmath59 @xmath60 @xmath61 and @xmath62 and with initial conditions such that @xmath63 can be reduced , by scaling space and time and the amplitudes , to the problem @xmath64 with @xmath65 and with initial conditions such that @xmath66 in this paper we have limited the discussion to the case @xmath67 and @xmath68 the case @xmath69 leads to a moving reaction front . a change of coordinates to a moving frame complicates the analysis , but the problem could still be treated with the methods presented here . choosing @xmath70 makes the equations mathematically simpler . as a consequence , as we have seen , the two equations for @xmath27 and @xmath28 can be reduced to just one equation for the sum @xmath71 since the equation for the difference @xmath72 can be solved explicitly . even though we do not expect the asymptotic behavior of the solution to change in any relevant way if @xmath73 , the strategy of proof would have to be changed considerably , since the equations can not be decoupled anymore in that case . before we state our results , we briefly discuss the expected dependence of the results on the parameter @xmath74 the case @xmath3 has been studied in detail in @xcite , where it is proved that in this case the reaction term ( [ reaction001 ] ) satisfies , for all @xmath75 @xmath76 where @xmath77 and @xmath78 and where @xmath79 is a smooth function that decays like @xmath80 for large values of @xmath81 it follows furthermore from the results in @xcite that the function @xmath54 is very small on the diffusive scale in the sense that for @xmath82 @xmath83 and all @xmath84@xmath85 the smallness of @xmath54 on the diffusive scale is easily understood by realizing that , for @xmath3 and for positive values of @xmath86 on the diffusive scale , i.e. , for @xmath87 equation ( [ system1 ] ) essentially reduces to @xmath88 with @xmath89 therefore , the function @xmath27 decays exponentially fast to zero on this scale , and similarly for @xmath28 for negative values of @xmath55 for @xmath90 however , equation ( [ system1 ] ) reduces , for @xmath87 essentially to @xmath91 with @xmath89 the solution of ( [ a002 ] ) has an asymptotic behavior that is radically different from the solution of ( [ a001 ] ) . in particular , for @xmath92 , the solution may even blow up in finite time if @xmath27 is not a positive function . note that , for @xmath93 odd , the nonlinear term in ( [ a002 ] ) is always a `` friction term '' , independent of the sign of @xmath94 and the case of @xmath93 odd will therefore turn out to be easier to treat than the case of @xmath93 even . it is well known @xcite that for @xmath95 and small bounded integrable initial conditions , the nonlinearity in ( [ a002 ] ) becomes irrelevant for large times in the sense that the solution converges to a multiple of @xmath96 which solves the linear equation @xmath97 we would therefore expect that , for @xmath4 the function @xmath54 is of the order @xmath98 on the diffusive scale . this turns out to be wrong . as we will prove below , @xmath54 is of the order @xmath99 for @xmath4 because @xmath54 converges on this scale pointwise to a function that is not integrable at the origin . this corresponds to a solution of ( [ a002 ] ) for which the nonlinear term is a marginal perturbation , i.e. , a solution with an amplitude of the order @xmath100 . we will see that one can take advantage of this fact , and a diffusive stability bound will be good enough to prove convergence of @xmath54 to its limit , but as a consequence , our results will be limited to the case @xmath101 the cases @xmath92 and @xmath102 are special and will not be discussed any further . the following theorem is our main result . [ main]for arbitrary but fixed @xmath5 @xmath103 there exist @xmath104 functions @xmath105 @xmath106 @xmath107 @xmath108 and a class of initial conditions ( specified at @xmath109 such that ( [ eqv ] ) has a unique solution @xmath50 that satisfies for all @xmath110 the bound @xmath111 where @xmath112 @xmath113 @xmath114 and @xmath115 this theorem is a local result , in the sense that the class of initial conditions will be a set of functions in a ( small ) neighborhood of the function @xmath116 @xmath117 in particular , our methods do not allow us to show that the solution with the `` natural '' initial condition @xmath118 at @xmath119 belongs to this set at @xmath48 we do expect , however , that this is the case , as has been proved for @xmath3 in @xcite . we note that , if an initial condition @xmath33 is such that @xmath120 for a certain @xmath121 then @xmath122 , if @xmath20 or @xmath123 if @xmath23 a priori , we do not need to consider such initial conditions , since in our model @xmath27 and @xmath28 represent particle densities , and the solutions @xmath27 and @xmath28 are positive if the initial conditions @xmath124 and @xmath15 are positive . as we will see , for @xmath103 it will not be necessary to impose that @xmath124 and @xmath15 be positive everywhere , and it will neither be necessary to impose that @xmath125 be an even function . as we will see , the functions @xmath126 and @xmath127 are small on the diffusive scale , i.e. , for @xmath128 @xmath129 and @xmath130 large , @xmath131 where @xmath132 if @xmath133 and @xmath134 if @xmath135 @xmath136 using the definition ( [ defv ] ) , ( [ defu ] ) for @xmath50 and @xmath137 we therefore find that for @xmath138 and @xmath130 large , @xmath139 and similarly for @xmath140 for @xmath141 in contrast to the case @xmath82 where only exponentially few particles reach the diffusive scale , the amount of particles decays only slowly for @xmath101 our results imply that , for large times , the density of the remaining particles is given by the function @xmath142 i.e. , it is independent of the initial conditions . as a corollary to theorem [ main ] we get a precise description of the reaction front @xmath54 on the reactive and the diffusive scale . this description will be given in section 4 , once we have defined the functions @xmath143 @xmath144 @xmath126 and @xmath127 in section 3 . in section 2 we explain our strategy for proving theorem [ main ] . this strategy is implemented in section 5 and section 6 . the appendix contains the proof of the existence of the functions @xmath145 @xmath126 and @xmath127 . consider functions @xmath50 of the form @xmath146 with @xmath147 as in theorem [ main ] , and @xmath148 for some @xmath149 with @xmath150 substituting ( [ ansatz ] ) into ( [ eqv ] ) leads to an equation for the function @xmath151 of the form @xmath152 for certain functions @xmath153 and @xmath154 and for @xmath155 some nonlinear map . we will show that if @xmath147 is defined correctly , @xmath156 large enough and @xmath157 small enough , then @xmath153 can be chosen positive and @xmath155 will be small , so that the solution of equation ( [ eqpsi ] ) will be bounded for large times by the corresponding solution of the inhomogeneous heat equation @xmath158 we will find that , with the right choice of @xmath159@xmath160 from which the bound ( [ bound ] ) will follow . we note that @xmath161 for @xmath162 so that contributions of initial conditions will become irrelevant for large times , i.e. , the solution @xmath151 becomes what is called `` slaved to the inhomogeneous term '' . in order to implement the strategy outlined in section 2 , we need a function @xmath163 that approximates the solution @xmath50 for large times sufficiently well , uniformly in @xmath55 since we would like to control the time evolution of equation ( [ eqpsi ] ) on @xmath164 this function @xmath163 needs to satisfy @xmath165 in order for @xmath50 to satisfy the boundary conditions ( [ limitv ] ) . furthermore , the inhomogeneous term @xmath166 in equation ( [ eqpsi ] ) contains second derivatives of @xmath167 and the function @xmath166 can therefore only be in @xmath168 if @xmath163 is at least twice differentiable . we now construct a function @xmath163 satisfying these requirements through a two length scale asymptotic expansion . to simplify the notation later on we use the convention that , unless stated otherwise , @xmath169 and @xmath170 and we will refer to @xmath171 as the diffusive length scale and to @xmath172 as the reactive length scale . the function @xmath147 is given by the first and second order terms of a so called `` matched asymptotic expansion '' . the `` matched '' refers to the fact that such an expansion contains functions that can not be associated uniquely with one of the length scales and can therefore be used to `` match '' the behavior at large distances of the shorter scale with the behavior at short distances of the larger scale . let @xmath173 and let @xmath174 @xmath175 and @xmath176 be as in theorem [ main ] . then , the functions @xmath177 @xmath178 @xmath179 and @xmath180 are of this form and will naturally show up in the function @xmath181 as a consequence , the representation ( [ vinfty ] ) for @xmath147 is not unique . if we choose ( as we will ) to compute the expansion for @xmath147 in the order of decreasing amplitudes , i.e. , if we first compute the term of order @xmath182 then the term of order @xmath183 and finally the term of order @xmath184 we get a representation of @xmath147 of the form @xmath185 where @xmath186 with @xmath187 a certain constant to be determined later . we note that , by definition , @xmath147 is a symmetric function , and it is therefore sufficient to consider positive values of @xmath86 if we choose appropriate boundary conditions at @xmath56 to ensure regularity . finally , since we will need to describe the asymptotic behavior of various functions near zero and infinity , we introduce the following notation . let @xmath188 be a continuous function from @xmath189 to @xmath190 @xmath191 a positive integer and @xmath192 real numbers . then , we say that @xmath193 near @xmath194 if @xmath195 and we say that @xmath196 near @xmath197 if @xmath198 to lowest order the function @xmath50 is asymptotic to @xmath200 , with @xmath41 as defined in ( [ defmu ] ) . we note that @xmath41 has near @xmath201 the expansion @xmath202 where @xmath203 and @xmath204 . furthermore , @xmath205 we make the ansatz @xmath207 which we substitute into equation ( [ eqv ] ) . we multiply the resulting equation with @xmath208 and take then the limit @xmath209 keeping @xmath172 fixed . this leads to the differential equation @xmath210 where @xmath172 is now considered a variable in @xmath211 since @xmath212 the correct boundary condition for @xmath213 at infinity is @xmath214 and at @xmath215 we impose @xmath216 which makes the function @xmath217 twice differentiable at @xmath194 since @xmath41 and @xmath218 are twice differentiable at zero and @xmath219 a proof of the following proposition can be found in the appendix . [ theoremeta]for @xmath220 @xmath221 there exists a unique function @xmath222 that satisfies equation ( [ eqeta ] ) with the boundary conditions ( [ etainfinity ] ) and ( [ etazero ] ) . the function @xmath223 is positive , and has near @xmath215 the expansion @xmath224 with positive coefficients @xmath225 @xmath226 and @xmath227 for @xmath172 large , @xmath223 is of the form @xmath228 for a certain constant @xmath229 with @xmath230 as in ( [ defdelta007 ] ) , @xmath231 and @xmath232 we note that @xmath233 the constants @xmath234 @xmath235 @xmath236 and @xmath237 are given in the appendix . the function @xmath238 is defined in ( [ etaphi1 ] ) in terms of @xmath239 we make the ansatz @xmath241 which we substitute into equation ( [ eqv ] ) . we multiply the resulting equation with @xmath242 and take then the limit @xmath209 keeping @xmath171 fixed . since @xmath243 and @xmath244 this leads to the differential equation for the function @xmath245 @xmath246 where @xmath171 is now considered as a variable in @xmath211 at @xmath201 we impose the boundary condition @xmath247 which removes the leading singularity of the function @xmath248 at @xmath249 as we will see , the sub leading singularity is proportional to @xmath250 which is not a twice differentiable function at @xmath201 ( except for @xmath251 where @xmath252 this singularity will be cancelled by imposing appropriate boundary conditions for the function @xmath253 the correct choice of boundary conditions for @xmath254 at infinity is somewhat less obvious . in the appendix we show that the condition @xmath255 is not sufficient to single out a unique function @xmath256 if @xmath254 does converge to zero at infinity , then it is asymptotic to a solution of the equation @xmath257 this linear equation is compatible with a ( very slow ) algebraic decay , @xmath258 or with a modified gaussian decay , @xmath259 with the algebraic decay being the generic case . it will be essential in later sections that @xmath260 decays rapidly at infinity , and we therefore impose the boundary condition @xmath261 a proof of the following proposition can be found in the appendix . [ cmu2]for all @xmath262 there exists a unique positive function @xmath263 that satisfies equation ( [ mu2 ] ) with the boundary conditions ( [ bmu2 - 1 ] ) and ( [ bmu2 - 2 ] ) . for @xmath171 small , the function @xmath254 is of the form @xmath264 with @xmath265 with @xmath266 and with @xmath267 as in proposition [ theoremeta ] . for @xmath171 large , the function @xmath254 decays rapidly in the sense that @xmath268 for some constant @xmath269 we make the ansatz @xmath271 which we substitute into equation ( [ eqv ] ) . we multiply the resulting equation with @xmath272 and take then the limit @xmath273 keeping @xmath274 fixed . this leads to the ( linear ) differential equation for @xmath275 @xmath276 ~. \label{varphi2}\ ] ] in order to compensate the sub leading singular behavior of @xmath277 near @xmath278 we make the ansatz @xmath279 which we substitute into equation ( [ varphi2 ] ) . for the function @xmath280 we get the equation @xmath281 since the function @xmath218 is regular near @xmath282 the solution @xmath280 turns out to be regular near @xmath282 too . therefore , the function @xmath283 is twice differentiable near @xmath56 if we impose at @xmath215 the boundary condition @xmath284 at infinity we need that @xmath285 we therefore require that @xmath286 a proof of the following proposition can be found in the appendix . [ cfi22]for all @xmath262 there exists a unique function @xmath287 that satisfies equation ( [ h ] ) with the boundary conditions ( [ bh-1 ] ) and ( [ bh-2 ] ) . near @xmath282 the function @xmath280 is of the form @xmath288 with certain coefficients @xmath289 and @xmath290 and for @xmath172 large @xmath280 is of the form @xmath291 with @xmath292 as defined in ( [ lambda0 ] ) , for a certain constant @xmath293 , and with @xmath294 as defined in proposition [ theoremeta ] . using the properties of the functions @xmath295 @xmath106 @xmath238 and @xmath108 we get from theorem [ main ] the following behavior of the reaction front @xmath296 let @xmath50 be as in theorem [ main ] , and @xmath54 as defined in ( [ reaction001 ] ) . then , for all @xmath297@xmath298 and for all @xmath129@xmath299 here , @xmath300 @xmath301 are as defined in proposition [ theoremeta ] and @xmath26 is as defined in ( [ asymmu2 ] ) . in order to simplify the notation we define the function @xmath303 @xmath304 the function @xmath305@xmath306 the function @xmath307@xmath308 and the function @xmath309@xmath310 the function @xmath147 in theorem [ main ] and in ( [ repres1 ] ) can then be written as @xmath311 let now @xmath312 then , @xmath313 therefore , substituting the ansatz @xmath314 into ( [ eqv ] ) leads to the following equation for the function @xmath315@xmath316 with the function @xmath317 @xmath318 the function @xmath154@xmath319 and the map @xmath320@xmath321 the function @xmath323 @xmath324 is strictly decreasing on @xmath325 and therefore @xmath326 for @xmath327 furthermore , the functions @xmath213 and @xmath41 are strictly positive and @xmath41 is strictly increasing . these properties imply that , for @xmath328 @xmath329 @xmath330 @xmath331 @xmath332 where @xmath333 next , since the functions @xmath334 and @xmath335 are bounded and since @xmath336 for @xmath337 we have that @xmath338 and as a consequence @xmath339 and @xmath340 are positive functions of @xmath86 for all fixed @xmath341 if @xmath342 large enough . [ propvodd]for @xmath93 odd , @xmath343 there exists @xmath344 such that for all @xmath345 the function @xmath346 is positive . the function @xmath347 is positive , for @xmath93 odd . as a consequence , for @xmath93 odd , equation ( [ equpsi1 ] ) is of the form indicated in section 2 , with @xmath348 and @xmath349 the rest of this section treats the case of @xmath93 even , which , as indicated in the introduction , is slightly more delicate . it can be skipped in a first reading or if the reader is only interested in the case of @xmath93 odd . so let @xmath93 be even . the idea is to split @xmath346 into its positive part @xmath350 @xmath351 and its negative part @xmath352 , and to show that @xmath353 is small enough so that it can be treated together with the nonlinear term . consider the function @xmath354 defined in ( [ defphi ] ) . the problem is that @xmath354 becomes negative for large values of @xmath86 , and that therefore @xmath346 becomes negative for large values of @xmath55 to understand why @xmath354 becomes negative , we note that the leading order term @xmath355 in the large @xmath172 asymptotics of @xmath356 is compensated by the leading order term @xmath357 in the large @xmath171 asymptotics of @xmath358 the leading order of @xmath354 at @xmath86 large is therefore given by the second order term in the large @xmath172 asymptotics of @xmath359 and the leading term in the large @xmath172 asymptotics of @xmath360 the first of these terms is proportional to @xmath361 and the second one is proportional to @xmath362 the corresponding proportionality constants @xmath363 and @xmath364 can be computed for @xmath365 and turn out to be negative . for @xmath366 these constants can not be obtained from asymptotic expansions , but numerical results show that they are in fact also negative in these cases . we do not need a proof of this numerical fact , because the following proposition is also correct for positive @xmath346 . [ propv]for @xmath93 even , @xmath103 there exists @xmath367 , such that the function @xmath368 , satisfies for all @xmath369 the bound @xmath370 the idea is to write @xmath354 as the sum of a function @xmath371 that is positive and a function @xmath372 that absorbs the asymptotic behavior at infinity . since @xmath373 for @xmath171 small , with @xmath374 there exists @xmath375 such that @xmath376 for all @xmath377 let @xmath378 to be chosen below , and let @xmath379 be the heaviside step function , i.e. , @xmath380 for @xmath381 and @xmath382 for @xmath23 then , we define the function @xmath383 by the equation @xmath384 and we set @xmath385 in order to prove that @xmath386 is positive , for @xmath387 large enough , we write @xmath388 where @xmath389 and @xmath390 @xmath391 is positive for @xmath392 since in this case @xmath393 and @xmath394 is positive for @xmath395 by definition of @xmath396 next we consider @xmath397 for @xmath398 we have that @xmath399 . but @xmath400 for all @xmath401 and all @xmath402 if @xmath403 is sufficiently large , since @xmath404 since @xmath127 is bounded , and since @xmath405 for @xmath172 large enough . finally , using the asymptotic properties of @xmath223 and @xmath406 we see that @xmath407 for @xmath408 if @xmath387 is chosen large enough . we now estimate the function @xmath409 from the definition of @xmath410 we get that @xmath411 and therefore , since @xmath371 is positive , we have the lower bound @xmath412 for some constant @xmath413 , from which ( [ v1bound ] ) follows . [ thm : inhomo ] let @xmath166 be as defined in ( [ equi ] ) , and let @xmath414 then , there exists a constant @xmath415 such that for all @xmath416@xmath417 the function @xmath166 is even , and it is therefore sufficient to bound it for @xmath418 the strategy of the proof will be to rewrite the function @xmath166 as a sum of functions of the form @xmath419 with @xmath420 and with @xmath421 and @xmath54 functions with appropriate asymptotic behavior at zero and infinity . each of the terms in the sum can then be estimated with the help of lemma [ lem : estimations ] below . in order to keep the notation as simple as possible , we suppress in what follows the arguments of functions whenever there is no risk of confusion . for @xmath422 the function @xmath423 is of the form @xmath424 where @xmath425 where @xmath426 and where @xmath427 in terms of the functions ( [ t1])([t6 ] ) we get that , for @xmath20 @xmath428 and therefore @xmath429 , where @xmath430 and where @xmath431 @xmath432 and @xmath433 are as defined above . since @xmath434 it remains to be shown that @xmath435 using the differential equations for @xmath145 @xmath223 and @xmath436 we find that @xmath437 where @xmath438 as defined above , where @xmath439 and where @xmath440 \\ & \hspace{3cm}+n(t^{-2\gamma } t_{1})^{n-1}(t^{-4\gamma } t_{3})~.\end{aligned}\ ] ] the functions @xmath441 and @xmath442 can be further decomposed as follows @xmath443 where @xmath444 and @xmath445 are as defined above , and where @xmath446 it remains to be shown that @xmath447 but this follows using the definitions . in order to characterize the behavior of a function near zero and infinity we introduce the following family of vector spaces . let @xmath448 and @xmath449 be two real numbers with @xmath450 then , we define @xmath451 to be the vector space of continuous functions @xmath54 from @xmath452 to @xmath190 for which the norm @xmath453 is finite . note that , if a function is in @xmath454 , then it is also in @xmath455 for any pair of numbers @xmath456 for which@xmath457 furthermore , if @xmath458 is in @xmath459 , and @xmath460 is in @xmath461 , then the product @xmath462 is in @xmath463 . the following lemma provides the tool that we use to estimate the terms on the right hand side of ( [ repi ] ) . [ lem : estimations ] [ lem : bornes ] let @xmath464 and @xmath465 , and assume that @xmath466 and that @xmath467 then , there is a constant @xmath468 such that for all @xmath416 @xmath469 where @xmath470 from ( [ cb3 ] ) it follows that either @xmath471 or @xmath472 in the first case we get using ( [ cb1 ] ) that @xmath473 @xmath474 @xmath475 and therefore @xmath476 and in the second case we get using ( [ cb2 ] ) that @xmath477 and therefore @xmath478 we now show that the right hand side in ( [ repi ] ) can be bounded by a sum of terms of the form @xmath479 for each of these terms we then show that the corresponding functions @xmath421 and @xmath54 satisfy the hypothesis of lemma [ lem : estimations ] . this then implies that @xmath480 @xmath481 for a certain @xmath482 depending on @xmath54 and @xmath483 it is therefore sufficient to prove that @xmath484 for all these terms in order to prove the inequality ( [ eqn : thminhomo ] ) . for @xmath485 @xmath486 we have the bounds @xmath487 and @xmath488 , @xmath489 , @xmath490 , @xmath491 , @xmath492 , @xmath493 , @xmath494 . the inequalities ( [ inequalities ] ) follow by using the triangle inequality and the asymptotic properties of the functions @xmath41 , @xmath495 , @xmath218 and @xmath238 . [ [ bound - on - the - function - a_2 . ] ] bound on the function @xmath496 . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + we have the bound @xmath497 where @xmath498 the function @xmath499 is in @xmath500 , and the function @xmath501 is in @xmath502 @xmath503 . since @xmath504 the inequalities ( [ cb1 ] ) and ( [ cb2 ] ) are satisfied and , since @xmath505 for @xmath506@xmath507 therefore , @xmath508 @xmath509 @xmath510as required . [ [ bound - on - the - function - a_3 . ] ] bound on the function @xmath432 . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + we have that @xmath511 where @xmath512 @xmath513 . since @xmath514 , and @xmath515 we have the bound @xmath516 with @xmath517 the function @xmath518 is in @xmath519 and the function @xmath520 is in @xmath521 . since @xmath522 the inequalities ( [ cb1 ] ) and ( [ cb2 ] ) are satisfied and @xmath523 . therefore @xmath524 @xmath525 @xmath526 as required . similarly , we have that @xmath527 with @xmath528 the function @xmath529 is in @xmath530 and the function @xmath531 is in @xmath532 . the inequalities ( [ cb1 ] ) and ( [ cb2 ] ) are satisfied and @xmath533 @xmath534 therefore , @xmath535 @xmath536 @xmath526 as required . finally , @xmath537 where @xmath538 the function @xmath539 is in @xmath540 , and the function @xmath541 is in @xmath542 . the inequalities ( [ cb1 ] ) and ( [ cb2 ] ) are satisfied and @xmath543 . therefore , @xmath535 @xmath544 @xmath526 as required . [ [ bound - on - the - function - a_4 . ] ] bound on the function @xmath433 . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + since the functions @xmath545 and @xmath546 are bounded , @xmath547 and @xmath548 we have that @xmath549 where @xmath550 the function @xmath539 is in @xmath551 , and the function @xmath552 is in @xmath553 . the inequalities ( [ cb1 ] ) and ( [ cb2 ] ) are satisfied and @xmath554 . therefore , @xmath524 @xmath555 @xmath526 as required . [ [ bound - on - the - function - a_5 . ] ] bound on the function @xmath438 . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + we have the bound @xmath556 where @xmath550 the function @xmath539 is in @xmath551 , and the function @xmath557 is in @xmath558 . the inequalities ( [ cb1 ] ) and ( [ cb2 ] ) are satisfied and @xmath543 . therefore , @xmath524 @xmath559 @xmath560 as required . [ [ bound - on - the - function - a_6 . ] ] bound on the function @xmath561 . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + we have the bound @xmath562 where @xmath563 the function @xmath564 is in @xmath565 and the function @xmath566 is in @xmath567 the inequalities ( [ cb1 ] ) and ( [ cb2 ] ) are satisfied and @xmath568 . therefore , @xmath524 @xmath569 @xmath526 as required . [ [ bound - on - the - function - a_7 . ] ] bound on the function @xmath570 . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + we have the bound @xmath571 where @xmath572 . the function @xmath573 is in @xmath574 and the function @xmath575 is in @xmath576 the inequalities ( [ cb1 ] ) and ( [ cb2 ] ) are satisfied and @xmath577 . therefore , @xmath524 @xmath578 @xmath526 as required . [ [ bound - on - the - function - a_8 . ] ] bound on the function @xmath579 . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + we have the bound @xmath580 where @xmath581 the function @xmath582 is in @xmath583 and the function @xmath584is in @xmath585 the inequalities ( [ cb1 ] ) and ( [ cb2 ] ) are satisfied and @xmath586 . therefore , @xmath587 @xmath588 @xmath589 as required . [ [ bound - on - the - functions - a_pq . ] ] bound on the functions @xmath590 . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + we have the bound @xmath591 where @xmath592 the function @xmath593 is in @xmath594 and the function @xmath595 is in @xmath596 . the inequalities ( [ cb1 ] ) and ( [ cb2 ] ) are satisfied , and @xmath597 therefore , @xmath598 , for @xmath599 and @xmath600 , for @xmath601 as required . this completes the proof of theorem [ thm : inhomo ] . @xmath602 equation ( [ equpsi1 ] ) is of the form ( [ eqpsi ] ) if we define the map @xmath603 by the equation @xmath604 with @xmath605 as defined in ( [ that ] ) and @xmath368 as defined in section [ sectionv ] . using the definitions , we see that @xmath155 can be written as , @xmath606 with @xmath607 let @xmath608 as in ( [ vpq ] ) . then , for all @xmath416 @xmath609 where @xmath610 the case @xmath611 follows from ( [ v1bound ] ) . let now @xmath612 since @xmath613 for all @xmath614 @xmath615 we find that @xmath616 furthermore , since @xmath617 near @xmath618@xmath619 since the function @xmath620 is bounded , it follows that @xmath621 @xmath622 with @xmath623 for @xmath624 we improve this bound using additional properties of the function @xmath625 namely , since @xmath626 we have that @xmath627 for functions @xmath628 in @xmath629 we use the norms @xmath630 , @xmath631 and @xmath632 , and we denote by @xmath633 the banach space of functions @xmath634 in @xmath635 for which the norm @xmath636 , @xmath637 is finite . let @xmath638 as in proposition [ propvodd ] and @xmath639 as in proposition [ propv ] , and consider , for fixed @xmath640 functions @xmath641 of the form @xmath642 with @xmath643 let @xmath644 be the fundamental solution of the differential operator @xmath645 and let , for given @xmath646 , the map @xmath647 be defined by the equation @xmath648 where @xmath649 and where @xmath650 with @xmath651 the integral equation @xmath652 is equivalent to the differential equation ( [ eqpsi ] ) with initial condition @xmath653 we note that , since the function @xmath153 is positive , the kernel @xmath644 is bounded pointwise by the fundamental solution @xmath654 of the heat equation , @xmath655 the following proposition makes theorem [ main ] precise . [ maint]let @xmath656 with @xmath657 as defined in ( [ eqn : thminhomo ] ) , and let @xmath403 be sufficiently large . then , for all @xmath646 with @xmath658 , the equation @xmath652 has a unique solution @xmath659 in the ball @xmath660 the solution of the integral equation will be dominated by @xmath661 and , as we will see , @xmath662 has been chosen such that @xmath663 the idea is therefore to show that , if @xmath403 is large enough to make the nonlinear part of the map @xmath647 small , and if @xmath664 , then the map @xmath647 contracts the ball @xmath665 into itself , which by the contraction mapping principle implies the theorem . we first show that @xmath647 maps the ball @xmath666 into itself . for the contribution coming from the initial condition we have @xmath667 and therefore @xmath668 we next estimate the norm @xmath669 let @xmath670 then , @xmath671 and therefore @xmath672 it remains to be shown that the nonlinearity is also bounded by @xmath673 for @xmath403 large enough . for @xmath674 we have , @xmath675 for @xmath611 we get , since @xmath676 @xmath677 and for @xmath624 we get @xmath678 and for the other cases we have @xmath679 and therefore @xmath680 if @xmath403 is large enough . using the triangle inequality we get that @xmath681 which proves that @xmath682 as claimed . we now show that @xmath647 is lipschitz . let @xmath126 and @xmath127 be in @xmath683 we have @xmath684 and therefore we get , using the same estimates as for ( [ normcase ] ) , that @xmath685 provided @xmath403 is large enough . this completes the proof of theorem [ maint ] . we first prove the existence of a unique positive solution of equation ( [ eqeta ] ) satisfying the boundary conditions ( [ etainfinity ] ) and ( [ etazero ] ) . then , we derive the results on the asymptotic behavior near zero and infinity . let , for @xmath687 , @xmath688 be the solution of the initial value problem on @xmath189 , @xmath689 then , there exists a unique @xmath690 such that the function @xmath691 is positive and satisfies @xmath692 . we first prove that @xmath693 is unique . given a function @xmath218 from @xmath189 to @xmath694 we define the function @xmath695 @xmath696 . assume that there are two values @xmath697 such that the functions @xmath698 and @xmath699 are positive and satisfy @xmath700 @xmath701 we first show that the function @xmath702 is positive for all @xmath703 namely , if we assume the contrary , then because @xmath704 , there must be a first @xmath705 such that @xmath706 furthermore , if @xmath707 then @xmath708 and therefore @xmath709 a contradiction . therefore @xmath710 and as a consequence @xmath711 are positive for all @xmath121 from which it follows that @xmath712 in contradiction with @xmath713 to prove the existence of a @xmath690 for which @xmath714 is positive and for which @xmath715 , we use the so called shooting method . note that , for any @xmath687 , the initial value problem ( [ 1 ] ) has a unique solution @xmath716 and since @xmath717 the function @xmath688 is strictly decreasing on @xmath718 for @xmath719 small enough . we will show that for small enough @xmath687 , the graph of @xmath688 intersects the real axis and @xmath688 becomes negative , whereas for @xmath720 large enough , @xmath688 has a minimum and then diverges to plus infinity . the ( unique ) point between those two sets is @xmath690 . define the two subsets @xmath721 and @xmath722 of @xmath325 @xmath723 we note that if @xmath724 and @xmath725 for some @xmath726 , then @xmath727 on any interval @xmath728 on which @xmath688 is defined , and a function @xmath729 with @xmath730 can therefore not converge to zero at infinity . furthermore , since the function @xmath731 is a solution of the differential equation ( [ 1 ] ) , it follows , since solutions are unique , that @xmath732 if @xmath733 and therefore the intersection of @xmath721 with @xmath722 is empty . the sets @xmath721 and @xmath722 are open , by continuity of the solution @xmath688 as a function of the initial data @xmath734 . we now show that @xmath721 is non empty and bounded , which shows that @xmath735 this @xmath736 is neither in @xmath721 nor in @xmath722 , and therefore the function @xmath737 is at the same time strictly positive and strictly decreasing , and therefore @xmath715 . to prove that @xmath721 is non empty , we fix any @xmath738 positive and choose @xmath739 small enough such that on @xmath740 $ ] the solution @xmath741 exists and is strictly decreasing . then , @xmath742 choose now @xmath743 and let @xmath744 be the corresponding solution . as before , we have that the function @xmath745 and its second derivative @xmath746 are positive on the interval @xmath747 and therefore , since @xmath748 @xmath749 @xmath750 we find that @xmath751 using the definition of @xmath752 we therefore find that @xmath753 therefore @xmath754 . we now prove that @xmath721 is bounded . for @xmath755 , let @xmath719 be the largest value ( possibly infinite ) such that on @xmath718 the solution @xmath688 exists and is strictly positive . then , @xmath756 is positive on @xmath757 and , therefore @xmath758 for @xmath759 . as a consequence , if the function @xmath688 exists on @xmath760,$ ] then @xmath761 using again that @xmath762 we then find that @xmath763 for @xmath764 $ ] , and therefore @xmath765 on @xmath766 $ ] , which implies that @xmath767 which is positive if @xmath768 therefore @xmath769 must be equal to zero for some @xmath770 any such @xmath734 therefore belongs to @xmath771 if the function @xmath688 ceases to exist before @xmath772 it must have been diverging to plus infinity for some @xmath773 which again implies that @xmath774 must have been equal to zero for some @xmath775 and the corresponding @xmath734 is in @xmath771 the function @xmath223 is regular at zero , and the coefficients of its taylor series at zero can be computed recursively . we have @xmath777 and @xmath778 and therefore we get using the differential equation that @xmath779 @xmath780 and @xmath781 the asymptotic behavior of @xmath223 at infinity is obtained as follows . assuming that @xmath223 behaves like @xmath782 at infinity we get from the differential equation that @xmath230 and @xmath267 are as defined in ( [ defdelta007 ] ) and ( [ deflambda007 ] ) , respectively . that this is indeed the correct leading behavior of @xmath223 at infinity can now be proved by using standard techniques based on repeated applications of lhpital s rule . see for example @xcite . since the proof is simple , but lengthy and quite uninteresting , we do not give the details here . once the leading behavior of @xmath223 at infinity has been established we make the ansatz @xmath783 . to leading order we get for the function @xmath784 the linear equation @xmath785 there is a certain constant @xmath786 such that the function @xmath787 @xmath788 is a particular solution of equation ( [ eqs ] ) . the solutions of the homogeneous equation associated with ( [ eqs ] ) are of the form @xmath789 , and using the definition ( [ deflambda007 ] ) for @xmath267 we find that @xmath790 for @xmath791 we have that @xmath792 and the asymptotic behavior of @xmath784 is therefore for @xmath791 of the form @xmath793 with @xmath794 , and of the form @xmath795 with some unknown coefficient @xmath363 for @xmath796 . it is tedious , but not difficult , to prove that this is indeed the correct second order behavior of @xmath223 at infinity . we omit the details . in order to study equation ( [ mu2 ] ) with boundary conditions ( [ bmu2 - 1 ] ) and ( [ bmu2 - 2 ] ) , we make the ansatz @xmath797 for the function @xmath798 we get the differential equation @xmath799 and the boundary conditions for @xmath798 are @xmath800 as indicated in section [ smu2 ] , a solution of equation ( [ mu2 ] ) that is defined on @xmath189 behaves at infinity either like @xmath802 or like @xmath803 the proof is similar to the one in @xcite . we omit the details . given the asymptotic behavior of @xmath254 at infinity , we find for the function @xmath798 at infinity either a behavior proportional to @xmath804 , or a behavior proportional to @xmath805 since @xmath806 we find that @xmath807 if and only if the boundary condition ( [ bcm1 ] ) is satisfied , and we will impose ( [ bcm3 ] ) from now on . we now discuss the asymptotic behavior of the function @xmath798 near zero . from equation ( [ eqh ] ) we see that @xmath808 exists if and only if @xmath809 , i.e. , if @xmath810 and if @xmath811 we then find , that @xmath812 , with @xmath292 as defined in ( [ lambda0 ] ) . by taking derivatives of equation ( [ eqh ] ) we find that @xmath813 , and that @xmath814 for some constant @xmath815 by taking further derivatives , one can recursively compute the taylor coefficients of a solution @xmath816 of equation ( [ eqh ] ) that is regular ( in fact , analytic ) in a neighborhood of zero . the solution @xmath816 does however not satisfy the boundary condition ( [ bcm3 ] ) . the solution of ( [ eqh ] ) that does satisfy ( [ bcm3 ] ) is of the form @xmath817 where @xmath818 with @xmath819 as defined in ( [ eqn : critexp ] ) . here , @xmath820 , with @xmath821 and with @xmath822 to be determined . the asymptotic form ( [ msing ] ) can be obtained by substituting the ansatz ( [ msing ] ) for @xmath798 into equation ( [ eqh ] ) . since @xmath823 we find from ( [ msing ] ) that near zero @xmath824 . we omit the details of the proof that the asymptotic behavior is as indicated . we now prove the existence of a function @xmath798 that satisfies equation ( [ eqh ] ) with the boundary conditions ( [ bcm0 ] ) and ( [ bcm3 ] ) . since the second derivative of the solution @xmath798 at zero is positive , and since @xmath798 converges to zero at infinity , there must be a first @xmath825 such that @xmath826 the basic idea is now to use this position @xmath827 and the value @xmath828 of @xmath798 at @xmath482 , as parameters in shooting arguments towards zero and infinity . the first shooting argument will allow us to define a curve @xmath829 of initial conditions @xmath830 in @xmath831 for which the boundary condition at zero is satisfied , and the second shooting argument will allow us to find on this curve an initial condition for which the boundary condition at infinity is satisfied as well . so , let @xmath832 be an initial condition . locally , i.e. , near @xmath833 there exists a solution @xmath834 of equation ( [ eqh ] ) . by definition , @xmath835 @xmath836 and therefore we get for the second derivative of @xmath834 at @xmath833 @xmath837 where @xmath838 and @xmath839 for initial conditions such that @xmath840 where @xmath841 and @xmath842 we therefore have that @xmath843 see fig . 1 for the graph of the function @xmath844 the function @xmath845 has a maximum at the point @xmath846 that satisfies the equation @xmath847 and the line @xmath845 divides the set of initial conditions into two subsets , a subset @xmath0 where @xmath848 and a subset @xmath849 where @xmath850 for initial conditions on @xmath845 we can compute @xmath851@xmath852 comparing with ( [ defxim ] ) we find that @xmath853 and we have that @xmath854 for @xmath855 we now construct the line @xmath829 for @xmath855 the proof is similar to the one in section [ seta777 ] . define the two subsets @xmath721 and @xmath722 of the interval @xmath863 , @xmath864 the intersection of @xmath721 with @xmath722 is by definition empty , and the sets @xmath721 and @xmath722 are open , by continuity of the solution @xmath865 as a function of the initial data @xmath734 . we now show that all @xmath720 sufficiently close to @xmath267 are in @xmath866 and that all @xmath734 sufficiently close to @xmath867 are in @xmath868 this implies that @xmath869 and @xmath870 is neither in @xmath721 nor in @xmath722 , and therefore the function @xmath871 satisfies @xmath872 for all @xmath873 and therefore @xmath874 since @xmath875 so let @xmath832 be an initial condition . then , @xmath834 satisfies the integral equation @xmath876 where @xmath877 and where @xmath878 @xmath879 is strictly negative for @xmath880 and @xmath881 and therefore we find , like in the proof in section [ theoremeta ] that any solution with an initial condition @xmath734 sufficiently close to @xmath882 will cross the line @xmath883 similarly , for an initial condition @xmath832 close to @xmath884 we can use that @xmath885 @xmath886 and that @xmath887 is strictly negative to show that the corresponding solution will cross the line @xmath844 this completes the proof of the existence of @xmath888 to prove uniqueness it is sufficient to use that @xmath889 for @xmath890 in the set @xmath26 ( see fig . 1 ) , and to integrate the difference of two solutions from their respective initial condition to zero , which leads to a contradiction , since both solutions have to be equal to @xmath267 at zero . finally , that @xmath829 is a continuous function follows from the continuity of @xmath834 as a function of @xmath734 and @xmath891 using the uniqueness of @xmath888 we now prove with a second shooting argument that solutions with initial conditions @xmath892 with @xmath893 become negative somewhere in the interval @xmath894 and that solutions with initial conditions @xmath895 , with @xmath896 stay positive and diverge to plus infinity . define the two subsets @xmath721 and @xmath722 of the interval @xmath900 , @xmath901 by definition , the intersection of @xmath721 with @xmath722 is empty , and the sets @xmath721 and @xmath722 are open , by continuity of the solution @xmath902 as a function of the initial data @xmath891 . we now show that all @xmath891 sufficiently close to @xmath903 are in @xmath866 and that all @xmath891 sufficiently close to @xmath846 are in @xmath868 this implies that @xmath904 is neither in @xmath721 nor in @xmath722 , and therefore the function @xmath905 is positive and decreasing for @xmath906 which implies that @xmath907 so let @xmath908 be an initial condition with @xmath909 with @xmath910 the proof that such an initial condition is in @xmath721 is rather lengthy and we therefore do not give the details here , but on a heuristic level it is easy to understand why such a solution is in @xmath721 . namely , near zero the asymptotics of the solution @xmath871 is @xmath911 where @xmath818 and where @xmath912 is such that @xmath913 . neglecting higher order terms we find that @xmath914 , and we conclude that @xmath915 therefore , @xmath916 and @xmath917 if @xmath891 is small enough . therefore , since @xmath918 for all @xmath919 we find that @xmath920 and therefore @xmath921 for some @xmath922 as claimed . next , let @xmath923 be an initial condition with @xmath924 with @xmath925 to be chosen below . we now show that such an initial condition is in @xmath771 for all @xmath926 we have the lower bounds @xmath927 and @xmath928 @xmath929 where @xmath930 @xmath931 and @xmath932 and for @xmath933 we therefore have the lower bound @xmath934 and it follows , using again the integral equation ( [ inteq ] ) , that @xmath935 diverges at ( or before ) infinity , that therefore @xmath936 for some @xmath937 which implies that @xmath938 provided @xmath939 for @xmath940 small enough and for @xmath93 large enough ( [ cond ] ) can be verified without too much difficulty . with the help of a computer one can show that ( [ cond ] ) is satisfied for the remaining @xmath941 for @xmath251 ( [ cond ] ) is not satisfied , since the above bounds on @xmath942 and @xmath943 are too weak . sufficiently good bounds can be obtained by dividing the interval @xmath944 in two pieces and by integrating lower bounds on each of the subintervals . we omit the details . finally , uniqueness of @xmath945 can be proved by integrating the difference of two solutions from @xmath946 to infinity , which , using the positivity of @xmath947 leads to a contradiction with the fact that both of the solutions converge to zero at infinity . the equation ( [ h ] ) for @xmath280 is linear . we therefore first construct two linearly independent solutions @xmath948 and @xmath949 for the corresponding homogeneous equation , which we then use to construct , using standard methods , a solution of ( [ h ] ) that satisfies the boundary conditions ( [ bh-1 ] ) and ( [ bh-2 ] ) . the homogeneous equation is @xmath950 where @xmath951 since the equation ( [ homh ] ) is linear , the integral equation for @xmath948 , @xmath952 has a positive solution that exists for all @xmath86 in @xmath211 by definition , we have near @xmath56 the behavior @xmath953 at infinity , the solution @xmath948 is asymptotic to a solution of the equation @xmath954 this equation is the same as the homogeneous part of equation ( [ eqs ] ) , and the leading order behavior of @xmath948 at infinity is therefore either proportional to @xmath955 or to @xmath956 with @xmath957 as defined in ( [ eqn : critexp ] ) . since @xmath948 is positive , we find using ( [ eqh1 ] ) , that @xmath958 for all @xmath86 in @xmath325 and therefore @xmath948 is near infinity of the form @xmath959 for some constant @xmath960 . a second solution of the homogeneous equation ( [ homh ] ) is @xmath961 near @xmath56 we have that @xmath962 and near infinity we find that @xmath963 where @xmath964 , and @xmath965 @xmath966 we note that @xmath967 therefore , the function @xmath968 , @xmath969 where @xmath970 and where @xmath971 satisfies equation ( [ h ] ) . near zero , the function @xmath628 is of the form @xmath972 , and therefore , using the behavior of @xmath948 and @xmath949 near zero , we find that @xmath973 is near zero of order @xmath974 , and @xmath845 is near zero of order @xmath975 the function @xmath968 is therefore of order @xmath976 near zero . at infinity , the function @xmath628 is of the form @xmath977 where @xmath978 and therefore the function @xmath845 is near infinity of the form @xmath979 , and @xmath973 is near infinity of the form @xmath980 , for some constant @xmath981 using these asymptotic behavior for @xmath973 , @xmath948 , @xmath845 , and @xmath290 we find for the function @xmath982 near infinity the behavior , @xmath983 in the last equality we have used the definition ( [ lambda0 ] ) for @xmath984 the function @xmath985@xmath986 solves the equation ( [ h ] ) , satisfies the boundary condition ( [ bh-1 ] ) , and since , as we show in the next section , the higher order terms in ( [ hpinfty ] ) converge to zero at infinity , it also satisfies the boundary condition ( [ bh-2 ] ) . by construction , the leading behavior of @xmath280 at infinity is @xmath987 . we therefore make the ansatz @xmath988 , and to leading order , we get for the function @xmath191 the linear equation @xmath989 for a certain constant @xmath990 . the general solution of equation ( [ bvarphi2 - 2 ] ) is @xmath991 with a certain constant @xmath992 and with @xmath819 , @xmath993 as defined in ( [ eqn : critexp ] ) . since @xmath994 but @xmath995 the coefficient of the term proportional to @xmath996 must be zero . therefore , since@xmath997 for all @xmath998 the asymptotic behavior of @xmath999 is always given by @xmath1000 we omit the details of the proof that this is indeed the correct second order behavior of @xmath191 at infinity . j. bricmont , a. kupiainen & g. lin : renormalisation group and asymptotics of solutions of nonlinear parabolic equations . _ communications in pure and applied mathematics _ , new york , ny issn 0010 - 3640 , * 47 * , 893 , ( 1994 ) .

using the two - loop approximation of chiral perturbation theory , we calculate the momentum and density dependent nuclear spin - orbit strength . this quantity is derived from the spin - dependent part of the interaction energy of a nucleon scattering off weakly inhomogeneous isospin symmetric nuclear matter . we find that iterated-exchange generates at saturation density ,mev , a spin - orbit strength at ofmev in perfect agreement with the empirical value used in the shell model . this novel spin - orbit strength is neither of relativistic nor of short range origin . the potential underlying the empirical spin - orbit strength becomes a rather weak one ,mev , after the identification as suggested by the present calculation . we observe however a strong-dependence of leading even to a sign change abovemev . this and other features of the emerging spin - orbit hamiltonian which go beyond the usual shell model parametrization leave questions about the ultimate relevance of the spin - orbit interaction generated by-exchange for a finite nucleus . we also calculate the complex - valued isovector single - particle potential in isospin asymmetric nuclear matter proportional to . for the real part we find reasonable agreement with empirical values and the imaginary part vanishes at the fermi - surface . 24.6 cm 17.0 cm -1.6 cm -0.6 cm -0.6 cm # 1#2 # 2 cm n. kaiser + physik department t39 , technische universitt mnchen , d-85747 garching , germany + _ email : nkaiser@physik.tu-muenchen.de_ pacs : 12.38.bx , 21.65.+f , 24.10.cn + keywords : effective field theory at finite density , nuclear spin - orbit interaction , complex single - particle potential in isospin asymmetric nuclear matter . 0.5 cm

nucl-th0206056

the introduction of the spin - orbit term into the nuclear single - particle hamiltonian by haxel , jensen , suess and goeppert - mayer @xcite in 1949 has been most decisive for the success of the nuclear shell model . only with a very strong and attractive spin - orbit potential one is , for example , able to explain the observed sequence of so - called magic numbers @xmath18 . the dynamical origin of the strong nuclear spin - orbit force has not been fully resolved even up to date . the analogy with the spin - orbit interaction in atomic physics gave the hint that it could be a relativistic effect . this idea has lead to the construction of the relativistic ( scalar - vector ) mean - field model @xcite . in this model the spin - independent nuclear potential ( of approximate depth @xmath19mev ) results from an almost complete cancelation of a very strong attraction generated by scalar ( @xmath20-meson ) exchange and a nearly equally strong repulsion generated by vector ( @xmath21-meson ) exchange . the corresponding spin - orbit term ( obtained by a non - relativistic reduction of the nucleon s dirac - hamiltonian ) comes out proportional to the coherent sum of the very large scalar and vector mean - fields . in this sense , the relativistic mean - field model gives a simple and natural explanation of the basic features of the nuclear shell model potential . refinements of relativistic mean - field models which include additional non - linear couplings of the scalar and vector fields or explicitly density - dependent couplings are nowadays widely and successfully used for nuclear structure calculations @xcite . the nuclear spin - orbit potential arises generally as a many - body effect from the underlying spin - orbit term in the ( free ) nucleon - nucleon scattering amplitude . the calculation of the tree level diagrams with one scalar - meson or one vector - meson exchange between nucleons gives indeed a spin - orbit term in the nn t - matrix proportional to @xmath22 , with @xmath23 denoting the nucleon mass . the nuclear spin - orbit potential corresponding to scalar and vector meson exchange is therefore obviously a truly relativistic effect . however , the quadratic reciprocal scaling of the spin - orbit nn - amplitude with the nucleon mass @xmath23 is not universal , and it changes if one considers the exchange of two mesons between nucleons , i.e. loop diagrams . for example , irreducible two - pion exchange gives rise to a spin - orbit term in the nn t - matrix proportional to @xmath24 ( see eqs.(22,23 ) in ref.@xcite ) and iterated one - pion exchange produces a spin - orbit term in the nn t - matrix which even scales linearly with the nucleon mass @xmath23 ( see eq.(33 ) in ref.@xcite ) . it is one of the chief purposes of this paper to investigate in detail the contributions from iterated one - pion exchange to the nuclear spin - orbit interaction . as already mentioned the latter arises from the spin - orbit term in the nn t - matrix as a many - body effect , e.g. in connection with a nuclear matter calculation . in a recent work @xcite , we have used chiral perturbation theory for a systematic treatment of the nuclear matter many - body problem . in this calculation the contributions to the energy per particle , @xmath25 , originate exclusively from one- and two - pion exchange between nucleons and they are ordered in powers of the fermi momentum @xmath26 ( modulo functions of @xmath27 ) . it has been demonstrated in ref.@xcite that the empirical saturation point and the nuclear matter compressibility @xmath28mev can be well reproduced at order @xmath29 in the chiral expansion with just one single momentum cut - off scale of @xmath30gev which parametrizes all necessary short range dynamics . most surprisingly , the prediction for the asymmetry energy , @xmath31mev , is in very good agreement with its empirical value . furthermore , as a nontrivial fact pure neutron matter is predicted to be unbound and the corresponding equation of state agrees roughly with that of sophisticated many - body calculations for low neutron densities @xmath32@xmath33 . in a subsequent work @xcite , the momentum and density dependent ( real ) single - particle potential @xmath34 ( i.e. the spin - independent average nuclear mean - field ) has been calculated in the same framework . it was found that chiral @xmath2- and @xmath14-exchange give rise to a potential depth for a nucleon at the bottom of the fermi sea of @xmath35mev . this value is in very good agreement with the depth of the empirical optical model potential and the nuclear shell model potential . nuclear matter at finite temperatures has been investigated in the same framework in ref.@xcite . there it was shown that chiral @xmath2- and @xmath14-exchange reproduce the first - order liquid - gas phase transition of isospin symmetric nuclear matter with a realistic value @xmath36mev of the critical temperature . our approach to the nuclear matter problem is in many respects different from most other commonly used ones , where one starts from a so - called realistic nn - potential . for example in the relativistic nuclear matter calculation of ref.@xcite the @xmath37 , @xmath38 and @xmath39-waves deliver more than 95% of the potential energy per particle . the finding that perturbative chiral pion - nucleon dynamics leads already to good nuclear matter and single particle properties hints at the fact that the detailed nn - interaction is of no more relevance . fine - tuning of the single cut - off scale @xmath40 to one nuclear matter observable ( the binding energy per particle @xmath41 ) is however still necessary in our present approach @xcite . it is the purpose of this work to calculate , using the same framework as in ref.@xcite , the momentum and density dependent nuclear spin - orbit strength @xmath0 . this quantity is derived from the spin - dependent part of the interaction energy @xmath42 of a nucleon scattering off weakly inhomogeneous isospin symmetric nuclear matter . we will present here analytical expressions for the contributions from @xmath2-exchange and iterated @xmath2-exchange to the spin - orbit strength @xmath0 . furthermore , we calculate in isospin asymmetric ( homogeneous ) nuclear matter the ( complex - valued ) isovector single - particle potential @xmath43 accompanied by the isospin double - asymmetry @xmath16 . our results can be summarized as follows : * at nuclear matter saturation density , @xmath3mev , @xmath2-exchange and iterated @xmath2-exchange generate a spin - orbit strength at @xmath4 of @xmath44mev@xmath6 . this result , which is dominated by the contribution of four hartree - type diagrams , is in perfect agreement with the empirical value of the spin - orbit strength @xmath45mev@xmath6 @xcite used in shell - model calculation of nuclei . the novel spin - orbit strength found here is neither of relativistic nor of short range origin . it is in fact linearly proportional to the nucleon mass @xmath23 and its inherent range is the pion compton wavelength @xmath46fm . the latter feature tempts to an unconventional interpretation of the strong nuclear spin - orbit interaction . the potential @xmath7 underlying the empirical spin - orbit strength @xmath47mev@xmath6 becomes a rather weak one , namely @xmath48mev , after the identification of the effective range @xmath49 with the pion compton wavelength , @xmath50 , as suggested by the present calculation . * we observe however a strong @xmath11-dependence of @xmath12 which leads even to a sign change above @xmath13mev . the calculated spin - orbit strength @xmath51 depends also strongly on the value of the pion mass and it shows a pronounced maximum around @xmath52mev . a further property of the spin - orbit hamiltonian emerging from our diagrammatic calculation is that it has ( in coordinate space ) terms proportional to @xmath53 as well as terms proportional to @xmath54 ( with @xmath55 the normalized radial density profile ) which get weighted differently at the surface of a finite nucleus . all such features of our calculation which go beyond the usual shell model parametrization of the spin - orbit hamiltonian leave questions about the ultimate relevance of the spin - orbit interaction generated by @xmath14-exchange for a finite nucleus . implementing the present results for @xmath0 into nuclear structure calculations will clarify the role of the nuclear spin - orbit interaction generated by @xmath14-exchange . * the real part of the isovector single - particle potential @xmath56 generated by chiral @xmath2- and @xmath14-exchange has a density dependence very similar to that of the asymmetry energy @xmath57 @xcite . at saturation density , @xmath58@xmath33 , we find a repulsive isovector potential of @xmath59mev . this prediction is comparable to the value @xmath60mev used in shell model calculations @xcite or the value @xmath61mev deduced from nucleon - nucleus scattering in the framework of the optical model @xcite . the momentum dependence of @xmath62 is non - monotonic in the interval @xmath63 . one observes a broad maximum at @xmath64mev where the ( real ) isovector single - particle potential has increased by about 30% to the value @xmath65mev . the imaginary part @xmath66 vanishes ( quadratically ) at the fermi - surface @xmath67 in accordance with luttinger s theorem @xcite . let us begin with recalling the spin - orbit hamiltonian of the nuclear shell model @xcite which is generally written in the form : @xmath68 here , @xmath69 is the conventional pauli spin - vector and @xmath70 is the orbital angular momentum of a nucleon . @xmath71 denotes the radial density distribution of a nucleus ( typically parametrized by a saxon - woods function ) . the empirical value of the nuclear spin - orbit strength is @xmath72 @xcite . this large and positive value of @xmath73 leads to a strongly attractive spin - orbit potential acting mainly at the surface of a nucleus . we wish to calculate the nuclear spin - orbit strength @xmath73 ( or an appropriate generalization of it ) in the systematic framework of chiral perturbation theory @xcite . the first observation one makes is that the spin - orbit interaction vanishes identically in infinite homogeneous nuclear matter since there is no preferred center in this system in order to define an orbital angular momentum . therefore one has to generalize the calculation of the single - particle potential in ref.@xcite to ( at least ) weakly inhomogeneous nuclear matter . the relevant quantity in order to extract the nuclear spin - orbit strength is the spin - dependent part of the interaction energy of a nucleon scattering off weakly inhomogeneous isospin symmetric nuclear matter from initial momentum @xmath74 to final momentum @xmath75 , which reads : @xmath76 the ( small ) momentum transfer @xmath77 is provided by the fourier - components of the inhomogeneous nuclear matter distribution . the density form factor @xmath78 plays the role of a probability distribution of these fourier - components . the form factor @xmath79 should be viewed as narrowly peaked function around @xmath80 with its fourier - transform equal to the ( slowly varying ) density profile , @xmath81 . using this relationship the spin - orbit hamiltonian @xmath82 in eq.(1 ) becomes equal to the fourier - transform of the product of the density form factor and the spin - dependent interaction energy : @xmath83 . consistent with the assumption of a weakly inhomogeneous nuclear matter distribution we keep in @xmath84 only linear terms in @xmath77 corresponding to small spatial density gradients . for practical purposes , this means that after isolating the proportionality factor @xmath77 in an explicit calculation the momentum and density dependent spin - orbit strength @xmath0 can be finally computed in the limit of homogeneous isospin symmetric nuclear matter ( characterized by its fermi momentum @xmath26 ) . ( 400,120 ) ( 50,0)(50,100 ) ( 43,47)(57,47 ) ( 43,53)(57,53 ) ( 50,25)(50,30 ) ( 50,70)(50,75 ) ( 10,50)[]@xmath85 ( 20,15)[]@xmath86 ( 22,85)[]@xmath87 ( 50,0)4 ( 50,100)4 ( 65,0)[]@xmath88 ( 65,100)[]@xmath89 ( 150,0)(150,100 ) ( 150,50)(200,50)55 ( 230,50)(30,0,180 ) ( 230,50)(30,180,360 ) ( 253,47)(267,47 ) ( 253,53)(267,53 ) ( 150,25)(150,30 ) ( 150,70)(150,75 ) ( 150,50)4 ( 200,50)4 ( 175,65)[]@xmath90 ( 125,15)[]@xmath74 ( 125,85)[]@xmath75 ( 350,0)(350,120 ) ( 350,60)(40,-90,90)6 ( 350,20)4 ( 350,100)4 ( 343,57)(357,57 ) ( 343,63)(357,63 ) ( 350,39)(350,40 ) ( 350,79)(350,85 ) _ fig.1 : left : the double line symbolizes the medium insertion for a weakly inhomogeneous many - fermion system @xmath91 defined by eqs.(3,4 ) . middle : the @xmath20- and @xmath21-exchange hartree graph . right : the @xmath2-exchange fock graph . _ in ref.@xcite the calculation of the single - particle potential in homogeneous isospin symmetric nuclear matter has been organized in the number of so - called medium insertions . the latter is a technical notation for the difference between the vacuum and in - medium nucleon propagator ( see eq.(3 ) in ref.@xcite ) . in the case of homogeneous nuclear matter a medium insertion in a self - energy diagram converts effectively a four - dimensional loop integration into an integral over a fermi - sphere of radius @xmath26 . the medium insertion for a ( non - relativistic ) many - fermion system is generally constructed from the sum over the occupied energy eigenstates as @xcite : @xmath92 the double line in the left picture of fig.1 symbolized this medium insertion together with the assignment of the out- and in - going nucleon momenta @xmath93 . the momentum transfer @xmath94 is provided by the fourier - components of the inhomogeneous matter distribution . semiclassical expansions @xcite give for a weakly inhomogeneous and spin - saturated many - fermion system : @xmath95 with @xmath96 the density form factor introduced after eq.(2 ) . the subleading @xmath97 term in eq.(4 ) will in fact never come into play in our diagrammatic calculation of the spin - dependent interaction energy @xmath84 to linear order in @xmath77 . as a first check on this formalism we evaluate the @xmath20- and @xmath98-exchange hartree diagram in fig.1 . we perform the non - relativistic @xmath99-expansion of the scalar / vector interaction vertex sandwiched between dirac - spinors for the out- and in - going nucleon ( of momentum @xmath100 ) until we obtain the spin - orbit like term @xmath101 . after that we can take the limit of homogeneous nuclear matter and perform the remaining integral over a fermi - sphere of radius @xmath26 which brings one factor of density @xmath102 . putting all pieces together we reproduce the familiar result : @xmath103 of the relativistic mean - field model @xcite . the contribution of the analogous @xmath20- and @xmath21-exchange fock diagrams to the nuclear spin - orbit strength can also be easily calculated with the help of the formalism outlined above . we obtain from these fock diagrams @xmath104 of the @xmath20-exchange contribution and @xmath105 of the @xmath21-exchange contribution written in eq.(5 ) . in this work our main interest is focussed on the nuclear spin - orbit interaction generated by chiral one- and two - pion exchange . in this section we present analytical results for the nuclear spin - orbit strength @xmath0 as given by chiral one- and two - pion exchange . we start with the @xmath2-exchange fock graph ( last diagram in fig.1 ) . in the static approximation the product of @xmath106-interaction vertices @xmath107 is spin - independent . a non - vanishing nuclear spin - orbit strength comes therefore only as a relativistic @xmath22-correction . isolating the @xmath108 factor from the product of fully relativistic pseudovector @xmath106-interaction vertices , performing the @xmath24-expansion , and integrating finally over a fermi - sphere of radius @xmath26 , we get from the @xmath2-exchange fock diagram : @xmath109 \nonumber \\ & & + \bigg[(x^2-u^2)^3-{3\over 2}(x^2-u^2)^2 + 6x^2 + { 1\over 2}\bigg ] l(x , u ) \bigg\ } \ , . \end{aligned}\ ] ] here , we have introduced the dimensionless variables @xmath110 and @xmath111 and the auxiliary function @xmath112 for the readers convenience we will present the spin - orbit strength at @xmath4 of each individual diagram in a separate formula , since in most cases the limit @xmath113 is quite non - trivial . for the @xmath2-exchange fock diagram we find the simple expression @xmath114 which gives numerically at saturation density @xmath3mev only about @xmath115 of the empirical value @xmath72 . ( 400,140 ) ( 20,110)[](a ) ( 0,0)(0,120 ) ( 0,30)(40,30)6 ( 0,90)(40,90)6 ( 40,30)(40,90 ) ( 40,60)(30,-90,90 ) ( 63,57)(77,57 ) ( 63,63)(77,63 ) ( 0,30)4 ( 0,90)4 ( 40,30)4 ( 40,90)4 ( 0,58)(0,62 ) ( 40,58)(40,62 ) ( 130,110)[](b ) ( 110,0)(110,120 ) ( 110,30)(150,30)6 ( 110,90)(150,90)6 ( 150,30)(150,90 ) ( 150,60)(30,-90,90 ) ( 143,57)(157,57 ) ( 143,63)(157,63 ) ( 110,30)4 ( 110,90)4 ( 150,30)4 ( 150,90)4 ( 103,57)(117,57 ) ( 103,63)(117,63 ) ( 180,62)(180,56 ) ( 110,18)(110,19 ) ( 110,109)(110,110 ) ( 240,110)[](c ) ( 220,0)(220,120 ) ( 220,30)(260,30)6 ( 220,90)(260,90)6 ( 260,30)(260,90 ) ( 260,60)(30,-90,90 ) ( 283,57)(297,57 ) ( 283,63)(297,63 ) ( 220,30)4 ( 220,90)4 ( 260,30)4 ( 260,90)4 ( 213,57)(227,57 ) ( 213,63)(227,63 ) ( 260,56)(260,66 ) ( 220,18)(220,19 ) ( 220,109)(220,110 ) ( 350,110)[](d ) ( 330,0)(330,120 ) ( 330,30)(370,30)6 ( 330,90)(370,90)6 ( 370,30)(370,90 ) ( 370,60)(30,-90,90 ) ( 393,57)(407,57 ) ( 393,63)(407,63 ) ( 330,30)4 ( 330,90)4 ( 370,30)4 ( 370,90)4 ( 363,57)(377,57 ) ( 363,63)(377,63 ) ( 370,77)(370,80 ) ( 370,38)(370,49 ) ( 330,59)(330,66 ) _ fig.2 : the iterated @xmath2-exchange hartree graphs . their isospin factor in symmetric nuclear matter is 6 . _ next , we come to the evaluation of the four hartree diagrams of iterated one - pion exchange shown in fig.2 . we start with the left graph with one medium insertion , labeled ( a ) . the relevant @xmath116 prefactor can be isolated already in the first step of the calculation from the product of @xmath106-interaction vertices @xmath117 at the open nucleon line . for all remaining parts of the diagram one can then take the limit of homogeneous nuclear matter ( i.e. @xmath118 ) . using the analytical results given in section 4.3 of ref.@xcite for the inner @xmath119-loop integral we can even perform the integral over a fermi - sphere of radius @xmath26 . altogether , we find the following closed form expression for the spin - orbit strength generated by the hartree diagram ( a ) : @xmath120 x l(x , u ) \bigg\ } \ , . \end{aligned}\ ] ] at zero nucleon momentum @xmath121 this simplifies to @xmath122 note also that @xmath123 in eq.(9 ) originates from the real part of the iterated @xmath2-exchange spin - orbit nn - amplitude ( eq.(33 ) in ref.@xcite ) evaluated in forward direction and integrated over a fermi - sphere of radius @xmath26 . we continue with the calculation of the hartree diagrams with two medium insertions , labeled ( b ) , ( c ) and ( d ) in fig.2 . in these diagrams , a ( small ) momentum transfer @xmath124 with @xmath125 occurs at each medium insertion . the important prefactors @xmath126 can again be isolated in the first step of the calculation from the product of @xmath127-interaction vertices at the open nucleon line . after that the vectors @xmath128 can be set to zero in all remaining components of these diagrams . when fourier - transformed with the density form factors to coordinate space each such momentum transfer @xmath129 leads to the expression : @xmath130 , with @xmath131 the density profile of weakly inhomogeneous nuclear matter . consistent with the assumption of a weakly inhomogeneous nuclear matter ( i.e. keeping only linear terms in small spatial gradients ) we can make the replacement : @xmath132 . we will come back to this point in the next section when discussing the results for the spin - orbit strength @xmath0 as well as their relevance for a finite nucleus . the essential conclusion from the previous considerations is that for the calculation of @xmath84 in weakly inhomogeneous nuclear matter each momentum transfer @xmath94 can be identified with @xmath77 . with the help of this rule and certain techniques to reduce six - dimensional principal value integrals over the product of two fermi - spheres of radius @xmath26 , we end up with the following result for the hartree diagram ( b ) : @xmath133 \bigg [ { 3s+2s^3 \over 1+s^2 } -3 \arctan s \bigg ] \,,\ ] ] with the auxiliary function @xmath134 . throughout this work the momentum @xmath11 is restricted to the interval @xmath135 . at zero nucleon momentum @xmath121 eq.(11 ) simplifies to @xmath136 similarly , we find for the hartree diagram ( c ) : @xmath137 ^ 2 } \,\bigg[2u \xi + ( u^2-\xi^2 ) \ln{u+\xi\over u-\xi } \bigg ] \,,\ ] ] @xmath138 \bigg\ } \ , , \end{aligned}\ ] ] and for the hartree diagram ( d ) : @xmath139 \bigg [ { 3\sigma+2\sigma^3 \over 1+\sigma^2 } -3 \arctan \sigma \bigg ] \,,\ ] ] with the auxiliary function @xmath140 . the limit @xmath113 of eq.(15 ) is quite non - trivial since it requires a careful treatment of the singular function re@xmath141 . the following representation is manifestly free of singularities , @xmath142 \bigg\ } \ , . \end{aligned}\ ] ] note also the similarity of the expressions eqs.(11,13,15 ) with the three - body single - particle potential given in eq.(11 ) of ref.@xcite . ( 400,140 ) ( 20,135)[](e ) ( 0,0)(0,140 ) ( 0,52.5)(35,-90,90)6 ( 0,87.5)(35,-90,90)6 ( 0,17.5)4 ( 0,52.5)4 ( 0,87.5)4 ( 0,122.5)4 ( -7,67)(7,67 ) ( -7,73)(7,73 ) ( 0,106)(0,111 ) ( 0,38)(0,41 ) ( 130,135)[](f ) ( 110,0)(110,140 ) ( 110,52.5)(35,-90,90)6 ( 110,87.5)(35,-90,90)6 ( 110,17.5)4 ( 110,52.5)4 ( 110,87.5)4 ( 110,122.5)4 ( 103,102)(117,102 ) ( 103,108)(117,108 ) ( 103,32)(117,32 ) ( 103,38)(117,38 ) ( 110,69)(110,76 ) ( 240,135)[](g ) ( 220,0)(220,140 ) ( 220,52.5)(35,-90,90)6 ( 220,87.5)(35,-90,90)6 ( 220,17.5)4 ( 220,52.5)4 ( 220,87.5)4 ( 220,122.5)4 ( 213,102)(227,102 ) ( 213,108)(227,108 ) ( 213,67)(227,67 ) ( 213,73)(227,73 ) ( 220,39)(220,41 ) ( 350,135)[](g ) ( 330,0)(330,140 ) ( 330,52.5)(35,-90,90)6 ( 330,87.5)(35,-90,90)6 ( 330,17.5)4 ( 330,52.5)4 ( 330,87.5)4 ( 330,122.5)4 ( 323,67)(337,67 ) ( 323,73)(337,73 ) ( 323,32)(337,32 ) ( 323,38)(337,38 ) ( 330,107)(330,111 ) _ fig.3 : the iterated @xmath2-exchange fock graphs . their isospin factor in symmetric nuclear matter is @xmath143 . _ next , we come to the evaluation of the four iterated @xmath2-exchange fock diagrams shown in fig.3 . we start with diagram ( e ) with one medium insertion . let us first consider the @xmath106-interaction vertices . the spin - dependent part of the matrix - product @xmath144 is @xmath145 . in the case of graph ( e ) one makes the assignment : @xmath146 and @xmath147 , with @xmath148 the loop momentum and @xmath149 , where @xmath150 belongs to the external nucleon line and @xmath151 to the internal nucleon line carrying the medium insertion . the important proportionality factor @xmath152 is now not produced by the @xmath106-interaction vertices as it was the case for the hartree diagrams in fig.2 . in order to isolate the factor @xmath77 we combine the previously mentioned @xmath153 term with the nucleon energy denominator ( which in the actual calculation results from the @xmath154-loop integration ) and employ the identity : @xmath155^{-1}=1-\vec l \cdot\vec q \,\,[\vec l \cdot ( \vec l+\vec q+\vec q\,)]^{-1}$ ] . the term coming along with @xmath156 in this decomposition finally loop - integrates to zero . from the second term one can now easily isolate the factor @xmath77 and take the limit to homogeneous nuclear matter of all remaining factors of the diagram . putting all pieces together we end up with the following representation for the nuclear spin - orbit strength generated by the iterated @xmath2-exchange fock diagram ( e ) : @xmath157 [ u^2-(2\xi - x)^2 ] \over 4\xi^2 ( 1 + 2\xi^2 ) } \nonumber \\ & & \times\big[(1 + 4\xi^2 ) \arctan 2\xi -4\xi^2(1+\xi^2 ) \arctan \xi \big ] \bigg\ } \ , , \end{aligned}\ ] ] @xmath158 \bigg\ } \,.\ ] ] note also that @xmath159 in eq.(17 ) originates from the real part of the iterated @xmath2-exchange spin - orbit nn - amplitude ( eq.(33 ) in ref.@xcite ) evaluated in backward direction and integrated over a fermi - sphere of radius @xmath26 . we continue with the computation of the fock diagrams with two medium insertion , labeled ( f ) and ( g ) in fig.3 . diagram ( f ) with a symmetrical arrangement of the two medium insertions leads to the following contribution to the nuclear spin - orbit strength : @xmath160 \nonumber \\ & & \times \bigg [ u(1+u^2+x^2)-[1+(u+x)^2][1+(u - x)^2 ] l(x , u)\bigg ] \nonumber \\ & & + \int_{-1}^1 dy\ , \int_{-1}^1 dz \,{y|z| \,\theta(y^2+z^2 - 1 ) \over x|y| \sqrt{y^2+z^2 - 1 } } \big [ \ln(1+s^2)-s^2 \big ] \big [ t-\arctan t \big]\bigg\ } \ , , \end{aligned}\ ] ] with the auxiliary function @xmath161 . at zero nucleon momentum @xmath121 the lengthy expression eq.(19 ) simplifies drastically to @xmath162 note the leading @xmath163-behavior of the expressions in eqs.(12,20 ) derived from diagrams with two medium insertions . the results eqs.(8,10,18 ) belonging to diagrams with one medium insertion , on the other hand , show a leading linear dependence on the density @xmath164 . finally , we have to evaluate the last two topologically distinct fock diagrams in fig.3 . since they contribute equally to the nuclear spin - orbit strength we have given both diagrams the same label ( g ) . in order to avoid very lengthy formulas we split their contribution to @xmath0 into a `` factorizable '' @xmath165 and a `` non - factorizable '' part @xmath166 . technically these two pieces are distinguished by the feature whether the nucleon propagator in the denominator can be canceled by terms from the product of @xmath106-interaction vertices in the numerator , or not . we find from the iterated @xmath2-exchange fock diagrams ( g ) with two medium insertions the following `` factorizable '' contribution to the nuclear spin - orbit strength : @xmath167 \big[x+(x^2 - 1-\xi^2)l(\xi , x ) \big ] \nonumber \\ & & + \bigg [ { 3\xi^4\over 2 } ( 4u^2 + 5x^2 - 3 ) + { \xi^2 \over 2}(5 + 10u^2 - 4x^2 - 14u^2x^2-x^4)-6\xi^6-u^2x^4 \nonumber \\ & & + { 1\over 2}(5 + 3u^4 + 3x^2-u^4x^2)+2u^2+x^4 - 3u^2x^2+{3\over2\xi^2 } ( 1+u^2)^2(1+x^2)^2 \bigg ] \nonumber \\ & & \times l(\xi , x ) l(\xi , u ) + \bigg [ 6 \xi^4 + { \xi^2 \over 2 } ( x^2 - 12u^2 - 3)+ ( 1+x^2)(u^2 - 1)\nonumber \\ & & -{3 \over 2 \xi^2 } ( 1+x^2)(1+u^2)^2 \bigg ] x l(\xi , u ) + \bigg [ 6\xi^4 + { \xi^2 \over 2 } ( 13 - 15x^2)\nonumber \\ & & + { 1 \over 2}(x^4+u^2x^2 + 3x^2 - 3u^2 - 2 ) -{3 \over 2 \xi^2}(1+x^2)^2(1+u^2 ) \bigg]u l(\xi , x ) \bigg\}\ , , \end{aligned}\ ] ] which turns at zero nucleon momentum @xmath121 into the form : @xmath168 + ( 9\xi^4 -6u^2 \xi^2 + 18\xi^2 -3u^4 -26u^2 -7 ) l(\xi , u ) \bigg\ } \ , . \end{aligned}\ ] ] the analytical evaluation of the `` non - factorizable '' parts from diagrams ( g ) terminates with two non - elementary integrations and we find the following representation for their contribution to the nuclear spin - orbit strength : @xmath169 \big[x - l(\xi , x)\big ] \nonumber \\ & & + \bigg [ { 1\over 2 } \ln(1 + \sigma^2 ) + 4\xi y \arctan\sigma + 2u^2 - 2\xi^2 -{5\over 2 } \sigma^2 \bigg ] \ln{|x+\xi y|\over |x-\xi y|}\nonumber \\ & & + \bigg [ 4\xi y \arctan \sigma+ { 1\over 2 } ( 1+\xi^2-x^2 ) \ln(1 + \sigma^2 ) + 2u^2 - 2\xi^2 \nonumber \\ & & + { \sigma^2 \over 2 } ( x^2 - 5-\xi^2 ) \bigg ] { 1\over r } \ln { |x r+ ( x^2 - 1-\xi^2)\xi y| \over |x r+ ( 1+\xi^2-x^2)\xi y| } \bigg\}\ , , \end{aligned}\ ] ] with the auxiliary function @xmath170 . at zero nucleon momentum @xmath121 eq.(23 ) turns into the ( singularity free ) form : @xmath171 \nonumber \\ & & + y\bigg [ \ln(1+\sigma^2 ) -\ln(1 + u^2-\xi^2)-2\xi y \sigma + { 2\xi^2 ( 1+\xi^2)(u^2-\xi^2 ) \over 1+u^2-\xi^2 } \bigg ] \nonumber \\ & & + 2\xi(1+\xi^2)(1+y^2 ) \big [ \sqrt{u^2 -\xi^2 } -\arctan\sqrt{u^2-\xi^2 } \big]\bigg\}\ , . \end{aligned}\ ] ] note again the similarity of the expressions in eqs.(19,21,23 ) with the three - body potential given in eq.(13 ) of ref.@xcite . we also like to emphasize that the techniques used to reduce six - dimensional principal value integrals over the product of two fermi - spheres of radius @xmath26 to at most double integrals have been checked rigorously in ref.@xcite via the hugenholtz - van - hove theorem . let us end this section with power counting considerations for the spin - dependent interaction energy @xmath84 . counting the quantities @xmath172 and @xmath173 collectively as small momenta , we deduce that the relativistic @xmath22-correction from the @xmath2-exchange fock graph is of fifth order , while all contributions from iterated @xmath2-exchange are of fourth order in small momenta . irreducible @xmath14-exchange gives also rise to a fifth order contribution to the ( real part of the ) single - particle potential ( see eq.(16 ) in ref.@xcite ) . to the spin - orbit nn - amplitude irreducible @xmath14-exchange contributes as a ( higher - order ) relativistic @xmath24-correction ( see eqs.(22,23 ) in ref.@xcite ) and it will therefore enter @xmath174 at sixth order in small momenta . we have also checked that the relativistic @xmath24-corrections to iterated @xmath2-exchange start to contribute to @xmath84 first at sixth order in small momenta . from all that we can conclude that the present calculation of the nuclear spin - orbit strength @xmath0 is complete up - to - and - including third order in small momenta . strictly speaking , this counting argument applies only to the long - range effects induced by chiral @xmath2- and @xmath14-exchange since it does not cover the possible short - range contribution , @xmath175 , which is of course also of third order in small momenta . from this point of view the @xmath176-exchange term eq.(5 ) ( together with the fock contributions ) provides a model for the a priori undetermined low - energy constant @xmath177 . in order to learn about the convergence of the chiral expansion of the nuclear spin - orbit strength one should calculate the contributions from irreducible @xmath14-exchange of fourth order in small momenta . work along this line is in progress . furthermore , we note that spin - orbit strength generated by the three - body force diagrams in fig.2 of ref.@xcite is ( formally ) of higher order in small momenta . these should also be evaluated with the formalism introduced in section 2 . [ cols="^,^,^,^,^,^,^,^,^,^",options="header " , ] for the numerical evaluation of the nuclear spin - orbit strength @xmath0 we use consistently the same parameters as in our previous works @xcite . we choose the value @xmath178 for the nucleon axial vector coupling constant . the weak pion decay constant has the value @xmath179mev and @xmath180mev and @xmath181mev are the masses of the nucleon and the ( neutral ) pion , respectively . in the second row of table1 , we present numerical values for the contributions of individual diagrams to the spin - orbit strength @xmath51 at nuclear matter saturation density @xmath3mev . as expected the relativistic @xmath22-correction from the @xmath2-exchange fock graph is a very small @xmath115 effect . the contributions of individual iterated @xmath2-exchange diagrams are surprisingly large . in several cases they even exceed the empirical value @xmath182mev@xmath6 in magnitude and moreover they are of alternating signs . the basic reason for these large values is the large scale enhancement factor @xmath23 ( the nucleon mass ) entering the iterated @xmath2-exchange . the proportionality factor @xmath23 stems from the energy denominator of such second - order diagrams which is a difference of small nucleon kinetic energies . adding up the entries in the second row of table1 one gets @xmath183 , which is in perfect agreement with the empirical value of the nuclear spin - orbit strength @xmath184mev@xmath185 @xcite . the predicted total sum is dominated by the contribution @xmath186 of the four hartree diagrams ( a ) , ( b ) , ( c ) and ( d ) ( see fig.2 ) . interestingly , the same feature , namely the numerical suppression of the iterated @xmath2-exchange fock diagrams against the hartree diagrams , holds also for the ( spin - independent ) nuclear mean - field @xmath34 studied in ref.@xcite . the novel spin - orbit strength generated ( almost completely ) by iterated @xmath2-exchange is neither of relativistic nor of short range origin . it is in fact linearly proportional to the nucleon mass @xmath23 and its inherent range is the pion compton wavelength @xmath187fm . the latter feature tempts to an unconventional interpretation of the strong nuclear spin - orbit interaction . the potential @xmath7 underlying the empirical spin - orbit strength @xmath188mev@xmath6 becomes a rather weak one , namely @xmath48mev , after the identification of the effective range @xmath49 with the pion compton wavelength , @xmath189fm , as suggested by the present calculation . as a side remark we consider @xmath190 in the chiral limit @xmath191 . in that case all occurring integrals can be performed analytically and we find the following simple expression : @xmath192 which gives at @xmath3mev the negative value @xmath193mev@xmath6 . in fig.4 , we show the pion mass dependence of the nuclear spin - orbit strength @xmath51 at nuclear matter saturation density ( and zero nucleon momentum ) . one observes a sign change and a pronounced maximum at @xmath194mev . this strong @xmath173-dependence of @xmath51 has its origin in the alternating signs of the large contributions from individual diagrams as well as their own specific @xmath173-dependence . for comparison , we show in fig.5 the pion mass dependence of the depth of the ( real ) single - particle potential @xmath195 calculated in ref.@xcite . one observes a weak variation of the potential depth @xmath195 by at most @xmath196 if @xmath173 runs from zero ( chiral limit ) to the physical value @xmath181mev . with increasing pion mass the attractive nuclear mean - field generated by chiral @xmath14-exchange gets however soon lost and it turns into repulsion above @xmath197mev . in fig.6 , we show by the full line the dependence of the calculated nuclear spin - orbit strength @xmath190 on the nucleon density @xmath198 . one observes in the region @xmath199@xmath33 an approximate linear growth of @xmath190 with the density as it is , for example , known from @xmath20- and @xmath21-exchange ( see eq.(5 ) ) . in a finite nucleus the spin - orbit force acts mainly on the surface where the density gradients are largest and the density has dropped to about half of the central density . the replacement @xmath200 ( instead of @xmath201 valid for weakly inhomogeneous nuclear matter ) describes then more realistically the situation for a finite nucleus . the dashed line in fig.6 shows the spin - orbit strength which results if the contributions from the diagrams with two medium insertion ( b ) , ( c ) , ( d ) , ( f ) and ( g ) are weighted with a factor @xmath202 . this different weighting leads to a substantial reduction of the total spin - orbit strength such that only about @xmath203 of the empirical value @xmath204mev@xmath6 are left at nuclear matter saturation density @xmath205@xmath33 . in fig.7 , we show by the full line the dependence of the calculated nuclear spin - orbit strength @xmath12 at saturation density @xmath206mev on the nucleon momentum @xmath11 for @xmath63 . one observes a very strong @xmath11-dependence which leads even a sign change of @xmath207 above @xmath13mev . again , this strong @xmath11-dependence of @xmath12 has its origin in the alternating signs of the large contributions from individual diagrams as well as their own specific @xmath11-dependence . the numerical values in the second and third row of table1 indicate how these contributions from individual diagrams change with the nucleon momentum from @xmath4 to @xmath208 . one also should keep in mind that the scale relevant for momentum dependences is here set by the pion mass , @xmath209mev . note that the ratio @xmath210 changes by two units from @xmath4 to @xmath211 . furthermore , the dashed line in fig.7 corresponds to the weighting of diagrams with two medium insertions with a factor 1/2 . it shows the same strong @xmath11-dependence including a sign change as the full line in fig.7 . let us briefly summarize our results . we have demonstrated here that the nuclear spin - orbit interaction is not necessarily a relativistic effect . iterated @xmath2-exchange ( i.e. loosely speaking , the @xmath2-exchange spin - spin and tensor force in second order ) generates large nuclear spin - orbit terms which in fact scale linearly with the nucleon mass @xmath23 . for weakly inhomogeneous nuclear matter and small nucleon momenta ( @xmath212mev ) the spin - orbit strength from iterated @xmath2-exchange agrees ( at saturation density ) well with the empirical shell model value @xmath213mev@xmath6 . the strong @xmath11-dependence of @xmath12 ( including a sign change ) and the different weighting of @xmath214 and @xmath215 at the nuclear surface leave however questions about the ultimate relevance of this @xmath14-exchange spin - orbit interaction for real nuclear structure . nuclear structure calculation which use the calculated spin - orbit strength @xmath0 as input are necessary in order to clarify the role of the spin - orbit interaction generated by @xmath14-exchange . in this section we generalize our previous calculation @xcite of the single - particle potential to isospin asymmetric ( homogeneous ) nuclear matter . any relative excess of neutrons over protons in the nuclear medium leads to a different single - particle potential for a proton and a neutron . this fact is expressed by the following decomposition of the single - particle potential in isospin asymmetric nuclear matter : @xmath216 \ , \tau_3 \ , \delta + { \cal o}(\delta^2 ) \,.\ ] ] here , @xmath217 is the ( complex ) single - particle potential in isospin symmetric nuclear matter . the term linear in the isospin asymmetry parameter @xmath218 defines the ( complex ) isovector single - particle potential @xmath219 , and @xmath220 for a proton or a neutron . the situation of an isospin asymmetric nuclear medium is realized by the simple substitution : @xmath221 in the medium insertion eq.(4 ) . here , @xmath222 denote the ( different ) fermi momenta of protons and neutrons . differences in comparison to the calculation of @xmath217 in ref.@xcite occur only with respect to isospin factors and the radii of the fermi - spheres , @xmath223 . in practise the isovector single - particle potential @xmath219 is obtained by differentiating the @xmath224-components of the diagrammatic expressions with respect to @xmath225 at @xmath226 . \i ) @xmath2-exchange fock diagram in fig.1 including the relativistic @xmath22-correction : @xmath228[1+(u - x)^2 ] } \bigg ] \bigg\ } \ , , \end{aligned}\ ] ] with @xmath110 and @xmath111 . the auxiliary function @xmath229 has been defined in eq.(7 ) . with @xmath232 and the auxiliary function @xmath233 has been defined after eq.(11 ) . @xmath234 ^ 2 } \bigg\ { s'(2\xi + x y ) \bigg[\xi \ln{u+\xi \over u-\xi}-2u\bigg ] \nonumber \\ & & -{2 s ' \over 1+(xy+\xi)^2}\bigg [ 2u \xi + ( u^2-\xi^2 ) \ln{u+\xi \over u-\xi } \bigg]- ( 2s+s'\ , ) u^2 \ln{u+\xi \over u-\xi } \bigg\ } \ , , \end{aligned}\ ] ] @xmath235 \nonumber \\ & & + \bigg[{\sigma^2\over ( 1+\sigma^2)^2 } ( 5\sigma^4 + 9\sigma^2 + 6 -4\sigma^3 \sigma')-6 \ln(1+\sigma^2)\bigg]\ln{|x+ \xi y|\over |x- \xi y| } \bigg\}\ , , \end{aligned}\ ] ] with @xmath236 and the auxiliary function @xmath20 has been defined after eq.(15 ) . the symbol @xmath237 stands for a principal value integral . \iii ) iterated @xmath2-exchange fock graphs in fig.3 : @xmath238 \bigg\ } \ , , \end{aligned}\ ] ] @xmath239 \bigg\ } \ , , \end{aligned}\ ] ] with the auxiliary function @xmath240[1+(u - x)^2 ] \,l(x , u)$ ] . @xmath241 \nonumber \\ & & + 5u \int_0^u d\xi \,\bigg [ { 1\over \xi}(1+x^2-\xi^2 ) l(x,\xi ) -1 \bigg ] \,{\partial g(\xi , u ) \over \partial u } \bigg\ } \ , , \end{aligned}\ ] ] @xmath242 \nonumber \\ & & \times \bigg [ \ln{|x+ \xi y|\over |x- \xi y| } + { 1 \over r } \ln{|x r + ( x^2-\xi^2 - 1 ) \xi y|\over|x r + ( \xi^2 + 1-x^2 ) \xi y| } \bigg ] + \xi \big[\ln(1+\sigma^2)- \sigma^2\big ] \nonumber \\ & & \times \bigg[{r ' \over r^2 } \ln{|x r + ( x^2-\xi^2 - 1 ) \xi y|\over|x r + ( \xi^2 + 1-x^2 ) \xi y| } + { y(1-x^2 + 3\xi^2 ) -x r ' \over r [ x r+(x^2-\xi^2 - 1 ) \xi y]}\nonumber \\ & & + { y(1-x^2 + 3\xi^2 ) + x r ' \over r [ x r + ( \xi^2 + 1-x^2 ) \xi y ] } - { 2 x y \over x^2 - \xi^2 y^2 } \bigg ] \bigg\}\ , , \end{aligned}\ ] ] with @xmath243 and the auxiliary function @xmath244 has been defined after eq.(23 ) . \iv ) irreducible @xmath14-exchange hartree and fock graphs : @xmath245 @xmath246 \xi \sqrt{1+\xi^2}\ln(\xi+\sqrt{1+\xi^2})\nonumber \\ & & + ( 17 + 242 g_a^2 - 787 g_a^4 ) \xi^2 -(3 + 14 g_a^2 + 15g_a^4 ) \xi^4 \nonumber \\ & & + \big [ 60(1 + 6g_a^2 - 15 g_a^4 ) \xi^2 + 4(1 + 10 g_a^2 - 59g_a^4 ) \xi^4 \big ] \ln{m_\pi \over 2\lambda } \ , . \end{aligned}\ ] ] \v ) power divergences specific for cut - off regularization : @xmath247 \,.\ ] ] the term linear linear in the cut - off @xmath40 stems from iterated @xmath2-exchange with a contribution of the hartree diagram ( a ) and the fock diagram ( e ) in the ratio @xmath248 . the term quadratic in the cut - off @xmath40 , on the other hand , originates from irreducible @xmath14-exchange . note that the @xmath11-independent contribution to @xmath227 in eq.(39 ) is just twice its contribution to the asymmetry energy @xmath57 ( see eq.(29 ) in ref.@xcite ) . this relative factor of 2 is typical for a momentum independent nn - contact interaction , to which the power divergences are completely equivalent , as emphasized in ref.@xcite . we use consistently the value @xmath249mev @xmath250 of the cut - off scale which has been fine - tuned in ref.@xcite to the binding energy per particle , @xmath251mev . in fig.8 , we show by the full line the total ( real ) isovector single particle potential @xmath252 of our calculation for a nucleon at rest ( @xmath4 ) as a function of the nucleon density @xmath253 . the shape of this curve is very similar to the asymmetry energy @xmath57 ( see fig.7 in ref.@xcite ) . in comparison to @xmath57 the scale on the ordinate is stretched by a factor of about 1.4 . interestingly , the ( real part of the ) isovector single - particle potential @xmath252 has its maximum close to the saturation density @xmath58@xmath33 . the actual value at that point is @xmath254mev . this prediction is comparable to the value @xmath255mev @xcite used in shell model calculations or the value @xmath256mev @xcite deduced from nucleon - nucleus scattering in the framework of the optical model . the dashed line in fig.8 shows the density dependence of the ( real ) isovector single - particle potential @xmath257 at the fermi surface @xmath17 . at that point the ( real ) isovector single - particle potential comes out always more repulsive than at @xmath4 . note also that the ( possibly unrealistic ) downward bending branches of the curves in fig.8 start at densities higher than those relevant for conventional nuclear physics . furthermore , we show in fig.9 the momentum dependence of the ( real ) isovector single - particle potential @xmath62 at saturation density @xmath258mev . the @xmath11-dependence of @xmath62 is non - monotonic in the interval @xmath63 . one observes a broad maximum at @xmath64mev where @xmath62 has increased by about 30% to the value @xmath259mev . note that in comparison to the spin - orbit strength @xmath12 shown in fig.7 the @xmath11-dependence of the real part of the isovector single - particle potential @xmath62 is very moderate . most useful would be nuclear structure calculations using the calculated isovector single particle potential @xmath227 as input . in this subsection , we discuss the imaginary part @xmath66 of the isovector single - particle potential . according to eq.(26 ) , it splits and shifts the half - width of a neutron - hole state and a proton - hole state in the fermi - sea ( with momentum @xmath260 ) by the amount @xmath261 . within the present calculation the imaginary part @xmath66 arises entirely from iterated one - pion exchange . it is advantageous to present analytical formulas only for complete classes of diagrams . we find from the iterated @xmath2-exchange hartree diagrams in fig.2 ( including the analogous graph with three medium insertions ) the following contribution to the imaginary isovector single - particle potential : @xmath262 \nonumber \\ & & - 2u^2 \big(1+{u\over x}\big ) \ln[1+(u+x)^2]+2u^2 \big({u\over x}-1\big ) \ln[1+(u - x)^2 ] \nonumber \\ & & + \int_{-1}^1 dy \,\bigg\ { 2u^2 \bigg [ { 2s^2+s^4 \over 1+s^2}-2\ln(1+s^2 ) \bigg ] + { ( x^2-u^2 ) s^5 s'\over 2(1+s^2)^2 } \nonumber \\ & & + \int_0^u d\xi\,\xi^2 \bigg[\delta(x-\xi|y|)\bigg(2\ln(1+\sigma^2)- { 2\sigma^2 + \sigma^4 \over 1+\sigma^2}\bigg ) + \theta(x-\xi|y| ) \nonumber \\ & & \times{1\over x } \bigg({\sigma^2 \over 1+\sigma^2}(6 + 9\sigma^2 + 5\sigma^4 -4 \sigma^3 \sigma'\,)-6\ln(1+\sigma^2)\bigg)\bigg ] \bigg\}\bigg\ } \,.\end{aligned}\ ] ] the iterated @xmath2-exchange fock diagrams in fig.3 ( again including the analogous graph with three medium insertions ) lead , on the other hand , to the following expression : @xmath263 \over \sqrt{(1+x^2-u^2)^2 + 4(u^2- x^2z^2)}}+{5u^2 \over x}\int_{(u - x)/2}^{(u+x)/2 } d\xi \ , { 1 + 4\xi^2 \over 1 + 2\xi^2}\nonumber \\ & & \times \ln(1 + 4\xi^2)+ \int_{-1}^1 dy \ , \bigg\{\int_{-1}^1 dz \ , { \theta(1-y^2-z^2)\over \pi \sqrt{1-y^2-z^2}}\,{s^3 s ' \over 1+s^2 } \big [ t^2-\ln(1+t^2)\big]\nonumber \\ & & + \int_0^u d\xi\ , { \xi^2 \over x } \,\bigg [ \theta(x-\xi |y| ) { \xi r'\over r^2}\big(\sigma^2-\ln(1 + \sigma^2)\big ) + \bigg(1-{1\over r}\bigg ) \bigg[\theta(x-\xi |y|)\nonumber \\ & & \times \bigg({\sigma^2\over 1+\sigma^2 } \big[\sigma \sigma ' \,[8 - 10\,\theta ( \xi - x ) ] + 3 + 5 \sigma^2 \big ] -3 \ln(1+\sigma^2 ) \bigg ) \nonumber \\ & & + x\ , \delta(x-\xi |y| ) \big(\ln(1+\sigma^2)-\sigma^2\big ) - { 2u^2 \sigma_x^3 \ , \theta(x-\xi ) \over ( \sigma_x- \xi y ) ( 1+\sigma_x^2 ) } \bigg]\bigg ] \bigg\}\bigg\ } \ , , \end{aligned}\ ] ] with the auxiliary function @xmath264 . if the delta - function @xmath265 is used to eliminate the @xmath266-integration in eqs.(40,41 ) the remaining @xmath267-integral extends over the restricted region @xmath268 . the sum of both contibutions eqs.(40,41 ) evaluated at zero nucleon momentum @xmath121 can even be written as a closed form expression : @xmath269 finally , we show in fig.10 the momentum dependence of the imaginary isovector single - particle potential @xmath270 at saturation density @xmath258mev . the associated value at zero nucleon momentum , @xmath271mev , agrees within @xmath196 with the isoscalar half - width @xmath272mev found in ref.@xcite . as a consequence of the decreasing phase space available for redistribution of a nucleon - hole state s energy , the curve in fig.10 drops with momentum @xmath11 and @xmath270 reaches zero at the fermi - surface @xmath211 . the exact vanishing of @xmath66 at the fermi - surface @xmath17 is even separately true for the class of iterated @xmath2-exchange hartree diagrams and the class of iterated @xmath2-exchange fock diagrams . the conditions @xmath273 and @xmath274 serve as an excellent ( analytical and numerical ) check on the involved calculations leading to eqs.(40,41 ) . in summary , we find that the predictions from chiral @xmath2- and @xmath14-exchange for the real part of the isovector single - particle potential @xmath227 agree fairly well with empirical values . the calculated imaginary part @xmath66 fulfills the constraints imposed by luttinger s theorem @xcite . 99 o. haxel , j.h.d . jensen and h.e . suess , _ phys . * 75 * ( 1949 ) 1766 ; m. goeppert - mayer , _ phys . * 75 * ( 1949 ) 1969 . serot and j.d . walecka , _ adv . * 16 * ( 1986 ) 1 ; and refs . therein . p. ring , _ prog . * 37 * ( 1996 ) 193 ; and refs . therein . s. typel and h.h . wolter , _ nucl . phys . _ * a656 * ( 1999 ) 331 ; and refs . therein . f. hofmann , c.m . keil and h. lenske , _ phys . rev . _ * c64 * ( 2001 ) 034314 ; and refs . therein . n. kaiser , r. brockmann and w. weise , _ nucl . phys . _ * a625 * ( 1997 ) 758 . n. kaiser , s. fritsch and w. weise , _ nucl . phys . _ * a697 * ( 2001 ) 255 . for an earlier similar approach see : m. lutz , b. friman and c. appel , _ phys . * b474 * ( 2000 ) 7 . n. kaiser , s. fritsch and w. weise , _ nucl . phys . _ * a700 * ( 2001 ) 343 . s. fritsch , n. kaiser and w. weise , nucl - th/0202005 , submitted to _ phys . * b*. r. brockmann and r. machleidt , _ phys . rev . _ * c42 * ( 1990 ) 1965 . a. bohr and b.r . mottelson , nuclear structure , vol.i , benjamin , ( 1969 ) , chapt.2.4 . g. eder , kernmaterie , spektrum akademischer verlag , ( 1995 ) , chapt.4.1 . hodgson , growth points in nuclear physics , vol.3 , pergamon press , ( 1981 ) , chapt.2 . luttinger , _ phys . _ * 121 * ( 1961 ) 942 . dreizler and e.k.u . gross , density functional theory , springer verlag , ( 1990 ) , chapt.5 . l. mnchow and r. reif , recent developments in the nuclear many - body problem , vol.1 , teubner verlagsgesellschaft , leipzig , ( 1985 ) , chapt.4.6 . pieper , v.r . pandharipande , r.b . wiringa and j. carlson , _ phys . rev . _ * c64 * ( 2001 ) 014001 .

we derive anomaly constraints for abelian and non - abelian discrete symmetries using the path integral approach . we survey anomalies of discrete symmetries in heterotic orbifolds and find a new relation between such anomalies and the so - called ` anomalous ' u(1 ) .

0805.0207

symmetries play a key role in the understanding of fundamental laws of physics . apart from continuous , in particular gauge , symmetries , discrete symmetries provide a useful tool in field - theoretic model building and arise often in top - down models . very much like continuous symmetries , discrete symmetries can be broken by quantum effects , i.e. have an anomaly @xcite . if this is the case , one expects that the corresponding conservation laws be violated through non - perturbative effects . the criteria for discrete symmetries to be non - anomalous , and thus to be exact , have been extensively studied in the abelian ( @xmath0 ) case @xcite . anomaly criteria for non - abelian discrete symmetries have been discussed first in specific examples @xcite . here , we use the path integral approach @xcite to derive anomaly constraints on non - abelian discrete symmetries . we follow the discussion of @xcite , and extend it such as to include gravitational anomaly constraints . we further re - derive the conditions for abelian discrete symmetries to be anomaly - free , using the path integral method . this derivation allows for an alternative , perhaps more intuitive understanding of the criteria , which does not rely on contributions from heavy states . we explore the issue of discrete anomalies in string compactifications , focusing on heterotic orbifolds . the question we seek to clarify is whether discrete anomalous symmetries can appear in string - derived models @xcite . the discrete symmetries on orbifolds reflect certain geometrical symmetries of internal space . since the geometrical operations , i.e. space group transformations , are embedded into the gauge group , one might suspect that the discrete anomalies are related to gauge anomalies . we find that this is indeed the case , specifically we find that the so - called ` anomalous ' 1 , which occurs frequently in heterotic orbifolds , determines the anomalies of discrete symmetries . the paper is organized as follows . in section [ sec : anomalyfreediscretesymmetries ] we first re - derive anomaly constraints for abelian discrete symmetries and then derive the constraints for non - abelian discrete symmetries , using the path integral approach . in section [ sec : orbifoldanomalies ] we consider heterotic orbifolds and identify a geometric operation on the orbifold , which we would like to refer to as ` anomalous space group element ' , as the source of all discrete anomalies . section [ sec : summary ] contains our conclusions . we also include four appendices where we present the calculation of anomalies of the dihedral group @xmath1 ( [ app : d4 ] ) , @xmath1 anomalies in a concrete model from the literature ( [ app : grimuslavoura ] ) and the anomaly coefficients in two concrete string models ( [ sec : krzanomalies ] & [ sec : bhlranomalies ] ) . consider a theory described by a lagrangean @xmath2 with a set of fermions @xmath3 $ ] , where @xmath4 denotes a field transforming in the irreducible representation ( irrep ) @xmath5 of all internal symmetries . a general transformation @xmath6 or , more explicitly , @xmath7 ~\to~ \left(\begin{array}{ccc } u^{(1 ) } & & 0\\ & \ddots & \\ 0 & & u^{(m ) } \end{array}\right)\ , \left[\begin{array}{c } \psi^{(1)}\\ \vdots\\ \psi^{(m)}\end{array}\right]\;,\ ] ] which leaves @xmath2 invariant ( up to a total derivative ) denotes a classical symmetry . by noether s theorem , _ continuous _ symmetries imply , at the classical level , conserved currents , @xmath8 . for instance , in the case of an abelian continuous symmetry one can define the charge @xmath9 which satisfies the conservation law @xmath10 . in the case of a discrete symmetry , the situation is similar . consider , for simplicity , an abelian discrete symmetry , i.e. @xmath11 . under this symmetry , the fermions of the theory transform as @xmath12 where ( by convention ) the discrete charges @xmath13 are integer and only defined modulo @xmath14 . if is a symmetry of @xmath2 , the corresponding charge @xmath13 is conserved modulo @xmath14.-parity , which implies that superpartners can only be produced in pairs . ] classical chiral symmetries can be broken by quantum effects , i.e. have an anomaly . specifically , consider a chiral transformation @xmath15 where @xmath16 with @xmath17 denoting the generators of the transformation , and @xmath18 is the left - chiral projector . it is well known that at the quantum level the classically conserved current @xmath19 is not necessarily conserved any more , that is ( cf . e.g.@xcite ) @xmath20 the anomaly @xmath21 can be derived using fujikawa s method , i.e. by calculating the transformation of the path integral measure @xcite , which in our case reads @xmath22 where the jacobian of the transformation is given by @xmath23 the anomaly function @xmath24 decomposes into a gauge and a gravitational part @xcite , @xmath25 the gauge part @xmath26 corresponds to the triangle diagram @xmath27gauge gauge ( figure [ fig : trianglediagrams ] ) . this anomaly is given by @xmath28\;. \label{eq : agauge}\ ] ] here @xmath29 $ ] is the field strength of the gauge symmetry , such that @xmath30 for a 1 symmetry , and @xmath31 denotes its dual . the trace ` tr ' runs over all internal indices . analogously , the gravitational part @xmath32 is the mixed @xmath27gravity gravity anomaly . it is known that it takes the form @xcite @xmath33\ ; , \label{eq : agrav}\ ] ] where the summation runs over the ( spin-1/2 ) fermions in the representations @xmath5 . the subscript ` @xmath34 ' indicates that each representation @xmath5 appears only once in the sum . @xmath35 denotes @xmath36 in the representation @xmath5 , and might therefore be thought of as a @xmath37 matrix such that ` @xmath38 ' is the standard ( matrix ) trace . the contribution of a single weyl fermion to the gravitational anomaly is given by @xcite @xmath39 to evaluate the anomaly , we split the set of all generators @xmath17 into generators of continuous symmetries @xmath40 and those of discrete symmetries @xmath41 . therefore , we shall discuss separately the two cases : 1 . anomalies of continuous symmetries with @xmath42 ; 2 . anomalies of discrete symmetries with @xmath43 . note that we ( implicitly ) assume that all symmetries are gauged . for the evaluation of the anomalies , it is useful to recall the powerful index theorems @xcite , which imply [ eq : indextheorems ] @xmath44 & \in & \mathbbm{z } \ ; , \label{eq : index1}\\ * \frac{1}{2}\,\int\!{\mathrm{d}}^4 x\ , \frac{1}{384 \pi^2}\,\frac{1}{2}\ , \varepsilon^{\mu\nu\rho\sigma}\ , r_{\mu\nu}{}^{\lambda\gamma}\ , r_{\rho\sigma\lambda\gamma } & \in & \mathbbm{z } \;,\label{eq : index2}\end{aligned}\ ] ] where @xmath40 are in the fundamental representation of a particular gauge factor @xmath45 . note that in our conventions @xmath46=\frac{1}{2}\delta_{ab}$ ] . the factor @xmath47 in eq . follows from rohlin s theorem @xcite , as discussed in @xcite . we start by reviewing the anomaly constraints for the continuous symmetries . they arise from demanding that @xmath48 vanish for arbitrary @xmath49 in order for the jacobian @xmath50 to be trivial . consider first the mixed @xmath51 anomaly , where @xmath45 is a non - abelian gauge factor with generators @xmath40 . representations under @xmath45 are denoted by @xmath52 . this anomaly can be related to diagram [ fig : anomalyu1-g - g ] . from equation and the index theorem , one finds that it only vanishes if @xmath53 } a_\mathrm{{\ensuremath{\mathrm{u}(1)}}-g - g}~\equiv~ \sum_{\boldsymbol{r}^{(f ) } } q^{(f)}\,\ell(\boldsymbol{r}^{(f)})~=~0\;.\ ] ] in analogy to equation , ` @xmath54 ' means that each representation @xmath52 is only summed once . @xmath55 denote the respective 1 charges . the dynkin indices @xmath56 are defined by @xmath57 \;.\ ] ] our conventions are such that @xmath58 for @xmath59 and @xmath60 for @xmath61 . consider next the @xmath62 anomaly , equation . from the index theorem , it vanishes if @xmath63 the sum ` @xmath64 ' indicates a plain summation over all fermions . in summary , we see that the continuous symmetries are non - anomalous if and only if the jacobian is trivial for arbitrary @xmath27 . now consider a discrete symmetry , i.e. @xmath43 where , by convention , @xmath66 takes only the discrete values @xmath67 and the eigenvalues of @xmath41 are integer . like before , we demand that @xmath50 be trivial . it is now important to note that the jacobian can also be trivial for non - zero arguments of the exponential . let us specify the conditions for this to happen . consider first the abelian case , i.e. a @xmath11 symmetry with @xmath68 . from the gauge and gravitational parts of the anomaly function , equations and , and the index theorems , we see that the jacobian is trivial if @xmath69 the factor 2 in front of the dynkin index in is due to our conventions ( @xmath46=\frac{1}{2}\delta_{ab}$ ] ) . this means that the constraints for a @xmath11 symmetry to be anomaly - free are [ eq : znconditions ] @xmath70 if @xmath14 is odd , we can always make the @xmath11 charges even by shifting them by integer multiples of @xmath14 . this explains why the sums in and can always be made integer . hence in the case of an odd @xmath14 one can replace @xmath71 by @xmath14 after a suitable shift of the charges . the constraints and coincide with the ones of the literature @xcite . we would like to emphasize that , in our derivation , we did not invoke the contributions from heavy majorana fermions . condition for even @xmath14 has been justified as follows @xcite : one can always introduce majorana fermions @xmath72 with @xmath0 charges @xmath71 . their contribution to the sum is @xmath71 , on the other hand the majorana mass term @xmath73 is allowed by the discrete symmetry . since @xmath34 can be arbitrarily large , @xmath72 can be ` removed from the theory ' . while the argument leads to the correct result , one might nevertheless wonder if the anomaly conditions change if one considers more constrained settings ( such as string - derived theories ) where extra degrees of freedom can not be introduced at will . our derivation shows that the anomaly conditions remain unchanged . ] rather , the anomaly constraints ( including the condition mod@xmath74 ) are a consequence of the index theorems and follow from demanding that the jacobian be trivial . we also note that in our approach one immediately sees that there are no cubic anomaly constraints for discrete symmetries , which is in agreement with @xcite . we now turn to non - abelian discrete symmetries @xmath75 . consider a specific transformation @xmath76 . since we are considering a discrete symmetry , there is a positive integer @xmath14 such that @xmath77 , i.e. @xmath76 is generating a @xmath11 symmetry ; we take @xmath14 to be the smallest such integer . denote the ( discrete ) representations of @xmath75 by @xmath78 . moreover , an element @xmath79 in a representation @xmath78 is given by @xmath80 with @xmath81 and @xmath82 having integer eigenvalues . in the evaluation of the anomaly functions , equations and , we note that @xmath83 $ ] takes the role of the @xmath11 charge . this charge , denoted by @xmath84 , can be expressed in terms of the group elements @xmath85 as ( cf . @xcite ) @xmath86 ~=~n\,\frac{\ln \det u(\boldsymbol{d}^{(f)})}{2\pi\,{\mathrm{i}}}\;.\ ] ] as usual , the @xmath11 charges @xmath84 are defined modulo @xmath14 only ( such that they can consistently be expressed through the multi - valued logarithm ) . from the index theorems , we find that demanding that the jacobian be trivial amounts to requiring [ eq : nonabeliandiscreteanomalyconstraints1 ] @xmath87 where the sum ` @xmath88 ' indicates that only over representations is summed which are non - trivial w.r.t.both @xmath45 and @xmath75 ; the symbol ` @xmath89 ' in means that the sum extends over all non - trivial representations @xmath78 . these constraints have to be fulfilled for each discrete transformation @xmath76 separately . however , elements with @xmath90 do not lead to anomalies , cf . equation . non - abelian discrete groups @xmath75 have more than one element . assume that we have verified that the constraints are fulfilled for @xmath91 . it is then obvious that this implies that for both elements @xmath92 and @xmath93 equations hold as well . this means that in practice one only has to check anomaly constraints for the generators of @xmath75 . in appendix [ app : d4 ] , we discuss @xmath1 anomalies as an example for non - abelian discrete anomalies . in appendix [ app : grimuslavoura ] we present a sample calculation . the anomaly constraints for discrete symmetries can be summarized as follows : 1 . anomalies of @xmath11 symmetries + [ eq : znanomalyconstraints ] @xmath94 2 . anomalies of non - abelian discrete symmetries @xmath75 : one has to verify that for all generators of @xmath75 the following two equations hold + [ eq : nonabeliananomalyconstraints ] @xmath95 + here , the sum @xmath89 extends over all non - trivial representations of @xmath75 , @xmath84 is defined in equation , and @xmath14 denotes the order of the generator . now we turn to study the implications of an anomalous discrete symmetry . one might envisage several scenarios in which such a symmetry appears . in what follows , we focus on a particular one : we start with a so - called ` anomalous ' 1 and break it to a discrete subgroup . later , in section [ sec : orbifoldanomalies ] , where we investigate string - derived models , we will attempt to realize different situations . in a fundamental theory , anomalies of a continuous symmetry are not acceptable . however , there is the well - understood situation in which a 1 factor appears ` anomalous ' , i.e. the usual anomaly conditions seem not to be satisfied . this is the case when the anomaly is canceled by the green - schwarz ( gs ) mechanism @xcite . to discuss this scenario , consider a supersymmetric gauge theory . under the ( ` anomalous ' ) @xmath96 transformation , the chiral superfields @xmath97 containing the chiral fermions @xmath98 and the vector superfield @xmath99 transform as @xmath100 the anomaly is canceled by the transformation of the dilaton @xmath101 ( or possibly a different chiral field ) , which gets shifted under the @xmath96 transformation as @xmath102 where @xmath103 is proportional to the trace of the generator of @xmath96 , @xmath104 , ( see below ) . the tree - level khler potential for the dilaton is @xmath105 as usual , the kinetic terms for the scalar components of @xmath101 arise from the corresponding @xmath75-term , @xmath106_d$ ] , i.e. @xmath107 where @xmath108 and @xmath109 . consider now the axionic shift , @xmath110 the kinetic term is invariant under this shift when @xmath111 is constant . however , as the parameter @xmath111 depends on @xmath112 for @xmath96 transformations , the kinetic term is not invariant under @xmath96 . to make it invariant , we have to introduce the terms , @xmath113 and @xmath114 , in the stckelberg form . that implies the @xmath96-invariant khler potential for the dilaton is @xmath115 which also includes the @xmath116-dependent fayet - iliopoulos ( fi ) d - term . it is convenient to define a normalized @xmath96 generator @xmath117 , whose charges @xmath118 fulfill the consistency conditions ( cf.@xcite ) @xmath119 our conventions are such that @xmath103 is positive . ( as before , @xmath64 means plain summation . ) the mixed @xmath120 anomaly coefficients , as defined in , have to satisfy the consistency conditions @xmath121 where @xmath122 denotes the ka - moody level of @xmath45 . for the green - schwarz mechanism to work , this relation has to hold for all gauge group factors . the first question is whether @xmath96 can be used to forbid couplings . to answer this question , consider a product of fields , @xmath123 , with @xmath124 . in the case of a usual 1 symmetry , @xmath123 can not denote an allowed coupling . however , in the case of @xmath96 , this conclusion does not apply ; instead one finds that the non - perturbative coupling @xmath125 with an appropriate @xmath126 can be induced ( cf . @xcite ) . in other words , the field @xmath127 transforms under the 1 with a charge that is opposite to @xmath128 . that means that @xmath96 does not forbid products of fields with @xmath124 . what can one say about products of fields @xmath123 with @xmath129 ? here the answer is that an anomalous 1 implies the existence of a fi @xmath75-term ( cf . equation ) . to obtain a supersymmetric vacuum , the fi term has to be canceled . that is , certain fields with net negative anomalous charge have to attain a vev in the vacuum . multiplication of @xmath123 by such fields can lead to allowed couplings , hence in supersymmetric vacua couplings of the type @xmath123 will generically be allowed . given these considerations , it is also clear what happens if one breaks @xmath96 to a discrete , anomalous subgroup . since @xmath96 is violated by terms of the form , also the discrete subgroup is expected not to be exact . gets broken to a @xmath11 subgroup which , however , is non - anomalous by the criteria . here , either the terms appear nevertheless , or there is a subclass of terms , which are forbidden by the non - anomalous @xmath11 , and where the coefficient happens to be zero . that is , if the second possibility is true , non - perturbative effects break @xmath96 to an non - anomalous @xmath11 subgroup . to find out which situation is realized would be , by itself , an interesting question , which is , however , beyond the scope of this study . ] an anomaly of an abelian discrete symmetry does not necessarily signal an inconsistency of the model . symmetries might just be accidental or approximate , and , therefore , need not to be gauged . further , if the anomalies are universal , they can be canceled by a green - schwarz mechanism . in practice , this means that they are broken by the vevs of certain fields ; in addition there are non - perturbatively induced terms with hierarchically small coefficients , as in . these small corrections might turn out to be a virtue rather than a problem in concrete models . we conclude this section by commenting on supersymmetric texture zeros @xcite , which go sometimes also under the term ` susy zero mechanism ' . it is stated that , due to holomorphicity of the superpotential , an anomalous 1 symmetry can enforce absence of certain couplings even though the symmetry is broken in supersymmetric vacua , where the fi @xmath75-term is canceled . let us briefly review the argument : cancellation of the fi term requires certain field with certain , say negative , sign of ` anomalous ' charge to attain a vacuum expectation value ( vev ) . now one might envisage a situation in which only fields with non - positive charges get a vev . consider then a combination of some other fields , @xmath123 , which has total negative anomalous charge . to be neutral w.r.t . the @xmath96 symmetry , this combination needs to be multiplied by fields with positive @xmath96 charge . however , so the argument goes , those fields do not attain vevs , and hence @xmath123 can not denote an allowed coupling . that is , couplings of the type @xmath123 appear to be absent . on the other hand , in many applications of the ` susy zero mechanism ' it is not possible to specify a symmetry that forbids those couplings . with what we have discussed above , we are able to resolve the puzzle : @xmath127 carries positive charge and hence couplings of the form @xmath130 can arise . the induced effective coupling is suppressed ( so that , as far as textures are concerned , a ` zero ' can be a good approximation ) , however , in contrast to what is often assumed , in general it is not related to the scale of supersymmetry breakdown . an interesting question is whether discrete anomalies occur in top - down constructions , in particular in string compactifications @xcite . since string theory is believed to be uv complete , one would expect that there are no ( uncanceled ) anomalies in this framework . while this has been extensively checked for continuous gauge symmetries , the case of discrete symmetries is somewhat more subtle . construction of string models exhibiting discrete anomalies would lead to a playground in which the ` quantum gravity effects ' , which are commonly believed to spoil the discrete conservation laws , can be specified in somewhat more detail than usual . specifically , we study anomalies of discrete symmetries in heterotic orbifold models . in our presentation , we mainly focus on the 6-ii orbifold , yet in our computations we also considered different orbifolds , so that our results are more generally valid . we start with a very brief review on orbifolds , summarize the essentials of ( discrete ) string selection rules , continue by relating the so - called ` anomalous 1 ' to a discrete transformation in compact space , which we refer to as the ` anomalous space group element @xmath131 ' , and conclude by relating anomalies in discrete symmetries to the anomaly in the discrete transformation @xmath131 . a heterotic orbifold emerges by dividing a six - dimensional torus @xmath132 by one of its symmetries @xmath111 @xcite ( see @xcite for a recent review ) . @xmath132 can be parametrized by three complex coordinates @xmath133 ( @xmath134 ) . then we denote @xmath135 . for example , in 6-ii orbifolds one has @xmath136 . a model is defined by the compactification lattice , the twist vector @xmath137 , the shift @xmath99 and the wilson lines @xmath138 . given these data , the massless spectrum ( at the orbifold point ) is completely determined ( for recent explicit examples see e.g. @xcite ) . a rather common feature of these constructions is the occurrence of a so - called ` anomalous 1 ' , @xmath96 , @xcite ( cf.section [ sec : discreteanomalies ] ) , which implies that , at one - loop , a fi @xmath75-term is induced @xcite . as we shall see , the ` anomalous ' 1 plays a prominent role in the discussion of discrete anomalies . couplings on heterotic orbifolds are governed by certain selection rules @xcite ( see also @xcite ) , some of which can be interpreted as discrete symmetries of the effective field theory emerging as ` low - energy ' limit in these constructions . these symmetries fall into two classes , depending on whether they reflect space group rules or @xmath139-charge ( or @xmath140-momentum ) conservation . the space group selection rules are stated by @xmath141 where we label the states entering the coupling by @xmath142 . @xmath143 is the space group element representing the string boundary condition with @xmath144 integer , @xmath145 are lattice vectors defining @xmath132 , and ` @xmath146 ' means that the product on the l.h.s . lies in the same equivalence class as the identity element . the rotational part of gives rise to the point group selection rule , and here we refer to it as the @xmath122-rule , which in 6-ii orbifolds reads @xmath147 the translational part can be rewritten as [ eq : spacegrouprules ] @xmath148 the quantum numbers @xmath149 , @xmath150 and @xmath151 specify the localization of the states on the orbifold ; we follow the conventions of @xcite . the space group rules can be interpreted as @xmath152 flavor symmetries , denoted @xmath153 in what follows . under this symmetry , each state comes with two 2 charges and one 3 charge , [ eq : qnspacegroup ] @xmath154 in models where certain wilson lines are absent , these symmetries combine with permutation symmetries of equivalent fixed points to non - abelian discrete flavor symmetries @xcite . as we are interested in anomalies , we focus on the abelian subgroups of these discrete symmetries ( cf . section [ sec : nonabeliandiscretanomalies ] ) . the discrete @xmath139-symmetries in 6-ii orbifolds based on the lie lattice @xmath156 are expressed by @xcite [ eq : h - momentumrules ] @xmath157 hereby , @xmath158 denotes the @xmath159 component of the @xmath140-momentum of the bosonic components of chiral superfields , @xmath160 where @xmath161 denote the so(6 ) shifted momenta of bosonic states and @xmath162 is the difference of oscillator numbers @xmath163 . for twisted sectors , it can be shown that @xmath164 , with @xmath165 being the smallest integer , such that @xmath166 . in orbifold constructions , @xmath167-duality transformations act as discrete reparametrizations of the moduli space . in general , there are three @xmath167-moduli @xmath168 ( @xmath134 ) , each of which corresponds to the @xmath169-th complex plane @xmath133 . for example , the moduli @xmath170 , @xmath171 and @xmath172 in 6-ii orbifolds correspond to the overall sizes of @xmath173 , @xmath174 and @xmath175 tori , respectively . modular symmetry is in a sense different from other symmetries , where moduli @xmath168 are singlets . under the modular symmetry , the moduli @xmath168 transform as @xmath176 where @xmath177 and @xmath178 . the khler potential @xmath179 of matter fields @xmath97 depends in general on the @xmath168 moduli as @xmath180 where the so - called modular weights @xmath181 are given by @xcite @xmath182 we require that the khler potential @xmath179 be invariant under . this implies that the matter fields with the modular weight @xmath181 transform under as the following chiral rotation : is not invariant under . @xmath167-duality invariance requires that the holomorphic superpotential @xmath183 transform as @xmath184 , such that the combination @xmath185 , which appears in the supergravity lagrangean , is invariant . ] once the @xmath168 attain vacuum expectation values , these symmetries are ( in general ) completely broken . that is , the @xmath167-duality symmetries are not expected to contribute to discrete symmetries which survive to low energies . according to the various discrete symmetries described in the previous subsections , we now define the corresponding anomaly coefficients . we further conduct a scan over many models , based on several orbifold geometries , and elicit whether there the symmetries of subsection [ sec : discreteorbifoldsymmetries ] are anomalous or not . let us start by studying anomalies in the @xmath188 symmetries . a special class of @xmath188 anomalies is given by @xmath189 where the sum extends over all non - trivial representations @xmath52 of a non - abelian gauge factor @xmath45 and the @xmath190 are defined in . @xmath191 is only defined up to twice the smallest non - vanishing dynkin index @xmath192 that appears , i.e. up to 1 if fundamental representations of @xmath193 groups are present . we have investigated various heterotic orbifolds , and find that , in general , they exhibit flavor anomalies ( see appendices [ sec : krzanomalies ] and [ sec : bhlranomalies ] for specific examples ) . the @xmath139 anomalies are given by @xcite @xmath194 with @xmath195 denoting the quadratic casimir . the sum extends over all irreps @xmath52 denoting the representation of the field @xmath196 w.r.t . the gauge factor @xmath45 . the discrete @xmath139 charges in this orbifold are only defined modulo @xmath197 . therefore , the anomalies can only be specified up to @xmath197 times twice the smallest non - vanishing dynkin index @xmath198 appearing in the sum in . we find empirically that the @xmath139 anomalies are not universal ( for specific examples see appendices [ sec : krzanomalies ] and [ sec : bhlranomalies ] ) . by considering the one - loop effective supersymmetric lagrangean , one finds that the gauge coupling constant is not invariant under the discrete modular group of @xmath167-duality transformations . the coefficients of this @xmath167-duality anomaly are given by @xcite @xmath200 where @xmath201 denotes the modular weight of the state @xmath97 w.r.t . the plane @xmath169 ( cf . ) . as is well known , @xmath167-duality anomalies can be canceled in two different ways . one part of it is removed by the green - schwarz mechanism whereas a second part only disappears after considering one - loop threshold corrections to the gauge coupling constants . only universal anomalies , i.e. those @xmath202 in with fixed @xmath169 whose values do not depend on @xmath45 , can be canceled by the green - schwarz mechanism . in contrast , cancellation of non - universal @xmath167-duality anomalies requires additionally threshold corrections . according to @xcite , in orbifold models the anomaly associated to the modulus @xmath168 is non - universal only if any of the orbifold twists acts trivially on the corresponding @xmath159 complex plane of the underlying six - torus . this means in particular , that for 6-ii orbifolds the anomalies of @xmath171 and @xmath172 are non - universal and therefore the associated moduli appear in the threshold corrections . further , since the orbifold twist acts non - trivially on the first complex plane , the @xmath170-anomaly must be universal to be completely canceled by the green - schwarz mechanism . we have conducted a scan over @xmath167-anomaly in 6-ii orbifold models , and confirm that only the @xmath170-anomalies are universal ( for our conventions for labeling the two - tori see @xcite ) . however , this does not imply that there are uncanceled @xmath167-anomalies in the other tori . rather , as we shall see in the next section , some @xmath167-anomalies are inherited from what we will call the ` @xmath122-anomaly ' , which can be canceled by the green - schwarz mechanism . discrete anomalies can also be canceled by the green - schwarz mechanism , just like in the @xmath96 case @xcite . under discrete transformation , the dilaton @xmath101 ( more precisely the axion ) gets shifted according to , . note that for the discrete transformation , the shift @xmath203 and @xmath111 are constant ( cf . @xcite ) , while for the anomalous @xmath96 the shift @xmath204 and @xmath205 are @xmath112-dependent . hence , both forms of the khler potential and are invariant under the anomalous discrete transformation . this implies that the term @xmath206 has a definite charge under the discrete transformation . then , stringy non - perturbative effects induce terms of the form @xmath207 , where the @xmath208 transform under ( anomalous ) discrete symmetries . these terms transform trivially ( although they appear to be forbidden by the discrete symmetry ) because the transformation of the fields gets compensated by the dilaton @xcite . note that a superpotential term @xmath207 has to transform trivially both for the anomalous @xmath96 and anomalous discrete symmetries . furthermore , anomaly cancellation by the green - schwarz mechanism requires that discrete anomalies be universal for different gauge group up to modulo the structure . we will examine the universality conditions for discrete anomalies in section [ sec : survey ] . in orbifolds there are certain quantum numbers like @xmath122 ( denoting the twisted sector ) , @xmath209 ( shifted @xmath210 momentum ) , @xmath211 ( shifted 8 momentum ) and oscillator numbers . from these , one can derive other useful quantum numbers such as the discrete @xmath139-charges and modular weights , as defined in equations and . it is hence clear that the derived quantum numbers are related . on the other hand , the discrete @xmath139-charges and modular weights represent discrete charges relevant for the string selection rules . clearly , since the @xmath212 charges derive from the same set of quantum numbers , the different @xmath212 symmetries entailing different string selection rules can not be completely independent . to see what this means , consider the discrete @xmath139-charges and the corresponding selection rule . at first sight , one might think that the @xmath213 , @xmath214 and @xmath215 rules in 6-ii orbifolds entail @xmath216 , @xmath217 and @xmath218 symmetries , respectively . however , it is obvious that , once the @xmath122-rule is satisfied , the discrete @xmath139 symmetries boil down to @xmath219 discrete symmetries . that is , one can factorize this subset of discrete symmetries as @xmath220_r\;.\ ] ] this raises the question whether the @xmath222 symmetry ( which is implied by the selection rule ) has an anomaly . to clarify this , define the @xmath122-anomalies as @xmath223 where the sum extends over all non - trivial representations of @xmath45 . similarly as for the flavor anomalies , the @xmath122-anomaly is only defined modulo twice the smallest non - vanishing dynkin index @xmath224 appearing in the sum in . condition implies that , if @xmath225 is not integer , one has a @xmath222 anomaly . now let us evaluate the @xmath213 anomaly , using the prescription of @xcite . one has @xmath226 where we have used that @xmath227 . the summations @xmath228 and @xmath229 extend , respectively , over untwisted and twisted representations of the gauge factor @xmath45 . this calculation shows that @xmath230 and @xmath225 are related . repeating the calculation for @xmath214 and @xmath215 yields @xmath231 [ eq : ar - ga - ga ] that is , whenever @xmath225 is non - zero , the @xmath232 anomalies can be fractional . in this subsection , we put the @xmath122- and @xmath233 anomalies into a greater perspective . it turns out that they can be related to the so - called ` anomalous 1 ' direction . denote the corresponding generator by @xmath234 . is determined , so that the first equality sign in represents a non - trivial condition which can be used to check the consistency of the model . ] obviously , @xmath234 is a function of the input , i.e. shift and wilson lines , @xmath235 this direction is fixed up to rescaling , our conventions are to normalize @xmath234 such that ( for @xmath236 ) @xmath237 where the sum extends over all states . above equation . in heterotic orbifolds , one can use the scalar product of the @xmath210 lattice , which also appears in . with this scalar product , @xmath238 fulfills @xmath239 . ] together with the other properties @xmath96 , this implies @xmath240 now perform a weyl rotation of the input , @xmath241 with @xmath242 and @xmath243 denoting the weyl group . this is nothing but a change of the basis , hence @xmath244 under . this fixes @xmath234 to be a linear superposition of @xmath99 and the @xmath138 with coefficients that are invariant under weyl transformations . because we are working on the lattice @xmath245 , this relation holds only up to lattice vectors , i.e.@xmath246 where @xmath247 is a lattice vector . this relation between @xmath234 and the orbifold parameters indicates that the presence of an anomalous u(1 ) can be attributed to a geometrical operation in the six dimensional compactified space . this transformation is then encoded in the space group element @xmath248 . we would like to comment that one can not trade @xmath249 for @xmath250 ( and vice versa ) as long as @xmath251 and @xmath252 with @xmath253 denoting the order of the wilson line . that is , the coefficients @xmath249 and @xmath250 are fixed mod @xmath14 and @xmath253 , respectively . further , if @xmath254 , one has @xmath255 , i.e. if @xmath249 or @xmath250 are non - zero , one can infer that @xmath236 , but the converse is in general not true . as we shall see in the next section , it turns out that the coefficients @xmath249 and @xmath256 are related to the @xmath122- and flavor anomalies . we have verified that the decomposition is possible , i.e. that there exist @xmath249 and @xmath250 such that @xmath257\in\lambda_{{\ensuremath{\mathrm{e}_{8}}}\times{\ensuremath{\mathrm{e}_{8}}}}$ ] , for several @xmath0 and @xmath258 orbifolds with and without wilson lines . as we have seen , not all discrete anomalies are independent in orbifold constructions . specifically , we found that the @xmath122- and @xmath139 anomalies are related by . given the decomposition , one is tempted to suspect that discrete anomalies are related to and determined by the coefficients @xmath249 and @xmath250 . to figure out whether this is so , we have conducted a scan over several thousands of models with various geometries and have calculated the @xmath122- , @xmath139- and @xmath167-duality anomalies . we obtain the following ( empirical ) relations : [ [ relation - between - the - boldsymbolk - anomaly - and - boldsymbolkmathrmanom ] ] relation between the @xmath221-anomaly and @xmath259 : + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + @xmath260 in particular , the @xmath225 anomalies are universal . furthermore , the mixed @xmath261 anomaly @xmath262 turns out to be always @xmath263 , thus consistent with the anomaly constraints . [ [ relation - between - boldsymbola_ensuremathmathbbmz_alphamathrmflavor - g - g - and - boldsymboln_alphamathrmanom ] ] relation between @xmath264 and @xmath265 : + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + @xmath266 @xmath267 these anomalies turn out to be universal for different gauge groups in the models under consideration . [ [ relation - between - the - boldsymbolk -- and - boldsymbolr_i - anomalies ] ] relation between the @xmath221- and @xmath268 anomalies : + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + only if there is a @xmath122-anomaly , the @xmath139 anomalies can be fractional . we find that the @xmath232-anomalies are ` inherited ' from the @xmath122-anomaly , specifically [ eq : rkanomaliesmod ] @xmath269 [ [ relation - between - the - boldsymbolk -- and - boldsymbolt - duality - anomalies . ] ] relation between the @xmath221- and @xmath199-duality anomalies . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + similarly to , we have found that the @xmath167-duality anomaly is related to the @xmath122-anomaly by [ eq : tkanomaliesmod ] @xmath270 these statements apply also to the models presented in appendices [ sec : krzanomalies ] and [ sec : bhlranomalies ] . [ [ relation - between - the - boldsymbolk -- boldsymbolt - duality - and - boldsymbolr_i - anomalies . ] ] relation between the @xmath221- , @xmath199-duality and @xmath268 anomalies . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the previous relations and imply [ eq : rtkanomaliesmod ] @xmath271 to summarize , we have conducted a search for discrete anomalies in heterotic orbifolds . as in previous searches @xcite , we find that all basic discrete anomalies are universal in the models we studied , and all anomalies can be canceled by the discrete green - schwarz mechanism . we identify previously unknown relations between the occurrence of discrete anomalies and the so - called ` anomalous 1 ' . the anomalous 1 is in one - to - one correspondence to the ` anomalous space group element ' @xmath131 , whose gauge embedding is the generator of the ` anomalous ' 1 . @xmath167-duality anomalies can be canceled by two ways : the green - schwarz mechanism and @xmath167-dependent threshold corrections as said in section [ sec : tanomalies ] . it is widely believed @xcite that @xmath167-dependent threshold corrections would be non - universal and there would be no certain relation among @xmath167-duality anomalies for @xmath168 , which appear in threshold corrections , e.g. @xmath171 and @xmath172 in 6-ii orbifolds . on the other hand , our ( empirical ) results , which have been checked in several thousands of models with different geometries , show that there exist certain relations among @xmath167-duality anomalies . that is , @xmath167-duality anomalies are related to some basic anomalies that are cancelled only by the green - schwarz mechanism . this issue will be studied in more detail elsewhere . as already mentioned , an ` anomalous ' 1 implies the existence of a fi term , which needs to be canceled in supersymmetric vacua ( as well as in settings with low - energy supersymmetry ) . that means that certain fields which have negative @xmath96 charges need to attain vacuum expectation values ; hence @xmath96 is broken in ( almost ) supersymmetric vacua . in other words , there are no ` anomalous - looking ' unbroken 1 factors . the requirement of keeping the @xmath75-terms of the other symmetries zero leads typically to a situation in which more than one field attains a vev and in which the various vevs are related . achieving @xmath75-flatness translates in the construction of gauge invariant monomials which carry net negative anomalous charge @xcite ( see @xcite for more details ) . one may wonder if one could break @xmath96 by canceling the fi term as usual while leaving the anomalous flavor symmetries intact . we have tried to do this in a large set of models with ` anomalous ' 1 ( including the models presented in @xcite ) , i.e. we searched for gauge invariant monomials with net negative charge under @xmath96 whose constituents transform trivially under the anomalous discrete symmetries . in most models it is hard , if not impossible , to find such a monomial . in other words , according to what we find , the requirement of keeping supersymmetry unbroken forces one not only to break the ` anomalous ' 1 , as is well known , but generically also implies that ` anomalous ' discrete symmetries get broken ( which is somewhat surprising because they do of course not have a @xmath75-term ) . however , in a couple of models we did find a monomial whose constituents transform trivially under some of the anomalous discrete symmetries . in these models , an anomalous @xmath272 subgroup of the original @xmath222 remains unbroken . we posted the details of the model at a web site @xcite . implications will be studied elsewhere . we have studied various aspects of discrete anomalies . we started by reproducing the well - known anomaly constraints for @xmath11 symmetries , taking a different route than usual , namely using the path integral approach . unlike in the conventional approach , our derivation does not rely on contributions from heavy majorana fermions ; only massless fermions enter the computation . we have used the path integral approach to derive anomaly constraints for non - abelian discrete symmetries ; the constraints are given in equation . in the second part of the study , we have explored discrete anomalies in string - derived models , focusing on heterotic orbifolds . we find that discrete anomalies can only occur if there is an ` anomalous ' 1 . one can then rotate the anomalous symmetries into two basic symmetries , corresponding to the rotational and translational part of the space group selection rules , i.e. the @xmath122 rule and @xmath273 rules . all other anomalies , such as @xmath232-anomalies and @xmath167-duality anomalies , derive from these basic anomalies . the coefficients of the basic anomalies are connected to an anomalous space group element , whose gauge embedding arises from the generator of the anomalous 1. we find that the basic anomalies are always universal , such that they might be canceled by the same green - schwarz mechanism that cancels the 1 anomaly . we have also searched for models where the ` anomalous ' 1 symmetry can be broken ( i.e. the fi term can be canceled ) without breaking the ` anomalous ' discrete symmetries . while it is hard to find a model with these properties , we could find a few examples in which an anomalous @xmath272 symmetry survives . the implication of these anomalous @xmath272 symmetries will be discussed elsewhere . of course , discrete and continuous symmetries that are broken by a suppressed vev , as is the case in the ` anomalous ' 1 , are known to be a useful tool in model building . indeed , our results indicate that in string models discrete cousins of ` anomalous ' 1 symmetries are frequently present , whereby , according to what we find , cancellation of the fi term triggers symmetry breakdown . since the fi term is loop suppressed , the vacuum expectation value of the field that breaks the symmetry can be small . the emerging approximate symmetries can play an important role in understanding the observed pattern of fermion masses and mixings . we acknowledge discussions with k. fujikawa , r.n . mohapatra , h.p . nilles and s. raby . we would like to thank the summer institute 2007 ( held at fuji - yoshida ) , where this work was initiated , and the aspen center for physics , where some of the work has been carried out , for hospitality and support . this research was supported by the grand - in - aid for scientific research no . 20540266 and no . 18540257 from the ministry of education , culture , sports , science and technology of japan , the dfg cluster of excellence origin and structure of the universe , the european union 6th framework program mrtn - ct-2004 - 503069 `` quest for unification '' , mrtn - ct-2004 - 005104 `` forcesuniverse '' , mrtn - ct-2006 - 035863 `` universenet '' and sfb - transregios 27 `` neutrinos and beyond '' and 33 `` the dark universe '' by deutsche forschungsgemeinschaft ( dfg ) . in this appendix , we discuss anomalies of the discrete symmetry @xmath1 . the @xmath1 symmetry is one of the simplest non - abelian discrete symmetries . flavor symmetry happens to occur in certain , potentially realistic string models @xcite , which have been constructed recently within the framework of heterotic orbifolds . ] the non - abelian finite group @xmath1 has eight elements , which can be written as products of the two generators @xmath275 and @xmath276 , i.e. @xmath277 @xmath1 has five irreps : @xmath278 , @xmath279 , @xmath280 , @xmath281 and @xmath282 . the action of @xmath275 and @xmath276 on these irreps is @xmath283 according to our discussion in section [ sec : nonabeliandiscretanomalies ] , all we need to do for @xmath1 anomalies is to study the anomalies for the group elements @xmath275 and @xmath276 ( or another combination ) . the @xmath1 flavor symmetry can appear from @xmath284-ii orbifold models @xcite ( and other orbifold models whose compact spaces include the 1d @xmath272 sub - orbifold ) . in @xmath284-ii orbifold models , the group element @xmath275 corresponds to @xmath285 or @xmath286 . there are two fixed points on the 1d @xmath272 sub - orbifold . massless spectra on these two fixed points are degenerate , when there is no wilson line on the 1d @xmath272 sub - orbifold . then , these modes correspond to @xmath278 and the group element @xmath276 corresponds to the permutation of these modes . in @xmath284-ii orbifold models , only the doublet @xmath287 and the trivial singlet @xmath288 can appear as fundamental modes . in this case , anomalies are constrained . we denote @xmath289 now note that @xmath290 for the doublet @xmath278 and @xmath291 . thus , all eight elements of the @xmath1 group can be written as products of @xmath275 and @xmath292 , and the generator @xmath292 does not lead to anomalies . that implies that all of @xmath1 anomalies originate from @xmath285 anomalies , that is , @xmath1 anomalies , e.g. anomalies for the permutation @xmath276 , appear in @xmath284-ii orbifold models only if there are @xmath285 anomalies , i.e. anomalies for the group element @xmath275 . the situation is the same for @xmath1 anomalies in heterotic orbifold models with the 1d @xmath272 sub - orbifold such as @xmath293 . the situation would change if we had heterotic orbifold models including non - trivial singlets of the @xmath1 flavor symmetry , in @xmath280 and @xmath281 , because in these representation the determinants of @xmath275 and @xmath276 differ . indeed , in heterotic orbifold models including the 2d @xmath218 sub - orbifold , non - trivial singlets can appear as fundamental modes @xcite . however , massless states corresponding to @xmath280 and @xmath281 are always degenerate . this can only be changed by introducing a wilson line , which , however , breaks the @xmath1 flavor symmetry . thus , in these models , non - trivial singlets @xmath280 and @xmath281 do not contribute to anomalies . therefore , the situation is the same as @xmath284-ii orbifold models , that is , all of @xmath1 anomalies originate from @xmath285 anomalies . in this appendix we present a sample calculation in order to demonstrate how the anomaly constraints can be applied . we will base the calculations on the grimus - lavoura model @xcite , which is not supersymmetric . the lepton and higgs fields are assigned the transformation properties displayed in table [ tab : grimuslavoura2 ] . l. m. krauss and f. wilczek , phys . * 62 * ( 1989 ) , 1221 . l. e. ibez and g. g. ross , phys . b260 * ( 1991 ) , 291295 . t. banks and m. dine , phys . * d45 * ( 1992 ) , 14241427 , [ hep - th/9109045 ] . p. h. frampton and t. w. kephart , int . phys . * a10 * ( 1995 ) , 46894704 , [ hep - ph/9409330 ] . k. fujikawa , phys . * 42 * ( 1979 ) , 1195 . k. fujikawa , phys . rev . * d21 * ( 1980 ) , 2848 . t. araki , prog . * 117 * ( 2007 ) , 11191138 , [ hep - ph/0612306 ] . m. dine and m. graesser , jhep * 01 * ( 2005 ) , 038 , [ hep - th/0409209 ] . r. a. bertlmann , oxford , uk : clarendon ( 1996 ) 566 p. ( international series of monographs on physics : 91 ) . l. alvarez - gaume and e. witten , nucl . * b234 * ( 1984 ) , 269 . l. alvarez - gaume and p. h. ginsparg , ann . * 161 * ( 1985 ) , 423 . k. fujikawa , s. ojima , and s. yajima , phys . rev . * d34 * ( 1986 ) , 3223 . k. fujikawa and h. suzuki , oxford , uk : clarendon ( 2004 ) 284 p. c. csaki and h. murayama , nucl b515 * ( 1998 ) , 114162 , [ hep - th/9710105 ] . l. e. ibez and g. g. ross , nucl . b368 * ( 1992 ) , 337 . l. e. ibez , nucl . phys . * b398 * ( 1993 ) , 301318 , [ hep - ph/9210211 ] . k. s. babu , i. gogoladze , and k. wang , nucl . * b660 * ( 2003 ) , 322342 , [ hep - ph/0212245 ] . h. k. dreiner , c. luhn , and m. thormeier , phys . rev . * d73 * ( 2006 ) , 075007 , [ hep - ph/0512163 ] . m. b. green and j. h. schwarz , phys b149 * ( 1984 ) , 117122 . t. kobayashi and h. nakano , nucl . phys . * b496 * ( 1997 ) , 103131 , [ hep - th/9612066 ] . t. banks and m. dine , phys . * d53 * ( 1996 ) , 57905798 , [ hep - th/9508071 ] . n. arkani - hamed , m. dine , and s. p. martin , phys . lett . * b431 * ( 1998 ) , 329338 , [ hep - ph/9803432 ] . m. leurer , y. nir , and n. seiberg , nucl . phys . * b420 * ( 1994 ) , 468504 , [ hep - ph/9310320 ] . p. bintruy , s. lavignac , and p. ramond , nucl . phys . * b477 * ( 1996 ) , 353377 , [ hep - ph/9601243 ] . l. j. dixon , j. a. harvey , c. vafa , and e. witten , nucl . phys . * b261 * ( 1985 ) , 678686 . l. j. dixon , j. a. harvey , c. vafa , and e. witten , nucl . b274 * ( 1986 ) , 285314 . s. frste , h. p. nilles , p. k. s. vaudrevange , and a. wingerter , phys . rev . * d70 * ( 2004 ) , 106008 , [ hep - th/0406208 ] . t. kobayashi , s. raby , and r .- j . zhang , nucl . * b704 * ( 2005 ) , 355 , [ hep - ph/0409098 ] . w. buchmller , k. hamaguchi , o. lebedev , and m. ratz , nucl . phys . * b785 * ( 2007 ) , 149209 , [ hep - th/0606187 ] . m. dine , n. seiberg , and e. witten , nucl . b289 * ( 1987 ) , 589 . s. hamidi and c. vafa , nucl . phys . * b279 * ( 1987 ) , 465 . l. j. dixon , d. friedan , e. j. martinec , and s. h. shenker , nucl . phys . * b282 * ( 1987 ) , 1373 . t. kobayashi and n. ohtsubo , int . a9 * ( 1994 ) , 87126 . o. lebedev , h. p. nilles , s. raby , s. ramos - snchez , m. ratz , p. k. s. vaudrevange , and a. wingerter , phys . rev . * d77 * ( 2007 ) , 046013 , [ arxiv:0708.2691 [ hep - th ] ] . t. kobayashi , h. p. nilles , f. plger , s. raby , and m. ratz , nucl . phys . * b768 * ( 2007 ) , 135156 , [ hep - ph/0611020 ] . t. araki , k. s. choi , t. kobayashi , j. kubo and h. ohki , phys . d * 76 * , 066006 ( 2007 ) [ arxiv:0705.3075 [ hep - ph ] ] . l. j. dixon , v. kaplunovsky , and j. louis , nucl . b329 * ( 1990 ) , 2782 . j. louis , talk presented at the 2nd international symposium on particles , strings and cosmology , boston , ma , mar 25 - 30 , 1991 . l. e. ibez and d. lst , nucl . phys . * b382 * ( 1992 ) , 305364 , [ hep - th/9202046 ] . j. p. derendinger , s. ferrara , c. kounnas , and f. zwirner , nucl b372 * ( 1992 ) , 145188 . f. buccella , j. p. derendinger , s. ferrara , and c. a. savoy , phys . lett . * b115 * ( 1982 ) , 375 . a. font , l. e. ibez , h. p. nilles , and f. quevedo , nucl . b307 * ( 1988 ) , 109 , erratum _ ibid . _ * b310*. g. cleaver , m. cvetic , j. r. espinosa , l. l. everett , and p. langacker , nucl . phys . * b525 * ( 1998 ) , 326 , [ hep - th/9711178 ] . o. lebedev , h. p. nilles , s. raby , s. ramos - snchez , m. ratz , p. k. s. vaudrevange , and a. wingerter , phys . b645 * ( 2007 ) , 88 , [ hep - th/0611095 ] . t. araki , t. kobayashi , j. kubo , s. ramos - snchez , m. ratz , and p. k. vaudrevange , _ additional material for discrete anomalies _ , 2008 , ` www.th.physik.uni-bonn.de/nilles/z6iiorbifold/anomalies/ ` .

we consider on - center and off - center observers in an inhomogeneous , spherically symmetric , isocurvature ( flat ) concentration of dark energy with typical size of a few gpc . such a concentration could be produced e.g. by a recently formed global monopole with core size that approaches the hubble scale . in this case we would have what may be called ` topological quintessence ' in analogy with the well - known _ topological inflation_. we show that the minimum comoving radius of such a dark energy inhomogeneity that is consistent with the union2 type ia supernovae ( snia ) data at the level is . as expected , the best - fit fractional dark energy density at the center , , approaches the corresponding value for large enough values of the inhomogeneity radius ( ) . using the union2 data , we show that the maximum allowed shift of the observer from the center of the inhomogeneity is about which respects the copernican principle . the model naturally predicts the existence of a preferred axis and alignment of the low cmb multipoles . however , the constraints on coming from the magnitude of the cmb dipole remain a severe challenge to the copernican principle and lead to even for an inhomogeneity radius as large as .

1103.4143

detailed cosmological observations made during the last decade ( see @xcite for some of the latest results ) have made it clear that the minimal cosmological model ( scdm ) based on the cosmological principle ( homogeneity , isotropy , validity of general relativity ) and with a purely radiation and matter content ( baryonic and cold dark matter ( cdm ) ) , is inconsistent with observations . the simplest extension of scdm able to reconcile it with the observed accelerated cosmic expansion ( ) involves the existence of a homogeneous , constant in time energy density that constitutes about @xmath11 of the total energy density of the universe and is known as the _ cosmological constant _ @xcite . a time - dependent generalization of this energy has been dubbed _ dark energy_. the repulsive gravitational properties of the cosmological constant dominate on large scales and may induce the observed acceleration . due to its simplicity and its consistency with most cosmological observations is currently the _ standard _ cosmological model , replacing scdm , which held that status in the early 90 s , when observational data were far less extensive and accurate . however , is also faced with some challenges , which originate from both theoretical and observational considerations . the two main theoretical challenges of the model are _ fine tuning _ and _ lack of theoretical motivation_. indeed , in order to make the model consistent with observations , the energy scale of the cosmological constant must be fine - tuned to an unnaturally small value , @xmath12 orders of magnitude smaller than the one anticipated from theoretical models of quantum field theory @xcite . from the observational point of view , even though is in agreement with the vast majority of data , there is a set of observations that appear to be mildly inconsistent with the model at the @xmath13 level . most of these observations , which we proceed to summarize , seem to be related to the existence of a preferred cosmological axis : * * planarity and alignment of the cmb low multipole moments : * it has been known since 2003 @xcite ( see @xcite for recent reviews ) that the quadrupole and octopole components of the cmb temperature perturbation maps ( the largest scale part of those maps ) are dominated by planar features which are unnaturally aligned with each other in the direction ( perpendicular to the planes ) @xmath14 @xcite . the probability that such combination of planarity and alignment occurs in a random gaussian map is less than @xmath15 @xcite . this peculiar feature has been attributed to a posteriori considerations @xcite , namely _ within a large number of features of a map , it is statistically anticipated that some of them will present large statistical fluctuations_. this type of argument would be stronger if the large fluctuations affected random uncorrelated features ( for example multipoles with @xmath16 and @xmath17 ) . in that case we would have an a posteriori arbitrary selection of features . the quadrupole - octopole moments , however , constitute _ all _ the information we have about the cmb temperature fluctuation maps on the largest angular scales . thus the observed alignment could also be restated by saying that _ the large scale features of the cmb maps are significantly more planar than anticipated for random gaussian maps_. the fact that the _ preferred _ plane is relatively close to the galactic plane ( which is usually masked due to noise , before the construction of the power spectrum is made ) is probably responsible @xcite for the relatively low values of the angular power spectrum components @xmath18 , @xmath19 , which is considered by some authors to be an independent puzzle @xcite . * * large scale velocity bulk flows : * recent studies @xcite have detected a large scale flow with a bulk velocity in the range @xmath20 $ ] km / s towards @xmath21 on scales @xmath22 . this detection is larger than the prediction and inconsistent with it at the level of @xmath23 . * * large scale alignment in the qso optical polarization data : * quasar polarization vectors are not randomly oriented over the sky with a probability often in excess of 99.9% . the alignment effect seems to be prominent along a particular axis with direction @xmath24 @xcite . * * profiles of cluster haloes : * predicts shallow , low concentration and density profiles , in contrast to some observations which point to denser cluster haloes @xcite . as announced , three of the above four puzzles are large - scale effects related to preferred cosmological directions ( cmb multipole alignments , qso polarization alignment and large scale bulk flows ) ; those directions , moreover , appear to be not far from each other , being approximately normal to the axis of the ecliptic poles @xmath25 and lying close to the ecliptic plane and the equinoxes . this coincidence has triggered investigations for possible systematic effects related to the cmb preferred axis but no significant such effects have been found @xcite . finally , it should be mentioned that , in addition to the discussed large - scale effects , the model also faces some issues on galactic scales ( missing satellites problem @xcite and the cusp / core nature of the central density profiles of dwarf galaxies @xcite ) . in spite of those problems , the construction of generalizations of has been motivated so far mainly by theoretical rather than observational considerations . some of the most popular ideas among those that go beyond are the following : * * time dependent dark energy @xcite : * within this approach , dark energy retains its homogeneous and isotropic character , but may present a time dependence . this time dependence aims at solving ( or at least alleviating ) the previously mentioned fine - tuning problem , i.e. the unnaturally small value of the current dark energy required by the observations . indeed , it turns out that a dark energy density which is larger at early times and evolves via proper dynamics to its present low value is easier to accommodate in theoretical models . a representative of this class of models is _ quintessence _ , where the role of dark energy is played by a dynamically evolving scalar field . however , for this evolution to be consistent with the observations , the scalar field must be assigned an extremely ( fine - tuned ) small mass , fact that makes the additional complexity introduced in this class of models be of questionable value . * * modified gravity : * the repulsive gravity provided by the cosmological constant on large scales can also be induced by modifications of general relativity in the context of a homogeneous and isotropic background . several such modifications have been suggested , including scalar tensor gravity @xcite , dgp models that involve gravitons leaking through large extra dimensions @xcite , @xmath26 gravity @xcite etc . these models are well motivated as complete geometric physical theories , but they have more degrees of freedom than and they also require significant fine - tuning in order to be consistent with both the observed cosmic accelerating expansion and the gravitational solar system constraints that prevent large local departures from general relativity . * * inhomogeneous and isotropic lemaitre - tolman - bondi void models @xcite : * within this framework , an additional dark energy component is not needed to secure consistency with the cosmological data that indicate accelerating expansion . the basic idea that lies behind them is to consider that the increased expansion rate occurs locally in space rather than at recent cosmological times , a fact that can be achieved by assuming a locally - reduced matter energy density @xcite . thus , the observer is placed close to the center of a giant void with dimensions of a few gpc @xcite . even though this approach is free of dark energy , it is by no means free of fine tuning . apart from the unnatural assumption of giant - size gpc voids , which are very unlikely to be produced in any cosmology , these models require the observer to be placed within a very small volume at the center of the void ( about @xmath27 of the total volume of the void ) . a slightly off - center observer , however , will naturally experience a preferred cosmological direction ( towards the center of the void ) , which may help to resolve some of the observational puzzles of discussed above . in the present study we consider yet an alternative approach for the generalization of . instead of breaking the time translation invariance of dark energy , as in the class of quintessence models , we break its space translation invariance ( homogeneity ) . we thus consider a spherically symmetric overdensity of dark energy with typical scale of a few gpc : _ inhomogeneous dark energy ( ide)_. this class of models adds degrees of freedom to in a manner that is generically different from the other three classes considered above . thus , it also has specific advantages as compared to them . in particular : * * physical mechanism : * there is a well - defined _ physical mechanism _ that can produce this type of spherically symmetric , hubble - scale ide . it is based on applying the principles of _ topological inflation _ @xcite to the case of late - time acceleration . according to the idea of topological inflation , the false vacuum energy of the core of a topological defect can give rise to accelerating expansion if the core size reaches the hubble scale when gravity starts dominating the dynamics . thus , for example , a recently formed global monopole with appropriate scale of symmetry breaking and coupling could naturally produce a hubble - scale , spherically symmetric , isocurvature dark energy overdensity . by analogy with topological inflation , this mechanism may be called _ topological quintessence_. * * naturally large scale : * whereas matter inhomogeneities ( voids ) of a few gpc are very unlikely to be formed through large scale structure mechanisms , ide on hubble scales is more natural and easy to construct . in the ide case ( also in contrast to inhomogeneous matter models ) , the homogeneous limit ( ) is consistent with observations and one interesting question is _ ` what is the smallest scale of inhomogeneity for which this consistency remains?'_. * * more natural off - center observers ( less fine tuning ) : * due to the arbitrarily large scale of ide ( supported by observations , as we will see ) , there is less fine tuning for the location of the observer with respect to the center of the spherical inhomogeneity . in fact , as the inhomogeneity scale approaches the horizon the model becomes indistinguishable from and there is no observational constraint for the location of the observer . * * natural generation of a preferred cosmological axis : * as we have seen , most of the observational challenges of hint towards the existence of a preferred cosmological direction . such a direction is naturally provided for an off - center observer in an inhomogeneous matter or dark energy model : _ the direction that connects the location of the observer with the center of the inhomogeneity_. the displacement of the observer naturally leads to an alignment of low cmb multipole moments and bulk velocity flows . the additional degrees of freedom ( with respect to ) introduced in this class of models are similar to the corresponding ones in models that break the time translation invariance of dark energy . we also need a scalar field , but now with spatial rather than temporal dependence . compared to matter void models , however , the ide setup requires more ( and less natural ) degrees of freedom . this is the price to pay for the reduction in the amount of fine tuning . the goal of the present study is to investigate some of the observational consequences and constraints of ide models . we focus on the constraints coming from snia data and the cmb low multipole moments . we also discuss the physical mechanism that could give rise to ide ( _ topological quintessence _ ) , although we postpone a complete analysis of this mechanism for a future study @xcite . the structure of this paper is the following : in the next section we derive the cosmological equations , light - like geodesics and luminosity distance for on and off - center observers in the presence of ide . we also discuss the basics of topological quintessence as a mechanism for producing inhomogeneous dark energy and derive the angular dependence of the cmb temperature fluctuations induced by the shift of the observer from the center of the inhomogeneity . in section 3 we use the predicted luminosity distance to constrain the model parameters using the union2 snia data @xcite . in the same section we compare the values of the low multipole moments ( dipole , quadrupole ) of cmb perturbations induced by the displacement of the observer from the center to the corresponding observed values and derive constraints on the parameters of the model . finally , in section 4 we conclude and discuss the implications and possible extensions of our results . in what follows we fix the present age of the universe to @xmath28gyr . in the best fit model , this would correspond to fixing @xmath29km/(smpc ) . we also refer to a comoving radial distance simply as ` distance ' . the line element of an expanding , spherically symmetric spacetime is given by the lemaitre - tolman - bondi ( ltb ) metric @xcite : @xmath30 we consider a spherically symmetric energy - momentum tensor : t_^= ( -(r)-_m(r , t),p_r(r),p_t(r),p_t(r ) ) [ enmom ] where @xmath31 is the matter density and we have allowed for a general inhomogeneous and spherically symmetric static fluid with : ( t_f)_^= ( -(r),p_r(r),p_t(r),p_t(r ) ) [ enmommon ] ( being @xmath32 the transverse pressure ) . this expression is motivated by the energy - momentum of a global monopole , which asymptotically is of the form @xcite : ( t_mon)_^= ( -/r^2,-/r^2,0,0 ) [ tmnmonas ] with @xmath33 the symmetry - breaking scale related to the formation of the monopole . the einstein equations @xmath34 may be obtained from equations ( [ ltbmet ] ) , ( [ enmom ] ) and read : @xmath35 @xmath36 @xmath37 and @xmath38 the primes and dots denote derivatives with respect to @xmath39 and @xmath40 , respectively . ( [ yht10 ] ) leads to : @xmath41 where @xmath42 is a constant of integration depending only on the coordinate @xmath39 and @xmath43 is a function associated with the spatial curvature . thus , the ltb metric ( [ ltbmet ] ) takes the form : @xmath44 taking ( [ ltbmet2 ] ) to ( [ yht11 ] ) we obtain : [ freq1 ] ^2 + 2 a + k(r ) = - 8 g p_r(r ) a^2 , which implies accelerating expansion for negative pressure and proper curvature profile . it is straightforward to check that this equation has the following first integral : @xmath45 this last equation may also be written as @xmath46\ ] ] provided we define : @xmath47 @xmath48 where @xmath49 ( @xmath50 is the present time ) , @xmath51 , @xmath52 ( which is positive for negative radial pressure ) and @xmath53 . note that the profile of the inhomogeneous dark energy pressure , @xmath54 , is in principle arbitrary , being determined by the physical mechanism that induced the inhomogeneous overdensity . the conservation of the energy - momentum tensor ( [ enmommon ] ) leads to the equations : [ conenmomr ] + ( + p_r ) + ( + p_t)&= & 0 , + p_r + ( p_r - p_t)&= & 0.[conenmomp ] in the special case of isotropic pressure ( @xmath55 ) and @xmath56 we obtain the homogeneous dark energy model which has been well studied . in what follows we will assume the presence of anisotropic pressure . a physical model that can give rise to an energy - momentum tensor of the form ( [ enmommon ] ) is a global monopole with a hubble - scale core embedded in an expanding spacetime . such a system is described by the action : @xmath57\,,\ ] ] where the higgs field has three components @xmath58 , and the potential is @xmath59 with @xmath33 the vacuum expectation value and @xmath60 a coupling constant . in this model , the global monopole field configuration @xmath61 is topologically stable and can not decay to the true vacuum @xmath62 in a continuous manner @xcite . these field configurations are expected to form during cosmological phase transitions when the vacuum manifold of a field theory has the topology of the sphere @xmath63 via the kibble mechanism @xcite . they are predicted to form initial isocurvature perturbations @xcite . the monopole solution has localized energy in a core where @xmath64 with thickness @xmath65 and @xmath66 reaches asymptotically the vacuum expectation value @xmath33 as a power law . the asymptotic form of the energy - momentum tensor is given by eq . ( [ tmnmonas ] ) . a recently formed global monopole would naturally have a core size comparable to the present hubble scale ^-1/2 ^-1h_0 ^ -1 ( 10 ^ -33 ev)^-1 [ monscale ] for @xmath67 we obtain @xmath68 , but this energy scale can increase significantly if we allow for small enough values of the coupling constant @xmath60 . such a field configuration could also be considered as a mechanism leading to spatial variation of fundamental constants ( like the fine structure constant ) through coupling with the spatially varying scalar field of the monopole . in fact there have been recent claims about detection @xcite of spatial variation of the fine structure constant along a particular direction in the sky . by varying the action ( [ gmaction ] ) with respect to the metric and with respect to the scalar field , we obtain the following dynamical equations : g__-12g_r=8m_pl^ 2 t _ , @xmath69 where @xmath70,\ ] ] is the energy - momentum tensor of the global monopole . for the metric ( [ ltbmet2 ] ) , its components are t_0 ^ 0= & -(r)= -(^22+^22x(r , t)^2+^2a(r , t)^2+v ( ) ) + t_0r= & + t_r^r= & p_r(r)=(^22+^22x(r , t)^2-^2a(r , t)^2-v ( ) ) + t_^= & p_t(r)=t_^=(^22-^22x(r , t)^2-v ( ) ) for a static global monopole they asymptotically reduce to the form ( [ tmnmonas ] ) and negative pressure is naturally obtained in the core where the potential energy dominates . such a static solution has been shown to exist in flat spacetimes @xcite and also in expanding spacetimes when the core size is smaller than the hubble scale ( fig.2a of ref . monopoles with cores larger than a critical fraction ( of @xmath71 ) of the hubble scale have been shown @xcite to become non - static , as their core starts expanding along with the background hubble expansion . such an expansion may also be induced by the false vacuum energy of the monopole core leading to _ topological inflation_. even though there has been no study of the embedded monopole in an expanding background where matter is also present it is conceivable that a similar behavior will also be present in that case i.e. a static solution will be supported for core sizes up to about the hubble scale and for a larger size , the core will start expanding along with the background especially when the monopole energy density of the core starts to dominate over the matter energy density . the behavior of such solutions in the presence of matter is an issue under investigation @xcite . between the tangent vector to the geodesic and the unit vector along the axis connecting the center with the observer . [ xifig ] , scaledwidth=48.0% ] in what follows we focus on the cosmological equation ( [ freq3 ] ) and use it to derive observational constraints for observers that lie at the center of the inhomogeneity and also for off - center observers . to achieve this goal we follow the following steps : 1 . we assume a dark energy profile @xmath54 in eq . ( [ freq3 ] ) , characterized by two main parameters : the scale @xmath5 of the inhomogeneity and the magnitude of @xmath3 at the center ( @xmath72 ) . 2 . we solve eq . ( [ freq3 ] ) assuming flatness ( @xmath73 ) with initial conditions taken at the present time @xmath50 as @xmath74 and evolving backwards in time @xcite . this initial condition is a gauge choice that corresponds to the usual setting @xmath75 made in homogeneous models . our assumption for flatness is based on the fact that the initial perturbations produced by topological defects are isocurvature while the universe is assumed to be flat before the formation of the hubble - scale global monopole . 3 . using the derived solution we solve the geodesic equations and find the light - like geodesics @xmath76 ) , @xmath77 , @xmath78 for on - center and off - center observers . the off - center observer geodesics depend on two additional parameters : the distance @xmath79 of the observer from the center and the angle @xmath80 between the tangent vector to the geodesic and the unit vector along the axis connecting the center with the observer @xcite . the value of the angle @xmath80 for each direction in the sky depends on the assumed position ( galactic coordinates @xmath81 ) of the center of the inhomogeneity ( i.e. @xmath82 , see fig . [ xifig ] ) . + _ supernovae _ 4 . using the derived geodesics we obtain the luminosity distance and other cosmological observable distances . these distances depend on the model parameters namely @xmath5 and @xmath83 for the central observer and in addition to @xmath79 and @xmath81 ( the galactic coordinates of the center ) for the off - center observer . we fit the derived forms of the luminosity distance to the measured luminosity distances of the union2 snia data and derive the best - fit values of the parameters @xmath5 , @xmath83 , @xmath79 and @xmath84 and their @xmath85 regions . + _ cosmic microwave background _ 6 . for the derivation of the additional cmb temperature anisotropy induced by the shift of the observer from the center , we use the geodesic equations to derive the dependence of the redshift of the last scattering surface on the angle @xmath80 . having found @xmath86 we find the angular dependence of the temperature as t(,r_obs)= [ tempprof ] where @xmath87 is the observed temperature of the last scattering surface . this is assumed to be uniform since we are focusing on the _ additional _ temperature fluctuations induced by the displacement of the observer . we then convert it to temperature fluctuations from the mean and expand in spherical harmonics to find the multipole coefficients . comparison with the measured values of the moments ( particularly of the dipole ) leads to constraints on @xmath79 . we will consider the following form for the profile of the inhomogeneity [ omegam ] _ m(r)&=&_m , + ( _ m , -_m , ) + [ omegax ] _ x(r)&=&_x , + ( _ x , -_x , ) , keeping in mind the anticipated physical model of a global monopole , we set @xmath89 , @xmath90 and since we are considering isocurvature inhomogeneities in a flat universe ( @xmath91 ) , we keep @xmath92 . in our formulae we keep @xmath93 and @xmath94 , even though these quantities turn out to vanish for the particular profiles we have considered . we also fix @xmath95 , corresponding to a fairly sharp transition from ` in ' to ` out ' . we are thus left with two parameters characterizing the profile : the radial size @xmath5 and the density at the center @xmath3 . in order to solve the dynamical equation ( [ freq3 ] ) we first determine @xmath96 in units of @xmath97 , assuming a simultaneous big - bang and , as previously mentioned , an age of the universe @xmath28gyr . thus we use : @xmath98 which can be easily derived from eq . ( [ freq3 ] ) . we then use this last equation and the initial condition @xmath74 to solve eq . ( [ freq3 ] ) and obtain numerically the form of @xmath88 and its derivatives . in the case of the on - center observer , spherical symmetry with respect to the observer is retained and therefore light travels radially towards the observer . this implies that @xmath99 along the geodesic . the remaining two light - like geodesics for radially _ incoming _ light rays are of the form @xcite @xmath100 @xmath101 where @xmath102 is the velocity on light in units of gpc / gyr used for consistency in order to obtain the luminosity distance in gpc . equations ( [ dtdz ] ) and ( [ drdz ] ) with initial conditions @xmath103 and @xmath104 determine @xmath105 and @xmath106 . the angular diameter and luminosity distance are then simply given by : @xmath107 the corresponding distance modulus , which is the difference between the apparent magnitude @xmath108 and the absolute magnitude @xmath109 , is given by @xmath110 since @xmath111 is obtained in @xmath112 . it is now straightforward to fit the predicted distance moduli to the observed ones from the union2 dataset and derive the best - fit values of parameters for the on - center observer . the light - like geodesics that go through the displaced off - center observer @xmath113 ( @xmath60 is the affine parameter ) , are derived by solving the geodesic equations in a coordinate system with origin at the center of the inhomogeneity @xcite ( see fig . [ xifig ] ) . taking into account the axial symmetry around the observer , it is clear that we will have @xmath114 along the geodesic . moreover , due to the the 4-velocity identity @xmath115 ( valid for light - like geodesics ) , the time geodesic reduces to a first order ordinary differential equation . the geodesic equation for @xmath78 also reduces to first order thanks to the spherical symmetry of the inhomogeneity ( conservation of angular momentum ) . finally , the geodesic for @xmath77 is second order , but it can be split in two first order equations by means of the convenient definition @xmath116 . the resulting system consists of the following five first order equations @xcite : @xmath117,\\ \frac{d p}{d\lambda } \;=\ ; & 2\dot{a}'\ , p\,\sqrt{\frac{p^{2}}{1-k(r)}\ , + \frac{j^{2}}{a^{2}\ , a'^{2}}}\\ & + \,c\,\frac{1-k(r)}{a^{3}a'}j^{2}- c\ , \left[\frac{k(r)'}{2 - 2k(r ) } + \frac{a''}{a'}\right ] p^{2}\ , , \end{aligned}\ ] ] where @xmath118 is the conserved angular momentum ( whose conservation emerges from the geodesic of @xmath78 ) . the initial conditions are specified at the time @xmath50 when the photon arrives at the observer s position , which is given by @xmath119 and @xmath120 since without loss of generality we may assume that the observer is displaced along the @xmath121 axis . the photon hits the observer at an angle @xmath80 relative to the @xmath121-axis . according to the metric ( [ ltbmet2 ] ) , the spatial components of the unit vector along this axis are v^i = ( 1,0,0 ) where the three components are in the @xmath39 , @xmath122 and @xmath123 direction , respectively . the unit tangent vector @xmath124 to the photon path at the present time @xmath50 is given by ^i = || ( , , ) = -(p , j / a^2,0 ) [ barui ] where @xmath125 and the first factor ensures normalization , @xmath126 ( notice that @xmath127 in view of the light - like geodesic condition @xmath128 ) . the minus sign in the above equation emerges because we can choose the affine parameter @xmath60 such that @xmath129 when @xmath130 and @xmath131 since the time decreases from @xmath50 as @xmath60 increases . the angle @xmath80 is then given by : [ cosxi ] = g_ij|u^i v^j = - . notice that at a time @xmath132 the angle @xmath80 does not represent the angle between the geodesic tangent and the @xmath121 axis but the angle between the geodesic tangent and the radial unit vector ( which coincides with the @xmath121 axis unit vector only at @xmath130 when @xmath133 ) . in what follows we refer to @xmath80 as the value obtained from eq . ( [ cosxi ] ) at the present time @xmath50 . we can now express the initial condition for @xmath134 in terms of the incident photon angle @xmath80 as p_0== [ p0 ] also using the fact that @xmath135 evaluated at @xmath130 is a unit vector , we find using the metric ( [ ltbmet2 ] ) and eq . ( [ barui ] ) j = a(r_obs , t_0)= r_obs [ j0val ] thus , once the incident angle @xmath80 and observer shift @xmath79 are specified , the initial conditions to the system ( [ offctrsyst ] ) are fixed as follows @xmath136 thus , the numerical solution of the system ( [ offctrsyst ] ) with initial conditions ( [ offctrinit ] ) is straightforward and leads to the geodesics @xmath137 , @xmath138 , @xmath139 and @xmath140 . by inverting the expression @xmath141 the affine parameter @xmath60 may be replaced by the redshift in @xmath40 , @xmath39 and @xmath122 . the luminosity distance is now obtained from the geodesics for the case of an off - center observer as @xcite @xmath142\ , \ ] ] where @xmath143 and the partial derivatives with respect to @xmath80 are obtained numerically by evaluating the geodesics at slightly different values of @xmath80 . it is easy to see that for the on - center observer we have @xmath144 and @xmath145 ( since @xmath146 ) and therefore we re - obtain the expression of the luminosity distance for the on center observer ( [ lumidi ] ) . in order to compare the predicted luminosity distance with the corresponding observed one through snia data , it is important to derive the dependence of the angle @xmath80 on the coordinates of each snia . let us assume an equatorial coordinate system , where coordinates are given by the right ascension @xmath147 and declination @xmath148 , since the directions of the snia of the union2 dataset are provided in this coordinate system @xcite . being the angle between the center - observer axis and the observer - snia direction , @xmath80 depends on both the coordinates of the inhomogeneity center @xmath149 and on the coordinates of each snia @xmath150 . it may be shown @xcite that the function @xmath151 is given by @xmath152,\ ] ] thus , equation ( [ xiacdccon ] ) translates the parameter @xmath80 to the coordinates of the center of the inhomogeneity @xmath149 and introduces the snia equatorial coordinates @xmath150 . this allows the fit to the snia union2 data and the derivation of the best fit parameter values @xmath153 . we assume isotropic temperature on the last scattering surface and focus only on those additional temperature anisotropies that emerge due to the shift of the observer from the center of the dark energy inhomogeneity . thus the cmb temperature measured by the off - center observer is given by eq . ( [ tempprof ] ) . in order to determine @xmath86 we proceed as follows : * we find the light - like geodesics for an on - center observer with fixed @xmath5 and @xmath3 and determine the time at decoupling @xmath154 where @xmath155 is determined by solving the equation @xmath156 . * for a given off - center observer shifted by @xmath79 and a given direction @xmath80 we find the light - like geodesics and determine @xmath157 by solving the equation @xmath158 where @xmath159 is the decoupling time as determined for the on - center observer . * we determine @xmath160 where @xmath161 is the redshift along the geodesic as determined in the previous subsection . in order to calculate the predicted temperature fluctuations we must also obtain the predicted mean value @xmath162 of the cmb temperature @xmath163 . notice that this value is not exactly equal to the temperature @xmath87 at the last scattering surface . the mean temperature @xmath162 is evaluated using eq . ( [ tempprof ] ) as @xcite [ meantemp ] |t(r_obs)d t(,r_obs ) = _ 0^d + the predicted angular dependence of the temperature fluctuations may now be easily evaluated using eq . ( [ meantemp ] ) as : [ tempfluct ] ( , r_obs ) & & + & = & where @xmath164 is defined as [ barz ] obtained by expanding the temperature fluctuations into spherical harmonics as [ dtexpy ] ( , r_obs ) = _ l , ma_lmy_lm ( , ) where the values of the multipole moments @xmath165 are obtained as [ cmbalm ] a_lm= _ 0 ^ 2_0^(,r_obs)y_lm^ * ( , ) d ddue to axial symmetry of the temperature perturbations , all the @xmath165 will vanish , except those with @xmath166 . the observed cmb dipole @xmath167 is of the order @xmath168 . by demanding that the predicted value of @xmath167 does not exceed @xmath168 we may obtain an upper bound on @xmath79 . there is a newtonian analytical approximation that can be made in order to obtain an expression for @xmath169 valid for small observer displacement @xmath79 . the cosmological setup corresponding to a displaced observer from the center of a spherical inhomogeneity may be replaced @xcite by that of an observer with a specific peculiar velocity @xmath170 . this peculiar velocity is due to the increased value of the hubble parameter @xmath171 inside the inhomogeneity compared to the corresponding value that would exist if the inhomogeneity was not present ( @xmath172 ) . it is given by q= = r_obs [ qdef ] for example for the profiles of eqs . ( [ omegam ] ) , ( [ omegax ] ) with @xmath173 , @xmath174 we have q0.07 [ qval ] this peculiar velocity leads to the observation of doppler shifted photon frequencies and to an anisotropic temperature profile @xmath175 . the ratio of this anisotropic temperature over the isotropic temperature @xmath176 observed by the central observer is @xcite = = [ temprat ] where the first equality is obtained by a doppler shift analysis and @xmath177 is the redshift of the last scattering surface as detected by the central observer . for small @xmath178 we find z_ls(,r_obs)z_ls , c ( 1-q ) [ zlsanal ] thus , using eq . ( [ zlsanal ] ) in eq . ( [ tempprof ] ) we have an approximate analytic expression for the predicted temperature profile which may lead to the evaluation of the predicted low multipole moments . we have checked that our numerical results presented in the next section are in rough agreement with the predictions of this newtonian analytical model . this model correctly predicts the qualitative angular dependence of @xmath179 on @xmath80 ( @xmath180 ) but we found it to be off by a factor @xmath181 on the amplitude @xmath178 for the models considered . in this section , we implement the methods described in the previous section to impose observational constraints on the parameters of the model . in order to do so we will use the union2 snia data and also cobe / wmap results for the observed cmb low multipole moments . , @xmath182 and @xmath1 contours in the parameter space @xmath183 for an on - center observer , obtained from eq . ( [ chi2 ] ) . the parameter values @xmath184 provide a good fit to the data and will be often assumed when dealing with the off - center case . the corresponding @xmath185 contours for the usual ltb model ( only matter+curvature , @xmath186 ) are also shown . note that the relation @xmath92 is obviously only valid for the ide model.[ontrfig],scaledwidth=48.0% ] the union2 snia data points are given , after the corrections have been implemented , in terms of the distance modulus _ ( z_i)m_(z_i ) - m [ mug].the parameters of the theoretical model are determined by minimizing the quantity : ^2_1 ( r_0,_x , , _ 0)= _ i=1^n [ chi2 ] , where @xmath187 and @xmath188 are the errors due to flux uncertainties , intrinsic dispersion of snia absolute magnitudes and peculiar velocity dispersion . these errors are assumed to be gaussian and uncorrelated . the theoretical distance modulus @xmath189 is given by eq . ( [ distmodonc ] ) and we have allowed for an offset @xmath190 which is dealt with either by marginalization or by minimization ( the choice does not affect the fit of the other parameters ) @xcite . in fig . [ ontrfig ] we show the @xmath85 , @xmath182 and @xmath1 contours in the parameter space @xmath183 ( where @xmath92 ) obtained from eq . ( [ chi2 ] ) after minimization with respect to the nuisance parameter @xmath190 . as could have been anticipated , there is no upper observational limit on the size of the inhomogeneity , since the homogeneous limit @xmath191 corresponds to the model , which is indeed consistent with the data . what is more interesting is the fact that the range @xmath192 is excluded by the union2 snia data at the @xmath1 level . the range of allowed values of @xmath3 is @xmath193 at the @xmath1 level . in fig . [ ontrfig ] we also show the corresponding @xmath185 contours for the usual ltb model ( only matter+curvature , @xmath186 ) . notice the reduced size of the allowed region in the parameter space and also the reduced quality of the fit ( @xmath194 ) as compared to the ide and @xmath195cdm models ( @xmath196 ) . in the remainder of the section we move to investigate the snia and cmb results for an off - center observer . since in that case there are three additional parameters , @xmath197 , we will consider a discrete set of values for @xmath198 , as shown in fig.[ontrfig ] , with special attention to the values @xmath199 . these values provide a good fit to the union2 data for the on - center observer , while being distinct from the @xmath195cdm limit . + in the case of the off - center observer , the theoretical distance moduli depend not only on @xmath200 , but also on @xmath201 ( see eqs . ( [ distmodonc ] ) , ( [ lumdistoff ] ) , ( [ xiacdccon ] ) ) . in order to show the effect of the additional parameters on the luminosity distance and the corresponding distance modulus , we show in fig . [ distmod ] the distance modulus as obtained from eqs . ( [ distmodonc ] ) , ( [ lumdistoff ] ) in the direction of the center ( @xmath202 ) and in the opposite direction ( @xmath203 ) , for various values of the observer shift @xmath79 and fixed @xmath184 . we also superpose the union2 data . clearly , values of @xmath204 in the direction away from the center do not provide a good fit . however , in the direction _ towards the center _ , the fit is good even for large @xmath79 approaching the size of the inhomogeneity . + [ fig9 ] in order to find the best - fit values for the parameters of the model , and assuming that @xmath198 are fixed , we need to minimize the following @xmath185 function : ^2_2 ( r_obs,_c,_c,_0)= _ i=1^n [ chi2off ] , where @xmath205 is obtained by combining eqs . ( [ distmodonc ] ) and ( [ lumdistoff ] ) . the contour plot of @xmath206 obtained for @xmath207 after conversion to galactic coordinates and minimization with respect to @xmath79 and @xmath190 is shown in fig . the best - fit coordinate values for the location of the center of the dark energy spherical inhomogeneity are @xmath208 where the errors correspond to a rough estimate of the @xmath85 region . the corresponding @xmath85 range for the observer shift is @xmath209 . in table [ tab1 ] and in fig . [ fig4 ] we also show the best - fit directions towards the inhomogeneity center obtained for the other values @xmath210 shown in fig . [ ontrfig ] . even though these directions are in a similar region in the sky , the uncertainty is clearly very large . despite of the existence of three additional parameters @xmath153 , the improvement of the fit compared to on - center case is marginal ( for @xmath211 we obtain @xmath212 at the best - fit point @xmath213 ) . the improvement with respect to the model is even smaller : we obtain @xmath214 for the same values of @xmath210 . we conclude that there is no statistically significant advantage of this model over . however , the direction of the mildly preferred axis is not far from the direction of the cmb dipole @xmath215 where we have allowed for identification of opposite directions . .best - fit direction @xmath216 and observer displacement @xmath79 from the center of the inhomogeneity , obtained for the various values of @xmath210 shown in fig . [ ontrfig ] . [ cols="^,^,^,^,^,^",options="header " , ] in fig . [ fig5 ] we show a contour plot of the best - fit value of @xmath79 as a function of the direction in the sky ( given by the location of the inhomogeneity center ) , in galactic coordinates . the largest best - fit value is @xmath217 corresponding to the direction @xmath218 . the maximum value of the observer shift @xmath79 consistent with the union2 data may be obtained by plotting @xmath219 after minimization with respect to @xmath220 . this plot is shown in fig . [ robsmax ] for the set of values @xmath210 shown in fig.[ontrfig ] . it shows that even values of @xmath79 as large as @xmath221 provide an acceptable fit to the snia data i.e. @xmath222 ( even though significantly worse than , which corresponds to @xmath223 ) . notice that shifting the observer by @xmath79 results only in a small improvement of the fit ( @xmath224 ) with respect to the on - center case . using eqs . ( [ cmbalm ] ) , ( [ tempfluct ] ) it is straightforward to obtain numerically @xmath225 , the additional cmb fluctuations dipole moment induced by the shift @xmath79 of the observer . this quantity is shown in fig . [ fig7 ] for various values of the inhomogeneity scale @xmath5 . in the same plot we show the value of the measured cmb dipole ( dashed line ) @xcite @xmath226 & & \label{cobedip2}\end{aligned}\ ] ] clearly , for @xmath227 the observer is confined to be in a sphere which is a fine tuned small spatial fraction of the dark energy inhomogeneity ( @xmath228 ) which implies severe fine tuning of the model as in the case of ltb models with matter . this is shown in fig . [ fig8 ] where @xmath229 is plotted for values of @xmath5 between 1 and 7 gpc . as shown in that figure , the fine tuning starts to reduce ( @xmath229 increases ) as the size of the inhomogeneity increases beyond @xmath230 . we anticipate it to disappear as the size of the inhomogeneity reaches @xmath231gpc ( the comoving distance to the last scattering surface ) . a potentially interesting observation of this class of models is the natural alignment of the low cmb multipoles since the model includes the existence of a preferred axis which connects the observer with the center of the inhomogeneity . the predicted cmb maps corresponding to the dipole quadrupole and octopole are shown in fig . [ fig9 ] ( b - d ) where we have selected the direction of the preferred axis to coincide with the direction of the observed cmb dipole and set @xmath232 mpc . to try to avoid complications related to the numerical precision of the computations ( complications that become manifest mainly for low values of @xmath79 ) , we have computed the quadrupole ( octopole ) in the following way : we first calculated its value ( using eqs . ( [ cmbalm ] ) , ( [ tempfluct ] ) ) for 30 values of @xmath79 between 0 and 1.2 gpc and then performed a quadratic ( cubic ) fit to those data . the values for @xmath233 and @xmath234 used in figs . [ fig9 ] and [ fig10 ] have been derived from those fits . notice that the predicted magnitude of the quadrupole is comparable with the observed value . however , the octopole is smaller than the observed value by a factor @xmath235 . even though the alignment is apparent , the fact that all the @xmath165 coefficients with @xmath236 vanish destroys all the planar features on the plane perpendicular to the preferred axis . such features could naturally appear if we had considered a _ lattice _ of spherical dark energy inhomogeneities . in fig . [ fig10 ] we show the quadrupole coefficient @xmath233 as a function of the observer shift @xmath79 . let us recall that the angular power spectrum is defined through : = _ ll_mmc_l . each @xmath237 can thus be computed by averaging the squares of the corresponding @xmath238 coefficients @xmath165 . however , in our case , due to the axial symmetry ( and assuming that the observed anisotropy were due only to our shift from the inhomogeneity center ) we have just : c_l= where , in addition , we should multiply by @xmath239k in order to get the @xmath237 s in units of @xmath240k@xmath241 , which is the usual convention . the yellow shaded region in the plot represents the value measured by wmap , @xmath242 @xcite , where the uncertainty only includes the measurement error . this corresponds to : @xmath243 we also plot the 68% ( gray ) and 95% ( blue ) confidence level regions according to the more complete statistical analysis carried out in @xcite . notice that the observed value of the quadrupole is reproduced for approximately the same shift of the observer @xmath79 as the one required to produce the observed dipole . this relatively large value of the quadrupole we have derived is in agreement with the results of ref . @xcite but is significantly larger than the value obtained in ref . @xcite for a purely matter void profile . we have considered a generalization of by allowing spatial instead of temporal dependence of the dark energy . an advantage of this class of models is that they naturally predict a preferred cosmological direction in agreement with the indications of some cosmological observations @xcite . we have considered isocurvature spherically symmetric dark energy inhomogeneities motivated from the possible recent formation of global monopoles with hubble - scale cores . we have found that such spherical dark energy inhomogeneities with radius less than about @xmath244 are excluded at the @xmath1 confidence level by the union2 snia data . inhomogeneities with larger scales are consistent with the snia data and with the cmb dipole provided that the observer remains close to the center of the inhomogeneity , at a distance smaller than a maximum distance @xmath7 . for the snia union2 data we find @xmath245 demanding @xmath246 . this is a factor four larger than the corresponding maximum observer shift in a ltb model based on a matter void implying that the fine tuning of the observer location is somewhat reduced in the case of inhomogeneous dark energy . this is partly due to the increase on the size of the observationally allowed inhomogeneities . the corresponding bound obtained from the cmb dipole however is not affected significantly . we find @xmath247 even for inhomogeneities as large as @xmath248 . this value of @xmath7 increases rapidly as the size of the inhomogeneity approaches the comoving radial distance to the last scattering surface @xmath249 . in this limit , the model practically reduces to . our results have demonstrated that hubble - scale inhomogeneities of dark energy are consistent with cmb and snia cosmological observations provided that the inhomogeneity scale is larger than about @xmath244 and the observer is kept close to the center . potential future extensions of this project include the following : * construct a concrete physical model that can give rise to such large - scale dark energy inhomogeneities . topological monopole configurations discussed in section 2b constitute a possible physical basis for such models . * in the context of the physical model mentioned above , one could try to derive predictions for other cosmological observations . for example , the coupling of electromagnetism or gravity to the inhomogeneous fields of the topological monopole could lead to a natural spatial variation of fundamental constants @xcite which would be more prominent along the preferred direction that connects the observer to the center of the inhomogeneity . a similar coupling could lead to a spatial coherent rotation of the optical polarization of quasars @xcite . * even though the model does predict an alignment of the low cmb multipoles , this alignment does not affect the multipoles with @xmath236 . however , this is no longer true if we consider a superposition of multiple spherical inhomogeneities with scales and distances between their centers of a few @xmath112 . such a lattice would keep the existence of a preferred axis and possible introduce concentric cycles in the maps @xcite while inducing at the same time alignment between multipoles @xmath165 with @xmath236 . the study of the observational predictions of such a lattice would be an interesting extension of this project . * hubble - scale dark energy inhomogeneities with different symmetries would also be potentially interesting . for example , global vortices would give rise to cylindrically symmetric inhomogeneities while a domain wall would give rise to planar symmetry . * it has been shown that galaxies in the inner region of a spherical inhomogeneity are seen by an off - center observer to have a bulk velocity flow relative to matter outside of the inhomogeneity . the derivation of the predicted profile of such velocity flows in models with inhomogeneous dark energy would also be an interesting extension of the present study . in conclusion , the introduction of spatial variation of dark energy density on hubble scales constitutes a natural and straightforward generalization of which is supported by concrete physical mechanisms . the detailed study of this class of models and its comparison with cosmological observations constitutes an interesting task that has been initiated by the present study . + * numerical analysis files : * the _ mathematica _ files used to produce the figures are available by e - mail upon request and they will also be available shortly at http://leandros.physics.uoi.gr/ide . we thank valerio marra for useful discussion at the early stages of this work . we have benefited from reading the publicly available code of ref . @xcite . 99 e. komatsu _ et al . _ [ wmap collaboration ] , astrophys . j. suppl . * 192 * , 18 ( 2011 ) [ arxiv:1001.4538 [ astro-ph.co ] ] . p. j. e. peebles and b. ratra , rev . phys . * 75 * , 559 ( 2003 ) [ arxiv : astro - ph/0207347 ] ; t. padmanabhan , phys . rept . * 380 * , 235 ( 2003 ) [ arxiv : hep - th/0212290 ] ; s. m. carroll , living rev . * 4 * , 1 ( 2001 ) [ arxiv : astro - ph/0004075 ] ; v. sahni , class . * 19 * , 3435 ( 2002 ) [ arxiv : astro - ph/0202076 ] . m. tegmark , a. de oliveira - costa and a. hamilton , phys . d * 68 * , 123523 ( 2003 ) [ arxiv : astro - ph/0302496 ] . k. land and j. magueijo , phys . lett . * 95 * , 071301 ( 2005 ) [ arxiv : astro - ph/0502237 ] . c. j. copi , d. huterer , d. j. schwarz and g. d. starkman , adv . astron . * 2010 * , 847541 ( 2010 ) [ arxiv:1004.5602 [ astro-ph.co ] ] . c. l. bennett _ et al . _ , astrophys . j. suppl . * 192 * , 17 ( 2011 ) [ arxiv:1001.4758 [ astro-ph.co ] ] . c. g. park , c. park and j. r. i. gott , astrophys . j. * 660 * , 959 ( 2007 ) [ arxiv : astro - ph/0608129 ] . l. r. abramo , a. bernui , i. s. ferreira , t. villela and c. a. wuensche , phys . d * 74 * , 063506 ( 2006 ) [ arxiv : astro - ph/0604346 ] . a. gruppuso and k. m. gorski , jcap * 1003 * , 019 ( 2010 ) [ arxiv:1002.3928 [ astro-ph.co ] ] . a. pontzen and h. v. peiris , phys . rev . d * 81 * , 103008 ( 2010 ) [ arxiv:1004.2706 [ astro-ph.co ] ] . a. rakic and d. j. schwarz , phys . d * 75 * , 103002 ( 2007 ) [ arxiv : astro - ph/0703266 ] . r. watkins , h. a. feldman and m. j. hudson , mon . not . soc . * 392 * , 743 ( 2009 ) [ arxiv:0809.4041 [ astro - ph ] ] . a. kashlinsky , f. atrio - barandela , d. kocevski and h. ebeling , astrophys . j. * 686 * , l49 ( 2009 ) [ arxiv:0809.3734 [ astro - ph ] ] . g. lavaux , r. b. tully , r. mohayaee and s. colombi , astrophys . j. * 709 * , 483 ( 2010 ) [ arxiv:0810.3658 [ astro - ph ] ] . d. hutsemekers , r. cabanac , h. lamy and d. sluse , astron . astrophys . * 441 * , 915 ( 2005 ) [ arxiv : astro - ph/0507274 ] ; d. hutsemekers , a. payez , r. cabanac , h. lamy , d. sluse , b. borguet and j. r. cudell , arxiv:0809.3088 [ astro - ph ] ; d. hutsemekers and h. lamy , astron . astrophys . * 367 * , 381 ( 2001 ) [ arxiv : astro - ph/0012182 ] . t. j. broadhurst , m. takada , k. umetsu , x. kong , n. arimoto , m. chiba and t. futamase , astrophys . j. * 619 * , l143 ( 2005 ) [ arxiv : astro - ph/0412192 ] . k. umetsu and t. broadhurst , astrophys . j. * 684 * , 177 ( 2008 ) [ arxiv:0712.3441 [ astro - ph ] ] . a. a. klypin , a. v. kravtsov , o. valenzuela and f. prada , astrophys . j. * 522 * , 82 ( 1999 ) [ arxiv : astro - ph/9901240 ] . b. moore , aip conf . proc . * 586 * , 73 ( 2001 ) [ arxiv : astro - ph/0103100 ] . p. madau , j. diemand and m. kuhlen , astrophys . j. * 679 * , 1260 ( 2008 ) [ arxiv:0802.2265 [ astro - ph ] ] . g. gentile _ et al . _ , mon . not . soc . * 351 * , 903 ( 2004 ) [ arxiv : astro - ph/0403154 ] ; g. gentile _ et al . _ , astrophys . j. * 634 * , l145 ( 2005 ) [ arxiv : astro - ph/0510607 ] ; j. d. simon _ et al . _ , astrophys . j. * 621 * , 757 ( 2005 ) [ arxiv : astro - ph/0412035 ] ; w. j. g. de blok , astrophys . j. * 634 * , 227 ( 2005 ) [ arxiv : astro - ph/0506753 ] . for a review , see e. j. copeland , m. sami and s. tsujikawa , int . j. mod . d15 * , 1753 ( 2006 ) [ arxiv : hep - th/0603057 ] . b. boisseau , g. esposito - farese , d. polarski and a. a. starobinsky , phys . lett . * 85 * , 2236 ( 2000 ) [ arxiv : gr - qc/0001066 ] ; s. nesseris and l. perivolaropoulos , phys . d * 75 * , 023517 ( 2007 ) [ arxiv : astro - ph/0611238 ] ; j. p. uzan , phys . d * 59 * , 123510 ( 1999 ) [ arxiv : gr - qc/9903004 ] . a. lue , phys . rept . * 423 * , 1 ( 2006 ) [ arxiv : astro - ph/0510068 ] . s. nojiri and s. d. odintsov , in _ problems of modern theoretical physics _ ( tspu publishing , tomsk , 2008 ) , p.266 [ arxiv:0807.0685 [ hep - th ] ] ; s. fay , s. nesseris and l. perivolaropoulos , phys . rev . d * 76 * , 063504 ( 2007 ) [ arxiv : gr - qc/0703006 ] ; v. faraoni , phys . d * 74 * , 023529 ( 2006 ) [ arxiv : gr - qc/0607016 ] . g. lemaitre , annales soc . math . astron . a * 53*:51 , 1933 . for an english translation , see : g. lemaitre , `` the expanding universe '' , gen . * 29 * , 641 ( 1997 ) . r. c. tolman , proc . sci . * 20 * , 169 ( 1934 ) . h. bondi , mon . not . soc . * 107 * , 410 ( 1947 ) . a. krasinski , `` inhomogeneous cosmological models '' , cambridge university press ( 1997 ) . h. alnes , m. amarzguioui and o. grn , phys . d * 73 * , 083519 ( 2006 ) [ arxiv : astro - ph/0512006 ] . a. vilenkin , phys . * 72 * , 3137 ( 1994 ) [ arxiv : hep - th/9402085 ] . j. grande and l. perivolaropoulos ( in preparation ) . r. amanullah _ et al . _ , astrophys . j. * 716 * , 712 ( 2010 ) [ arxiv:1004.1711 [ astro-ph.co ] ] . m. barriola and a. vilenkin , phys . lett . * 63 * , 341 ( 1989 ) . t. w. b. kibble , phys . rept . * 67 * , 183 ( 1980 ) . j. h. traschen , n. turok and r. h. brandenberger , phys . d * 34 * , 919 ( 1986 ) ; r. a. battye and j. weller , phys . d * 61 * , 043501 ( 2000 ) [ arxiv : astro - ph/9810203 ] . j. k. webb , j. a. king , m. t. murphy , v. v. flambaum , r. f. carswell and m. b. bainbridge , arxiv:1008.3907 [ astro-ph.co ] . n. sakai , h. a. shinkai , t. tachizawa and k. i. maeda , phys . rev . d * 53 * , 655 ( 1996 ) [ erratum - ibid . d * 54 * , 2981 ( 1996 ) ] [ arxiv : gr - qc/9506068 ] . v. marra and m. paakkonen , jcap * 1012 * , 021 ( 2010 ) [ arxiv:1009.4193 [ astro-ph.co ] ] . h. alnes and m. amarzguioui , phys . d * 74 * , 103520 ( 2006 ) , [ arxiv : astro - ph/0607334 ] . k. enqvist and t. mattsson , jcap * 0702 * , 019 ( 2007 ) [ arxiv : astro - ph/0609120 ] . k. tomita , astrophys . j. * 529 * , 26 ( 2000 ) [ arxiv : astro - ph/9905278 ] . c. quercellini , l. amendola , a. balbi , p. cabella and m. quartin , arxiv:1011.2646 [ astro-ph.co ] . m. blomqvist and e. mortsell , jcap * 1005 * , 006 ( 2010 ) [ arxiv:0909.4723 [ astro-ph.co ] ] . h. alnes and m. amarzguioui , phys . d * 75 * , 023506 ( 2007 ) [ arxiv : astro - ph/0610331 ] . m. blomqvist , j. enander and e. mortsell , jcap * 1010 * , 018 ( 2010 ) [ arxiv:1006.4638 [ astro-ph.co ] ] . s. nesseris and l. perivolaropoulos , phys . d * 72 * , 123519 ( 2005 ) [ arxiv : astro - ph/0511040 ] . c. h. lineweaver , in _ microwave background anisotropies _ , edited by f. r. bouchet _ et al . _ ( frontieres , gif - sur - yvette , 1997 ) [ arxiv : astro - ph/9609034 ] . l. campanelli , p. cea and l. tedesco , mod . a * 22 * , 1013 ( 2007 ) [ arxiv : astro - ph/0510825 ] . k. a. olive , m. peloso and j. p. uzan , phys . d * 83 * , 043509 ( 2011 ) [ arxiv:1011.1504 [ astro-ph.co ] ] . i. k. wehus and h. k. eriksen , arxiv:1012.1268 [ astro-ph.co ] . v. g. gurzadyan and r. s. penrose , arxiv:1011.3706 [ astro-ph.co ] .

in this article , we analyze the vertices , , and within the framework of the light - cone qcd sum rules approach in an unified way . the strong coupling constants and are important parameters in evaluating the charmonium absorption cross sections in searching for the quark - gluon plasmas , our numerical values of the and are compatible with the existing estimations although somewhat smaller , the symmetry breaking effects are very large , about . for the charmed scalar mesons and , we take the point of view that they are the conventional and mesons respectively , and calculate the strong coupling constants and with the vector interpolating currents . the numerical values of the scalar- and - coupling constants and are compatible with the existing estimations , the large values support the hadronic dressing mechanism . furthermore , we study the dependence of the four strong coupling constants , , and on the non - perturbative parameter of the twist-2 meson light - cone distribution amplitude . + z. g. wang , s. l. wan + department of physics , north china electric power university , baoding 071003 , p. r. china + department of modern physics , university of science and technology of china , hefei 230026 , p. r. china + pacs numbers : 12.38.lg ; 13.25.jx ; 14.40.cs * key words : * strong coupling constants , light - cone qcd sum rules

hep-ph0606002

the suppression of the @xmath22 production in relativistic heavy ion collisions maybe one of the important signatures to identify the possible phase transition to the quark - gluon plasma @xcite . the dissociation of the @xmath22 in the quark - gluon plasma due to color screening can lead to a reduction of its production , however , the @xmath22 suppression maybe already present in the hadron - nucleus collisions . it is necessary to separate the absorption of the @xmath22 by the nucleons and by the co - mover light mesons ( @xmath23 , @xmath19 , @xmath24 , @xmath25 , etc . ) before we can make a definitive conclusion about the formation of the quark - gluon plasma . it is of great importance to understand the @xmath22 production and absorption mechanisms in the hadronic matter . the values of the @xmath22 absorption cross sections by the light hadrons are not known empirically , we have to resort to some theoretical approaches . among existing approaches for evaluating the charmonium absorption cross sections by the light hadrons , the one - meson exchange model and the effective @xmath6 theory are typical @xcite . the detailed knowledge about the hadronic vertices or the strong coupling constants which are basic parameters in the effective lagrangians is of great importance . the discovery of the two strange - charmed mesons @xmath26 and @xmath27 with spin - parity @xmath28 and @xmath29 respectively has triggered hot debate on their nature , under - structures and whether it is necessary to introduce the exotic states @xcite . the mass of the @xmath26 is significantly lower than the values of the @xmath28 state mass from the quark models and lattice simulations @xcite . the difficulties to identify the @xmath26 and @xmath30 states with the conventional @xmath31 mesons are rather similar to those appearing in the light scalar mesons below @xmath32 . among the various explanations , the hadronic dressing mechanism is typical . the scalar mesons @xmath33 , @xmath34 , @xmath8 and @xmath9 may have bare @xmath35 , @xmath36 and @xmath11 kernels in the @xmath37wave states with strong coupling to the nearby threshold respectively , the @xmath38wave virtual intermediate hadronic states ( or the virtual mesons loops ) play a crucial role in the composition of those bound states ( or resonances due to the masses below or above the thresholds ) . the hadronic dressing mechanism ( or unitarized quark models ) takes the point of view that the @xmath34 , @xmath33 , @xmath8 and @xmath9 mesons have small @xmath39 , @xmath36 and @xmath11 kernels of the typical @xmath39 , @xmath10 and @xmath11 mesons size respectively . the strong couplings to the virtual intermediate hadronic states ( or the virtual mesons loops ) may result in smaller masses than the conventional scalar @xmath39 , @xmath10 and @xmath11 mesons in the constituent quark models , enrich the pure @xmath39 , @xmath10 and @xmath11 states with other components @xcite . those mesons may spend part ( or most part ) of their lifetime as virtual @xmath40 , @xmath14 and @xmath15 states @xcite . it is interesting to study the possibility of the hadronic dressing mechanism . in this article , we calculate the values of the strong coupling constants @xmath41 , @xmath42 , @xmath12 and @xmath13 within the framework of the light - cone qcd sum rules approach . the light - cone qcd sum rules approach carries out the operator product expansion near the light - cone @xmath43 instead of the short distance @xmath44 while the non - perturbative matrix elements are parameterized by the light - cone distribution amplitudes which classified according to their twists instead of the vacuum condensates @xcite . furthermore , we study the dependence of the strong coupling constants @xmath41 , @xmath42 , @xmath12 and @xmath13 on the coefficient @xmath18 of the twist-2 @xmath19 meson light - cone distribution amplitude @xmath45 , and estimate the values of the non - perturbative parameter . it is very difficult to determine the @xmath18 with the qcd sum rules , the values of the @xmath18 suffer from large uncertainties , as it concerns high dimension vacuum condensates which are known poorly @xcite . it is of great importance to determine the values directly from the experimental data . the article is arranged as : in section 2 , we derive the strong coupling constants @xmath41 , @xmath46 , @xmath47 and @xmath13 within the framework of the light - cone qcd sum rules approach ; in section 3 , the numerical results and discussions ; and in section 4 , conclusion . in the following , we write down the definitions for the strong coupling constants @xmath41 , @xmath46 , @xmath12 and @xmath13 , @xmath48 here the @xmath49 are the polarization vectors of the mesons @xmath50 and @xmath51 . we study the strong coupling constants @xmath52 , @xmath46 , @xmath12 and @xmath13 with the interpolating currents @xmath53 , @xmath54 , @xmath55 and @xmath56 in an unified way , and choose the two - point correlation functions @xmath57 and @xmath58 , @xmath59 the correlation functions @xmath60 can be decomposed as @xmath61 due to the lorentz covariance . in this article , we derive the sum rules with the tensor structures @xmath62 and @xmath63 respectively , and make detailed studies . according to the basic assumption of current - hadron duality in the qcd sum rules approach @xcite , we can insert a complete series of intermediate states with the same quantum numbers as the current operators @xmath64 ( @xmath65 ) and @xmath56 ( @xmath66 ) into the correlation function @xmath67 ( @xmath68 ) to obtain the hadronic representation . after isolating the ground states and the first orbital excited states contributions from the pole terms of the @xmath69 , @xmath50 and @xmath8 ( @xmath70 , @xmath51 and @xmath9 ) mesons , the correlation function @xmath71 ( @xmath72 ) can be expressed in terms of the strong coupling constants @xmath73 and the decay constants @xmath74 of the heavy mesons , the explicit expressions are presented in the appendix . we use the standard definitions for the decay constants @xmath74 ( @xmath75 , @xmath76 , @xmath77 , @xmath78 , @xmath79 , @xmath80 ) of the heavy mesons , @xmath81 the quarks @xmath82 and @xmath83 have finite and non - equal masses , the non - conservation of the vector currents @xmath66 and @xmath56 can lead to the non - vanishing couplings to the scalar mesons @xmath9 and @xmath8 beside the vector mesons @xmath51 and @xmath50 , we can study the properties of those mesons with the two interpolating currents @xmath66 and @xmath56 in an unified way . here we have not shown the contributions from the high resonances and continuum states explicitly as they are suppressed due to the double borel transformation . the numerical values of the fractions @xmath84 are less than @xmath85 and the corresponding spectral densities for the ground states are greatly suppressed , the tensor structures with @xmath86 are especially suitable for studying the first orbital excited states @xmath8 and @xmath9 with the vector currents . the numerical values of the fractions @xmath87 are about @xmath88 , the tensor structures with @xmath89 are especially suitable for studying the ground states @xmath50 and @xmath51 with the vector currents . now we carry out the operator product expansion near the light - cone @xmath43 to obtain the representation at the level of quark - gluon degrees of freedom for the correlation functions @xmath90 and @xmath72 . in the following , we briefly outline the operator product expansion for the correlation functions @xmath90 and @xmath72 in perturbative qcd theory . the calculations are performed at the large space - like momentum regions @xmath91 and @xmath92 , which correspond to the small light - cone distance @xmath43 required by the validity of the operator product expansion approach . we write down the propagator of a massive quark in the external gluon field in the fock - schwinger gauge firstly @xcite , @xmath93\right\}\ , , \end{aligned}\ ] ] here @xmath94 is the gluonic field strength , @xmath95 denotes the strong coupling constant . substituting the above @xmath82 quark propagator and the corresponding @xmath19 meson light - cone distribution amplitudes into the correlation functions @xmath90 and @xmath72 in eqs.(2 - 3 ) and completing the integrals over the variables @xmath96 and @xmath97 , finally we obtain the representation at the level of quark - gluon degrees of freedom , the explicit expressions are presented in the appendix . in calculation , we have used the two - particle and three - particle @xmath19 meson light - cone distribution amplitudes @xcite , the explicit expressions are also presented in the appendix . the parameters in the light - cone distribution amplitudes are scale dependent and can be estimated with the qcd sum rules approach @xcite . in this article , the energy scale @xmath98 is chosen to be @xmath99 . we perform the double borel transformation with respect to the variables @xmath100 and @xmath101 for the correlation functions @xmath102 and @xmath103 , and obtain the analytical expressions for those invariant functions , the explicit expressions are presented in the appendix . in order to match the duality regions below the thresholds @xmath104 and @xmath105 for the interpolating currents @xmath106(@xmath107 ) and @xmath108 ( @xmath65 ) respectively , we can express the correlation functions @xmath102 and @xmath103 at the level of quark - gluon degrees of freedom into the following form , @xmath109 then we perform the double borel transformation with respect to the variables @xmath100 and @xmath110 directly . however , the analytical expressions for the spectral densities @xmath111 are hard to obtain , we have to resort to some approximations . as the contributions from the higher twist terms are suppressed by more powers of @xmath112 or @xmath113 , the continuum subtractions will not affect the results remarkably , here we will use the expressions in eqs.(28 - 29 ) for the three - particle ( quark - antiquark - gluon ) twist-3 , twist-4 terms , and the two - particle twist-4 terms . in fact , their contributions are of minor importance , the dominating contributions come from the two - particle twist-2 and twist-3 terms involving the @xmath45 , @xmath114 and @xmath115 . we perform the same trick as refs.@xcite and expand the amplitudes @xmath45 , @xmath114 and @xmath115 in terms of polynomials of @xmath116 , @xmath117 then introduce the variable @xmath118 and the spectral densities are obtained . after straightforward but cumbersome calculations , we can obtain the final expressions for the double borel transformed correlation functions @xmath119 at the level of quark - gluon degrees of freedom below the thresholds . the masses of the charmed mesons are @xmath120 , @xmath121 , @xmath122 , @xmath123 , @xmath124 , and @xmath125 , the ratios are @xmath126 , @xmath127 , @xmath128 and @xmath129 @xcite . there exist overlapping working windows for the two borel parameters @xmath130 and @xmath131 . it s convenient to take the value @xmath132 , @xmath133 , @xmath134 , furthermore , the @xmath19 meson light - cone distribution amplitudes are known quite well at the value @xmath135 comparing with the values at the end - points . we can introduce the threshold parameter @xmath104 and make the simple replacement , @xmath136 to subtract the contributions from the higher resonances and continuum states @xcite , finally we obtain the following sum rules , @xmath137 @xmath138 @xmath139 @xmath140 @xmath141 @xmath142 the explicit expressions of the notations @xmath143 , @xmath144 , @xmath145 , @xmath146 , @xmath147 and @xmath148 are lengthy and given explicitly in the appendix . a slight different manipulation ( with the techniques taken in the ref.@xcite ) for the dominating contributions come from the terms involving the two - particle twist-2 and twist-3 light - cone distribution amplitudes @xmath45 , @xmath114 and @xmath115 leads to the sum rules with the same type as in ref however , those type sum rules are not stable with respect to the variations of the borel parameter @xmath149 , here we will not show the expressions explicitly for simplicity . it is not surprise that the qcd sum rules as a qcd model have both advantages and shortcomings . the input parameters are taken as @xmath150 , @xmath151 , @xmath152 , @xmath153 , @xmath154 , @xmath155 , @xmath156 , @xmath157 @xcite , @xmath158 , @xmath159 , @xmath160 , @xmath161 , @xmath162 . in this article , we take the values of the @xmath18 to be zero , and explore the dependence of the strong coupling constants @xmath4 , @xmath5 , @xmath16 and @xmath17 on this parameter . for the threshold parameter @xmath163 , we can use the experimental data as a guide , @xmath164 , @xmath165 @xcite , and choose the values @xmath166 to subtract the contributions from the high resonances and continuum states . the mass and width of the @xmath8 from belle and focus are @xmath167 , @xmath168 @xcite , @xmath169 , @xmath170 @xcite . the predictions from the constituent quark models are @xmath171 @xcite . the values of the mass from the two collaborations have the difference about @xmath172 , in this article , we take the value @xmath171 as input parameter , our final numerical results for the large strong coupling constant @xmath16 support smaller values for the @xmath8 if the same mechanism takes place for both the charmed scalar mesons @xmath8 and @xmath9 . furthermore , the strong coupling constant @xmath16 is not sensitive to the values of the @xmath173 , taking the values @xmath171 or @xmath174 can not change the conclusion qualitatively or quantitatively . for the threshold parameters @xmath175 , @xmath176 and @xmath177 , the experimental values of the masses are @xmath178 , @xmath179 and @xmath180 , the widths are very narrow @xcite . we can choose the values of the threshold parameters @xmath181 , @xmath182 and @xmath183 to subtract the contributions from the high resonances and continuum states . from figs.1 - 3 , we can see that the numerical values of the strong coupling constants @xmath4 and @xmath16 are not sensitive to the threshold parameters @xmath184 in those regions , the values we chosen here are reasonable . with the parameters @xmath149 and @xmath175 from eq.(10 ) . ] with the parameters @xmath149 and @xmath163 from eq.(12 ) . ] with the parameters @xmath149 and @xmath163 from eq.(14 ) . ] the values of the decay constants @xmath76 , @xmath75 , @xmath78 , @xmath185 , @xmath186 and @xmath79 vary in a large range , for example , @xmath187 , @xmath188 @xcite , @xmath189 , @xmath190 @xcite , @xmath191 @xcite , @xmath192 , @xmath193 @xcite , @xmath194 @xcite from the qcd sum rules ; @xmath195 , @xmath196 , @xmath197 , @xmath198 @xcite , @xmath199 , @xmath200 @xcite , @xmath201 , @xmath202 @xcite from the potential models ; @xmath203 , @xmath204 @xcite from the quark models , and @xmath205 from the experimental data @xcite . for a review of the values of the decay constants for the mesons @xmath70 and @xmath206 from the qcd sum rules and lattice qcd , one can consult the second article of the ref.@xcite . in this article , we take the following constraints for the decay constants , @xmath207 and choose the values , @xmath208 in numerical calculation , we observe that the values of the strong coupling constants @xmath4 , @xmath5 , @xmath16 and @xmath17 are sensitive to the six hadronic parameters , small variations of those parameters can lead to relatively large changes for the numerical values , refining the six hadronic parameters is of great importance . the borel parameters in eqs.(10 - 11 ) are taken as @xmath209 and @xmath210 , in those regions , the values of the strong coupling constants @xmath4 and @xmath5 are rather stable from the sum rules in eqs.(10 - 11 ) with the simple subtraction , which are shown , for example , in the fig.1 and figs.4 - 7 for the strong coupling constant @xmath4 , similar figures can be obtained if the values of the strong coupling constant @xmath5 are plotted . in this article , we only show the numerical values from the sum rules in eq.(10 ) , eq.(12 ) and eq.(14 ) explicitly for simplicity . the borel parameters in eqs.(12 - 15 ) are chosen as @xmath211 and @xmath212 , in those regions , the values of the strong coupling constants @xmath16 and @xmath17 are rather stable from the sum rules in eqs.(12 - 13 ) with the simple subtraction , which are shown in the fig2 , figs.4 - 6 and fig.8 for an illustration . however , the strong coupling constants @xmath16 and @xmath17 from the sum rules in eqs.(14 - 15 ) have a negative sign comparing with the corresponding ones from the sum rules in eqs.(12 - 13 ) , and much smaller absolute values . the fractions @xmath87 are about @xmath88 . in the sum rules in eqs.(10 - 11 ) , the ground state saturate condition can be safely satisfied below the threshold @xmath175 ( @xmath213 ) . the vector interpolating current @xmath214 ( @xmath215 ) has both non - vanishing couplings to the vector state @xmath50 ( @xmath51 ) and to the scalar state @xmath8 ( @xmath9 ) , there are two hadronic states , the ground state @xmath50 ( @xmath51 ) and the first orbital excited state @xmath8 ( @xmath9 ) in the channel @xmath36 ( @xmath216 ) below the threshold @xmath163 ( @xmath177 ) , the ground states @xmath50 and @xmath51 are not suppressed due to the factor @xmath88 , the sum rules in eqs.(14 - 15 ) are not suitable for studying the strong coupling constants @xmath16 and @xmath17 , our final numerical values support this assumption . we show this fact in the fig.3 for an illustration . we determine the values of the strong coupling constants @xmath4 and @xmath5 from the eq.(10 ) and eq.(11 ) respectively , then use those values as the input parameters , and calculate the values of the strong coupling constants @xmath16 and @xmath17 from the eqs.(12 - 15 ) respectively . the uncertainties of the five parameters @xmath217 , @xmath218 , @xmath219 , @xmath220 and @xmath221 can not lead to large uncertainties for the numerical values . the main uncertainties come from the ten parameters @xmath222 , @xmath223 , @xmath224 , @xmath225 , @xmath76 , @xmath78 , @xmath185 , @xmath186 , @xmath75 and @xmath79 , small variations of those parameters can lead to relatively large changes for the numerical values , which are shown in the figs.4 - 8 for an illustration . and @xmath226 with the parameters @xmath149 and @xmath222 from eq.(10 ) and eq.(12 ) respectively . , title="fig:",width=264 ] and @xmath226 with the parameters @xmath149 and @xmath222 from eq.(10 ) and eq.(12 ) respectively . , title="fig:",width=264 ] and @xmath226 with the parameters @xmath149 and @xmath75 from eq.(10 ) and eq.(12 ) respectively.,title="fig:",width=264 ] and @xmath226 with the parameters @xmath149 and @xmath75 from eq.(10 ) and eq.(12 ) respectively.,title="fig:",width=264 ] and @xmath226 with the parameters @xmath149 and @xmath224 from eq.(10 ) and eq.(12 ) respectively . , title="fig:",width=264 ] and @xmath226 with the parameters @xmath149 and @xmath224 from eq.(10 ) and eq.(12 ) respectively . , title="fig:",width=264 ] with the parameters @xmath149 and @xmath78 from eq.(10 ) . , width=264 ] with the parameters @xmath149 and @xmath186 , @xmath223 , @xmath225 from eq.(12 ) . , title="fig:",width=264 ] with the parameters @xmath149 and @xmath186 , @xmath223 , @xmath225 from eq.(12 ) . , title="fig:",width=264 ] with the parameters @xmath149 and @xmath186 , @xmath223 , @xmath225 from eq.(12 ) . , title="fig:",width=264 ] taking into account all the uncertainties , finally we obtain the numerical results for the strong coupling constants , @xmath227 which are shown in the figs.9 - 10 respectively . ( a ) and @xmath5(b ) with the parameter @xmath149 from eq.(10 ) and eq.(11 ) respectively.,title="fig:",width=264 ] ( a ) and @xmath5(b ) with the parameter @xmath149 from eq.(10 ) and eq.(11 ) respectively.,title="fig:",width=264 ] ( a ) and @xmath17(b ) with the parameter @xmath149 from eq.(12 ) and eq.(13 ) respectively . , title="fig:",width=264 ] ( a ) and @xmath17(b ) with the parameter @xmath149 from eq.(12 ) and eq.(13 ) respectively . , title="fig:",width=264 ] the strong coupling constants @xmath4 , @xmath5 , @xmath16 and @xmath17 can be related to the parameters @xmath228 and @xmath229 in the heavy - light chiral perturbation theory @xcite , @xmath230 here the @xmath231 are the heavy scalar mesons with @xmath28 , the @xmath232 are the heavy pseudoscalar mesons with @xmath233 , the @xmath234 are the heavy vector mesons with @xmath235 , and the @xmath23 stand for the light pseudoscalar mesons . the parameter @xmath228 has been calculated with the light - cone qcd sum rules @xcite , the quark models @xcite and extracted from the experimental data @xcite . the values vary in a large range , the corresponding values of the strong coupling constants @xmath4 and @xmath5 in the @xmath236 limit for the light pseudoscalar mesons are listed in the table.1 . from the table , we can see that our numerical results are compatible with the existing estimations , although somewhat smaller . the values of the strong coupling constants @xmath4 and @xmath5 are sensitive to the non - perturbative parameter @xmath18 , if we take a larger value rather than zero , larger values of the @xmath4 and @xmath5 are obtained . the @xmath237 and @xmath5 are more sensitive to the @xmath18 comparing with the @xmath226 and @xmath17 , which are shown in the fig.11 . in fact , the largest uncertainties come from the uncertainties of the @xmath18 , they are ideal channels to determine this parameter directly from the experimental data . once the experimental data for the values of the strong coupling constants @xmath237 and @xmath5 are available , powerful constraints can be put on the range of the parameter @xmath18 . if we take the values from the qcd sum rules as input parameters @xcite , @xmath238 and @xmath239 , very large values of the @xmath18 are obtained . the parameter @xmath229 has been estimated with the light - cone qcd sum rules @xcite , the quark models @xcite , adler - weisberger type sum rules @xcite , and extracted from the experimental data @xcite , the values are listed in the table.2 , from those values we can estimate the values of the corresponding strong coupling constants @xmath16 and @xmath17 in the @xmath236 limit for the light pseudoscalar mesons . the value of the dimensionless effective coupling constant @xmath240 from lattice qcd @xcite is somewhat smaller than the values extracted from the experimental data @xmath241 , here the @xmath242 is the decay width and the @xmath97 is the decay momentum . our numerical values @xmath243 and @xmath244 are compatible with the existing estimations in refs.@xcite , although somewhat smaller comparing with the values obtained in ref.@xcite with the scalar interpolating current for the @xmath9 meson , and about @xmath245 times as large as the energy scale @xmath125 , and favor the hadronic dressing mechanism . for a short discussion about the hadronic dressing mechanism , one can consult ref.@xcite , or one can consult the original literatures for the details @xcite . the large values of the strong coupling constants @xmath16 and @xmath17 obviously support the hadronic dressing mechanism , the @xmath8 and @xmath9 ( just like the scalar mesons @xmath34 and @xmath33 , see ref.@xcite ) can be taken as having small scalar @xmath10 and @xmath11 kernels of typical meson size with large virtual s - wave @xmath14 and @xmath15 cloud respectively . in ref.@xcite , the authors analyze the unitarized two - meson scattering amplitudes from the heavy - light chiral lagrangian , and observe that the scalar meson @xmath9 appears as the bound state pole with the strong coupling constant @xmath246 . our numerical results @xmath247 are smaller , the values of our previously work @xmath248 with the scalar interpolating current are more satisfactory @xcite . . numerical values of the parameter @xmath228 , and the corresponding values of the strong coupling constants @xmath4 and @xmath5 in the @xmath236 limit . here we have double the values of our numerical results and the ones from ref.@xcite due to the difference between the definitions for the strong coupling constants . [ cols="^,^,^,^",options="header " , ] ( a ) , @xmath5(b ) , @xmath226(c ) , @xmath17(d ) with the parameters @xmath149 and @xmath18 from eq.(10 ) , eq.(11 ) , eq.(12 ) , eq.(13 ) respectively . , title="fig:",width=264 ] ( a ) , @xmath5(b ) , @xmath226(c ) , @xmath17(d ) with the parameters @xmath149 and @xmath18 from eq.(10 ) , eq.(11 ) , eq.(12 ) , eq.(13 ) respectively . , title="fig:",width=264 ] ( a ) , @xmath5(b ) , @xmath226(c ) , @xmath17(d ) with the parameters @xmath149 and @xmath18 from eq.(10 ) , eq.(11 ) , eq.(12 ) , eq.(13 ) respectively . , title="fig:",width=264 ] ( a ) , @xmath5(b ) , @xmath226(c ) , @xmath17(d ) with the parameters @xmath149 and @xmath18 from eq.(10 ) , eq.(11 ) , eq.(12 ) , eq.(13 ) respectively . , title="fig:",width=264 ] in this article , we analyze the vertices @xmath0 , @xmath1 , @xmath2 and @xmath3 within the framework of the light - cone qcd sum rules approach in an unified way . the strong coupling constants @xmath4 and @xmath5 are important parameters in evaluating the charmonium absorption cross sections in searching for the quark - gluon plasmas , our numerical values of the @xmath4 and @xmath5 are compatible with the existing estimations although somewhat smaller , the @xmath6 symmetry breaking effects are very large , about @xmath7 , the approximation of the @xmath6 symmetry @xmath249 is not suitable @xcite . for the scalar mesons @xmath8 and @xmath9 , we take the point of view that they are the conventional @xmath10 and @xmath11 meson respectively , and calculate the strong coupling constants @xmath250 and @xmath13 within the framework of the light - cone qcd sum rules approach . the numerical values of the scalar-@xmath14 and -@xmath15 coupling constants @xmath12 and @xmath13 are compatible with the existing estimations although somewhat smaller , the large values support the hadronic dressing mechanism . just like the scalar mesons @xmath34 and @xmath33 , the scalar mesons @xmath8 and @xmath9 may have small @xmath10 and @xmath11 kernels of typical @xmath10 and @xmath11 mesons size respectively . the strong coupling to virtual intermediate hadronic states ( or the virtual mesons loops ) can result in smaller mass than the conventional scalar mesons @xmath10 and @xmath11 in the constituent quark models , enrich the pure states @xmath10 and @xmath11 with other components . the @xmath251 and @xmath26 may spend part ( or most part ) of they lifetime as virtual @xmath252 and @xmath253 states . furthermore , we study the dependence of the strong coupling constants @xmath4 and @xmath5 on the non - perturbative parameter @xmath18 of the twist-2 @xmath19 meson light - cone distribution amplitude . the values of the strong coupling constants @xmath237 and @xmath5 are more sensitive to the @xmath18 comparing with the @xmath226 and @xmath17 . the largest uncertainties come from the uncertainties of the @xmath18 , they are the ideal channels to determine the parameter directly from the experimental data . once the experimental data for the values of the strong coupling constants @xmath237 and @xmath5 are available , powerful constraints can be put on the range of the parameter @xmath18 . the explicit expressions of the correlation functions @xmath90 and @xmath72 in the hadronic representation , @xmath254 @xmath255 the explicit expressions of the correlation functions @xmath90 and @xmath72 at the level of quark - gluon degrees of freedom , @xmath256\right . \nonumber\\ & & + m_cf_km_k^2\int_0 ^ 1du\int_0^udt\frac{b(t)}{\left\{(q+up)^2-m_c^2\right\}^2 } \nonumber\\ & & + f_{3k}m_k^2\int_0 ^ 1dv \int_0 ^ 1d\alpha_g \int_0^{1-\alpha_g}d\alpha_s \frac{(2v-3)t(\alpha_u,\alpha_g,\alpha_s)}{\left\{\left[q+((1-v)\alpha_g+\alpha_s)p\right]^2-m_c^2\right\}^2 } \nonumber\\ & & -4m_cf_km_k^4\int_0 ^ 1dv v \int_0 ^ 1 d\alpha_g\int_0^{\alpha_g } d\beta\int_0^{1-\beta}d\alpha \frac{\phi(1-\alpha-\beta,\beta,\alpha)}{\left\{\left[q+(1-v\alpha_g)p\right]^2-m_c^2\right\}^3 } \nonumber \\ & & \left.+4m_cf_km_k^4\int_0 ^ 1 dv\int_0 ^ 1 d\alpha_g\int_0^{1-\alpha_g } d\alpha_s \int_0^{\alpha_s}d\alpha \frac{\phi(1-\alpha-\alpha_g,\alpha_g,\alpha)}{\left\{\left[q+((1-v)\alpha_g+\alpha_s)p\right]^2-m_c^2\right\}^3}\right\ } \nonumber\end{aligned}\ ] ] @xmath257\frac{1}{(q+up)^2-m_c^2}-\frac{2m_c^2}{\left[(q+up)^2-m_c^2\right]^2 } \right\ } \nonumber\\ & & + m_cf_k\int_0 ^ 1du \left\ { \frac{\phi_k}{(q+up)^2-m_c^2}-\frac{m_k^2m_c^2}{2}\frac{a(u)}{\left[(q+up)^2-m_c^2\right]^3 } \right\ } \nonumber\\ & & + f_{3k}m_k^2 \int_0 ^ 1dv\int_0 ^ 1d\alpha_g \int_0^{1-\alpha_g}d\alpha_s \frac { \left[(1-v)\alpha_g+\alpha_s\right]\left(2v -3\right)t(\alpha_u,\alpha_g,\alpha_s ) } { \left\{\left[q+((1-v)\alpha_g+\alpha_s)p\right]^2-m_c^2\right\}^2}\nonumber\\ & & -2f_{3k } \int_0 ^ 1dvv \int_0 ^ 1d\alpha_g \int_0^{1-\alpha_g}d\alpha_s t(\alpha_u,\alpha_g,\alpha_s)\frac{d}{du}\frac{1 } { ( q+up)^2-m_c^2}\mid_{u=(1-v)\alpha_g+\alpha_s}\nonumber \\ & & -4m_cf_km_k^4\int_0 ^ 1dvv \int_0 ^ 1 d\alpha_g\int_0^{\alpha_g } d\beta\int_0^{1-\beta}d\alpha \frac{(1-v\alpha_g)\phi(1-\alpha-\beta,\beta,\alpha)}{\left\{\left[q+(1-v\alpha_g)p\right]^2-m_c^2\right\}^3 } \nonumber \\ & & + 4m_cf_km_k^4\int_0 ^ 1 dv\int_0 ^ 1 d\alpha_g\int_0^{1-\alpha_g } d\alpha_s \int_0^{\alpha_s}d\alpha \frac{((1-v)\alpha_g+\alpha_s)\phi(1-\alpha-\alpha_g,\alpha_g,\alpha ) } { \left\{\left[q+((1-v)\alpha_g+\alpha_s)p\right]^2-m_c^2\right\}^3 } \nonumber \\ & & \left . -m_cf_km_k^2\int_0 ^ 1 dv\int_0 ^ 1 d\alpha_g\int_0^{1-\alpha_g } d\alpha_s \frac{\psi(\alpha_u,\alpha_g,\alpha_s ) } { \left\{\left[q+((1-v)\alpha_g+\alpha_s)p\right]^2-m_c^2\right\}^2}\right\ } \ , , \end{aligned}\ ] ] @xmath258 the light - cone distribution amplitudes of the @xmath19 meson , @xmath259 here the operator @xmath260 is the dual of the @xmath261 , @xmath262 , @xmath263 is defined as @xmath264 , @xmath265 and @xmath266 . the light - cone distribution amplitudes are parameterized as @xmath267c_2^{\frac{3}{2}}(2u-1)\right\}\ , , \nonumber \\ t(\alpha_i,\mu ) & = & 360 \alpha_u \alpha_s \alpha_g^2 \left \{1 + \lambda_3(\alpha_u-\alpha_s)+ \omega_3 \frac{1}{2 } ( 7 \alpha_g - 3 ) \right\ } \ , , \nonumber\\ v_{\parallel}(\alpha_i,\mu ) & = & 120\alpha_u \alpha_s \alpha_g \left ( v_{00}+v_{10}(3\alpha_g-1)\right)\ , , \nonumber \\ a_{\parallel}(\alpha_i,\mu ) & = & 120 \alpha_u \alpha_s \alpha_g a_{10 } ( \alpha_s-\alpha_u)\ , , \nonumber\\ v_{\perp}(\alpha_i,\mu ) & = & -30\alpha_g^2 \left\{h_{00}(1-\alpha_g)+h_{01}\left[\alpha_g(1-\alpha_g)-6\alpha_u \alpha_s\right ] \right . \nonumber\\ & & \left . + h_{10}\left [ \alpha_g(1-\alpha_g)-\frac{3}{2}\left(\alpha_u^2+\alpha_s^2\right)\right]\right\}\ , , \nonumber\\ a_{\perp}(\alpha_i,\mu ) & = & 30 \alpha_g^2 ( \alpha_u-\alpha_s ) \left\{h_{00}+h_{01}\alpha_g+\frac{1}{2}h_{10}(5\alpha_g-3 ) \right\ } , \nonumber\\ a(u,\mu)&=&6u(1-u)\left\ { \frac{16}{15}+\frac{24}{35}a_2 + 20\eta_3+\frac{20}{9}\eta_4 \right . \nonumber \\ & & + \left [ -\frac{1}{15}+\frac{1}{16}-\frac{7}{27}\eta_3\omega_3-\frac{10}{27}\eta_4\right]c^{\frac{3}{2}}_2(2u-1 ) \nonumber\\ & & \left.+\left [ -\frac{11}{210}a_2-\frac{4}{135}\eta_3\omega_3\right]c^{\frac{3}{2}}_4(2u-1)\right\}+\left\ { -\frac{18}{5}a_2 + 21\eta_4\omega_4\right\ } \nonumber\\ & & \left\{2u^3(10 - 15u+6u^2 ) \log u+2\bar{u}^3(10 - 15\bar{u}+6\bar{u}^2 ) \log \bar{u } \right . \nonumber\\ & & \left . + u\bar{u}(2 + 13u\bar{u})\right\ } \ , , \nonumber\\ g_k(u,\mu)&=&1+g_2c^{\frac{1}{2}}_2(2u-1)+g_4c^{\frac{1}{2}}_4(2u-1)\ , , \nonumber\\ b(u,\mu)&=&g_k(u,\mu)-\phi_k(u,\mu)\ , , \end{aligned}\ ] ] where @xmath268 here @xmath269 , @xmath270 and @xmath271 are gegenbauer polynomials , @xmath272 and @xmath273 @xcite . the explicit expressions of the borel transformed correlation functions @xmath274 and @xmath275 in the hadronic representation , @xmath276 @xmath277 here we have not shown the contributions from the high resonances and continuum states explicitly for simplicity . the explicit expressions of the borel transformed correlation functions @xmath274 and @xmath275 at the level of quark - gluon degrees of freedom , @xmath278\right . \nonumber\\ & & + m_cf_km_k^2\int_0^{u_0}dtb(t)-f_{3k}m_k^2 \int_0^{u_0}d\alpha_s \int_{u_0-\alpha_s}^{1-\alpha_s}d\alpha_g \frac{\left[2(\alpha_s - u_0)-\alpha_g\right]t(\alpha_u,\alpha_g,\alpha_s)}{\alpha_g^2 } \nonumber\\ & & -\frac{2m_cf_km_k^4}{m^2 } \int_{1-u_0}^1 d\alpha_g\int_0^{\alpha_g } d\beta\int_0^{1-\beta}d\alpha \frac{(1-u_0)\phi(1-\alpha-\beta,\beta,\alpha)}{\alpha_g^2 } \nonumber \\ & & + \frac{2m_cf_km_k^4}{m^2 } \left[\int_0^{1-u_0 } d\alpha_g\int^{u_0}_{u_0-\alpha_g } d\alpha_s \int_0^{\alpha_s}d\alpha + \int_{1-u_0}^1 d\alpha_g\int_{u_0-\alpha_g}^{1-\alpha_g } d\alpha_s \int_{0}^{\alpha_s}d\alpha\right ] \nonumber\\ & & \left.\frac{\phi(1-\alpha-\alpha_g,\alpha_g,\alpha)}{\alpha_g}\right\ } \nonumber \\ & + & p_\mu \exp\left\{-\frac{u_0(1-u_0)m_k^2+m_c^2}{m^2}\right\}\left\{\frac{f_km_k^2m^2}{m_s } u_0\phi_p(u_0 ) + m_cf_km_k^2 u_0\int_0^{u_0}dt b(t ) \right . \nonumber \\ & & -\frac{f_km_k^2}{6m_s}\left [ \phi_\sigma(u_0)m^2+\frac{d}{du_0}u_0\phi_\sigma(u_0)m^2 + 2m_c^2\phi_\sigma(u_0 ) \right ] \nonumber\\ & & + m_cf_k \left [ -\phi_k(u_0)m^2 + \frac{m_k^2m_c^2a(u_0)}{4m^2 } \right ] \nonumber\\ & & + 3u_0f_{3k}m_k^2\int_0^{u_0}d\alpha_s \int_{u_0-\alpha_s}^{1-\alpha_s}d\alpha_g \frac{t(\alpha_u,\alpha_g,\alpha_s)}{\alpha_g } \nonumber\\ & & + 2f_{3k}\left[m^2\frac{d}{du_0}-u_0m_k^2\right]\int_0^{u_0}d\alpha_s \int_{u_0-\alpha_s}^{1-\alpha_s}d\alpha_g \frac{\alpha_g+\alpha_s - u_0}{\alpha_g^2}t(\alpha_u,\alpha_g,\alpha_s ) \nonumber\\ & & -\frac{2u_0m_cf_km_k^4}{m^2 } \int_{1-u_0}^1 d\alpha_g\int_0^{\alpha_g } d\beta\int_0^{1-\beta}d\alpha \frac{(1-u_0)\phi(1-\alpha-\beta,\beta,\alpha)}{\alpha_g^2 } \nonumber \\ & & + \frac{2u_0m_cf_km_k^4}{m^2}\left[\int_0^{1-u_0 } d\alpha_g\int^{u_0}_{u_0-\alpha_g } d\alpha_s \int_0^{\alpha_s}d\alpha + \int_{1-u_0}^1 d\alpha_g\int_{u_0-\alpha_g}^{1-\alpha_g } d\alpha_s \int_{0}^{\alpha_s}d\alpha\right ] \nonumber\\ & & \left . \frac{\phi(1-\alpha-\alpha_g,\alpha_g,\alpha)}{\alpha_g } + m_cf_km_k^2 \int_0^{u_0 } d\alpha_s\int_{u_0-\alpha_s}^{1-\alpha_s } d\alpha_g \frac{\psi(\alpha_u,\alpha_g,\alpha_s ) } { \alpha_g}\right\}\ , , \\ b_m \pi_\mu^2&=&b_m \pi_\mu^1(u\longleftrightarrow 1-u ; \alpha_s \longleftrightarrow \alpha_u ) \ , , \end{aligned}\ ] ] here @xmath279 , @xmath280 . the explicit expressions of the notations @xmath143 , @xmath144 , @xmath145 , @xmath146 , @xmath147 and @xmath148 , @xmath281 \right\}\nonumber\\ & & + \exp\left\{-\frac{u_0(1-u_0)m_k^2+m_c^2}{m^2}\right\}\left\{m_cf_km_k^2 u_0\int_0^{u_0}dt b(t ) + \frac{f_km_k^2m_c^3a(u_0)}{4m^2 } \right . \nonumber \\ & & + 3u_0f_{3k}m_k^2\int_0^{u_0}d\alpha_s \int_{u_0-\alpha_s}^{1-\alpha_s}d\alpha_g \frac{t(\alpha_u,\alpha_g,\alpha_s)}{\alpha_g } \nonumber\\ & & + 2f_{3k}\left[m^2\frac{d}{du_0}-u_0m_k^2\right]\int_0^{u_0}d\alpha_s \int_{u_0-\alpha_s}^{1-\alpha_s}d\alpha_g \frac{\alpha_g+\alpha_s - u_0}{\alpha_g^2}t(\alpha_u,\alpha_g,\alpha_s ) \nonumber\\ & & -\frac{2u_0m_cf_km_k^4}{m^2 } \int_{1-u_0}^1 d\alpha_g\int_0^{\alpha_g } d\beta\int_0^{1-\beta}d\alpha \frac{(1-u_0)\phi(1-\alpha-\beta,\beta,\alpha)}{\alpha_g^2 } \nonumber \\ & & + \frac{2u_0m_cf_km_k^4}{m^2}\left[\int_0^{1-u_0 } d\alpha_g\int^{u_0}_{u_0-\alpha_g } d\alpha_s \int_0^{\alpha_s}d\alpha + \int_{1-u_0}^1 d\alpha_g\int_{u_0-\alpha_g}^{1-\alpha_g } d\alpha_s \int_{0}^{\alpha_s}d\alpha\right ] \nonumber\\ & & \left . \frac{\phi(1-\alpha-\alpha_g,\alpha_g,\alpha)}{\alpha_g } + m_cf_km_k^2 \int_0^{u_0 } d\alpha_s\int_{u_0-\alpha_s}^{1-\alpha_s } d\alpha_g \frac{\psi(\alpha_u,\alpha_g,\alpha_s ) } { \alpha_g}\right\ } \ , ; \end{aligned}\ ] ] @xmath282 \right\}\nonumber\\ & & + \exp\left\{-\frac{u_0(1-u_0)m_k^2+m_c^2}{m^2}\right\}\left\{m_cf_km_k^2 u_0\int_0^{u_0}dt b(t ) + \frac{f_km_k^2m_c^3a(u_0)}{4m^2 } \right . \nonumber \\ & & + 3u_0f_{3k}m_k^2\int_0^{u_0}d\alpha_s \int_{u_0-\alpha_s}^{1-\alpha_s}d\alpha_g \frac{t(\alpha_u,\alpha_g,\alpha_s)}{\alpha_g } \nonumber\\ & & + 2f_{3k}\left[m^2\frac{d}{du_0}-u_0m_k^2\right]\int_0^{u_0}d\alpha_s \int_{u_0-\alpha_s}^{1-\alpha_s}d\alpha_g \frac{\alpha_g+\alpha_s - u_0}{\alpha_g^2}t(\alpha_u,\alpha_g,\alpha_s ) \nonumber\\ & & -\frac{2u_0m_cf_km_k^4}{m^2 } \int_{1-u_0}^1 d\alpha_g\int_0^{\alpha_g } d\beta\int_0^{1-\beta}d\alpha \frac{(1-u_0)\phi(1-\alpha-\beta,\beta,\alpha)}{\alpha_g^2 } \nonumber \\ & & + \frac{2u_0m_cf_km_k^4}{m^2}\left[\int_0^{1-u_0 } d\alpha_g\int^{u_0}_{u_0-\alpha_g } d\alpha_s \int_0^{\alpha_s}d\alpha + \int_{1-u_0}^1 d\alpha_g\int_{u_0-\alpha_g}^{1-\alpha_g } d\alpha_s \int_{0}^{\alpha_s}d\alpha\right ] \nonumber\\ & & \left . \frac{\phi(1-\alpha-\alpha_g,\alpha_g,\alpha)}{\alpha_g } + m_cf_km_k^2 \int_0^{u_0 } d\alpha_s\int_{u_0-\alpha_s}^{1-\alpha_s } d\alpha_g \frac{\psi(\alpha_u,\alpha_g,\alpha_s ) } { \alpha_g}\right\}\mid_{\alpha_s \leftrightarrow \alpha_u } \ , ; \nonumber \\\end{aligned}\ ] ] @xmath283t(\alpha_u,\alpha_g,\alpha_s)}{\alpha_g^2 } \nonumber\\ & & -\frac{2m_cf_km_k^4}{m^2 } \int_{1-u_0}^1 d\alpha_g\int_0^{\alpha_g } d\beta\int_0^{1-\beta}d\alpha \frac{(1-u_0)\phi(1-\alpha-\beta,\beta,\alpha)}{\alpha_g^2 } \nonumber \\ & & + \frac{2m_cf_km_k^4}{m^2 } \left[\int_0^{1-u_0 } d\alpha_g\int^{u_0}_{u_0-\alpha_g } d\alpha_s \int_0^{\alpha_s}d\alpha + \int_{1-u_0}^1 d\alpha_g\int_{u_0-\alpha_g}^{1-\alpha_g } d\alpha_s \int_{0}^{\alpha_s}d\alpha\right ] \nonumber\\ & & \left.\frac{\phi(1-\alpha-\alpha_g,\alpha_g,\alpha)}{\alpha_g}\right\ } \ , ; \end{aligned}\ ] ] @xmath284t(\alpha_u,\alpha_g,\alpha_s)}{\alpha_g^2 } \nonumber\\ & & -\frac{2m_cf_km_k^4}{m^2 } \int_{1-u_0}^1 d\alpha_g\int_0^{\alpha_g } d\beta\int_0^{1-\beta}d\alpha \frac{(1-u_0)\phi(1-\alpha-\beta,\beta,\alpha)}{\alpha_g^2 } \nonumber \\ & & + \frac{2m_cf_km_k^4}{m^2 } \left[\int_0^{1-u_0 } d\alpha_g\int^{u_0}_{u_0-\alpha_g } d\alpha_s \int_0^{\alpha_s}d\alpha + \int_{1-u_0}^1 d\alpha_g\int_{u_0-\alpha_g}^{1-\alpha_g } d\alpha_s \int_{0}^{\alpha_s}d\alpha\right ] \nonumber\\ & & \left.\frac{\phi(1-\alpha-\alpha_g,\alpha_g,\alpha)}{\alpha_g}\right\ } \mid_{\alpha_s \leftrightarrow \alpha_u } \ , ; \end{aligned}\ ] ] @xmath285 \right\}\nonumber\\ & & + \exp\left\{-\frac{u_0(1-u_0)m_k^2+m_c^2}{m^2}\right\}\left\{m_cf_km_k^2 u_0\int_0^{u_0}dt b(t ) + \frac{f_km_k^2m_c^3a(u_0)}{4m^2 } \right . \nonumber \\ & & + 3u_0f_{3k}m_k^2\int_0^{u_0}d\alpha_s \int_{u_0-\alpha_s}^{1-\alpha_s}d\alpha_g \frac{t(\alpha_u,\alpha_g,\alpha_s)}{\alpha_g } \nonumber\\ & & + 2f_{3k}\left[m^2\frac{d}{du_0}-u_0m_k^2\right]\int_0^{u_0}d\alpha_s \int_{u_0-\alpha_s}^{1-\alpha_s}d\alpha_g \frac{\alpha_g+\alpha_s - u_0}{\alpha_g^2}t(\alpha_u,\alpha_g,\alpha_s ) \nonumber\\ & & -\frac{2u_0m_cf_km_k^4}{m^2 } \int_{1-u_0}^1 d\alpha_g\int_0^{\alpha_g } d\beta\int_0^{1-\beta}d\alpha \frac{(1-u_0)\phi(1-\alpha-\beta,\beta,\alpha)}{\alpha_g^2 } \nonumber \\ & & + \frac{2u_0m_cf_km_k^4}{m^2}\left[\int_0^{1-u_0 } d\alpha_g\int^{u_0}_{u_0-\alpha_g } d\alpha_s \int_0^{\alpha_s}d\alpha + \int_{1-u_0}^1 d\alpha_g\int_{u_0-\alpha_g}^{1-\alpha_g } d\alpha_s \int_{0}^{\alpha_s}d\alpha\right ] \nonumber\\ & & \left . \frac{\phi(1-\alpha-\alpha_g,\alpha_g,\alpha)}{\alpha_g } + m_cf_km_k^2 \int_0^{u_0 } d\alpha_s\int_{u_0-\alpha_s}^{1-\alpha_s } d\alpha_g \frac{\psi(\alpha_u,\alpha_g,\alpha_s ) } { \alpha_g}\right\ } \ , ; \end{aligned}\ ] ] @xmath286 \right\}\nonumber\\ & & + \exp\left\{-\frac{u_0(1-u_0)m_k^2+m_c^2}{m^2}\right\}\left\{m_cf_km_k^2 u_0\int_0^{u_0}dt b(t ) + \frac{f_km_k^2m_c^3a(u_0)}{4m^2 } \right . \nonumber \\ & & + 3u_0f_{3k}m_k^2\int_0^{u_0}d\alpha_s \int_{u_0-\alpha_s}^{1-\alpha_s}d\alpha_g \frac{t(\alpha_u,\alpha_g,\alpha_s)}{\alpha_g } \nonumber\\ & & + 2f_{3k}\left[m^2\frac{d}{du_0}-u_0m_k^2\right]\int_0^{u_0}d\alpha_s \int_{u_0-\alpha_s}^{1-\alpha_s}d\alpha_g \frac{\alpha_g+\alpha_s - u_0}{\alpha_g^2}t(\alpha_u,\alpha_g,\alpha_s ) \nonumber\\ & & -\frac{2u_0m_cf_km_k^4}{m^2 } \int_{1-u_0}^1 d\alpha_g\int_0^{\alpha_g } d\beta\int_0^{1-\beta}d\alpha \frac{(1-u_0)\phi(1-\alpha-\beta,\beta,\alpha)}{\alpha_g^2 } \nonumber \\ & & + \frac{2u_0m_cf_km_k^4}{m^2}\left[\int_0^{1-u_0 } d\alpha_g\int^{u_0}_{u_0-\alpha_g } d\alpha_s \int_0^{\alpha_s}d\alpha + \int_{1-u_0}^1 d\alpha_g\int_{u_0-\alpha_g}^{1-\alpha_g } d\alpha_s \int_{0}^{\alpha_s}d\alpha\right ] \nonumber\\ & & \left . \frac{\phi(1-\alpha-\alpha_g,\alpha_g,\alpha)}{\alpha_g } + m_cf_km_k^2 \int_0^{u_0 } d\alpha_s\int_{u_0-\alpha_s}^{1-\alpha_s } d\alpha_g \frac{\psi(\alpha_u,\alpha_g,\alpha_s ) } { \alpha_g}\right\ } \mid_{\alpha_s \leftrightarrow \alpha_u}\ , . \nonumber \\\end{aligned}\ ] ] this work is supported by national natural science foundation , grant number 10405009 , and key program foundation of ncepu . s. g. matinyan , b. muller , phys . rev . * c58 * ( 1998 ) 2994 ; k. l. haglin , phys . rev . * c61 * ( 2000 ) 031902 ; z. w. lin , c. m. ko , phys . rev . * c62 * ( 2000 ) 034903 ; a. sibirtsev , k. tsushima , a. w. thomas , phys . rev . * c63 * ( 2001 ) 044906 . s. godfrey and n. isgur , phys . rev . * d32 * ( 1985 ) 189 ; s. godfrey and r. kokoshi , phys . rev . * d43 * ( 1991 ) 1679 ; g. s. bali , phys . rev . * d68 * ( 2003 ) 071501 ; a. dougall , r. d. kenway , c. m. maynard and c. mc - neile , phys b569 * ( 2003 ) 41 . n. a. tornqvist , z. phys . * c68 * ( 1995 ) 647 ; m. boglione and m. r. pennington , phys . lett * 79 * ( 1997 ) 1998 ; n. a. tornqvist , hep - ph/0008136 ; n. a. tornqvist and a. d. polosa , nucl . phys . * a692 * ( 2001 ) 259 ; a. deandrea , r. gatto , g. nardulli , a. d. polosa and n. a. tornqvist , phys . lett . * b502 * ( 2001 ) 79 ; f. de fazio and m. r. pennington , phys b521 * ( 2001 ) 15 ; m. boglione and m. r. pennington , phys . rev . * d65 * ( 2002 ) 114010 . e. van beveren , g. rupp , phys . * 91 * ( 2003 ) 012003 ; d. s. hwang , d. w. kim , phys . lett . * b601 * ( 2004 ) 137 ; yu . a. simonov , j. a. tjon , phys . rev . * d70 * ( 2004 ) 114013 ; e. e. kolomeitsev , m. f. m. lutz , phys b582 * ( 2004 ) 39 ; j. hofmann , m. f. m. lutz , nucl . phys . * a733 * ( 2004 ) 142 . i. i. balitsky , v. m. braun and a. v. kolesnichenko , nucl . phys . * b312 * ( 1989 ) 509 ; v. l. chernyak and i. r. zhitnitsky , nucl . phys . * b345 * ( 1990 ) 137 ; v. l. chernyak and a. r. zhitnitsky , phys . * 112 * ( 1984 ) 173 ; v. m. braun and i. e. filyanov , z. phys . * c44 * ( 1989 ) 157 ; v. m. braun and i. e. filyanov , z. phys . * c48 * ( 1990 ) 239 . m. a. shifman , a. i. and vainshtein and v. i. zakharov , nucl . phys . * b147 * ( 1979 ) 385 , 448 ; l. j. reinders , h. rubinstein and s. yazaky , phys * 127 * ( 1985 ) 1 ; s. narison , qcd spectral sum rules , world scientific lecture notes in physics * 26*.

the reaction mechanism of projectile fragmentation at intermediate energies has been investigated observing the target dependence of the production cross sections of very neutron - rich nuclei . measurement of longitudinal momentum distributions of projectile - like fragments within a wide range of fragment mass and its charge was performed using a hundred - mev / nar beam incident on be and ta targets . by measurement of fragment momentum distribution , a parabolic mass dependence of momentum peak shift was observed in the results of both targets , and a phenomenon of light - fragment acceleration was found only in the be - target data . the analysis of production cross sections revealed an obvious enhancement of the target dependence except target size effect when the neutron excess is increased . this result implies the breakdown of factorization ( bof ) of production cross sections for very neutron - rich nuclei near the drip line .

nucl-ex0702050

during these two decades , unstable nuclear physics has become one of the most interesting fields of nuclear physics . nuclear fragmentation of heavy - ion beams is utilized for producing secondary beams of unstable nuclei far from @xmath1-stability . for designing experiments with the secondary beams , a good knowledge of production cross sections is essential . to deduce the production cross sections relevantly , an empirical parametrization of fragmentation cross sections ( epax ) is widely used in simulation programs for projectile - fragment separators @xcite . for consideration of reaction mechanism , fragmentation models based on abrasion - ablation scheme ( aa models ) are often applied to estimate the production cross sections . with the recent development of heavy - ion accelerators , the number of accessible nuclei lying far from @xmath1-stability line has been increasing . an example has been shown by the experimental findings of particle stability of @xmath2f @xcite , @xmath3ne @xcite , @xmath4na @xcite , @xmath5 mg @xcite , @xmath6al @xcite , and @xmath7si @xcite , as the most neutron - rich nuclei so far identified for the @xmath8=9@xmath914 elements . for the new isotope - search experiment , the accuracy to predict the production cross section of neutron - rich nuclei near the drip line is very important to discuss the particle stability of them . however , the predictive power of the epax parametrization and aa models is not strong enough for specific very neutron - rich nuclei @xcite . for instance , although tantalum is often experimentially used as a production target in order to earn the better yield of these nuclei , the target dependence of production cross sections is not taken into account in these models . instead , we first determined the production cross sections for the observed isotopes , which were then used to estimate the production cross sections and the expected yields for the non - observed isotopes . therefore , to enhance predictive powers for production cross sections , deeply understanding reaction mechanism as well as observing systematic behaviors of production cross sections are necessary . the validity of the epax parametrization and aa models has been mainly verified for medium and heavy mass fragments via multi - gev high - energy fragmentation reactions . the epax formula has been also used for the intermediate - energy experiments ( several tens @xmath10 mev ) since the formula can reproduce reasonably well even at intermediate enegies the production cross sections of stable and unstable nuclei near the @xmath1-stability line . a target dependence of fragment - production cross sections in relativistic heavy - ion collisions is limited only by taking account of the nuclear - size effect @xcite . however , the recent experiments show that the production yields of nuclei far from stability line are quite different from the prediction of epax formula , and strongly dependent on the @xmath11/@xmath8 ratio of the target @xcite . it is of great interest whether the cross sections measured with different targets factorize in projectile fragmentation at intermediate energies . in order to confirm the validity of factorization , a careful measurement of fragment momentum distribution is necessary for precise determination of the production cross sections . at relativistic energies , the shape of momentum distribution for an isotope was found to be a gaussian function @xcite and the width was well understood with a statistical model @xcite . on the other hand , the momentum distribution of fragments produced at intermediate energies has an asymmetric shape with a tail at low momentum side . a theoretical attempt was made to reproduce the asymmetric shape , by taking into account nucleon flows between projectile and target in the collision time @xcite . in this model , stochastic nucleon transfers using a monte carlo method and sequential evaporation were taken into account . this model can reproduce the low - momentum - side tails , however , due to a large friction force , the whole of predicted distributions are shifted toward the low momentum side much larger than the observed . this discrepancy of momentum distributions affords a large ambiguity to evaluate the production cross section from the measured yield of fragments . in addition , the measurement for very neutron - rich nuclei has been performed using very thick targets to earn the yields , so far . as distortions of momentum distributions due to target thickness , the momentum distributions of fragments can not be obtained from these data clearly . therefore , the measurement of momentum distribution of very neutron - rich nuclei has become important . in the present work , we focus on the target dependence of momentum distribution of projectile - like fragment ( plf ) produced by nuclear fragmentation reactions at an intermediate energy . to investigate the target dependence of the production cross sections systematically , we used two production targets of be and ta . to avoid distortions of momentum distributions due to the target thickness , we prepared relatively thin targets . we performed the experiment with the riken - rips to eliminate the primary beam and to collect the projectile - like fragments . the data of this experiment were taken in a wide range for fragment mass including very neutron - rich side ( @xmath12 @xmath13 3 ) toward the neutron drip - line and light mass ( @xmath14 @xmath15 3 ) , with a good statistics for momentum - distribution tails . in the following , we first describe the experimental setup and procedure in sec . ii . in sec . iii , the analysis of the data is described . in sec . iv , the experimental results are described , where the observed momentum distributions of fragments , the momentum peak shift , the high- and low - momentum side widths , and the production cross sections are presented in terms of target dependence . based on the results , we discuss the prefragment production mechanism in projectile fragmentation reactions . the projectile fragmentation experiment using a @xmath0ar beam was performed at the riken accelerator research facility . the measurement of momentum distributions of projectile - like fragments was carried out with the projectile fragment separator , rips @xcite . the @xmath0ar@xmath16 beam accelerated by the ring cyclotron with energies up to 90@xmath10 mev and 94@xmath10 mev irradiated a 95-mg/@xmath17 thick @xmath18be target and a 17-mg/@xmath17 thick @xmath19ta target , respectively . the primary beam intensity was monitored for normalization of fragment yields to obtain the momentum distribution . a plastic telescope consisted of three plastic scintillators with a size of 50@xmath2050@xmath200.5 mm@xmath21 was placed at a backward angle of 135 degrees and at a distance of 0.5 m from the target position . the plastic detectors counted yield rates of light particles produced with nuclear reactions at the production target . the three photomultiplier tubes ( pmts ) were mounted on the plastic scintillators one - by - one . the counting rate of triple coincidence was used to monitor the primary beam . the primary beam intensity was measured with the indirect method of plastic counters calibrated by a faraday cup . by changing the primary - beam intensity which ranges from @xmath22 to 1.0 of the full beam intensity , the calibration data were taken . the systematic error of the beam monitor was estimated to be 7% . the rips was used as a doubly achromatic spectrometer . projectile fragments produced at the production target and emitted at 0 were collected and transported to a double achromatic focal plane ( f2 ) . the momentum acceptance was set to be @xmath23=1% at a momentum dispersive focal plane ( f1 ) where left and right slits formed the rectangle window of momentum acceptance . the angular acceptance was set with a square window formed by four slits ( upper , lower , left and right ) , which were placed at the behind of the production target . the @xmath24 and @xmath25 angular acceptances were 25 mrad , which is narrower than the width of angular distribution of fragments in r.m.s . we use the constant value of 6.25@xmath26 msr as the solid angle of @xmath27 . momentum distributions of fragments were measured at 23 settings of magnetic rigidity ( @xmath28 ) over a range of 2.520@xmath94.068 tm using @xmath18be target . in the case of @xmath29ta target , the measurement of momentum distribution for each fragment was performed at 31 magnet settings as same as be case . when the magnetic field was changed for each run , the f2 image of secondary beam was measured by means of a parallel - plate avalanche counter ( ppac ) @xcite to confirm transmission , and the x - position of the beam at f2 was corrected to center precisely by tuning the d2 field . the beam position was monitored with an accuracy of 1 mm . the systematic error of magnetic rigidity was about 3@xmath30 from the ratio of 1 mm to 3.6 m. the difference between d1 and d2 magnetic fields was less than 0.05% . the identification of fragments was carried out event - by - event by means of measurement of time of flight ( tof ) and energy deposit ( @xmath31 ) for each fixed @xmath28 run with the 1% momentum slit . according to an estimation of charge state distribution @xcite , all fragments in flight are fully striped ( @xmath32@xmath33@xmath8 ) . under this assumption , the particle identification can be performed on the basis of the relations : @xmath34 where @xmath10 and @xmath8 are mass number and atomic number , respectively . the detectors of two 0.5-mm thick surface - barrier - type silicon counters ( ssd1 , ssd2 ) and a 0.5-mm thick plastic scintillation counter ( pl ) were installed for the @xmath31 and tof measurement at f2 . to use two silicon detectors allowed us to deduce @xmath8 number of the fragment independently , and to take correlations between them to achieve good @xmath35/@xmath11 ratio . both ssd1 and ssd2 have 48 mm@xmath2048 mm sensitive area which is wide enough to accept all particles reaching to f2 where fwhm of the beam profile is 6 mm . two photomultiplier tubes ( pmts ) were mounted on both sides of pl ( left and right ) . a timing of pl was determined with an average of the signals from the left and right . the tof of each fragment over 21.3-m flight path between the production target and the f2 was determined from the difference of timing signals between rf signal of the cyclotron and the pl timing . the tof resolution was measured with a faint beam to be 0.27 ns ( r.m.s . ) , which included the timing jitter of rf signal ( @xmath360.09 ns ) . thus , the intrinsic resolution for pmt was estimated to be 0.18 ns . in this section we describe the procedures of the data analysis , fitting of momentum distributions , and evaluation of the production cross sections . figure [ fig : detof ] shows a two - dimensional plot versus @xmath31 in ssd1 for one @xmath28 setting using the be target . a rejection of the background events was carried using a correlation gate between ssd1 and ssd2 . by use of the 3-@xmath37 gate by two ssds , we achieved the particle identification of fragments with low background events . fragment yields were obtained by counting the isotopes from the particle identification . figure [ fig : run31-proj ] shows the @xmath8- and @xmath10/@xmath8-projection of the particle identification at @xmath28=2.523 tm setting using the be target . the counting gate of isotopes was a rectangle region with @xmath383-@xmath37 of the resolution @xmath37(r.m.s . ) for @xmath8 and @xmath10/@xmath8 . + in the data acquired using the be target , the analyzed isotopes were @xmath39li , @xmath40be , @xmath41b , @xmath42c , @xmath43n , @xmath44o , @xmath45f , @xmath46ne , @xmath47na , @xmath48 mg , @xmath49al , @xmath50si , @xmath51p , @xmath52s , @xmath53cl , and @xmath54ar . most of these isotopes are neutron - rich ones . it should be noted that the @xmath55al,@xmath5si , and @xmath56p isotopes analyzed have neutron numbers larger than the projectile ( n @xmath15 23 ) , which are produced through neutron pick - up process . we analyzed from the ta - target data for @xmath57li , @xmath58be , @xmath59b , @xmath60c , @xmath61n , @xmath62o , @xmath63f , @xmath64ne , @xmath65na , @xmath66 mg , @xmath67al , @xmath68si , @xmath69p , @xmath70s , @xmath71cl , @xmath72ar , and @xmath73k . the potassium isotopes should be produced by the reaction with proton pick - up process . in order to obtain the doubly - differential cross section from each fragment yield , we estimated the transmission between f0 to f2 , and the reaction loss in the detectors . a monte carlo simulation by mocadi @xcite was performed under the realistic condition of rips using a reference beam of @xmath0ar . the value of transmission was obtained to be 95.3@xmath380.3% . the nuclear reaction loss of fragment in the detectors was evaluated with reaction cross sections calculated by a simple geometrical model . the reaction loss in the detectors was estimated to be less than 0.8% . after all , the systematic error was @xmath389% for evaluation of the fragment cross sections . momentum distributions of fragments have information for understanding reaction mechanisms . at relativistic energies , the projectile fragments have symmetric momentum distributions fitted to a gaussian form , of which width has been discussed with respect to the fermi motion of nucleons or temperatures of pre - fragments @xcite . the results obtained in this work are different from at relativistic energies . figure [ fig : mg30 ] shows a typical momentum distribution of this experiment . by comparison of the momentum distribution fitted with a gaussian function ( dotted curve ) , asymmetric feature of the distribution is clearly observed . the momentum distribution of projectile - like fragments produced at intermediate energies are generally asymmetric with a tail on the low momentum side @xcite . to deduce the most probable momentum and width from such skewed shapes , the momentum distributions have been fitted so far with several kinds of trial functions @xcite . since physical models have not been established for the low momentum tail , what kind of functions to be used is not unique . to study a systematics of low - momentum tail , the following asymmetric function with four free parameters is applied for the present data fitting , @xmath74 where @xmath75 is the most probable peak value of momentum in the distribution , @xmath76 and @xmath77 are momentum width in low and high energy side , respectively . in this fitting procedure , the maximum likelihood method was used to treat the small statistics at the tail parts of distribution . the result of fitting with the asymmetric gaussian function is shown with the solid curve in fig . [ fig : mg30 ] . by means of this method , we obtained fitting results of momentum distribution for all isotopes available in our data , though there is another kind of complexity in a few case . we found two components in the momentum distributions of very light fragments only for the be - target data . figure [ fig : typical ] shows the momentum distributions of the @xmath78be and @xmath79 mg isotopes from the be - target data . the both distributions are scaled as a function of velocity ( @xmath1 ) . two arrows in the figure indicate the velocities of projectile ( @xmath80 ) and center - of - mass system ( @xmath81 ) , respectively . in the distribution of @xmath79 mg ( fig . [ fig : typical](b ) ) , a single component is observed near the projectile velocity ( @xmath80 ) . on the other hand , the @xmath78be distribution ( fig . [ fig : typical](a ) ) shows two components at @xmath80 and @xmath81 . here , the component around @xmath80 is defined as a high - energy side of peak ( he ) , and that of @xmath81 is as a low - energy side of peak ( le ) . we made an attempt to fit the data using the asymmetric gaussian function for the he - component and a gaussian function for the le - component . the fitting results are drawn with solid curves . with increasing the fragment mass number , the le - component decreases very quickly . at last , no significant le - component has been observed for heavy fragments like the @xmath79 mg data . in our data , we have found clearly the le - component for the fragments with @xmath10 of 9@xmath3612 . the le - component has been observed for light fragments in ar+be reaction , while , the le - component has not been found in the momentum distributions of ar+ta reaction data in the momentum region where the experimental data were taken . when a light fragment like @xmath78b is produced in the ar+ta reaction , the impact parameter is much larger than in the case of the ar+be reaction . thus , we found the le - component only for the ar+be system . since we focus on the projectile fragmentation reaction , we discuss mainly the he - component in this paper . the production cross sections of fragments were evaluated with the fitting results of the longitudinal momentum distributions . the transverse momentum distributions are assumed to be a gaussian function with a width of @xmath82 , @xmath83 where @xmath84 is a parameter of the deflection effect , and @xmath85 are mass numbers of a projectile and a fragment , respectively . in ref . @xcite , the deflection parameters for two targets of @xmath86al and @xmath87au with 92.5 and 117.5@xmath10 mev @xmath88o beams . the results from their measurement were @xmath84=190@xmath9220 mev / c , found no large target dependence . taking account of an energy dependence reported in ref . @xcite , we used @xmath84 of 195 mev / c around 90@xmath10 mev for the present experiment . the ambiguity of transverse momentum distributions was taken into account as the systematic error of the cross sections . the overall systematic error for the production cross sections is shown in fig . [ fig : cs - syserror2 ] . the fitting results with the asymmetric gaussian function are presented in the following . first of all , the fitting results of momentum distributions are compared with several formulae taken from reaction models . the result of high momentum side width can be understood by the goldhaber model . however , we need the further discussion of the momentum peak shift and low momentum side width . next , we show the result of production cross sections . the result reveals the phenomena , breakdown of factorization ( bof ) , for production of very neutron - rich nuclei . we discuss the systematics of isotope production cross sections . the charge distribution of the cross sections for a fragment mass is obtained from our data , and compared with the epax formula . finally , we discuss the prefragment production mechanism in projectile fragmentation reactions to search for the origin of bof . in the nuclear fragmentation process , a part of the kinetic energy of projectile is converted into excitation energies of fragments , and the projectile velocity is decreased . this energy loss in projectile fragmentation reaction is called as ` momentum peak shift ' . the momentum peak shift is obtained from the difference of the projectile velocity and the most probable velocities of fragments . we present the result of momentum peak shift in a unit of energy per nucleon , which is proportional to square of velocity . this unit is convenient to discuss the kinetic energy consumption in the nuclear reaction . figure [ fig : mps ] shows the momentum peak shift of fragments produced in ar+be and ar+ta reactions . the solid lines for each isotope are drawn to guide the eye . the primary beam energies were corrected with the mean energy loss in the production target . the obtained projectile energies ( solid lines ) were taken to be 87.5@xmath10 mev and 93.8@xmath10 mev , respectively . the values of the most probable energies of fragments were also corrected with energy losses in the targets . the negative shift of data from the solid line of the projectile energy indicates the energy loss by nuclear fragmentation process . + the systematic error of momentum shift was estimated to be 0.9% in the unit of energy per nucleon . the deviation for a given element chain in the be target is comparable to the systematic error . as seen in fig . [ fig : mps ] , the measured momentum shifts for the be target ( a ) , compared with the case of the ta target ( b ) , vary widely . one could understand the deviation as an effect of the target thickness in atomic energy - loss process @xcite , because the target thickness of be is six times thicker than that of ta . the bounce of momentum peak shift as a function of fragment mass was observed in the results of both targets . the momentum peak shift increases when the number of removed nucleons @xmath89 is increased up to a half of projectile mass ( @xmath9020 ) . on the other hand , the momentum shift decreases when the mass loss @xmath91 is increased beyond 20 ( @xmath9220 ) . in short , we observed the maximum of momentum peak shift around @xmath14=20 . a phenomenon of fragment acceleration was observed in the be - target data . in fig . [ fig : mps](a ) , the solid line of 87.5@xmath10 mev corresponds to the primary beam energy . velocities of the fragments @xmath93li , @xmath58be , and @xmath94b are larger than the projectile velocity . the very light fragments are accelerated in the reaction process . we note that the acceleration phenomena was also observed in collisions of @xmath95u at 1 a gev with lead , reported by enqvist @xcite . on the other hand , no acceleration phenomenon for all fragments @xmath908 was observed in the ta - target data . except the acceleration phenomenon , no significant difference between both targets was observed in the momentum peak shift for the projectile - like fragments @xmath9020 . we found two features in the momentum peak shift . first , the maximum of momentum peak shift is observed for the fragments around @xmath9620 in fig . [ fig : mps ] for both targets . secondly , the acceleration phenomenon was observed in light fragments only for the be target . namely , the most probable velocities of light fragments are beyond the projectile velocity . in the following , we discuss the features observed in the momentum peak shift . the momentum peak shift has been investigated for long time . many of the reports have shown a linear mass dependence of the peak shift for fragments with @xmath97 . several formulae are proposed so far to reproduce the momentum peak shift for the heavy fragments produced in peripheral collisions . we compare our experimental results for the wide fragment mass range with the formulae and make an attempt to introduce a new picture to reproduce the parabolic mass dependence . in fig . [ fig : mps ] , four formulae are shown with dashed curves and the labels as greiner75 , kaufman82 , borrel85 and morrissey89 @xcite . for every formula , the momentum shift becomes large when the number of removed nucleons @xmath98@xmath10 is increased up . this tendency conflicts the present result in the region of @xmath9920 . to compensate for the deviation , j.a.winger et al . @xcite obtained a formula symmetrized mathematically with respect to @xmath100/2 ( dotted line ) . however , it still overestimates the momentum shift around @xmath101 . after all , all of formulae above can not predict the parabolic mass dependence . the previous investigations were mainly performed in the region of fragments close to the projectile mass . for whole fragment mass region observed in this work , a parabolic dependence , where the symmetric point is at the half mass of projectile , is observed . as borrel commented first in ref . @xcite , the symmetric behavior of the velocity shift implies another mechanism less costly , compared with the removal of individual nucleons . the mechanism may be a process that the projectile splits into two pieces . then we make a new formula which calculates the momentum shift based on the splitting picture . first , we assumed that the projectile splits in two fragments which are not exited state . by means of the empirical mass formula of weizscker - bethe @xmath102(@xmath11,@xmath8 ) , where ( @xmath11,@xmath8 ) is the neutron and charge number of nuclei @xcite , the energy loss in the splitting process is described as @xmath103 @xmath104 @xmath105 . the change of binding energy shows the parabolic mass dependence of peak shift . however , the energy loss by the splitting process gives only several tens percent of the value measured as momentum peak shift , quantitatively . next , we deduce a semi - empirical formula in consideration of the excitation energy of fragments . the splitting process consumes the kinetic energy of projectile due to the excitation . here we assume that the energy consumption is proportional to the number of pairs of nucleons destroyed in the reaction , where the nucleons are acting mutually by long range force . in this picture , the number of pairs of nucleons in projectile is counted with @xmath106=@xmath100(@xmath100@xmath9 1)/2 , where the mark of @xmath107 is the combinatorial which gives the number of ways of choosing m out of n. when the projectile splits into the spectator @xmath14 and the participant ( @xmath108 ) , the number of pairs of nucleons which decreased by reaction serves as @xmath14(@xmath109 ) . when one nucleon is removed ( @xmath14=@xmath1101 ) , the number of pairs which decreased is @xmath1101 , and the total energy loss can be defined as @xmath111 mev . the energy loss per a nucleon pair is @xmath111/(@xmath1101 ) mev on average . thus , the kinetic - energy loss in the splitting process can be written as , @xmath112 in the case of non - relativistic beam energy , the energy conservation between projectile and two pieces gives a new formula as follows : @xmath113 if we select the energy loss parameter of @xmath111=8 mev , this formula corresponds to borrel s for the case of one - nucleon removal . however , the value of @xmath111 may not be 8 mev . in fig . [ fig : mps ] , the @xmath75 values predicted by the formula were compared with the present data . the solid curve ( this work ) was drawn with @xmath111=12 mev to fit the data . the experimental data may support that @xmath111=12 mev is better than 8 mev . the acceleration effect can not be explained by a fragmentation model based on the abrasion - ablation picture , in which the projectile always loses the kinetic energy in the laboratory system for abrasion process of nucleons . the acceleration phenomenon may be peculiar to the @xmath18be(@xmath0ar , x ) reaction , namely inverse kinematics . no acceleration effect was found in the ta target data . by comparison between both targets , we found two particular features of momentum distributions for light fragments produced in the ar+be reaction . one feature is the acceleration phenomenon . the other feature is the existence of le - component . both phenomena coincide in our data . therefore , the acceleration phenomenon may be related to the le - component . f. auger et al . @xcite have suggested that the light fragments originate in division of a composite system . in their experiment of @xmath114kr+c and @xmath114kr+al collisions , the energy of the kr beam was 35@xmath10 mev , and relatively lower than our experiment . as a result of their experiment , two components of fission - like fragments were found in the velocity distribution of fragments with each mass number . the two components may be generated from a highly excited compound system since the average speed of them was equal to the center - of - mass ( cm ) speed of the incident system . in short , it is possible that the le - component and the he - component come from one source . as shown in fig . [ fig : accel ] , the velocities of the le - component observed in this work are not below but just on the cm velocity . in addition , the mean energy of he- and le- component is around 80@xmath10 mev . thus , the observed phenomenon is different from the work of f. auger et al . then , we suggest two source model that the le - component and the accelerated he - component come from two sources produced in different reactions , respectively . one source is an excited compound system from projectile and target nuclei running on the cm velocity . the source emits light particles homogeneously in angular space and becomes a fragment in the le - component . on the other hand , the other source is a hot projectile . the hot projectile is formed by a strong impact of target nucleus to convert the kinetic energy of projectile nucleus to the internal energy for the excitation of projectile nucleus . the hot projectile decays to two components again , which consists of a faster component and a slower component than the projectile velocity , like the one - source model . this two - source model predicts a third component in the momentum distribution of light fragments . such third component was not observed in our data . after all , the origin of the acceleration phenomenon has not been understood so far . neither the acceleration effect nor the le - component were found in the ta target data . the different situation from the be target comes from the large impact parameter . for a fragment mass , the impact parameter in the ar+ta reaction is larger than in the ar+be reaction when the fragment is formed by the geometrical cut . the hot compound system is not produced in the peripheral collision with the large impact parameter . therefore , the exotic phenomena may not be observed in the ta target data . + + the high momentum side widths @xmath77 as a function of fragment mass are shown in fig . [ fig : width](a ) and fig . [ fig : width - high - ta](a ) , for the be and ta targets , respectively . the high momentum side widths of all the observed fragments are compared to the formulation by goldhaber as follows , @xmath115 where @xmath85 is the mass number of the projectile and fragment , respectively . the solid curves are drawn with the reduced width @xmath116=90 mev / c from experimental results at relativistic energies @xcite . the deduced values of @xmath116 for the fragments in the mass range 9 to 36 are shown in fig . [ fig : width](b ) and fig . [ fig : width - high - ta](b ) . the dashed lines denote the mean value of @xmath116 = 93.5@xmath38@xmath117@xmath38@xmath118 for the be target , and 97.4@xmath38@xmath119@xmath38@xmath120 mev / c for the ta target , respectively . no significant difference between both targets was observed in the high momentum side widths . the light fragments @xmath14@xmath12113 produced by using the be target show slightly a deviation from the goldhaber model . the deviation is originating from the le - component and the fitting function to be used . to avoid a digression from the main purpose , we do not follow up the deviation further . these results are in good agreement with high - energy experiments @xcite . at relativistic energies , the reduced width @xmath116 is independent of the primary beam energy . at lower energies , the reduction of @xmath116 has been observed @xcite . the reduction mechanism has been argued by several theoretical works , where the reason is for example due to the pauli blocking . due to the effect , the @xmath116 has the energy dependence at 30 - 40@xmath10 mev @xcite , and becomes constant up to 90@xmath10 mev . the fact that the measured @xmath116 is the same as high - energy one is consistent with this picture . the results of momentum width at low momentum side are shown in fig . [ fig : width - low0 ] . the widths of low momentum side @xmath76 are plotted as a function of fragment mass with the statistical errors . we compared the results with the high momentum side width @xmath77 . instead of showing individual data of @xmath77 , the dashed curve calculated with the goldhaber model is presented . the systematic error of @xmath76 is estimated to be 8% , which is not shown in the figure . in the be - target data , each isotope chain has a mountain - style structure ( solid curves ) . on the other hand , the low momentum widths @xmath76 of the ta data may have no such a structure . one can understand the deviation as the target thickness effect . another reason for the mountain - style structure should be noted . the momentum distribution of light fragments has the le - component at the low momentum side only for the @xmath18be target data . in the fitting procedure by use of the asymmetric gaussian - like function , the le - component may also affect the results . therefore , the mountain - style structure of @xmath76 for fragment mass should be ignored as a systematic error . as seen in fig . [ fig : width - low0 ] , the measured widths of @xmath76 are twice wider than @xmath77 approximately . obviously , the large width can not be explained by the goldhaber model . the @xmath76 may have a linear dependence as a function of mass loss @xmath89 , which is very different from the parabolic feature of @xmath77 . no target dependence of @xmath76 between the be and ta targets was found in our data . as any models for the low momentum tail have not been established , it is difficult to discuss the low momentum tail only from the systematics of @xmath76 . yet we should comment that the low momentum tail may not depend on the target . in addition , another interesting point was found in the result . at the limit @xmath91@xmath122@xmath123 , the @xmath76 may not converge on 0 , but 300@xmath36400 mev / c . this feature is different from the goldhaber formula . even if the fragmentation reaction is the dominant process for production of fragments , these new observations may lead to an additional reaction process necessary for intermediate - energy reactions . + in fig . [ fig : width - low0 ] , a fitting line was first obtained from the ta data . the fitting lines are drawn in both figures . the experimental results ( @xmath76 ) and the fitting lines are located far from a prediction of the goldhaber model . what is the origin of the large width of low momentum side ? the large width is indeed produced by an energy - loss process in the nuclear reaction . the energy - loss process for @xmath76 may be different from the `` pure '' fragmentation process , because the width of low momentum side strongly depends on the beam energy . at relativistic energies , the momentum distribution becomes symmetric with @xmath76 = @xmath77 . for a low energy beam at 30@xmath10 mev , the large tail of low momentum side appears obviously @xcite . macroscopic friction process of nuclear dissipation is not the satisfactory mechanism for @xmath76 because the momentum peak @xmath75 , which has no significant energy dependence , may also be changed at the same time . the energy - loss process to produce the width of low momentum side may be explained by nucleon exchange reaction between target and projectile @xcite . for instance , transfer mechanism adds one nucleon to the projectile or fragment , the energy per nucleon @xmath124 may be changed as @xmath125 , where @xmath10 is the mass number before transfer reaction . since the velocity of projectile is reduced by the nucleon exchange reaction , the nucleon transfer may contribute the low momentum tail . when the projectile gives a nucleon to the target nuclei , the velocity of the remained nucleons in the projectile does not change significantly . like the effect of momentum peak shift , the change of potential energy in projectile affects slightly the velocity of projectile nucleus . on the other hand , when the projectile picks up a nucleon from the target nuclei , the projectile velocity should be deduced because nucleons in target nuclei are much slow on average at the laboratory system . thus , we suppose the nucleon exchange reaction as an origin of the low momentum tail . assumed that the probability of nucleon exchange is described as the poisson distribution and the average of number of transfer nucleons in a reaction is quite small , the transfer process does not contribute the peak shift of fragment - momentum distribution but the large width of low momentum tail . if the probability of transfer process is small on average , the peak shift of fragment - momentum distribution , which is related to the case of no nucleon transfer , does not suffer from the transfer process to be independent of the beam energy . on the other hand , the low momentum width is sensitive to the transfer probability . we next discuss the linear dependence of @xmath76 as a function of @xmath98@xmath10 . first , we try to explain this dependence in terms of the surface abraded . if the transfer probability is proportional to area of the surface , the probability may be described as a symmetric function with respect to the half of projectile mass . however , the observed behavior has a linear dependence . so , the abraded surface may not be directly related to the transfer process . secondly , we think about the overlap volume of projectile and target nuclei , i.e. , total number of nucleons in the participant region . the region has almost linear dependence as a function of @xmath14 . thus , the linear dependence of @xmath76 may be related to the volume of the overlap region . the transfer - like fragments of @xmath55al,@xmath5si,@xmath56p , and @xmath73k were observed at this experiment . these fragments have more neutrons or protons than the projectile nucleus . the fragments can not be produced with projectile fragmentation reactions . neutron and proton pick - up processes are necessary for production of the fragments . figure [ fig : tlf1336c4 ] shows the momentum distribution of @xmath55al acquired by using the be target . this isotope production needs at least one neutron pick - up process from target nucleus . we made an attempt to fit the momentum distribution with the asymmetric gaussian function . due to the lack of data at low - momentum tail , we assumed that the fitting parameters of momentum widths were fixed . the @xmath77 was given from the goldhaber formula with @xmath116=90 mev / c . the @xmath76 was assumed to be ( @xmath126 ) mev / c . this value was obtained from the systematics as shown in fig . [ fig : width - low0 ] . the fitting result is shown as a dotted curve in fig . [ fig : tlf1336c4 ] . the peak of the momentum distribution corresponds to 83.9@xmath10 mev . as the primary beam energy was measured as 87.5@xmath10 mev , the momentum peak shift was obtained as ( 3.6@xmath381.2 ) @xmath10 mev . the measured momentum peak shows clearly a larger shift toward low momentum side , compared with that of the same mass number , @xmath360.9@xmath10 mev . the large momentum shift of transfer - like fragments was also observed in the fragments of @xmath73k requiring to proton pick - up process in production . the target dependence of cross sections was investigated with the results of our experiments . we have obtained the cross sections in wide range of fragment charge for each fragment mass with small statistical and systematic errors , and for the same projectile ( @xmath0ar ) with two sets of targets ( @xmath18be and @xmath29ta ) . so we can investigate the validity of factorization for fragmentation reactions at intermediate energies . figure [ fig : csratio ] shows the ratios of cross sections for a fragment in ar+ta reactions to those in ar+be reactions . the cross sections are normalized with the experimental mass yield @xmath127(@xmath14 ) to eliminate the target size effect . the mass yields were obtained from the sum of fragment cross sections with the same mass number . the ratio is shown as a function of charge difference between the most stable charge @xmath128(@xmath14 ) and the fragment charge @xmath129 . thus , the target dependence of projectile fragmentation cross sections except the target size effect is represented for the wide range of isotopes near and far from the stability line . if the factorization is valid for production of an isotope , the ratio has no @xmath130 dependence . in fig . [ fig : csratio ] , the ratios near the @xmath1-stability line are constant . it corresponds that the factorization is valid for the production of the nuclei . according to the previous work @xcite , the factorization is valid for the isotopes in @xmath130@xmath1312 . on the other hand , the ratios increases when @xmath130 is increased . this deviation shows that the factorization hypothesis is clearly broken down for neutron - rich nuclei with ( @xmath130)@xmath152 . the epax was developed by using data of the spallation reactions and heavy - ion induced fragmentation reactions at several @xmath10 gev @xcite . the cross section of a fragment with mass @xmath10 and proton number @xmath8 produced by projectile fragmentation from a beam ( @xmath100 , @xmath132 ) impinging on a target ( @xmath133 , @xmath134 ) can be written as , @xmath135 where @xmath127(@xmath10 ) represents the mass yield which is the sum of the isobaric cross sections for fragments with mass number @xmath10 , and @xmath136(@xmath10,@xmath8 ) describes the charge distribution which means the cross - section distribution of a given fragment mass which has a maximum peak at @xmath137 . the charge dispersion @xmath138 is described as @xmath139 , @xmath137 and @xmath140 parameter . the most probable charge , @xmath137 , is written as , @xmath141 where @xmath128 is the @xmath1-stable charge for a fragment of mass number , @xmath10 @xcite , @xmath142 is a proton - excess between the stability line and the most probable line of fragmentation reaction , and @xmath143 is the so - called `` memory effect '' , i.e. , the influence of the projectile @xmath11/@xmath8 ratio on the fragment @xmath11/@xmath8 ratio . the @xmath139 is a function of fragment mass number , @xmath10 , which shows a fragment mass dependence of a steepness controlling cross sections of isotopes from the stability line to the drip lines . the mass dependence has been confirmed from many combinations of projectiles and targets . the @xmath140 parameter is given as a constant for all neutron - rich isotopes so far . using new data obtained mainly at the gsi / frs facility , the epax was recently modified @xcite slightly to tune the mass yield and the u parameter for proton - rich fragments in the vicinity of projectiles . the present paper focus on the neutron - rich fragments in a wide range . thus , this work gives the analysis and discussion using the epax formula based on the original epax that has relatively simple functions . the epax formula is valid for the `` limiting fragmentation '' regime , where the fragmentation process is no longer energy dependent . the energy dependence of fragmentation cross sections has been investigated by silberberg and tsao @xcite . we can see the similar energy dependence in the total reaction cross section . the total reaction cross section has been studied both theoretically and experimentally for more than 50 years . a detailed list of reference is found in ref . compare the cross section of @xmath144c+@xmath144c at 90@xmath10 mev with 900@xmath10 mev , the difference of cross sections is about 30% @xcite . this value is nearly equal to the systematic error of measured cross sections in this work . thus , we can assume that the limiting fragmentation hypothesis is valid in this work . the epax formula follows the factorization hypothesis . the mass yield has a target dependence , however it is limited to the target nuclear - size effect . the charge dispersion @xmath138 is independent of the target nucleus . in this work , however , the bof has been found in the production of very neutron - rich nuclei . an investigation of the charge distribution is necessary to find appropriate description for cross sections from our data . in the charge distribution of epax , the @xmath140 parameter is a constant of 1.65 for neutron - rich side , and has no target dependence . the value of @xmath140 parameter is very sensitive to the production cross sections of the isotopes far from the @xmath1-stability line . in the following , we reduce the @xmath140 parameter for each fragment mass with both targets from our data . at the same time , we investigate the target dependence of @xmath137 since @xmath137 is also sensitive to the cross sections of isotopes far from the @xmath1-stability line . a fitting of the production cross sections was performed using the epax function . the function is represented as the product of @xmath145 and @xmath138 . the charge distribution @xmath138 is characterized with the most probable charge @xmath137 , the slope constant @xmath140 and the width parameter @xmath139 . the @xmath139 parameter has slightly fragment - mass dependence and an effect to the slope of charge distribution . we have two kinds of @xmath140 parameters of @xmath146 for proton - rich side and @xmath147 for neutron - rich side . in the fitting procedure , the values of the @xmath146 and the @xmath139 parameters are fixed , originally given by the epax formula . the fitting procedure was performed with the experimental data of each fragment mass . first , we obtain the maxima of the charge distribution @xmath137(exp ) as a function of fragment mass number from the experimental data with both targets . the values are compared with the epax ones , the @xmath1-stability line , and @xmath11/@xmath8 ratio of projectile nuclei . next , the @xmath140 parameters are deduced from the data when we assume the same parametrization of @xmath137 for both of the targets . we have investigated the target dependence of @xmath137 . the measured cross sections for each fragment mass were fitted with @xmath148 . the mass yield @xmath127 is a constant for a fragment mass . fitting parameters were @xmath140 parameter in neutron - rich side , @xmath137 , and the mass yield . figure [ fig : csfit - zprob ] shows the charge distribution of fragment mass 29 , produced in ar+ta reaction . the @xmath137 is obtained as 13.47@xmath380.01(@xmath149)@xmath380.04(@xmath150 ) . similarly , the @xmath137 value for each fragment mass is deduced from our experimental data . the experimental @xmath137 is compared with the value from the epax formula . figure [ fig : chargedist2 ] shows the deviation of @xmath137(exp ) from @xmath137(epax ) for ar+be and ar+ta data . solid lines are drawn to guide the eye . the epax reproduces the @xmath137 very well , especially for the mass number between 20 to 35 . the deviation of @xmath137 is less than 0.2 . due to the lack of proton - rich side data , we obtained only three values of @xmath137 for the ar+be data in @xmath14=18 , 22 and 26 . to compare both data of be and ta targets shows no significant target dependence of @xmath137 . figure [ fig : chargedist ] shows the most probable charge in @xmath11/@xmath8 unit as a function of fragment mass for the be and ta targets . the most probable charge @xmath137 from the epax is close to the @xmath1-stable charge @xmath128 . the ratio @xmath151 of the projectile is 1.22 . for the composite system of projectile and target , the ratio is represented as @xmath152 , of which values are 1.23 and 1.43 for ar+be and ar+ta , respectively . the @xmath137(exp ) is well described as the @xmath137(epax ) as well as @xmath128 . the difference of @xmath137(epax ) and @xmath128 is mainly the memory effect @xmath143 . the memory effect is not clearly seen in our data . no significant difference of @xmath137(exp ) between the be and ta targets has been observed , however the bof has been found in the very neutron - rich nuclei . what does it mean ? it should be noted that the bof had never been found in the projectile fragmentation at other experiments . this fact is consistent with our observation of @xmath137(exp ) for the be and ta targets . to understand the charge distribution , we show the @xmath137 of projectile - like fragments produced by the reactions at low and intermediate energies . at low energies , the most probable charge shows the existence of two different reaction mechanisms . in ref . @xcite , cl isotopes were produced in 7@xmath10-mev @xmath0ar+@xmath153ni reaction and the contour plots of the cl isotope yield were drawn to the fragment mass and the kinetic energy . we clearly see two components , one corresponding to quasi elastic reactions centered at a high energy and a mass of 39 . the second one is centered around a mass of 36 and a small kinetic energy . the most probable charge obtained from the mass number is near the @xmath154 of both projectile and composite system which corresponds to the quasi elastic and deep inelastic reactions , respectively @xcite . an intermediate composite system has been shown as the result of a complete damping of the relative motion between the projectile and target nuclei . the projectile - like fragments are produced via binary nuclear system that collective effects dominate . on the other hand , at high energies , the process is dominated by individual nucleonic collisions that described as participant - spectator models . the values of most probable charge distinguish the reaction mechanisms to produce the projectile - like fragments . at intermediate energies , d.guerreau et al . @xcite reported the observation of the systematic shift of isotope distributions between two targets . the isotope distributions of fragment yield were measured in @xmath0ar+ni and @xmath0ar+au reactions at 44@xmath10 mev . the systematic shift of the isotope distributions for a given element of si in @xmath0ar+au reaction was observed about 0.3 mass unit towards the neutron rich side . however , their result is different from our data . the fitting results of our charge distributions have shown no significant difference of @xmath137(exp ) between the be and ta targets . we have already found that the factorization assumption is invalid for production of neutron - rich nuclei . the flexibility of the charge distribution except the most probable charge is the slope parameter . thus , we would seek for the origin of target dependence in the slope constant @xmath140 . due to the bof , the @xmath140 may change from a simple constant to a complex parameter which depends on target nuclei . as the target dependence of @xmath137 has not been found , we assume now that the @xmath137 can be described as the epax formula for both of the targets . we make an attempt to fit the data with the function of charge distribution . fitting parameters were @xmath140 in neutron - rich side and mass yield @xmath127 , and fragment mass dependence of @xmath140 is tried to be obtained . the fitting procedure is performed for the data of each fragment - mass group , where every 4 mass units are combined to avoid poor statistics . figure [ fig : uparam ] shows the @xmath140 parameters as a function of fragment mass for both targets . we plot the values with the systematic errors . the slope parameter @xmath140 of epax is a constant of 1.65 ( dashed lines ) . the average of @xmath140 is 1.62 for the ta target . the smaller @xmath140 gives larger values of production cross sections for neutron - rich nuclei . on the other hand , the @xmath140 parameter of be target is larger than 1.65 and shows fragment mass dependence . our data demonstrate the bof for very neutron - rich nuclei as the target dependence of @xmath140 parameter . we have obtained the modified epax formula for the nuclear fragmentation at an intermediate energy for both be and ta targets . to confirm the validity of our parametrization , we show an example of the production cross section predicted for extremely neutron - rich nuclei . figure [ fig : o24 ] shows the predictive power of the new parameterization . the dashed and solid curves are the charge distribution of mass-24 isobars produced in @xmath0ar+@xmath18be and @xmath0ar+@xmath29ta reactions , respectively . the solid box is the cross section of fragmentation channel for @xmath29ta(@xmath0ar,@xmath155o ) . this value acquired in another experiment @xcite is in agreement with our new parametrization clearly . we have investigated the nuclear fragmentation reaction in nucleus - nucleus collisions at intermediate energies . the momentum distributions of projectile - like fragments have been measured for various isotopes including very neutron - rich nuclei . the charge distributions of fragment cross sections acquired with the be and ta targets have revealed the bof for very neutron - rich nuclei . in order to explain the origin of bof , we discuss the mechanism of prefragment production in terms of the aa model . in the aa model , the production cross section of a final fragment is determined by the charge distribution and the excitation energy of prefragments . first , we discuss the charge distribution of prefragments on the basis of aa model . the abrasion model is a macroscopic description which gives a picture of a clean cut of the projectile nucleus by the target nucleus @xcite . the concept of geometrical separation in an overlap or participant zone does not specify its proton - to - neutron ratio . a few different methods to calculate the charge distribution of the prefragments are so far proposed , a correlation model @xcite , hypergeometrical model @xcite , and gdr model @xcite . after all , the charge distribution of prefragments from these methods depends on the projectile nuclei but has no target dependence . next , we discuss the excitation energy of prefragments . in an early model @xcite , the average excitation energy of a prefragment after abrasion was estimated from the additional surface energy due to the excess surface area . several studies have been made on the source of the excitation energy @xcite . according to the researches , the surface energy is not the main source of the excitation energy , but , one of the energy sources for prefragment excitation . at least , the target dependence of the excitation energy can be seen in the surface energy . figure [ fig : over2 ] shows a schematic view of prefragment production . the prefragments of the same mass number @xmath14 are produced by the projectile fragmentation reaction with target nuclei of different nuclear radius . if a target nucleus has a large size , the impact parameter is also large to produce the prefragment ( case a ) . on the other hand , the small size of a target nucleus caused the violent collision to form the prefragment with the same mass number ( case b ) . in the geometrical model , we calculated the surface energy of prefragments in ar+be and ar+ta reactions . as the result , large difference of @xmath156 is presented for both of the targets . as for the prefragments of @xmath157 , the values of @xmath156 are approximately 90 mev and 10 mev for the be and ta targets , respectively . the target dependence of prefragment excitation energies stems from the additional surface energies in the aa model . we observed no significant target dependence of momentum peak shift in the data for the projectile - like fragments of @xmath9020 . assumed that nucleons are emitted homogeneously from the hot prefragment in the ablation process , the velocity of a prefragment does not change on average when it becomes the final fragment . the results of momentum peak shift show that the energy losses in the fragmentation process for the fragment of @xmath158 were 60@xmath3680 mev for both of the targets . the energy losses are not correlated to the additional surface energies on the calculation . thus , we can not deny the possibility that the bof originates from the target dependence of the charge distributions for prefragments except the excitation energies . to apply this idea to the bof problem , we suggest two mechanisms of a fluctuation giving for the charge distributions of prefragments . one is the nucleon exchange process during the fragmentation process , and another is the effect from the difference of nucleon - nucleon total cross sections ( @xmath159,@xmath160 and @xmath161 ) . we observed the nucleon exchange process during the fragmentation process in our data . the fitting results of @xmath75 and @xmath77 carry the information at the abrasion stage and supporting that `` pure '' projectile fragmentation is a dominant process at intermediate energies . as for @xmath76 , the larger values of @xmath76 than @xmath77 may be related to the energy loss due to the transfer process . unlike the behavior of @xmath77 , @xmath76 has a linear dependence as a function of fragment mass . the linear dependence may be connected from both projectile and target nuclei . so , the source of transferred nucleons can be assumed to be the participant region . since the ta target nucleus is more neutron - rich than the be , the probability of neutron transfer should be large . the nucleon exchange reaction between projectile and target nuclei during the abrasion process may break the factorization of fragment production cross sections . the difference of the total cross sections between proton - to - proton ( @xmath159 ) and proton - to - neutron ( @xmath161 ) may bring the fluctuation of prefragment charge distributions . as a fact known well , the @xmath162 is larger than the @xmath160 and @xmath159 when the nucleon energy is less than 500 mev . in projectile fragmentation process , a neutron - rich target nucleus easily knocks out protons in the projectile nucleus , so that the production of neutron - rich prefragments is promoted . the discussions so far are qualitative . however , they gives several predictions . first , if the nucleon exchange process is just the reason of bof , the probability of nucleon exchange reaction becomes large at low energies . the nuclear fragmentation experiments with lower incident beam energies than at our experiment may reveal the large bof effect for very neutron - rich nuclei . secondly , if the difference of n - n reaction cross sections causes the bof , the nuclear fragmentation experiment with 100 - 800@xmath10 mev beams should be carried out to confirm the bof for very neutron - rich nuclei . since the cross section curves of @xmath159 and @xmath161 cross at 500@xmath10 mev , the difference of u - parameter between the be and ta targets might change to be reversed . the further investigations of fragmentation cross sections at around 500@xmath10 mev are of great interest . the projectile fragmentation reactions at intermediate energies have been investigated using a 90 - 94@xmath10 mev @xmath0ar beam at riken - rips . we paid our attention to the target dependence . measurement of longitudinal momentum distributions of projectile - like fragments within a wide range for fragment mass and its charge including very neutron - rich nuclei has been performed with @xmath18be and @xmath29ta targets . from the momentum distribution of fragments , a parabolic mass dependence of momentum peak shift was observed in the data of both targets , and a phenomenon of fragment acceleration was observed only in the be - target data . a linear dependence of the low momentum tail as a function of removed nucleons was found for both targets . as a possible origin of the low momentum tail , we suppose the nucleon exchange reaction . we observed large target dependence of the cross sections to produce very neutron - rich nuclei with ( @xmath130)@xmath152 . the deviation shows that the production cross sections of very neutron - rich nuclei far from the stability line do not factorize . the reaction mechanism was discussed , however the discussions so far are qualitative . it is much interesting to investigate the target dependence of the cross sections at higher energies . + the authors wish to thank the riken ring cyclotron staff and crew for their cooperation . we are also indebted to dr . artukh , dr . g.a . souliotis , and prof . t. shimoda for stimulating discussions . one of the authors ( m.n . ) acknowledges the finantial support of the riken special researchers basic science program . + j.a . winger et al . : instr . meth . b70 , 380 ( 1992 ) . n. iwasa et al . , nucl . instr . meth . b126 , 284 ( 1997 ) ; th.schwab , gsi report no . 91 - 10 ( 1991 ) . k. smmerer and b. blank : phys . rev . c * 61 * , 034607 ( 2000 ) . h. sakurai et al . , phys . lett . b * 448 * , 180 ( 1999 ) . h. sakurai et al . , phys . * c54 * , r2802 ( 1996 ) . m. notani et al . , phys . lett . b * 542 * , 49 ( 2002 ) . h. sakurai et al . , nucl . * a616 * , 311 ( 1997 ) . gaimard and k.h.schmidt : nucl . phys . a * 531 * , 709 ( 1991 ) . cummings et al . c * 42 * , 2530 ( 1990 ) . d. guerreau et al . c * 41 * , 937 ( 1990 ) . heckman et al . : * 28 * , 926 ( 1972 ) . goldhaber : phys . lett . * 53b * , 244 ( 1974 ) . l. tassan - got et al . phys . a * 524 * , 121 ( 1991 ) . t. kubo et al . instrum . methods phys . b * 70 * ( 1992 ) 309 . h. kumagai et al . : . methods phys . res . a * 470 * 562 ( 2001 ) . b. robinet et al . , nucl . instr . 190 , 197 ( 1981 ) . r. pfaff et al . c * 51 * , 1348 ( 1995 ) . d. bazin et al . a * 515 * , 349 ( 1990 ) . o. bacri et al . phys . a * 555 * , 477 ( 1993 ) . k. van bibber et al . : lett . * 43 * , 840(1979 ) . r. dayras et al . a * 460 * , 299 ( 1986 ) . dufour et al . : instrum . methods phys . res . a * 248 * ( 1986 ) 267 . t. enqvist et al . phys . a * 658 * , 47 ( 1999 ) . d.e . greiner et al . : lett . * 152 * , 35 ( 1975 ) . cumming et al . lett . * 44 * , 17 ( 1980 ) . cumming et al . c * 24 * , 2162 ( 1981 ) . kaufman et al . : c * 26 * , 2694 ( 1982 ) . morrissey et al . : phys . rev . c * 39 * , 460 ( 1989 ) . v. borrel et al . : z.phys . a * 314 * , 191 ( 1983 ) . v. borrel et al . : z.phys . a * 324 * , 205 ( 1986 ) . b. mottelson : `` nuclear structure '' in cours de lecole det de physique thorique des houches ( 1958 ) . f. auger et al . : phys . rev . c * 35 * , 190 ( 1987 ) . d. guerreau et al . phys . a * 447 * , 37 ( 1985 ) . mermaz et al . : z.phys . a * 324 * , 217 ( 1986 ) . a.g . artukh et al . : instrum . methods phys . res . a * 426 * ( 1999 ) 605 . c. grgoire and b. tamain : ann . fr . * 11 * , 323 ( 1986 ) . d. j. olson et al . rev . c * 28 * , 1602 ( 1983 ) . k. smmerer et al . c * 51 * , 1348 ( 1990 ) . p. marmier and e. sheldon : _ physics of nuclei and particles _ ( academic , new york and london ) , vol.i , 15 ( 1971 ) . r. silberberg and c.h . tsao : astrophys . j. suppl . * 25 * , 315 ( 1973 ) . r. silberberg et al . : astrophys . j. suppl . * 58 * , 873 ( 1985 ) . s. kox et al . : rev . c * 35 * , 1678 ( 1987 ) . m. buenerd et al . phys . a * 424 * , 313 ( 1984 ) . j. galin : j. phys . c * 5 , 37 * , 83 ( 1976 ) . m. lefort and ch . ng : ann . phys . * 3 * , 5 ( 1978 ) . d. guerreau et al . 131b * , 293 ( 1983 ) . y. yanagisawa et al . : lett . b * 566 * , 84 ( 2003 ) . j. gosset et al . rev . c * 16 * , 629 ( 1977 ) westfall et al . : lett . * 43 * , 1859 ( 1979 ) . friedman : phys . c * 27 * , 569 ( 1983 ) . morrissey et al . : phys . rev . c * 18 * , 1267 ( 1978 ) . j.d . bowman et al . : lbl report * lbl*-2908 ( unpublished ) ( 1973 ) . wilson et al . : instrum . methods phys . b * 18 * 225 ( 1987 ) . j. hfner et al . : phys . rep . * 125 * , 129 ( 1985 ) .

we present new results on instabilities in rapidly and differentially rotating neutron stars . we model the stars in full general relativity and describe the stellar matter adopting a cold realistic equation of state based on the unified sly prescription . we provide evidence that rapidly and differentially rotating stars that are below the expected threshold for the dynamical bar - mode instability , , do nevertheless develop a shear instability on a dynamical timescale and for a wide range of values of . this class of instability , which has so far been found only for small values of and with very small growth rates , is therefore more generic than previously found and potentially more effective in producing strong sources of gravitational waves . overall , our findings support the phenomenological predictions made by watts , andersson and jones on the nature of the low- instability as the manifestation of a shear instability in a region where the latter is possible only for small values of the . furthermore , our results provide additional insight on shear instabilities and on the necessary conditions for their development .

