article
stringlengths 0
745k
| abstract
stringlengths 0
92.7k
|
|---|---|
the evolution of galaxies is a complex process involving angular momentum redistribution , star formation , chemical evolution , feedback and numerous merging events . understanding how and why galaxies have the properties we observe is further complicated by the fact that observations are limited in a number of complicated ways , including dust obscuration and our inability to view a galaxy in all its projections .
one of the key observations of disk galaxy properties is the ubiquitous exponential surface brightness profile ( de vaucouleurs 1959 ; freeman 1970 ) .
freeman , and many others since , modeled galaxies as two component systems with : an inner spheroid where @xmath5 and an outer exponential disk component with @xmath6 .
it is now believed that the inner galaxy components are best modeled by a srsic function , @xmath7 , with @xmath8 ( macarthur et al .
2003 ; pohlen & trujillo 2006 , hereafter pt06 ) .
exponential surface brightness profiles have largely been explained as a relic of the initial angular momentum distribution of the gas which collapsed to form the observed stellar disk population ( fall & efstathiou 1980 ; ferguson & clarke 2001 ) .
while the exponential profile is certainly common among disk galaxies , many galaxies deviate significantly from it .
a large percentage of galaxies show a two - component exponential profile with a steep outer region ( pt06 ) . the transition between the two regimes as sometimes been referred to as the truncation or break radius . in order to avoid confusion in notation
we will refer to these profiles as two - component or broken exponentials .
the issue of broken profiles was first addressed by van der kruit ( 1979 ) and while improved observations have led to subsequent refinements in the analysis of these profiles , confusion still remains as to the nature of these profile excursions .
cosmological simulations coupled with semi - analytical models have been able to reproduce many of the observed features of galaxies .
recent high resolution simulations by governato et al .
( 2007 ) with gas cooling and feedback have shown a rotationally supported disk with an exponential density profile can be generated .
however , current models still exhibit striking failures .
for instance , the well - known angular momentum catastrophe may be caused by dynamical friction of infalling gas leading to the loss of angular momentum ( navarro & benz 1991 ; navarro & white 1994 ; navarro & steinmetz 1999 ) . smooth mass accretion , additional resolution and angular momentum conservation through feedback processes also yield disks that are too centrally concentrated . producing
a realistic disk galaxy still remains a challenging task for all galaxy modelers ( lake & carlberg 1988 ; katz 1992 ; navarro & white 1994 ; navarro & steinmetz 1997 ; thacker & couchman 2000 ; sommer - larsen et al . 2003 ; governato et al .
2004 ; kaufmann et al . 2007 ;
governato et al .
2007 ) . to date , no single galaxy model has ever matched simultaneously the size - velocity - luminosity and color luminosity of galaxies ( courteau et al .
2007 ; dutton et al . 2007 ) .
cosmological simulations of galaxies have yet to produce pure exponential disk galaxies as seen in nature ( ngc 300 ( bland - hawthorne et al .
2005 ) .
due to measurement limitations and the fact that the observed outer disks blend into the sky noise especially for face - on orientations , it is difficult to address the origin of double exponential profiles using observations alone .
simulations allow us to examine the evolution of the galaxy in time free of any observational constraints .
however , careful comparisons with observations are still required . while semi - analytical studies and n - body simulations have been performed in this context ( dalcanton , summers & spergel 1997 , hereafter dss97 ; mo , mao & white 1998 , hereafter mmw98 ; van den bosch 2001 , hereafter vdb01 ; debattista et al .
2006 ; hereafter d06 ; kauffman et al . 2007 ;
governato et al .
2007 ) , an exhaustive exploration of the full parameter space of galaxy structure models has not yet been attempted .
vdb01 conducted a semi - analytical study of the nature of galaxy density distributions , and focused specifically on the angular momentum catastrophe .
like most treatments that ignore angular momentum exchange between the bulge , disk and halo components , the disks produced by vdb01 were too centrally concentrated in both the stars and gas .
however , the stellar breaks in the exponential profiles were in close agreement with observations due to the use of a star formation threshold .
because the gas distribution was directly related to angular momentum , the subsequent development of a high central density in the disks drove the gas to the inner regions producing outer breaks interior to those observed .
furthermore the models of vdb01 did not produce the observed double exponential profile in the gas and stars , but were rather sharply truncated in the outer regions .
this work addresses the nature of physical two - component profiles in isolated galaxies free of cosmological evolution ( such as merging and accretion ) and whether these profiles can be attributed to dynamical processes in a disk or alternatively to star formation , or indeed both . as noted by d06 , a semi - analytical approach like that of vdb01 ,
can not account for the fundamental role of bars and secular evolution on the profiles .
the pioneering numerical study of d06 was however restricted to a small range of galaxy parameters .
d06 used an n - body approach to examine whether secular evolution could reproduce observed stellar density profiles .
d06 found that the density profiles may evolve considerably over time , particularly under the action of a bar .
furthermore , their study of broken exponentials relied on rigid halo models and also lacked a representation of star formation and feedback . for our study of galaxy density profiles
, we have explored a more extensive parameter space of the dark and baryonic matter in galaxies , with combinations of the structural parameters that control the various observed profile types and how those profiles evolve over time .
our study involves a sensitivity analysis of galaxy models based on five fundamental parameters , the spin parameter , @xmath1 , halo concentration , @xmath9 , mass or virial velocity , @xmath10 , disk mass fraction , @xmath0 , and the presence of a bulge with a mass fraction , @xmath11 .
our study considers strictly isolated galaxies free of gas infall or interactions .
while such models do not accurately represent nature , they enable a clear focus on the physical processes associated with secular evolution , that would otherwise be masked by the addition of merging histories and gas accretion . upon identifying the most significant parameters that control break development , one can then begin to consider a more complete environment ( roskar et al .
2007 , curir et al . 2006 ) .
this research makes use of an n - body / sph code , gadget-2 , developed and made available to us by volker springel ( see springel 2005 for a description of this code ) .
the star formation model uses the parameters of springel & hernquist ( 2003 ) and was held fixed to keep the exploration of the dynamical parameter space manageable .
the layout of this paper is as follows : in 2 we review recent works and discuss our motivations . in 3 , we describe our simulations and the choice of parameter space .
we present our findings in 4 and compare our results with observations in 5 .
a discussion and a summary of our main findings are presented in 6 .
recent theoretical models of galaxy formation have shown that the angular momentum , @xmath12 , is the principal determinant for the observable features of a galaxy ( dss97 ; vdb01 ; hernandez & cervantes - sodi 2006 ; hereafter hs06 ; dutton et al .
dss97 showed that the angular momentum distribution of a galaxy controls the shape of its rotation curve and its overall surface brightness .
low angular momentum galaxies are more centrally concentrated , have high surface brightness ( hsb ) and are globally unstable to the formation of bars and bulges .
conversely , high angular momentum galaxies have greater disk scale lengths , have lower surface brightnesses ( lsb ) and a higher dynamical mass - to - light ratio .
angular momentum is also central to determining if a system is dynamically stable to bar formation ( mmw98 ) .
the early simulations of efstathiou , lake & negroponte ( 1982 ) showed that disks are unstable to bar formation if : + @xmath13 where @xmath14 is the maximum rotation velocity of the disk , @xmath15 is the disk scale length and @xmath16 is the total disk mass . for a given disk mass ,
only a range of spin parameters would produce a stable disk . in the model of mmw98
the stability condition is recast as : + @xmath17 where @xmath18 and @xmath19 , @xmath20 and @xmath21 are functions introduced for a self - gravitating disk ( see eqs .
23 , 29 , 34 and 37 of mmw98 ) .
since the effect of disk self - gravity on @xmath15 and @xmath14 is weak , the condition for stability can be approximated by @xmath22 .
thus , a galaxy with a high disk mass fraction and low angular momentum will be dynamically unstable to bar formation .
we contrast this global dynamical stability estimator based on @xmath2 with the local toomre stability criterion in 4.3 .
the maximum specific angular momentum of the baryons in the disk may potentially cause disk breaks . if angular momentum redistribution does not occur in the disk ,
then the outermost stellar radius will reflect the maximum value of the angular momentum of the baryonic material . whether or not a galaxy develops a broken
exponential could be an important indicator of the angular momentum properties of the original halo .
thus , the outer disk could serve as an important diagnostic of properties of the proto - galaxy ( freeman & bland - hawthorne 2002 ) .
several ongoing problems have plagued our understanding of the role of angular momentum in disk formation .
the assumption of angular momentum conservation for the baryonic component is essential to avoid the so - called _ angular momentum catastrophe_. a simple expression of the disk angular momentum is given by @xmath23 ( mmw98 ) .
angular momentum conservation implies that if the disk mass increases , the disk scale length must also decrease ( assuming constant @xmath14 ) , producing small compact galaxies .
numerical simulations have shown that angular momentum loss yields disk galaxies that are an order of magnitude too small / compact ( navarro & benz 1991 ) . to prevent angular momentum loss and small disks ,
stellar feedback and ionizing uv background radiation definitely play some role ( thacker & couchman 2001 ; somerville 2002 ; governato et al .
recent work by kauffman et al .
( 2007 ) and naab et al .
( 2007 ) , which extends the work of okamoto et al .
( 2003 ) , have shown that increased numerical resolution also lessens the angular momentum catastrophe . nonetheless , current galaxy models simulated from cosmological initial conditions still fail to reproduce the observed galaxy scaling relations ( gnedin et al .
2006 ; dutton et al . 2007 ;
governato et al .
2007 ; dutton & courteau 2008 ) and more research is needed .
while the light profile of disk galaxies may generally be described by an exponential function , observed profiles are in fact much more complex due to the presence of a bulge , bar , spiral arms and transient features .
freeman ( 1970 ) already distinguished two types of disk profiles : the type i s that are closely exponential over all radii except for the possible presence of a bulge component , and type ii s that display a depression of the surface brightness between the outer bulge and the inner disk .
van der kruit ( 1979 ) and later van der kruit & searle ( 1981 , 1982 ) , found departures from a pure exponential profile at roughly 4 - 5 disk scale lengths .
this range of break radii was later systematically revised to lower values ( 2.2 - 3.6 disk scale lengths ) ( pt06 ) , thanks to a better understanding of observational issues .
specifically , these early studies used edge - on systems since the line of sight integration enhances the surface brightness of a break . however , line of sight integration distributes the light along the exponential decline meaning that the measured scale lengths in projection will systematically be smaller than in the face - on case ( pohlen et al .
2004 ) . for face - on galaxies alone ,
van der kruit ( 1988 ) and pohlen et al .
( 2002 ) found mean break radii at @xmath24 and @xmath25 disk scale lengths respectively .
there is considerable controversy surrounding what fraction of galaxies exhibit breaks , due in part to conflicting definitions for breaks and galaxy types .
for instance , unlike freeman ( 1970 ) , pt06 label type ii galaxies as having a break in the outer region ( but those two are not inconsistent ) .
erwin , beckman and pohlen ( 2005 ) also defined type iii galaxies as anti - truncated or upbending systems .
while sky subtraction errors ( underestimated sky level ) could explain away these profiles , bona fide type iii s are associated with galaxy mergers .
since our study focuses on the secular evolution of isolated galaxies , this galaxy type will not be investigated .
proper statistics of truncated systems will be provided elsewhere but suffice to say here that roughly two thirds of observed disk systems exhibit a two - component profile .
it is not surprising that this fraction should match that of barred systems since the bars and profile breaks are intimately connected ( as we shall see in 4.2 ) .
menendez - delmestre et al . (
2007 ) has shown that roughly @xmath26 of all nearby bright galaxies in the ir are barred .
broken exponentials are also ubiquitous at higher redshift .
prez ( 2004 ) found evidence for `` early truncation '' in a sample of 16 galaxies at @xmath27 with an average break radius @xmath28 3.5 disk scale lengths .
six of these distant galaxies had an average break radius @xmath291.8 disk scale lengths .
trujillo & pohlen ( 2005 ) studied 36 face - on galaxies in the hst ultra deep field galaxies of which 21 exhibited a two - component profile with break radii smaller than 1 - 3 kpc from observed local values .
these findings would suggest an increase in galaxy disk sizes by 25% since z @xmath301 ( somerville et al .
2008 ) . in summary ,
while there is considerable scatter in the frequency of galaxy types and the location of breaks , the fact that so many groups have identified broken exponentials using different samples and different methods , means that breaks are certainly _ real_. any complete theory of galaxy formation and evolution must account for the origin and range of observed surface brightness profiles .
both the early processes of disk assembly , which are dominated by the results of major and minor mergers , and the long term effects of secular evolution play central roles in defining many of the observed features of galaxy surface brightness profile .
non - axisymmetric perturbations , such as those generated by a bar , may play a significant role in shaping the observed light distributions .
as they evolve , bars drive spiral structure producing rings and driving gas to the center of galaxies triggering starbursts ( hohl 1971 ) .
bars are also thought to trigger bulge formation ( kormendy & kennicutt 2004 ) .
the action of the bar induces resonances in the disk gas ( inner lindblad resonance at the end of the bar and outer lindblad resonance at @xmath30 2.2 bar radii ) , and funnels gas to the inner regions of the galaxy ultimately yielding the formation of a `` pseudo - bulge '' ( kormendy 1993 ; courteau et al .
1996 ; kormendy & kennicutt 2004 ) .
bars have also been linked to broken exponential profiles .
anderson , bagget & bagget ( 2004 ) found from the study of surface brightness profiles for 218 spiral and lenticular galaxies that barred galaxies are four times more likely to exhibit breaks than unbarred ones .
they postulated that unbarred truncated galaxies might be produced by resonances driven by other asymmetries .
they proposed that lopsided features which were identified in 30% of the galaxies by zaritsky & rix ( 1997 ) might produce such resonances .
some authors have suggested that broken exponentials may reflect variations in disk intensity due to clumpy star formation , spiral arms or ring - like features ( narayan & jog 2003 ) .
erwin et al .
( 2008 ) found that 42% of a sample of 66 barred galaxies showed breaks .
they showed that these breaks may indeed be associated with the bar or resonances in the disk .
simulations have also shown that bars can induce a substantial amount of matter redistribution in the outer disk ( athanassoula & misiriotis 2002 ; valenzuela & klypin 2003 ; d06 ) .
d06 explored the evolution of disks under the presence of a bar and showed that breaks are produced interior to where the bar sheds its angular momentum and carried away by resonantly - coupled spiral arms . while their work highlighted the importance of secular evolution in determining the morphology of galaxies , it only covered a limited parameter space .
our study follows in their footsteps and explores secular evolution in general , not just that induced by bar formation .
there are two favored hypotheses for the physical origin of broken exponentials .
the _ collapse model _ of van der kruit ( 1987 ) links the break radius to the angular momentum of the protogalaxy .
kennicutt ( 1989 ) , on the other hand , suggested a _ threshold model _
whereby a dynamical critical threshold for star formation yields a visible break in the stellar luminosity profile .
the break radius would be located where the gas density is less than the critical value defined by the toomre criterion ( toomre 1964 ) : @xmath31 where @xmath32 is the radial velocity dispersion of the gas and @xmath33 is the epicyclic frequency of the stars and gas . the origin of a gas cloud that exceeds this critical density may be a product of swing - amplified or shear instabilities , ring instabilities or giant expanding shells in turbulence - compressed regions ( elmegreen & hunter 2006 ) .
stellar disks behave similarly to gaseous disks at long wavelengths ( rafikov 2001 ) .
the description of a two - component disk with a hot stellar fluid and a cold ism fluid is however more elaborate .
wang & silk ( 1994 ) added a contribution of the stellar component to the critical density and expressed the toomre - q parameter for the gas @xmath34 star model as : @xmath35 with @xmath36 .
thus , star formation can only occur in _ locally _ unstable regions of the disk , @xmath37 and the timescale of star star formation is related to the local dynamical time , @xmath38 , where @xmath39 is the orbital frequency .
@xmath4 also correlates with the star formation rate , with small values implying more rapid star formation ( wang & silk 1994 ) .
vdb01 considered a combination of the collapse and the threshold models in his semi - analytical approach to understand the density distribution of galaxies and the angular momentum catastrophe . in his model ,
the break in the gas density profile is directly related to the angular momentum of the protogalaxy whereas the break radius of the stars is determined by a threshold density predicted by the toomre criterion .
vdb01 inferred that stellar break radii were in good agreement with observations , due to the use of star formation threshold , but his gas breaks had too high a surface density in conflict with observations .
his models predicted a sharp cut - off of the gas close to the radii @xmath40 , where the h@xmath41i column density falls to @xmath42@xmath43 @xmath44 . in real galaxies , however , h@xmath41i gas is found far beyond @xmath40 .
like most treatments that ignore angular momentum exchange between the bulge , disk and halo components , the disks produced by vdb01 were too centrally concentrated in both the stars and gas .
because the gas distribution was directly related to angular momentum , the subsequent development of a high central density in the disks would drive the gas to the inner regions producing an outer break interior to those observed .
another possible explanation for the origin of the exponential profile lies in a mechanism that could drive the stellar profile towards an exponential regardless of the initial gas profile .
the fact that many observed gas profiles depart from an exponential profile at large radii further supports this idea ( ferguson & clarke 2001 ) .
lin & pringle ( 1987 ) first introduced the idea of viscosity to regulate the angular momentum redistribution and radial flows .
viscous models lead to matter funneling into a central concentration and the outer regions , subsequently moving outward , and forming an exponential profile .
ferguson & clarke ( 2001 ) did not acknowledge the development of two - component profiles but , as noted by pohlen ( 2002 ) , their figs . 2 and 5
do show break - like features at 4 - 5 disk scale lengths in agreement with observed values .
the timescale for development of the outer exponential profile can be long , comparable to the star formation timescale ( slyz et al .
2002 ) , especially in cases where the initial disk is highly irregular .
furthermore , irregular galaxies also exhibit exponential profiles , where there is little shear from viscous effects ( hunter & elmegreen 2006 ) .
elmegreen & hunter ( 2006 ) showed that a double exponential profile , in contrast to a sharp cut - off in a single exponential disk profile , could be produced by invoking gas turbulence in the outer regions of the disk . in the inner regions the star formation threshold is responsible for the inner disk component . in the outer regions turbulent compression , which allows for cloud formation and star formation despite sub - critical threshold values , can reproduce observed density profiles of disk galaxies with great fidelity .
their model thus explains the notion that hsb galaxies would have an increased ratio of the break radius to the inner exponential scale length , since unstable regions of the disk would extend further out , as compared to lsb galaxies . in their n - body
/ sph study d06 found that galaxy density profiles may evolve considerably over time , particularly under the action of a bar .
they surmised that previous investigations based on semi - analytic models ( vdb01 ) had failed to produce breaks since they lacked the necessary development of bars .
angular momentum redistribution was shown to produce both pure exponential profiles and profiles with realistic breaks .
purely collisionless systems matched observations poorly , but the addition of gas and star formation produced much more realistic systems with breaks and mass density excursions from the pure exponential initial disk profile similar to those observed in real galaxies .
d06 stressed that angular momentum can not be trivially related to a distribution of scale lengths , which is often assumed ( dss97 ; mmw98 ) .
rather , the evolution of the density profile after bar formation should depend primarily on the toomre - q parameter .
if q is small , a bar develops quickly .
if the disk is hotter , bar formation is suppressed and the disk scale length remains relatively unchanged .
the process of bar formation is thus directly related to the formation of breaks : the hotter disks that suppress bar formation will also preserve their pure exponential profile .
however , d06 acknowledge that breaks have been observed in unbarred galaxies and postulated that these galaxies may have been excited through some form of interaction ( although typically type iii or upbending profiles are associated with such events ) .
they found that bars could not be destroyed , thus preventing the subsequent development of unbarred truncated galaxy as well .
while their simulations achieved significant resolution , the approximation of purely collisionless dynamics in their break simulations may lack important physics such as the effects of star formation , star formation threshold and feedback ( see 3.4 ) .
they also explored a limited section of the available parameter space , completing only 12 simulations , and focusing on variations of the ratio of the disk scale height to the scale length ( @xmath450.025 - 0.2 ) , the toomre - q parameter ( @xmath4=1 - 1.6 ) , the halo type ( logarithmic or hernquist potential ) , the ratio of halo to disk scale length ( @xmath46 3.3 and 20.8 ) and the srsic index of the disk ( @xmath471.0 , 1.5 , 2.0 ) .
our study explores a more complete parameter space as well as the time evolution of the different features in the disk ( see 3 ) .
d06 s study of breaks also relied on rigid halo models and lacked star formation .
we improve upon these limitations in our models .
building on this body of previous work , we thus investigate the origin of two - component profiles in isolated galaxy disks using n - body / sph methods .
we use an exhaustive grid search to compare the relative importance of the halo parameters including the mass , concentration and angular momentum with disk and bulge parameters .
important physical processes such as star formation and feedback are included in our simulations using the n - body / sph models of sh03 for the gadget-2 code ( springel 2005 ) .
gadget-2 includes a star formation threshold which , unlike d06 , enables us to compare breaks produced by such a threshold with those resulting from matter redistribution in the disk . over half of our simulations
were evolved up to 10 gyr .
our study also addresses the time evolution of mass density profiles and their features . by comparing with observations , we can test the validity of our findings and constrain the parameters that control the observed features in galaxy density profiles .
our simulations rely on the gadget-2 code ( springel 2005 ) with star formation and feedback as described in springel & hernquist ( 2003 )
. the initial conditions of each galaxy model were generated with the initial conditions generator of springel & white ( 1999 ) .
the simulations were computed on the canadian sharcnet facilities at the university of waterloo . to reduce running times
four processors were used in parallel in each simulation and , on average , each simulation required 10 wallclock days to complete .
the softening length was chosen as 0.08 kpc for the star and gas particles and 0.18 kpc for the dark matter particles for all models .
these particular values are sufficiently small so as not to suppress bar formation ( kaufmann et al .
2007 ) .
gadget-2 uses a multiphase model for star formation which incorporates cooling and supernovae energy injection .
star formation is assumed to convert cold clouds into stars on a characteristic timescale @xmath48 ; a mass fraction , @xmath49 , of these stars are short - lived and die instantly as sne ( sh03 ) .
the star formation rate parameterization in gadget-2 is given as : @xmath50 , where @xmath51 is the density of cold star forming gas clouds .
the star formation timescale is proportional to the dynamical time of the gas : @xmath52 where @xmath53 is a model parameter and @xmath54 is the threshold density that depends on @xmath53 . in the models of sh03 , a value of @xmath55 gyr was adopted to reproduce the schmidt law . the star formation rate of an sph particle of current mass @xmath56 is given by @xmath57 , where @xmath58 is the density fraction of cold clouds . in a given time step , @xmath59 , a new star particle of mass @xmath60 is born provided that a random number between [ 0,1 ] is below @xmath61\right)\ ] ] @xmath62 is the initial mass of a gas particle and @xmath63 is an integer which expresses the potential number of star generations that each gas particle might spawn .
gadget-2 allows a gas particle to spawn multiple stars .
the total number of particles in the simulation will increase as stars are formed .
the temperature of the cold clouds , @xmath64 , and the supernovae mass fraction factor , @xmath49 , are adopted as 1000 k and 0.1 respectively . as sh03 note ,
the results are not sensitive to these values provided @xmath65 k and @xmath66 .
we do not assume a fixed gas mass density , @xmath67 , below which star formation is impeded ; rather , gadget-2 allows for a threshold density , @xmath54 , which is determined via the star formation timescale , @xmath48 , which sets the overall gas consumption timescale .
the star formation timescale , @xmath68 , is chosen by fine - tuning to match the global schmidt law of kennicutt ( 1998 ) .
a detailed overview of the resulting phase evolution between hot and cold gad can be found in sh03 .
the power index on @xmath69 was chosen to fit a value of 1.5 , in order for the star formation to be proportional to the local dynamical timescale .
this adjustment prompts the normalization coefficient to be lowered by a factor of 2 in order for the star formation rate to be unchanged at an intermediate density of @xmath70 m@xmath71 pc@xmath72 .
springel & hernquist ( 2005 ) showed that the best fitting timescale for star formation was 2.1 gyr .
this led not only to a good fit to the kennicutt slope , but also to the observed gas threshold density of @xmath73 m@xmath71 pc@xmath72 ( kennicutt 1998 ) .
springel & hernquist ( 2005 ) could not explain why this choice of @xmath68 reproduces both the kennicutt law and the gas threshold simultaneously .
the significance of this , perhaps fortuitous , coincidence remains unclear .
we also note our choice of a 30% gas fraction to allow for sufficient star formation in the disk over a 10 gyr evolution time
. the use of a star formation threshold will only affect the newly formed stars and not the initial stars in the model .
thus , when drawing comparisons with observations on the nature of a star formation threshold , we focus strictly on newly formed stars . because density profile breaks typically occur at several scale lengths , a sufficiently high number of particles is required to properly sample the outer disk .
however , the particle number should also allow the numerous simulations to be completed in a reasonable amount of time and yet not compromise any scientific value .
time constraints from our resource allocation lead us to choose a minimum of 1.4 million initial particles in the no - bulge case and 1.6 million initial particles in the presence of a bulge . to assess the impact of resolution on our simulations we also conducted lower resolution simulations , one simulation using 350 000 particles and another using 90 000 particles . fig .
[ res ] shows the impact of resolution on the simulated radial profiles for these three different resolutions , on a galaxy model with parameters @xmath74 , @xmath75 km s@xmath76 , @xmath77 , @xmath78 and @xmath79 .
profile features are essentially eliminated in the low resolution case of 90 000 particles ( red ) . in the intermediate range of 350 000 particles ( green ) ,
the break feature is evident , but the outer disk is poorly sampled . the highest achieved resolution can reproduce the break feature and provides adequate sampling in the outer region .
naturally , we have used as many particles as possible , as prescribed by the number of cpu hours available for the project .
d06 used collisionless particle models to investigate the development of a two - component profile .
our study extends this work , not only by considering a broader parameter space but also by including gas particles and star formation . in order to ascertain the effect , if any , of dissipational processes on the radial profiles , we ran several pure collisionless models and models with gas and star formation .
[ col ] shows a comparison of two such models .
the collisionless runs are shown in red , while those with gas and star formation are shown in blue for models after 5 and 10 gyr .
the radial profiles differ in several important ways : ( 1 ) the breaks in the collisionless runs tend to develop further out into the disk ; ( 2 ) more matter is distributed to the inner regions of the disk in the collisionless models ; and ( 3 ) the profiles evolve more rapidly in the collisionless case .
thus , gas and star formation does affect significantly the radial profiles and the development of a two - component profile .
our grid search involved permutations of the basic galaxy parameters listed in table 1 , resulting in a total of 162 galaxy models with gas and star formation ( see appendix a ) .
we focused on the five galaxy parameters , @xmath1 , @xmath9 , @xmath10 , @xmath0 and @xmath11 as a fundamental base to determine the rotation curve and the mass density profile of a galaxy .
for instance , it is known that the spin parameter , concentration of the halo , and its mass are fundamental to the development of bars within a galaxy ( athanassoula 2003 ) .
furthermore , the relationship between disk mass fraction and spin parameter determines whether a galaxy is bar unstable ( mmw98 ) .
finally , a significant concentration of mass in the center of a galaxy , such as classical bulge , can slow the formation of a bar or possibly impede its formation ( van den bosch 1998 ) .
all successful simulations were computed for a minimum of 5 gyr , and extended to 10 gyr , computer time and resources permitting .
the minimum evolution of 5 gyr allowed the disk to stabilize and form a substantial amount of stars . following an initial period of transient evolution
, we found that most simulated profiles had settled into a near final configuration after 3 - 4 gyr . upon completing the full permutations of model parameters
, we also ran a new suite of simulations to examine a finer grid of the parameter space .
this enabled the confirmation and refinement of various trends found with our coarser grid .
we varied @xmath1 from 0.005 to 0.1 in increments of 0.005 for those simulations ( see table 1 and appendix a for details ) .
.grid search parameters [ cols="^,^,^,^",options="header " , ]
we discuss the three methods used to determine the break radii of our simulated galaxies . considering a smooth galaxy density profile ,
the location of the break at the transition between the inner and outer disks may be found in principle at the peak of the second derivative of this profile .
however , simulated ( and observed ) profiles contain many local minima and maxima and the process of finding the peak in the second profile derivative is non - trivial .
hence , we developed a search method for the second derivative which proceeds from the inner regions of a surface density profile outwards to locate the second derivative maximum while avoiding maxima associated with raw ( unsmoothed ) and transient features .
[ [ break - radius - from - the - deviation - from - the - initial - mass - density - profile ] ] break radius from the deviation from the initial mass density profile ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ our second method involves comparing any evolved density profile to its initial profile .
this method is used only for the total and gas profiles , as the new stars have no initial profiles .
this technique can also not be applied to observationally derived density profiles since initial profiles are unknown .
our break identification involves starting from the outermost profile point and stepping inwards towards smaller radii and tagging the location where the current profile deviates steadily ( by more than 1.5 % ) from the initial profile .
a further test for the validity of these two methods is the comparison of the aforementioned break identifications with those estimated by - eye .
the inner break radius is roughly the transition from the outer bulge , if any , to the inner disk ; the outer break radius is the transition between the inner and outer disk components ( see fig . [ mark ] ) .
the slopes ( inverse scale lengths ) of the inner and outer profile components can be determined and the relative sharpness of the break is computed as the angle , @xmath80 , between the two fitted exponential lines .
[ mark ] shows a total ( stars @xmath34 gas ) mass density profile with the relevant parameters . while the two numerical methods predicted break points reasonably well
, their locations did not always closely match the estimates by - eye .
[ progcomp ] shows a comparison of the two numerical methods against eye - ball markings for the 10 gyr sample .
the discrepancies are especially acute for snapshots earlier than 3 gyr where transient features persist .
the comparison is significantly better for the 10 gyr models .
while close agreement is seen with gas break estimates , we find a greater dispersion for the total profile measurements .
the numerical techniques are clearly promising and worthy of further investigation .
however , for simplicity , we will rely on eye - ball estimates which , for most simulations , are in close agreement with the numerical predictions . | we investigate the evolution of mass density profiles in secular disk galaxy models , paying special attention to the development of a two - component profile from a single initial exponential disk free of cosmological evolution ( i.e. , no accretion or interactions ) . as the source of density profile variations , we examine the parameter space of the spin parameter , halo concentration , virial mass , disk mass and bulge mass , for a total of 162 simulations in the context of a plausible model of star formation and feedback .
the isolated galaxy models are based on the method of springel & white ( 1999 ) and were evolved using the n - body / sph code gadget-2 .
the initially pure exponential disks have a minimum of 1.4 million particles and most models were evolved over a period of 10 gyr .
we find that the slope of the outer density profile is in close agreement with that of the initial profile and remains stable over time , whereas the inner density profile slope evolves considerably as a result of angular momentum redistribution .
the evolution of the galaxy mass density profile , including the development of a two - component profile with an inner and outer segment , is controlled by the ratio of the disk mass fraction , @xmath0 , to the halo spin parameter , @xmath1 .
the location of the break between the two components and speed at which it develops is directly proportional to @xmath2 ; the amplitude of the transition between the inner and outer regions is however controlled by the ratio of halo concentration to virial velocity .
the location of the divide between the inner and outer profile does not change with time .
the condition for a two - component profile is roughly @xmath3 .
while the development of a two - component density profile is coupled to bar formation , not all barred galaxies develop a two - component profile .
a galaxy model showing a clear minimum toomre @xmath4 , normally linked to a double exponential in the stellar profile , may never exhibit any two - component feature thus yielding a fully evolved pure exponential disk .
[ firstpage ] galaxies : dynamics galaxies : spirals numerical simulations |
properties of finite - dimensional spin glasses are still under debate although the problem is essentially settled for the mean - field model @xcite .
outstanding problems for finite - dimensional spin glasses include the existence or absence of spin glass phase and whether or not the mean - field picture of the spin glass phase applies . another interesting , but less extensively studied , issue is the structure of the phase diagram , in particular where precisely the multicritical point is located and what the values are for the critical exponents characterizing the system behaviour at and away the multicritical point .
the present paper discusses this problem of the location of the multicritical point for finite - dimensional spin glass models by analytical methods .
a number of numerical investigations on this problem exist for various lattices .
however , it has been quite difficult to derive analytical results for regular finite - dimensional lattices until a few years ago when we succeeded in devising a method to predict the exact locations of the multicritical points for the square lattice ising and potts models and four - dimensional random plaquette gauge model using duality , gauge symmetry and the replica method @xcite . in the present paper
we generalize this theoretical framework so that it is applicable to a pair of mutually - dual lattices , for which our theory relates the pair of values of multicritical points of the two lattices .
the logic of our theory includes a step which is yet to be justified rigorously , and hence the status of our result is a conjecture at this moment .
nevertheless , our theory enables us to understand all the relevant available numerical data for the multicritical points derived independently by a number of authors .
also our theory satisfies necessary conditions which the exact solution should satisfy .
this paper is organized as follows . in the next section ,
we recall the basic formulation which was developed in our previous studies in order to fix the notation and set the stage for further developments in the following sections . in section 3 , a generalization of the theory to non - self - dual @xmath2 ( ising ) models with randomness
is discussed .
this argument is followed by section 4 , in which we further generalize the arguments to @xmath1 systems with chiral randomness .
the final section is devoted to conclusion and discussions .
in this section we briefly review the duality arguments in @xcite applied to the two - dimensional ( @xmath3 ) @xmath0 random bond ising model on the square lattice and the @xmath4 random @xmath2 lattice gauge model , by which we fix the notation and set a stage for generalizations in the following sections . after recalling the duality arguments for non - random systems
, we apply the idea to random systems .
we first elucidate the duality of a non - random @xmath2 ( ising ) system @xcite .
let us prepare a @xmath5-dimensional lattice and assign @xmath6 spins on @xmath7 dimensional elements @xmath8 on the lattice , which we denote by @xmath9 .
we consider a model on the lattice whose hamiltonian is given by @xmath10 where @xmath11 is the @xmath12-dimensional element on the lattice and @xmath13 is its boundary of dimension @xmath7 @xcite .
let @xmath14 denote the boltzmann factor for an element @xmath11 , @xmath15 where @xmath16 .
for the case of the usual ising model @xmath17 , @xmath18 is the bond ( edge ) boltzmann factor for parallel spins at both ends and @xmath19 is for anti - parallel spins .
then the partition function @xmath20 is a function of @xmath14 ; @xmath21 .
the dual model is defined on the dual lattice ( the definition of which is given in @xcite ) .
the dual boltzmann factor for the dual element @xmath22 is defined by the two - component fourier transformation of @xmath14 @xcite , @xmath23 for the present hamiltonian ( [ generalspin ] ) , the dual hamiltonian is given by @xmath24 which is of the same form as equation ( [ generalspin ] ) .
@xmath25 is the dual element of @xmath11 and has dimension @xmath26 .
next we derive the dual expression of the partition function for general dual pairs . the partition function
as a function of @xmath27 has the following property , @xmath28 here @xmath29 is the partition function of the original model and @xmath30 is for the dual model .
the symbol @xmath31 is a constant determined by the numbers of the elements on the lattice .
( see appendix . )
we give several examples of dual pairs below , which would be helpful to understand the definition of the dual transformation .
the arrows indicate duality relations . * @xmath32 2@xmath5 ising model on the square lattice @xmath33 2@xmath5 ising model on the square lattice ( self - dual ) * @xmath32 2@xmath5 ising model on the triangular lattice @xmath33 2@xmath5 ising model on the hexagonal lattice * @xmath34 4@xmath5 @xmath2 lattice gauge model on the hypercubic lattice @xmath33 4@xmath5 @xmath2 lattice gauge model on the hypercubic lattice ( self - dual ) * @xmath35 3@xmath5 ising model on the cubic lattice @xmath33
3@xmath5 @xmath2 lattice gauge model on the cubic lattice so far the discussions have not been restricted to self - dual models . for self - dual cases ,
@xmath29 and @xmath30 are the same function and the prefactor @xmath36 on the right hand side in ( [ o_d ] ) becomes a trivial constant which is negligible in the thermodynamic limit ( and is omitted in the following ) .
hence ( [ o_d ] ) is simplified to @xmath37 where the symbol @xmath38 is simplified to @xmath39 . from this expression it is seen that @xmath39 is invariant under the exchange @xmath40 and @xmath41 , which means self - duality of the partition function .
the critical point of a self - dual model is obtained , if it is unique , by the fixed - point condition for these boltzmann factors , @xmath42 , which yields @xmath43 .
it is clear that this transition point is shared by the 2@xmath5 ising model on the square lattice , the 4@xmath5 @xmath2 gauge model on the hypercubic lattice and their higher dimensional generalizations @xcite .
let us introduce randomness . to investigate critical points of random systems with the aid of the duality formalism
, we will utilize the technique of reference @xcite with some modifications .
the random model treated here is a system with @xmath2 variables and bimodal randomness .
the hamiltonian is written as @xmath44 where @xmath45 is a quenched random variable dependent on each element @xmath11 .
@xmath45 takes the value @xmath46 with probability @xmath47 and @xmath48 with @xmath49 . to treat random systems , we employ the standard replica method .
let us consider the @xmath50-replicated system and define the averaged boltzmann factor @xmath51 for an element @xmath11 , which corresponds to the configuration @xmath52 in @xmath53 replicas and @xmath48 in @xmath54 replicas .
the explicit form of @xmath51 is @xmath55 the @xmath50-replicated partition function is , after average over randomness , a function of these boltzmann factors , @xmath56_{\rm{av } } \equiv { \cal z}_n \ { x_{0}(p , k ) , x_{1}(p , k),\ldots , x_{n}(p , k ) \},\ ] ] where @xmath57_{\rm{av}}$ ] means random average .
we also define the dual boltzmann factor @xmath58 on the dual lattice .
the explicit forms are obtained by the two - component fourier transformation with the result @xmath59 where @xmath54 is a non - negative integer in the range @xmath60 .
the partition function satisfies a generalization of ( [ o_d ] ) , @xmath61 where @xmath62 is an appropriate constant .
now we restrict our attention to the case where the system is self - dual when we remove randomness ( e.g. 2@xmath5 @xmath0 random bond ising model on the square lattice ) . using ( [ 5a ] )
we can express the self duality of the @xmath50-replicated partition function for such a system as @xmath63 where an overall constant is neglected .
thus self duality is recognized by the fact that @xmath64 is invariant when the exchanges @xmath65 for all @xmath54 are performed simultaneously .
it is in general impossible to identify the transition point from the fixed - point condition of the duality relation ( [ selfdual ] ) , unlike the non - random case because the fixed - point conditions of all the variables @xmath66 are not satisfied simultaneously .
the authors of @xcite nevertheless developed an argument leading to a conjecture that the fixed - point condition of the leading boltzmann factor @xmath67 can well be the most plausible candidate to give the exact transition point of the random system at least on the nishimori line ( nl ) @xcite , @xmath68 , where enhanced symmetry simplifies the system properties significantly .
this prediction has been confirmed to be correct in the cases of @xmath69 and @xmath70 @xcite .
it has also been shown that numerical results for the quenched limit @xmath71 @xcite agree very well with the conjectured value of @xmath72 which is the solution to the formula obtained in the @xmath71 limit of @xmath73 on the nl : @xmath74 this conjecture also leads us to an interesting result that the multicritical points of the @xmath3 @xmath0 random bond ising model and the @xmath4 random @xmath2 lattice gauge model ( random plaquette gauge model ) are located at the same point on the @xmath47-@xmath75 plane @xcite .
this observation has also been confirmed numerically by a recent study @xcite .
determination of the multicritical point is also quite important from the standpoint of the quantum information theory , quantum memory in particular . to be specific , the value @xmath76 at the multicritical point of the @xmath3 random bond ising model is equivalent to the accuracy threshold of the @xmath3 toric code with perfect measurement @xcite , which is estimated to be @xmath77 from the above discussion .
in addition , the value @xmath76 at the multicritical point of the @xmath4 random gauge model also gives the accuracy threshold of the @xmath4 toric code ( or @xmath78 code with imperfect measurement ) @xcite , which is also determined as @xmath77 .
with the knowledge of the previous section in mind , we proceed to discussions on the non - self - dual cases with and without randomness . in this subsection
, we develop an argument for duality of generic non - self - dual @xmath2 models .
it is clear that a simple duality relation like ( [ o_d ] ) is not enough to determine the transition point .
nevertheless we show that , by generalizing the arguments in the previous section , one can still derive a relation between the transition points of a model and its dual .
first let us discuss non - random systems .
consider the product of partition functions of the original and dual models with inverse temperatures @xmath79 and @xmath80 , respectively . from ( [ o_d ] ) , one finds @xmath81 which indicates that the product is invariant under the simultaneous exchange @xmath82 and @xmath83 .
hence , if there is a unique transition point @xmath84 ( resp .
@xmath85 ) in the original ( resp . dual ) model , the relation between two critical points is expected to be given by @xmath86 which is invariant under the transformation above .
one can verify that this is equivalent to @xmath87 , which gives the correct relation between the two transition points .
now we move on to random systems . in this case
, we can express the duality between two random models ( [ 5a ] ) as follows , @xmath88 where @xmath89 and @xmath90 denote the probability of positive interaction and the inverse temperature for the original / dual models , respectively .
thus the product of the partition functions is invariant under the simultaneous exchange @xmath91 and @xmath92 for all @xmath54 .
the argument developed so far naturally suggests that the relation between the critical points of the original and dual systems is given by the fixed - point condition of the leading boltzmann factors at least on the nl .
explicitly , this condition reads @xmath93 in conjunction with the nl condition @xmath94 equation ( [ selfdual3 ] ) with ( [ nl ] ) is written in terms of @xmath95 and @xmath96 as @xmath97 we expect that the two multicritical points are related by this equation .
if the quenched ( @xmath98 ) limit is taken , this yields the relation , @xmath99 where @xmath100 therefore our main statement for the non - self - dual @xmath2 model is the following : the two critical points on the nls of the mutually - dual systems with quenched randomness are expected to be related by equation ( [ dualcriticalpoint ] ) .
there are many reasons to believe that our conjecture expressed in ( [ dual_mc_n ] ) and ( [ dualcriticalpoint ] ) is exact .
some of them are given in the rest of this paper .
for simplicity the discussion in the following is restricted to the @xmath3 random bond ising models on the mutually - dual lattices such as the hexagonal and triangular lattices , though the conjecture applies quite generally to arbitrary systems described by the hamiltonian ( [ generalrandomspin ] ) . before closing the present subsection
, we explain an explicit representation of the dual random models @xcite , which is necessary in the following discussions
. the ratios of dual boltzmann factors ( [ dualboltzmann ] ) to @xmath101 are @xmath102 these ratios of boltzmann factors are realized by a system with the following explicit boltzmann factors written in terms of ising spin variables @xmath103 where @xmath104 is the product of ising spin variables in the @xmath54th replica .
for example , in the case of the usual nearest neighbour interactions , @xmath104 stands for @xmath105 .
@xmath106 and @xmath107 are defined by @xmath108 from the expression ( [ dualrep ] ) , the dual of the @xmath0 random bond ising model can be interpreted as the model with the non - random ising interaction in each replica and the interaction between replicas .
if the condition of the nl , @xmath109 , is imposed , this turns to @xmath110 let us first show that the relation ( [ selfdual3 ] ) gives the exact answer even without the nl condition when @xmath111 .
equation ( [ selfdual3 ] ) is , for @xmath111 , @xmath112 > from the expression ( [ dualrep ] ) , it is found that the @xmath113 random bond ising model for @xmath111 , averaged over randomness , is regarded as a non - random ising model with coupling @xmath114 on the dual lattice .
for example , the random bond ising model on the 2@xmath5 triangular lattice with parameters @xmath115 and @xmath79 is equivalent to the non - random ising model on the hexagonal lattice , whose inverse temperature @xmath116 is given by @xmath117 where @xmath118 is defined by the second expression of ( [ kkp ] ) with @xmath47 replaced by @xmath119 .
conversely , the random bond ising model on the 2@xmath5 hexagonal lattice with parameters @xmath120 and @xmath80 corresponds to the non - random ising model on the triangular lattice , whose inverse temperature @xmath121 is @xmath122 as is well - known , the non - random ising models on the 2@xmath5 triangular and the hexagonal lattices are mutually - dual , and two critical inverse temperatures satisfy @xcite @xmath123 using equation ( [ dualkp ] ) we can confirm the equivalence between equations ( [ dualn1 ] ) and ( [ dualpuren1 ] ) .
note that the condition of the nl is not used , and this equivalence holds everywhere on the phase boundary .
next we discuss the two - replica case . equation ( [ selfdual3 ] ) for @xmath124 is @xmath125 in the rest of this subsection , we restrict our attention onto the nl . due to the dual representation ( [ dualrep2 ] ) for @xmath124 , we know that the random bond ising model is equivalent to the non - random four - state potts model on the dual lattice ; four states are constructed by the direct product of ising factors in two replicas @xmath126 , and the dual boltzmann factor is @xmath127 for @xmath128 and @xmath129 otherwise
. therefore the system corresponds to the four - state potts model on the dual lattice with coupling @xmath130 , namely , @xmath131 for example , the random bond ising model on the triangular lattice is equivalent to the non - random four - state potts model on the hexagonal lattice and vice versa .
here we denote the parameters of the random bond ising model on the triangular lattice by @xmath132 and that on the hexagonal lattice by @xmath133
. then the inverse temperatures of the corresponding four - state potts models @xmath134 are given by @xmath135 as mentioned above .
the dual coupling @xmath136 ( or @xmath137 ) is defined by equation ( [ dualkp ] ) .
the inverse critical temperatures of the four - state potts models on mutually - dual lattices satisfy @xcite @xmath138 it is straightforward to show that equation ( [ dualn2 ] ) is equivalent to equation ( [ dualpuren2 ] ) under the nl condition ( [ nl ] ) .
next is an argument for the @xmath140 limit .
the partition function for arbitrary @xmath50 is expressed as @xmath141_{\rm av } = [ e^{-n \beta n f(k)}]_{\rm av},\ ] ] where @xmath142 is the number of sites and @xmath143 is the free energy for the inverse temperature @xmath75 . if the replica number @xmath50 is taken to be quite large , the averaged partition function is dominated by the configuration of the smallest value of @xmath143 , which is realized by the perfect ferromagnetic bond configuration and its gauge equivalents . using this argument @xmath144
can be approximated as @xmath145 where @xmath146 is the free energy of the non - random ferromagnetic system .
therefore the random bond ising system for @xmath147 is interpreted as the non - random ferromagnetic ising system ( and its critical point is denoted by @xmath84 ) .
we also consider the random bond ising model on the dual lattice , and for @xmath140 we can regard it as the non - random ising model with the critical point @xmath148 .
these two critical points are related by the well - known equation , @xmath149 if we consider the @xmath140 limit of equation ( [ selfdual3 ] ) combined with the nl condition ( [ nl ] ) , we obtain @xmath150 which is equivalent to equation ( [ dualninfty ] ) . we have checked the relation ( [ dualcriticalpoint ] ) in the quenched limit numerically .
we executed monte carlo simulations for the @xmath3 random bond ising models on the hexagonal and the triangular lattices , a dual pair . to observe the criticality
, we make use of the non - equilibrium relaxation ( ner ) method @xcite , which yields the power - law behaviour of decreasing magnetization with monte carlo steps on criticality . for this method
we prepare all - up spins as the initial state and let them relax in each monte carlo step . in order to identify the multicritical point , we choose the parameters of the system to be on the nl and vary the parameter @xmath47 ( and @xmath75 is also varied accordingly ) .
the results are shown in figure 1 . from these results ,
the locations of the multicritical points are estimated at @xmath151 ( or @xmath152 ) for the hexagonal lattice and at @xmath153 ( or @xmath154 ) for the triangular lattice .
the two @xmath155 s sum up to @xmath156 , a consistent result with our conjecture in equation ( [ dualcriticalpoint ] ) .
spins and averaged the results over @xmath157 samples . the system at criticality is expected to yield a straight line in this log - log plot .
, title="fig:",width=593 ] [ nertrihex ] as already noted in section 2 , the random bond ising model and the random @xmath2 lattice gauge model on the @xmath78 cubic lattice are mutually dual .
thus we expect that the multicritical points of these two models will satisfy equation ( [ dualcriticalpoint ] ) .
this conjecture can be confirmed by numerical simulations of these @xmath78 systems .
in fact the multicritical points of these two models have already been estimated . for the @xmath78 random bond ising model the multicritical point
is estimated to be @xmath158 @xcite which yields @xmath159 . for the @xmath78 random gauge model
it is @xmath160 @xcite , giving @xmath161 . from these results the sum @xmath162 is about @xmath163 , a reasonable value in view of our expectation ( [ dualcriticalpoint ] ) . from the viewpoint of the quantum information theory ,
the accuracy threshold of the @xmath3 toric code with imperfect measurement can be determined by the value @xmath76 at the multicritical point of the @xmath78 random gauge model @xcite .
therefore it is an advantage of the present analysis that the duality gives an alternative way to determine the accuracy threshold when we utilize the result of the @xmath78 random bond ising model . from the result of the direct numerical analysis of the @xmath78 random gauge model @xcite and
the duality argument combined with the numerical result of the @xmath78 random bond ising model @xcite , the accuracy threshold is estimated to be about @xmath164 . in this and the next subsections we discuss the @xmath3 random bond ising model with anisotropic disorder .
this model is _ not _ a non - self - dual model , but the structure of
the problem is similar to the mutually - dual case .
we show that the duality formalism can be applied to this system and gives a conjecture for the critical points .
it should also be noted that the critical point of this model gives the accuracy threshold of the @xmath165 quantum repetition code with imperfect measurement @xcite .
first we review the duality of the anisotropic non - random ising model on the @xmath3 square lattice .
the hamiltonian is @xmath166 where @xmath167 and @xmath168 are uniform coupling constants .
the symbol @xmath169 denotes the set of horizontal bonds and @xmath170 is for vertical ones .
the partition function is @xmath171 where @xmath172 and @xmath173 .
the symbol @xmath27 is the boltzmann factor for the bond between @xmath174 and @xmath175 , @xmath176 next we define the dual boltzmann factor by the binary fourier transformation , @xmath177 using these boltzmann factors , we can express the duality of partition function as @xmath178 in this expression the boltzmann factors of the vertical and horizontal bonds are exchanged because a vertical bond is mapped to a horizontal bond on the dual lattice and vice versa .
it is obvious that the partition function is invariant under the exchange @xmath179 and @xmath180 , which is similar to the case of mutually - dual non - random systems .
the critical points are determined by the equation , @xmath181 which yields @xmath182 .
next we study the system with randomness .
the hamiltonian is @xmath183 where @xmath184 are random variables which depend on each bond and obey the probability distribution , @xmath185 for horizontal@xmath186 or vertical@xmath187 bond .
the averaged partition function is a function of averaged boltzmann factors @xmath188 and @xmath189 , @xmath190_{\rm{av } } \equiv { \cal z}_n \ { x_{0}(p_h , k_h),\ldots , x_{n}(p_h , k_h ) , x_{0}(p_v , k_v),\ldots , x_{n}(p_v , k_v ) \}.\ ] ] we also define the dual averaged boltzmann factors in the same way as in section 2 , @xmath191 using @xmath51 and @xmath192 , we can express the duality of the @xmath50-replicated partition function , @xmath193 which is invariant under the simultaneous exchange @xmath194 and @xmath195 for all @xmath54 .
the boltzmann factors of the vertical and horizontal bonds should be exchanged as in the non - random case . from the argument above
, we make a conjecture for the critical points on the nls , @xmath196 from the analogy with the non - random case ( [ dualnonrandomani ] ) or the non - self - dual random case ( [ selfdual3 ] ) . in this system ,
the nls are defined for the horizontal and vertical bonds , respectively , by @xmath197 we should note that the original system has four parameters @xmath198 and the two conditions of the nls reduce the number of independent variables to two .
consequently , the conjecture ( [ randomanicritical ] ) combined with the conditions of the nls ( [ khv ] ) does not fix the multicritical point on the phase diagram but determines the location of the `` critical line '' .
this fact is favourable for numerical verification of our conjecture because one parameter can be chosen freely even on criticality and on the nls .
we verify the validity of the conjecture in equation ( [ randomanicritical ] ) . for this purpose
the following property is useful ; the form of equation ( [ randomanicritical ] ) is completely the same as equation ( [ selfdual3 ] ) when the parameters are replaced as @xmath199 using this correspondence we can check the conjecture ( [ randomanicritical ] ) by the same argument as in mutually - dual systems with @xmath200 and @xmath140 . *
@xmath111 : the anisotropic random bond ising model is equivalent to the anisotropic non - random ising model , and equation ( [ dualpuren1 ] ) turns to the critical condition for the anisotropic non - random model when the replacement ( [ replacement ] ) is used . *
@xmath124 : the system is equivalent to the anisotropic non - random four - state potts model . using the replacement ( [ replacement ] ) , equation ( [ dualpuren2 ] ) becomes the critical condition for the anisotropic potts model .
* @xmath140 : the system is equivalent to the anisotropic non - random ising model and the critical point satisfies equation ( [ dualninfty ] ) using the correspondence ( [ replacement ] ) . if we consider the quenched limit @xmath98 , we obtain the relation , @xmath201 by the same argument as in mutually - dual systems ( [ dualcriticalpoint ] ) . here
we take @xmath202 as the two independent variables and eliminate @xmath203 using the conditions of the nls .
we expect that @xmath204 and @xmath205 will satisfy equation ( [ anicriticalpoint ] ) on criticality .
we have checked equation ( [ anicriticalpoint ] ) numerically .
the result is shown in figure 2 .
spins under all - up initial configuration .
the results are averaged over @xmath157 samples . for executing numerical calculation , we fixed @xmath205 and scanned @xmath204 in order to search the critical point .
the critical points obtained are shown with error bars .
the curve @xmath206 lies within the error bars for all @xmath205 .
, title="fig:",width=264 ] [ nerani ] the locations of the critical points expected from equation ( [ anicriticalpoint ] ) are in reasonable agreement with numerical results . using the ner with ordered and disordered initial configurations
, we also checked that the result for @xmath205 very close to @xmath46 ( or @xmath207 ) does not show discrepancy from the conjecture , which is not shown in the figure because the critical point could not be determined with sufficient precision due to rigid spin domains along one axis . thus we conclude that our conjecture is consistent with the numerical result for the anisotropic system as well .
we are allowed to interpret ( [ anicriticalpoint ] ) as the accuracy threshold of the @xmath165 quantum repetition code , given by the value of @xmath208 under the imperfect measurement rate @xmath209 .
in this section we consider a generalization of the formalism to a multi - valued spin model , the random @xmath1 model .
the @xmath1 model has @xmath210 states per site . in the non - random case
, this system in two dimensions has both aspects of the ferromagnetic ising model ( i.e. with ferromagnetic order ) and the xy model ( i.e. with the kosterlitz - thouless ( kt ) order ) for @xmath210 not too small @xcite .
the authors of reference @xcite showed that this model has three phases for @xmath211 , disordered , kt - like , and ferromagnetic , if there exists a kt phase in the continuum limit @xmath212 .
duality has been used to relate the two transition points .
the objective of the present section is to generalize the duality arguments in the previous section and apply them to the randomized version of the @xmath1 model .
as far as the authors know , this type of models have not been studied analytically in detail in the literature .
we consider the random chiral @xmath1 model , for which the partition function is written in the form , @xmath213 where @xmath214 is the interaction which satisfies cyclic condition , @xmath215 . here
@xmath216 are the site indexes , @xmath54 is the state variable which takes @xmath210 values , @xmath217 , with @xmath218 and @xmath219 is the random variable on each bond which also takes @xmath210 values , @xmath217 .
the probability that a random variable @xmath219 on a bond takes the value @xmath219 is denoted by @xmath220 ( @xmath221 ) .
a collection of them is denoted by @xmath222 in the sequel .
the summation @xmath223 is taken over state variables on all sites and the product @xmath224 is taken over all bonds .
only nearest neighbour interactions are assumed to exist on the square lattice in the present section .
our discussions are mostly for the @xmath1 version of the villain model @xcite , for which the boltzmann factor is given by @xmath225 we generalize the duality arguments to the random @xmath1 model .
let us average the @xmath50-replicated partition function .
the resulting function @xmath64 can be written in terms of the @xmath226 variables ( local boltzmann factors for neighbouring spin pairs ) , @xmath227 where @xmath228 ( @xmath229 ) .
the variable @xmath230 is the generalization of the boltzmann factor @xmath231 which appeared in section 2 .
if we set @xmath232 and @xmath233 , we obtain the relation between the present and previous boltzmann factors , @xmath234 we can define the dual model , for which the partition function can be written in terms of the dual variables .
they are defined by @xmath210-value fourier transformations @xcite @xmath235 where @xmath236 . a remark is in order .
the variables @xmath237 are already defined at the end of section 2 of reference @xcite in a different form from the ones given above .
the expressions ( 74)-(79 ) in @xcite are not in general valid and should be replaced by ( [ chi ] ) and ( [ chi_st ] ) .
now we define @xmath238 and apply the same arguments as in the previous section to identification of transition points of the @xmath1 model .
if uniqueness of the critical point is assumed , generalization of ( [ selfdual2 ] ) gives @xmath239 remember that the non - random model does not have a unique transition point for @xmath211 @xcite .
rather , two transition points are related to each other by the duality relation .
therefore let us consider the case where there exist two transition points for the random case .
if there are two critical points @xmath90 for a given @xmath240 , they may satisfy @xmath241 let us take the quenched ( @xmath71 ) limit of these relations .
when uniqueness is assumed ( @xmath242 ) , the condition reads @xmath243 if there are two critical points , this is replaced by @xmath244 if we consider the non - random case , where @xmath245 , @xmath246 and @xmath247 read @xmath248 if there are two critical points , equation ( [ 2cpp ] ) gives the relation between them , @xmath249 this agrees with the correct relation @xcite .
however , it is in general not expected that the condition ( [ 2cpp ] ) determines the whole shape of the phase boundary in the phase diagram of the random @xmath1 model .
restriction to the nl , where enhanced symmetry helps us to predict various exact results , is more likely to lead to reliable results .
consider a specific choice of @xmath220 , @xmath250 which enables us to obtain exact results using gauge symmetry under the nl condition @xcite , @xmath251 the above set @xmath252 on the nl is denoted by @xmath253 .
as in the previous section , we try to identify the relation between two transition points on the nl using the condition ( [ 2cpp ] ) .
let therefore @xmath254 denote the variables on the nl , @xmath255 our conjecture for the replicated systems is the following .
if uniqueness of the critical point is assumed , the location of it is determined by @xmath256 if there are two critical points @xmath90 on the nl , they satisfy @xmath257 we expect these relations to hold even in the quenched ( @xmath71 ) limit . in this limit
, the condition reads , corresponding to equation ( [ sdnl ] ) , @xmath258 and to equation ( [ 2cpnl ] ) @xmath259 it is interesting to note that these conditions are written in terms of the entropy function . this fact is suggestive of some underlying structure behind the scene , but we do not have a clear interpretation of this fact at the moment . it is convenient to define @xmath260 to facilitate more compact expressions for duality relation .
clearly , we have @xmath261 we also see @xmath262 so that we have @xmath263 hence equations ( [ sdnl])([2cpnlq ] ) can be rewritten in terms of @xmath264 .
it should be noticed that the function @xmath264 can be written in terms of @xmath265 , a theta function : @xmath266 where @xmath267 ( @xmath268 ) and @xmath269 .
this expression is useful for numerical evaluations of the transition points on the computer with special functions software implemented .
now we would like to discuss the plausibility of our conjectures ( [ sdnl ] ) - ( [ 2cpnlq ] ) .
the @xmath232 and the @xmath270 cases are equivalent to the ising model and the three state potts model , respectively , both of which have been discussed in detail in @xcite .
these are favourable facts to support our conjectures .
unfortunate aspect about the @xmath1 model is the difficultly of the analysis of the replicated model even for @xmath111 .
hence we do not give such evidence as we did for the @xmath2 case .
instead , the limit @xmath271 is discussed in the next subsection . when @xmath210 is large and the transition point is unique , it is reasonably expected that the transition point is of order @xmath273 . in fact , if we set @xmath274 and suppose @xmath210 is large , we find a consistent solution as follows .
the functions appearing in ( [ plk ] ) behave asymptotically as @xmath275 substitution of these expressions into ( [ sdnlq ] ) leads to @xmath276 if we approximate the summation by an integral , which should be valid for large @xmath210 , we find @xmath277 where we set @xmath278 .
we therefore have @xmath279 implying @xmath280 with @xmath281 being the base of natural logarithm .
hence , when uniqueness of the critical point is assumed , the asymptotic location is @xmath282 when there are two critical points @xmath90 , the above argument should be modified to some extent .
if we assume that both of @xmath79 and @xmath80 are of order @xmath273 , we may set @xmath283 ( @xmath284 ) and apply the same procedure as above .
the result is that @xmath79 and @xmath80 should be related through @xmath285 however , for the non - random case , it is known that the two transition points are not of order @xmath273 ; one is of order @xmath286 and the other @xmath287 @xcite .
since this is expected to persist for the random case ( see subsection [ ph_st ] below ) , equation ( [ 2cpqi ] ) may not necessarily capture the true asymptotics of the transition points . in this subsection
we discuss the phase structure of the random @xmath1 model . for the case of the non - random model
, it was shown that , when @xmath210 is sufficiently large , the simple ising - like two phase picture ( with ferromagnetic and paramagnetic phases ) is not possible @xcite .
the authors of @xcite showed that the transition point determined by the duality in the case of only two phases is inconsistent with a kind of griffith s inequality .
the discussions were based on the assumption of the existence of the kt transition in the continuous model @xmath212 .
we would like to address the issue of the full phase structure of the random @xmath1 model but let us restrict our main interest on the nl for a moment .
basically it is expected that the phase transitions on the nl are of a similar nature to the non - random case .
there are three phases ; disordered , kt and ferromagnetic .
this conclusion may be drawn by using the same arguments as for the non - random case .
there are , however , still some debates about the existence of the kt phase when randomness is introduced in the model @xcite .
hence , in this section , we give a different argument supporting the three - phase picture without assuming the existence of the kt transition in the continuous model . a crucial point in our arguments
is the proof of existence and non - existence of the ferromagnetic phase in appropriate parameter regions when @xmath288 .
first we show the existence of a ferromagnetic phase near the limit @xmath289 ( the ground state of the non - random system ) following @xcite .
let us first consider the non - random case .
the order parameter is bounded as @xmath290 where @xmath291 means thermal average , @xmath292 is the probability that the spin takes the value @xmath54 on each site , and @xmath293 is the number of sites such that @xmath294 .
all boundary spins have @xmath295 . as in @xcite , @xmath293 is bounded as @xmath296 where @xmath297 is 1 if the @xmath175th border ( separating a domain of sites from those with different @xmath54 s ) of length @xmath298 occurs in a configuration and 0 otherwise . the symbol @xmath299 stands for the number of possible borders of length @xmath298 .
the average @xmath300 is bounded by , for large @xmath210 , @xmath301 using @xmath302 where the factor @xmath303 comes from the possible number of boundary bonds , we find @xmath304 hence if we choose @xmath305 sufficiently small , the order parameter , the left - hand side of ( [ zq - orderparameter ] ) , is certainly positive . from this
follows that the ferromagnetic phase exists when @xmath306 with @xmath307 being some positive constant . following @xcite
, we can generalize the above argument to the random case .
at least for @xmath75 and @xmath308 very large , it is possible to prove the existence of a ferromagnetic phase .
let us take a border of spin configurations such that @xmath309 and notice that the thermal average @xmath310 is written as @xmath311 where the sum in the numerator is taken over states in which the @xmath175th border of length @xmath298 occurs . here
@xmath312 is the quenched randomness and occurs with probability @xmath313 ( @xmath314 ) approximately at each bond .
take @xmath315 ( @xmath50 to be distinguished from the replica number ) and consider the case where the number of such bonds with @xmath316 is @xmath317 . when @xmath318 , one restricts the summation in the denominator of ( [ xss ] ) to configurations which appear in the numerator and obtains a trivial upper bound @xmath319 .
on the other hand , when @xmath320 , one finds an upper bound @xmath321 with @xmath322 , where the summation in the exponent is over the bonds with @xmath323
. this is obtained by restricting the summation in the denominator in ( [ xss ] ) to configurations with all @xmath324 .
hence the configurational average @xmath325_{\rm av}$ ] is bounded as @xmath326_{{\rm av } } \leq \sum_{m=0}^{n-1 } \left[\begin{array}{c}2n\\m\end{array}\right ] \sum_{l_{i , j } } r^{\sum l_{i , j}^2 } \lambda^{2n - m-\sum l_{i , j}^2 } + \sum_{m = n}^{2n } \left[\begin{array}{c}2n\\m\end{array}\right ] q^m r^m.\ ] ] when @xmath327 , one has @xmath328 so that this can be replaced by @xmath326_{{\rm av } } \leq \sum_{m=0}^{n-1 } \left[\begin{array}{c}2n\\m\end{array}\right ] q^m r^m \lambda^{2n-2 m } + \sum_{m = n}^{2n } \left[\begin{array}{c}2n\\m\end{array}\right ] q^m r^m.\ ] ] then , if we choose @xmath329 , by almost the same argument as in the non - random case , @xmath330 can be made sufficiently small so that the order parameter does not vanish .
in particular , there exists a positive constant @xmath331 such that the order is nonzero when @xmath332 on the nl . the ferromagnetic phase is also expected when @xmath329 but we need a little tighter estimation to prove it . on the other hand , to show non - existence of ferromagnetic phase at high temperature , we use the arguments in @xcite .
we show below @xmath333_{\rm av}\right| \leq \langle e^{2\pi
i k_j / q } \rangle_{k_p,\rm{nonrandom } } \langle e^{2\pi i k_j / q } \rangle_{k , \rm{nonrandom}},\ ] ] where the average on the right - hand side is taken for the non - random model at inverse temperatures @xmath308 and @xmath75 , respectively . to verify this inequality
, we first notice @xmath334_{\rm av } \hspace{-3mm}&= & \left[\frac{\tr_{\{k\}}e^{2\pi ik_j / q } \prod_{\langle j_1,j_2 \rangle } e^{v_k(k_{j_1}-k_{j_2}-l_{j_1,j_2 } ) } } { \tr_{\{k\ } } \prod_{\langle j_1,j_2 \rangle } e^{v_k(k_{j_1}-k_{j_2}-l_{j_1,j_2 } ) } } \right]_{\rm av } \nonumber \\
\hspace{-2.5 cm } \hspace{-3mm}&= & \sum_{\{l\ } } \left ( \prod_{\langle j_1,j_2 \rangle } \frac{e^{v_{k_p}(l_{j_1,j_2})}}{\sum_{l_1,l_2}e^{v_{k_p}(l_{j_1,j_2})}}\right ) \frac{\tr_{\{k\}}e^{2\pi ik_j / q } \prod_{\langle j_1,j_2 \rangle } e^{v_k(k_{j_1}-k_{j_2}-l_{j_1,j_2 } ) } } { \tr_{\{k\ } } \prod_{\langle j_1,j_2 \rangle } e^{v_k(k_{j_1}-k_{j_2}-l_{j_1,j_2})}}.\end{aligned}\ ] ] applying the gauge transformation ,
@xmath335 we find @xmath336_{\rm av } = \sum_{\{l\ } } p_{\{l\ } } \langle e^{2\pi i\frac{\kappa_j}{q } } \rangle_{k_p } \langle e^{2\pi i\frac{k_j}{q } } \rangle_k,\ ] ] where @xmath337 denotes a certain probability distribution .
if we apply a very reasonable inequality that the value of the order parameter of the random system is smaller than its non - random counterpart with the same inverse temperature @xmath338 to each term in ( [ pl ] ) , we arrive at ( [ order_pure ] ) .
unfortunately we have not succeeded in proving the inequality ( [ ran_pure ] ) mathematically except for the @xmath232 case , which was already proved in @xcite . however , even in the absence of a formal mathematical proof , equation ( [ ran_pure ] ) should certainly be valid .
now , if we let @xmath79 denote the phase transition point of the non - random model at which the order vanishes , we see that the ferromagnetic phase can not exist when @xmath339 or @xmath340 .
in fact @xmath79 is known to be of order @xmath286 @xcite and we conclude that the ferromagnetic phase can not exist on the nl when @xmath341 with @xmath342 some positive constant .
notice that the above arguments imply the existence of ( at least one ) transition point(s ) between ferromagnetic and non - ferromagnetic phases .
now if there is a unique transition point , it is of order @xmath343 [ cf ( [ kqe ] ) ] . but this contradicts with the fact that the order vanishes when @xmath341 .
therefore we have shown that the simple ising - like two phase picture is not possible when @xmath210 is large enough .
this conclusion strongly suggests that there are three phases in the full phase diagram . a schematic phase diagram expected from the above arguments is given in figure 3
. model .
solid lines show expected phase boundaries.,title="fig:",width=226 ] [ phasezq ] as a remark , for @xmath210 not very large , the situation may be more subtle .
numerical simulation results suggest that something peculiar may be happening for these models ( e.g. ner analysis in @xcite ) .
further careful analyses are necessary on this point .
let us consider the maximally random case ( @xmath344 ) , in which the possibility of a finite temperature phase transition has been discussed in the literature @xcite .
here we give an argument against such a possibility using duality .
as in the case of the ising model @xcite , the @xmath50-replicated system with @xmath344 is related to the @xmath345-replicated system on the nl .
more precisely , it is easy to prove @xmath346 where @xmath347 stands for the averaged partition function ( [ zisotropy ] ) .
now let us assume the applicability of this relation down to the @xmath71 limit and discuss the phase transitions for the quenched model with @xmath344 .
note that the @xmath71 limit for @xmath344 is equivalent to taking the @xmath348 limit on the nl according to ( [ znzn-1 ] ) .
if we assume that there is a unique transition point , it is determined by the relation ( [ sdnl ] ) , which reads @xmath349 let us take the @xmath71 limit . if we set @xmath350 , we see @xmath351 so that ( [ km ] ) becomes @xmath352 this is nothing but a condition to determine the transition point for the non - random system when there is a unique transition point .
accordingly we know that @xmath353 by ( [ k1k2zq ] ) with @xmath242 and hence @xmath354 as @xmath71 .
this means that the phase transition , if it is unique , occurs at @xmath355 .
the same discussions can be applied to the case where there are two transition points . in any case
the conclusion is that the phase transitions , if there exist , occur only at @xmath355
. we should remember that our discussions were based on the subtle assumption about the applicability of the relation as @xmath71 ( @xmath348 on the nl ) and the assumption that ( [ sdnl ] ) or ( [ 2cpnl ] ) gives the criticality condition .
hence our conclusion has not been completely rigorously derived .
it would be interesting to clarify this point with other methods or arguments .
in this paper , we have generalized the duality argument combined with the replica method , which was originally applied to the @xmath3 random bond ising model on the square lattice , to a variety of random spin systems .
our main results are the conjectures on the transition points of systems with quenched randomness .
first we considered the random @xmath2 models .
we have given the conjecture ( [ dualcriticalpoint ] ) about the relation between the multicritical points of two models which are mutually dual .
besides exact computations for the replicated system with @xmath69 and @xmath70 , numerical simulations support our conjecture for the @xmath3 @xmath0 random bond ising model on the triangular / hexagonal lattices , the @xmath78 random ising / gauge models on the cubic lattice and the @xmath3 anisotropic random bond ising model on the square lattice .
we think it remarkable that a single theoretical framework makes it possible to derive a series of predictions to be compared favourably with many independent numerical simulations .
next we treated the random @xmath1 model .
we have shown that there are at least two transition points for sufficiently large @xmath210 .
the most probable scenario is that there are three phases in the phase diagram , paramagnetic , kt - like and ferromagnetic phases .
our arguments , however , do not assume the existence or some specific properties of the kt phase . by applying the duality argument to this model
, we have also given the conjectured relation ( [ 2cpnlq ] ) between the two transition points .
an interesting question is why we restrict ourselves to the nl .
the fixed - point condition of the leading boltzmann factor @xmath356 relates @xmath47 and @xmath75 and may give the whole shape of the phase boundary .
this is indeed the case for @xmath111 as well as for @xmath124 above the multicritical point .
we nevertheless have restricted ourselves to the nl in this paper because it is difficult to directly verify the ansatz on the whole part of the phase diagram numerically for many systems with quenched randomness .
it should also be kept in mind that the multicritical point is the place where two completely different types of symmetries , invariance under duality and gauge transformation , meet under the present conjecture .
this implies that the multicritical point has clearly distinguished symmetry features , which allows us to discuss this point on a different basis from other points of phase transition .
we believe that sufficient analytical and numerical evidence has been accumulated to support the validity of our conjecture .
it is an interesting future problem to provide a mathematically rigorous proof .
the authors would like to thank y ozeki and t nakamura for their suggestions and advice on numerical calculations and t hamasaki for his useful comments .
this work was supported by the grant - in - aid for scientific research on priority area `` statistical - mechanical approach to probabilistic information processing '' and the 21st century coe program at tokyo institute of technology `` nanometer - scale quantum physics '' .
ts is also supported by the grant - in - aid for young scientists ( b ) , the ministry of education , culture , sports , science and technology , japan .
in this appendix we calculate the prefactor in the duality relation of the partition function ( [ o_d ] ) in section 2 . here
we follow the derivation of the duality relation by wu and wang @xcite .
the partition function of the system described by the hamiltonian ( [ generalspin ] ) in section 2 is @xmath357 where @xmath358 @xmath359 is the boltzmann factor for an element @xmath11 with dimension @xmath12 .
for example , the usual two - body interaction on a bond has @xmath360 and the lattice gauge theory has @xmath361 .
@xmath362 is a modulo-2 spin variable which takes 0 or 1 , and @xmath363 is defined by @xmath364 modulo 2 .
let us define the dual boltzmann factor for the dual element @xmath25 by the binary fourier transformation , @xmath365 or conversely , @xmath366 where @xmath367 and @xmath368 is the number of ( @xmath7)-dimensional elements on the boundary of @xmath11 . inserting ( [ appfourier ] ) into ( [ appz ] ) and taking the sum over @xmath362 , we obtain a modulo-2 kronecker delta for each @xmath362 , @xmath369 it is useful to define the following symbol of a constrained sum , which stems from the kronecker deltas in ( [ appft2 ] ) , @xmath370 using this , ( [ appz ] ) becomes @xmath371 the prefactor can be simplified as @xmath372 where the number of @xmath317-dimensional elements is denoted by @xmath373 .
the partition function ( [ appz ] ) can also be rewritten using the boltzmann factors for the element @xmath11 , @xmath375 note that @xmath359 in ( [ appzdual1 ] ) is a function of @xmath362 as in ( 1 ) , but it is not in ( [ appzdual2 ] ) : it is redefined on each element @xmath11 as a function of @xmath363 itself . the usual sum over @xmath376 in ( [ appzdual1 ] ) can be written as in ( [ appzdual2 ] ) using the definition ( [ appsum ] ) if we consider a correct mapping from the configuration space defined by @xmath362 onto the one by @xmath363 .
the prefactor @xmath377 in ( [ appzdual2 ] ) is the degree of the ground - state degeneracy .
the mapping from the @xmath362 space to the @xmath363 space is not one - to - one . in ( [ appzdual2 ] )
the ground state configuration is @xmath378 for all @xmath11 for ferromagnetic interactions and is unique , while in ( [ appzdual1 ] ) the ground state is degenerate .
thus the mapping is @xmath377 to 1 .
for example , @xmath379 for @xmath360 because all - up and all - down states are degenerate . for @xmath361
, @xmath377 is dependent on the number of sites because the model has @xmath2 gauge symmetry in this case .
wegner calculated this degree of degeneracy in @xcite under general conditions and the result is @xmath380 where @xmath381 is a constant which depends on the topology of the lattice ( or the boundary condition ) and not on the number of elements .
if we use the generalized euler s relation for the number of lattice elements @xcite , @xmath382 where @xmath383 is also a constant dependent on the topology , @xmath384 becomes @xmath385 with @xmath386 being another constant determined by @xmath387 and @xmath12 .
we can derive the factor ( [ appng ] ) intuitively from the difference of the numbers of spin configurations between the expressions ( [ appzdual1 ] ) and ( [ appzdual2 ] ) . in ( [ appzdual1 ] )
the number of independent @xmath362 s is @xmath388 . in ( [ appzdual2 ] )
the number of independent @xmath363 s may appear to be @xmath389 , but we must take the number of constraints in ( [ appzdual2 ] ) into account , which is @xmath390 and should be subtracted from @xmath391 . however these constraints are redundant or not independent of each other because @xmath392 dimensional element is always on the boundary of @xmath393 dimensional element
. therefore we must subtract @xmath394 from the number of constraints @xmath390 .
however the number @xmath394 is also redundant again and @xmath395 must be subtracted from it , which is similar for higher dimensional elements
. then @xmath396 is calculated as @xmath397 which is equivalent to ( [ appng ] ) .
the constant @xmath386 will be the same as in ( [ appng ] ) . using ( [ appzorig ] ) , ( [ appzdual2 ] ) and ( [ appng ] ) , we obtain the duality relation between the original and the dual partition functions , @xmath398 where @xmath399 and @xmath400 is also a constant which depends on the topology .
@xmath31 is defined by @xmath401 remember that we use the modulo-2 spin variables here , while spins take the value @xmath402 in the main text .
now let us consider the self - dual case
. in this case @xmath5 must be even and @xmath403 .
furthermore the number of @xmath317 dimensional elements satisfies @xmath404 inserting this into the euler s relation ( [ appeuler ] ) , we obtain @xmath405 in addition , @xmath396 and @xmath406 become @xmath407 hence , @xmath408 therefore @xmath31 is just a trivial constant in the self - dual case and negligible in the thermodynamic limit . the factors @xmath409 and @xmath410 in ( [ appzdualrelation ] ) , which differ only by a trivial constant , do not concern when we derive the transition point from the relation @xmath42 .
mzard m , parisi g and virasoro m a 1986 _ spin glass theory and beyond _ ( world scientific ) ; young a p ( ed ) 1997 _ spin glasses and random fields _ ( world scientific ) ; nishimori h 2001 _ statistical physics of spin glasses and information processing : an introduction _ ( oxford ) nishimori h and nemoto k 2002 * 71 * 1198 maillard j -m , nemoto k and nishimori h 2003 * 36 * 9799 takeda k and nishimori h 2004 b * 686 * 377 wegner f 1971 * 12 * 2259 wu f and wang y 1976 * 17 * 439 nishimori h 1981 _ prog .
theor . phys . _
* 66 * 1169 singh r r p and adler j 1996 b * 54 * 364 ; aarao reis f d a , de queiroz s l a and dos santos r r 1999 b * 60 * 6740 ; honecker a , picco m and pujol p 2001 * 87 * 047201 ; merz f and chalker j t 2002 b * 65 * 054425 ; ito n and ozeki y 2003 _ physica _ a * 321 * 262 ; de queiroz s l a and stinchcombe r 2003 b * 68 * 144414 arakawa g , ichinose i , matsui t and takeda k _ preprint _ hep - th/0409076 , to appear in b kitaev a 2003 * 303 * 2 dennis e , kitaev a , landahl a and preskill j 2002 * 43 * 4452 ; wang c , harrington j and preskill j 2003 * 303 * 31 see e.g. itzykson c and drouffe j -m 1989 _ statistical field theory _ vol .
i ( cambridge ) ito n 1993 _ physica _ a * 192 * 604 ; ito n 1993 _ physica _ a * 196 * 591 ; ito n , matsuhisa t and kitatani h 1998 * 67 * 1188 ; ozeki y and ito n 1998 * 31 *
5451 ito n , ozeki y and kitatani h 1999 * 68 * 803 ohno t , arakawa g , ichinose i and matsui t 2004 b * 697 * 462 jos j v , kadanoff l p , kirkpatrick s and nelson d r 1977 b * 16 * 1217 elitzur s , pearson r b and shigemitsu j 1979 d * 19 * 3698 villain j 1975 * 36 * 581 ozeki y and nishimori h 1993 3399 korshunov s e 1993 b * 48 * 1124 tang l -h 1996 b * 54 * 3350 mudry c and wen x -g 1999 b * 549 * 613 griffiths r b 1964 * 136 * a437 horiguchi t and morita t 1982 * 15 * l75 horiguchi t and morita t 1981 * 14 * 2715 horiguchi t and morita t 1979 a * 74 * 340 ozeki y work in progress holme p , kim b j and minnhagen p 2003 b * 67 * 104510 | we present a conjecture on the exact location of the multicritical point in the phase diagram of spin glass models in finite dimensions . by generalizing our previous work ,
we combine duality and gauge symmetry for replicated random systems to derive formulas which make it possible to understand all the relevant available numerical results in a unified way .
the method applies to non - self - dual lattices as well as to self dual cases , in the former case of which we derive a relation for a pair of values of multicritical points for mutually dual lattices .
the examples include the @xmath0 and gaussian ising spin glasses on the square , hexagonal and triangular lattices , the potts and @xmath1 models with chiral randomness on these lattices , and the three - dimensional @xmath0 ising spin glass and the random plaquette gauge model . |
hierarchical clustering scenarios suggest that present day galaxies assembled from much smaller building blocks during the course of their evolution ( e.g. , peebles 1993 ) .
recent deep surveys have found small dwarf galaxies ( i.e. , building blocks ) at @xmath5 ( e.g. , pascarelle , windhorst , & keel 1998 ; ellis et al . 2001 ; taniguchi et al .
2003a ; santos et al .
2004 ; kneib et al .
2004 ; for a review see taniguchi et al .
at the present epoch , the assembly process continues with large galaxies capturing their satellite galaxies . in order to understand the mass assembly of galaxies ,
it is therefore important to investigate such minor merger processes in detail .
it is known that most galaxies in the local universe have satellite galaxies ( e.g. , zaritsky et al .
1997 and references therein ) .
satellite galaxies are expected to sink to the center of the potential of their host galaxies due to dynamical friction ( ostriker & tremaine 1975 ; tremaine 1981 ) ; however , satellites with long dynamical friction timescales are still found orbiting their hosts .
the satellites suffer from tidal disruption as they orbit in the galaxy s host dark matter halo .
therefore , the kinematic and morphological properties of the satellites can be used to investigate the structure and potential of the host dark matter halo .
the recent discovery of ultra compact dwarf ( ucds ) galaxies in the fornax cluster suggests that such tidal disruption processes can yield small satellite galaxies ( drinkwater et al .
2000a , 2000b ; bekki , couch , & drinkwater 2001 ; bekki et al
as dwarf galaxies orbit around the host galaxy , their outer stellar components are tidally removed .
this process is also referred to as galaxy threshing . recently , forbes et al .
( 2003 ) identified the first case of a galaxy threshing system in an early release image of the advanced camera for surveys ( acs ) on the hubble space telescope ( tran et al .
2003 ; de grijp et al .
the acs image clearly shows faint , long tidal tails emanating from a satellite galaxy around an edge - on spiral galaxy at @xmath6 .
although a pair of tidal tails is often found in major mergers ( arp 1966 ; toomre & toomre 1973 ) , it is rare to find such tails in on - going minor mergers . in this respect ,
the discovery of forbes et al .
( 2003 ) provides important information of tidal disruption of satellite galaxies around the host galaxy . following bekki et al .
( 2001 ) we use the term galaxy threshing for such on - going , tidally - disrupted satellite galaxies . here , we present the serendipitous discovery of a new potential galaxy threshing system found in the cosmos 2-square degree field ( scoville et al . 2006 ) .
the most important point is that we find clear evidence for the disruption of a small satellite galaxy .
destroyed systems are often observed , but faint tidal remnants are generally too faint to be detected in most cases .
we present the photometric properties of the system and discuss some implications .
we adopt a flat universe with @xmath7@xmath8 , @xmath7@xmath9 , and @xmath10 km s@xmath11 mpc@xmath11 throughout this paper , and use the ab system for optical magnitudes .
the cosmic evolution survey ( cosmos ) is a treasury program on the hubble space telescope ( hst ) , awarded a total of 640 hst orbits , to be carried out in two cycles ( 320 orbits in cycles 12 and 13 each ; scoville et al .
2006 ; koekemoer et al .
this is the largest amount of hst time ever , allocated to a single project .
cosmos is a 2 square degree imaging survey of an equatorial field in @xmath12(f814w ) band , using the advanced camera for surveys ( acs ) .
the cosmos hst survey alone can not be used to address all scientific questions without additional ground - based and space - based follow - up observations at multiple - wavelength .
we have been carrying out optical multi - band imaging surveys of the cosmos 2-square degree field centered at @xmath13(j2000 ) = @xmath14 and @xmath15(j2000 ) = @xmath16 ( taniguchi et al .
2006 ; see also capak et al .
2006 ) using suprime - cam , which consists of @xmath17 ccds with 2k @xmath18 4k pixels and has a pixel scale of @xmath19 pixel@xmath11 ( miyazaki et al .
2002 ) on the 8.2 m subaru telescope ( kaifu et al .
2000 ) . during our two observing runs in january and february 2004
, we obtained @xmath20 , @xmath21 , @xmath22 , @xmath23 , and @xmath24 band images of the whole cosmos field .
the individual ccd data were reduced and combined using our own data reduction software ( yagi et al .
2002 ) and iraf .
the combined images for individual bands were aligned and smoothed with gaussian kernels to match the atmospheric seeing conditions .
the fwhm of the psf of the final images has been matched to @xmath25 corresponding to the fwhm of the psf in the @xmath26 band image which had the worst seeing of all the data .
photometric calibrations are made using spectrophotometric standard stars .
details of the data reduction processes for the cosmos acs data are given in koekemoer et al .
in figure 1 , we show our five broad - band images of the galaxy threshing system .
the threshing system is a pair of interacting galaxies labeled a and b in the @xmath20 band image ( fig .
1 ) . galaxy a appears to be a giant elliptical galaxy while galaxy b appears to be a small satellite disk galaxy .
it is remarkable that very long tidal tails are emanating from galaxy b. these features are quite similar to those found in the first galaxy threshing system reported by forbes et al .
although we have not yet obtained optical spectroscopy of galaxies a and b , their physical association appears to be real , and thus this system is the second case of galaxy threshing . in figure 2
, we show an acs @xmath27 image of this system . in the high resolution hst image , there are many other satellite galaxies around galaxy a. this group has been identified as i d 10203 in sdss by merchn & zandivarez ( 2005 ) , they are all large , bright galaxies ( @xmath28 ) shown in figure 3 .
note that galaxy b is not contained in the sdss catalog , because the spectroscopic redshift of galaxy b was not obtained . in order to study structural properties of the two galaxies in the threshing system
, we examine their surface brightness distributions using the iraf `` ellipse '' program . in figure 4 , we show the surface brightness profiles of galaxy a as a function of @xmath29 .
it follows a de vaucouleur s @xmath29 law , suggesting that it is an elliptical galaxy , consistent with the sdss spectrum . in figure 5
, we show the surface brightness profile of galaxy b as a function of radius .
it shows an exponential radial profile suggesting that this galaxy is probably a disk galaxy .
next , we derive the apparent magnitudes of the two galaxies using galfit ( peng et al .
we apply a de vaucouleur s @xmath29 model for galaxy a and an exponential disk model for galaxy b. since the central region ( @xmath30 ) of galaxy a is saturated in our @xmath23 band image , we have masked out this region .
the magnitudes and their errors of both galaxies are summarized in table 1 .
the scale length of galaxy b and the effective radius of galaxy a are shown in table 2 and table 3 , respectively .
we also give results for the sersic fitting in table 4 .
the sersic index of galaxy a ranges from @xmath31 2.5 to @xmath31 3 , being smaller than 4 ( which corresponds to a de vaucouleur profile ) .
note that the central region of galaxy a is saturated in @xmath23 , and thus the sersic index may have large uncertainty . on the other hand
, galaxy b has sersic index around 1.4 in each filter .
these results seem to be basically consistent with our interpretation that galaxy a is an elliptical galaxy while galaxy b is a disk galaxy . at a redshift of 0.08 ( corresponding to a luminosity distance of 363.5 mpc ) , we estimate an absolute @xmath21 magnitude ( luminosity ) of @xmath32 for galaxy a , and of @xmath33 for galaxy b. we found a spectroscopic redshift of @xmath34 for galaxy a ( sdss j100003.2 + 020146.4 ) in the sdss spectroscopic data archive of third data release ( abazajian et al .
unfortunately , no spectroscopic information on the redshift is available for galaxy b. we derive a photometric redshift for galaxy b using the @xmath35 minimizing method ( e.g. , lanzetta , yahil , & fernndez - soto 1996 ) .
we generate sed ( spectral energy distribution ) templates for @xmath36 with a redshift bin of @xmath37 using the population synthesis model galaxev ( bruzual & charlot 2003 ) .
the seds of local galaxies are well reproduced by models whose star - formation rate declines exponentially ( @xmath38 models ) ; i.e. , @xmath39 , where @xmath40 is the age of the galaxy and @xmath38 is the time scale of star formation .
we adopt @xmath41 models with a salpeter initial mass function ( a power index of @xmath42 and a stellar mass range of @xmath43 ) and solar metallicity , @xmath44 .
we then calculate seds for ages of @xmath45 , 2 , 3 , 4 , and 8 gyr , corresponding to the sed of a starburst , irr , scd , sbc , and elliptical galaxy , respectively .
we adopt a visual extinction of @xmath46 .
note that our threshing system consists of basically nearby galaxies .
therefore , our results should not be affected by metallicity and @xmath47 .
applying this method , we obtain a photometric redshift of the galaxy b as @xmath48 , being consistent with the spectroscopic redshift of the galaxy a. for the rest of the paper , we adopt @xmath49 for the galaxy threshing system ( @xmath50 at @xmath51 ) .
we investigate the morphological properties of the two tidal tails emanating from galaxy b. as shown in figure 1 , the two long tidal tails are obvious in images of all bands , suggesting that they consist of stellar material .
the lengths of the two tidal tails are 58 arcsec for the tail extending towards the galaxy a and 30 arcsec for the counter tail ( east of galaxy b ) .
these values correspond to lengths of 102 kpc and 52 kpc , respectively in the plane of the sky .
the tail extending toward galaxy a appears to be narrower than the counter tail whose width increases with increasing distance from galaxy b. this property is also seen in the first galaxy threshing system found by forbes et al .
( 2003 ) . in order to study this tendency more quantitatively
, we examine the surface brightness distributions of the tails . in figure 6
, we show the @xmath52-band surface brightness profile perpendicular to the tails for the four spatial positions , a , b , c , and d that are indicated in the right panel of figure 6 .
as shown in the left panel of figure 6 , the tail extending to galaxy a is significantly narrower than the counter one .
the tail widths measured at 3@xmath53 and 5@xmath53 noise level for each position are given in table 5 .
the width at position d exceeds 10 kpc even then it is measured at the 5@xmath53 noise level .
however , the width of the tail extending to galaxy a is only @xmath54 6 - 7 kpc .
there are some knotty structures in the tidal tails .
it is likely that some of these may be tidal dwarf candidates ( e.g. , zwicky 1956 ; schweizer 1978 ; barnes & hernquist 1992 ; yoshida , taniguchi , & murayama 1994 ; duc & mirabel 1998 and references therein ) .
future spectroscopic identification will be necessary to confirm them .
we estimate the mean surface brightness of the tidal tails in a sky area above @xmath55 of the surface brightness in each band .
the errors are estimated using following relation ; @xmath56 @xmath57 is the total number of pixels of tidal tail , and @xmath58 is @xmath59 sky background noise .
the results are given in table 6 .
the mean surface brightness is fainter by @xmath31 1 magnitude than that of the tail found by forbes et al .
( 2003 ) .
we also estimate the total magnitude of the tidal tails in each band .
errors are estimated using following relation ; @xmath60 the results are given in table 6 .
it is found that the magnitude of the tails is comparable to that of galaxy b in each band . in figure 7 , we show the @xmath61 color distribution in the galaxy threshing system .
the mean @xmath61 color of galaxy a is @xmath62 , while , that of galaxy b is @xmath63 .
there is a small color difference between the main body of galaxy b and the two tidal tails .
this suggests that stars located in the tidal tails were dispersed from the main body of galaxy b , providing strong evidence that galaxy threshing has been occurring in this system .
the total magnitude of the tidal tails is comparable to that of the main body of galaxy b. we estimated that the luminosity of galaxy b is @xmath64 , and that of the tail is @xmath65 .
these luminosities are nearly the same .
this suggests that galaxy b lost about half of its former mass during the ( on - going ) threshing event .
we assume that galaxy a has a mass - to - light ratio of 10 typical for an elliptical galaxies and 3 for galaxy b. given the @xmath21-band luminosity in section 3.1 , we obtain @xmath66 and @xmath67 for a and b , respectively .
next , we estimate the probable pericenter distance ( @xmath68 ) for the galaxy threshing event .
as shown in figure 5 , galaxy b shows an exponential surface brightness profile for @xmath69 arcsec .
if we assume that the mass outside of this radius has been expelled during the previous encounter between the two galaxies , and we use the equation of tidal radius as a rough guide of the pericenter distance , we can then estimate the pericenter as follows , @xmath70 using the derived masses of the two galaxies , we obtain a pericenter distance of @xmath71 kpc .
bekki et al .
( 2003 ) made general numerical simulations of a galaxy threshing systems with @xmath72 kpc .
mayer et al . (
2001 ) used a similar value , @xmath73 kpc .
our estimate is consistent with the results of these numerical simulations .
the projected separation between two galaxies is 102 kpc .
if we adopt an average relative velocity between the two galaxies of 200 km s@xmath11 , we find that the previous encounter occurred @xmath74 yrs ago .
galaxy b found in this study will merge into galaxy a after a few billion years after ( see bekki et al .
2001 ) . even in a facial phase of such minor mergers , faint remnants could be found around the host galaxy .
recently , morphological evidence for such a final phase of minor mergers has been found around field red elliptical galaxies ( van dokkum 2005 ) .
deep optical imaging survey for ordinary - looking galaxies should provide more information on the global , and dynamical evolution of galaxies .
in this paper , we have presented the discovery of a new potential galaxy threshing system .
although merging systems and interacting systems are often observed , it is rare to find faint tidal tails in on - going minor mergers because their tidal tails are generally too faint to be detected .
our discovery is important because we find clear evidence for the disruption of a small satellite galaxy .
this is only the second well - defined galaxy threshing system found so far .
the hst cosmos treasury program was supported through nasa grant hst - go-09822 .
we wish to thank tony roman , denise taylor , and david soderblom for their assistance in planning and scheduling of the extensive cosmos observations .
we gratefully acknowledge the contributions of the entire cosmos collaboration consisting of more than 70 scientists . more information on
the cosmos survey is available + at * http://www.astro.caltech.edu/~cosmos*. it is a pleasure the acknowledge the excellent services provided by the nasa ipac / irsa staff ( anastasia laity , anastasia alexov , bruce berriman and john good ) in providing on - line archive and server capabilities for the cosmos datasets .
the cosmos science meeting in may 2005 was supported in part by the nsf through grant oise-0456439 .
we would like to thank the subaru telescope staff for their invaluable assistance .
iraf ( image reduction and analysis facility ) is distributed by the national optical astronomy observatory , which is operated by the association of universities for research in astronomy , inc .
, under cooperative agreement with the national science foundation .
this work was financially supported in part by the ministry of education , culture , sports , science , and technology ( nos .
10044052 and 10304013 ) , and by the japan society for the promotion of science ( 15340059 and 17253001 ) .
sss and tn are financially supported by the japan society for the promotion of science ( jsps ) through jsps research fellowship for young scientists .
cccccccc j095959.6 + 020206(b ) & 9 59 59.61 & + 2 02 06.6 & 20.87 & 20.15 & 19.85 & 19.56 & 19.28 + & & & & & & & + + j100003.2 + 020146(a ) & 10 00 03.23 & + 2 01 46.5 & 17.29 & 16.33 & 15.92 & 15.48 & 15.21 + & & & & & & & + cccccc 3@xmath78 & ( kpc ) & @xmath79 & @xmath80 & @xmath81 & @xmath82 + & ( arcsec ) & @xmath83 & @xmath84 & @xmath85 & @xmath86 + + 5@xmath53 & ( kpc ) & @xmath87 & @xmath88 & @xmath89 & @xmath90 + & ( arcsec ) & @xmath91 & @xmath92 & @xmath93 & @xmath94 + * fig .
4 * surface brightness profiles of the threshing galaxy .
note that the central region ( @xmath98 ) of the threshing galaxy is saturated in @xmath23 and has been masked out .
we also show the @xmath59 of sky background and seeing radius ( @xmath99 ) .
the values of background are shown in the table 6 . *
fig . 6 * surface brightness distribution perpendicular to the extent of the tidal tails as a function of position along the tails .
the investigated spatial positions are marked by bars in the right panel .
the surface brightness profile at each position is shown in the left panel .
* @xmath100 color map .
image size of this figure is @xmath95 .
outer parts of the threshing system ( in particular , galaxy a ) show redder colors .
however , this feature is due to poor signal - to - noise ratio because of much lower surface brightness . | we report on the discovery of a new potential galaxy threshing system in the cosmos 2 square degree field using the prime - focus camera , suprime - cam , on the 8.2 m subaru telescope .
this system consists of a giant elliptical galaxy with @xmath0 and a tidally disrupted satellite galaxy with @xmath1 at a photometric redshift of @xmath2 .
this redshift is consistent with the spectroscopic redshift of 0.079 for the giant elliptical galaxy obtained from the sloan digital sky survey ( sdss ) archive .
the luminosity masses of the two galaxies are @xmath3 and @xmath4 , respectively .
the distance between the two galaxies is greater than 100 kpc .
the two tidal tails emanating from the satellite galaxy extend over 150 kpc .
this system would be the second well - defined galaxy threshing system found so far . |
besides their classical and traditional applications in signal processing and communications @xcite ( cf . also further references in @xcite ) and speech recognition @xcite , hidden markov models have recently become indispensable in computational biology and bioinformatics @xcite as well as in natural language modelling @xcite and information security @xcite . at the same time , their spatial extensions , known as hidden markov random field models ( hmrfm ) , have also been immensely influential in spatial statistics @xcite , and particularly in image analysis , restoration , and segmentation @xcite . indeed , hidden markov models are called ` one of the most successful statistical modelling ideas that have [ emerged ] in the last forty years ' @xcite .
hm(rf)ms owe much of their success on the one hand to the persistence of the markov property of the unobserved , or hidden , layer in the presence of observed data , and on the other , to the richness of the observed system @xcite . namely , in bayesian terms , in addition to the prior , the posterior distribution of the hidden layer also possesses a markov property ( albeit generally inhomogeneous even with homogeneous priors ) , whereas the marginal law of the observed layer can still include global , i.e. non - markovian , dependence .
the markov property of the posterior distribution and the conditional independence of the observed variables given the hidden ones , have naturally led to a number of computationally feasible methods for inference about the hidden realizations as well as model parameters ( if any ) .
hmms are naturally a special case of _ graphical models _
@xcite , ( * ? ? ?