1001.5281

non - axisymmetric deformations of rapidly rotating bodies are rather generic phenomena in nature and can appear in a wide class of systems . particularly interesting within an astrophysical context are those deformations taking place in fluids that are self - gravitating and the literature on this has a long history dating back to the work of @xcite on incompressible newtonian uniformly rotating bodies . since then , the study of these instabilities has continued over the years both in newtonian gravity and in full general relativity . special attention has traditionally been paid to the study of @xmath3 instabilities , which are characterized by the exponential growth of @xmath3 deformations , where @xmath4 parametrizes the azimuthal dependence @xmath5 in a standard mode decomposition in spherical harmonics . most of the interest in this type of deformation in compact stars stems from the fact that it has the shortest growth time and leads to the emission of a strong gravitational - wave signal . the development of non - axisymmetric instabilities is commonly analyzed in terms of the quantity @xmath6 ( _ i.e. _ the ratio between the kinetic rotational energy @xmath7 and the gravitational potential energy @xmath8 ) , that provides a dimensionless measure of the amount of angular momentum that can be tapped to feed the development of the instabilities . this parameter plays an important role in what is possibly the most celebrated of the non - axisymmetric instabilities : the so - called _ dynamical _ bar - mode instability . this is an @xmath3 instability which takes place when the parameter @xmath1 is larger than a critical one , @xmath9 . in the case of a newtonian incompressible self - gravitating polytrope , for instance , the dynamical bar - mode instability develops for @xmath10 @xcite and is only weakly dependent on the considered polytropic index or whether the fluid is compressible . post - newtonian ( pn ) studies @xcite or fully general - relativistic ones @xcite correct this results only slightly , by reducing the threshold to somewhat lower values of the instability parameter . as an example , for a polytropic relativistic star with polytropic index @xmath11 , the accurate calculations reported in @xcite reveal that the critical value is @xmath12 and that a simple dependence on the stellar compactness allows one to track this threshold from the newtonian limit over to the fully relativistic one @xcite . the onset and development of the bar - mode instability has been traditionally studied by means of nonlinear 3d simulations of newtonian stars that are either unmagnetized @xcite and , more recently , also magnetized @xcite . in addition , pn and fully relativistic simulations have been performed and highlighted , for instance , that the persistence of the bar is strongly dependent on the degree of overcriticality and is generically of the order of the dynamical timescale . furthermore , generic nonlinear mode - coupling effects between the @xmath13 and the @xmath3 mode appear during the development of the instability and these can severely limit the persistence of the bar deformation and eventually suppress the bar deformation @xcite . these results have been recently confirmed by the perturbative calculations in @xcite . besides dynamical instabilities , which are purely hydrodynamical , _ secular _ instabilities are also possible in rotating compact stars and these are instead triggered by dissipative processes , such as viscosity or radiation emission . if , in particular , the dissipative mechanism is the emission of gravitational radiation , then the secular instability is also known as chandrasekhar - friedman - schutz or cfs instability @xcite . contrary to what their name may suggest , secular instabilities do not necessarily develop on secular timescales ( although they normally do ) and are characterized by having a much smaller threshold for the instability . once again , in the case of a newtonian polytrope , the critical secular instability parameter is as small as @xmath14 and thus much more easy to attain in astrophysical circumstances . although widely observed in numerical simulations , the physical conditions leading to a dynamical bar - mode instability are difficult to be encountered in standard astrophysical scenarios . such large values of the instability parameter , in fact , can not be easily attained in old and cold neutron stars , which have been brought into uniform rotation and thus to rather small values of @xmath1 . however , more recently these pessimistic prospects have been changed when a new @xmath3 instability has been discovered in differentially rotating newtonian stars @xcite for values of @xmath15 , therefore well below the expected values for a dynamical bar - mode instability . the most salient aspect of this new instability is that it appears in stars with a large degree of differential rotation and that it grows on a timescale which is longer but comparable with the dynamical one . this instability has been referred to as the `` low-@xmath2 instability '' and its dependence on the polytropic index and on the degree of differential rotation has been studied in @xcite . since then , the instability has been observed or discussed in a number of related studies @xcite , all of which have highlighted the possible occurrence of this type of instability during the collapse of a massive stellar core . despite the abundant numerical evidence on the development of this instability , the nature of these low-@xmath2 instabilities is still matter of debate and , most importantly , a sufficient criterion for its onset has not been derived yet . this instability has been studied in great detail by watts and collaborators @xcite , who have made a number of phenomenological predictions either using a toy shell - model first introduced in @xcite , or for a stellar model in newtonian gravity . overall , the work of watts and collaborators ( but see also @xcite ) recognizes the low-@xmath2 instabilities as the manifestation of a more generic class of instabilities , the _ shear instabilities _ @xcite , that is unstable oscillations that do not exist in uniformly rotating systems and are associated to the existence of a corotation band @xcite . watts , andersson and jones suggest , in particular , that a necessary condition for the development of the instability is `` corotation '' , that is the presence of a point at which the star rotates at the same pattern speed of the unstable mode @xcite . an alternative suggestion on the necessary conditions has been made also by ou and tohline @xcite , who instead associate the development of the instability to the presence of a minimum in the vortensity profile of the star . this minimum can then drive unstable not only the corotating @xmath3-modes but also the odd modes such as the @xmath13 and @xmath16-modes @xcite . in this interpretation , the growth time of the instability is proportional to distance between the corotation radius , _ i.e. _ the radial position at which the unstable mode corotates with the star , and the minimum of the vortensity . the purpose of this work is to shed some light on the development of shear instabilities and , in particular , to validate one prediction made , although not explicitly , by watts , andersson and jones . more specifically , we show that , for sufficient amounts of differential rotation , shear instabilities develop for _ any value _ of the instability parameter @xmath1 and also below the expected critical value for the dynamical bar - mode instability . we therefore provide evidence that the low-@xmath2 instability is not a new instability but rather the manifestation of a shear instability in a region where the latter is possible only for small values of @xmath1 . our analysis proceeds via the simulation in full general relativity of sequences of neutron star models having constant rest - mass and constant degrees of differential rotation , but with different amounts of rotation , _ i.e. _ with different values of @xmath1 . the neutron - star matter is described by a realistic equation of state ( eos ) defined by the unified sly prescription @xcite and we study the development of the non - axisymmetric instabilities from their linear growth up to the fully nonlinear development and suppression . interestingly , we find that depending on the degree of differential rotation , the shear instability leads either to the growth of a single modes ( for the low-@xmath1 models ) or to the simultaneous presence of up to three unstable modes ( for the high-@xmath1 models ) , which produce beatings in the growth of the overall @xmath3 deformation . special attention is also paid to the properties of the unstable modes and to their position within the corotation band or the vortensity profiles . in this way we are able to confirm both the necessary conditions proposed so far for the onset of the instability . in particular , we show that all the unstable modes are within the corotation band of the progenitor axisymmetric model ( _ cf . _ @xcite ) and that all of the unstable models have vortensity profiles with a local minimum ( _ cf . _ @xcite ) . the structure of the paper is as follows : in section [ sec : numerics ] we describe the numerical setting of our simulations , the eos we used and the initial models we generated . in section [ sec : methodology ] we describe the quantities and tools we used to monitor the evolution of the instability . in section [ sec : results ] we report the results of the simulations and section [ sec : conclusions ] is dedicated to conclusions and discussion . we use a spacelike signature @xmath17 and a system of units in which @xmath18 ( or in cgs units whenever more convenient ) . greek indices are taken to run from @xmath19 to @xmath20 , latin indices from @xmath21 to @xmath20 and we adopt the standard convention for the summation over repeated indices . in what follows we provide a brief overview of the numerical setup used in the simulations , of the realistic eos adopted and on the procedure followed for the construction of the initial axisymmetric models . we solve numerically the full set of einstein equations @xmath22 where @xmath23 and @xmath24 are the einstein tensor and the stress - energy tensor , respectively . the equations are solved within the `` 3 + 1 '' decomposition of spacetime , in which the @xmath25-dimensional metric @xmath26 is decomposed into the spatial metric @xmath27 , the lapse function @xmath28 and the shift vector components @xmath29 . the field equations , which then also provide an evolution for the extrinsic curvature tensor @xmath30 , are then coupled to those of general relativistic hydrodynamics @xmath31 where , in the case of a perfect - fluid , the stress - energy tensor is given by @xmath32 above @xmath33 is the fluid @xmath25-velocity , @xmath34 is the fluid pressure , @xmath35 the specific internal energy and @xmath36 the rest - mass density , so that @xmath37 is the energy density in the rest frame of the fluid . the set of hydrodynamics equation is then closed by a prescription for the properties of the matter in the form of a relation between the pressure and other quantities in the fluid , _ e.g. _ @xmath38 , and for which we have chosen a cold and realistic eos which will be discussed in the following section . the evolution of the spacetime was performed using the ` ccatie ` code , a three - dimensional finite - differencing code providing a solution of a conformal traceless formulation of the einstein equations ( see @xcite for the explicit expressions of the equations solved in the code and also @xcite for a more recent and improved implementation ) . the relativistic hydrodynamics equations , on the other hand , were solved using the ` whisky ` code , which adopts a flux - conservative formulation of the equations as presented in @xcite and high - resolution shock - capturing schemes or hrsc ( see @xcite for the explicit expressions of the equations solved in the code and also @xcite for a more recent extension of the code to mhd ) . the ` whisky ` code implements several reconstruction methods , such as total - variation - diminishing ( tvd ) methods , essentially - non - oscillatory ( eno ) methods @xcite and the piecewise parabolic method ( ppm ) @xcite . also , a variety of approximate riemann solvers can be used , starting from the harten - lax - van leer - einfeldt ( hlle ) solver @xcite , over to the roe solver @xcite and the marquina flux formula @xcite ( see @xcite for a more detailed discussion ) . all the results reported hereafter have been computed using the marquina flux formula and a ppm reconstruction . both the einstein and the hydrodynamics equations are solved using the vertex - centered adaptive mesh - refinement ( amr ) approach provided by the ` carpet ` driver @xcite . our rather basic form of amr consists of box - in - box structures centered on the origin of the coordinate system and with the finest grid covering the whole star at all times . the simulations reported here make use of @xmath25 levels of refinement , with the finest having a resolution of @xmath39 and the coarsest one a resolution of @xmath40 . the outer boundary was set relatively close to the star and at a distance of @xmath41 , _ i.e. _ at about @xmath42 times the size of the star . a reflection symmetry across the @xmath43 ( equatorial symmetry ) plane was used to reduce the computational costs , but not a rotational one around the @xmath44-axis ( @xmath45-symmetry ) as it would have artificially prevented the growth of odd-@xmath4 modes ( see discussion in @xcite ) . as mentioned above , the system of hydrodynamics equations needs to be closed by an eos relating the pressure with the other primitive variables , _ e.g. _ the rest - mass density . previous studies of the bar - mode instability , both in newtonian gravity and in general relativity , have been focused on the use of ideal fluids and analytic eoss , either in the form a of a `` polytropic '' ( and isentropic ) eos @xmath46 , or of an `` ideal - fluid '' and ( non - isentropic ) eos @xmath47 ( _ cf . _ discussion in @xcite ) . while these two descriptions are expected to provide results that are qualitatively correct , a more accurate modelling of these instabilities in compact stars necessarily requires a more physically - motivated description of the neutron - star matter . it is in this spirit that we have here considered a realistic eos , namely the unified sly eos @xcite , which models high - density and cold ( _ i.e. _ zero temperature ) matter via a skyrme effective potential for the nucleon - nucleon interactions . the sly eos , which describes via a single effective hamiltonian the neutron star s interior , is supplemented with the hp94 eos @xcite to describe the crustal matter and with the bps eos @xcite for lower density regions . this prescription results in a one - parameter eos in the form @xmath48 , where the sly eos is used for @xmath49 , the hp94 eos is used for @xmath50 and the bps eos for @xmath51 ( see also figure 1 of @xcite ) . in addition , at even lower densities the eos becomes temperature dependent ( and thus no longer a simple barotropic eos ) , but because these these regions are well below the threshold for the artificial atmosphere , we do not consider an additional prescription for @xmath52 . we recall , in fact , that our hrsc methods require the use of a tenuous atmosphere which fills the regions of the computational domain not occupied by the compact star . the threshold value for the rest - mass density of the atmosphere is chosen to be several orders of magnitude smaller than the maximum value and in our simulation a fluid element is considered to be part of the atmosphere if its rest - mass density @xmath36 satisfies @xmath53 . when this happens the fluid element is treated as a non - dynamical cold fluid described by a polytropic eos , @xmath54 , with @xmath11 and @xmath55 and its velocity is set to zero ( see @xcite for a more detailed discussion on the use of the atmosphere in the ` whisky ` code ) . the practical implementation of the realistic eos can take place in a number of different ways . the simplest is to use standard interpolation techniques , _ e.g. _ based on lagrangian polynomials , on the values of the published tables . while straightforward , the interpolations in this approach do not guarantee in general that the thermodynamics relations are fulfilled ( see @xcite for a thermodynamical preserving interpolation ) . in addition , the derivatives of the fields , _ e.g. _ of the pressure to evaluate the sound speed , are typically not available in the tables . furthermore , the use of high - order interpolation and/or finite differences can lead to undesirable spurious oscillations . a second approach that removes all of these problems , uses analytic fits that have been proposed for the pressure . as an example , ref . @xcite suggested to fit the specific internal energy of the unified sly eos table with the expression @xmath56 where @xmath57 , @xmath36 and @xmath35 are in cgs units , and the coefficients @xmath58 are @xmath59 ( see table @xmath60 of @xcite ) . equation ( [ eq : analfit ] ) is obtained from eq . ( 15 ) of @xcite after using @xmath61 , where @xmath62 is the baryon number density and @xmath63 is the mass of the nucleons . as discussed in @xcite , it is then possible to compute the pressure from the value of @xmath35 using the first principle of thermodynamics at @xmath64 @xmath65 and thus to have an evaluation of the pressure which is thermodynamically consistent . the differences between the fit and the table are typically less then @xmath66 . unfortunately , although apparently very convenient , the evaluation of the fitting formulas containing several exponential and logarithmic functions , turns out to be computationally rather expensive even if done in a optimized way . as a third approach , which combines the efficiency of a table search with the thermodynamical consistency of an analytic fit , consists of performing a simple linear interpolation among the tabulated values constructed from the analytic fit . besides being highly efficient , a linear interpolation also eliminates the spurious oscillations that arise , for instance , in the derivative of the pressure if high - order interpolation formulas are used . in this case , the interpolation error can be reduced simply by populating the analytically constructed tables with a large number of entries , _ e.g. _ @xmath67 in place of the @xmath68 which are typically available in published tables . this third approach is the one actually implemented in ` whisky ` and provides a speed up of about @xmath69 with respect to the evaluation of the pressure via the analytic fits and with comparable accuracy . the initial data for our simulations are prepared as stationary and axisymmetric equilibrium solutions for rapidly rotating relativistic stars @xcite . adopting spherical quasi - isotropic coordinates , the line element of the corresponding spacetime is @xmath70 where @xmath71 , @xmath72 , @xmath73 and @xmath74 are functions of @xmath75 and @xmath76 . moreover we assume the usual relativistic @xmath77-constant law of differential rotation and that amounts to assume an angular - velocity distribution of the form @xmath78 \ , , \label{eq : velocityprofile_gr}\ ] ] where @xmath79 is the coordinate equatorial stellar radius and the coefficient @xmath80 provides a measure of the degree of differential rotation . expression ( [ eq : velocityprofile_gr ] ) represents the general - relativistic equivalent of the simpler newtonian @xmath77-constant law @xcite @xmath81 clearly , @xmath82 corresponds to a star in uniform rotation , while @xmath83 corresponds to a star with increasing degree of differential rotation . as a reference , @xmath84 yields a star with an angular - velocity profile which varies of a factor @xmath85 between the center and the surface of the star ( _ cf . _ left panel of figure [ fig : ida ] ) . in practice , we have computed a very large number of initial models using the sly prescription for the eos for which we have computed _ baryonic _ mass @xmath86 , the gravitational mass @xmath87 , the angular momentum @xmath88 , the rotational kinetic energy @xmath7 and the gravitational binding energy @xmath8 defined as @xmath89 where @xmath90 is the square root of the determinant of three - dimensional metric @xmath27 and @xmath91 is the fluid lorentz factor . we stress that the definitions ( [ eq : def m])([eq : def w ] ) of quantities such as @xmath88 , @xmath7 , @xmath8 and @xmath1 are meaningful only in the case of stationary axisymmetric configurations and should therefore be treated with care once the rotational symmetry is lost . plane of the initial models computed with @xmath92 , with the filled circles representing those we have evolved numerically ( a similar behaviour is shown also by the models with @xmath93 . indicated with solid thin lines are isocontours of constant baryon mass models while indicated with a thick dashed line is the threshold to the dynamical bar - mode instability as computed for a @xmath11 polytrope @xcite . note that the threshold for the instability tends to increase for smaller rest - masses . right panel : the same initial models as in the left panel but shown in @xmath94 planes with isocontours of constant @xmath1 ( upper part ) or constant @xmath95 ( lower part ) . see table [ table : initial ] for a summary of the properties of the initial models.,title="fig:",scaledwidth=49.0% ] plane of the initial models computed with @xmath92 , with the filled circles representing those we have evolved numerically ( a similar behaviour is shown also by the models with @xmath93 . indicated with solid thin lines are isocontours of constant baryon mass models while indicated with a thick dashed line is the threshold to the dynamical bar - mode instability as computed for a @xmath11 polytrope @xcite . note that the threshold for the instability tends to increase for smaller rest - masses . right panel : the same initial models as in the left panel but shown in @xmath94 planes with isocontours of constant @xmath1 ( upper part ) or constant @xmath95 ( lower part ) . see table [ table : initial ] for a summary of the properties of the initial models.,title="fig:",scaledwidth=49.0% ] -.5 cm .main properties of the simulated stellar models . starting from the left : the name of the model , the differential rotation parameter @xmath80the instability parameter @xmath1 , the compactness @xmath95 , the central rest - mass density @xmath96 , the maximum of the density @xmath97 ( note that for models with @xmath93 , @xmath98 . ) , the ratio between the polar and the equatorial coordinate radii @xmath99 , the proper equatorial radius @xmath100 , the gravitational mass @xmath87 , the total angular momentum @xmath88 divided by the square of the gravitational mass . [ cols="^,^,^,^,^,^,^,^,^,^ " , ] out of this large set , we have then selected for the numerical evolution a number of models so as to build sequences of constant baryonic mass with @xmath101 , degree of differential rotation @xmath102 and values of the instability parameter @xmath1 ranging between @xmath103 and @xmath104 . each model in this sequence is supramassive , namely it has a mass which is larger than the maximum mass allowed for a corresponding nonrotating model , _ @xmath105 , although it is not hypermassive , namely it does not have mass which is above the maximum mass for a uniformly rotating model , _ i.e. _ @xmath106 . there are two different reasons why such a large - mass model has been chosen . the first one is that we are interested in the development of a shear instability in the metastable star produced by the merger of a binary system of two neutron stars . as shown by a number of authors and most recently in @xcite , the product of this merger is either a supramassive or a hypermassive neutron star . hence , our reference model has a mass which is sufficiently large so as to be a reasonable approximation to the product of a binary neutron star merger . the second reason is that a sufficiently massive model is necessary in order to reach values of the instability parameter which are above the expected threshold for dynamical bar - mode instabilities as computed in @xcite , _ i.e. _ @xmath107 . indeed , the maximum possible value for @xmath1 within the computed sequence is around @xmath108 and thus just above the threshold ( _ cf . _ long - dashed line in the left panel of figure [ fig : idb ] ) . note also that the threshold for the instability tends to increase for smaller rest - masses . overall , when comparing with equilibrium models generated with a polytropic eos with @xmath55 and @xmath11 ( see table 1 of @xcite ) the realistic eos models reach higher compactness ( models in @xcite typically have @xmath109 ) but lower values of @xmath1 ( _ i.e. _ @xmath110 for the models considered here , while @xmath111 for the polytropic models considered in @xcite ) . the whole space of parameters is shown in figure [ fig : idb ] , whose left panel reports the position in the @xmath112 plane of the initial models computed with @xmath92 , and where the filled circles represent those we have evolved numerically . indicated with solid thin lines are isocontours of constant baryon mass models while indicated with a thick dashed line is the threshold to the dynamical bar - mode instability as computed for a @xmath11 polytrope @xcite . the right panel reports the same initial models considered in the left one but shown in @xmath113 planes with isocontours of constant @xmath1 ( upper part ) or constant @xmath95 ( lower part ) . the main properties of the simulated models are also reported in table [ table : initial ] , where we also introduce our naming convention . any initial model is indicated as ` m.% . # ` , with ` % ` being replaced by the value of the differential - rotation parameter @xmath114 and ` # ` by the instability parameter @xmath1 . as an example , ` m.1.200 ` is the star with @xmath92 and @xmath115 . finally , shown in figure [ fig : ida ] are the angular - velocity profiles ( left panel ) and the rest - mass density profiles ( right panel ) of some representative models , namely ` m.1.150 ` , ` m.1.200 ` and ` m.1.250 ` . indicated with different symbols , which match the ones in the left panel of figure [ fig : corotation_band ] , are the normalized radial positions of the corotation radii , with the one for model ` m.1.150 ` being shown filled to help distinguish it from the others . although we postpone to sections [ sec : evolution ] and [ sec : necessary conditions ] the discussion of the implications of these corotation radii , two aspects of the initial data are worth emphasizing . the first one is that the amount of differential rotation for a given value of @xmath114 effectively decreases when increasing the instability parameter @xmath1 ( _ cf . _ left panel of figure [ fig : ida ] ) , thus resulting in a smaller corotation band for models with large @xmath1 . the second one is that all the initial models evolved are axisymmetric but have a `` toroidal - topology '' , namely have the maximum density @xmath97 that is not at the center of the star , and thus @xmath116 . this toroidal deformation increases with the rotation and thus with @xmath1 . and @xmath92 ( a similar behaviour is shown also by the models with @xmath93 ) . indicated with different symbols , which match the ones in the left panel of figure [ fig : corotation_band ] and [ fig : corot_rad ] , are the normalized radial positions of the corotation radii , with the one for model ` m.1.150 ` being shown filled to help distinguish it . right panel : the same as in the left panel but for the initial rest - mass density.,title="fig:",scaledwidth=49.5% ] and @xmath92 ( a similar behaviour is shown also by the models with @xmath93 ) . indicated with different symbols , which match the ones in the left panel of figure [ fig : corotation_band ] and [ fig : corot_rad ] , are the normalized radial positions of the corotation radii , with the one for model ` m.1.150 ` being shown filled to help distinguish it . right panel : the same as in the left panel but for the initial rest - mass density.,title="fig:",scaledwidth=49.5% ] -.5 cm a number of different quantities are calculated during the evolution to monitor the dynamics of the instability . among them is the quadrupole moment of the matter distribution @xmath117 which we compute in terms of the conserved density @xmath118 rather than of the rest - mass density @xmath36 or of the @xmath119 component of the stress energy momentum tensor . of course , the use of @xmath120 in place of @xmath36 or of @xmath119 is arbitrary and all the three expressions would have the same newtonian limit , though with different amplitudes for the gravitational waveforms produced ( see for example @xcite ) . however , we here adopt the form ( [ eq : defquadrupole ] ) because @xmath121 is a quantity whose conservation is guaranteed by the form chosen for the hydrodynamics equations . the quadrupole moment ( [ eq : defquadrupole ] ) can be conveniently used to quantify both the growth time of the @xmath3 instability @xmath122 and the oscillation frequency once the instability is fully developed @xmath123 . ( hereafter we will indicate respectively with @xmath124 and @xmath125 the growth time and frequencies of the @xmath126 unstable modes and we note that , as will be discussed later on , during the simulation a number of different modes appear , thus justifying the use of the upper index `` @xmath127 '' ) . in practice , we perform a nonlinear least - square fit of the @xmath128 component of the computed quadrupole @xmath129 and we generally use as fitting function a sum of @xmath130 ( usually three ) exponentially modulated cosines @xmath131 where @xmath132 . because we commonly have only about @xmath133 cycles in the time interval considered for the fits , extreme care needs to be applied when computing the growth time , especially when the oscillation frequencies and the growth times are close to each other . in these cases , in fact , variations of the initial phase of the modes @xmath134 can result in large variation of the growth times . in view of this , we will not report them . using three components of the quadrupole moment in the @xmath135 plane we can define the distortion parameters @xmath136 and @xmath137 , as well as the axisymmetric mode @xmath138 as @xmath139 so that the modulus @xmath140 and the instantaneous orientation of the bar are given by @xmath141 finally , as a useful tool to describe the nonlinear properties of the development and saturation of the instability , the rest - mass density is decomposed into its fourier modes @xmath142 as @xmath143 the phase @xmath144 essentially provides the instantaneous orientation of the @xmath4-th mode when the corresponding mode has a nonzero power . note that despite their denomination , the fourier modes ( [ eq : modes ] ) do not represent proper eigenmodes of oscillation of the star . while , in fact , the latter are well defined only within a perturbative regime , the former simply represent a tool to quantify , within the fully nonlinear regime , what are the main components of the rest - mass distribution . as a final comment we note that while all quantities ( [ eq : defquadrupole])([eq : phimodes ] ) are expressed in terms of the coordinate time @xmath145 and are not invariant measurements , the lengthscale of variation of the lapse function at any given time is always larger than twice the stellar radius at that time , ensuring that events on the same timeslice are also close in proper time . in what follows we first describe the dynamics of the shear instability as deduced from the numerical simulations and then contrast our results with the phenomenological predictions on the necessary conditions for its development . . for each panel the upper part reports the evolution of the bar - distortion parameter @xmath146 , while the lower part shows the evolution of the power in the different modes of the fourier decomposition of the rest - mass density @xmath147 . see text for details . note that we here report only some models and with @xmath92 as they are representative also of those with @xmath93.,title="fig:",scaledwidth=42.5% ] 0.5 cm . for each panel the upper part reports the evolution of the bar - distortion parameter @xmath146 , while the lower part shows the evolution of the power in the different modes of the fourier decomposition of the rest - mass density @xmath147 . see text for details . note that we here report only some models and with @xmath92 as they are representative also of those with @xmath93.,title="fig:",scaledwidth=42.5% ] . for each panel the upper part reports the evolution of the bar - distortion parameter @xmath146 , while the lower part shows the evolution of the power in the different modes of the fourier decomposition of the rest - mass density @xmath147 . see text for details . note that we here report only some models and with @xmath92 as they are representative also of those with @xmath93.,title="fig:",scaledwidth=42.5% ] 0.5 cm . for each panel the upper part reports the evolution of the bar - distortion parameter @xmath146 , while the lower part shows the evolution of the power in the different modes of the fourier decomposition of the rest - mass density @xmath147 . see text for details . note that we here report only some models and with @xmath92 as they are representative also of those with @xmath93.,title="fig:",scaledwidth=42.5% ] . for each panel the upper part reports the evolution of the bar - distortion parameter @xmath146 , while the lower part shows the evolution of the power in the different modes of the fourier decomposition of the rest - mass density @xmath147 . see text for details . note that we here report only some models and with @xmath92 as they are representative also of those with @xmath93.,title="fig:",scaledwidth=42.5% ] 0.5 cm . for each panel the upper part reports the evolution of the bar - distortion parameter @xmath146 , while the lower part shows the evolution of the power in the different modes of the fourier decomposition of the rest - mass density @xmath147 . see text for details . note that we here report only some models and with @xmath92 as they are representative also of those with @xmath93.,title="fig:",scaledwidth=42.5% ] -0.5 cm for the models described in figure [ fig : evolutiongeneral ] ( black continuous line ) and of the axisymmetric mode @xmath148 ( red dashed line ) . indicated with vertical lines ( long - dashed black , dotted blue and dot - dashed magenta ) are the peak frequencies reported in the left panel of figure [ fig : corotation_band ] within the corotation band and which are used in figure [ fig : corot_rad ] to mark the position of the corotation radii . finally , shown with a shaded rectangular area is the corotation band for all models.,title="fig:",scaledwidth=42.5% ] 0.5 cm for the models described in figure [ fig : evolutiongeneral ] ( black continuous line ) and of the axisymmetric mode @xmath148 ( red dashed line ) . indicated with vertical lines ( long - dashed black , dotted blue and dot - dashed magenta ) are the peak frequencies reported in the left panel of figure [ fig : corotation_band ] within the corotation band and which are used in figure [ fig : corot_rad ] to mark the position of the corotation radii . finally , shown with a shaded rectangular area is the corotation band for all models.,title="fig:",scaledwidth=42.5% ] for the models described in figure [ fig : evolutiongeneral ] ( black continuous line ) and of the axisymmetric mode @xmath148 ( red dashed line ) . indicated with vertical lines ( long - dashed black , dotted blue and dot - dashed magenta ) are the peak frequencies reported in the left panel of figure [ fig : corotation_band ] within the corotation band and which are used in figure [ fig : corot_rad ] to mark the position of the corotation radii . finally , shown with a shaded rectangular area is the corotation band for all models.,title="fig:",scaledwidth=42.5% ] 0.5 cm for the models described in figure [ fig : evolutiongeneral ] ( black continuous line ) and of the axisymmetric mode @xmath148 ( red dashed line ) . indicated with vertical lines ( long - dashed black , dotted blue and dot - dashed magenta ) are the peak frequencies reported in the left panel of figure [ fig : corotation_band ] within the corotation band and which are used in figure [ fig : corot_rad ] to mark the position of the corotation radii . finally , shown with a shaded rectangular area is the corotation band for all models.,title="fig:",scaledwidth=42.5% ] for the models described in figure [ fig : evolutiongeneral ] ( black continuous line ) and of the axisymmetric mode @xmath148 ( red dashed line ) . indicated with vertical lines ( long - dashed black , dotted blue and dot - dashed magenta ) are the peak frequencies reported in the left panel of figure [ fig : corotation_band ] within the corotation band and which are used in figure [ fig : corot_rad ] to mark the position of the corotation radii . finally , shown with a shaded rectangular area is the corotation band for all models.,title="fig:",scaledwidth=42.5% ] 0.5 cm for the models described in figure [ fig : evolutiongeneral ] ( black continuous line ) and of the axisymmetric mode @xmath148 ( red dashed line ) . indicated with vertical lines ( long - dashed black , dotted blue and dot - dashed magenta ) are the peak frequencies reported in the left panel of figure [ fig : corotation_band ] within the corotation band and which are used in figure [ fig : corot_rad ] to mark the position of the corotation radii . finally , shown with a shaded rectangular area is the corotation band for all models.,title="fig:",scaledwidth=42.5% ] -0.5 cm as mentioned when discussing the initial data , we have evolved numerically two sequences of constant baryonic mass , with @xmath149 . the first sequence has a higher degree of differential rotation ( _ i.e. _ @xmath92 ) and instability parameter ranging from @xmath150 to @xmath151 . the second sequence , on the other hand , has a smaller degree of differential rotation ( _ i.e. _ @xmath152 ) and instability parameter between @xmath153 and @xmath115 . for both sequences the highest value of @xmath1 considered is also very close to the highest attainable with our initial data code ( see figure [ fig : idb ] ) . to keep the computational costs to an affordable level we have evolved of all these models for @xmath154 and although this time window is in general insufficient to capture the suppression of the instability , is adequate to measure the frequencies of the unstable modes and provide a first estimate of the growth times . after analyzing the results following the method outlined in the previous section and focusing on the properties of the distortion parameters @xmath146 , @xmath155 and @xmath148 , we find that _ all of the models _ show an @xmath3 instability . the maximum distortions obtained during the simulation time are @xmath156 ( _ i.e. _ distortions of @xmath157 with respect to the axisymmetric progenitor ) , with the maximum values being reached for the models of the @xmath92 sequence . the models of the @xmath93 sequence , in fact , have in general much smaller distortions , with @xmath158 over the timescale over which they have been evolved . we believe this is simply the consequence of the fact that the models in this sequence have smaller growth rates ( see also the discussion in the next section ) . to show the general behaviour of the simulation we focus on the description of the result of the simulation for the @xmath92 sequence as this is representative also of the one with @xmath93 . in particular , in the upper part of each panel in figure [ fig : evolutiongeneral ] we show the time evolution of the distortion parameter @xmath159 computed with eq . ( [ eq : etas ] ) for six representative models : ` m.1.150 ` , ` m.1.160 ` , ` m.1.200 ` , ` m.1.220 ` , ` m.1.241 ` and ` m.1.250 ` . similarly , in the lower parts of each panel we show the corresponding evolution of the fourier - modes computed from eq . ( [ eq : modes ] ) for @xmath160 . note that for models with lower values of @xmath1 the bar - mode deformation is very similar to the one already discussed in @xcite , growing exponentially and with only one unstable mode appearing ( _ cf . _ first row of figure [ fig : evolutiongeneral ] ) . however , as the rotation is increased , the development of the instability is more complex and at least two unstable modes appear which develop in different parts of the star . these two modes have very similar but distinct frequencies and growth times , leading to a series of beatings in the evolutions of @xmath159 , whose irregular evolution makes the calculation of the growth times challenging ( _ cf . _ second row ) . as the rotation is increased towards the maximal values of @xmath1 , up to three distinct modes appear and the evolution of the instability is correspondingly more complex ( _ cf . _ third row ) . a similar behaviour is shown also by the evolution of the fourier modes . the two models with lower @xmath1 , in fact , show a clear growth of the @xmath3 mode and of the @xmath13 , with the latter becoming first comparable and then larger when the instability is suppressed . in models ` m.1.250 , m.1.241 ` , on the other hand , the @xmath3 and @xmath16 modes have comparable amplitude for a long period and then the @xmath3 becomes the dominant one ( this is more clear in model ` m.1.241 ` ) . finally , for larger values of @xmath1 , the @xmath13 mode never attains values comparable with either the @xmath3 or the @xmath16 , which instead control the evolution . figure [ fig : psd ] reports the power spectral density ( psd ) of @xmath146 and @xmath148 for the same @xmath161 models ( solid black lines ) and allows to appreciate how the spectrum changes as the instability parameter is increased . more specifically , it is very apparent that at low @xmath1 the spectra have only one peak , whose maximum is marked by the vertical blue dotted lines and that is present at all the values of beta considered ; this is the mode we refer to as @xmath162 ( _ cf . _ table [ table : results ] and figure [ fig : corotation_band ] ) . as @xmath1 increases , the amplitude of this peak decreases and it becomes the weakest for very large rotation rates . starting from model ` m.1.200 ` ( although a hint of a peak is present already in ` m.1.160 ` ) , a second peak appears at higher frequency and is marked with magenta vertical dot - dashed lines ; this is the mode we refer to as @xmath163 and although it never becomes the largest one , its amplitude increases with @xmath1 . finally , a third peak appears at low frequency in the spectra starting from model ` m.1.220 ` ( even though a hint is present also for model ` m.1.210 ` ) and is marked with black vertical long - dashed lines . this mode , which we refer to as @xmath164 , becomes the dominant one at high @xmath1 . note also that in each panel of figure [ fig : psd ] we report with a dashed blue box the corotation band that will be further discussed in the following section . as a final remark we note that the peaks in the spectrum of @xmath165 ( red dashed lines in figure [ fig : psd ] ) are related to axisymmetric @xmath166 modes and most likely to @xmath167-modes oscillation excited by the development of the instability . indeed , their frequencies @xmath168 match reasonably well the phenomenological fit for the @xmath167-mode frequencies of nonrotating neutron stars with realistic eoss computed in @xcite . because our sequences do not contain nonrotating configurations ( the rest mass is larger than the maximum one ) , this association is just qualitative and a more detailed investigation of the frequency spectra of the equilibrium models is necessary to confirm this suggestion . @cccccccccc model & @xmath169 & @xmath170 & @xmath171 & @xmath172 & @xmath173 & @xmath174 & @xmath175 & @xmath176 & @xmath177 + & ( khz ) & ( khz ) & ( khz ) & ( khz ) & ( khz ) & ( khz ) & ( khz ) & ( khz ) & ( khz ) + [ 1 mm ] ` m.1.140 ` & @xmath178 & @xmath179 & @xmath180 & @xmath179 & @xmath179 & @xmath181 & @xmath179 & @xmath182 & @xmath183 + ` m.1.150 ` & @xmath184 & @xmath179 & @xmath181 & @xmath179 & @xmath179 & @xmath185 & @xmath179 & @xmath186 & @xmath187 + ` m.1.160 ` & @xmath188 & @xmath179 & @xmath189 & @xmath190 & @xmath179 & @xmath191 & @xmath192 & @xmath193 & @xmath194 + ` m.1.180 ` & @xmath195 & @xmath179 & @xmath196 & @xmath197 & @xmath198 & @xmath199 & @xmath200 & @xmath201 & @xmath202 + ` m.1.200 ` & @xmath203 & @xmath204 & @xmath205 & @xmath206 & @xmath207 & @xmath208 & @xmath209 & @xmath210 & @xmath211 + ` m.1.210 ` & @xmath212 & @xmath213 & @xmath214 & @xmath215 & @xmath216 & @xmath217 & @xmath218 & @xmath219 & @xmath220 + ` m.1.220 ` & @xmath221 & @xmath222 & @xmath223 & @xmath224 & @xmath225 & @xmath226 & @xmath227 & @xmath228 & @xmath229 + ` m.1.230 ` & @xmath230 & @xmath231 & @xmath232 & @xmath233 & @xmath234 & @xmath235 & @xmath236 & @xmath237 & @xmath238 + ` m.1.241 ` & @xmath239 & @xmath240 & @xmath241 & @xmath242 & @xmath243 & @xmath244 & @xmath245 & @xmath246 & @xmath247 + ` m.1.244 ` & @xmath248 & @xmath249 & @xmath250 & @xmath251 & @xmath252 & @xmath253 & @xmath254 & @xmath255 & @xmath256 + ` m.1.247 ` & @xmath257 & @xmath258 & @xmath259 & @xmath260 & @xmath261 & @xmath262 & @xmath263 & @xmath264 & @xmath265 + ` m.1.250 ` & @xmath266 & @xmath267 & @xmath268 & @xmath269 & @xmath270 & @xmath271 & @xmath272 & @xmath273 & @xmath274 + ` m.2.125 ` & @xmath275 & @xmath276 & @xmath277 & @xmath179 & @xmath278 & @xmath279 & @xmath179 & @xmath280 & @xmath281 + ` m.2.150 ` & @xmath282 & @xmath283 & @xmath284 & @xmath179 & @xmath285 & @xmath286 & @xmath179 & @xmath287 & @xmath288 + ` m.2.175 ` & @xmath289 & @xmath290 & @xmath291 & @xmath179 & @xmath292 & @xmath293 & @xmath179 & @xmath294 & @xmath295 + ` m.2.200 ` & @xmath296 & @xmath297 & @xmath298 & @xmath179 & @xmath299 & @xmath300 & @xmath179 & @xmath301 & @xmath302 + to support the interpretation of these instabilities as shear instabilities we have considered whether the necessary conditions suggested by @xcite and by @xcite are met . we recall that watts and collaborators pointed out that an unstable configuration should have the unstable mode with a frequency within the corotation band . in the case of the newtonian expression ( [ eq : velocityprofile_newt ] ) , the corotation band is simpler to compute and for a mode with frequency @xmath303 and azimuthal number @xmath4 , this is simply given by @xmath304 we have therefore checked whether any of the unstable @xmath3 ( bar - modes ) with computed frequency @xmath305 has pattern speed velocity , @xmath306 , within the corotation band . a careful and rather involved analysis has indeed confirmed the prediction of watts and collaborators : namely , _ all the unstable modes are within the corotation band _ of the progenitor axisymmetric model . these results for the models with @xmath92 are listed in table [ table : results ] , where we report the edges of the general - relativistic corotation band in terms of angular frequency at the surface @xmath307 and on the axis @xmath177 , as well as the frequencies of the unstable @xmath3 modes obtained from the position of the peaks in the psd ( _ i.e. _ @xmath308 ) or from the nonlinear fitting of eq . ( [ eq : quadrupolefit ] ) ( _ i.e. _ @xmath309 ) . note that depending on the rate of rotation and the degree of differential rotation some frequencies may not be present in the corotation band . the data in table [ table : results ] is also shown in the left panel of figure [ fig : corotation_band ] , where we plot the position of the non - axisymmetric frequencies @xmath123 within the corotation band as a function of @xmath1 ; indicated with a filled symbol , to distinguish it from the others , is the @xmath310 frequency for model ` m.1.150 ` ( _ cf . _ figure [ fig : ida ] and [ fig : corot_rad ] ) . the right panel of the same figure shows the corresponding information and on the same scale but for the models with @xmath93 ( the inset shows instead a magnified view ) . clearly , for both sequences all the unstable modes are within the band and , as the stellar rotation rate increases , more unstable modes appear for the same value of @xmath1 . for models with @xmath92 and shown as a function of @xmath1 ; indicated with thick solid lines are the edges of the corotation band , _ i.e. _ the frequency interval between @xmath177 and @xmath176 . shown with a filled symbol , to help distinguish it from the others , is the @xmath164 frequency for model ` m.1.150 ` ( _ cf . _ figure [ fig : ida ] and [ fig : corot_rad ] ) . right panel : the same as in the left one and with the same scale but for models with @xmath93 ; the inset shows instead a magnified view . clearly , in both panels all the unstable modes are within the band and up to three unstable modes appear for the @xmath92 sequence with increasing @xmath1.,title="fig:",scaledwidth=49.5% ] for models with @xmath92 and shown as a function of @xmath1 ; indicated with thick solid lines are the edges of the corotation band , _ i.e. _ the frequency interval between @xmath177 and @xmath176 . shown with a filled symbol , to help distinguish it from the others , is the @xmath164 frequency for model ` m.1.150 ` ( _ cf . _ figure [ fig : ida ] and [ fig : corot_rad ] ) . right panel : the same as in the left one and with the same scale but for models with @xmath93 ; the inset shows instead a magnified view . clearly , in both panels all the unstable modes are within the band and up to three unstable modes appear for the @xmath92 sequence with increasing @xmath1.,title="fig:",scaledwidth=49.5% ] -.5 cm watts , andersson and jones also give a qualitative description , summarized in their figure 2 of @xcite , of how the unstable and stable models should be distributed in the ( @xmath311 ) plane . the considerations they make are particularly simple : for high degrees of differential rotation , that is at low values of @xmath80 , the corotation band is rather wide and there will be both an upper value and a lower value of @xmath1 between which the shear instability can develop ( these critical values of @xmath1 correspond to the entrance and exit of the unstable mode in the corotation band ) . this situation corresponds therefore to the one commonly encountered in numerical simulations , such as those in @xcite , and for which the shear instability takes place only for very small values of the instability parameter and on timescales that are much longer than the dynamical one . when moving to larger degrees of differential rotation , that is when going to smaller values of @xmath80 , the corotation band becomes larger and larger and the shear instability can develop essentially for all values of @xmath1 , merging with the dynamical bar - mode instability for @xmath312 . this is exactly what has been found here . conversely , when moving to smaller degrees of differential rotation , that is when going to higher values of @xmath80 , the corotation band becomes thinner and the shear instability can develop only for a smaller range of @xmath1 . as the differential rotation is further decreased and the star tends to rotate uniformly , the corotation band width vanishes , all the models are stable to the shear instability and subject only to the dynamical bar - mode instability for @xmath313 . this is indeed the case for the unstable models evolved in @xcite , which were purely ( bar - mode ) dynamically unstable and none of which had the unstable mode within the corresponding corotation band , but above it . in essence , therefore , there should be an intermediate range of @xmath80 for which the instability is absent at low @xmath1 , appears at intermediate values and then disappears again at high @xmath1 , thus defining an interval of values of @xmath1 for which the models are unstable . to validate also this prediction we simulated a sequence of models with smaller differential rotation and @xmath152 this seconds sequence has the same baryonic mass as that with @xmath314 , but with @xmath1 in a smaller range , namely between @xmath153 and @xmath115 ( note that @xmath115 is also very close to the largest value for which we could build an equilibrium model . ) unfortunately , also all of these models show an @xmath3 shear instability , with the unstable frequencies falling within the corotation band , as it was for the @xmath314 stars . of course , lack of evidence is not evidence of absence and the fact that we have not found stable models within our range of values of @xmath1 most likely means that these stable models have to be searched either for values of @xmath315 ( for this value of @xmath152 ) , or by moving towards higher values of @xmath80 , where the corotation band is less large . an obvious consequence of the phenomenological scenario described by @xcite and confirmed by the calculations reported here is that the idea of a low-@xmath2 instability is indeed misleading and it is instead more meaningful to think of a more generic shear instability that , depending on the degree of rotation and of differential rotation , may manifest itself on timescales that are comparable with the dynamical ones ( as in the cases reported here ) or on much longer ones ( as in the cases reported in @xcite ) . another phenomenological prediction made by watts , andersson and jones in @xcite , is that the growth times should be shorter in the center of the band and increase towards the edges ( the growth times should in fact diverge at the edges ) . as shown in figure [ fig : corotation_band ] , all of the simulated models do have unstable modes well inside the corotation band and this probably explains why we observe them develop on a timescale which is comparable with the dynamical one and not on much longer timescales as in the original finding of @xcite . we also find indications that the @xmath152 models , which generically have longer growth times , have unstable modes which occupy regions of the corotation band that are overall more central ( _ cf . _ right panel of figure [ fig : corotation_band ] ) and thus in contrast with what expected from @xcite . however , the difficulties mentioned above in computing an accurate estimate of the growth times and the intrinsic difficulties in determining what is the central part of the corotation band , prevent us from providing a more quantitative validation of this prediction . we now switch to consider our results in the light of the other necessary condition for the onset of the shear instability discussed in @xcite . as mentioned in the introduction , it has in fact been shown that the presence of a minimum in the profile of the vortensity is a necessary condition for a mode in corotation to be unstable @xcite . the intuitive description is that the vortensity well can act as a resonant cavity inside the star , amplifying the modes that happen to lay near its minimum @xcite . indeed , the growth rate of the unstable mode is expected to depend on the location of its corotation radius with respect to the vortensity profile , being proportional to the depth of its corotation radius inside the vortensity well @xcite . note that the mere existence of a local minimum in the vortensity can not be used as a sufficient condition for the occurrence of a shear instability . all of the unstable models considered in @xcite , in fact , do show a local minimum in the vortensity but do not have the unstable modes in the corotation band . for three representative models with small , medium and high values of @xmath1 . the different symbols match the ones in figure [ fig : ida ] and in the left panel of figure [ fig : corotation_band ] , with the one for model ` m.1.150 ` reported as filled , and show the actual position of the corotation radii . note that none of these coincides with the minimum of @xmath316 ; for compactness we have reported only the models with @xmath92 . right panel : normalized corotation radii for the different frequencies @xmath123 presented in the left panel of figure [ fig : corotation_band ] shown as a function of @xmath1 ; the same convention of the previous figures is used for the different symbols.,title="fig:",scaledwidth=49.5% ] for three representative models with small , medium and high values of @xmath1 . the different symbols match the ones in figure [ fig : ida ] and in the left panel of figure [ fig : corotation_band ] , with the one for model ` m.1.150 ` reported as filled , and show the actual position of the corotation radii . note that none of these coincides with the minimum of @xmath316 ; for compactness we have reported only the models with @xmath92 . right panel : normalized corotation radii for the different frequencies @xmath123 presented in the left panel of figure [ fig : corotation_band ] shown as a function of @xmath1 ; the same convention of the previous figures is used for the different symbols.,title="fig:",scaledwidth=49.5% ] -.5 cm to validate whether this newtonian condition holds also for the general - relativistic instabilities simulated here , we have computed for all the models the newtonian vortensity , which is defined as the ratio , along the radial cylindrical coordinate , between the radial vorticity and the density @xmath317 where @xmath318 is the radial cylindrical coordinate ( a fully general - relativistic definition of the vortensity is also possible but more complicated to compute and not significantly different from the newtonian one ) . when doing this we found that all the unstable models have vortensities with a local minimum in the star and this is shown in the left panel of figure [ fig : corot_rad ] , which reports the radial profiles of the vortensity @xmath316 for three representative models with small , medium and high values of @xmath1 . the different symbols match the ones in the left panel of figure [ fig : ida ] and [ fig : corotation_band ] , with the one for model ` m.1.150 ` reported as filled , and show the actual positions of the corotation radii . the right panel of the same figure , on the other hand , shows the normalized corotation radii for the different frequencies @xmath123 presented in the left panel of figure [ fig : corotation_band ] as a function of @xmath1 . note that none of these coincides systematically with the minimum of @xmath316 , nor with the maximum of the rest - mass density . all of the corotation radii , however , do move towards larger radial positions as the rotation rate is increased , exactly as does the minimum of the vortensity and the rest - mass maximum . although we can not confirm that modes whose corotation radius is closer to the minimum of the vortensity have systematically shorter growth times ( our data is not sufficiently accurate for this ) , we can compare the models with the higher degree of differential rotation @xmath92 with those having a smaller degree of differential rotation @xmath93 . in the first case the vortensity well is considerably deeper , with @xmath319 , while for the latter the vortensity has shallower wells with @xmath320 . since the models with @xmath321 have smaller growth rates , our results indicate therefore that also the existence of a local minimum in the vortensity can be taken as a necessary condition for the development of the shear instability and that the depth of the vortensity well can be used to estimate the growth of the instability . for many years the properties of rapidly rotating and self - gravitating fluids have been characterized by a complete analytic perturbative theory which provided , for instance , sufficient conditions for the development of instabilities ( see , for instance , @xcite for a collection of results ) . as numerical simulations have become increasingly accurate and stable on longer timescales , these predictions , both in newtonian theory and in full general relativity , have been verified , corrected and in some cases extended . our ability of modelling such configurations has now reached a maturity such that a number of new properties and instabilities have been `` discovered '' numerically , but which do not have behind a fully perturbative description . the main reason for this is that most of this phenomenology is the result of physical scenarios which are much more complex than the ones investigated perturbatively in the past , _ e.g. _ non - isolated systems with high differential rotation and exchanging mass and angular momentum , which are much more difficult to treat analytically . a most notable example of these complex and yet ubiquitous instabilities is the so - called `` low-@xmath2 instability '' , which was initially found in @xcite and then reproduced in a number of other different scenarios @xcite . as the phenomenological description of this instability provided by the simulations has become richer and richer , a full understanding of the mechanisms that lead to its development has lagged behind and it is presently unclear . we are therefore in a situation in which numerical simulations can probe regimes and conditions which are not yet accessible to perturbative calculations , and can guide the latter by confirming or refuting those scenarios that although possible do not find a realization in practice . this work has followed this spirit and has used fully general - relativistic calculations of rapidly and differentially rotating neutron stars modeled with a realistic eos to shed some light on the development of the low-@xmath2 instability . in particular we have concentrated our attention on validating an indirect prediction , made by watts , andersson and jones @xcite , who recognized the low-@xmath2 instability as the manifestation of a more generic class of instabilities associated to the existence of a corotation band @xcite , the _ shear instabilities _ , and that should develop for _ any value _ of the instability parameter @xmath1 when sufficient amounts of differential rotation are present . this is exactly what we have found in our simulations . more specifically , we have performed simulations of sequences of neutron - star models described by a realistic sly eos @xcite and having constant rest - mass and degrees of differential rotation , but with different amounts of rotation . in all cases considered we have found the development of a bar - mode instability growing on a dynamical timescale , even when the initial axisymmetric models were well below the critical limit for the dynamical bar - mode instability . these results , which match well the phenomenological scenario portrayed in @xcite , suggest therefore that the idea of a low-@xmath2 instability is indeed misleading and should be replaced by the more general one of shear instability . depending then on the degree of rotation and of differential rotation , the instability will develop on timescales that are comparable to the dynamical one ( as reported here ) or on much longer ones ( as reported in the first low-@xmath2 instability studies ) . special attention has also been paid to the properties of the unstable modes and to their position within the corotation band or the vortensity profiles . in particular , we have shown that all the unstable modes are within the corotation band of the progenitor axisymmetric model ( which is the necessary condition for the development of the instability proposed by @xcite ) and that all of the unstable models have vortensity profiles with a local minimum ( which is the necessary condition suggested by @xcite ) . finally , by comparing the growth times among models with different degree of differential rotation we have shown that there is a correlation , although not a strong one , between the depth of the vortensity well and growth rate of the instability , with the latter being larger for models with deeper wells . in summary , the results presented here shed some light on several aspects of shear instabilities , but they also reveal that more work is required , for instance , to distinguish between the predictions based on the corotation band and the ones based on the vortensity well , or to establish whether in effect they just represent two different ways of expressing the same physical conditions . clarifying these aspects will requires additional analytical and numerical modelling , and this will be part of our future research . it is a pleasure to thank anna watts for enlightening discussions and for sharing her notes with us , as well as nils andersson , ian jones , joel tohline for useful comments . we are grateful to nikolaous stergioulas for the code to compute the initial data used in these calculations and to e. schnetter and all the ` carpet / cactus ` developers . the computations were mainly performed on the damiana cluster at the aei and , in a part , on the parma s albert2 cluster and on the bcx cluster at cineca . this work was supported in part by the dfg grant sfb / transregio 7 `` gravitational - wave astronomy '' , by `` compstar '' , a research networking programme of the european science foundation , and by the della riccia foundation . 10 url # 1#1urlprefix[2][]#2 f and haensel p 2001 _ astronomy and astrophysics _ * 380 * 151167

shallow water equations are widely used at several scales for liquid flows when the depth is smaller than the longitudinal scale . they are based on assumptions on the velocity profile which allow to estimate its shape factor and the shear stress at the wall to close the system of equations . we present here a two layers decompositon between an ideal fluid and a viscous layer , in the spirit of ibl ( interactive boundary layer ) or ivi ( inviscid viscous interaction ) introduced in aeronautics . it means that the two layers interact . the displacement thickness of the viscous layer and the order of magnitude of the shear stress at the wall are supposed small and of same order of magnitude . we use this small parameter for expansion and obtain a coupled system of two interacting layers , the viscous layer being then a kind of apparent topography due to the displacement thickness . we show the link with classical shallow water equations . the assumption on the velocity profile shape is rejected in the viscous layer , which makes assumptions on profiles ( shape factor and wall shear stress ) more precise . we test the final system on some classical cases like the starting flow , the flow over bump at several froude regimes . we finally focus on the flow over a bump in subcritical flows . the computed wall shear stress in this configuration presents some characteristic features of the influence of the boundary layer , it depends on the topography , and moreover its maximum is reached before the top of the bump , which is impossible in classical shallow water equations . also , an additional term with the same magnitude as the shear stress appears in the system , which can be interpreted as a correction to the pressure . franois james , pierre - yves lagre and mathilde legrand +

1607.00923

many phenomena in maritime or fluvial hydraulics involve free surface flows in shallow waters . the so - called shallow water equations were introduced by saint - venant in 1871 @xcite for studying floods and tides . since then , the model has been widely extended and is used in the modelling and numerical simulation of a number of natural or manmade phenomena such as river flow @xcite , flood forecasting @xcite , pollutant transport @xcite , dam - break @xcite , tsunami @xcite , overland flow @xcite , soil erosion @xcite . the shallow water system can be obtained from the incompressible navier - stokes equations under several hypotheses , the main one being that the characteristic wavelength is much larger than the water depth . two consequences follow then : the hydrostatic pressure law , and the viscosity vanishing in the horizontal direction . next , to proceed from navier - stokes to shallow water , we integrate the equations along the vertical direction . at this point , we have to be careful about the vertical velocity profile , which on the one hand has to be approximated to deal with nonlinearities , but on the other hand drives the bottom boundary condition , hence the friction phenomena . two classical assumptions on the velocity profile lead to explicit integrations . first , a viscous poiseuille - like ( i.e. parabolic ) profile on the whole water depth gives rise to a linear ( with respect to the mean velocity ) friction term , sometimes referred to as laminar friction . even simpler , the constant vertical profile , somehow corresponding to an ideal fluid , has the main drawback that by construction the boundary condition at bottom disappears ( non penetration for an ideal fluid ) . hence there is no friction term in the integrated equations . this corresponds to the original model proposed by saint - venant , which he obtained by completely different means . friction has to be added afterwards , using empirical laws ( manning , chzy, ... )@xcite . the main drawback of these two points of view is the non - adaptability of the viscous term due to large velocity variations , so that the assumed profiles ( parabolic or flat ) do not hold . we propose here a more flexible coupled system based on a supplementary viscous layer study . more precisely the aim of this paper is to understand how the viscous no - slip boundary condition gives rise to the friction term in the integrated system . this is done by assuming the existence of a viscous layer above the ground , with thickness quantified by a parameter @xmath2 , which is related to the inverse of the reynolds number of the fluid . here @xmath3 will be small but not necessarily vanishing as in classical boundary layer analysis . above this layer lies an ideal fluid . integrating the incompressibility equation under this assumption leads to the same conservation equation on the fluid depth . on the contrary , the integration of the momentum equation exhibits major differences . indeed , it turns out that the order of magnitude of the friction term , which results from the parietal constraints , is precisely @xmath3 , while the above mentioned poiseuille profile leads to a @xmath4 order of magnitude ( see part [ subsec : sw ] ) . motivated by this precise quantification of the intensity of friction , we introduce a new closure for the flux momentum . it can be interpreted as an additional pressure law , with the same magnitude @xmath3 , so that both terms vanish for an ideal fluid , that is @xmath5 . at this stage , we obtain a system of two equations which are similar in structure with the usual shallow water system , but involving several additional unknowns functions precisely related to the viscous layer . following classical methodology in aerodynamics , see e.g. @xcite , the second step consists therefore in a careful analysis of this layer , through the prandtl equation , which is integrated along the vertical axis to obtain the so - called von krmn equation . it drives the evolution of the so - called displacement thickness @xmath6 ( see figure [ fig : domain ] ) , which is involved in the definition of the corrective pressure mentioned above , and can be interpreted as some physical thickness of the viscous layer . the system has to be complemented by the velocity equation of the ideal fluid , since it is involved in the von krmn equation . together with some assumptions on the vertical velocity profile in the viscous layer , this leads to a system of four equations . in practice we will discuss the effects for flows over short bumps . the acceleration induced by the bump will change a lot the basic flow so that the shape velocity profile is no longer a half poiseuille nor a flat one . this study aims to understand this kind of flows which are not taken into account by the shallow water equations themselves . the major limit of the following theory is the necessity of the ideal fluid / viscous layer decomposition . the outline of the paper is as follows . in a first section we recall the navier - stokes system , and state the long wave approximation which is convenient for shallow water approximation . next we turn to the viscous layer analysis , and derive prandtl and von krmn equations . vertical velocity profiles are also introduced . the third section is devoted to the derivation of the extended shallow water system leading to our new model ( [ finalmass])-([finalideal ] ) . in section 4 we derive some formal properties of the model , together with analytical steady state solutions . finally , we evidence several properties of the model by numerical simulations , the purpose of the paper not being to develop sharp discretizations . in this section we recall how classical models for shallow waters are obtained from navier - stokes equations . the first assumption is a long wave approximation , stating that indeed we deal with a thin layer of water . next , we integrate along the vertical direction , assuming a given velocity profile on the whole water depth . we consider a fluid in a time - dependant domain @xmath7 ( see figure [ fig : domain ] ) , where the first dimension @xmath8 is horizontal and with infinite extension , the second one @xmath9 is vertical . the fluid is limited by the ground , represented by a function @xmath10 , and some free surface @xmath11 . the ground is a given function ( no erosion or ground modification ) , whereas the free surface is an unknown of the problem . the water height is @xmath12 . in this study , the properties of the air above the free surface are completely neglected . , the bottom is a given function @xmath13 and @xmath14 is the free surface . two families of velocity profiles for the flow over the topography , first with the usual half - poiseuille description ( dashed ) ( see [ subsec : sw ] for the explanations ) , and second with the flat profile with a boundary layer ( plain ) . note the shear ( slope of the velocity at the wall ) is completely different in both descriptions.,width=415 ] our starting point to describe the liquid layer is the dimensionless navier - stokes equations for an incompressible newtonian fluid in laminar flow @xcite : @xmath15 where * @xmath16 is the velocity vector * @xmath17 is the reynolds number expressing the ratio between the inertia force and the viscosity * @xmath18 is the froude number , ratio between the kinetic and potential energies * @xmath19 is the pressure nondimensionalization has been made with the same characteristic length for the abscissa and the ordinate , and corresponds for instance to the mean water depth @xmath20 , that appears in the reynolds and froude numbers . we consider in this work only laminar flows because the asymptotics from the navier - stokes equations is clearer , and moreover the resulting description is quantitative . a similar study can be made with a modified reynolds tensor and we can hope it leads at least to some qualitative description . the system is complemented with the following boundary conditions : * bottom : no - slip @xmath21 for @xmath10 * free surface @xmath11 * * kinematic boundary condition : @xmath22 * * continuity of the stress tensor : @xmath23 , where @xmath24 is the stress tensor and @xmath25 is the normal to the free surface . up to now , no hypothesis has been taken into account for the size order of the characteristic quantities @xmath26 . we have in mind applications to rivers or coastal flows where the following conditions may be observed : * the horizontal velocity has small variation along the vertical ; * the reynolds number is large ; * the vertical velocity is small compared to the horizontal velocity . let us start by investigating the third one , which justifies the following scaling for the velocities : @xmath27 then the mass conservation equation ( [ mass ] ) enforces also a scaling for the space variables @xmath28 since @xmath29 hence there are two options for the variable scaling : 1 . long wave hypothesis . we introduce the aspect ratio of the system @xmath30 . the scaling is @xmath31 this scaling is the same as a nondimensionalization of the navier - stokes equations with a characteristic length @xmath32 for the domain and a characteristic physical height @xmath20 . this scaling has to be complemented with @xmath33 . so , the long wave hypothesis needs a long time study . furthermore , since @xmath34 , we have @xmath35 and so the physical slope needs to be small enough . thin layer scaling : @xmath36 this scaling restricts the study of small vertical velocity only to a thin water depth which tends to zero when @xmath37 . it is the classical scaling used in the boundary layer approach . so , the only scaling which is compatible with a small vertical velocity compared to the horizontal velocity is the long wave scaling . let us see the consequences for the set of equations ( [ mass])-([momentum2 ] ) and the boundary conditions . x}\tilde p+\frac{1}{re_h}\left[\varepsilon^2\partial_{\tilde x}^2\tilde u+\partial_{\tilde y}^2\tilde u\right]\\ \label{tildem2 } \varepsilon^2\left[\partial_{\tilde t}\tilde v+\tilde u\partial_{\tilde x}\tilde v+\tilde v \partial_{\tilde y}\tilde v\right]=&-\frac{1}{fr^2}-\partial_{\tilde y}\tilde p+\frac{\varepsilon}{re_h}\left[\epsilon^2 \partial_{\tilde x}^2\tilde v+\partial_{\tilde y}^2\tilde v\right]\\ \tilde u=\tilde v=&0 \qquad \textrm{at}\qquad \tilde y=\tilde{f_b}\nonumber \\ \tilde v=\partial_{\tilde t}\tilde \eta+&\tilde u\partial_{\tilde x}\tilde \eta \qquad \textrm{at}\qquad \tilde y=\tilde \eta\nonumber \\ \begin{pmatrix } \varepsilon \big((2\varepsilon \partial_{\tilde x}\tilde u-\tilde p)\partial_{\tilde x}\tilde \eta-\varepsilon \partial_{\tilde x}\tilde v\big)-\partial_{\tilde y}\tilde u\\ \varepsilon \big(\partial_{\tilde x}\tilde \eta(\partial_{\tilde y } \tilde u+\varepsilon \partial_{\tilde x}\tilde v)-2\partial_{\tilde y}\tilde v\big)-\tilde p \end{pmatrix } & = \begin{pmatrix}0\\0\end{pmatrix } \qquad \textrm{at } \qquad \tilde y=\tilde \eta \nonumber \end{aligned}\ ] ] taking an approximation at order @xmath39 leads to : * simplified version of the stress tensor continuity at the free surface : @xmath40 and @xmath41 at @xmath42 . * hydrostatic pressure with ( [ tildem2 ] ) and @xmath40 at the surface : @xmath43 this result for pressure at order @xmath39 is already observed , see for instance @xcite , chapter 5 . * cancellation of the viscosity in the @xmath8 direction in ( [ tildem1 ] ) : @xmath44 we emphasize here that the above properties , which are classical shallow water hypotheses , are brought out solely by the long wave approximation . to summarize , the long wave approximation of the navier - stokes equations consists of the following set of equations : @xmath45 this will be our reference system for the remaining of this article . hence , from now on we shall drop the tildes when referring to the variables and unknowns of the previous system of equations . notice that in this setting the effective reynolds number is @xmath46 , which takes into account the aspect ratio of the system @xmath47 . we shall come back more precisely on this when studying the prandtl equation below , but notice that the actual ideal fluid limit here corresponds to @xmath48 , leading to the hydrostatic euler system : @xmath49 as usual , the no - slip boundary condition is not relevant for the ideal fluid and has to be replaced by the non - penetration condition @xmath50 recovering some connection between the ideal fluid equations and the no - slip condition is precisely the aim of the viscous layer theory , which we will present in section [ sec : boundlayer ] . audusse et al . propose in @xcite to consider in the long - wave approximation of both euler and navier - stokes system several layers and integrate over each layer . similarly to finite volume numerical methods , matter and momentum fluxes are taken into account at each interface between the layers . therefore they are faced with a superposition of shallow water systems each one interacting with its neighbours . this can be interpreted as a discretization in the vertical direction of the two - dimensional systems . our method is different in that the two layers have different status , based on physical motivations . the relative thickness of each part is not fixed but evolves in time . the aim of this paragraph is to recall the classical way to obtain shallow water equations by vertical integration over the whole water depth , first in the long - wave approximation of the euler equations , then in the long - wave approximation of the navier - stokes equations . averaging the euler system ( [ eqflupcl1])-([eqflupcl5 ] ) over the water depth leads to the so - called shallow water model without friction . we introduce the natural averaged velocity . [ meanvel ] the averaged velocity @xmath51 is defined by @xmath52 . with this construction , @xmath51 is defined @xmath53-a.e . ( i.e. only for @xmath54 ) . then straightforward computations ( see section [ sec : esw ] below for details ) lead to @xmath55 the momentum equation above requires some closure for the flux . this can be achieved by assuming a flat velocity profile over the water depth , which is compatible with the ideal fluid assumption . this leads to @xmath56 , hence the momentum equation becomes @xmath57 equations and correspond to the equations derived by de saint - venant by completely different means , with the additional topography term . if we proceed in an analogous way from the navier - stokes equations in long wave scaling , we obtain a viscous shallow water where the water depth conservation is the same as before , but the momentum equation becomes @xmath58 once again , without the presence of complementary equations , a closure is needed for the @xmath59-profile over all the water depth . explicit formul are given in @xcite depending of the flow index of the fluid . focusing on a newtonian fluid with a negative constant slope , the balance between the friction and the driving force of the slope gives a parabolic solution for @xmath59 , the half - poiseuille or nusselt solution : @xmath60 with @xmath61 a positive constant . in the general case , even if the invariance is not conserved , a similar form is chosen for the velocity profile but the deviating is expressed with a variable amplitude instead of a constant one : @xmath62 . nevertheless , the explicit integration gives the following system : @xmath63 the friction term which naturally appears in this derivation is @xmath64 . it is indeed linear with respect to the mean velocity @xmath51 , and is usually referred to as the laminar friction term . an alternative way to reintroduce some viscosity in the inviscid shallow water equations ( [ swcm])-([swcmo ] ) is to put a posteriori some friction term similar to the one in the poiseuille - shallow water . a general form for such a friction term is @xmath65 to be consistent with the previous equation , but a large family of empirical friction laws involves a @xmath66 depending on @xmath51 as well , for instance chzy , manning , see @xcite for a bibliographical study . we turn now to the main step towards the model we look for . it mainly consists in dividing the fluid in two layers * an ideal fluid layer dealing with the free surface ; * a thin viscous layer with the no - slip condition at the bottom . between these layers lies an interface which connects the velocities and will be described below . in the first layer we take advantage of the explicit integration along the vertical , in the second we take into account the viscosity in the vertical direction , hence we recover some friction in the integrated equations . this section is devoted to the study of the viscous layer . we introduce a small parameter @xmath67 , whose magnitude will be specified below . it is related to the thickness of the viscous layer , but does not correspond to its actual physical value . we follow the classical strategy used in the so - called boundary layer theory , see e.g. @xcite,@xcite , except that in that case @xmath68 , whereas we keep a finite value here . the first step is to derive a set of equations , sometimes referred to as rnsp ( reduced navier - stokes / prandtl ) equations @xcite . since they are yet two - dimensional , the next step consists in writing another equation , by integrating the prandtl system over the viscous layer height . this leads to the so - called von krmn equation , where extra unknowns are introduced . to obtain a closed form , some assumptions have to be done on the vertical velocity profile , which are discussed in the last paragraph . starting from the navier - stokes equations with long wave scaling , we introduce the following change of variables , referred to as the prandtl shift : @xmath69 as a consequence , we have the relations @xmath70 velocities must be scaled in order to preserve the mass conservation , which rewrites here @xmath71 hence we choose @xmath72 and @xmath73 . the complete prandtl shift is therefore @xmath74 then equations , , and become @xmath75 equation together with leads to @xmath76 this leads to several possible scalings for @xmath77 in terms of @xmath46 , namely * if @xmath67 verifies @xmath78 , we recover the ideal fluid equations ; * if @xmath67 satisfies @xmath79 , we obtain @xmath80 . this equation is explicitly solved and from the continuity of the stress tensor and the adherence on the ground we obtain @xmath81 . so we do not consider this trivial case ; * the last possibility is @xmath82 . this analysis preserves as many as possible terms in the equations , and hence most of the physical phenomena described by those terms . it is referred to as `` dominant balance '' . this is why in what follows , we consider the scaling @xmath83 with this hypothesis , we obtain the prandtl equations written in viscous layer variables : _ xu+_y v=0&[eccl1 ] + _ t u+u _ x u + v _ y u = -_x p - + _ y^2 u & [ eccl2 ] + _ yp=-&[eccl3 ] + u = v=0 y=0 & [ eccl4 ] the prandtl equations we obtain are in the same form as those obtained directly with a boundary layer scaling on the navier - stokes equations ( @xcite , ch . vii ) except for the topography term in . [ remgradp ] notice that implies that @xmath84 in does not depend on @xmath85 . up to now , we do not have enough boundary conditions for the viscous layer . the natural connection consists in assuming that the velocity at the `` top '' of the viscous layer , that is for @xmath86 , coincides with the ground velocity for the ideal fluid , defined by @xmath87 , where @xmath16 is the velocity vector for the ideal fluid . more precisely , we impose the required boundary condition at infinity as @xmath88 this is meaningful since in the prandtl shift @xmath89 and @xmath90 . note that in classical boundary layer theory the limit known as `` asymptotic matching '' is used : @xmath91 . therefore we conclude this paragraph by giving the behaviour of @xmath92 , which is an easy consequence of equations and . [ eque ] the horizontal component @xmath92 satisfies the following equation : @xmath93 the von krmn equation ( see schlichting @xcite ) expresses the defect of velocity between the ideal fluid and the viscous layer . a classical way to obtain such an equation consists in writing the prandtl equation , then introducing the velocity defect @xmath94 , and finally integrating it on the viscous layer . following @xcite we introduce the following two integrated quantities . [ def : delta1 ] let @xmath51 be the averaged velocity ( definition [ meanvel ] ) . we define * the displacement thickness @xmath6 , defined by @xmath95 * the momentum thickness @xmath96 , defined by @xmath97 the displacement thickness expresses the distance by which the ground should be displaced to obtain an ideal fluid with velocity @xmath92 with the flow rate @xmath98 . in the same way , the momentum thickness accounts for the loss of momentum in the viscous layer . a simple computation leads to the following expressions for these quantities : @xmath99 thus we can recover the classical formul in the boudary layer scaling @xmath100 ( see @xcite ) @xmath101 [ vkcond ] the displacement and momentum thicknesses are ruled by the so - called von krmn equation : @xmath102 where @xmath103 and @xmath104 denotes the parietal constraints : @xmath105 notice that in this single equation three new unknowns are actually introduced , namely @xmath6 , @xmath96 and @xmath104 . we postpone the presentation of the closure assumptions until the next section . first we notice that together with the prandtl shift rewrites @xmath106 where we have used to rewrite the right - hand side . the difference between this equation and ( [ eccl2 ] ) gives @xmath107 through ( [ eccl1 ] ) and ( [ eccl4 ] ) , we can rearrange the term @xmath108 . furthermore with the addition and withdrawal of @xmath109 , we get @xmath110 the last term in the left - hand side is rewritten @xmath111 now we integrate over @xmath85 the resulting equation between @xmath112 and @xmath113 using the boundary condition : @xmath114 the last three terms of the left - hand side can be rewritten as @xmath115 . so , using also definition , the equation becomes : @xmath116 finally , since @xmath92 is independent of @xmath85 , we have the relations @xmath117 and since all the unknowns are independent of @xmath85 and @xmath118 , we can drop as well the bars in the derivatives . with all this , we obtain equation ( [ equedelta ] ) . the von krmn equation together with the equation on @xmath92 give only a partial representation of the boundary layer , since they involve four unknowns , namely @xmath92 , @xmath6 , @xmath96 , and @xmath119 which represents the parietal forces . to proceed further towards an integrated model , we need a closure for this system of equations . by definition , the viscous layer is the layer where the velocity @xmath120 varies from @xmath112 on the floor to @xmath92 . therefore we introduce a profile function @xmath121 as well as a scaling factor @xmath122 , chosen in such a way that @xmath123 quantifies the physical thickness of the viscous layer . as a general rule we wish to have @xmath124 therefore we choose @xmath125 , and @xmath126 such that @xmath127 hence @xmath128 therefore choosing @xmath123 amounts to fixing the form factor @xmath129 defined in ( e.g. @xmath130 corresponds to @xmath131 ) . now we can readily obtain a relationship between @xmath6 and @xmath96 : @xmath132 finally , the parietal constraints can also be expressed in terms of the profile @xmath121 : @xmath133 the collection of all the preceding relations leads to put the von krmn equation ( [ equedelta ] ) on the following form : @xmath134 at this stage , several shapes can be used for the profile , including turbulent ones . as far as laminar profiles are concerned , we refer to ( * ? ? ? * ch.x ) for elements of comparisons between different profiles . according to the velocity profile similarity principle over a flat plane at zero - incidence @xcite , we shall assume that @xmath121 depends solely on the @xmath135 variable . table [ l - pprofiles ] contains the results for two polynomials approximations : the piecewise linear one and the parabolic one . these two profiles do not allow the observation of separation for decelerated flows as their coefficients are fixed . adaptive profiles allowing flow separation can also be seen , for instance in @xcite for the integral interacting boundary layer equations ( ibl ) . .piecewise linear and parabolic profiles for a finite layer @xmath136 . note that for an infinite layer , the solution would be blasius solution where @xmath137 and @xmath138 . see schlichting for other possible choices . [ cols="^,^,^,^",options="header " , ] , a parabolic profile up to a finite @xmath123 , a blasius like profile . right the same profiles are plotted for a same @xmath6 focusing on the displacement thickness .,title="fig:",width=188 ] , a parabolic profile up to a finite @xmath123 , a blasius like profile . right the same profiles are plotted for a same @xmath6 focusing on the displacement thickness .,title="fig:",width=188 ] [ h=1 ] * in the case where @xmath139 , the shape factor @xmath140 satisfies @xmath141 since @xmath142 . * the limit case @xmath143 enforces a discontinuity at the origin for @xmath126 : @xmath144 for @xmath145 , @xmath146 , thus formally leading to a singular friction term in the equation . we are now in position to obtain the extended model we are looking for . the first step consists in integrating the mass and momentum conservation equations , as in the usual way to derive the shallow water equations from navier - stokes , see for instance @xcite . the resulting mass equation is the same , but several differences occur for momentum . indeed , the parietal constraints arise in the right - hand side at order @xmath147 in @xmath3 , motivating a new closure for the momentum flux . the momentum equation thus depends on the unknowns @xmath6 and @xmath96 , so the system has to be coupled with the von krmn equation we derived above . we recall that the natural averaged velocity in viscous shallow water is @xmath148 . from the incompressibility equations we recover the usual balance equation on the water depth @xmath53 . the mass conservation in its integrated form is @xmath149 we integrate the conservation equation between @xmath13 and @xmath14 @xmath150 now from the no - slip boundary condition we get @xmath151 , and the kinematic one gives @xmath152 since the ground is independent of time , and we recognize ( [ consvmassint ] ) . we turn now to the momentum balance equation . the main point to notice at this stage is the appearance of the parietal constraint @xmath104 with prefactor @xmath3 . the integrated momentum conservation is written : @xmath153 where @xmath104 is defined by . we apply the same process as for the equation of mass conservation . recall that the momentum equation together with the scaling can be rewritten @xmath154 we integrate in @xmath9 , and recall that @xmath155 does not depend on @xmath9 , see remark [ remgradp ] . hence we get @xmath156 exchanging the time derivative and the integral , and integrating by parts in the last term of the left - hand side gives @xmath157_{f_b}^{\eta}-\int_{f_b}^{\eta}u\partial_yv\,dy = -h\partial_xp-\overline \delta^2\partial_yu|_{y = f_b}\ ] ] finally we use the mass conservation to replace the @xmath158 term , and the parietal force is estimated thanks to the prandtl shift , since @xmath159 ( see above ) . therefore we obtain @xmath160_{f_b}^{\eta } = -h\partial_xp-\overline \delta\tau,\ ] ] which together with the boundary conditions gives ( [ consvmomint ] ) . both equations and remain true for an ideal fluid as well , since the terms at @xmath161 cancel thanks to the non penetration boundary condition . the viscous layer theory leads to a friction term in a laminar form but only over the height @xmath162 . also the order term is only @xmath163 whereas for the poiseuille - shallow water ( see equation ( [ pois ] ) ) it is at order @xmath164 . at this point we are faced with two equations and which are very similar to the usual shallow water equations , except that the friction term is not given explicitly in terms of @xmath51 and/or @xmath53 . a consistent expression with the model is obtained through the velocity profile , see above . the momentum equation remains unclosed : we need to give a closure for @xmath165 . the classical approximation for the shallow water system is based on the fact that for an ideal fluid with a low depth , we have @xmath166 as we said in section [ subsec : sw ] . relation does not give any closure for the momentum flux , it merely emphasizes that it can be entirely completed by the choice of @xmath96 . in this respect , we prove now that this expression for the flux is the most convenient , since it takes into account the effect of the viscous layer , and in some sense independantly of the choice for @xmath96 . the complete closure can be achieved by the study of the viscous layer , as we did above in section [ sec : viscprof ] . more precisely , we have the following two results , which clarify the role of the von krmn equation . the starting point is the system of three integrated equations : @xmath167 where @xmath168 is the momentum flux for which we seek a closure . we evidence now the fact that a convenient definition of @xmath168 allows to recover the von krmn equation from this system of integrated equations . [ close - flux ] let @xmath169 be solution to , and . assume @xmath6 is defined by . then @xmath6 and @xmath96 solve the von krmn equation if and only if there holds @xmath170 we start from the von krmn equation and introduce to obtain @xmath171 to this equation we add , the parietal term disappears , leading to @xmath172 developping the time derivative and simplifying with we obtain @xmath173 finally , we use to eliminate the time derivative , regroup terms and get @xmath174 so that , up to a constant which can be taken equal to zero by considering that the flux is zero when the velocity is zero , we have @xmath175 which together with gives precisely . conversely , we consider and use successfully the mass and momentum balance equations @xmath176 noting that @xmath177 we recover as required the von krmn equation . in the above proposition , we only use the three equations ( [ eq : mass])-([eq : fp ] ) and the physical definition ( [ tildeufctue ] ) of @xmath6 . however if we forget the physical definition , we obtain another formulation using the von krmn equation ( [ equedelta ] ) together with equations ( [ eq : mass])-([eq : fp ] ) . this approach gives back the physical definition in the sense of characteristics as explained in the next proposition . [ caract - delta ] let @xmath168 be defined by , and @xmath178 be a ( smooth ) solution to the system , , , together with the von krmn equation . denote by @xmath179 the thickness obtained using . then @xmath180 hence , the error is constant along the characteristics of the ideal fluid . in particular , if initially @xmath181 then it is true for all times . from @xmath182 now using the von krmn equation to eliminate @xmath96 , we obtain @xmath183 which is the desired result . in summary , putting together the results of proposition [ close - flux ] , the closure formulas and using the velocity profile in the viscous layer , and the hydrostatic law for the pressure , we obtain a system of three partial differential equations , repectively for the mass conservation , the momentum conservation , and the behaviour of the prefect fluid , namely _ th+_x(hu ) = 0&[finalmass ] + _ t(hu)+_x(+(h-|_1(1+))u_e^2 ) = --(0)&[finalmom ] + _ tu_e+u_e_xu_e = -(_xh+f_b)&[finalideal ] propositions [ close - flux ] and [ caract - delta ] show that there are two equivalent possibilities to compute @xmath6 in order to close the system . the first one consists in adding the algebraic relation , which gives somewhat an equation of state : @xmath184 the second one makes use of the von krmn equation , so that @xmath6 appears as another conserved quantity @xmath185 the relations between @xmath53 , @xmath51 , @xmath92 and @xmath6 through the displacement thickness and directly imply the following expressions for the momentum flux @xmath186 the last one , which can be rewritten as a ratio of fluxes : @xmath187 enhances the extension of the classical shallow water model without friction : taking abruptly @xmath188 in these equations , the first two equations are decoupled from the last one and we recover the classical momentum equation ( with a ratio equal to one ) . on the other hand , as soon as viscosity effects arise , that is @xmath189 , we have not only the friction term on the right - hand side but also a correction of order one in @xmath3 to the hydrostatic pressure . this is consistent with the order of magnitude of the friction term . in @xcite a similar correction in the flux of the shallow water system is proposed to improve the study of roll - waves . in the following , we choose to compute the model with the equations ( [ finalmass ] ) , ( [ finalmom ] ) , ( [ finalideal ] ) and ( [ vkprof ] ) instead of ( [ finalmass ] ) , ( [ finalmom ] ) , ( [ finalideal ] ) and ( [ finalcloseflux ] ) . indeed the equation ( [ vkprof ] ) gives more information about the evolution of the displacement thickness ( see part [ analsol ] ) . furthermore this choice is more stable for the illustrations in section [ num ] . from now , we limit ourselves to a constant shape for the velocity profile in the viscous layer ( piecewise linear or parabolic , see table [ l - pprofiles ] ) which expresses by the form parameter @xmath140 and the shear stress @xmath190 constants in the equations ( [ finalmom ] ) and ( [ vkprof ] ) . [ aero ] an aerodynamic type approach would not conserve the equation ( [ finalmom ] ) but instead ( [ finalmass ] ) , ( [ finalideal ] ) , ( [ finalcloseflux ] ) and ( [ vkprof ] ) . equation ( [ finalmom ] ) is the important point of this work which combines the aerodynamic approach and the shallow water one . furthermore , this approach is known as `` interactive boundary layer '' or `` viscous inviscid interaction '' in aerodynamics @xcite . we present in this section several analytical solutions which can be used for testing the numerical methods . first we provide two solutions to the von krmn equation , for short and long times , then a solution to the complete linearized problem . the few simulations of the esw model in this part are obtained with the numerical scheme presented in section [ num ] . in this part , we consider the von krmn equation in the following configuration , also studied by stewartson ( * ? ? ? 3 ) : * @xmath92 is constant , * @xmath191 , * @xmath140 , @xmath190 are constant . consequently the von krmn equation ( [ vkprof ] ) can be rewritten @xmath192 and it has to be complemented with the following initial and boundary value conditions : @xmath193 the solution is readily obtained by the method of characteristics : @xmath194 it is clear on this formula that for @xmath195 large enough the initial value of @xmath6 has no influence on the solution . figure [ blasiustime ] illustrates the creation and development of the viscous layer , with initial and boundary values @xmath196 and @xmath197 . when time evolves , @xmath198 captures indeed the blasius solution @xmath199 . notice that this solution of the von krmn equation with constant @xmath92 creates an unbounded viscous layer , without regards of the limitation of the water depth . hence it is clearly physically not valid for large @xmath8 or @xmath195 . in this context , a final time @xmath200 is enough to get the profile all over the domain @xmath201 $ ] . we turn now to the study of the von krmn equation with the aerodynamic point of view ( see remark [ aero ] ) for large times . more precisely we assume that @xmath6 is stationary , similar to the interactive boundary layer in @xcite . then equation becomes @xmath202 if we assume in addition that both @xmath53 and @xmath51 are constant , combined with the relation @xmath203 , we recover @xmath204 then , the large time von krmn equation is an ordinary differential equation in @xmath6 : @xmath205}.\ ] ] equation ( [ odesatu ] ) is consistent with the the blasius solution ( [ bla / ray ] ) when @xmath3 goes to @xmath112 and @xmath206 . this corresponds to a small value of the effective displacement thickness @xmath207 , hence to small values of @xmath8 . figure [ satut ] displays the three solutions we have at hand , and clearly evidences this behaviour . on the other hand , equation exhibits a stationary solution @xmath208 , and we observe indeed a saturation for large values of @xmath8 , see figure [ satuode ] . notice also that the smaller @xmath3 is , the larger is the domain size required to reach the stationary state . hence computing the stationary solution with the complete esw model may be very difficult . moreover it is not clear that actual solutions to the complete esw system satisfy @xmath53 and @xmath51 constant for large @xmath8 , so that , even if such saturation behaviour is physically plausible , this solution definitely might not be caught by the extended model , which can explain the difference in figure [ satut ] . the previous section evidences a steady solution for the displacement thickness for any parameter @xmath163 with the assumption that @xmath53 and @xmath51 are constant . in this part , we look for a steady form for all the unknowns on a flat ground . this steady form , depending on @xmath163 , is an explicit non constant solution . it is actually a solution to the linearized system of equations , over a particular steady state . for any unknown function @xmath209 , we look for an expansion @xmath210 , and drop all terms of order greater than 2 . we start from the steady equations , in their aerodynamics - like form ( see remark [ aero ] ) : @xmath211 first we notice the decomposition of the two following terms : @xmath212 @xmath213 this provides us with the required steady state . 1 . thanks to the relation ( [ relation ] ) , @xmath214 ; 2 . the equations ( [ stcont ] ) and ( [ stue ] ) , written only with @xmath215 and @xmath216 , imply that @xmath215 and @xmath216 are constants ( and so @xmath217 too ) ; 3 . consequently , equation ( [ stvk ] ) becomes @xmath218 which readily gives @xmath219 : @xmath220 notice that this expression is valid only if @xmath221 and @xmath222 , which is ensured by the choice of the profile @xmath121 . we turn now to the order 1 perturbation of the above steady state . 1 . with the equation ( [ stue ] ) , we get a relation between @xmath223 and @xmath224 : @xmath225 2 . the combination of the relation ( [ relation ] ) and the equation ( [ stcont ] ) we have @xmath226 introducing in this expression the relation ( [ firstorderhelp ] ) , we get @xmath227 + integrating in @xmath8 , we get an explicit form for @xmath223 in terms of @xmath219 : @xmath228 where @xmath229 is the local froude number defined by @xmath230 . thanks to the equation ( [ firstorderhelp ] ) we have @xmath231 as @xmath232 , we finally get @xmath233 4 . exploiting the equation ( [ stcont ] ) with @xmath53 and @xmath51 , we obtain from the above relation @xmath234 so , we have state a linearized non constant solution for the steady system ( [ stcont])-([stue ] ) : @xmath235 the aim of this last part is to illustrate in a simple way the behaviours of the different variables of the esw system . we are aware that more accurate analysis is mandatory to study the system . in a first step , we briefly present the numerical scheme we use . a second step will be devoted to a comparison between the esw solutions and those of the classical shallow water system ( sw ) . finally we will observe the evolution of the displacement thickness and the friction term with different parameters of the calculation . we recall and fix some notations in the general framework of a system of nonlinear conservation laws in the form @xmath236 where @xmath237 is the flux and @xmath238 the source term . recall that the aim of this part is not to establish a sharp discretization , but merely to illustrate some interesting features of the esw system . we apply a standard finite volume method with time splitting for the source terms . as regard the conservative equation without the source term we use a first - order explicit three points scheme ( @xcite ) : @xmath239 where @xmath240 . the choice of the numerical flux @xmath241 determines the scheme . we choose here a hll scheme . if we denote by @xmath242 the eigenvalues of @xmath243 , the flux @xmath241 , in our notations , is defined by : @xmath244 with @xmath245 and @xmath246 . the pressure term in the equation ( [ finalideal ] ) contains a spatial derivate of the water height @xmath53 . to fully respect the formulation above , this term should be included in @xmath243 . however this leads to a matrix whose eigenvalues are not known and so this complicates the computation . in the following we treat all the pressure term in equation ( [ finalideal ] ) as a source term . with this choice and the flux presented in ( [ usquarefcthu ] ) , @xmath247 , @xmath243 is a block diagonal matrix whose four eigenvalues can be computed and are comparable with the classical eigenvalues of the shallow water system : @xmath248 the hll scheme could be computed in this state . nevertheless , we have not the means to propose an equivalence of the classical hydrostatic reconstruction in order to match the usual simulations when @xmath249 . as the block diagonal form inspires a possible splitting of the equations without denaturing the global bahavior , we choose to do the computation of the flux in several steps all the while conserving the information of the eigenvalues as regards the speeds used in the hll scheme . 1 . the ideal velocity is firstly computed thanks to the equation ( [ finalideal ] ) with only a centered difference for the source term : @xmath250.\ ] ] where @xmath251 is linked to @xmath252 2 . let @xmath253 . the von krmn equation ( [ vkprof ] ) gives : @xmath254 with @xmath255 associated to @xmath256 3 . let @xmath257 . the two equations of conservation ( [ finalmass])-([finalmom ] ) are rewritten in the form : @xmath258 where @xmath259 . firstly , we concentrate only on the @xmath260 part which gives the classical shallow water equations and apply the ( first - order ) hydrostatic reconstruction . as the method is fully described in the literature ( see e.g. @xcite ) and is not specific to our case , we just set we construct @xmath261 for the shallow water part thanks to this way . the last step concerns only @xmath98 : @xmath262-\delta t \overline \delta \frac{((u_e)_i^{n+1})^2\alpha_2\varphi'(0)}{w_i^{n+1}}\ ] ] where @xmath241 is obtained from @xmath263 . to conclude the presentation of the used code , we precise that we take an uniformly discretized space @xmath264 $ ] where the cell length @xmath265 is got thanks to @xmath168 the number of points for the space mesh . as regards the time discretization , we use an adaptive cell length @xmath266 calculated at each time step with a cfl condition @xmath267 . the speed propagation @xmath268 is estimated with the four eigenvalues of the block diagonal matrix . as said previously , the velocity profile is fixed in the computations . the choice ( linear or parabolic ) is indicated in the captions of the figures . the quantities are made non dimensional with say meters , seconds ... so that @xmath269 . examples , except those with transcritical phenomena , are subcritical ( @xmath270 ) . as we saw before , theoretically the esw model includes the shallow water model if the effective reynolds number @xmath271 is infinite . so the first numerical verification is to check the concordance between the two models . to illustrate this point , we present three cases : a subcritical flow over a gaussian bump , and two transcritical flows , one without shock and one with shock @xcite . notice that analytic solutions exist for these three academic examples , see @xcite . however , we choose here to discretize the shallow water model with the same finite volume scheme as the esw system . in this case , the ground is a gaussian bump @xmath272 . initial conditions are taken equal to @xmath273 and for the boundary conditions @xmath274 on figure [ h - u_sv ] , we plotted the results for the shallow water model without friction and for the esw system . the simulation for @xmath275 answers our expectation since the shallow water curves are the same as those of esw for the heights and velocities . therefore we can observe that for a lower reynolds number , @xmath276 , the shallow water velocity is higher than the esw velocity . since a lower reynolds number expresses a larger viscosity , hence higher friction , this phenomenon is expected : the flow slows down . we complement the figures for the two reynolds number by simulations of the displacement thickness @xmath6 . + for the two transcritical flows , the studied topography is the one studied in @xcite : @xmath277\indic_{[8,12]}(x ) \qquad x\in [ 0,25].\ ] ] usually the initial conditions are a lake at rest . due to the naive computation of the esw , @xmath278 and @xmath279 is not allowed . for initial conditions of @xmath53 and @xmath51 we take the usual ones as in @xcite , and we choose small values for @xmath92 and @xmath6 . in each case ( see figure [ trans ] for the transcritical without shock case and figure [ transs ] for the transcritical with shock one ) , the simulations give a behavior of the surface @xmath14 and the velocity @xmath51 not disturbed on the bump for the two reynolds number in comparison with the shallow water curves but some changes on the flat parts of the ground appear for the smallest reynolds number presented . we find also this result for the displacement thickness @xmath6 . * * transcritical flow without shock * + the transcritical flow without shock ( see figure [ trans ] ) illustrates a subcritical upstream flow which becomes supercritical at the top of the bump . the initial conditions are : @xmath280 + for the boundary conditions , we choose : @xmath281 + + * * transcritical flow with shock . * in this part , we keep the same parabolic bump but we change the flow conditions so that the flow becomes supercritical on the top of the bump and then fluvial again after a hydraulic jump . compared to computations using the shallow water model , we recover the hydraulic jump at a position which is consistent with the classical sw model , but slightly depends on the value of @xmath77 , see figure [ transs ] , top . however , the stabilization of @xmath6 after the shock is not ensured for `` large '' values of @xmath77 , see figure [ transs ] , bottom . actually , we observe some time instability after the hydraulic jump for these values of @xmath77 , see figure [ transstime ] . this can question both the model and the numerical resolution . indeed concerning numerics , we are faced for @xmath6 with an equation where the transport operator contains a discontinuous velocity field , which can induce stability problems . on the other hand , we also reach the limits of the model itself since most likely some recirculation phenomena occur after the jump , and this is not accounted for in this model . + initial conditions : @xmath282 + for the boundary conditions , we choose : @xmath283 in part [ thstatsol ] , we established , on a flat ground , a stationary solution for our system linearized with respect to @xmath163 : @xmath284 where @xmath215 , @xmath217 , @xmath216 are constants and @xmath214 . to be satisfying , the numerical scheme presented in this section must be consistent with it . on the figure [ sollin - delta1 ] , we can observe the analytic linearized solution and the computed solutions for two values of @xmath163 . the entrance boundary condition is not sophisticated enough and so there is an initial difference which appears . nevertheless , we can observe a very good similitude for the value @xmath285 . on the other way , for @xmath286 the computed solution gets away from the analytic one with especially a different behaviour in space . this observation is consistent with the steady state for the von krmn equation presented in the section [ ssvke ] . the concordance observed between the shallow water model without friction and the esw system , we will now observe more precisely the behavior of the displacement thickness as well of the friction term in response to a parametrized bump . for the following simulations , we choose for the ground a gaussian profile : @xmath287 the initial conditions , by default , are on the form @xmath288 . the considered flow is subcritical , and as regards the boundary conditions we consider @xmath289 the following illustrations place themselves in case where the displacement thickness is developing . even if a gaussian bump locates in the middle of the simulation , the behavior for the displacement thickness stays the blasius solution [ bla / ray ] which will be more or less disturbed by the bump as seen in the figure [ global ] . this global growing shape may interfere with the local observations of the effects ( especially for the friction term @xmath104 because of large first values ) . to rectify this , we will present as well some figures rescaled by the blasius solution . due to the monoticity of the rescaling functions , the local extrema observed will be moved downstream in comparison with the displacement thickness or friction term without the rescaling . the plotted bump in the following figures is just an indicative one . the width is respected but not the amplitude nor the ordinate . the shallow water theory already allows to treat a viscosity effect thanks to added terms in the momentum equation ( see [ subsec : sw ] ) . however they do not fit with the expectations , that is to say a friction more significant just before the top of the bump as in @xcite . in this part , we will observe more in details the local extrema observed in the previous figure [ global ] . then we will compare with the laminar friction in the sw model . the figure [ fbump ] is only a zoom of figure [ global ] where the extrema are evidenced . we can observe that the local minimum for the displacement thickness is located before the top of the bump . even if the abscissa of the maximum of @xmath104 is moved on the right ( due to the growing of @xmath92 which compensate the decrease of @xmath290 ) , the local maximum is reached upstream the top of the bump . the comparison between sw with friction and esw models brings out an issue as the level of friction are different . the poiseuille - shallow water ( see section [ subsec : sw ] ) gives a friction term in @xmath291 whereas the esw one is in @xmath163 . the difference does not allow to plot the friction terms in the same window . also , in place of @xmath292 , we choose to take the laminar friction term for the shallow water model on the form @xmath293 where the coefficient @xmath294 is adapted by hand . we fix @xmath295 to fit the velocities between sw and esw simulations when @xmath296 ( recall that esw friction is @xmath297 ) . the consequence on the friction terms is presented in the figure [ frict ] . the laminar sw term has its maximum at the top of the bump which is not physically expected . due to the only dependence in the velocity and the height ( the two unknowns of the shallow water equations ) , other patterns would have the same defect . the parameter @xmath298 controls the thickness of the viscous layer . it is the key point of the esw model . also , we will observe here the influence of its value for the shape of the displacement thickness @xmath6 and the friction @xmath104 . the physical values are @xmath162 and @xmath297 , but to compare them we plot only @xmath6 and @xmath104 . on the flat ground before the bump , the displacement thickness is nearly conserved as we can see in the figure [ deltare ] , excepted for @xmath299 which is explained by the limitation of the scheme . around the bump a net difference appears even if it is more noticeable on the right of the bump . anyway @xmath6 gets down with the growing of @xmath163 but conserves its shape and the local minimum abscissa for @xmath6 is not influenced by @xmath163 . this statement holds as far as @xmath163 is not big since @xmath299 begins to show a significantly disruption from the beginning . taking a huger value for @xmath163 leads to great instabilities and the scheme not hold . + * bump amplitude . * not very surprisingly , the bump amplitude has a direct influence on the amplitude of variation of @xmath6 and @xmath104 . on the one hand , as it is seen in figure [ ampb ] , the smaller the amplitude , the closer @xmath6 is to the blasius solution . on the other hand , it can be observed that the solution departs from the blasius one after the bump when the amplitude increases . this phenomenon becomes important for large values of the amplitude ( see figure [ fig : boum ] ) , and might be bound with the boundary layer separation when the downstream slope of the bump becomes too high ( see for instance @xcite ) . we can not , with the chosen profiles here compute boundary layer separation . the figure [ ampb ] illustrates the limitation of the local extrema observation due to the square root shape ( or inverse of a square root for the friction term @xmath104 ) . when the amplitude tends to zero , an inflection point stays but it is difficult to locate the precise abscissa . even if the rescaled outline moves the abscissa of the extrema to the right , the figure [ ampbn ] allows to observe quantitatively the amplitude influence on @xmath6 and @xmath104 . the variation of the bump amplitude brings out another property for the displacement thickness . as far the amplitude is sufficient , the displacement thickness catches up always the same curve around the bump and that , for different initial values . in all the previous examples , @xmath6 was initialized by @xmath300 . this choice of a small value for @xmath6 was intended to observe the development of the viscous layer . starting with a constant value not morally equal to zero is interpreted as an established viscous flow in entry . in the figure [ entree ] , this initial value varies with a given bump . even if the blasius behavior described by the equation [ bla / ray ] begins at different values , around the bump the four curves are gathered . so there is an independence of the solution around the bump with respect to the initial value for the displacement thickness , even if the phenomenon can not be observed for every entrance and is limited by the bump amplitude choice . * bump width . * the bump width is quantified by the standard deviation @xmath301 in the gaussian profile . increasing the width of the bump increases as well the extent of the response of both @xmath6 and @xmath104 , see figures [ width ] and [ widthtau ] . the amplitude is also slightly decreased . but the most interesting phenomenon is that the local maximum of the friction moves upstream from the top of the bump . nevertheless , since the response of the displacement thickness spreads with the bump width , a too much broad bump leads only to an inflection point . rescaling functions in figures [ width ] and [ widthtau ] are more appropriated to observe the phenomenon . as we have seen in the section [ sec : viscprof ] , a key role for the esw model is played by the velocity profile @xmath121 . several choices can be made for a better answer when the viscous velocity significantly varies . a refined way would be an adjustable profile to react faster over a bump ( see @xcite ) . once again , in this paper we will restrict to the easiest computations : the linear and parabolic profiles . table [ l - pprofiles ] references the resultant values for @xmath129 , @xmath140 and @xmath190 the three parameters encountered in the system . in the following figures [ deltaprof ] and [ frictprof ] , we plotted the displacement thickness @xmath6 and the friction term @xmath104 for these two profiles . the parabolic profile presents an advanced reaction , in relation to the top of the bump , compared to the linear profile . the phenomenon is clearer for the displacement thickness than the friction term . this difference is explained by the expression of @xmath302 . @xmath190 and @xmath129 are constant for these profiles but @xmath92 , which is growing in this zone , plays a determinant role in the position of the local maximum . this paper concerns a shallow water model with an improved description in the long wave and large reynolds limit of the flow and the parietal friction . this model consists of a system of four equations ( [ finalmass ] ) , ( [ finalmom ] ) , ( [ finalideal ] ) and ( [ vkprof ] ) : two are similar to shallow water system ( mass and momentum conservation ) with a supplementary term describing momentum flux , one ideal fluid equation and the von krmn equation to describe the viscous layer development . the boundary layer acts as a new topography for the model ( eq . [ finalcloseflux ] ) . this model is an example of `` interactive boundary layer '' or `` viscous inviscid interaction '' . the flux in the momentum equation contains not only the hydrostatic pressure term @xmath303 but an additional correction term of the same order of magnitude as the friction . the friction term is no longer a combination of velocity and depth ( as in usual laws such as darcy or manning ) but the result of a viscous boundary layer like approach ( eq . ) . as it is , it actually depends on the topography , as evidenced by the examples provided in the last part of the work . in particular its maximum is reached before the summit of a bump . this effect is limited here , and it is a limitation of this work , because the velocity profiles used lead to constant shape and friction coefficients . as a consequence , possible boundary layer separation with recirculation downstream of the bump can not be observed . imposing a family of adjustable profiles will be the next step , this should enhance the effects of topography . results from @xcite in case of flows in pipes , and preliminary comparisons with multilayer shallow water @xcite make us confident . the main drawback of this model is the viscous layer / ideal fluid decomposition , which forbids the viscous layer to fill all the water depth far downstream of the bump . extra modelling in this direction and more examples have now to be worked out and tested , an effort has to be done as well to improve the numerical method . finally , even thoug this approach is restricted here to laminar flows , the ideas developed here can be extended with little modifications to mean turbulent profiles . e. audusse , f. bouchut , m .- o . bristeau , r. klein & b. perthame , _ a fast and stable well - balanced scheme with hydrostatic reconstruction for shallow water flows _ , siam journal on scientific computing , * 25 * ( 2004 ) , n@xmath305 6 , 2050 - 2062 e. audusse , m .- o . bristeau , b. perthame & j. sainte - marie . _ a multilayer saint - venant system with mass exchanges for shallow water flows . derivation and numerical validation _ , esaim : m2an , * 45 * ( 2011 ) , 169 - 200 e. berger , m. deverge , a. hirschberg , p .- y . lagre & c. vilain , _ characterization of the pressure drop in a 2d symmetrical pipe : some asymptotical , , numerical and experimental comparisons _ , zamm : z. angew . * 85 * ( 2005 ) , n@xmath305 2 , 141 - 146 j. burguete , & p. garca - navarro , _ implicit schemes with large time step for non - linear equations : application to river flow hydraulics _ , international journal for numerical methods in fluids , * 46 * ( 2004 ) , n@xmath305 6 , 607 - 636 o. delestre , s. cordier , f. james & f. darboux . _ simulation of rain - water overland - flow_. in e. tadmor , j .- g . liu and a. tzavaras , editors , proceedings of the @xmath306 international conference on hyperbolic problems , volume 67 of proceedings of symposia in applied mathematics , pages 537 - 546 , university of maryland , college park ( usa ) , 2009 . o. delestre , c. lucas , p .- a . ksinant , f. darboux , ch . laguerre , t .- vo , f. james & s. cordier , _ swashes : a compilation of shallow water analytic solutions for hydraulic and environmental studies _ , int . j. numer . fluids ( 2012 ) m. esteves , x. faucher , s. galle & m. vauclin . _ overland flow and infiltration modelling for small plots during unsteady rain : numerical results versus observed values_. journal of hydrology , 228(3 - 4 ) : 265 - 282 , 2000 . lagre & s. lorthois . _ the rns / prandtl equations and their link with other asymptotic descriptions . applications to the computation of the maximum value of the wall shear stress in a pipe_. int . j. eng sci . , volume 43/3 - 4 pp352 - 378 , 2005 . l. tatard , o. planchon , j. wainwright , g. nord , d. favis - mortlock , n. silvera , o. ribolzi , m. esteves & chi - hua huang . _ measurement and modelling of high - resolution flow - velocity data under simulated rainfall on a low - slope sandy soil_. journal of hydrology , 348(1 - 2 ) : 1 - 12 , 2008 .

the notions of convexity and convex polytopes are introduced in the setting of tropical geometry . combinatorial types of tropical polytopes are shown to be in bijection with regular triangulations of products of two simplices . applications to phylogenetic trees are discussed .

math0308254

the _ tropical semiring _ @xmath0 is the set of real numbers with the arithmetic operations of _ tropical addition _ , which is taking the minimum of two numbers , and _ tropical multiplication _ , which is ordinary addition . thus the two arithmetic operations are defined as follows : @xmath1 the @xmath2-dimensional space @xmath3 is a semimodule over the tropical semiring , with tropical addition @xmath4 and tropical scalar multiplication @xmath5 the semiring @xmath0 and its semimodule @xmath3 obey the usual distributive and associative laws . the purpose of this paper is to propose a tropical theory of convex polytopes . convexity in arbitrary idempotent semimodules was introduced by cohen , gaubert and quadrat @xcite and litvinov , maslov and shpiz @xcite . some of our results ( such as theorem [ polar ] and propositions [ intersectpolyt ] and [ vertexunique ] ) are known in a different guise in idempotent analysis . our objective is to provide a combinatorial approach to convexity in the tropical semiring which is consistent with the recent developments in tropical algebraic geometry ( see @xcite , @xcite , @xcite ) . the connection to tropical methods in representation theory ( see @xcite , @xcite ) is less clear and deserves further study . there are many notions of discrete convexity in the computational geometry literature , but none of them seems to be quite like tropical convexity . for instance , the notion of _ directional convexity _ studied by matouek @xcite has similar features but it is different and much harder to compute with . a subset @xmath6 of @xmath7 is called _ tropically convex _ if the set @xmath6 contains the point @xmath8 for all @xmath9 and all @xmath10 . the _ tropical convex hull _ of a given subset @xmath11 is the smallest tropically convex subset of @xmath3 which contains @xmath12 . we shall see in proposition [ tstc ] that the tropical convex hull of @xmath12 coincides with the set of all tropical linear combinations @xmath13 any tropically convex subset @xmath6 of @xmath3 is closed under tropical scalar multiplication , @xmath14 . in other words , if @xmath15 then @xmath16 for all @xmath17 . we will therefore identify the tropically convex set @xmath6 with its image in the @xmath18-dimensional _ tropical projective space _ @xmath19 basic properties of ( tropically ) convex subsets in @xmath20 will be presented in section 2 . in section 3 we introduce tropical polytopes and study their combinatorial structure . a _ tropical polytope _ is the tropical convex hull of a finite subset @xmath12 in @xmath20 . every tropical polytope is a finite union of convex polytopes in the usual sense : given a set @xmath21 , their convex hull has a natural decomposition as a polyhedral complex , which we call the _ tropical complex _ generated by @xmath12 . the following main result will be proved in section 4 : [ main ] the combinatorial types of tropical complexes generated by a set of @xmath22 vertices in @xmath20 are in natural bijection with the regular polyhedral subdivisions of the product of two simplices @xmath23 . this implies a remarkable duality between tropical @xmath18-polytopes with @xmath22 vertices and tropical @xmath24-polytopes with @xmath2 vertices . another consequence of theorem [ main ] is a formula for the @xmath25-vector of a generic tropical complex . in section 5 we discuss applications of tropical convexity to phylogenetic analysis , extending known results on injective hulls of finite metric spaces ( cf . @xcite , @xcite , @xcite and @xcite ) . we begin with two pictures of tropical convex sets in the tropical plane @xmath26 . a point @xmath27 is represented by drawing the point with coordinates @xmath28 in the plane of the paper . the triangle on the left hand side in figure [ tropconv ] is tropically convex , but it is not a tropical polytope because it is not the tropical convex hull of finitely many points . the thick edges indicate two tropical line segments . the picture on the right hand side is a _ tropical triangle _ , namely , it is the tropical convex hull of the three points @xmath29 and @xmath30 in the tropical plane @xmath26 . the thick edges represent the tropical segments connecting any two of these three points . we next show that tropical convex sets enjoy many of the features of ordinary convex sets . [ lotsofstuff ] the intersection of two tropically convex sets in @xmath3 or in @xmath20 is tropically convex . the projection of a tropically convex set onto a coordinate hyperplane is tropically convex . the ordinary hyperplane @xmath31 is tropically convex , and the projection map from this hyperplane to @xmath32 given by eliminating @xmath33 is an isomorphism of tropical semimodules . tropically convex sets are contractible spaces . the cartesian product of two tropically convex sets is tropically convex . we prove the statements in the order given . if @xmath6 and @xmath34 are tropically convex , then for any two points @xmath35 , both @xmath6 and @xmath34 contain the tropical line segment between @xmath36 and @xmath37 , and consequently so does @xmath38 . therefore @xmath38 is tropically convex by definition . suppose @xmath6 is a tropically convex set in @xmath3 . we wish to show that the image of @xmath6 under the coordinate projection @xmath39 is a tropically convex subset of @xmath32 . if @xmath40 then we have the obvious identity @xmath41 this means that @xmath42 is a homomorphism of tropical semimodules . therefore , if @xmath6 contains the tropical line segment between @xmath36 and @xmath37 , then @xmath43 contains the tropical line segment between @xmath44 and @xmath45 and hence is tropically convex . the same holds for the induced map @xmath46 . most ordinary hyperplanes in @xmath3 are not tropically convex , but we are claiming that hyperplanes of the special form @xmath47 are tropically convex . if @xmath36 and @xmath37 lie in that hyperplane then @xmath48 . this last equation implies the following identity for any real numbers @xmath49 : @xmath50 hence the tropical line segment between @xmath36 and @xmath37 also lies in the hyperplane @xmath51 . consider the map from @xmath51 to @xmath32 given by deleting the @xmath52-th coordinate . this map is injective : if two points differ in the @xmath33 coordinate they must also differ in the @xmath53 coordinate . it is clearly surjective because we can recover an @xmath52-th coordinate by setting @xmath54 . hence this map is an isomorphism of @xmath55-vector spaces and it is also an isomorphism of @xmath0-semimodules . let @xmath6 be a tropically convex set in @xmath3 or @xmath20 . consider the family of hyperplanes @xmath56 for @xmath57 . we know that the intersection @xmath58 is tropically convex , and isomorphic to its ( convex ) image under the map deleting the first coordinate . this image is contractible by induction on the dimension @xmath2 of the ambient space . therefore , @xmath58 is contractible . the result then follows from the topological result that if @xmath6 is connected , which all tropically convex sets obviously are , and if @xmath58 is contractible for each @xmath59 , then @xmath6 itself is also contractible . suppose that @xmath60 and @xmath61 are tropically convex . our last assertion states that @xmath62 is a tropically convex subset of @xmath63 . take any @xmath64 and @xmath65 in @xmath62 and @xmath49 . then @xmath66 lies in @xmath67 since @xmath6 and @xmath34 are tropically convex . we next give a more precise description of what tropical line segments look like . [ slopes ] the tropical line segment between two points @xmath36 and @xmath37 in @xmath20 is the concatenation of at most @xmath68 ordinary line segments . the slope of each line segment is a zero - one vector . after relabeling the coordinates of @xmath69 and @xmath70 , we may assume @xmath71 the following points lie in the given order on the tropical segment between @xmath36 and @xmath37 : @xmath72 between any two consecutive points , the tropical line segment agrees with the ordinary line segment , which has slope @xmath73 . hence the tropical line segment between @xmath36 and @xmath37 is the concatenation of at most @xmath68 ordinary line segments , one for each strict inequality in ( [ sortthedifferences ] ) . this description of tropical segments shows an important feature of tropical polytopes : their edges use a limited set of directions . the following result characterizes the _ tropical convex hull_. [ tstc ] the smallest tropically convex subset of @xmath20 which contains a given set @xmath12 coincides with the set of all tropical linear combinations ( [ alltropcomb ] ) . we denote this set by @xmath74 . let @xmath75 be the point in ( [ alltropcomb ] ) . if @xmath76 then @xmath36 is clearly in the tropical convex hull of @xmath12 . if @xmath77 then we write @xmath78 . the parenthesized vector lies the tropical convex hull , by induction on @xmath22 , and hence so does @xmath36 . for the converse , consider any two tropical linear combinations @xmath79 and @xmath80 . by the distributive law , @xmath81 is also a tropical linear combination of @xmath82 . hence the set of all tropical linear combinations of @xmath12 is tropically convex , so it contains the tropical convex hull of @xmath12 . if @xmath12 is a finite subset of @xmath20 then @xmath83 is a _ tropical polytope_. in figure [ throughex ] we see three small examples of tropical polytopes . the first and second are tropical convex hulls of three points in @xmath26 . the third tropical polytope lies in @xmath84 and is the union of three squares . one of the basic results in the usual theory of convex polytopes is carathodory s theorem . this theorem holds in the tropical setting . [ carat ] if @xmath36 is in the tropical convex hull of a set of @xmath22 points @xmath85 in @xmath20 , then @xmath36 is in the tropical convex hull of at most @xmath2 of them . let @xmath75 and suppose @xmath86 . for each coordinate @xmath87 , there exists an index @xmath88 such that @xmath89 . take a subset @xmath90 of @xmath91 composed of one such @xmath52 for each @xmath92 . then we also have @xmath93 , where @xmath94 . the basic theory of tropical linear subspaces in @xmath20 was developed in @xcite and @xcite . recall that the _ tropical hyperplane _ defined by a tropical linear form @xmath95 consists of all points @xmath96 in @xmath20 such that the following holds ( in ordinary arithmetic ) : @xmath97 just like in ordinary geometry , hyperplanes are convex sets : [ hypconv ] tropical hyperplanes in @xmath20 are tropically convex . let @xmath98 be the hyperplane defined by ( [ hyperplane ] ) . suppose that @xmath36 and @xmath37 lie in @xmath98 and consider any tropical linear combination @xmath99 . let @xmath52 be an index which minimizes @xmath100 . we need to show that this minimum is attained at least twice . by definition , @xmath101 is equal to either @xmath102 or @xmath103 , and , after permuting @xmath36 and @xmath37 , we may assume @xmath104 . since , for all @xmath105 , @xmath106 and @xmath107 , it follows that @xmath108 for all @xmath105 , so that @xmath109 achieves the minimum of @xmath110 . since @xmath36 is in @xmath98 , there exists some index @xmath111 for which @xmath112 . but now @xmath113 . since @xmath114 is the minimum of all @xmath115 , the two must be equal , and this minimum is obtained at least twice as desired . proposition [ hypconv ] implies that if @xmath12 is a subset of @xmath20 which happens to lie in a tropical hyperplane @xmath98 , then its tropical convex hull @xmath116 will lie in @xmath98 as well . the same holds for tropical planes of higher codimension . recall that every _ tropical plane _ is an intersection of tropical hyperplanes @xcite . but the converse does not hold : not every intersection of tropical hyperplanes qualifies as a tropical plane ( see @xcite ) . proposition [ hypconv ] and the first statement in theorem [ lotsofstuff ] imply : [ planesconvex ] tropical planes in @xmath20 are tropically convex . a theorem in classical geometry states that every point outside a closed convex set can be separated from the convex set by a hyperplane . the same statement holds in tropical geometry . this follows from the results in @xcite . some caution is needed , however , since the definition of hyperplane in @xcite differs from our definition of hyperplane , as explained in @xcite . in our definition , a tropical hyperplane is a fan which divides @xmath20 into @xmath2 convex cones , each of which is also tropically convex . rather than stating the most general separation theorem , we will now focus our attention on tropical polytopes , in which case the separation theorem is the farkas lemma stated in the next section . throughout this section we fix a finite subset @xmath117 of tropical projective space @xmath20 . here our goal is to study the tropical polytope @xmath119 . we begin by describing the natural cell decomposition of @xmath20 induced by the fixed finite subset @xmath12 . let @xmath36 be any point in @xmath20 . the _ type _ of @xmath36 relative to @xmath12 is the ordered @xmath2-tuple @xmath120 of subsets @xmath121 which is defined as follows : an index @xmath52 is in @xmath122 if @xmath123 equivalently , if we set @xmath124 then @xmath122 is the set of all indices @xmath52 such that @xmath125 and @xmath36 have the same @xmath92-th coordinate . we say that an @xmath2-tuple of indices @xmath126 is a _ type _ if it arises in this manner . note that every @xmath52 must be in some @xmath122 . let @xmath127 , @xmath128 and @xmath129 . there are @xmath130 possible types as @xmath36 ranges over the plane @xmath26 . the corresponding cell decomposition has six convex regions ( one bounded , five unbounded ) , @xmath131 edges ( @xmath132 bounded , @xmath133 unbounded ) and @xmath132 vertices . for instance , the point @xmath134 has @xmath135 and its cell is a bounded pentagon . the point @xmath136 has @xmath137 and its cell is one of the six vertices . the point @xmath138 has @xmath139 and its cell is an unbounded edge . our first application of types is the following separation theorem . [ farkas ] for all @xmath140 , exactly one of the following is true . \(i ) the point @xmath36 is in the tropical polytope @xmath141 , or \(ii ) there exists a tropical hyperplane which separates @xmath36 from @xmath142 . the separation statement in part ( ii ) means the following : if the hyperplane is given by ( [ hyperplane ] ) and @xmath143 then @xmath144 for all @xmath145 . consider any point @xmath140 , with @xmath146 , and let @xmath147 as before . we define @xmath148 there are two cases : either @xmath149 or @xmath150 . the first case implies ( i ) . since ( i ) and ( ii ) clearly can not occur at the same time , it suffices to prove that the second case implies ( ii ) . suppose that @xmath150 . then @xmath151 is empty for some index @xmath152 . this means that @xmath153 for @xmath154 . let @xmath155 be smaller than any of these @xmath22 positive reals . we now choose our separating tropical hyperplane ( [ hyperplane ] ) as follows : @xmath156 this certainly satisfies @xmath157 . now , consider any point @xmath158 in @xmath159 . pick any @xmath160 such that @xmath161 . by definition of the @xmath162 , we have @xmath163 for all @xmath105 , and there exists some @xmath92 with @xmath164 . these equations and inequalities imply @xmath165 therefore , the hyperplane defined by ( [ separator ] ) separates @xmath36 from @xmath142 as desired . the construction in ( [ nearestpointmap ] ) defines a map @xmath166 whose restriction to @xmath142 is the identity . this map is the tropical version of the _ nearest point map _ onto a closed convex set in ordinary geometry . such maps were studied in @xcite for convex subsets in arbitrary idempotent semimodules . if @xmath167 and @xmath168 are @xmath2-tuples of subsets of @xmath169 , then we write @xmath170 if @xmath171 for @xmath172 . we also consider the set of all points whose type contains @xmath6 : @xmath173 [ ineqs ] the set @xmath174 is a closed convex polyhedron ( in the usual sense ) . more precisely , @xmath175 let @xmath140 and @xmath176 . first , suppose @xmath36 is in @xmath174 . then @xmath177 . for every @xmath178 such that @xmath179 , we also have @xmath180 , and so by definition we have @xmath181 , or @xmath182 . hence @xmath36 lies in the set on the right hand side of ( [ x_spoly ] ) . for the proof of the reverse inclusion , suppose that @xmath36 lies in the right hand side of ( [ x_spoly ] ) . then , for all @xmath183 with @xmath184 , and for all @xmath105 , we have @xmath185 . this means that @xmath186 and hence @xmath187 . consequently , for all @xmath92 , we have @xmath188 , and so @xmath189 . as an example for lemma [ ineqs ] , we consider the region @xmath190 in the tropical convex hull of @xmath191 , @xmath192 , and @xmath193 . this region is defined by six linear inequalities , one of which is redundant , as depicted in figure [ reg213 ] . lemma [ ineqs ] has the following immediate corollaries . [ intersectregion ] the intersection @xmath194 is equal to the polyhedron @xmath195 . the inequalities defining @xmath196 are precisely the union of the inequalities defining @xmath174 and @xmath197 , and points satisfying these inequalities are precisely those in @xmath194 . [ boundedness ] the polyhedron @xmath174 is bounded if and only if @xmath198 for all @xmath199 . suppose that @xmath200 for all @xmath201 . then for every @xmath92 and @xmath105 , we can find @xmath184 and @xmath202 , which via lemma [ ineqs ] yield the inequalities @xmath203 . this implies that each @xmath204 is bounded on @xmath174 , which means that @xmath174 is a bounded subset of @xmath20 . conversely , suppose some @xmath122 is empty . then the only inequalities involving @xmath53 are of the form @xmath205 . consequently , if any point @xmath36 is in @xmath122 , so too is @xmath206 for @xmath207 , where @xmath208 is the @xmath92-th basis vector . therefore , in this case , @xmath174 is unbounded . [ regionfaces ] suppose we have @xmath126 , with @xmath209 . then if @xmath210 , @xmath197 is a face of @xmath174 , and furthermore all faces of @xmath174 are of this form . for the first part , it suffices to prove that the statement is true when @xmath34 covers @xmath6 in the poset of containment , i.e. when @xmath211 for some @xmath212 and @xmath213 , and @xmath214 for @xmath215 . we have the inequality presentation of @xmath174 given by lemma [ ineqs ] . by the same lemma , the inequality presentation of @xmath197 consists of the inequalities defining @xmath174 together with the inequalities @xmath216 by assumption , @xmath52 is in some @xmath217 . we claim that @xmath197 is the face of @xmath6 defined by the equality @xmath218 since @xmath174 satisfies the inequality @xmath219 , ( [ extraeq ] ) defines a face @xmath220 of @xmath6 . the inequality @xmath221 is in the set ( [ extraineqs ] ) , so ( [ extraeq ] ) is valid on @xmath197 and @xmath222 . however , any point in @xmath220 , being in @xmath174 , satisfies @xmath223 for all @xmath224 . adding ( [ extraeq ] ) to these inequalities proves that the inequalities ( [ extraineqs ] ) are valid on @xmath220 , and hence @xmath225 . so @xmath226 as desired . by the discussion in the proof of the first part , prescribing equality in the facet - defining inequality @xmath182 yields @xmath197 , where @xmath227 and @xmath228 for @xmath229 . therefore , all facets of @xmath174 can be obtained as regions @xmath197 , and it follows recursively that all faces of @xmath174 are of this form . [ alwaystype ] suppose that @xmath126 is an @xmath2-tuple of indices satisfying @xmath230 . then @xmath174 is equal to @xmath197 for some type @xmath34 . let @xmath36 be a point in the relative interior of @xmath174 , and let @xmath231 . since @xmath232 , @xmath34 contains @xmath6 , and by lemma [ regionfaces ] , @xmath197 is a face of @xmath174 . however , since @xmath36 is in the relative interior of @xmath174 , the only face of @xmath174 containing @xmath36 is @xmath174 itself , so we must have @xmath233 as desired . we are now prepared to state our main theorem in this section . [ celldecomp ] the collection of convex polyhedra @xmath174 , where @xmath6 ranges over all types , defines a cell decomposition @xmath234 of @xmath235 . the tropical polytope @xmath236 equals the union of all bounded cells @xmath174 in this decomposition . since each point has a type , it is clear that the union of the @xmath174 is equal to @xmath20 . by corollary [ regionfaces ] , the faces of @xmath174 are equal to @xmath237 for @xmath238 , and by corollary [ alwaystype ] , @xmath239 for some type @xmath240 , and hence @xmath237 is in our collection . the only thing remaining to check to show that this collection defines a cell decomposition is that @xmath194 is a face of both @xmath174 and @xmath197 , but @xmath241 by corollary [ intersectregion ] , and @xmath196 is a face of @xmath174 and @xmath197 by corollary [ regionfaces ] . for the second assertion consider any point @xmath140 and let @xmath242 . we have seen in the proof of the tropical farkas lemma ( proposition [ farkas ] ) that @xmath243 lies in @xmath142 if and only if no @xmath244 is empty . by corollary [ boundedness ] , this is equivalent to the polyhedron @xmath174 being bounded . the collection of bounded cells @xmath174 is referred to as the tropical complex generated by @xmath12 ; thus , theorem [ celldecomp ] states that this provides a polyhedral decomposition of the polytope @xmath245 . different sets @xmath12 may have the same tropical polytope as their convex hull , but generate different tropical complexes ; the decomposition of a tropical polytope depends on the chosen generating set , although we will see later ( proposition [ vertexunique ] ) that there is a unique minimal generating set . here is a nice geometric construction of the cell decomposition @xmath246 of @xmath20 induced by @xmath247 . let @xmath248 be the fan in @xmath20 defined by the tropical hyperplane ( [ hyperplane ] ) with @xmath249 . two vectors @xmath36 and @xmath37 lie in the same relatively open cone of the fan @xmath248 if and only if @xmath250 if we translate the negative of @xmath248 by the vector @xmath85 then we get a new fan which we denote by @xmath251 . two vectors @xmath36 and @xmath37 lie in the same relatively open cone of the fan @xmath252 if and only @xmath253 [ fans ] the cell decomposition @xmath254 is the common refinement of the @xmath22 fans @xmath255 . we need to show that the cells of this common refinement are precisely the convex polyhedra @xmath174 . take a point @xmath36 , with @xmath256 and define @xmath257 by letting @xmath258 whenever @xmath259 two points @xmath36 and @xmath37 are in the relative interior of the same cell of the common refinement if and only if they are in the same relatively open cone of each fan ; this is tantamount to saying that @xmath260 . however , we claim that @xmath261 . indeed , taking the negative of both sides of ( [ maxnotmin ] ) yields exactly the condition for @xmath52 being in @xmath262 , by the definition of type . consequently , the condition for two points having the same type is the same as the condition for the two points being in the relative interior of the same chamber of the common refinement of the fans @xmath263 . an example of this construction is shown for our usual example , where @xmath191 , @xmath192 , and @xmath193 , in figure [ tropfans ] . the next few results provide additional information about the polyhedron @xmath174 . let @xmath264 denote the undirected graph with vertices @xmath265 , where @xmath266 is an edge if and only if @xmath267 . [ dimension ] the dimension @xmath268 of the polyhedron @xmath174 is one less than the number of connected components of @xmath264 , and @xmath174 is affinely and tropically isomorphic to some polyhedron @xmath197 in @xmath269 . the proof is by induction on @xmath2 . suppose we have @xmath270 . then @xmath174 satisfies the linear equation @xmath271 where @xmath272 . eliminating the variable @xmath273 ( projecting onto @xmath274 ) , @xmath174 is affinely and tropically isomorphic to @xmath197 where the type @xmath34 is defined by @xmath275 for @xmath276 and @xmath277 . the region @xmath197 exists in the cell decomposition of @xmath274 induced by the vectors @xmath278 with @xmath279 for @xmath276 , and @xmath280 . the graph @xmath281 is obtained from the graph @xmath264 by contracting the edge @xmath266 , and thus has the same number of connected components . this induction on @xmath2 reduces us to the case where all of the @xmath122 are pairwise disjoint . we must show that @xmath174 has dimension @xmath68 . suppose not . then @xmath174 lies in @xmath20 but has dimension less than @xmath68 . therefore , one of the inequalities in ( [ x_spoly ] ) holds with equality , say @xmath282 for all @xmath283 . the inequality `` @xmath284 '' implies @xmath285 and the inequality `` @xmath286 '' implies @xmath287 . hence @xmath122 and @xmath151 are not disjoint , a contradiction . the following proposition can be regarded as a converse to lemma [ ineqs ] . [ convineqs ] let @xmath288 be any polytope in @xmath20 defined by inequalities of the form @xmath289 . then @xmath288 arises as a cell @xmath174 in the decomposition @xmath290 of @xmath20 defined by some set @xmath291 . define the vectors @xmath85 to have coordinates @xmath292 for @xmath293 , and @xmath294 . ( if @xmath295 did not appear in the given inequality presentation then simply take it to be a very large positive number . ) then by lemma [ ineqs ] , the polytope in @xmath20 defined by the inequalities @xmath296 is precisely the unique cell of type @xmath297 in the tropical conver hull of @xmath298 . the region @xmath174 is a polytope both in the ordinary sense and in the tropical sense . [ regconvex ] every bounded cell @xmath174 in the tropical complex generated by @xmath12 is itself a tropical polytope , equal to the tropical convex hull of its vertices . the number of vertices of the polytope @xmath174 is at most @xmath299 , and this bound is tight for all positive integers @xmath2 . by proposition [ dimension ] , if @xmath174 has dimension @xmath268 , it is affinely and tropically isomorphic to a region in the convex hull of a set of points in @xmath269 , so it suffices to consider the full - dimensional case . the inequality presentation of lemma [ ineqs ] demonstrates that @xmath174 is tropically convex for all @xmath6 , since if two points satisfy an inequality of that form , so does any tropical linear combination thereof . therefore , it suffices to show that @xmath174 is contained in the tropical convex hull of its vertices . the proof is by induction on the dimension of @xmath174 . all proper faces of @xmath174 are polytopes @xmath197 of lower dimension , and by induction are contained in the tropical convex hull of their vertices . these vertices are a subset of the vertices of @xmath174 , and so this face is in the tropical convex hull . take any point @xmath300 in the interior of @xmath174 . since @xmath174 has dimension @xmath2 , we can travel in any direction from @xmath36 while remaining in @xmath174 . let us travel in the @xmath301 direction until we hit the boundary , to obtain points @xmath302 and @xmath303 in the boundary of @xmath174 . these points are contained in the tropical convex hull by the inductive hypothesis , which means that @xmath304 is also , completing the proof of the first assertion . for the second assertion , we consider the convex hull of all differences of unit vectors , @xmath305 . this is a lattice polytope of dimension @xmath68 and normalized volume @xmath306 . to see this , we observe that this polytope is tiled by @xmath2 copies of the convex hull of the origin and the @xmath307 vectors @xmath308 with @xmath309 . the other @xmath68 copies are gotten by cyclic permutation of the coordinates . this latter polytope was studied by gelfand , graev and postnikov , who showed in ( * ? ? ? * theorem 2.3 ( 2 ) ) that the normalized volume of this polytope equals the catalan number @xmath310 . we conclude that every complete fan whose rays are among the vectors @xmath305 has at most @xmath311 maximal cones . this applies in particular to the normal fan of @xmath174 , hence @xmath174 has at most @xmath311 vertices . since the configuration @xmath312 is unimodular , the bound is tight whenever the fan is simplicial and uses all the rays @xmath308 . we close this section with two more results about arbitrary tropical polytopes in @xmath20 . [ intersectpolyt ] if @xmath142 and @xmath313 are tropical polytopes in @xmath20 then @xmath314 is also a tropical polytope . since @xmath142 and @xmath313 are both tropically convex , @xmath314 must also be . consequently , if we can find a finite set of points in @xmath314 whose convex hull contains all of @xmath314 , we will be done . by theorem [ celldecomp ] , @xmath142 and @xmath313 are the finite unions of bounded cells @xmath315 and @xmath316 respectively , so @xmath317 is the finite union of the cells @xmath194 . consider any @xmath194 . using lemma [ ineqs ] to obtain the inequality representations of @xmath174 and @xmath197 , we see that this region is of the form dictated by proposition [ convineqs ] , and therefore obtainable as a cell @xmath318 in some tropical complex . by proposition [ regconvex ] , @xmath318 is itself a tropical polytope , and we can therefore find a finite set whose convex hull is equal to @xmath194 . taking the union of these sets over all choices of @xmath6 and @xmath34 then gives us the desired set of points whose convex hull contains all of @xmath314 . [ vertexunique ] let @xmath319 be a tropical polytope . then there exists a unique minimal set @xmath12 such that @xmath245 . suppose that @xmath142 has two minimal generating sets , @xmath320 and @xmath321 . write each element of @xmath240 as @xmath322 . we claim that @xmath323 . consider @xmath324 and write @xmath325 if the term @xmath326 does not minimize any coordinate in the right - hand side of ( [ selfexpr ] ) , then @xmath327 is a linear combination of @xmath328 , contradicting the minimality of @xmath12 . however , if @xmath326 minimizes any coordinate in this expression , it must minimize all of them , since @xmath329 . in this case we get @xmath330 , or @xmath331 . pick any @xmath52 for which @xmath332 ; we claim that @xmath333 . indeed , if any other term in @xmath334 contributed nontrivially to @xmath335 , that term would also contribute to the expression on the right - hand side of ( [ selfexpr ] ) , which is a contradiction . consequently , @xmath336 , which means @xmath337 since both sets are minimal by hypothesis . like many of the results presented in this section , propositions [ intersectpolyt ] and [ vertexunique ] parallel results on ordinary polytopes . we have already mentioned the tropical analogues of the farkas lemma and of carathodory s theorem ( propositions [ carat ] and [ farkas ] ) ; proposition [ dimension ] is analogous to the result that a polytope @xmath338 of dimension @xmath268 is affinely isomorphic to some @xmath339 . proposition [ regconvex ] hints at a duality between an inequality representation and a vertex representation of a tropical polytope ; this duality has been studied in greater detail by michael joswig @xcite . every set @xmath340 of @xmath22 points in @xmath20 begets a tropical polytope @xmath341 equipped with a cell decomposition into the tropical complex generated by @xmath12 . each cell of this tropical complex is labelled by its type , which is an @xmath2-vector of finite subsets of @xmath91 . two configurations ( and their corresponding tropical complexes ) @xmath12 and @xmath240 have the same _ combinatorial type _ if the types occurring in their tropical complexes are identical ; note that by lemma [ regionfaces ] , this implies that the face posets of these polyhedral complexes are isomorphic . with the definition in the previous paragraph , the statement of theorem [ main ] has now finally been made precise . we will prove this correspondence between tropical complexes and subdivisions of products of simplices by constructing the polyhedral complex @xmath342 in a higher - dimensional space . let @xmath240 denote the @xmath343-dimensional real vector space @xmath344 . the natural coordinates on @xmath240 are denoted @xmath345 . as before , we fix an ordered subset @xmath346 of @xmath20 where @xmath347 . this defines the unbounded polyhedron @xmath348 [ triangtrop ] there is a piecewise - linear isomorphism between the tropical complex generated by @xmath12 and the complex of bounded faces of the @xmath343-dimensional polyhedron @xmath349 . the image of a cell @xmath350 of @xmath342 under this isomorphism is the bounded face @xmath351 of the polyhedron @xmath349 . that bounded face maps isomorphically to @xmath350 via projection onto the @xmath352-coordinates . let @xmath220 be a bounded face of @xmath353 , and define @xmath122 via @xmath184 if @xmath354 is valid on all of @xmath220 . if some @xmath355 or @xmath356 appears in no equality , then we can subtract arbitrary positive multiples of that basis vector to obtain elements of @xmath220 , contradicting the assumption that @xmath220 is bounded . therefore , each @xmath52 must appear in some @xmath122 , and each @xmath122 must be nonempty . since every @xmath355 appears in some equality , given a specific @xmath352 in the projection of @xmath220 onto the @xmath352-coordinates , there exists a unique @xmath37 for which @xmath357 , so this projection is an affine isomorphism from @xmath220 to its image . we need to show that this image is equal to @xmath174 . let @xmath352 be a point in the image of this projection , coming from a point @xmath358 in the relative interior of @xmath220 . we claim that @xmath359 . indeed , looking at the @xmath92th coordinate of @xmath352 , we find @xmath360 @xmath361 the defining inequalities of @xmath174 are @xmath362 with @xmath184 . subtracting the inequality @xmath363 from the equality in ( [ intype ] ) yields that this inequality is valid on @xmath352 as well . therefore , @xmath359 . similar reasoning shows that @xmath364 . we note that the relations ( [ outtype ] ) and ( [ intype ] ) can be rewritten elegantly in terms of the tropical product of a row vector and a matrix : @xmath365 for the reverse inclusion , suppose that @xmath366 . we define @xmath367 . this means that @xmath368 we claim that @xmath357 . indeed , we certainly have @xmath369 for all @xmath52 and @xmath92 , so @xmath370 . furthermore , when @xmath184 , we know that @xmath371 achieves the minimum in the right - hand side of ( [ expry ] ) , so that @xmath372 and @xmath373 is satisfied . consequently , @xmath357 as desired . it follows immediately that the two complexes are isomorphic : if @xmath220 is a face corresponding to @xmath174 and @xmath374 is a face corresponding to @xmath197 , where @xmath6 and @xmath34 are both types , then @xmath174 is a face of @xmath197 if and only if @xmath375 . however , by the discussion above , this is equivalent to saying that the equalities @xmath374 satisfies ( which correspond to @xmath34 ) are a subset of the equalities @xmath220 satisfies ( which correspond to @xmath6 ) ; this is true if and only if @xmath220 is a face of @xmath374 . so @xmath174 is a face of @xmath197 if and only if @xmath220 is a face of @xmath374 , which implies the isomorphism of complexes . the boundary complex of the polyhedron @xmath376 is polar to the regular subdivision of the product of simplices @xmath377 defined by the weights @xmath378 . we denote this regular polyhedral subdivision by @xmath379 . explicitly , a subset of vertices @xmath380 of @xmath377 forms a cell of @xmath381 if and only if the equations @xmath382 indexed by these vertices specify a face of the polyhedron @xmath353 . we refer to the book of de loera , rambau and santos @xcite for basics on polyhedral subdivisions . we now present the proof of the result stated in the introduction . .1 cm _ proof of theorem [ main ] : _ the poset of bounded faces of @xmath353 is antiisomorphic to the poset of interior cells of the subdivision @xmath381 of @xmath383 . since every full - dimensional cell of @xmath381 is interior , the subdivision is uniquely determined by its interior cells . in other words , the combinatorial type of @xmath353 is uniquely determined by the lists of facets containing each bounded face of @xmath353 . these lists are precisely the types of regions in @xmath342 by lemma [ triangtrop ] . this completes the proof . .3 cm theorem [ main ] , which establishes a bijection between the tropical complexes generated by @xmath22 points in @xmath20 and the regular subdivisions of a product of simplices @xmath383 , has many striking consequences . first of all , we can pick off the types present in a tropical complex simply by looking at the cells present in the corresponding regular subdivision . in particular , if we have an interior cell @xmath384 , the corresponding type appearing in the tropical complex is defined via @xmath385\,\mid\,(j , i)\in m\}$ ] . it is worth noting that via the cayley trick @xcite , theorem [ main ] is equivalent to saying that tropical complexes generated by @xmath22 points in @xmath20 are in bijection with the regular mixed subdivisions of the dilated simplex @xmath386 . this connection is expanded upon and employed in a paper with francisco santos @xcite . another astonishing consequence of theorem [ main ] is the identification of the row span and column span of a matrix . this result can also be derived from ( * ? ? ? * theorem 42 ) . [ polar ] given any matrix @xmath387 , the tropical complex generated by its column vectors is isomorphic to the tropical complex generated by its row vectors . this isomorphism is gotten by restricting the piecewise linear maps @xmath388 and @xmath389 . by theorem [ main ] , the matrix @xmath384 corresponds via the polyhedron @xmath390 to a regular subdivision of @xmath391 , and the complex of interior faces of this regular subdivision is combinatorially isomorphic to both the tropical complex generated by its row vectors , which are @xmath22 points in @xmath20 , and the tropical complex generated by its column vectors , which are @xmath2 points in @xmath392 . furthermore , lemma [ triangtrop ] tells us that the cell in @xmath390 is affinely isomorphic to its corresponding cell in both tropical complexes . finally , in the proof of lemma [ triangtrop ] , we showed that the point @xmath358 in a bounded face @xmath220 of @xmath390 satisfies @xmath393 and @xmath394 . this point projects to @xmath37 and @xmath352 , and so the piecewise - linear isomorphism mapping these two complexes to each other is defined by the stated maps . the common tropical complex of these two tropical polytopes is given by the complex of bounded faces of the common polyhedron @xmath390 , which lives in a space of dimension @xmath395 ; the tropical polytopes are unfoldings of this complex into dimensions @xmath396 and @xmath68 . theorem [ polar ] also gives a natural bijection between the combinatorial types of tropical convex hulls of @xmath22 points in @xmath20 and the combinatorial types of tropical convex hulls of @xmath2 points in @xmath392 , incidentally proving that there are the same number of each . this duality statement extends a similar statement in @xcite . figure [ dualtriang ] shows the dual of the convex hull of @xmath397 , also a tropical triangle ( here @xmath127 ) . for instance , we compute : @xmath398 this point is the image of the point @xmath399 under this duality map . note that duality does not preserve the generating set ; the polytope on the right is the convex hull of points @xmath400 , while the polytope on the left is the convex hull of points @xmath401 . this is necessary , of course , since in general a polytope with @xmath22 vertices is mapped to a polytope with @xmath2 vertices , and @xmath22 need not equal @xmath2 as it does in our example . we now discuss the generic case when the subdivision @xmath381 is a regular triangulation of @xmath383 . we refer to @xcite for the geometric interpretation of the _ tropical determinant_. [ thegenericcase ] for a configuration @xmath12 of @xmath402 points in @xmath20 with @xmath403 the following are equivalent : 1 . the regular subdivision @xmath404 is a triangulation of @xmath383 . no @xmath105 of the points in @xmath12 have projections onto a @xmath105-dimensional coordinate subspace which lie in a tropical hyperplane , for any @xmath405 . no @xmath406-submatrix of the @xmath407-matrix @xmath408 is tropically singular , i.e. has vanishing tropical determinant , for any @xmath409 . the last equivalence is proven in ( * ? ? ? * lemma 5.1 ) . we will prove that ( 1 ) and ( 3 ) are equivalent . the tropical determinant of a @xmath105 by @xmath105 matrix @xmath384 is the tropical polynomial @xmath410 . the matrix @xmath384 is tropically singular if the minimum @xmath411 is achieved twice . the regular subdivision @xmath404 is a triangulation if and only if the polyhedron @xmath353 is simple , which is to say if and only if no @xmath412 of the facets @xmath413 meet at a single vertex . for each vertex @xmath414 , consider the bipartite graph @xmath415 consisting of vertices @xmath416 and @xmath417 with an edge connecting @xmath355 and @xmath356 if @xmath414 lies on the corresponding facet . this graph is connected , since each @xmath355 and @xmath356 appears in some such inequality , and thus it will have a cycle if and only if it has at least @xmath412 edges . consequently , @xmath353 is not simple if and only there exists some @xmath415 with a cycle . if there is a cycle , without loss of generality it reads @xmath418 . consider the submatrix @xmath384 of @xmath408 given by @xmath419 and @xmath420 . we have @xmath421 , @xmath422 , and so on , and also @xmath423 . adding up all of these equalities yields @xmath424 . but consider any permutation @xmath425 in the symmetric group @xmath151 . since we have @xmath426 , we have @xmath427 . consequently , the permutations equal to the identity and to @xmath428 simultaneously minimize the determinant of the minor @xmath384 . this logic is reversible , proving the equivalence of ( 1 ) and ( 3 ) . if the @xmath22 points of @xmath12 are in general position , the tropical complex they generate is called a _ generic tropical complex_. these polyhedral complexes are then polar to the complexes of interior faces of regular triangulations of @xmath377 . all tropical complexes generated by @xmath22 points in general position in @xmath20 have the same @xmath25-vector . specifically , the number of faces of dimension @xmath105 is equal to the multinomial coefficient @xmath429 by proposition [ thegenericcase ] , these objects are in bijection with regular triangulations of @xmath430 . the polytope @xmath142 is equidecomposable @xcite , meaning that all of its triangulations have the same @xmath25-vector . the number of faces of dimension @xmath105 of the tropical complex generated by given @xmath22 points is equal to the number of interior faces of codimension @xmath105 in the corresponding triangulation . since all triangulations of all products of simplices have the same @xmath25-vector , they must also have the same interior @xmath25-vector , which can be computed by taking the @xmath25-vector and subtracting off the @xmath25-vectors of the induced triangulations on the proper faces of @xmath142 . these proper faces are all products of simplices and hence equidecomposable , so all of these induced triangulations have @xmath25-vectors independent of the original triangulation as well . to compute this number , we therefore need only compute it for one tropical complex . let the vectors @xmath85 , @xmath431 , be given by @xmath432 . by theorem [ ineqs ] , to count the faces of dimension @xmath105 in this tropical complex , we enumerate the existing types with @xmath105 degrees of freedom . consider any index @xmath52 . we claim that for any @xmath36 in the tropical convex hull of @xmath433 , the set @xmath434 is an interval @xmath435 , and that if @xmath436 , the intervals @xmath437 and @xmath435 meet in at most one point , which in that case is the largest element of @xmath437 and the smallest element of @xmath435 . suppose we have @xmath184 and @xmath438 with @xmath436 . then we have by definition @xmath439 and @xmath440 . adding these inequalities yields @xmath441 , or @xmath442 . since @xmath436 , it follows that we must have @xmath443 . therefore , we can never have @xmath184 and @xmath438 with @xmath436 and @xmath444 . the claim follows immediately , since the @xmath435 cover @xmath445 $ ] . the number of degrees of freedom of an interval set @xmath446 is easily seen to be the number of @xmath52 for which @xmath435 and @xmath447 are disjoint . given this , it follows from a simple combinatorial counting argument that the number of interval sets with @xmath105 degrees of freedom is the multinomial coefficient given above . finally , a representative for every interval set is given by @xmath448 , where if @xmath122 and @xmath449 have an element @xmath52 in common ( they can have at most one ) , @xmath450 , and if not then @xmath451 . therefore , each interval set is in fact a valid type , and our enumeration is complete . the number of combinatorially distinct generic tropical complexes generated by @xmath22 points in @xmath20 equals the number of distinct regular triangulations of @xmath377 . the number of respective symmetry classes under the natural action of the product of symmetric groups @xmath452 on both spaces is also the same . the 35 symmetry classes of tropical complexes generated by four points in @xmath26 . ] the symmetries in the group @xmath374 correspond to a natural action on @xmath377 given by permuting the vertices of the two component simplices ; the symmetries in the symmetric group @xmath453 correspond to permuting the points in a tropical polytope , while those in the symmetric group @xmath454 correspond to permuting the coordinates . ( these are dual , as per corollary [ polar ] . ) the number of symmetry classes of regular triangulations of the polytope @xmath455 is computable via jrg rambau s topcom @xcite for small @xmath22 and @xmath2 : @xmath456 for example , the @xmath457 entry of the table divulges that there are @xmath458 symmetry classes of regular triangulations of @xmath459 . these correspond to the @xmath458 combinatorial types of four - point configurations in @xmath26 , or to the 35 combinatorial types of three - point configurations in @xmath84 . these @xmath458 configurations ( with the tropical complexes they generate ) are shown in figure [ 35quads ] ; the labeling corresponds to rambau s labeling ( see @xcite ) of the regular triangulations of @xmath460 . a fundamental problem in bioinformatics is the reconstruction of phylogenetic trees from approximate distance data . in this section we show how tropical convexity might help provide new algorithmic tools for this problem . our approach augments the results in @xcite and it provides a tropical interpretation of the work on @xmath34-theory by andreas dress and his collaborators @xcite , @xcite , @xcite . consider a symmetric @xmath461-matrix @xmath462 whose entries @xmath463 are non - negative real numbers and whose diagonal entries @xmath464 are all zero . we say that @xmath465 is a ( finite ) _ metric _ if the triangle inequality @xmath466 holds for all indices @xmath178 . our starting point is the following easy observation : [ metricspace ] the symmetric matrix @xmath465 is a metric if and only if all principal @xmath467-minors of the negated symmetric matrix @xmath468 are tropically singular . both properties involve only three points , so we may assume @xmath469 , in which case @xmath470 the tropical determinant of this matrix is the minimum of the six expressions @xmath471 this minimum is attained twice if and only if it is attained by the last two ( identical ) expressions , which occurs if and only if the three triangle inequalities are satisfied . in what follows we assume that @xmath472 is a metric . let @xmath473 denote the tropical convex hull in @xmath474 of the @xmath2 row vectors ( or column vectors ) of the negated matrix @xmath475 . proposition [ metricspace ] tells us that the tropical polytope @xmath476 is always one - dimensional for @xmath477 . the finite metric @xmath472 is said to be a _ tree metric _ if there exists a weighted tree @xmath34 with @xmath2 leaves such that @xmath478 denotes the distance between the @xmath52-th leaf and the @xmath92-th leaf along the unique path between these leaves in @xmath34 . the next theorem characterizes tree metrics among all metrics by the dimension of the tropical polytope @xmath476 . it is the tropical interpretation of results that are quite classical and well - known in the phylogenetics literature . [ buneman ] for a given finite metric @xmath472 the following conditions are equivalent : 1 . @xmath465 is a tree metric , 2 . the tropical polytope @xmath476 has dimension one , 3 . all @xmath479-minors of the matrix @xmath480 are tropically singular , 4 . all principal @xmath479-minors of the matrix @xmath480 are tropically singular , 5 . for any choice of four indices @xmath481 , the maximum of the three numbers @xmath482 and @xmath483 is attained at least twice . the condition ( 5 ) is the familiar _ four point condition _ for tree metrics . the equivalence of ( 1 ) and ( 5 ) is a classical result due to various authors , including buneman @xcite and zaretsky @xcite . see equation ( b3 ) on page 57 in @xcite . suppose that the condition ( 5 ) holds . by the discussion in @xcite , this means that @xmath484 is a point in the tropical grassmannian of lines , in symbols @xmath485 . by * theorem 3.8 ) , the point @xmath480 corresponds to a tropical line @xmath486 in @xmath235 . the @xmath2 distinguished points whose coordinates are the rows of @xmath480 lie on the line @xmath487 . by corollary [ planesconvex ] , it follows that their tropical convex hull @xmath476 is contained in @xmath487 . this means that @xmath476 has dimension one , that is , ( 2 ) holds . suppose that ( 2 ) holds . then the tropical rank of the matrix @xmath480 is equal to two , by ( * ? ? ? * theorem 4.2 ) . this means that all @xmath488-minors of @xmath480 are tropically singular for @xmath489 . the case @xmath490 is precisely the statement ( 3 ) . obviously , the condition ( 3 ) implies the condition ( 4 ) . what remains is to prove the implication from ( 4 ) to ( 5 ) . for this we note that the tropical determinant of the @xmath491-matrix @xmath492 equals twice the minimum of @xmath493 and @xmath494 . ( it s the tropicalization of a @xmath491-pfaffian ) . the matrix is tropically singular if and only if the minimum is attained twice . if the five equivalent conditions of theorem [ buneman ] are satisfied then the metric tree @xmath34 coincides with the one - dimensional tropical polytope @xmath476 . to make sense of this statement , we regard tropical projective space @xmath20 as a metric space with respect to the infinity norm induced from @xmath3 , @xmath495 and we note that the finite metric @xmath465 embeds isometrically into @xmath476 via the rows of @xmath496 : @xmath497 we learned from @xcite that the tropical polytope @xmath476 first appeared in the 1964 paper @xcite by john isbell . for the proof of the following result we assume familiarity with results from @xcite and @xcite . the tropical polytope @xmath476 equals isbell s injective hull of the metric @xmath465 . according to lemma [ triangtrop ] , the tropical polytope @xmath476 is the bounded complex of the following unbounded polyhedron in the @xmath498-dimensional space @xmath499 : @xmath500 dress et al . @xcite showed that the injective hull @xmath501 of the finite metric @xmath465 coincides with the complex of bounded faces of the following @xmath2-dimensional unbounded polyhedron : @xmath502 what we need to show is that the two polyhedra have the same bounded complex . the metric @xmath465 satisfies the tropical matrix identity @xmath503 , because @xmath504 . this implies that any column vector @xmath37 of @xmath480 satisfies @xmath505 . consider any vertex @xmath358 of @xmath506 . then @xmath37 is a column vector of @xmath480 . equation ( [ troplinalg ] ) implies @xmath507 . hence every vertex of @xmath506 lies in the subspace defined by @xmath508 , and so does the complex of bounded faces of @xmath506 . therefore the linear map @xmath509 induces an isomorphism between the bounded complex of @xmath510 and the bounded complex of @xmath511 . theorem [ polar ] specifies an involution on the set of all tropical complexes . we are interested in the fixed points of this canonical involution . a necessary condition is that @xmath512 and @xmath12 is a symmetric matrix . the previous result and its proof can be reinterpreted as follows : a tropical complex @xmath142 is pointwise fixed under the canonical involution ( on the set of all tropical complexes ) if and only if @xmath142 is the injective hull of a metric on @xmath513 . in order for @xmath142 to be fixed under the canonical involution , it is necessary that @xmath514 . hence we can write @xmath515 for some non - negative square matrix @xmath465 . now , @xmath142 is fixed under the involution if and only if the identity @xmath516 holds . this identity is equivalent to @xmath465 being a metric . dress , huber and moulton @xcite emphasize that the tropical polytope @xmath476 records many important invariants of a given finite metric @xmath465 . for instance , the dimension of @xmath476 gives information about how far the metric is from being a tree metric . in practical biological applications of phylogenetic reconstruction , the distances @xmath478 are not known exactly , and @xmath476 appears to contain many ofs the various trees which are found by existing software for phylogenetic reconstruction . the dimension of the tropical complex @xmath517 can be characterized combinatorially by tropicalizing the sub - pfaffians of a skew - symmetric @xmath461-matrix . the tropical pfaffians of format @xmath491 specify the four point condition ( 5 ) in theorem [ buneman ] , while the tropical sub - pfaffians of format @xmath518 specify the six - point condition which is discussed in @xcite . the combinatorial study of _ @xmath105-compatible split systems _ can be interpreted in the setting of tropical algebraic geometry ( cf . @xcite , @xcite , @xcite ) as the study of the @xmath105-th _ secant variety _ in the grassmannian @xmath519 . tropical convexity provides a convenient language to study numerous extensions of the classical problem of tree reconstruction . as an example , imagine the following scenario , which would correspond to the grassmannian of planes in @xmath20 , denoted @xmath520 . suppose there are @xmath2 taxa , labeled @xmath521 , and rather than having a distance for any pair @xmath183 , we are now given a proximity measure @xmath522 for any triple @xmath523 . we can then construct a tropical polytope by taking the tropical convex hull of @xmath524 points as follows : @xmath525 under certain hypotheses , the tropical polytope @xmath142 can be realized as the complex of bounded faces of the polyhedron in @xmath3 defined by the inequalities @xmath526 . it provides a polyhedral model for the tree - like nature of the data @xmath527 . the case of most interest is when @xmath142 is two - dimensional in which case it plays the role of a _ two - dimensional phylogenetic tree_. the construction of this particular tropical polytope @xmath142 was pioneered by dress and terhalle in the important paper @xcite . there they discuss _ valuated matroids _ , which are essentially the points on the tropical grassmannian of @xcite , and they call @xmath142 the _ tight span of a valuated matroid_. we share their view that these tropical polytopes constitute a promising tool for phylogenetic analysis . this work was conducted while mike develin held the aim postdoctoral fellowship 2003 - 2008 and bernd sturmfels held the msri - hewlett packard professorship 2003/2004 . sturmfels also acknowledges partial support from the national science foundation ( dms-0200729 ) . gelfand , m.i . graev and a. postnikov : combinatorics of hypergeometric functions associated with positive roots , in _ arnold - gelfand mathematical seminars , geometry and singularity theory _ , ( eds . arnold , i.m . gelfand , m. smirnov and v.s . retakh ) , birkhuser , boston , 1997 , pp . 205221 . j. de loera , j. rambau and f. santos : _ triangulations of point configurations _ , algorithms and computation in mathematics , springer verlag , heidelberg , to appear . m. develin , f. santos and b. sturmfels : on the rank of a tropical matrix , arxiv : math.co/0312114 . a. dress , k.t . huber and v. moulton : an explicit computation of the injective hull of certain finite metric spaces in terms of their associated buneman complex , _ advances in mathematics _ * 168 * ( 2002 ) 128 . a. dress and w. terhalle : the tree of life and other affine buildings . proceedings of the international congress of mathematicians , vol . iii ( berlin , 1998 ) , _ documenta mathematica _ , extra vol . iii , 1998 , 565574 . kirillov : introduction to tropical combinatorics , in _ physics and combinatorics 2000 _ , proceedings of the nagoya 2000 international workshop , ( eds . kirillov and n. liskova ) , pp . 82150 , world scientific , 2001 . j. rambau : topcom ( triangulations of point configurations and oriented matroids ) , software available at . the numbers cited for regular triangulations of products of simplices can be found at http://www.zib.de / rambau / topcom / topcom - examples/. j. richter - gebert , b. sturmfels , and t. theobald : first steps in tropical geometry , arxiv : math.ag/0306366 , to appear in _ idempotent mathematics and mathematical physics _ , proceedings vienna 2003 , ( editors g.l . litvinov and v.p . maslov ) , american mathematical society , 2004 .

spectral line and continuum observations of the ionized and molecular gas in g20.08 - 0.14 n explore the dynamics of accretion over a range of spatial scales in this massive star - forming region . very large array observations of nh at angular resolution show a large - scale ( 0.5 pc ) molecular accretion flow around and into a star cluster with three small , bright h ii regions . higher resolution ( ) observations with the submillimeter array in hot core molecules ( chcn , ocs , and so ) and the vla in nh , show that the two brightest and smallest h ii regions are themselves surrounded by smaller scale ( 0.05 pc ) accretion flows . the axes of rotation of the large and small scale flows are aligned , and the timescale for the contraction of the cloud is short enough , 0.1 myr , for the large - scale accretion flow to deliver significant mass to the smaller scales within the star formation timescale . the flow structure appears to be continuous and hierarchical from larger to smaller scales . millimeter radio recombination line ( rrl ) observations at angular resolution indicate rotation and outflow of the ionized gas within the brightest h ii region ( a ) . the broad recombination lines and a continuum spectral energy distribution ( sed ) that rises continuously from cm to mm wavelengths , are both characteristic of the class of h ii regions known as `` broad recombination line objects '' . the sed indicates a density gradient inside this h ii region , and the rrls suggest supersonic flows . these observations are consistent with photoevaporation of the inner part of the rotationally flattened molecular accretion flow . we also report the serendipitous detection of a new nh ( 3,3 ) maser .

0910.2270

massive star - forming regions ( msfrs ) with o stars are usually identified by a group of hypercompact ( hc ) h ii or ultracompact ( uc ) h ii regions found together , deeply embedded in a dense molecular cloud @xcite . that several h ii regions are typically found within each star - forming region indicates that massive stars form together in small clusters . furthermore , the infrared luminosity and radio continuum brightness of the individual h ii regions suggest that some of them may themselves contain more than one massive star . thus , the spatial structure of massive star - forming regions is clustered and hierarchical : the star - forming regions contain a number of separate hc and uc h ii regions , each of which may in turn contain a few massive stars . low angular resolution , single - dish , molecular line surveys of msfrs show evidence for large scale contraction of the embedding molecular clouds @xcite . higher angular resolution observations of some of these regions identify velocity gradients consistent with rotation and inflow . in addition to the accretion flows seen on the large - scale ( @xmath5 pc ) of the embedding molecular cloud ( g10.6 - 0.4 : ho & haschick 1986 , keto et al . 1987a , keto 1990 ; g29.96 - 0.02 : olmi et al . 2003 ) , accretion flows are also seen on smaller ( @xmath6 pc ) scales around individual hc and uc h ii regions ( g10.6 - 0.4 : keto et al . 1988 , sollins et al . 2005a ; w3(oh ) : keto et al . 1987b , keto et al . 1995 ; w51e2 : zhang & ho 1997 , young et al . 1998 ; g28.20 - 0.05 : sollins et al . 2005b ; g24.78 + 0.08 : beltrn et al . 2004 , 2005 , 2006 , galvn - madrid et al . 2008 ; g29.96 - 0.02 : beuther et al . 2007 ) . it is unclear how the flows on different length scales are related . in the case of g10.6 - 0.4 , the cluster - scale accretion flow can be traced down from the largest cloud scale to the small scale of the brightest h ii region , but it is not known whether this holds for other objects . for example , in a survey of msfrs , selected on the basis of iras colors and specifically excluding those with h ii regions , multiple bipolar molecular outflows ( implying the presence of accretion flows ) are seen in random orientations @xcite . the different orientations of these smaller - scale flows suggest separate , individual centers of collapse . this comparison raises the question whether a large - scale coherent flow is required for the formation of the most massive stars , o stars ( @xmath7 @xmath8 ) capable of producing bright h ii regions , whereas b stars require only smaller scale flows . it is also unclear what happens in an accretion flow when the inflowing molecular gas reaches the boundary of an embedded h ii region . previous observations suggest that the h ii regions in an msfr that are surrounded by accretion flows , may be best understood as deriving from the continuous ionization of the accretion flow @xcite , rather than as a dynamically separate expanding bubble of ionized gas within the flow . part of the ionized gas may continue to the central star or stars and part escapes off the rotationally flattened accretion flow as a photoevaporative outflow perpendicular to the plane of rotation @xcite . the outflow is accelerated to supersonic speeds by the density gradient maintained by the stellar gravity @xcite . because the extent of an ionized outflow is generally larger than the region of ionized inflow , in most cases the outflow should be detected more easily than the inflow . h ii regions classified as broad recombination line objects " ( brlo ) @xcite show steep density gradients and supersonic flows @xcite , consistent with photoevaporation and acceleration . it is not known whether all brlo are associated with accretion . if the accretion surrounding an o star cluster is continuous from the largest to the smallest scales , this must be the case . there are only a handful of radio recombination line ( rrl ) observations that spatially resolve the ionized flow within an hc h ii region . velocity gradients consistent with outflow and rotation in the ionized gas have been previously reported for w3(oh ) @xcite , w51e2 @xcite , and g28.20 - 0.05 @xcite . observations of the very massive and spatially large g10.6 - 0.4 h ii region made at the vla in the highest possible angular resolution are able to map the inflowing ionized gas @xcite . in order to study the accretion dynamics over a range of scales in a msfr , from the cluster scale down to the scale of individual hc h ii regions and within the ionized gas , we set up a program of radio frequency molecular line , recombination line , and continuum observations at two telescopes and with several different angular resolutions . for this study we chose the massive star formation region g20.08 - 0.14 north ( hereafter g20.08n ) , identified by three uc and hc h ii regions detected in the cm continuum by @xcite . the total luminosity of the region is @xmath9 @xmath10 for a distance of 12.3 kpc . kpc ) is the most commonly quoted in the previous literature . in contrast , @xcite and @xcite report that this region is at the far kinematic distance ( @xmath11 kpc ) . we will assume the far distance throughout the rest of the paper . for reference , a scale of @xmath12 corresponds to @xmath13 au ( 0.03 pc ) . the total luminosity of the region was estimated to be @xmath14 @xmath10 assuming the near kinematic distance @xcite . correcting for the location at the far distance , the luminosity is @xmath9 @xmath10 . ] previous observations suggest accretion in the g20.08n cluster . molecular - line observations show dense gas embedding the h ii regions @xcite . molecular masers , generally associated with ongoing massive - star formation , have been detected in a number of studies ( oh : ho et al . 1983 ; h@xmath3o : hofner & churchwell 1996 ; and ch@xmath0oh : walsh et al . @xcite observed large - scale inward motions consistent with an overall contraction of the embedding molecular cloud . those authors also observed sio line profiles suggestive of massive molecular outflows , further evidence for accretion and star formation . the recombination line spectra show broad lines @xcite , presumably due to large , organized motions in the ionized gas . however , the previous observations do not have the angular resolution and the range of spatial scales needed to confirm the presence of accretion flows and study them in detail . in this paper we report on several observations of g20.08n and discuss our findings . we confirm active accretion within the cluster . furthermore , we find that the parsec - scale accretion flow fragments into smaller flows around the individual hc h ii regions , and that the gas probably flows from the largest scale down to the smallest scale . this continuous and hierarchical accretion may be necessary to supply enough mass to the small - scale flows to form o - type stars , in contrast to low- and intermediate - mass star - forming regions with stars no more massive than @xmath15 @xmath8 , where isolated accretion flows around individual protostars may be sufficient . we observed g20.08n on 2006 june 25 and july 6 with the submillimeter array @xcite in its very extended ( vex ) configuration . two sidebands covered the frequency ranges of @xmath16 ghz and @xmath17 ghz with a spectral resolution of @xmath18 @xmath19 . the h@xmath20 recombination line ( @xmath21 ghz ) was positioned in the upper sideband . the observations sampled baseline lengths from @xmath22 to @xmath23 k@xmath24 , sensitive to a range of spatial scales from @xmath25 to @xmath26 . the visibilities of each observation were separately calibrated using the sma s data calibration program , mir . table 1 lists relevant information on the calibrators . we used quasars for the absolute amplitude scale as well as the time - dependent phase corrections and frequency - dependent bandpass corrections . inspection of the quasar fluxes and comparison with their historical flux densities in the sma database suggest that flux calibration is accurate to better than @xmath27 . the calibrated data were exported to miriad for further processing and imaging . there were enough line - free channels in the 2-ghz passband to subtract the continuum in the @xmath28 domain . the line - free continuum was self - calibrated in phase , and the gain solutions were applied to the spectral line data . we list all the identified lines in table 2 . figure 1 shows the continuum - free spectra across the entire sidebands at the position of the 1.3-mm peak . to improve the sensitivity , the data were smoothed to a spectral resolution of 2 @xmath19 . the rms noise in our natural - weighted maps , made from the combined observations of both days , is @xmath29 @xmath30 for the single - sideband continuum and @xmath31 @xmath30 per channel ( 2 @xmath19 wide ) for the line data . spectral line observations of nh@xmath0 @xmath32 and ( 3,3 ) at two different angular resolutions were obtained with the very large array in the d and bna configurations ( projects as749 , as771 , and as785 ) . partial results from these observations were presented in @xcite . all the observations , except the vla - d ( 3,3 ) , were done with a bandwidth of 3.125 mhz ( @xmath33 @xmath19 ) divided in 64 spectral channels , each @xmath34 @xmath19 wide . the vla - d ( 3,3 ) observation was done with the same bandwidth divided in 128 spectral channels , each @xmath35 @xmath19 wide . the bandwidth covers the main hyperfine line and one line from the innermost satellite pair molecule is symmetric top with inversion , see @xcite for details . ] . the nh@xmath0 data are presented at a spectral resolution of 0.6 @xmath19 . the noise per channel in the final images was in the range of @xmath36 @xmath30 . in addition to the molecular line observations , we observed the ionized gas in the h66@xmath37 recombination line in the vla b configuration . the correlator was set up to cover a bandwidth of 12.5 mhz ( @xmath38 @xmath19 ) divided in 64 channels of 2.6 @xmath19 each . the rms noise per channel in the final image was @xmath39 @xmath30 . all three vla data sets were calibrated using standard procedures in the aips software . tables 1 and 2 summarize the relevant observational parameters . the continuum was constructed in the @xmath28 domain from line - free channels and was then self - calibrated . the gain solutions from self - calibration were applied to the line data . figure [ fig2 ] shows the 1.3-cm continuum ( contours ) obtained from the vla - bna observations overlaid with the 1.3-mm continuum from the sma - vex data ( color scale ) . at 1.3 cm we resolve the g20.08n system into the three components reported by @xcite . h ii region a is the brightest , westernmost peak . h ii region b is the slightly broader peak @xmath40 to the se of a. h ii region c is the more extended uc h ii further to the se . its brightest , eastern rim is detected at the @xmath41 @xmath30 level in our 1.3-mm observations . the continuum of h ii region a is unresolved at 1.3 cm ; at 1.3 mm it shows a core - halo morphology . the 1.3-mm core is unresolved ( gaussian fits yield a deconvolved size at half power fwhm @xmath42 ) . the low - intensity halo has a diameter of @xmath43 ( see fig . [ fig2 ] ) . the h@xmath20 emission is confined to the unresolved core ( [ ionized ] ) and the warm molecular gas ( [ smallscale ] ) coincides with the extended continuum halo . this indicates that the unresolved h ii region a is surrounded by a dust cocoon . h ii region b is barely resolved in the 1.3-cm vla bna map ( deconvolved fwhm @xmath44 ) . the peak position of h ii region a is identical at 1.3 cm and 1.3 mm : @xmath45 @xmath46 @xmath47 , @xmath48 @xmath49 @xmath50 , within the positional uncertainty of the reference quasars ( in the range @xmath51 ) . previous observations of the recombination lines at 2 and 6 cm @xcite put the h ii regions in g20.08n in the class known as `` broad recombination line objects '' ( although those observations did not have sufficient angular resolution to separate h ii regions a , b , and c ) . the large widths in cm - wavelength recombination lines are due to pressure broadening at high gas densities ( @xmath52 @xmath53 ) as well as unresolved supersonic motions @xcite . these h ii regions also have continuum spectral energy distributions ( seds ) that increase with frequency through the mm wavelengths , evidence for a steep density gradient in the ionized gas @xcite . our new vla and sma observations , at 1.3 cm and 1.3 mm respectively , extend the sed to millimeter wavelengths . we find that the flux density of h ii region a continues to rise from cm to mm wavelengths ( fig . [ fig2 ] ) , and we analyze this characteristic in detail below . the flux densities of h ii region b at 1.3 mm ( @xmath54 mjy ) and 1.3 cm ( @xmath55 mjy ) imply a spectral index ( @xmath37 , where @xmath56 ) of @xmath57 , roughly consistent with the expected index of @xmath58 of optically thin gas . h ii region c is more extended and most of the 1.3-mm flux is resolved out . at wavelengths shorter than 1 mm , thermal dust emission contributes significantly to the continuum . we estimate the relative contributions of dust and free - free emission at 1.3 mm from our recombination line observations ( described in [ ionized ] ) and the theoretically expected line - to - continuum ratio . in the optically thin limit ( a good approximation at 1.3 mm ) the free - free line - to - continuum flux ratio @xmath59 is given by the ratio of the opacities @xmath60 @xcite : @xmath61 = \frac{\pi h{^}3 e{^}2 } { ( 2\pi m_e k){^}{3/2 } m_e c } n_1{^}2 f_{n_1,n_2 } \phi_\nu \frac{n_e n_i}{t{^}{3/2 } } \nonumber \\ \times \exp{\biggl(\frac{e_{n_1}}{kt}\biggr ) } \bigl(1- e{^}{-h\nu /kt}\bigr ) , ~ ~ ~ ~ ~ ( 1a ) \nonumber \\ \biggl[\frac{\kappa_c}{\mathrm{cm}{^}{-1}}\biggr ] = 9.77 \times 10{^}{-3 } \frac{n_e n_i}{\nu{^}2 t{^}{3/2 } } \biggl[17.72 + \ln{\frac{t{^}{3/2}}{\nu } } \biggr ] , ~ ~ ~ ~ ~ ( 1b ) \nonumber \label{eq : kappac}\end{aligned}\ ] ] where all the units are in cgs , the physical constants have their usual meanings , @xmath62 for h @xmath20 , @xmath63 for @xmath37 lines , @xmath64 for hydrogen , @xmath65 000 k , and @xmath66 is the normalized line profile . the main source of uncertainty in equations [ eq : kappal ] and [ eq : kappac ] is the temperature of the ionized gas . assuming that the line profile is gaussian and correcting for 8 @xmath67 helium in the gas , the expected ratio at the line center is @xmath68 for @xmath69 k , or @xmath70 for @xmath71 000 k. the observed ratio is @xmath72 therefore , assuming that the rrls are in lte , the free - free contribution to the 1.3-mm flux of h ii region a is @xmath73 for the assumed temperature range . figure [ fig3 ] shows a model sed for h ii region a in which 70 @xmath67 ( 355 mjy ) of the 1.3-mm flux is produced by free - free from an h ii region with a density gradient and 30 @xmath67 ( 142 mjy ) by warm dust . assuming radiative equilibrium , we set the dust temperature @xmath74 to 230 k , the average temperature of the dense gas at the same scales ( [ small - parms ] ) . the modeling procedure is described in @xcite and @xcite . table 3 summarizes the model . the total gas mass inferred from the dust emission is too large ( @xmath75 @xmath8 ) for the h ii region alone , so most of the dust must be in the cocoon around the h ii region . the calculated mass range takes into account uncertainties in the dust emissivity , but not in the temperature . the density gradient derived for the ionized gas in h ii region a is @xmath76 , with @xmath77 . equilibrium between recombination and ionization in this model h ii region requires an ionizing flux equivalent to an o7.5 star ( using the computations of * ? ? ? * ) , although this ionizing flux could be made up of several stars of slightly later spectral type . the model for the sed assumes spherical geometry and a static h ii region with no inflow of neutral gas into the h ii region . in contrast , h ii region a is embedded in a rotationally flattened accretion flow ( see [ smallscale ] ) , so the determination of the stellar spectral type is only approximate . the large - scale molecular cloud is detected by the nh@xmath0 vla - d observations . from these data we find the presence of a parsec - scale accretion flow surrounding the cluster of h ii regions . in [ large - dyn ] we first estimate the systemic velocity @xmath78 of the cloud with respect to the local standard of rest ( lsr ) , and then analyze the line velocities to determine rotation and infall following the procedure used in measuring accretion flow velocities in previous papers ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) . in [ large - parms ] we derive the physical properties of the cloud : temperature , mass , ammonia abundance , and density . figure [ fig4 ] shows the channel maps of the @xmath79 main hyperfine line . the most notable feature is a velocity gradient along the major axis of the cloud , consistent with rotation . the redshifted emission is toward the ne , while the blueshifted gas is toward the sw . also , there is strong absorption against the free - free background of the h ii regions at the center of the cloud , which are unresolved at @xmath80 resolution . the separation between the emission peaks on either side of the absorption is @xmath81 , or 0.6 pc . the symmetry of the channel maps suggests a systemic velocity of about @xmath82 @xmath19 . a gaussian fit to the emission spectrum integrated in a box covering the entire cloud , and clipping out the redshifted absorption , gives a centroid velocity of @xmath83 @xmath19 . we adopt a systemic velocity for the parsec - scale cloud of @xmath84 @xmath19 , in agreement with that reported by @xcite ( 41.9 @xmath19 ) based on observations of cs @xmath85 . figure [ fig5 ] shows position - velocity ( pv ) diagrams across cuts at pa @xmath86 ( sw - ne , major axis ) and pa @xmath87 ( nw - se , minor axis ) . the rotation is seen in the sw - ne cut as a shift in the velocity of the emission contours from one side of the absorption to the other . the velocity offset with respect to @xmath78 seen in the emission contours increases inward , suggesting that the gas rotates faster with decreasing radius . under the assumption that the velocity gradient seen in emission along the major axis is dominated by rotation , the redshifted absorption in the pv diagrams ( fig . [ fig5 ] ) is also evidence for inward flow toward the central h ii regions , as there is an excess of redshifted absorption . this is more clearly seen in fig . [ fig6 ] , which shows the spectra toward the absorption center . the nh@xmath0 ( 2,2 ) main hyperfine absorption peak is redshifted by 2.3 kms@xmath88 with respect to the systemic velocity . an infall velocity @xmath89 @xmath19 is also seen in the nh@xmath0 ( 3,3 ) line , although the spectrum is contaminated by an nh@xmath0 ( 3,3 ) maser ( see description in fig . [ fig6 ] ) . the maser is confirmed by our high angular resolution nh@xmath0 observations ( [ 33maser ] ) . if the rotation were seen edge - on , there would not be a velocity gradient across the h ii region along the minor axis ( nw - se ) . at an oblique viewing angle a velocity gradient along the minor axis is created by the inflow . the velocity gradient along both the minor and major axes ( fig . 5 ) implies that the rotationally flattened flow is tipped with respect to the line of sight . the optical depth of the gas can be determined from the brightness ratios of the hyperfine lines . from their optical depth ratio the rotational temperature between the @xmath90 and @xmath91 transitions can be determined ( see * ? ? ? * ; * ? ? ? * for details of the procedure ) . this temperature can be considered as a first - order approximation to the kinetic temperature ( @xmath92 ) of the molecular gas . @xcite suggest that an improved estimate of the kinetic temperature ( @xmath92 ) is obtained by accounting for the populations in the upper states of the k - ladders rather than assuming that all states above the lowest are negligibly populated . including this correction , we obtain an average kinetic temperature @xmath93 k for the parsec - scale cloud . the high - resolution data discussed in [ smallscale ] show that higher gas temperatures are found closer to the h ii regions . the mean column density of nh@xmath0 is @xmath94 @xmath95 . we can determine the abundance of ammonia @xmath96 by comparing the nh@xmath0 and h@xmath3 column densities . the mass of the cloud can be estimated from the observed velocity dispersion ( @xmath97 kms@xmath88 ) , the radius of the cloud ( @xmath98 au ) , and the virial theorem : @xmath99 . similarly , from the observed velocity gradient , the rotation velocity is @xmath100 @xmath19 at a radius @xmath101 au . equating the centripetal and gravitational forces , the gas mass inside @xmath102 is @xmath103 @xmath8 . therefore , we estimate the average column density of h@xmath3 to be @xmath104 @xmath95 and @xmath105 , within a factor of three of the value estimated for g10.60.4 @xcite . the mean h@xmath3 density in the large - scale cloud is @xmath106 @xmath53 . our sma very extended ( vex ) and vla - bna observations provide a rich view of the molecular environment close to the h ii regions . the lines detected with the sma that have peak intensities @xmath107 and are not blended with any other line are listed in the top part of table 2 . we follow a similar outline as for the large - scale cloud . figure [ fig7a ] shows the channel maps of ocs @xmath108 , so@xmath3 @xmath109 , ch@xmath0cn @xmath110 , and ch@xmath0cn @xmath111 . these maps show that the line emission from these hot - core molecules is considerably brighter around h ii region a. all the molecules show a velocity gradient across this source , from the southwest to northeast , in a similar orientation as the larger , cluster - scale flow . we do not detect molecular emission around h ii region b. at this high angular resolution we are not sensitive to brightness temperatures of less than @xmath41 k ( @xmath112 ) for the mm lines . figure [ fig8 ] shows the velocity - integrated intensity ( moment 0 ) and the intensity - weighted mean velocity ( moment 1 ) maps of the four lines previously shown in fig . [ fig7a ] . the integrated emission of ch@xmath0cn @xmath113 is brightest in front of the h ii region , while for the other molecules the brightness peak is slightly offset from the continuum . these differences reflect the relative brightness of each molecule with respect to the continuum emission of h ii region a. the velocity gradient across h ii region a is also seen . figure [ fig9 ] shows the velocity - integrated ( moment 0 ) and velocity dispersion ( moment 2 ) maps for the same lines as fig . the line widths increase toward the continuum peak , indicating that unresolved motions increase closer to the h ii region . figure 10 shows the position - velocity ( pv ) diagrams for the lines of figure 7 in cuts at position angles pa @xmath114 and pa @xmath115 across the continuum peak of hii region a. the cuts at @xmath116 show the velocity gradient also seen in the channel maps of ocs and both of the ch@xmath0cn lines . the pv diagram of ch@xmath0cn 12(4)-11(4 ) has a feature suggestive of a velocity gradient in the perpendicular direction pa @xmath117 , with an excess of redshifted emission toward the nw . consistent with our interpretation of the nh@xmath0 vla - d data , this suggests inward motion in a rotationally - flattened flow that is seen not quite edge - on . the same redshifted feature is also marginally detected in the lower excitation ch@xmath0cn transition as well as the ocs line . however , the infall signature in emission is only tentative , and a clearer indication of infall at small scales comes from the redshifted nh@xmath0 absorption in the vla - bna data ( see below ) . in general , observational experience suggests that ch@xmath0cn , along with nh@xmath0 , is a reliable tracer of high - density molecular mass and accretion flows @xcite . ch@xmath0cn has recently been detected in the outflow of the nearby low - mass star l1157 @xcite , but at a very low brightness ( @xmath118 k ) . the distribution of the ocs molecule in our observations is very similar to the ch@xmath0cn , but the so@xmath3 velocities do not show the same pattern , and are more difficult to interpret . the so@xmath3 may be more easily affected by the excitation conditions , and part of the observed emission could arise from the shocked boundaries of outflows . from gaussian fits to the ch@xmath0cn @xmath119 and 4 emission lines at the position of the continuum peak , the systemic velocity at small scales is estimated to be @xmath120 . figure [ fig11 ] shows the ch@xmath0cn spectra and their gaussian fits . at subarcsecond angular resolution , our nh@xmath0 observations are sensitive to emission of brightness temperature above @xmath121 k. therefore the thermal nh@xmath0 is detected only in absorption against the bright continuum . as in [ large - parms ] , a comparison between the nh@xmath0 absorption line velocity and @xmath78 shows an inward velocity of @xmath122 in front of h ii region a ( figure [ fig12 ] ) . the nh@xmath0 ( 2,2 ) absorption line in front of h ii region b is redshifted by @xmath123 @xmath19 with respect to @xmath78 , implying inward motion and accretion toward h ii region b as well ( figure [ fig12 ] ) . the ( 3,3 ) absorption in front of h ii region b is mixed with nh@xmath0 maser emission ( [ 33maser ] ) , and the determination of the inward velocity is uncertain . more sensitive observations are needed to constrain the properties of the molecular gas around h ii region b. we estimate the dynamical mass @xmath124 ( gas plus stars ) within the smaller accretion flow in the same way as with the large - scale flow ( see [ large - parms ] ) . at a radius @xmath125 au the rotation velocity is @xmath126 @xmath19 . therefore , @xmath127 . this is consistent with the lower limit to the stellar mass , @xmath128 @xmath8 , required for ionization equilibrium ( [ sed ] ) . the estimate is also consistent with the gas mass derived from the mm continuum once the free - free contribution has been properly subtracted , @xmath129 ( [ sed ] ) . the mean h@xmath3 density is @xmath130 @xmath53 . we derive the temperature in the dense gas surrounding h ii region a from the rotational energy diagram of the lines of the ch@xmath0cn @xmath131 k - ladders ( see e.g. , * ? ? ? * ; * ? ? ? figure [ fig13 ] shows this diagram for two cases : one considering all the @xmath132 lines , and the other including only the @xmath133 lines , which have lower optical depths than the low - number k lines ( also , the @xmath134 lines are blended , separated by only 5.8 @xmath19 ) . the rotational temperature obtained for the former case is @xmath135 k , while for the latter it is @xmath136 k. the difference between the two values given above appears to be caused by optical depth effects ( the rotational diagram analysis assumes that the emission is optically thin ) . although our sensitivity level does not permit us to detect the lines of the isotopologue ch@xmath137cn and measure the optical depth of the ch@xmath0cn emission , the upper limits are not restrictive ( @xmath138 ) . high optical depths are also suggested by the flat slope of the k ladders ( fig . 1 ) . fitting the k ladders taking into account the opacities @xcite yields a kinetic temperature close to the lower estimation , @xmath139 k , and optical depths are @xmath140 ( keping qiu 2009 , personal communication ) . the rotational temperatures are higher if we use only the brightest pixels instead of averaging over all the emission . this suggests a temperature gradient toward the center of the h ii region . we derive a ch@xmath0cn column density of @xmath141 @xmath95 assuming @xmath142 k. comparison of the ch@xmath0cn column density with the dynamical mass implies an abundance @xmath143(ch@xmath0cn ) @xmath144 for the range of masses quoted above . abundance estimates in other msfrs cover a range of values : @xmath145 inside the orion hot core and @xmath146 outside @xcite ; @xmath147 inside the orion hot core and @xmath148 in the orion ridge @xcite ; @xmath149 in sgr b2(n ) @xcite ; @xmath150 in w51e1/e2 @xcite . the optical depth of nh@xmath0 in absorption toward h ii region a is @xmath151 , @xmath152 . the mean column density is @xmath153 @xmath95 . if the ammonia abundance at these small scales is in the range @xmath154 , then the molecular hydrogen column density is @xmath155 @xmath95 . this implies a molecular gas mass of @xmath156 @xmath8 . the kinetic temperature obtained from the nh@xmath0 is @xmath157 k , considerably cooler than that obtained for ch@xmath0cn and implying that most of the nh@xmath0 column density is further away from the h ii region than the ch@xmath0cn . the distribution of the warm molecular gas around h ii region a appears to be rotationally flattened ( figs . [ fig7a ] , [ fig8 ] , [ fig9 ] ) , but the observed size is slightly larger than the radius at which the accretion flow is expected to become centrifugally supported ( i.e. , a `` disk '' ) , @xmath158 au for a star of mass @xmath159 and @xmath160 . actually , the disk scale matches the size of the h ii region ( table 3 , [ ionized ] ) . the mass - inflow rate toward h ii region a can be estimated from the high - resolution nh@xmath0 absorption . from an inflow velocity of @xmath161 and spherical geometry , the mass - inflow rate is @xmath162 yr@xmath88 ( for the @xmath163 of molecular gas detected in ammonia absorption ) . this estimate may be an upper limit because the hot molecular core is flattened rather than spherical . in [ sed ] we inferred a density gradient inside the hc h ii region a. in this section , we derive the internal dynamics of this h ii region based on multifrequency rrls . the mm / sub - mm lines are especially important because they are much less affected by pressure broadening and preferentially trace denser gas . while subarcsecond resolution studies at wavelengths longer than 7 mm have been available for many years @xcite , similar studies at shorter wavelengths have had limited angular resolution ( e.g. , * ? ? ? @xcite presented the first results of high - frequency , high - resolution ( @xmath164 ) , multifrequency rrl observations in a sample of 5 msfrs with similar characteristics to g20.08n a. they were able to separate the contributions of pressure broadening and large - scale motions to the line width , even when the h ii regions were unresolved . we follow their procedure to analyze our rrl data . we observed the h30@xmath37 ( @xmath165 ghz ) and h66@xmath37 ( @xmath166 ghz ) lines at subarcsecond angular resolution ( see table 2 ) . because both the line - to - continuum ratio and the continuum intensity are lower at 22 ghz than at 231 ghz , the h66@xmath37 line is much weaker than the h@xmath20 line . this is somewhat alleviated by the better sensitivity of the vla , but the signal - to - noise ratio ( s / n)in the high - frequency line is still better . figure [ fig14 ] shows the moment 0 and moment 1 maps of the h@xmath20 line . although the emission is unresolved at half power , there is a slight indication of a velocity gradient in the ionized gas that agrees ( not perfectly ) with the rotation seen in ch@xmath0cn and ocs . figure [ fig15 ] shows the h@xmath20 and h@xmath167 spectra toward h ii region a and their gaussian fits . the h@xmath167 line shows evidence of a blueshifted wing , suggesting either inflow or outflow in addition to rotation . within the uncertainties of the fits , both the h66@xmath37 and h30@xmath37 lines have the same line width ( table 4 ) . assuming that the dynamical broadening @xmath168 ( caused by turbulence and ordered motions ) and the thermal broadening @xmath169 are gaussian , and that the pressure broadening @xmath170 is lorentzian , the rrl has a voigt profile with line width @xcite : @xmath171 where all the widths are fwhm . for the h@xmath20 line at 231.9 ghz the pressure broadening is less than the thermal broadening at electron densities below @xmath172 @xmath53 @xcite . our sed modeling indicates lower densities over most of the hii region . therefore , the observed line width can be attributed to thermal plus dynamical broadening . the electron temperature @xmath173 in uc h ii regions is typically @xmath174 000 k , with a small gradient as a function of galactocentric radius @xcite . we adopt @xmath175 k ( @xmath176 k for h30@xmath37 because of the low optical depth and @xmath177 filling factor ) , which translates into a thermal fwhm of @xmath178 . therefore , from eq . [ eq : voigt ] , we obtain a dynamical width of @xmath179 . from the velocity gradient ( fig . [ fig14 ] ) it is seen that @xmath180 @xmath19 of @xmath168 can be in the form of rotation . the rest could be caused by inflowing or outflowing ionized gas , as suggested by the blueshifted ( @xmath181 @xmath19 ) mean velocities of the rrls , and by the blue wing in the h@xmath167 spectrum ( fig . [ fig15 ] ) . most of the ionized gas that we see should be outflow . inflow inside the h ii region is expected within the radius where the escape velocity from the star exceeds the sound speed of the ionized gas . this is approximately the bondi - parker transonic radius @xcite , @xmath182 au @xmath183 , or about 190 au for h ii region a , assuming a sound speed @xmath184 kms@xmath88 and stellar mass @xmath159 . h ii region a extends out to @xmath185 au , so most of the gas is not gravitationally bound to the star and flows outward . in this model , the outflow is continuously supplied by photoevaporation off the rotationally flattened accretion flow ( t. peters et al . 2009 , in preparation ) . the somewhat misaligned velocity gradient in the ionized gas ( fig . 14 ) derives from a combination of the rotation and outflow blended together in the observing beam . low angular resolution ( hpbw @xmath186 ) , single - dish ( jcmt ) observations of sio ( 8 - 7 ) show evidence for large - scale molecular outflows in g20.08n @xcite . although the standard outflow tracers @xmath187co @xmath188 and @xmath137co @xmath188 are in our passband , we do not detect any co in emission . the @xmath28 coverage of our sma - vex observations is incapable of imaging structures larger than @xmath189 ; therefore , the co emission from the molecular cloud and molecular outflow must be on larger scales . @xmath187co and @xmath137co are seen in our data in absorption at the position of h ii region a , at several different velocities in the range @xmath190 . some co absorption features are at the same velocities as the hi absorption features of @xcite and are therefore due to foreground gas that is not related to g20.08n but rather to intervening galactic spiral arms . a handful of nh@xmath0 masers have been reported in the literature , always associated with massive star formation ( e.g. , * ? ? ? many of the known nh@xmath0 masers are from non - metastable ( @xmath191 ) transitions . the first clear detection of a metastable ( @xmath192 ) nh@xmath0 ( 3,3 ) maser was reported by @xcite toward dr 21(oh ) . most of the detections point toward a shock excitation origin for the population inversion , inasmuch as the maser spots are invariably associated with outflow indicators such as bipolar co and/or sio structures , class i methanol masers , and/or water masers @xcite . we report the serendipitous detection of a new nh@xmath0 ( 3,3 ) maser toward g20.08n . the maser spot is relatively weak , and is spatially centered at @xmath45 @xmath46 @xmath193 , @xmath48 @xmath49 @xmath194 , close in projection to the center of h ii region b. the maser spot is spatially unresolved even in uniform weighting maps of the vla - bna data ( hpbw = @xmath195 , pa = @xmath196 ) . if the deconvolved source size is limited to half the beam size , then the peak brightness temperature of the spot is constrained to @xmath197 k. the high intensity , together with the absence of similar emission in our ( 2,2 ) maps at high angular resolution , confirm the maser nature of the ( 3,3 ) emission . the spectral feature is also very narrow ( fig . [ fig16 ] ) , typical of maser emission , although it shows evidence of line wings . from a gaussian fit to the line profile , the velocity of the maser is @xmath198 ( accurate only to 20 % @xmath199 40 % ) . the fwhm is @xmath200 , after deconvolving the channel width of @xmath201 . owing to its position , it is probable that the maser is excited by h ii region b. we do not have sufficient data , however , to assert that it is excited in a shock . one of the central questions in star formation is whether star formation is bimodal " , i.e. , whether high - mass and low - mass stars form in a different way @xcite . analysis of recent observations suggests that accretion flows around protostars of all masses can be explained within the standard model of star formation as a combination of a thin disk inside a rotationally flattened envelope ( kumar & grave 2007 ; molinari et al . 2008 ; e. keto & q. zhang 2009 , in preparation ) . of course , once a protostar gains the mass and temperature of an o star , the formation of an h ii region within the accretion flow introduces new phenomena @xcite . yet not all molecular clouds form massive stars . in a comprehensive survey , @xcite found that clouds that form low - mass stars are uniformly distributed throughout the galactic disk , but those that form the clusters of the most massive stars , o stars capable of producing significant h ii regions , are associated with the galactic spiral arms . if the formation process is very similar ( with the addition of significant ionization for @xmath202 ) , then the difference may be found in the conditions in the molecular clouds . the observations of g20.08n reported here and of g10.6 - 0.4 reported previously @xcite suggest that one difference is that the molecular clouds surrounding young clusters of o stars are in a state of overall collapse whereas in star - forming regions without h ii regions we see only localized collapse . in g20.08n the h ii regions a , b , and c are surrounded by a common molecular cloud of radius @xmath203 pc and mass @xmath204 , which is rotating with a velocity of @xmath205 @xmath19 and contracting with an inward velocity of @xmath29 @xmath19 . this rotation and contraction constitute a large - scale accretion flow . the velocity of the inward flow is about equal to the rotational velocity implying that the gas is approximately in free - fall and not constrained by centrifugal force . within this larger flow are at least two smaller accretion flows around h ii regions a and b. the molecular core around h ii region a is bright enough to be detected in hot - core molecules in emission at subarcsecond angular resolution . the radius and mass of this core are 0.05 pc and @xmath206 . accretion in the core is indicated by rotation at a velocity of 3 or 4 @xmath19 and contraction of @xmath29 @xmath19 . we do not detect molecular emission around h ii region b , but the accretion inflow is inferred from nh@xmath0 absorption that is redshifted by @xmath207 with respect to the systemic velocity . in contrast , star - forming regions that contain only low mass stars or even stars as massive as type b do not appear to have this global collapse of the entire parent cloud . the observations of these regions suggest only localized accretion flows within individual clumps . for example , star - forming regions such as iras 19410 + 2336 @xcite , iras 05358 + 3543 @xcite , and afgl 5142 @xcite , have bolometric luminosities of at most a few times @xmath208 @xmath10 , consistent with type b protostars . there are a number of cores of size similar to those in g20.08n , but there is no reported evidence of a larger parsec - scale , accretion flow . in an analysis of a recent numerical simulation , @xcite also find that the formation of massive stars or clusters is associated with large - scale collapse involving thousands of @xmath8 and accretion rates of @xmath209 . in contrast , low- and intermediate - mass stars or clusters in their simulation are associated with isolated accretion flows that are a factor of 10 smaller in size , mass , and accretion rate . if global collapse of the host molecular cloud is necessary for the formation of o stars in clusters , then the association of o stars with galactic spiral arms may imply that compression of giant molecular clouds as they pass through galactic spiral arms may be the primary mechanism for initiating global collapse @xcite . the low - mass star - forming regions found by @xcite to be spread throughout the galaxy may not need such large - scale compression . the orientations of the large cloud - scale accretion flow and the core - scale flow around h ii region a are similar and the flows could be continuous ( we do not know the orientation of the flow around h ii region b because we can not detect the surrounding molecular emission ) . the molecular core around h ii region a contains only a few tens of @xmath8 , similar to the mass of a single o star . if the core is to form one or more o stars at less than 100 @xmath67 efficiency , its mass must be resupplied by the larger scale accretion flow . resupply is also suggested by the short dynamical or crossing timescale , given by the ratio of the size to the infall velocity , @xmath210 yrs . if the core is to last more than this , it must be resupplied by the larger scale flow , which has a dynamical timescale of @xmath211 yrs . if the core is not resupplied then the growing protostar may simply run out of gas before reaching the mass of an o star . a similar process of resupply is suggested in recent theoretical work . analytic arguments show that as an unstable cloud fragments , there should be a continuous cascade of mass from larger to smaller fragments as well as a cascade of kinetic energy @xcite . in recent numerical simulations of high mass star formation , @xcite find that most of the mass is supplied from outside a 0.1 pc core around the protostar . @xcite and t. peters et al . ( 2009 , in preparation ) also find in their simulations that as a massive core is consumed from the inside by an accreting protostar , the core continues to accrete mass from the outside . the accretion rate within the small - scale flow around h ii region a is @xmath212 to @xmath213 . similar accretion rates are reported for flows around other h ii regions such as g10.6 - 0.4 , @xmath214 yr@xmath88 @xcite , and w51e2 , @xmath215 to @xmath216 yr@xmath88 @xcite . in contrast , the accretion rates estimated for the cores in these msfrs without bright h ii regions are generally lower by one or two orders of magnitude , @xmath217 @xcite . a high accretion rate is necessary to form an o star . because accreting massive protostars begin core hydrogen burning well before reaching the mass of an o star , they evolve essentially as main - sequence stars of equivalent mass . numerical simulations of stellar evolution that include accretion predict that unless the accretion rate is high enough , a growing massive protostar will evolve off the main - sequence and explode before it reaches the mass of an o star @xcite . for a protostar to gain the mass of an o star , the rate at which accretion supplies fresh hydrogen to the growing protostar must be greater than the rate at which the star burns the hydrogen . at the upper end of the mass spectrum , @xmath218 , this rate is @xmath219 . the accretion flows in g20.08n are capable of supplying gas at the rate necessary to form massive o stars . in msfrs without o stars , the accretion rates may be too low for the stars to achieve the mass of an o star within their hydrogen - burning lifetimes . if the flow in g20.08n is continuous , then the observations show that the flow spins up as it contracts . ignoring projection effects , the magnitude of the specific angular momentum on the large scale is @xmath220 pc at @xmath221 pc , from the vla - d nh@xmath0 data , while from the vex sma data it is @xmath222 pc at @xmath223 pc around h ii region a. from these estimates @xmath224 % of the specific angular momentum in the large - scale flow is lost . in previous observations of g10.6 - 0.4 ( keto 1990 ) , we found that 97% of the angular momentum in that flow is lost between 1.5 and 0.02 pc . evidently , angular momentum is efficiently transferred outward and does not prevent collapse of the cloud . magnetic fields may be important in this process ( e.g. , * ? ? ? although there are no observations of the magnetic field in g20.08n , the field direction has been mapped in another msfr with hc h ii regions , w51e1/e2 . here , observations show that the magnetic field is uniform on the larger scale , 0.5 pc @xcite , while on the smaller scale , 0.03 pc , of the accretion flow onto w51e2 , the field is pinched into an hourglass shape with the accretion flow at the waist @xcite . thus ordered field lines extend from the h ii region - scale accretion flow to the large - scale molecular cloud , and if the field has enough strength , angular momentum could be transferred outward by the field . however , dust polarization observations do not give a direct estimate of the field strength , and it is also possible that the field is essentially passive and just carried along by the flow . the observation that the clouds in both w51e1/e2 and g20.08n are close to free - fall collapse implies that the magnetic field is not strong enough to support the clouds . in other words , these clouds are magnetically supercritical . numerical simulations of star formation that do not include magnetic forces show that angular momentum can be transferred by hydrodynamics alone . @xcite find that at any radius , there is both low and high angular momentum gas , and that pressure forces or shock waves can redistribute the angular momentum between fluid elements . lower angular momentum gas sinks inward and displaces higher angular momentum gas outward , resulting in a net outward flow of angular momentum . whether the specific angular momentum is transferred by hydrodynamics or magnetic forces , observations and simulations show that angular momentum is not conserved as a function of radius , and does not prevent the gas from flowing continuously from large to small scales in a rotating flow . we report radio and mm observations of the molecular and ionized gas toward the o - star cluster g20.08n , made with an angular resolution from @xmath225 pc to @xmath226 pc . our main findings can be summarized as follows : 1 . we find a large - scale ( @xmath203 pc ) accretion flow around and into a star cluster with several o - type stars , identified by one uc and two hc h ii regions . this flow is rotating and infalling towards its center . the two hc h ii regions are surrounded by smaller accretion flows ( @xmath227 pc ) , each of them with the signature of infall too . the brightest ( toward h ii region a ) is detected in mm emission lines , and rotates in concordance with the large - scale flow . the similar orientations of the flows at small and large scales , as well as their dynamical timescales ( @xmath210 yrs and @xmath211 yrs respectively ) , and masses ( @xmath228 and @xmath229 respectively ) , suggest that , if o stars are forming in g20.08n ( as it is observed ) , then the smaller scales ought to be resupplied from the larger scales . the same result has been found in recent numerical simulations of massive star formation in clusters . the brightest hc h ii region ( a ) has a rising sed from cm to mm wavelengths and broad hydrogen recombination lines . both characteristics suggest density gradients and supersonic flows inside the h ii region . a tentative velocity gradient is detected in the recombination line emission of this source , suggesting rotation and outflow in the ionized gas at the innermost scales . h ii region a can be interpreted as the inner part of the surrounding molecular accretion flow , with the observed ionization being produced by photoevaporation . ccccccc [ tab1 ] 2003 apr 28 & vla - d & 18 28 10.384 & @xmath230 28 48.65 & 3c454.3 & @xmath231 & @xmath232 + 2003 may 13 & vla - d & 18 28 10.384 & @xmath230 28 48.65 & 3c454.3 & @xmath233 & @xmath234 + 2003 oct 09 & vla - bna & 18 28 10.384 & @xmath230 28 48.65 & 3c454.3 & @xmath233 & @xmath234 + 2003 oct 10 & vla - bna & 18 28 10.384 & @xmath230 28 48.65 & 3c454.3 & @xmath233 & @xmath234 + 2006 jun 25 & sma - vex & 18 28 10.38 & @xmath230 28 48.60 & 3c273 & @xmath235 & @xmath235 + 2006 jun 25 & sma - vex & 18 28 10.76 & @xmath230 29 27.60 & 3c273 & @xmath235 & @xmath235 + 2006 jul 06 & sma - vex & 18 28 10.38 & @xmath230 28 48.60 & 3c454.3 & @xmath236 & @xmath236 + 2006 jul 06 & sma - vex & 18 28 10.76 & @xmath230 29 27.60 & 3c454.3 & @xmath236 & @xmath236 + 2007 oct 26 & vla - b & 18 28 10.400 & @xmath230 28 49.00 & 3c454.3 & @xmath237 & @xmath234 + 2007 oct 27 & vla - b & 18 28 10.400 & @xmath230 28 49.00 & 3c454.3 & @xmath237 & @xmath234 + + cccccc [ tab2 ] h & 66@xmath37 & 22.364178 & vla - b & @xmath238 ; 3 + nh@xmath0 & @xmath90 & 23.722633 & vla - d & @xmath239 ; 236 + nh@xmath0 & @xmath91 & 23.870129 & vla - d & @xmath240 ; 8 + nh@xmath0 & @xmath90 & 23.722633 & vla - bna & @xmath241 ; @xmath242 + nh@xmath0 & @xmath91 & 23.870129 & vla - bna & @xmath243 ; @xmath244 + @xmath137co & 2 - 1 & 220.398681 & sma - vex & @xmath245 ; @xmath246 + ch@xmath0cn & 12(4)-11(4 ) & 220.679297 & sma - vex & @xmath245 ; @xmath246 + ch@xmath0cn & 12(3)-11(3 ) & 220.709024 & sma - vex & @xmath245 ; @xmath246 + ch@xmath0cn & 12(2)-11(2 ) & 220.730266 & sma - vex & @xmath245 ; @xmath246 + so@xmath3 & 11(1,11)-10(0,10 ) & 221.965200 & sma - vex & @xmath247 ; @xmath246 + co & 2 - 1 & 230.538000 & sma - vex & @xmath248 ; @xmath249 + ocs & 19 - 18 & 231.060991 & sma - vex & @xmath248 ; @xmath249 + h & 30@xmath37 & 231.9009 & sma - vex & @xmath248 ; @xmath249 + + + ch@xmath0cn & 12(7)-11(7 ) & 220.539340 & sma - vex & @xmath245 ; @xmath246 + hnco & 10(1,9)-9(1,8 ) & 220.584762 & sma - vex & @xmath247 ; @xmath246 + ch@xmath0cn & 12(6)-11(6 ) & 220.594438 & sma - vex & @xmath245 ; @xmath246 + ch@xmath0cn & 12(5)-11(5 ) & 220.641096 & sma - vex & @xmath245 ; @xmath246 + ch@xmath0cn & 12(1)-11(1 ) & 220.743015 & sma - vex & @xmath245 ; @xmath246 + ch@xmath0cn & 12(0)-11(0 ) & 220.747265 & sma - vex & @xmath245 ; @xmath246 + ch@xmath3chcn & 24(0,24)-23(0,23 ) & 221.76598 & sma - vex & @xmath245 ; @xmath246 + @xmath137cs & 5 - 4 & 231.220768 & sma - vex & @xmath250 ; @xmath249 + + ccc [ tab3 ] hii radius ( au ) & & 2530 + electron density ( @xmath251 @xmath53 ) & & 1.4 + exponent & & 1.3 + gas mass ( @xmath8 ) & & 35 - 95 + spectral type & & o7.5 + stellar mass ( @xmath8 ) & & 34 + hii mass ( @xmath8 ) & & 0.05 + ccccc [ tab4 ] h & 66@xmath37 & @xmath252 & @xmath253 & @xmath254 + h & 30@xmath37 & @xmath255 & @xmath256 & @xmath257 + ch@xmath0cn & 12(4)-11(4 ) & @xmath258 & @xmath259 & @xmath260 + ch@xmath0cn & 12(3)-11(3 ) & @xmath261 & @xmath262 & @xmath263 + ch@xmath0cn & 12(2)-11(2 ) & @xmath261 & @xmath264 & @xmath265 + . the question mark ( _ ? _ ) in the upper sideband ( _ top _ frame ) might be a superposition of lines of ch@xmath0och@xmath0 and ch@xmath0ch@xmath3cn . s _ at the center of the lower sideband ( _ bottom _ frame ) could be from vibrationally excited ch@xmath0cn . _ in the upper sideband close to the so@xmath3 might be from ch@xmath0cch . ] to the se of a. h ii region c is the more extended emission to the se of b. the color scale goes linearly from 8 to 430 @xmath30 ( the rms noise in the mm image is 2 @xmath30 ) . contours are placed at @xmath266 @xmath267 @xmath30 , the noise of the cm image . the sma - vex beam ( empty ellipse ) encircles the vla - bna beam ( filled ellipse ) at the bottom - left of the image . ] and @xmath268 uncertainty expected in the vla and sma flux measurements , respectively . the _ red dashed _ line shows the flux of an h ii region with a density gradient . the _ black dotted _ line is the flux from the warm dust component . the _ solid blue _ line is the sum of the two components . the relative contributions of free - free ( 70 @xmath67 ) and dust ( 30 @xmath67 ) to the 1.3-mm flux were estimated from the observed h@xmath20 line - to - continuum ratio , and are marked with _ diamonds_. ] ( 3,3 ) observations . emission is in _ solid _ contours and absorption in _ dashed _ contours . the star covers the h ii regions shown in fig . contour levels are at @xmath269 @xmath270 @xmath30 . a clear velocity gradient in emission is seen from one side of the absorption to the other . the lsr systemic velocity of the molecular gas is @xmath271 @xmath19 . the original maps at 0.3 @xmath19 spectral resolution were smoothed to 0.6 @xmath19 for clarity . ] ( 2,2 ) and ( 3,3 ) from the vla - d data . the cuts were done at pa=45@xmath272 ( _ bottom _ row ) and pa=135@xmath272 ( _ top _ row ) . _ dashed _ contours are absorption , _ solid _ contours are emission . contouring is at @xmath273 @xmath267 @xmath30 . only one inner satellite line is covered in the velocity range . in the sw - ne cuts ( across the major axis of the cloud ) the difference in the velocity of the emission with respect to @xmath84 @xmath19 increases closer to the position center . this can be interpreted as spin up with decreasing distance from the center . however , the same trend is present in the nw - se cuts ( along the minor axis of the cloud ) , although only in the stronger , blueshifted side of the emission . this suggests that besides rotation , radial motions in the frame of the central stars are also present . ] spectra from the vla - d observations toward the center of the absorption in g20.08n . the vertical line marks the systemic velocity ( @xmath274 @xmath19 ) . the small peak near the middle of the ( 3,3 ) absorption is due to maser emission ( confirmed in the high - resolution , vla - bna data , see [ 33maser ] ) . the absorption peaks in the main ( 2,2 ) and ( 3,3 ) lines are redshifted with respect to @xmath78 , indicating the presence of inflow in the kinematics of the parsec - scale molecular cloud . the other weak absorption component seen in the ( 2,2 ) spectrum is one of the inner satellites . the absorption component at @xmath275 @xmath19 in the ( 3,3 ) spectrum is also an inner satellite . the weaker absorption at @xmath276 @xmath19 in the ( 3,3 ) spectrum may arise from an outflow . the blueshifted absorption in the main lines closest to @xmath78 might arise from blending with the rotation seen in emission . ] from the sma - vex observations . contours are @xmath277 @xmath278 ( negative in @xmath279 , and positive in @xmath280 ) . the peak intensity is @xmath281 . the numbers in the upper left corner indicate the central lsr velocity of the channel . the two stars mark the positions of h ii regions a ( west ) and b ( east ) . ] ( 2,2 ) and ( 3,3 ) spectra toward the absorption peak of h ii region a in the vla - bna observations . the vertical line marks the systemic velocity ( @xmath289 ) of the molecular gas at scales comparable to the h ii region . the centers of the absorption lines are redshifted with respect to @xmath78 , indicating inflow of molecular gas toward h ii region a at small scales . the ( 3,3 ) absorption is broader ( fwhm @xmath290 ) than the ( 2,2 ) ( fwhm @xmath291 ) , probably caused by larger motions closer to the center . _ bottom _ panel : nh@xmath0 ( 2,2 ) spectrum toward h ii region b. the absorption , considerably fainter than for h ii region a , is also redshifted . ] rrl emission toward g20.08n . contours are at @xmath292 @xmath293 @xmath294 . although the emission is unresolved at half power , the moment 1 map hints at the presence of a velocity gradient in the ionized gas similar to that seen in ch@xmath0cn and ocs . ] ( _ top _ panel ) and h@xmath167 ( _ bottom _ panel ) lines toward g20.08n a. error bars denote the @xmath287 noise in the channels . the channel spacing is @xmath288 for h@xmath20 and @xmath295 for h@xmath167 . the flux was integrated over a 0.5 square box centered on h ii region a. ]

the reparametrization symmetries of witten s vertex in ordinary or vacuum string field theories can be used to extract useful information about classical solutions of the equations of motion corresponding to d - branes . it follows , that the vacuum string field theory in general has to be regularized . for the regularization recently considered by gaiotto et al . , we show that the identities we derive , are so constraining , that among all surface states they uniquely select the simplest butterfly projector discovered numerically by these authors . the reparametrization symmetries are also used to give a simple proof that the butterfly states and their generalizations are indeed projectors . mit - lns-02 - 297 + mit - ctp-3249 + hep - th/0202139 anomalous reparametrizations and butterfly states in string field theory + + 0.5 cm _ center for theoretical physics , massachusetts institute of technology , + cambridge , ma 02139 , usa _

hep-th0202139

the open @xcite has so far resisted all the attempts to find analytically its solutions corresponding to the true closed string vacuum or lower dimensional d - branes . on the other hand the solutions are known numerically @xcite and provide a good evidence for the sen s tachyon conjectures @xcite . the vacuum ( vsft ) @xcite , designed to describe the physics around the true tachyon vacuum has been conjectured to have purely ghost kinetic term , which then allows for some exact calculations . canonical choice of a ghost kinetic term has been recently proposed by gaiotto , rastelli , sen and zwiebach @xcite based on some earlier results of hata and kawano @xcite . it has been further discussed by okuyama @xcite . generic feature of pure ghost kinetic terms in vsft is that they allow for simple construction of solutions in terms of projectors in the matter part of the algebra . it has been proved that such solutions indeed have the right ratios of tensions to be interpreted as d - branes @xcite . to determine however the absolute normalization turned out to be quite problematic . as we will show , on general grounds one should expect some sort of divergences in the normalization of the action . to deal with the problem , the authors of @xcite added a small term to the kinetic operator proportional to @xmath0 . with this regularization they observed that the numerical solution in the limit of removed regulator approaches a new type of a projector , called butterfly projector the poetic name comes from the shape of the surface which represents these kind of states in the conformal field theory . some properties of these states were also reported in @xcite . for other very recent development in the vsft see @xcite . in this paper we would like to get more insight into the problem of classical solutions and butterfly states in vsft by an extension of the analysis of our earlier work @xcite . the basic idea is to study implications of anomalous symmetries of the witten s vertex for classical solutions of the action . these techniques turn out to be remarkably useful in this new context . they show that in the vacuum one inevitably encounters some sort of singularities . for the regularized version of the theory we obtain infinite set of nontrivial identities which , as we show , are fulfilled for the simplest butterfly state . the identities are so constraining , that among all well defined surface states they uniquely pick up one particular state , which even happens to be a projector . it would be rather interesting if one could derive more identities which would then determine the projector uniquely in the the whole hilbert space , not necessarily in the subspace of surface states . it turns out to be possible to express any well defined surface state as a reparametrization of the vacuum in the form [ ufek ] u_f = e^k , which greatly simplifies calculations of some star products . here @xmath1 stands for some linear combination of the generators @xmath2 , which are known to act as exterior derivatives in the algebra . this fact has been shown in a recent paper @xcite for wedge states . it allowed to determine the star products of wedge states purely algebraically , without thinking much about gluing surfaces and various coordinate systems . particularly interesting class of states is the one of the form _ e^k . if the limit exists , then it gives states which are invariant under the reparametrization @xmath1 , i.e. they are annihilated by @xmath1 . these states are natural candidates for being projectors . for example the simplest butterfly surface state [ butt2 ] e^-l_2 = _ - e^k_2 is a projector . on the other hand the opposite limit [ butt2f ] e^+l_2 = _ + e^k_2 . is not a projector , we call it a ` false butterfly ' state . these kind of states ( [ butt2],[butt2f ] ) can be naturally generalized by replacing @xmath3 with @xmath4 . the paper is organized as follows : in section 2 we study the consequences of the anomalous symmetries of the witten s three vertex for the classical solutions of the ordinary , vacuum and regulated vacuum theories . we derive an infinite set of identities , which as we show are obeyed by the simplest butterfly state . we then show in section 3 , that the butterfly states and their generalization are indeed projectors . we manage also to calculate the star product @xmath5 for arbitrary @xmath6 $ ] . in section 4 we prove the uniqueness of the solution of the set of identities among the surface states . as a corollary we obtain a formula ( [ ufek ] ) . appendices contain some useful formulas for the exponentials of the virasoro operators , conservation laws for butterfly states and some numerical results for illustrative purposes . related work on butterfly states with some small overlaps with our work has been done independently by gaiotto , rastelli , sen and zwiebach and should appear simultaneously in @xcite . to start with , we shall recall first some identities for the tachyon condensate found in our earlier work @xcite . the action of witten s @xcite takes a nice form of chern - simons action @xmath7=-\frac{1}{\alpha ' g_o^2 } \left ( \frac 12 \aver{\phi , q \phi } + \frac 13 \aver{\phi,\phi*\phi } \right),\ ] ] where @xmath8 is the usual string brst operator . the star multiplication is given by a three vertex @xmath9 @xcite which satisfies various identities @xcite . for our purposes the most important are [ ansyms ] _ i=1 ^ 3 ( l_-n^x(i ) - l_n^x(i ) ) & = & 3 k_n^x , + _ i=1 ^ 3 ( l_-n^gh(i ) - l_n^gh(i ) ) & = & -3 k_n^x , + _ i=1 ^ 3 ( j_-n^(i ) + j_n^(i ) ) & = & 3 ( h_n^gh + 3_n,0 ) , valid for @xmath10 even and @xmath11 and @xmath12 denote matter or ghost virasoro operators and @xmath13 are ghost current generators . the constants @xmath14 and @xmath15 for @xmath10 even are given explicitly by k_n^x & = & ( -1)^n2 , + h_n^gh & = & - ( -1)^n2 . taking the variation of the action ( [ action ] ) under the infinitesimal variations of the @xmath16 [ variations ] & = & ( l_-n^x - l_n^x - k_n^x ) , + & = & ( j_-n + j_n - h_n^gh - 3_n,0 ) we see that they leave the cubic term in the action invariant . on the other hand we know that the total action should also be invariant , as long as @xmath16 satisfies equations of motion . combining these two facts we get from the kinetic term [ anids ] [ q , l_n^x ] & = & -k_n^x q , + [ q , j_n ] & = & h_n^gh q . these identities in the siegel gauge simplify to [ anidssiegel ] c_0 l_n^x & = & 1n k_n^x c_0 l_0^tot , + c_0 ( n j_n + l_n^tot ) & = & -h_n^gh c_0 l_0^tot , where @xmath17 is the total virasoro generator . they can be further rewritten as [ anidssiegel2 ] c_0 l_n^x & = & + ( -1)^ k , + c_0 l_n^gh & = & - ( -1)^ k , where @xmath18 and @xmath19 are virasoro operators of the twisted ghost conformal field theory @xcite with central charge @xmath20 . note that the operators @xmath21 are singlets under the @xmath22 symmetry in the siegel gauge , discussed in @xcite . the basic idea of the vacuum ( vsft ) is that when the witten s is expanded around the nonperturbative vacuum , its action can be again written in the form @xmath23=-\kappa_0 \left ( \frac 12 \aver{\psi , \qq \psi } + \frac 13 \aver{\psi,\psi*\psi } \right),\ ] ] where @xmath24 is the fluctuation around the new vacuum after suitable reparametrization and @xmath25 is some new brst - like operator which is nilpotent , acts as a derivation of the algebra and is conjectured to be purely ghost . recently it has been argued @xcite , that there are certain singular reparametrizations which lead naturally to one particular kinetic operator [ q - grsz ] = c_0 + _ k=1^(-1)^k ( c_2k + c_-2k ) . the major advantage of vsft with the purely ghost kinetic operator @xmath25 is that it allows for constructing d - branes in terms of projectors of the algebra . they have been shown exactly , using the methods of boundary conformal field theory and fock space oscillators , to have the right ratios of tensions @xcite . what is still missing , is to show that there is a solution corresponding to the original unstable d25-brane with the right tension . we will see that the analysis of the constraints analogous to those discussed in the previous subsection can shed some light on the nature of this d25-brane solution . let @xmath26 be a stationary point of the action ( [ vsftaction ] ) , i.e. a solution of the equations of motion @xmath27 . then the identities read [ anidsvsft ] [ , l_n^x ] & = & -k_n^x , + [ , l_n^gh ] & = & k_n^x , + [ , j_n ] & = & h_n^gh . for a generic @xmath25 which is purely ghost , the first equation implies that = 0 . it seems therefore to show that the tension of the d25-brane and of any other soliton is zero . the same conclusion follows for the particular choice ( [ q - grsz ] ) also from a combination of the second and third identity of ( [ anidsvsft ] ) in the siegel gauge . so , does this mean that the vsft can not describe finite tension d25-branes ? no , the argument merely shows that for the soliton solutions things become singular and one must be very careful . in the scenario advocated in @xcite for the emergence of purely ghost kinetic operator in vsft , the constant @xmath28 in front of the action becomes naturally singular . when the action is carefully regularized we may hope to get the right finite answer . one might be curious whether the above argument definitely rules out the possibility that one day we find completely regular vsft action with purely ghost kinetic term and with finite action solutions . as a possible example of what could happen , the following toy version of vsft has been proposed by @xcite . suppose we start with a string field of ghost number zero and consider an operator @xmath29 , which is a dimension zero primary of ghost number three . take an action of the form [ ghn0sft ] s = [ 12 ( y(i)+y(-i ) ) + 13 ( y(i)+y(-i ) ) ] . from the argument in @xcite it follows that for projectors which are surface states in the total virasoro algebra ( the sliver for example ) , the action is finite and equal to @xmath30 . so where does the argument breaks down ? it is quite simple to find few places in the case of the butterfly projector = e^-12 l_-2 , discovered in @xcite and discussed in more detail in section [ butterflies ] and in @xcite . easy check shows , that for example the conservation laws ( [ ansyms ] ) break down . indeed , for @xmath31 & & _ j=1 ^ 3 ( l_-2^x(j ) - l_2^x(j ) ) ( y(i)+y(-i ) ) = + & & 3 k_2^x ( y(i)+y(-i ) ) = - , where @xmath32 is the central charge in the matter sector . in section [ butterflies ] we will see that some things are actually ill defined , for instance the inner product @xmath33 and even the star product @xmath34 itself . to conclude , our argument rules out possibility of finite action solutions which are sufficiently regular , it allows however for less well behaved solutions as are the projectors . in @xcite the following regularized action of the vsft in the siegel gauge has been proposed [ regvsftaction ] s_a[]=-_0(a ) ( 12 + 13 ) . the original gauge fixed vsft is recovered in the limit @xmath35 . it is perhaps curious that the other limit @xmath36 corresponds to the ordinary witten s theory . one might therefore hope that both actions share some features , and that is indeed true . let us write the constraints on the solutions to the equation of motion . since we are in the siegel gauge , we may write only identities based on variations @xmath37 , which preserve the gauge [ anidsregvsft ] c_0 l_n^x & = & ( -1)^ a k , + c_0 ( n j_n + l_n^gh ) & = & - ( -1)^ a k , where k = c_0(1 + 1a l_0 ) . it was observed in @xcite that in the limit @xmath35 the numerical solution of the equations of motion representing the d25-brane seems to approach the following state [ grszsol ] = e^-12 ( l_-2^x + l_-2^gh ) c_1 . we will now show that this solution is compatible with the infinite set of identities ( [ anidsregvsft ] ) even for finite @xmath38 . an immediate question is then , whether the solution ( [ grszsol ] ) is a unique solution of this set of identities . we will prove in section [ uniqueness ] , that it is indeed unique among the surface states in the deformed cft . it is most likely not unique among non - surface states , since for example in the limit @xmath39 corresponding to ordinary theory , we have another solution . to deal with non - surface states , one would almost certainly need more identities , which would differ from ( [ anidsregvsft ] ) by taking expectation values of products of more operators . to check the identities we have to first regularize the solution ( [ grszsol ] ) = e^ s l_-2^x + s l_-2^gh c_1 keeping @xmath40 and @xmath41 , and let @xmath42 and @xmath43 approach @xmath44 . notice , that the regularized solution is not necessarily a solution of the regularized action . we are taking fixed finite @xmath38 and want to show that the identities are obeyed in the limit @xmath45 . let us focus first on the identity = - , which follows from ( [ anidsregvsft ] ) . by a straightforward calculation , whose details are given in appendices [ apptransf ] and [ reggrsz ] , we see that we have to require that in our limit [ fixedratio ] = is kept fixed . remarkably , making this single adjustment suffices to show that all other identities ( [ anidsregvsft ] ) are satisfied . using the formulas collected in appendix [ reggrsz ] we get [ ak ] ak = ( a c_0 + c_0 l_0^tot ) ()^ ^2 . here it is important , that we take first the limit @xmath45 for fixed @xmath38 , and only at the end when we get @xmath38 independent ratios we may send @xmath35 , if we wish so . we have also [ buttid ] c_0 l_2k^x & & ( -1)^k ()^ ^2 , + c_0 l_2k^gh & & -(-1)^k ()^ ^2 . the identities ( [ anidsregvsft ] ) now immediately follow from ( [ ak ] ) and ( [ buttid ] ) . let us make one more remark about the logic here . we have shown that with a finite regularization @xmath38 of the vsft action , the butterfly state exactly solves the infinite set of identities . in order to prove that , we had to regularize the state as we did . we have not shown , that it exactly obeys the equations of motion of the regularized theory . we will see in section [ detailedbutt ] however , that the regularized butterfly approximately solves the equations of motion of the regularized vsft , provided we identify @xmath46 and let it depend on @xmath38 in a definite way . butterfly states were introduced first in @xcite as surface states associated to the conformal mappings is related to the parameter @xmath47 of gaiotto et al . @xcite by @xmath48 . ] [ defbuttmap ] f_r(z ) & = & ( z ) + & = & ( z ) . various descriptions of the surface states have been developed recently in @xcite . let us review some definitions and properties we will need later . for a conformal field theory ( cft ) defined on upper half plane , take a map @xmath49 which maps real axis to real axis and the upper unit half disk one to one in the upper half plane . the surface state associated to the map @xmath50 is defined as a linear functional on the hilbert space , which is just a bra vector , satisfying [ defsurf ] = = , . here @xmath51 is the operator corresponding to the state @xmath52 through the operator state correspondence , and @xmath53 denotes its conformal transformation . the operators u_f = e^_n0 v_n l_n satisfying f ( z ) = u_f ( z ) u_f^-1 , form a representation of the conformal group on the hilbert space of the worldsheet cft . the coefficients @xmath54 in front of the virasoro operators @xmath55 can be determined by solving certain recursive equations . the vector field @xmath56 is also determined up to a constant by the equation @xmath57 . for recent review and discussion of related issues the reader is referred to @xcite . using conformal map h(z ) = , which maps the upper half plane with a coordinate @xmath58 one to one onto the unit disk with a coordinate @xmath59 we can evaluate the correlator ( [ defsurf ] ) as _ uhp = _ disk . the image of the unit upper half disk under the map @xmath60 is called local coordinate patch . for the simplest butterfly surface state corresponding to the map ( [ defbuttmap ] ) with @xmath61 the shape of the local patch is indicated in fig . [ figbutt2 ] . recall , that in this coordinate system the local coordinate patch for the sliver state looks just as a sliver . there is another coordinate system with a coordinate @xmath62 , in which the local patch is mapped to the right half disk . in this coordinate system the star product of surface states is represented simply by gluing the surfaces with the local patch detached . for the butterfly state , the part of the surface without the local patch looks as a butterfly , hence the name . for the sliver the surface in this coordinates looks as a helix of an infinite excess angle . for general @xmath63 , the butterfly states can be expressed explicitly as elements of the verma module ( representation space of the virasoro algebra ) [ buttr2 ] = u_r^e^- l_-2 , where @xmath64 are well known from the studies of wedge states @xcite . they are given by u_r^= e^2 a^ , where [ defa ] a= -12 l_0 + _ k=1^ l_2k and the dagger denotes the bpz conjugation @xmath65 . similar vector field has been also considered recently in @xcite . from the proof we are going to present in the next subsection it follows , that the family of butterfly projectors can be generalized also to include surface states = e^ l_-2k associated to conformal maps f(z)=. considered already in @xcite . an example of the corresponding surface for @xmath66 is given at fig . [ figbutt4 ] . these are direct generalizations of the state ( [ buttr2 ] ) for @xmath61 . once we establish that the states with @xmath61 are projectors , then the case @xmath67 follows by considering finite reparametrizations and will be discussed separately in subsection [ family ] . let us assume for simplicity , that our cft has overall central charge @xmath68 . the complications for nonzero central charge , as in the matter plus twisted ghost cft can be dealt with separately . we start then with the following simple identity [ etok2k ] e^k_2k = e^-2k l_-2k (2k)^- l_0 e^ 2k l_2k discussed more in appendix [ appgroup ] . using this formula we can write the butterfly state in a regulated form as e^(-)^k+1k_2k = e^2k l_-2k , where @xmath69 . the advantage of this expression for the butterfly is that @xmath4 are ( exterior ) derivations of the star algebra and thus we have in general [ transfstar ] e^k ( * ) = e^k * e^k , for any two string fields @xmath70 and @xmath71 . to prove that the butterfly is a projector we can write e^ l_-2k * e^ l_-2k = _ e^(-)^k+1k_2k * e^(-)^k+1k_2k = _ e^(-)^k+1k_2k u_3^. now formally holds [ reparu3 ] _ e^(-)^k+1k_2k u_3^= _ e^ l_-2k (2k)^- l_0 e^- l_2k u_3^ = e^ l_-2k , since all the terms in @xmath72 with nonzero level get multiplied with vanishing factors . there is one important potential flaw in this argument however . we could have repeated all the argument without bothering about the @xmath73 factor . this would imply the incorrect statement that @xmath74 is also a projector . what goes wrong in this case is that @xmath75 is divergent in the level expansion . when regulated , the divergences are just cancelled by the vanishing factors and we end up with something finite . to complete the proof we have to show that [ buttinvwedge ] _ e^-2k l_2k u_3^ = e^- l_2k u_3^ has finite unambiguous limit . for @xmath76 it can be proved by explicit calculation ( see below ) , for general @xmath77 we have to use the gluing theorem . in general for two conformal maps @xmath50 and @xmath78 the state @xmath79 is again a surface state . it can be calculated as follows : from the surfaces corresponding to the states @xmath80 and @xmath81 in the @xmath82 coordinate we detach the local coordinate patches and glue the remaining surfaces together along the segment @xmath83 . the part of the surface corresponding to @xmath81 represents now new coordinate patch and we may map it back to the canonical half - disk . the resulting surface is then the surface representation of the state @xmath79 . this procedure has been applied to the case of wedge states in @xcite , where it was shown to agree with other methods . in our present case we have to glue a wedge with opening angle @xmath84 ( after detaching the local coordinate patch ) with a surface associated to the map [ wrongbuttmap ] f(z)=. two examples of these surfaces are in fig . [ figbuttinv2 ] and [ figbuttinv4 ] . since this map is regular at the midpoint @xmath85 , which in the upper half plane coordinates is located at @xmath86 the associated surface is regular at this point . this is also seen at the pictures fig . [ figbuttinv2 ] and [ figbuttinv4 ] but is manifestly not true for the surfaces associated to butterfly states at fig . [ figbutt2 ] and [ figbutt4 ] . the midpoint @xmath85 is the only point where the wedge corresponding to @xmath87 has a singularity . gluing the wedge with the ` wrong sign ' butterfly surface associated to the map ( [ wrongbuttmap ] ) results in a surface with simple singularities which are just cuts and conical singularities which can be smoothed out . this proves that the state ( [ buttinvwedge ] ) is finite and well defined state and concludes thus the proof . on the other hand , changing the sign in the exponent of the butterfly states would violate the last argument . we would have to glue two surfaces , one which has a cut originating at the midpoint @xmath85 and other which has a conical singularity there . this is a source of divergences which cancel the vanishing factors in ( [ reparu3 ] ) . for instance the product like @xmath88 is finite and well defined but does not lead to @xmath74 . using the conservation laws ( [ butconslaw ] ) and the methods of @xcite we may find [ ubutt ] u_r e^12 ^2l_-2 = ()^- ()^- ()^ e^12 ^2 l_-2 . we have written the formula for general central charge @xmath89 , now we shall restrict to the case @xmath68 however . in general , if we know explicitly the surface state @xmath90}^\dagger\ket{0}$ ] we may calculate @xmath91}^\dagger\ket{0}$ ] using [ phigf ] = i i f i g^-1 i. this can be easily established by calculating @xmath92 for an arbitrary local field @xmath93 . there are two different ways of bringing it to the form @xmath94 whose equivalence requires ( [ phigf ] ) . note that with our definition the functions @xmath95 $ ] and @xmath96 $ ] are defined modulo @xmath97 transformation acting from the left , which is immaterial for the definition of the surface states . there is one preferred choice however and this is the reason why the first factor on the right hand side of ( [ phigf ] ) is @xmath98 . with this choice we have [ uudag ] u_f u_g^= u_^u _ as the operator statement and for the corresponding maps f i g^-1 i = i ^-1 i . the equation ( [ uudag ] ) has been used in @xcite for the algebraic proof of the gluing theorem . the maps @xmath95 $ ] and @xmath96 $ ] were constructed as series in an auxiliary parameter @xmath47 which controls the contribution of higher levels and which has to be sent to @xmath99 at the end of the calculation . taking f(z ) = (z ) , g(z ) = and using ( [ ubutt ] ) , ( [ phigf ] ) and ( [ etol2 ] ) we find easily [ phifgappl ] ( z ) & = & + [ phigfappl ] ( z ) & = & . specializing to @xmath100 and @xmath101 we find by expanding ( [ phigfappl ] ) in series and solving iteratively for the generating vector field as in @xcite e^ l_2 u_3^ = e^ l_-2 + l_-4 + l_-6 + l_-8 + . this supports our claim that it is a well defined state with finite coefficients and therefore the butterfly state @xmath102 is a true projector . let us now look at what happens to the wrong sign butterfly states under the star product . the function ( [ phigfappl ] ) clearly ceases to be holomorphic at @xmath103 in the limit @xmath104 for @xmath101 and the coefficients of the corresponding surface state diverge . multiplying the state by the middle factor in ( [ etok2k ] ) which is now (1-()^4)^ ()^l_0 has the effect of changing @xmath105 in ( [ phigfappl ] ) , which in our limit keeps the domain of holomorphicity around zero from shrinking to zero size . up to an irrelevant overall factor the function ( [ phigfappl ] ) with this substitution has a well defined limit . finally multiplying the state with @xmath106 gives a surface state associated to the map . using the recursive relations we determine again the generating vector field and the coefficients in front of the virasoro generators . the result is [ wrongbuttprod ] e^ l_-2*e^ l_-2 = e^ l_-2 + l_-4 - l_-6 + l_-8 + . results from independent numerical check at level 16 are summarized in appendix [ levelexpprod ] . note also that this state is annihilated by @xmath3 just as the false butterfly @xmath107 itself . if we keep multiplying these states we get a whole commutative subalgebra annihilated by @xmath3 , which resembles in many respects the subalgebra of wedge states , which are characterized by the fact , that they are annihilated by @xmath108 . the calculation we have just performed can be easily generalized . from the formulas ( [ etok2k ] ) and ( [ phigfappl ] ) we can calculate the star product of regularized butterflies . the star product e^-12 ^2l_-2 * e^-12 ^2l_-2 = u_f_3,^ is a surface state associated to the map [ f3gamma ] f_3,(z)= and therefore it looks as u_f_3,^ = e^v_2 l_-2 + v_4 l_-4 + , where the first two coefficients are v_2 & = & , + v_4 & = & . the multiple star products _ ( r-1 ) = u_f_r,^ correspond to maps which look just as ( [ f3gamma ] ) with the two factors of @xmath109 replaced by @xmath63 . from the formula then follows that the limit of the repeated multiplication of @xmath107 with itself is just the true buttefly state @xmath110 . let us consider now in detail the limit @xmath111 , in which the state @xmath112 approaches the butterfly projector . the product of two such states is a surface state associated to the map f(z ) = ()^ [+ (- ) + o(- ) ^2 ] . due to the fact that the second term in the bracket is just a change of the first under infinitesimal scaling proportional to z ( ) = we find easily e^s l_-2 * e^ s l_-2 & = & [ 1- (1-)(s+ ) l_-2 ] e^ s l_-2 + o(s+)^2 + & = & [ 1 + (1-)(s+ ) l_0 ] e^ s l_-2 + o(s+)^2 . we thus see that @xmath113 solves the equation of motion of ghost number zero * = (1+a^-1 l_0 ) in the limit @xmath35 , provided we identify [ as ] a^-1 = (1-)(s+ ) . this can be directly extended to the case of standard regularized vsft , where up to terms of order @xmath114 the solution is = e^s(l_-2^x + l_-2^gh ) c_1 with the relation between @xmath38 and @xmath42 being the same as in ( [ as ] ) . the normalization factor @xmath115 is for @xmath116 finite and calculable . we have thus fully proven the conjecture of @xcite for which we gave a lot of evidence in section [ constrregvsft ] . now we would like to prove our claim from section [ constrvsft ] that the star product @xmath34 is ill defined . to demonstrate it , we shall calculate the even level part of the regulated expression l_-2^x e^k_2 * e^k_2 and show that it is divergent in the limit @xmath69 . from ( [ butconslaw ] ) follows l_-2^x e^k_2 = -^2 2 ( k_2^x + 2 ) e^k_2 . using the @xmath117 conservation law and ( [ transfstar ] ) we get & & l_-2^x e^k_2 * e^k_-2 + e^k_-2 * l_-2^x e^k_-2 = + & & = -^2 2 [ k_2^x + 2- 3k_2 ] e^k_2 u_3^. using the relation ( [ reparu3 ] ) we arrive to _ [ l_-2^x e^-2l_-2 . & * & e^-2l_-2 + + e^-2l_-2 & * & . l_-2^x e^-2l_-2 ] = c (- - ) e^- l_-2 demonstrating that the star product in the square bracket is indeed divergent in the limit @xmath69 . to show that also the whole family of butterfly states = u_r^e^- l_-2 are projectors is rather easy . important observation is that @xmath118 given by ( [ defa ] ) annihilates the simplest butterfly a e^- l_-2 = 0 , thanks to the identity _ k=1^ = . therefore , using the formula e^-(a - a^ ) = u_1+e^^u_1+e^- from @xcite we get = e^-(r-1 ) ( a - a^ ) e^- l_-2 . the exponent @xmath119 is an exterior derivative in the algebra ( it is a linear combination of @xmath120 s ) and therefore one may use ( [ transfstar ] ) to show that all members of the butterfly family are projectors and all arise as reparametrizations of the vacuum . special cases are the @xmath121 limit , which gives the sliver , and the @xmath122 limit leading to an interesting state which resembles the identity of @xcite = (_n=2^e^- l_-2^n ) e^-l_-2 . it differs from the identity only in the sign of @xmath123 in the exponent . in section [ constraints ] we have shown that the simplest butterfly projector [ simplbutt ] = e^-12 ( l_-2^x + l_-2^gh ) c_1 does obey all the quadratic identities , which should be satisfied by all classical solutions of the regularized theory in a siegel gauge . an immediate question that arises is , to what extent is this solution unique . we are going to prove , that in the space of surface states the identities are so constraining that the solution ( [ simplbutt ] ) is indeed a unique solution . it turns out , that we do not need to study the whole set of identities , we will see that already in the matter sector alone the constraint [ qidsimpl ] l_2k ~(-1)^k is strong enough to disqualify all other surface states as possible solutions . the symbol @xmath124 means up to a possibly divergent normalization constant independent of @xmath77 . before we start the actual proof , we can easily demonstrate that the generalized butterfly projectors certainly do not obey the identities . indeed , for @xmath125 even e^s l_n l_m e^s l_-n = 0 m \\{ ,-n,0,n,2n,3n , } but e^s l_n l_n e^s l_-n = ( n^3-n ) s (1-(ns)^2)^- -1 0 . the reason is that if @xmath126 is not an integer multiple of @xmath10 , after normal ordering of the operators all the terms will have their @xmath127 weight nonzero , equal to @xmath126 modulo @xmath10 and therefore the expectation value will vanish . let us now describe a general method of testing the identity ( [ qidsimpl ] ) applicable to any surface state = u_f^ associated to a given conformal map @xmath49 . we are going to calculate the matrix elements [ ssevoft ] u_f l_-m u_f^ = u_f t(z ) u_f^. from the geometric discussion in section [ butterflies ] the state @xmath128 , which formally exists in the verma module for any @xmath49 holomorphic near origin , is meaningful in the cft sense only if @xmath49 maps one to one the upper unit half disk into the upper half plane . the projectors which seem to be characterized by the fact that @xmath129 are thus limiting cases . the identity and the false butterflies with @xmath130 are also limits of regular surface states . we shall therefore restrict to the regular surface states only and later discuss these interesting limits . we also have to require that u_f t(z ) = ft(z ) u_f is a well defined state in the level expansion . this is not automatic , since the left hand side is a series in positive and negative powers of @xmath58 . negatively moded virasoro generators when commuted to the left give rise to infinite series in mostly positive powers of @xmath58 , which then should sum to coefficients m . the convergence of these verma module coefficients follows from our assumptions about the map @xmath49 provided @xmath131 , where @xmath132 is the radius inside which the map is holomorphic and one to one into the upper half plane . since we need also @xmath133 being well defined , this implies @xmath134 . in order for @xmath135 to be absolutely convergent we need thus |z| ( r^-1,r ) r>1 . from what has been said now follows that the right location of the contour in ( [ ssevoft ] ) is inside the annulus of radii @xmath136 and @xmath137 . to calculate the integrand we need to use the transformation properties of the energy momentum tensor @xmath138 ft(z ) = (f(z))^2 t(f(z ) ) + s[f , z ] , where s[f , z ] = is the schwarzian derivative . we can then calculate [ utu ] & & u_f t(z ) u_f^ = ft(z ) u_f u_f^ = e^c ft(z ) u_^ = + & & = e^c i i ft(z ) = s [ i i f , z ] e^c , where the function @xmath139 $ ] was introduced and discussed in section [ detailedbutt ] . the central charge @xmath89 appearing in the normalization constant @xmath140 might in principle be different from that of the energy momentum tensor we insert , but that is just a trivial extension of our formulas . before we move on to the proof , let us have a look on two examples of the butterfly and sliver projectors . for regularized butterfly state associated to the function f_s(z ) = the function @xmath141 $ ] is given by ( z ) = . and i i f(z ) = . the schwarzian derivative is then s[i i f , z ] = - ( 1 + ) ( 1- ) . we thus see that the contour integral in ( [ ssevoft ] ) receives contribution from poles at @xmath142 . up to a normalization constant @xmath143 it is equal to s[i i f , z ] = c ( 2s)^k ( + ) , for @xmath144 even . for @xmath126 odd the integral vanishes . although the above calculation is rather straightforward , these results can be obtained in a much simpler way by working directly with the virasoro generators and conservation laws as we discuss in the appendices [ apptransf ] and [ reggrsz ] . notice also that in the limit @xmath145 the integral behaves like @xmath73 and we readily confirm the identities ( [ qidsimpl ] ) . the situation for the sliver is more complicated . even if we regularize the sliver by considering wedge state associated to the map [ wedgemap ] f_r(z ) = ( z ) for large but finite @xmath63 , this function still has branch cuts starting at @xmath146 and going all the way to infinity . therefore the expression @xmath135 does not absolutely converge for any values of @xmath58 . in the limit @xmath147 we can however circumvent this problem quite unambiguously as we shall see . the schwarzian derivative for the wedge states reads s[f_r , z ] = and the function @xmath148 $ ] is given by _ r(z ) = f_4-(z ) , as was found in @xcite . let us start by calculating u_f l_-m = s[f , z ] u_f + ( f(z))^2 t(f(z ) ) u_f . the contour encircles the origin and remains inside the unit circle . then we contract the expression with @xmath149 . up to the overall factor @xmath150 the first term gives [ slivid1term ] s[f , z ] , which is fine as far as concerns the integral . the second term is then s [ , if(z ) ] . note that the form of both terms can be understood as a consequence of the identity for the schwarzian derivative s[i i f , z ] = s[f , z ] + s [ , if(z ) ] , but we have chosen the stepwise derivation . the location of the contour in the second term is ambiguous . in fact it can not be deformed in any way to ensure the absolute convergence . we may take it along the unit circle , but then have to assume some additional convenient regularization for all @xmath63 . what saves us in the case of sliver is that s [ , if(z ) ] = - 2 ( ) ^2 uniformly vanishes for @xmath121 . whatever @xmath63 independent regularization we decide for , the second term in the limit will not contribute . therefore a unique contribution comes from the first term ( [ slivid1term ] ) and we find for the sliver @xmath151 [ sliverid ] l_2k = k ( -1)^k , which is clearly different from ( [ qidsimpl ] ) . the normalization factor @xmath152 , which is divergent has been also calculated in @xcite . we shall start by rewriting the set of quadratic identities ( [ anids ] ) or ( [ anidsregvsft ] ) in a compact form . consider a variation [ varcompact ] = ( t(z ) - t(-1/z ) ) _ 0 around a solution @xmath26 of the equations of motion of a given theory . here @xmath138 stands for any of the matter , ghost , twisted ghost energy momentum tensors or combinations thereof . for any of these choices we have = (2 + z ) t(z ) . the cubic term in the action is left invariant under the variation ( [ varcompact ] ) . this can be understood in two ways : first , this variation is an infinite sum of variations ( [ variations ] ) , in which the anomalies happen to cancel for a natural choice of regularization . second , the invariance of the cubic term can be derived from the overlap equations for the three vertex @xcite [ overlap ] [t^(s)(z ) - t^(s+1)(-1/z ) ] = 0 . they are strictly speaking true only for @xmath153 and with the prescription that in the first term @xmath58 approaches the unit circle from outside , whereas in the second term from interior . when contracted with regular states , these stringent conditions can be relaxed to some annulus around the unit circle . as a consequence of the equations of motion , if the cubic term is invariant under the variation , so has to be also the kinetic term . from that we find an important relation (2 + z ) t(z ) = 0 , whose unique solution is [ compid ] t(z ) = . this equation contains as much information as the original set of identities ( [ anids ] ) or ( [ anidsregvsft ] ) . the constant @xmath154 depends on the type of the energy momentum tensor and is also related to the anomaly . details depend also on whether we are in vacuum or ordinary theory . now we are going to solve the equation ( [ compid ] ) in the space of surface states discussed in section [ testingids ] . due to ( [ utu ] ) our problem reduces to the equation [ sgeq ] s[i i f , z ] = , where we denote = e^- c. although the equation ( [ sgeq ] ) looks as a third order nonlinear differential equation we can find easily all its solutions . denoting g(z ) = i i f(z ) we find the solution [ gsol ] g(z ) & = & h ( z^ ) , + g(z ) & = & h ( z ) = , where @xmath155 is an arbitrary @xmath156 map . note that for this reason the sign of the square root does not matter . the map @xmath155 is parameterized by three real parameters , just as many as the number of parameters necessary to specify the initial conditions of a third order differential equation . so we have managed to reduce the problem to the equation [ phiifeq ] i i f(z ) = h z^n , restricting for simplicity to the case @xmath157 and denoting @xmath158 . now , a bit surprisingly the equation ( [ phiifeq ] ) , for @xmath10 even at least , does not admit any regular solution . to see that , note that from ( [ phigf ] ) the left hand side of ( [ phiifeq ] ) commutes with the inversion i ( i i f ) = ( i i f ) i , but the right hand side never commutes , due to the fact that i h = h i ( -id ) can be never satisfied within @xmath156 group . what we can do however , is to find a sequence of approximations @xmath159 for @xmath160 whose limit is @xmath161 and hope that the corresponding sequence of @xmath162 s whose uniqueness we shall prove , have always the same limit . this is just a technical assumption of smooth dependence of @xmath50 on @xmath78 . the most natural sequence of approximations labelled with a continuous parameter @xmath38 is [ gndef ] g_n , a(z ) = ()^. for @xmath10 even and @xmath163 we have g_n , a(z ) = 1 + + o()^2 and this has indeed the form of @xmath164 up to the first order in @xmath165 . we would have obtained a similar result also for @xmath38 close to @xmath166 . the schwarzian derivative ( for all @xmath10 ) is s[g_n , a , z ] = - (1 + ) (1- ) and in the limit @xmath167 it clearly approaches @xmath168 . now we are going to determine the surface states , which actually solve our equation ( [ phiifeq ] ) with @xmath169 replaced by @xmath159 . there are two ways to proceed . based on our results from subsection [ butterex ] , we could guess and check that ( [ gndef ] ) requires f(z ) = and the limits @xmath167 correspond to the butterfly and false butterfly states . what about the uniqueness ? it seems that we are getting more solutions depending on @xmath10 , recall however , that different @xmath10 correspond to different @xmath154 which we take fixed in such a way , that the butterfly state @xmath170 solves the constraint . are there any other solutions ? we can prove uniqueness of the solution of the equation i i f(z ) = g(z ) for any given @xmath171 by the following reasoning : suppose there is another solution @xmath172 with @xmath173 $ ] such that i i f = i i f. since the operator corresponding to @xmath174 contains only negatively moded virasoro operators , we find @xmath175 and thus @xmath176 modulo some unimportant global rescaling . an alternative procedure is to recall from @xcite that the map ( [ gndef ] ) is generated by a vector field @xmath177 , which obeys @xmath178 . satisfying @xmath179 is generated by a vector field @xmath180 obeying @xmath178 , as can be seen from the equation @xmath181 . ] the operator which corresponds to the map @xmath182 is thus u_g_n = e^k_n . the dependence of @xmath47 on @xmath42 should be worked out separately . again , since the operator corresponding to @xmath174 acts as identity on the vacuum @xmath183 we have u_f = e^ e^k_n , where @xmath184 . from the formulas in appendix [ appgroup ] we easily find the relation between the parameters @xmath47 and @xmath42 s = 2k . as a byproduct of the above arguments we obtain an interesting statement . for any regular surface states @xmath128 associated to the maps specified above , there exists a star algebra derivative @xmath1 , which is a linear combination of the generators @xmath185 , such that e^k = e^- u_^-1^u_f = e^ u _ u_f^-1^and therefore [ byprod ] u_f = e^ e^k . this follows from the fact , that the function @xmath186 is generated by some @xmath1 and from the fact that @xmath187 is holomorphic around @xmath188 and hence the corresponding operator acts trivially on @xmath183 . the equation ( [ byprod ] ) has been checked on a number of examples like wedge states in @xcite and regularized butterflies in the present paper . we have seen that the anomalous @xmath120 symmetries , which are just consequences of the overlap equations for the string vertex , yield nontrivial information about the solutions of the ordinary or vacuum theories . in order to obtain finite action solutions in vsft with pure ghost kinetic operator , we generically need infinite normalization of the action . due to the singular nature of projectors it could happen that the normalization turns out to be finite . in any case it shows that some sort of regularization of the string field action is inevitable . solutions of the regularized version of vsft constructed by adding a term @xmath189 to the kinetic operator obey such a constraint , which among all the surface states admits unique solution which is just the simplest butterfly projector . the constraint , up to a single constant , is the same as for the ordinary open ( osft ) . in that case we know however , that we should look for the solution of the full equations of motion outside the family of surface states . the overlap equations encode all the information about the vertices . it would be interesting to see if we could derive more identities based on overlap equations for products of local operators . such identities could then uniquely specify all the coefficients in the solution @xcite , except for a single parameter . in a recent paper @xcite by gaiotto et al . the authors found an intriguing coincidence for the osft solutions . the coefficients in the matter sector agree within few percent with coefficients of a solution of a ghost number zero theory , whose equation of motion in the siegel gauge reads ( l_0 - 1)= - * . even though the solution is given purely in terms of total virasoro operators acting on the vacuum , we can still find our quadratic identities based on @xmath120 s in the matter sector . they happen to be the same as the matter part of such identities in the osft . whether one can disentangle the matter and ghost sectors and add more identities to fully prove the observed pattern remains to be seen . another issue discussed in the paper are the surface states and projectors . we have shown that @xmath120 star algebra derivatives can be useful to calculate some nontrivial star products and to find some projectors . it would be interesting to obtain full classification of all projectors and to clarify their relation to star algebra derivatives in general . i would like to thank l. rastelli , a. sen , w. taylor and b. zwiebach for useful discussions . this work has been supported in part by doe contract # de - fc02 - 94er40818 . let us take an @xmath190 algebra with a central extension satisfying the commutation relations [t_1,t_-1 ] & = & 2 t_0 + a , + [t_0,t_1 ] & = & -t_1 , + [t_0,t_-1 ] & = & t_-1 . then we have e^(t_1-t_-1 ) & = & e^- t_-1 ()^- ( 2 t_0 + a ) e^ t_1 , + e^(t_1+t_-1 ) & = & e^ t_-1 ()^- ( 2 t_0 + a ) e^ t_1 , + e^t_1 e^t_-1 & = & e^ t_-1 ( 1-)^- ( 2 t_0 + a ) e^ t_1 , + e^t_1 e^t_0 & = & e^t_0 e^ e^t_1 . these identities can be presumably established by a variety of means . we have found them by constructing explicitly an infinite dimensional representation built by successive actions of @xmath191 on some vacuum state @xmath192 which is annihilated by @xmath193 and @xmath194 and then using the method of differential equations described in @xcite . much simpler way to obtain them is by analytic continuation from analogous equations for @xmath195 group . when applied to subalgebras of the virasoro algebra through the identification t_1=1n l_n , t_-1= 1n l_-n , t_0=1n l_0 and a= , we find easily the following important relations e^k_1 & = & e^ l_-1 ()^-2 l_0 e^ l_1 , + [ etok2 ] e^k_2 & = & e^-2 l_-2 (2)^- l_0- e^ 2 l_2 , + e^k_2k & = & e^-2k l_-2k (2k)^- l_0- e^ 2k l_2k , where @xmath196 as usual . the regularized version of the simplest butterfly state @xmath197 is also a surface state , since as we know from @xcite [ etol2 ] e^s l_2 = u_. by straightforward use of virasoro commutation relations or by the general techniques of @xcite we may find various conservation laws for this state [ butconslaw ] ( l_0 - 2s l_-2 ) e^s l_-2 & = & 0 , + ( l_2 - ( 2s)^2 l_-2 ) e^s l_-2 & = & e^s l_-2 , + ( l_2k - ( 2s)^k+1 l_-2 ) e^s l_-2 & = & ( 2s)^k e^s l_-2 , k 1 . we keep the central charge @xmath89 arbitrary . applying simple method of differential equations ( see e.g. @xcite ) one can derive from these conservation laws formulas for the action of finite conformal transformation [ butfintransf ] e^v l_0 e^s l_-2 & = & e^s e^2v l_-2 , + e^v l_2 e^s l_-2 & = & (1- 4vs ) ^- ( l_-2 ) , + e^v l_2k e^s l_-2 & = & (1 - 2kv ( 2s)^k ) ^- ( l_-2 ) . by the same method , or using the formula ( [ etok2 ] ) from appendix [ appgroup ] as a shortcut , one can derive the action of a finite reparametrization generated by @xmath198 [ etoketol ] e^k_2 e^s l_-2 = ( 2- 2s 2)^- ( l_-2 ) . note , that from this formula follows for @xmath68 and @xmath199 _ e^k_2 e^s l_-2 = e^-12 l_-2 . this is quite analogous to ( [ reparu3 ] ) , in that infinite reparametrization of some set of states yields the butterfly projector . it can be proven along the same lines since the surfaces corresponding to @xmath200 and @xmath201 for @xmath199 can be glued together without gluing singularity to singularity . here we collect some useful formulas for the regularized butterfly states in a twisted conformal field theory = e^ s l_-2^x + s l_-2^gh c_1 . using ( [ butconslaw ] ) and ( [ butfintransf ] ) , one can easily establish for this state ( omitting the superscripts ) c_0 & = & ( 1 - 4s^2)^- ( 1 - 4s^2)^ , + c_0 l_0^x & = & c_0 , + c_0 l_0^gh & = & - c_0 , + c_0 l_0^tot & = & (- -1 ) c_0 , + c_0 l_2k^x & = & 26 ( 2s)^k ( + ) c_0 , + c_0 l_2k^gh & = & -2 ( 2s)^k ( + ) c_0 . in this appendix we present some numerical data showing how well the butterfly projectors behave in level expansion . we have calculated first three coefficients of the product @xmath202 up to level 16 . calculation up to level @xmath203 means , that we truncate the input state up to that level and then perform the star multiplication exactly . the results are @xmath204 for the product @xmath205 we have calculated the first three coefficients up to level 24 : @xmath206 the values for level @xmath207 were obtained in both cases by a simple fit @xmath208 . for some coefficients a different power of @xmath203 seems more accurate , but due to lack of data we refrain from more careful analysis . anyway , the numerical results are compatible with butterflies being projectors . finally as a check on our methods and formulas we present numerical calculation of the star product ( [ wrongbuttprod ] ) e^ l_-2*e^ l_-2 & = & e^ l_-2 + l_-4 - l_-6 + l_-8 + + & = & (1 + 0.3333 l_-2 + 0.0926 l_-4 - 0.0309 l_-6 + 0.0555 l_-2 l_-2 + . + & & . + 0.0154 ( l_-2 l_-4 + l_-4 l_-2 ) + 0.00617 l_-2 l_-2 l_-2 + ) . numerical calculation at level 16 gives e^ l_-2*e^ l_-2 & = & (1 + 0.3001 l_-2 + 0.0928 l_-4 - 0.0272 l_-6 + 0.0452 l_-2 l_-2 + . + & & + 0.0140 ( l_-2 l_-4 + l_-4 l_-2 ) + 0.00388 l_-2 l_-2 l_-2 + + & & . - 0.00022 l_-3 l_-3 + ) , which is at least roughly compatible with the exact answer . e. witten , `` noncommutative geometry and string field theory , '' nucl . * b268 * , 253 ( 1986 ) . a. sen and b. zwiebach , `` tachyon condensation in string field theory , '' jhep * 0003 * ( 2000 ) 002 [ arxiv : hep - th/9912249 ] . n. moeller and w. taylor , `` level truncation and the tachyon in open bosonic string field theory , '' nucl . b * 583 * ( 2000 ) 105 [ arxiv : hep - th/0002237 ] . a. sen , `` universality of the tachyon potential , '' jhep * 9912 * ( 1999 ) 027 [ arxiv : hep - th/9911116 ] . l. rastelli , a. sen and b. zwiebach , `` string field theory around the tachyon vacuum , '' arxiv : hep - th/0012251 . l. rastelli , a. sen and b. zwiebach , `` vacuum string field theory , '' hep - th/0106010 . d. gaiotto , l. rastelli , a. sen and b. zwiebach , `` ghost structure and closed strings in vacuum string field theory , '' arxiv : hep - th/0111129 . h. hata and t. kawano , `` open string states around a classical solution in vacuum string field theory , '' jhep * 0111 * ( 2001 ) 038 [ arxiv : hep - th/0108150 ] . k. okuyama , `` ghost kinetic operator of vacuum string field theory , '' jhep * 0201 * ( 2002 ) 027 [ arxiv : hep - th/0201015 ] . l. rastelli , a. sen and b. zwiebach , `` boundary cft construction of d - branes in vacuum string field theory , '' jhep * 0111 * ( 2001 ) 045 [ arxiv : hep - th/0105168 ] . k. okuyama , `` ratio of tensions from vacuum string field theory , '' arxiv : hep - th/0201136 . a. sen , `` ghosts and vacuum string field theory , '' talk at jhs/60 , caltech 2001 , http://quark.theory.caltech.edu/jhs60/program.html m. schnabl , `` constraints on the tachyon condensate from anomalous symmetries , '' phys . b * 504 * ( 2001 ) 61 [ arxiv : hep - th/0011238 ] . m. schnabl , `` wedge states in string field theory , '' arxiv : hep - th/0201095 . d. j. gross and a. jevicki , `` operator formulation of interacting string field theory i , ii , '' nucl . * b283 * , 1 ( 1987);nucl . phys . * b287 * , 225 ( 1987 ) . s. samuel , `` the physical and ghost vertices in witten s string field theory , '' phys . * b181 * , 255 ( 1986 ) . l. j. romans , `` operator approach to purely cubic string field theory , '' nucl . b * 298 * ( 1988 ) 369 . l. rastelli and b. zwiebach , `` tachyon potentials , star products and universality , '' jhep * 0109 * ( 2001 ) 038 [ arxiv : hep - th/0006240 ] . b. zwiebach , `` trimming the tachyon string field with su(1,1 ) , '' arxiv : hep - th/0010190 . l. rastelli , a. sen and b. zwiebach , _ private communication_. l. rastelli , a. sen and b. zwiebach , `` star algebra spectroscopy , '' arxiv : hep - th/0111281 . a. leclair , m. e. peskin and c. r. preitschopf , `` string field theory on the conformal plane . 1 . kinematical principles , '' nucl . phys . b * 317 * ( 1989 ) 411 . a. leclair , m. e. peskin and c. r. preitschopf , `` string field theory on the conformal plane . 2 . generalized gluing , '' nucl . b * 317 * ( 1989 ) 464 . i. ellwood , b. feng , y. he and n. moeller , `` the identity string field and the tachyon vacuum , '' jhep * 0107 * ( 2001 ) 016 [ hep - th/0105024 ] . a. s. schwarz and a. sen , `` gluing theorem , star product and integration in open string field theory in arbitrary background fields , '' int . j. mod . phys . a * 6 * ( 1991 ) 5387 . g. t. horowitz and s. p. martin , `` conformal field theory and the symmetries of string field theory , '' nucl . b * 296 * ( 1988 ) 220 . l. bonora , d. mamone and m. salizzoni , `` b field and squeezed states in vacuum string field theory , '' arxiv : hep - th/0201060 . d. gaiotto , l. rastelli , a. sen and b. zwiebach , `` patterns in open string field theory solutions , '' arxiv : hep - th/0201159 . h. hata , s. moriyama and s. teraguchi , `` exact results on twist anomaly , '' arxiv : hep - th/0201177 . i. y. arefeva , d. m. belov and a. a. giryavets , `` construction of the vacuum string field theory on a non - bps brane , '' arxiv : hep - th/0201197 . j. kluso , `` exact solutions of open bosonic string field theory , '' arxiv : hep - th/0202045 . m. r. douglas , h. liu , g. moore and b. zwiebach , `` open string star as a continuous moyal product , '' arxiv : hep - th/0202087 . d. gaiotto , l. rastelli , a. sen and b. zwiebach , _ to appear_.

we investigate the effects caused by noncommutativity of the phase space generated by two scalar fields that have non minimal conformal couplings to the background curvature in the frw cosmology . we restrict deformation of the minisuperspace to noncommutativity between the scalar fields and between their canonical conjugate momenta . then , the investigation is carried out by means of a comparative analysis of the mathematical properties ( supplemented with some diagrams ) of the time evolution of variables in a classical model and the wave function of the universe in a quantum perspective , both in the commutative and noncommutative frames . we find that the impose of noncommutativity causes more ability in tuning time solutions of the scalar fields and hence , has important implications in the evolution of the universe . we get that the noncommutative parameter in the momenta sector is the only responsible parameter for the noncommutative effects in the spatially flat universes . a distinguishing feature of the noncommutative solutions of the scalar fields is that they can be simulated with the well known three harmonic oscillators depending on three values of the spatial curvature . namely the free , forced and damped harmonic oscillators corresponding to the flat , closed and open universes , respectively . in this respect , we call them _ cosmical oscillators_. in particular , in closed universes , when the noncommutative parameters are small , the cosmical oscillators have analogous effect with the familiar beating effect in the sound phenomena . some of the special solutions in the classical model and the allowed wave functions in the quantum model make bounds on the values of the noncommutative parameters . the existence of a non zero constant potential ( i.e. a cosmological constant ) does not change time evolutions of the scalar fields , but modifies the scale factor . an interesting feature of well behaved solutions of the wave functions is that the functional form of the radial part is the same as the commutative ones within a given replacement of constants caused by the noncommutative parameters . furthermore , the noether theorem has been employed to explore the effects of noncommutativity on the underlying symmetries in the commutative frame . two of the six noether symmetries of spatially flat universes , in general , are retained in the noncommutative case , and one out of the three ones in non flat universes . -2.4 cm

1007.2499

scalar field theories have become generic playgrounds for building cosmological models related to particle physics @xcite , and more recently , have played very important contributions in the other branches of cosmology , e.g. , as a powerful tool for providing expanding accelerated universes . also , they have key roles in some models of the early cosmological inflation @xcite , and are viable favorite candidates for dark matter @xcite . indeed , scalar field cosmological models have extensively been studied in the literature , see , e.g. , refs . @xcite and references therein . unifying theories of interactions , such as the kaluza klein , supergravity and superstring theories , usually predicts non minimal couplings between geometry of space time and scalar fields . these theories mostly are derived from effective actions of string theories @xcite or from compactified kaluza klein theories in four dimensions @xcite . in both cases , resulting models have non minimal couplings between gravity and one ( or even more ) scalar field . actually , considering more than one scalar field has been viewed as an easier approach to study accelerating models . these cosmological ideas have also been employed in models of inflation to describe the early universe @xcite or an evolving scalar field known as quintessence @xcite . recently , double scalar tensor theories have been applied in extended gravitational theories , and have given successful descriptions of inflationary scenarios @xcite . besides , it has been shown @xcite that the reduction of a five dimensional brans dicke gravity into four dimensions is equivalent to a 4metric coupled to two scalar fields , which may account for the present accelerated expansion of the universe . also , effects of additional scalar fields plus independent exponential potentials have been considered in the literature @xcite . in cosmological models , scalar fields usually present degrees of freedom and appear as dynamical variables of corresponding minisuperspaces . this point can be viewed as relevance of noncommutativity in these models . the proposal of noncommutativity concept between space time coordinates , in the first time , was introduced in 1947 @xcite . thereafter , a mathematical theory , nowadays known as the noncommutative geometry ( * ncg * ) , has begun to take shape based on this concept since 1980 @xcite . noncommutativity among space time coordinates implies noncommutativity among fields as minisuperspace coordinates . such a kind of noncommutativity has gotten more attention in the literature . in the last decade , study and investigation of physical theories in the noncommutative ( * nc * ) frames , such as the string and m theory @xcite , have caused a renewed interest on noncommutativity in the classical and quantum perspectives . in particular , a novel interest has been developed in studying the nc classical and quantum cosmologies , e.g. refs . @xcite and references therein . also , deformation of the phase space structure has been viewed as an alternative path , in the context of cosmology , toward understanding quantum gravity @xcite . the influence of noncommutativity has been examined by formulation of a version of the nc cosmology in which a deformation of the minisuperspace coordinates @xcite@xcite or the minisuperspace momenta @xcite is required instead of the space time coordinates . from a qualitative point of view , noncommutativity in the minisuperspace coordinates ( space sector ) leads to general effects , however , a non trivial noncommutativity in the momentum sector introduces distinct and instructive effects in the behavior of dynamical variables . in this work , we consider homogeneous and isotropic cosmological models based on a particular ( multi)scalar tensor gravity theory with two scalar fields that have non minimal conformal couplings to the space time . the effects of noncommutativity on the phase space , generated by these fields plus the scale factor , are investigated . our approach is to build a nc scenario via a deformation conveyed by the moyal product @xcite in a classical and a quantum perspective , where their cosmological implications are more appreciated in the classical one . the procedure of quantization is proceeded by the wheeler dewitt ( * wd * ) equation for a wave function of the universe by the hamiltonian operator of the model via the generalized dirac quantization . we introduce noncommutativity between the scalar fields and between their canonical conjugate momenta , and will pay more attention to the outcome of results . here , we neglect noncommutativity between scalar fields with the scale factor , such effects can be found in , e.g. , refs . actually we treat the effects of noncommutativity by two parameters which are the nc parameters corresponding to the space and momentum sectors . then , we analyze the mathematical properties of solutions and look for relations , including the ranges and values , among the nc parameters for which particular or allowed solutions exist . the manuscript is organized as follows . in section @xmath0 , we specify our toy model and investigate the classical version within the commutative and nc frames . the quantum version of this model , by probing the universe wave functions , is considered in section @xmath1 , where the general properties of solutions in the commutative and nc frames are compared . in section @xmath2 , we employ the noether theorem and explore the effects of noncommutativity on the underlying symmetries in the commutative frame . conclusions are presented in the last section , and some necessary formulations have been furnished in the appendix . we consider a classical model consisting of a cosmological system presented by a four dimensional action with two non minimally scalar fields coupled to gravity in a friedman robertson walker ( * frw * ) universe . to specify the nc effects of the model , we first treat the commutative version and then consider the nc one in the following subsections . a general action for non minimally coupled double scalar fields can be written as @xcite @xmath3-v(\phi,\psi)\right\}d^4x\hspace{1mm},\ ] ] where @xmath4 is the determinant of the metric @xmath5 , @xmath6 is the ricci scalar , @xmath7 and @xmath8 are coupling and ( total ) potential functions of scalar fields , respectively . also , @xmath9 and @xmath10 are two typical functions coupled to the kinetic terms . we assume homogeneous scalar fields , that is @xmath11 and @xmath12 , in the following frw metric @xmath13 where @xmath14 is a lapse function , @xmath15 is a scale factor and @xmath16 specifies spatial geometry of the universe . we restrict our considerations to a non interacting conformally scalar field model . ansatz _ is obviously restrictive . however , a general reason for selecting such scalar fields is for simplicity . indeed , it allows exact solutions in simple cases , as those will be discussed in this work and are rich enough to be useful as a probe for significant modifications that ncg introduces in the classical and quantum cosmologies . thus in this case , one can set @xmath17 where @xmath18 is a function of its argument as @xmath19 . also , @xmath20 and @xmath21 is a non minimal coupling parameter with an arbitrary value that represents a direct coupling between the curvature and the scalar fields . the case @xmath22 obviously is the minimally coupling situation ; however in this work , we consider the conformal coupling case , i.e. @xmath23 , and employ the units @xmath24 and @xmath25 ( i.e. @xmath26 ) . we also , in general , consider the scalar fields that are not subject to any potential , , can also be achieved by non interacting ( inner ) potentials , e.g. @xmath27 , where @xmath28 is the cosmological constant . such a case is an important one in models with double scalar fields @xcite . ] and assume more simple cases with @xmath29 . however , at the end of this section , we will investigate the particular case of non zero constant values of the potential function ( i.e. the cosmological constant ) . now , based on these assumptions , substituting metric ( [ b2 ] ) into action ( [ b1 ] ) yields the lagrangian density @xmath30 where a total time derivative term has been neglected . by rescaling the scalar fields as @xmath31 lagrangian ( [ b3 ] ) reads @xmath32 then , the corresponding hamiltonian is @xmath33,\ ] ] where @xmath34 , @xmath35 and @xmath36 are the corresponding canonical conjugate momenta . for the conformal time gauge selection , namely @xmath37 , one gets @xmath38 therefore , the hamilton equations are @xmath39 solutions of equations ( [ b6 ] ) , with the hamiltonian constraint @xmath40 , depend on the curvature index ; actually their solutions are @xmath41 @xmath42 @xmath43 where @xmath44 s , @xmath45 s and @xmath46 s are constants of integration . as it is obvious , if one of the scalar fields vanishes and/or be equal to each other , i.e. @xmath47 , all equations will lead to those derived in the case of one scalar field in our previous work @xcite . we will compare these solutions with their nc analogs in the next section . in the classical physics , noncommutativity is achieved by replacing ordinary product with the so called moyal product , shown by the @xmath48 notation . this associative product on @xmath49dimensional phase space is defined between two arbitrary functions of general phase space variables , namely @xmath50 for @xmath51 , as @xcite @xmath52f(\zeta_1)g(\zeta_2 ) { \biggr|}_{\zeta_1=\zeta_2=\zeta}\ , , \ ] ] such that the symplectic structure is @xmath53 where @xmath54 . the matrix elements @xmath55 are assumed to be real , @xmath56 and @xmath57 are antisymmetric , and @xmath58 ( which can be written as a combination of @xmath56 and @xmath57 ) is symmetric . the modified ( or the @xmath59star deformed ) poisson brackets are defined , in terms of the moyal product , as @xmath60 hence , the modified poisson brackets of the phase space variables are @xmath61 for classical evolution , one starts by proposing that the nc nature of the minisuperspace variables can be encoded in a deformation of the poisson structure . the main advantage of this approach is that the hamiltonian does not need any modification . hence , for our model , we require that the algebra of the minisuperspace variables obeys relations ( [ a4 ] ) in terms of the moyal product defined in ( [ a1 ] ) . however , to introduce such deformations , it is more convenient to consider a linear and non canonical transformation on the classical phase space as ( see e.g. refs . @xcite ) @xmath62 then , if the variables of classical phase space obey the usual poisson brackets , i.e. @xmath63 and @xmath64 , the primed variables will yield @xmath65 with @xmath66 . consequently , to consider the noncommutativity effects , one can work with the ordinary poisson brackets ( [ a43 ] ) instead of the @xmath59star deformed poisson brackets ( [ a4 ] ) . indeed , transformation ( [ a42 ] ) allows an extension of the commutative classical mechanics to the nc one . in the geometrical language , the usual poisson brackets are mapped onto the modified poisson brackets through transformation ( [ a42 ] ) and vice versa . it is evident that , for a compatible extension , transformation ( [ a42 ] ) must be invertible , and this imposes a condition on the nc parameters which we will specify for our model in below . furthermore , let @xmath67 be the hamiltonian of a system including the commutative variables ; we shift the canonical variables through ( [ a42 ] ) and assume that the functional form of the hamiltonian in the nc case is still the same as the commutative one , i.e. @xmath68 this function is also defined on the commutative space and , obviously , equations of motion for unprimed variables are @xmath69 and @xmath70 . evidently , the effects due to the noncommutativity arise by terms including the parameters @xmath56 and @xmath57 . now , in our model , the following notations are adopted @xmath71 @xmath72 . and , in this work , we assume that the only non zero nc parameters are and @xmath73 while they commute with the scale factor is completely consistent with noncommutativity between the original scalar fields , @xmath74 and @xmath75 , while they also commute with the scale factor . ] @xmath76 hence @xmath77 , where @xmath78 and @xmath79 are constants . actually , in accordance with ( [ a42 ] ) , we consider the following transformation @xmath80 transformation ( [ c3 ] ) dictates that it is inverted when its determinant is not zero , that is when @xmath81 . then , by making transformation ( [ c3 ] ) in hamiltonian ( [ b5,2 ] ) , the nc hamiltonian is @xmath82 and equations of motion are @xmath83 and @xmath84 equations ( [ c9 ] ) show that , in general , the motion equations of the scalar fields are coupled in the nc case , and these equations reduce to the commutative equations ( [ b6 ] ) for @xmath85 , as expected . also , in the purposed model , the noncommutativity does not affect the time dependence of the scale factor and its solution is the same as the commutative case . the motion equations of the scalar fields , after eliminating momenta variables in ( [ c9 ] ) , are @xmath86 with the hamiltonian constraint @xmath87 @xmath88 is possible only when @xmath89 , and hence @xmath90 . ] or @xmath91 @xmath92 . space time geometries with @xmath93 only satisfy the latter constraint for any @xmath78 , while @xmath89 geometry fulfills the former constraint with @xmath90 ( and consequently , because of the inversion condition , @xmath94 ) , and the latter one with @xmath95 . in general , solutions of ( [ c13 ] ) depend on the sign of a quantity defined as @xmath96 . for @xmath97 , the sign of @xmath98 is always positive , and thus , the real solution of equations ( [ c13 ] ) can be written as @xmath99 where @xmath100 and @xmath101 are constants of integration subject to the corresponding hamiltonian constraint , and @xmath102 obviously , the corresponding terms in ( [ c16 ] ) are in @xmath103 phase difference . in the case of @xmath104 , the nc parameter @xmath78 does not actually appear in equations ( [ c13 ] ) , and hence , in a spatially flat frw universe , the scalar fields motion equations are affected only by the nc parameter @xmath79 . besides , one gets @xmath105 and @xmath106 , and thus solutions are similar to the motion of a free harmonic oscillator . however , the time dependence of solutions are modified from linear in the commutative case , to periodic in the nc case with the period of @xmath107 which allows the model to be adjusted more easily with the observational data . for @xmath108 geometry , the commutative solutions ( [ c14 ] ) are simple harmonics with the period of @xmath109 , whereas , the nc solutions ( [ c16 ] ) in general are not , though they still oscillate between two limits . incidentally , solutions ( [ c16 ] ) are symmetric with respect to the nc parameters , and results do not vary by replacement of their roles . besides , as @xmath110 , each solution of ( [ c16 ] ) can be considered as a forced ( driven ) harmonic oscillator , such that if the ratio of @xmath111 is a rational fraction , then solutions will be periodic ( as in the lissajous figures ) with angular frequency given by the greatest common divisor of @xmath112 and @xmath113 . otherwise , solutions are non periodic and never repeat themselves . a general condition that picks periodic solutions has been obtained in the appendix ( supplemented with a diagram ) . meanwhile , in the following special example , we provide a better insight about this situation . without loss of generality , let us consider the special case @xmath114 , with @xmath115 , for which solutions ( [ c16 ] ) read @xmath116 with initial conditions @xmath117 , @xmath118 and @xmath119 . the condition for the periodic solutions is @xmath120 , where @xmath121 and @xmath122 are positive integers and obviously @xmath123 . by solving this equation in terms of @xmath78 , one gets @xmath124 . this relation provides different values of @xmath78 parameter in terms of two integers , for which solutions ( [ c17 ] ) can be periodic . it is constructive to plot @xmath78 in terms of @xmath125 as a continuous quantity , is a rational number greater than unity , indeed , @xmath126 yields @xmath127 which resumes the commutative case . ] and this has been illustrated in fig . @xmath128 for range @xmath129 . as it is obvious , @xmath130 is always positive , thus the function @xmath131 changes monotonically with @xmath132 . also , when @xmath132 has very large values , the rate @xmath130 approximately decreases as @xmath133 . for the smallest allowed values of integers , i.e. @xmath134 and @xmath135 , we get the smallest value of @xmath78 parameter , i.e. @xmath136 for the periodic solutions . in solutions ( [ c17 ] ) , the terms @xmath137 and @xmath138 are envelopes of the corresponding curves . these types of oscillations , due to the periodic variation of the amplitude , are usually called beats . the well known example of such oscillations is the simple harmonic one , where a driven force causes mechanical beats . also , this situation can be simulated in the acoustical systems that produce a sound effect , known as beating @xcite , when @xmath139 is sufficiently small and the terms including @xmath140 in ( [ c17 ] ) are ignored . as a sample illustration of the beating effect for these _ cosmical oscillators _ , we have graphed a plot of @xmath141 , equations ( [ c17 ] ) , with its envelopes @xmath142 , in fig . @xmath0 ( right ) for numerical values @xmath143 and @xmath144 . as the figure illustrates , beats are more obvious when the envelopes are drawn . * fig . 1 * : the nc parameter @xmath78 , in @xmath108 case with @xmath145 , as a function of @xmath146 . the points whose @xmath132 are rational fractions correspond to the periodic scalar fields , and other points correspond to non periodic ones . * 2 * : the nc case for @xmath108 : periodic field ( left ) with @xmath147 , beating effect ( right ) with @xmath143 and @xmath114 including the cosmical oscillator ( solid line ) and envelops ( dashed lines ) . when the ratio of @xmath113 and @xmath112 is not a rational fraction , solutions are non periodic but , as periodic cases , their behaviors still depend on the values of the nc parameters . for example , if one constructs a plot with numerous or a few relative extremum in a given time interval , then iterative drawings will indicate that the separation between the high and low points increases when the nc parameters tend to smaller values . this property is intensified for values less than unity , which is illustrated in fig . @xmath1 for @xmath148 in equations ( [ c16 ] ) with constants @xmath149 and @xmath150 . from this point of view , the nc solutions have particular preference with respect to the corresponding commutative ones . * : non periodic field in the nc case for @xmath108 : with @xmath151 and @xmath152 ( solid line ) , and with @xmath153 and @xmath154 ( dashed line ) . for @xmath89 geometries , one has @xmath155 that can be either positive , or negative or zero depending on different choices of the nc parameters . that is , when ( @xmath156 and @xmath157 ) or ( @xmath158 and @xmath159 ) , @xmath98 is positive and hence one gets solutions of type ( [ c16 ] ) . however , in this case , the nc parameters have upper and lower bounds , and there is more restriction on finding the periodic solutions than in @xmath108 geometry . on the other hand , when @xmath98 is negative and @xmath160 , the real solutions of equations ( [ c13 ] ) are @xmath161 where @xmath162 and @xmath163 are constants of integration subject to the corresponding hamiltonian constraint . solutions ( [ c15 ] ) are constrained to the condition that both nc parameters simultaneously have a lower bound , namely @xmath78 and @xmath164 , or an upper bound , @xmath78 and @xmath165 . also , in the case @xmath108 , oscillations of the scalar fields in ( [ c15 ] ) have the phase difference of @xmath103 . these solutions , in contrast to their commutative analogs , are oscillating with a time hyperbolic amplitude that depends on the nc parameters and gives enough room for better adjustments . for instance , one can change the time interval between two successive zero points of the scalar fields , for the interval is @xmath166 . besides , increasing and decreasing amplitudes in ( [ c15 ] ) with respect to the time depend on initial conditions . as an example , if one chooses the initial condition @xmath167 , then solutions of ( [ c15 ] ) will oscillate with decreasing amplitudes . such solutions are similar to the damped harmonic oscillators with amplitudes proportional to @xmath168 as envelopes . the main characters of such an oscillator , namely the _ time decay _ and the _ natural frequency _ , can be described in terms of the nc parameters as @xmath169 and @xmath170 , respectively . another interesting point is that , if one assumes @xmath78 and @xmath171 then , solutions will damp quickly . inversely , to possess more lately damped oscillations , the best choice of the nc parameters in the range @xmath78 and @xmath165 is when one assumes @xmath172 and @xmath173 , in which each of the scalar fields has a maximum number of oscillations before the complete damping occurs . the diagram of such an oscillation is plotted in fig . @xmath2 for the @xmath141 of solutions ( [ c15 ] ) with @xmath144 for numerical values ( @xmath174 and @xmath175 ) as quick damping , and for ( @xmath176 and @xmath177 ) as late damping . when @xmath178 in @xmath89 geometry , solutions of ( [ c13 ] ) are decoupled hyperbolic time functions with the coefficient @xmath179 in the exponent , which is the only difference between the nc and commutative solutions . the case @xmath180 is also possible when @xmath90 or when @xmath181 , where solutions ( [ c15 ] ) are again periodic with the period of @xmath182 or @xmath183 , respectively . note that , the case @xmath89 with @xmath90 resembles the constraint condition @xmath88 with arbitrary @xmath79 , and however for @xmath181 one gets trivial constant solutions , but this value is not allowed . * fig . 4 * : damping scalar field in the nc case @xmath89 : late damping with @xmath176 and @xmath184 ( solid line ) , and quick damping with @xmath174 and @xmath185 ( dashed line ) . * non zero constant potential * let us obtain the equations of motion for the scale factor when one considers a non zero constant value for the potential function ( i.e. a cosmological constant ) in action ( [ b1 ] ) . thus , suppose the potential function is @xmath186 . in a manner similar to the free potential case , one gets the hamiltonian @xmath187 in the commutative case , and @xmath188 in the nc one . it is obvious that the hamiltonian equations for the scalar fields , derived from ( [ b5,111 ] ) and ( [ c41 ] ) , are the same as those obtained in the corresponding free potential case , namely equations ( [ b6 ] ) and ( [ c9 ] ) , respectively . but , the hamiltonian equations of the scale factor , for both the commutative and nc cases , are modified in the same manner as @xmath189 eliminating the momentum variable gives @xmath190 by integrating , one gets @xmath191 where @xmath192 is an integration constant . solution of equation ( [ cd ] ) can be written in terms of the jacobi elliptic functions , that is @xmath193 where the initial condition @xmath194 , @xmath195^{-1/2}.\ ] ] the jacobi function , on the right hand side ( [ cd1 ] ) , behaves as a sine function when @xmath59 is a real and @xmath196 is either a real or a pure imaginary number . therefore , for negative potentials in an arbitrary geometry , and also for positive potentials in the @xmath108 geometry when @xmath197 , the scale factor behaves periodically as a sinusoidal function . on the other hand , for positive potentials in @xmath198 geometries , and in the @xmath108 geometry when @xmath199 , the jacobi function has a complex value , and hence these situations are not allowed . in the limit @xmath200 for @xmath108 , one gets @xmath201 , @xmath202 and @xmath203 behaves as @xmath204 . hence , we get back to the free potential solution ( [ c14 ] ) , as expected . however , in such a limit , for @xmath205 , as @xmath206 , solution ( [ cd1 ] ) is not valid , and hence , one must consider equation ( [ cd ] ) . now , by regarding the limit @xmath200 in equation ( [ cd ] ) for @xmath205 , and solving the resulted equation , one again gets the free potential solutions ( [ c141 ] ) and ( [ c142 ] ) , respectively . although the effects of nc cosmology are mostly desired and important in classical approaches , but still it is instructive to investigate its quantum counterparts . in particular , if a universe has been commenced by a big bang or from a very very tiny scale , then it would be plausible that quantum behaviors should have significant implications in its evolution and cosmology . thus , in this section we proceed to quantize the cosmological model given by action ( [ b1 ] ) for free potential , such that the canonical quantization of the phase space leads to the wd equation , @xmath207 , where @xmath208 is the hamiltonian operator and @xmath209 is a wave function of the universe @xcite . we employ the usual canonical transition from classical to quantum mechanics via the generalized dirac quantization of the poisson brackets to quantum commutators , i.e. @xmath210 $ ] . then , as the classical approach , we investigate the general properties of the wave function in the commutative and nc frames of the quantum model in the following subsections . it is well known @xcite that , in the nc quantum mechanics , the original phase space and its symplectic structure are modified . that is , for the nc proposal of quantum cosmology , we assume that operators ( variables ) of the frw minisuperspace obey a kind of deformed heisenberg algebra like the ones in the nc quantum mechanics as @xcite @xmath237=i\theta_{ij}\ , , \hspace{.5cm}[\hat{x}_i,\hat{p}_j]=i(\delta_{ij}+\sigma_{ij } ) \hspace{.5cm}{\rm and}\hspace{.5cm}[\hat{p}_i,\hat{p}_j]=i\beta_{ij}\ , .\ ] ] the notations and definitions are the same as in the nc classical model . this kind of extended noncommutativity maintains symmetry between the canonical operators , and yields the usual heisenberg algebra in the limit @xmath56 and @xmath57 @xmath238 . as usual , this deformation can be redefined in terms of noncommutativity of minisuperspace functions with the moyal product defined in ( [ a1 ] ) . thus , the corresponding noncommutative wd equation can be written by replacing the operator product , in equation ( [ d1 ] ) , with the star product , namely @xmath239 . however , it is possible to reformulate equations in terms of the commutative operators with the ordinary product of functions if the new operators are introduced , again as our previous assumption @xmath240 clearly , if unprimed operators obey the usual heisenberg commutators , then the non zero primed operators will obey the deformed heisenberg commutators ( [ a8 ] ) in the form @xmath241=i\theta,\hspace{.5cm}[\hat{x}',\hat{p}'_x]=i(1+\theta\beta)=[\hat{y}',\hat{p}'_y]\quad { \rm and } \quad [ \hat{p}'_x,\hat{p}'_y]=4i\beta .$ ] therefore , operator transformation ( [ a911 ] ) can be regarded as a generalization of the usual quantum mechanics to the nc one . on the other hand , the inverse transformation of ( [ a911 ] ) are @xmath242 where @xmath243 . consequently , one can go from the usual commutators to the deformed ones and vice versa provided that again @xmath244 . as a result , the original equation , employing the new operators , reads @xcite @xmath245 hence , the noncommutative wd equation corresponding to the nc hamiltonian ( [ c4 ] ) is @xmath246\psi_{\rm nc}(a , x , y)=0\,,\ ] ] which in the polar coordinates ( [ d2 ] ) yields @xmath247\psi_{\rm nc}(a,\rho,\varphi)=0.\ ] ] when @xmath248 , by using the _ ansatz _ ( [ d4 ] ) , equation ( [ d10 ] ) is separable to @xmath249 and @xmath250b=0,\ ] ] where @xmath221 is a constant of separation . the scale factor part of the wave function , equation ( [ d11 ] ) , is similar to its commutative analog ( [ d41 ] ) , as expected . the radial part of the wave function , equation ( [ d12 ] ) , first reduces to its commutative analog , equation ( [ d42 ] ) , when @xmath85 , as again expected . secondly , in the case @xmath104 with @xmath114 , equation ( [ d12 ] ) once again reduces to equation ( [ d42 ] ) even if the nc parameter @xmath78 does exist . however , in @xmath104 , solutions of equation ( [ d12 ] ) do not depend on the nc parameter @xmath78 at all . namely in a flat frw universe , the @xmath79 parameter is the only responsible parameter for the nc effects . these properties are common with the classical model . comparing equation ( [ d12 ] ) with the commutative analog ( [ d42 ] ) shows that the functional form of the radial part of the wave function ( and hence , the whole wave function ) is the same as the commutative ones provided that the coefficients @xmath221 and @xmath16 in equation ( [ d42 ] ) are replaced by @xmath251 therefore , for @xmath93 and @xmath89 with @xmath95 , the nc eigenfunctions are in terms of the whittaker functions similar to the commutative solutions ( [ d71 ] ) . however , in the case @xmath89 , there are bounds on the nc parameters . for a better illustration , let us write ( [ d121 ] ) as @xmath252 since this modified curvature index , @xmath253 , can have positive , negative and zero values , the solutions again are similar to the commutative cases ( [ c1411 ] ) and ( [ c14111 ] ) , where @xmath254 also requires @xmath181 . but , for @xmath255 , the nc parameters must satisfy inequalities depending on the sign of @xmath253 . namely if @xmath256 , then there will be ( @xmath78 and @xmath165 ) or ( @xmath78 and @xmath164 ) , and if @xmath257 , then ( @xmath156 and @xmath165 ) or ( @xmath158 and @xmath164 ) . for instance , by choosing inequality ( @xmath78 and @xmath165 ) when @xmath256 , the corresponding wave packet can be written in the form @xmath258 where the lower limits , in contrast to the commutative wave packet ( [ d72 ] ) , are bounded by @xmath259 . in the case @xmath88 , namely when @xmath89 and @xmath90 ( hence @xmath94 ) , with the separation of variables @xmath260 , equation ( [ d10 ] ) reduces to two differential equations . one equation is similar to equation ( [ d11 ] ) , and the other one is @xmath261b(\rho,\varphi)=c\ , , \ ] ] where @xmath162 is a separation constant . as , in general , @xmath162 has a non zero value , equation ( [ d16 ] ) does not have a suitable solution for @xmath181 ( which itself is not allowed ) . for @xmath262 , one easily shows that its solution can be written as @xmath263 , where @xmath264 is an arbitrary function and @xmath265 . as this function is not well behaved when @xmath266 tends to zero , it is not allowed . in this section , we employ the noether theorem and explore the effects of noncommutativity on the underlying symmetries in the commutative frame . for this purpose , following refs . @xcite , one can define the noether symmetry as a vector field , say @xmath267 , on the tangent space of the phase space . in our model , it can be , in general , as @xmath268 such that the lie derivative of the lagrangian with respect to this vector field vanishes , i.e. @xmath269 for simplicity , we assume that unknown functions @xmath100 , @xmath101 and @xmath162 to be linear in terms of the configuration variables @xmath270 , @xmath271 and @xmath73 . the time derivative @xmath272 represents the lie derivative along the dynamical vector fields , which in our model is @xmath273 . now , in order to obtain constants of motion , let us rewrite equation ( [ b ] ) as @xmath274 by employing the lagrange equation , @xmath275@xmath276= @xmath277 , it reads @xmath278 which yields @xmath279 therefore , the constants of motion are @xmath280 for different unknown functions of @xmath100 , @xmath101 and @xmath162 . to obtain these functions , one can employ equation ( [ d ] ) or equation ( [ e ] ) which is more suitable in the hamiltonian formalism . to manage this , one can write equation ( [ e ] ) in terms of the poisson bracket , @xmath281 , as @xmath282p_a\nonumber\\ + \left[\frac{\partial b}{\partial a}\{a,{\cal h}\}+\frac{\partial b}{\partial x}\{x,{\cal h}\}+\frac{\partial b}{\partial y } \{y,{\cal h}\}\right]p_x+ \left[\frac{\partial c}{\partial a}\{a,{\cal h}\}+\frac{\partial c}{\partial x}\{x,{\cal h}\}+\frac{\partial c}{\partial y } \{y,{\cal h}\}\right]p_y=0.\end{aligned}\ ] ] this equation , in general , gives quadratic polynomials in terms of the momenta with coefficients being partial derivatives of @xmath100 , @xmath101 and @xmath162 with respect to the configuration variables . hence , the expression identically is equal to zero if and only if these coefficients vanish , which lead to a system of partial differential equations for @xmath100 , @xmath101 and @xmath162 . in the following subsections , we obtain such symmetries for the model in the commutative and nc cases . in this case the hamiltonian is given by relation ( [ b5,2 ] ) ; hence by substituting the corresponding poisson brackets into equation ( [ h ] ) , one gets @xmath283 let us first treat spatially non flat geometry , @xmath284 , for which equation ( [ i ] ) leads to @xmath285 a general solution of constraints ( [ j ] ) is @xmath286 where @xmath287 s are three constants of integration . for an easier representation of the commutation relations between symmetric vectors , let us change @xmath288 and @xmath289 in solution ( [ l ] ) . thus by ( [ g ] ) , three independent constants of motion are @xmath290 which are the well known angular momenta about the configuration variables . the corresponding symmetric vectors are @xmath291 which satisfy the lie algebra @xmath292=\varepsilon_{ij}{}^{k}x_k$ ] , where @xmath293 is the levi civita tensor . for @xmath104 , it is clear from equation ( [ i ] ) that the symmetries can be obtained from the last two constraints of ( [ j ] ) . thus , the corresponding solution is @xmath294 where @xmath295 s are six constants of integration . therefore , the six independent constants of motion are @xmath296 incidentally , in this case , a glance at hamiltonian ( [ b5,2 ] ) shows that all configuration variables are cyclic and consequently their corresponding momenta are constants of motion , i.e. @xmath297 to @xmath298 . in addition , hamiltonian ( [ b5,2 ] ) for @xmath104 ( actually the lagrangian @xmath299 ) is invariant under rotation about the @xmath270axis . thus , the angular momentum about this axis is conserved , i.e. @xmath300 . the corresponding symmetric vectors , in the flat geometry , are @xmath301 which satisfy @xmath302&=&0,\hspace{1.5cm}[x_1,x_3]=0,\hspace{1.3cm}[x_1,x_4]=0,\nonumber\\ \left[x_1,x_5\right]&=&x_3,\hspace{1.2cm}[x_1,x_6]=x_2,\hspace{1cm}[x_2,x_3]=0,\nonumber \\ \left[x_2,x_4\right]&=&-x_3,\hspace{.9cm}[x_2,x_5]=0,\hspace{1.3cm}[x_2,x_6]=x_1,\\ \left[x_3,x_4\right]&=&x_2,\hspace{1.2cm}[x_3,x_5]=x_1,\hspace{1cm}[x_3,x_6]=0,\nonumber \\ \left[x_4,x_5\right]&=&-x_6,\hspace{.9cm}[x_4,x_6]=x_5 , \hspace{1cm}[x_5,x_6]=x_4.\nonumber\end{aligned}\ ] ] now , let us find out which of the above symmetries survive in the nc case . here , the hamiltonian is given by relation ( [ c4 ] ) , and the required poisson brackets are given by equations ( [ c5 ] ) and ( [ c9 ] ) . substituting the corresponding poisson brackets into equation ( [ h ] ) gives @xmath303\nonumber \\ + \frac{1}{2}\left[(1+k\theta^2)\frac{\partial a}{\partial x}-\frac{\partial b}{\partial a}\right]p_ap_x+\frac{1}{2}\left[(1+k\theta^2)\frac{\partial a}{\partial y}-\frac{\partial c}{\partial a}\right]p_ap_y+\frac{1}{2}(1+k\theta^2)\left(\frac{\partial b}{\partial y}+ \frac{\partial c}{\partial x}\right)p_x p_y\nonumber\\ + ( k\theta+\beta)\left[\left(y\frac{\partial a}{\partial x}-x\frac{\partial a}{\partial y}\right)p_a+\left(y\frac{\partial b}{\partial x}-x\frac{\partial b}{\partial y}-c\right)p_x+\left(y\frac{\partial c}{\partial x}-x\frac{\partial c}{\partial y}+b\right)p_y\right]=0.\end{aligned}\ ] ] obviously , equation ( [ i1 ] ) yields extra restrictions on @xmath100 , @xmath101 and @xmath162 with respect to the commutative case ( [ i ] ) . when @xmath304 , is possible when ( @xmath89 and @xmath178 ) and or ( @xmath305 and any value of @xmath78 ) . ] by putting each coefficient in equation ( [ i1 ] ) equal to zero , one gets @xmath306 hence , one obtains the solution @xmath307 and @xmath308 where @xmath309 and @xmath310 are integration constants . thus , when @xmath311 , the only constant of motion is @xmath312 with the symmetric vector @xmath313 which is an especial case of commutative solution ( [ l ] ) with initial conditions @xmath314 . for @xmath104 ( with @xmath315 ) , in addition to @xmath316 , we have an another constant of motion , namely @xmath317 , which corresponds to the symmetric vector @xmath318 with the trivial lie algebra @xmath319=0 $ ] . this case also restricts commutative solution ( [ ac ] ) with initial conditions @xmath320 . therefore , in the @xmath304 case , the noncommutativity reduces the number of symmetries to one for @xmath284 and two for @xmath104 . when @xmath321 , the last row of constraints in ( [ p1 ] ) is omitted . in this case , one choice is @xmath89 and @xmath178 , where a non trivial ( i.e. @xmath322 ) linear solution exists if and only if @xmath323 . hence , by the sign change of functions , i.e. @xmath324 and @xmath325 ( or equivalently @xmath326 ) , the solution is the same as the commutative solution ( [ l ] ) , which again yields the constants of motion and symmetric vectors ( [ m ] ) and ( [ ae ] ) . however for @xmath89 and a special value of @xmath327 , it gives one solution ( i.e. ( [ r ] ) ) , but this case is not allowed due to the inversion condition @xmath81 . for another choice @xmath305 with any value of @xmath78 , it gives the commutative solution ( [ ac ] ) . that is , even though the noncommutativity is still present , the number of symmetries and constants of motion do not change with respect to the corresponding commutative case . note that , in all cases , irrespective of whether the noncommutativity exists or not , and for any value of the curvature index , the angular momentum about the @xmath270axis is conserved , as expected . besides , in the above considerations , we have not , in general , specified numerical values of the nc parameters . we have carried out an investigation for the role of ncg in cosmological scenarios , based on a four dimensional free potential ( multi)scalar tensor action of gravity , by introducing a nc deformation in the minisuperspace variables . the phase space is generated by two non interacting conformal scalar fields plus the scale factor with their canonical conjugate momenta . the scalar fields are non minimally coupled to geometry whose background is the frw metric , where the conformal time gauge evolutions have been studied . the noncommutativity has been introduced only between the scalar fields and between their canonical conjugate momenta via two nc parameters @xmath78 and @xmath79 , respectively . the investigation has been carried out for this toy model by means of a comparative mathematical analysis of the time evolution of the dynamical variables in the classical level and of the wave function of the universe in the quantum perspective , both in the commutative and nc frames . we have paid more attention to the outcome of results and have looked for the relations , including ranges and values , among the nc parameters for which particular or allowed solutions exist . in general , we have shown by this toy model that the purposed noncommutativity has important implications in the evolution of the universe , however , does not affect the time dependence of the scale factor , i.e. its solution is the same as the commutative case , as expected . also , we have found that one of the particularity of the nc parameter in the momenta sector , i.e. @xmath79 , is in the spatially flat frw universe , where it is the only responsible parameter for the nc effects in the classical and quantum frames . in the classical model , exact solutions have been obtained . one of the important aspects of the nc solutions is that they can be regulated with both nc parameters . for example , these parameters can be employed to adjust the time dependence of solutions with the experimental or observational data . a distinguished feature of the noncommutativity effects , which we call a cosmical oscillator , is that the scalar fields behave similar to ( or can be simulated with ) the three most important harmonic oscillators depending on three values of the spatial curvature . these are the free , forced and damped harmonic oscillators corresponding to the flat , closed and open universes , respectively . in the flat universes , the time dependence of solutions are modified from linear in the commutative case to periodic in the nc frame . in the closed universes , if the ration of frequencies of the scalar fields is a rational fraction , then solutions will be periodic . this condition restricts the nc parameters . when this ratio is not a rational fraction , the solutions are non periodic but their behaviors still depend on the values of the nc parameters . a plot with numerous ( or a few ) relative extremum in a given time interval indicates that the separation between the high and low points increases when the nc parameters tend to smaller values , and this property is intensified for values less than unity . from this point of view , the nc solutions have particular preference with respect to the corresponding commutative ones . furthermore , in the @xmath108 case , the solutions are symmetric with respect to the nc parameters , and the results do not vary by replacement of their roles . also , when the nc parameters are small , the cosmical oscillators have analogous effects with the familiar beating effects in the sound phenomena . for a better view on this situation , an example has been illustrated in the text . in the open universes , there are upper and lower bounds on the nc parameters . the quick and late damping of this case can be adjusted by the nc parameters . for example , for @xmath328 and @xmath329 , the scalar fields have maximum number of oscillations before the complete damping occurs . we have also shown that the existence of a non zero constant value of the potential function ( i.e. the cosmological constant ) does not change the time evolutions of the scalar fields , but it modifies the time dependence of the scale factor in a same manner for both the commutative and nc frames . indeed , we have obtained that in all allowed conditions the scale factor behaves as a sinusoidal function . in the quantum model , the exact solutions for the wave functions of the universe have also been obtained . the scale factor part of the wave function is similar to its commutative analog , as expected . one still expects that when the noncommutativity effects are turned on in the quantum scenario , they should introduce significant modifications in the scalar fields . however , an interesting feature of the well behaved solutions is that the functional form of the radial part of the nc wave function is the same as the commutative ones within the given replacements of constants caused by the nc parameters , although these replacements in turn may cause drastic effects . for example , the curvature index is modified , and in open universes , the allowed nc wave functions impose bounds on the nc parameters . finally , we have employed the noether theorem and have explored the effects of noncommutativity on the underlying symmetries in the commutative frame . we have shown that for spatially flat universes , there are six noether symmetries , and , in general , only two of them are retained in the nc case . in the special case @xmath305 , all symmetries survive regardless of the existence of @xmath78 parameter . for non flat universes , there are three noether symmetries in the commutative case , one of which is retained in the nc frame . however , in open universes , when the nc parameters have the values @xmath323 , the number and general form of symmetries do not change with respect to the commutative frame . the only difference is related to the sign of symmetries . conservation of the angular momentum about the scale factor axis is a common face between the commutative and nc cases , as expected in the purposed noncommutativity . we should emphasis that the scalar fields solutions are given after rescalings of the original fields , relations ( [ rescalingfields ] ) , and in the conformal time . consequently , the functional forms of the solutions are not as simple when translated to the original fields . however , any detection should be performed in a reference frame , and there are debates on the physical frame in the cosmological contexts . besides , the model is to be taken as a toy model only that still provides a valuable contribution to the assessment of the implications of the nc geometrical deformations of the phase space upon the dynamics of the cosmological model envisaged . however , a more realistic nc cosmological model may be achieved when one also involves the noncommutativity of the scalar fields with the scale factor , where the value of its time derivative in the form of the hubble parameter can be determined from observations . we first obtain a general relation among the nc parameters when the solutions ( [ c16 ] ) are periodic . then , we treat a specific periodic solution as an example . as mentioned in the text when @xmath108 , solutions ( [ c16 ] ) play as the forced oscillators , and the existence of periodic solutions is provided when the ratio @xmath111 is a rational fraction . in this case , the @xmath98 parameter , by its definition , is @xmath330 ; hence by the transformation reversibility constraint @xmath81 , one always has @xmath331 . therefore , by definitions ( [ c161 ] ) , we can take @xmath332 and @xmath333 , where @xmath121 and @xmath122 are two positive integers with @xmath123 , and @xmath234 is an arbitrary non zero positive constant . hence , again , definitions ( [ c161 ] ) give @xmath334 these relations impose a firm restriction on the values of the nc parameters ( or equivalently on allowed positive integers ) . indeed , the condition for periodic scalar fields is that , the values of the nc parameters must be such that there exist two positive integers satisfying relations ( [ ca ] ) . as @xmath78 and @xmath79 are real roots of the quadratic equation @xmath335 , relations ( [ ca ] ) lead to @xmath336 the above result implies that if one finds a pair of integers satisfying inequality ( [ caa ] ) , then the values of the nc parameters will be calculated from equations ( [ ca ] ) . let us treat a specific example for the above considerations . consider the equality in relation ( [ caa ] ) , which yields @xmath178 , which also by the constraint @xmath81 gives @xmath95 ( note that we have assumed @xmath337 ) . also , by definitions ( [ c161 ] ) , we have @xmath338 and @xmath339 . thus , by imposing the periodic condition as @xmath340 it leads to @xmath341 where @xmath196 is a real number greater than one . gives the commutative case . ] in the limit @xmath342 , one solution of ( [ caaaa ] ) yields @xmath343 , as expected ; however , the other one gives infinity that is not physically accepted . also , in the limit @xmath344 , we have @xmath345 , which , in general , may be viewed as a transition from the forced harmonic oscillator in the @xmath108 geometry to the simple one in the @xmath104 case . in another words , when the value of @xmath196 slowly gets bigger and bigger , then the driven force is gradually removed away from the cosmical oscillators . the time evolution of scalar field @xmath141 , equation ( [ c16 ] ) , is depicted in fig . @xmath0 ( left ) for the numerical value @xmath346 corresponding to @xmath151 , with superposition constants @xmath149 and @xmath150 . the period of solutions is the least common multiple of distinct periods , namely @xmath347 and @xmath348 . we thank the research office of the shahid beheshti university for financial support . this work is a part of b.m.s ph.d . thesis under supervision of m.f .. 9 d. boyanovsky , h.j . de vega , r. holman , d.s . lee and a. singh , _ phys . d _ * 51 * , 4419 ( 1995 ) , [ hep - ph/9408214 ] ; + k.g . zloshchastiev , _ phys . b _ * 450 * , 397 ( 1999 ) , [ hep - th/9911012 ] . a.h . guth _ * 23 * , 347 ( 1981 ) ; + a.d . linde _ phys . _ * 108b * , 389 ( 1982 ) . v. sahni and l. wang , _ phys . d _ * 62 * , 103517 ( 2000 ) , [ astro - ph/9910097 ] ; + d. bertacca , s. matarrese and m. pietroni , _ mod . phys . lett . a _ * 22 * , 2893 ( 2007 ) , [ astro - ph/0703259 ] . abreu , p. crawford and j.p . mimoso , _ class . grav . _ * 11 * , 1919 ( 1994 ) , [ gr - qc/9401024 ] ; + n. cruz and c. martine , _ class . * 17 * , 2867 ( 2000 ) , [ gr - qc/0007038 ] ; + s. fay , _ class . * 21 * , 1609 ( 2004 ) , [ gr - qc/0402104 ] ; + s. koh , _ j. korean phys . soc . _ * 49 * , s787 ( 2006 ) , [ astro - ph/0510030 ] ; + m. demianski , e. piedipalumbo , c. rubano and c. tortora , _ astron . astrophys . _ * 454 * , 55 ( 2006 ) , [ astro - ph/0604026 ] . barrow and k.e . kunze , _ chaos solitons fractals _ * 10 * , 257 ( 1999 ) , [ gr - qc/9807040 ] . t. appelquist , a. chodos and p.g.o . freund , _ modern kaluza klein theories _ , ( addison wesley , new york , 1987 ) ; + f.g . alvarenga and j.c . fabris , _ class . * 12 * , l69 ( 1995 ) . linde , _ phys . b. _ * 108 * , 389 ( 1982 ) . caldwell , r. dave and p.j . steinhardt , _ astrophys . space sci . _ * 261 * , 303 ( 1998 ) . s. capozziello and g. lambiase , _ grav . * 6 * , 164 ( 2000 ) , [ gr - qc/0003048 ] . s. nojiri and s.d . odintsov , _ phys . b _ * 659 * , 821 ( 2008 ) , [ hep - th/0708.0924 ] . l. qiang , y. ma , m. han and d. yu , _ phys d _ * 71 * , 061501 ( 2005 ) , [ gr - qc/0411066 ] . liddle , a. mazumdar and f.e . schunck , _ phys . d _ * 58 * , 061301 ( 1998 ) , [ astro - ph/9804177 ] . h. snyder , _ phys . rev . _ * 71 * , 38 ( 1947 ) . a. connes , _ c. r. acad . paris , ser . * 290 * ( 1980 ) ; + a. connes , _ noncommutative geometry _ , ( academic press , new york , 1994 ) ; + a. connes , _ j. math _ * 41 * , 3832 ( 2000 ) , [ hep - th/0003006 ] ; + a. connes , `` noncommutative geometry '' , _ math/0011193_. n. seiberg and e. witten , _ jhep _ * 9909 * , 032 ( 1999 ) , [ hep - th/9908142 ] . t. banks , w. fischler , s.h . shenker and l. susskind , _ phys . d _ * 55 * , 5112 ( 1997 ) , [ hep - th/9610043 ] . f. lizzi , g. mangano and g. miele , _ jhep _ * 0206 * , 049 ( 2002 ) , [ hep - th/0203099 ] ; + n. khosravi , s. jalalzadeh and h.r . sepangi , _ jhep _ * 0601 * , 134 ( 2006 ) , [ hep - th/0601116 ] ; + e. akofor , a.p . balachandran , s.g . jo , a. joseph and b.a . qureshi , _ jhep _ * 0805 * , 092 ( 2008 ) , [ astro - ph/0710.5897 ] ; + e. akofor , a.p . balachandran , a. joseph , l. pekowsky and b.a . qureshi , _ phys . d _ * 79 * , 063004 ( 2009 ) , [ astro - ph/0806.2458 ] . szabo , _ phys . _ * 378 * , 207 ( 2003 ) , [ hep - th/0109162 ] ; + c.k . zachos , d.b . fairlie and t.l . curtright ( editors ) , _ quantum mechanics in phase space _ , ( world scientific , singapore , 2005 ) ; + n. khosravi , h.r . sepangi and b. vakili , _ gen . rel . gravit . _ * 42 * , 1081 ( 2010 ) , [ gr - qc/0909.2487 ] . l.o . pimentel and c. mora , _ gen . grav . _ * 37 * , 817 ( 2005 ) , [ gr - qc/0408100 ] ; + l.o . pimentel and o. obregn , grav . _ * 38 * , 553 ( 2006 ) . h. garca compen , o. obregn and c. ramrez , _ phys . lett . _ * 88 * , 161301 ( 2002 ) , [ hep - th/0107250 ] . barbosa and n. pinto neto , _ phys . d _ * 70 * , 103512 ( 2004 ) , [ hep - th/0407111 ] . barbosa , _ phys . rev . d _ * 71 * , 063511 ( 2005 ) , [ hep - th/0408071 ] . n. khosravi , h.r . sepangi and m.m . jabbari , _ phys . b _ * 647 * , 219 ( 2007 ) , [ hep - th/0611236 ] ; + b. vakili , n. khosravi and h.r . sepangi , _ class . * 24 * , 931 ( 2007 ) , [ gr - qc/0701075 ] ; + n. khosravi , s. jalalzadeh , h.r . sepangi , _ int . j. mod * 16 * , 1187 ( 2007 ) , [ gr - qc/0702108 ] ; + a. bina , k. atazadeh , s. jalalzadeh , _ int . j. theor . phys . _ * 47 * , 1354 ( 2008 ) , [ gr - qc/0709.3623 ] ; + c. bastos , o. bertolami , n.c . dias and j.n . prata , _ phys . rev . d _ * 78 * , 023516 ( 2008 ) , [ gr - qc/0712.4122 ] ; _ j. phys _ * 174 * , 012053 ( 2009 ) [ hep - th/0812.3488 ] ; + w. guzmn , m. sabido and j. socorro , _ phys . d _ * 76 * , 087302 ( 2007 ) , [ gr - qc/0712.1520 ] ; `` on noncommutative minisuperspace and the friedmann equations '' , _ gr - qc/0812.4251_. b. malekolkalami and m. farhoudi , _ phys . b _ * 678 * , 174 ( 2009 ) , [ gr - qc/0911.2548 ] . m. chaichian , m.m . jabbari and a. tureanu , _ phys . lett . _ * 86 * , 2716 ( 2001 ) , [ hep - th/0010175 ] . a.e.f . djema and h. smail , _ commun . phys . _ * 41 * , 837 ( 2004 ) , [ hep - th/0309006 ] . kinsler , a.r . fery , a.b . coppens and j.v . sanders , _ fundamentals of acoustics _ , ( john wiley & sons , new york , 4th ed . m. abramowitz and i.a . stegun , _ handbook of mathematical functions with formulas , graphs , and mathematical tables _ , ( dover publications , new york , 9th printing 1972 ) ; + l.j . slater , _ confluent hypergeometric functions _ , ( cambridge university press , london , 1960 ) ; + h. buchholz , _ the confluent hypergeometric function _ , ( springer , berlin , 1969 ) . m. cavagli , v. de alfaro and t. filippov , _ int . a _ * 10 * , 611 ( 1995 ) , [ gr - qc/9402031 ] ; + j. feinberg and y. peleg , _ phys . d _ * 52 * , 1988 ( 1995 ) , [ hep - th/9503073 ] . t. curtright , d. fairlie and c. zachos , _ phys . d _ * 58 * , 025002 ( 1998 ) , [ hep - th/9711183 ] . s. fay , _ class . grav . _ * 18 * , 4863 ( 2001 ) , [ gr - qc/0309088 ] .

this note presents a discussion of the algebraic and combinatorial aspects of the theory of pure-sequences . various instances where pure-sequences appear are described . several open problems that deserve further investigation are also presented . the work for this paper was done while the first author was sponsored by the national security agency under grant number h98230 - 12 - 1 - 0204 , and by the simons foundation under grant # 208579 . + the work for this paper was done while the second author was sponsored by the national security agency under grant number h98230 - 12 - 1 - 0247 , and by the simons foundation under grant # 208869 . ] _ dedicated to david eisenbud on the occasion of his 65th birthday _

1204.5247

pure @xmath0-sequences are fascinating objects that arise in several mathematical areas . they have been the subject of extensive research , yet our knowledge about pure @xmath0-sequences is limited . the goal of this note is to survey some of the known results and to motivate further investigations by pointing out connections to various interesting problems . much of the material for this paper has been drawn from the recent monograph @xcite by the authors with mats boij and rosa mir - roig , cited in this article as bmmnz . we often quote results giving the name of the author(s ) and the year the result was published , so as to also hint at the history of ideas in the development of the theory of pure @xmath0-sequences . the multi - faceted interest in pure @xmath0-sequences is already indicated in their definition . on the one hand , a pure @xmath0-sequence can be defined as the vector whose entries record the number of monomials of a fixed degree in an order ideal generated by monomials of the same degree . on the other , a pure @xmath0-sequence is the hilbert function of a finite - dimensional graded algebra that is level , i.e. , its socle is concentrated in one degree , and has monomial relations . there is an extensive literature on monomial ideals and an extensive literature on level algebras . pure @xmath0-sequences form a bridge between the two theories , and we will outline the work on pure @xmath0-sequences from this point of view . but more than that , these sequences have a broad array of applications , and occur in many settings , largely combinatorial ones . in section [ sec : level ] , we review some of the basic results in the theory of pure @xmath0-sequences and focus on qualitative aspects of their shape . for instance , in some cases pure @xmath0-sequences are known to be unimodal ; that is , they are first weakly increasing , and once the peak is reached they are weakly decreasing . however , pure @xmath0-sequences may fail to be unimodal , even with arbitrarily many valleys . " we include a discussion of recent results on conditions that force unimodality . connections to various combinatorial problems are the subject of section [ sec : comb ] . face vectors of pure simplicial complexes are examples of pure @xmath0-sequences . in particular , the existence of certain block designs , such as steiner systems , is related to that of some pure @xmath0-sequences . as a special case , the existence of finite projective planes is equivalent to the existence of particular pure @xmath0-sequences . another challenging problem is stanley s conjecture that the @xmath2-vector of any matroid complex is a pure @xmath0-sequence . we discuss some recent progress . however , the conjecture remains open in general . in section [ sec : enum ] , we describe results on the enumeration of pure @xmath0-sequences , with a focus on asymptotic properties . in particular , it follows that , when the number of variables is large , `` almost all '' pure @xmath0-sequences are unimodal . we conclude this note with a collection of open problems , most of which are mentioned in the earlier sections . a finite , nonempty set @xmath3 of ( monic ) monomials in the indeterminates @xmath4 is called a _ monomial order ideal _ if , whenever @xmath5 and @xmath6 is a monomial dividing @xmath7 , then @xmath8 . the _ h - vector _ of @xmath3 is defined to be the vector @xmath9 counting the number of monomials of @xmath3 in each degree . a monomial order ideal , @xmath3 , is called _ pure _ if all maximal monomials of @xmath3 ( in the partial ordering given by divisibility ) have the same degree . a _ pure @xmath0-sequence _ is the @xmath2-vector of a pure monomial order ideal . for reasons that will be clear shortly , we call @xmath10 the _ socle degree _ of @xmath11 . the _ type _ of a pure @xmath0-sequence is the number of maximal monomials . notice that if we think of @xmath12 as the indeterminates of a polynomial ring @xmath13 $ ] over a field , the question of whether a given sequence is or is not a pure @xmath0-sequence does not depend on the choice of @xmath14 . thus , for many of our results it does not matter what we choose for @xmath14 . however , in some situations choosing a nice " field @xmath14 allows us to use special algebraic tools to say something about pure @xmath0-sequences , so in this case we make whatever additional assumptions we need for @xmath14 . we now let @xmath15 $ ] , where @xmath14 is an infinite field . we will consider standard graded artinian @xmath14-algebras @xmath16 , where @xmath17 will usually be a monomial ideal . without loss of generality we will assume that @xmath17 does not contain nonzero linear forms , so we will define @xmath18 to be the _ codimension _ of @xmath19 . let @xmath13 $ ] , and consider the action of @xmath20 on monomials of @xmath21 by contraction . by this we mean the action generated by @xmath22 for a monomial ideal @xmath23 , we define the _ inverse system _ to be the @xmath20-module @xmath24 . one can check that @xmath25 consists of the monomials not in @xmath17 ( identifying @xmath26 with @xmath27 ) , and as such it can be viewed as a monomial order ideal . recalling that for a standard graded algebra @xmath28 the _ hilbert function _ is defined to be @xmath29_t$ ] , we observe that the @xmath2-vector ( as defined above ) of the order ideal @xmath25 coincides with the hilbert function of @xmath28 . furthermore , @xmath25 is a pure monomial order ideal if and only if @xmath28 is a _ level _ algebra ; that is , the socle of @xmath28 ( i.e. , the annihilator of the homogeneous maximal ideal of @xmath28 ) is concentrated in one degree , called the _ socle degree _ of @xmath28 ; it is necessarily the degree of the maximal monomials of @xmath25 . the dimension of the socle as a @xmath14-vector space is equal to the type of the pure @xmath0-sequence . see @xcite for more details on inverse systems . thus the study of pure @xmath0-sequences boils down to a study of the possible hilbert functions of artinian monomial level algebras . in some cases we can rule out candidates for pure @xmath0-sequences by showing that there is not even a level algebra with that hilbert function , but more often we need to use the structure of monomial algebras themselves . the most basic tool is macaulay s theorem to determine if the sequence is even an @xmath0-sequence ; that is , to determine if it is the hilbert function of _ some _ artinian algebra . we refer to @xcite for details of macaulay s theorem , but we recall the statement . let @xmath30 and @xmath31 be positive integers . there exist uniquely determined integers @xmath32 such that @xmath33 this is called the _ @xmath31-binomial expansion of @xmath30_. we set @xmath34 and @xmath35 , for each @xmath31 . then macaulay s theorem is the following . [ macthm ] let @xmath19 be a standard graded algebra with hilbert function @xmath36 . then for all @xmath37 , @xmath38 . an _ @xmath0-sequence _ is a ( possibly infinite ) sequence of integers @xmath39 that satisfies the growth condition of theorem [ macthm ] for every value of @xmath40 . thus the @xmath0-sequences are the sequences that occur as the hilbert function of some standard graded algebra . the sequence @xmath41 is not a pure @xmath0-sequence because it is not even an @xmath0-sequence ( the growth from degree 4 to degree 5 is too big ) . similarly , @xmath42 is an @xmath0-sequence but it is not a pure @xmath0-sequence because it is not the hilbert function of a _ level _ algebra ( see @xcite ) . finally , @xmath43 is the hilbert function of a level algebra , but it is not a pure @xmath0-sequence @xcite because there is no _ monomial _ level algebra with this hilbert function . in fact , @xmath2 has been the first nonunimodal level hilbert function discovered in codimension 3 ( see the third author @xcite ) , and boyle @xcite has shown that this is in fact the smallest possible such hilbert function . so the challenge is to determine what additional conditions on an @xmath0-sequence are imposed by requiring that it be the hilbert function of an _ artinian level monomial _ algebra . the first result , due originally to stanley @xcite with subsequent proofs given by j. watanabe @xcite , ikeda @xcite , reid , roberts and roitman @xcite , herzog and popescu @xcite , and lindsey @xcite , concerns monomial complete intersections . ( note though that an equivalent property was proven earlier by de bruijn , van ebbenhorst tengbergen and kruyswijk @xcite . ) this result requires us to introduce here the notion of the weak and strong lefschetz properties . the consequence of this result for pure @xmath0-sequences of type 1 is perhaps not so critical , as alternative proofs could be given . but its influence in the study of the slp and the wlp in general , and on related topics in commutative algebra , can hardly be overstated . in the result below , and throughout the paper , we will call a sequence _ unimodal _ if it is nondecreasing up to some degree and then nonincreasing past that degree . we will call it _ strictly unimodal _ if it is strictly increasing up to some degree , then possibly constant for some range , then strictly decreasing until it reaches zero , and then zero past that point . unimodality is a central concept in combinatorics and combinatorial commutative algebra , but also in other branches of mathematics . see for instance the classical surveys of stanley ( @xcite , 1989 ) and brenti ( @xcite , 1994 ) . [ swrrr thm ] let @xmath15 $ ] , where @xmath14 has _ characteristic zero _ , and let @xmath17 be an artinian monomial complete intersection , i.e. , @xmath44 let @xmath45 be a general linear form . then for any positive integers @xmath31 and @xmath46 , the homomorphism induced by multiplication by @xmath47 , @xmath48_i \rightarrow [ r / i]_{i+d},\ ] ] has maximal rank . ( in particular , this is true when @xmath49 . ) as a consequence , a pure @xmath0-sequence of type 1 is strictly unimodal . when @xmath17 is an arbitrary artinian homogeneous ideal , the above maximal rank property for all @xmath31 and @xmath46 is called the _ strong lefschetz property ( slp ) _ , and the case @xmath49 is called the _ weak lefschetz property ( wlp)_. a consequence of the wlp is that in the range where @xmath50 is injective , we have a short exact sequence @xmath51_i \rightarrow [ r / i]_{i+1 } \rightarrow [ r/(i , l)]_{i+1 } \rightarrow 0.\ ] ] thus , the _ first difference _ @xmath52 is the hilbert function of a standard graded algebra in this range , that is , it is again an @xmath0-sequence . we then say that @xmath53 is a _ differentiable @xmath0-sequence _ in this range . it also follows from this sequence that if the wlp holds , then once the peak of the hilbert function is reached , the hilbert function will be nonincreasing , and therefore the whole hilbert function is unimodal . [ rem : positve - char ] simple examples show that theorem [ swrrr thm ] may fail in positive characteristic . it turns out that the question in which positive characteristics a monomial complete intersection has the weak or strong lefschetz property leads to unexpected connections to the problem of determining the number of certain plane partitions , lozenge tilings , or families of lattice paths ( see @xcite ) . one of the early important results on pure @xmath0-sequences is due to hibi @xcite . let @xmath11 be a pure @xmath0-sequence of socle degree @xmath10 . then @xmath54 whenever @xmath55 . this has the following two important consequences : * @xmath11 is _ flawless _ , i.e. , @xmath56 for all @xmath57 . * the `` first half '' of @xmath11 is nondecreasing : @xmath58 this latter result was later improved by the following algebraic _ @xmath59-theorem _ of hausel @xcite : [ hauselthm ] let @xmath19 be a monomial artinian level algebra of socle degree @xmath10 . if the field @xmath14 has characteristic zero , then for a _ general _ linear form @xmath45 , the induced multiplication @xmath60 is an injection , for all @xmath61 . in particular , over any field , the sequence @xmath62 is an @xmath0-sequence , i.e. , the `` first half '' of @xmath11 is a _ differentiable @xmath0-sequence_. we have the following additional results on differentiability from @xcite : [ thm : diff - o ] * every finite differentiable @xmath0-sequence @xmath11 is the first half " of some pure @xmath0-sequence . ( this is the converse of hausel s theorem . ) * in particular , any finite differentiable @xmath0-sequence is pure ( by truncation ) . * any nondecreasing pure @xmath0-sequence of socle degree @xmath63 is differentiable . it turns out that ( c ) is the best possible result in this direction : there exist nondecreasing pure @xmath0-sequences of any socle degree @xmath64 that are not differentiable . [ not diff ] we illustrate the preceding result with an example from @xcite . observe first that the @xmath2-vector @xmath65 is a pure @xmath0-sequence since it is the @xmath2-vector of the truncation of a polynomial ring in four variables , @xmath66 ; the pure order ideal arises using all 35 monomials of degree 4 in @xmath66 . the @xmath2-vector @xmath67 is also a pure @xmath0-sequence , since it is the order ideal generated by a monomial @xmath68 in four new variables . now we work in a polynomial ring in the eight variables @xmath69 , and we consider the pure order ideal generated by the above 36 monomials of degree 4 . the resulting @xmath2-vector @xmath2 is [ cols="^,^,^,^,^,^,^,^,^,^,^,^ " , ] since the first difference of @xmath70 is @xmath71 , which is not an @xmath0-sequence ( because @xmath72 ) , we have constructed the desired example . putting aside the class of nondecreasing pure @xmath0-sequences , we now turn to the question of unimodality . there are three factors that go into whether a nonunimodal example will exist : the codimension ( i.e. , the number of variables ) , the socle degree , and the type . an easy application of macaulay s theorem gives that any standard graded algebra of codimension two has unimodal hilbert function , and in fact it is not hard to show that if @xmath73/i$ ] is level ( monomial or not ) then the hilbert function is strictly unimodal ( see e.g. iarrobino @xcite , 1984 ) . hence the interesting questions arise for codimension @xmath74 . we will begin with some results involving the socle degree ( some of which will also bring in the codimension ) . it follows from hibi s result on flawlessness that any pure @xmath0-sequence of socle degree @xmath63 is unimodal . the next case , socle degree 4 , already is not necessarily unimodal , again thanks to an example from @xcite : observe that @xmath75 is a pure @xmath0-sequence , since it is the @xmath2-vector of the truncation of a polynomial ring in five variables , and as before @xmath76 is also pure , since it corresponds to the maximal monomial @xmath77 $ ] . hence , reasoning as above , we now consider one copy of @xmath78 and eleven copies of @xmath79 as @xmath2-vectors of pure @xmath0-sequences in twelve different rings , and we work in the tensor product of those rings . it follows that @xmath80 is a nonunimodal pure @xmath0-sequence of socle degree 4 . a natural question , then , is what is the smallest codimension for which nonunimodal pure @xmath0-sequences exist with socle degree 4 . this remains open . we do have the following results from @xcite , however . * all pure @xmath0-sequences of socle degree @xmath81 in three variables are unimodal . * all pure @xmath0-sequences of socle degree @xmath82 in four variables are unimodal . * in four or more variables , there exist nonunimodal pure @xmath0-sequences in all socle degrees @xmath83 . in @xcite , boij and the third author gave a nonunimodal pure @xmath0-sequence of codimension 3 and socle degree 12 . this is the smallest known example in codimension 3 . it follows from this and boyle s result that in codimension 3 , the only open cases are socle degrees 10 and 11 . notice that , in codimension 4 , the previous theorem leaves only open the socle degrees 5 and 6 . now we turn to questions involving the type . of course it follows immediately from theorem [ swrrr thm ] that pure @xmath0-sequences of type 1 are unimodal . what else can be deduced about pure @xmath0-sequences using the wlp ? a collection of results was obtained in @xcite which showed , in some sense , the limits of the wlp in the study of pure @xmath0-sequences . [ wlp situation ] over a field of characteristic zero the following hold : * any monomial artinian level algebra of type 2 in three variables has the wlp . thus a pure @xmath0-sequence of type 2 and codimension 3 is differentiable until it reaches its peak , is possibly constant , and then is nonincreasing until it reaches zero . * fix two positive integers @xmath18 and @xmath31 . then all monomial artinian level algebras of codimension @xmath18 and type @xmath31 possess the wlp if and only if at least one the following is true : * * @xmath84 or 2 ; * * @xmath49 ; * * @xmath85 and @xmath86 . the proof of ( a ) was surprisingly long and intricate . the main point of ( b ) is that in all other cases , we were able to show that artinian monomial level algebras exist that do _ not _ have the wlp . notice that theorems [ swrrr thm ] , [ hauselthm ] and [ wlp situation](a ) require that @xmath14 have characteristic zero . the statements about injectivity and surjectivity require this property of the characteristic , and indeed a great deal of research has been carried out to see what happens when @xmath14 has positive characteristic ; we refer to @xcite for an overview of these results . the consequences on the shape of the pure @xmath0-sequences are indeed characteristic free , as has been noted above , since the hilbert function of a monomial ideal does not depend on the characteristic . if one is studying all artinian level monomial algebras of fixed type , theorem [ wlp situation ] is a serious limitation on the usefulness of the wlp . however , in the study of pure @xmath0-sequences ( e.g. to determine combinations of codimension , type and socle degree that force unimodality ) it is still conceivable that the wlp will play a useful role . as a trivial example , we know that monomial complete intersections in any codimension possess the wlp , thanks to theorem [ swrrr thm ] . it is not known ( except in codimension @xmath63 , as noted below ) whether _ all _ complete intersections have this property , but the knowledge in the special case of monomial ideals is enough to say that all complete intersection hilbert functions are unimodal . perhaps a similar phenomenon will allow the wlp to continue to play a role in the study of pure @xmath0-sequences . a first approach using this philosophy was obtained by cook and the second author @xcite , where they _ lifted _ a monomial ideal to an ideal of a reduced set of points in one more variable , showed that the general artinian reduction has the wlp , and concluded that the hilbert function of the original monomial algebra is unimodal , regardless of whether it has the wlp or not . for this reason we mention a useful tool in studying the wlp , that was introduced by the first and second authors with t. harima and j. watanabe in @xcite and whose study was continued by h. brenner and a. kaid in @xcite . this is the study of the _ syzygy bundle _ , and the use of the grauert - mlich theorem ( see @xcite ) . it was used in @xcite to show that _ any _ complete intersection @xmath17 in @xmath87 $ ] over a field of characteristic zero has the wlp . the idea is to restrict to a general line , say one defined by a general linear form @xmath45 . the key is that the restricted ideal @xmath88 ( which now has codimension 2 , hence has a hilbert - burch matrix ) should have minimal syzygies in two consecutive degrees , at most . the idea of @xcite was that for height 3 complete intersections , this information can be obtained by considering the module of syzygies of @xmath17 , sheafifying it to obtain the syzygy bundle , and applying the grauert - mlich theorem to the general line defined by @xmath45 . of course this introduces questions about the semistability of the syzygy bundle , which we mostly omit here . in a more general setting , the following result from @xcite summarizes the idea nicely , at least for codimension three . let @xmath89 = r$ ] be an artinian homogeneous ideal whose syzygy bundle @xmath90 is semistable on @xmath91 . then * if the restriction of @xmath90 splits on a general line @xmath45 as @xmath92 then @xmath16 has the wlp . * if the restriction of @xmath90 splits on a general line @xmath45 as @xmath93 with @xmath94 and @xmath95 , then @xmath16 does not have the wlp . other applications of this approach , for higher codimension , can be found for instance in @xcite . returning to unimodality questions , tools other than the wlp will also be needed . we have the following theorems of boyle @xcite , which relied on decomposition results and an analysis of complete intersection hilbert functions but did not use the wlp . * in codimension 3 , all pure @xmath0-sequences of type 3 are strictly unimodal . * in codimension 4 , all pure @xmath0-sequences of type 2 are strictly unimodal . a natural question is whether all pure @xmath0-sequences of type 2 and arbitrary codimension are unimodal . also , for any fixed codimension , it is an interesting problem to determine which types force unimodality . the record " for the smallest known nonunimodal example in codimension 3 is type 14 , given in @xcite . finally , one can ask `` how nonunimodal '' a pure @xmath0-sequence can be . the answer is `` as nonunimodal as you want . '' we have : [ arb peaks ] for any integers @xmath96 and @xmath74 , there exists a pure @xmath0-sequence in @xmath18 variables which is nonunimodal and has exactly @xmath7 maxima . of course the price " in theorem [ arb peaks ] is paid in having a large socle degree and type . in fact , even _ cohen - macaulay @xmath97-vectors _ ( i.e. , the face vectors of cohen - macaulay simplicial complexes , which are a much smaller subset of pure @xmath0-sequences ) can be nonunimodal with arbitrarily many peaks ( see @xcite ) . this result considerably extends theorem [ arb peaks ] , even though , unlike for arbitrary pure @xmath0-sequences , here the number of variables becomes necessarily very large as the number of peaks increases . a good topic to build a bridge between the algebraic and the combinatorial sides of the theory of pure @xmath0-sequences is the _ interval conjecture_. the _ interval property _ ( ip ) was introduced in 2009 by the third author @xcite , where he conjectured its existence for the set of hilbert functions of level and , in a suitably symmetric way , gorenstein algebras . namely , the ip says that if two ( not necessarily finite ) sequences , @xmath2 and @xmath98 , of a class @xmath99 of integer sequences coincide in all entries but one , say @xmath100 for some index @xmath46 and some positive integer @xmath101 , then the sequences @xmath102 are also in @xmath99 , for all @xmath103 given that level and gorenstein hilbert functions are nearly impossible to characterize , the ip appears to be both a very natural property and one of the strongest structural results that we might hope to achieve for the set of such sequences . for example , it is proved in @xcite that the ip holds for all gorenstein hilbert functions of socle degree 4 . since gorenstein hilbert functions are symmetric , this means that , for any fixed @xmath18 , @xmath104 is gorenstein if and only if @xmath105 ranges between some minimum possible value , say @xmath106 , and @xmath107 . this latter is the maximum allowed by a polynomial ring in @xmath18 variables , and is achieved by the so - called _ compressed _ gorenstein algebras ( see , e.g. , the 1984 papers of frberg - laksov and iarrobino @xcite or the third author s works @xcite ) . notice that the existence of the ip for these hilbert functions is especially helpful in view of the fact that , for most codimensions @xmath18 , the value of @xmath106 is not known . in fact , such gorenstein hilbert functions are `` highly '' nonunimodal . asymptotically , we have @xmath108 as proved by these three authors in @xcite ( 2008 ) , solving a longstanding conjecture of stanley @xcite ( see also @xcite , 2009 , for some broad generalizations ) . the interval property is still wide open today for both level and gorenstein algebras . in a more combinatorial direction , in bmmnz 2012 the ip has then also been conjectured for : 1 ) pure @xmath0-sequences ( under the name `` icp '' ) ; and 2 ) the @xmath97-vectors of pure simplicial complexes , a topic of discussion of the next section . as for pure @xmath0-sequences , the icp has been proved in a number of special cases . most importantly , it is known when the socle degree is at most 3 , in any number of variables . the icp holds for the set of all pure @xmath0-sequences of socle degree @xmath109 . thanks to this result , h.t . h , e. stokes and the third author @xcite have recently developed a new approach leading to a proof of stanley s matroid @xmath2-vector conjecture in krull - dimension 3 , as we will see in the next section . while the icp remains open in most instances e.g. , in three variables it must be pointed out that , just recently , it has been _ disproved _ in the four variable case by constantinescu and varbaro ( see ( * ? ? ? * remark 1.10 ) ) , who found the following counterexample . [ cv ex ] consider the pure order ideals generated by @xmath110 and @xmath111 . their @xmath2-vectors are the pure @xmath0-sequences @xmath112 and @xmath113 . however , an exhaustive computer search over all sets of three monomials of degree 6 in four variables reveals that the sequence @xmath114 is not pure , contrary to the icp . it is worth remarking that @xmath115 is , however , a level @xmath2-vector , and so this does not provide a counterexample to the ip for arbitrary level algebras . indeed , @xmath2 is the @xmath2-vector of a level algebra in four variables whose inverse system is generated by two sums of sixth powers of six general linear forms each , and the sixth power of one general linear form . as for pure @xmath97-vectors , the ip is still wide open , and little progress has been made so far . in general , at this time it is still unclear what the exact scope of the interval property is , and if it can also be of use in other areas of combinatorial algebra or even enumerative combinatorics . it is well known to hold , e.g. , for the set of hilbert functions of graded algebras of any krull - dimension ( see macaulay s theorem ) , the @xmath97-vectors of arbitrary simplicial complexes ( the kruskal - katona theorem ) , and the @xmath97-vectors of cohen - macaulay complexes ( bmmnz 2012 ) . instead , the ip fails quite dramatically , for example , for matroid @xmath2-vectors , which are conjecturally another subset of pure @xmath0-sequences , as we will see in the next section ( we refer to bmmnz 2012 and @xcite for details ) . stanley and the third author @xcite recently looked at the ip in the context of _ @xmath18-differential posets _ , a class of ranked posets generalizing the young lattice of integer partitions and the young - fibonacci lattice . here , even though the ip fails in general , it might be a reasonable property to conjecture , for instance , for @xmath84 , which is the most natural class of differential posets . much of the motivation for the study of pure @xmath0-sequences comes from combinatorics , and in this section we give an overview of this side of the theory . in order to put in context the definition of a pure @xmath0-sequence given in the introduction , we quickly recall the notion of posets and order ideals . for an introduction to this theory , we refer to chapter 3 of stanley s new edition of `` ec1 '' ( @xcite , 2012 ) . a ( short for _ partially ordered set _ ) is a set @xmath99 equipped with a binary relation , `` @xmath116 '' that is : 1 ) reflexive ( i.e. , @xmath117 for all @xmath118 ) ; 2 ) antisymmetric ( @xmath119 implies @xmath120 ) ; and 3 ) transitive ( @xmath121 implies @xmath122 ) . an _ order ideal _ in a poset @xmath99 is a subset @xmath17 of @xmath99 that is closed with respect to `` @xmath123 '' that is , if @xmath124 and @xmath125 , then @xmath126 . thus , our monomial order ideals are the ( finite ) order ideals of the poset @xmath127 of all monomials in the polynomial ring @xmath128 $ ] , where the binary relation of @xmath127 is divisibility . notice that @xmath127 is a _ ranked _ poset , where the rank of a monomial is its degree in @xmath20 . therefore , macaulay s @xmath0-sequences are exactly the possible rank functions of the order ideals of @xmath127 , since every hilbert function satisfying macaulay s theorem can be achieved by a monomial algebra . similarly , as we have seen , a pure @xmath0-sequence is the rank function of some monomial order ideal whose _ generators _ ( i.e. , the antichain of maximal monomials ) are all of the same degree . another fundamental class of order ideals are those contained in the _ boolean algebra _ @xmath129 , the poset of all subsets of @xmath130 , ordered by inclusion . by identifying the integer @xmath46 with a vertex @xmath131 , the order ideals of @xmath129 are usually called _ simplicial complexes _ ( on @xmath18 vertices ) . the elements of a simplicial complex @xmath132 are dubbed _ faces _ , and the maximal faces are the _ facets _ of @xmath132 . the _ dimension of a face _ is its cardinality minus 1 , and the _ dimension of @xmath132 _ is the largest of the dimensions of its faces . notice that if we identify @xmath131 with a variable @xmath133 $ ] , then simplicial complexes also coincide with the order ideals of @xmath127 generated by _ squarefree _ monomials . in particular , if we define as _ pure _ those simplicial complexes whose facets have all the same dimension , then clearly their rank vectors , called _ pure @xmath97-vectors _ , are the special subset of pure @xmath0-sequences that can be generated by squarefree monomials . the simplicial complex @xmath134 is the order ideal of @xmath135 generated by @xmath136 and @xmath137 . thus , @xmath132 is a pure complex of dimension 2 , whose pure @xmath97-vector is @xmath138 . equivalently , @xmath139 is the pure @xmath0-sequence generated by the two squarefree monomials @xmath140 and @xmath141 . similarly to macaulay s theorem for arbitrary @xmath0-sequences , we know a characterization of the class of pure @xmath97-vectors thanks to the classical _ kruskal - katona theorem _ ( see e.g. @xcite ) . however , analogously , things become dramatically more complicated ( hopeless , we should say ) when it comes to attempting a characterization of _ pure _ @xmath97-vectors . in the last section of bmmnz 2012 , we have begun a study of pure @xmath97-vectors , but still very little is known today beyond what is known for arbitrary pure @xmath0-sequences . besides their obvious intrinsic importance simplicial complexes are a central object in algebraic combinatorics , combinatorial algebra and topology , just to name a few subjects pure @xmath97-vectors also carry fascinating applications . it is on their connections to finite geometries and design theory that we want to focus in the next portion of this section . it will follow from our discussion , as probably first observed by bjrner ( @xcite , 1994 ) , that a characterization of pure simplicial complexes and their @xmath97-vectors would imply , for instance , that of all steiner systems , and as a further special case , a classification of all finite projective planes , one of the major open problems in geometry . a _ steiner system _ @xmath142 is an @xmath18-element set @xmath143 , together with a collection of @xmath144-subsets of @xmath143 , called _ blocks _ , such that every @xmath145-subset of @xmath143 is contained in exactly one block . steiner systems are a special family of the so - called _ block designs_. we refer our reader to the two texts @xcite , where she can find a truly vast amount of information on combinatorial designs . for instance , a steiner _ triple _ system ( sts ) is a steiner system @xmath146 , while @xmath147 is dubbed a steiner _ system , where @xmath18 is the _ order _ of the system . since we are dealing with maximal sets of the same cardinality ( @xmath144 , in this case ) , it is clear that if we identify each element of @xmath143 with a variable @xmath27 , then the existence of steiner systems ( and similarly for other block designs ) will be equivalent to the existence of certain pure @xmath97-vectors . let us consider sts s of order 7 , i.e. , @xmath148 . constructing such a design is tantamount to determining a family of squarefree degree 3 monomials of @xmath149 $ ] , say @xmath150 , such that each squarefree degree 2 monomial of @xmath20 divides exactly one of the @xmath151 . clearly , since there are @xmath152 squarefree degree 2 monomials in @xmath20 , if the @xmath151 exist , then @xmath153 . in other words , @xmath148 exists if and only if @xmath154 is a pure @xmath97-vector . notice also that @xmath97 exists as a pure @xmath97-vector if and only if it exists as a pure @xmath0-sequence , since for seven degree 3 monomials to have a total of 21 degree 2 divisors , each needs to have exactly three linear divisors , i.e. , it must be squarefree . it is easy to see that an sts of order 7 , and so the pure @xmath97-vector @xmath155 , do indeed exist , using the monomial order ideal generated by : @xmath156 some simple numerical observations show that a necessary condition for an sts of order @xmath18 to exist is that @xmath18 be congruent to 1 or 3 modulo 6 , and it is a classical result of kirkman ( @xcite , 1847 ) that this is also sufficient . in other words , @xmath157 is a pure @xmath97-vector , if and only if it is a pure @xmath0-sequence , if and only if @xmath18 is congruent to 1 or 3 modulo 6 . a different , and more challenging , problem is the classification of all steiner systems , up to isomorphism . even the existence of particular systems sometimes brings into the story a nontrivial amount of interesting algebra . as an illustration , we mention here the case of the steiner systems @xmath158 , @xmath159 , @xmath160 , @xmath161 , and @xmath162 , which are intimately connected to the first sporadic finite simple groups ever discovered , called the _ mathieu groups _ ( see e.l . mathieu @xcite , 1861 and 1873 ) . these five groups denoted respectively by @xmath163 , @xmath164 , @xmath165 , @xmath166 , and @xmath167 in fact arise as the _ automorphism groups _ of the above steiner systems ( i.e. , the transformations of the systems that preserve the blocks ) . there exists only one sts of order 7 , which is called the _ fano plane _ ( see the figure below ) for reasons that will be clear in a minute . in other words , the seven monomials of the previous example are , up to isomorphism , the only possible set of generators for a pure order ideal in @xmath168 $ ] with @xmath169 as its pure @xmath0-sequence . ( 150,0 ) ( 223,148)@xmath170 ( 164,95)@xmath171 ( 256,95)@xmath172 ( 226,72)@xmath173 ( 134,37)@xmath174 ( 209,25)@xmath175 ( 286,37)@xmath176 also for @xmath177 , there exists a unique sts . however , it is reasonable to believe that the number of nonisomorphic sts increases extremely quickly for @xmath18 large . for instance , there are 80 nonisomorphic sts of order 15 , and there are @xmath178 of order 19 . similarly , the possible orders of steiner quadruple systems are known and nicely characterized , since the obvious necessary conditions again turn out to be also sufficient : @xmath147 exists if and only if its order @xmath18 is congruent to 2 or 4 modulo 6 , as proved by hanani ( @xcite , 1960 ) . in other words , reasoning as above , since there are four possible 3-subsets of any given 4-set , we have that @xmath179 is a pure @xmath97-vector if and only if it is a pure @xmath0-sequence , if and only if @xmath18 is congruent to 2 or 4 modulo 6 . however , as we increase the cardinality @xmath144 of the blocks , things become more and more obscure . this is due to the high complexity of computing combinatorial designs over a large vertex set , as well as to the lack of a general theory . already for steiner quintuple systems , the trivial necessary conditions ( @xmath18 congruent to 3 or 5 , but not to 4 , modulo 6 ) are no longer sufficient . for example , no steiner quintuple system @xmath180 exists ( see @xcite , 2008 ) . the smallest value of @xmath18 for which the existence of @xmath181 is currently open is @xmath182 . in other words , it is unknown whether @xmath183 is a pure @xmath0-sequence . perhaps the best - known family of examples of steiner systems is that of finite projective planes , so they deserve a special mention here . recall that a _ projective plane _ is a collection of points and lines such that any two lines `` intersect at '' exactly one point , and any two points `` lie on '' exactly one line ( one also assumes that there exist four points no three of which are collinear , in order to avoid uninteresting pathological situations ) . if the projective plane is finite , it can easily be seen that the number of points is equal to the number of lines , and that this number must be of the form @xmath184 , where the integer @xmath185 is the _ order _ of the plane . further , in a projective plane of order @xmath185 , any line contains exactly @xmath186 points , and by duality , any point is at the intersection of exactly @xmath186 lines . in other words , finite projective planes are the steiner systems @xmath187 . the reader may want to consult , e.g. , @xcite for an introduction to this area . thus , the above example of a steiner system @xmath148 , the fano plane , is the unique smallest possible projective plane . similarly to how we argued earlier in terms of pure @xmath97-vectors , one can show that a projective plane of order @xmath185 exists if and only if @xmath188 is a pure @xmath97-vector , if and only if it is a pure @xmath0-sequence . a major open problem in geometry asks for a classification of all finite projective planes , or even just of the possible values that @xmath185 may assume . conjecturally , @xmath185 is always the power of a prime , and it is a standard algebraic exercise , using finite field theory , to construct a projective plane of any order @xmath189 . the best general necessary condition known today on @xmath185 is still the following theorem from @xcite : if @xmath185 is the order of a projective plane and @xmath185 is congruent to 1 or 2 modulo 4 , then @xmath185 is the sum of two squares . thus , for instance , as a consequence of the bruck - ryser - chowla theorem , no projective plane of order 6 exists . in other words , @xmath190 is not a pure @xmath0-sequence . however , already ruling out the existence of projective planes of order 10 has required a major computational effort ( see lam @xcite , 1991 ) . the case @xmath191 is still open . notice that , at least for certain values of @xmath185 , the number of nonisomorphic projective planes of order @xmath185 can be very large , and a general classification seems entirely out of reach . the smallest @xmath185 for which there exists more than one nonisomorphic projective plane is @xmath192 , where the four possible cases were already known to o. veblen ( @xcite , 1907 ) . the second important application of pure @xmath0-sequences that we want to discuss brings our attention to a very special class of simplicial complexes , called _ matroid _ complexes . matroids are ubiquitous in mathematics , where they often show up in surprising ways ( see @xcite ) . the algebraic theory of matroids began in the same 1977 seminal paper of stanley @xcite that introduced pure @xmath0-sequences . a finite matroid can be naturally identified with a pure simplicial complex over @xmath193 , such that its restriction to any subset of @xmath143 is also a pure complex . one associates , to any given simplicial complex @xmath132 over @xmath194 , the following squarefree monomial ideal in @xmath195 $ ] , where @xmath14 is a field : @xmath196 @xmath197 is called the _ stanley - reisner ideal _ of the complex @xmath132 , and the quotient algebra @xmath198 is its _ stanley - reisner ring_. it is a standard fact of combinatorial commutative algebra ( see e.g. @xcite ) that the stanley - reisner ring @xmath198 of a matroid complex is cohen - macaulay and level , although of course of positive krull - dimension ( except in degenerate cases ) . thus , the @xmath2-vector of @xmath198 is level , since it is the @xmath2-vector of an artinian reduction of @xmath198 . however , even though @xmath198 is presented by monomials , notice that its artinian reductions will in general be far from monomial , for they require taking quotients by `` general enough '' linear forms . the following spectacularly simple conjecture of stanley ( @xcite , 1977 ) predicts that , for any matroid complex @xmath132 , we can nonetheless find _ some _ artinian monomial level algebra having the @xmath2-vector of @xmath198 as its @xmath2-vector : [ stst ] any matroid @xmath2-vector is a pure @xmath0-sequence . the problem of characterizing matroid @xmath2-vectors appears to be once again hopeless , and conjecture [ stst ] has motivated much of the algebraic work done on matroids over the past 35 years ( see , as a highly nonexhaustive list , @xcite ) . the main approach to conjecture [ stst ] has been , given the @xmath2-vectors of a certain class of matroids , to explicitly produce some pure monomial order ideals having those matroid @xmath2-vectors as their pure @xmath0-sequences . recently , h.t . h , e. stokes and the third author @xcite introduced a `` more abstract '' approach to conjecture [ stst ] . their main idea , inspired by the latest progress on pure @xmath0-sequences made in bmmnz 2012 , and in particular the proof of the interval conjecture ( icp ) in socle degree 3 , has been to try to reduce stanley s conjecture , as much as possible , to one on the properties of pure @xmath0-sequences , thus avoiding explicit construction of a monomial ideal for each matroid @xmath2-vector . the approach of @xcite has already led to a proof of conjecture [ stst ] for all matroid complexes of krull - dimension at most 2 ( the dimension 1 case , that had been the focus of a large portion of the thesis @xcite , simply followed in a few lines ) . all matroid @xmath2-vectors @xmath199 are pure @xmath0-sequences . more generally , the following is a first concrete , if still tentative , general approach to conjecture [ stst ] ( see @xcite ) . assuming conjecture [ stst ] holds for all matroid complexes whose deletions with respect to any vertex are cones ( which may not be too difficult to show with the techniques of paper @xcite ) , conjecture [ stst ] is true in general under the following two natural ( but still too bold ? ) assumptions : 1 . _ any matroid @xmath2-vector is differentiable for as long as it is nondecreasing . _ ( in fact , incidentally , would a _ @xmath59-element _ , that hausel @xcite and swartz @xcite proved to exist in the `` first half '' of a matroid , carry on all the way ? ) 2 . _ suppose that the _ shifted sum _ , @xmath200 , of two pure @xmath0-sequences @xmath2 and @xmath98 is differentiable for as long as it is nondecreasing . then @xmath201 is also a pure @xmath0-sequence . _ as we have seen above , pure @xmath0-sequences arise in several areas , yet their properties are not well understood , and there are other important questions that should be addressed even if a classification is not available . for example , one would like to estimate the number of pure @xmath0-sequences of given codimension and socle degree . what happens asymptotically ? moreover , we have seen that pure @xmath0-sequences can be as far from being unimodal as we want . nevertheless , one may ask : what are the odds for a pure @xmath0-sequence to be unimodal ? in order to discuss such questions , let us denote by @xmath202 , @xmath203 , and @xmath204 the sets of @xmath0-sequences , pure @xmath0-sequences , and differentiable @xmath0-sequences , respectively , that have codimension @xmath18 and socle degree @xmath10 . recall that given two functions @xmath205 , one says that @xmath97 is _ asymptotic _ to @xmath59 , and writes @xmath206 , if @xmath207 . all limits are taken for @xmath18 approaching infinity . consider now an @xmath0-sequence @xmath208 in @xmath209 . integrating it , that is , passing to @xmath210 , provides a differentiable @xmath0-sequence in @xmath204 . thus , since finite differentiable @xmath0-sequences are pure by theorem [ thm : diff - o](b ) , we have the following inclusions : @xmath211 results by linusson ( see @xcite ) imply that , for @xmath18 large , the cardinalities of @xmath209 and @xmath202 are asymptotically equal . it follows that in large codimensions almost all @xmath0-sequences are pure . more precisely , one has : [ almost ] fix a positive integer @xmath10 . then , for @xmath18 large , almost all @xmath0-sequences of socle degree @xmath10 are differentiable . namely , @xmath212 [ c_e ] where @xmath213 since pure @xmath0-sequences are hilbert functions of level algebras , we immediately get the following consequence . [ level ] fix a positive integer @xmath10 . let @xmath214 be the set of _ level _ hilbert functions of codimension @xmath18 and socle degree @xmath10 . then , for @xmath18 large , almost all level sequences are pure and unimodal , and @xmath215 the outcome changes drastically if we fix as additional parameter the socle type @xmath40 . since each monomial of degree @xmath40 is divisible by at most @xmath40 distinct variables , we observe : [ 000 ] let @xmath216 be the set of pure @xmath0-sequences of codimension @xmath18 , socle degree @xmath10 , and type @xmath40 . then @xmath217 for @xmath218 , this bound being sharp . note that , in contrast to this result , there is no analogous restriction on level hilbert functions . for example , gorenstein algebras have type 1 and admit any positive socle degree and codimension . in fact , asymptotically , the number of their hilbert functions is known . recall that an _ si - sequence _ of socle degree @xmath10 is an @xmath0-sequence that is symmetric about @xmath219 and differentiable up to degree @xmath220 . initially , stanley and iarrobino ( see @xcite ) had hoped that all hilbert functions of gorenstein algebras were si - sequences . although this is not true ( see @xcite for the first counterexample ) , it is _ almost true _ ! in fact , any differentiable @xmath0-sequence in @xmath221 can be extended to a symmetric sequence , so that the result is an si - sequence . moreover , every si - sequence is the hilbert function of some gorenstein algebra ( see @xcite ) . obviously , the first half of the hilbert function of a gorenstein algebra is an @xmath0-sequence . taken together , it follows that the number of gorenstein hilbert functions that are not si - sequences is negligible : [ gorsi ] fix a positive integer @xmath10 . let @xmath222 be the set of gorenstein hilbert functions of codimension @xmath18 and socle degree @xmath10 , and let @xmath223 be the set of si - sequences of codimension @xmath18 and socle degree @xmath10 . then , for @xmath18 large , almost all gorenstein hilbert functions are si - sequences . more precisely , @xmath224 returning to pure @xmath0-sequences , it would be very interesting to determine , or at least to find a good estimate of , the number of pure @xmath0-sequences of codimension @xmath18 , socle degree @xmath10 , and type @xmath40 . this seems a difficult problem . however , in the simplest case , where @xmath225 , there is an easy combinatorial answer . in fact , any pure @xmath0-sequence of type 1 is the hilbert function of a ( complete intersection ) algebra whose inverse system is a monomial of the form @xmath226 , where @xmath227 and @xmath228 for all @xmath46 . we may assume that @xmath229 , so that @xmath230 is a _ partition _ of @xmath10 . since it is easy to see that distinct partitions lead to different hilbert functions , we arrive at the following result . [ p_r ] @xmath231 , the number of partitions of the integer @xmath10 having exactly @xmath18 parts . we now consider a slightly different asymptotic enumeration question . fix positive integers @xmath10 and @xmath40 . since the number of monomials dividing @xmath40 monomials of degree @xmath10 is finite , there are only finitely many pure @xmath0-sequences of socle degree @xmath10 and type @xmath40 . denote their set by @xmath232 . determining @xmath233 exactly seems out of reach . however , one may hope to at least find its order for @xmath40 large . the first interesting case , namely @xmath234 , has been settled . [ thmasymptoticsocledegree3 ] let @xmath235 denote the number of pure @xmath0-sequences of socle degree 3 and type @xmath40 . then @xmath236 its proof gives some further information . consider a pure @xmath0-sequence @xmath237 . then @xmath238 by hibi s theorem and the fact that @xmath40 monomials of degree 3 are divisible by at most @xmath239 distinct quadratic monomials . it follows that @xmath240 . to see that the two functions are in fact asymptotically equal , notice that , for fixed @xmath40 , the possible values of @xmath18 and @xmath105 fall into one of the following three regions , illustrated in the figure below : region i : @xmath241 ; region ii : @xmath242 ; region iii : @xmath243 . ( 170,160)(10,10 ) ( 10,10)(0,1)170 ( 10,10)(1,0)170 ( 10,10)(1,1)160 ( 180,0)@xmath18 ( 0,180)@xmath105 ( 60,10)(0,1)150 ( 160,10)(0,1)150 ( 10,60)(1,0)150 ( 10,160)(1,0)150 ( 60,-2)@xmath40 ( 156,-2)@xmath239 ( -4,156)@xmath239 ( 0,60)@xmath40 ( 87,120)i ( 27,120)ii ( 15,40)iii using superscripts to denote the sets of pure @xmath0-sequences in each region , it is shown in bmmnz that @xmath244 the arguments use in a crucial way the interval property for pure @xmath0-sequences of socle degree 3 . unfortunately , this property fails in general . nevertheless it would be very interesting to extend the above results to socle degree @xmath245 . in this section we collect a few interesting problems that remain open in the area of pure @xmath0-sequences . most of them have been discussed in the previous sections , but some are related problems that were not addressed above . 1 . what is the largest type @xmath40 for which all pure @xmath0-sequences are unimodal ( independently of the codimension or socle degree ) ? even proving that @xmath246 , i.e. , that pure @xmath0-sequences of type 2 are unimodal in any codimension , would be very interesting . 2 . for a fixed codimension @xmath18 , what is the largest type @xmath40 for which all pure @xmath0-sequences are unimodal ? 3 . for a fixed codimension @xmath18 , what is the largest socle degree @xmath10 for which all pure @xmath0-sequences are unimodal ? 4 . what is the smallest codimension @xmath18 for which there exists a nonunimodal pure @xmath0-sequence of socle degree 4 ? 5 . determine asymptotically the number @xmath233 of pure @xmath0-sequences of socle degree @xmath10 and type @xmath40 , for @xmath40 large . what is the order of magnitude of @xmath233 ? 6 . the first example of a nonunimodal pure @xmath0-sequence ( due to stanley @xcite , 1977 ) was @xmath247 , which is in fact the @xmath97-vector of a cohen - macaulay simplicial complex , hence in particular a pure @xmath97-vector . what is the smallest number of variables @xmath30 ( i.e. , the number of vertices of the complex ) allowing the existence of a nonunimodal pure @xmath97-vector ? + a. tahat @xcite has recently determined the sharp lower bound @xmath248 for nonunimodal _ cohen - macaulay _ @xmath97-vectors of socle degree 5 ( the socle degree of stanley s original example ) , and has produced examples in larger socle degree with @xmath30 as low as 39 . 7 . stanley s twenty - fifth problem for the year 2000 imu volume `` mathematics : frontiers and perspectives '' @xcite asks , among a few other things : are all matroid @xmath97-vectors unimodal , or even log - concave ? what about matroid @xmath2-vectors ? + notice that matroid @xmath97-vectors are a much smaller subset of cohen - macaulay @xmath97-vectors . for a major recent breakthrough on this problem , see huh @xcite , and huh and katz @xcite ; as a consequence of their work , log - concavity ( hence unimodality ) is now known for all @xmath97-vectors ( see lenz @xcite , 2012 ) and @xmath2-vectors ( huh @xcite ) of _ representable _ matroids . notice also that , in general , the log - concavity of matroid @xmath2-vectors would imply the log - concavity of matroid @xmath97-vectors , as proved by dawson @xcite ( 1984 ) and lenz @xcite . we wish to thank david cook ii and richard stanley for helpful comments . the third author also thanks jrgen bierbrauer for an interesting discussion on the connections between group theory and steiner systems . we thank the referee for a careful reading of the paper . a. bjrner : _ nonpure shellability , @xmath97-vectors , subspace arrangements and complexity _ , in formal power series and algebraic combinatorics , " series formelles et combinatoire algbrique ( 1994 ) , dimacs series in discrete mathematics and theoretical computer science , volume 24 , 1994 , 2554 . a.v . geramita : _ inverse systems of fat points : waring s problem , secant varieties and veronese varieties and parametric spaces of gorenstein ideals _ , queen s papers in pure and applied mathematics , no . 102 , the curves seminar at queen s ( 1996 ) , vol . x , 3114 . e. mathieu : _ mmoire sur ltude des fonctions de plusieurs quantits , sur la manire de les former et sur les substitutions qui les laissent invariables _ , j. math . pures appl . ( liouville ) ( 2 ) , vi ( 1861 ) , 241323 . c. merino : _ the chip firing game and matroid complexes _ , discrete models : combinatorics , computation , and geometry ( 2001 ) , discrete math . proc . , aa , maison inform . math . ( mimd ) , paris ( 2001 ) , 245255 . c. okonek , m. schneider and h. spindler : `` vector bundles on complex projective spaces . '' corrected reprint of the 1988 edition . with an appendix by s. i. gelfand . modern birkhuser classics . birkhuser / springer basel ag , basel ( 2011 ) . j. watanabe : _ the dilworth number of artinian rings and finite posets with rank function _ , in : commutative algebra and combinatorics , advanced studies in pure math . 11 , kinokuniya co. north holland , amsterdam ( 1987 ) , 303312 .

a finite element - based modal formulation of diffraction of a plane wave by an absorbing photonic crystal slab of arbitrary geometry is developed for photovoltaic applications . the semi - analytic approach allows efficient and accurate calculation of the absorption of an array with a complex unit cell . this approach gives direct physical insight into the absorption mechanism in such structures , which can be used to enhance the absorption . the verification and validation of this approach is applied to a silicon nanowire array and the efficiency and accuracy of the method is demonstrated . the method is ideally suited to studying the manner in which spectral properties ( e.g. , absorption ) vary with the thickness of the array , and we demonstrate this with efficient calculations which can identify an optimal geometry .

1606.00390

photonic crystals , which consist of a periodically arranged lattice of dielectric scatterers , have attracted substantial research interest over the past decade @xcite . most commonly these structures are used to trap , guide , and otherwise manipulate pulses of light , and have led to a variety of important applications in modern nanophotonics , including extremely high - q electromagnetic cavities @xcite and slow - light propagation of electromagnetic pulses @xcite . however , a new and increasingly important application of photonic crystal structures is in the field of photovoltaics . it has long been known that structured materials can be used to achieve photovoltaic conversion efficiency beyond the yablonovitch limit @xcite , and indeed researchers have recently proposed photonic crystals @xcite and arrays of dielectric nanowires @xcite as inexpensive ways to create highly efficient absorbers . recent research in this area has been driven by advances in nanostructure fabrication @xcite concurrent with increased investment in renewable energy technology ( for the latest developments see @xcite ) . one aspect in which the use of photonic crystals in photovoltaics differs markedly from their employment in optical nanophotonics is the important role played by material absorption . for most nanophotonic applications absorption is an undesirable effect which must in general be minimized and can in many cases be neglected . this is in stark contrast to photovoltaics , in which the main aim is to exploit the properties of the structure to increase the overall absorption efficiency . modeling of absorbing photonic crystals has thus far been performed using direct numerical methods such as the finite - difference time - domain ( fdtd ) method @xcite , the finite element method ( fem ) @xcite or using the transfer matrix method @xcite . although these methods have produced valuable information about the absorption properties of such structures they do not allow us to gain direct physical insight into the mechanism of the absorption within them . furthermore , these calculations require substantial computational time and resources . here we present a rigorous modal formulation of the scattering and absorption of a plane wave by an array of absorbing nanowires , or , correspondingly , an absorbing photonic crystal slab ( fig . [ slab ] ) . the approach is a generalization of diffraction by capacitive grids formulated initially in refs @xcite for perfectly conducting cylinders . in contrast to conventional photonic crystal calculations the material absorption is taken into account rigorously , using measured values for the real and the imaginary parts of the refractive indices of the materials comprising the array . our formulation can be applied to array elements of arbitrary composition and cross - section . the semi - analytical nature of this modal approach results in a method which is quick , accurate , and gives extensive physical insight into the importance of the various absorption and scattering mechanisms involved in these structures . the method is based on an expansion in terms of the fundamental eigenmodes of the structure in each different region . within the photonic crystal slab the fields are expanded in terms of bloch modes , while the fields in free space above and below the slab are expanded in a basis of plane waves . these expansions are then matched using the continuity of the tangential components of the fields at the top and the bottom interfaces of the slab to compute fresnel interface reflection and transmission matrices . an important aspect of this approach is that the set of bloch modes must form a complete basis . this is not a trivial matter , as even for structures consisting of lossless materials the eigenvalue problem for the bloch modes is not formally hermitian . however , it is well known from the classical treatment of non - hermitian eigenvalue problems that a complete basis may be obtained by including the bloch modes of the adjoint problem . here , we use the fem to compute both the bloch modes and the adjoint bloch modes . because the mode computation may be carried out in two dimensions ( 2d ) like waveguide mode calculations , this routine is highly efficient . we note that a previous study @xcite , undertaken using a different method , failed to locate some of the modes , as is demonstrated in sec . [ section : dispersion ] ; we emphasis that the algorithm presented here is capable of generating a complete set of modes . in addition the fem allows us to compute bloch modes for arbitrary materials and cross sections . we note that this approach differs from earlier formulations @xcite in which the photonic crystal had to consist of an array of cylinders with a circular cross - section . with the modes identified , the transmission through , and reflection from , the slab can be computed using a generalization of fresnel reflection and transmission matrices , and the absorption is found using an energy conservation relation . the organization of the paper is as follows . the theoretical foundation of the method is given in sec . [ formulation ] while the numerical verification , validation and characterization of the absorption properties of a particular silicon nanowire array occurs in sec . [ applica ] . the details of the mode orthogonality , normalization , completeness as well as energy and reciprocity relations are given in the appendices . as mentioned in sec . [ intro ] , we separate the solution of the diffraction problem into three steps , one involving the consideration of the scattering of plane waves at the top interface and the introduction of the fresnel reflection and the transmission matrices for a top interface ( fig . [ slab ] ) . next we introduce the fresnel reflection and transmission matrices for the bottom interface by considering the reflection of the waveguide modes of the semi - infinite array of cylinders at the bottom interface . then the total reflection and transmission through the slab can be calculated using a fabry - perot style of analysis . the approach is based on the calculation of the bloch modes and adjoint bloch modes of an infinite array of cylinders . before doing so however , we first provide the field s plane wave expansions above and below the photonic crystal slab . in a uniform media such as free space , all components of the electromagnetic field of a plane wave must satisfy the helmholtz equation @xmath0 where @xmath1 is the free space wavenumber . here we consider the diffraction of a plane wave on a periodic square array of cylinders with finite length ( see fig . [ slab ] ) . in such a structure , the fields have a quasi - periodicity imposed by the incident plane wave field @xmath2 , where @xmath3 . that is , @xmath4 where @xmath5 and @xmath6 is a lattice vector , where @xmath7 and @xmath8 are integers . all plane waves of the form @xmath9 , must satisfy the bloch condition eq . ( [ qp ] ) and so @xmath10 . hence @xmath11 , where @xmath12 is an integer . it follows then that the coefficients @xmath13 and @xmath14 are discretized as follows @xmath15 and form the well known diffraction grating orders . we split the electromagnetic field into its transverse electric @xmath16 and the transverse magnetic @xmath17 components ( see , for example , ref . @xcite ) . for the transverse electric mode , the electric field is perpendicular to the plane of incidence , while for the transverse magnetic mode the magnetic field vector is perpendicular to the plane of incidence with the plane of incidence being defined by the @xmath18-axis and the plane wave propagation direction given by the vector @xmath19 . these te and tm resolutes are given by @xmath20 respectively , where @xmath21 . the te and tm plan wave modes are mutually orthogonal , and are normalized such that @xmath22 where the overline in @xmath23 denotes complex conjugation and the integration is over the unit cell . the general form of the plane wave expansions above and below the grating ( see fig . [ slab ] ) can then be written in terms of these @xmath16 and @xmath17 modes . following the nomenclature of ref . @xcite , the plane wave expansions take the forms @xmath24 \bm{r}^e_s(\bm{r } ) \nonumber\\ & + & \chi_s^{1/2 } \left [ f^{m-}_s e^{-i { \gamma}_s(z - z_0 ) } \right.\nonumber\\ & + & \left . f^{m+}_s e^{i { \gamma}_s(z - z_0 ) } \right ] \bm{r}^m_s(\bm{r } ) \label{epw}\\ \bm{e}_z\times \bm{h}_\perp(\bm{r } ) & = & \sum_{s } \chi_s^{1/2 } \left [ f^{e-}_s e^{-i { \gamma}_s(z - z_0 ) } \right.\nonumber\\ & - & \left . f^{e+}_s e^{i { \gamma}_s(z - z_0 ) } \right ] \bm{r}^e_s(\bm{r } ) \nonumber\\ & + & \chi_s^{-1/2 } \left [ f^{m-}_s e^{-i { \gamma}_s(z - z_0 ) } \right.\nonumber\\ & -&\left . f^{m+}_s e^{i { \gamma}_s(z - z_0 ) } \right ] \bm{r}^m_s(\bm{r } ) \label{kpw } \label{pwexpansion}\end{aligned}\ ] ] where @xmath25 and @xmath26 represent the amplitudes of transverse electric and magnetic component of the downward @xmath27 and upward @xmath28 propagating plane waves and @xmath29 where @xmath30 denotes a plane wave channel represented by the pair of integers @xmath31 . in the numerical implementation it is convenient to order the plane waves in descending order of @xmath32 . in eqs ( [ epw ] ) and ( [ kpw ] ) , @xmath33 is defined as @xmath34 and the factors @xmath35 are included to normalize the calculation of energy fluxes . the fem presented here is a general purpose numerical method which can handle the square , hexagonal or any other array geometry . the constitutive materials of the array can be dispersive and lossy . we first introduce the eigenvalue problem and then we present a variational formulation and the corresponding fem discretization . at a fixed frequency , the electric and magnetic fields of the electromagnetic modes satisfy maxwell s equations @xmath36 where @xmath37 and @xmath38 are the relative dielectric permittivity and magnetic permeability respectively . we express the time dependence in the form @xmath39 . the magnetic field @xmath40 has been rescaled as @xmath41 with @xmath42 , the impedance of free space . we now consider the electromagnetic modes of an array of cylinders of infinite length . the cylinder axes are aligned with the @xmath43axis . the dielectric permittivity and magnetic permeability of the array are invariant with respect to @xmath18 . from this translational invariance , we know that the bloch modes of the array have a @xmath43dependence @xmath44 and are quasi - periodic with respect to @xmath45 and @xmath46 . this reduces the problem of finding the modes to a unit cell @xmath47 in the @xmath48plane ( see fig . [ slab ] ) . as explained in section [ section : adjoint ] , this modal problem for the cylinder arrays is not hermitian and therefore the eigenmodes do not necessarily form an orthogonal set . however , by introducing the modes of the adjoint problem we can form a set of adjoint modes which have a biorthogonality property @xcite with respect to the primal eigenmodes . in order to introduce the adjoint problem , we first write maxwell s equations for conjugate material parameters @xmath49 and @xmath50 : @xmath51 the fields @xmath52 and @xmath53 have the same time dependence and quasi - periodicity as @xmath54 and @xmath40 . we define the adjoint modes as the conjugate fields @xmath55 and @xmath56 , and they satisfy maxwell s equations : @xmath57 therefore , the adjoint modes @xmath58 and @xmath59 satisfy the same wave equations as @xmath54 and @xmath40 , but have the opposite quasi - periodicity and @xmath60 time dependence . let @xmath47 denote a unit cell of the periodic lattice . within the array , a bloch mode is a nonzero solution of the vectorial wave equation @xmath61 which is quasi - periodic in the transverse plane with respect to the wave vector @xmath62 , and has exponential @xmath18 dependence , with the propagation constant @xmath63 , i.e. @xmath64 the longitudinal and transverse components of the electric field @xmath54 are respectively @xmath65 and @xmath66^t$ ] at the edges of the unit cell , the tangential component @xmath67 ( @xmath68 denotes a unit tangential vector to the unit cell boundary @xmath69 ) and the longitudinal component @xmath65 of a bloch mode satisfy the boundary conditions @xmath70 the quasi - periodicity of the components @xmath71 and @xmath72 of the magnetic field @xmath40 associated with @xmath54 is enforced as a natural boundary condition " of the fem . taking the exponential @xmath18 dependence into account by substituting ( [ mode1 ] ) into ( [ wave1 ] ) , one is led to the coupled partial differential equations @xmath73 where the operators @xmath74 and @xmath75 are the gradient and the divergence with respect to the transverse variables @xmath45 and @xmath46 ; the transverse curl operators are defined as @xmath76 { \mbox{\ensuremath $ -$ } } \frac{\partial f_z}{\partial x } \end{array } \right ] , \\ \nabla_{\perp } \times \bm{f}_{\perp } & = & \left ( \frac{\partial f_y}{\partial x } - \frac{\partial f_x}{\partial y } \right ) \bm{e}_z.\end{aligned}\ ] ] problem ( [ mode2 ] ) is a nonlinear eigenproblem with respect to the unknown @xmath63 since it involves both @xmath63 and @xmath77 . for @xmath78 , the substitution @xmath79 leads to a generalized eigenvalue problem involving only @xmath77 @xmath80 note that an eigenvalue @xmath81 of eq . ( [ mode3 ] ) corresponds to a pair of propagation constants @xmath82 and @xmath83 which are respectively associated with a _ upward propagating _ ( i.e. , towards @xmath84 ) wave , @xmath85 = [ \bm{e}_{\perp } , -i \ , \zeta^+ \hat{e}_z ] $ ] ( according to the scaling ( [ ez : scaled ] ) ) and a _ downward propagating _ wave @xmath86 = [ \bm{e}_{\perp } , -i \ , \zeta^- \hat{e}_z ] $ ] . as shown in ref . @xcite , except for a countable set of frequencies , in general all eigenvalues of the problem ( [ mode3 ] ) are nonzero . we remark that , in some cases , the mathematical analysis of the eigenproblem can be simpler and more elegant if eq . ( [ mode3 ] ) is rewritten in the following form @xmath87 for instance , as is explained below ( see eqs . ( [ primal : mode:1])([operator : m ] ) ) , the differential operators on the left and right hand sides of eq . ( [ mode3b ] ) are hermitian , in the case of lossless gratings while , for lossy gratings , the adjoint of each operator is the complex conjugate of the operator . however , all nonzero fields of the form @xmath88 become eigenmodes ( although most are non - physical ) of eq . ( [ mode3b ] ) associated with the zero eigenvalue . in order to avoid the unnecessary calculations of these non - physical modes , we have used a numerical implementation based on eq . ( [ mode3 ] ) . we now explain our convention used in the modal classification . taking into account the exponential @xmath18 dependence @xmath89 , if @xmath90 and @xmath91 ( propagating mode ) the upward travelling wave @xmath92 corresponds to the positive square root of @xmath77 , i.e. , @xmath93 ; otherwise if @xmath94 or @xmath95 ( evanescent mode ) the upward travelling wave @xmath92 corresponds to the mode such that @xmath96 decreases as @xmath18 increases , i.e. , @xmath97 is the complex square root of @xmath77 such that @xmath98 . in order to obtain the variational formulation corresponding to the problem eqs . ( [ mode3 ] ) and ( [ boundary ] ) we introduce the following functional spaces @xmath99 then if we multiply the first and second equations in eq . ( [ mode3 ] ) respectively by the complex conjugate of the test functions @xmath100 and @xmath101 , we obtain the variational formulation of the problem after integration by parts : _ find @xmath102 and @xmath103 such that @xmath104 and @xmath105 _ @xmath106 \left ( \nabla_{\perp } f_z , \mu^{-1 } \bm{e}_{\perp } \right ) - \left ( \nabla_{\perp } f_z , \mu^{-1 } \nabla_{\perp } \hat{e}_z \right ) \\ \hspace*{2 cm } + \ , k^2 \left ( f_z , \varepsilon \hat{e}_z \right ) = 0 , \end{array } \right.\ ] ] where @xmath107 represents the @xmath108 inner product @xmath109 let @xmath110 and @xmath111 be two finite dimensional approximation spaces to the functional spaces @xmath112 and @xmath113 respectively . the discretized problem is obtained by substituting @xmath114 for @xmath115 in the formulation of problem ( [ formul1 ] ) . we introduce sets of basis functions , respectively for the spaces @xmath110 and @xmath111 , and map these onto eq . ( [ formul1 ] ) to derive the discretized problem in matrix form @xcite : @xmath116 \left [ \begin{array}{c } \bm{e}_{\perp , n } \\ \hat{\bm{e}}_{z , n } \end{array } \right ] = \zeta_n^2 \left [ \begin{array}{cc } \bm{m}_{tt } & \bm{k}_{zt}^h \\ \bm{0 } & \bm{0 } \end{array } \right ] \left [ \begin{array}{c } \bm{e}_{\perp , n } \\ \hat{\bm{e}}_{z , n } \end{array } \right],\ ] ] where the superscript @xmath117 denotes the hermitian transpose ( or conjugate transpose ) operator . let @xmath118 and @xmath119 be the chosen basis functions for the spaces @xmath111 and @xmath110 respectively . for @xmath120 , the elements of the matrices @xmath121 and @xmath122 are defined as @xmath123 and , for @xmath124 and @xmath125 , the matrix @xmath126 is defined as @xmath127 while , for @xmath128 , the coefficients of the matrix @xmath129 are @xmath130 the generalized eigenvalue problem eq . ( [ matrix1 ] ) can be solved efficiently using the eigensolver for sparse matrices arpack @xcite . once the array modes @xmath131 are computed , we express a field @xmath54 inside the array by the modal expansion @xmath132 where the index @xmath133 counts out all the upward and downward propagating modes . in contrast , a formulation of maxwell s equations which is based on @xmath134 fields leads to a nonlinear eigenvalue problem @xcite which becomes inefficient to solve as the number of eigenvalues to be computed increases . the arpack library can be used to compute a few selected eigenvalues of large sparse matrices , and a shift and invert spectral transformation can be applied so that the numerical solutions converge to eigenvalues located within a desired region of the spectrum . here , by considering the plane wave dispersion relation eq . ( [ eq1:pw : dispersion ] ) , the target region includes the propagation constant @xmath63 near a reference value @xmath135 where the reference index @xmath136 is selected to be slightly higher than the largest real part of the slab refractive indices . we have chosen @xmath110 as the space of two - dimensional vector fields whose components are piecewise polynomials of degree @xmath137 while @xmath111 consists of piecewise continuous polynomials of degree @xmath138 ( in this paper @xmath139 ) . the vector fields of @xmath110 must be tangentially continuous across the inter - element edges of the finite element triangulation while their normal component is allowed to be discontinuous ( _ edge element _ ) . we now present some general principles which have guided our choice for the spaces @xmath110 and @xmath111 . it is known that , for a simply connected domain @xmath47 , the gradient , curl and div operators form an _ exact sequence _ : @xmath140 i.e. , the range of each operator coincides with the kernel of the following one . the derivation of this statement is based on the poincar lemma and for more details see ref . @xcite , for instance . for the sake of numerical stability @xcite , fem approximation spaces must be chosen such that the exact sequence is reproduced at the discrete level . in the context of waveguide mode theory , a scalar function takes the form @xmath141 and its gradient is @xmath142 \ , e^{i \ , \zeta \ , z}$ ] . if @xmath143 is a piecewise continuous polynomial of degree @xmath138 , i.e. , @xmath144 , then the two components of @xmath145 are piecewise polynomials of degree @xmath137 , i.e. , @xmath146 . we can then verify that , with our choice of @xmath110 and @xmath111 , the first exactness relation of eq . ( [ eq1:diagram ] ) , i.e. , @xmath147 , is also reproduced at the discrete level . we do not discuss the second exactness relation since , here , we do not have to build an approximation space for @xmath148 . we recall that in this paper we set @xmath139 , and approximate the transverse and longitudinal components using polynomials of degrees 2 and 3 respectively ; the construction of the basis functions for the fem spaces is described in ref . @xcite . modal orthogonality relations , or more correctly biorthogonality relations , in the case of the problem considered here , are important in determining the field expansion coefficients @xmath149 in eq . ( [ array : mode : expansion ] ) . although we may recast eq . ( [ matrix1 ] ) in a form in which each matrix is hermitian ( for the lossless case ) : @xmath150 \left [ \begin{array}{c } \bm{e}_{\perp , n } \\ \hat{\bm{e}}_{z , n } \end{array } \right ] = \zeta_n^2 \left [ \begin{array}{cc } \bm{m}_{tt } & \bm{k}_{zt}^h \\ \bm{k}_{zt } & \bm{k}_{zz } \end{array } \right ] \left [ \begin{array}{c } \bm{e}_{\perp , n } \\ n } \end{array } \right],\ ] ] this generalized eigenvalue problem @xmath151 is not hermitian in general , since for two hermitian matrices , @xmath152 and @xmath153 , the corresponding eigenproblem , @xmath154 is not necessarily hermitian since the product of two hermitian matrices is not , in general , hermitian . accordingly , the eigenmodes @xmath131 do not necessarily form an orthogonal set . it is also clear that the eigenproblem is not hermitian in the presence of loss . therefore we introduce the adjoint problem that we solve to obtain a set of adjoint modes which have a biorthogonality property @xcite with respect to the eigenmodes @xmath131 . in order to define the adjoint operators , we first write ( [ mode3b ] ) , the alternate form of eq . ( [ mode3 ] ) , as @xmath155 where @xmath156 and @xmath157 are the differential operators defined by @xmath158 \ ! , \\ \label{operator : m } \mathcal{m } \ , \bm{e } \ ! = \!\ ! \left [ \begin{array}{cc } - \mu^{-1 } \bm{e}_{\perp } & \mu^{-1 } \nabla_{\perp } \hat{e}_z \\ \ ! - \nabla_{\perp } \ ! \cdot \ ! ( \mu^{-1 } \bm{e}_{\perp } \ ! ) & \nabla_{\perp } \ ! ( \mu^{-1 } \nabla_{\perp } \hat{e}_z ) \ ! + \ ! k^2 \varepsilon \hat{e}_z \end{array } \ ! \right ] \ ! .\end{aligned}\ ] ] the adjoint operators @xmath159 and @xmath160 , with respect to the inner product eq . ( [ eq : inner : product ] ) are @xmath161 \ ! , \\ \mathcal{m}^{\dagger } \bm{f } \ ! = \!\ ! \left [ \begin{array}{cc } - \overline{\mu}^{-1 } \bm{f}_{\perp } & \overline{\mu}^{-1 } \nabla_{\perp } \hat{f}_z \\ \ ! - \nabla_{\perp } \ ! \cdot \ ! ( \overline{\mu}^{-1 } \bm{f}_{\perp } ) & \nabla_{\perp } \ ! ( \overline{\mu}^{-1 } \nabla_{\perp } \hat{f}_z ) \ ! + \ ! k^2 \overline{\varepsilon } \hat{f}_z \end{array } \ ! \right ] \!\ ! , \end{aligned}\ ] ] and follow from the definitions @xmath162 although for lossless media , we then have @xmath163 and @xmath164 , as in eq . ( [ matrix : sym2 ] ) , eigenproblem eq . ( [ primal : mode:1 ] ) is not hermitian and , as we see in section [ applica ] , complex values of @xmath165 can occur even for a lossless photonic crystal . the modes @xmath166 are the eigenmodes of the problem @xmath167 which satisfy the same quasi - periodic boundary conditions as @xmath131 . we now conjugate the boundary value problem eq . ( [ adjoint : mode:1 ] ) and take into account the fact that @xmath168 and @xmath169 . consistent with eqs ( [ eq2:a1 ] ) and ( [ eq2:a2 ] ) we redefine the adjoint mode as the eigenmode @xmath170 which satisfies the same partial differential equation as @xmath131 , i.e. , @xmath171 but with the quasi - periodic boundary conditions associated with the adjoint wavevector @xmath172 . this is convenient for the fem programming since the same subprograms can be used to handle the partial differential equations ( [ primal : mode:1 ] ) and ( [ adjoint : mode:2 ] ) while only a few lines of code are needed to manage the sign change for @xmath62 . the spectral theory for non - self - adjoint operators is difficult and in general less developed . in this paper we assume that the modes @xmath173 form a complete set and that the adjoint modes @xmath174 can be numbered such that @xmath175 , and the following biorthogonality relationship is satisfied @xmath176 in which @xmath177 $ ] can be obtained from the electric field @xmath178 $ ] using the relation @xmath179 which is derived directly from maxwell s equations ( [ a2 ] ) . a similar spectral property has been proven for a class of non - self - adjoint sturm - liouville problems ( see for instance , theorem 5.3 of ref . it is not clear if such a theorem can be extended to the vectorial waveguide mode problem , although it has been shown that the spectral theory of compact operators can be applied to a waveguide problem @xcite . however , our numerical calculations have generated modes which satisfy the biorthogonality relation and verify the completeness relations eqs ( [ mc5 ] ) and ( [ ip10 ] ) , in appendix [ mc ] , to generally within @xmath180 , and with even a better convergence when the number of plane wave orders and array modes , used in the truncated expansion , increase . we now establish the biorthogonality relation eq . ( [ eq1:biorthog ] ) using the operator definitions eqs . ( [ def : adjoint : l ] ) and ( [ def : adjoint : m ] ) of the adjoint , and eqs ( [ primal : mode:1 ] ) and ( [ adjoint : mode:2 ] ) . this relation may also be established directly from maxwell s equations , as is shown in appendix [ om ] . we begin with @xmath181 and , since @xmath131 and @xmath170 are eigenfunctions , we obtain @xmath182 now , by taking into account eq . ( [ def : adjoint : m ] ) , we can derive the following biorthogonality property : @xmath183 i.e. , @xmath184 the integrand of the field product in eq . ( [ eq2:biorthog ] ) can be expressed in term of the fields @xmath185 and @xmath186 as in eq . ( [ eq1:biorthog ] ) by noting @xmath187 since @xmath188 , ( [ primal : mode:1 ] ) and ( [ operator : l ] ) . the biorthogonality relation eq . ( [ eq2:biorthog ] ) is useful for the fem implementation of the field product since the product of two vectors @xmath189 and @xmath190 takes the form @xmath191 where @xmath192 is the discrete version of the operator @xmath157 and is the matrix on the right hand side of eq . ( [ matrix : sym2 ] ) . the calculation of the modes @xmath131 and the adjoint modes @xmath174 using this fem approach is highly efficient and numerically stable . in the following section we use modes of the structure for the field expansion inside the photonic crystal slab , and exploit the adjoint modes , which are biorthogonal to the primal modes , in the solution of the field matching problem in a least square sense . in this section we introduce the fresnel reflection and transmission matrices for photonic crystal - air interfaces and calculate the total transmission , reflection and absorption of a photonic crystal slab . first we introduce the fresnel reflection @xmath193 and transmission @xmath194 matrices for an interface between free space and the semi - infinite array of cylinders . we specify an incident plane wave field @xmath195 ( see eqs([epw])([kpw ] ) ) propagating from above onto a semi - infinite slab , giving rise to an upward reflected plane wave field @xmath196 and a downward propagating field of modes @xmath197 in the slab . the field matching equations between the plane wave expansions eqs ( [ epw ] ) and ( [ kpw ] ) and the array mode expansion eq . ( [ array : mode : expansion ] ) are obtained by enforcing the continuity of the tangential components of transverse fields on either side of the interface : @xmath198 equations ( [ eq12a ] ) and ( [ eq12b ] ) correspond to the continuity condition of the tangential electric field and magnetic field respectively . here @xmath199 denotes the downward tangential electric field component of mode @xmath133 while @xmath200 denotes the downward tangential magnetic field of mode @xmath133 , which satisfy the orthonormality relation @xmath201 in appendix [ mc ] we derive completeness relations for both the bloch modes basis and the plane wave basis . we now proceed to solve these equations in a least squares sense , using the galerkin - rayleigh - ritz method in which the two sets of equations are respectively projected on the two sets of basis functions . in this treatment , we project the electric field equation onto the plane wave basis and the magnetic field equation onto the slab mode basis to derive , in matrix form , @xmath202 where @xmath203,\nonumber\\ \qquad j^{e / m}_{sm } & = & \iint \overline{\bm{r}}^{e / m}_s \cdot \bm{e}_{m\perp } \,\ , ds , \\ \bm{j}^\dag & = & \left ( \begin{matrix } { \bm{j}}^{\dag e } \,\ , \bm{j}^{\dag m } \end{matrix } \right ) , \qquad \bm{j}^{\dag e / m } = \left [ j^{\dag e / m}_{ms } \right],\nonumber\\ \qquad j^{\dag e / m}_{ms } & = & \iint { \bm{r}}^{e / m}_s \cdot \bm{e}^{\dag}_{m\perp } \,\ , ds , \label{jmat } \\ \,\,\bm{\chi } & = & \mathrm{diag~ } \ { \chi_s \}.\end{aligned}\ ] ] then , defining the scattering matrices @xmath193 and @xmath194 according to the definitions @xmath204 we may solve eqs ( [ eqe12 ] ) and ( [ eqk12 ] ) to derive @xmath205 @xmath206 for a structure with inversion symmetric inclusions the following useful relations hold : @xmath207 , @xmath208 and @xmath209 . while the fresnel matrices @xmath193 and @xmath194 given in eqs ( [ eqr12 ] ) and ( [ eqt12 ] ) have been derived by presuming a plane wave field incident from above , identical forms are derived if we presume incidence from below , a consequence of the given symmetries of the modes . we now derive the slab - free space fresnel coefficients , assuming that we have a modal field @xmath197 incident from above and giving rise to a reflected modal field @xmath210 and a transmitted plane wave field @xmath211 below . this time , the field matching equations are @xmath212 and we again project the electric field equation onto the plane wave basis and the magnetic field equation on to the modal basis for the slab . accordingly , we derive @xmath213 then , defining the scattering matrices @xmath214 and @xmath215 according to @xmath216 and @xmath217 , we form @xmath218 these fresnel scattering matrices @xmath193 , @xmath194 , @xmath214 , @xmath215 can now be readily utilized to calculate the total transmission , reflection and the absorption of the slab . note that for inversion symmetric inclusions the following reciprocity relations hold : @xmath219 and @xmath220 . these relations hold independently of the truncation of the field expansions . in addition , in general , the following relations are also true : @xmath221 and @xmath222 . following the diagram in fig . [ fig : diagram ] , we may use the fresnel interface matrices to calculate the transmission and the reflection matrices for the entire structure from the following equations : @xmath223 where @xmath224 $ ] is the diagonal matrix which describes the propagation of the @xmath225 bloch mode inside the slab with a thickness @xmath226 . the transmission and reflection matrices defined as @xmath227 , @xmath228 can be deduced from the system eqs . ( [ sf1])([sf4 ] ) as @xmath229 the amplitudes of the transmitted and reflected fields are then given by , @xmath230 , @xmath231 , where @xmath232^t$ ] is the vector containing the magnitudes of components of the incident plane wave in the specular diffraction order , and @xmath233 is the angle between the electric vector and the plane of incidence . the absorptance , @xmath234 , is calculated by energy conservation as @xmath235,\end{aligned}\ ] ] where @xmath236 , @xmath237 are the diffraction order components of @xmath238 , @xmath239 and @xmath240 is the set of all propagating orders in free space . in the absence of absorption the following relations for the slab reflection @xmath241 and transmission @xmath242 matrices can be deduced ( for details see appendix c ) . given that the photonic crystal slab is up / down symmetric , the slab transmission @xmath243 and the reflection @xmath244 matrices for plane wave incidence from below are the same as for incidence from above : @xmath245 and @xmath246 . therefore the energy conservation relations take the form @xmath247 with @xmath248 denoting a diagonal matrix with the entries @xmath249 for the propagating plane wave channels and zeros for the evanescent plane wave channels , and @xmath250 being a diagonal matrix which has entries @xmath251 in the evanescent @xmath16 channels and @xmath252 in the evanescent @xmath17 channels , and zeros for the propagating channels . the semi - analytic expressions for the transmission eq . ( [ tr ] ) and the reflection eq . ( [ trr ] ) matrices for the slab can give important insight to improve the overall absorption efficiency . for instance , in eq . ( [ tr ] ) the matrices @xmath194 and @xmath215 represent the coupling matrices for a plane wave into and out of the slab , while the scattering matrix @xmath253 describes fabry - perot - like multiple scattering . as demonstrated in ref . @xcite , absorption is enhanced if first , there is a strong coupling ( @xmath194 ) , strong scattering amplitudes eq . ( [ fpm ] ) that increase the effective path in the slab multiple times , and the field strength is concentrated in the region of high absorption . in this section , we first use our mode solver to compute the dispersion curves of an array of lossless cylinders . then we apply our modal approach to analyze the absorption spectrum of an array of lossy cylinders ( silicon ) and we also examine the convergence of the method with respect to the truncation parameters . finally , we consider the example of a photonic crystal slab which exhibits fano resonances . though our method can be applied to inclusions of any cross section , we first consider an array of lossless circular cylinders , with dielectric constant @xmath254 ( alumina ) , and normalized radius @xmath255 , in an air background ( refractive index @xmath256 ) . we compute the propagation constant @xmath63 of the bloch modes defined in eq . ( [ mode1 ] ) using the vectorial fem . figure [ fig : dispersion:1 ] shows the dispersion curves corresponding to a periodic boundary condition in the transverse plane , i.e. , @xmath257 . the solid red curves indicate values of the propagation constant @xmath63 such that @xmath77 is real with positive values of @xmath77 corresponding to propagating modes , while negative values indicate evanescent modes . the dispersion curve for the fundamental propagating mode is at the lower right corner and starts at the coordinate origin . complex values of @xmath77 can also occur , even for a lossless system , and even with @xmath258 ; these modes , which occur in conjugate pairs , are shown by the dashed blue curves and are distinguished by a horizontal axis which is labeled @xmath259 . we can observe that the dashed blue curves connect a maximum point of a solid red curve to a minimum point of another solid red curve . this property of the dispersion of cylinders arrays was observed by blad and sudb @xcite . the dispersion curves in fig . [ fig : dispersion:1 ] are plotted using the same parameters as in fig . 4 of ref . all curves shown in fig . 4 of @xcite also appear in fig . [ fig : dispersion:1 ] of the present paper although our figure reveals many additional curves , for instance , near ( @xmath260 . in @xcite , the dispersion curves were plotted using a continuation method whose starting points are on the axis @xmath261 ; most of the curves which do not intersect this axis are missing in fig . 4 of @xcite but their solutions are required for the completeness of the modal expansion eqs ( [ eq12a ] ) and ( [ eq12b ] ) . for instance , if the eigenvalues @xmath262 are numbered in decreasing order of @xmath263 , then the index number of the eigenvalues , which appear near ( @xmath264 in fig . [ fig : dispersion:1 ] , are between 10 and 20 and , as explained in the convergence study of the next section , we typically need well over 20 modes to obtain good convergence . we now consider a silicon nanowire ( sinw ) array consisting of absorptive nanowires of radius @xmath265 , arranged in square lattice of lattice constant @xmath266 . this constitutes a dilute sinw array since the silicon fill fraction is approximately 3.1% . the dilute nature of the array can facilitate the identification of the modes which play a key role in the absorption mechanism @xcite . the height of the nanowires is @xmath267 . for silicon we use the complex refractive index of green and keevers @xcite . figure [ fig : dilute : sinw:1 ] shows the absorptance spectrum of the dilute sinw array , together with the absorptance of a homogeneous slab of equivalent thickness and of a homogeneous slab of equivalent volume of silicon . the absorption feature between 600 and 700 nm is absent in bulk silicon and is entirely due to the nanowire geometry . using our method we have identified some specific bloch modes which play a key role in this absorption behavior @xcite . at shorter wavelengths the absorption of the silicon is high and therefore the absorption of the slab does not depend on the slab thickness ( see fig.[fig : dilute : sinw:1 ] thin blue curve and thick red curve ) . note that that the geometry of the inclusion does not need to be circular since our fem based method can handle arbitrary inclusion shapes . indeed , in fig . [ fig : dilute : sinw:2 ] the absorptance spectrum of a sinw array consisting of square cylinders is analyzed and compared to the absorptance for circular cylinders of same period and cross sectional area . at long wavelengths , the absorption for the two geometries is the same , while at shorter wavelengths the absorption is slightly higher for the square cylinders . this can be explained by the field concentration at the corners of the square cylinders @xcite . the contour plot in fig . [ fig : dilute : sinw:3 ] shows the absorptance versus wavelength and the cylinder height @xmath226 for the circular sinw array . note that the nanowire height of @xmath267 used in fig . [ fig : dilute : sinw:1 ] is , indeed , in a region of high absorptance for the wavelength band @xmath268 $ ] . note that the propagation matrix @xmath269 is the only matrix in the expressions ( [ tr ] ) and ( [ trr ] ) for the reflection and transmission matrices which depends on the thickness @xmath226 and it can be easily updated when @xmath226 is varied for a fixed value of the wavelength . thus , the modal method is a very fast technique in exploring the dependence of the absorptance with respect to the thickness . indeed , the contour plot is obtained by computing the absorptance for 6001 height values uniformly spaced over the interval @xmath270 $ ] , for 409 wavelength values uniformly distributed over the interval @xmath271 $ ] ; a high sampling resolution is required in order to capture the oscillatory features which occur in the region @xmath272 $ ] . it took about 44 hours to generate the full results using 16 cores of a high performance parallel computer with 256 cores ( it is a shared memory system consisting of 128 processors intel itanium 2 1.6ghz ( dual core ) ) . if the absorptance had to be computed independently for the @xmath273 data points , this would have required many months of computer time . we have studied the convergence with respect to the truncation parameters of the plane wave expansions and array mode expansions in eqs ( [ eq12a ] ) and ( [ eq12b ] ) . the array modes are ordered in decreasing order with respect to @xmath274 . we have used a _ circular truncation _ for the plane wave truncation number @xmath275 , i.e. , for a given value of @xmath275 , only the plane wave orders @xmath276 such that @xmath277 are used in the truncated expansions ; this choice is motivated by the fact that , for normal incidence , it is consistent with the ordering of the array modes since the plane wave propagation constants are given by the dispersion relation @xmath278 ( see eq . ( [ eq1:pw : dispersion ] ) ) ; the propagation constants @xmath279 are also numbered in decreasing value of @xmath280 . figures [ fig : dilute : sinw:4 ] and [ fig : dilute : sinw:5 ] illustrate the convergence when the number of plane wave orders and the number of array modes used in modal expansions are increased . the wavelength is set to @xmath281 and the corresponding silicon refractive index is @xmath282 , which is taken from ref . the error is estimated by assuming that the result obtained with the highest discretization is exact " ( @xmath283 for @xmath284 and @xmath285 ) . the calculations are based on a highly refined fem mesh consisting of 8088 triangles and 16361 nodes . in fig . [ fig : dilute : sinw:5 ] , there is a sudden jump in error when @xmath286 ( isolated blue dot ) ; this is due to the chosen truncation cutting through a pair of degenerate eigenvalue @xmath287 i.e. , including one member of the pair but excluding the other ) . indeed the solution is well - behaved when @xmath288 . similar behavior has been observed when a pair of conjugate eigenvalues is cut . thus all members of a family of eigenvalues must be included together in the modal expansion , otherwise there is degradation of the convergence , which is particularly strong when it occurs for a low - order eigenvalue which makes a significant contribution to the modal expansions . in practice we expect the computed absorptance to have about three digits of accuracy when the truncation parameters of the plane wave expansions and array mode expansions are set respectively to @xmath289 ( giving 29 plane wave orders and @xmath290 basis functions for te and tm polarizations ) and @xmath291 , assuming adequate resolution of the fem mesh . the absorptance curve for the dilute sinw array in fig . [ fig : dilute : sinw:1 ] is obtained using these truncation parameters and an fem mesh which has 1982 triangles and 4061 nodes . figure [ fig : dilute : sinw:6 ] presents the absorptance spectrum for off - normal incidence ( @xmath292 ) . the absorptance is sensitive to the angle of incidence and the light polarization . compared with normal incidence , the absorptance peak in the wavelength band @xmath293 $ ] ( low silicon absorption ) has shifted to shorter wavelengths while the peak near @xmath294 ( high silicon absorption ) has shifted to longer wavelengths . fano resonances are well known from the field of particle physics @xcite , and they are observable also in photonic crystals @xcite . they are notable for their sharp spectral features and so serve as a good benchmark for the accuracy of new numerical methods . we have carried out a calculation of fano resonances using our modal formulation . we present here an example that was first studied by fan and joannopoulos @xcite . the photonic crystal slab consists of a square array of air holes in a background material of relative permittivity @xmath295 . figure [ slabt ] shows the transmittance of a photonic crystal slab as a function of the normalized frequency @xmath296 for a plane wave at normal incidence . the parameters of the slab considered in fig . [ slabt ] are identical to those in fig . 12(a ) of ref . @xcite , and the curves from the two figures are the same to visual accuracy . this is an additional validation of the approach presented here . the transmittance curve reveals a strong transmission resonance which is typical of asymmetric fano resonances . these resonances are very sharp and can be used for switching purposes @xcite . we have developed a rigorous modal formulation for the diffraction of plane waves by absorbing photonic crystal slabs . this approach combines the strongest aspects of two methods : the computation of bloch modes is handled numerically , using finite elements in a two - dimensional context where this method excels , while the reflection and transmission of the fields through the slab interfaces are handled semi - analytically , using a generalization of the theory of thin films . this approach can lead to results achieved using a fraction of the computational resources of conventional algorithms : an example of this is given in fig . [ fig : dilute : sinw:3 ] , in which a large number of different computations for different slab thicknesses were able to be computed in a very short time . although the speed - testing of this method against conventional methods ( fdtd and 3d finite element packages such as comsol ) is a subject for future work , we have demonstrated here the method s accuracy and its rapidity of convergence . the method also satisfies all internal checks related to reciprocity and conservation of energy , as well as reproducing known results from the literature in challenging situations , such as the simulation of fano resonances ( fig . [ slabt ] ) . the method is very general with respect to the geometry of the structure . we have demonstrated this by modeling both square and circular shaped inclusions . in addition , because the method is at a fixed frequency , it can handle both dissipative and dispersive structures in a straightforward manner , using tabulated values of the real and imaginary parts of the refractive index . this is in contrast to time - domain methods such as fdtd , or to some formulations of the finite element approach . though we have used a single array type here ( the square array ) , other types of structure ( such as hexagonal arrays ) can be dealt with by appropriately adjusting the unit cell @xmath47 , together with the allowed range of bloch modes . it is also easy to see how this method could be extended to multiple slabs containing different geometries , as well as taking into account the effect of one or more substrates ; this extension would involve the inclusion of field expansions for each layer , together with appropriate fresnel matrices , in the equation system ( [ sf1])-([sf4 ] ) . in principle this approach could then be used to study rods ( or holes ) whose radius or refractive index changed continuously with depth , provided the spacing between the array cells remained constant . our method has an important advantage over purely numerical algorithms in that it gives physical insight into the mechanisms of transmission and absorption in slabs of lossy periodic media . the explanation of the absorption spectrum in arrays of silicon nanorods is vital for the enhancement of efficiency of solar cells , however this spectrum is complicated , with a number of processes , including coupling of light into the structure , fabry - perot effects , and the overlap of the light with the absorbing material , all playing an important role . by expanding in the natural eigenmodes of each layer of the structure , it is possible to isolate these different effects and to identify criteria that the structure must satisfy in order to efficiently absorb light over a specific wavelength range . this , we have discussed in a recent related publication @xcite . this research was conducted by the australian research council centre of excellence for ultrahigh bandwidth devices for optical systems ( project number ce110001018 ) . we gratefully acknowledge the generous allocations of computing time from the national computational infrastructure ( nci ) and from intersect australia . here we prove the biorthogonality of the modes and adjoint modes . let us consider set of modes @xmath297 and adjoint modes @xmath298 of photonic crystal . these modes satisfy @xmath299 and @xmath300 we multiply both sides of ( [ me11 ] ) by @xmath301 and ( [ me21 ] ) by @xmath131 and correspondingly we multiply each set of ( [ me12 ] ) by @xmath302 and ( [ me22 ] ) by @xmath303 then adding these we deduce @xmath304 next we separate the transverse @xmath305 and longitudinal @xmath306 ( along the cylinder axes ) components according to @xmath307 after the substitution of ( [ nab ] ) into ( [ maine ] ) we obtain @xmath308 \nonumber \\ = \nabla_\perp \cdot \left [ \bm{e}_{n \perp } \times \bm{h}^\dag_{m \parallel } \right . + \left . \bm{e}_{n \parallel } \times \bm{h}^\dag_{m \perp } \right . \\ + \left . \bm{e}^\dag_{m \perp } \times \bm{h}_{n \parallel } \right . \bm{e}^\dag_{m \parallel } \times \bm{h}_{n\perp } \right ] . \label{efo}\end{aligned}\ ] ] taking into account the @xmath18-dependence on the modes given by the factors @xmath309 , @xmath310 and integrating ( [ efo ] ) over the unit cell we derive @xmath311 since the integral on the left hand side vanishes due to quasi - periodicity . we finally obtain @xmath312 which holds for arbitrary modes such that @xmath313 . the same relation holds for the counter propagating mode @xmath12 @xmath314 the minus sign in the superscript position indicates the direction of the propagation . taking into account the relations @xmath315 and @xmath316 we can rewrite ( [ efpm ] ) in the form @xmath317 after subtraction of relation ( [ efpmm ] ) from ( [ efpm ] ) the orthogonality relation takes form @xmath318 which states that two distinct modes propagating in the same direction are orthogonal . it is then clear that these modes can always be normalized such that @xmath319 from the field expansions we can derive the condition of the modal completeness . the plane waves can be expanded in the following forms : @xmath320 by projecting eq . ( [ mc1 ] ) on the modes @xmath321 and using the biorthogonality relations eq . ( [ orto4 ] ) we deduce ( see eq . ( [ jmat ] ) ) @xmath322 similarly we project ( [ mc2 ] ) onto the adjoint magnetic mode @xmath323 and deduce @xmath324 thus , @xmath325 next we project ( [ mc2 ] ) onto plane wave basis by multiplying both sides of ( [ mc2 ] ) by @xmath326 and integrating over the unit cell . we obtain @xmath327 the equation ( [ mc4 ] ) represents the completeness relation for the modes . if we introduce the vectors of matrices @xmath328 \text { and } \bm{k } = \left [ \begin{array}{l } \bm{k}^e \\ \bm{k}^m \end{array } \right],\end{aligned}\ ] ] where @xmath329 \text { and } \bm{k}^{e / m } = \left [ k_{s n}^{e / m } \right]\end{aligned}\ ] ] then the completeness relation eq . ( [ mc4 ] ) can be written in the matrix form @xmath330 where @xmath331 is the identity matrix . the completeness relation of the rayleigh modes can be established in a similar way . the transverse component of electric and magnetic modal fields can be represented as a series in terms of rayleigh modes in the region above the grid as @xmath332 by multiplying ( [ ip01 ] ) on ( [ ip02 ] ) and integrating we deduce @xmath333 the completeness relation eq . ( [ ip09 ] ) can be written in matrix form @xmath334 here we briefly outline the derivation of energy conservation relations for the situation when there is no absorption . the flux conservation leads to the certain relations between the fresnel interface reflection and transmission matrices . for some value @xmath18 we can write an expansion @xmath335 and similarly @xmath336 the downward flux is defined by @xmath337 = \hspace{2 cm } \label{aie3 } \\ \frac{1}{2}\left [ ( \bm{c}^- - \bm{c}^+)^h\bm{u}(\bm{c}^- + c^+)+(\bm{c}^-+\bm{c}^+)^h\bm{u}^h(\bm{c}^- - c^+)\right ] , \nonumber\end{aligned}\ ] ] where the matrix @xmath338 is given by @xmath339 the relation ( [ aie3 ] ) can be written in the form @xmath340\bm{v}\left [ \begin{array}{c } \bm{c}^- \\ \bm{c}^+ \\ \end{array } \right]\right\ } , \label{aie5}\end{aligned}\ ] ] @xmath357 the slab energy conservation relations can be found in the similar way as for the interface relations as above . furthermore the expressions of the energy relations are very similar to the interface relations . the only difference is now the transmission @xmath194 and the reflection @xmath193 matrices in ( [ a310 ] ) need to be replaced by the slab reflection and transmission matrices . the adjoint modes are defined by @xmath358 with anti - quasi - periodicity condition @xmath359 when there is no absorption @xmath360 given @xmath361 and @xmath362 are then self adjoint operators . note that even though the operators @xmath157 and @xmath156 are self adjoint ( when there is no absorption ) the eigenvalue problem eq . ( [ modeadj ] ) is not hermitian because the operator @xmath363 is not self adjoint in general . therefore the eigenvalues can be real representing propagating modes , as well as complex ( pure imaginary or complex ) representing evanescent modes . now from eq . ( [ modeadj ] ) we deduce @xmath364 where overline means complex conjugation . by comparing eq . ( [ modeadj2 ] ) and the original eigenvalue equation we deduce @xmath365 for the real eigenvalues @xmath63 we choose @xmath366 while for complex @xmath63 we must choose @xmath367 which will ensure that the downward evanescent propagating field is decaying . therefore we have @xmath368 for the propagating field and @xmath369 for the evanescent field . from maxwell s equations we deduce that @xmath370 thus , the equations ( [ efoo ] ) and ( [ efpm ] ) can be rewritten as @xmath371 therefore when @xmath372 , by adding and subtracting the relations ( [ efooa ] ) and ( [ efpma ] ) we may deduce @xmath373 which are the @xmath374 elements of the matrix @xmath338 introduced earlier . from ( [ efpma ] ) and for @xmath375 and real @xmath262 we deduce that the integral @xmath376 is real . when @xmath377 is pure imaginary and @xmath378 then from ( [ efpma ] ) we deduce that @xmath379 is pure imaginary . now let us consider the case where @xmath262 is complex . the @xmath338 matrix is defined by ( [ aie4 ] ) . for modes @xmath380 and @xmath381 from ( [ aie4 ] ) and ( [ modeadj7 ] ) we deduce @xmath382 by using the orthonormal condition @xmath383 we find @xmath384 . for the modes @xmath385 and @xmath386 from ( [ aie4 ] ) we obtain @xmath387 so the elements @xmath388 . this means that @xmath389 is a diagonal matrix with unit elements on only the corresponding to propagating modal part . s. nishimura , n. abrams , b. a. lewis , l. i. halaoui , t. e. mallouk , k. d. benkstein , j. van de lagemaat , and a. j. frank , `` standing wave enhancement of red absorbance and photocurrent in dye - sensitized titanium dioxide photoelectrodes coupled to photonic crystals , '' j. am . soc . * 125 * , 63066310 ( 2003 ) . r. c. mcphedran , d. h. dawes , l. c. botten , and n. a. nicorovici , `` on - axis diffraction by perfectly conducting capacitive grids , '' journal of electromagnetic waves and applications * 10 * , 10851111(27 ) ( 1996 ) . l. c. botten , r. c. mcphedran , n. a. nicorovici , and a. b. movchan , `` off - axis diffraction by perfectly conducting capacitive grids : modal formulation and verification , '' journal of electromagnetic waves and applications * 12 * , 847882(36 ) ( 1998 ) . b. c. p. sturmberg , k. b. dossou , l. c. botten , a. a. asatryan , c. g. poulton , c. m. de sterke , and r. c. mcphedran , `` modal analysis of enhanced absorption in silicon nanowire arrays , '' opt . express * 19 * , a1067a1081 ( 2011 ) . the dashed blue curves represent solutions such that @xmath391 is not zero . the complex solutions @xmath77 occur as conjugate pairs and , in order to differentiate the pairs , we use the term @xmath392 for the x - axis instead of @xmath393 . , width=302 ] increases for a fixed wavelength @xmath281 and waveguide truncation number @xmath395 . the computed absorptance for @xmath284 and @xmath285 is @xmath283 and this value is considered as `` exact '' and used to compute the error curve.,width=321 ] increases . the details given in the caption of fig . [ fig : dilute : sinw:4 ] also apply here except that the plane wave truncation number is fixed at @xmath284 . note that @xmath287 is a degenerate eigenvalue . the isolated blue dot corresponds to the truncation value @xmath286 where the error is unexpectedly high ; thus the truncation @xmath288 is instead used for the error curve . , width=321 ] off - normal orientated along @xmath45-axis ( azimuthal angle = 0 ) . the solid red and dashed blue curves represent an incidence by te - polarized and tm - polarized plane wave respectively . the absorption spectrum for normal incidence ( fig . [ fig : dilute : sinw:1 ] ) is also shown ( dotted black ) . , width=321 ]

despite of the topical engineering need and all scientific investments , the mathematical formulation of modeling elastic deformations in magnetic systems is not yet fully established . often , especially in electrical engineering applications , a model assuming small ( infinitesimal ) strains seems sufficient . to express such small - strain magneto - elastic problems in a suitable form for discretization methods , we present here a formulation in the framework of differential geometry . the given analysis shows algebraic similarity between small - strain elasticity and magnetism . this suggests that a class of magnetic , elastic , and magneto - elastic problems may be modeled in the same algebraic category , constituting suitable domain for discretizations .

1602.04966

a class of electromechanical problems couple magnetism and elasticity . a magneto - elastic coupled model is relevant for the development of electric motor , sensor and actuator technologies , and much work have been done to make such modeling available to engineers @xcite . in modeling , it is commonplace to rely on a variational formulation of the coupled problem . variational formulation is also adopted in a recent paper by bossavit @xcite . therein , a small - strain magneto - elastic model is built on an analogy between magnetism and small - strain elasticity , with emphasis on finite element implementation using whitney elements . in the present paper we consider the formulation of small - strain magneto - elastic problems using analysis on manifolds . accordingly , we will be precise about the mathematical structures used in different parts of the analysis . this is how mathematics deals with the same thing which is called coordinate system invariances in physics . for example , defining maxwell s equations on differentiable manifold means the corresponding coordinate equations are invariant to general coordinate transformations . on the other hand , to express certain equations in a coordinate independent manner , we will introduce additional structure on the manifold . example of an equation which we express by using additional structure that of a connection is the mechanical equilibrium equation . in short , the manifold theory serves to drive information from coordinates to well defined mathematical structures . this makes possible a condensed presentation , allowing rapid access to the magneto - elasticity problem for mathematically informed reader . as a prerequisite we assume that the reader is familiar with differential forms , and the basic tools of differential geometry . in particular , one recognizes the exterior derivative as the counterpart for the gradient , curl and divergence of vector analysis . for example , the gauss s law and ampere s law in magnetostatics are naturally expressed by using the exterior derivative , whose definition requires only the differentiable structure . on the other hand , the magnetic constitutive law requires more structure , and is expressed in the variational formulation by using a magnetic energy density object . for algebraic similarity between elasticity and magnetism , we will introduce the reader to vector- and covector - valued differential forms together with their natural product @xcite . these notions require only the differentiable structure . in addition , we present some related tools of differential geometry . among these there is the covariant exterior derivative , whose definition requires the additional structure of a connection @xcite . similarly to magnetism the constitutive law between stress and strain will be expressed in the variational framework by using an energy density object . when it comes to magneto - elasticity , a variational formulation involves , in general , magneto - elastic energy density . such magneto - elastic energy density constitutes a way to express `` coupled constitutive laws '' , and it implies that magneto - elasticity can not be meaningfully separated into `` magnetic '' and `` elastic '' parts . such genuine magneto - elastic coupling is often called magnetostriction @xcite . here , we give two alternative mathematical constructs for magneto - elasticity , both of them allowing magnetostriction . the given analysis shows that certain problems of magnetism , small - strain elasticity , and small - strain magneto - elasticity may be modeled by using geometric constructs that are in many ways similar algebraically . this makes it relevant to search for _ algebraic category _ for the modeling of a class of such problems . we argue that the detailed specification of an appropriate category should be made such that ( i ) it is general enough to cover important problems , and ( ii ) its discretization to the category of finite dimensional vector spaces can be realized . an analysis such as the one given here is a necessary step in the way to specify an appropriate category . to position the formulation of this paper into a more general context , we begin by introducing material bodies , the underlying ( classical ) space - time , and the notion of observer . these notions are taken from ref . @xcite , and are reproduced here ( using slightly different notation ) to make the paper self - contained . moreover , we specify here the observers allowed by the given formulation . in classical space - time , the notion of simultaneity has absolute meaning , and simultaneous events may be identified with a reference manifold . by observation of space - time we mean the process of making such an identification for all time instants . independently of observation , it is assumed that time `` passes '' homogeneously and distances between simultaneous events can be determined . it is noted that to formalize this intuition , there is no need to invoke affine euclidean structure ( cf . @xcite ) . given 4-dimensional manifold @xmath0 , and 3-dimensional manifold @xmath1 , a diffeomorphism @xmath2 is called a _ slicing_. we define @xmath3 for each @xmath4 . the manifold @xmath0 is _ classical space - time _ if * there is surjective map @xmath5 , called universal time , with everywhere non - vanishing derivative , such that @xmath6 is diffeomorphic to @xmath1 for all @xmath4 * there is a slicing @xmath7 which is compatible with time , that is , @xmath8 for all @xmath4 * there is a symmetric connection on @xmath0 by which @xmath9 is covariant constant * on each @xmath10 there is a riemannian metric @xmath11 , whose levi - civita connection @xmath12 coincides with the restriction of the space - time connection to @xmath10 . a slicing @xmath7 of classical space - time is called an _ observer_. an observer that is compatible with time ( item ( ii ) ) is called _ newtonian observer_. newtonian observer @xmath7 perceives simultaneous events at time @xmath13 using @xmath14 . this may be used to carry the metric @xmath11 and connection @xmath12 to @xmath1 . accordingly , we define the _ metric observed by @xmath15 on @xmath1 at time @xmath13 _ as @xmath16 , and the _ connection observed by @xmath15 on @xmath1 at time @xmath13 _ as @xmath17 ( defined using the push - forward by @xmath18 ) . it follows that @xmath19 is the levi - civita connection of @xmath20 . if there is a fixed ( time - independent ) metric @xmath21 on @xmath1 , newtonian observer @xmath15 is called _ rigid _ if @xmath22 for all @xmath23 . to describe material body in space - time , we let @xmath24 be a reference manifold ( with @xmath25 ) , and define _ world tube for @xmath24 _ as a map @xmath26 such that the map @xmath27 is an embedding , and fulfills @xmath28 , for all @xmath4 . the path @xmath29 is the _ world line for @xmath30_. the `` absolute '' velocity , or _ four - velocity _ , of the world line for @xmath30 is the velocity vector field of the world line , and its `` absolute '' acceleration , or _ four - acceleration _ , is given by taking the covariant derivative of the four - velocity along the world line , using the space - time connection . going through all points @xmath31 gives the four - velocity and four - acceleration of the world tube @xmath32 . in particular , newtonian observer @xmath7 is a world tube for @xmath1 , and has a well - defined four - velocity and four - acceleration . given a world tube @xmath33 and a newtonian observer @xmath7 , the _ motion of @xmath24 on @xmath1 relative to @xmath15 _ is the map @xmath34 defined by @xmath35 . the _ ( `` apparent '' ) velocity of @xmath32 relative to @xmath15 _ is the velocity vector field of @xmath36 ( defined by sewing together the velocity vector fields of the paths @xmath37 corresponding to all @xmath31 . the _ ( `` apparent '' ) acceleration of @xmath32 relative to @xmath15 _ is given by taking the covariant derivative of the relative velocities of material points along the paths @xmath37 , using the ( possibly time dependent ) connection observed on @xmath1 by @xmath15 . using this setup , it is straightforward to derive , in particular , the transformation of relative velocities and accelerations under change of observer . these transformations are not given here , for we will restrict the analysis to statics , and allow only _ inertial _ observers , namely , those whose velocity field is covariant constant and whose acceleration therefore vanishes . the following analysis serves as our opening to the modeling of magneto - elastic coupled problems , and it will need to be extended to allow for dynamics and non - inertial observers . in particular , this would require considering the non - tensorial transformation of accelerations and forces under change of observer ( see @xcite ) . in addition to restricting the analysis to small - strain elastostatics , a restriction is made to hyper - elastic constitutive laws , meaning that elastic energy density will be assumed . ( it is noted that large amount of work have been made on the geometric analysis of the general elastic problem and continuum mechanics in general that is free of these restrictive assumptions , see for example @xcite . ) an analysis that restricts to small - strain magneto - hyper - elasticity is justified by the significance of this model in important applications , such as electric motor , sensor , and actuator technologies . let us assume that a rigid inertial observer @xmath7 is given , and denote as @xmath38 the observed ( time - independent ) riemannian metric on @xmath1 . further , let us denote as @xmath34 the motion of @xmath24 in @xmath1 relative to @xmath15 , and as @xmath39 a stress - free reference configuration . intuitively , we will be concerned with an infinitesimal displacement of material points from the reference configuration . to formulate the problem on @xmath24 , we consider _ displacement _ @xmath40 as a vector field on @xmath24 . this is appropriate because we assume both @xmath24 and @xmath1 to be 3-dimensional . ( more general modelings , where @xmath24 and @xmath1 may be of different dimension , can be found in continuum mechanics literature . ) this is a slightly atypical choice of a modeling quantity , as the spatial displacement observed by @xmath15 is now the vector field @xmath41 on @xmath42 . in strict formal language , our displacement @xmath40 is a smooth section of the tangent bundle @xmath43 of @xmath24 ( with the projection @xmath44 ) , that is , a smooth map @xmath45 satisfying @xmath46 ( identity on @xmath24 ) . finally , in preparing to give the formulation on @xmath24 , let us introduce on @xmath24 the metric @xmath47 and the connection @xmath48 . it follows that @xmath49 is the levi - civita connection of @xmath50 . further , the metric @xmath50 is euclidean if and only if @xmath38 is euclidean , because @xmath24 and @xmath1 are of the same dimension . in the following we will often work with the category of vector bundles and vector bundle mappings @xcite . given two vector bundles @xmath51 and @xmath52 over @xmath24 , we denote as @xmath53 the vector bundle whose fiber above point @xmath31 is the vector space @xmath54 of linear maps from the fiber @xmath55 to the fiber @xmath56 . ( for further details , see for example @xcite . ) the ( infinite dimensional ) vector space of ( appropriately smooth ) sections of a vector bundle @xmath51 will be denoted as @xmath57 . to express small - strain elasticity and magnetism in analogous form , we bring fore the notion of _ displacement gradient_. intuitively , the displacement gradient serves to describe how the values of displacement are changed to the first order in the neighborhoods of points . to avoid introducing new notation , we will denote the displacement gradient by the symbol @xmath58 , which is usually reserved for strain tensor ( the symmetric part of displacement gradient , see section [ ela_energy_section ] below ) . formally , the displacement gradient is a section of the vector bundle @xmath59 , that is , a vector - valued 1-form on @xmath24 . accordingly , its value at each point @xmath31 belongs to the vector space @xmath60 of linear maps on the tangent space @xmath61 . to formalize the above intuition about the relation of displacement gradient to displacement , we need to compare the displacement vectors at neighboring points of @xmath24 an operation for which we use the connection @xmath49 on @xmath24 . since @xmath49 takes two smooth vector fields @xmath62 and @xmath63 to a third smooth vector field @xmath64 , which depends linearly on @xmath63 , and is function - linearthat is , the mapping @xmath65 is a structure preserving map on the @xmath66-module of vector fields , where @xmath66 is the ring of smooth real - valued functions on @xmath24 . ] in @xmath62 , we may conceive it as the linear map @xmath67 . the relation of displacement gradient @xmath58 to displacement @xmath40 is given as @xmath68 the above relation is invariant to changes of observer , provided that the observed connection is used . by this we mean the following . the change of observer corresponds to a diffeomorphism @xmath69 . the metric and connection on @xmath24 that correspond to the new observer are @xmath70 and @xmath71 . denoting as @xmath72 and @xmath73 the transformed displacement gradient and displacement , that is , @xmath74 and @xmath75 , we have @xmath76 . we recall that when @xmath51 and @xmath52 are vector bundles , with projections @xmath77 and @xmath78 , a vector bundle map @xmath79 induces a unique map @xmath80 such that @xmath81 holds @xcite . it is said that @xmath82 is a vector bundle map over @xmath83 . in the following , we will make use of a correspondence between smooth sections of @xmath53 and vector bundle maps @xmath84 over the identity on @xmath24 . this correspondence is bijective , associating to a section @xmath85 of @xmath53 the vector bundle map @xmath86 over the identity on @xmath24 whose restriction to the fiber @xmath55 is defined by @xmath87 for all @xmath31 , and vice versa . in the following , when we say that a section @xmath85 of @xmath53 operates on a section @xmath88 of @xmath51 we mean the map @xmath89 , taking @xmath88 to a section of @xmath52 . forces inside materials operate on virtual displacement vector fields to produce 3-forms that can be integrated to yield contributions to virtual work . accordingly , we introduce _ body force _ as a section of @xmath90 , where @xmath91 is the bundle of 3-covectors on @xmath24 . we use `` tilde '' in the notation as a reminder that one should really deal with twisted forms @xcite . ( twisted forms are treated with varying accuracy in this paper . if @xmath24 is orientable , and a specific orientation is selected , this issue may be omitted , as twisted p - forms become represented by p - forms . ) with hindsight , we will denote the body force as @xmath92 . for a virtual displacement vector field @xmath93 the body force @xmath92 produces the 3-form @xmath94 , which may be integrated over @xmath24 ( independently of orientation , or even when @xmath24 is not orientable , using integration of twisted forms ) . in addition to body forces we allow surface forces on the boundary @xmath95 of @xmath24 . using the natural inclusion @xmath96 we may consider the pull - back bundle @xmath97 . this is the vector bundle whose base manifold is @xmath95 and fiber above @xmath98 is ( identified with ) the tangent space @xmath99 . we take _ surface force _ @xmath100 as a section of @xmath101 , where @xmath102 is the bundle of 2-covectors on @xmath95 . accordingly , surface force can operate on sections of @xmath103 to produce 2-forms on @xmath95 . finally , for a section @xmath93 of @xmath43 , we denote as @xmath104 the section of @xmath103 defined by @xmath105 . global force should give the virtual work done under virtual displacements of the body @xmath24 . accordingly , given body force @xmath92 and surface force @xmath100 , we consider _ force _ @xmath106 as the linear functional on @xmath107 defined using integration as @xmath108 for all @xmath109 . the surface force may not be given beforehand on some part of @xmath95 , and then it will be defined on this part using the notion of stress ( section [ equilibrium_eq ] ) . stress will be defined in this paper through a constitutive law ( section [ stress ] ) , although it could be introduced without taking sides to the constitutive law , see for example @xcite . the relation is invariant to changes of _ inertial _ observers . ( for the general transformation of forces , see @xcite . ) in the context of small - strain and hyper - elasticity assumptions , an elastic energy density is assumed , which depends on the displacement only through the symmetric part of the displacement gradient . this means that rigid infinitesimal displacements will not affect elastic energy . when the space manifold is euclidean , such displacements are translations , rotations , and their combinations . let @xmath51 and @xmath52 be vector bundles over @xmath24 and @xmath79 a vector bundle map , that is , a smooth fiber preserving map which is linear in each fiber . this allows us to define a linear map @xmath110 . the symbol @xmath111 represents a functor from the category of vector bundles over @xmath24 to the category of vector spaces . in the following , @xmath82 will be a vector bundle map over the identity on @xmath24 . it will be constructed by using linear maps @xmath112 defined identically for all @xmath113 ( such that the resulting map @xmath82 will be smooth ) . denoting as @xmath114 and @xmath115 the bundle projections , the map @xmath82 is defined by @xmath116 for all @xmath117 and all @xmath31 . often the maps @xmath118 will be bijections , making @xmath82 a vector - bundle isomorphism @xcite , and @xmath119 a linear isomorphism . let us introduce _ elastic energy density _ as a fiber bundle morphism @xmath120 over the identity on @xmath24 . this means @xmath121 is a smooth fiber preserving map , that is , it maps vectors from @xmath122 to @xmath123 for all @xmath31 . ( see for example @xcite . ) therefore , its restriction to the fiber @xmath122 is a smooth map @xmath124 for each @xmath31 . in the most simple cases ( corresponding to linear constitutive laws ) these restricted maps will be quadratic ( as in section [ example ] below ) . taking the composition of @xmath121 and a section @xmath58 of @xmath59 we get a section of @xmath91 , and then we will say that @xmath121 operates on @xmath58 to produce the 3-form @xmath125 . to express the invariance to rigid displacements , let us decompose the displacement gradient into symmetric and antisymmetric parts using the metric @xmath50 . for this , we define linear maps @xmath126 identically for all @xmath31 , by setting @xmath127 . for the antisymmetric ( or skew ) part , we define linear maps @xmath128 by @xmath129 . then , for any @xmath130 we have the decomposition @xmath131 . the resulting vector bundle maps , and the linear maps given by @xmath111 , will be denoted by the same symbols . the classical small - strain tensor , defined on @xmath132 using the rigid observer s metric , is @xmath133 , where @xmath134 is the lie derivative . its relation to the displacement gradient is given by using @xmath135 and the naturality of the lie derivative with respect to push - forward , that is , @xmath136 . in the present formulation , the invariance of elastic energy to rigid displacements is taken into account by insisting @xmath137 not to depend on @xmath138 . elastic energy density @xmath121 lets us introduce _ elastic energy _ as the map @xmath139 where @xmath121 is considered as a ( non - linear ) map @xmath140 , and @xmath141 is a linear map @xmath142 . this expression for the elastic energy is invariant to changes of observer , provided that the observed connection is used . the transformed elastic energy density under @xmath69 is defined , using the pull - back of 3-forms , such that @xmath143 , and the elastic energy is given by integrating @xmath144 over @xmath24 . therefore , the variational procedure of the next section may be carried out in case of arbitrary observer that uses the observed connection . denote an arbitrary `` adjusted '' connection of the observer whose observed connection on @xmath24 is @xmath49 . because the difference @xmath145 transforms as a tensor , it is possible to define an `` adjusted '' elastic energy density @xmath146 , such that @xmath147 . this suggests that there is a connection - free manner of representing elastic energy and its variations , using first order jets of sections of @xmath43 . we will not pursue this point further in this paper . ] elastic energy variations appear through variations of elastic energy density . for the variations of elastic energy density that result from variations of displacement gradient , we take the derivatives of the restricted maps @xmath124 corresponding to all @xmath31 . at each point @xmath31 , the derivative @xmath148 has the operation @xmath149 the derivative depends only on the topologies on the spaces @xmath122 and @xmath150 ; not on any particular norms that induce the topologies @xcite . further , arbitrary norms may be used here , as in case of a finite dimensional vector space ( over real or complex numbers ) all norms induce the same topology @xcite . we may define , for each displacement gradient value @xmath58 ( section of @xmath59 ) , a section @xmath151 of the bundle @xmath152 by @xmath153 for all @xmath31 . the section @xmath151 can operate on variations of displacement gradient ( sections of @xmath59 ) to yield variations of elastic energy density ( sections of @xmath91 ) . now we are in the position to express a variational principle as the defining equation for the displacement @xmath40 , when the energy density @xmath121 , the body force @xmath92 , and surface force @xmath100 on some ( possibly empty ) part of @xmath95 , are specified . if @xmath100 is not specified at some part @xmath154 we specify @xmath155 on this boundary part . then , according to the basic principle , we require that the virtual deformation work done by the given forces coincides with the variation of elastic energy for all kinematically admissible virtual displacements . accordinglyvariations of elastic energy may be considered once a norm is given for the vector space @xmath107 , and its completion to a banach space is performed . a norm may be defined , for instance , by using the metric @xmath50 on @xmath24 to define an inner product for @xmath107 . then , the variation of elastic energy corresponding to virtual displacement @xmath109 is given by using the frchet derivative as @xmath156 ( assuming the differentiability of @xmath157 ) . let us assume that norms are given also for @xmath158 and @xmath159 by similar procedure , and that their completion is performed to obtain banach spaces . ( inner product for @xmath159 may be defined by @xmath160 , using the exterior product @xmath161 of differential forms and the hodge operator @xmath162 implied by the metric @xmath50 . similarly , inner products for both @xmath107 and @xmath158 may be defined by @xmath163 , using the product @xmath164 and the hodge operator @xmath165 defined later in this section . ) then , by using the chain rule , and taking into account the linearity of @xmath49 and @xmath141 , the variation of elastic energy becomes @xmath166 ( assuming the required differentiability ) . it is not difficult to show , assuming differentiability , that the frchet derivative of @xmath167 coincides with the ( pointwise ) derivative defined in . finally , for _ arbitrary metric _ to be allowed in this variational procedure , it is left to be shown that the differentiability of @xmath49 , @xmath121 , and @xmath141 , is independent of the metric conferred on @xmath24 by which the required norms are constructed . ] , the problem is to find @xmath168 , with @xmath169 predefined on @xmath154 , such that @xmath170 for all @xmath109 , with @xmath171 zero on @xmath172 . on the right - hand - side the virtual deformation work @xmath173 is evaluated according to . in equation , the following requirement concerns @xmath106 , when @xmath174 is predefined on all of @xmath95 . because of the invariance of elastic energy density to rigid displacements , this prescribed load must yield zero virtual work for rigid virtual displacements @xmath93 , that is , those satisfying @xmath175 . indeed , the energy density @xmath121 is of the form @xmath176 for some fiber bundle morphism @xmath177 , and hence , using the chain rule to the pointwise derivative , we have @xmath178 such that rigid virtual displacements produce zero elastic energy variations . given a vector bundle @xmath51 over @xmath24 , its dual bundle @xmath179 will be denoted as @xmath180 . the specific notation @xmath181 will be used for the cotangent bundle , that is , the dual bundle of @xmath43 . we will sometimes abuse the notation by using the same symbol for a section of a bundle and its value at a point . body force may be viewed as a covector - valued 3-form because of an isomorphism between @xmath90 and @xmath182 , where @xmath183 is the bundle of 3-vectors on @xmath24 . we define the linear isomorphism between @xmath184 and @xmath185 identically for all @xmath31 . the isomorphism assigns to the element @xmath186 of @xmath185 the element @xmath187 defined by @xmath188 for all @xmath189 and @xmath190 . accordingly , body force may also be considered as the section @xmath186 of @xmath182 . the generalization of this isomorphism to covector - valued p - forms is straightforward @xcite . similarly , surface force may be considered as the section @xmath174 of @xmath191 , where @xmath192 is the bundle of 2-vectors on @xmath95 , and @xmath193 is the pull - back of @xmath181 by @xmath194 . here , we will define stress as a covector - valued 2-form on @xmath24 , that is , as a section of @xmath195 , which will be able to produce energy density variations from given displacement gradient variations . this performance will be obtained by using an isomorphism between @xmath195 and @xmath152 . this will allow us to consider stress as a section of @xmath152 , which can operate on sections of @xmath196 to produce sections of @xmath91 . we define the linear isomorphism between @xmath197 and @xmath198 identically for all @xmath31 . the isomorphism assigns to @xmath199 the element @xmath200 defined by @xmath201 for all @xmath202 and @xmath203 . linearity and bijectivity of this map is verified in appendix [ isomorphism ] . since the above deals with twisted forms the triplet @xmath204 on the left hand side is in fact a twisted 3-vector , and it therefore comes with a transverse orientation . in our three dimensional case this is just a sign ( plus or minus ) . the tuples @xmath205 , @xmath206 and @xmath207 on the right hand side are then twisted 2-vectors , whose transverse orientations ( crossing directions ) are specified by @xmath208 , @xmath209 and @xmath210 , respectively , the sign being that of @xmath204 . the given isomorphism between the bundles @xmath195 and @xmath152 is a special case of the relation between `` variational stresses '' and `` cauchy stresses '' given in @xcite . we may now define _ stress _ @xmath211 as the section of @xmath195 whose corresponding section @xmath212 of the bundle @xmath152 coincides with the derivative of elastic energy density , that is @xmath213 which serves as a _ constitutive law _ between stress and displacement gradient . ( again , we note that it is possible to define stress in the general context , without taking sides to the constitutive law , see for example @xcite . ) the constitutive law is invariant to changes of observer , that is , we have @xmath214 , where @xmath215 . ( here , we use the push - forward of bivectors , and pull - back of 1-forms . ) this can be verified by considering the transformation rule of @xmath121 , applying the chain rule to the pointwise derivative , and making use of the naturality of the above defined isomorphism with respect to diffeomorphisms , that is , @xmath216 . let us now define a product for vector- and covector - valued forms which has similar properties as the exterior product of differential forms . this will motivate the above usage of the symbol @xmath164 . the product appears also in @xcite . we first define bilinear maps @xmath217 identically for all @xmath31 . let us denote as @xmath218 the set of all permutations @xmath219 of the index set @xmath220 that satisfies @xmath221 and @xmath222 . then , for @xmath223 and @xmath224 , their product @xmath225 is defined , for @xmath226 , by @xmath227 for all @xmath228 , where @xmath229 is the signature of the permutation . this defines the values of @xmath230 for simple ( p+q)-vectors . this is sufficient when @xmath25 . the bilinear map @xmath231 is defined similarly to fullfill graded anticommutativity @xmath232 in analogy to the exterior product of differential forms . by the above definition , we have a vector bundle map over the identity on @xmath24 from the bundle @xmath233 to @xmath234 , with the operation @xmath235 . also , we have a vector bundle map from @xmath236 to @xmath237 over the identity on @xmath24 . by using these maps we may construct the products @xmath238 and @xmath239 that are bilinear and fullfill graded anticommutativity . similar construction applies in case of forms defined on @xmath95 ( and is required to deal with traces of forms ) . we will sometimes regard sections of @xmath43 as vector - valued 0-forms , that is , as sections of the bundle @xmath240 , where @xmath241 is defined to be @xmath242 . this is simply achieved by assigning to @xmath243 the element of @xmath244 whose operation on @xmath245 produces @xmath62 . the restriction of forms to the boundary @xmath95 may be defined by using the natural inclusion @xmath96 . for a vector - valued 0-form @xmath246 its trace @xmath247 is the element of @xmath248 defined by @xmath249 . the trace defined here replaces , when we identify vector fields with vector - valued 0-forms , the restriction of vector fields defined in [ force ] . ( the required commutative diagram is readily checked . ) to define traces of covector - valued p - forms , we make use of the following procedure . for the element @xmath250 we denote as @xmath251 an element of @xmath244 such that @xmath252 . then , the trace of covector - valued p - form @xmath253 is the element @xmath254 defined , for @xmath255 , by @xmath256 for all @xmath250 , where @xmath257 on the right - hand - side is the trace of real - valued p - forms defined as the pull - back by @xmath194 . ( to pull back twisted forms , @xmath95 must be transverse orientable @xcite . ) from , it follows that @xmath258 . the trace of vector - valued p - forms may be defined similarly ( but is not required in this paper ) . assuming the stress form @xmath211 smooth enough , the variational ( or weak ) formulation has the equilibrium equation with boundary conditions as a pointwise ( or strong ) counterpart . we go through the derivation here , as it will be a feasible introduction to the covariant exterior derivative . we will sometimes view vector fields as vector - valued 0-forms , and vice versa , without making the difference explicit in the notation . considering , with the definitions and , and with the term @xmath259 added and subtracted . we have @xmath260 for all @xmath261 , with @xmath262 zero on @xmath154 . let us express the integrand in the rightmost term as @xmath263 , and replace the other boundary integral involving @xmath264 by an integral over @xmath24 by using the exterior derivative @xmath265 . we have @xmath266 for all @xmath261 , with @xmath262 zero on @xmath172 . the term @xmath267 contains derivatives of both @xmath211 and @xmath93 , but it turns out that the involved derivatives of @xmath93 are canceled when we subtract the term @xmath268 . therefore , the map @xmath269 is not only linear but function - linear - module of vector fields , where @xmath66 is as in footnote [ module1 ] . ] , and we have the remarkable situation that this map defines a section of @xmath90 . further , since @xmath270 is isomorphic to @xmath271 , there is a unique covector - valued 3-form , denoted here as @xmath272 , such that @xmath273 for all @xmath109 . using this in ( [ weak_ela ] ) , we get @xmath274 for all @xmath261 , with @xmath262 zero on @xmath172 . thus , with variations vanishing on @xmath95 , we gethere , we may use a metric on @xmath24 to define ( overloading the notation @xmath24 for the manifold @xmath275 ) an inner product on @xmath271 by @xmath276 , employing the hodge operator @xmath277 of subsection [ hodge ] . we then have the fact that @xmath278 for all @xmath279 implies @xmath280 . ] the pointwise equilibrium equation @xmath281 and with variations supported on @xmath282 , we get . ] the boundary condition @xmath283 finally , to define the force functional of section [ force ] , we set @xmath284 the above defines operator @xmath285 , which is linear by the linearity of exterior derivative and bilinearity of the product @xmath164 . this operator may be generalized to vector- and covector - valued p - forms as follows . the operator defined here is a special case of the derivatives defined in @xcite . see also @xcite . for covector - valued p - forms the covariant exterior derivative is the linear map @xmath286 defined by @xmath287 for all @xmath253 and @xmath243 . in symmetric fashion , for vector - valued p - forms the covariant exterior derivative is the linear map @xmath288 defined by @xmath289 for all @xmath290 and @xmath291 . in case of a vector - valued 0-form the derivative of ( [ cov_ext_der_vec ] ) reduces to the covariant derivative of the corresponding vector field . this is a straightforward consequence of the above definitions . therefore , as a replacement for we may write @xmath292 where @xmath293 . the derivative @xmath294 can be used to express curvature , as for vector fields @xmath295 we have @xmath296 where @xmath297 is the curvature of the connection @xcite . thus , in case of a flat connection we have @xmath298 for all @xmath299 . in this flat case we further have @xmath300 for general vector- and covector - valued p - forms , in analogy to the property @xmath301 of the exterior derivative @xmath265 . the metric @xmath50 on @xmath24 yields a procedure to map , for example , a vector - valued p - form linearly to a covector - valued twisted ( n - p)-form , where @xmath302 . this map can be thought of as a vector bundle map @xmath303 over the identity on @xmath24 , or the resulting linear map given by @xmath111 . ( the operation of @xmath111 on vector bundle maps was defined in the second paragraph of subsection [ ela_energy_section ] . ) we will define linear maps @xmath304 identically for all @xmath31 . for this , recall that the hodge operator for p - vectors is defined as follows . given @xmath305 , we consider the linear map @xmath306 defined by @xmath307 . selecting a unit n - vector @xmath219 for the basis of @xmath308 , this map may be identified with a linear map @xmath309 , and the riesz representation theorem implies the existence of unique @xmath310 , such that @xmath311 where @xmath312 is the inner product on @xmath313 implied by @xmath50 ( see for example @xcite ) . the twisted ( n - p)-vector represented by @xmath314 ( or @xmath315 ) is uniquely determined by @xmath62 , defining the operator @xmath316 . also , the ( n - p)-vector @xmath317 is uniquely determined by the twisted p - vector represented by @xmath318 , defining the operator @xmath319 . now , we may define the hodge operator for p - covectors , and further , for elements of @xmath320 . for instance , the operator @xmath165 in is defined by its operation on any @xmath321 as @xmath322 for all @xmath323 and @xmath189 . the definition has the following consequence . using a basis @xmath324 for @xmath61 , and its dual basis @xmath325 to express any @xmath321 as @xmath326 ( summation over repreated indices with @xmath327 ) , we have @xmath328 where @xmath329 is the hodge operator of p - covectors , and @xmath330 is the covariant version of @xmath331 . by reusing the notation @xmath332 for the inner product on @xmath320 , defined by @xmath333 we have @xmath334 where @xmath335 is the volume element implied by @xmath50 . the preceding yields the hodge operator @xmath336 similar construction yields , for instance , the hodge operator @xmath337 whose definition uses the inverse of the metric correspondence @xmath338 . in the following we will need the trace @xmath339 , taking vector - valued 1-forms linearly to real - valued functions . for this , linear maps @xmath340 are defined identically for all @xmath31 , by setting @xmath341 ( sum over @xmath342 ) . the definition is independent of the used basis . the resulting vector bundle map , and the linear map given by @xmath111 ( see the first paragraph of subsection [ local_forces ] ) , will also be denoted as @xmath339 . since a section @xmath58 of @xmath59 may be considered as a vector bundle map @xmath343 over the identity on @xmath24 , we can form the composition @xmath344 , which is again a vector bundle map @xmath345 over the identity on @xmath24 , and can therefore be considered as a section of @xmath59 . it will be denoted as @xmath346 . in the present paper we take a restricted view of electromagnetism , allowing only static observers ( with respect to the rigid inertial observer @xmath15 of section [ elastostatics ] ) that sense pure magnetic field . this way relativistic effects are avoided , and we may use the classical space - time model of section [ setting ] . in preparation to formulate the coupled magneto - elastic problem , we wish to choose for the problem domain a reference manifold @xmath347 that has the material body @xmath24 as its submanifold . for this , we need to transfer the magnetic quantities observed on @xmath1 to the reference manifold @xmath347 . for static formulation , this means taking into account the particular configuration of @xmath24 which corresponds to the observer s measurement of the magnetic quantities . in the following , the formulation will be given directly on a 3-dimensional reference manifold @xmath347 with boundary @xmath348 , and its relation to measurements will be considered in section [ example ] , where constitutive laws will be specified . we will emphasize the analogy of magnetostatics to the formulation of elasticity given above . we will be brief and show only the main points , as it is straightforward to fill in the details by using the analogy with the previous section . let us model _ magnetic induction _ @xmath349 as a 2-form , that is , an element of @xmath350 . the absence of magnetic charges is imposed by expressing @xmath349 as the exterior derivative of @xmath351 , that is , we require @xmath352 where @xmath353 will be called _ magnetic covector potential_. this equation is invariant to changes of allowable observers . for , the change of observer is represented by a diffeomorphism @xmath354 , and the exterior derivative is natural with respect to diffeomorphisms , that is @xmath355 . therefore , denoting as @xmath356 and @xmath357 the transformed magnetic induction and magnetic covector potential , that is , @xmath358 and @xmath359 , we have @xmath360 . to record the virtual work done by electrical energy sources to generate a variation @xmath361 of the magnetic covector potential , we introduce the _ current density _ as a section of @xmath362 . with hindsight , it is denoted as @xmath363 . ( as the notation suggests , there is an isomorphism between @xmath362 and @xmath364 , but this will be required only later . ) also , we introduce a _ surface magnetic field intensity _ as a section of @xmath365 , denoted as @xmath366 . let us then introduce _ magnetic drive force _ @xmath367 as the linear functional on @xmath368 defined by @xmath369 for all @xmath370 , where @xmath371 is the natural inclusion . the term @xmath366 may not be given beforehand on some part of @xmath348 , and then it will be defined on this part by using the notion of magnetic field intensity ( section [ ampere ] ) . the equation is invariant to changes of allowable observers . for a variational formulation of magnetism , we introduce _ magnetic energy density _ as a fiber bundle morphism @xmath372 over the identity on @xmath347 . therefore , at each point @xmath373 we have a smooth map @xmath374 ( quadratic in case of linear constitutive laws ) . magnetic energy density allows us to define _ magnetic energy _ as the map @xmath375 where @xmath376 is considered as a ( non - linear ) map @xmath377 , and integration is a map @xmath378 . this expression for magnetic energy is independent of allowable observers . the transformed magnetic energy density under @xmath354 is @xmath379 , and the magnetic energy is given by integrating @xmath380 over @xmath347 . therefore , the following variational procedure may be carried out in case of any allowable observer . to consider variations of magnetic energy density that result from variations of magnetic induction , we take the derivatives of the restricted maps @xmath381 corresponding to all @xmath373 . that is , we define , for each magnetic induction value @xmath349 ( section of @xmath382 ) , a section @xmath383 of @xmath384 by @xmath385 for all @xmath373 . the section @xmath383 can operate on magnetic induction variations ( sections of @xmath382 ) to produce energy density variations ( sections of @xmath386 ) . to state the variational principle for magnetostatics we assume that magnetic energy density @xmath376 and current density @xmath363 are specified on @xmath347 , and that @xmath366 is specified on some ( possibly empty ) part of @xmath348 . further , @xmath387 is predefined on the part @xmath388 where @xmath389 is not specified . then , we take as a basic principle that the virtual work done by external electrical energy sources coincides with the variation of magnetic energy for all admissible magnetic covector potential variations . accordinglysimilar consideration applies here as in footnote [ elastic_energy_note ] . ] , the problem is to find @xmath351 , with @xmath387 predefined on @xmath390 , such that @xmath391 for all @xmath370 with @xmath392 zero on @xmath390 . if @xmath367 is completely specified beforehand it must be done such that @xmath393 for all variations @xmath394 with @xmath395 . various numerical solution methods make use of the fact that ( electro)magnetism may be modeled by using exterior algebra equipped with the exterior derivative and a constitutive law @xcite . the exterior product @xmath396 may be constructed in an analogous manner to what we did with the product @xmath164 in section [ product ] . we only state here that @xmath161 is bilinear , graded anticommutative , and associative , and it makes ( letting @xmath397 ) the direct sum @xmath398 into a graded algebra . ( we have used the notations @xmath399 and @xmath400 . ) when equipped with the exterior derivative , we have an instance of differential graded algebra . we also have the graded algebra of forms on the boundary @xmath348 . then , the trace @xmath401 is an algebra homomorphism defined by pulling back forms on @xmath347 using @xmath371 . ( to pull back twisted forms the boundary @xmath348 must be transverse orientable . ) the same symbol @xmath257 will be used for the trace of forms to ( n-1)-dimensional submanifolds of @xmath347 ( material interfaces ) . for the constitutive law , we use the magnetic energy density @xmath376 , and define the _ magnetic field intensity _ @xmath402 by @xmath403 using the isomorphism between @xmath404 and @xmath405 provided by the exterior product . this constitutive law is invariant to changes of allowable observers , that is , we have @xmath406 , where @xmath407 . ( this can be shown by considering the transformation rule of @xmath376 , applying the chain rule to the pointwise derivative , and making use of the compatibility of pull - back with the exterior product . ) consequently , the derivation of the following ampere s law is valid in case of any allowable observer . we use the above algebraic structure to derive ampere s law and boundary conditions . to allow for discontinuities of @xmath408 on material interfaces , this 1-form is required to be only piecewise smooth on @xmath347 . then , equation , together with and , is equivalent to ampere s law @xmath409 on regularity regions where @xmath408 is smooth , the interface condition @xmath410 = 0\end{aligned}\ ] ] setting the discontinuity @xmath411 $ ] of @xmath412 to zero on the boundaries of these regularity regions ( assuming zero surface current on these interfaces ) , and the boundary condition @xmath413 the verification of this is similar to the calculation performed in subsection [ equilibrium_eq ] ( but easier since now we are dealing with real - valued forms ) . finally , to define the magnetic drive force of section [ var_magnetic ] , we set @xmath414 in this section , the above described elastic and magnetic problems are coupled by taking both the displacement gradient and the magnetic induction as system state variables . we allow magneto - elastic coupling both through constitutive behavior and through magnetic forces . for a variational formulation of magneto - elasticity , we recall that the whitney sum @xmath415 of @xmath416 and @xmath382 is the vector bundle whose fiber above @xmath373 is the direct sum @xmath417 ( see for example @xcite ) . then , we introduce _ magneto - elastic energy density _ as a fiber bundle morphism @xmath418 over the identity on @xmath347 . accordingly , when restricted to the fiber above @xmath373 , we have a smooth map @xmath419 . the 3-form produced when operating by @xmath121 on section @xmath420 of @xmath415 will be denoted as @xmath421 . to consider variations of energy density that result from displacement gradient variations , we take the partial derivatives of the restricted maps @xmath422 , corresponding to all @xmath373 , with respect to the first argument . accordingly , for given section @xmath423 of @xmath415 , we define a section @xmath424 of @xmath425 by @xmath426 for all @xmath373 . similarly , to consider variations of energy density that result from magnetic induction variations , we take partial derivatives with respect to the second argument . that is , for given section @xmath423 of @xmath415 , we define a section @xmath427 of @xmath428 by @xmath429 for all @xmath373 . next , we describe two different models for magneto - elasticity , having slightly different ranges of applicability . the model described here assumes not only small displacement gradients , but also small overall deformations , such that the stress - free reference configuration may be assumed to correspond with the observer s measurement of the magnetic quantities . ( really , the stress - free reference configuration correspond to zero magnetic field , and the observer s measurement of magnetic quantities takes place in the deformed equilibrium configuration , where magnetic forces are balanced by inner stresses in the material . ) in the given model , the modeler directly specifies a rule that gives magnetic stresses from the state variables of the coupled system . this will result in magnetic forces on the right hand side of the equilibrium equation . this means that the modeler should not include in the magneto - elastic energy density @xmath121 a particular mechanism by which magnetic field is affected by displacement gradient , namely , the mechanism that identifies the magnetic measurements to the deformed equilibrium configuration . for , this particular mechanism is already taken into account on the right hand side of the equilibrium equation by the prescribed magnetic forces . consequently , as the particular mechanism becomes excluded from the constitutive law , this construct assumes not only small displacement gradients in the material body , but also small overall deformations . for example , this rules out situations with `` long '' material bodies , which may deform significantly despite of small strains in the material body . for the coupling through magnetic forces , we proceed in two steps . first , we assume a fiber bundle morphism @xmath430 over the identity on @xmath347 , which will be called _ magnetic stress mapping_. in the simplest cases the maps @xmath431 will be quadratic in @xmath432 , and independent of @xmath433 ( section [ example ] ) . when operating on a section @xmath420 of @xmath434 this will give a covector - valued 2-form , denoted as @xmath435 . the above magnetic stress mapping allows the kind of magnetic stresses used typically in engineering , see for example @xcite . as the next step , we replace the force functional of section [ force ] by a map @xmath436 , defined such that @xmath437 ) \dot{\wedge } i^ * \delta \nu\end{aligned}\ ] ] for all @xmath438 . here , the term @xmath439 $ ] is the trace of @xmath440 taken from the outside of @xmath24 minus the trace taken from the inside . if @xmath174 is not given beforehand on some part of @xmath95 , it will be defined on this part by using stress . the coupled problem may now be stated as follows . let us be given magneto - elastic energy density @xmath121 , magnetic stress mapping @xmath441 , the mechanical ( as opposed to magnetic ) body force @xmath92 on @xmath24 , and the mechanical surface force @xmath100 on some part @xmath282 of @xmath95 , the current density @xmath363 on @xmath347 , and the surface magnetic field intensity @xmath366 on some part @xmath442 of @xmath348 . then , using the inclusions @xmath443 and @xmath444 , the problem is to find @xmath445 and @xmath351 , with @xmath446 predefined on @xmath172 and @xmath447 predefined on @xmath390 , such that @xmath448 for all @xmath449 with @xmath450 zero on @xmath172 , and @xmath451 for all @xmath370 with @xmath452 zero on @xmath390 . we may now use the differential graded algebra of exterior forms , and the calculus of vector- and covector - valued forms developed in section [ elastostatics ] , to derive the equilibrium equations of the coupled system . first , for strain @xmath58 and magnetic induction @xmath349 , we define @xmath453 by @xmath454 and @xmath402 by @xmath455 for convenience , we use the notation @xmath456 for @xmath440 . then , by using and in , we get the equilibrium equations @xmath457 \ \ \mbox{on } \partial m \setminus s_e . \label{equi_coup2}\end{aligned}\ ] ] by using and in , we get the familiar - . finally , to define the map @xmath106 , we set @xmath458 \ \ \mbox{on } s_e,\end{aligned}\ ] ] and to define the magnetic drive force , we set @xmath414 this construction allows for magnetostrictive behavior , in which case there is genuine constitutive coupling between magnetism and elasticity @xcite . here , it is further possible to decompose the energy density into two parts . that is , one can define `` elastic '' energy density @xmath459 and `` magnetic '' energy density @xmath460 as fiber bundle morphisms @xmath461 , such that @xmath462 where the sum is defined fiberwise ( using the vector space structure of the fibers of @xmath386 ) . for example , we may define @xmath459 independently of @xmath349 , by @xmath463 , and then @xmath460 is defined by @xmath464 . in case of magnetostrictive behavior , this @xmath460 depends on @xmath58 , and then the use of in the constitutive law yields a decomposition @xmath465 where the `` elastic '' stress @xmath466 is defined independently of @xmath349 by @xmath467 , and the `` magnetostrictive '' stress @xmath468 is defined by @xmath469 . this results in a `` magnetostrictive '' force in the equilibrium equations for @xmath466 . ( this is in addition to the magnetic force term in and . ) when magnetostrictive behavior is allowed the above decomposition of energy density is not unique . we may instead define @xmath460 independently of @xmath58 , by @xmath470 , and then @xmath459 is defined by @xmath471 . because now @xmath460 is independent of @xmath58 , no `` magnetostrictive '' forces appear . ( instead , one can decompose @xmath408 to get an additional `` current density '' term to the ampere s law written for the `` magnetic '' part of @xmath408 . ) in the model described here , all mechanisms by which magnetic field is affected by displacement gradient are included in the magneto - elastic energy density @xmath121 . the model allows taking into account that the measurement of magnetic quantities takes place in the deformed equilibrium configuration , and not in the stress - free reference configuration . accordingly , it may be used in case of small displacement gradients in the material body , and large overall deformations . this construction , when restricted from the beginning to euclidean geometry , is considered in @xcite . because the coupling is realized totally by using magneto - elastic energy density , the notion of magnetic force is not introduced in the first place . consequently , we use the force functional @xmath472 of subsection [ force ] . the coupled problem may be stated as follows . let us be given the magneto - elastic energy density @xmath121 , the mechanical ( as opposed to magnetic ) body force @xmath92 on @xmath347 ( zero on @xmath473 ) , and the mechanical surface force @xmath100 on some part @xmath282 of @xmath95 , the current density @xmath363 on @xmath347 , and the surface magnetic field intensity @xmath366 on some part @xmath442 of @xmath348 . then , we find @xmath474 and @xmath351 , with @xmath446 predefined on @xmath172 and @xmath447 predefined on @xmath390 , such that @xmath475 for all @xmath476 with @xmath450 zero on @xmath172 , and @xmath477 for all @xmath370 with @xmath452 zero on @xmath390 . we emphasize that and are not yet sufficient to solve for @xmath478 , and , as an additional requirement , @xmath40 is extended smoothly to @xmath473 , such that @xmath169 vanishes on @xmath348 . the energy density @xmath121 must be so specified that @xmath353 and @xmath40 on @xmath24 , and total system energy , will not depend on the particular extension of @xmath40 . ( for an example where the extension of @xmath40 is performed according to laplace equation , see @xcite . ) the equilibrium equations may be derived as follows . first , for strain @xmath58 and magnetic induction @xmath349 , we define @xmath453 by @xmath479 and @xmath402 by @xmath480 we assume @xmath211 only piecewise smooth to allow for `` magnetic surface forces '' . from , using and , we may now infer the equations @xmath481 & = \tilde{\tau } \ \ \mbox{on } \partial m \setminus s_e , \label{equi_coup2_2}\end{aligned}\ ] ] and from , using and , we get the familiar - . in the term @xmath482 $ ] is the trace of @xmath211 taken from the inside of @xmath24 minus the trace taken from the outside . finally , to define the force functional , we set @xmath483 \ \ \mbox{on } s_e,\end{aligned}\ ] ] and to define the magnetic drive force , we set @xmath414 in the present example , magneto - elastic coupling is due merely to magnetic forces ( no magnetostriction ) . also , we consider linear isotropic material behavior . the magneto - elastic energy density is expressed as the sum @xmath484 where the elastic and magnetic energy densities @xmath459 and @xmath460 are defined as follows . the term @xmath459 is defined independently of @xmath349 , such that @xmath485 where @xmath486 and @xmath487 are the lam constants , @xmath332 is the inner product of vector - valued 1-forms implied by the metric @xmath50 , and @xmath488 is the volume form implied by @xmath50 . the term @xmath460 is defined independently of @xmath58 , such that @xmath489 where @xmath490 is the reluctivity ( piecewise smooth function , taken here as element of the multiplicative ring of the module of 1-forms ) , and @xmath162 is the hodge operator of 2-forms implied by @xmath50 . this specifies the observer s measurement of magnetic quantities to the stress - free reference configuration . for the elastic constitutive law , we have @xmath491 , resulting in @xmath492 where @xmath493 is the identity vector - valued 1-form . one can verify that this constitutive law gives zero variations @xmath268 of elastic energy density in case @xmath494 is antisymmetric , that is , in case @xmath495 vanishes . for the magnetic constitutive law , we have @xmath496 , resulting in @xmath497 to specify the magnetic stress mapping @xmath441 , we use the interior product of differential forms by vector fields . for p - form @xmath498 and vector field @xmath62 , their interior product @xmath499 is the ( p-1)-form defined such that @xmath500 for all vector fields @xmath501 ( see for example @xcite ) . now the mapping @xmath441 is defined independently of @xmath58 , such that @xmath502 for all vector fields @xmath63 . with euclidean metric , this is the classical maxwell s stress tensor , as measured in the reference configuration . we give an example with euclidean geometry . as in the previous example , we assume negligible magnetostriction . this means that magneto - elastic coupling will be due merely to change in magnetic energy that results from change in geometry ( change in shape of the material body @xcite ) . as in the previous example , we restrict to linear isotropic material behavior . the magneto - elastic energy density is the sum of elastic and magnetic energy densities @xmath459 and @xmath460 . the term @xmath459 is defined by as above , and @xmath460 is defined such that @xmath503 where @xmath504 is the hodge operator of 2-forms implied by a metric @xmath505 that depends on @xmath50 and @xmath58 as follows . the metric @xmath505 is defined by @xmath506 where @xmath507 etc . denote the components of @xmath58 , and @xmath508 are the components of the identity vector - valued 1-form . the above energy density specifies the observer s measurement of magnetic quantities to the deformed equilibrium configuration . in the euclidean geometry , the displacement @xmath40 yields the change of configuration , and then the metric @xmath505 of , expressed in terms of @xmath509 , is the pull - back of @xmath50 by the change of configuration . ( for details , see @xcite , pp . 57 - 60 . ) observe that the metric @xmath505 is defined also in @xmath473 , because displacement is extended smoothly to this region . we may define @xmath456 by @xmath510 , and then we will get the equilibrium equations @xmath511 \ \ \mbox{on } \partial m \setminus s_e , \label{equi_example2}\end{aligned}\ ] ] just as equations and in section , but therein the dependence of @xmath456 on @xmath58 originates from material behavior . finally , having assumed small displacement gradients , we may neglect the term in which is quadratic in @xmath58 ( at least in the material body @xmath24 ) . using this approximation for the metric , and performing the derivation @xmath512 , yields magnetic stress @xmath456 which coincides with the familiar maxwell s stress tensor , as measured in the deformed equilibrium configuration . but if magnetic stresses are measured in the equilibrium configuration , so should the magnetic forces , and for this , we should use the magnetic forces @xmath513 , where @xmath514 is the levi - civita connection of @xmath505 . however , this does not arise from the above formalism . it seems that the setup of small - strain elasticity and the setup of taking account the change of geometry in magnetics does not perfectly fit together . a consistent formulation could be achieved by considering geometrically non - linear elasticity , where the balance of forces and stresses is enforced in the deformed equilibrium configuration . the formulation of small - strain elasticity by using some of the recent developments of differential geometry strengthens similarities between small - strain elasticity and magnetism . the given perspective point to the use in small - strain magneto - elasticity the same discretization methods that have already proved powerful in electromagnetic field computation . to make such a program systematic , we suggest searching for algebraic category where a class of magnetic , elastic , and magneto - elastic problems may be modeled , and which constitutes suitable domain for discretizations . let us verify that the map @xmath515 with operation @xmath235 , where @xmath516 is defined by @xmath517 for all @xmath202 and @xmath203 , is linear isomorphism . for linearity , let us select arbitrary @xmath518 , @xmath519 , @xmath202 , and @xmath203 . we have @xmath520 therefore , we have the equality of the 3-covectors @xmath521 and thus the equality @xmath522 for all @xmath518 and @xmath519 . for bijectivity , let us first select a basis @xmath523 for the tangent space @xmath61 , and denote as @xmath524 the dual basis of @xmath525 . then , the tensors @xmath526 , where @xmath527 with @xmath528 , constitute a basis for @xmath529 . further , since the tensors @xmath530 , where @xmath531 , constitute a basis for the vector space @xmath532 , the dual basis @xmath533 of @xmath534 is defined by @xmath535 then , a basis for @xmath536 is given by @xmath537 , where @xmath531 . in these bases the map @xmath235 is given by ( using summation over repeated indices with @xmath528 ) @xmath538 where @xmath539 is the permutation symbol . ( this symbol takes the value @xmath540 or @xmath541 when the indices @xmath542 form an even or odd permutation of @xmath543 , respectively , otherwise it takes the value zero . ) thus , the linear map becomes represented by diagonal @xmath544 matrix , with @xmath545 on the diagonal , which is a non - singular matrix . r. bustamante , a. dorfmann and r. w. ogden , numerical solution of finite geometry boundary - value problems in nonlinear magnetoelasticity . international journal of solids and structures * 48 * ( 2011 ) , 874 - 883 .

simple elastic network models of dna were developed to reveal the structure - dynamics relationships for several nucleotide sequences . first , we propose a simple all - atom elastic network model of dna that can explain the profiles of temperature factors for several crystal structures of dna . second , we propose a coarse - grained elastic network model of dna , where each nucleotide is described only by one node . this model could effectively reproduce the detailed dynamics obtained with the all - atom elastic network model according to the sequence - dependent geometry . through normal - mode analysis for the coarse - grained elastic network model , we exhaustively analyzed the dynamic features of a large number of long dna sequences , approximately bp in length . these analyses revealed positive correlations between the nucleosome - forming abilities and the inter - strand fluctuation strength of double - stranded dna for several dna sequences .

1511.02026

elastic network models of proteins , including all - atom models@xcite and coarse - grained models@xcite represent some of the simplest and most powerful types of theoretical models that can accurately reveal structure - dynamics relationships and the mechanisms underlying a protein s functional activities @xcite . such models have also been widely employed to accurately reproduce the temperature factors on the crystal structure of a protein via normal - mode analysis . these models can thus demonstrate the large and slow deformations of proteins that are essential for protein functions , but remain difficult to demonstrate via all - atom molecular dynamics simulations . along with proteins , dna is the most important biomolecule for the activities of all living organisms . recent developments in molecular biology have revealed that dna contains several functional regions . therefore dna is no longer considered to have the sole function in the storage of genetic information , but is now known to be actively involved in gene regulation@xcite , insulator activity@xcite , and in the construction of chromosomal architectures@xcite such as heterochromatins and topologically associated domains through nucleosome formation and protein bindings@xcite . the functional behavior of each strand of dna is determined not only by chemical aspects of the nucleotides and base pairs but also by its physical characteristics such as the structure and dynamics of the nucleotide sequences in each functional region . however , comprehensive understanding of the physical properties of nucleotide sequences lags far behind the knowledge accumulated of their biochemical aspects @xcite . since the last century , much progress has been made in revealing the physical aspects of dna using all - atom normal - mode analysis@xcite and molecular dynamics simulations@xcite . several coarse - grained models of dna ( and rna ) have also been proposed . some of these describe the detailed shape of each nucleotide using three or more particles@xcite , whereas others describe each base pair by simply one or two particles@xcite . molecular dynamics simulations and normal - mode analysis of these models have identified the flexibilities , nucleosome - forming abilities , and zip - unzip transitions of the double helices of some specific dna sequences from tens to a few hundred base pairs in length . although these methods have proven to be very powerful for the analysis of the physical aspects of dna , molecular dynamics simulations are not suitable for conducting exhaustive analyses of several sequences simultaneously owing to the high computational costs of such extensive simulations ; for example , analyses of whole genomic and whole possible sequences . moreover , the normal - mode analysis of all - atomic models of long dna sequences is also computationally heavy . alternatively , data - driven methods have been proposed for determining the mechanical properties of dna with respect to the helical parameters and local flexibilities of base pairs from x - ray crystal structure analysis and all - atom molecular dynamics @xcite . these methods also seem to be powerful and can be applied to the analysis of the mechanical properties of several dna sequences simultaneously . however , the methods thus far proposed have focused on static and local mechanical properties . therefore , they are not particularly useful for the study of the functional contribution of the dynamic and correlated motions of dna . the objective of this study was to construct a simple coarse - grained elastic network model of dna to allow for exhaustive analysis of the dynamic correlated motions of several long dna sequences . for this purpose , we first constructed a simple all - atom elastic network model of short dna sequences based on the method introduced by tirion@xcite for the modeling of protein dynamics . we confirmed that this model could accurately reproduce the temperature factors of some dna fragments from data obtained with x - ray crystal structure analysis . second , we developed a simple coarse - grained elastic network model of long dna sequences , where each nucleotide is described by only one node . we confirmed that this simplified model has lower computational costs but can nonetheless reproduce the nucleotide sequence - dependent dynamics revealed by the all - atom elastic network model . finally , through the normal - mode analysis of this coarse - grained model , we conducted an exhaustive analysis of the general features of the sequence - dependent structure - dynamics relationships among several dna sequences . we specifically focused on the dynamic properties of a large number of long dna sequences , approximately @xmath0 bp in length ( with respect to the length of the nucleosome - forming regions ) , for the genomes of some model organisms as well as for random sequences with several a , t , c , g ratios . through these analyses , we found that the dynamic aspects of dna are highly influenced by their sequences , and found positive correlations between the nucleosome - forming abilities and inter - strand fluctuations of double - stranded dna . in particular , we focus on the geometry dependencies of the dynamics of several dna sequences , since a recent study demonstrated that the geometry of dna sequences has a dominant contribution to their mechanical features@xcite . in the following arguments , for simplicity , we use `` sequence - dependent '' to mean `` dynamics that depend on sequence geometry '' . in order to construct elastic network models of dna , the basic dna structure first needs to be determined . in the following arguments , we construct two types of models : i ) a simple all - atom elastic network model ( aaenm ) to reproduce the characteristics of the crystal structures of dna , and ii ) a simple coarse - grain elastic network model ( cgenm ) to reproduce the characteristics of the aaenm . we obtained the basic dna structures in the following two ways for the respective purposes of constructing each model . for construction of the aaenm , we employed the atom coordinate sets of several naked dna crystal structures included in the protein data bank ( pdb ) as the given basic structures of the model . we used @xmath1 dna crystal structures containing only dna and water molecules under different conditions of crystallization , where all temperature factors are given as positive values ( table 1 ) . the suitability of the model was evaluated by comparisons of the temperature factors obtained between the model and those obtained from x - ray crystal structure analysis . the objective of our constructed arguments was to unveil the sequence - dependent dynamic features of several long dna sequences simultaneously . in recent crystal structure analysis , only shorter dna sequences ( i.e. , less than @xmath2 bp in length ) were studied . on the other hand , several types of helical parameter sets , base pair parameters , and base step parameter sets have been proposed through experiments or molecular dynamics simulations@xcite(table 2 , ) . here , we used the application x3dna@xcite to obtain the coordinates of each atom for any sequence from these helical parameters , and used these parameters to construct the aaenms and cgenms of longer dna sequences ( for detailed methods of the generation of the atom coordinates , see lu and olson @xcite ) . in the following arguments , we compare the characteristics of the aaenm and cgenm constructed by x3dna from three different helical parameter sets : i ) a parameter set obtained by an _ in vitro _ experiment and x - ray crystal structure analysis@xcite ( table 2 ) , ii ) a parameter set obtained only by the x - ray crystal structure analysis , and iii ) a parameter set obtained by all - atom molecular dynamics simulations@xcite ( ) . a simple all - atom elastic network model of double - stranded dna was constructed based on the model proposed by tirion@xcite . in this model , we regard all of the atoms of given dna sequences as the nodes comprising the elastic network . for the analysis of the crystal structures of dna involving water molecules , we also regard the oxygen atoms of water around dna as the nodes of the elastic network . we define the mass and position of atom @xmath3 as @xmath4 and @xmath5 ( @xmath6 ) , respectively . the potential @xmath7 of all atoms is given as @xmath8 here , @xmath9 is the position of atom @xmath3 of the basic dna structure , as described above . the first term indicates the interaction potential among atoms that are spatially closed in the basic dna structure ( fig . 1(a ) ) . here , @xmath10 refers to the van der waals radius of atom @xmath3 , @xmath11 is an arbitrary cut - off parameter that models the decay of interactions with distance , and @xmath12 indicates the heaviside function , where @xmath13 ( @xmath14 ) for @xmath15 ( @xmath16 ) . we assume @xmath17 , which is empirically considered to be an appropriate range for biomolecules , at least for proteins@xcite . the results of the arguments presented in this paper were qualitatively unchanged for the appropriate range of @xmath11 . in the crystal structures of dna , the motions of water molecules are also restricted by the crystal packing . thus , for all atoms , we assume that spatially closed pairs of atoms are connected by linear springs with their respective natural lengths . the elastic coefficient @xmath18 ( @xmath19 $ ] ) is a phenomenological constant , which is assumed to be the same for all interacting pairs . interacts with all nucleotides connected by the 11 bold curves ; and ( c ) @xmath20 , and @xmath21 . , title="fig:",width=302 ] interacts with all nucleotides connected by the 11 bold curves ; and ( c ) @xmath20 , and @xmath21 . , title="fig:",width=302 ] interacts with all nucleotides connected by the 11 bold curves ; and ( c ) @xmath20 , and @xmath21 . , title="fig:",width=302 ] the second term indicates the boundary effects of each dna and water molecule in each dna crystal structure . this term plays a crucial role for the analysis of the fluctuations of the crystal structure of dna , such as temperature factors , since the fluctuations of the nucleotides at the edges of dna and water molecules are restricted due to the following facts . in the crystal of dna , the motion of edges at upper and lower streams of each dna segment ( left and right edges in fig . 1(a ) ) is influenced by atoms of other adjacent segments in the long - axis direction of dna segments . moreover , the motions of water molecules around each dna segment are influenced by atoms of other adjacent water molecules or dna segments in the direction perpendicular to the longitudinal axis of the dna segment . thus , we need to consider the second term of eq . ( 1 ) to implement such effects , where @xmath22 indicates the strength of such effects for atoms in the edge nucleotides of each dna and water molecule , respectively . remarkably , as shown in the results section , the distributions of the temperature factor of atoms exhibited various patterns from the same sequence of crystallized dna ( and ) owing to the dependency on the conditions of crystallization . therefore , in order to compare the characteristics of the present dna model to those of the crystal structure of dna , appropriate values of @xmath23 need to be assigned to the atom @xmath24 that belongs to the edge base pairs and water molecules . for simplicity , we assign @xmath25 and @xmath26 to the atoms @xmath24 at the edge base pairs and the water molecules , respectively , and @xmath27 otherwise . it is noted that the internal structure of the crystal of dna is spatially anisotropic . thus , it is reasonable to assume that the interactions among local parts of the crystal of dna exhibit different strengths in different directions . accordingly , in general , the strengths of the restrictions of atoms belonging to the edge of dna differ from those of water molecules . thus , we assume that @xmath28 and @xmath29 are different values ( table 1 ) . for simplicity , the mass of each atom is assumed as a constant value ( @xmath30 $ ] , @xmath31 $ ] is avogadro s number ) . however , we confirmed that the results were almost identical when using the precise masses of the atoms . a simple coarse - grained elastic network model of double - stranded dna , where each nucleotide is described as one node , was constructed as follows ( fig . we define the coordinate of the c1 carbon of nucleotide @xmath3 , @xmath32 ( @xmath33 ) , as the position of nucleotide @xmath3 , and regard the motion of the c1 carbon as that of the nucleotide . here , we assume that the mass of the c1 carbon obeys @xmath34 $ ] . the potential @xmath7 of all nucleotides is given as @xmath35 here , @xmath36 is the position of nucleotide @xmath3 of the basic structure of dna , as defined above . we assume that nucleotides @xmath3 and @xmath37 belong to the same base pair . for nucleotide @xmath3 , the sum is restricted to the pair of nucleotides in the same base pair ( @xmath38 ) , in the neighboring base pairs ( @xmath39 ) , in the next neighboring bases pairs ( @xmath40 ) , and in the next - next neighboring bases in another strand ( @xmath41 ) ( fig . 1(b ) ) . the elastic coefficient @xmath42 ( @xmath19 $ ] ) is a phenomenological constant that is assumed to be the same for all interacting pairs . this model is considered as a simplified version of previously proposed one - site - per - nucleotide models@xcite . an overview of the theory of normal - mode analysis is provided in several recent studies@xcite . thus , we here briefly show the results of this analysis . for this analysis , we define @xmath43 ( @xmath44 , @xmath45 ) as a @xmath46-dimensional position vector , and @xmath47 as the position vector of the basic structure . here , @xmath48 for the aaenm and @xmath49 for the cgenm . the motions of small deviations of @xmath43 from @xmath47 , @xmath50 obey @xmath51 where @xmath52 and @xmath53 ( @xmath54 ) are the @xmath55-th largest eigenvalue and its eigenvector of the @xmath56 hessian matrix @xmath57 as @xmath58 we assume that the system is in thermodynamic equilibrium with temperature @xmath59 . thus , the amplitude @xmath60 is given as @xmath61 with boltzmann constant @xmath62 $ ] . using this solution , the mean square fluctuation of the @xmath3-th atom in the aaenm ( @xmath63 ) is obtained as @xmath64 with boltzmann constant , and the temperature factor is displayed as @xmath65 . here , @xmath66 represents the temporal average . for the cgenm , we define the mean square fluctuation of the @xmath67-th nucleotide ( @xmath68 ) as @xmath69 . to consider the motion of the @xmath67-th nucleotide in the aaenm , we define the average nucleotide motions as @xmath70 . using this vector , we define the mean square fluctuation of the @xmath67-th nucleotide as @xmath71 . for the motif @xmath72 of the @xmath67-th nucleotide ( @xmath73 sugar , base , and phosphoric acid ) , the average temperature factor of the motif ( @xmath74 ) is defined as the average of @xmath75 belonging to each motif of each nucleotide , @xmath76 . in order to evaluate the validity of the aaenm , we measured the correlation coefficient between the profile of the temperature factor obtained from pdb data ( via x - ray crystal structure analysis ) and that obtained from the aaenm based on this crystal structure . it is noted that the temperature factor profiles for some of the pdb data often include unnaturally large or small values . thus , the correlation coefficients were estimated using data excluding such outliers . in the present evaluations , the value @xmath77 was considered as an outlier if @xmath78 , where @xmath79 and @xmath80 are the average and standard deviation of @xmath81 , respectively , and @xmath82 is used based on the standard arguments of statistics . we also focused on the relationships between the fluctuations of each nucleotide in the aaenms and cgenms in the directions parallel to the base pair axis , parallel to the helix axis , and vertical to both the base pair and helix axes . here , we name the nucleotides in one and the other strand constructing the @xmath3-th base pair as the @xmath3-th and @xmath37-th nucleotide . we define the position vectors of the c1 atoms belonging to the @xmath3-th and @xmath37-th nucleotides as @xmath83 and @xmath84 , and consider @xmath85 @xmath86 and @xmath87 ( fig . it is noted that @xmath88 and @xmath89 are not orthogonal in general ; however , we confirmed that the angles of these vectors were always sufficiently close to @xmath90 rad for each @xmath3 . using these vectors , the mean square fluctuations of the @xmath3-th and @xmath37-th nucleotides of the aaenm in the directions parallel to the base pair axis are defined as @xmath91 and @xmath92 , those parallel to the helix axis are defined as @xmath93 and @xmath94 , and those in the torsional direction are defined as @xmath95 and @xmath96 , respectively . the mean square fluctuations of the @xmath3-th and @xmath37-th nucleotides of the cgenm in these same directions are obtained by @xmath97 and @xmath98 , @xmath99 and @xmath100 , and @xmath101 and @xmath102 , respectively . moreover , we consider the mean square fluctuation of the relative base position of each base pair of the cgenm in the three directions listed above given by @xmath103 , @xmath104 , and @xmath105 , respectively . the overall geometry of each modeled dna molecule is characterized by the ratios among the square root of the three principal components of the populations of atoms , @xmath106 , @xmath107 , and @xmath108 ( @xmath109 ) . here , @xmath110 , @xmath111 , and @xmath112 are obtained as eigenvalues of the covariant matrix @xmath113 where @xmath114 , @xmath115 is the position of @xmath3-th atom , @xmath116 is the position of the center of mass of a given dna molecule , and @xmath117 indicates the average for all @xmath3s . we evaluate the overall geometry of a given dna molecule using the linearity @xmath118 and the line symmetry with respect to the @xmath110-axis @xmath119 . here , @xmath120 is large when the dna is straightened , and @xmath121 is large ( small ) when the dna forms a three ( two)-dimensional curve with wide ( flat ) envelope . we employ pearson s correlation coefficients , @xmath122 , to evaluate the correlations among the profiles of temperature factors and several anisotropic fluctuations of atoms obtained by aaenm , cgenm , and experiments . the fluctuations of atoms of the aaenms of several short dna sequences were measured with normal - mode analysis . here , the basic structures of the dnas are given by the crystal structures of the naked dnas with the following pdb ids : 1bna , 7bna , 9bna , 1d91 , 1dc0 , 122d , 123d , 181d , and 330d ( table 1@xcite ) . to confirm the validities of the aaenms , the correlations between the distribution profiles of the temperature factor of atoms ( @xmath123 ) and the average temperature factor of the motifs ( @xmath124 ) in the crystal structures and those in the corresponding aaenms were measured . in the following arguments , we employ the optimal values of @xmath18 , @xmath125 , and @xmath26 for each crystal structure ( table 1 ) , which were manually found to maximize @xmath122 of @xmath123 between the results of the crystal structure analysis and those of the aaenm . by choosing the appropriate conditions for the atoms of the two edge base pairs and water molecules ( @xmath125 and @xmath126 ) for each pdb i d , @xmath123 of each aaenm exhibited a similar profile to that of the crystal structure with a significant correlation coefficient @xmath122 ( fig . 2(a ) , table 1 , and ) . therefore , the aaenm seemed to reproduce the overall structure of the temperature factor profile of each crystal structure well , although the detailed profiles among atoms showed some deviations . furthermore , we focused on the average temperature factor for the motifs , @xmath124 , as recently discussed @xcite . we found that the @xmath124 obtained with the aaenms could accurately reproduce those of the corresponding crystal structures with high correlation coefficients @xmath122 , where @xmath127 was obtained for most cases ( fig . 2(b ) , table 1 , and ) . these results demonstrate that the aaenm is a suitable model for describing the sequence - dependent fluctuations of the nucleotide motifs of several double - stranded dna sequences , despite the simplicity of model construction and its easy implementation . moreover , these results show that the sequence - dependent forms of dna have a dominant contribution to their overall flexibilities and fluctuations . nevertheless , the present aaenm could not accurately reproduce @xmath124 well for some crystal structures . these deviations are considered to have arisen from the following primary assumption : we only considered the effects of the restrictions by the packing of dnas in crystal form for two edge base pairs and water , whereas such effects have an influence on all atoms . therefore , obtaining and incorporating more detailed knowledge of the restrictions of bulk sequences caused by crystal packing should help to achieve a more accurate reproduction of the molecular fluctuations for all cases . ) and ( b ) average temperature factor of the motifs ( @xmath124 ) obtained by the aaenms ( black curve ) and crystal structure analysis ( csa , gray ( red ) curve ) of typical double - stranded dna ( pdb i d : 1bna ) . here , @xmath128 $ ] , @xmath129 , and @xmath130 . atom and motif indices in ( a ) and ( b ) are given in the same order as shown for ( c ) . @xmath122 indicates the pearson correlation coefficient of the profiles between the two curves.,width=302 ] the main objective of this study was to unveil the sequence - dependent dynamic correlated motions of several long dna sequences . we next describe these dynamics of dna sequences with longer lengths than considered in the previous subsection . for this purpose , we also constructed a coarse - grained model , which is often useful for focusing on the slow and large - scale dynamics of molecules that essentially influence their function . thus , to propose a coarse - grained model of long double - stranded dna , we evaluated whether the cgenm proposed provides an appropriate coarse - grained model of the present aaenm . we performed normal - mode analysis of the aaenm and corresponding cgenm for @xmath131 randomly chosen @xmath132-bp sequences , and compared the mean square fluctuations of the @xmath3-th nucleotide ( @xmath133 and @xmath134 ) in the directions parallel to the base pair axis ( @xmath135 and @xmath136 ) , parallel to the helix axis ( @xmath137 and @xmath138 ) , and in the torsional direction ( @xmath139 and @xmath140 ) . here , @xmath141 is employed , as in the previous study@xcite , and @xmath142 is assumed , which was manually found to provide the best fit of fluctuation profiles between aaenm and cgenm . here , the overall fluctuation profiles of cgenm are independent of the value of @xmath42 since @xmath42 influences only on the absolute values of fluctuations . independent of the sequences and employed helical parameters , we found that the fluctuations of each nucleotide in several directions were very similar between the aaenms and cgenms when these models are constructed with the same helical parameters , with average correlation coefficients @xmath122 @xmath143 ( fig . 3 , table 3 , , , and ) . -bp random dna sequence ( 5 - gaggctaaagtctatttagaccggagttgacgtggaagcccggctagtct - 3 ) . ( a ) @xmath133 and @xmath134 , ( b ) @xmath135 and @xmath136 , ( c ) @xmath137 and @xmath138 , and ( d ) @xmath139 and @xmath140 . helical parameter set ( i ) ( table 2 ) was used for both models . the nucleotide indices in ( a ) to ( d ) are given in the same order shown for ( e ) . @xmath122 indicates the pearson correlation coefficient of the profiles between the two curves.,title="fig:",width=226 ] -bp random dna sequence ( 5 - gaggctaaagtctatttagaccggagttgacgtggaagcccggctagtct - 3 ) . ( a ) @xmath133 and @xmath134 , ( b ) @xmath135 and @xmath136 , ( c ) @xmath137 and @xmath138 , and ( d ) @xmath139 and @xmath140 . helical parameter set ( i ) ( table 2 ) was used for both models . the nucleotide indices in ( a ) to ( d ) are given in the same order shown for ( e ) . @xmath122 indicates the pearson correlation coefficient of the profiles between the two curves.,title="fig:",width=226 ] -bp random dna sequence ( 5 - gaggctaaagtctatttagaccggagttgacgtggaagcccggctagtct - 3 ) . ( a ) @xmath133 and @xmath134 , ( b ) @xmath135 and @xmath136 , ( c ) @xmath137 and @xmath138 , and ( d ) @xmath139 and @xmath140 . helical parameter set ( i ) ( table 2 ) was used for both models . the nucleotide indices in ( a ) to ( d ) are given in the same order shown for ( e ) . @xmath122 indicates the pearson correlation coefficient of the profiles between the two curves.,title="fig:",width=226 ] -bp random dna sequence ( 5 - gaggctaaagtctatttagaccggagttgacgtggaagcccggctagtct - 3 ) . ( a ) @xmath133 and @xmath134 , ( b ) @xmath135 and @xmath136 , ( c ) @xmath137 and @xmath138 , and ( d ) @xmath139 and @xmath140 . helical parameter set ( i ) ( table 2 ) was used for both models . the nucleotide indices in ( a ) to ( d ) are given in the same order shown for ( e ) . @xmath122 indicates the pearson correlation coefficient of the profiles between the two curves.,title="fig:",width=226 ] -bp random dna sequence ( 5 - gaggctaaagtctatttagaccggagttgacgtggaagcccggctagtct - 3 ) . ( a ) @xmath133 and @xmath134 , ( b ) @xmath135 and @xmath136 , ( c ) @xmath137 and @xmath138 , and ( d ) @xmath139 and @xmath140 . helical parameter set ( i ) ( table 2 ) was used for both models . the nucleotide indices in ( a ) to ( d ) are given in the same order shown for ( e ) . @xmath122 indicates the pearson correlation coefficient of the profiles between the two curves.,title="fig:",width=226 ] it is noted that the present cgenm contains only one node per nucleotide , whereas the aaenm contains @xmath144 nodes ( atoms ) per nucleotide . this fact demonstrates that the computational costs of the cgenms are much lower than those of the aaenms , although the accuracies of the obtained statistical aspects are almost identical between these two models . thus , this cgenm could be used for exhaustive analysis and comparisons of the dynamic features of several sequence - dependent dnas related to protein binding affinities , functions of transcription regulation sequences , and nucleosome positioning@xcite . in the next subsection , we provide an example of such an analysis to determine the relationships between the nucleosome - forming abilities of several double - stranded dna sequences and their inter - strand dynamic features . nucleosome positioning is important not only for compacting dna but also for appropriate gene regulation . several recent studies have been performed for genome - wide nucleosome mapping and the identifications and predictions of nucleosome - forming and -inhibiting sequences for some model organisms@xcite . as an example of the applications of the cgenm to an exhaustive analysis of the sequence - dependent behavior of dna , we compared the dynamic features of dna sequences of @xmath0 bp that were predicted as nucleosome - forming or nucleosome - inhibiting sequences in the genome of budding yeast _ saccharomyces cerevisiae _ ( 5000 forming sequences and 5000 inhibiting sequences of @xmath145 bp)@xcite , nematode _ caenorhabditis elegans _ ( 2567 forming sequences and 2608 inhibiting sequences of @xmath146 bp ) , _ drosophila melanogaster _ ( 2900 forming sequences and 2850 inhibiting sequences of @xmath146 bp ) , and _ homo sapiens _ ( 2273 forming sequences and 2300 inhibiting sequences of @xmath146 bp)@xcite . the histograms of the average relative fluctuations of dnas for the three directions @xmath147 , @xmath148 , @xmath149 , @xmath150 , @xmath151 , @xmath152 , @xmath153 , and @xmath154 ( @xmath117 indicates the average for all @xmath3s . ) for the nucleosome - forming sequences and the nucleosome - inhibiting sequences are shown in fig . 4 and @xmath155 . here , we employed the helical parameter set ( i ) ( table 2 ) that was used in the coarse - grained molecular dynamics simulations by freeman et al . , which exhibited consistent results to some experiments@xcite . the histograms for budding yeast showed that that nucleosome - forming sequences tend to exhibit larger fluctuations in several directions compared to the inhibiting sequences . in particular , the histogram of @xmath152 for the nucleosome - forming sequences showed a clear shift in the direction toward larger values compared to that for the nucleosome - inhibiting sequences ( fig . 4 ) . for the other organisms , most of the histograms showed few differences between the nucleosome- forming and -inhibiting sequences . however , similar to the case of yeast , the distribution of @xmath152 for the nucleosome - forming sequences shifted largely in the direction of larger values compared to that for the nucleosome - inhibiting sequences in these organisms ( @xmath155 ) . these results indicate that the nucleosome - forming ability is highly correlated to the fluctuations of the inter - strand distances of dnas , in which sequences with larger fluctuations tend to form the nucleosome . and @xmath151 , ( b ) @xmath156 and @xmath152 , ( c ) @xmath157 and @xmath153 , and ( d ) @xmath158 , and @xmath154 , for nucleosome - forming and -inhibiting sequences of budding yeast_saccharomyces cerevisiae _ ( @xmath145 bp ) . helical parameter set ( i ) ( table 2 ) was used.,title="fig:",width=453 ] and @xmath151 , ( b ) @xmath156 and @xmath152 , ( c ) @xmath157 and @xmath153 , and ( d ) @xmath158 , and @xmath154 , for nucleosome - forming and -inhibiting sequences of budding yeast_saccharomyces cerevisiae _ ( @xmath145 bp ) . helical parameter set ( i ) ( table 2 ) was used.,title="fig:",width=453 ] and @xmath151 , ( b ) @xmath156 and @xmath152 , ( c ) @xmath157 and @xmath153 , and ( d ) @xmath158 , and @xmath154 , for nucleosome - forming and -inhibiting sequences of budding yeast_saccharomyces cerevisiae _ ( @xmath145 bp ) . helical parameter set ( i ) ( table 2 ) was used.,title="fig:",width=453 ] and @xmath151 , ( b ) @xmath156 and @xmath152 , ( c ) @xmath157 and @xmath153 , and ( d ) @xmath158 , and @xmath154 , for nucleosome - forming and -inhibiting sequences of budding yeast_saccharomyces cerevisiae _ ( @xmath145 bp ) . helical parameter set ( i ) ( table 2 ) was used.,title="fig:",width=453 ] similar to the previous arguments , we compared the dynamical features among the random @xmath145-bp dna sequences varying in average gc - contents ; the histograms , averages , and standard deviations of @xmath147 , @xmath151 , @xmath152 , @xmath153 , and @xmath154 were measured from 10,000 random sequences for each gc content ( fig . 5 and ) . in this case , @xmath147 exhibited a minimum at a gc content of @xmath159 , which indicates that the at - rich sequences tend to be more rigid than the gc - rich sequences . however , the fluctuations of sequences consisting of only a or t were as large as those of the gc - rich sequences . moreover , the gc content dependencies of @xmath151 , @xmath152 , @xmath153 , and @xmath154 showed different characteristics for gc contents larger or smaller than @xmath160 . in particular , the results for @xmath152 were similar for cases with a large gc content ( @xmath161 ) but monotonically decreased with a gc content with little variance . a recent experimental study showed that the probability of nucleosome formation tends to increase with increases in the gc content ratio @xcite . thus , the present results indicate that sequences with larger fluctuations of inter - strand distances tend to form the nucleosome , which is consistent with the results described above from the analysis of the genomes of the four model organisms . and ( b ) @xmath152 , and ( c ) averages and standard deviations of @xmath147 , @xmath151 , @xmath152 , @xmath153 , and @xmath154 for 10,000 samples of random @xmath145-bp sequences with an average gc content @xmath162 , @xmath163 , @xmath164 , @xmath165 , and @xmath166 . helical parameter set ( i ) ( table 2 ) was used.,title="fig:",width=377 ] and ( b ) @xmath152 , and ( c ) averages and standard deviations of @xmath147 , @xmath151 , @xmath152 , @xmath153 , and @xmath154 for 10,000 samples of random @xmath145-bp sequences with an average gc content @xmath162 , @xmath163 , @xmath164 , @xmath165 , and @xmath166 . helical parameter set ( i ) ( table 2 ) was used.,title="fig:",width=377 ] and ( b ) @xmath152 , and ( c ) averages and standard deviations of @xmath147 , @xmath151 , @xmath152 , @xmath153 , and @xmath154 for 10,000 samples of random @xmath145-bp sequences with an average gc content @xmath162 , @xmath163 , @xmath164 , @xmath165 , and @xmath166 . helical parameter set ( i ) ( table 2 ) was used.,title="fig:",width=453 ] finally , we focus on the relationships between the overall geometries and fluctuations of the considered dna sequences . the overall geometries of several dna sequences , the nucleosome - forming and -inhibiting dna sequences for four model organisms and random dna sequences with different gc contents , were evaluated using scatter plots of @xmath120 ( linearity ) and @xmath121 ( line symmetry ) ( fig . 6 and ) . it is noted that @xmath120 and @xmath121 are highly correlated . the average , standard deviation , and distributions of @xmath120 and @xmath121 exhibited slight but not significant deviations between the nucleosome - forming and -inhibiting sequences . for the random sequences , the average value of @xmath120 exhibited similar variations to @xmath147 with an increase in gc content . in particular , both values decreased with increases in gc content for gc contents @xmath167 , whereas they increased with increases in gc content for gc contents @xmath168 ( fig . 5(c ) and fig . 6(c ) ) . in fact , @xmath120 and @xmath169 were highly correlated for random dna sequences , regardless of their cg content . the pearson correlation coefficient for the 110,000 sequences analyzed above with gc contents @xmath170 was @xmath171 . it is noted that @xmath120 showed large dispersion for each gc content , and there were significant overlaps among @xmath120 distributions with different gc contents ( fig 6(b ) and 6(c ) ) . this fact indicates that different dna sequences can often show similar geometries , and such sequences also tend to show similar overall fluctuations . on the other hand , the fluctuations of inter - strand distances @xmath172 that may correlate to the nucleosome - forming ability did not correlate significantly to either @xmath120 or @xmath121 , with pearson correlation coefficients of @xmath173 and @xmath174 this fact indicates that the nucleosome - forming ability of dna sequences are not only determined by the overall dna geometries but also by their dynamic properties . and @xmath121 for ( a ) nucleosome - forming and -inhibiting sequences of budding yeast_saccharomyces cerevisiae _ ( @xmath145 bp ) , and for ( b ) random sequences with several gc contents ( @xmath145 bp ) . ( c ) averages and standard deviations of @xmath120 for 10000 samples of random @xmath145-bp sequences with an average gc content @xmath162 , @xmath163 , @xmath164 , @xmath165 , and @xmath166 . helical parameter set ( i ) ( table 2 ) was used.,title="fig:",width=453 ] and @xmath121 for ( a ) nucleosome - forming and -inhibiting sequences of budding yeast_saccharomyces cerevisiae _ ( @xmath145 bp ) , and for ( b ) random sequences with several gc contents ( @xmath145 bp ) . ( c ) averages and standard deviations of @xmath120 for 10000 samples of random @xmath145-bp sequences with an average gc content @xmath162 , @xmath163 , @xmath164 , @xmath165 , and @xmath166 . helical parameter set ( i ) ( table 2 ) was used.,title="fig:",width=453 ] in this study , simple elastic network models of double - stranded dnas were developed in order to perform an exhaustive analysis of several sequence - dependent dynamical features . first , we constructed a simple all - atom elastic network model that could reproduce the fluctuations of the motifs of each nucleotide ( sugar , phosphoric acid , and bases ) of several crystal structures of short dna sequences . second , we proposed a simple course - grained elastic network model that could reproduce the dynamic features of the long dna sequences obtained by the all - atom elastic network model . through exhaustive analysis of the dynamic features of several dna sequences with normal - mode analysis of the presented coarse - grained elastic network model , we found that the dynamic aspects of dna are highly influenced by the properties of nucleotide sequence such as gc content . we also found that the nucleosome - forming abilities of double - stranded dna exhibited positive correlations with their sequence - dependent inter - strand fluctuations . in the present study , we demonstrated the sequence - dependent dynamic features for several @xmath0-bp dna sequences to evaluate the relationships between the nucleosome - forming abilities and dna dynamics . of course , dna sequences longer than @xmath0 bp can also be analyzed using the presented course - grained model . moreover , course - grained molecular dynamics simulations can be performed to consider the large deformations of dna , such as formation of a super helix and nucleosome that are the basic structures of higher - order chromosome architectures , using the presented elastic network models with the excluded effect of the volume of each atom or each nucleotide . we are currently conducting these molecular dynamics simulations , and the results will be reported in the future . we did not consider the effects of solvents such as temperature and salt concentrations in the present elastic network models . therefore , we are also planning to attempt modifications of the models so that several solvent conditions can be incorporated in future work . the author is grateful to s. tate and y. murayama for fruitful discussions . conceived the study : aa , ns , hn . designed the models : ns , aa , si . performed the analysis : si , aa wrote the paper : aa , ns , hn . 10 tirion mm ( 1996 ) large amplitude elastic motions in proteins from a single - parameter , atomic analysis . phys rev lett 77 : 1905 - 1908 . doi : http://dx.doi.org/10.1103/physrevlett.77.1905 hinsen k , petrescu aj , dellerue s , bellisent - funel mc , kneller gr ( 2000 ) harmonicity in slow protein dynamics . chem phys 261 : 25 - 37 . doi:10.1016/s0301 - 0104(00)00222 - 6 atilgan ar , durell sr , jernigan rl , demirel mc , keskin o , et al . ( 2001 ) anisotropy of fluctuation dynamics of proteins with an elastic network model . biophys j 80 : 505 - 515 . doi:10.1016/s0006 - 3495(01)76033-x bahar i , rader aj ( 2005 ) coarse - grained normal mode analysis in structural biology . curr opin struct biol 15 : 586 - 592 . doi:10.1016/j.sbi.2005.08.007 bahar i , lezon tr , yang lw , eyal e ( 2010 ) global dynamics of proteins : bridging between structure and function . annu rev biophys 39 : 23 - 42 . doi : 10.1146/annurev.biophys.093008.131258 bahar i , lezon tr , bakan a , shrivastava ih ( 2010 ) normal mode analysis of biomolecular structures : functional mechanisms of membrane proteins . chem rev 110 : 1463 - 1497 . dykeman ec , sankey of ( 2010 ) normal mode analysis and applications in biological physics . j phys condens matter 22 : 423202 . flechsig h , mikhailov as ( 2010 ) tracing entire operation cycles of molecular motor hepatitis c virus helicase in structurally resolved dynamical simulations . proc natl acad sci ( usa ) 107 : 20875 - 20880 . atilgan c , okan ob , atilgan ar ( 2012 ) network - based models as tools hinting at nonevident protein functionality . annu rev biophys 41 : 205 - 25 . van wynsberghe aw , cui q ( 2006 ) interpreting correlated motions using normal mode analysis . structure 14 : 1647 - 1653 . doi:10.1016/j.str.2006.09.003 tama f , brooks cl ( 2006 ) symmetry , form , and shape : guiding principles for robustness in macromolecular machines . annu rev biophys biomol struct 35 : 115 - 33 . 0.1146/annurev.biophys.35.040405.102010 moritsugu k , smith jc ( 2007 ) coarse - grained biomolecular simulation with reach : realistic extension algorithm via covariance hessian . biophys j 93 : 3460 - 3469 . doi:10.1529/biophysj.107.111898 yang l , song g , jernigan rl ( 2009 ) protein elastic network models and the ranges of cooperativity . proc natl acad sci ( usa ) 106 : 12347 - 12352 . doi : 10.1073/pnas.0902159106 dehouck y , mikhailov as ( 2013 ) effective harmonic potentials : insights into the internal cooperativity and sequence - specificity of protein dynamics . plos comp bio 9 : e1003209 albert b , et al . ( 2008 ) molecular biology of the cell 5th edition . garland science , taylor & francis group , llc . latchman ds ( 2010 ) gene control . garland science , taylor & francis group , llc . fukue y , sumida n , nishikawa j , ohyama t ( 2004 ) core promoter elements of eukaryotic genes have a highly distinctive mechanical property . nucleic acids res 32 : 5834 - 5840 . doi : 10.1093/nar / gkh905 ohyama t ( 2005 ) dna conformation and transcription . new york , ny , usa : springer , landes bioscience . valenzuela l , kamakaka rt ( 2006 ) chromatin insulators . annu rev genet 40 : 107 - 138 . doi : 10.1146/annurev.genet.39.073003.113546 bushey am , dorman er , corces vg ( 2008 ) chromatin insulators : regulatory mechanisms and epigenetic inheritance . mol cell 32 : 1 - 9 . doi:10.1016/j.molcel.2008.08.017 geyer pk , clark i ( 2002 ) protecting against promiscuity : the regulatory role of insulators . cell mol life sci 59 : 2112 - 2127 . doi:1420 - 682x/02/122112 - 16 gaszner m , felsenfeld g ( 2006 ) insulators : exploiting transcriptional and epigenetic mechanisms . nat rev genet 7 : 703 - 713 . doi:10.1038/nrg1925 raab jr , kamakaka rt ( 2010 ) insulators and promoters : closer than we think . nat rev genet 11 : 439 - 446 . takagi h , inai y , watanabe s , tatemoto s , yajima m , akasaka k , et al . ( 2012 ) nucleosome exclusion from the interspecies - conserved central at - rich region of the ars insulator . j biochem 151 : 75 - 87 . doi : 10.1093/jb / mvr118 turner bm ( 2001 ) chromatin and gene regulation . blackwell science ltd . dekker j , rippe k , dekker m , and kleckner n ( 2002 ) capturing chromosome conformation . science 295 : 1306 1311 . doi : 10.1126/science.1067799 tolhuis b , palstra rj , splinter e , grosveld f , de laat w ( 2002 ) looping and interaction between hypersensitive sites in the active beta - globin locus . mol cell 10 : 1453 - 1465 . doi:10.1016/s1097 - 2765(02)00781 - 5 allis cd et al . ( 2007 ) epigenetics . cold spring harbor laboratory press . dixon jr , selvaraj s , yue f , kim a , li y , shen y , et al . ( 2012 ) topological domains in mammalian genomes identified by analysis of chromatin interactions . nature 485 : 376 - 380 . doi:10.1038/nature11082 apostolou e , ferrari f , walsh rm , bar - nur o , stadtfeld m , cheloufi s , et al . ( 2013 ) genome - wide chromatin interactions of the nanog locus in pluripotency , differentiation , and reprogramming . cell stem cell 13 : 1 - 14 . doi:10.1016/j.stem.2013.04.013 gibcus jh , dekker j ( 2013 ) the hierarchy of the 3d genome . mol cell 49 : 773 - 782 . doi:10.1016/j.molcel.2013.02.011 dekker j , marti - renom ma , mirny la ( 2013 ) exploring the three - dimensional organization of genomes : interpreting chromatin interaction data . nat rev gen 14 : 390 - 403 . doi:10.1038/nrg3454 bickmore wa , van steensel b ( 2013 ) genome architecture : domain organization of interphase chromosomes . cell 152 : 1270 - 1284 . doi:10.1016/j.cell.2013.02.001 lin d , matsumoto a , go n ( 1997 ) normal mode analysis of a double - stranded dna dodecamer d(cgcgaattcgcg ) . j chem phys 107 : 3684 - 3690 . doi : 10.1063/1.474724 matsumoto a , olson wk ( 2002 ) sequence - dependent motions of dna : a normal mode analysis at the base - pair level . biophys j 83 : 22 - 41 . doi:10.1016/s0006 - 3495(02)75147 - 3 lankas f , sponer j , langowski j , cheatham te 3rd ( 2003 ) dna basepair step deformability inferred from molecular dynamics simulations . biophys j 85 : 2872 - 2883 . doi:10.1016/s0006 - 3495(03)74710 - 9 fujii s , kono h , takenaka s , go n , sarai a ( 2007 ) sequence - dependent dna deformability studied using molecular dynamics simulations . nucl acids res 35 : 6063 - 6074 . doi : 10.1093/nar / gkm627 orozco m , noy a , p@xmath175rez a ( 2008 ) recent advances in the study of nucleic acid flexibility by molecular dynamics . curr opin struct biol 18 : 185 - 193 . doi:10.1016/j.sbi.2008.01.005 bomble yj , case da ( 2008 ) multiscale modeling of nucleic acids : insights into dna flexibility . biopolymers 89 : 722 - 731 . doi:10.1002/bip.21000 beveridge dl , cheatham te 3rd , mezei m ( 2012 ) the abcs of molecular dynamics simulations on b - dna , circa 2012 . j biosci 37 : 379 - 397 . doi:10.1007/s12038 - 012 - 9222 - 6 cheatham te 3rd , case da ( 2013 ) twenty - five years of nucleic acid simulations . biopolymers 99 : 969 - 977 . doi:10.1002/bip.22331 pasi m , maddocks jh , beveridge d , bishop tc , case da , cheathm t 3rd , et al . ( 2014 ) abc : a systematic microsecond molecular dynamics study of tetranucleotide sequence effects in b - dna . nucl acids res 42 : 12272 - 12283 . doi : 10.1093/nar / gku855 knotts iv ta , rathore n , schwartz dc , de pablo jj ( 2007 ) a coarse grain model for dna . j chem phys 126 : 084901 . doi : 10.1063/1.2431804 morriss - andrews a , rottler j , plotkin , ss ( 2010 ) . a systematically coarse - grained model for dna and its predictions for persistence length , stacking , twist , and chirality . j chem phys 132 : 035105 . doi : 10.1063/1.326999 dans pd , zida a , machado mr , pantano s ( 2010 ) a coarse grained model for atomic - detailed dna simulations with explicit electrostatics . j chem theor comp 6 : 1711 - 1725 . doi:10.1021/ct900653p freeman gs , hinckley dm , de pablo jj ( 2011 ) a coarse - grain three - site - per - nucleotide model for dna with explicit ions . j chem phys 135 : 165104 . doi : 10.1063/1.3652956 freeman gs , lequieu jp , hinckley dm , whitmer jk , de pablo jj ( 2014 ) dna shape dominates sequence affinity in nucleosome formation . phys rev lett 113 : 168101 . doi : http://dx.doi.org/10.1103/physrevlett.113.168101 freeman gs , hinckley dm , lequieu jp , whitmer jk , pablo jj ( 2014 ) coarse - grained modeling of dna curvature . j chem phys 141 : 165103 . doi : 10.1063/1.4897649 pinamonti g , bottaro s , micheletti c , bussi g ( 2015 ) elastic network models for rna : a comparative assessment with molecular dynamics and shape experiments . nucleic acids res . doi : 10.1093/nar / gkv708 savelyev a , papoian ga ( 2010 ) chemically accurate coarse graining of double - stranded dna . proc nat acad sci 107 : 20340 - 20345 . doi : 10.1073/pnas.1001163107 setny p , zacharias m ( 2013 ) elastic network models of nucleic acids flexibility . j chem theor comp 9 : 5460 - 5470 . doi : 10.1021/ct400814n gonzalez o , petkeviciute d , maddocks jhj ( 2013 ) a sequence - dependent rigid - base model of dna . j chem phys 138 : 055102 . doi : 10.1063/1.4789411 naome a , laaksonen a , vercauteren dp ( 2014 ) a solvent - mediated coarse - grained model of dna derived with the systematic newton inversion method . j chem theor comp 10 : 3541 - 3549 . doi:10.1021/ct500222s korolev n , luo d , lyubartsev ap , nordensklold l ( 2014 ) a coarse - grained dna model parameterized from atomistic simulations by inverse monte carlo . polymers 6 : 1655 - 1675 . doi:10.3390/polym6061655 go@xmath176i jr , fenollosa c , perez a , torrents d , orozco m ( 2008 ) dnalive : a tool for the physical analysis of dna at the genomic scale . bioinformatics 24 : 1731 - 1732 . doi : 10.1093/bioinformatics stolz rc , bishop tc ( 2010 ) icm web : the interactive chromatin modeling web server . nucleic acids res 38 : w254-w261 . doi : 10.1093/nar / gkq496 chen w , lin h , feng pm , ding c , zuo yc , chou kc ( 2012 ) inuc - physchem : a sequence - based predictor for identifying nucleosomes via physicochemical properties . plos one 7 : e47843 . doi : 10.1371/journal.pone.0047843 hospital a , faustino i , collepardo - guevara r , gonzalez c , gelpi jl , orozco m ( 2013 ) naflex : a web server for the study of nucleic acid flexibility . nucleic acids res 41 : w47-w55 . doi : 10.1093/nar / gkt378 guo sh , deng ez , xu lq , ding h , lin h , chen w ( 2014 ) inuc - pseknc : a sequence - based predictor for predicting nucleosome positioning in genomes with pseudo k - tuple nucleotide composition . bioinformatics 30 : 1522 - 1529 . doi : 10.1093/bioinformatics / btu083 petkevici@xmath177t@xmath175 d , pasi m , gonzalez o , maddocks jh ( 2014 ) cgdna : a software package for the prediction of sequence - dependent coarse - grain free energies of b - form dna . nucleic acids res 42 : e153 . doi : 10.1093/nar / gku825 olson wk , gorin aa , lu xj , hock lm , zhurkin vb ( 1998 ) dna sequence - dependent deformability deduced from protein dna crystal complexes . proc nat acad sci 95 : 11163 - 11168 . doi : 10.1073/pnas.95.19.11163 olson wk , colasanti av , li y , ge w , zheng g , zhurkin vb ( 2006 ) dna simulation benchmarks as revealed by x - ray structures . in : computational studies of rna and dna . springer netherlands : 235 - 257 . morozov av , fortney k , gaykalova da , studitsky vm , widom j , siggia ed ( 2009 ) using dna mechanics to predict in vitro nucleosome positions and formation energies . nucleic acids res 37 : 4707 - 4722 . doi : 10.1093/nar / gkp475 perez a , lankas f , luque fj , orozco m ( 2008 ) towards a molecular dynamics consensus view of b - dna flexibility . nucleic acids res 36 : 2379 - 2394 . doi : 10.1093/nar / gkn082 lavery r , zakrzewska k , beveridge d , bishop tc , case da , cheatham t , et al . ( 2010 ) a systematic molecular dynamics study of nearest - neighbor effects on base pair and base pair step conformations and fluctuations in b - dna . nucleic acids res 38 : 299 - 313 . doi : 10.1093/nar / gkp834 dans pd , perez a , faustino i , lavery r , orozco m ( 2012 ) . exploring polymorphisms in b - dna helical conformations . nucleic acids res 40 : 10668 - 10678 . doi : 10.1093/nar / gks884 lu xj , olson wk ( 2003 ) 3dna : a software package for the analysis , rebuilding and visualization of three dimensional nucleic acid structures . nucleic acids res 31:5108 - 5121 . doi : 10.1093/nar / gkg680 drew hr , wing rm , takano t , broka c , tanaka s , itakura k , et al . ( 1981 ) structure of a b - dna dodecamer : conformation and dynamics . proc natl acad sci usa 78 : 2179 - 2183 . holbrook sr , dickerson re , kim sh ( 1985 ) . anisotropic thermal - parameter refinement of the dna dodecamer cgcgaattcgcg by the segmented rigid - body method . acta crystallogr b 41:255 - 262 . doi:10.1107/s0108768185002087 westhof e ( 1987 ) . re - refinement of the b - dodecamer d(cgcgaattcgcg ) with a comparative analysis of the solvent in it and in the z - hexamer d(5brcg5brcg5brcg ) . j biomol struct dyn 5:581 - 600 . doi:10.1080/07391102.1987.10506414 kneale g , brown t , kennard o , rabinovich d ( 1985 ) . g.t base - pairs in a dna helix : the crystal structure of d(g - g - g - g - t - c - c - c ) . j mol biol 186 : 805 - 814 . doi:10.1016/0022 - 2836(85)90398 - 5 ng hl , kopka ml , dickerson re ( 2000 ) . the structure of a stable intermediate in the a ; - > b dna helix transition . proc natl acad sci usa 97:2035 - 2039 . doi : 10.1073/pnas.040571197 hahn m , heinemann u ( 1993 ) . dna helix structure and refinement algorithm : comparison of models for d(ccaggcm5ctgg ) derived from nuclsq , tnt and x - plor . acta crystallogr d 49 : 468 - 477 . doi:10.1107/s0907444993004858 sadasivan c , gautham n ( 1995 ) . sequence - dependent microheterogeneity of z - dna : the crystal and molecular structures of d(cacgcg).d(cgcgtg ) and d(cgcacg).d(cgtgcg ) . j mol biol 248 : 918 - 930 . doi:10.1006/jmbi.1995.9894 timsit y , vilbois e , moras d ( 1991 ) . base - pairing shift in the major groove of ( ca)n tracts by b - dna crystal structures . nature 354 : 167 - 170 . doi:10.1038/354167a0 segal e , fondufe - mittendorf y , chen l , thastrom ac , field y , moore ik , et al . ( 2006 ) . a genomic code for nucleosome positioning . nature 442 : 772 - 778 . doi:10.1038/nature04979 lee w , tillo d , bray n , morse rh , davis rw , et al . ( 2007 ) a high - resolution atlas of nucleosome occupancy in yeast . nat genet 39 : 1235 - 1244 . doi:10.1038/ng2117 schones de , cui k , cuddapah s , roh ty , barski a , wang z , et al.(2008 ) dynamic regulation of nucleosome positioning in the human genome . cell 132 : 887 - 898 . doi:10.1016/j.cell.2008.02.022 mavrich tn , ioshikhes ip , venters bj , jiang c , tomsho lp , qi j , et al . ( 2008 ) a barrier nucleosome model for statistical positioning of nucleosomes throughout the yeast genome . genome res 18 : 1073 - 1083 . doi : 10.1101/gr.078261.108 mavrich tn , jiang c , ioshikhes ip , li x , venters bj , zanton sj , et al . ( 2008 ) nucleosome organization in the drosophila genome . nature 453 : 358 - 362 . doi:10.1038/nature06929 tillo d , hughes tr ( 2009 ) . g+ c content dominates intrinsic nucleosome occupancy . bmc bioinform 1 : 442 . doi:10.1186/1471 - 2105 - 10 - 442 * helical parameter sets . * ( a ) helical parameter sets ( ii ) obtained by x - ray crystal structure analysis,@xcite , and ( b ) helical parameter sets ( iii ) obtained by all - atom molecular dynamics simulations.@xcite . * comparisons between cgenm and aaenm . * average and standard deviation of the correlation coefficients of 500 samples of random @xmath132-bp sequences between @xmath133 and @xmath134 , @xmath135 and @xmath136 , @xmath137 , and @xmath138 , and @xmath139 and @xmath140 . helical parameter sets ( ii ) and ( iii ) ( ) were used . * temperature factor of each atom . * the temperature factor of each atom obtained by the aaenms ( black curve ) and x - ray crystal structure analysis ( gray ( red ) curve ) of typical double - stranded dnas obtained from pdb i d ( a ) 7bna , ( b ) 9bna , ( c ) 1d91 , ( d ) 1dc0 , ( e ) 122d , ( f ) 123d , ( g ) 181d , and ( h ) 330d . parameters @xmath18 , @xmath125 , and @xmath126 are given in . * average temperature factor of motifs . * average temperature factor of motifs obtained by the aaenms ( black curve ) and x - ray crystal structure analysis ( gray ( red ) curve ) of typical double - stranded dnas obtained from pdb i d ( a ) 7bna , ( b ) 9bna , ( c ) 1d91 , ( d ) 1dc0 , ( e ) 122d , ( f ) 123d , ( g ) 181d , and ( h ) 330d . parameters @xmath18 , @xmath125 , and @xmath126 are given in . * comparisons between aaenm and cgenm . * comparisons of the fluctuations between each nucleotide in the aaenm ( black curves ) and cgenm ( gray ( red ) curves ) for a typical random @xmath132-bp dna sequence ( 5 - agtggtaaggcatggttctcgaatctcggtttatttacactgctgctcca - 3 ) . ( a ) @xmath133 and @xmath134 , ( b ) @xmath135 and @xmath136 , ( c ) @xmath137 and @xmath138 , and ( d ) @xmath139 and @xmath140 using helical parameter set ( ii ) . * comparisons between aaenm and cgenm . * comparisons of the fluctuations between each nucleotide in the aaenm ( black curves ) and cgenm ( gray ( red ) curves ) for a typical random @xmath132-bp dna sequence ( 5 - atatgctgtagagcgtcccgtccgcgcgttgtggttttttcggtgctcta - 3 ) . ( a ) @xmath133 and @xmath134 , ( b ) @xmath135 and @xmath136 , ( c ) @xmath137 and @xmath138 , and ( d ) @xmath139 and @xmath140 using helical parameter set ( iii ) . * histograms of the average fluctuations in_caenorhabditis elegans_. * histograms of the average fluctuations of ( a ) @xmath147 and @xmath151 , ( b ) @xmath156 and @xmath152 , ( c ) @xmath157 and @xmath153 , and ( d ) @xmath158 and @xmath154 for nucleosome - forming and nucleosome - inhibiting sequences of the nematode _ caenorhabditis elegans _ ( @xmath146 bp ) . helical parameter set ( i ) ( table 2 ) was used . * histograms of the average fluctuations in_drosophila melanogaster_. * histograms of the average fluctuations of ( a ) @xmath147 and @xmath151 , ( b ) @xmath156 and @xmath152 , ( c ) @xmath157 and @xmath153 , and ( d ) @xmath158 and @xmath154 for nucleosome - forming and nucleosome - inhibiting sequences of _ drosophila melanogaster _ ( @xmath146 bp ) . helical parameter set ( i ) ( table 2 ) was used . * histograms of the average fluctuations in_homo sapiens_. * histograms of the average fluctuations of ( a ) @xmath147 and @xmath151 , ( b ) @xmath156 and @xmath152 , ( c ) @xmath157 and @xmath153 , and ( d ) @xmath158 and @xmath154 for nucleosome - forming and nucleosome - inhibiting sequences of _ homo sapiens _ ( @xmath146 bp ) . helical parameter set ( i ) ( table 2 ) was used . * histograms of the average fluctuations with different gc contents . * histograms of ( a ) @xmath151 , ( b ) @xmath178 , and ( c ) @xmath179 for 10,000 samples of random @xmath145-bp sequences with different average gc contents . helical parameter set ( i ) ( table 2 ) was used . * overall geometries of long dna sequences in model organisms . * scatter plots of @xmath120 and @xmath121 for nucleosome - forming and -inhibiting sequences of ( a ) _ caenorhabditis elegans _ ( @xmath146 bp ) , ( b ) _ drosophila melanogaster _ ( @xmath146 bp ) , and ( c ) _ homo sapiens _ ( @xmath146 bp ) . helical parameter set ( i ) ( table 2 ) was used .

predictions for energy levels , production and decay rate of gluinonia , non - relativistic boundstates of gluinos , are presented . the potential between color - octet constituents is derived in next - to - next - to leading order and one - loop qcd corrections are derived for the production cross section and the decay rate into gluon jets . in addition we evaluate the decay rate into top quarks and into two photons . the signal - to - background ratio is estimated for the dominant decay mode and found to be around . for relatively light gluinos the bound states thus might be detectable . pacs numbers : 12.38.bx 12.60.jv 14.80.ly

0910.2612

with the turning - on of the lhc in the near future , one of the most important tasks of both general - purpose detectors atlas and cms will be the search for physics beyond the standard model , with one of the most promising candidates being supersymmetry . in the course of this program , which will obviously be adjusted and focused with increasing luminosity and with the first indications of `` new physics '' , the determination of the properties of newly detected particles will be a major task . indeed , only after the quantum numbers , the interaction and the mass of a new particle have been measured with high precision , this requirement can be considered fulfilled . in the present paper we want to concentrate on the gluino , the supersymmetric partner of the gluon , and discuss the possibility to determine its quantum numbers , i.e. its spin , color , majorana nature and its mass , through the investigation of non - relativistic boundstates of two gluinos , generally denoted `` gluinonia '' . the classification of gluinonia according to their quantum numbers and their qualitative properties , i.e. spectra and decay modes , as well as estimates of their production cross section at hadron colliders have been studied already more than two decades ago and presented in refs . @xcite concentrating at that time on gluino masses which are by now experimentally excluded . more recently gluinonia have been discussed in refs . hadronic transitions between different gluinonium levels and some rare annihilation decays were evaluated in @xcite . the production of vector and pseudoscalar states at the tevatron and the lhc was investigated in @xcite where it was argued that , given sufficiently good yet realistic jet mass resolution , pseudoscalar states could be detected for bound state masses up to 3 tev . a phenomenological study of the question , to which extent a signal for color - octet states could indeed be detected experimentally , has been performed in @xcite . in all these cases , however , the production cross section has been calculated in lowest order approximation only and a qualitative phenomenological potential for the gluino - gluino interaction has been employed . the topic has gained renewed interest in connection with split susy @xcite , ( see , e.g. , refs . @xcite ) . the detection of these bound states would provide important information about gluino properties , difficult to obtain through other means . since gluinonia only exist if gluinos do not decay too rapidly , say with decay rates less than a few gev , their observation would immediately provide at least an upper limit on @xmath1 . the energy spectrum of the gluinonium system , e.g. the mass difference between ground state and first radial excitation ( or open gluino production ) would be sensitive to the potential and thus to the color - octet nature of the constituents . the same is true for the production cross section which is proportional to the square of the wave function at the origin . gluinos are expected to decay through cascades into several jets plus the lightest supersymmetric particle ( lsp ) which escapes detection . for this reason the precise determination of its mass in direct gluino decays is limited by the missing mass or energy resolution . in contrast , once the mass of the boundstate and the dynamics of the interaction are known with sufficient precision , the mass of the constituents is fixed with the corresponding accuracy and in any case subject to a different systematic error . the energy levels of gluinonia were estimated already in the early papers employing phenomenological color - octet potentials , obtained from phenomenological quarkonium potentials by multiplication with the ratio of the corresponding casimir operators @xmath2 and the gluino mass was interpreted as an effective mass closely related to the pole mass . during recent years more refined studies have been performed to describe non - relativistic boundstates of color - triplet constituents . this includes the evaluation of the static two - loop potential @xcite , first steps toward the evaluation of the three - loop potential @xcite and the calculation of mass - suppressed terms proportional @xmath3 @xcite and @xmath4 @xcite . mass definitions better suited to control the long distance part of the potential have been introduced @xcite , and the evaluation of the spectra through analytic methods has been possible , using a perturbative approach to include higher order contributions from the potential @xcite . last but not least , finite - width effects were incorporated for top - quark production at electron - positron @xcite and hadron colliders @xcite . these new results can be applied also in studies of gluino bound states . in order to arrive at reliable predictions for the signals of gluinonium production and decay , the inclusion of a realistic qcd potential and next - to - leading order ( nlo ) corrections to boundstate production at the lhc is mandatory . considering and building on the significant progress in the perturbative treatment of non - relativistic boundstates , in particular quarkonium and positronium , predictions of similar quality can be made for boundstates with color - octet constituents . this is the main content of the present paper . we start with a brief overview and a discussion of the quantum numbers and qualitative properties of boundstates with majorana constituents and evaluate the qcd potential for a color - octet boundstate ( section [ sec : properties ] ) . subsequently , in section [ sec : spectroscopy ] , we calculate the spectrum and the wave function at the origin for the lowest - lying levels . these quantities are the necessary ingredients for a realistic prediction of gluinonium production and decay rates presented in section [ sec : production ] . assuming that squarks are relatively heavy , the dominant decay channel of the pseudoscalar singlet boundstate proceeds through two gluons , and it will be a difficult task to distinguish this signal from the irreducible background from the two - gluon jet continuum . in section [ sec : photons ] we consider as an alternative the decays into top quarks and two photons , which can proceed through virtual quark ( @xmath5 ) and squark ( @xmath6 ) intermediate states . one might , in principle , hope that the extremely clean @xmath7 signal sticks out of the irreducible background similarly to the higgs decay into two photons which has been identified as a promising signal significantly superior to the @xmath8 or two - gluon mode . similar considerations apply to the @xmath9 mode . in order that these modes can compete with the two - gluon channel , squarks should not be dramatically heavier than gluinos . on the other hand , to avoid dominance of the single - gluino decay , the decay mode @xmath10 should be kinematically forbidden , corresponding to @xmath11 . we therefore evaluate the corresponding decay rates and study if , given suitable choices for the respective masses , the two - photon mode might lead to a possible signal . in section [ sec : background ] a rough estimate of the dominant signal versus background is presented . section [ sec : conclusions ] contains our conclusions . let us start with a brief recapitulation of the quantum numbers of gluino boundstates and the corresponding color , spin and orbital momentum configurations @xcite . these differ from those of quarkonia due to the restrictions arising from the majorana nature of gluinos , and due to their different color assignment . two color - octet states can be combined into irreducible representations as follows @xmath12 where the index indicates the ( anti-)symmetry with respect to their color index . the interaction can be either attractive , repulsive or absent ( in lowest order ) . in lowest order the coefficient of the qcd potential is given by the expectation value of the product of the color generators @xmath13 , taken between two - particle states in the respective representation . this product , in turn , can be expressed by the eigenvalues of the quadratic casimir operator of the constituents , @xmath14 , and the boundstate in representation @xmath15 , @xmath16 : @xmath17=\frac{1}{2}\left(c_{{r}}-2c_{a}\right ) \ , . \label{fdotf}\end{aligned}\ ] ] the results are listed in tab . [ table1 ] . in the following we shall limit the discussion to the cases with negative coefficients , corresponding to attraction . .color interaction of two @xmath18 octets . [ cols="^,^,^,^",options="header " , ] the production rate of non - relativistic boundstates in hard collisions is proportional to the squared wave function at the origin ( for @xmath19 waves ) or its derivative ( for @xmath20 waves ) , the latter being significantly suppressed and of relative order @xmath21 . for this reason the following discussion will be limited to @xmath19 waves only . we , furthermore , anticipate that the experimental mass resolution and , eventually , the large natural width of the boundstate will lead to a sizable smearing of the `` narrow '' resonances . the signal to background ratio will be small and their detection difficult . for this reason we will consider color - singlet states only , where the level spacings and production rates are enhanced approximately by the square of the color coefficient listed in tab . [ table1 ] . let us now discuss the qcd potential . similarly to the case of heavy - quark boundstates it can be decomposed into the following two terms @xmath22 which are ordered according to their inverse powers in the constituent mass @xmath23 and are given in coordinate space in close similarity to the quarkonium potential ( see e.g. @xcite ) . for @xmath19 waves it reads @xmath24\biggr . \nonumber\\ & & \biggl.+\frac{\alpha_{s}(\mu^2)^2}{(4\pi)^2}\biggl[a_{2}+2\gamma_{e}\left(2 a_{1}\beta_{0}+\beta_{1}\right)+\beta_{0}^2\left(\frac{\pi^2}{3}+4\gamma_{e}^2\right)\biggr.\biggr . \nonumber\\ & & \biggl.\biggl.+\left(4a_{1}\beta_{0}+2\beta_{1}+8\gamma_{e}\beta_{0}^2\right)\ln(\mu r)+4\beta_{0}^2\ln(\mu r)^2\biggr]\biggr\ } \,,\nonumber\\ v_{\rm nc}(\vec{r}\,)&=&u_{1}(r)+u_{2}(\vec{r}\,)+u_{3}(r ) \,,\nonumber\\ u_{1}(r)&=&-\frac{\delta^2}{4m_{\tilde{g}}^3}+\frac{c_{a}\alpha_{s}(\mu^2)}{m_{\tilde{g}}^2 r}\delta \nonumber\,,\\ u_{2}(\vec{r}\,)&=&\frac{4\pi c_{a}\alpha_{s}(\mu^2)}{3m_{\tilde{g}}^2}\vec{s}^2\delta\left(\vec{r}\,\right ) \,,\nonumber\\ u_{3}(r)&=&-\frac{c_{a}^2\alpha_{s}^2(\mu^2)}{4m_{\tilde{g } } r^2 } \,,\label{vundu}\end{aligned}\ ] ] where @xmath25 c_{a}^2-\left[\frac{1798}{81}+\frac{56}{3}\zeta(3)\right]c_{a}t_{f}n_{f } \nonumber\\ & & -\left[\frac{55}{3}-16\zeta(3)\right]c_{f}t_{f}n_{f}+\left(\frac{20}{9}t_{f}n_{f}\right)^2 \,.\label{betaunda}\end{aligned}\ ] ] here @xmath26 stands for the strong coupling constant in the @xmath27 scheme , @xmath28 is the number of active quark flavors , @xmath14 , @xmath29 , @xmath30 , @xmath31 the euler constant , @xmath32 and @xmath33 is the constituent pole mass . for the coulomb part @xmath34 the transformation from color - triplet , where the result can be found in the literature , to constituents in an arbitrary representation @xmath15 , bound to a singlet state , can be understood as follows : in lo the overall normalization is changed by substituting @xmath35 . this is the only modification also in nlo , thus leaving the constant @xmath36 and the terms proportional to @xmath37 unchanged . for the case under discussion ( @xmath38 ) this is evident from fig . [ feynnlo ] : diagram ( a ) contributes proportional @xmath39 , diagram ( b ) proportional @xmath40 . its @xmath39 part can be combined with ( a ) and in total the ladder plus crossed ladder can be collected in the @xmath41 part of the potential . the @xmath42 part of diagram ( b ) contributes to the @xmath43 term in @xmath44 and in the constant @xmath36 . diagram ( d ) contributes proportional @xmath45 and is responsible for the @xmath46 terms in @xmath44 and @xmath36 . the corresponding considerations are also applicable to the two - loop case , see e.g. fig . [ feynnnlo ] . for diagrams with virtual gluons only ( e.g. fig . [ feynnnlo ] ( a ) ) , the substitution @xmath47 is valid throughout . for diagrams with exactly one dressed gluon propagator exchange ( e.g. fig . [ feynnnlo ] ( c ) , ( d ) , ( e ) ) , the same substitution is valid , as far as the coupling of the gluon to the constituent is concerned and only the external @xmath48 is replaced by @xmath49 . diagrams with three - gluon coupling through a triangular fermion loop ( fig . [ feynnnlo ] ( b ) ) are also proportional to @xmath50 and it is again the first factor , which is chosen to be @xmath51 in the case of quark- and @xmath52 in the case of gluino - constituents . diagrams similar to those of fig . [ feynnlo ] ( a)-(c ) , but with one dressed gluon ( e.g. fig . [ feynnnlo ] ( f ) ) can be handled like the nlo diagrams . in total , up to nnlo , the coulomb part of the potential for the binding of constituents in an arbitrary representation @xmath15 is obtained through the substitution @xmath47 in the overall factor . ( 350,50)(0,0 ) ( 0,25)[l](a ) ( 10,50)(80,50 ) ( 27,50)1.3 ( 63,50)1.3 ( 27,50)(27,0)26 ( 63,50)(63,0)26 ( 27,0)1.3 ( 63,0)1.3 ( 10,0)(80,0 ) ( 90,25)[l](b ) ( 100,50)(170,50 ) ( 117,50)1.3 ( 153,50)1.3 ( 117,50)(153,0)28 ( 153,50)(117,0)28 ( 117,0)1.3 ( 153,0)1.3 ( 100,0)(170,0 ) ( 180,25)[l](c ) ( 190,50)(260,50 ) ( 207,50)1.3 ( 243,50)1.3 ( 207,50)(225,25)24 ( 243,50)(225,25)24 ( 225,25)1.3 ( 225,25)(225,0)23 ( 225,0)1.3 ( 190,0)(260,0 ) ( 270,25)[l](d ) ( 280,50)(350,50 ) ( 315,50)1.3 ( 315,50)(315,37)21 ( 315,37)1.3 ( 315,25)(12,0,360 ) ( 315,13)1.3 ( 315,13)(315,0)21 ( 315,0)1.3 ( 280,0)(350,0 ) ( 280,140)(0,0 ) ( -10,110)[l](a ) ( 0,140)(80,140 ) ( 10,140)1.2 ( 40,140)1.2 ( 70,140)1.2 ( 10,140)(40,80)27 ( 40,140)(70,80)27 ( 70,140)(10,80)29 ( 10,80)1.2 ( 40,80)1.2 ( 70,80)1.2 ( 0,80)(80,80 ) ( 90,110)[l](b ) ( 100,140)(180,140 ) ( 115,140)1.2 ( 165,140)1.2 ( 165,140)(153,123)22 ( 115,140)(127,123)22 ( 153,123)1.2 ( 127,123)1.2 ( 140,99)1.2 ( 153,123)(127,123 ) ( 127,123)(140,99 ) ( 140,99)(153,123 ) ( 140,99)(140,80)22 ( 140,80)1.2 ( 100,80)(180,80 ) ( 190,110)[l](c ) ( 200,140)(280,140 ) ( 240,140)1.2 ( 240,140)(240,125)22 ( 240,125)1.2 ( 240,110)(15,0,360 ) ( 225,110)1.2 ( 255,110)(225,110)23 ( 255,110)1.2 ( 240,95)1.2 ( 240,95)(240,80)22 ( 240,80)1.2 ( 200,80)(280,80 ) ( -10,30)[l](d ) ( 0,60)(80,60 ) ( 40,60)1.2 ( 40,60)(40,45)22 ( 40,45)1.2 ( 40,30)(15,0,360 ) ( 32.5,42.99)1.2 ( 32.5,42.99)(32.5,17.01)23 ( 32.5,17.01)1.2 ( 40,15)1.2 ( 40,15)(40,0)22 ( 40,0)1.2 ( 0,0)(80,0 ) ( 90,30)[l](e ) ( 100,60)(180,60 ) ( 140,60)1.2 ( 140,60)(140,45)22 ( 140,45)1.2 ( 140,30)(15,0,180 ) ( 155,30)(125,30 ) ( 155,30)1.2 ( 125,30)1.2 ( 140,30)(15,180,270)22 ( 140,30)(15,270,360)22 ( 140,15)1.2 ( 140,15)(140,0)22 ( 140,0)1.2 ( 100,0)(180,0 ) ( 190,30)[l](f ) ( 200,60)(280,60 ) ( 210,60)1.2 ( 262.07,52.07)(270,60)-21 ( 210,0)(247.93,37.93)-26 ( 270,60)1.2 ( 255,45)(10,0,360 ) ( 262.07,52.07)1.2 ( 247.93,37.93)1.2 ( 210,0)1.2 ( 270,0)(210,60)-29 ( 270,0)1.2 ( 200,0)(280,0 ) the term @xmath53 of eq . ( [ vundu ] ) of order @xmath54 originates from a one - loop calculation , and the factor @xmath55 must be replaced by @xmath56 . the terms of order @xmath57 collected in @xmath58 and @xmath59 can be translated simply by replacing the global factor @xmath48 by @xmath49 . for our application the substitution @xmath60 is implicit . using the potential as specified in eqs . ( [ pot ] ) and ( [ vundu ] ) , binding energies , level spacings and bound - state wave functions are easily obtained by solving schrdinger s equation numerically for @xmath61 and adding the singular terms collected in @xmath62 and @xmath53 in perturbation theory or , alternatively , by evaluating energy level and wave functions in a perturbative series in @xmath63 . the analytic results both for binding energies @xmath64 and wave functions at the origin @xmath65 are listed in appendix [ app : energy ] . in numerical form the results for the two lowest levels are given by @xmath66\right\ } \,,\nonumber\\ e_{2}&\hspace{-0.1cm}=&\hspace{-0.2cm}-\frac{m_{\tilde{g}}c_{a}^2\alpha_{s}^2}{16}\left\ { 1+\alpha_{s}\left(2.44l_{2}+4.42\right)\right . \nonumber\\ & & \hspace{0cm}\hspace{1.95cm}\left.+\alpha_{s}^2\left[\left(4.47l_{2}^2 + 14.19l_{2}+20.54\right)_{c}+\left(11.95\right)_{nc}\right]\right\ } \,,\nonumber\\ \left|\psi_{1}\left(0\right)\right|^2&\hspace{-0.1cm}=&\hspace{-0.2cm}\frac{m_{\tilde{g}}^3c_{a}^3\alpha_{s}^3}{8\pi}\left\ { 1+\alpha_{s}\left(3.66l_{1}-0.43\right)\right . \nonumber\\ & & \hspace{-0.2cm}\hspace{1.8cm}\left.+\alpha_{s}^2\left[\left(8.93l_{1}^2 - 5.11l_{1}+5.83\right)_{c}+\left(27l_{1}+57.38\right)_{nc}\right]\right\ } \,,\nonumber\\ \left|\psi_{2}\left(0\right)\right|^2&\hspace{-0.1cm}=&\hspace{-0.2cm}\frac{m_{\tilde{g}}^3c_{a}^3\alpha_{s}^3}{64\pi}\left\ { 1+\alpha_{s}\left(3.66l_{2}-0.18\right)\right . \nonumber\\ & & \hspace{-0.2cm}\hspace{1.8cm}\left.+\alpha_{s}^2\left[\left(8.93l_{2}^2 - 3.86l_{2}+10.19\right)_{c}+\left(27l_{2}+29.53\right)_{nc}\right]\right\ } \,,\label{eundpsinum}\end{aligned}\ ] ] with @xmath67 and @xmath33 being the gluino pole mass . the first of the @xmath68 terms give the coulombic corrections ( @xmath69 ) the second ones the non - coulombic ones ( @xmath70 ) . here and below we use @xmath71 as defined in the @xmath27 scheme and the @xmath72 dependence from the numerical solution of the three loop renormalization group equation with the starting value @xmath73 . we use @xmath74 . this is well justified , as long as the characteristic scale @xmath75 is smaller than the top - quark mass , an assumption well justified for all gluino masses under consideration . top - quark effects could be included following @xcite . if not stated otherwise , in the following we use for the renormalization scale @xmath76 with @xmath77 corresponding to @xmath78 . we have convinced ourselves that the perturbative series and the numerical results are in very good agreement as far as @xmath34 is concerned . the subsequent analysis will therefore be based on the fully perturbative approach . in the pole mass scheme as function of twice the constituent mass . ] in the potential subtracted mass scheme as function of twice the potential subtracted mass . the curves have been obtained using the pole mass as input and evaluating both the potential subtracted mass ( using eq . ( [ mps ] ) ) and @xmath79 to a given order in @xmath26 . note that the dash - dotted and dashed curves are almost on top of each other . ] the groundstate energy @xmath80 is shown in fig . [ energiepol ] as a function of @xmath81 for gluino masses @xmath33 between @xmath82 and @xmath83 . the lo ( dotted ) , nlo ( dashed ) and nnlo predictions are displayed individually . for the nnlo result both the result for @xmath34 ( dash - dotted ) and the one including the corrections collected in @xmath58 , @xmath59 and @xmath53 ( solid ) are shown . as stated above , @xmath33 is understood as the constituent pole mass . the poor convergence of the perturbative series can partly be traced back to the large non - coulombic correction which appear for the first time in nnlo and amount to twice the nnlo coulombic one . qualitatively this behavior is quite similar to the one observed for quarkonia in @xcite . it remains to be seen to which extent inclusion of orders of @xmath84 and higher will stabilize these predictions . in fig . [ energieps ] we show as an alternative representation the predictions for the ground state energy , using as reference the potential subtracted mass @xcite , which is related to the pole mass through @xmath85\biggr . \nonumber\\ & & \biggl.\mbox{}+\left(\frac{\alpha_{s}(\mu)}{4\pi}\right)^2\biggl[a_{2}-\left(2a_{1}\beta_{0}+\beta_{1}\right)\left(\ln\left(\frac{\mu_{f}^2}{\mu^2}\right)-2\right)\biggr.\biggr . \nonumber\\ & & \biggl.\biggl.\mbox{}+\beta_{0}^2\left(\ln^2\left(\frac{\mu_{f}^2}{\mu^2}\right)-4\ln\left(\frac{\mu_{f}^2}{\mu^2}\right)+8\right)\biggr]\biggr\ } \ , . \label{mps}\end{aligned}\ ] ] in the following we adopt for the factorization scale @xmath86 , independent of @xmath87 ( with @xmath77 as before ) . ( note , that @xmath88 now also depends on @xmath87 . ) we refrain from listing explicitly the results for the energy levels in the potential subtracted scheme since the corresponding formulae are quite bulky . they can easily be obtained from eqs . ( [ eundpsinum ] ) and ( [ mps ] ) . the implementation of the potential subtracted mass leads to a significantly improved convergence of the perturbative series for @xmath89 as far as the coulombic part is concerned ( see fig . [ energieps ] ) and similarly for @xmath90 . however , the meson mass difference between the @xmath91 and the @xmath92 state , @xmath93 , is independent of the choice of the scheme and exhibits a poor convergence in both cases . the predictions for @xmath93 are shown in fig . [ anregungpol ] , adopting @xmath76 . the nnlo terms are evidently important and the poor convergence of the prediction for @xmath80 is reflected in this figure . in the on - shell scheme the @xmath91 binding energy is evidently closely related to the energy difference between the groundstate and the onset of open gluino production or , more specifically , the threshold for pair production of color - neutral @xmath94 hadrons . however , anticipating a mass resolution of several tens of gev at least , it will be difficult to resolve the densely distributed radial excitations . for this reason the energy difference between the groundstate and the first radial excitation is a convenient measure of the isolation of the @xmath91 state . as function of twice the constituent mass . ] production and decay of non - relativistic @xmath19-wave bound states are proportional to the square of the wave function at the origin . the analytic results are collected in the appendix [ app : energy ] , numerical results are listed in eq . ( [ eundpsinum ] ) and shown in fig . [ wellenfunktion ] . the enormous size in particular of the non - coulombic nnlo terms has also been observed for the top - antitop system and destabilizes the predictions @xcite . at present the size of these corrections must be considered as an estimate of the theory uncertainty . in the pole mass scheme ( a ) and the potential subtracted scheme ( b).,title="fig : " ] in the pole mass scheme ( a ) and the potential subtracted scheme ( b).,title="fig : " ] . ] the renormalization scale dependence of the lo , nlo and @xmath95 results for @xmath80 and @xmath96 are shown in figs . [ energiemu ] and [ wellenfunktionmu ] for a constituent mass of @xmath97 tev . the predictions are normalized to the lo prediction evaluated at the scale @xmath72 as defined above . a significant stabilization is observed with increasing order in @xmath63 and , furthermore , the higher order contributions are reasonably small for the preferred choice of the renormalization scale , at least , as far as the coulombic part is concerned . this is in particular true for the ground state energy parameterized in terms of the potential subtracted mass where close to @xmath98 gev the coulombic nnlo results even vanish . in the case of the wave function one observes around @xmath99 gev corrections of the order of a few percent from nlo and nnlo@xmath100 , however , a huge contribution from the non - coulombic terms . in the present context squarks are assumed to be significantly heavier than gluinos , such that the direct decay @xmath101 is forbidden . an extreme example is provided by split susy @xcite with @xmath102 and we shall adopt this simplifying mass assignment throughout this section . however , this assumption could be significantly relaxed . indeed , once @xmath103 is valid the gluinos have a very small decay rate , for @xmath104 of order @xcite @xmath105 \ , , \label{singledecay}\end{aligned}\ ] ] characterizing the rate for gluino decays into neutralinos ( similar estimates being applicable for decays into charginos ) such that the decay rate is far smaller than the annihilation decay . the evaluation of the rate for gluinonium annihilation into gluon ( and quark ) jets proceeds similar to the one for quarkonia . in born approximation the color factor is replaced as follows @xmath106 and another relative factor @xmath107 arises from the majorana nature of the gluinos , whence @xcite @xmath108 with @xmath109 . using the wave function obtained in lowest order perturbation theory leads to @xmath110 which provides a qualitative estimate of the full result . the perturbative corrections arising from virtual and real emission can be calculated similar to the ones for quarkonia @xcite . they reduce the scale dependence of the lowest order approximation and are complementary to the higher order corrections to the wave function . in total one finds @xmath111\biggr\ } \ , . \nonumber\\ \label{gammanlo}\end{aligned}\ ] ] state into gluons . ] here @xmath63 is understood in the theory with @xmath112 effective flavors and all quark masses , including @xmath113 , have been neglected . the @xmath114 term present in @xmath115 is interpreted in the @xmath74 effective theory . the predictions for @xmath116 and @xmath117 are shown in fig . [ zerfall ] , where the nnlo wave function as shown in fig . [ wellenfunktion ] ( excluding the non - coulombic contribution ) and @xmath118 for the renormalization scale in eq . ( [ gammanlo ] ) have been adopted . as anticipated , the decay rate of the bound state is small compared to the level spacing and large compared to @xmath119 , the decay rate of a single gluino , once @xmath120 . hence annihilation decays will constitute the dominant signal . however , it should be emphasized , that bound states will exist even for @xmath121 , as long as @xmath122 , which is true for the region @xmath123 . the bound state production cross section can be calculated similarly to the one for the pseudoscalar bound state of two top quarks ( usually called @xmath124 ) @xcite . in born approximation the reaction proceeds through gluon fusion and the radiative corrections involve additional @xmath125 , @xmath126 and @xmath127 initiated subprocesses . the final result is quite similar to the one for quarkonia production @xcite . we find @xmath128\biggr\ } \,,\nonumber\\ \hat{\sigma}_{gq}&=&\sigma_{0}\frac{\alpha_{s}}{\pi}\biggl\{-c_{f}\frac{z}{2}\ln z+c_{f}z+\frac{1}{2}p_{gq}(z)\left(\ln\left(\frac{4m_{\tilde{g}}^2(1-z)^2}{\mu^2}\right)-1\right)\biggr\ } \,,\nonumber\\ \hat{\sigma}_{g\overline{q}}&=&\hat{\sigma}_{gq } \,,\nonumber\\ \hat{\sigma}_{q\overline{q}}&=&\sigma_{0}\frac{\alpha_{s}}{\pi}\frac{32}{27}z(1-z ) \,,\ ] ] with @xmath129_{+}+z(1-z)-2\right ) \nonumber\\ & & + \left(\frac{11}{6}c_{a}-\frac{2}{3}n_{f}t_{f}\right)\delta(1-z ) \,,\nonumber\\ p_{gq}(z)&= & c_{f}\frac{1+(1-z)^2}{z } \,,\nonumber\\ f(z)&=&\frac{11z^5 + 11z^4 + 13z^3 + 19z^2 + 6z-12}{6z(1+z)^2}\biggr . \nonumber\\ & & \mbox{}+\,4\left(\frac{1}{z}+z(1-z)-2\right)\ln(1-z)+4\left[\frac{\ln(1-z)}{1-z}\right]_{+ } \nonumber\\ & & \mbox{}\biggl.+\left(\frac{2(z^3 - 2z^2 - 3z-2)(z^3-z+2)z\ln z}{(1+z)^3(1-z)}-3\right)\frac{1}{1-z } \,.\label{sigmaadd}\end{aligned}\ ] ] here @xmath130 , @xmath72 and @xmath131 are the renormalization and the factorization scales , @xmath63 is again defined in the @xmath112 flavor theory , @xmath132 are the altarelli - parisi splitting functions and @xmath133_{+}$ ] denotes the usual @xmath134 distribution . for the wave function at the origin we again adopt the nnlo result without the non - coulombic terms . the predictions for the production cross section are shown in fig . [ produktion ] where the parton distribution functions mstw2008lo ( mstw2008nlo ) @xcite have been used for the lo ( nlo ) calculation . leading order ( dashed ) and nlo ( solid curve ) predictions are evaluated using @xmath135 . state at @xmath136 . ] production cross section and decomposition into contributions from various subprocesses . the line labeled `` gg '' contains also the lo result . ] for a gluino mass of @xmath137 the @xmath72 dependence ( with @xmath138 ) of the lo and the nlo results is shown in fig . [ produktionmu ] , together with the decomposition of the nlo result into the contributions from the different subprocesses . from figs . [ produktion ] and [ produktionmu ] we learn that the nlo correction amounts to more than @xmath139 fairly independent of the renormalization scale . the dominant uncertainty , not displayed in these figures , arises from the wave function at the origin , with indications for an increase by another factor of two . for an anticipated integrated luminosity of @xmath140 between @xmath141 and @xmath142 events are expected for @xmath33 between @xmath143 and @xmath144 . it remains to be seen , if these can be separated from the hadronic continuum background . in the limit of extremely heavy squarks ( and conserved r - parity ) , corresponding to split susy , decay modes of single gluinos are extremely suppressed , such that gluinos ( confined in color - neutral hadrons like @xmath145 ) could even travel microscopic distances . in this case gluinonium annihilation into two gluons is the only relevant channel . however , once squark and gluino masses are comparable , new annihilation decays become possible . at first glance one might consider the mode @xmath146 , mediated by squark exchange . however , for massless quarks this process vanishes as a consequence of helicity conservation . for massive top quarks one finds @xmath147 using @xmath148 . the complete result for @xmath149 and including squark mixing is given in appendix [ app : ttbar ] . decay rate into top quarks . ] the result is plotted in fig . [ ratiott ] for different ratios of @xmath150 . branching ratios of up to @xmath151 are possible for gluino masses close to @xmath152 and squark masses not much bigger . for heavier squark and gluino masses the ratio drops quickly . as a second possibility we consider the loop induced decay into two photons . although suppressed by the electromagnetic couplings @xmath153 , the cleaner signal - to - background ratio might ( at least in principle ) compensate this disadvantage . for arbitrary squark mixing and masses the result is given in appendix [ app : gamgam ] . in the limit of massless quarks and degenerate squarks the decay rate simplifies considerably @xmath154 with a ratio @xmath155 around one to several @xmath156 the two - photon signal will not be detectable . assuming the two - gluon mode to dominate the decay of gluino bound states , the resonant signal must be isolated from the two - jet continuum . below we present an order of magnitude estimate for the signal - to - background ratio , following essentially the arguments given in ref . @xcite . for simplicity it is assumed that quark and gluon jets can be experimentally separated . in this case the background is dominated by gluon - gluon scattering . furthermore , we enhance the signal - to - background ratio by excluding small scattering angles in the partonic center - of - mass system , requiring @xmath157 . to isolate a resonance in the gluon - gluon channel , a good experimental resolution @xmath158 of the invariant dijet mass @xmath159 is essential . let us compare the differential cross section , integrated over the interval @xmath160 for background and signal , and integrated over the same angular range . this leads to the signal - to - background ratio @xmath161^{-1 } .\label{soverr}\end{aligned}\ ] ] . ] using @xmath162 for the dijet resolution @xcite , one obtains the signal - to - background ratio as shown in fig . [ back ] , varying between @xmath163 to @xmath164 . given event rates between @xmath165 and @xmath166 for @xmath33 between @xmath167 and @xmath168 the detection of gluino bound states might become feasible , in particular at a high luminosity version of the lhc . for a large class of supersymmetric extensions of the standard model gluinos will be sufficiently stable to form color - neutral non - relativistic boundstates . the nnlo potential between color - octet constituents forming color singlets has been derived . just as in leading and next - to - leading order the coulombic part of the nnlo potential is related to the quarkonium potential through the substitution @xmath169 in the overall factor . the transformation of the short distance terms involving an additional factor @xmath170 is only slightly more complicated and discussed in section [ sec : properties ] . based on this potential , predictions for the binding energy and the wave function at the origin of the lowest - lying boundstates and the first radial excitation have been derived . for gluino masses between @xmath152 and @xmath171 the @xmath91 binding energy ( in the pole mass scheme ) varies between @xmath172 and @xmath173 , the @xmath174 energy difference between @xmath175 and @xmath176 . for the case of gluinos being lighter than squarks the gluino decay rate is significantly smaller than the level spacing and the boundstates are thus well separated from the continuum . also the wave function at the origin , which enters bound state production and decay rates , has been evaluated in nnlo . for @xmath121 the decay rate of gluinonia into two gluon jets is evaluated in nlo and shown to dominate all other modes . for top squarks with masses comparable to @xmath33 the decay mode into @xmath177 may reach several percent , the decay through virtual quark loops into two photons is too small to be detectable . we evaluate the boundstate production cross section again in nlo . large positive corrections of around @xmath178 arise from the nnlo terms in the wave function at the origin and around @xmath179 from the nlo terms in the cross section . the production rate depends strongly on the gluino mass , with up to @xmath142 events for gluino masses close to the lower limit around @xmath144 ( for an integrated luminosity of @xmath140 ) dropping quickly to @xmath141 events only for @xmath180 . using standard assumptions for the dijet - mass resolution and assuming that gluon and quark jets can be separated , a signal - to - background ratio between @xmath181 and @xmath164 percent might be conceivable . for favorable gluino- and squark - mass combinations gluinonia might thus be detectable . * acknowledgements * m.s . thanks alexander penin for discussions about @xmath182 . this work was supported by the bmbf through grant no . 05h09vke . _ note added _ : after completion of this work we received a preprint by k. hagiwara and h. yokoya where the production of unstable gluinos in the threshold region is considered @xcite . for completeness we present the perturbative analytic results for the energy levels @xmath64 of the @xmath19 waves and the corresponding values of the squared wave function at the origin @xmath65 , truncating the perturbative series at nnlo . we present the results following the notation of @xcite . we define @xmath183 \,,\nonumber\\ \ , , \label{eundpsis}\end{aligned}\ ] ] and parameterize the corrections as @xmath184 where @xmath185 stands for the corrections to the coulombic part of the potential and @xmath186 for the remaining one . the lo values are given as @xmath187 the nlo and nnlo corrections are @xmath188 \nonumber\\ & & + a_{1}^2 + 2a_{2}+4s_{1}\beta_{1}+4a_{1}\beta_{0}\left(3s_{1}-1\right ) \nonumber\\ & & + \beta_{0}^2\left[s_{1}\left(12s_{1}-8-\frac{8}{n}\right)+16s_{2}-8ns_{3}+\frac{2\pi^2}{3}+8n\zeta(3)\right ] \,,\nonumber\\ e_{2}^{nc}&=&\frac{16\pi^2c_{a}^2}{n}\left(3-\frac{11}{16n}-\frac{2}{3}\vec{s}^2\right ) \,,\nonumber\\ f_{1}^{c}&=&6\beta_{0}\ln\left(\frac{n\mu}{m_{\tilde{g}}c_{a}\alpha_{s}}\right)+3a_{1}+2\beta_{0}\left(s_{1}+2ns_{2}-1-\frac{n\pi^2}{3}\right ) \,,\nonumber\\ f_{1}^{nc}&=&0 \,,\nonumber\\ f_{2}^{c}&=&24\beta_{0}^2\ln^2\left(\frac{n\mu}{m_{\tilde{g}}c_{a}\alpha_{s}}\right ) \nonumber\\ & & + \ln\left(\frac{n\mu}{m_{\tilde{g}}c_{a}\alpha_{s}}\right)\biggl\{-12\beta_{0}^2 + 6\beta_{1}+8\beta_{0}\left[3a_{1}+2\beta_{0}\left(s_{1}+2ns_{2}-1-\frac{n\pi^2}{3}\right)\right]\biggr\ } \nonumber\\ & & + 3a_{1}^2 + 3a_{2}+2a_{1}\beta_{0}\left(4s_{1}+8ns_{2}-7-\frac{4n\pi^2}{3}\right)+2\beta_{1}\left(s_{1}+2ns_{2}-1-\frac{n\pi^2}{3}\right ) \nonumber\\ & & + \beta_{0}^2\biggl[s_{1}\left(8s_{1}+16ns_{2}-20-\frac{12}{n}-\frac{8n\pi^2}{3}\right)+s_{2}\left(4n^2s_{2}+8 - 8n-\frac{4n^2\pi^2}{3}\right ) \biggr . \nonumber\\ & & \hspace{0.3cm}\biggl.+28ns_{3}-20n^2s_{4}-24ns_{2,1}+16n^2s_{3,1}+4+\frac{(3 + 4n)\pi^2}{3}+\frac{n^2\pi^4}{9}+20n\zeta(3)\biggr ] \,,\nonumber\\ f_{2}^{nc}&=&16\pi^2c_{a}^2\biggl\{3\ln\left(\frac{n\mu}{m_{\tilde{g}}c_{a}\alpha_{s}}\right)-3s_{1}+\frac{6}{n}-\frac{15}{8n^2}+\frac{21}{4}\biggr . \nonumber\\ & & \hspace{1.7cm}+\biggl.\vec{s}^2\left[-\frac{2}{3}\ln\left(\frac{n\mu}{m_{\tilde{g}}c_{a}\alpha_{s}}\right)+\frac{2}{3}s_{1}-\frac{4}{3n}-\frac{7}{9}\right]\biggr\ } \,.\label{eundf}\end{aligned}\ ] ] the constants @xmath189 and @xmath190 depend on the casimir operators @xmath191 and @xmath192 and are defined in eq . ( [ betaunda ] ) . the harmonic sums @xmath193 , the nested harmonic sums @xmath194 and the zeta - function @xmath195 are given as @xmath196 ( 275,50)(0,0 ) ( 0,0)(25,25)-2.53 ( 0,0)(25,25 ) ( 0,50)(25,25)2.53 ( 0,50)(25,25 ) ( 25,25)1.8 ( 25,25)(50,25)33 ( 50,25)1.8 ( 75,0)(50,25 ) ( 50,25)(75,50 ) ( 100,0)(137.5,10)-2.53 ( 100,0)(137.5,10 ) ( 100,50)(137.5,40)2.53 ( 100,50)(137.5,40 ) ( 137.5,40)1.8 ( 137.5,10)(137.5,40)3 ( 137.5,10)1.8 ( 137.5,40)(175,50 ) ( 175,0)(137.5,10 ) ( 118,25)[l]@xmath197 ( 200,0)(237.5,10)-2.53 ( 200,0)(237.5,10 ) ( 200,50)(237.5,40)2.53 ( 200,50)(237.5,40 ) ( 237.5,40)1.8 ( 237.5,40)(237.5,10)3 ( 237.5,10)1.8 ( 237.5,10)(251.5625,25 ) ( 237.5,40)(251.5625,25 ) ( 251.5625,25)(275,50 ) ( 275,0)(251.5625,25 ) ( 218,25)[l]@xmath197 in fig . [ qqbar ] the corresponding diagrams for the decay of two gluinos into top - antitop are shown . diagram @xmath198 drops out for the case of interest because the two gluinos are projected here onto a color - singlet state . the resulting amplitude for @xmath199 and @xmath200 reads @xmath201\left(p_{-}u_{h1}^{(\tilde{t})}-p_{+}u_{h2}^{(\tilde{t})}\right)v_{s_{2}}(k_{2},m_{t } ) \,,\label{ttbaramp}\end{aligned}\ ] ] where @xmath202 and @xmath203 are the spin indices of the top , the anti - top and of the gluinos , @xmath204 and @xmath205 are the left- and right - handed projectors as usual and the @xmath206 are the elements of the orthogonal squark - mixing matrix @xmath207 the relative minus sign between the diagrams @xmath199 and @xmath200 of fig . [ qqbar ] stems from the interchange of two fermionic operators . the amplitude in eq . ( [ ttbaramp ] ) has now , following the methods explained in refs . @xcite , to be projected onto the spin-@xmath208 state using the clebsch - gordon coefficients @xmath209 , where the squared brackets of eq . ( [ ttbaramp ] ) become due to the antisymmetry of the spin function @xmath210\left<0,0;s,\overline{s}\right > \nonumber\\ & = & 2\sum_{s,\overline{s}}\left[u_{\overline{s}}\left(\frac{p}{2},m_{\tilde{g}}\right)\overline{v}_{s}\left(\frac{p}{2},m_{\tilde{g}}\right)\right]\left<0,0;s,\overline{s}\right > \,.\label{spinproj}\end{aligned}\ ] ] now the calculation can be done as for the case of the decay into two gluons and the rate reads @xmath211 ( 455,60)(-8,0 ) ( 0,0)(30,10)-2.53 ( 0,0)(30,10 ) ( 0,55)(30,45)2.53 ( 0,55)(30,45 ) ( 30,10)1.8 ( 30,45)1.8 ( 30,10)(30,45)3 ( 65,10)(30,10 ) ( 30,45)(65,45 ) ( 65,45)(65,10 ) ( 65,10)1.8 ( 65,45)1.8 ( 65,10)(95,0)-33 ( 65,45)(95,55)33 ( 120,0)(150,10)-2.53 ( 120,0)(150,10 ) ( 120,55)(150,45)2.53 ( 120,55)(150,45 ) ( 150,10)1.8 ( 150,45)1.8 ( 150,45)(150,10 ) ( 150,10)(185,10)3 ( 185,45)(150,45)3 ( 185,10)(185,45)3 ( 185,10)1.8 ( 185,45)1.8 ( 185,10)(215,0)-33 ( 185,45)(215,55)33 ( 240,-5)(305,10)-2.56 ( 240,-5)(305,10 ) ( 240,55)(270,45)2.53 ( 240,55)(270,45 ) ( 270,10)1.8 ( 270,45)1.8 ( 270,10)(270,45)3 ( 305,10)(270,10)3 ( 270,45)(305,45 ) ( 305,45)(305,10 ) ( 305,10)1.8 ( 305,45)1.8 ( 270,10)(335,-5)-37 ( 305,45)(335,55)33 ( 360,0)(390,10)-2.53 ( 360,0)(390,10 ) ( 360,55)(390,45)2.53 ( 360,55)(390,45 ) ( 390,10)1.8 ( 390,45)1.8 ( 390,10)(390,45 ) ( 425,27.5)(390,10)3 ( 390,45)(425,27.5)3 ( 425,27.5)1.8 ( 425,27.5)(455,0)-34 ( 425,27.5)(455,55)34 treating all the virtual quarks in fig . [ 2gluto2gam ] as massless , the onto spin and color - projected amplitude has the form @xmath212 the sum in the first equation of ( [ ampgamgam ] ) goes over the flavors ( @xmath213 ) and the indices of the squarks ( @xmath214 ) . the @xmath215 are the momenta of the outgoing photons , the @xmath216 their polarization vectors and the @xmath217 the charges of the corresponding quarks and squarks . the definition of the @xmath218 -functions is the same as in ref . @xcite . 99 w. y. keung and a. khare , phys . d * 29 * ( 1984 ) 2657 . j. h. khn and s. ono , phys . b * 142 * ( 1984 ) 436 . j. t. goldman and h. haber , physica * 15d * , 181 ( 1985 ) . v. g. kartvelishvili , a. v. tkabladze and e. g. chikovani , sov . . phys . * 51 * , 546 ( 1990 ) [ yad . fiz . * 51 * , 859 ( 1990 ) ] . e. chikovani , v. kartvelishvili , r. shanidze and g. shaw , phys . d * 53 * , 6653 ( 1996 ) [ arxiv : hep - ph/9602249 ] . v. g. kartvelishvili , a. v. tkabladze and e. g. chikovani , z. phys . c * 43 * ( 1989 ) 509 . n. arkani - hamed and s. dimopoulos , jhep * 0506 * , 073 ( 2005 ) [ arxiv : hep - th/0405159 ] . g. f. giudice and a. romanino , nucl . b * 699 * , 65 ( 2004 ) [ erratum - ibid . b * 706 * , 65 ( 2005 ) ] [ arxiv : hep - ph/0406088 ] . w. kilian , t. plehn , p. richardson and e. schmidt , eur . j. c * 39 * , 229 ( 2005 ) [ arxiv : hep - ph/0408088 ] . k. cheung and w. y. keung , phys . d * 71 * , 015015 ( 2005 ) [ arxiv : hep - ph/0408335 ] . m. peter , phys . lett . * 78 * , 602 ( 1997 ) [ arxiv : hep - ph/9610209 ] . m. peter , nucl . b * 501 * , 471 ( 1997 ) [ arxiv : hep - ph/9702245 ] . y. schrder , phys . b * 447 * , 321 ( 1999 ) [ arxiv : hep - ph/9812205 ] . a. v. smirnov , v. a. smirnov and m. steinhauser , phys . b * 668 * ( 2008 ) 293 [ arxiv:0809.1927 [ hep - ph ] ] . s. n. gupta and s. f. radford , phys . d * 24 * , 2309 ( 1981 ) . s. n. gupta and s. f. radford , phys . d * 25 * , 3430 ( 1982 ) . s. titard and f. j. yndurain , phys . d * 49 * , 6007 ( 1994 ) [ arxiv : hep - ph/9310236 ] . b. a. kniehl , a. a. penin , m. steinhauser and v. a. smirnov , phys . d * 65 * , 091503 ( 2002 ) [ arxiv : hep - ph/0106135 ] . a. pineda and f. j. yndurain , phys . d * 58 * , 094022 ( 1998 ) [ arxiv : hep - ph/9711287 ] . a. pineda and j. soto , nucl . * 64 * , 428 ( 1998 ) [ arxiv : hep - ph/9707481 ] . a. v. manohar and i. w. stewart , phys . d * 62 * , 074015 ( 2000 ) [ arxiv : hep - ph/0003032 ] . m. beneke , phys . b * 434 * ( 1998 ) 115 [ arxiv : hep - ph/9804241 ] . a. h. hoang , m. c. smith , t. stelzer and s. willenbrock , phys . d * 59 * ( 1999 ) 114014 [ arxiv : hep - ph/9804227 ] . j. h. khn , a. a. penin and a. a. pivovarov , nucl . b * 534 * , 356 ( 1998 ) [ arxiv : hep - ph/9801356 ] . a. a. penin and m. steinhauser , phys . b * 538 * ( 2002 ) 335 [ arxiv : hep - ph/0204290 ] . b. a. kniehl , a. a. penin , v. a. smirnov and m. steinhauser , nucl . b * 635 * , 357 ( 2002 ) [ arxiv : hep - ph/0203166 ] . v. s. fadin and v. a. khoze , jetp lett . * 46 * , 525 ( 1987 ) [ pisma zh . teor . fiz . * 46 * , 417 ( 1987 ) ] . j. h. khn and p. m. zerwas , phys . rept . * 167 * , 321 ( 1988 ) . m. j. strassler and m. e. peskin , phys . rev . d * 43 * , 1500 ( 1991 ) . y. sumino , k. fujii , k. hagiwara , h. murayama and c. k. ng , phys . d * 47 * , 56 ( 1993 ) . m. jezabek , j. h. khn and t. teubner , z. phys . c * 56 * , 653 ( 1992 ) . a. h. hoang _ et al . _ , eur . j. direct c * 2 * , 1 ( 2000 ) [ arxiv : hep - ph/0001286 ] . a. h. hoang and c. j. reisser , phys . d * 71 * ( 2005 ) 074022 [ arxiv : hep - ph/0412258 ] . v. s. fadin , v. a. khoze and t. sjstrand , z. phys . c * 48 * , 613 ( 1990 ) . k. hagiwara , y. sumino and h. yokoya , phys . b * 666 * , 71 ( 2008 ) [ arxiv:0804.1014 [ hep - ph ] ] . y. kiyo , j. h. kuhn , s. moch , m. steinhauser and p. uwer , eur . j. c * 60 * ( 2009 ) 375 [ arxiv:0812.0919 [ hep - ph ] ] . m. melles , phys . d * 58 * , 114004 ( 1998 ) [ arxiv : hep - ph/9805216 ] . m. melles , phys . d * 62 * , 074019 ( 2000 ) [ arxiv : hep - ph/0001295 ] . m. beneke , y. kiyo and k. schuller , nucl . b * 714 * , 67 ( 2005 ) [ arxiv : hep - ph/0501289 ] . a. a. penin , v. a. smirnov and m. steinhauser , nucl . b * 716 * ( 2005 ) 303 [ arxiv : hep - ph/0501042 ] . y. kiyo and y. sumino , phys . rev . d * 67 * ( 2003 ) 071501 [ arxiv : hep - ph/0211299 ] . h. e. haber and g. l. kane , phys . * 117 * , 75 ( 1985 ) . j. h. khn and e. mirkes , phys . d * 48 * , 179 ( 1993 ) [ arxiv : hep - ph/9301204 ] . a. d. martin , w. j. stirling , r. s. thorne and g. watt , arxiv:0901.0002 [ hep - ph ] . the cms collaboration , cms pas sbm 07 001 . k. hagiwara and h. yokoya , arxiv:0909.3204 [ hep - ph ] . b. guberina , j. h. khn , r. d. peccei and r. rckl , nucl . b * 174 * , 317 ( 1980 ) . j. h. khn , j. kaplan and e. g. o. safiani , nucl . phys . b * 157 * , 125 ( 1979 ) . g. t hooft and m. j. g. veltman , nucl . b * 153 * , 365 ( 1979 ) .

for finsler metrics ( no reversibility assumed ) on closed orientable surfaces of genus greater than one , we study the dynamics of minimal rays and minimal geodesics in the universal cover . we prove in particular , that for almost all asymptotic directions the minimal rays with these directions laminate the universal cover and that the busemann functions with these directions are unique up to adding constants . moreover , using a kind of weak kam theory , we show that for almost all types of minimal geodesics in the sense of morse , there is precisely one minimal geodesic of this type .