* ch . 8) .
hmms , or one dimensional hmrfms , have been particularly popular not least due to the fact that linear order of the indexing set ( usually associated with time ) makes exploration of hidden realizations relatively straightforward from the computational viewpoint .
in contrast , higher dimensional hmrfms generally require approximate , possibly stochastic , techniques in order to compute optimal configurations of the hidden field @xcite .
in particular , _ maximum a posteriori _ ( map ) estimator of the hidden layer of an hmm is efficiently and exactly computed by a dynamic programming algorithm bearing the name of viterbi , whereas a general higher dimensional hmrfm would commonly employ a simulated annealing type method @xcite to produce approximate solutions to the same task .
we adopt the machine and statistical learning convention and therefore refer to the hidden and observed processes as @xmath0 and @xmath1 , respectively , in effect reversing the convention that is more commonly used in the hmm context . thus , let @xmath2 be a markov chain with state space @xmath3 , @xmath4 . even though we include inhomogeneous chains in most of what follows , for brevity we will still be suppressing the time index wherever this
does not cause ambiguity .
hence , we write @xmath5 for all transition matrices .
let @xmath6 be a process with the following properties .
first , given @xmath7 , the random variables @xmath8 are conditionally independent .
second , for each @xmath9 , the distribution of @xmath10 depends on @xmath11 ( and @xmath12 ) only through @xmath13 .
the process @xmath1 is sometimes called the _ hidden markov process _ ( hmp ) and the pair @xmath14 is referred to as a _ hidden markov model _ ( hmm ) .
the name is motivated by the assumption that the process @xmath0 ( sometimes called a _ regime _ ) is generally non - observable .
the conditional distribution of @xmath15 given @xmath16 is called an _ emission distribution _ , written as @xmath17 , @xmath18 .
we shall assume that the emission distributions are defined on a measurable space @xmath19 , where @xmath20 is usually @xmath21 and @xmath22 is the corresponding borel @xmath23-algebra .
without loss of generality , we assume that the measures @xmath17 have densities @xmath24 with respect to some reference measure @xmath25 , such as the counting or lebesgue measure
. given a set @xmath26 , integers @xmath27 and @xmath28 , @xmath29 , and a sequence @xmath30 , we write @xmath31 for the subsequence @xmath32 . when @xmath33 , it will be often suppressed .
thus , @xmath34 and @xmath35 stand for the fixed observed and unobserved realizations , respectively , of the hmm @xmath36 up to time @xmath37 .
any sequence @xmath38 is called a _
path_. we shall denote by @xmath39 the joint probability density of @xmath40 , i.e. @xmath41 overloading the notation , for every @xmath38 and for every sequence of observations @xmath42 , let @xmath43 and @xmath44 stand for the marginal probability mass function @xmath45 of path @xmath46 and probability density function @xmath47 of the data
@xmath42 , respectively .
it is a standard ( see , e.g. @xcite , ( * ? ? ?
* ch . 13 ) ) in this context to define the so - called _ forward _ and _ backward _ variables @xmath48 where @xmath49 and @xmath50 are the conditional densities of the data segments @xmath51 and @xmath52 , respectively , given @xmath53 .
_ segmentation _ here refers to estimation of the hidden path @xmath54 . treating @xmath54 as missing data @xcite , or parameters , a classical and by far the most popular solution to the segmentation problem is to maximize @xmath55 in @xmath38 . often , especially in the digital communication literature
@xcite , @xmath55 is called the likelihood function which might become potentially problematic in the presence of any genuine model parameters .
such `` maximum likelihood '' paths are also called _ viterbi paths _ or _ alignments _ after the viterbi algorithm @xcite commonly used for their computation .
if @xmath56 is thought of as the prior distribution of @xmath57 , then viterbi path also maximizes @xmath58 , the probability mass function of the posterior distribution of @xmath59 , hence the term ` _ maximum a posteriori ( map ) path _ ' . in spite of its computational attractiveness
, viterbi inference may be unsatisfactory for a number of reasons , including its suboptimality with regard to the number of correctly estimated states @xmath60 .
also , using the language of information theory , there is no reason to expect a viterbi path to be typical @xcite . indeed , `` there might be many similar paths through the model with probabilities that add up to a higher probability than the single most probable path '' @xcite .
the fact that a map estimate need not be representative of the posterior distribution has also been recently discussed in a more general context in @xcite .
atypicality of viterbi paths particularly concerns situations when estimation of @xmath54 is combined with inference about model parameters , e.g. transition probabilities @xmath61 @xcite .
even when estimating , say , the probability of heads from independent tosses of a biased coin , we naturally hope to observe a typical realization and not the constant one of maximum probability . an alternative and very natural way to estimate @xmath54 is by maximizing the posterior probability @xmath62 of each individual hidden state @xmath13 , @xmath63 .
we refer to the corresponding estimator as _ pointwise maximum a posteriori ( pmap)_. pmap is well - known to maximize the expected number of correctly estimated states ( section [ sec : risk ] ) , hence the characterization ` _ optimal accuracy _ ' @xcite . in statistics , especially spatial statistics and image analysis ,
this type of estimation is known as _ marginal posterior mode _
@xcite or _ maximum posterior marginals _
@xcite ( mpm ) estimation . in computational biology , this is also known as the _ posterior decoding _
( pd ) @xcite and has been reported to be particularly successful in pairwise sequence alignment @xcite . in the wider context of biological applications of discrete high - dimensional probability models ,
this has also been called _ consensus _ estimation , and in the absence of constraints , _ centroid _ estimation @xcite . in communications applications of hmms , largely influenced by @xcite , the terms ` _ optimal symbol - by - symbol decoding _ ' @xcite , ` _ symbol - by - symbol map estimation _ ' @xcite , and ` _ map state estimation _ ' @xcite have been used for this .
although optimal in the sense of maximizing the expected number of correctly estimated states , a pmap path might at the same time have low , in principle zero , probability @xcite .
it is actually not difficult to constrain the pmap decoder to _ admissible _ paths , i.e. of positive posterior probabilities as described in @xcite ( albeit in a slightly more general form allowing for state aggregation ) and also in subsection [ sec : pvdmore ] , , below .
a variation on this idea has been applied in @xcite for prediction of membrane proteins , giving rise to the term ` _ _ posterior viterbi decoding ( pvd ) _ _ ' @xcite .
pvd , however , maximizes the product @xmath64 @xcite ( and also below ) and not the sum @xmath65 , whereas the two criteria are _ no longer equivalent in the presence of path constraints _ ( subsection [ sec : pvdmore ] ) . in @xcite ,
a pmap decoder is proposed to obtain optimal pairwise sequence alignments .
the authors of @xcite use the term `` a legitimate alignment '' which suggests admissibility , but the description of the actual algorithm ( * ? ? ?
* section 3.8 ) appears to be insufficiently detailed to verify if the algorithm indeed enforces admissibility , or , if inadmissible solutions are altogether an issue in that context .
in many applications , e.g. gene identification , the pointwise ( e.g. nucleotide level ) error rate is not necessarily the main measure of accuracy , hence the constrained pmap need not be an ultimate answer .
together with the above problem of atypicality of map paths , this has been addressed by moving from single path inference towards _ envelops _ @xcite .
thus , for example , in computational biology the most common approach would be to aggregate individual states into a smaller number of semantic labels ( e.g. codon , intron , intergenic ) . in effect , this would realise the notion of path similarity by mapping many `` similar '' state paths to a single label path , or _ annotation _ @xcite . however , this leads to the problem of _ multiple paths _ , which in many practically important hmms renders the dynamic programming approach of the viterbi algorithm np - hard @xcite .
unlike the viterbi / map decoder , the pmap decoder handles annotations as easily as it does state paths , including the enforcement of admissibility @xcite .
a number of alternative heuristic approaches are also known in computational biology , but none appears to be fully satisfactory @xcite .
evidently , mapping optimal state paths to the corresponding annotations need not lead to optimal annotation and can actually give poor results @xcite .
overall , although the original viterbi decoder has still been the most popular paradigm in many applications , and in computational biology in particular , alternative approaches have demonstrated significantly higher performance , e.g. , in predicting various biological features .
for example , @xcite suggested the _
1-best _ algorithm for optimal labelling .
more recently , @xcite have demonstrated pvd to be superior to the 1-best algorithm , and not surprisingly , to the viterbi and pmap decoders , on tasks of predicting membrane proteins .
a starting point of this paper is that restricting the pmap decoder to paths of positive probability is but one of _ numerous ways to combine the useful features of the map and pmap path estimators_. indeed , as a sensible remedy against vanishing probabilities , in his popular tutorial @xcite rabiner briefly mentions maximization of the expected number of correctly decoded ( overlapping ) blocks of length two or three , rather than single states . with @xmath66 and @xmath67 being the block length and corresponding path estimate , respectively
, this approach yields viterbi inference as @xmath68 increases to @xmath69 ( with @xmath70 corresponding to pmap ) . therefore
, this approach could be interpreted as interpolating between the pmap and viterbi inferences .
intuitively , one might also expect @xmath71 to be strictly increasing with @xmath68 .
this is not exactly so as can be seen from example [ sec : example ] where @xmath72 .
however , we find the idea of interpolation between the pmap and viterbi inferences worth a further investigation . to the best of our knowledge ,
the only published work which explicitly proposes a solution to such interpolation is @xcite .
the approach of @xcite is algorithmic , directly based on continuous mappings , and also deserves an analysis which we present in subsection [ subsec : alg ] . in this paper
, we consider the segmentation problem in the more general framework of statistical learning .
namely , we consider sequence _ classifier _ mappings @xmath73 and optimality criteria for their selection . in section [ sec : risk ] , criteria for optimality of @xmath74 are naturally formulated in terms of risk minimization whereby @xmath75 , the _ risk of _
@xmath46 , derives from a suitable _ loss function_. in section [ sec : combrisk ] , we consider families of risk functions which naturally generalize those corresponding to the viterbi and pmap solutions ( subsection [ subsec : viterbipmap ] ) .
furthermore , as shown in section [ sec : bridge ] , these risk functions define a family of path decoders parameterized by an integer @xmath68 with @xmath76 and @xmath77 corresponding to the pmap and viterbi cases , respectively ( theorem [ k - block ] ) .
we also show the close connection between the aforementioned family of decoders and the rabiner @xmath68-block approach .
if needed , then the new family of decoders can easily be embedded into a yet wider class with a principled criterion of optimality .
all these decoders ( classifiers ) would only be of theoretical interest if they could not be easily calculated . in section
[ sec : combrisk ] , we show that all of the newly defined decoders can be implemented efficiently as a dynamic programming algorithm in the usual forward - backward manner with essentially the same ( computational as well as memory ) complexity as the pmap or viterbi decoders ( theorem [ dyn ] ) .
given a sequence of observations @xmath42 , we define the ( posterior ) _ risk _ to be a function @xmath78.\ ] ] naturally , we seek a state sequence with minimum risk : @xmath79 following the _ statistical decision and pattern recognition theories _ , the classifier @xmath80 will be referred to as the _ bayes classifier _
( relative to risk @xmath81 ) . within the same framework
, the risk is often specified via a _ loss - function _
@xmath82,\ ] ] interpreting @xmath83 as the loss incurred by the decision to predict @xmath84 when the actual state sequence was @xmath85 .
therefore , for any state sequence @xmath38 , the risk is given by @xmath86=\sum_{a^t\in s^t}l(a^t , s^t)p(a^t|x^t).\ ] ] the most popular loss function is the so - called _ symmetrical _ or _ zero - one _ loss @xmath87 defined as follows : @xmath88 we shall denote the corresponding risk by @xmath89 . with this loss , clearly @xmath90 thus @xmath91 is minimized by a viterbi path , i.e. a sequence of maximum posterior probability .
let @xmath92 stand for the corresponding classifier , i.e. @xmath93 with a suitable tie - breaking rule .
evidently , viterbi paths also minimize the following risk @xmath94 it can actually be advantageous to use the log - likelihood based risk since , as we shall see later , it leads to various natural generalizations ( sections [ sec : combrisk ] and [ sec : bridge ] ) . when sequences are compared pointwise , it is common to use additive loss - functions of the form @xmath95 where @xmath96 is the loss associated with classifying the @xmath12-th element @xmath97 as @xmath98 .
typically , for every state @xmath99 , @xmath100 .
it is not hard to see that , with @xmath101 as in , the corresponding risk can be represented as follows @xmath102 where @xmath103 .
most commonly , @xmath104 is again symmetrical , or zero - one , i.e. @xmath105 , where @xmath106 stands for the indicator function of set @xmath107 . in this case
, @xmath108 is naturally related to the _ hamming distance _ @xcite .
then also @xmath109 so that the corresponding risk is @xmath110 let @xmath111 stand for the bayes classifier relative to the @xmath112-risk .
it is easy to see from the above definition of @xmath112 , that @xmath111 delivers pmap paths , which clearly minimize the expected number of misclassification errors .
in addition to maximizing @xmath113 , @xmath114 also maximizes the _ pseudolikelihood _ @xmath115 , and therefore minimizes the following _ log - pseudolikelihood _ risk @xmath116 recall ( subsection [ sec : segment ] ) that pmap paths can be of zero probability ( i.e. not admissible ) . to ensure admissibility , @xmath112-risk
can simply be minimized over the admissible paths : @xmath117 assuming that @xmath118 , @xmath63 , @xmath119 , have been precomputed ( by the classical forward - backward recursion @xcite ) , the solution of can be easily found by a viterbi - like recursion @xmath120 where @xmath121 , @xmath122 .
the recursion is also equivalent to @xmath123 however , in the presence of path constraints , minimization of the @xmath112-risk is no longer equivalent to minimization of the @xmath124-risk .
in particular , the problem is not equivalent to the following problem ( _ posterior - viterbi decoding _ ) @xmath125 a solution to can be computed by a related recursion given in below @xmath126 recursion is clearly equivalent to @xmath127 although admissible minimizers of @xmath112 and @xmath124 risk are by definition of positive probability , this probability might still be very small .
indeed , in the above recursions , the weight @xmath128 is 1 even when @xmath61 is very small .
we next replace @xmath128 ( @xmath129 ) by the true transition ( initial ) probability @xmath61 ( @xmath130 ) in minimizing the @xmath124-risk ( i.e. maximization of @xmath115 ) .
then the solutions remain admissible and now also tend to maximize the prior path probability . with the above replacements , recursions and now solve the following _ seemingly unconstrained _ optimization problem ( see theorem [ dyn ] ) @xmath131\quad \leftrightarrow \quad \min_{s^t}\big[\bar{r}_1(s^t|x^t)+h(s^t)\big],\ ] ] where the penalty term @xmath132 is the prior log - likelihood risk which does not depend on the data .
the thereby modified recursions immediately generalize as follows : @xmath133 solving @xmath134,\ ] ] where @xmath135 is a regularization constant and @xmath136 ( see section [ sec : combrisk ] and theorem [ dyn ] ) . then , pvd , i.e. the problem solved by the original recursions and , can be recovered by taking @xmath137 sufficiently small .
( alternatively , the pvd problem can also be formally written in the form with @xmath138 and @xmath139 given , for example , by @xmath140 . )
what if the actual probabilities @xmath61 ( @xmath130 ) were also used in the optimal accuracy / pmap decoding , i.e. optimization - ?
it appears more sensible to replace the indicators @xmath128 ( @xmath129 ) with @xmath61 ( @xmath130 ) in ( and not in ) .
this solves the following problem : @xmath141\quad \leftrightarrow \quad \min_{s^t}\big[r_1(s^t|x^t ) + \r_{\infty}(s^t)\big].\ ] ] a more general problem can be written in the form @xmath142,\ ] ] where @xmath143 is some penalty function ( independent of the data @xmath42 ) .
thus , the problem of optimal accuracy / pmap decoding over the admissible paths is obtained by taking @xmath137 sufficiently small and @xmath136 .
( setting @xmath144 also reduces the problem back to . )
motivated by the previous section , we consider the following general problem @xmath145,\ ] ] where @xmath146 , @xmath147 , @xmath148 .
this is also equivalent to @xmath149,\ ] ] where @xmath150,\nonumber\\ & = -{1\over t}[\log \pi_{s_1}+ \sum_{t=1}^{t-1}\log p_{s_t s_{t+1}}+\sum_{t=1}^t \log f_{s_t}(x_t)],\nonumber \\
\bar{r}_{\infty}(s^t|x^t)&={1\over t}\log p(s^t|x^t),\quad\text{recalling}~\eqref{loglikerisk},\nonumber \\ & = \r_{\infty}(s^t , x^t)-{1\over t}\log p(x^t),\nonumber \\
\bar{r}_1(s^t)&:=-{1\over t}\sum_{t=1}^t\log p_t(s_t),\label{eqn : barr1prior}\\ \bar{r}_{\infty}(s^t)&=-{1\over t}\log p(s^t),\quad\text{recalling}~\eqref{eq : rinftyprior},\nonumber \\ & = -{1\over t}[\log \pi_{s_1}+
\sum_{t=1}^{t-1}\log p_{s_t s_{t+1}}].\end{aligned}\ ] ] the newly introduced risk @xmath151 is the prior log - pseudo - likelihood .
evidently , the combination @xmath152 corresponds to the map / viterbi decoding ; the combination @xmath153 yields the pmap case , whereas the combinations @xmath154 and @xmath155 give the _
maximum a priori _ decoding and _ marginal prior mode _ decoding , respectively . the case @xmath156 subsumes and the case @xmath157 is the problem @xmath158.\ ] ] thus , a solution to is a generalization of the viterbi decoding that allows one to suppress ( @xmath135 ) contribution of the data . _
it is important to note that with @xmath159 every solution of ( [ gen - problem ] ) is admissible . _
_ no less important , and perhaps a bit less obvious , is that @xmath160 also guarantees admissibility of the solutions _ , as stated in proposition [ prop : admit ] below .
[ prop : admit]let @xmath160 .
then , for _ almost every
_ realization @xmath40 of the hmm process @xmath161 , the minimized risk is finite and any minimizer @xmath46 is admissible , i.e. satisfies @xmath162 . without loss of generality ,
assume @xmath156 .
suppose the problem has no finite solution .
then for any @xmath46 with @xmath163 , we would have some @xmath12 , @xmath63 , such @xmath164 .
this would imply that @xmath165 for all @xmath46 with @xmath163 , contradicting the hypothesis that @xmath166 . now
, suppose that @xmath46 is a minimizer of but @xmath167 .
since @xmath163 , we would have some @xmath12 , @xmath63 , such @xmath168 .
this would imply @xmath169 , and subsequently that @xmath164 and @xmath170 , contradicting optimality of @xmath46 .
[ rem : pvd ] thus , note that the posterior - viterbi decoding @xcite can be obtained by either setting @xmath171 and taking @xmath172 , or setting @xmath156 and taking @xmath173 .
if the smoothing probabilities @xmath118 , @xmath174 and @xmath119 , have been already computed ( say , by the usual forward - backward algorithm ) , a solution to can be found by a standard dynamic programming algorithm .
let us first introduce more notation .
for every @xmath175 and @xmath119 , let @xmath176 note that the function @xmath177 depends on the entire data @xmath42 .
next , let us also define the following scores @xmath178 using the above scores @xmath179 and a suitable tie - breaking rule , below we define the back - pointers @xmath180 , terminal state @xmath181 , and the optimal path @xmath182.@xmath183 , \quad \hbox{when $ t=1,\ldots , t-1 $ ; } \nonumber \\ i_t&:= \arg\max_{i\in s}\delta_t(i).\\ \label{st } \hat s^{t}(j)&:=\left\ { \begin{array}{ll } i_1(j ) , & \hbox{when $ t=1 $ ; } \\
\big(\hat s^{t-1}(i_{t-1}(j)),j \big ) & \hbox{when $ t=2,\ldots , t$. } \end{array } \right.\end{aligned}\ ] ] the following theorem formalizes the dynamic programming argument ; its proof is standard and we state it below for completeness only .
[ dyn ] any solution to can be represented in the form @xmath184 provided the ties in are broken accordingly . with a slight abuse of notation , for every @xmath185 ,
let @xmath186,\ ] ] where @xmath187 and @xmath188 .
hence , @xmath189=u(s^t)\ ] ] and any maximizer of @xmath190 is clearly a solution to and .
next , note that @xmath191 for all @xmath119 , and that @xmath192 for @xmath193 and also @xmath185 . by induction on @xmath12 , these yield @xmath194 for every @xmath195 and for all @xmath119 .
clearly , every maximizer @xmath196 of @xmath190 over the set @xmath197 must satisfy @xmath198 , or , more precisely @xmath199 , allowing for non - uniqueness . continuing to interpret @xmath200 as a set
, recursion implies recursions and , hence any maximizer @xmath196 can indeed be computed in the form @xmath201 via the _ forward _ ( recursion ) -_backward _ ( recursion ) procedure .
1similarly to the generalized risk minimization of , the generalized problem of accuracy optimization can also be further generalized as follows : @xmath202,\ ] ] where risk @xmath203 is the error rate relative to the prior distribution .
this problem apparently can be solved by the following recursion @xmath204 where now @xmath205 as in the generalized _ posterior - viterbi decoding _
, here @xmath159 also implies admissibility of the optimal paths .
however , unlike in , @xmath160 is not sufficient to guarantee admissibility of the solutions .
we have been discussing a set of related ideas which allow us to balance path accuracy against path probabilities .
next , we extend this discussion by presenting a couple of notably different approaches . recall ( subsection [ sec : segment ] ) that rabiner s compromise between map and pmap is to maximize the expected number of correctly decoded pairs or triples of ( adjacent ) states . with
@xmath68 being the length of the overlapping block ( @xmath206 ) this means to minimize the conditional risk @xmath207 which derives from the following loss function : @xmath208 obviously , for @xmath76 this gives the usual @xmath112 maximization the pmap decoding which is known to fault by allowing inadmissible paths .
it is natural to think that minimizers of @xmath209 `` move '' towards viterbi paths `` monotonically '' as @xmath68 increases to @xmath69 . indeed , when @xmath210 , minimization of @xmath209 is equivalent to minimization of @xmath211 achieved by the viterbi decoding .
however , as example [ sec : example ] shows below , minimizers of are not guaranteed to be admissible for @xmath212 , which is a drawback of using the loss @xmath213 .
we now show that this drawback is easily overcome when the sum in is replaced by the product .
certainly , these problems are not equivalent , and in particular with the product in place of the sum the @xmath68-block idea works well .
namely , the longer the block , the larger the resulting path probability , which is also now guaranteed to be positive already for @xmath214 . moreover ,
this gives another interpretation of the risks @xmath215 ( see also remark [ rem : pvd ] above ) and , though perhaps less interestingly , the prior risks @xmath216 . let @xmath68 be a positive integer .
for the time being , let @xmath217 represent any first order markov chain on @xmath197 , and let us define @xmath218 thus @xmath219 where @xmath220 thus , @xmath221 is a natural generalization of @xmath124 ( introduced first for the posterior distribution in ) since when @xmath76 , @xmath222 .
[ k - block ] let @xmath68 be such that @xmath223
. then the following recursion holds @xmath224 note that @xmath225 next , for all @xmath226 such that @xmath227 , the markov property gives @xmath228 and @xmath229 hence , @xmath230 the second equality above also follows from the markov property .
taking logarithms on both sides and dividing by @xmath231 completes the proof .
now , we specialize this result to our hmm context , and , thus , @xmath43 and @xmath232 are again the prior and posterior hidden path distributions .
[ corollary0]let @xmath68 be such that @xmath223 .
for all paths @xmath38 the prior risks @xmath221 and @xmath233 satisfy . for every @xmath234 and for all paths @xmath38 , the posterior risks @xmath221 and @xmath233 satisfy .
@xmath235 clearly , conditioned on the data @xmath42 , @xmath59 remains a first order markov chain ( generally inhomogeneous even if it was homogeneous _ a priori _ ) .
hence , theorem [ k - block ] applies .
below , we focus on the posterior distribution and risks , even though the following would readily extend to any first order markov chain
. let @xmath236 be a classifier that minimizes @xmath237 .
thus , @xmath238 we refer to such classifiers as _ k - block _ pvd or _
k - block _ pmap as they naturally extend @xmath111 , the pmap / optimal accuracy / posterior decoder ( section [ subsec : viterbipmap ] ) . to be consistent with applications @xcite ,
the term ` @xmath68-block posterior - viterbi decoding ' , however , is perhaps more accurate given the use of the product - based risk @xmath239 as opposed to @xmath240 .
now , we present some properties of the new risks and decoders .
[ corollary ] for every @xmath234 , and for every @xmath38 , we have @xmath241 equation follows immediately from equation of corollary [ corollary0 ] .
inequality follows from inequalities below @xmath242 which in turn follow from equation of corollary [ corollary0 ] .
also , equation implies that for every @xmath243 , @xmath244 which , together with inequality , implies .
inequality means that the posterior path probability @xmath245 increases with @xmath68 .
equation is also of practical significance showing that @xmath236 is a solution to with @xmath246 , @xmath247 , @xmath171 , and as such can be computed in the same fashion for all @xmath68 ( see theorem [ dyn ] above ) .
thus , increasing @xmath68 increases @xmath124-risk , i.e. decreases the product of the ( conditional ) marginal probabilities of states along the path @xmath236 .
inequalities and clearly show that as @xmath68 increases , @xmath248 monotonically moves from @xmath111 ( pmap ) towards the viterbi decoder , i.e. @xmath249 .
however , the maximum block length is @xmath210 .
a natural way to complete this bridging of pmap with map is by embedding the collection of risks @xmath221 into the family @xmath250 via @xmath251 $ ] .
thus , extends to @xmath252 with @xmath253 and @xmath254 corresponding to the viterbi and pmap cases , respectively . given @xmath42 and a sufficiently small @xmath255 ( equivalently , large @xmath68 ) , @xmath236 , the minimizer of @xmath256 ( or , the right hand side of )
would produce a viterbi path @xmath257 ( since @xmath197 is finite ) .
however , such @xmath255 ( and @xmath68 ) would generally depend on @xmath42 , and in particular @xmath68 may need to be larger than @xmath69 , i.e. @xmath258 may be different from @xmath257 . at the same time , we clearly have @xmath259 on which we comment more in section below .
an alternative that does not involve the risk functions is to simply transform the forward and backward variables @xmath260 and @xmath261 defined in .
consider , for one example , the recursively applied power transformations given in below @xmath262^\frac{1}{q}f_j(x_{t+1}),\quad 1\le t < t \nonumber\\
\beta_{t}(j ; q ) & : = & \left[\sum_{i\in s}\left(p_{ji}f_i(x_{t+1})\beta_{t+1}(i ; q)\right)^q\right]^\frac{1}{q},\quad 1\le t < t
\\ \beta_t(j ; q)&:=&\beta_t(j)=1.\nonumber\end{aligned}\ ] ] clearly , @xmath263 and @xmath264 , for all @xmath119 and all @xmath265 .
thus , @xmath266 leads to the pmap decoding . using induction on @xmath12 and continuity of the power transform
, it can also be seen that the following limits exist and are finite for all @xmath267 and all @xmath265 : @xmath268 and @xmath269 , where for @xmath270 @xmath271 and therefore , any viterbi path @xmath272 has the following property : @xmath273 this has been already been pointed out by @xcite , who , to the best of our knowledge , @xcite have been the only group to publish on the idea of hybridization of the pmap and viterbi decoders via a continuous transformation . ignoring potential non - uniqueness of viterbi paths , @xcite state , based on , that the viterbi path can be found _ symbol - by - symbol_.
certainly , when viterbi paths are non - unique , symbol - by - symbol decoding based on can produce suboptimal , and in principle inadmissible , paths .
in contrast to viterbi , non - unique pmap paths ( in the absence of constraints ) can certainly be found symbol - by - symbol .
note also that for their hybrid of pmap with viterbi , @xcite use the following transformations : @xmath274 where @xmath275 ( in our notation ) and @xmath276 are required to be continuous on @xmath277 with finite limits as @xmath278 .
these @xmath279 are then substituted for by appropriate expressions in terms of the recursively transformed forward and backward variables .
the following points with regard to this hybridization idea have also motivated our present work : 1 .
transformations appear to be somewhat more sophisticated than the power transforms .
it appears that the only reason explicitly stated in @xcite for making their choice of transformation is to deliver the correct limits ( in that case pmap with @xmath280 and viterbi with @xmath281 ) . besides and , there are other ( single parameter ) transformations meeting this condition .
2 . implicitly , the authors of @xcite do recognize the usual problem of numerical underflow , but somehow appear to suggest that the rescaling trick @xmath282 would be sufficient to resolve this problem when computing their transformed variables .
we could not experimentally confirm this optimism even with simple models .
in fact , without addressing the scaling issue in full , the expressions and are short of defining practically meaningful path decoders .
true , it is not difficult to renormalize these , or similar , expressions while preserving their limiting behavior .
however , unlike in the original , i.e. untransformed , forward - backward algorithm , _ renormalization of the transformed forward and backward variables will alter the original decoders for intermediate values of the tuning parameter_. this can already be suspected by examining equations below @xmath283^\frac{1}{q}f_j(x_{t+1 } ) } { \sum_{l\in s}\left[\sum_{i\in s}\left(\tilde \alpha_t(i ; q)p_{il}\right)^q\right]^\frac{1}{q}f_l(x_{t+1})},\quad 1\le t < t
\nonumber\\ \tilde \beta_{t}(j ; q ) & : = & \frac{\left[\sum_{i\in s}\left(p_{ji}f_i(x_{t+1})\tilde \beta_{t+1}(i ; q)\right)^q\right]^\frac{1}{q } } { \sum_{l\in s}\left[\sum_{i\in s}\left(\tilde \alpha_t(i ; q)p_{il}\right)^q\right]^\frac{1}{q}f_l(x_{t+1})},\quad 1\le t <
\tilde \beta_t(j ; q)&:=&\beta_t(j)=1.\nonumber\end{aligned}\ ] ] which implement the usual rescaling @xcite .
3 . moreover , algorithmically defined estimators are generally hard to analyze rigorously @xcite . in our context ,
optimization criteria for intermediate members of the above interpolating families are indeed not clear , making it difficult to interpret the corresponding decoders .
this might be discouraging should such decoders be included in more complex inference cycles ( i.e. when any genuine model parameters are estimated as well , e.g. viterbi training @xcite ) .
other recursion schemes ( for example , cf .
@xcite for derin s formula ) can surely be experimented with in a similar manner . however , now more than ten years after appearance of @xcite , we find that _ the value of any such interpolation is yet to be demonstrated .
as already mentioned above , the symbol - by - symbol implementation of the transform - based hybrids is problematic when the solution is non - unique and full path probability is a factor .
given a classifier @xmath74 and a risk function @xmath81 , the quantity @xmath284 evaluates the risk when @xmath74 is applied to a given sequence @xmath42 . when @xmath74 is optimal in the sense of risk minimization , then @xmath285 .
we are also interested in the random variables @xmath286 .
thus , in @xcite , convergence of several risks of the viterbi decoding has been considered .
based on the asymptotic theory of viterbi processes @xmath287 @xcite , it has been shown that under fairly general assumptions on the hmm , the random variables @xmath288 , @xmath289 , @xmath290 as well as @xmath291 , @xmath292 and @xmath293 all converge to constant limits , _ a.s._. convergence of these risks obviously imply convergence of @xmath294 and @xmath295 the risks appearing in the generalized problems and , respectively .
actually , convergence of @xmath296 is also proved ( and used in the proof of convergence of @xmath297 ) .
hence , the risk in , evaluated at viterbi paths , converges as well . the limits _ asymptotic risks _ are ( deterministic ) constants that depend only on the model and evaluate the viterbi inference .
for example , let @xmath298 be the limit of @xmath288 , which is the asymptotic misclassification rate of the viterbi decoding .
thus , for big @xmath69 , the viterbi decoding makes about @xmath299 misclassification errors .
the asymptotic risks might be , in principle , found theoretically , but as the limit theorems show , the limiting risks can also be estimated by simulations . in @xcite , it has been also shown that under the same assumptions @xmath300 converges to a constant limit , say @xmath301 . in @xcite , @xmath302 has been also shown to converge .
clearly @xmath303 , and even if their difference is small , the number of errors made by the viterbi decoder in excess of pmap in the long run can still be found significant .
presently , we are not aware of a universal method for proving the limit theorems for these risks . convergence of the risks of the viterbi decoding is possible due to the existence of the so - called viterbi process ( see @xcite ) that has nice ergodic properties .
the question whether infinite pmap processes have similar properties , is still open .
therefore , convergence of @xmath304 was proven with a completely different method based on the smoothing probabilities .
in fact , all of the limit theorems obtained thus far have been proven with different methods .
we conjecture that these different methods can be combined so that convergence of the minimized combined risk or could be proven as well . in summary , as mentioned before , thus far convergence of the minimized combined risks has been obtained for trivial combinations only , i.e. with three of the four constants being zero .
note that while convergence of the intermediate case @xmath305\ ] ] with its minimizer @xmath306 is an open question , gives @xmath307 this , together with the _ a.s . _ convergence of @xmath289 , implies that in the long run , for most sequences @xmath42 , @xmath308 will not exceed @xmath309 by more than @xmath310 .
since this limit is finite , letting @xmath137 increase with @xmath69 , @xmath311 obviously approaches @xmath312 _ a.s .
_ , i.e. as the intuition predicts , the likelihood of @xmath313 approaches to that of @xmath314
certainly , the logarithmic risks , , , on the one hand , and the ordinary risks , , @xmath315 , , on the other , can be respectively combined into a single parameter family by , for example , the power transformation as shown below .
let @xmath217 for the moment be any probability distribution on @xmath197 .
@xmath316 thus , our two generalized problems and are naturally members of the same family of problems : @xmath317,\ ] ] where @xmath146 and @xmath318 , @xmath147 , and @xmath148 .
clearly , the dynamic programming approach of theorem [ dyn ] and immediately applies to any member of the above family with @xmath319 .
theorem [ k - block ] and corollaries [ corollary0 ] and [ corollary ] obviously generalize to higher order markov chains as can be seen below .
let @xmath217 represent a markov chain of order @xmath27 , @xmath320 , on @xmath197 .
then for any @xmath38 and for any @xmath321 , we have @xmath322 this is a straightforward extension of the proof of theorem .
the present risk - based discussion of hmm path inference also naturally extends to the problem of optimal _ labelling _ or _ annotation _ ( section [ sec : segment ] ) .
namely , the state space @xmath323 can be partitioned into subsets @xmath324 , @xmath325 ,
, @xmath326 , for some @xmath327 , in which case @xmath328 assigns label @xmath25 to every state @xmath329 .
the fact that the pmap problem is as easily solved over the label space @xmath330 as it is over @xmath197 has already been used in practice . indeed , adding the admissibility constraint , @xcite in effect average @xmath331 s within the label classes and
then use recursions to obtain the pmap labelling , say , @xmath332 , of admissible state paths .
this approach clearly corresponds to using the point loss @xmath333 in when solving @xmath334 .
importantly , our generalized problem also immediately incorporates the above pointwise label - loss in either the prior @xmath335 or posterior risk @xmath336 , or both .
since computationally these problems are essentially as light as and since @xcite report their special case to be useful in practice , we believe that the above generalizations offer yet more useful possibilities to practitioners . note the different kinds of averaging corresponding to different values of @xmath337 to be used with the @xmath112 risks : @xmath338 certainly , the choice of the basic loss functions , inflection parameters @xmath339 and weights @xmath340 of the respective risks , is application dependent , and can be tuned with the help of labelled data , using cross - validation .
finally , these generalizations are presented for the standard hmm setting , and work on extending them to more complex and practically more useful hmm - based settings ( e.g semi - markov , autoregressive , etc . ) is underway .
consider the following four - state mc transition matrix @xmath341 suppose observations @xmath342 and the emission densities @xmath24 @xmath343 are such that @xmath344 hence every admissible path begins and ends with 2 .
thus , to simplify the notation , we assume without loss of generality that @xmath345 . amongst the paths that begin and end with 2 , the paths whose probabilities are listed below , are
the only ones of positive ( prior ) probability ( the probabilities below are calculated up to the normalization constant ) : the viterbi paths here are @xmath357 . also note that , for every @xmath147 , @xmath358 and @xmath359 .
thus @xmath360 is also the unique pmap path .
minimization of @xmath361 over all @xmath362 here is equivalent to maximization of @xmath363 , and the optimal paths are @xmath364 and @xmath365 .
l. bahl , j. cocke , f. jelinek , and j. raviv .
optimal decoding of linear codes for minimizing symbol error rate ( corresp . ) .
_ information theory , ieee transactions on _ , 200 ( 2):0 284287 , 1974 .
url http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=1055186 . christopher m. bishop .
_ pattern recognition and machine learning_. information science and statistics .
springer , new york , 2006 .
isbn 978 - 0387 - 31073 - 2 ; 0 - 387 - 31073 - 8 .
doi : 10.1007/978 - 0 - 387 - 45528 - 0 .
url http://dx.doi.org/10.1007/978-0-387-45528-0 .
broa brejov , daniel g. brown , and tom vina .
the most probable annotation problem in hmms and its application to bioinformatics . _
j comput syst sci _
, 730 ( 7):0 1060 1077 , 2007 .
issn 0022 - 0000 .
doi : doi : 10.1016/j.jcss.2007.03.011 .
url http://www.sciencedirect.com / science / article / b6wj0 - 4n85bd6 - 1/2/ffd66145% 030eb2db405304f3de9bf6c4[http://www.sciencedirect.com / science / article / b6wj0 - 4n85bd6 - 1/2/ffd66145% 030eb2db405304f3de9bf6c4 ] .
bioinformatics iii .
chris burge and samuel karlin .
prediction of complete gene structures in human genomic dna .
_ journal of molecular biology _ , 2680 ( 1):0 78 94 , 1997 .
issn 0022 - 2836 .
doi : doi : 10.1006/jmbi.1997.0951 .
url http://www.sciencedirect.com / science / article / b6wk7 - 45vgf7t-9/2/0db81327% 54939b6d0d07e85a6276d801[http://www.sciencedirect.com / science / article / b6wk7 - 45vgf7t-9/2/0db81327% 54939b6d0d07e85a6276d801 ] .
olivier capp , eric moulines , and tobias rydn .
_ inference in hidden markov models_. springer series in statistics .
springer , new york , 2005 .
isbn 978 - 0387 - 40264 - 2 ; 0 - 387 - 40264 - 0 . with randal douc s contributions to chapter 9 and christian p. robert s to chapters 6 , 7 and 13 , with chapter 14 by gersende fort , philippe soulier and moulines , and chapter 15 by stphane boucheron and elisabeth gassiat .
yariv ephraim and neri merhav .
hidden markov processes .
_ ieee trans .
inform . theory _
, 480 ( 6):0 15181569 , june 2002 .
issn 0018 - 9448 .
special issue on shannon theory : perspective , trends , and applications .
piero fariselli , pier martelli , and rita casadio .
a new decoding algorithm for hidden markov models improves the prediction of the topology of all - beta membrane proteins . _
bmc bioinformatics _ , 60 ( suppl 4):0 s12 , 2005 .
issn 1471 - 2105 .
doi : 10.1186/1471 - 2105 - 6-s4-s12 .
j. hayes , t. cover , and j. riera .
optimal sequence detection and optimal symbol - by - symbol detection : similar algorithms . _ communications , ieee transactions on _ , 300 ( 1):0 152157 , january 1982 .
issn 0090 - 6778 .
url http://ieeexplore.ieee.org/xpl/freeabs_all.jsp?arnumber=1095391 .
g. ji and j. bilmes . backoff model training using partially observed data : application to dialog act tagging . in _ proc .
human language techn .
naacl , main conference _ , pages 280287 , new york city , usa , 2006 .
association for computational linguistics .
url http://www.aclweb.org/anthology/n/n06/n06-1036 .
dhiraj joshi , jia li , and james z. wang . a computationally efficient approach to the estimation of two- and three - dimensional hidden markov models .
_ ieee trans . image process .
_ , 150 ( 7):0 18711886 , 2006 .
issn 1057 - 7149 .
lukas kll , anders krogh , and erik l. l. sonnhammer . an hmm posterior decoder for sequence feature prediction that includes homology information .
_ bioinformatics _ , 210 ( suppl_1):0 i251257 , 2005 .
doi : 10.1093/bioinformatics / bti1014 .
url http://bioinformatics.oxfordjournals.org / cgi / content / abstract/21/suppl% _ 1/i251[http://bioinformatics.oxfordjournals.org / cgi / content / abstract/21/suppl% _ 1/i251 ] .
alexey koloydenko and jri lember .
infinite viterbi alignments in the two state hidden markov models .
_ acta comment .
_ , 0 ( 12):0 109124 , 2008 .
issn 1406 - 2283 .
8th tartu conf .
multivariate statist .
june 2007 .
steffen l. lauritzen .
_ graphical models _ , volume 17 of _ oxford statistical science series_. the clarendon press oxford university press , new york , 1996 .
isbn 0 - 19 - 852219 - 3 .
oxford science publications .
jri lember and alexey koloydenko .
http://projecteuclid.org/handle/euclid.bj/1202492790/[the adjusted viterbi training for hidden markov models ] .
_ bernoulli _ , 140 ( 1):0 180206 , february 2008 .
issn 1350 - 7265 .
doi : 10.3150/07-bej105 .
jia li , robert m. gray , and richard a. olshen .
multiresolution image classification by hierarchical modeling with two - dimensional hidden markov models .
_ ieee trans .
inform . theory _
, 460 ( 5):0 18261841 , 2000 .
issn 0018 - 9448 .
information - theoretic imaging .
shu lin and daniel j. costello jr . _ error control coding : fundamental and applications_. computer applications in electrical engineering .
prentice - hall , inc .
, englewood cliffs , new jersey 07632 , 1983 .
isbn 0 - 13 - 283796-x .
william h. majoros and uwe ohler . _ knowledge discovery and emergent complexity in bioinformatics _ ,
volume 4366/2007 of _ lecture notes in computer science _ , chapter advancing the state of the art in computational gene prediction , pages 81106 .
springer berlin / heidelberg , 2007 .
doi : 10.1007/978 - 3 - 540 - 71037 - 0_6 .
url http://www.springerlink.com/content/x7h8126560375r25 .
jose l. marroquin , edgar arce santana , and salvador botello .
hidden markov measure field models for image segmentation .
_ ieee trans . pattern anal .
_ , 250 ( 11):0 13801387 , 2003 .
issn 0162 - 8828 .
doi : http://dx.doi.org/10.1109/tpami.2003.1240112 .
joshua mason , kathryn watkins , jason eisner , and adam stubblefield . a natural language approach to automated cryptanalysis of two - time pads . in _
ccs 06 : proceedings of the 13th acm conference on computer and communications security _ , pages 235244 , new york , ny , usa , 2006 .
isbn 1 - 59593 - 518 - 5 .
doi : http://doi.acm.org/10.1145/1180405.1180435 .
c. a. mcgrory , d. m. titterington , r. reeves , and a. n. pettitt .
variational bayes for estimating the parameters of a hidden potts model . _
statistics and computing _ , 190 ( 3):0 329340 , 2009 .
issn 0960 - 3174 .
doi : http://dx.doi.org/10.1007/s11222-008-9095-6 .
h. ney , v. steinbiss , r. haeb - umbach , and _ et .
_ an overview of the philips research system for large vocabulary continuous speech recognition . _
. j. pattern recognit .
_ , 80 ( 1):0 3370 , 1994 .
patrick robertson , emmanuelle villebrun , and peter hoeher .
a comparison of optimal and sub - optimal map decoding algorithms operating in the log domain . in _ communications , 1995 .
icc 95 seattle , gateway to globalization , 1995 ieee international conference on _ , volume 2 , pages 10091013 , 1995 .
doi : 10.1109/icc.1995.524253 .
i. shu , l. hetherington , and j. glass .
baum - welch training for segment - based speech recognition . in _ proc .
ieee asru workshop _ ,
pages 4348 , st .
thomas , u. s. virgin islands , http://groups.csail.mit.edu/sls/publications/2003/asru_shu.pdf , 2003 .
a. viterbi .
http://ieeexplore.ieee.org/search/wrapper.jsp?arnumber=1054010[error bounds for convolutional codes and an asymptotically optimum decoding ] algorithm .
_ ieee trans .
inform . theory _
, 130 ( 2):0 260269 , april 1967 .
gerhard winkler .
_ image analysis , random fields and markov chain monte carlo methods _ , volume 27 of _ applications of mathematics ( new york)_. springer - verlag , berlin , second edition , 2003 .
isbn 3 - 540 - 44213 - 8 . a mathematical introduction , with 1 cd - rom ( windows ) ,
stochastic modelling and applied probability . | motivated by the continuing interest in discrete time hidden markov models ( hmms ) , this paper reexamines these models using a risk - based approach . simple modifications of the classical optimization criteria for hidden path inference lead to a new class of hidden path estimators .
the estimators are efficiently computed in the usual forward - backward manner and a corresponding dynamic programming algorithm is also presented .
a particularly interesting subclass of such alignments are sandwiched between the most common _ maximum a posteriori _ ( map ) , or viterbi , path estimator and the minimum error , or _
pointwise maximum a posteriori _ ( pmap ) , estimator .
similar to previous work , the new class is parameterized by a small number of tunable parameters . unlike their previously proposed relatives , the new parameters and class are more explicit and have clear interpretations , and bypass the issue of numerical scaling , which can be particularly valuable for applications .
risk , hmm , hybrid , interpolation , map sequence , viterbi algorithm , symbol - by - symbol , posterior decoding |
since the birth of modern cosmology the theoretical framework used to describe the evolution of the observed universe has been based on the einstein equations applied to a perfectly homogeneous and isotropic model filled with various types of fluid components .
it is universally believed , since the universe on large scales looks almost homogeneous , that the assumption of homogeneity is a good zeroth order approximation for describing our universe , and on top of this one usually adds linear fluctuations ( with initial amplitude of order @xmath0 , as measured by cmb experiments ) in order to describe the growth of density perturbations and the formation of structures .
some steps beyond this paradigm have been made for describing late time behaviour of small scale density perturbations ( which at late times become non - linear ) and more recently in the second order relativistic perturbation theory to catch new effects ( as non - gaussianities in the spectrum of primordial perturbations ) .
however , few efforts have been made to understand how reliable is the assumption of homogeneity .
of course there are corrections to this description due to perturbations : the nonlinearity of einstein equations implies that the presence of perturbations modifies the background .
the usual way of thinking implies that these corrections should be small : since they are nonlinear , the effect should appear as a second order effect , so they should be of order @xmath1 .
however this reasoning is too naive , because it does not take into account the fact that perturbations grow during the matter era . also , in recent years the exploration of the universe at redshifts of order unity has provided information about the time evolution of the expansion rate of the universe .
observations indicate a very unexpected fact : the universe is presently undergoing a phase of accelerated expansion @xcite .
the accelerated expansion is usually interpreted as evidence for a negative pressure component to the mass - energy density of this homogeneous model universe .
the nature of this `` dark energy '' fluid is a huge mystery .
the goal of this paper is to show that this interpretation can not be trusted , since the approximation of perfect homogeneous background becomes inadequate at late times due to the growth of perturbations and it fails at sufficiently late times . in other words the average description of our universe using friedmann equation at small redshift is likely to be incorrect , due to the effect of the inhomogeneities .
it is striking the fact that the friedmann equation fails around the same epoch at which the acceleration is observed .
so a universe filled only with matter , may not necessarily be in contrast with a late time acceleration : it might just be due to the unjustified use of the friedmann equation .
this would completely answer to the coincidence problem : at some time around structure formation the fluctuations become big enough to influence the background , even irrespectively of the initial amplitude of the fluctuations .
the fact that we are pointing out here is entirely due to subhorizon perturbations , and so it is completely different from what has been suggested in @xcite . in that work
it has been suggested that a large variance due to superhorizon modes can have an important effect on the observed expansion : if we live in a large void we might experience an expansion very different from the usual frw .
we analyze here instead the mean effect of inhomogeities on the expansion rate at perturbative level ( using the results of @xcite ) and estimate the mean correction which is unavoidably present in any region of the universe .
this may be interpreted as the contribution of inhomogeneous gravitational fields to the energy content of the universe .
it is in fact a pure nonlinear effect , due to the fact that gravitational fields gravitate .. we will analyze the contribution of the subhorizon modes , ignoring the influence of the unknown superhorizon modes .
we stress so that there are no free parameters here and in principle everything is calculable and predictable . to be able to catch this nonlinear effect ,
the first non - trivial order is second order . at this level
, we found that the mean contribution of the second order is @xmath2 .
it can also be shown that this contribution may act as an effective negative pressure component .
this gives rise to a first crucial observation : the effect is five orders of magnitude bigger than what is naively expected , and this is due to the growth of perturbations after matter - radiation equality .
we point out here that although this contribution may still appear small , one has to consider the full perturbative series . by a simple order of magnitude estimate
we show that in most of the history of the universe , the magnitude of higher orders is decreasing , and so the perturbative approach is likely to give a good description
. however , the main result of this paper is that the expansion parameter becomes of order @xmath3 at redshifts of order @xmath4 , and not @xmath0 as one might naively think .
this tells us that all the terms in the perturbative series are going to be of the same order at late times .
there is some uncertainty on what is exactly the redshift at which the expansion parameter becomes @xmath3 , but it is certainly close to present time . as we said , at the very same redshifts the universe
is observed to accelerate .
we speculate that the acceleration might just be the result of the fact that the series diverges , and that the present @xmath5 concordance model might just be regarded as an effective theory in which all the terms of the series have been resummed , giving rise to an effective negative pressure component .
we base our analysis on the results of @xcite , and we recall the notations used there . we consider a universe filled only by irrotational dust and choose the coordinates of an observer at rest with the dust ( _ i.e. , _ comoving coordinates ) , and with the same time coordinate for every point of the spatial hypersurface ( _ i.e. , _ synchronous coordinates ) .
this system of coordinates can be chosen if the universe is filled by a single pressureless component .
our approach is to start with a flat matter dominated universe ( so with an initial matter energy density fraction @xmath6 , where @xmath7 is the matter energy density , @xmath8 is the critical energy density , @xmath9 is the newton constant and @xmath10 is the hubble parameter ) , and compute perturbatively the correction .
however we are speculating that the correction itself drives the critical density @xmath8 away from @xmath7 due to the fact that @xmath10 does not evolve as in a matter dominated homogeneous universe , at late times .
so , when using the expression of the matter power spectrum ( shape factor and normalization ) for estimating the effect , we will not use as a present value @xmath11 we always mean present time . ] , but a smaller value , as measured by observations on the turnover point of the matter power spectrum . on the other hand , when solving einstein equations and getting the time evolution of the growth function for density perturbations we assume , in a perturbatively consistent way , matter domination , and this will be valid until the breakdown of the approximation ( that is @xmath12 of order 1 ) .
we will call @xmath13 the ( conformal ) time in this gauge , and @xmath14 the spatial coordinates , so that the metric has the form @xmath15 \label{metric}\ ] ] a prime will denote derivatives with respect to conformal time .
we also introduce a proper time by @xmath16 .
the metric will be expanded perturbatively ( @xcite ) expressing all the metric variables as a function of a single initial condition , given by the function @xmath17 , which is constant in time and related to @xmath18 , the first - order density perturbation , by the poisson equation , @xmath19 we assume initial conditions as given by inflation , neglecting initial vector and tensor modes ( @xcite )
. then we will be interested in the local expansion rate @xmath20 as seen by a comoving observer .
we are going to average this quantity over a finite region of size @xmath21 at fixed time @xmath13 with the following prescription : @xmath22 after having computed @xmath23 , we compare this with the naive expectation ( @xmath24 ) that an observer would have measuring the average energy density @xmath25 and applying the homogeneous friedmann equations .
so what we want to estimate is the quantity : @xmath26 also , we stress that here we are interested just in the mean corrections , which means that these are effects that are present in any realization of the region of radius @xmath21 .
so this means that the extension of the region @xmath21 is irrelevant for this discussion .
as we found in @xcite the mean correction is due to modes which are subhorizon , and in fact the main contribution comes from wavelenghts of @xmath27 mpc , and it is insensitive to the infrared and to the ultraviolet .
these variances around this mean values are also interesting , as they seem to depend on superhorizon modes ( see @xcite,@xcite ) , but we do not discuss them here and simply assume they are negligible ( this is certainly true for example if fluctuations with wavelenght bigger than the present hubble radius are negligible ) .
the result that we found consistently at second order in @xcite was : @xmath28 \ , , \label{risultatowithbt}\end{aligned}\ ] ] where @xmath29 .
let us comment on the various terms that are present here .
first , note the absence of terms that go as @xmath30 , without gradients . then , as long as we are interested only in the statistical mean , many terms are irrelevant .
indeed the first order term has obviously zero mean , and the terms in the last line are small if we perform the spatial averages on sufficiently large scales ( for example at horizon scales , they are completely irrelevant ) .
so , only two types of terms are potentially important : the terms with two spatial derivatives @xmath31 , and the terms with four spatial derivatives . as already noted in @xcite
, a term with four derivatives would give a correction of order 1 to the expansion at present times , already at second order .
in fact , as we are going to show again , the presence of gradients boosts the corrections : the more gradients we have , the bigger is the correction .
however , an important property to keep in mind is that the mean of a total spatial gradient is zero ( this can be easily verified going to fourier space ( following @xcite ) ) in other words , as long as we are interested only in the mean , we are allowed to derive by parts our expression .
so , doing that , we immediately see that the four derivative terms gives zero , and that the result can be written as : @xmath32 now let us comment on the two derivative term . symbolically we will call it @xmath33 term .
this contribution was found to be of order @xmath0 .
we stress now that this is _ not _ a small correction , in the sense that naively one could expect a second order term to be of order @xmath1 .
this is the crucial point : the presence of the two powers of @xmath34 boosts the contribution of five orders of magnitude approximatively at the present epoch .
the reason for this is that the scales @xmath35 which have had more time to grow are scales much smaller than the horizon , and so a factor @xmath36 gives a big boost .
we are going to show in detail why this happens , and show that this signals the breakdown of the perturbative expansion .
in fact , extending the estimates to analoguous terms that will appear at higher order in perturbation theory , we find that the expansion parameter is of order 1 at late times .
let us have a look at higher order terms .
of course , being interested only in the mean , and assuming the variable @xmath37 gaussianly distributed , only the even orders of perturbation are relevant .
so the next relevant order is the fourth order . for each order
the maximum number of spatial derivatives grows by two . as we said the more derivatives we have the bigger is the contribution to the mean .
so at fourth order the highest derivative term is @xmath38 . as we will show
, it is possible to see that such a term again yields a order @xmath3 possible correction . at this point
one might be tempted to speculate that at fourth order this may not be zero ( as in @xcite ) . however , the fact that the highest derivative term is a total spatial gradient at second order may not be accidental , and one might expect that at any order in perturbation theory the term with the highest number of derivatives always appear as a spatial gradient .
so , in order to be conservative , we will not rely on this speculation , and assume that all the terms with the highest number of derivatives are zero at any order .
also , as we discuss in section [ newton ] , the highest derivative terms are the newtonian terms and it is expected to get zero mean corrections at the newtonian level .
now , the next term at fourth order is @xmath39 , and by analogy with the second order term @xmath40 , we expect it to be nonzero .
so , generically at order @xmath41 , we assume terms like @xmath42 to be zero , while we keep terms like @xmath43 . for comparison with @xcite
, we stress again that here we are not relying on effects due to uv divergencies ( that is to non linear structures ) , since our considerations involve effects due to scales which are still linear today .
and we do not rely on newtonian terms , but our effect is purely relativistic ( see section [ newton ] ) .
so let us analyze the magnitude of the @xmath33 term . assuming the usual flat spectrum for @xmath37 , its value is given by ( @xcite ) : @xmath44 where @xmath45 is the amplitude of the spectrum of @xmath37 and it is given by @xmath46 , where we took into account of the present value of @xmath47 in the normalization ( @xcite ) .
then , @xmath48 is the bardeen , bond , kaiser , szalay ( bbks ) transfer function @xcite : @xmath49^{-\frac{1}{4 } } .\end{aligned}\ ] ] which describes the stagnation effect of the small scale perturbations that entered the horizon during the radiation era .
of course at small @xmath50 , @xmath51 , while at large @xmath50 , @xmath52 .
the integral receives the biggest contribution from the scale at which the integrand has a maximum , which corresponds to @xmath27 mpc . here
@xmath53 is the shape factor , defined for a flat universe in terms of the baryon fraction @xmath54 and the total value of @xmath55 as @xmath56 .
@xmath53 is sensitive especially to the value of @xmath57 and it is decreasing as @xmath57 increases .
finally , the integral in eq.([k^2 ] ) has the numerical value of @xmath58 . for comparison we may estimate a @xmath59 term (
although as we said it is not present in the result given by eq.([resultmean ] ) ) : @xmath60 such a result would imply a big correction already at second order .
in fact the factor @xmath61 is as large as ten or eleven orders of magnitude ( depending on the value of cosmological parameters ) . in addition , the integral is even logaritmically divergent in the ultraviolet , and so gets its main contribution from the small scales . even putting a cutoff at the nonlinear scales , the integral results in a @xmath62 factor . and
finally this correction increases with time twice as fast as the @xmath33 term , becoming big at late times .
now let us go to estimate the higher orders .
as we said , going to fourth order , we safely assume that the leading term is @xmath63 .
this is estimated as follows : @xmath64 here the term @xmath65 stands for one or more integrals in @xmath66 .
depending on what is the precise structure of the contraction of indices in the @xmath63 term , it may contain also ultraviolet divergent integrals which would give a larger contribution than the integral that we found at second order , and might be sensitive to the small scales .
however , in order to make a conservative estimate we do not rely on such speculations and we just assume the minimal value @xmath67 at the generic order @xmath41 ( where @xmath41 is even ) so we have : @xmath68 where @xmath69 is the expansion parameter : @xmath70
the crucial fact is that this expansion parameter becomes of order 1 for small @xmath12 ( the value depends on uncertainties on the cosmological parameters , mainly @xmath47 through @xmath53 ) : @xmath71 while already for @xmath72 , the terms of the series start decreasing , so one may assume that the series converges .
the expansion parameter is plotted in fig.[figuracdm ] , for different values of the cosmological parameters . in a similar way
one can also see that the highest @xmath34 terms ( that we set to zero ) are of the same order of @xmath59 at small redshifts .
now , we may speculate that the fact that the expansion parameter is of order @xmath3 just around @xmath73 is related to the appearance of acceleration in the observed universe .
if this is the case , this fully solves the coincidence problem .
even more : one can see that the time at which @xmath69 is of order @xmath3 depends on the amplitude of the initial fluctuations , so it may appear a coincidence that the initial value of @xmath0 is such that the expansion breaks down right today .
however for any initial amplitude the time at which the series breaks down is always around the time of structure formation , and since we can only live around that time , there is no coincidence at all . at this point , so , let us assume that the universe starts accelerating as a result of the divergence of the series .
this , in turn , stops the growth of the density perturbations , and so the expansion parameter itself is prevented to grow and it stabilizes around the value that it takes when the effect starts to be important ( as one can see in fig .
( [ figuralcdm ] ) ) .
it has been pointed out in @xcite and @xcite that a deceleration parameter defined in terms of a local expansion rate can never become negative in absence of vorticity and in absence of negative pressure fluids .
however this result does not apply here , since the expansion we are considering is not @xmath74 , but is @xmath23 . when trying to derive the same result to this quantity , one can see that new negative terms appear due to the time derivative of the @xmath75 , which is present in the definition of the spatial average . in order to show this explicitly
we need to introduce the expansion tensor , which reads @xmath76 this can be also decomposed into its trace ( which is @xmath77 ) and a traceless tensor ( the shear tensor ) : @xmath78 once we have defined the expansion rate as @xmath79 , its deceleration parameter is given by : @xmath80 the raychaudhuri equation gives : @xmath81 where @xmath82 . and where we assumed zero pressure and zero vorticity .
another useful relation is obtained taking the determinant of eqs .
( [ sigma ] ) and ( [ thetaij ] ) , which gives @xmath83 . using these relations it is straightforward to obtain the following : @xmath84 which is an exact equation , valid in general .
the first line corresponds to what was derived in @xcite , while the second line arises due to the presence of @xmath75 in the averages .
this clearly shows that there are negative contributions and so when the corrections overcome the background value , the deceleration parameter can change sign . _ the expansion parameter @xmath69 as a function of the redshift , for different values of the parameters ( @xmath85 , as taken from wmap only(@xcite ) ) .
the grey dashed line corresponds to the central value , while the red solid lines to @xmath86 ranges .
here we used the growth factor as in matter domination . for comparison
the green dotted line corresponds to @xmath87 .
the perturbative series breaks down at small @xmath12 when @xmath69 becomes of order 1 ( we also extended the plot up to negative @xmath12 in order to show that at some point the series badly diverges unavoidably.)_,scaledwidth=45.0% ] _ the expansion parameter @xmath69 as a function of the redshift , for different values of the parameters ( @xmath85 , as taken from wmap only(@xcite ) ) .
the grey dashed line corresponds to the central value , while the red solid lines to @xmath86 ranges .
here we used the growth factor for density perturbations of a @xmath5 model @xcite , in order to show that @xmath69 is prevented to grow and it stabilizes at late times , if the universe starts accelerating._,scaledwidth=45.0% ]
it is very interesting to note that ( see @xcite ) the highest @xmath34 terms ( that we neglected ) are also the terms with the highest power of @xmath88 ( we are using here the simple growth factor and background evolution of a pure matter dominated universe ) .
now , in an expansion in powers of the speed of light @xmath89 , each power of @xmath13 is accompanied by a power of @xmath89 .
so the highest @xmath34 terms are the newtonian terms ( @xcite ) .
we stress instead that the terms that we considered here , instead , are the post - newtonian ones .
the following example is quite illuminating in order to understand why the highest @xmath34 terms should give zero effect .
consider a homogeneous universe filled with many spherical inhomogeneous regions : outside these regions matter is uniformly distributed with energy density @xmath90 , and inside these regions matter density is inhomogeneously distributed ( but still maintaining the spherical symmetry and in such a way that the total amount of matter inside a sphere is equal to the amount of a same - sized sphere with homogeneous matter density @xmath90 ) .
now , in newtonian gravity , we can use the superposition principle and consider separately the effect of a single sphere . by gauss theorem
the only thing that matters is the total amount of matter inside the sphere , so the average expansion rate over a region that contains spheres is the same one of a perfectly homogenous universe with matter density @xmath90 .
the only way in order to guarantee that the effect is zero in this example is that the newtonian terms be total spatial gradients .
in fact in this case one can rewrite the spatial average as a flux of a vector field over a surface . taking the surface in such a way that it does not intersect any sphere ,
this gives zero because the vector field in the homogeneous region is zero .
this shows that newtonian terms have to be surface terms , and in order to have an effect one has to consider the purely relativistic terms ( and the post - newtonian approximation should be accurate enough ) .
so , what we are claiming is that the series ( in an expansion in the smallness of the metric elements ) of all the post - newtonian terms is diverging , and we may speculate that the resummation of all these terms may give rise to an accelerating universe . or in other words it may give rise to an effective theory in which the friedmann equations are still valid , but with an effective @xmath91 term . in order to prove this kind of `` phase transition '' at redshift of order 1 , so , one needs a treatment of the postnewtonian terms , which does not assume that the metric elements are small .
we have shown that , even being very conservative , the contribution of the inhomogeneities to the average expansion rate becomes perturbatively out of control at low redshifts .
this leads to the conclusion that the usual friedmann equation can not be trusted at late time .
so , instead of using the naive homogeneous model to fit the data , one should to take a nonperturbative approach to einstein equations in a inhomogeneous universe , before talking of dark energy .
we suggest that the failure of this equation might be the reason why we need to introduce mysterious components in order to fit the data .
we speculate that the resummation of the terms of the diverging series ( which may act as a negative pressure component ) might lead to an effective friedmann equation in which the series is resummed , giving an effective `` dark energy '' .
once this effect is taken into account , the perturbations themselves stop growing ( as happens in a @xmath5 model ) , and we speculate that the effect itself may be stabilized , reaching a fixed point , in which it might even be possible to predict the effective @xmath92 .
we would like to thank rocky kolb , sabino matarrese and toni riotto for previous work and discussions on the subject .
it is a pleasure to thank also niayesh afshordi , thorsten battefeld , robert brandenberger , paul chouha , fabrizio di marco , balaji katlai , natalia shuhmaher and alberto vallinotto for useful comments and discussions .
99 for a review of the observational evidence for the acceleration of the expansion of the universe , see , e.g. , t. padmanabhan , phys .
rept . * 380 * , 235 ( 2003 ) [ arxiv : hep - th/0212290 ] ; t. padmanabhan , arxiv : astro - ph/0411044 ; s. hannestad and e. mortsell , jcap * 0409 * , 001 ( 2004 ) ; a. melchiorri , l. mersini - houghton , c. j. odman and m. trodden , phys .
d * 68 * , 043509 ( 2003 ) [ arxiv : astro - ph/0211522 ] .
a. melchiorri , l. mersini - houghton , c. j. odman and m. trodden , phys .
d * 68 * , 043509 ( 2003 ) ; p. j. e. peebles and b. ratra , rev .
phys . * 75 * , 559 ( 2003 ) ; r. bean and a. melchiorri , phys .
d * 65 * , 041302 ( 2002 ) .
l. hui and u. seljak , in _ proceedings of the 173rd iau symposium _ , edited by c. s. kochanek and j. n. hewitt ( kluwer academic publishers , dordrecht , 1995 ) ; u. seljak and l. hui , in _ proceedings of clusters , lensing and the future of the universe _ ( asp conference series vol .
88 , v. trimble and a. reisenegger , eds . , 1995 ) , p 267 . | it is widely believed that the assumption of homogeneity is a good zero_th _
order approximation for the expansion of our universe .
we analyze the correction due to subhorizon inhomogeneous gravitational fields . while at early times this contribution ( which may act as a negative pressure component ) is perturbatively subdominant
, we show that the perturbative series is likely to diverge at redshift of order 1 , due to the growth of perturbations .
so , the homogeneous friedmann equation can not be trusted at late times .
we suggest that the puzzling observations of a present acceleration of the universe , may just be due to the unjustified use of the friedmann equation and not to the presence of a dark energy component .
this would completely solve the coincidence problem . |
statistical inference is the task of computing marginals ( or expectation values ) of complex multi - variate distributions . _
belief propagation _ ( bp ) is a generic method for accomplishing this task quickly but approximately , when the multivariate distribution factorizes according to a sparse graphical structure .
the advent of sparse graph codes and iterative bp decoding @xcite has naturally made decoding become an important case of this general problem .
the present paper builds on this connection by ` importing ' an algorithm that has been recently developed in the context of approximate counting and inference @xcite .
we will refer to the new algorithm as _ tree pruning _ ( tp ) _
decoding_. for a number of reasons the application of this method to decoding is non - trivial .
however , it is an interesting approach for the three following reasons . @xmath0
it provides a sequence of decoding schemes that interpolates continuously between bp and the optimal maximum a posteriori ( map ) decoding .
@xmath1 at each level of this sequence , the effect of loops of increasing length is taken into account .
@xmath2 we expect that an appropriate truncation of this sequence might yield a polynomial algorithm for map decoding on general graphs of bounded degree , for low enough noise levels .
preliminary numerical results are encouraging .
as for bp decoding , tp decoding aims at estimating the a posteriori marginal probabilities of the codeword bits .
unhappily , the relation between bp estimates and the actual marginals is in general poorly understood . in the case of random low - density parity - check ( ldpc )
codes and communication over memoryless channels , density evolution allows to show that , at small enough noise level , the bp bit error probability becomes arbitrarily small if the blocklength is large enough .
this implies that the distance between bp estimates and the actual marginals vanishes as well .
this result is not completely satisfactory , in that it relies in a crucial way on the locally tree - like structure of sparse random graphs .
this property does not hold for structured graphs , and , even for large graphs , it kicks in only at very large blocklengths .
in contrast to this , the algorithm considered in this paper accounts systematically for short loops .
it should therefore produce better performances , in particular in the error floor regime since this is dominated by small error events @xcite .
a convenient way of understanding the difference between bp and map decoding makes use of the so - called computation tree .
consider a code described by a factor graph @xmath3 whereby @xmath4 represents the variable nodes , @xmath5 the factor nodes , and @xmath6 the edges .
let @xmath7 , then the corresponding computation tree denoted by @xmath8 is the tree of non - reversing walks in @xmath9 that start at @xmath10 .
this gives a graph ( tree ) structure in a natural way : two nodes are neighbors if one is reached from the other adding a step .
bp uses the marginal at the root of @xmath8 as an estimate for the marginal distribution at @xmath10 on the original graph @xmath9 .
if @xmath9 contains short loops in the neighborhood of @xmath10 , the computation tree differs from @xmath9 in a neighborhood of the root and , as a consequence , the bp estimate can differ vastly from the actual marginal .
weitz @xcite made the surprising remark that there exists a simple way of pruning the computation tree ( and fixing some of its variables ) in such a way that the resulting root marginal coincides with the marginal on @xmath9 .
unhappily the size of the pruned tree , which we call the _ self - avoiding walk tree _ and denote by @xmath11 , is exponential in the size of the original graph @xmath9 . nevertheless , the tree can be truncated thus yielding a convergent sequence of approximations for the marginal at @xmath10 .
the complexity of the resulting algorithm is linear in the size of the truncated tree .
its efficiency depends on how sensitive is the root marginal to the truncation depth . applying this approach to decoding linear codes
poses several challenges : 1 .
weitz s construction is valid only for markov random fields ( mrfs ) with pairwise interactions and binary variables .
the decoding problem does not fit this framework .
2 . the original justification for truncating the self - avoiding walk tree followed the so - called ` strong spatial mixing ' or ` uniqueness ' condition .
this amounts to saying that the conditional marginal at the root given the variables at depth @xmath12 , depends weakly on the values of the latter .
this is ( most of the times ) false in decoding . for a ` good ' code ,
the value of bit @xmath10 in a codeword is completely determined by the values of bits outside a finite neighborhood around @xmath10 .
3 . even worse , we found in numerical simulations that the original truncation procedure performs poorly in decoding
. the self - avoiding walk tree construction has already motivated several applications and generalizations in the past few months .
jung and shah @xcite discussed its relation with bp , and proposed a distributed implementation .
mossel and sly @xcite used it to estimate mixing times of monte carlo markov chain algorithms . finally , and most relevant to the problems listed above , nair and tetali @xcite proposed a generalization to non - binary variables and multi - variable interactions .
while this generalizations does in principle apply to decoding , its complexity grows polynomially in the tree size .
this makes it somewhat unpractical in the present context . in this paper
we report progress on the three points above .
specifically , in section [ sec : decodingtree ] we use duality to rephrase decoding in terms of a generalized binary markov random field .
we then show how to generalize the self - avoiding walk tree construction to this context . in section [ sec :
truncation ] we discuss the problems arising from the original truncation procedure , and describe two procedure that show better performances .
numerical simulations are presented in section [ sec : simulations ] .
finally , one of the most interesting perspectives is to use tp as a tool for analyzing bp and , in particular , comparing it with map decoding .
some preliminary results in this direction are discussed in section [ sec : theoretical ] .
we should stress that a good part of our simulations concerns the binary erasure channel ( bec ) . from a practical point of view , tp decoding is not an appealing algorithm in this case .
in fact , map decoding can be implemented in polynomial time through , for instance , gaussian elimination .
the erasure channel is nevertheless a good starting point for several reasons . @xmath0
comparison with map decoding is accessible .
@xmath1 we can find a particularly simple truncation scheme in the erasure case .
@xmath2 some subtle numerical issues that exist for general channels disappear for the bec .
throughout this paper we consider binary linear codes of blocklength @xmath13 used over a binary - input memoryless channel .
let bm@xmath14 , where @xmath15 is a noise parameter , denote a generic channel .
assume that @xmath16 is the output alphabet and let @xmath17 denote its transition probability . with a slight abuse of terminology we shall identify a code with a particular parity - check matrix @xmath18 that represents it , @xmath19 therefore , the code is further identified with a tanner graph @xmath20
whose adjacency matrix is the parity - check matrix @xmath18 .
we will denote by @xmath21 the neighborhood of function ( check ) node @xmath22 , and write @xmath23 .
analogously , @xmath24 indicates the neighborhood of the variable node @xmath10 . the conditional distribution for the channel output @xmath25 given the input @xmath26 factorizes according to the graph @xmath9 ( also called factor graph ) .
it follows immediately from bayes rule that @xmath27 , where @xmath28 we denote by @xmath29 the marginal distribution at bit @xmath10 . _ symbol map decoding _ amounts to the following prescription , @xmath30 both bp and tp decoders have the same structure , whereby the marginal @xmath31 is replaced by its approximation , respectively @xmath32 or @xmath33 .
we call a _ generalized markov random field _ ( gmrf ) over the finite alphabet @xmath34 a couple @xmath35 , where @xmath36 is an ordinary graph over vertex set @xmath37 , and edge set @xmath38 .
further @xmath39 is a set of weights indexed by edges and vertices in @xmath40 , @xmath41 , @xmath42 .
notice that , unlike for ordinary mrfs , the edge weights in generalized mrfs are required to be non - negative . given a subset @xmath43 , the _ marginal _ of the gmrf @xmath44 on @xmath45 , is defined as the function @xmath46 , with entries @xmath47 when @xmath48 , we shall omit the subscript and call @xmath49 the _ weight _ of configuration @xmath26 . more generally , the _ expectation _ of a function @xmath50 can be defined as @xmath51 notice that @xmath52 is not ( and in general can not be ) normalized . in the sequel , whenever the relevant mrf has non - negative weights and is normalizable , we shall use words ` expectation ' and ` marginal ' in the usual ( normalized ) sense .
duality can be used to reformulate decoding ( in particular , the posterior marginals @xmath53 ) in terms of a gmrf .
more precisely , given a code with tanner graph @xmath54 , we define a gmrf on graph @xmath55 where @xmath56 and @xmath57 , proceeding as follows .
we let @xmath58 and associate variables @xmath59 to @xmath7 and @xmath60 to @xmath61 .
we then introduce the weights @xmath62 although next statement follows from general duality theory , it is convenient to spell it out explicitly .
the marginals of the a posteriori distribution defined in eq .
( [ eq : apdistr ] ) are proportional to the ones of the gmrf defined in eq .
( [ eq : gmrfdec1 ] ) and eq .
( [ eq : gmrfdec2 ] ) .
more precisely , we get @xmath63 $ ] .
it is immediate to prove a stronger result , namely that the distribution @xmath64 is proportional to @xmath65 , where @xmath66 is defined using eq .
( [ eq : gmrfdec1 ] ) and eq .
( [ eq : gmrfdec2 ] ) .
we have @xmath67 which is proportional to the right - hand side of eq .
( [ eq : apdistr ] ) .
this result , which derives from @xcite and @xcite , motivates us to extend weitz s construction to gmrfs . ) ) , while processing all graph vertices in a similar way .
] this is the object of the next section .
assume we are given a graph @xmath55 and a node @xmath68 .
we have already described the computation tree rooted at @xmath10 , which we denote by @xmath8 .
an ` extended self - avoiding walk ' ( saw ) on a graph @xmath55 , starting at @xmath68 is a non - reversing walk that never visits twice the same vertex , except , possibly , for its end - point .
the ` self - avoiding walk tree ' rooted at @xmath69 is the tree of all extended self - avoiding walks on @xmath40 starting at @xmath10 .
it is straightforward to see that @xmath11 is in fact a finite sub - tree of @xmath8 .
its size is bounded by @xmath70 , where @xmath71 is the maximum degree of @xmath40 , and @xmath72 the node number .
( 200,100 ) ( 0,0).,title="fig : " ] ( 210,250 ) ( 0,0 ) for the tanner graph of figure
[ fig : tanner ] , rooted at variable node @xmath73 . in this picture , each node of the self - avoiding tree is labeled by its projection onto @xmath37 . at `
terminated ' nodes we marked the value that the variable is forced to take .
, title="fig : " ] as an example , figure [ fig : sawtree ] shows a saw tree for the small graph @xmath40 depicted in figure [ fig : tanner ] .
( in this case , @xmath40 is the tanner graph of a repetition code of length @xmath74 . )
if we denote by @xmath75 the vertex set of @xmath11 , there exists a natural projection @xmath76 that preserves edges .
formally , @xmath77 maps a self - avoiding walk to its end - point .
notice that @xmath11 has two types of leaf nodes : @xmath0 nodes that are leaves in the original graph @xmath40 .
@xmath1 nodes that are not leaves in @xmath40 but corresponds to extended
self - avoiding walks that can not be further continued .
the latter case arises when the endpoint of the self - avoiding walk has already been visited ( i.e. , when a loop is closed ) .
we shall refer to nodes of the second type as _ terminated _ nodes .
indeed , the self - avoiding walk tree @xmath11 can be obtained from @xmath8 by the following _ termination _ procedure .
imagine descending @xmath8 along one of its branches .
when the same projection is encountered for the second time , terminate the branch .
formally , this means eliminating all the descendants of @xmath78 whenever @xmath79 for some ancestor @xmath80 of @xmath78 . given a gmrf @xmath35 , we can define a gmrf on @xmath11 in the usual way .
namely , to any edge @xmath81 , we associate a weight coinciding with the one of the corresponding edge in @xmath40 : @xmath82 . the analogous definition is repeated for any non - terminated node : @xmath83 .
finally , the choice of weight on terminated nodes makes use of the hypothesis that @xmath84 .
assume that the edges of @xmath40 are labeled using a given order , e.g. , a lexicographic order .
let @xmath78 be a terminated node with @xmath85 .
then the self - avoiding walk corresponding to @xmath78 contains a loop that starts and ends at @xmath86 .
we let @xmath87 ( respectively , @xmath88 ) if this loop quits @xmath86 along an edge of higher ( respectively , lower ) order than the one along which it enters @xmath86 .
the relevance of this construction is due to weitz who considered the case of _ permissive _ binary mrfs . by this
we mean that @xmath89 , @xmath90 , and , for any @xmath91 , there exists @xmath92 , such that @xmath93 for any @xmath94 with @xmath95 and @xmath96 .
( the latter is referred to as the ` permissivity ' condition . )
[ propo : weitz ] given a binary mrf @xmath35 , the marginal of @xmath97 with respect to @xmath35 is proportional to the root marginal on @xmath11 .
the problem with non - permissive mrfs and , _ a fortiori _ , with generalized mrfs , is that the tree model @xmath11 may not admit any assignment of the variables such that all the weights @xmath98 , @xmath99 are non - negative . as a consequence
the mrf on @xmath11 does not define a probability distribution and this invalidates the derivation in @xcite or @xcite .
even worse , the procedure used in these papers was based in keeping track of ratios among marginals , of the form @xmath100 .
when the mrf does not define a distribution , ill - defined ratios such as @xmath101 can appear .
let us stress that this problem is largely due to the ` termination ' procedure described above .
this in fact constrains the set of assignments with non - vanishing weight to be compatible with the values assigned at terminated nodes . in order to apply the self - avoiding walk construction to gmrfs
, we need to modify it in the two following ways .
@xmath0 we add further structure to @xmath11 . for any @xmath102 , let @xmath103 be the set of its _ children _
( i.e. , the set of extended self - avoiding walks that are obtained by adding one step to @xmath78 ) .
then we partition @xmath104 as follows .
let @xmath105 be two children of @xmath78 , and write them as @xmath106 , @xmath107 .
further , let @xmath108 .
then we write @xmath109 if there exists an extended self - avoiding walk of the form @xmath110 . here
we are regarding @xmath78 , @xmath111 as walks on @xmath40 ( i.e. , sequences of vertices ) and we use @xmath112 to denote the concatenation of walks .
it is not difficult to verify that @xmath113 is an equivalence relation .
the partition @xmath114 is defined to be the partition in equivalence classes under this relation .
@xmath1 we define the _ generalized root marginal _ of @xmath11 through a recursive procedure that makes it always well - defined .
first notice that , if @xmath40 is a tree rooted at @xmath10 , then the marginal at @xmath10 can be computed by a standard message passing ( dynamic programming ) procedure , starting from the leaves and moving up to the root .
the update rules are , for @xmath115 , @xmath116 where edges are understood to be directed towards the root . the marginal at the root
is obtained by evaluating the right hand side of eq .
( [ eq : upnodes ] ) with @xmath117 .
the generalized root marginal is defined by the same procedure but changing eq .
( [ eq : upnodes ] ) as follows .
given the partition @xmath104 described above , we let @xmath118 where we define @xmath119 through a _ concatenation _
let @xmath120 be the set of messages @xmath121 ordered according to the order of edges @xmath122 in @xmath40 .
then we let @xmath123 the reason for calling this a ` concatenation ' follows from the remark that , with the notations above , we have @xmath124 , @xmath125 , etc .
we refer to the discussion ( and proof ) below for a justification of this claim . as a consequence , the procedure in eq .
( [ eq : concatenation ] ) can be described as follows : write the components of @xmath126 in sequence , and eliminate repeated entries . with this groundwork
, we obtain the following generalization of weitz s result .
[ propo : gmrf ] given a gmrf @xmath35 , the marginal at @xmath68 with respect to @xmath35 is equal to the generalized root marginal on @xmath11 .
the proof is very similar to weitz s original proof in @xcite ; the difference is that special care must be paid to avoid ill - defined expressions .
the argument consists in progressively simplifying the graph @xmath40 ( rooted at @xmath10 ) until @xmath11 is obtained .
we shall represent these simplifications graphically .
consider the first step , corresponding to eq .
( [ eq : upnodes1 ] ) , with @xmath117 . the partition of @xmath103 in @xmath127 , @xmath128 , corresponds to a partition of of the subgraph @xmath129 ( obtained by eliminating from @xmath40 , @xmath10 as well as its adjacent vertices ) into connected components .
this correspondence is depicted below ( whereby gray blobs correspond to connected sub - graphs ) .
( -195,55)@xmath78 ( -95,55)@xmath130 ( -63,55)@xmath131 after factoring out the term @xmath132 , the definition of marginal in eq .
( [ eq : margdef ] ) factorizes naturally on such components , leading to eq .
( [ eq : upnodes1 ] ) consider now one of such components , call it @xmath133 , such as the one depicted below .
the corresponding generalized root marginal is computed using the concatenation rule , specified in eq .
( [ eq : concatenation ] ) .
( -225,55)@xmath78 ( -165,55)@xmath134 ( -130,55)@xmath135 ( -65,53)@xmath134 ( -42,53)@xmath135 ( -67,0)@xmath136 ( -50,0)@xmath137 in order to derive this rule , first consider the graph @xmath138 obtained from @xmath133 by replacing its root @xmath78 by @xmath139 copies @xmath140 , each of degree @xmath141 ( here @xmath142 denotes the degree of vertex @xmath80 ) .
each of the newly introduced vertices is adjacent to one of the edges incident on the root in @xmath133 .
further @xmath140 are labeled according to the ordering ( chosen at the beginning of the reduction procedure ) on the adjacent edges .
these @xmath143 nodes will be referred to as ` split nodes ' in the sequel . from the definition of marginal in eq .
( [ eq : margdef ] ) , and using the notation @xmath144 for the gmrf on @xmath145 , we have @xmath146 for @xmath147 .
this identity is represented as the first equality in the figure above .
next we replace the graph @xmath145 by @xmath143 copies of it , @xmath148 . with a slight abuse of notation
, we re - name @xmath134 the first of the @xmath143 ` split nodes ' in @xmath149 , @xmath135 the second in @xmath150 , and so on .
further we add node weights to the other ` split nodes , ' ( i.e. , the ones that remained un - named ) , either of the form @xmath151 ( forcing @xmath152 to take value @xmath136 ) or of the form @xmath153 ( forcing @xmath152 to take value @xmath137 ) .
more precisely , for any @xmath154 on @xmath155 we force to @xmath136 those split nodes that come before @xmath156 , and to @xmath137 the ones that come after . as a consequence ,
if we use @xmath157 for the gmrf @xmath158 , we have @xmath159 in particular , for any @xmath160 , @xmath161 . as a consequence of this fact and of eq .
( [ eq : splitting ] ) , we get @xmath162 this proves eq .
( [ eq : concatenation ] ) with @xmath163 ( second equality in the last figure above ) .
finally , eq . ( [ eq : upedges ] ) follows by considering the marginal of a node of degree @xmath141 , as @xmath140 in graphs @xmath149 , @xmath164 , and expressing it in terms of the marginal of its only neighbor .
this completes one full step of the procedure that breaks the loops through node @xmath10 . by recursively
repeating the same steps , the graph is completely unfolded giving rise to @xmath11 .
the self - avoiding walk tree @xmath11 appears as a convenient way to organize the calculation of the marginal at @xmath10 in the general case . in the case of permissive mrfs
this calculation coincides with a standard marginal calculation on the tree @xmath11 .
it is instructive to check this explicitly .
proposition [ propo : weitz ] is a special case of proposition [ propo : gmrf ] for permissive mrfs .
first notice that , for permissive mrfs , the self - avoiding walk tree construction yields a mrf on @xmath11 that defines a probability distribution ( non - negative and normalizable ) , whose marginals will be denoted as @xmath165 as well .
we have to prove that , in this case , the generalized root marginal is proportional to the ordinary marginal at the root of @xmath11 .
the crucial remark is that , because of permissivity , the messages are non - negative and , in particular , @xmath166 and @xmath167 . assume , without loss of generality , that @xmath168 .
we define the likelihood ratios on the @xmath11 tree @xmath169 , @xmath170 and @xmath171 .
the ratio @xmath172 is defined analogously in terms of @xmath173 . equation ( [ eq : upedges ] ) then implies @xmath174 eq .
( [ eq : upnodes1 ] ) yields on the other hand @xmath175 finally , using the remark that @xmath176 for @xmath177 , we get from eq .
( [ eq : concatenation ] ) @xmath178 putting the last two equations together @xmath179 it is now easy to check that , eq .
( [ eq : rationodes ] ) and eq .
( [ eq : ratioedges ] ) coincide with the appropriate recursive definition of probability marginal ratios on @xmath11 . proposition [ propo : gmrf ] does not yield an efficient way of computing marginals of gmrf .
the conundrum is that the resulting complexity is linear in the size of @xmath11 which is in turn exponential in the size of the original graph @xmath40 . on the other hand ,
it provides a systematic way to define and study algorithms for computing efficiently such a marginal .
the idea , proposed first in @xcite , is to deform @xmath11 in such a way that its generalized root marginal does not change too much , but computing it is much easier .
bp can be seen as an example of the approach mentioned at the end of the previous section . in this case
@xmath11 is replaced by the first @xmath12 generations of the computation tree , to be denoted by @xmath180 . in this case
the complexity of evaluating the generalized root marginal scales as @xmath12 rather than as @xmath181 .
a different idea is to cut some of the branches of @xmath11 in such a way to reduce drastically its size .
we will call _ truncation _ the procedure of cutting branches of @xmath11 .
it is important to keep in mind that truncation is different from the _ termination _ of branches when a loop is closed in @xmath40 . while termination is completely defined , we are free to define truncation to get as good an algorithm as we want . in the following
we shall define truncation schemes parametrized by an integer @xmath12 , and denoted as @xmath182 .
we will have @xmath183 for @xmath184 , thus recovering the exact marginal by proposition [ propo : gmrf ] . in order for the algorithm to be efficient
, we need to ensure the following constraints .
@xmath182 is ` small enough ' ( as the complexity of computing its generalized root marginal is at most linear in its size ) .
@xmath1 @xmath182 is ` easy to construct . ' for coding applications , this second constraint is somewhat less restrictive because the tree(s ) @xmath182 can be constructed in a preprocessing stage and not recomputed at each use of the code . in order to achieve the second goal
, we must define the partition @xmath185 of children of @xmath78 according to the subtree @xmath182 used in the computation .
consider two children of @xmath78 , which we denote by @xmath105 . in a similar way as for the @xmath186 in the complete tree case
, we write them as @xmath106 , @xmath107 , and define @xmath187 if there exists a descendant @xmath188 of @xmath189 in @xmath182 such that @xmath190 or a descendant @xmath191 of @xmath192 such that @xmath193 . the construction of the partition @xmath194 will be different whether communication takes place over erasure or general channels . the truncation procedure proposed in @xcite
amounts to truncating all branches of @xmath11 at the same depth @xmath12 , unless they are already terminated at smaller depth .
variables at depth @xmath12 ( boundary ) are forced to take arbitrary values .
the rationale for this scheme comes from the ` strong spatial mixing ' property that holds for the system studied in @xcite .
namely , if we denote by @xmath195 the normalized marginal distribution at the root @xmath10 given variable assignment at depth @xmath12 , we have @xmath196 uniformly in the boundary conditions @xmath197 , @xmath198 for some constants @xmath199 , @xmath200 .
it is easy to realize that the condition in eq .
( [ eq : ssm ] ) generically does not hold for ` good ' sparse graph codes .
the reason is that fixing the codeword values at the boundary of a large tree , normally determines their values inside the same tree . in other words
the tp estimates strongly depend on this boundary condition .
one can still hope that some simple boundary condition might yield empirically good estimates .
an appealing choice is to leave ` free ' the nodes at which the tree is truncated .
this means that no node potential is added on these boundary vertices .
we performed numerical simulations with this scheme on the same examples considered in the next section .
the results are rather poor : unless the truncation level @xmath12 is very large ( which is feasible only for small codes in practice ) the bit error rate is typically worse than under bp decoding . for decoding over the bec ,
a simple trick improves remarkably the performances of tp decoding .
first , construct a subtree of @xmath11 of depth at most @xmath12 by truncating at the deepest variable nodes whose depth does not exceed @xmath12 .
the partition @xmath194 is constructed using the equivalence class of the transitive closure of @xmath201 .
then run ordinary bp on this graph , upwards from the leaves towards the root , and determine all messages in this direction .
if a node @xmath78 is decoded in this way , fix it to the corresponding value and further truncate the tree @xmath11 at this node .
the resulting tree @xmath202 is not larger than the one resulting from fixed - depth truncation . for low erasure probabilities
it is in fact much smaller than the latter .
the above trick can not be applied to general bm channels .
we therefore resort to the following two constructions .
@xmath0 construction @xmath203 : define the distance @xmath204 between two variable nodes @xmath10 , @xmath86 to be the number of check nodes encountered along the shortest path from @xmath10 to @xmath86 .
let @xmath205 be the subgraph induced of variable nodes is the one including all those check nodes that only involve variables in @xmath206 .
] by variable nodes whose distance from @xmath10 is at most @xmath12
. then we let @xmath207 be the _ complete _ self - avoiding walk tree for the subgraph @xmath205 .
this corresponds to truncating @xmath11 as soon as the corresponding self - avoiding walk exits @xmath205 .
no forcing self - potential is added on the boundary .
a nice property of this scheme is that it returns the _ a posteriori _ estimate of transmitted bit @xmath208 given the channel outputs within @xmath205 , call it @xmath209 . as a consequence , many reasonable performance measures ( bit error probability , conditional entropy , etc . )
are monotone in @xmath12 @xcite . on the negative side
, the size of the tree @xmath203 grows very rapidly ( doubly exponentially ) with @xmath12 at small @xmath12 .
this prevented us from using @xmath210 .
@xmath1 construction @xmath211 : the tree @xmath203 constructed as in the previous approach is augmented by adding some descendants to those nodes that are terminated in @xmath203 . more precisely , below any such node @xmath78 in the mentioned tree , we add the first @xmath212 generations of the computation tree .
@xmath2 construction @xmath202 : we can implement a finer scheme for the general bm case .
this scheme operates on @xmath202 obtained by truncating all branches of @xmath11 at the same depth @xmath12 .
the description of this method is slightly lengthy .
we omit the details here and choose to present the numerical results in fig .
[ fig : exitcurvesconvc50bawgn ] of section [ sec : simulations ] .
for communication over the bec , the implementation of tp decoding as described in section [ sec : erasuretruncation ] is satisfying .
while it is simple enough for practical purpose , it permits us to depict performance curves that interpolate successfully between bp and map decoding .
the binary erasure channel , which we denote by bec@xmath213 if the erasure probability is @xmath214 , is appealing for a first study for the following reasons .
@xmath0 the accessibility of performance curves under map decoding allows for a careful study of the new algorithm . @xmath1
the tp decoder turns out to be ` robust ' with respect to changes in the truncation method , hence simpler to study . as an example for a generic bm channel
, we shall consider the binary - input additive white gaussian noise channel with standard deviation @xmath215 , which we denote by bawgn(@xmath216 ) .
let us stress that the tp decoder is not ( in general ) symmetric with respect to codewords .
this complicates a little the analysis ( and simulations ) which has to be performed for uniformly random transmitted codewords .
the erasure case is illustrated by three examples : a tail - biting convolutional code , the @xmath217 golay code and a sparse graph code . here
the comparison is done with bp after convergence ( ` infinite ' number of iterations ) and map as implemented through gaussian elimination .
tp decoder permits us to plot a sequence of performance curves , indexed by the truncation parameter @xmath12 . in all of the cases considered ,
tp improves over bp already at small values of @xmath12 . as @xmath12 increases ,
tp eventually comes very close to map .
the gain is particularly clear in codes with many short loops , and at low noise .
this confirms the expectation that , when truncated , tp effectively takes care of small ` pseudocodewords . '
the first example is a memory two and rate @xmath218 convolutional code in tailbiting form with blocklength @xmath219 .
the performance curves are shown in fig .
[ fig : ccm2_100 ] . the tp and
bp decoders are based on a periodic tanner graph associated with the tailbiting code with generator pair @xmath220 . more precisely , they are based on the graph representing the parity - check matrix with circulant horizontal pattern @xmath221 .
( 260,200 ) ( 0,0 ) ( 10,10 ) and blocklength @xmath219 .
black curve : bp decoding with @xmath222 .
red curve : map decoding ( bp and gaussian elimination ) .
blue curves : bp decoding with @xmath223 ( almost indistinguishable ) .
red curves : tp decoding with @xmath223 ( truncated tree ) . , title="fig : " ] ( 12,0)(24,0)[cb]@xmath224,@xmath225,@xmath226,@xmath227,@xmath228,@xmath229,@xmath230,@xmath231,@xmath232,@xmath233,@xmath234 ( 8,12)(0,23)[rc]@xmath235,@xmath236,@xmath237,@xmath238,@xmath239,@xmath240,@xmath241,@xmath242 ( 260,12)(0,0)[l]@xmath15 ( 8,185)(0,0)[l]@xmath243}\prob\{\hat{x}^{{{\ensuremath{\text{\tiny tp}}}}/{{\ensuremath{\text{\tiny bp}}}}/{{\ensuremath{\text{\tiny map}}}}}_i(y_{[n]})\neq x_i\}$ ] ( 89,125 ) ( 37,85 ) the second example is the standard ( perfect ) golay code with blocklength @xmath244 . it is shown in fig . [
fig : cgolay_23 ] .
( 260,200 ) ( 0,0 ) ( 10,10 ) golay code with blocklength @xmath245 .
blue curve : bp decoding with @xmath222 .
black curve : map decoding ( bp and gaussian elimination ) .
blue curves : bp decoding with @xmath246 .
red curves : tp decoding with @xmath246 ( truncated tree ) . , title="fig : " ] ( 12,0)(24,0)[cb]@xmath224,@xmath225,@xmath226,@xmath227,@xmath228,@xmath229,@xmath230,@xmath231,@xmath232,@xmath233,@xmath234 ( 8,12)(0,23)[rc]@xmath235,@xmath236,@xmath237,@xmath238,@xmath239,@xmath240,@xmath241,@xmath242 ( 260,12)(0,0)[l]@xmath15 ( 8,185)(0,0)[l]@xmath243}\prob\{\hat{x}^{{{\ensuremath{\text{\tiny tp}}}}/{{\ensuremath{\text{\tiny bp}}}}/{{\ensuremath{\text{\tiny map}}}}}_i(y_{[n]})\neq x_i\}$ ] the third example , an ldpc code with blocklength @xmath247 , is depicted in fig . [
fig : creg36_50 ] .
( 260,200 ) ( 0,0 ) ( 10,10 ) ldpc code with blocklength @xmath248 .
blue curve : bp decoding with @xmath222 .
black curve : map decoding ( bp and gaussian elimination ) .
blue curves : bp decoding with @xmath249 .
red curves : tp decoding with @xmath249 ( truncated tree ) . , title="fig : " ] ( 12,0)(24,0)[cb]@xmath224,@xmath225,@xmath226,@xmath227,@xmath228,@xmath229,@xmath230,@xmath231,@xmath232,@xmath233,@xmath234 ( 8,12)(0,23)[rc]@xmath235,@xmath236,@xmath237,@xmath238,@xmath239,@xmath240,@xmath241,@xmath242 ( 260,12)(0,0)[l]@xmath15 ( 8,185)(0,0)[l]@xmath243}\prob\{\hat{x}^{{{\ensuremath{\text{\tiny tp}}}}/{{\ensuremath{\text{\tiny bp}}}}/{{\ensuremath{\text{\tiny map}}}}}_i(y_{[n]})\neq x_i\}$ ] in the case of the bawgn channel , we consider a single example of code , the tail - biting convolutional code used above , and two truncation schemes , the constructions @xmath1 and @xmath2 described in section [ sec : truncationgeneral ] . our results are shown in fig . [
fig : ccm2_50_bawgn ] and fig .
[ fig : exitcurvesconvc50bawgn ] . the tp and bp decoders are based on the natural periodic tanner graph associated with the tailbiting code .
we run bp a large number of iterations and check the error probability to be roughly independent of the iterations number .
the map decoder is performed using bp on the single - cycle tailbiting trellis ( i.e. , bcjr on a ring @xcite ) .
we observe that the two schemes @xmath211 and @xmath202 with @xmath250 outperform bp .
unhappily , due to complexity constraints we were limited to small values of @xmath12 and therefore could not approach the actual map performances .
( 260,200 ) ( 0,0 ) ( 10,10 ) and blocklength @xmath248 .
dashed black curve : bp decoding with @xmath251 .
black curve : map decoding ( wrap - around bcjr ) .
blue curves : bp decoding with @xmath252 .
red curves : tp decoding with @xmath250 ( truncated tree , denoted by tp(@xmath253 ) , tp decoding on a ball of radius 2 ( scheme @xmath1 with no bp processing , denoted by @xmath254 ) and , tp decoding according to scheme @xmath1 ( with parameters as indicated by @xmath255 ) . , title="fig : " ] ( 10,0)(40,0)[cb]@xmath256,@xmath257,@xmath258,@xmath259,@xmath260,@xmath261,@xmath262 ( 8,12)(0,54)[rc]@xmath239,@xmath240,@xmath241,@xmath242 ( 260,12)(0,0)[l]@xmath263 ( 258,-4)(0,0)[l](db ) ( 8,185)(0,0)[l]@xmath243}\prob\{\hat{x}^{{{\ensuremath{\text{\tiny tp}}}}/{{\ensuremath{\text{\tiny bp}}}}/{{\ensuremath{\text{\tiny map}}}}}_i(y_{[n]})\neq x_i\}$ ] ( 120,92 ) ( 210,117 ) ( 200,89 ) ( 170,82 ) ( 260,200 ) ( 0,0 ) ( 10,10 ) and blocklength @xmath248 .
blue curve : bp decoding with @xmath251 .
black curve : map decoding ( wrap - around bcjr ) .
red curve : tp decoding according to scheme @xmath2 ( with parameter as indicated by @xmath264 using a suitable truncated tree ) .
, title="fig : " ] ( 11,0)(22,0)[cb]@xmath265,@xmath256,@xmath257,@xmath266,@xmath141,@xmath267,@xmath74,@xmath268,@xmath269,@xmath270,@xmath271,@xmath272 ( 8,12)(0,20)[rc]@xmath273,@xmath235,@xmath236,@xmath237,@xmath238,@xmath239,@xmath240,@xmath241,@xmath242 ( 260,12)(0,0)[l]@xmath263 ( 258,-4)(0,0)[l](db ) ( 8,185)(0,0)[l]@xmath243}\prob\{\hat{x}^{{{{\ensuremath{\text{\tiny tp}}}}}/{{{\ensuremath{\text{\tiny bp}}}}}/{{\ensuremath{\text{\tiny map}}}}}_i(y_{[n]})\neq x_i\}$ ] ( 190,114 ) ( 180,104 ) ( 170,94 )
one interesting direction is to use the self - avoiding walk tree construction for analysis purposes .
we think in particular of two types of developments : @xmath0 a better understanding of the relation between bp and map decoding , and @xmath1 a study of the ` inherent hardness ' of decoding sparse graph codes . while the first point is self - explanatory , it might be useful to spend a few words on the second .
the most important outcome of the theory of iterative coding systems can be phrased as follows .
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ there exist families of graphs ( expanders @xcite , random @xcite ) with diverging size and bounded degree , below a certain noise level , map decoding can be achieved in linear time up to a ` small error . ' _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ we think ( a formal version of ) the same statement to be true for _ any family of graphs _ with bounded degree . this can be proved for the erasure channel .
let @xmath274 be a family of tanner graphs of diverging blocklength @xmath13 , with maximum variable degree @xmath275 and check degree @xmath276 .
consider communication over bec@xmath14 with @xmath277 .
then , for any @xmath278 there exists a decoder whose complexity is of order @xmath279 and returning estimates @xmath280 such that @xmath281 .
the decoder consists in returning the map estimate of @xmath10 given the subgraph @xmath205 and the values received therein . consider the subgraph @xmath282 of @xmath283 obtained by removing non - erased bits .
the proof consists in an elementary percolation estimate on this graph , see @xcite .
it is easy to see that @xmath284 is upper bounded by the probability that the connected component of @xmath282 that contains @xmath10 is not - contained in @xmath205 .
this is in turn upper bounded by the number of paths between @xmath10 and a vertex at distance @xmath285 ( which is at most @xmath286 ) times the probability that one such path is completely erased ( which is @xmath287 ) .
therefore , for @xmath288 and @xmath289 , we get @xmath290 the proof is completed by taking @xmath291 , and noticing that @xmath205 can be decoded in time polynomial in its size , that is polynomial in @xmath292 .
the computation is repeated for each @xmath293 whence the factor @xmath13 .
we think that a strengthening ( better dependence on the precision @xmath294 ) and generalization ( to other channel models ) of this result can be obtained using the self - avoiding walk tree construction .
yi lu is supported by the cisco stanford graduate fellowship .
cyril masson is supported by the swiss national fund postdoctoral fellowship . | ` tree pruning ' ( tp ) is an algorithm for probabilistic inference on binary markov random fields .
it has been recently derived by dror weitz and used to construct the first fully polynomial approximation scheme for counting independent sets up to the ` tree uniqueness threshold . '
it can be regarded as a clever method for pruning the belief propagation computation tree , in such a way to exactly account for the effect of loops . in this paper
we generalize the original algorithm to make it suitable for decoding linear codes , and discuss various schemes for pruning the computation tree .
further , we present the outcomes of numerical simulations on several linear codes , showing that tree pruning allows to interpolate continuously between belief propagation and maximum a posteriori decoding . finally , we discuss theoretical implications of the new method . |
significant numbers of massive galaxies have been identified in the near - infrared with colors so red that they are not found in surveys that select galaxies at visual wavelengths .
such objects have been named _ extremely red objects _ ( eros ) . with surface densities similar to that of bright quasars ( hu & ridgeway , 1994 )
, they represent an important component of the population of high - redshift galaxies .
the identification of eros coincided with the development and implementation of infrared arrays for astronomical research , the first eros being noted in the k - band surveys of elston , rieke , & rieke ( 1988 , 1989 ) . since then
, many groups have identified objects with extremely red colors ( mccarthy , persson , & west 1992 ; eisenhardt & dickinson 1992 ; persson et al .
1993 , hu & ridgeway 1994 , soifer et al .
1994 , dey , spinrad & dickinson 1995 , djorgovski , et al .
1995 , among others ) , though often with widely different definitions of _ extremely red_. these objects were primarily identified in the fields around high - redshift active galactic nuclei , i.e. around radio galaxies and quasars . in this paper , we define the selection criteria for bright eros to be @xmath1 @xmath6 190 and with ( @xmath0-@xmath1 ) @xmath2 6@xmath3 . the broad - band spectral energy distributions ( seds ) of these galaxies were initially thought to be fitted best by an old stellar population , such as found in present - day elliptical galaxies .
in this scenario , a redshifted ( @xmath9 ) strong 4000 - break falling between the @xmath0 and @xmath10 filter bandpasses and the lack of any appreciable restframe uv light from a young population of stars are responsible for the extremely red colors .
an alternative interpretation is that these galaxies are starbursts or active galactic nuclei ( agn ) , perhaps triggered by a merging event . in this scenario
, the presence of significant quantities of interstellar dust hides the star - forming regions or agn , considerably reddening the observed seds .
these would also lie at redshifts @xmath11 , which shifts the restframe uv into the optical bandpass while still sampling the restframe optical / near - ir wavelengths in the k filter .
archetypal examples of each of these interpretations have been identified .
lbds 53w091 ( dunlop et al . 1996 ) is a faint radio galaxy at a redshift of 1.55 , derived from an absorption line spectrum obtained at restframe uv wavelengths .
spectral modeling suggests an elliptical galaxy spectrum with an age of up to 3.5 gyr ( dunlop et al .
1996 , spinrad et al .
1997 , but see bruzual & magris 1997 for an alternative viewpoint ) , implying a much higher redshift of formation for this system .
in contrast , hr10 ( hu & ridgeway 1994 ) has been identified as a dusty starforming galaxy , perhaps with an active nucleus , at @xmath12 ( graham & dey 1996 ) .
detections of this galaxy at submillimeter and millimeter wavelengths support the interpretation that the sed is dominated by the presence of dust ( cimatti et al .
1998 ) . in both cases ,
the eros are most likely to lie in the redshift range @xmath13 .
it is difficult to produce such red colors at lower redshifts , while higher redshift objects become exceedingly luminous .
if eros are dominated by old stellar populations , then massive galaxy formation was well underway at @xmath14 .
dust - dominated eros , however , imply that much of the massive galaxy formation could actually occur at late times , supporting hierarchical galaxy formation models .
the contrasting implications these two scenarios have for the history of galaxy formation is the primary driver behind continuing work on these objects .
we have therefore started a wide - area field survey to select an unbiased and statistically significant sample of eros for further study .
this paper presents initial results from the first of seven fields , obtained as part of the calar alto deep imaging survey ( cadis , meisenheimer et al .
1997 ) , each of which will cover @xmath15160 square arcminutes .
section 2 describes the observations and reductions .
section 3 includes a discussion of the sample selection , morphology , and surface and volume density of the eros identified here .
section 4 gives a brief summary of this paper .
the field observed in this paper is centered on @xmath16 @xmath5 16@xmath17 23@xmath18 29@xmath19 @xmath20 @xmath5 @xmath21 50@xmath22 47@xmath23 ( hereafter the 16h field ) .
it is one of 7 fields in the cadis survey ( meisenheimer et al .
1997 ) , chosen to be free of bright stars or galaxies , known galaxy clusters , and in regions of minimal extinction due to galactic cirrus .
all magnitudes quoted in this paper are referred to @xmath24 lyra .
the @xmath1 imaging was obtained with the omega - prime camera ( bizenberger et al .
1998 ) at the prime focus of the calar alto 3.5 m telescope on ut 1996 june 7 - 8 and ut 1996 august 22 - 29 .
the majority of these nights were photometric .
omega - prime is a direct imaging camera equipped with a hawaii 1024@xmath25 hgcdte array .
the image scale is 0.396" / pixel ( 6.75 arcminute square field of view ) .
data were taken in a 2@xmath262 mosaic to cover as much of the cadis survey field ( 16@xmath22 diameter ) as possible .
the @xmath1 filter ( @xmath27@xmath28 m , @xmath29@xmath28 m , see wainscoat and cowie 1991 ) was used because its bluer bandpass relative to a standard k filter significantly reduces the thermal background seen by the array .
the data were reduced with a standard infrared reduction algorithm in iraf .
a sky frame was constructed from 6 to 12 temporally adjacent images , scaled to have the same median counts before averaging all but the 1 or 2 most extreme values in each pixel , effectively removing stars from the final sky image . the sky frame
was then scaled to and subtracted from the data frame and the result divided by a normalized dome flat to remove pixel - to - pixel fluctuations in quantum efficiency .
on photometric nights , standard stars from the ukirt faint standards list ( casali & hawarden 1992 ) were observed every 2 - 3 hours at several different airmasses to determine both a photometric zero point and an extinction correction .
individual images were then corrected for extinction and put onto a common photometric scale .
night to night variations of the zero points were under 0.1 magnitude , giving a measure of the systematic uncertainty in the @xmath1 calibrations .
as iraf was not able to handle either the large number of frames ( 1209 ) or the large size of the data set ( 4.8 gb ) , a procedure for idl was developed to stack the individual images into the final mosaic .
known bad pixels were ignored and only good pixels were included in the mosaic .
the software used a multi - pass sigma - clipping algorithm to reject cosmic rays and other transient phenomena ( e.g meteors and satellites ) .
the final mosaic is 2419 @xmath26 2546 pixels ( 1596 @xmath26 1680 ) , including regions of higher noise around the edge where coverage was incomplete .
the deepest portion of the mosaic , with per pixel exposure times greater than 6000 seconds , covers the central @xmath1512.5 arcmin square with a seeing of 1.1 arcsec full - width at half - maximum ( fwhm ) .
a grayscale image of the exposure time map , along with the relative fields of view for the optical and infrared data , is shown in figure [ xtmap ] .
the central portion of the @xmath1 mosaic has a 5@xmath4 point source detection limit of @xmath1 @xmath5 205 , within an aperture diameter of 22 , twice the seeing fwhm .
limits for extended sources are up to 050075 brighter , depending on the object morphology .
the @xmath0-band imaging was obtained with the cafos reimaging camera at the calar alto 2.2 m telescope on ut 1996 may 15 ( 5@xmath26500s , 13 seeing ) and ut 1997 february 5 ( 4@xmath26700s 16 seeing ) .
cafos was equipped with a loral 2048@xmath25 ccd ( 0.33@xmath23/15@xmath28 m pixel ) for the may observations , and with a site 2048@xmath25 ccd ( 0.53@xmath23/24@xmath28 m pixel covering a circular field of view @xmath1516 arcminutes in diameter ) for the remaining observations .
the @xmath0 filter used in cadis ( @xmath30 nm , @xmath31 nm ) is slightly narrower and bluer than a standard johnson @xmath0 filter .
this difference has no significant impact on the goals of this survey .
the data were reduced with the midas software package . following bias subtraction and flatfielding , multiple exposures obtained on a single night
were shifted into coincidence , bad pixels and cosmic rays masked out , then summed to form a deep image for that night .
flux calibration of the @xmath0-band data was established relative to two stars in the 16h field for which we have good spectrophotometry .
these stars were calibrated against an hst uv spectrophotometric standard ( agk@xmath32 ) as well as 3 stars from oke ( 1990 ) .
the magnitudes of the two 16h field reference stars in the @xmath0 filter were determined by integrating their calibrated spectra over the @xmath0 filter bandpass and instrumental transmission curve .
the 2@xmath4 limits were established within an aperture diameter twice the seeing fwhm , scaling from the per - pixel noise in the stacked data .
a full description of the calibration of the cadis optical data set can be found in fockenbrock ( 1998 ) .
objects were separately identified on the deep @xmath0 and @xmath1 images using the source extractor software ( bertin & arnouts 1996 ) and the resulting lists merged into a master catalog .
this catalog was then used to extract photometric magnitudes for all objects in each individual exposure ( mosaics from each night for the @xmath1 data ) using photometry software ( meisenheimer & rser 1986 ) incorporating a gaussian weighting function , designed to compensate for variations in seeing between individual exposures .
the procedure recovers accurate _ colors _ for all objects , but underestimates the total magnitude for extended objects . for the faint galaxy images from which the sample of eros was drawn , typically subtending only 1 - 2@xmath23 , the underestimate is of order 0.1 magnitude .
a final list of objects was selected from the master catalog , containing objects with @xmath1 magnitudes above the 5@xmath4 limit of 205 that also appear on at least one of the two @xmath0 images ( from the loral and site ccds ) .
figure [ xtmap ] shows a contour map and image of the exposure time , overlaid with the outline of the survey area .
the full area imaged in both @xmath1 and @xmath0 covers 154 square arcminutes .
we select objects with extremely red colors from the color - magnitude diagram for the full survey area , shown in figure [ fullcmd ] .
we define the selection criteria for bright eros as objects with @xmath1 @xmath6 190 and ( @xmath0-@xmath1 ) @xmath2 6@xmath3 .
a more detailed plot of this region of the color - magnitude diagram is shown in figure [ zoomcmd ] .
eight objects in our survey field meet these selection criteria . their positions , magnitudes and colors are summarized in table [ ero_data ] , while finding charts from the @xmath1 data are given in figure [ fcs ] .
the magnitude limit defining the bright sample of eros is partially driven by the current data : most eros brighter than @xmath1 @xmath6 190 are detected above the 2@xmath4 limit in @xmath0 while most of those fainter than this are not .
this also represents the practical limit for obtaining followup near - infrared spectroscopy at 8m - class telescopes ( moorwood & spyromilio 1997 ) with exposure times of order 1 hour . in the absence of significant dust extinction
, the reddest galaxy colors will come from the oldest stellar populations at any redshift . a present - day elliptical galaxy sed under a no - evolution model peaks in color at @xmath33 ( ignoring the lyman break at higher redshift ) , with an @xmath0-@xmath1color of about 6 . for an old stellar population ,
this color is produced by a combination of the cosmological k - corrections and a strong 4000 - break lying between the two bandpasses . since
all galaxies should be bluer than this limit , selecting objects with @xmath0-@xmath1 @xmath2 6@xmath3 will isolate a sample of galaxies with extremely red colors .
the specific choice of @xmath0-@xmath1@xmath2 60 is , however , also motivated by historical precedence , approximating the limits used to identify other eros in the literature .
such objects were identified as lying away from the locus of other field objects in the color - magnitude diagram . however , the areas surveyed , and thus the number of objects detected , were generally too small to address this point statistically .
figure [ colorhist ] displays histograms of the @xmath0-@xmath1colors for objects with 16@xmath3 @xmath6 @xmath1 @xmath6 20@xmath3 in 1@xmath3 bins .
table [ colorhist_data ] lists the mean color and fwhm for the gaussian least - square fits which are also plotted in figure [ colorhist ] .
for example , in the range 185 @xmath6 @xmath1 @xmath6 195 , the bin containing most of the eros , the mean color is 392 and the standard deviation is 111 .
an ( @xmath0-@xmath1 ) @xmath2 6@xmath3selection criterion therefore corresponds to colors that deviate by more than about two standard deviations from the mean color .
the extremely red galaxies selected here are thus the reddest 2% of the population and should consist of the oldest and/or the dustiest high redshift galaxies .
there is a distinct trend to redder colors at fainter magnitudes , with the mean ( @xmath0-@xmath1 ) color becoming 07 redder between @xmath1 @xmath5 165 and @xmath1 @xmath5 195 .
this continues the trend noted at brighter magnitudes in ( i - k ) colors by huang et al .
( 1997 ) . over the range sampled by our data ,
this trend probably reflects an increasing fraction of high redshift galaxies .
there are an additional 43 objects with ( @xmath0-@xmath1 ) @xmath2 6@xmath3 in the range 190 @xmath34 @xmath1 @xmath6 200 .
these fainter eros lie only in the central 120 square arcminutes of the survey field , where the overlap of the two sets of @xmath0-band data reach deep enough to select eros with @xmath1 @xmath6 200 .
we note that since both the color and color dispersion for the objects in the 16@xmath35 field increase with increasing magnitudes , the eros in this fainter magnitude range may no longer have truly extreme colors .
the current optical data do not have sufficient depth to address this more quantitatively .
we chose to set the detection limit in the @xmath0 data at 2@xmath4 to ensure that all of the eros are selected , at the expense of some possible contamination of the sample . in practice , contamination is likely to be low . consider a malmquist bias vector defined on the color - magnitude diagram ( figure [ zoomcmd ] ) . for objects near the ero selection region , it would be dominated by the uncertainties in the @xmath0-band magnitudes , since most of the objects there are well - detected at @xmath1 but near the 2@xmath4 detection limit in @xmath0 .
the malmquist bias vector would thus point to _ bluer _ colors and slightly brighter @xmath1 magnitudes .
inspection of figure [ zoomcmd ] shows that there are a comparable number of objects which could scatter into the ero region as are already there , so contamination of the sample is likely to be low and would be coming from objects with redder colors .
the radial brightness profiles of the eros were compared to those of several stars near each object to determine whether the eros are resolved .
five of the 8 bright eros are easily resolved , with fwhms ranging from 14 to 23 in 11 seeing , and are therefore likely to be galaxies .
subtracting the intrinsic seeing in quadrature gives deconvolved fwhms ranging from 09 to 20 .
such sizes are similar to both hr10 ( 07 , hu & ridgeway 1994 ) and 53w091 ( 13 , dunlop et al .
1996 ) . a sixth object , c16-ero3 , is formally resolved , but at a low statistical significance , with a deconvolved fwhm @xmath6 04 .
the remaining 2 bright eros in the 16h field are unresolved .
their broadband colors indicate that they are low mass galactic stars .
one has subsequently been spectroscopically confirmed as a low mass star of spectral type l1 ( wolf et al .
the second is well - detected in the cadis filters ( @xmath1@xmath5 1632 ) , with colors fully consistent with a star of spectral type m9 ( wolf et al .
1998 ) , though without spectroscopic confirmation we can not completely rule out that the object may be a compact galaxy or agn at high redshift . finding charts for these two objects can be found in wolf et al .
( 1998 ) , and they are not considered further in this paper .
only one of the six resolved eros is significantly non - circular ( c16-ero1 ) .
the deconvolved fwhms along the major and minor axes are 19 and 11 , respectively , with the major axis oriented at a position angle ( pa ) of 143@xmath36 , east of north .
there is also a companion ero with @xmath1 @xmath37 190 , 28 away at pa @xmath5 126@xmath36 .
the proximity of the second ero and the apparent alignment between the pa of c16-ero1 and the direction to the fainter galaxy suggest that this might be an interacting system .
c16-ero1 and c16-ero2 also make the closest pair of eros in the current sample , separated by 392 at pa @xmath5 1255 , possibly indicating the presence of a cluster or group .
one other object ( c16-ero4 ) has a fainter ero companion nearby ( 49 away at pa @xmath5 127@xmath36 ) , though there is no evidence for interaction in the current data set . of the fainter 43 eros ,
37 are either clearly or marginally resolved .
below k@xmath38 195 , it becomes difficult to distinguish between unresolved and slightly extended sources on the survey data .
nevertheless , the fainter eros in this field are clearly dominated by galaxies . the 6 bright
, extragalactic eros identified here were found in an area of 154arcmin@xmath25 , yielding a surface density of 0.039 @xmath7 0.016 arcmin@xmath8 ( poisson uncertainties ) .
this density is a factor of four higher than found by hu & ridgeway ( 1994 ) , estimated from 100arcmin@xmath25 of i and @xmath1 data from the hawaii deep k - band surveys ( cowie et al .
in their survey , hu & ridgeway found no galaxies as red as hr10 , implying a surface density of @xmath390.01arcmin@xmath8 .
the difference is not statistically significant due to the currently large uncertainties from the small number statistics as well as a possible contribution from field - to - field variations .
combining these data with those from future fields obtained as part of this survey will reduce this uncertainty to @xmath1515% . still , the surface density of eros is comparable to or higher than that of bright ( @xmath40@xmath3 ) quasars , making the eros a significant population .
dey , spinrad , and dickinson ( 1995 ) suggest that eros are more common along the line of sight to or clustered with high redshift radio galaxies and quasars .
estimates of the surface density range up to 10 - 100 times that of the field eros .
if the eros are at the same redshifts as the high - redshift agn , then many of them would have luminosities significantly in excess of @xmath41 . with the eros in the foreground
there is the possibility of a magnification bias , where the ero ( or a group or cluster containing it ) is acting as a gravitational lens for the background agn .
this higher surface density in the fields of high - redshift agn may also reflect what we term an `` observational bias . ''
this is a combination of ( _ i _ ) agn were a common target for observations in the near - ir , ( _ ii _ ) unusual objects of any sort were noted ( here objects with extremely red colors ) when the data were published , and ( _ iii _ ) earlier generations of near - ir imagers had fields of @xmath151arcmin@xmath25 .
a systematic measurement of the surface density in agn fields still remains to be done .
guiderdoni et al .
( 1997 , their model c ) predict the surface density ( @xmath42arcmin@xmath43 of high - redshift dusty star - forming galaxies observed at far - infrared wavelengths that is remarkably similar to the surface density of bright eros found here , suggesting that eros may be dusty starbursts .
the predictions were based on local ultra - luminous infrared galaxies , actively forming stars but completely shrouded by dust .
the detection of significant quantities of dust in hr 10 ( cimatti et al . 1998 ) supports this picture . including the additional 43 eros with @xmath1 @xmath6 20@xmath3
, the surface density is [email protected] arcmin@xmath8 .
as noted above , the fainter sample is not well defined over the survey area , and this is only an estimate to the true surface density of faint eros .
the depth of the @xmath0 band data is currently the limiting factor , and deeper imaging will allow us to make a stronger statement about these fainter eros .
we can derive a volume density of eros if we can estimate the redshift range over which they occur .
this volume density would then be directly comparable to known populations of galaxies , perhaps providing some insight into their nature .
if the ero population is dominated by old ( elliptical ) galaxies , the red color would be coming from a strong 4000 break .
this spectral break comes primarily from balmer continuum limit and ca h@xmath44k absorption lines and can significantly redden an ( @xmath0-@xmath1 ) color when the break lies between the two filter bandpasses . for the @xmath0 filter used here
, this implies a redshift beyond @xmath45 .
if , on the other hand , the ero population is dominated by dusty starburst galaxies , the red color would be due to dust extinction reddening a flat sed .
a color of ( @xmath0-@xmath1 ) @xmath15 6@xmath3 can be produced at @xmath46 by a v - band extinction of @xmath47 , assuming the seaton ( 1979 ) milky way extinction law . in the grayer extinction law of calzetti
et al . ( 1994 ) , derived empirically from a sample of local starburst galaxies , a factor of two more extinction would be required to give similar results .
the steeply rising extinction curves at shorter wavelengths , however , imply that less dust extinction would be required to produce the same reddening at higher redshifts . in practice ,
a moderate continuum decrement across the 4000 break is also seen in relatively young stellar populations , contributing to the red color and relaxing the extinction requirements .
we therefore adopt a redshift of @xmath45 as the lower redshift limit for both types of galaxies . for the high redshift limit ,
consider an l@xmath48 galaxy at @xmath49 .
it would have an apparent magnitude of @xmath1 @xmath50 203 ( h@xmath51kms@xmath52mpc@xmath52 and q@xmath53 ) , assuming only passive evolution ( evolution and k - corrections from poggianti 1997 and fioc & rocca - volmerange 1997 give similar results ) .
our selection of bright eros at @xmath1 @xmath6 19@xmath3 is thus already sampling the luminosity function well above l@xmath48 at @xmath49 . beyond this redshift
, the galaxies would have to be intrinsically as luminous as bright quasars to be selected as eros .
we therefore adopt the range @xmath54 $ ] where eros are most likely to be found with the selection criteria used here .
the comoving volume covered in this redshift range over our survey field in the assumed cosmology is 3.55@xmath55mpc@xmath56 .
for the six bright eros in our sample , this yields a volume density of @xmath57mpc@xmath58 .
extending the redshift range down to @xmath59 would only increase the comoving volume by 15% .
increasing the upper redshift limit to @xmath60 would approximately double the sampled volume .
the volume density derived here is , therefore , not very sensitive to the assumed redshift range .
we can compare this estimate of the volume density of eros with that of other known types of objects . for quasars ( m@xmath61 ) , the volume density in a similar redshift range ( boyle 1991 ) is only @xmath62mpc@xmath58 , a factor of 8 lower than the bright ero density .
a simple extrapolation of the boyle luminosity function to m@xmath63 ( it is strictly only valid for m@xmath64 ) is required to reach space densities similar to those of the eros .
thus the _ bright _ ero galaxies appear to be considerably more numerous than quasars , and perhaps as common as seyfert is in the local universe .
the eros , however , have been compared to both old ellipticals and young dusty starbursts , and it is perhaps more interesting to compare them to these groups of objects .
the red galaxies from the canada - france redshift survey ( cfrs , lilly et al . 1995 ) represent a population of objects which to not undergo significant evolution from @xmath65 to the present , similar to what would be expected of the eros if they are dominated by old elliptical galaxies .
the integrated luminosity function for the cfrs red galaxies reaches a volume density comparable to the eros at a luminosity of @xmath154l@xmath48 .
thus , if the ero population is dominated by old elliptical galaxies , then the eros are very massive and suggest that the most massive systems collapsed early in the history of the universe .
such galaxies at @xmath66 could also mark the location of high redshift ( forming ? ) clusters or groups of galaxies .
luminous infrared galaxies are often powered by strong starbursts or agn , but emit most of their bolometric luminosity at restframe far - infrared wavelengths due to the presence of significant quantities of dust .
the ero population could be high - redshift analogs to these luminous infrared galaxies if the eros are dominated by objects like hr10 .
the integrated luminosity function for the luminous infrared galaxies ( soifer & neugebauer 1991 ) reaches a volume density comparable to the eros at a bolometric luminosity of a few@xmath67 , corresponding to @xmath1510l@xmath48 .
thus , if the ero population is dominated by young , dusty starbursts similar to the luminous infrared galaxies , then many massive galaxies were undergoing formative mergers at relatively late times .
we have surveyed an area of 154 sq .
arcmin to a 5@xmath4 point - source limit of @xmath1 @xmath5 205 , representing the largest contiguous field imaged in the near infrared at this depth to date . from these data ,
we have selected 8 eros with @xmath1 @xmath6 190 and with ( @xmath0-@xmath1 ) @xmath2 6@xmath3 .
six of the 8 bright eros are resolved at @xmath1 and are , therefore , most likely to be galaxies at high redshift ( @xmath68 ) .
these eros have a surface density of 0.039 @xmath7 0.016 arcmin@xmath8 , or , with simple assumptions about their range of redshifts , a volume density comparable to that of seyfert galaxies .
interpretations as to the nature of these galaxies range from old , evolved elliptical galaxies to young dusty starbursts or agn .
the remaining 2 eros are low - mass galactic stars , either main sequence m stars or brown dwarfs ( wolf et al . 1998 ) .
such low - mass stars can be seen to distances well beyond 100 pc , depending on their intrinsic luminosity .
morphological considerations alone rule out the possibility that the majority of the eros with @xmath1 @xmath6 20@xmath3 are low - mass stars .
the true count of such stars in high - latitude fields will provide interesting constraints on galactic structure and the faint end of the local stellar luminosity function .
eros appear to be a significant population of galaxies absent from samples derived from observations at visual wavelengths
. future work on these objects will distinguish between the old elliptical and young dusty starburst interpretations .
observations at submillimeter ( ground based ) and mid- to far - infrared ( sirtf , ngst ) wavelengths will be able to quantitatively determine how much dust , if any , is present .
in either case , the eros are likely drawn from a population of massive galaxies at high redshift . whichever interpretation prevails
, the eros will provide important clues to the formation and evolution of massive galaxies .
we are grateful to the team that constructed the omega camera , particularly m. mccaughrean and p. bizenberger , without whom the wide field , deep k@xmath22 survey would not have been possible .
we also thank the staff at the calar alto observatory for support during the observing runs , and mpia for support of the cadis key project .
this research was supported by the max - planck - society .
bertin , e. & arnouts , s. 1996 , a&as , 117 , 393 .
bizenberger , p. , mccaughrean , m. , thompson , d. , & birk , c. , 1998 , spie conference proceedings .
3354 , p. 825 .
boyle , b. j. 1991 , in _ texas / eso - cern symposium on relativistic astrophysics , cosmology , and fundamental physics _
, eds . j. d. barrow , l. mestel , & p. a. thomas , ( n.y.acad.sci.:new york ) , p. 14 .
bruzual , g. a. & magris , g. c. 1997 , iau symp . 183 .
calzetti , d. , kinney , a. l. , & storchi - bergmann , t. 1994 , , 429 , 582 .
casali , m. m. & hawarden , t 1992 , _ jcmt ukirt newsletter _ , 4 , 33 .
cimatti , a. , andreani , p. , rttgering , h. & tilanus , r. 1998 , , 392 , 895 .
cowie , l. l. , gardner , j. p. , hu , e. m. , songaila , a. , hodapp , k .- w . , &
wainscoat , r. j. 1994 , , 434 , 114 .
dey , a. , spinrad , h. , & dickinson , m. 1995 , , 440 , 515 .
djorgovski , s. , soifer , b. t. , pahre , m. a. , larkin , j. e. , smith , j. d. , neugebauer , g. , smail , i. , matthews , k. , hogg , d. w. , blandford , r. d. , cohen , j. , harrison , w. , & nelson , j. 1995 , , 438 , l13 .
dunlop , j. , peacock , j. , spinrad , h. , dey , a. , jimenez , r. , stern , d. , & windhorst , r. 1996 , , 381 , 581 .
eisenhardt , p. & dickinson , m. 1992 , , 399 , l47 .
elston , r. , rieke , g. h. , & rieke , m. j. 1989 , , 341 , 80 .
elston , r. , rieke , m. j. , & rieke , g. h. 1988 , , 331 , l77 .
fioc , m. , & rocca - volmerange , b. 1997 , , 326 , 950 .
fockenbrock , r. 1998 , ph.d .
thesis , ruprecht karls universitt , heidelberg , germany .
graham , j. r. & dey , a. 1996 , , 471 , 720 .
guiderdoni , b. , bouchet , f. r. , puget , j .- l .
, lagache , g. & hivon , e. 1997 , , 390 , 257 .
hu , e. m. & ridgeway , s. e. 1994 , , 107 , 1303 .
huang , j .- s .
, cowie , l. l. , gardner , j. p. , hu , e. m. , songaila , a. , & wainscoat , r. j. 1997 , , 476 , 12 .
lilly , s. j. , tresse , l. , hammer , f. , crampton , d. , & lefevre , o. 1995 , , 455 , 108 .
mccarthy , p. j. , persson , s. e. , & west , s. c. 1992 , , 386 , 52 .
meisenheimer , k. & rser , h .- j .
1986 , in _ the optimization of the use of ccd detectors in astronomy _ , eds . j .-
baluteau & s. dodorico , ( eso : garching ) , p. 227 .
( eso - ohp : garching bei mnchen ) , p. 227 .
meisenheimer , et al .
1997 , in _ the early universe with the vlt _ , ed .
j. bergeron ( springer : heidelberg ) , 165 .
moorwood , a. f. m. , & spyromilio , j. 1997 , in _ the early universe with the vlt _ , ed .
j. bergeron ( springer : heidelberg ) , 21 .
oke , j. b. 1990 , , 99 , 1621 .
persson , s. e. , mccarthy , p. j. , dressler , a. , matthews , k. 1993 , in _ the evolution of galaxies and their environment , _ ed .
d. hollenbeck et al . , nasa conference publication # 3190 , p. 78 .
poggianti , b. m. 1997 , supp .
ser . , 122 , 399 .
seaton , m. j. 1979 , , 187 , 73p .
soifer , b. t. , matthews , k. , djorgovski , s. , larkin , j. , graham , j. r. , harrison , w. , jernigan , g. , lin , s. , nelson , j. , neugebauer , g. , smith , g. , smith , j. d. , & ziomkowski , c. 1994 , , 420 , l1 .
soifer , b. t. , & neugebauer , g. 1991 , aj , 101 , 354 .
spinrad , h. , dey , a. , stern , d. , dunlop , j. , peacock , j. , jimenez , r. , & windhorst , r. 1997 , , 484 , 581 .
wainscoat , r. & cowie , l. l. 1991 , aj , 103 , 332 .
wolf , c. , mundt , r. , thompson , d. , chaffee , beckwith , s. v. w. , f. , fockenbrock , fried , j. , r. , hippelein , h. , huang , j .- s . ,
von kuhlmann , b. , leinert , c. , meisenheimer , k. , phleps , s. , rser , h .- j . , &
thommes , e. 1998 , , 338 , 127 .
cccccc c16-ero1 & 16:24:01.22 @xmath4455:38:49.0 & 18.56@xmath69 & 26.04@xmath70 & 7.48 & 20 + c16-ero2 & 16:24:04.99 @xmath4455:38:26.2 & 18.90@xmath71 & 24.97@xmath72 & 6.07 & 10 + c16-ero3 & 16:24:28.83 @xmath4455:43:10.4 & 19.00@xmath73 & 25.14@xmath74 & 6.14 & 04 + c16-ero4 & 16:24:43.84 @xmath4455:49:40.6 & 18.98@xmath69 & 25.17@xmath70 & 6.19 & 10 + c16-ero5 & 16:25:05.53 @xmath4455:39:54.4 & 18.92@xmath75 & 25.64@xmath76 & 6.72 & 11 + c16-ero6 & 16:25:13.11 @xmath4455:48:30.4 & 18.78@xmath77 & 24.86@xmath78 & 6.08 & 09 + c16-lms1 & 16:25:00.63 @xmath4455:44:44.3 & 18.57@xmath79 & 26.00@xmath70 & 7.43 & 00 + c16-lms2 & 16:23:57.24 @xmath4455:46:27.5 & 16.32@xmath80 & 23.06@xmath81 & 6.74 & 00 + ccc 160 @xmath6 @xmath1 @xmath34 170 & [email protected] & [email protected] + 170 @xmath6 @xmath1 @xmath34 180 & [email protected] & [email protected] + 180 @xmath6 @xmath1
@xmath34 190 & [email protected] & [email protected] + 185 @xmath6 @xmath1 @xmath34 195 & [email protected] & [email protected] + 190 @xmath6 @xmath1 @xmath34 200 & [email protected] & [email protected] + | we present initial results from a field survey for extremely red objects ( eros , defined here as ( @xmath0-@xmath1 ) @xmath2 6@xmath3 ) covering 154 square arcminutes of sky , from the first of 7 deep , wide - field @xmath1 images obtained as part of the calar alto deep imaging survey ( cadis ) .
the 5@xmath4 point source detection limits are @xmath1
@xmath5 205 and @xmath0 @xmath5 250 , while extended - source limits are up to 050075 brighter .
we identify a total of 8 bright eros with @xmath1 @xmath6 190 .
six of these bright eros are resolved and are likely to be galaxies , while the remaining 2 are unresolved , with colors consistent with their being low - mass galactic stars .
we derive a surface density for the 6 bright , extragalactic eros of [email protected] arcmin@xmath8 , higher by a factor of 4 than previous values .
we estimate that the volume density of bright eros to be as high as that of nearby seyfert galaxies . |
statistical behavior of the foreign exchange ( fx ) markets and price fluctuations in currency have been the subject of studies in recent years @xcite , @xcite . the high - frequency data for financial markets has made it possible to investigate market dynamics on timescales as short as 1 min , a value close to the minimum time needed to perform transaction in the market .
it was observed that the short - term price fluctuations in fx market , for example , between us dollar and german mark , has the same statistical behavior as the velocity differences in hydrodynamic turbulence @xcite .
probability distributions in turbulence well fit the experimental data by superposition of the gaussians with log - normal distributions of its variances @xcite,@xcite .
the convergence of the velocity differences distributions toward a gaussian shape corresponds to a decrease of the log - normal variance for increasing of spatial distances @xmath0 . in the fx fluctuation dynamics
the statistical distributions of the price difference separated by time @xmath1 was elaborated by the theoretical model of a levy walk or levy flight @xcite,@xcite .
i have developed a new stochastic dynamic theory for the fx market .
the langevin type stochastic differential equation is proposed .
i introduce the random force with some general characteristics in order to fit the theoretical predictions of my model with observed statistical dependencies .
it is shown that the developed model describes exactly the
experimental data of the dynamics of a price index of the new york stock exchange .
the probability distributions of the standard & poor s 500 index differences @xmath2 are reproduced exactly by the @xmath3 .
the new model describes well the scaling behavior of the probability of return @xmath4 as a function of @xmath1 ( for definitions , see @xcite ) .
we will describe the dynamics of the fx price @xmath5 by the stochastic differential equation @xmath6 where @xmath7 is the random force .
the quantity which were interested about is the price differences separated by the time scale @xmath1 , @xmath8 the distribution function of stochastic process @xmath9 is defined as follow @xmath10 where ... means the averaging over the all possible realizations of the random force @xmath7 .
let us construct the general stochastic force @xmath7 by the following way @xmath11 here @xmath12 are the random amplitudes , @xmath13 is the response function , @xmath14 are the homogeneously distributed ( on time interval @xmath15 $ ] ) moments of time , the number @xmath16 of which obeys the poisson law .
thus , the averaging includes three statistically independent averaging procedures : \1 . averaging over random amplitudes
@xmath12 , @xmath17 , @xmath18 where @xmath19 is the probability distribution of amplitudes @xmath12 . \2 . averaging over @xmath14 on time interval @xmath20 , @xmath21 \3 . averaging over random numbers @xmath16 of time moments @xmath14 , @xmath22 where @xmath23 and @xmath24 is the density of points @xmath14 on time interval @xmath20 . taking into account the definition eq.([eq3 ] ) and performing the averaging in accordance with eqs.([eq5])-([eq7 ] ) we will have @xmath25 where we put notation @xmath26 for simplicity the eq.([eq8 ] ) is considered for the market currency situation when @xmath19 is factorized as follow @xmath27 where the gaussian distribution @xmath28 is given by @xmath29 we also choose the response function in the form @xmath30 the response function describes the influence of a piece of information which has become available at the delay time @xmath31 on the decision of a trader to propose or accept a price change .
the parameter @xmath32 is the character scale of the time delay .
we also have considered the response function in the form @xmath33 where @xmath34 is a new parameter .
for example , for @xmath35 we have @xmath36 where @xmath37 thus , we derive the new @xmath3 distribution ( see eqs .
( [ eq8 ] ) , ( [ eq9 ] ) ) starting from the differential stochastic equation eq.([eq1 ] ) . as it was mentioned the levy stable distribution
is widely applied to this problem .
the comparison between @xmath3 transformed to the form @xmath38 @xmath39 the levy stable distribution @xcite , @xcite , @xcite @xmath40 and the gauss distribution @xmath41 is shown in fig.1 .
figure 2 is a comparison of scaling properties of @xmath42 , and @xmath43 it is interesting that the _ laskin _ stochastic dynamics and phenomenological approach based on the _ levy _ distribution give the same scaling for the probability of return @xmath4 of the s&p 500 index variations as a function of the @xmath1 @xcite .
the main goal of the developed theory is to predict the behavior of currency exchange market in the future and based on these prediction to develop the efficient currency market strategy . as we will see
further it is useful to generalize the eqs.([eq1 ] ) , ( [ eq4 ] ) in following way @xmath44 where @xmath45 is a price dissipative coefficient the financial mean of which will be discussed further .
let us define the new pdf @xmath46 as follow @xmath47 where @xmath48 is the formal solution of the eq.([eq19 ] ) . and @xmath49 using the eqs.([eq19 ] ) , ( [ eq20 ] ) it is easy to obtain the evolution equation for pdf @xmath46 @xmath50 @xmath51 the initial condition for this equation has the form @xmath52 the eq.([eq21 ] ) , ( [ eq22])will serve as the main equations to predict the behavior of fx market in the future .
it is easy to see that the solution of the problem eqs.([eq21 ] ) and ( [ eq22 ] ) can be written as @xmath53 @xmath54 where @xmath55 is characteristic function of the random amplitude @xmath56 .
it is well - known @xcite that the pdf @xmath28 and characteristic function @xmath55 are connected each other @xmath57 and @xmath58 the eqs.([eq21 ] ) and ( [ eq22 ] ) allow one to study the evolution problems in fx currency market .
some analytically solvable evolution problems will be demonstrated in the next publication .
the new dynamic stochastic theory of fx currency market dynamics is developed .
we have established the new _ laskin _ distribution which well describes the observed statistical dependencies of the price difference separated by time scale @xmath1 and the _ laskin _ pdf which allows elaborate evolution of the currency market and predict behavior of market dynamics .
the theoretical predictions based on the _ laskin _ distribution are compared with phenomenological the _ levy _ distribution based approach and observed statistical dependencies .
it is interesting that the _ laskin _ stochastic dynamics and phenomenological approach based on the _ levy _ distribution give the same statistical and scaling dependencies
. 1 r.n .
mantega and h.e .
stanley , scaling behaviour in the dynamics of economic index , nature , v.*376 * 1995 , p.46 - 49 .
s. ghashghaie , w.breymann , j. peinke , p.talkner , y.dodge , turbulent cascades in foreign exchange markets , nature , v.*381 * , 1996 , p.767 - 770 .
b. castaing , y. gagne and e. hopfinger , velocity probability density functions of high reynolds number turbulence , physica * d46 * , 1990 , p.177 - 200 p. levy , _ theorie de laddition des variables aleatoires _ , gauthiere - villars , paris , 1937 w. feller , _ an introduction to probability theory and its applications _ ( wiley , new york , 1971 )
fig.1 . comparison between @xmath3 , @xmath59 and @xmath60 ( see the definitions eqs.([eq16 ] ) , ( [ eq17 ] ) , ( [ eq18 ] ) ) .
for @xmath3 the parameters are @xmath61 , @xmath62 , @xmath63 and @xmath64 .
for @xmath65 the parameters are @xmath66 and @xmath67 . for @xmath60 , @xmath68 ( see , @xcite ) . | a new stochastic theory of a foreign exchange markets dynamics is developed . as a result
we have the new probability distribution which well describes statistical and scaling dependencies experimentally observed in foreign exchange markets in recent years .
the developed dynamic theory is compared with well - known phenomenological levy distribution approach which is widely applied to this problem . it is shown that the developed stochastic dynamics and phenomenological approach based on the levy distribution give the same statistical and scaling dependencies . _
pacs _ number(s ) : 02.50 ey , 05.40 .
fb tcilatex |
one of the main goals of nuclear physics is to understand to what extent a nucleus differs from a loosely bound system of quasi - independent nucleons .
when nucleons are very close spatially , corresponding to rare high momentum components of the single particle wave function , many interesting and potentially exotic configurations can arise .
one way to look for such configurations is with reactions that are significantly sub - threshold to production from a free nucleon . of all such reactions ,
photoproduction of charmonium is one of the cleanest because the charm quark content of a nucleon is expected to be small compared to the light quarks . in light meson photoproduction , the quark content of the mesons can originate in the nuclear target , while in the case of charmonium photoproduction , the quark - interchange mechanism is essentially absent , and the reaction must proceed via gluon exchange in order for color to be conserved .
in addition , the heavy mass of the charm quark ( about 1.5 gev ) ensures a hard scale to the problem , making it more tractable in qcd . the goal of this experiment was to study the production mechanisms in the extreme conditions of matter that may be relevant in heavy ion collisions .
these conditions are ensured by using a photon beam energy well below photoproduction threshold on a free nucleon . in the quasi - free picture where a photon interacts with a single off - shell nucleon in a nucleus
, the nucleon three - momentum @xmath1 must be pointing anti - parallel to the photon direction ( @xmath2 ) for the invariant mass of the photon - nucleon system @xmath3 to be above the threshold value of @xmath4 gev@xmath5 , where @xmath6 is the nucleon mass and @xmath7 is the @xmath0 mass .
an additional constraint is that the missing energy @xmath8 can not be too high . specifically , we calculated @xmath3 for a given photon energy @xmath9 using @xmath10 from eq .
[ eq : s ] , it is clear that larger values of @xmath8 correspond so smaller values of @xmath3 , for fixed values of @xmath11 .
the highest value of @xmath8 that is kinematically allowed is shown as a function of the magnitude of @xmath1 for photon energies of 4 , 5 , 6 , and 8 gev in fig . [
fig : pmem ] ( for the case of @xmath1 anti - parallel to the photon direction ) .
since @xmath8 must be greater than zero , this leads to the conclusion that for @xmath0 photoproduction , the minimum nucleon momentum @xmath11 increases as the photon energy decreases .
in particular the minimum momentum for 4 gev photons is 1.15 gev , for 6 gev photons it is 0.35 gev , and for 8 gev photons it is 0.05 gev . as discussed below , it is generally thought that the region where few - nucleon short - range correlations , hidden color configurations , and other short - range effects play a significant role for momenta larger than 0.35 gev @xcite , corresponding to a photon energy of 6 gev .
this was therefore chosen as the ideal energy for the present experiment . in practice
, we used a slightly lower energy due to accelerator limitations at the time that the data were taken . another way to realize the importance of short - range correlations comes from the observation that the photon threshold to produce @xmath0
meson is 5.7 gev for a deuteron target at rest , and 4.8 gev for a triton or @xmath12he target .
thus , two - nucleon and particularly three - nucleon correlations in the carbon nucleus are needed to kinematically permit @xmath0 photoproduction with photons with energies below 6 gev . in the language of light - front ( infinite momentum frame ) qcd , photon energies of 4 , 6 , and 8 gev correspond to minimal light - cone fractions @xmath13 , 1.4 , and 1.0 respectively , where @xmath14 . in the limit of very large quark masses
the cross section of onium production in the impulse approximation is expressed through the light - cone density matrix of the nucleus as @xmath3 in eq .
[ eq : s ] is @xmath15 . for the @xmath0 case constraints due to the recoil mass
contribute as well .
the kinematics of the present experiment ( represented by the open circles in fig .
[ fig : pmem ] ) all lie in the range @xmath16 , corresponding to the region of multi - nucleon correlations @xcite .
the recoil effects in these kinematics essentially remove the contribution of two - nucleon short range correlations , as in this case the recoil energy is taken by one nucleon ( this is the kinematic domain which was studied in recent jlab experiments using the @xmath17 reaction @xcite and inclusive scattering @xcite at @xmath18 ) .
the dominant contribution originates from the region where at least two nucleons balance the struck nucleon momentum , i.e. three or more nucleon correlations , which were studied so far only in @xmath19 measurements @xcite . as a function of the negative @xmath2 component of missing momentum @xmath11 , for 4 , 5 , 6 , and 8 gev photons , for the case of quasi - free photoproduction of @xmath0 mesons from a nucleon bound in a nucleus .
the circles represent the relative distribution of events predicted for the present experiment for the specific model described in the text [ although no events were actually observed ] .
the values of light cone fraction @xmath20 are indicated for each photon energy.,width=384 ] the high @xmath1 momentum region is where few - nucleon short - range correlations are known to be important @xcite , potentially leading to significantly enhanced yields compared to a simple quasi - free model . in the hard - scattering picture
@xcite , this could correspond to strong contributions of twist 3-gluon exchange , compared to the minimal 2-gluon exchange needed to ensure the @xmath0 color singlet final state .
the influence of intrinsic charm contributions , or hidden color contributions , could potentially increase the cross section by an order of magnitude above the expectations of a quasi - free model ( see fig .
5 of ref .
another possibility to enhance the sub - threshold cross section is a diagram in which two gluons are exchanged to two different nucleons .
this type of process could become important in the sub - threshold region , because each of the gluons could have a much lower momentum fraction than if the pair of gluons came from a single nucleon . in @xmath21 collisions
, it has been observed that anti - protons and kaons are produced on nuclear targets at substantially lower energies than is kinematically possible on free nucleons @xcite .
scattering on a single nucleon in the nucleus would , at these energies , require a single - nucleon momentum in the vicinity of 800 mev .
while the @xmath21 data can be fit assuming such high momenta are sufficiently likely , this assumption leads to an underestimate of sub - threshold production in @xmath22 collisions by about three orders of magnitude @xcite .
there are at least two qualitatively different scenarios for the observed sub - threshold production of anti - protons @xcite .
one scenario is that the projectile strikes a local hot spot " with a high energy density in the nucleus .
the effective mass of the scatterer is high , lowering the kinematic threshold .
alternatively , the momentum required to create the anti - proton is not transferred locally , but picked up in an extended longitudinal region .
establishing either scenario would teach us something qualitatively new about rare , highly excited modes of the nucleus .
sub - threshold photoproduction can help to identify the correct mechanism , because the @xmath23 component of the photon is almost point - like at the charm threshold and below .
effects due to the shrinking effective size of a hadron probe near threshold are eliminated .
the @xmath23 pair is created locally , within a proper time @xmath24 .
the extended acceleration scenario is thus not effective for charm photoproduction .
if significant sub - threshold charm photoproduction occurs ( beyond what can be ascribed to the quasi - free model ) this selects the hot spot scenario .
to model quasi - free photoproduction from a heavy nucleus , we use a convolution integral similar to the model for the deuteron described in ref .
@xcite @xmath25 where the integral over nucleon momentum @xmath1 and energy @xmath8 is limited to the kinematically allowed region , as described above , and the photon - nucleon center of mass energy squared @xmath3 and momentum transfered @xmath26 are functions of @xmath8 , @xmath1 , and photon energy @xmath9 .
the elementary amplitude is far off shell ( @xmath27 for the interacting nucleon is @xmath28 ) , which may lead of a significant suppression of the cross section , not taken into account in the present analysis .
the photon flux @xmath29 is given by @xmath30 , where @xmath31 in the present experiment is the sum of the target thickness in radiation lengths divided by two , and an effective quasi - real electroproduction factor @xcite of 0.02 .
the cross section @xmath32 is for @xmath0 photoproduction from a free nucleon , and we assume for simplicity that neutron and proton cross sections are equal . the flux term @xcite @xmath33 averages to about 1.7 for the kinematics of this experiment ( see fig . [
fig : pmem ] ) .
the function @xmath34 is a carbon spectral function @xcite .
the spectral function is defined as the probability of finding a nucleon of momentum @xmath35 and separation energy @xmath8 in the nucleus . in this picture , the only unknowns are the model for @xmath32 and the carbon spectral function at high @xmath1 and @xmath8 .
there are no published data on sub - threshold photoproduction , but it is useful to review the existing data just above threshold on a free nucleon , as this provides the baseline for sub - threshold predictions . the existing data below 20 gev come from cornell @xcite using 9.3 to 11.8 gev photons , and from slac @xcite from 13 to 21 gev .
these experiments detected lepton pairs from the @xmath0 decay to provide relatively background - free measurements .
additional unpublished data from slac @xcite detected only a single lepton , leading to relatively large background subtractions .
the experiments typically parameterize the data according to @xmath36 .
what is remarkable is that , while @xmath37 has values of 3 to 5 gev@xmath38 at high energy , which are characteristic of diffractive processes , the values drop rapidly near threshold , with cornell @xcite quoting a value of only @xmath39 gev@xmath38 for an 11 gev photon energy .
it was pointed out by ref .
@xcite , the actual slope of the data seems to be more like 1.5 gev@xmath38 , still quite small .
the cornell value is more than a factor of two below the slac value of @xmath40 gev@xmath38 at 19 gev .
it is difficult to reconcile the two experiments with a smooth fit , assuming the exponential form corresponds to an effective form factor .
one way to resolve this is to assume that @xmath41 scales as a dipole form factor squared of the form @xmath42 @xcite .
a reasonably good fit to all data up to photon energies of 100 gev can be found with @xmath43 gev@xmath5 . since each experiment measured over a limited range of @xmath26 , and @xmath44 increases near threshold ,
a natural explanation for the variation of @xmath37 with photon energy can be found .
we found a reasonable dipole fit to the low energy data is given by @xmath45
nb / gev@xmath5 , where @xmath26 is in units of gev@xmath5 . in the qcd picture of laget
@xcite , this almost flat dependence on photon energy corresponds to three - gluon exchange dominance , while 2-gluon exchange should have an extra factor of @xmath46 , where @xmath47 is the momentum fraction . in the definition of ref .
@xcite , @xmath48 is unity for at the threshold for photoproduction . in the approach of ref .
@xcite , a different definition of @xmath47 is used which yields a maximum value of about 0.8 , so factors such as @xmath49 do not cause a large threshold suppression . in summary
, there is a very large uncertainty in how to extend fits to existing free - nucleon cross section measurements below 11 gev .
this is discussed in more detail below in the context of cross section limits from the present experiment .
the experiment was performed in 2004 in hall c at the thomas jefferson national accelerator facility ( jlab ) .
the layout of the spectrometers and detectors is indicated in fig .
[ exp_setup ] . the @xmath0 search data taking used a 5.76 gev continuous - wave ( cw ) electron beam with a typical current of 60 @xmath50a , and lasted for eight days .
the integrated beam charge on target was 27 coulombs .
the electron beam impinged on a narrow solid @xmath51c target with a thickness ( in the beam direction ) of 2.5 gm / cm@xmath5 , or 0.06 radiation length ( r.l . ) .
the effective photon flux per electron was @xmath53 , where @xmath9 is the photon energy , obtained by considering that on average half of the real photons produced in the target are usable , plus an effective 2% from small - angle electroproduction .
the experiment used the high momentum spectrometer ( hms ) in coincidence with the short orbit spectrometer ( sos ) to measure lepton pairs from @xmath0 decays ( 6% branching ratio for each of @xmath54 and @xmath55 ) .
the magnets of the hms were configured to detect positive charged particles while the sos had negative polarity .
the hms central momentum was set at 3.5 gev/@xmath56 and the central scattering angle at 24 degrees .
the sos had a central momentum set near its maximum value of 1.7 gev/@xmath56 and the central angle set to 53 degrees .
these settings were chosen to optimize the acceptance for forward - angle @xmath0 mesons decaying to lepton pairs , with the assumption that the angular distribution of the lepton pairs is given by @xmath57 @xcite , where @xmath58 is the center - of - mass ( c.m . )
decay angle .
the optimization was done assuming the elementary cross section model had the form @xmath45 nb / gev@xmath5 and the spectral function of benhar @xcite ( with one of the high @xmath11 extrapolations discussed below ) to generate a sample of quasi - free @xmath0 mesons .
the choice of spectrometer polarities was made to minimize backgrounds and random coincidences .
the cw nature of the electron beam was essential to reducing accidental coincidences to an acceptable level .
electrons in the sos and positrons in the hms were identified by requiring a signal of several photoelectrons in threshold gas cherenkov counters , with the threshold momentum for pions or heavier particles to produce cherenkov light set above the maximum accepted particle momentum in the given spectrometer .
in addition , at least 70% ( 90% ) of the particle energy was required to be deposited in an array of lead glass blocks in the sos ( hms ) .
these cuts reduced contamination from other particles to a negligible level , while preserving an efficiency of over 90% of electron - positron pairs .
separating muons from pions was more problematic . in the case of the hms spectrometer ,
the pion threshold in the gas cherenkov counter was set just below the maximum accepted momentum ( i.e. , about 8% above the central momentum ) .
the threshold for muons was 25% lower than for pions , and therefore 17% below the central momentum of the spectrometer ( which has a nominal momentum acceptance of @xmath59 ) . with approximately 10 photo - electrons ( pe ) for fully relativistic particles ( such as electrons )
, the muons produced on average 2 pe to 4 pe over the spectrometer acceptance . with a threshold of 1 pe
, the efficiency was varied from 85% to 98% , depending on muon momentum .
the original cherenkov counter in the sos was designed to be sub - atmospheric , hence could not achieve the correct index of refraction to trigger on muons .
it was therefore replaced with the slac 1.6 gev spectrometer cherenkov counter modified for the scattered particle envelope in the sos .
this detector was filled with c@xmath60f@xmath61 and was pressurized to trigger on muons , but not on pions .
additional muon identification was provided by requiring minimum ionizing pulse heights in each of the four layers of the lead glass shower counters .
the overall efficiency for muon pairs was estimated to be 80% .
drift chambers in each spectrometer were used to measure particle momenta with a resolution of better than 0.2% , and scattering angles with an accuracy of 1 to 3 mr . the resulting resolution on the di - lepton mass is approximately 10 mev .
scintillator paddles in each spectrometer were used for triggering , background rejection , and additional particle identification .
the last two features are illustrated in fig .
[ fig : coin ] , which shows the number of events as a function of @xmath62 , the time difference between the sos and hms relative to that expected for di - lepton pairs ( the time difference between electron - positron and di - muon pairs is negligible on the scale used ) .
the top panel ( fig .
[ fig : coin]a ) is for the case of a cleanly identified electron in the sos , and all events in the hms ( dominated by protons and pions ) .
the peak near zero is due to pions , and the peak near -4 ns is due to protons .
the observed peak widths are about 0.5 ns ( 1 @xmath63 ) .
the random accidental background is very small . the electron - positron time difference spectrum is plotted in fig .
[ fig : coin]b .
only one event is observed in the cut region @xmath64 ns ( illustrated as the vertical dashed lines ) , and there are no events in the electron - proton or accidental coincidence regions . the di - muon time difference spectrum is plotted in fig .
[ fig : coin]c .
only one event is observed in the cut region @xmath64 ns but in this case there are 26 events in the lepton - proton peak region .
the latter are due to protons that produced a signal in the cherenkov counter ( either from knock - on electrons , or from scintillation light ) .
a few accidental - in - time events are also visible . after putting a cut around the time - difference peaks shown in fig .
[ fig : coin ] , the reconstructed mass spectra using the hms and sos particles ( @xmath65 ) was determined and are shown in fig . [ fig : mass ] . for electron - pion coincidences , shown in fig . [ fig : mass]a , the mass spectrum is smooth , and covers the mass range @xmath66 gev .
this basically illustrates the phase space acceptance of the two spectrometers , and shows good acceptance at the @xmath0 mass of 3.097 gev . as illustrated in fig .
[ fig : mass]b , the one coincident @xmath67 event has an invariant mass of only @xmath68 gev , more than 40 @xmath63 from the @xmath0 mass .
thus this event is background : either an electron - pion event ( with the pion mis - identified as a positron ) , or a wide - angle pair conversion of a bremsstrahlung photon . as illustrated in fig . [
fig : mass]c , the one coincident @xmath69 event has an invariant mass of only @xmath70 gev , more than 30 @xmath63 from the @xmath0 mass .
thus this event is also background : mostly likely an accidental coincidence , but it could also be a wide - angle pair conversion of a bremsstrahlung photon or a process involving particle mis - identification , or the decay of pions to muons .
mass.,width=240 ] in order to facilitate the calculation of the predicted number of events for a given sub - threshold cross section model , we give here the @xmath0 acceptance corresponding to the hms and sos setting used in this experiment .
this was determined by a detailed monte carlo simulation of the spectrometers , convoluted with the di - lepton decays of @xmath0 mesons with a particular momentum and polar angle with respect to the beam axis .
the results are shown in table [ tab : acc ] , and indicate a maximum detection probability of about @xmath71 for momenta near 4.3 gev and angles less than 3 degrees . [ cols= " > , > , > , > , > , > , > , > , > , > " , ] [ accept ] several calibration runs were performed to check the kinematic settings and verify that the spectrometers were fully functional .
the first calibration check was to set up the spectrometers for coincident @xmath72 running where @xmath73 could be either a positively - charged pion , proton , or kaon .
a liquid hydrogen target replaced the carbon target and the hms was moved forward to 13 degrees .
three sos angles were used , @xmath74 , @xmath75 , and @xmath76 .
these central kinematics corresponded to four momentum transfers of @xmath77 7.8 , 4.6 , and 2.1 gev@xmath5 , and invariant mass @xmath78 , 2.0 , and 2.5 gev , respectively . in all cases , the peaks in the missing mass spectra were at the proper positions and the widths were understood .
as an example , using proton cuts on the hms arm , the missing mass spectrum for @xmath79 shows prominent @xmath80 and @xmath81 peaks at the correct locations ( as seen in previous hall c experiments @xcite ) . as a further check of the luminosity
, accidental and multi - pion continuum backgrounds were subtracted from the data to produce experimental yields .
these yields were corrected for radiative effects and spectrometer acceptances to produce differential cross sections .
good agreement was found between the measured and simulated differential cross sections @xcite .
this indicates that the target density and electron beam current determinations were well understood .
the other significant calibration check was to set the spectrometers to measure lepton pair decays from other lower mass vector mesons .
in particular , we centered the spectrometers to detect @xmath82 pairs from @xmath83 and @xmath84 meson decays . for these runs
a beryllium target was used with the sos set to 25 degrees and a central momentum of 0.9 gev , and the hms set to 11 degrees and a central momentum of 1.95 gev . in spite of the very small branching ratios of about @xmath85 ,
a clear @xmath86 peak with the expected full width of 8 mev was seen on top of a 140 mev wide @xmath87 peak , as illustrated in fig .
[ fig : om ] .
the number of detected @xmath83 events is consistent with the predicted number from a monte carlo simulation that had as much as possible in common with the simulation of @xmath0 events .
( indicated by arrow ) and @xmath88.,width=240 ]
since no @xmath0 mesons were detected in this experiment , here we try to quantify what the implications are for elementary photoproduction cross section models , in the context of the single - particle convolution model described above , for four choices of the prescription used to extend the spectral function of benhar @xcite beyond @xmath89 gev ( where there is insufficient data from @xmath90 to constrain the function ) .
we show two `` eyeball '' extrapolations of the probability distribution integrated over @xmath8 in fig .
[ fig : sf ] .
we label this as `` high '' and `` low '' according whether they are are higher or lower at large @xmath11 .
as it happens , these extrapolations are quite similar to the two extrapolations ( for nuclear matter ) shown in fig . 1 of ref . @xcite . as discussed in ref .
@xcite , the `` high '' extrapolation should probably be considered as an upper limit , and probably will over - estimate yield predictions .
we used two choices for the @xmath8 distribution : a ) `` freeze '' the @xmath8 distribution to that of benhar et al .
@xcite at @xmath91 gev ( the highest provided in this fit ) ; b ) shift the @xmath8 distribution by @xmath92 , with @xmath11 in units of gev .
the latter corresponds to following the `` ridge '' observed at lower @xmath8 .
using the nuclear matter prescription of eq .
10 in ref .
@xcite should give results somewhere between these two cases , as their prescription involves shifting the @xmath8 distribution along the same `` ridge '' , but having the width increase with increasing @xmath11 .
following ref .
@xcite , we also investigated scaling the spectral function by a relativistic correction factor @xmath93 and found the results to change by less than 20% .
we used three different free nucleon cross sections ( as motivated by the discussion in the introduction ) : @xmath94 where @xmath31 and @xmath37 are free parameters , and we used @xmath48 for each model , we varied the @xmath26-slope parameter @xmath37 within a reasonable range and , for each value of @xmath37 , determined @xmath31 such that the total cross section would agree with the cornell measurement @xcite of 0.7 nb at @xmath95 gev .
the predicted counts are shown in fig .
[ fig : limits ] as a function of @xmath37 for each of the three models for four high-@xmath11 extrapolations of the spectral function .
the predictions for model iii are quite a bit lower than models i and ii , due to the factor of @xmath49 .
the sensitivity to the spectral function extrapolation is also largest for model iii , because on average higher values of @xmath11 are probed . for all models ,
`` shifting '' the @xmath8 distributions for @xmath96 gev makes a difference of typically a factor of two in predicted rates .
not shown in the figure is the effect of assuming @xmath97 ( i.e. , a virtual photon interacting with an almost on - shell nucleon ) , as was initially assumed in the planning phase of the present experiment .
this assumption results in predicted rates higher by approximately two orders of magnitude than if a more realistic @xmath8 distribution is used . while the predicted number of counts is below our 87% confidence level observation of less than 2 counts ( shown as the hatched bands in fig .
[ fig : limits ] ) for all the combinations considered here , it is not impossible that for some cross sections models ( i and ii in particular ) , something like the `` hot spot '' scenario discussed in the introduction could result in a prediction of more than two counts . in the quasi - free picture , this could correspond to considerably lower average values of @xmath8 than in the benhar spectral function .
while we are not aware of any detailed theoretical predictions for sub - threshold @xmath0 photoproduction specifically , we can use the calculation of charm ( @xmath98 ) photoproduction of braun and vlahovic @xcite as a possible guide . using pqcd to evaluate the photon - gluon fusion process
, they predict a total cross section of 0.25 fb / nucleon for carbon for 5.5 gev photons .
this corresponds to approximately 0.02 events for our experimental conditions .
this is similar to our predictions using model iii .
the non - observation of sub - threshold @xmath0 photoproduction on carbon in the present experiment is consistent with predictions of quasi - free production with a variety of reasonable elementary free nucleon cross sections models near threshold , and educated guesses for the high missing momentum and missing energy spectral function distributions in carbon . for a given set of assumptions on the cross section and spectral function choices , upper limits
could be set on exotic mechanisms that could potentially enhance the sub - threshold cross sections ( such as gluon exchange to two different nucleons , hidden color configurations , 3-gluon exchange , etc . ) .
the interpretation of the present experiment will be greatly aided with precision measurement of the elementary free nucleon cross section near threshold , planned at jefferson lab @xcite once beam energies up to 11 gev are available . the authors wish to thank j.m .
laget for theoretical support .
this work is supported in part by research grants from the u.s .
department of energy and the u.s . national science foundation .
the south african group was acknowledges the support of the national research foundation .
the southeastern universities research association operates the thomas jefferson national accelerator facility under the u.s .
department of energy contract deac05 - 84er40150 . | a search was made for sub - threshold @xmath0 production from a carbon target using a mixed real and quasi - real bremsstrahlung photon beam with an endpoint energy of 5.76 gev .
no events were observed , which is consistent with predictions assuming quasi - free production .
the results place limits on exotic mechanisms that strongly enhance quasi - free production . |
we begin with a brief description of weighted projective lines and their categories of coherent sheaves .
let @xmath0 be an algebraically closed field , let @xmath1 be the projective line over @xmath0 , let @xmath2 be a ( possibly empty ) collection of distinct closed points of @xmath1 , and let @xmath3 be a _ weight sequence _ , that is , a sequence of positive integers .
the triple @xmath4 is called a _ weighted projective line_. geigle and lenzing @xcite have associated to each weighted projective line a category @xmath5 of coherent sheaves on @xmath6 , which is the quotient category of the category of finitely generated @xmath7-graded @xmath8-modules , modulo the serre subcategory of finite length modules . here
@xmath7 is the rank 1 additive group @xmath9 and @xmath10 / ( x_i^{p_i } + \la_{i1 } u - \la_{i0 } v ) , \ ] ] with grading @xmath11 and @xmath12 , where @xmath13 $ ] in @xmath1 .
geigle and lenzing showed that @xmath5 is a hereditary abelian category with finite dimensional hom and ext spaces .
the free module @xmath8 yields a structure sheaf @xmath14 , and shifting the grading gives twists @xmath15 for any sheaf @xmath16 and @xmath17 .
every sheaf is the direct sum of a torsion - free sheaf and a finite length sheaf .
a torsion - free sheaf has a finite filtration by line bundles , that is , sheaves of the form @xmath18 .
the finite length sheaves are easily described as follows .
there are simple sheaves @xmath19 ( @xmath20 ) and @xmath21 ( @xmath22 , @xmath23 ) satisfying for any @xmath24 that @xmath25 if and only if @xmath26 , and the only extensions between them are @xmath27 for each simple sheaf @xmath28 and @xmath29 there is a unique sheaf with length @xmath30 and top @xmath28 , which is _ uniserial _ , meaning that it has a unique composition series .
these are all the finite length indecomposable sheaves .
categories of the form @xmath5 for some weighted projective line @xmath6 play a special role in the study of abelian categories .
this follows from a theorem of happel @xcite which we now explain .
consider a connected hereditary abelian category @xmath31 that is @xmath0-linear with finite dimensional hom and ext spaces .
suppose in addition that @xmath31 admits a tilting object , that is some object @xmath32 with @xmath33 such that @xmath34 and @xmath35 imply @xmath36 .
thus the functor @xmath37 into the category of modules over the endomorphism algebra @xmath38 induces an equivalence @xmath39 of derived categories .
there are two important classes of such hereditary abelian categories admitting a tilting object : module categories over path algebras of finite connected quivers without oriented cycles , and categories of coherent sheaves on weighted projective lines .
happel s theorem then states that there are no further classes .
more precisely , an abelian category @xmath31 as above is , up to a derived equivalence , either of the form @xmath40 for some finite connected quiver @xmath41 without oriented cycles or of the form @xmath5 for some weighted projective line @xmath6 .
the following treatment of coherent sheaves on weighted projective lines is based on a list of axioms ( extending the list in happel s theorem ) which we postpone until [ se : cohx ] . before that we discuss in some detail the necessary background material : abelian categories , derived categories , tilting theory , expansions of abelian categories , and coherent sheaves on @xmath1 .
@xmath42\\ \s1\ar[rrd]\ar[r]&\s2\ar[r]\ar[ru]&\s4\ar[r]&\s6\ar[r]&\s7\\ & & \s5\ar[ru ] } \\ \text{\centerline{\sc leitfaden}}\end{gathered}\ ] ] these notes are based on a seminar and a course held at the university of paderborn in the first half of 2009 . both authors wish to thank the participants for their interest and for stimulating discussions related to the topic of these notes .
in particular , we are grateful to dirk kussin and helmut lenzing for their advice .
it is a pleasure to thank marco angel bertani - kland and david ploog for their detailed comments .
most of the material presented here is taken from the existing literature .
an exception is [ se : extensions ] , where the concept of an ` expansion of abelian categories ' is introduced .
the previously unpublished proof of theorem [ th : gabriel ] is due to yu ye .
a category @xmath31 is _ additive _ if every finite family of objects has a product , each morphism set @xmath43 is an abelian group , and the composition maps @xmath44 are bilinear . given a finite number of objects @xmath45 of an additive category @xmath31 , there exists a _
direct sum _
@xmath46 , which is by definition an object @xmath47 together with morphisms @xmath48 and @xmath49 for @xmath50 such that @xmath51 , @xmath52 , and @xmath53 for all @xmath54 .
note that the morphisms @xmath55 and @xmath56 induce isomorphisms @xmath57 given any object @xmath47 in @xmath31 , we denote by @xmath58 the full subcategory of @xmath31 consisting of all finite direct sums of copies of @xmath47 and their direct summands . a _ decomposition _
@xmath59 of an additive category @xmath31 is a pair of full additive subcategories @xmath60 and @xmath61 such that each object in @xmath31 is a direct sum of two objects from @xmath60 and @xmath61 , and @xmath62 for all @xmath63 and @xmath64 .
an additive category @xmath31 is _ connected _ if it admits no proper decomposition @xmath59 .
a functor @xmath65 between additive categories is _ additive _ if the induced map @xmath66 is linear for each pair of objects @xmath67 in @xmath31 .
the _ kernel _ @xmath68 of an additive functor @xmath65 is by definition the full subcategory of @xmath31 formed by all objects @xmath47 such that @xmath69 .
the _ essential image _
@xmath70 of @xmath65 is the full subcategory of @xmath71 formed by all objects @xmath72 such that @xmath72 is isomorphic to @xmath73 for some @xmath47 in @xmath31 .
an additive category @xmath31 is _ abelian _ if every morphism @xmath74 has a kernel and a cokernel , and if the canonical factorization @xmath75 ^ -{\p'}&a\ar[r]^-\p\ar[d ] & b\ar[r]^-{\p''}&\operatorname{coker}\p\\ & \operatorname{coker}\p'\ar[r]^-{\bar\p}&\operatorname{ker}\p''\ar[u]}\ ] ] of @xmath76 induces an isomorphism @xmath77 . given an abelian category @xmath31 , a finite sequence of morphisms @xmath78 in @xmath31 is _ exact _ if @xmath79 for all @xmath80 . an additive functor @xmath65 between abelian categories is _ exact _
if @xmath81 sends each exact sequence in @xmath31 to an exact sequence in @xmath71 .
\(1 ) let @xmath82 be a right noetherian ring .
the category @xmath83 of finitely generated right modules over @xmath82 is an abelian category .
\(2 ) let @xmath0 be a field and @xmath41 a quiver .
the category @xmath84 of finite dimensional @xmath0-linear representations of @xmath41 is an abelian category . throughout ,
all categories are supposed to be _ skeletally small _ , unless otherwise stated .
this means that the isomorphism classes of objects form a set .
subcategories are usually full subcategories and closed under isomorphisms
. functors between additive categories are always assumed to be additive .
the composition of morphisms is written from right to left , and modules over a ring are usually right modules .
let @xmath85 be an abelian category .
a non - empty full subcategory @xmath86 of @xmath31 is called a _ serre subcategory _ provided that @xmath86 is closed under taking subobjects , quotients and extensions .
this means that for every exact sequence @xmath87 in @xmath31 , the object @xmath47 belongs to @xmath86 if and only if @xmath88 and @xmath89 belong to @xmath86 .
the kernel of an exact functor @xmath90 between abelian categories is a serre subcategory of @xmath31 . given a serre subcategory @xmath86 of @xmath31 , the _ quotient category _
@xmath91 of @xmath31 with respect to @xmath86 is defined as follows .
the objects in @xmath91 are the objects in @xmath31 . given two objects @xmath67 in @xmath31
, there is for each pair of subobjects @xmath92 and @xmath93 an induced map @xmath94 .
the pairs @xmath95 such that both @xmath96 and @xmath97 lie in @xmath86 form a directed set , and one obtains a direct system of abelian groups @xmath98 .
we define @xmath99 and the composition of morphisms in @xmath31 induces the composition in @xmath91 .
is a category , in particular that the composition of morphisms is associative .
this requires some work ; see @xcite . ]
the _ quotient functor _ @xmath100 is by definition the identity on objects .
the functor takes a morphism in @xmath101 to its image under the canonical map @xmath102 .
[ le : mor ] each morphism @xmath103 in @xmath91 is of the form @xmath104 for some pair @xmath95 of subobjects with @xmath96 and @xmath97 in @xmath86 and some morphism @xmath105 in @xmath31 , where @xmath106 and @xmath107 denote the canonical morphisms in @xmath31 . for each morphism @xmath103 in @xmath91
, there is by definition a pair @xmath95 of subobjects and a morphism @xmath105 in @xmath31 such that the following diagram commutes .
@xmath108&b\ar[d]^{q\pi}\\ a'\ar[u]^{q\iota}\ar[r]^-{q\p}&b / b'}\ ] ] now observe that for each object @xmath109 in @xmath31 the inclusion @xmath110 induces a bijection @xmath111 .
thus @xmath112 is invertible .
analogously , one shows that @xmath113 is invertible .
the following result summarizes the basic properties of a quotient category and the corresponding quotient functor .
[ pr : quotient ] let @xmath31 be an abelian category and @xmath86 a serre subcategory . 1 .
the category @xmath114 is abelian and the quotient functor @xmath115 is exact with kernel @xmath116 .
2 . let @xmath65 be an exact functor between abelian categories .
if @xmath117 , then there is a unique functor @xmath118 such that @xmath119 .
moreover , the functor @xmath120 is exact .
\(1 ) it follows from the construction that the morphism sets of the quotient category @xmath91 are abelian groups .
also , the quotient functor induces linear maps between the morphism sets and it preserves finite direct sums . thus the quotient category and the quotient functor are both additive .
the quotient functor sends a morphism in @xmath31 to the zero morphism if and only if its image belongs to @xmath86 .
thus @xmath116 .
let @xmath121 be a morphism in @xmath91 as in . denote by @xmath122 the kernel and by @xmath123 the cokernel of @xmath76 in @xmath31 . then the kernel of @xmath124 is @xmath125 , whereas the cokernel of @xmath124 is @xmath126 .
it follows that the category @xmath91 is abelian and that the quotient functor preserves kernels and cokernels .
\(2 ) the functor @xmath118 takes an object @xmath47 to @xmath73 and a morphism of the form @xmath127 as in to @xmath128 .
note that @xmath129 and @xmath130 are isomorphisms in @xmath71 , since @xmath81 is exact and @xmath117 .
the functor @xmath120 is additive and the description of ( co)kernels in ( 1 ) shows that @xmath120 preserves ( co)kernels .
thus @xmath120 is exact .
the quotient functor @xmath131 is the universal functor that inverts the class @xmath132 of morphisms @xmath133 in @xmath31 with @xmath134 and @xmath135 in @xmath86 .
more precisely , for any class @xmath28 of morphisms in @xmath31 , there exists a universal functor @xmath136 $ ] such that 1 .
the morphism @xmath137 is invertible for every @xmath138 , and 2 .
every functor @xmath139 such that @xmath140 is invertible for each @xmath138 admits a unique functor @xmath141\to \b$ ] such that @xmath142 .
the category @xmath143 $ ] is the _ localization _ of @xmath31 with respect to @xmath28 and is unique up to a unique isomorphism ; see ( * ? ? ?
* i.1 ) .
[ le : serre ] let @xmath86 be a serre subcategory of @xmath31 .
the quotient functor @xmath115 is the universal functor that inverts all morphisms in @xmath132 .
therefore @xmath144=\a/\c.\ ] ] from proposition [ pr : quotient ] it follows that @xmath145 inverts all morphisms in @xmath132 .
now let @xmath139 be a functor such that @xmath140 is invertible for each @xmath146 .
then for each pair @xmath67 of objects in @xmath31 and each pair of subobjects @xmath92 and @xmath147 with @xmath96 and @xmath97 in @xmath86 , the map @xmath66 factors through the canonical map @xmath94 .
thus there are induced maps @xmath148 which induce a unique functor @xmath149 such that @xmath150 .
it follows that @xmath151=\a/\c$ ] .
\(1 ) let @xmath82 be a commutative noetherian ring and @xmath152 the localization with respect to a prime ideal @xmath153 .
the localization functor @xmath154 sending a @xmath82-module @xmath155 to @xmath156 is exact and induces an equivalence @xmath157 . roughly speaking ,
restriction of scalars along the morphism @xmath158 yields a quasi - inverse .
\(2 ) let @xmath82 be a right noetherian ring and @xmath159 an idempotent .
the functor @xmath160 sending a @xmath82-module @xmath155 to @xmath161 is exact .
the kernel @xmath162 identifies with @xmath163 and @xmath32 induces an equivalence @xmath164 .
the functor @xmath165 yields a quasi - inverse .
\(3 ) let @xmath82 be a right artinian ring .
given a set @xmath166 of simple @xmath82-modules , the @xmath82-modules @xmath155 having a finite filtration @xmath167 with each factor @xmath168 isomorphic to one of the simples @xmath166 form a serre subcategory of @xmath83 . moreover , each serre subcategory of @xmath83 arises in this way and is therefore of the form @xmath163 for some idempotent @xmath169 .
we collect some further properties of abelian quotient categories . [
le : small ] let @xmath31 be an abelian category that is not supposed to be skeletally small , and let @xmath86 be a serre subcategory . then the following are equivalent : 1 . the category @xmath31 is skeletally small .
the categories @xmath86 and @xmath91 are skeletally small .
in addition , @xmath170 and @xmath171 are sets for all @xmath172 and @xmath173 .
one direction is clear .
so suppose that @xmath86 and @xmath91 are skeletally small , and that extensions with objects in @xmath86 form sets .
first observe that the morphisms in @xmath174 form a set for each pair of objects @xmath67 .
here one uses that the subobjects @xmath92 with @xmath88 or @xmath96 in @xmath86 form , up to isomorphism , a set , since @xmath86 is skeletally small .
next observe that for each object @xmath47 in @xmath31 , there is only a set of isomorphism classes of objects @xmath72 with @xmath175 in @xmath91 .
this follows from the fact that each isomorphism @xmath103 in @xmath91 is represented by a chain @xmath176 of epis and monos in @xmath31 with kernel and cokernel in @xmath86 ; see lemma [ le : mor ] . here
one uses that @xmath86 is skeletally small and that extensions with objects in @xmath86 form sets . from this
it follows that the isomorphism classes of objects in @xmath31 form a set , since the quotient @xmath91 has this property .
the following example gives an abelian category @xmath31 with a serre subcategory @xmath86 such that @xmath86 and @xmath91 are skeletally small but @xmath31 itself is not .
let @xmath0 be a field and @xmath41 a quiver with set of vertices @xmath177 and a proper class of arrows @xmath178 .
each arrow of @xmath41 corresponds to a canonical element of @xmath179 , where @xmath180 denotes the simple representation supported at the vertex @xmath181 .
these extensions are linearly independent and yield pairwise non - isomorphic two - dimensional representations .
the functor @xmath182 sending a representation of @xmath41 to the corresponding vector space at vertex @xmath183 induces an equivalence @xmath184 , and @xmath162 is equivalent to @xmath185 .
an abelian category is called _ noetherian _ if each of its objects is _ noetherian _ ( i.e.satisfies the ascending chain condition on subobjects ) . [
le : noetherian ] let @xmath31 be an abelian category and @xmath86 a serre subcategory .
then the following are equivalent : 1 .
the category @xmath31 is noetherian .
the categories @xmath86 and @xmath91 are noetherian , and each object in @xmath31 has a largest subobject that belongs to @xmath86 .
\(1 ) @xmath186 ( 2 ) : suppose @xmath31 is noetherian
. then @xmath86 is noetherian .
also , @xmath91 is noetherian because each ascending chain of subobjects in @xmath91 can be represented by an ascending chain of subobjects in @xmath31 ; see lemma [ le : mor ] .
noetherianess implies that each non - empty set of subobjects has a maximal element .
in particular , each object has a subobject that is maximal among all subobjects belonging to @xmath86 .
\(2 ) @xmath186 ( 1 ) : let @xmath187 be an ascending chain of subobjects in @xmath31 . using that @xmath91 is noetherian , there exists some integer @xmath188 such that @xmath189 belongs to @xmath86 for all @xmath190 .
let @xmath191 be the maximal subobject of @xmath192 belonging to @xmath86 .
then the chain @xmath193 becomes stationary since @xmath86 is noetherian .
it follows that the original chain of subobjects of @xmath47 becomes stationary .
thus @xmath47 is noetherian .
next we give an example of an abelian category @xmath31 with a serre subcategory @xmath86 such that @xmath86 and @xmath91 are noetherian but @xmath31 itself is not .
the ring @xmath194}$ ] is well known to be left but not right noetherian .
consider the abelian category @xmath83 of finitely presented ( right ) @xmath82-modules and the functor @xmath195 sending a @xmath82-module @xmath155 to @xmath196 with @xmath197}$ ] . then @xmath162 is equivalent to @xmath198 and @xmath32 induces an equivalence @xmath199 .
the next lemma provides the analogue of a noether isomorphism for abelian categories .
the proof is straightforward .
[ le : noeth ] let @xmath31 be an abelian category and @xmath200 a pair of serre subcategories such that @xmath201 . then the following holds : 1 .
the inclusion @xmath202 identifies @xmath203 with a serre subcategory of @xmath204 .
2 . the quotient functor @xmath205 induces an isomorphism @xmath206 recall that a non - zero object @xmath28 of an abelian category is _ simple _ if @xmath28 has no proper subobject @xmath207 .
the next lemma says that a quotient functor preserves this property if the object is not annihilated .
the proof is straightforward .
[ le : quotient_simple ] let @xmath31 be an abelian category and @xmath86 a serre subcategory . if @xmath28 is a simple object not belonging to @xmath86 , then @xmath28 is simple in @xmath91 and the quotient functor induces an isomorphism @xmath208 .
let @xmath31 be an abelian category .
in some cases , the quotient functor @xmath209 with respect to a serre subcategory @xmath86 admits a right adjoint .
then the perpendicular category @xmath210 provides another description of the quotient category @xmath91 . for any class @xmath86 of objects in @xmath31 , its _ perpendicular categories _
are by definition the full subcategories @xmath211 [ le : perp ] let @xmath31 be an abelian category and @xmath86 a serre subcategory
. then the following are equivalent for an object @xmath72 in @xmath31 : 1 .
the object @xmath72 belongs to @xmath210 .
2 . the quotient functor induces a bijection @xmath212 for every object @xmath47 in @xmath31 .
the map @xmath213 is bijective for every morphism @xmath133 in @xmath31 with @xmath134 and @xmath135 in @xmath86 .
\(1 ) @xmath186 ( 2 ) : for each pair of subobjects @xmath92 and @xmath93 such that both @xmath96 and @xmath97 lie in @xmath86 , the map @xmath94 is bijective . thus the quotient functor induces a bijection @xmath212 .
\(2 ) @xmath186 ( 3 ) : the quotient functor sends a morphism @xmath133 in @xmath31 with @xmath134 and @xmath135 in @xmath86 to an isomorphism in @xmath91 .
thus @xmath214 is bijective . using the bijections in ( 2 )
it follows that @xmath215 is bijective .
\(3 ) @xmath186 ( 1 ) : let @xmath109 be an object in @xmath86 .
then @xmath216 induces a bijection @xmath213 .
thus @xmath217 .
let @xmath218 an exact sequence in @xmath31 .
the morphism @xmath219 induces a bijection @xmath213 , and therefore @xmath220 splits .
thus @xmath221 .
we need the following elementary lemma about pairs of adjoint functors . [
le : adjoint ] let @xmath65 and @xmath222 be a pair of functors such that @xmath223 is a right adjoint of @xmath81
. denote by @xmath224 the class of morphisms @xmath133 in @xmath31 such that @xmath140 is invertible
. then the following are equivalent : 1 .
the functor @xmath81 induces an equivalence @xmath225{\xrightarrow}{\sim}\b$ ] .
the functor @xmath223 is fully faithful .
the adjunction morphism @xmath226 is invertible for each @xmath227 .
see ( * ? ? ?
* i.1.3 ) .
the next result provides a useful criterion for an exact functor to be a quotient functor .
moreover , it describes the right adjoint of a quotient functor .
[ pr : perp ] let @xmath65 be an exact functor between abelian categories and suppose that @xmath81 admits a right adjoint @xmath222
. then the following are equivalent : 1 .
the functor @xmath81 induces an equivalence @xmath228 .
the functor @xmath81 induces an equivalence @xmath229 .
the functor @xmath223 induces an equivalence @xmath230 .
the functor @xmath223 is fully faithful .
moreover , in that case @xmath231 and @xmath232 .
let @xmath224 denote the class of morphisms @xmath133 in @xmath31 such that @xmath140 is invertible .
then it follows from lemma [ le : serre ] that @xmath225=\a/\operatorname{ker}f$ ] .
\(1 ) @xmath186 ( 2 ) : the functor @xmath81 induces a full and faithful functor @xmath233 by lemma [ le : perp ] . for each @xmath234
, we have @xmath235 by lemma [ le : adjoint ] , and @xmath236 belongs to @xmath237 by lemma [ le : perp ] .
thus @xmath81 induces an equivalence @xmath229 .
\(2 ) @xmath186 ( 3 ) : note that @xmath238 .
thus @xmath223 induces a functor @xmath239 which is a right adjoint of the equivalence @xmath233 .
now one uses that an adjoint of an equivalence is again an equivalence .
\(3 ) @xmath186 ( 4 ) : an equivalence is fully faithful .
\(4 ) @xmath186 ( 1 ) : use lemma [ le : adjoint ] .
observe that ( 3 ) implies @xmath231 .
in particular , @xmath240 .
the other inclusion follows from the isomorphism @xmath241 .
the next result characterizes the fact that the quotient functor admits a right adjoint .
[ pr : section ] let @xmath31 be an abelian category and @xmath86 a serre subcategory
. then the following are equivalent : 1 .
the quotient functor @xmath131 admits a right adjoint @xmath242 .
every object @xmath47 in @xmath31 fits into an exact sequence @xmath243 such that @xmath244 and @xmath245 .
3 . the quotient functor induces an equivalence @xmath246 . in that case
the functor @xmath247 sending @xmath47 to @xmath88 is a right adjoint of the inclusion @xmath248 , and the functor @xmath249 sending @xmath47 to @xmath191 is a left adjoint of the inclusion @xmath250 .
\(1 ) @xmath186 ( 2 ) : we apply proposition [ pr : perp ] .
suppose that the quotient functor @xmath251 admits a right adjoint @xmath223 .
the functor @xmath81 inverts the adjunction morphism @xmath252 , since @xmath253 by lemma [ le : adjoint ] .
the exactness of @xmath81 then implies that @xmath254 and @xmath255 belong to @xmath86 .
the object @xmath191 belongs to @xmath256 by construction .
\(2 ) @xmath186 ( 3 ) : the quotient functor induces a fully faithful functor @xmath257 by lemma [ le : perp ] .
this functor is an equivalence , because each object @xmath47 in @xmath91 is isomorphic to one in its image via the isomorphism @xmath258 .
\(3 ) @xmath186 ( 1 ) : choose a quasi - inverse @xmath259 of the equivalence @xmath260 .
for each @xmath47 in @xmath31 and @xmath72 in @xmath91 , there are bijections @xmath261 the first map is bijective by lemma [ le : perp ] and the second is bijective because @xmath253 .
thus @xmath223 is right adjoint to the quotient functor @xmath131 . for any @xmath109 in @xmath86 ,
the induced map @xmath262 is bijective . therefore sending @xmath47 to @xmath88
provides a right adjoint of the inclusion @xmath248 . on the other hand , for any @xmath72 in @xmath210 ,
the induced map @xmath263 is bijective . therefore sending @xmath47 to @xmath191
provides a left adjoint of the inclusion @xmath250 . the objects @xmath88 and @xmath89 occurring in represent certain functors defined on @xmath86 .
we have @xmath264 where @xmath88 is viewed as a subobject of @xmath47 .
let @xmath31 be an abelian category .
for a pair of objects @xmath67 and @xmath265 , let @xmath266 denote the group of extensions in the sense of yoneda .
set @xmath267 and @xmath268 for @xmath269 .
note that there are composition maps @xmath270 for all @xmath271 .
the _ projective dimension _ of an object @xmath47 is by definition @xmath272 dually , one defines the _ injective dimension _
@xmath273 .
[ lem : hereditary ] let @xmath31 be an abelian category . for @xmath47 in @xmath31 and @xmath274
the following are equivalent : 1 .
@xmath275 is right exact .
2 . @xmath276 .
@xmath277 for all @xmath190 .
4 . @xmath278 .
it suffices to show that ( 1 ) and ( 2 ) are equivalent ; the rest is straightforward .
we use the long exact sequence for @xmath279 .
\(1 ) @xmath186 ( 2 ) : fix an element @xmath280 which is represented by an exact sequence @xmath281 let @xmath109 be the image of @xmath282 and write @xmath283 as the composite of extensions @xmath284 and @xmath285 . for the connecting morphism @xmath286 induced by @xmath287 , we have @xmath288 since @xmath275 is right exact .
thus @xmath289 .
\(2 ) @xmath186 ( 1 ) : clear .
the _ global dimension _ of @xmath31 is by definition the smallest integer @xmath290 such that @xmath291 . as usual ,
the dimension is infinite if such a number @xmath188 does not exist .
we denote this dimension by @xmath292 and observe that it is equal to @xmath293 and @xmath294 .
the category @xmath31 is called _ hereditary _ provided that @xmath295 . for a right noetherian ring @xmath82 , the global dimension of the module category @xmath83
is called the ( right ) _ global dimension _ of @xmath82 and denoted by @xmath296 .
\(1 ) let @xmath82 be the ring of integers @xmath297 or the polynomial ring @xmath298 $ ] over a field @xmath0 .
then @xmath83 is hereditary .
more generally , @xmath83 is hereditary if @xmath82 is a dedekind domain .
\(2 ) for a field @xmath0 and a quiver @xmath41 , the category of representations @xmath84 is hereditary .
\(3 ) let @xmath31 be a hereditary abelian category and @xmath86 a serre subcategory
. then @xmath86 and @xmath91 are again hereditary .
let @xmath31 be an abelian category .
an object @xmath47 of @xmath31 has _ finite length _ if there exists a finite chain of subobjects @xmath299 such that each quotient @xmath300 is a simple object .
such a chain is called a _ composition series _ of @xmath47 .
a composition series is not necessarily unique but its length is an invariant of @xmath47 by the jordan - hlder theorem ; it is called the _ length _ of @xmath47 and is denoted by @xmath301 .
note that an object has finite length if and only if it is both _ artinian _ ( i.e. satisfies the descending chain condition on subobjects ) and _ noetherian _ (
i.e. satisfies the ascending chain condition on subobjects ) .
every object of finite length decomposes essentially uniquely into a finite direct sum of indecomposable objects with local endomorphism rings .
this follows from the krull - remak - schmidt theorem .
the objects of finite length form a serre subcategory of @xmath31 which is denoted by @xmath302 .
the abelian category @xmath31 is called a _ length category _ if @xmath303 .
let @xmath31 be a length category .
the _ ext - quiver _ or _
gabriel quiver _ of @xmath31 is a valued quiver @xmath304 which is defined as follows .
the set @xmath305 of vertices is a fixed set of representatives of the isomorphism classes of simple objects in @xmath31 . for a simple object @xmath28 , let @xmath306 denote its endomorphism ring , which is a division ring .
observe that @xmath307 carries a natural @xmath308-@xmath306-bimodule structure for each pair @xmath309 in @xmath305 .
there is an arrow @xmath310 with valuation @xmath311 in @xmath312 if @xmath313 with @xmath314 and @xmath315 .
we write @xmath316 if @xmath317 .
the following observation is easily proved .
[ le : comp ] a length category is connected if and only if its ext - quiver is connected .
\(1 ) let @xmath82 be a commutative noetherian local ring with maximal ideal @xmath318 .
then @xmath319 equals the category of @xmath318-torsion modules and @xmath320 is the unique simple @xmath82-module .
\(2 ) for a field @xmath0 and a quiver @xmath41 , the category of representations @xmath84 is a length category .
suppose that @xmath41 has no oriented cycle and no pair of parallel arrows .
then the ext - quiver of @xmath84 is isomorphic to @xmath41 , with valuation @xmath321 for each arrow .
let @xmath85 be a length category .
an object @xmath47 is _ uniserial _ provided it has a unique composition series .
note that any non - zero uniserial object is indecomposable .
moreover , subobjects and quotient objects of uniserial objects are uniserial .
the length category @xmath31 is called _ uniserial _ provided that each indecomposable object is uniserial .
the following result characterizes uniserial categories in terms of their ext - quivers .
[ th : gabriel ] a length category @xmath31 is uniserial if and only if for each simple object @xmath28 , we have @xmath322 the proof of this result requires some preparations and we begin with some notation .
let @xmath47 be any object in @xmath31 .
we denote by @xmath323 the intersection of all its maximal subobjects and let @xmath324 .
analogously , @xmath325 denotes the sum of all simple subobjects of @xmath47 .
[ le : gabriel ] let @xmath31 be a length category and suppose that holds for each simple object @xmath28 .
let @xmath326 be a non - split extension such that @xmath47 is uniserial and @xmath28 is simple .
then we have the following : 1 .
every epimorphism @xmath327 induces an isomorphism @xmath328 2 .
the object @xmath16 is uniserial .
given any non - split extension @xmath329 in @xmath330 , there exists an isomorphism @xmath331 such that @xmath332 .
\(1 ) it is sufficient to consider the case @xmath333 .
moreover , it is sufficient to show that the induced map @xmath334 is a monomorphism , since @xmath335 by .
we use the long exact sequence which is obtained from the short exact sequence @xmath336 by applying @xmath337 . if @xmath338 , then the induced map @xmath339 is a monomorphism . now assume that @xmath340 . using induction on the length
, we have that @xmath341 .
next observe that @xmath342 since @xmath47 is uniserial .
thus implies @xmath343 , and a dimension argument shows that the connecting morphism @xmath344 is an isomorphism .
thus from the long exact sequence we infer that the natural map @xmath345 is a monomorphism .
\(2 ) it suffices to show that every proper subobject @xmath346 is contained in @xmath47 .
otherwise we have an induced extension @xmath347 .
thus the inclusion @xmath348 induces a non - zero map @xmath349 . composing this with the isomorphism @xmath350 from ( 1 ) gives a non - zero map @xmath351 which is induced by the composite @xmath352 .
this is impossible .
\(3 ) we use induction on the length of @xmath47 . the case
@xmath353 follows from the equality @xmath354 . if @xmath355 , choose a maximal subobject @xmath92 and let @xmath356 .
it follows from ( 1 ) that the canonical morphism @xmath357 induces an isomorphism @xmath358 taking @xmath220 to @xmath359 .
there is an isomorphism @xmath360 such that @xmath361 since @xmath362 .
we claim that @xmath363 extends to an isomorphism @xmath364 satisfying @xmath365 .
this implies @xmath366 , and therefore @xmath332 .
thus it remains to construct @xmath367 . to this end
consider the non - split extension @xmath368 .
the induction hypothesis yields an isomorphism @xmath369 such that @xmath370 since @xmath371 .
this gives the isomorphism @xmath367 satisfying @xmath372 . [ le : uniserial ] let @xmath31 be a length category and suppose that holds for each simple object @xmath28 .
for two uniserial objects @xmath47 and @xmath72 the following are equivalent : 1 .
2 . @xmath373 and @xmath374 .
3 . @xmath375 and @xmath374 .
the condition is self - dual .
thus it suffices to show the equivalence ( 1 ) @xmath376 ( 2 ) .
this equivalence follows from lemma [ le : gabriel ] using induction on the length @xmath301 .
suppose first that @xmath31 is uniserial .
choose a simple object @xmath28 and assume that @xmath377 then there exists an extension @xmath378 with @xmath379 such that for each non - zero morphism @xmath380 with @xmath381 , the induced extension @xmath382 does not split .
it is not difficult to check that @xmath16 is indecomposable and has at least two different composition series .
thus @xmath16 is not uniserial which is a contradiction .
now assume that holds for each simple object @xmath28 and fix an indecomposable object @xmath47 .
we show by induction on @xmath301 that @xmath47 is uniserial .
the case @xmath353 is clear .
thus we choose an exact sequence @xmath383 with @xmath28 simple and fix a decomposition @xmath384 into indecomposable objects .
note that each @xmath385 is uniserial by our hypothesis .
if @xmath386 , then @xmath47 is uniserial by lemma [ le : gabriel ] .
now assume @xmath387 .
denote by @xmath388 the pushout of @xmath220 along the projection @xmath389 .
note that @xmath390 ; otherwise @xmath385 is isomorphic to a direct summand of @xmath47 .
therefore @xmath391 for all @xmath181 , and lemma [ le : gabriel ] implies @xmath392 for all @xmath181 .
assume that @xmath393 .
then we have an epimorphism @xmath394 by lemma [ le : uniserial ] which induces an isomorphism @xmath395 by lemma [ le : gabriel ] .
moreover , there exists an isomorphism @xmath396 such that @xmath397 .
consider the morphism @xmath398 with @xmath399 , @xmath400 and @xmath401 for @xmath402 .
we have @xmath403 by construction , and therefore @xmath404 is isomorphic to a direct summand of @xmath47 .
this is a contradiction .
thus @xmath47 is uniserial .
let @xmath31 be a uniserial category .
choose a complete set of representatives of the isomorphism classes of indecomposable objects in @xmath31 and denote it by @xmath405 .
an object in @xmath405 with top @xmath28 and length @xmath188 is denoted by @xmath406}$ ] .
analogously , we write @xmath407}$ ] for an object in @xmath405 with socle @xmath28 and length @xmath188 . for each simple object @xmath28 , we have a chain of monomorphisms @xmath408}\rightarrowtail s_{[2]}\rightarrowtail s_{[3]}\rightarrowtail \cdots\ ] ] which is either finite or infinite .
dually , there is a chain of epimorphisms @xmath409}\twoheadrightarrow s^{[2]}\twoheadrightarrow s^{[1]}=s.\ ] ] a morphism @xmath74 in an additive category is called _ irreducible _ if @xmath76 is neither a split monomorphism nor a split epimorphism and if for any factorisation @xmath410 the morphism @xmath411 is a split monomorphism or @xmath412 is a split epimorphism .
[ le : irr ] let @xmath31 be a uniserial category . for a morphism @xmath74 between indecomposable objects ,
the following are equivalent : 1 .
the morphism @xmath76 is irreducible .
the object @xmath413 is simple .
3 . there exists a simple object @xmath28 and an integer @xmath188 such that @xmath76 is , up to isomorphism , of the form @xmath407}\rightarrowtail s_{[n+1]}$ ] or @xmath414}\twoheadrightarrow s^{[n]}$ ] .
\(1 ) @xmath186 ( 2 ) : an irreducible morphism is either a monomorphism or an epimorphism .
it suffices to discuss the case that @xmath76 is an epimorphism ; the other case is dual .
if @xmath415 and @xmath416 is a simple subobject , then @xmath76 can be written as composite @xmath417 of two proper epimorphisms .
this is a contradiction , and therefore @xmath413 is simple .
\(2 ) @xmath186 ( 3 ) : clear .
\(3 ) @xmath186 ( 1 ) : it suffices to consider the morphism @xmath418}\twoheadrightarrow s^{[n]}$ ] ; the dual argument works for @xmath407}\rightarrowtail s_{[n+1]}$ ] . let @xmath414}{\xrightarrow}{\a } x{\xrightarrow}{\b}s^{[n]}$ ] be a factorization and fix a decomposition @xmath419 into indecomposable objects .
then @xmath420 is an epimorphism for at least one index @xmath421 .
it follows from lemma [ le : uniserial ] that @xmath422}$ ] for some @xmath423 .
if @xmath424 , then @xmath425 is an isomorphism , and therefore @xmath426 is a split epimorphism .
otherwise , we obtain a factorization @xmath414}{\xrightarrow}{\a_{i_0 } } x_{i_0}{\xrightarrow}{\b'_{i_0}}s^{[n+1]}\twoheadrightarrow s^{[n]}$ ] of the epimorphism @xmath420 .
it follows that @xmath427 is an epimorphism and hence an isomorphism .
thus @xmath428 is a split monomorphism .
[ re : inj ] let @xmath31 be a uniserial category and @xmath28 a simple object .
suppose there is a bound @xmath188 such that @xmath429 for each indecomposable object with @xmath430 .
then each indecomposable object @xmath47 of length @xmath188 with @xmath430 is injective , since @xmath431 for every simple object @xmath32 in @xmath31 by lemma [ le : gabriel ] .