1404.0573

we begin by fixing some notation . we assume throughout the paper that @xmath0 is a closed orientable surface of genus @xmath1 . on @xmath0 , there exists a ( hyperbolic ) riemannian metric @xmath2 of constant curvature @xmath3 . the riemannian universal cover of @xmath0 , denoted by @xmath4 , is identified with the poincar disc model , i.e. @xmath5 and @xmath6 , where the geodesics @xmath7 are circle segments meeting @xmath8 orthogonally . let @xmath9 be the distance function on @xmath10 induced by @xmath2 . we write @xmath11 for the group of deck transformations @xmath12 with respect to the covering @xmath13 , which extend to naturally to the `` boundary at infinity '' @xmath14 . let @xmath15 be the set of all oriented , unparametrized geodesics @xmath7 with respect to @xmath2 . we can think of @xmath15 as @xmath16 , associating to @xmath17 its pair of endpoints @xmath18 on @xmath14 . moreover , set @xmath19 we consider any finsler metric @xmath20 , which is assumed to be invariant under @xmath21 ( cf . definition [ def finsler ] ) . we write @xmath22 and @xmath23 for the geodesic with respect to @xmath24 defined by @xmath25 . the main object of interest in this paper are rays and minimal geodesics , that is geodesics @xmath26 , respectively that minimize the length with respect to @xmath24 between any of their points . in @xcite , h. m. morse studied the global behavior of minimal geodesics and showed that minimal geodesics lie in tubes around hyperbolic geodesics @xmath17 of finite width @xmath27 , where @xmath27 is a constant depending only on @xmath24 and @xmath2 . in particular , minimal geodesics @xmath28 have well - defined endpoints @xmath29 at infinity , i.e. on @xmath14 ; the analogous result holds for rays . for @xmath30 set @xmath31 morse studied in particular the behavior of minimal geodesics in @xmath32 , where @xmath33 is fixed under a non - trivial group element @xmath34 ; we will recall these results in subsection [ section periodic ] . however , morse left open the finer structure in the asymptotic directions which are not fixed by an element of @xmath21 and the author is not aware of any other work in the literature in this direction . the purpose of this paper is to fill this gap , i.e. to study the structure of @xmath32 for general @xmath35 . while we are still not able to give the structure for all @xmath33 , we will be able to do it for `` most '' @xmath33 . another novelty in this paper , compared to the work of morse , is the use of finsler instead of riemannian metrics . it was observed by e. m. zaustinsky @xcite , that the results of morse carry over to these much more general systems . moreover , it is known that finsler metrics can be used to describe the dynamics of arbitrary tonelli lagrangian systems in high energy levels , cf . @xcite . the sets @xmath36 are bounded by two particular minimal geodesics . the following lemma is theorem 8 of @xcite . [ bounding geodesics morse ] for @xmath17 , there are two particular non - intersecting , minimal geodesics @xmath37 in @xmath36 , such that all minimal geodesics in @xmath36 lie in the strip in @xmath10 bounded by @xmath37 . as a rule , we will always assume that @xmath38 lies left of @xmath39 with respect to the orientation of @xmath40 . we can now state our first result , saying that for most asymptotic directions @xmath35 , the bounding geodesics @xmath41 of the various @xmath42 do not intersect . the proof makes use of weak kam theory , but does not rely on the group action by @xmath21 . in fact , theorem [ thm countable unstable ] and its corollary [ cor thm countable unstable ] hold for any finsler metric @xmath24 on @xmath10 , which is uniformly equivalent to the norm of @xmath2 ( cf . definition [ def finsler ] ) . [ thm countable unstable ] for all but countably many @xmath35 , the set @xmath43 is a lamination of @xmath10 ( i.e. no curves in @xmath44 intersect transversely ) . it would be desirable to know if the set @xmath44 is a lamination of @xmath10 for all @xmath35 ; we would then gain more insight in the structure of @xmath32 for all @xmath33 , cf . theorem [ thm a ] below . from this we will deduce the following . [ cor thm countable unstable ] for almost all @xmath45 with respect to the lebesgue measure , the set @xmath36 consists of precisely one minimal geodesic . for the statement of our next results , we need the following definitions . a minimal ray @xmath46 is called _ forward unstable _ , if there is no minimal ray @xmath47 with @xmath48 and @xmath49 , which is not a subray of @xmath50 . * instability of minimal geodesics was studied in a paper @xcite of w. klingenberg , and some of our results appear in @xcite . however , @xcite contains errors and the results which we prove here , in particular the existence of unstable geodesics , remained open . our work will be independent of @xcite . * it is easy to see ( cf . lemma [ partial instability bounding ] ) that the unique geodesic in @xmath36 in corollary [ cor thm countable unstable ] is forward ( and backward ) unstable . for @xmath51 define the _ ( forward ) width _ of @xmath52 , respectively by @xmath53 [ bem width def ] due to the morse lemma ( cf . theorem [ morse lemma ] ) , @xmath54 is bounded by the global constant @xmath55 . as we will see , @xmath56 implies that all minimal rays in @xmath32 are forward unstable and in particular that the set @xmath44 in theorem [ thm countable unstable ] is a lamination of @xmath10 , while @xmath57 implies the forward instability of all minimal geodesics in @xmath36 . moreover , @xmath56 implies a uniqueness result for weak kam solutions and busemann functions . hence , the main task in this paper is to show @xmath56 for directions @xmath35 , which are not fixed by elements of @xmath21 . while theorem [ thm countable unstable ] and its corollary did not depend on @xmath21 , the following results rely strongly on the @xmath21-action . assuming a certain behavior of a background geodesic @xmath42 under @xmath21 , we can calculate @xmath54 . [ dense width 0 ] if @xmath58 contains a hyperbolic geodesic , which has a dense forward orbit in the hyperbolic unit tangent bundle of @xmath59 , then @xmath56 . in particular , for almost all @xmath35 with respect to the lebesgue measure , we have @xmath56 . note that theorem [ dense width 0 ] has the following analogue for the case where @xmath0 is the @xmath60-torus @xmath61 ( cf . @xcite for the riemannian , @xcite for the finsler case ) : for all irrational rotation directions ( which have full lebesgue measure in @xmath14 ) , the set of minimal geodesics with this direction is a lamination of the universal cover @xmath62 and the weak kam solutions with irrational directions are unique up to adding constants . for completeness and to exemplify the concept of width , we state the following proposition , which follows directly from the results of morse . [ width periodic intro ] if @xmath63 is an axis of a non - trivial group element @xmath34 ( i.e. @xmath64 ) , then @xmath65 in particular , @xmath66 if and only if there is only one geodesic in @xmath36 . a direct generalization of periodicity is recurrence ; we say that @xmath17 is forward recurrent , if the projection of @xmath67 has a forward recurrent orbit in the hyperbolic unit tangent bundle of @xmath59 . for the case @xmath56 the structure of @xmath32 was explained in remark [ bem width def ] above . if @xmath68 , we have the following results . [ thm intro recurrent ] let @xmath35 and @xmath68 . 1 . there are at most two forward recurrent geodesics in @xmath58 , 2 . if @xmath69 is forward recurrent , then @xmath70 and there is a forward unstable geodesic in @xmath36 , 3 . if @xmath33 satisfies the conclusion of theorem [ thm countable unstable ] ( i.e. if the set of bounding geodesics @xmath44 with direction @xmath33 is a lamination of @xmath10 ) , then there can be at most one forward recurrent direction in @xmath58 and if there is a forward recurrent direction @xmath42 , then @xmath65 if @xmath63 is forward recurrent , then there exists forward unstable minimal geodesic in @xmath36 . if @xmath66 , then all minimal geodesics in @xmath36 are forward unstable by proposition [ w=0 implies unique weak kam ] . if @xmath71 , then the claim follows from theorem [ thm intro recurrent ] ( 2 ) . in mather theory , one studies minimal geodesics @xmath72 ( more generally : action minimizers of lagrangian systems ) , such that @xmath73 is a graph over its projection in @xmath0 . this means that in the universal cover , the curves @xmath74 and @xmath75 are equal or disjoint for all @xmath34 , and @xmath50 is called _ simple_. note that , if a minimal geodesic @xmath50 in @xmath36 is simple , then so is its background geodesic @xmath40 . in this case we have the following result . [ thm simple ] if @xmath58 contains a simple hyperbolic geodesic , which is not the axis of any @xmath76 , then @xmath56 . it is not clear to us whether the conclusion of theorem [ thm countable unstable ] is true for all @xmath35 . if it would hold , then this would clarify the structure of all @xmath32 , as seen in the following theorem . [ thm a ] if @xmath24 is such that for all @xmath35 the set of bounding geodesics @xmath44 with direction @xmath33 defined in theorem [ thm countable unstable ] is a lamination of @xmath10 , then @xmath56 for all @xmath35 that are not fixed under any @xmath76 . the set @xmath44 is a lamination of @xmath10 for all @xmath35 , e.g. , if @xmath24 has no conjugate points , cf . theorem 12.1 in @xcite . in particular , if @xmath24 is riemannian with non - positive curvature , then the only flat strips of uniformly positive width are periodic . by theorem [ thm countable unstable ] , this assumption of theorem [ thm a ] is always `` almost '' true . [ bem high dim ] the main concepts in this paper work in any dimension , in particular the weak kam theory for manifolds of hyperbolic type developed in section [ section weak kam ] and the concept of width in section [ section width ] . e.g. , if in a closed manifold @xmath0 carrying a riemanniam metric @xmath2 of strictly negative curvature there exists a hyperbolic geodesic @xmath40 with respect to @xmath2 , such that in @xmath36 there is only one minimal geodesic , then @xmath56 for almost all @xmath33 in the so - called gromov boundary ( with respect to the lebesgue measure , identifying the gromov boundary with a sphere @xmath77 ) , cf . the arguments for the proof of theorem [ dense width 0 ] . here @xmath54 has to be defined differently , for instance by setting @xmath78 most results , however , are strongest in dimension two , hence we stick to this case to simplicity the exposition . * structure of this paper * in section [ section finsler ] , we recall basic definitions and properties as well as the morse lemma . section [ section weak kam ] is devoted to the proof of theorem [ thm countable unstable ] ; here we study so - called weak kam solutions , which are used to prove theorem [ thm countable unstable ] and its corollary . in section [ section width ] , the concept of width of asymptotic directions is introduced to obtain theorems [ dense width 0 ] , [ width periodic intro ] , [ thm intro recurrent ] , [ thm simple ] and [ thm a ] . in appendix [ appendix ] , we make additional remarks on weak kam solutions in dimension two ; the results of the appendix are , however , not used in this paper . we write @xmath79 for the canonical projection , @xmath80 denotes the zero section and @xmath81 the fibers . [ def finsler ] a function @xmath20 is a _ finsler metric on @xmath10 _ , if the following conditions are satisfied : 1 . ( smoothness ) @xmath24 is @xmath82 off @xmath80 , 2 . ( positive homogeneity ) @xmath83 for all @xmath84 , 3 . ( strict convexity ) the fiberwise hessian @xmath85 of the square @xmath86 is positive definite off @xmath80 for all @xmath87 . we say that @xmath24 is _ uniformly equivalent _ to @xmath2 , if @xmath88 we say that @xmath24 is _ invariant under @xmath89 _ , the group of deck transformations @xmath21 with respect to the covering @xmath13 , if @xmath90 note that uniform equivalence of @xmath91 is implied by invariance under @xmath21 due to compactness of @xmath0 . for the rest of this paper , we fix the finsler metric @xmath24 on @xmath10 , which is assumed to be uniformly equivalent to @xmath2 . write @xmath92 for the unit tangent bundle of @xmath24 . the _ geodesic flow _ @xmath93 of @xmath24 is given by @xmath94 , where @xmath23 is the @xmath24-geodesic defined by @xmath25 . we write @xmath95 for the @xmath24-length of absolutely continuous ( @xmath96 ) curves @xmath97\to x$ ] and @xmath98\to x ~ c^{ac } , ~ c(0)=x , c(1)=y \}\ ] ] for the @xmath24-distance . note that if @xmath24 is not reversible , i.e. if not @xmath99 for all @xmath100 , we have @xmath101 in general . a @xmath96 curve segment @xmath97\to x$ ] with @xmath102 a.e . is said to be _ minimal _ , if @xmath103 . curves @xmath46 , @xmath104\to x$ ] , @xmath28 are called _ forward rays , backward rays , minimal geodesics _ , respectively , if each restriction @xmath105}$ ] is a minimal segment . set @xmath106\to x \text { is a backward ray } \ } , \\ { \mathcal{r}}_+ & : = \{v \in sx : c_v:[0,\infty)\to x \text { is a forward ray } \ } , \\ { \mathcal{m } } & : = \{v \in sx : c_v:{\mathbb{r}}\to x \text { is a minimal geodesic } \}.\end{aligned}\ ] ] we will in this paper mainly be concerned with forward rays , the results for backward rays being completely analogous . the following lemma is a key property of rays . it excludes in particular successive intersections of rays and shows that asymptotic rays can cross only in a common initial point . the idea of the proof is classical ; it can be found in @xcite , lemma 2.21 . [ crossing minimals ] let @xmath107 with @xmath108 . assume that @xmath109 for some @xmath110 , but @xmath111 . then , for all @xmath112 and sufficiently large @xmath113 , @xmath114 the following theorem , due to h. m. morse , is the cornerstone of our investigation and arises from the uniform equivalence of @xmath24 and the negatively curved @xmath2 . the fact that the morse lemma also holds in the finsler case was first observed by e. m. zaustinsky @xcite . due to klingenberg @xcite , the morse lemma holds in any dimension . [ morse lemma ] let @xmath91 be uniformly equivalent . then there exists a constant @xmath115 depending only on @xmath91 with the following property . for any two points @xmath116 , the hyperbolic geodesic segment @xmath117\to x$ ] from @xmath118 to @xmath119 and any minimal segment @xmath120\to x$ ] from @xmath118 to @xmath119 we have @xmath121 } d_h({\gamma}[0,d_h(x , y)],c(t ) ) \leq d.\ ] ] if @xmath46 is a forward ray , then there exists a hyperbolic ray @xmath122 , and conversely , if @xmath122 is a hyperbolic ray , then there exists a forward ray @xmath46 , such that @xmath123 the analogous statements hold for backward rays and minimal geodesics . with the morse lemma we can associate asymptotic directions to rays . namely , if @xmath46 is a ray , choose any hyperbolic ray @xmath122 , such that @xmath124 and write @xmath125 . note that @xmath126 if and only if @xmath127 in the euclidean sense in @xmath128 . recall that @xmath15 is the set of oriented , unparametrized hyperbolic geodesics @xmath7 and that @xmath129 are the sets of initial vectors of forward and backward rays and minimal geodesics , respectively . for @xmath130 we set @xmath131 [ remark bounding geod ] due to the morse lemma we have @xmath132 the following lemma shows that the usual topology of @xmath14 is connected with the topology in @xmath133 . recall that we identified @xmath10 with the interior of the unit disc in @xmath134 . for the closure @xmath135 we take the usual topology induced by @xmath134 . [ as dir cont rays ] if @xmath136 with @xmath108 , then @xmath137 in @xmath14 . let @xmath138 be the hyperbolic geodesics connecting @xmath139 to @xmath140 , respectively . the morse lemma shows that for large @xmath113 the hyperbolic geodesics @xmath141 stay at bounded @xmath9-distance on long subsegments , their length increasing to @xmath142 with @xmath113 . in the limit , by @xmath143 , we obtain a hyperbolic limit geodesic initiating from @xmath144 , with bounded distance from @xmath145 and hence from @xmath40 . by the uniqueness of such @xmath40 , we obtain @xmath146 with respect to the euclidean hausdorff distance in @xmath134 and in particular their endpoints converge in @xmath14 . we continue to assume that @xmath147 are uniformly equivalent . for the proof of theorem [ thm countable unstable ] , we will use certain functions , that have been studied in various situations throughout the literature ; in particular , they arise as busemann functions , as we shall see below . we use the language of a. fathi s weak kam theory ( cf . @xcite for an introduction ) . a function @xmath148 is called a _ forward weak kam solution of direction @xmath35 _ , written @xmath149 , if the following two conditions hold : 1 . @xmath150 for all @xmath116 , 2 . for all @xmath87 there exists a forward ray @xmath46 with @xmath151 , @xmath152 , @xmath102 and @xmath153 for all @xmath154 . we write @xmath155 for the set of all @xmath156 , such that @xmath157 for all @xmath158 , call the ray @xmath159 _ @xmath160-calibrated _ and set @xmath161 the set of all forward weak kam solutions is denoted by @xmath162 we write @xmath163 and call @xmath164 the _ asymptotic direction _ of @xmath165 . the set of backward weak kam solutions @xmath166 is defined analogously , with analogous sets @xmath167 and @xmath168 for @xmath169 and the asymptoptic direction @xmath170 we will show in lemma [ as dir well - def ] , that @xmath171 is well - defined . note that it follows from the definition ( condition ( 1 ) ) that @xmath172 and @xmath173 for all @xmath174 . the following lemma is well - known in weak kam theory and probably also in the theory of busemann functions . if @xmath148 is differentiable in @xmath87 , write @xmath175 here @xmath176 denotes differentiation along the fiber , @xmath177 is the legendre transform associated to @xmath24 . note that if @xmath24 is riemannian , then @xmath178 is the usual gradient . [ lemma fathi ] let @xmath165 . then for all @xmath179 , @xmath160 is differentiable in @xmath180 . if @xmath160 is differentiable in @xmath87 , then @xmath181 note that @xmath155 is forward - invariant by the geodesic flow @xmath182 and that by lemma [ lemma fathi ] the set @xmath183 for @xmath184 is a graph over the zero section @xmath80 ( it is even locally lipschitz by theorem 4.11.5 in @xcite ) . we will refer to this fact as the _ graph property _ of @xmath155 . define the lagrangian @xmath185 , then @xmath186 is a tonelli lagrangian ( the fact that it is only @xmath187 in @xmath80 does not matter in this situation ) . write @xmath188 for the action and consider ma s potential @xmath189\to x ~ c^{ac } , c(0)=x , c(t)=y \right\}.\ ] ] we claim that @xmath190 . this shows that @xmath160 is dominated by @xmath186 in the sense of fathi @xcite , such that the lemma is just a reformulation of theorem 4.3.8 in @xcite . proof of the claim : observe that the action of any closed curve in @xmath10 with respect to @xmath191 is non - negative and , using proposition 5.11 in @xcite , one finds for each @xmath116 a time @xmath192 and a @xmath96 curve @xmath193\to x$ ] with @xmath194 , such that @xmath195 . the energy function of @xmath196 is just @xmath196 itself and @xmath50 is also critical for the action with respect to @xmath196 when fixing the connection time , hence we obtain @xmath197 . consider for @xmath198 the reparametrizations @xmath199\to x$ ] , @xmath200 . by minimality of @xmath201 under all @xmath202 and homogeneity of @xmath24 , one easily shows @xmath203 and hence @xmath102 . thus , by minimality of @xmath50 with respect to @xmath204 @xmath205 on the other hand , if @xmath120\to x$ ] is a minimal geodesic segment from @xmath118 to @xmath119 with @xmath102 , we obtain the other inquality : @xmath206 [ as dir well - def ] if @xmath149 , then @xmath207 . let @xmath208 and @xmath179 . by definition , there exists a minimal ray @xmath46 with @xmath209 and @xmath210 . by lemma [ lemma fathi ] we find @xmath211 , i.e. also @xmath212 . we have the following corollary to lemma [ lemma fathi ] . [ j = j then u = u ] if @xmath213 with @xmath214 , then @xmath215 is constant . let @xmath216 be the set where both @xmath217 are differentiable . @xmath218 has full measure by rademacher s theorem ( weak kam solutions are lipschitz by definition ) . lemma [ lemma fathi ] and @xmath214 show @xmath219 for all @xmath220 . the claim follows . we now show @xmath221 . there is a classical way to construct weak kam solutions by considering so - called busemann functions or horofunctions . for this , let @xmath222 and @xmath223 be a sequence with @xmath224 . any @xmath225 limit @xmath226 of the sequence of functions @xmath227 is called a horofunction . if @xmath228 and @xmath229 , then @xmath160 is called the busemann function of @xmath230 . [ existence horofunction ] if @xmath231 with @xmath232 , then for the sequence of functions @xmath233 there exist @xmath225 limit functions @xmath226 . moreover , any limit @xmath160 belongs to @xmath234 . in particular , associating the busemann function @xmath235 to @xmath236 shows @xmath237 there exists a constant @xmath238 , such that @xmath239 due to uniform equivalence of @xmath147 by assumption . let @xmath240 , then by the triangle inequality @xmath241 and by symmetry of @xmath9 , the functions @xmath242 are lipschitz with lipschitz constant @xmath238 . using @xmath243 for all @xmath113 , the arzela - ascoli theorem shows that the @xmath242 have a @xmath225 convergent subsequence with limit in @xmath244 . after passing to a subsequence , we now assume @xmath245 . we then obtain for @xmath116 as before @xmath246 now let @xmath87 and @xmath247\to x$ ] be a minimal segment from @xmath118 to @xmath248 with @xmath249 . choose a convergent subsequence of @xmath250 with limit @xmath251 , then @xmath159 is a minimal ray . similarly , let @xmath252 connect @xmath253 to @xmath248 with @xmath254 . all @xmath255 and hence the limits @xmath256 have uniformly bounded distance by the morse lemma . setting @xmath257 , we have @xmath258 for all so obtained @xmath251 . we have to show @xmath259 for @xmath158 . the triangle inequality for @xmath260 and @xmath261 for @xmath262 show @xmath263 . hence , using minimality of @xmath264 , we have @xmath265 so far , everything in this section works for manifolds of hyperbolic type ( i.e. possessing a metric @xmath2 of negative curvature ) in any dimension . for the next proposition , we need @xmath266 . recall the definition of the bounding geodesics @xmath37 of @xmath36 in lemma [ bounding geodesics morse ] . in each @xmath267 there are similar bounding geodesics . [ bounding geodesics for rays ] given @xmath174 and @xmath268 , there exist two particular non - intersecting , minimal geodesics @xmath269 in @xmath270 , such that all minimal geodesics in @xmath270 lie between @xmath269 . moreover , for any @xmath87 , there exists a unique @xmath268 , such that @xmath118 lies between @xmath269 . we first show that for each @xmath268 , there is at least one geodesic in @xmath270 . for this , we take @xmath271 and a ray @xmath272 in @xmath155 initiating from @xmath248 . each @xmath264 has bounded distance from @xmath40 , independently of @xmath113 , by the morse lemma . with @xmath262 , we obtain as a limit a minimal geodesic in @xmath270 . by the closedness of @xmath273 and the graph property of @xmath267 , there are rightmost and leftmost geodesics @xmath269 in @xmath270 ; the first claim follows . now let @xmath87 and assume that @xmath274 . then there are two minimal geodesics @xmath275 in @xmath267 lying closest to @xmath118 on either side by the closedness of @xmath276 . we have to show that @xmath277 and assume the contrary . choose then a hyperbolic geodesic @xmath278 with @xmath279 in the open segment @xmath280 between @xmath281 , such that @xmath282 . there exists @xmath283 , such that @xmath284 can not intersect and @xmath285 lies between @xmath275 , contradicting the assumption that @xmath275 are closest to @xmath118 . we repeat the definition of instability . a forward ray @xmath46 is called _ forward unstable _ , if for all minimal rays @xmath47 with @xmath48 and @xmath49 we have @xmath286 , i.e. @xmath287 is a subray of @xmath50 . backward instability of backward rays @xmath104\to x$ ] is defined analogously . a minimal geodesic is called _ unstable _ , if it is both forward and backward unstable . instability can be characterized by weak kam solutions . [ unstable aubry ] for @xmath288 , the minimal ray @xmath159 is forward unstable if and only if @xmath289 in particular , the set of forward unstable rays @xmath290 is closed . a similar set @xmath291 appears in the setting of a. fathi s weak kam theory . we called it the _ forward aubry set of direction @xmath33_. it is not clear to us , whether @xmath292 for all @xmath33 . we do not expect any kind of continuity of @xmath293 . let @xmath288 be forward unstable , @xmath179 and @xmath149 . for @xmath294 , we find a ray @xmath46 in @xmath155 , such that @xmath151 . but due to @xmath207 and the instability of @xmath230 , we have @xmath295 , showing by closedness of @xmath155 , that @xmath296 . conversely , let @xmath297 , @xmath179 and assume that @xmath46 is a ray in @xmath32 initiating from @xmath298 . the busemann function @xmath160 of @xmath50 then belongs to @xmath299 , while by assumption @xmath208 . lemma [ lemma fathi ] shows differentiability of @xmath160 in @xmath300 and @xmath211 , i.e. @xmath50 is a subray of @xmath230 . for later reference , we prove the following lemma , showing a partial instability of the bounding geodesics @xmath41 of @xmath36 from lemma [ bounding geodesics morse ] . the lemma was basically already known to morse . [ partial instability bounding ] if @xmath63 and @xmath301 is the closed strip between @xmath37 , then any ray @xmath46 initiating in @xmath302 with @xmath303 lies entirely in @xmath302 . the analogous statement holds for backward rays with @xmath304 . suppose @xmath46 is a ray with @xmath305 and @xmath303 , leaving @xmath302 . consider a sequence of minimal segments @xmath264 from @xmath306 to @xmath307 , converging to a minimal geodesic @xmath50 in @xmath36 by the morse lemma . but due to minimality of @xmath308 , it has to lie left of @xmath38 , contradicting the fact that all minimal geodesics in @xmath36 lie right of @xmath38 by definition . we prove a lemma . recall the definition of @xmath309 in proposition [ bounding geodesics for rays ] for @xmath310 . [ countable 1 ] let @xmath35 and @xmath311 , then for all but countably many @xmath312 we have @xmath313 , which is then forward unstable . let @xmath314 @xmath186 defines a closed lamination of @xmath10 by the graph property of @xmath167 , hence all open sets @xmath315 are pairwise disjoint . moreover , for any @xmath316 with @xmath317 , the strip between the @xmath309 is an element of @xmath318 . the first part of the lemma follows , as long as @xmath318 is countable . but as @xmath10 is the union of countably many compact sets @xmath319 , and each @xmath319 can contain at most countably many disjoint open sets . the claim follows . for the forward instability of @xmath320 , suppose @xmath287 is a forward ray in @xmath321 initiating from @xmath50 . the nearby @xmath322 converge to @xmath50 , as @xmath323 converges to @xmath40 , while they have different points at @xmath324 . hence @xmath287 would have to be intersected twice by a minimal geodesic , contradiction . take a dense sequence @xmath325 and write @xmath326 for the set of points @xmath327 of @xmath328 and such that @xmath36 contains a forwards unstable minimal geodesic . then @xmath329 has a countable complement in @xmath14 by lemma [ countable 1 ] . set @xmath330 which still has a countable complement in @xmath14 . if @xmath331 , then for all @xmath113 , there exists a forward unstable , minimal geodesic @xmath332 , where @xmath333 is the hyperbolic geodesic from @xmath334 to @xmath33 . if @xmath42 is arbitrary , choose two subsequences @xmath335 of @xmath334 , such that @xmath336 with @xmath337 in the counterclockwise orientation of @xmath14 . writing @xmath338 for the forward unstable geodesic found in @xmath339 , we find limits @xmath340 of @xmath341 in @xmath36 . since the @xmath342 can not intersect the bounding geodesics @xmath41 by instability , we have @xmath343 . since @xmath291 is closed , the theorem follows . let @xmath344 be the set of @xmath33 satisfying the conclusion of theorem [ thm countable unstable ] , then @xmath345 has full lebesgue measure , since it has a countable complement . for @xmath346 and any @xmath347 , both @xmath348 belong to @xmath267 by construction , so we can apply lemma [ countable 1 ] , showing @xmath349 for all but countably many @xmath69 . hence , if we write @xmath350 for the set of points @xmath351 of @xmath42 with @xmath36 consisting of only one minimal geodesic , each @xmath350 with @xmath346 again has full lebesgue measure in @xmath14 . we obtain @xmath352 for the whole section [ section width ] , we assume that @xmath24 is invariant under the group @xmath21 of deck transformations with respect to the covering @xmath13 . note that then @xmath24 can be considered a finsler metric on @xmath0 . we concentrate of the forward behavior of forward rays , while all results have analogons for backward rays . as we remarked earlier , the morse lemma is the cornerstone of this work . it allows us to define the width @xmath54 of @xmath35 , which is special for surfaces of genus @xmath1 . it does not work for the 2-torus , even though the morse lemma also holds in this case due to g. a. hedlund @xcite , but does work in closed manifolds of arbitrary dimension admitting a riemannian metric @xmath2 of negative curvature . hence the finiteness of the following objects reflects once more the hyperbolic background structure of @xmath10 . recall the definition of the bounding geodesics @xmath41 of @xmath36 in lemma [ bounding geodesics morse ] . [ def width ] for @xmath30 set @xmath353 we call @xmath354 , the _ ( forward ) width _ of @xmath355 , respectively . [ width bounded ] let @xmath356 which is finite by the morse lemma . then @xmath357 one can study uniqueness of weak kam solutions using the notion of width as follows . [ w=0 implies unique weak kam ] if @xmath56 for @xmath35 , then there is up to adding a constant only one forward weak kam solution @xmath149 , i.e. for any @xmath149 we have @xmath358 in particular , if @xmath56 , then @xmath359 , i.e. all rays @xmath46 with @xmath152 are forward unstable . first observe , that all @xmath41 with @xmath42 are positively unstable . namely , if @xmath46 is a ray initiating e.g. from @xmath39 , then @xmath360 has to lie inside the strip between @xmath37 by lemma [ partial instability bounding ] ; now @xmath57 and lemma [ crossing minimals ] show that @xmath50 is a subray of @xmath39 . this shows that the bounding geodesic @xmath309 in @xmath267 for @xmath347 ( cf . proposition [ bounding geodesics for rays ] ) coincide with @xmath41 . let now @xmath361 , @xmath208 , @xmath179 and @xmath362 with @xmath363 . by the instability of the @xmath41 , the rays @xmath230 and @xmath285 lie in a common gap of @xmath291 , and hence between a pair @xmath37 for some @xmath42 . again @xmath57 and lemma [ crossing minimals ] show that @xmath285 is a subray of @xmath230 . by closedness of @xmath364 we find @xmath365 , showing @xmath214 by symmetry . corollary [ j = j then u = u ] shows that @xmath215 is constant . the widths defined above satisfy a useful upper semi - continuity property , from which we can deduce theorem [ dense width 0 ] from the introduction . for this , we use the group action by @xmath21 . [ def positive gamma ] we say that a sequence @xmath366 converges to @xmath17 , if for the pairs of endpoints @xmath367 with respect to the euclidean metric in @xmath368 . a sequence @xmath369 is _ positive for @xmath17 _ , if there exists a sequence @xmath370 with @xmath371 ( with respect to the euclidean distance in @xmath128 ) and a compact set @xmath372 , such that @xmath373 for all @xmath113 . intuitively , a sequence @xmath374 is positive for @xmath63 , if @xmath375 describes the behavior of @xmath376 in the compact quotient @xmath0 , as @xmath377 . [ width semi - cont ] let @xmath378 and @xmath379 be a positive sequence for @xmath40 , such that @xmath380 . then @xmath381 . let @xmath382 . as @xmath374 is positive for @xmath40 , any other @xmath383 will converge under the @xmath374 to the same @xmath384 ( @xmath385 are asymptotic with respect to @xmath9 , which is invariant under the isometries @xmath374 ) . hence it is enough to show @xmath386 . under @xmath374 , the minimal geodesics @xmath37 have as a limit two geodesics @xmath387 in @xmath388 ( theorem 7 in @xcite ) , which have @xmath389 since @xmath374 is positive , the claim follows . the set @xmath318 of @xmath35 , such that the geodesics @xmath42 have a dense forward orbit in the hyperbolic unit tangent bundle of @xmath0 , has full lebesgue measure ( the geodesic flow is ergodic with respect to the lebesgue measure in the hyperbolic unit tangent bundle ; theorem 1.7 in @xcite shows that almost all forward orbits are dense ) . for @xmath331 , take any @xmath69 and a sequence @xmath369 positive for @xmath40 , such that @xmath390 , where @xmath384 is such that @xmath388 contains only one minimal geodesic . such @xmath384 exist by corollary [ cor thm countable unstable ] . but then @xmath391 and lemma [ width semi - cont ] proves that @xmath56 for all @xmath331 . intuitively the above proof shows that , if @xmath35 is the point at infinity of a forward dense hyperbolic geodesic , then the complicated behavior of the geodesics in @xmath58 forces all rays in @xmath32 to come close to each other at infinity . otherwise , the hyperbolic topology of @xmath0 would `` peel '' the geodesics away from each other , as the geodesics move around the surface . observe that the techniques in the proof of theorem [ dense width 0 ] work in any dimension , as long as there is one @xmath17 with @xmath392 ( cf . also remark [ bem high dim ] in the intoduction ) . the existence of @xmath40 with @xmath392 in general hyperbolic manifolds is , however , not clear to us . we have two more techniques to derive @xmath57 , that we will use below . [ one - sided approx ] suppose @xmath393 and @xmath369 is positive for @xmath40 with @xmath380 , such that @xmath394 in @xmath10 . then @xmath57 . first assume that @xmath375 and @xmath384 have no points at infinity in common for infinitely many @xmath113 . the bounding geodesics @xmath41 converge ( after taking a subsequence if necessary ) under @xmath374 to one bounding geodesic , @xmath395 , say ( no geodesic in @xmath396 can intersect @xmath395 , cf . theorem 4 in @xcite ) . this shows that for large @xmath113 , the two geodesics @xmath397 are close . using that @xmath374 is positive for @xmath40 , the claim follows . now let @xmath398 for infinitely many @xmath113 , w.l.o.g . for all @xmath113 . then we find @xmath399 , such that @xmath400 fixes @xmath327 . since a periodic hyperbolic geodesic can not approach any other hyperbolic geodesic , @xmath40 is not the axis of @xmath401 . theorem [ morse periodic ] below shows that @xmath37 are asymptotic . finally , suppose @xmath402 for infinitely many @xmath113 , w.l.o.g . for all @xmath113 . replace @xmath40 by @xmath403 , such that we can assume @xmath404 . now it follows that all @xmath374 fix @xmath33 and there exists @xmath76 , such that @xmath405 for a sequence @xmath406 . but for such @xmath374 , there can not be a sequence @xmath407 , which is moved under @xmath374 into a given compact set , contradicting the assumption that @xmath374 is positive for @xmath40 . [ periodic approx ] let @xmath378 , @xmath369 positive for @xmath40 and @xmath384 be the axis of some @xmath76 , such that @xmath390 and @xmath408 for all @xmath113 . then @xmath57 . by lemma [ one - sided approx ] , we can assume that @xmath409 for all @xmath113 . assume that @xmath410 in the counterclockwise orientation of @xmath14 , the other case being analogous . then @xmath411 and @xmath412 have to intersect in some point @xmath413 . choose @xmath414 , such that @xmath415 lies in a compact set and assume w.l.o.g . that @xmath416 . the backward rays @xmath417 $ ] converge w.l.o.g . to a backward ray @xmath104\to x$ ] with @xmath151 and @xmath418 by @xmath390 . by theorem [ morse periodic ] ( 3 ) below , @xmath50 is a subray of @xmath412 and hence @xmath419 converges to @xmath412 . note that the new sequence @xmath420 is again positive for @xmath40 ( the points @xmath248 converge to @xmath421 ) and under @xmath422 , the geodesic @xmath38 converges to a minimal geodesic @xmath423 lying at finite distance left of the limit @xmath412 of the @xmath264 . but by definition of @xmath38 being the leftmost minimal geodesic in @xmath388 , we also obtain @xmath424 . the claim follows . [ bem periodic approx ] assume in lemma [ periodic approx ] , that additionally @xmath375 and @xmath384 do not have the same point at @xmath324 for all but finitely many @xmath113 . then , as all geodesics in @xmath425 converge under @xmath374 to the same periodic geodesic @xmath384 , the proof of lemma [ periodic approx ] , combined with lemma [ one - sided approx ] , actually shows @xmath66 . if on the other hand @xmath375 and @xmath384 do have the same point at @xmath324 , then @xmath327 is a fixed point of @xmath21 . we will discuss this situation in the next subsection . the following theorem due to morse completely clarifies the structure of @xmath426 , if @xmath35 is fixed by some non - trivial @xmath34 , cf . figure [ fig periodic ] . recall the definition of the bounding geodesics @xmath41 of @xmath36 in lemma [ bounding geodesics morse ] . [ morse periodic ] let @xmath76 be a prime element with hyperbolic axis @xmath63 . then 1 . the set @xmath427 of initial vectors to minimal geodesics in @xmath36 that are invariant under @xmath401 contains the bounding geodesics @xmath37 and any minimal geodesic in @xmath36 invariant under some @xmath428 with @xmath429 is also invariant under @xmath401 itself , hence belongs to @xmath427 , 2 . any ray in @xmath430 is asymptotic to a minimal geodesic in @xmath427 ; moreover , no geodesic in @xmath36 can be asymptotic in both its senses to a single periodic minimal geodesic in @xmath427 , 3 . no ray @xmath230 with @xmath431 has @xmath432 , 4 . between any pair of neighboring minimal geodesics @xmath275 in the closed set @xmath433 ( @xmath434 lying right of @xmath435 with respect to the orientation of @xmath40 ) , there are minimal geodesics @xmath436 in @xmath36 , such that @xmath437 is asymptotic to @xmath434 for @xmath438 and asymptotic to @xmath435 for @xmath377 ; @xmath439 has the opposite behavior . writing @xmath440 for the sets of all initial conditions of such `` heteroclinics '' , we have @xmath441 and the two sets @xmath442 are laminations of @xmath10 . note that morse proved the above statements in the riemannian setting . however , one finds that his arguments do not rely on the riemannian character of the metric @xmath24 . in @xmath443 in theorem [ morse periodic ] for @xmath35 being a fixed point of some @xmath76 . minimal geodesics with respect to @xmath24 are depicted in black , their corresponding background geodesics in gray , `` per '' labels the hyperbolic axis of @xmath401 . [ fig periodic ] ] theorem [ morse periodic ] has the following corollary , which we stated as proposition [ width periodic intro ] in the introduction . [ width periodic ] let @xmath35 be a fixed point of a group element @xmath76 and let @xmath17 be the hyperbolic axis of @xmath401 . then 1 . for the width we have @xmath444 in particular , @xmath56 if and only if there is only one minimal geodesic in @xmath36 . if @xmath302 is the closed strip between @xmath37 , then @xmath445 theorem [ morse periodic ] shows that any ray @xmath230 with initial condition @xmath446 is asymptotic to the periodic bounding geodesic @xmath41 with @xmath447 or @xmath448 corresponding to the connected component of @xmath144 in @xmath449 . ( 1 ) follows and , using @xmath450 , the proof of proposition [ w=0 implies unique weak kam ] shows that any ray @xmath230 with @xmath446 is forward unstable and hence belongs to @xmath291 . theorem [ morse periodic ] ( 3 ) shows that also the periodic geodesics in @xmath427 are forward unstable , i.e. @xmath291 contains the set on the right hand side . now let @xmath451 , such that @xmath452 . then @xmath144 lies in an open strip @xmath453 bounded by two neighboring periodic minimal geodesics @xmath275 in @xmath427 , and there are two heteroclinic minimal geodesics @xmath454 between @xmath275 ( theorem [ morse periodic ] ( 4 ) ) . if @xmath300 is asymptotic to @xmath434 , say , and if @xmath437 is asymptotic to @xmath435 as @xmath377 , we find @xmath455 , such that @xmath456 intersects @xmath457 . this shows that @xmath230 is not forward unstable , hence does not belong to the aubry set @xmath291 by proposition [ unstable aubry ] . in this subsection we prove theorem [ thm intro recurrent ] from the introduction ( cf . theorem [ recurrent thm 1 ] and its corollaries [ recurrent thm 2 ] , [ recurrent thm 3 ] ) . recall that a geodesic ray @xmath46 is forward recurrent , if there exists a sequence @xmath369 and a sequence of times @xmath458 , such that @xmath459 converges to @xmath460 with respect to the euclidean hausdorff metric in @xmath134 . this definition is equivalent to the usual definition of forward recurrence of the projected geodesic in @xmath461 . the idea behind the results of this subsection is that the structure of @xmath32 , if @xmath33 is the endpoint of a forward recurrent geodesic @xmath17 , is very similar to the case where @xmath33 is fixed by some @xmath76 . if @xmath56 , then @xmath32 has a simple structure by proposition [ w=0 implies unique weak kam ] , hence we assume @xmath68 . for the whole subsection , fix @xmath35 with @xmath68 . [ recurrent thm 1 ] if @xmath69 is forward recurrent , then 1 . there is a forward unstable geodesic in @xmath36 , 2 . @xmath462 . presently , we do not know if @xmath37 are recurrent , if @xmath40 is recurrent . let @xmath369 be positive for @xmath40 , such that @xmath463 . by lemma [ width semi - cont ] , @xmath464 . on the other hand , @xmath465 by definition , so ( 2 ) follows . moreover , the minimal geodesics @xmath397 have limits @xmath466 in @xmath36 with @xmath467 ( @xmath468 follows from @xmath469 ; if the infimum would be @xmath470 , then also @xmath471 ) . next , we observe that w.l.o.g . , @xmath40 is non - periodic and using lemma [ one - sided approx ] , the approximation @xmath463 can be assumed to satisfy @xmath472 for all @xmath113 . assume that @xmath473 in the counterclockwise orientation of @xmath14 , the other case being analogous . if @xmath46 is a ray initiating from @xmath474 , it can not lie right of @xmath39 by lemma [ partial instability bounding ] . suppose @xmath50 is not a subray of @xmath39 , then for @xmath184 , @xmath475 has a uniformly positive distance from @xmath474 by lemma [ crossing minimals ] . by @xmath476 , there exist times @xmath477 , such that @xmath478 . for large enough @xmath479 , we then find a point @xmath480 and hence for large @xmath113 a point @xmath481 in the closed strip bounded by @xmath482 ( using the asymptotic behavior of @xmath483 ) . a contradiction is reached using lemma [ crossing minimals ] , cf . figure [ fig recurrent unstable ] . this shows that @xmath39 is forward unstable and hence ( 1 ) holds . . [ fig recurrent unstable ] ] [ bem recurrent thm 1 ] we saw in the proof of theorem [ recurrent thm 1 ] that one bounding geodesic @xmath41 is positively unstable , say @xmath39 . let @xmath484 be the points @xmath485 , such that @xmath486 in the counterclockwise order of @xmath14 . then for all @xmath487 with @xmath488 we have @xmath450 and hence all forward rays initiating in the connected component @xmath489 of @xmath490 touching @xmath484 with endpoint @xmath33 are forward unstable ( by the arguments in the proof of proposition [ w=0 implies unique weak kam ] ) . proof : under a positive sequence @xmath374 for @xmath40 with @xmath463 , both bounding geodesics @xmath491 have limits in @xmath36 right of the limit @xmath492 by the instability of @xmath39 . we saw in the proof of theorem [ recurrent thm 1 ] , that @xmath434 and @xmath39 come close to each other near @xmath324 . also the limits of @xmath493 both come close to @xmath39 , since @xmath483 and @xmath493 do not intersect . @xmath450 follows . [ recurrent thm 2 ] there are at most two positively recurrent hyperbolic geodesics in @xmath58 . suppose the contrary and let @xmath494 with @xmath495 all be forward recurrent , distinct and such that @xmath496 are increasing with respect to the counterclockwise orientation of @xmath14 . but remark [ bem recurrent thm 1 ] shows that @xmath450 for all @xmath384 on one side of @xmath497 , contradicting @xmath498 in theorem [ recurrent thm 1 ] ( 2 ) . the next corollary shows that for all but countably many @xmath35 with @xmath68 , such that there is a forward recurrent @xmath42 , the set @xmath32 has the same structure as in the case where @xmath33 is fixed by some @xmath76 , cf . corollary [ width periodic ] ( the only difference being that there can be more forward unstable rays between @xmath37 ) . [ recurrent thm 3 ] if @xmath33 satisfies the conclusion of theorem [ thm countable unstable ] ( i.e. all bounding geodesics @xmath41 with @xmath382 are forward unstable ) , then there can be at most one forward recurrent direction in @xmath58 and if there is a forward recurrent direction @xmath69 , then 1 . for the width we have @xmath499 2 . if @xmath302 is the open strip between @xmath37 , then @xmath500 we now by corollary [ recurrent thm 2 ] that there can be at most two recurrent directions @xmath58 , which both have maximal width . arguing as in remark [ bem recurrent thm 1 ] and using our assumption that the bounding geodesics of the two directions can not intersect , the width of one of the two recurrent directions would vanish . items ( i ) and ( ii ) follow from theorem [ recurrent thm 1 ] and remark [ bem recurrent thm 1 ] . in this short subsection we prove theorem [ thm simple ] , i.e. if @xmath35 is not a fixed point of @xmath21 and @xmath58 contains a geodesic @xmath40 , which is simple ( i.e. @xmath501 are disjoint for all @xmath76 ) , then @xmath56 . let @xmath17 be simple and @xmath502 , such that @xmath33 is not fixed under any @xmath76 . choose any positive sequence @xmath369 , such that @xmath375 converges to some @xmath503 . since @xmath40 is simple , also @xmath384 is simple . if @xmath384 is the axis of some @xmath76 , apply lemma [ periodic approx ] and remark [ bem periodic approx ] , which shows @xmath56 in this case . if @xmath384 is not an axis , there exist @xmath504 positive for @xmath384 , such that @xmath505 converge to some @xmath506 with @xmath507 , since @xmath384 is simple . apply lemma [ one - sided approx ] , showing @xmath450 . now apply lemma [ width semi - cont ] to obtain @xmath508 . on the 2-torus , the only directions that admit more than one weak kam solution are the rational directions , i.e. those which have a periodic geodesic . * question . * do we have @xmath56 for all @xmath35 not being fixed under any non - trivial group element of @xmath21 ? in this subsection we show that the answer is affirmative , if we make an additional assumption . * assumption ( a ) . * for any @xmath35 , the union @xmath44 of all bounding geodesics @xmath37 with @xmath509 is a lamination of @xmath10 , i.e. the bounding geodesics pairwise do not intersect . \(a ) is fulfilled , e.g. , if @xmath24 has no conjugate points ( in this case all geodesics are unstable , cf . theorem 12.1 in @xcite ) and theorem [ thm countable unstable ] shows that ( a ) is always `` almost '' fulfilled . [ thm assumption a ] if @xmath24 satisfies the assumption ( a ) , then @xmath56 for all @xmath35 not being fixed under any non - trivial group element of @xmath21 . the idea in the following proof is similar to an idea in section 3 of @xcite . . [ fig aussmption_a ] ] figure [ fig aussmption_a ] depicts the following arguments . set @xmath510 assume @xmath511 and choose a small @xmath184 and @xmath17 with @xmath512 let @xmath369 be a positive sequence for @xmath40 , such that @xmath513 for some @xmath503 . we may assume w.l.o.g . , that @xmath514 and @xmath384 is not an axis , since otherwise @xmath57 ( lemmata [ one - sided approx ] and [ periodic approx ] ) . we assume moreover that @xmath515 for all @xmath113 in the counterclockwise orientation of @xmath14 , the other case being analogous . by lemma [ width semi - cont ] , we have @xmath516 there is a point of intersection @xmath248 of the geodesics @xmath517 . let @xmath504 be a positive sequence for @xmath384 , such that @xmath518 and such that @xmath519 converges to some @xmath87 . let @xmath520 be a point of intersection of @xmath521 , then @xmath522 ( the limit of @xmath523 lies left of @xmath395 ) . for @xmath262 , the segment of @xmath524 between @xmath525 becomes a backward ray in the strip between @xmath526 in @xmath527 and since by the assumption ( a ) , no minimal geodesics of different type can intersect , the limit of @xmath524 also belongs to @xmath527 . but now we have in the forward end of @xmath527 to strips of width @xmath528 , one bounded by @xmath529 and the other bounded by @xmath530 , showing @xmath531 and hence @xmath506 is an axis . lemma [ periodic approx ] then shows @xmath450 , a contradiction . in this appendix we prove results about weak kam solutions , which are special in dimension two . even though the results are not used in the paper , the techniques might prove useful in the future . as before , we only assume that @xmath147 are uniformly equivalent for the morse lemma to hold . [ j(u ) semi - cont ] the sets @xmath532 for fixed @xmath222 are sequentially compact in the @xmath225 topology and if @xmath533 with @xmath534 in @xmath225 , then @xmath535 moreover , the asymptotic direction @xmath171 is continuous . by @xmath536 with @xmath238 as in the proof of lemma [ existence horofunction ] , all @xmath174 are equi - lipschitz . it now follows from the arzela - ascoli theorem that any sequence @xmath537 has a convergent subsequence with limit in @xmath244 . assume @xmath534 . if @xmath116 , then @xmath538 let now @xmath539 , such that @xmath108 for some @xmath156 . then by @xmath540 and @xmath541 , we have @xmath542 and hence @xmath543 fixing @xmath544 shows that for all @xmath118 we find a ray @xmath159 with @xmath545 , such that @xmath546 . moreover , if we take a subsequence of @xmath242 , such that @xmath547 for some @xmath35 , then lemma [ as dir cont rays ] shows that @xmath548 for any so obtained @xmath251 and hence @xmath149 . the same arguments show that @xmath549 . conversely , if @xmath208 , let @xmath179 and @xmath539 , such that @xmath550 . lemma [ lemma fathi ] shows @xmath551 and since @xmath552 is closed , we have @xmath553 , hence @xmath554 . [ bounding horofunctions ] let @xmath222 . for each @xmath35 , there exist two unique @xmath560 with the following property : for all sequences @xmath561 with @xmath547 and @xmath562 , any @xmath225 limit lies in @xmath563 . more precisely , assuming the counterclockwise orientation of @xmath14 , we have @xmath564 analogously , for @xmath565 with @xmath566 we have @xmath567 . left : @xmath569 starts off @xmath230 to the left . right : @xmath569 starts off @xmath230 to the right . the gray circles indicate the ( almost ) intersections that contradict lemma [ crossing minimals ] . [ fig bounding_kam ] ] let @xmath570 and assume that in the counterclockwise orientation of @xmath14 , we have @xmath571 . choose any forward weak kam solutions @xmath572 vanishing in a fixed reference point @xmath222 and let @xmath573 be limit functions of the @xmath574 , respectively ( after passing to a subsequence , using continuity of @xmath555 and compactness of @xmath575 ) . we claim that @xmath214 , which also shows @xmath576 by corollary [ j = j then u = u ] . by symmetry , we need to show only @xmath577 . let @xmath296 and @xmath179 . we claim @xmath578 , which shows @xmath365 , since @xmath179 is arbitrary and @xmath364 is closed . by lemma [ j(u ) semi - cont ] , there exist @xmath579 with @xmath108 . by @xmath580 , we find analogously @xmath581 with @xmath582 , converging to a vector @xmath583 . using lemma [ crossing minimals ] , the sketches in figure [ fig bounding_kam ] show a contradiction by considering two cases , if @xmath584 . the above proof shows : if @xmath585 , such that @xmath108 with @xmath586 , then @xmath587 . one can think of @xmath588 as the set of rightmost vectors in @xmath32 , @xmath589 as the leftmost vectors in @xmath32 . e.g. for @xmath593 , consider any sequence of @xmath594 , where @xmath595 , such that @xmath596 in the counterclockwise orientation of @xmath14 . proposition [ bounding horofunctions ] and lemma [ j(u ) semi - cont ] show that @xmath597 and if @xmath598 , @xmath230 has to lie asymptotically left of @xmath38 , which is prohibited by lemma [ partial instability bounding ] . we close by the following nice property of the bounding weak kam solutions . assume that @xmath24 is invariant under @xmath21 and let @xmath599 be the covering map with differential @xmath600 . write @xmath601 for the non - wandering set of @xmath602 in @xmath603 . hence , if one is interested only in recurrent dynamics ( e.g. in the supports of invariant measures ) , one can restrict the attention to the dynamics in the above set , which in the universal cover consists of only two @xmath602-invariant graphs over @xmath80 for each direction @xmath35 . in another paper @xcite , we exploited this idea to calculate the topological entropy of @xmath605 . if @xmath606 , then there exist by definition @xmath607 and @xmath608 , such that @xmath609 in @xmath461 . lifted to @xmath10 , we find @xmath369 , such that @xmath610 and @xmath611 converge to @xmath612 . if the asymptotic direction @xmath613 is a fixed point of @xmath21 , then @xmath612 belongs to the set above by the results in subsecction [ section periodic ] . in the other case , if @xmath33 is not fixed by @xmath21 , the asymptotic directions of @xmath614 and @xmath615 can not all be equal to @xmath33 by the presence of the non - trivial @xmath374 . proposition [ bounding horofunctions ] now proves the claim . _ weak kam theorem in lagrangian dynamics , preliminary version number 10_. preprint ( 2008 ) , available online at http://www.math.u-bordeaux1.fr/~pthieull/recherche/kamfaible/publications.html .

graphene sheets encapsulated between hexagonal boron nitride ( hbn ) slabs display superb electronic properties due to very limited scattering from extrinsic disorder sources such as coulomb impurities and corrugations . such samples are therefore expected to be ideal platforms for highly - tunable low - loss plasmonics in a wide spectral range . in this article we present a theory of collective electron density oscillations in a graphene sheet encapsulated between two hbn semi - infinite slabs ( hbn / g / hbn ) . graphene plasmons hybridize with hbn optical phonons forming hybrid plasmon - phonon ( hpp ) modes . we focus on scattering of these modes against graphene s acoustic phonons and hbn optical phonons , two sources of scattering that are expected to play a key role in hbn / g / hbn stacks . we find that at room temperature the scattering against graphene s acoustic phonons is the dominant limiting factor for hbn / g / hbn stacks , yielding theoretical inverse damping ratios of hybrid plasmon - phonon modes of the order of- , with a weak dependence on carrier density and a strong dependence on illumination frequency . we confirm that the plasmon lifetime is not directly correlated with the mobility : in fact , it can be anti - correlated .

1408.1653

hexagonal boron nitride ( hbn ) , a wide - bandgap insulator , has recently emerged as a sort of `` magic '' substrate for exfoliated graphene sheets . early on , it was demonstrated @xcite that hbn is a much better substrate than @xmath2the ordinary substrate @xcite for much of the early work in graphene physics and devices because its surface is much flatter and because it presents a much smaller number of charged impurities . exfoliated graphene sheets deposited on hbn ( g / hbn ) or graphene sheets that are encapsulated in hbn ( hbn / g / hbn ) have therefore much larger mobilities @xcite than those that are deposited on @xmath2 . subsequently , vertical stacks @xcite comprising graphene and hbn have been used for proof - of - concept devices such as field - effect tunneling transistors @xcite and fundamental studies of electron - electron interactions @xcite . more recent experimental work @xcite has demonstrated that hbn substantially alters the electronic spectrum of the massless dirac fermion ( mdf ) @xcite carriers hosted in a nearby graphene sheet . indeed , when graphene is deposited on hbn , it displays a moir pattern @xcite , a modified tunneling density of states @xcite , and self - similar transport characteristics in a magnetic field @xcite . this spectral reconstruction of the mdf energy - momentum dispersion relation is , however , only relevant in the case of long - wavelength moir superlattices ( superlattice period @xmath3 ) , which occur when the twist angle between the graphene and hbn crystals is small . short - wavelength superlattices yield changes of the mdf spectrum at dopings that are not achievable by electrostatic gating . finally , we would like to mention that the authors of ref . have demonstrated that hbn / g / hbn samples , in which the role of contact resistance is minimized by using a suitable geometry , can display very large mobilities , which are solely limited by scattering of electrons against graphene s acoustic phonons . according to boltzmann - transport theory @xcite , this scattering mechanism yields @xcite a mobility @xmath4 that decreases like @xmath5 with increasing carrier density @xmath6in good agreement with experimental data @xcite and , therefore , a drude dc transport scattering time @xmath7 that decreases like @xmath8 in the same limit . we remind the reader that , in the same theoretical framework and by virtue of screening , scattering against charged impurities yields a mobility that increases with increasing carrier density @xcite . high - quality vertical heterostructures comprising graphene and hbn crystals may have a large impact on the success of graphene plasmonics @xcite , an emerging field of research that has recently attracted a great deal of attention . the goal of graphene plasmonics is to exploit the interaction of infrared light with dirac plasmons " ( dps)the self - sustained density oscillations of the mdf liquid in a doped graphene sheet @xcite for a variety of applications such as infrared @xcite and terahertz @xcite photodetectors , strong light - matter interactions @xcite , enhanced light absorption @xcite and bio - sensing @xcite . interest in graphene plasmonics considerably increased after two experimental groups @xcite showed that the dp wavelength is much smaller than the illumination wavelength , allowing an extreme concentration of electromagnetic energy , and that it is easily gate tunable . these early experiments , based on scattering - type near - field optical spectroscopy ( s - snom ) , were not optimized to minimize dp losses and therefore maximize the plasmon inverse damping ratio . microscopic calculations targeting the role of electron - electron interaction effects @xcite beyond the random phase approximation ( rpa ) and charged impurity scattering @xcite indicate that losses can be strongly reduced by using hbn rather than @xmath2 as a substrate . indeed hbn has both a larger static dielectric constant , thus suppressing the strength of electron - electron interactions , and a much lower impurity concentration than @xmath9 . the impact of electron - phonon scattering on the lifetime of hpp modes was recently addressed in ref . , where the authors showed that the inverse damping ratio of mid - infrared hpp modes is strongly limited by scattering against optical phonons of both the sio@xmath10 substrate and the graphene sheet . the hpp mode damping rate was estimated by introducing a self - energy correction in the _ local _ conductivity @xmath11 , which took into account electron - impurity , electron - phonon , and edge scattering . the contribution of electron - phonon interactions to the hpp lifetime was assumed to be independent of momentum . the dependence of the damping rate on momentum stemmed from the contribution due to the scattering against the edges of the sample . in this article we present a theoretical study of the performance of hbn / g / hbn stacks for applications in the field of graphene plasmonics . more precisely , we present a microscopic theory of the damping rate of hpp modes in a graphene sheet encapsulated between two hbn semi - infinite slabs . we focus on scattering of hpp modes in a graphene sheet against i ) graphene s acoustic phonons and ii ) hbn optical phonons , two sources of scattering that are expected to play a key role in limiting the lifetime of collective density oscillations in hbn / g / hbn stacks . besides achieving good quantitative agreement with recent experimental work @xcite , we confirm the important fact that the plasmon lifetime is not necessarily correlated with the mobility @xcite ( i.e. with the transport lifetime that controls the uniform dc conductivity ) . more accurately , the plasmon lifetime is controlled by the non - local conductivity @xmath12 , where @xmath13 is the wavevector and @xmath14 is the plasmon frequency . in the present case , retaining both the wavevector- and frequency - dependence of the non - local conductivity , we are able to show that the plasmon lifetime is in fact _ anti - correlated _ with the transport mobility . indeed , while the mobility decreases with increasing carrier density @xcite , the plasmon lifetime shows exactly the opposite behavior . this article is organized as follows . in sect . [ sect : qualityfactor ] we introduce the hpp mode inverse damping ratio @xmath15 and relate it to the microscopic _ non - local _ dynamical conductivity @xmath16 of a 2d electron liquid embedded in a medium with a generic frequency - dependent dielectric function @xmath17 . in sect . [ sect : theory ] we present our microscopic theory of the hpp mode dispersion relation and losses in hbn / g / hbn stacks due to electron - phonon scattering . in sects . [ sect : intrinsic_acoustic ] and [ sect : optical_substrate ] we describe the details of the electron - phonon interactions we have considered , i.e. scattering of 2d mdfs against graphene s acoustic phonons and hbn optical phonons , respectively . finally , in sect . [ sect : summary ] we report a summary of our main results and conclusions . we have presented a number of relevant technical details in four appendices , with the aim of making our article as self - contained as possible . in this section we briefly derive general expressions for the inverse damping ratio @xmath15 of a self - sustained oscillation in the density channel of an electron liquid @xcite . to connect our theoretical results to the observables in s - snom experiments @xcite , we assume the mode frequency to be a purely real quantity , which is fixed by the illumination frequency @xmath18 . the mode damping rate @xmath19 is encoded in the imaginary part of the _ complex _ collective mode momentum @xmath20 . more precisely , we define @xmath19 as @xmath21 the inverse damping ratio @xmath15 is the inverse of the damping rate , i.e. @xmath22 . on the general grounds of linear response theory @xcite , the dispersion of hpp modes is calculated by solving the equation @xmath23 where @xmath24 is an effective electron - electron interaction screened by a suitable substrate dielectric function @xmath17 and @xmath25 is the proper density - density linear response function @xcite . in eq . ( [ eq : dressedinteraction ] ) @xmath26 is the 2d fourier transform of the bare coulomb interaction . the precise functional dependence of @xmath17 on the illumination frequency @xmath18 is not specified in this section . we stress that @xmath27 is defined as the solution of eq . ( [ eq : collective_modes_definition ] ) at fixed illumination frequency @xmath18 , while @xmath13 is hereafter a generic wavevector . the causal ( i.e. retarded ) density - density response function @xmath28 can be expressed in terms of the non - local frequency - dependent conductivity @xmath16 as follows @xcite @xmath29 in the long - wavelenght @xmath30 limit and in two spatial dimensions @xcite , @xmath31 . here , @xmath32 is the fermi wave number . since we are interested in describing scattering of collective modes against weak disorder , we approximate the non - local conductivity in the following generalized drude form @xcite : @xmath33 in the spirit of the rpa @xcite , @xmath34 is the drude weight of a system of non - interacting 2d mdfs , @xmath35 , with @xmath36 the mdf fermi energy , @xmath37 the universal optical conductivity @xcite , and @xmath38 the number of fermion flavors @xcite in graphene . corrections beyond rpa to the drude weight , stemming from the lack of galilean invariance of the 2d mdf model , have been worked out in refs . and will be neglected in this work for the sake of simplicity . using eq . ( [ eq : chi_nn_sigma_relation ] ) in eq . ( [ eq : collective_modes_definition ] ) and solving for @xmath39 we get @xmath40 in the limit @xmath41 and neglecting terms containing the product @xmath42 \re e[\sigma(q_1,\omega)]$ ] , eq . ( [ eq : q_p_sigma_epsilon ] ) allows us to write @xmath43}{\re e[\varepsilon_{\rm s}(\omega ) ] } + \frac{\re e [ \sigma(q_1,\omega)]}{\im m [ \sigma(q_1,\omega)]}~,\ ] ] or , equivalently , @xmath44}{\re e[\varepsilon_{\rm s}(\omega ) ] } + \frac{\re e [ \sigma(q_1,\omega)]}{\im m [ \sigma(q_1,\omega)]}}~.\ ] ] eq . ( [ eq : q_definition ] ) is the most important result of this section . the first term on the right - hand side of eq . ( [ eq : gamma_p_definition ] ) , @xmath45}{\re e[\varepsilon_{\rm s}(\omega)]}~,\ ] ] encodes the contribution to the damping rate that is solely controlled by the dissipative component of the substrate dielectric function @xmath17 evaluated at the illumination frequency . the processes that are responsible for this dissipation have nothing to do with the presence of the graphene layer . on the other hand , the second term on the right - hand side of eq . ( [ eq : gamma_p_definition ] ) , can be considered as an _ intrinsic _ damping rate , which involves electronic processes in the graphene layer . indeed , expanding the right - hand side of eq . ( [ eq : sigma_def ] ) in the weak - scattering @xmath46 limit , we immediately find @xmath47}{\im m [ \sigma(q,\omega ) ] } \simeq \frac{1}{\omega \tau(q,\omega)}\ ] ] and @xmath48 any microscopic theory of the collective mode damping rate requires the calculation of the quantity @xmath49 . combining eqs . ( [ eq : chi_nn_sigma_relation ] ) and ( [ eq : sigma_def ] ) we find that , in the weak - disorder @xmath46 limit , @xmath50~.\ ] ] this is a very convenient expression @xcite that will be used below to calculate damping rates due to electron - phonon scattering . in the remainder of this article we set @xmath51 . we consider the following model hamiltonian @xmath52 the first term in eq . ( [ eq : hamiltonian ] ) describes @xmath53-electrons in graphene at the level of a one - orbital tight - binding ( tb ) model . to keep the model as simple as possible , we set to zero all the hopping parameters but the nearest - neighbor one . the low - energy mdf limit will be taken only at the very end of the calculation , _ after _ carrying out all the necessary algebraic manipulations . as extensively discussed in refs . , this procedure allows us to avoid problems associated with the introduction of a rigid ultraviolet cut - off , which breaks gauge invariance @xcite . the non - interacting tb hamiltonian reads @xmath54 where the field operator @xmath55 ( @xmath56 ) creates ( annihilates ) an electron with bloch momentum @xmath57 , belonging to the sublattice @xcite @xmath58 . the quantity @xmath59 is defined as @xcite @xmath60 , -\im m\left[e^{-i { \bm k}\cdot{\bm \delta}_i}\right]\right ) ~.\end{aligned}\ ] ] here @xmath61 is the nearest - neighbor tunneling amplitude , while @xmath62 ( @xmath63 ) are the vectors which connect an atom to its three nearest neighbors , _ i.e. _ @xmath64 , @xmath65 , and @xmath66 . here is the carbon - carbon distance in graphene . the sum over @xmath57 in eq . ( [ eq : sm_non_int_h ] ) is restricted to the first brillouin zone ( bz ) and the pauli matrices @xmath68 ( @xmath69 ) operate on the sublattice degrees of freedom . the tb problem posed by the hamiltonian ( [ eq : sm_non_int_h ] ) can be easily solved analytically @xcite . one finds the following eigenvalues @xmath70 , with @xmath71 . these two bands touch at two inequivalent points ( @xmath72 and @xmath73 ) in the hexagonal bz . the low - energy mdf model is obtained from eq . ( [ eq : sm_non_int_h ] ) by taking the limit @xmath74 , while keeping the product @xmath75 constant . in this limit @xmath76 , where @xmath77 is the density - independent fermi velocity . it turns out to be more convenient to work in an eigenstate representation , in which the tb hamiltonian reads @xmath78 where @xmath79 ( @xmath80 ) creates ( annihilates ) an electron in the single - particle eigenstate with eigenvalue @xmath70 , with @xmath71 . the second term in eq . ( [ eq : hamiltonian ] ) represents long - range coulomb interactions between electrons . in the same representation as in eq . ( [ eq : tb_eigenstate ] ) , the coulomb hamiltonian reads @xcite @xmath81 where the density operator is @xmath82 and @xmath83 is the 2d _ discrete _ fourier transform of the real - space coulomb interaction , which is a periodic function of the reciprocal - lattice vectors . finally , in eq . ( [ eq : sm_density_op ] ) we have introduced the density vertex " @xmath84 with @xmath85 $ ] . here @xmath86 denotes the cartesian components of the vector @xmath59 . in the low - energy mdf limit , @xmath87 , where @xmath88 is the angle between @xmath57 and the @xmath89 axis . note that in writing eq . ( [ eq : sm_interaction_h ] ) we have neglected the one - body operator proportional to the total number of particles , which avoids self - interactions @xcite , since it has no effect on the calculations that we will carry out below . the lifetime of collective modes outside the single particle - hole continuum is indeed determined by two - particle excitations only , which are generated by _ two - body _ operators . finally , the third and fourth term in eq . ( [ eq : hamiltonian ] ) represent the bare phonon hamiltonian and the electron - phonon interaction hamiltonian , respectively . these are given by @xmath90 and @xmath91 here @xmath92 ( @xmath93 ) is the creation ( annihilation ) operator of a phonon belonging to the branch @xmath94 , with momentum @xmath95 measured from the bz center , and energy @xmath96 . here note that eq . ( [ eq : h_ph ] ) is completely general and can be used to describe either graphene s intrinsic phonons , which are bound to the 2d graphene plane , or 3d phonons traveling in hbn . in the former ( latter ) case the phonon momentum @xmath95 is a 2d ( 3d ) vector . in the 3d case we write @xmath98 , where @xmath99 represents the projection of the 3d vector @xmath95 on the graphene plane , while @xmath100 is the component of @xmath95 perpendicular to it . in eq . ( [ eq : h_e_ph ] ) we defined the electron - phonon interaction vertex @xmath101 , which will be specified later in sect . [ sect : intrinsic_acoustic ] and [ sect : optical_substrate ] . we stress again that also in eq . ( [ eq : h_e_ph ] ) the vector @xmath95 can be either 2d or 3d , according to the phonon modes of interest . in the case of optical phonons in the hbn substrate , only the component of the momentum along the graphene plane is conserved in the electron - phonon interaction . in what follows we concentrate on a doped graphene sheet . for the sake of definiteness , we assume the system to be @xmath6-doped . as usual , results for a @xmath102-doped system can be easily obtained by appealing to the particle - hole symmetry of the model defined by eq . ( [ eq : hamiltonian ] ) . as bulk graphite , hbn is a layered material : its response to electric fields is therefore highly anisotropic . let @xmath103 be the axis perpendicular to the hbn planes , while @xmath89 and @xmath104 denote two orthogonal directions parallel to the hbn plane . the hbn dielectric tensor in this basis is diagonal and has the following uniaxial form @xcite , @xmath105 the dependence on frequency of the components @xmath106 of the dielectric tensor of bulk hbn is usually parametrized in the following form @xcite @xmath107 with @xmath108 . the parameters entering in eq . ( [ eq : bn_dielectric ] ) are listed in table [ table_parameters ] and have been taken from recent measurements @xcite on high - quality bulk hbn . here @xmath109 and @xmath110 are the static and high - frequency dielectric constants , respectively , while @xmath111 is the bulk transverse optical phonon frequency in the direction @xmath112 . the bulk longitudinal optical phonon frequency @xmath113 satisfies the lyddane - sachs - teller ( lst ) relation @xcite @xmath114 . .the parameters entering the bulk hbn dielectric functions in eq . ( [ eq : bn_dielectric ] ) . these values have been extracted from ref . .[table_parameters ] [ cols="<,^,^",options="header " , ] in fig . [ fig : four ] we show our results for the hpp mode inverse damping ratio due to scattering against hbn optical phonons . in fig . [ fig : four]a ) @xmath15 is plotted as a function of density @xmath6 and for a fixed illumination wavelength ( solid line ) . by comparing the numbers for @xmath15 with those shown in fig . [ fig : one ] , we conclude that the result is completely dominated by the extrinsic contribution @xmath115 evaluated at @xmath116 . the dashed line in fig . [ fig : four]a ) indicates the result that one obtains by artificially forcing @xmath117 . in fig . [ fig : four]b ) we show the same quantities as in panel a ) but this time plotted as functions of the illumination frequency @xmath18 and for a fixed carrier density @xmath118 . in these plots the temperature has been fixed at the value @xmath119 : since the hbn optical phonon energy scales are very high , we do not expect any significant temperature dependence in the relevant range @xmath120 . we have presented a theory of hybrid plasmon - phonon modes in a graphene sheet encapsulated between two hexagonal boron nitride ( hbn ) semi - infinite slabs ( hbn / g / hbn ) . by using linear response theory and the random phase approximation , we have calculated the dispersion relation of hybrid plasmon - phonon modes that stem from the hybridization between graphene dirac plasmons and hbn optical phonons . the uniaxial hyperbolic nature of hbn yields three plasmon - phonon branches separated by two reststrahlen bands . we have carried out a detailed study of the inverse damping ratio of these plasmon - phonon modes . we have considered two possible sources of scattering limiting their lifetime : scattering against graphene s acoustic phonons and hbn optical phonons . we have discovered that scattering against intrinsic acoustic phonons is the dominant limiting factor in hbn / g / hbn stacks and that it yields theoretical inverse damping ratios of hybrid plasmon - phonon modes that lie in the range @xmath0-@xmath121 at room temperature , with a weak dependence on carrier density and a strong dependence on illumination wavelength . while the current work focuses on room temperature for its relevance to applications , the inverse damping ratio is expected to increase when temperature is lowered and the scattering of electrons with intrinsic acoustic phonons is suppressed . numerical calculations on the temperature dependence of the inverse damping ratio at a fixed carrier density and illumination frequency will be shown elsewhere @xcite . our theoretical predictions indicate that hbn / g / hbn stacks can be very fruitfully used as a low - loss and gate tunable platform for plasmonics in the mid - infrared spectral range . we will show somewhere else @xcite that our results are in excellent quantitative agreement with recent s - snom measurements . it is a great pleasure to thank andrea tomadin and leonid levitov for many useful discussions . this work was supported by the doe grant de - fg02 - 05er46203 ( a.p . and g.v . ) , a research board grant at the university of missouri ( a.p . and g.v . ) , the e.u . through the graphene flagship ( contract no cnect - ict-604391 ) program ( m.c . , m.l . , f.h.l.k . , and m.p . ) , and the italian ministry of education , university , and research ( miur ) through the programs firb - futuro in ricerca 2010 " - project plasmograph ( grant no . rbfr10m5bt ) and progetti premiali 2012 " - project abnanotech ( m.p . ) . m.c . acknowledges also the support of miur - firb2012 - project hybridnanodev ( grant no . rbfr1236vv ) . in this appendix we derive the effective electron - electron interaction in a graphene sheet embedded between two semi - infinite uniaxial dielectrics . we assume to have a graphene sheet lying on the @xmath122 plane and surrounded by two semi - infinite uniaxial dielectrics filling the half spaces @xmath123 and @xmath124 . to determine the effective electron - electron interaction we solve the electrostatic problem @xmath125 \phi({\bm r},z ) = -4\pi e n({\bm r } ) \delta(z)\ ] ] for the 3d electrical potential @xmath126 . here @xmath127 is a 2d vector in the @xmath122 plane , @xmath128 , and @xmath129 is the charge density on the graphene sheet . the in - plane @xmath130 and out - of - plane @xmath131 components of the hbn dielectric tensor have been defined earlier in eq . ( [ eq : bn_dielectric ] ) . we solve eq . ( [ eq : app_electrostatic_problem ] ) in the two half - spaces @xmath124 and @xmath123 with the _ ansatz _ @xmath132 substituting eq . ( [ eq : app_ansatz_es ] ) in eq . ( [ eq : app_electrostatic_problem ] ) we find , for @xmath133 , @xmath134 \phi_{{\bm q}_\perp } = 0~,\ ] ] which is solved by the choice @xmath135 note that @xmath100 becomes imaginary when the product @xmath136 . in this case the _ ansatz _ ( [ eq : app_ansatz_es ] ) describes electromagnetic waves that are not bounded to the graphene sheet and that propagate in all the three spatial directions . the quantity @xmath137 is determined by matching the @xmath123 and @xmath124 solutions at @xmath122 with the `` metallic '' boundary condition imposed by the presence of graphene , i.e. @xmath138 = - 4\pi e~,\ ] ] where @xmath139 is the electron charge . the requested 2d effective electron - electron interaction @xmath140 is simply given by @xmath141 with @xmath142 we start from the definition of the finite - temperature ( matsubara ) linear response function @xmath143 where @xmath144 stands for the average over the finite - temperature density matrix and @xmath145 indicates the imaginary - time ordering operator . we assume that @xmath146 and @xmath147 are bosonic operators built with an even number of fermionic operators . in the frequency representation @xmath148 reads as follows : @xmath149 where @xmath150 is a bosonic matsubara frequency and @xmath151 is the inverse temperature . more specifically , @xmath152 and @xmath153 where the operator @xmath154 has been introduced earlier in eq . ( [ eq : upsilon_approx ] ) . since @xmath155 in eq . ( [ eq : chi_ab_omega_def ] ) , we can remove the imaginary - time ordering operator @xmath145 on the right - hand side of eq . ( [ eq : chi_ab_tau_def ] ) . furthermore , we approximate @xmath156 { \hat \upsilon}_{{\bm q},{\bm q}'_\perp}(\tau ) \nonumber\\ & \times & ( { \hat a}^\dagger_{-{\bm q}'',\nu ' } + { \hat a}_{{\bm q}'',\nu ' } ) { \hat \upsilon}_{-{\bm q},-{\bm q}''_\perp } \big\rangle \nonumber\\ & \simeq & \big\langle\ ! \big[{\hat a}^\dagger_{{\bm q}',\nu}(\tau ) + { \hat a}_{-{\bm q}',\nu}(\tau)\big ] ( { \hat a}^\dagger_{-{\bm q}'',\nu ' } + { \hat a}_{{\bm q}'',\nu ' } ) \!\big\rangle \nonumber\\ & \times & \big\langle\ ! { \hat \upsilon}_{{\bm q},{\bm q}'_\perp}(\tau ) { \hat \upsilon}_{-{\bm q},-{\bm q}''_\perp } \!\big\rangle \nonumber\\ & = & \delta_{{\bm q}',{\bm q } '' } \delta_{\nu,\nu ' } { \cal d}^{({\rm ph})}_\nu({\bm q}',\tau ) \chi_{\upsilon_{{\bm q},{\bm q'}_\perp } \upsilon_{-{\bm q},-{\bm q}'_\perp}}^{\cal t}(\tau ) ~,\end{aligned}\ ] ] where we retained only the wick pairings that are dominant in the large-@xmath157 limit . using the last line of eq . ( [ chi_tau_approx ] ) in eq . ( [ eq : chi_ab_omega_def ] ) and the identity @xmath158 which is valid for any bosonic matsubara frequency @xmath159 , we get @xmath160 the matsubara sum in the last line of eq . ( [ eq : chi_jnjn_omega ] ) can be transformed into an integral over a contour in the complex plane that encircles the poles of the bose - einstein occupation factor @xmath161 . in doing so , we exclude the branch cuts of @xmath162 which occur for @xmath163 . after the analytical continuation to real frequencies we find @xmath164 \nonumber\\ & \times & \big [ f_{++}(\omega',\omega-\omega ' ) - f_{--}(\omega',\omega-\omega ' ) \big ] \nonumber\\ & + & n_{\rm b}(\omega ' ) \big [ f_{+-}(\omega',\omega-\omega ' ) - f_{-+}(\omega',\omega-\omega ' ) \big ] \big\ } ~ , \nonumber\\\end{aligned}\ ] ] where @xmath165 and @xmath166 . note that the term in square bracket in the last line of eq . ( [ chi_jn_jn_causal_2 ] ) is purely imaginary . since it is multiplied by the imaginary unit , its contribution to the integral is purely real . taking the imaginary part of eq . ( [ chi_jn_jn_causal_2 ] ) we finally find @xmath167 \nonumber\\ & \times & \im m\chi_{\upsilon_{{\bm q},{\bm q'}_\perp } \upsilon_{-{\bm q},-{\bm q}'_\perp}}(\omega - \omega ' ) ~.\end{aligned}\ ] ] in the limit @xmath168 , and for @xmath169 , eq . ( [ im_chi_jn_jn_causal ] ) becomes @xmath170 in this appendix we show that , in the limit @xmath171 and @xmath172 , eq . ( [ eq : lifetime_final_t ] ) reproduces the dc transport time @xmath7 for scattering of electrons against graphene s acoustic phonons as found e.g. in ref . . in the limit above , the matrix element @xmath173 defined in eq . ( [ eq : upsilon_approx ] ) reduces to @xmath174 substituting eq . ( [ eq : app_matr_elem_transport_limit ] ) in eq . ( [ eq : lifetime_final_t ] ) we therefore get @xmath175\im m~[\chi_{j_\alpha j_\beta}^{(0 ) } ( { \bm q}',\omega ' ) ] ~.\end{aligned}\ ] ] we now recall @xcite that , in a homogeneous and isotropic electron liquid , the current - current linear response tensor can be decomposed in the following manner : @xmath176 where @xmath177 and @xmath178 are the so - called @xcite longitudinal and transverse current - current response functions , respectively . using eq . ( [ eq : current_decomposition ] ) we can write eq . ( [ eq : transport_time_1 ] ) in the following form : @xmath179\im m~[\chi_{nn}^{(0 ) } ( q',\omega')]~.\end{aligned}\ ] ] here we used eq . ( [ eq : continuity_equation ] ) to express the longitudinal current - current response function in terms of the density - density response function . using eq . ( [ eq : acoustic_phonon_interaction ] ) in eq . ( [ eq : transport_time_2 ] ) and the imaginary part of the phonon propagator in the absence of phonon damping , i.e. @xmath180 & = & -\pi[\delta(\omega'-\omega_{\rm ph}(q ' ) ) \nonumber \\ & - & \delta(\omega'+\omega_{\rm ph}(q'))]~,\end{aligned}\ ] ] we find @xmath181~.\end{aligned}\ ] ] note that the two terms on the right - hand side of eq . ( [ eq : phonon_propagator_nodamping ] ) give identical contributions to the integral in eq . ( [ eq : transport_time_2 ] ) : this explains the factor two on the right - hand side of eq . ( [ eq : transport_time_3 ] ) . we now specialize eq . ( [ eq : transport_time_3 ] ) to the case in which the phonon frequency is much smaller than both temperature and fermi energy . in this case we use the following approximate expressions : @xmath182 and @xmath183 & \to & -\theta(2 k_{\rm f } - q')\frac{\omega'}{v_{\rm f } q ' } \frac{n_{\rm f } k_{\rm f}}{2\pi v_{\rm f } } \nonumber \\ & \times & \sqrt{1-\frac{q'^2}{4k_{\rm f}^2}}~.\end{aligned}\ ] ] after some straightforward algebra eq . ( [ eq : transport_time_3 ] ) gives @xmath184 which coincides with the result reported in ref . . in this appendix we derive the interaction hamiltonian between electrons in graphene and optical phonons in hbn . in what follows we shorten our notation by setting @xmath185 . following ref . we start from the following equations @xmath186 which describe the coupling between the electric field @xmath187 and lattice motion , which is encoded in the displacement field @xmath188 between the two atoms in the hbn unit cell . the latter induces a polarization @xmath189 , which adds to the electric field to produce the electric displacement @xmath190 . finally , @xmath191 is the reduced mass and @xmath192 the number of cells in the unit volume . [ eq : stroscio_ft ] @xmath195 e_\ell({\bm q } , \omega ) ~ , \label{eq : stroscio_ft_2 } \\ & e_\ell({\bm q},\omega ) = -i q_\ell \phi({\bm q},\omega ) ~ , \label{eq : stroscio_ft_3 } \\ & q_\perp \varepsilon_x(\omega ) e_\perp({\bm q},\omega ) + q_\parallel \varepsilon_z(\omega ) e_\parallel({\bm q},\omega ) = 0 ~. \label{eq : stroscio_ft_4}\end{aligned}\ ] ] eq . ( [ eq : stroscio_ft_3 ] ) implies that the electric field is purely longitudinal , _ i.e. _ it is parallel to @xmath95 . combining eqs . ( [ eq : stroscio_ft_3 ] ) and ( [ eq : stroscio_ft_4 ] ) we get the following necessary condition @xmath196 to make sure that @xmath197 vanishes . the solutions @xmath198 of the previous equation are the so - called `` extraordinary '' phonon frequencies @xcite . these modes are neither longitudinal nor transverse , except in the limiting cases of @xmath95 oriented parallel or perpendicular to the @xmath103 axis . indeed , eq . ( [ eq : stroscio_ft_1 ] ) implies that in general the lattice displacement is neither parallel nor perpendicular to the electric field and therefore to @xmath95 . recall that @xmath199 and @xmath95 are always parallel . we now define @xmath200 as the lattice displacement associated to the normal mode characterized by the frequency @xmath96 . since normal modes are orthogonal to each other , combining eqs . ( [ eq : stroscio_ft_1 ] ) and ( [ eq : stroscio_ft_3 ] ) we get the electrostatic potential @xmath201 once this quantity is quantized , it completely determines the electron - phonon interaction hamiltonian , which is given by @xmath202 . in what follows we proceed to quantize eq . ( [ eq : phi_u ] ) . ( [ eq : stroscio_ft_1 ] ) implies that the displacement can be written as @xmath203 ^ 2 } \frac{q_j^2}{q^2 } \right]^{-1/2 } \nonumber\\ & \times & \frac{\omega_\ell \sqrt{\epsilon_{\ell,0 } - \epsilon_{\ell,\infty}}}{\omega_\ell^2 - \omega_{{\rm ph},\nu}^2({\bm q } ) } \frac{q_\ell}{q } u_{\nu}({\bm q } ) ~,\end{aligned}\ ] ] where @xmath204 is the modulus of the displacement vector @xmath205 , _ i.e. _ @xmath206 . the prefactor of @xmath204 on the right - hand side of eq . ( [ eq : u_ell_components ] ) is uniquely determined by the angle between the vectors @xmath205 and @xmath95 , and can easily be derived from eq . ( [ eq : stroscio_ft_1 ] ) . quantizing the lattice displacement in the usual manner , i.e. @xmath207 , with @xmath208 the volume of the system , we obtain @xmath209 ^ 2 } \frac{q_j^2}{q^2 } \right]^{-1/2 } \nonumber\\ & \times & \frac{\omega_\ell \sqrt{\epsilon_{\ell,0 } - \epsilon_{\ell,\infty}}}{\omega_{{\rm ph},\nu}^2({\bm q } ) - \omega_\ell^2 } \frac{q_\ell}{q } \big ( a_{{\bm q},\nu } + a^\dagger_{-{\bm q},\nu } \big ) ~.\end{aligned}\ ] ] with the help of eq . ( [ eq : bn_dielectric ] ) it is easy to show that the term in square brackets on the right - hand side of eq . ( [ eq : u_ell_def ] ) gives @xmath210 ^ 2 } \frac{q_j^2}{q^2 } \nonumber\\ & = & \frac { \big [ q_\parallel^2 \partial_\omega \varepsilon_z(\omega ) + q_\perp^2 \partial_\omega \varepsilon_x(\omega ) \big]_{\omega=\omega_{{\rm ph},\nu}({\bm q})}}{2 q^2 \omega_{{\rm ph},\nu}({\bm q } ) } ~. \nonumber\\\end{aligned}\ ] ] when eq . ( [ eq : square_bracket_derivative_epsilon ] ) is used in eq . ( [ eq : u_ell_def ] ) we obtain @xmath211_{\omega=\omega_{{\rm ph},\nu}({\bm q})}^{1/2}}~. \nonumber\\\end{aligned}\ ] ] combining eqs . ( [ eq : phi_u ] ) and ( [ eq : u_ell_def_2 ] ) we finally get @xmath212_{\omega=\omega_{{\rm ph},\nu}({\bm q } ) } } } \nonumber\\ & \times & \big ( a_{{\bm q},\nu } + a^\dagger_{-{\bm q},\nu } \big ) ~. \nonumber\\\end{aligned}\ ] ] this in turn implies that the electron - phonon hamiltonian is @xmath213 where @xmath214_{\omega=\omega_{{\rm ph},\nu}({\bm q } ) } } } ~.\ ] ] let us briefly comment this equation . first of all , if the system is isotropic , _ i.e. _ @xmath215 , the frequencies of the normal modes become @xmath216 ( we omit the direction index , since the two are equivalent ) . the second solution , corresponding to a purely transverse mode , should be disregarded ( see below ) . substituting @xmath217 one immediately recovers the usual frhlich hamiltonian . moreover , when @xmath95 is parallel ( perpendicular ) to the @xmath103 axis @xmath218 coincides with @xmath219 ( @xmath220 ) , while @xmath221 is equal to @xmath222 ( @xmath223 ) . this in turn implies that the mode @xmath224 is purely longitudinal , while @xmath225 describes a transverse optical phonon . however , the denominator of eq . ( [ eq : el_phon_hamilt ] ) , _ i.e. _ @xmath226_{\omega=\omega_{{\rm ph},\nu}({\bm q})}$ ] , diverges when @xmath96 coincides with one of the transverse frequencies , thus excluding the transverse modes from the electron - phonon interaction . by adding the phonon - mediated electron - electron interactions , derived from eq . ( [ eq : el_phon_hamilt ] ) , to the bare coulomb potential one finds the effective 3d electron - electron interactions @xmath227_{\omega=\omega_\nu({\bm q } ) } } ~ , \nonumber\\\end{aligned}\ ] ] which turns out to be equal to @xmath228 note that it is possible to recover the effective 2d electron - electron interaction in eq . ( [ eq : app_effective_interaction ] ) by integrating eq . ( [ eq : v_eff_electrostatics ] ) over @xmath100 . moreover , eqs . ( [ eq : v_eff_frohlich ] ) and ( [ eq : v_eff_electrostatics ] ) allows us to write @xmath229 this equation is used in sect . [ sect : optical_substrate ] , where the integral @xmath230 is rewritten as @xmath231 with the help of eq . ( [ eq : app_stilde_def ] ) . ( [ eq : frohlich_reduction_formula ] ) in particular implies that @xmath232 & = & \im m \left[\frac{2\pi e^2}{q_\perp \sqrt{\varepsilon_x(\omega ) \varepsilon_z(\omega)}}\right ] \nonumber\\ & \equiv & \im m\big[v(q_\perp,\omega)\big ] ~.\end{aligned}\ ] ] 77 c.r . dean , a.f . young , i. meric , c. lee , l. wang , s. sorgenfrei , k. watanabe , t. taniguchi , p. kim , k.l . shepard , and j. hone , http://dx.doi.org/10.1038/nnano.2010.172[nature nanotech . * 5 * , 722 ( 2010 ) ] . a.k . geim and k.s . novoselov , http://dx.doi.org/10.1038/nmat1849[nature mater . * 6 * , 183 ( 2007 ) ] . novoselov and a.h . castro neto , http : dx.doi.org/10.1088/0031 - 8949/2012/t146/014006[phys . t146 * , 014006 ( 2012 ) ] ; f. bonaccorso , a. lombardo , t. hasan , z. sun , l. colombo , and a.c . ferrari , http://dx.doi.org/10.1016/s1369-7021(13)70014-2[mater . today * 15 * , 564 ( 2012 ) ] ; a.k . geim and i.v . grigorieva , http://dx.doi.org/10.1038/nature12385[nature * 499 * , 419 ( 2013 ) ] . l. britnell , r.v . gorbachev , r. jalil , b.d . belle , f. schedin , a. mishchenko , t. georgiou , m.i . katsnelson , l. eaves , s.v . morozov , n.m.r . peres , j. leist , a.k . geim , k.s . novoselov , and l.a . ponomarenko , http://dx.doi.org/10.1126/science.1218461[science * 335 * , 947 ( 2012 ) ] . elias , r.v . gorbachev , a.s . mayorov , s.v . morozov , a.a . zhukov , p. blake , l.a . ponomarenko , i.v . grigorieva , k.s . novoselov , f. guinea , and a.k . geim , http://dx.doi.org/10.1038/nphys2049[nature phys . * 7 * , 701 ( 2011 ) ] . gorbachev , a.k . geim , m.i . katsnelson , k.s . novoselov , t. tudorovskiy , i.v . grigorieva , a.h . macdonald , s.v . morozov , k. watanabe , t. taniguchi , l.a . ponomarenko , http://dx.doi.org/10.1038/nphys2441[nature phys . * 8 * , 896 ( 2012 ) ] . yu , r. jalil , b. belle , a.s . mayorov , p. blake , f. schedin , s.v . morozov , l.a . ponomarenko , f. chiappini , s. wiedmann , u. zeitler , m.i . katsnelson , a.k . geim , k.s . novoselov , and d.c . elias , http://dx.doi.org/10.1073/pnas.1300599110[proc . natl . acad . ( usa ) * 110 * , 3282 ( 2013 ) ] . m. yankowitz , j. xue , d. cormode , j.d . sanchez - yamagishi , k. watanabe , t. taniguchi , p. jarillo - herrero , p. jacquod , and b.j . leroy , http://dx.doi.org/10.1038/nphys2272[nature phys . * 8 * , 382 ( 2012 ) ] . ponomarenko , r.v . gorbachev , g.l . elias , r. jalil , a.a . patel , a. mishchenko , a.s . mayorov , c.r . woods , j.r . wallbank , m. mucha - kruczynski , b.a . piot , m. potemski , i.v . grigorieva , k.s . novoselov , f. guinea , v.i . falko , and a.k . geim , http://dx.doi.org/10.1038/nature12187[nature * 497 * , 594 ( 2013 ) ] . dean , l. wang , p. maher , c. forsythe , f. ghahari , y. gao , j. katoch , m. ishigami , p. moon , m. koshino , t. taniguchi , k. watanabe , k.l . shepard , j. hone , and p. kim , http://dx.doi.org/10.1038/nature12186[nature * 497 * , 598 ( 2013 ) ] . b. hunt , j.d . sanchez - yamagishi , a.f . young , m. yankowitz , b.j . leroy , k. watanabe , t. taniguchi , p. moon , m. koshino , p. jarillo - herrero , and r.c . ashoori , http://dx.doi.org/10.1126/science.1237240[science * 340 * , 1427 ( 2013 ) ] . castro neto , f. guinea , n.m.r . peres , k.s . novoselov , and a.k . geim , http://dx.doi.org/10.1103/revmodphys.81.109 [ * 81 * , 109 ( 2009 ) ] . j. xue , j. sanchez - yamagishi , d. bulmash , p. jacquod , a. deshpande , k. watanabe , t. taniguchi , p. jarillo - herrero , and b.j . leroy , http://dx.doi.org/10.1038/nmat2968[nature mater . * 10 * , 282 ( 2011 ) ] . r. decker , y. wang , v.w . brar , w. regan , h .- z . tsai , q. wu , w. gannett , a. zettl , and m.f . crommie , http://dx.doi.org/10.1021/nl2005115[nano lett . * 11 * , 2291 ( 2011 ) ] . l. wang , i. meric , p.y . huang , q. gao , y. gao , h. tran , t. taniguchi , k. watanabe , l.m . campos , d.a . muller , j. guo , p. kim , j. hone , k.l . shepard , and c.r . dean , http://dx.doi.org/10.1126/science.1244358[science * 342 * , 614 ( 2013 ) ] . s. das sarma , s. adam , e.h . hwang , and e. rossi , http://dx.doi.org/10.1103/revmodphys.83.407 [ * 83 * , 407 ( 2011 ) ] . hwang and s. das sarma , http://dx.doi.org/10.1103/physrevb.77.115449 [ * 77 * , 115449 ( 2008 ) ] . koppens , d.e . chang , and f.j . garca de abajo , http://dx.doi.org/10.1021/nl201771h[nano lett . * 11 * , 3370 ( 2011 ) ] ; a.n . grigorenko , m. polini , and k.s . novoselov , http://dx.doi.org/10.1038/nphoton.2012.262[nature photon . * 6 * , 749 ( 2012 ) ] ; t. stauber , http://dx.doi.org/10.1088/0953-8984/26/12/123201[j . phys . : condens . matter * 26 * , 123201 ( 2014 ) ] ; t. low and p. avouris , http://dx.doi.org/10.1021/nn406627u[acs nano * 8 * , 1086 ( 2014 ) ] ; f.j . garca de abajo , http://dx.doi.org/10.1021/ph400147y[acs photonics * 1 * , 135 ( 2014 ) ] . see , for example , b. wunsch , t. stauber , f. sols , and f. guinea , http://dx.doi.org/10.1088/1367-2630/8/12/318[new j. phys . * 8 * , 318 ( 2006 ) ] ; e.h . hwang and s. das sarma , http://dx.doi.org/10.1103/physrevb.75.205418 [ * 75 * , 205418 ( 2007 ) ] ; m. polini , r. asgari , g. borghi , y. barlas , t. pereg - barnea , and a.h . macdonald , http://dx.doi.org/10.1103/physrevb.77.081411 [ * 77 * , 081411(r ) ( 2008 ) ] ; a. principi , m. polini , and g. vignale , http://dx.doi.org/10.1103/physrevb.80.075418 [ * 80 * , 075418 ( 2009 ) ] . m. freitag , t. low , w. zhu , h. yan , f. xia , and p. avouris , http://dx.doi.org/10.1038/ncomms2951[nature commun . * 4 * , 1951 ( 2013 ) ] . l. vicarelli , m.s . vitiello , d. coquillat , a. lombardo , a.c . ferrari , w. knap , m. polini , v. pellegrini , and a. tredicucci , http://dx.doi.org/10.1038/nmat3417[nature mater . * 11 * , 865 ( 2012 ) ] ; d. spirito , d. coquillat , s.l . de bonis , a. lombardo , m. bruna , a.c . ferrari , v. pellegrini , a. tredicucci , w. knap , and m.s . vitiello , http://dx.doi.org/10.1063/1.4864082 [ * 104 * , 061111 ( 2014 ) ] . koppens , d.e . chang , and f.j.g . de abajo , http://dx.doi.org/10.1021/nl201771h[nano lett . * 11 * , 3370 ( 2011 ) ] . s. thongrattanasiri , f.h.l . koppens , and f.j.c . de abajo , http://dx.doi.org/10.1103/physrevlett.108.047401 [ * 108 * , 047401 ( 2012 ) ] . kravets , f. schedin , r. jalil , l. britnell , r.v . gorbachev , d. ansell , b. thackray , k.s . novoselov , a.k . geim , a.v . kabashin , and a.n . grigorenko , http://dx.doi.org/10.1038/nmat3537[nature mater . * 12 * , 304 ( 2013 ) ] . y. li , h. yan , d.b . farmer , x. meng , w. zhu , r.m . osgood , t.f . heinz , and p. avouris , http://dx.doi.org/10.1021/nl404824w[nano lett . * 14 * , 1573 ( 2014 ) ] . z. fei , a.s . rodin , g.o . andreev , w. bao , a.s . mcleod , m. wagner , l.m . zhang , z. zhao , m. thiemens , g. dominguez , m.m . fogler , a.h . castro neto , c.n . lau , f. keilmann , and d.n . basov , http://dx.doi.org/10.1038/nature11253[nature * 487 * , 82 ( 2012 ) ] . j. chen , m. badioli , p. alonso - gonzlez , s. thongrattanasiri , f. huth , j. osmond , m. spasenovi , a. centeno , a. pesquera , p. godignon , a. zurutuza elorza , n. camara , f.j . garca de abajo , r. hillenbrand , and f.h.l . koppens , http://dx.doi.org/10.1038/nature11254[nature * 487 * , 77 ( 2012 ) ] . a. principi , g. vignale , m. carrega , and m. polini , http://dx.doi.org/10.1103/physrevb.88.195405 [ * 88 * , 195405 ( 2013 ) ] . a. principi , g. vignale , m. carrega , and m. polini , http://dx.doi.org/10.1103/physrevb.88.121405 [ * 88 * , 121405(r ) ( 2013 ) ] . h. yan , t. low , w. zhu , y. wu , m. freitag , x. li , f. guinea , p. avouris , and f. xia , http://dx.doi.org/10.1038/nphoton.2013.57[nature photon . * 7 * , 394 ( 2013 ) ] . et al_. , to appear soon on arxiv . gerber , s. berweger , b.t . ocallahan , and m.b . raschke , http://dx.doi.org/10.1103/physrevlett.113.055502 [ * 113 * , 055502 ( 2014 ) ] . g.f . giuliani and g. vignale , _ quantum theory of the electron liquid _ ( cambridge university press , cambridge , 2005 ) . d. pines and p. nozires , _ the theory of quantum liquids _ ( w.a . benjamin inc . , new york , 1966 ) . kotov , b. uchoa , v.m . pereira , f. guinea , and a.h . castro neto , http://dx.doi.org/10.1103/revmodphys.84.1067 [ * 84 * , 1067 ( 2012 ) ] . abedinpour , g. vignale , a. principi , m. polini , w .- k . tse , and a.h . macdonald , http://dx.doi.org/10.1103/physrevb.84.045429 [ * 84 * , 045429 ( 2011 ) ] . levitov , a.v . shtyk , and m.v . feigelman , http://dx.doi.org/10.1103/physrevb.88.235403 [ * 88 * , 235403 ( 2013 ) ] . r. geick , c.h . perry , and g. rupprecht , http://dx.doi.org/10.1103/physrev.146.543[phys . rev . * 146 * , 543 ( 1966 ) ] . yu and m. cardona , _ fundamentals of semiconductors _ ( springer - verlag , berlin , 1996 ) . caldwell , a. kretinin , y. chen , v. giannini , m.m . fogler , y. francescato , c.t . ellis , j.g . tischler , c.r . woods , a.j . giles , m. hong , k. watanabe , t. taniguchi , s.a . maier , and k.s . novoselov , http://xxx.lanl.gov/abs/1404.0494[arxiv:1404.0494 ] . b. amorim , j. schiefele , f. sols , and f. guinea , http://dx.doi.org/10.1103/physrevb.86.125448 [ * 86 * , 125448 ( 2012 ) ] . s. dai , z. fei , q. ma , a.s . rodin , m. wagner , a.s . mcleod , m.k . liu , w. gannett , w. regan , k. watanabe , t. taniguchi , m. thiemens , g. dominguez , a.h . castro neto , a. zettl , f. keilmann , p. jarillo - herrero , m.m . fogler , and d.n . basov , http://dx.doi.org/10.1126/science.1246833[science * 343 * , 1125 ( 2014 ) ] . a. poddubny , i. iorsh , p. belov , and y. kivshar , http://dx.doi.org/10.1038/nphoton.2013.243[nature photon . * 7 * , 948 ( 2013 ) ] . koch , th . seyller , and j.a . schaefer , http://dx.doi.org/10.1103/physrevb.82.201413 [ * 82 * , 201413(r ) ( 2010 ) ] . hwang , r. sensarma , and s. das sarma , http://dx.doi.org/10.1103/physrevb.82.195406 [ * 82 * , 195406 ( 2010 ) ] . mahan , _ many - particle physics _ , ( plenum press , new york , 1993 ) . tse and s. das sarma , http://dx.doi.org/10.1103/physrevb.79.235406 [ * 79 * , 235406 ( 2009 ) ] . viljas and t.t . heikkil , http://dx.doi.org/10.1103/physrevb.81.245404 [ * 81 * , 245404 ( 2010 ) ] . t. stauber , n.m.r . peres , and f. guinea , http://dx.doi.org/10.1103/physrevb.76.205423 [ * 76 * , 205423 ( 2007 ) ] . f.t . vasko and v. ryzhii , http://dx.doi.org/10.1103/physrevb.76.233404 [ * 76 * , 233404 ( 2007 ) ] . kubakaddi , http://dx.doi.org/10.1103/physrevb.79.075417 [ * 79 * , 075417 ( 2009 ) ] . r. bistritzer and a.h . macdonald , http://dx.doi.org/10.1103/physrevlett.102.206410 [ * 102 * , 206410 ( 2009 ) ] . park , n. bonini , t. sohier , g. samsonidze , b. kozinsky , m. calandra , f. mauri , and n. marzari , http://dx.doi.org/10.1021/nl402696q[nano lett . * 14 * , 1113 ( 2014 ) ] . t. sohier , m. calandra , c - h . park , n. bonini , n. marzari , and f. mauri , http://xxx.lanl.gov/abs/1407.0830[arxiv:1407.0830 ] . more precisely , the value of @xmath233 is determined by constrainining the theoretical mobility @xmath4 , calculated on the basis of eq . ( [ eq : mobility ] ) , to yield the value @xmath234 at a carrier density @xmath235 and a temperature @xmath236 . this implies a deformation potential @xmath237 . once the value of the deformation potential has been found , we can make predictions for @xmath4 as a function of carrier density . b.c . lee , k.w . kim , m. dutta , and m.a . stroscio , http://dx.doi.org/10.1103/physrevb.56.997 [ * 56 * , 997 ( 1997 ) ] . a. principi _ et al . _ , to be submitted .

in this paper we study various particle physics effects of a light , scalar dark energy field with chameleon - like couplings to matter . we show that a chameleon model with only matter couplings will induce a coupling to photons . in doing so , we derive the first microphysical realization of a chameleonic dark energy model coupled to the electromagnetic field strength . this analysis provides additional motivation for current and near - future tests of axion - like and chameleon particles . we find a new bound on the coupling strength of chameleons in uniformly coupled models . we also study the effect of chameleon fields on higgs production , which is relevant for hadron colliders . these are expected to manufacture higgs particles through weak boson fusion , or associated production with a or . we show that , like the tevatron , the lhc will not be able to rule out or observe chameleons through this mechanism , because gauge invariance of the low energy lagrangian suppresses the corrections that may arise . diags

0911.1267

we now have compelling evidence that the rate of expansion of the universe is accelerating , requiring the universe to be dominated by an unknown form of matter , known as ` dark energy , ' characterized by the equation of state @xmath2 . this dark energy could be a cosmological constant unnaturally tuned at the level of 1 part in @xmath3 @xcite , or it may be associated with the potential energy of a scalar field @xcite . if a scalar field drives the accelerated expansion then it must be extremely light , with mass of order @xmath4 . the existence of light scalar fields results in new , long range `` fifth forces , '' which are tightly constrained by experiment @xcite . to avoid these constraints the energy scale controlling the coupling of such scalar fields to matter must be many orders of magnitude larger than the planck scale @xcite . explaining why this coupling is so weak is a major problem for dynamical dark energy models . these problems can be circumvented by allowing the scalar field mass mass to be determined by a _ chameleon _ mechanism @xcite . this makes the field heavy in a dense environment , but light in vacuum . such scalar fields hide from experimental searches for fifth forces @xcite and modifications of gravity @xcite in a novel way : in the interior of a massive object the chameleon is very heavy , and has a correspondingly short interaction length . an observer outside the body feels a scalar force sourced only by a thin shell of matter at the surface of the object . this suppresses unwanted fifth forces @xcite . observations impose few constraints on the strength of the coupling of the scalar field to the standard model , allowing the energy scale of the coupling to be many orders of magnitude below the planck scale @xcite . chameleonic scalar fields have been shown to have a successful classical phenomenology . their equation of state depends on the energy density of the universe . in the early universe the chameleon behaves as an additional matter component , but at late times it has an equation of state with roughly @xmath5 . this enables it to drive an era of accelerated expansion @xcite . weak constraints on the initial conditions ensure that the chameleon does not disrupt the dynamics of the early universe @xcite . in this paper we discuss these theories in the framework of quantum mechanics . this is of interest because the cosmological constant problem is essentially quantum mechanical . there is no profit in replacing the cosmological constant by some other theory which is equally unnatural once quantum corrections are taken into account . the chameleon mechanism makes it mandatory for the dark energy scalar field to interact with conventional matter , and these interactions are potentially very strong . they may help us unravel the microphysics of dark energy if their indirect consequences could be measured @xcite . if the interactions are sufficiently strong , however , then an unambiguous _ direct _ consequence may be observable : new , light scalar quanta must be present in the beam pipe of any particle accelerator , raising the prospect of observing dark energy in the laboratory . fully consistent quantum mechanical realizations of the chameleon have not yet been constructed , and may suffer from the same naturalness and tuning difficulties as quintessence @xcite . in this paper we do not attempt to address naturalness concerns , or the quantum - mechanical construction of the model . instead , supposing such models to exist , we determine constraints which are imposed by experiment . real or virtual chameleon - like particles will certainly be produced in particle collisions if dark energy couples to the electroweak gauge bosons . in this paper , we will provide a microscopic derivation of this coupling for any chameleon - like model . the models which furnish chameleon - like scalar particles at low energy are effective theories , valid below some energy scale @xmath6 . in ref . @xcite , corrections to electroweak precision observables from particles with chameleonic or axionic couplings were analysed , without making a commitment to any specific choice of physics at energies above @xmath6 . it was shown that very weak constraints on the energy scale of the chameleon coupling to matter ensure that the chameleon does not have an observable effect on measurements of the @xmath0-width . for processes involving fermions and @xmath7 gauge bosons , it was shown that large effects could always be absorbed into renormalizations of the fermi constant , @xmath8 , and the gauge boson masses . when higgs processes are included it is no longer clear that large effects can be hidden in this way , potentially allowing higgs production to function as a diagnostic of dark energy physics and its coupling to the standard model . in this paper , we focus on the consequences of such dark energy quanta for higgs production . this is a key target for both the _ large hadron collider _ ( lhc ) and the _ tevatron_. the details of higgs production depend on which couplings occur in the theory . to achieve a successful chameleon phenomenology we assume conformal couplings to matter . this has implications for interactions with gauge boson kinetic terms , which will be described in more detail in [ sec : scalar - corrections ] below . the standard model gauge bosons have conformally invariant kinetic terms and develop no couplings to a scalar field of this type . in this paper we show that violations of conformal symmetry , arising from couplings to matter species , can be communicated to the kinetic term by quantum corrections . therefore , a coupling _ is _ generated at energies below @xmath6 via loops of charged heavy particles . this coupling has important phenomenological consequences which will be briefly recalled . in particular we will show that this coupling induces a coupling to the electromagnetic field and in so doing provides a motivation for the effects probed in quantum laser experiments @xcite . our results apply for any light scalars with the requisite chameleon - like couplings , whatever their origin . what scale , @xmath6 , should be associated with these new degrees of freedom ? a conservative choice would be the gut scale , @xmath9 gev , which may be connected with an early inflationary stage , or a seesaw explanation of neutrino mass @xcite . if so , heavy neutrinos of mass @xmath6 might exist but would be sterile , having no @xmath7 quantum numbers . this would not lead to the required coupling . alternatively , if supersymmetry is realized in nature then @xmath6 could be associated with the scale at which it is spontaneously broken , perhaps of order @xmath10@xmath11 tev . in any supersymmetric completion of the standard model there exist fermionic partners of @xmath12 , @xmath1 , @xmath0 and the higgs , known as gauginos and higgsinos . ( for a review see @xcite and references therein . ) the mass eigenstates of these particles ( `` charginos '' ) would naturally be of order @xmath6 . in this article we remain agnostic about the nature of whatever particles circulate within loops . we give our calculation in a form which can be specialized immediately to the minimal supersymmetric standard model ( mssm ) , but our conclusions are general and do not depend on the details of a specific implementation . in any case , the calculation we describe can be adapted easily to any heavy particle carrying the requisite quantum numbers . the outline of this paper is as follows . in [ sec : axionic ] we introduce our model and show that a coupling between gauge bosons and dark energy is generated at low energy by integrating out heavy particles . some of the phenomenology associated with this coupling is presented . in [ sec : higgs ] we briefly review higgs production at a hadron collider , emphasizing the role of vertices between the electroweak gauge bosons and the higgs . in [ sec : scalar - corrections ] we compute corrections to the higgs production rate arising from the low energy theory written down in [ sec : axionic ] . we show that the effective lagrangian includes new contact interactions between the gauge bosons and the higgs . at scales smaller than @xmath13 these contact interactions resolve into heavy particle loops . we compute the correction to higgs production from both effects . in [ sec : conclusions ] we conclude by arguing that the effect of chameleon - like particles on higgs production is expected to be rather small . throughout this paper , we adopt units in which @xmath14 . we set the reduced planck mass , @xmath15 , to unity . our metric convention is @xmath16 . khoury & weltman @xcite suggested that a scalar field , @xmath17 , might evade detection and yet remain light in vacuum if it coupled to the matter species , @xmath18 , via a set of conformally rescaled metrics,[multiblock footnote omitted ] @xmath19 in this formula , @xmath20 is the spacetime metric , and @xmath21 is a coupling scale that is not necessarily related to the cutoff controlling the validity of the theory , @xmath6 . the einstein frame action is @xmath22 where @xmath23 is the lagrangian describing the dynamics of the matter fields , @xmath18 . the dynamics of @xmath17 are controlled by an effective potential @xmath24 , which depends explicitly on the environment through @xmath23 and satisfies @xmath25 where the sum is over all metrics and @xmath26 . the conformal functions @xmath27 and the potential @xmath28 must be chosen to realise a successful chameleon mechanism . this requires that the effective potential has a minimum . another requirement is the existence of a _ thin shell _ effect @xcite , which is realized if the scalar field potential and coupling functions are chosen in such a way that large variations in the mass of the field can occur as the matter content of the local environment changes . classical realizations of this idea have been constructed , but it is not known whether quantum realizations exist , or what condition the @xmath27 must satisfy once quantum corrections are taken into account . in what follows we will neglect the indices @xmath29 and @xmath30 for simplicity , but it is not necessary to assume that the dark energy field couples with the same strength to all matter fields . we work in the einstein frame , in which the gravitational part of the action is that of general relativity . if the scalar field couples conformally with universal strength , then a classically equivalent description can be obtained by transforming to the jordan frame in which all effects of the scalar field reappear in the gravitational sector . in the classical theory , physical observables are independent of this choice . in the quantum theory , however , a non - trivial jacobian may be necessary to connect the einstein- and jordan - frame measures in the path integral @xcite . in this paper we work in the einstein frame from the outset , and neglect terms arising from the jacobian . these make additional contributions to the coupling between gauge - boson kinetic terms and the dark energy scalar . we hope to return to this question in a future publication . despite appearances , chameleonic couplings of the form do not necessarily give rise to large variations in fundamental constants or particle masses @xcite . this follows because the minimum of the effective potential is a cosmological attractor . the location of the minimum drifts only slowly , owing to time evolution of the matter density . this implies that variations in the scalar - dependent particle masses and fundamental constants are correspondingly small . the kinetic terms of spin-1 particles are conformally invariant , and are left inert under the substitution @xmath31 in eq . . however , particle physics is not conformal ; any coupling to other matter species will generically break this invariance . if so , quantum corrections will cause the kinetic terms to depend on @xmath32 . in this section we compute these threshold corrections , leading to an effective theory which describes the interaction of gauge bosons and the dark energy field at low energy . the necessary corrections are depicted in fig . [ fig : chi - diagrams ] , in which a heavy fermionic particle circulates in the loop . fermions couple to the metric via a vielbein , @xmath33 , which satisfies @xmath34 there is an inverse vielbein , @xmath35 , satisfying @xmath36 and @xmath37 . the indices @xmath38 , @xmath39 , transform under a rigid lorentz symmetry and can be coupled to dirac @xmath12-matrices . the action for a dirac fermion , @xmath40 , with large mass @xmath41 , can be written @xmath42 where @xmath43 is a gauge - covariant derivative and @xmath44 is the spinor conjugate to @xmath40 . we suppose that @xmath40 transforms under an abelian symmetry with gauge coupling constant @xmath45 , so that @xmath46 where @xmath47 $ ] . note that the spin connexion , @xmath48 , transforms non - trivially under conformal rescalings . the calculation will be generalized to non - abelian symmetries below eq . . conformal invariance is broken by the fermion mass , @xmath41 . like the mass of any canonical field , @xmath41 varies under conformal transformations like @xmath49 . if all energy scales ran in the same way with the vacuum expectation value of the scalar field then the effects of this coupling would never be observable , but this is not the case . both the planck scale , @xmath50 , and the scale controlling the strength of the scalar coupling to matter , @xmath51 , are unchanged by variations in the scalar field . we take the dark energy scalar to have a spatially - independent vacuum expectation value @xmath52 , around which we quantize small fluctuations @xmath53 . ( for this reason our calculation can not be applied to very large spacetime volumes in which @xmath17 may develop appreciable gradients . ) after rescaling the fermion fields to have canonical kinetic terms at leading order , the interaction between @xmath40 and these fluctuations can be written @xmath54 where @xmath55 and we have defined a conformally transformed mass @xmath56 . nevertheless , we emphasize that the physical mass of the fermion is @xmath41 . a prime @xmath57 denotes the derivative of a function with respect to its argument . in each diagram of fig . [ fig : chi - diagrams ] , operators @xmath58 , @xmath59 and @xmath60 are inserted on the external legs , with all momenta flowing inwards . [ fig : chi - diagrams ] depicts the `` crossed '' diagram , which corresponds to reversing the sense of momentum flow in the fermion loop . it can be obtained from fig . [ fig : chi - diagrams ] by the simultaneous replacements @xmath61 and @xmath62 . ( 110,60 ) ( 110,60 ) accounting for both diagrams we find that the correlation function @xmath63 can be written @xmath64 where ` @xmath65 ' denotes a trace over dirac indices . the @xmath66 integral is divergent , and such integrals are not guaranteed to be invariant under the rigid shift @xmath67 . in the absence of a shift symmetry , the integral can depend on the labeling of momenta within the loop . we apply dimensional regularization to maintain gauge invariance , after which it can be checked that the result is insensitive to the routing of momentum through the diagram . . also contains a potentially gauge - violating zero derivative term , which would be proportional to @xmath68 in real space . this term vanishes when the @xmath66 integral is evaluated using any gauge - invariant regulator . discarding terms which are proportional to the equations of motion , we find that eq . can be reproduced from an effective lagrangian of the form @xmath69 where @xmath70 denotes terms containing four or more derivatives . . applies for each species of heavy fermion in the theory . if there are many such fermions , each contributes with its own @xmath51 and @xmath71 . if @xmath45 , @xmath72 and @xmath73 are of order unity for all species , the coupling will be dominated by the fermion with largest @xmath51 . similar couplings would be induced by scalar particles , such as heavy sleptons . the physical effect of the coupling in eq . could also be understood in the jordan frame . there , one would typically neglect the effect of dark energy because of its tiny coupling , of order @xmath74 . in the model we are considering , this neglect would be unjustified . the scalar field is more strongly coupled , and its effects would be manifest in the curvature of spacetime caused by particles participating in an interaction . the curvature scale would be associated with energies @xmath75 , which would be larger than that typically associated with gravitational phenomena . in this calculation , the gauge field @xmath76 was taken to be abelian . however , it is clear that the same calculation generalizes immediately to non - abelian fields for which the heavy fermion , @xmath77 , transforms in the fundamental representation . consider any gauge group with generators @xmath78 , so that @xmath79 . the @xmath78 may be normalized to satisfy @xmath80 for an arbitrary constant @xmath81 , given that ` @xmath82 ' denotes a trace over indices in the gauge group . . therefore applies equally in a non - abelian theory after the substitution @xmath83 . . will be accompanied by more complicated corrections which couple @xmath84 to all powers of @xmath53 . resumming this expansion , we generate a coupling of the form @xmath85 for some function @xmath86 and coupling scale @xmath87 . . allows us to estimate the scale @xmath87 , @xmath88 in the following we will assume , for simplicity , that no coupling to gluons is generated . the coupling to electromagnetism will give a small scalar field dependent contribution to the mass of atoms . if the theory were not of the chameleonic type , this would potentially lead to a strong violation of the weak equivalence principle . for chameleons , this will be heavily suppressed by the thin shell mechanism . . shows that , in a conformally - coupled theory , even if contact interactions involving @xmath84 are not present at high energies , they will inevitably be generated after passing the mass threshold of any matter species which couples both to @xmath17 and the gauge field . therefore , laboratory and astrophysical bounds can not be evaded merely by taking the @xmath89 interaction to be absent , although their interpretation becomes model dependent . we believe this to be the first microphysical derivation of the coupling in eq . . this coupling means that chameleons are also axion - like particles : they couple to photons in an analogous way to the peccei quinn axion , and therefore have a similar phenomenology . they are only ` axion - like ' because the mass and couplings of the field are not related as they are for a standard axion , and because we are considering a scalar ( rather than pseudo - scalar ) field . observational constraints on the couplings of axion - like particles are very tight . however , applying these constraints to chameleonic fields is not straightforward because the mass of the chameleon field depends on its environment . the strongest constraints on axion - like particles come from their production in the cores of stars , but a chameleon field becomes very massive in the interior of the sun and therefore its production through scattering processes is suppressed . the constraints on chameleons from such observations are discussed in more detail in ref . @xcite . for similar reasons chameleonic fields are not constrained by so - called `` light shining through walls '' searches for axion - like particles @xcite . this is because the chameleon becomes heavy in the wall and is reflected by it rather than passing though . on the other hand , constraints that come from observing the behaviour of axion - like particles purely in diffuse environments can be applied to chameleonic fields . strong constraints come from laboratory @xcite and astrophysical searches @xcite . indeed , it is possible that certain astronomical observations may be explained most simply by including light scalars which couple as chameleons @xcite . under certain circumstances it is possible to translate the stringent bounds obtained from electromagnetic probes , discussed above , into bounds on the matter coupling . we assume a minimal model in which the dark energy couples to matter with a uniform strength @xmath51 , irrespective of species . this coupling is subject to only mild restrictions , depending on the precise self - interaction potential which is chosen for the chameleon field . even where such restrictions exist , they typically require @xmath51 to be no smaller than the ordinary scale of nuclear physics , @xmath90 @xcite . for @xmath91 and gauge coupling @xmath92 , which is roughly the scale of the @xmath93 and @xmath94 couplings of the standard model , we find @xmath95 therefore , constraints on @xmath87 can be translated to limits on the matter coupling , @xmath51 . unfortunately this constraint is highly model - dependent , and can be weakened arbitrarily by decreasing the gauge coupling @xmath45 . strong constraints are obtained only when @xmath45 can be determined by other means . the strongest bound on @xmath87 follows from observations of the polarization of starlight in the milky way , yielding @xmath96 for models in which the mass of the field satisfies @xmath97 in the interstellar medium @xcite . , also come from the pvlas @xcite and gammev @xcite experiments , and apply to models where the field is sufficiently light in the vacuum tube and sufficiently heavy in the walls of the tube to prevent it escaping from the experiment . ] in models where eq . applies , it follows that there is a new , stronger bound on the matter coupling , , corresponding to @xmath98 , which should be easily satisfied for the chameleon field in all relevant backgrounds . ] @xmath99 if the strength of the coupling of the scalar field to the gauge bosons is strong , we might also expect to see the scalar field in particle colliders . because the chameleonic field is light in the vacuum of a particle collider it can not be integrated out of the theory . in ref . @xcite the effects of dark energy on precision electroweak observables were discussed . corrections from processes with scalar fields in the final state were shown to contribute only if the scale of the coupling is low . the best constraints come from observations of the width for @xmath0-decay , and corrections to this from the scalar field are invisible if @xmath100 . it was also shown that large corrections due to the scalar field are screened in all @xmath101 fermion scattering interactions due to a combination of gauge invariance and the structure of boson / lepton couplings in the standard model . the only processes in the electroweak sector to which this screening theorem does not apply are those involving higgs bosons . for this reason , we might wonder whether coupling a dark energy scalar to the higgs boson would lead to a large enhancement of higgs production in particle colliders . in the next section we will show that this loophole can be closed , and that one can expect all electroweak processes to be screened from chameleon corrections . below the electroweak symmetry breaking scale of order 1 tev , interactions of the lightest neutral higgs , @xmath102 , with the @xmath0 boson are described by cubic and quartic couplings @xcite , @xmath103 the @xmath0 mass is @xmath104 gev and the fermi constant , @xmath8 , is measured experimentally to be @xmath105 gev@xmath106 @xcite . is numerically correct for a minimal standard model higgs . in a two higgs doublet model these couplings are shared between the two neutral scalars , leading to suppression by a numerical factor . , where @xmath51 parametrizes the ratio of higgs vacuum expectation values , @xmath107 , and @xmath30 is an angle which occurs when diagonalizing the higgs mass matrix . @xcite . in a simple mssm , @xmath108 may be near unity . ] analogous interactions for the @xmath1 are obtained by the substitutions @xmath109 and @xmath110 . a summary of the methods for higgs production at a hadron collider can be found in the recent book by kilian @xcite . ( see also ref . @xcite . ) ( 80,60 ) ( 80,60 ) one especially important production mechanism is ` weak boson fusion , ' @xmath111 , shown in fig . [ fig : higgs - production ] , where @xmath28 is any of the vector bosons @xmath112 . this process is generally sub - dominant to the rate of gluon gluon fusion , but benefits from more accurate background subtraction . the precursor bosons originate within spectator quarks , as shown in fig . [ fig : higgs - production ] and discussed in more detail in [ sec : effective - w ] below . following a boson fusion event , the spectator quarks are disrupted and initiate transverse hadronic jets . another interesting mechanism is ` associated production , ' shown in fig . [ fig : higgs - production ] , in which two quarks fuse to form a @xmath113 resonance that subsequently decays via `` higgsstrahlung '' into an on - shell @xmath113 and a higgs , @xmath114 . then its primary decays will be to @xmath115 pairs . weak boson fusion and associated production rely on couplings of the higgs to two vector bosons . these vertices are present in the standard model and in many theories of beyond - the - standard - model physics , although in certain cases they can be tuned to be absent . in this section , we study the possibility that corrections due to dark energy can change the relationship of these vertices to the other measurable parameters of the lagrangian . our analysis applies for a large class of models in which these two - boson couplings are represented by eq . and its generalization to @xmath116 interactions . the processes of weak boson fusion and associated production can be studied using conventional perturbation theory , but alternative methods exist . an especially useful tool is the _ effective @xmath113 approximation _ ( see , eg . , ref . @xcite ) . for weak boson fusion , this means that the vector boson precursors are taken to be on - shell partons within colliding hadrons . in the parton picture , a hadron consists of three valence quarks which are surrounded by a sea of virtual particles . a probe which samples this sea at sufficiently high resolution has a chance to resolve the virtual quanta , rather than valence quarks . since quarks participate in the electroweak interaction , @xmath0 and @xmath1 bosons will be found within the virtual sea . its precise composition can be determined by solving a system of dglap - like equations , which can be thought of as an approximate boltzmann hierarchy @xcite . in this picture , calculations involving hadron collisions with vector boson intermediate states simplify considerably . in the remainder of this paper we will work within the parton picture and the effective @xmath113 approximation . the utility of this description is that a complex cross - section can be factorized into a sequence of simpler subprocesses @xcite . dropping the contribution from @xmath0 bosons , the effective cross - section for weak boson fusion can be written @xcite @xmath117 where @xmath118 are transverse and longitudinal polarization modes , and @xmath119 is the centre of momentum energy of the collision . the functions @xmath120 can be calculated or measured experimentally . eq . shows that the remaining chameleon corrections can be determined from the on - shell rate @xmath121 , taken to occur at centre of momentum energy @xmath122 . in what follows , we will concentrate on modifications to this quantity . the rate @xmath125 depends on the amplitude of a green s function describing effective @xmath126 and @xmath127 vertices . this green s function captures the appearance of new physics in eq . . in many theories of physics beyond the standard model , the form of the cubic and quartic vertices in eq . are unmodified . in these theories , corrections to the self - energy diagrams of @xmath0 , @xmath1 , @xmath12 and @xmath128 particles disrupt tree - level relationships between the particle masses @xmath129 , the fine structure constant @xmath30 , and the fermi constant @xmath8 . this disruption can be summarized using _ oblique parameters _ introduced by peskin & takeuchi @xcite and refined by maksymyk , london & burgess @xcite . in our model there are both oblique and non - oblique corrections . to summarize our results we use the oblique notation of refs . @xcite and note differences explicitly where they occur . consider the green s function describing an effective @xmath124 vertex , from which the @xmath130 result can be derived after trivial modifications . the s - matrix element must depend on @xmath131 where the wavefunction renormalizations arise on transition to the @xmath132-matrix . we work in unitary gauge , where the goldstone modes associated with the @xmath93 higgs doublet are absorbed as longitudinal polarizations of @xmath0 and @xmath1 . accounting for oblique corrections , the @xmath0 propagator is @xmath133 where @xmath134 takes the form @xmath135 in eqs . , @xmath66 stands for either @xmath136 or @xmath137 . the quantity @xmath138 is defined as follows . we choose @xmath139 to be the sum of all one - particle - irreducible graphs connecting an ingoing and an outgoing @xmath0 . in vacuum this has a unique tensorial decomposition , @xmath140 we neglect the term involving @xmath141 and can therefore drop superscript s without ambiguity , so that @xmath143 . it is this quantity which appears in eq . . if the mass of external fermions is at most @xmath144 , this neglect is equivalent to dropping powers of @xmath145 . it is likely to be a good approximation provided the higgs is not too heavy : for a higgs lighter than the top mass , @xmath146 gev , decay into a top quark is kinematically forbidden . therefore , @xmath147 is at most of order @xmath148 to @xmath149 . including oblique corrections , the @xmath0 mass becomes @xmath150 a similar formula can be written for the @xmath113 propagator , making the replacements @xmath110 and @xmath151 . quantities with a tilde , such as @xmath152 and @xmath153 , refer to the value of these parameters in the absence of oblique corrections . with the same conventions , the fermi constant satisfies @xmath154 @xmath8 parametrizes the strength of the weak force near zero momentum transfer . allowing for these shifts , the decay rate @xmath155 is related to the pure standard model rate by the rule @xmath156 for @xmath113 bosons the decay rate is @xmath157 where , as above , @xmath158 is the fine structure constant . eqs . and have been written in terms of the conventional oblique quantities @xmath28 and @xmath113 , which are defined to satisfy @xcite @xmath159 in addition , we have introduced a new quantity @xmath160 which is a measure of the higgs wavefunction renormalization , @xmath161 if the dark energy coupling scale is greater than the typical scale of electroweak processes , @xmath162 , we expect @xmath28 and @xmath113 to be negligible @xcite . the impact of new physics is therefore contained entirely in @xmath160 . eqs . determine the sensitivity of weak boson fusion and higgsstrahlung to new physics . this sensitivity is measured by the higgs oblique parameter , @xmath160 . in this section we make a quantitative estimate of its magnitude . to do so , we must be precise about the corrections @xmath138 and @xmath163 which modify the standard model prediction . in [ sec : low - energy - oblique ] we determine these quantities in a low energy chameleon - type model coupled to the gauge bosons . we calculate oblique corrections to the production rate , and show that they are sensitive to the high energy completion of the theory . in [ sec : charginos - higgs ] we compute non - oblique corrections generated by integrating out heavy fermions . these are described by a new quartic coupling between the higgs field and the gauge bosons . a dark energy field induces both straight and oblique corrections to the vacuum polarizations of the higgs and gauge bosons . in ref . @xcite it was argued that the straight corrections effectively divide into processes involving `` chameleonstrahlung , '' where dark energy particles are produced but escape the detector , and a collection of `` bridges , '' `` daisies '' and `` rainbows '' which dress the bare processes of the standard model . at leading order , these dressings are momentum independent . chameleonstrahlung was shown to give constraints roughly comparable to those arising from oblique corrections . in this paper we focus on oblique corrections only . after electroweak symmetry breaking , we can parameterize the interactions of eq . by adopting an effective @xmath0 boson lagrangian of the form employed in ref . @xcite , @xmath164 the functions @xmath86 and @xmath165 should satisfy @xmath166 , but depend on the details of ultra - violet physics . more precisely , they are derived from eq . and similar higher - order diagrams involving more powers of @xmath53 . likewise , the couplings @xmath51 and @xmath167 are inherited from whatever heavy particles are integrated out to generate this interaction . working with a sharp momentum cutoff , the @xmath0 vacuum polarization was determined in ref . @xcite and found to be @xmath168 \\ \nonumber \mbox { } + ( x k^2 + \gamma m_z^2)^2 \left [ - \frac{1}{2 } \frac{\lambda^2}{\lambda^2 + \sigma_z^2 } + \frac{1}{2 } \ln \left ( 1 + \frac{\lambda^2}{\sigma_z^2 } \right ) \right ] \\ \mbox { } - \frac{\omega}{2 } ( k^2 + \epsilon m_z^2 ) \left [ \frac{\lambda^2}{2 } - \frac{m_\chi^2}{2 } \ln \left(1 + \frac{\lambda^2}{m_\chi^2 } \right ) \right ] \bigg\ } , \label{eq : self - energy } \end{aligned}\ ] ] where @xmath169 satisfies @xmath170 and @xmath171 . the parameters @xmath172 and @xmath12 are defined by @xmath173 also , @xmath174 represents @xmath175 where @xmath176 is the mass of the dark energy fluctuation @xmath53 . near @xmath177 , @xmath178 has the approximate form @xmath179 we must determine the vacuum polarization of the higgs . as above , we assume this to follow from a conformal coupling to @xmath17 . in the standard model , this would give the coupling @xmath180 ^ 2 \big ) \right\ } , \label{eq : higgs - couple}\ ] ] where @xmath181 and @xmath86 are the gauge fields of the unbroken @xmath93 and @xmath182 symmetries , respectively ; @xmath183 are a set of appropriately normalized generators of @xmath93 ; and @xmath184 is the generator of @xmath94 . @xmath128 is an @xmath93 higgs doublet , and @xmath38 is a standard parameter of the quartic higgs potential , related to the higgs mass by the rule @xmath185 . in many models the phenomenological couplings @xmath86 , @xmath165 and @xmath186 will be closely related , but for the present we leave them arbitrary . if no relationship exists between the couplings , we find that unitarity is not respected at tree level in two - body scattering of gauge bosons @xcite at energy scales above @xmath187^{-1/2}$ ] . may arise from new dark energy exchange diagrams in two - body scattering . ] in models containing more than one higgs doublet we assume that eq . continues to give a good approximation to the couplings of the lightest neutral higgs . ( 80,60 ) ( 80,60 ) the higgs vacuum polarization @xmath188 can be computed . the one - loop contributions are shown in fig . [ fig : polarizations ] , and depend on the following vertices : @xmath189 @xmath190 the diagram in fig . [ fig : polarizations ] corresponds to a vacuum polarization @xmath191 , \end{aligned}\ ] ] where @xmath192 is a feynman parameter and we have rotated to euclidean signature . in analogy with eq . , @xmath193 is defined so that @xmath194 the diagram in fig . [ fig : polarizations ] contributes @xmath195 carrying out the @xmath196 integrals , we find @xmath197 \\ \nonumber \hspace{-8 mm } \mbox { } + \big\ { b_h^{\prime 2 } x^2 k^4 + m_h^2 \bar{c}_h ' ( 2 \bar{b}_h ' x k^2 + m_h^2 \bar{c}_h ' ) \big\ } \left [ - \frac{1}{2 } \frac{\lambda^2}{\lambda^2 + \sigma_h^2 } + \frac{1}{2 } \ln \left ( 1 + \frac{\lambda^2}{\sigma_h^2 } \right ) \right ] \\ \hspace{-8 mm } \mbox { } - \frac{\bar{b}_h}{2 } ( \bar{b}''_h k^2 + m_h^2 \bar{c}''_h ) \left [ \frac{\lambda^2}{2 } - \frac{m_\chi^2}{2 } \ln \left(1 + \frac{\lambda^2}{m_\chi^2 } \right ) \right ] \bigg\ } . \label{eq : higgs - vac } \end{aligned}\ ] ] from eqs . and it is possible to compute @xmath160 , the parameter which summarizes the oblique dependence of the rates @xmath155 and @xmath198 . we find @xmath199 + \mbox{finite terms of order $ \or\left ( \beta_h^2 { m_{\mathrm{ew}}}^2 \right)$ } . \label{eq : r}\ ] ] at leading order in the divergence , it is independent of @xmath186 . we note , however , that a dependence on @xmath186 persists among those terms which are finite in the limit @xmath200 . these finite terms are of order @xmath201 and can be neglected when @xmath202 . we conclude that the oblique correction is very small , unless the divergent part of eq . can contribute a significant effect . what is the meaning of the divergent term in eq . ? one must be wary when reasoning with power - law divergences , because they can be ascribed no invariant significance . for example , they are absent in dimensional regularization . in ref . @xcite it was found that similar divergences could be absorbed in renormalizations of @xmath8 and the vector boson masses , @xmath203 and @xmath204 . the divergence in eq . can not be absorbed in this way . it expresses a sensitivity to whatever physics completes the low - energy theory containing the standard model and the effective interaction , eq . . it is not a prediction that large effects should be observed in a particle collider . the same divergence arises in all models with this low - energy limit , irrespective of what physics takes place at high energy . to fix its value with confidence , we must know the details of the high energy completion . in the next section we compute its value for a model in which the low energy theory is obtained by integrating out heavy fermions with susy - like couplings . if it is obtained as the low energy effective theory of some complete uv physics , eq . will be accompanied by other interactions which can not be neglected . the most important is a new contact interaction between gauge bosons and the higgs field , and in what follows we estimate its effect on higgs production . we use the size of the contribution to higgs production from this straight corrections to estimate the size of the cutoff that controls corrections due to interactions with dark energy . gauge invariance constrains which operators can appear in the low - energy theory . in the minimal scenario we are considering , the higgs field is in the fundamental @xmath205 representation of @xmath93 . by construction , a field strength term such as @xmath206 transforms in the adjoint representation . therefore the lowest order non - trivial interaction with the higgs must involve @xmath207 , making @xmath208 a gauge invariant dimension - six operator . if this operator is present in the low - energy effective theory , it will modify our expectation for higgs production . accordingly , we must evaluate its coefficient . the prediction is model - dependent . consider a minimal scenario , where the heavy charged particles are fermions which have vertices with the lightest neutral higgs of the form @xmath209 where @xmath29 and @xmath210 label different species of fermion and the strength of the coupling is parametrized by @xmath211 . the operator @xmath212 projects onto the left - chirality half of a dirac spinor , and @xmath213 is its conjugate . the @xmath214 should be chosen real and symmetric if cp violation is to be avoided . in a supersymmetric standard model , the @xmath215 will be charginos and neutralinos . these have couplings to the lightest neutral higgs of the form , with @xmath214 determined by the various factors which diagonalize the chargino and neutralino mass matrices . explicit expressions can be found in a.7 of ref . @xcite . in this case , the chargino and neutralino loops would be accompanied by heavy slepton loops which we do not calculate . there is no reason to expect the slepton contribution to be larger than the neutralino or chargino terms , so we anticipate that the fermion contribution alone is representative . . gives rise to effective operators depicted in fig . [ fig : higgs - diagrams ] . we denote a fermion of species @xmath29 by a straight doubled line , and species @xmath210 by a wiggly doubled line . these diagrams must be summed over all @xmath29 and @xmath210 . ( 110,50 ) ( 110,50 ) ( 110,50 ) ( 110,50 ) the correlation function @xmath216 satisfies @xmath217 where @xmath218 is the mass of species @xmath29 and the symmetrizations are nested , so that @xmath219 is carried out after the simultaneous exchanges @xmath61 and @xmath62 . as above , the @xmath66 integral is divergent and potentially dependent on the labeling of momenta inside the loop , but once the integral has been regularized it can be checked that this dependence vanishes . after a tedious calculation , we find that eq . can be reproduced using the following effective lagrangian , @xmath220 where @xmath221 and the functions @xmath222 satisfy @xmath223 \\ \hspace{-9 mm } \mbox { } - 2 x_{ij } \gamma_{ij } \left [ 4 - 6 x_{ij}^2 - 9 x_{ij}^4 + 14 x_{ij}^6 - 3 x_{ij}^8 + 6 x_{ij}^2 ( 3 - 6 x_{ij}^2 + x_{ij}^4 ) \ln x_{ij } \right ] , \\ \fl\nonumber \zeta_2(x_{ij } ) \equiv 2 \sigma_{ij } \left [ 7 - 47 x_{ij}^2 + 63 x_{ij}^4 - 25 x_{ij}^6 + 2 x_{ij}^8 - 12 x_{ij}^4 ( 6 - 5 x_{ij}^2 + x_{ij}^4 ) \ln x_{ij } \right ] \\ \hspace{-9 mm } \mbox { } - 4x_{ij } \gamma_{ij } \left [ 2 + 9 x_{ij}^2 - 18 x_{ij}^4 + 7 x_{ij}^6 + 6 x_{ij}^2 ( 3 - x_{ij}^4 ) \ln x_{ij } \right ] , \\ \fl\nonumber \zeta_3(x_{ij } ) \equiv -2 \sigma_{ij } \left [ 4 - 23 x_{ij}^2 + 63 x_{ij}^4 - 49 x_{ij}^6 + 5 x_{ij}^8 + 12 x_{ij}^6 ( 5 - x_{ij}^2 ) \ln x_{ij } \right ] \\ \hspace{-9 mm } \mbox { } + 2x_{ij } \gamma_{ij } \left [ 1 - 9 x_{ij}^2 - 9 x_{ij}^4 + 17 x_{ij}^6 - 12 x_{ij}^4 ( 3 + x_{ij}^2 ) \ln x_{ij } \right ] . \end{aligned}\ ] ] the quantities @xmath224 and @xmath225 are defined by @xmath226 to exhibit its gauge invariance , eq . has been written in terms of a conventionally normalized @xmath93 doublet @xmath128 , which coincides with the field - space orientation of the lightest neutral higgs . it will be accompanied by higher derivative terms , represented by @xmath70 , which have been neglected . there is also a term of the form @xmath227 , with @xmath228 a divergent constant , whose role is to renormalize the charge @xmath45 . we discard this term and take @xmath45 to be the renormalized charge . at higher order in @xmath128 , arbitrary powers of @xmath207 may be generated . these will lead to higher - dimension operators which couple polynomials of @xmath128 and its derivatives to the gauge field , but at leading order we can restrict our attention to eq . . to determine the correct order of magnitude for @xmath229 , the cutoff used in the calculation of the oblique corrections in [ sec : low - energy - oblique ] , it is safest to match to whatever theory controls physics in the ultra - violet @xcite . in the present case , this is summarized by eq . . using the effective interactions in eq . it is possible to compute the enhancement to higgs production due to these straight corrections and the represent this in the form of an oblique correction . this will give an estimate of the size of the cut off needed in the calculation of the oblique corrections in [ sec : low - energy - oblique ] . as a reasonable approximation , we take the @xmath93 doublet , @xmath128 , to develop a vacuum expectation value of order the standard model scale , @xmath230 . we take neutral excitations around this condensate to be representative of the interaction with the lightest neutral higgs . choosing the gauge field to be the abelian vector associated with the @xmath0 , the resulting effective lagrangian is @xmath231 h ( \partial_a z_a \partial^a z^b - \partial_a z_b \partial^b z^a ) \\ \hspace{3.4 cm } \mbox { } - [ \zeta_2(x_{ij } ) + \zeta_3(x_{ij } ) ] \partial^2 h z^a z_a + \zeta_3(x_{ij } ) h z^b \partial^2 z_b \big\ } . \end{aligned}\ ] ] in the limit where the relative velocity of the colliding vector bosons goes to zero , approximation , the invariant magnitude of any momenta will be of order @xmath232 . this implies that although the result may be modified by factors of @xmath233 , it is unlikely that we commit a gross error by specializing to the zero - velocity limit . ] and defining @xmath234 we find @xmath235 despite appearances eq . is dimensionless , because each @xmath236 has the same dimension as @xmath8 , [ mass]@xmath106 . the magnitude of this correction varies with the mass ratio , @xmath237 , approaching zero as @xmath238 but asymptoting to an approximate constant for large or small ratios . in a typical supersymmetric standard model the chargino and neutralino masses are undetermined , but provided there is not total mass degeneracy we expect that this threshold correction generates a contribution represented by a cutoff of order @xmath239 this is typically a rather small number , leading to an essentially negligible enhancement in the higgs production rate . in this paper we have derived the low energy effective theory which governs interactions between the gauge bosons of the electroweak sector and a dark energy scalar field . the dark energy is taken to have conformal couplings to the matter species of the standard model . it is possible that couplings of this type allow so - called `` chameleon '' behaviour , in which the field dynamically adjusts its mass to be large in regions of high average density , and small elsewhere . if such theories exist then they would lead to an unambiguous prediction of light dark energy quanta interacting in the beam pipe of any particle accelerator . our low energy theory applies strictly in any model containing heavy charged fermions , but a very similar effective lagrangian would apply for a model containing heavy charged particles of any spin . as a specific example , any supersymmetric standard model must contain charginos and neutralinos . these carry the quantum numbers of the @xmath7 gauge group and have masses of order the supersymmetry - breaking scale . however , our calculation is not restricted to the supersymmetric case . one might worry that the presence of chameleonic quanta would change our predictions for the outcome of particle physics experiments . in ref . @xcite we argued this did not happen for any process without higgs quanta . in this paper we have extended our argument to include higgs production . in particular , using the low energy theory we have computed the oblique corrections to each of the higgs , @xmath0 and @xmath1 propagators at energies below the mass , @xmath41 , of the heavy charged fermions . such corrections modify the rates @xmath155 and @xmath240 , where @xmath102 is the lightest neutral higgs , and in principle could change the rate of production of this particle at a hadron collider . we find that the corrections diverge quadratically with the scale chosen as the cut off for the effective theory . this does not predict a large enhancement to the production of higgs bosons from interactions with dark energy , instead indicating that the interactions with dark energy make this process sensitive to the uv physics . other contributions exist , generated by processes taking place at high energy , which are integrated out of the low - energy description . we determine these `` threshold corrections '' by integrating out heavy fermion loops which mediate interactions between the lightest neutral higgs and the gauge bosons . the scale of these corrections can then be reinterpreted as a cutoff in an oblique calculation , of order @xmath241 , and therefore leads to at most small effects . when @xmath41 is much larger than the standard model scale @xmath230 it is entirely negligible . in an unconstrained theory , we might have anticipated a cutoff of order @xmath242 , because at this scale the effective dark energy theory becomes invalid . if this were true , it would be possible to contemplate corrections to the higgs production rate of @xmath233 or larger , which would lie within the discovery reach of the lhc . unfortunately , the corrections are much smaller . large terms could only arise from the relevant operator @xmath243 , but its appearance is forbidden by gauge invariance above the scale of electroweak symmetry breaking . therefore , we expect the coefficient of this term to be at most @xmath230 , rather than @xmath244 . instead , the leading correction comes from the operators @xmath245 , @xmath246 and @xmath247 . these are dimension - six operators , because of the @xmath93 nature of the higgs doublet . accordingly the cutoff is suppressed by @xmath248 , making it small . this must be typical of any uv correction because there is no relevant operator we can write down which will couple @xmath17 to the gauge fields . to get a larger effect , it appears to be necessary to break the gauge invariance of the theory . this does not rule out the possibility that dark energy could be responsible for an enhancement in the higgs production rate , but such a scenario would apparently require exotic physics . in the context of particle colliders where strong magnetic fields are present , the coupling derived in [ sec : axionic ] implies a coupling of the chameleon to synchrotron radiation . this would lead to emergence of chameleon - like particles due to the primakov effect . for most chameleon theories , the large mass assumed by chameleon - like particles in a dense environment implies that the beam pipe acts as a reflecting wall , preventing dark energy particles from leaking out . hence collider experiments would actually take place in a dark energy bath . the analysis of this phenomenon is left for future work . we would like to thank ben allanach , neil barnaby , cliff burgess , douglas shaw and james stirling for helpful discussions . cb is supported by the german science foundation ( dfg ) under the collaborative research centre ( sfb ) 676 . acd and ds are supported by stfc and the cambridge centre for theoretical cosmology . aw is supported by the cambridge centre for theoretical cosmology . 10 c. wetterich , _ the cosmon model for an asymptotically vanishing time dependent cosmological ` constant ' _ , _ astron . astrophys . _ * 301 * ( 1995 ) 321328 , [ http://xxx.lanl.gov/abs/hep-th/9408025 [ arxiv : hep - th/9408025 ] ] . i. zlatev , l .- m . wang , and p. j. steinhardt , _ quintessence , cosmic coincidence , and the cosmological constant _ , _ phys . * 82 * ( 1999 ) 896899 , [ http://xxx.lanl.gov/abs/astro-ph/9807002 [ arxiv : astro - ph/9807002 ] ] . j. khoury and a. weltman , _ chameleon fields : awaiting surprises for tests of gravity in space _ , _ phys . * 93 * ( 2004 ) 171104 , [ http://xxx.lanl.gov/abs/astro-ph/0309300 [ arxiv : astro - ph/0309300 ] ] . d. f. mota and j. d. barrow , _ local and global variations of the fine structure constant _ , _ mon . not . * 349 * ( 2004 ) 291 , [ http://xxx.lanl.gov/abs/astro-ph/0309273 [ arxiv : astro - ph/0309273 ] ] . d. f. mota and d. j. shaw , _ evading equivalence principle violations , astrophysical and cosmological constraints in scalar field theories with a strong coupling to matter _ , _ phys . * d75 * ( 2007 ) 063501 , [ http://xxx.lanl.gov/abs/hep-ph/0608078[arxiv:hep-ph/0608078 ] ] . p. brax , c. van de bruck , a .- c . davis , j. khoury , and a. weltman , _ detecting dark energy in orbit : the cosmological chameleon _ , _ phys . * d70 * ( 2004 ) 123518 , [ http://xxx.lanl.gov/abs/astro-ph/0408415 [ arxiv : astro - ph/0408415 ] ] . a. weltman , _ hunting for chameleons in alp searches _ , in _ proceedings of the 4th patras workshop on axions , wimps and wisps _ , 2008 . http://xxx.lanl.gov/abs/0809.4293[arxiv:0809.4293 ] . isbn 978 - 3 - 935702 - 26 - 3 . c. burrage , _ searching for axion - like - particles in the sky _ , _ nuclear physics b proceedings supplements _ * 194 * ( 2009 ) 190195 . new horizons for modern cosmology : proceedings of the galileo galilei institute conferences on dark matter and dark energy . a. upadhye , j. h. steffen , and a. weltman , _ constraining chameleon field theories using the gammev afterglow experiments _ , _ phys . * d81 * ( 2010 ) 015013 , [ http://xxx.lanl.gov/abs/0911.3906 [ arxiv:0911.3906 ] ] . p. brax , c. burrage , a .- c . davis , d. seery , and a. weltman , _ collider constraints on interactions of dark energy with the standard model _ , _ jhep _ * 09 * ( 2009 ) 128 , [ http://xxx.lanl.gov/abs/0904.3002 [ arxiv:0904.3002 ] ] . p. brax , c. van de bruck , a .- c . davis , d. f. mota , and d. j. shaw , _ testing chameleon theories with light propagating through a magnetic field _ , _ phys . rev . _ * d76 * ( 2007 ) 085010 , [ http://xxx.lanl.gov/abs/0707.2801[arxiv:0707.2801 ] ] . p. brax , c. van de bruck , and a .- c . davis , _ compatibility of the chameleon - field model with fifth- force experiments , cosmology , and pvlas and cast results _ , _ phys . * 99 * ( 2007 ) 121103 , [ http://xxx.lanl.gov/abs/hep-ph/0703243[arxiv:hep-ph/0703243 ] ] . a. mirizzi , g. g. raffelt , and p. d. serpico , _ photon - axion conversion as a mechanism for supernova dimming : limits from cmb spectral distortion _ , _ phys . * d72 * ( 2005 ) 023501 , [ http://xxx.lanl.gov/abs/astro-ph/0506078 [ arxiv : astro - ph/0506078 ] ] . h. gies , d. f. mota , and d. j. shaw , _ hidden in the light : magnetically induced afterglow from trapped chameleon fields _ , _ phys . * d77 * ( 2008 ) 025016 , [ http://xxx.lanl.gov/abs/0710.1556 [ arxiv:0710.1556 ] ] . m. ahlers , a. lindner , a. ringwald , l. schrempp , and c. weniger , _ alpenglow - a signature for chameleons in axion - like particle search experiments _ , _ phys . rev . _ * d77 * ( 2008 ) 015018 , [ http://xxx.lanl.gov/abs/0710.1555[arxiv:0710.1555 ] ] . c. burrage , a .- c . davis , and d. j. shaw , _ detecting chameleons : the astronomical polarization produced by chameleon - like scalar fields _ , _ phys . * d79 * ( 2009 ) 044028 , [ http://xxx.lanl.gov/abs/0809.1763[arxiv:0809.1763 ] ] . davis , c. a. o. schelpe , and d. j. shaw , _ the effect of a chameleon scalar field on the cosmic microwave background _ , _ phys . * d80 * ( 2009 ) 064016 , [ http://xxx.lanl.gov/abs/0907.2672 [ arxiv:0907.2672 ] ] . c. p. burgess , s. godfrey , h. konig , d. london , and i. maksymyk , _ a global fit to extended oblique parameters _ , _ phys . lett . _ * b326 * ( 1994 ) 276281 , [ http://xxx.lanl.gov/abs/hep-ph/9307337 [ arxiv : hep - ph/9307337 ] ] .

consider the definition of quasilocal energy stemming from the hamilton - jacobi method as applied to the canonical form of the gravitational action . we examine in the standard `` small - sphere limit , '' first considered by horowitz and schmidt in their examination of hawking s quasilocal mass . by the term _ small sphere _ we mean a cut , level in an affine radius , of the lightcone belonging to a generic spacetime point . as a power series in , we compute the energy of the gravitational and matter fields on a spacelike hypersurface spanning . much of our analysis concerns conceptual and technical issues associated with assigning the zero - point of the energy . for the small - sphere limit , we argue that the correct zero - point is obtained via a `` lightcone reference , '' which stems from a certain isometric embedding of into a genuine lightcone of minkowski spacetime . choosing this zero - point , we find the following results : ( i ) in the presence of matter|_{p } + { \rm o}(r^{4})$ ] and ( ii ) in vacuo|_{p } + { \rm o}(r^{6})$ ] . here , is a unit , future - pointing , timelike vector in the tangent space at ( which defines the choice of affine radius ) ; is the matter stress - energy - momentum tensor ; is the bel - robinson gravitational super stress - energy - momentum tensor ; and denotes `` restriction to . '' hawking s quasilocal mass expression agrees with the results ( i ) and ( ii ) up to and including the first non - trivial order in the affine radius . the non - vacuum result ( i ) has the expected form based on the results of newtonian potential theory .

gr-qc9810003

consider einstein s non - covariant first - order action@xcite , the 4-integral of a `` bulk '' lagrangian which is quadratic in the christoffel symbols and thus often called the `` @xmath12 action . '' starting with the einstein action , one applies standard techniques associated with noether s theorem in order to derive , among other things , an energy definition in general relativity : namely , the 2-integral of an einstein superpotential over some generic 2-surface @xmath13 in spacetime @xmath14 . the einstein energy is well known to be ambiguously defined because it depends on the choice of background coordinates . nevertheless , one may use the einstein construction to define sensible notions of total gravitational energy . indeed , consider the scenario of asymptotic flatness , say , towards future null infinity @xmath15 . in this case , @xmath13 tends to a round , infinite - radius , 2-sphere cut of @xmath15 , and the ( now suitably unique ) choice of asymptotically cartesian coordinates ensures that the einstein energy agrees with the accepted trautman - bondi - sachs ( tbs ) notion of total energy.@xcite however , were we to offer the einstein definition as the energy contained within some _ quasilocal _ ( that is , finite ) 2-surface @xmath13 , we would still be confronted with the task of choosing a physically meaningful set of background coordinates . the only natural choice would be coordinates which are partially adapted to the embedding of @xmath16 . however , such a choice wrecks the agreement between the einstein and tbs energies as @xmath13 tends towards @xmath15 . in fact , choosing such @xmath13-adapted coordinates , one finds that the einstein energy blows up in the said limit . similar statements can be made regarding other approaches for defining energy , which trade coordinate ( that is , holonomic - frame ) ambiguity for ambiguity of a different stripe , e. g. tetrad ( or rigid ) frame , spin frame , or auxiliary vector ( or spinor ) fields . the traditional party line regarding these issues is the following : _ there is no over - arching rule , applicable for all quasilocal 2-surfaces , for selecting a ( suitably unique ) background frame _ ; whence gravitational quasilocal energy is not well - defined . to what extent does the stubborn presence of frame ambiguity in the quasilocal context point to a gap in our understanding of gravitational energy ? to address this question and sharpen our thoughts on these issues , let us consider a covariant version of the einstein construction . employing a straightforward field - theoretic generalization of the hamilton - jacobi ( hj ) method@xcite , one may derive from a covariant action functional an expression for canonical quasilocal energy ( qle ) in general relativity.@xcite we call this definition of qle _ canonical _ , because , owing to its intimate connection with hj theory , this qle is also the on - shell value of the gravitational hamiltonian for the choice of unit lapse function and vanishing shift vector at the system boundary.@xcite we also note that the canonical qle is the thermodynamic internal energy in a thermodynamical description of a ( relativistic ) self - gravitating system.@xcite the analysis that leads to the canonical qle runs along a somewhat different line than the one followed in a noether - type analysis , but it also leads to a concept of energy which is not unique . indeed , _ as is always the case with energy _ , the canonical qle is defined only up to the choice of a zero - point . the zero - point ambiguity may be traced to a freedom present in the action principle . namely , one may always add to the action any functional of the fixed boundary data without affecting the variational principle . as with the situation above , if the goal is to obtain agreement with the accepted notions of gravitational energy at spacelike or null infinity , then there is a suitably unique choice of energy zero - point@xcite , whereas at the quasilocal level there seems to be no preferred choice .- adapted coordinates for the einstein definition . like before , such a choice , leading to an infinite energy , wrecks the agreement between the canonical quasilocal and tbs energies in the large - sphere null limit . ] while at first sight this seems no better or worse than the situation encountered above , notice that now the ambiguity in the energy has a physical interpretation , and , moreover , is a field - theoretic generalization of the standard ambiguity present in the hj definition of energy in ordinary mechanics . we may now restate the emphasized portion of the party line above as follows : _ there is no over - arching rule , applicable for all quasilocal 2-surfaces , for selecting a ( suitably unique ) zero - point . _ taking this statement at face value , we claim that it is the physicist s job to select an appropriate energy zero - point , guided by the principle that the selection should be appropriate for the physics of the problem at hand . we would like to point out that this is a common enough state of affairs in general relativity , a many - faceted theory known for its wealth of possible boundary conditions . indeed , by way of analogy consider the search for solutions of the einstein field equations . in practice , relativists certainly do not attempt to find _ the _ general solution , rather they attempt to find solutions given some additional physical input ( boundary conditions , symmetries , etc . ) . in practice , the same such additional input is needed to associate a meaningful qle with a particular quasilocal 2-surface . bearing these points in mind , we recall the form of the canonical qle : @xmath17 where we adopt geometrical units ( in which both newton s constant and the speed of light are set to unity ) , @xmath13 is a closed 2-surface , @xmath18 is the proper area element on @xmath13 , and @xmath19 is the mean curvature of @xmath13 as embedded in a spanning 3-surface @xmath5 . it is important to realize that this @xmath0 , while obtained as a proper surface integral over @xmath13 , is the energy of the gravitational and matter fields which live on @xmath5 , that is to say , @xmath0 is a functional of the initial data of @xmath5 . this concept of energy is rooted in the @xmath20 view of spacetime geometry , and for a fixed @xmath13 it is slightly sensitive to the choice of spanning @xmath5 . [ more precisely , @xmath0 depends on the equivalence class of spanning 3-surfaces determined by a unit timelike vector on ( and pointing orthogonal to ) @xmath13 . ] this sensitivity is quite analogous to the observer dependence of energy in special relativity , and _ a priori _ we expect its presence.@xcite in general , the term @xmath21 represents an arbitrary ( local ) function of the intrinsic metric @xmath22 of @xmath13 and corresponds to the freedom to assign the qle zero - point . notice that this freedom corresponds to a proper surface integral of what is effectively a free function of two variables . this freedom , stemming from the field - theoretic character of gravity , is rather more subtle than the freedom in simple mechanics of simply adding a constant to the energy . for our analysis here we only consider two energy zero - points , one determined by _ lightcone reference _ and the other by _ euclidean reference_. we make these concepts precise below . in this paper we examine the canonical qle in the standard `` small - sphere limit , '' first considered by horowitz and schmidt in their classic examination of hawking s quasilocal mass . by the term _ small sphere _ we mean a cut @xmath1 , level in an affine radius @xmath2 , of the lightcone @xmath3 belonging to a generic spacetime point @xmath4 . as a power series in @xmath2 , we compute the energy @xmath0 of the gravitational and matter fields on a spacelike hypersurface @xmath5 spanning @xmath1 . much of our analysis concerns conceptual and technical issues associated with assigning the zero - point of the energy , and , therefore , particularly elucidates the points raised in the first two paragraphs above . for the small - sphere limit , we argue that the correct zero - point is obtained via the aforementioned lightcone reference , which stems from a certain isometric embedding of @xmath1 into a genuine lightcone of minkowski spacetime and amounts to fixing @xmath23\!{}^{-1}\right\ } \label{lcmeancurvature}\ ] ] in the above qle definition . here @xmath24 is the laplacian operator and @xmath25 is twice the gaussian curvature of @xmath1 ; therefore , as advertised , this choice for @xmath26 depends solely on the intrinsic geometry of @xmath1 . notice that , due to the presence of the square root in this expression , one expects this choice for @xmath26 to be valid only for ( topologically spherical ) 2-surfaces possessing everywhere positive gaussian curvature ( as is the case both for the small spheres we study here ) . choosing the proper surface integral of this choice for @xmath27 as the energy zero - point , we find the following results for small spheres : in the presence of matter @xmath28|_{p } + { \rm o}(r^{4 } ) \label{result_one}\ ] ] and in vacuo @xmath29|_{p } + { \rm o}(r^{6 } ) { \ , } \label{result_two}\ ] ] here , @xmath8 is a unit , future - pointing , timelike vector in the tangent space at @xmath4 ( which defines the choice of affine radius ) ; @xmath9 is the matter stress - energy - momentum tensor ; @xmath10 is the bel - robinson gravitational super stress - energy - momentum tensor ; ] where @xmath30 is the weyl tensor , @xmath31 is the left - dual of the weyl tensor , and @xmath32 is the self - dual @xmath33 [ anti - self - dual @xmath34 part of the weyl tensor . further curvature conventions are discussed in the appendix . ] and @xmath11 denotes `` restriction to @xmath4 . '' it is interesting to note that , when integrated , the bel - robinson `` energy '' in ( [ result_two ] ) has been proven to be very useful already in the sense of a mathematical `` energy '' in the study of existence of solutions of hyperbolic equations in einsteinian gravity.@xcite it is noteworthy that it shows up also in the physical limit given in ( [ result_two ] ) . although the full physical significance of the physical limit is not known to us , we think that these mathematical and physical properties go beyond mere coincidences . for both the vacuum and non - vacuum cases , hawking s quasilocal mass expression agrees with the canonical qle results ( [ result_one ] ) and ( [ result_two ] ) up to and including the first non - trivial order in the affine radius . we find this result rather striking in light of the fact that the hawking mass has no apparent connection with the gravitational action or hamiltonian . we show that the non - vacuum result ( [ result_one ] ) has the expected form based on the results of newtonian potential theory . compare the small - sphere limit considered here with the large - sphere limit when @xmath1 tends to a cut of @xmath15 , in which case we know that the choice of euclidean reference yields agreement between @xmath0 and the tbs energy.@xcite in both limiting cases , the 2-surface @xmath1 of interest is a cut , level in an affine radius @xmath2 , of an outgoing congruence of null geodesics . there is , however , a crucial distinction to be made . in the small - sphere case , @xmath1 arises as the cut of a genuine lightcone , while in the large - sphere case this is generally not true . we find it remarkable that this distinction can be mirrored in the choice of zero - points . to grasp this point , consider first the euclidean reference for either limit . this reference involves an _ embedding of @xmath1 into a flat euclidean 3-space @xmath35 . now , in both limiting scenarios @xmath1 is , in general , slightly distorted from perfect roundness . therefore , with @xmath35 viewed in turn as an inertial slice of minkowski space @xmath36 , the enveloped @xmath37 can not be the cut of a genuine lightcone of @xmath36 . ( technically , in this case the outward null congruence associated with @xmath1 is not shear - free , but the lightcones of @xmath36 are shear - free . ) therefore , in the general large - sphere scenario , neither the @xmath1 embedding into the physical spacetime @xmath14 nor its embedding into the reference spacetime @xmath36 corresponds to a lightcone embedding [ that is , in neither case is @xmath1 the cut of a lightcone ] . of course , for the small - sphere limit we may employ either the euclidean reference or the lightcone reference . of these two choices , the lightcone employs flat _ spacetime _ to put the reference space on an equal footing with the small sphere construction in the physical spacetime . to define the lightcone reference , we first isometrically embed @xmath1 into the lightcone @xmath38 of a point @xmath39 , and then select a certain 3-surface @xmath40 spanning @xmath1 [ so that @xmath41 . the details of this construction , eventually leading to the expression ( [ lcmeancurvature ] ) , are found in subsection 2.a . now , having defined the lightcone reference ( tailored to the small sphere limit ) and found the resulting closed - form expression ( [ lcmeancurvature ] ) , we may now invert the question , asking whether or not a lightcone - reference @xmath26 defines a qle ( [ brownyorkqle ] ) possessing a correct large - sphere limit towards @xmath15 . we discuss this issue in appendix a. the organization of this paper is as follows . in section 1 we lay the foundations for our examination of the small - sphere limit . we describe the geometry of the limit in subsection 1.a , fix some general conventions in subsection 1.b , and make some general observations concerning the embedding of a @xmath42-surface in minkowski spacetime in subsection 1.c . no choice of energy zero - point is made in section 1 , although the results of subsection 1.c are used in the subsequent sections to construct zero - points . in section 2 we study the small - sphere limit , subject to the choice of lightcone reference , and derive the results ( [ result_one ] ) and ( [ result_two ] ) . in section 3 we discuss the relationships between the main results ( [ result_one ] ) , ( [ result_two ] ) and results from newtonian potential theory . appendix a contains discussions of the small - sphere limit subject to the choice of euclidean reference , and the large - sphere limit towards @xmath15 subject to the choice of lightcone reference . throughout our analysis , we use the newman - penrose ( np ) formalism@xcite , with which we assume the reader is familiar . in appendix b we collect various conventions and results associated with the np formalism which are used in the main parts of the paper . choose a generic spacetime point @xmath43 , as well as a unit , future - pointing , timelike vector @xmath8 which lies in @xmath44 , the tangent space at @xmath4 . we may think of @xmath8 as the instantaneous four - velocity of an eulerian observer at @xmath4 . let @xmath45 represent the future lightcone generated by the null geodesics emanating from @xmath4 . label the generators of @xmath3 by coordinates @xmath46 , or equivalently by @xmath47 , where @xmath48 is the stereographic coordinate . further , choose the affine parameter @xmath2 along the generators of the lightcone which at the point @xmath4 satisfies the following conditions : @xmath49 and the null tangent @xmath50 to the affinely parameterized generators obeys @xmath51 . by the term _ small sphere _ we shall mean a 2-surface @xmath52 level in the coordinate @xmath2 . provided that we restrict our attention to small enough values of @xmath2 , we need not be troubled by conjugate points and each 2-surface @xmath1 will be only slightly distorted from perfect roundness . on our lightcone @xmath3 we construct a null tetrad @xmath53 as follows . we take @xmath54 as the inward null normal to each @xmath1 ( normalized so that @xmath55 ) , and assume that the complex space leg @xmath56 points everywhere tangent to each @xmath1 . further , via enforcement of the condition @xmath57 , we remove the freedom to perform @xmath2-dependent rotations of the complex dyad . together with the geodetic property of @xmath58 , this implies that the spin coefficient @xmath59 [ cf . ( [ spincoefficients]a ) ] . we also consider a standard pseudo - orthonormal tetrad @xmath60 at the point @xmath4 , in terms of which we have the following expansions at @xmath4 : @xmath61{\ , } . \eqnum{\ref{triad}b } \addtocounter{equation}{1}\end{aligned}\ ] ] capital latin letters , e. g. @xmath62 , denote pseudo - orthonormal tetrad indices and run over the values @xmath63 . the expansions @xmath64 and @xmath65 in eq . ( [ triad ] ) show that the components @xmath66 and @xmath67 are essentially the first four spherical harmonics . standard orthogonality properties of the spherical harmonics then yield the following identities : @xmath68 { \ , } , \eqnum{\ref{averages}c } \addtocounter{equation}{1}\end{aligned}\ ] ] which prove quite useful for reducing many of the integral expressions encountered below . here we make use of the convenient notation @xmath69 , where @xmath70 is the area element of a unit - radius round sphere and the integrations in eq . ( [ averages ] ) are over the unit - radius round sphere . let us now define the _ physical energy surface density _ @xmath71 , whose proper surface integral is the total unreferenced quasilocal energy . our construction requires that we fix a 3-dimensional hypersurface @xmath5 spanning @xmath1 , or , more precisely , an equivalence class of spanning 3-surfaces determined by the choice of a unit , future - pointing , timelike vector @xmath8 on ( and pointing orthogonal to ) @xmath1 . we choose @xmath72 which at @xmath4 agrees with the four - velocity introduced in the last paragraph . in terms of the convergences @xmath73 and @xmath74 of the null normals defined in the appendix eqs . ( [ spincoefficients]i , k ) , the mean curvature of @xmath1 as embedded in @xmath5 is given by @xmath75 we shall write @xmath76 for the unreferenced qle associated with the physical slice @xmath5 in spacetime @xmath14 . likewise , we introduce the _ reference energy surface density _ @xmath27 , and with it define the reference contribution to the qle , @xmath77 as mentioned , @xmath78 is a functional solely of the intrinsic geometry of @xmath1 , although as yet we have made no definite choice for this functional . in section 2 we choose the specific functional stemming from the lightcone reference , while in section 3 we choose the one stemming from the euclidean reference . the difference @xmath79 is the total referenced qle ( [ brownyorkqle ] ) . this subsection collects a few basic results concerning the reference embedding @xmath80 . we shall use these results later when constructing a particular reference energy surface density . for notational convenience here and in what follows , we often use a sans serif notation for objects associated with minkowski spacetime @xmath36 , and we may write @xmath81 in place of @xmath82 . we restrict our attention to _ minkowski - spacetime references_. that is to say , the reference energy surface density @xmath83 is determined via an auxiliary isometric embedding of the 2-surface @xmath1 into a 3-dimensional hypersurface @xmath84 which is itself contained in minkowski spacetime @xmath36 . physically , this means that we _ assign _ the zero value of the energy to that portion of the slice @xmath84 contained within @xmath1 . at the end of the day , our expressions for @xmath81 depend only on the @xmath1 2-metric @xmath22 ( with @xmath85 as @xmath1 indices ) , and , therefore , there is technically no need to consider the reference spacetime . nevertheless , in order to motivate and derive our choices , we begin with the chain of inclusions @xmath86 and an overall spacetime point of view associated with it . let us collect a few definitions . take @xmath87 as the unit future - pointing normal of @xmath84 as embedded in @xmath36 . let @xmath88 be the extrinsic curvature tensor associated with the spacelike , outward - pointing , unit normal of @xmath1 as embedded in @xmath84 . denote by @xmath89 the extrinsic curvature tensor ( with @xmath90 as @xmath84 indices ) of @xmath84 associated with @xmath87 . projection into @xmath1 of all of the free indices of @xmath89 defines an extrinsic curvature tensor @xmath91 on @xmath1 associated with @xmath92 . tangential - normal projection with respect to @xmath1 of @xmath89 defines a covector @xmath93 on @xmath1 . if @xmath1 is to arise as a 2-surface in @xmath36 , then along with its intrinsic metric @xmath22 the triple @xmath94 of extrinsic data must obey certain constraints . these are integrability criteria relating the intrinsic and extrinsic data of @xmath1 to the ( vanishing ) components of the @xmath36 riemann tensor . among the constraints for @xmath80 are the following : is widespread in this paper , we use parenthesis when a variable , say @xmath25 , is raised to a power . hence , @xmath95 and @xmath96 , while @xmath97 always denotes the @xmath98 coefficient in the expansion @xmath99 . the only exception to this rule will be the radius @xmath2 itself . as there is no possibilty for confusion , we use , for example , @xmath100 to mean @xmath101 . ] @xmath102 here @xmath25 is the scalar curvature of @xmath1 , and the colon denotes covariant differentiation intrinsic to @xmath1 . we shall not have need to consider the other embedding constraints . throughout this section , the term @xmath26 stands for the explicit expression ( [ lcmeancurvature ] ) given in the introduction , and the functional ( [ e_ref ] ) is fixed accordingly . to construct the lightcone reference ( see the figure at the end of this paper ) , consider the notations of subsection 1.c and assume that @xmath1 is a cut of a genuine lightcone @xmath38 belonging to a point @xmath39 . in this case , the geodesic congruence @xmath38 is sheer - free , which means that the complex shear @xmath103 of @xmath38 vanishes . here @xmath104 and @xmath105 are complex components with respect to the @xmath1 dyad @xmath106 , respectively capturing the trace - free pieces of @xmath88 and @xmath107 . the vanishing of @xmath108 thus implies that the trace - free piece of @xmath88 equals minus the trace - free piece of @xmath107 ; whence eq . ( [ embeddingconstraints]a ) becomes @xmath109 we have no need to consider eq . ( [ embeddingconstraints]b ) in this subsection.b ) and the other embedding constraints not appearing above would be differential equations determining the remaining @xmath1 extrinsic data from @xmath22 and our choices for @xmath110 and @xmath81 . ] we define the convergence @xmath111 of the outward null normal to @xmath112 , and with it rewrite eq . ( [ tracesgone ] ) as @xmath113 now , we are going view @xmath114 as some specified function . specification of @xmath115 ( indirectly ) fixes a slice @xmath84 ( or , more precisely , an equivalence class of slices ) spanning @xmath1 , which in turn defines @xmath81 . with the last equation and some algebra , we find @xmath116 as the expression for @xmath81 in terms of the function @xmath115 which is not yet specified . now , plugging both the radial expansion ( [ gausscurvature ] ) for @xmath25 and some radial expansion for @xmath115 into eq . ( [ lightcone_k ] ) , we obtain a radial expansion for @xmath81 , concerning which we make the following crucial observations . first ( relevant for the non - vacuum case ) , a reference convergence of the form @xmath117 determines an expansion for @xmath81 up to and including @xmath98 ( which is actually higher than the order needed to get the non - vacuum limit ) . second ( relevant for the vacuum case ) , a reference convergence of the form @xmath118 determines an expansion for @xmath81 up to and including @xmath119 . since we wish our final expression for @xmath81 to serve as a proper reference term , we demand that @xmath120 be built purely from the 2-metric @xmath22 of @xmath1 ( so that @xmath81 is ) . this restriction alone hardly fixes the choice of @xmath115 ( and @xmath81 ) . however , for both scenarios of interest the geometry of the embedding of @xmath1 into the _ physical spacetime _ @xmath14 suggests a natural choice . let us write down our choice for @xmath115 , verify that it leads to eq . ( [ lcmeancurvature ] ) , and finally discuss why it is physically meaningful . we pick @xmath121 { \ , } , \label{refrhochoice}\ ] ] where @xmath122 is the full `` eth '' operator on @xmath1 [ cf . ( [ vacuumthop ] ) and ( [ nonvacuumthop ] ) ] and on weight - zero scalars the @xmath1 laplacian is @xmath123 . with this choice for @xmath115 ( and hence this choice for @xmath84 ) , we obtain from eq . ( [ lightcone_k ] ) the same closed - form expression , @xmath124\!{}^{-1}\right\ } { \ , } , \label{lightconek_ref}\ ] ] as given in eq . ( [ lcmeancurvature ] ) for @xmath125 times the reference energy surface density . now , let us argue that eq . ( [ lightconek_ref ] ) defines a physically sensible reference surface density for the small - sphere limit . using eqs . ( [ vacuumthop ] ) and ( [ nonvacuumthop ] ) for the @xmath122 operator and the radial expansion ( [ gausscurvature ] ) for @xmath25 , we find ( whether in vacuo or not ) that ( [ refrhochoice ] ) satisfies @xmath126 where @xmath127 is the unit - radius round - sphere `` eth '' operator . hence , we have at least @xmath128 for the reference convergence . to start with , consider the non - vacuum case and glance at the expansion ( [ nonvacexpns]a ) for the convergence @xmath73 associated with the physical embedding @xmath129 . we see that through @xmath130 our choice for @xmath115 agrees with the physical @xmath73 . now turn to the vacuum case and notice that the expansion ( [ vacexpns]a ) for the convergence @xmath73 associated with the physical embedding obeys @xmath131 . combining eq . ( [ serifrminus ] ) with the vacuum identity piece of the @xmath1 curvature scalar . one finds @xmath132 [ cf . eqs.([vaccurv]a ) and ( [ zerobianchi]d ) ] . but , as shown in the final part of the appendix , the radial bianchi identities [ cf . eq . ( [ zerobianchi]d , e , f , g ) ] imply that @xmath133 which establishes the result . ] @xmath134 we have @xmath135 ; and , therefore , now through @xmath136 , our choice for @xmath115 agrees with the physical @xmath73 . to sum up , we can state that , whether in vacuo or not , our choice ( [ refrhochoice ] ) for @xmath115 determines an embedding of @xmath1 into @xmath137 which would seem closely related both intrinsically and _ extrinsically _ to the physical embedding of @xmath1 into @xmath45 . let us now put aside the issue of motivation , and simply expand our choice ( [ lightconek_ref ] ) for @xmath81 in powers of @xmath2 . since @xmath138 in non - vacuum , from eq . ( [ lightcone_k ] ) we get @xmath139 for the non - vacuum radial expansion of eq . ( [ lightconek_ref ] ) . likewise , since @xmath118 in vacuo , from eq . ( [ lightcone_k ] ) we now get @xmath140 for the vacuum radial expansion of eq . ( [ lightconek_ref ] ) . turning our attention to the non - vacuum scenario , we begin by using the appendix eqs . ( [ nonvacexpns]a , f ) for the convergences @xmath141 and @xmath73 of the null normals to determine the following expansion for @xmath125 times the physical energy surface density : [ cf . ( [ kequals_2mu+rho ] ) ] @xmath142 + { \rm o}(r^{2 } ) { \ , } , \label{matter_k}\ ] ] here , @xmath143 , @xmath144 , and @xmath145 are @xmath146 limits of standard curvature terms from the np formalism . respectively , they are proportional to a certain null - tetrad component of the weyl tensor @xmath147 at @xmath4 , a certain null - tetrad component of the ricci tensor @xmath148 at @xmath4 , and the ricci scalar @xmath149 at @xmath4 [ cf . eqs . ( [ npcurvature]c , f , k ) ] . because we have adapted our null tetrad to the lightcone @xmath3 , the components @xmath143 and @xmath144 are in fact angle - dependent , i. e. they are functions of @xmath47 ; however , as the limit of a scalar function , @xmath150 is not angle dependent . next , we substitute both eq . ( [ matter_k ] ) and the appendix eq . ( [ nonvacuumds ] ) for the surface element @xmath18 of @xmath1 into the basic expression ( [ e_phy ] ) for @xmath151 , and expand out the result in order to obtain @xmath152 + { \rm o}(r^{4 } ) { \,}. \label{anothere_phy}\ ] ] like before , @xmath153 , an angle - dependent term , is proportional to a certain null - tetrad component ( [ npcurvature]g ) of the ricci tensor at @xmath4 . we can perform the integrations remaining in eq . ( [ anothere_phy ] ) . the spherical average of the real part of @xmath143 vanishes identically . indeed , the @xmath146 limit of eq . ( [ npcurvature]c ) shows that @xmath154 , and by symmetry the average of @xmath155 must yield terms either proportional to @xmath156 or containing @xmath157 . therefore , that the average in question vanishes follows from the index symmetries and trace - free character of the weyl tensor . for the next integration we can straightaway write @xmath158 . to evaluate the unit - sphere averages of the other terms in the integrand , we must use the identities ( [ averages]a , b ) . the @xmath146 of eq . ( [ npcurvature]g ) is @xmath159|_{p}$ ] , which shows that the average of @xmath153 is readily obtained with eq . ( [ averages]a ) . by rewriting the @xmath146 limit of eq . ( [ npcurvature]k ) as @xmath160|_{p}$ ] , we likewise obtain the average of @xmath144 with eq . ( [ averages]b ) . adding together the individual results for these integrations , we find @xmath161\right|_{p } + { \rm o}(r^{4 } ) \label{e_physeries}\ ] ] as the final limiting expression for the unreferenced qle . to obtain an expansion analogous to eq . ( [ e_physeries ] ) for @xmath78 , we must first explicitly compute the radial expansion for @xmath125 times the reference energy density . putting together eqs . ( [ nonvack_ref ] ) and ( [ nonvaccurv_0 ] ) , we get @xmath162 + { \rm o}(r^{2 } ) { \,}. \label{finalk_ref}\ ] ] now , substitute eqs . ( [ finalk_ref ] ) and ( [ nonvacuumds ] ) into eq . ( [ e_ref ] ) , do some algebra , and perform a few simple integrations ( with sw denoting _ spin weight _ ) . then with @xmath127 representing the `` eth '' operator on the unit - radius round sphere , we have @xmath163 . ] in order to reach @xmath164 like before , we use eq . ( [ averages]a ) to compute the remaining integral , thereby finding the desired result : @xmath165\right|_{p } + { \rm o}(r^{4 } ) { \ , } , \label{e_refseries}\ ] ] which should be compared with eq . ( [ e_physeries ] ) . we may now easily obtain the introduction s result ( [ result_one ] ) for the non - vacuum limit . indeed , combination of eqs . ( [ e_physeries ] ) and ( [ e_refseries ] ) yields the following result for the total qle ( [ totalqle ] ) : @xmath166 \right|_{p } + { \rm o}(r^{4 } ) { \,}.\ ] ] notably , this result for @xmath0 is valid whether or not the field equations hold , i. e. it is a geometric identity . however , with einstein s equations @xmath167 we immediately arrive at the introduction s result ( [ result_one ] ) . the derivation of the vacuum limit proceeds along the same lines as those just considered for the non - vacuum limit . with the appendix eqs . ( [ vacexpns]a , f ) for the vacuum - case convergences @xmath141 and @xmath73 , we compute the radial expansion for the physical energy surface density , [ cf . ( [ kequals_2mu+rho ] ) ] @xmath168 + { \textstyle \frac{1}{6}}r^{2}\left [ { { \eth}{}_{{{\scriptscriptstyle 0}}}}\bar{\psi}^{1}_{1 } + { \bar{{\eth}}{}_{{{\scriptscriptstyle 0}}}}\psi^{1}_{1}\right ] \nonumber \\ & & + r^{3 } \left[{\textstyle \frac{1}{180 } } \psi^{0}_{0}\bar{\psi}{}^{0}_{0 } - { \textstyle \frac{1}{20 } } { { \eth}{}_{{{\scriptscriptstyle 0}}}}\left ( \bar{\psi}{}^{0}_{0}\psi^{0}_{1 } + 4 \bar{\psi}^{2}_{1}\right ) - { \textstyle \frac{1}{20 } } { \bar{{\eth}}{}_{{{\scriptscriptstyle 0}}}}\left(\psi^{0}_{0 } \bar{\psi}{}^{0}_{1 } + 4 \psi^{2}_{1}\right ) - { \textstyle \frac{1}{2 } } { \cal r}^{2}\right ] + { \rm o}(r^{4 } ) { \ , } . \label{vack_phy}\end{aligned}\ ] ] next , we substitute this expansion along with the expansion ( [ vacuumds ] ) for the vacuum - case surface element into eq . ( [ e_phy ] ) , do some algebra , and perform a few simple integrations . these steps lead to @xmath169 + { \rm o}(r^{6 } ) { \ , } . { \,}\ ] ] a glance at the explicit expression ( [ vaccurv]c ) for the coefficient @xmath97 [ of the @xmath98 term in the expansion ( [ gausscurvature ] ) for @xmath25 ] shows that the unit - sphere average of the term within the parenthesis above vanishes . to evaluate the final unit - sphere average , we first note that the @xmath146 limit of the square of eq . ( [ npcurvature]a ) is in fact @xmath171|_{p}$ ] . this may be checked by an explicit computation using our definition ( [ belrobinson ] ) of the bel - robinson tensor . evidently then , we may use eq . ( [ averages]c ) to find @xmath172\right|_{p } + { \rm o}(r^{6 } ) \label{vacuume_phy}\ ] ] as the desired limit expression for @xmath151 in vacuo . turning now to the calculation of @xmath78 , we first put together eqs . ( [ basicvack_ref ] ) and ( [ vaccurv ] ) , in order to get the following explicit expansion : @xmath173 \nonumber \\ & & + { \textstyle \frac{5}{12}}r^{2}\left [ { { \eth}{}_{{{\scriptscriptstyle 0}}}}\bar{\psi}^{1}_{1 } + { \bar{{\eth}}{}_{{{\scriptscriptstyle 0}}}}\psi^{1}_{1}\right ] + { \textstyle \frac{1}{2 } } r^{3}\left [ \left({\textstyle \frac{1}{45 } } \psi^{0}_{0}\bar{\psi}{}^{0}_{0 } - { \cal r}^{2}\right ) - { \textstyle \frac{1}{45 } } \psi^{0}_{0}\bar{\psi}{}^{0}_{0}\right ] + { \rm o}(r^{4 } ) { \ , } . \label{vack_ref}\end{aligned}\ ] ] we substitute this expansion as well as the expansion ( [ vacuumds ] ) into eq . ( [ e_ref ] ) , and again do some algebra and a few simple integrations , thereby obtaining @xmath174 + { \rm o}(r^{6 } ) { \,}.\ ] ] finally , calculations identical to those just performed for @xmath151 establish that @xmath175\right|_{p } + { \rm o}(r^{6 } ) { \ , } , \label{vacuume_ref}\ ] ] and the difference of ( [ vacuume_phy ] ) and ( [ vacuume_ref ] ) immediately gives the introduction s result ( [ result_two ] ) . it is interesting to compare our main results with analogous results from newtonian potential theory . the newtonian interpretation for the introduction s result ( [ result_one ] ) is straightforward . consider a pressureless ball of fluid of radius @xmath2 and constant ( volume ) mass density @xmath176 . the total newtonian mass for the ball is @xmath177 . if we identify this radius with the affine radius that appears in the result ( [ result_one ] ) , then a further identification between the newtonian mass @xmath14 and the quasilocal energy @xmath0 yields the correspondence @xmath178|_{p}$ ] this is a solid result , fully in accord with the non - existence of pure gravitational field energy " at a point , since only matter energy contributes in this limit . the newtonian analog of the result ( [ result_two ] ) is not clear . nevertheless , we find the following result rather interesting . consider again the ball of fluid , but now let us compute the newtonian gravitational field energy within the ball . recall that the energy density of the newtonian gravitational field is given by @xmath179 , where @xmath180 is the flat - space gradient operator and @xmath181 is the newtonian potential . inserting the appropriate expression for @xmath181 into the energy density and integrating over the interior of the ball , we obtain @xmath182 for the gravitational potential energy inside the ball . comparison of this result with the small - sphere result ( [ result_two ] ) , we see that both expressions depend on the fifth power of radius and both contain a numerical factor of @xmath183 . we do not know at this time if the close resemblance between these results has any physical significance . for discussions we thank h. balasin , t. g. concannon , n. murchadha , and l. b. szabados . in particular , srl thanks l. b. szabados for comments which led to this investigation and n. murchadha for discussions concerning the newtonian interpretation of the vacuum limit . this work was begun while srl visited the research institute for particle & nuclear physics in budapest , hungary , and srl wishes to express his appreciation to this institute both for financial support and a hospitable stay in budapest . this work has been supported by the national science foundation of the usa ( nsf grant # phy-9413207 to the university of north carolina ) , and in part by the `` fonds zur frderung der wissenschaftlichen forschung '' in austria ( fwf project 10.221-phy ) . one can also carry out our analysis of the small - sphere limit for the choice of euclidean reference . to define the euclidean reference , again consider the notations of subsection i.c , but now assume that @xmath84 is a flat inertial hyperplane @xmath35 in @xmath36 . in this case , @xmath184 and the constraints given in eqs . ( [ embeddingconstraints]a , b ) are simply @xmath185 to analyze these equations , it proves useful to split @xmath186 into its trace and trace - free pieces . this is readily achieved by working with the components of @xmath186 with respect to the null dyad @xmath187 . indeed , @xmath188 and @xmath189 respectively capture the trace and trace - free pieces of @xmath88 , and in terms of these quantities the eqs . ( [ scalarandvector]a , b ) become @xmath190^{2 } - { { \mathsf k}}{}_{mm } { { \mathsf k}}{}_{\bar{m}\bar{m } } - { \textstyle \frac{1}{2}}{\cal r } & = & 0 \label{scalarandvector2 } \eqnum{\ref{scalarandvector2}a}\\ { \eth}{{\mathsf k}}{}_{m\bar{m } } - \bar{{\eth } } { { \mathsf k}}{}_{mm } & = & 0 { \,}. \eqnum{\ref{scalarandvector2}b } \addtocounter{equation}{1}\end{aligned}\ ] ] for the case of euclidean reference , we are unable to obtain a closed - form expression for @xmath81 ( neither in the non - vacuum nor vacuum cases ) . nevertheless , it is evident from eqs . ( [ scalarandvector]a , b ) that the full extrinsic curvature tensor @xmath88 ( and hence its trace piece ) is determined solely by the intrinsic metric on @xmath1 . one expects that eqs . ( [ scalarandvector2]a , b ) may be solved for @xmath88 , provided @xmath1 is only slightly distorted from perfect roundness . for small enough values of @xmath2 , the expansion ( [ gausscurvature ] ) for @xmath25 assures us that this is indeed the case . moreover , the solution @xmath88 should be unique up to euclidean translations and rotations . lacking a closed - form expression for @xmath81 , we have obtained radial expansions in both the non - vacuum and vacuum scenarios . for both scenarios , we obtain such expansions by first plugging into eqs . ( [ scalarandvector2]a , b ) the _ anstze _ @xmath191 along with appropriate ( vacuum or non - vacuum , as the case may be ) expansions for both @xmath122 and @xmath25 . we have performed this calculation , and find that the choice of euclidean reference also establishes ( [ result_one ] ) in the non - vacuum case . however , with the choice of euclidean reference , the vacuum small - sphere limit of the qle is not related directly to the bel - robinson tensor . our closing comments concern the large - sphere limit , in which case @xmath1 tends to a round infinite - radius 2-sphere cut of @xmath15 . thus , we now consider a spacetime @xmath14 which is asymptotically flat towards @xmath15 and a corresponding system of bondi coordinates @xmath192 . here @xmath193 is a retarded time coordinate , and @xmath2 is an affine radius similar to before . now @xmath1 arises as a cut , level in @xmath2 , of an outgoing null hypersurface @xmath194 ( in general not a genuine lightcone ) , and twice the gaussian curvature of @xmath1 has the following expansion in powers of inverse @xmath2 : @xmath195 where the coefficient @xmath196 may be expressed in terms of the asymptotic shear @xmath197 of @xmath194.@xcite if we use formula ( [ tracesgone ] ) ( appropriate for a lightcone embedding ) along with the `` rest - frame '' choice @xmath198 , then we have @xmath199 although this expression differs from ( [ lcmeancurvature ] ) , it is also a lightcone reference . from ( [ simple_reference ] ) we find the following radial expansion for the reference mean curvature : @xmath200 which agrees through @xmath201 with the @xmath26 expansion obtained in ref . @xcite via the euclidean reference . hence , all of the results found in ref . @xcite are also valid for the this choice of lightcone reference . in particular , the lightcone - referenced qle agrees with the tbs energy in a suitable null limit . moreover , in the same limit the `` smeared '' version of the qle [ which incorporates a lapse function into the definition ( [ brownyorkqle ] ) ] agrees with geroch s supermomentum ( when the latter is evaluated in a bondi conformal frame ) . see ref . @xcite for further details . to differentiate between the lightcone and euclidean references in the large - sphere limit , one could examine multipole - moment terms ( which arise at higher powers of inverse radius ) for stationary spacetimes . we hope to return to this issue elsewhere . throughout our discussion , we have taken @xmath202 as the signature of the spacetime metric . we have also adopted the index conventions of ref . @xcite for curvature . with these choices we shall define the standard newman - penrose ( np ) connection and curvature coefficients such that our np equations match those listed by newman and tod in the appendix of ref . @xcite . as we use a somewhat non - standard np formalism , let us list some of our basic conventions explicitly . label our null frame as @xmath203 and name its associated connection coefficients as follows : @xmath204 where , for example , @xmath205 . for the small - sphere limit the null frame is adapted to the lightcone @xmath3 as described in the first section , and , as a result , the following vanish identically for our construction : @xmath206 as we retain the freedom to perform @xmath2-independent rotations of the space leg @xmath56 , we shall use those elements of the geroch - held - penrose ( ghp ) formalism@xcite pertaining both to spin - weighted scalars and the `` eth '' operator @xmath122 on @xmath1 . if a quantity @xmath207 transforms as @xmath208 under the rotation @xmath209 ( with @xmath210 independent of @xmath2 ) , then @xmath207 is said to be a spin - weighted scalar of spin weight @xmath211 ( in symbols , sw@xmath212 ) . as the extents of our null normals @xmath58 and @xmath54 have been fixed once and for all , we have no need to consider the concept of boost weight . next , consider the spacetime ricci scalar as well as the components of the weyl and ricci tensors with respect to the null tetrad . with these define the following standard pieces of the spacetime curvature : @xmath213 as mentioned above , with these conventions our np equations are exactly those listed in the appendix of ref . @xcite ; however , we shall consider all possible simplification of these equations afforded by eq . ( [ vanishinggammas ] ) and the use of @xmath122 [ cf . ( [ nonvacuumthop ] ) and ( [ vacuumthop ] ) ] . for the small - sphere limit examined in this paper , we consider the pullbacks to the lightcone @xmath3 of the curvature components ( [ npcurvature ] ) and assume that each pullback may be expanded as a power series in @xmath2 along the lightcone . that is , we assume @xmath214 we shall use these expansions for the curvature components along with the radial np field equations in order to obtain radial expansions along @xmath3 for those spin coefficients ( [ spincoefficients ] ) used in our analysis . we consider the non - vacuum and vacuum cases separately . up to the appropriate order in the affine radius , we have confirmed the non - vacuum asymptotic expansions of the spin coefficients given by kelly _ et al_@xcite and by dougan.@xcite for completeness we recall that this list is @xmath215 + { \rm o}(r^{2 } ) \eqnum{\ref{nonvacexpns}c } \\ \beta & = & - r^{-1 } \bar{\alpha}{}^{0 } - { \textstyle \frac{1}{6 } } r\left[\bar{\alpha}{}^{0 } \phi^{0}_{00 } - 3\psi^{0}_{1 } - \alpha^{0}\psi^{0}_{0}\right ] + { \rm o}(r^{2 } ) \eqnum{\ref{nonvacexpns}d } \\ \lambda & = & { \textstyle \frac{1}{6}}r \left[5 \phi^{0}_{20 } + \bar{\psi}{}^{0}_{0}\right ] + { \rm o}(r^{2 } ) \eqnum{\ref{nonvacexpns}e } \\ \mu & = & - { \textstyle \frac{1}{2 } } r^{-1 } + { \textstyle \frac{1}{2 } } r \left [ \psi^{0}_{2 } + \bar{\psi}{}^{0}_{2 } + 2 \lambda^{0 } + { \textstyle \frac{2}{3}}\phi^{0}_{11 } - { \textstyle \frac{1}{3}}\phi^{0}_{00}\right ] + { \rm o}(r^{2 } ) \eqnum{\ref{nonvacexpns}f } { \ , } , \addtocounter{equation}{1}\end{aligned}\ ] ] where @xmath216 . for the scenario at hand the expansions ( [ nonvacexpns]c , d ) determine the expansions for @xmath217 and @xmath218 up to @xmath98 . actually , we need only know that @xmath219 in order to obtain the @xmath220 coefficient for @xmath221 ; however , for completeness we have explicitly given the @xmath220 coefficient for @xmath222 . as they are not needed in this paper , we do not list the expansions for @xmath223 and @xmath224 . for the vacuum scenario , we need some of the required spin - coefficient expansions out to an order higher than given in either of refs . we obtain the following list : @xmath225 - { \textstyle \frac{1}{12 } } r^{2 } \left[\bar{\psi}{}^{1}_{1 } + \bar{\alpha}{}^{0 } \bar{\psi}{}^{1}_{0}\right ] \nonumber \\ & & + { \textstyle \frac{1}{360 } } r^{3}\left[7 \alpha^{0 } \psi^{0}_{0 } \bar{\psi}{}^{0}_{0 } + 8 \bar{\psi}{}^{0}_{0}\psi^{0}_{1 } - 18 \bar{\alpha}{}^{0 } \bar{\psi}{}^{2}_{0 } - 3 { { \eth}{}_{{{\scriptscriptstyle 0}}}}\bar{\psi}{}^{2}_{0}\right ] + { \rm o}(r^{4 } ) \eqnum{\ref{vacexpns}c } \\ \beta & = & - r^{-1 } \bar{\alpha}{}^{0 } + { \textstyle \frac{1}{6 } } r \left[3\psi^{0}_{1 } + \alpha^{0 } \psi^{0}_{0}\right ] + { \textstyle \frac{1}{12 } } r^{2 } \left[4\psi^{1}_{1 } + \alpha^{0 } \psi^{1}_{0}\right ] \nonumber \\ & & - { \textstyle \frac{1}{360 } } r^{3 } \left[7 \bar{\alpha}{}^{0 } \psi^{0}_{0 } \bar{\psi}{}^{0}_{0 } - 20 \psi^{0}_{0}\bar{\psi}{}^{0}_{1 } - 18 \alpha^{0 } \psi^{2}_{0 } - 15 { \bar{{\eth}}{}_{{{\scriptscriptstyle 0}}}}\psi^{2}_{0}\right ] + { \rm o}(r^{4 } ) \eqnum{\ref{vacexpns}d } \\ \lambda & = & - { \textstyle \frac{1}{6}}r\bar{\psi}{}^{0}_{0 } + { \rm o}(r^{2 } ) \eqnum{\ref{vacexpns}e } \\ \mu & = & - { \textstyle \frac{1}{2}}r^{-1 } + { \textstyle \frac{1}{2 } } r \left[\psi^{0}_{2 } + \bar{\psi}{}^{0}_{2}\right ] + { \textstyle \frac{1}{3 } } r^{2 } \left [ \psi^{1}_{2 } + \bar{\psi}{}^{1}_{2}\right ] \nonumber \\ & & + r^{3 } \left[{\textstyle \frac{1}{360 } } \psi^{0}_{0}\bar{\psi}{}^{0}_{0 } - { \textstyle \frac{1}{40 } } { { \eth}{}_{{{\scriptscriptstyle 0}}}}\left ( \bar{\psi}{}^{0}_{0}\psi^{0}_{1 } + 4 \bar{\psi}^{2}_{1}\right ) - { \textstyle \frac{1}{40 } } { \bar{{\eth}}{}_{{{\scriptscriptstyle 0}}}}\left(\psi^{0}_{0 } \bar{\psi}{}^{0}_{1 } + 4 \psi^{2}_{1}\right ) - { \textstyle \frac{1}{4 } } { \cal r}^{2}\right ] + { \rm o}(r^{4 } ) { \ , } , \eqnum{\ref{vacexpns}f } \addtocounter{equation}{1}\end{aligned}\ ] ] where the coefficient @xmath97 [ cf . ( [ gausscurvature ] ) ] found in @xmath221 is written out explicitly below in eq . ( [ vaccurv]c ) . also , the unit - sphere `` eth '' operator @xmath127 is defined below in eq . ( [ ethnaught ] ) . as before , eqs . ( [ vacexpns]c , d ) determine expansions for @xmath217 and @xmath218 up to @xmath226 , and we do not need the expansions for @xmath223 and @xmath224 . the particular form of the @xmath119 coefficient @xmath227 in the expansion ( [ vacexpns]f ) for @xmath228 plays a crucial role in our calculation of the qle s vacuum limit . therefore , let us sketch how to obtain the given form of @xmath227 , assuming that we have already determined both @xmath221 up to and including @xmath98 and the remaining expansions ( [ vacexpns ] ) as given . rather than making a straightforward appeal to the np field equations ( as we have done to obtain the other spin coefficients ) , we shall instead take advantage of the particular geometry of our construction and derive this coefficient via the geometric identity @xcite @xmath229 @xmath230 is the complex gauss curvature of @xmath1 , and the ricci scalar of @xmath1 is simply @xmath231 . since we work here in vacuo and both @xmath73 and @xmath221 are real , eq . ( [ complexgauss ] ) implies that @xmath232 now , into this equation insert both eq . ( [ muexpansion ] ) and the expansions ( [ vacexpns]a , b , e ) , and then isolate the @xmath98 piece of the resulting expression , in order to establish that @xmath233 to work this relationship into the desired form , we appeal to the following vacuum np bianchi identity : @xmath234 where @xmath235 and the action of @xmath236 on @xmath237 is defined as the conjugate of @xmath238 [ with @xmath122 as in eq . ( [ vacuumthop ] ) and sw@xmath239 . the given form of eq . ( [ bianchi ] ) is specific to the geometry of our construction , but it can be deduced from the ( more general ) appendix eq . ( a.4c ) given in ref . @xcite . now , plug the expansions ( [ curvexpns ] ) and ( [ vacexpns ] ) as well as the radial expansion for @xmath236 into the bianchi identity ( [ bianchi ] ) , thereby obtaining a tower of identities ( one identity at each power of @xmath2 ) . isolate the particular identity determined at @xmath220 in the tower , and into this equation make repeated substitutions with ( [ zerobianchi]a , b ) . these steps lead to @xmath240 which , upon substitution into eq . ( [ muthree ] ) , establishes the chosen form of the coefficient @xmath227 found in the @xmath221 expansion ( [ vacexpns]f ) . in this final appendix subsection we consider the intrinsic geometry of @xmath1 and collect radial expansions along @xmath3 for the `` eth '' operator @xmath122 , the surface area element @xmath18 , and the intrinsic curvature scalar @xmath25 [ the gaussian curvature of @xmath1 being twice @xmath25 ] . again , we consider the non - vacuum and vacuum cases separately . we obtain the desired intrinsic - geometry expansions as follows . first , writing the complex space leg of the null tetrad as@xcite @xmath241 we determine the @xmath2 expansions along @xmath3 for the tetrad coefficients @xmath242 and @xmath243 from the np commutator equations,@xcite @xmath244 and the radial expansions for the spin coefficients [ eqs . ( [ nonvacexpns]a , b , c , d ) or eqs . ( [ vacexpns]a , b , c , d ) as the case may be ] . the given forms of these equations are particular to how our null frame has been adapted to the lightcone @xmath3 . let us first consider the @xmath122 operator . with both eqs . ( [ nonvacexpns]c , d ) and the non - vacuum radial expansions for @xmath242 and @xmath243 , we compute the non - vacuum radial expansion for @xmath122 , @xmath245 + { \rm o}(r^{2 } ) { \ , } . \label{nonvacuumthop}\end{aligned}\ ] ] here , as above , @xmath246 is the standard np notation for the directional derivative along @xmath56 , @xmath247 denotes the spin weight of some spin - weighted scalar @xmath248 on which @xmath122 acts , and in terms of @xmath249 we define @xmath250 a similar calculation shows that the vacuum @xmath122 operator is @xmath251 + { \rm o}(r^{4 } ) { \ , } , \label{vacuumthop}\end{aligned}\ ] ] with @xmath246 , @xmath211 , and @xmath127 as before . next , we substitute the radial expansions we find for @xmath242 and @xmath243 into area element of @xmath1 , @xmath252 here @xmath70 is the standard unit - radius round sphere area element . for the non - vacuum scenario @xmath253 { \ , } , \label{nonvacuumds}\ ] ] while for the vacuum scenario @xmath254 { \ , } . \label{vacuumds}\ ] ] these expansions may be checked with the well - known formula @xmath255 for the change in the surface - area element.@xcite let us turn to the gauss curvature of @xmath1 . an order - by - order examination of the identity @xmath256 [ where sw@xmath257 , but @xmath248 is otherwise arbitrary , and we assume the expansion @xmath258 determines the following expansion for the intrinsic curvature of @xmath1 : @xmath259 in examining eq . ( [ gausscommutator ] ) , we must use the appropriate expansion for @xmath122 , eq . ( [ nonvacuumthop ] ) or eq . ( [ vacuumthop ] ) as the case may be . for the non - vacuum scenario , we find the first relevant coefficient to be @xmath260 { \ , } . \label{nonvaccurv_0}\ ] ] for the vacuum case we need more coefficients . tedious but straightforward calculation yields the set : @xmath261 we remark that , along the way to obtaining these coefficients of @xmath262 from eq . ( [ gausscommutator ] ) , one finds terms involving derivatives of the coefficients of @xmath248 . however , one knows that such terms must vanish , as the right - hand side of eq . ( [ gausscommutator ] ) involves no derivatives of @xmath248 . therefore , one may simply discard these terms . a detailed examination ( order by order in @xmath2 ) of the bianchi identities ( non - vacuum or vacuum as the case may be ) shows that such terms indeed vanish . consider the vacuum bianchi identities as given in in the appendix ref . for the situation at hand , we may simplify these equations by writing them in terms of @xmath122 rather than @xmath246 , and we do so . then , considering the equations order by order in @xmath2 , we find the following at the lowest order in @xmath2 : @xmath263 from these one can derive results such as @xmath264 . respectively , the top and bottom boxes depict the physical and reference spacetimes . in the top box the shaded surface spanning @xmath1 is the 3-surface @xmath5 , while in the bottom box the shaded surface spanning @xmath1 is the 3-surface @xmath84 . whether viewed as the intersection @xmath265 or the intersection @xmath266 , the 2-surface @xmath1 has the same intrinsic 2-metric . our limit construction gives us @xmath5 , but we must choose @xmath84 ; and , moreover , our choice of @xmath84 must be determined solely by the intrinsic 2-metric of @xmath1 . our choice and its physical motivation are described in subsection 2.a . however , we note here that , whenever @xmath1 is at all distorted from perfect roundness ( as it generally will be ) , @xmath84 is not flat euclidean 3-space @xmath35 ( because the intersection of @xmath35 with the genuine lightcone @xmath38 would be a round sphere ) . choosing such a lightcone reference , we _ assign _ the zero value of the energy to that ( shaded ) portion of @xmath84 contained within @xmath1 , and compute the energy of ( the shaded portion of ) @xmath5 relative to this zero - point .

we develop a new set of models for intermediate - metallicity single stellar populations in the blue / optical region and use those models to determine the spectroscopic age of 47 tuc . the models are based on a moderately high - resolution ( 1.8 fwhm ) empirical spectral library , state - of - the - art theoretical isochrones from m. salaris and the most recent set from the padova group , and new semi - empirical calibrations between fundamental stellar parameters and observables . model line - strengths include all corrections for deficiencies of the stellar library that are described in paper i. we highlight the importance of correctly modeling the luminosity function ( lf ) of the cluster at the level of the giant branch , in order to achieve a good reproduction of the integrated spectrum ; agreement between the spectroscopic age and the age based on the cluster s color - magnitude diagram ( cmd ) is achieved only if the _ observed _ lf is used rather than the theoretical ones , which either do not include agb stars ( salaris ) or underpredict the total number counts of bright giants in the cluster by a factor of two ( padova ) . after all corrections are made , the cmd and the spectroscopic ages ( from and ) are in close agreement : 1112 gyrs for salaris isochrones and 13 gyrs for padova . the difference between the model ages is due to the inclusion of atomic diffusion in the salaris models . previously older spectroscopic ages were due to the underestimate of the number of red giants and/or the use of isochrones that neglected the effects of he - diffusion and-enhancement . uncertainties in spectroscopic age determinations of old stellar populations stem from a number of effects , the most important of which are the and _ [ fe / h]_-scales of the giant stars used in the stellar library , the lf on the upper giant branch , and the assumed metallicity of the target stellar population itself . a 1 gyr uncertainty in age results from uncertainties of 75 k in the-scale of the library giants , 0.1 dex in the level of the giant - branch lf , and 0.1 dex in the assumed _ [ fe / h ] _ of either the target stellar population or the assumed zeropoint of the metallicity scale of the stellar library . a similar underestimate in the bright giant lf , if it exists in current super - solar metallicity models , would cause spectroscopic ages of elliptical galaxies inferred from such models to be too high by approximately 30% .

astro-ph0109365

in paper i of this series ( schiavon et al . 2002 ) , we performed an empirical synthesis of the integrated spectrum of the galactic metal - rich globular cluster 47 tucanae . the model spectrum was computed directly from a color - magnitude diagram ( cmd ) of the cluster , being thus essentially independent of inputs from theoretical isochrones . in that paper we achieved good agreement between models and observations in the equivalent widths of two balmer lines and several metal lines . in order to reach such agreement , we needed to correct the model line indices for two limitations of the spectral library adopted in our models , namely , the relative paucity of sufficiently metal - poor giants and the absence of cn - strong stars , which are known to be present in the cluster . we also had to require that the iron abundance of 47 tuc be _ [ fe / h]_=0.75 , which is slightly below and yet within the errorbars of the standard value determined by carretta & gratton ( 1997 ) ( [ fe / h]=0.7 ) . in this second paper of the series , we turn to the more challenging task of making models based on theoretical isochrones . all the corrections described above , which were inferred in paper i , are applied to the models computed here . moreover , we subtract from the observed indices the contribution by blue stragglers , which are not included in the theoretical isochrones . it was shown in paper i that such corrections are important only for @xmath6 and , to a lesser degree , @xmath7 . the effect of blue stragglers on all indices studied in this paper is shown in table [ tbl-4 ] . the index values adopted in this paper are those listed in the second row of table 3 of paper i. the resultant model spectra computed from the theoretical isochrones are used to deduce the cluster s age . such a test is of fundamental importance for stellar population models . a key requirement is that the same age has to be obtained from the fit of the theoretical isochrone to the cluster s cmd , and from the fit of theoretical spectra to the integrated spectrum of the cluster ( the so - called spectroscopic age ) . this is especially interesting in view of the exceedingly high spectroscopic age ( @xmath8 20 gyrs ) that was obtained for 47 tuc from the equivalent width of @xmath0 by gibson et al . the mismatch between spectroscopic and cmd - based ages may even be more general , because cohen , blakeslee & rhyzov ( 1998 ) have found very high spectroscopic ages for a sample of galactic gcs , on the basis of the equivalent width of @xmath1 . more recently , vazdekis et al . ( 2001 ) showed that a lower spectroscopic age ( @xmath2 13 gyrs ) is obtained when isochrones that take into consideration effects such as @xmath3-enhancement and diffusion of heavy elements are used . however , vazdekis et al . still obtained a residual difference of @xmath2 4 gyrs between their spectroscopic and cmd - based ages , in the sense that the spectroscopic age was higher . in the present paper , we approach this problem as follows . we first infer an age for the cluster by fitting theoretical isochrones to its cmd . we adopt the isochrones by salaris and collaborators , which are the same as those employed by vazdekis et al . ( 2001 ) . for comparison purposes , we perform the same exercise adopting @xmath3-enhanced isochrones by salasnich et al . ( 2000 , hereafter , padova isochrones ) . the isochrones are compared to cluster data in both the cmd and luminosity function ( lf ) domains . a major conclusion is that the raw isochrones adopted do not reproduce the cluster s lf correctly . they underestimate the number counts of giants brighter than the horizontal branch by roughly a factor of 2.5 , in part because the theoretical isochrones do not include agb stars . when we adopt the padova isochrones , which do include agb stars , the mismatch is reduced by @xmath2 0.1 dex . we then infer corrections to the model predictions by forcing the theoretical lf to match the observed one . such corrections are shown to be very important , implying a decrease of the order of 34 gyrs in the spectroscopic age inferred from balmer lines . the lf of giant stars is thus a key input in deriving accurate spectroscopic ages of old stellar populations . the revised spectroscopic ages inferred when adopting the `` lf - corrected '' isochrones are in excellent agreement with those obtained from the fits to the cmd . via this careful two - step procedure in papers i and ii , we now feel that the mysterious spectroscopic age discrepancy for 47 tuc has been resolved . the paper is organized as follows . in section [ isocmd ] the theoretical isochrones are compared to data on 47 tuc in both the cmd and lf domains . in section [ isoobs ] we compare observed line - indices to model predictions computed from both the original and the `` lf - corrected '' isochrones . model uncertainties due to errors in the input stellar parameters are discussed in section [ unc ] , and our conclusions are summarized in section [ conc ] . before comparing the isochrone - based model spectra with the observations , we call attention to important features of the theoretical isochrones that are crucial to interpreting our results . in figure [ fig1 ] the cmd of 47 tuc ( howell , guhathakurta & gilliland 2000 ) is overlaid with isochrones from salaris and collaborators for 8 , 10 , 12 and 14 gyrs . conversion from the theoretical plane ( @xmath9,@xmath4 ) to the observational plane [ @xmath10,_(b v ) _ ] was performed according to the recipe described in paper i , and adopting a distance modulus of @xmath11=13.33 and _ e(b v)_=0.04 ( kaluzny et al . 1998 ) . these isochrones are described in vazdekis et al . ( 2001 ) and were kindly provided by m. salaris . they are computed taking into consideration diffusion of heavy elements and an @xmath3-enhanced mixture ( _ [ fe / h]_=0.7 , _ [ @xmath3/fe]_=+0.4 ) . these were the isochrones used by vazdekis et al . ( 2001 ) to infer a spectroscopic age of @xmath2 14 gyrs for 47 tuc , on the basis of the measurement of the @xmath7 index ( for a definition of that index , see figure 16 of paper i ) . the age inferred from their fit to the cmd of the cluster is significantly younger : @xmath2 10 gyrs . in spite of this discrepancy , vazdekis et al . ( 2001 ) greatly improved over previous spectroscopic age determinations , which gave ages in excess of 20 gyrs ( gibson et al . in particular , they showed that @xmath3-enhancement and diffusion of heavy elements cause a reduction in both the cmd and spectroscopic ages of 47 tuc by several gyrs . from figure [ fig1 ] it can be seen that the best match to the position of the cluster s turn - off in the cmd is achieved for an isochrone with age 11 - 12 gyrs . this is of course dependent on the reddening assumed for the cluster . an error of 0.01 mag in _ e(b v ) _ results in an uncertainty of @xmath2 1 gyr in the age inferred from cmd - fitting . that is the reason why we infer a slightly older age than vazdekis et al . , even though we are using the same set of isochrones , as their adopted _ e(b v ) _ is 0.01 mag higher than ours . the isochrone for 11 gyrs matches very well the color of the turn - off of the cluster . however , it predicts a subgiant branch which is too bright by 0.05 mag . there also are mismatches in the color of the giant branch . the model isochrone is too red by @xmath2 0.03 mag for giants brighter than v @xmath2 14 and too blue by a comparable amount at the base of the giant branch . adopting the alonso et al . ( 1999 ) calibration of @xmath4s vs. @xmath12 , rather than the recipe of paper i , leads to a slightly better match in the base of the giant branch , but it worsens the match at @xmath13 , where our calibration provides a good match to the data . such mismatches , if due to errors in the @xmath4s of the isochrones , may have a non - negligible impact on the ages inferred from the fit of the integrated spectrum of the cluster , as we show below . it is also important to notice that agb stars are not included in the salaris isochrones . it will be shown in section [ isoobs ] that this also has an important impact on the integrated line indices . in figure [ fig2 ] we repeat the cmd of 47 tuc , this time comparing the data with the padova isochrones . theoretical quantities ( @xmath4 and @xmath14 ) were transformed into the observational plane following the same recipes as adopted to the salaris isochrones shown in figure [ fig1 ] . the padova isochrones were computed assuming the same chemical composition as adopted in the computation of the salaris isochrones . however , the padova isochrones do not consider the effect of diffusion of helium and heavier elements . as explained by vazdekis et al . ( 2001 ) , inclusion of heavy - element diffusion results in a slightly cooler and fainter turn - off for a given stellar mass and chemical composition . in fact , the turn - off of the 11 gyrs isochrone from salaris is about 0.16 mag fainter and 150 k cooler than the turn - off of the padova isochrone for a similar age ( 11.2 gyrs ) . for this reason the age that best fits the turn - off color and magnitude of 47 tuc , according to the padova isochrones , is somewhere between 12.5 and 14.1 gyrs , being thus slightly older than the one inferred from the salaris isochrones ( figure [ fig1 ] ) . evidence for the occurence of atomic diffusion in the sun has been presented in a number of studies . for instance , basu , pinsonneault & bahcall ( 2000 ) show that diffusion of heavy elements is required if models are to fit the data on the profiles of sound speed , density and adiabatic index in the sun . for more metal - poor stars , lebreton et al . ( 1999 ) showed that the hr diagram from hipparcos data for stars with 1.05 @xmath15 _ [ fe / h ] _ @xmath15 0.45 can be reconciled with the predictions of stellar evolution only if diffusion of heavy elements is considered . however , the extent to which atomic diffusion affects stellar structure and evolution is still under debate . it has been argued that the constancy of the li abundance in metal - poor dwarfs with a wide range of metallicities ( the so - called lithium plateau , spite & spite 1982 ) suggests that diffusion effects may be partially inhibited ( deliyannis & demarque 1991 ) . on the other hand , salaris & weiss ( 2001 ) showed that models with fully efficient diffusion may be consistent with the lithium plateau if the errors on lithium abundances and sample incompleteness are duly accounted for . according to vazdekis et al . ( 2001 ) , the consideration of atomic diffusion in its full extent reduces the turn - off age of 47 tuc by @xmath2 1 gyr . an important constraint on the extent of diffusion affects has also been placed by gratton et al . ( 2001 ) , who found that _ [ fe / h ] _ is the same for stars at the turn - off and at the base of the giant branch of the metal - poor globular cluster ngc 6397 . models by chaboyer et al . ( 2001 ) show that the latter result implies an inhibition of diffusion in the outer layers of stellar models , such that the reduction of cmd - based ages due to atomic diffusion is only 4% ( @xmath2 0.5 gyr in the case of 47 tuc ) . in summary , though we tend to favor the age based on salaris models ( 11 gyrs ) , the main emphasis of this work is on making sure that the spectroscopic age we obtain is _ consistent _ with the one based on the fit of the cluster turn - off , for whatever set of isochrones is employed . in figure [ fig2 ] it also can be seen that the padova isochrones do not perfectly fit the color of the giant branch , the mismatch being similar ( if slightly worse ) to the one mentioned above for the salaris isochrones : padova isochrones are bluer than the observations by @xmath2 0.04 mag in the giant branch . we note that the padova isochrones do include the agb phase , but it is also displaced to the blue , as compared to the observations , by the same amount as the giant branch . the position of the hb is well matched by the padova isochrones , but its extension is underpredicted . matching a cmd such as the one shown in figure [ fig1 ] is just the first step when comparing a given theoretical isochrone with cluster data ; matching the lf of the cluster is also extremely important . in fact , the lf of giant stars has a large and not often appreciated impact on the integrated spectrum in general , and on line indices in particular . in figure [ fig3 ] , we compare the lf measured from the data with the lf of the first - ascent giant - branch predicted by the best - fitting salaris theoretical isochrone of age 11 gyrs ( thick line ) . the observed lf is in good agreement with data on 47 tuc from other sources in the literature , namely hesser et al . ( 1987 ) , kaluzny et al . ( 1998 ) , and zoccali et al . ( 2000 ) ( the last being based roughly on the same data as this work ) . the theoretical lf is a good match to the observed one up to the horizontal branch . for brighter stars , however , the salaris lf underpredicts the observations by @xmath2 0.4 dex on average ( the mismatch is slightly lower at @xmath16 and higher at the location of the agb bump , at @xmath17 ) . the lf mismatch is in part due to the fact that the salaris lf does not include agb stars . however , on purely theoretical grounds it is hard to attribute the mismatch solely to the lack of these stars , as current stellar evolution models predict a ratio of @xmath2 0.15 between the lifetimes in the agb and hb phases ( cassisi et al . 2001 ) . at the agb bump ( v @xmath2 13.1 ) , the observed number of agb stars is roughly the same as first - ascent giants : in the interval 12.8 @xmath18 13.2 , there are 30 agb and 32 first - ascent rgb stars . thus , adding agb stars would imply roughly a factor of two ( 0.3 dex ) increase , in the raw theoretical lf at the agb - bump level . the difference between the observed and theoretical lfs at the level of the agb bump amounts to @xmath2 0.5 dex ( @xmath19 in figure [ fig3 ] ) , which leaves an extra 0.2 dex mismatch between the pure rgb observed lf and the theoretical one . for brighter stars , it becomes increasingly difficult to distinguish between agb and first - ascent rgb stars , but it is doubtful that the ratio changes very much . at fainter magnitudes , below the agb bump , there are even fewer observed agb stars . therefore we conclude that only at most 2/3 of the mismatch between the salaris lf and the data at the level of the agb bump could be due to the lack of agb stars in the theoretical isochrone , and that fraction is probably lower in other regions of the agb , where agb stars are even more sparsely distributed . in short , some increase in the model prediction for the number of model _ first - ascent _ giants above the hb seems also to be needed . this conclusion can be checked by comparing to the predicted lf of the padova isochrones , which include both first - ascent and agb stars explicitly . this comparison is shown in figure [ fig4 ] which illustrates theoretical lfs based on the padova isochrones for 11.2 gyrs . the thick solid line represents the total padova lf , while the thick dash - dotted line represents the same lf after removal of the agb stars . for reference , we also reproduce the original salaris lf from figure [ fig3 ] , now as a thin line . we note that because the padova isochrones are provided in coarser mass steps , the signature of the horizontal branch in the padova lf is broader than in the case of the salaris lf , but the total number of hb stars is similar . two conclusions can be drawn from this figure . first , we call attention to the very good agreement between the two sets of pure rgb theoretical isochrones ( thin line and thick dashed line ) , which means that both groups predict similar lifetimes for first - ascent giant stars . this has also been shown by zoccali & piotto ( 2000 ) . second , and more importantly , even when the agb stars in the padova isochrones are considered ( thick solid line ) , the number counts above the horizontal branch are still underpredicted by @xmath2 0.2 - 0.3 dex . this confirms our suspicion above that the theoretical number of first - ascent giants above the hb is also too smal . evidently the shortfall is similar in both the salaris and padova models . we recall that our observed lf is in good agreement with that of other works , like hesser et al . ( 1987 ) , kaluzny et al . ( 1998 ) and zoccali et al . ( 2000 ) . in particular , the lfs of hesser et al . and kaluzny et al . do not refer to the central parts of the cluster , so that it is very unlikely that the lf mismatch is due to an environmental effect , or that it may be afflicted by observational errors due to crowding . another feature that can affect the observed number ratio of giant to main sequence stars is binarity . throughout this discussion , we have adopted the turn - off region ( @xmath20 ) as the normalization point to compare the theoretical lf to the observations . however , if the number counts at the level of the turn - off and upper main sequence are underestimated due to the presence of binary stars , this could be responsible for part of the lf mismatch of figures [ fig3 ] and [ fig4 ] . the color distribution of stars within narrow magnitude ranges of the main sequence of 47 tuc indeed are not symmetrical , presenting an extended tail towards redder @xmath12 , which is a well - known signature of binary contamination . in fact , albrow et al . ( 2001 ) have determined a binary fraction of 14% @xmath5 4% from observations of upper main sequence stars in the core of 47 tuc . in a study of binarity effects on the lf of ngc 6752 , rubenstein & bailyn ( 1999 ) found that the turn - off luminosity function needs to be corrected upwards by @xmath2 0.1 dex , if the binary fraction is 22% at about 1 mag below the turn - off . we therefore conclude that , since the binary fraction of 47 tuc is lower than that of ngc 6752 , at a comparable position in the main sequence , the correction on the lf of 47 tuc is lower than 0.1 dex , and thus is not able to account for the mismatch seen in figures [ fig3 ] and [ fig4 ] . moreover , we note that a correction for the effect of binaries , while bringing a slight improvement to the match in the upper giant branch , would bring the lf into disagreement in the lower giant branch . the situation regarding the match of stellar evolution models to the lfs of first - ascent giants in clusters is controversial . that the theoretical models do not provide a perfect match to the number counts of giants in some clusters has been pointed out some time ago ( stetson 1991 , bergbusch & vandenberg 1992 , bolte 1994 , sandquist et al . 1996 , sandquist et al . 1999 ) . in a recent paper , langer , bolte & sandquist ( 2000 ) found a comparable mismatch ( _ i.e. _ @xmath2 0.2 dex underestimate above the hb ) between observed and theoretical lfs of first - ascent rgb stars for the metal - poor clusters m5 and m30 . in particular , they note that the models underpredict the number of stars at different levels of the giant branch for the two clusters . for m30 , the mismatch happens for giants below the rgb bump , whereas for m5 the models underpredict the number of giants brighter than the rgb bump , similarly to what we found for 47 tuc . langer et al . suggest that the mismatch can be due to deep mixing , which would bring fresh fuel to the hydrogen - burning shell , thus prolonging the lifetimes of first - ascent rgb stars . on the other hand , zoccali & piotto ( 2000 ) compared the observed lfs of a number of clusters over a broad range of metallicities to theoretical models by different groups , including padova , straniero et al . ( 1997 ) , previous models by salaris and collaborators ( salaris & weiss 1998 ) . contrary to the conclusions of langer et al . ( 2000 ) , they found rather good agreement for all clusters more metal - poor than _ [ fe / h ] _ @xmath2 0.7 . however , in their figure 5 a slight mismatch ( up to @xmath21 0.2 dex ) , in the same sense as found in this work , can be seen in the lf of 47 tuc for @xmath22 . while there still is controversy concerning the ability of theoretical models to predict the lf of first - ascent giants , the agb lf is largely unexplored . to our knowledge , there is as yet no direct detailed comparison in the literature between number counts of agb stars in clusters and model lfs . this is partly due to the paucity of statistically representative samples of cluster agb stars in the literature , mostly on account of their short lifetimes . on the theoretical side , agb models are subject to important uncertainties that may affect lifetime predictions ( cassisi et al . 2001 ) , mostly related to the unknown efficiency of macroscopic mechanisms , like convection and diffusion , and to the input physics of the models ( equation of state , opacities and nuclear cross sections ) . on the calibration side , the matter is further complicated by the uncertainties in the bolometric corrections of m giants , especially for the most metal - rich clusters , where molecular opacities in the v band , and in particular their dependence on metallicity , are difficult to model ( see , for instance , westera et al . 2002 ) . maraston ( 1998 ) compared model predictions , based on the fuel consumption theorem , to the fractional contribution of rgb and agb stars from ngc 6528 and ngc 6553 to the integrated light in the v band , finding good agreement . this result is difficult to reconcile with our findings , although maraston s models are based on a different set of theoretical models . moreover , maraston s test does not disentangle the effect of the number of agbs and their v - band luminosities , which are both subject to significant uncertainties . taken at face value , maraston s result suggests that there may not be a lf mismatch for clusters of near - solar metallicity , although this is at odds with the finding , by zoccali & piotto ( 2000 ) , that model lfs are a poor match to the number counts of near - solar metallicity clusters in the upper giant branch . regardless of the origin of the lf discrepancy in figure [ fig3 ] , its presence must be considered in order to account correctly for all evolutionary stages in the cmd of figure [ fig1 ] . we also want to compare spectroscopic ages based on both the raw and corrected lfs , in order to compare our results to those of vazdekis et al . ( 2001 ) , who used the salaris isochrones and ( apparently ) the associated raw lf . this will be discussed in section [ isoobs ] . another important issue concerns the interplay between the @xmath4-scale of the model isochrones and the @xmath4 @xmath23 _ ( b v ) _ relation used in the conversion of the theoretical isochrones to the observational plane . as noted above , the salaris isochrone that best matches the cluster s turn - off ( @xmath2 11 gyrs ) presents a mismatch of as much as @xmath2 0.03 mag in the color of the red giant branch . however , even a successful fit to the giant branch _ ( b v ) _ color in the cmd of a cluster would _ not _ mean that both the theoretical @xmath4-scale and the @xmath4 @xmath23 _ ( b v ) _ are individually correct . as stressed by weiss & salaris ( 1999 ) and salaris , cassisi & weiss ( 2002 ) , a suitable _ combination _ of a wrong @xmath4-scale with a wrong @xmath4 @xmath23 color relation can produce a successful match to the giant branch . for the same reason , the slight mismatch of 0.03 mag between the theoretical isochrone for 11 gyrs and the data on the giant branch of 47 tuc could be due to a wrong @xmath4-scale in the isochrone , to systematic errors in the @xmath4 @xmath23 _ ( b v ) _ calibration , or to a combination of the two . to quantify the importance of this effect , let us suppose that the assumed @xmath4 @xmath23 _ ( b v ) _ relation is correct and that the mismatch is due solely to a systematic error in the theoretical @xmath4s of the giant stars , as given by the isochrones . in this case , we can estimate by how much the @xmath4s of the giants would need to be increased ( decreased ) in order to shift the upper ( lower ) giant branch by 0.03 mag to the blue in figure [ fig1 ] . with the aid of the @xmath4 vs. _ ( b v ) _ calibration derived in paper i ( figure 4 ) , we estimate that the @xmath4s of the giants need to be increased ( decreased ) by @xmath2 40 k in order to match the upper ( lower ) giant branch . in the case of the padova isochrones , the mismatch is different , the theoretical isochrone being systematically @xmath2 0.030.05 mag bluer than the observations all along the giant branch . thus , under the same assumption as above , @xmath4s predicted by the isochrones would need to be systematically decreased by 4080 k in order to match the observed giant branch . we note that such mismatches are not implausible in view of discrepancies among current theoretical @xmath4-scales predicted by different groups , which amount to as much as 200 k ( see figure 2 of schiavon , barbuy & bruzual 2000 , and discussion in salaris et al . 2002 ) . the consequence of such a shift for our model predictions is discussed in section [ unc ] . this section compares model line indices based on theoretical isochrones to the observations . here we briefly summarize the most important ingredients of our models , referring the reader to paper i for details . our models are based on a moderately - high resolution ( 1.8 fwhm ) stellar spectral library ( jones 1999 ) and two sets of isochrones : those by salaris and collaborators , and the ones by salasnich et al . ( 2000 ) , which are computed adopting the same chemical mixture as the former but do not consider diffusion of heavy elements . the observed integrated spectra of 47 tuc are described in paper i. the procedure to compute integrated spectra from an observed cmd was described in paper i. the procedure used here is the same except that the number of stars per isochrone point is given by theory plus a salpeter imf , while in the computations from the cmd it is given by the observed number of stars . another difference refers to the @xmath4s , which are provided directly by the isochrones , so that they do not rely on a _ ( b v ) _ vs. @xmath4 relation . however , we do need to resort to such a relation when computing the weighted , fluxed spectrum for every isochrone point . our first goal in computing isochrone - based models is to recalculate the integrated model indices of vazdekis et al . ( 2001 ) using the same isochrones and spectral library , but adopting the new set of relations derived in paper i between fundamental stellar parameters ( @xmath4 and @xmath9 ) and various observables ( colors , absolute magnitudes , spectra , and line indices ) . vazdekis et al . adopt the calibrations of alonso et al . ( 1995 , 1996 , 1999 ) for all evolutionary stages . we showed in paper i that our calibrations are consistent with alonso et al.s for dwarfs but are roughly 70 k warmer at fixed _ ( b v ) _ for giants . another difference is the stellar parameters of the library stars , which were determined through different procedures . we will first concentrate on the computations based on the salaris isochrones , then contrast these results with the computations from the padova isochrones . the model spectra are smoothed and rebinned to the resolution and dispersion of the observed spectra of 47 tuc , and the line indices are measured using the definitions given in paper i. corrections are applied as in paper i , namely , corrections due to the paucity of metal - poor giants in the library , the slightly lower metallicity of 47 tuc ( from _ [ fe / h ] _ = 0.7 to 0.75 ) , and the effect of cn - strong stars . the isochrones adopted are computed for _ [ fe / h]_=0.7 and _ [ @xmath3/fe]_=+0.4 . however , in paper i we inferred corrections to bring line index predictions to _ [ fe / h]_=0.75 , which is the effective fe abundance corresponding to our model predictions . we also correct the observed indices to eliminate the contribution by blue stragglers , which of course are not included in the isochrones . the corrected ews are given in table 4 of paper i. percentage corrections for blue stragglers in each index are given in table [ tbl-4 ] . the first comparison is to salaris models without agb stars . figures [ fig5 ] and [ fig6 ] show a comparison between these models versus observed indices in a number of index @xmath23 index plots . according to these figures , different balmer lines imply different ages . @xmath1 and @xmath7 are best fitted for an age of 14 gyrs , while @xmath6 requires an even older age . most importantly , best - fitting spectroscopic ages are always older ( by at least @xmath2 3 gyrs ) than that obtained from the fit to the color and absolute magnitude of the cluster turn - off ( 1112 gyrs ) . the models also provide good fits to all metal lines , although the age for which the best fit is achieved again varies from one metal line to the next . note that the model prediction for _ ( b v ) _ is also too blue compared to the observations : for an age of 11 gyrs , the model predicts _ ( b v)_=0.79 , which is 0.07 mag bluer than the observed value ( see table 3 of paper i ) . this again implies that an older age is needed to reconcile with the observed color . on balance , these spectroscopic ages are consistent with the ones obtained by vazdekis et al . ( 2001 ) on the basis of the same set of isochrones and spectral library but using a different set of stellar parameters and calibrations . figures [ fig7 ] and [ fig8 ] reproduce the data of figures [ fig5 ] and [ fig6 ] but now using model predictions based on the padova isochrones . in order to bring all isochrones to the same footing , we first compare results based on padova isochrones after removing their agb components . the basic conclusion from figures [ fig7 ] and [ fig8 ] is that spectroscopic ages inferred from padova isochrones are larger than those based on salaris isochrones by a little more than 2 gyrs . such a difference is significantly higher than expected based on fits to the position of the cluster turn - off ( figures [ fig1 ] and [ fig2 ] ) , which yielded an age difference ( due to heavy element diffusion ) of @xmath2 1 gyr . the small discrepancy is plausibly explained by the systematically too warm giant branch in the padova models , which forces an older overall spectroscopic age . the temperature difference at the level of the horizontal branch is of the order of 50 k and increases up to the tip of the giant branch , where it reaches @xmath2 100 k. this difference in the temperature of the red giant branch is also responsible for another feature worthy of notice in figures [ fig12 ] and [ fig13 ] : that the metal lines are all weaker when padova isochrones are used ( compare with figures [ fig10 ] and [ fig11 ] ) . as discussed in section [ isocmd ] , both sets of isochrones employed in this work underestimate the number of giant stars brighter than the horizontal branch . in the case of the salaris isochrones , which do not take into account agb stars , the mismatch is of the order of 0.4 dex , while in the case of the padova isochrones , which do include agbs , it is slightly lower , of the order of 0.20.3 dex . to evaluate the impact of these underestimates on our computations , we apply empyrical corrections to both theoretical lfs to bring them into agreement with the observations . we first correct the salaris isochrones in order to bring their predicted lf into agreement with the observations ; discussion of the padova isochrones is postponed to the next sub - section . the corrected salaris lf for a model with 11 gyrs is represented in figure [ fig3 ] as the thin line . the consequence of correcting our model predictions for this effect is illustrated as the arrow in figure [ fig9 ] for the @xmath24@xmath1 diagram . since the giant light contribution is increased , the integrated spectrum becomes redder , metal lines become stronger , and hydrogen lines become weaker . with this correction to the lf , the spectroscopic age based on @xmath1 is reduced from @xmath25 14 to 11 gyr , in excellent agreement with the isochrone age in figure [ fig1 ] . we checked the size of the corrections by recomputing the integrated spectrum directly from the cmd as in paper i , but excluding agb / rgb stars so as to force agreement with the theoretical uncorrected luminosity function . differences between the line indices obtained in paper i and the ones computed with the exclusion of agb / rgb stars are essentially the same as the ones obtained adopting the two lfs shown in figure [ fig3 ] , which validates the corrections estimated here . figures [ fig10 ] and [ fig11 ] compare the lf - corrected model predictions with the observations for the other indices . virtually all line indices are well fit for ages ranging from 9 to 14 gyrs . the best fitting age for @xmath7 is 1112 gyrs , in agreement with the age inferred from @xmath1 and cmd - fitting . the notable exception is @xmath6 , for which the best - fitting age is slightly higher : @xmath2 14 gyrs . the latter result was anticipated in paper i , because the model spectrum computed directly from the cmd of the cluster also overestimates @xmath6 , and when we considered the correction of @xmath6 due to the effect of cn - strong stars , the mismatch was further increased . as discussed in section 5.3 of paper i , we have reason to believe that our correction for this effect on @xmath6 may be in error . in any case it was already clear in paper i that @xmath6 was aberrant . all the metal - line indices are well fit by models with ages of 11 - 12 gyrs , including ca4227 , which was greatly improved after correction for the effect of cn - line contamination , derived in paper i. finally , the integrated _ ( b v ) _ of the cluster is also matched for an age of 11 gyrs . the ratio of the contribution to the integrated light of turn - off stars to giant stars ( hb stars excluded ) provides a good gauge of the impact of the underestimate of the number of bright stars in the theoretical isochrone . in paper i we computed the fractional contribution to the integrated light by different evolutionary stages at a number of reference wavelengths . those numbers are given in table 5 of paper i. they are fairly robust , since they are based on a statistically representative cmd of the cluster . from table 5 of paper i , it can be seen that the ratio of turn - off light to giant light is 0.50 and 0.28 in the regions of @xmath6 and @xmath1 respectively . if the uncorrected theoretical lf is adopted , those figures change to 0.77 and 0.47 respectively . we have just shown that a 0.4 dex underestimate in the contribution of red giants to the integrated light leads to an age overestimate of about 34 gyrs for fixed hydrogen line equivalent width . it is interesting to note that this age difference is the same as the one found by vazdekis et al . ( 2001 ) between their spectroscopic and cmd - based age determinations . the fact that the amount of red giant light contribution has such an important effect on spectroscopic age estimates may be surprising , given that it is often said in the literature that spectral features in the integrated light of old stellar populations evolve with age mainly as a function of the temperature of the turn - off stars . though this statement is strictly correct , it neglects the basic fact that a significant fraction , if not most , of the _ continuum _ light of old stellar populations even in the blue is provided by giant stars ( see the discussion in section 5.5 of paper i ) . thus , our results show that the luminosity function is a key constraint on population synthesis models , upon which our ability to correctly predict the ages of old stellar populations from their integrated light is strongly dependent . in summary , we conclude that the too - large spectroscopic age of 47 tuc found in previous works is due to a combination of reasons : the neglect of important effects like @xmath3-enhancement and he - diffusion ( as previously shown by vazdekis et al . 2001 ) , and the underestimate of the contribution of bright red giants to the integrated light of the cluster . in this section we compare the observed absorption line ews with model predictions based on padova isochrones , now including agb stars . as in the previous section , the lf above the horizontal branch has been empyrically corrected in order to reproduce the observed number counts . the comparisons are displayed in figures [ fig12 ] and [ fig13 ] , where it can be seen that the spectroscopic age inferred when the lf - corrected padova isochrones are adopted is about 14 gyrs , for both @xmath1 and @xmath0 ( @xmath6 again being older ) . based on section [ modorig ] , we expect that the padova isochrones will yield a spectroscopic age that is roughly 12 gyrs older than the one that is inferred when using the salaris isochrones . aside from these systematically older ages , it is important to emphasize that , for @xmath1 and @xmath0 we achieve generally good consistency between spectroscopic and cmd - based ages for the padova models , just as we did with the salaris models . as for the model predictions for @xmath6 they are again much higher than the observations , thus leading to too old ages according to this index . we also performed computations using the original padova isochrones , i.e. , including agb stars with their theoretically predicted lf ( solid line in figure [ fig4 ] ) , which underpredicts the observed lf of bright giants by 0.20.3 dex . the spectroscopic age inferred in this case is @xmath2 23 gyrs older than the one inferred above , and thus is in disagreement with the cmd - based age . therefore , we again stress the importance of correctly matching the giant branch luminosity function in order to predict spectroscopic ages which agree with the ones based on the fit of the cluster cmd . as discussed in previous sections , it is hard to quantify to which extent the lf mismatch is due to errors in the models predictions of agb and first - ascent rgb stars . while agb stars may have larger uncertainties , we have reasons to believe that the predictions for first - ascent giants may be also in error , as noted in section [ lfun ] . in this section we gauge the effects of uncertainties in the @xmath4 and _ [ fe / h]_-scales adopted as model inputs . as discussed in paper i , the @xmath4 and _ [ fe / h]_-scales of metal - poor stars from different sources in the literature may differ by up to @xmath2100k and 0.1 dex . we also argued in section [ isocmd ] that the @xmath4s predicted by theoretical isochrones may be in error by comparable amounts . our set of stellar parameters was determined from strmgren photometry for the dwarfs and from calibrations of line - indices @xmath23 stellar parameters for the giants . our @xmath4 and metallicity - scales for the library giants are , by construction , tied to the stellar parameter calibrations of soubiran et al . ( 1998 ) ( see paper i for details ) . as in paper i , we resort to fitting functions in order to assess the changes in line indices that result from slight changes in model inputs . the stellar parameters tested were the @xmath4 and _ [ fe / h]_-scales of dwarfs and giants separately , plus the assumed _ [ fe / h ] _ of 47 tuc . for these computations , giants are considered to be all stars with @xmath26 6000 k and @xmath27 3.6 , and dwarfs are in the same @xmath4 interval but with @xmath28 3.6 ( thus including subgiants ) . from the discussion of systematic effects on these stellar parameters carried out in paper i , we chose to vary the @xmath4s of giants and dwarfs by @xmath5 75 and @xmath5 50 k respectively . the _ [ fe / h]_s of giants and dwarfs were varied by @xmath5 0.1 and @xmath5 0.05 dex respectively , while the _ [ fe / h ] _ of 47 tuc was varied by @xmath5 0.1 dex . the results are displayed in table [ tbl-4 ] , where we give the percentage variation of each line index as a function of model input changes . to aid the reader in judging the importance of these effects , we provide in the last row of table [ tbl-4 ] the percentage variations due to an increment of 1 gyr in age ( from 10 to 11 gyrs ) . the best way to interpret the numbers in table [ tbl-4 ] is to compare the percentage variation in each index with the variation due to the age change given in the last row . for instance , @xmath6 varies by @xmath5 9% for an uncertainty of @xmath5 50 k in the @xmath4s of dwarfs . it varies by 7% when age changes from 10 to 11 gyrs . therefore , the @xmath5 50 k uncertainty in the @xmath4s of dwarfs implies a @xmath29 1.3 gyr uncertainty in the age inferred from @xmath6 at the metallicity of 47 tuc . before discussing systematic effects in the input stellar parameters , we assess the effect of the mismatch of the theoretical isochrones to the red giant branch of 47 tuc in figures [ fig1 ] and [ fig2 ] . in section [ cmd ] we saw that the salaris isochrone is bluer by @xmath2 0.04 mag at the lower giant branch and redder by a comparable amount at the upper giant branch . in the case of the padova isochrones , the theoretical prediction is bluer than the observations through all the rgb , by @xmath2 0.03 mag . such color differences , if due solely to temperature effects , translate into @xmath4 mismatches of @xmath5 50 k in the case of the salaris isochrones and + 70 k in the case of the padova isochrones , according to the calibration derived in paper i. in the case of the salaris isochrones , correcting for these mismatches amounts to negligible changes in the line index predictions . in fact , the percentage variations in all indices correspond to less than 0.4 gyr , so that we do not need to perform any correction to our model predictions due to the slight mismatch in the red giant branch . this is because there is a compensation between the opposite @xmath4 mismatches in the lower and upper giant branch . in the case of the padova isochrones , as the model rgb is systematically bluer than the observations , the effect on spectroscopic age predictions is not negligible , leading to systematic age overestimates of about 1 gyr . in section [ padova ] we inferred a spectroscopic age of 14 gyrs from @xmath1 and slightly above that value from @xmath0 . correcting these predictions for the systematic temperature error in the rgb would bring them into better agreement with the age inferred from the position of the turn - off found in section [ cmd ] ( 12.5 @xmath15 t @xmath15 14.1 gyrs ) . uncertainties due to systematic errors in the input parameters of dwarf stars are smaller than the variations of the input parameters of giants . thus , they have only a relatively minor impact on model predictions , at the old ages of gcs . except for @xmath6 , all line indices vary by less than 3% when varying the @xmath4 and _ [ fe / h]_-scales of dwarfs . taking @xmath1 as illustrative , this translates to roughly a 0.5 gyr uncertainty in age due to the @xmath4-scale of dwarfs and a 0.25 gyr uncertainty due to their _ [ fe / h]_-scale . by comparison , uncertainties of 75 k or 0.1 dex in the input @xmath4 and _ [ fe / h]_-scales of giants produce a 1 gyr uncertainty in the age inferred from either @xmath1 or @xmath6 . this is particularly important in view of the current discrepancies in the @xmath4-scales of giant stars among different sets of the most recent theoretical isochrones , which amount to 200 k ( salaris et al . 2002 , schiavon , barbuy & bruzual 2000 ) . such a large discrepancy is still lingering , mostly because the position of the theoretical red giant branch is dependent on poorly constrained inputs of stellar evolutionary models , such as the mixing length parameter and the t(@xmath30 ) relation adopted as boundary condition for the interior models . on the other hand , the present work may be viewed as an important _ consistency check _ on the current theoretical salaris @xmath4s , as table [ tbl-4 ] shows that an error of @xmath5 200 k in the @xmath4s of the giants would cause a mismatch of @xmath293 gyr between the cmd and the spectroscopic ages , which is not seen . this is a valuable independent confirmation of the correctness of the salaris @xmath4-scale , at least for giants . the input metallicity of 47 tuc is also an important source of uncertainty for spectroscopic age determination . varying _ [ fe / h]_@xmath31 by @xmath50.1 dex ( the nominal errorbar quoted by carretta & gratton 1997 ) causes the spectroscopic age to vary by roughly @xmath5 1 gyr , as measured from all balmer lines . the effect of blue stragglers is also not negligible . if the observations are not corrected for the contribution of blue stragglers , ages are underestimated by slightly more than 1 gyr when measured from @xmath6 . the effect on @xmath7 and @xmath1 is less important , leading to underestimates of @xmath2 0.6 and 0.5 gyrs respectively . in this paper , we compared model single stellar population spectra to the integrated spectrum of the metal - rich galactic globular cluster 47 tuc . the model spectra were computed from two sets of state - of - the - art theoretical isochrones . we tested the sensitivity of our predictions to variations in a number of key model inputs related to the spectral library employed , the theoretical isochrones , and the relations between fundamental and observed stellar parameters . our main conclusions can be summarized as follows : different sets of isochrones predict slightly different ages when compared to the cluster data on the color - magnitude diagram . in the case of the salaris set , the isochrone providing the best match to the turn - off of the cluster has an age of 11 - 12 gyrs . vazdekis et al . ( 2001 ) , using the same set of isochrones , obtained an age younger by about 1 gyr . the difference is minor and is due to the different reddenings assumed . when the isochrones of salasnich et al . ( 2000 ) are used , a slightly older age is obtained ( 12.5 - 14 gyrs ) the reason being that the latter set of isochrones does not consider diffusion of heavy elements . a final decision as to which set of isochrones is best awaits a better knowledge of the extent to which atomic diffusion is operating in main sequence stars . when compared to the observations in the luminosity function domain , both sets of theoretical isochrones adopted in this work underestimate the number of giant stars brighter than the horizontal branch . in the case of the salaris isochrones , which do not include agb stars , the mismatch amounts to 0.30.4 dex . the padova set , even though it does include agb stars , is also offset , though by a smaller amount ( @xmath2 0.20.3 dex ) . it is unclear to us what causes this discrepancy . at the level of the agb bump , roughly 2/3 of the mismatch seems to be due to the lack of agb stars in the isochrones , this fraction being probably lower in other parts of the upper giant branch . however , even after performing an approximate correction for the absence of agb stars , there appears to be a small residual discrepancy between models and observations for first - ascent giants . we conclude that theoretical predictions to the luminosity functions of both the rgb and the agb may be too low , though it is difficult to quantify by how much in each case . spectroscopic ages inferred from @xmath1 and @xmath0 when adopting isochrones without any agb stars are systematically older , by at least 3 gyrs , than the ages inferred from the fit to the cluster s cmd . moreover , metal - lines are systematically too weak and the integrated _ ( b v ) _ is too blue by 0.07 mag . empirically correcting the theoretical lfs to match the observed luminosity function reduces the spectroscopic age to 1112 gyrs in the case of the salaris isochrones , and @xmath2 13 gyrs in the case of the padova ones as determined from @xmath1 and @xmath0 ( the latter is obtained after correction for a systematic mismatch of the cluster s rgb by padova isochrones ) . such revised ages are now in excellent agreement with the ages inferred from fitting both sets of isochrones to the cmd . also _ ( b v ) _ and metal - lines studied are very well fit by the lf - corrected models . the best value found from the fit to @xmath6 is still higher by @xmath2 3 gyrs than found for the other balmer lines . this mismatch is in the same sense found in paper i , where @xmath6 computed directly from the cmd - based synthesis was relatively stronger than @xmath0 or @xmath1 . as @xmath6 is more strongly affected by c , n abundance anomalies , we suggested in paper i that the discrepancy was due to our failure to properly model and correct for cn lines within the passband and continuum windows of the index . we estimate the impact on spectroscopic ages of uncertainties in a number of model inputs , such as the @xmath4 and _ [ fe / h]_-scales of dwarfs and giants , the luminosity function of the red giant branch and the metallicity of the cluster . spectroscopic age - dating of old stellar populations is mostly influenced by uncertainties coming from the @xmath4 and _ [ fe / h]_-scale of giant stars , as well as the luminosity function of the upper red giant branch and the metallicity adopted for the cluster . using @xmath1 , a 1 gyr uncertainty is caused by an error of 75 k in the giant @xmath4-scale , a 0.1 dex error in the giant - branch luminosity function , and a 0.1 dex error in either the _ [ fe / h]_-scale of giants or the _ [ fe / h ] _ of the stellar population . because the uncertainties in the parameters of dwarf stars are lower , their impact on model predictions is generally less important . the contribution by blue stragglers to the integrated line indices is important only for @xmath6 and slightly less so for @xmath7 . when @xmath6 is corrected for the contribution of blue stragglers , we obtain an age 1 gyr older . in the case of @xmath7 , the age inferred is 0.5 gyr older . in summary , we conclude that the exceedingly high spectroscopic ages found for 47 tuc in previous works was due to : 1 ) the use of theoretical isochrones that neglect important effects such as he - diffusion and @xmath3-enhancement , as shown by vazdekis et al . ( 2001 ) , and 2 ) the underestimate of the number of giant stars above the horizontal branch , which was partly due to the omission of agb stars from the isochrones , and partly to a genuine apparent shortfall in the predicted numbers of agb stars and , most likely , first - ascent giants as well . another requirement , which is discussed in paper i , is on the iron abundance of the cluster . models provide a good match if _ [ fe / h ] _ is 0.05 dex lower than found by carretta & gratton ( 1997 ) . such a shift in _ [ fe / h ] _ is well within the errorbars of carretta & gratton s analysis and the uncertainties in the _ [ fe / h]_-scale of our stellar library . with regard to our long term goal of constructing metal - rich stellar population models for elliptical galaxies , we have made considerable progress . first , the ages from @xmath1 and @xmath7 agree , after the latter is corrected for the effect of c , n abundance variations ( see paper i ) . second , all metal lines are correctly predicted for the same ages obtained from balmer lines . third , the giant temperature scale of the salaris models looks to be generally correct . on the other hand , @xmath6 is shown to be a problematic feature , embedded as it is in a forest of cn lines . this problem is likely to be further exacerbated in elliptical galaxies , which have notoriously strong cn features . moreover , the present models do not include variations in type ia vs. type ii supernovae element ratios ( trager et al . 1998 ) , nor do they probe the high - metallicity regime needed for ellipticals . finally , we have identified an error in the upper giant branch luminosity function of current evolutionary models , which underestimate the aggregate number of luminous red giants in 47 tuc by roughly a factor of 2 . we have furthermore shown that an accurate lf for these stars is needed for accurate spectroscopic age determination . a major unknown is whether the red - giant excess of 47 tuc extends to all metal - rich old populations , and what its magnitude might be . testing the luminosity function in old metal - rich populations has emerged as an important prerequisite for stellar population studies . these and further improvements are left for future works . we would like to thank maurizio salaris for making his isochrones available . stefano covino is thanked for the red integrated spectrum of 47 tuc . we would also like to thank alexandre vazdekis , maurizio salaris , manuela zoccali , achim weiss and mike bolte for helpful discussions . the referee , brad gibson , is thanked for valuable suggestions that greatly improved this paper . r.p.s . thanks the hospitality of the physics dept . of the university of north carolina , chapel hill , where part of this work was developed . likewise , j.a.r . thanks the astronomy department at uc , santa cruz for hospitality during a visit in which part of this work was developed . this work has made extensive use of the simbad database . r.p.s . acknowledges support provided by the national science foundation through grant gf-1002 - 99 and from the association of universities for research in astronomy , inc . , under nsf cooperative agreement ast 96 - 13615 , and cnpq / brazil , for financial help ( 200510/99 - 1 ) . this research has also been supported by nsf grant ast-9900720 to the university of north carolina , and by nsf grants ast-9529098 and ast-0071198 to the university of california , santa cruz . albrow , m.d . , gilliland , r.l . , brown , t.m . , edmonds , p.d . , guhathakurta , p. & sarajedini , a. 2001 , apj , 559 , 1060 alonso , a. , arribas , s. & martnez - roger 1995 , , 297 , 197 alonso , a. , arribas , s. & martnez - roger 1996 , , 313 , 873 alonso , a. , arribas , s. & martnez - roger 1999 , , 139 , 335 basu , s. , pinsonneault , m.h . & bahcall , j.n . 2000 , apj , 529 , 1084 bergbusch , p.a . & vandenberg , d.a . 1992 , apjs , 81 , 163 bolte , m. 1994 , apj , 431 , 223 carretta , e. & gratton , r.g . 1997 , , 121 , 95 cassisi , s. , castellani , v. , deglinnocenti , s. , piotto , g & salaris , m. 2001 , a&a , 366 , 578 chaboyer , b. , fenton , w.h . , nelan , j.e . , patnaude , d.j . & simon , f.e . 2001 , apj , 562 , 521 cohen , j.g . , blakeslee , j.p . & ryzhov , a. 1998 , apj , 496 , 808 deliyannis , c.p . & demarque , p. 1991 , apj , 379 , 216 . gibson , b.k . , madgwick , d.s . , jones , l.a . , da costa , g.s . & norris , j.e . 1999 , 118 , 1268 gratton , r.g . 2001 , a&a , 369 , 87 hesser , j.e . , harris , w.e . , vandenberg , d.a . , allwright , j.w.b . , shott , p. & stetson , p.b . 1987 , , 99 , 739 howell , j.h . , guhathakurta , p. & gilliland , r.l . 2000 , , 112 , 1200 jones , l.a . 1999 , phd thesis , university of north carolina kaluzny , j. , wyzocka , a. , stanek , k.z . & krzemiski , w. 1998 , aca , 48 , 439 langer , g.e . , bolte , m. & sandquist , e. 2000 , apj , 529 , 936 lebreton , y. , perrin , m .- n . , cayrel , r. , baglin , a. & fernandes , f. 1999 , a&a , 350 , 587 maraston , c. 1998 , mnras , 300 , 872 richer , h.b . , et al . 1996 , , 463 , 602 rose , j.a . 1994 , 107 , 206 rubenstein , e.p . & bailyn , c.d . 1999 apj , 513 , l33 salaris , m. & weiss , a. 1998 , 335 , 943 salaris , m. & weiss , a. 2001 , 376 , 955 salaris , m. , cassisi , s. & weiss , a. 2002 , pasp , 114 , 375 salasnich , b. , girardi , l. , weiss , a. & chiosi , c. 2000 , a&a , 361 , 1023 sandquist , e. , bolte , m. , stetson , p.b . & hesser , j.e . 1996 , apj , 470 , 910 sandquist , e. , bolte , m. , langer , g.e . , hesser , j.e . & mendes de oliveira , c. 1999 , apj , 518 , 262 schiavon , r.p . , barbuy , b. & bruzual a. , g. 2000 , 532 , 453 schiavon , r.p . , faber , s.m . , rose , j.a . & castilho , b.v . 2002 , apj , submitted ( paper i ) soubiran , c. , katz , d. & cayrel , r. 1998 , , 133 , 221 spite , f. & spite , m. 1982 , a&a , 115 , 357 stetson , p.b . 1991 , in the formation and evolution of star clusters , ed . k. janes ( asp conf 13 ) , 88 straniero , o. , chieffi , a. & limongi , m. 1997 , apj , 490 , 425 trager , s.c . , worthey , g. , faber , s.m . burstein , d. & gonzlez , j.j . 1998 , apjs , 116 , 1 vazdekis , a. , salaris , m. , arimoto , n. & rose , j.a . , 2001 , apj , 549 , 274 weiss , a. & salaris , m. 1999 , 346 , 897 westera , p. , lejeune , t. , buser , r. , cuisinier , f. & bruzual , g. 2002 , a&a , 381 , 524 zoccali , m. & piotto , g. 2000 , a&a , 358 , 943 zoccali , m. , cassisi , s. , bono , g. , piotto , g. , rich , r.m . & djorgovski , s.g . 2000 , apj , 538 , 289 @xmath5 4 & @xmath29 11 & @xmath5 1.5 & @xmath29 4 & @xmath5 4.5 & @xmath29 6 & @xmath29 8.5 & @xmath29 3.5 & @xmath32 75 k + @xmath5 9 & @xmath29 1 & @xmath5 1.5 & @xmath29 1.5 & @xmath5 1.5 & @xmath29 1 & @xmath29 1 & @xmath29 @xmath15 1 & @xmath33 50 k + @xmath29 11 & @xmath5 2 & @xmath5 2 & @xmath5 5 & @xmath5 3 & @xmath5 3.5 & @xmath5 5.5 & @xmath5 4.5 & @xmath34_{giants } \pm$ ] 0.1 + @xmath29 2.5 & @xmath5 1 & @xmath5 1 & @xmath5 1.5 & @xmath5 @xmath15 1 & @xmath5 1 & @xmath5 1 & @xmath5 1 & @xmath34_{dwarfs } \pm$ ] 0.05 + @xmath29 7.5 & @xmath5 2.5 & @xmath5 4 & @xmath5 6 & @xmath5 5 & @xmath5 4 & @xmath5 5 & @xmath5 5 & @xmath34_{47 tuc } \pm$ ] 0.1 + 30 & + 10 & -6 & + 10 & -10 & + 5 & + 12 & + 11.5 & lf correction + 9.5 & + 1 & 2 & + 1 & 2 & @xmath15 1 & @xmath15 1 & @xmath15 1 & no bs + 7 & + 4 & 3 & + 4 & 4 & + 4 & + 2 & + 2 & 10 @xmath35 11 gyrs +

in this paper , we report our studies on the gaseous and chemical properties of a relatively large sample ( 53 members ) of blue compact dwarf galaxies ( bcds ) . the results of correlations among the oxygen abundance , stellar mass , gas mass , baryonic mass , and gas fraction are present both for e- and i - type bcds , which are classified according to loose & thuan ( 1985 ) and show elliptical and irregular outer haloes , respectively . these correlations of i - type bcds show similar slopes to those of e - type ones . however , in general , e - type bcds are more gas - poor and metal - rich than i - type ones at a given baryonic mass . based on these results , we suggest that e - type bcds , at least a part of them , and i - type ones might be likely at different evolutionary phases and/or having different progenitors . our investigation of the correlation between oxygen abundance and gas fraction shows that bcds appear to have not evolved as isolated systems , but to have experienced some gas flows and/or mergers .