\(1 ) let @xmath82 be a dedekind domain . the finitely generated torsion modules over @xmath82 form a uniserial category .
we denote this category by @xmath432 because it coincides with the category @xmath319 of finite length objects of @xmath83 .
let @xmath433 denote the set of prime ideals .
the functor which takes a @xmath82-module @xmath155 to the family of localizations @xmath434 induces an equivalence @xmath435 \(2 ) let @xmath0 be a field and @xmath436 $ ] an irreducible polynomial . for each @xmath274 ,
the finitely generated @xmath298/(p^n)$]-modules form a uniserial category with a unique simple object .
\(3 ) let @xmath0 be a field and @xmath41 a quiver having no oriented cycle .
the category of representations @xmath84 is uniserial if and only if for each vertex @xmath437 of @xmath41 , there is at most one arrow starting at @xmath437 and at most one arrow ending at @xmath437 .
let @xmath0 be a commutative ring .
a category @xmath31 is _ @xmath0-linear _ if each morphism set is a @xmath0-module and the composition maps are @xmath0-bilinear .
a functor between @xmath0-linear categories is _ @xmath0-linear _ provided that the induced maps between the morphism sets are @xmath0-linear .
a @xmath0-linear category is _ hom - finite _ if each morphism set is a @xmath0-module of finite length .
suppose now that @xmath31 is a @xmath0-linear abelian category .
then the extension groups are naturally modules over @xmath0 , and @xmath31 is called _ ext - finite _ if @xmath438 is of finite length over @xmath0 for all @xmath67 in @xmath31 and @xmath290
. fix a field @xmath0 and a hom - finite @xmath0-linear abelian category @xmath31 .
the category @xmath31 is said to satisfy _
_ serre duality _ _ appear in algebraic geometry ; see also 3.4 . ]
if there exists an equivalence @xmath439 with functorial @xmath0-linear isomorphisms @xmath440 for all @xmath441 in @xmath31 , where @xmath442 denotes the standard @xmath0-duality .
the functor @xmath367 is called _ serre functor _ or _ auslander - reiten translation_. note that a serre functor is @xmath0-linear and essentially unique provided it exists ; this follows from yoneda s lemma . recall that @xmath302 denotes the full subcategory consisting of all finite length objects in @xmath31
. denote by @xmath443 the full subcategory consisting of all objects @xmath47 in @xmath31 satisfying @xmath444 for all @xmath445 in @xmath302 .
[ pr : serreduality ] let @xmath31 be a hom - finite @xmath0-linear abelian category and suppose @xmath31 admits a serre functor @xmath367 .
then the following holds : 1 .
the category @xmath31 is hereditary .
the category @xmath31 has no non - zero projective or injective objects .
a noetherian object @xmath47 has a unique maximal subobject @xmath445 of finite length .
moreover , @xmath445 is a direct summand of @xmath47 and @xmath446 belongs to @xmath443 .
4 . for each indecomposable object @xmath47 in @xmath31
, there is an almost split sequence @xmath447 .
5 . for each object
@xmath47 in @xmath31 , we have @xmath448 .
\(1 ) for each object @xmath47 , the functor @xmath449 is right exact .
thus the category @xmath31 is hereditary by lemma [ lem : hereditary ] .
\(2 ) let @xmath47 be a projective object .
then @xmath450 . thus @xmath451 and therefore @xmath36 .
the dual argument works for injective objects .
\(3 ) choose a maximal subobject @xmath445 of finite length .
then @xmath446 belongs to @xmath443 , and every finite length subobject of @xmath47 is contained in @xmath445 . in particular
, @xmath445 is unique .
we have @xmath452 by serre duality , and therefore @xmath445 is a direct summand of @xmath47 .
\(4 ) let @xmath47 be an indecomposable object .
the endomorphism ring @xmath453 is local and we denote by @xmath318 its maximal ideal .
choose any non - zero @xmath0-linear map @xmath454 such that @xmath455 vanishes on @xmath456 .
the map @xmath455 corresponds via serre duality to a non - split short exact sequence @xmath457 .
we claim that @xmath220 is an almost split sequence . for this one
needs to show that each morphism @xmath458 factors through the morphism @xmath459 , provided that @xmath428 is not a split epimorphism .
thus one needs to show that @xmath460 .
the element @xmath461 corresponds via serre duality to @xmath462 which sends @xmath463 to @xmath464 .
thus @xmath460 , since @xmath465 belongs to @xmath318 .
\(5 ) this is clear from the definitions .
next observe that a serre functor @xmath367 on @xmath31 restricts to a serre functor on the subcategory @xmath302 of finite length objects .
the following result describes the structure of a length category with serre duality .
let us recall the shape of the relevant diagrams .
@xmath466&2\ar@{-}[r]&3\ar@{-}[r]&{\dots}\ar@{-}[r]&n\ar@{-}[r]&n+1\ar@{- } @/^1pc/[lllll]}\\[3ex ] \bba_\infty^\infty&\colon\;\;\ ; \xymatrix{\dots \ar@{-}[r]&\scriptstyle{\bullet}\ar@{-}[r ] & \scriptstyle{\bullet}\ar@{-}[r]&\scriptstyle{\bullet}\ar@{-}[r ] & \scriptstyle{\bullet}\ar@{-}[r]&{\dots}}\end{aligned}\ ] ] [ pr : serre_finlen ] let @xmath31 be a hom - finite @xmath0-linear length category and suppose @xmath31 admits a serre functor @xmath367 .
then @xmath31 is uniserial .
the category @xmath31 admits a unique decomposition @xmath467 into connected uniserial categories with serre duality , where the index set equals the set of @xmath367-orbits of simple objects in @xmath31 .
the ext - quiver of each @xmath468 is either of type @xmath469 ( with linear orientation ) or of type @xmath470 ( with cyclic orientation ) .
we apply the criterion of theorem [ th : gabriel ] to show that @xmath31 is uniserial . to this end
fix a simple object @xmath28 .
then @xmath471 for some @xmath472 if and only if @xmath473 .
moreover , @xmath474 .
thus the category @xmath31 is uniserial .
the structure of the ext - quiver of @xmath31 follows from the condition .
the serre functor acts on @xmath305 and the set of @xmath367-orbits @xmath475 is the index set of the decomposition @xmath467 into connected components ; see lemma [ le : comp ] .
the ext - quiver of @xmath468 is of type @xmath469 if the corresponding @xmath367-orbit is infinite . otherwise , the ext - quiver of @xmath468 is of type @xmath476 where @xmath477 equals the cardinality of the @xmath367-orbit .
let @xmath31 be a hom - finite @xmath0-linear length category and suppose @xmath31 admits a serre functor .
then a complete set of representatives of the isomorphism classes of indecomposable objects of @xmath31 is given by @xmath478}\mid s\in\si_0,\ , n\ge 1\}$ ] and also by @xmath479}\mid s\in\si_0,\ , n\ge 1\}$ ] ; see remark [ re : inj ] . [ ex : cyclic ] let @xmath0 be a field and @xmath41 a quiver of extended dynkin type @xmath480 with cyclic orientation
. denote by @xmath481 the full subcategory of @xmath84 consisting of all nilpotent representations .
then @xmath31 satisfies serre duality and the ext - quiver of @xmath31 is isomorphic to @xmath41 , with valuation @xmath321 for each arrow . note that the serre functor on @xmath31 has order @xmath477 and every simple object has endomorphism algebra @xmath0 .
in fact , a connected hom - finite @xmath0-linear length category with serre duality satisfying these properties is equivalent to @xmath482 .
to each abelian category is associated its derived category .
this section provides a brief introduction .
we present the definition and discuss two cases where one has a convenient description : if the abelian category is hereditary , then each complex is isomorphic to its cohomology . on the other hand ,
if there are enough projective objects , then one can compute morphisms in the derived category by passing to the homotopy category of projective objects .
let @xmath31 be an additive category .
cochain complex _ in @xmath31 is a sequence of morphisms @xmath483 such that @xmath484 for all @xmath485 .
we denote by @xmath486 the category of cochain complexes , where a morphism @xmath487 between cochain complexes consists of morphisms @xmath488 with @xmath489 for all @xmath485 .
a _ chain complex _ in @xmath31 is a sequence of morphisms @xmath490 such that @xmath491 for all @xmath485 .
any chain complex may be viewed as a cochain complex by changing its indices , and vice versa .
thus we often confuse both concepts and simply use the term _ complex_. a morphism @xmath487 between complexes is _ null - homotopic _ if there are morphisms @xmath492 such that @xmath493 for all @xmath485 .
the null - homotopic morphisms form an _ ideal _ @xmath494 in @xmath486 , that is , for each pair @xmath495 of complexes a subgroup @xmath496 such that any composition @xmath497 of morphisms in @xmath486 belongs to @xmath494 if @xmath76 or @xmath124 belongs to @xmath494 . the _ homotopy category _
@xmath498 is the quotient of @xmath486 with respect to this ideal .
thus @xmath499 for every pair of complexes @xmath495 .
now let @xmath31 be an abelian category .
the _ cohomology _ of a complex @xmath500 in degree @xmath188 is by definition @xmath501 , and each morphism @xmath487 of complexes induces a morphism @xmath502 . the morphism @xmath76 is a _ quasi - isomorphism _ if @xmath503 is an isomorphism for all @xmath485 .
note that two morphisms @xmath504 induce the same morphism @xmath505 , if @xmath506 is null - homotopic .
the _ derived category _
@xmath507 of @xmath31 is obtained from @xmath498 by formally inverting all quasi - isomorphisms . to be precise
, one defines @xmath508\ ] ] as the localization of @xmath498 with respect to the class of all quasi - isomorphisms .
the full subcategory consisting of objects that are isomorphic to a complex @xmath500 such that @xmath509 for almost all @xmath485 is denoted by @xmath510 .
an object @xmath47 in @xmath31 is identified with the complex @xmath511 concentrated in degree zero , and this complex is also denoted by @xmath47 . given any complex @xmath500 in @xmath31 and @xmath512 , we denote by @xmath513 $ ] the shifted complex with @xmath514^n = x^{n+p}\quad\textrm{and}\quad d^n_{x[p]}=(-1)^pd^{n+p}_x.\ ] ] this operation induces an isomorphism @xmath515 and is called _
shift_. the derived category @xmath507 is an additive category with some additional structure : it is a triangulated category in the sense of verdier @xcite .
for instance , any exact sequence @xmath516 in @xmath31 induces an exact triangle @xmath517 $ ] in @xmath507
. given two abelian categories @xmath31 and @xmath518 , a functor @xmath519 is by definition a _ derived equivalence _ if it is an _ equivalence of triangulated categories _ , that is , @xmath81 is an equivalence , there is a functorial isomorphism @xmath520\cong f(x[1])$ ] for each @xmath500 in @xmath507 , and @xmath81 preserves exact triangles .
the following statement justifies the study of derived categories .
[ pr : ext ] let @xmath67 be objects in @xmath31 .
then @xmath521)\quad\text{for all}\quad n\in \mathbb{z}.\ ] ] for the case that @xmath31 has enough injectives or enough projectives , see ( * ? ?
* corollary 10.7.5 ) . for
the general case , see ( * ? ? ?
* iii.3 ) .
let @xmath31 be a hereditary abelian category , that is , @xmath522 vanishes .
in this case , there is an explicit description of all objects and morphisms in @xmath510 via the ones in @xmath31 .
every complex @xmath500 is completely determined by its cohomology because there is an isomorphism between @xmath500 and the following complex with trivial differential .
@xmath523 to construct this isomorphism , note that the vanishing of @xmath524 implies the existence of a commutative diagram @xmath525&x^{n-1}\ar[r]\ar[d]&e^n\ar[r]\ar[d ] & h^nx\ar[r]\ar@{=}[d]&0\\ 0\ar[r]&{\operatorname{im}}d^{n-1}\ar[r]&\operatorname{ker}d^n\ar[r]&h^nx\ar[r]&0}\ ] ] with exact rows .
we obtain the following commutative diagram .
\cdots\ar[r]&0\ar[u]\ar[r]\ar[d]&x^{n-1}\ar[u]\ar[r]\ar@{=}[d]&e^n\ar[u]\ar[r]\ar[d ] & 0\ar[u]\ar[r]\ar[d]&\cdots\\ \cdots\ar[r]&x^{n-2}\ar[r]&x^{n-1}\ar[r]&x^{n}\ar[r]&x^{n+1}\ar[r]&\cdots } \ ] ] the vertical morphisms yield two morphisms in @xmath510 . the upper one is a quasi - isomorphism , and the lower one induces a cohomology isomorphism in degree @xmath188 .
this yields for each @xmath485 a morphism @xmath527\to x$ ] in @xmath510 and therefore the following description of @xmath500 .
[ le : der_hereditary ] let @xmath31 be a hereditary abelian category and @xmath500 a complex in @xmath31 . in @xmath510
there is a ( non - canonical ) isomorphism @xmath528{\stackrel{\sim}{\longrightarrow } } x.\ ] ] the morphism is a quasi - isomorphism by construction . for an abelian category @xmath31 one defines its _ repetitive category _
@xmath529 $ ] as the additive closure of the union of disjoint copies @xmath530 $ ] of @xmath31 with morphisms @xmath531,\ , b\in\a[q]\ ] ] and composition given by the yoneda product of extensions .
it follows from proposition [ pr : ext ] that the family of functors @xmath530\to\bfd^b(\a)$ ] ( @xmath485 ) sending an object @xmath47 to @xmath532 $ ] induces a fully faithful functor @xmath533{\stackrel{{}}{\longrightarrow}}\bfd^b(\a).\ ] ] [ co : der_hereditary ] the canonical functor @xmath529\to\bfd^b(\a)$ ] is an equivalence for any hereditary abelian category @xmath31 .
we describe the derived category of an abelian category @xmath31 in terms of its projective objects .
the crucial observation is the following .
[ le : der_proj ] let @xmath495 be a pair of complexes in @xmath31 .
suppose that each @xmath534 is projective and @xmath509 for @xmath535 .
then the map @xmath536 is bijective .
see for example ( * ? ? ?
* corollary 10.4.7 ) .
let @xmath537 denote the full subcategory of @xmath31 consisting of all objects that are projective . denote by @xmath538 the full subcategory of complexes @xmath500 in @xmath539 such that @xmath509 for @xmath535 and @xmath540 for almost all @xmath485 .
one says that @xmath31 has _ enough projectives _ if each object in @xmath31 is the quotient of some projective object .
[ pr : der_equiv_proj ] let @xmath31 be an abelian category having enough projectives .
then the canonical functor @xmath541 induces an equivalence @xmath542 the functor @xmath543 is by definition the identity on objects , and @xmath81 is fully faithful by lemma [ le : der_proj ] .
it is clear that each object in the image of @xmath81 is isomorphic to one in @xmath510 . to show that each complex @xmath500 in @xmath510 is isomorphic to one in the image of @xmath81 , we use induction on @xmath544 note that each bounded complex @xmath545 fits into an exact triangle @xmath546 $ ] such that @xmath547 and @xmath548 .
if @xmath549 , say @xmath550 , then @xmath551)$ ] where @xmath552 denotes a projective resolution of @xmath534 .
such a resolution exists since @xmath31 has enough projectives .
if @xmath553 , then the induction hypothesis implies that the morphism @xmath554 is up to isomorphism of the form @xmath555 for some morphism @xmath556 in @xmath538 . completing the morphism @xmath76 to an exact triangle @xmath557 $ ] shows that @xmath500 belongs to @xmath70 since @xmath558 .
tilting provides a method to relate a category of coherent sheaves to a category of linear representations .
for instance , a result of beilinson @xcite establishes for the category @xmath559 of coherent sheaves on the projective @xmath188-space over a field @xmath0 an equivalence of derived categories @xmath560 via a tilting object @xmath32 in @xmath559 .
, the object @xmath561 is not a tilting object in the strict sense of these notes ; see proposition [ pr : coh_tilt ] . ] in this section let @xmath0 be a field and @xmath31 a @xmath0-linear abelian category that is ext - finite .
we show that each tilting object @xmath32 in @xmath31 provides an equivalence of derived categories @xmath562 as in the example above .
the principal reference for this result is @xcite , even though the proof given here is somewhat more direct , avoiding the formalism of torsion pairs and t - structures .
fix an object @xmath32 in @xmath31 .
the object @xmath32 is called _ tilting object _ , provided that 1 .
@xmath563 , 2 .
@xmath33 , and 3 .
@xmath564 implies @xmath36 for each object @xmath47 in @xmath31 .
a morphism @xmath565 in @xmath31 is called _
right @xmath32-approximation _ of @xmath47 if it induces an epimorphism @xmath566 and @xmath567 belongs to @xmath568 .
an exact sequence @xmath569 is called _
universal @xmath32-extension _ of @xmath47 if it induces an epimorphism @xmath570 and @xmath567 belongs to @xmath568 .
such approximations and extensions exist for all @xmath47 in @xmath31 , since @xmath31 is ext - finite .
finally , set @xmath571 [ le : tilt_obj ] let @xmath572 be a tilting object .
then the following holds : 1 .
let @xmath573 be a right @xmath32-approximation .
then @xmath574 is in @xmath575 , @xmath576 , and @xmath577 .
2 . let @xmath569 be a universal @xmath32-extension
. then @xmath578 .
the objects in @xmath575 are precisely the factor objects of objects in @xmath568 .
\(1 ) write the sequence @xmath579 as composite of two exact sequences @xmath580 and @xmath581 .
then apply @xmath582 to both sequences .
\(2 ) apply @xmath582 to the sequence @xmath569 .
\(3 ) clearly , each factor of an object in @xmath568 belongs to @xmath575 . for the other implication one uses ( 1 ) .
let @xmath32 be an object in @xmath31 and @xmath38 .
we consider the functor @xmath583 this functor induces an equivalence @xmath584 and admits a left adjoint @xmath585 given a @xmath82-module @xmath155 with projective presentation @xmath586 , the cokernel of the corresponding morphism @xmath587 in @xmath568 is by definition @xmath588 . for @xmath589 ,
denote by @xmath590 the @xmath181-th left derived functor of @xmath591 and set @xmath592 [ le : tilt_tor ] let @xmath572 be a tilting object .
then @xmath593 for @xmath594 .
let @xmath595 and choose a projective resolution @xmath596 apply @xmath591 and set @xmath597 .
the induced morphism @xmath598 is a right @xmath32-approximation for @xmath589 , and therefore @xmath599 belongs to @xmath575 for @xmath594 by lemma [ le : tilt_obj ] .
thus @xmath600 is an epimorphism for @xmath594 , and this implies @xmath593 .
[ le : tilt_eq ] let @xmath572 be a tilting object
. then @xmath601 and @xmath591 restrict to equivalences between @xmath575 and @xmath602 .
fix objects @xmath603 and @xmath604 .
we need to show that the adjunction morphisms @xmath605 are invertible .
choose an exact sequence @xmath606 such that the induced morphism @xmath607 is a right @xmath32-approximation of @xmath608 for each @xmath609 .
such a sequence exists by lemma [ le : tilt_obj ] .
the functor @xmath582 sends the sequence @xmath220 to a projective resolution of the @xmath82-module @xmath610 . applying then @xmath591 gives back @xmath220 , that is
, the adjunction morphism @xmath611 is invertible .
moreover , @xmath610 belongs to @xmath602 .
now choose an exact sequence @xmath612 such that @xmath552 is projective .
note that @xmath613 belongs to @xmath602 by lemma [ le : tilt_tor ] .
the sequence @xmath614 is exact since @xmath604 , and the sequence @xmath615 is exact since @xmath616 belongs to @xmath575 .
thus there is the following commutative diagram with exact rows .
@xmath617 & m ' \ar[r ] \ar[d]^{\eta_{m ' } } & p \ar[r ] \ar[d]^{\eta_p } & m \ar[r ] \ar[d]^{\eta_m } & 0 \\ 0 \ar[r ] & \operatorname{hom}_\a(t , m'\otimes_\la t ) \ar[r ] & \operatorname{hom}_\a(t , p\otimes_\la t ) \ar[r ] & \operatorname{hom}_\a(t , m\otimes_\la t ) \ar[r ] & 0 } \ ] ] the morphism @xmath618 is an isomorphism and it follows that @xmath619 is an epimorphism for all @xmath155 in @xmath602 . in particular , @xmath620 is an epimorphism . using the snake lemma
, it follows that @xmath619 is an isomorphism .
let @xmath47 be an object in @xmath31 .
an _ @xmath568-resolution _ of @xmath47 is by definition a complex @xmath621 together with a quasi - isomorphism @xmath622 such that each @xmath623 belongs to @xmath568 .
[ le : tilt_ex ] let @xmath572 be a tilting object and @xmath622 an @xmath568-resolution of an object @xmath624 .
then @xmath625 for all @xmath578 and @xmath626 .
use induction on @xmath188 and dimension shifting .
[ le : tilt_ext ] let @xmath572 be a tilting object .
then the functor @xmath591 induces an isomorphism @xmath627 for all @xmath628 in @xmath602 and @xmath290 .
choose a projective resolution @xmath629 of @xmath155 .
note that @xmath630 by lemma [ le : tilt_eq ] , since @xmath631 belongs to @xmath602 .
then we obtain the following sequence of isomorphisms .
@xmath632 the last isomorphism follows from lemma [ le : tilt_ex ] , since @xmath633 is an @xmath568-resolution . for a tilting object @xmath32 in @xmath31 ,
let us define the derived functor @xmath634 by taking a complex @xmath552 of projective @xmath82-modules with bounded cohomology to @xmath635 ; see proposition [ pr : der_equiv_proj ] .
the cohomology of @xmath636 is bounded , since @xmath637 for @xmath594 by lemma [ le : tilt_tor ] .
[ th : tilt ] let @xmath31 be a @xmath0-linear abelian category that is ext - finite .
let @xmath32 be a tilting object in @xmath31 and @xmath38 .
then the functor @xmath638 is an equivalence of triangulated categories and its right adjoint @xmath639 is a quasi - inverse .
we do not give the formal definition of the derived functor @xmath639 ; all we use is the fact that it is a right adjoint of @xmath640 .
set @xmath641 .
we identify objects in @xmath83 and @xmath31 with complexes that are concentrated in degree zero .
for instance , @xmath642 for each @xmath155 in @xmath602 .
we need to show that for each pair of complexes @xmath495 in @xmath83 , the induced map @xmath643 is bijective . set @xmath644 and
note that each bounded complex @xmath545 fits into an exact triangle @xmath645 $ ] such that @xmath646 and @xmath647 . using the five lemma and induction on @xmath648
, one shows that @xmath649 is bijective .
the case @xmath650 follows from lemma [ le : tilt_ext ] . to be precise , one uses that each @xmath82-module @xmath155 fits into an exact sequence @xmath651 with @xmath652 in @xmath602 , which yields an exact triangle @xmath653 $ ] in @xmath654 .
next we show that each object in @xmath510 is isomorphic to one in the image of @xmath655 .
in fact , it suffices to show that each object in @xmath31 belongs to the essential image @xmath656 , since @xmath656 is a triangulated subcategory and @xmath510 is generated ( as a triangulated category ) by the objects from @xmath31 .
it follows from lemma [ le : tilt_obj ] that each object @xmath47 in @xmath31 fits into an exact triangle @xmath517 $ ] with @xmath657 in @xmath575 . on the other hand
, each @xmath47 in @xmath575 belongs to @xmath656 , since @xmath658 by lemma [ le : tilt_eq ] .
\(1 ) let @xmath659 be two objects in @xmath31 with @xmath660 .
then @xmath32 is a tilting object if and only if @xmath567 is a tilting object .
\(2 ) let @xmath0 be a field and @xmath82 a finite dimensional @xmath0-algebra .
then any free @xmath82-module of finite rank is a tilting object in @xmath83 .
\(3 ) let @xmath0 be a field and @xmath661 the path algebra of a finite quiver @xmath41 without oriented cycles .
for each vertex @xmath662 let @xmath663 denote the corresponding idempotent .
fix a vertex @xmath421 which is not a sink and consider the following short exact sequence @xmath664 where the direct sum runs over all arrows starting at @xmath421 and each morphism @xmath665 is given by multiplication with the corresponding arrow @xmath666 .
set @xmath667 for each vertex @xmath668 .
then @xmath669 is a tilting object of @xmath83 .
let @xmath31 be an abelian category . denote by @xmath670 the free abelian group generated by the isomorphism classes of objects in @xmath31 . let @xmath671 be the subgroup generated by @xmath672-[y]+[z]$ ] for all exact sequences @xmath673 in @xmath31 .
the _ grothendieck group _ @xmath674 of @xmath31 is by definition the factor group @xmath675 . [
le : gr_gr1 ] let @xmath31 be a length category . then @xmath674 is a free abelian group and the isomorphism classes of simple objects in @xmath31 form a basis .
let @xmath500 be an object in @xmath31 and @xmath676 a composition series .
then @xmath672=[x_1/x_0]+\cdots + [ x_n / x_{n-1}]$ ] in @xmath674 .
the jordan - hlder theorem implies the uniqueness of this expression .
let @xmath677 be a triangulated category .
denote by @xmath678 the free abelian group generated by the isomorphism classes of objects in @xmath677 .
let @xmath679 be the subgroup generated by @xmath672-[y]+[z]$ ] for all exact triangles @xmath680 $ ] in @xmath677 .
the _ grothendieck group _ @xmath681 of @xmath677 is by definition the factor group @xmath682 .
[ le : gr_gr2 ] let @xmath31 be an abelian category .
the inclusion @xmath683 induces an isomorphism @xmath684 .
each exact sequence @xmath673 in @xmath31 induces an exact triangle @xmath680 $ ] in @xmath510 .
this gives a morphism @xmath685 .
the inverse map sends the class @xmath672 $ ] of a complex @xmath500 to @xmath686 $ ] .
let @xmath0 be a field and @xmath82 a finite dimensional @xmath0-algebra .
we denote by @xmath687 the standard @xmath0-duality .
the _ nakayama functor _
@xmath688 identifies the category of projective @xmath82-modules with the category of injective @xmath82-modules .
note that @xmath689 for every finitely generated projective @xmath82-module @xmath552 , since both functors are left exact and agree on @xmath82 .
this isomorphism induces for every bounded complex @xmath500 of finitely generated projective @xmath82-modules a sequence of natural isomorphisms @xmath690 where the first and the last isomorphism follow from lemma [ le : der_proj ] .
a hom - finite @xmath0-linear triangulated category @xmath677 is said to satisfy _ serre duality _ if there exists an equivalence @xmath691 of triangulated categories with functorial @xmath0-linear isomorphisms @xmath692 for all @xmath495 in @xmath677 .
the functor @xmath367 is called a _ serre functor_. note that a serre functor is @xmath0-linear and essentially unique provided it exists ; this follows from yoneda s lemma .
[ pr : serre_alg ] let @xmath82 be a finite dimensional @xmath0-algebra
. then @xmath654 satisfies serre duality if and only if @xmath82 has finite global dimension .
if the global dimension of @xmath82 is finite , then every bounded complex in @xmath83 is quasi - isomorphic to a bounded complex of finitely generated projective @xmath82-modules . thus serre duality for @xmath654 follows from the isomorphism .
the converse follows immediately from lemmas [ le : asymp ] and [ le : gldim ] below .
[ le : asymp ] let @xmath31 be an abelian category and @xmath693 . then the following holds : 1 .
@xmath694)=0 $ ] for @xmath695 .
2 . @xmath694)=0 $ ] for @xmath535 , if @xmath31 has finite global dimension .
use induction on @xmath648 , where @xmath696 for any complex @xmath697 .
the case @xmath698 is clear , since @xmath699)\cong\operatorname{ext}^n_\a(a , b)$ ] for all objects @xmath67 in @xmath31 ; see proposition [ pr : ext ] . [
le : gldim ] let @xmath82 be a finite dimensional algebra and @xmath700 a set of representatives of the isomorphism classes of simple @xmath82-modules
. then @xmath701 if and only if @xmath702 for all @xmath703 .
use that each @xmath82-module @xmath155 has a finite filtration @xmath704 such that @xmath168 is semisimple for all @xmath181 .
[ le : tilt_gldim ] let @xmath31 be an ext - finite @xmath0-linear abelian category and suppose there exists a tilting object @xmath32 . if @xmath31 has finite global dimension , then @xmath705 has finite global dimension .
let @xmath706 .
the functor @xmath639 provides an equivalence @xmath707 of triangulated categories by theorem [ th : tilt ] .
thus we have @xmath708 for @xmath535 and each pair @xmath309 of simple @xmath82-modules by lemma [ le : asymp ] .
it follows from lemma [ le : gldim ] that the global dimension of @xmath82 is finite .
[ pr : serre ] let @xmath31 be a @xmath0-linear abelian category that is ext - finite and admits a tilting object .
then the following are equivalent : 1 .
the category @xmath31 is hereditary and has no non - zero projective object .
the category @xmath31 satisfies serre duality .
\(1 ) @xmath186 ( 2 ) : let @xmath32 be the tilting object and @xmath38 .
then @xmath82 has finite global dimension by lemma [ le : tilt_gldim ] , and therefore @xmath654 has serre duality by proposition [ pr : serre_alg ] .
there is an equivalence @xmath707 by theorem [ th : tilt ] , and this yields a serre functor @xmath709 .
now let @xmath67 be objects in @xmath31 and view them as complexes concentrated in degree zero
. then @xmath710),\ ] ] and it remains to show that @xmath711)$ ] for all @xmath712 .
any complex @xmath500 in @xmath31 is quasi - isomorphic to @xmath713 $ ] since @xmath31 is hereditary ; see lemma [ le : der_hereditary ] .
assume that @xmath47 is indecomposable .
then @xmath714\cong\bar a[d]$ ] for some @xmath715 and some object @xmath191 in @xmath31 .
we claim that @xmath716 .
first observe that @xmath717)\neq 0 $ ] for some object @xmath72 , since @xmath47 is non - projective .
thus @xmath716 or @xmath718 .
the case @xmath718 is impossible since @xmath719 .
thus @xmath714 $ ] is concentrated in degree zero .
\(2 ) @xmath186 ( 1 ) : see proposition [ pr : serreduality ] .
let @xmath0 be a field and @xmath31 a @xmath0-linear abelian category .
suppose that @xmath31 is ext - finite and of finite global dimension .
the _ euler form _ associated to @xmath31 is by definition the bilinear form @xmath720 with @xmath721,[b]\rangle=\sum_{n\geq 0}(-1)^n{\operatorname{dim}}_k\operatorname{ext}_\a^n(a , b).\ ] ] suppose that @xmath674 is a free abelian group of finite rank and fix a basis @xmath722 .
the _ discriminant _ of the euler form is then by definition the determinant of the matrix @xmath723 and we denote it by @xmath724 .
note that this value does not depend on the choice of the basis since the matrix is defined over @xmath297 .
[ le : euler_fin ] let @xmath82 be a finite dimensional @xmath0-algebra of finite global dimension .
then the euler form associated to @xmath83 is non - degenerate .
set @xmath725 .
let @xmath726 be representatives of the isomorphism classes of simple @xmath82-modules and choose a projective cover @xmath727 for each @xmath181 .
then @xmath728,\dots,[p_r]$ ] form a basis of @xmath674 , since each @xmath180 has a finite projective resolution .
let @xmath729 $ ] be a non - zero element of @xmath674 and pick an index @xmath730 such that @xmath731
. then @xmath732\rangle=\a_{j}{\operatorname{dim}}_k\operatorname{hom}_\la(p_{j},s_{j})\neq 0 $ ] .
thus @xmath733 is non - degenerate .
let @xmath677 be a @xmath0-linear triangulated category .
suppose that @xmath677 is hom - finite and that @xmath734)=0 $ ] for each pair of objects @xmath495 and @xmath735 . the _ euler form _ associated to @xmath677 is by definition the bilinear form @xmath736 with @xmath737,[y]\rangle=\sum_{n\in\bbz}(-1)^n{\operatorname{dim}}_k\operatorname{hom}_\t(x , y[n]).\ ] ] it is clear from these definitions that the isomorphism @xmath738 from lemma [ le : gr_gr2 ] is an _ isometry _
, that is , @xmath739 now let @xmath31 and @xmath71 be @xmath0-linear abelian categories that are ext - finite and of finite global dimension . a @xmath0-linear equivalence @xmath740 of triangulated categories induces an isometry @xmath741 which is defined by the commutativity of the following diagram .
@xmath742^\sim\ar[d]_{\p_\a}&&k_0(\b)\ar[d]^{\p_\b}\\ k_0(\bfd^b(\a))\ar[rr]^{k_0(f)}&&k_0(\bfd^b(\b))}\ ] ] this has the following consequence .
[ pr : euler ] let @xmath31 be a @xmath0-linear abelian category that is ext - finite and of finite global dimension .
suppose that @xmath31 has a tilting object .
then the grothendieck group @xmath674 is free of finite rank and the euler form associated to @xmath31 is non - degenerate .
we identify @xmath674 with @xmath743 , where @xmath38 for a tilting object @xmath32 in @xmath31 .
then the grothendieck group @xmath674 is free of finite rank by lemma [ le : gr_gr1 ] , and the euler form is non - degenerate by lemma [ le : euler_fin ] .
the following examples show that the existence of a tilting object is an essential assumption for the euler form to be non - degenerate .
[ ex : degen ] let @xmath31 be a @xmath0-linear length category that is hom - finite and satisfies serre duality .
suppose that the grothendieck group has finite rank , and let @xmath744 be a representative set of the simple objects .
then @xmath745\rangle=0 $ ] for @xmath746 $ ] and all @xmath181 .
thus the euler form is degenerate .
let @xmath16 be a smooth elliptic curve over some algebraically closed field .
then the category of coherent sheaves on @xmath16 is hereditary and satisfies serre duality , but the corresponding euler form is degenerate .
in fact , any two simple sheaves @xmath309 satisfy @xmath747\rangle=\langle-,[t]\rangle$ ] , but @xmath748\neq[t]$ ] if @xmath28 and @xmath32 are concentrated in different points of @xmath16 . indeed , @xmath748\neq [ t]$ ] follows from ( * ? ? ?
ii , exercise 6.11 ) and the fact that the set of closed points is naturally identified with a subset of the picard group ( * ? ? ?
iv , example 1.3.7 ) .
next we collect some further properties of the grothendieck group and its euler form . [
le : euler_tilt ] let @xmath31 be a @xmath0-linear abelian category that is hereditary , ext - finite , and has a non - degenerate euler form .
suppose also that @xmath749\neq 0 $ ] for each non - zero object @xmath47 in @xmath31 .
then an object @xmath32 is a tilting object if and only if @xmath33 and the classes of the indecomposable direct summands of @xmath32 generate @xmath674 .
suppose first that @xmath32 is a tilting object with @xmath38 .
the isomorphism @xmath750 identifies ( the classes of ) the indecomposable direct summands of @xmath32 with the indecomposable projective @xmath82-modules .
now one uses that the indecomposable projective @xmath82-modules generate @xmath743 ; see the proof of lemma [ le : euler_fin ] conversely , suppose that @xmath33 and that the indecomposable direct summands of @xmath32 generate @xmath674
. then there exists for any non - zero object @xmath47 in @xmath31 some indecomposable direct summand @xmath567 of @xmath32 such that @xmath751,[a]\rangle\neq 0 $ ] .
thus @xmath752 , and it follows that @xmath32 is a tilting object .
a full subcategory @xmath71 of an abelian category @xmath31 is called _ exact abelian _
if @xmath71 is an abelian category and the inclusion functor is exact .
[ le : groth_onepoint ] let @xmath31 be an abelian category and @xmath71 an exact abelian subcategory such that the inclusion admits an exact left adjoint .
let @xmath753 .
then @xmath754 , where @xmath755 denotes the image of the canonical map @xmath756 .
let @xmath757 be the inclusion and @xmath758 its left adjoint .
observe that @xmath759 is a serre subcategory of @xmath31 by proposition [ pr : perp ] .
thus the inclusion @xmath248 induces a linear map @xmath756 .
we have @xmath760 and therefore @xmath761 identifies @xmath762 with a direct summand of @xmath674 .
the kernel of @xmath763 equals @xmath755 , since there is an exact sequence @xmath764 for each object @xmath47 in @xmath31 with @xmath765 in @xmath86 ; see proposition [ pr : section ] .
the following lemma describes more specifically the term @xmath766 in the decomposition @xmath767 .
[ le : euler_onepoint ] let @xmath31 be a @xmath0-linear abelian category that is ext - finite and of finite global dimension .
suppose that @xmath674 is free of finite rank .
let @xmath71 be an exact abelian subcategory such that the inclusion admits an exact left adjoint and @xmath768 is equivalent to @xmath769 for some division ring @xmath770 .
then @xmath771 and @xmath772 let @xmath757 be the inclusion and denote by @xmath758 its left adjoint .
then @xmath773 for some simple object @xmath28 with @xmath774 ; see proposition [ pr : perp ] .
lemma [ le : groth_onepoint ] implies that @xmath775 $ ] .
observe that @xmath776\neq 0 $ ] for @xmath777 in @xmath297 , since @xmath778,[s]\rangle\neq 0 $ ] .
thus @xmath779 . the formula for @xmath780 follows since @xmath781,[b]\rangle=0
$ ] for every object @xmath72 in @xmath71 .
in this section we introduce the concept of expansion and contraction for abelian categories . roughly speaking , an expansion is a fully faithful and exact functor @xmath782 between abelian categories that admits an exact left adjoint and an exact right adjoint .
in addition one requires the existence of simple objects @xmath783 and @xmath784 in @xmath31 such that @xmath785 , where @xmath71 is viewed as a full subcategory of @xmath31 .
in fact , these simple objects are related by an exact sequence @xmath786 in @xmath31 such that @xmath28 is a simple object in @xmath71 . in terms of the ext - quivers of @xmath31 and @xmath71
, the expansion @xmath782 turns the vertex @xmath28 into an arrow @xmath787 . on the other hand
, @xmath71 is a contraction of @xmath31 in the sense that the left adjoint of @xmath782 identifies @xmath784 with @xmath28 , whereas the right adjoint identifies @xmath783 with @xmath28 . in the following we use the term ` expansion ' but there are interesting situations where ` contraction ' yields a more appropriate point of view .
so one should think of a process having two directions that are opposite to each other .
further material about expansions can be found in @xcite .
let @xmath31 be an abelian category . recall that a full subcategory @xmath71 of @xmath31 is called exact abelian
if @xmath71 is an abelian category and the inclusion functor is exact .
now let @xmath788 be a fully faithful and exact functor between abelian categories .
it is convenient to identify @xmath71 with the essential image of @xmath181 , which means that @xmath71 is an exact abelian subcategory of @xmath31 .
we call the functor @xmath181 a _ left expansion _
if the following conditions are satisfied : 1 . the functor @xmath782 admits an exact left adjoint .
the category @xmath768 is equivalent to @xmath769 for some division ring @xmath770 .
@xmath789 for all @xmath790 .
the functor @xmath782 is called _ right expansion _ if the dual conditions are satisfied .
[ le : leftmax ] let @xmath757 be a left expansion of abelian categories
. denote by @xmath758 its left adjoint and set @xmath759 . 1 .
the category @xmath86 is a serre subcategory of @xmath31 satisfying @xmath753 and @xmath791 .
the composite @xmath792 is an equivalence and the left adjoint @xmath758 induces a quasi - inverse @xmath793 .
@xmath794 for all @xmath172 , @xmath234 , and @xmath290 .
part ( 1 ) and ( 2 ) follow from proposition [ pr : perp ] . it remains to prove ( 3 ) .
the case @xmath795 is clear . for @xmath265
, the isomorphism sends a class @xmath796 $ ] in @xmath797 to @xmath798 $ ] in @xmath799 , where @xmath800 is the unit of the adjoint pair and @xmath801 denotes the pullback of @xmath802 along @xmath803 .
an object @xmath28 in @xmath31 is called _ localizable _ if the following conditions are satisfied : 1 .
the object @xmath28 is simple .
2 . @xmath804 and @xmath805 are of finite length over @xmath806 for all @xmath172 .
3 . @xmath807 and @xmath808 for all @xmath172 .
an object @xmath28 is _ colocalizable _ if the dual conditions are satisfied .
[ le : localizable ] let @xmath31 be an abelian category and @xmath71 an exact abelian subcategory .
then the following are equivalent : 1 .
the inclusion @xmath782 is a left expansion .
2 . there exists a localizable object @xmath809 such that @xmath810 .
\(1 ) @xmath186 ( 2 ) : let @xmath28 be an indecomposable object in @xmath768 .
then @xmath28 is a simple object and @xmath811 since @xmath812 is semisimple . for each object
@xmath47 in @xmath31 , we use the natural exact sequence @xmath813 with @xmath814 and @xmath815 .
this sequence induces the following isomorphisms .
@xmath816 here we use the condition @xmath817 .
it follows that @xmath804 and @xmath805 are of finite length over @xmath806 .
now observe that the functor sending @xmath47 to @xmath818 is right exact .
thus @xmath808 by lemma [ lem : hereditary ] .
finally , @xmath810 follows from proposition [ pr : perp ] .
\(2 ) @xmath186 ( 1 ) : a left adjoint @xmath758 of the inclusion @xmath782 is constructed as follows .
fix an object @xmath47 in @xmath31 .
there exists an exact sequence @xmath819 for some @xmath820 such that @xmath821 since @xmath805 is of finite length over @xmath806 .
now choose a morphism @xmath822 such that the induced map @xmath823 is surjective and let @xmath191 be its cokernel .
it is easily checked that the composite @xmath824 is the universal morphism into @xmath825 .
thus we define @xmath826 .
next observe that the kernel and cokernel of the adjunction morphism @xmath827 belong to @xmath828 for each object @xmath47 in @xmath31 .
moreover , @xmath86 is a serre subcategory of @xmath31 since @xmath28 is simple and @xmath807 .
thus we can apply proposition [ pr : section ] and infer that the composite @xmath829 is the quotient functor with respect to @xmath86 .
therefore the left adjoint @xmath758 is exact .
we have @xmath830 by proposition [ pr : perp ] , and @xmath337 induces an equivalence @xmath831 .
thus the inclusion @xmath782 is a left expansion .
a fully faithful and exact functor @xmath782 between abelian categories is by definition an _ expansion _ of abelian categories if the functor is a left and a right expansion .
let @xmath788 be an expansion of abelian categories .
then we identify @xmath71 with the essential image of @xmath181 .
we denote by @xmath758 the left adjoint of @xmath181 and by @xmath832 the right adjoint of @xmath181 .
we choose an indecomposable object @xmath783 in @xmath768 and an indecomposable object @xmath784 in @xmath833 .
thus @xmath834 and @xmath835 . finally , set @xmath836 .
an expansion @xmath757 is called _ split _ if @xmath837 .
if the expansion is non - split , then the exact sequences for @xmath783 and @xmath784 are of the form @xmath838 for some integers @xmath839 . in lemma
[ le : extsimple ] , we see that @xmath840 .
let @xmath782 be an expansion of abelian categories .
then the following are equivalent : 1
. the expansion @xmath782 is split .
@xmath841 for some serre subcategory @xmath86 of @xmath31 .
3 . @xmath71 is a serre subcategory of @xmath31 .
\(1 ) @xmath186 ( 2 ) : take @xmath842 .
\(2 ) @xmath186 ( 3 ) : an object @xmath172 belongs to @xmath71 if and only if @xmath843 for all @xmath844 .
thus @xmath71 is closed under taking quotients and extensions .
the dual argument shows that @xmath71 is closed under taking subobjects .
\(3 ) @xmath186 ( 1 ) : if the expansion is non - split , then the sequences in show that @xmath71 is not a serre subcategory .
[ le : extsimple ] let @xmath788 be a non - split expansion of abelian categories
the object @xmath845 is a simple object in @xmath71 and isomorphic to @xmath846 .
2 . the functor @xmath758 induces an equivalence @xmath847 .
the functor @xmath832 induces an equivalence @xmath848 .
\(1 ) let @xmath849 be a non - zero morphism in @xmath71 .
adjunction takes this to a monomorphism @xmath850 in @xmath31 since @xmath784 is simple .
applying @xmath758 gives back a morphism which is isomorphic to @xmath76 .
this is a monomorphism since @xmath758 is exact .
thus @xmath851 is simple .
now apply @xmath832 to the first and @xmath758 to the second sequence in .
note that by adjunction @xmath852 .
then we have @xmath853 this implies @xmath840 and therefore @xmath854 .
\(2 ) we have a sequence of isomorphisms @xmath855 which takes a morphism @xmath76 to @xmath856 .
thus @xmath758 induces an equivalence @xmath857 .
\(3 ) follows from ( 2 ) by duality .
an expansion @xmath782 of abelian categories determines a division ring @xmath770 such that @xmath768 and @xmath833 are equivalent to @xmath769 ; we call @xmath770 the _ associated division ring_. fix an expansion @xmath757 with associated division ring @xmath770 .
we identify the perpendicular categories of @xmath71 with @xmath769 via the equivalences @xmath858 .
there are inclusions @xmath859 and @xmath860 with adjoints @xmath861 and @xmath862 .
these functors yield the following diagram .
@xmath863|-{i}&&\,\a\ , \ar@<1.0ex>[rr]|-{j_\rho}\ar@<-1.0ex>[rr]|-{k_\la } \ar@<1.2ex>[ll]^-{i_\rho}\ar@<-1.2ex>[ll]_-{i_\la } & & \,{\operatorname{mod}}\delta\ar@<2.2ex>[ll]^-{k}\ar@<-2.2ex>[ll]_-{j } } \ ] ] note that this diagram induces a recollement of triangulated categories @xcite : @xmath864|-{\bfd^b(i)}&&\,\bfd^b(\a)\ , \ar[rr]|- { } \ar@<1.2ex>[ll]^-{\bfd^b(i_\rho)}\ar@<-1.2ex>[ll]_-{\bfd^b(i_\la ) } & & \,\bfd^b({\operatorname{mod}}\delta)\ar@<1.2ex>[ll]^-{\bfd^b(k)}\ar@<-1.2ex>[ll]_-{\bfd^b(j ) } } \ ] ] indeed , the labeled functors are part of a recollement , and therefore the right adjoint of @xmath865 is isomorphic to the left adjoint of @xmath866 , both of which are isomorphic to the quotient functor of @xmath510 with respect to the triangulated subcategory @xmath867 .
let @xmath757 be an expansion .
the left adjoint @xmath758 induces a bijection between the isomorphism classes of simple objects of @xmath31 that are different from @xmath783 , and the isomorphism classes of simple objects of @xmath71 .
on the other hand , all simple objects of @xmath31 correspond to simple objects of @xmath71 via @xmath181 .
all this is made precise in the next lemma .
[ le : simple ] let @xmath757 be an expansion of abelian categories . 1 .
if @xmath28 is a simple object in @xmath71 and @xmath868 , then @xmath869 is simple in @xmath31 and @xmath870 .
2 . there is an exact sequence @xmath871 in @xmath31 , provided the expansion @xmath782 is non - split .
if @xmath28 is a simple object in @xmath31 and @xmath872 , then @xmath873 is simple in @xmath71 .
moreover , @xmath874 if @xmath875 .
\(1 ) let @xmath876 be a subobject .
then @xmath877 shows that @xmath878 .
thus @xmath879 , and therefore @xmath880 belongs to @xmath881 . on the other hand , @xmath882 . thus @xmath883 ,
and it follows that @xmath869 is simple .
finally observe that @xmath884 for every object @xmath47 in @xmath71 .
\(2 ) take the exact sequence in .
\(3 ) this is a general fact : a quotient functor @xmath131 takes each simple object of @xmath31 not belonging to @xmath86 to a simple object of @xmath91 ; see lemma [ le : quotient_simple ] . here , we take @xmath759 and identify @xmath758 with the corresponding quotient functor .
if @xmath875 , then @xmath885 and therefore @xmath886 is simple by ( 1 ) . thus the canonical morphism @xmath887 is an isomorphism .
the ext - groups of most simple objects in @xmath31 can be computed from appropriate ext - groups in @xmath71 .
this follows from an adjunction formula ; see lemma [ le : leftmax ] .
the remaining cases are treated in the following lemma .
[ le : ext_simple ] let @xmath757 be a non - split expansion of abelian categories .
1 . @xmath888 .
2 . @xmath889 for @xmath265 .
\(1 ) applying @xmath890 to the first sequence in yields the isomorphism @xmath891 .
the other isomorphism is dual .
\(2 ) we have @xmath892 where the first isomorphism follows from lemma [ le : leftmax ] and the second from the first sequence in .
[ pr : onepoint_length ]
let @xmath757 be an expansion of abelian categories . 1 .
the functor @xmath181 and its adjoints @xmath758 and @xmath832 send finite length objects to finite length objects .
the restriction @xmath893 is an expansion of abelian categories .
the induced functor @xmath894 is an equivalence .
\(1 ) follows from lemma [ le : simple ] and ( 2 ) is an immediate consequence .
\(3 ) let @xmath759 .
the functor @xmath758 induces an equivalence @xmath793 .
moreover , @xmath895 and @xmath758 identifies @xmath896 with @xmath897 by ( 1 ) .
it follows from lemma [ le : noeth ] that @xmath758 induces an equivalence @xmath898 .
this is a quasi - inverse for the functor @xmath894 induced by @xmath181 .
the ext - quiver @xmath899 of @xmath31 can be computed explicitly from the ext - quiver @xmath900 , and vice versa .
the following statement makes this precise .
let @xmath757 be a non - split expansion of abelian categories .
the functor induces a bijection @xmath901 and for each pair @xmath902 the following identities : @xmath903 combine lemmas [ le : leftmax ] , [ le : simple ] , and [ le : ext_simple ] .
the following diagram shows how @xmath900 and @xmath899 are related . into an arrow linking @xmath783 with @xmath784 justifies the term ` expansion of abelian categories ' . ]
@xmath904 \(1 ) let @xmath0 be a field and @xmath905 a quiver of extended dynkin type @xmath480 with cyclic orientation .
consider the category @xmath906 of all finite dimensional nilpotent representations .
fix a simple object @xmath28 .
this object is ( co)localizable and @xmath907 .
thus the inclusion @xmath908 is an expansion , and @xmath825 is equivalent to @xmath909 .
the following diagram depicts the shape of the _ auslander - reiten quiver _ of @xmath910 .
thus the vertices represent the indecomposable objects , and there is an arrow between two indecomposable objects if and only if there exists an irreducible morphism ; see lemma [ le : irr ] .
@xmath911 note that the two dotted lines are identified and that there are @xmath477 simple objects which sit at the bottom ( @xmath912 in this example ) .
the serre functor @xmath367 induces an automorphism of order @xmath477 , that is , @xmath913 .
the indecomposables not belonging to @xmath825 are represented by circles .
thus @xmath28 and @xmath914 disappear , while @xmath915 becomes a simple object in @xmath825 .
suppose that @xmath916 .
then @xmath915 is a ( co)localizable object of @xmath825 and this gives another expansion @xmath917 .
iterating this formation of perpendicular categories yields a chain @xmath918 of expansions such that @xmath919 is equivalent to @xmath920 for each @xmath181 .
the category @xmath921 has a unique simple object @xmath922 which the inclusion @xmath923 sends to @xmath414}$ ] .
the induced map @xmath924 sends the class @xmath925 $ ] to @xmath926$].$ ] in @xmath674 again explains the term ` expansion of abelian categories ' . ] \(2 ) let @xmath0 be a field and consider a finite quiver @xmath41 without oriented cycles having two vertices @xmath927 that are joined by an arrow @xmath928 which is the unique arrow starting at @xmath929 and the unique arrow terminating at @xmath930 .
@xmath931 we obtain a new quiver @xmath932 by identifying @xmath929 and @xmath930 and removing @xmath220 .
let @xmath933 and consider the full subcategory @xmath71 of representations such that @xmath220 is represented by an isomorphism .
note that @xmath71 is equivalent to @xmath934 .
the simple representations @xmath935 and @xmath936 are ( co)localizing objects of @xmath31 and they are joined by an almost split sequence @xmath937 . the auslander - reiten formulae @xmath938 imply that @xmath939 .
thus the inclusion functor @xmath782 is an expansion
. given a non - split expansion @xmath782 , the corresponding simple objects @xmath783 and @xmath784 in @xmath31 are related by an auslander - reiten formula .
[ pr : arformula ] let @xmath782 be a non - split expansion of abelian categories and @xmath770 its associated division ring .
then @xmath940 where @xmath941 denotes the standard duality .
in particular , any non - split extension @xmath942 is an almost split sequence .
recall that @xmath943 and @xmath835 .
fix an object @xmath47 in @xmath31 and consider the corresponding exact sequence @xmath944 with @xmath765 in @xmath945 and @xmath191 in @xmath71 .
the morphism @xmath946 induces the first and the third isomorphism in the sequence below , while the second isomorphism follows from the isomorphism @xmath891 in lemma [ le : ext_simple ] .
@xmath947 the isomorphism @xmath948 follows from the first by duality .
the argument given in the proof of proposition [ pr : serreduality ] shows that any non - split extension @xmath942 is an almost split sequence .
expansions of abelian categories respect decompositions of abelian categories .
the following lemma is a precise formulation of this fact .
[ le : decomp ] let @xmath757 be a non - split expansion of abelian categories . 1 .
if @xmath59 is a decomposition , then there exists a decomposition @xmath949}\colon \b_1\amalg\b_2\to\a_1\amalg\a_2 $ ] such that one of @xmath950 and @xmath951 is a non - split expansion and the other is an equivalence .
if @xmath952 is a decomposition , then there exists a decomposition @xmath949}\colon \b_1\amalg\b_2\to\a_1\amalg\a_2 $ ] such that one of @xmath950 and @xmath951 is a non - split expansion and the other is an equivalence .
if @xmath953 is an equivalence of abelian categories , then @xmath954}\colon \b\amalg\b'\to\a\amalg\a'$ ] is a non - split expansion .
therefore @xmath31 is connected if and only if @xmath71 is connected .
we identify @xmath71 with the essential image of @xmath181 and choose a simple object @xmath28 in @xmath31 with @xmath810 .
\(1 ) the decomposition of @xmath31 restricts to a decomposition of @xmath71 by taking @xmath955 for @xmath956 .
now suppose without loss of generality that @xmath28 belongs to @xmath60 .
then @xmath950 is a non - split expansion and @xmath951 is an equivalence .
\(2 ) let @xmath957 and suppose without loss of generality that @xmath915 belongs to @xmath958 . then @xmath959 and this yields a decomposition @xmath59 if we set @xmath960 and @xmath961 .
it follows that @xmath950 is a non - split expansion and @xmath951 is an equivalence .
\(3 ) clear .
we compute the global dimension for an expansion of abelian categories . [
le : cohdim ] let @xmath757 be a non - split expansion of abelian categories . 1 .
@xmath31 has non - zero projective objects if and only if @xmath71 has non - zero projective objects .
\(1 ) use the adjunction formula for @xmath963 from lemma [ le : leftmax ] together with the fact that @xmath964 for all @xmath47 in @xmath768 .
note also that @xmath965 by lemma [ le : ext_simple ] .
\(2 ) we use the general fact that a functor between abelian categories preserves projectivity if it admits an exact right adjoint .
thus @xmath181 and its left adjoint @xmath758 preserve projectivity .
note that there are no non - zero projectives in the kernel of @xmath758 by lemma [ le : ext_simple ] .
next we discuss ext - finiteness for expansions of abelian categories .
[ le : extfin ] let @xmath0 be a commutative ring and @xmath757 a non - split expansion of @xmath0-linear abelian categories . then @xmath31 is ext - finite if and only if @xmath71 is ext - finite .
we use the adjunction formula for @xmath963 from lemma [ le : leftmax ] . note that these isomorphisms are @xmath0-linear since we assume the functor @xmath181 to be @xmath0-linear .
it is clear that @xmath71 is ext - finite if @xmath31 is ext - finite . to prove the converse ,
fix a simple object @xmath28 in @xmath31 such that @xmath810 , and an arbitrary object @xmath109 in @xmath31 .
then @xmath966 for some simple object @xmath915 in @xmath71 ; see lemma [ le : extsimple ] .
thus @xmath806 is of finite length over @xmath0 , and it follows that @xmath967 is of finite length over @xmath0 for all @xmath820 since the object @xmath28 is localizable ; see lemma [ le : localizable ] . on the other hand
, @xmath968 is of finite length over @xmath0 for all @xmath72 in @xmath71 by the adjunction formula for @xmath963 from lemma [ le : leftmax ] .
now choose @xmath47 in @xmath31 and apply @xmath969 to the natural exact sequence @xmath813 with @xmath765 in @xmath970 and @xmath191 in @xmath71 .
it follows that @xmath971 is of finite length over @xmath0 for all @xmath820 .
in this section we discuss the category of coherent sheaves on the projective line @xmath1 over an arbitrary base field @xmath0 .
this is a hereditary abelian category with finite dimensional hom and ext spaces .
moreover , the category satisfies serre duality and admits a tilting object .
various structural properties can be derived from these basic facts . the projective line @xmath1 is covered by two copies of the affine line @xmath972 .
using this fact , we identify sheaves on @xmath1 with ` triples ' , that is , pairs of modules over the polynomial ring @xmath298 $ ] which are glued together by a glueing morphism . in this way ,
basic properties of sheaves are easily derived from properties of modules over a polynomial ring in one variable .
every coherent sheaf is the direct sum of a torsion - free sheaf and a finite length sheaf .
an indecomposable torsion - free sheaf is a line bundle and an indecomposable finite length sheaf is uniserial .
the simple sheaves are parametrized by the closed points of @xmath1 , and for each simple sheaf @xmath28 and @xmath973 there is a unique sheaf with length @xmath974 and top @xmath28 .
this yields a complete classification of all indecomposable sheaves .
a classical theorem of serre identifies the category of coherent sheaves on @xmath1 with the quotient category of the category of finitely generated @xmath297-graded @xmath975$]-modules modulo the serre subcategory of finite length modules .
we use the concept of dehomogenization to pass from graded @xmath975$]-modules to modules over @xmath298 $ ] . in geometric terms , this reflects the passage from the projective to an affine line .
let @xmath0 be a field and @xmath972 the affine line over @xmath0 .
the polynomial ring @xmath298 $ ] is the ring of regular functions and the category of coherent sheaves @xmath976 is equivalent to the module category @xmath977 $ ] via the global section functor .
let @xmath978 $ ] denote the set of prime ideals of @xmath298 $ ] .
note that @xmath298 $ ] is a principal ideal domain .
thus irreducible polynomials correspond to non - zero prime ideals by taking a polynomial @xmath552 to the ideal @xmath979 generated by @xmath552 .
a _ closed point _ of @xmath972 is by definition a non - zero prime ideal @xmath980 and the _ generic point _
is the zero ideal .
the following result describes the category @xmath981 $ ] of torsion modules and the corresponding quotient category .
[ pr : aff ] let @xmath298 $ ] be the polynomial ring over a field @xmath0 . 1 .
the functor which sends a @xmath298$]-module @xmath155 to its family of localizations @xmath982}$ ] induces an equivalence @xmath983{\stackrel{\sim}{\longrightarrow}}\coprod_{0\neq\frp\in\operatorname{spec}k[x]}{\operatorname{mod}}_0(k[x]_\frp).\ ] ] 2 .
the localization functor @xmath977\to{\operatorname{mod}}k(x)$ ] induces an equivalence @xmath984}{{\operatorname{mod}}_0 k[x]}{\stackrel{\sim}{\longrightarrow}}{\operatorname{mod}}k(x).\ ] ] \(1 ) the assertion follows from standard properties of finitely generated modules over principal ideal domains .
note that the quasi - inverse functor takes a family of modules @xmath985}$ ] to @xmath986 .
\(2 ) set @xmath987/{\operatorname{mod}}_0 k[x]$ ] .
the kernel of the localization functor @xmath988}k(x)$ ] is the category @xmath981 $ ] .
thus @xmath32 induces a faithful functor @xmath989 .
the functor is dense , since a @xmath990-module of rank @xmath974 is isomorphic to @xmath991^r)$ ] . to show that @xmath992 is full
, it suffices to show that @xmath992 induces a surjective map @xmath993,k[x]){\stackrel{{}}{\longrightarrow}}\operatorname{hom}_{k(x)}(k(x),k(x))\cong k(x).\ ] ] given any non - zero polynomial @xmath436 $ ] , let @xmath994\to k[x]$ ] denote the multiplication by @xmath552 .
the kernel and cokernel of @xmath995 belong to @xmath981 $ ] , and therefore @xmath995 becomes invertible in @xmath31 . thus @xmath996 .
it follows that @xmath997 is surjective .
let @xmath0 be a field and @xmath1 the projective line over @xmath0 .
we view @xmath1 as a scheme and begin with a description of the underlying set of points .
let @xmath975 $ ] be the polynomial ring in two variables with the usual @xmath297-grading by total degree .
denote by @xmath998 $ ] the set of homogeneous prime ideals of @xmath975 $ ] that are different from the unique maximal ideal consisting of positive degree elements .
note that @xmath975 $ ] is a two - dimensional graded factorial domain .
thus homogeneous irreducible polynomials correspond to non - zero homogeneous prime ideals by taking a polynomial @xmath552 to the ideal @xmath979 generated by @xmath552 .
a _ closed point _ of @xmath1 is by definition an element @xmath999 in @xmath998 $ ] , and the _ generic point _
is the zero ideal . using homogeneous coordinates , a _
rational point _ of @xmath1 is a pair @xmath1000 $ ] of elements of @xmath0 subject to the relation @xmath1000=[\alpha \la_0:\alpha \la_1]$ ] for all @xmath1001 , @xmath1002 .
we identify each rational point @xmath1000 $ ] with the prime ideal @xmath1003 of @xmath975 $ ] . using the identification @xmath1004
, we cover @xmath1 by two copies @xmath1005 $ ] and @xmath1006 $ ] of the affine line , with @xmath1007 $ ] .
more precisely , the morphism @xmath975\to k[y]$ ] which sends a polynomial @xmath552 to @xmath1008 induces a bijection @xmath1009\smallsetminus\{(x_0)\}{\stackrel{\sim}{\longrightarrow}}\operatorname{spec}k[y];\ ] ] see [ se : dehom ] for details .
analogously , the morphism @xmath975\to k[y^{-1}]$ ] which sends a polynomial @xmath552 to @xmath1010 induces a bijection @xmath1009\smallsetminus\{(x_1)\}{\stackrel{\sim}{\longrightarrow}}\operatorname{spec}k[y^{-1}].\ ] ] based on the covering @xmath1011 , the category @xmath1012 of coherent sheaves admits a description in terms of the following pullback of abelian categories @xmath1013\ar[d]&\operatorname{coh}u'\ar[d]\\\operatorname{coh}u''\ar[r]&\operatorname{coh}u'\cap u '' } \ ] ] where each functor is given by restricting a sheaf to the appropriate open subset ; see ( * ? ? ? * chap .
vi , prop . 2 ) . more concretely , this pullback diagram has , up to equivalence , the form @xmath1014\ar[d]&{\operatorname{mod}}k[y]\ar[d]\\ { \operatorname{mod}}k[y^{-1}]\ar[r]&{\operatorname{mod}}k[y , y^{-1}]}\ ] ] where the category @xmath31 is defined as follows .
the objects of @xmath31 are triples @xmath1015 , where @xmath613 is a finitely generated @xmath1016$]-module , @xmath1017 is a finitely generated @xmath1018$]-module , and @xmath1019 is an isomorphism of @xmath1020$]-modules . here , we use for any @xmath1021-module @xmath155 the notation @xmath1022 to denote the localization with respect to an element @xmath1023 . a morphism from @xmath1015 to @xmath1024 in @xmath31 is a pair @xmath1025 of morphisms , where @xmath1026 is @xmath1016$]-linear and @xmath1027 is @xmath1018$]-linear such that @xmath1028 . given a sheaf @xmath1029 on @xmath1
, we denote for any open subset @xmath1030 by @xmath1031 the _ sections _ over @xmath1032 . the functor which sends a coherent sheaf @xmath1029 on @xmath1 to the triple @xmath1033 gives an equivalence @xmath1034 . the description of a sheaf @xmath1029 on @xmath1011 in terms of its restrictions @xmath1035 , @xmath1036 , and @xmath1037 is classical ; see ( * ? ? ? * chap .
vi , prop . 2 ) .
thus it remains to observe that taking global sections identifies @xmath1038 $ ] , @xmath1039 $ ] , and @xmath1040 $ ] . from now on
we identify the categories @xmath1012 and @xmath31 via the above equivalence .
let @xmath1041 $ ] denote the category of finitely generated @xmath297-graded @xmath975$]-modules and let @xmath1042 $ ] be the serre subcategory consisting of all finite length objects .
there is a functor @xmath1043{\stackrel{{}}{\longrightarrow}}\operatorname{coh}\bbp^1_k\ ] ] that takes each graded @xmath975$]-module @xmath155 to the triple @xmath1044 , where @xmath1045 acts on the degree zero part of @xmath1046 via the identification @xmath1004 , the variable @xmath1047 acts on the degree zero part of @xmath1048 via the identification @xmath1049 , and the isomorphism @xmath1050 equals the obvious identification @xmath1051_{x_1/x_0}=[(m_{x_1})_0]_{x_0/x_1}$ ] .
[ pr : serre ] the functor induces an equivalence @xmath1052}{{\operatorname{mod}}_0^\bbz k[x_0,x_1]}{\stackrel{\sim}{\longrightarrow}}\operatorname{coh}\bbp^1_k.\ ] ] we refer to @xcite for the proof .
it is clear from the definition that the functor is exact having kernel @xmath1053}$ ] .
this fact yields the induced functor which is faithful by construction . for any @xmath485 and @xmath1054 in @xmath1012 , denote by @xmath1055 the _ twisted sheaf _
@xmath1056 , where @xmath1057 is the map @xmath1058 followed by multiplication with @xmath1059 . given a module @xmath155 in @xmath1060 $ ] , the _ twisted module _
@xmath1061 is obtained by shifting the grading , that is , @xmath1062 for @xmath1063 .
note that the functor is compatible with the twist functors defined on @xmath1041 $ ] and @xmath1012 .
the category @xmath1012 is a hom - finite @xmath0-linear abelian category satisfying serre duality .
more precisely , there is a functorial @xmath0-linear isomorphism @xmath1064 see ( * ? ? ?
* iii.7 ) .
a sheaf @xmath1015 in @xmath1012 is called _ locally free _ or _ vector bundle _ if @xmath613 and @xmath1017 are free modules over @xmath1016 $ ] and @xmath1018 $ ] respectively .
we denote the full subcategory of vector bundles in @xmath1012 by @xmath1065 .
the _ structure sheaf _ is the sheaf @xmath1066,k[y^{-1 } ] , \operatorname{id}_{k[y , y^{-1}]})$ ] . for any pair @xmath1067
, we have a natural bijection @xmath1068_{n - m}{\stackrel{\sim}{\longrightarrow}}\operatorname{hom}(\oc(m),\oc(n)).\ ] ] the map sends a homogeneous polynomial @xmath552 of degree @xmath1069 to the morphism @xmath1025 , where @xmath1070\to k[y]$ ] is multiplication by @xmath1008 and @xmath1071\to k[y^{-1}]$ ] is multiplication by @xmath1010 .
this bijection is a special case of the next result .
let @xmath1072 $ ] and denote by @xmath1073 the category of finitely generated projective @xmath297-graded @xmath1021-modules .
note that @xmath1021 is a graded local ring since the homogeneous elements of positive degree form the unique maximal homogeneous ideal .
thus finitely generated projective @xmath1021-modules are up to isomorphism of the form @xmath1074 .
[ pr : grothendieck ] the functor induces an equivalence @xmath1075{\stackrel{\sim}{\longrightarrow}}\operatorname{vect}\bbp^1_k.\ ] ] in particular , each locally free coherent sheaf on @xmath1 is isomorphic to a sheaf of the form @xmath1076 .
we need to show that the functor is fully faithful when it is restricted to @xmath1073 , where @xmath1077 $ ] .
every finitely generated projective @xmath1021-module is up to isomorphism of the form @xmath1074 .
thus it suffices to show that induces a bijection @xmath1078 for each pair @xmath1067 .
but this is clear since the map coincides with the bijection .
it remains to show that the functor is dense , that is , each locally free coherent sheaf is isomorphic to one of the form @xmath1079 .
for this we refer to @xcite .
an elementary proof is based on an argument due to birkhoff @xcite . a vector bundle @xmath1054 is basically determined by an invertible matrix over @xmath1020 $ ] , which represents the isomorphism @xmath1058 .
now one uses the fact that such a matrix can be transformed into a diagonal matrix with entries @xmath1080 by multiplying it with an invertible matrix over @xmath1016 $ ] from the right and an invertible matrix over @xmath1018 $ ] from the left .
this yields an isomorphism @xmath1081 .
a sheaf @xmath1015 in @xmath1012 is called _ torsion _ if @xmath613 and @xmath1017 are torsion modules over @xmath1016 $ ] and @xmath1018 $ ] respectively .
we denote the full subcategory of torsion sheaves in @xmath1012 by @xmath1082 .
note that each sheaf @xmath1054 admits a unique maximal subobject @xmath1083 in @xmath1012 that is torsion .
one obtains @xmath1083 by taking the torsion parts of @xmath613 and @xmath1017 respectively .
clearly , @xmath1084 is locally free .
each coherent sheaf @xmath1029 on @xmath1 admits an essentially unique decomposition @xmath1085 such that @xmath1086 is torsion and @xmath1087 is locally free .
the torsion sheaves are precisely the objects of finite length in @xmath1012 .
we use elementary facts about finitely generated modules over @xmath1016 $ ] and @xmath1018 $ ] respectively . take @xmath1088 and @xmath1089 .
serre duality implies @xmath1090 , since there are no non - zero morphisms from torsion to locally free sheaves .
thus the canonical exact sequence @xmath1091 splits .
let @xmath1054 be a coherent sheaf .
if @xmath1029 is torsion , then it is a finite length object in @xmath1012 , since the corresponding torsion modules @xmath613 and @xmath1017 have finite length . on the other hand
, the structure sheaf has infinite length , since it admits factor objects of arbitrary length .
thus each non - zero locally free sheaf is of infinite length , by proposition [ pr : grothendieck ] .
let @xmath1021 be a @xmath297-graded commutative ring and @xmath1023 a non - nilpotent element of degree @xmath183 .
the _ dehomogenization _ of @xmath1021 with respect to @xmath437 is the ring @xmath1092 .
[ le : dehom ] the canonical morphism @xmath1093 induces an isomorphism @xmath1094 .
moreover , @xmath1095 induces a bijection between the set of homogeneous ( prime ) ideals of @xmath1021 modulo which the element @xmath437 is regular and the set of all ( prime ) ideals of @xmath1092 . the morphism @xmath1095 induces a morphism @xmath1096 and its kernel is the ideal generated by @xmath1097 .
now observe that @xmath1098 , since @xmath437 is in @xmath28 a unit of degree @xmath183 . the bijective correspondence between ideals of @xmath1021 and @xmath1092 sends an ideal @xmath1099 to @xmath1100 , and its inverse sends an ideal @xmath1101 to @xmath1102 .
let us consider the dehomogenization of the polynomial ring @xmath975 $ ] with respect to the variable @xmath1103 .
then lemma [ le : dehom ] implies an isomorphism @xmath1104/(x_0 - 1)\cong ( k[x_0,x_1]_{x_0})_0=k[y]\ ] ] via the identification @xmath1004 .
denote by @xmath1105\to k[y]$ ] the canonical morphism which sends a polynomial @xmath552 to @xmath1008 .
the morphism @xmath1095 induces a bijection @xmath1106\smallsetminus\{(x_0)\}{\stackrel{\sim}{\longrightarrow}}\operatorname{spec}k[y]\ ] ] and for any prime ideal @xmath1107 in @xmath998 $ ] an isomorphism @xmath1108_\frp)_0{\stackrel{\sim}{\longrightarrow}}k[y]_{\pi(\frp)}.\ ] ] here , @xmath975_\frp$ ] denotes the homogeneous localization with respect to @xmath980 which inverts all homogeneous elements not lying in @xmath980 .
the following lemma describes the dehomogenization for graded modules over @xmath975 $ ] with respect to @xmath1103 .
this functor induces an equivalence if one passes to the localization with respect to a prime ideal @xmath1107 .
[ le : dehom_mod ] let @xmath1107 be in @xmath998 $ ] .
the functor sending a graded @xmath975$]-module @xmath155 to @xmath1109 induces the following commutative diagram of exact functors @xmath1110\ar[rr]^-f\ar[d]&&{\operatorname{mod}}k[y]\ar[d]\\ { \operatorname{mod}}^\bbz k[x_0,x_1]_\frp\ar[r]^-{f'_\frp}&{\operatorname{mod}}(k[x_0,x_1]_\frp)_0\ar[r]^-{f''_\frp}&{\operatorname{mod}}k[y]_{\pi(\frp ) } } \ ] ] where the vertical functors are the localization functors with respect to @xmath980 and @xmath1111 respectively .
moreover , the functors @xmath1112 and @xmath1113 are equivalences .
the composite @xmath975\to k[y]\to k[y]_{\pi(\frp)}$ ] induces a morphism @xmath975_\frp\to k[y]_{\pi(\frp)}$ ] and therefore @xmath81 composed with localization at @xmath1111 induces a functor @xmath1114_\frp{\stackrel{{}}{\longrightarrow}}{\operatorname{mod}}k[y]_{\pi(\frp)}.\ ] ] this functor can be written as composite @xmath1115 .
the first functor @xmath1112 takes a graded @xmath975_\frp$]-module to its degree zero part ; it is an equivalence since @xmath975_\frp$ ] is strongly graded .
the second functor @xmath1113 is an equivalence thanks to the isomorphism .
there are analogous results for the dehomogenization of @xmath975 $ ] with respect to @xmath1116 which is denoted by @xmath1018 $ ] .
next we describe the category @xmath1117 of torsion sheaves more explicitly .
note that @xmath1117 is uniserial because it is a length category with serre duality ; see proposition [ pr : serre_finlen ] .
the category decomposes into connected abelian categories , and each component has a unique simple object since it is equivalent to the category of finite length modules over a local ring .
[ pr : coh_p ] 1 .
the functor induces an equivalence @xmath1118}{\operatorname{mod}}^\bbz_0 k[x_0,x_1]_\frp{\stackrel{\sim}{\longrightarrow}}\operatorname{coh}_0\bbp^1_k\ ] ] 2 .
the functor taking a sheaf @xmath1015 to @xmath1119 } k(y)\cong m''\otimes_{k[y^{-1 } ] } k(y^{-1})$ ] induces an equivalence @xmath1120 \(1 ) we denote the functor by @xmath81 .
dehomogenization with respect to @xmath1103 and @xmath1116 , respectively , induces the functors @xmath1121 , @xmath1122 , and @xmath1123 , which make the following diagram commutative .
\ar[rrr]\ar[ddd]\ar[rd]^-{f } & & & \underset{\frp\neq ( x_0)}{\coprod}{\operatorname{mod}}^\bbz_0 k[x_0,x_1]_\frp \ar@{-}[d]\ar[rd]^-{f_0}\\ & \operatorname{coh}_0\bbp^1_k\ar[rrr]\ar[ddd]&&{}\ar[dd]&{\operatorname{mod}}_0 k[y]\ar[ddd]\\ \\ \underset{\frp\neq ( x_1)}{\coprod}{\operatorname{mod}}^\bbz_0 k[x_0,x_1]_\frp \ar@{-}[r]\ar[rd]^-{f_1}&{}\ar[rr ] & & \underset{(x_0)\neq\frp\neq ( x_1)}{\coprod}{\operatorname{mod}}^\bbz_0 k[x_0,x_1]_\frp \ar[rd]^-{f_{0,1}}\\ & { \operatorname{mod}}_0 k[y^{-1}]\ar[rrr]&&&{\operatorname{mod}}_0 k[y , y^{-1}]}\ ] ] here , @xmath980 runs through all non - zero prime ideals in @xmath998 $ ] .
note that the front square and the back square are pullback diagrams of abelian categories .
the functors @xmath1121 , @xmath1122 , and @xmath1123 are equivalences .
this follows from proposition [ pr : aff ] and lemma [ le : dehom_mod ] in combination with the bijection .
thus @xmath81 is an equivalence .
\(2 ) the functor @xmath1125 is exact and its kernel is the category @xmath1117 of torsion sheaves .
thus there is an induced functor @xmath1126 which is faithful .
the structure sheaf is the unique indecomposable object , and the argument given in the proof of proposition [ pr : aff ] shows that the functor is an equivalence .
given a point @xmath1127 $ ] of @xmath1 , the _ associated local ring _
@xmath1128 is by definition @xmath1129_\frp)_0 $ ] .
denote by @xmath1130 the _ residue field _ at @xmath980 which is by definition the residue field of the local ring @xmath1128 .
note that dehomogenization induces isomorphisms @xmath1131_{\frp'}&\text{if } \frp\neq ( x_0),\\ k[y^{-1}]_{\frp''}&\text{if } \frp\neq ( x_1 ) , \end{cases}\ ] ] where @xmath1132\mid p\in \frp\}\quad\text{and}\quad\frp''=\{p(y^{-1},1)\in k[y^{-1}]\mid p\in \frp\}.\ ] ] the _ stalk _ of a sheaf @xmath1054 at @xmath980 is the @xmath1128-module @xmath1133 where @xmath1134 and @xmath1135 are identified via @xmath1058 , if @xmath1136 for @xmath1137 . the _ support _ of @xmath1029 is by definition @xmath1138\mid\f_\frp\neq 0\}.\ ] ] note that a torsion sheaf @xmath1029 admits a unique decomposition @xmath1139}\f_{\{\frp\}}\ ] ] such that each @xmath1140 is a sheaf supported at @xmath980 .
this follows directly from properties of torsion modules over a polynomial ring in one variable .
the functor sending a sheaf @xmath1029 to the family @xmath1141}$ ] provides an equivalence @xmath1142}{\operatorname{mod}}_0\oc_{\bbp^1_k,\frp}.\ ] ] note that this yields a quasi - inverse for the equivalence from proposition [ pr : coh_p ] if the functor is composed with the family of equivalences @xmath1143_\frp$ ] from lemma [ le : dehom_mod ] .
let @xmath980 be a closed point and choose a homogeneous irreducible polynomial @xmath552 of degree @xmath1144 that generates @xmath980 .
the bijection identifies for each integer @xmath973 the polynomial @xmath1145 with a morphism @xmath1146 .
denote by @xmath1147 the cokernel of this morphism .
thus there is an exact sequence @xmath1148 let @xmath980 be a closed point of @xmath1 and @xmath973 an integer . an indecomposable sheaf in @xmath1012
has support @xmath1149 and length @xmath974 if and only if it is isomorphic to @xmath1147 .
we compute the support and the length of @xmath1147 .
we may assume that @xmath1107 .
the morphism @xmath76 is given by multiplication with @xmath1150 and @xmath1151 respectively .
thus for each point @xmath1152 , the stalk morphism @xmath1153 is an isomorphism , and therefore @xmath1154 . on the other hand ,
the @xmath1128-module @xmath1155 is isomorphic to @xmath1156 , where @xmath1157 denotes the maximal ideal of @xmath1128 .
note that this module is indecomposable .
thus @xmath1147 has length @xmath974 and is indecomposable .
the equivalence shows that an indecomposable torsion sheaf is uniquely determined by its support and its length .
[ re : hom - simple ] let @xmath1158 be the simple sheaf concentrated at @xmath980 .
the sequence induces an isomorphism @xmath1159 .
moreover , we have an isomorphism @xmath1160 of algebras .
let @xmath1161 denote the _ projective linear group _ , that is , the group of invertible @xmath1162 matrices over @xmath0 modulo the subgroup of matrices of the form @xmath1163}$ ] .
any element @xmath1164}$ ] in @xmath1161 induces an automorphism @xmath975{\xrightarrow}{\sim}k[x_0,x_1]$ ] by sending @xmath1165 to @xmath1166 ( @xmath1167 ) .
this yields a map @xmath1168 into the automorphism group of the projective line . recall that a rational point @xmath1000 $ ] is identified with the prime ideal @xmath1169 .
then the automorphism corresponding to @xmath133 sends a rational point @xmath1000 $ ] to @xmath1170 $ ] .
[ pr : aut ] the map @xmath1168 is an isomorphism . it is clear that the map is injective .
we provide an inverse map as follows .
let @xmath1171 be an automorphism .
this morphism sends rational points to rational points .
in particular for @xmath1167 the inverse @xmath1172 sends the prime ideal @xmath1173 to an ideal of the form @xmath1174 for some homogenous irreducible polynomial @xmath1175 in @xmath1176 $ ] .
let @xmath1177 and denote by @xmath1178 its image under @xmath76 .
then @xmath76 induces isomorphisms of affine lines @xmath1179=u'_0 { \stackrel{\sim}{\longrightarrow } } u_0=\operatorname{spec}k[x_1/x_0]\ ] ] and @xmath1180=u'_1{\stackrel{\sim}{\longrightarrow } } u_1=\operatorname{spec}k[x_0/x_1]\ ] ] which preserve the origins .
thus there are non - zero scalars @xmath927 such that @xmath1181 here , @xmath1182 denotes the induced morphism between the rings of regular functions ( @xmath1137 ) .
the morphisms @xmath1183 and @xmath1184 agree on @xmath1185 , and therefore @xmath1186 .
it follows that @xmath76 is given by the linear transformation @xmath1187}$ ] in @xmath1161 .
the category @xmath1012 admits a tilting object which is actually unique up to a shift and up to multiplicities of its indecomposable direct summands .
[ pr : coh_tilt ] an object @xmath32 in @xmath1012 is a tilting object if and only if @xmath1188 for each @xmath485 , the endomorphism algebra of the tilting object @xmath1189 is isomorphic to the kronecker algebra @xmath82 ( i.e. the path algebra of the quiver @xmath1190\ar@<-.5ex>[r]&\cdot}$ ] ) , and this yields a derived equivalence @xmath1191 . consider @xmath1192 .
we apply the bijection and serre duality .
this yields @xmath1193 .
let @xmath1029 be an indecomposable sheaf .
if @xmath1029 is torsion , then @xmath1194 ; see [ se : support ] .
if @xmath1029 is locally free , say @xmath1195 , then @xmath1196 if @xmath626 , and @xmath1197 if @xmath269 .
thus @xmath1192 is a tilting object , and its endomorphism algebra equals the kronecker algebra . from this
it follows that any object @xmath32 in @xmath1012 satisfying is a tilting object .
now let @xmath32 be any tilting object in @xmath1012 .
then @xmath32 is locally free since any non - zero torsion sheaf @xmath1029 has @xmath1198 .
another application of the bijection and serre duality yields the condition .
the derived equivalence is a consequence of theorem [ th : tilt ] .
the grothendieck group of @xmath1012 is free of rank two and the corresponding euler form is non - degenerate .
following work of lenzing @xcite , we describe the abelian categories that arise as categories of coherent sheaves on weighted projective lines .
we provide two different approaches : a list of axioms and a description in terms of expansions of abelian categories .
the axioms basically say that these abelian categories are hereditary and noetherian , admit a tilting object , and have no non - zero projective objects .
we collect some direct consequences of these axioms .
in particular , we investigate the quotient category modulo the serre subcategory of finite length objects ; it is a length category with a unique simple object .
an abelian category satisfying these axioms has a grothendieck group that is free of finite rank .
we show that the rank is minimal if and only if the category is equivalent to the category of coherent sheaves on the projective line .
the axioms are invariant under forming expansions .
moreover , an expansion increases the rank of the grothendieck group by one .
thus the formation of expansions reflects the insertion of weights for specific points of the projective line .
these observations provide the basis for the final description of categories of coherent sheaves on weighted projective lines . throughout this section
we fix an arbitrary field @xmath0 .
a _ weighted projective line _ over a field @xmath0 is by definition a triple @xmath1199 , where @xmath2 is a ( possibly empty ) collection of distinct closed points of the projective line @xmath1 , and @xmath3 is a _ weight sequence _ , that is , a sequence of positive integers . in this work
we make the additional assumption that the closed points of @xmath1200 are supposed to be _
this assumption simplifies our exposition .
in fact , there is no substantial difference between this case and the case where the field @xmath0 is algebraically closed .
we refer to the introduction for the definition of the category @xmath5 of coherent sheaves on a weighted projective line @xmath6 .
let us remark that since the field @xmath0 is not necessarily algebraically closed , the @xmath7-graded algebra @xmath8 , even up to isomorphism , might depend on the choice of the homogeneous coordinates @xmath1201 for each @xmath1202 $ ] .
however , up to equivalence the associated category @xmath5 of coherent sheaves is independent of this choice ; see corollary [ co : wpl_wf ] .
let @xmath31 be a @xmath0-linear abelian category .
we consider the following set of axioms : 1 .
the category @xmath31 is skeletally small , connected , and ext - finite .
the category @xmath31 is noetherian , that is , each object of @xmath31 is noetherian .
the category @xmath31 is hereditary and has no non - zero projective object .
4 . the category @xmath31 has a tilting object .
the euler form associated to @xmath31 is non - degenerate and has discriminant @xmath1203 .
let us collect the basic properties of a category satisfying ( h1)(h4 ) so that we can use them from now on freely without any further reference .
recall that @xmath302 denotes the full subcategory consisting of all finite length objects in @xmath31 and that @xmath443 is the full subcategory consisting of all objects @xmath47 in @xmath31 satisfying @xmath444 for all @xmath445 in @xmath302 .
one should think of objects in @xmath302 as _ torsion objects _
, whereas the objects in @xmath443 are _ torsion - free _ or _
vector bundles_. a @xmath0-linear abelian category @xmath31 satisfying _ ( h1)(h4 ) _ has the following properties : 1 .
the category @xmath31 admits a serre functor @xmath1204 .
the grothendieck group @xmath674 is free of finite rank and the euler form associated to @xmath31 is non - degenerate .
every object in @xmath31 is a direct sum of an object in @xmath302 and an object in @xmath443 .
4 . the category of finite length objects admits a decomposition @xmath1205 into connected uniserial categories .
\(1 ) follows from proposition [ pr : serre ] , ( 2 ) from proposition [ pr : euler ] , ( 3 ) from proposition [ pr : serreduality ] , and ( 4 ) from proposition [ pr : serre_finlen ] .
the noetherianness of @xmath31 implies that @xmath302 is non - trivial .
on the other hand , @xmath1206 because a length category with serre duality and a grothendieck group of finite rank has a degenerate euler form ; see example [ ex : degen ]
. denote by @xmath305 a set of representatives of the isomorphism classes of simple objects of @xmath31 .
we identify the set @xmath1207 of @xmath367-orbits with the index set @xmath1208 of the decomposition @xmath1205 . for each @xmath1209 , let @xmath1210 denote the number of isomorphism classes of simple objects of @xmath1211 .
each @xmath1210 is finite .
more precisely , @xmath1212 is bounded by the rank of @xmath674 .
first observe that for two simple objects @xmath1213 , we have @xmath1214 if and only if @xmath1215 ; see theorem [ th : gabriel ] and proposition [ pr : serreduality ] .
now choose for each @xmath1209 a simple object @xmath1216 and let @xmath1217 .
the set @xmath1218 admits a linear ordering such that @xmath1219 implies @xmath1220,[t]\rangle = 0 $ ] .
it follows that the corresponding classes @xmath748 $ ] , @xmath1221 , are linearly independent in @xmath674 ; see lemma [ le : lin ] below .
thus @xmath1222 is bounded by the rank of @xmath674 .
[ le : lin ] let @xmath223 be an abelian group and @xmath1223 a subset .
suppose there is a non - degenerate bilinear form @xmath76 on @xmath223 and a linear ordering on @xmath500 such that @xmath1224 for all @xmath1225 and @xmath1226 for all @xmath1227 in @xmath500 .
then @xmath500 is linearly independent .
straightforward .
[ le : linear - rank ] there exists a linear map @xmath1228 such that 1 .
@xmath1229)\geq 0 $ ] for all @xmath172 ; 2 .
@xmath1229)=0 $ ] if and only if @xmath1230 ; 3 .
@xmath1231)=\rho ( [ a])$ ] for all @xmath172 . for each @xmath1209 choose a simple object @xmath1232 and let @xmath1233 $ ] .
choose elements @xmath1234 from @xmath1208 such that the subgroup of @xmath674 generated by @xmath1235 contains each @xmath1236 and set @xmath1237 .
serre duality implies that @xmath1238,w\rangle=0 $ ] for all @xmath47 in @xmath302 .
if @xmath47 is a non - zero object in @xmath443 , then @xmath47 has a simple quotient , say @xmath19 , by noetherianness . using that @xmath1239 for all @xmath445 in @xmath302 by serre duality , this yields @xmath1238,w_x\rangle>0 $ ] and therefore @xmath1240,w\rangle>0 $ ] .
thus the map @xmath1241 has the desired properties .
the abelian category @xmath1242 is a length category .
let @xmath47 be an object in @xmath31 .
we prove by induction on @xmath1229)$ ] that @xmath47 has finite length in @xmath1242 . if @xmath1229)=0 $ ] , then @xmath36 in @xmath1242 .
now suppose @xmath1229)>0 $ ] .
the category @xmath1242 is noetherian and therefore each non - zero object has a simple quotient .
thus there exists a subobject @xmath92 such that @xmath96 is simple in @xmath1242 . then @xmath1243 )
< \rho([a])$ ] and therefore @xmath88 has finite length in @xmath1242 .
it follows that @xmath47 has finite length in @xmath1242 . for each object
@xmath47 in @xmath31 , we denote by @xmath1244 the length of @xmath47 in @xmath1242 and call it the _ rank_. this function extends to a linear map @xmath1245 .
this linear map is surjective and satisfies the conditions in lemma [ le : linear - rank ] .
indeed , such a map is unique by proposition [ pr : k - split ] .
let @xmath31 be a @xmath0-linear abelian category satisfying ( h1)(h4 ) .
an indecomposable object in @xmath31 of rank one is called _
line bundle_. thus line bundles are precisely the objects in @xmath443 of rank one ; they form the building blocks of the category @xmath443 .
let us collect their basic properties .
[ pr : lb_filtr ] every object @xmath47 in @xmath443 admits a filtration @xmath1246 of length @xmath1247 such that each factor @xmath300 is a line bundle .
we proceed by induction on @xmath188 .
the case @xmath1248 is clear . if @xmath916 , choose a monomorphism @xmath1249 in @xmath1242 with simple cokernel .
this morphism is represented by a morphism @xmath1250 in @xmath31 such that @xmath1251 and @xmath92 are subobjects with @xmath1252 and @xmath88 in @xmath302 ; see lemma [ le : mor ] .
it follows that @xmath1253 since @xmath1254 . passing from @xmath1255 to the image of @xmath76
, we may assume that @xmath1256 .
the cokernel @xmath1257 is simple in @xmath1242 .
thus there is a decomposition @xmath1258 with @xmath1259 of finite length and @xmath1260 a line bundle . forming the pullback of the exact sequence @xmath1261 along the inclusion @xmath1262 yields a monomorphism @xmath1263 with cokernel @xmath1260 .
clearly , @xmath1264 belongs to @xmath443 and has rank @xmath1265 .
[ le : filtr ] let @xmath47 be a non - zero object in @xmath443 and suppose that @xmath1266 for some simple object @xmath28 . 1 .
there exists a chain of monomorphisms @xmath1267 in @xmath31 such that @xmath1268 and @xmath1269 for all @xmath589 .
2 . for each @xmath290
, there exists an exact sequence @xmath1270 in @xmath31 such that @xmath1271 and @xmath1272=\sum_{i=0}^n[\tau^{-i } s]$ ] .
\(1 ) choose a non - split exact sequence @xmath1273 .
for each simple object @xmath32 , the induced morphism @xmath1274 is an isomorphism .
thus @xmath1275 belongs to @xmath443 .
serre duality implies that @xmath1276 .
thus we can iterate the construction and obtain a sequence of morphisms @xmath1277 \(2 ) apply ( 1 ) by taking the composite @xmath1278 for the morphism @xmath103 .
[ le : lineb ] let @xmath1279 be line bundles and @xmath1280 an exact sequence such that @xmath28 is simple .
then for each @xmath1209 , @xmath1281 if and only if @xmath1282 .
note that @xmath1283 if and only if @xmath1284 for some simple object @xmath1232 .
the assumptions imply @xmath1285 and @xmath1286 .
if @xmath32 is a simple object not lying in the @xmath367-orbit of @xmath28 , then @xmath1287 induces an isomorphism @xmath1288 .
[ le : lineb1 ] let @xmath1279 be line bundles .
then the following are equivalent : 1 . for each @xmath1209 , @xmath1281 if and only if @xmath1282 .
there exists @xmath1209 such that @xmath1283 and @xmath1289 .
there exists @xmath485 such that @xmath1290 in @xmath1242 .
\(1 ) @xmath186 ( 2 ) : observe that @xmath1283 if and only if @xmath1291 for some simple object @xmath1292 .
thus noetherianness of @xmath1293 implies that @xmath1283 for some @xmath1209 .
\(2 ) @xmath186 ( 3 ) : choose a simple object @xmath28 and a non - zero morphism @xmath1294 .
modulo some power of @xmath367 , there is a non - zero morphism @xmath1295 . forming the pullback of @xmath76 and @xmath411 , one obtains the following commutative diagram with exact rows .
@xmath1296&k\ar@{=}[d]\ar[r]&p\ar[d]\ar[r]&l'\ar[d]^{\p'}\ar[r]&0\\ 0\ar[r]&k\ar[r]&l\ar[r]^\p&s\ar[r]&0 } \ ] ] if the top row splits , then @xmath1297 and therefore @xmath1298 in @xmath1242 .
otherwise , @xmath1299 .
thus @xmath1300 in @xmath1242 .
\(3 ) @xmath186 ( 1 ) : suppose that @xmath1290 in @xmath1242 .
then there is a subobject @xmath1301 with @xmath1302 of finite length and there is a monomorphism @xmath1303 with cokernel of finite length . for each @xmath1209
, it follows then by iterating lemma [ le : lineb ] that @xmath1283 if and only if @xmath1289 .
[ pr : lmorph ] let @xmath1293 be a line bundle and @xmath1209 .
then we have @xmath1283 .
let @xmath1304 be the set of all @xmath1305 such that @xmath1306 , and let @xmath1307 .
we show that @xmath1308 .
let @xmath1309 be the class of line bundles @xmath1310 such that @xmath1290 in @xmath1242 for some @xmath485 , and let @xmath1311 be the class of all remaining line bundles .
denote by @xmath468 ( @xmath1312 ) the full subcategory consisting of objects @xmath172 having a filtration with factors in @xmath1313 or @xmath1314 .
there are no non - zero morphisms between objects from different @xmath468 s , and therefore also no extensions by serre duality .
this follows from lemma [ le : lineb1 ] and the fact that each non - zero morphism between line bundles in @xmath31 induces an isomorphism in @xmath1242 . on the other hand
, each indecomposable object belongs to one of the @xmath468 s .
this is clear for objects of finite length .
an object @xmath47 from @xmath443 has a finite filtration @xmath1315 such that each factor @xmath300 is a line bundle ; see proposition [ pr : lb_filtr ] .
the factors belong to a single @xmath1313 since there are no non - split extensions between different @xmath1313 s .
thus @xmath59 , but this implies @xmath1316 since @xmath31 is connected .
[ le : lb ] let @xmath1293 be a line bundle and @xmath1317
. then there are monomorphisms @xmath1318 and @xmath1319 such that @xmath88 belongs to @xmath443 and the cokernel of @xmath1319 belongs to @xmath302 .
moreover , each non - zero morphism @xmath1320 is a monomorphism .
the object @xmath47 has a simple quotient since it is noetherian .
thus @xmath1321 for some simple object @xmath28 by serre duality .
depending on the value of @xmath1322,[a]\rangle$ ] and using that @xmath1293 admits a non - zero morphism to the @xmath367-orbit of @xmath28 by proposition [ pr : lmorph ] , we choose in lemma [ le : filtr ] the number @xmath188 sufficiently big so that there exists an exact sequence @xmath1323 in @xmath31 with @xmath1271 , @xmath1272=\sum_{i=0}^n[\tau^{-i } s]$ ] , and @xmath1324,[b]\rangle= \langle [ l],[a]\rangle+\langle [ l],[c]\rangle>0.\ ] ] now set @xmath1325 . if @xmath1326 is a non - zero morphism , then @xmath1327 .
thus @xmath1256 since @xmath1293 belongs to @xmath443 .
the discussion of line bundles yields further properties of @xmath1242 and @xmath674 .
[ pr : k - split ] let @xmath31 be a @xmath0-linear abelian category satisfying _
then the following holds : 1 . the abelian category @xmath1242 has , up to isomorphism , a unique simple object .
each non - zero object @xmath47 in @xmath31 satisfies @xmath749\neq 0 $ ] in @xmath674 .
3 . let @xmath1293 be a line bundle in @xmath31
. then @xmath1328\oplus k'_0(\a)$ ] , where @xmath1329 denotes the image of the canonical map @xmath1330 .
\(1 ) let @xmath1279 be line bundles in @xmath31 . there are monomorphisms @xmath1331 and @xmath1332 with both cokernels in @xmath302 , by lemma [ le : lb ] .
thus @xmath1333 in @xmath1242 .
\(2 ) let @xmath47 be a non - zero object .
if @xmath47 is not of finite length , then @xmath1334}\neq 0 $ ] and therefore @xmath749\neq 0 $ ] .
now suppose that @xmath47 is of finite length .
it follows from proposition [ pr : lmorph ] that @xmath1335 for some line bundle @xmath1293 . on the other hand , @xmath1336 by serre duality .
thus @xmath1337,[a]\rangle\neq 0 $ ] , and it follows that @xmath749\neq 0 $ ] .
\(3 ) we have @xmath1338\cap k'_0(\a)=0 $ ] since @xmath1339 and @xmath1340 for all @xmath1341 .
we show by induction on the rank that each class @xmath749 $ ] belongs to @xmath1338 + k'_0(\a)$ ] .
this is clear if @xmath1342 .
if @xmath1343 , then there is an exact sequence @xmath1344 such that @xmath1345 in @xmath1242 ; see lemma [ le : lb ] .
it follows that @xmath1346 , and therefore @xmath1347=[l]+[a'']$ ] belongs to @xmath1338 + k'_0(\a)$ ] .
finally observe that @xmath749-[a']$ ] belongs to @xmath1329 .
an immediate consequence is the fact that the rank of @xmath674 is at least two .
let @xmath31 be a @xmath0-linear abelian category that is hom - finite and hereditary .
an object @xmath47 is called _ exceptional _ if @xmath1348 and @xmath453 is a division ring .
[ le : hr ] let @xmath67 be indecomposable objects in @xmath31 .
if @xmath1349 , then each non - zero morphism @xmath103 is a monomorphism or an epimorphism .
let @xmath74 be a non - zero morphism and @xmath1350 its canonical factorization .
we obtain the following commutative diagram with exact rows @xmath525&a\ar[r]\ar[d]_-{\p'}&e\ar[r]\ar[d]&b/{\operatorname{im}}\p\ar[r]\ar@{=}[d]&0\\ 0\ar[r]&{\operatorname{im}}\p\ar[r]^-{\p''}&b\ar[r]&b/{\operatorname{im}}\p\ar[r]&0}\ ] ] since @xmath1351 is right exact .
the induced exact sequence @xmath1352 splits .
thus @xmath411 or @xmath412 is a split monomorphism since @xmath453 is local . in the first case
@xmath76 is a monomorphism , and in the second case @xmath76 is an epimorphism . let us collect some immediate consequences .
[ pr : hr1 ] let @xmath31 be a @xmath0-linear abelian category that is hom - finite and hereditary .
then the following holds : 1 . an indecomposable object @xmath47 satisfying @xmath1348 is exceptional .
2 . let @xmath67 be non - isomorphic exceptional objects .
suppose that @xmath1353 and @xmath1349
. then @xmath843 or @xmath1354 .
3 . assume further that @xmath31 is ext - finite .
let @xmath67 be exceptional objects and @xmath749=[b]$ ] in @xmath674
. then @xmath175 .
we apply lemma [ le : hr ] and use the fact that for each indecomposable object , an endomorphism is either nilpotent or invertible .
in particular , an endomorphism that is a monomorphism or an epimorphism is invertible . from this , ( 1 ) and ( 2 ) are clear .
\(3 ) observe that @xmath1355 since @xmath1240,[b]\rangle>0 $ ] .
let @xmath74 be a non - zero morphism and @xmath1356 . applying the right exact functor @xmath1357 to the inclusion @xmath1358
shows that @xmath1359 .
thus @xmath1360,[a]\rangle=\langle [ b'],[b]\rangle>0 $ ] .
composing a non - zero morphism @xmath1361 with the epimorphism @xmath1362 induced by @xmath76 yields an isomorphism since @xmath453 is a division ring .
thus @xmath76 is a monomorphism .
the dual argument shows that @xmath76 is an epimorphism .
thus @xmath175 .
we consider expansions of abelian categories that satisfy the axioms ( h1)(h4 ) .
[ le : reduction ] let @xmath31 be a @xmath0-linear abelian category satisfying _ ( h1)(h4)_. for a full subcategory @xmath71 of @xmath31 , the following are equivalent : 1 .
the inclusion @xmath782 is a non - split expansion of abelian categories .
2 . there exists a simple object @xmath28 such that @xmath1363 and @xmath810 .
apply lemma [ le : localizable ] .
if @xmath782 is an expansion , then @xmath1364 for some localizable object @xmath28 , and serre duality implies @xmath1363 .
conversely , if @xmath28 is simple and @xmath1365 , then @xmath28 is a ( co)localizable object with @xmath1366 .
thus the inclusion @xmath1367 is an expansion , and this is non - split since @xmath31 is connected .
let @xmath757 be a non - split expansion of @xmath0-linear abelian categories satisfying ( h1)(h4 ) .
then @xmath181 restricts to an expansion @xmath1368 by proposition [ pr : onepoint_length ] .
let @xmath1369 and @xmath1370 be the decompositions into connected uniserial categories ; see proposition [ pr : serre_finlen ] .
[ pr : onepoint_comp ] let @xmath757 be a non - split expansion of @xmath0-linear abelian categories satisfying _
( h1)(h4)_. there exists a bijection @xmath1371 and @xmath1372 such that @xmath181 restricts to an expansion @xmath1373 and to equivalences @xmath1374 for all @xmath1375 in @xmath1376 .
apply lemma [ le : decomp ] .
next we investigate the existence of tilting objects for expansions of abelian categories .
[ le : onepoint_tilt ] let @xmath757 be a non - split expansion of @xmath0-linear abelian categories that are ext - finite . 1 .
if @xmath71 admits a tilting object , then @xmath31 admits a tilting object .
2 . suppose in addition that @xmath31 is hereditary and has no non - zero projective object .
if @xmath31 admits a tilting object , then @xmath71 admits a tilting object .
\(1 ) let @xmath32 be a tilting object in @xmath71 .
choose an exact sequence @xmath1377 in @xmath31 with @xmath28 in @xmath1378 such that the induced map @xmath1379 is an epimorphism .
we claim that @xmath1380 is a tilting object for @xmath31 . the formula @xmath1381 for all @xmath172 and @xmath820 implies @xmath1382 .
therefore @xmath1383 .
the construction of @xmath567 implies @xmath1384 , and @xmath1385 is clear since @xmath1386 and @xmath1387 .
we have @xmath1388 and therefore @xmath1389 .
finally , assume that @xmath1390 for some @xmath47 in @xmath31 ( @xmath1391 ) .
the condition @xmath1392 implies that @xmath47 belongs to the image of @xmath181 , say @xmath1393 , and that @xmath1394
. then @xmath1395 implies @xmath1396 .
thus @xmath1032 is a tilting object .
\(2 ) let @xmath32 be a tilting object in @xmath31 .
we intend to show that @xmath1397 is a tilting object for @xmath71 .
the formula @xmath1398 for all @xmath234 and @xmath820 shows that @xmath1399 and that @xmath1400 ( @xmath1391 ) implies @xmath1396 .
it remains to show that @xmath1401 .
in fact , it is equivalent to show that @xmath1402 .
we proceed by cases .
first assume that @xmath1403 .
thus the adjunction morphism @xmath1404 is an epimorphism , and therefore @xmath1402 since @xmath1405 is right exact .
next assume that @xmath1406 .
then @xmath1407 follows since @xmath1408 and @xmath1409 belong to @xmath1378 .
finally , assume that @xmath1410 .
we apply the auslander - reiten formula from proposition [ pr : arformula ] and have @xmath1411 .
thus there is an epimorphism @xmath1412 , and this implies @xmath1413 . the category
@xmath31 admits a serre functor @xmath1414 by proposition [ pr : serre ] , and the auslander - reiten formula implies @xmath1415 .
thus @xmath1416 .
the object @xmath1417 is a tilting object for @xmath31 , and the above argument shows that @xmath1418 is a tilting object for @xmath71 .
the following result says that the concept of an expansion of abelian categories is compatible with the list of axioms ( h1)(h5 ) .
[ th : reduction ] let @xmath0 be a field and @xmath782 a non - split expansion of @xmath0-linear abelian categories with associated division ring @xmath0
. then @xmath31 satisfies _ ( h1)(h5 ) _ if and only if @xmath71 satisfies _
( h1)(h5)_. in that case the rank of @xmath762 is one less than that of @xmath674 .
we provide the references for each axiom .
the final assertion about the rank of @xmath762 follows from lemma [ le : euler_onepoint ] .
( h1 ) lemmas [ le : small ] , [ le : decomp ] , and [ le : extfin ] .
( h2 ) lemma [ le : noetherian ] .
( h3 ) lemma [ le : cohdim ] .
( h4 ) lemma [ le : onepoint_tilt ] .
( h5 ) proposition [ pr : euler ] and lemma [ le : euler_onepoint ] .
we give a criterion so that an equivalence of derived categories @xmath1419 restricts to an equivalence @xmath1420 .
let @xmath31 be a @xmath0-linear abelian category satisfying ( h1)(h4 ) and consider its bounded derived category @xmath510 . recall that @xmath1421\ ] ] with non - zero morphisms @xmath1422\to\a[j]$ ] only if @xmath1423 since @xmath31 is hereditary ; see corollary [ co : der_hereditary ] .
the isomorphism @xmath1424 yields a rank function @xmath1425 .
note that for each complex @xmath500 concentrated in degree @xmath188 , we have @xmath1426}\geq 0 $ ] .
let @xmath32 be an indecomposable object in @xmath443 and view it as a complex concentrated in degree zero .
define @xmath1427 to be the class of indecomposable objects @xmath1428 of rank one such that @xmath1429 for some indecomposable object @xmath1430 of rank one satisfying @xmath1431 .
[ le : l(t ) ] let @xmath32 be an indecomposable object in @xmath443 .
then the objects in @xmath1427 are precisely those that are isomorphic to a line bundle in @xmath31 , viewed as a complex concentrated in degree zero .
let @xmath1310 be a complex in @xmath510 that is concentrated in one degree , say @xmath188 . if @xmath1432 , then @xmath795 or @xmath1433 .
if the rank of @xmath1310 is positive , then @xmath795 .
the same argument shows that every indecomposable object @xmath1293 in @xmath510 of rank one and satisfying @xmath1429 is isomorphic to a line bundle , viewed as a complex concentrated in degree zero .
conversely , let @xmath1293 be a line bundle .
then there exists a non - zero morphism @xmath1434 for some object @xmath567 that admits an exact sequence @xmath1435 such that @xmath109 has finite length ; see lemma [ le : lb ] .
the pullback of @xmath1436 and @xmath1434 yields a line bundle @xmath1310 with non - zero morphisms to @xmath32 and @xmath1293 .
thus @xmath1293 belongs to @xmath1427 .
[ le : tilt_zero ] let @xmath32 be an indecomposable object in @xmath443 .
then the following are equivalent for an indecomposable object @xmath500 in @xmath510 : 1 .
@xmath1437 for all @xmath712 .
2 . @xmath1438 } \geq 0 $ ] and @xmath1439 for some @xmath1440 .
we apply lemma [ le : l(t ) ] . because @xmath500 is indecomposable
, there exists an integer @xmath188 such that @xmath1437 for all @xmath1441 .
\(1 ) @xmath186 ( 2 ) : if @xmath795 , then @xmath1438 } \geq 0 $ ] and
@xmath1442 for some line bundle @xmath1293 in @xmath31 .
thus @xmath1439 for some @xmath1440 .
\(2 ) @xmath186 ( 1 ) : if @xmath1439 for some @xmath1440 , then @xmath795 or @xmath1433 .
it follows that @xmath795 if @xmath500 has positive rank . if the rank of @xmath500 is zero , then @xmath1443 has finite length and therefore @xmath1444 .
thus @xmath795 .
[ pr : der_equiv ] let @xmath1445 be abelian categories satisfying _
( h1)(h4 ) _ with tilting objects @xmath572 and @xmath1446 .
suppose that @xmath1447 and that the induced equivalence @xmath1448 preserves the rank .
then @xmath31 and @xmath518 are equivalent categories .
we identify @xmath1449 and obtain equivalences @xmath1450 this yields an equivalence @xmath1451 taking @xmath32 to @xmath567 .
the functor @xmath81 preserves the rank since the indecomposable direct summands of @xmath32 form a basis of @xmath1452 .
it follows from lemma [ le : tilt_zero ] that @xmath81 identifies @xmath31 with @xmath518 .
let @xmath31 be a @xmath0-linear abelian category satisfying ( h1)(h4 ) .
call @xmath31 _ homogeneous _ if @xmath1453 for each object @xmath47 of finite length , or equivalently , @xmath1454 for each simple object @xmath28 . this property together with the conditions ( h1)(h5 ) characterizes the category @xmath1012
. the following characterization of the property of @xmath31 to be homogeneous will be useful .
[ pr : homog ] let @xmath31 be a @xmath0-linear abelian category satisfying _
( h1)(h4)_. then the following are equivalent : 1 . if @xmath28 is a simple object in @xmath31 , then @xmath1455 .
2 . if @xmath67 are finite length objects in @xmath31 , then @xmath1238,[b]\rangle=0 $ ] .
3 . the rank of @xmath674 equals two .
if @xmath47 has infinite length and @xmath28 is a simple object in @xmath31 , then @xmath1456 .
\(1 ) @xmath186 ( 2 ) : it suffices to show that @xmath1220,[t]\rangle=0 $ ] for each pair of simple objects @xmath309 .
the equality @xmath1457,[t]\rangle=0 $ ] is an immediate consequence of serre duality if @xmath1458 .
\(2 ) @xmath186 ( 3 ) : let @xmath1329 be the image of the canonical map @xmath1330 .
we have @xmath1459 by proposition [ pr : k - split ] . now apply lemma [ le : z2 ] below .
\(3 ) @xmath186 ( 1 ) : suppose there exists a simple object @xmath28 such that @xmath1363 .
then @xmath1364 satisfies ( h1)(h4 ) and @xmath1460\oplus k_0(\b)$ ] , by theorem [ th : reduction ] and lemma [ le : reduction ] .
if @xmath71 is homogeneous , then @xmath1461 by the first part of the proof .
otherwise , we proceed as before and reduce to the homogeneous case . in any case , the rank of @xmath674 is at least @xmath1462 .
\(1 ) @xmath186 ( 4 ) : this follows from proposition [ pr : lmorph ] since each infinite length object @xmath47 admits a subobject @xmath88 such that @xmath96 is a line bundle .
\(4 ) @xmath186 ( 1 ) : suppose there exists a simple object @xmath28 such that @xmath1363
. then @xmath1463 yields an expansion @xmath782 by lemma [ le : reduction ] , and this induces an equivalence @xmath894 by proposition [ pr : onepoint_length ] .
thus any infinite length object in @xmath31 yields one in @xmath71 , say @xmath47 , satisfying @xmath1464 by construction .
[ le : z2 ] let @xmath223 be a free abelian group of finite rank with a non - degenerate bilinear form @xmath76 .
suppose there is a subgroup @xmath1465 such that @xmath1466 and @xmath1226 for all @xmath1467
. then @xmath1468 .
the assumption on @xmath1469 implies that for each pair of non - zero elements @xmath1467 , there are non - zero integers @xmath1470 with @xmath1471 .
this implies @xmath1472 since @xmath1469 is free .
[ le : homog_except ] let @xmath31 be a @xmath0-linear abelian category satisfying _ ( h1)(h4 ) _ and suppose that @xmath31 is homogeneous
. then the following holds : 1 .
each line bundle @xmath1293 is exceptional .
2 . there exists a simple object @xmath28 such that in @xmath674 the class of each simple object @xmath1473 is of the form @xmath1474=n\cdot [ s]$ ] for some @xmath1475 .
\(1 ) set @xmath1476 $ ] and choose a second basis vector @xmath1477 of @xmath1478 lying in the image of the map @xmath1330 ; see proposition [ pr : k - split ] . for @xmath1479 in @xmath674 , we have @xmath1480 since @xmath1481 and @xmath1482 . from the existence of a tilting object @xmath32 in @xmath31
, it follows that @xmath1483 since @xmath1484,[t]\rangle>0 $ ] .
now observe that @xmath1485 is divisible by @xmath1486 since the endomorphism ring of any line bundle is a division ring by lemma [ le : lb ] .
thus @xmath1487 .
\(2 ) choose a simple object @xmath28 such that @xmath1488,[s]\rangle$ ] is minimal .
given any simple object @xmath1473 , there are integers @xmath1489 with @xmath1490,[s']\rangle = q\cdot d+r$ ] and @xmath1491 . applying proposition [ pr : lmorph ] and lemma [ le : filtr ]
, we obtain extensions @xmath1492 and @xmath1493 such that @xmath16 , @xmath1494 are line bundles , @xmath1272=q\cdot [ s]$ ] , and @xmath1495=[s']$ ] .
hence @xmath1496,[e']\rangle= \langle [ l],[l]\rangle+\langle [ l],[s']\rangle - q\cdot\langle [ l],[s]\rangle>0.\ ] ] this gives an exact sequence @xmath1497 , where @xmath81 is an object of finite length and @xmath1490,[f]\rangle = r < d$ ] .
the minimality of @xmath1490,[s]\rangle$ ] implies @xmath1498 , and therefore @xmath1474=q\cdot [ s]$ ] . the class @xmath748 $ ] in lemma [ le : homog_except ] yields a generator for the image of the map @xmath1330 .
thus for any finite length object @xmath47 in @xmath31 , there exists some @xmath820 with @xmath749=n\cdot [ s]$ ] .
we call this number the _ degree _ of @xmath47 and observe that it is independent of the choice of @xmath28 .
next we describe tilting objects for an abelian category that satisfies ( h1)(h5 ) and is homogeneous .
[ pr : homog_tilt ] let @xmath31 be a @xmath0-linear abelian category satisfying _ ( h1)(h5 ) _ and suppose that @xmath31 is homogeneous .
let @xmath1293 be a line bundle and @xmath28 a simple object of degree one
. then @xmath1499 let @xmath1500 be a non - split extension . then @xmath1501 is a tilting object and its endomorphism algebra is isomorphic to the kronecker algebra ( i.e. the path algebra of the quiver @xmath1502\ar@<-.5ex>[r]&\cdot}$ ] ) . moreover , the simple objects of degree one are precisely the objects that arise as the cokernel of a non - zero morphism @xmath1287 .
it follows from proposition [ pr : k - split ] and lemma [ le : homog_except ] that @xmath1503 $ ] and @xmath748 $ ] form a basis of @xmath674 .
the corresponding matrix @xmath1504,[l ] \rangle&\langle [ l],[s ] \rangle\\ \langle [ s],[l ] \rangle&\langle [ s],[s ] \rangle\end{smallmatrix}\right]}$ ] has determinant @xmath1203 .
thus @xmath1505,[s ] \rangle=-\langle [ s],[l ] \rangle$ ] .
this implies @xmath1506 and @xmath1507 .
the space @xmath1508 is a module over @xmath1509 and over @xmath806 .
it follows that @xmath1510 and @xmath1511 .
next we show that @xmath1512 is a tilting object .
an application of @xmath1513 to @xmath1280 yields @xmath1514 , while application of @xmath1515 implies @xmath1516 .
thus @xmath1517 .
for any non - zero object @xmath47 in @xmath31 , we have @xmath1322,[a ] \rangle\neq 0 $ ] or @xmath1518,[a]\rangle\neq 0 $ ] since @xmath1503 $ ] and @xmath1519 $ ] form a basis of @xmath674 and @xmath749\neq 0 $ ] .
thus @xmath32 is a tilting object .
a simple computation shows that @xmath1520 , while @xmath1521 .
thus @xmath705 is isomorphic to the kronecker algebra .
let @xmath1522 be a non - zero morphism .
this is a monomorphism since @xmath1293 is a line bundle .
the cokernel @xmath1257 is of finite length since @xmath1293 and @xmath1310 have the same rank .
the degree of @xmath109 is one since @xmath1272=[l']-[l]=[s]$ ] . in particular
, @xmath109 is simple .
now let @xmath1473 be a simple object of degree one .
choose a non - split extension @xmath1523 .
then @xmath16 is a line bundle and therefore exceptional by lemma [ le : homog_except ] .
we have @xmath1524=[l']$ ] in @xmath674 since @xmath748=[s']$ ] , and it follows from proposition [ pr : hr1 ] that @xmath1525 .
thus @xmath1473 arises as the cokernel of a morphism @xmath1287 .
the next theorem provides an axiomatic description of the category @xmath1012 .
[ th : homog ] let @xmath31 be a @xmath0-linear abelian category satisfying _ ( h1)(h5 ) _ and suppose that @xmath31 is homogeneous
. then @xmath31 is equivalent to @xmath1012 . the categories @xmath1012 and @xmath31 admit each a tilting object such that its endomorphism algebra is isomorphic to the kronecker algebra , see propositions [ pr : coh_tilt ] and [ pr : homog_tilt ] . note that in both cases the indecomposable direct summands of a tilting object have rank one .
now apply proposition [ pr : der_equiv ] .
the following theorem characterizes the abelian categories that arise as categories of coherent sheaves on weighted projective lines in the sense of geigle and lenzing @xcite .
[ th : axioms ] let @xmath0 be a field and @xmath31 a @xmath0-linear abelian category
. then the following are equivalent : 1 .
the category @xmath31 satisfies _ ( h1)(h5)_. 2 .
there is a finite sequence @xmath1526 of full subcategories such that @xmath921 is equivalent to @xmath1012 and @xmath1527 is a non - split expansion of @xmath1528 with associated division ring @xmath0 for @xmath1529 .
3 . the category @xmath31 is equivalent to @xmath5 for some weighted projective line @xmath1199 .
\(1 ) @xmath186 ( 2 ) : suppose that @xmath31 satisfies ( h1)(h5 ) .
the rank of @xmath674 is finite , say @xmath188 .
so one constructs a filtration @xmath1530 of length @xmath1531 by reducing the rank of the grothendieck group as follows .
if @xmath31 is homogeneous , then @xmath31 is equivalent to @xmath1012 by theorem [ th : homog ] . otherwise , there is a simple object @xmath28 such that @xmath1532 by proposition [ pr : homog ] . then put @xmath1533 and the inclusion @xmath1534 is a non - split expansion by lemma [ le : reduction ] .
moreover , @xmath1535 satisfies ( h1)(h5 ) by theorem [ th : reduction ] , and the associated division ring is @xmath0 by lemma [ le : euler_onepoint ] .
note that the rank of @xmath1536 is one less than that of @xmath674 .
so one proceeds and constructs a sequence of subcategories @xmath919 .
the process stops after @xmath974 steps when @xmath921 is homogeneous .
\(2 ) @xmath186 ( 1 ) : suppose that @xmath31 admits a filtration @xmath1530 such that @xmath921 is equivalent to @xmath1012 and @xmath1527 is a non - split expansion of @xmath1528 with associated division ring @xmath0 for @xmath1537 .
the discussion in
[ se : coh ] shows that @xmath1012 satisfies ( h1)(h5 ) .
an iterated application of theorem [ th : reduction ] yields that @xmath31 satisfies ( h1)(h5 ) .
\(2 ) @xmath186 ( 3 ) : suppose again that @xmath31 admits a sequence @xmath1530 of expansions such that @xmath921 is equivalent to @xmath1012 .
this yields a fully faithful exact functor @xmath1538 , and it follows from proposition [ pr : onepoint_comp ] that this functor identifies the index set of the decomposition @xmath1539}{\operatorname{mod}}_0\oc_{\bbp^1_k,\frp}\ ] ] into connected components with the index set of the decomposition @xmath1205 . thus there is a canonical bijection between the set of closed points of @xmath1 and the set @xmath1208 .
moreover , if @xmath1209 is a point with @xmath1540 , then the corresponding closed point @xmath980 of @xmath1 is rational since the residue field of the corresponding local ring @xmath1128 equals @xmath0 .
this follows from the fact that in the filtration @xmath1526 the associated division ring of each expansion equals @xmath0 .
let @xmath1200 be the finite collection of points @xmath1541 , viewed as points of @xmath1 , and denote by @xmath1542 the corresponding sequence of positive integers @xmath1210 .
then there exists a tilting object @xmath32 such that @xmath1543 ; see proposition [ pr : tiltobj ] below . on the other hand ,
let @xmath1544 be the weighted projective line that is determined by the parameters @xmath1200 and @xmath1542 .
the category @xmath5 of coherent sheaves on @xmath6 admits the following tilting object @xmath1545}\oplus\dots\oplus s_1^{[p_1 - 1]})\oplus \dots \oplus(s_n^{[1]}\oplus\dots\oplus s_n^{[p_n-1]})\ ] ] with endomorphism algebra @xmath1546 , where the notation is taken from the introduction with @xmath1547 ; see ( * ? ? ?
* example 4.4 ) .
this yields a derived equivalence @xmath1548 which restricts to an equivalence @xmath1549 by proposition [ pr : der_equiv ] .
\(3 ) @xmath186 ( 1 ) : see @xcite .
let @xmath31 be a @xmath0-linear abelian category satisfying ( h1)(h5 ) .
\(1 ) the reduction to the homogeneous case in the proof of theorem [ th : axioms ] shows that the rank of the grothendieck group of @xmath31 is @xmath1550 .
\(2 ) let @xmath1209 and @xmath1540 .
then @xmath1211 is equivalent to the category of finite dimensional nilpotent representations of a quiver of extended dynkin type @xmath1551 with cyclic orientation ; see example [ ex : cyclic ] .
let @xmath0 be a field and @xmath31 a @xmath0-linear abelian category satisfying ( h1)(h5 ) .
we construct a tilting object and compute its endomorphism algebra , which is a _ squid algebra _ in the sense of brenner and butler @xcite . given a collection @xmath1552 of distinct rational points @xmath1553
$ ] of @xmath1 , and a sequence @xmath3 of positive integers , we define @xmath1546 to be the finite dimensional associative algebra given by the quiver @xmath1554}}$ } \put(24,54){$\scriptstyle{s_2^{[p_2 - 1]}}$ } \put(24,74){$\scriptstyle{s_1^{[p_1 - 1]}}$ } \put(54,2){$\scriptstyle{s_n^{[p_n-2]}}$ } \put(54,54){$\scriptstyle{s_2^{[p_2 - 2]}}$ } \put(54,74){$\scriptstyle{s_1^{[p_1 - 2]}}$ } \put(126,2){$\scriptstyle{s_n^{[1]}}$ } \put(126,54){$\scriptstyle{s_2^{[1]}}$ } \put(126,74){$\scriptstyle{s_1^{[1]}}$ } \put(42,2){}\put(42,54){}\put(42,74){}\end{picture}\ ] ] modulo the relations @xmath1555 [ pr : tiltobj ] a @xmath0-linear abelian category satisfying _ ( h1)(h5 ) _ admits a tilting object with endomorphism algebra isomorphic to @xmath1546 for some pair @xmath1556 . fix a @xmath0-linear abelian category @xmath31 satisfying ( h1)(h5 ) .
we apply theorem [ th : axioms](2 ) and follow its proof .
thus there exists a sequence @xmath1557 of expansions such that @xmath71 is equivalent to @xmath1012 .
let @xmath1558 be the collection of distinct points @xmath1209 with @xmath1540 and set @xmath3 with @xmath1559 for each @xmath181 .
choose line bundles @xmath1293 and @xmath1310 in @xmath71 forming a tilting object @xmath1501 for @xmath71 , and choose a basis @xmath1560 of @xmath1561 ; see proposition [ pr : homog_tilt ] .
the inclusion @xmath1562 restricts to a family of inclusions @xmath1563 ; see proposition [ pr : onepoint_comp ] . note that each inclusion @xmath1563 is a composite of @xmath1564 non - split expansions
thus there are simple objects @xmath1565 and @xmath1566 such that @xmath1567}$ ] . here
, @xmath1568}$ ] denotes the uniserial object with top @xmath180 and length @xmath1564 , and we use that an expansion sends a specific simple object to an object of length two ; see lemma [ le : simple ] .
in particular , @xmath1569}$ ] belongs to @xmath768 for @xmath1570 .
note that @xmath1571 since the division ring of each expansion is @xmath0 . in @xmath1012 , a simple object with trivial endomorphism ring
has degree one .
thus each simple object @xmath1572 fits into an exact sequence @xmath1573 by proposition [ pr : homog_tilt ] , and this yields a collection @xmath2 of rational points @xmath1574 $ ] in @xmath1 . moreover , there are canonical morphisms @xmath1575}\twoheadrightarrow s_i^{[p_i-1]}$ ] in @xmath31 satisfying the relations @xmath1576 .
it is straightforward to verify that the object @xmath1577}\oplus\dots\oplus s_1^{[p_1 - 1]})\oplus \dots \oplus(s_n^{[1]}\oplus\dots\oplus s_n^{[p_n-1]})\ ] ] is a tilting object for @xmath31 , using the criterion of lemma [ le : euler_tilt ] .
indeed , @xmath1517 , the indecomposable direct summands of @xmath32 yield a basis of @xmath674 , and @xmath749\neq 0 $ ] for each object @xmath1578 by proposition [ pr : k - split ] .
finally , one checks that @xmath705 is isomorphic to @xmath1546 . here
, one uses that each epimorphism @xmath1568}\to s_i^{[j]}$ ] induces an isomorphism @xmath1579}){\xrightarrow}{\sim } \operatorname{hom}_\a(l\oplus l',s_i^{[j]})$ ] .
the canonical algebras in the sense of ringel @xcite provide a link between weighted projective lines and the representation theory of finite dimensional algebras .
in fact , geigle and lenzing constructed in @xcite for each weighted projective line @xmath1199 a tilting object in @xmath5 such that its endomorphism algebra is isomorphic to the canonical algebra with the same parameters ; see example [ ex : can ] .
it turns out that this tilting object is somehow canonical . from this
it follows that the parameters @xmath1200 and @xmath1542 can be reconstructed from the category @xmath5 . to be more precise
, we consider an abelian category @xmath31 that is equivalent to @xmath5 for some weighted projective line @xmath6 .
for each line bundle @xmath1293 in @xmath31 one constructs a canonical tilting object @xmath1580 such that its endomorphism algebra is a canonical algebra .
then one shows that for each pair of line bundles @xmath1279 there is a sequence of tubular mutations in the sense of lenzing and meltzer @xcite that yields an equivalence @xmath1581 taking @xmath1293 to @xmath1310 and therefore @xmath1580 to @xmath1582 . in particular , the endomorphism algebras of @xmath1580 and @xmath1582 are isomorphic and therefore an invariant of @xmath31 .
finally one observes that the parameters @xmath1200 and @xmath1542 form an invariant of the canonical algebra @xmath1583 . throughout this section
we fix an arbitrary field @xmath0 .
consider a collection @xmath1552 of distinct rational points @xmath1553 $ ] of @xmath1 , and a sequence @xmath3 of positive integers .
we define the _ canonical algebra _
@xmath1584 to be the finite dimensional associative algebra given by the quiver @xmath1585 modulo the relations is empty . in that case
@xmath1584 equals the kronecker algebra . ]
[ th : tiltobj ] a @xmath0-linear abelian category satisfying _ ( h1)(h5 ) _ admits a tilting object @xmath32 such that @xmath32 is a direct sum of line bundles and the endomorphism algebra of @xmath32 is isomorphic to @xmath1584 for some pair @xmath1556 . fix a @xmath0-linear abelian category @xmath31 satisfying ( h1)(h5 ) .
we adapt the proof of proposition [ pr : tiltobj ] and modify the tilting object constructed in that proof as follows .
consider for each point @xmath1587 with @xmath1588 the exact sequence @xmath1589}{\stackrel{{}}{\longrightarrow}}0\ ] ] in @xmath31 , where @xmath1590 .
we form successively the pullback along the chain of monomorphisms @xmath1591}\rightarrowtail(\tau^{-1}s_i)_{[2]}\rightarrowtail\cdots \rightarrowtail ( \tau^{-1 } s_i)_{[p_i-1]}\rightarrowtail ( \tau^{-1}s_i)_{[p_i]}=s_i^{[p_i]}\ ] ] and obtain the following commutative diagram with exact columns .
@xmath1592&0\ar[d]&&0\ar[d]&0\ar[d]\\ l\ar[d]\ar@{=}[r]&l\ar[d]\ar@{=}[r]&\hdots\ar@{=}[r ] & l\ar[d]\ar@{=}[r]&l\ar[d]^{\p_i}\\ l_i^{(1)}\ar[d]\ar[r]&l_i^{(2)}\ar[d]\ar[r]&\hdots\ar[r]&l_i^{(p_i-1)}\ar[d]\ar[r]&l'\ar[d]\\ ( \tau^{-1}s_i)_{[1]}\ar[d]\ar[r]&(\tau^{-1}s_i)_{[2]}\ar[d]\ar[r]&\hdots\ar[r]&(\tau^{-1 } s_i)_{[p_i-1]}\ar[d]\ar[r]&(\tau^{-1}s_i)_{[p_i]}\ar[d]\\ 0&0&&0&0 } \ ] ] note that each object @xmath1593 is a line bundle .
it is straightforward to verify that the object @xmath1594 is a tilting object for @xmath31 , following the line of arguments in the proof of proposition [ pr : tiltobj ] .
let @xmath31 be a @xmath0-linear abelian category satisfying _
( h1)(h5)_. then there exists for some pair @xmath1556 an equivalence of derived categories @xmath1595 let @xmath31 be a @xmath0-linear abelian category satisfying ( h1)(h4 ) .
the category @xmath443 consisting of the vector bundles in @xmath31 determines together with its exact structure the category @xmath31 .
in fact , every object @xmath47 in @xmath31 admits a presentation @xmath1596 with @xmath1597 in @xmath443 . introducing appropriate morphisms between complexes in @xmath443
, one can make these presentations functorial .
[ le : vb_pres ] every object in @xmath31 is a factor object of an object in @xmath443 .
every object of @xmath31 decomposes into an object of @xmath443 and an object of finite length .
thus it suffices to show that for an indecomposable object @xmath47 of finite length there is an epimorphism @xmath459 with @xmath16 in @xmath443 .
we use induction on the length @xmath301 of @xmath47 .
up to certain power of @xmath367 , the case @xmath353 follows from proposition [ pr : lmorph ] .
assume that @xmath355 .
take a maximal subobject @xmath92 . by the induction hypothesis
there is an epimorphism @xmath1598 .
note that @xmath1599 induces an epimorphism @xmath1600 .
in particular , we obtain the following commutative diagram with exact rows . @xmath1601 & e ' \ar[d]^{\phi } \ar[r ] & e \ar[r ] \ar[d ] & a / a ' \ar@{=}[d]\ar[r ] & 0\\ 0\ar[r ] & a ' \ar[r ] & a \ar[r ] & a / a ' \ar[r ] & 0 . } \ ] ] note that the upper row does not split since @xmath47 is indecomposable .
thus the morphism @xmath1602 induces a bijection @xmath1603 for each simple object @xmath28 .
it follows that @xmath16 belongs to @xmath443 , and therefore @xmath47 is a quotient of an object in @xmath443 .
the following result says that the subcategory @xmath443 determines @xmath31 .
however , it is important to notice that one uses the exact structure on @xmath443 that is inherited from @xmath31 . to be precise ,
a morphism of complexes in @xmath443 is by definition a _ quasi - isomorphism _ if it is a quasi - isomorphism of complexes in @xmath31 .
[ pr : vb_pres ] let @xmath31 be a @xmath0-linear abelian category satisfying _
the inclusion @xmath1604 induces an equivalence @xmath1605{\stackrel{\sim}{\longrightarrow}}\bfd^b(\a).\ ] ] the subcategory @xmath443 of @xmath31 is closed under forming extensions and taking kernels of epimorphisms .
moreover , each object @xmath47 in @xmath31 fits into an exact sequence @xmath1606 with each @xmath385 in @xmath443 ; see lemma [ le : vb_pres ] . with these properties
, the assertion follows from ( * ? ? ?
iii , prop .
2.4.3 ) .
[ co : vb_pres ] let @xmath1607 be an equivalence and suppose that a sequence @xmath220 in @xmath443 is exact if and only if @xmath1608 is exact .
then @xmath81 extends uniquely to an equivalence @xmath1581 .
we apply proposition [ pr : vb_pres ] .
thus the assumption on @xmath81 implies that the functor extends to an equivalence @xmath1609 .
this equivalence restricts to an equivalence @xmath1581 because @xmath31 identifies with the full subcategory consisting of complexes @xmath1610 with @xmath1597 in @xmath443 and @xmath1611 a monomorphism ; see lemma [ le : vb_pres ] .
let @xmath31 be a @xmath0-linear abelian category satisfying ( h1)(h5 ) and @xmath1205 be the decomposition of @xmath302 into connected uniserial components
. for each @xmath1209 denote by @xmath1612 the direct sum of a representative set of simple objects in @xmath1211 and set @xmath1613 .
note that @xmath1614 is a hom - finite semisimple abelian category with finitely many simple objects .
thus each additive functor @xmath1615 is representable .
this observation yields two functors @xmath1616 such that for each object @xmath47 in @xmath31 @xmath1617 fix an object @xmath47 in @xmath443 .
the identity morphism of @xmath1618 corresponds to a morphism @xmath1619 .
this is an epimorphism and we complete it to an exact sequence @xmath1620 . on the other hand , the identity morphism of @xmath1621 corresponds to an exact sequence @xmath1622 . it is easily checked that this defines two functors @xmath1623 .
the following lemma shows that @xmath1624 and @xmath1625 yield equivalences @xmath1626 .
[ le : mutat ] let @xmath1627 be an exact sequence in @xmath31 with @xmath32 in @xmath1614 for some @xmath1209 .
then the following are equivalent : 1 .
@xmath220 induces an iso @xmath1628 and @xmath47 belongs to @xmath443 .
2 . @xmath220 induces an iso @xmath1629 and @xmath47 belongs to @xmath443 .
3 . @xmath220 induces an iso @xmath1630 and @xmath88 belongs to @xmath443 .
\(1 ) @xmath376 ( 2 ) : apply serre duality and observe that @xmath1631 .
\(1 ) & ( 2 ) @xmath186 ( 3 ) : let @xmath28 be any simple object in @xmath31 and apply @xmath337 to @xmath220 . using ( 1 ) and the fact that @xmath47 belongs to @xmath443 , it follows that @xmath1632 .
thus @xmath88 is in @xmath443 .
now let @xmath28 be any object in @xmath1614 and apply @xmath1633 to @xmath220 .
this yields the following exact sequence @xmath1634 where @xmath428 is an isomorphism if and only if @xmath1635 is an isomorphism .
thus ( 3 ) holds .
\(3 ) @xmath186 ( 2 ) : the object @xmath47 is in @xmath443 since @xmath443 is closed under taking subobjects .
the rest follows as before by choosing @xmath28 in @xmath1614 and applying @xmath1633 to @xmath220 .
the functors @xmath1624 and @xmath1625 form a pair of mutually inverse equivalences @xmath1626 .
moreover , @xmath1624 and @xmath1625 take exact sequences in @xmath443 to exact sequences .
the first assertion is an immediate consequence of lemma [ le : mutat ] . for the exactness
observe that @xmath1636 and @xmath1637 are exact when restricted to @xmath443 .
the exactness of @xmath1624 and @xmath1625 then follows from the @xmath1638 lemma . using corollary [ co : vb_pres ] , the functors @xmath1639 yield equivalences @xmath1581 .
these functors are called _ tubular mutations _ and were introduced by lenzing and meltzer @xcite .
[ co : mutat ] the equivalences @xmath1640 extend to a pair of mutually inverse equivalences @xmath1581 .
next we show that for each pair of line bundles @xmath1279 there exists a sequence of tubular mutations taking @xmath1293 to @xmath1310 .
we need the following proposition which is of independent interest .
[ pr : hom_lb ] let @xmath1293 be a line bundle
. for each @xmath1209 there exists up to isomorphism a unique simple object @xmath19 in @xmath1211 such that @xmath1641 .
moreover , any non - zero morphism @xmath1642 induces an isomorphism @xmath1643 . for the purpose of this proof , call a line bundle @xmath1293 _ hom - simple _ if the assertion of the proposition holds for @xmath1293 .
we begin by showing that a specific line bundle is hom - simple .
fix a sequence of subcategories @xmath1644 as in theorem [ th : axioms ] and choose a line bundle @xmath1293 in @xmath921 .
note that the inclusion @xmath923 sends @xmath1293 to a line bundle of @xmath31 by proposition [ pr : onepoint_length ] . for each @xmath1209
, there is at least one simple object @xmath1232 with @xmath1641 by proposition [ pr : lmorph ] . on the other hand ,
the intersection of @xmath1645 with @xmath1211 is a serre subcategory having @xmath1646 simple objects .
in particular , @xmath1513 vanishes on them .
thus there is a unique simple object @xmath19 in @xmath1211 with @xmath1641 .
let @xmath1647 denote the right adjoint of the inclusion @xmath923 .
then @xmath1648 is a simple object by lemma [ le : simple ] . choose a non - zero morphism @xmath1642 .
this yields the following commutative square , where @xmath1649 denotes the adjunction morphism .
@xmath1650&\operatorname{hom}_{\a^0}(l , i_\rho s_x)\ar[d]^{\operatorname{hom}_\a(l,\eta)}\\ \operatorname{end}_{\a}(s_x)\ar[r]\ar[u]^{i_\rho}&\operatorname{hom}_{\a}(l , s_x)}\ ] ] the map @xmath1651 is an isomorphism since @xmath1293 belongs to @xmath921 , and @xmath832 induces an isomorphism @xmath1652 since @xmath832 is a quotient functor and @xmath19 is simple ; see lemma [ le : quotient_simple ] .
finally observe that @xmath921 is equivalent to @xmath1012 .
thus the induced map @xmath1653 is an isomorphism because we may assume that @xmath1293 corresponds to the structure sheaf ; see remark [ re : hom - simple ] .
it follows that @xmath1654 .
having shown the assertion for a specific line bundle , we apply lemma [ le : lb ] to verify the assertion for an arbitrary line bundle .
thus we need to show that for any pair @xmath1279 of line bundles and each monomorphism @xmath1522 with cokernel in @xmath302 , the object @xmath1293 is hom - simple if and only if @xmath1310 is hom - simple . using induction on the length of the cokernel of @xmath76
, we may assume that the cokernel is simple .
the exact sequence @xmath1655 induces for each simple object @xmath32 an exact sequence @xmath1656 if @xmath1293 is hom - simple , then a simple calculation shows that @xmath1635 is an isomorphism .
thus @xmath428 is an isomorphism and it follows that @xmath1310 is hom - simple .
the same argument shows that @xmath1293 is hom - simple if @xmath1310 is hom - simple .
the following result is due to kussin ; see ( * ? ? ? * proposition 4.2.3 ) .
[ pr : shift ] let @xmath1279 be two line bundles in @xmath31 .
then there exists an equivalence @xmath1581 taking @xmath1293 to @xmath1310 .
in particular , each line bundle is exceptional .
we apply lemma [ le : lb ] .
thus we may assume that there is an exact sequence @xmath1657 with @xmath109 of finite length .
using induction on the length of @xmath109 , we may even assume that @xmath109 is simple .
suppose that @xmath109 belongs to @xmath1211 with @xmath1209 .
then @xmath220 induces an isomorphism @xmath1658 by proposition [ pr : hom_lb ] and serre duality .
thus @xmath1659 and the tubular mutation given by @xmath1625 sends @xmath1293 to @xmath1310 ; see corollary [ co : mutat ] .
let @xmath0 be a field and @xmath31 a @xmath0-linear abelian category satisfying ( h1)(h5 ) .
we associate to @xmath31 a weight function and show that it is an invariant of @xmath31 which determines @xmath31 up to equivalence . in the following
we identify the projective linear group @xmath1161 with @xmath1660 ; see proposition [ pr : aut ] . a _ weight function _
@xmath1661 is a map which assigns to each closed point @xmath437 of @xmath1 a positive integer @xmath1662 such that @xmath1663 for almost all @xmath437 .
two weight functions @xmath1664 are _ equivalent _ if there exists some linear transformation @xmath1665 such that @xmath1666 for every closed point @xmath437
. given a collection @xmath1667 of distinct closed points @xmath1668 , and a sequence @xmath3 of positive integers , there is associated a weight function @xmath1669 , where @xmath1670 for @xmath1671 and @xmath1672 otherwise .
it follows from our convention that each weight function @xmath1673 corresponding to a weighted projective line satisfies @xmath1663 if the point @xmath437 is not rational .
[ th : canon_tiltobj ] let @xmath31 be a @xmath0-linear abelian category satisfying _ ( h1)(h5)_. for each line bundle @xmath1293 , there exists a canonically defined tilting object @xmath1580 which is unique up to isomorphism .
moreover : 1 .
the object @xmath1580 determines parameters @xmath1556 such that @xmath1674 .
the parameters @xmath1556 depend on a choice and any other choice gives parameters @xmath1675 such that the weight functions @xmath1669 and @xmath1676 are equivalent .
2 . let @xmath155 be a second line bundle .
then there exists an equivalence @xmath1581 taking @xmath1293 to @xmath155 .
thus @xmath1583 is isomorphic to @xmath1677 and the associated weight functions are equivalent .
let @xmath1558 be the collection of distinct points @xmath1209 with @xmath1540 and set @xmath3 with @xmath1559 for each @xmath181 .
we apply proposition [ pr : hom_lb ] and choose for each @xmath181 a simple object @xmath1678 such that @xmath1679 .
thus @xmath1680 by serre duality , and lemma [ le : filtr ] yields a chain of monomorphism @xmath1681 with @xmath1682 for all @xmath730 .
note that each object @xmath1593 is a line bundle and therefore exceptional .
the class @xmath1683=[l]-\sum_{j=1}^{p(x)}[\tau^{j}s_i]$ ] does not depend on @xmath181 since the inclusion @xmath924 sends the class of the simple object @xmath1684 of @xmath1685 to @xmath1686 $ ] ; see lemma [ le : simple ] . here
, we use that @xmath1684 is a simple object of degree one ; so its class in @xmath1687 is independent of @xmath181 by lemma [ le : homog_except ] .
thus the object @xmath1688 does not depend on @xmath181 by proposition [ pr : hr1 ] , and we denote it by @xmath1310 .
the construction of the @xmath1593 is parallel to the construction in the proof of theorem [ th : tiltobj ] , and we refer to the commutative diagram illustrating it .
next observe that each object @xmath1593 depends only on @xmath1293 and @xmath1165 because its class has this property and the object is exceptional .
thus the object @xmath1689 depends up to isomorphism only on @xmath1293 .
in particular , the object equals up to equivalence the tilting object constructed in the proof of theorem [ th : tiltobj ] , because for each pair of line bundles @xmath628 there exists an equivalence @xmath1581 taking @xmath155 to @xmath631 , by proposition [ pr : shift ] .
it follows that @xmath1580 is a tilting object for @xmath31 .
\(1 ) choosing a basis @xmath1560 of @xmath1561 , we obtain rational points @xmath1574 $ ] in @xmath1 such that @xmath1590 , where @xmath1690 is the composite of the morphisms in .
note that each @xmath1690 depends on the choice of the @xmath1691 , but it is unique up to a non - zero scalar .
it follows that @xmath1583 is isomorphic to @xmath1584 , where @xmath2 .
any other choice of the basis @xmath1560 gives another set of parameters @xmath1692 and a linear transformation @xmath1665 such that @xmath1693 for each @xmath181 .
thus the weight function @xmath1669 is unique up to equivalence .
\(2 ) apply proposition [ pr : shift ] .
there is an analogue of theorem [ th : canon_tiltobj ] with the canonical algebra @xmath1584 replaced by the squid algebra @xmath1546 .
the following example gives an explicit description of the tilting object for the category of coherent sheaves on a weighted projective line ; see ( * ? ? ?
* proposition 4.1 ) .
[ ex : can ] let @xmath1694 be a weighted projective line . then @xmath1695 is the tilting object for @xmath5 that is associated to the line bundle @xmath14 .
the endomorphism algebra is isomorphic to the canonical algebra @xmath1584 .
it is now a consequence of theorem [ th : canon_tiltobj ] that a @xmath0-linear abelian category satisfying ( h1)(h5 ) is determined up to equivalence by its associated weight function .
[ co : wpl_wf ] for two weighted projective lines @xmath1694 and @xmath1696 over a field @xmath0 , the following are equivalent : 1 .
the weight functions @xmath1669 and @xmath1676 are equivalent .
the algebras @xmath1584 and @xmath1697 are isomorphic .
the categories @xmath5 and @xmath1698 are equivalent .
\(1 ) @xmath186 ( 2 ) : suppose that @xmath1669 and @xmath1676 are equivalent via some linear transformation @xmath1699}$ ] in @xmath1161 .
thus we may assume that the points of @xmath1200 and @xmath1692 are related via @xmath1700 for @xmath1701 .
the algebra @xmath1584 is generated by a collection of arrows @xmath1702 ; analogously @xmath1697 is generated by arrows @xmath1703 .
we obtain an isomorphism @xmath1704 by defining @xmath1705 , @xmath1706 and @xmath1707 ( @xmath1701 ) .
\(2 ) @xmath186 ( 3 ) : the category @xmath5 admits a tilting object @xmath32 with endomorphism algebra @xmath1584 , which is obtained from the tilting object @xmath1545}\oplus\dots\oplus s_1^{[p_1 - 1]})\oplus \dots \oplus(s_n^{[1]}\oplus\dots\oplus s_n^{[p_n-1]})\ ] ] by modifying it as in the proof of theorem [ th : tiltobj ] .
analogously , @xmath1698 admits a tilting object @xmath567 with endomorphism algebra @xmath1697 .
if both algebras are isomorphic , then proposition [ pr : der_equiv ] implies that @xmath5 and @xmath1698 are equivalent .
\(3 ) @xmath186 ( 1 ) : suppose that there is an equivalence @xmath1708 .
we apply theorem [ th : canon_tiltobj ] .
thus the functor @xmath81 takes for any line bundle @xmath1293 of @xmath5 the canonically defined tilting object @xmath1580 to @xmath1709 . the associated weight function for @xmath1580 is equivalent to @xmath1669 , whereas for @xmath1709 it is equivalent to @xmath1676 .
it follows that @xmath1669 and @xmath1676 are equivalent .
in this section we list a few topics which have attracted interest in the past , and which are areas of present research .
the lists of papers is certainly not complete and we refer to the references in the listed papers for more information .
w. geigle and h. lenzing , a class of weighted projective curves arising in representation theory of finite - dimensional algebras , in _ singularities , representation of algebras , and vector bundles ( lambrecht , 1985 ) _ , 265297 , lecture notes in math . ,
1273 , springer , berlin , 1987 .
graded singularities : the study of weighted projective lines in terms of graded singularities ( maximal cohen - macaulay modules , vector bundles , the triangulated category of singularities in the sense of buchweitz and orlov ) .
h. lenzing and j. a. de la pea , spectral analysis of finite dimensional algebras and singularities , in _ trends in representation theory of algebras and related topics _ , 541588 , eur .
soc . , zrich , 2008 .
kac s theorem , hall algebras : the theorem characterizes the dimension types of indecomposable coherent sheaves over weighted projective lines in terms of loop algebras of kac - moody lie algebras ; the proof uses hall algebras .
m. auslander , i. reiten and s.o .
smal , _ representation theory of artin algebras _ , cambridge univ . press ,
cambridge , 1995 .
a. a. belinson , coherent sheaves on @xmath1711 and problems in linear algebra , funktsional .
i prilozhen .
* 12 * ( 1978 ) , no . 3 , 6869 .
a. a. belinson , j. bernstein and p. deligne , faisceaux pervers , in _ analysis and topology on singular spaces , i ( luminy , 1981 ) _ , 5171 ,
astrisque , 100 , soc .
france , paris , 1982 .
g. d. birkhoff , a theorem on matrices of analytic functions , math . ann .
* 74 * ( 1913 ) , no . 1 , 122133 . s. brenner and m. c. r. butler , generalizations of the bernstein - gelfand - ponomarev reflection functors , in _ representation theory , ii ( proc .
second internat .
carleton univ . ,
ottawa , ont .
, 1979 ) _ , 103169 , lecture notes in math . , 832 , springer , berlin , 1980
x. w. chen and h. krause , expansions of abelian categories , arxiv:1009.3456 .
e. enochs , s. estrada and b. torrecillas , an elementary proof of grothendieck s theorem , in _
abelian groups , rings , modules , and homological algebra _ , 6773 ,
chapman & hall / crc , boca raton , fl , 2006 .
p. gabriel , des catgories abliennes , bull .
france * 90 * ( 1962 ) , 323448 .
gabriel and m. zisman , _ calculus of fractions and homotopy theory _ , springer - verlag new york , inc . ,
new york , 1967 .
w. geigle and h. lenzing , a class of weighted projective curves arising in representation theory of finite - dimensional algebras , in _ singularities , representation of algebras , and vector bundles ( lambrecht , 1985 ) _ , 265297 , lecture notes in math . , 1273 , springer , berlin , 1987 . w. geigle and h. lenzing , perpendicular categories with applications to representations and sheaves , j. algebra * 144 * ( 1991 ) , no . 2 , 273343
a. grothendieck , sur la classification des fibrs holomorphes sur la sphre de riemann , amer . j. math . * 79 * ( 1957 ) , 121138 .
d. happel , a characterization of hereditary categories with tilting object , invent .
* 144 * ( 2001 ) , no . 2 , 381398 .
d. happel , i. reiten and s. o. smal , tilting in abelian categories and quasitilted algebras , mem .
amer . math .
* 120 * ( 1996 ) , no .
575 , viii+ 88 pp .
r. hartshorne , _ algebraic geometry _ , springer , new york , 1977 .
d. kussin , graduierte faktorialitt und die parameterkurven tubularer familien , dissertation , paderborn , 1997 .
h. lenzing , hereditary noetherian categories with a tilting complex , proc .
* 125 * ( 1997 ) , no . 7 , 18931901 .
h. lenzing , hereditary categories , in _ handbook of tilting theory _
, 105146 , cambridge univ .
press , cambridge , 2007 .
h. lenzing and h. meltzer , sheaves on a weighted projective line of genus one , and representations of a tubular algebra , in _ representations of algebras ( ottawa , on , 1992 ) _ , 313337 , amer
soc . , providence , ri , 1993 .
h. lenzing and h. meltzer , tilting sheaves and concealed - canonical algebras , in _ representation theory of algebras ( cocoyoc , 1994 ) _ , 455473 , amer .
providence , ri , 1996 .
h. meltzer , tubular mutations , colloq .
* 74 * ( 1997 ) , no . 2 , 267274 . c. m. ringel , _ tame algebras and integral quadratic forms _
, lecture notes in math . , 1099 , springer , berlin , 1984 . c. m. ringel , the canonical algebras , in _ topics in algebra , part 1 ( warsaw , 1988 ) _
, 407432 , pwn , warsaw , 1990 .
serre , faisceaux algbriques cohrents , ann . of math .
( 2 ) * 61 * ( 1955 ) , 197278 .
verdier , des catgories drives des catgories abliennes , astrisque no . 239 ( 1996 ) , xii+253 pp . c. a. weibel , _ an introduction to homological algebra _ , cambridge univ . press , cambridge , 1994 . | these notes provide a description of the abelian categories that arise as categories of coherent sheaves on weighted projective lines .
two different approaches are presented : one is based on a list of axioms and the other yields a description in terms of expansions of abelian categories .
a weighted projective line is obtained from a projective line by inserting finitely many weights .
so we describe the category of coherent sheaves on a projective line in some detail , and the insertion of weights amounts to adding simple objects .
we call this process ` expansion ' and treat it axiomatically .
thus most of these notes are devoted to studying abelian categories , including a brief discussion of tilting theory .
we provide many details and have tried to keep the exposition as self - contained as possible . |
according to the models of explosive nucleosynthesis , much of the iron comes from type ia supernovae , while @xmath4 elements such as o and mg come from type ii supernovae . because the difference in lifetime of the progenitors , it is generally considered that the iron enrichment delays relative to @xmath4 elements by 12 billion years ( @xcite ; @xcite , 1998 ) .
if feii / mgii , the relative strengths of feii emission lines and the mgii @xmath2 doublet , reflects the fe / mg abundance ratio , there will be a break in feii / mgii at high redshift . despite of much efforts made by many observational groups
( e.g. , @xcite ; @xcite ; @xcite , 2003 ; @xcite , 2004 ; @xcite ; @xcite ; @xcite ; @xcite , 2008a ; @xcite ) , there have been found no signs of such a break ; feii / mgii looks constant from low - redshift up to @xmath5 with large scatter .
no break in feii / mgii might reflect a significantly shorter delay - time of [email protected] gyr , as suggested by @xcite , @xcite , and @xcite .
the expected break can also be obscured by non - abundance effects .
simulations of feii emitting regions , assuming either photoionization or shocks , imply that the fe abundance is not only one parameter which controls the feii strength , but several non - abundance factors can also affect it .
such non - abundance factors include spectral energy distribution ( sed ) of the central source , strength of the radiation field , and the gas density of broad emission line region ( belr ) clouds .
recently , @xcite and @xcite pointed out that a large microturbulence velocity may be responsible for strong feii emission .
@xcite have studied non - abundance factors by using spectra of a low - redshift sample of 14 quasars , covering wide rest - wavlengths 1000@xmath17300 , and claimed that the feii strength correlates with the mass of the central black hole , the line width , and the x - ray photon index . in this paper
, we present spectra of six quasars at @xmath6 , and compare with those in the low - redshift sample . throughout this paper ,
a cosmology with @xmath7 , and @xmath8 is assumed .
six quasars were selected for optical spectroscopy from the catalog by @xcite . according to the catalog , these are luminous with @xmath9
@xmath10 at @xmath11 @xmath12 , bright enough to take optical feii emission lines through near - infrared spectroscopy at a later opportunity .
gmos on gemini - south telescope are used in the long - slit mode with grating r150_g5326 and order sorting filter og515_g0330 .
wavelengths observed are in 5400@xmath19800 , corresponding to rest wavelengths 1800@xmath13300 at @xmath6 quasars .
the slit width is 1.0 and the spectral resolution is 3.286 pixel@xmath13 . the grating was centered at 8150 for the first three exposures and changed to 8250 for the following exposures , filling up the gaps of the ccd chip array .
wavelengths were calibrated using the cuar arc lamp taken at the both central wavelengths of the grating .
ltt1788 ( @xcite ) was used for flux - scaling .
the observing log is summarlized in table [ tab : obslog ] .
individual spectral frames were processed using the gemini iraf package ver1.9.1 .
sky of individual frames is subtracted with gsskysub , and then scombine was used to combine into the final spectral frame .
the corrections for telluric absorption were not applied .
reduced spectra are shown in figure [ fig : spec ] .
redshifts are determined from fit of mgii emission line . with these measured redshifts ,
absolute @xmath14 magnitudes are calculated assuming a cosmology with @xmath7 , and @xmath8 .
redshifts and absolute @xmath14 magnitudes are also listed in table [ tab : obslog ] . [ cols="<,^,^,^,^,^,^,^ " , ]
in table [ tab : measure ] , the feii emission line flux _ feii_@xmath15 , the mgii line flux _
mgii(total ) _ , the fwhm ( mgii ) , the 3000 luminosity @xmath16 , and the blackhole mass @xmath17 derived from equation ( [ eq : mclure ] ) , are given . note
that , for b0226 - 104 , mgii emission line is heavily affected by telluric absorption line , so that we did not measure mgii flux and mgii fwhm . a plot of _
feii_@xmath15/_mgii(total ) _ against @xmath17 is given in figure [ fig : femg]@xmath18 and 3000 luminosity @xmath16 in figure [ fig : femg]@xmath19 .
analyzing 14 low - redshift quasars , @xcite found the correlation between the flux ratio feii(uv)/mgii and the blackhole mass .
this relation is shown by the dotted line in figure [ fig : femg]@xmath18 .
filled circles are our quasars at @xmath20 and open circles are low - redshift quasars at @xmath3 by @xcite .
our quasars have an absolute luminosity of @xmath21 , which are much luminous than the low - redshift quasars having an absolute luminosity of @xmath22 .
as can be seen in figure [ fig : femg]@xmath18 , the five @xmath0 quasars do not follow the correlation found by @xcite .
all of them have feii(uv)/mgii greater than expected from the tsuzuki s correlation . what is the cause of the large feii(uv)/mgii value in the @xmath0 quasars relative to the low - redshift quasars ? because the five @xmath0 quasars are much luminous than the 14 low - redshift quasars , the luminosity effect is examined .
as shown in figure [ fig : femg]@xmath19 , the luminosity effect is not responsible for the large feii(uv)/mgii value in the @xmath0 quasars .
the _ real _ cause would be evolution in feii(uv)/mgii or non - abundance effects such as the sepectral energy distribution of the continuum from the central source , the strength of the radiation field and the gas density of blr clouds as well as the microturbulence of blr gas ( @xcite ; @xcite ) .
further investigations are required using large samples of quasars .
this work was financially supported in part by grant - in - aid for scientific research ( 17104002 ) and specially promoted reserch ( 20001003 ) and the japan - australia research cooperative program from jsps .
baldwin , j. a. , ferland , g. j. , korista , k. t. , hamann , f. , & lacluyz , a. 2004 , apj , 615 , 610 dietrich , m. , hamann , f. , appenzeller , i. , & vestergaard , m. 2003 , apj , 596 , 817 dietrich , m. , appenzeller , i. , vestergaard , m. , & wagner , s. j. 2002 , apj , 564 , 581 elston , r. , thompson , k. l. , & hill , g. j. 1994 , nature , 367 , 250 freudling , w. , corbin , m. r. , & korista , k. t. 2003 , apj , 587 , l67 friaa , a. c. s. , & terlevich , r. j. 1998 , mnras , 298 , 399 granato , g. l. , de zotti , g. , silva , l. , bressan , a. , & danese , l. 2004 , apj , 600 , 580 hamann , f. , & ferland , g. 1993 , apj , 418 , 11 hamuy , m. , suntzeff , n. b. , heathcote , s. r. , walker , a. r. , gigoux , p. , & phillips , m. m. 1994 , pasp , 106 , 566 iwamuro , f. , kimura , m. , eto , s. , maihara , t. , motohara , k. , yoshii , y. , & doi , m. 2004 , apj , 614 , 69 iwamuro , f. , motohara , k. , maihara , t. , kimura , m. , yoshii , y. , & doi , m. 2002 , apj , 565 , 63 kawara , k. , murayama , t. , taniguchi , y. , & arimoto , n. 1996 , apj , 470 , l85 kurk , j. d. , et al .
2007 , apj , 669 , 32 maiolino , r. , juarez , y. , mujica , r. , nagar , n. m. , & oliva , e. 2003 , apj , 596 , l155 matsuoka , y. , kawara , k. , & oyabu , s. 2008a , apj , 673 , 62 matsuoka , y. , oyabu , s. , tsuzuki , y. , & kawara , k. 2007 , apj , 663 , 781 matteucci , f. , & recchi , s. 2001 , apj , 558 , 351 mclure , r. j. , & jarvis , m. j. 2002 , mnras , 337 , 109 pei , y. c. 1992 , apj , 395 , 130 schlegel , d. j. , finkbeiner , d. p. , & davis , m. 1998 , apj , 500 , 525 tsuzuki , y. , kawara , k. , yoshii , y. , oyabu , s. , tanab , t. , & matsuoka , y. 2006 , apj , 650 , 57 verner , e. , bruhweiler , f. , verner , d. , johansson , s. , & gull , t. 2003 , apj , 592 , l59 vron - cetty , m .-
, & vron , p. 2003 , a&a , 412 , 399 yoshii , y. , tsujimoto , t. , & kawara , k. 1998 , apj , 507 , l113 yoshii , y. , tsujimoto , t. , & nomoto , k. 1996 , apj , 462 , 266 | we present spectra of six luminous quasars at @xmath0 , covering rest wavelengths 1600@xmath13200 .
the fluxes of the uv feii emission lines and mgii @xmath2 doublet , the line widths of mgii , and the 3000 luminosity were obtained from the spectra .
these quantities were compared with those of low - redshift quasars at @xmath3 studied by tsuzuki et al . in a plot of
the feii(uv)/mgii flux ratio as a function of the cental black hole mass , feii(uv)/mgii in our @xmath0 quasars is systematically greater than in the low - redshift quasars .
we confermed that luminosity is not responsible for this excess .
it is unclear whether this excess is caused by rich fe abundance at @xmath0 over low - redshift or by non - abundance effects such as high gas density , strong radiation field , and high microturbulent velocity .
galaxies : abundances galaxies : active line : formation quasars : emission lines |
the study of the internal dynamics of stellar systems plays an essential role in astronomy . from the observed positions and velocities of the stars in galaxies and globular clusters it is possible to infer their total ( dark+luminous ) mass distribution , which , in particular , provides information on the presence and properties of dark halos and massive black holes . in turn
, this structural knowledge constrains theories for the formation and evolution of these systems .
the dynamical state of a stellar system is determined by its phase space distribution function , @xmath3 , which counts the stars as a function of position @xmath4 and velocity @xmath5 . typically , however , only three of the six phase - space coordinates are available observationally : the projected sky position @xmath6 , and the velocity @xmath7 along the line of sight ( los ) .
proper motion observations can provide the additional velocities @xmath8 , but such data are generally not available ( with the notable exception of some galactic globular clusters ) . to make progress with the limited information available , the dynamical theorist is often forced to make simplifying assumptions about geometry ( e.g. , that the system is spherical ) or about the velocity distribution ( e.g. , that it is isotropic ) .
such assumptions can have strong effects on the inferred mass distribution ( @xcite ) . to obtain the most accurate results
it is therefore important to make models that are as general as possible .
of particular importance for collisionless , unrelaxed systems such as galaxies is to constrain the velocity anisotropy using available data , rather than to assume it a priori . in a collisionless system
the distribution function satisfies the collisionless boltzmann equation .
analytical methods to find solutions of this equation usually rely on the jeans theorem , which states that the distribution function must depend on the phase - space coordinates through integrals of motion ( quantities that are conserved along a stellar orbit ) . in a spherical system
all integrals are known analytically , namely , the energy @xmath9 and the components of the angular momentum vector @xmath10 .
analytical models for spherical systems are therefore fairly easily constructed . in an axisymmetric system things
are more complicated ( e.g. , @xcite ) .
only two integrals are known analytically , @xmath9 and the vertical component @xmath11 of the angular momentum vector denote the coordinates intrinsic to the axisymmetric stellar system , with the plane @xmath12 being the equatorial plane , and @xmath13 the symmetry axis .
these relate via the inclination @xmath14 to the observable coordinates @xmath6 on the plane of the sky ( aligned , respectively , along the projected major and minor axes of the stellar system ) , and @xmath15 the line - of - sight direction , positive in the direction away from us .
] , but there is generally a third integral for which no analytical expression exists .
therefore , it is not generally possible to construct an axisymmetric model analytically .
the special class of so - called ` two - integral ' ( @xmath16 ) models ( e.g. , @xcite ) has its uses ( e.g. , @xcite ) , but these have an isotropic velocity distribution in their meridional plane , which need not be a good fit to real dynamical systems . the most practical way to model
a general axisymmetric system is to do it numerically .
while a few methods exist to do this ( e.g. , @xcite ) , the most common approach uses schwarzschild s ( 1979 ) method .
one starts with a trial guess for the gravitational potential @xmath17 and then numerically calculates an orbit library that samples integral space in some complete and uniform way .
the orbits are integrated for several hundred orbital periods , and the time - averaged intrinsic and projected properties ( density , los velocity , etc . )
are stored as the integration progresses .
the construction of a model consists of finding a weighted superposition of the orbits that : ( 1 ) reproduces the observed stellar or surface brightness distribution on the sky ; and ( 2 ) reproduces all available kinematical data to within the observational error bars .
additional constraints can be added to enforce that the distribution function in phase space be smooth and reasonably well behaved , e.g. , through regularization or by requiring maximum entropy .
several axisymmetric schwarzschild codes have been developed in the last decade ( e.g. , @xcite ) .
these codes deal with the situation in which information on the line - of - sight velocity distribution ( losvd ) is available for a set of positions on the projected plane of the sky .
this is the case , e.g. , when the kinematical data are from long - slit or integral - field spectroscopic observations of unresolved galaxies .
the optimization problem for such data can be reduced to a linear matrix equation for which one needs to find the least - squares solution with non - negative weights @xcite .
one dimension of the matrix corresponds to the number of orbits in the library , while the other corresponds to the number of ( luminosity , kinematical and regularization ) constraints that must be reproduced .
both dimensions are typically in the range @xmath18@xmath19 .
nonetheless , efficient numerical algorithms exist to find the solution , which yield the orbital and the velocity distribution of the model , as well as the @xmath20 of the fit to the kinematical data .
the procedure must then be iterated with different gravitational potentials , to determine the potential that provides the overall best @xmath20 .
the existing codes have been used and tested extensively ( e.g. , @xcite ) .
some questions remain , e.g. , about the importance of smoothing in phase space , the exact meaning of the confidence regions determined using @xmath21 contours , and , in some situations , valid concerns have been raised regarding whether the available data contain enough information so as to warrant the conclusions of the schwarzschildmodeling @xcite .
nevertheless , on the whole schwarzschild codes have now been established as an accurate and versatile tool to study a wide range of dynamical problems .
a disadvantage of the existing codes is that they can not be easily applied to the large class of problems in which the kinematical observations come in the form of discrete velocity measurements , rather than as losvds .
this is encountered , e.g. , when modeling the dynamics of galaxies at large radii , where the low - surface brightness prevents integrated - light spectroscopy .
the only available data are then often of a discrete nature , e.g. , via the los velocities of individual stars in galaxies of the local group ( e.g. , @xcite ) , or via planetary nebulae ( e.g. , @xcite ) and globular clusters ( e.g. , @xcite ) surrounding giant ellipticals .
the kinematical data available for clusters of galaxies , consisting of redshifts for individual galaxies , are of a similarly discrete nature ( e.g. , @xcite ) .
the typical datasets in all these cases consist of tens to hundreds of los velocities .
galactic globular clusters constitute another class of object for which kinematical data is often available only as discrete measurements , rather than in the form of losvds . from ground - based observations , data sets of individual los velocities
can be available for up to thousands of stars in these systems ( e.g. , @xcite ) , and for @xmath22 cen it has been possible to assemble large samples of proper motions as well @xcite . with the capabilities of _ hst _ , accurate proper motion data sets with up to @xmath23 stars are now becoming available for several more galactic globular clusters ( e.g. , @xcite ) .
note that discrete datasets do not necessarily provide better or worse information than datasets obtained from integrated - light measurements .
both types of data have their advantages and disadvantages . for discrete datasets , for example
, interloper contamination can be a problem ( see also the end of section [ sec : logl ] below ) .
by contrast , for integrated - light measurements , it is often difficult to constrain the wings of the losvd due to uncertainties associated with continuum subtraction .
which type of data is most appropriate and most easily obtained depends on the specific object under study .
this is therefore not a question that we address in this paper .
instead , we focus on the issue of how to best analyze discrete data , if that happens to be what is available
. analyses of discrete datasets have often been more simplified than the analyses that are now common for integrated - light data .
for example , the observations are analyzed using the jeans equations ( e.g. , @xcite ) , often with the help of data binning to calculate rotation velocity and velocity dispersion profiles ( see , however , the `` spherical '' schwarzschild models of m87 of @xcite ) .
the disadvantage of such an approach is that not all the information content of the data is used , including information on deviations of the velocity histograms from a gaussian .
such deviations are important because they constrain the velocity dispersion anisotropy of the system ( e.g. , @xcite ) .
this anisotropy is an important ingredient in some existing controversies , e.g. regarding the presence of dark halos around elliptical galaxies @xcite .
loss of information can be avoided when large numbers of datapoints are available , as is often the case for globular clusters .
it is then possible to create velocity histograms for binned areas on the projected plane of the sky , after which analysis can be done with existing schwarzschild codes ( e.g. , @xcite ) .
while this is possible for large datasets , such an approach is not viable for the more typical , smaller datasets that are often available .
the availability of schwarzschild codes that can fully exploit the information content of such smaller datasets would therefore be valuable to advance this subject .
motivated by these considerations we set out to adapt our existing schwarzschild code @xcite to deal with discrete datasets .
this does not constitute a trivial change , since it changes the constrained superposition procedure from a linear matrix problem to a more complicated maximum likelihood one .
for each observed velocity of a particle in the system the question becomes : what is the probability that this velocity would have been observed if the model is correct ?
the overall likelihood of the data , given a trial model , is the product of these probabilities for all observations .
such likelihood problems have previously been solved for spherical systems @xcite and the special class of axisymmetric @xmath24 systems @xcite . however , for the axisymmetric schwarzschild modeling approach the problem corresponds to finding the minimum of a function in a space with a dimension of @xmath18@xmath19 .
we show in this work , via the schwarzschildmodeling of simulated datasets , that this problem can indeed be solved successfully and efficiently .
moreover , we follow @xcite and implement in our new code the ability to calculate and fit proper motions in addition to los velocities .
applications of the code to real datasets will be presented in forthcoming papers .
the structure of the paper is as follows . in section [ sec : logl ]
we phrase the new problem of fitting a schwarzschildmodel to a dataset of discrete velocities ( of one , two , or three dimensions ) of individual kinematic tracers in terms of a likelihood formalism .
section [ sec : code ] describes the implementation of the discrete fitting procedure into our existing schwarzschildcode . at the same time , we summarize here the major steps involved in the construction of the probability matrix that describes the likelihood of a given kinematic data point belonging to some particular orbit of the library .
we then present in section [ sec : tools ] sets of simulated data that we use for the purpose of testing the performance of the discrete schwarzschildcode .
we also describe the known input distribution functions from which these data were drawn .
the application of the code to the simulated datasets is presented in section [ sec : tests ] .
we present a thorough analysis of the accuracy with which our discrete schwarzschildcode recovers the known distribution function , mass - to - light ratio and inclination used to generate the simulated data .
finally , in section [ sec : end ] we summarize our findings and present our conclusions .
in the schwarzschildscheme the properties of every orbit @xmath25 in the orbit library are computed and stored .
the modeling consists in finding the superposition of orbital weights @xmath26 , i.e. , the fraction of particles in the system residing in each orbit , that best reproduces some set of constraints .
the weights are written as squares to ensure that they never become negative .
linear constraints are of the form @xmath27 here @xmath28 is a constraint value that needs to be reproduced , @xmath29 is its uncertainty , and @xmath30 is its model prediction @xmath31 the matrix @xmath32 represents here , for orbit @xmath25 , the probability distribution corresponding to the constraint @xmath33 .
the constraints are generally one of the following : ( a ) the integrated light ( surface brightness ) of a stellar population in some aperture number in the projected plane of the sky , necessary to reproduce an observational measurement of the surface brightness ; ( b ) the mean los velocity , velocity dispersion , or for data of sufficient quality , a higher - order gauss - hermite moment in some aperture number in the projected plane of the sky , necessary to reproduce an observational measurement of the stellar kinematics ; ( c ) the integrated mass in some meridional @xmath34 plane grid point , necessary to provide a consistent model ; ( d ) a combination of distribution function moments in some meridional @xmath34 plane grid point , if a model with a particular dynamical structure is desired ( e.g. , one may want a model with @xmath35 equal to zero in order to simulate a two - integral @xmath24 model ) ; ( e ) a combination of orbit weights , if regularization constraints are desired to enforce smoothness of the model in phase space ( e.g. , one can set the n - th order divided difference of adjacent orbit weights to zero , with an uncertainty @xmath36 that measures the desired amount of smoothing ) .
it is natural to choose the best - fitting model to be the one that produces the maximum likelihood . to determine the likelihood we need to write down an expression for the probability of measuring @xmath33 among all its possible values
. to do this , we recall that any model is not an attempt to reproduce a set of observations to infinite accuracy , but instead to do it within the uncertainty @xmath37 . for observational constraints , such as those in ( a ) and ( b ) above , @xmath37 is equal to the measurement uncertainty . for other constraints , such as those in ( c)-(e ) above
, @xmath37 can be used as a forcing parameter that compels how accurately the likelihood needs to peak around a particular value of @xmath33 . if one assumes that these uncertainties have a normal ( gaussian ) distribution , then the probability we are interested in is given by @xmath38 . \ ] ] the combined probability for the simultaneous occurrence of all @xmath39 linear constraints is then given by the product of the single probabilities , @xmath40 .
using equation ( [ eq.non.linear.term ] ) , the logarithm of this linear part of the likelihood is therefore @xmath41 the first sum on the right - hand side of this expression does not depend on the orbital weights @xmath26 and , therefore , does not affect the likelihood maximization .
the second term has the exact form of the @xmath20 statistic .
maximizing the likelihood is therefore equivalent to the minimization of this @xmath20 .
this can be done by finding the solution of the set of equations ( [ eq.constraint ] ) and ( [ eq.gamma.def ] ) , which can be rewritten as an overdetermined matrix equation . this matrix equation can be solved with the use of standard non - negative least - squares ( nnls ) algorithms ( see @xcite for a detailed description ) . in the case of discrete data
, however , the introduction of constraints of a `` non - linear '' type is inevitable in order to adequately exploit the entire information content available , avoiding restrictive simplifications and loss of information due to binning .
this occurs because the individual probabilities do not necessarily have the simple , gaussian form of equation ( [ eq.non.linear.term ] ) .
the procedure for finding the maximum likelihood then can not be cast as the solution of a linear matrix equation anymore .
suppose we have a kinematic dataset consisting of discrete measurements which we are trying to model using the schwarzschildtechnique .
let @xmath42 be the phase - space probability distribution of any given orbit , properly averaged azimuthally , and normalized such that @xmath43 .
we use @xmath44 to denote a vector of up to six euclidean spatial and velocity coordinates .
whenever @xmath44 is shorter than 6 elements , it is understood that the distribution has been marginalized over the missing dimensions .
then the total probability of drawing a particle from a superposition of orbits representing the whole system is @xmath45 we now need to consider the total probability of the ensemble of @xmath46 particles with kinematic information that constitute our discrete dataset . before this , however , it is necessary to make the distinction , in the language of probabilities , between the possible modes of sampling of the tracers available in a system of particles .
the two main possibilities depend on whether the particles are randomly or non - randomly drawn from their spatial distribution , and we may refer to these , respectively , as random positional sampling and incomplete positional sampling . additionally , particles may be drawn with or without velocity information , thus adding up to a total of four possibilities .
the case with incomplete positional sampling and no velocity information , however , does not provide any useful constraint to the analysis and therefore we restrict the discussion to the remaining three cases . for particles drawn randomly from the spatial distribution with no velocity information ,
the probability @xmath47 is @xmath48 where * r * represents a 2 or 3 dimensional position .
this type of dataset could result from imaging of the resolved populations of a stellar system , where the positional information could be used as actual constraints .
this would force the model to fit the underlying spatial distribution of discrete tracers , instead of making use of a parametrization of the ( continuous ) brightness profile of the system .
of course , a dataset without velocity information can not by itself constrain the dynamical state or the mass of the system . in the case of random positional sampling
including velocity information , particles are randomly drawn from both the spatial and velocity distributions . in this case , @xmath47 has the form @xmath49 where @xmath50 is the same as above and @xmath51 represents a general 1 , 2 , or 3 dimensional velocity
. this would be the case when being able to obtain the velocities of particles in a given field without introducing any spatial or velocity bias , such as the proper motions of all stars ( brighter than some magnitude limit ) in a sufficiently sparse stellar cluster , or when los velocities are obtained for a complete ( or possibly magnitude - limited ) set of globular clusters or planetary nebulae in a galaxy .
in contrast , having _ incomplete _ positional sampling means that the particles are drawn from a velocity distribution , with _ a priori _ fixed positions .
this can occur , for example , when because of the usually limited availability of telescope time and resources , los velocities are measured only for stars within some distance from the photometric major or minor axes of a galaxy , or when because of the finite size of fibers in a fiber - fed spectrograph , not all the potentially observable kinematic tracers in the field can be actually acquired .
incomplete positional sampling also arises when , even though particles can be randomly drawn spatially , this is the case only for a limited area .
this occurs , for example , when the observations have to avoid the innermost regions of a galaxy or globular cluster , where , because of crowding , stars can not be individually resolved . in these case
, @xmath47 has the form @xmath52 where , rather than just @xmath26 , the effective weights when summing together the individual orbital distributions are @xmath53 .
once the individual probabilities for all possible cases of spatial sampling that comprise the data have been properly assigned , we can proceed to the construction of the total probability of observing the entire dataset .
let @xmath54 and @xmath55 be the number of observational data points obtained under the mode of random positional sampling without and with velocity information , respectively , and @xmath56 the number of data points obtained with incomplete positional sampling with kinematic information .
then , the total probability is simply the product of the individual probabilities , with logarithm given by @xmath57 .
using equations ( [ eq.prob.r ] ) to ( [ eq.prob.rfix ] ) , and adopting the abbreviated notation @xmath58 , @xmath59 , and @xmath60 ( all known for each orbit @xmath25 and particle @xmath14 from the orbit library calculation ; see 3 ) , the quantity to maximize becomes @xmath61 joining the results in equations ( [ eq : loglinear ] ) and ( [ eq.total.likelihood.1 ] ) , the complete log - likelihood for a general application of the schwarzschildmethod , which is the full expression to be maximized with respect to the orbital weights @xmath62 , is the sum of the log - likelihoods for linear and discrete constraints @xmath63 finding the maximum likelihood corresponds to finding the solution of @xmath64 = 0 , for all @xmath65 . denoting @xmath66 ,
the expression for the first derivative is @xmath67 one important question that remains is regarding the estimation of confidence regions around the parameters of the best - fitting model , i.e. , the ( statistical ) uncertainties around the likelihood maximum in the case of non - linear constraints . recalling that maximizing @xmath68 is equivalent to minimizing the quantity @xmath69 , it is easy to realize that , if the probabilities involved in equation ( [ eq.total.likelihood.1 ] ) were all of gaussian form , then @xmath70 would simply reduce to the well known @xmath20 statistic , as we have already seen for the case with linear constraints in equation ( [ eq : loglinear ] ) .
when dealing with non - linear constraints , however , the likelihood does not reduce to a simple @xmath20 form .
nevertheless , one still can use another well known theorem of statistics which , used before by @xcite and @xcite , states that the `` likelihood - ratio '' statistic @xmath71 does tend to a @xmath20 statistic in the limit of large @xmath46 , with the number of degrees of freedom equal to the number of free parameters that have not yet been varied and chosen so as to optimize the fit .
therefore , the likelihood - ratio statistic reduces to the @xmath72 statistic for @xmath73 , even though the probabilities in equation ( [ eq.total.likelihood.1 ] ) are not all individually gaussian .
since in the present work we explore datasets consisting of 100 kinematic measurements or more , the condition of large @xmath46 should be reasonably fulfilled .
therefore , following the likelihood - ratio statistic , we assume @xmath74 , and compute the confidence regions around the best - fit parameters in the usual way ( e.g. , @xcite ) , i.e. , with the @xmath75 error for a single parameter corresponding to wherever @xmath76 , and so forth . other approaches to quantify the uncertainties exist as well , e.g. , using bayesian statistics , but these are generally more difficult to implement ( e.g. , @xcite ) .
the equations described above assume that any possible `` interloper '' contaminants have already been removed , and that the targets with observed velocities that enter the likelihood equations all belong to the system under study . for realistic datasets , contamination by interlopers can certainly be a problem @xcite ; i.e. , targets that happen to lie close to the line - of - sight of the stellar system under study and are difficult to reject from the sample .
however , efficient interloper rejection schemes do exist for various types of samples and these have been well - described in the literature @xcite .
moreover , the use of empirically - calibrated selection criteria ( independent of the measured velocity ) can produce extraordinarily clean samples for kinematic analysis @xcite .
either way , interloper rejection is best discussed in the context of specific data sets .
we therefore do not discuss it further in the present paper .
interloper rejection for discrete data sets can also be built in as part of the likelihood analysis @xcite , so a simple modification of the likelihood equations given above could deal with interlopers explicitly .
however , we have not yet explored this in the present context .
given equations ( [ eq.logl ] ) and ( [ eq.1st.deriv ] ) , fitting a schwarzschildmodel to the data requires the following two steps : ( a ) determination of all the individual probabilities @xmath77 and matrix elements @xmath32 , so that the only unknowns in equation ( [ eq.1st.deriv ] ) are the coefficients @xmath78 ; and ( b ) performing the maximization of the total likelihood ,
i.e. , finding the set of orbital weights @xmath78 that satisfies @xmath64 = 0 , for all @xmath65 , and therefore best fits the available constraints .
the elements of the matrix @xmath32 , corresponding to the linear constraints discussed in [ sec : logl ] , are calculated in the same way as in the old ( continuous ) implementation of the code , and for them we refer to @xcite , @xcite and @xcite . in what follows
we concentrate on the probabilities @xmath77 associated with the discrete treatment that is the subject of this work .
the matrix elements @xmath77 in equation ( [ eq.1st.deriv ] ) , which keep track of the probability that orbit @xmath25 of the library would have produced the measurement @xmath14 of the dataset ( each @xmath25 corresponding to some combination of the three integrals of motion @xmath9 , @xmath79 , and @xmath1 ) , are stored as the orbit in question is being computed .
that is , at every time step during the orbit integration , we check whether the position and velocity along the orbit is consistent with any of the observational datapoints . to accomplish this , it is necessary to implement some degree of _ smoothing _ , both in position and velocity space , since otherwise the probability of having a particle on an orbit at exactly the observed position and velocity would be infinitesimally small . smoothing in the spatial coordinates
is accomplished through the definition of an _ aperture _ around the position of each particle in the dataset , with the size of the aperture controlling the amount of smoothing .
the optimal aperture size will be somewhat dependent on the sampling characteristics of the data .
in general , apertures should not be too small , or otherwise few time steps during orbit integration will fall on any one of them .
this would lead to large shot noise in the computed probabilities @xmath77 , unless the orbits are integrated for very long times .
nor should the apertures be too large , so that information on the orbital structure of the model is not erased by excessive spatial smoothing .
the choice of aperture shape is arbitrary and a matter of numerical convenience .
we adopt square apertures as in previous implementations of the code ( long - slit observations naturally produce data for rectangular apertures ) , and set their sizes to a user - supplied fraction of @xmath80 , the radius in the projected plane at the aperture s position . once the spatial apertures are defined , and every time the projected position along the orbit being integrated falls within an aperture , we need to keep track of whether the orbital velocity matches the observed velocity . in the old ( continuous ) implementation ,
the losvd was computed and stored for every orbit @xmath25 at each aperture @xmath14 , with the size of the bins in the histogram determining the amount of smoothing in velocity space . in our discrete treatment of the problem , @xmath77 would simply be the histogram value for the bin that contains the observed velocity
. a direct , though information ally incomplete , generalization of this implementation to kinematical data in three - dimensions would be to keep track of two additional histograms at each aperture to account for @xmath81 and @xmath82 .
this has been done by @xcite and @xcite , who calculated moments of the three model velocity distributions and fitted them to those obtained from binning los and proper - motion observations of stars in @xmath22 cen and m15 , respectively ( note that these studies still handle the data in a continuous fashion , by reducing the initially discrete datasets to binned velocity distributions at a number of apertures on the sky , an approach only possible thanks to the very large number of stars with measured velocities in these systems ) . while reproducing the three - dimensional mean velocities and dispersions of the stars in a stellar system
is already an improvement over all previous implementations of the schwarzschildtechnique , doing so is nevertheless a simplification of the problem .
the reason is that it implicitly assumes that the three velocity components are independent of each other , i.e. , it does not account for the fact that there is a velocity ellipsoid whose cross terms are , in the most general case , not identical to zero .
the most complete treatment would be to store a cube with entries for all possible combinations of @xmath83 , and do this at each spatial aperture where there is kinematical data available .
this implementation would be , however , expensive in terms of memory storage and , moreover , not absolutely necessary , simply because we are not interested in the entire probability cube .
instead , we only need probabilities in the cases when the model velocities are close to the observed ones . thus , in the framework of velocity histograms or full velocity cubes , and because of the discrete nature of the data ,
the large majority of the bins or entries would be filled with weights that do not affect the likelihood in equation ( [ eq.logl ] ) .
therefore , we adopt an approach in which , instead of storing velocity histograms or cubes , every time an orbit @xmath25 passes through an aperture @xmath14 with kinematical data , we add a gaussian contribution to @xmath77 .
this contribution is centered on the observed ( any - dimensional ) velocity and has a dispersion that reflects the measurement errors , and if desired , any amount of extra velocity smoothing .
thus , denoting the actually measured components of the particle s velocity in aperture @xmath14 as @xmath84 and their associated uncertainties as @xmath85 , with @xmath86 corresponding to @xmath87 , @xmath88 , and @xmath89 , the multiplicative contribution @xmath90 to the probability has the form @xmath91},\end{aligned}\ ] ] where @xmath92 is the component @xmath93 of the velocity of a test particle on orbit @xmath25 .
the quantity @xmath94 is the numerical smoothing assigned to velocity component @xmath93 .
whenever a particular component @xmath93 of the velocity is not available , we set @xmath95 .
finally , in order to account for the fact that we represent a continuous orbit by a discrete sequence of time steps , we weigh this gaussian factor by multiplying it by the timestep @xmath96 . therefore , for every orbit @xmath25 , and every time the orbit integration falls within an aperture , the probability is increased according to @xmath97 when the integration of orbit @xmath25 is done , the @xmath77 elements for all datapoints ( apertures ) are written to a file for later use by the algorithm that performs the maximization of the likelihood . in most practical applications one can set @xmath98 , since the error bars @xmath85 on the data already provide sufficient natural smoothing for numerical efficiency .
we do this throughout the rest of this paper . however , we note that there may be situations in which non - zero @xmath94 may be beneficial . for example , if the observational errors @xmath85 are much smaller than the velocity dispersions @xmath37 of the system .
it then takes very long integrations to beat down the shot noise in the orbital distributions @xmath77 .
addition of a numerical smoothing @xmath94 with @xmath99 can then speed up the calculations without affecting the accuracy of the results .
the approach of equations ( [ eq.weights ] ) and ( [ eq.pij ] ) assumes that the errors @xmath85 for the different datapoints are uncorrelated .
sometimes this is not true , as in the case of the proper motions of stars in the globular cluster @xmath22 cen , where relative rotation between the old photographic plates used in their derivation produce an artifact overall rotation of the cluster @xcite .
if problems like these can not be removed before modeling , a more complicated treatment than the one described here will be necessary .
the non - linear nature of the discrete problem addressed in this paper requires the use of a non - linear optimizer , and there is no guarantee of a unique optimum .
after experimentation with various optimization algorithms , we settled on the toms 500 conjugate gradient optimizer of @xcite .
this code uses the function value and gradient to optimize along successive vectors ( lines ) in the space of the orbital weights , choosing the optimization direction at every pass in a manner that attempts to minimize the number of such line minimizations needed ( see chapter 10 of @xcite for details on conjugate gradient methods ) . in our code
, we rely on the fact that the majority of orbits do not contribute to any particular linear constraint , or to the likelihood of any particular observational datum . in the notation of equation ( [ eq.1st.deriv ] ) , the linear constraints @xmath100 , and also the @xmath101 , are sparse matrices .
accordingly , the code to evaluate the gradient in equation ( [ eq.1st.deriv ] ) is written to store and evaluate only non - zero terms of @xmath100 and @xmath102 , reducing the computational burden by a factor of four or five . to evaluate convergence and estimate the proximity of our final likelihood maximum to the true ( possibly local ) maximum
, we plot the magnitude of the improvement of the likelihood @xmath103 as a function of the number of function evaluations @xmath46 .
see figure [ fig : mkfitin ] .
we find that @xmath103 is well represented by an exponential relation @xmath104 , where @xmath105 .
therefore , the future change in the likelihood if the optimizer were allowed to run forever would be @xmath106 , where @xmath107 is the current change in likelihood at step @xmath108 . in practice
, we terminate the optimization at @xmath109 , so that we expect to be within an additive factor of @xmath110 of the true likelihood maximum .
this typically occurs after a number @xmath111 of function evaluations .
the final accuracy is merely linear in the exponential coefficient @xmath112 , so that this accuracy estimate should be reasonably robust .
we ran a variety of tests in order to establish whether the algorithm has a tendency of finding local extrema as opposed to global ones .
in particular , for some of the test cases to be discussed later in
[ sec : tests ] , we started the iterative algorithm from different initial conditions , to verify that the solutions thus obtained were always in ( statistical ) agreement . also , as will be shown in
[ sec : tests ] , we find that the algorithm recovers the properties of known input models with reasonable accuracy . while this does not prove that the schwarzschildcode can not end up in a local maximum , at least it shows that the code does not end up in ( potential ) local maxima that are far from the correct solution . in practice
we usually start the maximization procedure from a homogeneous set of initial mass weights .
we also investigated whether the convergence to a solution can be sped up by starting the iterative process from initial conditions that may already be reasonably close to the final solution .
for example , we ran tests starting from a set of weights corresponding to a two - integral df of the form @xmath24 that already fit the light ( surface brightness ) profile followed by the input data .
such a solution is easily obtained as the nnls solution of a matrix equation .
we found that the same final answer was reached in essentially the same number of iterations .
in order to test the performance of our discrete schwarzschildcode , we generate sets of simulated data drawn from a known phase - space distribution function ( df ) .
unlike the case of using actual observations of a real stellar system , this approach offers the advantage of unambiguously knowing in advance the input properties underlying the data , which an optimally - working code should be able to `` recover '' .
it also provides flexibility by allowing the possibility of adapting the input data at will in order to test different aspects of the code ( [ sec : tests ] ) .
we discuss here the construction of various sets of pseudo - data and the properties of the underlying models .
our simulated input data are obtained from a set of @xmath24 dfs derived by @xcite , with the methodology for drawing n - body initial conditions from these dfs described in @xcite .
the models have a constant mass - to - light ratio @xmath113 , and have neither a central black hole or extended dark halo .
they provide good fits to available photometric and kinematic observations of the galaxy m32 over the radial range from @xmath114 arcsec . however , this property has no bearing on the present analysis .
we only use the fact that there is a known df , and not that this df resembles any realistic stellar system .
a two - integral @xmath24 df provides a useful test case ( see also @xcite , @xcite ) , and does not mean that the model results would be less valid for more general dfs . also , the use of a constant @xmath113 is motivated only to simplify the test environment .
central black holes ( e.g. , @xcite ) and extended dark halos ( e.g. , @xcite ) can be easily implemented in any schwarzschildcode .
the luminous mass density is assumed to be axisymmetric and is parameterized according to @xmath115^{\beta}[1+(m / c)^2]^{\gamma},\ ] ] with @xmath116 . here
, @xmath117 is the ( constant ) intrinsic axis ratio , related to the projected ( observed ) axis ratio @xmath118 via the inclination angle @xmath14 , @xmath119 .
the parameters in equation ( [ eq : light ] ) are set to @xmath120 , @xmath121 , @xmath122 , @xmath123 , @xmath124 , @xmath125 , and @xmath126 , with the @xmath127-band luminosity density @xmath128@xmath129pc@xmath130 , and @xmath131 the mass - to - light ratio in the @xmath127-band and in solar units .
the adopted distance is 0.7 mpc .
the models share the property of appearing the same in projection on the sky , but correspond to different intrinsic axis ratios as determined by the inclination angle @xmath14 .
the even part @xmath132 of the df @xmath24 is uniquely determined by the mass density @xmath133 ( e.g. , @xcite ) .
to specify the odd part @xmath134 of the df we follow @xcite and write @xmath135,\ ] ] with @xmath136 being the angular momentum of a circular orbit of energy @xmath9 in the equatorial plane ( @xmath137 ) , and the auxiliary function @xmath138 defined by @xmath139 & \mbox{$(u < 0)$}. \end{array}\right.\ ] ] the choice of the parameters @xmath140 and @xmath141 determines the degree of streaming of the dataset .
these free parameters can have values in the ranges @xmath142 and @xmath143 , with @xmath140 controlling the fraction of stars in the equatorial plane with clock - wise rotation , and @xmath141 controlling the behavior of the stellar streaming with orbital shape .
the family of functions @xmath138 is shown in figure 1 of @xcite
. combinations of @xmath144 that fit data for m32 are also discussed in that paper . here
we explore a variety of input datasets with different amounts of mean streaming and test the recovery of these properties by our discrete schwarzschildcode .
we generated 6 different datasets to test our discrete schwarzschildcode .
by dataset we mean a number of particle @xmath6 positions on the sky with corresponding proper motions @xmath145 and los velocities @xmath146 .
for two chosen inclinations on the sky , @xmath147 and @xmath148 , we produced three datasets resembling systems with varying degrees of rotation : a non - streaming system ( @xmath149 and @xmath150 ) , a maximally - streaming system ( @xmath151 and @xmath152 ) , and a third system with intermediate streaming ( @xmath151 and @xmath153 ) .
we label our different datasets as 90ns , 90is , and 90ms to indicate the non - streaming , intermediate - streaming , and maximally - streaming cases of @xmath147 , respectively .
similarly , for the @xmath148 case , we have the 55ns , 55is , and 55ms datasets .
the mass - to - light ratio used to generate the datasets is @xmath154 for @xmath147 and @xmath155 for @xmath148 , in units of @xmath156/@xmath157 .
although we examined the performance of our schwarzschildcode with tests that involve all of the six simulated datasets introduced above , we chose to use the 55is dataset to show most of our results .
figure [ fig : data55is ] shows some projections of the phase - space coordinates for the 55is dataset . in order to quantitatively
judge the performance of the three - integral schwarzschildcode , it is desirable to make a comparison between the properties of the input df ( i.e. , that from which the pseudo - data were obtained ) and those of the fitted df ( i.e. , that found as the solution to the fitting or minimization problem ) . it is important to note in this context that the direct output of our schwarzschildcode is not in the form of a proper df @xmath158 , but rather in the form of `` mass weights '' @xmath159 associated to each set of integrals of motion @xmath160 that uniquely define an orbit .
the relation between the df and the orbital mass weights is through a volume element dependent on the three integrals and an integration over the 3-dimensional space associated to the particular orbit ( see @xcite for a detailed discussion ) .
such a conversion can be done in schwarzschild codes ( e.g. , @xcite ) , but this is not necessary for the goals of the present paper .
we therefore limit ourselves to the comparison between the input and the fitted orbital mass weight distributions , which from now on we denote by @xmath161 and @xmath162 , respectively .
to validate the weights @xmath162 returned by the schwarzschildcode , we need to know the weights @xmath161 for the model df @xmath24 .
this is not a simple problem in the absence of an analytic expression for @xmath1 .
however , two related functions are more easily accessible .
the first is @xmath163 , defined as the projection of @xmath161 over the @xmath164 plane ( i.e. , integrated over @xmath1 ) . having the means of drawing n - body initial conditions from the df @xcite , we know that the energy and z - component of the angular momentum of each particle are given by @xmath165 and @xmath166 , respectively .
therefore , @xmath163 is easily obtained by binning a large n - body dataset @xmath167 in @xmath9 and @xmath79 .
the second related function that is easily accessible is @xmath168 , the distribution of mass weights for an @xmath24 model of axial ratio @xmath117 and a power - law density profile with logarithmic slope @xmath70 in a spherical kepler potential .
this function is calculated analytically in de bruijne et al .