1212.5159

dwarf galaxies are by far the most numerous galaxies in the universe ( e.g. mateo 1998 ) and play a crucial role in our understanding of the formation and evolution of galaxies . they are proposed to be the building - blocks in hierarchical scenarios of galaxy formation ( kauffmann et al . dwarf galaxies are typically metal - poor ( e.g. mateo 1998 ) , and have low stellar masses . the low metallicities of these systems make them having similar conditions of high - redshift galaxies . therefore , dwarf galaxies at the local universe can be used as a proxy to reveal the physical properties of some galaxies at the early universe . based on their morphological properties and luminosity , dwarf galaxies can be classified into different types , e.g. , blue compact dwarfs ( bcds ) , dwarf ellipticals ( des ) , dwarf irregulars ( dis ) , and etc . bcds , which are the focus of the current paper , have dramatically different properties compared to normal dwarf galaxies ( zwicky 1966 ; gil de paz et al . 2003 , hereafter g03 ) . they are small , gas - rich ( hi mass fraction typically higher than 30% ; please see section 2.3 and 3.1 ) extragalactic objects . being metal - poor ( @xmath0 , here @xmath1 ( allende prieto et al . 2001 ; asplund et al . 2009 ) ; e.g. terlevich et al . 1991 ; izotov et al . 1994 ; izotov & thuan 1999 ; hunter & hoffman 1999 ) , these galaxies are allowing us to study the star formation and chemical enrichment in a near - primordial environment . as such , they are important to our understanding of galaxy formation and evolution . bcds are spectroscopically characterized by a faint , blue optical continuum accompanied , in most cases , by intense and narrow emission lines , due to the intense star formation activity distributed in one to several star - forming regions ( cairs et al . . des have smooth visual appearances and low gas content , and they represent the bright , gas - poor dwarf galaxies which have little star formation ( e.g. binggeli 1994 ) . by contrast , dis are characterized by a lack of spiral arms and an irregular visual appearance ( see review by hunter 1997 ) . they have low dust content ( e.g. , calzetti 2001 ) and large gas reservoirs ( see , e.g. , hoffman et al . central surface brightness increases with increasing luminosity in all of these types of dwarf galaxies ( see recent review of tolstoy et al . 2009 ) , thus it seems natural to consider a possible evolutionary relationship between bcds , dis and des ( e.g. thuan 1985 ; ferguson & binggeli 1994 ; papaderos et al . 1996 ) . however , the relation between bcds , dis , and des / dsphs remains unclear . current knowledge bearing on this question comes mainly from analysis of the photometric properties ( such as structure , surface brightness profile , and color ) of these three types of dwarf galaxies ( e.g. thuan 1985 ; james 1994 ; papaderos et al . 1996 ; noeske et al . 2000 ; gil de paz & madore 2005 ; vaduvescu et al . the results from different authors seem to be contradictory . for example , thuan ( 1985 ) suggests that bcds and dis are the same type of galaxy , differing only in their present star formation rate , and they can evolve into des . however , james ( 1994 ) obtains an opposite result that bcds are probably des undergoing starburst activities , while dis are a fundamentally distinct population from them . the results in papaderos et al . ( 1996 ) show that bcds have a brighter central surface brightness and a smaller exponential scale length than dis and des at a given @xmath2-band luminosity . however , gil de paz & madore ( 2005 ) claim that at least @xmath315% of bcds in their sample have structural properties compatible with being des ( also see papaderos et al . 1996 ) experiencing a burst of star formation , whereas vaduvescu et al . ( 2006 ) conclude that bcds have similar structures to dis . recently , vaduvescu & mccall ( 2008 ) suggest that there are strong connections among bcds , dis , and des since these galaxies lie precisely on a fundamental plane defined by the @xmath4-band absolute magnitude , the central surface brightness , and the neutral hydrogen line width ( also see mccall et al . 2012 ) . some other attempts to address this issue rely on the chemical evolution of dwarf galaxies . richer & mccall ( 1995 ) show that [ o / fe ] for a number of field des are larger than those in field dis at similar luminosities and conclude that there is unlikely an evolutionary connection between des and dis . vaduvescu et al . ( 2007 ; hereafter vmr07 ) argue that the evolution of bcds has been similar to that of dis since they have the same metallicity - gas fraction correlations . however , the sample in vmr07 is too small ( 14 bcds ) , and therefore might suffer large uncertainties . moreover , the correlation between metallicity and the gas fraction for bcds and dis shows large dispersions ( see figure 10 in vmr07 ) . in our previous works , we have studied the luminosity ( or mass)-metallicity ( _ l - z _ ) relation ( zhao et al . 2010 ; hereafter zgg10 ) , stellar population and star formation ( zhao et al . 2011 ; hereafter zgg11 ) for a sample of bcds . our results show that bcds are old systems currently experiencing starbursts , and follow a well - defined _ l - z _ relation . based on these two works and g03 , along with hi data compiled from literature ( see reference in table 1 ) , we can further investigate the evolution of gaseous and chemical properties of bcds . in this paper , we will explore correlations among different galactic parameters , i.e. oxygen abundance , stellar mass , gas mass , baryonic mass , and gas mass fraction . we also compare these relations for morphologically different types of bcds and dis . our investigations on such correlations may shed some light on the answer to the question of whether there is an evolutionary connection between bcds , dis and des . the paper is organized as following : section 2 describes the sample of bcds and the data . in section 3 , we first calculate the gas masses using the compiled hi fluxes and derive the stellar masses , and then present several relationships involving stellar mass , gas mass and oxygen abundance . we also do some comparisons between dis and bcds in this section . based on the results shown in section 3 , we discuss possible evolutionary connections between different types of dwarf galaxies , and gas flows in bcds , in section 4 . a brief summary is given in the last section . lcccccccccccc haro 2 & 31.67 & 13.39 & 12.87 & ne & @xmath5 & @xmath6 & @xmath7 & 9.43 & 2 & @xmath8 & @xmath9 & 1 + haro 3 & 30.79 & 13.22 & 12.61 & ie & @xmath10 & @xmath6 & @xmath11 & 8.92 & 2 & @xmath12 & @xmath13 & 1 + haro 4 & 29.73 & 15.59 & 15.44 & ii & @xmath14 & @xmath6 & @xmath15 & 7.01 & 4 & @xmath16 & @xmath17 & 1 + haro 8 & 31.02 & 14.27 & 13.20 & ne & @xmath18 & @xmath6 & @xmath19 & 8.54 & 2 & @xmath20 & @xmath21 & 2 + haro 14 & 30.73 & 13.65 & 12.91 & ie & @xmath22 & n2 & @xmath23 & 8.93 & 1 & @xmath24 & @xmath25 & 1 + i zw 18 & 30.50 & 16.05 & 16.24 & i0 & @xmath26 & @xmath6 & @xmath27 & 6.77 & 4 & @xmath28 & @xmath29 & 2 + i zw 123 & 30.34 & 15.42 & 14.85 & ne & @xmath30 & @xmath6 & @xmath31 & 7.62 & 2 & @xmath32 & @xmath33 & 1 + ii zw 40 & 29.96 & 11.87 & 11.10 & iim & @xmath34 & @xmath6 & @xmath35 & 8.08 & 1 & @xmath36 & @xmath37 & 2 + ii zw 70 & 31.36 & 14.84 & 14.29 & ii & @xmath38 & @xmath6 & @xmath39 & 8.37 & 1 & @xmath40 & @xmath41 & 1 + ii zw 71 & 31.36 & 14.45 & 13.54 & ii & @xmath42 & @xmath6 & @xmath43 & 8.74 & 3 & @xmath44 & @xmath45 & 1 + vii zw 403 & 28.41 & 14.11 & 13.58 & ie & @xmath46 & @xmath6 & @xmath47 & 7.31 & 1 & @xmath48 & @xmath49 & 2 + mrk 05 & 30.60 & 15.13 & 14.56 & ie & @xmath50 & @xmath6 & @xmath51 & 8.12 & 1 & @xmath52 & @xmath53 & 2 + mrk 67 & 30.77 & 16.10 & 15.34 & ne & @xmath54 & @xmath6 & @xmath55 & 7.75 & 2 & @xmath56 & @xmath57 & 5 + mrk 86 & 29.20 & 12.07 & 11.49 & ier & @xmath58 & n2 & @xmath59 & 8.98 & 1 & @xmath60 & @xmath61 & 2 + mrk 108 & 31.69 & 15.15 & 14.66 & iim & @xmath62 & @xmath6 & @xmath63 & 8.09 & 1 & @xmath64 & @xmath65 & 2 + mrk 178 & 28.11 & 14.15 & 13.60 & ie & @xmath66 & @xmath6 & @xmath67 & 7.14 & 2 & @xmath68 & @xmath69 & 2 + mrk 209 & 28.82 & 14.15 & 13.94 & ie & @xmath70 & @xmath6 & @xmath71 & 7.25 & 1 & @xmath72 & @xmath73 & 2 + mrk 324 & 32.01 & 15.17 & 14.60 & ne & @xmath74 & n2 & @xmath75 & 8.76 & 1 & @xmath76 & @xmath77 & 5 + mrk 328 & 31.75 & 14.93 & 14.18 & ne & @xmath78 & @xmath6 & @xmath79 & 8.89 & 1 & @xmath80 & @xmath81 & 5 + mrk 409 & 31.64 & 14.37 & 13.33 & ier & @xmath82 & n2 & @xmath83 & 9.58 & 3 & @xmath84 & @xmath85 & 2 + mrk 450 & 30.57 & 14.44 & 13.65 & ie & @xmath86 & @xmath6 & @xmath87 & 8.48 & 1 & @xmath88 & @xmath89 & 2 + mrk 475 & 29.93 & 16.20 & 15.67 & ne & @xmath90 & @xmath6 & @xmath91 & 7.44 & 2 & @xmath92 & @xmath93 & 4 + mrk 600 & 30.82 & 14.85 & 14.82 & ie & @xmath94 & @xmath6 & @xmath95 & 7.81 & 1 & @xmath96 & @xmath97 & 2 + mrk 900 & 31.37 & 14.17 & 13.56 & ne & @xmath98 & @xmath6 & @xmath99 & 8.96 & 1 & @xmath100 & @xmath101 & 4 + mrk 996 & 31.88 & 15.01 & 14.08 & ne & @xmath102 & n2 & @xmath103 & 9.11 & 1 & @xmath104 & @xmath105 & 4 + mrk 1313 & 32.51 & 16.02 & 15.50 & i0 & @xmath106 & @xmath6 & @xmath107 & 8.63 & 3 & @xmath108 & @xmath109 & 5 + mrk 1329 & 31.02 & 14.08 & 13.38 & ie & @xmath110 & @xmath6 & @xmath111 & 8.64 & 1 & @xmath112 & @xmath113 & 5 + mrk1416 & 32.64 & 16.32 & 15.75 & ii & @xmath114 & @xmath6 & @xmath115 & 8.34 & 4 & @xmath116 & @xmath117 & 4 + mrk1423 & 31.55 & 14.90 & 13.52 & ne & @xmath118 & n2 & @xmath119 & 8.85 & 1 & @xmath92 & @xmath120 & 4 + mrk 1450 & 30.83 & 15.75 & 15.09 & ne & @xmath62 & @xmath6 & @xmath121 & 7.82 & 2 & @xmath122 & @xmath123 & 7 + mrk 1480 & 32.18 & 16.17 & 15.56 & ne & @xmath124 & @xmath6 & @xmath125 & 7.93 & 3 & @xmath126 & @xmath127 & 7 + mrk 1481 & 32.18 & 16.19 & 15.57 & iim & @xmath128 & n2 & @xmath129 & 8.46 & 3 & @xmath130 & @xmath131 & 7 + ngc 2915 & 27.78 & 11.93 & 10.96 & ie & @xmath132 & n2 & @xmath133 & 8.30 & 1 & @xmath134 & @xmath135 & 3 + ngc 4861 & 30.50 & 12.68 & 11.91 & iic & @xmath136 & @xmath6 & @xmath137 & 8.53 & 1 & @xmath138 & @xmath139 & 2 + sbs1054 + 504 & 31.52 & 16.08 & 15.46 & ne & @xmath140 & @xmath6 & @xmath141 & 8.29 & 3 & @xmath142 & @xmath143 & 7 + sbs1147 + 520 & 31.39 & 16.95 & 15.98 & ne & @xmath102 & n2 & @xmath144 & 8.02 & 4 & @xmath145 & @xmath146 & 7 + sbs1331 + 493 & 29.98 & 14.87 & 14.16 & ie & @xmath147 & @xmath6 & @xmath148 & 8.00 & 4 & @xmath149 & @xmath150 & 4 + sbs1415 + 437 & 30.02 & 15.43 & 14.77 & iic & @xmath151 & @xmath6 & @xmath152 & 7.74 & 4 & @xmath153 & @xmath154 & 4 + sbs1428 + 457 & 32.74 & 15.42 & 14.67 & ii & @xmath155 & @xmath6 & @xmath156 & 9.22 & 3 & @xmath157 & @xmath158 & 7 + ugca 184 & 31.81 & 15.99 & 15.80 & ii & @xmath159 & @xmath6 & @xmath160 & 7.97 & 3 & @xmath161 & @xmath162 & 4 + ugca 412 & 33.07 & 15.55 & 14.62 & ne & @xmath163 & @xmath6 & @xmath164 & 9.21 & 4 & @xmath165 & @xmath166 & 4 + um 133 & 31.91 & 15.41 & 14.51 & iic & @xmath167 & @xmath6 & @xmath168 & 8.77 & 4 & @xmath169 & @xmath170 & 5 + um 323 & 32.26 & 16.09 & 15.24 & ie & @xmath171 & @xmath6 & @xmath172 & 8.49 & 3 & @xmath108 & @xmath109 & 5 + um 408 & 33.58 & 17.46 & 16.70 & ne & @xmath173 & @xmath6 & @xmath174 & 8.46 & 4 & @xmath175 & @xmath176 & 5 + um 439 & 30.73 & 14.77 & 14.09 & ie & @xmath177 & @xmath6 & @xmath178 & 7.83 & 3 & @xmath179 & @xmath180 & 5 + um 452 & 31.45 & 15.25 & 14.07 & ne & @xmath181 & @xmath6 & @xmath182 & 8.80 & 3 & @xmath183 & @xmath184 & 5 + um 455 & 33.64 & 17.02 & 16.26 & ii & @xmath185 & @xmath6 & @xmath186 & 8.66 & 4 & @xmath187 & @xmath188 & 5 + um 491 & 32.20 & 15.54 & 14.95 & ne & @xmath22 & n2 & @xmath189 & 8.68 & 3 & @xmath190 & @xmath191 & 5 + um 533 & 30.34 & 14.63 & 13.64 & ie & @xmath192 & @xmath6 & @xmath193 & 8.55 & 4 & @xmath194 & @xmath195 & 5 + vcc 0130 & 31.02 & 17.05 & 16.27 & ie & @xmath22 & n2 & @xmath196 & 7.65 & 3 & @xmath197 & @xmath198 & 6 + vcc 0459 & 31.02 & 14.95 & 14.13 & ii & @xmath199 & @xmath6 & @xmath200 & 8.48 & 1 & @xmath201 & @xmath202 & 6 + vcc 0655 & 31.02 & 13.32 & 12.12 & ier & @xmath203 & n2 & @xmath204 & 9.43 & 1 & @xmath205 & @xmath206 & 6 + vcc 0848 & 31.02 & 15.03 & 14.10 & iim & @xmath207 & @xmath6 & @xmath11 & 8.43 & 1 & @xmath208 & @xmath209 & 6 + using a unified concept of bcd galaxy , g03 have compiled a bcd sample ( including 105 members ) from several exploratory studies . the galaxies in their sample were selected by putting forward a new set of quantitative classification criteria . this new set of criteria is a combination of the galaxy s color , morphology and luminosity . briefly , a bcd galaxy has to fulfill the following three criteria ( referring to _ blue , compact _ and _ dwarf _ , respectively ) : ( 1 ) @xmath210 , where @xmath211 and @xmath212 are the peak surface brightness of _ b_- and _ r_-band , respectively , ( 2 ) @xmath213 mag arcsec@xmath214 , and ( 3 ) the absolute magnitude @xmath215 mag ( see g03 for details ) . to study the chemical evolution of bcds , we need to know their metallicities and gaseous properties . based on the sample in g03 , zgg10 have selected a subsample of bcds whose metallicities have been measured or having sdss spectroscopic data . therefore , the sample we used here is derived from zgg10 and a galaxy is selected if its hi data has been given in literature . the final sample contains 53 bcd galaxies , whose basic properties are listed in table 1 . the distance , @xmath2 and @xmath216 photometric data and morphological type ( except three galaxies ; see below ) are all adopted from g03 . except for several galaxies whose distance had been determined by measuring the magnitude of the tip of the red giant branch ( rgb ) , the majority were computed using the galactic standard of rest velocity of the galaxies assuming a hubble constant , @xmath217 , of 70 km s@xmath218 mpc@xmath218 ( freedman et al . 2001 ) . for the photometric data , the foreground galactic extinctions have been corrected in g03 , using @xmath219 values determined following schlegel et al . ( 1998 ) and the galactic extinction law of cardelli et al . ( 1989 ) with @xmath220 . we did not attempt to correct for internal extinction , since this is known to vary spatially in bcds ( e.g. , cannon et al . 2002 ; garca - lorenzo et al . 2008 ; cairs et al . the internal extinction obtained through balmer lines and/or continuum fitting based on the aperture spectra can only be applied to a small part of a galaxy . in addition , it was derived only for about 40% of the total sample . the morphological types of these bcds were classified according to loose & thuan ( 1986 ) . however , for mrk 005 ( cairs et al . 2001b ; amorn et al . 2009 ) , mrk 178 ( schulte - ladbeck et al . 2000 ; noeske et al . 2003 ) and mrk 1329 ( noeske et al . 2003 ; vaduvescu et al . 2006 ) , we adopt their morphological types as ie , based on the deeper optical and/or ir observations in the above references . please refer to g03 for details of the aforementioned properties . in the following , we refer to all ne and ie bcds ( showing an outer elliptical envelope ) collectively as e - type " bcds , and refer to all ii bcds ( showing an irregular outer halo ; here including two i0 bcds ) as ii or i - type " bcds unless otherwise stated . the oxygen abundances are taken from zgg10 ( also see references therein ) . in the majority of the cases the oxygen abundances have been determined on the basis of electronic temperatures measured from the [ oiii]@xmath221 line ( i.e. @xmath6-method ) . in a few cases @xmath6-based abundances are not available , we adopt the n2-based oxygen abundances , which have been derived with the empirical n2-o / h relation proposed by denicol et al . ( 2002 ) ( i.e. n2-method , where n2@xmath222([nii]@xmath2236584/h@xmath224 ) ) . as shown in zgg10 , for the bcd sample studied here , metallicities derived with the n2-method are generally consistent with those derived with the @xmath6 method , and have an accuracy of @xmath225 dex ( also see denicol et al . 2002 ; pettini & pagel 2004 and salzer et al . 2005 ) . please see zgg10 and references therein for details . in figure 1 we plot the integrated @xmath226 color and the absolute @xmath2 magnitude ( @xmath227 ) distributions for the g03 sample and our subsamples using different histograms . we can see that ours have similar distributions of @xmath226 and @xmath227 to the g03 sample , except that our ii subsample contains a higher fraction of galaxies with @xmath228 mag and a lower fraction of galaxies with @xmath229 . therefore , our subsamples can be representative of the total sample of g03 . however , please note that none of these samples are complete , although they covers a large range of galactic parameters . magnitude ( upper ) and ( @xmath230 ) color ( bottom ) for the three bcd samples . our subsamples ( hatched filled and solid open ) show similar distributions to the g03 sample ( dashed open).,scaledwidth=48.0% ] in order to explore whether their is an evolutionary link between bcds and dis by means of comparing their gaseous and chemical properties , we also include a sample of field dis adopted from lee et al . ( 2003 ) in our analysis . lee et al . ( 2003 ) sample is compiled from the literature and consists of 22 dis . the data for this di sample , including metallicity , gas mass and stellar mass , are all adopted directly from lee et al . the metallicities are determined by the direct method , and the gas masses are derived using the same method as described below . whereas for the stellar mass , we will check the consistency in the next section . the stellar mass , @xmath231 , of a galaxy , can be derived from its luminosity and mass - to - light ratio ( @xmath232 ) , which can be estimated by comparing observed galaxy colors ( e.g. bell & de jong 2001 ) , broadband spectral energy distributions ( e.g. walcher et al . 2008 ) , or full optical spectra ( e.g. cid fernandes et al . 2007 ) with predications from population synthesis models . for 21 galaxies in our sample , their stellar masses have been derived in zgg11 by using optical spectra and stellar population synthesis code , starlight ( cid fernandes et al . 2005 , 2007 ) . during the fitting process , the bruzual & charlot ( 2003 ) models were adopted , and the spectra of the simple stellar population library were computed with the salpeter ( 1955 ) initial mass function ( imf ) , padova 1994 models , and the stelib library ( le borgne et al . 2003 ) , by limiting the metallicities to @xmath233 for this metal poor sample . stellar masses obtained with this method are represented by 2 ( using integrated spectra ; moustakas & kennicutt 2006 ) or 3 ( using sdss fiber spectra , and the fraction of the total galaxy luminosity in the sdss _ i _ band inside the fiber had been used to correct for the aperture effect ) in column 10 of table 1 . please refer to zgg11 for more details . for the rest galaxies , we use the @xmath232 ratio - color relations given by bell & de jong ( 2001 ) , which are derived by using the bruzual & charlot ( 2003 ) model with a salpeter imf , to convert luminosity to stellar mass . due to the ongoing starburst activities in bcds , we prefer to use the near infrared luminosity ( @xmath234 ) and @xmath235 to calculate the stellar mass . this is because that the @xmath236-band luminosity is relatively less sensitive to recent star formation activity and dust extinction , and can well approximate the stellar mass of a galaxy . the @xmath235 ratio - color relation , as given by bell & de jong ( 2001 ) , is applied to correct for the dependence of @xmath235 on star formation history . for our metal - poor bcds , we adopt the result derived with metallicity @xmath237 . in addition , 15 galaxies in our sample have both optical spectra and nir photometric data . therefore , we can compare the derived stellar masses . we find that the nir - based stellar masses are systematically larger ( 0.11 dex with a dispersion of 0.23 dex ) than spectrum - based ones . therefore , we subtract this systematic difference from the nir - based stellar masses . the results obtained with this method are also listed in table 1 , as represented by 1 in column 10 . eleven galaxies of our sample have neither optical spectra nor nir photometry . for these galaxies we use @xmath216-band data and the @xmath238-color relation in bell & de jong ( 2001 ) to calculate their stellar masses . the @xmath216-band data might be more affected by recent star formation and extinction than nir data . to check to what extent , we compare the @xmath216-band - based stellar masses with spectrum - based ones for 21 galaxies . we find that , in general , @xmath216-band - based stellar masses are 0.16 dex ( with a dispersion of 0.27 dex ) larger than spectrum - based ones . hence , a correction of -0.16 dex is applied to the @xmath216-band - based stellar masses , as represented by 4 in column 10 of table 1 . as shown above , there exist some offsets between the stellar masses derived with different methods . the most possible reason causing this offset is the star formation histories ( sfhs ) , as discussed in gallazzi & bell ( 2009 ) . for galaxies with busty ( or continuous ) sfhs , such as bcds , the one color - based mass - to - light ratio is generally overestimated by 0.1 - 0.2 dex ( gallazzi & bell 2009 ) , which agrees well with our results . the dust might also play a role in this discrepancy since we do not correct for the internal extinction for the optical data . however , dust extinction decreases the luminosity and reddens the optical color . to first - order , these effects cancel out , and the estimated stellar mass of a galaxy is almost unaffected even if a moderate internal extinction of @xmath239 mag is assumed ( using the cardelli et al . ( 1989 ) extinction law ) . the uncertainty in the derived stellar mass depends majorly on the uncertainties of models , galaxy evolution prescription , and sfhs , which are all of order of 0.1 - 0.2 dex or less ( bell & de jong 2001 ) . therefore , we adopt a statistical uncertainty of 0.3 dex for our results . the stellar masses for the di sample are derived with a two - component ( young " plus old " ) method ( lee et al . 2003 ) . to check whether this method is consistent with ours , we calculated the @xmath240 using the @xmath241 color and @xmath242-color relation given by bell & de jong ( 2001 ) for the di sample , and then compared them with those derived with the two - component method . we found that the one - color based @xmath242 is systematically larger by 0.14 dex , with a scatter of 0.11 dex , than that derived with the two - component method . this is generally in accordance with the result described above . therefore , we conclude that the methods used to determine the stellar masses for the di and bcd sample are consistent . the total mass of atomic hydrogen , @xmath243 , in solar masses , can be derived from the hi 21 cm observations by using the following equation ( roberts 1975 ; roberts & haynes 1994 ) : @xmath244 where @xmath245 is the 21 cm flux integral in jy km s@xmath218 and @xmath246 is the distance in mpc . to account for helium and other metals , the total gas mass in solar masses is given by @xmath247 where x is the fraction of the gas mass in the form of hydrogen , here adopted to be 0.733 ( lee et al . 2003 ) in consideration of the similar range of metallicity and the consistency between the di and bcd samples . in table 1 we include the logarithm of the total gas mass for our bcd sample , calculated from @xmath245 ( collected from literature ; see references in table 1 ) and the distance modulus included in the second column . the gas fraction , which is a natural parameter for quantifying to what degree a galaxy has exhausted its available fuel supply , is defined as @xmath248 where the total mass in baryons , @xmath249 , is taken to be the mass of gas and stars . the gas fraction is a distance - independent quantity , because the gas mass and stellar mass are both derived from electromagnetic fluxes . derived gas fractions for the sample of bcds are summarized in table 1 . for the bcds in our sample , about 75% have their @xmath250 , and the median value of @xmath251 is 0.57 . the total gas mass and stellar mass in our bcd sample both range from @xmath252 to @xmath253 @xmath254 . figure 2 shows the distribution of total gas mass as a function of stellar mass for e - type and i - type bcds . the solid line is a reference line for the case when the two qualities are the same . we find that , almost all i - type bcds lie above the @xmath255 line , whereas more than half of the e - type bcds just below this line , indicating that i - type bcds tend to have larger gas masses than e - type bcds at a given stellar mass . for i - type bcds , except for the apparent merging system , mrk 108 , whose gas content is extraordinarily high , the gas mass well correlates with the stellar mass , namely galaxies with a larger stellar mass also contains higher gas masses . the e - type bcds seem to have a similar @xmath256 relation when @xmath257 @xmath254 . while at the high - mass end ( @xmath258 ) , galaxies tend to have lower gas contents comparing to the lower mass counterparts . as a comparison , in figure 2 we also plot the data for field dis which are adopted from lee el al . ( 2003 ) . in this plot , field dis seem to distribute similarly to bcds as a whole , but the @xmath256 relation of field dis has a smaller dispersion than that of bcds . ratio , and the cross in the bottom right gives the typical error bar of the bcd sample.,scaledwidth=48.0% ] since we are interested in the potential for connections / differences between different types of dwarfs , it is informative to consider the gas fraction of a galaxy . in figure 3 we plot the number distribution of gas fraction , @xmath259 , in e- and i - type bcds , and field dis . as mentioned above , i - type bcds are generally more gas - rich than e - type bcds . the separation between these types of bcds is more striking in figure 3 . the @xmath260 for e - type bcds shows a wide distribution and ranges from @xmath261 to @xmath262 , and has a median value of @xmath263 . however , the distribution of @xmath260 for i - type bcds is much narrower , and looks similar to that for field dis . i - type bcds concentrate between @xmath264 ( 14 out of 17 ) , while relatively more field dis have @xmath265 . the median values of @xmath260 are @xmath266 and @xmath267 , for i - type bcds and field dis , respectively . to quantify the significance of the difference or similarity of the gas content among these different types of galaxies , we perform a nonparametric statistical test , the kolmogorov - smirnov ( k - s ) test . the k - s test for the i - type sample against the e - type sample rejects the null hypothesis ( that these two data sets can be derived from a same parent population ) at a @xmath268 significance level . on the other hand , the k - s tests give probabilities of @xmath269 , @xmath270 and @xmath271 , for the dis vs i- and e - type bcds , all bcds as a whole , data sets , respectively . therefore , we conclude that the gas fraction distributions of i- and e - type bcds are much different , whereas the gas fraction distributions of dis and bcds seem to be indistinguishable . nevertheless , we caution that such analysis is possibly limited by the small numbers of the samples . we show the gas fraction of different types of galaxies plotted against their stellar mass in figure 4 . at the low - mass end ( @xmath272 ) , in general , the gas mass is roughly independent on the stellar mass ( i.e. proportional to the baryonic mass ) . at higher masses , however , the relation exhibits a break and becomes considerably steeper since the gas has been more efficiently converted to stars in higher stellar mass galaxies . from figure 4 we can see that most bcds and field dis distributes similarly in the @xmath259-@xmath231 plot . however , at a given stellar mass , the gas fractions for e - type bcds have a larger span and a lower mean value than those for i - type bcds and field dis , which confirms that e - type bcds are generally more gas - poor . star - forming dwarf galaxies are observed to be well mixed chemically throughout their high surface brightness components ( kobulnicky & skillman 1996 , 1997 ) , having no significant gradients or inhomogeneities in oxygen abundances measured from hii regions located at different galactocentric radii ( e.g. lee & skillman 2004 ) . in figure 5 we show the metallicity distributions for e- and i - type bcds and dis . it seems that all of these different types of dwarfs show a double - peaked distribution of metallicity . however , e - type bcds have a larger median oxygen abundance ( @xmath273 ) than i - type bcds and dis ( both @xmath274 ) . in the following we explore the chemical evolution for bcds by investigating in detail the correlations between their oxygen abundance and gas mass , baryonic mass , and gas fraction . ) as a function of stellar mass for bcds and dis . the cross in the bottom right gives the typical error bar of the bcd sample . more massive systems have lower gas fractions.,scaledwidth=48.0% ] figure 6 shows the metallicity as a function of the total gas mass . albeit with large scatters , a galaxy with a greater gas mass is generally more abundant in metals . if we look into different types of galaxies , the scatter will reduce much . a unweighted least - squares linear fit , using a geometrical mean functional relationship ( isobe et al . 1990 ) , to the two bcd subsamples , gives , @xmath275 @xmath276 for e- and i - type bcds respectively . the fitted trends are shown as ( purple ) solid and ( blue ) dashed lines in figure 6 , and the rms deviations of the data from the relationships are 0.35 and 0.34 dex in ( o / h ) , for e- and i - type bcds , respectively . while for the di sample , we obtain the following @xmath277 relation @xmath278 with a rms scatter of 0.26 dex in ( o / h ) , and is shown with the ( red ) dot - dashed line in figure 6 . although the slopes of the @xmath279 relations for i- and e - type bcds are almost the same , and there are larger scatters in these relations , it is clear that e - type bcds have higher mean oxygen abundance than i - type ones at a given gas mass , suggesting that they are depleted in gas relative to i - type bcds of comparable metallicity . field dis have a shallower @xmath279 relation than bcds , but are mixed with bcds in the plot . combining this difference with the fact that i - type bcds are more gas - rich than e - type bcds ( see section 3.2 ) , it might indicate a possible difference in potential for the progenitors and/or evolutionary phases of e- and i - type bcds . figure 7 plots the oxygen abundance as a function of the total baryonic mass . for e - type bcds , unlike figure 6 , this figure shows a much better correlation between metallicity and baryonic mass . a geometric linear fit to the e - type sample ( 36 members ) gives the following relation @xmath280 with a rms scatter of 0.25 dex in ( o / h ) . the fitted relation is plotted with a solid line . , with the cross showing the typical error bar of the bcd sample . , scaledwidth=48.0% ] we also do the same fits to the i - type bcd and di samples , which give @xmath281 and @xmath282 with rms scatters of 0.31 dex and 0.22 dex in ( o / h ) , for the i - type bcds and dis , respectively . the dashed and dot - dashed lines superposed in figure 7 are the respective fitting results for i - type bcds and dis . these results show that the baryonic mass - metallicity relations for i- and e - type bcds have similar slopes , which are a bit steeper than that of field dis . however , e - type bcds appear to lie at a higher oxygen abundance comparing with i - type ones at a given baryonic mass , which again seems to suggest that e- and i - type bcds might be at different evolutionary phases and/or have different progenitors . in fact , bcds also have steeper _ l - z _ and mass - metallicity ( _ m - z _ ) relations than dis ( zgg10 ; zahid et al . kunth & stlin ( 2000 ) show that there exist offsets between the @xmath2 band _ l - z _ relations of bcds , dis and des / dsphs , in the sense that at a given @xmath2 luminosity it becomes relatively more metal - deficient following the des / dsphs , dis and bcds sequence . this offset persists even only comparing the metallicities of the old populations in dis and des / dsphs ( grebel et al . woo et al . ( 2008 ) find the same phenomenon in the _ m - z _ relations of dis and des / dsphs . in the current work , however , we find that dis distribute in between the two types of bcds in the baryonic - metallicity relation . therefore , the evolutionary connections , if exist , between des , dis , and bcds may be complicated . the relation between metallicity and gas fraction can be used to check different chemical evolution models . the closed - box model predicts a linear correlation between the logarithm of the inverse gas fraction @xmath283 and the oxygen abundance by number , @xmath284 , namely ( e.g. lee et al . 2003 ) @xmath285 where @xmath286 is the yield of oxygen and we have assumed that the fraction of gas in the form of hydrogen is @xmath287 ( solar value ) . a comparison of the observed oxygen abundances to those predicted by closed - box models ( thick solid line ) is shown in figure 8 , which indicates that few of the galaxies in these bcd and di samples agree with a closed - box chemical evolution . the vast majority of galaxies appear to have oxygen abundances less than that expected by closed - box models ( with a salpeter imf ) , which imply varying imf , and/or gas inflows / outflows in these galaxies . it is also interesting to find that i - type bcds and dis occupy a similar region in this plot , whereas e - type bcds extend to the most metal - rich and gas - poor region . this indicates that ( 1 ) these galaxies follow a similar chemical evolution path ( e.g. experiencing similar gas flows ) , but at different phase ( e.g. e - type bcds are more evolved than dis and i - type bcds ) ; or ( 2 ) e - type bcds experience different gas flows from dis and i - type bcds ( see below ) . the effects of inflows and/or outflows have been described by many authors ( e.g. , edmunds 1990 ; dalcanton 2007 ; erb 2008 ; matteucci 2008 ; spitoni et al . in particular , dalcanton ( 2007 ) shows that inflows are very efficient in reducing the effective yield , @xmath288 , for gas - poor galaxies while outflows are more effective in changing @xmath288 in gas - rich galaxies . erb ( 2008 ) and matteucci ( 2008 ) introduce a model in which galaxies have inflows / outflows proportional to the star formation rate ( sfr ) . this model adopts the instance recycling and mixing approximations , and the infalling gas is assumed to have no metal , while the enriched outflows is considered to have the same metallicity as the interstellar medium ( ism ) . in this model , we can change two free parameters , the amount of outflow @xmath289 and inflow @xmath290 in unit of the sfr of the galaxy , to reproduce the observed oxygen abundance . two more parameters can be varied , if necessary , i.e. , the true yield @xmath291 ( i.e. imf ) and the fraction @xmath292 of mass still locked in stars . in figure 8 we also plot the predicted metallicity - gas fraction relations from this model using different parameters . we adopt @xmath286 as 1 and @xmath293 of the true oxygen yield predicted by the closed - box model with a salpeter imf(@xmath294 ; meynet & maeder 2002 ) , and vary @xmath289 and @xmath290 to reproduce the dependence of oxygen abundance on the inverse gas fraction . the parameter @xmath292 is a slowly decreasing function of time , as more stars leave the main sequence , but can vary in quite a small range . it is expected to be @xmath295 at @xmath296 yr and @xmath297 at @xmath298 yr ( bruzual & charlot 2003 ) for sub - solar metallicity galaxies with a salpeter imf . given the age of these galaxies ( see , e.g. zgg11 ) , we have used @xmath299 , in despite of that the results are not very sensitive to the chosen value . for a salpeter imf , as shown in figure 8 , the pure inflow model ( black dotted lines ) with a variable rate can not explain the observed metallicity - gas fraction relation , ( especially for e - type bcds ) , which is consistent with the finding in spitoni et al . small inflow rate predicts a higher metallicity than the observed one , whereas high inflow rate can not reach the gas - poor region . pure outflow model ( thin solid lines ) or the combination of outflow and inflow models ( dashed and dot - dashed lines ) can reproduce the observed distribution , but the pure outflow model may need very high flow rate . the best solution could be a combination of gas flows and a variable imf . ; meynet & maeder 2002 ) . the thin solid ( dashed ) lines are pure outflow ( inflow ) models , with an effective yield 1 ( black ) and @xmath293 ( blue ) of the true yield . the rest ( noncontinuous ) lines show the results of applying the model in erb ( 2008 ) with values of @xmath289 , @xmath290 and @xmath286 in the labels . the labeled galaxy , mrk 108 , which is an apparent merging system , appears to be interpreted by none of these models.,scaledwidth=48.0% ] in this section we will mainly discuss the possible evolutionary connections between bcds , des and dis . there have been several evolutionary scenarios , both from observational and theoretical aspects , between bcds , des , and dis ( see papaderos et al . 1996 for a review ) . however , the question of whether there is an evolutionary connection between these three types of dwarf galaxies have not been achieved a solid answer yet . as mentioned in section 1 , some apparently contrary results are present in previous studies on the basis of observational measurements . for the bcds in our sample , metallicity correlates with gas mass and total baryonic mass , in the sense that more massive systems are more abundant in metals . for these two correlations , i - type bcds show similar slopes to e - type ones . however , i - type bcds are generally more metal - poor and gas - rich than e - type ones at a given gas / baryonic mass . although field dis seem to have a shallower slope relative to bcds , they are well mixed with bcds in the distributions of @xmath279 and @xmath300 plots . moreover , in the plot of metallicity - gas fraction relation , i - type bcds generally occupy a region differing from e - type ones . our results may shed some light on the evolutionary connection between the three types of dwarf galaxies . i- and e - type bcds might be at different evolutionary phases and/or have different progenitors . for example , if some e - type bcds are formed from des ( e.g. silk et al . 1987 ) , whereas i - type ones are formed from dis , then this part of e - type bcds should be more evolved than i - type ones . this scenario is also supported by the fact that many e - type bcds show structural properties compatible with des ( papaderos et al . 1996 ; gil de paz & madore 2005 ) . in addition , field dis show chemical and gaseous properties in between i- and e - type bcds , which suggests that they could be connected to both types of bcds ( e.g. thuan 1985 ; davies & phillipps 1988 ) . as discussed in papaderos et al . ( 1996 ) , the interchanges between dis , bcds and des need gas outflows / inflows to get involved , which is very consistent with our result that most bcds should have experienced gas flows during their lives . indeed , observations have provided ( direct ) evidence for gas flows in bcds ( e.g. , papaderos et al . 1994 ; martin 1996 , 1999 ; martin et al . 2001 ; elson et al . 2009 ; van eymeren et al . 2009 ; lpez - schez et al . 2010 ) . using x - ray data , papaderos et al . ( 1994 ) estimate a galactic wind flowing at a velocity of @xmath301 km s@xmath218 for the bcd vii zw 403 , whereas the estimated escape velocity for this galaxy is only 50 km s@xmath218 . martin et al . ( 2001 ) have shown that the galactic wind in ngc 1569 carries away almost all the oxygen produced in the current starburst . more recently , elson et al . ( 2009 ) present their hi observations for ngc 2915 , and find a gas component clearly lagging in velocity relative to the main gas disk of the galaxy . this might be evidence for gas accretion from inter - galactic space onto the outer galaxy disk . van eymeren et al . ( 2009 ) report their detailed kinematic studies on the neutral ( hi ) and ionized ( h@xmath292 ) gas components in a nearby bcd , ngc 4861 . the authors find that both in hi and h@xmath292 the galaxy shows several outflows . using multi - wavelength data , another work by this group ( lpez - schez et al . 2010 ) have studied the environment of a sample of nearby bcds , and shown that all bcds present evident interacting features in their hi component despite the environment in which they reside . in particular , ic 4870 , which has an elliptical outer component , is accreting hi clouds . we have presented a detailed studies on the gaseous and chemical properties of a sample of 53 bcds , which is selected from g03/zgg10 sample . according to their morphological information , this sample of bcds is divided into two subsamples : e- and i - type bcd samples . an e - type bcd contains an elliptical outer component while an i - type bcd has an irregular outer halo . for both types of bcds , various correlations among metallicity , gaseous , stellar and baryonic mass , and gas fraction , are investigated . we find that in general , ( 1 ) more massive systems are more metal - rich , ( 2 ) a galaxy having a larger stellar mass also contains more gas , but a smaller gas fraction , ( 3 ) i- and e - type bcds have similar slopes of the @xmath279 and @xmath300 relations , but i - type bcds are generally more gas - rich and metal - poor . based on the comparisons of these correlations , we suggest that i- and ( at least a part of ) e - type bcds might be likely at different evolutionary stages and/or having progenitors . by comparing the metallicity - gas fraction relation for our sample with those predicted by chemical evolution models , we conclude that most bcds should have experienced some gas inflows / outflows and/or mergers . the authors gratefully acknowledge the anonymous referee for his / her constructive comments and suggestions that improved this paper . research for this project is supported by the national natural science foundation of china under grants 10833006 and 10903029 . this research has made use of the nasa / ipac extragalactic database ( ned ) , which is operated by the jet propulsion laboratory , california institute of technology , under contract with the national aeronautics and space administration . allende prieto , c. , lambert , d. l. , & asplund , m. 2001 , , 556 , l63 amorn , r. , alfonso , j. , aguerri , j. a. l. , muoz - tun , c. , & cairs , l. m. 2009 , a&a , 501 , 75 asplund , m. , grevesse , n. , sauval , a. j. , & scott , p. 2009 , , 47 , 481 bell , e. f. , & de jong , r. f. 2001 , , 550 , 212 binggeli , b. 1994 , in dwarf galaxies , ed . g. meylan & p. prugniel ( garching : eso ) , 13 bruzual , g. , & charlot , s. 2003 , , 344 , 1000 calzetti , d. 2001 , pasp , 113 , 1449 cannon , j. m. , skillman , e. d. , garnett , d. r. , & dufour , r. j. 2002 , , 565 , 931 cardelli , j. a. , clayton , g. c. , & mathis , j. s. 1989 , , 345 , 245 cairs , l. m. , caon , n. , vlchez , j. m. , gonzlez - prez , j. n. , & muoz - tun , c. 2001a , apjs , 136 , 393 cairs , l. m. , vlchez , j. m. , gonzlez - prez , j. n. , iglesias - pramo , j. , & caon , n. 2001b , apjs , 133 , 321 cairs , l. m. , caon , n. , zurita , c. , kehrig , c. , roth , m. , & weilbacher , p. 2010 , , 520 , 90 cid fernandes , r. , asari , n. v. , sodr , l. , stasiska , g. , mateus , a. , torres - papaqui , j. p. , & schoenell , w. 2007 , mnras , 375 , 16 cid fernandes , r. , mateus , a. , sodr , jr , l. , stasiska , g. , & gomes , j. m. 2005 , mnras , 358 , 363 dalcanton , j. j. 2007 , , 658 , 941 davies , j. i. , & phillipps , s. 1988 , , 233 , 553 denicol , g. , terlevich , r. , & terlevich , e. 2002 , , 330 , 69 edmunds , m. g. 1990 , , 246 , 678 elson , e. , de blok , e. , kraan - korteweg , r. c. 2009 in proceedings of panoramic radio astronomy : wide - field 1 - 2 ghz research on galaxy evolution , ed . g. heald & p. serra , 51 erb , d. k. 2008 , , 674 , 151 ferguson , h. c. , & binggeli , b. 1994 , a&a rev . , 6 , 67 freedman , w. l. , et al . 2001 , , 553 , 47 gallazzi , a. , & bell , e. f. 2009 , , 185 , 253 garca - lorenzo , b. , cairs , l. m. , caon , n. , monreal - ibero , a. , & kehrig , c. 2008 , , 677 , 201 gavazzi , g. , boselli , a. , van driel , w. , & k. oneil 2005 , a&a , 429 , 439 gil de paz , a. , madore , b. f. , & pevunova , o. 2003 , , 147 , 29 ( g03 ) gil de paz , a. , & madore , b. f. 2005 , , 156 , 345 gordon , d. , & gottesman , s. t. 1981 , , 86 , 161 grebel , e. k. , gallagher , iii , j. s. , & harbeck , d. 2003 , , 125 , 1926 hoffman , g. l. , salpeter , e. e. , farhat , b. , roos , t. , williams , h. , & helou , g. 1996 , , 105 , 269 huchtmeier , w. k. , krishna , g. , & petrosian , a. 2005 , a&a , 434 , 887 huchtmeier , w. k. , & richter , o .- g . 1988 , a&a , 203 , 237 hunter , d. a. 1997 , pasp , 109 , 937 hunter , d. a. , & hoffman , l. 1999 , , 117 , 2789 isobe , t. , feigelson , e. d. , akritas , m. g. , & babu , g. j. 1990 , , 364 , 104 izotov , y. i. , & thuan , t. x. 1999 , , 511 , 639 izotov , y. i. , thuan , t. x. , & lipovetsky , v. a. 1994 , , 435 , 647 james , p. a. 1994 , , 269 , 176 kauffmann , g. , white , s. d. m. , & guiderdoni , b. 1993 , mnras , 264 , 201 kobulnicky , h. a. , & skillman , e. d. 1996 , , 471 , 211 kobulnicky , h. a. , & skillman , e. d. 1997 , , 4789 , 636 kunth , d. , & stlin , g. 2000 , a&arv , 10 , 1 le borgne , j .- f . , 2003 , a&a , 402 , 433 lee , h. , mccall , m. l. , kingsburgh , r. l. , ross , r. , & stevenson , c. c. 2003 , , 125 , 146 lee , h. , & skillman , e. d. 2004 , , 614 , 698 loose , h .- h . , & thuan , t. x. 1986 , in star forming dwarf galaxies and related objects , ed . d. kunth , t. x. thuan , & j. t. t. van ( gif - sur - yvette : editions frontires ) , 73 lpez - snchez , a. r. , koribalski , b. , van eymeren , j. , esteban , c. , popping , a. , & hibbard , j. 2010 , in galaxies in isolation : exploring nature versus nurture , ed . l. verdes - montenegro , a. del olmo , & jack sulentic ( san francisco : astronomical society of the pacific ) , 65 martin , c. l. 1996 , , 465 , 680 martin , c. l. 1999 , , 513 , 156 martin , c. l. , kobulnicky , h. a. , & heckman , t. m. 2002 , , 574 , 663 mateo , m. 1998 , , 36 , 435 matteucci , f. 2008 , arxiv : 0804.1492 mccall , m. l. , vaduvescu , o. , pozo nunez , f. , barr dominguez , a. , fingerhut , r. , unda - sanzana , e. , li , b. , & albrecht , m. 2012 , a&a , 540 , 49 meynet , g. , & maeder , a. 2002 . a&a , 390 , 561 moustakas , j. , & kennicutt , r. c. , jr . 2006 , apjs , 164 , 81 noeske , k. g. , guseva , n. g. , fricke , k. j. , izotov , y. i. , papaderos , p. , & thuan , t. x. 2000 , a&a , 361 , 33 noeske , k. g. , papaderos , p. , cairs , l. m. , & fricke , k. j. 2003 , a&a , 410 , 481 papaderos , p. , fricke , k. j. , thuan , t. x. , & loose , h .- h . 1994 , a&a , 291 , l13 papaderos , p. , loose , h .- h . , fricke , k. j. , & thuan , t. x. 1996 , a&a , 314 , 59 pettini , m. , & pagel , b. e. j. 2004 , , 348 , l59 richer , m. g. , & mccall , m. l. 1995 , , 445 , 642 roberts , m. s. 1975 , in galaxies and the universe , ed . a. sandage , m. sandage , & j. kristian ( chicago : univ . chicago press ) , 309 roberts , m. s. , & haynes , m. p. 1994 , , 32 , 115 salpeter , e. e. 1955 , , 121 , 161 salzer , j. j. , lee , j. c. , melbourne , j. , hinz , j. l. , alonso - herrero , a. , & jangren , a. 2005 , , 624 , 661 salzer , j. j. , macalpine , g. m. , & boroson , t. a. 1989 , 70 , 479 schlegel , d. j. , finkbeiner , d. p. , & davis , m. 1998 , , 500 , 525 schulte - ladbeck , r. e. , hopp , u. , greggio , l. , & crone , m. m. 2000 , , 120 , 1713 silk , j. , wyse , r. f. g. , & shields , g. a. 1987 , , 322 , l59 spitoni , e. , calura , f. , matteucci , f. , & recchi , s 2010 , a&a , 514 , 73 telles , e. , melnick , j. , & terlevich , r. 1997 , , 288 , 78 terlevich , r. , melnick , j. , masegosa , j. , moles , m. , & copetti , m. v. f. 1991 , a&as , 91 , 285 thuan , t. x. 1985 , , 299 , 881 thuan , t. x. , lipovetsky , v. a. , martin , j .- m . , & pustilnik , s. a. 1999 , a&as , 139 , 1 thuan , t. x. , & martin , g. e. 1981 , , 247 , 823 tolstoy , e. , hill , v. , & tosi , m. 2009 , , 47 , 371 vaduvescu , o. , & mccall , m. l. 2008 , a&a , 487 , 147 vaduvescu , o. , mccall , m. l. , & richer , m. g. , 2007 , , 134 , 604 ( vmr07 ) vaduvescu , o. , richer , m. g. , & mccall , m. l. 2006 , , 131 , 1318 van eymeren , j. , marcelin , m. , koribalski , b. s. , dettmar , r .- j . , bomans , d. j. , gach , j .- l . , & balard , p. 2009 , a&a , 505 , 105 walcher , c. j. , et al . 2008 , a&a , 491 , 713 woo , j. , courteau , s. , & dekel , a. 2008 , , 390 , 1453 zahid , h. j. , bresolin , f. , kewley , l. j. , coil , a. l. , & dav , r. 2012 , , 750 , 120 zhao , y. , gao , y. , & gu , q. 2010 , , 710 , 663 ( zgg10 ) zhao , y. , gu , q. , & gao , y. 2011 , , 141 , 68 ( zgg11 ) zwicky , f. 1966 , , 143 , 192

detailed anisotropic ( hab and hc ) resistivity and specific heat measurements were performed on online - grown ybptin and solution - grown ybpt in single crystals for temperatures down to 0.4 k , and fields up to 140 kg ; hab hall resistivity was also measured on the ybpt in system for the same temperature and field ranges . all these measurements indicate that the small change in stoichiometry between the two compounds drastically affects their ordering temperatures ( t k in ybptin , and k in ybpt in ) . furthermore , a field - induced quantum critical point is apparent in each of these heavy fermion systems , with the corresponding critical field values of ybpt in ( h around 35 - 45 kg and h kg ) also reduced compared to the analogous values for ybptin ( h kg and h kg ) .

cond-mat0506425

in recent years , stoichiometric yb - based heavy fermion compounds have raised a lot of interest , particularly due to the limited number of such systems known to date . ybptin is one of the few examples of such heavy fermion systems@xcite , as indicated by measurements performed on single crystal samples extracted from on - line melts of the polycrystalline material . based on the existing data@xcite , ybptin has a relatively low magnetic ordering temperature ( t@xmath8 k ) and an enhanced electronic specific heat coefficient ( @xmath9 mj@xmath10mol k@xmath11 ) , whereas the magnetic entropy at t@xmath12 amounts to only about 60@xmath13 of the @xmath14 value expected for a doublet ground state . in light of these observations , this system appears to be qualitatively similar to the other recently studied yb - based heavy fermion antiferromagnets , ybrh@xmath15si@xmath15@xcite , and ybagge@xcite , the latter compound being isostructural to ybptin . in ybrh@xmath15si@xmath15 , magnetic order occurring at very low temperature ( below 70 mk)@xcite was associated with a low entropy release ( around 0.01 @xmath16)@xcite . for the ybagge@xcite compound , the temperature associated with the magnetically ordered state , whereas still fairly low ( @xmath17 k ) , is enhanced compared to that of ybrh@xmath15si@xmath15 ; also , the magnetic entropy at the ordering temperature is larger than in ybrh@xmath15si@xmath15 , but still less than 10@xmath13 of @xmath14 . thus these two compounds can be regarded as systems with small moment ordering . having an even higher ordering temperature and magnetic entropy at t@xmath12 , ybptin seemed a good candidate to further study a progression from small moment to reduced moment ordering , in stoichiometric yb - based heavy fermion compounds , with field - induced quantum critical point . in contrast to a classical phase transition at finite temperatures , driven by temperature as a control parameter with thermal fluctuations , a quantum phase transition is driven by a control parameter c other than temperature ( _ e.g. _ , c = pressure , doping or magnetic field ) at t = 0 , with quantum mechanical fluctuations . such a control parameter tunes the system from a magnetically ordered state towards a state without magnetic order , at zero temperature , crossing a quantum critical point . due to the hybridization of the 4f electrons and the conduction electrons in heavy fermion ( hf ) systems , which can be modified by any one of the aforementioned c control parameters , the ( hf ) compounds are very suitable to study quantum critical behavior . moreover , close to the critical value c@xmath18 which drives the ordering temperature close to zero , pronounced deviations from the fermi liquid - like fl behavior can occur . this has been observed in a large number of hf systems where c = doping or pressure , and only a few doped systems have been field - tuned through a qcp@xcite . to date , ybrh@xmath15si@xmath15@xcite and ybagge@xcite are the only stoichiometric yb - based hf compounds in which a field - induced quantum critical point qcp has been observed . the heavy fermion character of the ybptin system has already been reported@xcite ; in this paper we will discuss the existence of a field - induced qcp in ybptin which renders it very similar to the other two stoichiometric , yb - based heavy fermions . recently we presented anisotropic low - field susceptibility measurements , as well as specific heat data in zero applied field , on solution - grown single crystals of the rptin series , including ybptin@xcite . no features indicative of magnetic order could be identified in our magnetization measurements on single crystals of the r = yb compound , down to t @xmath19 2 k. however , a well - defined peak at @xmath20 k in the zero - field specific heat data suggests that this compound orders magnetically below 2.1 k , just above the low temperature limit of our magnetization measurements . as we will present in this paper , detailed transport and thermodynamic measurements down to 0.4 k not only confirm the magnetic ordering in this compound below 2.1 k , but also suggest that another phase transition might exist in this system around 1 k. however , our measurements were in part inconsistent with previous reports on ybptin single crystals : whereas kaczorowski _ _ et al.__@xcite have also presented magnetization data showing no features associated with magnetic ordering above 1.7 k , in their specific heat measurements three anomalies can be observed ( at 3.1 , 2.3 , and 1.2 k ) , all at slightly different temperatures than in our data . furthermore , the low - field susceptibility data presented by trovarelli _ _ et al.__@xcite are consistent with an ordering temperature around 3.4 k , also confirmed by their resistivity and specific heat measurements . in order to address these apparent discrepancies between our data on solution grown single crystals and the two previous reports on on - line grown single crystals@xcite , we reproduced the growth as described by trovarelli _ _ et al.__@xcite . the anisotropic specific heat and transport measurements on our single crystals extracted from the melt confirmed the existence of magnetic phase transitions at 3.4 k and 1.4 k , as observed previously . in this paper we will try to examine the differences between solution and on - line grown single crystals , given that a small pt - deficiency occurred in the former types of crystals , leading to a stoichiometry closest to ybpt@xmath1 in , and no disorder could be detected in the on - line grown crystals . also , given that heavy fermion compounds with small moment ordering are likely to be driven to a quantum critical point qcp by disorder or applied field , we will study the evolution of both ybpt@xmath1 in and ybptin towards a field - induced qcp . detailed anisotropic measurements of specific heat and resistivity on the two types of crystals , for fields up to 140 kg and temperatures down to 0.4 k , will be used for the comparison between the two types of crystals . additional hall effect measurements were performed on the solution grown single crystals for the same temperature and field ranges , allowing us to further explore the effects of the qcp in this compound . after briefly describing the methods used for synthesizing the two types of ybptin single crystals , and the measurement techniques used for the sample characterization , we will present the results of single crystal x - ray diffraction . next we will compare the low field data for the two types of compounds , pointing out the similarities , as well as the significant differences in their physical properties . next we will present the higher field data for both types of crystals , for @xmath21 and @xmath22 . this will allow us to follow , in both systems , the progression from reduced magnetic moment order to the quantum critical point qcp , as driven by the application of increasing magnetic field . also , we will extend the zero - field comparison between the two ybptin systems to their field- dependent properties , and will try to identify common features as well as possible effects of the site deficiency . ybptin is one of the rptin compounds , reported to crystallize in zrnial hexagonal structure@xcite , space group p@xmath23 . single crystals of this compound were obtained by two methods : from high - temperature ternary solution , as described in ref . [ 15 ] , and from stoichiometric on - line melt@xcite . the initial concentration used for the solution growth was yb@xmath24pt@xmath25in@xmath26 , and the ternary solution , sealed in a 3-cap ta crucible@xcite under partial argon pressure , was slow - cooled from 1200@xmath27 c to 1000@xmath27 c over approximately 100 hours . subsequently , the excess liquid solution was decanted , and the resulting hexagonal rods were briefly etched in hcl to remove residual flux from the surface . for the on - line melt , 1:1:1 atomic ratios of yb , pt and in were sealed in a ta crucible under partial argon atmosphere . after briefly heating the sealed crucible up to approximately 1650@xmath27 c in an induction furnace , the crucible was sealed in a quartz ampoulle and annealed at 700@xmath27 c for 120 hours . small rods were subsequently extracted from the resulting crystalline conglomerate . anisotropic magnetization measurements as a function of temperature and applied field m(h , t ) were performed in a quantum design mpms squid magnetometer ( t @xmath28 k , h@xmath29 kg ) . temperature- and field - dependent specific heat c@xmath30(h , t ) , resistivity @xmath31(h , t ) and hall resistivity @xmath32(h , t ) measurements were taken in a quantum design ppms-14 instrument with he-3 option , for temperatures down to 0.4 k , and applied magnetic fields between 0 and 140 kg . for the anisotropic ( _ i.e. , _ @xmath33 or @xmath34 ) specific heat measurements a relaxation technique with fitting of the whole calorimeter ( sample with sample platform and puck ) was used . the sample platform and grease background , measured separately for all necessary ( h , t ) values , was subtracted from the sample response . also , the non - magnetic contribution to the total specific heat was estimated based on measurements on flux - grown luptin , used as the non - magnetic analogue of ybptin . the single crystals of luptin were grown using similar initial composition ( r@xmath35pt@xmath35in@xmath36 ) and temperature profile ( slow cooling from 1190@xmath27c to 800@xmath27c over @xmath19100 hours ) that yielded fully - occupied rptin single crystals for the lighter members of the series@xcite . we therefore anticipate that the luptin crystals are also fully - occupied , and we used this system as the non - magnetic analogue for our both yb compounds . it should be noted that we made the assumption that the non - magnetic background is insensitive to the 2@xmath13 pt - deficiency in the solution - grown compound . the specific heat for luptin was measured in the same temperature range as for ybptin and ybpt@xmath1 in , for 0 and 140 kg applied field , and was found to be virtually field - independent over the measured temperature range . a standard ac four - probe resistance technique ( @xmath37 hz , @xmath38 ma ) was used for the field- and temperature - dependent resistivity measurements . given the rod - like geometry of the samples , the current was always flowing along the crystallographic c - axis , whereas the field was either parallel ( for longitudinal magnetoresistance ) , or perpendicular ( for transverse magnetoresistance ) to the direction of current flow . the hall resistivity measurements were performed on a sample that was polished down to a plate - like shape . this sample geometry restricted the hall resistivity measurements to the field applied within the hexagonal crystallographic plane ( @xmath39 ) . the hall resistivity @xmath32(h , t ) was measured for field perpendicular to both the current and the hall voltage directions . in order to minimize the inherent ( small ) misalignment of the voltage contacts , these measurements were taken for two opposite directions of the applied field , h and -h , and the odd component , ( @xmath32(h ) - @xmath32(-h))@xmath40 was taken as the hall resistivity . for each of the two systems ( ybptin and ybpt@xmath1 in ) , we will compare the similar data for the the two orientations of the applied field ( _ i.e. _ , for @xmath39 and @xmath41 ) , as well as the resulting t - h phase diagrams . the criteria used for determining the points on all presented phase diagrams will be discussed in more detail in the appropriate sections , but in general they were maxima in the c@xmath30(h , t ) data , and derivative maxima or onset values for the transport data . for @xmath39 , the maximum temperature for which the @xmath42 could be fit to a straight line was the criterium used in determining the upper limit of this fermi liquid - like region . additional hall resistivity measurements were performed for the field applied within the basal plane , for the ybpt@xmath1 in ( flux - grown ) compound . various criteria were used for determining critical points similar to the hall line for the ybrh@xmath15si@xmath15@xcite and ybagge@xcite systems . for both solution and on - line grown crystals , the crystal structure was confirmed by powder x - ray diffraction , with no detectable impurity peaks . however , additional single crystal x - ray measurements were performed on the two types of ybptin compounds . crystals with dimensions @xmath43 and @xmath44 were extracted from the flux and on - line grown samples respectively . room - temperature x - ray diffraction data were collected on a stoe ipdsii image plate diffractometer with mo k@xmath45 radiation , and were recorded by taking @xmath46 scans in @xmath47 in the full reciprocal sphere . the range of 2 @xmath47 extended from @xmath48 to @xmath49 . numerical absorption corrections for both crystals were based on crystal face indexing , followed by a crystal shape optimization . structure solution and refinement were done using the shelxtl program . the crystallographic and structural data are summarized in tables [ t1 ] - [ t2 ] . .atomic coordinates and equivalent isotropic displacement parameters ( @xmath50 ) for the flux - grown ybptin system . u(eq ) is defined as one third of the trace of the orthogonal u@xmath51 tensor . space group _ p @xmath52 2 m _ , a @xmath53 , c @xmath54 , r = 0.0273 , r@xmath55 = 0.0598 . [ cols="^,^,^,^,^,^ " , ] a high temperature factor was inferred for pt(1 ) of the flux - grown crystal during final stages of the refinement , which is usually indicative of possible atomic deficiencies , symmetry reduction or superstructure formation . no superstructure reflections were observed , and symmetry reduction did not resolve the issue . however , relaxing pt(1 ) site occupancy resulted in a statistically significant deficiency ( 0.06(1 ) , table [ t1 ] ) , and led to improvements in the pt(1 ) temperature factor and overall residual r value . we also tested for possible deficiencies on the other atomic sites in this crystal , but refined occupancies for yb ( 1.00(6 ) ) , pt(2 ) ( 1.00(6 ) ) and in ( 1.02(2 ) ) did not suggest presence of atomic deficiencies . thus the composition of the flux - grown crystal can be written as ybpt@xmath1 in , noting that the pt@xmath1 value reflects the pt stoichiometry of the whole unit cell , not just of the pt(1 ) site . in contrast to the flux - grown sample , the pt(1 ) temperature factor of the on - line grown crystal had a reasonable value . relaxing the pt(1 ) occupancy yielded only a slight and statistically insignificant deficiency of 0.014(9 ) ( table [ t2 ] ) . occupancy refinement for other atomic sites showed no deviations from unity . although small deficiencies on the pt(1 ) site could not be excluded , the ybptin formula is a good presentation of the composition of the on - line crystal in terms of sensitivity of our x - ray diffraction experiments . this difference in the stoichiometry of the solution and on - line grown samples is consistent with ybptin having a small width of formation extending towards the pt - deficient side ; given that the initial melt composition is very pt - poor ( _ i.e. _ , yb@xmath24pt@xmath25in@xmath26 ) , it is expected to be sensitive to such a small width of formation . a closer look at the atoms positions given in tables [ t1]-[t2 ] suggests that two compounds might be mirror images of each other ; whereas racemic twinning could not be excluded for either the solution or the on - line grown compounds , no evidence of the existence of both `` left '' and `` right '' structures in each system could be found . the lattice parameters and unit cell volume for the solution - grown , ybpt@xmath1 in crystals were @xmath56 , @xmath57 @xmath58and @xmath59 @xmath60 . the analogous unit cell dimensions on the on - line grown crystals were slightly smaller : @xmath61 , @xmath62 @xmath58and @xmath63 @xmath60 . anisotropic magnetization measurements are presented in fig.[f01 ] for both the ybpt@xmath1 in ( full symbols ) and the ybptin ( open symbols ) compounds . as can be observed in fig.[f01]a , the paramagnetic susceptibility indicates moderate anisotropy for both systems ( with @xmath64 at the lowest temperature ) , and no clear sign of magnetic ordering down to @xmath65 k. the anisotropic m(h ) isotherms show a continuous increase of the magnetization , with a trend towards saturation above 40 kg ( fig.[f01]b ) for h applied within the ab - plane ; the axial magnetization remains linear and significantly smaller than m@xmath66 up to our maximum field available for these measurements ( 55 kg ) . whereas qualitatively there is an overall similarity between the corresponding data of the two compounds , the absolute values of both susceptibility and field - dependent magnetization are slightly larger for ybpt@xmath1 in than for ybptin . we believe the @xmath19 10 to 20@xmath13 difference to be too large to have been caused by weighing errors alone , and thus we conclude that it may reflect the different kondo temperatures and exchange coupling due to the change of stoichiometry in the two compounds . the zero - field specific heat and resistivity data shown in fig.[f02 ] are consistent with magnetic ordering in both compounds , however at different temperatures : two well defined peaks at t@xmath67 k and 1.4 k are visible in the ybptin c@xmath30(t ) data ( open symbols , fig.[f02]a ) , whereas only one peak can be distinguished , around 2 k , in the ybpt@xmath1 in data ( full symbols ) . these transition temperatures are marked by the vertical dotted lines for the former compound , and by one dashed line for the latter . it can be seen that the corresponding resistivity measurements ( fig.[f02]b ) show changes in slope around the respective transition temperatures . another noticeable difference between ybptin and ybpt@xmath1 in manifests in the resistivity values ( fig.[f02]b and [ frt300 ] ) with the ones for the former compound being approximately three times smaller in the latter one . the error bars shown for the lowest temperature @xmath31 values give a caliper of the uncertainty in estimating the resistivity values for the two compounds , further confirming the aforementioned difference . the larger residual resistivity in ybpt@xmath1 in is consistent with the additional disorder ( _ i.e. _ , site disorder ) or presence of additional vacancies in this type of crystals . thus the zero - field measurements indicate a dramatic effect of the small pt - deficiency on the ordered state in ybptin : the upper transition is shifted down in temperature in the pt - deficient compound , whereas a second one is clearly identifiable only in ybptin . in order to explore the differences between these two samples more thoroughly , a systematic study of the field - dependence of @xmath68 and c@xmath69 was undertaken . * @xmath39 * + the low - temperature specific heat data for the on - line grown compound ybptin is shown in fig.[f11]a , for various values of the applied field . as already seen , two sharp peaks present in the h = 0 data can be associated with magnetic phase transitions at t @xmath70 k and 1.4 k. as the applied field is increased , these transitions ( indicated by small arrows in fig.[f11]a ) move to lower temperatures , and eventually drop below 0.4 k around 20 kg , and 60 kg respectively . trovarelli _ _ et al.__@xcite have reported similar measurements up to 80 kg , which are consistent with our data ; however , their study did not include a systematic analysis of the h - t phase diagram or the potential quantum critical behavior in this compound . from the linear extrapolation of the zero - field c@xmath71 _ vs. _ t@xmath11 data from t @xmath19 5 k down to t = 0 ( dotted line in the inset , fig.[f11]a ) , the electronic specific heat coefficient @xmath72 can be roughly estimated as @xmath73 mj / mol k@xmath11 . the magnetic component of the specific heat is defined as @xmath74(ybptin)@xmath75(luptin ) , and is shown in fig.[f11]b as @xmath76 _ vs. _ t@xmath11 , for the same field values as before . when the magnetic specific heat is plotted in c@xmath77 vs. @xmath78 t coordinates ( fig.[f11a ] ) , a reduced region of logarithmic divergency ( non - fermi liquid nfl behavior ) is apparent ; however this linear region in c@xmath79 is more ambiguous than in other heavy fermion compounds displaying nfl behavior ( _ e.g. _ , ybrh@xmath15si@xmath15@xcite and ybagge@xcite ) . because of a downturn in the high field data ( h @xmath80 50 kg [ fig.[f11a]b ] ) around 5 k for h = 50 kg , the largest logarithmic divergency which occurs for h @xmath19 60 kg , is limited to only a fraction of a decade in temperature ( 1.5 k @xmath81 t @xmath81 6.5 k ) . the above observations suggests that a qcp may exist around a critical field value h@xmath4 just above 60 kg , but the presence of a nfl region at intermediate field values is less clearly defined than in the previously studied yb - based heavy fermion compounds . one of the expressions considered in the scaling analysis at a qcp@xcite is the cross - over function @xmath82~/~t$ ] _ vs. _ @xmath83 . in the case of ybrh@xmath15si@xmath15@xcite and ybagge@xcite , the @xmath83 range over which universal scaling was observed in high fields corresponded to 1@xmath10 t @xmath81 3 k@xmath84 , and 1.2 k@xmath84 respectively ( with @xmath85 ) . due to the slightly enhanced magnetic ordering temperature @xmath86 k in ybptin , the analogous @xmath87 t range is drastically reduced ( 1@xmath10 t @xmath81 0.3 k@xmath84 ) , making the unambiguous determination of the critical exponent @xmath88 essentially impossible . low - temperature resistivity curves for different values of the applied magnetic field are shown in fig.[f12 ] . the @xmath89 data are consistent with the presence of two magnetic phase transitions at low fields ; the small arrows indicate these transition temperature values , as determined from maxima in the d@xmath90 . both these transitions are suppressed by increasing applied field . whereas the upper transition can still be detected for h @xmath91 kg , a field of about 20 kg is sufficient to drive the lower one below our base temperature of 0.4 k. it is worth noting that the critical field h@xmath92 kg , determined from the @xmath31(t , h ) data as the field required to suppress the magnetic ordering , is close to the position of the qcp as inferred based on the c@xmath30 data ( fig.[f11]-[f11a ] ) . as fig.[f12sqab ] indicates , when the magnetic field is increased beyond h@xmath4 , the temperature dependence of the resistivity is ambiguous ; particularly at fields just above h@xmath4 ( h @xmath93 100 kg ) , it is difficult to distinguish between linear or quadratic behavior of the resistivity as a function of temperature . consequently we leave the complete analysis of the temperature - dependent resistivity outside the ordered state for the discussion section . when magnetoresistance measurements ( @xmath94 ) are performed ( fig.[f13 ] ) three features are apparent at very low temperatures , as indicated for t = 0.4 k by the small arrows . the inset shows @xmath31(h ) at t = 0.8 k , to exemplify how these transition temperatures were inferred from these data . for t @xmath80 1 k , the two lower transitions merge and the resulting one is still distinguishable up to approximately 1.4 k. the upper transition moves down towards zero field as the temperature approaches t@xmath95 k. based on the above field- and temperature - dependent thermodynamic and transport measurements , a @xmath96 phase diagram for @xmath21 can be constructed ( fig.[f14 ] ) : in zero magnetic field , two magnetic phase transitions can be observed , around t@xmath95 k and t@xmath97 k. increasing magnetic field splits the lower transition into two separate ones around h @xmath98 kg ; one of these phase lines drops towards our lowest temperature at almost constant field , whereas the second one has a slower decrease with field , such that it approaches t = 0.4 k around h = 20 kg . in a similar manner , the upper transition is driven down towards 0 around @xmath99 kg . + * @xmath41 * + fig.[f20 ] presents specific heat data for ybptin , @xmath100 , for fields up to 80 kg . high torques on this sample for this orientation of the field prevented us from completing these measurements up to 140 kg . moreover , as will be shown below , there are significant discrepancies between the transition temperatures determined even from the intermediate - field specific heat data and transport measurements ( _ i.e. _ , for h @xmath80 40 kg ) . this observation prompts us to suspect that significant torques may have changed the sample orientation for the specific heat measurements , even for fields significantly lower than 80 kg . in zero field , we can confirm the two magnetic transitions observed before , at t@xmath95 k , and t@xmath97 k respectively ; as the small arrows indicate , the lower - t transition is driven down in field , and falls below 0.4 k for h @xmath101 kg , whereas the upper transition persists above 80 kg . the temperature and field dependent resistivity data ( fig.[f21]-[f21rh ] ) indicate a much slower suppression of the magnetic order with the applied field . in fig.[f21]a , @xmath68 curves are shown , with the arrows indicating the transition temperatures as determined from @xmath102 . fig.[f21]b presents a subset of these derivatives , to illustrate the criteria for determining the temperatures : for the lower transition , a peak in @xmath102 broadens as field is increased , and disappears for h @xmath103 80 kg ; the upper transition is marked by a step in these derivatives , which also broadens as h increases . at the highest measured field ( _ i.e. _ , 140 kg ) we are unable to distinguish between a very broad step ( with a possible transition temperature marked by the small arrow ) or a cross - over in corresponding @xmath102 . the field - dependent resistivity data ( fig.[f21rh ] ) are indicative of a low temperature transition consistent with that seen in @xmath68 , with the critical fields determined from onsets . given the above c@xmath30(t , h ) and @xmath31(t , h ) data , we suspect that magnetic fields h @xmath103 20 kg deform the four wires supporting the he-3 specific heat platform used for the c@xmath30(t , h ) measurements , whereas the resistivity sample appears to be well held in place by grease on the rigid platform . consequently , at high fields , the two sets of data ( c@xmath30(t , h ) and @xmath31(t , h ) ) may not correspond to the same orientation of the field ( @xmath41 ) , yielding different transition temperature values for the corresponding applied fields . as a result , in constructing the t - h phase diagram for h@xmath104 ( fig.[f22 ] ) , we will consider the t@xmath105 values as determined from the @xmath31(t , h ) data up to h = 140 kg , and only the h @xmath106 20 kg ones based on specific heat measurements . also shown are error bars for points determined from @xmath31(t ) data at several field values ( _ i.e. _ , for h = 20 , 80 , 100 , 120 and 140 kg ) , and for the point obtained from @xmath31(h ) at our minimum temperature ( t = 0.4 k ) ; these give a caliper of the errors bars in determining the points on this phase diagram for the whole field and temperature range . two transitions can be observed in fig.[f22 ] , at low fields , around 3.4 k , and 1.4 k respectively . as h is being increased , the low temperature line slowly approaches t = 0 around h @xmath107 kg . the step in @xmath102 associated with the upper transition ( fig.[f21]b ) broadens as the field increases , resulting in increasingly large error bars in determining these transition temperatures . as already mentioned , it is uncertain if the transition persists up to h = 140 kg , or if cross - over occurs between 120 and 140 kg . * @xmath39 * + given the differences between ybptin and ybpt@xmath1 in evidenced by both thermodynamic and transport data ( figs.[f01]-[frt300 ] ) , it is desirable to compare similar measurements on the two compounds , and to study the effect of the small stoichiometry change on the field - induced qcp . consequently , in fig.[f03 ] we present the low - temperature specific heat data of ybpt@xmath1 in , for various values of the applied field h @xmath108 . a well - defined peak at t @xmath109 k in the h @xmath110 data may be associated with the magnetic ordering of this compound . as the applied magnetic field is increased , this transition ( indicated by small arrows ) moves to lower temperatures , and eventually drops below 0.4 k around 35 kg . from the linear extrapolation of the h = 0 data from t @xmath111 k down to t = 0 ( inset , fig.[f03]a ) , we can estimate the electronic specific heat coefficient @xmath72 as @xmath112 mj / mol k@xmath11 . fig.[f03]b the magnetic specific heat data for the same field values , @xmath74(ybpt@xmath1in)@xmath75(luptin ) , plotted as c@xmath113 _ vs. _ t@xmath11 . fig.[f03b ] shows the magnetic specific heat as @xmath114 _ vs. _ @xmath115 , for the same field values as before . a logarithmic divergence can be observed in these data , with the largest temperature region where @xmath116 occurring around h = 35 kg ( dotted line ) . however , on the next measured curve ( _ i . , for h = 40 kg ) the linear region extends over a comparable temperature interval , at slightly higher temperatures than in the h = 35 kg case . it thus appears that the largest temperature region ( close to a decade ) for the logarithmic divergency of the c@xmath113 data may occur for some intermediate field value ( 35 kg @xmath81 h @xmath81 40 kg ) . these data could be described as @xmath117 , with the ranges for @xmath118 and t@xmath119 determined from the linear fits on the h = 35 and 40 kg curves : 420 mj@xmath10mol k@xmath120 430 mj@xmath10mol k@xmath11 , and 14.7 k @xmath103 t@xmath121 13.5 k. the above observations seem consistent with a qcp in this compound with critical field just above 35 kg . fig.[f04 ] shows the low - temperature resistivity data for various values of the applied magnetic field . a maximum in d@xmath90 , associated with the magnetic ordering can be identified at t @xmath122 k in the h = 0 data , and is marked by small arrow in fig.[f04]a ; as the field is increased above 35 kg , small arrows indicate that this transition temperature drops below 0.4 k , consistent with the specific heat data . as was the case of ybptin , in ybpt@xmath1 in the temperature dependence of the resistivity outside the ordered state is ambiguous . from figs.[f04]b-[f04sqab ] it is unclear whether the resistivity is linear or quadratic in field for h @xmath123 , particularly for intermediate field values ( h @xmath106 55 kg ) . a detailed analysis of the resistivity data above the critical field will be performed in the discussion section , for both compounds . this should allow us to better clarify the position of the qcp and the existence of a fermi liquid - like ( fl ) regime in these systems . transverse magnetoresistance measurements were taken at constant temperatures ranging from 0.4 k to 10 k. as shown in fig.[f05 ] , two different temperature regimes can be identified in these data : for t @xmath124 k ( fig.[f05]a ) , two transitions can be distinguished . the small arrows mark the positions for these two transitions for t = 0.4 k , whereas the inset illustrates how these critical field values were determined . as the temperature increases up to about 1.8 k , the upper transition moves down in field and broadens , whereas the position of the lower one seems almost unaffected by the change in temperature . for temperatures higher than 2 k ( fig.[f05]b ) , the magnetoresistance isotherms display only a broad feature that looks more like a cross - over rather than a transition . using the detailed c@xmath30(t , h ) and @xmath31(t , h ) measurements discussed above , the ybpt@xmath1 in t - h phase diagram for h@xmath125 can be constructed . as can be seen in fig.[f06 ] , it is qualitatively similar to the corresponding t - h phase diagram for the stoichiometric compound ( fig.[f14 ] ) : in ybpt@xmath1 in magnetic ordering occurs at around 2.2 k. an almost field independent phase line is apparent in the @xmath94 data around 8 kg , and it appears to persist close to the magnetic ordering temperature . increasing applied field drives the higher transition towards t = 0 at a critical field values around 35 kg . + * @xmath41 * + the similarities observed previously between ybptin and ybpt@xmath1 in are also present for the @xmath41 direction , as the c@xmath30(t , h ) and @xmath31(t , h ) measurements indicate . as in the case of the stoichiometric compound , significant torques on the specific heat platform and sample for h @xmath80 50 kg may be the cause of the different transition temperature values , as determined by the two data sets mentioned above . therefore we will only take into consideration c@xmath30(t , h ) data for h @xmath81 50 kg ( fig.[f17 ] ) . similar to the @xmath39 measurements ( fig.[f03 ] ) , the @xmath41 c@xmath30 curves reveal a magnetic transition around 2.1 k for h = 0 ( fig.[f17 ] ) , which drops to @xmath19 2 k for h = 40 kg , before the sample torques significantly ; small arrows indicate the position of the transition temperature for the three curves shown in fig.[f17 ] . the temperature and field dependent resistivity data ( fig.[f18 ] ) indicate a slow suppression of the magnetic order with the applied field . fig.[f18]a shows the @xmath31(t ) curves in various applied fields , with the large circles marking the phase transition as determined from d@xmath90 ; the inset illustrates how the transition temperature was determined for h = 0 . the critical field required to suppress this transition below our base temperature appears to be around 120 kg . given the limited temperature range at these high fields , for h = 130 and 140 kg we were unable to distinguish a linear or quadratic temperature dependence of the resistivity . the magnetoresistance isotherms are presented in fig.[f18]b , and the large squares on this plot also indicate the high - t magnetic phase transition ; in the inset , a few d@xmath126 curves are shown to illustrate how the critical field values for the transition were determined . based on the c@xmath30(t , h ) and @xmath31(t , h ) presented above , the @xmath41 t - h phase diagram for ybpt@xmath1 in can be obtained , as shown in fig.[f19 ] . at low fields , this t - h phase diagram is consistent with the in - plane one for this compound : a magnetic transition is apparent around 2.2 k , but the possible second one around 1.0 k is not visible in the h@xmath0c measurements ; the t @xmath127 2.2 k transition is driven down in temperature by increasing applied fields , and it approaches our base temperature ( _ i.e. _ , 0.4 k ) around 120 kg . lack of measurements below 0.4 k or above 140 kg limits our ability to probe the existence of a qcp in this orientation , similar to the one at h@xmath128 kg for @xmath39 . for both the ybptin and ybpt@xmath1 in compounds a number of similar properties , as well as systematic differences , could be distinguished . first , the small stoichiometry difference was apparent from single crystal x - ray measurements , which also resulted in slightly reduced lattice parameters and unit cell volumes in the ybptin system . furthermore , the resistivity values were shifted towards higher values in the pt - deficient compound ( figs.[f02]b-[frt300 ] ) for the whole temperature ( t = 0.4 - 300 k ) and field ranges ( h = 0 - 140 kg ) of our measurements . high enough fields suppressed the magnetic order in both systems , at least for field applied within the basal plane ; in a similar manner to the t - scale in these compounds , the critical field value was reduced in ybpt@xmath1 in by comparison to the analogous one in ybptin . for @xmath129 , the field values required to suppress the magnetic order in ybpt@xmath1 in and ybptin were close to , or respectively higher than our maximum available field ( _ i.e. _ , 140 kg ) ; this precluded us from studying the low temperature properties of the two compounds outside the ordered state , for this orientation of the field . for @xmath39 nfl - like behavior seems to occur in ybpt@xmath1 in for intermediate field values above the ordering temperatures , as indicated by the logarithmic divergence of the c@xmath113 data ( fig.[f03b]a ) . in ybptin the nfl region is less clearly defined in the specific heat , as its logarithmic divergence was limited to a small temperature range by a downturn in the c@xmath130 data ( fig.[f11a ] ) towards high t. even more ambiguous was the temperature dependence of the resistivity data in both compounds , for h @xmath123 . below we are presenting a detailed analysis of these data , aimed at distinguishing between , if existent , the nfl and fl regimes , and providing a more accurate estimate of the qcp position in each of the two compounds . given the ambiguity already mentioned in the power law that best describes @xmath68 data above the ordered state , we attempt to determine the exponent @xmath88 when the temperature dependence of the resistivity is given by : @xmath131 , where @xmath132 . this is best done by determining the slope of the @xmath133 data on log - log plots at various field values . however , the accuracy of the resulting @xmath134 values is highly dependent on accurate estimates of the residual resistivity values @xmath135 . in turn , the estimates of the @xmath135 values are made difficult by their enhanced values compared to the over - all resistivity values ( fig.[frt300 ] ) . consequently , we determine the residual resistivity values using two criteria : firstly , from fits of the low - t resistivity data to a @xmath136 function ; secondly , given that figs.[f04]b-[f04sqab ] ( for ybpt@xmath1 in ) , or fig.[f12sqab ] ( for ybptin ) suggests that low - t @xmath68 data are close to being either linear or quadratic in t , we also choose ( for mathematical rather than physical reasons ) a second - order polynomial function to determine @xmath137 as a function of field : @xmath138 . for each field value , the average @xmath137 resulting from the two aforementioned criteria is used to calculate @xmath132 plotted in fig.[f04sqcd]a and [ f12sqcd]a respectively , on a log - log scale . the large data scattering towards the lowest temperatures , particularly at high fields , is due to the fact that @xmath139 is calculated as the difference between two comparably large values , @xmath68 and @xmath137 . as a consequence , the values of the exponent @xmath88 are being determined from ( i ) linear fits shown as solid lines , for @xmath132 larger than @xmath19 0.1 @xmath140 ( marked by the dashed horizontal line ) , and ( ii ) from similar linear fits down to the lowest measured temperature ( dotted lines ) . together with the exponent values originally determined from fitting the low - t data to @xmath136 functions , we can now determine the @xmath88 values shown in fig.[f04sqcd]b and [ f12sqcd]b , as full symbols , with their corresponding error bars . in the case of ybpt@xmath1 in ( fig.[f04sqcd]a ) , the resistivity is linear on the log - log scale , up to some maximum temperatures t@xmath141 marked with large triangles , with a large change in slope for 60 kg @xmath81 h @xmath81 70 kg . this is equivalent to the exponent @xmath88 having a discontinuous change from @xmath142 below 60 kg ( _ i.e. , _ for h @xmath103 h@xmath4 ) , to @xmath143 in higher fields ( h @xmath144 h@xmath4 ) , as can be seen in fig.[f04sqcd]b . also shown in fig.[f04sqcd]b are the temperatures t@xmath141 below which the low - temperature @xmath145 manifests a fermi liquid - like exponent of @xmath146 2 , analogous to the cross - over previously reported for ybrh@xmath15si@xmath15@xcite and ybagge@xcite . the departure from linearity on the resistivity plots can be determined up to a temperature range , and not a unique temperature value , which results in finite error bars on these cross - over temperatures . however , a monotonic increase of the cross - over temperatures with field is observed , with the corresponding error bars also increasing . when adding this cross - over line to the h - t phase diagram ( see below fig.[f09 ] , full triangles ) , it appears to converge with the magnetic order upper boundary ( solid line ) around the qcp between 45 and 70 kg . when we describe the ybptin low - t resistivity data as @xmath147 for h @xmath103 h@xmath4 , we encounter a situation very similar to that seen in ybpt@xmath1 in : the logarithmic plot in fig.[f12sqcd]a , linear at low t , is indicative of a jump in slope for an applied field between 70 kg and 80 kg , where the slope gives the values of the exponent @xmath88 . it appears that the exponent @xmath88 has a discontinuous variation with field ( fig.[f12sqcd]b ( left ) ) from @xmath148 for h @xmath103 h@xmath4 , to @xmath149 when h @xmath144 h@xmath4 . this suggests a nfl - like regime just above the magnetically ordered state , with a cross - over to a fl state above 100 kg , making this compound fairly similar to the pt - deficient one , but also to the previously reported ybrh@xmath15si@xmath15@xcite and ybagge@xcite . as observed before for ybpt@xmath1 in , due to the data scattering , the departure of the low temperature resistivity plots from the fl - like @xmath146 2 behavior at the highest field allows us to determine a cross - over temperature range , rather than a temperature value . this results in finite error bars on the cross - over temperatures , as seen in fig.[f12sqcd]b ( right ) . however , an increase of the cross - over temperatures with field is observed , both in fig.[f12sqcd]b , and also in the revised h - t phase diagram ( fig.[f14beta ] ) ( full symbols ) . based on the low temperature thermodynamic and transport measurements , ybptin and ybpt@xmath1 in can be regarded as yb - based heavy fermion compounds with long range , possibly reduced moment ordering that can be driven through a field - induced quantum critical point , similar to the previously studied yb - based hf systems , ybrh@xmath15si@xmath15@xcite and ybagge@xcite . in the latter two compounds , the hall effect served as an additional tool for characterizing the qcp and its effects on the finite - temperature properties of these materials . in order to further explore the field - induced qcp our yb - based compounds , field - dependent hall resistivity measurements were performed for temperatures up to 300 k. however , the data on the stoichiometric compound was noisy , given the limited ( small ) crystal size , and meaningful hall resistivity data could only be collected for the solution - grown compound for h @xmath108 . these measurements are shown in fig.[f07 ] . we determined the hall coefficient r@xmath150 for ybpt@xmath1 in as @xmath151 , in the low - field ( h @xmath152 kg ) and high - field ( h @xmath153 kg ) regimes , shown in fig.[f07 ] for various t. ( this alternative definition for the r@xmath154 was preferred to @xmath155 due to the scattering of the data at low fields as seen later in fig.[f08]b ) . the results are shown in fig.[f10 ] on a semi - log scale . the high - h points ( squares ) show the expected leveling off of r@xmath154 as @xmath156 , whereas the low - h data appear to have a stronger temperature dependence . at temperatures higher than 25 k , the two data sets merge ( as the @xmath157 data become roughly linear for the whole field range [ fig.[f07]a ] ) . these data indicate that the field - dependent hall resistivity will be non - trivial and potentially of interest at low temperatures . in the field - dependent hall resistivity measurements shown in fig.[f07 ] , two possible regimes can be distinguished : a high - t regime ( t @xmath158 k ) , where a monotonic decrease with field can be observed , despite the scattering of the data ( fig.[f07]a ) , and a low - t regime ( t @xmath159 k ) , for which a minimum in the @xmath160h ) data appears and sharpens as the temperature decreases ( fig.[f07]b ) . for 5 k @xmath106 t @xmath106 25 k , the hall resistivity curves show a fairly broad minimum ( fig.[f07]b ) , marked by the large gray dots , which moves down in field with decreasing temperature . below 5 k , this minimum is more and more pronounced , and is almost unaffected by the change in temperature . coleman _ _ et al.__@xcite have indicated that r@xmath150(c ) data ( where c is a control parameter , _ i.e. , _ h in our case ) can be used to distinguish between two possible qcp scenarios : diffraction off of a critical spin density wave sdw ( manifesting as a change in the slope of r@xmath150(c ) at c@xmath18 ) or a breakdown of the composite nature of the heavy electron ( signaled by the divergence of the slope of r@xmath150(c ) at c@xmath18 ) . in our case , it is not clear what definition of the hall coefficient should be used for comparison with the theory ( _ i.e. _ , either r@xmath161 or r@xmath162 ) since the magnetic field h is itself the control parameter c. therefore in fig.[f08 ] we are presenting the r@xmath150(h ) curves determined using both of the aforementioned definitions . ( fig.[f08]a ) , the hall coefficient is linear up to a field value which varies non - monotonically with t , as indicated by the dashed line . this line is shifted to higher field values with respect to the hall resistivity line ( fig.[f08]a ) , and is marked here by the large diamonds . when using the r@xmath150(h)@xmath164 definition for the hall coefficient ( fig.[f08]b ) , the maximum - h points on the low - field linear fits could be used as the criterium for defining the hall line ( indicated by the dotted line ) . regardless of what criterium is being used , the hall measurements define a new phase line , distinct from any of the ones inferred from either the c@xmath30(t , h ) or the @xmath31(h , t ) data . fig.[hallht ] shows the @xmath96 phase diagram with the phase lines determined from the @xmath31(h , t ) ( h @xmath165 ) or the c@xmath30(t , h ) data ( lines ) , together with the cross - over line and the hall line inferred from the various definitions ( symbols ) . the line resulting from the hall resistivity @xmath32 data ( diamonds ) seems to persist even below t@xmath12 , down to our lowest t. however , the other two criteria used for the definition of the hall line ( r@xmath166 or h@xmath141 for the linear fit of r@xmath167 ) delineate lines , which more closely resembles the hall line observed for either ybrh@xmath15si@xmath15@xcite and ybagge@xcite as it appears to converge with the `` coherence '' line ( _ i.e. _ , the line defining the cross - over between the non - fermi liquid and fermi - liquid like regimes ) , and the high - t magnetic ordering phase boundary , at the qcp , around h@xmath4 . it should be noted , though that , unlike ybagge , for which the hall line is clearly not associated with the saturation of the yb moments@xcite , for ybpt@xmath1 in the hall line appears to be close to the saturation fields , at least at the lowest temperatures . based on the presented data , we can conclude that , in both ybpt@xmath1 in and ybptin , when magnetic field is increased above @xmath168 , the system first enters a nfl state , followed , at higher fields , by a fl - like state : the specific heat has a logarithmic divergence ( figs.[f03b ] and [ f11a ] ) and the resistivity shows a @xmath169 functional dependence ( fig.[f04sqcd]a , b and [ f12sqcd]a , b ) , with @xmath88 having a discontinuous change at intermediate field values , from @xmath148 to @xmath170 2 . moreover , from hall resistivity measurements on the pt - deficient compound , we were able to determine another totally distinct line in the t - h phase diagram ( fig.[f09 ] ) , convergent , towards the lowest temperature , with the cross - over line and the high - t magnetic order line . all these observations lead us to believe that a qcp exists in both compounds , around @xmath171 kg for ybpt@xmath1 in , and 60 kg for ybptin . to further study the nature of the field - induced qcp in the two systems , we analyze the field dependence of the electronic specific heat coefficient @xmath72 . also , in order to extend the comparison of the disordered state in our systems and in previously studied compounds ( ybrh@xmath15si@xmath15@xcite and ybagge@xcite ) , we analyze the field - dependence of the coefficients a in the @xmath172 ( @xmath88 = 2 ) resistivity , even though in both ybptin and ybpt@xmath1 in the resistivity was not exactly quadratic for h @xmath144 h@xmath4 , but @xmath146 2 . the field - dependent electronic specific heat coefficient @xmath72 could be estimated at low t , outside the ordered state , and these values are shown in fig.[f03a ] . for both compounds , @xmath173 was taken as the corresponding c@xmath77 value at t @xmath174 k , such as to avoid the low - t upturn in the highest field data . a drastic decrease of @xmath72 ( almost an order of magnitude ) can be observed in fig.[f03a ] for both compounds , for fields above their respective @xmath168 . the field dependence of the coefficient a of the quadratic resistivity is shown in fig.[f23]a , b for ybptin and ybpt@xmath1 in respectively . a @xmath175 divergence can be observed for both compounds , with @xmath176 estimated from the fit as 64.4 kg for ybptin , and 56.0 kg for ybpt@xmath1 in . this critical field value for the former system is consistent with the @xmath39 specific heat and resistivity data ( figs.[f11]-[f14 ] ) , which suggested that @xmath177 60 kg . in the case of ybpt@xmath1 in , where nfl behavior was observed between 35 kg and 55 kg , the critical field value can only be determined to a range within the above two field values ; the @xmath178 kg determined from the a _ vs. _ @xmath175 fit ( fig.[f23]b ) is close to the above field range , thus consistent with the specific heat and resistivity data ( figs.[f03]-[f06],[f04sqcd]-[f09 ] ) . the proportionality @xmath179 between the resistance coefficient a and the electronic specific heat coefficient @xmath72 is emphasized by the logarithmic plots in the insets in fig.[f23 ] . in the case of the ybptin compound ( fig.[f23]a , inset ) , the solid line represents the kadowaki - woods ratio @xmath180 cm @xmath10 ( mj / mol k)@xmath11 . such a value is close to that observed for many heavy fermion systems@xcite , but slightly higher than the @xmath181 ratio reported for ybrh@xmath15si@xmath15@xcite ; this , in turn , was larger than the values observed for most yb - based intermetallic compounds@xcite . the corresponding value was larger still in ybagge@xcite , and it would appear that this is a common feature of yb - based materials with field - induced nfl - like behavior . when we turn to the pt - deficient compound ( fig.[f23]b , inset ) , a smaller @xmath181 ratio is indicated by the solid line : @xmath182 cm @xmath10 ( mj / mol k)@xmath11 . this value is again one order of magnitude larger than that expected for many yb - based compounds@xcite , and several times smaller than the ratio observed for the stoichiometric ybptin compound ( fig.[f23]a ) . in light of observations of suppression of the a coefficient due to site disorder@xcite , we may have to consider the @xmath181 ratio for ybpt@xmath1 in as a reduced value from the enhanced ratio observed for the stoichiometric compound . the detailed field- and temperature - dependent measurements presented here allowed us to confirm that ybptin is a heavy fermion compound , as has been presented by trovarelli et al.@xcite . in addition , we showed that a field - induced qcp exists in this material , at least for @xmath39 , with @xmath183 kg . in addition to the magnetic field used as a control parameter , we showed that a small pt - deficiency introduced in this system had effects consistent with positive pressure applied to yb - based heavy fermion compounds . thus , in the ybpt@xmath1 in compound we also see a suppression of the magnetic ordering by applied magnetic field ; in addition , the small disorder also suppresses the ordered state ( both t@xmath12 and h@xmath105 have smaller values in this compound than the similar ones in ybptin ) . nfl regions can be observed in both our compounds for h @xmath103 h@xmath4 , characterized by linear @xmath184 ( @xmath170 1 ) and logarithmic divergency of the specific heat . for h @xmath144 h@xmath4 , the exponent @xmath88 approaches 2 , suggesting a cross - over to a low - t fl state at these field values , similar to ybrh@xmath15si@xmath15@xcite and ybagge@xcite . as the critical field required to suppress the magnetic order in ybpt@xmath1 in and ybptin in the c direction appeared to exceed our maximum field , experiments to higher fields would be desirable in order to extend the comparison with ybrh@xmath15si@xmath15 and ybagge to the effects of anisotropy on the field - induced qcp . ames laboratory is operated for the u.s . department of energy by iowa state university under contract no . this work was supported by the director for energy research , office of basic energy sciences . in ( full symbols ) , with the resistivity error bars shown for the lowest temperature ( t = 0.4 k ) ; the transition temperatures are indicated by two dotted lines ( for ybptin ) and one dashed line ( for ybpt@xmath1 in ) . note that the error bars in ( b ) are geometrical , and thus they scale proportionally at higher temperatures.,width=340 ] , with the arrows indicating the peak positions associated with magnetic phase transitions ; inset : the h = 0 c@xmath185(t@xmath11 ) curve , with its linear fit below 5 k ( dotted line ) extrapolated down to t = 0 to provide a rough estimate of the electronic specific heat coefficient @xmath73 mj / mol k@xmath11 ( b ) low - temperature part of the c@xmath113(t@xmath11 ) curves for various values of the applied field.,width=302 ] _ vs. _ t for ( a ) h = 0 - 60 kg and ( b ) h = 60 - 140 kg . all curves ( except for the h = 140 kg one ) are shifted up by multiples of 100 mj@xmath10mol k@xmath11 . the dotted line ( for h = 60 kg ) is a guide to the eye for the largest region of logarithmic divergency of c@xmath113.,width=340 ] ( h ) isotherms for ybptin , for h @xmath108 and t = 0.4 - 3.6 k , @xmath186 t = 0.2 k ; the arrows point to the transition fields at the lowest temperature , with the inset exemplifying how these critical fields are determined for t = 0.8 k.,width=340 ] specific heat data for ybptin , for h = 0 , 10 , 20 , 40 , 60 and 80 kg ; small arrows indicate the positions of peaks possible associated with magnetic phase transitions ( dotted arrows : peaks on c@xmath187 data for possibly torqued sample [ see text]).,width=340 ] ( t ) data for ybptin taken at different applied fields , for h @xmath188 . ( b ) @xmath102 curves for h = 20 , 80 , 100 , 120 and 140 kg . on both plots , the small arrows indicate the magnetic transition temperatures.,width=340 ] t - h phase diagram for ybptin , as determined from resistivity ( full symbols ) and specific heat ( open symbols ) data . error bars on points from @xmath31(t ) data at h = 20 , 80 , 100 , 120 and 140 kg , and from @xmath31(h ) at t = 0.4 k shown as a caliper of the errors in determining the points on this phase diagram.,width=340 ] in , for h @xmath108 , with the arrows indicating the positions of the peaks associated with magnetic ordering ; inset : the h = 0 c@xmath185(t@xmath11 ) curve , with its linear fit below 5 k ( dotted line ) extrapolated down to t = 0 to provide a rough estimate of the electronic specific heat coefficient @xmath73 mj / mol k@xmath11 ( b ) low - temperature part of the c@xmath113(t@xmath11 ) curves for various values of the applied field.,width=340 ] _ vs. _ t for ( a ) h = 0 - 40 kg and ( b ) h = 40 - 140 kg . all curves ( except for the h = 140 kg one ) are shifted up by multiples of 200 mj@xmath10mol k@xmath11 . the dotted lines ( for h = 35 and 40 kg ) are guides to the eye for the linear regions on the c@xmath113 curves for h possibly just above and below h@xmath4 ( see text).,width=340 ] in low - temperature resistivity data for h @xmath108 , for ( a ) h = 0 , 10 , 20 , 25 , 30 , 35 and 40 kg and ( b ) h = 40 , 42.5 , 45 , 50 , 55 , 60 , 70 , 80 , 100 , 120 and 140 kg . small arrows in ( a ) indicate the possible ordering temperatures , whereas in ( b ) the solid lines are guides to the eye for the potentially linear regions on the h = 42.5 , 45 , 50 and 55 kg curves.,width=332 ] in @xmath31(t@xmath11 ) data , for ( a ) h = 42.5 , 45 , 50 , 55 , 60 and 70 kg and ( b ) h = 70 , 80 , 100 , 120 and 140 kg , with solid lines as guides to the eye for the potentially linear regions for the h @xmath80 45 kg curves.,width=309 ] ( h ) isotherms for ybpt@xmath1 in , for h @xmath108 and ( a ) t = 0.4 - 1.0 k ( @xmath186 t = 0.1 k ) , 1.2 - 2.2 k ( @xmath186 t = 0.2 k ) , and ( b ) t = 2.2 - 3.0 k ( @xmath186 t = 0.2 k ) , 4 , 6 , 8 and 10 k ; the arrows point to the transition fields at t = 0.4 k , with the inset exemplifying how these critical fields were determined.,width=340 ] phase diagram for ybpt@xmath1 in , as determined based on specific heat ( open symbols ) and transport ( full symbols ) data . the error bars on the high - field points represent the uncertainty in determining the cross - over temperatures.,width=340 ] in low - temperature resistivity data for h @xmath188 , for h = 0 - 140 kg ( @xmath186h = 10 kg ) , with the large dots marking the possible magnetic ordering temperature ; inset : h = 0 d@xmath189dt , with the small arrow marking the transition temperature . ( b ) @xmath31(h ) isotherms for t = 0.4 - 3.0 k ( @xmath186 t = 0.1 k ) . the critical fields for the possible phase transition were determined as local maxima in d@xmath126 ( as illustrated in the inset for t = 1.2 - 1.6 k ) , and are shown as large squares.,width=340 ] in @xmath190 _ vs. _ @xmath191 data for h = 45 , 50 , 55 , 60 , 70 , 80 , 100 , 120 and 140 kg , with their respective linear fits at very low t ( solid lines ) . the large triangles indicate the temperatures at which deviations from linearity occur . ( b ) exponent @xmath88 ( left ) and cross - over temperature t@xmath141 ( right ) as a function of h , with the corresponding error bars.,width=332 ] _ vs. _ @xmath191 data for h = 70 , 80 , 100 , 120 and 140 kg , with their respective linear fits at very low t ( solid lines ) . the large triangles indicate the temperatures at which deviations from linearity occur . ( b ) exponent @xmath88 ( left ) and cross - over temperature t@xmath141 ( right ) as a function of h , with their corresponding error bars.,width=332 ] in phase diagram ( h @xmath108 ) . the solid lines correspond to the phase lines from the t - h phase diagram in fig.[f06 ] , and the symbols delineate the new phase lines as determined from the @xmath192 data , using various criteria as described in the text.,width=340 ] ) . the solid lines correspond to the phase lines from the t - h phase diagram in fig.[f14 ] , and the symbols delineate the new phase line as determined from the @xmath192 data , using various criteria as described in the text.,width=340 ] in ( h @xmath108 ) for ( a ) t = 25 - 300 k and ( b ) t = 0.4 - 20 k ( except for the t = 0.4 k and 25 k , all curves are shifted up by multiples of 0.2 @xmath193 for clarity ) . the lines represent the phase transitions from the phase diagram in fig.[f06 ] ; the large diamonds indicate the @xmath32 minima.,width=340 ] in phase diagram ( h @xmath108 ) . the solid lines now represent the phase lines from the t - h phase diagram in figs.[f06 ] and [ f09 ] , and the symbols delineate the new phase line as determined from the hall resistivity data , using various criteria as described in the text.,width=340 ] t@xmath11-resistivity coefficient a as a function of field , for ( a ) ybptin and ( b ) ybpt@xmath1 in ( symbols ) ; the solid line represents a 1@xmath194 ) fit , from which the expected critical field values @xmath176 were estimated ( see text ) . inset : log plot of a vs. @xmath72 ( symbols ) , with a linear fit ( solid line ) used to estimate the a @xmath195 ratio.,width=340 ]

the hole - doped antiferromagnetic spin- two - leg ladder is an important model system for the high- superconductors based on cuprates . using the technique of self - similar continuous unitary transformations we derive effective hamiltonians for the charge motion in these ladders . the key advantage of this technique is that it provides effective models explicitly in the thermodynamic limit . a real space restriction of the generator of the transformation allows us to explore the experimentally relevant parameter space . from the effective hamiltonians we calculate the dispersions for single holes . further calculations will enable the calculation of the interaction of two holes so that a handle of cooper pair formation is within reach .

1103.0162

the hole - doped antiferromagnetic spin-@xmath0 two - leg ladder is used as a model system , theoretically as well as experimentally , for the two - dimensional cuprate superconductors . because the spin ladder is quasi one - dimensional it can more easily be treated by numerical and analytical approaches than the full two - dimensional model . for this reason the spin ladder was subject to many theoretical investigations in recent years @xcite . from a strong coupling perspective it is reasonable to describe the doped spin ladder by a @xmath2-@xmath3-model @xcite . in this paper , our aim is to derive an effective hamiltonian which describes not only the motion of the magnetic degrees of freedom , i.e. , the triplons @xcite , but also of the charges , i.e. , of the doped holes . this task is more challenging than the description of the triplons alone because the two kinds of excitations interact strongly and their energy ranges are not separated but they strongly overlap . hence the excitations are not infinitely long - lived but they decay , at least in certain regions of the brillouin zone . the technique employed here is the approach of self - similar continuous unitary transformations ( scut ) . in this approach the mapping of the true ground state to a vacuum of elementary excitations is constructed systematically , details will be explained in sect . [ chap : scut ] . the basic idea is to implement a transformation that adjusts itself during the procedure of the diagonalisation depending on the current form of the hamiltonian at this instant of the continuous transformation . the change of the hamiltonian induced by the transformation is determined by the current magnitude of the non - diagonal elements . the flowing hamiltonian is given in second quantization , i.e. , it has a certain structure in terms of elementary creation and annihilation operators @xcite . this formulation incorporates the linked cluster property @xcite automatically . the scut systematically defines a set of differential equations for the prefactors of the monomials of the creation and annihilation operators . the key issue is to construct the flow to the effective hamiltonian in a robust way . due to the energetically overlapping states this goal required modifications @xcite with respect to the previously performed unitary transformations . thereby we achieved the successful calculation of the dispersion of a single hole . in sects . [ chap : hole disp without ring exc ] and [ chap : hole disp with ring exc ] we investigate the dispersions of single hole excitations obtained from the effective hamiltonians , which are results of the scut . we also compare our results to the results of other methods . sects . [ chap : summ ] and [ chap : outlook ] conclude this article with a summary and perspectives for future work . ladder consisting of the upper leg u , the lower leg l and the rungs @xmath4 . the magnetic couplings are the nearest neighbor coupling on the rungs @xmath5 , the nearest neighbor coupling along the legs @xmath6 and the four spin coupling @xmath7 for the ring exchange . the nearest neighbor hopping is described by the constants @xmath8 ( on the rungs ) and @xmath9 ( along the legs ) . the symmetry axis of the ladder is depicted by the dashed line . ] the spin-@xmath0 ladder is realised in the so called telephone number compounds @xmath10 @xcite . for these systems superconductivity can be detected under high pressure @xcite . interactions between the ladders are weak because they result from @xmath11 exchange @xcite . thus the ladders can be considered to be isolated from one other . even if the interladder coupling is taken into account , the strong frustration of the lattice causes the system to be effectively one - dimensional @xcite . the dispersions of the magnetic excitations of the complete layer with interladder coupling are similar to those of a single spin ladder . important magnetic properties of the ladders are investigated in refs . . the focus of the present article is the lightly doped spin ladder because this will allow us to address the dynamics of charges . of course , this system has been investigated previously by many techniques , for instance exact diagonalization @xcite , density - matrix renormalization @xcite , high - order series expansion @xcite , diagrammatic approaches @xcite , and quantum monte carlo @xcite . hence much is known about the dynamics of a single hole , i.e. , its dispersion , its spectral density and to some extent also about the bound states of two holes . the aim of the present paper is to establish an approach which generates an effective model for the motion of holes and their interaction with magnetic excitations and other holes . such a model can be expressed in terms of creation and annihilation operators which create and annihilate holes , i.e. , charge excitations , and triplons , i.e. , magnetic excitations . the techniques used so far do not yield such an effective model but numerical data for eigen energies and spectral weights . for this reason we here use the complementary approach of self - similar continuous unitary transformations . the spin sites in our model ( see fig . [ ladder ] ) are the @xmath12 orbitals of the copper atoms coupled via the @xmath13 or @xmath14 orbitals of the oxygen atoms , which hybridise with the copper @xmath12 orbitals so that an antiferromagnetic superexchange @xcite is possible . the hamiltonian in the usual form of a @xmath2-@xmath3-model , for derivations see ref . and references therein , reads @xmath15 \nonumber\\ & & - t_{\perp } \sum_{n , \alpha=\mathrm{u , l } } c^{\dagger}_{\sigma , \alpha , n } c^{\phantom{\dagger}}_{\sigma , \overline{\alpha } , n } - t_{\parallel } \sum_{n , \alpha=\mathrm{u , l } } c^{\dagger}_{\sigma , \alpha , n } c^{\phantom{\dagger}}_{\sigma , \alpha , n\pm1}. \nonumber \end{aligned}\ ] ] the nearest neighbor terms with the couplings @xmath5 and @xmath6 ( see fig . [ ladder ] ) are not sufficient to describe the magnetic interactions in the system . because the hybridization path around the @xmath16 square plaquettes is strong , the influence of the four - spin interactions belonging to these plaquettes ( i.e. two neighboring rungs ) is not negligible @xcite . these interactions are referred to as ring exchange ( also cyclic exchange ) with the coupling @xmath7 ( see fig . [ ladder ] ) . actually the complete ring exchange also includes two - spin terms for all two - spin combinations of the four spins on two neighboring rungs . yet the contributions from the terms coupling the spins along the rungs and parallel to the legs are merged with the nearest neighbor terms so that the coupling constants @xmath5 and @xmath6 include these contributions , whereas the terms coupling the spins diagonally can be neglected because their prefactor is only of the order of @xmath17 of @xmath5 @xcite . the complete representation of the ring exchange using spin operators can be found in ref . . the hopping of the spins is described by the constants @xmath8 and @xmath9 ( see fig . [ ladder ] ) . the creation and annihilation operators @xmath18 and @xmath19 already incorporate hardcore properties so that double occupancy is forbidden . we define the dimensionless quantities @xmath20 which are normalized to @xmath5 . it is convenient to treat the ladder as a one - dimensional chain of rungs . there are nine possible local states on a rung @xmath4 of the ladder . @xmath21 the states without holes are the singlet ( [ singlet ] ) and the triplet states ( [ tripletx]-[tripletz ] ) , for which we use the representation from ref . because the symmetry in spin space is manifest in this representation . the states consisting of one hole and one spin @xmath0 ( [ hole1]-[hole4 ] ) have the quantum numbers @xmath22 , which is the parity of the state with respect to the symmetry axis of the ladder ( see dashed line in fig . [ ladder ] ) , and @xmath23 , which indicates whether the @xmath24 component of the spin is up or down . the double hole state ( [ 2holes ] ) is neglected here because only one - hole states will be considered . the local energy of @xmath25 is also larger than the energy of the other states . therefore @xmath25 is expected to be also less important for ladders with low , but macroscopic doping , though it will influence the quantitative results for hole - hole scattering and hole - hole binding . for the application of the continuous unitary transformation a representation using the creation and annihilation operators for the local states ( [ singlet]-[hole4 ] ) is indicated . the complete hamiltonian ( [ hamiltonian ] ) in the respresentation using these operators can be found in ref . . note that the local rung states incorporate hardcore properties . also the fermionic states exhibit additional hardcore properties which simply result from the fact that each rung can only be in _ one _ particular state . the excitations are actually dominated by the local states as long as the couplings between the rungs are small . the magnetic excitations are triplons , i.e. , triplet states dressed with the magnetic interactions with their environment @xcite . we do not want to use the term `` magnon '' for these magnetic excitations because this term is usually associated with quasiparticles in systems that exhibit long - range magnetic order ( which is not the case for the spin ladder ) . moreover triplons feature the appropriate threefold degeneracy based on their @xmath26 character @xcite . we use the method of continuous unitary transformations ( cut ) to obtain an effective hamiltonian @xcite . instead of applying only a single constant unitary transformation that diagonalises the hamiltonian at once or several constant unitary transformations successively , a unitary transformation depending on a continuous parameter is applied to the hamiltonian . this transformation adjusts itself permanently during its application . discrete transformations must be known explicitly before we can apply them , whereas for the continuous transformation it is sufficient to set up the infinitesimal generator of the transformation . the unitary operator @xmath27 depending on the continuous parameter @xmath28 transforms the hamiltonian by @xmath29 the generator of the transformation is defined by @xmath30 the generator defines the properties of the transformation . the derivative of @xmath31 with respect to @xmath28 is given by the so - called flow equation @xmath32\,\mathrm{.}\ ] ] this is actually a system of differential equations ( generically highly coupled ) for the coefficients of the operators appearing in @xmath33 . to achieve the effective hamiltonian the flow equation has to be solved . for a self - similar application of the cut @xcite the hamiltonian is represented by a sum of different operators @xmath34 , which are products of local operators in second quantization , multiplied by prefactors @xmath35 . @xmath36 the operators @xmath34 determine the structure of @xmath33 . this is the reason why this formulation of the cuts is called self - similar . in particular , we express the operators @xmath34 in terms of creation and annihilation operators of charge excitations ( holes ) and spin excitations ( triplons ) in the manner introduced previously @xcite . a reference state , the vacuum , has to be chosen with respect to which these creation and annihilation processes are defined . in our case , this state is the product of local singlets on the rungs of the ladder . for the undoped ladder this state is a convenient starting point as seen in many high - order perturbative @xcite and diagrammatic approaches @xcite . since we focus in this work on single hole dynamics the use of the product of local singlets on the rungs as reference vacuum is well justified . the differential equations for the @xmath37 are given by the flow equation in the form @xmath38 because an infinite system yields an infinite number of differential equations which can not be solved numerically , we have to apply a truncation to @xmath33 so that a closed system of differential equations is obtained . we choose a real space truncation because the correlation length is finite for the spin ladder due to the energy gap @xcite . the concrete truncation scheme is explained in sect . [ chap : trunc_scheme ] . it should also be noted that the truncation is only restricted to operators . there is no truncation of the hilbert space of the states . thus although operators affecting higher numbers of particles are omitted , the number of particles that can be treated is arbitrarily large . to illustrate this we consider the action of a term in second quantisation , e.g. , @xmath39 , which annihilates one particle and creates two this term acts not only on the one - particle subspace , but also on the subspaces with more than one particle , e.g. it changes a four - particle into a five - particle state . the solution of the flow equation is a rather straightforward numerical integration . the convergence is monitored during the integration . since all contributions to the generator decrease in case of convergence , the absolute values of the concerning coefficients are squared and summed . the square root of this sum is defined as residual off - diagonality ( rod ) . it is the norm of the generator and also a measure for the convergence . the rod is expected to tend to zero for @xmath40 . note that the term `` off - diagonality '' is meant in a broader sense , i.e. the definition of the generator determines which elements shall be kept for @xmath40 and these elements are defined as `` diagonal '' parts of the hamiltonian . the rods depicted in the following are always normalized to the initial rod at @xmath41 . with the decrease of the rod the wanted effective hamiltonian is approached . if the relative rod falls below a certain threshold specifying the precision of the result ( usually @xmath42 ) , the integration can be considered to be completed . the number of coefficients treated in our actual calculations is of the order up to @xmath43 for the doped spin ladder . induced by the pc generator : each coloured ( @xmath44 , @xmath45)-block represents contributions to @xmath31 creating @xmath44 particles after annihilating @xmath45 particles . ] the particle conserving or pc generator @xmath46 was introduced by mielke for band matrices @xcite for @xmath47 and simultaneously by knetter and uhrig for many - body hamiltonians where @xmath48 counts the number of elementary excitations @xcite . the indices @xmath49 and @xmath50 label the transition from state @xmath50 to state @xmath49 . these states are eigenstates of an operator @xmath51 , which counts the number of elementary excitations , i.e. , quasi - particles . the eigen value of @xmath51 for the state @xmath49 is @xmath48 . the differences between the pc and the wegner generator are discussed by means of simple hamiltonians in ref . . in particular the convergence behavior is examined . the wegner generator @xmath52 $ ] @xcite always leads to a fixed point , but it does not necessarily aim at a quasi - particle picture because degeneracies may hinder the diagonalisation . the pc generator is not sensitive to degeneracies , but the induced transformation does not always converge for infinite systems . due to the sign function the generator ( [ pc generator ] ) contains only terms of the hamiltonian that change the particle number . terms which decrease the particle number acquire an additional minus sign . the resulting transformation yields an effective hamiltonian for @xmath40 which conserves the particle number ( see fig . [ pc cut blocks ] ) . the asymptotic behavior @xcite of the non - particle - conserving terms , i.e. , @xmath53 , is dominated by @xmath54 because @xmath55 tends to zero for @xmath40 if @xmath53 , the transformation tries to sort the eigenenergies @xmath56 according to the particle number @xmath48 , i.e. , @xmath57 holds true for large @xmath28 . overlapping energies of states with different quasiparticles can hinder the convergence of the flow . this problem can be solved by excluding terms from the generator @xcite . the ground state or gs generator @xmath58 introduced in ref . includes only terms that couple to the vacuum @xmath59 of excitations so that in the effective hamiltonian only this vacuum is decoupled from the remaining hilbert space . then it has become the ground state . this is sufficient for the calculation of the hole dispersion for the spin ladder if the ground state can be continuously linked to the state without magnetic excitations for a given hole configuration . this is the case for a single hole at a given momentum close to the minimum of its band . and @xmath60 . the cyan ( gray ) dashed lines do not pose a problem because they do not cross the one - particle dispersion . convergence problems can only be induced by the terms that act on the states with the cyan ( gray ) dotted dispersions , which cross the one - particle dispersion . ] but even the gs generator encounters convergence problems if the interaction between the rungs becomes too strong , i.e. , @xmath61 , @xmath62 . hence we have developed a generator adaption with a real space restriction of the generator terms . many terms that remain after the real space truncation of the hamiltonian act on states with discrete energies within the continua ( see fig . [ generator truncation ] ) . the divergence of the scut in case of an overlap is not induced by all terms that mediate between overlapping continua , but only by a part of these terms . thus a stricter real space truncation of the hamiltonian can induce convergence . however , a stricter truncation causes a larger error for the results of the scut . therefore we do not apply a stricter truncation to the hamiltonian but we choose a stricter generator . we emphasize that a restriction of the generator does not imply an approximation . it only changes the direction of the unitary rotation . this restriction of the generator is not based on the total extension of the terms , but on the extension of the creation operators and on the extension of the annihilation operators separately . let us consider the example of an overlap between one- and two - particle energies shown in fig . [ generator truncation ] . the terms @xmath63 and @xmath64 are responsible for transitions between the one- and the two - particle subspace . we choose to restrict the generator based on @xmath65 , which is the distance between the two particles which are created or annihilated . because the one - particle state does not have an extension , it is not relevant for the restriction of the generator . all terms that have a larger @xmath65 than a certain @xmath66 are excluded from the generator . note that they are still part of the hamiltonian as long as they meet the truncation criteria for the hamiltonian . the transformation induced by this restricted generator does not try to sort all eigenenergies ( cf . [ sorting of the eigen values ] ) , but only those which are captured by the terms in the generator . therefore the flow may also converge in case of overlapping energies . the price to be paid is that the subspaces affected by the omitted terms are not completely decoupled from the remaining hilbert space . thus either an additional diagonalization or a diagrammatic analysis has to be applied to the effective hamiltonian from this scut or the results have to be considered as upper limit for the actual results . the restriction of the generator can be applied to the pc generator or to any of its adaptions , e.g. , to the gs generator . the finite energy gap @xmath67 of the triplons @xcite is equivalent to a correlation function which is exponentially decreasing with respect to the distance . hence a truncation in real space is appropriate . the extension in real space shall be used as a measure for the physical importance of a term of the hamiltonian . in our quasi one - dimensional spin ladder the extension of a term is defined as the difference between the smallest and the largest rung index of the local operators within the term . the simplest way of truncating would be one maximal extension in real space for all terms . terms exceeding this limit would be omitted . but this approach is not reasonable in our case . the number of possible terms increases much more strongly with the maximal extension for terms consisting of more local operators . however , terms consisting of less operators are usually more important . hence higher extensions should be taken into account for terms with less operators . for instance , the coefficient of a one - particle hopping term consisting of two local operators is usually larger than the coefficient of a two - particle interaction term consisting of four local operators if they both have the same extension . therefore , for the pure triplon terms different maximal extensions @xmath68 are defined in units of the rung distance where @xmath4 is the total number of local operators of the term under study . because only terms changing the particle number by two or conserving the particle number occur in the initial magnetic hamiltonian , only the @xmath68 with even @xmath4 are meaningful . by @xmath69 we denote the maximal extension for @xmath70 hole state operators . it does not matter if the term concerned contains additional triplon operators . for terms with triplon and hole state operators , @xmath69 affects only the hole state operators . the total maximal extension @xmath71 for these mixed terms depends on the number of triplon operators @xmath72 . the parameter @xmath73 denotes the maximal number of triplons interacting with holes . note that @xmath71 with odd @xmath72 have to be taken into account because the triplon number is changed by an odd number if the hole state parity is altered . additionally a maximal triplon number @xmath74 and a maximal hole number @xmath75 are defined for the operators , i.e. , terms affecting higher triplon numbers than @xmath74 or higher hole numbers than @xmath75 are completely omitted . this truncation scheme is not compulsory . other schemes could be implemented which use different classifications for the groups of terms sharing the same maximal extension . since the size of the system of differential equations for the doped ladder grows exponentially with increasing extensions , the truncation for our calculations is fairly strict . the parameters @xmath76 , @xmath77 , @xmath78 , @xmath79 and @xmath80 were used for the pure triplon terms because these maximal extensions are sufficient for the undoped case up to @xmath61 . for the terms including hole operators the truncation used for the scut is given by @xmath81 , @xmath82 , @xmath83 , @xmath84 , @xmath85 , @xmath86 , @xmath87 , @xmath88 and @xmath89 . all results shown in the sequel are obtained with these parameters unless stated otherwise . calculated with the pc generator ; left : @xmath90 , right : @xmath91 . the curves coincide well with the series expansion results from . ] the dispersion of a single hole in the absence of triplons can be easily derived from the effective hamiltonian by means of fourier transformation . the part of the hamiltonian to be diagonalised @xmath92 contains the one - hole terms restricted by @xmath93 and characterised by the coefficients @xmath94 . the one - hole dispersion @xmath95 only depends on the parity @xmath22 and it is degenerate concerning the spin @xmath23 . if @xmath96 and @xmath97 are small while @xmath98 , the dispersion reads @xmath99 in the following we only consider the isotropic case @xmath100 . for small values of the parameters @xmath101 , @xmath102 and @xmath103 the deviations from eq . ( [ hole dispersion for small parameters ] ) are actually small . in fig . [ x=0 lambda=0.25,0.5 ] two values of @xmath103 are considered for @xmath98 . the result for @xmath90 already exhibits deviations from the relation ( [ hole dispersion for small parameters ] ) for small hopping constants . both bandwidths are smaller than @xmath104 . the odd band is narrower than the even band . moreover the odd band is shifted upwards by less than @xmath103 , while the even band is shifted downwards by slightly more than @xmath103 . , @xmath105 calculated with the gs generator and the pc generator . the results for the even bands coincide so that the curves can not be distinguished . the cyan ( gray ) shaded area is the continuum formed by one triplon and one even hole state . ] the increase of @xmath103 to @xmath106 yields a result with obvious deviations relative to eq . ( [ hole dispersion for small parameters ] ) . the odd band becomes lower and narrower with growing @xmath103 in this region . the cosine shape of both dispersions is for @xmath90 as well as for @xmath91 not deformed by higher harmonics . the results of the series expansion @xcite exhibit the same behavior for these parameters in good agreement with the scut results . however , it has to be pointed out that the convergence of the scut is much worse for @xmath91 than for @xmath90 . the residual off - diagonality ( rod ) defined in sect . [ chap : scut ] and used as a measure for the convergence is decreasing very slowly for @xmath91 . while the rod is smaller than @xmath107 at @xmath108 for @xmath90 , it is still larger than @xmath109 at @xmath108 for @xmath91 . both rods are decreasing exponentially for large @xmath28 . theses rods are not shown here . for @xmath98 , @xmath105 the pc generator yields results for the even band that still agree very well with the series expansion . for the odd band there are no series expansion results available . [ x=0 lambda=1 ] shows the comparison of the results for the pc generator and for the gs generator . the even band results lie on top of each other , but the odd band exhibits deviations for @xmath110 . while the gs generator produces an almost featureless dispersion ( compared to the even band ) , the pc generator causes a pronounced maximum at @xmath111 . these differences are due to the position of the lower boundary of the continuum formed by one triplon and one even hole state . ] ( see also fig . [ x=0 lambda=1 ] ) . this is the odd continuum due to the odd parity of the triplon . an overlap between the odd dispersion and this continuum is present for the pc result . the deviating odd band from the gs generator avoids this overlap . for both generators the even and the odd band cross at @xmath112 . while the even band keeps its cosine shape , the second harmonic for the odd band is no longer negligible for both the pc and the gs generator . , @xmath105 induced by the pc generator ( solid line ) and by the gs generator ( dashed line ) . ] for @xmath98 , @xmath105 the transformation does not converge for the pc generator , while the gs generator induces convergence ( see fig . [ x=0 lambda=1 rod ] ) . the overlap hinders the convergence for the pc generator . the kink at @xmath113 in the gs rod is probably due to numerical inaccuracies that are amplified via a feedback within the flow equation . since the relative rod is already @xmath114 at the kink we can stop the transformation at this point and neglect the remaining off - diagonal terms , i.e. , we consider the transformation to be converged . for the calculation of the hole dispersions the pc scut was stopped at @xmath115 where the rod is still @xmath116 and thus not negligible . therefore the result from the gs generator is more trustable than the result from the pc generator . the parameters @xmath98 , @xmath117 lead to divergence for either generator . here we want to consider @xmath119 . for @xmath120 the energy of the hole states , which is independent of momentum @xmath121 and parity @xmath22 , decreases with increasing @xmath101 . already for small finite @xmath103 the deviations from the simple cosine shape appear for the odd band . this can be seen in fig . [ x=0.5 lambda=0.25,0.5 ] where the one - hole dispersions for @xmath122 and @xmath90 are depicted . the odd band agrees well with the series expansion ; only slight deviations at @xmath123 occur . for the even band no series expansion data are available . , @xmath124 , @xmath90 calculated with the pc generator . the series expansion result for the odd band is also shown , while the result for the even band is not available . right panel : one - hole dispersion for @xmath122 , @xmath124 , @xmath91 calculated with the pc generator and the gs generator . for case a ) in the legend the usual truncation was used . for case b ) in the legend @xmath93 was increased to @xmath125 , @xmath126 and @xmath127 to @xmath128 as well as @xmath129 and @xmath130 to @xmath131 . the dotted curves are extrapolations from the series expansion data . ] for @xmath122 and @xmath91 , see also fig . [ x=0.5 lambda=0.25,0.5 ] , the result for the even band agrees again well with the series expansion result @xcite , but the odd band behaves differently . only for large @xmath121 the behavior is similar although also in this region the scut result is slightly lower . the series expansion result exhibits a local maximum at @xmath123 and a global minimum at @xmath132 , while the scut result hardly changes in the region @xmath133 . the gs generator yields the same result as the pc generator apart from minimal deviations ( @xmath134 ) at @xmath123 . however , the gs generator allows us to extend the truncation scheme : @xmath93 was increased to @xmath125 , @xmath126 and @xmath127 to @xmath128 as well as @xmath129 and @xmath130 to @xmath131 . in the result the shape of the odd band changes mainly for small @xmath121 . thus the odd dispersion still changes with increasing maximal extensions . , @xmath124 , @xmath91 . the odd hole dispersion was derived from scut using the pc generator ( cyan / gray solid ) and by series expansion ( cyan / gray dotted ) . the cyan ( gray ) shaded area is the continuum ( derived from the same scut calculation ) . ] we can understand these differences between scut and series expansion by looking at the position of the continuum formed by one triplon and one even hole state ( see fig . [ x=0.5 lambda=0.5 continuum ] ) . the series expansion result crosses the lower boundary of this continuum at @xmath135 . assuming that the series expansion yields a result close to the actual odd hole dispersion it is cogent that the scut based on the pc generator is not able to sort the eigen values properly for small @xmath121 because creating one triplon does not necessarily increase the energy . since the problematic overlap concerns the zero- and the one - triplon space , the use of the gs generator is no remedy here . this may come as a surprise at first since it was shown in ref . that the gs generator always implies a robust flow . this statement applies if the hole is put in its band minimum because this is by construction the lowest eigen state of the ladder doped with one hole . but for other energies of the hole dispersion this does not need to be the case so that the energy of one hole may lie higher than the energy of one hole and one triplon . hence the gs generator yields mainly the same result and suffers from the same problems as the pc generator . a robust extrapolation of the series can still yield reasonable results because the series is not fully sensitive to the overlap of states at higher values of the expansion parameters . , @xmath124 , @xmath91 calculated with the pc generator and the gs generator . the dashed curves are extrapolations from the series expansion data . ] , @xmath124 , @xmath91 . the odd hole dispersion was derived from scut using the pc generator ( cyan / gray solid ) , using the gs generator ( cyan / gray dotted ) and by series expansion ( cyan / gray dashed ) . the cyan ( gray ) shaded area is the continuum ( derived from the same pc scut calculation ) . ] for @xmath61 and @xmath91 ( see fig . [ x=1 lambda=0.5 ] ) the scut results for the odd band again show distinct deviations from the series expansion results . the even band exhibits a maximal deviation at @xmath123 of only @xmath136 . the deviations of the energetically higher odd band are larger , but still well tolerable for the gs generator . all in all the deviations are not as pronounced as for @xmath122 and @xmath90 . a comparison with the continuum formed by one even hole and one triplon ( see fig . [ x=1 lambda=0.5 continuum ] ) shows that an overlap exists around @xmath111 for the series expansion result and for the gs result . this explains again the deviations in this region . as this overlap is not as strong as the overlap for @xmath122 and @xmath91 the deviations are smaller . even if there is no actual overlap , the continuum is at least very close . hence it is to be expected that the actual odd hole dispersion is lowered for @xmath137 due to this reason . , @xmath124 , @xmath105 calculated with the pc generator and the gs generator . the dashed curve is the result from the series expansion , for which only the even band is available . the dashed - dotted curves are exact diagonalization results . ] for @xmath61 and @xmath105 ( see fig . [ x=1 lambda=1 ] ) no one - hole dispersion with odd parity is available from series expansion . thus we look at the even band including results from exact diagonalization . for the exact diagonalization a finite ladder with @xmath138 rungs was examined and the resulting eigenenergies were fitted by a sum of three cosine terms for both bands , see eq . ( [ fourier series ] ) . the pc result for the even band differs completely from the gs result . for the odd band the pc result lies below the gs result and the deviations grow with increasing momentum . the comparison with the series expansion and with the exact diagonalization suggests that the gs result is more reliable because the deviations are smaller than @xmath139 for the even band . also for the odd band the gs result is closer to the exact diagonalization result . the odd band is more difficult to assess due to the vicinity to the odd continuum . it should be noted that the diagrammatic approach from ref . yields bands that show qualitative deviations from the scut and from the exact diagonalization concerning the shape for small @xmath121 . the authors state that this regime actually exceeds the applicability of their approach . the quantum monte carlo result from ref . is in very good agreement with our result . for the even band the agreement is even excellent and comparable to the agreement between scut and exact diagonalization . , @xmath124 , @xmath105 . left : the odd hole dispersion was derived from scut using the pc generator ( cyan / gray solid ) and using the gs generator ( cyan / gray dotted ) . the cyan ( gray ) shaded area depicts the continuum with one even hole and one triplon derived from the same gs scut calculation . right : the even hole dispersion was derived from scut using the pc generator ( black solid ) , using the gs generator ( black dotted ) and by series expansion ( black dashed ) . the black shaded area depicts the continuum with one odd hole and one triplon derived from the same gs scut calculation . ] the odd band result from the gs scut enters the continuum formed by one triplon and one even hole state for @xmath140 ( see fig . [ x=1 lambda=1 continuum ] ) , while the pc result lies always below this continuum . this overlap is also present in the exact diagonalization result and in the quantum monte carlo result @xcite . a comparison of the even one - hole dispersion and the even continuum formed by one triplon and one odd hole state ( see also fig . [ x=1 lambda=1 continuum ] ) supports the assumption that the gs result is more reliable than the pc result . the dispersion induced by the gs scut exhibits a shape that appears as if it were formed by the lower boundary of the approaching continuum . this is plausible because the vicinity of the continuum lowers the dispersion . the pc result , however , stays away from the continuum at @xmath123 and at the boundary of the brillouin zone . the overlap with the continuum around @xmath141 occurs in the pc calculation only due to truncation . at this point we compare the convergence behavior for @xmath122 , @xmath91 ; @xmath61 , @xmath91 and @xmath61 , @xmath105 for the pc generator . the corresponding rods are depicted in fig . [ rod comparison ] . all three curves exhibit a kink with non - converging behavior afterwards . this is typical for cumulating rounding errors because of a symmetry breaking due to numerical inaccuracies . such a symmetry breaking is not unlikely because the full spin symmetry could not be used explicitly . it is surprising that the rod for @xmath61 , @xmath91 reaches a lower minimum than the one for the lower parameter set @xmath122 , @xmath91 . this seems to contradict the expectation that larger values of the interdimer processes imply a more difficult cut . but inspecting the slope of the rods in fig . [ rod comparison ] before reaching their minima one sees that these slopes , i.e. , the convergence velocities , are indeed largest for the smallest values for the interdimer processes . for @xmath61 and @xmath143 the parameters are entering a region which is expected to reflect realistic relations of the constants in the telephone number compounds . the results for @xmath61 and @xmath117 are shown in fig . [ x=1 lambda=2 ] . again we compare with results from series expansion and from exact diagonalization for @xmath138 rungs . , @xmath124 , @xmath117 calculated with the pc generator and the gs generator . the graph also shows the exact diagonalization data ( ed ) and the series expansion data ( se ) @xcite . for the latter only the even band is available . the even band results are black , the odd band results are cyan ( gray ) . ] approximate analytic results were obtained in ref . by perturbation theory improved by a variational ansatz . these results lie even above the series expansion results but confirm the qualitative shape for the even band . the pc result for both the even and the odd band is again very distinct from the other results comparable to @xmath61 , @xmath105 . the gs result , however , is in better agreement with the data from the series expansion and especially with the exact diagonalization result in accordance with our previous observations . apart from the pc result the dispersions exhibit the same features . the even band has a global maximum at @xmath123 and a local maximum at @xmath111 , while it is vice versa for the odd band . because both bands lie in the same energy range , they cross in the middle between @xmath123 and @xmath111 . the exact diagonalization predicts the crossing to be at @xmath144 , but the gs scut finds the crossing at @xmath145 . the even band calculated by series expansion is located above both the series expansion and the gs scut result for all @xmath121 . this is further evidence that the extrapolation used to correct the bare series underestimates the lowering of the band induced by the hybridisation with the hole - triplon continuum . , @xmath124 , @xmath117 . left : the odd hole dispersion was derived from pc scut ( cyan / gray solid ) , gs scut ( cyan / gray dashed ) and by exact diagonalization ( cyan / gray dotted ) . the cyan ( gray ) shaded area depicts the continuum with one even hole and one triplon ( derived from the same gs scut calculation ) . the cyan ( gray ) dash dot dot line is the lower boundary of the continuum formed by two triplons and one odd hole . right : the even hole dispersion was derived from pc scut ( black solid ) , gs scut ( black dashed ) , by exact diagonalization ( black dotted ) and by series expansion ( black dashed - dotted ) . the black shaded area depicts the continuum with one odd hole and one triplon ( derived from the same gs scut calculation ) . the black dash dot dot line is the lower boundary of the continuum formed by two triplons and one even hole . ] let us consider the continua formed by one hole and one triplon . the continua consisting of one hole and one triplon are compared to the one - hole dispersions in fig . [ x=1 lambda=2 continuum ] . the continua do not overlap with the hole dispersions , but they are very close to them . the only exception is the series expansion result for the odd band which actually exhibits an overlap with the continuum formed by a triplon and an even hole state . , @xmath124 , @xmath117 induced by the pc generator ( solid line ) and by the gs generator ( dashed line ) . ] the behavior of the rod yields further evidence why the gs results should be preferred to the pc results for this parameter regime . [ x=1 lambda=2 rod ] depicts the evolution of the rod during the flow for both generators . not only that the pc scut converges extremely slowly , the shape of the curve for @xmath146 is an indicator for a problem with respect to the sorting of the eigenenergies . it is a generic behavior of the rod that the sorting of the eigen values ( cf . ( [ sorting of the eigen values ] ) ) is reflected by features for small values of @xmath28 . when the sorting is completed the rod decreases exponentially with a constant rate . for the gs generator the decrease of the rod attains this rate not later than at @xmath147 . before this point the decrease is slower . before it decreases with a constant rate , the sorting usually does not pose a problem . ] the kink of the gs rod at @xmath148 with the subsequent upturn is again most probably due to cumulating numerical inaccuracies . this is no real problem because the gs rod has already fallen below a value of less than @xmath149 at @xmath148 and can hence be neglected . however , the pc rod exhibits several humps and a pronounced rise at @xmath150 before a decrease with a constant rate is achieved . this is a typical indication for a suppressed divergence that would actually occur for a less strict truncation . if such a behavior occurs the transformation is susceptible to truncation errors . the physical origin of these problems is the strong overlap between the one - hole - one - triplon continuua and the one - hole - two - triplon continua ( see fig . [ x=1 lambda=2 continuum ] ) . results from exact diagonalization are available for @xmath61 and @xmath151 , but even the gs generator does not induce convergence for this case . however , if we apply the generator restriction introduced in sect . [ chap : gen_rest ] to the gs generator , it yields convergence for @xmath152 . then some coupling between one hole and one hole plus one triplon @xmath153 is not eliminated completely . because the hamiltonian is not diagonalised with respect to the terms omitted from the generator , the hole dispersions we obtain from a fourier transformation are only upper limits for the actual result . but this is a minor problem since the remaining matrix elements are small . another way of improvement ( not followed here ) would be to use self - consistent born approximation to deal with the remaining coupling @xmath153 . , @xmath124 , @xmath117 . right panel : @xmath61 , @xmath124 , @xmath151 . the black lines represent the even hole dispersion and the cyan ( gray ) lines represent the odd hole dispersion . ] here we compare the results for upper limits from the restricted generator for @xmath61 and @xmath117 with the gs results before we investigate the results for @xmath61 and @xmath151 . the left panel of fig . [ truncated generator results ] shows the comparison of the one - hole dispersions for @xmath61 and @xmath117 . we see that for @xmath152 the upper boundary from the restricted gs generator is close to the result from the full gs generator . in the right panel of fig . [ truncated generator results ] the results from the restricted gs generator are compared to the exact diagonalization results for @xmath61 and @xmath151 . for the even hole dispersion the agreement between the result from the restricted gs generator and the exact diagonalization result is almost perfect . also the agreement for the odd hole dispersion is good .. ] the deviations are comparable to the deviations of the result by the full gs generator from the exact diagonalization result for @xmath61 and @xmath117 . the investigation of the rod shows that the restricted gs generator yields a faster convergence than the full gs generator for @xmath61 and @xmath117 ( see fig . [ truncated generator rod ] ) . for @xmath61 and @xmath151 the rod diverges for the full generator , while the restricted generator induces convergence ( see also fig . [ truncated generator rod ] ) . note that the kinks of the rods in fig . [ truncated generator rod ] with an increase afterwards are again probably due to a feedback of numerical inaccuracies . but all the kinks appear at values where the rod is already smaller than @xmath154 . so the flow can be considered to be converged at the kinks for practical purposes . the terms left out from the restricted generator still contribute to the hamiltonian after the transformation . these contributions yield an estimate of the difference between the actual energy and the upper boundary for the energy resulting from the restricted generator . for the @xmath155 terms of the form @xmath156 or @xmath157 the sum over their squared coefficients is @xmath158 ; its square root yields @xmath159 . rungs ( discrete points ) for @xmath61 , @xmath105 . the lower boundaries of the continua calculated by gs scut are shown as dashed lines . the results for the even hole state are black , the results for the odd hole state are cyan ( gray ) . ] to understand the deviations between scut and exact diagonalization we also compare our results to the complete spectrum from exact diagonalization . in fig . [ ed al x=1 lambda=1 ] the results of the exact diagonalization for various finite ladders with @xmath160 rungs are compared to the results from the gs generator for @xmath61 and @xmath105 . for the even band the lowest lying eigen value can be clearly distinguished from the larger eigen values , which are the precursor of the continuum . this holds true for all momenta . in the case of the odd band the lowest eigen value is very close to the higher ones for large momenta . our calculation of the continuum predicts that the dispersion merges with the continuum in this region . also the quantum monte carlo result for the spectral weight @xcite exhibits no peak below the continuum around @xmath137 . our result for the even band is in excellent agreement with the exact diagonalization result . around @xmath161 our result for the odd band lies above the exact diagonalization result while it lies below the exact diagonalization result around @xmath162 . rungs ( discrete points ) for @xmath61 , @xmath117 . the lower boundaries of the continua calculated by gs scut are shown as dashed lines . the results for the even hole state are black , the results for the odd hole state are cyan ( gray ) . ] in fig . [ ed al x=1 lambda=2 ] we compare the results of the exact diagonalization with the results from the gs generator for @xmath61 and @xmath117 . for the even band for large momenta the lowest lying eigen value is clearly distinguishable from the larger eigen values , which are again the precursor of the continuum . the same is true for the odd band for small momenta . but for small momenta in case of the even band and for large momenta in case of the odd band the lowest eigen value is very close to the higher ones . hence the distinction between hole dispersion and continuum becomes difficult in these regions . it is even questionable if they are actually distinct or if the dispersion merges with the continuum . in these regions the deviations between scut and exact diagonalization are the largest . we conclude that the scut suffers from truncation errors if the dispersion runs close to continua or even enters them . also the lower boundary of the continuum that we calculated from the gs result is higher at @xmath123 for the even band and at @xmath111 for the odd band than we would expect from the exact diagonalization data . ( solid lines ) and exact diagonalization for various numbers or rungs @xmath160 ( symbols ) for @xmath61 , @xmath151 . the dashed lines are estimates for the lower boundaries of the continua based on the result from the restricted gs generator . the results for the even hole state are black , the results for the odd hole state are cyan ( gray ) . ] for @xmath61 and @xmath151 the exact diagonalization results are shown in fig . [ ed al x=1 lambda=3 ] for various finite ladders with @xmath160 rungs . the odd hole dispersion is more difficult to distinguish from the continuum than for @xmath61 and @xmath117 . also for small momenta the lowest eigen value is very close to the higher ones . the hole dispersions from the restricted gs generator with @xmath163 is also depicted in fig . [ ed al x=1 lambda=3 ] . in the region @xmath164 this dispersion lies below the exact diagonalization data . , @xmath165 and various values for @xmath102 . ] , @xmath166 and various values for @xmath102 . ] , @xmath167 and various values for @xmath102 . ] since the ring exchange is needed for an adequate description of the experimentally available systems , we also investigate the influence of the ring exchange on the doped ladder , see ref . and references therein . a typical value is @xmath168 @xcite . as for the examination of the anisotropic hopping we take the reliable results for @xmath169 , @xmath165 ; @xmath122 , @xmath166 and @xmath61 , @xmath170 without ring exchange as starting point for our investigation . for the first case @xmath169 , @xmath165 the influence of the ring exchange is the weakest ( see fig . [ x=0 lambda=0.5 ring exchange ] ) . the odd band hardly changes . it is slightly lowered only the maximum is decreased more strongly . for the even band , however , the changes are more pronounced . it is strongly lowered so that a band crossing occurs and the shape changes . a local miminum appears at @xmath123 developing into the global minimum with growing @xmath102 . [ x=0.5 lambda=0.25 ring exchange ] depicts the case @xmath122 , @xmath166 . both bands are lowered stronger for small @xmath121 than for large @xmath121 . also the shape of both bands changes . while the minimum of the even band moves to @xmath171 for @xmath172 and a local maximum occurs at @xmath111 , the local maximum at @xmath123 of the odd band moves to @xmath145 and a local minimum at occurs @xmath123 . for @xmath167 and @xmath61 only the gs results are discussed because the pc generator yields no conclusive results for these parameters without ring exchange . the resulting one - hole dispersions are shown in fig . [ x=1 lambda=2 ring exchange ] . the increase of @xmath102 yields a lowered dispersion for both bands . the shape of the bands is conserved , but for the odd band the decrease is pronounced around @xmath111 . even an increase of the energy can be observed for the even band around @xmath111 . the aim of the present paper was to establish an approach for doped low - dimensional quantum antiferromagnets which yields an effective model for the motion of charges . to this end , we employed a self - similar continuous unitary transformation ( scut ) to lightly doped spin ladders . the method is particularly suited to provide effective models beyond numerical results for special quantities . spin ladders are a particularly well suited system because their magnetic excitations are very well understood @xcite . we are able to calculate the one - hole dispersions by means of scut . the agreement with results from series expansion is very good for small parameters . even in the regime @xmath61 , @xmath124 , @xmath173 the agreement with the exact diagonalisation results is still very good for the even band and satisfactory for the odd one . the hole dispersions are strongly influenced by the triplons . because the one - hole dispersion with odd parity has a larger local energy which is influenced more strongly by the higher lying continuum , it changes more explicitly with increasing hopping constants . therefore the deviations from the simple cosine shape are more pronounced for the odd band . these deviations grow if either the magnetic coupling or the strength of the hopping is increased : * the broadening of the odd band is slowing down , then turning into a narrowing . finally the shape changes completely under the influence of the second harmonic so that the maximum at @xmath123 is only local , the total maximum occurs at @xmath111 and the minimum lies between @xmath174 and @xmath175 . * the shift upwards is also slowing down and then turning into a shift downwards due to the influence of the continuum formed by one triplon and one even hole state . in the regime around @xmath61 and @xmath170 the average odd dispersion is approximately as low as the average even dispersion . for the even band the deviations from the cosine shaped dispersion consist essentially in the growth of the second harmonic . for @xmath61 and @xmath167 the minimum moves from @xmath111 into the region @xmath176 . a local maximum at @xmath111 occurs . but the absolute maximum remains at @xmath123 . the combination of these effects for @xmath61 and @xmath170 yields a crossing of the two dispersions . the crossing point lies between the minima of the bands . however , the pc generator is not applicable for the scut if the parameters are in this region . this is suggested by the deviations from the series expansion and from exact diagonalisation as well as by the peculiar convergence behavior of the scut . in the regime @xmath177 and @xmath178 the pc generator is no longer suitable because the convergence of the flow is hindered by the overlap between the one - hole - one - triplon continuua and the one - hole - two - triplon continua . the pc results deviate from the series expansion and exact diagonalisation results . furthermore the convergence is very slow and exhibits features that indicate problems in the sorting of the eigen values . the remedy is to use the gs generator which only decouples the zero - triplon subspace from the remaining hilbert space . then the hole dispersions are similar to the exact diagonalisation results . although there are small deviations from the series expansion , the agreement is astonishingly good . because the exact diagonalisation studies a ladder with fourteen rungs , finite size effects are present so that a part of the deviations stem from them . the remaining deviations are probably due to truncation errors . another aspect in favour of the gs generator is the satisfactory convergence behavior . the treatment of the case @xmath179 is problematic even for the gs generator because the convergence is hindered by the overlap between the odd one - hole dispersion and the continuum formed by one even hole and one triplon . to achieve convergence in this regime we have developed the following restriction for the gs generator . a term affecting the hole - triplon continuum is omitted , if the distance between the triplon and the hole state on which the term acts is larger than @xmath66 . strictly , the effective hamiltonian yields an upper boundary for the hole dispersions . the comparison of the results from the full gs generator and from the restricted gs generator for @xmath61 and @xmath167 shows that the upper boundary given by the result from the restricted generator is already very close to the result from the full generator for @xmath163 . for @xmath61 and @xmath180 this restriction induces convergence , while the flow diverges for @xmath181 . the estimates we obtain for the hole dispersions agree again well with the exact diagonalisation results . if @xmath182 and @xmath183 or @xmath184 the bands exhibit a similar shape and relative position to each other so that still a crossing at @xmath185 occurs . but the energy is lowered and the bandwith of both bands is increased by a factor of @xmath186 if @xmath103 is changed from 2 to 3 . the results from the restricted gs generator yield an estimate for the even band which is in good agreement with the exact diagonalisation result , while the odd hole dispersion exhibits deviations that are very similar to the deviations between the exact diagonalisation and the full gs generator for @xmath187 . the case @xmath188 is a special one . both the pc and the gs generator ( even with increased maximal extensions ) exhibit deviations from the series expansion results for the odd hole state . these deviations stem apparently from the closeness ( or even overlap ) of the continuum formed by one even hole and one triplon . if the lowering of the odd band is overestimated by the scut or underestimated by the series expansion is not clear . the ring exchange leads to a lowering of the hole dispersions . this effect is least pronounced for the even band around @xmath111 . the deformation of the band shape is most pronounced for the odd band . summarizing , the essential achievement of this paper is the development of a unitary transformation which allows us to disentangle the motion of the doped charges , the holes , and the magnetic excitations , triplons . thereby an effective model for the motion of the holes and the triplons has been derived . the challenge was to systematically derive this effective model even in the experimentally relevant regime @xmath189 where the hopping takes about three times the value of the magnetic exchange constants . this challenge could only be met by modification of the unitary transformation in the spirit of what was done in ref . . ladders doped with more than one hole have not yet been treated by the scut . to do so the approach presented here has to be extended to include also interaction terms between two holes . this step will be subject of future research because it will address the key question how large the attractive forces between two doped charges are . in view of the cooper pair formation in doped cuprate superconductors this is a very interesting issue . of course , numerical and diagrammatic results for bound states of two holes exist @xcite . but so far no effective model has been derived which incorporates the attractive potential explicitly . the fact that we could realize a systematic mapping of the initial doped spin ladder to an effective model for the motion of single excitations is encouraging for the next step incorporating the hole - hole interaction . the additionally required resources , for instance the increase in the number of coefficients , are significant . we estimate that about six times more terms have to be kept track of . still this should be realizable in the near future . we expect that a substantial gain in the understanding of cooper pair formation and thus of superconductivity in doped mott insulators will be possible . we are indebted to a.m. luchli for the provision of the exact diagonalization data shown in figs . [ ed al x=1 lambda=1 ] , [ ed al x=1 lambda=2 ] , and [ ed al x=1 lambda=3 ] . we thank kai p. schmidt , carsten raas , tim fischer , simone a. hamerla and nils drescher for fruitful discussions . the initial stage of this project was supported by the graduiertenkolleg `` materials and concepts for quantum information processing '' ( gk726 ) funded by the deutsche forschungsgemeinschaft . @xmath105 in case of the odd band at @xmath123 and at @xmath190 . the cyan ( gray ) circles depict the exact diagonalization results for finite ladders with various numbers of rungs @xmath160 . the solid black line is a linear extrapolation with respect to @xmath191 and the dashed line is an extrapolation based on exponential saturation ( see eq . ( [ exponential saturation ] ) ) . for comparison the result from the gs scut is also shown ( solid cyan / gray ) . ] to understand the deviations between scut and exact diagonalization in case of the odd band we also study the finite size scaling of the exact diagonalization . first we consider the exact diagonalization results for @xmath61 and @xmath105 shown in fig . [ ed al x=1 lambda=1 ] . around @xmath161 our result for the odd band lies above the exact diagonalization result , while it lies below the exact diagonalization result around @xmath162 . an investigation of the finite size scaling for the exact diagonalization data in these regions is difficult because we have two points at most for an extrapolation . thus we consider @xmath123 , where an extrapolation is conclusive . the result of this extrapolation is used to support an extrapolation in the region where the deviations are observed . the left panel of fig . [ finite size scaling x=1 lambda=1 ] shows the finite size scaling for @xmath123 using two kinds of extrapolation . the first is a simple linear extrapolation with respect to @xmath191 , while the second assumes an exponential saturation with increasing @xmath160 so that @xmath192 holds true for the difference @xmath193 from the limit for @xmath194 . the correlation length @xmath195 is determined to be @xmath196 by this extrapolation . the correlation length can be used to apply the second extrapolation also for @xmath190 where only two results are obtained by exact diagonalization . the finite size scaling for @xmath190 in case of the odd band is investigated by both extrapolations in the right panel of fig . [ finite size scaling x=1 lambda=1 ] . the extraploation results at @xmath123 are still close to the gs scut result and the extrapolation results at @xmath190 are in good agreement with the gs scut result . the linear extrapolation is even in excellent agreement with our result . note that the points from the @xmath197 calculation are omitted for the extrapolations because they deviate from the behavior of the other points due to the very small size . @xmath117 in case of the odd band at @xmath123 and at @xmath190 . the cyan ( gray ) circles depict the exact diagonalization results for finite ladders with various numbers of rungs @xmath160 . the solid black line is a linear extrapolation with respect to @xmath191 and the dashed line is an extrapolation based on exponential saturation ( see eq . ( [ exponential saturation ] ) ) . for comparison the result from the gs scut is also shown ( solid cyan / gray ) . ] for @xmath61 and @xmath117 the scut result for the dispersions comes close to the extrapolation of exact diagonalization results in the regions where the continuum is distinguishable from the hole dispersion . this is most obvious for the odd band at @xmath123 . the left panel of fig . [ finite size scaling x=1 lambda=2 ] shows the finite size scaling for this case using the same extrapolations like for @xmath61 and @xmath105 . we determine @xmath195 to be @xmath198 by the extrapolation based on exponential saturation . the correlation length is again used to apply this kind of extrapolation also for @xmath190 . the finite size scaling for @xmath190 in case of the odd band is investigated by both extrapolations in the right panel of fig . [ finite size scaling x=1 lambda=2 ] . for both momenta the linear extrapolation comes close to the scut result , but also the extrapolation based on exponential saturation does not exhibit large deviations from the scut result . the deviation for @xmath123 is @xmath199 and the deviation for @xmath190 is @xmath200 . the points from the @xmath197 calculation are omitted for the extrapolations because they deviate from the behavior of the other points due to the small size of the system . for @xmath61 and @xmath151 the tendency of the finite size scaling at @xmath190 indicates that the exact diagonalization results overestimate the energy of the odd band in this region so that one can expect that a proper finite size scaling yields a dispersion that lies completely below the upper boundary provided by the restricted scut energies . however , an extrapolation from the two points ( @xmath201 and @xmath202 ) at @xmath190 is not conclusive . we have several points for an extrapolation for @xmath123 and @xmath111 , but there the distinction between continuum and dispersion is difficult . because the gs scut diverges without restriction of the generator , we actually expect that a strong overlap of the energies of single holes and of the energies of hole plus triplon states is present .

we propose a simple prescription to calculate the solar neutrino survival probability in the quasivacuum oscillation ( qvo ) regime . such prescription is obtained by matching perturbative and exact analytical results , which effectively take into account the density distribution in the sun as provided by the standard solar model . the resulting analytical recipe for the calculation of is shown to reach its highest accuracy ( in the whole qvo range ) when the familiar prescription of choosing the solar density scale parameter at the mikheyev - smirnov - wolfenstein ( msw ) resonance point is replaced by a new one , namely , when is chosen at the point of `` maximal violation of adiabaticity '' ( mva ) along the neutrino trajectory in the sun . the mva prescription admits a smooth transition from the qvo regime to the msw transition one . we discuss in detail the phase acquired by neutrinos in the sun , and show that it might be of relevance for the studies of relatively short timescale variations of the fluxes of the solar lines in the future real - time solar neutrino experiments . finally , we elucidate the role of matter effects in the convective zone of the sun .

hep-ph0011306

the solar neutrino problem @xcite , emerging as a deficit of the observed solar neutrino rates @xcite with respect to standard solar model ( ssm ) predictions @xcite , can be explained through neutrino flavor oscillations @xcite , possibly affected by the presence of matter @xcite ( see @xcite for reviews of oscillation solutions and @xcite for recent analyses ) . the analysis of neutrino oscillations requires a detailed study of the evolution of the flavor states @xmath4 , @xmath5 being any linear combination of @xmath6 and @xmath7 , along the neutrino trajectory . in the simplest case , the active states @xmath4 are assumed to be superpositions , through a mixing angle @xmath8 $ ] , of two vacuum mass eigenstates @xmath9 , characterized by a mass squared gap @xmath10 ( see , e.g. , @xcite ) . the corresponding neutrino evolution equation in the flavor basis , @xmath11 involves then the following vacuum @xmath12 and matter @xmath13 terms in the hamiltonian @xmath14 , @xmath15 with the usual definitions for the neutrino wavenumber in vacuum , @xmath16 and for the neutrino potential in matter , @xmath17 @xmath18 being the electron density at the point of neutrino trajectory in the sun , located at radial distance @xmath19 from the sun center . the @xmath18 distribution is usually taken from ssm calculations @xcite . the hamiltonian @xmath14 is diagonalized in the matter eigenstate basis @xmath20 through the mixing angle in matter @xmath21 defined by @xmath22 @xmath23 being the neutrino wavenumber in matter , @xmath24 in this work , we focus on approximate solutions of eq . ( [ evol ] ) for neutrino flavor transitions at relatively small values of @xmath25 @xcite , characterizing the so - called quasivacuum oscillation ( qvo ) @xcite regime ev@xmath26/mev ) corresponds to the vacuum oscillation ( vo ) regime @xcite . in this work , we do not explicitly distinguish the vo range , but simply treat it as part of the qvo range . let us note that vo solutions of the solar @xmath3 problem are disfavored by the current data @xcite . ] , @xmath27 in order to define our goals more precisely , we recall that , both in the qvo regime and in the `` mikheyev - smirnov - wolfenstein '' ( msw ) @xcite regime , which takes place at @xmath28 ev@xmath26/mev , the solar @xmath29 survival probability at the earth surface can be written as @xmath30 where @xmath31 is the average probability , @xmath32 and @xmath33 is an oscillating term @xcite ( see also @xcite ) , @xmath34 the main ingredients of the above equations are : @xmath35 the mixing angle in matter @xmath36 at the @xmath29 production point @xmath37 , with @xmath38 in the qvo regime ; @xmath39 the level crossing ( or jump ) probability @xmath40 of the transition from @xmath41 at @xmath42 to @xmath43 at the surface of the sun ( @xmath44 ) ; and the phase @xmath45 accumulated on the @xmath3 path from @xmath37 to the earth surface , @xmath46 a.u . , @xcite , they can always be implemented through a modification of the expression for @xmath47 , see the discussion in @xcite . ] where the phases @xmath48 , @xmath49 , have a simple physical interpretation @xcite : @xmath50 $ ] , @xmath51 being the probability amplitude of the transition @xmath52 between the initial matter - eigenstate @xmath53 and the final vacuum - mass eigenstate @xmath54 . the phase @xmath55 can also be decomposed ( see , e.g. , appendix a ) as a sum of a `` solar '' phase @xmath56 acquired on the path in the sun ( @xmath57 $ ] ) and a `` vacuum '' phase @xmath58 , acquired in vacuum ( @xmath59 $ ] ) , @xmath60 as a consequence of eqs . ( [ p0p1])([csiv ] ) , solving the neutrino evolution equation ( [ evol ] ) basically reduces to calculating @xmath40 and @xmath56 . this task can be accomplished through numerical integration of eq . ( [ evol ] ) for @xmath57 $ ] , leading to `` exact '' solutions . however , suitable approximations to the exact results are also useful , both to speed up the calculations and to clarify the inherent physics . the starting point of such approximations is usually the analytical solution of eq . ( [ evol ] ) for the case of exponential density @xmath18 in the sun @xcite ( see also @xcite ) . deviations of the ssm density from the exponential profile can then be incorporated by an appropriate prescription for the choice of the scale height parameter @xmath61^{-1}\ , \label{r0x } % r0x\ ] ] characterizing the realistic change of the electron density along the neutrino trajectory in the sun . the exponential density analytical results are derived assuming that @xmath62 const . a well - known and widely used prescription for precision calculations of the jump probability @xmath40 in the msw regime is to use the `` running '' scale height parameter @xmath63 in eq . ( [ r0x ] ) , where for given @xmath25 and for @xmath64 , @xmath19 is chosen to coincide with the msw resonance point , @xmath65 @xcite . in the present work , we show how the analytical solution ( sec . ii ) and the `` running resonance '' prescription for @xmath2 ( sec . iii ) can be smoothly extended from the msw range down to the qvo range , and we give a justification for our procedure . the extension is first achieved by matching the resonance prescription to a perturbative expression for @xmath40 , in the limit of small @xmath66 ( sec . a more satisfactory match is then reached ( sec . iv ) by replacing the resonance prescription , @xmath67 , which can be implemented only for @xmath68 , with the `` maximal violation of adiabaticity '' ( mva ) prescription , @xmath69 , where for given @xmath25 and any given @xmath8 $ ] , @xmath70 is the point of the neutrino trajectory in the sun at which adiabaticity is maximally violated ( or , more precisely , the adiabaticity function has a minimum ) . we also show how the phase @xmath56 can be easily and accurately calculated in the qvo range through perturbative expressions at @xmath71 , and discuss the conditions under which @xmath56 might play a phenomenological role ( sec . v ) . in sec . vi it is shown that the perturbative results essentially probe the low - density , convective zone of the sun . our final prescriptions for the calculation of @xmath40 and @xmath56 , summarized in sec . vii , allow to calculate @xmath0 using eqs . ( [ p0p1])([p1 ] ) with an accuracy @xmath72 in the whole qvo range . all technical details are confined in appendixes a e . while this work was being completed , our attention was brought to the interesting work @xcite , where the qvo range is also investigated from a viewpoint that , although being generally different from ours , partially overlaps on the topic of adiabaticity violation . we have then inserted appropriate comments at the end of sec . in this section we recall the analytical expressions for @xmath40 and @xmath56 , valid for an exponential density profile , @xmath75 where @xmath76 and @xmath2 are derived from a best fit to the ssm profile @xcite ( see fig . 1 in @xcite ) , @xmath77 assuming @xmath18 as in eq . ( [ nexp ] ) , the neutrino evolution equation ( [ evol ] ) can be solved exactly in terms of confluent hypergeometric functions @xcite . the associated expression for @xmath40 can be simplified @xcite by making a zeroth order expansion of the confluent hypergeometric functions in the small parameter @xmath78 and by using the asymptotic series expansion in the inverse powers of the large parameter @xmath79 where @xmath80 ( bulk of the neutrino production zone ) . then one obtains the well - known `` double exponential '' form for @xmath40 @xcite , @xmath81 concerning the phase @xmath56 , to zeroth order in @xmath82 one gets a compact formula @xcite ( without using the asymptotic series expansion in @xmath83 ) , @xmath84+k(r_\odot - x_0)\ , \label{xianalyt } % xianalyt\end{aligned}\ ] ] where @xmath85 both eqs . ( [ pcanalyt],[xianalyt ] ) are valid at any @xmath86 , including @xmath87 . made at the beginning of @xcite , which is equivalent to take @xmath88 for the usual choice @xmath89 , was basically functional to obtain @xmath90 in a certain region of the parameter space ( as it was implied by the results of the homestake experiment ) . however , such restriction does not play any role in the derivation of the eqs . ( [ pcanalyt ] ) and ( [ xianalyt ] ) in @xcite , although this was not emphasized at the time . this has also been recently noticed in @xcite . ] let us note that the expression ( [ p0p1 ] ) for the probability @xmath0 with the average probability @xmath31 and the oscillating term @xmath33 given by eqs . ( [ p0])([csiv ] ) , ( [ pcanalyt ] ) and ( [ xianalyt ] ) was shown @xcite to assume the correct form in the vo , msw transition and `` large @xmath91 '' ( averaged oscillations ) regimes . in this section , we show how the resonance prescription for calculation of @xmath40 ( and @xmath0 ) , valid in the msw regime and in the first octant of @xmath86 , can be modified to obtain accurate values for @xmath40 also for @xmath92 and for @xmath93 . the resonance prescription in the msw range is based on the following approximations : @xmath35 oscillations are assumed ( and where shown in @xcite ) to be averaged out , so that effectively @xmath94 and @xmath56 becomes irrelevant @xcite . ] ; @xmath39 @xmath40 is taken from eq . ( [ pcanalyt ] ) , but with a variable scale height parameter @xmath95 [ eq . ( [ r0x ] ) ] , with @xmath19 `` running '' with the resonance ( res ) point @xmath96 @xcite , @xmath97 where @xmath98 is defined by the resonance condition @xcite @xmath99 when applicable . in the absence of resonance crossing , it was customary to take @xmath100 in the msw regime ( see , e.g. , @xcite ) . in particular , in the msw analysis of @xcite , @xmath100 was taken in the second octant of @xmath86 , where @xmath101 and eq . ( [ omegares ] ) is never satisfied . the indicated resonance prescription for the calculation of @xmath40 is known to work very well over at least three decades in @xmath25 ( @xmath102@xmath103 ev@xmath26/mev ) @xcite , with a typical accuracy of a few @xmath104 in @xmath0 . however , in the lowest msw decade ( @xmath105@xmath106 ev@xmath26/mev ) , such prescription is not very accurate both in the first octant @xcite ( where it tends to underestimate the effective value of @xmath2 ) and in the second octant @xcite ( where @xmath40 is small but not exactly zero ) . it was found numerically in @xcite that , for small @xmath66 , a relatively small and constant value of @xmath2 ( @xmath107 ) provided a better approximation to @xmath31 in the first octant were obtained in @xcite by using the density @xmath18 provided by the bahcall - ulrich 1988 standard solar model @xcite . ] ; this observation has been also discussed and extended to the second octant in @xcite ( where @xmath108 is used for @xmath109 ) . here we derive ( and improve ) such prescriptions by means of perturbation theory . the key result is worked out in appendix b , where we find a perturbative solution of the neutrino evolution equation ( [ evol ] ) , in the limit of small values @xmath66 , by treating the vacuum term @xmath110 as a perturbation of the dominant matter term @xmath111 in the hamiltonian [ eq . ( [ hsplit ] ) ] . the @xmath112 solution or @xmath113 . however , for simplicity , we use the notation @xmath114 , @xmath109 etc . , instead of @xmath115 , @xmath116 , etc . ] can be expressed in terms of the following dimensionless integral , @xmath117 where @xmath118 , and @xmath119 is given by eq . ( [ v ] ) . in particular , the effective value of @xmath2 ( in the limit of small @xmath66 ) is given by @xmath120 , namely , @xmath121 independently of @xmath86 ( i.e. , both in the first and in the second octant ) . the above perturbative results show that the asymptotic ( @xmath109 ) effective value for @xmath2 depends upon a well - defined integral over the density profile @xmath18 . using the ssm profile for @xmath18 @xcite , we find that @xmath122 the same value is obtained through exact numerical calculations . the appearance of integrals over the whole density profile indicates that , for small @xmath66 , the behavior of @xmath40 becomes nonlocal , as was also recently noticed in @xcite . we further elaborate upon the issue of nonlocality in sec . vi , where we show that the @xmath112 perturbative results are actually dominated by matter effects in the convective zone of the sun ( @xmath123 ) , where the function @xmath18 resembles a power law rather than an exponential . in order to match the usual resonance prescription [ @xmath124 with the value @xmath125 in the regime of small @xmath66 , we observe that , for the ssm density distribution @xcite it is @xmath126 at @xmath127 . thus , we are naturally led to the following `` modified resonance prescription , '' @xmath128 where `` otherwise '' includes cases with @xmath129 , for which @xmath98 is not defined . such a simple recipe provides a description of @xmath40 which is continuous in the mass - mixing parameters , and is reasonably accurate both in the qvo range ( @xmath130 ev@xmath26/mev ) and in the lowest msw decade ( @xmath131@xmath106 ev@xmath26/mev ) . figure 1 shows isolines of @xmath40 in the bilogarithmic plane charted by the variables interval somewhat beyond the qvo range , in order to display the smooth transition to the msw range . ] @xmath132 $ ] ev@xmath26/mev and @xmath133 $ ] . was introduced in @xcite to chart both octants of the solar @xmath3 mixing angle @xmath86 in logarithmic scale . ] the solid lines refer to the exact numerical calculation of @xmath40 , while the dotted lines are obtained through the analytical formula for @xmath40 [ eq . ( [ pcanalyt ] ) ] , supplemented with the modified resonance prescription [ eq . ( [ modres ] ) ] . also shown are , in the first octant , isolines of resonance radius for @xmath134 and 0.904 . the maximum difference between the exact ( numerical ) results and those obtained using the analytic expression for @xmath40 , eq . ( [ pcanalyt ] ) , amounts to @xmath135 , and is typically much smaller . since @xmath40 is not a directly measurable quantity , we propagate the results of fig . 1 to probability amplitudes observable at the earth , namely , to the average probability @xmath31 [ eq . ( [ p0 ] ) ] and to the prefactor of the oscillating term @xmath33 [ eq . ( [ p1 ] ) ] . is separately studied in sec v. ] figure 2 shows isolines of @xmath31 for ssm density , derived numerically ( solid lines ) and by using the analytic expression for @xmath31 and the modified resonance prescription ( dotted lines ) . the maximum difference is @xmath136 . figure 3 shows , analogously , isolines of @xmath137 . the difference amounts to @xmath138 . from figs . 13 , the modified resonance prescription for @xmath2 [ eq . ( [ modres ] ) ] emerges as a reasonable and remarkably simple approximation to the exact results for @xmath40 , valid in both the msw and the qvo regimes . however , it does not reproduce the exact behavior of @xmath40 with the requisite high precision of few % for @xmath139 ev@xmath26/mev and @xmath140 . this difference can be understood and removed , to a large extent , through the improved prescription discussed in the next section . in this section we generalize and improve eq . ( [ modres ] ) , by replacing the point of resonance @xmath141 with the point where adiabaticity is maximally violated ( @xmath70 ) , more precisely , where the adiabaticity function has its absolute minimum on the neutrino trajectory . let us recall that the validity of the resonance prescription @xmath142 is based on the fact that @xmath143 in a relatively shorth part of the @xmath3 trajectory , where the propagation is locally nonadiabatic . the resonance condition , however , can be fulfilled only in the first octant of @xmath86 . the most general condition for nonadiabaticy , as introduced already in the early papers @xcite on the subject , has instead no particular restriction in @xmath86 @xcite . such alternative condition can be expressed , in the basis @xmath20 relevant for the calculation of @xmath40 , in terms of the ratio between the diagonal term @xmath144 and the off - diagonal term ( @xmath145 ) in the ( traceless ) hamiltonian . more specifically , a transition is nonadiabatic if the ratio @xmath146 satisfies the inequality @xmath147 at least in one point of the neutrino trajectory in the sun . if @xmath148 is large along the whole trajectory , @xmath149 , the transition is adiabatic . the minimal value of @xmath148 identifies the point of `` maximum violation of adiabaticity , '' @xmath70 , @xmath150 we show in appendix b , that , along the solar @xmath3 trajectory , the mva point is uniquely defined , for any @xmath86 in both octants . in particular , such point can be unambiguously characterized through the condition @xmath151 in the first octant , in general , @xmath70 differs from @xmath98 , although one can have @xmath152 in some limiting cases ( see appendix b for a more general discussion ) . for instance , as eq . ( [ adfunction ] ) indicates , the two points practically coincide , @xmath153 , in the case of nonadiabtic transitions at small mixing angles ( @xmath154 ) . indeed , let us consider for illustration the simplified exponential case of @xmath155 . in this ( `` exp '' ) case it is easy to find from eq . ( [ adfunction ] ) that : @xmath156 obviously , at small mixing angles ( @xmath157 ) we have @xmath158 . however , this is no longer true for the nonadiabatic transitions at large mixing angles in the qvo regime we are interested in . in the latter case , as it follows from eq . ( [ xmvares ] ) , @xmath159 . we discuss in appendix b the more realistic case of ssm density . as far as the calculation of @xmath40 is concerned , it turns out that @xmath160 in the limits of small @xmath86 ( or of large @xmath66 ) . therefore , the mva condition smoothly extends the more familiar resonance condition in both directions of large mixing and of small @xmath66 , which are relevant to pass from the msw to the qvo regime . the inequality @xmath161 , derived for exponential density , persists in the qvo range for the realistic case of ssm density , implying that @xmath162 with @xmath63 defined as in eq . ( [ r0x ] ) . as a conseqence , a difference arises in the value of @xmath40 if @xmath98 is replaced by @xmath70 in the prescription for calculating @xmath2 . explicitly , our mva prescription reads @xmath163 figure 4 shows curves of iso-@xmath2 obtained through eq . ( [ modmva ] ) . the value of @xmath2 presents weak variations ( within a factor of two ) in the whole mass - mixing plane and , by construction , smoothly reaches the plateau @xmath125 for @xmath164 ev@xmath26/mev . figures 5 , 6 , and 7 are analogous to figs 1 , 2 and 3 , respectively , modulo the replacement of the resonance condition [ eq . ( [ modres ] ) ] with the mva condition [ eq . ( [ modmva ] ) ] . the mva prescription clearly improves the calculation of @xmath40 , @xmath31 , and @xmath137 , with an accuracy better than a few percent in the whole plane plotted : @xmath165 , @xmath166 , @xmath167 . we conclude that the analytical formula for @xmath40 [ eq . ( [ pcanalyt ] ) ] , valid for an exponential density , can be applied with good accuracy to the case of ssm density , provided that the scale height parameter @xmath2 is chosen according to the mva prescription , eq . ( [ modmva ] ) . a final remark is in order . we agree with the author of @xcite about the fact that , in order to understand better the behavior of @xmath40 in the qvo regime , the concept of adiabaticity violation _ on the whole neutrino trajectory _ is to be preferred to the concept of adiabaticity violation _ at the resonance point_. however , we do not share his pessimism about the possibility of using the running value @xmath168 for accurate calculations of @xmath40 : indeed , figs . 5 - 7 just demonstrate this possibility . such pessimism seems to originate from the observation that , as @xmath66 decreases , @xmath40 starts to get nonlocal contributions from points rather far from @xmath70 @xcite . in our formalism , this behavior shows up in the @xmath109 limit [ eq . ( [ r0limit ] ) ] , where , as mentioned in the previous section , the effective value of @xmath2 gets contributions from an extended portion of the density profile [ see also sec . vi and appendix c for further discussions ] . our prescription ( [ modmva ] ) , however , effectively takes this fact into account , by matching the `` local '' behavior of @xmath2 for large @xmath66 [ @xmath169 with the `` nonlocal '' behavior of @xmath2 at small @xmath66 [ @xmath125 ] . in conclusion , the mva prescription , appropriately modified [ eq . ( [ modmva ] ) ] to match the @xmath109 limit , allows a description of @xmath40 which is very accurate in the whole qvo range , and which smoothly matches the familiar resonance prescription up in the msw range . in this section , we discuss the last piece for the calculation of @xmath0 , namely , the solar phase @xmath56 . as we will see , this phase can significantly affect the quasivacuum oscillations of almost `` monochromatic '' solar neutrinos ( such as those associated to the @xmath171be and _ pep _ spectra ) . indeed , there might be favorable conditions in which the phase @xmath56 ( often negligible in current practical calculations ) could lead to observable effects and should thus be taken into account . first , let us observe that , in the qvo range , the size of the solar phase @xmath56 is of @xmath172 , as indicated by eq . ( [ xianalyt ] ) for exponential density @xcite , and also confirmed through numerical calculations for ssm density @xcite . figure 9 shows , in particular , exact results for the ratio @xmath173 , as a function of @xmath25 , using the ssm density . it appears that , neglecting @xmath56 with respect to the vacuum phase @xmath58 , is almost comparable to neglect @xmath174 as compared with @xmath175 . remarkably , there are cases in which corrections of @xmath176 are nonnegligible , e.g. , in the study of time variations of @xmath0 over short time scales @xcite , induced by the earth orbit eccentricity ( @xmath177 ) . in fact , the fractional monthly variation of @xmath178 from aphelion to perihelion is @xmath179 . real - time experiments aiming to detect time variations of the @xmath3 flux in monthly bins might thus test terms of @xmath180 , as also emphasized at the end of sec . vi in the work @xcite . secondly , let us recall that the oscillating term @xmath181 gets averaged to zero when the total phase @xmath55 is very large @xcite . the approximation @xmath182 holds in the msw regime , but it becomes increasingly inaccurate ( and is eventually not applicable ) as @xmath66 decreases down to the qvo regime . in order to understand when @xmath55 starts to be observable ( at least in principle ) one can consider an optimistic situation , namely , an ideal measurement of @xmath0 with a real - time detector having perfect energy resolution , and monitoring the flux from narrowest solar @xmath3 spectra ( the be and _ pep _ neutrino lines ) . in this case , the most important and basically unavoidable source of smearing is the energy integration @xcite over the @xmath3 lineshape . production zone is irrelevant in the qvo regime @xcite , see also appendix a. ] it has been shown in @xcite ( see also @xcite ) that such integrations effectively suppresses the oscillating term @xmath33 at the earth through a damping factor @xmath183 , calculable in terms of the @xmath3 lineshape . figure 8 shows the factor @xmath183 for the _ pep _ line and the for two be lines , characterized by average energies @xmath184 , 863.1 , and 385.5 kev , respectively . the be and _ pep _ lineshapes [ having @xmath185 kev widths ] have been taken from @xcite and @xcite , respectively . figure 8 proves that @xmath55 is observable , at least in principle , in the whole qv range @xmath130 ev@xmath26/mev , as far as the narrowest be line is considered . of course , the observability of @xmath55 becomes more critical ( or even impossible ) by increasing the initial energy spread ( e.g. , for continuous @xmath3 spectra ) or by performing measurements ( like in current experiments ) with additional and substantial detection smearing in the energy or time domain @xcite . with the above caveats in mind , we set out to describe accurately @xmath56 in the whole qvo range . this task is accomplished in appendixes d and e where , by means of the same perturbative method applied earlier to @xmath40 , we obtain the @xmath112 and @xmath71 expressions for @xmath56 , respectively . the final perturbative result is @xmath186 which is in excellent agreement with the exact result for @xmath56 shown in fig . 9 , in the whole qvo range . the @xmath112 coefficient in eq . ( [ csikk ] ) is just the real part of the integral @xmath187 in eq . ( [ ietafirst ] ) ( see appendix d ) and is already sufficient for an accurate description of @xmath56 in the qvo range . remarkably , the right magnitude of this coefficient can also be obtained from the analytical expression ( [ xianalyt ] ) , which would give @xmath188 ( see appendix d ) . we also keep the small @xmath71 term in eq . ( [ csikk ] ) , because it neatly shows that @xmath56 starts to become @xmath86-dependent for increasing values of @xmath25 , consistently with the exact numerical results of fig . figure 10 shows the error one makes on the phase @xmath56 , by using increasingly better approximations , for two representative values of @xmath189 . the error is given as the absolute difference between approximate and exact results for @xmath56 , in units of a period ( @xmath190 ) . the lowest possible approximation is simply to neglect @xmath174 with respect to @xmath175 , so that @xmath191 and @xmath192 , namely , neutrino oscillations are effectively started at the sun center ( `` empty sun '' ) . a better approximation is to start oscillations at the sun edge , @xmath193 and @xmath194 , by assuming that the high sun density damps oscillations up to the surface @xcite ( `` superdense sun '' ) . such two approximations ( and especially the first ) are widely used in phenomenological analyses . figure 10 shows , however , that they can produce a considerable phase shift ( even larger than a full period for the case of `` empty '' sun ) in the upper qvo range . the fact that the ssm density is neither @xmath195 nor @xmath196 is correctly taken into account through the @xmath112 term in eq . ( [ csikk ] ) , and even more accurately through the full @xmath71 expression for @xmath56 in eq . ( [ csikk ] ) , as evident from fig . 10 . the errors estimated in fig . 10 show that different approximations for @xmath56 can affect calculations for future experiments . in particular , the use of the familiar `` empty sun '' or `` superdense sun '' approximations can possibly generate fake phase shifts in time variation analyses for real - time detectors sensitive to neutrino lines , such as borexino @xcite or kamland @xcite . according to the estimates in @xcite , such two experiments might be sensitive to monthly - binned seasonal variations for @xmath197 . figure 10 shows that , in the upper part of such sensitivity range , the empty sun approximation gives @xmath56 ( and thus @xmath55 ) totally out of phase , as compared with exact results . analogously , the superdense sun approximation induces a phase shift that , although smaller than in the previous case , can still be as large as @xmath198 @xmath199 in the quoted sensitivity range , and can thus produce a big difference [ @xmath200 in the calculation of @xmath33 . it is an unfortunate circumstance , however , that the dominant term in @xmath56 is proportional to @xmath66 . this fact implies that , neglecting @xmath201 in the total phase @xmath202 , is basically equivalent to introduce a small bias of the kind @xmath203 . on one hand , such bias is sufficient to produce a substantial difference in @xmath181 when @xmath204 approaches @xmath205 ev@xmath26/mev , and is thus observable in principle . on the other hand , @xmath91 is not known a priori , but must be derived from the experiments themselves , and it will be hardly known with a precision of @xmath206 for some time . therefore , although @xmath56 may produce a big effect at _ fixed _ values of @xmath91 , it might be practically obscured by uncertainties in the _ fitted _ value of @xmath91 . in conclusion , we have found a simple and accurate expression for the solar phase @xmath56 , to be used in the calculation of the total phase @xmath207 . the solar phase @xmath56 produces effects of @xmath176 , which can be nonnegligible in high - statistics , real - time experiments sensitive to short - time variations associated to neutrino lines , such as borexino and kamland , as was also emphasized in @xcite and more recently in @xcite . in such context , we recommend the use of eq . ( [ csikk ] ) for precise calculations of @xmath0 at fixed value of @xmath91 , although the observability of @xmath56 certainly represents a formidable challenge . in this section we elaborate upon the @xmath112 perturbative results discussed previously for @xmath40 and @xmath56 ( and detailed in appendix c and d ) . in particular , we show that they are related to the density in the convective zone of the sun , corresponding to @xmath208 . the @xmath112 results crucially depend upon the quantity @xmath187 , which is defined in eq . ( [ ietafirst ] ) as an integral over an oscillating function , having the density @xmath18 as inner argument . numerical evaluation of such integral for ssm density gives @xmath209 ( see appendix c ) . it turns out that the largest contribution to @xmath187 comes from the outer regions of the sun , where the integrand oscillates slowly , while in the inner regions the integrand oscillates rapidly , with vanishing net contribution to the real and imaginary parts of @xmath187 . numerical inspection shows that the value of @xmath187 is dominated by the @xmath210 range , which happens to correspond to the convective zone of the sun @xcite . therefore , it is sufficient to consider such zone to estimate @xmath187 . in the convective zone , the density profile @xmath18 is better described by a power law rather than by an exponential function ( see , e.g. , @xcite ) . a good approximation to the ssm density for @xmath211 is : @xmath212 with @xmath213 and @xmath214 mol/@xmath215 . by adopting such expression for the density , and using the position @xmath216 , we get the following expression for @xmath187 , @xmath217 where @xmath218 and @xmath219 . once again , we note that the above integrand gives a very small contribution outside the convective zone @xmath220 , since the exponent becomes large . therefore , the upper limit can be shifted from @xmath221 to @xmath222 without appreciable numerical changes , and with the advantage that the result can be cast in a compact analytical form , reproducing the exact ssm numerical result with good accuracy : @xmath223 using the above equation and the results of appendixes c and d , the small-@xmath66 limit for @xmath2 and @xmath56 can be explicited as @xmath224 such results show that the effective values of @xmath2 and of @xmath56 at small @xmath66 are connected to the parameters @xmath225 describing the power - law dependence of @xmath18 in the convective zone of the sun [ eq . ( [ np ] ) ] . therefore , while for relatively large @xmath66 neutrino oscillations in matter probe the inner `` exponential bulk '' of the solar density profile , for small @xmath66 they mainly probe the outer , `` power - law '' zone of convection . finally , let us notice that , in the range @xmath226 , where the `` running '' mva prescription @xmath227 is applicable , the power law in eq . ( [ np ] ) leads to @xmath228 , so that @xmath229 at the matching point @xmath230 , consistently with the prescription in eq . ( [ modmva ] ) . we summarize our best recipe for calculation of @xmath0 as follows . in the qvo regime , for any given value of the mixing angle @xmath86 and of the neutrino wavenumber @xmath232 , the @xmath29 survival probability reads @xmath233 where @xmath234 the value of @xmath40 can be evaluated with high precision by using the analytical form @xmath235 provided that the density scale parameter @xmath2 is calculated as @xmath236 where @xmath70 is the point where adiabaticity is maximally violated along the neutrino trajectory in the sun ( see appendix b ) . such expression for @xmath40 smoothly joins the more familiar resonance prescription when passing from the qvo to the msw regime . the value of the sun phase @xmath56 can be calculated with high precision by using the @xmath71 perturbative result , valid for @xmath237 ev@xmath26/mev , @xmath238 we have shown that such phase can play a role in precise calculations of @xmath0 in the qvo range . it is not necessary to extend the calculation of @xmath56 in the msw range , where effectively @xmath239 , and @xmath56 is not observable even in principle . notice that the neutrino production point @xmath37 does not appear in the calculation of both @xmath40 and @xmath56 in the qvo range . finally , we have checked that the above recipe allows the calculation of @xmath0 with an accuracy @xmath240 ( and often much better than a percent ) in both octants of @xmath86 for the whole qvo range ( @xmath241 ev@xmath26/mev ) . figure 11 shows , as an example , a graphical comparison with the exact results for @xmath0 at the exit from the sun ( @xmath44 ) . at the earth ( @xmath242 ) can not be usefully plotted in the range of fig . 11 , due to its rapidly oscillating behavior . ] it appears at glance that our recipe represents an accurate substitute to exact numerical calculations of @xmath0 , in the whole qvo range @xmath243 ev@xmath26/mev . we have worked out a simple and accurate prescription to calculate the solar neutrino survival probability @xmath0 in the quasivacuum oscillation regime . such prescription adapts the known analytical solution for the exponential case to the true density case , as well as to the perturbative solution of neutrino evolution equations in the limit of small @xmath25 . the accuracy of the prescription is significantly improved ( up to @xmath244 in @xmath0 in the whole qvo range ) by replacing the familiar prescription of choosing the scale height parameter @xmath2 at the msw resonance point by a new one : @xmath2 is chosen at the point of maximal violation of adiabaticity ( mva ) along the neutrino trajectory in the sun . such generalization preserves a smooth transition of our prescription from the qvo to the msw oscillation regime , where the two prescriptions practically coincide . we show that at sufficiently small @xmath245 in the qvo regime , the effective value of @xmath2 is determined by an integral over the electron density distribution in the sun , @xmath18 : @xmath246 where @xmath118 . the main contribution in the above integral is shown to come from the region corresponding to the convective zone of the sun , @xmath247 . thus , if quasivacuum oscillations take place , solar neutrino experiments might provide information about the density distribution in the convective zone of the sun . we also discuss in detail the phase acquired by neutrinos in the sun , whose observability , although possible in principle , poses a formidable challenge for future experiments aiming to observe short timescale variations of fluxes from solar @xmath3 lines . a.m. acknowledges kind hospitality at sissa during the initial stage of this work . s.t.p . would like to acknowledge the hospitality of the aspen center for physics where part of this work was done . the work of e.l . , a.m. , and a.p.was supported in part by the italian murst under the program `` fisica astroparticellare . '' the work of s.t.p . was supported in part by the eec grant erbfmrxct960090 and by the italian murst under the program `` fisica teorica delle interazioni fondamentali . '' in this appendix we recall the derivation of the basic equations for @xmath0 given in the introduction , together with further details relevant for appendixes b e . given an initial solar @xmath29 state @xmath248 ( @xmath249 being the transpose ) , its final survival amplitude @xmath250 at the detector can be factorized as @xmath251 where the @xmath252 matrices act as follows ( from right to left ) : ( i ) @xmath253 rotates ( at @xmath42 ) the initial state @xmath254 into the matter mass basis ; ( ii ) @xmath255 is a generic @xmath256 parametrization for the evolution of the matter mass eigenstates @xmath20 from the origin ( @xmath42 ) up to the exit from the sun @xmath257 where @xmath258 , in terms of the so - called crossing probability @xmath259 and of two generic phases @xmath260 and @xmath261 ; @xmath262 @xmath263 evolves the mass eigenstates in vacuum along one astronomical distance @xmath175 , with @xmath58 defined in eq . ( [ csiv ] ) ; and @xmath264 @xmath265 finally rotates the mass basis back to the flavor basis at the detection point ( @xmath242 ) . , since they turn out to be nonnegligible only in the msw regime , when oscillations are averaged out and @xmath94 , see @xcite and references therein . ] after some algebra , the resulting @xmath29 survival probability can be expressed as the sum of an `` average term '' @xmath31 plus four `` oscillating terms '' @xcite , @xmath266 where @xmath267 \ , \label{ap3}\\ % ap3 p_4 & = & \frac{1}{2}\sin2\omega_m^0\sin2\omega\cos(\xi_v-2\alpha ) \ .\label{ap4 } % ap4\end{aligned}\ ] ] it has been shown in @xcite that the last three terms @xmath268 can , in general , be safely neglected for practical purposes , so that one can take @xmath269 which , together with the identification @xmath270 leads to eqs . ( [ p0p1])([p1 ] ) . phase notation of @xcite through the following identifications : @xmath271 ; @xmath272 ; @xmath273 ; and @xmath274 . the notation in the present work explicitly factorizes out the contribution of the vacuum phase @xmath58 for @xmath275 $ ] . ] in the qvo regime relevant for our work ( @xmath276 ev@xmath26/mev ) , the negligibility of @xmath268 is evident from the fact that @xmath277 for @xmath278 , so that the @xmath268 prefactor @xmath279 is negligibly small , while the nonvanishing terms @xmath280 can be written as : @xmath281 it was shown in @xcite that the above equations are not spoiled by earth matter effects , ) and ( [ p1qv ] ) in this work coincide with eq . ( 28 ) in @xcite , modulo the identification @xmath282 , valid in the qvo regime . ] as they turn out to be negligible in the qvo range a fortunate circumstance that considerably simplifies the calculations . we have verified the applicability of the approximations in eq . ( [ p0qv ] ) and ( [ p1qv ] ) in the qvo regime , by checking that our exact results do not change in any appreciable way by setting @xmath283 from the start . in particular , the results of the numerical integration of the evolution equations vary very little by forcing the initial state to be @xmath41 rather than @xmath29 . we have also verified that none of our figures changes in a graphically perceptible way in the qv range , by moving the point @xmath37 within the production zone @xmath284 not only radially but also for off - center trajectories .- independence of @xmath285 implies that no smearing over the production zone is necessary , as also emphasized in @xcite . ] therefore , in the following appendixes , we will neglect corrections of @xmath286 in the qvo regime , and just set @xmath287 from the start , without any appreciable loss of accuracy . given eq . ( [ aee ] ) , and the fact that the initial state can be taken , to a high degree of accuracy , @xmath288 , the probability amplitudes to find the two mass eigenstates at @xmath44 ( @xmath289 ) simply read @xmath290 in terms of flavor state transition amplitudes @xmath250 and @xmath291 at @xmath44 , the jump probability reads @xmath292 and the phase reads ) and ( [ csiqv ] ) can also be obtained from eqs . ( 13a)(13d ) , ( 15 ) , and ( 19 ) in @xcite . ] @xmath293\ . \label{csiqv } % csiqv\end{aligned}\ ] ] we conclude this appendix by quoting useful expressions for some dimensionless quantities which appear in the calculation of @xmath40 and @xmath55 : @xmath294 in this appendix we characterize the condition of maximum violation of adiabaticity along the neutrino trajectory . we also show that the mva condition reduces to the familiar resonance condition in appropriate limits . in the matter eigenstate basis , the neutrino evolution equation @xmath295 is governed by the hamiltonian @xmath296 where @xmath297 . the evolution is adiabatic when the diagonal term dominates over the off - diagonal term , so that the ratio of such elements [ the function @xmath148 defined in eq . ( [ adfunction ] ) ] is large . the `` maximal violation of adiabaticity '' along the neutrino trajectory is reached at the point @xmath298 where @xmath148 is instead minimized , @xmath299 the above mva condition can be rewritten as @xmath300 since @xmath301 increases monotonically from its value at the production point ( @xmath302 ) to its vacuum value @xmath303 , the above condition characterizes the ( flex ) point of fastest increase for the function @xmath304 , which is thus uniquely defined at any @xmath8 $ ] . the mva condition ( [ mva2 ] ) can be compared with the usual resonance condition , @xmath305 which can be fulfilled only for @xmath306 $ ] . to understand under which circumstances the two conditions practically coincide ( @xmath307 ) , it is useful to rewrite the mva condition in the following , equivalent form : @xmath308 the above condition reduces to eq . ( [ res ] ) in the case linear density @xmath309 ( which , however , is not applicable to the sun ) . it also reduces to eq . ( [ res ] ) for very small mixing ( @xmath310 ) , which explains why the simple resonance condition has been very frequently used in the literature , to replace the slightly more complicated mva condition . notice also that , for increasing @xmath66 , @xmath40 becomes nonnegligible only at small mixing angles . therefore , as far as the calculation of @xmath40 is concerned , the mva and resonance conditions also merge at large @xmath66 . in all other cases , it turns out that @xmath311 for @xmath312 $ ] . for @xmath313 $ ] , @xmath98 is simply not defined . we conclude by providing a handy approximation to the value of @xmath70 , which may be used to by - pass the calculation of derivatives of @xmath119 , at the price of a small loss of accuracy in the calculation of @xmath0 . as discussed before , the mva condition characterizes the flex point of the curve @xmath314 , which increases from @xmath315 to @xmath316 in the qvo range . we have verified that , in the parameter region where @xmath40 is nonnegligible , the flex point @xmath70 is close to the half - rise point @xmath317 where @xmath318 assumes the average value between its two extrema , defined as @xmath319 the above definition leads to the condition @xmath320 which clearly reduces to the resonance condition in the small @xmath86 limit . using @xmath321 as a substitute for @xmath168 in the mva prescription [ eq . ( [ modmva ] ) ] provides a final accuracy of @xmath322 in @xmath0 in the whole qv regime ( only slightly worse than the accuracy quoted at the end of sec . in this appendix we study the limit @xmath109 for @xmath40 , in both cases of exponential and ssm density . such limit gives an asymptotic value for the effective scale parameter @xmath2 , relevant for the lower part of the qv regime . the positions in eqs . ( [ x00 ] ) and ( [ pi/2 ] ) are adopted . we also introduce a normalized radial coordinate , @xmath324 in the case of exponential density , eq . ( [ pcanalyt ] ) gives , at first order in @xmath66 , @xmath325 where @xmath326 [ eq . ( [ r0exp ] ) ] . in the case of ssm density , we demonstrate that eq . ( [ pck ] ) is formally preserved , modulo the replacement @xmath327 the crucial observation is that , for @xmath328 ev@xmath26/mev and ssm density , we have @xmath329 for most of the @xmath3 trajectory inside the sun ( @xmath330 ) before the rapid density drop at the border . this suggest a perturbative solution of eqs . ( [ evol ] ) and ( [ hv+hm ] ) with the vacuum term @xmath110 playing the role of a small perturbation , as compared with the dominant matter term @xmath111 . the perturbative expansion is expected to be accurate when the second term in eq . ( [ pck ] ) is small , namely , for @xmath331 . at 0th order in @xmath110 , the evolution operator from @xmath332 to @xmath333 is trivial in the flavor basis ( @xmath111 being diagonal ) , @xmath334 & 0 \\ 0 & \exp\left[+\frac{i}{2}\eta(1,0)\right ] \end{array}\right)\ , \label{t0 ' } % t0'\end{aligned}\ ] ] where @xmath335 at 1st order in @xmath110 ( i.e. , in @xmath66 ) , one gets an improved evolution operator @xmath336 , @xmath337\ , \label{t1 } % t1\ ] ] where the correction matrix @xmath338 reads @xmath339 by applying @xmath336 to the initial flavor state @xmath340 , one gets the flavor state at @xmath44 , @xmath341\ , \label{flavor } % flavor\ ] ] where the @xmath112 correction is explicited . the basic integral , @xmath342 = c+is\ , \label{ieta } % ieta\ ] ] appearing in eq . ( [ flavor ] ) , can be numerically evaluated using the ssm density , @xmath343 finally , using eqs . ( [ pcqv],[flavor ] ) and keeping only @xmath112 terms , one gets @xmath344 which reproduces eq . ( [ pck ] ) , up to the anticipated replacement @xmath345 we have independently checked the value @xmath346 by numerical studies of the @xmath109 limit for the quantity @xmath347 , with @xmath40 derived exactly rather than perturbatively . ( namely , @xmath348 ) , by fitting his exact results for small @xmath66 . in this appendix we have given an explicit derivation of @xmath349 , based on perturbation theory , and independent on numerical fits . ] we observe that , for the ssm density , the scale parameter @xmath63 happens to be equal to the value in eq . ( [ r0 ] ) at @xmath350 , @xmath351 such relation proves useful to match our analytical and perturbative results around the critical zone of the `` knee '' of the ssm density , which occurs just at @xmath352 . as a final exercise , we prove that @xmath353 in the exponential limit @xmath354 , as it should be . the proof uses the assumptions of large initial density and small final density , as discussed for eqs . ( [ small ] ) and ( [ large ] ) . then one changes the integration variable from @xmath355 to @xmath356 in the calculation of @xmath357 [ eq . ( [ s ] ) ] , and exploits the fact that @xmath358 is very small ( large ) for @xmath333 ( @xmath332 ) : @xmath359 in this appendix we study the small-@xmath66 limit of the phase @xmath56 , by using the same strategy adopted in the previous section for @xmath40 . in particular , let us consider the @xmath112 expansion of the analytical phase in eq . ( [ xianalyt ] ) , @xmath360 where is the euler constant . ] @xmath361 = 0.116\ . \label{c ' } % c'\ ] ] we show that eq . ( [ csik ] ) , valid for exponential density , remains formally unchanged also for the ssm density , modulo the replacement @xmath362 , where @xmath363 is given by eq . ( [ c ] ) . in fact , using eqs . ( [ csiqv ] ) and ( [ flavor ] ) , and keeping only the @xmath112 terms , we get @xmath364 such result is independently confirmed by studying the @xmath109 limit of the quantity @xmath173 , with @xmath56 derived from exact ( rather than perturbative ) calculations . finally , we show that @xmath365 in the limit of exponential density , as it should be . with manipulations similar to those used to derive eq . ( [ boh ] ) we get : @xmath366\\ & = & 1-\frac{r_0}{r_\odot}\int^{r_0 v(0)}_0\frac{1-\cos\phi}{\phi } \label{diverg}\\ % diverg & \simeq & 1-\frac{r_0}{r_\odot}[\gamma_e+\ln r_0v(0 ) ] = c'\end{aligned}\ ] ] where , in the divergent integral of eq . ( [ diverg ] ) , we have kept only the two leading terms [ @xmath367 and @xmath368 which do not decrease when the upper limit @xmath369 becomes large . in this appendix , we present an @xmath71 perturbative expression for @xmath56 , which effectively accounts for the @xmath86-dependence of the exact results ( as evident in the upper part of the qvo range in fig . 9 ) . we basically iterate the calculation of the evolution operator @xmath370 ( see appendix c ) at second order in the perturbation term @xmath110 . we omit the ( somewhat lengthy ) derivation and present the final @xmath71 result , @xmath371 \label{csi2k}\\ % csi2k & = & 0.130\ , ( k r_\odot ) + 1.67 \times 10^{-3 } ( k r_\odot)^2\,\cos 2\omega\ , \label{csi2k ' } % csi2k'\end{aligned}\ ] ] where @xmath372 is defined in eq . ( [ ieta ] ) , while @xmath373 and @xmath374 are defined as @xmath375 \\ & = & ( 12.45 + i\ , 7.07)\times 10^{-2}\ , \label{ieta'}\\ i''_\eta & = & \int^1_0 d\rho \int^\rho_0 d\rho ' \ , \exp[i\,\eta(1,\rho ' ) ] \\ & = & ( 0.58 + i\ , 1.25)\times 10^{-2}\ , \label{ieta''}\end{aligned}\ ] ] and have been numerically evaluated for ssm density . the @xmath112 term reproduces , of course , the results given in the previous section . the @xmath71 behavior of eq . ( [ csi2k ] ) is not captured by the analytical phase written in the form of eq . ( [ xianalyt ] ) , which contains only odd powers in a @xmath66-expansion . the reason can be traced to the fact that eq . ( [ xianalyt ] ) was derived in @xcite under the assumption of zero density at @xmath44 ( which is not exactly fulfilled for exponential density ) . preliminary studies indicate that removal of such assumption gives the correct @xmath71 behavior of the analytical phase [ including the @xmath376 dependence displayed in eq . ( [ csi2k ] ) ] . such studies are proceeding and will be presented after completion . b. pontecorvo , zh . eksp . 53 * , 1717 ( 1967 ) [ sov . jetp * 26 * , 984 ( 1968 ) ] ; z. maki , m. nakagawa , and s. sakata , prog . * 28 * , 675 ( 1962 ) ; v. gribov and b. pontecorvo , phys . b * 28 * , 493 ( 1969 ) . talks by d. montanino and by a. palazzo at _ now 2000 _ , 2nd europhysics neutrino oscillation workshop ( conca specchiulla , italy , 2000 ) , to appear in the proceedings . transparencies available at http://www.ba.infn.it/@xmath377now2000 see , e.g. , v. barger , k. whisnant , and r.j.n . phillips , phys . d * 24 * , 538 ( 1981 ) ; s.l . glashow and l.m . krauss , phys . b * 190 * , 199 ( 1987 ) ; a. acker , s. pakvasa , and j. pantaleone , phys . d 43 , 1754 ( 1991 ) ; p.i . krastev and s.t . petcov , phys . b * 285 * , 85 ( 1992 ) and _ ibid . _ b * 299 * , 99 ( 1993 ) ; n. hata and p. langacker , phys . d * 56 * , 6107 ( 1997 ) ; b. fad , g.l . fogli , e. lisi , and d. montanino , astropart . phys . * 10 * , 93 ( 1999 ) . mikheyev and a.yu . smirnov , in _ moriond 87 _ , proceedings of the 7th moriond workshop on new and exotic phenomena ( les arcs , france , 1987 ) , edited by o. fackler and j. trn thanh vn ( editions frontires , gif - sur - yvette , france , 1987 ) , p. 405 . a. messiah , in _ moriond 86 _ , proceedings of the 6th moriond workshop on massive neutrinos in astrophysics and in particle physics ( tignes , france , 1986 ) , edited by o. fackler and j. trn thanh vn ( editions frontires , gif - sur - yvette , france , 1986 ) , p. 373 . fig . 1 . the crossing probability @xmath40 , calculated for the ssm density in the parameter space @xmath378 . the three lowest decades in @xmath25 characterize the quasivacuum oscillation regime . solid curves : exact numerical results . dotted curves : approximate results , using the analytical resonance prescription . also shown are isolines of resonance radii . the radius @xmath379 represents the matching point of the resonance prescription with the perturbative results valid for small values of @xmath380 . fig . 2 . the average survival probability @xmath31 , as obtained through exact numerical calculations ( solid curves ) and through the analytical resonance prescription ( dotted curves ) . the oscillating term prefactor @xmath137 , as obtained through exact numerical calculations ( solid curves ) and through the analytical resonance prescription ( dotted curves ) . isolines of the effective density scale parameter @xmath2 ( in units of @xmath174 ) , as derived through the prescription of maximum violation of adiabaticity ( mva ) , matched with the perturbative result valid at small values of @xmath25 ( @xmath381 ) . fig . 5 . the crossing probability @xmath40 , as obtained through exact numerical calculations ( solid curves ) and through the analytical mva prescription ( dotted curves ) . also shown are isolines of mva radii . the radius @xmath379 represents the matching point of the mva prescription with the perturbative results valid at small values of @xmath25 . fig . 6 . the average survival probability @xmath31 , as obtained through exact numerical calculations ( solid curves ) and through the analytical mva prescription ( dotted curves ) . fig . 7 . the oscillating term prefactor @xmath137 , as obtained through exact numerical calculations ( solid curves ) and through the analytical mva prescription ( dotted curves ) . damping factors for the oscillating term @xmath181 , as calculated for the be and _ pep _ solar neutrino lines ( @xmath382 being their average energy ) . the damping factors completely suppress oscillations at the earth for @xmath383 above the quasivacuum range . fig . 9 . the solar phase @xmath56 , in units of @xmath384 , as obtained from exact numerical calculations for some representative values of @xmath86 . notice the @xmath86-independent limit for @xmath385 . fig . 10 . the absolute error induced by approximate calculations of the solar phase @xmath56 , as compared with the exact numerical calculations , in units of @xmath190 . see the text for details . fig . 11 . application of our final analytical recipe for @xmath40 and @xmath56 to the calculation of @xmath386 at the exit from the sun ( dotted lines ) , as compared with the corresponding exact numerical results ( solid lines ) .

we study branching random walks in random environment on the-dimensional square lattice , . in this model , the environment has finite range dependence , and the population size can not decrease . we prove limit theorems ( laws of large numbers ) for the set of lattice sites which are visited up to a large time as well as for the local size of the population . the limiting shape of this set is compact and convex , and the local size is given by a concave growth exponent . also , we obtain the law of large numbers for the logarithm of the total number of particles in the process . + * short title : * branching random walks in random environment + * keywords : * shape theorem , subadditive ergodic theorem , transience , growth exponent , population size + * ams 2000 subject classifications : * primary 60k37 ; secondary 60j80 , 82d30 universit paris 7 , ufr de mathmatiques , case 7012 , 2 , place jussieu , f75251 paris cedex 05 , france e - mail : ` comets@math.jussieu.fr ` , url : ` http://www.proba.jussieu.fr/\simcomets ` instituto de matemtica e estatstica , universidade de so paulo , rua do mato 1010 , cep 05508090 , so paulo sp , brasil e - mail : ` popov@ime.usp.br ` , url : ` http://www.ime.usp.br/\simpopov `

0709.2926

we start with an informal description of the model we study in this paper . particles live in @xmath4 and evolve in discrete time . at each time , every particle is substituted by ( possibly more than one ) offspring which are placed in neighboring sites , independently of the other particles . the rules of offspring generation depend only on the location of the particle . the collection of those rules ( so - called the _ environment _ ) is itself random , it is chosen randomly before starting the process , and then it is kept fixed during all the subsequent evolution of the particle system . this model considered in this paper was introduced in @xcite . the random environment here affects both branching and transition mechanisms , and ( as opposed to the models of @xcite ) the _ immediate _ descendants of a particle are not supposed to be independent . in @xcite we proved a dichotomy for recurrence / transience , depending only on the support of the environmental law , and we gave sufficient conditions for recurrence and transience . in the recurrent case , the tails of the hitting times are studied and a shape theorem is obtained in a preliminary form . the recurrence / transience classification was later completed in @xcite . we refer also to @xcite for other work related to branching random walks in random environment . now , let us look at the subject of this paper from a different angle . for @xmath5 , let @xmath6 be the transition probabilities from @xmath7 to its nearest neighbors , and @xmath8 . consistently with the notation introduced later in this section , we denote by @xmath9 the collection of coefficients @xmath10 ( as explained below , @xmath11 stands for the mean offspring in @xmath7 ) , and by @xmath12 the corresponding discrete markov operator , @xmath13.\ ] ] in this paper we will study for @xmath14 the solution @xmath15 of the equation @xmath16 it is easily checked , for instance by the discrete feynman - kac formula , that the solution @xmath17 is given by the expectation @xmath18 of the number @xmath19 of particles in @xmath20 at time @xmath21 in a discrete - time branching random walk starting from a single individual located at site @xmath7 at time 0 . the evolution rule of this branching random walk is that particles at @xmath7 branch with an average of @xmath11 children which then move independently to a neighboring site randomly chosen from @xmath6 . we will be interested in the case where the coefficients @xmath22 are given by a stationary and finitely dependent random field . the model has other possible formulations . in the case of continuous time , the above equation becomes the parabolic partial differential equation @xmath23 with @xmath24 the branching rate . in our case the mean number of offspring is greater than or equal to @xmath25 and bounded from above , or , equivalently , @xmath24 is nonnegative and bounded . in the case where @xmath6 are the simple random walk transition probabilities @xmath26 , so that @xmath27 is the standard laplace operator , this equation is known as the parabolic anderson problem , and has also continuous - space versions , see @xcite and @xcite . these models have motivated a huge scientific activity , with particular interest on localization and intermittency ( e.g. @xcite ) and survival analysis @xcite , leading to fine pictures in the different cases of bounded or unbounded @xmath28 s . we stress that the markov operator @xmath12 is random in the present paper , a case that seems not to have been studied so far . moreover , @xmath12 is non symmetric , this makes the model non - reversible with a possibly large drift , and prevents the use of a spectral theory as in the above references . the present paper represents a step towards the understanding of the equation ( [ eq : discrete_anderson ] ) . we will prove that the solution @xmath17 behaves like @xmath29 as @xmath30 ( see remark [ rem_complete_shape ] where we comment on the function @xmath31 ) . we will even get into the finer study of the population size @xmath32 itself , and we prove the same asymptotics under the assumption that particles can not die . hence , the growth of the population is rather smooth at the coarse scale @xmath33 . the occurrence of shape theorems and local growth rate goes back at least to @xcite and @xcite for branching random walks , and to @xcite in random environment . in fact , our model is slightly more general than described above . as mentioned in the beginning of this section , the branching and the displacement may be dependent . also , the environment that we consider is not necessarily independent , but we allow for a finite range dependence . we believe that the results extend to more general dependence , but considering this case would require an important additional technical work . * the model . * we now describe the model , keeping the notations of @xcite whenever possible . let @xmath34 and @xmath35-s be the coordinate vectors of @xmath4 . we use the notation @xmath36 for @xmath37 ( or @xmath38 ) . define the distance between two sets @xmath39 ( or @xmath40 ) by @xmath41 fix a finite set @xmath42 such that @xmath43 for all @xmath44 . define @xmath45 and for @xmath46 put @xmath47 ; note that @xmath48 for all @xmath46 . furthermore , let @xmath49 be the set of all probability measures @xmath50 on @xmath51 : @xmath52 then , suppose that @xmath53 is a stationary ergodic random field , and denote by @xmath54 the probability and expectation with respect to @xmath22 . throughout this paper we suppose that this field is also _ finitely dependent _ , that is , the following condition holds : * condition i. * there exists a positive number @xmath55 such that for any two sets @xmath40 with @xmath56 the following holds : the sigma - algebra generated by @xmath57 is independent under @xmath58 from the sigma - algebra generated by @xmath59 . note that @xmath60 corresponds to the case of independent identically distributed environment . the collection @xmath61 is called _ the environment_. given the environment @xmath22 , the evolution of the process is described in the following way : start with one particle at some fixed site of @xmath4 . at each integer time the particles branch independently using the following mechanism : for a particle at site @xmath38 , a random element @xmath62 is chosen with probability @xmath63 , and then the particle is substituted by @xmath64 particles in @xmath65 for all @xmath66 . note that the population never gets extinct , since every individual has at least one direct descendant . for the process starting from one particle at @xmath7 , let us denote by @xmath67 the number of particles in @xmath68 at time @xmath21 . define the random variable @xmath69 i.e. , @xmath70 is the total number of particles at moment @xmath21 for the process starting from @xmath7 . we denote by @xmath71 the ( so - called quenched " ) probability and expectation for the process starting from @xmath7 in the fixed environment @xmath22 . we use the notation @xmath72 = { { \mathbb e}}\,{{\mathtt p}_{\omega}}^x[\,\cdot\,]$ ] for the annealed law of the branching random walk in random environment , and @xmath73 for the corresponding expectation . also , sometimes we use the symbols @xmath74 without the corresponding superscripts when it can create no confusion ( e.g. when the starting point of the process is indicated elsewhere ) . throughout this paper , and often without recalling it explicitly , we suppose that the two conditions below are fulfilled : * condition b. * @xmath75>0.\ ] ] * condition ue . * for some @xmath76 , @xmath77 = 1.\ ] ] condition b ensures that the model can not be reduced to random walk without branching , and condition ue is a natural uniform ellipticity condition which ensures that the walk is really @xmath0-dimensional . in this paper , the weaker ellipticity condition e of @xcite will usually not be enough for our purposes . in fact , we believe that most of our results do not generally hold if one only assumes condition e. for technical reasons we need also the following two conditions : * condition d. * there exists a positive constant @xmath78 such that the expectation of the total number of the immediate descendants of any particle is at most @xmath78 , i.e. , @xmath79=1 $ ] . * condition a. * there exist @xmath80 with @xmath81 even and @xmath82 such that @xmath83>0 $ ] . we refer to condition a as the `` aperiodicity condition '' because , without it , the process starting from the origin would live on even sites at even times , and on odd sites at odd times . if ( with @xmath84 of condition a ) a site @xmath68 is such that @xmath85 , we say that this site is an _ aperiodic _ site , and when @xmath86 , we say that this site is @xmath87-aperiodic . we briefly mention that suitably adjusted versions of all our results are also true without condition a ; the proofs are either the same or even simpler , since we do not have to care about searching for the aperiodic sites , e.g. the proofs of theorems [ t_shape_reach ] and [ t_complete_shape ] . [ rem : usuel ] a particular case of the model considered here is the usual construction of the branching random walk , that was already mentioned in the beginning of this paper : for each @xmath7 , specify the transition probabilities @xmath88 , @xmath66 , and branching probabilities @xmath89 , @xmath90 . a particle in @xmath7 is first substituted by @xmath91 particles with probability @xmath89 , then each of the offspring jumps independently to @xmath65 with probability @xmath88 . the pairs @xmath92 are chosen according to some i.i.d . field on @xmath4 . in our notations , @xmath93 is a mixture of multinomial distributions on @xmath94 : @xmath95 * statement of the results . * all through the paper , we will assume the above five conditions . following @xcite , we define the notions of transience and recurrence : [ def_rec ] for the particular realization of the random environment @xmath22 , the branching random walk is called recurrent if @xmath96=1.\ ] ] otherwise , the branching random walk is called transient . the following result is a consequence of propositions 1.2 , 1.3 , and theorem 1.6 of @xcite , and also theorem 3.2 of @xcite ; in section [ s_ind_rw ] , we comment about the validity of ( i ) and ( ii ) below in the non - i.i.d . case @xmath97 . [ rec / tr ] we have either : * for @xmath58-almost all @xmath22 , the branching random walk is recurrent , in which case @xmath98=1 $ ] for all @xmath38 , or : * for @xmath58-almost all @xmath22 , the branching random walk is transient , in which case @xmath98=0 $ ] for all @xmath38 . the classification criterion : in the case of an i.i.d . environment , i.e. , @xmath60 , the branching random walk in random environment is transient if and only if there exist @xmath99 and @xmath100 such that @xmath58-almost surely we have @xmath101 where @xmath102 is the mean number of particles sent to @xmath66 by one particle at the origin . it is interesting to observe that , in the case of an i.i.d . environment , one has a simple explicit criterion of transience / recurrence for the branching random walk in random environment ; however , for the many - dimensional random walk _ without branching _ in random environment the problem of finding such a criterion is still far from being solved . in @xcite one can find more evidence that branching makes random walks in random environment `` simpler '' ; see , for instance , the results about the tails of first hitting times . now , we are ready to formulate the main results of this paper . in what follows , for any @xmath103 , @xmath104 is the closure of @xmath105 , and @xmath106 is the interior of @xmath105 . first , we obtain a shape result about the sites that _ can _ contain a particle at time @xmath21 . [ t_shape_reach ] there exists a compact convex set @xmath107 such that for any @xmath108 , for almost all @xmath22 there exists @xmath109 such that @xmath110 = 1\ ] ] and @xmath111 < 1 \qquad \text { for all } x\in{{\mathbb z}}^d\cap ( 1-{\varepsilon})mf\ ] ] for all @xmath112 . next , we characterize the ( quenched ) mean local number of particles . [ t_local_size ] there exists a convex downwards , continuous function @xmath113 such that for any closed @xmath114 @xmath115 as @xmath116 . the function @xmath31 is called the local growth exponent for the branching random walk . denote @xmath117 ; by the convexity of @xmath31 , the set @xmath118 is compact and convex . all through , convex functions will mean convex downwards . [ t_rec_trans ] the branching random walk in random environment is recurrent if and only if @xmath119 . note that the border case @xmath120 is transient . note also that , from the property ( [ sup_beta>0 ] ) below , one obtains that @xmath119 if and only if @xmath121 . thus , given @xmath118 , one can determine whether the branching random walk is recurrent or transient . the next result does not only tell us , similarly to theorem 1.10 of @xcite , where the particles are located at time @xmath21 , but it also gives an important information about the local size of the population . [ t_complete_shape ] for any closed @xmath122 , @xmath123 as @xmath116 , a.s . [ rem_complete_shape ] ( equivalence of models . ) the local growth exponent @xmath31 is defined by ( [ shape_exp ] ) . by definition it only depends on the quenched expectation of the number of particles , and further , only on the mean number @xmath124 of particles sent in one step to @xmath68 by one particle at the origin . indeed , @xmath125 where @xmath126 is the environment shifted by @xmath127 , and where the sum ranges over all sequences @xmath128 with @xmath129 . in particular , for any mapping @xmath130 from @xmath49 to itself such that @xmath131 for all @xmath132 , the two branching random walks in the environments @xmath22 and @xmath133 are equivalent , in the sense that they have the same local behavior at the logarithmic scale . fine details of the branching and displacement do not matter , under the above five conditions . moreover , for results that only concern the expected number of particles ( such as theorem [ t_local_size ] ) we really do not need the assumption that any particle produces at least one offspring ; this can be substituted by a weaker assumption @xmath134 @xmath58-a.s . finally , we formulate a result about the total size of the population : [ t_total_size ] the function @xmath31 has the property @xmath135 for the total size of the population @xmath136 , it holds that @xmath137 a.s . , as @xmath116 . note that , by ( [ sup_beta>0 ] ) , @xmath138 , and @xmath139 by convexity . from theorems [ t_local_size ] and [ t_complete_shape ] it follows that for @xmath58-almost all @xmath22 , for any @xmath108 @xmath140&=1.\end{aligned}\ ] ] as opposed to theorem 1.10 of @xcite , here we prove this result both in transient and recurrent cases . [ ex1 ] * ( constant branching ) * more information can be obtained in the particular case when there exists a constant @xmath141 such that @xmath142 in this case , the expected size of the population is @xmath143 . this is the case for instance when @xmath144 does not depend on @xmath50 , and then the branching random walk is equivalent to a tree - indexed markov chain ( since in this case we can suppose also that the immediate descendants jump independently , and the offspring distribution does not depend on the site , cf . remark [ rem_complete_shape ] ) . in the general case , we can define the transitions @xmath145 ( recall that @xmath146 only depends on @xmath147 ) , and consider the random walk in random environment @xmath148 with transition probabilities @xmath149=p^{\omega_x}(y)$ ] for @xmath150 . we see here that @xmath151 , and therefore , by theorem [ t_local_size ] , @xmath152\big]\ ] ] with @xmath153 $ ] the integer part of @xmath154 ( coordinatewise ) . the limit can be expressed in terms of the quenched large deviation rate function @xmath155 , which have been studied in the nestling case in @xcite and in complete generality in @xcite : it holds that @xmath156 these references are for i.i.d . environment and estimate the probability of sets in the scale @xmath21 instead of the probability of points , but one can see that they apply to our discussion here . this example shows that the convex function @xmath31 is not necessarily strictly convex . indeed , it is known that the rate function has a flat horizontal part if the random walk in random environment is nestling with a non - zero speed @xmath157 , in which case @xmath158 for @xmath159 in the whole interval with endpoints @xmath160 and @xmath157 . finally , it is straightforward to see from the nearest neighbor jumps case , that the shapes @xmath161 and @xmath118 may have `` facets '' , i.e. , flat parts on their boundaries ( cf . e.g. example 7 of @xcite ) . the rest of this paper is organised as follows . first , in section [ s_ind_rw ] we recall some concepts of @xcite , such as recurrent seeds and induced random walks . in section [ s_proof_shape_reach ] we study the set of sites which can be reached up to time @xmath21 ( this amounts , basically , to applying the subadditive ergodic theorem ) . in section [ s_proof_t_local_size ] we prove the results related to the _ expected _ local population size , and in section [ s_proof_t_complete_shape ] we study the local population size itself . finally , in section [ s_proof_exp_growth ] we prove the equation ( [ sup_beta>0 ] ) , i.e. , that the expectation of the total number of particles grows exponentially ( somewhat surprisingly , this is one of the most difficult results of this paper ) , and then we prove theorem [ t_total_size ] . to begin , we introduce some more basic notations . we denote by @xmath162 the set of rational numbers , and define @xmath163 . let @xmath164 to be the maximal jump length , i.e. , @xmath165 and let @xmath166 be the @xmath0-dimensional discrete ball with respect to the @xmath167-norm : @xmath168 as in section 2.1 of @xcite , we define now the notion of _ induced _ random walk in random environment associated with the branching random walk in random environment . defining @xmath169 we consider some probability measure @xmath170 on @xmath171 with marginal @xmath58 on @xmath172 . a stationary random field @xmath173 with the law @xmath170 defines our branching random walk as above , coupled with a random walk in random environment with transition probability @xmath174 from @xmath7 to @xmath65 . in words , we pick randomly one of the children in the branching random walk . to keep things simple , we will drop the tilde from the notations @xmath170 . in this paper , we need only the so - called uniform induced random walk , for which the measure @xmath175 is defined as follows : @xmath175 is uniform on the locations @xmath176 . an important idea that will be repeatedly employed in this paper is to use the uniform ellipticity of the walk in order to reveal some independence in the environment : because of condition ue , the uniform induced random walk is uniformly elliptic as well , and so sometimes it makes its steps `` without looking at the environment '' . a similar construction can be found in @xcite . specifically , let us consider the uniform induced random walk @xmath177 ( @xmath20 stands for the starting location of this random walk ) . according to ( [ prob_ind_rw ] ) , the transition probabilities for @xmath177 are : @xmath178 = \sum_{v : v_y\geq 1}\frac{\omega_x(v)}{|\{u : v_u\geq 1\}|},\ ] ] which means that , in the case when a particle has more than one offspring in the branching random walk ( i.e. , it produces a configuration @xmath157 with @xmath179 ) , the next ( relative ) location for the uniform induced random walk is chosen uniformly among the locations @xmath176 . by condition ue , this induced random walk is uniformly elliptic in the sense that @xmath180\geq { \hat{\varepsilon}}_0 \qquad \text{for all } e\in\{\pm e_i , i=1,\ldots , d\}\ ] ] where @xmath181 . let @xmath182 be a sequence of i.i.d . random variables with values in @xmath183 , such that @xmath184={\hat{\varepsilon}}_0 $ ] for @xmath185 and of course @xmath186= 1 - 2d{\hat{\varepsilon}}_0 $ ] . we still keep the symbol @xmath187 to denote the probability on the enlarged probability space where both the branching random walk and the sequence @xmath188 are defined . set @xmath189 . let us enumerate the elements of the set @xmath190 in some order , so that @xmath191 . we can now construct the induced random walk as follows : [ def_seed ] fix a finite set @xmath201 containing 0 , and @xmath202 with @xmath203>0 $ ] . with @xmath204 , the pair @xmath200 is called a seed . we say that @xmath22 has a @xmath200-seed at @xmath14 ( or that a @xmath200-seed occurs in @xmath20 ) if @xmath205 and that @xmath22 has a @xmath200-seed in the case @xmath206 . we call @xmath20 the center of the seed . as in @xcite , it is easy to see that with probability @xmath25 the branching random walk visits infinitely many distinct @xmath200-seeds ( this can be done by showing that the uniform induced random walk does so ) . we now give the argument . with @xmath207 , at any time a.s . there exist subsequent times @xmath208 such that at time @xmath209 the uniform induced random walk is situated in a location @xmath7 which is at distance @xmath210 away from its range up to time @xmath21 , without looking at the environment ( that is , @xmath211 ) . then , by condition i , the environment can be constructed inside the translate @xmath212 independently from all what done before , and so the probability to generate a @xmath200-seed at site @xmath7 is a positive constant . by the borel - cantelli lemma , with probability 1 an infinite number of @xmath200-seeds will be visited . then , still following @xcite , we define the branching random walk _ restricted _ to set @xmath213 simply by discarding all particles that step outside @xmath214 , and write @xmath215 for corresponding probability and expectation . next , we consider a shortened version of definition 2.5 from @xcite : [ rec_seed ] let @xmath216 be a finite subset of @xmath4 with @xmath217 . let @xmath218 be a probability distribution on @xmath219 with mean larger than 1 , i.e. , @xmath220 with @xmath221 , @xmath222 , @xmath223 . an @xmath200-seed is called @xmath218-recurrent if for any @xmath22 such that @xmath224 , we have @xmath225 \geq \sum_{j = i}^\infty p_j\ ] ] for all @xmath226 . by `` _ free _ '' particles we mean that none is the descendant of another one . then , it is straightforward to see that all the discussion of section 2.2 of @xcite readily extends to the case of finitely dependent environment as well . in particular , the recurrence is equivalent to the existence of recurrent seeds ; this fact will be used several times in this paper . for arbitrary @xmath227 , @xmath228 $ ] ( @xmath229 is from condition ue ) , define @xmath230 so that @xmath231 is the minimal number of steps necessary for a particle in @xmath7 to send an offspring to @xmath68 , with the condition also that this should happen with big enough probability on each step . by condition ue , it is immediate that @xmath232 clearly , this family of random variables has the subadditive property : for any @xmath233 , @xmath234 $ ] , and _ any _ @xmath22 @xmath235 consider any @xmath236 and define @xmath237 where @xmath238 is the smallest positive integer such that @xmath239 . with ( [ t<||_1 ] ) and ( [ t_subadd ] ) , the subadditive ergodic theorem ( see e.g. theorem 2.6 of chapter vi of @xcite ) shows that the ( nonrandom ) limit in ( [ eq_def_mu ] ) exists a.s . and in @xmath240 ; by ( [ t_subadd ] ) , this limit verifies @xmath241 , @xmath242 , for any @xmath243 . moreover , since the jumps are bounded , we have @xmath244 and finally @xmath245 . then , by continuity one can define @xmath246 for any @xmath247 in such a way that @xmath248 is a norm on @xmath249 . let @xmath250 clearly , for any @xmath228 $ ] , the set @xmath251 is compact and convex , and @xmath252 . by definition , for any @xmath253 and @xmath22 it holds that @xmath254 when @xmath255 . so , we have that @xmath256 for @xmath257 . for the rest of this paper , denote @xmath258 ( and this is the compact convex set we are looking for in theorem [ t_shape_reach ] ) . the following lemma shows that the family @xmath251 is continuous in @xmath87 : _ proof : _ by covering @xmath161 with finitely many small disks , we see it is enough to prove that , for any @xmath261 , @xmath262 = 0,\ ] ] where @xmath238 is the smallest positive integer such that @xmath239 . observe that , by the subadditive ergodic theorem , the left - hand side of ( [ eq_cont_delta ] ) is equal to @xmath263.\ ] ] note that , by ( [ t<||_1 ] ) , @xmath231 depends only on a finite piece of the environment , so @xmath264 as @xmath265 . now , fix an arbitrary @xmath108 and choose @xmath266 in such a way that @xmath267 by ( [ monot_delta ] ) and ( [ shodimost_t ] ) , there exists @xmath268 such that @xmath269 for all @xmath270 . so , we obtain that @xmath271 \leq 2{\varepsilon},\ ] ] which implies ( [ eq_cont_delta ] ) . by a standard argument ( see e.g. @xcite ) one can show that , for any @xmath108 , @xmath275 for all @xmath21 large enough . in particular , since @xmath276 for all @xmath21 large enough , the first claim of theorem [ t_shape_reach ] follows . in order to prove the second claim , let us define by @xmath277 the set of sites that can be reached in _ exactly _ @xmath21 steps ( with our usual restriction on the probabilities of the steps ) . clearly , if @xmath278 , then @xmath279 . denoting also @xmath280^d$ ] , we intend to prove that , for any @xmath108 and almost all @xmath22 @xmath281 for all @xmath21 large enough . let ( recall the definition of @xmath282 from ( [ def_kn ] ) ) @xmath283 since the random environment is finitely dependent , one obtains that there are some positive constants @xmath284 such that for all @xmath285 @xmath286 \geq 1-c_1n^d\exp(-c_2 n^{-d/2}).\ ] ] by borel - cantelli lemma , @xmath287=1.\ ] ] 1 . @xmath291 is even . then it is trivial to obtain that @xmath292 ( one can complete the path of length @xmath293 which ends in @xmath7 by @xmath294 ) . @xmath291 is odd . suppose also that @xmath21 is so large that @xmath295 . since @xmath290 occurs , there exists an aperiodic site @xmath68 such that @xmath296 . then , to complete the path of length @xmath293 which ends in @xmath7 , essentially one goes from @xmath7 to @xmath68 in exactly @xmath297 steps , then jumps from @xmath68 to some @xmath298 with @xmath299 even , then goes back to @xmath7 ( and then , if necessary , one puts @xmath294 to the end of the path ) . [ l_def_beta ] for any @xmath300 , @xmath301 , the following quantity @xmath302 is well - defined and is a.s . constant : @xmath303 where @xmath238 is the smallest positive even integer number such that @xmath304 . _ proof : _ the expected number of particles @xmath305 has a supermultiplicative property : for any @xmath233 , @xmath306 @xmath307 so the family of random variables @xmath308 is superadditive . note , however , that the random variables of the latter family may assume the value @xmath309 . suppose first that @xmath310 . then , from condition ue we obtain that @xmath311 and so , taking condition d into account , the existence of the limit in ( [ eq_def_beta ] ) immediately follows from the subadditive ergodic theorem . however , if one wants to apply the subadditive ergodic theorem to the family @xmath312 in the case @xmath313 , there is the following difficulty : there may be some @xmath314 , such that with positive probability it happens that @xmath315 , which means that @xmath316 ( even @xmath317>0 $ ] ) . so , for the case @xmath313 we need a different approach . for the rest of the proof of lemma [ l_def_beta ] we suppose that @xmath313 and let @xmath87 be such that @xmath318 . define @xmath319 ( in principle , @xmath302 could depend also on @xmath22 , but in the next few lines we will show that it is a.s . constant ) . by ( [ incl_rn ] ) , @xmath58-a.s . there exists @xmath320 such that ( recall ( [ def_rn ] ) ) @xmath321 for all @xmath322 . using ( [ supermult ] ) , we obtain that for all @xmath322 @xmath323 for all @xmath324 such that @xmath325 , which means that @xmath326 for all @xmath322 . since the sequence @xmath327 is stationary ergodic , this shows that the upper limit in ( [ def_beta_a ] ) is a.s . constant . now , our goal is to prove that @xmath328 choose @xmath329 in such a way that @xmath330 ; let @xmath331 be the smallest positive integer such that @xmath332 . fix a small @xmath333 . by ( [ incl_rn ] ) , for all @xmath21 large enough we have @xmath334 . recall that , if @xmath278 , then @xmath335 , so @xmath336 now , we divide the integer interval @xmath339 into some subintervals and some singletons using the following algorithm . begin with @xmath340 ; inductively , let @xmath341 be the smallest integer not yet assigned . if the event @xmath342 occurs , then there exists @xmath343 such that @xmath344 . in this case we add the interval @xmath345 to our collection ( and then pass to @xmath346 ) . on the other hand , if the event @xmath347 occurs , then we declare @xmath341 to be a singleton ( and then pass to @xmath348 ) . as a result of this procedure , we obtain @xmath349 intervals @xmath350 , @xmath351 , and @xmath352 singletons @xmath353 . for each of the above intervals , we have @xmath354 , so that @xmath355 @xmath351 . then , condition ue implies that @xmath356 for any @xmath324 . so , abbreviating @xmath357 , @xmath358 , @xmath359 , we obtain from ( [ supermult ] ) , ( [ pervyj_shag ] ) , ( [ good_interval ] ) , and ( [ reach_small_steps ] ) that @xmath360 by construction of the intervals , we have @xmath361 so , by birkhoff s theorem @xmath362)(1-k_1r\alpha);\ ] ] then , @xmath363(1-k_1r\alpha).\ ] ] take @xmath364 small so that @xmath365 . note that @xmath366 , so , by ( [ oc_phi2 ] ) and ( [ prob_h_to_0 ] ) , @xmath367\big ) < 1-k_1\alpha\ ] ] if @xmath368 is so large that @xmath369 \leq k_1(r-1)\alpha ( 1-k_1r\alpha)^{-1 } ( \|a\|-1)^{-1}$ ] . thus , using ( [ interm_1 ] ) , we obtain @xmath370 the inequality ( [ liminf_beta_a ] ) now follows from ( [ prob_h_to_0 ] ) , ( [ oc_phi1 ] ) ( note also that , trivially , @xmath371 ) , and ( [ oc_phi2 ] ) . letting @xmath372 , this concludes the proof of lemma [ l_def_beta ] . _ proof : _ consider @xmath375 such that @xmath376 , and note that there exists @xmath259 such that @xmath377 . now , we have to prove that for any @xmath378 , @xmath379 let @xmath380 , @xmath381 , and @xmath382 . use the abbreviation @xmath383 note that , by ( [ incl_rn ] ) , lemma [ l_def_beta ] , and the bounded convergence theorem , for any @xmath384 it holds that ( with @xmath385 ; we use the convention @xmath386 ) @xmath387 from the supermultiplicative property ( [ supermult ] ) we obtain @xmath388 since @xmath389 taking expectations in ( [ eq_conv_proof ] ) and applying ( [ bounded_conv ] ) , we obtain ( [ eq_conv ] ) ( note also the following elementary fact : if @xmath390 a.s . , then @xmath391 $ ] ) . _ proof of theorem [ t_local_size ] : _ consider a closed set @xmath114 and fix any @xmath108 . there exists @xmath259 such that @xmath392 . clearly , for any small enough @xmath393 there exist @xmath394 such that @xmath395 let @xmath396 , and let @xmath397 , @xmath398 . then , by lemma [ l_def_beta ] , @xmath399 for all @xmath21 large enough , @xmath398 . using ( [ covering- ] ) , we obtain that for any @xmath400 there exists @xmath91 such that @xmath401 . then , analogously to the proof of the second claim of theorem [ t_shape_reach ] , we can show that on the event @xmath402 @xmath403 for all @xmath21 large enough ( note that @xmath404 ) . now , from ( [ supermult ] ) , ( [ snizu ] ) , and ( [ ostatok ] ) we obtain that @xmath405 so , from the uniform continuity of @xmath31 in @xmath406 ( cf . lemma [ l_convex_beta ] ) we obtain that @xmath407 suppose that there exists @xmath400 such that @xmath411 by ( [ covering+ ] ) , there exists @xmath91 such that @xmath412 . then , on the one hand , lemma [ l_def_beta ] implies that @xmath413 for all @xmath21 large enough , and , on the other hand , @xmath414 this leads to a contradiction when @xmath415 is small enough , and thus we obtain ( [ limsup_local_size ] ) . the proof of theorem [ t_local_size ] is completed . _ proof of theorem [ t_rec_trans ] : _ first , note that recurrence implies the existence of @xmath416-recurrent seeds ( cf . lemma 3.1 of @xcite and definition [ rec_seed ] above ) . such seeds give rise to a supercritical galton - watson process that survives with positive probability ( see @xcite for details ) and so the expected number of the particles at the origin grows exponentially , thus showing that @xmath119 . on the other hand , if @xmath119 , then theorem [ t_local_size ] implies that , with positive @xmath58-probability , there exists @xmath417 ( possibly depending on @xmath22 ) such that @xmath418 . this implies the existence of a recurrent seed . indeed , denote @xmath419 and take @xmath420 . choose a small @xmath421 in such a way that @xmath422 for any @xmath423 such that @xmath424 for all @xmath425 . then , @xmath426 is a recurrent seed , and so the branching random walk is recurrent . fix @xmath108 and consider any @xmath400 . by theorem [ t_local_size ] , there exists @xmath427 ( which does not depend on @xmath68 ) such that @xmath428 for all @xmath429 . we write @xmath430 \leq \frac{{{\mathtt e}_{\omega}}\eta^0_n(y)}{\exp((\beta(y / n)+2{\varepsilon})n ) } \leq e^{-{\varepsilon}n},\ ] ] so , by borel - cantelli lemma , @xmath431 * step 1 : * first of all , we establish that , for any @xmath108 @xmath435>0 \qquad\text { for $ { { \mathbb p}}$-almost all $ { \text{\boldmath $ { \omega}$}}$}.\ ] ] choose @xmath259 such that @xmath436 , then choose a positive @xmath437 in such a way that @xmath438 , then let @xmath439 . abbreviate @xmath440;\ ] ] recall that , by ( [ incl_rn ] ) , @xmath441 as @xmath116 . by virtue of ( [ bounded_conv ] ) , @xmath442 so one can choose @xmath266 such that @xmath443 and also @xmath444 @xmath445 now , we construct a branching process in random environment @xmath446 in the following way . here , @xmath447 stands for the size of @xmath448th generation of this process . with respect to the original process , the particles of @xmath448th generation are in @xmath449 at time @xmath450 defined below ( note that it means that @xmath451 , but the equality should not necessarily hold true , there may be also some particles in @xmath449 at time @xmath450 which do not belong to this branching process in random environment ) . specifically , the initial particle is considered the particle of @xmath160th generation , and we set @xmath452 , @xmath453 . inductively , consider the @xmath454 particles of @xmath455th generation , situated in @xmath456 at the moment @xmath457 . then , the particles of @xmath448th generation are their descendants which are in @xmath449 at time @xmath450 , where @xmath458 by birkhoff s theorem , @xmath459 and , by condition ue and ( [ 2nd_restr_n_1 ] ) , @xmath460 assume that @xmath421 is so small that @xmath461 . since ( by condition ue and using the fact that @xmath438 ) @xmath462 \geq \min\{{\varepsilon}_0^{k_1\|a\|n_1},\delta^{k_1(1-h)n_1}\},\ ] ] one can use e.g.theorem 1 of @xcite , or theorem 5.5 and proposition 6.2 of @xcite to obtain that @xmath463>0 \qquad \text { $ { { \mathbb p}}$-a.s.},\ ] ] and @xmath464 = 1.\ ] ] on the event @xmath465 one can choose @xmath466 ( depending on @xmath22 ) such that @xmath467 and @xmath468 for all @xmath469 . then , at the moment @xmath470 we have at least @xmath471 particles in @xmath472 . by condition ue , each of those particles has a descendant in @xmath472 at time @xmath473 with probability at least @xmath474 . so , using the large deviation bound for the binomial distribution ( cf . e.g.formula ( 34 ) of @xcite ) , we obtain that for some positive @xmath475 @xmath476}~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ \\ & \geq 1-\exp(-c_1e^{c_2m}).\end{aligned}\ ] ] using condition ue again , we obtain ( [ po(liminf)>0 ] ) . * step 2 : * now , let us show that ( [ po(liminf)>0 ] ) implies ( [ liminf_fixed_a ] ) . this is easy in the case when the branching random walk is recurrent . indeed , in this case it can be shown that a.s . the origin will be visited by infinitely many `` free '' particles ( i.e. , none of them is a descendant of another ; see @xcite for more details ) . each of those particles gives rise to a copy of the branching process in random environment constructed above ( they use the same environment , but are conditionally independent ) ; so , with probability @xmath25 at least one of them survives , and from this we obtain ( [ liminf_fixed_a ] ) in the recurrent case . however , this argument does not work in the case when the branching random walk is transient , so we present a general argument that works in both cases . abbreviate @xmath477 . let @xmath478 the key observation is that the branching process @xmath479 constructed above depends only on the environment inside @xmath480 . in particular the probability in ( [ surv_bpre ] ) only depends on @xmath481 . suppose that @xmath482 for some @xmath324 ; this means that up to ( `` real '' ) time @xmath483 the branching process in random environment became extinct . then , the idea is the following : with positive probability it happens that a particle goes outside the already explored part of @xmath22 `` without revealing more environment '' , and we can construct a new branching process in random environment , independent of the previous one . at the moment @xmath483 ( when we know that the `` initial '' branching process @xmath479 in random environment became extinct ) , let us remove all particles except one from the process @xmath484 ( for definiteness , choose the remaining particle uniformly among the particles that are present at time @xmath483 ) . let @xmath20 be the `` initial '' ( i.e. , at time @xmath483 ) position of this particle , note that @xmath485 . we let this particle perform the uniform induced random walk @xmath177 ( i.e. , immediately removing from the process other particles that may eventually appear ) until some random moment @xmath486 defined below . let @xmath487 be the set of sites from where the construction analogous to the construction of the above branching process @xmath479 may depend on already revealed pieces of the environment . clearly , if @xmath493 , then @xmath494 . from condition ue it follows that a.s . there exists an infinite sequence @xmath495 such that @xmath496 for all @xmath341 . indeed , if @xmath497 is such that @xmath498 , then the fact that @xmath499 is large enough guarantees that there is @xmath341 such that @xmath500 $ ] . since from any @xmath501 the particle can perform any given @xmath502 unit steps with @xmath199-value @xmath25 with uniformly positive probability ( and so it can go out of @xmath503 with at least that probability without revealing more environment ) , this shows that @xmath504 a.s . now , at time @xmath505 we can start an independent copy @xmath506 of the branching process in random environment @xmath479 constructed above . if it happens that @xmath506 dies out as well , repeating this construction one can start another independent copy @xmath507 , and so on . eventually , one of those branching processes in random environment will survive , and so ( [ liminf_fixed_a ] ) follows from ( [ growth_bpre ] ) . now , let us prove ( [ limsup_complete ] ) . we use the same method as in the proof of the first part of theorem [ t_local_size ] . namely , suppose that @xmath259 is such that @xmath392 . for any small enough @xmath393 there exist @xmath394 such that ( [ covering- ] ) holds . let @xmath396 , and let @xmath508 , @xmath398 . by ( [ liminf_fixed_a ] ) , provided that @xmath509 are chosen in such a way that @xmath510 we have @xmath511 for all @xmath21 large enough , @xmath398 . using ( [ covering- ] ) , we obtain that for any @xmath400 there exists @xmath91 such that @xmath401 . then , we can show that on the event @xmath402 @xmath512 \geq \delta_0{\varepsilon}_0^{n - k_im_i}.\ ] ] so , since the particles that are in @xmath513 at time @xmath514 act independently , again using a large deviation bound for the binomial distribution together with ( [ snizu_complete ] ) and ( [ prob_omega_reach ] ) , we obtain that for some positive constants @xmath515 @xmath516 \geq 1 - \exp(-c_3e^{c_4n}),\ ] ] and this proves ( [ limsup_complete ] ) . theorem [ t_complete_shape ] now follows from ( [ liminf_complete_shape ] ) and ( [ limsup_complete ] ) . first , we prove the property ( [ sup_beta>0 ] ) . for the case of recurrent branching random walk , ( [ sup_beta>0 ] ) immediately follows from theorem [ t_rec_trans ] , so we only need to consider the case of transient branching random walk . without restriction of generality , one can assume that any particle can generate at most two particles at the next moment , and that there is @xmath517 such that the probability of generating two offspring is ( depending on the location ) @xmath160 or @xmath518 , i.e. , @xmath519=1 , \qquad { { \mathbb p}}[\omega_0(|v|=2)\in\{0,h\}]=1.\ ] ] this is no restriction since it can be easily shown that any branching random walk in random environment ( satisfying condition b ) dominates some branching random walk in random environment satisfying ( [ reduce_brw ] ) . also , it is clear that the latter branching random walk in random environment can be defined in such a way that condition ue still holds ( possibly with another constant ; however , to keep the notations simple , we will assume in our argument that condition ue holds with @xmath229 ) . then , we can make a further simplifying assumption : we suppose that _ immediate _ descendants of a particle jump independently ( so that we are in the situation considered e.g. in @xcite ) , which means that , for a particle in @xmath7 , we first decide if the particles generate 1 or 2 offspring , and then each of these offspring jump independently with probabilities @xmath522 , y\in x+{{\mathfrak a}})$ ] . here , @xmath523 is the uniform induced random walk starting from @xmath7 . the reason why we can make this assumption without loss of generality is that the two branching random walks are in some sense equivalent , as mentioned in remark [ rem_complete_shape ] . [ l_noreturn ] there exists @xmath526 , depending only on @xmath518 and @xmath229 , such that , for any transient branching random walk satisfying ( [ reduce_brw ] ) and with independent immediate descendants , we have @xmath527 , for all @xmath38 and for @xmath58-almost all @xmath22 . _ proof of lemma [ l_noreturn ] : _ by stationarity , it is enough to prove that @xmath528=1 $ ] . we argue by contradiction . the idea is that , when @xmath529 is small enough , the branching random walk should be recurrent . we split the space of environments in three parts , and consider three cases accordingly . suppose that @xmath530 , i.e. , @xmath160 is a branching site . abbreviating @xmath531 $ ] , we have @xmath532 , so @xmath533 = \frac{-(1\!-\!h)+\sqrt{(1\!-\!h)^2 + 4h \psi_\omega(0)}}{2h}.\ ] ] thus , @xmath534 $ ] can be made arbitrarily close to @xmath160 by making @xmath529 small . now , if @xmath535 < \frac{h}{1+h},\ ] ] then , as in the proof of theorem [ t_rec_trans ] , it is straightforward to show that there exists a recurrent seed and so the branching random walk is recurrent . we suppose that @xmath536 , but there is a branching site in @xmath537 . in this case we can choose a branching site there which is closest to the origin , so that it is accessible from the origin by a path of non - branching sites . this means that there exist @xmath538 , @xmath539 such that @xmath540 , @xmath541 , @xmath542 are non - branching , and @xmath543 is a branching site ( and we have also @xmath544 ) . then , again we obtain that if @xmath529 is too small , then the branching random walk should be recurrent . for this , proceed as follows . first , define @xmath545 so that @xmath546=1-\psi_\omega(0)$ ] . at the moment @xmath547 ( provided @xmath548 ) , consider one of the particles which are in @xmath160 and let @xmath549 then , repeat this procedure to define @xmath550 on @xmath551 , @xmath552 on @xmath553 , and so on . clearly , @xmath554=(1-\psi_\omega(0))^m$ ] . being @xmath555 the uniform induced random walk starting from @xmath160 , define the event @xmath556 i.e. , the initial particle goes straight to @xmath543 , branches there , and then the `` first '' descendant goes straight back to the origin ( note that we do not assume anything about the second descendant ) . clearly , we have @xmath557 \geq \frac{h{\varepsilon}_0^{2k}}{1-\psi_\omega(0 ) } \geq h{\varepsilon}_0^{2(\rho-1)}.\ ] ] on the event @xmath105 , the second particle ( generated in @xmath543 ) goes to @xmath160 with probability at least @xmath558 . take @xmath324 such that @xmath559 : if @xmath529 is small enough , then @xmath560 . then we obtain that there exists large enough @xmath561 ( which actually depends on @xmath22 ) such that up to time @xmath561 the mean number of `` free '' particles that visit the origin is greater than @xmath25 . ( recall that , by `` _ free _ '' particles we mean that none of them is the descendant of another one , see definition [ rec_seed ] . ) as above , this implies the existence of recurrent seeds and so the branching random walk is recurrent . suppose that @xmath521 for all @xmath562 , so that there are no branching sites in @xmath563 . let @xmath564 be the annulus ( or extended external boundary ) of @xmath563 . denote by @xmath565\ ] ] the probability of ever having particles in @xmath566 for the process starting from @xmath7 . for any @xmath567 , we see from condition ue that @xmath568 , so @xmath569 at this point , the important observation is that , for any @xmath567 , the quantity @xmath570 is a function of the environment outside @xmath563 , and so it is independent of @xmath147 by condition i. so , supposing that @xmath529 is small enough , one can construct a recurrent seed as follows . define @xmath571,\ ] ] and note that @xmath572 as @xmath573 . then , suppose that @xmath574 , where ( a small number ) @xmath575 is such that @xmath576 choose first large enough @xmath238 and then small enough @xmath421 in such a way that @xmath577 for all @xmath423 such that @xmath578 for all @xmath579 ( here , @xmath580 denotes the ball of radius @xmath421 in @xmath49 with respect to any fixed metrics ) . by the contradiction assumption and by ( [ inf_g(xk ) ] ) , the set of such @xmath423 s has positive @xmath58-probability . then , the idea is the following : put a branching site at the origin and suppose that in a sufficiently large region around @xmath160 ( excluding @xmath563 ) the environment is `` close '' to the environment above . more precisely , we consider the @xmath200-seed with @xmath581 and @xmath582 condition i implies that @xmath583>0 $ ] . now , from each site in @xmath563 , the uniform induced random walk goes to @xmath160 without leaving @xmath563 with probability at least @xmath558 , and any particle in @xmath584 sends at least one descendant to @xmath584 at least with probability given by ( [ choose_k0 ] ) . suppose without restriction of generality that @xmath585 . then , the probability that any particle starting from @xmath94 sends at least one descendant to @xmath160 is at least , for the coin 2 the probability of head is @xmath586 . we flip the coins in an alternate fashion , i.e. , 1,2,1,2,1,2, then , the probability that the first head comes from the coin 1 is equal to @xmath587 . ] @xmath588 which is greater than @xmath589 by ( [ eq_df_tilde_delta ] ) . so , we obtain a recurrent seed . we continue the proof of theorem [ t_total_size ] . let @xmath590 be the range of the uniform induced random walk up to time @xmath21 . we now prove that , due to lemma [ l_noreturn ] , @xmath591 is of order @xmath21 with at least constant probability . let @xmath592 be the number of times that the uniform induced random walk starting from @xmath7 visits @xmath68 . using lemma [ l_noreturn ] , we obtain that @xmath593 for all @xmath38 and for @xmath58-almost all @xmath22 . on the other hand @xmath594 so , taking expectation , @xmath595\\ & \leq ( 1+\theta^{-1}){{\mathtt e}_{\omega}}|u_n|.\end{aligned}\ ] ] since trivially @xmath596 , using the fact that for any random variable @xmath597 with @xmath598 a.s . and @xmath599 it is true that @xmath600\geq b/(2a)$ ] , we obtain that , for @xmath58-almost all @xmath22 , @xmath601 \geq ( 1+\theta^{-1})^{-1}/2.\ ] ] consider the evolution of the branching random walk up to time @xmath21 , and let us enumerate the @xmath602 particles that are present at time @xmath21 in random order ( i.e. , select one particle at random and attach the label `` 1 '' to it , then select one of the unlabelled particles and put the label `` 2 '' to it , and so on ) . we define @xmath603 to be the number of bifurcations on the path from the root to the particle ( of the @xmath21th generation ) labelled @xmath91 on the genealogical tree of the branching random walk ( see figure [ f_gen_tree ] ) . let us prove now that , on the event that the range of the uniform induced random walk is linear in @xmath21 , with large probability @xmath610 will be linear as well . to this end , we construct a set @xmath611 in the following way : first , we have @xmath612 . inductively , suppose that the set @xmath613 was already constructed . then , @xmath614 if and only if the following holds : define the cubes @xmath619 note that @xmath620 let @xmath453 , @xmath621 . inductively , define for @xmath622 @xmath623 and @xmath624 is such that @xmath625 . observe that , for any @xmath615 such that @xmath626 we have @xmath627 so @xmath628 for all @xmath629 . since @xmath630 , we have @xmath631 let @xmath632 be the sigma - algebra generated by @xmath633 . by definition of the set @xmath634 , @xmath635 \geq h{\varepsilon}_0^\rho { { \mathbb p}}[\omega_0(|v|=2)=h].\ ] ] write @xmath636 , where @xmath637 for all @xmath629 . abbreviate @xmath638 , @xmath639 , @xmath640 $ ] , and @xmath641 . the property ( [ annealed_next_branching ] ) implies that , for some @xmath642 , @xmath643 \geq 1-e^{-c_1n}.\ ] ] so , using ( [ range > cn ] ) and ( [ phi_range ] ) together with the elementary inequality @xmath644 \geq 1-\frac{{{\mathbf p}}[a^c]}{{{\mathbf p}}[b]}$ ] , we obtain for some @xmath645 that @xmath646 & \geq { { \mathbf p}}[{{\mathsf b}{\mathsf r}}(\zeta_n , n)\geq\alpha_4 n\mid |u_n|\geq \alpha_1n ] { { \mathbf p}}[|u_n|\geq \alpha_1n]\\ & \geq { { \mathbf p}}[{{\mathsf b}{\mathsf r}}(\zeta_{\phi_{{\lfloor}\alpha_1\alpha_2n{\rfloor}}},\phi_{{\lfloor}\alpha_1\alpha_2n{\rfloor}})\geq \alpha_4 n \mid |u_n|\geq \alpha_1n ] \\ & \quad { } \times { { \mathbf p}}[|u_n|\geq \alpha_1n]\\ & \geq c_2.\end{aligned}\ ] ] therefore , with @xmath58-probability at least @xmath647 , @xmath648 \geq c_3 \ , .\ ] ] let @xmath649 be the sigma - algebra generated by the branching random walk up to time @xmath21 . applying ( [ eq_prob_zeta ] ) , we obtain that , for such @xmath22 s , @xmath650 = \sum_{i:{{\mathsf b}{\mathsf r}}(i , n)\geq\alpha_4 n } 2^{-{{\mathsf b}{\mathsf r}}(i , n ) } \leq 2^{-\alpha_4n}{{\mathcal z}}^0_n\,,\ ] ] so , taking expectations , @xmath651 . as observed in the beginning of this section , this proves ( [ sup_beta>0 ] ) .

the space of stability conditions on a triangulated category is naturally partitioned into subsets of stability conditions with a given heart . if has finite length and simple objects then has a simple geometry , depending only on . furthermore , bridgeland has shown that if is obtained from by a simple tilt , i.e. by tilting at a torsion theory generated by one simple object , then the intersection of the closures of and has codimension one . suppose that , and any heart obtained from it by a finite sequence of ( left or right ) tilts at simple objects , has finite length and finitely many indecomposable objects . then we show that the closures of and intersect if and only if and are related by a tilt , and that the dimension of the intersection can be determined from the torsion theory . in this situation the union of subsets , where is obtained from by a finite sequence of simple tilts , forms a component of the space of stability conditions . we illustrate this by computing ( a component of ) the space of stability conditions on the constructible derived category of the complex projective line stratified by a point and its complement .

0909.0552

bridgeland introduced the notion of a stability condition on a triangulated category @xmath5 in @xcite and proved that the set of locally - finite stability conditions forms a complex manifold @xmath6 . this space provides a geometric picture of many aspects of the category and carries a natural action of its automorphisms . each stability condition determines an abelian heart in the triangulated category . if we denote the subset of locally - finite stability conditions with heart @xmath1 by @xmath0 then the @xmath0 partition @xmath6 . if the heart @xmath1 has finite length and @xmath2 simple objects then @xmath7 where @xmath8 in this case @xmath9 is non - empty and of codimension one when @xmath3 is a simple left tilt of @xmath1 ( * ? ? * lemma 5.5 ) , i.e. when @xmath3 is obtained from @xmath1 by tilting in the sense of @xcite at a torsion theory generated by a simple object . we generalise this result as follows . fix a heart @xmath1 and assume that any heart obtained from it by a finite sequence of simple tilts has finite length and finitely many indecomposable objects . in particular each of these hearts therefore has finitely many simple objects and finitely many torsion theories . in this case we show that @xmath10 if and only if @xmath3 is the left tilt of @xmath1 at some torsion theory . the codimension of the intersection is governed by the size of the torsion theory , lower codimension corresponding to a smaller torsion theory ( see corollary [ intersection corollary ] for a precise statement ) . proposition [ constraints ] says that points in the boundary of @xmath11 correspond to limiting central charges of stability conditions in @xmath0 for which the central charge of no semi - stable object vanishes . furthermore , the union of subsets of stability conditions with hearts obtained from @xmath1 by sequences of simple tilts forms a connected component of @xmath6 ( theorem [ component thm ] ) . together these results yield a description of a component of @xmath6 in terms of the combinatorics of tilting in @xmath5 . in brief , [ torsion theories ] , [ tilting ] and [ tilting at simples ] provide the necessary background on the relation between @xmath12-structures , torsion theories and tilting in triangulated categories . [ stability conditions ] recalls the definition and basic properties of stability conditions . this material is standard and is included for expository purposes only . the main results , outlined above , are in [ main results ] . they are illustrated in [ constructible ] in which we compute ( a component of ) the space of stability conditions on the constructible derived category of @xmath13 stratified by a point and its complement . ( algebraists may prefer to view this example as the derived category of representations of the bound quiver @xmath14_c & \cdot \ar@/{}_{.5pc}/[l]_v } \qquad vc=0\ ] ] and symplectic geometers as the subcategory of the derived asymptotic fukaya category of @xmath15 generated by the zero section and a cotangent fibre . ) this example is sufficiently simple to allow explicit computation and yet complicated enough to exhibit all the interesting features of the results . finally [ coherent ] discusses the contrasting case of coherent sheaves on @xmath13 , following the detailed treatment in @xcite . in this example the hearts do not have finitely many indecomposables and we no longer obtain an entire component of the space of stability conditions by iterating simple tilts . i m grateful to ivan smith and michael butler for several helpful conversations and to tom bridgeland for his comments and for pointing me towards the example of coherent sheaves on @xmath13 as a useful counterpoint . i am indebted to the referee who spotted several gaps and errors in the original version . we fix some notation . let @xmath5 be an additive category . we write @xmath16 to mean @xmath17 is an object of @xmath5 . we will use the term _ subcategory _ to mean strict , full subcategory . when @xmath18 is a subcategory we write @xmath19 for the subcategory on the objects @xmath20 and similarly @xmath21 for @xmath22 . when @xmath1 and @xmath3 are subcategories of @xmath5 we write @xmath23 for the subcategory on objects which lie in both @xmath1 and @xmath3 . suppose @xmath5 is triangulated with shift functor @xmath24 $ ] . exact triangles in @xmath5 will be denoted either by @xmath25 $ ] or by a diagram @xmath26 & & b \ar[dl ] \\ & c \ar@{-->}[ul ] & } \ ] ] where the dotted arrow denotes a map @xmath27 $ ] . we will always assume that @xmath5 is essentially small so that isomorphism classes of objects form a set . given sets @xmath28 of objects for @xmath29 let @xmath30 denote the ext - closed subcategory generated by objects isomorphic to an element in some @xmath28 . we will use the same notation when the @xmath28 are subcategories of @xmath5 . a _ @xmath12-structure _ on a triangulated category @xmath5 is a subcategory @xmath31 such that @xmath32 \subset { \mathcal{d}}$ ] and for each @xmath16 there is an exact triangle @xmath33 $ ] with @xmath34 and @xmath35 . the subcategory @xmath36 $ ] is abelian ( * ? ? ? * thorme 1.3.6 ) and is known as the _ heart _ of the @xmath12-structure . it is more common to define a @xmath12-structure to be a pair @xmath37 of sub - categories satisfying a short list of conditions , see @xcite . this definition is equivalent to the one above if we put @xmath38 and @xmath39 $ ] . the subcategory @xmath40 is sometimes referred to as a @xmath12-category or an aisle . it follows from the existence of the triangle that @xmath40 is right admissible , i.e. that there is a right adjoint @xmath41 to the inclusion @xmath42 and that @xmath43 is left admissible with left adjoint @xmath44 to the inclusion @xmath45 . these adjoints are referred to as truncation functors . the exact triangle associated to an object @xmath17 is unique ( up to isomorphism ) and can be written @xmath46\ ] ] where the first two maps come respectively from the counit and unit of the adjunctions . the truncations give rise to cohomological functors @xmath47 to the heart @xmath1 where @xmath48)[-n]$ ] and @xmath49)[1-n]$ ] . a @xmath12-structure @xmath40 is _ bounded _ if any object of @xmath5 lies in @xmath50 \cap { \mathcal{d}}^\perp[n]$ ] for some @xmath51 . in the sequel we will always assume that @xmath12-structures are bounded . this has two important consequences . firstly , a bounded @xmath12-structure is completely determined by its heart @xmath1 ; the @xmath12-structure is recovered as @xmath52 , { \mathcal{a}}[2 ] , \ldots \rangle.\ ] ] secondly , the inclusion @xmath53 induces an isomorphism @xmath54 of grothendieck groups . closely related to this is the easy but important fact that if @xmath55 are hearts of ( bounded ) @xmath12-structures then @xmath56 . a @xmath12-structure is said to be _ faithful _ if there is an equivalence @xmath57 where @xmath58 is the bounded derived category of the heart @xmath1 . note that , in general , there is not even a naturally defined functor @xmath59 extending the inclusion @xmath60 . however , if such an @xmath61 does exist then it is an equivalence if and only if the ext groups ( in @xmath5 ) between any two objects in @xmath1 are generated in degree @xmath62 ( * ? ? ? * chapter 5 , theorem 3.7.3 ) . in the sequel , we do not assume that @xmath12-structures are faithful ( although some of those we consider will be ) . a _ torsion theory _ in an abelian category @xmath1 is a subcategory @xmath63 such that every @xmath64 fits into a short exact sequence @xmath65 for some @xmath66 and @xmath67 . objects of @xmath63 are known as torsion objects and objects of @xmath68 as free objects ; the motivating example is the subcategories of torsion and free abelian groups . torsion theories are more commonly defined as pairs @xmath69 of subcategories such that @xmath70 whenever @xmath66 and @xmath71 and every @xmath64 sits in a short exact sequence @xmath72 with @xmath66 and @xmath71 , see for example ( * ? ? ? * definition 1.1 ) . the definitions are equivalent : if @xmath73 is a torsion theory then @xmath74 . the short exact sequence @xmath75 is unique up to isomorphism . the first term determines a right adjoint to the inclusion @xmath76 and the last term a left adjoint to the inclusion @xmath77 . it follows that @xmath63 is closed under factors , extensions and coproducts and that @xmath78 is closed under subobjects , extensions and products . torsion theories in @xmath1 form a poset with the ordering @xmath79 . the ( bounded ) @xmath12-structures on @xmath5 also form a poset under the ordering @xmath80 . [ hrs ] let @xmath5 be a triangulated category , and @xmath81 a @xmath12-structure with heart @xmath1 . then there is a canonical isomorphism between the poset of torsion theories in @xmath1 and the interval in the poset of @xmath12-structures consisting of @xmath12-structures @xmath82 with @xmath83 $ ] . denote the cohomological functors associated to @xmath40 by @xmath84 . given a torsion theory @xmath63 in @xmath1 define a subcategory @xmath85 \rangle = \langle c \in { \mathcal{c}}\ |\ h^ic=0 \ \textrm{for}\ i>1 \ \in { \mathcal{t}}\rangle.\ ] ] happel , reiten and smal ( * ? ? ? * proposition 2.1 ) show that this is a @xmath12-structure . it is clear that @xmath83 $ ] . conversely , given a @xmath12-structure @xmath82 with @xmath86 $ ] we define @xmath87 = \langle a\in{\mathcal{a}}\ |\ a = h^1e \ \textrm{for some } \ e\in { \mathcal{e } } \rangle.\ ] ] beligiannis and reiten ( * ? ? ? * theorem 3.1 ) show that this is a torsion theory in @xmath1 . these constructions preserve the partial orders and are mutually inverse . let @xmath1 be the heart of a @xmath12-structure @xmath40 . by the above a torsion theory @xmath63 in @xmath1 determines a new @xmath12-structure @xmath88 \rangle$ ] : we say this new @xmath12-structure is obtained from the original by _ left tilting at @xmath63 _ and denote its heart by @xmath89 . explicitly @xmath90 \rangle \\ = \ { c \in { \mathcal{c } } \ | \ \textrm{$h^{0}c \in { \mathcal{f } } , h^1c\in { \mathcal{t}}$ and $ h^ic=0 $ for $ i \neq 0,1 $ } \},\ ] ] where @xmath84 denotes the @xmath91th cohomology functor with respect to the initial @xmath12-structure and @xmath68 . a torsion theory also determines another @xmath12-structure @xmath92 by a ` double dual ' construction . this is the shift by @xmath24 $ ] of the other : we say it is obtained by _ right tilting _ at @xmath63 and denote the new heart by @xmath93 . intuitively , @xmath89 is obtained from @xmath1 by replacing @xmath63 by its shift @xmath94 $ ] and @xmath93 is obtained by replacing @xmath78 by @xmath95 $ ] . left and right tilting are inverse to one another : @xmath78 is a torsion theory in @xmath89 and right tilting with respect to this we recover the original heart @xmath1 . [ simplicity criterion 1 ] let @xmath63 be a torsion theory in the heart @xmath1 of a @xmath12-structure . then any simple object in @xmath89 lies either in @xmath68 or in @xmath94 $ ] and 1 . @xmath71 is simple in @xmath96 there are no exact triangles @xmath97 \qquad \textrm{or } \qquad t'[-1 ] \to f\to f'\to t'\ ] ] with @xmath98 and @xmath99 and all non - zero ; 2 . @xmath100 \in { \mathcal{t}}[-1]$ ] is simple in @xmath96 there are no exact triangles @xmath101 \qquad \textrm{or } \qquad t'[-1]\to t[-1 ] \to f'\to t'\ ] ] with @xmath102 and @xmath103 and all non - zero . since @xmath78 is a torsion theory in @xmath89 with @xmath104 $ ] every object @xmath105 fits into a short exact sequence @xmath106 \to 0 $ ] , and therefore can only be simple if it lies in either @xmath78 or @xmath94 $ ] . the necessity of the conditions is evident . we prove sufficiency only in the second case , the argument for the first case is similar . suppose @xmath100\in t[-1]$ ] is not simple in @xmath89 . then there must be a proper subobject , and since @xmath94 $ ] is the free part of a torsion theory in @xmath89 this subobject must also be in @xmath94 $ ] , i.e. we have a short exact sequence @xmath107\to t[-1 ] \to b\to 0\ ] ] with @xmath99 and @xmath105 . there is a short exact sequence @xmath108\to 0 $ ] with @xmath71 and @xmath109 . if @xmath110 then @xmath111 is a proper quotient of @xmath12 and we can change our choice of @xmath112 ( by enlarging it ) to obtain a short exact sequence @xmath107\to t[-1 ] \to t''[-1]\to 0\ ] ] in @xmath89 . on the other hand , if @xmath113 then @xmath114 and we obtain a short exact sequence @xmath115\to t[-1]\to f\to 0\ ] ] in @xmath89 . these sequences yield the claimed exact triangles . we leave the reader to formulate and prove the analogue for right tilts . we will be particularly interested in the smallest changes that can be made in @xmath12-structures by tilting . assume that the heart @xmath1 has finite length and finitely many simple objects . to spell this out , @xmath1 is both artinian and noetherian so that every object has a finite composition series . by the jordan - hlder theorem , the graded object associated to such a composition series is unique up to isomorphism . the simple objects form a basis for the grothendieck group , which is isomorphic to @xmath116 , where @xmath2 is the number of simple objects . under this assumption each simple object @xmath117 determines two torsion theories , namely @xmath118 and @xmath119 . these are respectively minimal and maximal ( in the sense of the poset of torsion theories ) non - trivial torsion theories in @xmath1 . to harmonise our notation with @xcite we use the shorthand @xmath120 and @xmath121 . we will refer to @xmath122 and @xmath123 as _ simple tilts _ of @xmath1 . for the remainder of this section we make the following stronger assumption . [ assumption1 ] the heart @xmath1 and all hearts obtained from it by finite sequences of simple tilts have finite length and finitely many simple objects . ( note that they all have the same number of simple objects , since the classes of these form a basis for the grothendieck group . ) [ simplicity criterion 2 ] suppose @xmath124 is simple . for each simple @xmath125 in @xmath1 there are unique , up to isomorphism , simples @xmath126 and @xmath127 fitting into exact triangles @xmath100\to a\to l_sa \to t$ ] with @xmath128 and @xmath129 $ ] with @xmath130 respectively . each simple in @xmath122 , respectively @xmath123 , is either @xmath131 $ ] , respectively @xmath132 $ ] , or of the form @xmath133 , respectively @xmath134 , for a unique simple @xmath135 in @xmath1 . moreover , @xmath136}l_sa\cong a$ ] and @xmath137}b\cong b$ ] . the cases are similar so we deal only with the left tilt . it follows immediately from lemma [ simplicity criterion 1 ] that @xmath131 $ ] is simple in @xmath122 , and that every other simple must lie in @xmath138 . suppose that @xmath139 is simple in @xmath1 . if @xmath135 is _ not _ simple in @xmath122 then by lemma [ simplicity criterion 1 ] it has @xmath131 $ ] as a subobject . quotienting by this , and repeating if necessary , we obtain @xmath140 with @xmath141,b)}=0 $ ] . we use the assumption that @xmath122 has finite length to show this process of quotienting must terminate . since @xmath138 is a torsion theory in @xmath122 the quotient @xmath142 of @xmath135 is in @xmath138 too , in particular @xmath143 . by construction , the composition factors of @xmath142 in @xmath1 are @xmath135 and a number ( possibly zero ) of copies of @xmath144 . it follows that there is no exact triangle @xmath145\ ] ] with @xmath146 and both non - zero . therefore , by lemma [ simplicity criterion 1 ] , the object @xmath142 is simple in @xmath122 . the class of @xmath142 in the grothendieck group is @xmath147+r[s]$ ] for some @xmath148 . it follows that there is a unique such simple quotient ( otherwise the simples in @xmath122 would not form a basis ) . set @xmath149 . considering classes in the grothendieck group again , we see that @xmath150 implies @xmath151 . thus , by counting , each simple in @xmath122 other than @xmath131 $ ] is of the form @xmath133 for some simple @xmath64 . the last assertion follows from noting that @xmath136}l_sa$ ] is the unique simple in @xmath1 fitting into a triangle @xmath152\to r_{s[-1]}l_sa\to l_sa \to a'\ ] ] with @xmath153 . since @xmath135 also fits into the second term of this triangle , uniqueness tells us that @xmath136}l_sa\cong a$ ] . the proof that @xmath154}b\cong b$ ] is similar . if we define @xmath155 $ ] and @xmath156 $ ] then we obtain natural bijections between the simples in @xmath1 and those in @xmath122 and @xmath123 respectively . there are no such natural bijections for more general tilts . in the example we compute in [ constructible ] it is possible to perform a sequence of simple tilts beginning and ending at the same heart for which the composite of the above natural bijections is not the identity . the next lemma shows that tilting left by a torsion theory containing only finitely many indecomposables is equivalent to performing a sequence of simple left tilts . there is an obvious analogue for right tilts . [ factorisation ] suppose @xmath157 is a finite sequence of objects in @xmath1 such that @xmath158 is simple in @xmath159 . then @xmath160 where @xmath161 . furthermore , any left tilt at a torsion theory containing finitely many ( isomorphism classes of ) indecomposables is obtained in this way as a composition of simple left tilts . write @xmath162 for @xmath1 and @xmath163 for @xmath159 . denote the @xmath12-structure with heart @xmath1 by @xmath164 and so on . by proposition [ hrs ] @xmath165 \rangle= \cdots = \langle { \mathcal{d}}_{{\mathcal{a}}_0 } , a_0[-1],\ldots , a_{n-1}[-1 ] \rangle = \langle { \mathcal{d}}_{\mathcal{a } } , { \mathcal{t}}[-1 ] \rangle\ ] ] where @xmath161 . hence @xmath166 as claimed . now suppose @xmath167 is a ( non - zero ) torsion theory in @xmath162 containing only finitely many indecomposables . since @xmath168 is closed under factors it must contain at least one simple object , @xmath169 say . we claim that @xmath170 is a torsion theory in @xmath171 with corresponding free theory @xmath172\rangle$ ] where @xmath173 . to see this , consider @xmath174 . there is a short exact sequence @xmath175 in @xmath171 with @xmath176 and @xmath177\rangle$ ] . since @xmath178 there is then a short exact sequence @xmath179 in @xmath162 with @xmath180 and @xmath181 . furthermore , @xmath12 must be in @xmath182 as @xmath183 is , so that @xmath184 . the three objects in the second short exact sequence all lie in @xmath171 so that it is , in fact , short exact in @xmath171 too . hence @xmath12 is a subobject of @xmath142 in @xmath171 and the quotient is an extension of @xmath185 by @xmath186 and so lies in @xmath187 . since @xmath188 the claim follows . note that @xmath189 ( although we do not claim that @xmath190 is a torsion theory in @xmath162 ) . repeating the argument we obtain a nested sequence of full subcategories in @xmath168 . since @xmath168 has only finitely many indecomposables , and at each stage we remove at least one indecomposable , namely the simple @xmath191 , we must eventually reach @xmath192 . by the first part @xmath193 is a torsion theory in @xmath162 with @xmath194 clearly @xmath195 . in fact they must be equal , so that @xmath196 as required . if not choose @xmath197 but not in @xmath198 . we may assume that @xmath199 . if it is not then there is a monomorphism @xmath200 for some @xmath201 and we may replace @xmath142 by the cokernel , which must also lie in @xmath168 but not @xmath198 . iterating , and using the fact that @xmath162 has finite length to ensure termination , we may indeed assume @xmath202 . we may also assume @xmath142 is indecomposable ; if not replace it with a summand , at least one of which can not be in @xmath198 . hence there is an indecomposable @xmath203 which contradicts @xmath192 . the actions of automorphisms of @xmath5 commute with simple tilts . more precisely , if @xmath204 is an automorphism , i.e. a triangulated equivalence , then @xmath205 see ( * ? ? ? * lemma 8.2 ) . in the example in [ constructible ] the triangulated category @xmath5 will be equipped with a duality , a triangulated anti - equivalence @xmath206 such that @xmath207 . thus every @xmath12-structure will have an associated dual @xmath12-structure , given a torsion theory @xmath63 in the heart @xmath1 of a @xmath12-structure there will be a dual torsion theory @xmath208 in the heart @xmath209 of this dual @xmath12-structure and left tilting @xmath1 with respect to @xmath63 is dual to right tilting @xmath209 with respect to @xmath208 . in particular if @xmath144 is a simple object of @xmath1 then @xmath210 . let @xmath5 be a triangulated category and @xmath211 be its grothendieck group . a _ stability condition _ @xmath212 on @xmath5 ( * ? ? ? * definition 1.1 ) consists of a group homomorphism @xmath213 and full additive subcategories @xmath214 of @xmath5 for each @xmath215 satisfying 1 . if @xmath216 then @xmath217 where @xmath218 ; 2 . @xmath219 $ ] for each @xmath215 ; 3 . if @xmath216 and @xmath220 with @xmath221 then @xmath222 ; 4 . [ hn filtration ] for each nonzero object @xmath16 there is a finite collection of triangles @xmath223 & & c_1 \ar[r ] \ar[dl ] & \cdots \ar[r ] & c_{n-1 } \ar[rr ] & & c_n = c \ar[dl]\\ & b_1\ar@{-->}[ul ] & & & & b_n \ar@{-->}[ul ] } \ ] ] with @xmath224 where @xmath225 . the homomorphism @xmath226 is known as the _ central charge _ and the objects of @xmath214 are said to be _ semi - stable of phase @xmath227_. the objects @xmath228 are known as the _ semi - stable factors _ of @xmath17 . we define @xmath229 and @xmath230 . the _ mass _ of @xmath17 is defined to be @xmath231 . a stability condition @xmath212 determines a bounded @xmath12-structure on @xmath5 with heart the extension - closed subcategory generated by semi - stables with phases in @xmath232 $ ] . conversely , if we are given a bounded @xmath12-structure on @xmath5 together with a stability function on the heart with the harder narasimhan property the abelian analogue of property [ hn filtration ] above then this determines a stability condition on @xmath5 ( * ? ? ? * proposition 5.3 ) . we say a stability condition is _ locally - finite _ if we can find @xmath233 such that the quasi - abelian category @xmath234 , generated by semi - stables with phases in @xmath235 , has finite length ( see ( * ? ? ? * definition 5.7 ) ) . the set of locally - finite stability conditions can be topologised so that it is a , possibly infinite - dimensional , complex manifold , which we denote @xmath6 ( * ? ? ? * theorem 1.2 ) . there is a linear subspace @xmath236 such that the projection @xmath237 is a local homeomorphism on @xmath238 . the slogan is that deformations of the central charge lift to deformations of the stability condition . the topology arises from the ( generalised ) metric @xmath239 which takes values in @xmath240 $ ] . the group @xmath241 of automorphisms acts continuously on the space @xmath6 of stability conditions with an automorphism @xmath242 acting by @xmath243 [ c action ] there is also a free action of @xmath244 on any space of stability conditions given by @xmath245 this action preserves the semi - stable objects , but changes their phases and masses . the action of @xmath246 corresponds to the action of @xmath24 $ ] . a stability condition determines a family of torsion theories in the heart of any @xmath12-structure . [ stab - fam - tor ] let @xmath1 be the heart of some @xmath12-structure on @xmath5 and @xmath247 a stability condition . then for each @xmath248 there is a torsion theory @xmath249 in @xmath1 . set @xmath250 . each @xmath64 has a filtration in terms of semi - stable objects with respect to @xmath247 . combining the semi - stable factors of phase @xmath251 into @xmath252 we obtain a short exact sequence @xmath253 in @xmath1 with @xmath254 and @xmath255 , i.e. with @xmath256 and @xmath257 . since there are no non - zero morphisms from a semi - stable of phase @xmath251 to one of phase @xmath258 we see that @xmath259 . our strategy for understanding the space of stability conditions is based on @xcite . let @xmath260 be the subset of locally - finite stability conditions with heart @xmath1 . clearly these subsets partition the space of stability conditions . if @xmath1 has finite length and @xmath2 simple objects then a stability condition with heart @xmath1 is determined by any non - zero choice of central charge with phase in @xmath232 $ ] for each of the @xmath2 simples . hence the subset @xmath0 is isomorphic to @xmath261 where @xmath262 we can think of the @xmath0 for such hearts @xmath1 as tiles covering part of the space of stability conditions . the next result shows that performing a simple tilt corresponds to moving to an adjacent tile . [ tiling ] let @xmath263 be the heart of a bounded @xmath12-structure and suppose @xmath1 has finite length and @xmath2 simple objects . then , if @xmath264 is the codimension @xmath62 subset for which the simple @xmath144 has phase @xmath62 and all other simples in @xmath1 have phases in @xmath265 , @xmath266 when @xmath267 the codimension one faces of @xmath0 and @xmath268 are glued using a linear map expressing the change of basis in @xmath211 from a basis of simples in @xmath1 to a basis of simples in @xmath3 . it follows from lemma [ simplicity criterion 2 ] that this gluing map @xmath269 where @xmath270 is the strict upper half - plane , is given by a matrix of the form @xmath271 where @xmath272 for @xmath273 . given proposition [ tiling ] it is natural to ask : can we obtain an entire component of the space of stability conditions by iterating simple tilts ? do the tiles @xmath0 and @xmath274 share a boundary for a general torsion theory @xmath63 ? in general the answers are both negative , but in this section we show that the answers are affirmative in the presence of some ( quite strong ) assumptions . fix a heart @xmath1 and let @xmath275 be the set of hearts obtained from @xmath1 by a finite sequence of simple tilts . in order that we can continue to perform simple tilts indefinitely we assume that assumption [ assumption1 ] holds , i.e. that each heart @xmath163 has finite length and finitely many simple objects . the subset @xmath260 is naturally stratified with a stratum of codimension @xmath276 corresponding to a collection of @xmath276 simples @xmath277 in @xmath1 . the corresponding stratum @xmath278 consists of those @xmath279 for which the @xmath280 have phase @xmath62 and all other simples have phases in @xmath265 . the stratum is isomorphic to @xmath281 . suppose @xmath117 is simple and @xmath282 . we say @xmath283 is a _ simple left tilt about @xmath247 _ if @xmath284 and a _ simple right tilt about @xmath247 _ if @xmath285 . note that the subset of stability conditions for which @xmath144 is semi - stable is closed ( * ? ? ? * proposition 8.1 ) , in particular @xmath144 is semi - stable for any @xmath286 so that these definitions make sense . if @xmath283 is a simple left tilt about @xmath282 then @xmath287 . there is an obvious analogue for simple right tilts . since @xmath288 and @xmath284 the stability condition @xmath247 is contained in the closure of the stratum @xmath289 in @xmath0 , and hence @xmath290 too by proposition [ tiling ] . [ intersection corollary ] suppose @xmath291 is a torsion theory in @xmath1 containing only finitely many indecomposables . then @xmath292 and in particular @xmath293 . by lemma [ factorisation ] we can write @xmath294 where @xmath158 is simple in @xmath159 and left tilting at @xmath63 is equivalent to tilting left successively at the simples @xmath295 and @xmath296 . each @xmath297 is in @xmath298 so that @xmath299 is a simple left tilt about @xmath247 for any @xmath300 . [ tilting about stratum ] suppose @xmath3 is obtained from @xmath1 by a finite sequence of left tilts about @xmath282 . then there is a torsion theory @xmath63 in @xmath1 such that @xmath301 . the proof is by induction on the length of the sequence of left tilts about @xmath247 . it is clear for a sequence of length @xmath62 . suppose it is true for all sequences of length @xmath302 and that @xmath3 is obtained from @xmath1 by a sequence of @xmath303 tilts at simples about @xmath247 . let @xmath304 be the heart obtained after the first @xmath302 left tilts . by induction there is a torsion theory @xmath305 in @xmath1 such that @xmath306 . the heart @xmath3 is obtained from @xmath304 by a simple left tilt left at some @xmath307 . by lemma [ simplicity criterion 1 ] either @xmath308 $ ] or @xmath309 . in the latter case @xmath310 and the result follows from proposition [ hrs ] . it remains to show that @xmath311 $ ] . suppose it is , so that @xmath312 $ ] for some @xmath313 . then for @xmath314 we have @xmath315 and , as @xmath316 is continuous ( * ? ? ? * proposition 8.1 ) , and @xmath286 , @xmath317 on the other hand , @xmath142 is simple in @xmath304 hence semi - stable in the closure @xmath318 . we can left tilt at @xmath142 about @xmath247 so @xmath319 which is a contradiction . [ nested remark ] the statement of the lemma is false for more general sequences of simple left ( or right ) tilts ; for instance the heart @xmath320}l_s{\mathcal{a}}$ ] is not a left tilt of @xmath1 at any torsion theory . a sequence of left tilts about @xmath247 corresponds to a strictly increasing nested sequence of torsion theories in @xmath1 . in order to obtain the next result , a converse to corollary [ intersection corollary ] , we need an additional assumption , whose purpose is to ensure that we can not tilt about a stability condition indefinitely . [ assumption2 ] each heart @xmath163 has finite length and only finitely many indecomposables . assumption [ assumption2 ] implies that each @xmath163 has finitely many simples and finitely many torsion theories . this follows because simple objects are indecomposable and because a torsion theory is completely determined by the indecomposables it contains . in particular assumption [ assumption2 ] implies assumption [ assumption1 ] . [ intersection corollary converse ] suppose assumption [ assumption2 ] holds , that @xmath321 are simple objects in @xmath1 and @xmath322 . then @xmath323 for some @xmath324 . choose @xmath325 . if @xmath326 $ ] for every simple @xmath327 then @xmath328 which implies @xmath329 and the result is immediate . assume then that there is a simple @xmath327 with @xmath330 . hence @xmath331 too . we repeat this step until we can no longer right tilt about @xmath247 . the process must terminate : by the analogue of lemma [ tilting about stratum ] for right tilts any such sequence is equivalent to a single right tilt at some torsion theory and by remark [ nested remark ] longer sequences correspond to strictly larger torsion theories . since , by assumption [ assumption2 ] , there are only finitely many torsion theories we can not continue indefinitely . when we can no longer right tilt about @xmath247 then we have reached @xmath1 . hence @xmath332 for some @xmath63 . by construction @xmath333 for @xmath66 so @xmath291 as required . naively , one might hope that the closure of @xmath0 is obtained by allowing the central charges of arbitrary subsets of simples to tend to real values , i.e. that @xmath334 stratified in the obvious fashion . however , there are constraints on the ways in which central charges of simples can degenerate to real values . these arise because the set of stability conditions for which an object is semi - stable is closed ( * ? ? ? * proposition 8.1 ) and the central charge of a semi - stable must be non - zero . thus , for example , if @xmath335 is short exact ( and not split ) where @xmath336 are distinct simples then @xmath135 becomes semi - stable as @xmath337 and @xmath338 . hence degenerations to the boundary of @xmath0 with @xmath337 and @xmath339 are forbidden . these constraints mean that certain hyperplanes are excised from the naively expected strata of codimension @xmath340 . ( there are no constraints of this kind when a set of central charges of simples all become either positive or negative reals . ) [ constraints ] under assumption [ assumption2 ] , a stability condition @xmath212 in the boundary of @xmath11 is determined by allowing the central charge of a stability condition in the interior of @xmath0 to degenerate in such a way that the charges of a non - empty set of simples become real . the only constraint on this degeneration is that there is no @xmath64 with @xmath341 which is semi - stable for each of a sequence of stability conditions in @xmath0 with central charges @xmath342 . it follows from ( * ? ? ? * lemma 5.2 ) that any stability condition in the boundary of @xmath11 corresponds to such a degeneration of the central charge . it remains only to prove that there are no other constraints . let @xmath343 be the limit of a sequence of central charges of stability conditions in @xmath0 . suppose that @xmath344 where @xmath321 are simples in @xmath1 , that @xmath345 for any other simple @xmath144 in @xmath1 and that @xmath346 for any @xmath135 which is semi - stable for a sequence of stability conditions in @xmath0 whose central charges limit to @xmath226 . start at @xmath1 and repeatedly right tilt at simples @xmath144 for which @xmath347 . by the same argument as in the proof of corollary [ intersection corollary converse ] this process terminates at a heart @xmath348 where @xmath63 is a torsion theory in @xmath1 . then @xmath226 is the limit of central charges of stability conditions in @xmath4 . this implies @xmath349 for any @xmath327 . if in addition @xmath142 is simple then we must have @xmath350 , otherwise we could right tilt at @xmath142 . it follows that @xmath350 for any @xmath327 . recall that @xmath351 \rangle$ ] where @xmath352 . it follows immediately that @xmath353 for any @xmath66 and @xmath354)=-\mathcal{z}(f ) \in { \mathbb{h}}\cup \{0\}$ ] for any @xmath355 . on the other hand @xmath356 because @xmath357 so @xmath358 . we now show @xmath359 for any @xmath327 . it then follows that @xmath226 is the central charge of a stability condition in @xmath4 and that this stability condition is in the boundary of @xmath11 , i.e. this degeneration of the central charge corresponds to a degeneration of stability conditions as claimed . first suppose @xmath360 and @xmath361 . consider a short exact sequence @xmath362 in @xmath1 . since @xmath142 is torsion so is @xmath252 . it follows that @xmath363 and @xmath364 . in fact we must have @xmath365 for some quotient @xmath252 . if not , then @xmath366 and @xmath367 for all such short exact sequences and this implies that @xmath142 is semi - stable for all stability conditions in @xmath0 whose central charges are close to @xmath226 . by assumption @xmath368 in this situation , which is a contradiction . therefore , we can choose a quotient @xmath252 with @xmath365 and repeat the argument . after finitely many steps we obtain a simple quotient @xmath144 with @xmath369 which again contradicts the assumptions on @xmath226 . we conclude that @xmath370 . the second case is when @xmath371 $ ] and @xmath361 . a similar argument , involving passing to subobjects rather than quotients , leads to a contradiction . hence @xmath368 when @xmath142 is in @xmath63 or in @xmath95 $ ] . the result follows . this gives us a good understanding of the closure @xmath11 . it also follows from the proof that @xmath372 and hence that @xmath373 [ component thm ] under assumption [ assumption2 ] the union @xmath374 is a component of the space of stability conditions . we show that @xmath374 is both open and closed as a subset of the space of stability conditions . let @xmath375 and consider the ball @xmath376 . suppose @xmath377 , say @xmath378 is a stability condition in the intersection . consider the torsion theory ( recall lemma [ stab - fam - tor ] ) @xmath379 in @xmath3 with free theory @xmath380 . let @xmath381\rangle$ ] . it is clear that @xmath382 $ ] for any @xmath383 . since @xmath384 we have @xmath385 for any @xmath386 . hence @xmath387\cap{\mathcal{d}}_{\mathcal{a}}^\perp[1]$ ] , or equivalently @xmath388 $ ] , and there is a torsion theory @xmath305 in @xmath1 for which @xmath306 . therefore @xmath389}l_{{\mathcal{t}}'}{\mathcal{a}}\ ] ] is obtained from @xmath1 by tilting twice , first left and then right . ( of course one or both of these tilts may be trivial . ) under assumption [ assumption2 ] it follows that @xmath390 for some @xmath91 . so @xmath391 and in particular @xmath374 is open . furthermore if @xmath392 then @xmath393 is in @xmath376 for some @xmath247 . by the above , it is in @xmath374 which is therefore closed . assumption [ assumption2 ] guarantees that the tiling of the component containing stability conditions with heart @xmath1 by the subsets @xmath394 is locally - finite . the results of this section depend upon assumption [ assumption1 ] and , for the most part , on the stronger assumption [ assumption2 ] . we make some remarks on two special situations in which the property of having finitely many indecomposables , or the weaker property of having finite length and finitely many simples are inherited by a tilted heart . we will see examples of each in the next section . for the remainder of this section we assume that @xmath5 is @xmath276-linear for some field @xmath276 and of finite type , i.e. @xmath395 is finite dimensional for any objects @xmath17 and @xmath396 . [ spherical remarks ] the first special situation arises when the simple @xmath144 is a _ spherical object_. this means that there is some @xmath397 , the _ dimension _ of the spherical object @xmath144 , such that @xmath398 if @xmath399 or @xmath400 and vanishes otherwise , and that for any @xmath16 the pairing @xmath401 is perfect . a spherical object defines an automorphism , often called a twist , @xmath402 of @xmath5 characterised by the triangle @xmath403,\ ] ] where the first map is evaluation , see @xcite . suppose that @xmath404 and that for @xmath405 @xmath406 for any simple @xmath407 in @xmath1 . then by considering the long exact sequence obtained from ( [ twist formula ] ) upon applying @xmath408 we see that @xmath409 for @xmath399 and @xmath62 . hence , by lemmas [ simplicity criterion 1 ] and [ simplicity criterion 2 ] , we have @xmath410 . furthermore @xmath411 $ ] so that when @xmath412 the tilted heart @xmath122 is the image of @xmath1 under the automorphism @xmath402 . thus it inherits all the properties of @xmath1 , in particular if @xmath1 has finitely many indecomposables then so does @xmath122 . clearly we can repeat this construction starting from the tilted heart with @xmath131 $ ] in place of @xmath144 . [ excellent remarks ] the second special situation arises when we have an _ excellent collection _ in the sense of @xcite ( where the full details of what follows can be found ) . this is a full , strong exceptional collection @xmath413 of objects in @xmath5 with an additional property that ensures that all mutations of the collection are also excellent . it follows that the braid group acts on the set of excellent collections of length @xmath2 in @xmath5 by mutations . each excellent collection in @xmath5 determines a faithful @xmath12-structure whose heart we denote @xmath414 . this heart has finite length and precisely @xmath2 simples , which have a canonical ordering , say @xmath415 in which @xmath416 unless @xmath417 . proposition 3.5 of @xcite states that , for @xmath418 , @xmath419 where @xmath420 is the @xmath91th standard generator of the braid group . in particular the tilted heart @xmath421 also arises from an excellent collection and so we can repeat the process . note , that this theory tells us nothing about the tilt at the first simple @xmath422 , so that it only guarantees a partial version of assumption [ assumption1 ] ( and does not help in verifying assumption [ assumption2 ] at all ) . a more refined situation is discussed in @xcite , in which one has control over tilts at all simples so that assumption [ assumption1 ] holds , but this does not apply to either of the examples in the next section . our main example , in [ constructible ] , is the constructible derived category of @xmath13 stratified by a point and its complement . the perverse sheaves are the heart of a @xmath12-structure on this category which satisfies assumption [ assumption2 ] , although this is only apparent after some computation . we therefore obtain a combinatorial description of a component of the space of stability conditions as a locally - finite tiling by subsets @xmath423 . this description enables us to show that this component is isomorphic to @xmath424 as a complex manifold , and to compute the subgroup of automorphisms of the category which fix this component . as a counterpoint we consider stability conditions on the coherent derived category of @xmath13 in [ coherent ] . the space of these was computed in @xcite and is also isomorphic to @xmath424 . the kronecker heart satisifes assumption [ assumption1 ] but not the stronger assumption [ assumption2 ] . the simple tilting process starting from this heart leads to a tiling of a dense subset of the space of stability conditions , but not of an entire component . the tiling is very different from that for the constructible sheaves , for example it is not locally - finite . this is a reflection of the fact that the kronecker heart is of tame type ( i.e. has one - dimensional families of indecomposables ) , rather than of finite type as are the hearts arising in the constructible derived category . let @xmath425 be @xmath13 stratified by a point @xmath426 and its complement @xmath427 and let @xmath428 be the constructible ( with respect to this stratification ) derived category of sheaves of complex vector - spaces on @xmath425 . the first task is to construct some @xmath12-structures on @xmath428 . both strata are simply - connected so the only @xmath12-structures on @xmath429 and @xmath430 are the usual ones ( with heart the local systems ) and their shifts . let @xmath431 and @xmath432 be the inclusions . then there are functors @xmath433 with respective left and right adjoints @xmath434 and @xmath435 and @xmath436 and @xmath437 obeying the usual identities , see e.g. @xcite . thus we have ` gluing data ' and given a _ perversity _ , i.e. a pair @xmath438 indexing @xmath12-structures on @xmath430 and @xmath429 , we can construct a @xmath12-structure on @xmath428 with @xmath439 see @xcite . denote the heart of this @xmath12-structure by @xmath440}}$ ] . verdier duality acts on these @xmath12-structures via @xmath441 so that @xmath442 } } = { { \mathcal{p}}^{-r}_{}{[-s]}}$ ] . the category @xmath443}}$ ] has finite length and two simple objects ( one for each stratum ) . the simple objects in @xmath443}}$ ] are @xmath444 $ ] and @xmath445 & \textrm{if } & r>0\\ { \mathbb{c}}_x[s - r+1 ] & \textrm{if } & r=0 \\ \jmath_*{\mathbb{c}}_u[s - r+1 ] & \textrm{if } & r<0.\end{aligned}\ ] ] for @xmath446 this follows from ( * ? ? ? * chapter 7 , proposition 1.10.1 ) . for @xmath447 if @xmath448}}$ ] then the cohomological vanishing conditions imply @xmath449 so that the triangle @xmath450\ ] ] splits to give @xmath451 . it is then clear that the simple objects are as described . a verdier dual argument applies in the case @xmath452 . the category @xmath443}}$ ] is semi - simple if @xmath453 . when @xmath454 the heart is ( up to a shift ) either the category of constructible sheaves on @xmath425 ( when @xmath455 ) or its verdier dual ( when @xmath456 ) . in either case the heart is equivalent to the representation category of the quiver @xmath457 with two vertices and one arrow . the interesting case ( the only one in which the @xmath12-structure is faithful ) is when @xmath458 in which the heart is , up to a shift , the category @xmath459 of perverse sheaves on @xmath425 . this is equivalent to the category of representations of the quiver with relations @xmath460 : @xmath461_c & \cdot \ar@/{}_{.5pc}/[l]_v } \qquad vc=0,\ ] ] see , for example , ( * ? ? ? * chapter 7 , 2.6 ) . the auslander reiten quiver of the representation category is shown below . its vertices are the ( isomorphism classes of ) the indecomposable representations and its arrows correspond to irreducible maps between them , i.e. to maps which can not be written as composites except where one factor is an isomorphism . the dotted lines indicate the almost - split short exact sequences . see ( * ? ? ? * chapter vii ) for more details on auslander reiten quivers . @xmath462 & & \\ & \jmath _ ! { \mathbb{c}}_u[1 ] \ar[dr ] \ar[ur ] \ar@{--}[rr ] & & \jmath_*{\mathbb{c}}_u[1 ] \ar[dr ] & \\ { \mathbb{c}}_x \ar[ur ] \ar@{--}[rr ] & & { \mathbb{c}}_x[1 ] \ar[ur ] \ar@{--}[rr ] & & { \mathbb{c}}_x . } \ ] ] the simple object @xmath463 $ ] is spherical of dimension @xmath464 , and furthermore @xmath465,{\mathbb{c}}_x ) } = 0 = { { \mathrm{ext}^{i}}({\mathbb{c}}_x,{\mathbb{c}}_x[1])}\ ] ] unless @xmath466 . hence , by the remarks on page , the twist @xmath467}$ ] satisfies @xmath468 } { \mathcal{p}}^0.\ ] ] we write @xmath469 for this twist because geometrically it arises from a dehn twist . nadler and zaslow @xcite construct an equivalence @xmath470 between the real - analytically constructible derived category of a manifold @xmath471 and the derived asymptotic fukaya category of its cotangent bundle . ( the objects of the asymptotic fukaya category are lagrangians which are legendrian at @xmath472 , i.e. asymptotic to a union of cotangent fibres , hence the name . ) the equivalence takes a constructible sheaf to a lagrangian smoothing of its characteristic cycle . taking @xmath473 we can restrict to obtain an equivalence between @xmath428 and the full subcategory of @xmath474 generated by the cotangent fibre @xmath475 and zero section , which are the characteristic cycles of @xmath476 and @xmath477 respectively . dehn twisting about the zero section gives an automorphism of @xmath474 . algebraically this automorphism is given by twisting about the corresponding object @xmath463 $ ] in @xmath428 . using these properties we can show that the set of hearts @xmath481}}$ ] for @xmath482 is closed under simple tilts . all of these hearts are distinct , except for those identified by the third property . each of these hearts is either semi - simple , isomorphic to the constructible sheaves or to the perverse sheaves . thus assumption [ assumption2 ] is satisfied . by theorem [ component thm ] this is a list of the hearts of stability conditions in a component of the space of stability conditions . [ diagram explanation ] we describe the closure @xmath483 of the set of stability conditions with heart the perverse sheaves ( other cases are similar ) . the strata of codimension @xmath484 are described by proposition [ tiling ] ; we focus on the strata of codimension @xmath464 . by proposition [ constraints ] points of these correspond to degenerations @xmath226 of the central charge for which the charges of both simples become real , and the charge of no semi - stable vanishes . these limiting central charges correspond to points in the real plane coordinates @xmath485 and @xmath486)$ ] with certain points deleted , see figure [ bdy picture ] . the deleted points form half - lines ; we label these with the object which is semi - stable for stability conditions with nearby central charges and whose charge vanishes on that half - line . each of the codimension @xmath464 strata in @xmath483 is the ( unique ) codimension @xmath464 stratum in @xmath4 for some right tilt @xmath487 , and each such right tilt occurs . in figure [ bdy picture ] the half - lines divide the plane into regions corresponding to codimension @xmath464 strata and each is labelled by the corresponding right tilt . hearts labelling adjacent regions are related by simple tilts the transition in the direction of the arrows is given by right tilting at the object labelling the corresponding half - line . in particular the two simples in the heart corresponding to a region are the two objects labelling the bounding half - lines , but with objects on inward pointing arrows shifted by @xmath24 $ ] . explicit stability conditions with central charges and hearts as indicated can be constructed , thus verifying the results of [ main results ] in this case . [ c2thm ] there is a free action of @xmath490 on the component @xmath488 of the space of stability conditions , where @xmath244 and @xmath469 act as described on p. the quotient is isomorphic to @xmath491 and @xmath492 . the method is similar to that in @xcite . since @xmath469 has infinite order and does not preserve any of the hearts it acts freely . the action of @xmath244 is free and preserves semi - stables , whereas @xmath469 changes them , so the action of @xmath493 is also free . each orbit contains a unique stability condition in the subset satisfying to see this note that any orbit intersects @xmath499 . for @xmath500 both @xmath476 and a shift of @xmath494 $ ] are simple in @xmath501 and hence semi - stable . for the same reason @xmath476 is semi - stable in @xmath489 but in this subset @xmath494 $ ] is semi - stable if and only if @xmath502 ) \geq \varphi({\mathbb{c}}_x).\ ] ] however , if @xmath497 ) < \varphi({\mathbb{c}}_x)$ ] then applying @xmath469 we get a stability condition in @xmath503 for which @xmath497 ) > \varphi({\mathbb{c}}_x)$ ] and it follows that @xmath476 and @xmath494 $ ] are semi - stable . thus each orbit contains a stability condition for which @xmath476 and @xmath494 $ ] are semi - stable . using the action of @xmath244 we can rotate and rescale so that the second condition holds . conversely , when the first condition is satisifed the heart of the stability condition must be one of the @xmath501 for @xmath504 or @xmath505 . furthermore if it is @xmath459 or @xmath505 then @xmath502 ) \geq \varphi({\mathbb{c}}_x)\ ] ] ( otherwise @xmath494 $ ] is not semi - stable ) . it follows that @xmath497)\geq 1 $ ] when the first two conditions are satisifed . for phases there are no restrictions . if the phase is @xmath62 then the heart is @xmath505 when @xmath498)\in ( 0,1)$ ] and @xmath459 when @xmath498)\in ( 1,\infty)$ ] . the case @xmath498 ) = 1 $ ] does not occur for this implies @xmath507 which is forbidden since @xmath477 is semi - stable . furthermore , if @xmath498)=m\in ( 1,\infty)$ ] then acting by @xmath508 we obtain a new stability condition , again satisfying the first two conditions and with @xmath497)= 1 $ ] , but with @xmath498)=1/m \in ( 0,1)$ ] . hence , as claimed , each orbit contains a stability condition satisfying all three conditions . it is easy to check that no orbit meets the subset satisfying the three conditions in more than one point . we identify @xmath509 with @xmath491 by first mapping a stability condition satisfying the three conditions to @xmath510 ) + i \pi \varphi(\jmath_!{\mathbb{c}}_u[1 ] ) \in \ { x+iy \in { \mathbb{c}}\ | \ y > \pi \ \textrm{or}\ y=\pi , x<0 \}\ ] ] and then applying @xmath511 . this is holomorphic because the complex structure on a space of stability conditions comes from that on the space of central charges . since @xmath512 $ ] is simple in @xmath501 when @xmath513 we can choose @xmath514 satisfying the three conditions for which the mass and phase of @xmath494 $ ] take any given values in @xmath515 and @xmath516 $ ] respectively . we can also find @xmath517 satisfying the three conditions such that @xmath494 $ ] has phase @xmath62 and any given mass in @xmath265 . it follows that the above map is surjective . furthermore , the induced gluing of the boundary is exactly that arising from the group action . and verdier duality on figure [ graph ] . shift moves the diagram one place to the right . duality rotates it by @xmath519 about the perverse sheaves , swapping the upper and lower layers , and @xmath469 rotates by @xmath519 about the central spine and shifts by half a place to the left . shift and @xmath469 preserve tilting relationships whereas duality reverses them . , width=264 ] as an application we determine the subgroup of the automorphism group of the category which preserves the component @xmath488 . ( it is conjectured that spaces of stability conditions are connected ; if this were known we could determine the entire automorphism group . ) it is convenient to determine the subgroup of @xmath520-automorphisms first , i.e. to allow functors such as @xmath24 $ ] as well as triangulated automorphisms . suppose @xmath242 is a @xmath520-automorphism . since @xmath242 must map @xmath0 to @xmath521 we can check from the combinatorial description that @xmath522 $ ] preserves the perverse sheaves for some @xmath302 and @xmath2 . an automorphism which preserves the perverse sheaves induces an action on the auslander reiten quiver . in this case it must act as the identity on the vertices , as there are no other symmetries ; up to a natural isomorphism we may assume it fixes the indecomposable perverse sheaves . since there is at most one irreducible map between each pair of indecomposables , the automorphism can only act by rescaling each of these maps by an element of @xmath491 . by a further natural isomorphism we can assume these rescalings are all trivial except for one , for concreteness say that of the map @xmath523 \to { \mathbb{c}}_x$ ] . ( we use the fact that the sum of the two composites from @xmath494 $ ] to @xmath523 $ ] is zero here . ) therefore the subgroup of automorphisms ( up to natural isomorphism ) which preserve the perverse sheaves is isomorphic to @xmath491 . hence @xmath524 with @xmath491 acting trivially on the space of stability conditions and @xmath525 generated by @xmath469 and @xmath24 $ ] . the _ bona fide _ automorphisms are the subgroup @xmath526 . this description makes it easy to identify the serre functor ; it is @xmath527 . let @xmath528 be the coherent derived category of @xmath13 . the space of stability conditions was computed in @xcite and is isomorphic to @xmath424 . we use the notion of excellent collection see the remarks on page and @xcite in this case @xmath529 . the corresponding heart @xmath530 is equivalent to the representation category of the kronecker quiver @xmath531 \ar@/{}^{.5pc}/[r ] & \cdot } \ ] ] and we refer to it as the kronecker heart . it has two simple objects @xmath532 and @xmath533 $ ] ( in that order in the canonical ordering ) . right tilting at the first leads to the heart corresponding to the mutated collection : @xmath534 write @xmath535 for @xmath536 . repeating we see that @xmath537 . hence by corollary [ intersection corollary ] @xmath538 and in fact is of ( real ) dimension two . this yields the picture of the ( infinitely many ) codimension two strata in @xmath539 shown in figure [ kronecker bdy ] . assumption [ assumption2 ] is false for the kronecker heart , and we do not obtain an entire component of the space of stability conditions by tilting from it . there is a stability condition with heart @xmath541 and central charge @xmath542 . the semi - stables of phase @xmath227 are the semi - stable ( in the usual sense ) coherent sheaves of slope @xmath543 . rotating this central charge we obtain stability conditions in @xmath544 ( * ? ? ? * proposition 2.4 ) . it follows that these stability conditions are in the same component . however , the coherent sheaves are not obtained from the kronecker heart by any _ finite _ sequence of simple tilts , although they are the left tilt of it at the torsion theory @xmath545 \ | \ d\geq 1 \rangle$ ] . this torsion theory contains infinitely many indecomposables , but it is still true that @xmath546 and in fact has the expected codimension , two . to see this we note that the standard stability condition on the coherent sheaves degenerates to one in which the phase of @xmath547 is @xmath548 for @xmath549 and @xmath62 for @xmath550 ( and all @xmath547 remain semi - stable ) . the heart is therefore the kronecker heart . we can freely choose the masses of @xmath532 and @xmath533 $ ] in this degeneration and so obtain a ( real ) dimension two intersection as claimed . it is not however true that corollary [ intersection corollary converse ] holds without any assumptions on the torsion theory . from figure [ kronecker bdy ] we can see that @xmath551)\cap \overline{u({\mathcal{a}}_0 ) } = \emptyset\ ] ] even though @xmath552 $ ] for @xmath553\ |\ d\geq 1\rangle$ ] . the degenerations of central charges which ` should ' give stability conditions in the intersection are forbidden because the charges of semi - stable ( shifted ) torsion sheaves vanish . if we allow hearts obtained from a given one by any finite sequence of tilts ( rather than just simple tilts ) then we obtain an entire component of the space of stability conditions in this example . however , some tilts lead to hearts which are not the heart of any stability condition , see ( * ? ? ? * remark 3.5 ) . proposition [ constraints ] still holds degenerations of the central charge for which the charge of no semi - stable vanishes lift to degenerations of stability conditions even though the given proof is invalid .

we study the collective modes of a low - energy continuum model of the quasi - two - dimensional electron liquid in a layer of the organic compound(bedt - ttf)i in a perpendicular magnetic field . as testified by zero magnetic field transport experiments and _ ab initio _ theory , this material hosts both massless and massive low - energy carriers , the former being described by tilted and anisotropic dirac cones . the polarizability of these cones is anisotropic , and two sets of magnetoplasmon modes occur between any two cyclotron resonances . we show that the tilt of the cones causes a unique intervalley damping effect : the upper hybrid mode of one cone is damped by the particle - hole continuum of the other cone in generic directions . we analyse how the presence of massive carriers affects the response of the system , and demonstrate how doping can tune(bedt - ttf)i between regimes of isotropic and anisotropic screening .