( 1996 ; their equation ( 38 ) ) , and has the form @xmath169.\ ] ] here , @xmath70 is the logarithmic slope of the mass distribution and @xmath170 is a known function .
the spherical kepler potential is of course only an accurate approximation to our model at asymptotically large radii .
nonetheless , we can combine @xmath163 and @xmath168 to obtain a reasonable approximation for @xmath161 throughout the system , namely @xmath171 with @xmath172}{\int j_{\lambda}\left[l_z / l_{z,{\rm max}}(e),i_3\right]{\rm d } i_3}.\ ] ] for @xmath70 we take the slope of the mass distribution of equation ( [ eq : light ] ) at @xmath173 , the radius of the circular orbit of energy @xmath9 in the equatorial plane @xmath174 .
the function @xmath175 in equation ( [ eq : zetain ] ) is correct ( i.e. , reduces to @xmath176 ) when projected on the @xmath164 plane , and has approximately the correct distribution over @xmath1 at fixed @xmath177 .
in this way , we construct sets of orbital mass weights for each of our 6 simulated datasets described in [ sec : data ] .
using all the kinematic ( pseudo ) datasets and their corresponding input dfs described in [ sec : tools ] we now proceed to examine how accurately the discrete schwarzschildcode can recover the properties of the galaxy models used to generate the input datasets . by _ recovery _ we mean to determine how close or how far is the obtained solution from the known df , known mass - to - light ratio @xmath113 , and known inclination @xmath14 of the galaxy model corresponding to the simulated dataset that was provided as input to the code . at the same time
, we investigate the reliability of the uncertainties provided by the code on each of these properties . in the general case of modeling real observations of an actual stellar system , the true radial mass density profile is not known a priori and is typically described following some parameterization . since mass may not necessarily follow light , or may do so in some complicated way , different plausible mass models should be attempted , as well as allowing for possible variations of the mass - to - light ratio with position . for the purposes of the present tests , however , the underlying mass distribution is assumed to be perfectly known from equation ( [ eq : light ] ) , except for the value of @xmath113 .
therefore , the assumed parameterization for the mass distribution is only a 1-parameter family , and includes the `` correct '' distribution @xmath178 . in applications to real data , higher - parameter families may be necessary , and there is no guarantee that any member of the family would provide a good approximation to the true underlying distribution .
the results of our tests are examined via three different exercises , which can be performed on each of the 6 different input datasets , providing a good baseline to judge the performance of our discrete schwarzschildcode .
first , we explore the recovery of the internal orbital structure of the input dataset ( i.e. , the input df , or more specifically , the input mass weights @xmath179 ) by feeding the code with the correct inclination and mass - to - light ratio @xmath113 ( [ sec : getdf ] ) .
second , we fix the inclination to the correct value of the input dataset and study whether the code finds the minimum of the @xmath21 function at the correct value of @xmath113 ( [ sec : getml ] ) . and
third , we explore grids of schwarzschildmodels with different ( @xmath14,@xmath113 ) combinations to study how well these two properties are recovered when they are both assumed unknown ( [ sec : grids ] ) .
we run all the above exercises for various subsets of each of our 6 datasets in order to explore the results as a function of relevant observational variables , particularly the size of the input dataset and the type of kinematical constraints available ( i.e. , only los velocities , only proper motions , or the complete three - dimensional velocities ) .
this adds even more elements for a thorough assessment of the code s performance .
it also provides insights into the types of datasets that will be necessary to constrain @xmath14 or @xmath113 to some given uncertainty in realistic situations .
our schwarzschildcode has the capability of computing and storing , during a single orbit integration , the orbital properties for a series of different values of @xmath113 .
thus , during the construction of the orbit library , different values of @xmath113 are converted into a dimensionless factor @xmath180 that multiplies all our original velocities , thus with @xmath113 scaling simply as @xmath181 .
we stress that this allows us to explore several values of @xmath113 while computing only one orbit library . in our tests ,
we explore models for velocity factors in the range @xmath182 . given that our galaxy models with different inclinations have slightly different mass - to - light ratios @xmath131
, the use of this dimensionless representation also facilitates the visualization of the results in [ sec : getml ] and [ sec : grids ] .
the correct ( input ) value of @xmath113 is always at @xmath183 . at each of its different steps ,
the schwarzschildcode requires the user to specify several settings ( or dials ) that control a corresponding number of tasks and functions of the modeling procedure .
here we list the settings that we use for our standard run .
we concentrate on the settings that are new to the discrete implementation .
all other settings that are needed to fit a schwarzschildmodel ( e.g. , the resolution and limits of the polar grids used to compute the gravitational potential @xmath17 ; the required numerical accuracies in the fitting of the mass in the meridional plane and/or the projected plane of the sky ; etc . )
are identical to previous implementations of the code , so for those we refer to @xcite and @xcite . at the heart of the schwarzschildmethod
lies the generation of a comprehensive library of orbits that should be representative of all types of orbits possible in the gravitational potential under study .
this is achieved by adequately sampling the ranges of values that the three integrals of motion @xmath160 can acquire , each set of values uniquely determining one possible orbit . in this work
we build models using two libraries that only differ in their size .
most of our runs consist of the generation of initial conditions and libraries with @xmath184 orbits , obtained by sampling the available integral space with 20 energies @xmath9 , 14 angular momenta @xmath79 ( 7 positive and 7 negative ) , and 7 third integrals @xmath1 . in order to study the dependency of the results on the size of the orbit library ,
we also compute schwarzschildmodels using a much larger orbit library , with @xmath185 combinations of @xmath160 .
the energy @xmath9 is sampled via the corresponding radius @xmath186 of the circular orbit of that energy ( that with maximum angular momentum ) in the equatorial plane @xmath174 .
this radius is logarithmically sampled from a minimum value that we choose to be much smaller than the spatial resolution of the data , to a maximum value set much beyond the point at which most of the mass of the input distribution is actually encompassed .
since totally unconstrained by the data , therefore , the few first and last energy bins will mostly be of no interest ( i.e. , no mass gets assigned to them in the process of optimization ) .
the vertical component of the angular momentum , @xmath79 , is linearly sampled using the variable @xmath187 , where @xmath188 and @xmath189 is the angular momentum of the circular orbit with energy @xmath9 .
while orbits with both positive and negative @xmath79 are included in the library , the latter need not be individually integrated because they are simply obtained by reversing the velocity vector at each point along the orbit .
the third integral @xmath1 is sampled via an angle @xmath190 , where @xmath191 determines the position at which the `` thin tube '' orbit at the given @xmath177 touches its zero - velocity curve ( defined by the equation @xmath192 , where @xmath193 is the effective gravitational potential ; see @xcite for a detailed presentation ) .
finally , in order to help alleviate the discrete nature of the numerical orbit library , some extra radial smoothing of the orbits is performed by randomly generating a small variation to the energy and computing and storing the contribution to the probabilities from the `` new '' orbit with integrals @xmath194 .
it is possible to implement similar smoothing in @xmath79 and @xmath1 as well ( e.g. , @xcite ) , but we leave this for a future version of our code .
this energy smoothing is repeated , at each timestep , for 7 random @xmath195 values .
azimuthal averaging is also performed by randomly drawing 7 @xmath196 values at each timestep . smoothing in phase - space
is accomplished with the use of apertures ( [ sec : pij ] ) .
the size of the ( squared - shaped ) spatial apertures are defined as a fraction of @xmath80 , the distance to the center of the stellar system in the projected plane , and we set this fraction to 10% . in velocity space , and for most practical applications ,
the measurement errors @xmath85 themselves will provide sufficient `` natural '' smoothing for numerical purposes .
thus , we set the factors @xmath94 in equation [ eq.weights ] to zero for all our tests ( see also discussion in
[ sec : pij ] ) . in practice ,
the optimal value of @xmath94 will depend on the characteristics of the data ( particularly the size of the velocity errors ) as well as the stellar system under study .
when dealing with actual data , therefore , at least a few different values should be tried in order to explore their impact on the results .
additionally , the extra smoothing provided by @xmath94 can also be useful to explore the validity of the quoted errors in any given application .
the uncertainties @xmath85 in the los velocities and/or proper motions are in practice determined by the details of the observations and , since obtained by different techniques ( spectroscopy versus astrometry ) , are of different size in general .
furthermore , the uncertainties in the velocities tangential to the plane of the sky are affected by the uncertainty in the distance to the stellar system under study . here
, however , since we deal with simulated data , we assume kinematical data of nowadays typical good quality , and simply set all these errors to a moderate and arbitrary value of @xmath197@xmath198 .
the large majority of our tests were done on the simulated datasets as described in
[ sec : data ] _ without _ the addition of simulated observational errors ( i.e. , random gaussian deviates with dispersion @xmath85 ) .
this simplification was made early on in our project , based on the fact that the velocity errors should not matter much as long they are much smaller than the average one - dimensional velocity dispersion of the system under study .
however , we realized later that this does induce a slight bias in our estimated mass - to - light ratios .
our typical simulated datasets have dispersions of 48.4@xmath198and 46.3@xmath198 , for the 55is and 90is cases , respectively .
therefore , by not adding any random velocity errors , the one - dimensional velocity dispersion of the pseudo - data that we actually analyzed is too small by a factor @xmath199 . as a consequence of the virial theorem , it follows that we should expect to infer a mass - to - light ratio that is too small by a factor of @xmath200 , corresponding to about 2.2% and 2.4% for the 55is and 90is datasets , respectively . instead of rerunning all our calculations , which would have been computationally expensive , we therefore simply corrected for this bias _
so when studying the recovery of the mass - to - light ratio in [ sec : getml ] and [ sec : grids ] , instead of comparing the inferred values to the value @xmath131 of the input model , we compare to the slightly smaller @xmath201 .
this quantity is @xmath202 for @xmath203 and @xmath204 for @xmath205 .
the sizes of currently existing kinematic datasets of discrete nature range from a few hundred datapoints ( red giants in local group dwarf galaxies , planetary nebulae in the outskirts of giant ellipticals ) to a few thousands ( stars in galactic globular clusters , systems of globular clusters around giant ellipticals ) .
for our standard tests we adopt datasets with 1000 kinematic observational points , although we also study the consequences of studying datasets with sizes ranging from 100 to 2000 datapoints . in these tests ,
the small-@xmath46 datasets are subsets of the largest dataset ( @xmath206 ) , which means that there will be some correlation between the results of experiments done as a function of the number of available observations .
this approach , we note , is of no substantial difference than having all the datasets of different @xmath46 but within the same simulation to be completely disjoint . the progression with @xmath46
should still follow the expected @xmath207 statistical - convergence behavior ( see fig . [
fig : errors_ml ] and [ sec : getml ] ) .
the generation of one of our smaller @xmath208 orbit libraries , simultaneously storing discrete probabilities for a set of 1000 observational points with both los velocities and proper motions , takes 2.5 hours on a 3.6 ghz , pentium 4 , 64-bit cpu with 2 gb memory .
an additional 0.5 hours are needed to find the maximum likelihood fit to the data . in practice
, these steps must be iterated over a grid of gravitational potential parameterizations . in order to determine whether the best - fitting solution obtained by the discrete schwarzschildcode actually resembles the properties of the input data , we start by making detailed comparisons between the input and fitted dfs . to do this
, we feed the code with the correct inclination and mass - to - light ratio @xmath113 used to generate the input datasets , and compare the fitted mass weights @xmath209 to those corresponding to the input data , @xmath161 , approximated using equation ( [ eq : zetain ] ) .
we use datasets with 1000 los velocities and proper motions , and present results for both the small and big orbit libraries detailed in [ sec : settings ] .
the comparison is best achieved via the analysis of corresponding one- and two - dimensional projections of the cubes of mass weights @xmath162 and @xmath161 , obtained by integrating over two and one of the integrals of motion , respectively ( figs .
[ fig:1dplots ] to [ fig:2dplot55is ] ) . also , we make comparisons of two - dimensional @xmath210 slices of both cubes at selected values of the energy ( fig . [
fig : ebins55is ] ) . for al of these projections
we quantify the agreement between fits and input data by computing the rms and median absolute deviation of the quantity @xmath211 , i.e. , the difference between fit and input mass weights normalized by the input mass weights .
these statistics are listed for schwarzschildmodels run on all our input datasets in table 1 .
since the rms can be biased disproportionately by a small number of large outliers , in our discussion below we use preferentially the median absolute residual .
figure [ fig:1dplots ] shows , for the 55is case , the integrated mass weights as a function of each of the three integrals of motion , for both the input dataset and the discrete schwarzschildfit . inside the region actually constrained by kinematic data ( containing 99.83% of the total mass ) , the mean absolute deviations between the fitted and input distributions of mass weights are 3% , 16% , and 18% , for the integrated distributions as a function of @xmath9 , @xmath11 , and @xmath1 , respectively . as listed in table 1 , similar numbers are obtained for the other 5 simulated datasets , with the agreement between both distributions as a function of energy always better than 5% . as a function of @xmath79
, the largest disagreement actually corresponds to the one shown in figure [ fig:1dplots ] , the 55is case .
it goes down to 7% for our case of closest agreement , the case labeled 55ns .
the net rotation inherent to the 55is dataset ( reflected in the middle panel by all the mass weights with positive @xmath79 being larger than those with negative @xmath79 ) is clearly reproduced by the schwarzschildfit . as a function of the third integral ,
the median absolute deviation varies from 16% for the 55ns case to up to 25% for the 90ns case .
note that , since we are showing orbital mass weights instead of the actual df , the @xmath1 distributions are not expected to be constant over @xmath1 , even though the input df underlying all simulated datasets is of the form @xmath24 .
next , integrating only over @xmath1 , we show in figures [ fig:2dplot55ns ] and [ fig:2dplot55is ] the agreement between the fitted and input sets of mass weights as a function of @xmath9 and @xmath79 , for the 55ns and 55is cases , respectively .
the upper panels of these figures show the results of the schwarzschildfit ( @xmath209 ) and the lower panels the original input distributions ( @xmath175 ) .
the left - hand panels show the results for a @xmath160 library of @xmath212 orbits , 8 times larger ( i.e. , finer ) than that of the right - hand panels , which correspond to our standard case of @xmath208 orbits . only the energy range constrained by the respective sets of data is shown .
black corresponds to zero weight , and the white ( brightest ) color in each pair of panels ( fit and model , or upper and lower ) has been assigned to the maximum orbital weight among the two panels , so that the comparison between fits and models is made using the same color scale . both figures
[ fig:2dplot55ns ] and [ fig:2dplot55is ] show that the main features of the input @xmath164 distributions of mass weights are well reproduced by the 3-integral schwarzschildfits . in particular , the mean streaming properties of both datasets are satisfactorily recovered . in figure
[ fig:2dplot55ns ] , the two prominent phase - space blobs occupying symmetrical locations on the negative and positive sides of the @xmath79-axis correspond well with the non - rotating overall nature of the 55ns dataset .
moreover , this is recovered by both the models with standard and large orbit libraries ( right- versus left - hand panels ) .
similarly , in figure [ fig:2dplot55is ] , the single phase - space blob at positive @xmath79 with a pronounced elongation towards negative @xmath79 ( in light blue and blue ) , indicative of the rotating nature of the 55is case , is reproduced by the schwarzschildfit as well .
the median absolute deviations between the fitted and input @xmath164 distributions , always restricted to the energy range constrained by the data , are 14% and 19% for the 55ns and 55is cases , respectively ( table 1 ) . in figure
[ fig : ebins55is ] we show the 3-dimensional distributions of mass weights of our 55is case in the form of a series of @xmath210 planes at different energies . here again , the upper panels show the results of the discrete schwarzschildfit ( @xmath209 ) , the lower panels the distribution of mass weights corresponding to the input data ( @xmath179 ) , and the color scale is set up in the same way as in the @xmath164 figures . as the energy @xmath9 is sampled via the radius @xmath186 of the circular orbit ( its value in arcmin indicated at the top of each pair of panels )
, this series of planes shows the variation of the @xmath210 distribution with increasing distance from the center of the galaxy .
the fraction of the total mass at each energy slice is given as a percentage at the bottom of each panel .
the bottom panels of figure [ fig : ebins55is ] indicate that , in the inner regions ( inside 0.2 arcmin ) , most of the mass in the 55is dataset is concentrated in orbits with @xmath79 near zero .
the corresponding upper panels show that the schwarzschildfit recovers this @xmath213 component , but it distributes more weight than the input model into orbits with positive @xmath79 .
these inner regions , nevertheless , have a relatively low mass content in comparison with regions at larger radii . as the radius increases
, the @xmath214 region of phase - space gets progressively depleted of stars in favor of orbits with high @xmath79 .
this transition is reasonably well reproduced by the schwarzschildsolution , and the agreement between fit and input data becomes better at large radii , at which point most of the mass at each energy is concentrated in orbits of high @xmath79 .
note also that a common characteristic of figures [ fig:2dplot55ns ] , [ fig:2dplot55is ] , and [ fig : ebins55is ] is that schwarzschildfits typically present mass weight distributions that appear broader ( more extended ) and less peaked than the corresponding distributions displayed by the pseudo - data .
the effect is most obvious among the right - most panels of figure [ fig : ebins55is ] , where one can see that the @xmath210 mass - weight distributions of the input data ( lower panels ) have higher peaks and overall sharper features than the corresponding fitted distributions ( upper panels ) .
this is an expected effect and is due to the combined smoothing of the fitted distribution introduced both by the ( necessary ) use of velocity apertures for the computation of likelihoods ( see [ sec : pij ] ) , and by the regularization constraints imposed in order to enforce smoothness in phase space .
while the first smoothing is particular to our discrete implementation , the second is a well - known procedure common to most schwarzschildcodes .
models without regularization tend to be unrealistically noisy @xcite and unreliable for parameter estimation @xcite .
thus , although we choose to plot the input distribution of mass weights as they actually are , the most fair of comparisons would be one in which the schwarzschildfit is compared with a smoothed version of the original mass weight distribution describing the input data .
we explored this by convolving the input distribution of mass weights with a ( circular ) gaussian kernel , and then computing the same statistics shown in table 1 ( but this time using the smoothed version of the input distribution ) for different widths of the gaussian kernel .
we have verified that indeed it is possible to find a kernel width for which the agreement between fit and input data is best , improving both the rms and mean absolute deviations of table 1 by factors between 1.2 and 1.5 .
finally , we also note that the comparison in figure [ fig : ebins55is ] might be affected by the accuracy of the approximation in equation ( [ eq : zetain ] ) , which means that the values in table 1 are actually upper limits to the true accuracy of the schwarzschildfits . from these tests
we conclude that our discrete schwarzschildcode can successfully recover the original df inside the region constrained by the kinematic data , at least for the case in which the inclination and mass - to - light ratio are assumed known . for a large range of potential applications of a schwarzschildcode , such as investigating dark matter halos in galaxies , the most important property
that one is interested in measuring with confidence is the mass - to - light ratio . in the present tests ,
this quantity is a scalar , @xmath113 , although in more general applications it could be a function of radius . in this section
we study in detail the capacity of our code to infer the correct @xmath113 when the inclination of the system is assumed known .
tests were performed for a number of input datasets in order to investigate the dependence of the results on key observational variables such as the number of kinematic measurements and the type of kinematic constraints available ( i.e. , only - los velocities , only proper motions , as well as both los velocities and proper motions ) .
all models in this section were computed using our small orbit library , with @xmath208 combinations of @xmath160 .
the results of these experiments are summarized in figures [ fig : mlparabn]-[fig : errors_ml ] . for the 90is and 55is cases and using full 3-dimensional velocity information , figure [ fig : mlparabn ]
shows the @xmath21 parabolae obtained when applying the discrete schwarzschildcode with a number of @xmath113 values , distributed around the correct one ( @xmath215 ) , for datasets of varying sizes .
the quantity @xmath216 along the ordinate denotes the velocity scaling ; @xmath217 corresponding to a schwarzschildmodel with the input value @xmath215 , defined in [ sec : settings ] .
the zero point of the vertical axis ( both in figures [ fig : mlparabn ] and [ fig : mlparab2 ] ) is arbitrary , but the difference @xmath72 between points on the same curve has its usual statistical meaning , and indeed we compute the ( random ) uncertainties on the determination of @xmath113 directly from them .
figure [ fig : mlparabn ] shows that the difference @xmath72 between points on the same curve becomes larger ( the parabolae become narrower ) as the number of available kinematic measurements increases .
the determination of the best - fit @xmath113 also depends on the type of available kinematic measurements .
this is illustrated in figure [ fig : mlparab2 ] , where we plot the @xmath21 parabolae obtained when considering only los velocities , only proper - motions , or the full 3-dimensional velocity information .
all cases are for the 55is dataset with 1000 kinematic measurements . in this case , the @xmath72 parabolae become narrower as the number of available velocity components increases .
furthermore , the statistical errors are generally smaller for larger datasets , as well as when more velocity components are available .
this is shown in figure [ fig : errors_ml ] , where we plot the behavior of the best - fit @xmath113 and its uncertainties as a function of @xmath218 , where @xmath46 is the number of datapoints .
the uncertainties @xmath219 displayed in the upper panel of figure [ fig : errors_ml ] represent the @xmath75 intervals around the minimum of the parabolae in figures [ fig : mlparabn ] and [ fig : mlparab2 ] , and are defined as half the distance between the points on the curve where @xmath220 with respect to the minimum .
the statistical errors scale roughly as @xmath207 over an interval of 1.3 dex in @xmath221 .
also , the errors in the best - fit @xmath113 associated to datasets with only proper - motions ( triangles ) are smaller than those associated to only - los datasets ( open circles ) for any value of @xmath46 . in other words ,
our discrete schwarzschildcode satisfies the fundamental statistical expectation that it should become easier for the method to distinguish between models with different @xmath113 when the amount of observational information is larger . in the case of datasets with the full 3-d velocity information
, the 55is uncertainties do not quite seem to follow the @xmath207 behavior expected from statistics .
we attribute this to our tests having reached a fundamental floor due to the discrete nature of the models , a limit that can not be overcome by increasing the number @xmath46 of available measurements .
this can cause an apparent flattening with respect to the regular @xmath207 behavior at large @xmath46 . to test the robustness of the errors estimated as above , we performed the following exercise . selecting 10 different ( disjoint ) realizations of the n - body data ( for the 55is case with 1000 measurements of only line - of - sight velocities , the case most often found in practice ) , we repeated the exercise of figures [ fig : mlparabn ] and [ fig : mlparab2 ] and computed discrete schwarzschildmodels for a set of different @xmath113 values distributed around the correct input one .
this was done using our small orbit library .
we obtained an average best - fit @xmath113 of 2.46 ( less than @xmath75 away from the input value , @xmath222 ) , with an rms of 0.074 ( corresponding to about 3% ) .
when computing the statistical uncertainties using the @xmath223 parabolae as described above , the average @xmath75 error in the best - fit @xmath113 of the set of experiments turns out to be 0.204 , equivalent to a fractional error of 8% .
this is a factor of 2.5 larger than the scatter in the results from multiple independent realizations of the pseudo - data .
this gap is smaller when additional information about the individual kinematics of the tracers is available . indeed , repeating the above exercise for the same datasets but now using two - dimensional proper - motions instead of only line - of - sight velocities , the average error in @xmath113 computed from our @xmath21 parabolae is 0.112 , a factor of 1.7 larger than the scatter of the best - fit values , which was 0.067 .
therefore , we conclude that our error estimation using @xmath72 is conservative . despite the smaller statistical errors for the case with proper - motions alone
, the bottom panel of figure [ fig : errors_ml ] indicates that the best - fit @xmath113 is closer to the real value , @xmath215 , for the case with only - los velocities .
while the best - fit @xmath113 from datasets with only los velocities are well within @xmath75 of @xmath215 for any @xmath46 , this is not the case for the datasets with only proper - motions , with best - fit @xmath113 values that are @xmath224 away from @xmath215 .
still , the formal best - fit @xmath113 for the case of full 3-d velocities ( thick squares ) is on average within @xmath225 of the real value , @xmath215 , corresponding to better than @xmath226% accuracy .
one contribution to the small systematic bias in @xmath113 may come from the fundamental nature of inverse problems in general ( of which schwarzschildmodeling is an example ) , namely , that there may not necessarily be a unique solution : it may be possible to change the mass profile and the df without appreciably changing the model predictions .
if such is the case and there are multiple solutions , we do not necessarily expect a flat @xmath223 profile ( i.e. , with a number of equally acceptable solutions containing the correct one ) , most likely because of numerical noise and discretization effects . while we can not rule this out
, our results do show that this probably does not affect the recovered mass - to - light ratio at more than the @xmath227% level ( based on figure [ fig : errors_ml ] , built with models using our smaller orbit library ) . unless superb data are available , random uncertainties are likely larger than such systematic errors .
currently , the only exception to this are some galactic globular clusters , for which thousands of proper motions are being measured .
however , such systems are often closer to spherical than galaxies , and hence one expects any theoretical degeneracies to be smaller .
alternatively , numerical noise in the orbit library may be the cause of this systematic bias in @xmath113 seen in the bottom panel of figure [ fig : errors_ml ] .
numerical noise may be reduced in part by the use of larger orbit libraries .
indeed , we show in [ sec : grids ] below that a substantially larger orbit library tends to produce more accurate results overall .
the likelihood ratio statistic @xmath21 in figures [ fig : mlparabn ] and [ fig : mlparab2 ] allows us to find the best - fit model parameters and their confidence intervals .
however , it does not shed light on the question whether the best - fit model is actually statistically consistent with the data .
the likelihood @xmath68 of the best - fit model also can not be used for this purpose .
there is no theorem of mathematics that states what the value of @xmath68 should be for a statistically acceptable model , given that the underlying velocity distributions from which the particles are drawn are not known a priori ( and are not generally gaussian ) .
nonetheless , many other statistics can be defined to address this issue once the best - fit model has been found .
for example , the velocity moments of the best - fit model can be calculated ( as a function of position on the sky ) , and statistics can be defined that assess whether these moments are consistent with the observed data .
alternatively , one can draw random realizations of the data from the best - fit model and use a kolmogorov - smirnov test to assess whether the data and the realization are consistent with being drawn from the same underlying distribution .
we have explored a subset of these approaches and these suggested that the best - fit models are indeed statistically consistent with the pseudo - data they were designed to fit . in general
neither the mass - to - light ratio nor the inclination of a stellar system under study are known in advance , and thus one has to explore models with several combinations of both parameters in a search for those values that provide the best fit to the data . in this section we present and discuss the results of running the discrete schwarzschildcode on grids of @xmath228 values to study whether the correct input combination is recovered . as in [ sec : getml ] , we perform tests on datasets with different types of kinematic constraints ( los velocities and/or proper motions ) . the results of tests are presented in figures [ fig:55is_los_mu ] and [ fig : grid_55_90 ] .
they show @xmath72 contours that result when computing discrete schwarzschildmodels on a grid of @xmath228 values , including the correct input combination , for a variety of input datasets of the 55is and 90is cases .
the goodness - of - fit parameter @xmath72 shown in these plots is obtained by first rebinning with a much finer grid the @xmath228 space explored by the models actually calculated ( indicated by small dots ) , and then computing the values on this new grid via interpolation between the nearest calculated models .
we then determine the minimum on the finer grid ( whose location is indicated by the star ) and subtract it from all grid points to obtain the @xmath72 parameter , for which contours are shown .
as in the case of figures [ fig : mlparabn ] and [ fig : mlparab2 ] , the mass - to - light ratio is parameterized by the dimensionless velocity scaling @xmath229 , so that the input value corresponds to @xmath230 .
we start by showing in figure [ fig:55is_los_mu ] the results of running grids of models for input datasets composed of only - los velocities and only proper motions , in both cases for the 55is case with 1000 observational datapoints , and using our small orbit library with @xmath208 combinations of @xmath160 .
overall , and in agreement with the results of figure [ fig : mlparab2 ] discussed in
[ sec : getml ] , the @xmath72 contours indicate that proper motions ( bottom panel ) better constrain the best - fit @xmath228 combination than a dataset with only - los velocities ( upper panel ) .
the @xmath231 uncertainties ( thick contours ) obtained from the only - los dataset are twice as large than those from the proper motions alone ( 31% and 16% , respectively ) . the input mass - to - light ratio @xmath215 is adequately recovered by both datasets ( to within the @xmath75 confidence region ) .
the best - fit inclination , however , is offset from the actual input value @xmath148 for both datasets , although somewhat closer to the correct value in the case of proper motions only .
the @xmath231 uncertainties in the best - fit inclination are @xmath232 and @xmath233 for the only proper motions and only los cases , respectively .
difficulties in constraining the inclination using schwarzschildmodeling of stellar kinematics have been encountered in the past .
a good recent example is that of @xcite who , based on integrated stellar los velocity profiles and ionized gas observations of the e4 galaxy ngc 2974 , carried out a study analogous to the present one by constructing simulated observations of this galaxy , which they feed to their `` continuous '' ( as opposed to discrete ) schwarzschildcode in order to study the recovery of the input mass - to - light ratio and inclination .
they find that even with artificially perfect input kinematics the inclination is very poorly constrained . the same conclusion is reached when attempting to fit the actually observed los velocity profiles with schwarzschildmodels , so stellar los velocity profiles provide weak constraints on the inclination of this system , a statement they are confident about because the actual inclination of ngc 2974 is known from observations of its extended disc of neutral and ionized gas in rapid rotation .
while one could expect that the availability of proper motion measurements in addition to los velocities would enhance the ability of the models to obtain useful constraints on the inclination of a stellar system in general , the reality is that the current state - of - the - art of schwarzschildmodeling does not have a definitive answer on this issue yet .
as recent studies of the kinematics of stars in globular clusters seem to indicate , the chances of success are highly dependable on the quality and quantity of available data on the system under study ( compare , for example , the results of @xcite and @xcite regarding the best - fit inclinations of @xmath22 cen and m15 , respectively ) .
there are at least two factors that may contribute to the difficulty in recovering the inclination from stellar kinematics : degeneracies inherent to schwarzschildmodels , and numerical noise .
first , there is no guarantee that inclinations other than the correct one must fit the data worse .
indeed , in their modeling of high signal - to - noise integral - field data of ngc 2974 , @xcite already observe that the differences between schwarzschildmodels with different inclinations are smaller than the differences between the best - fitting model and the data , which they interpret as indication of a fundamental degeneracy in the recovery of the inclination with three - integral models .
numerical noise , on the other hand , is a consequence of schwarzschildmodels being in the end only discrete representations of a smooth , continuous distribution of possible orbits , and it could be argued that this discreteness might have a more negative effect for high inclinations .
for example , even a simple and smooth circular orbit presents cusps or discontinuities when viewed close to edge - on .
the turning points of such an orbit may get smoothed out differently for different inclinations .
the issue of degeneracy , nevertheless , can be avoided in those cases where the inclination is known to be uniquely determined by the data .
this is the case , e.g. , in the situations where the following conditions are met : ( 1 ) the kinematical dataset consists of proper motion measurements and los velocities , ( 2 ) the stellar system is reasonably close to axisymmetric , and ( 3 ) there exists an independent measurement of the distance @xmath234 to the system .
as first used in practice by @xcite , the inclination then follows directly from the following relationship between the mean los velocity ( in units of @xmath198 ) and the mean proper motion along the short axis ( in units of masyr@xmath235 ) , @xmath236 where @xmath234 is the distance in kpc , and the brackets denote an integration along the line - of - sight .
this relation is true at each projected position @xmath6 in any axisymmetric system , and has been successfully applied to the galactic globular clusters @xmath22 cen and m15 @xcite . here , in order to explore the applicability of this simple relationship , we take advantage of our a priori knowledge of the correct inclination for our simulated datasets , and study the circumstances under which the use of equation ( [ eq : inclination ] ) provides an accurate result . unlike the case of integrated light measurements ( where @xmath237 is simply the average of the losvd at
any given projected position on the sky ) , in the context of discrete datasets neither @xmath237 nor @xmath238 are quantities that can be rigorously obtained from the data at any given @xmath6 .
both quantities may , nevertheless , be approximated by averaging a number of kinematic measurements that fall within one or more apertures of a given size around projected positions @xmath6 . following this , we applied equation ( [ eq : inclination ] ) to a series of subsets of our 6 simulated datasets with varying number of kinematic measurements , and verified that indeed the correct inclination is reproduced provided : ( a ) the system is rotating ( otherwise , while the relation is still valid , both averages are nearly zero and hence the inclination is not really constrained ) ; ( b ) most of the datapoints are not located close to the minor axis ( where rotation velocities are too small ) ; and ( c ) the averages are computed from a sufficiently large number of kinematical measurements ( so that the error in @xmath239 is not too large )
. these are conditions that are certainly fulfilled by datasets on some galactic globular clusters , currently the only class of stellar system for which there are 3-dimensional kinematic information available .
therefore , in those cases , equation ( [ eq : inclination ] ) can be safely applied .
the schwarzschildmodeling can then concentrate on recovering the more interesting properties such as the orbital structure and mass - to - light ratios , which we have shown are successfully recovered when the inclination is assumed known . to better understand the problem of numerical noise , we explored the dependence of the results on the size of the orbit library used to construct the schwarzschildmodels .
we did this for cases with 1000 datapoints with complete three - dimensional velocities , so that because of equation ( [ eq : inclination ] ) we know that there is no theoretical degeneracy in inclination . figure [ fig : grid_55_90 ] shows the @xmath72 contours resulting from fits of schwarzschildmodels using our standard library of @xmath208 orbits ( upper panels ; same library size as in figure [ fig:55is_los_mu ] ) in comparison with fits that use a library 8 times larger , i.e. , one with @xmath212 orbits ( lower panels ) .
we show results for the 55is ( left - hand panels ) and 90is ( right - hand panels ) cases . in all four panels of figure [ fig : grid_55_90 ] ,
the best - fit mass - to - light ratio is always within @xmath75 of the input value @xmath215 , with the exception of the 55is case with the bigger library ( lower left ) , where they agree at the @xmath225 level .
the size of the confidence regions on the mass - to - light ratio does not change significantly when the orbit library is increased in size .
therefore , we conclude that libraries of @xmath208 orbits are large enough to properly constrain the mass - to - light ratio ( provided that one uses regularization as we do here ; see @xcite ) .
this provides further justification for our use of this library size in section [ sec : getml ] .
the top left panel in figure [ fig : grid_55_90 ] is directly comparable to the two panels of figure [ fig:55is_los_mu ] , but now with three components of velocities observed , instead of just one or two , respectively .
consistent with the results in figures [ fig : errors_ml ] and [ fig:55is_los_mu ] , we see that the addition of an extra component of velocity decreases the size of the confidence regions .
more interestingly , a secondary minimum in @xmath223 appears close to the @xmath228 values for the correct input model .
this suggests that indeed all three components of velocity may be necessary to uniquely constrain the inclination of an axisymmetric stellar system .
the bottom left panel shows the effect of increasing the orbit library size .
there is now only a single minimum , centered at an inclination that agrees with the input value at the @xmath240 level .
the right panels in figure [ fig : grid_55_90 ] show the situation for the 90is case . with the small library ( top right ) ,
the best - fit inclination is at @xmath241 , substantially far from the input value . when the orbit library size is increased ( lower right ) , the best - fit shifts to @xmath242 .
this is only @xmath243 from the correct input value , which may well be acceptable for many realistic applications . on the other hand , the best fit and the input value are inconsistent at the many sigma level , which is certainly reason for some concern .
a possible cause for this is that the turning points of orbits in edge - on systems have very sharp edges in projection .
therefore , larger grid sizes than we have used may be necessary to correctly represent them in all the necessary detail . however , we have not explored this further for two reasons .
first , information on all three velocity components may be necessary to be able to uniquely constrain the inclination .
if that is available , then use of equation ( [ eq : inclination ] ) will be more accurate and efficient than use of schwarzschild modeling .
second , in practice one is generally much more interested in the mass distribution than in the inclination .
figure [ fig : grid_55_90 ] shows that the mass - to - light ratio is correctly recovered , even when the inclination is systematically biased . in conclusion ,
our tests demonstrate that the recovery of the most important properties of the system ( its orbital structure and the mass - to - light ratio ) by our discrete schwarzschildmodels is robust .
correct recovery of the inclination appears to be the most complicated aspect of the modeling .
sufficient observational data must be available and a large enough orbit library must be used .
our code can then adequately recover the inclination of sufficiently inclined systems .
however , for edge - on systems there remains a systematic inclination bias of @xmath244 that we have been unable to resolve .
this is the primary shortcoming of our new approach that was unearthed by the pseudo - data tests that we have presented .
this may be a generic property of schwarzschild codes , since other authors have also reported difficulties in recovering inclinations .
either way , this is not believed to be a significant limitation for most potential practical applications of our code .
discrete kinematic datasets , composed of velocities of individual tracers ( e.g. , red giants , planetary nebulae , globular clusters , galaxies , etc . ) , are routinely being assembled for a variety of stellar systems of all scales ( [ sec.intro ] ) .
these include not only los - velocity surveys .
high - quality proper - motion databases already exist for galactic globular clusters , and future facilities hold the promise of providing the same for stars in the nearest galaxies .
however , the most sophisticated tools typically being used in the modeling of these observations were actually developed for the analysis of kinematic data in the form of losvds , a rather different type of velocity information than the case of the velocities of kinematic tracers on a one - by - one basis . as a consequence ,
the information content of any particular dataset of a discrete nature is likely not being fully exploited .
we thus have developed a specific tool for the modeling of discrete datasets , which we have presented in this paper along with detailed tests of its performance based on the modeling of simulated data .
the new tool consists of a schwarzschildorbit - superposition code that , adapted from the implementation of @xcite , can handle any number of ( one- , two- , or three - dimensional ) velocities of individual kinematic tracers without relying on any binning of the data . under the only assumptions that the system is in steady - state equilibrium ( i.e.
, the gravitational potential is not changing in time ) and may be well approximated as axisymmetric , the code finds the distribution function ( a function of the three integrals of motion @xmath9 , @xmath79 , and @xmath1 ) that best reproduces the observations ( the velocities of the tracers as well as the overall light distribution ) in a given potential .
the fact that the distribution function is free to have any dependence on the three integrals of motion allows for a very general description of the orbital structure , thus avoiding common restrictive assumptions about the degree of ( an)isotropy of the orbits .
unlike previous implementations of the schwarzschildtechnique , we cast the problem of finding the best superposition of orbits using a probabilistic approach , i.e. , by building a likelihood function representing the probability that the entire set of measurements would have been observed assuming a particular form for the gravitational potential ( [ sec : logl ] ) . in this case , and in contrast with the old continuous versions , the dependence of the likelihood function on the orbital weights is non - linear , and the optimization problem can not be reduced to a linear matrix equation .
instead , it becomes a problem of the maximization of a likelihood with respect to the set of weights associated to all possible combinations of the integrals @xmath160 that comprise the orbit library ( [ sec : logl ] ) , and which accounts for the observed positions and ( any - dimensional ) velocities of all particles in the dataset , including their uncertainties ( [ sec : pij ] ) .
after extensive testing , a conjugate gradient algorithm was found to converge satisfactorily to the correct solution and was adopted for the remaining tests of the code s overall performance ( [ sec.mkfitin ] ) . in order to assess the reliability of our discrete schwarzschildcode
, we applied it to several sets of simulated data , i.e. , artificially generated kinematic observations obtained from a model of an axisymmetric galaxy of which the orbital structure , mass distribution , and inclination are known in advance .
pseudo - datasets were generated from a two - integral phase - space distribution function with varying degrees of overall rotation , types of velocity information ( only - los , only proper motions , and both ) , total number of particles , and for two different inclinations on the plane of the sky ( [ sec : data ] ) . using the various simulated datasets , we studied the recovery of the input orbital structure or df , mass - to - light ratio , and inclination . for the purposes of these tests
, we assumed complete knowledge of the radial profile of the underlying mass distribution and a mass - to - light ratio that remains constant as a function of radius .
these restrictions are easily ( and must be ) lifted when modeling data on real systems , in which case one needs to explore a range of plausible underlying potentials and allow for variations of the mass - to - light ratio to properly account for the possibility of central black holes and dark halos . inside the region
constrained by data , we find that the distribution function ( represented by the corresponding distributions of orbital mass weights ) and streaming characteristics of the input datasets are satisfactorily recovered by the schwarzschildfits when the correct inclination and mass - to - light ratio are known ( figs .
[ fig:1dplots ] to [ fig : ebins55is ] ) .
as measured by the mean absolute deviations between the integrated weight distributions , the agreement between the fitted and the input orbital weight distributions as a function of @xmath9 , @xmath79 , and @xmath1 is typically of the order of 3% , 10% , and 20% , respectively ( the numbers for our worst case being 5% , 16% , and 25% ) . when eliminating the dependence on @xmath1 , the agreement between the fitted and input @xmath164 distributions is of the order of 15% , with the net rotational behavior of the input datasets cleanly recovered ( figs .
[ fig:2dplot55ns ] and [ fig:2dplot55is ] ) .
thus , we conclude that the discrete schwarzschildcode can successfully recover the orbital structure of the system under study .
assuming that the inclination of the system on the plane of the sky is known , we quantified the recovery of the input mass - to - light ratio as a function of the size of the input dataset ( fig .
[ fig : mlparabn ] ) and of the type of kinematic information available ( fig .
[ fig : mlparab2 ] ) .
we studied both the best - fit value as well as the uncertainty in its determination ( fig .
[ fig : errors_ml ] ) .
the statistical expectation of better results when the amount of observational information is larger ( either regarding the number of datapoints or the number of velocity components ) is clearly reproduced by our discrete schwarzschildmodels .
for the smallest datasets used in our testing ( @xmath245 ) , and regardless of whether using only - los velocities , only proper motions , or both , the best - fit mass - to - light ratio is within 5 - 10% of the input value , with formal @xmath75 uncertainties of the order of 15% . when increasing either the number of available measurements or the number of measured velocity components , the mass - to - light ratio is always recovered to better than @xmath246 accuracy , with the corresponding random ( @xmath75 ) uncertainties in the range of 5 - 10% .
the discrete schwarzschildcode , therefore , recovers the mass - to - light ratio of the input datasets to satisfactory levels of accuracy .
the recovery of both the mass - to - light ratio and inclination when neither of these quantities are known in advance ( as is usually the case with real observations ) was studied using a grid of discrete schwarzschildmodels , exploring also the dependence on the type of velocity components available ( fig .
[ fig:55is_los_mu ] ) .
we find that the mass - to - light ratio was again successfully recovered , but the best - fit inclination was not identified correctly using small orbit libraries .
we found that this was remedied by better sampling the available @xmath160 integral space using a larger orbit library ( fig .
[ fig : grid_55_90 ] ) . for our input datasets with @xmath148 ,
the best - fit inclination obtained by our models with a large orbit library is @xmath247 , while for input datasets with @xmath147 we obtain a best - fit model with @xmath242 .
given the known difficulty of schwarzschildmodels in general for determining the inclination of stellar systems , and considering the low relative importance of this parameter compared to other properties such as the orbital structure and the mass - to - light ratio , we regard this small disagreement for the high inclination datasets as acceptable . in summary , we have shown that our new schwarzschildcode , designed to adequately handle modern datasets composed of discrete measurements of kinematic tracers , doing this without any loss of information due to data binning or restrictive assumptions on the distribution function , is able to constrain satisfactorily the orbital structure , mass - to - light ratio , and inclination of the system under study
. applications to data for galactic globular clusters and nearby de galaxies will be presented in future papers .
these are only two examples of a large range of dynamical problems in astronomy to which a discrete schwarzschildcode like ours can be applied , so we expect this new tool will contribute to the better understanding of stellar systems in general . we are happy to thank marla geha and raja guhathakurta for their continued interest in the present work and its extension to the study of actual galaxies using their unique data on dwarf ellipticals
. we also thank glenn van de ven for very useful discussions , his interest in the progress of this project and , last but not least , for his invaluable help with idl routines .
this paper also benefited by comments from davor krajnovic , aaron romanowsky , and david merritt .
thanks also to george meylan for his help with the writing of the hst theory proposal specified below , and to the anonymous referee , whose comments and suggestions improved the presentation of the paper .
this work was carried out as part of hst theory project # 9952 and was supported by nasa through a grant from stsci , which is operated by aura , inc .
, under nasa contract nas 5 - 26555 .
batsleer , p. , & dejonghe , h. 1993 , , 271 , 104 batsleer , p. , & dejonghe , h. 1995 , , 294 , 693 bender , r. , et al .
2005 , , 631 , 280 binney , j. , & mamon , g. a. 1982 , , 200 , 361 binney , j. , & tremaine , s. 1987 , princeton , nj , princeton university press , 1987 cappellari , m. , verolme , e. k. , van der marel , r. p. , kleijn , g. a. v. , illingworth , g. d. , franx , m. , carollo , c. m. , & de zeeuw , p. t. 2002 , , 578 , 787 cappellari , m. , et al . 2006 , , 366 , 1126 ct , p. , et al . 2001 , , 559 , 828 ct , p. , mclaughlin , d. e. , cohen , j. g. , & blakeslee , j. p. 2003
, , 591 , 850 cretton , n. , de zeeuw , p. t. , van der marel , r. p. , & rix , h .- w . 1999 , , 124 , 383 cretton , n. , rix , h .- w . ,
& de zeeuw , p. t. 2000 , , 536 , 319 cretton , n. , & emsellem , e. 2004 , , 347 , l31 davies , r. i. , et al .
2006 , , 646 , 754 dehnen , w. , & gerhard , o. e. 1994 , , 268 , 1019 dekel , a. , stoehr , f. , mamon , g. a. , cox , t. j. , novak , g. s. , & primack , j. r. 2005 , , 437 , 707 de bruijne , j. h. j. , van der marel , r. p. , & de zeeuw , p. t. 1996 , , 282 , 909 de lorenzi , f. , debattista , v. p. , gerhard , o. , & sambhus , n. 2007 , , 376 , 71 douglas , n. g. , et al .
2002 , , 114 , 1234 douglas , n. g. , et al .
2007 , arxiv : astro - ph/0703047 gebhardt , k. , pryor , c. , williams , t. b. , hesser , j. e. , & stetson , p. b. 1997 , , 113 , 1026 gebhardt , k. , et al .
2000 , , 119 , 1157 gebhardt , k. , et al . 2003 , , 583 , 92 geha , m. , guhathakurta , p. , rich , r. m. , & cooper , m. c. 2006 , , 131 , 332 gerhard , o. e. 1993 , , 265 , 213 gerhard , o. , jeske , g. , saglia , r. p. , & bender , r. 1998 , , 295 , 197 gerssen , j. , van der marel , r. p. , gebhardt , k. , guhathakurta , p. , peterson , r. c. , & pryor , c. 2002 , , 124 , 3270 gilbert , k. m. , et al .
2006 , , 652 , 1188 gilbert , k. m. , et al .
2007 , , 668 , 245 kleyna , j. t. , wilkinson , m. i. , evans , n. w. , & gilmore , g. 2001 , , 563 , l115 kleyna , j. , wilkinson , m. i. , evans , n. w. , gilmore , g. , & frayn , c. 2002 , , 330 , 792 krajnovi , d. , cappellari , m. , emsellem , e. , mcdermid , r.m . , & de zeeuw , p.t .
2005 , , 357 , 1113 kunkel , w. e. , demers , s. , irwin , m. j. , & albert , l. 1997 , , 488 , l129 kunkel , w. e. , demers , s. , & irwin , m. j. 2000 , , 119 , 2789 okas , e. l. 2002 , , 333 , 697 okas , e. l. , & mamon , g. a. 2003 , , 343 , 401 okas , e. l. , mamon , g. a. , & prada , f. 2005 , , 363 , 918 magorrian , j. 2006 , , 373 , 425 magorrian , j. , et al .
1998 , , 115 , 2285 mayor , m. , et al .
1997 , , 114 , 1087 mclaughlin , d. e. , anderson , j. , meylan , g. , gebhardt , k. , pryor , c. , minniti , d. , & phinney , s. 2006 , , 166 , 249 mcnamara , b. j. , harrison , t. e. , & anderson , j. 2003 , , 595 , 187 merritt , d. 1993 , , 413 , 79 merritt , d. , & saha , p. 1993
, , 409 , 75 merritt , d. , meylan , g. , & mayor , m. 1997 , , 114 , 1074 merritt , d. 1999 , , 111 , 129 press , w. h. , teukolsky , s. a. , vetterling , w. t. , & flannery , b. p. 1992
, cambridge university press , 1992 , 2nd ed .
reijns , r. a. , seitzer , p. , arnold , r. , freeman , k. c. , ingerson , t. , van den bosch , r. c. e. , van de ven , g. , & de zeeuw , p. t. 2006 , , 445 , 503 richstone , d. o. , & tremaine , s. 1984 , , 286 , 27 richtler , t. , et al .
2004 , , 127 , 2094 rix , h .- w . , de zeeuw , p. t. , cretton , n. , van der marel , r. p. , & carollo , c. m. 1997 , , 488 , 702 romanowsky , a. j. , & kochanek , c. s. 2001 , , 553 , 722 romanowsky , a. j. , douglas , n. g. , arnaboldi , m. , kuijken , k. , merrifield , m. r. , napolitano , n. r. , capaccioli , m. , & freeman , k. c. 2003 , science , 301 , 1696 schwarzschild , m. 1979 , , 232 , 236 shanno , d.f .
, & phua , k.h . 1980 , acm transactions on mathematical software , 6 , 618 simon , j. d. , & geha , m. 2007 , , 670 , 313 suntzeff , n. b. , & kraft , r. p. 1996
, , 111 , 1913 syer , d. , & tremaine , s. 1996 , , 282 , 223 teodorescu , a. m. , mndez , r. h. , saglia , r. p. , riffeser , a. , kudritzki , r .-
p . , gerhard , o. e. , & kleyna , j. 2005 , , 635 , 290 thomas , j. , saglia , r. p. , bender , r. , thomas , d. , gebhardt , k. , magorrian , j. , & richstone , d. 2004 , , 353 , 391 valluri , m. , merritt , d. , & emsellem , e. 2004 , , 602 , 66 van de ven , g. , van den bosch , r.c.e . ,
verolme , e.k .
, & de zeeuw , p.t . 2006 , , 445 , 513 van den bosch , r. , de zeeuw , t. , gebhardt , k. , noyola , e. , & van de ven , g. 2006 , , 641 , 852 van der marel , r. p. , & franx , m. 1993 , , 407 , 525 van der marel , r.p . , evans , n.w . ,
rix , h .- w . ,
white , s.d.m .
, & de zeeuw , p.t . 1994 , , 271 , 99 van der marel , r. p. , de zeeuw , p. t. , & rix , h .- w .
1997 , , 488 , 119 van der marel , r.p . ,
sigurdsson , s. , & hernquist , l. 1997 , , 487 , 153 van der marel , r.p . ,
cretton , n. , de zeeuw , p.t . , & rix , h .- w .
1998 , , 493 , 613 van der marel , r. p. , magorrian , j. , carlberg , r. g. , yee , h. k. c. , & ellingson , e. 2000 , , 119 , 2038 van der marel , r. p. , alves , d. r. , hardy , e. , & suntzeff , n. b. 2002 , , 124 , 2639 van der marel , r. p. , & van dokkum , p. g. 2006 , astro - ph/0611571 vandervoort , p.o . 1984 , , 287 , 475 van leeuwen , f. , le poole , r. s. , reijns , r. a. , freeman , k. c. , & de zeeuw , p. t. 2000 , , 360 , 472 verolme , e. k. , & de zeeuw , p. t. 2002 , , 331 , 959 walker , m. g. , mateo , m. , olszewski , e. w. , bernstein , r. , wang , x. , & woodroofe , m. 2006 , , 131 , 2114 wilkinson m.i . ,
kleyna , j.t . ,
evans , n.w . , gilmore g. , 2001 , mnras , 330 , 778 wilkinson , m. i. , kleyna , j. t. , evans , n. w. , gilmore , g. f. , irwin , m. j. , & grebel , e. k. 2004 , , 611 , l21 wojtak , r. , & okas , e. l. 2007 , , 377 , 843 wojtak , r. , okas , e. l. , mamon , g. a. , gottlber , s. , prada , f. , & moles , m. 2007 , , 466 , 437 wu , x. 2007 , astro - ph/0702233 wu , x. , & tremaine , s. 2006 , , 643 , 210 ccrlcrlcrlcrlcrl 55ns & & 4.5138 & 0.4531 & & 0.2207 & 0.1404 & & 0.0908 & 0.0359 & & 0.1080 & 0.0719 & & 0.2634 & 0.1591 + 55is & & 6.6525 & 0.5630 & & 0.3677 & 0.1915 & & 0.0939 & 0.0309 & & 0.1912 & 0.1598 & & 0.2460 & 0.1855 + 55ms & & 3.1524 & 0.4208 & & 0.2123 & 0.1207 & & 0.0884 & 0.0334 & & 0.1845 & 0.0903 & & 0.2505 & 0.1122 + 90ns & & 3.7560 & 0.5939 & & 0.2683 & 0.1633 & & 0.1419 & 0.0425 & & 0.2202 & 0.1365 & & 0.2361 & 0.2491 + 90is & & 2.7956 & 0.6410 & & 0.6253 & 0.1629 & & 0.1366 & 0.0347 & & 0.4826 & 0.1178 & & 0.2566 & 0.2364 + 90ms & & 1.4893 & 0.4927 & & 0.2298 & 0.1714 & & 0.1216 & 0.0384 & & 0.1456 & 0.1149 & & 0.2953 & 0.2333 + | we present a new schwarzschildorbit - superposition code that is designed to model discrete datasets composed of velocity measurements of individual kinematic tracers in a dynamical system .
this constitutes an extension of previous implementations that can only address continuous data in the form of ( the moments of ) velocity distributions , thus avoiding potentially important losses of information due to data binning .
furthermore , the code can handle any combination of available velocity components , i.e. , only line - of - sight velocities , only proper motions , or a combination of both .
it can also handle a combination of discrete and continuous data .
the code determines the combination of orbital mass weights ( representing the distribution function ) as a function of the three integrals of motion @xmath0 and @xmath1 that best reproduces , in a maximum - likelihood sense , the available kinematic and photometric observations in a given axisymmetric gravitational potential .
the overall best fit is the one that maximizes the likelihood over a parameterized set of trial potentials .
the fully numerical approach ensures considerable freedom on the form of the distribution function @xmath2 .
this allows a very general modeling of the orbital structure , thus avoiding restrictive assumptions about the degree of ( an)isotropy of the orbits .
we describe the implementation of the discrete code and present a series of tests of its performance based on the modeling of simulated ( i.e. , artificial ) datasets generated from a known distribution function .
we explore pseudo - datasets with varying degrees of overall rotation and different inclinations on the plane of the sky , and study the results as a function of relevant observational variables such as the size of the dataset and the type of velocity information available .
we find that the discrete schwarzschildcode recovers the original orbital structure , mass - to - light ratio , and inclination of the input datasets to satisfactory accuracy , as quantified by various statistics .
the code will be valuable , e.g. , for modeling stellar motions in galactic globular clusters , and modeling the motions of individual stars , planetary nebulae , or globular clusters in nearby galaxies .
this can shed new light on the total mass distributions of these systems , with central black holes and dark matter halos being of particular interest . |
it is an honor , of course , but also a pleasure for me to join in this celebration of murray gell - mann s sixtieth birthday and to address such a distinguished audience .
murray was my teacher and more recently we have worked together in the search for a quantum framework within which to erect a fundamental description of the universe which would encompass all scales from the microscopic scales of the elementary particle interactions to the most distant reaches of the realm of the galaxies from the moment of the big bang to the most distant future that one can contemplate .
such a framework is needed if we accept , as we have every reason to , that at a basic level the laws of physics are quantum mechanical .
further , as i shall argue below , there are important features of our observations which require such a framework for their explanation .
this application of quantum physics to the universe as a whole has come to be called the subject of quantum cosmology .
the assignment of the organizers was to speak on the topic `` where are our efforts leading ? ''
i took this as an invitation to speculate , for i think that it is characteristic of the frontier areas of science that , while we may know what direction we are headed , we seldom know where we will wind up .
nevertheless , i shall not shrink from this task and endeavor , in the brief time available , to make a few remarks on the future of quantum cosmology .
the point of view that i shall describe owes a great deal to my conversations with murray .
one can not contemplate the history of physics without becoming aware that many of its intellectual advances have in common that some idea which was previously accepted as fundamental , general , and inescapable was subsequently seen to be consequent , special , and dispensable .
further , this was often for the following reason : the idea was not truly a general feature of the world , but only _ perceived _ to be general because of our special place in the universe and the limited range of our experience .
in fact , it arose from a true physical fact but one which is a special situation in a yet more general theory . to quote murray himself from a talk of several years ago : `` in my field an important new idea ..... almost always includes a negative statement , that some previously accepted principle is unnecessary and can be dispensed with . some earlier correct idea ....
was accompanied by unnecessary intellectual baggage and it is now necessary to jettison that baggage.''@xcite in cosmology it is not difficult to cite previous examples of such transitions @xcite .
the transition from ptolemaic to copernican astronomy is certainly one .
the centrality of the earth was a basic assumption of ptolemaic cosmology .
after copernicus , the earth was seen not to be fundamentally central but rather one planet among others in the solar system .
the earth was in fact distinguished , not by a law of nature , but rather by our own position as observers of the heavens .
the idea of the central earth was excess baggage .
the laws of geometry of physical space in accessible regions obey the euclidean laws typified by the pythagorean theorem on right triangles
. laws of physics prior to 1915 , for example those governing the propagation of light , incorporated euclidean geometry as a fundamental assumption . after
einstein s 1915 general theory of relativity , we see euclidean geometry not as fundamental , but rather as one possibility among many others . in einstein
s theory , the geometry in the neighborhood of any body with mass is non - euclidean and this curvature gives a profound geometrical explanation for the phenomena we call gravity . on the very largest scales of cosmology , geometry is significantly curved and it is the dynamics of this geometry which describes the evolution of the universe .
euclidean geometry is the norm for us , not because it is fundamental , but only because our observations are mostly local and because we happen to be living far from objects like black holes or epochs like the big bang .
the idea of a fixed geometry was excess baggage .
in each of these examples there was a feature of the current theoretical framework which was perceived as fundamental but which in truth was a consequence of our particular position and our particular time in the universe in a more general theoretical framework .
our description was too special .
there was excess baggage which had to be discarded to reach a more general and successful viewpoint .
thus , in our effort to predict the future of quantum cosmology , the question naturally arises : which features of our _ current _ theoretical framework reflect our special position in the universe and which are fundamental ? which are excess baggage ?
we live at a special position in the universe , not so much in place , as in time .
we are late , living some ten billion years after the big bang , a time when many interesting possibilities for physics could be realized which are not easily accessible now .
moreover , we live in a special universe .
ours is a universe which is fairly smooth and simple on the largest scales , and the evidence of the observations is that if we look earlier in time it is smoother and simpler yet .
there are simple initial conditions which are only one very special possibility out of many we could imagine
. the question i posed above can therefore be generalized : which features of our current theoretical framework are fundamental and which reflect our special position in the universe _ or _ our special initial conditions .
are there natural candidates for elements of the framework which might be generalized ? is there excess baggage ?
this is the question that i would like to address today for quantum cosmology .
first , let us review the current quantum mechanical framework today as it was developed in the twenties , thirties and forties , and as it appears in most textbooks today .
it is familiar enough that there is no certainty in this world and that therefore we must deal in probabilities . when probabilities are sufficiently good we act .
it was nt certain that my plane to los angeles would nt be blown up , but i came because i thought that the probability was sufficiently low .
experimentalists ca nt be certain that the results of their measurements are not in error , but they publish because they estimate the probability of this as low .
and so on .
it was the vision of classical physics that fundamentally the world _ was _ certain and that the use of probability was due to lack of precision .
if one looked carefully enough to start , one could with certainty predict the future .
since the discovery of quantum mechanics some sixty years ago , we have known that this vision is only an approximation and that probabilities are inevitable and fundamental .
= 3.5 in indeed , the transition from classical to quantum physics contains a good example of excess baggage . in this case
it was classical determinism .
classical evolution was but one possibility out of many although by far the most probable evolution in the great majority of situations we were used to .
quantum mechanics constructs probabilities according to characteristic rules . as a simple example let us consider the the quantum mechanics of a slowly moving particle such as an electron in a solid .
its path in spacetime is called its history . in classical mechanics
the time history of the electron is a definite path determined by the electron s initial conditions and its equation of motion .
thus , given sufficiently precise initial conditions , a later observation of position can yield only a single , certain , predictable result . in quantum mechanics ,
given the most precise possible initial conditions , all paths are possible and all possible results of the observation may occur .
there is only a probability for any one of them .
the probability of an observation of position at time @xmath0 yielding the value @xmath1 , for example , is constructed as follows : a complex number , called the amplitude , is assigned to each path .
this number has the form exp@xmath2path@xmath3 $ ] where the action @xmath4 the inertial properties of the electron together with its interactions and @xmath5 is planck s constant .
the amplitude to arrive at @xmath1 at @xmath0 is the sum of the path amplitudes over all paths which are consistent with the initial conditions and which end at @xmath1 at @xmath0 ( figure 1 ) .
@xmath6\ , .
\label{twoone}\ ] ] the sum here is over all paths consistent with the initial conditions and ending at position @xmath1 at time @xmath0 .
for example , if the position of the particle were actually measured at a previous time but not in between , one would sum over all paths which connect that position with @xmath1 .
the probability to arrive at @xmath1 at @xmath0 is the absolute square of this amplitude .
in essence , this is feynman s sum over histories formulation of the rules of quantum mechanics .
there is another , older , way of stating these rules called the hamiltonian formulation of quantum mechanics . here
, the amplitude for observing the electron at @xmath1 at time @xmath0 is called the wave function @xmath7 .
if , as above , we can construct the amplitude @xmath8 at one time , we can also employ this construction at all other times .
the time history of the wave function gives a kind of running summary of the probability to find the electron at @xmath1 .
the wave function is thus the closest analog to the classical notion of `` state of a system at a moment of time '' .
the wave function obeys a differential equation called the schrdinger equation @xmath9 where @xmath10 is an operator , called the hamiltonian , whose form can be derived from the action .
it summarizes the dynamics in an equivalent way .
thus , although the electron s position does not evolve according to a deterministic rule , its wave function does .
feynman s sum over histories formulation of quantum mechanics is fully equivalent to the older hamiltonian formulation for this case of slowly moving particles .
one can calculate the wave function from for two nearby times and demonstrate that it satisfies the schrdinger equation .
one can use either formulation of quantum mechanics in this and many other situations as well .
more generally than the probability of one observation of position at one time , the probabilities of time _ sequences _ of observations are of interest .
just such a sequence in needed , for example , to check whether or not the electron is moving along a classical path between the initial conditions and time @xmath0 .
the amplitude for positive answers to checks of whether the electron is located inside a sequence of position intervals is the sum of the amplitude for a path over all paths which thread these intervals ( figure 2 ) .
the joint probability for the sequence of positive answers is the square of this amplitude .
= 3.5 in .
this means that the probability of finding the particle at @xmath1 at @xmath0 depends not only on the initial conditions , but also on what measurements were carried out between the initial conditions and @xmath0 . even if one does nt know the results of these measurements ,
the probabilities at @xmath0 will still depend on whether or not they were carried out .
thus , there are in quantum mechanics different rules for evolution depending on whether measurements occurred or did not . in the hamiltonian formulation
this means that the wave function evolves by the schrdinger equation only in between measurements . at a measurement it evolves by a different rule the notorious `` reduction of the wave packet '' . in some circumstances it makes no difference to the probabilities whether prior measurements were carried out or not .
for example , the classical limit of quantum mechanics occurs when the initial conditions are such that only a single path to @xmath1 the classical one contributes to the sum over histories .
then , as long as sufficiently crude measurements are considered , it makes no difference to the probability of @xmath1 at @xmath0 whether they are made or not .
the electron follows the classical evolution .
in such cases the classical history is said to `` decohere '' .
the framework of quantum mechanics described in the previous section was the starting point for the `` copenhagen '' interpretations of this subject .
an idea characteristic of the copenhagen interpretations was that there was something external to the framework of wave function and schrdinger equation which was necessary to interpret the theory .
various expositors put this in different ways : bohr @xcite spoke of alternative descriptions in terms of classical language .
landau and lifshitz @xcite emphasized preferred classical observables .
heisenberg and others @xcite stressed the importance of an external observer for whom the wave function was the most complete summary possible of information about an observed system .
all singled out the measurement process for a special role in the theory . in various ways
, these authors were taking as fundamental the manifest existence of the classical world that we see all about us .
that is , they were taking as fundamental the existence of objects which have histories whose probabilities obey the rules of classical probability theory and , except for the occasional intervention of the quantum world as in a measurement , obey deterministic classical equations of motion .
this existence of a classical world , however it was phrased , was an important part of the copenhagen interpretations for it was the contact with the classical world which mandated the `` reduction of the wave packet '' .
the copenhagen pictures do not permit the application of quantum mechanics to the universe as a whole . in the copenhagen interpretations
the universe is always divided into two parts : to one part quantum mechanical rules apply . to the other part classical rules
apply @xcite .
it was everett who in 1957 first suggested how to generalize the copenhagen framework so as to apply to cosmology @xcite .
his idea ( independently worked out by murray in 1963 ) was to take quantum mechanics seriously and apply it to the universe as a whole .
he started with the idea that there is one wave function for the universe always evolving by the schrdinger equation and never reduced .
any observer would be part of the system described by quantum mechanics , not separate from it .
everett showed that , if the universe contains an observer , then its activities measuring , recording , calculating probabilities , etc .
could be described in this generalized framework .
further , he showed how the copenhagen structure followed from this generalization when the observer had a memory which behaved classically and the system under observation behaved quantum mechanically . yet , everett s analysis was not complete .
it did not explain the manifest existence of the classical world much less the existence of something as sophisticated as an observer with a classically behaving memory .
classical behavior , after all , is not generally exhibited by quantum systems .
the subsequent , post - everett , analysis of this question involves a synthesis of the ideas of many people . out of many , i might mention in particular the work of joos and zeh @xcite , zurek @xcite , griffiths @xcite , and latterly murray , val telegdi , and myself .
it would take us too far to attempt to review the mechanisms by which the classical world arises but we can identify the theoretical feature to which its origin can , for the most part , be traced . a classical world can not be a general feature of the quantum mechanics of the universe , for the number of states which imply classical behavior in any sense is but a poor fraction of the total states available to the universe .
classical behavior , of course , can be an approximate property of a _
particular _ state as for a particle in a wave packet whose center moves according to the classical equations of motion .
but , the universe _ is _ in a particular state ( or a particular statistical mixture of states ) .
more exactly , particular quantum initial conditions must be posed to make any prediction in quantum cosmology .
it is to the particular features of the initial conditions of the universe , therefore , that we trace the origin of today s classical world and the possibility of such information gathering and utilizing systems as ourselves .
in retrospect , the copenhagen idea that a classical world or an observer together with the act of measurement occupy a fundamentally distinguished place in quantum theory can be seen to be excess baggage .
these ideas arose naturally , in part from our position in the late universe where there _ are _ classically behaving objects and even observers . in part
, they arose from the necessary focus on laboratory experiments as the most direct probe of quantum phenomena where there _ is _ a clear distinction between observer and observed .
however , these features of the world , while true physical facts in these situations , are not fundamental .
quantum mechanically , the classical world and observers are but some possible systems out of many and measurements are but one possible interaction of such systems out of many .
both are unlikely to exist at all in the very early universe .
they are , in a more general post - everett framework , _ approximate _ features of the late universe arising from its particular quantum state .
the classical reality to which we have become so attached by evolution is but an approximation in an entirely quantum mechanical world made possible by specific initial conditions .
the originators of the copenhagen interpretation were correct ; something beyond the wave function and the schrdinger equation _ is _ needed to interpret quantum mechanics .
but , that addition is not an external restriction of the domain to which the theory applies .
it is the initial conditions of the universe specified within the quantum theory itself .
heat flows from hot to cold bodies , never from cold to hot .
this is the essence of the second law of thermodynamics that entropy always increases . in the nineteenth century
this law was thought to be strict and fundamental @xcite .
a direction of time was distinguished by this law of nature .
after the success of the molecular theory of matter , with its _ time symmetric _ laws of dynamics , the increase in entropy was seen to be _ approximate _ one possibility out of many others although the overwhelmingly most probable possibility in most situations .
a direction of time was distinguished , not fundamentally , but rather by our position in time relative to initial conditions of simplicity and our inability to follow molecular motion in all accuracy .
the idea of a fundamental thermodynamic arrow of time was excess baggage .
exchanging the thermodynamic arrow of time for simple initial conditions might seem to be exchanging one asymmetry for another .
`` why '' , one could ask , `` was the universe simple in the past and not in the future ? '' in fact , this is not a question .
there is no way of specifying that direction in time which is the past except by calling it the direction in which the universe is simple .
its not an arrow of time which is fundamental , but rather the fact that the universe is simple at one end and not at the other @xcite .
the hamiltonian formulation of quantum mechanics also possesses a similar distinction between past and future . from the knowledge of the state of a system at one time _ alone _
, one can predict the future but one can not , in general , retrodict the past @xcite .
this is an expression of causality in quantum theory . indeed , quantum mechanics in general prohibits the construction of history .
the two slit experiment is a famous example ( figure 3 ) .
unless there was a measurement , it is not just that one ca nt say for certain which slit the electron went through ; it is meaningless even to assign a probability . in this asymmetry between past and future the notion of `` state '' in quantum mechanics
is different from that of classical physics from which both future and past can be extrapolated . in hamiltonian quantum mechanics ,
the past is the direction which is comprehensible by theory ; the future is the direction which is generally predictable .
= 3.5 in as long as the distinction between future and past in cosmology is fundamental , it is perhaps reasonable to formulate a quantum cosmology using a quantum mechanics which maintains this same distinction as fundamental .
however , it would seem more natural if causality were an empirical conclusion about the universe rather than a prerequisite for formulating a theory of it .
that is , it would seem more natural if causality were but one of many options available in quantum mechanics , but the one appropriate for this particular universe .
for example , one might imagine a universe where both initial _ and _ final conditions are set .
the generalization of hamiltonian quantum mechanics necessary to accommodate such situations is not difficult to find @xcite .
in fact , it is ready to hand in the sum over histories formulation described briefly in section ii .
the rules are the same , but both initial and final conditions are imposed on a contributing history . in the context of such a generalization ,
the arrow of time in quantum mechanics , and the associated notion of state @xcite would not be fundamental .
rather they would be features of the theory arising from the fact that those conditions which fix our particular universe are comprehensible at one of its ends but not at the other .
gravity governs the structure and evolution of the universe on the largest scales , both in space and time . to pose a quantum theory of cosmology , therefore , we need a quantum theory of gravity . the essence of the classical theory of gravity
einstein s general relativity is that gravity is curved spacetime . we thus need a quantum theory of space and time themselves in which the geometry of spacetime will exhibit quantum fluctuations . finding a consistent , manageable quantum theory of gravity has been one of the goals of theoretical research for the last thirty years . the chief problem has been seen as finding a theory which puts out more predictions then there are input parameters .
but , in cosmology there is also a conflict with the existing framework of quantum mechanics .
these difficulties are called `` the problem of time '' @xcite time plays a special and peculiar role in the familiar hamiltonian formulation of quantum mechanics .
all observations are assumed to be unambiguously characterized by a single moment of time and we calculate probabilities for `` observations at one moment of time '' .
time is the only observable for which there are no interfering alternatives ( as a measurement of momentum is an interfering alternative for a measurement of position ) .
time is the sole observable not represented in the theory as an operator but rather enters the schrdinger equation as a parameter describing evolution .
if the geometry of spacetime is fixed , external and classical , it provides the preferred time of quantum mechanics .
( more exactly , a fixed background spacetime provides a family of timelike directions equivalent in their quantum mechanics because of relativistic causality . )
if the geometry of spacetime is quantum mechanical fluctuating and without definite value then it can not supply a notion of time for quantum mechanics .
there is thus a conflict between the familiar hamiltonian quantum mechanics with a preferred time and a quantum theory of spacetime .
this is the `` problem of time '' . the usual response to this difficulty has been to keep hamiltonian quantum mechanics but to give up on spacetime .
there are several candidates for a replacement theory .
perhaps space and time are separate quantum mechanically with the beautiful synthesis of minkowski and einstein emerging only in the classical limit @xcite .
perhaps spacetime needs to be augmented by other , now hidden , variables which play the role of a preferred time in quantum mechanics @xcite .
perhaps spacetime is simply a totally inappropriate notion fundamentally and emerges only in the classical limit of some yet more subtle theory .
each of these ideas could be right .
i would like to suggest , however , that there is another alternative .
this is that the notion of a preferred time in quantum mechanics is another case of excess baggage . that the hamiltonian formulation of quantum mechanics is simply not general enough to encompass a quantum theory of spacetime . that the present formulation of quantum mechanics is but an approximation to a more general framework appropriate because of our special position in the universe .
it is not difficult to identify that property of the late universe which would have led to the perception that there is a preferred time in quantum mechanics . here ,
now , on all accessible scales , from the smallest ones of the most energetic accelerators to the largest ones of the furthest seeing telescopes , spacetime _ is _ classical .
the preferred time of quantum mechanics reflects this true physical fact .
however , it can only be an approximate fact if spacetime is quantum mechanical .
indeed , like all other aspects of the classical world , it must be a property of the particular quantum initial conditions of the universe .
the present formulation of quantum mechanics thus may be only an approximation to a more general framework appropriate because of particular initial conditions which mandate classical spacetime in the late universe . what is this more general quantum mechanics ?
feynman s sum over histories supplies a natural candidate @xcite i shall now sketch why .
we saw earlier that feynman s approach was equivalent to the hamiltonian one for particle quantum mechanics .
it is also for field theory .
a little thought about our example shows , however , that this equivalence arises from a special property of the histories that they do not double back on themselves in time .
thus , the preferred time enters the sum over histories formulation as a restriction on the allowed histories . in a quantum theory of spacetime
the histories are four dimensional spacetimes geometries @xmath11 with matter fields @xmath12 living upon them .
the sums over histories defining quantum amplitudes therefore have the form : @xmath13/ \hbar \right ) \label{fiveone}\ ] ] where @xmath4 is the action for gravity coupled to matter .
the sums are restricted by the theory of initial conditions and by the observations whose probabilities one aims to compute .
the interior sum in @xmath14/ \hbar \right ) \label{fivetwo}\ ] ] defines a usual quantum field theory in a temporarily fixed background spacetime @xmath11 .
the requirement that the fields be single valued on spacetime is the analog of the requirement for particles that the paths do not double back in time .
this quantum field theory thus possesses an equivalent hamiltonian quantum mechanics with the preferred time directions being those of the background @xmath11 . when the remaining sum over @xmath11 is carried out the equivalence with any hamiltonian formulation disappears .
there is no longer any fixed geometry , any external time , to define the preferred time of a hamiltonian formulation .
all that is summed over
. there would be an _ approximate _
equivalence with a hamiltonian formulation were the initial conditions to imply that for large scale questions in the late universe , only a single geometry @xmath15 ( or an incoherent sum of geometries ) contributes to the sum over geometries .
for then , @xmath16/ \hbar\right ) \approx \sum_{\phi ( x ) } \exp \left(i\ , s[{\hat { \cal g } } , \phi ( x)]/ \hbar\right ) \label{fivethree}\ ] ] and the preferred time can be that of the geometry @xmath15 .
it would be in this way that the familiar formulation of quantum mechanics emerges as an approximation to a more general sum over histories framework appropriate to specific initial conditions and our special position in the universe so far from the big bang and the centers of black holes .
the preferred time of familiar quantum mechanics would then be another example of excess baggage . in declaring the preferred time of familiar quantum mechanics excess baggage , one is also giving up as fundamental a number of treasured notions with which it is closely associated .
these include the notion of causality , any notion of state at a moment of time , and any notion of unitary evolution of such states .
each of these ideas requires a background spacetime for its definition and in a quantum theory _ of _ spacetime there is none to be had .
such ideas , however , would have an approximate validity in the late universe on those scales where spacetime is classical .
many will regard discarding these notions as too radical a step but i think it no less radical than discarding spacetime
one of the most powerful organizing features of our experience .
in this discussion of the excess baggage which has been discarded to arrive at a quantum mechanical framework general enough to apply to the universe as a whole , we have progressed from historical examples , through reasonable generalizations , to topics of current research and debate . even in the last case
, i was able to indicate to you candidates for the necessary generalizations .
i would now like to turn to features of the theoretical framework where this is not the case ; where there may be excess baggage but for which i have no clear candidates for the generalizations .
a candidate for excess baggage is the idea that there is anything `` initial '' about the conditions which are needed in addition to the laws of dynamics to make predictions in the universe .
our idea that conditions are initial comes from big bang cosmology in which the universe has a history with a simple beginning ten and some billion years ago .
the global picture of spacetime is that there is a more or less uniform expansion everywhere from this moment .
perhaps the very large scale structure of spacetime is very different .
perhaps , as a. linde has suggested @xcite there are many beginnings
. our local expanding universe might be just one inflationary bubble out of many others . perhaps , therefore , the idea that conditions should reflect a simple state at one end of the universe is excess baggage arising from our confinement to one of these bubbles .
conditions in addition to dynamics would still be needed for prediction but they might be of a very different character @xcite . in the framework we have discussed so far there are three kinds of information necessary for prediction in quantum cosmology : first , there are the laws of dynamics summarized by the action .
second , there is the specification of initial ( or other ) conditions .
third , there are our specific observations .
is it fundamental that the process of prediction be structured in this way ?
our focus on laboratory science , i believe , is the origin of the idea that there is a clean separation between dynamics and conditions .
dynamics are governed by the laws of nature ; that which we are seeking , not that which we control .
by contrast , the conditions represented in the experimental arrangement are up to us and therefore not part of the laws of nature .
this ideal of control is not truly realized in practice in the laboratory , but in cosmology it is never realized .
the initial conditions of the universe are most definitely not up to us but must be specified in the theoretical framework in the same law - like way as the dynamics .
a law of initial conditions has , therefore , become one of the central objectives of quantum cosmology @xcite . however , is it not possible that the distinction between conditions and dynamics is excess baggage arising from the focus on laboratory science to which our limited resources have restricted us ?
is it not possible that there are some more general principles in a more general framework which determine both the conditions and dynamics ?
is it not possible , as it were , that there is a principle which fixes both the state and the hamiltonian ?
recent work by hawking @xcite , coleman @xcite , giddings and strominger @xcite among others gives some encouragement to this point of view .
they show that in a quantum theory of spacetime which includes wormholes small handles on spacetime connecting perhaps widely separated spacetime regions that there is a closer connection between initial conditions and dynamics than had been previously thought .
specifically , the initial conditions determine the form of the hamiltonian that we see at energies below those on which spacetime has quantum fluctuations .
indeed , they may determine a range of possibilities for the hamiltonian only one of which is realized in our large scale universe .
dynamics and initial conditions are thereby entwined at a fundamental level .
the last candidate for excess baggage that i want to discuss is the idea that the laws of physics , and in particular laws of initial conditions , are unique , apart from the universe , apart from the process of their construction , and apart from us @xcite scientists , like mathematicians , proceed as though the truths of their subjects had an independent existence .
we speak , for example , of `` discovering '' the laws of nature as though there were a single set of rules by which the universe is run with an actuality apart from this world they govern . most honestly , the laws of physics are properties of the collective data that we have about the universe . in the language of complexity theory ,
this data is compressible .
there are computational algorithms by which the data can be stored in a shorter message .
take , for example , an observed history of motion of a system of classical particles .
this history could be described to a given accuracy by giving the position and momentum of each particle at a suitably refined series of times . however
, this message can be compressed to a statement of the positions and momenta at _ one _ time plus the equations of motion .
some of these initial values might be specified by observation but most , for a large system , will be specified by theory and statistically at that . for large systems
the result is a much shorter message .
further , we find for many different systems that the data can be compressed to the form of initial conditions plus the _ same _ equations of motion .
it is the universal character of this extra information beyond the initial conditions which gives the equations of motion their law - like character . similarly in quantum mechanics .
thus we have : @xmath17 the laws of physics , therefore , do not exist independently of the data .
they are properties of our data much like `` random '' or `` computable '' is a property of a number although probably ( as we shall see below ) in a less precise sense @xcite .
this characterization of compression is incomplete .
any non - physicist knows that one ca nt just take a list of numbers and compress them in this way .
one has to take a few courses in physics and mathematics to know what the symbols mean and how to compute with them .
that is , to the list on the right should be appended the algorithms for numerical computation for practical implementation of the theory , that part of mathematics which needed to interpret the results , the rules of the language , etc . etc .
yet more properly we should write : @xmath18 what confidence do we have that our data can always be compressed in this way ?
i have discussed the possibility that a clear division between initial conditions and equations of motion may be excess baggage .
bob geroch and i have discussed similar questions for the division between algorithms and the rest of physical theory @xcite .
however it is subdivided , what guarantee do we have that the resulting theory will be unique , independent of its process of construction , independent of the specific data we have acquired ? very little it seems .
as we move down the list on the right we encounter more and more items which seem particular to our specific history as a collectivity of observers and to our specific data .
it is an elementary observation that there are always many theories which will fit a given set of data just as there are many curves which will interpolate between a finite set of points .
further , when the data is probabilistic as it is in quantum theory , there is always the possibility of arriving at different theories whatever criteria are used to distinguish them . beyond this ,
however , what confidence do we have that different groups of observers , with different histories , with growing but different sets of specific data , will , in the fullness of time , arrive at the same fundamental theory ?
i do not mean to suggest that the theories might vary because they are _ consequences _ of specific history , for that is not science .
rather the question is whether there observationally indistinguishable theories which are different because of the processes of their discovery
. there may be specifically cosmological reasons to expect non - uniqueness in theories of initial conditions .
a theory of initial conditions , for example , must be simple enough that it can be stored within the universe .
if the initial conditions amounted to some particular complex specification of the state of all matter this would not be possible .
the act of constructing theories may limit our ability to find them .
the gravitational effect of moving a gram of matter on sirius by one centimeter in an unknown direction is enough to ruin our ability to extrapolate classically the motions of a mole of gas particles beyond a few seconds into the past @xcite . in view of this
there must be many theories of initial conditions rendered indistinguishable simply by our act of constructing them .
it would be interesting , i think , to have a framework which dispensed with the excess baggage that the laws of physics were separate from our observations of the universe , a framework in which the inductive process of constructing laws about the universe was described in it , and in which our theories were seen as but one possibility among many .
the assignment of the organizers was not just to speak on the subject `` where are our efforts leading ? '' .
they also wanted to know `` in fifty or one hundred years time how do you think today s efforts will appear ? '' .
i have been bold enough to try their first question , i shall not be foolish enough to essay the second .
i shall , however , offer a hope , and that is this : that in the future this might be seen a the time when scientists began to take seriously the idea that it was important to consider the universe as a whole and science as a unity , the time when they began to take seriously the search for a law of how the universe started , began to work out its implications for science generally , and began to discard the remainder of our excess baggage .
the author is grateful for the support of the john simon guggenheim foundation and the national science foundation ( nsf grant phy 85 - 06686 ) during the preparation of this essay .
he would also like to thank the department of applied mathematics and theoretical physics , cambridge university for hospitality while it was being written .
a number of developments in physics are considered from this point of view by the philosopher c.a .
hooker in _
bell s theorem , quantum theory , and conceptions of the universe _
, ed . by m. kafatos ( kluwer , boston , 1989 ) . for clear statements of this point of view
see f. london and e. bauer , _ la thorie de lobservation en mcanique quantique _ , hermann , paris ( 1939 ) ; r.b .
peierls , in _
symposium on the foundations of modern physics _
, ed . by p. lahti and p. mittelstaedt , world scientific , singapore ( 1985 ) . some have seen this division as motivation for restricting the domain of application of quantum mechanics and introducing new non - linear laws which apply in a broader domain .
examples are wigner s discussion of consciousness [ e.p .
wigner in _ the scientist speculates _
, ed . by i.j .
good , basic books , new york ( 1962 ) ] and penrose s discussion of gravitation [ r. penrose , in _ quantum concepts of space and time _ , ed . by c. isham & r. penrose , oxford university press , oxford ( 1986 ) ] .
the original paper is h. everett , _ rev .
* 29 * , 454 ( 1957 ) .
there is a useful collection of papers developing the everett approach in b. dewitt and n. graham , _ the many worlds interpretation of quantum mechanics _
, princeton university press , princeton ( 1973 ) .
a lucid exposition is in r. geroch , _ nos _ * 18 * 617 ( 1984 ) .
even a casual inspection of these references reveals that there is considerable diversity among the various everett points of view .
the arrow of time in quantum mechanics is discussed in many places .
see , for example , y. aharonov , p. bergmann and j.l .
lebovitz , 1410 , 1964 and r. penrose in _ general relativity : an einstein centenary survey _ , ed . by s.w . hawking and w. israel ( cambridge university press , cambridge , 1979 ) . for reviews
see j.a .
wheeler , in _ problemi dei fondamenti della fisica _ , _ scuola internazionale di fisica `` enrico fermi '' _ , corso 52 , ed .
by g. toraldo di francia , north holland , amsterdam ( 1979 ) ; k. kucha , in _ quantum gravity 2 _ , ed . by c. isham , r. penrose , and d. sciama , clarendon press , oxford , p. 329
ff ( 1981 ) .
for recent views see the articles and discussion in @xcite .
few of the proposals currently under investigation as candidates for the conditions of the universe genuinely have the character of `` initial '' conditions in the sense of posing conditions at a definite moment of time . indeed ,
characteristically , as described in section v , the notion of a single time breaks down in the early universe .
rather these proposals fix something like the quantum state of the universe for all times . but
also typically these proposals imply simplicity in the early universe and a global spatial structure which consists of a single inflationary bubble .
the no boundary " proposal @xcite is a case in point .
the point of the remarks in section vi.a is that there may be other possibilities .
for reviews of these efforts from the author s perspective see j.b .
hartle in _ proceedings of the international conference on general relativity and cosmology _ , goa india , 1987 , ( cambridge university press , cambridge,1989 ) ; and in _ proceedings of the 12th international conference on general gravitation _ , boulder , colorado , 1989 ( cambridge university press , cambridge , 1989 ) . for
a bibliography of papers on this subject as of 1991 see j.j .
halliwell at the end of his lectures in _ quantum cosmology and baby universes : proceedings of the 1989 jerusalem winter school for theoretical physics _ , eds .
s. coleman , j.b .
hartle , t. piran , and s. weinberg , world scientific , singapore ( 1991 ) pp .
159 - 243 .
wheeler has for many years stressed the mutability " of the laws of physics from a viewpoint which would entail much more fundamental revisions in physics than those suggested here but yet which has something in common with them .
see , j.a . wheeler in _ foundational problems in the special sciences _
, ed . by r.e .
butts and k.j .
hintikka , d. reidel , dordrecht ( 1977 ) and j.a .
wheeler , _
ibm j. res .
dev . _ * 32 * , 4 ( 1988 ) . for a general discussion of the status of the laws of physics and references to other points of view
see , p.c.w . davies
what are the laws of nature " in _ the reality club no .
2 _ , ed . by j.
brockman , lynx communications , new york ( 1989 ) .
as discussed by h. zeh in _ die physik der zeitrichtung _ , _
springer lecture notes in physics _ * 200 * , springer , heidelberg ( 1984 ) [ translated and updated as _ the physical basis of the direction of time _
, springer - verlag , berlin , 1992 ] .
he attributes the remark to e. borel , _ le hasard _ , alcan , paris , ( 1924 ) . | many advances in physics have in common that some idea which was previously accepted as fundamental , general , and inescapable was subsequently seen to be consequent , special , and dispensable .
the idea was not truly a general feature of the world , but only _ perceived _ to be general because of our special place in the universe and the limited range of our experience .
it was _ excess baggage _ which had to be jettisoned to reach a more a more general perspective .
this article discusses excess baggage from the perspective of quantum cosmology which aims at a theory of the universe s quantum initial state .
we seek to answer the question ` which features of our current theoretical framework are fundamental and which reflect our special position in the universe or its special initial condition ? '
past instances of cosmological excess baggage are reviewed such as the idea that the earth was at the center of the universe or that the second law of thermodynamics was fundamental .
examples of excess baggage in our current understanding are the notion that measurement is central to formulating quantum mechanics , a fundamental quantum mechanical arrow of time , and the idea that a preferred time is needed to formulate quantum theory .
we speculate on candidates for future excess baggage . |
henkin was an extraordinary insightful professor with a talent for exposition , and he devoted considerable effort to writing expository papers .
i will mention three of them , while trying to convince you to read them with your students as a source of mutual inspiration .
this is the title of a wonderful expository paper @xcite , which leon henkin published in _ science _ in 1962 , and subtitled : _ an old thesis of russell s is reexamined in the light of subsequent developments in mathematical logic .
_ i recommend that you to give this paper to your students , not only because the historical view provided is comprehensive and synthetic but also because it shows the henkin s characteristic style ; namely , the ability to strongly catch your attention from the start .
_ how does he achieve it ?
_ you might wonder . in that particular
paper , henkin tells us that his interest in logic began at the age of 16 , when _ ` ( i ) came across a little volume of bertrand russell entitled * mysticism and logic * ' .
_ in the introduction henkin cites russell s radical thesis , that _ _
` mathematics was nothing but logic ' _ _ together with the companion thesis _ `
that logic is purely tautological ' _ , and he describes the strong reaction against his thesis by the academic community : _ ` aux armes , citoyens du monde mathmatique ! ' _
henkin then devotes the first section of the paper to explain the two main ideas that could help explain how russell arrived at his conclusion .
the first was the lengthy effort to achieve a _
` systematic reduction of all concepts of mathematics to a small number of them ' _ , and the second was _ ` the systematic study by mathematical means of the laws of logic which entered into mathematical proofs ' .
_ henkin relates the work of frege , peirce , boole and schrder , during the second half of the nineteenth century , with the two efforts mentioned above , and identifies them as the primary _
raison detre _ of _ principia mathematica .
_ in the following section , entitled _ from russell to gdel _ , henkin explains the introduction of semantic notions by tarski , as well as the formulation and proof of the completeness theorem for propositional logic by post and for first - order logic by gdel .
_ ` this result of gdel s is among the most basic and useful theorems we have in the whole subject of mathematical logic ' . _ but , henkin also explains how , in 1931 , the hope of further extension of this kind of completeness was _ _
` dashed by gdel himself [ ... ] ( he ) was able to demonstrate that the system of * principia mathematica * , taken as a whole was incomplete ' . _ _ immediately after , and anticipating what the reader might be thinking , henkin dispels the hope of finding new axioms to repair the incompleteness phenomenon . in the section
entitled _ consistency and the decision problem _ , henkin analyzes these important notions and also explains how _ `
gdel was able to show that the questions of consistency and completeness were very closely linked to one another .
[ ... ] if a system such as the * principia * were truly consistent , then in fact it would not be possible to produce a sound proof of this fact ! ' .
_ in the following section , named _ logic after 1936 _ , henkin describes how alonzo church proved that no decision procedure is available for first - order logic , and he devotes the rest of the paper to set theory , recursive functions , and algebraic logic .
henkin ends the paper with a section where he analyzes _
russell s thesis in perspective_. _
_ henkin was awarded the _ chauvenet prize _ in 1964 for this paper .
the prize is described as a _ mathematical association of america _ award to the author of an outstanding expository article on a mathematical topic by a member of the association . in april 1 , 1963 , henkin received a very interesting letter from bertrand russell . in it , russell thanked henkin for _ ` your letter of march 26 and for the very interesting paper which you enclosed . ' _ right at the beginning russell declared : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ it is fifty years since i worked seriously at mathematical logic and almost the only work that i have read since that date is gdel s .
i realized , of course , that gdel s work is of fundamental importance , but i was pussled by it .
it made me glad that i was no longer working at mathematical logic .
if a given set of axioms leads to a contradiction , it is clear that at least one of the axioms is false .
does this apply to school - boysarithmetic , and if so , can we believe anything that we were taught in youth ?
are we to think that 2 + 2 is not 4 , but 4.001 ?
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ he then went on explaining his _ ` state of mind ' _ while whitehead and he were doing the principia and added : _ ` both whitehead and i were disappointed that the principia was almost wholly considered in connection with the question whether mathematics is logic . ' _
russell ended the letter with a request : _ `
if you can spare the time , i should like to know , roughly , how , in your opinion , ordinary mathematics or , indeed , any deductive system is affected by gdel s work . ' _ according to annellis : _
` henkin replied to russell at length with an explanation of gdel s incompleteness results , in a letter of july 1963 , specifically explaining that gdel s showed , not the inconsistency , but the incompleteness of the [ principia ] system . ' _ in a personal communication henkin affirmed that _ on mathematical induction _
@xcite , published in 1960 , was the favorite among his articles because it had a somewhat panoramic nature and was not directed exclusively to specialists .
he wrote : _ _ [ ... ] _ but my little paper on induction models from 1960 , which has always been my favorite among my expository papers_. in it , the relationship between the induction axiom and recursive definitions is studied in depth .
_ why do i so strongly recommend that you ask your students to read this paper ?
_ from my point of view , it is the best paper on logic to offer students as a first reading of a real - life article .
the paper is especially interesting because henkin describes something that would never appear in a formal article : his motivation .
it seems that henkin was trying to convince a mathematical colleague about why a given argument about the existence of recursive operations was completely wrong , even though at first sight it might seem convincing . before going into the details of the wrong argument that motivated the whole paper , henkin explains peano arithmetic .
peano axiomatized the theory of natural numbers . to do so , he started out from indefinable primitive terms in particular , those of natural number , zero , and the successor function- and by means of three axioms he synthesized the main facts . among those axioms is that of induction , which states that any subset of natural numbers , closed by the successor operation and to which zero belongs , is precisely the set of all natural numbers .
although the axioms for the theory of natural numbers are very important , the most interesting theorems of the theory did not stem from them alone because in most of the theorems , operations of addition , multiplication , etc .
are used .
peano thought that after axiomatizing a theory it did not suffice to organize the facts by means of axiomatic laws ; it was also necessary to organize the concepts , using definition laws .
in particular , we can define addition recursively by means of : 1 .
@xmath0 2 .
@xmath1 , for all @xmath2 of @xmath3 however , this definition must be justified by a theorem in which the existence of a unique operation that will satisfy the previous equations must be established . a poor argument to prove this is as follows
let us choose any @xmath4 .
we define a subset @xmath5 of @xmath3 , by placing all @xmath6 elements in @xmath5 for which @xmath7 is defined by the previous equations .
it is not difficult to see that @xmath8 and that @xmath9 . now using the induction axiom ,
we conclude that @xmath10 and thus that @xmath7 is defined for all @xmath6 . _
why is this proof wrong ?
_ this was the question that henkin s colleague posed him .
henkin tried to convince him that because the argument was designed to establish the existence of a function @xmath11 ( @xmath12 in the example ) , it is incorrect to assume in the course of the argument that we have such a function .
henkin made him see that in the proof only the third axiom was used and that , if correct , the same reasoning could be used not only for models that satisfy all peano axioms but also for those that satisfy only the induction one .
henkin called these induction models and proved that in them not all recursive operations are definable . for example
, exponentiation fails .
induction models turn out to have a fairly simple mathematical structure : there are standard ones that is , isomorphic to natural numbers but also non - standard ones .
the latter also have a simple structure : either they are cycles , in particular @xmath13 modulo @xmath14 , or they are what henkin calls spoons because they have a handle followed by a cycle .
the reason is that the induction axiom is never fulfilled alone , since it requires peano s first or second axiom .
this does not mean that peano s axioms are redundant , as it is well known that they are formally independent ; _ i.e. _ , each one is independent of the other two .
if you take a look at the list of documents leon henkin left us , the first published paper , _ the completeness of first order logic _
tiposhenkin:49 , corresponds to his well known result , while the last , _ the discovery of my completeness proof _ @xcite , _ _ is a extremely interesting as autobiography , thus ending his career with a sort of fascinating loop .
_ i claim that reading the last paper is a must .
_ as you know , leon henkin left us an important collection of papers , some of them so exciting as his proof of the completeness theorem both for the theory of types and for first - order logic .
he did so by means of an innovative and highly versatile method , which was later to be used in many other logics , even in those known as non - classical . in his 1996 paper , we learn about the process of discovery , which observed facts he was trying to explain , and why he ended up discovering things that were not originally the target of his enquiries .
thus , in this case we do not have to engage in risky hypotheses or explain his ideas on the mere basis of the later , cold elaboration in scientific articles .
it is well known that the _ logic of discovery _ differs from what is adopted on organizing the final exposition of our research through their different propositions , lemmas , theorems and corollaries .
we also learn that the publication order of his completeness results ( tiposhenkin:49 and @xcite ) is the reverse of his discovery of the proofs .
the completeness for first - order logic was accomplished when he realized he could modify the proof obtained for type theory in an appropriate way .
we consider this to be of great significance , because the effort of abstraction needed for the first proof ( that of type theory ) provided a broad perspective that allowed him to see beyond some prejudices and to make the decisive changes needed to reach his second proof .
in manzanocomphenkin2014 you can find a detailed commentary of henkin s contribution to the resolution and understanding of the completeness phenomena . in 1967
henkin published two very relevant expository papers on for the subject we are considering here , _ truth and provability _ and _ completeness _ , which were published in _ philosophy of science today _
morgenbesser1967 .
[ [ truth - and - provability ] ] _ truth and provability _ + + + + + + + + + + + + + + + + + + + + + + + in less than 10 pages , henkin gives a very intuitive introduction to the concept of truth and its counterpart , that of provability , in the same spirit of tarski s expository paper _ truth and proof _ @xcite .
the latter was published in _
scientific american _ two years after henkin s contribution .
this not so surprising as henkin had by then been in berkeley working with tarski for about 15 years and the theory of truth was tarski s contribution .
the main topics henkin introduces ( or at least touches upon ) are very relevant .
they include the _ use / mention _ distinction , the desire for _ languages with infinite sentences _ and the need for a _ recursive definition of truth _ , the _ language / metalanguage _ distinction , the need to avoid reflexive paradoxes , the concept of _ denotation _ for terms , and the interpretation of _ quantified formulas_. he also explains what an _ axiomatic _ _
theory _ is and how it works in harmony with a _ deductive calculus_. properties such as _ decidability _ and _ completeness / incompleteness of a theory _ are mentioned at the end .
i admire the way these concepts are introduced , with such lan , and the chain henkin establishes , which shows how each concept is needed to support the next .
[ [ completeness-1 ] ] _ completeness _ + + + + + + + + + + + + + + in this short expository paper henkin explores the complex landscape of the notions of completeness .
he introduces the notion of logical completeness both weak and strong as an extension of the notion already introduced of completeness of an axiomatic theory .
this presentation differs notably from the standard way these notions are introduced today where , usually , the completeness of the logic precedes the notion of completeness of a theory and , often , to avoid misunderstandings , both concepts are separated as much as possible , as if relating them were some sort of terrible mistake or even anathema .
gdel s incompleteness theorem is presented , as well as its negative impact on the search for a complete calculus for higher - order logic .
the paper ends by introducting his own completeness result for higher - order logic with general semantics .
the utilitarian way henkin uses to justify his general models as a way of sorting the provable sentences from the unprovable ones in the class of valid sentences ( in standard models ) is very peculiar .
henkin was often described as a social activist , he labored much of his career to boost the number of woman and underrepresented minorities in the upper echelons of mathematics .
he was also very aware that we are beings immersed in the crucible of history from which we find it hard to escape , an awareness he brought to the very beginning of his interesting article about the teaching of mathematics @xcite : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ waves of history wash over our nation , stirring up our society and our institutions .
soon we see changes in the way that all of us do things , including our mathematics and our teaching .
these changes form themselves into rivulets and streams that merge at various angles with those arising in parts of our society quite different from education , mathematics , or science .
rivers are formed , contributing powerful currents that will produce future waves of history .
the great depression and world war ii formed the background of my years of study ; the cold war and the civil rights movement were the backdrop against which i began my career as a research mathematicians , and later began to involve myself with mathematics education.(in @xcite page 3 ) _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ in this paper he gave both a short outline of the variety of educational programs he created and/or participated in , and interesting details about some of them .
in particular he discussed the following six : 1 .
1957 - 59 , * nsf * summer institutes .
the national science foundation is an independent federal agency created by congress in 1950 . as you can read in their web page , _
http://www.nsf.gov/about/_ , its aim was _ ` to promote the progress of science ; to advance the national health , prosperity , and welfare ; to secure the national defense '_. nowadays , nsf is the _ ` only federal agency whose mission includes support for all fields of fundamental science and engineering , except for medical sciences . ' _
nsf s strategic plan includes investing in science , engineering , and education for the nation s future . in @xcite henkin
related this initiative to historical facts : _ ` the launching of sputnik demonstrated superiority in space travel , and our country responded in a variety of ways to improve capacity for scientific and technical developments ' _ in 1957 , henkin involved himself in several nsf s programs , he served as a lecturer of several courses .
these programs were designed to improve high school and college mathematics instruction and were directed to mathematics teachers .
henkin explained that the variety of attitudes toward mathematics of the teachers attending the courses was amazing , and that the experience gave him a view of the nature of instruction around the country .
the subject of his courses was the _
axiomatic foundation of number systems_. one of his aims was to get students to understand the idea of a proof because he believed that it could help students in the effort of finding proofs of their own , in a much better way than the mere understanding of the steps that constitutes a proof .
2 . 1959 - 63 , * maa * math films .
the * * m**athematical * * a**ssociation of * * a**merica was established one century ago , in 1915 . as you can read in their web page , _
http://www.maa.org/_ , _
` over our first century , maa has certainly grown , but continues to maintain our leadership in all aspects of the undergraduate program in mathematics ' .
_ long before internet resources became available , the maa made movies . as henkin said : _ sensing a potential infusion of technology into mathematics instruction , maa set up a committee to make a few experimental films . _
[ ... ] _ the committee approached me in 1959 - 60 with a request to make a filmed lecture on mathematical induction which could be shown at the high - school - senior / college - freshman level .
i readily agreed. _
the film was part of the mathematics today series , and was shown on public television in new york city and in high schools . in @xcite henkin explained the preparation of the film , both from a technical point of view and from a methodological and pedagogical perspective .
he attributed the lack of understanding of the induction principle at the undergraduate level to the current formulation as a mathematical principle , and he proposed to use it as _ ` a statement about sets of numbers satisfying two simple conditions ; formulated in this way , it is a fine vehicle for giving students practice in forming and using sets of numbers to show that all natural numbers possess various properties ' _ 3 .
1961 - 64 cupm .
the mathematical association of america s * * c**ommittee on the * * u**nder- graduate * * p**rogram in * * m**athematics ( cupm ) is charged with making recommendations to guide mathematics departments in designing curricula for their undergraduate students . in the sixties ,
the cupm proposed courses to be taken by elementary teachers . in @xcite henkin said _ ` some of my colleagues and i began , for the first time , to have classroom contact with prospective elementary teachers , and that led , in turn , to in - service programs for current teachers .
i learned a great deal from teaching teachers - students ; i hope they learned at least half as much as i ! '
activities to broaden opportunity .
the sixties is the term used to describe the counterculture and revolution movement that took place in several places in the u.s.a . and europe .
berkeley students were taking energic actions against segregation in southeastern u.s.a . as well as against military actions in vietnam .
in @xcite henkin said _ in the midst of this turmoil i joined in forming two committees at berkeley which enlarged the opportunity of minority ethnic groups for studying mathematics and related subjects .
_ [ ... ] _ we noted that while there was a substantial black population in berkeley and the surrounding bay area , our own university student body was almost lily white and the plan to undertake action through the senate was initiated _ in 1964 , leon henkin and jerzy neyman , a world - famous polish - american statistician from berkeley university , started a program at berkeley to increase the number of minority students entering college from bay area high schools .
henkin told us that the inicitiave came after neyman participation in _ ` the maa s visiting mathematician program in fall 1963 .
he lectured in southerns states where , by law , whites and blacks studied in separate colleges . upon returning to berkeley he told some of his friend that first - rate students were being given a third - rate education ' _ henkin and neyman undertake actions through the senate , and in 1964 the senate established a committee with the desired effect .
the committee recruited promising students and offered them summer programs to study mathematics and english .
if they persisted in the program , they were offered special scholarships . in the same year , 1964 , henkin heard a talk by a berkeley high school teacher , bill johntz .
after that , henkin was invited to see him in action , while he was teaching mathematics to elementary students from low - income neighborhoods , and realized that johntz was able to raise great enthusiasm in the class .
significantly , students enjoyed and actively engaged in the process of learning , and they became integrally involved in their own education .
he was using a socratic group - discovery method modeled after the filmed teaching of david page , a university of illinois mathematics professor .
the method was working well , and they recruited university mathematics students as well as engineers as teachers , after some training .
the program was called project _ seed _ special elementary education for the disadvantaged .
this program is still alive , as you can see in their web page , _
http://projectseed.org/. _ 5 .
1960 - 68 . teaching teachers , teaching kids . in this paper henkin
described several conferences on school mathematics as well as several projects and courses he was involved in .
the following paragraph caught my eye : _ ` after i began visiting elementary school classes in connection with ctfo , i came to believe that the emotional response of the teachers to mathematics was of more importance to the learning process of the students than the teacher s ability to relate the algorithms of arithmetic to the axioms of ring theory ' . _
1968 - 70 .
open sesame : the lawrence hall of science .
the lawrence hall of science , a science museum in berkeley , was created in honor of the 1939 novel prize winner ernest orlando lawrence . as you can read in their web page,@xmath15__`we
have been providing parents , kids , and educators with opportunities to engage with science since 1968 . ' _ _ + according to henkin s tale _ ` in 1968 , the newly appointed director , professor of physics alan portis , decided to transform the museum into a center of science and mathematics education , whose functions would be integrated with graduate research programs directed by interdisciplinary group of faculty . '
_ to help in his endeavor , he gathered a group of faculty from a variety of science departments interested in science education . _ ` these faculty members proposed a new , interdisciplinary ph.d program under the acronym sesame special excellence in science and mathematics education .
entering students were required to have a masters degree in mathematics or in one of the sciences .
courses and seminars in theories of learning , cognitive science , and experimental design were either identified in various departments , or created ' .
_ nitsa movshovitz - hadar , a student from the technion in israel , was admitted in the sesame program , she wrote her thesis under the direction of leon henkin .
nitsa is one of the contributors of the book , _ the life and work of leon henkin : essays on his contributions_. 99 anellis , irving .
review of bertrand russell , towards the principles of mathematics , 1900 - 02 , edited by gregory h. moore , and bertrand russell , foundations of logic , 1903 - 05 edited by alasdair urquhart with the assistance of albert c. lewis .
mod . log . 8
( 2001 ) , no . 3 - 4 , 5793 .
http://projecteuclid.org/euclid.rml/1081173771 .
henkin , l. : the roles of action and of thought in mathematics education one mathematician s passage . in : fisher , n. d. , keynes , h. b. , wagreich , ph .
d. ( eds . ) changing the culture : mathematics education in the research community .
cbms issues in mathematics education , vol . 5 , pp . 3 - 16 .
american mathematical society in cooperation with mathematical association of america , providence ( 1995 ) manzano , m. henkin on completeness . in : manzano ,
m. et als ( eds . ) .
_ the life and work of leon henkin : essays on his contributions_. pp .
149 - 175 .
studies in universal logic .
springer international publishing .
switzerland ( 2014 ) | this paper is divided into two sections . in the first
i give reasons for strongly recommending reading some of henkin s expository papers . in the second
i describe leon henkin s work as a social activists in the field of mathematics education , as he labored in much of his career to boost the number of women and underrepresented minorities in the upper echelons of mathematics . |
suppose we have a numerical model , @xmath0 , for a dynamical process : @xmath1 where @xmath2 is the state at time indexed by @xmath3 .
then it is almost certain that such a numerical model @xmath0 will contain errors .
a large amount of uncertainty quantification has revolved around providing bounds of the accuracy of the numerical scheme used to approximate the underlying mathematical or statistical problem @xcite .
this ignores one large issue : the underlying mathematical problem does not represent all of the features of the real world system in which one is interested . for example , when modelling wind around a wind farm , drag induced by individual trees / plants will likely not be included . instead , some approximate homogenised quantity will be used . as another example , when modelling carbon fibre composites , specific manufacturing defects will not be included in the model until measurements of the materials response under various loading conditions are incorporated .
the ubiquitous quote on this issue is from @xcite : `` all models are wrong but some are useful '' . the problem we address is how to estimate the distribution of the errors that are made by the mathematical model in simulating a physical system . to get a measure of
how far a mathematical model is deviating from the real world , it is necessary to use observations from the system as an independent source of information .
typically we shall never be able to fully observe a system ( in time and/or space ) and so one has to be careful how to compare a small set of observations with the larger model of the system . the theoretical framework to do this
rigorously is well established : using data assimilation to numerically implement bayes theorem ( * ? ? ? * for example ) .
a physical system that occurs in nature has only one realisation .
we refer to that realisation as the _ truth _ and denote it @xmath4 .
if we use such a truth with the numerical model of the system we have the following evolution equation , @xmath5 here @xmath6 is the model error term which is a realisation of a stochastic process which occurs at time @xmath3 .
we wish to estimate , through the use of data assimilation , the properties of the distribution that @xmath6 follows .
the first two moments of this unknown distribution , its mean and covariance , are denoted @xmath7 and @xmath8 respectively .
knowledge of @xmath6 can be fed back into the dynamical model @xmath0 as a way to systematically improve the model .
for example , directly subtracting @xmath7 from the model would be an example of a bias correction method ( e.g. * ? ? ? * ) , whereas the sources of these errors can be attributed to components of the model equations @xcite and thus can be corrected to improve the mathematical model .
we use any or all prior information we have , along with observations of the system to obtain a _
`` best estimate '' _ of the truth .
this is the data assimilation stage and results in what is referred to as the _
`` analysis '' _ , @xmath9 .
there are many different techniques for data assimilation , each which make various assumptions or simplifications to bayes theorem .
see @xcite or @xcite for an overview of the field .
once we have @xmath9 we define our approximation to the model error at timestep @xmath3 , @xmath10 , with the equation @xmath11 this makes no assumptions on how @xmath9 is calculated .
however , we can immediately see that one traditional method of data assimilation is inappropriate to estimate the @xmath6 term . to estimate @xmath6 , @xmath9
should not be obtained from strong constraint 4dvar ( sc-4dvar ) . in sc-4dvar the analysis at time @xmath12 , @xmath13 is given as @xmath14 here , @xmath15 is a background estimate for the state at time @xmath12 , assumed to be gaussian with background error covariance matrix @xmath16 .
observations of the system @xmath17 are taken at @xmath18 separate times @xmath19 .
the observation is related to the system through the observation equation @xmath20 where the observation errors @xmath21 are assumed to be gaussian with an observation error covariance matrix @xmath22
. then at subsequent times in the assimilation window , @xmath23 and thus @xmath24 .
having an estimate for @xmath8 is important as it plays a vital role in many modern data assimilation methods . in the variational method of data assimilation known as weak constraint 4dvar ( wc-4dvar ) , in addition to the background and observation terms of sc-4dvar
, a third term is introduced to penalise the addition of model error @xcite , which is implicitly assumed to follow a gaussian distribution .
@xmath25 in this formulation , @xmath26 is the initial state which the method seeks to find , along with model error terms @xmath27 .
the first term penalises the difference in the initial state @xmath28 and an initial guess known as the background term @xmath15 , weighted by the inverse of the background error covariance matrix @xmath16 .
the second term penalises discrepancies between observations of the system @xmath29 and the model equivalent of the observations at the corresponding time , @xmath30 , all weighted by the appropriate observation error covariance matrix @xmath22 .
this notation hides the fact that @xmath31 etc .
the final term penalises the introduction of model error at each model timestep @xmath32 , weighted by the model error covariance matrix @xmath8 . in particle filters , such as the equivalent weights particle filter @xcite , @xmath8
is used as part of the proposal density to bring the ensemble closer to the observations .
it also appears in the computation of weights associated with each particle .
@xcite showed that the lack of information regarding @xmath8 is the limiting factor to applying particle filters in large - scale geophysical systems .
much of the work to estimate @xmath8 has been undertaken at the european centre for medium - range weather forecasting ( ecmwf ) in the context of implementing a form of wc-4dvar for operational numerical weather forecasting .
the lack of knowledge of @xmath8 is one of the main reasons that such weak - constraint methods are not operational .
@xcite notes `` weak - constraint 4d - var has never been implemented fully with a realistic forecast model because of the computational cost and because of the lack of information to define the model - error covariance matrix required to solve the problem '' .
it has been shown @xcite that for the atmospheric case , the background error covariance matrix ( @xmath16 ) is not successful as an approximation to the model - error covariance matrix ( @xmath8 ) . as an _ ad hoc _ approximation to @xmath8 , @xcite suggests the following . at time @xmath3 , get an ensemble of analysis states @xmath33 . as each ensemble member should be indistinguishable from the truth , each forecast @xmath34
could be considered a `` possible evolution of the atmosphere from the true state '' . from this , trmolet concluded that the differences in these forecast increments could be interpreted as `` an ensemble of possible realisations of model error '' . the real world , however , has only one realisation .
hence one needs to compare these forecasts not with themselves , but with the observations instead . in this way
, one discovers not just where the model diverges based on its initial conditions , but where it has failed to capture the realisation of the underlying stochastic dynamics . a smoother is a sequential method of data assimilation that gives an estimate of @xmath9 for all @xmath3 , not simply at timesteps where observations are present .
recently , the concept of using a forecast from a lag-1 smoother estimate to approximate the model error covariances has been investigated @xcite .
this work only considers estimating the @xmath8 the problem in observation space , and do not look at the full information available to them in state space .
this is physically motivated as in the case where only the observed variables of the dynamical system are well constrained , only the projection of @xmath8 into the observation space will be accurate .
such derivations from a smoother trajectory have been investigated by @xcite in the context of parameter and parameterization estimation .
when the model error can be attributed to incorrect coefficients in the formulation of the model equations then it has been shown that the first moment of the derived model error can be used to determine the necessary corrections to the parameters used in the numerical model .
this can be extended to estimation of parameterizations when the first moment of the model error is regressed onto different functions of the model state vector .
however higher order moments have not been investigated .
we now set out on the main path of this paper : to find a bound on the error in the approximation of @xmath7 and @xmath8 in terms of the accuracy of @xmath9 in approximating @xmath4 .
the main assumption we shall make is that the dynamical model @xmath0 is ( lipschitz ) continuous .
if the model were not continuous at @xmath35 then regardless of how close we approximate this argument , the resulting model trajectory could be wildly different .
this would lead to a poor approximation of the model error .
[ thm : beta]suppose @xmath0 is lipschitz with constant @xmath36 and @xmath37
. then @xmath38 .
@xmath0 lipschitz @xmath39 s.t . @xmath40 .
@xmath41 hence @xmath42 as @xmath43 then latexmath:[\ ] ] it is now possible to obtain a bound on the error in the covariance of the model error distribution in terms of the analysis error . for clarity
, the notation @xmath73 also refers to the @xmath57 component of a vector that already has a subscript .
[ thm : q ] suppose @xmath0 is lipschitz with constant @xmath36 .
suppose further that , for all @xmath3 , the analysis satisfies @xmath74 and @xmath75 then the corresponding estimated sample error covariance satisfies @xmath65 . by theorem [ thm : beta ] , if @xmath74 then @xmath76 similarly if @xmath77 then @xmath78 by corollary [ cor : mu ] if @xmath79 then @xmath80 and if @xmath81 then @xmath82 hence we can use lemma [ lemma : q ] . thus @xmath65 .
corollary [ cor : mu ] shows that if one desires to know the mean of the model error to an accuracy of @xmath83 , then the analysis of that variable must be within @xmath84 of the truth at every timestep .
contrast this with the result from theorem [ thm : q ] : to estimate the variance of a variable to within @xmath83 , the analysis of that variable must be within @xmath85 of the truth at each timestep
. with the factor of @xmath86 in the denominator , this can be seen as approximately an order of magnitude more accuracy needed in the analysis to achieve the same accuracy in the approximation . moreover , the accuracy in the variance depends on the mean value of that variable .
the larger the mean value , the more accuracy required in the analysis to give the same error in the variance .
this may be more easily understood conversely : the larger the mean value in a variable , the larger the error in the estimation of its variance .
in this section we apply the model error moment estimation method as detailed in the previous sections to the lorenz 96 system @xcite .
this is a very simple example to show numerically that the errors in the approximations of @xmath7 and @xmath8 are dependent on the quality of the analysis @xmath9 .
the lorenz 96 system is given by the following ode : @xmath87 where @xmath88 with cyclic boundary conditions .
we choose @xmath89 .
we prescribe the truth to evolve as with @xmath90 where @xmath91 and @xmath92 where @xmath93 @xmath94 and @xmath95 have been chosen to be _ interesting _ and @xmath95 in particular , for implementation purposes , has a rather simple symmetric square root .
we take observations of the true state at every timestep .
the observations are related to the state via the equation @xmath96 here we consider @xmath97 and will show two different experiments with observations of varying accuracy .
the first experiment will have @xmath98 , and represents very informative observations .
the second experiment will have @xmath99 , and represents observations containing less information . to get an analysis @xmath9
, we perform a simple 3dvar at each timestep .
that is , @xmath100 we use @xmath101 as a particularly uninformative prior .
the model is evolved for 3000 timesteps , giving us 2999 samples of @xmath10 .
figure [ fig : hov ] shows hovmller diagrams of the analysis error as a result of the data assimilation process .
the higher precision observations lead to significantly lower estimation errors . using these analyses , estimates for @xmath102 and @xmath103 are calculated .
results for @xmath104 are shown in figures [ fig : mu_8 ] to [ fig : qe_8 ] and for @xmath105 are shown in figures [ fig : mu_3 ] to [ fig : qe_3 ] .
these results show that the second order moment of model error , @xmath8 , is much more sensitive to the quality of the analysis than the first order moment , @xmath7 .
this is consistent with the theory set out in the previous section .
0.48 for the two different precision observations .
observe that the 3dvar algorithm can reduce the error in the state by @xmath106 orders of magnitude when the observations are more precise . ]
0.48 for the two different precision observations .
observe that the 3dvar algorithm can reduce the error in the state by @xmath106 orders of magnitude when the observations are more precise . ] 0.48 , title="fig : " ] 0.48 , title="fig : " ] 0.33 , scaledwidth=113.0% ] 0.33 , scaledwidth=113.0% ] 0.33 , scaledwidth=113.0% ] 0.33 , scaledwidth=113.0% ] 0.33 , scaledwidth=113.0% ] 0.33 , scaledwidth=113.0% ] 0.48 , title="fig : " ] 0.48 , title="fig : " ] 0.33 , scaledwidth=113.0% ] 0.33 , scaledwidth=113.0% ] 0.33 , scaledwidth=113.0% ] 0.33 , scaledwidth=113.0% ] 0.33 , scaledwidth=113.0% ] 0.33 , scaledwidth=113.0% ]
this paper has considered using the problem of estimating the moments of model error of a dynamic model .
the model error is approximated by the difference of the analysis and the model forecast of the analysis at the previous timestep .
bounds for the errors in both the first and second moment of the approximated model error compared to the sample model error are derived , in terms of the error in the analysis from the truth .
it is shown that to achieve the same error estimation in the second moment compared with the first , the analysis must be an order of magnitude closer to the truth .
numerical experiments were conducted to elucidate how the quality of the analysis affects the estimation of both the mean and the covariance of the model error .
it is shown numerically with the lorenz 96 system that the estimation of the covariance of the model error is much more sensitive to the quality of the data assimilation than the estimation of the mean of the model error . | using a dynamical model to make predictions about a system has many sources of error .
these can include errors in how the model was initialised but also errors in the dynamics of the model itself .
for many applications in data assimilation , probabilistic forecasting , or model improvement , these model errors need to be known over the timestep of the model , not over a time - averaged period . using a forecast from a state that combines observational information as well as prior information we can gain an approximation to the statistics of the model errors on the timescale of the model that is required . here
we give bounds on the errors in the estimation of the mean and covariance of the errors in the model equations in terms of the errors made in the state estimation .
this is the first time that such a result has been derived .
the result shows to what extent the state estimation must constrain the analysis in order to obtain a specified error on the mean or covariance of the model errors .
this is particularly useful for experimental design as it indicates the necessary information content required in observations of the dynamical system .
* keywords : * model error estimation , data assimilation , analysis forecast , model error covariance estimation |
ever since the realization @xcite , shortly after the discovery of the heterotic string @xcite , that it provides a promising venue wherein the standard model may be embedded , string phenomenology was born .
indeed , the @xmath0 gauge group of the heterotic string encompasses a natural gauge unification of the standard model @xmath1 group .
a mathematically succinct and physically appealing approach was thereby engendered , where the @xmath0 string was compactified on a calabi - yau threefold @xmath2 .
the tangent bundle @xmath3 of @xmath2 admits , by the calabi - yau nature , an @xmath4 connection and @xmath0 is thus broken to an @xmath5 gut group , the commutant of @xmath4 therein .
this was the initial setup and the @xmath5 particle content is conveniently computed by the cohomology groups taking value in @xmath3 , which , by standard hodge decomposition of khler manifolds @xcite , are simply the hodge numbers of @xmath2 : @xmath6 now , the 27 of an @xmath5 gut theory contains the entirety of the standard model fermions , hence the net number of generations is simply ( the absolute value of ) the difference between the above two terms , which , by rudimentary topology , is half the euler number of @xmath2 .
our desire , is that there be three net generations , in accord with experimental observations .
therefore , one of the first questions in string phenomenology was whether there exists a calabi - yau threefold ( cy3 ) with euler number @xmath7 : @xmath8 homogeneous polynomials in @xmath9 . here , complete intersection means that the dimension of the ambient space exceeds the number @xmath10 of defining equations by precisely 3 , i.e. , @xmath11 .
moreover , the calabi - yau condition of vanishing first chern class of @xmath3 translates to @xmath12 .
subsequently , each manifold can be written as an @xmath13 configuration matrix ( to which we may sometimes adjoin the first column , designating the ambient product of projective spaces , for redundant clarity ) : @xmath14_{m \times k \ , } \quad \begin{array}{l } k = \sum\limits_{r=1}^m n_r-3 \ , \\
\sum\limits_{j=1}^k q^{r}_{j } = n_r + 1 \ , \ \forall \ ; r=1 , \ldots , m \ .
\end{array}\ ] ] the authors of @xcite proved that there can be only a finite number of such matrices and spent a number of years , in the nascent days of computer power , to establish a complete classification .
the most famous example is , of course , the manifold @xmath15_{-200}^{1,101}$ ] ( or simply @xmath16 $ ] ) , known commonly as the * quintic*. we have marked the hodge numbers as superscript and euler number , subscript . unfortunately , none of the 8000 or so cicys has euler number @xmath7 .
nevertheless , it was soon realized by tian and yau that freely - acting quotients of these spaces could have the correct property and the manifold @xmath17 } / \iz_3 $ ] has topological numbers @xmath18 .
it was thus not surprising that this manifold became central to string phenomenology in the early days @xcite .
nowadays , @xmath5 gut theories are less favoured than their @xmath19 or @xmath20 counterparts , which can then be broken to the ( minimally supersymmetric ) standard model ( mssm ) by turning on discrete wilson lines .
furthermore , @xmath18 has 6 entire generations of anti - families , another cumbersome feature .
modern heterotic phenomenology focuses on @xmath19 or @xmath20 guts and their subsequent breaking to the mssm .
now , obtaining these gauge groups is relatively straight - forward : one simply recalls that their commutants in @xmath0 are , respectively , @xmath19 and @xmath21
. therefore , endowing the cy3 @xmath2 with an @xmath22 bundle @xmath23 with @xmath24 will generalize the traditional setup where @xmath25 ( now known as `` standard embedding '' ) to give these gut theories ( now known as `` general embedding '' ) @xcite .
the adjoint 248 of @xmath0 branches accordingly : @xmath26 the associated particle content is likewise captured by the various vector - bundle - valued cohomology groups : @xmath27 next , breaking the gut to the mssm gauge group is also elementary ( what is difficult , as we shall see , is obtaining the precise spectrum and couplings ) : one enriches the above structure of the pair @xmath28 with a discrete wilson line @xmath29 such that the commutant of @xmath29 in @xmath19 or @xmath20 is @xmath1 ( with a possibility of an extra @xmath30 factor ) . on a calabi - yau manifold with
non - trivial ( discrete ) fundamental group @xmath31 , one can turn on a @xmath29-wilson line and compute the @xmath29-equivariant cohomology groups for @xmath23 , in conjunction with the action of the wilson line , to obtain the final particle spectrum .
this has been developed in a programme led by ovrut , donagi , pantev et al.over the past decade : cf .
@xcite ( q.v . brief review in @xcite ) and @xcite .
this search for the mssm , wherein the particle content is encaptured by bundle cohomology , the moduli , bundle endomorphisms , and yukawa couplings , trilinear compositions amongst the cohomology groups , has been a substantial challenge and a healthy dialogue for collaborative efforts between physicists and mathematicians .
a preliminary success of the aforementioned endeavour culminated in 2005 where , sifting through some number of suitable vector bundles over so - called elliptically - fibered cy3s ( another large class of manifolds ) , two complementary heterotic mssm models with no exotic particles , no anti - generations , exactly one pair of higgs and reasonable yukawa couplings were found after years of search .
one @xcite was based on an @xmath21 bundle coupled with a @xmath32 wilson - line , giving rise to , via an @xmath20 gut , the mssm with an @xmath33 , the other @xcite , on an @xmath19 bundle coupled with a @xmath34 wilson - line , giving the mssm via an @xmath19 gut .
both models are based ( respectively by equivariant @xmath32 and @xmath34 actions ) on a parent self - mirror @xmath35 manifold , which is a @xmath36-fibration over a nineth del pezzo surface , which is itself an elliptic fibration over @xmath37 .
remarkably , this @xmath35 covering manifold , from which both quotients descend , resides in the intersection between the database of elliptically fibered threefolds and cicys and it was an astute observation , initially made by candelas , that it is , in fact , intimately related to the cover of the tian - yau manifold : @xmath38 } \
, \quad x_{-18}^{14,23 } = { \left [ \begin{matrix } 1&3&0\\ 1&0&3\\ \end{matrix } \right ] } \ .\ ] ] why indeed should the two most favoured manifolds from two completely different data - sets , spanning over a decade of search , be connected by so simple a transposition of configuration matrices ?
the _ observatio curiosa _ prompted the investigations in @xcite wherein the two manifolds , together with the bi - cubic cicy , @xmath39}$ ] , as well as their various quotients were shown to be conifold transition of each other .
the bundles thereon , were proposed to be linked by so - called * transgressions * , where a bundle on one is taken , being trivial on the conifold blowup @xmath37 cycles , to another bundle on the other after the conifold transition .
indeed , all cicys are related by conifold transitions and it is a conjecture known as * reid s fantasy * that all cy3s are related by blow - up / down transitions generalizing the conifold ; it is thus natural to speculate that all ( stable ) vector bundles on all cy3s should transgress perhaps the plethora of heterotic vacua are all inter - laced after all ! to have a glimpse of this multitude , a plot was made of all the known cy3s to date in @xcite , and in the spirit of @xcite as well as the pioneering work in mirror symmetry by candelas et al .
, the ordinate and abscissa are taken , respectively , as the sum and the difference ( euler number ) of the two hodge numbers .
\(a ) ( horizontal ) versus @xmath40 ( vertical ) of all the some @xmath41 known calabi - yau threefolds ; ( b ) magnification of the bottom corner in ( a).,title="fig : " ] ( b ) ( horizontal ) versus @xmath40 ( vertical ) of all the some @xmath41 known calabi - yau threefolds ; ( b ) magnification of the bottom corner in ( a).,title="fig : " ] this is presented in part ( a ) of figure [ f : cyplot ] .
there are about 500 million known calabi - yau manifolds , including the cicys , the elliptically fibered , isolated examples , etc . ; the largest family by far , however , is the impressive list compiled over the 1990 s , of smooth hypersurfaces in toric four - folds , performed in the lovely work of kreuzer - skarke @xcite , consisting of 473,800,776 inequivalent cy3s , with 30,108 distinct pairs of hodge numbers contributing to figure [ f : cyplot ] ( a ) .
there are certainly many intriguing observations to be made on the plot , such as whether there can exist a cy3 whose euler number exceeds 960 in magnitude .
more immediate to our concern , however , is that the centre of the plot has multiple occupation , that is , each point represents up to hundred of thousands of cy3s . nevertheless , as two decades of construction
have taught us , in conjunction with subsequent work in @xcite , that the bottom corner of the plot is rather sparse . zooming into this *
special corner * , presented in part ( b ) of the figure , shows rather pretty patterns .
importantly , these low - hodge - number manifolds comprises of the few `` good '' examples on which bundles begetting exact mssm spectrum live and mutually transgress .
is this corner an _ oasis _ in the seeming supererogation of string vacua , wherein a world like ours could blossom , while the remaining plenitude are no more than a swap - land @xcite ?
short of a _ selection principle _ , an immediate method of approach to the special corner is not obvious , instead , a synthetic rather than analytic perspective may prove to be conducive .
this has been undertaken over the past few years @xcite , where large classes of bundles are constructed over large data - sets of cy3s and then those with mssm properties , carefully selected .
such a philosophy of algorithmic scan is indeed facilitated by the rapid advances in computational algebraic geometry as well as its implementation on ever - faster machines @xcite ; the cross - pollination of this subject with theoretical and mathematical physics has also recently been a healthy endeavour @xcite .
we therefore wish to explicitly construct a large number of special unitary vector bundles @xmath23 on cy3s ; quite a few physical constraints can instantly be imposed .
[ [ supersymmetry ] ] supersymmetry : + + + + + + + + + + + + + + + we are aiming for the mssm , and hence require @xmath42 susy in the low - energy 4-dimensional theory , which in turn implies that @xmath23 admits a holomorphic connection @xmath43 satisfying the hermitian - yang - mills ( hym ) equations : @xmath44 , a generalization of ricci - flatness for calabi - yau manifolds . solving these pdes
explicitly are currently impossible .
luckily , we are saved by the donaldson - uhlenbeck - yau theorem @xcite , which states that on each ( poly-)stable holomorphic vector bundle , there exits a unique hym connection .
let us not delve into the details of stability , but only emphasize that this can be construed as a purely algebraic condition , without recourse to hard analysis .
some immediate and calculationally important consequences are that the zeroth and highest cohomologies of a stable bundle @xmath23 vanish : @xmath45 ; this , coupled with * serre duality * for any bundle @xmath23 on a cy3 , that @xmath46 , imply the following @xcite : @xmath47 [ [ net - generations ] ] net generations : + + + + + + + + + + + + + + + + + the vanishing conditions , together with the atiyah - singer index theorem on @xmath2 which generalizes the statement for the euler number , yield that @xmath48 consequently , this gives us an expression for the net number of generations of particles , analogous to : @xmath49 we will require that this number be a multiple of three , say @xmath50 , where @xmath51 is the order of a possible freely acting group @xmath52 on @xmath2 , so that upon descending to the quotient manifold @xmath53 , there would be precisely 3 generations .
indeed , in order that @xmath52 be a free action , @xmath51 must necessarily ( but not sufficiently ) divide the euler number @xmath54 .
[ [ anomaly - cancellation ] ] anomaly cancellation : + + + + + + + + + + + + + + + + + + + + + + to ensure green - schwarz anomaly cancelation @xcite , it is standard to set @xmath55 , where @xmath56 is the ricci form on @xmath2 , that is , @xmath57 .
however , one could allow m5-branes in the bulk , in a heterotic m - theory hoava - witten setup @xcite , which could wrap effective holomorphic 2-cycles ( i.e. , actual curves ) .
hence , one could allow that the difference @xmath58 correspond to an effective class .
finally , since we are dealing with su - bundles , @xmath59 .
thus , in sum , we have constraints on all the three chern classes of @xmath23 .
our strategy is clear . we will attempt to construct a large class of bundles satisfying the above constraints over a substantial data - set of cy3s .
the most systematic method is the so - called * monad construction * over projective varieties @xcite , developed by algebraic geometers in the later half of the last century .
in fact , all vector bundles over projective spaces can be so obtained , which is perhaps why such a leibnizian name , with regard to the universality implied , was originally chosen .
it is thus not surprising that these bundles have been used sporadically but fruitfully in a physical context since the early days of string model building @xcite .
the building block to a monad is a * line bundle * , the simplest vector bundle on any manifold .
on cicys , a line bundle is easy to describe and so let us begin with this historical data - set @xcite
. denote @xmath60 , as is standard , the @xmath51-th power of the hyperplane bundle @xmath61 over @xmath62 , then its first chern class is @xmath63 with @xmath64 being the khler class of @xmath62 .
subsequently , one could restrict this to any projective variety @xmath2 to obtain @xmath65 with the proviso that @xmath2 also has only a single khler class , as descended from the ambient @xmath62 .
such cases of when @xmath66 of the ambient and variety itself are both unity is called * cyclic*. in general , our ambient space @xmath67 for cicys are products of @xmath68 projective spaces in which @xmath10 homogeneous polynomials define @xmath2 .
we shall call the cases where @xmath69 as * favourable * ; here the khler classes descend completely from @xmath67 to @xmath2 . in this case
we can write line bundles over @xmath70 as @xmath71 with corresponding restriction to the cicy , @xmath2 .
finally , we have the kodaira vanishing theorem @xcite which states that for a positive line bundle @xmath72 on a calabi - yau manifold @xmath2 , @xmath73 for all @xmath74 . for cicys ,
a positive line bundle is one for which each @xmath75 entry above is positive .
this vanishing will be of significant aid to us in computing the spectrum and couplings later .
therefore , our starting point will be monads constructed from positive line bundles over favourable cicys . thus prepared , we can define a monad as the bundle @xmath23 which resides in a short exact sequence ( i.e. , a free resolution of length 2 ) : @xmath76 here , short exactness implies that @xmath77 and that @xmath78 .
the map @xmath79 is explicitly a matrix of polynomials ; e.g. , on @xmath80 the @xmath81-th entry is a homogeneous polynomial of degree @xmath82 .
moreover , our positivity requirement implies that all integers @xmath83 be strictly positive . in summary , the physical constraints in
[ s : cons ] manifest themselves as a list of combinatorial conditions on the integers @xmath83 : 1 .
bundle - ness : @xmath84 for all @xmath85 and the map @xmath79 can be taken to be _ generic _ so long as exactness of the sequence is ensured ; 2 .
su - bundle : @xmath86 ; 3 .
anomaly cancellation : @xmath87 ; 4 .
three generations : @xmath88 , with @xmath51 a divisor of @xmath54 . upon finding such monads , we shall then prove stability and compute the particle spectrum and yukawa couplings .
these are extremely technical and difficult problems and have been a major hindrance to existing literature if one were to calculate them for large classes on the order of thousands or more .
a break - through in the method of attack of the programme in @xcite is the extensive use of the recent rapid advances in computer algebra , especially in algorithmic algebraic geometry @xcite .
we shall not dwell too much in this review on these technicalities and shall only mention that for stability , at least for cyclic manifolds , a so - called hoppe s criterion reduces the problem to computing cohomology groups .
to arrive at the various cohomologies , the standard approach here is to use spectral sequence induced from the koszul resolution of @xmath23 , whereby simplifying the problem to known cohomologies ( via bott - borel - weil and knneth @xcite ) on the ambient product of projective spaces .
it was shown in @xcite that positive bundles satisfying the above list of constraints is finite in number , amounting to about 7000 on only a small number of cicys .
we present the statistics thereof below and the histograms of @xmath89 , i.e. , the number of generations , in figure [ f : pos ] : [ cols= " < " , ] + rank 3 & 5680 & 3091 & 458 & 19 + rank 4 & 1334 & 207 & 96 & 2 + rank 5 & 104 & 52 & 5 & 0 + total & 7118 & 3350 & 559 & 21 + \(a ) , of the 7118 positive monads found over 36 favourable cicys : the horizontal axis is @xmath90 and the vertical , the number of bundles ; ( b ) the same data set , but only taking into account monads with @xmath91 for some positive integer @xmath51 , such that @xmath51 divides the euler number of the corresponding cicy . , title="fig:",width=211 ] ( b ) , of the 7118 positive monads found over 36 favourable cicys : the horizontal axis is @xmath90 and the vertical , the number of bundles ; ( b ) the same data set , but only taking into account monads with @xmath91 for some positive integer @xmath51 , such that @xmath51 divides the euler number of the corresponding cicy . , title="fig:",width=211 ] emboldened by our ability to classify and further intrigued by the rarity of occurrence of anything akin to the mssm , we enlarged our data - set to include the so - called semi - positives ; these are monads with possible zeros in the entries .
currently , short of strong isomorphism theorems we do not have a finiteness result even though the numbers are expected to be so ; preliminary scans have produced a substantially larger set . on these bundles
an initial scan was performed and single one was so far found @xcite to have the exact mssm spectrum .
interestingly , the base manifold turned out to be the bi - cubic , @xmath92 @xmath93{c}\mathbb{p}^2\\\mathbb{p}^2\end{array}\left|\begin{array}[c]{ccc}3 \\3 \end{array}\right . \right]^{2,83}_{-162}$ ] , again one of the guises of the tian - yau manifold !
specifically , this new heterotic standard model , to be added to the other two within the special corner , has no anti - generations , no exotic particles , one pair of mssm higgs and an @xmath33 .
it is obtained from an @xmath21 monad bundle @xmath94 giving first an @xmath20 gut , which is subsequently broken by a @xmath32 wilson line to give the desired @xmath95 group . in order to turn on this wilson line
we ensured that there exists a freely acting @xmath32-group , which was found in @xcite , giving a quotient manifold @xmath96 . to guarantee that the bundle @xmath23 also descends to the quotient
, it was further shown that admits an equivariant structure with respect to the group action , and hence the quotient is also a bona fide bundle .
armed with the advances in contemporary algebraic geometry , encouraged by the possibility of extensively utilizing computer algebra coupled with physical intuition and insight in order to sift through vast data - sets , and spurred by the observation and speculation of transgressions within a special corner in the space of stringy vacua , our spirits are enheartened by optimism .
for the first time , the prospects of an systematic and thorough expedition of the heterotic landscape , a venture dreamt of for two decades , seem within our grasp .
we have explored the oldest geography of the cicys , and already found a rare occurrence of the mssm ; meanwhile , reconnaissance into the further territory of the elliptic manifolds and most importantly the toric hypersurfaces are well under way @xcite .
the glimpse into a barren expanse of a _ terra incognita _
, harbouring perhaps a rare treasure cove bejewelled by universes close to our own , enlivens our souls .
onwards we must march .
_ ad catharinae sanctae alexandriae et ad majorem dei gloriam ... + _ the author would like to lend this opportunity to gratefully acknowledge his oxford friends l. anderson , p. candelas , x. de la ossa , j. gray , s .- j .
lee , a. lukas , b. szendri , as well as his old philadelphia confreres v. braun , r. donagi , b. ovrut , and t. pantev for many years of fruitful and enjoyable collaborations on this subject .
he is most indebted to the gracious patronage of merton college , oxford , whose fellowship bestowed upon him is held dearly to his heart , and he carouses to the health of the parting warden , professor dame jessica rawson as he salutes the new , professor sir martin taylor . _
stet fortuna domus ! _
p. candelas , x. de la ossa , y. h. he and b. szendroi , `` triadophilia : a special corner in the landscape , '' adv .
theor . math .
* 12 * , 2 ( 2008 ) [ arxiv:0706.3134 [ hep - th ] ] .
a. ananthaswamy , special report , _ new scientist _ , jan 5th , 2008 . )
p. candelas , a. m. dale , c. a. lutken and r. schimmrigk , `` complete intersection calabi - yau manifolds , '' nucl .
b * 298 * , 493 ( 1988 ) .
a. lukas , b. a. ovrut , k. s. stelle and d. waldram , `` the universe as a domain wall , '' phys .
d * 59 * , 086001 ( 1999 ) [ arxiv : hep - th/9803235 ] .
r. donagi , b. a. ovrut , t. pantev and r. reinbacher , `` su(4 ) instantons on calabi - yau threefolds with @xmath98 fundamental group , '' jhep * 0401 * , 022 ( 2004 ) [ arxiv : hep - th/0307273 ] .
r. donagi , y. h. he , b. a. ovrut and r. reinbacher , `` the particle spectrum of heterotic compactifications , '' jhep * 0412 * , 054 ( 2004 ) [ arxiv : hep - th/0405014 ] .
y. h. he , `` gut particle spectrum from heterotic compactification , '' mod .
a * 20 * , 1483 ( 2005 ) .
r. donagi , y. h. he , b. a. ovrut and r. reinbacher , `` higgs doublets , split multiplets and heterotic @xmath99 spectra , '' phys .
b * 618 * , 259 ( 2005 ) [ arxiv : hep - th/0409291 ] .
r. donagi , y. h. he , b. a. ovrut and r. reinbacher , `` the spectra of heterotic standard model vacua , '' jhep * 0506 * , 070 ( 2005 ) [ arxiv : hep - th/0411156 ]
. v. braun , y. h. he , b. a. ovrut and t. pantev , `` a heterotic standard model , '' phys .
b * 618 * , 252 ( 2005 ) [ arxiv : hep - th/0501070 ] .
v. braun , y. h. he , b. a. ovrut and t. pantev , `` a standard model from the e(8 ) x e(8 ) heterotic superstring , '' jhep * 0506 * , 039 ( 2005 ) [ arxiv : hep - th/0502155 ] . v. braun , y. h. he , b. a. ovrut and t. pantev , `` vector bundle extensions , sheaf cohomology , and the heterotic standard model , '' adv .
theor . math .
* 10 * , 4 ( 2006 ) [ arxiv : hep - th/0505041 ] .
v. braun , y. h. he , b. a. ovrut and t. pantev , `` the exact mssm spectrum from string theory , '' jhep * 0605 * , 043 ( 2006 ) [ arxiv : hep - th/0512177 ] .
v. braun , y. h. he and b. a. ovrut , `` yukawa couplings in heterotic standard models , '' jhep * 0604 * , 019 ( 2006 ) [ arxiv : hep - th/0601204 ] .
v. braun , y. h. he and b. a. ovrut , `` stability of the minimal heterotic standard model bundle , '' jhep * 0606 * , 032 ( 2006 ) [ arxiv : hep - th/0602073 ] .
l. b. anderson , y. h. he and a. lukas , `` monad bundles in heterotic string compactifications , '' jhep * 0807 * , 104 ( 2008 ) [ arxiv:0805.2875 [ hep - th ] ] .
l. b. anderson , `` heterotic and m - theory compactifications for string phenomenology , '' arxiv:0808.3621 [ hep - th ] .
y. h. he , s. j. lee and a. lukas , `` heterotic models from vector bundles on toric calabi - yau manifolds , '' arxiv:0911.0865 [ hep - th ]
. l. b. anderson , j. gray , a. lukas and b. ovrut , `` stability walls in heterotic theories , '' jhep * 0909 * , 026 ( 2009 ) [ arxiv:0905.1748 [ hep - th ] ] .
greuel , g. pfister , and h. schnemann , _ singular : a computer algebra system for polynomial computations _ , centre for computer algebra , university of kaiserslautern ( 2001 ) .
available at http://www.singular.uni-kl.de/ j. gray , y. h. he , v. jejjala and b. d. nelson , `` exploring the vacuum geometry of n = 1 gauge theories , '' nucl .
b * 750 * , 1 ( 2006 ) [ arxiv : hep - th/0604208 ] .
j. gray , y. h. he , a. ilderton and a. lukas , `` stringvacua : a mathematica package for studying vacuum configurations in string phenomenology , '' comput .
commun .
* 180 * , 107 ( 2009 ) [ arxiv:0801.1508 [ hep - th ] ] .
y. h. he , `` vacuum geometry and the search for new physics , '' _ in the proceedings of 9th workshop on non - perturbative quantum chromodynamics , paris , france , 4 - 8 jun 2007 , pp 04_.
k. uhlenbeck and s .- t .
yau , _ on the existence of hermitian - yang - mills connections in stable vector bundles _ , _ comm .
pure appl . math .
_ * 39 * ( 1986 ) , no .
s , suppl .
frontiers of the mathematical sciences : 1985 ( new york , 1985 ) .
r. blumenhagen , r. schimmrigk and a. wisskirchen , `` ( 0,2 ) mirror symmetry , '' nucl .
b * 486 * , 598 ( 1997 ) [ arxiv : hep - th/9609167 ] .
r. blumenhagen , `` target space duality for ( 0,2 ) compactifications , '' nucl .
b * 513 * , 573 ( 1998 ) [ arxiv : hep - th/9707198 ] . | we review the recent programme undertaken to construct , systematically and algorithmically , large classes of heterotic vacua , as well as the search for the mssm therein .
specifically , we outline the monad construction of vector bundles over complete intersection calabi - yau threefolds , their classification , stability , equivariant cohomology and subsequent relevance to string phenomenology .
it is hoped that this top - down algorithmic approach will isolate special corners in the heterotic landscape . |
rotational bands and deformed shapes in @xmath5ar .
energies of the g.s .
( spherical shape ) and sd bands @xcite ( ellipsoidal shape ) , and the energies of hd band from the quasimolecular resonances observed in the @xmath2c+@xmath1 mg ( open rectangles ) and @xmath6o+@xmath7ne ( full rectangles ) reactions ( dinuclear shape ) are plotted as a function of j(j+1 ) @xcite.,height=359 ] the observation of resonant structures in the excitation functions for various combinations of light @xmath0-cluster ( n = z ) nuclei in the energy regime from the barrier up to regions with excitation energies of e@xmath8 = 20 - 50 mev remains a subject of contemporary debate @xcite .
these resonances have been interpreted in terms of nuclear molecules @xcite .
the question whether quasimolecular resonances always represent true cluster states in the compound systems , or whether they may also simply reflect scattering states in the ion - ion potential is still unresolved @xcite . in many cases ,
these resonant structures have been associated with strongly - deformed shapes and with clustering phenomena , predicted from the nilsson - strutinsky approach @xcite , the cranked @xmath0-cluster model @xcite , or other mean - field calculations @xcite .
of particular interest is the relationship between superdeformation ( sd ) and nuclear molecules @xcite , since nuclear shapes with major - to - minor axis ratios of 2:1 have the typical ellipsoidal elongation ( with quadrupole deformation parameter @xmath9 @xmath10 0.6 ) for light nuclei @xcite .
furthermore , the structure of possible octupole - unstable 3:1 nuclear shapes ( with @xmath9 @xmath10 1.0 ) - hyperdeformation ( hd ) - for actinide nuclei has also been widely discussed @xcite in terms of clustering phenomena .
a typical example of the possible link between quasimolecular bands and sd / hd shapes @xcite is given in fig . 1 for @xmath5ar @xcite .
large quadrupole deformations ( @xmath9 = 0.6 - 1.0 ) and @xmath0-clustering in light n = z nuclei are known to be general phenomena at low excitation energy . for high angular momenta and higher excitation energies ,
very elongated shapes are expected @xcite to occur in @xmath0-like nuclei for a@xmath11 = 20 - 60 .
these predictions come from the generalized liquid - drop model , taking into account the proximity energy and quasi - molecular shapes @xcite ( as in the cluster models @xcite ) .
in fact , highly deformed shapes and sd rotational bands have been recently discovered in several such n = z nuclei , in particular , @xmath5ar using @xmath4-ray spectroscopy techniques @xcite .
hd bands in @xmath5ar , illustrated in fig . 1 as quasimolecular bands observed in @xmath2c+@xmath1 mg ( open rectangles ) and @xmath6o+@xmath7ne ( full rectangles ) reactions @xcite , and their related ternary clusterizations are predicted theoretically @xcite . with the exception of
the cluster decays of @xmath12ni @xcite and @xmath13zn @xcite recently studied using charged particle spectroscopy @xcite , no evidence for ternary breakup has yet been reported @xcite in light nuclei ; the particle decay of @xmath5ar sd bands and of other highly excited bands ( displayed as stars in fig .
1 ) is still unexplored .
the main binary reaction ( i.e. @xmath0-transfer ) channel of the @xmath1mg+@xmath2c , for which both resonant effects @xcite ( see open rectangles in fig . 1 ) and orbiting phenomena @xcite have been observed , is investigated in this work by using charged particle-@xmath4-ray coincidence techniques .
results on ternary fission in @xmath3s+@xmath1 mg are also discussed .
photograph ( top ) showing a mask in place in front of one arm of the brs telescope ( top ) ; calibrated two - bidimensional position x versus y ( in mm ) spectrum ( bottom ) using a 210 @xmath14g/@xmath15 @xmath16au target with the 162 mev @xmath3s beam . the relative intensity ( counts )
is shown on the side bar.,height=472 ] the study of charged particle-@xmath4-ray coincidences in binary reactions in inverse kinematics is a unique tool in the search for extreme shapes related to clustering phenomena . in this paper , we investigate both the @xmath1mg+@xmath2c and the @xmath3s+@xmath1 mg reactions by using the binary reaction trigger spectrometer ( brs ) @xcite in coincidence with the euroball iv ( eb ) @xmath4-ray spectrometer @xcite installed at the vivitron tandem facility of strasbourg .
the @xmath1 mg and @xmath3s beams were produced and accelerated by the vivitron with beam intensities kept constant at approximately 5 pna . for the @xmath1mg+@xmath2c study the targets consisted of 200 @xmath14g/@xmath15 thick foils of natural c with an incident beam energy of e@xmath17 = 130 mev an excitation energy range up to e@xmath18 = 30 mev in @xmath1 mg is covered .
for both reactions the brs , in conjunction with eb , gives access to a novel approach for the study of nuclei at large deformations as described below .
the brs associated with eb combines as essential elements two large - area ( with a solid angle of 187 msr each ) heavy - ion gas - detector telescopes in a kinematical coincidence setup at forward angles . a photograph of one of the brs telescope is shown in fig .
2 along with a two - dimensional spectrum obtained with a mask during a @xmath3s+@xmath16au calibration run at 163.5 mev @xcite .
the two telescope arms are mounted symmetrically on either side of the beam axis , each covering the forward scattering angle range 12.5@xmath19 - 45.5@xmath19 , i.e. @xmath20 = 29@xmath19 @xmath21 16.5@xmath19 .
for this reason the 30 tapered ge detectors of eb @xcite were removed .
out - of - plane angular correlation yields for binary decay ( first spectrum ) and for respective non - binary emission channels measured from the @xmath3s+@xmath1 mg reaction at 163.5 mev with different missing charges @xmath22z.,height=453 ] @xmath6o high - energy excited states populated in the @xmath6o+@xmath7ne exit - channel of the @xmath1mg+@xmath2c reaction at 130 mev .
doppler - shift corrections have been applied for o fragments detected in the brs .
the three blue arrows show the respective first escape peak positions of the 6.13 mev , 6.92 mev and 7.12 mev @xmath4-ray transitions in @xmath6o .
the new partial level scheme of @xmath6o is plotted in the inset.,height=283 ]
fig . 3 shows out - of - plane angular correlations measured for the @xmath3s+@xmath1 mg reaction at 163.5 mev @xcite .
a carefull analysis of the fragment total kinetic energies was undertaken to isolate true ternary events from binary fission components due to significant oxygen and carbon contaminants ( see refs .
@xcite for more details ) having different q - values .
the narrow peak of the first angular correlation with @xmath22z = 0 comes from the binary nature of the fragmentation process .
the two other narrow distributions , also peaked at 180@xmath23 for missing charges @xmath22z = -4 ( -2@xmath0 ) and -6 ( -3@xmath0 ) , define the coplanar ( or collinear ) ternary fission with a small out - of - plane momentum .
the widening of the correlation width as observed with increasing @xmath22z for the underlying broad components is well understood as a statistical @xmath0-emission process where 1 - 4@xmath0 particles are emitted from the fully accelerated fragments . for the @xmath1mg+@xmath2c @xmath0-transfer reaction , the identifications of all @xmath4 rays in @xmath7ne were achieved @xcite .
two previously unobserved transitions in @xmath6o from the decay of the 3@xmath24 state at 11.09 mev are clearly visible in the @xmath4-ray spectrum of fig . 4 , have been identified for the first time .
the new the partial level scheme is proposed in the inset of fig .
we note that , thanks to the excellent resolving power of the eb+brs set - up , the respective first escape peak positions of the 6.13 mev , 6.92 mev and 7.12 mev @xmath4-ray transitions in @xmath6o are also apparent in this spectrum . with appropriate doppler - shift corrections applied to oxygen fragments identified in the brs
, it has been possible to extend the knowledge of the level scheme of @xmath6o at high energies @xcite , well above the @xmath2c+@xmath0 threshold , which is given in fig . 4 for the sake of comparison .
new information has been deduced on branching ratios of the decay of the 3@xmath24 state of @xmath6o at 11.085 mev @xmath21 3 kev ( which does not @xmath0-decay because of non - natural parity @xcite , in contrast to the two neighbouring 4@xmath24 states at 10.36 mev and 11.10 mev , respectively ) to the 2@xmath24 state at 6.92 mev ( 54.6 @xmath21 2 @xmath25 ) and a value for the decay width @xmath26 fifty times lower than the one given in the literature @xcite , it means @xmath27 @xmath28 0.23 ev . the third state with non - natural parity ( 0@xmath29 ) belonging to the ( 0@xmath29 , 3@xmath24 , 4@xmath24 ) triplet near 11 mev has a known transition to the 1@xmath29 state which is not observed experimentally in our work .
this 3@xmath24 state result is important as it is the highest known @xmath4-decaying level for the well studied @xmath6o nucleus @xcite .
however , other experimental techniques will have to be used in the near future ( such as the paris / gaspard projects @xcite in preparation for the forthcoming spiral2 facility at ganil ) to search for the bose - einstein condensation ( bec ) @xmath0-particle state in @xmath6o ( an equivalent @xmath0+hoyle state ) predicted to be the 0@xmath30 state at about 2 mev above the 4@xmath0-particle breakup threshold @xcite .
the connection of @xmath0-clustering , quasimolecular resonances phenomena and extreme deformations ( sd , hd , ... ) @xcite can be discussed in terms of the aspects of @xmath4-ray spectroscopy of binary and/or ternary fragments .
exclusive data were collected with the binary reaction spectrometer ( brs ) in coincidence with euroball iv installed at the vivitron tandem facility of strasbourg .
new @xmath4-ray spectroscopy results on @xmath6o from binary alpha - transfer reactions has been obtained from the @xmath1mg+@xmath2c reaction . in @xmath3s+@xmath1 mg , ternary events can be interpreted as ternary cluster decay from a @xmath12ni composite system at high angular momenta through hyper - deformed shapes .
the search for extremely elongated configurations ( hd ) in rapidly rotating medium - mass nuclei , which has been pursued exclusively using @xmath4-spectroscopy , can be performed in conjunction with charged particle spectroscopy .
we thank the staff of the vivitron for providing us with good @xmath1 mg and @xmath3s stable beams and the euroball group of strasbourg for the excellent support in carrying out all the experiments with the brs .
this work was supported by the french in2p3/cnrs , the german ministry of research ( bmbf grant under contract nr.06-ob-900 ) , and the ec euroviv contract hpri - ct-1999 - 00078 .
s.t . would like to express his gratitude and warm hospitality during his three month stay in strasbourg to the ires and , he is grateful for the financial support obtained from the in2p3 , france . d.g.j .
acknowledges receipt of an epsrc advanced fellowship .
9 w. greiner , j. y. park , and w. scheid , _ nuclear molecules _
world scientific ( 1995 ) . c. beck , y. abe , n. aissaoui , b. djerroud , and f. haas , phys .
c * 49 * , 2618 ( 1994 ) . g. leander and s. e. larsson , nucl .
a239 * , 93 ( 1975 ) .
a. berg and l .- o .
jonsson , z. phys .
* 349 * , 205 ( 1994 ) . s. marsh and w. d. rae , phys .
b * 180 * , 185 ( 1986 ) .
h. flocard , p. h. heenen , s. j. krieger , and m. s. weiss , prog .
* 72 * , 1000 ( 1984 ) .
gupta , s.k .
patra , p.d .
stevenson , c. beck , and w. greiner , j. phys.(london ) * g35 , 075106 ( 2008 ) ; and references therein . c. beck , int .
e * 13 * , 9 ( 2004 ) ; see also c. beck , arxiv:*nucl - th/0401005(2004 ) . c. beck , nucl .
a738 * , 24 ( 2004 ) ; see also c. beck , arxiv:*nucl - ex/0401004(2004 ) .
j. cseh , a. algora , j. darai , and p.o .
hess , phys .
c * 70 , 034311 ( 2004 ) .
a. v. andreev , g. g. adamian , n. v. antonenko , s. p. ivanova , s. n. kulin , and w. scheid , eur .
j. a * 30 * , 579 ( 2006 ) . c. beck and p. papka , _
( to be published ) ; and references therein .
g. royer , j. phys.(london ) g * 21 * , 249 ( 1995 ) .
j. zhang , a.c .
merchant , and w.d.m .
rae , phys .
c * 49 , 562 ( 1994 ) .
et al . , phys .
lett . * 85 , 2693 ( 2000 ) .
a. algora , j. cseh , j. darai , and p. o. hess , phys .
b * 639 * , 451 ( 2006 ) . w. von oertzen _
et al . , eur .
j. a * 36 * , 279 ( 2008 ) . c. wheldon
_ et al . , nucl
a811 * , 276 ( 2008 ) .
v. zherebchevsky , w. von oertzen , d. kamanin , b. gebauer , s. thummerer , ch .
schulz , and g.royer , phys.lett .
b * 646 , 12 ( 2007 ) . w. von oertzen , v. zherebchevsky , b. gebauer , ch .
schulz , s. thummerer , d. kamanin , g.royer , and th .
wilpert , phys . rev .
c * 78 , 044615 ( 2008 ) ; see also w. von oertzen et al .
, arxiv:*0811.0018(2008 ) .
w. von oertzen _
ser . : * 111 , 012051 ( 2008 ) .
s. j. sanders , a. szanto de toledo , and c. beck , phys .
* 311 * , 487 ( 1999 ) ; see also s. j. sanders , a. szanto de toledo , and c. beck , arxiv:*nucl - ex/9904009(1999 ) . c. beck _
et al . , nucl . phys . *
a734 * , 453 ( 2004 ) ; see also c. beck _
et al . , arxiv:*nucl - ex/0309007(2003 ) . c. beck _
et al . , int . j. mod
e * 17 , no : 10 ( november 2008 ) _
( to be published ) ; see also c. beck _
et al . , arxiv:*0808.2405(2008 ) .
m. endt , atomic data and nucl . data tables * 55 * , 171 ( 1993 ) .
url : http://www.nndc.bnl.gov/ d. a. bromley , h. e. gove , j. a. kuehner , a. e. litherland , and e. almqvist , phys . rev . * 114 * , 758 ( 1959 ) .
url : http://gaspard.in2p3.fr/. y. funaki , t. yamada , h. horiuchi , g. roepke , p. schuck , and a.tohsaki , phys .
* 101 , 082502 ( 2008 ) . *
* _ _ * _ * _ _ * * _ * * * _ _ * _ * _ * * * * | the occurence of `` exotic '' shapes in light n = z @xmath0-like nuclei is investigated for @xmath1mg+@xmath2c and @xmath3s+@xmath1 mg .
various approaches of superdeformed and hyperdeformed bands associated with quasimolecular resonant structures with low spin are presented . for both reactions ,
exclusive data were collected with the binary reaction spectrometer in coincidence with euroball iv installed at the vivitron tandem facility of strasbourg .
specific structures with large deformation were selectively populated in binary reactions and their associated @xmath4-decays studied .
the analysis of the binary and ternary reaction channels is discussed .
address = institut pluridisciplinaire hubert curien , umr7178 , cnrs - in2p3 et universit louis pasteur ( strasbourg i ) , b.p .
28 , f-67037 strasbourg cedex 2 , france address = institut pluridisciplinaire hubert curien , umr7178 , cnrs - in2p3 et universit louis pasteur ( strasbourg i ) , b.p .
28 , f-67037 strasbourg cedex 2 , france address = institut pluridisciplinaire hubert curien , umr7178 , cnrs - in2p3 et universit louis pasteur ( strasbourg i ) , b.p .
28 , f-67037 strasbourg cedex 2 , france address = institut pluridisciplinaire hubert curien , umr7178 , cnrs - in2p3 et universit louis pasteur ( strasbourg i ) , b.p .
28 , f-67037 strasbourg cedex 2 , france address = institut pluridisciplinaire hubert curien , umr7178 , cnrs - in2p3 et universit louis pasteur ( strasbourg i ) , b.p .
28 , f-67037 strasbourg cedex 2 , france address = institut pluridisciplinaire hubert curien , umr7178 , cnrs - in2p3 et universit louis pasteur ( strasbourg i ) , b.p .
28 , f-67037 strasbourg cedex 2 , france address = institut pluridisciplinaire hubert curien , umr7178 , cnrs - in2p3 et universit louis pasteur ( strasbourg i ) , b.p .
28 , f-67037 strasbourg cedex 2 , france address = institut pluridisciplinaire hubert curien , umr7178 , cnrs - in2p3 et universit louis pasteur ( strasbourg i ) , b.p .
28 , f-67037 strasbourg cedex 2 , france address = institut pluridisciplinaire hubert curien , umr7178 , cnrs - in2p3 et universit louis pasteur ( strasbourg i ) , b.p .
28 , f-67037 strasbourg cedex 2 , france address = institut pluridisciplinaire hubert curien , umr7178 , cnrs - in2p3 et universit louis pasteur ( strasbourg i ) , b.p .
28 , f-67037 strasbourg cedex 2 , france address = institut pluridisciplinaire hubert curien , umr7178 , cnrs - in2p3 et universit louis pasteur ( strasbourg i ) , b.p .
28 , f-67037 strasbourg cedex 2 , france address = hahn - meitner - institut , glienicker str . 100 , d-14109 berlin , germany address = hahn - meitner - institut , glienicker str .
100 , d-14109 berlin , germany address = hahn - meitner - institut , glienicker str .
100 , d-14109 berlin , germany address = hahn - meitner - institut , glienicker str .
100 , d-14109 berlin , germany address = infn - lab .
nationali di legnaro and dipartimento di fisica , i-35020 padova , italy address = infn - lab .
nationali di legnaro and dipartimento di fisica , i-35020 padova , italy address = infn - lab .
nationali di legnaro and dipartimento di fisica , i-35020 padova , italy address = infn - lab .
nationali di legnaro and dipartimento di fisica , i-35020 padova , italy address = infn - lab .
nationali di legnaro and dipartimento di fisica , i-35020 padova , italy address = school of physics and chemistry , university of surrey , guildford , surrey , gu2 7xh , uk address = department of physics , university of york , york , yo10 5dd , uk address = subatech , in2p3-cnrs et universit - ecole des mines , 4 rue a. kastler , f-44307 nantes |
narrow line active galactic nuclei form a distinct class of agn on the basis of the properties of their optical / uv spectrum : full width at half maximum ( fwhm ) of the hydrogen lines and other lines in the range 500 - 2000 km s@xmath7 , intense high ionization lines , and intense feii multiplets ( osterbrock & pogge , @xcite ; shuder & osterbrock , @xcite ) . the weakest of these agn , the narrow line seyfert 1 ( nls1 ) , and the somewhat brighter agn with absolute optical luminosity above but close to the lowest limit for quasars of @xmath9 = -23.4 have been extensively studied in the x - ray range ( e.g. laor , fiore , elvis et al . , 1994 ;
boller , brandt & fink , 1996 ; laor et al . , 1997 ) . among all agn ,
the narrow lines agn tend to have the steepest soft x - ray spectra ( rosat ) , and some of them show fast , large amplitude soft x - ray variability with occasionally giant outbursts ( e.g. grupe 1996 and references therein ; boller et al 1996 ) . in the harder 2 - 10 kev band
a comparative study of a large sample of nls1 and broad line seyfert 1s revealed that the 2 - 10 kev asca spectral slopes of nls1 are significantly steeper than those of broad line agn ( brandt , mathur & elvis 1997 ) .
recent bepposax observations of a selected sample of bright nls1 ( comastri , fiore , guainazzi et al .
1998 ; comastri , brandt , leighly et al . 1998 ) over the broad 0.1 - 10 kev range indicate that a two component model provides an adequate description of the x ray continuum .
the relative strengths and slopes of the two components are different from those of broad line seyfert 1s .
the nls1 are characterized by a stronger soft excess and , in general , have a steeper medium energy x - ray power law with respect to normal seyfert 1s , but in pg 1404 + 226 the medium energy spectral index is not very different from that of classical seyfert 1s while its soft excess is strong , very steep and rapidly variable like in other nls1 .
the spectral behaviour of nls1 suggests that the soft x - ray flux can not be due only to disk reprocessing unless there is highly anisotropic emission . in the framework of the thermal models for the x - ray emission in seyfert 1s ( haardt & maraschi 1993 ) a strong soft component could lead to a strong compton cooling of the hot corona electrons and to a steep hard tail
this hypothesis is also supported by the similarities between nls1 spectra and those of galactic black hole candidates in their high states first suggested by pounds , done and osborne ( 1995 ) and czerny et al .
( 1996 ) .
the high states of galactic black hole candidates are thought to be triggered by increases in the accretion rate possibly reaching values close to the eddington limit . in this case
the disk surface is expected to be highly ionized , in good agreement with the observation of a h - like fe edge in tons180 ( comastri et al .
it should also be noted that a high @xmath10 ratio would also be consistent with the narrowness of the optical lines in nls1 if the optical line producing region is virialized as suggested by laor et al .
( 1997 ) the quasar pg 1404 + 226 ( @xmath11 = 0.098 , v = 15 ) is one of the brightest members of the class of narrow line agn .
the optical spectrum of pg 1404 + 226 displays the characteristics of nls1 with fwhm ( h@xmath12 ) @xmath13 830 km s@xmath7 and strong fe ii emission ( boroson & green 1992 ) .
+ the observations with the rosat pspc ( @xmath2 0.12.0 kev ) showed a very steep spectrum ( @xmath14 ) with rapid flux ( a factor 2 in 10 hours ) and spectral variability typical of nls1 in the x - ray range , and revealed a complex absorption feature around 0.8 - 1.0 kev ( ulrich & molendi 1996 , thereafter um96 ) .
time resolved spectral analysis showed the data to be consistent with a shift of the absorption feature to higher energy when the source brightens ( um96 ) .
we have obtained a 40 ks asca observation of pg 1404 + 226 in order to investigate the x - ray spectrum over a larger energy range and at higher energy resolution than was possible with rosat . at the time of the asca observations the absorption feature was around 1 kev , at an energy definitely higher than that of the ovii and oviii edges at 0.74 and 0.87
kev commonly seen in asca and rosat spectra of agn ( e.g. reynolds 1997 ) .
preliminary results can be found in comastri , molendi & ulrich ( 1997 , hereafter cmu97 ) .
the asca and rosat data have been fit separately since they have not been obtained at the same epoch .
the challenge presented by pg 1404 + 226 is the identification of the absorption at @xmath15 kev .
brandt et al ( 1994 ) found a similar feature at 1.15 kev in ark 564 and considered several interpretations : neon edge , iron l edges , and outflowing material which would raise the energy edge of oviii ( 0.870 kev ) to the observed energy .
they find none of them to be satisfactory : neon edge because there is only a narrow range of ionization parameter where it would be stronger than the oviii edge ; the iron l edges would produce absorption at other somewhat higher energy ( 1.358 kev ) ; and the outflow seems unlikely because the source of its kinetic energy is unclear .
otani et al ( 1996 ) found a feature at @xmath8 kev in iras 13224 - 3809 for which they suggest an interpretation in terms of outflow , or alternatively and considering that the the x - ray flux of iras 13224 - 3809 varied by a factor 50 in two days `` the ionization state of the medium is far from equilibrium due to the violent variability '' .
krolik & kriss ( 1995 ) drew attention to the fact that because the ionization timescales of some ions may be as long as the variability timescales in agns , the ionic abundances indicated by the transmission spectra may not be well described by ionization equilibrium " .
this point has recently been re - investigated by nicastro et al .
( 1999a ) who pointed out also that ( i ) recombination times can be longer than photoionization times resulting in gaseous absorbers which are over - ionized with respect to the equilibrium ionization state and ( ii ) collisional ionization could be of comparable importance to photoionization .
hayashida ( 1997 ) found an edge near 1 kev in h0707 - 495 for which he also suggests an identification with a blushifted oviii absorption edge . for pg 1404 + 226 ,
cmu97 suggested an overabundance of iron , an interpretation which has the advantage to link the origin of the 1 kev absorption to the intense feii lines present in the optical / uv of this quasar and also of iras 13224 - 3809 and ark 564 .
the interpretation where the absorption near @xmath8 kev originates in hot gas outflowing at velocities in the range 0.2 - 0.5c has also been proposed by leighly et al ( 1997 ) .
this paper presents an analysis of the asca observations , a re - analysis of the rosat spectra with our warm absorber ( wa ) models calculated with cloudy ( ferland 1993 ) , and an analysis of hst spectra which we have obtained in order to search for uv absorption / emission lines .
the data reduction was carried out in a standard manner and like in um96 the data were split into high - state and low - state depending on count rate .
the analysis was performed with our warm absorber models based on cloudy .
ccccl state & log @xmath16 & log @xmath17 & @xmath18 & model + hs / ls & & & 4.1/2.0 & single pl + hs / ls & 1.0/0.7 & 23.7/23.4 & 2.5/0.9 & standard warm absorber + hs / ls & 1.0/0.7 & 23.8/23.5 & 2.2/0.9 & emission+reflection added + hs / ls & 0.7/0.3 & 23.3/22.9 & 1.3/0.9 & ne = 4@xmath19solar abundance + hs / ls & 0.6/0.4 & 23.1/23.1@xmath20 & 1.2/1.0 & additional 0.1 kev soft excess@xmath21 + @xmath20
fixed to value derived for hs ; @xmath21 abundances reset to solar the warm absorber is assumed to be of constant density , of solar abundances ( if not mentioned otherwise ) , and to be illuminated by the continuum of the central point - like energy source .
the spectral energy distribution from the radio to the gamma - ray region consists of piecewise powerlaws with , in particular , an energy index in the euv , @xmath4@xmath22 , of 1.4 and an x - ray photon index @xmath23@xmath24 of 1.9 . a black - body - like soft excess is added in some models .
the column density n@xmath25 of the warm material ( i.e. the total column density in hydrogen ) and the ionization parameter u are determined from x - ray spectral fits ( table 1 ) .
u is a measure of the number rate of ionizing photons above the lyman limit and is defined by @xmath26 where @xmath27 is the number rate of incident photons above the lyman limit , @xmath28 is the distance between central source and absorber , @xmath29 is the speed of light , and @xmath30 is the hydrogen density ( fixed to 10@xmath31 @xmath32 unless noted otherwise ) . as detailed below , several models give acceptable fits to the low - state spectrum but none gives a really good fit to the high - state spectrum ( which has a higher signal / noise ratio ) .
a simple power law plus cold absorption gives a very poor fit .
the cold absorbing column generally tends to slightly underpredict the galactic value .
this may reflect some low energy calibration uncertainties , or the presence of an additional spectral component in the form of a very soft excess . adding a soft excess parameterized as a black body which contributes only at very low energies ( e.g. in ngc4051 , komossa & fink 1997 ; ton s180 , fink et al .
1997 ) improves the fit which , however , remains unsatisfactory .
we have examined a family of models ( labelled a , b, .. e ; table 1 ) which include a warm absorber plus a power law component with a photon index @xmath23@xmath24 fixed at 1.9 .
_ model a : _ the inclusion of a ` standard ' warm absorber model clearly improves the fit but does not yield an acceptable @xmath33 , with systematic residuals remaining near 1 kev .
the reason is that , even for very high ionization parameters , the model has a strong oxygen absorption at 0.87 kev , even stronger than the neon absorptions ( nevii at @xmath34 1.1 kev and nex at @xmath34 1.36 kev ) .
this is exacerbated at high - state because the high - state data require absorption around the location of the neon edges to dominate .
( the same results hold for a steeper underlying powerlaw . )
+ _ model b : _ the addition of the emission and reflection components of the warm material , calculated for a covering factor of 0.5 results in a slightly improved fit , but the high - state data are still not well matched .
one way to make the neon absorption dominate over the oxygen absorption is to introduce deviations from solar abundances , with overabundant neon , or underabundant oxygen . +
several deviation factors were studied between an oxygen abundance of up to o = 0.2 x solar and a neon abundance of ne = 4 x solar .
these models strongly improve the quality of the fit and the values of @xmath33 reach acceptable values ( for the rosat spectra ) .
the best fit has an overabundance of neon of @xmath35 times the solar value .
+ we note that while non solar abundances have been reported in a number of agn / quasars ( e.g. hamann & ferland 1993 , netzer 1997 ) the ratio o / ne is expected to remain close to its solar value in all known astrophysical situations ( on the other hand , netzer occasionally depletes _ oxygen only _ in his photoionisation models , e.g. marshall et al .
1993 ) . in any event ,
the model with overabundance of neon is not compatible with the asca data ( section 3.2.2 ) .
_ model d : _ motivated by the asca evidence for the presence of a soft excess , a sequence of models was calculated with an additional hot bb component of kt = 0.1 kev .
this component was included in the ionizing spectral energy distribution that illuminates the warm absorber , i.e. the change in ionization structure of the warm material was self - consistently calculated .
a successful description of both , high- and low - state data is possible with solar abundances .
note that two - component models _ without edge _ , specifically , power law + blackbody and power law + bremstrahlung were found to provide satisfactory fits of the rosat data at low state .
for the high state data the power law + blackbody model gives @xmath33 = 67/20 ( um96 ) .
+ _ model e : _ u and
n@xmath25 have above been treated as free parameters , and the fits tend to give slightly different column densities in low- and high - state data , whereas this quantity is not expected to vary within short time - scales ( 10 hours or less ) .
we therefore have searched for the best fit model ( model e ) to the high- and low - state data in which the column density is identical in both states and only the ionization parameter u and the strength of the bb component are allowed to vary between states .
we find a successful model with log(n@xmath25 ) = 23.1 , log(u@xmath36 ) = 0.6 , log(u@xmath37 ) = 0.4 , and a contribution of the blackbody to the ionization parameter of log(u@xmath38 ) = log(u@xmath39 ) - 0.9 in the high - state data and log(u@xmath38 ) = log(u@xmath39 ) -1.5 in the low - state data ( fig . 1 ; u@xmath38 and u@xmath39 are the ionization parameters related to the blackbody component , and powerlaw component , respectively .
u is a measure of the number rate of ionizing photons above the lyman limit .
( 1 ) for the definition of u ) .
pg 1404 + 226 was observed by asca ( tanaka , inoue & holt 1994 ) july 13 - 14 , 1994 with the gas imaging spectrometers ( gis2/gis3 ) for a total effective exposure time of 35000 s , and with the solid state imaging spectrometers ( sis0/sis1 ) for about 29000 s. lccccc model & @xmath23 or @xmath40 @xmath20 & @xmath23 @xmath21 & @xmath41/@xmath42 @xmath43 & @xmath44/@xmath45 @xmath46 & @xmath47/dof + pl & @xmath48 & ... & ... & ... & 438.3/146 + pl+pl & @xmath49 & @xmath50 & ... & ... & 291.5/144 + br+pl & @xmath51 & @xmath52 & ... & ... & 235.7/144 + bb+pl & 126 @xmath53 4 & @xmath54 & ... & ... & 196.0/144 + bb+pl+edge & @xmath55 & @xmath56 & @xmath57 & @xmath58 & 158.6/142 + cutbb+pl & @xmath59 & @xmath60 & @xmath61 & 51@xmath62 & 159.5/142 + @xmath20 power law slope or thermal model temperature ( in ev ) for the soft component ; + @xmath21 power law photon index for the hard component ; + @xmath43 edge energy / exponential cutoff energy ( in kev ) ; + @xmath46 edge optical depth / e - folding energy of the exponential cutoff in ev the sis was operating in 1-ccd mode and all the data were collected in faint mode .
standard criteria for the good selection intervals have been applied ( i.e. cut - off rigidity @xmath63 7 for gis and @xmath63 6 for sis , minimum elevation angle from the earth @xmath63 5 degrees and minimum bright earth angle @xmath63 25 degrees for sis ) as well as dfe and echo corrections .
the background subtracted count rates for pg 1404 + 226 are 0.049,0.038,0.014,0.017 c / s in s0 , s1 , s2 , s3 respectively .
large flux variations have been detected . in panels
a and b of fig .
2 we show the soft ( 0.52.0 ) and hard ( 2.010 kev ) sis0 light curves .
the sis0 light curve in the soft band ( fig .
2a ) shows a factor @xmath2 4 of amplitude variability with a doubling timescale of @xmath2 8000 s. the variability pattern in the two bands is well correlated suggesting that the hard and soft fluxes varied in the same way .
a detailed spectral resolved temporal analysis is hampered by the low counting statistics especially at high ( @xmath63 2 kev ) energies . in order to have some indications on the spectral variability
we have performed a hardness ratio analysis for the high ( first part of the observation ) and the low ( second part of the observation ) state shown in figure 2 . the hardness ratio ( fig .
2c ) has been defined as hr = ( h - s)/(h+s ) where h and s are the counts in the 210 and 0.52 kev bands respectively .
the results for the high and low state are @xmath64 and @xmath65 suggesting that the spectrum is rather soft in both states .
the 210 kev flux is rather low and as a consequence both the 2 - 10 kev light curve and the hardness ratio light curve are noisy with no clear evidence of spectral variability in contrast to the spectral change during the rosat observations .
the spectral analysis has thus been performed on the spectrum accumulated over the entire duration of the asca observation .
the average luminosity is comparable with the rosat low state but the absorption feature is at high energy as during the rosat high state .
+ gis and sis spectra were binned with more than 20 counts / bin in the 0.710 kev and 0.510 kev energy ranges respectively .
since the spectral parameters obtained by fitting the four detectors separately were all consistent within the errors , the data were fitted simultaneously to the same model . given the very low galactic column density toward
pg 1404 + 226 ( @xmath66 @xmath67 ; elvis , lockman & wilkes 1989 ) and the low sensitivity of asca detectors for small column densities , all the spectral fits have been performed with @xmath68 .
this choice is also corroborated by the rosat results ( um96 ) .
a single power law model clearly does not provide an acceptable fit ( table 2 ) the power law fits reported in table 2 suggest that at least two components are required to fit the overall spectrum . we note that the slope of the hard power law , which could not have been detected by rosat , is consistent with the average quasar slope ( comastri et al .
1992 ) while the slope below 2 kev is consistent with the um96 findings . in order to model the broad band ( 0.410 kev ) spectrum we tried 4 types of double component
fits ( see table 2 for details ) : a double power law , bremsstrahlung plus power law , blackbody plus power law , and cut - off blackbody plus power law ( the cut - off is of the form exp @xmath69 $ ] for energies larger than the cutoff energy @xmath70 kev .
the depth of the cut - off is related to @xmath71 , a small value correponding to a very steep decline of the intensity ; here @xmath72 kev ) .
the first two models give unacceptable fits .
the only two acceptable descriptions of the data are the cut - off blackbody plus power law ( with a cut - off at 0.94 kev ) and the blackbody plus power law plus absorption edge model around 1 kev .
the addition of the absorption edge to the simple blackbody plus power law model produces an improvement in the fit which is significant at @xmath73 ( f - test ) and the residuals are featureless in agreement with the results of leighly et al ( 1997 ) . as absorption edges are the signature of warm absorbing material , we fit the warm absorber model available in xspec 9.0 ( the model is ` absori ' and the iron abundance is a free parameter ; it was developed by p. magdziarz & a. zdziarski following done et al 1992 and zdziarski , ghisellini , george et al .
1990 ) . given the relatively large numbers of free parameters in this model we have fixed the temperature of the warm material at @xmath74 k ( see for example reynolds & fabian 1995 ; the fits are not sensitive to the temperature , a change of log_t _ from 4.5 to 6 would make no significant difference ) and the iron abundance at the solar value .
in addition we have considered only the 0.53.0 kev energy range in order to avoid contamination from the high energy component .
we were not able to find any acceptable solution as can be judged from figure 3 .
leaving the iron abundance free to vary the improvement is significant at @xmath75% ( f - test ) and the residuals featureless ( fig 4 ) .
the resulting parameters are : @xmath76 erg cm s@xmath7 , @xmath77 @xmath67 , iron abundance @xmath63 25 solar and power law photon slope of @xmath78 .
( all the quoted errors are at 90% confidence for one interesting parameter , @xmath79 + 2.7 ) . the shape of residuals around 67 kev are suggestive of iron line emission ( fig 5 ) .
with the addition of a narrow line the improvement in the fit is , however , significant only at the 2 @xmath80 level .
the derived parameters are @xmath81 kev and ew = @xmath82 ev .
it is interesting to note that the huge equivalent width of the iron line is qualitatively consistent ( see figure 3 in reynolds , fabian & inoue , 1995 ) with the supersolar iron abundance found by fitting the data with the warm absorber model .
we have also attempted to fit some warm absorber models ( cloudy based , and with blackbody + power law continuum ) to the asca spectrum and have found no satisfactory fit - the reason being that the edge - like feature at 1 kev is too deep and too narrow for these models .
it has been recently shown that the spectra emerging from ionized gas are strongly modified by resonant absorption if the dispersion velocity in the gas is of the order of 100 km s@xmath7 or greater ( nicastro , fiore , matt 1999b ) .
the x ray warm absorber features are also strongly dependent from the shape of the ionizing continuum . for soft x
ray spectra as steep as the one observed for pg 1404 + 226 several resonance absorption lines from fe l , mg , si and s are predicted between 1 and 2 kev
. such blend of lines would appear as negative residuals in low resolution asca spectra .
the @xmath2 1 kev absorption feature of pg 1404 + 226 could be , at least in part , accounted for by resonant absorption .
signatures of this process have been looked for in the hst uv spectrum ( see below ) .
finally , we have also considered the possibility that the absorption at 1.07 kev is the blueshifted oviii edge whose rest frame energy is 0.87 kev . we have calculated the optical depth expected for the neon edges at rest energies 1.1 and 1.36 kev which would be blueshifted to 1.29 and 1.55 kev respectively .
for the 1.1 kev line the instrumental upper limit on @xmath44 is 0.14 ( 90% confidence level ) and is 0.27 for the 1.36 kev edge .
we note that with a ` standard ' warm absorber in relativistic outflow the oviii edge is never the `` only '' edge in the spectrum . if the absorber is lowly ionized , there is ovii co - existing and if it is very highly ionized , ovii becomes negligible , but one cannnot avoid having some neon absorption .
although the exact ratios of the optical depths @xmath44 depend on the input parameters , we thought it would be useful to check whether we could detect any other edge , or check how strict the upper limits are .
we take ngc 4051 as basis for comparison . in ngc 4051
, the absorber is rather highly ionized , with oviii being the strongest edge with @xmath44= @xmath83 , then nex with @xmath44=@xmath84 . the ovii edge is weaker with @xmath44=0.35 ; komossa & fink , 1997 .
using the ratio @xmath44@xmath85/@xmath44@xmath86 as typical for the relative depths in a highly ionized absorber , we can conclude that our non - detection of edges other than the one near 1.1 kev in pg1404 + 226 is still consistent with a standard warm absorber i.e. our upper limits on @xmath44 are not strict enough to rule out a warm absorber in relativistic outflow .
four spectra were taken in february 1996 with hst / fos and gratings g130 , g190h , g270 and g400 with integration times 2300 , 530 , 120 and 120 seconds respectively .
the total observed wavelength range covered is 1087 - 4773 with a nominal resolution of 1300 ( fig.6 ) .
the spectra were taken through the 0.86 arcsec diameter aperture .
standard reduction procedures were performed .
the wavelength scale of the spectrum taken with g190h was shifted by + 1.5 , to be consistent with the other spectra .
pg 1404 + 226 was observed with iue on july 1994 , 2 days after the rosat observations .
the spectrum , swp 51419 , has a total integration time of 315 minutes ( accumulated in 12 parts ) , and was taken through the large aperture and at low dispersion .
compared with the hst spectra taken 18 monthts later the continuum flux in july 1994 is @xmath34 1.3 times brighter in the common observed wavelength range 1265 to 1400 but the ly@xmath4 line kept the same intensity . the modest s / n and spectral resolution of the iue spectrum would prevent the detection of the absorption lines seen in the hst spectra .
no change in the emission / absorption profile of ly@xmath4 ( fig.7 ) can be detected by comparing the iue spectrum and the fos - g130 spectrum rebinned at 2 .
table 3 lists the emission line intensities ( h@xmath87 = 50 km s@xmath7 mpc@xmath7 , q@xmath87 = 0 , distance = 617 mpc ) .
we note the presence of some weak emission lines : ( 1 ) an unidentified line at 1175 ( rest wavelength 1070 ) noticed in a few other quasar spectra ( laor et al 1995 ; hamann et al 1997 ) ( 2 ) a line at 1290 which we identifiy with [email protected] .
llr@lr@l & & & + & & & + identification & @xmath89 & & + ovi 1033.83 & 1135.15 & t & & t & + heii 1085.15 & 1191.5 & 1&.2 & 5&.4 + ciii 1175.70 & 1290.0 & 1&.2 & 5&.4 + ly@xmath4 1215.67 & 1334.8 & 54 & & 245 & + nv 1240.15 & 1361.7 & & & & + oi 1304.46 & 1432.25 & 2&.5 & 11&.3 + cii 1334.53 & 1465.0 & 1&.2 & 5&.4 + siiv 1396.76 & 1533.6 & 3&.5 & 15&.9 + niv ] 1486.50 & 1632.2 & & & & + civ 1549.05 & 1700.85 & 11&.5 & 52&.2 + heii 1640.50 & 1801.3 & & & & + siiii ] 1892.03 & 2074.75 & siiii]+&ciii ] 9.9 & 45&.0 + ciii ] 1908.73 & 2093.9 & + mgii 2795.53 & 3069.0 & 4&.5 & 20&.4 + mgii 2802.70 & 3078.1 & & & & + h@xmath90 & 4945.4 & 1&.0 & 4&.5 + h@xmath91 + [ oiii ] &
& 1&.3 & 5&.9 + + this line is seen in hst spectra of izw1 and laor et al ( 1997 ) suggest that it is produced by resonance scattering of continuum photons by ciii@xmath88 ions , a mechanism which requires large velocity gradients ( @xmath34 1000 km s@xmath7 ) within _ each _ emitting cloud of the blr .
we identify two absorption systems in the hst spectra which , in the source frame , are separated by @xmath92 km s@xmath7 . in the ` blue ' system the absorption lines appear in the blue flank of the ly@xmath4 and civ emission lines at 800 km s@xmath7 from the peak . in the ` red ' system
the absorption lines appear on the red flank of the emission lines at 1100 km s@xmath7 from the peak .
it is known that in radio quiet agn / quasars , the high ionization lines such as civ are blueshifted with respect to the systemic velocity by a few hundred to a few thousand km s@xmath7 ( e.g. van groningen 1987 , corbin 1995 , sulentic et al .
1995 ) this blueshift being generally interpreted as evidence for a wind outflowing from the face of the accretion disk turned towards us . on this basis
, we argue that the red absorption system is close to zero velocity in the source frame while the blue system originates from material which has an outflow velocity of @xmath6 km s@xmath7 ( towards us with respect to the quasar ) .
llll @xmath93 ( ) & identification & @xmath11 meas . &
ew ( ) + meas .
& & & meas .
+ + 1206.5 & siiii 1206.50 & & 0.6 + 1259.8 & siii 1260.42 & & t @xmath20 + 2795.5 & mgii 2795.53 & & 0.8 + 2802.5 & mgii 2802.70 & & 0.95 + 2851.7 & mgi 2852.13 & & t + + 1282.5 & ciii*1175.70 & 0.09084 & 0.6 + 1327.0 & ly@xmath4 1215.57 & 0.091579 & 1.2 @xmath21 + 1352.4 & nv 1238.81 & 0.091700 & 0.6 + 1356.3 & nv 1242.79 & 0.091330 & 0.2 + 1521.2 & siiv 1393.76 & 0.09144 & 0.3 + & siiv 1402.77 & &
@xmath20 + 1688.5 & civ 1548.19 & 0.090630 & 1.1 @xmath21 + 1691.5 & civ 1550.76 & 0.090750 & 0.8 + + 1334.60 & ly@xmath4 1215.57 & 0.097830 & 0.95 @xmath43 + 1360.65 & nv 1238.81 & 0.098360 & 0.4 + 1365.45 &
nv 1242.79 & 0.098690 & 0.25 + & siiv 1393.76 & & + 1538.5 & siiv 1402.77 & 0.0967 & t + 1699.1 & civ 1548.19 & 0.097480 & 0.55 + & civ 1550.76 & & + 2860.0 & feii2599.4 ? & 0.10 & 0.7 + * footnotes : * @xmath20 @xmath21 @xmath43 several factors hamper the measurement of the absorption lines : modest s / n , limited resolution of fos and uncertainties affecting the profile of the emission lines .
the absorption lines were measured assuming a plausible reconstruction of the top of the emission lines .
still it is not possible to obtain sufficiently accurate ly@xmath4 absorption profile , and nv and civ doublet ratios to ascertain whether these lines are optically thin or thick and to estimate the covering factor ( some of the galactic lines do not reach zero either ) .
higher spectral resolution is needed to elucidate these important points . with this caveat ,
the measures are given in table 4 .
particularly interesting is the absorption line at 1282.5 ( fwhm of 2 and ew of @xmath34 0.7 ) , which could be ciii@xmath881175 in the @xmath11 = 0.091 system .
the agreement in redshift is good .
this line is present in the iue and hut spectra of ngc 4151 ( bromage et al 1985 , kriss et al 1992 ) .
bromage et al ( 1985 ) argue that absorption by the excited metastable level of ciii@xmath881175 and its strength relative to civ 1549 require this level to be collisionally populated in a high density medium with n@xmath94 = @xmath95 @xmath32 . in ngc 4151
, the ew of ciii@xmath881175 is between 0.7 and 1.0 times the ew of civ1548,1550 while in pg1404 + 226 , it is between 0.35 and 0.5 the ew of civ1548,1550 . note that an appealing alternative for the identification of this line is the _ civ doublet blueshifted _ by 0.3 with respect to pg1404 + 226 .
the line width , only @xmath34 2 , however , argues against this interpretation . there is no other candidate for highly blueshifted absorption lines in the spectrum .
the @xmath11 = 0.098 system is consistent with being produced in the halo of the host galaxy or of a nearby companion but could also be intrinsic to the nucleus .
as for the @xmath11 = 0.091 system the probable presence ( which needs verification ) of the ciii@xmath881176 line suggests that the system is intrinsic to the quasar and forms in an outflowing wind with a velocity of 1900 km s@xmath7 .
\(a ) the ew of the uv absorption lines expected from the warm absorber have been calculated for the statistically acceptable models of the rosat data .
they are given in table 5 , separately for low- and high - state as a measure of the uncertainty arising from the continuum variability ( non simultaneous uv and x - observations , non- equilibrium of the gas , etc .
see section 1 and nicastro et al .
1999a ) .
cccccccl state & log @xmath16 & log @xmath96 & & model + & & & neviii & ovi & fexxi & fexiv & + hs / ls & 0.7/0.3 & 23.3/22.9 & 1.6 & 0.1 & 1.7 & - & ( c ) : ne = 4@xmath19solar abundance + hs / ls & 0.7/0.5 & 23.5/23.4 & 1.3 & 0.2 & 2.5 & 0.01 & o = 0.4@xmath19solar abundance + hs / ls & 0.6/0.4 & 23.1/23.1 & 0.1 & 0.05 & 1.2 & - & ( e ) : additional 0.1 kev soft excess + we find that the civ1550 and nv 1240 absorption lines are weak at low state , and negligible at high state ( because c and n are highly ionized ) .
there is always some absorption by hydrogen due to the large column density in h. \(b ) following spitzer ( 1978 ) , a standard curve of growth was calculated for velocity parameters b = 20,60,100,140 km s@xmath7 ( figs .
the predicted equivalent widths of ly@xmath4 , civ , nv , and ( ovi ) were then compared with those derived from the analysis of the hst spectrum .
we find that at the rosat low - state ( which is close to the state of the source at the time of the asca observations ) , and for b about 60 km s@xmath7 , there is a rough match between the ly@xmath4 and civ absorption lines produced by the warm absorber and the observed lines at @xmath11 = 0.098 and @xmath11 = 0.090 ( models c - e ) .
the nv absorption line from the warm absorber is , however , always weaker than observed ( n is too highly ionized in the models ) . including an additional euv bump will further increase the level of ionization , thus not changing the above conclusions .
the different models differ most strongly in ovi , so this may be the most restrictive line , but it falls just outside the hst range.in conclusion , we find no single - phase medium which can produce both the uv and the x - ray absorption lines .
\(c ) in the case of the best fit to the asca data i.e. the model with fe overabundant by a factor of @xmath2 22 over solar , the degree of ionization is not known but it is likely to be too high for the production of the ciii@xmath97 absorption line .
thus , in this case also , the uv and x - ray absorbers are very probably in different gaseous phases . tab .
5 gives the intensity of the strongest lines emitted by the absorber within the wavelength range of the hst spectra and in the optical range .
the calculations were performed with a density of log n@xmath25 = 9.5 .
the line neviii@xmath93774 was added to the list because of recent reports of its detection in high-@xmath11 quasars ( hamann et al .
table 5 is meant to provide an order of magnitude estimate of which lines may be important / detectable in the future .
the actual strength of the lines depends on the covering factor of the warm absorber ; total coverage was assumed for the values in table 5 . in the recent years
it has been realized that gas outflow is ubiquitous in agn .
it occurs under different gas phases : broad emission line gas ( the most highly ionized lines are the most blueshifted ) and uv / optical absorption lines ( always observed at rest or blueshifted ) .
it is likely that the x - ray absorption features also , originate in outflowing gas at velocities comparable with or higher than those of the uv emission / absorption gas - this can not presently be ascertained because of the insufficient energy resolution of the x - ray instruments .
[ this is not counting here the extraordinary blueshift of the x - ray absorbing gas if the 1 kev features are blueshifted ovii or ovii edges ] .
the question as to whether uv absorption lines and x - ray absorption lines / edges tend to be present together or separately in agn has been adressed by ulrich ( 1988 ) and more recently by a number of authors ( e.g. schartel et al .
1997 , crenshaw 1997 , mathur 1997 , shields & hamann 1997 , kriss et al 1996 ) . in some agn
, the data appear consistent with a single - phase , photoionized plasma producing the uv absorption and the ovii and oviii edges ( ngc 3783 shields & hamann 1997 ; 3c212 , 3c351 and ngc 5548 mathur 1997 ) . in contrast , in other agn , the properties of the uv and the x - ray absorbers imply the presence of multiphase media ( ngc 3516 kriss et al 1996 ; mcg-6 - 3 - 15
otani et al 1996 ) .
the present analysis indicates that pg1404 + 226 is another such case of multiphase absorbing medium ( but the non - simultaneity of the observations has to be kept in mind ) . with the detection of x - ray and uv absorption in pg1404 + 226
the statistical association between the presence of uv absorption lines and x - ray absorption edges becomes stronger ( ulrich 1988 , mathur wilkes & elvis 1998 ) .
the two absorbers could be two different gaseous phases , partaking in the same outflow but differing by their physical conditions , velocity and radial distance to the central black hole .
the main results of our analysis can be summarized as follows : \1 ) the x - ray spectrum of pg 1404 + 226 is variable by a factor 4 in @xmath98 10@xmath99s and is characterized by a strong soft excess below 2 kev whose luminosity ( @xmath100 erg s@xmath7 in the 0.42.0 kev energy range ) is about a factor 3 greater than the 210 kev luminosity ( @xmath101 ) . the soft excess emission can be described with a high temperature optically thick blackbody ( @xmath102 ev ) .
optically thin models are ruled out combining the observed luminosity with the dimension of the region derived from the variability timescale .
\2 ) the residuals around 1 kev can be best described by an absorption edge at @xmath103 kev , _ not _ consistent with being caused by highly ionized oxygen at rest in the quasar frame .
a possible explanation could be either in terms of iron overabundance , as suggested by the warm absorber fits and the extremely high ew of the iron k@xmath4 line , or by resonant absorption in a turbulent gas .
the interpretation of a blueshifted oxygen edge in a relativistically outflowing gas is less likely and not supported by the optical uv data .
x ray observations of nls1 at high spectral resolution with xmm and _ chandra _ will allow to clarify the origin of the 1 kev absorption detected in pg 1404 + 226 and 3 other nls1 .
\3 ) two systems of absorption lines separated by @xmath6 km s@xmath7 are identified in the fos / hst spectra in the lines of ly@xmath4 , civ and nv .
one system is located to the red of the emission line peaks . considering that in most radio quiet qsos
the highly ionized emission lines are blueshifted ( as part of an outflow from the face of the accretion disk turned toward us ) we argue that this absorption system is nearly at rest in the agn frame .
its properties are consistent with this absorber being produced in the halo of the host galaxy or a companion . as for the system blueshifted by @xmath6 km s@xmath7 , the very probable detection of ciii@xmath97@xmath931175 ( which has been seen in absorption only in ngc 4151 - bromage et al .
1985 , kriss et al . 1992 ) indicates that this system is intrinsic to the quasar .
\4 ) with the detection of x - ray and uv absorption in pg1404 + 226 the statistical association between the presence of uv absorption lines and x - ray absorption edges first suggested by ulrich ( 1988 ) is becoming clearer ( mathur wilkes & elvis 1998 ) .
we may be seeing two different absorbers with different physical conditions and locations but both being parts of a grand design outflow .
the differences observed among absorption features in various agn are likely to result from intrinsic differences in the properties of the gaseous outflows , from differences in the aspect angles and from the shape of the x ray spectrum .
we thank gary ferland for providing cloudy .
ac acknowledges partial support by the italian space agency ( asi ) under the contract asi - ars-96 - 70 and by the italian ministry for university and research ( murst ) under grant cofin98 - 02 - 32 .
k. acknowledges support from the verbundforschung under grant no .
50 or 93065 , and p.c . acknowledges support from nasa through contract nas5 - 26670 .
boller t. , brandt w.n .
, fink h. , 1996 , a&a , 305 , 53 boroson t.a .
, green r.f . , 1992 , apjs , 80 , 109 brandt w.n .
, fabian a.c . , nandra k. , reynolds c. s. , brinkmann w. , 1994 , mnras 271 , 958 brandt w.n . , mathur s. , elvis m. , 1997 , mnras 285 , l25 bromage g.e .
, boksenberd a. , clavel j. , et al .
1985 , mnras , 215 , 1 comastri a. , setti g. , zamorani g. , et al 1992 , apj 384 , 62 comastri a. , molendi s. , ulrich m .- h . , 1997 , ( cmu97 ) in _ x - ray imaging and spectroscopy of cosmic hot plasmas _ , f. makino , k. mitsuda ( eds ) , 279 comastri a. , fiore f. , guainazzi m. , et al 1998 , a & a 333 , 31 comastri a. , brandt , w. n. , leighly , k. m. , et al .
1998 , proceedings of the conference _ the active x - ray sky , results from bepposax and rxte _ , scarsi l. et al eds nuclear physics b suppl .
69/1 page 523 corbin m.r . , 1995 , apj , 447 , 496 crenshaw d.m . ,
1997 in _ mass ejection from active galactic nuclei _
, arav n. , shlosman i. , weymann r.j .
, eds , asp conf . series , vol . , 128 , 121 czerny b. , witt h. j. , zycki p. t. , 1996 integral workshop , _ the transparent universe _ , st .
malo , september 1996 done c. , mulchaey j. s. , mushotzky r. f. , arnaud k. a. , 1992 , apj 395 , 275 elvis m. , lockman f.j .
, wilkes b.j . , 1989 , aj , 97 , 777 ferland g.j .
1993 , university of kentucky , physics department , internal report fink h. , walter r. , schartel n. , engels d. , 1997 , a&a , 317 , 25 fiore f. , matt g. , cappi m. , et al 1998 , mnras , 298 , 103 grupe d. , 1996 , phd thesis ( university of gttingen ) haardt f. , maraschi l. , 1993 , apj , 413,507 hamann f. ferland g.j .
1993 , apj , 418 , 11 hamann f. , cohen r.d . , shields j.c .
, burbidge e.m . , junkkarinen v. , crenshaw d.m . , 1997 ,
apj , 496 , 761 hayashida k. , 1997 in _ emission lines in active galaxies : new methods and techniques _ peterson b.m . , cheng f .- z . , wilson a.s .
, eds , asp conference series vol .
113 , 40 vol .
113 , 40 komossa s. , fink h. , 1997 a&a , 322 , 719 kriss g.a .
, davidsen a. f. , blair w. p. , et al . , 1992 , apj , 392 , 485 kriss g.a . , espey b. r. , krolik j. h. , et al 1996 , apj , 467 , 622 krolik j.h . , kriss g.a . , 1995 ,
apj , 447 , 512 laor a. , fiore , f. , elvis , m. , et al 1994 , apj 435 , 611 laor a. , bahcall j. n. , jannuzi b. t. , et al 1995 , apjs 99 , 1 laor a. , januzzi b.t . , green r.f .
, boroson t.a . , 1997 in _ emission lines in active galaxies : new methods and techniques _
peterson b.m .
, cheng f .- z . , wilson a.s . , eds , asp conference series vol . 113 , 116 leighly k.m . , mushotzky r.f . , nandra k. , forster k. , 1997 , apj , 489 , l25 marshall f. e. , netzer h. , arnaud k. a. , 1993 , apj , 405 , 168 mathur s. , 1997 in _ mass ejection from active galactic nuclei _ asp conf . series , vol .
, 128 , 161 arav n. , shlosman i. , weymann r.j . , eds asp conf .
series , vol .
, 128 , 161 mathur s. , wilkes b. , elvis m. , 1998 , apj , 503 , l23 netzer , h. , turner t. j. , 1997 apj , 488 , 694 nicastro f. , fiore f. , perola g. c. , elvis m. , 1999a , apj 512 , 184 nicastro f. , fiore f. , matt g. , 1999b , apj 517 , 108 osterbrock d.e .
, pogge r. , 1985 , apj . , 297 , 166 otani c. , kii t. , miya k. , 1996 in _ intern .
conf . on x - ray astron . and astrophys . , mpe report , 263 _ zimmermann h .- u . , trmper j.e .
, yorke h. , eds . ,
491 pounds k.a .
, done c. , osborne j.p . , 1995 , mnras , 277 , l5 reynolds c.s . , fabian a.c . , 1995 , mnras , 273 , 1167 reynolds c.s . ,
fabian a.c . , inoue h. , 1995 , mnras , 276 , 1311 reynolds c.s . , 1997 , mnras 286 , 513 schartel n. , komossa s. , brinkmann w. , fink h.h . , trmper j. , wamsteker w. , 1997 , a&a 320 , 421 shields j.c . , hamann f. , 1997 , apj , 481 , 752 shuder j.m . ,
osterbrock d.e . , 1981 , apj . ,
250 , 55 spitzer l. , 1978 , physical processes in the interstellar medium .
john wiley sons , new york sulentic j.w . ,
marziani p. , zwitter t. , calvani m. , 1995 , apj , 438 , l1 tanaka y , inoue h. , holt s.s .
, 1994 , pasj 46 , l37 ulrich m .- h .
, 1988 , mnras , 230 , 121 ulrich m .- h .