1406.7081

the layered organic conductor @xmath0(bedt - ttf)@xmath1i@xmath2 , an intensively investigated member of the ( bedt - ttf)@xmath1i@xmath2 family,@xcite recently enjoys renewed interest due to the presence of dirac cones in the low - energy band structure under high hydrostatic pressure ( above 15 kbar)@xcite or under uniaxial strain ( above 3 kbar along the @xmath3-axis).@xcite ( for reviews , see refs . , , and ) . this material is a bulk crystal with a quasi - two - dimensional ( 2d ) character , as it consists of weakly coupled conductive bedt - ttf [ bis(ethylenedithio)-tetrathiafulvalene ] layers and insulating i@xmath4 anion layers . the existence of the gapless conical valleys at the fermi energy under sufficiently high hydrostatic pressure or uniaxial strain is testified by three experimental findings : ( i ) the conductivity parallel to the layers is roughly constant from room temperature down to about 2 k , while both the carrier density ( measured via the hall coefficient ) and the mobility change by almost six orders of magnitude in opposite directions.@xcite ( ii ) the conductivity perpendicular to the layers is greatly enhanced by a perpendicular magnetic field , and shows a complex dependence on the parallel magnetic field component.@xcite ( iii ) the shubnikov - de haas oscillations and the integer quantum hall effect have been observed recently.@xcite these observations have been interpreted in terms of spin- and valley - degenerate massless dirac fermions , assuming that only two exterior layers are hole - doped by the substrate . notice that none of the above observations are particularly sensitive to the geometry of the dirac cones ; both the density of states and the landau level spectrum are qualitatively identical for generic cones and upright isotropic ones , which naturally arise in the low - energy spectrum of graphene@xcite and on the surface of topological insulators.@xcite our knowledge of the geometric properties of the conical valleys in @xmath0(bedt - ttf)@xmath1i@xmath2 mainly comes from theory , e.g. , tight - binding modeling of the highest occupied molecular orbitals at 3/4 filling,@xcite extended hubbard model calculations,@xcite and _ ab initio _ band structure calculations.@xcite on the other hand , theory also predicts@xcite a quadratic band maximum at the @xmath5 point of the first brillouin zone , and transport measurements by monteverde _ _ et al.__@xcite have identified carrier densities @xmath6 @xmath7 ( electrons ) and @xmath8 @xmath7 ( holes ) in the linear and the quadratic pockets , respectively , which indicates an accidental overall hole - doping of their samples . the presence of a hole - doped quadratic band maximum is , at first glance , at odds with findings ( iii ) above,@xcite and difficult to reconcile with ( i ) , although the smaller mobility of the massive carriers may be important in this connection . hence we believe a discussion of properties that can in principle probe the presence of massive carriers is timely . recent theoretical surveys by mori@xcite have identified several other potential massless dirac materials among layered organic conductors . in some of them , @xmath9(bedt - ttf)@xmath1i@xmath2 , there might be a tiny gap.@xcite in other cases , e.g. , @xmath10-(bedt - ttf)@xmath1khg(scn)@xmath11 , either the apex of the cones is far off the fermi energy , and the contact points give rise to electron - hole pockets due to overtilt if the contact points approach the fermi energy under pressure or strain , or exhibit charge - density - wave structures that are incompatible with the zero - gap property in the accessible range of the phase diagram.@xcite one material , @xmath10-(bedt - ttf)@xmath1nh@xmath11hg(scn)@xmath11 has no charge order and possesses dirac points that reach the fermi energy under pressure,@xcite but the linear bands become isotropic ( like in graphene ) , making it less interesting from the point of view of tilt - related electronic properties we are focusing on in this study . here , we study the particle - hole and collective excitations of a continuum model of the quasi-2d electron gas in @xmath0(bedt - ttf)@xmath1i@xmath2 in a magnetic field . the electronic coulomb interaction is taken into account within the random phase approximation ( rpa ) . our continuum model includes both generic massless dirac valleys and a massive one , such as those that are expected in the vicinity of the fermi energy from the topmost two bands in @xmath0(bedt - ttf)@xmath1i@xmath2 . we restrict our attention to the low - energy and low - momentum features of the density - density response , for which the structure of the highest two bands beyond these valleys is inessential . to be specific , we will use the band parameters of @xmath0(bedt - ttf)@xmath1i@xmath2 under high pressure from the literature , and focus on the qualitative features that stem from the anisotropy and the tilt of the dirac cones and the presence of massive low - energy carriers . thus we extend the zero magnetic field analysis by nishine _ et al._,@xcite who , neglecting the massive carriers , have found two plasmon modes , and interpreted the appearance of the second mode as a consequence of plasmon filtering.@xcite our work generalizes analogous theoretical studies of plasmonic excitations in graphene.@xcite in graphene in a perpendicular magnetic field , apart from the upper hybrid mode ( uhm ) , which is the zero - field plasmon mode modified by the magnetic field,@xcite one can readily identify linear magnetoplasmons , which run parallel to the boundary that separates the intra- and interband excitations in the frequency - momentum plane . although the latter mode is essentially the coalescence of broadened interband excitations , its huge spectral weight justifies its identification as a distinct collective mode of massless dirac electrons . the uhm starts in the forbidden region at low momenta and merges into the linear magnetoplasmon mode in the interband particle - hole excitation region . we are particularly interested in the effects of the anisotropy and the tilt of the dirac cones , and include the massive holes to assess the relative importance of the two carrier types in the response . indeed , the valley degeneracy of the plasmonic modes and the uhm in the presence of a magnetic field is lifted due to the tilt of the dirac cones in opposite directions , in contrast to graphene , which has isotropic and upright dirac cones . most saliently , the coulomb coupling between the different uhms , which we investigate in the rpa , reveals a particular _ intervalley damping _ effect that we discuss in contrast to a single - valley approximation , where coulomb interaction is taken into account only within a single cone and the plasmonic excitations can be associated to a particular valley . the same mechanism must be present in other 2d systems with valleys of _ tilted _ dispersion relations . the article is structured as follows . in sec . [ adatok ] we review the low - energy band structure of @xmath0(bedt - ttf)@xmath1i@xmath2 , introduce the three - valley model , and discuss the structure of the landau states for all carriers . in sec . [ interactions ] we briefly comment on the interaction coupling in the distinct valleys . in sec . [ methods ] we describe the methods and approximations involved in the calculation of the density - density response and the dielectric function . section [ results ] discusses the results . we provide a summary and discuss the possibly relevant experimental probes in sec . [ conclu ] . details of the calculation are delegated to the appendices . @xmath0(bedt - ttf)@xmath1i@xmath2 has a triclinic crystal structure with strongly pressure - dependent lattice parameters.@xcite regarding one layer as a 2d crystal , the unit cell is oblique , with the primitive vectors enclosing an angle @xmath12 . the length of the primitive vectors change monotonically from @xmath13 and @xmath14 to @xmath15 and @xmath16 as the pressure is increased from ambient pressure to 17.6 kbar.@xcite see fig . [ unitcell ] for a sketch of the first brillouin zone with the distinguished points and directions we refer to later . schematic view of the first brillouin zone of @xmath0(bedt - ttf)@xmath1i@xmath2 . the high symmetry points , the location of the massive valley ( squares ) , and that of the dirac cones ( dots ) are indicated . the arrow from point @xmath17 indicates the smallest steepness of cone @xmath17 . directions in the momentum plane will be related to this angle , @xmath18 . the dashed ellipse around @xmath17 represents an equipotential line . ] the low - energy band structure can be derived from a tight - binding model that involves the four relevant highest occupied molecular orbitals of the four different bedt - ttf molecules in the unit cell . among the four bands , only the upper two play a role as the filling is 3/4 at charge neutrality.@xcite these two bands may have contact points only in the absence of a stripe charge order.@xcite at ambient pressure the stripe charge order occurs up to a phase transition around 135 k,@xcite but it is suppressed by pressure or strain@xcite . in the non - stripe phase two tilted massless dirac cones form in the vicinity of the contact points at low - symmetry time - reversal related points @xmath19 and @xmath17 [ see fig . [ unitcell ] ] . at even higher pressure or strain , the dirac points coalesce at the @xmath20 point,@xcite where a gap opens . in addition to the linear valleys , there is a band maximum at the time - reversal symmetric point @xmath5 on the edge of the first brillouin zone , which may host massive holes . in the present work we use a _ three - valley model _ that keeps a pair of generic massless dirac cones of opposite tilt and a massive valley with appropriate momentum / energy cutoffs . naturally , this approach is justified only as long as we study low - momentum and low - energy features . we will show that this assumption is justified for our purposes because both the uhm and the linear magnetoplasmons occur within this range . the massless dirac carriers are suitably described by the minimal weyl hamiltonian@xcite using four parameters : @xmath21 where @xmath22 ( @xmath23 ) represents the cone at @xmath17 ( @xmath19 ) . this hamiltonian obviously respects time - reversal symmetry , @xmath24 . hamiltonian @xmath25 has spin - degenerate bands , but the dispersions in valleys @xmath26 differ . the inclination of the dirac cone is determined by the combined effect of the tilt and the anisotropy . by the _ anisotropy _ of the dirac cone we mean the difference between @xmath27 and @xmath28 , and we characterize it by the parameter @xmath29 by its _ tilt _ we mean that the constant energy slices are not concentric because @xmath30 . for convenience , we will also use a rescaled and rotated coordinate system,@xcite defined by the transformation @xmath31 rescaling the @xmath32 coordinate removes the anisotropy . the rotation brings the @xmath33 coordinate in the tilt direction if we choose @xmath34 after some straightforward algebra and a unitary transformation,@xcite the weyl hamiltonian can be written as @xmath35 where we have introduced the dimensionless parameters @xmath36 to quantify the tilt . notice @xmath37 , and that @xmath38 corresponds to the case of graphene . to be specific , we will use kobayashi _ et al . _ s estimate of the velocity parameters@xcite : @xmath39 which yield @xmath40 , @xmath41 , @xmath42 , and @xmath43 . various other velocity values are available in the literature,@xcite and they change considerably under pressure.@xcite we emphasize that we will focus on qualitative features that hardly depend on this particular choice . the landau levels ( lls ) of weyl bands have been derived both semiclassically@xcite and from a full quantum mechanical treatment.@xcite the spectrum is reminiscent of graphene , albeit with a reduced effective fermi velocity : @xmath44 where @xmath45 is the magnetic length , @xmath46 is the applied magnetic field , and @xmath47 is an integer . the corresponding orbirals in the landau gauge are given in appendix [ linorbit ] . we consider the linear approximation valid around each of the dirac points separately in a circular region , whose radius is chosen as about 1/8 of the side of the first brillouin zone [ @xmath48 , consistently with the band structure obtained in refs . , and . using the velocities in eq . ( [ velo ] ) , the energy cutoff @xmath49 and the number of available lls @xmath50 are @xmath51}.\end{aligned}\ ] ] based on monteverde _ et al . _ s electron density data,@xcite the fermi wave vector in the linear valleys is tiny , @xmath52 m@xmath53 , using @xmath54 for valley and spin degeneracy . see fig . [ landaulevels ] for a schematic view of the landau level structure . notice that , while the massive valley breaks the particle - hole symmetry , the landau levels of the linear valleys are particle - hole symmetric . the restoration of particle - hole symmetry for the ( tilted ) massless carriers is due to the magnetic field , and it arises because the cyclotron motion covers the complete isoenergy contours of the dispersion relation . the effect of the tilt is therefore only to decrease the level spacing by the factor @xmath55 in eq . ( [ landaulevel]).@xcite the massive valley is centered around the @xmath5 point at the brillouin zone edge ( see fig . [ unitcell ] ) . it is taken as isotropic with an effective mass@xcite @xmath56 in terms of the free electron mass @xmath57 . the valley is hole - like , a paraboloid of revolution open from below . the hamiltonian of the massive band is @xmath58 and the landau level spectrum is given by @xmath59 where @xmath60 is the appropriate cyclotron frequency and @xmath61 is an integer . ( notice that these nonnegative integers actually number hole lls . ) combining the carrier densities measured by monteverde _ _ et al.__@xcite with eqs . ( [ velo ] ) and ( [ holemass ] ) , the top of the massive band is about @xmath62 above the dirac point of the massless valleys . the corresponding eigenstates in the landau gauge @xmath63 are @xmath64 where @xmath65 is a hermite polynomial . the quadratic approximation is valid in a circle around point @xmath5 in momentum space , with a radius that is estimated as 17.5 % of the side of the first brillouin zone . this yields the cutoff energy @xmath66 and the number of lls is @xmath67 as @xmath68}.\end{aligned}\ ] ] the fermi mometum @xmath69 of the massive band@xcite is @xmath70 m@xmath53 , using @xmath71 for spin degeneracy . the relative strength of the interaction for each carrier type is characterized by the ratio between the interaction energy scale @xmath72 and the kinetic energy scale @xmath73 at the characteristic length scale @xmath74 . here , @xmath75 is the relative dielectric constant of the material . the kinetic energy scale depends on the carrier type . for the massive carriers , @xmath76 , and the ratio @xmath77 depends on the fermi wave vector , @xmath78 where @xmath79 is the bohr radius . for the last number we have used the fermi momentum as estimated for a specific sample in subsec . [ massivemodel ] . this ratio is traditionally called the wigner - seitz radius . in contrast to the parabolic bands , the kinetic energy of massless dirac carriers scales in the same manner as the interaction energy , @xmath80 , hence there is no characteristic length such as the bohr radius . indeed , the ratio between interaction and kinetic energy is independent of the electron density , and it may be characterized by a `` fine structure constant of @xmath0(bedt - ttf)@xmath1i@xmath2 '' @xmath81 where @xmath82 is the fine structure constant of quantum electrodynamics , and @xmath83 is the speed of light . notice that the average fermi velocity @xmath84 m / s is an order of magnitude smaller than the corresponding velocity in graphene.@xcite this is the origin of the rather large value of the ratio between the interaction and the kinetic energy . in view of this high value of the energy ratio for both the massless and massive carriers in @xmath0(bedt - ttf)@xmath1i@xmath2 , one may expect the formation of correlated phases , such as the wigner crystal . for the conventional 2d electron gas ( 2deg ) , @xmath85 is required to reach the wigner crystal phase of the massive carriers.@xcite this would require in turn a dielectric constant @xmath86 . to the best of our knowledge there is no available experimental value for @xmath75 in @xmath0(bedt - ttf)@xmath1i@xmath2 , but we expect it to be such as to rule out the wigner crystal . in the linear valleys , the high value of the dielectric constant compensates for the small fermi velocity , in which case , just like in graphene , one would not expect an instability of the semimetalic phase.@xcite throughout the article we assume that the wigner crystal can be discarded , and use @xmath87 . we identify the collective modes in the random phase approximation . assuming that all but the topmost two bands are inert , the density - density response @xmath88 is determined by the bare polarizability @xmath89 as @xmath90 here @xmath91 is the bare coulomb interaction , and the bare propagator @xmath92 is a @xmath93 matrix in the original lattice model related to the amplitudes on the four bedt - ttf molecules in the primitive cell . reliable band structure information , however , is only available near the contact points of the topmost two bands . hence we use the three - valley model introduced in sec . [ adatok ] , and approximate the polarizability accordingly . in eq . ( [ chidef ] ) , @xmath89 picks up a contribution only if @xmath94 and @xmath95 specify one filled and one empty state . as we will restrict our attention to @xmath96 ev , we can ignore the cases when both of these points are outside the vicinities of the @xmath19 , @xmath17 and @xmath5 points , respectively , assuming that the fermi energy is near the contact points . we can also ignore the cases where the states @xmath94 and @xmath95 are in distinct valleys , as long as we focus on small momenta @xmath97 . this approximation is justified because the coulomb interaction intervenes in the dielectric function @xmath98 , and suppresses intervalley contributions in comparison to intravalley contributions at fixed @xmath99 by a factor @xmath100 near the characteristic fermi momentum @xmath101 . furthermore , for small doping @xmath102 , we can also ignore the transitions that involve a state in the range of validity of the linear / quadratic approximations ( subsecs . [ masslessmodel ] and [ massivemodel ] ) and a state outside of this domain . thus we can safely approximate the bare polarizability for our limited purposes as the sum of the contributions of intravalley particle - hole pairs : @xmath103 where @xmath104 is the polarizability contribution from intravalley transitions in valley @xmath105 , with @xmath106 . further , when studying features of a wave length much larger than the unit cell , the distribution of amplitudes in the unit cell can be ignored . using the wave functions in eq . ( [ masseigenstates ] ) for massive carriers , their contribution is@xcite @xmath107 where we have used the form factors of the 2deg , @xmath108 for @xmath109 , whereas for @xmath110 , @xmath111 here @xmath112 denotes an associated laguerre polynomial . similarly , using the wave functions in eqs . ( [ nagypszi]-[nagypszivege ] ) for massless carriers , the contribution from valley @xmath26 is @xmath113 the corresponding form factors @xmath114 are defined in appendix [ applinpol ] . using the expressions of the form factors , one can check that @xmath115 and @xmath116 are related by the change of the sign of the tilt @xmath117 . for illustration purposes , we consider occasionally the polarizability , the dielectric function and the density - density response functions stemming from only one or two valleys , even if such model systems are unphysical . thus @xmath118 and @xmath119 are defined in terms of @xmath120 in an obvious manner , where @xmath105 is either @xmath19 , @xmath17 , or @xmath121 for the respective valley . moreover , we also consider cases with only two valleys taken into account . sometimes this is a physically relevant situation , e.g. , when the system is electron - doped or charge - neutral and the magnetic field is large , sometimes a theoretical contrast . then we define @xmath122 and again @xmath123 and @xmath124 follow in analogy to eqs . ( [ chirpa ] ) and ( [ dielectricfunc ] ) . the particle - hole excitation spectrum ( phes ) and the collective modes of @xmath0(bedt - ttf)@xmath1i@xmath2 at low energies are determined by the two massless dirac cones and the massive hole pocket . the way the three valleys contribute to the density - density response of a layer depends on the doping and the perpendicular magnetic field . for significant electron doping the massive band is full at zero temperature and only the massless carriers contribute to the transport ( c.f . [ edoping ] ) . at charge neutrality , the chemical potential is between the dirac point and the top of the massive band in zero magnetic field . by turning on the magnetic field , the central four - fold degenerate @xmath125 ll of the massless valleys is fixed at the dirac points , but the energy @xmath126 of the topmost @xmath125 ll of the massive valley decreases . let @xmath127 be the field when the @xmath128-th ll of the massive valley coincides with the @xmath47-th ll of the conical valleys . for @xmath129 , the @xmath125 ll of the dirac valleys is half - filled and the completely electron - filled massive valley is inert , c.f . [ landaulevels](a ) . the response is entirely due to the massless valleys , similarly to the electron - doped case ( subsec . [ edoping ] ) . in the interval @xmath130 the chemical potential lies between the @xmath125 ll of the linear band and the @xmath125 ll of the massive band ( @xmath131 , @xmath132 ) , c.f . [ landaulevels](b ) . the excess electrons in the massless valleys ( two per flux quanta ) exactly compensate for the excess holes in the massive pocket . now all valleys contribute to the density - density response ( subsec . [ threevalleys ] ) . for @xmath133 , the @xmath134 ll of the massive valley becomes empty and the four - fold degenerate @xmath134 ll of the massless dirac valleys is half - filled ; the chemical potential is set somewhere in the ( naturally broadened ) @xmath134 ll of the massless dirac fermions , c.f . [ landaulevels](c ) . combining refs . and , we estimate @xmath135 t and @xmath136 t. schematic view of the landau level structure of the two dirac cones and the massive valley , with the chemical potential at charge neutrality indicated . ( a ) @xmath137 t , ( b ) @xmath130 , and ( c ) @xmath138 t. ] in significantly hole - doped samples all valleys contribute . with a significant electron - doping in mind , we set @xmath139 , for which @xmath140 , i.e. , the massive band is full and inert for all realistic values of the magnetic field @xmath46 . to highlight the direction - dependent effects , we first consider the response of a single tilted massless dirac cone , as described by the weyl hamiltonian in eqs . ( [ hamiltoni ] ) and ( [ hamiltoni_2 ] ) . although this model is not directly related to a concrete physical situation , it reveals some basic phenomena associated with the tilt of the dirac cones , such as the direction - dependent plasmonic dispersions and the characterization of the different regions of allowed and forbidden particle - hole excitations . this preliminary analysis within the single - cone approximation thus helps us understand the effect of the coulomb coupling between the two different valleys , discussed in sec . [ bothcones ] . regarding the possibility of particle - hole excitations , the @xmath141 plane can be divided into several regions . here we follow the notation nishine _ _ et al.__@xcite have introduced for the phes in zero magnetic field . the regions of possible intraband ( a - region ) and interband ( b - region ) particle - hole excitations are separated by the boundary line @xmath142 . due to the opposite tilt , the various regions and their boundaries differ for the two cones in any general direction . both regions are divided into subregions , as explained below . ( color online ) the regions and subregions of the @xmath141 plane from the point of view of ( a ) cone @xmath17 , and ( b ) cone @xmath19 in a particular direction @xmath143 . ( c , d ) cuts of cones @xmath17 and @xmath19 in the same direction . in the direction @xmath144 , the cones @xmath19 and @xmath17 are interchanged , such that the panels ( a , c ) would correspond to cone @xmath19 and ( b , d ) to cone @xmath17 . the distinguished energies @xmath145 , momentum @xmath146 , and asymptotes @xmath147 are indicated . the gray shading indicates the forbidden regions 1b and 3a , where there are no particle - hole excitations in the absence of electron - electron interactions . ] figures [ nishinetiltdirection](a ) and ( b ) depict the @xmath141 plane in the direction @xmath143 , where @xmath148 is the angle of the momentum . in this direction the steepness of cone @xmath17 is minimal , while that of cone @xmath19 is maximal . first consider cone @xmath17 in this direction [ figs [ nishinetiltdirection](a ) and ( c ) ] . zero - energy excitations must have a momentum transfer less than the major axis @xmath146 of the ellipsoidal fermi surface . excitations with higher momenta require a minimal energy , which defines the upper boundary @xmath149 of the forbidden subregion 3a [ c.f . fig [ nishinetiltdirection](a , b ) ] , where @xmath150 . excitations with zero momentum transfer must be interband . for @xmath151 , there are two special energies . the smallest one is denoted @xmath152 in fig . [ nishinetiltdirection ] . below this energy no @xmath151 excitation is possible . if we increase the total momentum of the particle - hole pair in the positive manner , we must move the hole in the @xmath153 direction [ see arrow at @xmath154 in fig [ nishinetiltdirection](c ) ] . then the excitation energy increases , which explains the rise of the boundary line @xmath154 for small momenta . by thus decreasing the hole s wave vector , one follows the asymptote @xmath155 in fig [ nishinetiltdirection](a ) , which ultimately merges into the boundary line @xmath156 . starting from @xmath157 , one can decrease the excitation energy with positive total momentum ; this branch corresponds to @xmath156 and follows asymptote @xmath158 at small momenta , which merges into @xmath154 in the intraband region . at larger values of the momentum , @xmath156 asymptotically follows @xmath155 . the reason why the boundaries @xmath154 and @xmath156 deviate from the asymptotes @xmath147 and do not intersect each other is due to the 2d nature of the excitations , i.e. , one needs to consider transitions outside of the one - dimensional cut examined so far . between @xmath154 and @xmath142 no particle - hole excitations exist , thus region 1b is forbidden , just like 3a for intraband excitations . if one considers excitations of cone @xmath19 in the same momentum direction , the shapes of boundaries @xmath156 and @xmath154 differ considerably [ fig [ nishinetiltdirection](b ) ] . the smallest energy of a @xmath151 excitation is still @xmath152 . however , in contrast to cone @xmath17 , it is possible to decrease the energy from @xmath152 by moving the hole in the @xmath153 direction [ fig [ nishinetiltdirection](d ) ] , and the corresponding branch defines boundary @xmath154 . starting from @xmath157 one can only increase the excitation energy with a negative total momentum , and one thus obtains the branch corresponding to @xmath156 . naturally , the two cones are related by time - reversal symmetry , which also changes the sign of @xmath159 . the role of the @xmath17 and @xmath19 cones are therefore interchanged in the @xmath160 direction , and the regions of cone @xmath17 ( @xmath19 ) are then those in fig . [ nishinetiltdirection](b , d ) [ fig . [ nishinetiltdirection](a , c ) ] . these zero - field considerations correctly describe the regions of allowed transitions also for non - zero magnetic fields since the wave vector in the phes is that of ( neutral ) electron - hole pairs . it therefore remains a good quantum number also in the presence of a magnetic field . figures [ egydirection1](a ) and ( b ) show @xmath161 and @xmath162 in the direction of @xmath17 s smallest velocity @xmath143 , where we have also plotted the boundaries @xmath163 , @xmath164 , @xmath165 , and @xmath166 , the latter corresponding to the frequency @xmath142 for the right and the left cone , respectively . notice that we consider a finite broadening ( inverse quasiparticle lifetime ) @xmath167 . whereas treating @xmath168 as an energy - independent constant is a crude approximation,@xcite we use it here only as a phenomenological parameter that renders the structure of the phes more visible . , the other parameters are @xmath169 t and @xmath170 . the first two panels consider the cones individually ; ( a ) shows @xmath161 in the direction of its maximal tilt @xmath18 ( or @xmath162 in the direction @xmath160 ) , and ( b ) @xmath162 in the direction @xmath18 ( or @xmath161 in the direction @xmath160 ) . panel ( c ) shows @xmath171 in direction @xmath18 , which is the response of the total system for electron doping , or for @xmath137 t at charge neutrality . the straight lines are the boundaries of regions relevant at @xmath172 ; c.f . [ nishinetiltdirection](a , b ) . notice that in panel ( c ) , we have only depicted the lines @xmath166 for the two cones , as a guide to the eye . ] , the other parameters are @xmath169 t and @xmath170 . the first two panels consider the cones individually ; ( a ) shows @xmath161 in the direction of its maximal tilt @xmath18 ( or @xmath162 in the direction @xmath160 ) , and ( b ) @xmath162 in the direction @xmath18 ( or @xmath161 in the direction @xmath160 ) . panel ( c ) shows @xmath171 in direction @xmath18 , which is the response of the total system for electron doping , or for @xmath137 t at charge neutrality . the straight lines are the boundaries of regions relevant at @xmath172 ; c.f . fig . [ nishinetiltdirection](a , b ) . notice that in panel ( c ) , we have only depicted the lines @xmath166 for the two cones , as a guide to the eye . ] , the other parameters are @xmath169 t and @xmath170 . the first two panels consider the cones individually ; ( a ) shows @xmath161 in the direction of its maximal tilt @xmath18 ( or @xmath162 in the direction @xmath160 ) , and ( b ) @xmath162 in the direction @xmath18 ( or @xmath161 in the direction @xmath160 ) . panel ( c ) shows @xmath171 in direction @xmath18 , which is the response of the total system for electron doping , or for @xmath137 t at charge neutrality . the straight lines are the boundaries of regions relevant at @xmath172 ; c.f . [ nishinetiltdirection](a , b ) . notice that in panel ( c ) , we have only depicted the lines @xmath166 for the two cones , as a guide to the eye . ] as shown in fig . [ egydirection1](a ) , the uhm is present with a considerable spectral weight in the originally forbidden region 1b of cone @xmath17 in its tilting direction @xmath143 . for cone @xmath19 , on the other hand , the spectral weight in region 1b in the same direction is definitely smaller [ fig . [ egydirection1](b ) ] . as shown in fig . [ epsilonfig ] , where we have plotted a zoom of @xmath162 in region 1b , the spectral weight could nevertheless be increased considerably by lowering the dielectric constant @xmath75 . however , a low spectral weight is expected in the physically relevant situation , for an estimate @xmath173 ( c.f . sec . [ interactions ] above ) . the reason is that the weakening of the bare interaction pushes the zeros of the rpa dielectric function to higher momenta at fixed energy ; eventually the uhm is forced to the borderline of region 1b for cone @xmath19 in the given direction . for cone @xmath17 the tendency is similar but less significant , c.f . [ egydirection1](a ) . . we show @xmath162 in the direction opposite to cone @xmath19 s maximal tilt , @xmath143 , for ( a ) @xmath174 , ( b ) @xmath175 and ( c ) @xmath176 . the physical parameters are the same as in fig . [ egydirection1 ] , whose panel ( b ) depicts @xmath170 . ] after leaving region 1b , the uhm merges into the linear magnetoplasmon mode in region 2b . the concentration of the spectral weight near @xmath142 is in accordance with the @xmath172 limit@xcite and the @xmath177 behavior of graphene.@xcite recall that the uhm arises due to the modification of the classical plasmons by cyclotron motion.@xcite it does not require interband excitations , thus it forms by a transfer of spectral weight from the intraband phes to the originally forbidden region 1b . the regular `` island '' structures , i.e. , the alternation of parts with high and low spectral weight in the @xmath141-plane , which is observable at high energies in fig . [ egydirection1 ] and subsequent figures , evolve from interband particle - hole excitations as the magnetic field is turned on . their structure in momentum space is related to the structure of high - index landau orbitals involved and it is exactly of the same origin as in the case of nontilted cones : the form factors in eqs . ( [ form1 ] ) to ( [ form2 ] ) contain laguerre polynomials with many zeros for transitions between deep - lying and high - energy landau levels.@xcite to summarize , the density - density response of a generic massless dirac cone in a perpedicular magnetic field exhibits an upper hybrid mode and linear magnetoplasmons with anisotropic velocities , as a plausible generalization of the graphene case.@xcite fig . [ egydirection1](c ) shows @xmath171 for both cones in a fixed direction @xmath143 of the momentum plane . notice that for the electron - doped case this is actually the total density - density response . the uhm of cone @xmath17 disappears , though it was the most dominant part of the response in the single - cone approximation . the uhm of cone @xmath19 is still present , though with a reduced spectral weight in its own forbidden region . its linear magnetoplasmon mode manifests itself , although at high energies ( @xmath17 s region 2b ) it is surrounded by the interband particle - hole excitations of cone @xmath17 , which have a modest spectral weight in the single cone approximation . both modes are approximately in the same place as they were when only one cone was considered . for interpretation , compare fig . [ nishinetiltdirection](c ) , where we sketch the @xmath141 plane in a fixed direction @xmath143 . the forbidden region of cone @xmath19 lies entirely in that of cone @xmath17 , hence no damping results from particle - hole excitations of either cones here . the picture is dramatically different for cone @xmath17 : its forbidden region overlaps with the damped region of cone @xmath19 , i.e. , the uhm of cone @xmath17 is strongly damped by particle - hole excitations in cone @xmath19 . for this reason , the uhm of cone @xmath17 disappears entirely . in the opposite direction the roles of @xmath19 and @xmath17 are , of course , interchanged . therefore , the interaction of the two cones leads to a strong direction - dependent damping , which results in the complete suppression of the uhm of one cone where the other has particle - hole excitations of high spectral weight . this phenomenon , which we refer to as _ intervalley damping _ , is also visible in the imaginary part of the dielectric function ( not shown ) . the linear magnetoplasmons of each cone are situated in the particle - hole continuum of their own phes respectively , where they are already damped . they do not overlap with each other in the @xmath141 plane in the shown direction @xmath143 . therefore , we expect the dominance of the one with higher spectral weight at a particular @xmath141 . figure [ egydirection1](c ) testifies the dominance of the linear magnetoplasmon mode of cone @xmath17 in this direction , in agreement with its higher spectral weight in the single cone model . this explains the reappearance of the linear magnetoplasmons at larger momenta , and the disappearance of the linear magnetoplasmons of cone @xmath19 for intermediate momenta . figure [ egydirection3 ] shows the density - density response in the direction @xmath178 , i.e. , perpendicular to the direction of the minimal steepness of cone @xmath17 ( maximal steepness of cone @xmath19 ) . the response of the individual cones are almost identical [ panels ( a ) and ( b ) ] , and their forbidden regions practically coincide ( perfect coincidence occurs at a nearby angle ) . intervalley damping is therefore absent in this direction . the uhm and the linear magnetoplasmon mode of the two - valley system [ panel ( c ) ] are where they would be for a single cone ; albeit with an increased amplitude . in order to illustrate the phenomenon of intervalley damping in a more precise and quantitative manner , consider the bare polarizability ( [ chi0teljes ] ) of a multivalley model , which is generically written as @xmath179 where the sum runs over all different valleys . one may thus rewrite the rpa dielectric function ( 21 ) as @xmath180 where @xmath181 is the rpa dielectric function within a single - valley model , where only the valley @xmath182 is taken into account . the single - valley model therefore yields a good approximation for the collective modes [ given by the zeros of @xmath183 if @xmath184 in the region of interest , i.e. for @xmath185 obtained from the solution @xmath186 . this precisely means that the spectral weight for particle - hole excitations in the other valleys @xmath187 vanishes in this region , or else that the collective modes of @xmath182 survive in the forbidden regions of the other cones , as observed in our calculations . notice that the phenomenon of intervalley damping is absent if all individual polarizabilities @xmath188 are identical , e.g. , in the absence of a tilt of the ( albeit anisotropic ) dirac cones . in this case , the rpa dielectric function ( [ eq : epsrpa ] ) simply becomes @xmath189 for @xmath190 identical valleys . therefore , @xmath191 has only a single zero , at a slightly larger frequency as compared to the single - valley approximation because of the enhanced coupling @xmath192 . this situation is encountered , e.g. , when one takes into account the spin degeneracy in conventional electron systems ( @xmath193 ) or in graphene with non - tilted dirac cones with a fourfold spin - valley degeneracy ( @xmath194 ) . the tilt of the dirac cones , or more generally the broken @xmath195 symmetry in a single valley , is thus the basic ingredient for the mechanism of intervalley damping . finally , we note that nishine _ _ et al.__@xcite saw the effect of intervalley damping at zero magnetic field , interpreted it as plasmon filtering,@xcite and offered a detailed analysis of the angular dependence of strength of the lower - lying plasmon mode . the two mechanisms are essentially the same , despite the slightly different formulations . , ( b ) @xmath196 , ( c ) @xmath197 in a fixed direction @xmath178 on the momentum plane . @xmath169 t , @xmath139 , @xmath170 . the lines denote the same boundaries as in figs . [ nishinetiltdirection ] and [ egydirection1 ] . ] , ( b ) @xmath196 , ( c ) @xmath197 in a fixed direction @xmath178 on the momentum plane . @xmath169 t , @xmath139 , @xmath170 . the lines denote the same boundaries as in figs . [ nishinetiltdirection ] and [ egydirection1 ] . ] , ( b ) @xmath196 , ( c ) @xmath197 in a fixed direction @xmath178 on the momentum plane . @xmath169 t , @xmath139 , @xmath170 . the lines denote the same boundaries as in figs . [ nishinetiltdirection ] and [ egydirection1 ] . ] undoped samples with @xmath198 and hole - doped samples allow us to study the contribution of the quadratic valley . in this subsection we investigate how this third valley damps the collective modes of the linear valleys , and how the collective mode of the massive valley appears on alongside the excitations of the massless valleys . the particle - hole continuum of the massive valley is bounded by@xcite @xmath199 in the same manner as for the linear bands , a 2d quadratic band has a plasmon mode that disperses @xmath200 , and in a magnetic field , this mode is becomes the gapped uhm.@xcite notice that the separation of the lls scales as @xmath201 and @xmath202 for the massless and the massive carriers , respectively . one would therefore expect naively that the lls of the quadratic valley are generically less dense in energy than those of the dirac cones . however , due to the values for the fermi velocities of the dirac carriers and for the band mass of the massive carriers , the ll separation of the massive carriers is much lower than that in the tilted dirac cones in the magnetic - field range discussed here . furthermore , the relative position of the collective modes of the massive and the massless valleys and the particle - hole continua of the three valleys are sensitive to both the magnetic field @xmath46 and the doping level . hence we discuss two representative cases , one at charge neutrality and relatively small magnetic fields @xmath198 ( fig . [ massdirection1 ] ) , and one at heavy hole - doping ( fig . [ massdirection2 ] ) . the two situations correspond to the sketches in figs . [ landaulevels](b ) and ( c ) , respectively . the first case is studied in fig . [ massdirection1 ] , which depicts the density - density response at @xmath131 and @xmath132 , which is adequate at charge neutrality when @xmath130 ; c.f . [ landaulevels](b ) . now the particle - hole continuum of the quadratic valley lies well below the linear magnetoplasmon mode of the massless dirac carriers in both cones in the direction @xmath143 , for which the difference between the velocities in the two cones is the most pronounced . this is demonstrated in fig . [ massdirection1](a , b ) , where one of the cones are disregarded for visibility reasons . in addition to the characteristic frequency @xmath166 , we have also sketched the lines @xmath203 and @xmath204 , which delimit the particle - hole continuum of the massive carriers . thus the particle - hole excitations of the massive carriers do not overlap with those of the dirac valleys , and its spectral weight is comparable to the dominant linear magnetoplasmon modes of the latter . the uhm of the massive holes is not damped either , as it lies at very small energies , well below the cyclotron frequency of the massless carriers . ( the former is characterized by the energy scale @xmath205 ev , while the latter is given by @xmath206 ev at @xmath207 t. ) the total , physical density - density response in fig . [ massdirection1](c ) indicates that the response of @xmath0(bedt - ttf)@xmath1i@xmath2 at low frequencies is determined by the massive carriers . in the higher frequency range the dirac carriers dominate , with intervalley - damped collective modes , as discussed in [ bothcones ] above , but no damping due to the massive carriers . the reason is the separation of energy scales at low momenta , and not any difference in the density of states . in momentum space . the topmost filled band is @xmath208 in the massless valleys and @xmath209 in the massive valley . other parameters are @xmath210 t and @xmath170 . panel ( a ) shows the @xmath211 of an imaginary system that omits cone @xmath19 , ( b ) shows @xmath212 omits cone @xmath17 , while ( c ) shows the complete @xmath213 . the units are s / m@xmath214 ; normalization by the density of states is not applied . the straight line @xmath215 is the boundary between the interband and intraband excitations of massless carriers , and the curved ones [ @xmath216 in eq . ( [ demarcate ] ) ] demarcate the particle - hole continuum of massive carriers for @xmath172 . ] in momentum space . the topmost filled band is @xmath208 in the massless valleys and @xmath209 in the massive valley . other parameters are @xmath210 t and @xmath170 . panel ( a ) shows the @xmath211 of an imaginary system that omits cone @xmath19 , ( b ) shows @xmath212 omits cone @xmath17 , while ( c ) shows the complete @xmath213 . the units are s / m@xmath214 ; normalization by the density of states is not applied . the straight line @xmath215 is the boundary between the interband and intraband excitations of massless carriers , and the curved ones [ @xmath216 in eq . ( [ demarcate ] ) ] demarcate the particle - hole continuum of massive carriers for @xmath172 . ] in momentum space . the topmost filled band is @xmath208 in the massless valleys and @xmath209 in the massive valley . other parameters are @xmath210 t and @xmath170 . panel ( a ) shows the @xmath211 of an imaginary system that omits cone @xmath19 , ( b ) shows @xmath212 omits cone @xmath17 , while ( c ) shows the complete @xmath213 . the units are s / m@xmath214 ; normalization by the density of states is not applied . the straight line @xmath215 is the boundary between the interband and intraband excitations of massless carriers , and the curved ones [ @xmath216 in eq . ( [ demarcate ] ) ] demarcate the particle - hole continuum of massive carriers for @xmath172 . ] figure [ massdirection2 ] shows the second case , a strongly hole - doped situation with @xmath217 and @xmath218 . now the massive carriers are definitely involved in the damping . in direction @xmath143 , the forbidded region 1b of cone @xmath17 overlaps with the particle - hole continuum of the massive valley [ fig . [ massdirection2](a ) ] . the uhm , which stems from the dirac carriers , is still obtained at larger energies than that associated with the massive carriers . ( at @xmath219 t the cyclotron frequency is characterised by the energy scale @xmath220 ev for the massive carriers , while by @xmath221 ev for the massless ones . ) the latter is clearly discernible in fig . [ massdirection2](a ) , and it is damped once it enters the overlap region between the two lines @xmath222 and @xmath223 , which is not a forbidden region for particle - hole excitations of the massive carriers . furthermore , it is clear from fig . [ massdirection2](a ) that the uhm avoids entering this region and rather approaches asymptotically the border @xmath222 , as one expects for the uhm in a single - band model of massive carriers.@xcite however , the situation is slightly more involved in the present case , where one observes a coupling between the uhm with the linear magnetoplasmons of the massless dirac carriers , which gives rise to a modulation of the spectral weight along the uhm . this coupling is reminiscent to the so - called bernstein modes discussed in the framework of graphene in a strong magnetic field.@xcite for cone @xmath19 ( or cone @xmath17 in the @xmath144 direction ) , on the other hand , the forbidden region of the massive valley overlaps with the 1a intraband region of the massless valley . the uhm is therefore strongly damped and almost invisible at small momenta but reappears at @xmath224 , c.f . [ massdirection2](b ) . this reappearance of the uhm , even if it occurs in a region of possible intraband particle - hole excitations of the massless dirac carriers , may be understood from the particular chiral properties of the latter . indeed , the spectral weight of particle - hole excitations , in the case of dirac fermions , is concentrated around the lines @xmath142 due to the suppression of backscattering,@xcite such that the uhm is barely damped once it is further away from @xmath142 . in the regions where the allowed regions ( particle - hole continua ) overlap , the one with the greater spectral weight dominates . this is visible in fig . [ massdirection2](a ) , where the particle - hole continuum of the massive valley is almost as bright as the more concentrated linear magnetoplasmon mode of the dirac fermions . this also applies to the total , physical density - density response , fig . [ massdirection2](c ) , where the spectral weight is mostly concentrated on the excitations of the massive valley , although the linear magnetoplasmon of cone @xmath17 become visible at large momenta . , but in a hole - doped sample . the topmost filled massless landau level is @xmath225 , while among the massive lls it is @xmath226 . other parameters are @xmath169 t and @xmath170 . for notations [ massdirection1 ] . notice that one of the curves demarcating the particle - hole continuum of the massive valley for @xmath172 , @xmath227 in eq . ( [ demarcate ] ) , is not visible because @xmath228 is outside of the presented momentum range . ] , but in a hole - doped sample . the topmost filled massless landau level is @xmath225 , while among the massive lls it is @xmath226 . other parameters are @xmath169 t and @xmath170 . for notations [ massdirection1 ] . notice that one of the curves demarcating the particle - hole continuum of the massive valley for @xmath172 , @xmath227 in eq . ( [ demarcate ] ) , is not visible because @xmath228 is outside of the presented momentum range . ] , but in a hole - doped sample . the topmost filled massless landau level is @xmath225 , while among the massive lls it is @xmath226 . other parameters are @xmath169 t and @xmath170 . for notations [ massdirection1 ] . notice that one of the curves demarcating the particle - hole continuum of the massive valley for @xmath172 , @xmath227 in eq . ( [ demarcate ] ) , is not visible because @xmath228 is outside of the presented momentum range . ] the interplay between massless carriers in tilted anisotropic dirac cones and massive holes in a roughly isotropic pocket gives rise to a remarkable doping - dependent angular dependence of the screening properties . [ static ] shows the real part of the static bare polarizability @xmath229 , and the dielectric function @xmath230 in the rpa , respectively . for electron - doping or charge neutrality at @xmath231 t , the bare polarizability is entirely due to the linear bands [ panel ( a ) ] . this anisotropy is naturally inherited by the rpa dielectric function [ panel ( b ) ] , which means that screening due to the discussed bands is anisotropic . if the magnetic field is reduced , @xmath198 , however , massive carriers can also be polarized at zero temperature . because of their higher density of states at small energies , their contribution is dominant in the total polarizability , which shows a very weak angular dependence [ panel(c ) ] . as a result , screening is almost isotropic ; c.f . fig . [ static](d ) . the suppression of the screening anisotropy of the tilted cones by the massive valley naturally becomes stronger in hole - doped samples . ( color online ) the static screening properties of @xmath0(bedt - ttf)@xmath1i@xmath2 at charge neutrality . upper row : @xmath219 t , the massive valley is completely filled and inert . lower row : @xmath207 t , the topmost landau level of the massive valley is empty . left panels : @xmath229 in s / m@xmath214 , right panels : @xmath230 . we show several directions in the momentum plane , specified by the angle @xmath232 relative to @xmath18 , the tilting direction of cone @xmath17 . the grey line in panel ( c ) shows the polarizability of the massive valley , which is dominant and suppresses any anisotropy due to the tilted dirac cones if @xmath198 . ] in this work we have studied the low - energy magnetic excitations of the quasi-2d electron gas in a three - valley system of two tilted and anisotropic massless dirac cones and a massive hole pocket , with an emphasis on direction - dependent effects and the properties for which massive carriers are relevant . this model is a realistic representation of a layer of the organic conductor @xmath0(bedt - ttf)@xmath1i@xmath2 at high pressure or uniaxial strain , but there is some ambiguity regarding the band parameters and even the presence of massive carriers . we have found that the tilt of the cones causes a direction - dependent intervalley damping of the upper hybrid modes of the dirac valleys , while the linear magnetoplasmons are less affected . the magnetoplasmons of the massive band may coexist with those of the massless ones , depending on doping and the strength of the magnetic field . the latter also tunes the system between isotropic and anisotropic screening regimes . the experimental study of the collective modes in ribbon samples@xcite or by local probes,@xcite successfully applied for graphene , may be challenging under high pressure , but grating couplers might be usable . on the other hand , the layered structure of these organic conductors recommends itself to inelastic light scattering experiments,@xcite which have successfully clarified the intrasubband plasmon modes of multilayer gaas-(alga)as heterostructures . the elaboration of collective modes in the presence of interlayer coupling is delegated to future work . this research was funded by the hungarian scientific research funds no . k105149 . c. t. was supported by the hungarian academy of sciences . supercomputer facilities were provided by national information infrastructure development institute , hungary . j. s. was supported by campus france and acknowledges hospitality from laboratoire de physique des solides . here follow the morinari _ et al . _ s ref . , merely fixing some apparent bugs in the formulas . we use the landau gauge @xmath233 , and work in the coordinate system introduced in eq . ( [ newcoord ] ) . the transformed real - space coordinates @xmath234 are @xmath235 for @xmath236 in cone @xmath17 the landau orbitals are @xmath237,\end{gathered}\ ] ] while for the @xmath238 landau level , @xmath239 for @xmath236 in cone @xmath19 , the orbitals are @xmath240,\end{gathered}\ ] ] while for the @xmath125 , @xmath241 we have used the subformulas ( @xmath242 ) @xmath243 and @xmath244 with the argument @xmath245 notice that @xmath246 depends on both the ll index ( including its sign ) and the cone @xmath247 . also , it is identical for the @xmath248 and @xmath249 parts of @xmath250 in eqs . ( [ nagypszi ] ) and ( [ nagypszil ] ) . here we calculate the bare polarizability of the linear valleys by standard methods , using the orbitals that eqs . ( [ nagypszi ] ) to ( [ nagypszivege ] ) specify in the rotated , rescaled coordinate system , c.f . ( [ newcoord ] ) and ( [ newcoord2 ] ) . the field operators are ( @xmath242 ) : @xmath251 suppressing spin to avoid clutter , the ( gauge - dependent ) bare green s function has a matrix structure , @xmath252 where @xmath253 and @xmath254 . by fourier - transformation , @xmath255 now we can evaluate the bare polarizability in eq . ( [ chidef ] ) , involving the matrix structure of the green s functions in the trace . we obtain eq . ( [ chi0lin ] ) , with the form factors @xmath256 defined as @xmath257 for @xmath258 . similarly for @xmath259 , @xmath260 and finally , for @xmath261 , @xmath262 here we have introduced the functions @xmath263 , of @xmath264 . for @xmath265 they are defined as @xmath266 similarly , for @xmath267 the definition is @xmath268 in the above definitions @xmath269 it is easy to check that @xmath270^\ast$ ] . k. kajita , t. ojiro , h. fujii , y. nishio , h. kobayashi , a. kobayashi , and r. kato , j. phys . . jpn . * 61 * , 23 ( 1992 ) ; n. tajima , m. tamura , y. nishio , k. kajita , and y. iye , j. phys . . jpn . * 69 * , 543 ( 2000 ) ; n. tajima , s. sugawara , r. kato , y. nishio , and k. kajita , phys . lett . * 102 * , 176403 ( 2009 ) . p. r. wallace , phys . rev . * 71 * , 622 ( 1947 ) ; j. w. mcclure , phys . rev . * 108 * , 612 ( 1957 ) ; j. c. slonczewski and p. r. weiss , phys . rev . * 109 * , 272 ( 1958 ) ; k. s. novoselov , a. k. geim , s. v. morozov , d. jiang , m. i. katsnelson , i. v. grigorieva , s. v. dubonos , and a. a. firsov , nature * 438 * , 197 ( 2005 ) ; y. zhang , y .- w . tan , h. l. stormer , and p. kim , nature * 438 * , 201 ( 2005 ) . notice that there are four bedt - ttf molecules and two i@xmath4 ions in a unit cell . the latter capture two out of the eight electrons in the highest occupied molecular orbitals of the bedt - ttf s , which comprise the conducting layer . notice the tension arises because the quadratic valley has a maximum near the fermi energy , thus it has little effect if the sample is electron - doped , but as its set of landau levels is denser than that of the linear valleys , several of its lls are emptied for hole - doping before the first lls of the linear valleys are affected . kenneth w .- k . shung , phys . rev . b * 34 * , 979 ( 1986 ) ; t. ando , j. phys . soc . jpn . * 75 * , 074716 ( 2006 ) ; b. wunsch , t. stauber , f. sols , and f. guinea , new j. phys . * 8 * , 318 ( 2006 ) ; e. h. hwang and s. das sarma , phys . b * 75 * , 205418 ( 2007 ) . h. seo , j. phys . . jpn . * 69 * , 805 ( 2000 ) ; y. takano , k. hiraki , h. m. yamamoto , t. nakamura , and t. takahashi , j. phys . chem . solids , * 62 * , 393 ( 2001 ) ; r. wojciechowski , k. yamamoto , k. yakushi , m. inokuchi , and a. kawamoto , phys . rev . b * 67 * , 224105 ( 2003 ) ; y. takano , k. hiraki , y. takada , h. m. yamamoto , and t. takahashi , j. phys . soc . jpn . * 79 * , 104704 ( 2010 ) ; m. hirata , k. ishikawa , k. miyagawa , k. kanoda , and m. tamura , phys . b * 84 * , 125133 ( 2011 ) . g. f. giuliani and g. vignale , _ quantum theory of the electron liquid _ ( cambridge university press , cambridge , 2005 ) ; j. slyom , _ fundamentals of the physics of solids , volume iii_. ( springer - verlag berlin heidelberg , 2010 ) . z. fei , a. s. rodin , g. o. andreev , w. bao , a. s. mcleod , m. wagner , l. m. zhang , z. zhao , m. thiemens , g. dominguez , m. m. fogler , a. h. castro neto , c. n. lau , f. keilmann , and d. n. basov , nature * 487 * , 82 ( 2012 ) . d. olego , a. pinczuk , a. c. gossard , and w. wiegmann , phys . b * 25 * , 7867(r ) ( 1982 ) ; j. k. jain and p. b. allen , phys . lett . * 54 * , 2437 ( 1985 ) ; r. sooryakumar , a. pinczuk , a. gossard , and w. wiegmann , phys . rev . b * 31 * , 2578(r ) ( 1985 ) ; g. fasol , n. mestres , h. p. hughes , a. fischer , and k. ploog , phys . rev . lett . * 56 * , 2517 ( 1986 ) .

in this paper we prove schur s conjecture in , which states that any diameter graph in on vertices may have at most cliques of size . we obtain an analogous statement for diameter graphs on a sphere of radius . the proof rests on the following statement , conjectured by f. mori and j. pach : given two unit regular simplices on vertices in , either they share vertices , or there are vertices such that . the same holds for unit simplices on a-dimensional sphere of radius greater than . * msc : * 52c10 . + 0.2 cm * keywords : * schur s conjecture , diameter graph , unit simplices in .

1402.3694

one of the classical problems in discrete geometry , raised by p. erds in 1946 @xcite , is the following : given @xmath2 points in the plane , how many unit distances they may determine ? the key definition related to the question of p. erds is that of a _ unit distance graph_. a graph @xmath1 is a _ unit distance graph _ in @xmath0 if its set of vertices is a finite subset of @xmath0 and the edges are formed by the pairs of vertices which are at unit distance apart . in terms of distance graphs the question is to determine the maximal number of edges in a planar unit distance graph on @xmath2 vertices . in this paper we focus on the questions of this type for _ diameter graphs_. a graph @xmath11 is a _ diameter graph _ in @xmath0 , if @xmath12 is a finite set of diameter 1 , and edges of @xmath1 are formed by the pairs of vertices that are at unit distance apart . diameter graphs arise naturally in the context of the finite version of the famous borsuk s problem ( see , e.g. , @xcite for the survey on borsuk s problem ) , which is stated as follows : is it true that any ( finite ) set of unit diameter in @xmath0 can be partitioned into @xmath13 subsets of strictly smaller diameter ? the finite version is equivalent to the following question concerning diameter graphs : is it true that any diameter graph @xmath1 in @xmath0 satisfies @xmath14 ? a question about diameter graphs analogous to the question from the first paragraph has a simple answer : any set of @xmath2 points in the plane generates at most @xmath2 diameters , or any diameter graph on @xmath2 vertices in the plane has at most @xmath2 edges . this was proved by h. hopf and e. pannwitz in @xcite . interestingly , this result leads to a simple proof of the fact that borsuk s question for finite sets in the plane have a positive answer . indeed , it is easy to derive combinatorially that any graph @xmath1 on @xmath2 vertices with at most @xmath2 edges and such that any of its subgraphs has at least as many vertices as edges satisfies @xmath15 . a. vzsonyi conjectured that any diameter graph in @xmath16 on @xmath2 vertices has at most @xmath17 edges . again , it is easy to see that borsuk s conjecture for finite sets in @xmath16 follows from this statement . vzsonyi s conjecture was proved independently by b. grnbaum @xcite , a. heppes @xcite and s. straszewicz @xcite . an interesting generalization of this result to the case of @xmath18-th diameters was obtained by f. mori and j. pach @xcite . while the maximum number of edges in a diameter graph in @xmath19 is linear in the number of vertices , it becomes quadratic already in @xmath20 . to put the discussion in a more general context , we introduce the following notations . denote by @xmath21 ( @xmath22 ) the maximum number of cliques of size @xmath23 in a diameter ( unit distance ) graph on @xmath2 vertices in @xmath0 . erds @xcite studied @xmath24 and @xmath25 for different @xmath3 . he showed that for @xmath26 we have @xmath27 swanepoel @xcite determined @xmath24 for fixed even @xmath28 and sufficiently large @xmath2 depending on @xmath3 and determined @xmath25 for @xmath26 and sufficiently large @xmath2 . functions @xmath29 @xmath22 for @xmath30 and similar functions were studied in several papers . in particular , the following conjecture was raised in @xcite : [ gypsp ] we have @xmath31 for @xmath32 . this was proved by h. hopf and e. pannwitz for @xmath33 in @xcite and for @xmath34 by z. schur et . al . in @xcite . in the latter paper the authors also proved that @xmath35 . in @xcite p. mori and j. pach progressed towards resolving this conjecture . namely , they showed that schur s conjecture holds in the following special case : [ thph ] given a diameter graph @xmath1 on @xmath2 vertices in @xmath0 , the number of @xmath3-cliques in @xmath1 does not exceed @xmath2 , provided that any two @xmath3-cliques share at least @xmath7 vertices . as it turns out , schur s conjecture and related questions are tightly connected with analogous questions for spherical sets . the spherical analogues were studied in some papers . in particular , in the paper @xcite v. bulankina et al . noted that the statement of theorem [ thph ] holds for spheres of large radii : given a diameter graph @xmath1 on @xmath2 vertices in a @xmath3-dimensional sphere @xmath4 with radius @xmath5 , the number of @xmath3-cliques in @xmath1 does not exceed @xmath2 , provided that any two @xmath3-cliques share at least @xmath7 vertices ( theorem 4 from @xcite ) . moreover , they showed that schur s conjecture holds for @xmath36 for @xmath5 . to be precise , we formulate schur s conjecture for spheres separately : [ gypsph ] any diameter graph @xmath1 on @xmath2 vertices ( and with edges of unit euclidean length ) on a sphere @xmath37 with @xmath5 has at most @xmath2 @xmath3-cliques . in the paper @xcite a. kupavskii studied properties of diameter graphs in @xmath20 . the following theorem completes the description of the quantity @xmath38 for different @xmath23 : [ th10 ] 1 . for @xmath39 we have @xmath40 ( in corollary 3 from @xcite the same was proved for sufficiently large @xmath2 . ) 2 . for all sufficiently large @xmath2 we have @xmath41 3 . ( schur s conjecture in @xmath20 ) for all @xmath42 we have @xmath43 . in @xcite the first author also studied diameter graphs on @xmath44 with @xmath5 . in particular , he showed that an analogue of vzsonyi s conjecture holds for spheres . in the next section we present our main results and discuss related questions . in section [ sec3 ] we introduce the basic objects that are used in the proof . in section [ sec4 ] we present the proofs of the results . the main result of this paper is the proof of schur s conjecture both in the euclidean space and on the sphere in the general case : [ thmain ] schur s conjecture holds + 1 . in the space @xmath0 , + 2 . on the sphere @xmath4 of radius @xmath45 the proof of the first part actually relies heavily on the second part , so the questions for the euclidean space and for the sphere are indeed interconnected . + * remark * note that throughout the article by a @xmath18-simplex in @xmath0 we mean a set of @xmath46 vertices in @xmath0 in general position . + next we discuss several questions mentioned in the paper @xcite . since the authors of @xcite proved theorem [ thph ] , they naturally raised the following problem : [ ph1 ] any two unit regular simplices on @xmath3 vertices in @xmath0 must share at least @xmath7 vertices , provided the diameter of their union is 1 . we follow the approach suggested by f. mori and j. pach and confirm this conjecture ( and its spherical version ) in our paper , which together with theorem [ thph ] and its spherical analogue from @xcite , mentioned in the previous section , gives us the proof of schur s conjecture both in the space and on the sphere . another problem the authors of @xcite raised deals with irregular simplices . [ ph2 ] let @xmath47 and @xmath48 be two simplices on @xmath3 vertices in @xmath0 with @xmath49 , such that all their edges have length at least 1 . then there exist @xmath50 such that @xmath51 . by slightly modifying the proof of theorem [ thmain ] it is not difficult to obtain the following theorem : [ thirr ] consider a regular unit simplex @xmath52 and a simplex @xmath53 in @xmath0 ( or on @xmath4 with @xmath5 ) , where the second simplex satisfies the property @xmath54 for @xmath55 . then either these two simplices share @xmath7 vertices , or @xmath56 for some @xmath57 . this theorem solves conjecture [ ph2 ] in a stronger form in the case where one of the two simplices is regular . we omit the proof , but the main additional ingredient needed is that the radius of the smallest ball that contains the simplex @xmath58 is at least as big as for a regular unit @xmath59-simplex , provided that @xmath54 for @xmath55 . this , in turn , is an easy application of kirszbraun s theorem ( see @xcite for a short and nice proof ) : let @xmath60 be a subset of @xmath61 , where @xmath61 is @xmath0 , @xmath62 or @xmath63 ( a @xmath3-dimensional hyperbolic space ) . then any nonexpansive map @xmath64 can be extended to a nonexpansive map @xmath65 . a nonexpansive map @xmath66 is a map which satisfies @xmath67 for any @xmath68 . after having prepared the first version of this paper , we came across a paper by h. maehara @xcite , in which the author studies a seemingly unrelated concept of sphericity of a graph : given a graph @xmath1 , the sphericity of @xmath1 is the minimum dimension in which the vertices of the graph can be represented as unit spheres in such a way that two spheres intersect ( or touch ) iff the corresponding vertices are connected by an edge . in the paper @xcite the author discusses the sphericity of complete bipartite graphs . and , as it turned out , the main result of the paper is , in fact , the proof of conjecture [ ph2 ] , 20 years before the conjecture was formulated ! for a bit more on maehara s result in the context of schur s conjecture see paragraph [ secp1 ] . finally , in the paper @xcite the authors raised the following general problem : [ ph3 ] for a given @xmath3 , characterize all pairs @xmath69 of integers such that for any set of @xmath18 red and @xmath23 blue points in @xmath0 we can choose a red point @xmath70 and a blue point @xmath71 such that @xmath72 is at least as large as the smallest distance between two points of the same color . for @xmath73 and @xmath74 it is not difficult to construct an example of two regular unit simplices in @xmath0 on @xmath18 and @xmath23 vertices respectively , such that the distance between any two vertices from different simplices is smaller than 1 , which we describe at the end of the next section . ( an analogous , but somewhat different , example appeared in the latter version of the paper @xcite . ) we think that in this case one can not take a larger @xmath23 , thus , we conjecture the following . given two unit simplices in @xmath0 , one on @xmath13 vertices , the other on @xmath75 vertices , either they share a vertex , or the diameter of their union is strictly larger than 1 . given a hyperplane @xmath76 , we denote by @xmath77 and @xmath78 two closed half - spaces ( half - spheres in the spherical case ) that are determined by @xmath76 . the following object is very important for understanding diameter graphs : [ opr1 ] a reuleaux simplex @xmath79 in @xmath0 _ is a set formed by the intersection of the balls @xmath80 of unit radius with centers in @xmath81 , where @xmath82 s are the vertices of a unit simplex in @xmath0 . in the case @xmath34 we call this object a reuleaux tetrahedron , and in the case @xmath33 we call it a reuleaux triangle . _ we denote the @xmath59-dimensional spheres of unit radii with centers in @xmath83 ( the boundary spheres of @xmath84 ) by @xmath85 . a reuleaux simplex is a spherical polytope , so one can naturally partition the boundary of a reuleaux simplex into spherical faces of different dimensions : the vertices of the underlying simplex are the zero - dimensional faces , the arcs that connect the vertices are the one - dimensional faces and so on . we discuss it in more details a bit later in this section . the analogous definition could be given in the case of @xmath4 , @xmath5 . in this case we call the body a _ spherical reuleaux simplex_. the only thing one have to keep in mind is that on a @xmath3-dimensional sphere we still consider spherical reuleaux simplices on @xmath13 vertices . note that , by jung s theorem , on a @xmath3-dimensional sphere of radius @xmath86 one can have a regular unit @xmath87-simplex , which is , however , impossible for other radii ( and , in particular , impossible for @xmath88 ) . for a given set @xmath89 we denote its interior by @xmath90 . in the paper we use several times the following simple observation . [ obs0 ] consider two @xmath3-balls @xmath91 in @xmath0 of radii @xmath92 , correspondingly . denote their boundary spheres by @xmath93 . assume that @xmath94 and that @xmath95 and @xmath96 intersect in a @xmath97-dimensional sphere . denote the hyperplane that contains @xmath98 by @xmath99 and assume that the centers of @xmath93 are in the same closed halfspace @xmath100 with respect to @xmath99 . the other halfspace we denote by @xmath101 . then @xmath102 and @xmath103 . moreover , @xmath104 and @xmath105 figure 1 see fig . 1 , illustrating the observation . we apply the observation above to deduce the following two lemmas . [ lem0 ] consider a reuleaux simplex @xmath106 with the set of vertices @xmath81 and a ball @xmath107 with a boundary sphere @xmath95 , circumscribed around the @xmath3-simplex @xmath108 . then @xmath79 lies inside @xmath107 , moreover , @xmath109 denote by @xmath110 the hyperplane that passes through the vertices @xmath111 . we denote the closed halfspace defined by @xmath110 and that contains @xmath112 by @xmath113 . this halfspace contains the convex hull of the vertices @xmath108 . by @xmath114 we denote the other halfspace . applying the part of observation [ obs0 ] about interiors to the balls @xmath115 , @xmath107 , we get that @xmath116 . after going through all possible values of @xmath117 , we get that @xmath118 on the other hand , @xmath119 is just a convex hull of the points @xmath108 , and it is for sure contained in @xmath107 , moreover , it intersects @xmath95 only in its vertices @xmath83 . [ lem01 ] consider a reuleaux simplex @xmath106 with the set of vertices @xmath81 . then the intersection of @xmath79 with the hyperplane @xmath76 that passes through @xmath120 is a reuleaux simplex @xmath121 with vertices @xmath122 . we have @xmath123 and we have to prove that @xmath124 . thus , it is enough to show that @xmath125 . denote the circumscribed ball of the reuleaux simplex @xmath126 by @xmath127 and its boundary sphere by @xmath96 . then @xmath128 , and the statement of the lemma follows from the first claim of lemma [ lem0 ] . we as well need some knowledge about the structure of a reuleaux simplex @xmath79 as a spherical polytope . let the vertices of @xmath79 be @xmath83 and their convex hull be denoted by @xmath129 . the boundary of the reuleaux simplex @xmath79 can be partitioned into relatively open spherical regions of different dimensions in the following way . consider all spheres @xmath60 that are formed as intersections of several spheres out of @xmath130 . for example , the intersection of the first @xmath3 spheres is a two - point set , with one of its points being @xmath131 . we only exclude the intersection of all the spheres , which is empty . we denote the set of all such spheres @xmath60 by @xmath132 . for each point on the boundary of @xmath79 we may find the sphere from @xmath132 of minimal dimension that contains it . the set of all points from the boundary of @xmath79 that correspond to a given @xmath133 we call a face . it is defined by the set of strict quadratic inequalities and , thus , is naturally a relatively open set . the center of each sphere @xmath133 coincides with the center of some of the faces of @xmath129 . namely , if @xmath134 , where @xmath135 , then the center of @xmath60 is the centerpoint of the @xmath136-dimensional face with the vertices @xmath137 . on the other hand , @xmath60 contains all the vertices from @xmath138 and , since @xmath60 is a @xmath139-dimensional sphere and points @xmath140 are in general position , these points determine @xmath60 . we call these points the vertices of the face . each face of @xmath79 in the sense above is a connected set , moreover , it is obtained from the face of @xmath129 with the same set of vertices via projection from the center @xmath141 of @xmath129 on the boundary of @xmath79 . we verify this property in what follows . fix a vertex set of the face @xmath142 of @xmath79 . w.l.o.g . , it is @xmath143 . the points of @xmath142 on the boundary of @xmath79 are the ones that , first , are at distance 1 from @xmath144 and , second , are at distance strictly less than 1 from @xmath145 . similarly to what we have said in the previous paragraph , it follows from the first condition that all these points lie in the flat @xmath146 , which is an affine hull of the points @xmath147 ( we may use the centerpoint of the face @xmath148 of @xmath129 instead of @xmath141 ) . since @xmath149 has the same length for all @xmath150 , the second condition is equivalent to the fact that each point @xmath151 from @xmath142 satisfies the angular inequality @xmath152 for all @xmath153 . each of these inequalities is satisfied in a halfspace , defined by the hyperplane that passes through @xmath141 and vertices @xmath154 . these halfspaces bound the face @xmath143 in @xmath155 and , as we have showed above , they also bound the face with the same set of vertices in @xmath156 . moreover , they pass through @xmath141 , which concludes the proof of the statement about the central projection . we formulate the findings of the last three paragraphs in a lemma : [ lem02 ] in the notations introduced above , consider a reuleaux simplex @xmath106 . then its boundary may be split into relatively open connected spherical regions ( faces ) , each of which corresponds to an intersection of several spheres out of @xmath130 . the face @xmath142 of @xmath79 that corresponds to the intersection of the spheres @xmath157 lies on the sphere with the center in the centerpoint @xmath158 of the ( relatively open ) face @xmath159 of @xmath129 with vertices @xmath144 : @xmath160 . the minimal flat that contains @xmath142 , is an affine hull of points @xmath161 . moreover , the face @xmath142 is equal to the central projection of @xmath159 from the center of @xmath79 to the boundary of @xmath79 . next we define the object which is of a particular importance for the paper : [ opr1 ] a rugby ball @xmath162 in @xmath0 _ is a set formed by the intersection of the balls @xmath80 of unit radius with centers in @xmath163 , where @xmath82 s are the vertices of a unit @xmath59-simplex in @xmath0 . _ we omit the analogous definition of a _ spherical rugby ball_. note the difference between a reuleaux simplex and a rugby ball . in the latter case , we consider the intersection of @xmath3 balls instead of @xmath13 in the former . the intersection of the hyperplane @xmath76 that passes through @xmath120 , and the corresponding rugby ball is a reuleaux simplex of codimension 1 . the rugby ball is symmetric with respect to @xmath76 . moreover , consider a reuleaux simplex @xmath79 on the vertices @xmath83 , the rugby ball @xmath162 on the vertices @xmath164 , and the hyperplane @xmath76 containing vertices @xmath120 . suppose that @xmath165 in the following lemma we prove that @xmath166 . we denote this body by @xmath167 . [ lem03 ] in the notations introduced above , we have @xmath166 . since @xmath168 it is obviously sufficient to show that @xmath169 consider a ball @xmath107 circumscribed around @xmath79 . by using the same argument as in lemma [ lem0 ] , we get that @xmath170 . on the other hand , applying observation [ obs0 ] to @xmath171 , we get that @xmath172 . now we describe the construction mentioned in the end of the previous section . take a regular simplex on @xmath13 vertices in @xmath0 as the red points . next , construct the reuleaux simplex on the the red points and choose @xmath173 midpoints @xmath174 of some @xmath23 pairwise disjoint arcs that connect the vertices of the reuleaux simplex . it could be checked that the distance between the midpoints of two such arcs is strictly bigger than 1 . to see this , one have to consider a coordinate representation of the simplex @xmath108 in the hyperplane @xmath175 in @xmath176 and calculate the coordinates of a middle of an arc that connects two vertices of the corresponding reuleaux simplex . thus , if we consider the simplex on @xmath174 and contract it a little , we will get a simplex on vertices @xmath177 with all vertices inside the reuleaux simplex and with all sides greater than 1 . we take @xmath178 as the blue points , which together with the red points gives us the desired example . the proof of theorem [ thmain ] is based on induction and the following auxiliary theorem , which is of interest by itself : [ thcom ] given a diameter graph @xmath1 + 1 . in the space @xmath0 , @xmath49 ; + 2 . on the sphere @xmath4 of radius @xmath179 @xmath49 , + any two @xmath3-cliques in @xmath1 must share a vertex . as we came across the paper @xcite we were thinking whether or not to try to give a proof of theorem [ thcom ] using maehara s result . the euclidean case of theorem [ thcom ] is almost equivalent to theorem 2 from @xcite ( which is equivalent to conjecture [ ph2 ] ) . however , to get part 1 of theorem [ thcom ] from conjecture [ ph2 ] , one has to replace the non - strict inequality on distances by a strict one . it turns out that this seemingly technical detail is not easy to overcome . if applied directly , conjecture [ ph2 ] says only that , if there are two regular unit @xmath59-simplices in a diameter graph in @xmath0 that do not share a vertex , then there must be at least one edge between them . this is clearly not sufficient ( and it is fairly easy to reduce theorem [ thcom ] to this case ) . the proof of maehara , however , may be modified to give a proof of theorem [ thcom ] in the euclidean case , but becomes significantly more complicated . as for the spherical case , we do not even know . it may also be the case that it is possible to apply maehara s technique , but we believe that in sum it would not simplify the proof of theorem [ thcom ] . therefore , we decided to leave the proof as it is and not to utilize maehara s ideas . in this subsection we describe how to derive theorem [ thmain ] from theorem [ thcom ] . consider two @xmath3-cliques in a diameter graph @xmath1 in @xmath0 ( or on @xmath37 with @xmath5 ) . then , by theorem [ thcom ] , these two cliques must share a vertex . all the remaining vertices of the two simplices must lie on the @xmath59-dimensional unit sphere @xmath95 with the center in the common vertex of the two simplices . the vertices on @xmath95 form two @xmath97-dimensional unit simplices , and , since the graph on the sphere is still a diameter graph , we can again apply theorem [ thcom ] and obtain that they must share another vertex . finally , when we obtain that any two @xmath3-cliques must share @xmath7 common vertices we apply a spherical analogue of theorem [ thph ] from @xcite , and schur s conjecture is proved . we only have to verify the following : the spheres that we obtain in such a process always have radius greater than @xmath10 . this was shown to be true in the paper @xcite ( lemma 4 ) . we state this fairly easy lemma and present its proof for completeness . [ lemrad ] consider a @xmath3-dimensional sphere @xmath180 of radius @xmath181 and a unit simplex @xmath79 on @xmath18 vertices @xmath182 with all its vertices on @xmath95 . then the intersection @xmath183 of the sphere @xmath95 and the @xmath18 unit spheres with centers in @xmath182 is a sphere of radius @xmath184 . we assume that the sphere is embedded into the euclidean space , and we work in that space . denote by @xmath185 the center of the sphere @xmath96 , circumscribed around @xmath79 . by jung s theorem , the radius @xmath186 of @xmath96 is equal to @xmath187 . so , the radius @xmath188 of the sphere @xmath189 , which is the intersection of @xmath18 unit spheres with centers in @xmath182 is @xmath190 note that the center of @xmath189 is also @xmath191 . denote the center of @xmath95 by @xmath141 . then the center @xmath151 of @xmath183 lies on the segment @xmath192 of length @xmath71 . since @xmath182 lie on @xmath95 , we have @xmath193 . suppose @xmath151 splits the segment @xmath192 into the parts of length @xmath194 respectively . then , since @xmath195 , we get @xmath196 . we also have @xmath197 so we get @xmath198 therefore , @xmath199 because it is easy to see that @xmath200 . our main goal is to prove theorem [ thcom ] . the proof of this theorem also goes by induction . the statement of the theorem is known to be true for @xmath16 and @xmath44 for @xmath5 ( see the papers @xcite , @xcite ) . in what follows we reduce the problem for the @xmath3-dimensional space or for the @xmath3-dimensional sphere to the analogous problem for the @xmath59-dimensional sphere . since the base of the induction is already verified , this concludes the proof of the theorem . it is tempting to give a unified proof of theorem [ thcom ] , in which the euclidean and the spherical cases are both treated at the same time . but , on the other hand , it would make the proof more difficult to understand , so we decided to give a proof for the euclidean case first , and then to describe the differences and peculiarities of the spherical case in a separate subsection . we begin with the following important lemma : [ lemimp ] take a reuleaux simplex @xmath79 in @xmath0 and the hyperplane @xmath76 containing the vertices @xmath120 of @xmath79 . consider the body @xmath167 . suppose @xmath201 , and suppose that the projection @xmath202 of @xmath191 on the hyperplane @xmath76 lies inside the convex hull @xmath129 of @xmath164 . then @xmath203 with the equality possible in the following two cases : 1 . one of the vertices @xmath204 coincides with one of @xmath164 . the vertex @xmath151 lies in the hyperplane @xmath76 on the border of a reuleaux simplex @xmath205 , constructed on the vertices @xmath164 . at the same time the projection @xmath202 of the vertex @xmath191 on the hyperplane @xmath76 must lie on @xmath206 . it is enough to consider the case when none of @xmath207 coincide with the vertices of @xmath79 . consider the projections @xmath208 of @xmath207 on the hyperplane @xmath76 ( see fig . we have two possibilities : + [ tdplot_main_coords , scale=4 , cube/.style = very thick , black , grid/.style = very thin , gray , axis/.style=->,blue , thick , plane_pnt/.style = color = black!30!red , fill = black!30!red , pnt/.style = color = black , fill = black ] ( a1 ) at ( 0,0,0 ) ; ( a2 ) at ( 1,0,0 ) ; ( a3 ) at ( 0.5,sqrt(3)/2,0 ) ; ( a4 ) at ( 0.5,sqrt(1/12),sqrt(2/3 ) ) ; ( b12 ) at ( @xmath209 ) ; ( b23 ) at ( @xmath210 ) ; ( b13 ) at ( @xmath211 ) ; ( b14 ) at ( @xmath212 ) ; ( b24 ) at ( @xmath213 ) ; ( b34 ) at ( @xmath214 ) ; \(v ) at ( 0.2,0.2,0.3 ) ; ( v ) at ( 0.2,0.2,0 ) ; \(w ) at ( 0.7,0.6,0.2 ) ; ( w ) at ( 0.7,0.6,0 ) ; ; ; ; ; ( , , 0 ) ( , , 0 ) ; ( , , 0 ) ( , , 0 ) ; ( , , 0 ) ( , , 0 ) ; ( , , 0 ) ( , , 0 ) ; ( a1 ) ( a2 ) ; ( a2 ) ( a3 ) ; ( a3 ) ( a1 ) ; ( v ) ( w ) ; ( v ) ( v ) ; ( w ) ( w ) ; ( a3 ) arc ( 180 - 60:180:1 ) ; ( a2 ) arc ( 0:60:1 ) ; ( a1 ) arc ( -90 - 30:-90 + 30:1 ) ; ( @xmath215 ) (@xmath216(@xmath215)!.5!(@xmath217)@xmath218 ) (@xmath217 ) ; ; ( @xmath219 ) (@xmath220(@xmath219)!.5!(@xmath221)@xmath218 ) (@xmath221 ) ; ; ; ( a1 ) circle ( em ) ; ( a2 ) circle ( em ) ; ( a3 ) circle ( em ) ; ( v ) circle ( em ) node[right ] @xmath202 ; ( v ) circle ( em ) node[right ] @xmath191 ; ( w ) circle ( em ) node[below left ] @xmath222 ; ( w ) circle ( em ) node[left ] @xmath151 ; figure 2 \1 . since @xmath202 lies inside @xmath129 and the maximum of the distances from a fixed point to the points of a polytope is attained on the vertices of the polytope , there exists a vertex of @xmath129 , say , @xmath224 , such that @xmath225 and we obtain the following chain of inequalities : @xmath226 the proof of the lemma in this case is complete . assume for a moment that we obtained the inequality ( in this case not strict ) similar to the one in the previous case : there exists a vertex of @xmath129 , say , @xmath224 , such that @xmath228 then we obtain the statement of the lemma from the similar chain of inequalities : @xmath229 unfortunately , we do not have any strict inequality in this chain . however , it is clear that the inequality @xmath230 is strict unless @xmath231 . moreover , it is easy to show that the inequality @xmath232 is strict if @xmath151 is not on the border of @xmath205 , which proves that the equality may hold only if the first part of condition 2 from the lemma is satisfied . indeed , if @xmath151 is in @xmath205 , but not on the border , then change @xmath151 to the point of intersection of the ray @xmath233 with the border of @xmath205 . the distance between @xmath191 and @xmath151 will increase , which makes the inequality strict . therefore , to conclude the proof of lemma [ lemimp ] it is sufficient to show that such a vertex @xmath224 exists ( and to verify that the part of condition 2 from the lemma concerning @xmath202 holds as well ) . we formulate it as a separate lemma : [ lemrelo ] consider a reuleaux simplex @xmath205 in @xmath234 with the set of vertices @xmath164 and two points @xmath235 different from the vertices of @xmath205 , so that @xmath202 lies in the convex hull @xmath129 of @xmath236 then there exists @xmath117 so that @xmath237 , with the equality possible only in case if @xmath202 lies on the boundary of @xmath129 . lemma [ lemrelo ] is proved using a repetitive application of yet another lemma : [ lemred ] consider a closed half - space @xmath238 in @xmath0 bounded by a hyperplane @xmath239 . let @xmath240 be a sphere with center in @xmath158 , where @xmath241 . let @xmath183 be an open region on @xmath240 , @xmath242 consider two points @xmath243 @xmath244 . then one can find a point @xmath245 such that * @xmath246 , if @xmath247 ; * @xmath248 , if @xmath249 . figure 3 the equality from the statement of the lemma is obvious since @xmath183 and @xmath250 both lie on @xmath240 . as for the inequality , consider the two - dimensional plane @xmath251 that contains the points @xmath252 ( see fig . 3 ) . the line @xmath253 splits the plane into two closed halfplanes @xmath254 . let @xmath255 . there is at least one point @xmath245 in @xmath256 , which is different from @xmath257 . then we have the inequality for the angles @xmath258 and , thus , @xmath259 as we have already said , we apply lemma [ lemred ] repeatedly . we may assume that @xmath260 the boundary of a reuleaux simplex can be partitioned into the open spherical regions of different dimensions , as it is stated in lemma [ lem02 ] . we first find an open spherical region @xmath183 that contains @xmath151 and the sphere @xmath60 of minimal dimension that contains @xmath183 . we denote its center by @xmath158 . then we project @xmath129 on the flat @xmath99 of minimum dimension that contains @xmath60 . assume that @xmath99 is an affine hull of the points @xmath261 ( see lemma [ lem02 ] ) . then @xmath262 . the projection is fairly simply arranged . any of the points @xmath182 project into @xmath158 . this is true due to the fact that for any @xmath263 we have @xmath264 to the contrary , all the points @xmath265 project to themselves , since they lie in @xmath266 it is clear that the projection @xmath267 of @xmath202 falls into the projection of @xmath129 . in @xmath99 consider the hyperplane @xmath239 which passes through @xmath158 and is parallel to the hyperplane that contains @xmath265 . the vertices @xmath268 lie in the halfspace @xmath238 in which the whole projection of @xmath129 lie ( recall that @xmath269 ) . now it is possible to apply lemma [ lemred ] and find a point @xmath270 on an open spherical region of a smaller dimension such that @xmath271 moreover , if @xmath272 then @xmath267 can not coincide with @xmath158 , which means that the inequality is strict . by pythagoras theorem we get @xmath273 ( and a strict inequality in the case when @xmath274 ) . reducing the dimension of the spherical region which contains the current image of @xmath151 step by step , we eventually arrive at a vertex of @xmath129 , which concludes the proof of both lemma [ lemrelo ] and [ lemimp ] . consider a diameter graph @xmath1 and two @xmath3-cliques @xmath275 , @xmath276 in @xmath1 . denote by @xmath164 the vertices of @xmath275 . form a rugby ball @xmath162 on @xmath275 and denote the hyperplane containing @xmath275 by @xmath76 . the following step is essential for the proof . consider a @xmath3-dimensional ball @xmath107 , circumscribed around the clique @xmath275 ( it has a center in the center @xmath141 of the clique @xmath275 and radius @xmath277 but has one dimension more than a normal circumscribed ball ) . denote the boundary sphere of @xmath107 by @xmath95 . as usually , denote the @xmath59-dimensional spheres of unit radii with centers in @xmath164 ( the boundary spheres of @xmath278 ) by @xmath279 . the set @xmath280 for any @xmath281 is a sphere that lies in the hyperplane @xmath110 orthogonal to @xmath76 . indeed , it is true due to the fact that both centers ( @xmath141 and @xmath112 ) lie in @xmath76 . this together with observation [ obs0 ] gives us the following crucial observation . [ obs01 ] in the notations introduced above , whenever a point lies in @xmath282 , its projection on the hyperplane @xmath76 falls inside the convex hull @xmath129 of @xmath120 . if a point lies in @xmath283 , then its projection falls strictly inside @xmath129 . suppose that there are at least two vertices @xmath284 of @xmath276 in @xmath285 . if one of them , say @xmath286 , does not lie in @xmath107 , then , by observation [ obs01 ] , its projection on @xmath76 falls strictly inside @xmath129 , and we are done . indeed , checking the conditions in lemma [ lemimp ] that allow @xmath287 to hold , one sees that condition 2 does not take place since the projection of @xmath286 falls strictly inside @xmath129 , so the first condition must hold and , consequently , one of the vertices @xmath288 must coincide with one of the vertices of @xmath275 . the same reasoning apply for @xmath289 . now we are left with two possibilities . + * ( i ) * on both sides of the hyperplane @xmath76 we have at least two points of @xmath276 , or all vertices of @xmath276 lie on one side . this case , which seems to be essential , actually has a short resolution . in this case all points from @xmath276 lie inside the ball @xmath107 , and we are able to use some of its properties . namely , we know that , since @xmath276 is a clique of size @xmath3 , then the radius of the minimal ball that contains the clique equals the radius of @xmath107 ( even though it has a smaller dimension ) . this means that the center of that minimal ball must coincide with @xmath141 , and all the points of @xmath276 must in fact lie on @xmath95 ( otherwise the minimal ball will have a smaller radius ) . this , by observation [ obs01 ] , gives us that all the points of @xmath276 are projected inside @xmath129 . we may then apply lemma [ lemimp ] . the next step is to check the two conditions from the lemma that allow the equality to hold . condition 1 gives that one of the vertices of @xmath276 is as well the vertex of @xmath275 . condition 2 gives that @xmath151 must fall into the set @xmath290 . we note that that @xmath291 is a reuleaux simplex in the hyperplane @xmath76 and @xmath292 is its circumscribed sphere . thus , by lemma [ lem0 ] , we conclude that @xmath293 therefore , in any case some of the vertices of @xmath276 must coincide with some of @xmath164 and we are done in case ( i ) . + * ( ii ) * the other possibility is that exactly one vertex , say , @xmath286 , lies in @xmath77 , while the others lie in @xmath78 . moreover , we may assume that @xmath294 , otherwise , all the vertices of @xmath276 lie inside @xmath107 and we argue as in the previous case . our treatment of this case is as follows . we start with a configuration with @xmath3-cliques @xmath295 without common vertices of the type described above . we perturb the first clique , obtaining a valid configuration of two simplices @xmath296 without common vertices , provided that the initial configuration was valid . thus , if we obtain a contradiction at some point , it means that the initial configuration was as well impossible . we try to perturb the simplex @xmath275 so that @xmath286 will get to the top of the rugby ball @xmath162 , constructed on the perturbed @xmath275 , or , in other words , that @xmath286 will form unit distances with all the ( possibly perturbed ) vertices @xmath297 note that we do not modify @xmath276 . here is the procedure . suppose the distance between @xmath286 and @xmath224 is strictly less than 1 . we rotate @xmath224 around the vertices @xmath298 , which are fixed . the possible trajectory of @xmath224 is a circle , and we push @xmath224 towards @xmath78 . denote the image of @xmath224 by @xmath202 . + we stop the rotation procedure if one of the following two events happen : + * event 1 . * the distance between @xmath202 and @xmath286 is equal to 1 . + * event 2 . * some of @xmath299 fall on the hyperplane @xmath300 , which is the hyperplane that passes through @xmath301 . + before analyzing these two possibilities , we have to mention two facts that we use . denote by @xmath127 the image of @xmath107 under the rotation , and similarly define @xmath302 , @xmath96 , @xmath303 , and @xmath304 . the first fact is that @xmath286 stays inside @xmath302 , moreover , @xmath305 . the first statement holds since we do not move @xmath306 and because of event 1 . as for the second statement , the proof of it relies on the following simple observation : [ obs1 ] consider two points @xmath307 in @xmath0 and a hyperplane @xmath308 that passes through the middle of the segment @xmath309 and is orthogonal to it . denote by @xmath310 the closed halfspace bounded by @xmath308 and that contains @xmath311 . then for any point @xmath312 we have @xmath313 . consider the hyperplane @xmath251 , which contains the intersection of @xmath95 and @xmath96 . it passes through the vertices @xmath306 and through the middle of the segment @xmath314 . we can apply observation [ obs1 ] for the hyperplane @xmath251 and the centers of balls @xmath91 , which are obviously symmetric with respect to @xmath251 . therefore , for the halfspace formed by @xmath251 that contains @xmath191 , which we denote by @xmath315 , we have @xmath316 , while for the other ( denoted by @xmath256 ) it is the other way around . next , one can see that @xmath317 this is due to the fact that any point from @xmath318 projects from above inside the convex hull @xmath129 of the vertices of the reuleaux simplex @xmath319 ( see observation [ obs01 ] ) . we have @xmath320 . since the rotation is made continuously , we may for sure assume that the angular distance between @xmath191 and @xmath202 is less than @xmath321 . therefore , any point that is projected on @xmath76 inside @xmath129 from above , lies in @xmath315 ( see fig . 4 , where points that are projected inside @xmath129 from above , lie in the shaded rectangle ) . from @xmath322 it follows that @xmath323 , and , consequently , @xmath305 . figure 4 the second fact is that no vertex among @xmath299 can escape from @xmath324 without falling onto @xmath300 first . this fact is proved in a similar fashion . consider the hyperplane that contains the intersection of @xmath325 and @xmath326 actually , it is again @xmath251 , moreover , due to observation [ obs1 ] , we have @xmath327 . indeed , since the spheres @xmath328 do not change , we only have to look at the intersection of @xmath329 and @xmath330 , and we fall into a situation which is similar to the one considered in the previous paragraph . therefore , the only way for a point @xmath331 to escape @xmath302 is to fall onto @xmath251 first . but this is not possible , because any position of @xmath251 was a position of @xmath300 at the earlier stage of rotation , so the point @xmath331 has to fall onto @xmath300 first . we go on to the analysis . suppose that event 2 happens , and the point @xmath332 from @xmath276 falls onto @xmath300 . then we have two vertices of @xmath276 in @xmath333 , and by lemma [ lemimp ] we are done unless @xmath332 coincides with one of the vertices of @xmath304 . moreover , we are done unless @xmath334 , which means that before the rotation @xmath332 lay on the same arc as @xmath202 . however , the projection of this arc on the hyperplane @xmath76 is a straight segment connecting the vertex @xmath224 and the center @xmath141 . therefore , the projection of @xmath332 should fall inside @xmath129 , and , since there is more than one point in @xmath78 , we are done by lemma [ lemimp ] . thus , the second event either leads us to a contradiction , if we assume that there is no common vertex of @xmath275 and @xmath276 , or gives the common vertex of @xmath275 and @xmath276 . if event 1 happens , then we take another vertex of @xmath275 and proceed in the same way . finally , assume that we finished the procedure and have not found the common vertex of @xmath275 and @xmath276 yet . denote by @xmath335 the images of the vertices of @xmath275 after all the rotations . then @xmath286 forms a unit @xmath3-simplex with @xmath335 . in this case all the vertices @xmath336 lie on the unit sphere with center in @xmath286 . we are left to show that a unit @xmath59-simplex and a unit @xmath97-simplex on a @xmath59-dimensional sphere of radius greater than @xmath10 must share a common vertex , which follows from the ( second part of the ) statement of theorem [ thcom ] by induction . in what follows we will work on a @xmath3-dimensional sphere @xmath337 of radius greater than @xmath10 . spherical geometry is very similar to euclidean . to make the proof work in this case , one should , more or less , only change the notation : planes are changed to diametral ( or great ) spheres , halfspaces to hemispheres , balls to spherical caps . we will often use the euclidean names for the spherical objects , e.g. , say `` a plane '' instead of `` a diametral sphere '' . this should not cause confusion , since we will mostly work in terms of the internal spherical geometry . however , when it is convenient , we will think of the sphere as a subset of a euclidean space , and interpret points of @xmath337 as vectors . in the next several paragraphs we will point out some properties of spherical geometry that we will use in the proof . for an introduction to elementary spherical geometry we refer the reader to chapter 1 of the book due to l. fejes tth @xcite . for a systematic treatment of spherical geometry , that by far covers all the material used in this paper we refer to @xcite . * there is a natural way to assign dimensions to spherical planes , such that the definition will work the same way as in the euclidean case . namely , the dimension of a diametral sphere is equal to the dimension of the minimal euclidean plane that contains it . note that a spherical line consists of two points . * 2 . * for a flat ( diametral sphere ) @xmath251 denote by @xmath338 the maximal flat such that any vector from @xmath338 is orthogonal to any vector in @xmath251 . if @xmath251 is a hyperplane ( diametral sphere of codimension 1 ) , by @xmath339 we denote the closed half - spaces bounded by @xmath251 . * 3 . * for a given hyperplane @xmath251 and an arbitrary point @xmath340 we can define the projection of a point @xmath191 to @xmath251 . consider the two - point set @xmath338 . then the projection @xmath202 of @xmath191 on @xmath251 is the closest intersection point to @xmath191 of the great circle that goes through @xmath338and @xmath191 with the plane @xmath251 . we define the reflection @xmath341 with respect to a given hyperplane @xmath251 . for any given point @xmath191 in @xmath340 we consider the great circle that contains @xmath338 and @xmath191 , and find a point @xmath342 on that circle , which is symmetric to @xmath191 with respect to the projection of @xmath191 on @xmath251 . as for the @xmath338 , the reflection interchanges the two points in @xmath338 . using reflections , it is easy to introduce the notion of orthogonality to a hyperplane , which would be convenient for us . namely , a plane @xmath99 is orthogonal to a hyperplane @xmath251 , if @xmath343 . * suppose that we have a @xmath18-sphere @xmath183 on @xmath337 , which is not diametral . it is easy to show that any such sphere is contained in a ( spherical ) plane @xmath251 of dimension @xmath344 . indeed , taking a euclidean point of view , for any @xmath18-sphere there is a @xmath345-sphere with center in the center of @xmath337 that contains @xmath183 . this is the desired plane ( of dimension one higher than the dimension of the sphere by definition ) . note that this is the minimal plane that contains @xmath183 . * for points in an open hemisphere @xmath346 of @xmath337 one can easily define the distance between two points as the shorter angle between the corresponding vectors . in particular , the distance between any point in @xmath251 and any point in @xmath338 is @xmath347 . we denote the spherical distance between @xmath348 by @xmath349 . * one could define an angle between the two intersecting arcs as the dihedral angle between the corresponding vector planes . for three distinct points @xmath350 we denote by @xmath351 the angle between the arcs @xmath352 and @xmath353 . * there is a version of pythagoras theorem for spherical triangles . namely , given a right spherical triangle @xmath354 in @xmath346 with @xmath355 one have @xmath356 moreover , pythagoras theorem is a corollary of the spherical cosine law : @xmath357 out of this one can deduce the following statement : suppose we are given three distinct points @xmath350 and the angle @xmath351 between the arcs @xmath352 and @xmath353 is at least @xmath347 . assume moreover , that @xmath358 are less than @xmath347 . then @xmath359 . we need the notion of a convex hull of the set of points @xmath360 . the straightforward way to define it is by using the euclidean interpretation . it is simply the intersection of the cone formed by vectors corresponding to @xmath361 and @xmath337 . note that the boundary of such a convex hull is formed by planes ( diametral spheres ) . the distance in @xmath337 , that corresponds to euclidean distance 1 , we denote by @xmath362 . note that , since in our case @xmath337 has radius greater than @xmath10 , we have @xmath363 . suppose we are given a diameter graph on @xmath337 , which contains two simplices @xmath275 and @xmath276 on @xmath3 vertices . we consider the spherical rugby ball @xmath162 , formed by vertices of @xmath275 , and the diametral sphere @xmath76 that contains @xmath275 . the proof stays almost the same as in the euclidean case . we describe all the differences in what follows . all the notations are translated to this case from the euclidean case . first , we show that the spherical rugby ball @xmath162 is contained in one of the open hemispheres of @xmath337 . we denote such a hemisphere by @xmath346 . consider the unit ball @xmath329 with the center in @xmath224 ( one of the vertices of @xmath275 ) . on one hand , since the radius of @xmath337 is bigger than @xmath364 , @xmath329 is contained in the open hemisphere @xmath346 with the center in @xmath224 . on the other hand , surely , @xmath365 . consider a spherical reuleaux simplex @xmath79 and a point @xmath311 inside the ( spherical ) convex hull of its vertices . then an open halfsphere with the center in @xmath311 contains @xmath79 . this is due to the fact that @xmath79 is contained in the intersection of open halfspheres with centers in the vertices of @xmath79 , therefore , if we think about the vector interpretation of the situation , any vector @xmath366 from @xmath79 has positive scalar products with any of the vectors representing the vertices . since @xmath311 is a convex combination of the vectors representing the vertices , @xmath311 and @xmath366 have positive scalar product . we formulate it as the first part of the following observation : [ obs2 ] 1 . consider a spherical reuleaux simplex @xmath79 on @xmath337 and a point @xmath311 that lies in the spherical convex hull of the vertices of @xmath79 . then for any @xmath367 we have @xmath368 . + 2 . consider any diametral hypersphere @xmath251 on @xmath337 and any point @xmath311 , @xmath369 . let @xmath370 be the projection of @xmath311 to @xmath251 . then @xmath371 . lemma [ lemimp ] holds for the spheres . here is what we have to check : * i. * we use the following implication extensively : consider the points @xmath372 and @xmath373 for some plane @xmath251 . denote by @xmath202 the projection of @xmath191 to @xmath251 . if @xmath374 for some @xmath222 in @xmath375 , then @xmath376 this statement follows from the spherical pythagoras theorem ( point 9 from the previous subsection ) and from observation [ obs2 ] . the proof of the lemma in the spherical case should start from one of the following chain of inequalities : either @xmath377 or @xmath378 here @xmath208 are the projections of @xmath204 to the base of the rugby ball . the second inequality in both chains is justified using the previous point ( and also the next two points ) . in the euclidean case the choice of one chain of inequalities depends on which point among @xmath207 is closer to @xmath76 ( this gives us the first inequality in both chains ) . in the spherical case we can also choose one chain , since @xmath379 and by the last property from point 9 from the previous subsection , combined with observation [ obs2 ] , @xmath380 . the angle inequality holds since the vertices @xmath381 form a quadrangle with two right angles , while the sum of the angles in a quadrangle on the sphere is more than @xmath382 . * we have to verify that lemma [ lemred ] works in the spherical case . the proof goes word for word , but we present it here with the proof for the seek of clarity . [ lemreds ] consider a closed halfsphere @xmath238 in @xmath337 bounded by a diametral hypersphere @xmath239 . let @xmath240 be a sphere with center in @xmath158 , where @xmath241 . let @xmath183 be an open region on @xmath240 , @xmath242 consider two points @xmath243 @xmath244 . then one can find a point @xmath245 such that * @xmath383 , if @xmath247 ; * @xmath384 , if @xmath249 . the equality from the statement of the lemma is obvious since @xmath183 and @xmath250 both lie on @xmath240 . as for the inequality , consider the two - dimensional diametral sphere @xmath251 that contains the points @xmath252 . the diametral circle @xmath253 splits @xmath251 into two closed halfspheres @xmath254 . let @xmath255 . there is at least one point @xmath245 in @xmath256 , which is different from @xmath257 . then we have the inequality for the angles @xmath385 and , by the spherical law of cosines ( [ cos ] ) we get that @xmath386 this is because the first summand in ( [ cos ] ) stays the same in both cases , while in the second one the only change is that the last multiple gets smaller . note that we do not need to put any restrictions on @xmath387 , as it is done in the end of point 9 of the previous subsection . * iv . * the rest of the proof of the spherical version of lemma [ lemimp ] goes the same , since the projections of the set @xmath129 behave the same as in the euclidean case . + having finished the description of the changes in the proof of lemma [ lemimp ] in the spherical case , we return to the proof of the theorem . as in the euclidean case , spheres @xmath280 lie in the hyperplane @xmath110 , which is orthogonal to @xmath76 . the first thing we note is that @xmath280 is a @xmath97-sphere that is not diametral . it would be diametral only if both @xmath95 and @xmath388 are diametral , which is not the case . thus , by point 6 from the previous subsection , the minimal plane that contains @xmath280 is of codimension 1 . next , note that @xmath389 this is due to the fact that both centers of @xmath95 and @xmath388 lie in @xmath76 . hence , the same should hold for @xmath110 , and by point 5 from the previous subsection we obtain the desired orthogonality . combining the spherical version of lemma [ lemimp ] and the fact that @xmath110 is orthogonal to @xmath76 with the euclidean proof , we almost come to the last case in which we have one vertex of @xmath276 in @xmath390 while the rest lie in @xmath391 . however , we should make sure that the circumscribed ball considerations still work in the spherical case . indeed , suppose we are given two unit simplices @xmath275 , @xmath276 on @xmath3 vertices . suppose @xmath276 lies inside the @xmath3-dimensional ball @xmath107 of diameter @xmath392 , which is a ball of minimal diameter that contains @xmath275 . denote by @xmath393 the circumscribed ball for @xmath276 . then , if @xmath107 and @xmath393 do not coincide , the intersection @xmath394 is contained in a ball of smaller radius . indeed , if we choose an arbitrary point @xmath395 on the open segment connecting the centers @xmath396 of @xmath107 and @xmath393 , respectively , then by the last property in point 9 from the previous subsection , combined with the fact that @xmath397 for any @xmath398 we have @xmath399 since either @xmath400 or @xmath401 is at least @xmath347 . we obtain that @xmath276 is contained in a ball of radius strictly smaller than @xmath392 , which is impossible . thus , the centers of @xmath107 and @xmath393 coincide , and the rest of the argument in this case works exactly as for the euclidean space . returning to the last part of the proof , one can translate it word for word from the euclidean case . we note that the plane @xmath251 satisfies the following equation : @xmath402 , which is why @xmath403 . indeed , @xmath404 and @xmath405 is less than a halfball , while @xmath406 is bigger than a halfball . similar reasoning applies for the inclusion @xmath407 . finally , consider the case when @xmath332 falls into @xmath300 and , moreover , @xmath332 coincides with @xmath202 . to conclude the proof , we need to check that the arc on which @xmath408 lie projects inside the spherical convex hull of vertices @xmath120 . this is equivalent to the statement that the circle @xmath409 and the sphere @xmath95 touch at @xmath224 ( in the plane @xmath76 ) . in that case the circle would lie in the exterior of the ball @xmath107 . the fact about touching is clear , so the point @xmath332 lie in @xmath283 , and we can apply the spherical analogue of lemma [ lemimp ] . the proof is complete . we are grateful to filip mori and jnos pach for several interesting and stimulating discussions on the problem , eyal ackerman for pointing out the paper @xcite , ido shahaf for his valuable remarks on the presentation of the paper and for providing us with figure 2 , to arseniy akopyan and roman karasev for valuable discussions and for bringing kirszbraun s theorem to our attention , and anonymous referees for numerous valuable comments that helped to significantly improve the presentation of the paper . alekseevskij , e.b . vinberg , a.s . solodovnikov , _ geometry of spaces of constant curvature _ ( english . russian original ) geometry . ii : spaces of constant curvature , encycl . 29 , 1 - 138 ( 1993 ) ; translation from itogi nauki tekh . probl . mat . napravleniya 29 , 5 - 146 ( 1988 ) . f. mori and j. pach , _ large simplices determined by finite point sets _ , beitr . algebra geom . 54 ( 2013 ) , n1 , 4557 . a.m. raigorodskii , _ around borsuk s conjecture _ , j. of math . , 154 ( 2008 ) , n4 , 604 623 . z. schur , m. a. perles , h. martini , y. s. kupitz , _ on the number of maximal regular simplices determined by n points in @xmath0 _ , discrete and computational geometry , the goodman- pollack festschrift , aronov etc . eds . , springer , 2003 .

in this paper we present a general procedure allowing for the reduction or expansion of any network ( considered as a weighted graph ) which maintains the spectrum of the network s adjacency matrix up to a set of eigenvalues known beforehand from its graph structure . this procedure can be used to establish new equivalence relations on the class of all weighted graphs ( networks ) where two graphs are equivalent if they can be reduced to the same graph . we show such transformations are useful in determining whether a dynamical network i.e. a network of interacting dynamical systems has a globally attracting fixed point . + e - mail : bunimovich@math.gatech.edu and bwebb@math.gatech.edu

1010.3272

real world networks , i.e. those found in nature and technology , typically have a complicated irregular structure and consist of a large number of highly interconnected dynamical units . coupled biological and chemical systems , neural networks , social interacting species , and the internet are only a few such examples @xcite . because of this complexity , the first approach to capturing the global properties of such systems has been to model them as graphs whose nodes represent elements of the network and the edges define a topology ( graph of interactions ) between these elements . this a principally a static approach to modeling networks in which only the structure of the network s interactions ( and perhaps weights associated to them ) are analyzed . an additional line of research has been the investigation of the dynamical properties of such networks . this has been done by modeling network via interacting dynamical systems @xcite . important processes that are studied within this framework include synchronization @xcite or contact processes , such as opinion formation and epidemic spreading . these studies give strong evidence that the structure of a network can have a substantial impact on its dynamics @xcite . regarding this connection , the impact of a dynamical network s underlying structure on its dynamics has been of recent concern . in this the spectrum of the network s adjacency matrix has emerged as a key quantity in the study of a variety of dynamical networks . for example , systems of interacting dynamical units are known to synchronize depending only on the dynamics of the uncoupled dynamical systems and the spectral radius of the network adjacency matrix @xcite . moreover , the eigenvalues of a network are important for determining if the dynamics of a network is stable @xcite . to aid in understanding this interplay between the structure ( graph of interactions ) and dynamics of a network we introduce in this paper the concept of an _ isospectral graph transformation_. such transformations allow one to modify the structure of a network at the level of a graph while maintaining properties related to the network s dynamics . mathematically , an isospectral graph transformation is a graph operation that modifies the structure of a graph while preserving the eigenvalues of the graph s ( weighted ) adjacency matrix . we note here that besides modifying interactions , such transforms can also reduce or increase the number of nodes in a network . as not to violate the fundamental theorem of algebra , isospectral graph transformations preserve the spectrum of the graph ( specifically number of eigenvalues ) by allowing edges to be weighted by rational functions . we note that by allowing such weights it may appear that our procedure is trading the complexity of the graph s structure for complex edge weights . however , this is not the case . in fact , it is often possible isospectrally reduce a graph in such a way that the reduced graph s edges weights belong to a particularly nice set e.g. positive integers , real numbers , or the for unweighted graphs the weight set @xmath1 ( see theorem [ maintheorem ] ) . for a graph @xmath2 with vertex set @xmath3 this is done by reducing or expanding @xmath2 with respect to specific subsets of @xmath3 known as structural sets ( see section 3.1 for exact definitions ) . as a typical graph has many different structural sets it is possible to consider different isospectral transformations of the same graph @xmath2 as well as sequences of such transformations as a reduced graph generally has its own structural sets . in this regard , the flexibility of this procedure is reflected in the fact that for a typical graph @xmath2 it is possible to sequentially reduce @xmath2 to a graph on any nonempty subset of its original vertex set . this follows from an existence and uniqueness result regarding sequences of isospectral reductions . that is a typical graph @xmath2 can be uniquely reduced to a graph on any nonempty subset of its vertices ( see theorem [ theorem-1 ] and theorem [ theorem0.1 ] ) . because of this uniqueness , isospectral graph reductions can be used to induce new equivalence relations on the set of all graphs under some mild assumptions . namely , two graphs ( networks ) are _ spectrally equivalent _ if they can by isospectrally reduced to one and the same graph . in the study of complex networks , there is a need to find ways of reducing complexity while maintaining some important network characteristic(s ) . this amounts to course graining or finding the right scale at which to view the network . isospectral network transformations , specifically reductions , serve this need by allowing one to view a network as smaller network with essentially the same spectrum . finding the right scale here then amounts to finding the right structural set over which to reduce the network . of primary interest is the fact that isospectral network transformations suggest other useful transformations on interacting dynamical systems i.e. dynamical networks . an important example in this paper is a _ dynamical network expansion _ in which a dynamical network is modified in a way that preserves its dynamics but alters its associated graph structure . such transforms provide a new tool for the study of the interplay between the structure ( topology ) and dynamics of dynamical networks . much as general dynamical systems are investigated via change of coordinates , isospectral network transforms introduce a mechanism for rearranging the specific ( graph ) structure of a system while preserving its dynamics in an essential way . by so doing the original network s dynamics can be investigated by studying the transformed dynamical network . specifically , we generalize the results given in @xcite by showing in this paper that the existence of a global attracting fixed point of a network can be established by investigating one of its expansions when direct investigation of the network does not . the structure of the paper is as follows . in section 2 we present notation and some general definitions . section 3 contains the description of an isospectral graph transformation and results on sequences of such transformations . graph transformations over fixed weight sets and spectral equivalence are treated in section 4 . the proof of results contained in sections 3 and 4 are then given in section 5 . in section 6 the behavior of dynamical networks is investigated via dynamical network expansions and the condition of global stability is obtained . section 7 contains some concluding remarks . in this paper we are primarily interested in the topology ( graph structure ) of dynamical networks . we consider the most general class of networks ( with fixed topology i.e. the graph of interactions does not change in time ) , namely those networks whose graph of interactions are finite , directed , with loops and with edge weights in the set @xmath4 . such graphs form the class @xmath5 . a graph @xmath6 is an ordered triple @xmath7 where @xmath3 and @xmath8 are the _ vertex set _ and _ edge set _ of @xmath2 respectively . moreover , @xmath9 gives the _ edge weights _ of @xmath2 where @xmath10 if and only if @xmath11 . the set @xmath4 is defined as follows . let @xmath12 $ ] be the set of polynomials in the complex variable @xmath13 with complex coefficients . the set @xmath4 is the set of rational functions of the form @xmath14 where @xmath15 $ ] , @xmath16 and @xmath17 have no common factors ( roots ) , and @xmath18 . we note that the set @xmath4 is a subfield of the field of rational functions over @xmath19 . if the vertex set of the graph @xmath7 is labeled @xmath20 we denote an edge from @xmath21 to @xmath22 by @xmath23 . the matrix @xmath24 defined entrywise by @xmath25 is then the _ weighted adjacency matrix _ of @xmath2 . the _ spectrum _ or _ eigenvalues _ of a matrix @xmath26 are the solutions including multiplicities to the equation @xmath27 for the graph @xmath2 we let @xmath28 be the spectrum of @xmath29 . the spectrum of a matrix with entries in @xmath4 is therefore a generalization of the spectrum of a matrix with complex entries . moreover , as @xmath30 is the ratio of two polynomials @xmath31 let @xmath32 be the solutions including multiplicities of the equation @xmath33 . both @xmath28 and @xmath32 are _ lists _ of numbers . specifically , @xmath34 where @xmath35 is the multiplicity of the solutions @xmath36 to ( [ eq1 ] ) i.e. eigenvalues , @xmath37 the number of distinct solutions , and @xmath38 the elements in the list . the list @xmath32 has a similar formulation . in this section we formally describe the isospectral reduction process and the associated isospectral graph expansions . we then give specific examples of this method and present some results regarding sequences of isospectral reductions and spectral equivalence of graphs . as a typical graph @xmath6 can not be reduced over any arbitrary subset of its vertices ( by means of a single reduction ) we define the set of _ structural sets _ of a graph or those vertex sets for which this is possible . if @xmath39 where @xmath3 is the vertex set of a graph we write @xmath40 . also , a _ loop _ is an edge that begins and ends at the same vertex . [ def1 ] for @xmath7 let @xmath41 be the graph @xmath2 with all loops removed . the nonempty vertex set @xmath39 is a _ structural set _ of @xmath2 if @xmath42 contains no cycles and @xmath43 for each @xmath44 . we denote by @xmath45 the set of all structural sets of @xmath2 . to describe how a graph @xmath7 is reduced over the structural set @xmath46 we require the following standard terminology . a _ path _ @xmath47 in the graph @xmath7 is an ordered sequence of distinct vertices @xmath48 such that @xmath49 for @xmath50 . in the case that the vetices @xmath51 are distinct , but @xmath52 , then @xmath47 is a _ cycle_. moreover , we call the vertices @xmath51 of @xmath47 the _ interior _ vertices of @xmath47 . for @xmath7 with structural set @xmath53 let @xmath54 be the set of paths or cycles with no interior vertices in @xmath55 from @xmath21 to @xmath22 . furthermore , let @xmath56 we call the set @xmath57 the _ branches _ of @xmath2 with respect to @xmath55 . if @xmath58 for @xmath59 and @xmath60 let @xmath61 if @xmath62 then @xmath63 . we say @xmath64 is the branch product of @xmath65 . let @xmath7 with structural set @xmath66 . the graph @xmath67 where @xmath68 if @xmath69 and @xmath70 is the _ isospectral reduction _ of @xmath2 over @xmath55 . [ maintheorem ] if @xmath46 then @xmath71 . to make sense of the formula in theorem [ maintheorem ] we note that @xmath72 since the graph @xmath73 contains no ( nonloop ) cycles . hence , @xmath74 are the values at which the loops weights of @xmath75 equal @xmath13 and @xmath76 the values at which these weights are undefined . hence , both @xmath74 and @xmath75 equal @xmath13 can be read directly from the graph @xmath2 . figure 1 gives an example of an isospectral reduction of the graph @xmath2 over the set @xmath77 . here , using our list notation , one can compute the spectrum @xmath78 . as @xmath75 has loop weights @xmath79 @xmath80 , @xmath81 , and @xmath82 then @xmath83 and @xmath84 . theorem [ maintheorem ] then implies that @xmath85 . we note that it is possible to reduce a graph without losing any eigenvalues . for instance , if @xmath86 where @xmath87 for @xmath88 it follows that @xmath89 . hence , for any @xmath90 , @xmath91 since @xmath92 and @xmath93 . algebraically an isospectral reduction of a graph can be described as follows . if @xmath46 then the vertices of @xmath2 can be ordered such that @xmath94 has the block form @xmath95 where @xmath96 corresponds to the vertices in @xmath75 and is a triangular matrix with nonzero diagonal . the matrix @xmath96 can be made triangular from the fact that @xmath73 contains no ( nonloop ) cycles whereas the nonzero diagonal follows from the condition that @xmath43 for @xmath44 ( see _ frobenius normal form _ in @xcite ) . hence , @xmath96 is invertable . using the identity @xmath97 it follows that @xmath98 given @xmath99 , for some @xmath100 the isospectral reduction , @xmath101 is the graph with adjacency matrix @xmath102 ( see proof of lemma ( [ lemma1.2 ] ) in section 5 ) . moreover , as @xmath96 corresponds to the vertex set @xmath75 then by equation ( [ eq1.1 ] ) @xmath103 nonlin1.eps ( 28.5,-3)@xmath2 ( 4,10.5)@xmath104 ( 14,10.5)@xmath105 ( 75,10.5)@xmath105 ( 28.5,11.5)@xmath106 ( 28.5,2.5)@xmath107 ( 42,10.5)@xmath108 ( 93,10.5)@xmath108 ( 53,10.5)@xmath109 ( 83,12)@xmath110 ( 85,1.5)@xmath111 ( 65,7)@xmath110 ( 100.5,7)@xmath112 ( 82,-3)@xmath101 it is natural to define an _ isospectral graph expansion _ of a graph @xmath2 as a graph @xmath113 where @xmath114 for some @xmath115 . such expansions can be carried out by simply expanding edges into branches or multiple branches with the correct products and sums ( e.g. ( [ eq0.9 ] ) and ( [ eq1.0 ] ) ) . however , in contrast to graph reductions , such expansions are nonunique . for unique expansions see section 4.1 . as any reduction @xmath101 of a graph @xmath6 is again a graph in @xmath5 it is natural to consider sequences of reductions of a graph as well as to what degree a graph can be reduced . to do so we extend our notation to an arbitrary sequence of reductions . for @xmath7 suppose @xmath116 such that @xmath117 , @xmath118 and @xmath119 if this is the case we say @xmath120 _ induces a sequence of reductions _ on @xmath2 with _ final vertex set _ @xmath121 and write @xmath122 . [ theorem3]*(commutativity of reductions ) * for @xmath7 suppose @xmath123 . if @xmath124 then @xmath125 induces a sequence of reductions on @xmath2 where @xmath126 . that is , the final vertex set in a sequence of reductions completely specifies the reduced graph irrespective of the specific sequence of reductions if this vertex set is a structural set . to address whether a graph @xmath7 may be reduced via some sequence of reductions to a graph on an arbitrary subset @xmath127 we note the following if @xmath128 for some @xmath129 then @xmath130 since @xmath21 induces no cycles in @xmath41 . hence , any single vertex with this property can be removed from the graph . if it is somehow known a priori that no loop of @xmath2 or any sequential reduction of @xmath2 has weight @xmath13 ( which is always the case in applications ) then @xmath2 can be sequentially reduced to a graph on any subset of its vertex set . to give such a condition we define the set @xmath131 . for @xmath15 $ ] where @xmath132 let the function @xmath133 . let @xmath134 be the set of graphs with the property that for any @xmath135 , @xmath136 for each entry . [ lemma1 ] if @xmath135 and @xmath46 then @xmath137 . in particular , no loop of @xmath2 and no loop of any reduction of @xmath2 can have weight @xmath13 . by the reasoning above , if @xmath135 then @xmath2 can be ( sequentially ) reduced to a graph on any subset of its vertex set . moreover , the following holds . [ theorem-1]*(existence and uniqueness of reductions over any set of vertices ) * let @xmath7 be in @xmath138 . then for any nonempty @xmath127 any sequence of reductions on @xmath2 with final vertex set @xmath139 reduces @xmath2 to the unique graph @xmath140=(\mathcal{v},\mathcal{e},\mu)$ ] . moreover , at least one such sequence always exists . the notation @xmath141 $ ] in the previous theorem is intended to emphasize the fact that @xmath139 need not be a structural set of @xmath2 . moreover , this notation is well defined given that there is a unique reduction of @xmath2 over @xmath139 in the sense of theorem [ theorem-1 ] . if @xmath142 then @xmath135 . therefore , any graph with complex weights can be uniquely reduced to a graph on any nonempty subset of its vertex set . this is of particular importance for the estimation of spectra of matrices with complex entries in @xcite . here we consider specific types of vertex sets that are defined independent of any graph of @xmath5 . our motivation is that every nonempty graph in @xmath138 will have a unique reduction with respect to this type of vertex set . this will in turn allow us to partition the graphs in @xmath143 according to their reductions . two weighted digraphs @xmath144 , and @xmath145 are _ isomorphic _ if there is a bijection @xmath146 such that there is an edge @xmath23 in @xmath147 from @xmath21 to @xmath22 if and only if there is an edge @xmath148 between @xmath149 and @xmath150 in @xmath151 with @xmath152 . if the map @xmath153 exists it is called an _ isomorphism _ and we write @xmath154 . moreover , we note that if two graphs are isomorphic then their spectra are identical . * ( uniqueness and equivalence relations)*[theorem0.1 ] suppose for any graph @xmath7 in @xmath155 that @xmath156 is a rule that selects a unique nonempty subset @xmath157 . then @xmath156 induces an equivalence relation @xmath158 on the set @xmath155 where @xmath159 if @xmath160\simeq\mathcal{r}_{\tau(h)}[h]$ ] . as an example consider the following rule . for a graph @xmath7 where @xmath135 let @xmath161 be the set of vertices of minimal out degree . if @xmath162 then by theorem [ theorem-1 ] @xmath163 $ ] is uniquely defined and this process may be repeated until all vertices of the resulting graph have the same out degree . as the final vertex set of this sequence of reductions is uniquely defined then the relation of having an isomorphic reduction via this rule induces an equivalence relation on @xmath131 . nonlin2.eps ( 1.5,30.5)@xmath105 ( 1.5,4)@xmath106 ( -3.5,17)@xmath164 ( 8.5,17)@xmath165 ( 1.5,-2)@xmath2 ( 49,-2)@xmath113 ( 38.5,30.5)@xmath105 ( 38.5,4)@xmath106 ( 60.5,30.5)@xmath108 ( 60.5,4)@xmath107 ( 95.5,30.5)@xmath108 ( 95.5,4)@xmath107 ( 95.5,-2)@xmath166 ( 102,17)@xmath164 ( 83,17)@xmath167 note that the relation of simply having isomorphic reductions is not transitive . that is , if @xmath168\simeq\mathcal{r}_t[h]$ ] and @xmath169\simeq\mathcal{r}_v[k]$ ] it is not necessarily the case that there are sets @xmath170 and @xmath171 , subsets of the vertex sets of @xmath2 and @xmath166 respectively , such that @xmath172\simeq\mathcal{r}_y[k]$ ] . for instance , in figure 2 both @xmath168=\mathcal{r}_s[h]$ ] and @xmath173=\mathcal{r}_t[k]$ ] where @xmath174 and @xmath175 . however , one can quickly check that for no subsets @xmath176 and @xmath177 are @xmath172\simeq\mathcal{r}_y[k]$ ] . a general rule @xmath156 that selects a unique set of vertices from each graph is needed to overcome this intransitivity in general . the graph reductions and expansions of the previous section modify not only the graph structure but also the weight set associated with each graph . that is , if @xmath7 and @xmath67 then typically @xmath178 . thus , one may think that our procedure simply shifts the complexity of the graph s structure to its set of edge weights . however , this is not the case . in this section we introduce a procedure of reducing and expanding a graph while maintaining its set of edge weights where , as before , the procedure preserves the spectrum of the graph up to a known set ( list ) . such transformations are of particular importance in section 6 where dynamical network expansions are discussed . given a graph @xmath7 and @xmath46 , two branches @xmath179 are said to be _ independent _ if they have no interior vertices in common . moreover , if the branch @xmath58 let @xmath180 be the ordered sequence @xmath181 [ def10 ] for @xmath7 with structural set @xmath53 let @xmath182 be the graph where @xmath183 and the following holds : + ( i ) there are bijections @xmath184 for all @xmath185 such that if @xmath186 then @xmath187 . + ( ii ) the branches in @xmath188 are pairwise independent . + ( iii ) each vertex of @xmath189 is on a branch of @xmath188 . as this uniquely defines @xmath189 , up to a labeling of vertices , we call this graph the _ branch expansion _ of @xmath2 over @xmath55 . essentially , @xmath189 is the graph @xmath2 in which every pair of branches with respest to @xmath55 have been made independent . [ theorem-2 ] let @xmath7 and @xmath46 . if each vertex of @xmath2 is on a branch of @xmath57 then @xmath2 and @xmath189 have the same weight set , @xmath190 , and @xmath191 . moreover , @xmath192 where @xmath35 is the number of branches in @xmath57 containing @xmath21 . as an example of a branch expansion consider the graph @xmath113 in figure 3 where @xmath77 . given that @xmath193 one can compute @xmath194 via theorem [ theorem-2 ] . nonlin3.eps ( 0,20)@xmath105 ( 15,20)@xmath106 ( 15,-0.5)@xmath108 ( 0,-0.5)@xmath107 ( 9,-3)@xmath113 ( 43.5,11.5)@xmath104 ( 55,11.5)@xmath105 ( 70,13)@xmath109 ( 70,3)@xmath195 ( 85,11.5)@xmath108 ( 98,11.5)@xmath196 ( 68,-3)@xmath197 using techniques from both isospectral graph reductions and branch expansions it is often possible to reduce a graph over some fixed subset of weights @xmath198 while again preserving the graph s spectrum up to some known set ( list ) . to make this precise we require the following . let @xmath2 and @xmath113 be graphs such that @xmath199 for some @xmath46 , @xmath115 . then we say @xmath2 and @xmath113 are _ spectrally equivalent_. as an example , the graphs @xmath2 in figure 1 and @xmath113 in figure 3 are spectrally equivalent since @xmath200 for @xmath77 . if @xmath2 and @xmath113 are spectrally equivalent , @xmath201 , and both @xmath2 and @xmath113 have weight sets in some subset @xmath202 we say @xmath113 is a reduction of @xmath2 over the weight set @xmath203 . if @xmath204 then @xmath113 is an expansion of @xmath2 over the weight set @xmath203 . for @xmath7 and any @xmath202 containing @xmath205 an example of an expansion of @xmath2 over the weight set @xmath203 is any branch expansion of @xmath2 . finding reductions over some set @xmath203 is more complicated . however , if @xmath203 is a unital subring of @xmath4 and we restrict our attention to specific types of structural sets it is often possible to simply construct a reduction of @xmath2 over the set @xmath203 as follows . * the @xmath206-construction * for @xmath7 let @xmath207 be the collection of all structural sets of @xmath2 such that @xmath208 if @xmath209 for each @xmath210 . suppose @xmath7 and @xmath208 where each vertex of @xmath2 is on a branch of @xmath57 . if @xmath53 then for @xmath211 let @xmath212 be a branch of maximal length for each @xmath213 . as a first intermediate step let @xmath214 be the graph @xmath189 with all but the branches @xmath215 removed i.e. @xmath216 suppose each @xmath215 is denoted by the sequence of vertices @xmath217 where @xmath218 is the length of @xmath215 or number of edges . note each @xmath219 and @xmath220 . we then reweight the edges of @xmath214 in the following way . make the edge from @xmath221 to @xmath222 have weight @xmath223 for each @xmath224 and set all other edges of @xmath214 to weight 1 . next , for each branch @xmath225 from @xmath21 to @xmath22 attach an edge from vertex @xmath21 to vertex @xmath226 in @xmath214 with weight @xmath227 . if two or more branches correspond to the same edge we reduce these to a single edge with weight equal to the sum of the corresponding branch products . we call the resulting graph @xmath228 or the @xmath229 of @xmath2 over @xmath55 . by construction , if @xmath2 has weights in @xmath203 , some unital subring of @xmath4 , then so does @xmath228 . [ theorem10 ] let @xmath2 have weights in the unital subring @xmath202 and @xmath208 . then @xmath2 and @xmath228 are spectrally equivalent and have the same nonzero spectrum . moreover , if @xmath230 then @xmath228 is a reduction of @xmath2 over @xmath203 otherwise it is an expansion . nonlin4.eps ( 0,-2)@xmath105 ( 44,-2)@xmath106 ( 15,15)@xmath108 ( 29,15)@xmath107 ( 22,10)@xmath104 ( 22,-3)@xmath166 ( 54,-2)@xmath105 ( 98,-2)@xmath106 ( 68,15)@xmath108 ( 83,15)@xmath107 ( 75,-3)@xmath231 an example of an @xmath229 is the graph @xmath231 in figure 4 , which is constructed from the graph @xmath166 over @xmath232 having weights in the unital subring @xmath233 . in this section we give proofs of the theorems and lemma in sections 3 and 4 of this paper . our strategy will be to develop an analogous theory of matrix reductions from which such results will follow . let @xmath234 and @xmath235 be nonempty . if the set @xmath236 where @xmath237 then the matrix @xmath96 given by @xmath238 for @xmath239 is the _ principle submatrix _ of @xmath240 indexed by @xmath241 . in what follows , we will use the convention that if @xmath236 is an indexing set of a matrix then @xmath237 . moreover , if @xmath242 the principle submatrix of @xmath234 indexed by @xmath241 then there exists a unique permutation matrix @xmath47 such that @xmath243 where @xmath96 is the principle submatrix of @xmath240 indexed by @xmath244 . assuming @xmath96 is invertable , then by ( [ eq1.2 ] ) @xmath245 if this is the case we call the matrix @xmath246 the reduction of @xmath240 over @xmath241 and write @xmath247 . in the case that @xmath248 we let the matrix @xmath249 . moreover , if @xmath250 can be reduced over the index set @xmath251 we write the reduction of @xmath250 over @xmath251 by @xmath252 , and so on . we formulate the following lemma related to the reduction of matrices . [ lemma1.1 ] let @xmath234 and @xmath253 be nonempty . if the principle submatrix of @xmath240 indexed by @xmath254 is upper triangular with nonzero diagonal and the sets @xmath255 then @xmath256 . without loss in generality suppose @xmath257 such that the matrix @xmath240 has block form @xmath258 where @xmath96 is an upper triangular principle submatrix of @xmath240 indexed by @xmath259 with nonzero diagonal . then for @xmath260 there exists a unique permutation matrix @xmath47 such that @xmath261 where @xmath262 , @xmath263 , and @xmath242 are the principle submatrices indexed by @xmath264 , and @xmath265 respectively . moreover , both @xmath262 and @xmath263 are upper triangular matrices with nonzero diagonal as they are principle submatrices of @xmath96 and are therefore invertable . it then follows that @xmath266 the claim is then that the matrix @xmath267 is an upper triangular matrix possessing a zero diagonal . to see this let the sets @xmath268 , @xmath269 , and @xmath270 . as the inverse of an upper triangular matrix is upper triangular @xmath271 for @xmath272 . hence , @xmath273 as @xmath274 then each @xmath275 . for each @xmath276 note that similarly @xmath277 . moreover , if @xmath278 then @xmath279 since @xmath96 is upper triangular . if @xmath280 then @xmath281 implying @xmath282 for the same reason . then , as @xmath283 , it follows that @xmath284 verifying the claim . @xmath285 is therefore an upper triangular matrix with nonzero diagonal . hence , @xmath285 is invertible and @xmath286 can be reduced over @xmath265 . in particular , @xmath287 where @xmath288 . to verify that @xmath289 note that @xmath240 has block form @xmath290\ ] ] where @xmath291 is the principle submatrix of @xmath47 indexed by @xmath259 . therefore , @xmath292^{-1}q = q^{-1}\left [ \begin{array}{cc } ( a_0-b_0a_1^{-1}c_0)^{-1 } & -a_0^{-1}b_0\gamma^{-1}\\ -\gamma^{-1}c_0a_0^{-1 } & \gamma^{-1}\\ \end{array } \right]q.\ ] ] note the matrix @xmath293 by the woodbury matrix identity @xcite and is therefore well defined . from this @xmath294^{-1 } qb\\ = & d-\left [ \begin{array}{cc } c_2 & c_1 \end{array } \right ] \left [ \begin{array}{cc } ( a_0-b_0a_1^{-1}c_0)^{-1 } & -a_0^{-1}b_0\gamma^{-1}\\ -\gamma^{-1}c_0a_0^{-1 } & \gamma^{-1}\\ \end{array } \right ] \left [ \begin{array}{c } b_2\\ b_1 \end{array } \right]\\ = & d - c_2(a_0^{-1}-a_0^{-1}b_0\gamma c_0a_0^{-1})b_2-c_1\gamma c_0a_0^{-1}b_2-\\ & \hspace{1.85 in } c_2a_0^{-1}b_0\gamma c_0a_0^{-1}b_1+c_1\gamma b_1.\end{aligned}\ ] ] from ( [ eq2.1 ] ) it follows that @xmath295 for @xmath296 the same argument can be repeated to show @xmath297 as @xmath286 has the same form as @xmath240 , i.e. its upper left hand block is an upper triangular matrix with nonzero diagonal . hence , @xmath298 . the lemma follows by further extending @xmath299 to the nested sequence @xmath255 by repeated use of the above argument . to establish that graph reductions are analogous to matrix reductions we prove the following lemma . [ lemma1.2 ] if @xmath300 is a structural set of @xmath6 then @xmath101 is the graph with adjacency matrix @xmath301 . without loss in generality let @xmath302 . from the discussion in section 3.1 , if @xmath55 is a structural set of @xmath7 then @xmath94 has the block form @xmath303 where @xmath96 is the principle submatrix of @xmath94 indexed by @xmath304 and is an upper triangular matrix with nonzero diagonal . with this in mind , let @xmath305 and @xmath306 for @xmath307 . if @xmath308 for @xmath88 then , proceeding by induction , for @xmath309 @xmath310^t\frac{1}{\omega_{11}-\lambda}[\omega_{12},\dots,\omega_{1n}]\big)\ ] ] where @xmath311 is the principle submatrix of @xmath94 indexed by @xmath312 . hence , @xmath313 for @xmath314 . conversely , note that the set @xmath315 consists of at most the two branches @xmath316 and @xmath317 i.e. the branches from @xmath21 to @xmath22 with and without interior vertex @xmath105 . as @xmath318 and @xmath319 if the branches @xmath320 and @xmath321 are respectively in @xmath315 it then follows from ( [ eq2.5 ] ) and ( [ eq1.0 ] ) that @xmath322 . suppose then for some @xmath323 that @xmath324 . since the principle submatrix of @xmath94 indexed by @xmath325 is upper triangular with nonzero diagonal and @xmath326 an application of lemma [ lemma1.1 ] implies that @xmath327 . hence , @xmath328 letting @xmath329 then by the argument above @xmath330 for @xmath331 . as @xmath332 then the entries of @xmath333 are given by @xmath334 . observe that @xmath335 where @xmath336 and @xmath337 are the branches in @xmath338 that contain and do not contain the interior vertex @xmath339 respectively . it then immediately follows that @xmath340 implying @xmath341 . on the other hand , any @xmath342 can be written as @xmath343 where @xmath344 and @xmath345 . hence , equation ( [ eq0.9 ] ) then implies @xmath346 conversely , if @xmath344 and @xmath345 then @xmath347 . this follows from the fact that @xmath348 and @xmath349 share no interior vertices since otherwise @xmath350 would contain a cycle , which is not possible since @xmath351 . therefore , @xmath352 moreover , @xmath353 since @xmath354 contains at most the cycle @xmath355 given that @xmath356 . from ( [ eq2.6 ] ) and ( [ eq2.7 ] ) it therefore follows that @xmath357 implying @xmath358 . by induction it follows that @xmath359 by setting @xmath360 . if @xmath361 is a structural set of @xmath2 then let @xmath362 be the index set associated with @xmath55 . hence , if @xmath46 is indexed by @xmath241 then @xmath75 is indexed by @xmath254 and there is a unique permutation matrix @xmath47 such that @xmath363 where @xmath96 is the principle submatrix of @xmath94 indexed by @xmath254 . therefore , @xmath364 and it follows from lemma [ lemma1.2 ] and ( [ eq1.2 ] ) that @xmath365 letting @xmath366 then theorem [ maintheorem ] follows by observing that @xmath74 and @xmath76 are the solutions to @xmath367 and @xmath33 respectively . we now give proofs of the theorems on isospectral transformations . for a proof of theorem [ theorem3 ] we have the following . suppose @xmath116 where @xmath121 is a structural set of @xmath7 and let @xmath368 be the index set associated with @xmath369 for @xmath370 . by lemma [ lemma1.2 ] the graph @xmath371 has adjacency matrix @xmath372 . hence , and another application of lemma [ lemma1.2 ] , the graph @xmath373 has adjacency matrix @xmath374 continuing in this manner , it follows that @xmath375 has adjacency matrix given by @xmath376 which by lemma [ lemma1.1 ] is equivalent to the matrix @xmath377 . given that @xmath377 is the adjacency matrix of @xmath378 then again by use of lemma [ lemma1.2 ] it follows that @xmath379 . we now give a proof of lemma [ lemma1 ] . let @xmath380 and @xmath381 be in @xmath4 such that @xmath382 . as @xmath383 then the result follows from equation ( [ eq0.9 ] ) and ( [ eq1.0 ] ) . in order to prove theorem [ theorem0.1 ] we require the following lemma that essentially implies that the order in which any two vertices are removed from a graph @xmath135 does not effect the final reduced graph . [ lemma-2 ] suppose @xmath135 with vertex set @xmath20 for @xmath384 . if @xmath385 then @xmath386 . let @xmath135 and @xmath308 for @xmath88 . then @xmath94 can be written in the block form @xmath387\ ] ] is the principle submatrix of @xmath94 indexed by @xmath388 . note the assumption @xmath135 implies that both @xmath389 , @xmath390 . moreover , if @xmath391 then @xmath392 however , equation ( [ eq2.5 ] ) implies in this case that @xmath393 hence , @xmath394 , which is not possible as it contradicts the conclusion of lemma [ lemma1 ] . therefore , @xmath395 implying @xmath396 is well defined . letting @xmath397 for @xmath398 and @xmath399 then repeated use of ( [ eq1.2 ] ) implies @xmath400 . moreover , as @xmath401 then it follows that the graphs @xmath386 since they have the same adjacency matrices . to simplify the proof of theorem [ theorem0.1 ] we note the following . if @xmath120 induces a sequence of reductions on @xmath7 then @xmath375 can alternately be written as @xmath402 for @xmath403 . this notation is meant to indicate that at the @xmath404th reduction we remove the vertices @xmath405 from the graph @xmath406 for each @xmath407 where @xmath408 . with this notation in place we give a proof of theorem [ theorem0.1 ] . if @xmath409 and @xmath410 then by the discussion in section 3.2 the graph @xmath411 is defined . moreover , for any @xmath412 , let @xmath413 where @xmath414 . then by lemma [ lemma-2 ] it follows that @xmath415 which in turn implies that @xmath416 by repeatedly switching the order of any two vertices as above it follows that for any bijection @xmath417 @xmath418 suppose @xmath419 induces a sequence of reductions on @xmath2 . if @xmath420 and each @xmath421 then theorem [ theorem3 ] implies @xmath422 if @xmath423 also induces sequence of reductions on @xmath2 where @xmath424 and each @xmath425 then similarly , @xmath426 if @xmath427 then @xmath428 . there is then a bijection @xmath429 such that @xmath430 implying @xmath431 by use of equation ( [ eq2.8 ] ) . this completes the proof . the notation @xmath432=\mathcal{r}(g;s_1,\dots , s_m)$ ] is then well defined for any graph @xmath7 in @xmath131 and nonempty @xmath433 . moreover , if for any @xmath434 , @xmath157 is nonempty then the relation @xmath159 if @xmath160\simeq\mathcal{r}_{\tau(h)}[h]$ ] is an equivalence relation on @xmath131 . to see this , note that as the rule @xmath156 specifies a unique nonempty graph @xmath160 $ ] for any @xmath434 , then @xmath158 is an equivalence relation given that @xmath435 is reflexive , symmetric , and transitive hence , theorem [ theorem0.1 ] holds . we now give a proof of theorem [ theorem-2 ] . let @xmath7 and @xmath46 such that each vertex of @xmath2 is on a branch of @xmath57 . it follows then that each @xmath436 is part of the ordered sequence @xmath180 for some @xmath437 . given the bijection @xmath184 in definition [ def10 ] if @xmath182 then @xmath438 . hence , @xmath439 . similarly , as each vertex of @xmath189 is on a branch of @xmath188 then @xmath440 implying the weight sets of @xmath2 and @xmath189 are equal . moreover , the assumption that each vertex of @xmath2 and @xmath189 lies on a branch of @xmath441 and @xmath442 respectively implies together with ( i ) in definition [ def10 ] that @xmath443 . similarly , @xmath191 follows from the fact that branches of equal length in @xmath441 and @xmath442 are in bijective correspondence and that the branches in @xmath442 are independent . to compare the spectrum of @xmath2 and its expansion @xmath189 note equation ( [ eq3.3 ] ) and the fact that @xmath444 together imply @xmath445 by ( [ eq1.1 ] ) it then follows that @xmath446 @xmath447 where @xmath35 is the number of distinct branches in @xmath57 containing @xmath21 . this completes the proof . for theorem [ theorem10 ] we give the following proof . let @xmath7 and @xmath448 where each vertex of @xmath2 is on a branch of @xmath57 . by construction , it follows that there exists bijections @xmath449 for all @xmath88 such that if @xmath450 then @xmath451 and @xmath452 . hence , if @xmath453 then @xmath454 implying @xmath455 therefore , @xmath456 or @xmath2 and @xmath228 are spectrally equivalent . given that @xmath208 and @xmath457 then @xmath74 and @xmath458 are lists of zeros and both @xmath76 and @xmath459 are the empty set . by theorem [ maintheorem ] it then follows that @xmath28 and @xmath460 differ by at most a list of zeros . moreover , since in the construction of @xmath228 , @xmath461 then we have that @xmath462 as each @xmath215 has @xmath463 interior vertices and is independent of the branch @xmath464 for each @xmath398 and @xmath213 . as @xmath465 then @xmath228 is a reduction of @xmath2 over @xmath203 if @xmath466 and an expansion otherwise for any unital subring @xmath202 containing @xmath205 . in this section we consider discrete - time dynamical networks @xcite and use techniques from the previous sections to analyze their dynamics . in particular , we develop the notion of a _ dynamical network expansion _ and show that this notion allows for improved estimates on whether the given ( unexpanded ) network has a unique global attractor . dynamical networks or networks of interacting dynamical systems are composed of ( i ) local dynamical systems which have their own ( local intrinsic ) dynamics , ( ii ) interactions between these ( elements of the network ) local systems , and ( iii ) the graph of interactions ( topology of the network ) . let @xmath467 and @xmath468 be maps on the compact metric space @xmath469 where @xmath470 let @xmath471 denote the direct product of the local systems @xmath472 over @xmath241 on the compact metric space @xmath473 where @xmath474 is the @xmath475 metric . [ interaction ] a map @xmath476 is called an _ interaction _ if for every @xmath477 there exists a nonempty collection of indices @xmath478 and a continuous function @xmath479 that satisfies the following lipschitz condition for constants @xmath480 @xmath481 for all @xmath482 where @xmath483 is the restriction of @xmath484 to @xmath485 . then the ( interaction ) map @xmath486 is defined as follows : @xmath487 the constants @xmath488 in definition [ interaction ] form the matrix @xmath489 where the entry @xmath490 if @xmath491 . as in @xcite we assume that interactions are lipschitz . that is , each @xmath488 can be thought of as a lipschitz constant of the function @xmath486 measuring the maximal expansiveness of the function @xmath492 in the @xmath404th coordinate . in particular , if the interaction @xmath486 is continuously differentiable on @xmath170 then the constants @xmath493 satisfy condition ( [ eq2.3 ] ) where @xmath494 is the matrix of first partial derivatives of @xmath486 . [ netdef ] the superposition @xmath495 generates the dynamical system @xmath496 which is a _ dynamical network_. using notation similar to graphs , let @xmath497 $ ] i.e. @xmath498 moreover , for any @xmath499 let @xmath500 denote the _ spectral radius _ of the matrix @xmath96 . [ stability ] if @xmath501 then the dynamical network @xmath496 has a globally attracting fixed point . for @xmath502 and @xmath503 @xmath504 it then follows that enterywise @xmath505\leq m_n\left[\begin{array}{c } d\big(x_1,y_1\big)\\ \vdots\\ d\big(x_n , y_n\big ) \end{array}\right].\ ] ] @xmath506\leq m_n\left[\begin{array}{c } d\big(n(x)_1,n(y)_1\big)\\ \vdots\\ d\big(n(x)_n , n(y)_n\big ) \end{array}\right]\leq m^2_n\left[\begin{array}{c } d\big(x_1,y_1\big)\\ \vdots\\ d\big(x_n , y_n\big ) \end{array}\right]\ ] ] then by induction @xmath507^t||_{\infty}\ ] ] for all @xmath508 . note that @xmath509 if and only if @xmath510 ( see @xcite chapter 8) . then for any @xmath502 , @xmath511 under the assumption that @xmath510 . hence , @xmath512 as @xmath513 implying the sequence @xmath514 is cauchy and therefore convergent . if @xmath515 then it follows that @xmath516 for all @xmath517 completing the proof . for an interaction @xmath518 we say that @xmath486 _ stabilizes _ the local systems @xmath471 if the local systems are unstable i.e. @xmath519 but the dynamical network @xmath495 has a globally attracting fixed point . if @xmath520 and @xmath495 has a globally attracting fixed point we say @xmath486 maintains the _ stability _ of @xmath471 . the following is then a corollary to theorem [ stability ] . [ col2 ] for the dynamical network @xmath495 let @xmath521 . if @xmath522 then the interaction @xmath486 stabilizes ( or maintains stability ) the local systems @xmath471 . before proving this we need the following . for @xmath523 we write @xmath524 we note that if @xmath525 then @xmath526 ( see chapter 8 @xcite ) . we now give a proof of corollary [ col2 ] . as @xmath527\leq l\lambda^t$ ] then @xmath528 . moreover , given that @xmath529 then @xmath530 . this in turn implies @xmath531 . hence , if @xmath522 then by theorem [ stability ] the interaction @xmath486 stabilizes the local systems @xmath471 . the constant @xmath532 related to the interaction @xmath486 can then be thought of as ( an estimate of ) the interaction s _ stability factor _ or its ability to stabilize a set of local systems @xmath471 . in the following section we will use techniques analogous to branch expansions ( see section 4.1 ) to obtain improved estimates of an interaction s stability factor . the goal of this section is to develop a method of transforming a general dynamical network to one with a different graph of interactions but dynamically equivalent to the original system in the sense described below . the transformed network can then be investigated in order to infer dynamical properties of the original untransformed network . for @xmath495 the dynamical network @xmath533 can be considered to be the interaction @xmath534 , i.e. the composition of the local systems and their interaction . for simplicity then , in this section we consider only the interaction @xmath486 or equivalently the dynamical network @xmath486 with the understanding that all results extend to general dynamical networks of the form @xmath495 . to every interaction @xmath518 and some choice of constants @xmath535 satisfying ( [ eq2.3 ] ) there is a corresponding _ graph of interactions _ @xmath536 which is the graph with adjacency matrix @xmath537 . for @xmath538 we let @xmath539 be the index set of @xmath55 and let @xmath540 moreover , for @xmath484 let @xmath541 where @xmath542 . [ def20 ] for @xmath543 let @xmath544 be the function @xmath545 where each @xmath546 is replaced by @xmath547 , each @xmath548 by @xmath549 , and so on for all indices @xmath550 . we call the function @xmath551 the _ restriction of the interaction _ @xmath486 to @xmath55 . as @xmath552 contains no cycles this recursion can not continue indefinitely , hence @xmath544 is well defined . furthermore , the restriction of the interaction @xmath486 to @xmath55 is given by @xmath553 where @xmath554 as an example consider the interaction @xmath113 given by @xmath555.\ ] ] here , @xmath113 has the graph of interactions @xmath556 given in figure 5 ( left ) where @xmath557 if @xmath558 is a function of the variable @xmath559 and each @xmath557 has weight @xmath488 . as @xmath174 is a structural set in @xmath560 then for @xmath484 @xmath561.\ ] ] importantly , the function @xmath562 is not an interaction as its range @xmath563 is not a subset of its domain @xmath564 . however , the functions @xmath113 and @xmath565 have the same dynamics restricted to @xmath55 in the following sense . [ lemma10 ] for the interaction @xmath486 let @xmath538 . then for any @xmath484 @xmath566 for @xmath567 . let @xmath484 and @xmath568 . note that by construction @xmath569 . hence , if each @xmath570 nested in @xmath544 is recursively replaced by @xmath571 for all @xmath550 then the result is the function @xmath572 where @xmath573 as @xmath574 then @xmath575 . as @xmath484 was arbitrary , replacing it by @xmath576 for fixed @xmath567 implies @xmath566 completing the proof . nonlin5.eps ( 0,2)@xmath105 ( 44,2)@xmath106 ( 15,19)@xmath108 ( 29,19)@xmath107 ( 20,-2)@xmath577 ( 21,3)@xmath578 ( 21,8.5)@xmath579 ( 4,13)@xmath580 ( 21,18.5)@xmath581 ( 38,13)@xmath582 ( 14,13)@xmath583 ( 75,14)@xmath584 ( 54,2)@xmath105 ( 98,2)@xmath106 ( 68,19)@xmath585 ( 83,19)@xmath586 ( 73.5,-2)@xmath587 ( 75,3)@xmath588 ( 75,8.5)@xmath589 ( 61,13)@xmath164 ( 75.5,18.5)@xmath164 ( 92,13)@xmath590 ( 68,10.5)@xmath164 ( 83.5,10.5)@xmath591 we now proceed to define the expansion of the interaction @xmath486 with respect to @xmath55 in terms of its restriction @xmath592 . to do so we first index the interior vertices of @xmath593 by @xmath594 and let the map @xmath595 be a bijection . by abuse of notation we set @xmath596 if @xmath597 . for @xmath568 define @xmath598 to be the variables that appear in the function @xmath544 including multiplicities . then each @xmath599 is nested in some sequence of functions @xmath600 inside of @xmath544 corresponding to the branch @xmath601 hence @xmath602 . modify the function @xmath544 by changing this @xmath603 . if this is done over all variables in @xmath598 we call the resulting function @xmath604 and set @xmath605 @xmath606 where each @xmath607 for @xmath608 . here @xmath609 if @xmath610 . for the interaction @xmath486 suppose @xmath538 . we call the function @xmath611 the _ expansion _ of @xmath486 over @xmath55 . an expansion is then the interaction @xmath612 where @xmath613 we describe this construction for the interaction @xmath113 in the previous example depicted in figure 5 . for the interaction @xmath113 , @xmath614 ; @xmath615 ; @xmath616 ; and @xmath617 constitute the set @xmath593 hence , @xmath618 letting @xmath619 then @xmath620,\ ] ] where @xmath621 , @xmath622 , and @xmath623 . the expansion of @xmath113 over @xmath55 is then given by @xmath624.\ ] ] if the constants @xmath625 satisfy condition ( [ eq2.3 ] ) for @xmath626 then @xmath587 is the graph in figure 5 ( right ) where @xmath627 , @xmath628 , @xmath629 , as each of @xmath630 . moreover , we note that as unweighted graphs @xmath587 is the branch expansion of @xmath577 with respect to @xmath55 . [ 2ndmajor ] for the interaction @xmath486 suppose @xmath538 . then the following hold : + ( 1 ) for @xmath484 let @xmath631 where @xmath632 . if @xmath633 then the restriction @xmath634 for @xmath567 . + ( 2 ) if @xmath486 satisfies ( [ eq2.3 ] ) for the constants @xmath488 then there are constants @xmath625 for @xmath635 satisfying ( [ eq2.3 ] ) such that @xmath636 . we prove parts ( 1 ) and ( 2 ) of theorem [ 2ndmajor ] separately . for part ( 1 ) we give the following . let @xmath568 and @xmath567 . then @xmath637 by changing the variable @xmath603 for all @xmath638 . however , note that @xmath632 . hence , evaluating @xmath639 at @xmath640 is equivalent to changing each @xmath641 for each variable @xmath642 of the function @xmath639 . as this sequence of operations takes @xmath643 then , for @xmath644 , it follows that @xmath645 lemma [ lemma10 ] then implies that @xmath646 for @xmath567 . before giving a proof of part ( 2 ) we introduce the following notation . for the branch @xmath647 let @xmath648 moreover , for @xmath568 define the sets @xmath649 , @xmath650 , and @xmath651 lastly , define @xmath652 to be the function in the @xmath653 argument of @xmath604 . similarly , define @xmath654 to be the function in the @xmath655 argument of @xmath652 , and so on . we now give a proof of theorem [ 2ndmajor ] part ( 2 ) . for @xmath656 and @xmath597 @xmath657 as each @xmath658 . as @xmath659 for some choice of @xmath660 then the constants @xmath661 satisfy condition ( [ eq2.3 ] ) for @xmath635 for each @xmath662 . before considering the case where @xmath568 we first observe the following . let @xmath663 . as @xmath664 then for @xmath665 @xmath666 for @xmath667 it follows that @xmath668 as the sum @xmath669 then continuing inductively @xmath670 as @xmath671 then for @xmath568 the constants @xmath672 satisfy condition ( [ eq2.3 ] ) for @xmath635 . therefore , ( [ eqproof1 ] ) and ( [ eqproof2 ] ) give a complete set of the constants @xmath625 for @xmath635 . let @xmath673 . by construction there are one - to - one correspondences @xmath674 given as follows . suppose @xmath675 . if @xmath610 then @xmath676 . hence , @xmath677 implying @xmath678 by ( [ eqproof2 ] ) . thus , @xmath679 . if @xmath680 then @xmath681 is the branch corresponding to @xmath682 . equations ( [ eqproof1 ] ) and ( [ eqproof2 ] ) then imply @xmath683 as @xmath684 then @xmath679 . therefore , @xmath685 . as @xmath538 and @xmath686 by construction then theorem [ maintheorem ] implies that @xmath536 and @xmath687 have the same nonzero spectrum . hence , @xmath688 for this choice of constants . this completes the proof . part ( 1 ) of theorem [ 2ndmajor ] gives a correspondence between the dynamics of @xmath486 and its expansion @xmath635 . this can be used to infer dynamical properties of the untransformed network via its expansions as follows . [ 3rdmajor ] if @xmath689 where @xmath625 are constants satisfying ( [ eq2.3 ] ) for @xmath635 then the interaction @xmath486 has a globally attracting fixed point . assuming that @xmath689 then theorem [ stability ] implies the expansion @xmath635 has a globally attracting fixed point @xmath690 . for @xmath484 define @xmath691 where @xmath632 as in part ( 1 ) of theorem [ 2ndmajor ] . hence , @xmath692 as @xmath693 then similarly @xmath694 as @xmath695 . from definition [ def20 ] it follows for each @xmath543 that the function @xmath696 for any @xmath697 . here @xmath698 is the maximal length of any path or cycle from @xmath21 to @xmath22 with interior vertices in @xmath75 where @xmath21 is taken over all @xmath699 . by an argument similar to that in lemma [ lemma10 ] we have @xmath700 from the observation that @xmath701 as @xmath695 for all @xmath702 then this together with ( [ eq22 ] ) implies @xmath703 for all @xmath543 as @xmath544 is continuous over its domain . hence , @xmath486 has a globally attracting fixed point . simply put , if a network expansion @xmath635 has a globally attracting fixed point then the same holds for @xmath486 . this is based primarily on part ( 1 ) of theorem [ 2ndmajor ] . however , part ( 2 ) of theorem [ 2ndmajor ] implies that determining whether or not @xmath486 has this property is best accomplished by applying theorem [ stability ] to the expansion @xmath635 instead of @xmath486 itself . importantly , this method of using network expansions generalizes the method used in @xcite for determining whether a network has a unique global attractor . let @xmath704 be the standard logistic map and let the function @xmath705 where both are restricted to the interval @xmath706 $ ] . consider the previously introduced interaction @xmath113 ( see figure 5 ) with the specific form @xmath707.\ ] ] note the constants @xmath708 satisfy equation ( [ eq2.3 ] ) for @xmath113 and are given by @xmath709.\ ] ] as @xmath710 theorem [ stability ] does not imply that @xmath113 has a globally attracting fixed point . however , the expansion over @xmath174 of @xmath113 given by @xmath711.\ ] ] where @xmath712 is such that @xmath713.\ ] ] here , as before , @xmath714 . as @xmath715 theorem [ 3rdmajor ] implies that @xmath113 does infact have a globally attracting fixed point . as a final remark in this section , we note that it is possible to sequentially expand an interaction ( equivalently a dynamical network ) and thereby potentially improve ones estimate of whether the interaction has a globally attracting fixed point . moreover , sequential expansions allow for improved estimates of an interactions stability factor . for instance , if the interaction @xmath626 in example 1 were expanded over the vertex set @xmath716 then @xmath717 for the constants @xmath718 . that is , from this calculation the interaction @xmath113 has a stability factor less than or equal to @xmath719 . the goal of this paper is twofold ; introduce isospectral graph transformations and use these transformations to investigate the dynamic properties of dynamical networks with arbitrary topologies ( i.e. graph of interactions ) . for the first , we described the general process of isospectral graph reductions in which a graph is collapsed around a specific set of vertices as well as the inverse process of isospectral expansion . we then considered sequences of isospectral reductions and showed that under mild assumptions a typical graph could be uniquely reduced to a graph on any subset of its vertex set . as is shown in @xcite , such reductions and sequences of reductions can be used to improve all existing eigenvalue estimates @xcite for the general class of matrices with complex entries . the main result of @xcite is that such eigenvalue estimates improve as a graph is reduced using the method of reduction given in this paper . these estimates of gershgorin et al . can be traced back to work done by lvy , desplanques , minkowski , and hadamard @xcite on diagonally dominant matrices . in this application of isospectral transformations , the flexibility and commutativity of such reductions is particularly useful . moreover , isospectral reductions are also relevant to the theory of networks as they provide a flexible means of course graining networks ( i.e. viewing the network at some appropriate scale ) while simultaneously introducing new equivalence relations on the space of all networks . the second type of transformations we presented in this paper were isospectral graph transformations over fixed weight sets . such transformations allow one to transform an arbitrary graph while preserving the graph s set of edge weights and spectrum along with the its spectrum . motivated by this procedure , we introduced dynamical network expansions . dynamical network expansions modify a dynamical network in a way that preserves network dynamics but alters its associated graph structure ( graph of interactions ) . in this paper we demonstrated that this procedure allows one to establish global stability of a more general class of dynamical networks than considered in @xcite . that is , such expansions provide more information and can be used to obtain stronger results regarding network dynamics than the original ( unexpanded ) network . lastly , the results of the present paper introduce various approaches to simplifying a graph s structure while maintaining its spectrum . therefore , these techniques can be used for optimal design , in the sense of structure simplicity of dynamical networks with prescribed dynamical properties ranging from synchronizability to chaoticity @xcite . we would like to thank c. kirst and m. timme for valuable comments and discussions . this work was partially supported by the nsf grant dms-0900945 and the humboldt foundation .

we present an analysis and interpretation of oscillation spectra for all 69 sx phoenicis stars discovered in the field of the cluster . for most of the stars we have reliable absolute magnitudes and colors . except of one , or perhaps two , objects , the stars are cluster members . their pulsational behaviour is very diversified . multiperiodic variability with at least part of the excited modes being nonradial is most common but there are also many cases of high amplitude , presumably radial mode , pulsators . in a number of such cases we have evidence for two radial modes being excited . parameters of radial mode pulsators are in most cases consistent with standard evolutionary models for stars in the mass range . however , in four cases we have evidence that the masses are significantly lower than expected . three objects show frequency triplets that may be interpreted in terms of rotational frequency splitting of modes . implied equatorial velocities of rotation are from 10 to over 100 km / s . nearly all measured frequencies fall in the ranges predicted for unstable modes . two cases of low frequency variability are interpreted as being caused by tidal distortion induced by close companions . [ firstpage ] stars : sx phe - stars : variables globular clusters : individual : cen

astro-ph0412523

sx phoenicis stars ( sxps ) are the population ii short period pulsators , in most respects similar to much more numerous @xmath3 scuti stars , which are the population i objects . both types occupy the low luminosity end of the cepheid instability strip in the h - r diagram . it is not clear whether there are any systematic differences in pulsation properties between the two types . the incidence of high amplitude pulsation once seemed much higher among sxps but now , with improved cluster photometry , many low amplitude sxps are being detected ( pych et al . 2001 , mazur et al . 2003 , kaluzny & thompson 2003 , kaluzny et al . 2004 ) . what makes sxps interesting in a wider context is that they are blue straggler stars ( bss ) found in a large number in globular and old open clusters . this means that they have unusual life history which is not well understood as it can not be explained in terms of the standard single star evolution scenario . the problem of bss origin is being debated for decades . historically , the first explanation was that the objects are products of evolution of the mass receiving component in close binary systems ( mccrea 1964 ) . later scenarios involving mergers of main sequence stars were mostly considered . the scenarios included mergers of primordial binaries after a gradual decrease of the orbit ( e.g. carney et al . 2001 ) as well as direct stellar collisions in dense cluster cores ( e.g. lombardi et al . it is not clear how much imprints of the past evolution should be left in individual bss . this depends on the efficiency of chemical element and angular momentum mixing - the processes still poorly understood in stellar interior physics . understanding the origin and internal structure of bss is a challenging task for stellar evolution theory . the answer is of interest not only for this field but also for globular cluster research . as emphasized by lombardi & rasio ( 2002 ) , collisions and mergers of stars which lead to bss formation , play an important role in evolution of these systems . pulsation data on sxps are potential sources of accurate constraints on models of bss . each measured frequency of an identified oscillation mode is such a constraint . admittedly , mode identification is difficult unless we have evidence that the excited modes are radial . double mode radial pulsators seem quite frequent among sxps . the two accurately measured numbers yield strong constraints on stellar mass and heavy element content in the interior . data on nonradial mode frequencies could be even more interesting . in particular , determination of the frequency splitting provides certain mean value of the rotation rate in the interior . furthermore , there are nonradial modes whose frequencies are very sensitive to the extent of the element mixing beyond the convective core boundary . of great interest in the context of the debated bss origin is seeking evidence for presence of close companions manifesting themselves through cyclic period variations or tidally induced variability . what makes sxps interesting for stellar pulsation theory is the large diversity of their pulsation forms . in a relatively narrow period range , extending from less than half to two hours , we encounter both the multimode low - amplitude pulsation , typical for main sequence dwarfs , and the monomode high - amplitude pulsation , typical for cepheids . according to linear nonadiabatic calculations , there are many unstable modes in sxps . however , finite amplitude development of the instability is not understood . this is the most outstanding unsolved problem of the theory . a sample of sxps with well constrained mean parameters may provide a key information what makes a star to become a dwarf- or a giant - type pulsator . the globular cluster @xmath2 centauri houses the largest number of sxps of all systems in our galaxy ( kaluzny et al . 2004 ) . having a large number of objects with well determined luminosities is an obvious advantage . the fact that the stars can not be assumed coeval and of the same chemical composition ( rey et al . 2000 ) is a complicating factor . the cluster is atypical . data on sxps may prove useful in disentangling its evolution . in the next section we survey observational data on all sxps in @xmath2 cen which includes frequency analysis . in section 3 we compare mean photometric parameters of sxps with the corresponding values calculated for standard evolutionary models . we also provide information on pulsation properties of these models based on the linear nonadiabatic analysis . section 4 is a star by star analysis of individual objects in which we compare their observational properties with theoretical models . in section 5 we summarize our results . @xmath2 centauri contains the most numerous population of variable stars among globular clusters of the galaxy ( clement et al . however until the mid 1990ies there was only one sx phe - type variable known in the field of the cluster . it was v65 , which in fact turned out later to be a foreground star . however , not much later kaluzny et al . ( 1996 , 1997 ) reported a discovery of 25 sxps , most of which must be members of the cluster . subsequent work done by kaluzny et al . ( 2004 ) increased the number of sxps in the field of @xmath2 centauri to 69 , making it the richest in these variables among all globular clusters of our galaxy . kaluzny et al . ( 2004 ) provide extensive photometry for 61 sxps collected during two years . this photometry contains from 532 to 755 @xmath4 and over 150 @xmath5 measurements for each variable . it is an excellent data base for analyzing multi - mode behavior and extracting information on physical properties of individual objects . .basic properties of sx phe - type variables in field of @xmath2 cen.[t1 ] [ cols="<,<,^,^,^,^,^",options="header " , ] in addition , @xmath2 cen is the first globular cluster for which the case project ( thompson et al . 2001 ) determined the distance . the analysis of photometric and spectroscopic data for an eclipsing binary ogle gc17 yielded an apparent distance modulus of @xmath6 mag ( kaluzny et al . it is the most precise and reliable distance determination for this cluster and will be used in our work for calculating the absolute magnitudes of sxps belonging to @xmath2 cen . we will also use the mean color data dereddened with the color excess @xmath7 mag ( schlegel , finkbeiner & davis , 1998 ) to place individual objects in the h - r diagram . the full list of sxps located in the field of @xmath2 centauri containing main periods , amplitudes , mean magnitudes and colors is provided in table 1 . 1 shows color - magnitude diagram around blue stragglers region of @xmath2 cen . points and open circles denote constant stars and sxps , respectively . not all stars in the sxps region are variables . this is hardly surprising , as we have the same situation in the @xmath3 scuti domain . presumably these apparently constant objects are very low amplitude pulsators . many sx phe stars exhibit multiple periods hence their analysis poses specific challenges . arguably the most advanced project in terms of both extent of its photometry and multitude of detected periods , the whole earth telescope ( wet ) , employed either the prewhitening _ or _ synthesis methods in their analysis ( kepler , 1993 ) . from the statistical point of view the two methods correspond to gauss - seidel ( g - s ) and newton - raphson ( n - r ) solutions of the least squares problem ( lsq ) . on one hand , a current astronomical practice favors the synthesis ( n - r ) method , employing the covariance matrix with large extra - diagonal coefficients . presence of close frequencies and/or their combinations in the synthesis method yields near singular normal equations and may produce large near cancelling terms in the solution , creating the @xmath8 problem and yielding solutions with excess amplitudes . on the other hand , the method recommended to deal with singularity in least squares fitting ( lsq ) is by means of the singular value decomposition ( svd , e.g. press et al . , 1986 ) . in the svd procedure the near singular terms of the covariance matrix are set small , resulting in choice of the solution with minimum norm ( minimum amplitudes in the present context ) . this resembles somewhat the prewhitening method , where implicitly all extra - diagonal terms are set zero ( gray & desikachary , 1973 ) . we are unaware of any paper comparing merits of the prewhitening and synthesis methods in rigid mathematical ( statistical ) terms . we employ the prewhitening and synthesis methods in parallel to analyze our sx phe light curves . we employed no more than 6 harmonics in prewhitening and in most cases just 2 or 3 harmonics for the @xmath9 mode and one harmonic for other modes . the values of ( fundamental ) frequencies were adjusted by nonlinear iterations , their harmonics / combinations were tied to the base frequencies . the resulting frequencies are listed in table 2 . the both methods yielded consistent frequencies for all strong detections , corresponding to the aov statistics @xmath10 . we observed no adverse effect of reappearance of the frequencies once prewhitened . ambiguity , if any , arose occasionally due to aliasing and it manifested with equal strength in both the prewhitening and synthesis methods . such cases are marked in table 2 with superscript @xmath11 . the novelty in our approach , as compared with wet , is in use of the multi - harmonic aov periodogram ( schwarzenberg - czerny , 1996 ) . it employs orthogonal functions and is able to combine power from harmonics ( nh=2 harmonics were combined in practice ) . importance of use of orthogonal functions in period search was argued by lomb ( 1976 ) , ferraz - mello ( 1981 ) , scargle ( 1982 ) and foster ( 1994 ) . advantage in combining harmonics is two - fold : they contribute extra power for the base frequency and at the same time reduce power in the residuals . any method sensitive to the harmonics ( e.g. phase folding and binning and the present case of fourier fit ) is prone to produce sub - harmonics in the periodogram . the subharmonic pose little problem in practice as they are easily identified by tight packing of aliases . there exist exact relations between power in the data , root - mean square ( rms ) of the residuals and the aov statistics @xmath12 employed in the present case ( e.g. schwarzenberg - czerny , 1999 ) . detection of frequencies poses special case of hypothesis testing in statistics . in our case the detection criterion @xmath13 roughly corresponded to mode amplitude exceeding 4 times its standard deviation @xmath14 . any frequencies near / below that detection limit are deemed unsafe and indicated with superscript @xmath15 in table 2 . the purpose of listing these marginal detections ( @xmath16 ) is solely to indicate deviations of residuals from pure noise and not to ponder on frequencies . an example of our results for nv324 is shown in fig . the panels show anova periodograms ( schwarzenberg - czerny 1996 ) . the uppermost panel is computed for the original data and reveals the dominant peak with its aliases . the remaining ones are computed for the data after subtraction of all so far identified frequencies and significant combinations of thereof . 3 shows the light curves of nv324 associated with all six periods found . each light curve , phased with the particular period , was prewhitened with five remaining periods and their harmonics . note the asymmetry of the three upper light curves which is typical for pulsating stars . the window function is a concept associated with the power spectrum . strictly speaking it does not apply for the anova periodogram . however , in fig . 2 numerous window - like structures are present around principal frequencies . we refrain here from discussion of amplitudes . on one hand theory results on amplitudes are unreliable ( sect . [ s4 ] ) . on the other hand lsq fits of multifrequency light curves often suffer from large correlation of parameters , indicating large error ellipses despite small variances . we address this problem only indirectly here by relying our analysis on a large sample of stars with very extensive coverage of observations . application of two independent reduction methods benefitted us in that we checked against any computation errors as well as gained insight into repeatability and reliability of our results . lccccccc star & @xmath9 & @xmath17 & @xmath18 & @xmath19 & @xmath20 & @xmath21 & @xmath22 + v65 & 15.9430(2 ) & 20.2287(1 ) & 21.2182(8 ) & 24.5199(8 ) & 36.1720(8)@xmath11 & - & - + v194 & 21.1964(1 ) & 27.1633(5 ) & 43.7932(9 ) & 20.7471(6)@xmath11 & 22.1462(6)@xmath15 & - & - + v195 & 15.2692(1 ) & 15.2747(5 ) & 15.2668(6 ) & - & - & - & - + v197 & 21.2219(2 ) & 21.3904(5 ) & 21.3129(5 ) & 27.1838(8 ) & - & - & - + v198 & 20.7548(3 ) & - & - & - & - & - & - + v199 & 16.0548(2 ) & - & - & - & - & - & - + v200 & 20.1935(1 ) & 26.5308(5 ) & - & - & - & - & - + v204 & 20.2529(1 ) & 26.1552(5 ) & - & - & - & - & - + v217 & 18.7755(5 ) & - & - & - & - & - & - + v218 & 22.8627(5 ) & 27.0008(6)@xmath11 & 43.6865(9 ) & - & - & - & - + v219 & 25.8611(6 ) & 26.6204(8 ) & 28.2532(8 ) & 3.2836(3 ) & - & - & - + v220 & 18.9083(2 ) & 24.3161(5 ) & 23.8450(7 ) & - & - & - & - + v221 & 27.6749(4 ) & 26.9153(5 ) & 32.8848(5 ) & 27.6867(6 ) & 27.6630(9 ) & - & - + v225 & 20.5600(2 ) & 26.4145(5 ) & 26.8212(6 ) & - & - & - & - + v226 & 26.4185(5 ) & 27.4932(6 ) & 15.4167(5 ) & 25.4523(8 ) & - & - & - + v227 & 26.1605(5 ) & 25.4535(9 ) & 28.6334(8 ) & - & - & - & - + v228 & 25.0923(5 ) & 24.3732(7 ) & 26.7906(7)@xmath11 & 30.0974(9 ) & - & - & - + v229 & 26.6429(3 ) & 51.6073(6 ) & - & - & - & - & - + v231 & 26.6778(4 ) & 27.1269(6 ) & 27.3509(7 ) & 36.6960(6)@xmath15 & - & - & - + v233 & 27.3691(2 ) & 2.9205(5 ) & - & - & - & - & - + v237 & 15.2433(2 ) & 24.1162(5 ) & 15.2013(8 ) & - & - & - & - + v238 & 24.5096(4 ) & 24.1318(8 ) & 25.9960(7 ) & 24.2163(7)@xmath11 & 39.5205(7)@xmath15 & - & - + v249 & 28.6149(4 ) & 35.0000(9 ) & 29.6107(8 ) & 28.6212(8 ) & - & - & - + v250 & 24.6142(3 ) & 37.6484(8 ) & - & - & - & - & - + v252 & 21.4488(4 ) & - & - & - & - & - & - + v253 & 25.0196(5 ) & 25.7372(6 ) & 33.8173(6 ) & 31.1610(7)@xmath11 & - & - & - + nv294 & 56.3901(6 ) & 28.9864(6 ) & 52.0186(6 ) & 48.4195(6)@xmath23 & - & - & - + nv295 & 54.8509(5 ) & 28.3558(6 ) & 53.9590(6)@xmath11 & 31.5356(6)@xmath11 & - & - & - + nv296 & 45.1948(7 ) & 40.3318(7)@xmath11 & 38.6563(7 ) & 42.3111(8 ) & 30.5764(7)@xmath15 & - & - + nv297 & 29.5022(7 ) & 30.7973(7 ) & 31.8181(7 ) & 30.6969(9 ) & - & - & - + nv298 & 30.2674(4 ) & 31.5946(4 ) & - & - & - & - & - + nv299 & 29.0352(6 ) & 33.2050(8 ) & - & - & - & - & - + nv300 & 29.7935(6 ) & - & - & - & - & - & - + nv301 & 28.2141(5 ) & 27.8536(6 ) & 35.3744(7 ) & 51.6717(9 ) & - & - & - + nv302 & 28.1538(7 ) & - & - & - & - & - & - + nv303 & 27.8162(7 ) & - & - & - & - & - & - + nv304 & 27.6698(8 ) & - & - & - & - & - & - + nv305 & 27.3469(5 ) & 22.6529(6 ) & 19.0827(7 ) & 15.9932(7)@xmath15 & - & - & - + nv306 & 26.0387(5 ) & 40.7315(6 ) & - & - & - & - & - + nv307 & 25.9719(5 ) & - & - & - & - & - & - + nv308 & 25.6507(5 ) & 26.3387(5 ) & 26.0989(7)@xmath11 & 49.6777(8 ) & 42.2776(8)@xmath15 & - & - + nv309 & 25.1601(6 ) & 37.8653(9 ) & 36.0886(8 ) & 25.5930(8)@xmath11 & 30.2689(9)@xmath11 & 24.5741(9)@xmath15 & 38.8636(7)@xmath11 + nv310 & 24.8894(8 ) & 25.9192(6 ) & - & - & - & - & - + nv311 & 24.1469(6 ) & 36.6940(5 ) & - & - & - & - & - + nv312 & 23.0802(5 ) & 36.7987(8 ) & - & - & - & - & - + nv313 & 23.8958(4 ) & - & - & - & - & - & - + nv314 & 23.7405(5 ) & 23.7468(5 ) & 36.4025(6 ) & - & - & - & - + nv315 & 23.6513(5 ) & 23.1521(4 ) & 22.8341(6 ) & 34.3549(6)@xmath11 & - & - & + nv316 & 23.5827(8 ) & - & - & - & - & - & - + nv317 & 23.4525(5 ) & 22.7566(6)@xmath11 & 24.9835(6 ) & 24.4356(7 ) & - & - & - + nv318 & 22.8673(7 ) & - & - & - & - & - & - + nv319 & 20.4323(4 ) & - & - & - & - & - & - + nv320 & 21.1894(5 ) & - & - & - & - & - & - + nv321 & 21.0591(5 ) & 21.5337(7 ) & 20.1156(9 ) & - & - & - & - + nv322 & 20.8524(6 ) & 26.8419(6 ) & 20.4605(6 ) & 27.5321(7)@xmath11 & 43.3720(6)@xmath11 & 25.0637(9)@xmath11 & 35.9001(6)@xmath11 + nv323 & 20.2615(6 ) & 15.7478(6 ) & - & - & - & - & - + nv324 & 19.4953(4 ) & 30.2983(7 ) & 30.3112(9 ) & 17.4876(8 ) & 18.4348(9)@xmath11 & 30.2828(7)@xmath11 & - + nv325 & 18.6761(5 ) & 19.0173(7 ) & 19.0683(5 ) & 36.4063(7 ) & 36.3059(6)@xmath11 & 34.6931(7)@xmath11 & - + nv326 & 17.5728(6 ) & 22.5375(6 ) & - & - & - & - & - + nv327 & 16.4904(3 ) & 33.2019(9 ) & - & - & - & - & - + nv328 & 11.1231(6 ) & - & - & - & - & - & - + model calculations were made with warsaw - new jersey stellar evolution code which is a modern version of paczyski s ( 1970 ) code developed mainly by m. kozowski and r. sienkiewicz to include new opacity and eos data -both from the opal project ( iglesias & rogers 1996 , rogers , swenson & iglesias 1996 , respectively ) - mean effect of rotation , and overshooting . oscillation properties were calculated with modernized version of dziembowski s ( 1977 ) code , which includes effects of rotation . our default hypothesis is that stellar structure is adequately described by standard stellar models , that is that there is no imprints of mass exchange or merging in their current structure . when discussing individual objects we ask whether the data are consistent with this hypothesis . the test is possible if we have a plausible identification of at least one peak in the oscillation spectra with a radial mode . if a measured period is found longer than that of fundamental mode in the model consistent with the star position in the h - r diagram and this position also excludes driving of @xmath24-modes , then we may suspect that the object has lower mass than implied by its luminosity . we consider interpretation of three close frequencies in terms of rotational splitting of @xmath1 mode frequencies . the self - consistency of the hypothesis is checked by comparing the observed and calculated multiplet structures . the frequency distance between the extreme components is a measure of mean rotation rate , which is mode - dependent because different modes probe different parts of the star . the consistency test is possible if nearly uniform rotation is assumed . then we can evaluate the departure from symmetry induced by higher order effects of rotation and compare the result with observations . even if we can not identify individual peaks in oscillation spectra , still we can compare the frequency ranges where the peaks occur with the range of unstable modes in the models . this is of interest because the presence of oscillations left of the blue edge may indicate a substantial helium enhancement , which is expected in certain models of bss formation . peaks at frequencies much lower than that of the fundamental radial mode are of our interest too . these could be attributed to high order @xmath24-modes . it is now a debated problem whether such modes may be excited in @xmath3 scuti stars ( handler 1999 , breger et al . 2002 , dupret et al . 2004 ) . even more interesting possibility is that such peaks represent tidally induced changes ( aerts et al . 2002 , handler et al . 2002 ) . figure 4 shows evolutionary tracks and positions of sxps from @xmath2 cen in the color - magnitude diagram . tracks were calculated with two sets of the chemical composition parameters @xmath25 and @xmath26 . the range of @xmath27 values is consistent with metal abundances observed among the stars belonging to the cluster ( rey et el . the range of masses , which is given in the caption , was chosen to cover the range of stellar @xmath28 and @xmath29 values . the tracks were converted from the @xmath30 @xmath31 to the @xmath32 @xmath29 diagram with the use of kurucz s ( 1998 ) tabular data based on his stellar atmosphere model calculations . the blue edge was determined by means of our linear stability analysis . in this plot , we chose the first overtone because this mode is most frequently excited in the stars considered . the fundamental mode blue edge is redder by some 0.022 mag in @xmath33 and that of the second overtone is bluer by 0.014 mag relative to the that of the first overtone . these lines depend on mode frequency but not on degree @xmath34 . the adopted red edge is empirical and based on @xmath3 scuti data ( rodriguez et al . 2000 ) . we see in fig . 4 that the blue edge is sensitive to the assumed metal content . not surprisingly , the sensitivity to helium content is even stronger . after all , it is helium that does most of the mode driving . increasing the helium abundance from @xmath35 to 0.50 causes the blueward shift of the first overtone blue edge by 0.05 - 0.06 mag . note that most of our stars fall within the instability strip . however , there are few objects which are redder than the red edge but admittedly their positions are quite uncertain . it may depend on metallicity . there is only one star ( nv321 , not shown in the plot ) which is far bluer than the blue edge but , as we comment in the next section , its colour is very uncertain . thus , we see no evidence for a substantial helium enhancement in stellar outer layers . the period - luminosity relations calculated for the first overtone along the tracks are shown in fig . similar relations for other radial modes are obtained by the shifts given in the caption . the relations depend on the colour and metal content . thus , application of sxps as standard candles seems problematical . we will use the plots shown in figs . 4 and 5 to discuss the parameters of individual objects and their mode identification . 6 shows the period ratios for consecutive radial mode pairs as functions of the longer period in the pair ( hereafter prp diagram ) . the relation for the first two radial modes is known as the _ petersen diagram_. we may see that , just as in the case of double mode cepheids , the @xmath27-value is important but the pattern of the period period ratio relation is more complicated . for each object , the first question we asked is whether any of the peaks in the frequency spectrum could be identified with any of radial modes . stars with two radial modes exited are of special interest for constraining stellar parameters . with this in mind we first asked wheteher in our sample we find objects with the period ratios that are consistent with the expected values for consecutive radial modes . fig . 7 , which is similar to fig . 5 of poretti ( 2003 ) for the high amplitude delta scuti stars , shows that we have many candidates for two radial mode pulsators . however , since the ranges for each period ratio are rather wide , we can not use fig . 7 alone as a tool for a definite mode identification . the witdh of the band for the specfic period ratio arises from the uncertainty in metal abundance , in mass and in the effective temperature . how these parameters affect the period shows is shown in fig . thus , both figures must be used jointly . the agreement with the model values was regarded as a hint . if none of the period pairs gave the ratio consistent with the values seen in the prp diagram , we concluded that at most only one peak in the oscillation spectrum may be associated with a radial mode . the case of high amplitude pulsation was considered as another hint for the radial mode identification . it was then checked using the position of the star in the h - r and pl diagrams . if there were no high amplitude peaks , we checked only whether frequency of any of the peak in the power spectrum agrees with any frequency of an unstable radial mode ( or its close nonradial neighbour ) in the model . * v65 * : the first sxps found in the field of @xmath2 cen . it is out of our interest because it does not belong to the cluster physically . the variable v65 is the star number 60026 in the proper motion study of van leeuwen et al . the probability of membership for this variable is 0% . * v194 * : this is a very high amplitude pulsator ( amp = 0.51 mag ) . positions of @xmath9 and @xmath17 modes on the prp diagram ( fig . 6 ) show that these could be radial fundamental and first overtone modes . according to the standard models the @xmath27 value would be intermediate between 0.002 and 0.0002 . then the mass would be close to 1.05 @xmath36 according to its position in the h - r and pl diagrams . its evolutionary status would be near the end of core hydrogen burning . the remaining peaks could be identified only with nonradial modes . we emphasize that high frequency peak @xmath18 is still within the calculated instability range . * v195 * : this is another high amplitude pulsator ( amp = 0.38 mag ) . the shape of the light curve suggests the fundamental radial mode identification . assuming standard evolutionary model , the implied mass is close to 0.95 @xmath36 if @xmath37 and close to 1.10 @xmath36 if @xmath38 . the two close peaks around the dominant frequency are reminiscent of the pattern seen in blazhko rr@xmath39 stars . * v197 * : there are three possible explanations of the frequency pattern found in v197 . first , the ratio between @xmath9 and @xmath19 frequencies is 0.781 which may indicate that they correspond to the radial fundamenal and first overtone modes . but the star is too red for this hypothesis . the second possibility is that only one of three highest peaks might correspond to the radial mode and it could be only the second overtone and then the implied mass would be around 0.9 @xmath36 and @xmath27 should be low . the two peaks close to the main frequency may only be identified as nonradial modes . and alternatively , the three modes might be interpreted as an @xmath1 triplet . then the rotational period would be around 12 days and rotational velocity would be about 10 km / s . * * : the mean parameters are inconsistent with the value of period . at this high temperature the period corresponds to a high order @xmath24-mode which could not be excited . perhaps it is a low mass object . however , the color estimate is very uncertain in this case ( the error in @xmath33 is about 0.1 mag ) . * v199 * : the highest amplitude sxps in @xmath2 cen . no other significant frequencies were detected at the level above 0.005 mag . undoubtedly , we see a fundamental mode pulsator . the implied mass must be about 1.0 @xmath36 and the metallicity must be low . * v200 * : the shape of the light curve points to the fundamental radial mode identification . the second peak is significantly lower , too low to consider the star as a classical double - mode pulsator . the period ratio of 0.761 could be consistent with the first overtone interpretation of @xmath17 only if the metallicity is much higher than that considered for stars of @xmath2 centauri . another problem is too high luminosity for the period . perhaps the object is a foreground galactic star . * v204 * : a high amplitude double - mode pulsator with period ratio equal to 0.774 which indicates metallicity @xmath38 . adopting standard models , we get from the h - r and pl diagrams the mass of 1.15 @xmath36 and the shell hydrogen burning evolutionary status . * v217 * : a monoperiodic and intermediate amplitude sxps . in the standard model with low metallicity and mass around 0.90 @xmath36 the frequency corresponds to the first overtone of radial pulsation or nearby nonradial mode . * v218 * : the period ratio between two dominant modes falls into the range of the 3rd to 2nd radial overtones ratio . however , for standard models , this identification is inconsistent with the position in the h - r and pl diagrams , which indicates that @xmath9 could be only the fundamental or a nearby nonradial mode . such an identification implies mass range @xmath40 @xmath36 and @xmath27 around 0.0002 . another possibility is that the star is significantly undermassive . * v219 * : the power spectrum of the star is very much like that of a typical main sequence @xmath3 scuti star , that is , there are many low amplitude peaks spread over wide range of frequencies . such spectra are extremely difficult to interpret . the main frequency is consistent with fundamental radial mode identification if the metallicity is low and the mass is around 1.00 @xmath36 . the star position in the h - r diagram is near zams , which indicates that the star could be a recent merger of two stars of comparable mass . the low frequency @xmath41 c / d is troublesome . definitely too low for @xmath42-modes . the star is unlikely to excite @xmath24-modes , such as seen in @xmath43 doradus stars , because it is one of the hottest in our sample . perhaps we see ellipsoidal light variations . this is an interesting possibility with ramifications to the problem of bss origin . the implied orbital period of 0.609 d would be significantly shorter than any of the bss in the preston and sneden ( 2000 ) list . * v220 * : a moderate amplitude multimode pulsator with period ratio between two dominant frequencies equal to 0.778 . it implies metallicity close to @xmath38 , which is also consistent with the position of the star in the h - r and pl diagrams and leads to mass estimate of 1.10 @xmath36 . there is a close neighbour of the _ bona fide _ 1st overtone and it may be interpreted only in terms of nonradial modes . * v221 * : many low amplitude peaks , none two of them may be interpreted as two radial modes . high frequencies combined with @xmath29 value indicate excitation of high order @xmath42-modes . * v225 * : a high amplitude pulsator ( amp = 0.22 mag ) with period ratio between two dominant frequencies equal to 0.778 . it implies metallicity close to @xmath38 , which is also consistent with the position of the star at the h - r and pl diagrams and leads to mass estimate of 1.15 @xmath36 . there is a close neighbour of the 1st overtone candidate which could be interpreted as nonradial mode . * v226 * : a high amplitude pulsator ( amp = 0.17 mag ) . frequency @xmath44 c / d may correspond to the first overtone , if the metallicity is rather high and the mass is around 1.05 @xmath36 . then the frequencies @xmath17 and @xmath19 close to @xmath9 could be associated with nonradial modes . the frequency @xmath18 is too low to correspond to a nonradial mode located in the neighborhood of the radial fundamental mode ; therefore , it could be one of the pure @xmath24-modes . * v227 * : frequency @xmath45 c / d may correspond to the first overtone according to the positions of the star in the h - r and pl diagrams . the implied mass estimate is 1.1 @xmath36 and the metallicity around @xmath38 . frequencies @xmath17 and @xmath18 could be nonradial modes in the vicinity of the first overtone radial mode . * v228 * : a moderate amplitude multimode pulsator with period ratio between the frequencies @xmath9 and @xmath19 equal to 0.834 . looking at the prp diagram one can find that they can be the first and second overtone radial modes but only in the case of low metallicity . the frequencies @xmath17 and @xmath18 correspond then to nonradial modes in the vicinity of the first overtone . position of the star on the h - r and pl diagrams agrees with assumption of a low metallicity and indicates the mass of 0.90 @xmath36 . * v229 * : it is one of the bluest sxps in our sample , much bluer than kaluzny et al . ( 1996 ) found . since they relay on @xmath46 colours , we checked the object carefully for blending effects and found none . our colour implies that the dominant period is too long to correspond to the fundamental mode . the problem which is similar to the cases of v198 and v226 , where we do not expect excitation of pure @xmath24-modes in so hot star . * v231 * : the dominant frequency @xmath47 c / d could be due to the radial fundamental mode only for @xmath37 . in this case , the mass of the star is between 0.95 and 1.00 @xmath36 . the frequencies @xmath17 and @xmath18 could be due the excitation of two nearby nonradial modes . the frequency @xmath48 c / d is associated with nonradial mode placed between the first and the second overtone . * v233 * : the dominant frequency @xmath49 c / d could be due to the radial first overtone mode only if @xmath50 . in this case , the mass of the star would be around 0.95 @xmath36 . the low frequency peak at @xmath51 c / d could indicate that the star is , along with v219 , a candidate for an ellipsoidal variable . * v237 * : the high amplitude ( 0.27 mag ) peak at @xmath52 c / d could be identified as the radial fundamental mode . such an identification is consistent with the positions in the h - r and pl diagrams for low @xmath27 models . then the mass of the star is about 0.95 @xmath36 . if @xmath9 is indeed the fundamental mode then the expected position of the second overtone radial mode is around 24 c / d . thus , we conclude that the frequency @xmath53 c / d could correspond to the second overtone . the peak at @xmath54 c / d is due to a nonradial mode in the vicinity of fundamental frequency . * v238 * : again , the dominant frequency @xmath55 c / d could be due to the excitation of the radial fundamental mode only if @xmath27 is low and the mass of the star is 0.95 @xmath36 . the frequencies @xmath17 , @xmath18 and @xmath19 may be attributed to nearby nonradial modes . the expected position of the second overtone radial mode is around 38 c / d thus the frequency @xmath56 c / d would be due to the nonradial @xmath42-mode . * v249 * : a moderate amplitude multimode pulsator with the period ratio between two dominant frequencies equal to 0.818 . it is consistent with an assumption that @xmath9 and @xmath17 correspond to the first and second overtone radial modes . this requires high @xmath27 values , somewhat higher than @xmath38 . however this interpretation is inconsistent with the star position in the h - r and pl diagrams . * v250 * : a double mode pulsator with @xmath9 and @xmath17 being the radial fundamental and the second overtone modes , respectively . the position at the h - r and pl diagrams indicates a low metallicity and the mass between 0.90 and 0.95 @xmath36 . * v252 * : a monoperiodic and low amplitude star . the position of the star in the h - r and pl diagrams is consistent with the fundamental radial mode identification only for a low metallicity and the mass between 0.85 and 0.90 @xmath36 . * v253 * : a moderate amplitude multimode pulsator . the dominant frequency could be the fundamental radial mode only for our low metallicity models . in this case , the star has mass around 1.00 @xmath36 . the frequency @xmath17 corresponds to a nonradial mode in the vicinity of fundamental peak and frequencies @xmath18 and @xmath19 to some high order nonradial @xmath42 modes . * nv294 * : this is the shortest period sxps known . up to now , the shortest period variable of this type was v10 from ngc6397 ( kaluzny and thompson 2003 ) . nv294 is a multimode and very low amplitude object . with our h - r and pl diagrams the frequency @xmath57 c / d may be associated with the radial fundamental mode , in a model of high metallicity and mass around 1.2 @xmath36 . in this picture , the frequencies @xmath9 , @xmath18 and @xmath19 correspond to the high order nonradial @xmath42-modes . * nv295 * : this is another very short period , low amplitude and multimode object . the frequency @xmath58 c / d could correspond to the radial second overtone mode but only for low values of @xmath27 . in this case , the position of the star at the h - r and pl diagrams indicates the mass between 0.90 and 0.95 @xmath36 . the frequencies @xmath9 , @xmath18 and @xmath19 could correspond to nonradial modes . * nv296 * : another object similar to nv294 and nv295 . there are five frequencies present in the light curve of the star . with our models only the lowest amplitude peak at frequency @xmath20 might be identified with a radial mode . it would be the second overtone and in this case the metallicity of the star would be around @xmath38 and the mass would be between 1.10 and 1.15 @xmath36 . in such a case , other frequencies would correspond to nonradial high order @xmath42-modes . * nv297 * : a low amplitude multimode object with four relatively close peaks . one of them could be identified with the second overtone radial mode and then the star mass would be around 1.20 @xmath36 at @xmath38 and about 1.05 @xmath36 at @xmath37 . we consider possibility that three of the four frequencies represent a rotationally splitted @xmath1 triplet . in the plausible interpretation @xmath17 would be a retrograde and @xmath19 would be the prograde mode . then the frequency separation would imply rather fast rotation ( over 100 km / s ) . however , the calculated position of the @xmath59 component does not agree with the position of any of the two remaining peaks . it was much closer to the @xmath19 . * nv298 * : the dominant peak of the two observed in the star may be identified , in our models , with the fundamental radial mode if the metallicity is rather high and the mass is between 1.15 and 1.20 @xmath36 . the second close frequency must then be associated with a nonradial mode . * nv299 * : the dominant peak could be the radial first overtone mode in a star of a relatively high metallicity and mass around 1.05 @xmath36 . then the frequency @xmath17 would correspond to a nonradial mode located between the radial first and second overtones . the star is slightly too red for this interpretation but it is located in a crowded region of the cluster and its @xmath4 magnitude zero point is determined with the accuracy of around 0.04 mag . thus the star may be in fact even 0.05 mag bluer . * nv300 * : a monoperiodic and low amplitude sxps . the star could be the radial fundamental mode pulsator only if we assume the high metallicity and the mass of around 1.15 @xmath36 . * nv301 * : a multimode and low amplitude sxps . one of the two dominant frequencies may be the radial second overtone mode , according to our models . the star has then a high metallicity and the mass around 1.05 @xmath36 . the frequencies @xmath18 and @xmath19 should correspond to the high order nonradial @xmath42-modes . * nv302 * : a monoperiodic and low amplitude star . the frequency @xmath60 c / d could be the first overtone radial mode but only for our @xmath38 models . in this case , the mass of the star is between 1.10 and 1.15 @xmath36 . it is also possible that @xmath9 is the second overtone . this requires low @xmath27 and , in this case , the mass of the star is between 0.90 and 0.95 @xmath36 . * nv303 * : a monoperiodic and very low amplitude star . the only frequency determined could be associated with the first radial overtone . then the star has low metallicity and mass slightly less than 1.0 @xmath36 . * nv304 * : a similar pulsator to nv303 . again , the first radial overtone interpretation possible , however in this case the models imply high metallicity and mass around 1.05 @xmath36 . * nv305 * : a multiperiodic object with interesting power spectrum . the amplitude increases systematically with the frequency . the period ratio between two dominant peaks is equal to 0.828 - a value which could be the ratio between the first and second radial overtones , but only for very low metallicity . it is supported by the position of the star in the h - r and pl diagrams implying the mass in range of @xmath61 . the frequencies @xmath18 and @xmath19 would correspond to nonradial peaks in the vicinity of the fundamental radial mode . * nv306 * : one of the faintest sxps in our sample . the dominant frequency could be due to the radial fundamental mode . with this interpretation , the metallicity of the object is low and its mass is between 0.90 and 0.95 @xmath36 . the expected position of the second overtone radial mode is close to 40 c / d , thus the second peak could be connected with this mode . * nv307 * : a monoperiodic and moderate amplitude star . our models imply that the excited mode must be of higher order than two , which seems to be in conflict with the star relatively low temperature . * nv308 * : a multiperiodic object . of the three dominant peaks at low frequency one could be identified with the first overtone for our low metallicity models with the mass of about 0.90 @xmath36 . we also considered the possible interpretation of these three peaks in terms of rotationally splitted @xmath1 mode . in this case , the problem is opposite to that accounted in nv297 - the observed asymmetry is too high for the rotation rate of about 50 km / s implied by the distance between @xmath9 and @xmath17 . * nv309 * : the star has one of the richest frequency spectrum . we determined as many as seven frequencies . the @xmath62 ratio could be the frequency ratio between the first and second overtones of radial pulsations . this interpretation is however in conflict with the h - r position , where the star is too red . for this star we obtained a consistent interpretation of three peaks ( @xmath21 , @xmath9 and @xmath19 ) as a rotational splitted @xmath1 triplet . for our model with @xmath37 and mass of 0.95 @xmath36 the separation between @xmath19 and @xmath21 implies rotational velocity of 60 km / s . with this velocity the centroid component , displaced from the center by the second order effect of rotation , comes close to @xmath9 . this triplet is not far from the frequency of the first radial overtone . the peak @xmath20 would then be located in the vicinity of the second overtone while the highest frequency peaks ( @xmath18 and @xmath17 ) of the third overtone . * nv310 * : the only two frequencies are close to the expected position of the radial second overtone , for our models with high metallicity and mass around 1.10 @xmath36 . * nv311 * : the two frequencies measured in this star , @xmath63 c / d and @xmath64 c / d , may be due to the radial fundamental and the second overtone modes , respectively . unfortunately , we can not verify this hypothesis because we do not know the exact magnitude and color of the star , which is located in very crowded region of the cluster . * nv312 * : one of the brightest sxps in @xmath2 cen . the frequency @xmath65 c / d could be associated with the radial second overtone mode but only for high metallicity . in this case , the mass of the star is around 1.2 @xmath36 . the peak at @xmath66 c / d can be interpreted as the high order @xmath42-mode . * nv313 * : a monoperiodic , high amplitude and faintest sxps in the field of @xmath2 cen . the high amplitude may suggest that the star is a radial mode pulsator . the problem is that in our standard models , the star with the observed effective temperature and luminosity would have the fundamental radial mode period by factor 1.3 shorter . the star is also one of the hottest in our sample and we certainly do not expect excitation of @xmath24-modes in such a star . in fact , we would rather expect higher order modes excitation , therefore this object is likely to be undermassive by factor of 1.7 or more . * nv314 * : the frequencies @xmath67 c / d and @xmath68 c / d could be the radial fundamental and second overtone modes , respectively , but only if the star would be bluer by 0.06 mag . it is possible , because the star is located in the crowded region ( three bright companions ) and zero point in @xmath4-band is poorly determined . * nv315 * : one of the brightest sxps in @xmath2 cen with a rich oscillation spectrum . the photometric data combined with our model parameters imply that even the lowest frequency , @xmath9 , is higher than the second overtone radial mode . the preference to the high frequency mode excitation seems to be in conflict with the low temperature of the star , the one of the lowest in our sample . * nv316 * : a monoperiodic and low amplitude star . the period is too long for fundamental radial mode , according to our models . perhaps the star is another case of an undermassive object . * nv317 * : a low amplitude multiperiodic star . assuming our standard models , the frequency @xmath69 c / d could be attributed to the radial second overtone mode in a star with a rather high metallicity and mass of around 1.05 @xmath36 . three other peaks ( @xmath17 , @xmath19 and @xmath18 ) in this model can be interpreted as a rotationally splitted @xmath1 triplet , which implies rotational velocity of about 120 km / s . alternatively , another combination of three peaks ( @xmath9 , @xmath19 and @xmath18 ) might be interpreted as an @xmath1 triplet , which would imply rotational velocity of about 80 km / s . * nv318 * : a low amplitude and monoperiodic star . the frequency @xmath9 could be due to the radial second overtone mode in a model with high metallicity and mass between 1.05 and 1.10 @xmath36 . * nv319 * : a moderate amplitude and monoperiodic star . the only frequency could be that of the radial first overtone mode in a star of low metallicity and the mass between 0.85 and 0.90 @xmath36 . * nv320 * : it is almost identical to the previous object . * nv321 * : a multiperiodic and moderate amplitude star . the frequencies @xmath9 , @xmath17 and @xmath18 form an asymmetric triplet which perhaps might be interpreted as rotationally splitted @xmath1 mode . unfortunately , the variable is located in very crowded field and the mean @xmath5 and @xmath4 magnitudes are very uncertain . * nv322 * : a moderate amplitude variable with one of the richest power spectrum . the ratio between two dominant frequencies is 0.777 which may suggest that they are fundamental and first overtone radial modes . we have no color determination for this star and we can not exploit this hypothesis . * nv323 * : a double mode and low amplitude pulsator . surprisingly , we found the same period ratio of 0.777 as in previous case . however , we have a problem with interpretation of the frequencies in terms of the fundamental and first overtone radial pulsations . the period ratio suggests the metallicity of @xmath70 . our models imply the h - r position by some of 0.06 mag bluer than observed . * nv324 * : a high amplitude and multiperiodic star . its rich power spectrum is shown in fig . 2 . according to our models , the dominant frequency is close to the radial first overtone for low metallicity and mass of around 1.05 @xmath36 . the frequencies @xmath17 , @xmath18 and @xmath21 are all close to that of the third radial overtone mode . * nv325 * : a moderate amplitude and multiperiodic object . the dominant frequency may be interpreted , according to our models , as due to the second radial overtone in a star with mass between 1.15 and 1.20 @xmath36 if @xmath38 or between 1.00 and 1.05 @xmath36 if @xmath37 . presence of three high frequency modes is surprising in view of rather low temperature and high luminosity of the object . * nv326 * : this is a classical double mode star with @xmath71 c / d as radial fundamental and @xmath72 c / d as the first overtone modes . the frequency ratio of 0.780 nicely agrees with that of fundamental and first overtone in an intermediate metallicity object . further , the periods and the metallicity agree with the position of the star in the h - r and pl diagrams . the implied mass is then between 0.95 and 1.00 @xmath36 . * nv327 * : it is one of the bluest stars in our sample . the frequency @xmath9 may be associated with the fundamental mode but only if the star is significantly undermassive . * nv328 * : a low amplitude and monoperiodic star . it is the reddest star in our sample . the only frequency , according to our models , may be interpreted as the fundamental radial mode , implying the mass around 1.00 @xmath36 at @xmath37 or mass between 0.85 and 0.90 @xmath36 at @xmath38 . in most cases the observed properties of sxps in @xmath2 centauri may be explained in terms of standard evolutionary models and their linear nonadiabatic properties . in a number of cases , where the observational evidences pointed to radial mode excitation , we derived the masses of the objects . they all fall into the range @xmath73 . the implied evolutionary status covers both the main sequence and early post - main sequence evolutionary phases . only in four cases the observed parameters were inconsistent with the standard models and could be reconciled with the undermassive objects . such objects could arise in several situations . they can appear as a result of mergers , if there is a mass loss or element mixing . they could also arise as the result of mass exchange in binary systems . this possibility may be contemplated only in the case of very large mass loss on the lower red giant branch . applicable models would have rather low mass ( @xmath74 ) , most of which contained in the helium core . such models were once considered by dziembowski and kozowski ( 1974 ) for high amplitude @xmath3 scuti stars , called then dwarf cepheids . we found two cases of stars with long periodicities that may be interpreted as due to the tidally induced distortion . this could be an indirect evidence for binarity . possible implications for the mechanism of bss formation are interesting . therefore , the two objects deserve further studies to check the interpretation . this could be done by searching for induced pulsation period changes ( current data do not allow to reject nor confirm the proposal ) or by means of radial velocity measurements . our search for rotationally split triplets resulted in three plausible cases . for nv309 the inferred equatorial velocity would be about 60 km / s , and for nv317 it would be about 80 or 120 km / s , depending on three frequency peaks involved . for v197 the interpretation of all three close peaks as @xmath1 triplet results in equatorial rotational velocity of about 10 km / s . we found considerable diversity in pulsational behaviour ranging from that typical for main sequence @xmath3 scuti stars characterized by low amplitude and multimode variability to high amplitude cepheid - like monomode pulsations . this is quite amazing bearing in mind the fact that the spread in @xmath4 brightness is only about one magnitude . to give the picture of the observed diversity , fig . 8 shows the phased light curves of the monoperiodic pulsators from our sample . examples of @xmath3 scuti - like behaviour are objects like v219 , nv294 , nv295 and nv296 . these are among the faintest and the hottest stars in our sample . amongst the faintest objects there are also some relatively high amplitude pulsators ( v198 , v249 and nv313 ) but all of them are suspected for being undermassive . the star v199 , whose light curve resembles those of cepheids or rr@xmath39 stars , is amongst most luminous objects . other high amplitude objects are multiperiodic . among them there are five classical double mode pulsators i.e. stars with fundamental and the first overtone excited . these objects were crucial for deriving constrains on masses and metallicities . in three of them we detected additional , apparently nonradial modes . coexistence of high amplitude radial modes with low amplitude nonradial modes was first seen by walraven , walraven and balona ( 1992 ) in the @xmath3 scuti star ai vel ( see also poretti 2003 ) . in some of our cases the nonradial modes are located close to radial modes in the frequency spectrum but not in all , just like in ai vel . three clean cases of spectra consisting of a dominant peak surrounded by close low amplitude peaks are v195 , v197 and nv321 . these are possible analogs of blazhko rr@xmath39 stars . we are thankful to dr . poretti , the referee , for pointing mis - identification of @xmath75 frequency in nv324 . this work was supported by the kbn grants number 1 p03d 006 27 to a. olech , 5 p03d 030 20 to w.a . dziembowski , 5 p03d 012 20 to a.a . pamyatnykh , and 5 p03d 004 21 to j. kaluzny . a. schwarzenberg - czerny would like to acknowledge generous grant by clifford and mary corbridge trust of cambridge , england . 99 aerts c. , handler g. , arentoft t. , vandenbussche b. , medupe r. , sterken c. 2002 , mnras , 333 , l35 breger m. et al . 2002 , mnras , 329 , 531 carney b.w . , latham d.w . , laird j.b . , grant c.e . , morse j.a . , 2001 , aj , 122 , 3419 dupret m .- a . , grigahcne a. , garrido r. , gabriel m. , scuflaire r. 2004 , a&a , 414 , l1 dziembowski , w. & kozowski , 1974 , acta astron . , 24 , 245 . dziembowski w. , 1977 , acta astron . , 27 , 203 eadie w.t . , drijard d. , james f.e . , 1971 , statistical methods in experimental physics , amsterdam , north - holland . ferraz - mello , s. , 1981 , aj , 86 , 619 foster , g. , 194 , aj , 109 , 1889 gray d. f. & desikachary k. , 1973 , apj 181 , 523 . handler g. 1999 , mnras , 309 , l19 handler g. et al . 2002 , mnras , 333 , 262 iglesias c.a . , rogers f.j . 1996 , apj , 464 , 943 kaluzny j. , kubiak m. , szymaski m. , udalski a. , krzemiski w. , & mateo m. 1996 , a&as , 120 , 139 kaluzny j. , kubiak m. , szymaski m. , udalski a. , krzemiski w. , & mateo m. , 1997a , a&as , 122 , 471 kaluzny j. , thompson i.b . , krzemiski w. , olech a. , pych w. , mochejska b. , 2002 , in van leeuwen f. , hughes j.d . , piotto g. , eds . , asp conf . 265 , `` omega centauri , a unique window into astrophysics '' , astron . , san francisco , p. 155 kaluzny j. , thompson i.b . , 2003 , aj , 125 , 2534 kaluzny j. , olech a. , thompson i.b . , pych w. , krzemiski w. , schwarzenberg - czerny a. , 2004 , a&a , 424 , 1101 lomb , 1976 , a&ss 39 , 477 lombardi j.c . , jr . , rasio f.a . , 2002 , in shara m.m , asp conference proceedings , vol . 263 , `` stellar collisions , mergers and their consequences '' , san francisco : astronomical society of the pacific , p. 35 lombardi j.c . , warren j.s . , rasio f.a . , silla a. , warren a.r . , 2002 , apj , 568 , 939 kepler , s.o . , 1993 , in proceedings of second wet workshop , baltic astronomy , 2 , 515 kurucz r.l . , 1998 , in bedding t.r . , booth a.j . , davis j. eds . , proceedings of iau symposium 189 , `` the interaction between observation and theory '' , kluwer , dordrecht , p. 217 mazur b. , krzemiski w. , thompson i.b . , 2003 , mnras , 340 , 1205 mccrea w.h . , 1964 , mnras , 128 , 147 paczyski b. , 1970 , acta astron . , 20 , 47 poretti e. , 2003 , a&a , 409 , 1031 pych w. , kaluzny j. , krzemiski w. schwarzenberg - czerny a. , thompson , i. b. , 2001 , a&a , 367 , 148 rey s .- c . , lee y .- w . m . , walker a. , baird s. , 2000 , aj , 119 , 1824 rodriguez e. , lpez - gonzlez m.j . , lpez de coca p. , 2000 , in breger m. , montgomery m.h . , asp conf . 210 , `` delta scuti and related stars '' , astron . pac . , san francisco , p. 499 rogers f.j . , swenson f.j . , iglesias c.a . 1996 , apj , 456 , 902 scargle , j.h . , 1982 , apj , 263 , 835 schlegel d.j . , finkbeiner d.p . , davis m. 1998 , apj , 500 , 525 schwarzenberg - czerny a. , 1996 , apj letters , 460 , l107 schwarzenberg - czerny a. , 1999 , apj , 516 , 315 thompson i.b . , kaluzny j. , pych w. , burley g. , krzemiski w. , paczyski b. , persson s.e . , preston g.w . , 2001 , aj , 121 , 3089 van leeuwen f. , le poole r.s . , reijns r.a . , freeman k.c . , de zeeuw p.t . 2000 , a&a , 360 , 472 walraven th . , walraven j. , balona l.a . , 1992 , mnras , 254 , 59

we obtain phase diagrams of regular and irregular finite connectivity spin - glasses . contact is firstly established between properties of the phase diagram and the performances of low density parity check codes ( ldpc ) within the replica symmetric ( rs ) anstz . we then study the location of the dynamical and critical transition of these systems within the one step replica symmetry breaking theory ( rsb ) , extending similar calculations that have been performed in the past for the bethe spin - glass problem . we observe that , away from the nishimori line , in the low temperature region , the location of the dynamical transition line change within the rsb theory , in comparison with the ( rs ) case . for ldpc decoding over the binary erasure channel we find , at zero temperature and rate an rs critical transition point located at while the critical rsb transition point is located at , to be compared with the corresponding shannon bound . for the binary symmetric channel ( bsc ) we show that the low temperature reentrant behavior of the dynamical transition line , observed within the rs anstz , changes within the rsb theory ; the location of the dynamical transition point occurring at higher values of the channel noise . possible practical implications to improve the performances of the state - of - the - art error correcting codes are discussed . + pacs numbers : 89.90+n , 89.70+c,05.50+q +

cond-mat0507458

the survival of ferromagnetic ordering under the disruption of frozen random fields @xcite and the onset of spin - glass ( sg ) order in systems characterized by random competing interactions @xcite are two central problems in the statistical mechanics of systems with quenched randomness . when properly rephrased , these two problems turn out to relate to core problems in other , quite different disciplines . insight into these two closely related problems in any given system is obtained from the corresponding phase diagram as a function of the physical parameters , such as temperature and disorder strength . finite - connectivity mean - field sg models @xcite and their _ p_-spin counterpart @xcite are important for two distinct reasons . firstly , despite being mean - field in nature , they are believed to share common properties with finite - dimensional spin - glass systems @xcite . secondly , the statistical mechanics of sg model systems with fixed finite connectivity relates to the study of various hard computational problems . in the last two decades the analysis of finite - connectivity sg systems has also offered new tools for understanding the very nature of several optimization problems _ and _ improve algorithmic performances @xcite . in this paper we first obtain the phase diagram of finite - connectivity mean - field sg models . we then investigate the relation between features of the phase diagram and the performance of low density parity check ( ldpc ) codes , analyzing the effects induced by the replica symmetry breaking ( rsb ) nature of solutions obtained for these finite - connectivity sg models @xcite . in order to answer these questions , we exploit the relation between finite - connectivity sg systems and a class of exactly solvable models ; we employ an efficient computational method that has been shown to be quite effective for understanding several features of finite - dimensional sg @xcite . we will consider a generalized model unifying the frameworks for a few distinct problems . particular attention will be given to two special cases : a ) the bethe sg which has been extensively studied under both replica symmetric ( rs ) and rsb anstze @xcite and recently considered in its ferromagnetically biased version @xcite . b ) gallager ( ldpc ) error - correcting codes @xcite . while the former is well known within the statistical physics community , many in this community are less familiar with the latter and its links to the physics of disordered systems . reliable transmission of information in noisy conditions is one of the basic problems of modern communication technology . correspondingly , one of the central problems of information theory is channel coding . methods to achieve reliable information transmission rely on the introduction of structured redundancy to the original message in order to compensate for corruption due to noise . shannon derived rigorous bounds on the level of redundancy ( the code rate ) required to enable error - free communication for a given channel noise @xcite . however , shannon s theorems are non - constructive and do not offer a practical coding scheme . different methods , based on various redundancy construction recipes , have been proposed over the years @xcite . despite the fact that error - correcting codes are now widely used in a variety of applications , current performances of most methods are significantly below shannon s bound . one family of error - correcting codes that have been shown to provide close - to - optimal performances is that of gallager s ldpc codes @xcite . in these codes the structured redundancy is introduced through parity checks of boolean sums of randomly sampled message bits . the relation between the parity - check error - correcting codes and sg models , and consequently between quite different disciplines as statistical physics and information theory , has been pointed to in the seminal work of sourlas @xcite . decoding in ldpc codes corresponds to a class of sg models defined by an underlying lattice geometry that reflects the low density character of the code construction . the importance of this very last point , that made the above correspondence useful and of practical relevance was not understood until recently @xcite . methods developed in the statistical mechanics of diluted sg under the rs anstz have been successfully employed to compute macroscopic properties of such systems for different parameter values ; reasonable agreement with observed decoding bounds have been reported @xcite . to our knowledge however , the only phase diagram that has been evaluated for ldpc codes is in the limit of large coordination / large multi - spin interaction numbers @xcite , which also relates to the phase diagram of the random energy model ( rem ) @xcite . in the following section we introduce the class of ising spin systems to be analyzed and its links to the bethe sg and ldpc decoding problems this will be followed by section [ sec : rs ] where we review the rs equations and the corresponding solutions are obtained using a computational method adopted from the study of @xmath5-dimensional hierarchical models . as part of this section we will present the method and briefly discuss the relation between finite - connectivity sg and @xmath5-dimensional hierarchical models . in section [ sec : rsb ] we introduce the equations behind the rsb theory of finite - connectivity sg studied here and present results obtained for the bethe sg problem and ldpc error - correcting codes under two different transmission channels : the binary erasure channel ( bec ) and the binary symmetric channel ( bsc ) . in section [ sec : disc ] we discuss the results obtained and point to future research directions . the two different models examined in this work stem from a unified physical system of @xmath6-spin interaction ising model on a lattice of fixed connectivity @xmath7 . the hamiltonian describing the system takes the form @xmath8 where @xmath9 represents @xmath10 binary variables , @xmath11 a set of indices related to interaction @xmath12 and @xmath13 the corresponding connectivity tensor with the following properties : @xmath14 the parameter @xmath15 is related to the strength of the random exchange interaction . the various models considered in this work differ in the connectivity parameters and type of bias used , which corresponds to an additional component of the hamiltonian ( [ h_0 ] ) . the bethe sg has been extensively studied in the past @xcite and represents a special case of the hamiltonian ( [ h_0 ] ) where one is only considering pairwise interactions , @xmath16 . bethe sg with ferromagnetically biased exchange interaction @xmath15 , which we will focus on here , has been recently studied in @xcite , where the distribution of the quenched random variable @xmath15 takes the form @xmath17 and where the parameter @xmath18 $ ] is @xmath19 for a ferromagnet and @xmath20 for the unbiased bethe sg . we consider a simple communication model whereby @xmath21 bit messages are encoded using ldpc codes and then communicated through two channel types : @xmath22 the binary symmetric channel ( bsc ) where bits are flipped with probability @xmath23 during transmission and @xmath24 the binary erasure channel ( bec ) where bits are received with probability @xmath25 and fail to arrive with probability @xmath23 . a gallager ldpc code is defined by a binary matrix @xmath26 $ ] , concatenating two very sparse matrices known to both sender and receiver , with @xmath27 ( of dimensionality @xmath28 ) being invertible ; the matrix @xmath29 is of dimensionality @xmath30 . ldpc codes can be employed over various finite fields @xcite ; we will restrict the treatment in this paper to binary symbols . encoding refers to mapping the original @xmath21 dimensional binary message vector @xmath31 to an @xmath10 dimensional binary code word @xmath32 ( @xmath33 ) by taking the product @xmath34 , where all operations are performed in the field @xmath35 and are indicated by @xmath36 . the generator matrix is @xmath37 \mbox { ( mod 2)}$ ] , where @xmath38 is the @xmath39 identity matrix , implying that @xmath40 and that the first @xmath21 bits of @xmath41 are set to the message @xmath42 . in _ regular _ gallager codes the number of non - zero elements in each row of @xmath43 is chosen to be exactly @xmath6 . the number of elements per column is then @xmath44 , where the code rate is @xmath45 ( for unbiased messages ) . we will mostly focus here on regular code constructions . the corruption process that is most intuitively linked to physical systems is the bsc , where the encoded vector @xmath41 is corrupted by flip noise represented by the vector @xmath46 with components independently drawn from @xmath47 . the received vector takes the form @xmath48 . decoding is carried out by multiplying the received message by the matrix @xmath43 to produce the _ syndrome _ vector @xmath49 from which an estimate @xmath50 for the noise vector can be inferred . an estimate for the original message is then obtained as the first @xmath21 bits of @xmath51 . the bayes optimal estimator ( also known as _ marginal posterior maximiser _ ( mpm ) ) for noise vector bits is defined as @xmath52 . computing the mpm estimate directly is computationally hard , and is typically approximated in practice by employing an iterative inference algorithm such as belief propagation @xcite . the connection to statistical physics becomes clear when the field @xmath35 is replaced by ising spins @xmath53 and mod @xmath54 sums by products @xcite . the syndrome vector acquires the form of a multi - spin coupling @xmath55 where @xmath56 and @xmath57 . the @xmath6 indices of nonzero elements in the row @xmath12 of @xmath58 are given by @xmath59 and where @xmath60 is the connectivity tensor is given in equation ( [ tensor ] ) . after gauging @xmath61 the multi - spin coupling coefficient vanishes @xmath62 and the parity check component of the hamiltonian takes the form ( [ h_0 ] ) with @xmath63 forcing all candidate vectors to obey the parity checks . this notation is used to distinguish the temperature from the parity check constraints and to facilitate the finite temperature study of error - correcting codes . prior knowledge of the biased noise vector is added through a second component to the hamiltonian ( [ h_0 ] ) . after gauging , the new term takes the form @xmath64 where @xmath65 is the nishimori condition @xcite ( where @xmath66 ) , which corresponds to the correct prior for the variables @xmath67 . decoding at the nishimori condition corresponds to the mpm estimation @xcite . this gives rise to the hamiltonian @xmath68 for the bsc . another important estimator one should mention is the _ maximum a posteriori _ ( map ) estimator defined as @xmath69 . this corresponds to zero temperature decoding @xcite . codeword bits transmitted over a bec have a probability @xmath25 of being received ( in tact ) . encoding is carried out similarly to the bsc , but decoding is based on inferring the codeword itself , rather than the noise vector . in the case of the bec , solution vectors are , by definition , codewords and should therefore obey @xmath70 , where @xmath71 are the received vectors ( with a certain number of unknown bits ) . an estimate of the original message is then obtained by inferring the missing bits using mpm . similar practical methods to those described in the bsc may be employed also in this case . the statistical mechanics formulation of the problem is similar to that of the bsc , except for the second component added to the hamiltonian ( [ h_0 ] ) , that takes the form @xmath72 \ .\ ] ] code performance in the bec will be considered only at the zero temperature limit as in @xcite . a review of ldpc codes and their link to statistical mechanics has been presented in @xcite . the analysis carried out here is a natural extension of similar calculations done in the past for the bethe sg @xcite model and @xmath23-spin ising models @xcite . we therefore briefly review the rs theory of the model described by hamiltonian ( [ rfrb ] ) as both the bethe sg and ldpc decoding can be considered as special cases of the same model . to compute the free energy of the problem , consider the replica equality , @xmath73 where @xmath74 is the analytical continuation in the interval @xmath75 $ ] of the replicated partition function @xcite . following the standard derivation , e.g. as in @xcite , and employing the rs anstz , one obtains the following expression for the replica symmetric free - energy as a function of the local field distribution @xmath76 ( details can be found in appendix a ) @xmath77 there are both site and bond contributions , @xmath78 where @xmath79 denotes an average over the two sources of randomness in ( [ rfrb ] ) , and @xmath80 we denote @xmath81 where @xmath82 \nonumber \\ \beta a ( \ { x_j \},j)&= & \frac{1}{2 } \log [ r(+)r(- ) ] \ .\end{aligned}\ ] ] the polynomials @xmath83 are defined via , @xmath84 , \label{poly}\ ] ] where @xmath85 . the above expression , in a similar form , has already been obtained in @xcite , and the equivalence with their expression is shown in appendix a. we prefer the current formulation because of its manifest simplicity ( the two contributions to the free - energy coming from sites and bonds are here in evidence ) and because it depends on the function @xmath86 , which will play a central role in the rsb case , examined in section [ sec : rsb ] . the above writing also introduces the polynomials @xmath83 , already well known in the context of hierarchical lattices @xcite . evaluating the saddle point equations of the above free - energy expression is the basic premise of the rs theory , so that the crux of the quenched randomness problem ( [ rfrb ] ) lies in the convolution @xcite , @xmath87 where @xmath88 , @xmath89 ; and where the recursion relations are defined via @xmath90 we can also rewrite for convenience the above convolution with the aid of the auxiliary distribution @xmath91 , as a set of two coupled equations ( cfr . appendix a ) , @xmath92 to briefly explain the link between ldpc decoding and the replica formulation we will restrict this discussion to the bsc . the replica analysis as well as decoding in other noisy channels follow a similar path @xcite . following the derivation of @xcite we write saddle point equations for the ldpc decoding problem , considering a special case of the hamiltonian ( [ rfrb ] ) and the saddle point equations ( [ crux2 ] ) , where @xmath93 is deterministic @xmath94 most of the numerical results obtained so far are obtained along the nishimori line @xmath95 . here we investigate the phase diagram at @xmath96 prior temperature @xmath97 , for different choices of the parameters @xmath7 and @xmath6 . a typical case we will discuss in detail is the one corresponding to @xmath98 , @xmath99 . we will check the consistency between the results obtained and true decoding experiments , using belief propagation ( bp ) @xcite , at and away from the nishimori condition . we will not review bp in the paper as it is a well known message passing algorithm @xcite with strong links to the bethe approximation and its variants @xcite . the iterative algorithm is a microscopic equivalent of equations ( [ crux3 ] ) . the set of ( rs ) equations ( [ crux2 ] ) have been considered and solved in a variety of different ways already . this section introduces to a different method . we propose an alternative approach to existing methods ( e.g. population dynamics or monte carlo techniques ) which solves equations of the form ( [ crux ] ) efficiently and accurately . the main reason why it has not been used to study finite - connectivity sg models so far is because of the different context in which it was firstly introduced @xcite . the method we propose allows to determine the phase diagram of diluted systems of the type studied here and analyze the nature and exact location of the phase boundaries . it borrows tools and concepts from a different branch of research , namely the area of hierarchical models @xcite , to obtain accurate phase diagrams . the iterative method we now describe has a corresponding computational effort that scales as @xmath100 , n being the number of bins introduced and @xmath6 the multi - spin interaction parameter . equation ( [ crux ] ) complicates , after a few iterations , even a simple initial local field distribution . the distribution @xmath76 and its conjugate distribution @xmath101 are represented via histograms . each histogram is characterized by two quantities , @xmath102 and @xmath76 , where the latter is the associated probability . equation ( [ crux3 ] ) is the convolution of @xmath103 distributions , taking into account the distribution of the random field @xmath104 , contained in the average @xmath105 . this is numerically constructed from pairwise convolutions of distributions . for a @xmath1 rate gallager code , e.g. , three pairwise convolutions are required . a pairwise convolution ( or binary operation ) is achieved as follows : + ( 1 ) the histograms are placed on a grid properly drawn at each iteration step of the algorithm . all histograms that fall within the same grid cell are combined in such a way as to preserve the averages across the domain of the local field ( and its conjugated ) distribution(s ) . the histograms that falls outside the grid , representing a pre - determined small fraction of the overall probability , are similarly combined into a single histogram . in this procedure , a histogram that falls within a narrow band of a grid boundary is proportionally shared between neighboring bins . the gridding is done separately for @xmath106 and @xmath107 . + ( 2 ) two gridded distributions are convoluted according to equation ( [ crux2 ] ) regenerating the original number of histograms . + the convolution of @xmath108 distributions is achieved , from pairwise convolutions , as follows : ( 1 ) the pairwise convolution contained in the first equation of ( [ crux2 ] ) is cycled once for , e.g. , a code rate @xmath1 . ( 2 ) the resulting distribution is pairwise - convoluted with the noise channel distribution @xmath109 . ( 3 ) the resulting distribution is further self - convoluted according to the second equation in ( [ crux2 ] ) . this completes the entire process , that is then iterated until convergence is achieved . multiple cycling operations are also an interesting option being related to , e.g. , the source compression problem . the flows of the probability distribution @xmath76 determine the phase diagram of the system . most of our results have been obtained considering @xmath110 independent bins , corresponding to the flows of @xmath111 quantities ( the corresponding number of elementary operations considered is the number of bins to the power of @xmath112 ) . each of the two grids at @xmath106 and @xmath107 have e.g. @xmath113 bins on each side of the average value . + similar set of equations to ( [ crux ] ) have been discussed in a variety of different contexts . a very important one is that of real - space renormalization - group theory @xcite @xcite @xcite . eventhough the algebraic form of the convolution corresponding to these two problems is the same , the two problems are indeed very different . in the case of hierarchical models the distribution of the exchange interaction is studied , while in the problem we are discussing in this work one considers the distribution of the local field . this reflects the difference in the two underlying geometries . the @xmath5-dimensional hierarchical lattice include loops that are one of the essential characteristics that makes hierarchical lattices better candidates , within exactly solvable models , to study the problem of finite - dimensional sg @xcite . the similarity between equations ( [ crux3 ] ) and those related to hierarchical models is indicative of a deeper relation between the two ( both geometries are hierarchical in the sense of ref . @xcite ) and will be the subject of future work . the calculated phase diagram for the gallager ldpc code of rate @xmath1 and a bsc of flip rate @xmath23 as a function of the prior temperature @xmath97 are shown if fig . [ ldpc_phase ] . we report both the location of the dynamical noise transition @xmath114 , defined by the noise level at which suboptimal solutions emerge , and critical noise transition @xmath115 , where the ferromagnetic solution ceases to be dominant , for all ranges of temperature @xmath97 . the phase diagram was calculated considering the flows of @xmath111 quantities dictated by the convolution ( [ crux ] ) . as mentioned before the above number should be taken to the power @xmath112 to obtain the number of elementary operations within , e.g. , the second equation of ( [ crux2 ] ) . the location of the dynamical and critical transition values can be determined by analyzing the free - energy and the complexity as a function of the channel noise level for different values of the temperature @xmath116 ( fig . [ bscrsfree ] ) . we will refer noise level that corresponds to the emergence of sub - optimal solutions as the dynamical transition . at any value @xmath117 , the computed free - energy ( [ free - energy ] ) reduces to the ferromagnetic free - energy @xmath118 . consider first temperature values in the proximity of the nishimori line . the trajectories exhibit a crossover to an unstable distribution first , and then flow to the ferromagnetic sink below @xmath119 ( corresponding to a decoding success ) or to a stable non - ferromagnetic distribution ( corresponding to a decoding failure ) above it . a similar behavior is observed for any temperature above the nishimori line , the stable distribution being different for each values of the temperature @xmath120 . rather different behavior is observed below the nishimori line , where the unstable distribution is a strong coupling distribution , meaning that the average local field strength increases but the rescaled distribution remains constant . finally , the trajectory flows again , either to the ferromagnetic sink or to a fixed local field distribution , which is stable under iteration of the recursion ( [ crux3 ] ) . the computed free - energy , according to the convolution ( [ crux3 ] ) , is higher than the ferromagnetic free - energy for all noise levels @xmath121 . above the critical noise value @xmath115 , the computed free - energy becomes lower than the ferromagnetic free - energy . the location of the dynamical transition has been determined , at any temperature , considering the flows , under iteration , of the convolution ( [ crux ] ) , while the location of the critical noise threshold has been determined from the free - energy ( [ free - energy ] ) as the point where the computed and the ferromagnetic free - energies coincides . in the inset of fig . [ ldpc_phase ] we also report in square boxes the results of real decoding experiments obtained as follows . we iterated the probabilistic decoding algorithm @xcite , at different values of the noise channel @xmath23 , for an ensemble of @xmath111 networks of @xmath122 and @xmath123 nodes respectively . the inset shows the calculated mean values , the corresponding error being smaller than the corresponding symbol size . a precise estimate of the dynamical transition point has been obtained considering the percentage of decoding success within the network ensemble as a function of the noise level , for different system sizes , @xcite . the decoding experiments were performed for different choices of the temperature . along the nishimori line ( see fig . [ ldpc_phase ] ) we find the value of @xmath124 in excellent agreement with the reported value for regular @xmath1 gallager codes @xcite . notice that the reentrant nature of both the dynamical and critical boundaries , as we found solving the convolution ( [ crux2 ] ) , is confirmed by the decoding experiments . in the range of noise levels @xmath125 , decoding is possible only in a range of finite temperatures , as indicated in the inset of fig . [ ldpc_phase ] . the boundary between decodeable and undecodeable regions reaches its maximal value at the nishimori temperature ; zero temperature decoding fails at the same noise level . the reentrant nature of the critical phase boundary increases below the nishimori line , but eventually weakens as the temperature is decreases further , a phenomenon that has already been discussed in a variety of different contexts @xcite . the fact that mpm and zero temperature bp / rs decoding do not coincide is not in contradiction with the results obtained in @xcite which indicate that map and mpm decoding provide the same critical transition point . clearly , zero temperature bp / rs decoding does not provide map results due to rsb effects . the latter and their effect on the phase diagram and on the reentrant nature of the phase boundaries will be further discussed in section [ sec : rsb ] . an independent calculation has been performed at zero temperature . in this case the recursion relations ( [ recursion ] ) simplify and one finds limiting expressions for the two functions @xmath126 , @xmath127 . implementing numerically such asymptotic expressions would require an independent calculation where one explicitly changes the recursion relations ( [ recursion ] ) to their asymptotic @xmath128 form . the importance of the polynomials @xmath129 becomes clear at this point . at any temperature considered , we distinguish the finite and zero temperature contributions , corresponding to factorize the largest exponent in the sum ( [ poly ] ) , so that the limit of zero temperature is particularly simple to handle in our approach and no explicit changes are required in the recursion relations ( [ recursion ] ) . we observe that at zero temperature the local field distribution is not peaked on integers as one might expect . we find that the relative weight of the background behind the integer - peaked distribution weakens as one consider the rsb solution , but restricting the distribution of the local fields to integer values , which results in a modification of the recursion relations ( [ recursion ] ) leads to different results . for ldpc ( rs ) decoding in the bsc this leads to an erroneous dynamical transition value of @xmath130 , marked in fig . [ ldpc_phase ] by a black filled dot , well below the reported zero temperature value of @xmath131 . similar behavior occurs in the binary erasure channel problem if one simply neglects the background distribution , as we will discuss in section [ sec : rsb ] . the initial condition @xmath132 , at @xmath133 , has been used as initial condition for the bsc analysis . a dependence of the solution of the convolution ( [ crux3 ] ) on the initial condition for the local field has been observed . we have studied the above convolution ( [ crux3 ] ) for different choices of the initial condition in the form @xmath134 the choice of @xmath135 , where @xmath23 is the flip rate of the channel results in the phase diagram of fig . [ initial_cond ] . _ any _ choice of initial condition with @xmath136 ( including the case @xmath137 considered above ) leads to the dynamical transition line ( dashed line in fig . [ ldpc_phase ] ) , except for the choice @xmath135 which leads to a different dynamical transition . the location of the critical transition remains unchanged . for instance , using the initial condition @xmath135 , the dynamical transition along the nishimori line is increased to @xmath138 , well above the known dynamical transition point . the physical interpretation of this behavior remains to be explained . since it results from a simple change in the initial conditions of ( [ crux ] ) , one might argue that a properly modified decoding algorithm would enable us to decode above current @xmath114 values ; this is a possibility that we can not rule out and further research is needed in order to elucidate this point . we where able to perform a second set of decoding experiments that coincide with the dynamical transition line in fig . [ initial_cond ] but only for noise level values below the value @xmath139 . the reentrant nature of the dynamical transition line ( the dynamical transition at zero temperature being @xmath140 ) is reduced in this case . for a constant @xmath6 value one can select non - integer connectivity values of average @xmath141 and consider a modified form of the convolution ( [ crux3 ] ) that involves a linear combination of terms corresponding to different connectivity values . non - integer values of the multi - spin interaction parameter @xmath6 can also be easily accommodated in ( [ crux3 ] ) . it has been reported that irregular low - density parity - check codes outperform regular constructions @xcite . we considered a generalized form of the convolution ( [ crux3 ] ) , ( an explicit form is not provided for brevity ) where both connectivity @xmath7 and multi - spin interaction @xmath6 take non - integer values such that the rate @xmath142 is preserved . in fig . [ irregular ] we report the dynamical transition line for the regular construction @xmath143 ( @xmath1 ) already considered in the previous section together with the dynamical transition lines for irregular constructions of two values @xmath99 and @xmath144 and @xmath145 and @xmath98 , such that @xmath146 ( left to right ) and @xmath147 . the dynamical transition value is observed to increase for the irregular constructions . this does not give a direct information on the specific pattern of irregularity that one should consider but provides quantitative information on the average connectivity / multi - spin interaction values one should investigate further . as for the regular case discussed earlier , a phase diagrammatic approach to the problem of irregular ldpc codes is important to understand the reason behind the successful decoding properties of irregular ldpc codes . a similar behavior to the one shown in fig . [ irregular ] , but less pronounced , has also been observed for ldpc decoding with the initial conditions ( [ ic ] ) when @xmath135 . we considered the finite - connectivity model ( [ rfrb ] ) for the values of the connectivity @xmath149 and @xmath99 . despite the fact that it does not correspond to any gallager ldpc code construction it is interesting to consider these connectivity parameters for two reasons : 1 ) they can represent code constructions in other variants of the ldpc family such as sourlas or mn codes @xcite , and as such may also be linked to other ldpc - based applications . 2 ) they give rise to a sg behavior which is of interest . in fig . [ c9k4 ] we report the corresponding phase diagram obtained considering the convolution ( [ crux3 ] ) . an sg phase occurs in this case . at low values of the flip rate @xmath23 the convolution ( [ crux ] ) flows to the ferromagnetic sink . in the vicinity of the dynamical transition line ( marked in fig [ c9k4 ] as the solid black line ) the trajectories under iteration of the above convolution cross over to an unstable distribution and then either flow to the ferromagnetic sink , below the dynamical transition , or to a stable non - ferromagnetic distribution . for temperatures below the nishimori line , at high enough values of the flip rate , the convolution instead flows to a spin - glass distribution . the role of rsb for the problem ( [ rfrb ] ) is the subject of this and the following sections . the one - step rsb solution for the bethe sg , as a special limiting case of the hamiltonian ( [ rfrb ] ) , has been studied in detail long ago . in this case it is well known that the rs solution is incorrect @xcite . recently , the onset of rsb phenomena in some limiting cases of the problem ( [ rfrb ] ) has been considered , relying on the variational approximation @xcite ( for a recent formulation see @xcite ) . consider the equation for the global order parameter @xmath150 , firstly introduced for the bethe sg by mottishaw , @xcite in the general case of a multi - spin interaction parameter @xmath6 , @xmath151 \big ] . \nonumber \\ \label{gopk}\end{aligned}\ ] ] where @xmath152 labels different replicas of the system . the above expression holds in general , whether replica symmetry is broken or not ( see appendix b for the derivation ) . in the rs approximation one assumes that different replicas are equivalent and the functional order parameter @xmath150 depends only on the variable @xmath153 . in the limit of @xmath154 , can be shown to satisfy , @xmath155 . \nonumber \\ \label{rs}\end{aligned}\ ] ] this equation corresponds to the equation that have been written for the bethe sg , as soon as @xmath16 and for vanishing message prior temperature @xmath156 @xcite . a derivation of equation ( [ rs ] ) is found in appendix b , where we show the equivalence of the above expression with the set of recursion relations ( [ crux ] ) , for the local field and conjugate distributions . in the one - step rsb approximation one assumes instead that the functional order parameter @xmath150 depends only on the variables @xmath157 , where @xmath158 is the replica index , @xmath159 and @xmath160 . in this case one finds that the corresponding equation for the functional order parameter @xmath161 becomes @xmath162 \times \nonumber \\ \big [ \frac{m}{2 } \sum_m a ( \ { r_{mj } \},j ) \big ] , \nonumber \label{rsb}\end{aligned}\ ] ] where the normalization is given by @xmath163 the corresponding one - step rsb convolution follows , assuming that the order parameter factorize according to @xmath164 the assumptions behind the _ factorized _ anstz ( [ factorized ] ) , where the local field distribution is independent from the intra - group index , are discussed in @xcite . in this case the one - step rsb convolution ( [ rsb ] ) become @xmath165 \times \nonumber \\ & & \delta \big ( \hat{x}-u ( \ { x_j\},j ) \big ) , \label{crux_rsb}\end{aligned}\ ] ] where the normalization @xmath166 is given by @xmath167,\ ] ] and where @xmath168 . the above equations are valid , under the assumption ( [ factorized ] ) , at finite temperatures and arbitrary connectivity values @xmath7 and @xmath6 . after its very first formulation @xcite the convolution ( [ crux_rsb ] ) has been studied in a range of different contexts , from random @xmath6-sat to the coloring problem @xcite . using the identity , @xmath169 = \big [ \frac { 2 \cosh ( u ( \ { x_j\},j ) ) } { \prod_j \cosh ( \beta x_j ) } \big ] ^ { \mu } , \label{identity}\ ] ] and substituting it in equation ( [ crux_rsb ] ) one recovers the one - step rsb equations that have been also derived using the variational approach @xcite . this shows that the so called frozen variational anstz , considered in @xcite and the factorized anstz of @xcite are equivalent . even though there exists some degree of arbitrariness @xcite in the way one redistributes the re - weighting between the two equations ( [ crux_rsb ] ) , we prefer the notation that involves the exponential of the energy term @xmath170 , rather than the less obvious ( and numerically more difficult to implement ) parallel expression that follows considering the right hand side of ( [ identity ] ) @xcite . the simplified expression of the free - energy presented in section iii was obtained in virtue of the same identity ( [ identity ] ) , and the same degree of freedom writing the free - energy was already discussed for the bethe sg @xcite ( the right hand side of ( [ identity ] ) being , at @xmath16 the function @xmath171 discussed in ref . @xcite ) . we apply the method described in section iii to study the one - step rsb convolution ( [ crux_rsb ] ) of the bethe sg , which has been extensively studied and recently considered also in its ferromagnetically biased version @xcite . in this case the distribution of the quenched random variable @xmath15 is @xmath172 and where the parameter @xmath18 $ ] is @xmath19 for a ferromagnet and @xmath20 for the unbiased bethe sg . the bethe sg corresponds indeed to the original problem ( [ h_0 ] ) at @xmath16 . we will present results obtained at zero temperature . in the zero temperature limit we find the following expression for the free energy @xmath173 , @xmath174 where explicit expressions for the site and link contributions to the ground state energy are @xmath175 corresponding , at @xmath176 and @xmath16 , to the rs equation ( [ free - energy ] ) in the zero temperature limit and neglecting the prior term in ( [ rfrb ] ) . similar expression have been discussed in @xcite . in the limit of zero temperature one finds @xmath177 and @xmath178 , being @xmath179 and @xmath180 . in what follows we present results obtained under the @xmath181 anstz as well as for the full distribution . the relative importance of non - integer values in the distribution that solves the above convolution ( [ crux_rsb ] ) is rather limited in this case . the background behind the integer peaked values weakens as one consider the one - step rsb case , and may disappear in a fully replica symmetry broken scenario @xcite . in fig . [ bethe_gse ] we show the ground state energy as a function of the replica symmetry breaking parameter @xmath182 for different values of the ferromagnetic bias using the integer peaked anstz and in the case of the full local field distribution . the ground state energy one extrapolates including the background calculation is higher than the one obtained within the three peak anstz @xcite . in fig . [ bethe_compl ] we show the corresponding complexity or configurational entropy , that can be obtained as a parametric plot in terms of the rsb parameter @xmath183 , the free - energy @xmath184 being its legendre transform @xcite . we observe a positive complexity for all reported values of @xmath185 . at higher values of the ferromagnetic bias @xmath186 a positive complexity is not observed anymore . one should also consider that the above values should be extrapolated in the limit of large number of bins , so that one expects a slightly higher value @xmath187 . in fig . [ bethe_mag ] we show the magnetization for in the replica symmetric and one - step replica symmetry broken solution at different values of the ferromagnetic bias . at values @xmath188 one recovers the results known for the unbiased bethe sg . the solid line indicates the magnetization computed in the replica symmetric approximation ( notice that this is a computed magnetization for a finite number of bins , so that it approaches the expected value of @xmath189 from below ) , while the square boxes represent the magnetization in the one step rsb approximation without considering the background . for each value of @xmath190 , we extrapolated the magnetization that corresponds to the value of @xmath191 where the ground state energy has a maximum . similarly , the dashed line points to the results obtained considering the true convolution ( [ crux ] ) , that includes non integer values of the local field . the putative size of the so called @xmath192 , previously discussed @xcite , if any , is rather small and do not agree with previous estimates ( we observe the magnetization within the rsb calculation to vanish at values of @xmath193 , if one considers a large enough number of bins within the calculation , implying that the size of the mixed phase , if any , is vanishingly small ) . it should be said however that the calculation of @xcite considered a full one step rsb calculation , in the sense of @xcite , where the factorized approximation is relaxed . to investigate this point better , we performed the following : instead of implementing the convolution ( [ crux ] ) within the factorized approximation @xcite , we choose a population of @xmath194 distributions ( including a site dependence ) . at each step of the iteration process we consider pairwise - convolutions of local field distributions corresponding to different sites , selected at random ; after a transient regime , we obtain the equilibrium population . waiting a number of iteration steps of the order of the population size , according to our calculation , the local field distributions become site independent , meaning that the factorized approximation provides similar results to those of the full rsb anstz . if one compares fig . [ bethe_compl ] with the corresponding plot in ref . @xcite , where a similar calculation of the complexity is presented , we see that the location of the maximum of the complexity as a function of @xmath182 _ decreases _ for increasing @xmath190 values ( see fig . [ bethe_compl ] ) , as one might intuitively expect , increasing the ferromagnetic bias . in the case of the bec discussed in section ii , the form of the convolution ( [ crux3 ] ) simplifies and the distribution of the local field , and its conjugated distribution become bimodal . despite its simplicity the binary erasure channel problem is of practical relevance @xcite . for the bec one writes the convolution ( [ crux3 ] ) as @xmath195 where @xmath196 is a delta function located at @xmath197 , @xmath198 is the distribution of the effective local field and @xmath199 its conjugate distribution . the resulting convolution @xcite involving the two distributions @xmath198 and @xmath199 , in the one - step rsb case , is given by @xmath200 \times \nonumber \\ & & \delta \big ( \hat{x}-u(\ { x_i\},j ) \big ) , \nonumber \\ \label{crux_bec}\end{aligned}\ ] ] where @xmath201 are the binomial coefficients of order @xmath112 and @xmath202 the set of normalization coefficients defined by @xmath203.\ ] ] one can readily check that the above convolution reduces to the variational approximation @xcite using identity ( [ identity ] ) . we note that the variational anstz , as already discussed for the case of the bsc , coincides with the factorized anstz ( [ factorized ] ) @xcite . we consider an explicit expression for the one step rsb free - energy that involves the binomial coefficients of order @xmath6 . within the factorized hypothesis @xcite , the expression for the free - energy is easily computed as a function of the rsb parameter @xmath182 . we present results for the bec at zero temperature for values of the code rate r=1/4 and @xmath204 ( note that we also consider in what follows the integer peak anstz for the local field distribution @xcite in order to compare with the general case of non - integer local fields ) . in the inset of fig . [ bec_1 ] we show the energy as a function of the erasure probability @xmath23 for the rate value @xmath204 within the integer valued local field approximation . notice that the integer value approximation corresponds to having only three bins in our approach of section iii . in fig . [ bec_2 ] we compare the results within and without the integer approximation for a code rate @xmath1 . the dynamical transition is defined as the point where sub - optimal solutions emerge the corresponding complexity being non zero , as already discussed for the bsc channel ( see fig . [ bec_2 ] ) . the location of the dynamical transition is also the point where the recursion ( [ crux_rsb ] ) flows to a subdominant distribution , rather than the ferromagnetic sink . this coincides with the point where algorithms devised for real decoding experiments fail to decode the original message . within the rs framework , we find that the location of the dynamical transition changes significantly if non - integer values of the local field are taken into account . within the three peak anstz the local field distribution is simply defined as @xcite @xmath205 the calculation we performed within the integer local field approximation ( for rates @xmath204 and @xmath206 in fig . [ bec_1 ] and fig . [ bec_2 ] respectively ) might suggest to interpret the point @xmath207 for the rate value @xmath204 as the dynamical transition point @xcite . if however one compares ( see fig . [ bec_2 ] ) with the results of the local field distribution that now includes non - integer values ( the thick solid line corresponding to @xmath208 ) we see that the location of the dynamical transition is rather different . moreover the behavior of the local field probability @xmath209 around the value @xmath210 , within the @xmath211 anstz is rather suspicious . indeed , including non - integer values of the local field turns out to be crucial for having a quantitative description of the dynamical transition and possibly explains the disagreement with numerical experiments that has been reported lately for the r=1/2 bec problem @xcite . if one avoids the _ three peak _ anstz , the nature of the flows dictated by the recursion ( [ crux_bec ] ) is smooth as it always leads to a well - behaved behavior of the probability densities @xmath212 and @xmath213 . up to this point all considerations are still within the rs approximation . moreover , considering the full convolution ( [ crux_bec ] ) as we do in what follows , the location of the dynamical transition in the one - step rsb solution becomes higher than that obtained within rs , which arguably implies that one may be able to construct the algorithmic analog of the rsb recursion ( [ crux_bec ] ) to improve decoding performances . similarly , within the rsb calculation , the location of the critical transition is increased for the rate @xmath1 considered above and found to be very close to the shannon bound value @xmath214 ( cfr . [ bec_5 ] ) . + we iterated the convolution ( [ crux_bec ] ) within the one - step rsb approximation described in section vi and computed the free - energy as a function of the rsb parameter @xmath182 for several values of the erasure probability @xmath23 , above the dynamical transition point @xmath215 as reported in fig . [ bec_3 ] . in this range of @xmath23 values , the free - energy grows as a function of @xmath182 , dropping to the ferromagnetic free - energy value ( being @xmath216 for the bec ) at some critical @xmath217 . at higher values of the erasure probability @xmath23 , the free - energy eventually drops to zero after having reached its maximum ( see fig . [ bec_3 ] ) . according to the standard recipe of maximizing the free - energy with respect to @xmath12 @xcite , we extrapolate the value of the dynamical transition within the one - step rsb calculation , finding the value of @xmath218 , indicated in fig . [ bec_2 ] by the dotted line , well above the rs value @xmath219 . the corresponding complexity is shown in fig . [ bec_4 ] , for different values of the erasure probability considered . we observe two branches , a physical ( right ) and an unphysical one ( left ) as explained in ref . @xcite . from the former branch we extrapolated the rsb energy value corresponding to the noise channel probability . notice that above the rs dynamical transition we do observe , for values of @xmath220 that the second branch , that is the relevant one , is not present . this can be explained because the corresponding curves for the free - energy ( fig . [ bec_3 ] ) do not reach a maximum and drop instead to the ferromagnetic energy value ( @xmath221 ) . the dynamical transition within the rsb calculation increases . if one considers the free - energy as a function of the rsb parameter @xmath182 at values of the erasure probability slightly higher that the rs critical transition point , one finds that the maximum is located at positive energies , implying that the critical transition point also changes under the one step rsb calculation . in this way [ bec_5 ] ) we extrapolated the value of the critical transition , within the rsb approximation , to @xmath222 . note that shannon s bound for the specific code rate considered is @xmath223 . in comparison , the critical rs transition point @xmath224 . the fact that the dynamical and critical transition points increase within the one - step rsb calculation might be surprising at first . in the next section we will show that a similar situation is observed for the bsc at temperatures below the nishimori condition @xcite , where it is known that the rs solution is dominant @xcite . it is also reasonable to believe that the above one step rsb calculation is exact for this problem @xcite @xcite . the above findings concerning the location of the critical ( rsb ) transition seem to support this belief . in this section we return to the bsc ( [ rfrb ] ) and employ the same considerations discussed for the bec in the previous section . we consider the convolution ( [ crux_rsb ] ) along the nishimori line @xmath225 . the location of the dynamical transition has been discussed in section [ sec : rs ] and we found the value of @xmath226 within the rs approximation , in agreement with probabilistic decoding estimates . not surprisingly @xcite , if one looks at the free - energy as a function of the rsb parameter @xmath12 for values of the bsc noise probability @xmath227 , one finds that the free - energy decreases ( see the inset in fig . [ bsc_1 ] ) , indicating that the rs solution is dominant , so that the dynamical transition is the same in the one - step rsb case . @xmath228 . however , at temperatures well below the nishimori line instead , i.e. @xmath229 , we observe the following facts : above the rs dynamical transition @xmath230 the free - energy increases as a function of the rsb parameter @xmath182 , dropping to the corresponding ferromagnetic free - energy at some value @xmath217 , before reaching a maximum . this means that the second , physical branch of the complexity is not present , the analysis following the same lines as for the bec calculation . at higher values of the flip rate , i.e. @xmath231 the second branch of the complexity ( see fig . [ bsc_2 ] ) appears , indicating a non - zero complexity and that we are above the rsb dynamical transition . the corresponding complexity for the ldpc decoding at @xmath232 is reported in fig . [ bsc_2 ] . these findings imply that the location of the dynamical transition is improved with respect with the rs value . the same behavior has been observed , progressively reducing in importance at higher values of the temperature @xmath233 and @xmath234 , while at @xmath235 , corresponding to the nishimori line , this phenomenon is no longer observed . the inset of fig . [ bsc_1 ] shows the computed free - energy as a function of the rsb parameter @xmath182 at the nishimori temperature channel noise @xmath236 . we conclude observing that , below the nishimori line , the reentrant nature of the dynamical transition is not observed within the rsb calculation . we introduced a new method to compute the phase diagram of finite - connectivity sg model systems and discussed the connection with ldpc codes . we extended the one - step rsb theory , originally formulated for the bethe sg problem , to the gallager ldpc decoding for two different type of corruption processes , bec and bsc . for the bec at zero temperature we found that the location of the dynamical transition increases with respect to the rs case . our estimate for the rsb critical transition value is a mere one percent away from capacity , differently from the rs estimate . the critical and dynamical properties of ldpc decoding have also been studied for the case of a bsc . the reentrant nature of the ldpc phase diagram and the corresponding changes that rsb implies in its low temperature properties have also been studied . we found that below the nishimori line the dynamical transition increases within the rsb theory . possible implications to improve the decoding performances of the state - of - the - art error correcting codes are one of the perspectives we wish to address . we would like to construct the algorithmic analog of the convolution ( [ crux_rsb ] ) in order to improve the decoding performances as predicted by the rsb theory . the method we used to determine the phase diagrams along this work is very promising . it allows to study , without loss of continuity , rather different problems , ranging from the @xmath5-dimensional hierarchical sg to finite - connectivity mean - field models . in this sense it would be very interesting to apply this method to clarify a few open issues . on the one side we would like to investigate , within the one - step rsb theory , the finite temperature phase diagram of the bethe sg in a magnetic field . on the other , it would be interesting to explore the presence , if any , of rsb phenomena in @xmath5-dimensional sg hierarchical model systems . this work was supported in part by the european community s human potential programme under contract number hprn - ct-2002 - 00319 . y. imry and s. ma , phys . * 35 * , 1399 ( 1975 ) . s.f . edwards and p.w . anderson , j. phys . f * 5 * , 965 ( 1975 ) . k. wong and d. sherrington , j. phys . a * 20 * , l785 ( 1987 ) . m. mzard and g. parisi , europhys . lett . * 3 * , 1067 ( 1987 ) . p. mottishaw , europhys . lett . * 4 * , 333 ( 1987 ) . c. dedominicis and p. mottishaw , j. phys . a * 20 * , l375 ( 1987 ) . c. dedominicis and y.y . goldschmidt , j. phys . a * 22 * , 399 ( 1989 ) ; c. dedominicis and y.y . goldschmidt , phys . b * 41 * , 2184 ( 1990 ) . h. rieger and s. kirkpatrick , phys . b * 45 * , 9772 ( 1992 ) . r. monasson , j. phys . a * 31 * , 513 ( 1998 ) . y.y . goldschmidt and p.y . lai , j. phys a * 23 * , l775 ( 1990 ) . m. mzard and g. parisi , eur . j. b * 20 * , 217 ( 2001 ) ; m. mzard and g. parisi , j. stat . phys . * 111 * , 1 ( 2003 ) . m. mzard , g. parisi , and r. zecchina , science * 297 * , 812 ( 2002 ) . dewar and p. mottishaw , j. phys . a * 21 * , l1135 ( 1988 ) . g. migliorini and a.n . berker , phys . b * 57 * , 426 ( 1998 ) . t. castellani , f. krzakala , and f. ricci - tersenghi , cond - mat 0403053 . gallager low density parity check codes , ire trans . theory * it-8 * , 21 ( 1962 ) ; r.g . gallager , low density parity check codes , research monograph series * 21 * , cambridge , ma : mit press ( 1963 ) . shannon , mathematical theory of communication . j. * 27 * , ( pt i ) 379 , ( pt ii ) 623 ( 1948 ) . t. cover and j. thomas , elements of information theory , new york : wiley ( 1991 ) . n. sourlas , nature * 339 * , 693 ( 1989 ) . y. kabashima and d. saad , europhys . lett . * 45 * , 97 ( 1999 ) . r. vicente , d. saad , and y. kabashima , phys . e * 60 * , 5352 ( 1999 ) . a. montanari , eur . j. b , * 23 * , 121 ( 2001 ) . b. derrida , phys . b * 24 * , 2613 ( 1981 ) . k. nakamura , y. kabashima and d. saad , europhys . 56 * , 610 - 616 , ( 2001 ) . mackay , ieee trans . theory * 45 * , 399 ( 1999 ) . h. nishimori , j. phys . c * 13 * , 4071 ( 1980 ) . h. nishimori , j. phys . japan * 62 * , 2973 ( 1993 ) . y. iba , j. phys . a * 32 * , 3875 ( 1999 ) . s. franz , m. leone , a. montanari , and f. ricci - tersenghi , phys . e * 66 * 046120 ( 2002 ) . y. kabashima and d. saad , j. phys . a * * , r1 ( 2004 ) . d. bowman and k. levin , phys . b * 25 * , 3438 ( 1982 ) . y. fu and p.w . anderson , j. phys . a * 19 * , 1605 ( 1986 ) . m. mzard , g. parisi and m. virasoro , spin glass theory and beyond , singapore : world scientific ( 1987 ) . t.murayama , y. kabashima , d.saad and r. vicente , phys . e * 62 * , 1577 ( 2000 ) . t. tanaka and d. saad , j. phys . a * 36 * , 11143 ( 2003 ) . f. kschischang and b. frey , ieee j. selected areas commun . * 2 * , 153 ( 1998 ) . j. pearl , probabilistic reasoning in intelligent systems , san francisco , ca : morgan kaufmann publishers inc . yedida , w.t . freeman , and y. weiss , advances in neural information processing systems * 13 * , t.k . leen , t.g . dietterich and v. treps ( eds . ) , cambridge , ma : mit press ( 2001 ) . m. opper and d. saad ( editors ) , advanced mean field methods - theory and practice , cambridge , ma : mit press ( 2001 ) . d. andelman and a.n . berker , phys . b * 29 * , 2630 ( 1984 ) . m. hinczewski and a.n . berker , cond - mat 0507293 . berker and s.r . mackay , phys . b * 33 * , 4712 ( 1986 ) . m. kaufman and b. griffiths , phys . b * 24 * , 496 ( 1981 ) ; b. griffiths and m. kaufman , phys . rev . b * 26 * , 5022 ( 1982 ) . i. kanter and d. saad , j. phys . a * 33 * , 1675 ( 2000 ) . t. richardson and r. urbanke , ieee trans . theory , * 47 * , 599 ( 2001 ) . s.r mackay , a.n . berker , and s. kirkpatrick , phys . lett . * 48 * , 767 ( 1982 ) . c. amoruso and a.k . hartmann , cond - mat 0401464 . j. van mourik , d. saad and y. kabashima , phys . e * 66 * , 026705 ( 2002 ) . y. kabashima , cond - mat 0506311 . s. franz , m. leone , f. ricci - tersenghi , and r. zecchina , phys . * 87 * , 127 ( 2001 ) . m. mzard and r. zecchina , phys . e , * 66 * , 56126 ( 2002 ) . luby , m. mitzenmacher , m.a . shokrollahi , and d.a . spielman , ieee trans . theory * 47 * , 569 ( 2001 ) . t. richardson and r. urbanke , codes , systems , and graphical models , b. marcus and j. rosenthal ( eds . ) , new york , ny : springer ( 2001 ) . h. nishimori and d. sherrington , disordered and complex systems , p. sollich , a. coolen , l. hughston and r.streater ( eds . ) , new york , ny : aip , 67 ( 2001 ) ; p. le doussal and a.b . harris , phys . lett * 61 * , 625 ( 1988 ) . considering the replicated partition function , @xmath237 one can perform the average over the corresponding underlying geometry in ( [ partition_f ] ) introducing an integral representation for the finite - connectivity constrain @xmath238 in the spirit of @xcite , defining the set of overlaps @xmath239 and conjugated functional order parameter @xmath240 , @xcite . this leads , under the rs anstz , @xmath241 to the following expression for the free energy , @xmath242 \nonumber \\ & \times & \ln \big [ 1 + \tanh ( \beta j ) \prod_{j=1}^k \tanh ( \beta x_j ) \big ] \nonumber \\ & + & c \int dx d \hat{x } \pi(x ) \hat { \pi } ( \hat{x } ) \log \big ( 1+\tanh ( \beta x ) \tanh ( \beta \hat{x } ) \big ) \nonumber \\ & + & c \int dy \hat { \pi } ( \hat{x } ) \log \cosh ( \beta \hat{x } ) \nonumber \\ & - & e _ { \zeta } \int \prod_{l=1}^c dy_l \hat { \pi}(x_l ) \big [ 2 \cosh \beta ( \sum_{l=1}^c \hat{x}_l + f \zeta ) \big ] . \label{free - energy2}\end{aligned}\ ] ] the corresponding saddle point equations ( [ crux2 ] ) are derived differentiating the above expression with respect to the local field @xmath243 and its conjugated distribution @xmath244 . in order to obtain the free - energy ( [ free - energy ] ) , the second and third term in ( [ free - energy2 ] ) can be combined via equation ( [ crux2 ] ) , to obtain @xmath245 the fourth term can be written as @xmath246 where we wrote @xmath247 and where the last term in the free - energy expression above is nothing but @xmath248 . it is easy to see that several terms cancels out , together with the first term in equation ( [ free - energy2 ] ) , to obtain equation ( [ free - energy ] ) . similar simplifications and their interpretation were discussed already in the case of the bethe sg @xcite . consider the set of coupled equations ( [ crux2 ] ) , @xmath249 for the case of @xmath16 , with no random bias , corresponding to the bethe sg problem , let us define the laplace transform of the conjugated local field distribution as @xmath250 substituting the first equation of ( [ crux2b ] ) into the second and then into ( [ gop2 ] ) , one finds @xmath251 . \nonumber \\\end{aligned}\ ] ] which is the expression for the global order parameter found in @xcite and corresponds to expression ( [ rs ] ) for @xmath16 . the expression for arbitrary @xmath6 values of the multi - spin interaction simply involves multiple convolutions of the mottishaw global order parameter . the meaning of the global order parameter ( [ gop2 ] ) can be understood observing that , e.g. in the case of the bethe sg considered here ( in a uniform magnetic field ) , the partition function is given by @xcite @xmath252 [ g_m ( \sigma_o^ { \alpha } ) ] ^c,\ ] ] where @xmath253 \big ] . \\ \label{rsb2}\end{aligned}\ ] ] the subscript @xmath254 indicates the dependence of the global order parameter on the sub - shell tree structure . if one is interested in the thermodynamical limit , it is expected that the fixed point of the above recursion determines the equilibrium properties of the system . a delicate point is the choice of the boundary conditions which are well known @xcite to play an important role in this problem . the corresponding extension of equation ( [ rsb2 ] ) to the case of arbitrary @xmath6 values has been reported in equation ( [ gopk ] ) .

our knowledge of the diffuse emission that encompasses active regions is very limited . in the present paper we investigate two off - limb active regions , namely ar10939 and ar10961 , to probe the underlying heating mechanisms . for this purpose we have used spectral observations from hinode / eis and employed the emission measure ( em ) technique to obtain the thermal structure of these diffuse regions . our results show that the characteristic em distributions of the diffuse emission regions peak at and the cool - ward slopes are in the range 1.4 - 3.3 . this suggests that both low as well as high frequency nanoflare heating events are at work . our results provide additional constraints on the properties of these diffuse emission regions and their contribution to the background / foreground when active region cores are observed on - disk .

1409.1447

our knowledge on how the solar corona is heated and maintained at a million degrees kelvin , when the photosphere below is at 5800 kelvin , is still far from being comprehensive . active regions ( ars ) are ideal observing targets to probe the underlying heating mechanisms as they are the locations of profound heating processes . in addition , they show a wide distribution of physical parameters . accurate measurements of such parameters are critical in the formulation and constraint of coronal heating theories . for extended discussions on coronal heating , refer to , e.g. , @xcite and @xcite . topologically , ars possess different structures like the core loops primarily seen at @xmath1 3 - 5 mk rooted in moss regions which are seen primarily at around 1 - 2 mk ( * ? ? ? * ; * ? ? ? * ; * ? ? ? * and references therein ) , warm loops at @xmath1 1 mk @xcite , the cool fan structures at the edges of the ars at @xmath2 1 mk @xcite . in addition to the visible loop structures , there is a substantial amount of diffuse emission in and around the active regions @xcite . diffuse emission regions may be defined as regions with no resolvable structures . they may , however , appear to be loop - like structures at higher resolution . in fact , the emissions from confined loops structures are just about 20 - 30% higher than the background / foreground diffuse emission @xcite . @xcite investigated the density and temperature structure of a limb ar . the authors reported that ar plasmas are multi - thermal and the electron densities fall off as a function of distance from the core , as was also obtained using the white light observations of the corona . ar heating has often been debated between effectively steady heating ( high frequency nanoflares ) and impulsive heating ( low frequency nanoflares ) . in the former scenario , the delay between heating events is smaller than the cooling / draining timescale of the plasma , leading to conditions that are similar to constant heating . impulsive heating , however , suggests that the delay between heating events is longer than the cooling / draining timescale of the plasma , i.e. the plasma gets some time to cool down before another heating event takes place . the properties of 1 mk warm loops seem to favor impulsive heating @xcite . however , the heating of core loops is a matter of strong debate @xcite . recent analysis suggest that active regions during the early part of their evolution seem to show an em distribution ( emd ) that is consistent with impulsive heating , while during the latter part of their evolution , the variability of the core becomes more gentle and the em distribution is more consistent with high - frequency nanoflare heating @xcite . the study of diffuse emission has not been explored in great detail . the main aim of this paper is to study and probe the heating mechanism in the diffuse part of active regions . direct observations of heating processes are not possible yet , as these events happen on scales much smaller than the resolvable limits with the available present day instrumentation @xcite . emission measure ( em ) diagnostics have been advocated ( see e.g. , * ? ? ? * ) to be one of the possible indirect modes of studying the heating mechanisms among others such as doppler shifts @xcite and the recently developed time lag analysis @xcite . here , we have employed the technique of em to study the diffuse emission in active regions . in addition , we estimate the em(t ) distribution of topologically different regions in off limb _ ar 10939 _ and _ ar 10961 _ , i.e. , warm and core loop structures and compare with the em(t ) of the diffuse emission regions with the aim to probe the thermal structure in diffuse regions and thereby their heating mechanisms . @xcite presented a revised radiometric calibration , due to the degradation of detector s response of the eis instrument over time , since the launch date of the hinode mission . the revised calibration has been accepted by the eis team and has been provided in the solarsoftware package . in the present work , we have applied the revised calibration on intensities to obtain the em and compared with that obtained using the unrevised intensities . the rest of the paper is organised as follows : in section [ obs ] , we describe the data used in this study and the reduction procedures applied ; in section [ db ] , we discuss about an instrumental effect , the diffraction bands that are observed in our datasets ; we describe our data analysis method and present our results in section [ analysis ] followed by a summary and discussion in section [ conc ] . lcccrrrr + & + ion & wavelength & @xmath3&intensity & m2 & m5 & m8 & m10 + & ( @xmath4 ) & ( k ) & correction & & & & + & & & factors & & & & + & 268.990 & 5.65 & 1.11 & 132 @xmath5 2 & 287 @xmath5 3 & 596 @xmath5 3 & 966 @xmath5 5 + & 275.350 & 5.80 & 1.14 & 110 @xmath5 1 & 215 @xmath5 2 & 426 @xmath5 2 & 631 @xmath5 4 + & 185.210 & 5.80 & 1.35 & 336 @xmath5 4 & 626 @xmath5 6 & 1325 @xmath5 8 & 2278 @xmath5 13 + & 186.600 & 5.80 & 1.39 & 215 @xmath5 3 & 404 @xmath5 4 & 895 @xmath5 5 & 1352 @xmath5 9 + & 197.860 & 5.90 & 1.16 & 128 @xmath5 1 & 176 @xmath5 1 & 251 @xmath5 1 & 247 @xmath5 2 + & 188.500 & 5.90 & 1.43 & 195 @xmath5 2 & 271 @xmath5 3 & 400 @xmath5 3 & 438 @xmath5 4 + & 174.531 & 6.05 & 1.57 & 1852 @xmath5 82 & 2193 @xmath5 97 & 2785 @xmath5 96 & 2746 @xmath5123 + & 184.537 & 6.05 & 1.35 & 794 @xmath5 6 & 910 @xmath5 7 & 1154 @xmath5 7 & 1265 @xmath5 10 + & 188.217 & 6.15 & 1.45 & 2393 @xmath5 20 & 2664 @xmath5 25 & 3207 @xmath5 25 & 3673 @xmath5 35 + & 180.401 & 6.15 & 1.47 & 1513 @xmath5 6 & 1618 @xmath5 7 & 1927 @xmath5 8 & 2406 @xmath5 11 + & 258.374 & 6.15 & 1.35 & 882 @xmath5 5 & 1033 @xmath5 5 & 1292 @xmath5 5 & 2334 @xmath5 9 + & 261.060 & 6.15 & 1.29 & 406 @xmath5 3 & 456 @xmath5 3 & 569 @xmath5 3 & 865 @xmath5 5 + & 195.120 & 6.20 & 1.00 & 1093 @xmath5 3 & 1310 @xmath5 4 & 1558 @xmath5 4 & 2054 @xmath5 6 + & 192.394 & 6.20 & 1.13 & 2781 @xmath5 4 & 3179 @xmath5 6 & 3727 @xmath5 6 & 5227 @xmath5 10 + & 202.040 & 6.25 & 1.00 & 2757 @xmath5 9 & 3216 @xmath5 10 & 3608 @xmath5 9 & 4147 @xmath5 13 + & 203.827 & 6.25 & 1.02 & 1712 @xmath5 9 & 2595 @xmath5 10 & 3623 @xmath5 10 & 9613 @xmath5 20 + & 274.204 & 6.30 & 1.10 & 1629 @xmath5 5 & 2300 @xmath5 6 & 3293 @xmath5 6 & 6290 @xmath5 11 + & 284.163 & 6.35 & 1.32 & 6527 @xmath5 16 & 11322 @xmath5 20 & 21790 @xmath5 25 & 39060 @xmath5 44 + & 256.700 & 6.40 & 1.42 & 383 @xmath5 4 & 622 @xmath5 5 & 1444 @xmath5 7 & 3234 @xmath5 13 + & 262.980 & 6.45 & 1.25 & 344 @xmath5 3 & 779 @xmath5 4 & 2539 @xmath5 6 & 4337 @xmath5 10 + & 193.870 & 6.55 & 1.04 & 18 @xmath5 1 & 94 @xmath5 1 & 435 @xmath5 2 & 541 @xmath5 3 + & 201.000 & 6.65 & 0.99 & 84 @xmath5 10 & 61 @xmath5 3 & 280 @xmath5 3 & 313 @xmath5 5 + lcccrrrr + & + ion & wavelength & @xmath3&intensity & m3 & m7 & m9 & m11 + & ( @xmath4 ) & ( k ) & correction & & & & + & & & factors & & & & + & 275.350 & 5.80 & 1.23 & 19 @xmath5 1 & 37 @xmath5 1 & 136 @xmath5 2 & 132 @xmath5 2 + & 185.210 & 5.80 & 1.35 & 43 @xmath5 2 & 104 @xmath5 4 & 341 @xmath5 5 & 578 @xmath5 7 + & 174.531 & 6.05 & 1.57 & 939 @xmath5 84 & 1171 @xmath5122 & 1569 @xmath5127 & 1988 @xmath5153 + & 184.537 & 6.05 & 1.35 & 275 @xmath5 3 & 351 @xmath5 5 & 591 @xmath5 6 & 726 @xmath5 7 + & 188.217 & 6.15 & 1.45 & 682 @xmath5 3 & 781 @xmath5 5 & 1190 @xmath5 6 & 1614 @xmath5 8 + & 195.120 & 6.20 & 1.00 & 1471 @xmath5 3 & 1724 @xmath5 4 & 2367 @xmath5 5 & 3289 @xmath5 7 + & 202.040 & 6.25 & 1.00 & 1522 @xmath5 5 & 1883 @xmath5 8 & 2331 @xmath5 9 & 2889 @xmath5 11 + & 203.827 & 6.25 & 1.02 & 574 @xmath5 4 & 920 @xmath5 7 & 1739 @xmath5 9 & 3992 @xmath5 15 + & 274.204 & 6.30 & 1.18 & 604 @xmath5 3 & 953 @xmath5 5 & 1383 @xmath5 6 & 2555 @xmath5 9 + & 284.163 & 6.35 & 1.42 & 1689 @xmath5 6 & 3240 @xmath5 11 & 5258 @xmath5 14 & 10343 @xmath5 22 + & 262.980 & 6.45 & 1.34 & 57 @xmath5 2 & 164 @xmath5 3 & 409 @xmath5 4 & 953 @xmath5 6 + for the present analysis , we have used the off - limb _ ar 10939 _ and _ ar 10961 _ datasets obtained on 2007 jan 26 and 2007 july 08 , respectively , from the extreme ultraviolet imaging spectrometer ( eis ; * ? ? ? * ; * ? ? ? * ) on board hinode . eis rastered the the _ ar 10939 _ with a 1 slit over an area of 128 @xmath6 128 , with 26.5 sec exposure at each position and with a step size of 1 . the observing study of _ ar 10939 _ was a full spectral observation . similarly , _ ar 10961 _ was rastered with a 1 slit over an area of 256 @xmath6 256 , with 15 sec exposure at each position and with a step size of 1 . this observing sequence consisted of 17 spectral windows and out of which 11 spectral lines have been used in the current study . tables [ eis_linelist_ar10939_1 ] @xmath7 [ eis_linelist_ar10961_1 ] show the list of the spectral lines used in this study , along with their wavelengths and peak formation temperatures . the peak formation temperatures were obtained from recent chianti ionization equilibrium calculations ( chianti v7.1 ; * ? ? ? * ; * ? ? ? the data are reduced using the standard procedure _ _ eis_prep.pro _ _ and are corrected for eis slit , tilt and satellite orbital variation . using _ _ _ , a single gaussian line fit is applied to the data to derive the intensity maps , except for the 268.99 , 197.86 , 188.22 , 195.12 , 203.83 , 193.87 and 201.00 lines . in the latter cases , multiple gaussian fitting is applied to subtract the blended line contributions . the 256 line is one of the eis core lines , with the lowest formation temperature available in the eis spectral range . however , especially in the case of active regions , the interpretation of this line gets complicated by blends with 256.37 , 256.41 and 256.42 @xcite . hence , we do not include the 256 line in this work . the rest of the available lines are very weak in the off - limb structures and hence could not be used . the 185.21 ( @xmath8 ) line is blended with the line which peaks in the ar core plasma and de - blending is essential to estimate the plasma parameters . in _ ar 10961 _ data , the respective line could not be de - blended because of the non availability of other line observations . while in the case of _ ar 10939 _ , de - blending is performed by estimating the contribution of to the line through observing the 195.27 line . disentangling the diffuse emission regions from the rest of the active region is a non - trivial task , which is very crucial for this work . hence it is essential to track the active region from the center of the sun to the limb , which may give some ideas regarding the evolution of different structures in the active region . the six images in fig . [ fig1 ] display the intensity images of _ ar 10961 _ obtained in the line from 2007 aug 01 to 2007 aug 08 , tracking the evolution of the active region from disk center to the limb . the spectral line , whose peak formation temperature at @xmath9 , is the strongest fe line in the eis / hinode spectrum at ar conditions and the core structures that are primarily at few million degrees can be well studied at this temperature regime . thus by tracking the particular ar core at temperatures , the diffuse regions could be confidently constrained to be well outside the core structures . the intensity corrections required for the revised radiometric calibration were obtainied using _ eis_ltds.pro _ ( available in the ssw libraries ; * ? ? ? * ) , which provides a factor by which the revised intensities differ with the unrevised ones . the revised intensities have been obtained by applying these correction factors to the initial intensities obtained from the data . from here onwards , we will be calling the initial intensities and the respective em distributions as unrevised intensities and unrevised emds . while the intensities corrected according to the revised radiometric calibration and the corresponding em distributions will be referred as revised intensities and revised emds . the reduced data of both the above mentioned off - limb active regions show diffraction bands ( fig . [ fig5 ] ) in the quiet regions beside the active regions . these bands are clearly seen in high temperature lines ( - ) , being most prominent in line ( fig.[fig5 ] ) . these are probably similar to the cross - shaped diffraction patterns seen in aia onboard sdo and trace during flares ( see e.g. , * ? ? ? * ; * ? ? ? * ) . in our cases , the scattered light forms bands rather than the crossed - spikes patterns . this is probably due to the active region being a broad structure of enhanced brightness at the off - limb when compared to the flaring regions ( young , p. , 2013 : private communication ) . the bands always trace the edges of the active regions and may arise due to the large intensity contrast produced by the active region over the off - limb background values . for _ ar 10939 _ , intensity at the successive bands falls off approximately as 5% , 3% , 1% and 0.3% of the brightest part of the ar core ( @xmath10 39000 ergs @xmath11 s@xmath12 sr@xmath12 ) at distances of approximately 10 , 20 , 30 , and 40 . some of the observed emission at the locations of the bands comes from actual on - disk solar features , so these percentages represent upper limits on the diffracted core component . we use this information later to estimate the possible contamination in the diffuse emission above the limb . these factors mentioned above would depend on how big / bright the source is in that particular raster . while the _ ar 10961 _ raster shows only a trace of such bands as the exposure time is much shorter ( 15 sec ) than the exposure time of the _ ar 10939 _ raster ( 26.5 sec ) . _ ar 10961 _ data has another active region in the full field of view and it would be inappropriate to estimate the intensities of such bands as they have contamination from near - by structures . 197@xmath13 , in order to help us in tracking any ar structure through the solar rotation . though all the ar data have been plotted with the same y - axis scale , the scalings in the x - axis are different . this is because the apparant heliographic longitude ( the solar x coordinate ) of the ar changes vastly over the solar rotation , unlike the latitude ( the solar y coordinate ) . ] the motivation behind this work is to probe the heating mechanism for the diffuse emission regions . off - limb ar data are ideal to spectroscopically study the topologically different areas within the active region , especially the diffuse emission regions , as these regions could be isolated without much contamination from the core of the active regions as well as warm 1 mk loops . a detailed description on the spectroscopic techniques for deriving the physical parameters like electron density and temperature along with the em of the emitting plasma has been given in @xcite and @xcite . emd is a function of temperature and density of the emitting plasma . the observed intensity can be correlated with the emission measure as follows . @xmath14 where , @xmath15 is the elemental abundance and @xmath16 is the contribution function containing all the relevant atomic parameters like transition probabilities , ionization fraction etc . earlier works showed that , for ars , em obeys a power law @xmath17 up to a peak near 3 mk @xcite . while plotting the em with respect to temperature on a log - log scale , @xmath18 is the slope of a best fitted straight line up to the peak emission measure . the slope @xmath18 represents the temperature distribution of multi - thermal plasma along the line - of - sight . for both the ars , _ is used to choose a sample of topologically different areas within active regions like hot core loops , comparatively cooler warm loops and the diffuse emission regions , by masking pixels in polygon mode . it is crucial to make sure that each box contains only one type of feature , i.e. , they should not have any contamination from other regions . we have used _ eis_mask_spectrum.pro _ to retrieve the spatially averaged spectra over each chosen pixel group ( box ) . abs(xx1[1:29 ] ) estimation of emission measure from averaged spectra is non trivial . as the masked regions are selected manually , the box sizes are not the same for all the regions . it is important to understand the effect of such varying box size in the estimation of em , in order to compare our results with the other similar works . hence , we choose a set of seven randomly oriented boxes each having a sub - box over _ ar 10961_. thus , in total there are 14 regions as shown in fig . [ fig2 ] ( left ) for the analysis and their respective em plots are shown in fig . [ fig2 ] ( right ) . the em plot clearly shows that choosing different boxes with different sizes , in a particular area in the field of view , does not vastly change the em . the color scheme ( loadct , 5 ) , which we have chosen for images ( fig . [ fig2 ] ( left ) and fig . [ fig3 ] & [ fig4 ] ( top right ) ) , is found to be the best in representing the ar . the transition of color from blue to yellow distinctly represents the topologically different regions in the ar from the diffuse parts on the outskirts of the ar towards the core , respectively . [ fig3 ] ( top panel ) and [ fig4 ] ( top panel ) display the intensity images for _ ar 10939 _ and _ ar 10961 _ , respectively , obtained in ( left ) , ( middle ) and ( right ) lines . intensity image of _ ar 10939 _ ( fig . [ fig3 ] : top right ) are over - plotted with the 11 masked regions ( m1 , m2 , ... , m11 ) that are chosen for the estimation of emission measure . we have divided these regions into different sets based on their locations . set 1 is comprised of the first four regions , namely m1 , m2 , m3 & m4 , which are selected in the diffuse emission region well above the ar core . set 2 is comprised of regions m5 , m6 , & m7 , which are located at the boundary between the core and the diffuse emission region . set 3 is comprised of regions m8 , m9 , m10 and m11 being in the core of the active region . similarly , the intensity image of _ ar 10961 _ ( fig . [ fig4 ] : top right ) is over - plotted with the selected 12 masked regions ( m1 , m2 , ... , m12 ) . these regions are again divided into four groups following the same criteria as discussed above in the case of _ ar 10939_. initially , averaged spectra have been obtained from all the masked regions from _ ar 10939 _ and _ ar 10961 _ in all the spectral windows mentioned in table [ eis_linelist_ar10939_1 ] and table [ eis_linelist_ar10961_1 ] , respectively . spectral lines are fitted with a modified version of _ eis_autofit.pro _ to read the output structures from _ eis_mask_spectrum.pro_. for _ ar 10939 _ , estimated intensities for a sample masked region from each set ( m2 - set1 , m5 - set2 , m8 & m10 - set3 ) are given in the table [ eis_linelist_ar10939_1 ] and the intensities of the rest of the masked regions are given in table [ eis_linelist_ar10939_2 ] . similarly for _ ar 10961 _ , estimated intensities for a sample region from each set of masked regions ( m3 - set1 , m7 - set2 , m9 & m11 - set3 ) are given in the table [ eis_linelist_ar10961_1 ] and the intensities of the rest are given in table [ eis_linelist_ar10961_2 ] . the masked regions discussed in the table [ eis_linelist_ar10939_1 ] and the table [ eis_linelist_ar10961_1 ] fall as a strip , with which variations of spectroscopic parameters as a function of distance from the core of the ar can be explored . the em is derived using the method discussed by @xcite . even though there are many true inversion dem codes currently available , this method is objective and provides consistent results when compared to the other methods . also , the results obtained using pottasch method are similar to those from the other dem inversion codes . in order to derive the emission measure , it is necessary to provide electron density as an input parameter . therefore , we have used ( 202.04 @xmath19203.83 ) lines to derive the densities for each region , as they are one of the best coronal density diagnostics available to eis . table [ table_dens ] shows the estimated densities of all the masked regions choosen over both the ars . for each spectral line , the em at the peak formation temperature , @xmath20 , is approximated by assuming that the contribution function is constant and equal to its average value over the temperature range @xmath210.15 and zero at all other temperatures . we use the ionization equilibrium values as given in chianti ( v7.1 ; * ? ? ? * ; * ? ? ? * ) and consider both photospheric @xcite and coronal @xcite abundances . rrrrrr & & + & masked & density & masked & density + & regions & ( n@xmath22 @xmath23 ) & regions&(@xmath24 @xmath23 ) + set1 & m1 & 8.0e+08 & m1 & 5.9e+08 + & m2 & 8.6e+08 & m2 & 5.3e+08 + & m3 & 8.5e+08 & m3 & 5.4e+08 + & m4 & 7.1e+08 & m4 & 5.7e+08 + & & & m5 & 5.5e+08 + set2 & m5 & 1.2e+09 & m6 & 7.8e+08 + & m6 & 1.1e+09 & m7 & 7.1e+08 + & m7 & 1.2e+09 & m8 & 7.5e+08 + set3 & m8 & 1.5e+09 & m9 & 1.1e+09 + & m9 & 1.5e+09 & m10 & 1.3e+09 + & m10 & 6.5e+09 & m11 & 2.4e+09 + & m11 & 4.1e+09 & m12 & 2.4e+09 + as listed in tables [ eis_linelist_ar10939_1],[eis_linelist_ar10961_1 ] , [ eis_linelist_ar10939_2 ] and [ eis_linelist_ar10961_2 ] , the intensities in each spectral line show an increasing trend from the diffuse emission regions ( set1 ) towards the core of the ar ( set3 ) . the electron densities also show an increasing trend from set 1 to set 3 . in each set , the derived electron densities are remarkably similar among the choosen masked regions ( table [ table_dens ] ) , except for the m10 @xmath7 m11 ( set3 ) . the regions show higher densities when compared to m8 @xmath7 m9 ( set3 ) and also vary among themselves , which could be because of the possible contaminations from the underlying moss regions . these trends observed in intensities and densities suggest that similar structures were chosen in each set and also discriminate the regions chosen among different sets to be topologically different . the middle rows ( row 2 and 3 ) in figure [ fig3 ] display the em distribution for all the different sets of masked regions chosen over _ ar 10939 _ , estimated with unrevised and revised intensities , respectively . in both the cases , the em distribution for sets 2 and 3 are shown in the middle and right panels of rows 2 and 3 . the emds belonging to set3 are plotted with two different symbols . the regions m8 and m9 are plotted with plusses and m10 and m11 are plotted with triangles . this is to make a distinction between core regions without and with possible moss regions contamination . though almost all the curves of set2 and set3 show very similar trend , the emds for all the regions in set3 lie above those for set2 . in addition , the peaks of the em curves for all the regions in set3 are at larger temperatures than those of set2 regions . the peak of the emission measure for set3 mostly lies at around @xmath25 . however , the em distribution for regions m10 and m11 in set3 ( plotted with triangles ) also shows a definite peak at @xmath0 which is characteristic of moss regions @xcite . in comparison , the em distribution of set2 peaks at a lower temperature of @xmath26 . the revised radiometric calibration broadens the peak of two ( m5 @xmath7 m6 ) out of the three masked regions choosen in this set , with the em decreasing more slowly with temperature in the range @xmath27 . while , the third one ( m7 ) which shows a double peak with the original calibration , namely at @xmath0 and @xmath28 , shows only the second peak at @xmath28 prominently with the revised calibration . we have studied the slopes of the em and hot to warm ratios for the regions in set2 and set3 . the ratio of hot to warm emission is determined by taking the ratio between the peak em and the em obtained at and . the slope , however , is obtained between the temperature corresponding to peak em and the em of at @xmath29 . the slope of unrevised em for sets 2 & 3 ranges between 2.3 @xmath7 4.2 and the ratios of hot to warm emission varies between 3.8 @xmath7 7.3 when warm emission is considered at @xmath29 . these ratios increase to the range between 9 @xmath7 28 when considering as warm emission . the revised calibration makes the slope comparatively shallower in the range of 1.5 @xmath7 3 , with the hot to emission varying between 2.8 @xmath7 5.4 , and the hot to varying between 8 @xmath7 24 . + + + + the left panels of row 2 @xmath7 3 in figure [ fig3 ] show the emission measure distribution of diffuse regions classified as set1 , comprising m1 , m2 , m3 & m4 . both em distribution curves ( unrevised and revised ) for all four regions are strikingly similar and show a definite peak at @xmath0 , i.e. , at the peak formation temperature of , except that the peak of revised emds is broader at @xmath7 temperatures . for each of these curves , the estimated value of emd slopes , @xmath18 , along with the emission measure ratio of the hot plasma to the warm plasma ( and ) are given in the table [ table_slope_emratio ] . the slopes are computed between @xmath0 and @xmath29 . the active region spectra for _ ar 10939 _ are full spectral scans , and the obtained emd provide a complete picture of the thermal structure of the regions . therefore , we conclude that the peak of the em distribution for diffuse regions lies at @xmath30 . the masked regions choosen in the sets 1 , 2 , and 3 ( m8@xmath7m9 ) are located approximately 50 `` , 25 '' , 10 " from the brightest part of the core ( where set3 : m10@xmath7m11 fall ) . using the results of section 3 , we estimate a maximum contamination from diffracted core emission at these locations in to be less than 0.3% , 2% , and 5% respectively . this is about 117 , 780 and 1950 ergs @xmath11 s@xmath12 sr@xmath12 , which is less than 2% , 7% , and 9% of the actual observed intensity at these locations . therefore , it is safe to conclude that the contamination from the diffracted emission does not significantly impact our em measurements . if one assumes that the diffuse emission seen above the core in set1 is equivalent to the diffuse emission present along the line of sight in the studied core regions in set3 , i.e. , equivalent to the foreground / background emission along the line of sight , then it is possible to subtract the diffuse contribution from set3 to obtain the emission from the core plasma itself . when we do this using an averaged emd from all the masked regions in set1 , the em slopes of the core regions ( set3 ) increase by 0.64 to 0.94 , i.e. , the distributions become steeper . the bottom row ( row 4 ) in figure [ fig3 ] shows the best linear fits to a sample diffuse region ( left ) , a core region ( middle ) and the same core region after background subtraction ( right ) , with the corresponding slopes . the core observations may also be contaminated by warm loops in the foreground and background , so this must be considered only a partial correction . because of their characteristic emd peaking at , slope corrections using diffuse region emd are expected to make the core region distribution steeper . we caution , however , that the magnitude of the correction is very uncertain . both the emissivity and the line - of - sight depth of the diffuse plasma could be considerably different at the altitude of set3 than at the altitude of set1 . ar core emds ( set 3 ) show a clear rollover at about @xmath31 and this is most likely due to the contamination from the underlying moss regions , whose emds are known to peak at @xmath0 @xcite . the emd slopes would get steeper , by about a factor 0.2 ( set 3a ) and 0.8 ( set 3b ) , when estimated with respect to this rollover peak , rather than with respect to the em peak . since we are interested primarily in the coronal loop top emission and not the emission from footpoint moss regions , the slopes presented here are estimated with respect to the peak em . @xcite studied _ ar 10961 _ when it was near disk center , when full spectral scan observations were available . their goal was to measure the em slope of inter - moss regions ( i.e. , of the ar core ) . because of the on - disk observing geometry , lines - of - sight through the inter - moss regions included contributions from higher altitude diffuse emission above the core . to estimate to the magnitude of these contributions , they examined additional lines - of - sight outside the core and made adjustments to account for gravitational stratification . the diffuse emission was thereby estimated to contribute between 5 and 40@xmath32 of the total em at @xmath33 along the line of sight through the core . our limb measurements of this same active region give compatible results . [ fig4 ] shows that , at this temperature , the diffuse emission ( set1 ) is roughly 20@xmath32 of the core emission ( set3 ) . the boundary region emission ( set2 ) is roughly 30@xmath32 of the core emission . this suggests that the method used by @xcite to remove the diffuse component from their core measurements is reasonable . they estimated a potential error in the em slope of less than 0.1 . we caution that the observing geometries are completely different for these two studies and that a horizontal integration through a point above the limb may be much different from vertical integration through the same point . in addition , we could not study the effect of gravitational stratification at the height of the ar core , even by assuming the diffuse emission to be unresolved loop structures . this is because , measuring the loop length is not possible in our case taking the diffuse nature of the region into account . hence our results may not be comparable with the results of @xcite . our earlier on - disk study of _ ar 10961 _ @xcite benefited from a full spectral scan , which we do not have for our present limb study of this region . the lack of spectral lines hotter than @xmath34 prevents us from fully characterizing the core plasma , but we can obtain a rather complete picture of the diffuse and boundary region plasma . similar to the analysis performed for _ ar 10939 _ , we have chosen different regions ( shown in the top right panel fig . [ fig4 ] ) and have grouped them in different sets , namely set1 , set2 and set3 . the unrevised ( row 2 ) and the revised ( row 3 ) emds for set 2 and set 3 are plotted in the middle and right panels , and those for set 1 are plotted in the left panels in fig . similar to active region ar 10939 , we have plotted the em for m9 and m10 with pluses and regions m11 and m12 with triangles to differentiate between with regions with and without possible moss contribution . the emds for set2 and set3 appear very similar to those for _ ar 10939 _ , except that we do not find the peak above @xmath34 . this is because we do not have spectral lines at those temperatures ( above temperature ) in this study . ( @xcite found that the core em peaked at @xmath25 when observed on the disk . ) therefore , these data are not adequate for studying the active region core heating and also to probe the variation of slopes of ar core with background subtractions using an averaged diffuse region emd , as done with the _ ar 10939_. however , this region is still useful for studying the diffuse region due to the fact that the em for diffuse region peaks at @xmath0 as established earlier for _ ar 10939 _ using the full spectral scan ( assuming that the active regions are similar ) . + + + the bottom left plot in fig . [ fig4 ] shows the em distribution for diffuse regions named as set1 , comprising m1 , m2 , m3 , m4 & m5 . the peak of emd is again at at @xmath0 . we determined the slope of em curves ( between @xmath0 and @xmath29 ) for all these five regions in set1 and studied the ratio of hot ( ) to warm ( as well as ) emission . the values are provided in table [ table_slope_emratio ] . fig . [ fig4 ] shows that the em curves of set1 ( high altitude diffuse emission ) differ from the em curves of set3 ( core ) of _ ar 10961 _ by @xmath35log em @xmath1 0.3 - 0.6 for @xmath366.2 , which corresponds to roughly a factor of 2 - 3 . similarly , em curves of set1 differ from set3 of _ ar 10939 _ by a factor of 2 - 4 . the em distribution itself is comparable or slightly larger for _ ar 10939 _ ( fig . [ fig3 ] ) when compared with _ ar 10961_. [ table_slope_emratio ] ccccccccccc & & & & @xcite + & masked & slope & / & / & & slope & / & / + & regions & ( @xmath37 ) & & & & ( @xmath37 ) & & + _ ar 10939 _ & m1 & 2.90 & 3.1 & 19.7 & & 2.04 & 2.3 & 17.2 + & m2 & 3.08 & 3.2 & 21.6 & & 2.22 & 2.4 & 18.8 + & m3 & 2.95 & 3.3 & 9.5 & & 2.08 & 2.5 & 8.3 + & m4 & 2.87 & 3.2 & 14.1 & & 2.01 & 2.4 & 12.3 + _ ar 10961 _ & m1 & 2.31 & 2.5 & 15.4 & & 1.44 & 1.8 & 12.5 + & m2 & 2.79 & 2.3 & 37.9 & & 1.92 & 2.4 & 30.8 + & m3 & 3.45 & 4.4 & 55.5 & & 2.58 & 3.2 & 45.1 + & m4 & 2.31 & 2.6 & 13.0 & & 1.45 & 1.9 & 10.6 + & m5 & 4.15 & 6.2 & 108.2 & & 3.29 & 4.6 & 87.9 + it has now been established that there is a significant amount of plasma at higher coronal temperatures which does not appear in any structured form , such as distinct loops , when viewed with current instrumentation . rather , the plasma appears as diffuse emission spreading over a large area . the emission from coronal loops is only about 20 - 30% higher than this fuzzy emission ( see e.g. , * ? ? ? * ; * ? ? ? therefore , just explaining the heating of structured loops is not enough for understanding the heating of solar active regions in particular and the corona in general . this study aims at understanding the characteristics of these diffuse regions and their possible heating scenario . keeping this science goal in mind , we have probed two active regions , _ ar 10939 _ and _ ar 10961 _ , observed observed off limb on 2007 jan 26 and 2007 july 08 , respectively , using emission measure diagnostics techniques . we have distinguished the diffuse emission region from the rest of the ar by tracking the ar from the disk center to the off - limb , which provided us with a better constraint on selecting topologically different areas within the ar . we have chosen different sets of masked regions , designated 1 , 2 @xmath7 3 , to isolate , respectively , diffuse emission regions , the boundary between the diffuse regions and the core , and the core itself . emds have been obtained for each of the masked regions using the pottasch method ( see e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? we also compare the emd of such topologically different regions . [ table_avgslope ] ccccc masked regions & + & & + & no bg & bg subtracted & no bg & bg subtracted + set1 & 2.95 & - & 2.08 & - + set2 & 2.97 & - & 2.42 & - + set3a & 2.42 & 3.24 & 2.26 & 3.25 + set3b & 3.52 & 4.29 & 2.84 & 3.55 + in emd , the slope @xmath18 and the emission measure ratio of hot and warm plasma may indicate the possible mechanisms responsible for heating the plasma . for example , a shallower slope suggests that along a given line of sight , there are comparable amounts of hot ( @xmath1 few mk ) and warm ( @xmath1 1 mk ) plasma co - existing , implying a low frequency nanoflare heating , often called simply impulsive heating . in comparison , a steeper slope indicates much less warm plasma than hot plasma , implying a high frequency nanoflare heating , often referred to as steady heating , where the plasma is maintained at a given temperature . much importance has been given to the estimation of the slope @xmath18 for active region cores and has always been debated between low and high frequency nanoflare heating @xcite . while the diffuse emission associated with the ar has been poorly understood . @xcite showed that the uncertainty in the estimation of slope can be @xmath5 1 and depends on both the atomic physics uncertainties and the number of spectral lines used to constrain the distribution . this may represent the upper limit of the errors involved . @xcite suggested the total uncertainty involved in the em estimation to be @xmath1 50 % , which includes the radiometric and line fitting errors associated with intensities , and the uncertainties associated with the atomic physics and the pottasch method . the estimated power law slope @xmath18 and the emission measure ratios of all the diffuse emission regions studied here are given in the table [ table_slope_emratio ] . the emd of all these regions show that the emission measure peaks at @xmath0 . the distributions show a monotonic increase from @xmath8 to @xmath0 and then starts to decrease at higher temperatures . in general , 8 out of the 9 diffuse emission regions analysed here showed a power law slope in between 2.3 @xmath7 3.5 , except for the region m5 from _ ar 10961 _ , which shows a much higher slope ( @xmath18=4.18 ) and em ratios ( 108 when taken with and 6.2 when taken with ) than the others . implication of the revised radiometric calibration vastly changes the earlier results . it produced comparatively shallower slopes in the range of 1.4 & 2.6 and lowers the hot to warm plasma ratio . the emd of the core regions , which can only be studied for the _ ar 10939 _ , shows consistency with previous studies . the peak of the em is found to be around @xmath38 which is also consistent with the em peak obtained in previous studies @xcite . a sample of masked regions chosen from both the active regions , that are discussed in tables [ eis_linelist_ar10939_1 ] @xmath7 [ eis_linelist_ar10961_1 ] , can be used to probe the variation of spectroscopic parameters like intensity , density etc . , along with the emd slope ( @xmath18 ) as a function of distance from the ar core which is just over the limb . these samples comprise of m2 ( set1 ) , m5 ( set2 ) , m8 @xmath7 m10 ( set3 ) for the _ ar 10939 _ and m3 ( set1 ) , m7 ( set2 ) , m9 @xmath7 m11 ( set3 ) for the _ ar 10961 _ , that fall as a strip which includes both the core plasma and the diffuse emission plasma above the core . tables [ eis_linelist_ar10939_1 ] @xmath7 [ eis_linelist_ar10961_1 ] clearly show that the intensities and densities decrease from the core towards the diffuse regions , as a function of distance from the limb , in agreement with @xcite . the power law slope of those sample masked regions from the _ ar 10939 _ are 3.0 ( m2 ) , 3.2 ( m5 ) , 2.4 ( m8 ) @xmath7 2.7 ( m10 ) . similarly , the slope of sample masked regions from the _ ar 10961 _ are 3.4 ( m3 ) , 3.5 ( m7 ) , 3.4 ( m9 ) @xmath7 4.0 ( m11 ) . we , however , note that the slopes determined in the core of the active region for ar 10961 does not represent the true emd slope as a full spectrum for this ar was not available . the highestest temperature line in the available raster is , whereas the core emission peaks higher at @xmath34 . note also that these slopes were measured with the uncorrected data . slopes measured with the revised calibration are smaller by about 0.9 , at least for the diffuse regions where we have studied the difference ( table [ table_slope_emratio ] ) . based on the uncorrected data , we conclude that the slope does not have an obvious dependence on distance from solar surface . table [ table_avgslope ] shows the average slopes of emds , with and without background / foreground subtraction , for the _ ar 10939_. a background / foreground subtraction from the unrevised and revised core emd using , an averaged unrevised and revised emd of the studied diffuse regions respectively , steepened the slope by @xmath1 0.8 @xmath7 0.85 . we stress that the background / foreground subtraction is uncertain due to the unknown variation of diffuse emission with altitude ( both its horizontal line - of - sight depth and its emissivity ) . the difference in the slopes between the two ars can be attributed to the fact that they are two different ars at different instances in their lifetime . at the time of observations , _ ar 10939 _ and _ ar 10961 _ are 6 days and 13 days old ( solar monitor ) . the heating properties of active regions ( low versus high frequency nanoflares ) have been shown to depend on the age of the ar @xcite and the magnetic field strength of the ar @xcite . significant errors can come from the difference in the number of spectral lines constraining the distribution @xcite . the slopes of the emd curves and the hot to warm ratios for diffuse regions obtained in this study is within the range of values obtained for active regions cores . these results suggest that the diffuse emission regions are heated and maintained in similar way as the hot emission in the core of active regions . the slope is similar among the different parts of the active regions ( table [ table_avgslope ] ) . the em slopes of the diffuse emission outside the core are generally in the range of 2 . thus , they are consistent with both low frequency and high frequency heating , especially considering the uncertainties in the measurements . however , more detailed analysis with many active regions and theoretical modelling should be performed before making any definite conclusions . the results obtained in this study provide further constraints on the properties of diffuse emission in active regions . hinode is a japanese mission developed and launched by isas / jaxa , collaborating with naoj as a domestic partner , nasa and stfc ( uk ) as international partners . sci- entific operation of the hinode mission is conducted by the hinode science team organized at isas / jaxa . this team mainly consists of scientists from institutes in the partner countries . support for the post - launch operation is provided by jaxa and naoj ( japan ) , stfc ( u.k . ) , nasa , esa , and nsc ( norway ) . chianti is a collaborative project involving re- searchers at nrl ( usa ) ral ( uk ) , and the universities of cambridge ( uk ) , george mason ( usa ) , and florence ( italy ) . the work of jak was funded by the nasa supporting research program . acknowledges s.t.f.c . we acknowledge useful discussions at the issi on active region heating . we thank dr . peter young for his valuable comments and discussions . we thank the anonymous referee for thoughtful comments and a careful reading . bradshaw , s. j. , klimchuk , j. a. , and reep , j. w. 2012 , , 758 , 53 brooks , d. h. and warren , h. p. 2009 , , 703 , 10 culhane , j. l. , et al . 2007 , , 243 , 19 culhane , j. l. , et al . 2005 , advances in space research , 36 , 1494 dadashi , n. ; teriaca , l. ; tripathi , d. ; solanki , s. k. ; wiegelmann , t. 2012 , , 548 , 115 del zanna , g. , mason , h. e. 2003 , , 406 , 1089 del zanna , g. 2013 , , 555 , 47 del zanna , g. , tripathi . , d , subramanian , s. , odwyer and mason , h. e. 2013 , , submitted dere , k. p. 1982 , , 77 , 77 dere , k. p. and mason , h. e. 1993 , , 144 , 217 dere , k. p , et al . 1997 , a & a supplement series , 125 , 149 feldman , u. 1992 , physica scripta , 46 , 202 grevesse , n. and sauval , a. j. 1998 , space science reviews , 85 , 161 guennou , c. , auchre , f. , klimchuk , j. a. , bocchialini , k. and parenti , s. 2013 , , 774 , article i d . 31 , 13 jordan , c. , wilson , r. , 1971 , assl vol . 27 : physics of the solar corona , 219 klimchuk , j. a. 2006 , , 234 , 41 klimchuk , j. a. 2009 , the second hinode science meeting : beyond discovery - toward understanding asp conference series , vol . 415 . edited by b. lites san francisco : astronomical society of the pacific , 2009 , p.221 landi , e. ; young , p. r. ; dere , k. p. ; del zanna , g. ; mason , h. e. 2013 , , 763 , 9 lang , james ; kent , barry j. ; paustian . 2006 , applied optics ip , 45 , issue 34 , 8689 lin , a. c. ; nightingale , r. w. ; tarbell , t. d. 2001 , , 198 , 385 mason , h. e. and fossi , b. c. monsignori 1994 , , issn 0935 - 4956 , 6 , 123 odwyer , b. , del zanna , g. , mason , h. e. , sterling , a. c. , tripathi , d. , young , p. r. 2011 , , 525 , a137 pottasch , s. r. 1963 , , 137 , 945 raftery , c. l. ; krucker , s. and lin , r. p. 2011 , , 743 , l27 reale , f. 2010 , living rev . solar phys , 7 , 5 reep , j. w. , bradshaw , s. j. , and klimchuk , j. a. 2013 , , 764 , 193 schrijver , c. j. et al . 1999 , , 187 , 261 schmelz , j. t. and pathak , s. 2012 , , 756 , 126 tripathi , d. , mason , h. e. , dwivedi , b. n. , del zanna , g. , and young , p. r. 2009 , , 694 , 1256 tripathi , d. , mason , h. e. , del zanna , g. and young , p. r. 2010 , , 518 , 42 tripathi , d. , mason , h. e. , klimchuk , j. a. 2010 , , 723 , 713 tripathi , d. , klimchuk , j. a. , mason , h. e. 2011 , , 740 , 111 tripathi , d. , mason , h. e. , klimchuk , j. a 2012 , , 753 , 37 ugarte - urra , i. , warren , h. p. , brooks , d. h. 2009 , , 695 , 642 ugarte - urra , i. and warren , h. p. 2012 , , 761 , 21 viall , n. m. , and klimchuk , j. a. 2011 , , 738 , 24 viall , n. m. , and klimchuk , j. a. 2012 , , 753 , 35 viall , n. m. , and klimchuk , j. a. 2013 , , 771 , 115 warren , h. p. , winebarger , a. r. , and mariska , j. t. 2003 , , 593 , 1174 warren , h. p. , ugarte_urra , i. , young , p. r. , and stenborg , g. 2011 , , 727 , 58 warren , h. p. , winebarger , a. r. , brooks , d. h. 2012 , , 759 , 141 winebarger , a. r. , warren , h. , van ballegooijen , a. , deluca , e. , golub , l. 2002 , , 567 , 89 winebarger , a. r. , schmelz , j. t. ; warren , h. p. ; saar , s. h. ; kashyap , v. l. 2011 , , 740 , 2 winebarger , amy r. , tripathi , d. , mason , h. e. ; del zanna , g. 2013 , , 767 , 107 young , p. r. , et al . 2007 , , 59 , 857 young , p. r. , odwyer , b. , mason , h. e. 2012 , , 744 , 14 [ ! h ] & 127@xmath5 2 & 291@xmath5 2 & 133@xmath5 2 & 312@xmath5 3 & 169@xmath5 2 & 537@xmath5 3 & 407@xmath5 4 + & 111@xmath5 1 & 229@xmath5 2 & 109@xmath5 1 & 241@xmath5 2 & 125@xmath5 1 & 383@xmath5 2 & 279@xmath5 2 + & 274@xmath5 3 & 522@xmath5 4 & 304@xmath5 4 & 707@xmath5 6 & 552@xmath5 6 & 1267@xmath5 8 & 1253@xmath5 11 + & 182@xmath5 2 & 362@xmath5 3 & 196@xmath5 3 & 451@xmath5 4 & 276@xmath5 4 & 789@xmath5 5 & 665@xmath5 7 + & 139@xmath5 1 & 159@xmath5 1 & 90@xmath5 2 & 157@xmath5 1 & 113@xmath5 1 & 196@xmath5 1 & 167@xmath5 2 + & 199@xmath5 2 & 244@xmath5 2 & 139@xmath5 1 & 250@xmath5 2 & 172@xmath5 2 & 326@xmath5 3 & 292@xmath5 4 + & 1937@xmath5 76 & 1861@xmath5 71 & 1464@xmath5 80 & 2076@xmath5 85 & 2136@xmath5 99 & 2726@xmath5102 & 3043@xmath5130 + & 761@xmath5 5 & 699@xmath5 5 & 515@xmath5 5 & 809@xmath5 6 & 737@xmath5 6 & 1056@xmath5 7 & 1186@xmath5 10 + & 2315@xmath5 18 & 2139@xmath5 17 & 1715@xmath5 18 & 2526@xmath5 22 & 2433@xmath5 23 & 3006@xmath5 25 & 3894@xmath5 36 + & 1401@xmath5 5 & 1294@xmath5 4 & 1035@xmath5 5 & 1474@xmath5 7 & 1449@xmath5 7 & 1817@xmath5 8 & 2585@xmath5 12 + & 838@xmath5 5 & 682@xmath5 4 & 490@xmath5 4 & 855@xmath5 4 & 908@xmath5 5 & 1198@xmath5 5 & 2384@xmath5 9 + & 392@xmath5 3 & 309@xmath5 3 & 235@xmath5 3 & 384@xmath5 3 & 393@xmath5 3 & 522@xmath5 3 & 921@xmath5 5 + & 1031@xmath5 3 & 925@xmath5 3 & 708@xmath5 3 & 1135@xmath5 3 & 1118@xmath5 3 & 1394@xmath5 4 & 2374@xmath5 6 + & 2615@xmath5 4 & 2477@xmath5 4 & 1923@xmath5 4 & 2949@xmath5 5 & 2861@xmath5 6 & 3561@xmath5 6 & 5508@xmath5 10 + & 2622@xmath5 8 & 2524@xmath5 7 & 1951@xmath5 8 & 2933@xmath5 9 & 2863@xmath5 9 & 3380@xmath5 9 & 4775@xmath5 14 + & 1507@xmath5 7 & 1546@xmath5 7 & 1001@xmath5 7 & 2362@xmath5 9 & 2334@xmath5 9 & 3485@xmath5 11 & 9171@xmath5 20 + & 1195@xmath5 4 & 1237@xmath5 4 & 894@xmath5 4 & 1881@xmath5 5 & 2069@xmath5 6 & 3031@xmath5 7 & 6578@xmath5 12 + & 1449@xmath5 4 & 1529@xmath5 4 & 1131@xmath5 5 & 2225@xmath5 5 & 2342@xmath5 6 & 3219@xmath5 6 & 5863@xmath5 11 + & 6262@xmath5 14 & 6346@xmath5 14 & 5223@xmath5 15 & 11958@xmath5 19 & 12059@xmath5 21 & 18695@xmath5 25 & 24015@xmath5 36 + & 330@xmath5 4 & 294@xmath5 3 & 275@xmath5 4 & 600@xmath5 5 & 987@xmath5 6 & 1234@xmath5 7 & 2095@xmath5 11 + & 277@xmath5 2 & 301@xmath5 2 & 287@xmath5 2 & 825@xmath5 4 & 959@xmath5 4 & 1814@xmath5 5 & 2380@xmath5 7 + & 15@xmath5 1 & 13@xmath5 0 & 15@xmath5 1 & 81@xmath5 1 & 73@xmath5 1 & 206@xmath5 2 & 203@xmath5 2 + & 83@xmath5 10 & 93@xmath5 10 & 82@xmath5 13 & 52@xmath5 2 & 35@xmath5 2 & 104@xmath5 3 & 114@xmath5 4 + & 62@xmath5 1 & 25@xmath5 1 & 85@xmath5 1 & 11@xmath5 1 & 121@xmath5 2 & 50@xmath5 1 & 126@xmath5 2 & 153@xmath5 3 + & 121@xmath5 3 & 49@xmath5 2 & 158@xmath5 3 & 29@xmath5 3 & 258@xmath5 5 & 141@xmath5 4 & 333@xmath5 5 & 543@xmath5 10 + & 1292@xmath5123 & 1139@xmath5 94 & 1568@xmath5102 & 817@xmath5106 & 1769@xmath5129 & 1380@xmath5118 & 2036@xmath5118 & 2535@xmath5200 + & 434@xmath5 5 & 324@xmath5 4 & 481@xmath5 4 & 219@xmath5 3 & 580@xmath5 6 & 446@xmath5 5 & 691@xmath5 6 & 1016@xmath5 10 + & 767@xmath5 5 & 709@xmath5 4 & 924@xmath5 4 & 689@xmath5 4 & 974@xmath5 5 & 978@xmath5 5 & 1517@xmath5 6 & 2097@xmath5 12 + & 1389@xmath5 3 & 1397@xmath5 3 & 1619@xmath5 3 & 1599@xmath5 3 & 1865@xmath5 4 & 1921@xmath5 4 & 2793@xmath5 5 & 3681@xmath5 9 + & 1317@xmath5 7 & 1351@xmath5 6 & 1552@xmath5 6 & 1724@xmath5 7 & 1819@xmath5 9 & 1945@xmath5 8 & 2662@xmath5 9 & 3117@xmath5 15 + & 541@xmath5 6 & 502@xmath5 4 & 610@xmath5 5 & 655@xmath5 5 & 985@xmath5 8 & 1008@xmath5 7 & 2229@xmath5 9 & 4356@xmath5 20 + & 430@xmath5 3 & 390@xmath5 2 & 456@xmath5 2 & 495@xmath5 3 & 785@xmath5 4 & 789@xmath5 3 & 1623@xmath5 5 & 3112@xmath5 10 + & 528@xmath5 4 & 499@xmath5 3 & 591@xmath5 4 & 621@xmath5 4 & 917@xmath5 6 & 948@xmath5 5 & 1631@xmath5 6 & 2614@xmath5 11 + & 1543@xmath5 7 & 1327@xmath5 6 & 1434@xmath5 6 & 1447@xmath5 7 & 3300@xmath5 12 & 3193@xmath5 10 & 5219@xmath513 & 8706@xmath5 26 + & 57@xmath5 3 & 41@xmath5 2 & 52@xmath5 2 & 41@xmath5 2 & 224@xmath5 4 & 177@xmath5 3 & 402@xmath5 4 & 723@xmath5 7 +

the question for linear stability of spatially periodic waves for the boussinesq equation ( the cases ) and the klein - gordon - zakharov system is considered . for a wide class of solutions , we completely and explicitly characterize their linear stability ( instability respectively ) , when the perturbations are taken with the same period . in particular , our results allow us to completely recover the linear stability results , in the limit , for the whole line case .

1202.2133

in this paper we will be interested in the stability of spatially periodic waves for certain models , which involve second derivative in time . our interest will be mainly in two pde - the boussinesq equation and the klein - gordon - zakharov system , although the methods that we develop here will certainly find applications in other related models . the cauchy problem for the boussinesq equation , with periodic boundary conditions , is @xmath3,\ ] ] where @xmath4 will for the most part be @xmath5 . this is a model that was derived by boussinesq , @xcite , for @xmath6 , but was subsequently studied by many authors , both in the periodic and whole line context . we now review the current results regarding the well - posedness properties of the boussinesq equation . while we have a very satisfactory theory for the local solutions , see below , the the global well - posedness does not hold . more precisely , even if one requires smooth compactly supported data , the solutions may develop singularities in finite time , @xcite . this makes the stability questions , which is the main subject of this article even more relevant and interesting . in the whole line scenario , local well - posedness was established by bona and sachs , in the sobolev spaces @xmath7 , @xcite . further contributions were made by tsutsumi and mathashi , @xcite , linares , @xcite ( who also showed global existence for small data ) . farah , @xcite has shown well - posedness in @xmath8 , when @xmath9 and the space @xmath10 is defined via @xmath11 . kishimoto and tsugava , @xcite have finally shown well - posedness for all @xmath12 , which is likely to be sharp . regarding the case of periodic boundary conditions , fang and grillakis , @xcite who have established local well - posedness in @xmath13 , @xmath14 ( when @xmath15 in ) . this result was improved for @xmath6 to @xmath9 by farah and scialom , @xcite . oh and stefanov have recently shown local well - posedness in @xmath13 , @xmath16 , @xcite . our other main object of investigation will be the klein - gordon - zakharov system , which is given by in front of the non - linear term @xmath17 is non - standard , but rather adopted for convenience of our presentation . in particular , it helps create a self - adjoint linearized operator , which otherwise can be achieved via a simple change of the time variable . ] @xmath18\\ n_{tt}- n_{xx}={\frac{1}{2}}(|u|^2)_{xx } \\ \end{array } \right.\ ] ] this system describes the interaction of a langmuir wave and an ion sound wave in plasma . more precisely , @xmath19 is the fast scale component of the electric field , whereas @xmath20 denotes the deviation of ion density , @xcite . the system is locally well - posed in various function spaces , ( see ginibre - tsutsumi - velo , @xcite and ozawa - tsutaya - tsutsumi , @xcite ) . in our previous paper , @xcite , we have shown that the cauchy problem is locally well - posed ( both in priodic and whole line contexts ) in @xmath21 , whenever @xmath22 . in @xcite and @xcite , the authors have shown that small data produces solutions that persist globally , whereas large solutions are generally expected to blow up in finite time . the stability of periodic traveling waves have been studied extensively in the last decade . the nonlinear stability of periodic waves for koreteweg - de vries eqtuation based on the jacobi elliptic functions of cnoidal type was considered in @xcite , and for modified korteweg - de vries and nonlinear schrodinger equations of dnoidal type in@xcite . in @xcite , the authors considered the nonlinear stability of periodic waves for generalized benjamin - bona - mahony equation . recently , arruda @xcite considered the nonlinear stability of periodic traveling waves for boussinesq equation . the approach is based on the theory developed in @xcite and @xcite for the stability of solitary waves . in this paper , we consider the linearized stability of periodic traveling waves for the boussinesq model ( the cases @xmath0 ) and the klein - gordon - zakharov system . these are the cases of second order in time models , for which we can explicitly write the solutions in elliptic functions ( and moreover , we can explicitly compute the relevant portion of the spectrum of the linearized operators ) . while this certainly helps in the analysis , we believe that our results should be generalizable to all values of @xmath23 . the main tool is the theory for linearized stability for such models , developed recently by the second and third author , @xcite . the paper is organized as follows . in section [ sec:11 ] , we present the construction of our main object of study - the periodic traveling waves . this is not a new material by any means , but we do it in order to single out the solutions of interest , and to introduce some notations . in section [ sec:21 ] , we setup the linear stability problem , after which we present the main results . in section [ sec:2 ] , we outline the theory for linearized stability for second order ode in @xcite , and point out to the relevant spectral theoretic results about their linearized operators . in section [ sec:133 ] , we prove the main results - theorems [ theo:1 ] , [ theo:2 ] about the boussinesq model , while in section [ sec:5 ] , we prove theorem [ theo:10 ] about the kgz system . in this section , we show a glimpse of the construction of the periodic waves - in section [ sec:1.1 ] for the boussinesq equation ( when @xmath0 ) and in section [ sec:1.2 ] for the kgz system . applying the traveling wave ansatz , one sees that there is an one - parameter family of traveling waves of the form @xmath24 , which obey the equation @xmath25=0 , \ \ 0\leq x\leq t. $ ] whence , there exists @xmath26 , so that @xmath27 by the periodicity of @xmath28 , we conclude that @xmath29 and thus , we have a family of waves satisfying @xmath30 we now construct solutions of in various cases of interest , most notably @xmath6 and @xmath31 . this material is not new , but in order to introduce the particular parametrization that is convenient for us , we include a sketch of the construction for completeness . we will consider only the symmetric case @xmath32 . for the nonlinearity , @xmath33 , we have ( with @xmath34 ) @xmath35 therefore , @xmath36 hence the periodic solutions are given by the periodic trajectories @xmath37 of the hamiltonian vector field @xmath38 where @xmath39 the level set @xmath40 contains two periodic trajectories if @xmath41 , @xmath42 and a unique periodic trajectory if @xmath43 . here we consider the cases where @xmath44 and @xmath45 . to express @xmath46 through elliptic functions , we denote by @xmath47 the positive solutions of @xmath48 . then @xmath49 and one can rewrite ( [ kdv1a ] ) as @xmath50 introducing a new variable @xmath51 via @xmath52 , we transform ( [ kdv1b ] ) into @xmath53 where @xmath54 , @xmath55 , @xmath56 are positive constants ( @xmath57 ) given by @xmath58 therefore @xmath59 by the above formulas , @xmath60 the fundamental period of the cnoidal wave @xmath46 in ( [ kdv2 ] ) is @xmath61{1-\kappa^2+\kappa^4}}{\sqrt{w } } } , \ ; \ ; \ ; t\in \left ( { \frac{2\pi}{\sqrt{w } } } , \infty \right).\ ] ] here and below , @xmath62 and @xmath63 denote the elliptic integrals of the first and second kind in a legendre form . further , we will use the following relations @xmath64 [ lkdv1 ] for any @xmath41 and @xmath65 , there is a constant @xmath66 such that the periodic traveling solution ( [ kdv2 ] ) has period @xmath1 . the function @xmath67 is differentiable . it is easily seen that the period @xmath1 is a strictly increasing function of @xmath55 : @xmath68{1-\kappa^2+\kappa^4}k(\kappa))&= & \frac{\kappa(2\kappa^2 - 1)k(\kappa)+2(1-\kappa^2+\kappa^4)k'(\kappa)}{4\sqrt[4]{1-\kappa^2+\kappa^4}^3}\\ \\ & = & \frac{2(1-\kappa^2+\kappa^4)e(\kappa)+(1-\kappa^2)(\kappa^2 - 2)k(\kappa)}{4\kappa ( 1-\kappa^2)\sqrt[4]{1-\kappa^2+\kappa^4}^3}>0.\end{aligned}\ ] ] given @xmath69 and @xmath70 in their range , consider the functions @xmath71 , @xmath72 and @xmath73 given by the formulas and . we obtain @xmath74 further , using that @xmath75,we have @xmath76 we see that @xmath77 , whence the implicit function theorem implies the result . we consider the `` symmetric '' case @xmath32 only , with @xmath78 . denote @xmath79 . multiplying by @xmath80 and integrating implies @xmath81 hence the periodic solutions are given by the periodic trajectories @xmath37 of the hamiltonian vector field @xmath38 where @xmath82 then there are two possibilities : * ( _ outer case _ ) : for any @xmath43 the orbit defined by @xmath37 is periodic and oscillates around the eight - shaped loop @xmath83 through the saddle at the origin . * ( _ left and right cases _ ) : for any @xmath84 there are two periodic orbits defined by @xmath37 ( the left and right ones ) . these are located inside the eight - shaped loop and oscillate around the centers at @xmath85 , respectively . we will consider the left and right cases of duffing oscillator only . in these cases , denote by @xmath47 the positive roots of the quartic equation @xmath86 . then , up to a translation , we obtain the respective explicit formulas @xmath87 note that the fundamental period @xmath1 may be written as follows @xmath88 note that this is a two parameter family of solutions , parametrized explicitly in this case by @xmath89 , although we shall need a different parametrization . in fact , we would like to think of this family as being parametrized ( implicitly ) in terms of @xmath1 and @xmath69 , where these two will be independent of each other . we have the following [ lmkdv1 ] for @xmath90 , there is a constant @xmath66 such that the periodic traveling - wave solution @xmath91 determined by @xmath92 has a period @xmath1 . in addition , the function @xmath67 is differentiable . for the proof , see lemma 3.1 in @xcite . we are looking for @xmath1 periodic traveling solutions of the klein - gordon - zakharov system , . thus , we take the ansatz @xmath93 , where we take the speed @xmath94 . plugging into , we obtain the following relation between @xmath95 and @xmath28 @xmath96 two integrations in @xmath97 imply @xmath98 for some constants @xmath99 . by the periodicity we have that @xmath100 , whence @xmath101 . for simplicity , we shall consider the case @xmath32 only . that is @xmath102 returning back to the other equation in and using the relation , we obtain the following equation for @xmath103 @xmath104 thus , as in ( 1.1.2 ) after multiplying by @xmath103 and integrating we get @xmath105 denote by @xmath106 the positive roots of the polynomial @xmath107 then ( [ k:5a ] ) can be written in the form @xmath108 and the solution of is given by @xmath109 where @xmath110 moreover , @xmath111 since @xmath112 has fundamental period @xmath113 , then the solution @xmath103 has fundamental period @xmath114 . in terms of @xmath115 , given by @xmath116 [ lk ] for any @xmath41 and @xmath117 , there is a constant @xmath66 such that the periodic traveling solution ( [ k:10 ] ) has period @xmath1 . the period @xmath1 is a strictly increasing function of @xmath118 : @xmath119={\frac{k'(\kappa)+e'(\kappa)}{\sqrt{2-\kappa^2}}}>0.\ ] ] from ( [ k:10a ] ) and ( [ k:10c ] ) , we have @xmath120 @xmath121 the implicit function theorem then implies the result . we now set up the linear stability / instability problem for . set the ansatz @xmath122 and ignore all terms @xmath123 . we get @xmath124 , where @xmath125 note that this operator @xmath126 is not self - adjoint . however , if we introduce the variable @xmath127 , we get the following linearized equation in terms of @xmath128 , @xmath129=0.\ ] ] note that @xmath130={\partial}_x[h[z]]$ ] , where @xmath131 thus , the linearized equation becomes @xmath132=0 $ ] . in our considerations , we say that the wave @xmath103 is spectrally unstable , exactly when there is an exponentially growing mode , that is a pair @xmath133 and a @xmath1 periodic function @xmath134 , so that @xmath135=0 $ ] . this of course implies upon integration that for some constant @xmath136 @xmath137 integrating in @xmath138 $ ] and taking into account that both @xmath139 are exact derivatives , impies that @xmath140 . thus letting @xmath141 implies that @xmath142 and @xmath143 these arguments motivate the following [ defi:1 ] we say that the traveling wave @xmath103 is spectrally / linearly unstable , if there exists an @xmath1 periodic , * mean value zero * function @xmath144 and @xmath145 , so that @xmath146 the question for linear stability of equations in the form @xmath147 or what is equivalent ( at least in this case ) to the solvability of @xmath148 in @xmath149 has been addressed in a recent paper by the second and third author , @xcite . note that the self - adjoint operator @xmath150 that appears in is in the form @xmath151 here @xmath152 is the ubiquitous standard second order schrdinger operator , which appears in the linearization of the generalized kdv equation around its traveling wave solution @xmath103 . this observation will be crucial for the spectral properties of the operator @xmath150 as the properties of @xmath152 are generally well - known , at least for the cases into consideration , @xmath0 . we linearize the kgz system as follows . we take @xmath153 and @xmath154 and ignore the contributions of all quadratic and higher order terms . we obtain the following _ linear system _ for the corrections @xmath155 , @xmath156 further , introducing a new mean - value zero function @xmath128 , so that @xmath157 and @xmath158 . the second equation in becomes @xmath159=0,\ ] ] whence integrating in @xmath97 yields @xmath160 for some function @xmath161 . observe however , that in our choice of @xmath128 , we have required that + @xmath162 . thus , integrating the last equation in @xmath138 $ ] yields that all integrals on the left are zero , which is mean value zero ] @xmath163 , whence @xmath164 . we have shown that one can rewrite the linear stability problem as follows @xmath165 where @xmath166 and @xmath167 where @xmath168 clearly , the operator @xmath150 is self - adjoint , when considered over the domain @xmath169\times h^2_0[0,t ] \subset l^2[0,t]\times l^2_0[0,t]\ ] ] so that @xmath170\times l^2_0[0,t]$ ] . note that , when considering this spectral problem , our basic hilbertian space will be @xmath171\times l^2_0[0,t]$ ] , instead of the usual @xmath171\times l^2[0,t]$ ] . our main results are described in the next theorems . note that in all of them , the issue is linear stability for the said families of spatially periodic solutions , _ when the perturbation is taken to be periodic with the same period as the underlying solution_. the issue for stability / instability , when perturbations are taken to be with period of the form @xmath172 , @xmath20 integer is a more complicated one . clearly , the instability results continue to apply in this case , but it may very well be that some , previously stable waves ( in the context of the same period perturbations ) become unstable , when perturbed by @xmath172 periodic functions . [ theo:1 ] let the nonlinearity in has the form @xmath78 . then , the two parameter family of dnoidal solutions , described in , are linearly stable , @xmath173 where @xmath174[(2-{\kappa}^2 ) e({\kappa})-2(1-{\kappa}^2)k({\kappa})]}{(2-{\kappa}^2)(e^2({\kappa})-(1-{\kappa}^2 ) k({\kappa } ) ) } } , { \kappa}\in ( 0,1),\ ] ] where @xmath175 are elliptic integrals of the first and second kind in a legendre form . , together with its terminal value @xmath176 as a reference , width=302,height=226 ] we now give a different formulation of the main result . let @xmath177 . then , the waves described in are one parameter family of waves , having a fundamental period @xmath1 , which can be parametrized by @xmath178 , ( note that @xmath179 are in one - to - one relation given by ) . now , theorem [ theo:1 ] asserts that the stable waves in this family are exactly those with @xmath180 , where @xmath181 is determined as follows . let @xmath182 be the unique solution of @xmath183 then , @xmath184 . . the red line is @xmath185 . note that the range of this function is @xmath186.,width=302,height=226 ] our next theorem concerns the quadratic case [ theo:2 ] let the nonlinearity in has the form @xmath33 . then the periodic solutions , are linearly stable , @xmath187 where @xmath188 } { f(\kappa)+256k^4(\kappa)f'(\kappa)g(\kappa)(1-\kappa^2+\kappa^4)+{\frac{4096 k^6({\kappa})(1-{\kappa}^2+{\kappa}^4)^{3/2}(f'({\kappa } ) g({\kappa}))^2}{1 - 16\sqrt{1-\kappa^2+\kappa^4}k^2(\kappa)f'(\kappa)g(\kappa ) } } } } \end{aligned}\ ] ] and @xmath189 , \\ & & g({\kappa})=\frac{1}{128{\frac{d[k^4(\kappa)(1-\kappa^2+\kappa^4)]}{d\kappa}}}. \end{aligned}\ ] ] , together with its terminal value @xmath190,width=302,height=226 ] an alternative formulation is the following . fix @xmath191 and consider the family of cnoidal solutions , described in . this is a one - parameter family of solutions and @xmath192 are related by ] , having fundamental period @xmath1 and indexed by say @xmath192 , where @xmath193 . then , the linearly stable waves with period @xmath1 are exactly those , for which @xmath194 where @xmath195 is determined as follows . take @xmath182 to be the unique solution of the algebraic equation @xmath196{1-{\kappa}^2+{\kappa}^4}\sqrt{4+\tilde{f}({\kappa})}=t.\ ] ] then @xmath197 . { 1-{\kappa}^2+{\kappa}^4}\sqrt{4+\tilde{f}({\kappa})}:(0,1)\to { \mathbf r^1}$ ] . the red line is @xmath198 . note that the range of this function is @xmath199.,width=302,height=226 ] * remark : * using the results of theorem [ theo:1 ] and theorem [ theo:2 ] , one can reconstruct the results on linear stability of the whole line waves , @xcite @xmath200^{{\frac{1}{p-1 } } } sech^{{\frac{2}{p-1}}}\left({\frac{\sqrt{1-c^2 } ( p-1)}{2 } } \xi\right).\ ] ] recall that the results of @xcite predict linear stability if and only if @xmath201 . take @xmath31 . then , the periodic waves described in in the limit @xmath202 correspond to the whole line waves described in . observe that since @xmath203 and @xmath204 , we conclude easily @xmath205 , whence @xmath206 similarly , one can check that for @xmath6 ( more precisely if @xmath207 ) ) , we have @xmath208 thus , we obtain the results in @xcite for @xmath0 as a corollary . our next result concerns the kgz system . [ theo:10 ] the kgz system has a two - parameter family of traveling wave solutions + @xmath209 , described in and . these waves are stable , @xmath210 and @xmath211 where @xmath212 and the function @xmath213 is defined in . in @xmath214 . the green line is the terminal value of @xmath176.,width=302,height=226 ] if we take a limit as @xmath215 , we have @xmath216 since @xmath215 corresponds to the case @xmath217 or the case of the whole line , this allows us to conclude that the corresponding whole line solitons are stable , provided @xmath218 . this was indeed the conclusion in @xcite , so we are able to deduce this result , as a consequence of theorem [ theo:10 ] . once again , we provide an alternative interpretation of theorem [ theo:10 ] . let @xmath219 be a fixed period . then , there exists a one parameter family of periodic waves with period @xmath1 , described in , . more precisely , this family may be parameterized and @xmath192 are related by ] by @xmath192 with the following restrictions on @xmath192 : if @xmath220 , then @xmath221 , otherwise if @xmath222 , @xmath94 . then , the stable waves in this family are given by @xmath223 where @xmath224 is found as the unique solution ( see graph below ) of the algebraic equation @xmath225 . note that its range is @xmath226.,width=302,height=226 ] in this section , we give a precise statements of the results of @xcite , concerning the linear stability of or what is equivalent the solvability of . we assume the following about the self - adjoint operator @xmath150 : @xmath227=1 \\ { \mathcal h}\psi_0=0 , \ \ \dim[ker({\mathcal h})]=1\\ \|\psi_0\|=1 \end{array } \right.\ ] ] next , we require that for all @xmath228 ( note @xmath229 will be invertible ) @xmath230 finally , we require @xmath231 note that the last identity ensures that @xmath232 maps real - valued functions into real - valued functions . the following theorem , in a more general form , appears as theorem 1 in @xcite . [ theo:5 ] let @xmath150 be a self - adjoint operator on a hilber space @xmath232 . assume that it satisfies the structural assumptions , as well as the the reality assumption . then , if @xmath233,\psi_0 ' \rangle}\geq 0 $ ] , one has a solution to for all values of @xmath234 , that is instability in the sense of definition [ defi:1 ] . otherwise , supposing @xmath233,\psi_0 ' \rangle}<0 $ ] , * the problem has solutions , if @xmath235 satisfies the inequality @xmath236,\psi_0 ' \rangle}}}=:\mu^*({\mathcal h})\ ] ] * the problem does not have solutions ( i.e. stability ) , if @xmath237 satisfies the reverse inequality @xmath238 * remarks : * 1 . theorem [ theo:5 ] appears in @xcite as a result about the stability of , but we state it in its equivalent form for solvability of . 2 . in the applications that we consider , we restrict our attention to the hilbert space @xmath239 $ ] or @xmath240\times l^2_0[0,t]$ ] , depending on the situation that we are in . we review and state the main results regarding the spectral theory for the second order schrdinger operators , arising in the linearization around boussinesq waves . we consider again the cases @xmath6 and @xmath31 separately . in this section , we present some spectral results for @xmath152 that will be useful in the sequel . the first is a technical lemma , that will be used to establish the simplicity of the zero eigenvalue for @xmath241 . [ l1 ] @xmath242 in this case @xmath243 and @xmath244 . the spectral properties of the operator @xmath245 in @xmath246 $ ] are well - known @xcite . the first three(simple ) eigenvalues and corresponding eigenfunctions of @xmath247 are @xmath248>0\\ \\ \mu_1=0\\ \psi_1(x)=dn ( x;\kappa)sn(y;\kappa)cn(\alpha x;\kappa)={\frac{1}{2}}{\frac{d}{dy}}cn^2(y;\kappa)\\ \\ \mu_2=\kappa^2 - 2 + 2\sqrt{1-\kappa^2 + 4\kappa^4}>0\\ \psi_2(x)=dn ( x;\kappa)[1-(1 + 2\kappa^2+\sqrt{1-\kappa^2 + 4\kappa^4})sn^2(y;\kappa ) ] . \end{array}\ ] ] since the eigenvalues of @xmath249 and @xmath247 are related by @xmath250 , it follows that the first three eigenvalues of the operator @xmath249 , equipped with periodic boundary condition on @xmath246 $ ] are simple and @xmath251 . the corresponding eigenfunctions are @xmath252 and @xmath253 . note that @xmath254 and @xmath255 and hence , we can take inverse in ( [ kdv3 ] ) @xmath256 taking dot product with @xmath257 yields the relation @xmath258 differentiating ( [ kdv1 ] ) with respect to @xmath192 , we get @xmath259=2c\varphi_c$ ] , whence @xmath260 entering this last formula in the expression for @xmath261 , @xmath262 using ( [ kdv2a ] ) and that @xmath263 $ ] , we get @xmath264={\frac{1}{t}}f(\kappa),\ ] ] where @xmath265.\ ] ] now , we need to compute @xmath266 thus , to compute @xmath267 , we differentiate with respect to @xmath69 the following relation @xmath268 obtained from and . we get @xmath269 where @xmath270}{d\kappa}}}}.\ ] ] from the above relations , and , and we have @xmath271.\ ] ] the expression in the brackets above is strictly positive ( see the graphic below ) which proves the lemma . $ ] , width=302,height=226 ] consider @xmath272 we use ( ) to rewrite the operator @xmath249 in an appropriate form . from the expression for @xmath273 from ( [ mkdv2 ] ) and the relations between the elliptic functions @xmath274 , @xmath275 and @xmath276 , we obtain @xmath277\ ] ] where @xmath278 . it is well - known @xcite that the first five eigenvalues of @xmath279 , with periodic boundary conditions on @xmath280 $ ] , where @xmath281 is the complete elliptic integral of the first kind , are simple . these eigenvalues , with their corresponding eigenfunctions are as follows @xmath282 \nu_{1}=1+k^{2 } , & \psi_{1}(y)=cn(y , k)dn(y , k ) = sn'(y , k),\\[1 mm ] \nu_{2}=1 + 4k^{2 } , & \psi_{2}(y)=sn(y , k)dn(y , k ) = -cn'(y , k),\\[1 mm ] \nu_{3}=4+k^{2 } , & \psi_{3}(y)=sn(y , k)cn(y , k ) = -k^{-2}dn'(y , k),\\[1 mm ] \nu_{4}=2 + 2k^{2}+2\sqrt{1-k^{2}+k^{4 } } , & \psi_{4}(y)=1-(1+k^{2}+\sqrt{1-k^{2 } + k^{4}})sn^{2}(y , k ) . \end{array}\ ] ] it follows that the first three eigenvalues of the operator @xmath249 , equipped with periodic boundary condition on @xmath246 $ ] ( that is , in the case of left and right family ) , are simple and @xmath283 . the corresponding eigenfunctions are @xmath284 . thus , we have proved the following [ prop:3 ] the linear operator @xmath152 defined by @xmath285 has the following spectral properties : * the first three eigenvalues of @xmath152 are simple . + * _ the second eigenvalue of @xmath152 is @xmath286 , which is simple . + _ * the rest of the spectrum consists of a discrete set of eigenvalues , which are strictly positive . next , we verify the following technical result . [ l2 ] the operator @xmath152 verifies @xmath287 this statement was needed and proved in the work of deconinck - kapitula , @xcite , but we repeat the short argument for completeness . first we will prove that for @xmath288 , @xmath289 using the expression for @xmath247 and @xmath290 , we get @xmath291 it follows that for @xmath292 , @xmath293 observe however that @xmath294 , which means that @xmath295 , whenever @xmath296 ( since @xmath297 ) . by , this implies that @xmath298 . from ( theorem 2.15 , in @xcite ) , the eigenfunctions of the operator @xmath152 form an orthonormal basis of @xmath171 $ ] and hence we compute @xmath299 by expanding @xmath257 in the eigenfunction expansion . note that all terms corresponding to mean value zero eigenfunctions disappear ( since @xmath298 ) . hence , the expansion for @xmath299 has only two non - zero terms . more precisely , we have @xmath300 \end{aligned}\ ] ] where we have used the formulas @xmath301 ; \\ & & \int_{0}^{2k(\kappa)}{sn^4(y)}dy={\frac{2}{3\kappa^4}}[(2+\kappa^2)k(\kappa)-2(1+\kappa^2)e(\kappa)]\end{aligned}\ ] ] and @xmath302,\\ \\ & + & { \frac{(1+\kappa^2-\sqrt{1-\kappa^2+\kappa^4})^2}{3\kappa^4 } } [ ( 2+\kappa^2)k(\kappa)-2(1+\kappa^2)e(\kappa ) ] ; \\ b_4(\kappa)&= & k(\kappa)-{\frac{2(1+\kappa^2+\sqrt{1-\kappa^2+\kappa^4})}{\kappa^2}}[k(\kappa)-e(\kappa)]\\ \\ & + & { \frac{(1+\kappa^2+\sqrt{1-\kappa^2+\kappa^4})^2}{3\kappa^4 } } [ ( 2+\kappa^2)k(\kappa)-2(1+\kappa^2)e(\kappa ) ] . \end{aligned}\ ] ] from the graph of the function @xmath303 below , we realize that @xmath304 and hence lemma [ l2 ] is established . , width=302,height=226 ] lemma [ l1 ] and lemma [ l2 ] allow us to verify the important property about the simplicity of the zero eigenvalue for the operator @xmath305 . [ cor : p ] let @xmath6 or @xmath31 . then , the operator @xmath306 ( corresponding to @xmath307 ) defined in has zero as a simple eigenvalue in @xmath308 $ ] , with an eigenfunction + @xmath309 . first @xmath310 is easily seen to be an eigenfunction , since @xmath311=-{\partial}_x { \mathcal l}[{\varphi}_c']=0,\ ] ] since @xmath312 is an eigenfunction for @xmath152 . regarding uniqueness , let @xmath313 $ ] , so that @xmath314 . it follows that @xmath315=c = const.\ ] ] since @xmath316 and @xmath317 , we can resolve the last equation as @xmath318 . thus @xmath319 whence @xmath320 , since @xmath321 by lemma [ l2 ] . it follows that @xmath322 is an eigenvector for @xmath152 . thus , @xmath323 , by proposition [ prop:3 ] . but then , @xmath324 $ ] , since we are in the space @xmath325 and the uniqueness is established . first , as in the boussinesq case , we show that the operator @xmath150 has a simple eigenvalue at zero . in addition , we identify the unique ( up to a multiplicative constant ) eigenfunction of @xmath150 . recall that in our considerations , we work with the space @xmath308 $ ] , that is the second component will contain only functions with mean value zero . this proposition mirrors closely the corresponding statement of proposition 8 in @xcite , with a few notable differences . [ prop : k1 ] the self - adjoint operator @xmath150 introduced in has an eigenvalue at zero , which is simple . in addition , the unique ( up to a multiplicative constant ) eigenfunction is given by @xmath326 . let @xmath327 be an eigenvector corresponding to a zero eigenvalue , that is @xmath328 . in other words , @xmath329 integrating the second equation in @xmath97 implies that for some constant @xmath330 , we have @xmath331 whence the equation for @xmath332 becomes @xmath333 we will show that @xmath334 and then @xmath335 for some constant @xmath336 . to that end , recall the defining equation for @xmath103 , namely and differentiate it with respect to @xmath97 . we get @xmath337 following the usual analogy with the kdv equation , we introduce the second order differential operator @xmath338 using that @xmath339 and @xmath340 , we get @xmath341 where @xmath278 . it follows that the first three eigenvalues of the operator @xmath249 , equipped with periodic boundary condition on @xmath246 $ ] , are simple and @xmath342 . the corresponding eigenfunctions are @xmath284 , where @xmath343 and @xmath344 are given in ( 3.2.2 ) . in particular , the kernel of @xmath152 is spanned by @xmath312 , i.e. @xmath345 $ ] . going back to , we can rewrite it as @xmath346+c_0 { \varphi}_c=0.\ ] ] note that all solutions to this last equation are given by @xmath347,\ ] ] where @xmath336 is an arbitrary scalar , since @xmath348= ker({\mathcal l})$ ] . plugging this last formula in the equation for @xmath349 yields @xmath350\right)+c_0= -{\frac{d}{w}}{\varphi}{\varphi}'+c_0\left({\frac{{\varphi}{\mathcal l}^{-1}[{\varphi}]}{w}}+1\right)\ ] ] integrating the last expression in @xmath138 $ ] and using the periodicity yields @xmath351 \rangle}}{w}}+t\right)=0.\ ] ] thus , if we verify that @xmath352 \rangle}\neq - t w$ ] , we would conclude from that @xmath334 , whence @xmath353 . furthermore , @xmath354 , whence @xmath355 recall however that the @xmath356 in the formula above is uniquely determined by the fact that @xmath349 has mean value zero ( i.e. @xmath357 $ ] ) , whence @xmath358 thus , proposition [ prop : k1 ] is established modulo the following + + * fact : * @xmath359 \rangle}>-{\frac{tw}{3}}$ ] , so in particular @xmath352 \rangle } > - t w$ ] + + from ( [ k:5 ] ) , we have @xmath360 and thus @xmath361 on the other hand , differentiating @xmath362 , with respect to @xmath69 ( and dividing by @xmath69 ) and using to express @xmath363 , results in @xmath364=2{\varphi}''-{\frac{1}{w}}{\varphi}_c={\frac{1}{w}}{\varphi}_c+{\frac{1}{w}}{\mathcal l}{\varphi}_c,\ ] ] taking @xmath365 in the last identity yields @xmath366 since @xmath367 , we compute @xmath368 . therefore , @xmath369- \|{\varphi}_c\|^2= \\ & = & { \frac{w}{2 } } { \partial}_w [ \frac{16w^2}{t } e(\kappa)k(\kappa)]- \frac{16w^2}{t } e(\kappa)k(\kappa)=\frac{8w^3}{t}}{\frac{d}{d\kappa}}[k(\kappa)e(\kappa)]{\frac{d\kappa}{dw } \end{aligned}\ ] ] to compute @xmath267 , note that from ( [ k:10a ] ) , we have @xmath370 , whence @xmath371 differentiating ( [ k:19a ] ) respect to @xmath69 , we get @xmath372}}.\ ] ] thus , using , @xmath373 { \frac{(2-\kappa^2)k^2(\kappa)}{{\frac{d}{d\kappa}}[(2-\kappa^2)k^2(\kappa)]}}= \\ & = & -w t\left[2 { \frac { \frac{d}{d\kappa}[k(\kappa)e(\kappa ) ] } { { \frac{d}{d\kappa}}[(2-\kappa^2)k^2(\kappa ) ] } } \right]\end{aligned}\ ] ] looking at the graph below , we realize that since @xmath374 } { { \frac{d}{d\kappa}}[(2-\kappa^2)k^2(\kappa ) ] } } \geq 2 { \frac { \frac{d}{d\kappa}[k(\kappa)e(\kappa ) ] } { { \frac{d}{d\kappa}}[(2-\kappa^2)k^2(\kappa)]}}$ ] , we have that @xmath375 which establishes the claim . } { { \frac{d}{d\kappa}}[(2-\kappa^2)k^2(\kappa ) ] } } $ ] .,width=302,height=226 ] the next thing one needs to establish , in order to apply theorem [ theo:5 ] is that the operator @xmath150 for the kgz system ( defined in ) has a simple negative eigenvalue . this result should be compared with the corresponding statement in proposition 9 in @xcite for the whole line case . [ prop : k2 ] the operator @xmath150 , defined in has a simple negative eigenvalue . consider the eigenvalue problem in the form @xmath376 for some @xmath377 . as in proposition [ prop : k1 ] , this can be rewritten as . this means in particular that the statement for existence of negative eigenvalue is invalid , unless the second component of the hilbert space is @xmath308 $ ] ] @xmath378 form the second equation , we may resolve for @xmath349 @xmath379.\ ] ] this last formula requires a bit of justification , but basically since @xmath380 $ ] is guranteed to have mean value zero , it suffices to define @xmath381 ( where @xmath382 ) on @xmath171 $ ] by @xmath383:= \sum_{n=-\infty}^\infty { \frac{a_n}{a^2 + 4\pi^2 w { \frac{n^2}{t^2 } } } } e^{2\pi i n { \frac{x}{t}}},\ ] ] whence the formula for @xmath349 in makes sense . in fact , @xmath308 $ ] is invariant under the action of and @xmath381 and hence @xmath384 $ ] . in fact , we use to deduce the following formula for @xmath385 @xmath386= -{\frac{{\varphi}f}{w } } + { \frac{a^2}{w } } ( a^2-w { \partial}_x^2)^{-1 } [ { \varphi}_c f].\ ] ] we plug this in the first equation of to obtain the following equation for @xmath332 @xmath387 = 0.\ ] ] introduce a one - parameter family of self - adjoint operators @xmath388.\ ] ] in order to finish the proof of proposition [ prop : k2 ] , we need to establish that there is unique @xmath389 , so that the operator @xmath390 has an eigenvalue zero and such an eigenvalue is simple . we shall first show that there exists @xmath389 , so that @xmath390 has an eigenvalue at zero and then we show that this eigenvalue zero is simple for @xmath390 . to that end , we will establish the following + * claim : * for @xmath391 , we have @xmath392 . assuming the validity of the claim , let us finish the proof of proposition [ prop : k2 ] . define @xmath393 to be the minimal eigenvalue for @xmath394 , that is @xmath395 clearly , the function @xmath396 is continuous ( in fact , more generally , @xmath397 is continuous as a function from @xmath398)$ ] ) ) . moreover , as a consequence of the claim , @xmath393 is a strictly increasing function of its argument . in order to show the existence of @xmath399 , it will suffice to show that @xmath396 changes sign in @xmath400 ( and hence vanishes at some @xmath389 ) . but at @xmath32 , we have @xmath401 that was considered before . since we have checked that @xmath152 has a ( simple ) negative eigenvalue , say @xmath402 it follows that @xmath403 . on the other hand , by the claim @xmath404 in particular , for every @xmath405 , we have @xmath406 , whence @xmath407 . thus , for any @xmath408 , the function @xmath393 changes sign ( exactly once ) in the interval @xmath409 . we have shown that there is @xmath410 . now , we have to show the second part of the proposition , namely that @xmath411 is an isolated eigenvalue for @xmath390 . let @xmath412 be the eigenvector for the simple negative eigenvalue for @xmath413 . note that by proposition [ prop:1 ] , the second eigenvalue of @xmath152 is zero , which means that @xmath414 . in addition , by the claim , we have that @xmath415 , and thus , by the courant maxmin principle for the second eigenvalue @xmath416 it follows that @xmath417 , while @xmath418 , which shows the simplicity of the zero eigenvalue for @xmath390 . it now remains to establish the claim . by the form of the operators @xmath394 , it suffices to show for all trig polynomials @xmath419 $ ] that for @xmath420 , @xmath421,f \rangle}\geq b^2{\langle { \varphi}_c ( b^2-w { \partial}_x^2)^{-1 } [ { \varphi}_c f],f \rangle}\ ] ] but letting @xmath422 be the fourier coefficients of @xmath423 , that is @xmath424 or + @xmath425 , we see that @xmath426,f \rangle } & = & \sum_n { \frac{b^2 } { b^2 + 4\pi^2 w { \frac{n^2}{t^2}}}}|a_n|^2\leq \\ & \leq & \sum_n { \frac{a^2 } { a^2 + 4\pi^2 w { \frac{n^2}{t^2}}}}|a_n|^2= a^2{\langle { \varphi}_c ( a^2-w { \partial}_x^2)^{-1 } [ { \varphi}_c f],f \rangle } , \end{aligned}\ ] ] where we have used that @xmath427 , whenever @xmath41 , @xmath391 . now that we have the solutions , we need to check that the operator @xmath428 satisfies the requirements in theorem [ theo:5 ] , after which , we need to compute the index @xmath429 . we collect the needed results in the following propositions . ( spectral properties of @xmath428 ) [ prop:1 ] the operator @xmath305 , introduced in , satisfies , , for @xmath0 . our next result gives a precise formula for the index @xmath430 . [ prop:2 ] we have 1)for @xmath6 @xmath431 where @xmath432 is defined in below . \2 ) for @xmath433 @xmath434[(2-{\kappa}^2 ) e({\kappa})-2(1-{\kappa}^2)k({\kappa})]}{(2-{\kappa}^2)(e^2({\kappa})-(1-{\kappa}^2 ) k({\kappa}))}}}.\ ] ] * note : * the function under the square root is positive for all values of @xmath435 . we now finish the proof of theorem [ theo:1 ] , based on the results of propositions [ prop:1 ] and [ prop:2 ] . let @xmath31 . we apply theorem [ theo:5 ] , whence we get stability , provided @xmath436 . thus , we need to resolve the inequality @xmath437 where we have taken @xmath174[(2-{\kappa}^2 ) e({\kappa})-2(1-{\kappa}^2)k({\kappa})]}{(2-{\kappa}^2)(e^2({\kappa})-(1-{\kappa}^2 ) k({\kappa}))}},\ ] ] we obtain for the interval of the stable speeds @xmath438 as stated in theorem [ theo:1 ] . similarly , for @xmath6 , we have according to proposition [ prop:2 ] that @xmath439 whence we conclude similarly that @xmath440 is a necessary and sufficient condition for stability of the corresponding traveling wave . we note that standard arguments imply the validity of , since @xmath150 is fourth order operator . the reality condition is also trivially satisfied as all our potentials are real - valued . the hard to check condition is . we need to check that the operator @xmath150 has a simple eigenvalue at zero . this was indeed the conclusion of corollary [ cor : p ] . thus , it remains to show that the operator @xmath150 defined in has a simple negative eigenvalue and proposition [ prop:1 ] will follow . this is a non - trivial fact . interestingly enough , this was needed ( and proved in our paper @xcite ) , when we have considered the transverse instability of the same spatially periodic waves in the kadomtsev - petviashvili ( kp ) and modified kp models , that is exactly for the operators considered here , corresponding to the cases @xmath207 and @xmath78 . more precisely , we have derived a necessary condition , so that the first two eigenvalues of @xmath441 satisfy @xmath442 which was verified for both @xmath0 . this is exactly what is needed here . the interested reader may consult section 4.1 in @xcite for full and complete proof of these facts . we compute the index of stability @xmath443 . we have @xmath444 where @xmath445 let @xmath446=\varphi'_c$ ] . it follows that @xmath447 , for some constant @xmath70 . hence @xmath448 thus @xmath449 from ( [ kdv9a ] ) , we have that @xmath450 , whence @xmath451 combining ( [ kdv5 ] ) , ( [ kdv9 ] ) , and ( [ kdv9c ] ) yields @xmath452 from ( [ kdv1 ] ) after integrating @xmath453 we use to see @xmath454.\ ] ] next , we use , , and to compute @xmath455 \end{aligned}\ ] ] finally , using the formulas for @xmath456 allows us to find @xmath457}{t}}\ ] ] putting all formulas together @xmath458 } } \times \\ \\ & & \times \left[-f(\kappa)-256k^4(\kappa)f'(\kappa)g(\kappa)(1-\kappa^2+\kappa^4)-{\frac{4096 k^6({\kappa})(1-{\kappa}^2+{\kappa}^4)^{3/2}(f'({\kappa } ) g({\kappa}))^2}{1 - 16\sqrt{1-\kappa^2+\kappa^4}k^2(\kappa)f'(\kappa)g(\kappa)}}\right ] \end{aligned}\ ] ] thus , if we assign the function @xmath459 } { f(\kappa)+256k^4(\kappa)f'(\kappa)g(\kappa)(1-\kappa^2+\kappa^4)+{\frac{4096 k^6({\kappa})(1-{\kappa}^2+{\kappa}^4)^{3/2}(f'({\kappa } ) g({\kappa}))^2}{1 - 16\sqrt{1-\kappa^2+\kappa^4}k^2(\kappa)f'(\kappa)g(\kappa)}}}}.\ ] ] we get @xmath460 . thus , the index formula holds as stated in proposition [ prop:2 ] , namely @xmath461 in this section , we compute the index of stability . for that we need to consider first @xmath462 where @xmath463 . thus , we need to compute @xmath464 $ ] . let @xmath465={\varphi}_c'$ ] . it follows that @xmath466 for some constant @xmath70 . we conclude that @xmath467 note that @xmath468 is well - defined , since @xmath317 , which spans @xmath469 . thus @xmath470 differentiating ( with @xmath32 ) with respect to @xmath192 yields @xmath471 from , we get @xmath472 and @xmath473 \\ = -{\frac{4}{t}}{\frac{e^2(\kappa)-(1-\kappa^2)k^2(\kappa)}{\kappa ( 1-\kappa^2)}}{\frac{d\kappa}{dw}}.\ ] ] in addition , note that @xmath474=0.\ ] ] and hence @xmath475 from the relations , we have @xmath476 which after differentiating with respect to @xmath69 allows us to express @xmath477 thus @xmath478 using that @xmath479 , we obtain @xmath480 since @xmath481 , we get @xmath482 from the above relations and , we get @xmath483 thus , @xmath484[(2-{\kappa}^2 ) e({\kappa})-2(1-{\kappa}^2)k({\kappa})]}{(2-{\kappa}^2)(e^2({\kappa})-(1-{\kappa}^2 ) k({\kappa}))}}},\end{aligned}\ ] ] which is exactly the claim of proposition [ prop:2 ] . we have already checked the conditions on the operator @xmath150 , defined in in section [ sec:3.3 ] . namely , we have established the simplicity of the eigenvalue at zero in proposition [ prop : k1 ] and then , we have verified the existence and simplicity of a single negative eigenvalue . it now remains to compute the index @xmath443 , after which , we obtain a characterization of the linear stability by theorem [ theo:5 ] , namely @xmath436 . assuming the validity of proposition [ prop : k15 ] , we now finish the proof of theorem [ theo:10 ] . to that end , observe that since @xmath490 , we have instability whenever @xmath491 . for @xmath492 , we need to solve the inequality @xmath493 which results in the following necessary and sufficient condition for linear stability @xmath494 this is exactly the statement of theorem [ theo:10 ] . now we will estimate the index of stability @xmath495 , @xmath496 where @xmath497 is so that @xmath498 . thus , we need to compute @xmath499 we have @xmath500 integrating once in the second equation yields @xmath501 where @xmath502 is a constant of integration and needs to be determined . the first equation becomes @xmath503 or @xmath504 on the other hand , taking derivative with respect to @xmath69 in ( [ k:5 ] ) yields @xmath505 from ( [ k:16 ] ) and ( [ k:17 ] ) , we have @xmath506 and hence @xmath507 plugging this in ( [ k:15a ] ) and integrating , we get @xmath508 using that @xmath509 and , we have @xmath510 = \\ & = & { \frac{8w}{t}}\left [ 2e(\kappa)k(\kappa)-{\frac{(2-\kappa^2)k^2(\kappa){\frac{d}{d\kappa}}[e(\kappa)k(\kappa)]}{{\frac{d}{d\kappa}}[(2-\kappa^2)k^2(\kappa)]}}\right ] . \end{aligned}\ ] ] the above formula allow us to express @xmath502 as a function of @xmath511 only @xmath512 - 2(2-\kappa^2)k^2(\kappa)\frac{d}{d\kappa}[e(\kappa)k(\kappa ) ] } { ( 2-\kappa^2)k^2(\kappa)\frac{d}{d\kappa}[(2-\kappa^2)k^2(\kappa)]-2(2-\kappa^2)k^2(\kappa)\frac{d}{d\kappa}[e(\kappa)k(\kappa)]}\\ \\ & = & { \frac{(2-\kappa^2)e^2(\kappa)-8(1-\kappa^2)e(\kappa)k(\kappa)+2(1-\kappa^2)(2-\kappa^2)k^2(\kappa ) } { 2(2-\kappa^2)^2e(\kappa)k(\kappa)-2(1-\kappa^2)(2-\kappa^2)k^2(\kappa)-2(2-\kappa^2)e^2(\kappa)}}. \end{aligned}\ ] ] now , @xmath513 from ( [ k:5 ] ) and the expression for @xmath332 and @xmath385 , we get @xmath514 where @xmath515}}{\frac{d}{d\kappa}}\left [ { \frac{k(\kappa)}{2-\kappa^2}}i_1\right]\right];\\ \\ j_4 & = & { \frac{tc_1}{2w^3}}\langle { \varphi}_c^2 , { \varphi}_c^2\rangle=32c_1{\frac{k(\kappa)}{2-\kappa^2}}i_1;\\ \\ j_5 & = & { \frac{t(1-c_1)}{w}}\langle { \varphi}_c , { \frac{d{\varphi}_c}{dw}}\rangle=8(1-c_1 ) \left [ 2k(\kappa)e(\kappa)-{\frac{(2-\kappa^2)k^2(\kappa){\frac{d}{d\kappa}}[e(\kappa)k(\kappa)]}{{\frac{d}{d\kappa}}[(2-\kappa^2)k^2(\kappa)]}}\right ] . \end{aligned}\ ] ] in addition , we have @xmath516 using that @xmath517 and @xmath518 , we get @xmath519=\\ \\ & = & { \frac{8w}{3t}}{\frac{k(\kappa)}{2-\kappa^2}}[2(2-\kappa^2)e(\kappa)-4(1-\kappa^2)k(\kappa ) ] . \end{aligned}\ ] ] and @xmath520+{\frac{16k(\kappa)}{2-\kappa^2}}i_1- { \frac{16}{2-\kappa^2}}e^2(\kappa)\right]=\\ \\ & = & { \frac{16 w}{t}}\left[e(\kappa)k(\kappa)-{\frac{1-\kappa^2}{2-\kappa^2}}k^2(\kappa)-{\frac{1}{2-\kappa^2}}e^2(\kappa)\right ] . \end{aligned}\ ] ] combining the above relations yields @xmath521}}\\ : = { -\frac{n(\kappa)}{w } } \ ] ] j. boussinesq , _ therie des ondes et des remous qui se propagent le long dun canal rectangulaire horizontal , en communiquant au liquide continu dans ce canal des vitesses sensiblement pareilles de la surface au fond , _ _ j. math . pures appl . _ * 17 * ( 1872 ) , no . 2 , p. 55108 . t. ozawa , k. tsutaya , y. tsutsumi , _ normal form and global solutions for the klein - gordon - zakharov equations , _ _ ann . henri poincar , analyse non linaire _ * 12 * ( 1995 ) , p. 459503 . t. ozawa , k. tsutaya , y. tsutsumi , _ well - posedness in energy space for the cauchy problem of the klein - gordon - zakharov equations with different propagation speeds in three space dimensions . _ ann . _ * 313 * ( 1999 ) , no . 1 , p. 127140 .

it has been known for more than 30 years that star formation in giant molecular clouds ( gmcs ) is slow , in the sense that only of the gas forms stars every free - fall time . this result is entirely independent of any particular model of molecular cloud lifetime or evolution . here we survey observational data on higher density objects in the interstellar medium , including infrared dark clouds and dense molecular clumps , to determine if these objects form stars slowly like gmcs , or rapidly , converting a significant fraction of their mass into stars in one free - fall time . we find no evidence for a transition from slow to rapid star formation in structures covering three orders of magnitude in density . this has important implications for models of star formation , since competing models make differing predictions for the characteristic density at which star formation should transition from slow to rapid . the data are inconsistent with models that predict that star clusters form rapidly and in free - fall collapse . magnetic- and turbulence - regulated star formation models can reproduce the observations qualitatively , and the turbulence - regulated star formation model of krumholz & mckee quantitatively reproduces the infrared - hcn luminosity correlation recently reported by gao & solomon . slow star formation also implies that the process of star cluster formation can not be one of global collapse , but must instead proceed over many free - fall times . this suggests that turbulence in star - forming clumps must be driven , and that the competitive accretion mechanism does not operate in typical cluster - forming molecular clumps .

astro-ph0606277

@xcite first pointed out that star formation in giant molecular clouds ( gmcs ) happens surprisingly slowly . comparing the mass of gmcs in the galaxy with the total galactic star formation rate implies that no more than @xmath0 of the gas can form stars for each cloud free - fall time . this result is sufficiently surprising that numerous theories have been proposed to explain it , ranging from the idea that strong magnetic fields ( e.g. * ? ? ? * ) or turbulence ( e.g. * ? ? ? * ) within clouds inhibit star formation to the idea that galactic - scale gravitational instability regulates star formation ( e.g. * ? ? ? * ) to the idea that gmcs are , contrary to most observational estimates to date @xcite , actually gravitationally unbound ( e.g. * ? ? ? following @xcite , we define the dimensionless star formation rate per free - fall time @xmath1 as the fraction of an object s mass that it converts into stars per free - fall time at the mean density of that object . the @xcite argument shows that @xmath2 for gmcs . this provides a powerful constraint on models of gmcs . for example , it rules out the early gmc model of @xcite , a simplified version of which is that gmcs are spheres of gas of density @xmath3 that are in free - fall collapse . the clouds reach a singularity in a time @xmath4^{1/2}$ ] , at which point their mass is converted into stars . for a population of such clouds @xmath5 , which is inconsistent with the @xcite result . an important observational question , and a crucial test for theories of how star formation is regulated , is to what densities and length scales @xmath1 remains much smaller than unity . in other words , is there a density at which something like the @xcite free - fall collapse model becomes reasonable ? as an example , consider observing a star - forming region using a molecular tracer sensitive to gas at densities of @xmath6 @xmath7 , where @xmath8 is the number density of hydrogen nuclei , to estimate the total mass of gas at such high densities . this is larger than the mean density of `` cores '' seen both observationally ( e.g. * ? ? ? * ) and in simulations of star formation regulated by turbulence ( e.g. * ? ? ? * ) , and is roughly the density at which models of magnetically - regulated star formation predict that gas will completely decouple from the magnetic field and enter free - fall collapse ( e.g. * ? ? ? . at such high densities protostellar outflows can probably stop at most half the gas from reaching a star @xcite , and the thermal pressure in gas at @xmath9 @xmath7 is considerably higher than the typical ram pressure of the turbulence in gmcs , so gas at such high densities is largely impervious to external perturbations . thus , regardless of the model of star formation one adopts , one would expect that almost all of the gas at densities @xmath10 @xmath7 is part of gravitationally - bound , collapsing objects that have largely decoupled from the background turbulent flow . in the absence of effective internal support or external disturbance , order unity of the gas at such high densities is likely to be incorporated into a star within one free - fall time . for this reason , essentially all models of star formation predict that the total mass of gas at densities @xmath10 @xmath7 , divided by the free - fall time of this gas , should yield a value comparable to the total star formation rate in the region over which the mass is measured . instead of @xmath11 as for gmcs , i.e. slow star formation , one would obtain @xmath12 , i.e. rapid star formation . however , different models make different predictions about the shape of the curve of @xmath1 versus density in between @xmath0 at the characteristic density of gmcs , @xmath13 @xmath7 , and @xmath14 at a density @xmath10 @xmath7 . these different predictions correspond to different models of the physical scale at which gas both decouples from the background flow and ceases to be supported by internal feedback mechanisms , and thus transitions from slow to rapid star formation . at one extreme , magnetic regulation models such as those of @xcite , neglecting for the moment turbulent enhancement of the ambipolar diffusion rate ( e.g. * ? ? ? * ) , predict that star formation only becomes rapid once gas decouples from the magnetic field , a process that does not even begin until densities of @xmath15 @xmath7 . at the other extreme , @xcite argue that star clusters form from gas clumps at densities of @xmath16 @xmath7 that undergo a free - fall collapse in which at least @xmath17 of their mass is converted into stars @xcite , so star formation should be rapid at this density or higher . in this model , the decoupling scale corresponds to the transition from globally unbound structures ( gmcs ) to globally bound structures ( protoclusters ) . thus , extending the @xcite calculation of star formation rate divided by free - fall time to higher densities , in hopes of identifying a scale at which there is a transition from slow to rapid star formation , provides a means of distinguishing between models of how star formation is regulated . it is important at this point to differentiate the concepts of the _ rate _ and _ efficiency _ of star formation , and the _ lifetime _ of star - forming clouds . unfortunately these terms are often confused in the literature , and there are no standard definitions , so we describe here the definitions we use in this paper . the star formation rate @xmath18 is the easiest to define , since it is simply the instantaneous conversion rate of gas into stars within some volume . if we pick a density threshold @xmath3 , we can then define the dimensionless star formation rate per free - fall time for the gas above that density threshold , @xmath19 $ ] , where @xmath20 is the mass of material inside the volume of density @xmath3 or greater . in contrast , the lifetime @xmath21 of star - forming cloud is somewhat more ambiguous . we take it to mean the total duration during which a cloud is visible in a tracer sensitive to densities of @xmath3 or more . note that our definition neglects the complication that something that starts as a single cloud may in the course of its evolution break up into multiple pieces , so the visible lifetime and the dynamical lifetime may be different . finally , by the efficiency @xmath22 we mean the fraction of gas mass @xmath20 that is converted into stars by a cloud over its lifetime @xmath23 . again , we neglect the complication that clouds are not closed boxes , so @xmath20 is likely to be time - dependent due to continuing accretion or mass loss . ideally @xmath22 should be computed using a lagrangian definition of the cloud mass , i.e. all the mass that reached density @xmath3 or more at some point . however , this is generally not a direct observable . roughly speaking , the rate , efficiency , and lifetime are related by @xmath24 . consequently , one can define a rough time - averaged value for @xmath1 in a cloud @xmath25 $ ] , i.e. the mass fraction converted into stars divided by the cloud lifetime in free - fall times . one could also describe @xmath26 as being the efficiency per free - fall time , since it measures a fraction of mass converted into stars . however , we refer to it as a rate because it is measured in amount per unit time , whereas efficiency in the literature most commonly refers to the total fraction of mass converted into stars , not the amount per unit time . moreover , since the time - averaged definition is ambiguous anyway due to uncertainties in exactly what is meant by the efficiency and the lifetime , we will generally use the instantaneous value of @xmath1 , which is well defined and , as we show , directly observable . the reason for making all these definitions explicit it to point out that observational constraints on @xmath1 by themselves do not directly constraint @xmath27 or @xmath28 , and vice versa . numerous authors have used various observational techniques to estimate @xmath27 in clusters and gmcs ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? however , with the exception of [ clustertime ] , we will not discuss the question of cloud lifetime at all in this paper . when we refer to rapid versus slow star formation , what we mean is @xmath12 or @xmath29 , not is @xmath30 or @xmath31 . these questions are conceptually distinct . as an example , note that in the @xcite picture of gmcs as collapsing spheres , the evolution time scale is always the cloud free - fall time , @xmath32 . however , the model still gives @xmath5 , and we would therefore describe it as a rapid star formation . in contrast , @xcite argue that star formation lasts only one crossing time of a molecular cloud , so again @xmath33 , but that during this time only @xmath0 of the mass turns into stars . we would describe this as slow star formation , since @xmath34 , even though the cloud evolution time is similar to that in the @xcite model . the @xcite argument rules out the @xcite free - fall collapse model for gmcs , but is consistent with the @xcite model . thus , we emphasize that _ in this paper we remain completely agnostic on the question of molecular cloud lifetime_. in this paper we consider star formation in several classes of object . infrared dark clouds ( irdcs ) are regions of high extinction seen in absorption against the galactic infrared background @xcite . irdcs are clearly associated with star formation , and in at least some cases irdcs have massive protostars embedded within them @xcite . several authors have suggested ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) that irdcs are the progenitors of star clusters . within irdcs , at still higher densities , are dense molecular clumps . these objects may be observed in a variety of molecular transitions with high critical densities , and we consider two here : hcn(1 - 0 ) @xcite and cs(5 - 4 ) @xcite . molecular clumps seen in these two transitions are often associated with water masers and other signs of massive , clustered star formation . our goal is to determine @xmath1 for each of these increasingly dense gas tracers . we also determine this quantity for the orion nebula cluster using a completely different method , which provides an independent check on our estimates . the remainder of this paper proceeds as follows : in [ sfrest ] we use a variety of observations to derive @xmath1 for our objects and construct a plot of @xmath1 versus characteristic density to search for signs of a transition from slow to fast star formation . in [ implications ] , we compare our results to theoretical models for the regulation of star formation , and also point out some implications for the process of star cluster formation . finally , in [ conclusions ] we summarize our conclusions , and suggest directions for future work . let @xmath35 be the total star formation rate in a galaxy , and consider star formation occuring in objects of class x. let @xmath36 be the fraction of galactic star formation that occurs in these objects , @xmath37 be their total mass in the galaxy , and @xmath38 be their typical free - fall time . then @xmath39 in this section we use observations to estimate all the factors on the right hand side , and thus determine the dimensionless star formation rate @xmath40 in several classes of object . we emphasize that , as we vary the typical density of the objects we consider , both @xmath38 and @xmath37 will vary as we go to denser and denser clouds , the free - fall time will decline , pushing @xmath40 downward , but the amount of mass in objects of that density will also decrease , pushing @xmath1 back up . ultimately , we wish to determine which of these two effects dominate , or whether they roughly balance . in addition to determining a best value for @xmath40 , we also attempt to make a rough estimate our uncertainties by estimating the uncertainty on each parameter in equation ( [ sfrffeqn ] ) . we do this from data where possible and from more general considerations where not , and we add the uncertainty factors in quadrature . this method of combining errors is formally appropriate only if the errors are independent and gaussian - distributed , neither of which are likely to be completely true , but the procedure does give us a sense of by how much our estimates could vary . the unknown on the right hand side of ( [ sfrffeqn ] ) that has been studied most heavily is the star formation rate . numerous authors have discussed methods of inferring the star formation rate from various observables ( e.g. * ? ? ? * ; * ? ? ? * and references therein ) , so here we only summarize aspects of this discussion that we will apply directly in what follows . in the milky way , @xcite find @xmath41 @xmath42 yr@xmath43 based on catalogs of galactic hii regions . star formation is distributed in an exponential disk with scale radius @xmath44 kpc and sharp cutoffs at 3 kpc and 11 kpc in galactocentric radius . for this distribution , approximately @xmath45 of galactic star formation occurs within the solar circle . the dominant uncertainty in this estimate , roughly 0.3 dex @xcite , is the shape of the stellar initial mass function ( imf ) , since hii regions only trace the massive stellar population and one must extrapolate to estimate the total mass . estimates of the star formation rate based on chemical modelling @xcite also give star formation rates consistent with 3 @xmath42 yr@xmath43 to within a factor of 2 . for external galaxies , a commonly - used tracer of star formation is far - infrared ( fir ) light . while fir is usually not the preferred tracer of star formation in normal spiral galaxies like the milky way , it is detectable over a very wide range of sources , from ultraluminous infrared galaxies ( ulirgs ) to individual cluster - forming clumps in the milky way , making it possible to look for correlations over a wide range of luminosities ( e.g. * ? ? ? * ; * ? ? ? @xcite finds that in optically - thick starbursts the ir luminosity and star - formation rate are related by @xmath46 the dominant uncertainty is the age of the stellar population , but as we discuss in more detail below , for extragalactic sources this is only @xmath17 . the uncertainty is considerably larger in normal spiral galaxies , where old stellar populations contribute a significant luminosity , and where only a fraction of the light is reprocessed into infrared . based on comparisons of multiple tracers of the star formation rate in a range of galaxy types , @xcite find that ( [ irsfrconversion ] ) probably underestimates the star formation rate in normal spirals . we instead adopt @xmath47 in normal spirals , which is approximately consistent with the mean in the sample of @xcite for low star - formation rate ( @xmath48 @xmath42 yr@xmath43 ) systems . there is , however , a much larger scatter in this relation than in the corresponding relation for starburst systems . several authors ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) use infrared luminosities to estimate the star formation rates within individual cluster - forming gas clumps in the milky way . these objects have very high surface densities , so essentially all the light is reprocessed into ir and the uncertainty for spiral galaxies does not apply . however , we will not directly compare to these data because there is a much larger uncertainty arising from the age of the stellar population . in systems @xmath49 myr old no stars have disappeared through supernovae , so the massive stellar population can not yet have reached equilibrium between formation and destruction . to study the magnitude of the uncertainty that this effect induces , we use starburst99 version 5.0 @xcite to compute the evolution of light to star formation rate and light to mass ratios for systems with constant star formation rates . figure [ lt ] shows the results . the dashed lines show models using the @xcite imf , while the dotted lines use an imf with the @xcite slope of @xmath50 from 0.1 @xmath42 to 120 @xmath42 ; for all other starburst99 parameters , we use the defaults . as the plot shows , at ages of a few myr or more , the luminosity per unit star formation rate is quite insensitive to both age and imf , so luminosity is a good indicator of star formation rate . however , at younger ages the luminosity in the starburst99 calculation traces total stellar mass more closely than star formation rate , so the light is not a good indicator of star formation rate . one might think it possible to break this degeneracy by independently determining the age of the stellar population , and then using the luminosity to infer the stellar mass and thus the star formation rate . however , a more detailed treatment of very young systems than starburst99 provides shows that infrared observations alone are not sufficient to constrain the stellar mass in systems @xmath51 myr old . in such young systems , a significant fraction of the stars may not yet have contracted to the main sequence , in which case they will be more luminous than equal mass stars on the main sequence would be . accretion luminosity may further enhance the radiative output above what would be found for a non - accreting population of the same mass . there are also factors that reduce the luminous output compared to an older population of the same mass . massive stars require @xmath52 myr to assemble @xcite , so systems younger than this will be missing their contribution to the light . in systems of @xmath53 @xmath42 or less , poor sampling of the massive tail of the stellar imf produces a large scatter , and causes the median system to be less luminous per unit mass or per unit star formation rate than a larger population would be . we explore how these effects change the light output of a young cluster using the model of @xcite . in figure [ lt ] , the solid lines show the light output from a simulated cluster with a final stellar mass of @xmath54 @xmath42 forming at a constant star formation rate of @xmath55 @xmath42 yr@xmath43 . this model uses the same imf we use for the salpeter - slope starburst99 calculation ( the dotted line in the figure ) , but it includes accretion luminosity , pre - main sequence evolution , finite star formation times , and discrete sampling of the imf . it does not include any post - main sequence evolution . the thick central line shows the median of 400 runs with different samplings of the imf , and the thin lines above and below it show the tracks that bound @xmath56 of the runs . as the plots show , the median cluster @xmath51 myr in age will be much less luminous per unit star formation rate than a galactic stellar population that is @xmath57 myr old and contains enough stars to fully sample the imf , but the spread can be an order of magnitude for a 1000 @xmath42 cluster . as the cluster mass increases the effects of discrete sampling decrease , causing the luminosity spread to decrease and the median to rise . even for clusters massive enough to sample the full imf , though , neither the light to mass ratio nor the light to star formation rate ratio stay constant at ages @xmath51 myr , so light is a poor indicator of either stellar mass or star formation rate . while we can not use infrared luminosity to study the star formation rate in galactic star - forming gas clumps directly , there is an observational correlation between infrared and molecular luminosity for such objects @xcite from which we can learn a great deal . we discuss how to interpret this correlation in light of our results in [ clustertime ] . finally , note that there are few star clusters where one can estimate the star formation rate directly by placing a large number of stars on the hr diagram and using pre - main sequence tracks to estimate the cluster mass and age spread ( e.g. * ? ? ? * ; * ? ? ? since this requires luminosity and temperature determinations for many stars , it is posssible only in systems without too much extinction , which limits this technique to low density regions ( e.g. taurus ) or somewhat older regions where most of the initial gas is gone ( e.g. the orion nebula cluster , onc ) . we discuss the onc in more detail in [ oncsec ] , and also refer readers to @xcite for a detailed discussion of other techniques by which one may estimate the formation times of star clusters . as discussed above , irdcs are likely to be the progenitors of star clusters . @xcite catalog the infrared dark clouds in the inner galaxy . this enables us to compute the star formation rate per free fall time in irdcs by comparing to the total star formation rate in the inner galaxy , roughly @xmath58 @xmath42 yr@xmath43 with a factor of 2 uncertainty . we should note that our calculation of the star formation rate in irdcs is entirely different from the one given @xcite . their estimate assumes an unknown time scale and efficiency of star formation . the calculation we give here does not rely on any such assumptions . first we must estimate the fraction of star formation that occurs in irdcs . @xcite find that @xmath59 of galactic star formation occurs in clusters . if a cloud were visible during the entire star formation process as an irdc , this would imply @xmath60 . however , clouds that have too many embedded protostars will not be infrared dark , and will therefore cease to be visible as irdcs . this probably does not occur until most of the stars have formed , though , so we adopt @xmath61 as a reasonable guess , while acknowledging that it could be a bit smaller . being slighly more conservative about our uncertainty than @xcite , we adopt @xmath62 as the plausible range of possibilities , i.e. anywhere from 60% to 100% of star formation inside the solar circle takes place in irdcs , corresponding to a factor of 1.25 uncertainty on our estimate of @xmath63 . note that @xmath64 gives the largest possible value of @xmath1 for irdcs , so any uncertainties that move @xmath65 out of our plausible range can only make @xmath1 smaller , not larger . @xcite find that the total mass of irdcs in the inner galaxy is @xmath66 @xmath42 based on the measured properties of a sub - sample of irdcs observed in 1.2 mm continuum emission , and an estimate of the detection efficiency for the msx irdc survey @xcite . this mass estimate is probably uncertain by a factor of several , because it is unclear how representative the clouds surveyed by @xcite are of the entire irdc population . the sample consists of the darkest clouds ( darkness measures a combination of column density and degree of background illumination ) from a sample with known kinematic distances . this selection introduces an unknown bias in the mass estimate , so we consider the irdc mass estimate to be uncertain by factors of a few . the free - fall time in irdcs is @xmath67^{1/2}$ ] , where @xmath3 is the mean density . for the 38 irdcs in the sample of @xcite , we define an effective radius @xmath68 , where @xmath69 is the angular area within the @xmath70 detection threshold of the cloud taken from the catalog of @xcite , and @xmath71 is the distance estimate taken from rathborne et al . we take the mean density to be @xmath72 . this procedure is fairly uncertain , since the location of the @xmath70 contour depends on the background emission , and the morphology is filamentary rather than round for a significant minority of clouds . nonetheless , we can make a rough estimate for @xmath3 and @xmath73 , and check for any systematic variations with irdc size . we plot the derived free - fall times and densities in figure [ tffirdc ] . as the plot shows , there is a spread of a factor of @xmath74 , but no clear systematic trend . we adopt as our characteristic free - fall time the mass - weighted harmonic mean @xmath75 corresponding to a number density of hydrogen nuclei @xmath76 @xmath7 . the range of free - fall times covering the central 67% of the irdcs is @xmath77 myr , so the uncertainty is roughly a factor of 2.6 . similarly , the central 67% of irdc densities covers a range of @xmath78 @xmath7 , an uncertainty of a factor of 6.6 . we now have all the necessary figures to plug into equation ( [ sfrffeqn ] ) : the fraction of star formation inside the solar circle taking place in irdcs is @xmath63 , with a @xmath79 uncertainty ; the total star formation rate inside the solar circle is @xmath58 @xmath42 yr@xmath43 , with a factor of 2 uncertainty ; the free - fall time in irdcs is @xmath80 myr , with a factor of 4 uncertainty ; the total mass of irdcs in the inner galaxy is @xmath81 @xmath42 , with a hard - to - quantify uncertainty , which we take to be a factor of 3 . combining these estimates gives a fiducial value @xmath82 . combining the errors as described in [ sfrest ] gives an uncertainty of a factor of 4.6 . the molecular transition hcn(1 - 0 ) has a critical density of @xmath83 @xmath7 ( c.f . * ? ? ? * who give the critical density in terms of number density of hydrogen molecules rather than hydrogen nuclei ) , where @xmath84 is the line - center optical depth in the escape probability approximation , and is therefore an excellent tracer of the dense molecular regions associated with star formation . @xcite observe a large sample of normal spiral galaxies and luminous and ultraluminous infrared galaxies in hcn(1 - 0 ) and show that the hcn luminosity correlates well with the ir luminosity , following @xmath85 where the figure quoted is the mean plus or minus the standard deviation . galactic cores with luminosities above @xmath86 @xmath87 observed in hcn by @xcite show a similar correlation , with almost exactly the same mean , although the median is lower than the median of the extragalactic sample by a factor of 2.5 . since infrared luminosity is a tracer of star formation rate , and hcn(1 - 0 ) is a tracer of molecular gas , this relation is a direct measure of the star formation rate per unit mass in gas of densities traced by hcn , and thus of @xmath1 in that gas . we use the difference in medians between the galactic and extragalactic samples as an estimate of the uncertainty factor in the ir - hcn correlation , although we note that this is almost certainly an overestimate for the extragalactic sample because that data set is much more tightly correlated than the galactic sample . to derive the total gas mass that corresponds to a given luminosity @xmath88 in the hcn(1 - 0 ) line , @xcite observe in hcn(1 - 0 ) a sample of star - forming clumps with known virial masses determined from optically - thin c@xmath89s emission . ( other methods of estimating the mass of these objects that do not assume virial balance , and make other assumptions about the emission , e.g. @xcite , @xcite , @xcite , give masses that are comparable to or better than the uncertainty in the mass estimate we adopt below . ) they find a median ratio @xmath90 , and a mean of 11 . @xcite perform radiative transfer calculations under a variety of assumptions about the temperature and distribution of the hcn - emitting gas , and arrive at conversion factors of @xmath91 . for normal spiral galaxies , we adopt a conversion of @xmath92 , the mean value measured by @xcite . to be conservative we estimate that the conversion is uncertain by a factor of 3 , corresponding to a range that is larger than the range of @xmath93 spanned by the various estimates . recent hco@xmath94 observations suggest that chemical changes triggered by x - rays may enhance the hcn abundance in galaxies with an agn @xcite . this reduces the mass estimate for a given hcn luminosity by a factor of @xmath95 in starbursts , most of which have at least a small agn . for starbursts we therefore modify our light to mass conversion to @xmath96 , again with roughly a factor of 3 uncertainty . combining this with the conversion from ir luminosity to star formation rate discussed in [ irsfsection ] for both starburst galaxies and normal spirals gives @xmath97 . adding the uncertainties of a factor of 2.5 on the ir - hcn correlation , 3 on the hcn to mass conversion , and 1.3 on the ir to star formation rate conversion gives a combined uncertainty of a factor of 4.1 . the free - fall time depends on the mean density , for which a rough guess based on the optical thickness of the hcn emission is @xmath98 @xmath7 . the true density of the hcn - emitting gas could be lower or higher depending on optical depth and beam - filling effects , as appears to be the case with cs(5 - 4 ) emission ( see the discussion below in [ cssection ] ) . however , this is unlikely because the mass to luminosity ratio for hcn(1 - 0 ) inferred from radiative transfer calculations is very close to that inferred from correlation of hcn luminosity with virial masses for objects seen in the galaxy . this suggests that the hcn emission for galactic hcn clumps is close to beam - filling and is not extremely self - absorbed , so the density of the objects being observed is unlikely to be very different than @xmath98 @xmath7 . if anything , the observations of @xcite suggest that the mean density may be slightly larger than @xmath99 @xmath7 ( also j. wu , 2006 , private communication ) . we adopt a mean density of @xmath100 @xmath7 as a fiducial value , giving @xmath101 myr , and estimate that the uncertainty is no more than a factor of 2 , corresponding to a factor of 4 change in the density . finally , since the extra - galactic observations include all hcn(1 - 0 ) emission from the target galaxy , and all star - formation occurs at densities high enough to be traced by hcn emission , we set @xmath102 . using our fiducial estimates for @xmath103 , @xmath104 , and @xmath105 in ( [ sfrffeqn ] ) gives @xmath106 . the factor of 4.1 uncertainty in our value of @xmath103 , combined with the factor of 2 uncertainty in our free - fall time , gives a complete uncertainty of a factor of 4.6 . the cs(5 - 4 ) line has a critical density of @xmath107 @xmath7 , and thus traces gas at even higher densities than hcn(1 - 0 ) . @xcite and @xcite survey in cs emission lines a sample of regions selected from sources reported in the arcetri h@xmath108o maser catalog @xcite that are thought based on iras colors to be associated with star formation . from the mean cs(5 - 4 ) luminosity of their targets and the sky coverage fraction of the maser catalog , they estimate that the total cs(5 - 4 ) luminosity of the galaxy is at least @xmath109 @xmath87 . this is a lower limit only , because the maser catalog may not be complete over the region of sky it covers , there is probably at least some galactic cs(5 - 4 ) emission that is not correlated with water masers , and @xcite only detect 75% of their targets in cs(5 - 4 ) , while they detect 90% in all the transitions for which they look . nonetheless , this lower limit on luminosity gives a lower limit on the amount of gas in the galaxy that is sufficiently dense to produce cs(5 - 4 ) emission , which we can in turn use in equation ( [ sfrffeqn ] ) to obtain an upper limit on @xmath1 at cs densities . first we must estimate the cs(5 - 4 ) `` x '' factor to convert the luminosity to a mass . like hcn(1 - 0 ) , star - forming clumps are usually optically thick in cs(5 - 4 ) , so it is reasonable to expect that such an x factor might exist . figure [ mvirlcs ] shows the virial mass @xmath110 ( determined from optically - thin c@xmath89s emission ) versus @xmath111 for the 57 objects in the @xcite sample . the correlation is well fit by the line @xmath112 over more than two orders of magnitude in mass . this implies a total galactic mass in cs clumps of @xmath113 @xmath42 . the uncertainty of the fit , done in the logarithm of the data , is a factor of 3.3 , and we adopt this as the uncertainty in our luminosity to mass correlation . however , as we discuss more below , the virial mass gives the lowest mass of the three estimators tested by @xcite , by roughly a factor of 3 , so the true mass is much more likely to be larger than our value rather than smaller . note that this mass estimate strongly suggests that the true cs mass in the milky way is substantially larger . cs and hcn emission often come from overlapping regions in the milky way @xcite , and in such overlapping regions the area emitting cs is not vastly smaller than the area emitting hcn . however , the total mass of cs gas that our estimate suggests is smaller than the mass of hcn - emitting gas that @xcite see in galaxies like the milky way by @xmath114 orders of magnitude . the most likely explanation is incompleteness of the cs survey . to estimate the free - fall time , we must know the gas density . @xcite , based on radiative transfer modeling , find a mean density of @xmath115 @xmath7 , roughly the critical density . however , this average is intensity - weighted , so it may overestimate the true mean . @xcite and @xcite find that the virial mass for objects in their survey is systematically smaller than the mass estimated by assuming that all the gas is at the density inferred from the radiative transfer calculations . based on the difference in mass estimates , they conclude that the filling factor of gas at densities of @xmath116 @xmath7 or higher is typically @xmath117 . thus , the density of @xmath115 @xmath7 probably represents an upper limit on the true density . we therefore compute the mean density for the 57 clumps in the shirley et al . catalog from the virial mass and deconvolved half - peak radius ( shirley et al.s @xmath118 ) to compare to the result of the radiative transfer calculations . we show our derived densities and the corresponding free - fall times in figure [ tffcs ] . the mass - weighted harmonic mean free - fall time for the cs clumps ( computed from equation [ harmonicmean ] ) is @xmath119 myr , and we adopt this as our fiducial value . the range @xmath120 myr covers the central 67% of the free - fall times , so we take the uncertainty to be a factor of 2 . the density corresponding to our adopted mean free - fall time is @xmath121 @xmath7 , and the range in free - fall times corresponds to a density range of a factor of 1.4 . finally , the estimated cs luminosity comes from sources associated with water masers , which arise only in regions of massive , clustered star formation . this implies a maximum of @xmath122 . alternately , we would obtain the same result by assuming that all star - forming regions produce cs(5 - 4 ) emission and raising our estimate of @xmath123 by a factor of @xmath124 to account for the fraction of cs(5 - 4 ) emission associated with water masers . in either case , our value for either @xmath125 or @xmath123 is an upper limit because we do not know what fraction of clustered star formation is associated with water maser emission . combining @xmath125 , @xmath123 , and @xmath126 in equation ( [ sfrffeqn ] ) gives @xmath127 , with an uncertainty of a factor of 3.6 . again , it seems likely that the true value is significantly lower than this upper limit . we can add one more point for a specific object . the orion nebula cluster ( onc ) is the only star cluster that has been studied well enough so that it is possible to make reasonable estimates of the stellar mass , the formation time , and the properties of the progenitor gas system , allowing direct determination of @xmath1 . this is quite useful because it provides an estimate of @xmath1 that does not depend on conversions from luminosities to masses or star formation rates , and thus is subject to completely different systematic errors than the methods we have used thus far . the total stellar mass of the onc is 4600 @xmath42 @xcite . @xcite analyze several lines of evidence to conclude that formation of the onc took place over at least @xmath128 myr , and possibly longer see @xcite . the age estimate comes from a combination of pre - main sequence fitting , which is fairly reliable for stars @xmath129 myr in age , and from the age of a dynamical ejection event . we refer readers to @xcite for the detailed evidence . the free - fall time in the current stellar system is @xmath130 myr , but it must have been smaller in the progenitor gas system , since some of the mass has been expelled by the trapezium stars , and the cluster may be expanding . uncertainties in these processes lead to a range of mass estimates for the progenitor , ranging from @xmath131 @xmath42 if the onc today is marginally unbound and expanding @xcite to @xmath132 @xmath42 if the onc is currently bound and non - expanding @xcite , although the latter would imply an extraordinarily high star formation efficiency . if we neglect the possibility that the cluster has undergone significant expansion since expelling its gas , this implies that the free - fall time in the progenitor gas system was @xmath133 myr , corresponding to a density @xmath134 @xmath7 . it seems unlikely that star formation began when most of the gas in the onc was spread over a much larger region at lower density , both because there is no correlation between the age and spatial distribution for stars in the onc @xcite and because there must have been a system sufficiently dense to lead to dynamical interactions and ejection of massive stars 2.5 myr ago @xcite . combining the initial gas mass , final stellar mass , and initial free - fall time implies that in orion @xmath135 . the true value is likely to be towards the low end of this range , since even if the onc is bound it has almost certainly undergone some expansion from its original gaseous state , meaning that the free - fall time we have adopted is probably too large . we summarize our results by plotting the dimensionless star formation rate @xmath1 versus characteristic density in figure [ sfrffn ] . in addition to our points for cs clumps , hcn clumps , irdcs , and the onc , we can add a point for gmcs as a whole , in which the typical density and free - fall time are @xmath136 @xmath7 and @xmath137 myr @xcite . the total mass of gmcs in the galaxy is roughly @xmath138 @xmath42 @xcite , so using the same argument as in [ irdcsfr ] gives @xmath139 . @xcite give a much more detailed calculation of this value , but for simplicity we adopt rough numbers here . since the gmc population in the milky way is reasonably well characterized , the uncertainty in @xmath1 is dominated by uncertainty in the star formation rate , so it is about a factor of 2 . the range of densities covered by star - forming gmcs is roughly a factor of 3 , so we take this range and the range in @xmath1 to roughly define our uncertainty for gmcs . the star formation rate per free - fall time is a key prediction of any model for a physical mechanism that regulates the star formation rate . the most important result summarized by figure [ sfrffn ] , which any successful theory must be able to explain , is that there is no evidence for a transition to rapid star formation out to densities of @xmath140 @xmath7 . even given the large uncertainties in the data , this conclusion is quite firm . if any of the objects we have analyzed were bound , detatched from the large - scale turbulent flow , and collapsing without significant impedance from feedback , we would expect @xmath12 , while even the top of the uncertainty range for the data point with the highest value of @xmath1 is below 0.1 . only the cs data is consistent with a transition for rapid star formation , and this is an upper limit for which there is strong evidence to suggest that the true value is significantly lower . a second conclusion , which is much more tentative , is that we do not see any significant change in @xmath1 with density at all . given the uncertainties , obviously we are insensitive to changes in @xmath1 smaller than roughly an order of magnitude , and the somewhat higher point for the onc suggests that it is possible there is some modest rise in @xmath1 in star clusters . any such rise with density , though , must be fairly small over the range of density we have explored . in summary , the data are _ consistent _ with @xmath1 of a few percent regardless of density , and _ require _ @xmath141 to densities of at least @xmath142 @xmath7 . the characteristic scale at which one transitions from slow to fast star formation must lie at even higher densities . here we investigate how well various models of star formation explain this result , and discuss a few of its implications . one proposed explanation for the low star formation rate in molecular clouds is that gmcs are not bound by gravity @xcite . in this model , gmcs are transient over - densities created by turbulence or spiral shocks in the atomic ism . since they are confined by ram pressure , they re - expand and disperse in roughly one dynamical time . the star formation rate appears low because only a small fraction of the mass is gravitationally bound and can collapse , but that mass which can collapse will form stars rapidly . feedback is not effective at slowing collapse or establishing an equilibrium , and only stops the star formation process at its end by rapidly unbinding the cluster @xcite . star formation in protoclusters is rapid but is halted after a short period by feedback . thus , the star formation rate is high , even if the duration is short , and this should be detectable as a rise in @xmath1 at the typical density of protoclusters . simulations based on these premises allow us to compare these models to the observational data . before discussing individual simulations , we note that we can not analyze the simulation data as thoroughly as we have analyzed the observational data . ideally , one would take a snapshot of a simulation at various points in time , compute the mass @xmath20 that is denser than @xmath3 , make a plot of @xmath143 , and compare that to the instantaneous star formation rate in the simulation . this is effectively what we have done with the observational data . however , simulation papers generally do not report @xmath20 . they only give a single density , usually the initial ( uniform ) density , and report a total mass of stars formed over the course of the entire simulation . thus we perform our calculations using @xmath20 for the initial density at the initial time , and compare the mean star formation rate over the entire simulation . this should give a rough average value that can stand in for a more detailed calculation of @xmath143 as a function of density and time . however , we encourage authors of future simulations of gmc evolution or star cluster formation to compute @xmath143 for at least a few time slices to facilitate comparison to observations . @xcite estimate from their simulations that unbound gmcs would convert @xmath144 of their mass into stars in @xmath145 free - fall times , giving @xmath146 . in contrast , @xcite simulate a marginally - bound @xmath54 @xmath42 clump with an initial radius of @xmath147 pc , giving @xmath148 @xmath7 , roughly the properties of an hcn clump . there is no feedback , and the turbulence decays freely . they find that , after @xmath149 initial free - fall times , @xmath150 of the mass has gone into stars , giving @xmath151 . this is probably an underestimate of the true value of @xmath1 in this scenario , since the simulation starts with a uniform density and thus star formation does not really begin for roughly 1 free - fall time . if we instead measured from the time when the first star formed to the time when the last one formed , we would find an even larger value of @xmath1 . this is consistent with other simulations in which the turbulence is weak . for example , @xcite model formation of a star cluster by simulating a periodic box in which the gravitational potential energy greatly exceeds the kinetic energy , and find that their simulation converts @xmath152 of the initial mass into stars in 1.8 free - fall times , giving @xmath153 . the simulation is scale - free , but klessen & burkert suggest that the model should be reasonable for a region with a typical density of @xmath154 @xmath7 . @xcite perform a parameter study of regions with turbulent driving of various strengths and find that when the turbulent driving is weak so that the system is unstable to global collapse , @xmath155 is a typical value . ( as we discuss below in [ virturb ] , @xcite find that if the turbulent driving is strong , the simulations produce lower values of @xmath1 that agree much better with the observations . ) we place points from the simulations of @xcite , @xcite , and @xcite , and from one of the weakly - driven turbulence simulations of @xcite , on figure [ sfrffn ] . this indicates a clear observational problem for these models . they produce roughly the correct star formation rate at the gmc scale , but are inconsistent with the data on the irdc and hcn scale at the order of magnitude level . they are consistent with the cs data only if the value of @xmath1 in cs clumps is near its upper limit . we conclude that models where star formation occurs in unbound gmcs but freely collapsing dense clumps , or where the turbulence in those clumps is driven but at very sub - virial levels , are ruled out by the data . star formation in dense clumps can not occur through global collapse . we emphasize again that one can not avoid this problem by hypothesizing a feedback process that rapidly destroys clumps once they have turned a relatively small fraction of their mass into stars , but retaining the picture of star formation occuring in clumps that are undergoing global collapse . even if feedback destroyed collapsing clumps , while they existed they would still be forming stars at a rate much larger than @xmath1 of a few percent , which is inconsistent with the data . only if destruction by feedback prevents global collapse from starting can it produce the observed value of @xmath1 . in this case , though , estimates that cluster - forming clumps turn @xmath17 of their gas into stars @xcite require that global collapse be held off for much longer than a free - fall time . we discuss the implications of this in more detail in [ clustertime ] . one final caveat worth mentioning in our interpretation of these simulations is that once a significant fraction of the particles in an sph simulation have been accreted by sink particles , the effective resolution of that code , and its ability to follow the hydrodynamics , may be significantly degraded . it is unclear how large this effect is , and whether it would tend to increase or decrease @xmath1 . however , since the effect is only a problem late in the simulations , it seems unlikely that it could shift @xmath1 by more than a few percent . the criticism of unbound gmc models applies in a much weaker form to models in which the star formation rate is determined by large - scale global gravitational instability in a galactic disk ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? these models postulate that the star formation rate in galaxies is determined by the rate at which gravitational instability in the galactic disk drives gas to high densities , and that the strength of this gravitational instability determines the star formation rate . this is probably at least partly true , in that the star formation rate is set partly by the supply of molecular gas , which is in turn controlled by gravitational instability . however , roughly one third of the mass of the milky way s ism is in molecular clouds . clearly the rate - limiting step in star formation is not at the stage of converting atomic ism to molecular gas , but somewhere at higher density . for this reason , the gravitational instability models provide only a partial description of the star formation process . in practice , this limitation is reflected in the star formation prescriptions that simulations of galactic - scale gravitational instability adopt . these simulations can not resolve structures of molecular cloud size or smaller . instead , when such structures form the simulations use a recipe to determine the amount of gas that turns into stars . these prescriptions only reproduce the correct star formation rate if they correspond to rates @xmath156 in the dense gas . for example , @xcite assume for their highest resolution simulations that in cells with densities above @xmath157 @xmath7 , 50% of the gas is converted into stars per free - fall time of that gas . this corresponds to @xmath158 . however , examining their figure 8 shows that this leads to a star formation rate that is an order of magnitude larger than the observed @xcite law without feedback , and a half an order of magnitude too large if they include feedback . they also find that even reducing their value of @xmath1 by a factor of 10 is not by itself sufficient to reproduce the kennicutt law . this is not surprising based on our results that @xmath1 is at most a few percent in gas at densities of @xmath157 @xmath7 . reproducing the observed star formation law would require a value for @xmath1 of this order . in the simulations of @xcite , the authors are able to reproduce the @xcite law by adopting a prescription for star formation whereby collapsed mass is assumed to form stars with a 30% `` efficiency '' , in the sense that they count @xmath159 of the gas that is accreted onto sink particles in the simulation as stars . the remaining 70% of the mass stays in the sink particles , but is assumed not to form stars . there is no simple way to convert this recipe into a value of @xmath1 , since it specifies that the star formation rate in any particle is 30% of the particle s current accretion rate , and we do not have available the full time history of accretion for each particle . even if we did have this information , interpretation would be diffuclt because the gas assumed not to form stars remains locked up in sink particles indefinitely , rather than being ejected back into the ism by feedback as happens in real star clusters . this greatly reduces the supply of gas available to form stars . nonetheless , the fact that these simulations need to adopt a subgrid model that significantly reduces the collapsed mass counted as stars in order to reproduce the @xcite law is consistent with our point that one can only explain the observed rates of star formation if there is some mechanism that inhibits star formation even in the densest gas that the simulations can resolve . the reason that these simulations reproduce the observed star formation rate only if they have @xmath156 is clear in light of figure [ sfrffn ] : if the gas in irdcs or hcn gas were collapsing on its free - fall time , the star formation rate in the milky way would exceed the observed value by two orders of magnitude . to obtain the correct overall star formation rate , one must adopt a star formation rate @xmath156 even in the densest gas that the simulations are currently capable of resolving . while this is an entirely reasonable approach in simulations , a theoretical understanding of the star formation rate requires an explanation why @xmath156 even in gas that is already collapsed to densities above the critical density of hcn(1 - 0 ) , five orders of magnitude higher than the mean in the galactic ism . this point does not mean that global gravitational instability is unimportant in regulating star formation , simply that it can not be the sole agent . thus , models such as these are complementary to models of @xmath1 in dense gas such as those of @xcite . another possible explanation for low star formation rates is strong magnetic fields ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) . if star - forming clouds are magnetically subcritical , which is controversial on both observational and theoretical grounds @xcite , then they can not collapse before ambipolar diffusion allows the magnetic field to slip out of the gas , and star formation proceeds on the ambipolar diffusion time scale instead of the free - fall time scale . for a uniform gas , the time scale required for ambipolar diffusion to decouple the gas and the field is @xcite @xmath160 where @xmath161 and @xmath3 are the characteristic size scale and density of the cloud , @xmath162 is the alfvn speed , @xmath163 @xmath164 g@xmath43 s@xmath43 is the ion - neutral drag coefficient , and @xmath165 @xmath166 g@xmath167 is the cosmic ray ionization constant . ( in high column density environments like irdcs , interstellar uv photons can not penetrate and ionization is dominated by cosmic rays . ) if we assume that this describes the star formation timescale , then the star formation rate should obey roughly @xcite @xmath168 where @xmath169 is the fraction of the mass in a prestellar core that reaches the star rather than being ejected by outflows @xcite . to determine what this implies for @xmath1 , we need to know how the magnetic field behaves in objects of varying densities . while there are very few direct observations of magnetic fields for extremely dense objects of the sort we are considering , observations do show that magnetic fields obey the correlation @xmath170 , where @xmath171 is the velocity dispersion of the region , @xmath172 is the ratio of the object s mass - to - flux ratio to the critical value @xmath173 , @xmath174 is a constant of order unity the depends on the cloud s internal density distribution , and @xmath175 @xcite . this correlation is what one would expect if , in one direction , cloud self - gravity were balanced by magnetic plus turbulent pressure . if we use this correlation in ( [ sfrffmag ] ) , and re - write the relation in terms of the virial parameter @xmath176 , we find that all the dependence on dimensional quantities drops out and we are left with @xmath177 the uncertainty of this calculation is probably more than an order of magnitude , so we should not pay particular attention to the coefficient , and we will not attempt to place points for magnetic regulation models on figure [ sfrffn ] . major contributors to the uncertainty are ambiguities in the definition of the length and mass scales @xmath161 and @xmath178 , the lack of three - dimensional numerical simulations to determine how well equation ( [ sfrffmag ] ) holds on scales larger than a core collapsing to form a single star system , and the possibility that the cosmic ray ionization rate may vary substantially between galaxies . perhaps most significantly , equation ( [ tadest ] ) assumes that ambipolar diffusion in a turbulent medium is not substantially faster than in a quiescent one , an assumption that may well fail ( e.g. * ? ? ? * ; * ? ? ? if turbulence shortens the ambipolar diffusion time scale so that @xmath179 , as some simulations suggest it might , then magnetic regulation models would produce a value of @xmath1 that is too high . whether this occurs or not is a question that will have to be addressed through three - dimensional simulations with self - gravity and non - ideal mhd . given these uncertainties , we can not really say whether the rate of star formation in magnetically subcritical clouds subject to ambipolar diffusion is quantitatively consistent with the observations . however , the lack of dependence of @xmath1 on any properties but the virial parameter implies that we expect the magnetic @xmath1 to be roughly the same in all virialized objects , a prediction we can compare to observations . for our observed objects , we know that gmcs with masses @xmath180 @xmath42 , hcn clumps , and cs clumps are roughly virialized @xcite , so @xmath181 for them . for irdcs we lack kinematic information from optically thin molecular emission , and therefore we can not directly determine the velocity dispersion and the virial parameter . if we assume that these objects are virialized , then the magnetic regulation model is broadly consistent with observations that @xmath1 is roughly constant . we can not make a stronger statement than this because the magnetic regulation model can not currently make more specific predictions . note that it is not critical for this purpose that the objects in question truly be in exact virial equilibrium , since they almost certainly are not . we are using the virial parameter only as a way of parameterizing the strength of the turbulence relative to the strength of gravity , so all that we require is that the objects have virial parameters @xmath182 , i.e. that they not be either completely collapsing or completely non - self - gravitating . a fourth idea to explain the low star formation rate is that turbulence inhibits collapse . this idea has a long history , and we refer readers to @xcite for a thorough review . recently , @xcite synthesized simulations and analysis of turbulence - inhibited star formation by @xcite and @xcite to derive an estimate that turbulence produces a star formation rate @xmath183 , where @xmath184 and @xmath185 are the virial parameter and mach number of a star - forming gas cloud . the estimate is based on a derivation of the fraction of mass that is unstable to collapse in a medium that has the density and velocity structure common to supersonic isothermal turbulence , and is calibrated against simulations . its uncertainty is probably a factor of a few , stemming from uncertainty in the effects of magnetic fields and from the uncertain approximation @xmath186 . for gmcs , hcn clumps , and cs clumps , typical values are @xmath187 and @xmath188 @xcite . as with the magentic models , exact virial equilibrium is not required , simply that the kinetic and potential energies be comparable . we plot the predicted value of @xmath1 for this range of parameters in figure [ sfrffn ] . given the uncertainties in both the observations and the theoretical calculation , there is reasonable agreement . simulations of star formation in turbulent gas for which the turbulent energy is roughly equal to the gravitational potential energy , either because it is driven artificially or because the simulations include feedback from protostellar outflows , also give values of @xmath1 that are broadly consistent with the data . we show three examples in figure [ sfrffn ] : @xcite simulate a periodic box in which the kinetic and turbulent energies are approximately equal and the turbulence is driven to keep the level of turbulence roughly constant . they find a star formation rate of @xmath189 at a resolution of @xmath190 cells . ( since the simulation uses periodic boundary conditions , the true gravitational potential energy is not well - defined , so we can not state that the virialization is more than approximate . ) the simulations are dimensionless , but they suggest scaling to a typical density of @xmath191 @xmath7 . similarly , in scale - free simulations with strong turbulent driving , @xcite and @xcite find @xmath34 . @xcite find @xmath192 in a simulation with a central density of @xmath193 @xmath7 that includes protostellar outflows from forming stars , which keep the kinetic and gravitational potential energies approximately equal . a number of other simulations in which the turbulence is roughly in virial balance with the gravitational energy produce star formation rates in reasonable agreement with observations over the period in the simulations for which that balance is maintained . if the turbulence is sufficiently strong , simulations can achieve @xmath34 ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? we note that @xmath1 by itself does not appear to provide a way of distinguishing the length scale at which turbulence is driven , since simulations in which turbulence is driven primarily at large scales and ones in which it is driven primarily at small scales both seem capable of producing low values of @xmath1 . some authors ( e.g. * ? ? ? * ) emphasize the length scale of the driving as critical to determining whether the turbulence can keep @xmath1 low , and @xcite show that the characteristic distinguishing driving that gives @xmath12 from driving that produces @xmath156 is how the jeans length in the gas compares to the sonic length of the turbulence , defined as the length scale for which the turbulent velocity dispersion equals the sound speed . other diagnostics , such as the imf and clustering statistics ( e.g. * ? ? ? * ; * ? ? ? * ) , the linewidth - size relation of the gas ( e.g. * ? ? ? * ) , and the line luminosity of molecular clouds @xcite suggest that turbulence in molecular clouds is primarily driven on large scales . in this case @xcite show that the condition distinguishing cases with @xmath12 from those with @xmath194 reduces to the statement that clouds with @xmath195 have low values of @xmath1 , while those with @xmath196 have @xmath12 . thus , the combination of observations of @xmath1 and those showing that most turbulent energy is on large scales imply that turbulence can only reproduce the observed values of @xmath1 if it is driven strongly enough to produce @xmath182 . another important implication of this work is for the formation timescale of star clusters . the star formation efficiency of clusters , defined as the fraction of the initial gas mass that forms stars over the lifetime of the gas clump , is thought to be @xmath197 @xcite . since the observational data show that @xmath1 is at most a few percent , this implies that the cluster formation process must take at least @xmath198 ( @xmath199 crossing times ) . the recent observation by @xcite that individual hcn clumps in the galaxy with ir luminosities @xmath200 @xmath87 lie on the same @xmath201 correlation as entire galaxies also provides indirect evidence that the cluster formation time scale is comparatively long . wu et al . suggest that luminous hcn clumps fall on the galactic ir - hcn correlation because they contain enough stars to sample the imf fully . our results showing a large scatter in the light output for small clusters support this conjecture . however , our results on the age - dependence of the luminosity suggest that a mass large enough to sample the imf is not sufficient by itself to place a cluster on the extragalactic ir - hcn correlation . since star formation in the galaxies surveyed by @xcite has been ongoing for @xmath202 myr , their ir luminosities trace the star formation rate , and the @xmath201 correlation is therefore a statement of the star formation rate per unit mass in hcn gas . if the galactic hcn clumps observed by @xcite typically survived for only @xmath14 free - fall time , @xmath203 myr , then as figure [ lt ] shows their luminosity per unit star formation rate would be quite different than that of an older galactic population , and they would not lie on the galactic correlation . one would only expect galactic hcn clumps to follow the extragalactic @xmath204 correlation if they are massive enough to sample the imf well _ and _ if they are @xmath205 myr old , which means that they must have existed for at least @xmath206 , and probably closer to @xmath207 . these two lines of evidence provide additional support for the argument that star clusters form in near - equilibrium gas clumps presented by @xcite . tan et al . discuss the implications of this finding in more detail , but we note here two of the most significant . first , formation times of @xmath208 or longer appear inconsistent with models in which turbulence in protocluster gas is freely decaying . simulations show that if turbulence is allowed to decay freely ( e.g. * ? ? ? * ) or never contained energy comparable to the gravitational potential energy ( e.g. * ? ? ? * ) , the star formation process generally ends in only about @xmath209 , with @xmath210 of the mass in stars . only if the turbulence is continually driven can there be enough gas left up to @xmath207 for the star formation process to continue ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? the long cluster formation time scales suggested by the observations therefore imply that either the simulations of decaying turbulence are incorrect or ( more likely ) the turbulence is continually driven . this is consistent with on observations by @xcite and @xcite , and with simulations of protostellar feedback by @xcite . we note again that the observations of @xmath1 alone put no constraints on how or at what scale the turbulence is driven . it can be from internal sources of feedback on small scales , or from a turbulence cascade in which most of the power is on the size scale of an entire gmc . a second implication is for the mechanism responsible for the stellar initial mass function . @xcite have recently emphasized that the competitive accretion mechanism ( * ? ? ? * and references therein ) can only operate in the context of a strongly sub - virial gas clump that is undergoing global collapse and converting order unity of its gas mass into stars in a free - fall time . @xcite concur that the question of whether competitive accretion occurs or not depends on the extent to which gas and stellar velocities in a forming star cluster remain tightly coupled for the majority of the formation process , which in turn depends on whether feedback is effective at driving turbulence , which would change gas velocities but not stellar velocities . if turbulent driving keeps clusters virialized for many free - fall times and limits @xmath1 to @xmath211 , then competitive accretion can not occur within those clusters . the observations we discuss here provide evidence for low values of @xmath1 and long cluster formation times , and therefore provide strong evidence against the possiblity that competitive accretion determines the imf . we present observational evidence for two surprising conclusions , one shown quite strongly and the other more tentatively . first , and very clearly , _ star formation in dense gas is slow . _ the time required to convert all the gas into stars , the depletion time , is longer than the free - fall time by at least and order of magnitude , and probably closer to two orders of magnitude . there is no evidence that gas at densities up to @xmath212 @xmath7 resides primarily in objects that are undergoing collapse and rapid star formation . second , and more tentatively , _ the ratio of free - fall time to the depletion time is independent of the characteristic density of the star - forming object in question_. the data at present are uncertain , and are insensitive to changes of less than roughly an order of magnitude . nonetheless , it is interesting that this ratio varies by no more than an order of magnitude from the typical density of gmcs to the typical density of hcn(1 - 0)-emitting gas , a range of nearly 3 decades in density . these observations are a strong constraint for theories of star formation , and seem difficult to explain in the context of models in which there is a transition from slow , unbound star formation to rapid , bound star formation somewhere between the gmc scale and the protocluster scale . slow star formation in cluster - forming gas also implies that clusters require many free - fall times to assemble , as recently argued by @xcite on other grounds . models in which star formation takes place in virialized objects and is inhibited by strong magnetic fields are qualitatively consistent with the data , and models in which star formation is inhibited by turbulence are both qualitatively and quantitatively consistent with observations . in the future it would be extremely useful to improve the data on which figure [ sfrffn ] is based . one way to do this would be to perform detailed studies of other young clusters and obtain data comparable to that for the onc . this would provide a method of estimating @xmath1 that is independent of luminosity conversions and does not suffer from concerns about the completeness of galactic surveys . another improvement in the data could come from an unbiased survey of cs(5 - 4 ) emission in the milky way or in another galaxy , which would allow us to replace the upper limit on @xmath1 we derive here with an actual estimate . this would be particularly valuable because the upper limit on @xmath1 from cs(5 - 4 ) is well above the estimate from hcn(1 - 0 ) , even though the densities are not very different . unbiased cs observations could likely bring down this point . another improvement would be extend the data to higher densities . to accomplish this will require observations either of external galaxies or relatively complete surveys of the milky way in molecular transitions that trace densities @xmath213 @xmath7 . determining masses from the luminosities in these transitions will probably require high resolution follow - up observations of galactic sources in optically thin isotopomers so that the luminosity may be correlated against a virial mass . while this is a significant observational challenge , such surveys might make it possible to identify a scale at which star formation transitions from slow to fast , a crucial datum in understanding the star formation process . we thank n. evans , c. mckee , and j. wu for helpful discussions , and the referee , ralf klessen , for comments that improved the quality of the paper . support for this work was provided by nasa through hubble fellowship grant hsf - hf-01186 awarded by the space telescope science institute , which is operated by the association of universities for research in astronomy , inc . , for nasa , under contract nas 5 - 26555 ( mrk ) .

in recent years moment - closure approximations ( ma ) of the chemical master equation have become a popular method for the study of stochastic effects in chemical reaction systems . several different ma methods have been proposed and applied in the literature , but it remains unclear how they perform with respect to each other . in this paper we study the normal , poisson , log - normal and central - moment - neglect mas by applying them to understand the stochastic properties of chemical systems whose deterministic rate equations show the properties of bistability , ultrasensitivity and oscillatory behaviour . our results suggest that the normal ma is favourable over the other studied mas . in particular we found that ( i ) the size of the region of parameter space where a closure gives physically meaningful results , e.g. positive mean and variance , is considerably larger for the normal closure than for the other three closures ; ( ii ) the accuracy of the predictions of the four closures ( relative to simulations using the stochastic simulation algorithm ) is comparable in those regions of parameter space where all closures give physically meaningful results ; ( iii ) the poisson and log - normal mas are not uniquely defined for systems involving conservation laws in molecule numbers . we also describe the new software package moca which enables the automated numerical analysis of various ma methods in a graphical user interface and which was used to perform the comparative analysis presented in this paper . moca allows the user to develop novel closure methods and can treat polynomial , non - polynomial , as well as time - dependent propensity functions , thus being applicable to virtually any chemical reaction system .

1508.01737

biochemical reactions systems frequently comprise species with low copy numbers of molecules which leads to strong stochastic effects @xcite . under well - mixed and dilute conditions , the chemical master equation ( cme ) is the accepted description of the dynamics of such systems @xcite . for all but the most simple systems , however , no analytic solutions of the cme are known . the standard approach in this case is to use the stochastic simulation algorithm ( ssa @xcite ) , a popular monte - carlo method that samples from the solution of the cme . however , the ssa is computationally expensive and becomes infeasible for all but the smallest systems , in particular if some of the species occur in high molecule numbers with many reactions happening per unit time . while the derivation of a reduced cme enforcing time scale separation may help in some cases @xcite , analytical approximations are still an important alternative for the exploration of chemical systems . using the cme one can derive ordinary differential equations for the moments of the numbers of molecules of each species in the system . in general the equation for a given moment is coupled to the equations of higher order moments giving rise to an infinite hierarchy of equations which can not be solved @xcite . a popular method to approximate the moments of the cme are moment - closure approximations ( ma ) @xcite . the latter usually express moments above a certain order in terms of lower order moments , thereby closing the moment equations which can then be solved either analytically or numerically . several different moment - closure methods have been proposed in the literature . the most popular is the normal ma ( also called cumulant neglect ma " ) , which sets all cumulants above a certain order to zero , thus corresponding to a normal distribution @xcite . if all cumulants above order @xmath0 are set to zero we speak of the `` normal ma of order @xmath0 '' . several other mas have been proposed to close the moment equations ; some common types are the poisson ma @xcite , the log - normal ma @xcite and the central - moment - neglect ma ( cmn - ma ) @xcite . the purpose of this paper is twofold : ( i ) an empirical comparison of the predictions of different types of mas when applied to chemical reaction systems , and ( ii ) the presentation of a new user - friendly software package which enables the automatic derivation and analysis of mas . mas are an ad - hoc approximation and there is no straightforward way to predict their accuracy . while several different ma methods have been proposed @xcite and successfully applied @xcite in the literature , there are few studies analysing and comparing their performance . in @xcite the log - normal ma was found to be more accurate than the normal ma for a gene cascade network for one parameter set . in @xcite the accuracy of the normal ma has been investigated for general monostable systems in the limit of large volumes using the system size expansion . however , the accuracy of mas for small to intermediate volumes remains unknown and in particular how different ma methods perform with respect to each other . moreover , it is unknown under which conditions mas give physically meaningful results . in an empirical study @xcite formulated a set of validity conditions guaranteeing mas to give physically meaningful approximations to the moments of the cme . we will adopt these validity conditions here . specifically , whenever the cme has a stationary solution , we require the mas to have a single positive and globally attractive fixed point , and their time trajectories to stay non - negative and finite for all times and all initial conditions . in @xcite it was found that the normal ma fails to satisfy these validity conditions for certain systems and parameter regimes . it was shown that the normal ma can give rise to unphysical behaviour outside of this regime , such as negative mean values or variances , divergent time trajectories , unphysical oscillations and unphysical bistability , thus not allowing for a physical interpretation in these cases . it remains unclear if this is also the case for other moment - closure schemes and how their ranges in parameter space for which they are valid ( if they exist ) compare to each other . in this article , we apply the normal , poisson , log - normal and cmn - mas to several chemical reaction systems . we confine our analysis to mas of second order , since these are the most used in practice . we study their qualitative behaviour in the sense of the validity conditions stated before and compare their quantitative accuracy with exact stochastic simulations @xcite . it should be stressed that `` validity '' and `` accuracy '' are not unrelated properties , since one can only speak of a method s accuracy when it gives physically valid results . yet physically meaningful results can be quantitatively highly inaccurate . therefore , `` validity '' is a necessary but not sufficient condition for `` accuracy '' . in this study , we first use the different ma methods to study stochasticity in a system whose large volume limit is deterministically bistable . next , we investigate how well the ma methods can capture the influence of noise in a protein - phosphorylation system whose deterministic system shows ultrasensitivity . and finally , we use the mas to study the role of stochasticity in a system whose deterministic system is oscillating and which becomes entrained by an external force for a finite time interval . the derivation of the moment equations from the cme and the subsequent application of moment - closures is conceptually a straightforward task . practically , however , it becomes extremely cumbersome if more than one species is involved and if one considers higher - order mas . suppose for example a system of three species for which we want to compute the fourth - order normal ma equations . taking symmetries into account , this leads to 34 moment equations which have to be derived from the cme . these will have to be closed , and several fifth - order moments ( and potentially higher - order moments ) will have to be replaced in terms of lower - order moments . obviously , this task quickly becomes unfeasible to do manually . moreover , the numerical analysis of ma equations is not straightforward , and there is no user - friendly software package available allowing non - expert users to derive and analyse mas . to our knowledge , there are three software packages available in the literature for moment - closures : the matlab toolbox stochdyntools @xcite which allows the derivation of ma equations using several different closure schemes for mass - action chemical systems , i.e. , those with polynomial propensity functions ; the python package momentclosure @xcite which allows the same but only using the normal moment closure and has the facility to export the ma equations to a maple file for further analysis ; and a matlab toolbox presented in @xcite which allows to use normal moment closure to second order for mass - action chemical systems . for the application of all three packages , the user needs to be familiar with the respective programming language and the numerical analysis is not automated . in this article we present the mathematica package moca ( moment - closure analysis ) which was used for the presented numerical analysis . moca significantly extends the applicability and functionality of the two software packages stochdyntools and momentclosure @xcite as well as the software package presented in @xcite . it allows the non - expert user to apply and compare different moment - closure schemes in a graphical user interface ( gui ) without any coding necessary . it implements the normal , poisson , log - normal and cmn - ma and in addition allows the user to define his own novel moment - closure schemes . it extends the applicability to reaction systems with non - polynomial or time - dependent propensity functions . these can either be functions in time or given by discrete time points , for example obtained from experiments . all functions are available either in a gui or as code version for more experienced users , making the usage of moca maximally flexible . moca can perform steady state analysis with parameter scans , numerically integrate the ma equations in time and allows to export tables and figures to various commonly used formats . this paper is structured as follows . section [ sec_methods ] introduces the theoretical background for general moment - closure schemes and defines the particular ma methods analysed in this work . the numerical analysis of the various mas is then presented in section [ sec_numerical_analysis ] . next , we introduce the software package moca in section [ sec_moca ] . we explain the user input format and demonstrate its capabilities . we finish by summarising our results and concluding in section [ sec_conclusion ] . consider a chemical reaction system with species @xmath1 ( @xmath2 ) and @xmath3 chemical reactions : @xmath4 here , @xmath5 is the rate constant of reaction @xmath6 . we define the elements of the stoichiometric matrix @xmath7 as @xmath8 under well - mixed and dilute conditions the dynamics of the system is governed by the chemical master equation ( cme ) @xcite : @xmath9 here , @xmath10 is the joint probability distribution at time @xmath11 , where @xmath12 is the state vector of the system and @xmath13 is the number of molecules of species @xmath1 . @xmath14 is the @xmath15th column vector of the matrix @xmath7 and @xmath16 is the propensity function of reaction @xmath15 . for reactions described by the law of mass - action , the propensity is polynomial and defined as @xcite : @xmath17 here , @xmath18 denotes the volume of the compartment in which the reaction occurs . if in addition @xmath19 , which basically means that not more than two molecules react which each other in a single reaction ( at most a second - order reaction ) , we call reaction @xmath6 an `` elementary reaction '' . higher - order reactions do not really occur under conditions found in living cells and although they can often give a useful description of a system , they should really be interpreted as an effective approximate description of a set of elementary reactions , valid only under certain conditions . multiplying with @xmath20 and summing over all @xmath13 @xmath21 leads to the time evolution equation of the moment @xmath22 : @xmath23 here , @xmath24 denotes the expectation with respect to @xmath10 . accordingly , the first two moments obey @xmath25.\end{aligned}\ ] ] we see that , unless all @xmath16 are a zeroth or first - order polynomial in @xmath26 , the evolution equation of a certain moment depends on higher order moments , i.e. , the equations are not closed . the idea underlying the class of moment - closure approximations studied in this work is to express all moments above a certain order @xmath0 as functions of lower - order moments . the latter is typically done by assuming the distribution of the system to have a particular functional form , for example a normal distribution . this decouples the equations of the moments up to order @xmath0 from higher - order moments , which then allows one to numerically integrate the moment equations . we refer to such a moment - closure as `` ma of order @xmath0 '' . let @xmath27 denote the raw or `` normal '' moments , central moments and cumulants of order @xmath28 of the system , respectively . we call @xmath29 a `` diagonal moment '' if @xmath30 for all @xmath31 , and a `` mixed moment '' otherwise , and similarly for central moments and cumulants . in eq . @xmath32 is the cumulant generating function defined as @xmath33 we note that all three types of moments are respectively invariant under permutations of their indices . therefore , only one representative combination of each permutation class has to be considered . taking this symmetry into account significantly reduces the number of variables and moment equations . we adopt here the convention that the indices are ordered from small to large , i.e. , for a moment @xmath29 we have @xmath34 . expressing the moment - closure functions in terms of cumulants rather than raw moments often gives shorter expressions . the equations for the cumulants can then be rearranged to give equations for the raw moments . we consider here the following ma methods : * normal moment - closure " ( also called cumulant neglect moment - closure " in the literature ) : all cumulants above order @xmath0 are set to zero , i.e. , @xmath35 * poisson moment - closure " : the cumulants of a one - dimensional poisson distribution are all equal to the mean value . we assume here the multi - variate distribution to be a product of uni - variate poisson distributions . accordingly , for the poisson ma of order @xmath0 we set all diagonal cumulants to the corresponding mean and all mixed cumulants to zero , i.e. , @xmath36 * log - normal moment - closure " : let @xmath37 and @xmath7 be the mean vector and covariance matrix of a multi - dimensional normal random variable . then the logarithm of the latter has a multivariate log - normal distribution and its moments can be expressed in terms of @xmath37 and @xmath7 as @xcite @xmath38 where @xmath39 , where @xmath40 is the number of @xmath41 s having the value @xmath42 . this allows to express @xmath37 and @xmath7 in terms of the first two moments @xmath43 and @xmath44 which then in turn allows to express higher - order moments in terms of the latter , too . * cmn moment - closure " : all central moments above order @xmath0 are set to zero : @xmath45 each of the equations can then be used to express the raw moments above order @xmath0 in terms of lower order moments and thus close the moment equations according to the corresponding ma . we note that the normal ma , poisson ma and cmn - ma can be equivalent depending on the reaction system ( see later for examples of such systems ) . the normal moment - closure has been used in the field of biochemical reactions for example in @xcite and is probably the most commonly used one . it has been considered together with the poisson and log - normal ma for the one - dimensional stochastic logistic model in @xcite . the log - normal moment - closure technique has been proposed in @xcite . in @xcite it has been shown that the assumption of a log - normal distribution is equivalent to a derivative matching " closure . the cmn - ma has also been called a low dispersion moment - closure " in @xcite . as an example , consider a reaction system of the michaelis menten type : @xmath46 where @xmath47 is the free enzyme , @xmath7 is the substrate , @xmath48 is the enzyme - substrate complex and @xmath49 the product . the sum of the numbers of @xmath47 and @xmath48 molecules is constant at all times since each enzyme is either in the free @xmath47 or complex @xmath48 state . let @xmath50 denote the total number of enzyme molecules . assuming mass - action kinetics , the propensity vector is given by @xmath51 where @xmath18 is the volume of the system and we have defined @xmath52 and @xmath53 . here , @xmath54 and @xmath55 denote the copy number of substrate @xmath7 and free enzymes @xmath47 , respectively , and we have used the fact that the number of complex molecules @xmath48 is @xmath56 to eliminate the corresponding variable . the stoichiometric matrix is defined in eq . and reads for system @xmath57 the corresponding cme is obtained by substituting eqs . and in eq . leading to @xmath58 multiplying with @xmath59 and @xmath60 and summing over all @xmath54 and @xmath55 gives the following equations for the first two moments @xmath61 recall that the moments are invariant under index permutations and thus @xmath62 does not have to be considered explicitly . we see that the equations of the mean @xmath63 and @xmath64 depend on the second moment @xmath65 . the equation of the latter depends on the third moments @xmath66 and @xmath67 and similarly the equations for @xmath68 and @xmath69 . it can easily be seen that this applies also to all higher order moments , i.e. , the time - evolution equation of a moment of order @xmath28 depends on moments of order @xmath70 . therefore , the system of equations is not closed and can not be solved directly . now consider the normal ma which sets all cumulants above a certain order to zero . if we aim at closing the equations to second - order , we have to set the third - order cumulants to zero @xmath71 expressing the cumulants in terms of raw moments , this allows one to find expressions of the third - order moments in terms of first and second - order moments . for @xmath66 , for example , this reads @xmath72 and similarly for the other third - order moments . replacing the third - order moments accordingly in eqs . - closes the equations . we give here the resulting equations transformed to central moments : @xmath73 we recently formulated validity conditions guaranteeing physically meaningful predictions of ma approximations @xcite and analysed the validity of the normal ma for several example systems . we briefly review these conditions here . for a system for which the cme has a stationary solution , the exact moments of the system converge to a single steady - state in the limit of long times . therefore , for the mas to be valid moment approximations , we require convergence to a single steady - state in the limit of long times , too . moreover , the trajectories should preserve a positive mean and even central moments in the molecule numbers for all times and for all sensible initial conditions . note that this is also the case for deterministic bistable systems and deterministic oscillatory systems . if the cme converges to a stationary solution , the resulting moments are unique , even if the deterministic rate equations are bistable . moreover , while single ssa trajectories oscillate for a deterministic oscillatory system , the moments of the cme converge to fixed points because single ssa trajectories get out of phase over time . in the following we analyse different mas with respect to these validity conditions and compare their quantitative accuracy with ssa simulations . in @xcite it has been shown that for the deterministic bistable schlgl model @xcite , the normal ma gives physically meaningful results only for an intermediate range of volumes . for smaller volumes it shows negative or diverging trajectories , while it becomes bistable for larger ones . the ssa , in contrast , has a globally attractive positive fixed point and non - negative time trajectories for all volumes . here , we study the stochastic properties of the minimal elementary reaction system whose rate equations show bistability @xcite @xmath74 we added the first reaction to the ones given in @xcite to prevent the stochastic system from having an absorbing state for zero molecule numbers . depending on the parameter values the deterministic rate equations become bistable for this system . all parameter sets used in this section are chosen such that this is the case . we assume mass - action kinetics here . since the reactions in eqs . - are of order two or lower , their rate functions are polynomials up to order two in the species variables . this means that the time evolution equations of the second - moments depend on the third - order moments , but not on higher - order moments . we thus have to express the third - order moments in terms of first and second - order moments to close the equations to second order . recall that the second - order normal and cmn - mas set all cumulants and central moments above order two to zero , respectively ( c.f . eqs . and ) . since the third - order cumulant and third - order central moment are identical , the second - order normal ma and cmn - ma are thus equivalent for the reaction system in eq . this is of course a general result , i.e. , for chemical reaction systems with elementary reactions and mass - action kinetics ( i.e. , reactions up to order two and polynomial propensity functions ) , the second - order normal ma and second - order cmn - ma are identical . we thus analyse the normal , poisson and log - normal ma here . on log - scale obtained from steady - state analysis for the bistable reaction system in eqs . - for the parameters @xmath75 and @xmath76 . we shift the points slightly to make coinciding points distinguishable . we find that all three mas give a physical result of a single positive stable fixed point only on an intermediate range of volumes . the latter is significantly smaller for the log - normal ma than for the normal and poisson mas . ] qualitatively , we find a similar behaviour for the three different ma methods . as for the bistable system analysed in @xcite using the normal ma , we find that there exists an intermediate regime of volumes for the three mas to be valid , i.e. , they have a single globally attractive positive fixed point , and we find that the moments become bistable ( and hence physically meaningless ) above this regime . interestingly , however , we find that when increasing the volume further all three mas become tristable , i.e. , have three positive stable fixed points , see figure [ fig_bistable_number_of_fixed_points ] . this means the mas have more positive stable fixed points than the rate equations here , the latter being bistable independent of the volume , and thus the mas have no physical interpretation anymore whatsoever . in @xcite it has been shown that for monostable systems , the normal ma becomes equivalent to the rate equations for the means in the limit of large volumes . one can easily show that the result also applies to the poisson , log - normal and cmn - ma . here , we find numerically that the tristability remains for volumes up to @xmath77 , which suggests that the convergence of the mas to the res in the limit of large volumes does not hold for deterministic bistable systems . figure [ fig_bistable_time_course ] shows the time trajectories for the mas for different volumes , verifying that the mas can indeed have one , two or three positive stable fixed points depending on the volume . and different initial conditions for the parameters @xmath75 and @xmath76 . the dashed and dotted lines indicate the respective positive and stable fixed points of species @xmath49 and @xmath78 . depending on the volume , the mas have one , two or three positive stable fixed points . ] the table in figure 3 lists the endpoints of the validity interval for the mas for ten different parameter sets on logarithmic scale . the figure below visualises these . we observe that the log - normal ma has a much smaller validity range than the other two mas . the normal and poisson ma most of the time have a similar upper bound while the lower bound is generally smaller for the poisson ma . we thus find that in terms of validity , the log - normal ma performs significantly worse than the other two ma schemes for the reaction system studied here . [ cols="^,^,^,^,^,^,^,^,^,^,^,^,^ " , ] on logarithmic scale for different parameter sets for the bistable reaction system in eqs . - . @xmath79 and @xmath80 denote the left and right end of the validity interval , respectively . we have only checked for fixed points down to a volume of @xmath81 . the term @xmath82 " thus indicates that the lower boundary of the corresponding validity interval is smaller than @xmath81 . bottom : visualisation of the validity interval on logarithmic scale in the volume for the same ten parameter sets as used in the table . for a lower bound smaller than @xmath81 the lines have an arrow pointing to the left we find that the log - normal ma s range of validity is significantly smaller than that of the normal and poisson mas . ] [ table_bistable ] we next compare the prediction of the different ma schemes and of the rate equations for the mean copy numbers of species @xmath49 and species @xmath78 in steady state with results obtained from exact stochastic simulations using the ssa . the latter have been performed using the software package ina @xcite . figure [ fig_bistable_deviation_species_x ] shows the mean values of species @xmath49 as a function of the volume for the ten parameter sets used in figure 3 . the corresponding figures for species @xmath78 look very similar and are not shown here . the result is divided by the corresponding ssa result . the range of volumes shown corresponds roughly to the validity range of the normal and poisson ma . we observe here again that the mas become bistable for larger volumes and that the validity interval of the log - normal ma is significantly smaller than the one of the normal and poisson ma . we find that the mas overestimate the mean copy numbers and that the deviation from the ssa result increases for decreasing volumes . where two or all three mas are valid and thus comparable , the accuracy is similar with the log - normal ma being slightly more accurate than the other two and the normal ma being slightly more inaccurate than the poisson ma . note , however , that for most parameter sets the log - normal ma s range of validity is significantly smaller than that of the other mas . for large volumes , the mas have two positive stable fixed points converging to the two positive stable fixed points of the rate equations . the exact result obtained from ssa simulations agrees with the larger of these two fixed points . the third fixed point of the mas for large volumes seems to always lie between the two of the rate equations . while it lies exactly in the middle for the normal and poisson ma , it is very close to the lower one for the log - normal ma . we find the same behaviour for all parameter sets . note though that this can not be seen for all parameter sets in figure [ fig_bistable_deviation_species_x ] due to the small plot range . on logarithmic scale for the bistable reaction system in eqs . the parameter sets are the same as in the table in figure 3 . the values are divided by the corresponding result obtained from stochastic simulations using the ssa . the horizontal dashed line thus indicates the exact value . for the ssa result @xmath83 samples were simulated for each point . ] next , we study an enzyme catalysed protein - phosphorylation system with the reactions @xmath84{\quad a_1 \quad } e_1p \xrightarrow{\quad k_1 \quad } p^ * + e_1 , \\ \label{reactions_ultra2 } & p^ * + e_2 \xrightleftharpoons[\quad d_2 \quad]{\quad a_2 \quad } e_2p^ * \xrightarrow{\quad k_2 \quad } p + e_2.\end{aligned}\ ] ] this system shows ultrasensitivity for certain parameter values @xcite , namely when the enzymes are saturated , i.e. , most enzymes are on average in the complex state . here , @xmath85 and @xmath86 denote the non - phosphorylated and phosphorylated forms of the protein , respectively , @xmath87 and @xmath88 the phosphorylating and de - phosphorylating enzymes , respectively , and @xmath89 and @xmath90 the respective protein - enzyme - complexes . in @xcite the authors studied the dependence of the ratio of phosphorylated to non - phosphorylated proteins as a function of @xmath91 with @xmath92 and @xmath93 in a deterministic system , where @xmath94 and @xmath95 are the conserved total numbers of the respective enzymes in the system . assuming a hill - type response curve , the corresponding hill coefficient is often used to quantify the steepness of the response . the authors here speak of an ultrasensitive response " whenever the response is steeper than a michaelis - menten response , i.e. , has a hill coefficient of larger than unity . we study here the effect of noise on the ultrasensitive response and again compare moment - closure results with ssa simulations . the latter have been performed using the software package ina @xcite . first , however , we describe a surprising non - uniqueness of the poisson and log - normal ma and study the validity of the different ma schemes . as we have explained below eq . , the second - order normal and second - order cmm - ma are identical for elementary reaction systems with mass - action kinetics . since this is the case here , we study the normal , poisson and log - normal mas in the following . the studied reaction system in eqs . and has six species : @xmath96 , and three conservation laws : the total number of proteins and the total numbers of the respective enzymes , i.e. , @xmath97 , @xmath98 and @xmath99 , are conserved , where we use the symbol for the species also as the corresponding molecule number variable in a slight abuse of notation . the conservation laws allow one to reduce the system to three variables , which is obviously of computational advantage . there are two ways of obtaining the reduced moment - closure equations : arguably , the standard approach would be to start from the reduced cme , compute the reduced moment equations and subsequently apply the moment closure . alternatively , one may start from the full cme , compute the moment - closure equations and afterwards reduce the equations by taking the conservation laws into account . one may expect , or require , the two approaches for a sensible moment - closure scheme to be equivalent . it is easy to show that this is indeed the case for the normal and cmn moment - closures . however , we find here that this is not the case for the poisson and log - normal ma . we thus conclude that _ the poisson and log - normal mas are generally not uniquely defined if one reduces a system according to conservation laws in molecule numbers _ , a clear flaw of these methods . the reason for the non - uniqueness of the ma equations is that while the moment - equations depend on _ diagonal _ higher - order moments if one starts from a reduced cme , no such dependence is found if the ma equations are derived from the full cme . while the normal and cmn - mas treat diagonal and non - diagonal moments equivalently , the poisson and log - normal mas do not do so , thus leading to the issue of non - uniqueness . we explain this in more detail in appendix a. one consequence of this non - uniqueness is that certain symmetries of the system are broken . looking at the reaction system in eqs . and one sees that the system is symmetric under exchanging species labels and reaction constants , @xmath100 and @xmath101 and @xmath102 and @xmath103 . this means that for @xmath104 and @xmath105 the mean values of @xmath85 and @xmath86 , @xmath87 and @xmath88 , as well as @xmath89 and @xmath106 should be respectively equal . we find that this is indeed the case for the normal and cmn moment - closure , and also for the poisson and log - normal mas if one derives the equations starting from the full cme . if one applies the poisson and log - normal mas to the reduced cme , however , _ they do break the symmetry_. we conclude that one should be careful when using the poisson or log - normal ma for systems with conservation laws . in case the mas are non - unique it is favourable to first derive the mas before applying the conservation laws . in the following we will study the opposite cases , i.e. , if the poisson and log - normal ma are applied to the reduced cme , which would be normally the standard approach . as in @xcite we define @xmath92 and @xmath93 . the authors in @xcite studied the dependence of the fraction of the protein number in the phosphorylated state as a function of @xmath91 using deterministic rate equations . the authors call this response ultrasensitive " whenever it is steeper than michaelis - menten response , meaning a hill - coefficient larger than one . here , we would like to study the effect of noise on the response and investigate how different moment - closures perform for this system . to this end , we compute the mean value of the phosphorylated protein @xmath86 in steady state using the different methods of the protein on a grid in @xmath91 with all the other parameters fixed and fit a hill function @xmath107 to the result , where @xmath108 and @xmath109 are the dissociation constant and the hill - coefficient , respectively . for the protein phosphorylation system in eqs . and the blue and orange curve are hill - functions fitted to the points of the re and normal ma , respectively . the poisson and log - normal mas have only few positive stable fixed points in the response region making a sensible fit impossible . the used parameters are @xmath110 and @xmath111 , where @xmath112 and @xmath113 are the total number of enzyme @xmath87 , the total number of enzyme @xmath88 and the total number of proteins in the system , respectively . for the ssa result @xmath83 samples were simulated for each point . ] we find that the normal ma and rate equations are valid for all @xmath91 for all chosen parameter sets , whereas the poisson and log - normal ma are not for certain parameter regimes , i.e. , they do not always have a positive stable fixed point . figure [ fig_ultra_example ] visualises the fitting procedure for one parameter set . while the rate equations and normal ma are stable on the whole considered response region in @xmath91 , the poisson and log - normal mas are unstable for the major part of the region . we obtain only one and two values in the response region , respectively . the poisson and log - normal mas thus do not allow a sensible estimate of the response - steepness via a fit of a hill - function . and of @xmath91 for the protein phosphorylation system in eqs . and for five different parameter sets . if we write @xmath114 with @xmath115 , @xmath116 and @xmath117 , where @xmath113 is the total protein number and @xmath18 is the volume , the parameter sets are given by set 1 = @xmath118 , set 2 = @xmath119 , set 3 = @xmath120 , set 4 = @xmath121 and set 5 = @xmath122 . the blue regions indicate that the methods have no positive stable fixed point . the yellow regions indicate where a positive stable fixed points exists _ and _ the time trajectories converge with initial condition being the fixed point of the rate equations . the green regions show where the time trajectories diverge despite the existence of a positive stable fixed point , which means that the fixed point is only locally attractive . ] figure [ fig_ultra_validity ] visualises the validity of the rate equations , normal , poisson and log - normal mas as a function of the total enzyme number and @xmath91 for five different parameter sets . the figure indicates where the methods have a positive stable fixed point and where not . in addition , when a positive stable fixed points exists , we solve the time - dependent mas with the initial condition being the fixed point of the rate equations for the corresponding parameters , and the figure indicates the regions where these diverge despite the existence of a positive stable fixed point . this thus indicates the sensitivity of the different methods to initial conditions . while the rate equations and normal ma are stable and the time trajectories converge everywhere , the poisson and log - normal ma do so only in subregions of the parameter space . note that we do not make any statements about _ unstable _ fixed points here since we investigated the convergence of time - trajectories only for one fixed initial condition . the divergence of the time - trajectories in the green region suggest that there exists an unstable positive fixed point , but the same might be true in some parts of the yellow region despite the convergence of time - trajectories . in conclusion , we find that the normal ma performs significantly better than the poisson and log - normal ma for the studied system in terms of validity . for the protein phosphorylation system in eqs . and . the ssa result is shown as a solid black line . as explained in the main text , for some parameter values the poisson and log - normal ma do not allow to estimate a hill function due to instability . in such cases we set the hill coefficient to zero . for the ssa result @xmath83 samples were simulated for each point . ] next , we compare the hill coefficient obtained from the different methods with the results obtained from ssa simulations as a function of the total enzyme number @xmath123 for the five parameter sets defined in the caption of figure [ fig_ultra_validity ] . the ssa simulations were performed using the software package ina @xcite . if a method did not allow to estimate a hill coefficient for some @xmath123 we set the corresponding value to zero . figure [ fig_ultra_hill ] illustrates the results . first of all , we find that the rate equations overestimate the hill coefficient for all @xmath123 , with a larger deviation for small @xmath123 , which means that the noise in the system significantly reduces the steepness of the response . for small @xmath123 the hill coefficient estimated from the rate equations becomes up to four times larger then the ssa result ( set 4 in figure [ fig_ultra_hill ] ) . whenever they allow to estimate a hill coefficient , the moment - closure approximations are more accurate than the rate equations . while the normal and poisson mas underestimate the response , i.e. , overestimate the influence of noise , the log - normal overestimates the response . accuracy wise the three methods perform very similar , the poisson ma perhaps being slightly more accurate than the other two . however , this slightly higher accuracy of the poisson ma does not overcome its disadvantage of instability described in section [ sec_ultra_validity ] . next , we study the brusselator , a well known deterministic oscillating chemical system given by @xcite @xmath124{\quad c_3 \quad } x.\end{aligned}\ ] ] depending on the parameter values , the deterministic rate equations show sustained oscillations , damped oscillations or overdamped oscillations . single ssa trajectories may show sustained oscillations , while ensemble averages of the ssa always show damped or overdamped oscillations due to the dephasing of independent trajectories . therefore , a ma can only be interpreted as a valid moment approximation if its trajectories show damped or no oscillations . in @xcite it has been shown that for a parameter set for which the system in eq . is a deterministic oscillator , the normal ma is valid only for an intermediate range of volumes , with unphysical sustained oscillatory trajectories for larger volumes and either oscillatory or otherwise unphysical trajectories ( i.e. , divergent or negative trajectories ) for smaller volumes . here , we want to first study the validity of the different ma methods for different parameter sets , and then analyse their behaviour if the system becomes entrained by an external force . note that the first reaction in is trimolecular , which means that the corresponding propensity function is of third order in the molecule numbers ( c.f . the time - evolution equation of the second - order moments thus depend on the third and fourth - order moments ( c.f . therefore , since the fourth - order central moments and fourth - order cumulants are not identical ( in contrast to the third - order ones ) , the normal and cmn - mas are not equivalent for the reaction system in and we thus analyse all four mas separately in the following . ( blue line ) and @xmath78 ( orange line ) for several volumes for the brusselator system in eq . for the parameters @xmath125 . the blue and red curve denote the mean of species @xmath49 and species @xmath78 , respectively . while the normal , poisson and cmn - mas give physically meaningful results , i.e. , damped oscillations , for an intermediate range of volumes , the log - normal ma fails to do so for all volumes . to minimise the possibility of numerical effects we computed the shown results using the ode integration methods adams " , backward differentiation formula " , explicit runge kutta " , implicit runge kutta " , explicit midpoint " and stiffness switching " and varied the step sizes over several orders or magnitude , all giving the same results . ] we study here the validity of the mas for three different parameter sets defined in the caption of figure [ brusselator_validity_line_plot ] . similar to the findings in @xcite , we find that all four mas are only valid on an intermediate regime of volumes . however , unexpectedly , for the log - normal ma we can not find such a regime . figure [ brusselator_time_validity ] shows the time trajectories of the moments for the different mas for four different volumes for one fixed parameter set . while the normal , poisson and cmn - mas diverge for small volumes , are monostable for intermediate volumes and show sustained oscillations for large volumes , the log - normal switches directly from divergent to oscillatory behaviour . we estimated the range of validity for the three different parameter sets for fixed initial conditions of unity for the mean values of both species and zero variance . figure [ brusselator_validity_line_plot ] shows the ranges of validity on logarithmic scale in the volume . while the poisson and normal ma have a finite range of volumes where they lead to physically meaningful results for all parameter sets , the cmn - ma has a vanishing one for one parameter set and the log - normal for all parameter sets . in logarithmic scale . the used parameters for @xmath126 are set 1 @xmath127 , set 2 @xmath128 and set 3 @xmath129 . if the range of validity has length zero we plot a single point at zero . by range of validity " we mean the range of volumes for which the mas give physically meaningful ( i.e. , non - negative and converging ) time - trajectories . ] in systems biology it is frequently of interest to study systems where one or several propensity functions are time - dependent . for example , circadian oscillators are often modelled by a deterministic oscillatory system with an imposed periodic propensity function modelling the influence of an external light input @xcite . here , we want to study the performance of the different ma schemes for such a system in the stochastic setting . to this end , we modify the rate constant @xmath130 of the second reaction in eq . such that it varies over time from @xmath131 to @xmath132 times the chosen mean value in a sinusoidal way , i.e. , @xmath133 where @xmath134 is the fixed mean value of @xmath130 and the frequency @xmath135 of the sine curve is chosen to be the oscillation frequency of the deterministic system . after ten periods , we switch off the time dependence and fix @xmath130 to its mean value . since we have a time - dependent propensity function here , we can not use the ssa to simulate the system . we therefore use extrande , a recently developed exact mc method to sample from the solution of cmes with time - dependent rate functions @xcite . figure [ brusselator_entrainment ] shows the time trajectories for the rate equations , extrande simulations and the different ma methods . we find that the rate equations show sustained oscillations after entrainment , while the extrande results show damped or overdamped oscillations . the normal and poisson ma behave qualitatively the same way as the extrande and are thus valid moment approximations for the chosen parameter values . quantitatively they differ quite significantly from the extrande result , however . they underestimate the mean values , show oscillations with larger amplitudes during entrainment and a weaker damping after entrainment . looking at figure [ brusselator_entrainment ] one finds that these effects are stronger for the respective smaller volume for each parameter set . the normal and poisson ma thus _ underestimate _ the influence of noise here . the log - normal and cmn - mas fail everywhere to provide a physical result . for the the former this may to be expected , since it also failed to do so in the case without entrainment . interestingly , however , the cmn - ma is invalid even for parameters for which it is valid in the case without external input . overall , the normal and poisson ma seem to perform significantly better for this system than the log - normal and cmn - ma . with entrainment for two different volumes for each parameter set . the blue and orange lines denote the mean values of species @xmath49 and @xmath78 , respectively . the external input gets switched on at time @xmath136 and switched off after ten oscillation periods of the deterministic system ( which depends on the given parameter set ) . while the normal and poisson mas give physically meaningful results ( i.e. , non - negative and converging time - trajectories ) for an intermediate range of volumes , the log - normal and cmn - mas fails to do so for all volumes . for the extrande result we simulated @xmath137 samples for set 1 and @xmath83 samples for set 2 and set 3 , respectively . ] the mathematica package moca implements the investigated four moment - closure approximations , as well as deterministic rate equations , in a graphical user interface and is freely available in the supplemental material @xcite . in contrast to other available moment - closure software packages @xcite , moca does not only derive the closure equations but also automatically performs numerical analysis of the derived equations , making the methods available to non - expert users . the results are automatically visualised and can be exported to various formats . moca extends the applicability over existing moment - closure packages to * non - polynomial propensity functions * time - dependent propensities functions * propensities defined on discrete time points ( e.g. measured fluctuating external parameter ) note that while non - polynomial propensities can often give a useful description of a system , they should really be interpreted as an effective approximate description of a set of elementary reactions , valid only under certain conditions @xcite . for these type of propensities the software applies a taylor expansion of the propensity around the mean value to a specified order as proposed in @xcite . these different features make moca applicable to virtually any reaction system with arbitrary propensity functions . in addition to the different moment - closure methods described above , moca allows the user to define his own moment - closure method , providing an easy way to develop novel moment - closure schemes . and the rate constants @xmath138 into the parameters @xmath139 as defined below eq . . ] to use the package , the file moca.m needs to be placed in the same folder as the mathematica notebook that will be used for the analysis figure [ input1 ] shows an example input for the corresponding notebook for the reaction system defined in . the first two lines , which set the path and load the package , respectively , have to be executed without any modification . next , the number of species and the stoichiometric matrix have to be specified and assigned to the variables * ns * and * stochmatrix * , respectively , as depicted in the third and fourth line in figure [ input1 ] . the propensity vector and stoichiometric matrix are given in eqs . and , respectively . the number of species * ns * has to be equal to the number of rows of * stochmatrix*. next , the parameter vector * parameters * and the propensity vector called * propensity * need to be specified , as done in the fifth and sixth input lines in figure [ input1 ] . the species variables have to be denoted by an @xmath140 " with the species index as a subscript . all terms in the propensity function that are not species variables or numerical values have to be listed as parameter in * parameters*. this is all the input needed if dealing with time - independent propensity functions and when using the gui . note that the propensities do not need to be of mass - action , i.e. , polynomial type , they can have any analytical form . for using the coding version of moca , deterministic rate equations and time - dependent propensity functions , as well as for the definition of moment - closures , see the corresponding tutorial files in the supplemental material @xcite . the command * steadystate * has to be evaluated in the notebook for the gui to appear . the user can choose the closure method , closure order , expansion order and specify parameter values . for changes to apply the user needs to press the little update button " in the top right corner . ] there are four functions available to be used within a gui . they simply need to be typed into the notebook and evaluated to open the corresponding gui : * * deriveequations * : derives the ma equations for central moments for general parameters and allows to assign numerical values to the parameters . * * steadystate * : numerically searches for positive and stable fixed points of the ma equations . * * steadystatevaryparameter * : same as * steadystate * but with one parameter varied over a grid specified by the user . the resulting table can be exported into a csv " ( `` comma - separated values '' ) file . * * timetrajectory * : solves ma equations numerically in time for numerical parameter values and plots the result . the result can be exported as a figure to various formats or evaluated on a grid in time and stored in a csv " file . figure [ analysis_mm_equations ] shows the gui that appears after typing and evaluating * deriveequations*. the user can interactively choose a moment - closure method , the closure order as well as the expansion order . by expansion order " we mean the expansion of the propensity functions around the mean value as proposed in @xcite . this is only necessary for non - polynomial rate functions . for exclusively polynomial rate functions , the expansion does not make a difference as long as its order is equal to or higher than the maximum order of the propensity polynomials . finally , it is possible to assign numerical values to the parameters . the equations only become updated when the small update bottom " in the top right corner is clicked . this is also true for the functions described in the following , i.e. , changes in the input are only applied after clicking the update bottom " . the function * steadystate * allows to numerically compute positive stable fixed points of the ma equations . it has the same input parameters as the function * deriveequations * described before , with the difference that the parameters have an initial numerical value . for some parameter values , the method can not find a positive and stable fixed point . however , this does not necessarily mean that the numerical algorithm fails . in @xcite it has recently been shown that ma equations can indeed have no positive and stable fixed point for certain bimolecular reaction system ( even though the ssa and rate equations do have positive stable fixed points ) . the authors also showed that mas can have more than one positive stable fixed point , in which case * steadystate * function may give more than one result . the function * steadystatevaryparameter * also searches for positive stable fixed points but varies a user specified parameter over a user specified grid . the corresponding gui is shown in figure [ analysis_mm_ss_grid ] . the resulting table can be exported to a text file in csv " format to the same folder where the notebook is located . the final function * timetrajectory * solves the ma equations numerically in time and plots the result . figure [ analysis_mm_time ] shows the corresponding gui . in addition to method specifications and values for parameters , the user can specify initial conditions for the mean values of the species ( higher order central moments are set to zero initially , i.e. , deterministic initial conditions ) , the final time point and the plot order specifying up to which order moments should be plotted . the result can either be exported as a figure to various formats or into a csv " text file where the solution is evaluated on a time grid with user specified time spacing @xmath141 . the gui commands described above are also available as mathematica functions allowing more experienced mathematica users a more flexible application of the methods . see the example files in the supplemental material @xcite for details on how to use these functions . in this paper , we compared the second - order normal , poisson , log - normal and cmn - mas for several reaction systems with respect to their qualitative behaviour ( if they give physically meaningful results ) and their quantitative accuracy ( how well they approximate results obtained from exact stochastic simulations ) whenever they give physically meaningful results . while we found no significant difference in quantitative accuracy between the four mas , the ranges in parameter space for which the mas gave physically meaningful results were significantly larger for the normal ma suggesting that the normal ma is favourable over the other studied mas . we emphasise that the presented results are exclusively based on numerical analysis and although we confirmed the results for a wide range of parameter sets and several example systems , we can not expect them to hold in general for all parameter sets or chemical reaction systems . in @xcite for example , it has been found for a single parameter set for one chemical reaction system that the log - normal ma is significantly more accurate than the normal ma . however , for non - linear systems , our results suggest that the mas give physically meaningful results only above a certain critical volume if the system is deterministically monostable , and only for intermediate volumes if the system is not deterministically monostable . by physically meaningful " we mean the validity conditions proposed in @xcite which are : ( i ) the mean values and even central moments of a system should stay non - negative and finite for all times and they should converge to a steady state whenever the cme has a steady state solution , ( ii ) the moments are unique in the sense that the same steady - state moments can be reached from all initial conditions , and ( iii ) the moments do not exhibit sustained oscillations in the limit of long times ( unless there is an external time - dependent input ) . in @xcite it has been found that the normal ma does not satisfy ( i ) for small volumes for several non - linear reaction systems , and that it does not satisfy conditions ( ii ) and ( iii ) for large volumes for deterministic bistable and oscillatory chemical systems , respectively . here , we performed a similar analysis for four different ma methods . we first studied a deterministically bistable system , i.e. , a system whose rate equations have two positive stable fixed points . interestingly , we find that the mas have _ three positive stable fixed points _ for large volumes , thus allowing no physical interpretation . surprisingly , we found that , for an enzyme - catalised reaction , _ the poisson and log - normal mas were not uniquely defined_. our analysis suggests that this may indeed be generally the case for systems with conservation laws , a flaw not shared by the other two mas . finally , we studied a deterministically oscillatory system with and without an external periodic input . in both cases we found that the normal and poisson mas are valid only for an intermediate range of volumes , becoming unstable for smaller volumes and undergoing unphysical sustained oscillations for larger volumes . curiously , the cmn - ma behaves like this only for some of the studied parameter sets , and the log - normal for none of these , i.e. , _ there is no range of volumes where the latter two mas give physically meaningful results_. in conclusion , our results taken together do not favour one ma over the others in terms of accuracy , but suggest that the normal ma is favourable over the other mas , in the sense that the range of parameter space where it gives physically meaningful results is considerably larger than that of the other mas . finally , we presented the software package moca which was used for the numerical analysis of the various mas . moca allows one to derive and analyse moment - closure approximations for systems with polynomial , non - polynomial as well as time - dependent propensities . moca implements the normal " or cumulant - neglect " , the poisson " , the log - normal " and the cmn " closures as well as user - defined moment - closure schemes and automatises the numerical analysis . it allows non - expert users to apply moment - closure methods in a user - friendly graphical user interface . g.s . acknowledges support from the european research council under grant mlcs 306999 . 10 r. grima and s. schnell , essays biochem . * 45 * , 41 ( 2008 ) . d. t. gillespie , annu . phys . chem . * 58 * , 35 ( 2007 ) . d. t. gillespie , j. phys . chem . * 81 * , 2340 ( 1977 ) . c. a. gmez - uribe , g. c. verghese , and a. r. tzafriri , j. chem . phys . * 129 * , 244112 ( 2008 ) . d. t. gillespie , y. cao , k. r. sanft , l. r. petzold , j. chem . phys . * 130 * , 064103 ( 2009 ) . d. a. mcquarrie , j. app . prob . * 4 * , 413 ( 1967 ) . r. grima , j. chem . phys . * 136 * , 154105 ( 2012 ) . l. ferm , p. ltstedt and a. hellander , j. sci . comput . * 34 * , 127 ( 2008 ) . m. ullah and o. wolkenhauer , j. theor . biol . * 260 * , 340 ( 2009 ) . c. a. gomez - uribe and g. c. verghese , j. chem . phys . * 126 * , 024109 ( 2007 ) . a. ale , p. kirk and m. p. h. stumpf , j. chem . * 138 * , 174101 ( 2013 ) . i. nasell , theor . . biol . * 64 * , 233 ( 2003 ) . m. j. keeling , j. theor . biol . * 205 * , 269 ( 2000 ) . j. hespanha , proceedings of the 3rd international symposium on communications , control and signal processing ( ieee ) * 142 * , 2063 ( 2008 ) . a. singh , and j. p. hespanha , decision and control , 2006 45th ieee conference on . ieee , 2006 . e. lakatos , a. ale , p. kirk , and m. p. h. stumpf , j. chem . phys . * 143 * , 094107 ( 2015 ) . p. milner , c. s. gillespie and d. j. wilkinson , stat . * 23 * , 287 ( 2013 ) . c. zechner et al . usa * 109 * , 8340 ( 2012 ) . a. singh , and j. p. hespanha , ieee trans autom control * 56 * , 414 ( 2011 ) . d. schnoerr , g. sanguinetti , r. grima , j. chem . phys . * 141 * , 084103 ( 2014 ) . j. hespanha . stochdyntools - a matlab toolbox to compute moment dynamics for stochastic networks of bio - chemical reactions . available at http://www.ece.ucsb.edu/ hespanha , 2007 . c. gillespie , iet systems biology * 3 * , 52 ( 2009 ) . p. azunre , c. gmez - uribe , and g. verghese , iet systems biology * 5 * , 325 - 335 ( 2011 ) . n. g. van kampen , _ stochastic processes in physics and chemistry _ ( elsevier , 2007 ) . e. l. crow and k. shimizu , eds . _ lognormal distributions : theory and applications . _ ( new york : dekker , 1988 ) . f. schlgl , z. phys . * 253 * , 147 ( 1972 ) . t. wilhelm , bmc systems biology * 3 * , 90 ( 2009 ) . p. thomas , h. matuschek and r. grima , plos one * 7 * , e38518 ( 2012 ) . a.goldbeter and d. e. koshland , proc . natl , * 78 * , 6840 ( 1981 ) . i. prigogine and r. lefever , j. chem . phys . * 48 * , 1695 ( 1968 ) . r. lefever , g. nicolis and p. borckmans , j. chem . , faraday trans . 1 * 84 * , 1013 ( 1988 ) . p. o. westermark et al . , plos comput biol * 5 * , 1000580 ( 2009 ) . c. troein et al . , the plant journal * 66 * , 375 ( 2011 ) . g. kurosawa , a. goldbeter , j. theor . biol . * 242 * , 478 ( 2006 ) . m. voliotis , p. thomas , r. grima and c. bowsher , _ stochastic simulation of biomolecular networks in dynamic environments _ ( 2015 ) ( submitted ) . see supplemental material at http://grimagroup.bio.ed.ac.uk/publications.html for software package moca , as well as tutorial and example files . thomas , a. v. straube and r. grima , j. chem . phys . * 135 * , 181103 ( 2011 ) . here , we investigate in detail , the non - uniqueness of the poisson and log - normal mas for systems with conservation laws . to this end , we consider the simple reversible reaction system @xmath142{\quad k_1 \quad } c.\end{aligned}\ ] ] we now compute the ma equations by applying the conservation laws of the system once _ after _ , and once _ before _ closing the equations . this approach involves obtaining the moment equations from the cme and subsequently imposing the conservation laws on the resulting moment equations . the stoichiometric matrix @xmath7 and propensity functions @xmath143 and @xmath144 of the two elementary reactions for this system read ( c.f . @xmath145 where @xmath146 and @xmath147 denote the copy numbers of species @xmath148 and @xmath149 , respectively . the corresponding time - evolution equations for the first and second - order moments can be obtained from eqs . and . for @xmath150 and @xmath151 , for example , they read @xmath152 note that due to the term @xmath153 in @xmath143 , the equation for @xmath68 depends on the third - order moment @xmath66 , but not on any _ diagonal _ third - order moment , i.e. , not on @xmath154 or @xmath155 . the same is of course true for the equations of the other second - order moments : they do not depend on a diagonal third - order moment . this means that the second - order normal and poisson mas are equivalent , since they differ only in their expressions for diagonal moments ( c.f . eqs . - ) . the corresponding second - order normal and poisson mas for @xmath63 and @xmath68 are obtained by setting the corresponding third - order cumulant @xmath156 to zero which leads to @xmath157 and thus gives @xmath158 and similarly for the other first and second - order moments . note that the system has two conservation laws @xmath159 to simplify the following equations , let us assume @xmath160 , which implies @xmath161 . the system of moment equations of three variables can thus be reduced to a system with only one variable , since all moments of first and second order can be expressed in terms of @xmath162 and @xmath68 using eqs . ( [ app_cons_1 ] ) and ( [ app_cons_2 ] ) . for example , we have @xmath163 and @xmath164 and similarly for the other first and second - order moments . the resulting equations for @xmath63 and @xmath68 are thus closed and read @xmath165 note that these are the resulting second - order ma equations for both the normal _ and _ the poisson ma . alternatively , we can start from the reduced cme with species @xmath166 and @xmath149 eliminated , whose stoichiometric matrix and propensity functions are given by @xmath167 note that due to the term @xmath168 , the time - evolution equation for the second - order moment @xmath68 depends on the _ diagonal _ third - order moment @xmath169 ( all moments are diagonal here of course , since we deal with a system with only one variable ) . the corresponding equations for the first two moments can be obtained using eqs . and and read @xmath170 for closing these equations to second order , we need to express @xmath171 in terms of @xmath63 and @xmath68 . the corresponding expression is now not the same anymore for the normal and poisson mas . for the normal ma we have @xmath172 . inserting the latter into eq . one obtains the same result as in eqs . and which we obtained by applying the conservation laws _ after _ closing the equations . in contrast , if we apply the poisson ma , which sets @xmath173 , the resulting equation for @xmath68 is _ not _ equal to eq . . the reason for this is that the poisson ma does not treat diagonal and non - diagonal moments equivalently . here , this means that the replacements of @xmath169 and @xmath66 differ from each other if one sets the index @xmath174 to @xmath175 in the expression for @xmath66 . since the same is true for the log - normal ma , the latter also gives differing results depending if the equations are closed before or after the conservation laws are applied . since the normal and cmn - ma do treat diagonal and non - diagonal moments equivalently ( so the expressions for @xmath169 and @xmath66 are the same after setting @xmath174 to @xmath175 ) , these mas do not suffer from this flaw .

we present a general procedure for constructing lattices of qubits with a hamiltonian composed of nearest - neighbour two - body interactions such that the ground state encodes a cluster state . we give specific details for lattices in one- , two- , and three - dimensions , investigating both periodic and fixed boundary conditions , as well as present a proof for the applicability of this procedure to any graph . we determine the energy gap of these systems , which is shown to be independent of the size of the lattice but dependent on the type of lattice ( in particular , the coordination number ) , and investigate the scaling of this gap in terms of the coupling constants of the hamiltonian . we provide a comparative analysis of the different lattice types with respect to their usefulness for measurement - based quantum computation .

0805.2980

there is currently considerable interest in preparing exotic quantum states of many - body systems which can be used as resource states for measurement - based quantum computation ( mbqc ) that is , quantum computation that proceeds solely through local adaptive measurements on single quantum systems @xcite . the canonical example of such a resource state is the cluster state @xcite , which is a universal resource for mbqc on suitable lattices or graphs @xcite . it may be possible to prepare such a cluster state dynamically in atomic systems such as an optical lattice @xcite or using single photons @xcite however , one exciting possibility is that such resource states might be the non - degenerate ground state of a `` natural '' hamiltonian lattice system . if the system is gapped , then one simply needs to cool it sufficiently in order to obtain the desired state ( although , even for gapped systems , this cooling process may be difficult @xcite ) . consider the cluster state on a lattice @xmath0 , defined as the unique @xmath1 eigenstate of a set of stabilizer operators @xmath2 , where @xmath3 ( @xmath4 ) is the pauli @xmath5 ( @xmath6 ) operator at site @xmath7 and where @xmath8 denotes that @xmath9 is connected to @xmath7 by a bond in the lattice @xmath0 . the hamiltonian @xmath10 with @xmath11 an energy constant , has the cluster state as its unique ground state @xcite . in addition , this system is gapped ( the gap is @xmath12 ) , and such a system can be cooled efficiently @xcite . however , for any non - trivial lattice or graph , this hamiltonian involves many - body interactions , as opposed to the two - body interactions that occur frequently in nature . an obvious question , then , is whether it is possible to realize any given highly - entangled quantum state as the ground state of a hamiltonian with only two - body interactions . et al . _ @xcite proved that this is not possible in general , and nielsen @xcite used this result to prove that a cluster state on a computationally universal ( i.e. , two - dimensional or higher ) lattice can not arise as the ground state of a hamiltonian with only two - body interactions . however , investigations into quantum complexity theory @xcite have demonstrated that cluster states ( and other such states that are universal ) can be _ approximated _ by the ground state of a local two - body hamiltonian . the key idea is to make use of `` mediating '' ancilla qubits to create an effective many - body coupling out of two - body interactions . the problem with such methods is that the detailed parameters in the perturbing hamiltonian must be controlled with a precision that increases with the size of the system @xcite , making such approaches impractical for the task of creating cluster states on large lattices . using an alternate method based on the idea behind projected entangled pair states ( peps ) @xcite , bartlett and rudolph @xcite proved that it was possible to obtain a state that closely approximates an _ encoded _ cluster state on a square lattice using a hamiltonian with only two - body nearest - neighbour interactions . in addition , they proved that mbqc can proceed using such an encoded resource state , still requiring only adaptive single - qubit measurements . in this paper , we present a general method for constructing two - body nearest - neighbour hamiltonian systems for which the ground state encodes a cluster state , based on the techniques of @xcite . our rigorous application of perturbation theory reveals errors in the calculation of the energy gap for the square lattice investigated in @xcite ( although these errors do not affect their key result ) and we provide a correct treatment of this case . we also investigate the cluster state on a one - dimensional line ( useful for illustration , as well as for its application as a quantum wire @xcite ) , a hexagonal lattice in two - dimensions a universal mbqc resource with the best scaling of the energy gap in perturbation , and the cubic lattice in three - dimensions a resource state for which fault - tolerance thresholds have been found @xcite . we explicitly characterise the effects of fixed boundary conditions on the lattice , proving that such boundary conditions do not affect the main result . finally , we provide an outline of a proof that this method yields an encoded cluster state as the ground state on any graph . our general method relies on the fact that the cluster state is simply represented as a projected entangled pair state ( peps ) , also known as a valence - bond solid state . the peps representation @xcite is a powerful and often compact method of describing the state of a many - body system . consider a regular lattice @xmath0 of qubits , with coordination number @xmath13 ( i.e. , @xmath13 bonds connect each qubit to other sites on the lattice ) . a pep state on @xmath0 can be constructed by assigning a pair of _ virtual _ quantum systems of dimension @xmath14 to each bond on the lattice , each pair prepared in a maximally - entangled state , and then applying a projection @xmath15 to the @xmath13 virtual systems associated with each site . the cluster state ( and a wide variety of other states of interest ) require only @xmath16 for their representation ; in what follows , we restrict our attention to this case where the virtual systems are qubits . in addition , we choose the maximally - entangled state of these virtual qubits to be the two - qubit cluster state @xmath17 where @xmath18 . with this convention , the cluster state has a simple peps representation @xcite corresponding to the projection operator @xmath19 at each site , with @xmath13 zeros ( ones ) in @xmath20 ( @xmath21 ) , and the states @xmath22 and @xmath23 forming a basis for the resulting qubit at each site . as an example , consider the peps representation of the cluster state on a two - dimensional square lattice . there are four bonds emanating from every site in a square lattice , and so each site possesses four virtual qubits . virtual qubits connected by a bond are placed in the state @xmath24 , and then a projection @xmath25 is applied . the resultant state , @xmath26 is a cluster state on the square lattice . the essential idea of the method presented in this paper is to mimic the peps construction procedure using a physical two - body hamiltonian , wherein the `` virtual '' qubits are real physical systems and the resulting peps state is _ encoded _ into logical qubits . consider a regular lattice . let @xmath0 denote the set of sites , each with coordination number @xmath13 . we assign @xmath13 qubits to each site , and label with a double index @xmath27 , with @xmath28 and @xmath29 . ( the choice of the second label @xmath30 is completely arbitrary . ) let @xmath31 and @xmath32 denote the pauli @xmath5 and @xmath6 operators for the @xmath30th qubit at site @xmath28 . following the peps construction , if a site @xmath7 is connected to a site @xmath9 by a bond ( denoted @xmath33 ) , then we associate qubit @xmath27 and @xmath34 for some @xmath30 to this bond . ( we note that this notation can become problematic with certain periodic conditions , but it should be clear from the context how to adjust it appropriately . ) our _ peps hamiltonian _ is defined as follows . at each site , we require a two - body hamiltonian with a two - dimensional ground state space spanned by @xmath35 and @xmath36 . for this , we choose a site hamiltonian @xmath37 which is of ising form @xmath38 where @xmath39 denotes that qubits @xmath30 and @xmath40 are connected according to some graph structure . aside from being connected , the specific form of this graph is relatively unimportant ; however its structure will affect the energy levels of the hamiltonian . for example , for two - dimensional lattices , it is natural to choose a ring structure . between sites , we define a different two - body interaction of the form @xmath41 where @xmath42 is the site connected to @xmath7 via bond @xmath30 . with this hamiltonian , on every bond in the lattice @xmath33 there are two terms : @xmath43 and @xmath44 . note that the terms in @xmath45 stabilise @xmath24 , and therefore @xmath46 is the ground state of @xmath45 . this product of maximally - entangled states is the starting point of the peps construction . the site hamiltonian @xmath37 is meant to `` implement '' the peps projection by ensuring that the qubits at a site act as a single logical qubit ; to do so , the site hamiltonian @xmath37 must be much stronger than the bond hamiltonian @xmath45 . one is then lead to consider the ground state of the hamiltonian @xmath47 where @xmath48 , which is suitable for perturbative analysis in @xmath49 . in @xcite , this procedure was applied to a square lattice . the terms in the perturbation combine at higher orders to yield the stabilisers of the logical cluster state , and the resulting low - energy theory of the lattice is governed by an effective hamiltonian of the form of eq . ( [ eq : clusterhamiltonian ] ) . furthermore , the gap to the next excited state is finite and independent of the size of the lattice . of course , because this is a perturbative approach there will now be corrections to the unperturbed logical eigenstates which will not be in the logical space . so the _ exact _ cluster state will not be obtained . however , these errors will be small ( occurring with probability @xmath50 , as we will show ) and so a state arbitrarily close to the cluster state can be obtained . in what follows , this procedure is generalised to other lattice structures important in quantum computation . we now present a simple duality transformation that maps the hamiltonian to one describing uncoupled sites . consider decomposing as a sum of commuting terms @xmath51 , as @xmath52 where @xmath53 note that @xmath54=0 $ ] for @xmath55 . define the unitary transformation @xmath56 to be the application of a @xmath57 gate @xmath58 to every bond on the lattice . the action of this transformation on the bond terms in the above hamiltonian is @xmath59 . transforming each term @xmath51 thus yields @xmath60 which is localized entirely to the site @xmath7 . thus , the duality transformation @xmath56 yields a hamiltonian of uncoupled sites , where each site hamiltonian takes the form of a transverse - field ising model on some connected graph . with this mapping , the spectrum of the hamiltonian can be calculated explicitly , with the hamiltonian term @xmath61 at each site , for example by using a jordan - wigner transformation . we note at this point that each hamiltonian @xmath61 has a nondegenerate ground state for all @xmath62 ; thus , our peps hamiltonian on the full lattice will also possess a nondegenerate ground state for @xmath62 . for each site @xmath7 , define the operator @xmath63 that is , @xmath64 is the tensor product of @xmath65 for every qubit at site @xmath7 together with @xmath66 on every neighbouring site . it is straightforward to show that all such operators commute with the hamiltonian of eq . , @xmath67=0\ , , \quad \forall\ \mu \in \mathcal{l}\,,\ ] ] and with each other , @xmath68=0\ , , \quad \forall\ \mu,\nu\in\mathcal{l}\,.\ ] ] thus , if @xmath69 has a nondegenerate ground state ( as is the case for the peps hamiltonian with @xmath70 ) , it must also be a simultaneous eigenstate of all operators @xmath64 . using the duality transformation @xmath56 defined above , we find that @xmath71 using the well - known solution to the transverse - field ising model with hamiltonian ( [ eq : tfim ] ) , we find that the ground state for @xmath62 is the @xmath1 eigenstate of this operator . ( this can be inferred by the fact that the ground state is clearly the @xmath1 eigenstate of @xmath72 in the limit @xmath73 . ) thus , we have that the ground state of the peps hamiltonian for @xmath62 is the simultaneous @xmath1 eigenstate of all operators @xmath64 , @xmath28 . the operators @xmath64 , then , can be viewed as _ encoded cluster stabilizers _ , and the ground state for @xmath62 as an encoded cluster state . unfortunately , for @xmath62 , this encoding is no longer in the ground state space of @xmath37 spanned locally at sites by the states @xmath35 and @xmath36 . our perspective is to consider the encoding to be fixed in this space and view the ground state of the peps hamiltonian as the desired locally - encoded cluster state plus perturbative corrections in @xmath74 . these concepts are best illustrated through a simple example . first we illustrate this approach on the simplest lattice : a one - dimensional line of qubits with periodic boundary conditions . we demonstrate explicitly that the perturbative procedure yields a low - energy effective hamiltonian on the logical qubits of the form of eq . ( [ eq : clusterhamiltonian ] ) , and that an approximate cluster state is obtained as the non - degenerate ground state . the basic steps outlined in this example for the one - dimensional line , suitably generalized , will be applicable to more complex lattice structures . consider a one - dimensional line consisting of @xmath75 qubit sites with periodic boundary conditions ( i.e. , a ring ) . the coordination number of this lattice is @xmath76 , and thus our construction requires two qubits to be placed at each site . this lattice structure is illustrated in fig . [ fig:1dline ] . the hamiltonian is that of eq . ( [ hamilt ] ) . we first investigate the energy eigenvalue spectrum of the unperturbed hamiltonian @xmath77 . because @xmath77 is a sum of terms , each of the form @xmath78 acting on a single site , the energy spectrum can be determined by analysing each site individually . at a single site , there are two energy levels . the ground - state is degenerate , two - dimensional , and spanned by the states @xmath79 the ground state space of the unperturbed hamiltonian at each site , then , can be viewed as a _ logical qubit_. note also that this ground state space is , by construction , the logical subspace for the cluster state peps projection . the energy of this ground state space is @xmath80 . the first excited state at each site is also two - dimensional , has an energy of @xmath81 , and is spanned by the states @xmath82 and @xmath83 . with the spectrum of @xmath77 at each site , we now describe the spectrum of the unperturbed hamiltonian @xmath77 on the entire lattice . the lattice ground - state space is spanned by product states of all of the individual sites in the ground state ( i.e. in the logical space ) . this ground - state space has energy @xmath84 , is @xmath85-dimensional , and is spanned by all logical states of @xmath75 qubits . we denote this space @xmath86 . the first - excited space is @xmath87-dimensional , and has energy @xmath88 . thus , for the unperturbed hamiltonian @xmath77 , the gap from the ground to first - excited space is @xmath89 . the second - excited space has energy @xmath90 , and so on . these energies will serve as the zeroth - order energies in perturbation theory for the total hamiltonian . we now turn to perturbation theory and determine the effect of the term @xmath91 in the hamiltonian ( [ hamilt ] ) . we will show that this term lifts degeneracy of the ground state , and that the logical cluster state arises as the unique ground state ( although we also show that there are perturbative corrections to this state ) . for details of our use of perturbation theory and notation , see the appendix . let the @xmath92th - order energy correction to the @xmath40th state in @xmath86 be denoted by @xmath93 . let @xmath94 be the projection onto the degenerate ground state space of the unperturbed hamiltonian @xmath77 , i.e. , onto the `` logical '' space @xmath86 . define @xmath95 to be the projection onto the `` illogical space '' ( denoted @xmath96 ) and let the projection onto the @xmath40th unperturbed excited level be denoted @xmath97 . to obtain a conceptual view of the perturbation it is useful to see the effect of @xmath45 on a single site . the hamiltonian @xmath45 is a sum of terms of the form @xmath98 ; however , each of the pauli operators in such a term act on different sites , and so we must consider the action of @xmath65 and @xmath66 separately . because the logical space @xmath86 is spanned by @xmath99 and @xmath100 , the action of @xmath65 will move a site out of the logical space ; the action of @xmath66 will not , and simply induce a phase . the possible actions by the @xmath65 part of @xmath45 at a single site are shown in fig . [ fig:1denergy ] . terms in @xmath45 on a single site.,width=226 ] the first - order corrections to the energy are governed by the operator ( see eq . ( [ a11 ] ) in the appendix ) @xmath101 specifically , the first order energy corrections @xmath102 to the ground state are the eigenvalues of this operator . because all of the terms in @xmath45 contain a single @xmath65 , they all move a state in the logical space to the first excited state ( i.e. , @xmath103 ) . thus , @xmath104 , and there is no first - order correction to the energies . the second - order corrections are governed by the operator @xmath105 where the expression has been simplified using @xmath106 . the operator @xmath107 maps states from the ground state space to the first excited space and then back to ground state space . by investigating the different ways of returning to the logical space after just two @xmath65 operations , it is clear that there are two possible contributions to this term : 1 . if @xmath108 in @xmath45 is applied twice to the same bond , it yields the identity . the first @xmath65 can be applied to any of the qubits and then must be applied again to the same qubit , so there are @xmath109 of these terms . 2 . if @xmath65 is applied to each of the two qubits at a site ( and corresponding @xmath66 operations to qubits in the neighboring sites ) , then the lattice remains in the ground state . we can apply the first @xmath65 to either of the two qubits at the site and so there will be two of these terms that occur at each site . explicitly , this case will be a term applied to the logical space of the form @xmath110 where @xmath33 . + the operator @xmath111 , which acts only on the logical space , can be determined explicitly as follows . note that the product of two @xmath65 operators on a single site @xmath7 ( one on each physical qubit ) , restricted to the logical space , is equivalent to a _ logical _ @xmath5 operator @xmath112 also , a single @xmath66 operator acting on either of the two physical qubits at a site @xmath9 , restricted to the logical space , is equivalent to a _ logical _ @xmath6 operator @xmath113 thus , @xmath2 . this operator is a ( logical ) stabilizer of the cluster state on this lattice . therefore , we have @xmath114 substituting this result into eq . ( [ 2ndord ] ) and using @xmath115 gives @xmath116 the energies @xmath117 are the eigenvalues of @xmath118 and the corresponding eigenstates of @xmath118 will be the zeroth - order energy eigenstates after the degeneracy is lifted . next , we identify the basis which diagonalises @xmath118 ; this is straightforward given the expression ( [ 2ndord2 ] ) . as the cluster state is the simultaneous @xmath1 eigenstate of all stabilizer operators @xmath119 , the _ logical cluster state _ on this lattice , denoted by @xmath120 , is an eigenstate of @xmath118 . similarly , the other eigenstates of @xmath118 are also just the simultaneous eigenstates of the stabilizers @xmath119 ( as all such stabilizers commute ) . explicitly , let @xmath121 denote the logical cluster state with a logical @xmath6-operator ( called a _ @xmath6-error _ ) applied to the sites @xmath122 . using the anti - commutation relations of the pauli matrices , @xmath121 is the @xmath123 eigenstate of @xmath124 and the @xmath1 eigenstate of @xmath119 for @xmath125 . therefore , the @xmath85 states of the form @xmath121 will be eigenstates of @xmath118 . furthermore , these states are orthogonal , as each pair of states will have a differing eigenvalue for least one of the @xmath119 operators . in summary , the set of states @xmath126 , running over logical @xmath6-errors at all possible sites , forms an orthogonal basis of @xmath86 and diagonalises @xmath118 . the eigenvalue spectrum of @xmath118 is then straightforward to calculate using the properties of stabilisers . from the form of @xmath118 in eq . ( [ 2ndord2 ] ) , the lowest energy eigenstate will be the cluster state @xmath120 , because it is an eigenstate of all stabilisers in the sum @xmath127 with eigenvalue @xmath1 . thus the second - order correction for the energy associated with the cluster state is @xmath128 next , consider a state @xmath129 , a cluster state with a single @xmath6-error at the site @xmath130 . this state is also an eigenstate of all stabilizers in the sum @xmath131 with eigenvalue @xmath1 _ except _ the stabilizer @xmath132 for which it has eigenvalue @xmath123 . therefore , @xmath133 because there are @xmath75 states of the form @xmath134 , this @xmath135 excited space is @xmath75-dimensional . similarly , the @xmath92th excited space up to @xmath136 is @xmath137-dimensional and ( to zeroth order ) is spanned by states obtained from @xmath120 by @xmath92 logical @xmath6-errors . higher order corrections can be calculated by following a similar procedure . as noted in sec . [ subsec : exactsol ] , this hamiltonian can be easily solved exactly , with a ground state energy given by @xmath138 there is an energy gap @xmath139 to the first excited space ; all higher levels have energy @xmath140 . note that @xmath11 is independent of @xmath75 , the size of the lattice . intuitively , then , one may associate logical @xmath6 errors on any site with a fixed energy @xmath11 each . in summary , we have shown that the non - degenerate ground state of the hamiltonian @xmath141 is the cluster state , to zeroth order in @xmath74 , with an energy gap to the second excited state scaling as @xmath142 . we have shown that , to zeroth order in @xmath74 , the ground state of the system is the logical cluster state @xmath120 . however , the perturbation will also modify the energy eigenstates from their unperturbed states . to first order in @xmath45 , the perturbed ground state @xmath143 is given ( up to normalization ) as @xmath144 where @xmath145 ^ 2 p_l\,.\ ] ] for this perturbation , @xmath146 ; however , there exist states @xmath147 such that @xmath148 . note that @xmath45 is a sum of @xmath109 terms of the form @xmath149 acting across a bond . each of these terms applied to @xmath120 gives an excited state of the form @xmath150 using the anti - commutation relations of the pauli matrices , the terms in @xmath37 act on @xmath151 as @xmath152 hence @xmath153 , and therefore the states @xmath151 are in the first excited space of @xmath77 . ( [ operator1 ] ) also shows that @xmath154 and @xmath155 for @xmath156 are eigenvectors with different eigenvalues for the operator @xmath157 and thus they are orthogonal . however , recalling from earlier that @xmath158 stabilises @xmath120 , we have that @xmath159 and thus @xmath160 . hence @xmath161 , and @xmath162 there are @xmath75 states in the above sum , which determines the normalization . thus , we can calculate the fidelity @xmath163 of the ground state with the exact cluster state , which in this case is found to be @xmath164 this fidelity decays rapidly for increasing @xmath75 , which is unsurprising given that it is comparing quantum states on a large lattice and is an extensive quantity . for any lattice system with @xmath75 large , this fidelity is known to scale as @xmath165 , where @xmath166 is an intensive quantity that can be interpreted as the average fidelity per site @xcite . precisely , @xmath167 which is found to satisfy @xmath168 thus @xmath169 , which is independent of @xmath75 . this result demonstrates that the ground state is `` close '' to the ideal cluster state , as quantified by a large average fidelity per site , for @xmath170 . although it serves as an illustrative example of the techniques presented in this paper , the cluster state on a line is not a universal resource for mbqc ; a higher - dimensional lattice is required . in this section , we apply the perturbative procedure to lattice structures that are interesting from a mbqc perspective , and comment on their utility . for a hexagonal lattice in two dimensions with @xmath75 sites and periodic boundary conditions , the coordination number is 3 , and we require three physical qubits at each site ( see fig . [ fig:2dhexenergy ] ) . the hamiltonian for the lattice is again given by eq . ( [ hamilt ] ) . we investigate the spectrum of the unperturbed hamiltonian @xmath77 by considering its action at a single site , where the three site qubits interact via the ising coupling on a ring . the ground - state is degenerate , two - dimensional , and spanned by the states @xmath171 which encode our logical qubit . the energy of this ground state space is @xmath172 . the first excited state is six - dimensional , and has an energy of @xmath81 . thus , for the entire lattice of @xmath75 sites , the ground - state space has energy @xmath173 , is @xmath85-dimensional , and is spanned by all logical states of @xmath75 qubits . the first - excited space is @xmath174-dimensional , and has energy @xmath175 . we now turn to perturbation theory . it is again useful to obtain a conceptual view of the effect of @xmath45 on a single site , as illustrated in fig . [ fig:2dhexenergy ] . as the ground state space is spanned by @xmath176 and @xmath177 at each site , only the action of @xmath65 ( and not @xmath66 ) will map states out of the logical ground state space . the possible actions by the @xmath65 part of @xmath45 at a single site are shown in fig . [ fig:2dhexenergy ] . once again , @xmath104 and there is no first - order correction to the energies . the second order corrections @xmath117 are the eigenvalues of the operator @xmath118 defined in eq . ( [ opert ] ) . to evaluate the operator @xmath107 , we examine fig . [ fig:2dhexenergy ] and the ways of returning to the logical space after just two applications of @xmath45 . it is clear that there is only one possible contribution : if a @xmath108 in @xmath45 is applied twice to the same bond , this will yield the identity . the first @xmath65 can be applied to any of the qubits and then must be applied again to the same qubit , so there are @xmath178 such terms . hence @xmath179 using this result in eq . ( [ 2ndord2 ] ) as well as @xmath180 gives @xmath181 this operator simply acts as the identity on the logical space and so there is a constant second - order correction to the ground - state energy an energy shift given by @xmath182 the energy degeneracy of the ground state has still not been broken at second order and we must proceed to third order . the third order corrections @xmath183 are the eigenvalues of the operator @xmath184 given by @xmath185 ^ 2 p_l \nonumber \\ & = \frac{p_l v p_1 v p_1 v p_l}{(e_0^{(0)}-e_1^{(0)})^2}\,,\end{aligned}\ ] ] where the expression has been simplified using @xmath106 . with three applications of the perturbation @xmath45 , the operator @xmath186 maps states out of the ground state space and then back again via the first excited space . again investigating fig . [ fig:2dhexenergy ] , it is only possible for the lattice to remain in the ground state after three perturbation terms if @xmath65 operators are applied to each of the three qubits at a site ( and , through @xmath45 , the corresponding @xmath66 operators to the qubits on each of the neighboring sites ) . that is , this case will be a term applied to the logical space of the form @xmath187 where @xmath42 is the site connected to @xmath27 by a bond . just as in the case of the line , the operator @xmath64 acts on the logical space as @xmath188 , a logical cluster - state stabilizer operator . the three qubits at the site can be ordered in @xmath189 possible ways , and so there will be @xmath189 of these terms that occur at each site . therefore , @xmath190 and @xmath191 once again , as in the line , the set of @xmath85 states @xmath192 , running over logical @xmath6-errors on the cluster state @xmath120 at all possible sites , forms an orthogonal basis of @xmath86 and diagonalises @xmath184 . the cluster state @xmath120 is the unique lowest eigenstate of @xmath184 . the third - order correction for the energy associated with this state is @xmath193 again , this case is simple enough to analyze analytically ; the ground state of the hamiltonian @xmath141 has energy @xmath194 the @xmath92th excited space up to @xmath136 is @xmath137-dimensional and is spanned ( to zeroth order ) by states obtained from @xmath120 with @xmath92 logical @xmath6 errors . these states have energy @xmath140 where @xmath195 we can also calculate the first - order corrections to the ground state @xmath120 , by finding states @xmath196 such that @xmath148 . as before , define @xmath197 . by determining the effect of each of the terms in @xmath77 on @xmath151 it is clear that they are in the first excited space of @xmath77 and are orthogonal to each other . to first order in @xmath74 , @xmath198 comparing this ground state with the ideal cluster state , we find that the average fidelity per site @xmath166 is bounded by @xmath199 . we now repeat the above procedure for a 2d square lattice with @xmath75 sites and periodic boundary conditions . this case was originally examined in @xcite ; however , our detailed derivation reveals some errors in their calculation of the perturbed energies and the gap . the coordination number of this lattice is @xmath200 , and so four physical qubits are necessary at each site ( see fig . [ fig:2dsquareenergy ] ) . the hamiltonian for the lattice is again given by eq . ( [ hamilt ] ) , again with a ring of four qubits coupled via an ising interaction . there are now three energy levels of @xmath77 at a single site . the ground - state space of @xmath77 at a single site is spanned by the states @xmath201 and the energy of this ground state space is @xmath202 . the first excited state is twelve - dimensional , and has an energy of @xmath203 . the second excited state is two - dimensional and has a energy of @xmath204 . so , for the entire lattice of @xmath75 sites , the ground - state space has energy @xmath205 , is @xmath85-dimensional , and is spanned by all logical states of @xmath75 qubits . the first - excited space has energy @xmath206 and the second - excited space has energy @xmath207 . the possible actions by the @xmath65 part of @xmath45 at a single site are shown in fig . [ fig:2dsquareenergy ] . we now follow the identical procedure as done previously , and find @xmath208 that is , there are no first- or third - order corrections to the energy ; at second - order there is a constant energy shift @xmath209 to the ground state ; at fourth - order the degeneracy is broken . in the expression for @xmath210 , the first term is recognized as proportional to the cluster hamiltonian : the sum of stabilisers of the cluster state @xmath119 . therefore the set of @xmath85 states @xmath192 , running over logical @xmath6-errors at all possible sites on the cluster state @xmath120 , is an orthogonal basis of @xmath86 which diagonalises @xmath210 . the cluster state @xmath120 is the unique lowest eigenstate of @xmath210 , because it is an eigenstate of all stabilisers in the sum @xmath211 with eigenvalue @xmath1 . the fourth - order correction for the energy associated with this state is @xmath212 we note that this result differs , by numerical factors , from the result of @xcite . ( the error in @xcite arises from missing contributions to the perturbation operator @xmath210 in eq . ( [ theta4 ] ) . ) higher - order corrections follow in a similar fashion , and a complete analytic solution for the ground state energy is found to be @xmath213 the @xmath92th excited space up to @xmath136 is @xmath137-dimensional and is spanned ( to zeroth order ) by states obtained from @xmath120 by @xmath92 logical @xmath6 errors . these states have energy @xmath140 , where @xmath214 once again we can also calculate the first - order corrections to the ground state @xmath120 is calculated to be @xmath215 where @xmath216 . comparing this ground state with the ideal cluster state , we again find that the average fidelity per site @xmath166 is bounded by @xmath217 . we apply the now familiar procedure to a cubic lattice in three dimensions with @xmath75 sites and periodic boundary conditions . the coordination number is 6 , and so six physical qubits are necessary at each site ( see fig . [ fig:3dcubeenergy ] ) . the hamiltonian for the lattice is given in eq . ( [ hamilt ] ) , where we arrange the @xmath218 qubits on the vertices of a octahedron , and place @xmath78 couplings between all qubits connected by an edge of the octahedron , as in fig . [ fig:3dcubeenergy ] . there are four energy levels of @xmath77 at a single site ; the ground - state is degenerate , two - dimensional , and spanned by the states @xmath219 the energy of this ground state space is @xmath220 ; the first excited state space has energy @xmath202 ; the second excited state space has energy @xmath203 ; the third excited state space has energy @xmath204 . the possible actions of @xmath65 at a single site are shown in fig . [ fig:3dcubeenergy ] . again following our general perturbative procedure we find @xmath221 and @xmath222 thus , there are two constant energy shifts at second and fourth order of @xmath223 in the expression for @xmath224 , the degeneracy is broken by the terms @xmath225 which are the cluster stabilizers . the set of @xmath85 states @xmath192 , running over logical @xmath6-errors at all possible sites of the cluster state @xmath120 , forms an orthogonal basis of @xmath86 which diagonalises @xmath224 . the cluster state @xmath120 is the unique lowest eigenstate of @xmath224 . the sixth - order correction for the energy associated with this state is @xmath226 therefore , to sixth order , the energy of the ground state is @xmath227 the @xmath92th excited space up to @xmath136 is @xmath137-dimensional and is spanned ( to zeroth order ) by states obtained from @xmath120 by @xmath92 logical @xmath6 errors . these states have energy @xmath140 , where @xmath228 once again we can also calculate the first - order corrections to the ground state @xmath120 is calculated to be @xmath229 where @xmath216 . comparing this ground state with the ideal cluster state , we find that the average fidelity per site @xmath166 is bounded by @xmath230 . the cluster states on the three lattice types examined in this section ( the square , hexagonal and cubic lattices ) are all universal resources for quantum computation . in each case , it has been shown above that the perturbative procedure produces a non - degenerate ground state which approximates an encoded cluster state on the lattice . we chose to investigate each of these lattice structures because each has a unique relevance to the study of mbqc . the 2d square lattice is the canonical example for use in cluster - state quantum computing and was the original lattice structure presented in @xcite . this lattice is also the most easily accessible to experimental investigation in cold atomic systems @xcite . a hexagonal lattice was also examined above because ( as argued in @xcite ) the perturbative procedure produces a cluster state with the largest energy gap for a given ratio @xmath74 . we discuss the implications of this observation below . finally , recent work @xcite has shown that fault - tolerant thresholds can be found for mbqc if the lattice used is @xmath231-dimensional . following on from the discussion in @xcite , we now compare the results for each lattice and relate it to its usefulness for quantum computation . there are two sources of error when using the ground state obtained in the perturbative procedure for cluster - state quantum computation . first , note that errors will arise because the ground state of the system is not exactly the cluster state , but contains perturbative corrections ( cf . ( [ corr3 ] ) , ( [ corr4 ] ) , ( [ corr6 ] ) ) . in each case the ground state is given by a superposition of the cluster state with other first - excited states corresponding to `` errors '' @xmath232 applied to all bonds on the cluster state independently . this error rate is quantified by the average fidelity per lattice site @xmath166 , defined by eq . , which was explicitly bounded in all of the above examples . this bound takes the general form @xmath233 where @xmath234 is a constant of order one which depends on the lattice . this bound tells us that , for @xmath170 , the ground state is very close to the cluster state , and that the error rate for the independent @xmath232 is less than @xmath235 . because we require @xmath170 , this error probability will be small . the effect , and possible error correction , for such two - qubit correlated errors has not yet been investigated , but the independence and locality of the errors makes them amenable to existing error correction techniques . in particular , we note that such an error can be identified by checking sites , each of which _ should _ be in the code space spanned by @xmath35 and @xmath36 . errors of the form @xmath236 will cause a correctable error to this code space ( which must also include a phase correction to the appropriate neighbouring site ) provided that the lattice has coordination number @xmath237 , i.e. , for lattices of higher dimension than the 1-d line . this correction scheme would require measurements of multiple qubits , and it would be worthwhile to investigate whether such error correction could be performed using single - qubit measurements . the main difference arising in the calculations for each lattice structure , however , is the order in the perturbation theory at which the ground - state degeneracy is broken . this occurs at third order for the hexagonal lattice , fourth order for the square lattice and sixth order for the cubic lattice . in general , the order at which perturbation theory breaks the ground - state degeneracy is given by the coordination number of the lattice . this result leads directly to a dependence of the energy gap @xmath11 on the coordination number @xmath13 of the lattice , as @xmath238 in all cases the energy gap @xmath11 is independent of the size of the lattice , i.e. the system is gapped . given that the rate at which the thermal state of this system will exhibit @xmath6-errors depends explicitly on the size of this gap , the system will be less sensitive to these errors if the energy gap @xmath11 is made larger . the hexagonal lattice will have the largest energy gap , as is consequently less sensitive to thermal errors . it should be noted , however , that methods to identify and correct for such thermal errors ( and pauli errors in general ) within the mbqc paradigm currently exist only for 3-dimensional lattices @xcite . ( see also @xcite . ) the 2-d lattices ( hexagonal and square ) may not allow for error correction of such thermal errors using only single - qubit measurements ; this remains a key open question . we note that there exists a trade - off between these two types of errors when using the state for mbqc . increasing the value of @xmath74 will reduce the probability of thermal errors at a given temperature but also perturb the ground state away from the cluster state . the perturbative approach has so far been successful in producing the cluster state on each lattice type with periodic boundary conditions . we now analyse the effect of placing fixed boundaries on the lattice . we first examine a line with fixed endpoints . the interior sites still have coordination number @xmath239 , and so we require two physical qubits at these sites . however , the boundary sites will consist of just a single physical qubit . denote the number of interior sites by @xmath75 , so that there are @xmath240 sites in the entire line . in addition , denote the two boundary sites by the labels @xmath241 and @xmath242 . the hamiltonian for this lattice will remain that of eq . ( [ hamilt ] ) , where we do not place any site hamiltonian term on the boundary sites . the unperturbed energy spectrum at each of the interior sites is unchanged from the periodic boundary case ( as in fig . [ fig:1denergy ] ) . the two boundary qubits , however , have zero unperturbed energy . the spectrum of the unperturbed hamiltonian for the entire line is therefore a @xmath243-dimensional ground - state space with energy @xmath244 , and is spanned by all logical states of @xmath240 qubits . the first - excited space is @xmath245-dimensional , and has energy @xmath246 . we now turn to perturbation theory . note that , for the two boundary qubits , a single application of @xmath65 maps the logical space onto itself . thus , due to the contributions from the boundary qubits , there is now a first order correction to the energy @xmath247 where @xmath248 and as usual @xmath2 . in particular , the cluster stabilizers for the end sites are given by the product of an @xmath5 operator on the boundary site with a single @xmath6 operator on its sole neighbour . the first - order corrections to the ground - state energy , @xmath102 , are the eigenvalues of @xmath249 . this operator is diagonal in our familiar basis for @xmath86 of the @xmath243 states @xmath192 , running over logical @xmath6-errors at all possible sites . the states in this basis which are the @xmath1-eigenstates of both @xmath250 and @xmath251 will be the lowest eigenvalues of @xmath249 . by the property of stabilisers , the subspace @xmath252 which is stabilised by @xmath253 and @xmath251 is @xmath85-fold degenerate , and so the degeneracy is reduced by a factor of 4 at first order . the cluster state @xmath120 is contained in this subspace , as are all states with logical @xmath6-errors anywhere _ except _ on the boundary . thus the lowest energy space @xmath254 has a first - order correction given by : @xmath255 the next highest energy level includes states which have a @xmath6-error at either of the boundary sites @xmath256 or @xmath257 but not both . these @xmath258 states have @xmath259 , which is a gap of @xmath260 above the space @xmath254 . the second highest energy level will include states which have @xmath6-errors at both boundary sites , and in this case @xmath261 . the second order correction is calculated in an identical manner to the case with periodic boundary conditions . we have @xmath262 the above operator is already diagonal in our chosen basis ( the states of the form @xmath121 ) . of the states in @xmath254 , the cluster state @xmath120 is the unique @xmath1 eigenstate of all the stabilisers in the above sum , and so will be the lowest eigenvalue of @xmath118 . thus the second - order correction for the energy associated with this state is @xmath128 so in the case of fixed boundary conditions , the cluster state is still the ground state produced ( to zeroth order ) , with energy to 2nd order of @xmath263 a state @xmath121 with @xmath264 @xmath6-errors at boundary sites and @xmath265 @xmath6-errors at interior sites will have energy : @xmath266 where @xmath267 and @xmath268 . provided @xmath269 , the energy gap will be @xmath270 , the same energy gap which was obtained with the periodic boundary conditions . the first order corrections to this ground state will be given by @xmath271 and the bound on @xmath166 , the average fidelity per site , remains the same as for periodic boundary conditions . consider a square lattice of dimension @xmath272 . we define the number of interior ( non - boundary ) sites to be @xmath75 , and so @xmath273 and the total number of sites is @xmath274 . interior sites have coordination number @xmath200 , edge sites coordination number @xmath231 , and corner sites coordination number @xmath239 , determining the number of qubits at each site . denote the set of corner boundary sites by @xmath275 , the set of edge boundary sites by @xmath276 , and the set of interior sites by @xmath277 . each of the three type of sites will have a different unperturbed spectrum at each , corresponding to figs . [ fig:1denergy ] , [ fig:2dhexenergy ] and [ fig:2dsquareenergy ] respectively . the zeroth order energies of the lattice are now much more complicated due to the presence of these three different types of sites . the ground state of @xmath77 for the entire lattice , spanned by all possible logical states , is @xmath278-dimensional ground - state space with energy @xmath279 . the next four excited states separated by a energy gaps of @xmath89 . at first - order in the perturbation , @xmath280 , and thus there is still no first - order correction to the energies . at second , third , and fourth order , we have @xmath281 ( in this expression , @xmath282 is the number of @xmath6-errors at sites in @xmath275 relative to the cluster state . it appears is this expression because @xmath210 depends on the second - order energies . ) this operator is diagonal in the familiar basis @xmath192 . the corresponding corrections to the lowest energy ground - state energy are @xmath283 thus , the non - degenerate ground state of the system is the cluster state @xmath120 , to zeroth order , with an energy to fourth - order given by @xmath284 a state @xmath121 obtained from the cluster state by @xmath282 @xmath6-errors at sites in @xmath275 , @xmath285 @xmath6-errors at sites in @xmath276 and @xmath286 @xmath6-errors at sites in @xmath277 will have energy @xmath287 where @xmath288 for the range @xmath289 , the energy gap will be @xmath290 , the same energy gap obtained using periodic boundary conditions . in summary , the perturbative procedure is still successful on the line and square lattice with fixed boundary conditions , producing an approximate cluster state as the non - degenerate ground state . furthermore , the energy gap to the first - excited space is unchanged from the case with periodic boundary conditions . extending these results to other lattices is straightforward . we have shown that the perturbative procedure presented here is successful in producing a non - degenerate ground state that approximates the cluster state on all of the lattice types examined so far . in fact , the cluster state on any lattice type with any boundary conditions , or more generally on any graph , can be approximated using this method . we now outline a proof of this result . for any graph , we place at each site a number of physical qubits equal to the coordination number ( the number of bonds connecting that site to others ) and take the hamiltonian as in eq . ( [ hamilt ] ) . we note that the form of the site hamiltonian @xmath37 needs only yield a two - dimensional degenerate ground state spanned by @xmath35 and @xmath36 of all qubits at each site ; aside from this requirement , its precise form is quite flexible . we first show that the operators produced at each order in perturbation theory will always possess the cluster state as an eigenstate , and more generally are diagonalized by the set of cluster states with @xmath6 errors . note that the operators that arise at each order of the perturbation theory are linear combinations of terms of the form @xmath291p_l\,,\ ] ] for some integers @xmath292 and @xmath293 , where @xmath294 . now , @xmath45 is a sum of operators @xmath295 where @xmath296 , and therefore @xmath297 will be a sum of operators which map states in the logical space through the illogical spaces ( by applications of @xmath298 over various bonds ) and then return it to the logical space . from the pauli operator commutation relations , we note that every term @xmath298 that is applied to the logical space either commutes or anticommutes with the terms in @xmath37 and so it will always yield an eigenstate of @xmath37 . thus , successive applications maps the logical space to eigenspaces of definite unperturbed energy , and the term @xmath299 will just be a multiplicative constant . furthermore , the fact that each application of @xmath298 keeps the system in some eigenstate of @xmath37 means that all the terms in the sum are of the form @xmath300 , where @xmath301 is some product of the @xmath298 . now , suppose @xmath301 does not commute with all the terms in @xmath37 ( i.e. it anticommutes with at least one of them ) , then the effect of this term will be that @xmath301 maps logical states to illogical ones ( the resultant state will have a @xmath123 eigenvalue for at least one term in @xmath37 , whereas the logical space is the @xmath1 eigenstate of all the terms in @xmath37 ) . thus in this case @xmath302 . the only non - zero operators in the perturbation theory will be of the form @xmath300 where @xmath301 commutes with all the terms in @xmath37 ( i.e. all the @xmath303 ) . but then , if @xmath301 commutes with each term in @xmath37 it commutes with @xmath94 , the projection onto the logical subspace . thus , we have the eigenvalue relation @xmath304\notag\\ & = p_l k |c_2\rangle|c_2\rangle\cdots|c_2\rangle \notag\\ & = p_l |c_2\rangle|c_2\rangle\cdots|c_2\rangle = |c\rangle\end{aligned}\ ] ] where the last line follows because @xmath305 , and @xmath301 is a product of @xmath298 terms . hence , we have shown that all the terms that arise at each order in the perturbation theory stabilise the cluster state . this certainly shows that , to zeroth order , the cluster state is one of the eigenstates selected out of the degeneracy by the perturbation . we now show that it is the non - degenerate ground state . each term of the form @xmath300 stabilises the cluster state , and the eigenvalues of @xmath300 are restricted to @xmath306 . therefore , all that must be checked is that the sign in front of @xmath300 in the perturbation theory is negative to ensure that the cluster state is selected as the _ ground state_. suppose @xmath301 is a product of @xmath307 @xmath298 terms . then the operator @xmath300 will first appear at the @xmath307th order as a term in @xmath308^{m-1 } p_l$ ] . the @xmath309 operators will always contribute a sign @xmath310 to the energy correction . furthermore , each @xmath298 carries with it a negative sign in the definition of @xmath45 , contributing a further @xmath311 to the the energy correction . therefore , each @xmath300 will always appear in the energy correction with a negative sign , thus selecting the cluster state as the ground state . moreover , because it is clear that the cluster state stabilisers @xmath225 , will always arise as one of the @xmath300 terms in the perturbation theory , the ground state must be non - degenerate ( as the state stabilised by these operators is unique ) . hence we have shown what we set out to prove : that the cluster state on any lattice type can be approximated using this method . in fact , this perturbative approach can also be further generalised to approximate other states with a peps description . the existence of gapped quantum many - body systems , with hamiltonians consisting of only two - body nearest - neighbour interactions , for which the ground state encodes a cluster state allowing universal mbqc is an exciting result for the potential realisation of a quantum computer . the obvious avenue for future investigation is whether existing natural or artificial materials exist with interactions similar to those described here . as we have shown , as the perturbation parameter @xmath74 becomes larger , the ground state begins to deviate from the cluster state due to perturbative corrections . in this work , we have analysed these corrections as a source of error . it may also be fruitful , however , to analyse the usefulness of the finite @xmath74 ground state for mbqc in terms of the performance of a universal set of quantum gates , as in @xcite . although we do not believe the model investigated here exhibits a phase transition at any @xmath74 ( this is however an open question ) , it may nevertheless be possible that the usefulness of the ground state for mbqc undergoes some form of sharp transition @xcite . this work was supported by the australian research council . we thank sergey bravyi for identifying that the peps hamiltonian ground state was a simultaneous @xmath1 eigenstate of the encoded stabilizer operators @xmath64 , and for identifying their role in obtaining exact solutions to this model . we thank andrew doherty , terry rudolph and stein olav skrvseth for helpful discussions . we briefly outline the formalism of degenerate perturbation theory and the notation that we use , closely following ref . @xcite . suppose that the hamiltonian has the form @xmath312 , and that the eigenvalue problem has been solved exactly for @xmath37 . the corrections brought about by the introduction of the perturbation @xmath45 can then be approximated by a power series expansion in @xmath45 . for perturbation theory to converge , the magnitude of the largest eigenvalue of @xmath45 must be smaller than that of @xmath37 . suppose the unperturbed spectrum has a degenerate subspace @xmath86 with energy @xmath313 and we are interested in finding out how this energy degeneracy is broken . after the perturbation has been applied , denote the perturbed eigenstates of this subspace by @xmath314 and the perturbed energies by @xmath315 , for @xmath316 , i.e. , @xmath317 denote the projection onto the degenerate subspace @xmath86 as @xmath94 and define @xmath318 . then we can decompose @xmath314 as @xmath319 , where @xmath320 and @xmath321 . applying this decomposition to eq . ( [ a1 ] ) , we have : @xmath322 note that @xmath323 and therefore that @xmath324 . multiplying eq . ( [ a2 ] ) by @xmath94 and @xmath325 respectively , we obtain @xmath326 eq . ( [ a4 ] ) has the formal solution @xmath327 which can be substituted back into eq . ( [ a3 ] ) to obtain @xmath328 where @xmath329 this equation allows us to determine perturbed energy at any order of the perturbation theory . so far no approximations have been made . to implement the perturbation theory , it is necessary to expand @xmath330 as a power series in @xmath45 . we use @xmath331^{m}\bigr)(e_i - h_0)^{-1}\,.\end{aligned}\ ] ] the energies @xmath315 in these expressions must also be expanded as power series , @xmath332 . then we have @xmath333^{-1}(e_l^{(0)}-h_0)^{-1}\notag\\ & = \lambda\omega\end{aligned}\ ] ] where we have defined the operators @xmath334 the operator @xmath330 can then be expressed as @xmath335^{m } \lambda\omega \overline{p}_l v p_l\,,\ ] ] where we have used the fact that @xmath325 commutes with @xmath309 and @xmath336 . additionally , @xmath336 must be further expanded out as a power series @xmath337 we can now identify the terms in eq . ( [ a10 ] ) of each order . specifically , denote the terms in @xmath330 of @xmath234th order by @xmath338 , so that @xmath339 . then , when approximated to @xmath92th order in @xmath45 , eq . ( [ a6 ] ) becomes @xmath340 which is an eigenvalue equation over the subspace @xmath341 . ( we note that @xmath342 because @xmath343 at zeroth order and the perturbation is assumed small ) . the energy corrections to @xmath92th order @xmath344 are the eigenvalues of the operator @xmath345 . furthermore , the eigenstates that are selected by the perturbation to break the degeneracy ( to zeroth order ) are just the eigenvectors of @xmath345 corresponding to each eigenvalue . note that @xmath338 depends on lower order energy corrections and so each lower order correction must be calculated before proceeding to the higher order corrections . at each stage we must insist that the @xmath92th order energies differ from the @xmath346th order energies only by an amount @xmath92th order in @xmath45 , which removes any non - physical solutions . to determine the explicit form of the @xmath338 , we simply substitute eq . ( [ a10l ] ) into eq . ( [ a10 ] ) and identify the terms of the required order . clearly @xmath347 , and so to first order in @xmath45 , eq . ( [ a7 ] ) becomes @xmath348 thus , the first order energy corrections to states in @xmath86 are just the eigenvalues of the matrix @xmath349 . in the non - degenerate case we see that eq . ( [ a11 ] ) reduces to the well - known expression @xmath350 . if this eigenvalue spectrum is still degenerate , then the degeneracy is not completely broken at first order . it is then necessary to go to second order @xmath351 again we can examine whether the degeneracy has been broken at this stage by examining the eigenvalues and eigenvectors of the above operator @xmath352 . if not , we must continue to proceed to higher orders . the formulae for higher orders become increasingly complex , but they simplify if we assume that @xmath353 , which will be true for all the cases of interest that we investigate in this paper . for example we find ( with @xmath353 ) @xmath354 ^ 2 p_l \notag\\ \theta^{(4)}&= p_l v [ \omega \overline{p}_l v ] ^3 p_l- e_i^{(2 ) } p_l v\omega^2 \overline{p}_l v p_l\,,\end{aligned}\ ] ] and so forth . there are two additional points that must be noted . first , from eq . ( [ a10 ] ) , it can be concluded that @xmath355 are the eigenvalues of the operator @xmath356 for each order @xmath357 . however , the energies @xmath358 are not generally the eigenvalues of the operators @xmath338 ; this will only be true in general when all operators @xmath338 can be simultaneously diagonalised . fortunately , for the systems investigated in this paper , it can be shown that the @xmath338 commute with each other and therefore the energies @xmath358 are indeed the eigenvalues of the operators @xmath338 . second , there are some general properties we can note about the form of the @xmath338 operators . from the form of eq . ( [ a10 ] ) and the series expansion in eq . ( [ a10l ] ) , it is clear that all terms in the expansion of @xmath330 are proportional to operators of the form @xmath359 for some @xmath360 . this result is used in sec . [ sec : generalgraph ] . the above analysis allows the zeroth order energy eigenstates to be determined . we now direct our attention to the first order corrections to these states . suppose we have determined that @xmath361 to zeroth order for some @xmath362 using the above method . from eq . ( [ a5 ] ) and using eqs . ( [ a8 ] ) and ( [ a9 ] ) we have , to first order in @xmath45 , @xmath363 where @xmath147 are the eigenstates of @xmath37 with energy @xmath364 . determining @xmath365 to first order is somewhat more complicated : even though @xmath366 to zeroth order , it is possible for first order corrections to come from other states @xmath367 . however , if the energy degeneracy between @xmath368 and @xmath369 is only broken at order @xmath370 , then one must go to the order @xmath371 equations just to determine the _ first _ order eigenstate corrections to @xmath365 . in this case , @xmath372 but @xmath373 . then , at order @xmath371 , eq . ( [ a7 ] ) reads @xmath374 taking the inner product with @xmath369 and rearranging gives latexmath:[\[\begin{aligned } \langle l |\psi_i\rangle_l&=\frac{\langle l |\theta^{(m_l+1 ) } |\psi_i\rangle_l -e_i^{(m_l+1)}\langle l |\psi_i\rangle_l}{e_i^{(m_l)}-e_l^{(m_l ) } } \notag\\ & = \frac{\langle l |\theta^{(m_l+1 ) } @xmath45 . therefore , to first order in @xmath45 , we have latexmath:[\[\begin{gathered } \label{a19 } |\psi_i\rangle = |i\rangle + \sum_{|j\rangle \in \mathcal{h}_{\bar{l } } } \frac { \langle j | v |i\rangle } { e_l^{(0)}-e_j^{(0)}}|j\rangle \\ + \sum_{|l\rangle \in \mathcal{h}_l , l \neq i } \frac{\langle l |\theta^{(m_l+1 ) }

we have obtained a moderate resolution spectrum of the quasar pks 0528 - 250 with the red channel spectrograph on the multiple mirror telescope ( mmt ) in order to study a damped ly absorption line system at z = 2.8115 . we obtain a new upper limit for the co column density for the z = 2.8108 velocity component in the z = 2.8115 damped ly system . the ionization of different species in this component rules out a quasar spectral energy distribution ( sed ) as the ionization field , and implies an ultraviolet radiation field intensity a few times that of the milky way value . the estimated total number density is n(h ) 20 . the physical size for the z = 2.8108 component implied by these models is about 40 parsecs . the ionization of different species also suggests a structure with a hot intercloud medium associated with a h i cloud in this component , that is , most low ionized ions are from the cold medium where photoionization and photodissociation dominates . the highly ionized species may be from the intercloud medium where collisional ionization dominates . we also present newly identified ni ii absorption lines in the z = 2.1408 and z = 2.8115 damped ly systems . the derived depletion of nickel by dust confirms previous results that the dust - to - gas ratio in these two damped ly systems is about 10% of the milky way ratio . millimeter wavelength observations obtained at the nrao 12 meter telescope provide new upper limits on co ( 3 - 2 ) emission in the z = 2.8115 damped ly system .

astro-ph9708104

damped ly@xmath0 quasar absorption line systems may be the high redshift analogs of present - day galactic disks ( wolfe _ et al . they dominate the cosmological mass density of neutral gas in the universe and hence trace the bulk of material available for star formation at high redshifts ( lanzetta 1993 ) . if the physical conditions in the damped ly@xmath0 systems are similar to those in galactic interstellar clouds , the damped ly@xmath0 systems should show absorption from low - ionization species like c i , trace elements like ni , zn and cr and molecules like h@xmath3 and co. most of these species have been detected in absorption in the damped ly@xmath0 systems ( e.g. foltz _ et al . _ 1988 ; meyer _ et al . _ 1986 , 1987 , 1989 ; pettini _ et al . _ 1994 , 1996 ; ge , bechtold & black 1997 ; ge & bechtold , 1997 ) , with the notable exception of co. detection of molecules at high redshift would provide a strong diagnostic for the conditions in the clouds that harbor them . to detect absorption lines from c i , ni ii , zn ii , cr ii and co , high signal - to - noise ratio ( s / n ) spectra are required because the expected lines are very weak . pks 0528 - 250 is one of the brightest quasars known , m(1450)=17.73 ( sargent _ et al . _ 1989 ) , and so is a good object for detailed work . toward pks 0528 - 250 , there are two damped ly@xmath0 systems at z = 2.1408 and z = 2.8115 . the z = 2.8115 damped ly@xmath0 system has n(h i ) @xmath4 @xmath5 and is one of the highest column density damped ly@xmath0 systems known ( mller & warren 1993 ) . the z = 2.8115 damped ly@xmath0 system is a good candidate to detect c i and co. foltz _ ( 1988 ) have detected h@xmath3 absorption associated with it . lanzetta ( 1994 , private communication ) and songaila & cowie ( 1995 ) have further confirmed this detection . the redshift for the h@xmath3 absorption is z = 2.8118 ( lanzetta 1994 , private communication ) . the column density of molecular hydrogen , n(h@xmath3 ) = 2.6@xmath6 @xmath5 , a factor of 10 larger than the upper limits for other damped systems ( songaila & cowie 1995 ; levshakov _ et al . _ 1992 and references therein ) . the molecular fraction f=2n(h@xmath3)/[2n(h@xmath3)+n(h i)]=1.2@xmath7 is similar to the milky way diffuse clouds ( savage _ et al . previous observations show that the h@xmath3 , co and c i column densities are correlated with each other in the milky way diffuse clouds , e.g. n(h@xmath3)/n(c i ) is @xmath8 - @xmath9 and n(c i)/n(co ) is @xmath10 ( e.g. federman _ et al . thus , we expect that the c i and co column densities are high enough to be detectable . mller & warren ( 1993 ) detected ly@xmath0 emission at z = 2.8121 . the detected ly@xmath0 line luminosity of 1.1@xmath11 ergs s@xmath12 corresponds to a star formation rate ( sfr ) of @xmath13 m@xmath14yr@xmath12 , if there is no attenuation by dust ( we adopt q@xmath15 , h@xmath16 km s@xmath12mpc@xmath12 ) . the minimum size of the absorber is about 18@xmath17 kpc . to study the nature of the z = 2.8115 damped ly@xmath0 system we have conducted two kinds of observations : ( 1 ) high signal - to - noise optical spectroscopy covering the strongest absorption bands of the co a - x system ( 0 - 0 , 1 - 0 , 2 - 0 , 3 - 0 , 4 - 0 , 5 - 0 , 6 - 0 ) , c i @xmath18 1560 and some ni ii lines ; and ( 2 ) millimeter observations of co ( j=3 - 2 ) emission at 90 ghz . the observations are described in section 2 . identifications of absorption lines are discussed in section 3 . we discuss our results in section 4 and give a summary in the final section . the observations of pks 0528 - 250 were obtained on february 14 , 1994 with the loral 800@xmath191200 ccd attached to the red channel spectrograph on the mmt ( schmidt _ et al . the 1200 l / mm grating was used in first order with 1@xmath20 slit , which resulted in spectral resolution of 2.0 ( fwhm ) , measured from the comparison lamp lines . the seeing was 0.5@xmath21 ( fwhm ) . we took three 40 minute exposures with wavelength coverage from 5200 to 6100 . the quasar was stepped by a few arcseconds along the slit between each exposure to smooth out any residual irregularities in the detector response which remained after flat - fielding . an exposure of a he - ne - ar - cd - hg lamp and a quartz lamp were done before and after each exposure of the object to provide an accurate wavelength reference , a measure of the instrumental resolution , and a flat - field correction . bias frames were also obtained . the spectra were reduced in the standard way using iraf . the summed spectrum weighted by the signal - to - noise from the three exposures is shown in figure 1 . the signal - to - noise ratio is about 50 in most of the wavelength coverage . the continuum was fitted iteratively using a cubic spline routine . first the spectrum was divided into some moderately wide bins ( 40 pixels ) and averaged in each bin . then a cubic spline was fit to these averaged points . points which deviated negatively by more than two standard deviations were flagged as possible significant lines and rejected for subsequent iterations of the fit . this process was repeated about six times until the fit did not change , i.e. until the number of flagged pixels above the fit was about equal to the number of pixels below the fit and the reduced chi - squared of all unflagged pixels was near unity ( bechtold 1994 ) . significant lines were found by measuring the equivalent width in a bin equal to 2.46 times the number of pixels of the fwhm of the comparison lines , and its error ( c.f . young , sargent & boksenberg 1982 ) . all such bins where the equivalent width was significant at greater than or equal to 3 @xmath22 were flagged . the final equivalent width of each absorption line was measured by specifying the wavelength interval for each feature by hand . the central wavelengths , equivalent widths and their errors are listed in table 1 . the central wavelength of each line in table 1 is the centroid , weighted by the depth of each pixel in the line profile below the continuum ( c.f . bechtold 1994 ) . the error for the central wavelength shown in this table is from the uncertainty in the measurement of the equivalent width . we neglected the uncertainty from the wavelength calibration because the error from the wavelength calibration is very small , typically less than 0.02 . all reported wavelengths are vacuum and have been corrected to the heliocentric frame . our millimeter observations of the z = 2.8108 damped ly@xmath0 system toward pks 0528 - 250 were made in january 1994 at the nrao 12 meter telescope . all of the observations were done with the dual polarization sis 3 mm mixer receiver , which provided a system temperature of 260 k at 90 ghz . the central frequency was set to 90.74093 ghz corresponding to the heliocentric redshifted co ( j=3 - 2 ) emission frequency of the z = 2.8108 damped ly@xmath0 system . the telescope half - power beam width ( hpbw ) , @xmath23 , was 68@xmath21 . the frequency calibration was checked with observations of orion a. the pointing was checked with observations of q 0607 - 157 . the spectra were obtained using absolute position switching between the source position and a reference position 3@xmath24 away in azimuth . we used a 256 channel filter bank spectrometer with a spectral resolution of 2mhz . we obtained 498 minutes of integration on source in four nights ; scans were six minutes in length . each scan had the same three bad channels removed before the scans were summed . two noisy scans were omitted from the final summation . before taking the mean of a channel , we eliminated channels with more than a 10@xmath22 deviation from a robust median . at that point , a final 1@xmath22 deviation of each channel was calculated ( walker _ et al . the final spectrum is shown in figure 2 , where a single linear baseline has been fitted ( by iteratively throwing out channels with intensities greater than 3.5@xmath22 ) and then removed . in figure 2 , the 1@xmath22 error is represented as a dashed line . the results are discussed in section 4.4 . table 1 lists identifications of absorption lines in the optical spectrum of 0528 - 250 . the redshifts of absorption lines are consistent with those identified by previous authors ( morton _ et al . _ 1980 ; meyer & york 1987 ; sargent _ et al . _ 1989 ) . a few weak lines in table 1 have not been identified and may be spurious . the z = 2.671 and z = 2.673 c iv absorption line systems are newly identified by us . we describe the different redshift systems separately in the following . z@xmath25 the redshift agreement is good between the two components of the mg ii doublet . the equivalent widths for these two lines are about half the values obtained by sargent _ the discrepancy may be caused by line blending in their data . z@xmath26 the equivalent width of al ii @xmath18 1671 from this damped ly@xmath0 system is consistent with that measured by other authors ( morton _ et al . _ 1980 ; sargent _ et al . we identify the absorption line at @xmath27 as al iii @xmath18 1863 . the other member of the doublet , al iii @xmath18 1860 at @xmath28 , is blended with the strong absorption line si ii @xmath18 1526 from the z@xmath29 damped ly@xmath0 system . we have confirmed the identifications of two weak absorption lines of ni ii @xmath18 1709 , ni ii @xmath18 1741 by meyer _ ( 1989 ) and we identify another weak ni ii transition , ni ii @xmath18 1751 . we discuss these further in section 4.2 . z@xmath30 the equivalent widths of the c iv doublet are statistically consistent with those obtained by other authors ( morton _ et al . _ 1980 ; sargent _ et al . we have not detected absorption from si ii @xmath18 1526 in our spectrum . z@xmath31 this c iv absorption line system is tentatively identified here . there is a ly@xmath0 absorption line at @xmath32with w@xmath33 corresponding to z@xmath34 ( morton _ et al . _ 1980 ) . however , it is blended with the si iv @xmath18 1402 line of the z@xmath35 damped ly@xmath0 system . the absorption lines at 5614.02 and 5622.57 could also be the absorption lines of co ii @xmath36 1472 , 1475 from the z = 2.8115 damped ly@xmath0 system . z@xmath37 these two c iv absorption line systems are first identified here . the c iv 1550 absorption line from the z = 2.6710 system is blended with the c iv 1550 line from the z = 2.6732 system . the z = 2.6732 system has a moderate ly@xmath0 absorption line in the ly@xmath0 forest at @xmath38 with w@xmath39 corresponding to z@xmath40 ( morton _ et al . _ 1980 ) . z@xmath29 the equivalent widths of the strong absorption lines si iv @xmath18 1393 , 1402 , si ii @xmath18 1526 , c iv @xmath18 1548 , 1550 are consistent with earlier measurements of morton _ et al . _ ( 1980 ) , meyer & york ( 1987 ) and sargent _ et al . _ we have confirmed that the z@xmath29 system consists of at least two components with z@xmath41 ( a@xmath42 ) and z@xmath43 ( a@xmath3 ) respectively ( morton _ et al . _ 1980 ) , corresponding to a velocity separation of 45 km s@xmath12 . we have detected three ni ii absorption lines from the z@xmath29 system at better than the 3@xmath22 level . the absorption line at @xmath44 is identified as ni ii @xmath18 1370.13 at z = 2.8119 instead of c iv @xmath18 1548.20 at z = 2.3733 by chen & morton ( 1984 ) because no absorption line was detected for c iv @xmath18 1550.77 at the same redshift , and the equivalent width should be at least 0.32 if both members of the doublet are unsaturated , compared to our 5@xmath22 upper limit of 0.26 . the ni ii @xmath18 1370 and ni ii @xmath18 1454 lines exhibit the same two redshift components as the low ionized ions si ii , fe ii , oi and al ii lines . the absorption line of ni ii @xmath18 1467 , which is weaker than ni ii @xmath18 1370 , 1454 , does not exhibit two velocity components probably due to limited s / n and resolution . the lines of the ionized species , c iv @xmath18 1548 , 1550 and si iv @xmath18 1393 , 1402 are seen from the z@xmath41 system only ( morton _ et al . _ 1980 ; meyer & york 1987 ) . there is an absorption line at the wavelength corresponding to c i @xmath18 1560.31 from the z = 2.8115 damped ly@xmath0 system at about the 4@xmath22 significance level . the rest frame equivalent width of this line is @xmath45 m which is lower than the 2@xmath22 upper limit of 50 m given by meyer _ if this is the c i absorption line , then the column density of c i is @xmath46 @xmath5 , assuming that the line is on the linear part of the curve - of - growth . however , much higher quality spectrum with the resolution of about 8 kms@xmath12 and s / n @xmath47 50 from the keck hires by songaila _ et al . _ ( 1995 , private communication ) rules out this identification . their spectrum does not show any absorption lines from c i @xmath18 1277.24 and c i @xmath18 1328.83 to the 3 @xmath22 upper limit of n(c i ) @xmath48 @xmath5 , which is much smaller than that from our data . the 4 @xmath22 absorption feature in our mmt spectrum is probably a statistical artifact . therefore , in what follows , we adopt the measurements by songaila _ et al . _ we have not detected absorption of the excited si ii * @xmath18 1533 from the a@xmath42 component in the z = 2.8115 damped ly@xmath0 system . the 3 @xmath22 upper limit of equivalent width is 0.118 which will be used to derive the number density of hydrogen in this component ( see section 4.1 ) . the absorption of si ii * @xmath18 1533 from the a@xmath3 component in the z = 2.8115 damped system is blended with the al iii @xmath18 1863 line from the z = 2.4108 damped ly@xmath0 system . the absorption line at @xmath49 is consistent with the co a - x 1447.36 absorption line at z@xmath43 . while co @xmath18 1447.36 is one of the strongest co absorption lines ( e.g. field _ et al . _ 1983 ; sheffer _ et al . _ 1992 ) , we failed to detect the co a - x @xmath36 1477.52 , 1509.70 absorption lines which have similar strengths . the absorption line at 5519.40 could also be the co ii @xmath18 1448 line from the a@xmath42 component of the z = 2.8115 damped system . then the absorption line at 6000.53 could be co ii @xmath18 1574 lines from the same component . however , we also failed to detect co ii @xmath36 1466 , 1480 in our spectrum which have similar strengths ( morton 1991 ) . therefore , the identification for the absorption at @xmath18 = 5519.40 is uncertain . we have constructed models using the cloudy program ( ferland 1993 ) in order to investigate the ionization of the observed species , the spectrum of the uv ionizing radiation field , and the molecular fraction of the damped ly@xmath0 absorption line system at z = 2.8115 . the column densities for the species observed in the a@xmath42 and a@xmath3 components ( z = 2.8108 , z = 2.8132 ) of the z = 2.81 damped ly@xmath0 system are shown in table 2 . because the resolution of the available spectra is insufficient to resolve the profiles of the lines listed in table 2 , we use a curve - of - growth analysis to infer column densities . we adopt the doppler parameter b = 60 km s@xmath12 for lines from the a@xmath42 component and b = 30 km s@xmath12 for lines from the a@xmath3 component which are obtained from the curve - of - growth of fe ii ( meyer _ et al . if the absorption lines from the a@xmath42 and a@xmath3 components are blended with each other , we adopt b = 100 km s@xmath12 obtained from fitting the lyman lines ( foltz _ et al . _ 1988 ) . however , the si iv and c iv doublets in the a@xmath42 component suggest that the doppler parameter b = 60 km s@xmath12 derived for fe ii lines does not apply . a higher doppler parameter b = 80 km s@xmath12 gives a better fit . therefore , we adopt b = 80 km s@xmath12 to calculate column densities of high ionization species , c iv , si iii , si iv , n v , o vi in the a@xmath42 component ( table 2 ) . the b - values indicate there are probably several velocity components blended with each other in each line . however , the uncertainties in the column densities from single b - value curve - of - growth analysis may be less than a factor of 2 ( jenkins 1986 ) . in fact , our derived column densities are consistent with the results given by songaila & cowie ( 1995 ) , and lu _ et al . _ ( 1996 ) based on higher resolution spectra . as the input to cloudy , we adopt a metallicity of 10% of the solar value for all the elements in both a@xmath42 and a@xmath3 components of the z = 2.8115 damped ly@xmath0 system . we also consider the relative depletions by dust grains to zn ( about 0.6 - 0.7 in the logarithm unit ) of heavily depleted elements such as al , si , ca , fe and ni ( meyer _ et al . _ 1989 ) . the dust - to - gas ratio is 10% of the milky way and the constituents of dust grains are assumed the same as the milky way . the total neutral hydrogen column density n(h i ) for the two components is @xmath50 @xmath5 ( mller & warren 1993 ) . it is obtained by fitting the profile of the damped ly@xmath0 absorption line including the ly@xmath0 emission of the quasar because the redshift of the damped ly@xmath0 system ( z = 2.8115 ) is about the same as that of the quasar ( z = 2.765 , morton _ et al . _ 1980 ; mller & warren 1993 ) . the neutral hydrogen column density for each of the two components in the z = 2.8115 damped ly@xmath0 system is uncertain . however , similar absorption of dominant ions of most species , such as o i , si ii , s ii , fe ii , ni ii , in these two components suggest that the total neutral hydrogen column density in each component is also similar ( cf . table 2 ) . we take the total column density as the approximate value for each of the components in the following discussion . there is therefore at least a factor of 2 uncertainty in the results from cloudy from the uncertainty in the neutral hydrogen column density in each velocity component . since the damped ly@xmath0 system is possibly near the quasar , the uv flux from the quasar can affect the ionization . we therefore have considered five types of spectral energy distributions ( sed ) . figure 3 shows these seds , along with the ionization potential of several ions . we use a power - law with index @xmath51 for @xmath52 , where @xmath53 , measured for pks 0528@xmath54250 by sargent _ et al . _ we extrapolate the power to longer wavelengths which has little affect on our results . the quasar energy distribution at wavelengths shorter than the lyman limit is uncertain . we therefore consider two different quasar seds : a(1 ) and a(2 ) ( figure 3 ) . in these two models , the upper limit to the x - ray flux from the einstein ipc at 1.0 kev is adopted for the wavelength longer than 0.2 kev ( wilkes _ et al . _ 1994 ) . in a(1 ) , we assume a power - law with index between the lyman limit and 0.2 kev of @xmath55 ; this is the hardest spectrum that seems reasonable . in a(2 ) , the sed between the lyman limit and 0.2 kev is a power law connecting these two points . we consider the models for the evolution of galactic sed s , as calculated by bruzual & charlot ( 1993 ) . the shape of the uv sed in the @xmath56 region does not change after about 0.1 gyr if the sfr is constant , so we consider the sed of 1 gyr old galaxy as representative . this is shown in the figure 3 as model b. for an age less than 0.1 gyr the sed is similar to the sed of an instantaneous starburst galaxy , which we consider in the next model . compared with a 1 gyr old constant sfr galaxy sed model , this model lacks hard uv photons because of the lack of agb stars . this is shown in figure 3 as model c. we adopt the parameterization of the milky way radiation field by black ( 1987 ) . this includes starlight , cosmic background radiation , thermal emission of dust , hot interstellar gas heated by supernova remnants , stellar winds , and uv radiation from extragalactic sources such as qsos and nearby active galaxies . the energy distribution between 13.6 ev and 54 ev is a power law extrapolated from other wavelengths . this is shown in figure 3 as model d. the sed is a single power - law from radio to x - ray wavelengths . the energy distribution between the lyman limit and 1100 is harder than the milky way sed , the constant sfr galaxy sed and the starburst galaxy sed , but is much softer than the quasar seds . this is shown in figure 3 as model e. figure 4 shows the cloudy results for low ionization species . the ordinate is the column density of the various species , and the abscissa is the ionization parameter , u = @xmath57(h)/n(h)c , where @xmath57(h ) is the surface flux of hydrogen - ionizing photons , n(h ) is the total hydrogen number density ( ionized , neutral , and molecular hydrogen ) , and c is the speed of light . cloudy shows that the column densities of o i , s ii , fe ii , si ii and ni ii , which are the dominant ions of those species in the h i dominant regions , are not sensitive to different sed models over a large range of ionization parameter . since all sed models give similar results for these ions , we only show the result from model e in figure 4 . the abundance of these ions from cloudy are consistent with the observed ones within about a factor of two indicating that we have a reasonable set of assumptions for the abundance and depletions . figure 5 shows the calculated and observed ratios of n(h@xmath3)/n(c i ) in the a@xmath42 component of z = 2.8115 damped system . the common trend in the figure of n(h@xmath3)/n(c i ) vs. n(c i ) can be explained by molecular hydrogen chemistry . the main destruction process of h@xmath3 , photodissociation through the process of spontaneous radiative dissociation by the lyman and werner bands , is initiated by discrete line absorptions . the lines are very narrow . thus , the lines rapidly become optically thick , and then h@xmath3 shields itself effectively , so that most of the hydrogen forms h@xmath3 ( e.g. van dishoeck 1990 ) . this is why we see a rapid increase of n(h@xmath3)/n(c i ) vs. n(c i ) when the uv flux drops below a certain threshold . after that , the abundance of c i still increases as the uv decreases , but the abundance of h@xmath3 increases slowly due to lack of formation material , hydrogen , which has already been changed into h@xmath3 . this is why the ratio of n(h@xmath3)/n(c i ) decreases after the rapid increase phase . the best fit to the observed ratios is the starburst model ( b ) and ( c ) . model ( d ) and ( e ) can fit the observed ratios within about 3@xmath22 . the quasar sed models ( model a(1 ) and a(2 ) ) can not explain the observed ratios . the observed ratios are best explained by the soft uv energy distribution models , which suggest stars contribute most of the ionizing radiation field in the a@xmath42 component of z = 2.8115 damped ly@xmath0 system . thus , the sed model where the uv flux is mainly contributed by pks 0528 - 250 can be ruled out , i.e. the distance between the a@xmath42 component and the quasar is probably large enough ( @xmath58 1 mpc ) that the uv flux from the quasar does not dominate the uv background radiation field in the a@xmath42 component . figure 6 shows the cloudy results for the high ionization species assuming the total neutral hydrogen column density used in figure 4 . none of the model seds can explain the c iv , n v , si iv , o vi and si iii column densities . hence the absorption lines for these high ionization species are not coming from the same material as the absorption for the low ionization species . however , assuming the equilibrium between collisional ionization and radiative and dielectronic recombination , we find that t @xmath59 k can explain the relative ratios of these species ( shull & van steenberg 1982 ) . the a@xmath42 component may be a combination of a disk - like system and a halo , or warm and cold h i clouds associated with a hot intercloud medium . using the cloudy model , we can further estimate the average value of the uv radiation field along the sightline to the quasar in the a@xmath42 component . in order to calculate this , we need to know the number density of neutral hydrogen . we estimate this density using the relative population ratio of the excited state si ii@xmath60 @xmath18 1533 to the ground state si ii @xmath18 1526 . in the h i dominant region with density less than @xmath61 @xmath2 , this population ratio can be expressed as @xmath62 where @xmath63 , @xmath64 , @xmath65 and @xmath66 are , respectively , the electron , proton , hydrogen - atom and molecular hydrogen number densities , @xmath67 is the collision rate , and @xmath68 is the spontaneous transition probability ( bahcall & wolf 1968 ) . cloudy shows that the relative electron density @xmath69 is about @xmath70 at the outside to @xmath71 at the center of the cloud , and the equilibrium temperature is from @xmath72 k in the outside to @xmath73 k in the center . we therefore can roughly neglect the contributions of the fine structure excitation by the proton and molecular hydrogen collisions ( bahcall & wolf 1968 ) . then , @xmath74 where @xmath75 @xmath76 ( bahcall & wolf 1968 ) , and @xmath77 @xmath2 s@xmath12 ( dufton & kingston 1991 ) . taking the observed ratio @xmath78 , t @xmath79 k and @xmath80 , we get @xmath81 @xmath2 . cloudy gives a best fit log u @xmath822.9 and @xmath542.6 for the model b and c , respectively . thus , the upper limit of the neutral hydrogen density implies an upper limit of 1 - 2@xmath83 @xmath5 s@xmath12 for the hydrogen - ionizing photon flux in the a@xmath42 component . this upper limit is about 100 times larger than the milky way value of @xmath84 @xmath5s@xmath12 ( black 1987 ) . because si ii@xmath60 traces warm phase regions outside of the cloud ( cf . morton 1975 ) , the above result is more suitable for outside region of the cloud . for the central part of the absorber cloud , the population ratio of the h@xmath3 j = 4 to j = 0 levels , n(j=4)/n(j=0 ) = 4.5@xmath85 ( songaila & cowie , 1995 ) , provides @xmath86 s@xmath12 , where @xmath87 is the h@xmath3 formation rate , n = n(h ) + 2n(h@xmath3 ) ( see jura 1975 for details ) . we scaled the h@xmath3 formation rate to r @xmath88 @xmath89s@xmath12 , assuming the dust - to - gas ratio is 10% that of the milky way . the density is then @xmath90 20 @xmath2 . we therefore can use this value and other measurements to estimate the physical properties in the central part of the cloud and compare them with the cloudy results . in the central cloud where the kinetic temperature is around 100 k ( songaila & cowie 1995 ) , c ii provides most of the electrons ( e.g. morton 1975 ) , so @xmath91 where we have assumed a spatially homogeneous distribution of c ii ions , electrons and hydrogen atoms . thus , @xmath92 @xmath2 . the equation for c i photoionization equilibrium can be expressed as @xmath93 where we have also assumed a homogeneous distribution of c i ; the photoionization rate , @xmath94 ( s@xmath12 ) , is a function of the radiation field intensity , i ; and @xmath95 ( @xmath96 s@xmath12 ) is the radiative recombination rate coefficient , which is a function of kinetic temperature . the kinetic temperature in the h@xmath3 containing cloud is between 74 k and 270 k , where 74 k is derived from the population ratio of the h@xmath3 in the first excited rotational state to the ground rotational state and 270 k is derived from the b = 1.5 km s@xmath12 for the h@xmath3 absorption lines ( songaila & cowie 1995 ) . therefore we choose @xmath97 @xmath89s@xmath12 ( e.g. snow 1977 ) . from the measurements , we derive n(c i)/n(c ii)@xmath98 . putting these together , we get @xmath99 s@xmath12 which is a factor of few larger than typical value of the @xmath100 s@xmath12 in the milky way ism ( de boer _ et al . this value is consistent with what we expect for the outer region of the cloud . thus , all results derived here are consistent with those derived from the cloudy analysis . we can further use these estimates of the physical parameters to estimate the size of the z = 2.8108 cloud . the averaged value for the electron density is derived to be @xmath101 @xmath2 , n(c ii)@xmath47 n(e)@xmath102 , thus , the physical size for the z = 2.8108 cloud , is @xmath103 pc . for the a@xmath3 component , the lack of highly ionized ions may mean that the a@xmath3 component is not associated with any halo gas . the photoionization models described above can not explain the lack of h@xmath3 and c i , along with the possible detection of co , and special circumstances for molecule formation may apply . in summary , we can account for the z = 2.8108 absorption line system ( a@xmath42 ) containing both hot and cold , neutral gas . the cloud medium contributes most absorption lines of low ionization species , while the hot medium contributes most absorption lines of highly ionized species . this structure is similar to that of the interstellar diffuse clouds in the milky way . the uv flux intensity in this absorber is a few times that of the typical value in the milky way ism . pettini _ et al . _ ( 1994 ) have surveyed about a dozen damped ly@xmath0 systems to measure the abundance of zn and cr and have claimed that the metallicity of the damped ly@xmath0 systems at z @xmath1 2 is about 1/10 of the solar value . they used the relative abundance [ cr / zn ] to deduce the dust - to - gas ratio in the damped systems , which is about 1/10 of the milky way . this result is consistent with that from the dust reddening measurements of quasar spectrum slope by pei _ however , lu _ et al . _ ( 1996 ) recently claimed that the overabundance of zn relative to cr may be intrinsic to the stellar nucleosynthesis in these absorbers instead of dust depletion ( however , see pettini _ et al . _ 1996 ; smith _ et al . _ 1996 ) . here , we assume that the underabundance of cr and ni relative to zn is caused by dust depletion . compared with cr , ni is perhaps a better element to use to deduce depletion . like zn and cr , ni is an iron group element and it traces fe to about [ fe / h ] @xmath104 or even lower ( wheeler _ et al . _ 1989 ; ryan _ et al . ni is also more heavily depleted than cr in the milky way ( jenkins , 1987 ) : 0.35% ni remains in the gas phase compared to 0.62% for cr in the diffuse cloud toward @xmath105 per ( cardelli _ et al . _ 1991 ) . because the ionization potential of ni i is 7.635 ev which is smaller than that of h i , but the ionization potential of ni ii is 18.168 ev which is larger than that of h i , ni ii is the dominant species of ni in the neutral hydrogen dominant region . consequently , the ratio of n(ni ii)/n(h i ) can reflect that of ni / h gas phase abundance without the need to account for unobserved ion stages . for example , under the condition that best fits the ionization structure of the a@xmath42 component in the z = 2.8115 damped ly@xmath0 system , model ( b ) , log u = @xmath106 and n(h)=1 @xmath2 , n(ni ii)/n(ni ) = 97.7% . moreover , there are more uv absorption lines of ni ii than cr ii . if we can observe more than two weak absorption lines of ni ii , we can use a curve - of - growth analysis to get a more accurate estimate of the column density . another potential advantage of using ni ii instead of cr ii is that the ni ii transitions are observable with ground - base telescopes until redshifts about 5 without strong sky emission because the strong uv transitions of ni have rest wavelengths between 1400 and 1700 compared to @xmath1 2000 for cr and zn . our measured equivalent widths of ni ii @xmath18 1709 , ni ii @xmath18 1741 and ni ii @xmath18 1751 in the z@xmath26 system correspond to column densities 3.4@xmath107 @xmath5 , 3.8@xmath108 @xmath5 and 3.2@xmath109 @xmath5 , respectively , when we adopt a doppler parameter b = 30 km s@xmath12 ( meyer _ et al . they are consistent with the values obtained by meyer _ the average column density of ni ii is 3.5@xmath110 @xmath5 . the h i column density in this system is 5@xmath111 @xmath5 ( morton _ et al . thus , ni / h is 7.0@xmath112 , implying that ni is depleted by no more than a factor of 27@xmath1134 in the z@xmath26 damped ly@xmath0 system with respect to the solar ratio of 1.9@xmath114 ( withbroe 1971 ) or [ ni / h ] @xmath115 . the upper limit of n(zn ii ) , 2.6@xmath116 @xmath117 , measured by meyer _ ( 1989 ) indicates that the metallicity in the 2.1408 system is @xmath118 10% of solar abundance . the measured residual depletion in the z = 2.1408 damped ly@xmath0 system is [ ni / zn]@xmath119 . for comparison , [ ni / zn ] @xmath120 in the diffuse cloud toward @xmath105 per of the milky way which has n(h ) @xmath121 @xmath5 ( cardelli _ et al . this is the typical value in the milky way ( c.f . jenkins 1987 ) . the limit for [ ni / zn ] implies that dust grains in the z = 2.1408 damped ly@xmath0 system contain about 70% of the total ni , while dust in our galaxy contain about 99% of the ni . thus , the dust - to - gas ratio is @xmath1187% of that in the milky way . the low dust content implies an inefficient molecular hydrogen formation rate , which suggests low total molecular mass . hence , if the measured dust - to - gas ratio is typical of the interstellar medium in the z = 2.1408 damped ly@xmath0 system , our result is consistent with nondetection of co emission by wiklind & combes ( 1994 ) , who obtained an upper limit for the total molecular mass of @xmath122 m@xmath14 integrating over 1100 km s@xmath12 , but not consistent with the huge amount of molecular mass , m@xmath1237@xmath124 m@xmath14 , reported by brown & vanden bout ( 1993 ) . the column densities for the newly identified ni ii @xmath18 1370 , ni ii @xmath18 1454 and ni ii @xmath18 1467 lines in the z = 2.8115 damped ly@xmath0 system are shown in table 2 . the average column density of ni ii is 6.5@xmath125 @xmath5 in the a@xmath42 component and 3.3@xmath126 @xmath5 in the a@xmath3 component . together , we get 9.8@xmath127 @xmath5 for the two components which is consistent with the column density obtained by meyer _ et al . _ ( 1989 ) obtained by measuring the equivalent widths of ni ii @xmath18 1709 and @xmath18 1741 . the h i column density for the 2 velocity components together in the damped system is 2.24@xmath128 @xmath5 ( mller & warren 1993 ) . thus , ni / h is 4.4@xmath112 , so that ni is depleted by no more than a factor of 43@xmath129 with respect to the solar value or [ ni / h ] @xmath130 . the depletion of zn is [ zn / h ] @xmath131 ( meyer _ et al . _ 1989 ) , so the metallicity is about 10% of the solar value . the residual depletion of ni relative to zn , [ ni / zn ] @xmath132 which means about 80% of the ni is contained in dust grains . consequently , the dust - to - gas ratio is about 8% of that of the galaxy if dust grains cause the depletion of ni relative to zn . in figure 7 , we show our spectra which cover the five ultraviolet co bands ( a - x 0 - 0 , 1 - 0 , 2 - 0 , 3 - 0 , 4 - 0 ) . we mark the central positions of the co lines of the two velocity components ( z = 2.8108 , 2.8132 ) in each spectrum . the co 3 - 0 @xmath18 1447 line exhibits absorption at 5887.47 at the expected position of z = 2.8132 within the uncertainties . if this identification is correct , the co column density is 4.7@xmath133 @xmath5 assuming the line is unsaturated . however , the absorption line at 5887.47 could also be the co ii 1448 line at z = 2.8117 . the corresponding column density is @xmath134 @xmath5 , which is consistent with the upper limit obtained by songaila & cowie ( 1995 ) . in order to improve the co column density limit in the a@xmath42 component , we constructed a composite co spectra from the 5 co lines using a method similar to that employed by levshakov _ ( 1989 ) . to obtain the composite spectrum , we first shifted the spectral regions containing the co absorption bands to rest frame 1477.52 the wavelength of the co a - x 2 - 0 band , which has the largest oscillator strength in the co a - x system ( field _ et al . _ 1983 ) , then normalized to the continuum , and averaged with each pixel weighted by the inverse of its variance . in order to avoid the possible effect of telluric absorption at @xmath18 = 5892.08 and z = 2.8108 c iv absorption at 5900.16 we used the fitted continuum value instead of the observed flux in our coaddition . the final composite spectrum is shown in figure 7 . no absorption of co from the a@xmath42 component is found . the 3@xmath22 upper limit to the equivalent width in the rest frame is 17 m . using the weighted mean of the oscillator strengths of the considered bands , we derive an upper limit on the co column density of the a@xmath42 component of n(co)@xmath135 @xmath5 . thus the ratio n(co)/n(h i)@xmath136 for the a@xmath42 component , provided that the total neutral hydrogen absorption in the z = 2.8115 damped ly@xmath0 system is from this component . figure 2 shows the summed spectrum of co ( j=3@xmath1372 ) from the damped ly@xmath0 system z = 2.8115 toward pks 0528 - 250 . we have not detected any emission from the z = 2.8108 and z = 2.8132 components . the 3@xmath22 upper limit of the observed integrated line intensity i@xmath138 k km s@xmath12 for the a@xmath42 component , and 0.142 k km s@xmath12 for the z = 2.8132 component , where @xmath139 = 100 km s@xmath12 is the integrated velocity interval around the absorption components , and @xmath140 is the number of channels in the velocity interval of 100 km s@xmath12 . the co luminosity for the z = 2.8115 damped ly@xmath0 system can be expressed , @xmath141 where @xmath142 is the brightness temperature of the source , @xmath143 is the solid angle subtended by the source , and @xmath144 is the angular size distance . @xmath146 is the luminosity distance , @xmath147\}.\ ] ] the observed temperature of the source @xmath148 is defined as @xmath149 @xmath150 is the normalized power pattern , and the factor ( 1+z ) is from the expansion of the universe . moreover , the angular size of about 2.5@xmath21 for the ly@xmath0 emission region in the z = 2.8115 damped ly@xmath0 system ( mller & warren 1993 ) is much smaller than the beam size of the telescope of 68@xmath21 , so if the co emission region has similar size as the ly@xmath0 emission region , then t@xmath151 can be further expressed as @xmath152 where @xmath153 is the telescope beam solid angle . then the co line luminosity can be expressed as @xmath154 the molecular mass can be written as @xmath155 where @xmath0 is the co to h@xmath3 conversion factor . adopting the standard conversion factor , @xmath156 ( sanders _ et al . _ 1991 ) , we obtain a 3@xmath22 upper limit for the total molecular mass of @xmath157 in the a@xmath42 component , and 1.88@xmath158 in the a@xmath3 component . these upper limits are about a factor of two better than previous observations of wiklind & combes ( 1994 ) applying the same model assumptions to their t@xmath159 . however , the adopted standard conversion factor could introduce some uncertainty in the total molecular mass derived here . it can be easily shown that the conversion factor @xmath0 for gravitationally bound ( virialized ) clouds , @xmath160 . hence , the total molecular mass in the z = 2.8115 damped ly@xmath0 system could be overestimated if the clouds in it have a higher brightness temperature than the milky way clouds , whereas the total mass could be underestimated if the clouds have higher density . furthermore , there is a weak dependence of @xmath0 on the co abundance , @xmath161^{-1/4}$ ] ( radford _ et al . if the co abundance in the z = 2.8115 damped ly@xmath0 system is considerably lower than the milky way molecular clouds , the abundance effect will lead to an underestimate of the total amount of the molecular gas . from these considerations , the uncertainty in estimating the molecular mass for the z = 2.8115 damped ly@xmath0 system is a factor of a few . for comparison , the total molecular mass in the milky way is 2@xmath162 ( e.g. sanders _ 1991 ) , the total molecular mass in the ultraluminous ir galaxy arp 220 is @xmath163 ( e.g. solomon _ et al . _ 1990 ) , and the total molecular mass of the hyperluminous ir galaxy f10214 + 4724 at z = 2.286 , is @xmath164 ( e.g. downes et al . thus , the derived upper limit of the molecular mass in the z = 2.8115 damped ly@xmath0 system is higher than any ultraluminous ir galaxy . on the other hand , our optical observations suggest that this damped ly@xmath0 system is an analog of a milky way - like normal galaxy at high redshift . in order to detect the molecular mass in this system if it is similar to the milky way , an rms / channel of @xmath110@xmath165 mk is needed , using the same telescope size , system temperature and bandpass . this requires an integration time 10,000 times longer than obtained here . however , interferometers could be used to reach interesting detection levels ( e.g. omont _ et al . _ 1996 ; ohta _ et al _ 1996 ) . our main results are the following . the ionization of different species in the a@xmath42 component of the z = 2.8115 damped ly@xmath0 system rules out the possibility that the quasar uv flux dominates the uv ionizing field in this component , i.e. the quasar should be more than @xmath166 mpc away from the damped ly@xmath0 system . the ionization further suggests that the shape of the uv background radiation field in this component is similar to the milky way . the uv field intensity is a few times that of the milky way value which further implies that the sfr in this component is probably similar to the sfr of the milky way if the total mass in this component is similar to the milky way . this result is consistent with the results from ly@xmath0 observations ( mller & warren 1993 ) . the physical size for this component is about 40 pc . the ionization of different species suggests a structure with a hot intercloud medium associated with a h i cloud in this component , i.e. most lowly ionized ions are from the cold medium where photoionization and photodissociation dominates . most highly ionized ions are from the intercloud medium where collisional ionization dominates . we have not detected co absorption in the rest uv spectrum from the a@xmath42 component of the z = 2.8115 damped system . the 3@xmath22 upper limit of co column density is n(co)@xmath167 @xmath5 for this component . the ratios of the n(co)/n(h i ) and n(co)/n(h@xmath3 ) are in the range of the values for the milky way diffuse clouds ( federman _ et al . _ 1980 ) . the absorption line at 5519.40 could be a co absorption line from the a@xmath3 component ; however , it also could be a co ii absorption line from a@xmath42 component . higher resolution and higher signal - to - noise observations are needed to confirm the identification of this line . analysis of photoionization and photodissociation of h@xmath3 and c i suggests that the ratio n(h@xmath3)/n(c i ) is a good indicator of the shape of the radiation field in the ism of damped ly@xmath0 absorbers . newly identified ni ii absorption lines show that the dust - to - gas ratios in the z = 2.1408 and z = 2.8115 damped ly@xmath0 systems are about 10% of that of the milky way , which are consistent with previous results . we have not detected co ( 3 - 2 ) mm emission from the z = 2.8115 damped ly@xmath0 system . the 3@xmath22 upper limits for the mass of h@xmath3 in the a@xmath42 and a@xmath3 components in the z = 2.8115 damped system are @xmath157 and @xmath168 , respectively . the results still can not rule out the possibility that the z = 2.8115 damped ly@xmath0 system is similar in molecular mass to ultraluminous ir galaxies . we are grateful to dr . a. songaila cowie for providing important comments and data in advance of publication . we thank dr . j. shields for helpful discussions . we thank dr . g. ferland for providing his cloudy program . we thank dr . g. bruzual and dr . s. charlot for providing their galaxy isochrone synthesis spectral evolution library ( gissel ) . we also wish to thank the staffs of mmto and nrao for all of their help . this research was supported by nsf ast-9058510 and nasa grant nagw-2201 . bahcall , j. n. & wolf , r. a. 1968 , apj , 152 , 701 bechtold , j. 1994 , apjs , 91 , 1 bergeson , s.d . , & lawler , j.e . 1993 , apj , 414 , l137 black , j. h. 1987 , in interstellar processes ( reidel publishing company ) , eds . hollenbach , d. j. & thronson , jr . h. a. 731 brown , r. l & vanden bout , p. a. 1993 , apj , 412 , l12 bruzual , g. a. & charlot , s. 1993 , apj , 405 , 538 cardelli j. a. , savage , b. d , bruhweiler , f. c. , smith , a. m. , ebbets , d. c. , sembach , k. r. , & sofia , u. j. 1991 , apj , 337 , l57 chen , j. & morton , d. c. 1984 , mnras , 208 , 167 de boer , k. s. , koppenaal , k. , & pottasch , s. r. , 1973 , a&a , 28 , 145 downes , d. , solomon , p. m. , & radford , s. j. e. , 1995 , apj , 453 , l65 dufton , p. l. & kingston , a. e. 1991 , mnras , 248 , 827 federman , s. r. , glassgold , a. e. , jenkins , e. b. & shaya , e. j. 1980 , apj . 242 , 545 ferland , g. j. 1993 , univ . of kentucky , department of physics and astronomy internal report . field , r. w. , benoist dazy , o. , lavollee , m. , lopez - delgado , r. & tramer , a. 1983 , j. chem . 78 , 2838 foltz , d. b. , chaffee , f. h. & black , j. h. 1988 , apj , 324 , 267 ge , j. , bechtold , j. , & black , j.h . 1997 , apj , 474 , 67 ge , j. , & bechtold , j. 1997 , apj , 477 , l73 jenkins , e. b. 1986 , apj , 304 , 739 jenkins , e. b. 1987 , in interstellar processes ( reidel publishing company ) , eds . hollenbach , d. j. & thronson , jr . h. a. 533 lanzetta , k. m. 1993 , in the environment and evolution of galaxies ( kluwer academic publishers ) , ed . shull , j. m. & thronson , jr . h. a. 237 levshakov , s. a. , foltz , c. b. , chaffee , f. h. & black , j. h. 1989 , aj , 98 , 2052 levshakov , s. a. , chaffee , f. h. , foltz , c. b. & black , j. h. 1992 , a&a , 262 , 385 lu , l. , sargent , w.l.w . , & barlow , t.a . , 1996 , apjs , in press meyer , d. m. , black , j. h. , chaffee , f. h. , foltz , c. b. & york , d. g. 1986 , apj , 308 , l37 meyer , d. m. & york , d. g. 1987 , apj , 319 , l45 meyer , d. m. , welty , d. e. & york , d. g. 1989 , apj , 343 , l37 mller , p. & warren , s. j. 1993 , a&a , 270 , 43 morton , d. c. , 1975 , apj , 197 , 85 morton , d. c. , chen , j. , wright , a. e. , peterson , b. a. , jauncey , d. l. 1980 , mnras , 193 , 399 morton , d. c. , york , d. g. & jenkins , e. b. 1988 , apjs , 68 . 449 morton , d. c. , 1991 , apjs , 77 , 119 ohta , k. , yamada , t. kohno , k. , akiyama , m. , & kawabe , r. 1996 , nature , 382 , 426 pei , y.c . , fall , s.m . , & bechtold , j. 1991 , apj , 378 , 6 pettini , m. , smith , l. j. , hunstead , r. w. & king , d. l. 1994 , apj , 426 , 79 pettini , m. , king , d.l . , smith , l.j . , & hunstead , r. , 1996 , preprint radford , s. j. e. , solomon , p. m. & downes , d. 1991 , apj , 368 , l15 ryan , s. g. , norris , j. e. & bessell , m. s. 1991 , aj , 102 , 303 sanders , d. b. scoville , n. z. & soifer , b. t. 1991 , apj , 370 , 158 sargent , w. l. w. , steidel , c. c. & boksenberg , a. apjs , 1989 , 69 , 703 savage , b. d , bohlin , r. c. , drake , j. f. & budich , w. 1977 , apj , 216 , 291 schmidt , g. d. , weymann , r. j. & foltz , c. b. 1989 , pasp , 101 , 713 sheffer , y. , federman , s. r. , lambert , d. l. & cardelli , j. a. 1992 , apj , 397 , 482 shull , j. m. & van steenberg , m. 1982 , apjs , 48 , 95 smith , l.j . , pettini , m. , king , d.l . , & hunstead , r.w . , 1996 , preprint snow , t. p. , 1977 , apj , 216 , 724 solomon , p. m. radford , s. j. e. & downes , d. 1990 , ap j , 348 , l53 songaila , a. & cowie , l. l. 1995 , aj submitted van dishoeck , e. f. 1990 , in molecular astrophysics ( cambridge university press ) , ed . hartquist , t. w. 55 walker , c. e. _ et al . _ 1996 , in preparation wheeler , j. c , sneden , c. & truran , jr . , j. w. 1989 , ara&a , 27 , 279 wiklind , t. & combes , f. 1994 , a&a , 288 , 41 wilkes , b. _ et al . _ 1994 , apjs , 92 , 53 withbroe , g. l. 1971 , in the mengal symposium ( nbs spec . pub . 353 ) , ed . gebbie , k. b. ( washington : gpo ) , 127 wolfe , a. m. , turnshek , d. a. , smith , h. e. , & cohen , r. d. 1986 , apjs , 61 , 249 young , p. j. , sargent , w. l. w. , & boksenberg , a. 1982 , apj , 252 , 10 figure 1.the spectrum of pks 0528 - 250 obtained with the mmt red channel spectrograph . the lower curve presents the 1 @xmath22 error as derived from count statistics in the object and night sky spectra . the features identified at better 3@xmath22 significance in table 1 are indicated by tick marks . figure 2.the summed co(3@xmath1372 ) millimeter wavelength spectrum of pks 0528 - 250 observed with nrao 12 meter telescope . the intensity scale is main - beam brightness temperature ( k ) and the velocity scale is the offset ( km s@xmath12 ) from the optical redshift , z = 2.81086 . the two velocity components are marked . a fit to the baseline has been subtracted . the upper line presents the 1 @xmath22 error derived from all scans after removing bad scans . figure 3.spectral energy distributions adopted for photoionization calculations . a(1 ) and a(2 ) are quasar seds . 1 gyr old constant sfr galaxy sed . 0.001 gyr old instantaneous starburst galaxy sed . ( d ) . milky way sed . power - law sed with @xmath169 . ionization potentials for several ions are marked . figure 4.ionization models for the h i dominant region with n(h i ) = @xmath50 @xmath5 . the ordinate is the column density of various ions , the abscissa is the log of the ionization parameter u. the figure shows results of dominant ions , o i , fe ii , si ii , s ii and ni ii in the model e power - law sed model with @xmath169 . solid lines are from cloudy and dotted lines are the observed values . figure 5. the relative ratios of n(h@xmath3)/n(c i ) vs. n(c i ) . the results are shown for sed models described in figure 3 . the vertical lines show the upper limit ( 3@xmath22 ) of n(c i ) . the horizontal lines show the lower limit ( 3@xmath22 ) of the ratio of n(h@xmath3)/n(c i ) . figure 6. model predictions for the highly ionized species , o vi , n v , c iv , si iv , si iii for the different sed models described in figure 3 . solid lines are results from cloudy . dotted lines are the observed values . the cloudy predictions for o vi and n v are out of the range for the sed model ( c ) . o vi is out of the range for the sed model ( b ) . figure 7.spectral regions including the co a - x 0 - 0 , 1 - 0 , 2 - 0 , 3 - 0 , 4 - 0 absorption bands from the z = 2.8108 and 2.8132 components in the rest frame of the z = 2.8108 absorber toward pks 0528 - 250 . the positions of the absorption bands are indicated by tick marks . the lowest right panel is a composite spectrum obtained by combining the 2 - 0 band with the 0 - 0 , 1 - 0 , 3 - 0 , 4 - 0 bands which are shifted to the wavelength of the 2 - 0 band . species & @xmath18 & f@xmath171 & w@xmath172 & n(@xmath5)@xmath173 & n(@xmath5)@xmath174 & w@xmath172 & n(@xmath5)@xmath175 + al ii ( a@xmath42+a@xmath3)@xmath176 & 1670.788 & 1.88&2.100&@xmath177&&2.100&@xmath178 + c i & 1560.310 & 0.0822 & & @xmath179@xmath180 & & @xmath1810.032 & @xmath1811.8@xmath182 + c ii ( a@xmath42+a@xmath3+a@xmath183)@xmath176 & 1334.532 & 0.118 & 2.29 & 1.7@xmath184 & & 2.29 & 1.7@xmath184 + c ii@xmath60 & 1335.702 & 0.118 & & & & 0.312 & 3.6@xmath185 + c iii ( a@xmath42+a@xmath3)@xmath176 & 977.020 & 0.768 & 1.89 & 7.6@xmath6 & & 1.89 & 7.6@xmath6 + c iv & 1548.202 & 0.194&1.28 & 5.6@xmath186 & 1.3@xmath186 & & + c iv&1550.774&0.0970 & 0.976&2.0@xmath186 & 1.1@xmath186 & & + si i@xmath187 & 1845.520 & 0.229 & @xmath1810.105&@xmath1811.6@xmath188&&@xmath189&@xmath1811.6@xmath182 + si ii & 1808.013 & 0.00208 & 0.397&8.5@xmath186&&0.094&1.7@xmath186 + si ii@xmath60@xmath187 & 1533.431 & 0.13 & @xmath1810.031 & @xmath1811.2@xmath182 & & & + si iii & 1206.500 & 1.66 & 1.14&1.9@xmath186&3.6@xmath185 & & + si iv&1393.755 & 0.528 & 0.995 & 7.9@xmath185&3.0@xmath185 & & + si iv&1402.770 & 0.262 & 0.748 & 4.6@xmath185&3.0@xmath185 & & + s i@xmath187&1425.030&0.181&@xmath1810.031&@xmath1819.5@xmath116&&@xmath1810.031&@xmath1819.5@xmath190 + s ii&1250.583 & 0.00535 & 0.157 & 2.7@xmath186 & & 0.131 & 2.6@xmath191 + & 1253.808 & 0.0107 & 0.315 & ... & & 0.236 & ... + s iii ( a@xmath42+a@xmath3)@xmath176&1012.504 & 0.0355 & 0.622 & 3.2@xmath186 & & 0.622 & 3.2@xmath186 + o i&1302.169 & 0.0486 & 1.36 & 2.1@xmath184 & & 0.70 & 1.1@xmath192 + o vi & 1031.927 & 0.13 & 0.97 & 4.0@xmath6&5.2@xmath186 & & + n v & 1238.821 & 0.157&0.247 & @xmath193 & @xmath194 & & + fe ii&2249.879 & 0.0018 & 0.160 & 1.8@xmath186 & & 0.046 & 3.6@xmath195 + & 2260.781 & 0.0028 & 0.260 & ... & & 0.046 & ... + & 2374.461&0.0395&1.201& ... &&0.342 & ... + ni ii & 1741.549&0.068 & 0.116 & 6.9@xmath182&&0.072 & 3.8@xmath182 + & 1709.600 & 0.047&0.101 & ... & & 0.056 & ... + & 1370.132 & 0.100 & 0.103 & ... & & 0.062 & ... + & 1454.842 & 0.0515 & 0.065 & ... & & 0.027 & ... + ni ii ( a@xmath42+a@xmath3)@xmath176 & 1467.756 & 0.0149 & 0.038 & 1.3@xmath185 & & 0.038 & 1.3@xmath185 + co ii ( ? ) & 1448.011 & 0.04516 & 0.032 & 3.9@xmath182 & & & + co a - x ( ? ) & 1447.359 & 0.0360 & & & & 0.032 & 6.4@xmath182 + co ( composite)@xmath187&&&@xmath1810.017&@xmath196 & & & + h@xmath3 & & & & @xmath197 & & & +

we give an explicit demonstration that the derivative expansion of the qed effective action is a divergent but borel summable asymptotic series , for a particular inhomogeneous background magnetic field . a duality transformation gives a non - borel - summable perturbative series for a time dependent background electric field , and borel dispersion relations yield the non - perturbative imaginary part of the effective action , which determines the pair production probability . resummations of leading borel approximations exponentiate to give perturbative corrections to the exponents in the non - perturbative pair production rates . comparison with a wkb analysis suggests that these divergence properties are general features of derivative expansions and effective actions . = 8.5 in .5 cm

hep-th9902064

the effective action plays a central role in quantum field theory . here we consider the one - loop effective action in quantum electrodynamics ( qed ) for electrons in the presence of a background electromagnetic field : @xmath1 where @xmath2 , and @xmath3 is a fixed classical gauge potential with field strength tensor @xmath4 . when the background is a static magnetic field , the effective action @xmath5 is equal to minus the effective energy of the electrons in that background ; and when the background is an electric field , @xmath5 has an imaginary part which determines the pair - production rate for electron - positron pair creation @xcite . for a uniform background field strength @xmath6 , the effective action @xmath5 can be computed exactly @xcite . for more general backgrounds , with @xmath7 _ not _ constant , the situation is more complicated . one standard approach is to make a `` derivative expansion '' @xcite ( or `` gradient expansion '' @xcite ) which is a formal perturbative expansion in increasing numbers of derivatives of @xmath7 : @xmath8+s^{(2)}[f,(\partial f)^2]+\dots \label{derivative}\ ] ] in this paper we address two questions concerning the qed effective action @xmath5 in a nonuniform background . first , we consider the convergence / divergence properties of the perturbative derivative expansion in ( [ derivative ] ) . second , we ask how such a perturbative expansion can lead to corrections to schwinger s nonperturbative pair - production rate ( computed for a constant background ) when there are inhomogeneities in the background electric field . we can answer these questions precisely by considering some exactly solvable cases with special inhomogeneous backgrounds @xcite . the derivative expansion is found to be a divergent asymptotic series , and the rate of divergence at high orders can be used to compute the corresponding non - perturbative imaginary part of the effective action when the background is a time - dependent electric field . this divergence of the derivative expansion is not a bad thing ; it is completely analogous to generic behavior that is well known in perturbation theory in both quantum field theory and quantum mechanics . for example , dyson @xcite argued that qed perturbation theory can not be convergent as an expansion in the fine structure constant @xmath9 , because the theory is unstable when @xmath9 is negative . ( the divergent and asymptotic nature of field theoretic perturbation theory was also found in scalar @xmath10 theories @xcite ) . in fact , the euler - heisenberg - schwinger qed effective action in a constant background , and the qed effective action in the inhomogeneous backgrounds discussed in this paper , provide explicit demonstrations of dyson s argument . our analysis of the convergence / divergence properties of the qed effective action uses borel summation @xcite , a mathematical tool that can be used to relate the rate of divergence of high orders of perturbation theory to non - perturbative decay and tunneling rates , thereby providing a bridge between perturbative and non - perturbative physics . other well known explicit cases of this connection appear in quantum mechanical examples such as the anharmonic oscillator @xcite and the stark effect @xcite , and in quantum field theory in semi - classical analyses of scalar field theories @xcite and asymptotic estimates of large orders of qed perturbation theory @xcite . for an excellent review of a broad range of examples , see ref . typically one finds that in a stable situation ( _ i.e. _ , no tunneling or decay processes ) perturbation theory is divergent , with expansion coefficients that alternate in sign and grow factorially in magnitude . on the other hand , in an unstable situation , perturbation theory is generally divergent with coefficients that grow factorially in magnitude but do not alternate in sign . borel summation is an approach to the summation of divergent series that makes physical sense out of these two different types of behavior . we shall see that the divergence of the derivative expansion can be understood naturally in this borel framework . in addition to these theoretical considerations of understanding the connections between the perturbative derivative expansion and non - perturbative pair - production rates , another motivation for this work is provided by the attempt to observe electron - positron pair creation due to qed vacuum effects in the presence of strong electric fields . schwinger s constant field pair - production rate is far too small to be accessible with present electric field strengths . however , the constant field approximation is somewhat unrealistic , and so one can ask how this rate is modified by a time variation of the electric background . for sinusoidal time variation brzin and itzykson found a wkb result with fairly weak frequency dependence @xcite , while balantekin _ et al _ have applied group theory and uniform wkb to electric backgrounds with more general time dependence @xcite . the qed effective action has recently been computed @xcite as a closed form ( single integral ) expression for the particular time - dependent electric background with @xmath11 , and in this paper we consider the numerical implications of this result for pair - production rates in such a background . it is important to note that another related approach to observing pair - production is to use highly relativistic electrons as intermediate states , as has been done in recent experiments @xcite . finally , we note that the derivative expansion ( [ derivative ] ) is an example of an _ effective field theory _ expansion @xcite , such as is used in operator product expansions @xcite and chiral perturbation theory @xcite . in the effective field theory approach , the mass @xmath12 of the electrons ( which are ` integrated out ' in the one - loop approximation ) sets an energy scale , and the physics at energies @xmath13 should be described by a low energy effective action with the formal expansion @xmath14 where @xmath15 is an operator of dimension @xmath16 . for the case of the qed effective action , the simplest way to produce higher - derivative operators in such an expansion is to take higher powers of the field strength @xmath17 . thus , even the leading ( _ i.e. _ constant background ) term @xmath18 $ ] in the derivative expansion ( [ derivative ] ) is itself an effective field theory expansion of the form ( [ eff ] ) . this is simply the euler - heisenberg effective action which is a perturbative series expansion in powers of @xmath19 . this series is known to be divergent @xcite , and there are important physical consequences of this divergence , as we review below . another way to produce higher dimension operators in the expansion ( [ eff ] ) is to include derivatives of @xmath17 , with each derivative balanced by an inverse power of @xmath12 . this is what is done in the derivative expansion ( [ derivative ] ) . thus , we can view the derivative expansion ( [ derivative ] ) as a ` double ' series expansion , both in powers of @xmath17 and in derivatives of @xmath17 . in this paper we study the divergence properties of such an expansion . it has been suggested , based on the behavior of the constant field case @xcite , that the effective field theory expansion ( [ eff ] ) is generically divergent . here we provide an explicit demonstration of this divergence for inhomogeneous background fields . for energies well below the scale set by the fermion mass @xmath12 , the divergent nature of the effective action is not important , as the first few terms provide an accurate approximation . however , the divergence properties do become important when the external energy scale approaches the fermion mass scale @xmath12 , and/or when the inhomogeneity scale becomes short compared to a characteristic scale of the system . the divergence is also important for understanding how the non - perturbative imaginary contributions to the effective action arise from _ real _ perturbation theory . in section ii we review briefly the mathematical technique of borel summation , and in section iii we apply this to the euler - heisenberg - schwinger constant - background effective action . section iv gives the borel summation analysis of the particular exactly solvable cases with inhomogeneous magnetic and electric backgrounds . in section v we show how this is related to a wkb analysis , and section vi contains some concluding remarks . in this section we review briefly the basics of borel summation @xcite . consider an asymptotic series expansion of some function @xmath20 @xmath21 where @xmath22 is a ( small ) dimensionless perturbation expansion parameter and the @xmath23 are real coefficients . in an extremely broad range of physics applications @xcite it has been found that perturbation theory leads not to a convergent series but to a divergent series like ( [ exp ] ) in which the expansion coefficients @xmath23 have large - order behavior of the form @xmath24 \qquad\qquad ( n\to\infty ) \label{general}\end{aligned}\ ] ] for some real constants @xmath9 , @xmath25 , and @xmath26 . when @xmath27 , the perturbative expansion coefficients @xmath23 alternate in sign and their magnitude grows factorially . borel summation is a particularly useful approach for this case of a divergent , but alternating series . we shall see below that non - alternating series must be treated somewhat differently . consider , for example , the asymptotic series ( [ exp ] ) with @xmath28 , and @xmath27 . this series is clearly divergent for any value of the expansion parameter @xmath29 . borel summation of this divergent series can be motivated by the following formal procedure . write @xmath30 and then formally interchange the order of summation and integration , to yield @xmath31 \qquad\qquad ( g\to 0^+ ) \label{borel}\end{aligned}\ ] ] this integral is convergent for all @xmath32 , and so can be used to _ define _ the sum of the divergent series @xmath33 . to be more precise , the formula ( [ borel ] ) should be read from right to left : for @xmath34 , we can use laplace s method @xcite to make an asymptotic expansion of the integral , and we obtain the asymptotic series in ( [ exp ] ) with expansion coefficients @xmath28 . the borel integral ( [ borel ] ) can be analytically continued off the @xmath22 axis and in this case is in fact @xcite an analytic function of @xmath29 in the cut @xmath29 plane : @xmath35 . thus , we can use a simple dispersion relation ( using the discontinuity across the cut along the negative @xmath29 axis ) to _ define _ the imaginary part of @xmath20 for negative values of the expansion parameter : @xmath36 \qquad\qquad ( g\to 0^+ ) \label{imag}\end{aligned}\ ] ] note , of course , that an alternating series with negative @xmath29 is the same as a _ non - alternating _ series with positive @xmath29 . if the expansion coefficients in ( [ exp ] ) are _ non - alternating _ ( with @xmath22 ) then the situation is very different , both physically and mathematically . formal application of borel summation yields @xmath37 \qquad\qquad ( g\to 0^+ ) \label{non}\end{aligned}\ ] ] this borel integral has a pole on the integration contour . however , a principal parts prescription for such a pole gives an imaginary part in agreement with ( [ imag ] ) . this imaginary contribution is non - perturbative ( it clearly does not have an expansion in positive powers of @xmath29 ) and has important physical consequences . generically , the borel technique signals the possible presence of such non - perturbative physics if the perturbative expansion coefficients grow rapidly ( factorially ) in magnitude and are non - alternating . the associated dispersion relations provide a bridge between the perturbative physics [ i.e. the @xmath23 s ] and the non - perturbative imaginary parts [ i.e. the @xmath38 factors ] . we will see explicit examples of this below . we should note at this point that for a general divergent series these borel dispersion relations may be complicated by the appearance of additional poles and/or cuts in the complex @xmath29 plane . physically interesting poles , known as renormalons , are indeed found in certain resummations of perturbation theory for both qed and qcd @xcite . here , for the one - loop qed effective action in a fixed external background we find that we do not encounter such poles . the borel summation construction discussed above for the case @xmath28 , generalizes in the obvious way to the case where the perturbative coefficients are @xmath39 , which corresponds to the leading - order growth indicated in ( [ general ] ) : @xmath40 \qquad\qquad ( g\to 0^+ ) \label{genborel}\end{aligned}\ ] ] the corresponding imaginary part for negative values of the expansion parameter is @xmath41 \qquad\qquad ( g\to 0^+ ) \label{genimag}\end{aligned}\ ] ] notice that the parameter @xmath42 affects the exponent , while the combination @xmath43 is important for the prefactor . these last two formulas , the borel integral ( [ genborel ] ) and the borel dispersion relation ( [ genimag ] ) , will be used repeatedly below . now consider applying this borel summation machinery to qed effective actions . effective actions can be expanded perturbatively in terms of the coupling constant @xmath44 , and also in terms of derivatives of the background field strength @xmath7 . to begin , we review the well - known euler - heisenberg - schwinger effective action which corresponds to a uniform background field strength ; thus the only expansion is in terms of the perturbative coupling constant @xmath44 . we consider first a magnetic background , and then we consider an electric background . for a uniform background _ magnetic _ field of strength @xmath45 , the exact renormalized effective action can be expressed as a ` proper - time ' integral @xcite @xmath46 the @xmath47 term is a subtraction of the zero field ( @xmath48 ) effective action , while the @xmath49 subtraction corresponds to a logarithmically divergent charge renormalization @xcite . the @xmath50 factor is the space - time volume factor . it is straightforward to develop , for small @xmath51 , an asymptotic expansion of this integral : @xmath52 \qquad\qquad ( \frac{eb}{m^2}\to 0^+ ) \label{eh}\end{aligned}\ ] ] here the @xmath53 are bernoulli numbers @xcite . the perturbative series ( [ eh ] ) is the euler - heisenberg @xcite perturbative expression for the qed effective action in a uniform magnetic background @xmath45 . it is an expansion in powers of the coupling @xmath44 , with the @xmath54 power of @xmath44 being associated with a one - fermion - loop diagram with @xmath16 external photon lines [ we have not included the divergent @xmath55 self - energy term as it contributes to the bare action by charge renormalization ] . note that only even powers of @xmath56 appear in the perturbative expansion ( [ eh ] ) . this is due to charge conjugation invariance ( furry s theorem ) . the expansion ( [ eh ] ) is also an expansion in inverse powers of @xmath57 , as is familiar for an effective field theory action ( [ eff ] ) , with the higher dimensional operators in the expansion simply being higher powers of @xmath58 . the euler - heisenberg perturbative effective action ( [ eh ] ) is not a convergent series . rather , it is an asymptotic series of the form ( [ exp ] ) with expansion parameter @xmath59 the expansion coefficients in ( [ eh ] ) alternate in sign [ because : @xmath60 , and grow factorially in magnitude : @xmath61 the growth of these coefficients is of the form indicated in the example in ( [ general ] ) , with @xmath62 , and @xmath63 . therefore , the euler - heisenberg series ( [ eh ] ) is an example of a divergent but borel summable series . if we keep just the leading large - n behavior for the coefficients @xmath23 indicated in ( [ growth ] ) , then we can immediately read off from the borel summation formula ( [ genborel ] ) the leading - order borel approximation for the sum of the divergent series ( [ eh ] ) : @xmath64 it is straightforward to evaluate this integral numerically , and one finds approximately @xmath65 agreement with the exact answer ( [ proper ] ) even when the perturbative expansion parameter @xmath29 is as large as 50 , as is shown in figure 1 . ( note that ( [ growth ] ) suggests it is perhaps more ` natural ' to take the expansion parameter to be @xmath66 , so we have plotted the leading borel approximation ( [ firstborel ] ) for @xmath29 up to @xmath67 . ) .5 cm but in this case we can do much better , because we are in the unusual situation of knowing the _ exact _ perturbative expansion coefficients @xmath23 _ for all _ @xmath16 ( not simply their leading - order growth ) : @xmath68 for each @xmath69 in this sum , the coefficient is once again of the leading form in ( [ general ] ) , with @xmath70 , and @xmath63 . thus , we can apply the borel summation formula ( [ genborel ] ) directly to yield @xmath71 the sum over @xmath69 gives successive corrections to the leading borel approximation in ( [ firstborel ] ) . the contributions with one , two and three terms are plotted in figure 1 , each compared to the exact result ( [ proper ] ) . note that only three terms are needed to obtain @xmath72 accuracy , even when the expansion parameter @xmath29 is as large as 50 . in fact , the expansion @xcite @xmath73 together with the fact that @xmath74 , shows that the borel integral ( [ fullborel ] ) agrees precisely with the schwinger proper - time result ( [ proper ] ) . that is , schwinger s formula ( [ proper ] ) can be viewed as the borel sum of the ( divergent ) euler - heisenberg perturbative series ( [ eh ] ) . or , in other words , the euler - heisnberg perturbative series ( [ eh ] ) can be obtained by an asymptotic expansion of schwinger s integral representation formula ( [ proper ] ) , when @xmath51 is small . to get a sense of the size of the expansion parameter ( [ param ] ) appearing in the euler - heisenberg series , it is instructive to re - instate factors of @xmath75 and @xmath76 : @xmath77 the first equality in ( [ realparam ] ) expresses @xmath29 in terms of the square of the ratio of the cyclotron energy @xmath78 to the electron rest mass energy @xmath79 , while the second equality expresses it in terms of the fourth power of the ratio of the electron compton wavelength @xmath80 to the `` magnetic length '' scale @xmath81 set by the magnetic field . the critical magnetic field strength at which the dimensionless parameter @xmath29 in ( [ realparam ] ) is order 1 is @xmath82 gauss . this is well above currently available laboratory static magnetic field strengths , which are approximately @xmath83 gauss , in which case @xmath84 is _ extremely _ small . however , the critical field @xmath85 is comparable to the scale of magnetic field strengths observed in astrophysical objects such as supernovae and neutron stars which can have magnetic fields of the order of @xmath86 gauss @xcite . now consider the euler - heisenberg - schwinger effective action in a uniform background _ electric _ field of strength @xmath87 , instead of the uniform magnetic background @xmath45 . perturbatively , the only difference is that @xmath58 is replaced by @xmath88 , which amounts to changing the sign of the expansion parameter @xmath29 in ( [ param ] ) . therefore , in a uniform electric background , the euler - heisenberg perturbative effective action ( [ eh ] ) becomes a _ non - alternating _ series @xmath89 ( recall that @xmath90 , so @xmath91 is non - alternating . ) this series is clearly divergent and since the coefficients are non - alternating , it is not borel summable . nevertheless , using the borel dispersion relations we can extract the imaginary part of the effective action . if we keep just the leading large-@xmath16 growth ( [ growth ] ) of the expansion coefficients , then we can immediately read off from the borel dispersion relation result ( [ genimag ] ) the leading behavior of the imaginary part of the effective action in the electric background : @xmath92 \qquad\qquad ( \frac{ee}{m^2}\to 0^+ ) \label{leadingimag}\end{aligned}\ ] ] this imaginary part has direct physical significance - it gives half the electron - positron pair production rate in the uniform electric field @xmath87 @xcite . actually , as in the magnetic case , we can do better than just the leading behavior ( [ leadingimag ] ) . combining the expansion coefficients ( [ zeta ] ) with the borel dispersion formula ( [ genimag ] ) we immediately find @xmath93 \qquad\qquad ( \frac{ee}{m^2}\to 0^+ ) \label{fullimag}\end{aligned}\ ] ] which is schwinger s classic result @xcite . note that in the electric case the relevant small dimensionless parameter is [ compare with ( [ realparam ] ) ] @xmath94 which is ( up to a factor of @xmath95 ) the ratio of the work done by the electric field @xmath87 accelerating a particle of charge @xmath44 through an electron compton wavelength , to the energy required for pair production . for typical electric fields this is a very small number , so the exponential factors in ( [ leadingimag ] ) and ( [ fullimag ] ) are extremely small . the critical electric field at which the non - perturbative factors become significant is @xmath96 . this is still several orders of magnitude beyond the field obtainable in current lasers @xcite . to conclude this section , we stress that this constant - field case provides an explicit example of dyson s argument @xcite that qed perturbation theory can not be convergent as a series in the fine structure constant @xmath97 , because this would mean that the stable vacuum , with @xmath9 positive , is smoothly connected to the unstable vacuum , with @xmath9 negative , at least in a small neighborhood of the origin . the perturbative euler - heisenberg series in ( [ eh ] ) and ( [ ehe ] ) are expansions in powers of @xmath98 and @xmath99 , respectively . changing from a magnetic background to an electric background involves replacing @xmath100 with @xmath101 , which amounts to changing the sign of @xmath102 ( i.e. , the fine structure constant ) , since @xmath44 always appears as @xmath56 or @xmath103 . if the euler - heisenberg perturbative series were convergent , then the change from @xmath100 to @xmath101 would not produce any non - perturbative imaginary part in the effective action . thus there would be no pair production and we would miss the genuine physical instability of the qed vacuum in an external electric field . hence the euler - heisenberg perturbative series _ can not _ be convergent . so far , we have re - phrased well - known qed results in the language of borel summation . now we turn to the main point of this paper , which is to go beyond the euler - heisenberg - schwinger constant field results for the qed effective action . perturbatively , this leads to a derivative expansion ( [ derivative ] ) which is a formal expansion in increasing numbers of derivatives of the background field strength @xmath104 + s^{(2)}[f_{\mu\nu},\partial_\mu f_{\nu\rho}]+ \dots \label{der}\end{aligned}\ ] ] where @xmath105 involves no derivatives of the background field strength @xmath7 , while the first correction @xmath106 involves two derivatives of the field strength , and so on . the increasing powers of derivatives are balanced by increasing powers of @xmath107 . unfortunately , it is very difficult to say anything precise about the convergence or divergence of such a derivative expansion because it is not an actual series , as in ( [ exp ] ) , in terms of a dimensionless expansion parameter . for a general background there is a rapid proliferation of the number of independent terms with a given number of derivatives of the field strength ( see @xcite for the first order and @xcite for higher orders ) . even a first order derivative expansion calculation is quite non - trivial . moreover , for a general background field , it is extremely difficult to estimate and compare the magnitude of the various terms in the derivative expansion ( [ der ] ) . so a perturbative analysis to high orders in a derivative expansion appears prohibitively difficult for a general background field strength . this makes it difficult to reconcile a perturbative derivative expansion calculation with the calculation of the non - perturbative imaginary part of the effective action for an electric background . we explore this question below . as a first step towards overcoming these obstacles , we can consider restricted classes of special backgrounds for which the derivative expansion reduces to a manageable form . the qed effective action has recently been computed _ exactly _ for either ( not both ) of the following special inhomogeneous background magnetic @xcite and electric fields @xcite : @xmath108 it has of course long been known that the dirac equation is exactly solvable for such backgrounds , a fact that has permitted many authors to study the qed effective action in these backgrounds @xcite . the new feature of @xcite is that all the momentum traces have been performed so that the effective action is expressed as a simple integral representation [ involving a _ integral ] , as in schwinger s classic result ( [ proper ] ) for the uniform background field . this then permits the expansion of the effective action as a true series , whose convergence / divergence properties can be studied in detail . on the right side of ( [ inhom ] ) , @xmath109 and @xmath110 are constant vectors . thus , @xmath111 points in a fixed direction in space and is static , but its magnitude varies in the @xmath112 direction , with a characteristic length scale @xmath113 that is arbitrary . similarly , @xmath114 points in a fixed direction in space and is spatially uniform , but its magnitude varies in time , with a characteristic time scale @xmath115 that is arbitrary . these electric and magnetic fields satisfy the homogeneous maxwell equations , but not the inhomogeneous ones , so classically we should think of them as being supported by external currents . within a quantum path integral they simply correspond to some particular vector potential @xmath116 . note that in the limits @xmath117 and @xmath118 we regain the uniform field cases relevant for the euler - heisenberg - schwinger effective action . we therefore expect that , for the inhomogeneous backgrounds ( [ inhom ] ) , the derivative expansion of the effective action should correspond to an expansion for large @xmath113 and large @xmath115 . we concentrate first on the magnetic field case , and then use a duality transformation @xmath0 to convert to the electric field case . since each derivative of the magnetic field in ( [ inhom ] ) produces a factor of @xmath119 , a natural dimensionless expansion parameter for the derivative expansion in the magnetic case is ( restoring factors of @xmath75 and @xmath76 ) @xmath120 this dimensionless parameter is the square of the ratio of the magnetic length scale @xmath81 ( which is set by the peak magnetic field magnitude @xmath121 ) to @xmath113 , the length scale of the spatial inhomogeneity of the magnetic field . alternatively , we can combine this with the dimensionless perturbation parameter ( [ realparam ] ) of the constant field case , to obtain another dimensionless expansion parameter @xmath122 which is essentially ( up to factors of @xmath123 ) the square of the ratio of the compton wavelength of the electron to the inhomogeneity scale @xmath113 . thus , for a magenetic background with a macroscopic inhomogeneity scale @xmath113 , the ratio @xmath124 is extremely small . in this form , we clearly recognize the derivative expansion as an expansion in inverse powers of @xmath57 , as in a general effective field theory expansion ( [ eff ] ) . indeed , these expectations are borne out by the exact renormalized effective action @xcite , which has the double perturbative expansion ( we set @xmath76 and @xmath75 to 1 again ) @xmath125 in ( [ exact ] ) it is understood that the double sum excludes the @xmath126 term , as this term contributes to the logarithmically divergent charge renormalization of the bare action @xcite . we emphasize that the summation indexed by @xmath127 in ( [ exact ] ) corresponds precisely to the orders of the derivative expansion . this has been verified @xcite by comparison with independent derivative expansion calculations of the leading and first - correction term , computations that were done using the proper - time method @xcite . for example , to make the comparison with the leading term of the derivative expansion we simply take the euler - heisenberg constant field answer ( [ eh ] ) , replace @xmath45 by @xmath128 and do the @xmath112 integrals , using the fact that @xmath129 . this reproduces the @xmath130 term in ( [ exact ] ) . a similar argument @xcite holds for the first correction term in the derivative expansion , which reproduces the @xmath131 term in ( [ exact ] ) . it is important to note that the perturbative expression ( [ exact ] ) for the effective action is an explicit double sum ( with no remaining integrals ) in terms of two dimensionless parameters . one parameter @xmath51 characterizes the perturbative expansion in powers of the coupling @xmath44 , while the other parameter @xmath124 characterizes the derivative expansion . moreover , all the expansion coefficients are known exactly . thus , we can apply to this effective action the standard techniques for the analysis of divergent series ( such as borel summation ) . consider a fixed order @xmath127 of the derivative expansion . this still involves a perturbative expansion in powers of the coupling @xmath44 . for @xmath130 , from ( [ exact ] ) we see that the perturbative expansion in the magnetic case is @xmath132 the expansion coefficients alternate in sign and grow factorially with @xmath69 : @xmath133 \qquad\qquad , k\to \infty \label{bgrowth}\end{aligned}\ ] ] thus , we are in the situation described by ( [ general ] ) . therefore , this leading order ( @xmath130 ) in the derivative expansion is a divergent but borel summable series [ this is hardly surprising , as we already showed in the previous section that the euler - heisenberg effective action is itself a divergent but borel summable series ] . applying the borel summation formula ( [ genborel ] ) , the leading borel approximation for the series ( [ bzero ] ) is @xmath134 \qquad\qquad ( \frac{eb}{m^2}\to 0^+ ) \label{bzb}\end{aligned}\ ] ] it is instructive to compare this with the leading borel approximation ( [ firstborel ] ) to the effective action for a _ uniform _ magnetic background @xmath45 . to make this comparison we replace the uniform background @xmath45 in ( [ firstborel ] ) with the inhomogeneous background @xmath135 , and then perform the @xmath112 integration . thus @xmath136\ , \psi(\frac{1}{2},-1;\frac{m^2}{eb } s ) \label{bzbcheck}\end{aligned}\ ] ] where @xmath137 is the confluent hypergeometric function , with integral representation @xmath138 noting that @xmath139 for large @xmath140 , we see that ( [ bzbcheck ] ) is indeed in agreement with ( [ bzb ] ) when the perturbative expansion parameter @xmath51 is small . now consider the @xmath130 term of the derivative expansion for the inhomogeneous electric background @xmath141 . perturbatively , we replace @xmath58 by @xmath88 in the expansion ( [ bzero ] ) , so that the expansion coefficients are now non - alternating . thus , in the electric case the series is divergent but _ not _ borel summable . nevertheless , we can use the borel dispersion relations ( [ genimag ] ) to compute the imaginary part of the effective action . a direct application of ( [ genimag ] ) leads to @xmath142 \qquad\qquad ( \frac{ee}{m^2}\to 0^+ ) \label{ezb}\end{aligned}\ ] ] now compare this with the leading order borel result ( [ leadingimag ] ) for a uniform electric background @xmath87 . replacing @xmath87 by the inhomogeneous field @xmath141 and then performing the @xmath143 integration , we find @xmath144\ , \psi(\frac{1}{2},-1;\frac{m^2 \pi}{ee } ) \qquad\qquad ( \frac{ee}{m^2}\to 0^+ ) \label{ezbcheck}\end{aligned}\ ] ] when the perturbative parameter @xmath145 is small , this agrees precisely with the resummed answer in ( [ ezb ] ) . a similar analysis for the @xmath131 term ( i.e. the first derivative expansion correction term ) in ( [ exact ] ) shows that in the magnetic case the perturbative expansion in powers of @xmath146 is borel summable and has a leading borel approximation : @xmath147 \label{boneb}\end{aligned}\ ] ] we can compare this with the first correction in the derivative expansion which has been computed independently using proper - time methods @xcite : @xmath148^{\prime\prime\prime}\ , \int d^4x \ , { ( e b^\prime(x))^2\over eb(x ) } \ , \exp\left [ -\frac{m^2}{eb(x)}s\right ] \label{fo}\end{aligned}\ ] ] with the inhomogeneous background @xmath149 , the space - time integrals can be done to yield @xmath150^{\prime\prime\prime } \ , \exp\left [ -\frac{m^2}{eb}s\right]\ , \psi(\frac{3}{2},0;\frac{m^2}{eb}s ) \label{fd}\end{aligned}\ ] ] for small perturbative parameter @xmath51 this reduces ( after some integrations by parts in @xmath151 ) to the expression ( [ boneb ] ) which was obtained by borel summation of the @xmath131 term of the double series ( [ exact ] ) . in the electric case , the @xmath131 term in ( [ exact ] ) is a perturbative series expansion that is divergent but not borel summable , as the expansion coefficients are non - alternating . we can compute the imaginary contribution to the effective action using the borel dispersion relation result ( [ genimag ] ) @xmath152 \qquad\qquad ( \frac{ee}{m^2}\to 0^+ ) \label{efe}\end{aligned}\ ] ] this should be compared to the first - order derivative expansion result from a field - theoretic calculation @xcite @xmath153 \label{efefirst}\end{aligned}\ ] ] with the inhomogeneous background @xmath154 , the space - time integrals can be done to yield @xmath155 \qquad\qquad ( \frac{ee}{m^2}\to 0^+)\nonumber\\ & = & { l^3 m^6\over 32\sqrt{\pi } \tau e^2 e^2 } \exp\left[-\frac{m^2\pi}{ee}\right]\,\psi(\frac{3}{2},3;\frac{m^2\pi}{ee } ) \label{efecheck}\end{aligned}\ ] ] which agrees precisely with ( [ efe ] ) when @xmath145 is small . having verified explicitly that the borel techniques work for the first two orders of the derivative expansion , for both the magnetic and electric background , we now turn to the higher orders @xmath156 of the derivative expansion . these are very difficult to compute with field theory techniques @xcite . nevertheless , for the particular inhomogeneous backgrounds in ( [ inhom ] ) , the exact result ( [ exact ] ) contains _ all _ orders in the derivative expansion , and so it is a simple matter to study the divergence properties of each order @xmath127 of the derivative expansion . from ( [ exact ] ) , the @xmath157 order derivative expansion contribution to the effective action is ( @xmath158 ) @xmath159 for fixed @xmath127 this contribution is itself a perturbative series expansion in terms of the dimensionless parameter @xmath160 , with expansion coefficients @xmath161 these coefficients alternate in sign and grow factorially with @xmath69 , for any fixed @xmath158 . therefore , for the inhomogeneous magnetic background in ( [ inhom ] ) , each order of the derivative expansion is a divergent but borel summable asymptotic series . using the leading large @xmath69 behavior in ( [ dc ] ) together with the borel integral ( [ borel ] ) , the leading borel approximation to the @xmath157 order of the derivative expansion is ( for @xmath158 ) @xmath162\over 1+s^2/\pi^2 } \ , \frac{1}{j!}\ , \left({-m^4 s^3\over 4\pi^2 \lambda^2 e^3 b^3}\right)^j \qquad\qquad ( \frac{eb}{m^2}\to 0^+ ) \label{jb}\end{aligned}\ ] ] note the remarkable fact that these leading borel approximations , for each order @xmath127 of the derivative expansion , can be resummed into an exponential . the @xmath130 term must be treated separately [ because the sum in ( [ exact ] ) begins at @xmath163 when @xmath130 , because of charge renormalization ] . combining the @xmath130 result ( [ bzero ] ) with the @xmath158 result ( [ jb ] ) we find @xmath164 \label{bresum}\end{aligned}\ ] ] the first term is a finite charge renormalization , on top of the usual infinite charge renormalization for the uniform field case @xcite . in the second term , we see the interesting result that the leading borel approximations to each order of the derivative expansion _ exponentiate _ when they are resummed . a similar phenomenon occurs with the electric background . for any @xmath158 , a straightforward application of the borel dispersion result ( [ genimag ] ) , using the growth estimate of the coefficients in ( [ dc ] ) gives @xmath165 \ , \frac{1}{j!}\ , \left({m^4 \pi\over 4 \tau^2 e^3 e^3}\right)^j \qquad\qquad ( \frac{ee}{m^2}\to 0^+ ) \label{ej}\end{aligned}\ ] ] in fact , comparing with ( [ ezb ] ) we see that this result for the imaginary part also holds for @xmath130 . thus , resumming the derivative expansion , the result immediately exponentiates ) does not affect the imaginary part of the effective action in the electric case . ] to @xmath166 \qquad\qquad ( \frac{ee}{m^2}\to 0^+ ) \label{eresum}\end{aligned}\ ] ] it is very interesting to see that the leading borel approximations to each order of the derivative expansion can be resummed into an exponentiated form . thus , resumming the leading borel contributions amounts to a resummed perturbative modification of the non - perturbative exponent @xmath167 in the schwinger uniform field result ( [ leadingimag ] ) . clearly , the modification of the exponent derived in ( [ eresum ] ) is much more significant than a modification of the prefactor , which is all that is obtained by looking at a single ( low ) order of the derivative expansion @xcite . we stress that we are able to exponentiate the corrections because in this case we know the large order perturbative behavior for _ every _ order of the derivative expansion . finally , notice the appearance of the dimensionless parameter @xmath168 in the correction to the exponent in ( [ eresum ] ) . we will address the significance of this parameter below . in the previous section we took the leading borel approximation to the perturbative expansion [ the sum over @xmath69 in ( [ exact ] ) ] and then performed the derivative expansion [ the sum over @xmath127 in ( [ exact ] ) ] exactly . actually , it is possible to re - sum the perturbative @xmath69 expansion in ( [ exact ] ) , and express it as an integral . that is , we can write the effective action as a single ( derivative expansion ) series : @xmath169 where the expansion coefficients @xmath170 are now _ functions _ of the parameter @xmath51 . to study the divergence properties of this derivative expansion we need to know the rate of growth , for large @xmath127 , of the expansion coefficients @xmath170 appearing in ( [ dd ] ) . for @xmath171 there is a simple integral representation for these coefficients [ this amounts to summing the perturbative expansion in the double series ( [ exact ] ) , thereby reducing ( [ exact ] ) to the single series in ( [ dd ] ) ] : @xmath172 ds\end{aligned}\ ] ] here , @xmath173 is the standard hypergeometric function @xcite : @xmath174 for any @xmath127 it is straightforward to evaluate these coefficients @xmath170 numerically for various values of the dimensionless parameter @xmath51 . we find that the coefficients @xmath170 alternate in sign and grow in magnitude ( for large @xmath127 ) like @xmath175 , with some real @xmath9 , @xmath26 . this is illustrated in figure 2 , where the ratio of successive magnitudes @xmath176 shows a clear quadratic growth , for various values of @xmath51 . this shows that the derivative expansion itself ( [ dd ] ) is a divergent but borel summable asymptotic series . .5 cm in this section we apply the leading borel approximation to the derivative expansion , for a given order of the perturbative expansion , and then re - sum the perturbative expansion . first , take a fixed order @xmath177 of the perturbative expansion in ( [ exact ] ) @xmath178 this is a derivative expansion in terms of the dimensionless parameter @xmath124 , with expansion coefficients @xmath179 thus , the coefficients @xmath180 alternate in sign and grow factorially with @xmath127 , for any fixed @xmath177 . therefore , for any fixed order of perturbation theory in @xmath51 , the derivative expansion is a divergent but borel summable asymptotic series . now consider resumming the leading borel summation of each order @xmath69 of the perturbative expansion for the inhomogeneous electric background in ( [ inhom ] ) . from ( [ pt ] ) and the borel dispersion relation ( [ genimag ] ) the imaginary part of the effective action at order @xmath69 of the perturbative expansion is @xmath181 this is in fact valid for all @xmath182 . resumming these leading borel contributions gives the leading behavior , for large @xmath183 ( as is appropriate for the derivative expansion ) , @xmath184 \qquad\qquad ( \frac{1}{m\tau}\to 0^+ ) \label{presum}\end{aligned}\ ] ] once again , we see that the resummation of the leading borel contributions exponentiates , producing an exponent that is modified from that found in the schwinger uniform field result ( [ leadingimag ] ) . however , the exponential behavior in ( [ presum ] ) is very different from the exponential behavior in ( [ eresum ] ) , and indeed from the exponential behavior in the uniform case ( [ leadingimag ] ) . to understand this difference , we first recall that the exponential behavior in ( [ eresum ] ) was obtained by resumming the leading borel approximations to each order of the derivative expansion , while the exponential behavior in ( [ presum ] ) was obtained by resumming the leading borel approximations to each order of the perturbative expansion . that is , to obtain ( [ eresum ] ) we take the leading borel approximation for the @xmath69 summation in ( [ exact ] ) , for each fixed @xmath127 , and then resum over @xmath127 ; while to obtain ( [ presum ] ) we take the leading borel approximation for the @xmath127 summation in ( [ exact ] ) , for each fixed @xmath69 , and then resum over @xmath69 . the difference between these two approaches is governed by the relative size of the two dimensionless expansion parameters . in the perturbative expansion we assume that @xmath145 is small , and in the derivative expansion we assume that @xmath185 is small . the distinction between the two answers ( [ eresum ] ) and ( [ presum ] ) depends on the dimensionless combination : @xmath186 there are two natural regimes of interest : @xmath187 in the non - perturbative regime , @xmath188 implies that @xmath189 therefore , in this regime we expect the leading exponential contribution to @xmath190 to be the schwinger uniform field factor @xmath191 , as indeed is found in ( [ eresum ] ) . the resummation of leading borel approximations derived in ( [ eresum ] ) gives corrections to this leading exponent @xmath192 \label{bc1}\ ] ] which has the form of a small correction in terms of the parameter @xmath168 , which is small in this non - perturbative regime . on the other hand , in the perturbative regime , @xmath193 implies that @xmath194 in this regime , the exponential factor @xmath195 dominates the schwinger factor @xmath196 and gives a new leading contribution to @xmath190 . the resummation of leading borel approximations in ( [ presum ] ) gives corrections to this leading exponent @xmath197 \label{bc2}\ ] ] where in the perturbative regime the parameter @xmath198 is small . now ask the question : how does the time dependence of the inhomogeneous electric background in ( [ inhom ] ) modify schwinger s constant field result ( [ leadingimag ] ) ? the answer depends critically on how the characteristic time scale @xmath115 of the inhomogeneity relates to the time scale @xmath199 set by the peak electric field @xmath87 . this then determines [ see ( [ gov ] ) ] the relative magintude of the two expansion parameters @xmath185 ( corresponding to the derivative expansion ) and @xmath145 ( corresponding to the perturbative expansion ) . in the non - perturbative regime , the smaller of the two parameters is the derivative expansion parameter : @xmath200 . thus , we apply the leading borel approximation to the perturbative expansion [ the sum in powers of @xmath145 ] , and then resum the derivative expansion [ the sum in powers of @xmath185 ] exactly . this is exactly what was done in deriving the result ( [ eresum ] ) . in the perturbative regime , the smaller of the two parameters is the perturbative parameter : @xmath203 . thus , we apply the leading borel approximation to the derivative expansion [ the sum in powers of @xmath185 ] , and then resum the perturbative expansion [ the sum in powers of @xmath145 ] exactly . this is exactly what was done in deriving the result ( [ presum ] ) . note that , in each case , our ability to treat the remaining sum _ exactly _ relied on the fact that the leading borel approximations came out in a form that could be exponentiated . it is not a priori obvious that this dramatic simplification had to occur . however , in the next section we will see that this exponentiation is very natural in terms of a wkb formulation . in a general background , @xmath204 , the effective action is too complicated to permit such a detailed borel analysis as has been done in the previous sections for the uniform background and for the special inhomogeneous backgrounds in ( [ inhom ] ) . clearly , we do not know the spectrum of the dirac operator for a general background , so some sort of approximate expansion method , such as the derivative expansion , is required . but the formal derivative expansion ( [ der ] ) is not a series expansion , because more and more ( independent ) tensor structures appear with each new order of the derivative expansion . this is because with more derivatives there are more indices to be contracted in various ways . another way of saying this is that the inhomogeneity of a general background can not be characterized by a single ( or even a finite number of ) scale parameter(s ) , such as the length scale @xmath113 or the time scale @xmath115 in ( [ inhom ] ) . another problem is that in general it is difficult to estimate the size of various terms in such a derivative expansion when the background field strength @xmath7 is an arbitrary function of space - time . so , even if we could organize the derivative expansion into a sensible series , it would be difficult to estimate the magnitude of the coefficients at very high orders in the series , as is needed for a borel analysis . nevertheless , it is still instructive to consider a further generalization of the particular inhomogeneous backrounds in ( [ inhom ] ) . in this section we relax the condition that we know the exact spectrum of the dirac operator , but keep the restriction that the backgrounds only depend on one space - time coordinate . this has the effect of reducing the spectral problem to that of an ordinary differential operator . as is clear from ( [ action ] ) , the effective action is determined by the spectrum of the operator @xmath205 { \bf 1 } + \frac{e}{2}f_{\mu\nu } \sigma^{\mu\nu } \label{op}\ ] ] where @xmath206 $ ] . if we restrict our attention to inhomogeneous backgrounds that point in a fixed direction in space ( say , the @xmath140 direction ) and depend on just one space - time coordinate , then the operator in ( [ op ] ) can be diagonalized with a suitable gauge choice and a suitable dirac basis . for example , the spatially inhomogeneous magnetic field @xmath207 can be realized with the vector potential @xmath208 . then , in the standard dirac representation @xcite for the gamma matrices , the operator @xmath209 is diagonal , with diagonal entries ( appearing twice each on the diagonal ) : @xmath210 similarly , the time - dependent electric field @xmath211 can be realized with the vector potential @xmath212 . then , in the standard chiral representation @xcite for the gamma matrices , the operator @xmath209 is diagonal , with diagonal entries ( appearing twice each on the diagonal ) : @xmath213 therefore , in each case ( [ genb ] ) and ( [ gene ] ) , the spectrum is determined by a one - dimensional ordinary differential operator . however , note the appearance of the factors of @xmath214 in the electric case ( [ ediag ] ) . this shows immediately the fundamental difference between a magnetic background and an electric background . in the magnetic case , the eigenvalues of the associated ordinary differential operator are real , while for the electric case , the eigenvalues of the associated ordinary differential operator have an imaginary part . ( note that in the magnetic case the boundary condition for the ordinary differential operator is for solutions that decay at @xmath215 , while in the electric case we seek solutions going like @xmath216 at @xmath217 , corresponding to the particle / antiparticle pair @xcite . the reader is encouraged to check all this explicitly for the simple case of a uniform background . ) schwinger s uniform field case corresponds to choosing the function @xmath218 appearing in ( [ genb ] ) and ( [ gene ] ) to be @xmath219 , while the inhomogeneous backgrounds in ( [ inhom ] ) correspond to choosing @xmath220 . it is well known that in each case the spectrum of the associated ordinary differential operator in ( [ bdiag ] ) and ( [ ediag ] ) is exactly solvable @xcite . this explains why it is possible to compute the effective action exactly for these backgrounds . for the more general fields in ( [ genb ] ) and ( [ gene ] ) it is not possible to find the exact spectrum . however , we can still use a wkb approach to approximate the spectrum . for a time - dependent , but spatially uniform , electric background this leads to the following wkb expression for the imaginary part of the effective action @xcite @xmath221 where the wkb exponent is @xmath222 here @xmath223 and @xmath224 for the electric field in ( [ gene ] ) . the integration in ( [ omega ] ) is between the turning points of the integrand [ this expression is somewhat symbolic - in practice , the evaluation of @xmath225 requires careful phase choices , depending on the form of the function @xmath226 @xcite ] . for the constant field case , with @xmath219 , one finds @xmath227 then the momentum integrals in ( [ wkb ] ) can be done [ recall the density of states integral : @xmath228 to yield the familiar schwinger result ( [ fullimag ] ) . in the inhomogeneous electric field @xmath141 , which corresponds to @xmath220 , we can also compute @xmath225 exactly @xcite : @xmath229 it is a straightforward , but somewhat messy , computation to check that by doing the momentum integrals in ( [ wkb ] ) with this expression for @xmath225 , one arrives at the exact integral representation , derived in @xcite , for the effective action in this inhomogeneous background . the wkb expression gives the _ exact _ result in this case because the uniform wkb approximation gives the exact spectrum of the differential operators ( [ bdiag ] ) and ( [ ediag ] ) when @xmath220 @xcite . now consider this wkb exponent @xmath225 in ( [ exactomega ] ) in the non - perturbative and perturbative limits ( [ regimes ] ) . in the non - perturbative limit we can expand @xmath225 in inverse powers of @xmath44 as @xmath230 \label{long}\ ] ] in the perturbative limit , we obtain instead @xmath231 \label{short}\ ] ] doing the momentum trace over @xmath232 , the momentum in the direction of the field , in the wkb expression ( [ wkb ] ) modifies the prefactor , but not the exponent . the traces over the transverse momenta can be done by approximating @xmath233 . this effectively replaces @xmath234 with @xmath12 in ( [ long ] ) and ( [ short ] ) , and the @xmath235 integrals in ( [ wkb ] ) contribute to the prefactor . then comparing ( [ long ] ) and ( [ short ] ) with ( [ bc1 ] ) and ( [ bc2 ] ) , we see that we regain exactly the borel resummed results ( [ eresum ] ) and ( [ presum ] ) obtained for the two extreme limits ( [ regimes ] ) . thus , this wkb analysis explains why we found two different expressions , ( [ eresum ] ) and ( [ presum ] ) , by borel resummation of the double series ( [ exact ] ) , depending on the relative size of the two dimensionless expansion parameters . to conclude this discussion of the wkb approach , it is instructive to compare the @xmath11 case with the case of an oscillating electric field @xmath236 which was studied in detail using wkb methods by brzin and itzykson @xcite . this is not an exactly solvable case , but wkb provides a semiclassical result . here @xmath237 and so the wkb exponent is @xmath238 ^ 2 } \ , dt \label{sinwkb}\ ] ] while this integral can not be done in closed form , one can consider the non - perturbative and perturbative limits @xcite . in the non - perturbative regime , where@xmath239 , @xmath240 \label{wkblow}\ ] ] which is clearly analogous to the non - perturbative limit ( [ long ] ) of the @xmath11 case . thus , in this regime , the schwinger pair production rate has a modified exponent @xmath241 \label{lowmod}\ ] ] on the other hand , in the perturbative regime , where @xmath242 , @xmath243 which is very different from the perturbative limit ( [ short ] ) of the @xmath11 case . in this regime , the wkb pair production rate for an oscillating time - dependent electric background becomes @xcite @xmath244 this is a perturbative expression , with the perturbative parameter @xmath245 raised to a power @xmath246 which is the number of photons of frequency @xmath247 required to match the pair creation energy @xmath248 . so , while the non - perturbative results ( [ long ] ) and ( [ wkblow ] ) are very similar for the cases @xmath249 and @xmath250 respectively , the corresponding perturbative results are very different from one another for these two time - dependent electric backgrounds . this difference can be traced to the different perturbative limits ( [ short ] ) and ( [ wkbhigh ] ) of the wkb exponent @xmath225 . physically , this is not so surprising if we recall that the perturbative limits , @xmath193 and @xmath251 , can also be thought of as short - pulse and high - frequency limits ( respectively ) , in which case the @xmath252 and @xmath253 profiles of the electric background are significantly different . in conclusion , we have shown that the qed effective action is a divergent double series for the inhomogeneous magnetic background @xmath254 and for the inhomogeneous electric background @xmath249 . in particular , we have also demonstrated that for these inhomogeneous background fields the derivative expansion is itself divergent . borel summation techniques have been used to relate the rate of divergence to the non - perturbative imaginary part of the effective action , which determines the pair - production rate in the time - dependent electric background . remarkably , the leading borel approximations exponentiate to yield corrections to the familiar exponent appearing in the constant field case . these resummations can also be explained using a wkb analysis of the imaginary part of the effective action . finally , we conclude by considering the question of borel summability of the perturbative effective action when the inhomogeneous background has the more general form in ( [ genb ] ) or ( [ gene ] ) . the exponential form of the wkb expression ( [ wkb ] ) for the imaginary part of the effective action is very suggestive of the exponential imaginary parts found from the borel dispersion relation ( [ genimag ] ) . indeed , writing @xmath255 we can read the borel dispersion relation backwards [ with @xmath256 , @xmath257 , and @xmath258 to obtain the corresponding asymptotic expansion of the real part of the effective action @xmath259 where in the magnetic case @xmath260 . this expression is consistent with the result from the resolvent method ( although to be strictly correct we need to specify carefully phase conventions for @xmath225 @xcite ) . for example , in the constant magnetic case @xmath261 , and it is easy to verify [ recalling that @xmath262 that the expansion ( [ back ] ) reproduces the euler - heisenberg expansion ( [ eh ] ) . interestingly , the expansion ( [ back ] ) is a reorganization of the usual perturbative expansion of the effective action , and this reorganized form already makes manifest the generic divergence properties of the effective action , since the bernoulli numbers have leading behavior @xmath263 . thus , we see that in the magnetic case the expansion ( [ back ] ) has coefficients that alternate in sign and grow factorially in magnitude . in this formal sense , the divergence properties of the effective action discussed in this paper , for the special cases with @xmath264 and @xmath265 , appear to extend to more general backgrounds . however , we caution that this argument is formal , as it neglects the possible appearance of other poles and/or cuts in the borel plane which might invalidate the naive use of the borel dispersion relation ( [ genimag ] ) . nevertheless , it is suggestive to associate the non - perturbative wkb expression ( [ wkb ] ) for the imaginary part of the effective action with the divergent expansion ( [ back ] ) for the real part . this work has been supported by the u.s . department of energy grant de - fg02 - 92er40716.00 , and by the university of connecticut research foundation . we thank carl bender for helpful discussions . v. weisskopf , `` uber die elektrodynamik des vakuums auf grund der quantentheorie des elektrons '' , kong . dans . math - fys . medd . xiv no . 6 ( 1936 ) , reprinted in _ quantum electrodynamics _ , j. schwinger ( ed . ) ( dover , new york , 1958 ) . for some more recent analyses of the constant field case , see : s. blau , m. visser and a. wipf , `` analytic results for the effective action '' , int . j. mod . phys . * a6 * , 5409 ( 1991 ) ; r. soldati and l. sorbo , `` effective action for dirac spinors in the presence of general uniform electromagnetic fields '' , phys . lett . * b426 * , 82 ( 1998 ) . i. kosztin , s. kos , m. stone and a. j. leggett , `` free energy of an inhomogeneous superconductor : a wave function approach '' , phys . b * 58 * , 9368 ( 1998 ) ; s. kos and m. stone , `` the gradient expansion for the free - energy of a clean superconductor '' , cond - mat/9809182 . g. dunne and t. hall , `` an exact @xmath267 effective action '' , phys . b * 419 * , 322 ( 1998 ) . this paper generalizes an earlier result for @xmath268 : d.cangemi , e.dhoker , and g.dunne , `` effective energy for @xmath268 with semi - localized static magnetic field : a solvable model '' , phys . d * 52 * , r3163 ( 1995 ) . c. a. hurst , `` the enumeration of graphs in the feynman - dyson technique '' , proc . a * 214 * , 44 ( 1952 ) , `` an example of a divergent perturbation expansion in field theory '' , proc . 48 * , 625 ( 1952 ) ; w. thirring , `` on the divergence of perturbation theory for quantized fields '' , helv . * 26 * , 33 ( 1953 ) . c. m. bender and t. t. wu , `` large - order behavior of perturbation theory '' , phys . * 27 * , 461 ( 1971 ) , `` anharmonic oscillator . ii . a study of perturbation theory in large order '' , phys . rev . d * 7 * , 1620 ( 1973 ) . c. itzykson , g. parisi and j - b . zuber , `` asymptotic estimates in quantum electrodynamics '' , phys . d * 16 * , 996 ( 1977 ) ; r. balian , c. itzykson , g. parisi and j - b . zuber , `` asymptotic estimates in quantum electrodynamics . ii '' , phys . d * 17 * , 1041 ( 1978 ) . a. chodos , `` proper - time methods in the presence of non - constant background fields '' , in proceedings of grsey memorial conference , _ strings and symmetries _ ( istanbul , 1994 ) , lecture notes in physics * 447 * , g. aktas et al ( eds . ) ( springer , 1995 ) .

group lifting structures are introduced to provide an algebraic framework for studying lifting factorizations of two - channel perfect reconstruction fir filter banks . the lifting factorizations generated by a group lifting structure are characterized by abelian groups of lower and upper triangular lifting matrices , an abelian group of unimodular gain scaling matrices , and a set of base filter banks . examples of group lifting structures are given for linear phase lifting factorizations of the two nontrivial classes of two - channel linear phase fir filter banks , the whole- and half - sample symmetric classes , including both the reversible and irreversible cases . this covers the lifting specifications for whole - sample symmetric filter banks in parts 1 and 2 of the iso / iec jpeg 2000 still image coding standard . the theory is used to address the uniqueness of lifting factorizations . with no constraints on the lifting process , it is shown that lifting factorizations are highly nonunique . when certain hypotheses developed in the paper are satisfied , however , lifting factorizations generated by a group lifting structure are shown to be unique . a companion paper applies the uniqueness results proven in this paper to the linear phase group lifting structures for whole- and half - sample symmetric filter banks . = 1 * final revision , reformatted for arxiv.org * filter bank , wavelet , unique factorization , polyphase , lifting , linear phase filter , group .

1310.2206

this paper studies two - channel finite impulse response ( fir ) perfect reconstruction filter banks @xcite . figure [ ds_poly ] depicts the polyphase - with - advance representation of a filter bank @xcite . a lifting factorization @xcite decomposes the polyphase matrices @xmath0 and @xmath1 into upper and lower triangular _ lifting matrices _ , a variant of a well - known result from linear algebra ( e.g. , ( * ? ? ? * proposition vii.2.11 ) ) stating that any matrix over a euclidean domain can be diagonalized by elementary matrices . the construction of such factorizations via the euclidean algorithm was given for general fir perfect reconstruction filter banks in @xcite and was subsequently refined for linear phase filter banks in @xcite . these latter works were motivated by the iso / iec jpeg 2000 standard @xcite , which specifies whole - sample symmetric ( ws , or fir type 1 linear phase ) wavelet filter banks in terms of _ _ half-__sample symmetric ( hs , or fir type 2 ) lifting filters . ( the connection between filter banks and wavelet transforms is well - known and will not be treated here ; see @xcite . ) elementary matrix decompositions are known to be nonunique and it is also known that , in general , lifting factorizations of filter banks are similarly nonunique . nonetheless , this paper introduces a framework for lifting factorizations , called a _ group lifting structure _ , that allows us to prove _ uniqueness _ of lifting steps under carefully stated hypotheses . this is an extension of the group - theoretic approach to lifting introduced in @xcite . a companion paper @xcite applies the theory developed here to the two principal classes of linear phase filter banks , the ws and hs classes shown in figure [ ws - hs ] . it is proven in @xcite that there is _ only one way _ to factor a given ws filter bank into hs lifting filters as specified in @xcite , even though the given filter bank has many other lifting factorizations into different ( e.g. , nonlinear phase ) lifting filters . as shown in @xcite , factoring hs filter banks into linear phase lifting steps is more complicated than in the ws case , but @xcite contains analogous results on the uniqueness of linear phase lifting factorizations for hs filter banks . a follow - on paper currently in preparation studies the group - theoretic structure of lifting factorizations and determines the groups ( up to isomorphism ) arising in unique lifting factorizations . this work is being applied to develop a group - theoretic numerical framework for optimal filter bank design , a goal motivated by the observation that perfect reconstruction filter banks do not form vector spaces . examples of nonunique lifting factorizations are ubiquitous . consider our old friend , the haar filter bank , normalized with a polyphase determinant of 1 and a lowpass dc response of 1 : @xmath2[c]{,c15c , } 1/2 & 1/2 \\ -1 & 1 \end{ieeeeqnarraybox*}\right]\,.\ ] ] the lifting factorization of ( [ haar ] ) in @xcite and ( * ? ? ? * annex h ) is @xmath3[c]{,c15c , } 1 & 1/2 \\ 0 & 1 \end{ieeeeqnarraybox*}\right ] \left [ \begin{ieeeeqnarraybox*}[{\fontsize{11}{11}\selectfont}][c]{,c15c , } 1 & 0 \\ -1 & 1 \end{ieeeeqnarraybox*}\right]\,.\ ] ] the haar matrix has another factorization , however , with different lifting steps and a diagonal gain scaling matrix : @xmath4[c]{,c15c , } 1/2 & 0 \\ 0 & 2 \end{ieeeeqnarraybox*}\right ] \left [ \begin{ieeeeqnarraybox*}[{\fontsize{11}{11}\selectfont}][c]{,c15c , } 1 & 0 \\ -1/2 & 1 \end{ieeeeqnarraybox*}\right ] \left [ \begin{ieeeeqnarraybox*}[{\fontsize{11}{11}\selectfont}][c]{,c15c , } 1 & 1 \\ 0 & 1 \end{ieeeeqnarraybox*}\right]\,.\ ] ] similar observations are made in @xcite regarding nonunique factorizations of the 4-tap/4-tap ( @xmath5 ) orthogonal wavelet filter bank , and that example can be extended to exhibit nonunique lifting factorizations for the entire family of orthogonal daubechies ( @xmath6 ) wavelet filter banks . to appreciate just how fundamentally nonunique lifting factorizations are , equate ( [ haar_lifting_steps ] ) and ( [ alt_haar_lifting_steps ] ) and move the lifting steps from ( [ haar_lifting_steps ] ) over to the right end of ( [ alt_haar_lifting_steps ] ) . use ( * ? ? ? * section 7.3 ) to factor @xmath7 into lifting steps , yielding factorization ( [ identity_lift ] ) . rcl & = & + & & [ identity_lift ] the formulas in ( * ? ? ? * section 7.3 ) can be used to generate infinitely many other lifting factorizations of the identity . in spite of the general nonuniqueness of elementary matrix decompositions , work on the jpeg 2000 standard raised a number of questions regarding uniqueness of _ linear phase _ lifting factorizations . for instance , the lifting factorization of ws filter banks in ( * ? ? ? * theorem 9 ) involves left matrix `` downlifting '' of a ws transfer matrix into hs lifting steps . one is naturally led to ask whether the same sequence of hs lifting filters is obtained for right matrix factorizations or , more generally , for a mix of left and right factorization operations . for example , the software accompanying @xcite uses the euclidean algorithm to compute nine different lifting factorizations of a 7-tap/5-tap cohen - daubechies - feauveau ws wavelet filter bank based on cubic b - splines @xcite . one factorization ( [ cdf7 - 5 ] ) , computed in ( * ? ? ? * section 5.4 ) in synthesis matrix form , uses hs lifting filters exclusively ( the other eight factorizations involve non - symmetric lifting filters ) . l 2 ^ -9/2 + = [ cdf7 - 5 ] + this exact same linear phase lifting is obtained using both left and right matrix downlifting factorization as in @xcite . other questions about unique factorization arise when factoring hs filter banks , which contain an hs lowpass filter and a half - sample _ anti_-symmetric ( ha , or fir type 4 ) highpass filter , as in figure [ ws - hs](b ) . hs filter banks factor using whole - sample antisymmetric ( wa , or fir type 3 ) lifting filters and concentric equal - length hs base filter banks ( * ? ? ? * section vi ) . for instance , a 6-tap/10-tap hs wavelet filter bank , @xmath8 , can be derived from the same halfband spectral factorization as the 9-tap/7-tap cohen - daubechies - feauveau ws wavelet filter bank in @xcite , ( * ? ? ? * figure 8.8 ) . annex h.4.1.2.1 of @xcite has a factorization of @xmath8 in which it is lifted from a concentric 6-tap/6-tap hs base filter bank by a second - order wa lifting step . at the time that @xcite was written , it was unknown whether it was possible to lift @xmath8 from the haar filter bank using wa lifting filters . the same question was raised about a 10-tap/18-tap hs filter bank that is lifted from a 10-tap/10-tap hs base filter bank in ( * ? ? ? * annex h.4.1.2.2 ) by a fourth - order wa lifting filter . it would have simplified the writing of ( * ? ? ? * annex h ) if all hs filter banks could be lifted by wa lifting filters from a single base filter bank such as the haar , but our results will show this to be impossible . our approach , which is algebraic rather than algorithmic , ignores the mechanism used to _ generate _ lifting factorizations ( e.g. , the euclidean algorithm in @xcite or the `` downlifting '' approach in @xcite ) and instead constrains the universe of allowable lifting steps . under suitable hypotheses , we shall prove uniqueness within precisely defined classes of lifting factorizations . for example , using the results of this paper , we prove in @xcite that one always obtains the same hs lifting filters when factoring a ws filter bank into hs lifting steps , regardless of how the factorization is constructed . in @xcite we also prove uniqueness of the wa lifting filters and the equal - length hs base filter bank that arise in the factorization of a suitably normalized hs filter bank . both results follow from theorem [ thm : unique_factorization ] of the present paper , a result that does not depend on linear phase characteristics . as in @xcite , we make extensive use of the group - theoretic structure of lifting factorizations . readers may be surprised by the absence of traditional ring - theoretic methods for establishing unique factorization results over polynomial domains . while ring - theoretic techniques such as grbner bases @xcite have been used for constructing wavelet filter banks , filter banks are elements of _ matrix _ polynomial rings , which are not integral domains . more significantly , perfect reconstruction filter banks ( including lifting matrices ) are _ units _ in the ring of @xmath9 matrices over the laurent polynomials , @xmath10 $ ] . this means that traditional algebraic methods based on ideal theory can tell us nothing about uniqueness of lifting factorizations . of course , it is also surprising that unique factorization results can be proven using group - theoretic techniques since everything divides everything in a group , but this only highlights the importance of carefully formulating the universe of admissible lifting factors for a given class of polyphase matrices . the author considered recasting the problem in terms of rings of _ causal _ matrix polynomials , but doing so would have required reworking much of the basic theory of lifting and would have distracted attention from the results presented here . fortunately , the group - theoretic approach proved adequate for deriving the results of interest , and it probably resulted in a more elementary level of algebraic machinery than would have been used in a ring - theoretic approach . moreover , the technical crux of the group - theoretic approach , which is the proof of an order - increasing property , would probably be required in a ring - theoretic treatment as well . the reader should note that this paper does not require an extensive background in group theory . the basic material group axioms , subgroups , and homomorphisms found in any introductory text on `` abstract algebra '' will suffice ; e.g. , @xcite . daubechies and sweldens @xcite point out that the euclidean algorithm can generate many different lifting factorizations for a given filter bank , and they pose the analysis and exploitation of nonuniqueness as `` an interesting topic for future research . '' nonuniqueness of lifting factorizations was studied in @xcite to find optimal integer - to - integer liftings for a given reversible filter bank . recent research by zhu and wickerhauser @xcite exploits nonuniqueness to construct liftings that use only `` nearest - neighbor '' data ( i.e. , first - order lifting filters of the form @xmath11 or @xmath12 ) in order to minimize memory fetches in hardware implementations . their work shows that some lifting factorizations are much more ill - conditioned than others , which likely affects rate - distortion scalability in source coding applications . as far as the author is aware , however , there are no prior results establishing _ uniqueness _ of lifting steps for _ any _ classes of lifting factorizations . unlike @xcite , which also apply group theory to multirate discrete - time systems , the present paper makes no use of the specialized theory of _ finite _ groups . moreover , our results are not constrained to finite signals of length @xmath13 : the papers @xcite use group - theoretic tools to analyze the multiscale structure of the underlying space of finite - length signals , whereas we employ group - theoretic methods to analyze the algebraic structure of multirate fir filter banks acting on general discrete - time signals with no length constraints . section [ sec : filter ] reviews filter banks and lifting . section [ sec : abelian ] introduces the group - theoretic concepts that provide the framework for our uniqueness results , which are based on the notion of _ group lifting structures_. section [ sec : linearphase ] presents examples of group lifting structures for factoring ws and hs filter banks into linear phase lifting filters , including the elasf class of reversible hs filter banks lifted from the haar filter bank that was defined by m. adams . in section [ sec : uniqueness ] we define the key _ polyphase order - increasing property _ that a group lifting structure must satisfy to ensure unique lifting factorizations , and we prove the main uniqueness theorem . section [ sec : concl ] contains concluding remarks . this section reviews notation and introduces several important concepts and tools for working with lifting factorizations . throughout the paper , fir filters are written as ( possibly noncausal ) laurent polynomials , @xmath14\,.\ ] ] [ defn : supportinterval ] the _ support interval _ of an fir filter , denoted @xmath15\subset\mathbb{z}\;,\ ] ] is the smallest closed interval of integers containing the support of the filter s impulse response or , equivalently , the largest closed interval for which @xmath16 and if @xmath17\ ] ] then the _ order _ of @xmath18 is @xmath19 the analysis polyphase - with - advance decomposition of a filter ( * ? ? ? ( 9 ) ) is @xmath20 and its _ analysis polyphase vector representation _ is rcl ( z ) & & = _ n = c^d(n)z^-n,[poly_fltr ] + ( n ) & & [ poly_fltr_impulse ] in this paper we generally work with analysis ( rather than synthesis ) polyphase - with - advance filter bank representations , so henceforth `` polyphase '' will mean `` analysis polyphase - with - advance '' unless stated otherwise . the polyphase representation ( [ poly_fltr ] ) , ( [ poly_fltr_impulse ] ) has the _ polyphase support interval _ @xmath21 \,,\ ] ] which differs from the scalar support interval ( [ supp_int ] ) for the same filter . the _ polyphase order _ of ( [ poly_fltr ] ) is @xmath22 these definitions generalize for fir filter banks , @xmath23 . decompose each filter , @xmath24 , @xmath25 , into its polyphase vector representation , @xmath26 , as in ( [ poly_fltr ] ) and form the polyphase matrix : rcl ( z ) & & = _ n = c^d ( n)z^-n,[poly_matrix_def ] + ( n ) & & [ poly_matrix_impulse_def ] note that we use bold italics for column vectors and bold roman ( upright ) fonts for matrices . the polyphase support interval of the filter bank in ( [ poly_matrix_def ] ) , ( [ poly_matrix_impulse_def ] ) is defined to be @xmath27 \label{fb_suppint}\ ] ] and the polyphase order is @xmath28 @xmath0 is an _ fir perfect reconstruction filter bank _ ( i.e. , an fir filter bank with fir inverse ) if and only if @xmath29 the set of all such filter banks forms a multiplicative group , @xmath30 , called the _ fir filter bank group _ ( * ? ? ? * theorem 1 ) . a polyphase matrix is called _ unimodular _ @xcite if @xmath31 in ( * ? ? ? * theorem 1 ) it was noted that the filter banks satisfying ( [ ds_fir_pr ] ) form a normal subgroup , @xmath32 , of the fir filter bank group . we refer to @xmath33 , which readers may recognize as , as the _ unimodular group_. ( this meaning of _ unimodular _ is different from its use in group representation theory , where a group is called unimodular if its left and right translation - invariant measures are the same . ) daubechies and sweldens @xcite proved that any unimodular fir transfer matrix has a _ lifting factorization _ : @xmath34 we also refer to ( [ anal_lift_cascade ] ) as a _ lifting cascade_. the diagonal matrix , @xmath35 , is a _ unimodular gain scaling matrix _ with _ scaling factor _ @xmath36 . the lifting matrices @xmath37 are upper or lower triangular with ones on the diagonal and a lifting filter , @xmath38 , in the off - diagonal position . in the factorization corresponding to figure [ irr_anal_lift ] , for instance , the lifting matrix for the step containing @xmath39 ( which is a lowpass update ) is upper triangular , and the matrix for the second step ( a highpass update ) is lower triangular . [ defn : update_char ] the _ update characteristic _ of a lifting step ( or lifting matrix ) is a binary flag , @xmath40 or @xmath41 , indicating which polyphase channel is being updated by the lifting step . for instance , the update characteristic , @xmath42 , of the first lifting step in figure [ irr_anal_lift ] is `` lowpass , '' which is coded with a zero : @xmath43 . similarly , the update characteristic of the second lifting step is `` highpass '' ( coded with a one : @xmath44 ) . the update characteristic , denoted @xmath45 , is defined similarly for each lifting matrix , @xmath37 , in a lifting cascade ( [ anal_lift_cascade ] ) . ( the term `` update characteristic '' is taken from ( * ? ? ? * annex g ) . ) to cover cases that lift one filter bank to another , we do not assume all lifting cascades start from the identity ( i.e. , are `` lifted from the lazy wavelet '' ) . instead , we often work with _ partially factored _ lifting cascades relative to some _ base _ filter bank @xmath46 , @xmath47 with corresponding scalar filters @xmath48 and @xmath49 . since the lifting matrices multiply @xmath46 on the left , ( [ lift_from_b ] ) should be called a _ left _ lifting cascade . this is the primary case of interest , however , so `` partially factored lifting cascade '' will mean _ left _ cascade , as in ( [ lift_from_b ] ) , unless specified otherwise . moreover , we usually ignore the lifting steps involved in factoring @xmath46 and only study the steps @xmath37 to the left of @xmath46 in ( [ lift_from_b ] ) . cascades like ( [ anal_lift_cascade ] ) that are completely factored into lifting steps are regarded as the special case @xmath50 . lifting cascades will often be written recursively : rcl ( z ) & = & ( 1/k , k)^(n-1)(z ) , + ^(n)(z ) & = & _ n(z)^(n-1)(z ) [ recursive_cascade ] + ^(-1)(z ) & & ( z ) . this recursive formulation is the basis for the definition of the key polyphase order - increasing property in section [ sec : uniqueness : order - increasing ] . we can easily rule out a couple of degenerate cases that preclude uniqueness of lifting factors . [ defn : irreducible ] a lifting cascade ( [ lift_from_b ] ) is _ irreducible _ if all lifting steps are nontrivial ( @xmath51 ) and there are no consecutive lifting matrices with the same update characteristic ( i.e. , they strictly alternate between lower and upper triangular ) . for example , filter banks are signaled in jpeg 2000 part 2 codestreams as irreducible lifting cascades lifted from the identity ( @xmath50 ) . irreducibility is an attribute of lifting _ cascades _ rather than an attribute of individual matrices , making it qualitatively different from the notion of irreducible _ elements _ in a ring . requiring irreducibility poses no loss of generality since any lifting cascade can be simplified into an irreducible cascade using matrix multiplication . while irreducibility eliminates numerous trivial obstacles to uniqueness of lifting factors , irreducibility alone does not imply uniqueness , as the preceding examples show . in fact , by inserting factorizations of the identity such as ( [ identity_lift ] ) into a lifting of a given fir perfect reconstruction filter bank , it follows that every filter bank has infinitely many different irreducible lifting factorizations . such extreme nonuniqueness also trivializes questions like : `` can @xmath0 be lifted ( irreducibly ) from the haar filter bank ? '' just create a partial lifting factorization like , e.g. , _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ [ any lifting of @xmath52[inverse of ( [ haar_lifting_steps])][lifting ( [ identity_lift])]@xmath53 _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ and simplify to irreducible form . this trivial example can be generalized to yield the following pathological result . [ prop : nonuniqueness ] if @xmath1 and @xmath0 are any fir perfect reconstruction filter banks then @xmath1 can be irreducibly lifted from @xmath0 in infinitely many different ways . after comparing the above nonuniqueness results to the formulas in ( * ? ? ? * section 7.3 ) for factoring unimodular gain scaling matrices into lifting steps , it is tempting to dismiss the examples in section [ sec : intro : nonuniqueness ] and proposition [ prop : nonuniqueness ] ( as well as the nonunique factorizations of orthogonal filter banks in @xcite , which also involve zeroth - order lifting filters ) as a pathology somehow resulting from the use of constant or zeroth - order lifting filters , which are units in @xmath54 $ ] . to see that the failure of uniqueness can not be attributed to the use of zeroth - order lifting filters , we have the following example . [ exmp : nonunique_3_factors ] for all @xmath55 with @xmath56 , @xmath57 , the fourth - order unimodular transfer matrix @xmath58[5pt]\\ \frac{b^2-c^2}{b^2}(1+bz^{-1 } ) & \frac{b^2-c^2}{b^2}\left(1 - \frac{b^2c^2}{c^2-b^2}z^{-2}\right)\ieeestrut[14pt][8pt ] \end{array}\right]\ ] ] has two different irreducible lifting factorizations , @xmath59 specifically , @xmath60 , @xmath61 , and @xmath62 are lower triangular and @xmath63 , @xmath64 , and @xmath65 are upper triangular . the lifting filters are all first - order laurent polynomials : lcl s_0(z ) = 1+bz^-1 , & & s_0(z ) = ( 1-bz^-1 ) , + s_1(z ) = b^-2z^2(1-bz^-1 ) , & & s_1(z ) = ( 1+bz^-1 ) , + s_2(z ) = -c^2z^-2(1+bz^-1 ) , & & s_2(z ) = ( 1-bz^-1 ) . taking @xmath66 , @xmath67 in ( [ nonunique_3_factors ] ) leads to a reversible lifting of the identity ( [ identity_lift2 ] ) with linear phase lifting filters . rcl & = & + & & [ identity_lift2 ] + & & in short , without some nontrivial , carefully crafted limits on the scope of allowable lifting operations , there is no hope of obtaining unique lifting factorizations . a uniqueness theorem is always given relative to some universe of feasible solutions ; i.e. , it states that there is only one solution _ in that universe_. this makes it necessary to preface a unique factorization theorem by specifying the universe of factors with respect to which factorizations are unique . in this section , universes of lifting factorizations will be defined in terms of abelian groups of lifting matrices , abelian groups of gain scaling matrices , and sets ( not generally groups ) of base filter banks . define one - to - one mappings @xmath68\rightarrow\mathscr{n}\ ] ] that map the lifting filter @xmath69 $ ] to lifting matrices : @xmath70[c]{,c15c , } 1 & s(z ) \\ 0 & 1\ieeestrut[10pt][1pt ] \end{ieeeeqnarraybox*}\right ] \quad\mbox{and}\quad \lambda(s ) \equiv \left [ \begin{ieeeeqnarraybox*}[{\fontsize{11}{11}\selectfont}][c]{,c15c , } 1 & 0\ieeestrut[10pt][1pt]\\ s(z ) & 1\ieeestrut[10pt][1pt ] \end{ieeeeqnarraybox*}\right]\,.\ ] ] it is easy to show that @xmath71 and @xmath72 are homomorphisms from the additive group @xmath73 $ ] into the multiplicative group @xmath33 . [ lem : groupiso ] for any additive subgroup of the laurent polynomials , @xmath74 $ ] , the images @xmath75 and @xmath76 are abelian subgroups of @xmath33 that are isomorphic to @xmath77 . [ defn : lifting_matrix_groups ] for additive groups of laurent polynomials @xmath78 $ ] , the groups @xmath79 and @xmath80 are called the _ lifting matrix groups _ generated by @xmath81 and @xmath82 . note that @xmath83 and @xmath84 need not be isomorphic to one another ; i.e. , it may be that @xmath85 . another component of lifting is an abelian _ gain scaling group _ , @xmath86 , of unimodular constant matrices , @xmath87 with the product @xmath88 . @xmath86 need not contain the matrices ( [ gain_matrix ] ) corresponding to _ all _ @xmath36 ; @xmath86 could be defined in terms of any multiplicative subgroup of @xmath89 let @xmath86 act on @xmath33 via inner automorphisms : for a given @xmath90 define the inner automorphism @xmath91 , @xmath92 this is equivalent to the intertwining relation @xmath93 homomorphisms commute with inversion so there is no ambiguity in writing @xmath94 we will write `` @xmath95 '' as a shorthand for @xmath96 . note that @xmath95 is the _ inverse _ of the `` gain conjugation operator , '' @xmath97 , defined in ( * ? ? ? * section iv - a.1 ) . this makes the mapping @xmath98 a homomorphism of @xmath86 onto a subgroup @xmath99 . applying @xmath95 to a generic matrix gives ( [ lifting_conjugation ] ) : @xmath100[c]{,c15c , } a & b\\ c & d \end{ieeeeqnarraybox*}\right]= \left [ \begin{ieeeeqnarraybox*}[{\fontsize{11}{11}\selectfont}][c]{,c15c , } a & k^{-2 } b\ieeestrut[10pt][2pt ] \\ k^2 c & d\ieeestrut[10pt][1pt ] \end{ieeeeqnarraybox*}\right]\,.\ ] ] [ defn : d_invariance ] a group @xmath101 is _ @xmath86-invariant _ if all of the inner automorphisms fix the group @xmath102 ; i.e. , so that . this is equivalent to saying that @xmath86 lies in the normalizer of @xmath102 in @xmath33 : @xmath103 [ lem : d - invariance ] given an additive group @xmath77 of lifting filters , the groups @xmath75 and @xmath104 are @xmath86-invariant if and only if @xmath77 is closed under multiplication by @xmath105 for all @xmath106 . follows immediately from ( [ lifting_conjugation ] ) . the final component in a universe of lifting factorizations is a set , @xmath107 , of base filter banks from which other filter banks are lifted using lifting matrices in @xmath83 and @xmath84 . using a set rather than a single base filter bank will allow us to prove uniqueness of base filter banks as well as uniqueness of lifting matrices . @xmath107 need not form a group , as will be seen in the case of hs filter banks . we can now define a universe of lifting factorizations , called a _ group lifting structure . _ [ defn : liftingstructures ] a _ group lifting structure _ , @xmath108 , is an ordered four - tuple , @xmath109 where @xmath86 is a gain scaling group , @xmath83 and @xmath84 are upper and lower triangular lifting matrix groups , and @xmath110 . the _ lifting cascade group _ , @xmath111 , generated by @xmath108 is the subgroup of @xmath33 generated by @xmath83 and @xmath84 : @xmath112{0pt}{12pt}\mathbf{s}_1\cdots\mathbf{s}_k \,:\ , k\geq 1,\;\mathbf{s}_i\in\mathscr{u\cup l}\right\}\,.\ ] ] we say @xmath108 is a _ @xmath86-invariant group lifting structure _ if @xmath83 and @xmath84 , and therefore @xmath111 , are @xmath86-invariant groups . the set of all filter banks generated by @xmath108 is @xmath113{0pt}{12pt}\mathbf{dcb } : \mathbf{d}\in\mathscr{d},\;\mathbf{c}\in\mathscr{c}\,,\;\mathbf{b}\in\mathfrak{b}\right\}\,.\ ] ] we will regard elements of @xmath114 both as individual transfer matrices and as partially factored lifting cascades of the form ( [ lift_from_b ] ) with gain matrix @xmath106 , lifting matrices @xmath115 , and base @xmath116 . the statement that @xmath0 `` has a ( group ) lifting factorization in @xmath108 '' will mean that @xmath117 . moreover , since any group lifting factorization can be simplified via matrix multiplication into an irreducible lifting factorization with lifting matrices in the same groups , we obtain the following elementary result . [ prop : irreduciblefactorizations ] @xmath0 has a group lifting factorization in @xmath108 if and only if it has an _ irreducible _ lifting factorization in @xmath108 . [ defn : scaledliftinggroup ] the _ scaled lifting group _ , @xmath118 , generated by @xmath108 is the subgroup of @xmath33 generated by @xmath86 and @xmath111 : rcl & & = + & = & \ { ' '' '' _ 1_k : k1,_i}. [ scaled_lifting_group ] [ [ remarks ] ] remarks + + + + + + + when two adjacent factors in ( [ scaled_lifting_group ] ) have the form @xmath119 for @xmath120 and @xmath90 , we can use ( [ intertwine_gain ] ) to rewrite the product as @xmath121 where @xmath122 . in this way an arbitrary element of @xmath118 can be written @xmath123 if @xmath108 is @xmath86-invariant then @xmath124 for all @xmath125 , so ( [ prod_dc ] ) implies @xmath126 this shows that @xmath127 embodies the group - theoretic aspects of @xmath114 even if @xmath128 and @xmath114 are not groups . work in preparation analyzes the group - theoretic structure of @xmath127 and @xmath129 in detail . we now give examples of group lifting structures that parameterize the factorization of linear phase filter banks into linear phase lifting steps , based on the factorization theory in @xcite . it would also be nice to have examples of group lifting structures for paraunitary filter banks @xcite since paraunitary transfer matrices naturally form a group , but lifting is not as well - suited for paraunitary matrices as it is for ws or hs matrices . the problem is not existence of lifting factorizations : the theory in @xcite ensures that one can always factor paraunitary matrices into lifting cascades , and a three - step lifting structure is given in ( * ? ? ? * section 7.2 ) for arbitrary @xmath9 rotations , yielding a general lifting parameterization for paraunitary filter banks via givens rotations . rather , the problems stem from the fact that individual lifting matrices are never paraunitary in their own right . this means that lifting factorization of paraunitary matrices necessarily takes place _ outside _ of the paraunitary group . consequently , there does not appear to be any simple way of defining a group lifting structure that algebraically enforces the paraunitary property . moreover , it means that the paraunitary property is not robust with respect to quantization effects in finite - precision lifting implementations . this is in marked contrast to the cases of ws and hs filter banks . in ( * theorem 3 ) it was shown that the family of delay - minimized ws filter banks with real impulse responses are defined by an intertwining relation , @xmath130 with @xmath131 . filter banks satisfying ( [ ws_intertwining ] ) form a subgroup of @xmath30 called the _ ws group_. [ defn : uni_ws_group ] the _ unimodular ws group _ , @xmath132 , is the intersection of the ws group and @xmath33 ; i.e. , @xmath132 is the group of all real fir transfer matrices that satisfy both ( [ ds_fir_pr ] ) and ( [ ws_intertwining ] ) . [ exmp : uni_ws_lifting_structure ] _ group lifting structure for the unimodular ws group . _ define the following groups of half - sample symmetric real laurent polynomials : rcl _ 0 & & \{s(z)[z , z^-1 ] : s(z^-1)=zs(z)},[ws_g0 ] + _ 1 & & \{s(z)[z , z^-1 ] : s(z^-1)=z^-1s(z)}.[ws_g1 ] according to ( * ? ? ? * theorem 9 ) , every unimodular ws filter bank factors completely using lifting filters from @xmath81 and @xmath82 . specifically , the lifting matrix groups are and @xmath133 . since unimodular ws filter banks factor completely ( i.e. , can be `` lifted from the lazy wavelet '' ) using matrices in @xmath83 and @xmath84 , set @xmath134 . let @xmath135 and @xmath136 , which is @xmath137-invariant by lemma [ lem : d - invariance ] . in terms of @xmath138 , the ws lifting factorization theorem ( * ? ? ? * theorem 9 ) can be stated succinctly as rcl & = & + & = & where @xmath139 and @xmath140 . [ [ irreversible - ws - filter - banks - in - jpeg2000 ] ] irreversible ws filter banks in jpeg 2000 + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + jpeg 2000 part 2 ( * ? ? ? * annex g ) allows user - defined irreversible ws filter banks specified as irreducible lifting cascades in @xmath141 . jpeg 2000 does not accommodate _ all _ possible filter banks in @xmath141 , however , because the standard imposes a lowpass dc gain normalization requirement ( * ? ? ? * annex g.2.1 ) , @xcite : @xmath142 the irreversible ws filter banks permitted by ( * ? ? ? * annex g ) form a proper subset of @xmath141 : each unnormalized lifting cascade @xmath143 that satisfies @xmath144 must be paired with the unique gain scaling factor @xmath145 , which implies ( [ lowpass_dc_gain ] ) . see @xcite for an analysis of jpeg 2000 gain normalization . the ws filter banks satisfying the normalization ( [ lowpass_dc_gain ] ) do _ not _ form a subgroup of @xmath132 . for instance , the filter bank @xmath146 has @xmath147 with lifting filter @xmath148 the normalized polyphase matrix @xmath149 satisfies ( [ lowpass_dc_gain ] ) but @xmath150 does not , hence such matrices do not form a group . [ exmp : rev_ws_lifting_structure ] _ group lifting structure for the reversible ws group . _ in contrast to the group @xmath132 , whose definition is independent of the notion of lifting , the group of _ reversible _ ws filter banks , @xmath151 , is defined directly in terms of lifting @xcite . let @xmath152 be the ring of dyadic rationals , @xmath153 and define groups of dyadic symmetric laurent polynomials : rcl _ 0 & & \{s(z)[z , z^-1 ] : s(z^-1)=zs(z)},[ws_r_g0 ] + _ 1 & & \{s(z)[z , z^-1 ] : s(z^-1)=z^-1s(z)}.[ws_r_g1 ] the groups of reversible lifting matrices are @xmath154 and @xmath155 . since gain scaling is not used in reversible filter bank implementations , let @xmath156 . set @xmath157 and @xmath158 with @xmath159 . the _ reversible ws group _ is the group _ defined _ by @xmath160 note that @xmath151 , which is a subgroup of the unimodular ws group , @xmath132 , consists of _ linear _ filter banks and does not include specifications for the nonlinear rounding operations involved in reversible _ implementations _ of these filter banks . [ [ reversible - ws - filter - banks - in - jpeg2000 ] ] reversible ws filter banks in jpeg 2000 + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + jpeg 2000 is more restrictive regarding user - defined reversible ws filter banks in ( * ? ? ? * annex g ) than it is regarding irreversible ones because reversible implementations may not use gain scaling operations . this means that the normalization ( [ lowpass_dc_gain ] ) must be implicit in the choice of dyadic hs lifting filters . for instance , the 5-tap/3-tap legall - tabatabai spline wavelet analysis filter bank @xcite is specified in ( * ? ? ? * annex f.4.8.1 ) and ( * ? ? ? * annex g.4.2.1 ) using the dyadic lifting factorization @xmath161[c]{,c15c , } 1 & ( 1+z^{-1})/4 \\ 0 & 1 \end{ieeeeqnarraybox*}\right ] \left [ \begin{ieeeeqnarraybox*}[{\fontsize{11}{11}\selectfont}][c]{,c15c , } 1 & 0 \\ -(z+1)/2 & 1 \end{ieeeeqnarraybox*}\right]\,,\ ] ] which satisfies ( [ lowpass_dc_gain ] ) . note that ( [ ws_one_step ] ) provides an example of a dyadic lifted ws filter bank that does _ not _ satisfy ( [ lowpass_dc_gain ] ) without a gain scaling renormalization and therefore can not be implemented reversibly within the constraints of @xcite . [ [ remarks-1 ] ] remarks + + + + + + + while one can define reversible integer - to - integer filter banks by inserting rounding operations into floating - point lifting cascades @xcite , this produces finite - precision sensitivities that could cause problems in communications applications where data is encoded on one platform and decoded on another . for this reason , prevailing practice ( e.g. , @xcite ) has been to restrict reversible implementations to dyadic filter banks . ( a side benefit is the opportunity to replace floating point arithmetic with integer arithmetic and bit - shifts . ) this creates difficulties in factorization theory , however , since @xmath162 $ ] is not a euclidean domain . this means that lifting factorization of a transfer matrix over @xmath162 $ ] may not yield lifting filters in @xmath162 $ ] . for instance , if @xmath163 and @xmath164 then the transfer matrix @xmath0 in example [ exmp : nonunique_3_factors ] of section [ sec : filter : irreducible ] has matrix components in @xmath162 $ ] , but the factorization @xmath165 does _ not _ have dyadic lifting filters even though it could have been constructed using the euclidean algorithm . concentric delay - minimized hs filter banks ( in which both impulse responses are centered at @xmath166 ) are characterized in ( * ? ? ? ( 37 ) ) by the relation @xmath167 where @xmath168[c]{,c15c , } 1 & 0 \\ 0 & -1 \end{ieeeeqnarraybox*}\right ] \quad\mbox{and}\quad \mathbf{j } \equiv \left [ \begin{ieeeeqnarraybox*}[{\fontsize{11}{11}\selectfont}][c]{,c15c , } 0 & 1 \\ 1 & 0 \end{ieeeeqnarraybox*}\right ] \,.\ ] ] unlike the case of ws filter banks , ( [ hs_symmetry ] ) does _ not _ define a matrix group , which is reflected in the fact that the base filter banks for the hs group lifting structure do not form a group . [ defn : conc_hs_class ] the _ unimodular hs class _ , @xmath169 , is the class of all real fir transfer matrices that satisfy both ( [ ds_fir_pr ] ) and ( [ hs_symmetry ] ) . [ exmp : uni_hs_lifting_structure ] _ group lifting structure for the unimodular hs class . _ let @xmath170 . define the following group of whole - sample antisymmetric real laurent polynomials : @xmath171 \,:\ , s(z^{-1})=-s(z)\right\}\,.\ ] ] let @xmath172 and @xmath173 . the set of concentric equal - length base filter banks for the unimodular hs class is @xmath174 set @xmath175 and @xmath176 . then @xmath177 is @xmath137-invariant by lemma [ lem : d - invariance ] , and the hs lifting factorization theorem ( * ? ? ? * theorem 14 ) can be stated succinctly as @xmath178 [ exmp : rev_hs_lifting_structure ] _ group lifting structure for the reversible hs class . _ as with ws filter banks , the class of _ reversible _ hs filter banks , @xmath179 , is defined in terms of lifting . define the following group of dyadic antisymmetric laurent polynomials : @xmath180 \,:\ , s(z^{-1})=-s(z)\right\}\,.\ ] ] the groups of lifting matrices are @xmath181 and @xmath182 . defining the set @xmath183 of reversible hs base filter banks is problematic . as in ( [ b_hs ] ) they must be concentric equal - length hs filter banks , but in light of the remarks at the end of example [ exmp : rev_ws_lifting_structure ] we can not simply take the set of all such filter banks with dyadic coefficients because we have no guarantee that they will have lifting factorizations over @xmath162 $ ] . instead , define @xmath183 as the set of all matrices in @xmath184 that _ have _ lifting factorizations over @xmath162 $ ] . ( any scaling matrix , @xmath185 , for which @xmath186 can be factored into dyadic lifting steps using the formulas in ( * ? ? ? * section 7.3 ) . also , note that by ( * ? ? ? * theorem 13 ) any lifting factorization of a concentric equal - length , delay - minimized hs base filter bank requires some lifting filters that are _ not _ whole - sample antisymmetric . ) let @xmath187 and set @xmath188 . the _ reversible hs class _ is defined to be @xmath189 [ exmp : elasf ] _ the elasf family of m. adams . _ the ph.d . dissertation of m. adams @xcite defines a class of reversible lifted hs filter banks referred to as the _ even - length analysis / synthesis filter _ ( elasf ) family . adams studied constraints on the rounding rules used in reversible filter bank implementations resulting from the requirement that a reversible implementation yield symmetric or antisymmetric integer subbands in symmetric pre - extension schemes @xcite . ignoring issues related to rounding rules , which are not relevant to this paper , we can construct a group lifting structure for the linear ( non - rounded ) elasf family . all elasf filter banks are lifted from the haar ( [ haar ] ) so @xmath190 and all are reversible so @xmath187 . the lifting filters for the elasf family are the same antisymmetric dyadic filters ( [ rev_wa_polys ] ) used in @xmath191 . let @xmath192 be the corresponding group lifting structure . adams also specifies the reversible lifting factorization ( [ haar_lifting_steps ] ) for the haar base filter bank , and when the first wa lifting step ( @xmath61 in ( [ lift_from_b ] ) ) is upper triangular he advocates combining it with the upper triangular step in ( [ haar_lifting_steps ] ) before the first rounding operation to minimize quantization effects ( see ( * ? ? ? * figure 3 ) ) . since we are only interested in linear ( non - rounded ) filter banks , however , we regard ( linear ) elasf filter banks as being defined by a partially factored lifting cascade ( [ lift_from_b ] ) lifted from a single base filter bank , @xmath53 , using antisymmetric dyadic lifting filters . the set of all ( linear ) elasf filter banks , @xmath193 , generated by @xmath192 is therefore a subset of the reversible hs class , @xmath179 , and every group lifting factorization in @xmath192 is also a group lifting factorization in @xmath191 . we now establish sufficient conditions on a group lifting structure to imply that all of the matrix factors ( the lifting matrices , @xmath37 , and the gain scaling and base matrices ) in irreducible group lifting factorizations are unique modulo one trivial degree of freedom . from an engineering perspective , factorization ( [ identity_lift ] ) is of no practical interest because it does nt synthesize a high - order filter bank from low - order building blocks . ( for instance , ( [ identity_lift ] ) is not a factorization one would arrive at via the euclidean algorithm . ) writing nonunique lifting factorizations as factorizations of the identity , as in ( [ identity_lift ] ) and ( [ identity_lift2 ] ) , raises an interesting point . given a lifting of the identity , if some partial product @xmath194 has nonzero order then the order of subsequent partial products must eventually _ decrease _ so that the final product ( which is @xmath195 ) has order zero . this suggests that lifting structures that only generate `` order - increasing '' cascades will also generate _ unique _ factorizations , an idea that will be made rigorous in theorem [ thm : unique_factorization ] . evidently , though , the notion of order reduction embodied in the euclidean algorithm _ is not sufficiently strong _ to imply uniqueness when factorizations are constructed via the euclidean algorithm as in @xcite . we therefore need to define an appropriate notion of `` order - increasing '' lifting cascades . the key is to focus on the _ polyphase _ orders of the intermediate transfer matrices @xmath194 rather than on the polynomial orders of the intermediate scalar filters . [ defn : orderincreasing ] a lifting cascade ( [ lift_from_b ] ) is _ strictly polyphase order - increasing _ ( usually shortened to just _ order - increasing _ ) if the order ( [ fb_order ] ) of each intermediate polyphase matrix ( [ recursive_cascade ] ) is strictly greater than the order of its predecessor : @xmath196 a group lifting structure , @xmath108 , will be called order - increasing if every irreducible cascade in is order - increasing . [ exmp : non - order - increasing ] _ ws filter bank with a * non*-order - increasing hs lifting cascade . _ define the _ causal _ ( nonunimodular ) lazy wavelet filter bank using @xmath197 and @xmath198 ; i.e. , instead of lifting from the identity as in example [ exmp : uni_ws_lifting_structure ] we are lifting from @xmath199[c]{,c15c , } 1 & 0\ieeestrut[10pt][1pt ] \\ 0 & z^{-1}\ieeestrut[10pt][1pt ] \end{ieeeeqnarraybox*}\right ] \,,\ ] ] whose order is one . lift @xmath200 to a higher - order ws filter , @xmath201 , using the hs lifting filter @xmath202 : @xmath203 the symmetry of @xmath204 is chosen to leave @xmath205 symmetric about @xmath206 , just like @xmath207 . the lifted polyphase matrix ( [ recursive_cascade ] ) is rcl ^(0)(z ) & = & + & = & , whose polyphase order is still one ( no increase ) even though the order of the highpass scalar filter increased from zero to two . despite the similarities between this example and example [ exmp : uni_ws_lifting_structure ] , we show in @xcite that the group lifting structure for the unimodular ws group , @xmath132 , defined in example [ exmp : uni_ws_lifting_structure ] _ is _ strictly polyphase order - increasing , a somewhat nontrivial task . as noted above , any lifting cascade can be simplified to irreducible form by matrix multiplication . since @xmath83 and @xmath84 are groups , every transfer matrix generated by @xmath108 has an irreducible lifting factorization in our main result shows that such factorizations are `` unique modulo rescaling '' if @xmath108 is order - increasing ( per definition [ defn : orderincreasing ] ) and @xmath86-invariant . [ thm : unique_factorization ] suppose that @xmath208 is a @xmath137-invariant , order - increasing group lifting structure . let @xmath0 be a transfer matrix generated by @xmath108 , and suppose we are given two irreducible group lifting factorizations of @xmath0 in : @xmath209 then ( [ unprimed_cascade ] ) and ( [ primed_cascade ] ) satisfy the following three properties : rcl n & = & n , [ nprimeequalsn ] + (z ) & = & _ ( z)k / k , [ defn_alpha ] + _i(z ) & = & _ _ i(z ) [ almost_unique_factors ] if , in addition , @xmath46 and @xmath210 share a nonzero matrix entry at some point @xmath211 then the factorizations ( [ unprimed_cascade ] ) and ( [ primed_cascade ] ) are _ identical ; _ i.e. , @xmath212 , @xmath213 , and @xmath214 it also follows that @xmath212 if either of the scalar filters , @xmath48 or @xmath49 , shares a nonzero value with its primed counterpart ; e.g. , if the base filter banks have equal lowpass dc responses . [ [ remarks-2 ] ] remarks + + + + + + + the type of lifting equivalence described above is the subject of the following definition . [ defn : modulorescaling ] two factorizations of @xmath0 that satisfy ( [ nprimeequalsn])([almost_unique_factors ] ) are said to be _ equivalent modulo rescaling . _ if _ all _ irreducible group lifting factorizations of @xmath0 are equivalent modulo rescaling for _ every _ @xmath0 generated by @xmath108 , we say that irreducible factorizations in @xmath108 are _ unique modulo rescaling . _ we now prove theorem [ thm : unique_factorization ] . without loss of generality , assume that @xmath215 equate ( [ unprimed_cascade ] ) and ( [ primed_cascade ] ) , and combine the gain scaling matrices on the unprimed side : l _ _ n-1(z ) _ 0(z)(z ) = + _n-1(z)_0(z)(z),[simplify_1 ] where @xmath216 and @xmath217 . move the lifting matrices @xmath37 from the left - hand side of ( [ simplify_1 ] ) to the right - hand side using ( [ intertwine_gain ] ) and ( [ gamma_ainv ] ) : l _ ( z ) = + _ ^-1_0(z)_^-1_n-1(z ) _n-1(z)_0(z)(z).[simplify_2 ] since @xmath83 and @xmath84 are @xmath86-invariant groups , the matrices @xmath218 are still in @xmath219 . simplify ( [ simplify_2 ] ) to irreducible form via matrix multiplication . since ( [ unprimed_cascade ] ) and ( [ primed_cascade ] ) are irreducible , simplification can only occur if @xmath220 and @xmath221 have the same update characteristic . if so , combine them into a single lifting matrix , and if it reduces to the identity , i.e. , if @xmath222 then combine @xmath223 and @xmath224 , etc . simplification will eventually terminate in an irreducible cascade , @xmath225 for some @xmath226 . the cascade on the right - hand side of ( [ simplify_3 ] ) is still in since @xmath227 for all @xmath125 . if @xmath228 then the order - increasing property of @xmath108 implies that a contradiction . therefore we must have @xmath229 , which means we have pairwise cancellation of all of the matrices @xmath218 and @xmath230 in the reduction of ( [ simplify_2 ] ) to irreducible form . this proves uniqueness modulo rescaling ; i.e. , conclusions ( [ nprimeequalsn ] ) , ( [ defn_alpha ] ) and ( [ almost_unique_factors ] ) . if @xmath231 , then ( [ defn_alpha ] ) implies @xmath212 and ( [ unique_factors ] ) follows . the case of @xmath232 sharing a value with its primed counterpart , e.g. , @xmath233 , is similar . a new algebraic framework , the group lifting structure , has been introduced for studying lifting factorizations of fir perfect reconstruction filter banks . the primary focus of the paper is the uniqueness of the matrix factors arising in lifting factorizations . with no constraints on the lifting process , proposition [ prop : nonuniqueness ] shows that any fir perfect reconstruction filter bank can be irreducibly lifted from any other fir perfect reconstruction filter bank in infinitely many different ways . virtually all such factorizations are mathematical pathologies , however , and are of no interest for engineering applications . to impose constraints on the universe of feasible lifting factorizations for a given class of filter banks , we define group lifting structures in terms of abelian groups of lower and upper triangular lifting matrices , an abelian group of gain scaling matrices , and sets ( not necessarily groups ) of base filter banks . if a group lifting structure satisfies a polyphase order - increasing hypothesis and if the groups of lifting matrices are invariant under the action of the gain scaling group , the main result of the paper , theorem [ thm : unique_factorization ] , shows that irreducible group lifting factorizations are unique modulo , at most , rescaling . theorem [ thm : unique_factorization ] is used in @xcite to prove uniqueness results for both reversible and irreversible lifting factorizations of ws and hs filter banks parameterized by the linear phase group lifting structures defined in section [ sec : linearphase ] of the present paper . this scope includes the specification of ws filter banks in jpeg 2000 and m. adams elasf class of reversible hs filter banks . the key to these proofs is establishing the order - increasing property for the associated group lifting structures . i. c. daubechies , _ ten lectures on wavelets _ , ser . cbms - nsf regional conf . series in appl . math . , ( univ . mass.lowell , june 1990).1em plus 0.5em minus 0.4emphiladelphia : soc . indust . , 1992 , no . 61 . c. m. brislawn , `` group lifting structures for multirate filter banks , ii : linear phase filter banks , '' los alamos national lab , tech . laur-09 - 3049 , may 2009 , to appear in _ ieee trans . signal process . _ m. maslen and p. abbott , `` automation of the lifting factorisation of wavelet transforms , '' _ computer physics communications _ , 127 , no . 2 - 3 , pp . 309326 , 2000 . [ online ] . available : http://www.sciencedirect.com / science / article / b6tj5 - 40fgct9-g/2/502cc126% fa01e35997405bd199000fa4[http://www.sciencedirect.com / science / article / b6tj5 - 40fgct9-g/2/502cc126% fa01e35997405bd199000fa4 ] r. foote , g. mirchandani , d. rockmore , d. healy , and t. olson , `` a wreath product group approach to signal and image processing part i : multiresolution analysis , '' _ ieee trans . signal process . _ , vol . 48 , no . 1 , 102132 , jan . 2000 . g. mirchandani , r. foote , d. rockmore , d. healy , and t. olson , `` a wreath product group approach to signal and image processing part ii : convolution , correlation , and applications , '' _ ieee trans . signal process . _ , vol . 48 , no . 3 , pp . 749767 , mar . 2000 . a. zandi , j. d. allen , e. l. schwartz , and m. boliek , `` compression with reversible embedded wavelets , '' in _ proc . data compress . _ 1em plus 0.5em minus 0.4emsnowbird , ut : ieee computer soc . , mar . 1995 , pp . 212221 . d. legall and a. tabatabai , `` subband coding of digital images using symmetric short kernel filters and arithmetic coding techniques , '' in _ intl . conf . speech , signal process.__1em plus 0.5em minus 0.4emnew york city : ieee signal process . soc . , apr . 1988 , pp . 761764 . christopher m. brislawn ( m91sm05 ) received the b.s . degree in 1982 from harvey mudd college , claremont , ca , and the ph.d . degree in 1988 from the university of colorado boulder , both in mathematics . he was a visiting assistant professor at the university of southern california from 1989 to 1990 . he joined los alamos national laboratory ( lanl ) , los alamos , nm , as a postdoc in 1990 and became a permanent staff member in 1993 . currently he is a scientist in group ccs-3 of the computer , computational and statistical sciences division . his research interests include wavelet transforms , digital filter banks , communications coding , and statistical signal processing . from 1991 to 1993 he coauthored the wavelet / scalar quantization specification for compression of digitized fingerprint images with the u.s . federal bureau of investigation . from 1999 to 2003 he served as lanl s principal member in working group l3.2 of the international committee for information technology standards and led a lanl team that participated in writing the iso / iec jpeg 2000 standard ( iso 15444-x ) . his team worked on parts 1 and 2 of the standard , and he coauthored the proposal to create jpeg 2000 part 10 ( extensions for three - dimensional data ) , serving as the first editor of part 10 . in 20072008 he represented lanl on the motion imagery standards board for the national geospatial intelligence agency . in addition to his technical work , dr . brislawn has mentored 15 graduate students and 4 postdocs at lanl and has co - supervised one ph.d . dissertation for the university of texas austin .

external gravitational fields induce phase factors in the wave functions of particles . the phases are exact to first order in the background gravitational field , are manifestly covariant and gauge invariant and provide a useful tool for the study of spin - gravity coupling and of the optics of particles in gravitational or inertial fields . we discuss the role that spin - gravity coupling plays in particular problems .

0709.0819

the study of the interaction of spin with inertia and gravity has received a strong impulse from the work of bahram mashhoon @xcite . his work has stimulated the research on which we report below . covariant wave equations for scalar and vector bosons , for spin-1/2 fermions @xcite and spin-2 @xcite particles can be solved exactly to first order in the metric deviation . the background gravitational and inertial fields appear in the solutions as phase factors multiplying the wave function of the corresponding field - free equations . the phases can be calculated with ease for most metrics . we summarize the solutions for vector , tensor bosons and fermions in sections ii , iii and iv . in the same sections we also extract the spin - gravity interaction @xcite from the gravity - induced phases . the optics of the particles is derived in sections v. in sections vi - viii we discuss the relevance of the mashhoon coupling to muon @xmath0 experiments , discrete symmetries and neutrino helicity transitions . the conclusions are contained in section ix . photons in gravitational fields are described by the maxwell - de rahm equations @xcite @xmath1 which reduce to maxwell equations @xmath2 when @xmath3 , or when , as in lensing , the wavelength @xmath4 of @xmath5 is much smaller than the typical radius of curvature of the gravitational background . in ( [ a ] ) and ( [ b ] ) @xmath6 indicates covariant differentiation . we use units @xmath7 . in what follows we consider only ( [ b ] ) and its generalization to massive , charge - less , spin-1 particles @xmath8 in the hope that among the atoms and molecules to be used in interferometry some indeed satisfy ( [ c ] ) @xcite . following previous work @xcite-@xcite , we show below that equations ( [ b ] ) and ( [ c ] ) can be solved exactly to first order in the metric deviation @xmath9 , where @xmath10 and @xmath11 is the minkowski metric of signature -2 . to first order in @xmath12 , ( [ b ] ) and ( [ c ] ) become @xmath13 where ordinary differentiation of a quantity @xmath14 is equivalently indicated by @xmath15 or @xmath16 . in deriving ( [ d ] ) and ( [ e ] ) , we have used the lanczos - de donder gauge condition @xmath17 in the massless case , the field @xmath18 satisfies the condition @xmath19 it is convenient to impose ( [ g ] ) also in the case of a massive particle . equations ( [ b ] ) and ( [ c ] ) can be handled simultaneously . their solution is @xmath20 \nonumber \\ & & + \frac{1}{2}\int_{p}^{x}dz^{\lambda } \gamma_{\alpha\lambda}(z ) \partial^{\alpha}a_{\mu}(x)\nonumber \\ & & -\frac{1}{2}\int_{p}^{x}dz^{\lambda}(\gamma_{\mu\lambda , \sigma}(z)- \gamma_{\sigma\mu,\lambda}(z)-\gamma_{\sigma\lambda,\mu}(z ) ) a^{\sigma}(x)\ , , \label{h}\end{aligned}\ ] ] where @xmath21 satisfies the equation @xmath22 in the case of ( [ d ] ) , and @xmath23 when ( [ h ] ) is a solution of ( [ e ] ) . in ( [ h ] ) @xmath24 is a fixed reference point and @xmath25 a generic point along the particle s worldline . we can prove that ( [ h ] ) is an exact solution to first order in @xmath12 by straightforward differentiation . the first two integrals in ( [ h ] ) represent by themselves a solution of the klein - gordon equation @xmath26 . the additional terms are related to spin . in fact ( [ h ] ) can be re - written in the form @xmath27 @xmath28 where @xmath29 the rotation matrices @xmath30 satisfy the commutation relation @xmath31=i\epsilon_{ijk}s_k$ ] . by applying stokes theorem to the r.h.s . of ( [ h ] ) we find @xmath32 where @xmath33 is the total angular momentum of the spin-1 particle , @xmath34 is the linearized riemann tensor , and @xmath35 is the surface bound by the closed path along which the integration is performed . the weak field approximation @xmath36 does not fix the reference frame completely . the transformations of coordinates @xmath37 , with @xmath38 also small of first order , are still allowed and lead to the `` gauge '' transformations @xmath39 . equation ( [ n ] ) therefore indicates that solution ( [ h ] ) is covariant and also gauge invariant . it also follows from ( [ n ] ) that the term containing @xmath40 in ( [ l ] ) does not contribute to integrations over closed paths , behaves as a gauge term and may therefore be dropped . the spin - gravity coupling is contained in the third term on the r.h.s . of ( [ l ] ) . its time integral part is @xmath41 . since for rotation @xmath42 , one gets @xmath43 , where @xmath44 . in general , one may write @xmath45 , which must now be applied to a solution of the field free equations . one finds @xmath46 and , for particles polarized parallel or antiparallel to @xmath47 , @xmath48 , as in @xcite.@xmath49 is the energy observed by the co - rotating observer . for spin-@xmath50 fields , the simplest equation of propagation is derived in @xcite and is given by @xmath51 in lensing , the second term in ( [ mtw ] ) may be neglected when the wavelength @xmath52 associated with @xmath53 is smaller than the typical radius of curvature of the gravitational background @xcite . we consider here massless or massive spin-@xmath50 particles described by the equation @xcite @xmath54 to first order in @xmath55 , ( [ spin2 ] ) can be written in the form @xmath56 where @xmath57 is the linearized ricci tensor of the background metric and @xmath58 is the corresponding christoffel symbol of the first kind . it is easy to prove , by direct substitution , that a solution of ( [ wf ] ) , exact to first order in @xmath55 , is represented by @xmath59 \\ \nonumber + \frac{1}{2 } \int_{p}^{x}dz^{\lambda}\gamma_{\alpha\lambda}\left(z\right)\partial^{\alpha}\phi_{\mu\nu}\left(x\right)+ \int_{p}^{x}dz^{\lambda}\gamma_{\mu\lambda,\sigma}\left(z\right)\phi_{\nu}^{\sigma}\left(x\right)+ \int_{p}^{x}dz^{\lambda}\gamma_{\nu\lambda,\sigma}\left(z\right ) \phi_{\mu}^{\sigma}\left(x\right),\end{aligned}\ ] ] where @xmath60 satisfies the field - free equation @xmath61 and the gauge condition ( [ f ] ) has been used . equation ( [ solution ] ) can be written in the form @xmath62\phi_{\mu\nu}+ \\ \nonumber - \frac{i}{2}\int_{p}^{x}dz^{\lambda}\left(\gamma_{\alpha\lambda,\beta}\left(z\right)-\gamma_{\beta\lambda,\alpha}\right)s^{\alpha\beta}\phi_{\mu\nu } \left(x\right ) - \frac{i}{2 } \int_{p}^{x}dz^{\lambda}\gamma_{\beta\sigma,\lambda}\left(z\right)t^{\beta\sigma}\phi_{\mu\nu}\left(x\right)\,,\end{aligned}\ ] ] where @xmath63 from @xmath64 one constructs the rotation matrices @xmath65 that satisfy the commutation relations @xmath66=i \epsilon_{ijk}s_{k}$ ] . the spin - gravity interaction is therefore contained in the term @xmath67.\ ] ] the solution ( [ solution ] ) is invariant under the gauge transformations @xmath68 . if , in fact , we choose a closed integration path @xmath69 , stokes theorem transforms the first three integrals of ( [ sol ] ) into the gauge invariant result @xmath70 where @xmath71 is the surface bound by @xmath72 , and @xmath73 is the total angular momentum of the particle . for the same path @xmath69 the integral involving @xmath74 in ( [ sol ] ) it behaves like a gauge term and is therefore dropped . the helicity - rotation coupling for massless , or massive spin-@xmath50 particles follows immediately from the @xmath75 term in ( [ sol ] ) . in fact , the particle energy is changed by virtue of its spin by an amount given by the time integral of this spin term @xmath76 that must then be applied to a solution of ( [ ff ] ) . for rotation about the @xmath77-axis , @xmath78 , we find @xmath79 and the energy of the particle therefore changes by @xmath80 , where the factor @xmath81 refers to the particle s helicity , as discussed by ramos and mashhoon @xcite . equation ( [ hrc ] ) extends their result to any weak gravitational , or inertial field . the effect of ( [ sg ] ) on @xmath82 can be easily seen in the case of a gravitational wave propagating in the @xmath83-direction and represented by the components @xmath84 $ ] and @xmath85 $ ] . for an observer rotating about the @xmath86-axis the metric is @xmath87 . then the two independent polarizations @xmath88 and @xmath89 are transformed by @xmath90 into @xmath91 and @xmath92 . for closed integration paths and vanishing spin , ( [ sol ] ) coincides with the solution of a scalar particle in a gravitational field , as expected . this proves the frequently quoted statement @xcite that gravitational radiation propagating in a gravitational background is affected by gravitation in the same way that electromagnetic radiation is ( when the photon spin is neglected ) . the behavior of spin-1/2 particles in the presence of a gravitational field @xmath93 is determined by the covariant dirac equation @xmath94\psi(x)=0\,,\ ] ] where @xmath95 , @xmath96 is the spin connection and the matrices @xmath97 satisfy the relations @xmath98 . both @xmath99 and @xmath100 can be obtained from the usual constant dirac matrices by using the vierbein fields @xmath101 and the relations @xmath102 where @xmath103 $ ] . equation ( [ diracequation ] ) can be solved exactly to first order in @xmath104 . this is achieved by first transforming ( [ diracequation ] ) into the equation @xcite,@xcite,@xcite @xmath105{\tilde \psi}(x)=0\,,\ ] ] where @xmath106 by multiplying ( [ diraceqtrasf ] ) on the left by @xmath107 , we obtain the equation @xmath108 whose solution @xmath109 is exact to first order . the operator @xmath110 is defined as @xmath111\hat{l}^{\alpha\beta}(z)+ \frac{1}{2}\int_p^x dz^\lambda\gamma_{\alpha\lambda}\hat{k}^\alpha\,,\ ] ] @xmath112=\left((x^\alpha - z^\alpha)\hat{k}^\beta- ( x^\beta - z^\beta)\hat{k}^\alpha\right)\psi_0(x)\ , , \qquad [ \hat{k}^\alpha , \psi_0(x)]=i\partial^\alpha\psi_0\,,\ ] ] and @xmath113 satisfies the usual flat space - time dirac equation . @xmath114 and @xmath115 are the angular and linear momentum operators of the particle . it follows from ( [ exactsolution ] ) and ( [ psitilde ] ) that the solution of ( [ diracequation ] ) can be written in the form @xmath116 and also as @xmath117 where @xmath118 is of first order in @xmath119 . the factor @xmath120 on the r.h.s . of ( [ psisolution2 ] ) is required by the condition that both sides of the equation agree when the gravitational field vanishes . it is useful to re - derive some known results from the covariant dirac equation . on multiplying ( [ diracequation ] ) on the left by @xmath121 and using the relations @xmath122=-\frac{1}{4}\sigma^{\alpha\beta}(x)r_{\alpha\beta\mu\nu}\,,\ ] ] and @xmath123=-\frac{i}{4}\ , \sigma^{\alpha\beta}(x)r_{\alpha\beta\mu\nu}\,,\ ] ] we obtain the equation @xmath124 in ( [ relation2 ] ) and ( [ kgeq+r ] ) @xmath125 $ ] and @xmath126 is the ricci scalar . on applying stokes theorem to a closed space - time path @xmath127 and using ( [ relation1 ] ) , we find that @xmath128 changes by @xmath129 where @xmath130 is the total momentum of the particle . equation ( [ phase ] ) shows that ( [ psisolution2 ] ) is gauge invariant . the spin - rotation coupling derived by mashhoon by extending the hypothesis of locality can be now derived rigorously from the solution found . choose a cylindrical coordinate ( @xmath131 , @xmath132 , @xmath133 , @xmath134 ) for an inertial frame @xmath135 . an observer at rest in a frame @xmath136 rotating with a constant angular velocity @xmath137 relative to @xmath135 will follow the world line @xmath138 . consider an orthogonal tetrad consisting of the observer s four - velocity @xmath139 and the triad @xmath140 ( @xmath141 ) normal to the world line . by using the local tetrad @xcite @xmath142 , where @xmath143 , one can construct a vierbein field @xmath144 along the world line of the observer @xmath145 . it is then easy to calculate the spinor connection @xmath146 . in calculating the energy , only the component @xmath147 is necessary . by using the dirac representation for the @xmath148-matrices , one obtains @xmath149 and from @xmath150 also @xmath151 , where @xmath152 has the usual plane wave form . besides the contribution due to the coupling of the orbital angular momentum to rotation , which gives the sagnac effect @xcite , one obtains the spin - rotation coupling @xmath153 , and also , for spin polarizations parallel or antiparallel to the direction of rotation , one obtains , @xmath154 , as shown by mashhoon . the present result is exact and follows from the general form of the solution ( [ psisolution2 ] ) . it also agrees with those of hehl and ni @xcite and @xcite . the mashhoon effect is obviously a prime candidate for experiments with accelerators and will be discussed at length below . according to ( [ phase ] ) , both angular momentum and spin couple to a weak gravitational field in the same way . this confirms that , unlike the electromagnetic case , the gyro - gravitational ratio of a spin-1/2 particle is 1 , as shown in @xcite . a classical charge @xmath155 moving in a circle with angular momentum @xmath156 forms a current loop of magnetic moment @xmath157 , which gives the gyromagnetic factor @xmath158 . the magnetic moment of a charged particle depends therefore on the ratio @xmath159 and , for a rotating object , on the space distributions of charge and mass . for a quantum particle , the dirac equation indicates that @xmath160 . the corrections to @xmath160 come from quantum electrodynamics where the electron can be pictured at any instant as a bare particle in interaction with a cloud of virtual photons . qualitatively , if the charge remains associated with the electron , part of the mass energy is carried by the photon cloud resulting in a slight increase for the value @xmath161 of the electron itself . in the gravitational case , however , the gyro - gravitational ratio of the spin-1/2 particle is @xmath162 . this suggests , according to @xcite , that the internal distributions of the gravitational mass , associated with the interaction , and of the inertial mass , associated with the angular momentum , equal each other . in the geometrical optics approximation , valid whenever @xmath163 , where @xmath164 is the momentum of the particle , the interaction between the angular momentum of the source and the particle s spin vanishes . this interaction is quantum mechanical in origin . then the geometrical phase @xmath165 is sufficient to reproduce the classical angle of deflection @xcite , as it should , because ( [ phig ] ) coincides with the first two integrals in ( [ h ] ) and ( [ sol ] ) . we can therefore treat photons , gravitons and fermions simultaneously when spin is neglected . more detailed calculations involving neutrinos are given in @xcite . if , e.g. , we choose a gravitational background represented by the lense - thirring metric @xcite , @xmath166 , @xmath167 , @xmath168 , and @xmath169 , with @xmath170 , @xmath171 and @xmath172 is related to the angular momentum of the gravitational source . in particular , if the source rotates with angular velocity @xmath173 , then @xmath174 , @xmath175 . without loss of generality , we assume that the particles are massless and propagate along the @xmath134-direction , hence @xmath176 . using plane waves for the field free solution , the phase of the wave equation becomes @xmath177 + \frac{1}{2}\int_p^x dz^\lambda \gamma_{\alpha\lambda}(z ) k^\alpha\,.\ ] ] we can define the particle momentum as @xmath178 it is easy to show that @xmath179 satisfies the eikonal equation @xmath180 . for the lense - thirring metric , @xmath181 is given by @xmath182dz'\right]+k\int_p^q dz ' \phi \\ & & -\frac{k}{2}\int_p^q \left[\left ( ( x - x')(h_{1 , \,z'}-2h_{3,\ , x'})+(y - y')(h_{2 , \,z'}-2h_{3,\ , y'})\right)dz ' \right . \nonumber \\ & & \left . \quad -\left((x - x')h_{1,\ , x'}+(y - y')h_{1,\ , y'}\right)dx ' -\left((x - x')h_{2,\ , x'}+(y - y')h_{2,\ , y'}\right)dy'\right ] \nonumber \\ & & + \frac{k}{2}\int_p^q\left[2h_3dz'+h_1dx'+h_2dy'\right]\,,\end{aligned}\ ] ] where @xmath183 is the point at which the particles are generated , and @xmath184 is a generic point along their space - time trajectory . the components of the momentum are therefore @xmath185 we then have @xmath186 where @xmath187 is the component of the momentum orthogonal to the direction of propagation of the particles . since only phase differences are physical , it is convenient to choose the space - time path by placing the particle source at distances that are very large relative to the dimensions of the lens , and the generic point is located along the @xmath134 direction . we therefore replace @xmath184 with @xmath134 , where @xmath188 . using the expression for @xmath189 we find that their contribution is negligible and ( [ k1])-([k3 ] ) simplify to @xmath190 from ( [ k1bis])-([k3bis ] ) we can determine the deflection angle @xmath133 . let us analyze the case of non - rotating lenses , i.e. @xmath191 . we get @xmath192 where @xmath193 . by defining the deflection angle as @xmath194 it follows that @xmath195 in the limit @xmath196 we obtain the usual einstein result @xmath197 a general expression for the index of refraction @xmath198 can also be derived from ( [ gph]),([photonmomentum ] ) and @xmath199 . we now consider the propagation of light and gravitational waves in a background metric represented by @xmath200 , @xmath201 , where @xmath202 is the gravitational potential of the lens and @xmath132 the distance from @xmath203 to the particle . wave optics effects can be seen by using the type of double slit arrangement indicated in fig.1 . we will also use a solution @xmath204 in the form of a plane wave @xmath205 and neglect spin effects . this limits the calculation @xcite of the phase difference to the first two terms in ( [ 9 ] ) and ( [ 10 ] ) . we also assume for simplicity that @xmath206 , so that propagation is entirely in the @xmath207)-plane and the set - up is planar . the corresponding wave amplitude @xmath208 is therefore @xmath209 + \frac{1}{2}\left[\int_s^x dz^0 \gamma_{00 } \pi^0+\int_s^x dz^2 \gamma_{22 } \pi^2+\int_s^x dz^{3}\gamma_{33}\pi^{3}\right ] \right\}\ , , \label{*}\end{aligned}\ ] ] where @xmath210 , @xmath211 , and we have taken into account the fact that @xmath212 plays no role in the planar arrangement chosen . the phase must now be calculated along the different paths sp+po and sl+lo taking into account the values of @xmath213 in the various intervals . geometry of a two - image gravitational lens . the solid lines represent the particle paths between the particle source at @xmath214 and the observer at @xmath215 . @xmath216 is the spherically symmetric gravitational lens . @xmath217 and the particle paths lie in the same plane.,scaledwidth=90.0% ] the total change in phase is @xcite @xmath218 all integrations in ( [ * ] ) can be performed exactly and the results can be expressed in terms of physical variables @xmath219 , @xmath220 , @xmath221 , @xmath222 , and @xmath223 or lensing variables @xmath224 , @xmath225 , @xmath226 , @xmath227 , @xmath228 , and @xmath229 . we find @xmath230 where @xmath231 , @xmath232 , @xmath233 and @xmath234 . as an example , let us consider the simpler case @xmath235 , from which we obtain ( see fig.1 ) @xmath236 . a simple calculation shows that the probability density of finding a photon , or graviton , at @xmath237 is typically @xmath238 measurements of the interaction of spin with rotation have been carried out in the case of photons using signals from global positioning system satellites @xcite and data published in @xcite can be re - interpreted @xcite as due to the coupling of earth s rotation to the nuclear spins in mercury . the spin - rotation effect is also consistent with a small depolarization of electrons in storage rings @xcite . we show below that the same coupling is of particular interest in experiments with storage rings . it is essential in getting the correct @xmath240-dependence in @xmath241 experiments @xcite and eliminates the need of _ ad hoc _ , phenomenological arguments . the effect is conceptually important . it extends to the quantum level the classical coupling of rotation to the intrinsic angular momentum of a body and simplifies the treatment of rotational inertia . it also yields different potentials for different particles and for different spin states and can not , therefore , be considered universal @xcite . before discussing its connection with @xmath0 experiments , it is useful to briefly recall the usual experimental setup . the experiment @xcite involves muons in a storage ring a few meters in diameter , in a uniform vertical magnetic field . muons on equilibrium orbits within a small fraction of the maximum momentum are almost completely polarized with spin vectors pointing in the direction of motion . as the muons decay , those electrons projected forward in the muon rest frame are detected around the ring . their angular distribution thence reflects the precession of the muon spin along the cyclotron orbits . the calculations are performed in the rotating frame of the muon and do not therefore require a relativistic treatment of inertial spin effects @xcite . then the vierbein formalism yields @xmath242 and @xmath243 where @xmath244 and @xmath245 are the three - acceleration and three - rotation of the observer and , in the chiral representation of the usual dirac matrices , @xmath246=i \left(\matrix { \sigma^i & 0 \cr 0 & -\sigma^i \cr } \right)\,.\ ] ] the second term in ( [ 1 ] ) represents the mashhoon term . the first term drops out . in fact , by symmetrization we obtain @xmath247+\frac{i\hbar}{2}\left(\vec{a}\cdot \vec{\alpha}\right).\ ] ] the last term in ( [ simp ] ) and the first term in ( [ 1 ] ) therefore cancel each other . the remaining first order contributions in @xmath244 and @xmath245 to the dirac hamiltonian add up to @xcite @xmath248 \\ & & -{\vec \omega}\cdot \left({\vec l}+\frac{{\vec \sigma}}{2}\right)\ , . \nonumber\end{aligned}\ ] ] for simplicity all quantities in @xmath249 are taken to be time - independent . they are referred to a left - handed tern of axes rotating about the @xmath250-axis in the clockwise direction of motion of the muons . the @xmath251-axis is tangent to the orbits and in the direction of the muon momentum . the magnetic field is @xmath252 . only the mashhoon term and the magnetic moment interaction then couple the helicity states of the muon . the remaining terms contribute to the overall energy @xmath253 of the states , and we indicate by @xmath254 the corresponding part of the hamiltonian . before decay the muon states can be represented as @xmath255 where @xmath256 and @xmath257 are the right and left helicity states of the hamiltonian @xmath254 and satisfy the equation @xmath258 the total hamiltonian reduces effectively to @xmath259 , where @xmath260 @xmath261 represents the total magnetic moment of the muon and @xmath262 is the bohr magneton . electric fields used to stabilize the orbits and stray radial electric fields can also affect the muon spin . their effects can however be cancelled by choosing an appropriate muon momentum @xcite and will not be considered in what follows . the coefficients @xmath263 and @xmath264 in ( [ 2 ] ) evolve in time according to @xmath265 where @xmath216 is the matrix @xmath266\ ] ] and @xmath267 represents the width of the muon . the non - diagonal form of @xmath216 ( when @xmath268 ) implies that rotation does not couple universally to matter . @xmath216 has eigenvalues @xmath269 and eigenstates @xmath270\ , , \nonumber \\ |\psi_2 > & = & \frac{1}{\sqrt{2}}\,\left[-i|\psi_+>+|\psi_->\right ] \ , . \nonumber\end{aligned}\ ] ] the muon states that satisfy ( [ 4 ] ) , and the condition @xmath271 at @xmath272 , are @xmath273|\psi_+ > \right . \\ & & \left . + \left[e^{-i{\tilde \omega } t } + e^{i{\tilde \omega } t}\right]|\psi_- > \right\ } \ , , \nonumber\end{aligned}\ ] ] where @xmath274 the spin - flip probability therefore is @xmath275\ , . \nonumber\end{aligned}\ ] ] the @xmath267-term in ( [ 7 ] ) accounts for the observed exponential decrease in electron counts due to the loss of muons by radioactive decay @xcite . the spin - rotation contribution to @xmath276 is represented by @xmath277 which is the cyclotron angular velocity @xmath278 @xcite . the spin - flip angular frequency is then @xmath279 which is precisely the observed modulation frequency of the electron counts @xcite . this result is independent of the value of the anomalous magnetic moment of the particle . the cancellation of the dirac value of the magnetic moment contribution by the mashhoon term must therefore take place for all spin-1/2 particles in a similar physical set - up . hence , it is the spin - rotation coupling that generates the correct @xmath280 factor in @xmath137 by exactly cancelling , in @xmath281 , the much larger contribution @xmath262 that fermions with no anomalous magnetic moment produce . the cancellation is made possible by the non - diagonal form of @xmath216 and is therefore a direct consequence of the violation of the equivalence principle . it has , of course , been argued that this principle does not hold true in the quantum world @xcite . recently , discrepancies between the experimental and standard model values of @xmath282 have been observed with very high accuracy @xcite . the most precise data yet give @xmath283 for the negative muon @xcite , and @xmath284 for the positive muon @xcite . this discrepancy can be used to set upper limits on @xmath183 and @xmath285 invariance violations in spin - rotation coupling @xcite . the possibility that discrete symmetries in gravitation be not conserved has been discussed in the literature . attention has in general focused on the potential @xmath286,\ ] ] which applies to a particle of generic spin @xmath287 . the first term , introduced by leitner and okubo @xcite , violates the conservation of @xmath183 and @xmath285 . the same authors determined the upper limit @xmath288 from the hyperfine splitting of the ground state of hydrogen . the upper limit @xmath289 was determined in @xcite from sn 1987a data . the corresponding potential violates the conservation of @xmath183 and @xmath127 . conservation of @xmath127 and @xmath285 is violated by the last term , while ( [ 1a ] ) , as a whole , conserves @xmath290 . there is , as yet , no upper limit on @xmath291 . these studies are extended here to the mashhoon term . before decay , the muon states can be represented as in ( [ 2 ] ) where @xmath292 and @xmath293 again are the right and left helicity states of the hamiltonian @xmath294 defined in the previous section . assume now that the coupling of rotation to @xmath292 differs in strength from that to @xmath293 . then the mashhoon term can be modified by means of a matrix @xmath295 that reflects the different coupling of rotation to the two helicity states . the total hamiltonian now is @xmath296 , where @xmath297 a violation of @xmath183 and @xmath285 in ( [ 4a ] ) would arise through @xmath298 . the constants @xmath299 and @xmath300 are assumed to differ from unity by small amounts @xmath301 and @xmath302 . the coefficients @xmath263 and @xmath264 in ( [ 2 ] ) evolve in time according to @xmath303 where @xmath304 and @xmath267 represents , as before , the width of the muon . the spin - rotation term is off - diagonal in ( [ 6b ] ) and does not therefore couple to matter universally . it violates hermiticity as shown in @xcite and , in a general way , by scolarici and solombrino @xcite . it also violates @xmath285 , @xmath183 and @xmath305 , while nothing can be said about @xmath290 conservation which requires @xmath306 to be hermitian . because of the non - hermitian nature of ( [ 4a ] ) , one expects @xmath267 itself to be non - hermitian . the resulting corrections to the width of the muon are , however , of second order in the @xmath307 s and are neglected . @xmath216 has eigenvalues @xmath308 where @xmath309 and eigenstates @xmath310 then the muon states ( [ 2 ] ) are @xmath311,\end{aligned}\ ] ] where the condition @xmath312 has been applied . the spin - flip probability is therefore @xmath313.\end{aligned}\ ] ] when @xmath314 , ( [ 9 ] ) yields @xcite @xmath315,\ ] ] that provides the appropriate description of the spin - rotation contribution to the spin - flip transition probability . notice that the case @xmath316 ( no spin - rotation coupling ) yields @xmath317\ ] ] and does not therefore agree with the results of the @xmath280 experiments . hence the necessity of accounting for spin - rotation coupling whose contribution cancels the factor @xmath318 in ( [ a1 ] ) . substituting @xmath319 into ( [ 9 ] ) , one finds @xmath320.\ ] ] one may attribute the discrepancy between @xmath321 and @xmath322 to a violation of the conservation of the discrete symmetries by the spin - rotation coupling term in ( [ 4 ] ) . the upper limit on the violation of @xmath323 and @xmath305 is derived from ( [ 11 ] ) assuming that the deviation from the current value of @xmath322 is wholly due to @xmath324 . the upper limit is therefore @xmath325 in the case of negative muons @xcite and of @xmath326 for positive muons @xcite . at the same time the two values of @xmath327 can be thought of as due to a different coupling strength between rotational inertia and the two helicity states of the muon . then the values of @xmath328 and @xmath329 can be determined from @xmath330 in ( [ 9 ] ) according to the equations @xmath331 and @xmath332 equations ( [ 11a ] ) and ( [ 11b ] ) have the approximate solutions @xmath333 and @xmath334 more precise , numerical solutions give @xmath335 and @xmath336 . these values are significant in view of the precision with which @xmath337 have been determined . it then follows that the coupling of rotation to positive helicity is larger than that to negative helicity , which agrees with the value @xmath338 for both @xmath339 and @xmath340 . this also means that the violation of @xmath24 and @xmath341 is relatively stronger for positive helicity at a level @xmath342 and that the spin - rotation interaction is an inherent source of @xmath24 and @xmath285 violation . in this section , it is convenient to write the left and right neutrino wave functions in the form @xmath343 where @xmath344 represents the pauli matrices , @xmath345 are eigenvectors of @xmath346 corresponding to negative and positive helicity and @xmath347 , @xmath348 . this notation already takes into account the fact that if @xmath349 are the helicity states , then we have @xmath350 for relativistic neutrinos . in general , the spin precesses during the motion of the neutrino . this can be seen , for instance , from the contribution @xmath150 in @xmath351 . the expectation value of the contribution of @xmath352 to the effective mechanical momentum can in fact be re - written in the form @xmath353 where @xmath354 . equation ( [ omega1 ] ) represents the spin - rotation coupling for the lense - thirring metric . here rotation is provided by the gravitational source , rather than by the particles themselves . we now study the helicity flip of one flavor neutrinos as they propagate in the gravitational field produced by a rotating mass @xcite . the neutrino state vector can be written as @xmath355 where @xmath356 and @xmath357 is an affine parameter along the world - line . in order to determine @xmath358 and @xmath359 , we can write ( [ psisolution ] ) as @xmath360 where @xmath361 and @xmath362 is the corresponding solution in minkowski space - time . the latter can be written as @xmath363\,.\ ] ] strictly speaking , @xmath364 should also be normalized . however , it can be shown @xcite that @xmath365 is already of @xmath366 , can only produce higher order terms and is therefore unnecessary in this calculation . from ( [ neutrinorl ] ) , ( [ psisolution3 ] ) and ( [ psizero ] ) we obtain @xmath367 an equation for @xmath368 can be derived in an entirely similar way . if we consider neutrinos which are created in the left - handed state , then @xmath369 , and we obtain @xmath370 where @xmath371 . as remarked in @xcite , @xmath372 need not be a null vector if we assume that the neutrino moves along an `` average '' trajectory . we also find , to lowest order , @xmath373 where @xmath374 are the usual christoffel symbols , and @xmath375.\ ] ] in what follows , we compute the probability amplitude ( [ rtl ] ) for neutrinos propagating along the @xmath134 and the @xmath83 directions explicitly . for propagation along the @xmath134-axis , we have @xmath376 and @xmath377 . as in section iii , we choose @xmath378 . we get @xmath379 summing up , and neglecting terms of @xmath380 , ( [ rtl ] ) becomes @xmath381 the contributions to @xmath382 vanish . as a consequence @xmath383 and the probability amplitude for the @xmath384 transition is of @xmath385 , as expected . integrating ( [ dadz ] ) from @xmath386 to @xmath134 , yields @xmath387\\ & = & \frac{m}{e}\frac{gm}{2b}\left[1+\frac{z}{r}-i\frac{2\omega r^2b^2}{5r^3}\right]\,.\nonumber\end{aligned}\ ] ] it also follows that @xmath388\,.\ ] ] the first of the two terms in ( [ |alfa - z| ] ) comes from the mass of the gravitational source . the second from the source s angular momentum and vanishes for @xmath389 because the contribution from @xmath390 to 0 exactly cancels that from 0 to @xmath391 . in fact , if we consider neutrinos propagating from 0 to @xmath391 , we obtain @xmath392\,.\ ] ] according to semiclassical spin precession equations @xcite , there should be no spin motion when spin and @xmath393 are parallel as in the present case . this is a hint that rotation of the source , rather than of the particles , should produce a similar effect . the probabilities ( [ |alfa - z| ] ) and ( [ alfa infinito ] ) mark therefore a departure from expected results . they are however small of second order . both expressions vanish for @xmath394 , as it should for a stationary metric . in this case , in fact , helicity is conserved @xcite . it is interesting to observe that spin precession also occurs when @xmath395 vanishes @xcite . in the case of ( [ |alfa - z| ] ) the mass contribution is larger when @xmath396 , which , close to the source , with @xmath397 , becomes @xmath398 and is always satisfied . in the case described by ( [ alfa infinito ] ) , the rotational contribution is larger if @xmath399 which effectively restricts the region of dominance to a narrow strip about the @xmath400-axis in the equatorial plane , if the source is compact and @xmath395 is relatively large . in this case , we take @xmath376 , @xmath401 . as in section iii , the calculation can be simplified by assuming that the motion is in the equatorial plane with @xmath402 , @xmath403 . we then have @xmath404 summing up , and neglecting terms of @xmath380 , ( [ rtl ] ) becomes @xmath405 the contributions to @xmath382 again vanish and we get @xmath406\sim{\cal o}(m / e)\,.\ ] ] integrating ( [ dadx ] ) from @xmath386 to @xmath83 , we obtain @xmath407 and @xmath408 obviously , the mass contribution is the same as for propagation along the @xmath134-axis . however , the two cases differ substantially in the behavior of the angular momentum term . in this case , in fact , this term is even , so it does not vanish for @xmath389 . if we consider neutrinos generated at @xmath409 and propagating to @xmath410 , we find @xmath411 the mass term is larger when @xmath412 . at the poles @xmath413 and the mass term dominates because the condition @xmath414 is always satisfied . the angular momentum contribution prevails in proximity of the equatorial plane . the transition probability vanishes at @xmath415 . covariant wave equations for massless and massive particles can be solved exactly to first order in @xmath55 . the solutions are covariant and invariant with respect to the gauge transformations of the electromagnetic field and of @xmath416 and are known when a solution of the free wave equation is known . the external gravitational field only appears in the phase of the wave function . we have shown that the coupling of spin to inertia and gravitation follow from the solutions given . this allows a unified treatment of the interaction of gravity with spin and angular momentum without requiring ad hoc procedures . according to equations ( [ l ] ) and ( [ sol ] ) , the spin term @xmath417 finds its origin in the skew - symmetric part of the space - time connection . in the case of fermions , @xmath90 is accounted for by the spinorial connection . the terms that contain @xmath90 gives rise to the skrotskii effect for both electromagnetic @xcite and gravitational waves @xcite . from the phases we have derived the geometrical optics of the particles and verified that their deflection is that predicted by general relativity . in addition , the background gravitational field acts as a medium whose index of refraction can be calculated for any metric from ( [ gph ] ) , ( [ photonmomentum ] ) and @xmath418 . a more detailed treatment of the geometrical optics of single flavor neutrinos can be found in @xcite where we also calculate corrections due to the neutrino mass . for propagation parallel to the axis of rotation of the source , the rotation corrections vanish at infinity . not so for propagation perpendicular to the axis of rotation . we can finally conclude that the validity of covariant wave equations in an inertial - gravitational context finds support in experimental verifications of some of the effects they predict @xcite , in tests of the general relativistic deflection of light rays and also in the phase wrap - up in global position system measurements @xcite . we have then asked ourselves the question whether , beside the phase wrap - up in gps , there is a wider role for spin - gravity coupling in physics . in particular , we have considered muon @xmath280 experiments and helicity transitions in neutrino physics . we have found that spin - rotation coupling is largely responsible for producing the correct @xmath280 factor in the spin - flip angular frequency @xmath419 . measurements of this factor already provide the most stringent test yet of einstein s time dilation formula @xcite . however , muons in storage rings are also rotating quantum gyroscopes and inertia must be an essential ingredient of their description in experiments of high and ever increasing sensitivity . possibly these experiments also concern problems like the violation of the equivalence principle in quantum mechanics and the conservation of discrete symmetries in inertia - gravitation . we have , in fact , shown that a slightly anomalous inertial contribution to ( [ 11 ] ) at a level of the values @xmath420 and @xmath421 of section vii can produce violations of the discrete symmetries at the same level . the upper limits on the violations of @xmath24 and @xmath341 that can be reached by @xmath0 experiments are in fact as sensitive as those obtained by other means @xcite , but with an important difference . the @xmath280 measurements are performed in strictly controlled laboratory conditions rather than in astrophysical situations . in derivations based on the covariant dirac equation , the coupling of inertia and gravitation to spin is identical to that for orbital angular momentum . a suggestive interpretation of this result is that the internal distributions of the gravitational mass , associated with the interaction , and of inertial mass , associated with the angular momentum , equal each other . this is no longer so when @xmath422 . there is almost a similarity , here , with the electromagnetic case where @xmath423 is required by the dirac equation , but not by quantum electrodynamics . the deviations of @xmath424 and @xmath425 from unity that are consistent with @xmath241 experiments are both of the order of @xmath282 , or @xmath426 , and differ from each other by @xmath427 . while small values of @xmath428 do not give rise to measurable mass differences in macroscopic objects @xcite , violations of the discrete symmetries can have interesting astrophysical and cosmological implications . next , we have calculated the helicity transition amplitudes of ultra - relativistic , single flavor neutrinos as they propagate in a lense - thirring field . these transitions are interesting because at high energies chirality states are predominantly helicity states and right - handed neutrinos do not interact @xcite . the transition probabilities are of @xmath429 . two directions of propagation have again been selected and the results contain contributions from both mass and angular momentum of the source . the transitions also occur in the absence of rotation or with spin parallel to rotation , which is unexpected on semiclassical grounds . the mass contributions predominate when the neutrinos propagate from @xmath430 to @xmath431 ( and matter effects are neglected ) , provided the impact parameter @xmath432 . there is , however , a narrow region about the axis of propagation in the equatorial plane where the @xmath433 contribution is larger . the rotational contribution behaves differently in the two cases . it vanishes as @xmath434 for propagation along @xmath134 , but not so as @xmath435 in the second case . in addition , when the neutrinos propagate from @xmath436 to @xmath437 , the mass term dominates in the neighborhood of the poles , while the contribution of @xmath433 is larger close to the equator , with no attenuation at @xmath438 . in @xcite we have also calculated gravity induced , two - flavor oscillations and derived the relative equation and effective hamiltonian . the transition probabilities do indeed oscillate for the lense - thirring metric , and the curvature of space - time enters the oscillation probability through the gravitational red - shift of the local energy @xmath439 and the proper distance @xmath440 . the results presented in this paper can be applied to a number of problems in astroparticle physics and cosmology @xcite . for instance , an interesting question is whether gravity induced helicity and flavor transitions could effect changes in the ratio @xmath441 of the expected fluxes at earth . lepton asymmetry in the universe @xcite also is an interesting problem . it is known that the active - sterile oscillation of neutrinos can generate a discrepancy in the neutrino and antineutrino number densities . the lepton number of a neutrino of flavor @xmath442 is defined by @xmath443 , where @xmath444 is the number density of neutrinos ( antineutrinos ) and @xmath445 is the number density of photons at temperature @xmath341 . as noted above , the gravitational field generates transitions from left - handed ( active ) neutrinos to right - handed ( sterile ) neutrinos . if , in primordial conditions , ( [ alfa infinito ] ) and ( [ pinfinitox ] ) become larger , then helicity transitions may contribute in some measure to lepton asymmetry . finally , we have recently re - examined the behavior of the spin - gravity interaction and found that gravity can distinguish between chirality and helicity @xcite . we have also found that the spin - gravity interaction can distinguish between dirac and majorana wave packets @xcite . a spin - flip does in fact change a majorana neutrino into an antineutrino and behaves like a charge conjugation operation . the spin - gravity couplings discussed in our work make use of the weak field approximation in which gravity enters as a non - dynamical field . there are , of course , physical situations in which this approach can be trusted and the general agreement between the quasi - classical and the quantum mechanical approaches has been established @xcite . one could then be tempted to extend our approximation procedure to any order in the metric deviation , as suggested by equations ( [ n ] ) , ( [ phasephi ] ) and ( [ phase ] ) . this would however lead to inconsistencies that can only be removed , as shown by deser @xcite , by making use of the full non - linear apparatus of general relativity . furthermore , it is assumed , in calculating the phases induced by gravity , that all possible particle paths reduce , in the average , to the phase integration paths . this approximation worsens the more `` quantum mechanical '' particles and gravity become . an additional point concerns the use of the locality hypothesis in replacing non - inertial frames with inertial ones . as shown by mashhoon @xcite , this hypothesis has limitations and important consequences for the measuring process . for standard accelerated measuring devices , for instance , it entails the introduction of a maximal acceleration @xcite . in the absence of inertial frames , of a complete formulation of spin - gravity couplings in curved space - time and of direct experimental observations , assigning a role to curvature in spin related problems may then prove difficult , though worthy of investigation . a final question regards the validity of the equivalence principle in spin - gravity interactions . in @xmath280 experiments the interaction of spin with gravity depends on the relative direction of spin and rotation of the source . as such it is not universal and may be regarded as violating those formulations of the equivalence principle that hinge on universality . but a more fundamental violation has been introduced in section vii where the strength of the coupling itself depends on the helicity of the particle @xcite . this is a violation of the post - newtonian equivalence principle , recently discussed by silenko and teryaev @xcite , by which a particle s spin and angular momenta precession frequencies coincide . this principle must be the object of rigorous experimental verifications and the authors themselves suggest a number of tests to find even more precise upper limits than those determined by using muon @xmath280 experiments @xcite . in a closely related paper @xcite , the same authors show that the spin - gravity dipole coupling term found by obukhov @xcite using the eriksen - korlsrud transformation @xcite does not lead to observable effects . recently , impressive technical developments in the field of masers @xcite have succeeded in placing an upper limit of @xmath446 on violations of lorentz and @xmath290 symmetries . higher sensitivities are expected to be reached in the near future . since the interaction energy for a spin-@xmath447 particle in the rotation field of earth is of the same order of magnitude , these developments bode well for the physics of spin - gravity interactions . 99 b. mashhoon , phys . * 61 * , 2639 ( 1988 ) . b. mashhoon , phys . a * 143 * , 176 ( 1990 ) . b. mashhoon , phys . a * 145 * , 147 ( 1990 ) . b. mashhoon , phys . lett . * 68 * , 3812 ( 1992 ) . y. q.cai , g. papini , class . quantum grav . * 6 * , 407 ( 1989 ) . y. q. cai , g. papini , phys . 66 * , 1259 ( 1991 ) ; * 68 * , 3811 ( 1992 ) . d. singh , g. papini , nuovo cimento b * 115 * , 233 ( 2000 ) . g. lambiase , g. papini , r. punzi and g. scarpetta , phys . d * 71 * , 073011 ( 2005 ) . g. papini , phys . d * 75*,044022 ( 2007 ) . hehl , f.w . and lmmerzahl , c. , 1992 , _ relativistic gravity research _ , adited by j. ehlers and g. schfer , springer , berlin . c. s. misner , k. s. thorne and j.a . wheeler , _ gravitation _ ( freeman , san francisco , 1973 ) . riehle , f. , kister , th . , witte , a. , helmcke , j. , bord@xmath448,ch . j. , 1991 , _ phys . lett . _ * 67 * , 177 ; bord@xmath448 , ch.j . , 1989 , _ phys . lett . a _ * 140 * , 10 ; bord@xmath448 , ch.j . lett . a _ * 188 * , 187 ; bord@xmath448 , ch.j . , 1997 , in _ atom interferometry _ , edited by p. berman , academic press , london r. takahashi , t. nakamura , apj * 595 * , 1039 ( 2003 ) . j. ramos , b. mashhoon , phys . d * 73 * , 084003 ( 2006 ) . k. s. thorne in _ three hundred years of gravitation _ , edited by s. w. hawking and w. israel ( cambridge university press , cambridge , 1987 ) , p. 361 . g. papini in _ relativity in rotating frames _ edited by guido rizzi and matteo luca ruggiero ( kluwer science publishers , dordrecht , 2004 ) , pp . 335 - 359 . f. w. hehl , and w .- t . ni , phys . rev . d , * 42 * , 2045 ( 1990 ) . c. g. de oliveira and j. tiomno , nuovo cimento * 24 * , 672 ( 1962 ) . j. audretsch , j. phys a : math . gen . * 14 * , 411 ( 1981 ) . l. kannenberg , ann . ( n.y . ) * 103 * , 64 ( 1977 ) . g. lambiase , r. punzi , g. scarpetta and g. papini , _ general relativity and gravitational physics , 16th sigrav conference _ , edited by g. esposito , g. lambiase , g. marmo , g. scarpetta , g. vilasi ( american institute of physics , new york , 2005 ) , p. 82 . j. lense and h. thirring , z. phys . * 19 * , 156(1918 ) ; ( english translation : b. mashhoon , f.w . hehl and d.s . theiss , gen . * 16 * , 711(1984 ) ) . neil ashby , living reviews in relativity , 1 ( 2003 ) : in _ gravitation and relativity at the turn of the millennium _ , edited by n. dadhich and j. narlikar ( iuacc , pune , 1998 ) , pp.231 - 258 . b. j. venema , p. k. majumder , s. k. lamoreaux , b. r. heckel and e. n. fortson , phys . lett . * 68 * , 135 ( 1992 ) . b. mashhoon , phys . lett . a , * 198 * , 9 ( 1995 ) . j. s. bell and j. leinaas , nucl . b , * 284 * , 488 ( 1987 ) . g. papini , g. lambiase , phys . a * 294 * , 175 ( 2002 ) . b. mashhoon , class . quantum grav . * 17 * , 2399 ( 2000 ) . j. bailey @xmath449 @xmath450 , nucl . phys . * b150 * , 1 ( 1979 ) . farley and e. picasso , advanced series in high - energy physics , vol . 7 , quantum electrodynamics , ed . t. kinoshita , world scientific , singapore , 1990 , p. 479 . lewis ryder , j. phys . a : math * 31 * , 2465 ( 1998 ) . f. j. farley and e. picasso , ann . nucl . part . * 29 * , 243 ( 1979 ) . c. lmmerzahl , gen . * 28 * , 1043 ( 1996 ) . h. n. brown @xmath449 @xmath450 , muon ( g-2 ) collaboration , phys . lett . * 86 * , 2227 ( 2001 ) . w bennett @xmath449 @xmath450 , muon ( g-2 ) collaboration , phys . lett . * 89 * , 101804 ( 2002 ) ; phys . lett . * 92 * , 161802 ( 2004 ) . h. n. brown @xmath449 @xmath450 , muon ( g-2 ) collaboration , phys . d * 62 * , 091101 ( 2002 ) ; phys . lett . * 86 * , 2227 ( 2001 ) . g. lambiase and g. papini , phys . rev.d * 70 * , 097901 ( 2004 ) . g. lambiase and g. papini , in _ muons : new research _ , edited by j. caparthy ( nova science publishers , new york 2005 ) , pp . j. leitner and s. okubo , phys . rev . * 136 * , b1542 ( 1964 ) . almeida , g.e.a . matsas and a.a . natale , phys . d * 39 * , 677 ( 1989 ) . g. papini , phys . d * 65 * , 077901 ( 2001 ) . g. scolarici , l. solombrino , phys . a * 303 * , 239 ( 2002 ) . cardall , g.m . fuller , phys . d * 55 * , 7960 ( 1997 ) . bryan w. montague , phys . rep . * 113 * , 1 ( 1984 ) . d. singh , n. mobed , and g. papini , j. phys . a : math . gen . * 37 * , 8329 ( 2004 ) . bibitemaldov r. aldrovandi , g.e.a . matsas , s.f . novaes and d. spehler , phys . d * 50 * , 2645 ( 1994 ) . h. casini and r. montemayor , phys . d * 50 * , 7425 ( 1994 ) . g. v. skrotskii , dokl . nauk sssr * 114 * , 73 ( 1957 ) [ sov . * 2 * , 226 ( 1957 ) ] . schiff , phys . * 1 * , 254 ( 1958 ) . hari dass , phys . * 36 * , 393 ( 1976 ) ; ann . phys.(ny ) * 107 * , 337 ( 1977 ) . s. a. werner , j .- l . staudenmann and r. colella , phys . lett . * 42 * , 1103 ( 1979 ) . l. a. page , phys . lett . * 35 * , 543 ( 1975 ) . r. colella , a.w . overhauser , s.a . werner , phys . lett . * 34 * , 1472 ( 1975 ) . u. bonse and t. wroblewski , phys . lett . * 51 * , 1401 ( 1983 ) . f. combley , f.j.m . farley , j.h . field and e. picasso , phys . lett . * 42 * , 1383 ( 1979 ) . cooper , m.j . alguard , r.d . ehrlich , v.w . hughes , h. kobayakawa , j.s . ladish , m.s . lubell , n. sasao , k.p.schler and p.a . souder and d.h . coward , r.h . miller , c.y . prescott , d.j . sherden and c.k . sinclair and g. baum and w. raith and k. kondo , phys . * 42 * , 1386 ( 1979 ) . g. papini , proc . of the @xmath451 canadian conference on general relativity and relativistic astrophysics , edited by r.b . mann and r.g . mclenaghan , world scientific , singapore , 1994 , pp . 107 - 119 . dolgov , phys . rep . * 370 * , 333 ( 2002 ) . d. singh , n. mobed and g. papini , phys . lett . * 97 * , 041101 ( 2006 ) ; * 98*,069002 ( 2007 ) . a. j. silenko and o. v. teryaev , phys . d * 71 * , 064016 ( 2005 ) . s. deser , gen . . grav . * 1 * , 9 ( 1970 ) , also available from gr - qc/0411023 . b. mashhoon , in _ black holes : theory and observation _ , edited by f. w. hehl , c. kiefer and r. metzler ( springer , berlin , 1998 ) , pp . 269 - 284 . e. r. caianiello , lettere nuovo cimento * 41 * , 370 ( 1984 ) . e. r. caianiello , rivista nuovo cimento * 15 * , no . 4 ( 1992 ) . g. papini in _ imagination and rigor . essays on eduardo r. caianiello s scientific heritage _ , edited by s. termini ( springer , new york , 2006 ) , pp . 119 - 132 . g. lambiase , g. papini , r. punzi and g. scarpetta , class . quantum grav . * 23 * , 1347 ( 2006 ) . a. j. silenko and o. v. teryaev , gr - qc/0612103 . n. obukhov , phys . lett . * 86 * , 192 ( 2001 ) . e. eriksen and m. korlsrud , nucl . b * 18 * , 1 ( 1960 ) . r. l walsworth , _ tests of lorentz symmetry in the spin coupling sector _ , in lecture notes in physics * 702 * , pp . 493 - 505 , springer , berlin , 2006 .

the santa barbara cluster comparison project revealed that there is a systematic difference between entropy profiles of clusters of galaxies obtained by eulerian mesh and lagrangian smoothed particle hydrodynamics ( sph ) codes : mesh codes gave a core with a constant entropy whereas sph codes did not . one possible reason for this difference is that mesh codes are not galilean invariant . another possible reason is the problem of the sph method , which might give too much `` protection '' to cold clumps because of the unphysical surface tension induced at contact discontinuities . in this paper , we apply the density independent formulation of sph ( disph ) , which can handle contact discontinuities accurately , to simulations of a cluster of galaxies , and compare the results with those with the standard sph . we obtained the entropy core when we adopt disph . the size of the core is , however , significantly smaller than those obtained with mesh simulations , and is comparable to those obtained with quasi - lagrangian schemes such as `` moving mesh '' and `` mesh free '' schemes . we conclude that both the standard sph without artificial conductivity and eulerian mesh codes have serious problems even such an idealized simulation , while disph , sph with artificial conductivity , and quasi - lagrangian schemes have sufficient capability to deal with it .

1604.04868

@xcite conducted a comparison project of numerical simulations of the formation of a massive cluster of galaxies in a cold dark matter ( cdm ) universe in which the results of twelve independent simulation codes were compared . both eulerian mesh codes and lagrangian smoothed particle hydrodynamics ( hereafter sph ; * ? ? ? * ; * ? ? ? * ) codes were used in this comparison project . in this project , they took into account only gravity and hydrodynamics so that the comparison was simplified . this project is known as the `` santa barbara cluster comparison project '' . although most of global properties reasonably converged , there was a systematic difference in entropy profiles of the clusters in simulations with different methods . the results of the eulerian mesh codes showed the central core in entropy profiles , whereas those of sph codes did not ( see figure 18 in their paper ) . later studies confirmed this tendency ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? @xcite interpreted that this difference might reflect the difference in the treatment of shocks of sph and mesh codes . @xcite pointed out that there are `` fundamental differences '' between eulerian mesh and lagrangian sph codes ( see also * ? ? ? * ; * ? ? ? this difference is originated from the fact that the standard formulation of sph ( hereafter ssph , where @xmath0 and @xmath1 are the mass and density of a particle , as the discrete volume element for the discretization of the equations . ] ) has difficulties when dealing with contact discontinuities . since ssph requires the differentiability of the density field , the density around a contact discontinuity has a large error . in particular , the density of particles in the low - density side of the contact discontinuity is overestimated by a large factor . this error propagates to the pressure evaluation , resulting in the _ unphysical surface tension _ and the suppression of fluid instabilities ( see figure 1 in * ? ? ? we can solve this problem by using the volume element which does not depend on the density @xcite , by introducing extra dissipation terms @xcite , by using the integration form to evaluate the derivatives of physical quantities @xcite , by employing the non - standard equation of motion which breaks newton s third law @xcite , and by adopting the godunov sph @xcite where physical quantities are smoothed twice @xcite . @xcite used a diffusive quantity , @xmath2 , instead of the mass and mass density , to formulate the sph approximation . each particle has the @xmath3 , and from that its density is calculated using the usual formula to calculate mass density from mass of particles in sph . since all of the sph codes used in the santa barbara cluster comparison project were based on ssph , the cuspy entropy profile obtained with sph codes might be , at least partly , the artifact caused by the unphysical surface tension . in the hierarchical structure formation scenario , a cluster of galaxies grows through mergers of a number of small building blocks which contain cold , low - entropy gas . the unphysical surface tension of ssph tends to protect cold gas clumps in the building blocks against the fluid instabilities and the ram - pressure stripping , as demonstrated by the `` blob test '' of @xcite . as a result , cold clumps might reach the center of the cluster due to the dynamical friction with keeping their low entropy state , resulting in a cuspy entropy profile of the cluster . some of techniques described above have been applied to the cluster formation simulations in order to investigate their effect on the central entropy structure . @xcite added a diffusion term to the energy equation which mimics the turbulence mixing and performed the santa barbara cluster simulations . they found that the entropy core could be formed , but the size and depth of the entropy core depended strongly on the diffusion coefficient and numerical resolutions ( see their figures 12 and 13 ) . @xcite carried out a set of comparison tests of the cluster formation in a @xmath4cdm universe with sph and an adaptive mesh refinement ( amr ) codes , focusing on the formation of the entropy core . in their comparison tests , they used not only ssph but also their new sph , sphs @xcite which avoids multivalues of physical quantities by artificial diffusion terms with a higher order dissipation switch and increases the force accuracy by adopting a higher order kernel with a large number of neighbor particles . in sphs runs , the entropy core was formed in the cluster center and its size was consistent to that obtained by an amr code , ramses @xcite . this good agreement is achieved when the maximum dissipation parameter @xmath5 ( see their figure 6 ) . on the other hand , in the case with @xmath6 , the entropy profile has a core , but it is smaller and the core entropy is lower than those with converged values . in the case with @xmath7 , it is close to that found with ssph . recently , @xcite reported a new comparison test of a numerical cluster in a @xmath4cdm universe . they showed that sph codes with the artificial conductivity ( hereafter ac ) term and sphs give entropy profiles which are close to those given by amrs ( i.e. , cored profiles ) , although they did not provide the results of the parameter tests . the agreement with the results of amr calculations does not guarantee the validity of the results , since eulerian schemes have their own limitations . an obvious one comes from its eulerian nature . in order to integrate the motion and internal energy of fast - moving cold gas clump , very high accuracy is required . the internal energy of a gas clump can be many orders of magnitudes smaller than its kinetic energy in the reference frame of the mesh . numerical error comes from the kinetic energy , but it affects the internal energy of the clump . consider the case that a molecular cloud moves in a galaxy following the galactic potential . the typical temperature of the cloud is @xmath8 while the virial temperature which corresponds to the typical kinematic energy is @xmath9 for a galaxy of the milky way size . the energy difference between the thermal and kinetic energies of the cloud becomes about five orders of magnitude . very high accuracy is , thus , required to keep the structure of the clump , and this requirement in turn results in the requirement of high spacial resolution and small time step . otherwise , a large error is induced by the loss of significant digits and/or negative pressure . a remedy for this problem is the dual energy formulation @xcite . in this formulation , both total ( @xmath10 ) and internal ( @xmath11 ) energies are solved independently . then , either @xmath11 or @xmath12 , where @xmath13 is the kinetic energy , is used to evaluate pressure , depending on the fraction of the thermal energy . for the high mach number fluid , e.g. , @xmath14 , @xmath11 is adopted to evaluate pressure . otherwise , since @xmath15 , the loss of significant digits occurs . another difficulty is that the eulerian formulation is not the galilean invariant , which leads to the numerical diffusion . an example of this problem can be seen in figure 14 of @xcite . in eulerian codes , an initially hydrostatic cluster decays after 1 gyr evolution , in particular in enzo @xcite with zeus @xcite and flash @xcite , when a translational velocity was added . in principle , by going to higher spatial resolution , we can reduce the rate of decay by reducing the effect of the numerical dissipation @xcite . nevertheless , compared to lagrangian codes with the same effective accuracy , the computational cost would be much higher . the other way is to adopt higher spatial order schemes , such as essentially non - oscillatory polynomial interpolation ( eno ; * ? ? ? * ) , weighted eno ( weno ; * ? ? ? * ) , and constrained interpolation profile ( cip ; * ? ? ? * ) schemes . however , these schemes are computationally expensive and it is difficult to combine high - order schemes with amr in a consistent way . the latter is serious because joint interfaces of finer and coarse meshes lose their accurracy @xcite and thus careful treatment is necessary . recently , novel numerical schemes have been developed and used in the field of numerical astrophysics . @xcite developed a moving mesh scheme which is free from the above problem of the eulerian mesh codes . in @xcite , he carried out simulations of the santa barbara cluster with `` arepo '' , an implementation of his moving mesh scheme , and reported that the entropy core was formed . the size of the core is smaller than those obtained by amr simulations and the entropy at the core is lower than those obtained by amr simulations . @xcite developed a new code `` gizmo '' which implemented a meshless method and found that in cluster simulations with gizmo entropy cores similar to those of arepo is formed . these results are quite encouraging . however these two codes use similar approaches , such as quasi - lagrangian nature using riemann solvers . further tests with different schemes are important . in this paper , we report the results of santa barbara cluster simulations obtained with the density independent formulation of sph ( disph ) @xcite . since disph is free from the unphysical surface tension , we can expect that it gives better results compared to ssph . in addition , since disph do not introduce artificial dissipations , we can expect that the result is not affected by free parameters . the structure of this paper is as follows . in [ sec : disph ] we give a brief review of the formulation of ssph and disph and their advantages and disadvantages . we then describe the initial condition and our numerical methods in [ sec : method ] . the comparison is carried out in [ sec : results ] and discussion is given in [ sec : discussion ] . in the sph method , all physical quantities are evaluated using the kernel approximation over the field discretized by particles . the kernel approximation for a physical quantity @xmath16 is given by @xmath17 where @xmath18 and @xmath19 are position vectors , @xmath20 is a function with a compact support called the kernel function , and @xmath21 is the kernel size . we approximate this spatial integration by the summation over particles expressed as @xmath22 where @xmath23 and @xmath24 are the values of @xmath16 at the positions of particles @xmath25 and @xmath26 , respectively , and @xmath27 is the volume element associated with particle @xmath26 . in ssph , we use @xmath28 where @xmath29 and @xmath30 are the mass and density of particle @xmath26 , respectively . from eqs . and , by substituting @xmath1 into @xmath16 , we have @xmath31 note that eq . gives the _ smoothed _ estimate of the density field . thus , the density near a contact discontinuity suffers a very large error . consider the case that we evaluate the density around a contact discontinuity using eq . the density of a particle at the low - density side is over - estimated , while that of a particle at the high - density side is under - estimated . the absolute amount of the error is similar at the both sides . the relative error is much larger at the low - density side , simply because the true density is smaller . this large overestimate of the density results in equally large overestimate of the pressure for low - density particles near the contact discontinuity , which then causes a strong repulsive force . this repulsive force works as an unphysical surface tension ( e.g. , * ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? if the distribution of the internal energy is sufficiently smooth , this unphysical surface tension does not show up even with the smoothed density . the ac term introduced by @xcite automatically reduces the pressure jump by spreading the internal energy . . ] a switch is always used to avoid unnecessary diffusion . in disph , unlike ssph , the pressure ( or its arbitrary function ) is used to evaluate the volume element @xcite . the pressure field is smooth everywhere , except at the shock front . in sph , the shock is handled by the artificial viscosity ( hereafter av ) which smooths the velocity field and prevents multivalues . as a by - product , we have a smoothed pressure distribution in sph simulations , even at the shock front . therefore , we can expect that the problem of ssph at the contact discontinuity is solved if we use a volume element based on the pressure . this is the basic idea of disph . the simplest form of the fundamental equation used in disph can be obtained by using the volume element @xmath32 , where @xmath33 and @xmath34 are the energy density and the internal energy of particle @xmath26 , respectively . substituting @xmath35 into @xmath16 in eq . and using the energy density based volume element , we have @xmath36 since the pressure is smooth at contact discontinuities and even at the shock front , we can obtain smoothed values without large errors . note that , if one uses disph for a non - ideal gas , one needs to use pressure directly instead of the energy density @xcite , because it is no longer smooth . we note that this formulation itself does not introduce any dissipation . in disph , as a trade - off for using of the smoothed pressure for the formulation , the density distribution has a jump at a contact discontinuity if the internal energy distribution there is not smooth ( see figure 2 in * ? ? ? * ) . while this jump does not affect the motion of particles , it is necessary to be aware of this `` unphysical '' density jump when one uses this density to evaluate physical quantities , such as radiative cooling as is noted by @xcite and @xcite , the simplest solution is to use the smoothed density . in the same manner as ssph , disph can conserve the total energy , linear and angular momentum . if we use a variation principle for the derivation of sph equations @xcite , we can obtain a set of conservative sph equations @xcite . the equations in this derivation have the so - called `` grad-@xmath21 '' term . moreover , if we carefully constructs energy and momentum equations , we can obtain another set of the conservative equations which do not involve the grad-@xmath21 term ( see section 5.6 in * ? ? ? here we adopt disph with the grad-@xmath21 term , which is more robust under the strong shock . in appendix [ sec : method : timeintegration ] , we show the result of the three - dimensional collapse test @xcite . the relative energy error from the beginning and the end of the simulation is @xmath37% , which is comparable to the previous studies ( e.g. , * ? ? ? * ; * ? ? ? the initial condition we used here is basically the same as that used in @xcite . the volume of their initial condition is @xmath38 in the comoving space and the initial redshift is @xmath39 . the @xmath40 overdensity is imposed at the center of the volume so that a massive cluster will be formed at the center . there are two set of initial conditions with different resolutions : @xmath41 and @xmath42 particles . the cosmology is the standard cdm and the hubble parameter is @xmath43 where @xmath44 . we integrated the formation and evolution of the santa barbara cluster with a vacuum boundary condition in the physical space . we cut the central spherical volume with the radius of @xmath45 in the comoving space and remapped it to the physical space at @xmath46 . then , we imposed the hubble flow corresponding to the initial redshift . following @xcite , we assumed @xmath47 and @xmath48 . since the original initial conditions of @xcite are for dm only simulations , we divided each particle into two particles which have the same position and velocity of the original particle but have different masses , i.e. , that one has @xmath49 and the other has @xmath50 of the original particle mass , for simulations involving both dm and gas . the number of particles , the mass of particles and the spatial resolutions ( softening lengths for dm and baryon : @xmath51 and @xmath52 ) are summarized in table [ tab : parameters ] . for both of low- and high - resolution models , we performed two runs , one with the standard sph , and the other with disph . we carried out five high - resolution ssph runs involving ac . these five runs adopted different values for the maximum dissipation parameter . in addition , two extra models which combine the low mass and the high spatial resolutions ( mixed 1 ) and the high mass and the low spatial resolutions ( mixed 2 ) are adopted in order to investigate the effect of the resolutions further . we also carried out two runs with ssph and disph for each model . thus , in total we performed thirteen runs . [ cols=">,^,^,^,^,^",options="header " , ] + we used a parallel @xmath53-body / sph code asura @xcite for this experiment . the gravitational interactions between particles were calculated by the tree method @xcite . the tree with grape method @xcite was used . the parallelization method of the gravity calculation was that proposed by @xcite . we adopted the symmetrized plummer potential and its multipole expansion @xcite so that we can calculate the gravitational interactions between particles with different values of softening lengths by a single tree . the opening angles for the ordinary three dimensions and softening lengths were set to @xmath54 following @xcite . the phantom - grape library was used for the calculation of particle - particle interactions in order to accelerate the calculation @xcite . the equation of state of the ideal gas with the specific heat ratio @xmath55 was used . in order to handle shocks , the av term proposed by @xcite was used . in addition to this , we used an av switch proposed by @xcite in which the value of the av parameter , @xmath56 , changes in the range of @xmath57 . following @xcite , a modification which accelerates the increase of @xmath56 at lower @xmath56 regime and slows it at higher @xmath56 regime was also used . the balsara limiter @xcite was adopted in order to reduce the shear viscosity . we employed the wendland c4 kernel which is free from the paring instability @xcite . the number of neighbor particles was kept within @xmath58 . this higher order kernel with a larger number of neighbor particles can reduce the so - called e0 error @xcite . the time integration was carried out by a second order scheme ( see appendix [ sec : method : timeintegration ] for details ) . the individual and hierarchical time - step methods were used @xcite . for sph particles , the time - step limiter , which enforces the time - step difference among neighboring particles to be small enough to follow strong shocks , was used @xcite . the fast scheme , which allows each particle to have different time - steps for gravitational and hydrodynamical interactions and integrates these two interactions independently , was also adopted @xcite . when we used the fast scheme , the computational time ( wall - clock time ) became half of what it was without fast ( we used 128 cpu cores of cray xc30 in this comparison ) . this gain is almost the same as that reported in @xcite , although simulation setups in both runs were completely different . following previous studies @xcite , we define a dense region which is formed at the center of the simulation volume and with the mean density is 200 times higher than the background density as the `` santa barbara cluster '' . we express the radius where the mean density becomes 200 times of the background density as @xmath59 and call it the virial radius . the procedure to find the center and size of the santa barbara cluster is as follows . first , we find a particle of which total energy is the lowest and adopt it as the center of the cluster of galaxies . then , we sort all of particles based on the distance from the center of the cluster in ascending order . we calculate the mean density from the center to the outer part following the particle order , and we halt this operation when the mean density becomes less than 200 times of the background density . we regard the distance from the center to the particle position where we halt this operation as the virial radius . this simple strategy works well for the santa barbara test in which a single massive cluster is formed near the center of the simulation volume . in order to draw radial profiles , we use a density weighted center instead of the halo center since the peak of the baryon distribution does not always coincide with the position of particle with the lowest total energy . we use the following definition as the density weighted center @xcite : @xmath60 where we only take care of the gas particles in the virial radius . for this density @xmath30 , we just used the ssph - estimated density . the kernel function and the number of neighbor particles are the same as those described in [ sec : method : numerical ] . in this section , we compare the properties of the santa barbara cluster obtained by simulations with ssph and disph , and then investigate the origin of the difference in the entropy profile . we first describe the global properties of the santa barbara cluster in [ sec : results : globalproperties ] . the density , temperature and entropy distributions in the clusters obtained by simulations with ssph and disph at @xmath61 are compared in [ sec : results : snapshots ] . in [ sec : results : radialprofile ] , we compare radial profiles . we investigate how the central entropy core / cusp formed in [ sec : results : entropy : formation ] . in [ sec : results : ac ] , we investigate the contribution of the ac term for the cluster entropy profile . in order to avoid confusion , we use the label `` sph '' to indicate the results of ssph on figures and `` disph '' for the results of disph . the average of virial radii of the santa barbara clusters of the four runs ( the high- and low - resolution runs with ssph and disph ) is @xmath62 . this radius is somewhat larger than those obtained in previous studies @xcite . the total mass within @xmath59 is @xmath63 which is @xmath64 larger than those obtained by @xcite . these differences are probably due to the difference in the treatment of the boundary condition . we used the open boundary condition , while all runs in @xcite adopted the periodic boundary condition . since the radius of the initially imposed overdensity region is @xmath65 , the cut - off radius of @xmath45 would be insufficient to remove boundary effects . nonetheless , as we see later , the central region of the cluster does not seem to be affected by the boundary condition . in the following , we call the santa barbara cluster just `` cluster '' , for brevity . we found that the clusters obtained by disph contain more gas than those obtained by ssph . the gas fractions of the cluster with disph are @xmath66 which are close to the results of high resolution mesh codes in @xcite and also the cosmic averaged value of @xmath67 for this model . on the other hand , those of ssph are @xmath68 which are close to those of ssph runs in @xcite , @xmath69 . figure [ fig : virial : density ] shows the gas density maps of the four runs within @xmath59 at @xmath61 . these density maps are not very different . however , when we inspect these maps carefully , we find several differences . first , we can see that the central densities in runs with ssph are higher than those with disph . the density of the innermost regions of the clusters with ssph is one order of magnitude higher than that with disph . second , for both of ssph and disph runs , the central density of the cluster is higher for runs with high mass resolution ( large number of particles ) . figure [ fig : virial : temperature ] shows the temperature distribution within the cluster of four runs . we can see that the clusters of ssph runs contain more small - scale substructures compared to those of disph runs . figure [ fig : virial : entropy ] shows the entropy distributions ; the definition of the entropy we adopted in this paper is @xmath70 where @xmath71 and @xmath1 are the temperature in the unit of kelvin and the density in the unit of @xmath72 , respectively , and this definition is the same as that used in @xcite . it is obvious that the values of the central entropy in ssph runs are much lower than those in disph runs . the clusters in disph runs have no or a little amount of gas with entropy below @xmath73 , while the clusters in ssph runs have significant amounts of such gas at the center . it is also clear that the entropy distributions of the cluster in disph runs are smoother than those of clusters in ssph runs . in other words , the entropy distribution indicates that the gas in the clusters of ssph runs is not well mixed . figure [ fig : radialprofiles ] shows the radial profiles of the cluster at @xmath61 . in this figure , we used two high - resolution results . in this plots , `` sb '' denotes the averaged profiles in @xcite , `` arepo '' is the result of a moving mesh code arepo by @xcite . as expected , with this formalism the entropy generation is slightly suppressed @xcite . on the other hand , @xcite pointed out that spurious heating from riemann solver errors affects an entropy profile when this formalism turned off . hence , we adopt the original data shown in @xcite . ] , and `` nyx '' denotes the result of an amr code nyx by @xcite . -body / gas dynamics code designed for large scale cosmological simulations . in this code , the fluid evolution is solved by a finite volume method with a block structured amr . the dual energy formulation is used . the riemann solver is used for the evaluation of fluxes . according to these features , nyx is one of good representatives of the state - of - the - art amr codes . the entropy profile of the santa barbara cluster with nyx is almost identical to those with enzo @xcite and art @xcite . ] we note that these data were taken from figures in their papers : figure 18 in @xcite for sb , figure 45 in @xcite for arepo ( we adopt the result labeled @xmath41 ) , and figure 7 in @xcite for nyx . we also note that the entropy profile with gizmo reported by @xcite is comparable to that with arepo . the density profiles of dm ( the top - left panel of figure [ fig : radialprofiles ] ) of all four runs agree very well . the central density of gas of the ssph run is significantly higher than those of the other three runs . here , `` sb '' , `` arepo '' and our disph results seem to agree well . note , however , that the `` sb '' plot is the average of 7 sph and 5 mesh - based scheme runs , and sph models in @xcite gave the result similar to our ssph run , while the mesh runs gave a central density even lower than arepo / disph results . thus , the agreement between our disph result and the arepo result is rather good , whereas all ssph runs gave high central densities and amr runs , lower density . from the temperature profile ( the left - bottom panel of figure [ fig : radialprofiles ] ) and density profile ( the top - right panel ) of the gas component , we can see that the standard sph gives a central cusp with the temperature decreasing inward , while arepo and disph give constant - density , constant - temperature cores . mesh runs in @xcite also gave constant - density cores , but their cores are systematically larger than those obtained by arepo and our disph . the averaged `` sb '' profile of the temperature shows a sign of central decrease , since it is the average of sph ( decreasing ) and mesh ( increasing ) profiles . we can see the same tendency in the entropy profile ( the bottom - right panel ) . in figure [ fig : entropy : comparison ] , we plot the entropy profiles of our low - resolution runs , as well as those of our high - resolution runs and three results from the literature ( `` sb '' , `` arepo '' and `` nyx '' ) . we can see that for disph runs , the difference in resolution does not make much difference in the entropy profile ; the entropy profiles with disph are consistent with that obtained by a moving mesh code arepo . on the other hand , our low - resolution ssph run gave a central entropy significantly higher than that of our high - resolution ssph run . our low - resolution ssph result is similar to those of the ssph simulations in @xcite and @xcite . thus , it is clear that the `` core '' in these previous ssph results is due to the limitations in the numerical resolution . from this figure , we find that the central entropy in the high - resolution disph run is slightly higher than that in the low - resolution disph run , although we saw the opposite sense from the entropy maps ( figure [ fig : virial : entropy ] ) . this is because there is an offset of @xmath74 between the position of the entropy minimum and the plot center ( the density weighted center ) in the high - resolution disph run , and it makes the rather flatter central entropy profile . thus , there is no discrepancy . in order to assess which parameters , i.e. , mass and spatial resolutions , are more crucial to the final entropy profile of the cluster , we carried out two extra models with disph and ssph which mix resolutions ; ( 1 ) with the high mass / low spatial resolutions ( `` mixed resolution 1 '' ) and ( 2 ) with the low mass / high spatial resolutions ( `` mixed resolution 2 '' ) . as shown in figure [ fig : entropy : mixed ] , the former run reproduced the entropy profile of the high - resolution ssph run , while the latter run did that of the low - resolution ssph run . a slight difference is found only in the very central region ( @xmath75 ) . hence , in the ssph runs , the central structure of the entropy profile is sensitive to the small scale power of the initial density fluctuations . there is no significant influence on the results with disph . the good agreement between results with disph , arepo , and gizmo is quite encouraging . however , the analysis so far is only limited at the date of @xmath61 . in the next subsection , we compare the results obtained by disph and ssph methods deeply , by focusing on the formation process of the cluster . the cluster grew rapidly while @xmath76 , and then experienced a major merger around @xmath77 , and evolved mildly until @xmath61 . we follow these typical phases of evolution to understand how and when the entropy core and cusp are established . figure [ fig : entropy : threeepochs ] shows the radial entropy and density profiles for the high - resolution disph and ssph runs . we here plot them for three epochs , before the major merger ( @xmath78 ) , after the major merger ( @xmath79 ) , and at the end of the simulation ( @xmath61 ) . for the run with disph , at @xmath78 , the entropy profile has a relatively small core . then , the core entropy and core size increase and these quantities are kept unchanged until @xmath61 . the entropy profile at @xmath79 is comparable to that at @xmath61 ; the last major merger event triggers this core entropy increase ( see below ) . the entropy of the intermediate region of @xmath80 slightly increased during @xmath81 . the evolution of the density profile is similar , but in the opposite direction . the core density decreases by almost a factor of three from @xmath78 to @xmath79 , and it remains unchanged until @xmath61 . the evolution of the radial profile in the ssph run is quite different from that in the disph run . at @xmath78 , the entropy profile with ssph has a small core which is comparable to that of the disph run . this entropy core becomes larger until @xmath79 . however the core entropy is much lower than that with the disph run . at @xmath61 , the entropy core vanishes nearly . the density profile follows this evolution similarly but in the opposite direction . at the end of the simulation the density profile has a cusp . evolution of the radius of the entropy core and the averaged core entropy in the runs with disph and ssph are shown in figures [ fig : entropy : coreradius ] and [ fig : entropy : coreentropy ] . the definition of the core radius is described in appendix [ sec : method : core ] . the averaged core entropy is measured using the entropy of gas in the core radius . their evolutions can be divided into two phases and the transition takes place after the major merger epoch ( @xmath82 ) . in the disph runs , two results agree with each other very well . in the early phase ( @xmath83 ) , the radius of the entropy core increases from @xmath84 ( @xmath85 ) to @xmath86 ( @xmath82 ) and the averaged core entropy increases from @xmath87 ( @xmath85 ) to @xmath88 ( @xmath82 ) . then the growth of the core radius is stopped and the core is essentially unchanged until @xmath61 . in the ssph runs , the evolution is completely different from those of disph runs . when we compare the high resolution runs , the core radius of the ssph run is smaller than that of the disph run in the initial phase ( @xmath89 ) . major merger triggered the rapid increase , but at @xmath82 the core is still smaller than that of disph runs . second , the core size shrinks and the averaged core entropy decreases in the late phase ( @xmath90 ) . these figures suggest that the low - entropy material is supplied to the central region , most likely via minor mergers in the late phase . the low resolution ssph run follows the evolution of the high resolution one , although its goes between disph runs and the high resolution ssph runs because of lower resolution . in [ sec : majormerger ] and [ sec : minormerger ] , we further investigate the evolution of the cluster s central entropy induced by the major and minor mergers . figure [ fig : entropy : merger : disph ] shows the evolution of the distributions of pressure and entropy in the cluster in the high resolution disph run during the major merger phase ( @xmath91 ) . a smaller cluster consisting of the low - entropy material approaches from the top - right corner ( @xmath92 ) , it merges with the main cluster ( @xmath93 ) and after several crossing times these two clusters completely merge to form a single cluster with the relaxed structure ( @xmath94 and @xmath95 ) . in this run , the entropy at the cluster center increases due to shocks ( see bottom panels ) . before the merging event , the central entropy is lower than @xmath73 , but it becomes higher than @xmath73 after the merger event ( see also figure [ fig : entropy : coreentropy ] , @xmath96 at @xmath97 whereas @xmath98 at @xmath82 ) . the pressure maps are smooth throughout this event ( top panels ) . in the ssph run , we also observe the increase of the central entropy whereas the absolute value is lower than that in the disph run , as we see figure [ fig : entropy : merger : ssph ] . there are sharp edges in the distribution of the low - entropy ( cold ) gas clumps , and therefore the mixing of low- and high - entropy components is much weaker in the ssph run compared to that in the disph run . pressure maps show sharp gaps which are not found in those in the disph run . thus , these gaps are outcomes of the unphysical surface tension of ssph , resulting in the suppression of the entropy generation due to shocks . here , we focus on the late stage of the evolution of the cluster . figure [ fig : entropy : merger : minor ] shows the evolution of the radial entropy distribution . the main sequence of the entropy distribution , where the relative fraction is high , corresponds to the radial entropy profile we saw in , e.g. , figure [ fig : entropy : comparison ] . in addition to the main sequence , there are a number of low entropy components . these low - entropy components are clumps in and around the cluster . the evolution of the cold gas clumps are quite different in the disph run and ssph run . in the case of ssph run ( bottom panels ) , there are many clumps with minimum entropy around @xmath99 , within the distance range @xmath100 . on the other hand , in the case of the disph run , there are only a few clumps within the distance @xmath101 , and their minimum entropy is much higher . thus , in the case of the ssph runs , cold gas clumps survive in the cluster and fall to the cluster center , forming low - entropy core , while in the case of the disph run , cold gas clumps are disrupted before they reach to the center of the cluster . in order to clarify the difference of the late stage evolution of the disph run and that of the ssph run , we here focus on the evolution of one clump . for this analysis , we first pick up low entropy component within @xmath59 . the threshold entropy is @xmath102 which is slightly lower than the core entropy of the disph runs at @xmath61 . with this threshold , we found a number of low entropy clumps . in order to select one representative clump from them , we imposed the following conditions : ( 1 ) the clump is approaching to the cluster center , ( 2 ) the mass is sufficiently large so that we can track its evolution , ( 3 ) the clump reaches to the center of the cluster before @xmath61 , and ( 4 ) the clump can be found in both runs at the same position . the representative clump we picked up is found at @xmath103 and its total mass is @xmath104 at @xmath105 . figure [ fig : target : disph ] shows the evolution of this low - entropy clump in the high - resolution disph run . from the particle distribution maps ( upper panels ) , we can see that this representative clump enters the cluster center forming a bow - shock . its front surface is being stripped due to fluid instabilities , resulting in the complete destruction of the clump by @xmath106 ( recall the blob test in * ? ? ? the pressure distribution across the interface between the clump and the intracluster medium is always smooth . figure [ fig : target : ssph ] shows the evolution of the same clump in the high - resolution ssph run . there is a clear gap both in the particle and pressure distributions in the front of the clump . this gap is induced by the unphysical surface tension , protecting this clump from the development of fluid instabilities at the surface and ram - pressure stripping . thus , the low entropy gas in this clump can sink to the cluster center . we compare the evolutions of this representative clump in disph and ssph runs in figure [ fig : target : evolutions ] . top , middle , and bottom panels are the distance from the cluster center , the size , and the averaged entropy of the gas particles of the clump , respectively . before the pericenter passage , clumps in two runs are almost identical . when it passed the pericenter , the orbital decay starts . the orbital decay in the ssph run is much faster than that in the disph run . clumps expand when passing the pericenter . the clump in the disph run is more extended compared to that in the ssph run . these two results tell us that the clump in the ssph run keeps a compact structure much longer . the increase of the mean entropy starts slightly before the pericenter passage in both runs . at @xmath61 , the mean entropy of the clump in disph is @xmath107 higher than that in ssph . this indicates that the shock heating is more efficient in the disph run . however , behind the fact that ssph has unphysical surface tension , we should say that the shock heating in ssph is more inefficient . disph can capture the intrinsic shocks . this difference in the shock heating efficiency explains why cold gas clumps can reach to the center of the cluster in the ssph run but are disrupted in the disph run . this tendency that the clumps in the ssph run survive longer at the center of the cluster compared to those in the disph run can be seen in figure [ fig : target : stack ] . here , we used subfind clumps ( see appendix [ sec : method : subfind ] ) and stacked data during @xmath108 ( total 17 snapshots ) . generally , as clumps move closer to the center of the cluster , they loose their masses due to tidal and ram - pressure stripping . at the same distance from the cluster center , in particular for @xmath109 , we can see the clear tendency that clumps in ssph are more massive than those in disph . thus , the evolution of the inner part of the cluster strongly affected by the choice of the sph scheme . while clump surviving time is different , there is no significant difference in mass functions with different sph schemes ( see appendix [ sec : results : massfunction ] ) . it has been pointed out by @xcite that the cluster core entropy decreased at the low-@xmath110 with ssph . they measured the evolution of entropy at the central region ( @xmath111 ) of clusters of galaxies in a @xmath4cdm universe . they found that the core entropy of the ssph run decreased continuously for @xmath112 whereas those of amr and sphs runs were roughly constant . although they did not follow the individual clumps , they explained that the change of the central entropy found in low-@xmath110 period is due to the accretion of the low - entropy clumps induced by the minor mergers . our study supports their result and this process works in the evolution of the santa barbara cluster in previous studies as well . recently , the ac term has become widely used in sph simulations in order to remove the unphysical surface tension ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? , we investigate the effects of ac on the cluster formation and evolution . similar studies are found in the literature ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? according to the numerical experiments by @xcite and @xcite , it is expected that the entropy core appears if the ac term is introduced into ssph and the core size and core entropy depend on the maximum value of the conduction control parameter @xmath113 . thus , in this section , we highlight the contribution of @xmath113 to the thermal structure and the entropy profile . here , we adopt the following functional form for the ac : @xmath114 where @xmath115 , @xmath116 , @xmath117 is pressure , and @xmath118 $ ] . following @xcite , we evaluate the ac coefficient of particle @xmath25 , @xmath119 using the following equation ; @xmath120 where @xmath121 is evaluated by the sph manner and @xmath119 is allowed to evolve from @xmath122 to @xmath113 . this ac term is essentially the same as those adopted in the previous studies ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? we tested several functional forms and switches and found that the conclusion in this subsection does not affected by them . we show five high - resolution ssph runs with different values of @xmath123 : @xmath124 and @xmath125 . figure [ fig : ac : temp ] shows the temperature maps of five runs with the ac term , as well as that of the run without ac . from this figure , we can see that the temperature structure gets blurred when @xmath113 increases . we find that there is a similar effect in both the density and entropy maps . this blurring effect has an influence on the entropy profile of the cluster , as we see next . figure [ fig : ac : entropy ] shows the radial entropy profiles with and without ac . apparently , the entropy profile is affected by ac . the entropy profile with @xmath126 is cuspy ; it has a slightly higher entropy than that without ac in @xmath127 . on the other hand , those with @xmath128 have cores . these core entropies increase with increasing @xmath129 and the increase stalls when @xmath130 . the core entropy with @xmath131 is comparable to those with disph and arepo whereas that with @xmath132 is rather close to that obtained by a mesh code . in our test , unlike previous study @xcite , the core entropies with @xmath130 are slightly lower than those obtained by mesh codes . this difference might come from the detailed implementations . time evolutions of averaged core entropies for five ac runs are shown in figure [ fig : ac : entropy : core ] . the overall evolutions of the runs with @xmath133 are similar to those of disph runs ; the increase of the core entropy is fast until @xmath82 and then it stops . on the other hand , the evolution with @xmath134 is close to that without ac , whereas this run has a slightly higher entropy within @xmath135 . the final core sizes are @xmath136@xmath137 for the runs with @xmath138 ( recall that they are @xmath139 in disph runs ) . although ssph runs with ac ( @xmath138 ) have slightly larger entropy cores , the effect of the ac term looks similar to that of disph ; the essential difference is that ac requires the new free parameter @xmath123 . since the numerical results depend on this parameter , one needs to pay attention to determine its value . it is of course possible that the calibrated @xmath113 for one experiment is inadequate for another experiment . in this paper , we studied the formation of the entropy core in the santa barbara cluster . previous studies ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) have demonstrated that simulations with eulerian mesh codes make the entropy core in the cluster center while those with lagrangian sph codes do not . the original paper @xcite argued that the origin of this difference came from the different treatment of shocks in the sph and mesh codes . however , both eulerian mesh codes and lagrangian sph codes have their own , intrinsic limitations which might affect the entropy evolution . this implies that neither mesh results nor sph results is correct . @xcite and @xcite reported that new lagrangian fluid schemes made entropy cores but the depth of their cores is deeper than those obtained by eulerian codes . the clusters simulated with disph formed prominent entropy cores and its size and absolute value of the core entropy was insensitive to the adopted resolutions . during the cluster formation , a number of low entropy clumps accreted and these clumps brought low entropy materials to the cluster center . in the disph runs , the ram pressure stripping and fluid instabilities disrupted these clumps efficiently in disph runs . the entropy of these clumps increased due to hydrodynamical shock and they built the entropy core . the size of the entropy core at @xmath61 is @xmath139 and the depth of that is @xmath140 . these values are comparable with those obtained by the original version of the moving mesh code @xcite and the mesh free code @xcite . the values of the core entropy obtained in these simulations are about unity lower than those with eulerian mesh codes . this systematic difference is probably due to the numerical diffusion intrinsically existing in eulerian codes ( e.g. , * ? ? ? * ; * ? ? ? * ) . in principle , the effect of the numerical diffusion can be reduced by increasing the spatial resolution @xcite . however , whether this strategy is really useful for cosmological simulations or not is unclear since this approach is quite time - consuming . we were able to reproduce the cuspy entropy profile when we used ssph and the results with ssph depended strongly on the adopted mass resolution . when we employ low mass resolution , there is a small core which is comparable with the previous simulations . when we adopt the high mass resolution , the central entropy became much lower and the cusply profile appeared . by following the evolution of the central entropy , we found that the central entropy in ssph runs affected by the minor mergers involving the low - entropy gas . low - entropy gas in simulations with ssph is `` protected '' against ram - pressure stripping , as found by @xcite . the evolution of low - entropy clumps in disph is completely different from that in ssph . the cuspy entropy profile found in the ssph runs is artifact . we can obtain the entropy core in the santa barbara cluster with just changing the volume element of the formulation and without introducing any diffusions . there are several studies in which entropy core was obtained by introducing dissipations to sph @xcite . in their studies , they could obtain cored entropy profiles . we also confirmed it by our numerical experiments . as is reported by previous studies , the core size depended on the diffusion coefficient of the conductivity . our method has advantages over these methods since there is no need to calibrate the diffusion coefficients . in addition , there is no time and spacial scales which controls the degree of the dissipation strength if we used disph , since formulation of disph itself can remove the unphysical surface tension . we thus conclude that disph is a good alternative to ssph for the cosmological structure formation . there is room to improve the implementation of the av term and switches . for instance , a sophisticated switch is important in order to deal with a rotating system @xcite . recently , @xcite showed a survey of the implementation of the av . in their study , they showed that the run with the combination of the von - neumann - richtmyer - landshoff type av term velocity gradient @xcite can maintain the keplerian disk for @xmath141 orbits . we need to investigate the feasibility of these methods in structure formation simulations . we thank the anonymous referee . her / his insightful comments improve our manuscript greatly . t.r.s . thanks takashi okamoto and kohji yoshikawa for useful discussion . t.r.s . also thanks lucy kwok and junko saitoh who helped in writing the manuscript . numerical simulations were carried out on the cray xc30 system in the center for computational astrophysics at the national astronomical observatory of japan . this work is supported by grant - in - aid for scientific research ( 26707007 ) of japan society for the promotion of science and strategic programs for innovative research of the ministry of education , culture , sports , science and technology ( spire ) . this work is supported in part by mext spire and jicfus . the philosophy of the introduction of ac is to ensure all physical quantities will be smooth everywhere . at a contact discontinuity , ssph makes a pressure blip if the internal energy does not have a smoothed profile consistent with the smoothed profile of density . to guarantee the consistent profile of the internal energy , ac spreads the internal energy until the pressure profile becomes smooth . this term works very well in the standard tests and widely used . however , one meets a difficulty when one applies ac to a contact discontinuity consisting of a fluid with different chemical compositions . consider the case in which there are two contacting fluids and they have different chemical compositions . here , we assume that they have different mean molecular weights , @xmath142s . thus , the internal energy is a function of @xmath142 and temperature @xmath71 ; @xmath143 . moreover , these two fluids are initially in the pressure and temperature equilibrium state . disph can reproduce the original pressure and temperature equilibrium state without any difficulty . on the other hand , ssph can not reproduce it ; a pressure jump appears across the contact interface . ssph with the ac term gives a pressure equilibrium state by changing the internal energy of particles around the contact interface . since @xmath143 , this change of @xmath144 induced by ac is regarded as the changes of @xmath71 and @xmath142 . unfortunately , we do not know an appropriate procedure to spread @xmath71 and @xmath142 . hence , we can not predict exact quantities of @xmath142 and @xmath71 at the contact interface . there are two possible limits : ( 1 ) @xmath71 changes but @xmath142 does not , and ( 2 ) @xmath142 alters but @xmath71 does not . the former leads to undesirable chemical reactions if one is solving the chemical reactions , since they depend on @xmath71 . the latter is the compulsive mixing of chemical compositions . we show a simple , one - dimensional numerical experiment regarding this problem . the initial condition is as follows . the computational domain is @xmath145 with a periodic boundary condition . the fluid at @xmath146 has @xmath147 while the other has @xmath148 . both pressure and temperature have constant values in the whole region . since @xmath149 , the density and internal energy have different values in these two sub - domains . here the initial pressure and temperature are unity and @xmath150 . the density is normalized so that @xmath151 . we solved this system until @xmath152 sound crossing time . we used two schemes : one was disph and the other was ssph with the ac term . here , we adopt the ac term described in [ sec : results : ac ] and @xmath153 . figure [ fig : ac ] shows the results of ssph with the ac term and disph at time @xmath154 ( about 100 crossing time ) . in the ssph run with the ac term , the initial pressure jump is smeared by smoothing the distribution of the internal energy ( the top - right panel in figure [ fig : ac ] ) . if we consider this change of the internal energy as the change of the temperature , we have the temperature distribution with jumps at the contact discontinuities shown in the middle - right panel . on the other hand , if we regard it as the change of @xmath142 , we obtain the smoothed distribution of @xmath142 as shown in the bottom - right panel . in disph , there is no difficulty in dealing with this system ( panels in the left column ) . a system consisting of materials with different chemical compositions is not an unphysical situation . for instance , for a giant impact simulation , one solves collisions of core(iron)-mantle(silicon based rock ) planet embryos @xcite . at the core - mantle boundary , there is a jump of the composition . in addition , at the contact interface at the colliding time there is also a jump of the chemical composition . if we use the ac term for this simulation , the boundary materials are mixed and/or the temperature at these boundaries changes . disph for a non - ideal gas @xcite can deal with these boundaries without any difficulty @xcite . the time integration of a single step from @xmath155-th step to @xmath156-th step where the time step is @xmath157 is carried out by the following procedure : 1 . advance the velocity @xmath158 from @xmath155-th step to @xmath159-th step using the acceleration @xmath160 : @xmath161 2 . drift the position @xmath162 from @xmath155-th step to @xmath156-th step using @xmath163 : @xmath164 3 . evaluate the predictor of the velocity at @xmath156-th step using @xmath160 : @xmath165 4 . evaluate the predictor of the specific internal energy at @xmath156-th step using the time derivative of @xmath166 , @xmath167 : @xmath168 5 . compute @xmath169 and @xmath170 using @xmath171 , @xmath172 , and @xmath173 . advance @xmath163 by the rest of the half step using @xmath169 : @xmath174 7 . compute @xmath175 using both @xmath167 and @xmath176 : @xmath177 \delta t.\ ] ] the manner to integrate the internal energy described above is heun s method ( or the modified euler s method or the second order runge - kutta method ) . thus , our integration scheme is categorized as the second order method . for dm particles , this time integration scheme is reduced the standard second order symplectic scheme , the leap - frog scheme . as a demonstration , we show the results of the adiabatic collapse test with this time integration scheme . this test is one of the standard benchmark tests of self - gravitating fluid @xcite . a number of simulation codes showed results of this test ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? the typical error of the total energy in this test is less than @xmath178% . the initial condition of this test was generated as follows . first , we prepared gas particles located on uniform grid points . then , by stretching their relative distances from the center of the coordinates , we made the gas distribution with a radial density profile of @xmath179 , where @xmath180 is the distance from the center of the coordinates . the gas particles in @xmath181 are adopted to make a gas sphere . we assumed that they do not have a bulk velocity at the beginning of the simulation . the total mass and the gravitational constant are unity . the specific initial thermal energy is @xmath182 and @xmath55 . since the initial kinetic energy is zero and the thermal energy is less than the absolute value of the potential energy , this gas sphere collapses due to self - gravity and eventually reaches a state of equilibrium . here , we represented this gas sphere with the 61432 equal - mass sph particles . the gravitational softening length was set to @xmath183 . we carried out four runs . two are disph and ssph , which were used in this paper . in these runs , the kernel size was determined so that the neighbor number is in the range of @xmath58 . however , it is pointed out that the use of the constant neighbor number often makes sudden change of the interaction radius , resulting in the fluctuations of physical quantities . thus , the other two runs adopted the smoothed neighbor number method for the evaluation of kernel size @xcite . here , both disph and ssph were used . we express our fiducial cases as disph and ssph , and the other two cases with the smoothed neighbor number method as disph(s ) and ssph(s ) . the wendland kernel c4 was adopted . the self - gravity was solved by using the tree with grape method . the opening angle was set to @xmath54 . time step for each particle is determined following @xcite . figure [ fig:3dcollapse : evolution ] shows the evolution of radial profiles of density , pressure , and radial velocity of the disph run . three representative epochs are adopted . except for the central region ( @xmath184 ) , our result is comparable with a 1d ppm result . the other three cases , ssph , disph(s ) , and ssph(s ) , show almost identical results . in figure [ fig:3dcollapse : all ] , we show time evolutions of the kinetic , thermal , potential and total energies . again , we show the result of the disph run . the overall evolution is comparable with that obtained in previous studies and the other three cases show identical results . figure [ fig:3dcollapse : total ] represents time evolutions of the total energies in four runs . these total energies change significantly at time @xmath185 when the gas sphere collapses , while they keep almost a constant value at other times . we define the energy error as follows : @xmath186 where @xmath187 and @xmath188 are the total energy of a system at time @xmath189 and @xmath190 , respectively . energy errors evaluated by the above equation are 0.19% for disph , 0.22% for ssph , 0.14% for disph(s ) , and 0.16% for ssph(s ) , respectively . thus our time integration scheme works well . the fact that there is no significant difference among them indicates that the choices of the definition of the neighbor numbers and sph types are not crucial for the energy conservation in our simulations . in order to compare sizes of entropy core and their time evolutions among different simulations , we need a methodology to evaluate a core size from an entropy profile . as is shown in the text , the evolutions of core size and entropy are highly time - variable . in order to follow this dynamically evolving structure , we adopt the technique used in the field of stellar cluster simulation . below , we describe our method to define the entropy core . first , we calculate the radially averaged entropy profile ; it is expressed as @xmath191 to @xmath65 from the density center . the constant interval with the logarithmic scale we calculate the entropy of each particle and add it to a bin which covers the position of the particle . we obtain the radial entropy profile of the cluster by taking the average of entropy in each bin . the averaged entropy at the given @xmath192 is calculated by the linear interpolation of the binning values . if @xmath193 , we use @xmath194 as the averaged entropy at the given @xmath192 . ] , where @xmath192 is the distance from the density center , @xmath195 . we , then , pick up particles of which entropy is in the range of @xmath196 and evaluate the core radius using these particles . the reason why we sampled particles from the narrow region is to exclude the contamination of small , low - entropy clumps ( see also figure [ fig : entropy : merger : minor ] ) . without this process , we were unable to obtain a reasonable size of the entropy core . the following equation is used in order to evaluate the core radius : @xmath197 where @xmath198 is the entropy of particle @xmath199 and @xmath200 . the index @xmath199 runs particles of which entropy are in the range of @xmath201 and @xmath202 . we use a very strong power of the inverse of the entropy as the weight function to determine the entropy core . the outer part of the entropy profile of clusters of galaxies , typically in the range of @xmath203 , follows @xmath204 @xcite . therefore , a weight which is stronger than @xmath205 is suitable in order to evaluate of the core . after intense tests , we found that the weight , @xmath206 , is suitable in order to evaluate the core radius of the radial entropy profile . we adopt the subfind algorithm @xcite in order to extract self - bound clumps in the santa barbara cluster . the procedure is as follows . first , we calculate the density of all particles within @xmath207 using eq . the density of gas particles is evaluated by counting the contribution of gas particles while that of dm particles is calculated by counting the contribution of dm particles . the density of each particle is evaluated by the contribution from the nearest @xmath58 particles . here , we refer these nearest particles `` local particles '' . after the density evaluation , we sort all of particles in the descending order of the density . we , then , identify isolate clumps following the density distribution . we repeatedly compare the densities of particles with those in the densities of their local particles from the highest to the lowest density particles and apply the following operation . here , we focus a particle @xmath25 . if the particle @xmath25 is the highest density particle in its local particles , we set the particle @xmath25 as the peak of an isolate clump ( the core of a new clump ) . if the particle @xmath25 is the second highest density particle or it has two higher density particles which are in the same clump , we add the particle @xmath25 to the clump which includes the highest density particle . if the particle @xmath25 has more than two higher density particles and the two highest density particles are in different clumps , we tag these two clumps as _ the subfind clump candidates _ since the particle @xmath25 is at the saddle point which separate two clumps ( two density peaks ) . then , we merge these clumps with a single larger clump and attach particle @xmath25 to this clump . the last operation is necessary in order to extract subfind clump candidates which are located at the envelop of this merged clump . as the final step , we remove unbound particles from the subfind clump candidates iteratively . we first calculate the total energy ( the self - gravitational potential and the kinematic energy where the mean velocity of each clump is set to zero ) of each particle in a clump . then if the clump includes particles of which total energy is larger than zero , we remove these particles from the clump . we do not remove more that @xmath50 of the particles at once so that we can carry out this process stably . we continue this procedure until the clump consists only of the bound particles . we adopt clumps which contains more than 64 bound particles . we apply this procedure to all clumps and get a full catalog of subfind clumps . in this section we investigate the mass functions of substructures . the subclumps are extracted using the subfind algorithm ( see [ sec : method : subfind ] ) . figure [ fig : massfunction ] shows mass functions of substructures . mass functions of the runs with the same resolutions have almost identical profiles . all of mass functions asymptotically follow the power law profile with the index of @xmath208 . it is not surprising that there is no impact on substructure mass functions with different schemes , since the amount of the baryonic component is @xmath209 of the total mass . mass functions of the baryonic component is shown in figure [ fig : gasmassfunction ] . not all of substructures hold the baryonic component due to the stripping . thus , the number of baryonic substructures is very small compared to that of the total substructures . even though the evolution of the gas clumps in the central area of the cluster is quite different for two runs , this difference is little or no effect on the mass function . this is probably because in either case that lifetime of gas clump in the central area is short . our mass functions do not show convergence with different mass resolutions . we have checked mass functions with mixed resolutions and found that , at least toward the high mass end , the functional form of mass functions depends on the mass resolution . thus , the discrepancy between high - mass resolution runs and low - mass resolution runs comes from the non - linear effect induced by the smallest scale density fluctuations .

the nature of some grb host galaxies has been investigated by means of chemical evolution models of galaxies of different morphological type following the evolution of the abundances of h , he , c , n , o ,-elements , ni , fe , zn , and including also the evolution of dust . by comparing predictions with abundance data , we were able to constrain nature and age of grb hosts . we also computed a theoretical cosmic dust rate , including stellar dust production , accretion and destruction , under the hypotheses of pure luminosity evolution and strong number density evolution of galaxies . we suggest that one of the three grb hosts is a massive proto - spheroid catched during its formation , while for the other two the situation is more uncertain , although one could perhaps be a spheroid and the other a spiral galaxy . we estimated the chemical ages of the host galaxies which vary from 15 to 320 myr . concerning the cosmic effective dust production rate in an unitary volume of the universe , our results show that in the case of pure luminosity evolution there is a first peak between redshift z=8 and 9 and another at , whereas in the case of strong number density evolution it increases slightly from z=10 to and then it decreases down to z=0 . finally , we found tha the total cosmic dust mass density at the present time is : in the case of pure luminosity evolution and in the case of number density evolution . [ firstpage ] supernovae gamma - ray bursts galaxy evolution .

1407.6905

most lgrbs are associated with massive stellar explosions and consequently they are linked to ongoing star formation . savaglio et al . ( 2009 ) , studying a sample of grb host galaxies ( hgs ) , concluded that there is no clear evidence that hgs are peculiar galaxies and that they are instead similar to normal star - forming galaxies , both in the local and in the early universe . for this reason , understanding the nature of hgs provides constraints on the grbs progenitors : the collapsar model ( woosley 1993 ; woosley & heger 2006 ) predicts a low metallicity environment , typical of dwarf galaxies , because of the reduced angular momentum loss and mass loss via strong winds at the surface of the star , while in other progenitor models there is no such an assumption ( e.g. ouyed et al . 2005 ; fryer & heger 2005 ) . many attempts have been made to characterize the hosts of grbs ( e.g. fruchter et al . ( 1999,2006 ) , castro cern et al . ( 2006 , 2010 ) , le floch et al . ( 2006 ) , modjaz et al . ( 2008 ) , levesque et al . ( 2010 ) , perley et al . ( 2013 ) and boissier et al . in particular , berger et al . ( 2003 ) and le floch et al . ( 2006 ) , using mid - ir and radio observations , showed that some grb host galaxies are characterized by a large amount of star formation obscured by dust and in general they have attempted to understand the properties of grb hosts , but only few of them have used powerful chemical evolution indicators as we are going to do in this work . moreover , the importance of dust is also confirmed by the large fraction of grbs occurring in dusty environments ( e.g. perley et al . 2009 , 2013 ; greiner et al . 2011 ; hatsukade et al . 2012 ) . dust is present in almost every astrophysical environment affecting the comprehension of their physical properties , such as their star formation rate and abundance ratios . dust grains are made up of metals , primarily c , o , mg , si , fe , ni , s , ca which enriched the interstellar medium ( ism ) during particular stellar evolutionary phases ( e.g. asymptotic giant branch ( agb ) stars and core collapse supernovae . interesting dust evolution studies in galaxies have been performed by dwek ( 1998 ) , dwek et al . ( 2007 , 2011 ) , calura et al . ( 2008 ) and schurer et al . ( 2009 ) . in those papers , models of galactic chemical evolution including dust were presented . the basic idea of identifying grb host galaxies by means of chemical evolution models is based on the fact that galaxies of different morphological type should show different abundance patterns . such different patterns derive from their different star formation histories ( sfhs ) . calura et al . ( 2009 ) presented a first approach were the sfhs of the hgs of grbs were derived by comparison between observed abundances and model results ; to do that they adopted the calura et al . ( 2008 ) models including dust evolution . in the following years , fan et al . ( 2010 ) also deduced the nature of some hosts of grbs using chemical models but the dust was not considered . both those papers concluded that the hosts of the studied grbs were probably irregular galaxies . clearly , the presence of dust can affect the abundances derived from the gas in galaxies and therefore it is very important to include the possibility of dust condensation in chemical evolution models . in this paper , by adopting the dust evolution as in calura et al . ( 2008 ) , we aim at constraining the nature and sfh of some hgs by means of detailed chemical evolution models taking into account the variation in the abundance patterns . moreover , we aim at computing the dust evolution in each type of galaxy , including processes of stellar dust production , accretion of dust onto preexisting grains and dust destruction . as a byproduct , we are going to compute the cosmic effective dust production rate in an unitary volume of the universe by taking into account the contributions from all galaxies . section [ models ] describes the chemical evolution models used in this work and the nucleosynthesis prescriptions ; in section [ dust ] the prescriptions for the supernova rates and the dust production , accretion and destruction rates in galaxies are presented . in section [ sample ] we briefly show the observational data . in section [ res ] the results for dust production , accretion and destruction in galaxies of different morphological type as well the results deriving from the comparison between our theoretical predictions and observational abundance data of grb 081008,120327a and 120815 host galaxies are shown . the definition and computation of the cosmic effective dust production rates , in different frameworks of galaxy formation , are presented in section [ cosmic ] , together with the calculation of the present time cosmic dust density . finally , in section [ concl ] , some conclusions are drawn . we follow the evolution of the abundances of several chemical species and of the dust content of spirals , elliptical and irregular galaxies using detailed self - consistent chemical evolution models able to reproduce the majority of the properties of the different types of galaxies . in all models the instantaneous recycling approximation is relaxed and the stellar lifetimes are taken into account . the models all assume that galaxies form by gas accretion which accumulates a final mass called infall mass ( @xmath5 ) . the model takes into account the feedback from sne and stellar winds giving rise to a galactic outow when the gas thermal energy exceeds the binding energy of gas . in our models the sfr is a simple schmidt s ( 1963 ) law : @xmath6 where the quantity @xmath7 is the star formation efficiency ( sfe ) , namely the inverse of the typical time - scale for star formation , expressed in @xmath8 , and it has been varied as a function of the galaxy morphological type ( see section 5 ) . @xmath9 is the gas mass fraction expressed as @xmath10 , the ratio between the gas mass and the infall mass . dividing by the infall mass it is only for a normalization purpose and all the other variables are also normalized in the same way . because of this , @xmath11 . all the models used in this paper consider only one gas phase . the chemical evolution equation for a given chemical element i takes the following form : @xmath12 where @xmath13 is the fractional mass of the element _ i _ at the time _ t _ in the ism . , and @xmath14 represents the mass fraction of element _ i _ in the gas , i.e. the abundance by mass . the first item on the right hand side of the equation , @xmath15 , represents the mass change of the element _ i _ caused by the formation of new stars ; @xmath16 corresponds to the material returned through stellar winds or sn explosions in the ism . the last two terms take into account the infall of primordial gas @xmath17 and the outflow @xmath18 , respectively . in particular , the accretion rate of an element _ i _ follows the equation : @xmath19 where @xmath20 represents the abundance of the element _ i _ in the infalling gas that we assume to be primordial , @xmath21 is the normalization constant constrained to reproduce the present time total mass ( gas plus stars ) , i.e. @xmath22 , and @xmath23 is the accretion timescale expressed in @xmath24 and defined as the characteristic time at which half of the total mass of the system has been assembled . the rate of gas lost via galactic winds for each specific element _ i _ is assumed to be proportional to the amount of gas present at the time _ t _ through the free parameter @xmath25 : @xmath26 we compute in detail the contributions by low mass stars , type ia and type ii sne to the chemical enrichment of the ism . the main physical quantities related to these processes are the stellar yields , representing the amount of both newly formed and pre - existing elements ejected into the ism by stars of all masses at their death . the yields used in this paper are as follows : 1 . for low and intermediate mass stars , lims ( i.e. with masses @xmath27 ) we make use of the yields provided by van den hoek @xmath28 groenewegen ( 1997 ) as a function of metallicity . 2 . for massive stars ( i.e. with masses larger than @xmath29 ) and type ia sne , we adopt the empirical yields suggested by franois et al . ( 2004 ) , which are substantially based on the woosley & weaver ( 1995 ) and iwamoto et al . ( 1999 ) yields , respectively , and are tuned to best fit the abundances in the solar neighborhood . according to dwek ( 1998 ) , calura et al . ( 2008 ) and pipino et al . ( 2011 ) , the equation describing the evolution of the element _ i _ in the dust is similar to the equation [ chem - ism ] . let us define @xmath30 as the abundance by mass of the element @xmath31 at the time @xmath32 locked up in the dust ; the quantity @xmath33 represents the normalized mass of the element @xmath31 at the time @xmath32 in the dust . by means of the following equation it is possible to understand how a specific chemical element evolves in the dust : @xmath34 where , * @xmath35 describes the enrichment rate of the element @xmath31 where each term of the equation corresponds to a contribution coming from a different progenitor : @xmath36 dm \\ % \nonumber \\ & + ( 1-a)\int_{m_{b_m}}^{8 m_{\odot}}\psi(t-\tau_{m } ) \delta^{sw}_{i } q_{mi}(t-\tau_m)\phi(m)dm + \\ % \nonumber \\ & ( 1-a)\int_{8 m_{\odot}}^{m_{b_m}}\psi(t-\tau_{m } ) \delta^{ii}_{i } q_{mi}(t-\tau_m)\phi(m)dm \\ % \nonumber \\ & + \int_{m_{b_m}}^{m_u}\psi(t-\tau_m ) \delta^{ii}_{i } q_{mi}(t-\tau_m ) \phi(m)dm % \nonumber\\ \end{aligned}\ ] ] + the first term on the right takes into account the enrichment due to stars in the @xmath37 $ ] mass range , where @xmath38 is the minimum mass contributing , at a given time @xmath32 , to chemical enrichment and @xmath39 is the minimum total binary mass allowed for binary systems giving rise to type ia sn ( @xmath40 ) , while the mass @xmath41 represents the maximum total mass for a binary system to give rise to a sn ia ( @xmath42 ) ( see matteucci & greggio 1986 ) . the initial mass function ( imf ) used is the salpeter ( 1955 ) one : @xmath43 the quantity @xmath44 ( where @xmath45 is the lifetime of a star of mass @xmath46 ) contains all the information about stellar nucleosynthesis for elements either produced or destroyed inside stars or both ( talbot and arnett 1973 ) . the second term represents the contribution from binary systems that become type ia supernovae ; @xmath21 is the unknown fraction of binary stars giving rise to snia and is fixed by reproducing the observed present time snia rate ; @xmath47 indicates the lifetime of the secondary star of the binary system and consequently , the explosion timescale is taken into account . the quantity @xmath48 is the ratio between the secondary component of the binary system ( i.e. the less massive one ) and the total mass of the system @xmath49 and @xmath50 is the distribution function of this ratio . the third and fourth terms represent the contribution to dust from single stars in the mass range @xmath51 and @xmath52 where there are binary systems producing snia , single agb stars and core - collapse sne . * @xmath53 is the rate at which an element _ i _ is removed from the ism gas phase by accretion onto preexisting dust particles in molecular cloud , while @xmath54 is the timescale for the accretion process that will be described in section [ accr ] . * @xmath55 rapresents the amount of the element _ i _ returned to the ism gas phase , owing to the dust destroyed by sne , on the timescale of destruction @xmath56 ( see [ destr ] ) . * @xmath57 takes into account possible ejection of dust into the inter galactic medium ( igm ) by means of galactic winds triggered by sne explosions . the wind starts when the thermal energy of gas equates the binding energy of gas . in our calculations , we assume a differential wind in which the metals are preferentially lost ( see bradamante et al . the quantities @xmath58 represent the dust condensation efficiencies as defined by dwek ( 1998 ) and are relative to the low and intermediate mass stars which eject dust through stellar winds ( sw ) , the type ii sne ( ii ) and the type ia sne ( ia ) . for the model of the ellipticals we assume that after the onset of the galactic wind the star formation is quenched and the galaxy starts evolving passively , whereas for spirals and irregulars the star formation continues even after the wind . the evolution of dust is strongly dependent on the sn rates ( snrs ) . in our computation we assume , as contributors to the total dust amount of a galaxy , both core collapse ( cc ) and type ia supernovae . ccsne are divided in type ii and ib / c sne ; in particular type ii sne , originate from single stars with mass larger than @xmath59 while sne ib / c come from wolf - rayet ( wr ) stars , namely stars which have lost most of their h and he envelope and with masses larger than @xmath60m_{\odot}$ ] , whose value depends on the initial stellar metallicity . it has been suggested that they could be both single wr stars and massive stars in binary systems in the mass range 14.8 - 45@xmath61 , ( yoon et al . 2010 ; grieco et al . owing to the larger mass of the progenitors , ccsne explode on short timescales : from few million to several tenths of million years . type ia sne are believed to originate from low and intermediate mass stars , in particular from c - o white dwarfs in binary systems . in particular , to describe type ia supernova progenitors , we assume the single degenerate ( sd ) scenario by whelan & iben ( 1973 ) . in this scenario , a c - o white dwarf accretes mass from a non - degenerate companion until it reaches the chandrasekhar mass ( @xmath62 ) and explodes via c - deagration , leaving no remnant . type ia sne end their lives on timescales ranging from @xmath63 up to an hubble time or more and consequently the snia rate is related to the past star formation history ( sfh ) of a galaxy , whereas the ccsne reflect the birth rate of massive stars , i.e. the galactic star formation rate ( sfr ) . for this reason , different types of galaxies have different snrs . the total snr used in our models is the sum of the three individual rates , that may be approximated as : @xmath64 @xmath65 @xmath66dm_{b}\ ] ] the sn ib / c rate is composed by a first term evaluating the single progenitor channel and a second integral that takes into account the contribution from wrs in binary system . the factor f represents the fraction of massive binary stars producing type ib / c sne . this parameter is chosen to be equal to 0.15 , ( see calura & matteucci , 2006 ) . this value is motivated by the facts that half of the massive stars are possibly in binary systems , and the fraction of massive stars in close binary system is @xmath67 , i.e. similar to the close binary frequency predicted for low mass systems ( jeffries & maxted , 2005 ) . therefore , the estimated value for this parameter is @xmath68 . there are still many uncertainties in the yield of dust from sne and agb stars . the general definition of dust grains divide the elements that suffer depletion into dust ( i.e. refractory elements ) into silicate dust , composed by o , mg , si , s , ca , fe and ni , and into carbon dust , composed only by c. following dwek ( 1998 ) the dust sources considered in our work are : * _ low and intermediate mass stars ( lims ) _ : in these stars , dust is produced during the agb phase . we assume that dust formation depends mainly on the composition of the stellar envelopes . if @xmath69 and @xmath70 represent the o and c mass fractions in the stellar envelopes , respectively , we assume that stars with @xmath71 @xmath72 1 are producers of silicate dust , i.e. dust particles composed by o , mg , si , s , ca , ni , fe . on the other hand , c rich stars , characterized by @xmath71 @xmath73 1 , are producers of carbonaceous solids , i.e. carbon dust ( draine 1990 ) . being @xmath74 and @xmath75 the total ejected mass and the dust mass formed by the stars as functions of the initial mass @xmath46 for the element @xmath31 , respectively , we assume that for stars with @xmath71 @xmath73 1 : + + @xmath76\\$ ] + with @xmath77 and + + @xmath78 + for all the other elements . for stars with @xmath71 @xmath72 1 in the envelope , we assume + + @xmath79 + + @xmath80 + + with @xmath81 for mg , si , s , ca , ni , fe and + + + @xmath82 + + with @xmath83 being the mass of the @xmath31 element in atomic mass units . * _ supernovae ( sne ) _ : sne are potentially the most important source of interstellar dust . core collapse supernovae ( see section [ snrate ] and in particular type ii supernovae ( sne ii ) are considered to be one of the main grain production sources in the universe ( e.g. nozawa et al . 2003 ) . assuming that all the refractory elements precipitate with a mean efficiency of @xmath84 , a typical @xmath85 sn can produce about @xmath86 of dust ( woosley & weaver 1995 ; nomoto et al . the first evidence for dust condensation was observed in sn 1987a ( danziger et al . 1989 , lucy et al . 1989 ) ; formation of dust was then observed in several ccsne , i.e. sn 2003gd ( meikle et al . 2007 ) , 2004et ( kotak et al . 2009 ) , 2004dj ( szalai et al . 2011 ) , 2007od ( andrews et al . 2010 ) and type ib / c sn 2006jc ( nozawa et al . 2008 ; mattila et al . 2008 ; tominaga et al . 2008 and sakon et al . 2009 ) , with an estimated mass of recently formed dust between @xmath87 . + from a theoretical point of view , the amount of dust formed after a ccsn explosion was evaluated by the models of kozasa et al ( 1989 ) , todini & ferrara ( 2001 ) and nozawa et al . ( 2003 ) in the range @xmath88 . this prediction was confirmed by numerical models of bianchi & schneider ( 2007 ) , kozasa et al . ( 2009 ) and silvia et al . the discrepancy between observations and theories is still unclear and also the study of sn remnants could not solve the question of the amount of dust produced by sne . using far - infrared and sub - millimeter data , many groups estimated the amount of dust in cas a ( dunne et al . 2003 ; krause et al . 2004 ; rho et al . their results varied between @xmath89 and @xmath90 , while the values for kepler snr showed differencies from @xmath91 ( morgan et al . 2003 ) down to @xmath92 ( blair et al . recently , matsuura et al . ( 2011 ) report far - ir and sub - mm observations of sn1987a in the large magellanic cloud . the observations revealed the presence of cold dust grains with a temperature of @xmath93 . the intensity and spectral energy distribution of the emission suggests a dust mass of @xmath94 . it has been supposed that not only ccsne but also sneia can be possible producers of dust grains , especially fe grains ( tielens 1998 ) . + following calura et al . ( 2008 ) , we assume the same prescriptions for both sne types : + m_dust , c(m ) = ^ia , ii_c[m_c , ej(m ) ] + + with @xmath95 ; + m_dust , i(m ) = ^ia , ii_im_i , ej(m ) + + with @xmath96 for fe , ni + @xmath97 for mg , si , ca and + with @xmath98 for s ; + @xmath99 + there are still many uncertainties on the amount of dust produced by type ia sne . however , gomez et al . ( 2012 ) , on the basis of herschel pacs and spire photometry at @xmath100 have reported the existence of dust in two type ia sn remnants : kepler and tycho . in particular , they detected a warm dust component in kepler s remnant with temperature @xmath101 and mass @xmath102 ; similarly for the tycho s remnant , they detected warm dust at @xmath103 with mass @xmath104 . dust grains can grow by accretion of metals present in the ism onto preexisting grain cores within dense molecular clouds ( dwek 1998 , inoue 2003 ) . there are more observations indicating the existence of large , micrometre - sized dust grains in dense molecular clouds and suggesting a need for significant dust growth in the early universe ( pagani et al . 2010 , michalowski et al . moreover , dust growth appears to dominate over dust destruction also in the local , present - day universe ( hirashita 1999 ; inoue 2003 ; hirashita & kuo 2011 ) ; other evidence for dust accretion comes from the observed infrared emission of cold molecular clouds ( flagey et al . 2006 ) . following dwek ( 1998 ) , we consider that the key parameter in dust accretion is the dust accretion timescale that , for a given element @xmath31 , can be expressed as an increasing function of the dust mass : @xmath105 where @xmath106 for the timescale @xmath107 , typical values span from @xmath108 yr , of the order of the lifetime of a typical molecular cloud , up to @xmath109 yr . in this paper , we assume that the timescale @xmath107 is constant for all elements , with a value of @xmath110 yr . + the main mechanism for dust destruction is by sputtering in the high interstellar shocks driven by sne ( mckee 1989 ; jones et al 1994 ) . following dwek et al . ( 2007 ) , the dust - destruction timescale is independent of the dust mass and equal to : @xmath111 where @xmath112 is the gas mass density , @xmath113 is the effective ism mass that is completely cleared of dust by each sn event , and @xmath114 is the sum of core collapse and type ia sn rates . dwek et al . ( 2007 ) consider the grain destruction efficiency , corresponding to the range of uncertainty in the lifetime of the interstellar dust grains , as an unknown , and adopt @xmath113 as a free parameter of the model . regarding the milky way like galaxies , for values of @xmath115 , corresponding to the average density of the galactic ism , they get @xmath113 = 1100 - 1300 @xmath61 for an equal mix of silicate and graphite grains . in order to constrain the sf history of the hgs we have chosen from the literature the following hgs with detailed observed abundance data : grb 081008 ( delia et al . 2011 ) , 120327a ( delia et al . 2014 ) and 120815 ( kruhler et al . 2013 ) . in table [ grb_data ] are shown the observational data for each hgs studied in our analysis . concerning the [ o / fe ] and [ al / fe ] abundances of grb 081008 ( first column ) , these values are considered as lower limits because of the lines saturation ( see delia et al . 2011 for more details ) . the same problem is in the c , o and al abundances in fact in the delia et al . ( 2014 ) paper , the authors warn the reader explaining that despite the highly saturated , cii @xmath1161334 , oi @xmath1161039 , oi @xmath1161302 and alii @xmath1161670 transitions have not been excluded from the analysis , because they are the only lines describing cii , oi and alii . .observational data collected for grb 081008 ( delia et al . 2011 ) , 120327a ( delia et al . 2014 ) and 120815 ( kruhler et al . 2013 ) . note that all the abundances have been normalized to the same solar abundances , namely those of asplund et al . ( 2009 ) . [ cols="^,^,^,^",options="header " , ] as the dust content in the universe is the result of different and competitive processes , we study the individual contribution of accretion , destruction and production in each morphological type of galaxy using the model e00 with @xmath117 and sfe=@xmath118 for an elliptical , the model s00 with @xmath119 and sfe=@xmath120 for a spiral and finally the model c00 with @xmath121 and sfe=@xmath122 for an irregular galaxy . we focus on the effects that different star formation histories have on the evolution of the chemical abundances in the ism and how these abundances are affected by dust in galaxies of different morphological type as well as on the different competitive rates involving the dust . in particular , understanding how the dust rates evolve in time is important in the computation of the global amount of dust and the cosmic rate inferred from our models . the main assumption concerning the dust rate is on the dust yields describing the fractions of each element followed in our chemical evolution code , which is condensed into dust and restored into the ism by low and intermediate mass stars , type ia and ii sne . the prescription used to infer the dust production rate are summarized in table [ models - param ] . the rate of dust production in units of @xmath123 is shown in [ prodrate ] and describes the amount of dust produced during the whole galactic evolution for the elliptical , spiral and irregular galaxy model . in figures [ accrrate ] and [ destrrate ] are shown the rates of accretion and destruction , respectively , in unit of @xmath123 as functions of time for the same models shown in figure [ prodrate ] and with the same color set . as one can see from figures [ prodrate ] , [ accrrate ] and [ destrrate ] , the rate of dust production in ellipticals is maximum at the beginning of their evolution but also the destruction rate is maximum since they both depend on the sfr , which is assumed to be very high in the first phases of the evolution of these objects ( see fig the dust accretion rate instead is quite high in spirals when the sfr reaches a maximum , whereas accretion is active in ellipticals only during the starburst and it stops after the occurrence of a wind which devoids the galaxy of gas and dust . on the other hand , the rate of destruction of dust in ellipticals continue even after the star formation has stopped , since it depends on sn explosions and type ia sne are active until the present time even in absence of star formation . these different behaviours are responsible for the shape of the cosmic effective dust production rate that will be described in the next section . in order to understand the different behaviour of the chemical evolution models developed here , we compare our model results with the abundances measured in the hosts of the three grbs considered . our main goal was to derive the history of star formation and therefore the nature of the hosts comparing the predictions of [ x / fe ] vs.[fe / h ] for several chemical elements ( c , n , o , mg , s , si , ni and zn ) . among these species , n and zn are not affected by dust while all the others , together with iron , contribute to the total amount of dust . in figure [ 081008 ] , we show the results for the host of grb 081008 at @xmath124 studied by delia et al . ( 2011 ) . in order to assess the importance of dust in this kind of analysis , in figure [ 081008 ] ( only ) , are shown the predictions for @xmath0 -elements , zn and ni relative to fe both in the models with dust ( left column ) and without dust ( right column ) . it is worth noting that while the ratio [ si / fe ] does not change in the two cases due to the fact that both elements are depleted in dust , for other ratios , such as [ zn / fe ] the situation is different . for example , the curve relative to the irregular galaxy is much lower in the case without dust depletion . the reason for that resides in the fact that zn is not refractory , whereas fe is . in particular , in the irregular model the amount of fe in dust is large due to the less efficient dust destruction process relative to the other galaxy types . this is because the destruction rate depends on the sfr which is quite low in irregulars.therefore , this implies that at any given metallicity the [ zn / fe ] ratio increases relative to the case with no dust depletion . on the other hand , in the spiral and elliptical models , the [ zn / fe ] is not much altered since they are characterised by a dust destruction rate which is at least comparable to the accretion rate . in the case without dust the models are allabove the data and this trend suggests that the subtraction of fe by dust is necessary to explain the data . looking at all the abundance ratios together in figure [ 081008 ] , we can not derive a firm conclusion on the nature of the host of this grb : in fact , while the [ ni / fe ] ratio is inconsistent with all curves , the [ si / fe ] and [ zn / fe ] seem to indicate a spiral galaxy . in figure [ 120327a ] are shown the predictions and the data for the host of grb 120327a at @xmath125 studied by delia et al . ( 2014 ) . in this case we show only the models with dust for several elements . concerning n and zn it is possible to see an inversion in the models : the irregular galaxy model ( c00 ) predicts larger n and zn abundances with respect to the fe than the other two models . to understand the reason of this behaviour , we need to remember that n and zn are not affected by dust while fe is strongly depleted . indeed , the higher is the star formation efficiency of a galaxy , the higher is the dust grain destruction . therefore , the model with lower sfe ( green solid line ) predicts higher abundances in these two particular cases . the trends of [ si , s , zn / fe ] vs. [ fe / h ] seem to suggest that the host galaxy is a system with high mass and sfr , namely either a spiral or a proto - spheroid . concerning the [ n / fe ] ratio it should be said that the n yields are still uncertain and in particular the secondary / primary nature of this element . in particular , here n is considered secondary from massive stars , while there are suggestions that n could be produced in a primary fashion in massive stars but only at very low metallicity ( see chiappini et al . if this is true , it could change the shapes of the models . finally , in figure [ 120815 ] , we report the results for the host of grb 120815,at @xmath126 observed by kruhler et al . ( 2013 ) ; the elements analyzed here are si , s , ni and zn and all seem to indicate again a massive spheroid with strong sf . in fact , to best fit the abundances in this host we adopted model e01 with sfe=20@xmath127 , which is the maximum value we explored . it is interesting to note the high value of [ zn / fe ] which means that a higher quantity of dust is present along the line of sight of this grb . in our analysis , although many uncertainties are still present in the adopted stellar yields and in the observational data and no firm conclusions on the grb hosts can be drawn , two of the three grb host galaxies taken into account seem to favor either a spiral or a spheroidal host , while the third grb ( grb120815 ) seems to be more confidently hosted by a proto - spheroid . this result , if confirmed in the future by more data , would be in contrast with the claim that most of the hgs are small , star - forming and metal - poor galaxies ( see , e.g. , christensen et al . 2004 , fruchter et al . 2006 and levesque et al . probably , the apparent discrepancy is due to the galaxy sample used to infer the hgs properties , composed mainly of low redshift objects ( @xmath128 ) . in particular , recent works based on high redshift surveys seem to suggest a more complex view ( e.g. kruhler et al . 2011 ; hunt et al . 2011 ; savaglio et al . 2012 ; elliott et al . 2013 , perley et al . 2013 ) , where the majority of the hgs seem to suffer strong star formation . this result is not unexpected recalling that grbs are associated with the death of massive and short - lived stars that trace the star formation inside the host . the explanation lies in the evolution / changing of the star forming tracers as a function of redshift : in the local universe the star formation is mainly traced by small , blue galaxies , while at early epochs it is dominated by massive spheroids and elliptical galaxies . further evidence in favor of this hypothesis comes from the study of the so called dark grbs . these grbs are found at high redshift and they have in general a high metallicity . this is a very interesting fact suggesting that these grbs arise from active star formation in high redshift massive spheroids . since ellipticals show in general old stars , that means that we are catching ellipticals in formation . the characteristics behavior of [ @xmath0/fe ] ratios in spheroids shows a long plateau where they are overabundant relative to fe and the sun . this is due to the time - delay model which explains the behaviour of abundance ratios on the basis of the production of different elements in different stars with different timescales . in particular , @xmath0-elements are mainly produced in ccsne restoring them on short timescales , whereas fe is produced mainly in type ia sne and restored on a large range of timescales going from 30 myr to a hubble time ( matteucci 2001 ; 2012 ) . this implies that , if the sfr is very efficient , the ccsne can produce a substantial amount of fe by themselves and when type ia sne start restoring the bulk of fe the abundance of this element in the ism is already high . this produces high [ @xmath0/fe ] ratios at high metallicity . the contrary occurs when the sfr is quite low ; the enrichment of fe from core - collapse sne is very modest and when type ia sne start to restore the bulk of fe , the abundance of this element in the ism is still quite low . this produces low [ @xmath0/fe ] ratios at low metallicity . therefore , the diagrams [ x / fe ] vs. [ fe / h ] can be in principle very useful tools to infer the nature and the ages of unidentified objects . in the chemical evolution models , the abundance ratios between two elements formed on different timescales can be used as cosmic clocks@xmath129 and provide us with information on the roles of lims and sne in the enrichment process ( matteucci 2001 ) . as already mentioned , the study of abundance ratios such as [ @xmath0/fe ] is quite useful , since the @xmath0-elements are produced on short timescales by type ii sne , whereas the fe - peak elements and nitrogen are produced on long timescales by type ia sne and low and intermediate - mass stars , respectively ; we use the abundance ratios of elements formed on different timescales to set important constraints on the age and the nature grb hgs . we define the age of the galaxy @xmath130 as the time passed between the galaxy formation epoch and the grb observation time ; this gives us a quantitative measure of the time required for the galaxy to evolve and reach , at the time of observation of the grb , the chemical abundances observed during the afterglow episode . practically , it is possible to infer it from the simultaneous comparison between the [ @xmath0/fe ] vs [ fe / h ] and the same ratio as a function of the galactic time , or redshift , in a @xmath131cdm cosmological framework . applying the method to the three grbs analyzed in this paper , we find an age of 50 myr for the galaxy hosting the grb 120327a , 0.32 gyr for the grb081008 host and 15myr for the grb120815 host . these young ages indicate that we find grbs to be `` young '' galaxies , namely galaxies in which the star formation is very active , probably suffering their first major star formation episode , but they are not necessary typical of the main population of galaxies that could have started forming stars already much before . the cosmic rate is defined in an unitary comoving volume of the universe . this definition is necessary to study the rate at high redshift where the morphology of the observed galaxies is not known . the cosmic rate refers , in fact , to a mixture of galaxies which can be different at every redshift . in particular , in order to evaluate the amount of dust observed at high redshift , we have to take into account the cosmic dust cycle : dust is produced in stars and it is then blown off in a slow wind or a massive star explosion . the dust is then recycled in the clouds of gas between stars and some of it is consumed when the next generation of stars begins to form . we compute the cosmic dust rate in units of @xmath132 by adopting the same method applied to the computation of cosmic star formation rate , as described in the paper of grieco et al . ( 2012a ) . in that paper the cosmic sfr ( csfr ) has been computed by taking into account the sf histories of galaxies of different morphological type as those shown in fig . [ sfr ] . in particular the csfr was defined as : @xmath133 where the sfrs have been then convolved with the number density for each type of galaxy . the quantity @xmath134 is the galaxy number density , expressed in units of @xmath135 , and @xmath136 identifies a particular galaxy type . in the grieco et al . ( 2012a ) paper , the main assumption was that the galaxy number density is constant in time and equal to the present time one for each galaxy type . here we define , in the same way , the cosmic dust rate ( cdr ) : @xmath137 \cdot n_{k}^ { * } } , \ ] ] where @xmath138 is the total production rate of all the elements considered in the dust model ( o , mg , si , s , ca , ni , fe ) , and @xmath139 and @xmath140 are the total accretion and destruction rates , respectively . in order to study different scenarios for the evolution of galaxy number density , here we assume : @xmath141 as suggested by vincoletto et al . in particular , we explore two competitive scenarios : 1 . the _ pure luminosity evolution _ , ple , @xmath129 corrisponding to the case proposed in the grieco et al . ( 2012a ) ; the cdr is obtained by assuming that all galaxies started forming stars at the same time and that there is no evolution in the galaxy number density from early epoch to the present time . as a consequence , @xmath142 and @xmath143 for each galaxy type . 2 . the_density evolution_@xmath129 scenario , de , where the @xmath144 parameter evolves with redshift , as shown in table [ cdrparam ] , where the value of @xmath145 , the same in both cases , and the values of @xmath146 for each galaxy type , are reported . this particular combination of values is selected in order to obtain a csfr which reproduces the trend of a hierarchical galaxy formation scenario ( e.g. menci et al . 2004 ) . in this picture , the majority of ellipticals formed in a large redshift range and by mergers of spirals , with only a small percentage of them forming by means of a strong starburst at high redshift . clearly the csfrs for the two scenarios are extreme cases , which predict completely different galaxy formation patterns ( see fig . 6 of grieco et al . 2012a , where both cases are shown ) . because of this , we checked that the integrals of the csfrs in the two cases give a similar stellar density at the present time . @xmath147 & @xmath148 & @xmath149 + + + @xmath150 & @xmath151 & @xmath152 + in figure [ zcdustr ] , we show the predicted cosmic effective dust production rate as a function of redshift ( black solid line ) , which includes production , accretion and destruction for each galaxy type . this cosmic effective dust production rate , expressed in @xmath153 , here has been computed by means of eq . [ cdr ] , under the assumption of no number density evolution of galaxies . we remind that with no number density evolution we predict a very high star formation rate at high redshift due to the starbursts occurring during the early phases of the proto - spheroid formation ( see grieco et al . in the same plot are also shown , separately , the cosmic stellar dust production , dust accretion and dust destruction rates as functions of time . fixing the redshift of galaxy formation at @xmath154 , the predicted total cosmic effective dust production rate ( black curve ) shows a a peak at very high redshift followed by a decrease and then another higher peak at @xmath155 . this behaviour is due to the predominance of proto - spheroids at high redshift where the early high sn production induces first an increase in the dust rate but then a decrease due to the high destruction rate related to the sne themselves . the peak at @xmath1 is then due to the predominance of the star formation rate in the spirals , because since then , in this scenario , the ellipticals evolve passively . the cosmic effective dust production rate then decreases strongly for @xmath156 down to redshift @xmath157 and stays constant afterwards until @xmath158 . it is interesting to note that accretion and destruction rates almost compensate each others during the whole cosmic time . in figure [ zcdr_de ] we show instead the cosmic dust rates as functions of redshift in the case of strong number density evolution . it is interesting to note that the cosmic effective dust production rate increases slightly from high redshift down to @xmath159 and then decreases down to @xmath158 . in general , the evolution of the effective dust production rate is smoother than in the ple case . clearly , the two different galaxy evolution scenarios produce different results concerning the cosmic dust evolution . finally , integrating in time the cosmic effective dust production rate , shown in figures 8 and 9 , we can obtain the cumulative cosmic dust mass density in the universe as a function of redshift @xmath160 . in a @xmath131cdm cosmological frame with the critical density being : @xmath161 , for a hubble constant of @xmath162 we can compute @xmath163 and its evolution as a function of redshift , as shown in figure 10 for the two scenarios . the present time values of @xmath163 from figure 10 are : @xmath164 for the ple scenario , and @xmath165 for the de scenario . fukugita ( 2011 ) estimated a present time cosmic dust density produced by stars in the universe of @xmath166 , whereas the value estimated from observations is @xmath167 . other estimates ( loeb & haiman 1997 ; corasaniti 2007 ; inoue & kamaya 2004 ; fukugita 2011 ; menard & fukugita 2012 ) all suggest a value between @xmath168 and @xmath169 . our values are higher than these estimates and the reason could be that our stellar dust production is probably overestimated , in particular the assumed yield of dust from sneia could be overestimated , since there is no clear observational evidence for dust production in these sne . however , we checked the effect of suppressing type ia sne as dust producers and the effect on @xmath163 is negligible . on the other hand , it is reasonable that the observational estimate of the cosmic dust density is a lower limit . finally , the fact that @xmath163 is larger by a factor of two in the de scenario is probably due to the fact that spiral galaxies dominate , at variance with the ple scenario , where the number of spheroids is comparable to spirals . in the ple scenario , the elliptical galaxies evolve very quickly and their star formation is quenched by galactic winds , whereas in the de scenario ellipticals form by mergers of spirals where star formation continues until the present time and the dust production evolves more smoothly . therefore , in the de case more dust is expected . moreover , a slightly higher cosmic density of stars is predicted for the de scenario . we find : @xmath170 for the ple case and @xmath171 for the de case . in fukugita & peebles ( 2004 ) this quantity is estimated to be @xmath172 . . ] as a function of redshift in the ple and de scenarios . continuous black line refers to the ple scenario and the dashed red line the de scenario . ] in this paper , we present a method , based on chemical evolution models , to identify the hosts of grbs . our idea comes from the following considerations : if grbs are produced by the collapse of massive stars , they are presumably originating in galaxies where the bulk of massive star formation is taking place . from an observational point of view , almost in all the cases where the galaxy hosting the grbs is observed , the burst position suggests that grbs originate in regions where star formation is taking place ( bloom et al 2002 ) . in this context , the determination of the nature of grb host galaxies and measurements of their star formation rate not only provide evidence in favour of or against the collapsar model , but also provide insight on the nature and origin of star formation in the early universe . we have presented a comparison between predicted and observed abundances and abundance ratios in some galaxies hosting grbs ( grb081008 , grb120327a and grb120815 ) . to do that we have adopted detailed chemical evolutionary codes for galaxies of different morphological type including dust evolution . the second aim of this study was to understand the effect of dust on the observed abundance patterns and to make predictions on the cosmic dust content as a function of the cosmic epoch . our main conclusions can be summarized as follows : * we have studied the evolution of the abundances of c , n , o , mg , si , s , ni and zn in galaxies of different galaxies ( ellipticals , spirals , irregulars ) . our results suggest that one of the galaxies hosting grb120815 is probably a massive proto - spheroid catched during the active star formation phase , whereas for the other two galaxies hosting grb120327a and grb081008 , no firm conclusion can be drawn , although some of the data seem to suggest spiral or proto - spheroid systems with high mass and star formation rate . in order to derive better conclusions we would need more precise stellar and dust yields . in particular , for some chemical elements the stellar yields are still quite uncertain ( see romano et al . the result of grb120815 is particularly important since it is the first time that abundances measured in hosts of grbs indicate early type galaxies . our result , if confirmed by more precise data , suggest that grbs can be hosted in any type of galaxy suffering active star formation . obviously , local spheroids which do not show any active star formation can not host grbs . we have also estimated the ages of the grb hosts from chemical evolution and found that they are all young with ages ranging from 15 to 320 myr . * we have shown that it is very important to include dust in the chemical evolution of galaxies although many uncertainties are still present in the recipes for the formation and destruction of dust . here we have considered core - collapse sne , agb stars and type ia sne . we stress that the dust is produced very early in the evolution of galaxies since ccsne have lifetimes from few million to several tenths of million years and the massive agb stars ( 7 - 8 @xmath61 ) have lifetimes of the order of 25 - 40 myr . valiante et al . ( 2009 ) also showed the importance of taking into account agb stars to explain the dust in high redshift qsos . another possible source of dust is the condensation of dust grains in quasar winds ( elvis et al . 2002 ) , which has also been supported by observations ( markwick - kemper et al . 2007 ) . we did not include this source here since pipino et al . ( 2011 ) have shown that it contributes negligibly to the total dust production in a massive elliptical , which is the typical qso host . * we have computed the cosmic effective dust production rate , namely the rate of net dust production in an unitary volume of the universe , by adopting the results of our models and assuming either no number density evolution or strong number density evolution of galaxies , as suggested by the hierarchical clustering scenario for galaxy formation . they represent two extreme cases . the cosmic star formation rate in these two scenarios ( see grieco et al . 2012a ) is different but the present time cosmic stellar density , which is the integral of the csfr , is roughly the same . to compute the cosmic effective dust production rate , we took into account the rates of production from stars , destruction and accretion of dust in galaxies of different morphological type . we assumed as a redshift of galaxy formation @xmath154 . our results show that in the pure luminosity evolution framework there are two peaks , one very close to the galaxy formation redshift and another stronger one at @xmath155 . then the dust rate decreases down to z=0 . on the other hand , in the number density evolution framework the dust rate increases slightly from the beginning to @xmath159 and then decreases down to z=0 . * finally , we have estimated the cosmic dust density in the universe at the present time , @xmath173 in both scenarios and the results are : @xmath174 in the case of pure luminosity evolution and @xmath175 in the case of number density evolution of galaxies . previous estimates ( loeb & haiman 1997 ; corasaniti 2007 ; inoue & kamaya 2004 ; fukugita , 2011 ; menard & fukugita , 2012 ) had suggested @xmath176 . our values are larger than previous estimates and the reasons for this can be found in an overestimate of the dust production in our galaxy models due either to too large dust yields from stars ( sneii , ia and agb ) or to the poor knowledge of the other processes invoving dust such as destruction and accretion . on the other hand , we can not exclude that the amount of observed dust is underestimated . therefore , these particular results should be taken with caution . to conclude , it is worth noticing that in our analysis , two of the three grb host galaxies taken into account are well fitted by a massive proto - spheroid model . this result seems to be in contrast with the claim that most of the hgs are small , star - forming and metal - poor galaxies . probably , the apparent discrepancy is due to the galaxies sample used to infer the hgs properties composed mainly of low redshift objects ( z @xmath177 2 ) . the explanation for this difference lies in the evolution / change of the star forming galaxies as a function of redshift : in the local universe the star formation is mainly traced by small , blue galaxies while at early epochs it is dominated by massive proto - spheroids . v.g . , f.m . and f.c . , acknowledge financial support from prin miur2010 - 2011 , project the chemical and dynamical evolution of the milky way and local group galaxies , prot . n. 2010ly5n2 t . and f.m . also acknowledge financial support from trieste university through the project fra2011 on cosmic star formation rate and cosmic chemical enrichment . v.g . and f.m . thank a. pipino and e. spitoni for many useful discussions . finally , we would like to thank an anonymous referee for his / her careful reading of the paper and very important suggestions . andrews j. e. , gallagher j. s. , clayton g. c. et al . , 2010 , apj , 715 , 541 asplund , m. , grevesse , n. , sauval , a. j. & scott , p. , 2009 , ara&a , 47 , 481 berger e. , penprase b.e . , cenko s.b . , kulkarni s.r . , , steidel c.c . & reddy n.a . , 2006 , apj , 642 , 979 bianchi s. & schneider r. , 2007 , mnras , 378 , 973 blair w. p. , ghavamian p. , long k. s. et al . , 2007 , apj , 662 , 998 bloom , j. s. , kulkarni , s. r. , price , p. a. , et al . 2002 , apjl , 572 , l45 boissier , s. , salvaterra , r. , le floch , e. , et al . , 2013 , a&a , 557 , a34 bradamante f. , matteucci f. & dercole a. , 1998 , a&a , 337 , 338 calura f. & matteucci f. , 2006 , apj , 652 , 889 calura f. , pipino a. & matteucci f. 2008 , a&a , 479 , 669 calura f. , dessauges - zavadski m. , prochaska j. x. & matteucci f. , 2009 , apj , 693 , 1236 chiappini c. , hirschi r. , meynet g. et al . , 2006 , a&a , 449 , l27 chomiuk , l. , povich , m.s . , 2011 , aj , 142 , 197 castro cern , j. m. , michaowski , m. j. , hjorth , j. , et al . , 2006 , apjl , 653 , l85 castro cern , j. m. , michaowski , m. j. , hjorth , j. , et al . , 2010 , apj , 721 , 1919 christensen l. , hjorth j. & gorosabel j. , 2004 , a&a , 425 , 913 corasaniti p. s. , 2007 , newar , 51 , 332 danziger i. j. , gouiffes c. , bouchet p. & lucy l. b. , 1989 , iauc , 4746 , 1 de cia a. , ledoux c. , savaglio s. , schady p. & vreeswijk p. m. , 2013 , a&a , 560 , a88 delia v. , campana s. , covino s. et al . , 2011 , mnras , 418 , 680 delia v. , fynbo j. p. u. , goldoni p. et al . , 2014 , arxiv:1402.4026 draine b. t. , 1990 , in asp conf . 12 , the evolution of the interstellar medium , ed . l. blitz ( san francisco : asp ) , 193 dunne l. , eales s. , ivison r. , morgan h. & edmunds m. , 2003 , nature , 424 , 285 dwek e. , 1998 , apj , 501 , 643 dwek e. , galliano f. & jones a. p. , 2007 , nuovo cimento b serie , 122 , 959 dwek e. & cherchneff i. , 2011 , apj , 727 , 63 fan x. l. , yin j. & matteucci f. , 2010 , a&a , 521 , a73 flagey n. , boulanger f. , verstraete l. et al . , 2006 , a&a , 453 , 969 franois p. , matteucci f. , cayrel r. et al . 2004 , a&a , 421 , 613 fryer c. l. & heger a. , 2005 , apj , 623 , 302 fruchter a. s. et al . , 2006 , nature , 441 , 463 fruchter , a. s. , thorsett , s. e. , metzger , m. r. , et al . 1999 , apjl , 519 , l13 fukugita m. , 2011 , arxiv:1103.4191 fukugita , m. , & peebles , p. j. e. , 2004 , apj , 616 , 643 harris , j . , zaritsky , d. , 2009 , aj , 138 , 1243 hatsukade b. , hashimoto t. , ohta k. et al . , 2012 , apj , 748 , 108 hirashita h. & kuo t .- , 2011 , mnras , 416 , 1340 hunt l. , palazzi e. , rossi a. et al . , 2011 , apj , 736 , l36 kimura h. , mann i. & jessberger e. k. , 2003 , apj , 582 , 846 kozasa t. , hasegawa , h. & nomoto k. , 1989 , apj , 344 , 325 kozasa t. , nozawa t. , tominaga n. , et al . , 2009 , cosmic dust - near and far , 414 , 43 kotak r. , meikle w. p. s. , farrah d. et al . , 2009 , apj , 704 , 306 krause o. , birkmann s. m. , rieke g. h. et al . , 2004 , nature , 432 , 596 krhler t. , greiner j. , schady p. et al . , 2011 , a&a , 534 , a108 krhler t. , ledoux c. , fynbo j. p. u. et al . , 2013 , a&a , 557 , a18 le floch e. , charmandaris v. , forrest b. et al . , 2006 , gamma - ray bursts in the swift era , 836 , 528 levesque e. m. , kewley l. j. , berger e. & zahid h. j. , 2010 , aj , 140 , 1557 loeb a. & haiman z. , 1997 , apj , 490 , 571 lucy l. b. , danziger i. j. , gouiffes c. & bouchet p. , 1989 , iau colloq . 120 : structure and dynamics of the interstellar medium , 350 , 164 matsuura m. , dwek e. , meixner m. et al . , 2011 , science , 333 , 1258 matteucci f. , greggio l. , 1986 , a&a , 154 , 279 matteucci f. , 1994 , a&a , 288 , 57 matteucci f. , 2001 , the chemical evolution of the galaxy , assl , kluwer academic publisher matteucci f. , 2003 , arxiv : astro - ph/0203340 matteucci f. , 2012 , chemical evolution of galaxies : , astronomy and astrophysics library . isbn 978 - 3 - 642 - 22490 - 4 . springer - verlag berlin heidelberg , 2012 mattila s. , meikle w. p. s. , lundqvist p. et al . , 2008 , mnras , 389 , 141 markwick - kemper f. , gallagher s. c. , hines d. c. & bouwman j. , 2007 , apjl , 668 , l107 marzke r.o . , geller m.j . , huchra j.p . & corwin h.g . , jr . , 1994 , aj , 108 , 437 mckee c. f. , 1989 , in allamandola l. j. , tielens a. g. g. m. , eds , interstellar dust , proc . iau symposium 135 . kluwer , dordrecht , p. 431 meikle w. p. s. , mattila s. , pastorello a. et al . , 2007 , apj , 665 , 608 mnard b. & fukugita m. , 2012 , apj , 754 , 116 menci n. , cavaliere a. , fontana a. et al . , 2004 , apj , , 604 , 12 michaowski m. j. , murphy e. j. , hjorth j. et al . , 2010 , a&a , 522 , a15 modjaz m. , kewley l. , kirshner r. p. et al . , 2008 , aj , 135 , 1136 morgan h. l. & edmunds m. g. , 2003 , mnras , 343 , 427 nomoto k. , tominaga n. , umeda h. , kobayashi c. & maeda k. , 2006 , nuclear physics a , 777 , 424 nozawa t. , kozasa t. , umeda h. , maeda k. & nomoto k. , 2003 , apj , 598 , 785 nozawa t. , kozasa t. , tominaga n. et al . , 2008 , apj , 684 , 1343 pagani l. , steinacker j. , bacmann a. , stutz a. & henning t. , 2010 , science , 329 , 1622 perley d. a. , cenko s. b. , bloom j. s. et al . , 2009 , aj , 138 , 1690 perley d. a. , levan a. j. , tanvir n. r. et al . , 2013 , apj , 778 , 128 pipino a. , fan x. l. , matteucci f. et al . 2011 , a&a , 525 , a61 sakon i. , onaka t. , wada t. et al . , 2009 , apj , 692 , 546 savaglio s. , glazebrook k. & le borgne d. , 2009 , apj , 691 , 182 savaglio s. , rau a. , greiner j. et al . , 2012 , mnras , 420 , 627 schmidt m. , 1963 , apj , 137 , 758 schurer a. , calura f. , silva l. et al . , 2009 , mnras , 394 , 2001 silvia d. w. , smith b. d. & shull j. m. , 2010 , apj , 715 , 1575 stanek k. z. , gnedin o. y. , beacom j. f. et al . , 2006 , aca , 56 , 333 szalai t. , vink j. , balog z. et al . , 2011 , a&a , 527 , a61 valiante r. , schneider r. , bianchi s. & andersen a. c. , 2009 , mnras , 397 , 1661 van den hoek l.b . & groenewegen m.a.t . , 1997 , a&as , 123 , 305v vincoletto l. , matteucci f. , calura f. , silva l. & granato g. , 2012 , mnras , 421 , 3116 wolf c. & podsiadlowski p. , 2007 , mnras , 375 , 1049 woosley s. e. , 1993 , apj , 405:273 - 277 woosley s. e. & weaver t. a. , 1995 , apjs , 101 , 181 woosley s. e. e heger a. , 2006 , in gamma ray bursts in the swift era , washington , d.c . , eds . holt s. , gehrels n. and nousek j. a. , aip 836 , 398