, molendi s. , 1996 , apj , 457 , 77 ( um96 ) van groningen e. 1987 , a&a , 186 , 103 zdziarski a. , ghisellini , g. , george , i.m .
et al . 1990 , apj , 363 , l1 | we present and discuss our observations of the narrow line quasar pg1404@xmath0226 ( @xmath1 = 0.098 ) with asca and hst , and a re - analysis of our earlier observations with rosat .
the soft x - ray spectrum is very steep and displays an absorption feature ( edge or line at @xmath2 1.1 kev ) .
we have applied a variety of models to the asca and rosat spectra without finding a completely satisfactory fit , and the identification of the edge remains uncertain .
a satisfactory fit of the asca spectrum assuming that the edge is produced by highly ionized iron ( using the code absori in xspec ) is obtained with an overabundance of iron by a factor @xmath3 compared to solar , a suggestion supported by the extremely high equivalent width of the fe k@xmath4 line at 6.4 kev . a warm absorber model ( based on cloudy ) fitting the absorption feature with nevii - nex edges and assuming a peculiar oxygen / neon abundance ratio is consistent with the rosat data but not the asca data . finally , it is also possible that the observed edge is caused by a oviii or ovii edge or line , blueshifted by @xmath5 = 0.2 to 0.5 depending on the specific identification , as has been suggested previously for 2 other narrow line quasars , but there are no other features in the uv and x - ray spectra in support of this suggestion .
two systems of uv absorption lines , one nearly at rest in the source frame , the other blueshifted by @xmath6 km s@xmath7 are identified in the hst / fos spectra .
photoionization models indicate that the uv absorption and the @xmath8 kev absorption are probably caused by absorbers with different physical conditions .
pg1404@xmath0226 is one more case of agn where both uv and x - ray absorption features are detected , thereby increasing further the significance of the previously noted statistical association of the two types of absorbers . |
reference frames are one of the most basic notions in physics . almost every time a measurement is performed or a formula is written down , a reference frame has been implicitly used .
the choice of reference frame can either be thought of as abstract labelling of space - time , or a description relative to some physical laboratory with a particular set of rulers and clocks . in the latter case ,
the laboratory is always taken to be much heavier than the system under observation , with a well defined velocity and position .
in other words , it is assumed to be fundamentally _
classical_. however , in their seminal papers @xcite , aharonov and susskind showed that the concept of reference frame can be suitably accommodated in quantum theory .
although our everyday reference frames are classical for all practical purposes , aharonov and susskind have shown that theoretically nothing prevents the idea of a _ quantum reference frame_. these early works were followed by @xcite and culminated in the work of aharonov and kaufherr @xcite that is of particular relevance to us here .
since then much progress has been made by investigating the role of quantum reference frames from both foundational and applied viewpoints .
several works have investigated fundamental properties of quantum reference frames @xcite .
also , quantum reference frames have been studied in quantum information @xcite , in particular in the context of quantum communication @xcite and entanglement detection @xcite . in the present paper
we go back to the original problem discussed in @xcite , namely that of relative positions and momenta , and ask what physics looks like from an isolated quantum reference frame . in particular , we consider states composed of a small number of quantum systems , then promote one of these systems to the role of a _ quantum observer _ and examine the description of the state from its internal perspective .
for example , we will consider experiments performed within a freely - floating quantum rocket , from the perspective of the rocket itself , making no mention of any external reference frame whatsoever .
we find that we are easily lead into paradoxes when trying to analyse experiments from the perspective of such a quantum observer . by careful reanalysis
, we resolve these paradoxes and in doing so learn much about the physics of quantum reference frames .
let us consider a simple interference experiment , in which a particle is sent in superposition towards an interferometer .
although such experiments usually involve photons , for simplicity here we consider the particle to be non - relativistic .
the interferometer consists of two mirrors , a beam splitter , and two detectors , as depicted in fig .
[ f : mz ] ( a ) .
we take these components to be rigidly fixed together , but free to move relative to the laboratory ( which we treat as a classical external reference frame ) .
we initially place the centre of the interferometer at the origin of the laboratory , and denote the position of the left and right path of the particle by @xmath0 and @xmath1 respectively .
the initial state of the interferometer and particle relative to the laboratory can then be written @xmath2 where @xmath3 denotes a narrow wavepacket for the interferometer / particle centred at @xmath4 ( in the horizontal coordinate ) .
this description of the experiment is from the point of view of the laboratory , but what we are actually interested in is the description of this experiment from the perspective of the interferometer and the particle themselves .
since neither the interferometer nor the particle interacts with the laboratory we should not need to use it to describe the experiment . to this end we ask the following questions : a ) what does the interferometer see ?
b ) what does the particle see ? let us begin by first clarifying precisely what we mean by our terminology that the the interferometer or particle ` sees ' .
we use this rather colloquial terminology to mean the description of the physics relative to the system in question , i.e. the relative positions and momenta of the rest of the system .
assuming that the interferometer is much heavier than the particle , the answers are as we might expect .
the interferometer sees the particle approaching it in a superposition of positions @xmath0 and @xmath1 , while the particle sees the inverse the interferometer approaching it in a superposition of positions @xmath1 and @xmath0 .
now consider the slightly different experimental setup shown in fig .
[ f : mz ] ( b ) . in this case
, the particle is positioned at the origin of the laboratory , while the centre of the interferometer is placed in a superposition of positions @xmath0 and @xmath1 .
the state of the joint system for this second experiment is @xmath5 one might think that as far as the particle and interferometer are concerned this second situation would look the same as the first experiment , with the interferometer seeing the particle approaching it in a superposition of positions @xmath0 and @xmath1 , and the particle seeing the interferometer in a superposition of @xmath1 and @xmath0 as before . since it is only the particle and interferometer that are relevant , and not the external frame , we would expect the two experiments to yield identical results . indeed ,
if we were to place the detectors instead at the location of the mirrors then both situations would be equivalent ; in both cases both detectors would be equally likely to click .
surprisingly , however , if we follow the evolution in both cases we see a dramatic difference .
this is illustrated from the perspective of the laboratory in figure 2 . in the first experiment ,
the two particle paths interfere at the beam splitter and only the left detector clicks , while in the second experiment there is no interference between the photon paths and both detectors are equally likely to click .
as all perspectives must agree on the probability of detector clicks , the second experiment must also look different to the first from the perspective of the particle and the interferometer .
in other words , experiments ( a ) and ( b ) are not equivalent , contrary to what we expected .
but why not ? to find out , we consider in more detail the correct description of the particle relative to the interferometer for both experiments . to do this we change from laboratory co - ordinates to the relative coordinate of the particle as seen by the interferometer , @xmath6 , and the centre of mass , @xmath7 .
these are given by @xmath8 where @xmath9 and @xmath10 are the mass and position of the interferometer / particle , and @xmath11 .
one should note that when changing co - ordinates subtle effects due to the finite width of the wavepackets emerge . consider a state of the form @xmath12 , by which we actually mean @xmath13 , i.e. two wavepackets , centred around @xmath14 and @xmath15 , hence the centre of mass
is located around @xmath16 and the relative co - ordinate is centred around @xmath17 .
however , in this state there is an uncertainty in the position of each particle , hence there is an uncertainty in the position of the centre of mass and relative distance ; in fact the centre of mass and relative position will in general be entangled .
more precisely , in relative co - ordinates the state @xmath13 becomes @xmath18 which will in general not factorise as there will be entanglement between the _ fluctuations _ of the centre of mass and relative coordinate .
entanglement between the centre of mass and relative distance will turn out to be crucial in what follows .
however , it is not this entanglement between the fluctuations but the entanglement that appears when we consider superpositions of localised wavepackets , such as @xmath19 that is crucial in our paper . in other words ,
our concern is the entanglement between the _ mean _ positions of the centre of mass and relative co - ordinates , not the entanglement of the fluctuations . for the sake of clarity , from here on we thus work under the following approximation , @xmath20 , which amounts to ignoring the entanglement between the fluctuations .
importantly , this approximation has no consequences on the physics we present below ; for a more detailed analysis , including finite width effects , see the appendix . given the above , in relative co - ordinates the two experimental situations are described by the states @xmath21 the crucial point is to realise that there are degrees of freedom that are accessible to the internal observer and degrees of freedom which are not and to distinguish between them . on the one hand , it is clear that the position of the centre of mass is a degree of freedom that is not accessible inside as it explicitly refers to the external reference frame , i.e. the laboratory .
indeed the position of the centre of mass can neither be determined nor altered by any internal observer that has no access to any external frame . on the other hand ,
the relative position between the particle and interferometer is certainly accessible as it does not refer to the laboratory ( or any other external frame ) .
therefore to correctly describe the physics of the internal observer we must trace over the inaccessible degrees of freedom , i.e. the centre of mass .
it is the reduced state , @xmath22 which accurately describes the state of the particle relative to the interferometer .
similarly by taking instead @xmath23 we find the state of the interferometer relative to the particle . now ,
if we consider the regime where the mass of the interferometer is much greater than that of the particle , that is @xmath24 , we require @xmath25 then we see that the states ( 5 ) and ( 6 ) become , approximately , @xmath26 thus we see that in experiment ( a ) upon tracing over the centre of mass we are left in the pure state @xmath27 .
in this situation the interferometer sees the particle in a coherent superposition , hence we observe interference and only one of the detectors clicks . on the other hand however , in experiment ( b ) upon tracing over the centre of mass we are left in a mixed state @xmath28 representing an equal chance of the particle impinging upon the left and right mirrors . here
no interference can take place and hence both detectors are equally likely to click .
note that the relative state of the interferometer as seen by the particle is exactly the same , up to the sign of the relative position , and thus the same conclusions also hold from this perspective . hence both internal observers will make the same predictions as the external observer about the experimental outcome .
we see that the reason why the internal observers make different predictions in the two experiments is due to the fact that relative states in each experiment are different ; one is a coherent superposition , whilst the other is a mixture .
we are thus led to realise that the centre of mass plays a crucial role when considering what the interferometer or particle sees , for it is the entanglement of the relative coordinate with the centre of mass coordinate which makes the two states ( 5 ) and ( 6 ) completely different .
below , we provide another thought - experiment which shows explicitly how interference disappears whenever the centre of mass is entangled with the internal degrees of freedom .
this is in order to make it clear that the disappearance of interference is due to the entanglement between the centre of mass and relative co - ordinate which was already present in the initial state of the system , and not due to its subsequent interactions , such as the reflection of the particle by the mirrors .
consider the situation depicted in fig .
[ rocket ] in which a particle of mass @xmath29 moves freely inside a stationary rocket of mass @xmath30 . after the preparation of their initial state ,
the particle and the rocket no longer interact with the external world .
for an external observer this is a _ quantum _ rocket ,
i.e. the position of the rocket s centre of mass is given by a narrow wavepacket centred at the origin of the external frame , with a quantum uncertainty @xmath31 .
suppose the particle has been prepared in a coherent superposition relative to the external observer .
the two wave packets composing the superposition move towards each other with constant average momentum @xmath32 .
initially , they are symmetrically located at a distance @xmath1 apart from the origin of the external frame .
a quantum rocket is prepared in a well localised state , whose centre of mass is here depicted by the narrow red wavepacket . inside the rocket is a particle in superposition , here depicted as the blue wavepackets , initially located a distance @xmath1 from the centre of the rocket , with momemtum @xmath32 towards the centre . at a time @xmath33 when the wavepackets interfere either ( a ) the centre of mass is sufficienty localised to resolve the fringes or ( b ) the centre of mass is too uncertain to resolve them .
[ rocket ] ] after a period of time @xmath34 , given by @xmath35 the two wave packets meet at the origin and from the external observer s perspective , they form an interference pattern corresponding to the wavelength @xmath36 , in units such that @xmath37
. then the question arises whether an observer inside the rocket would be able to see the interference pattern .
intuitively we expect the uncertainty in the rocket s position at the time the two wavepackets interfere to be the limiting factor on the ability of the internal observer to measure the interference , that is the rocket can only resolve details significantly larger than its uncertainty @xmath38 relative to the external frame .
thus , this implies that the internal observer can only see the interference pattern when the uncertainty in the rocket s position is much smaller than the wavelength @xmath39 of the particle , namely @xmath40 where @xmath38 denotes the variance in the position of the rocket at the instant @xmath34 .
this is situation ( a ) in fig .
[ rocket ] . on the other hand , in situation ( b )
when @xmath41 , although from the external frame of reference there is still interference , from inside the rocket this can no longer be seen .
what we shall now see is that whether we end up in situation ( a ) or ( b ) depends only upon whether or not the centre of mass of the rocket is entangled with the relative co - ordinates .
let us assume that the rocket is initially in a minimum uncertainty wave packet ( @xmath42 ) .
using the approximation @xmath43 we now re - write eq .
as @xmath44 note that both terms on the left - hand side are positive , hence each must be much smaller than the right - hand side . using equation
this implies that @xmath45 this means that the interference pattern can indeed be resolved by an observer in the rocket provided that the initial uncertainty in the rocket s position obeys these relations .
given that we take the mass of the rocket to be much larger than the particle , then @xmath46 and @xmath47 , and eq .
implies that @xmath48 here @xmath49 represents by how much the centre of mass is displaced from the origin when the particle is prepared to the left or to the right .
therefore the meaning of ( 13 ) is that this displacement is smaller than the original uncertainty in the centre of mass , hence the centre of mass is insensitive to whether the particle is to the left or the right , i.e. it decouples from the relative position .
again we see that the entanglement between the centre of mass and the relative co - ordinates determines whether or not there will be interference
. however , this is by no means the end of the story , as we shall see in the next section .
in the previous section we have seen that the centre of mass , in particular its entanglement to the relative co - ordinates , plays a central role in the physics of quantum reference frames .
however , we shall see now that this is not the only issue in understanding quantum reference frames .
contrary to the situation considered so far we shall envisage a thought experiment in which the centre of mass is physically irrelevant , yet we run into an even more intriguing paradox .
let us consider now two particles , of mass @xmath50 and @xmath51 , prepared in the following entangled state : @xmath52 where the state @xmath53 denotes that particle @xmath54 is sharply localized at position @xmath4 relative to the laboratory , and @xmath55 is an arbitrary relative phase .
we choose the positions @xmath14 and @xmath15 such that @xmath56 and hence the centre of mass of the particles is at the origin of our co - ordinate system , as shown in fig .
[ f : particles]a .
we also denote the distance between the particles by @xmath57 , which allows us to write @xmath58 so far the description of the system has been given relative to an external reference frame .
again we would like to promote particle 1 to the role of a _ quantum observer _ and understand how _ it _ would describe the state of particle 2 .
( a ) the two particles are in an entangled state , such that the centre of mass of the system is unentangled from the relative co - ordinates .
( b ) a third , rather innocuous , particle is added in the picture , leaving the centre of mass unentangled with the relative co - ordinates . ]
thus we move again from absolute to relative positions .
we therefore describe our system using the position of the centre of mass @xmath59 and the relative position @xmath60 , given by @xmath61 we can rewrite the state of the system in these new co - ordinates as @xmath62 now we are in position to describe what particle 1 would observe without access to the external reference frame , i.e. by tracing over the centre of mass .
first , from eq .
we see that the position of the centre of mass and the relative position actually decouple , i.e. the global state is a product state .
therefore we conclude that particle 1 sees particle 2 in a pure state .
importantly this implies that particle 1 can get access to the phase @xmath55 by interacting with particle 2 alone , i.e. without access to the external reference frame .
now let us bring a third , and rather innocuous , particle into the game
. this particle has mass @xmath63 , and is located at position @xmath64 relative to the external frame ( see fig .
particles 1 and 2 are exactly as previously . in the external reference frame , the global state of the system
is now given by @xmath65 in order to understand what particle 1 would observe , we move again to relative co - ordinates , given by @xmath66 where @xmath67 . note that @xmath60 and @xmath68 are the positions of particle 2 and 3 relative to particle 1 .
the state of the system , expressed in these new
co - ordinates is @xmath69 again , let us distinguish those degrees of freedom which are accessible to particle 1 from those which are not .
first of all , we note that the inaccessible degree of freedom , i.e. the position of the centre of mass , factorizes once again .
however , the surprising fact now is that the accessible degrees of freedom the positions of particles 2 and 3 relative to particle 1 are now entangled .
consequently it appears that particle 1 must do a joint measurement on particles 2 and 3 to determine the phase @xmath55 .
if particle 1 can not access particle 3 , for instance when these two particles are far apart , it seems that we should trace over @xmath59 and @xmath68 , giving the reduced state @xmath70 which contains no information about the phase whatsoever .
here the paradox emerges . since the innocuous particle 3 is completely uncorrelated from particles 1 and 2 , adding it into the description of the system should clearly not change the physics .
however from the above discussion , it appears that the addition of particle 3 leads to dramatically different situations from the perspective of the internal observer .
on the one hand , when considering only particles 1 and 2 , we concluded that particle 1 can gain information about the phase @xmath55 . on the other hand , when particle 3 is included ( but inaccessible to particle 1 ) , it appears that particle 1 can not extract any information about the phase @xmath55 .
clearly this is in flagrant conflict with the predictions of the external observer , who would argue that particle 1 should be able to extract exactly the same information about @xmath55 in both cases , as it is a property of particles 1 and 2 alone . in the following subsections
we shall identify the optical illusion which is actually present and hence solve the paradox . in doing so we will learn more about the nature of quantum observers .
one may suspect that the root of the problem resides in the phase measurement ; indeed above we never explicitly showed how the phase could be measured . in the following we address this aspect in detail .
we focus initially on the case of two particles , for which we concluded that particle 1 has access to the phase , and see how particle 1 would proceed .
let us consider again the state of the system in relative co - ordinates , i.e. eq . .
measuring the relative position of particle 2 provides no information about the phase . instead
, the information about the phase is encapsulated in the relative momentum , @xmath71 where @xmath72 is the reduced mass . however
, this information is encoded in an intricate way : the expectation values of the moments of momentum , @xmath73 , are all independent of the relative phase .
the function of momentum whose expectation is sensitive to the phase is the relative shift operator @xmath74 where @xmath75 @xcite .
this operator has the effect of shifting the position of particle 2 versus particle 1 by a distance @xmath76 , causing one wavepacket in the original state to overlap the other in the shifted state , and revealing information about @xmath55 in its expectation value : @xmath77 where in the first line we use the approximation .
note that the shift operator is non - hermitian , the corresponding hermitian operators are @xmath78 and @xmath79 . in practice instead of measuring them directly , one would perform an interference experiment . from the perspective of the external observer , we can also understand this result , by noting that @xmath80 corresponding to a shift of particle 1 by @xmath81 , and particle 2 by @xmath82 in the external reference frame . from equation ( [ phi ] ) we see that @xmath83 so far everything appears perfectly consistent .
now let us move to the case of three particles . clearly , particle 1 can still measure the relative velocity of particle 2 , even when it has no access to particle 3 .
thus particle 1 can still measure the relative shift operator @xmath84 .
furthermore , still holds in the three particle case , hence using the state as seen by the external observer , we have @xmath85 this fits with the view that particle 1 can measure the phase without interacting with particle 3 .
the paradox arises when we look at the same measurement in terms of the relative state .
if @xmath86 only acts on the relative coordinate of particle 2 , it will see only the reduced state given by , which has no dependence on the phase .
we now present the solution to the paradox .
the crucial ( and surprising ) observation is that @xmath86 actually shifts the relative coordinate of particle 3 as well as that of particle 2 .
the root of the problem is that we have been considering relative position and momentum operators which are not canonically conjugate to each other .
when we considered the case of only two particles @xmath87 was the momentum canonically conjugate to @xmath88 .
however , when we move to the situation with three particles , this is no longer the case , as @xmath87 does not commute with @xmath89 .
indeed it can be immediately checked that @xmath90 = i\frac{\mu_{12}}{m_1 } \neq 0 .
\end{aligned}\ ] ] at this point it is important to recall that canonical observables play a fundamental role , because they are the building blocks of the hilbert space and underlie its structure .
specifically , the tensor product structure of the hilbert space is defined through a set of canonical observables ; the eigenstates of these observables define the natural basis . from this point of view
, it now appears judicious to identify the relative momentum operators canonically conjugate to the relative position operators @xmath91 , @xmath88 and @xmath89 . that is , we are looking for operators @xmath92 , and @xmath93 satisfying the canonical commutation relations @xmath94=i\delta_{\alpha\beta}$ ] .
a lengthy but straightforward calculation leads to the following set of operators : @xmath95 next , let us re - express the relative shift operator @xmath96 in terms of these canonically conjugate momentum operators : @xmath97 where we have used the fact that @xmath98 . from the above expression , it becomes clear that @xmath86 acts _ non - locally _ on the hilbert space spanned by the relative co - ordinates @xmath60 and @xmath68 , shifting both co - ordinates .
previously we argued that the inaccessibility of particle 3 by particle 1 would allow us to trace over @xmath68 .
however , this argument does not hold , because the relative shift operator @xmath99 also affects @xmath68
. we can now finally understand from particle 1 s perspective how it can extract the phase from the entangled state @xmath100 , as @xmath101 and therefore @xmath102 in agreement with .
this solves the paradox . from particle 1 s perspective
, it seemed impossible to retrieve the phase without interacting with particle 3 , due to the entanglement present .
however , here we see that an operation that particle 1 can carry out on particle 2 alone ( essentially a measurement of its relative velocity ) , also indirectly affects the relative position of particle 3 , allowing the phase to be extracted even when particle 3 is physically inaccessible .
the final piece of the puzzle is to understand physically the meaning of the apparent non - locality of the shift operator ; more precisely why it acts not only on the relative co - ordinate @xmath60 but also on @xmath68 .
the key observation is that particle 1 has a finite mass .
thus when it shifts particle 2 it also necessarily shifts itself , since the centre of mass of particle 1 and 2 must remain fixed . this shift in particle 1 s position changes the relative position of particle 3 versus particle 1 . in other words , when particle 1 performs an interference experiment on particle 2 , the relative distance between particles 1 and 3 is also necessarily changed .
finally , it is worth noting that in the limit where particle 1 becomes very massive , and thus experiences no back - action , the non - locality of the hilbert space disappears , i.e. the observables @xmath103 and @xmath104 become once again canonically conjugate , as seen from eq . .
the above illustrates that the hilbert space has a rich structure . in the two particle case the hilbert space possesses the normal bipartite structure that we expect ,
namely @xmath105 , where @xmath106 describes the centre of mass degrees of freedom and @xmath107 describes the degrees of freedom of particle 2 relative to particle 1 . in particular , the eigenstates of relative position and of relative momentum belong to the relative hilbert space , @xmath108 , @xmath109 , , and our statements can be taken to refer to these . ] and each can be written as a superposition of the other , e.g. @xmath110 , where @xmath111 .
one is tempted to think that when we add a third particle and go to relative co - ordinates all we have to do is simply add on the hilbert space for the relative degrees of freedom for this third particle , i.e. that @xmath112 . however
, this can not be the case indeed from eq .
we already know that the operator @xmath87 does not commute with @xmath89 , hence they can not live in two different hilbert spaces .
the hilbert spaces @xmath107 and @xmath113 therefore can have no meaning .
in other words it is now no longer possible to have a hilbert space such as @xmath107 which contains both eigenstates of the relative position and momentum of particle 2 relative to particle 1 ; this also implies that the decomposition @xmath110 is no longer possible .
in fact there exists two different decompositions of the hilbert space , @xmath114 and @xmath115 . in the first decomposition , the spaces are spanned by the eigenstates of relative position , @xmath88 and @xmath89 and of their conjugate momenta , @xmath116 and @xmath93 . in the second decomposition
the spaces are spanned by the eigenstates of relative momenta , @xmath87 and @xmath117 and their conjugate position , which we shall denote @xmath118 and @xmath119 . following from the above
, it is insightful to reverse the problem , that is to start from the relative momentum operators and find explicitly their conjugate relative positions .
let us first complete the set of relative momentum operators : @xmath120 where @xmath121 is the total momentum , and @xmath87 and @xmath117 are the relative momenta of particles 2 and 3 versus particle 1 .
now we look for the relative position operators canonically conjugate to these relative momentum operators ( i.e. satisfying @xmath122=i\delta_{\alpha\beta}$ ] ) .
again , a lengthy but straightforward calculation leads to @xmath123 with @xmath124 .
re - writing the state in terms of these new @xmath125 co - ordinates , we find that @xmath126 where @xmath127 .
note that with these co - ordinates , the state is fully separable . in this basis , the shift operator acts ` locally ' on the @xmath118 coordinate : @xmath128 and hence @xmath129 as expected in conclusion , we have seen that the relative position and relative momentum operators , which are the physically relevant operators from the perspective of an internal observer , are not canonically conjugate .
this has the consequence that some operators ( such as shifting particle 2 relative to particle 1 ) have an apparently ` non - local ' effect .
consequently , one can not trace over the relative positions of inaccessible particles to give a description of the remaining accessible degrees of freedom .
note that in the classical case , although there is no entanglement , it is still true that the usual relative positions and momenta are not canonically conjugate .
there are two crucial distinctions however : firstly , a quantum measurement of @xmath87 must disturb @xmath89 due to their non - commutativity , whilst classically a measurement can always be taken to produce an arbitrarily small disturbance .
secondly , the classical relative positions and momenta of particle 2 and 3 can all be independently specified , whilst quantum mechanically there is no state in which @xmath87 and @xmath89 are both simultaneously well defined .
hence in quantum theory , the relative state is inherently non - local , with particle 2 and particle 3 intertwined .
in all the cases we have considered , we have begun by defining the state in an external frame , and then moved to relative co - ordinates .
when we refer to the centre of mass coordinate , we are really describing its position in this external frame .
but does our analysis depend on which external frame we initially choose ?
do we require an external frame of reference at all ?
if so , what properties should it have ?
these issues are of particular importance in section ii , where the experimental results depend critically on whether the relative coordinate is entangled with the centre of mass coordinate or not .
the first thing to note is that if the centre of mass is in a superposition of different places in relation to one frame of reference it is also in a superposition relative to any other frame of reference that is classically related to the first frame , i.e. which has a well defined displacement and speed relative to this frame a galilean transformation .
so certainly we do not single out one particular frame amongst all the frames related by a galilean transformation .
however , in quantum mechanics , one might consider alternative ` non - classical ' transformations of the external reference frame such as moving to a frame in a superposition of different locations relative to the original frame @xcite .
looking at figure ( 1b ) it is possible to imagine a frame of reference in which the interferometer is not in a superposition but located at the origin just go to a frame of reference which is in a superposition versus the original one and correlated with the location of the interferometer . in this frame of reference , the position of the centre of mass will be in the same place in each branch of the superposition , and hence it will factorise .
does it mean that we are now able to measure interference ?
of course , whether or not we are able to see interference is just a question of internal measurements and should be independent of the external reference frame , so there is an inconsistency .
does this therefore mean that such reference frames which are in a superposition are not permitted ?
is it the case that there exists an absolute _ set _ of reference frames which are related by galilean transformations but all other frames which are in superposition relative to them do not exist ?
we seem to have run into a serious problem .
these problems have several aspects , and below we will address them in turn .
the first issue is whether we can determine the existence or properties of the external frame from inside the system alone .
i.e. without interacting directly with the external frame . unsurprisingly , the answer to this question is no . as in classical mechanics
, we are at liberty to describe any physical situation only in terms of relative co - ordinates . for each of the physical situations we consider ,
the correct relative description is obtained by tracing out the centre of mass degree of freedom , yielding a state of only the relative degrees of freedom .
the hamiltonian can also always be decomposed into two commuting terms , one of which is a function of the total momentum alone , and the other is a function of only the relative positions and momenta . keeping only the second term will correctly describe the evolution of all the relative degrees of freedom without reference to the centre of mass . however , our results highlight two curious aspects of this approach : ( i ) that some situations which can be described by pure states in an external frame must be described using mixed states in terms of relative co - ordinates @xcite .
( ii ) that the concept of locality is much harder to pin down in terms of relative co - ordinates , as the relative position and momentum of one particle do not commute with those of the other particles .
hence we can not simply trace out the relative co - ordinates of particles we are not interested in , and operations by the observer on one particle will affect the relative co - ordinates of the others .
another interesting consideration is that of transforming between different external reference frames , rather than eliminating the external frame entirely .
there are two ways in which we can think of an external reference frame , either ( i ) as an abstract labelling of space - time , or ( ii ) as a physical laboratory containing rulers and clocks . in case ( i )
transforming between reference frames is a purely mathematical transformation , a simple relabelling of the co - ordinates . in case ( ii ) the issue of transformation is more subtle .
there are actually two subcases .
the first way ( iia ) is to consider one single physical frame and going to a different frame being a purely mathematical transformation on the co - ordinates indicated on this physical frame .
for example , the physical frame may be a ruler , and we can go to a displaced frame by simply relabelling the marks of the ruler . the second way ( iib ) is to have two actual physical systems .
for example , we could take two rulers , one displaced relative to the other .
classically these different approaches are completely equivalent to one another .
quantum mechanically however , the situation is more complicated .
let us illustrate this in the example of section ii , where we analysed a particle s interaction with an interferometer in two different situations .
in the second case ( depicted in fig .
( 1b ) ) , the interferometer was prepared in a superposition of locations in the external reference frame , whilst the particle had a well - defined location .
consider now moving to a new reference frame in which the interferometer is localised near the origin .
let us see what happens if we attempt to change co - ordinates via a simple relabellings as in ( i ) or ( iia ) .
the description of the state in the new reference frame can not be given by a pure state , because the only pure state in which the particle can be found with equal probability at @xmath1 and @xmath0 and the interferometer always found at the origin is of the form up to some well defined relative phase between @xmath130 and @xmath131 .
this state however leads to different physical predictions than those made in the original frame , namely that there is interference in the interferometer .
alternatively , we might seek to describe the state in the new reference frame using a mixed state , but then the transformation between reference frames must be irreversible , and we would not be able to transform back to the first reference frame and recover the original description .
all of these problems stem from the fact that it makes no sense to relabel in superposition .
superposition and entanglement are properties of the states of physical objects
it is the physical object whose states can be in superposition or entangled ; there is no notion of abstract mathematical relabellings being in superposition . given the above difficulties , we now turn to the case ( iib ) , in which different _ external _ frames are considered as distinct physical laboratories .
we take these laboratories to be non - interacting rigid quantum objects of very large mass , such that the positions and momenta of objects in the system relative to the laboratory obey essentially the usual canonical commutation relations ( consider ( [ q3p2 ] ) in the limit as @xmath132 ) .
we can now approach the paradox discussed at the beginning of this section .
consider the situation depicted in figure ( 1b ) but now modified to include two physical frames , the original frame @xmath133 and a second frame @xmath134 , in which the interferometer is located at the origin . in
order for the interferometer to be located at the origin in @xmath134 it must be the case that it is actually entangled with the position of the interferometer . as we will show below , given that we now consider explicitly two frames of reference , the predictions made from each of them are in fact consistent , and there is no longer a paradox . as seen from frame @xmath133 ,
the state of the particle , interferometer and @xmath134 is @xmath135 note that this state is now different from ( [ eqn : setup_b ] ) , which did not include the frame @xmath134 . rewriting this in terms of the relative co -
ordinates @xmath136 and @xmath137 of the particle and interferometer relative to the @xmath134 , we obtain @xmath138 tracing out the centre of mass , we see that the description of the state in @xmath134 is the mixed state @xmath139 , which will not yield interference . hence in this case
the description relative to the original external frame @xmath133 and the new frame @xmath134 both give the same physical predictions .
let us now modify the set - up in figure ( 1a ) in a similar manner , introducing two frames , the original frame @xmath133 and a new frame @xmath134 in which the particle is located at the origin .
this modified set - up is now completely equivalent to the modified set - up of figure ( 1b ) with the role of the two frames interchanged . as such there can not be interference in the modified set - up ( 1a ) .
this may seem surprising , since in the original set - up there was interference . indeed , the state now , instead of being ( [ eqn : setup_a ] ) is given by @xmath140 as seen from @xmath133 .
if the frame @xmath134 is much more massive than the interferometer , this state can be expressed in relative co - ordinates as @xmath141 by tracing out the centre of mass , we obtain a mixed state from the perspective of the interferometer , hence interference will be lost . ) , as the position of the lab acts as a ` which - way ' detector for the photon .
] introducing the second physical frame @xmath134 into the state has therefore _ changed _ the physical predictions ( destroying the interference ) .
this may seem surprising , since all the measurements we perform involve only the particle and the interferometer and we do not touch the frame @xmath134 during the experiment .
why does then the mere existence of the frame @xmath134 change the physics ?
this stems from the fact that we wanted to introduce the frame @xmath134 in such a way that it is entangled with our system . in the classical case
we can introduce physical frames at will without any need to interact with the system and hence without changing its behaviour .
however , in order to prepare an entangled state , there must be an interaction between the system and the frame .
this is when the system is affected .
finally , we can consider introducing a quantum frame that is not entangled with the system , but is in a superposition of locations .
after tracing out the centre of mass , the situation observed from the perspective of such a frame will be the same as if it were prepared in a mixture of locations .
this will not change the physical predictions relating to internal properties of the system ( such as interference in our interferometry experiments ) , but it will nevertheless constitute an irreversible transformation of the original state of the system , mapping pure states to mixed states and losing information .
we investigated the physics of quantum reference frames by studying simple situations , involving only a few particles , whereby we promoted one of these particles to the role of a quantum observer .
we then asked the question of how this quantum observer would describe the rest of the system .
we first considered the case in which one of the particles is in superposition .
we argued that the centre of mass , in particular whether it is entangled to the relative co - ordinates or not , is a central issue .
second , we considered the situation in which two particles of the system are entangled such that the centre of mass and relative position are unentangled .
we then introduced a third particle which led us to realise that the hilbert space of the quantum observer is inherently non - local .
finally we discussed the issue of an absolute quantum frame of reference and showed that in quantum theory , introducing external reference frames can affect the behaviour of the system being studied in particular , when the additional reference frame and the system are taken to be entangled . in the future
it would be interesting to explore further the consequences of the non - local structure of the hilbert space of quantum observers .
furthermore it would be important to go beyond the one - dimensional case as new physical quantities arise in this setting , for example in two - dimensions and above we must also consider angular momentum .
we acknowledge financial support from cnpq / brazil , the uk epsrc , the projects qip - irc and qap .
ajs acknowledges support from the royal society .
99 y. aharonov and l. susskind , phys . rev . *
155 * , 1428 ( 1967 ) .
y. aharonov and l. susskind , phys . rev . *
158 * 1237 ( 1967 ) y. aharonov and g. carmi , found .
phys . * 3 * , 493 ( 1973 ) .
y. aharonov and g. carmi , found .
physics * 4 * , 75 ( 1974 ) .
y. aharonov and t. kaufherr , phys .
d * 30 * , 368 ( 1984 ) .
d. poulin and j. yard , new j. phys .
* 9 * , 156 ( 2007 ) .
boileau , l. sheridan , m. laforest and s.d .
bartlett , j. math . phys . * 49 * , 032105 ( 2008 ) . f. girelli and d. poulin , phys .
d * 77 * , 104012 ( 2008 ) .
g. gour and r. spekkens , new j. phys . *
10 * , 033023 ( 2008 ) .
g. gour , i. marvian and r. spekkens , phys .
a * 80 * , 012307 ( 2009 ) .
j. oppenheim and b. reznik , arxiv:0902.2361 [ hep - th ] s.d .
barlett , t. rudolph , and r. spekkens , rev .
phys . * 79 * , 555 ( 2007 ) .
s. massar and s. popescu , phys .
lett . * 74 * , 1259 ( 1995 ) .
n. gisin and s. popescu , phys .
lett . * 83 * , 432 ( 1999 ) .
a. peres and p. f. scudo , phys .
lett . * 86 * , 4160 ( 2001 ) .
e. bagan , m. baig , a. brey , r. munoz - tapia and r. tarrach , phys .
a * 63 * , 052309 ( 2001 ) .
bartlett , t. rudolph , r. spekkens and p.s .
turner , new j. phys . * 11 * , 063013 ( 2009 ) .
ioannou , and m. mosca , arxiv:0903.5156 .
f. costa , n. harrigan , t. rudolph , and c. brukner , new j. phys . * 11 * , 123007 ( 2009 ) .
liang , n. harrigan , s.d .
bartlett , and t. rudolph , phys .
lett . * 104 * , 050401 ( 2010 ) .
y. aharonov and d. rohlich , _ quantum paradoxes quantum theory for the perplexed _ , wiley - vch verlag gmbh & co. kgaa , weinheim ( 2005 ) .
y. aharonov , h. pendleton , and a. petersen , int .
* 2 * , 213 ( 1969 ) .
a. kiteav , d. mayers and j. preskil , phys .
a * 69 * , 052326 ( 2004 ) s.d .
bartlett , t. rudolph and r. spekkens , int .
. inf . * 4 * , 17 ( 2006 ) .
in this appendix , we further discuss the subtle issues of entanglement due to finite wavepacket width that were introduced in section [ sec:2_particle ] .
we note in the main text that a product state of two wavepackets in the external frame of reference will generally become entangled when expressed in terms of relative co - ordinates . for clarity
, we have made an approximation in which this entanglement of microscopic fluctuations is ignored , such that a product state in the external reference frame remains a product state when transformed into relative co - ordinates .
we therefore write @xmath142 throughout .
note that although fluctuations in the centre of mass and relative coordinate only extend over a small physical distance , this sense of approximation does not correspond to equivalence between the states for all measurements ( as the trace distance would ) . in particular , fine - grained measurements on the scale of a wavepacket might be able to distinguish @xmath12 from any particular product state @xmath143 .
however , such measurements are not relevant to our arguments .
furthermore , for some special choices of initial wavepacket in the two particle case , our approximation holds exactly .
for example consider a product state of two gaussian wavepackets in the external frame : @xmath144 i.e. particle 1 is centred around the position @xmath14 and has uncertainty @xmath145 and particle 2 is centred around @xmath15 with uncertainty @xmath146 .
moving to relative co - ordinates , using equations ( [ eqn : x_cm])-([eqn : x_r ] ) from the main text , we find the state transforms into @xmath147 where @xmath148 hence in general , the relative and centre of mass coordinate will be entangled .
however , if we choose @xmath149 we obtain @xmath150 and recover a product state . by taking the initial wavepackets to be gaussians satisfying @xmath149 and the transformed wavepackets to be gaussians of widths @xmath151 and @xmath152 , we could therefore eliminate our approximation and have @xmath153 . however , we emphasise that our central observations extend beyond this special case .
all we really require are that wavepackets are sufficiently small , and that all wavepackets for the same particle in an initial state have the same form ( i.e. they are translated copies of each other ) . with three particles ,
the centre of mass will again decouple from the relative co - ordinates when the initial wavepackets are gaussians satisfying @xmath154 .
however , in this case there will remain some residual entanglement between the relative co - ordinates @xmath60 and @xmath68 in the transformed state .
if we wished to include this entanglement explicitly , we could rewrite ( [ psi ] ) as @xmath155 where @xmath156 represents an entangled wavepacket in the relative co - ordinates , centred on @xmath157 and @xmath158 .
the reduced state of @xmath159 given by ( [ red_state ] ) would then be modified to @xmath160 where @xmath161 are mixed wavepackets centred on @xmath162 , and we have used the fact that @xmath163 . note that the phase is still unobservable in this state , and hence the conclusions are identical to those in the approximated case . | we investigate the physics of quantum reference frames . specifically , we study several simple scenarios involving a small number of quantum particles , whereby we promote one of these particles to the role of a quantum observer and ask what is the description of the rest of the system , as seen by this observer ?
we highlight the interesting aspects of such questions by presenting a number of apparent paradoxes . by unravelling these paradoxes
we get a better understanding of the physics of quantum reference frames . |
galaxy - galaxy lensing uses the distortions of background galaxies to study the mass distribution around foreground galaxies . in a typical lens situation ,
the shear from a foreground lens is only weak .
therefore , galaxy - galaxy lensing can only study dark matter halos of galaxies statistically by averaging over thousands of lens galaxies . for reviews on galaxy - galaxy lensing see @xcite and @xcite .
we use the maximum - likelihood technique proposed by @xcite .
first , we have to identify lenses and source galaxies which we do based on accurate photometric redshifts .
next , we adopt a specific lens model to calculate for each background galaxy the shear contributions from each foreground galaxy within a certain annulus . the estimated shear is compared to the observed shapes of the sources for a range of input parameters of the lens model and those parameters which maximize the likelihood are determined . here
, we use the singular isothermal sphere ( sis ) and the navarro - frenk - white ( nfw ) profile to model the lenses .
we adopt @xmath3 and @xmath4 .
for our investigation we use the combo-17 survey ( @xcite ) which is a deep survey with very good imaging quality and accurate photometric redshifts .
all data are taken with the wide field imager at the mpg / eso 2.2-m telescope on la silla , chile .
the survey consists of 4 fields of which 3 are used here .
the limiting magnitude is @xmath5 .
deep @xmath6-band observations were taken in the best seeing conditions ( below 0.8 " psf ) .
observations in @xmath7 and 12 medium - band filters are used to derive restframe colours and accurate photometric redshifts with @xmath8 at @xmath9 and @xmath10 at @xmath11 .
these allow us to select both lenses and sources based on their redshifts and to select and study subsamples of lens galaxies based on their restframe colours .
in the following , sources are all galaxies with @xmath12 and @xmath13 .
lenses are galaxies with @xmath12 , @xmath14 . the shear of a specific lens galaxy on a specific source galaxy is only considered if @xmath15 for that lens - source pair .
further , the projected separation between lens and source must be smaller than @xmath16 at the redshift of the lens when the lens is modelled as sis . at this @xmath17
we obtain the tightest constraints . when modelling lenses by nfw profiles we extend the maximum separation to @xmath18 to ensure that the region around the virial radius is probed .
due to the size of galaxy images , a minimum angular separation of 8 " between lenses and sources is required in order to avoid that shape measurements of sources are biased by the light of the lenses .
we investigate the lens sample as a whole and additionally split it into red and blue subsamples based on restframe colours .
galaxies with @xmath19 define the blue sample while all other galaxies are in the red sample ( @xcite ) .
the density profile of the sis is given by @xmath20 where @xmath21 is the velocity dispersion and @xmath22 is the distance from the center of the lens .
we assume that the velocity dispersion depends on the luminosity of a galaxy , @xmath23 , where @xmath24 is a characteristic luminosity measured in the restframe sdss @xmath22-band .
this is the tully - fisher or faber - jackson relation which was derived for galaxies on smaller scales than used here from their rotation curves or stellar velocity dispersion .
the left panel of figure [ sis ] shows 1- , 2- and 3-@xmath25 contours for the model parameters @xmath26 and @xmath27 derived for the whole lens sample .
the best - fit parameters with 1-@xmath25 errors are @xmath28 and @xmath29 .
these values agree very well with expectations from e.g. rotation curve measurements .
we want to compare this result to the galaxy - galaxy lensing measurement from the red - sequence cluster survey ( rcs , @xcite ) which probes lens galaxies in a comparable redshift range and uses a similar modelling .
@xcite find @xmath30 at fixed @xmath31 .
however , @xcite use a characteristic luminosity of @xmath32 measured in the @xmath33-band instead of the @xmath22-band as we do here . from the restframe luminosities of galaxies in combo-17 we estimate that galaxies with @xmath32 have @xmath34 .
further , @xcite use pairs with projected separations up to 2 corresponding to about @xmath35 . although for the sis model the velocity dispersion should be independent of radius , we find a decline of the fitted @xmath26 with increasing @xmath17 .
using @xmath36 and @xmath35 we measure @xmath37 and @xmath38 in very good agreement with the rcs result .
the error on @xmath26 is about 5 times smaller for the rcs than for combo-17 . given the about 60 times larger area of the rcs this is not surprising .
the uncertainties should also be influenced by the qualitiy of the redshift information .
the measurement from the rcs uses observations in a single passband only and does therefore not have redshift estimates for individual objects . in @xcite
we find that the velocity dispersion can be well constrained even in the absence of redshift information .
redshifts for individual lens galaxies reduce the errors on @xmath26 by only 15% .
however , they are essential for measuring the dependence of the velocity dispersion ( or mass ) on luminosity the errors on @xmath27 increase by a factor of 2.5 when omitting the lens redshifts .
individual redshifts for the sources are not important as long as the redshift distribution is known .
the right panel of figure [ sis ] shows likelihood contours for the blue and red subsamples . while no significant change in @xmath27 is seen , a 2-@xmath25 difference in @xmath26 is seen between the two lens populations .
the best - fit velocity dispersions are @xmath39 for the blue sample and @xmath40 for the red sample , respectively .
the red sample consists of 2579 galaxies , the blue sample of 9898 galaxies .
although only about 20% of the lenses are red , this subsample gives even tighter constraints than the blue subsample .
this shows clearly that most of the galaxy - galaxy lensing signal comes from red galaxies .
next , we model lens galaxies by nfw profiles .
the density profile is given by @xmath41 .
@xmath42 is a characteristic scale radius at which the density profile changes from @xmath43 to @xmath44 .
@xmath45 is related to the concentration @xmath46 .
the virial radius @xmath47 is defined by @xmath48 . here , the virial radius is the radius inside which the mean density is 200 times the mean density of the universe .
the relation between @xmath45 and @xmath46 is fixed by this definition .
unfortunately , the definition of the virial radius is not unique .
often , the critical density of the universe instead of its mean density is referred to or overdensities different from 200 are used .
these differences have to be kept in mind when comparing results from the nfw profile .
first , we try to measure the virial radius @xmath50 and the concentration @xmath46 at fixed @xmath31 , see fig .
[ nfw_c ] .
the virial radius can be constrained well while on the concentration we can only derive lower limits .
this implies an upper limit on the scale radius , @xmath51 .
this is at all considered lens redshifts smaller than the imposed minimum angular separation between lenses and sources of 8 " .
therefore , we can not expect to be sensitive to @xmath42 or @xmath46 . in the following , we fix @xmath52 which is at the lower end of the values allowed by our measurement .
note that when defining the virial radius as radius inside which the mean density is 200 times the critical density of the universe ( instead of its mean density as done here ) this would refer to @xmath53 .
correspondingly , the virial radii and virial masses which we are going to derive would be smaller in that case by about 40% and 20% , respectively .
figure [ nfw ] shows 1- , 2- and 3-@xmath25 contours for the virial radius @xmath50 and @xmath27 for the whole lens sample and for the blue and red subsamples .
averaged over all lenses , the best - fit parameters with 1-@xmath25 errors are @xmath54 and @xmath55 . for the blue sample
we find @xmath56 and @xmath57 , for the red sample @xmath58 and @xmath59 . between the blue and red subsamples we measure a 1-@xmath25 difference in the virial radius as well as in @xmath27 .
.constraints on dark matter halos of galaxies modelled by nfw profiles .
the virial radius @xmath50 and @xmath27 are fitted quantities ( see fig .
[ nfw ] ) , the virial mass @xmath60 , the virial mass - to - light ratio @xmath61 and the rotation velocity at the virial radius , @xmath62 , are calculated from @xmath50 .
@xmath63 gives the scaling of @xmath61 with luminosity , @xmath64 .
the maximum rotation velocity @xmath65 and the radius of the maximum rotation velocity @xmath66 are calculated for a concentration @xmath52 . [
cols="<,^,^,^,^,>,>,^,^ " , ] table [ table ] gives an overview of the measured parameters ( @xmath50 , @xmath27 ) and calculated quantities like the virial mass @xmath60 , virial mass - to - light ratio @xmath61 and the scaling between @xmath61 and luminosity .
again , we compare our results to those from other data sets .
@xcite find from the rcs @xmath67 at @xmath32 . at the corresponding @xmath68 measured in the restframe
@xmath22-band we find @xmath69 .
@xcite obtain @xmath70 at @xmath71 from the sdss where @xmath72 is measured in the sdss restframe @xmath22-band .
however , @xcite define the virial radius as radius inside which the mean density is 200 times the critical density of the universe .
our corresponding result using their value of @xmath72 and their definition of the virial radius is @xmath73 .
figure [ fields ] shows likelihood contours from fitting the sis model as in sect .
[ sect : sis ] .
clearly , the derived constraints on the velocity dispersion @xmath26 are not consistent for the individual fields .
while the a901 field gives very tight constraints , from the cdfs we can only derive an upper limit on @xmath26 .
the deviation from the measurement that uses all three fields together ( @xmath74 ) is 1-@xmath25 towards higher values for the a901 field and 2-@xmath25 towards lower values for the cdfs field .
only the s11 field is consistent with the overall measurement .
these three survey fields were already selected to be very different .
the s11 field is the only random field .
it contains the cluster abell 1364 at @xmath75 by chance .
the a901 field was chosen because it contains a supercluster with the three components abell 901a , 901b and 902 at @xmath76 .
the cdfs field contains the chandra deep field south and was chosen because of its emptiness . due to these selection criteria one might suspect that the measured galaxy - galaxy lensing signal is mostly due the foreground clusters
however , our lens sample only uses galaxies at @xmath77 and should thus not include cluster galaxies . therefore , the foreground clusters in the a901 and s11 field should not have any influence on our measurement . by including the additional shear from the foreground clusters we indeed confirm this assumption
further , we find that the shear from an additional cluster in the a901 field at @xmath78 does not induce a significant shift in the velocity dispersion . including the shear from this higher - redshift cluster does however increase the uncertainties by about 20% .
another suspected reason for the differences between the three fields is the imaging quality .
the sum image of the a901 field has the best quality with a psf of 0.74 `` .
the psf of the other two sum images is 0.88 '' .
if image quality had a dominant effect , then the results from the s11 field and the cdfs field should be consistent .
for the cdfs field we have several independent sum images available from different observing runs with very different seeing conditions and exposure times which yield consistent lensing signals . therefore we rule out image quality as possible explanation for the discrepant measurements .
the most probable explanation we find for the deviating results comes from the number counts in the different fields .
the number of lenses in the fields is 4636 ( a901 ) , 4268 ( s11 ) and 3573 ( cdfs ) .
therefore , one expects the tightest constraints from the a901 field .
the difference in the derived velocity dispersion could be due to differences in the composition of the lens samples .
indeed , the fraction of red lenses is 23.5% in the a901 field , 20.5% in the s11 field and only 17.2% in the cdfs field . given our findings in sect .
[ sect : sis ] we expect a higher velocity dispersion and tighter constraints with increasing fraction of red galaxies . | we study the dark matter halos of galaxies with galaxy - galaxy lensing using the combo-17 survey .
this survey offers an unprecedented data set for studying lens galaxies at @xmath0 including redshift information and spectral classification from 17 optical filters for objects brighter than @xmath1 .
so far , redshifts and classification for the lens galaxies have mainly been available for local surveys like the sloan digital sky survey ( sdss ) .
further , redshifts for the source galaxies have typically not been available at all but had to be estimated from redshift probability distribution which for faint surveys even had to be extrapolated . to study the dark matter halos we parametrize the lens galaxies as singular isothermal spheres ( sis ) or by navarro - frenk - white ( nfw ) profiles . in both cases
we find a dependence of the velocity dispersion or virial radius , respectively , on lens luminosity and colour .
for the sis model , we are able to reproduce the tully - fisher / faber - jackson relation on a scale of @xmath2 . for the nfw profile
we also calculate virial masses , mass - to - light ratios and rotation velocities .
finally , we investigate differences between the three survey fields used here . |
supermassive black holes ( smbhs ; @xmath1 ) are inferred to reside in most galactic nuclei @xcite . when efficiently supplied with gas
, these objects can produce some of the most luminous sources in the universe @xcite .
recent panchromatic surveys have shown that close dual agn are comparatively more luminous at high energies than their isolated counterparts , suggesting that the merger process is intricately tied to the feeding history of the waltzing smbhs @xcite .
gas supply to central smbhs during galaxy mergers directly affects the growth and luminosity of these objects .
therefore , understanding how mass is bestowed to smbhs results in predictions of their number density and luminosity distribution @xcite . in cosmological simulations of merging galaxies , some form of the classical bondi - hoyle - lyttleton ( bhl ) accretion prescription is usually implemented to estimate smbh feeding rates @xcite .
the use of this recipe assumes that the properties of gas at 50 - 100 pc ( generally determined by the resolution scale length ) accurately models the mass accretion rate onto the smbh @xcite . a significant fraction of smbhs are expected to be embedded in nuclear star clusters ( nscs ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) . because these nscs are typically more massive than the central smbh , they can significantly alter the gas flow at scales which are commonly unresolved in cosmological simulations ( @xmath2 ; * ? ? ?
as a result , nscs could provide an efficient mechanism for funneling gas towards the central black hole . accurately determining the mass accretion history of smbhs
has important consequences not only for their growth history , but also for the evolution of the host galaxy @xcite .
for this reason , pinning down the dominant mechanism by which gas at large scales is funneled into the smbh is essential for understanding their role in mediating galaxy evolution . in this paper
, we make use of simulations to investigate how the accretion rate of smbhs might be enhanced during galaxy mergers when they are embedded in massive and compact nscs .
hydrodynamical models with multiple levels of refinement are used to capture both the large scale gas flow around the smbh and nsc complex as well as the small scale accretion onto the central sink .
we model the smbh and the nsc as static potentials and simulate the structure of supersonic gas flows within their combined gravitational field in order to quantify the gas accretion rate in the inner tenths of parsecs .
simulations are performed for adiabatic and isothermal flows as well as for wide range of nsc radii and ambient conditions .
these well - resolved , three dimensional simulations are used to generate sub - grid accretion prescriptions for larger scale simulations , thus providing a more accurate estimate of the feeding rates of smbhs in cosmological simulations as well as a better determination of the expected luminosities of dual agns .
we test the effects of this accretion rate enhancement on the growth of smbhs in the galaxy merger simulations of @xcite and discuss what combination of gas and gravitational potential parameters results in a significant augmentation to the mass accretion rate .
this paper is organized as follows . in section [ section : anal ] we review the commonly used sub - grid prescriptions for calculating the mass accretion rates onto central massive black holes during galaxy merger simulations and suggest a modification in the presence of a nsc . a numerical scheme aimed at calculating black hole growth at high resolution , accounting for the gravitational focusing effects of nsc ,
is presented in section [ section : methods ] .
the conditions necessary for nscs to collect ambient gas and , in turn , enhance the mass accretion rate of smbhs are derived in section [ sec : cond ] . in section [ section : mergermods ] we use the central gas densities from realistic sph cosmological merger simulations to estimate the augmentation to smbhs mass accretion rates in the presence of a nsc as a function of time .
we summarize our findings in section [ section : summary ] .
smbhs can trigger nuclear activity only as long as they interact with the surrounding gas .
the way in which gas flows into a smbh depends largely on the conditions where material is injected . the mass accretion rate into smbhs in cosmological simulations is commonly estimated using analytical prescriptions based on the large scale gas structures in the centers of the simulated merging galaxies .
the bhl formalism assumes the gas is accreted spherically symmetrically @xcite : @xmath3 where @xmath4 is the relative velocity of the object through the external gas whose sound speed and density are given by @xmath5 and @xmath6 , respectively @xcite .
the flow pattern is dramatically altered if the inflowing gas has a small amount of angular momentum . if the inflowing gas is injected more or less isotropically from large @xmath7 , but has specific angular momentum per unit mass such that @xmath8 , then the quasi - spherical approximation will break down and the gas will have sufficient angular momentum to orbit the smbh ( here @xmath9 is the smbh s schwarzschild radius ) .
viscous torques will then cause the gas to sink into the equatorial plane of the smbh . in recent works ,
such departures from spherical symmetry have been accounted for by assuming the accretion proceeds through a disk : @xmath10 where @xmath11 is the average surface gas density of the accretion disk , @xmath12 is the average rotational angular frequency , and @xmath13 is a dimensionless parameter dictating the strength of turbulent viscosity in the disk @xcite .
both prescriptions rely on an understanding of the gas properties at @xmath14100 pc in order to determine the mass accretion rate onto the central smbh .
as discussed in @xcite , a nsc moving at a low mach number through relatively cold gas can drastically increase the gas density in its interior with respect to that of the external medium .
analytical estimates suggest that in the central regions of a nsc the expected density enhancement is given by @xmath15^{1/(\gamma-1 ) } , \label{eq : rhostationary}\ ] ] where the core radius , @xmath16 , is related to the cluster velocity dispersion by @xmath17 .
the gas properties characterized here by @xmath6 and @xmath18 are the density and sound speed at @xmath19 . here
@xmath20 denotes the adiabatic index of the flow and @xmath21 designates the mass of the nsc @xcite .
given this density enhancement in the core of the nsc , the accretion rate of the embedded smbh would be amplified by a factor of @xmath22 where @xmath23 is the sound speed of the gas within the nsc and we have assumed here that accretion proceeds at the classical bondi rate @xcite .
even a small amount of angular momentum can make a big difference , breaking the spherical symmetry of the inflowing gas and yielding an accretion disk instead of the radial flow assumed in equation [ eq : rhostationary ] .
if we instead suppose that accretion onto the central smbh proceeds by viscous torques and require the average surface density of the disk to be proportional to @xmath24 , the mass accretion rate can be approximated as @xcite @xmath25 with @xmath26 , and @xmath27}{c_\infty^2 r_{\rm c } [ 1 + \mu_\infty^2 ] } \right)^{\frac{1}{\gamma-1 } } \left ( \frac{\mu_\infty^2}{\mu_\infty^2 + 1}\right)^{\frac{3}{2}},1\right\}. \label{eq : rho}\ ] ] here @xmath24 is the modified density in the nsc s interior and @xmath28 is the mach number of the smbh+nsc complex with respect to the ambient gas .
the resultant accretion rate would , in this limit , be amplified by a factor @xmath29 .
contrary to the classical bhl case , the motion of the smbh+nsc complex with respect to the ambient medium does not result in a considerable change in the rate of mass accretion onto the smbh .
this is because equation [ eq : rho ] accounts for the protection provided by the large scale nsc which forms a quasi - hydrostatic envelope around the smbh .
having said this , the strength of this protection is indeed slightly diminished because the capture radius decreases due to the relative motion of the potential @xcite . as a result ,
the expected enhancement in the gas density of the nsc core is indeed moderately smaller than the one given by the stationary formula ( equation [ eq : rhostationary ] ) .
more importantly , we have assumed here that the influence of the nsc dominates the gravitational potential , which implies @xmath30 . when @xmath31 , the formation of a quasi - hydrostatic envelope is inhibited .
the presence of a compact and massive nsc with @xmath30 is thus required in order to significantly increase the accretion rate onto the central smbhs during a galaxy merger .
the degree by which the accretion rate is enhanced by the presence of a nsc is studied with numerical simulations in the remaining of the paper .
to examine the ability of a nsc to collect ambient gas and , in turn , enhance the mass accretion rate onto the central smbh , we simulate the smbh + nsc complex as a gravitational potential @xmath32 moving through ambient gas with flash , a parallel , adaptive mesh refinement hydrodynamics code @xcite .
a smooth potential , given by @xmath33 provides an accurate description of the nsc potential given the cluster mass , @xmath21 , and radius , @xmath34 @xcite .
the gravitational potential of the smbh is given by @xmath35 and is modeled by a sink term .
here , we assume the gravitational potential is static - @xmath36 does not grow - a valid approximation since the mass accreted during the simulation is small .
we use inflow boundaries to simulate the nsc s motion through the central galaxy medium . in order to accurately resolve the mass accretion rate on the black hole smbh sink sizes
are taken to be within hundredths of sonic radii , @xmath37 , where @xmath38 following the prescriptions of @xcite .
models are run from an initially uniform background until a steady density enhancement forms within the nsc , which usually takes tens of core sound crossing times .
convergence tests with higher refinement and longer run times produce models which show similar central densities and mass accretion rates to those depicted here .
we compute both models with adiabatic ( @xmath39 ) and nearly isothermal ( @xmath40 ) equation of states in order to test the effects of cooling . to adequately resolve the small scale sink , the significantly more extended nsc s core , and the even larger scale flow structures that develop around the cluster , a sizable level of refinement
is required on the amr grid ( commonly 14 levels of refinement ) . as the minimal time step in our simulation
is determined by the gas flow on small scales , necessary runtimes would need to be prohibitively large in order to simulate a resolved sink and the large scale gas structure until a steady state density enhancement forms within the nsc .
instead , we construct the accretion rates onto our model smbh+nsc systems from a set of three simulation setups , as depicted in figure [ fig : simsexplained ] with parameters summarized in the first row of table [ table : sims ] .
figure [ fig : simsexplained ] depicts the flow of nearly isothermal gas ( @xmath40 ) in the vicinity of a smbh ( @xmath41 , @xmath42 pc ) surrounded by a massive nsc ( @xmath43 , @xmath44 pc , @xmath45 ) .
the gas has a sound speed of @xmath46 and the smbh + nsc complex is moving at a mach number of @xmath47 with respect to the ambient gas .
the ratio of @xmath43 is consistent with observations : @xmath48 @xcite .
however , as noted in section [ section : anal ] , as the ratio of @xmath49 decreases so does the ability of a nsc to significantly alter the flow of gas onto the embedded smbh ( for a given @xmath50 ) .
thus , the model depicted in figure [ fig : simsexplained ] provides an example in which the presence of a massive nsc can drastically alter the flow properties around the central smbh .
the first of these simulations , labeled _ initial _ in the density contours and mass accretion rate plots shown in figure [ fig : simsexplained ] , is a small scale simulation which follows the gas flow as the central density enhancement begins to form in the nsc core while simultaneously tracking the mass accretion rate onto the fully resolved central sink .
once the large scale bowshock begins to interact with the boundaries of the computational domain ( at about three to five core sound crossing times ) the simulation is halted .
we concurrently simulate the same initial setup in a larger box ( about fifty core radii ) at a lower resolution and follow the density build up within the nsc s core without the presence of a sink .
we label this simulation as _ no sink _ in figure [ fig : simsexplained ] .
because the presence of a sink has a minimal effect on the build up of mass in the core , the central density in the large scale simulation , albeit lower resolution , agrees well with central density evolution observed in the _ initial _ simulation .
while there is no explicit sink in these second set of simulations , the mass accretion rate onto a central black hole can be relatively accurately inferred from the gas properties within the nsc s core as shown in figure [ fig : simsexplained ] . once a steady state density enhancement has formed in the central regions of the nsc , these large scale simulations are then refined further until a central sink is resolved by at least 16 cells .
this level of refinement was chosen to allow the unsteady mass accretion rate to converge , as argued by @xcite .
we refer to these simulations as _ steady state _ in figure [ fig : simsexplained ] . by using the set of three simulations discussed here for each smbh+nsc system
, we can resolve both the larger scale flow around the nuclear star cluster and the small scale flows into the accreting smbh . because we are not explicitly including radiative cooling , the simulations depicted in figure [ fig : simsexplained ] can be easily rescaled to consider a wide range of nsc properties .
for example , for a fixed @xmath49 , the structure of the flow will remain unchanged provided the ratio @xmath51 remains constant @xcite . in figure
[ fig : scaled ] we compare the flow in and around a _
smbh + nsc complex ( @xmath41 , @xmath52 , @xmath53 ) moving through hot gas ( @xmath54 , @xmath55 , @xmath39 ) with the flow surrounding a _ lighter _ system ( @xmath56 , @xmath57 , @xmath58 ) moving through cold gas ( @xmath59 , @xmath55 , @xmath39 ) .
the parameters of the less massive model are chosen such that the ratio @xmath60 remains the same .
because this ratio is constant , the flow is nearly identical .
it is important to note that while the mass accretion rate onto the central smbh is higher for the more massive system , the accretion rate enhancement with respect to that of a black hole without the nsc , @xmath61 , is the same between the two simulations .
the set of simulations shown in figure [ fig : simsexplained ] ( figure [ fig : scaled ] ) depict the gas flow around a compact nsc slowly transversing through nearly isothermal ( adiabatic ) gas . however , the enhancement in the black hole s accretion rate depends not only on the thermodynamical conditions of the ambient gas but also on the properties of the nsc .
the remainder of this paper is thus devoted to calculating the necessary conditions for large central density enhancements in the centers of smbh + nsc systems and determining whether or not these conditions persist during a galaxy merger .
in order for the gravitational potential of the nsc to alter the local gas flow before it is accreted onto the central smbh , the nsc must be moving relatively slowly through cold gas , a condition shown in @xcite to be equivalent to requiring
that @xmath62 for this reason , knowledge of the velocity dispersion of typical nscs is vital in determining whether or not central gas densities will form in their cores ( provided that @xmath63 ) . surveys of nscs show half light radii in the range @xmath64 pc @xcite , which assuming a plummer model ( with @xmath65 ) , gives a range in core radii of @xmath66 pc .
these results combined with the observed mass range of approximately @xmath67 @xcite naturally result in a large range of possible velocity dispersions : @xmath68 .
intuitively , we expect more massive systems to have larger core radii , a fact which is born out in several surveys of nsc properties as depicted in figure [ fig : msigma ] @xcite .
if we narrow our sample size by restricting our calculations to systems observed to contain both a nsc and a smbh the mass - size relation of nscs is less well constrained as clearly depicted in figure [ fig : msigma ] .
measuring the mass of theblack hole and the mass and velocity dispersion of the nsc requires high resolution observations , which are only available for a handful of systems , with the best currently known example being our own milky way . in the galaxy s nuclei ,
the black hole mass has been measured to be @xmath69 @xcite . with a nsc mass of @xmath70 and a half light radius of @xmath71 pc
, one finds @xmath72 pc and @xmath73 @xcite .
the observed nsc velocity dispersions in galaxies containing black holes , which are in the range @xmath74 , are similar in magnitude to both the sound speed of the surrounding gas and the relative gas velocities of smbhs during galaxy mergers @xcite . as a result ,
the conditions necessary for efficient mass accumulation are commonly satisfied , an assertion that we will quantify below by making use of detailed galaxy merger simulations .
figure [ fig : modbondiadia ] shows how the accretion rate onto a smbh is modified by the presence of a nsc satisfying @xmath75 . in all calculations ,
the flow is assumed to behave adiabatically . without the presence of a nsc , the bowshock penetrates close to the sink boundary
. however , in the presence of a nsc , the bowshock forms at the outer boundary of the cluster s core . as discussed in @xcite and @xcite ,
this effectively mitigates the effects of the gas motion on the central sink and the accretion proceeds as a nearly radial inward flow .
this added protection results in a moderate mass accretion rate enhancement onto the central sink , which are slightly larger for a more compact nsc .
as the flow accumulates in the nsc s potential , a quasi - hydrostatic envelope builds up around the central sink , whose central density increases with the compactness of the nsc . for an adiabatic flow , this density enhancement
is accompanied by an increase in the sound speed of the flow such that @xmath76 . as a result
, the accretion rate for a stationary sink is not expected to be aided by the increase in central density in an adiabatic flow .
the moderate increase in mass accretion rates for an adiabatic flow , produced mainly by changes in the flow structure , is shown by both our analytical ( see equations [ eq : mdot ] and [ eq : rho ] ) and simulation results . on the other hand , the central density enhancement enabled by the presence of a nsc when
@xmath77 can be accompanied by a drastic increase in mass accretion rate if the gas is permitted to cool .
this allows the central density to grow without a mitigating increase in the local sound speed of the gas .
in the near isothermal ( @xmath78 ) calculations depicted in figure [ fig : modbondiiso ] , we have @xmath79 and the presence of a nsc results in an enhancement of about an order of magnitude in the mass accretion rate onto the central sink . here
, the accretion rate fluctuates around a mean value , as the isothermal gas can collapse to much smaller scale structures than in the adiabatic case , providing the central sink with much larger temporal changes in the amount of accreted gas . figures [ fig : modbondiadia ] and [ fig : modbondiiso ] together demonstrate the important effects that both the equation of state and the compactness of the nsc can have on the mass accreted by the central smbh . in what follows
, we make use of cosmological simulations to estimate the range of gas and nsc properties conducive to large enhancements in the mass accretion rate onto the central smbhs during galaxy mergers .
the majority of galaxies ( @xmath80% ) are expected to harbor nuclear star clusters @xcite , and therefore large enhancements in accretion rates onto smbhs are possible during typical galaxy mergers when conditions are favorable ( i.e. @xmath81 ) .
if , in addition , during the merger the gas in the central regions cools efficiently , the increase in the mass accreted by the smbh can be significant . to determine if and when these conditions are satisfied during a merger we examine the gas properties in galaxy merger simulations from @xcite .
these full - scale sph simulations of major mergers include cooling , star formation and associated feedback , and feedback from a supermassive black hole . of interest for this work
is the model _ fidnof _ , which places two galaxies on a prograde parabolic orbit with unaligned spins and a merger mass ratio of 1:1 .
both galaxies in the simulation have a total mass of @xmath82 ( including the dark matter halo ) , and have @xmath83 particles .
the plummer equivalent gravitational force softening was @xmath84 pc .
figure [ fig : snapshots ] shows the gas properties in the central core regions of the galaxies in the _ fidnof _ model .
this range of sound speeds and densities represent the average values of a subset of particles within the accretion radii , defined as four times the simulation s gravitational softening length : @xmath85 pc @xcite . to estimate the average properties of the gas ( @xmath18 , @xmath86 , @xmath6 ) flowing toward the smbh we use only particles within a @xmath87 conical region in front of the smbh s velocity vector . while the average gas parameters are relatively insensitive to the exact value of the opening angle of the cone , the amount of inflowing gas is slightly underestimated using this method as it ignores the material which can be accreted from behind the direction of motion of the smbh .
given the estimated average properties of the gas flow at large scales , we expect the effects of the nsc in altering the mass accretion history of the central smbh to be most prominent when the gas is able to cool efficiently , as argued in section [ sec : cond ] . in a merger simulation , the condition for efficient cooling is established when the sound crossing time across the accretion radius is longer than the cooling time of the gas : @xmath88 .
if this condition holds , the cold gas can be significantly compressed and , as a result , lead to a large density enhancement in the core of the nsc .
we can estimate the accretion timescale of the smbh as @xmath89 where @xmath28 is the mach number of the large scale flow .
the cooling time can be written as @xmath90 $ ] , where @xmath91 is the internal energy of the gas , @xmath92 is the cooling rate of the gas at a temperature @xmath93 and metallicity @xmath94 , and @xmath95 and @xmath96 are the electron and neutral hydrogen number densities , respectively . in galaxy merger simulations ,
the condition @xmath88 is generally satisfied ( figure [ fig : snapshots ] ) although for a particular run , the average gas properties can fluctuate between the cooling and non - cooling regimes as the merger progresses .
this is illustrated in the simulation snapshots _ _ , _ _ and _ _ taken from the _ fidnof _ model of @xcite at early ( @xmath97 gyrs , @xmath98 gyrs ) and late ( @xmath99 gyrs ) times in the merger process , which are depicted in figure [ fig : snapshots ] . as a consequence , there may be times during the galaxy merger when cooling rates within the accretion radius are high and the mass accretion rate can be heavily augmented by the presence of a nsc , provided that @xmath100 .
since we are not treating the feedback from the black hole explicitly , we use both adiabatic ( inefficient cooling ; efficient feedback ) and isothermal ( efficient cooling ; inefficient feedback ) simulations to illustrate the effects of the surrounding nsc on the gas flow as a whole and the importance of the @xmath100 condition .
figure [ fig : nocool ] shows the gas in the inner regions of a model where @xmath101 . here ,
a moderately massive black hole ( @xmath102 ) with and without a surrounding nsc ( @xmath103 and @xmath104 ) propagates through a background medium with @xmath105 200km / s and @xmath106 ( similar to the gas properties found in simulation snapshot _ _ of figure [ fig : snapshots ] ) .
because @xmath101 , the gas flow around the smbh is not altered by the presence of the nsc and the mass accretion rates change only minimally between the model with and without the nsc , even when cooling is efficient .
this is corroborated by the results of the near isothermal and adiabatic simulations , which are shown in figure [ fig : nocool ] .
when the smbh+nsc complex propagates into a region in parameter space where cooling is efficient and the condition @xmath107 is satisfied , the presence of a nsc can dramatically increase the mass supply onto the smbh .
figure [ fig : cool ] shows the gas flow in the inner regions of a nsc where cooling is predicted to be efficient .
similar to figure [ fig : nocool ] , the smbh + nsc complex is characterized by @xmath102 and @xmath103 ( @xmath108 ) but in this case it propagates through a background medium with @xmath105 100 km / s and @xmath109 ( similar to those found in simulation snapshot _ _ of figure [ fig : snapshots ] ) . here , even without the presence of a nsc ,
the accretion rate is significantly higher than that derived from figure [ fig : nocool ] due to the isothermal equation of state , higher density and lower sound speed of the medium .
however , in these efficient cooling conditions , the presence of a nsc can have drastic effects on the mass feeding rate enhancement when compare to the case without a nsc .
this is evident when comparing the evolution of the mass accretion rate calculated in the adiabatic and near isothermal simulations .
we note here that in both regimes , the prescription laid out by equations ( [ eq : mdot ] ) and ( [ eq : rho ] ) provides a relatively good estimate of the steady state mass accretion rate onto the smbh ( see figures [ fig : nocool ] and [ fig : cool ] ) . in what follows
, we will assume the validity of such prescription in order to estimate the growth history of smbhs in galaxy mergers . to establish the mass feeding history of merging smbhs embedded in nscs ,
we use the gas properties in the central regions of simulated merging galaxies . figure [ fig : subgrid ] shows how the growth history of the central smbhs , as derived from the _ fidnof _ simulation of @xcite , is altered by the presence of a nsc . in both galaxies ,
a nsc with @xmath103 is assumed to reside in each galactic center at the start of the simulation .
these values are consistent with observations of nsc around smbhs @xcite .
the calculation assumes that the flow is able to cool efficiently ( @xmath40 ) and that the nsc is unable to grow as the mass of the smbh increases such that the mass accretion prescription reverts to the unmodified version ( equation [ eq : cpmmacc ] ) when @xmath110 . at early times
, the presence of the nsc enables the central smbh to grow quicker than it would if it was in isolation due to the increase in gas density accreted by the central smbhs as described by equation [ eq : mdot ] . as the galaxy merger progresses and the mass of the smbh quickly increases above @xmath21 , the gravitational influence of the cluster ceases to be relevant and the mass accretion prescription reverts to that used in the simulation ( equation [ eq : cpmmacc ] ) .
because in our formalism @xmath111 , the growth rate of the pre - merger smbh hosting a nsc is reduced when compared to the one in isolation , whose is initially lighter . the swifter rate of growth for the unmodified smbh implies that it is able to reach a similar total mass to that surrounded by a nsc before the merger takes place . despite having only a brief impact , this early growth spurt induced by the presence of a nsc results in a different early growth and feeding history for the pre - merger smbhs .
we find that the initial properties of the nsc have an enduring effect on the luminosity and mass assembly history of the early time pre - merger smbhs .
we note here that we have assumed that the nsc s mass remains unchanged during the entire simulation .
if a larger stellar concentration is able to form around the growing smbh , then the evolving nsc could have a longer lasting impact .
an understanding of how matter can be funneled to galactic nuclei is essential when constructing a cosmological framework for galaxy evolution .
when modeling galaxy evolution , an implementation of a sub - grid mass accretion prescription to estimate the gas flow onto smbhs is required for the sake of computational efficiency . in simulations of merging galaxies , some form of the classical bondi - hoyle - lyttleton ( bhl ) accretion prescription is usually implemented to estimate the smbh s feeding rate .
this prescription assumes that the properties of gas at hundreds of parsecs accurately determine the mass accretion rate . in this paper
we argue that nscs , a common component of galactic centers at parsec scales , can provide an efficient mechanism for funneling gas towards the smbh at scales which are commonly unresolved in cosmological simulations . for the conditions expected to persist in the centers of merging galaxies , the resultant large central gas densities in nscs should produce enhanced accretion rates onto the embedded smbhs , especially if cooling is efficient . because these nscs are typically more massive than the central smbh
, they can significantly alter the gas flow before being accreted . while the model shown in figure [ fig : subgrid ] results in a modest increase in the final mass of the merged smbh , the presence of nscs result in faster smbh growth rates and higher bolometric luminosities than predicted by the standard bhl formalism .
obviously , these calculation are incomplete and would improve with a self - consistent implementation of feedback .
it has been suggested that the interplay between smbh mass and host galaxy properties indicates that black hole feedback during mergers alters the properties of gas at galactic scales in order to shape these observed correlations @xcite .
progress in our understanding of these processes and higher resolution simulations will be necessary before we can conclude that quasar feedback is in fact an essential ingredient . with more accurate simulations of the growth of smbhs surrounded by nsc in galaxy mergers
, we can better constrain the relevant physics responsible for the @xmath112 and @xmath113 relations from comparisons to observational data .
+ + we acknowledge helpful discussions with doug lin , elena donghia , anil seth , morgan macleod and lars hernquist .
we also thank the anonymous referee for constructive comments .
support was provided by the david and lucile packard foundation and nsf grant ast-0847563 , simons foundation and nasa grant nnx11ai97 g .
ccccccccc1a & 2.5 & 2.5 & 280 & 5.3 & @xmath114 & @xmath47 & 1.31 + 2a ( _ heavy _ ) & 2.5 & 2.5 & 280 & 5.3 & @xmath115 & 1.33 & 2.12 + 2b ( _ light _ ) & 1.0 & 1.0 & 177 & 5.3 & @xmath115 & 1.33 & 2.12 + 4a & 0 & 2.5 & naked bh & naked bh & @xmath115 & 1.33 & naked bh + 4b & 2.5 & 2.5 & 198 & 10.6 & @xmath115 & 1.33 & 1.50 + 4c & 2.5 & 2.5 & 280 & 5.3 & @xmath115 & 1.33 & 2.12 + 5a & 0 & 2.5 & naked bh & naked bh & 1.1 & 1.64 & naked bh + 5b & 2.5 & 2.5 & 198 & 10.6 & 1.1 & 1.64 & 0.93 + 5c & 2.5 & 2.5 & 280 & 5.3 & @xmath114 & @xmath47 & 1.31 + 7a & 0 & 0.1 & naked bh & naked bh & @xmath114 & @xmath116 & naked bh + 7b & 0.1 & 0.1 & 115 & 2.0 & @xmath114 & @xmath116 & 0.58 + 7c & 0 & 0.1 & naked bh & naked bh & @xmath115 & @xmath116 & naked bh + 7d & 0.1 & 0.1 & 115 & 2.0 & @xmath115 & @xmath116 & 0.58 + 8a & 0 & 0.1 & naked bh & naked bh & @xmath114 & @xmath116 & naked bh + 8b & 0.1 & 0.1 & 115 & 2.0 & @xmath114 & @xmath116 & 1.2 + 8c & 0 & 0.1 & naked bh & naked bh & @xmath115 & @xmath116 & naked bh + 8d & 0.1 & 0.1 & 115 & 2.0 & @xmath115 & @xmath116 & 1.2 frank , j. , king , a. , & raine , d. j. 2002 , accretion power in astrophysics , by juhan frank and andrew king and derek raine , pp . 398 .
isbn 0521620538 .
cambridge , uk : cambridge university press , february 2002 . , | during galaxy mergers the gas falls to the center , triggers star formation , and feeds the rapid growth of supermassive black holes ( smbhs ) .
smbhs respond to this fueling by supplying energy back to the ambient gas .
numerical studies suggest that this feedback is necessary to explain why the properties of smbhs and the formation of bulges are closely related .
this intimate link between the smbh s mass and the large scale dynamics and luminosity of the host has proven to be a difficult issue to tackle with simulations due to the inability to resolve all the relevant length scales simultaneously . in this paper
we simulate smbh growth at high - resolution with _ flash _ , accounting for the gravitational focusing effects of nuclear star clusters ( nscs ) , which appear to be ubiquitous in galactic nuclei . in the simulations ,
the nsc core is resolved by a minimum cell size of about 0.001 pc or approximately @xmath0 of the cluster s radius .
we discuss the conditions required for effective gas funneling to occur , which are mainly dominated by a relationship between nsc velocity dispersion and the local sound speed , and provide a sub - grid prescription for the augmentation of central smbh accretion rates in the presence of nscs . for the conditions expected to persist in the centers of merging galaxies , the resultant large central gas densities in nscs should produce drastically enhanced embedded smbh accretion rates - up to an order of magnitude increase can be achieved for gas properties resembling those in large - scale galaxy merger simulations .
this will naturally result in faster black hole growth rates and higher luminosities than predicted by the commonly used bondi - hoyle - lyttleton accretion formalism . |
the discovery of topologically non - trivial states of matter opens up a new realm of knowledge for fundamental condensed matter physics . unlike conventional materials , these topological insulators " exhibit metallic surface states that are protected by time reversal symmetry , while maintaining an insulating bulk electronic structure
this leads to a variety of novel properties including odd number of surface dirac fermions , strict prohibition of back - scattering , etc . , paving the way to potential technical breakthroughs in e.g. quantum computing process via the application of spintronics@xcite .
recently , extensive theoretical and experimental efforts have led to the realization of such fascinating behaviors in e.g. the hgte quantum wells@xcite , the bi@xmath3sb@xmath4 system@xcite and the bi@xmath5x@xmath6 ( x = te , se ) binary compounds@xcite .
numerous half - heusler ternary compounds have been proposed , theoretically , to be potential new platforms for topological quantum phenomena@xcite , where the inherent flexibility of crystallographic , electronic and superconducting parameters provide a multidimensional basis for both scientific and technical exploration .
the experimental determination of their topological class would set the basis for possible spintronic utilization and further studies on the interplay between the topological quantum phenomena versus e.g. the magnetic@xcite , superconducting@xcite and heavy fermionic@xcite behaviors .
theoretically , the topological insulators experience a gapless surface state protected by time reversal symmetry and thus are robust against scattering from local impurities .
such a surface state is one half " of a normal metal in that the surface bands are strongly spin - polarized , forming a unique spin helical texture@xcite . on the other hand
, the kramers theorem requires that the spin be degenerate at the kramers points - @xmath7-points of the surface brillouin zone where time reversal symmetry is preserved@xcite . at the interface between , say , a normal spin - orbit system and vacuum ,
the spin - polarized surface bands connect pairwise ( kramers pair ) , crossing the chemical potential @xmath8 an even number of times between two distinct kramers points . at the interface between a topologically non - trivial material and vacuum
, however , one expects the surface bands to cross @xmath8 an odd number of times@xcite .
in this paper we present a systematic survey on the surface electronic structure of half - heusler compounds _ _ r__ptbi ( _ r _ = lu , dy , gd ) using angle resolved photoemission spectroscopy ( arpes ) .
our results show clear spin - orbit splitting of the surface bands that cross the chemical potential , which is nicely reproduced in the full - potential augmented plane wave calculation for a surface electronic state .
the kramers degeneracy of spin is unambiguously detected at both the @xmath1 and @xmath2 points .
no direct indication of topologically non - trivial behavior is detected , except for the fact that there is a weak fermi crossing in the close vicinity to the @xmath1 point , making the total number of crossings five . in the surface band calculation , however , this inner crossing is explained by two spin - orbit splitting bands that are very close to each other , forming another kramers pair . in this band configuration
, the total berry phase would be zero for the half - heusler systems , and they would not be topologically non - trivial .
the detailed topological class of this family of materials thus remains an open question , requiring a detailed spin - resolved arpes study with ultra - high momentum resolution and a direct calculation of the topological invariants based on the first principle band structure .
inset of ( c ) enhanced the arpes intensity near @xmath2 and @xmath9 for better visibility of the bands .
( d ) calculational band structure with respect to ( c ) .
( e)-(f ) expanded figures for ( b ) and ( d ) , respectively , showing six fermi crossings .
panel ( e ) is rotated by 30@xmath10 with respect to ( b).,width=307 ]
single crystals of _ _ r__ptbi ( _ r _ = lu , dy , gd ) were grown out of a bi flux and characterized by room temperature power x - ray diffraction measurements@xcite .
the crystals grow as partial octahedra with the ( 111 ) facets exposed .
typical dimensions of a single crystal are about @xmath11 @xmath12 .
the arpes measurements were performed at beamline 10.0.1 of the advanced light source ( als ) , berkeley , california using a scienta r4000 electron analyzer .
vacuum conditions were better than @xmath13 torr .
all arpes data was taken at @xmath14 k , above the magnetic ordering temperatures of all compounds@xcite .
the energy resolution was set at @xmath15 15 mev .
all samples were cleaved _ in situ _ , yielding clean ( 111 ) surfaces in which atoms arrange in a hexagonal lattice .
high symmetry points for the surface brillouin zone are defined as @xmath16 , @xmath17 and @xmath18 with unit momentum @xmath19 , where @xmath20 is the lattice constant for each type of crystals .
we emphasize here that no stress or pulling force is felt by the samples , which ensures that the measured data reveals the intrinsic electronic structure of the single crystals .
we begin this survey in fig . 1 by showing the fermi maps of the three half - heusler compounds _ _ r__ptbi ( _ r _ = lu , dy , gd ) . previous theoretical calculations for the bulk electronic structure@xcite suggested that the kramers crossing at the @xmath1 point happens very close to @xmath8 ; the fermi surface reduces to a single point ( dirac point ) at @xmath1 . the data in fig .
1 shows that , at least in the ( 111 ) cleaving plane , this is not the case .
instead there are several bands crossing @xmath8 in the vicinity of both the @xmath1 and @xmath2 points .
the overall fermi surface for all three half - heusler compounds are similar , indicating a similar cleaving plane and band structure for all members . by comparing the band structure measured at the ( 111 ) surface with results of band calculations for gdptbi ( fig .
2 ) , we find the cleaving plane to be bi(111 ) , marked by a red parallelogram in fig .
1(a ) . a closer look at fig .
1(c)-(e ) reveals that the @xmath1 pockets have different sizes for different heusler members .
for example the circular @xmath1 pockets in luptbi are larger in size than those in gdptbi .
this indicates a different effective electron occupancy for different members of the half - heusler family .
one should also note that in fig .
1(e ) the inner of the two bright @xmath1 pockets is hexagonal in shape , reminiscent of the hexagonal shape of the dirac cone in bi@xmath5te@xmath6 ( ref .
@xcite ) , which is explained by higher order terms in the @xmath21 hamiltonian@xcite .
this hexagonal shape is very nicely reproduced in the calculation [ fig .
2(b ) ] . for clarifying the topological class of the half - heuslers ,
two immediate questions follow the observations in fig . 1 : ( 1 ) are the observed bands actually arising due to the sample surface ? ( 2 ) exactly how many times do the bands intersect the chemical potential along the @xmath1-@xmath2 line segment ? [ red box in fig . [
fig1](c ) ] .
data is taken on luptbi samples .
( a)-(d ) binding energy dependence of band structure near @xmath2 .
map location in the surface brillouin zone is shown in panel ( e ) .
( f ) theoretical band map at the chemical potential for gdptbi . ( g),(h )
band maps for two perpendicular directions marked by red lines in ( g ) .
there are in total two fermi crossings along the @xmath1-@xmath2 line segment at the vicinity of @xmath2.,width=336 ] fig .
2 shows the comparison between the arpes data and a calculational surface state in gdptbi . for both the band structure and fermi surface calculation
, we used a full - potential linear augmented plane wave ( fplapw ) method@xcite with a local density functional@xcite .
the crystallographic unit cell is generated such that the ( 111 ) direction of the _ fcc _ brillouin zone points along the @xmath22-axis .
for calculation of the surface electronic structure , supercells with three unit cell layers and a 21.87 a.u .
vacuum is constructed .
we calculated band structures of all six possible surface endings ( gd - bi - pt - bulk , gd - pt - bi - bulk , bi - gd - pt - bulk , bi - pt - gd - bulk , pt - gd - bi - bulk , pt - bi - gd - bulk ) ; only the bi - pt - gd - bulk construction shows good agreement with experiment [ fig .
2(b ) , ( d)-(f ) ] .
structural data were taken from a reported experimental result@xcite . to obtain the self - consistent charge density , we chose 48 @xmath7-points in the irreducible brillouin zone , and set @xmath23 to 7.5 , where @xmath24 is the smallest muffin - tin radius and @xmath25 is the plane - wave cutoff .
we used muffin - tin radii of 2.5 , 2.4 and 2.4 a.u .
for gd , bi , and pt respectively . for the non - magnetic - state calculation valid for comparison with arpes results at 15 k , the seven 4__f _
_ electrons of gd atoms were treated as core electrons with no net spin polarization .
the atoms near the surface ( bi , pt , gd ) were relaxed along the @xmath22-direction until the forces exerted on the atoms were less than 2.0 mry / a.u .. with this optimized structure , we obtained self - consistency with 0.01 mry / cell total energy convergence .
after that , we calculated the band structure and two dimensional fermi surface in which we divided the rectangular cell connecting four @xmath9-points by @xmath26 , yielding 1681 @xmath7-points . even at first glance , fig .
2 gives the impression of remarkable agreement between theory and experiment .
all basic features observed by arpes - the overall shape and location of the fermi pockets [ fig .
2(a)-(b ) ] , the binding energies of the bands [ fig .
2(c)-(d ) ] - are well reproduced by the calculation .
the main point of this figure , however , is the fact that band calculations show a total of six fermi crossings along the @xmath1-@xmath2 line segment , which is an even number and is not directly consistent of the proposed strong topological insulating phenomenon@xcite .
it should be noted that , in order to take into account the spin - orbit splitting , relativistic effects are applied to the calculation .
similar calculations reproduce clear topological insulating behavior in bi@xmath5te@xmath6 thin films@xcite .
the excellent agreement shown in fig .
2 also implies the validity of such calculation in half - heusler compounds .
in fact traces for the inner two crossings is also found in the arpes data , where they appear to be one single crossing , most likely due to finite momentum resolution [ leftmost part in fig .
2(c ) , see also fig .
3(d)-(h ) ] . in fig .
3 we prove that the observed bands come from the sample surface .
this is done by scanning the incident photon energy along both @xmath1-@xmath9 and @xmath1-@xmath2 high symmetry directions .
varying the photon energy in arpes effectively changes the momentum offset along the direction perpendicular to the sample surface . in our case , this direction corresponds to @xmath27 or the ( 111 ) direction of the _ fcc _ brillouin zone .
3(a)-(b ) show that all observed bands form straight lines along the @xmath27 direction , a clear indication for the lack of @xmath27 dependence . in fig .
3(c ) we compare this to a calculated fermi surface map for the _ bulk _ bands , along the same direction as in fig .
the difference is clear : the bulk bands are dispersive along the @xmath28-@xmath29 direction ; and most of the experimentally observed bands are not present in the calculation . in figs .
3(d)-(h ) we pay special attention to the bands crossing @xmath8 near @xmath1 by showing the band structure for four different photon energies . in total
there are at least three fermi contours surrounding @xmath1 , the outer two being a lot brighter than the inner one ( or two , see discussion for fig .
2 ) . as shown in fig .
3(h ) , these three ( or four ) bands cross @xmath8 at exactly the same @xmath7 positions for all photon energies .
therefore all of them are surface bands . the data in fig .
3 thus show , unambiguously , that a metallic surface electronic state exists in the half - heusler compounds . the exact number of fermi crossings along the @xmath1-@xmath2 line segment is also examined in fig .
4 . the main conclusion for fig . 4
is that there are also three ( or four ) visible fermi crossings at the vicinity of @xmath1 between these two kramers points .
we show these bands on the luptbi and gdptbi samples . both on the band dispersion maps [ figs .
4(a)-(b ) ] and the momentum distribution curves [ mdcs , figs .
4(c)-(d ) ] we see that there are two bright hole - like bands almost parallel to each other , and a much weaker inner band with lower fermi velocity .
this inner band is not easy to see in the band maps ( nontheless indicated by green arrows ) , but is clearly visible in the mdcs by small intensity peaks tracing down from the one marked by a green bar [ also marked by a green color in figs .
4(e)-(f ) ] .
the same band also exists in the @xmath1-@xmath9 direction [ figs .
3(d)-(h ) ] . same as the discussion for figs . 2 and 3 ,
this inner crossing is reproduced in the band calculation by two closely located spin - orbit - splitting bands that form a kramers pair .
the brighter parallel bands form a second kramers pair of opposite spins . in fig .
4(e)-(f ) we show the linear extrapolation of the two brighter bands .
in gdptbi they are likely to reduce to a dirac point at about 0.4 ev above @xmath8 .
if the total number of crossing is four , such a configuration will give zero contribution to the total berry phase .
5 we examine the bands near the @xmath2 point . the @xmath7-space location of the arpes maps [ figs .
5(a)-(d ) ] is shown in fig .
panels 5(g)-(h ) present the band dispersion maps for two cuts crossing @xmath2 , whose positions are marked in panel 5(f ) with the band calculation result .
5(a)-(d ) show that the @xmath2 bands form a very special shape . at high binding energies
[ @xmath30 ev , fig .
5(d ) ] , two u - shape bands are well separated . as binding energy decreases
these two bands merge into each other and hybridize to form a central elliptical contour and two curly - bracket - like segments .
the segments near each @xmath2 points link together , forming another large fermi contour enclosing the zone center @xmath1 .
it is clear from fig .
5(g)-(h ) that there are two fermi crossings in both the @xmath1-@xmath9 and @xmath1-@xmath2 directions .
the special shape of the fermi surface is formed by two bands that are likely to be members of another kramers pair .
kramers degeneracy of spin happens at @xmath31 mev below @xmath8 .
all this features are obtained with our calculation for the surface states [ fig 2(b ) and 2(e ) ] . these two bands also give zero contribution to the total berry phase . in summary
, we performed an arpes survey on the electronic structure of three half - heusler compounds _ _ r__ptbi ( _ r _ = lu , dy , gd ) which are proposed to be topological insulators .
our result show unambiguously that these materials have a metallic surface state markedly different from the calculational result on the bulk electronic structures .
this surface state is reproduced with high accuracy in our band calculations .
both experiment and theory reveal several bands that cross the fermi level .
knowledge of the exact number of these bands is possibly limited by experimental momentum resolution .
no direct consistency with the proposed strong topological insulating behavior is found in the arpes results . for final determination of their topological classes ,
both an apres measurement of ultrahigh @xmath7-resolution and a direct calculation of the first chern number as a topological invariant @xcite are in need .
we thank s .- c . zhang and j. schmalian for instructive discussions as well as sung - kwan mo for grateful instrumental support at the als .
ames laboratory was supported by the department of energy - basic energy sciences under contract no .
de - ac02 - 07ch11358 .
als is operated by the us doe under contract no .
de - ac03 - 76sf00098 .
d. hsieh , y. xia , d. qian , l. wray , j. h. dil , f. meier , j. osterwalder , l. patthey , j. g. checkelsky , n. p. ong , a. v. fedorov , h. lin , a. bansil , d. grauer , y. s. hor , r. j. cava and m. z. hasan , nature ( london ) * 460 * , 1101 ( 2009 ) .
p. blaha , k. schwarz , g. k. h. madsen , d. kvasnick and j. luitz , wien2k , an augmented plane wave plus local orbitals program for calculation crystal properties ( k. schwarz , tu wien , austria , 2001 ) isbn 3 - 9501031 - 1 - 2 . | rare - earth platinum bismuth ( @xmath0ptbi ) has been recently proposed to be a potential topological insulator . in this paper
we present measurements of the metallic surface electronic structure in three members of this family , using angle resolved photoemission spectroscopy ( arpes ) .
our data shows clear spin - orbit splitting of the surface bands and the kramers degeneracy of spins at the @xmath1 and @xmath2 points , which is nicely reproduced with our full - potential augmented plane wave calculation for a surface electronic state .
no direct indication of topologically non - trivial behavior is detected , except for a weak fermi crossing detected in close vicinity to the @xmath1 point , making the total number of fermi crossings odd . in the surface band calculation , however , this crossing is explained by another kramers pair where the two splitting bands are very close to each other .
the classification of this family of materials as topological insulators remains an open question . |
ognitive radio ( cr ) , since the name was coined by mitola in his seminal work @xcite , has drawn intensive attentions from both academic ( see , e.g. , @xcite and references therein ) and industrial ( see , e.g. , @xcite and references therein ) communities ; and to date , many interesting and important results have been obtained . in cr networks ,
the secondary users or crs usually communicate over the same bandwidth originally allocated to an existing primary radio network . in such a scenario ,
the cr transmitters usually need to deal with a fundamental tradeoff between maximizing the secondary network throughput and minimizing the resulted performance degradation of the active primary transmissions .
one commonly known technique used by the secondary users to protect the primary transmissions is _ opportunistic spectrum access _ ( osa ) , originally outlined in @xcite and later introduced by darpa , whereby the secondary user decides to transmit over a particular channel only when all primary transmissions are detected to be off . for osa ,
an enabling technology is to detect the primary transmission on / off status , also known as _ spectrum sensing _ , for which many algorithms have been reported in the literature ( see , e.g. , @xcite and references therein ) .
however , in practical situations with a nonzero misdetection probability for an active primary transmission , it is usually impossible to completely avoid the performance degradation of the primary transmission with the secondary user osa .
another approach different from osa for a cr to maximize its throughput and yet to provide sufficient protection to the primary transmission is allowing the cr to access the channel even when the primary transmissions are active , provided that the resultant interference power , or the so - called _ interference temperature _ ( it ) @xcite , @xcite , at each primary receiver is limited below a predefined value .
this spectrum sharing strategy is also referred to as spectrum underlay @xcite , @xcite or horizontal spectrum sharing @xcite , @xcite . with this strategy , _ dynamic resource allocation _
( dra ) becomes essential , whereby the transmit powers , bit - rates , bandwidths , and antenna beams of the secondary transmitters are dynamically allocated based upon the channel state information ( csi ) in the primary and secondary networks . a number of papers have recently addressed the design of optimal dra schemes to achieve the point - to - point cr channel capacity under the it constraints at the primary receivers ( see , e.g. , @xcite-@xcite ) . on the other hand , since the cr network is in nature a multiuser communication environment , it will be more relevant to consider dra among multiple secondary users in a cr network rather than that for the case of one point - to - point cr channel . deploying the interference - temperature constraint as a practical means to protect the primary transmissions , the conventional network models such as the multiple - access channel ( mac ) , broadcast channel ( bc ) , interference channel ( ic ) , and relay channel ( rc )
can all be considered for the secondary network , resulting in various new cognitive network models and associated problem formulations for dra ( see , e.g. , @xcite-@xcite ) .
it is also noted that there has been study in the literature on the information - theoretic limits of the cr channels by exploiting other types of `` cognitions '' available at the cr terminals different from the it , such as the knowledge of the primary user transmit messages at the cr transmitter @xcite , @xcite , the distributed detection results on the primary transmission status at the cr transmitter and receiver @xcite , the `` soft '' sensing results on the primary transmission @xcite , and the primary transmission on - off statistics @xcite . in this paper , we focus on the single - input single - output ( siso ) or single - antenna fading cognitive mac ( c - mac ) and cognitive bc ( c - bc ) for the secondary network , where @xmath0 secondary users communicate with the base station ( bs ) of the secondary network in the presence of @xmath1 primary receivers . it is assumed that the bs has the perfect csi on the channels between the bs and all the secondary users , as well as the channels from the bs and each secondary user to all the primary receivers .
thereby , the bs can implement a centralized dynamic power and rate allocation scheme in the secondary network so as to optimize its performance and yet maintain the interference power levels at all the primary receivers below the prescribed thresholds .
an information - theoretic approach is taken in this paper to characterize the maximum sum - rate of secondary users averaged over the channel fading states , termed as _ ergodic sum capacity _
, for both the fading c - mac and c - bc . the ergodic sum capacity can be a relevant measure for the maximum achievable throughput of the secondary network when the data traffic has a sufficiently - large delay tolerance .
as usual ( see , e.g. , @xcite ) , we consider both the long - term ( lt ) transmit - power constraint ( tpc ) that regulates the _ average _ transmit power across all the fading states at the bs or each of the secondary user , as well as the short - term ( st ) tpc that is more restrictive than the lt - tpc by limiting the _ instantaneous _ transmit power at each fading state to be below a certain threshold .
similarly , we also consider both the lt interference - power constraint ( ipc ) that regulates the resultant average interference power over fading at each primary receiver , and the st - ipc that imposes a more strict instantaneous limit on the resultant interference power at each fading state .
the major problem to be addressed in this paper is then to characterize the ergodic sum capacity of the secondary network under different combinations of lt-/st - tpc and lt-/st - ipc .
apparently , such a problem setup is unique for the fading cr networks .
moreover , we are interested in investigating the conditions over each case for the optimality of the dynamic time - division - multiple - access ( d - tdma ) scheme in the secondary network , i.e. , when it is optimal to schedule a single secondary user at each fading state for transmission to achieve the ergodic sum capacity .
these optimality conditions for d - tdma are important to know as when they are satisfied , the single - user decoding and encoding at the secondary bs becomes optimal for the c - mac and c - bc , respectively .
this can lead to a significant complexity reduction compared with the cases where these conditions are not satisfied such that the bs requires more complex multiuser decoding and encoding techniques to achieve the ergodic sum capacity .
information - theoretic studies can be found for the deterministic ( no fading ) siso - mac and siso - bc in , e.g. , @xcite , and for the fading ( parallel ) siso - mac and siso - bc in , e.g. , @xcite-@xcite and @xcite-@xcite , respectively .
in addition , d - tdma has been shown as the optimal transmission scheme to achieve the ergodic sum capacity of the fading siso - mac under the lt - tpc at each transmitter @xcite , @xcite .
thanks to the duality result on the capacity regions of the gaussian mac and bc @xcite , the optimality of d - tdma is also provable for the fading siso - bc to achieve the ergodic sum capacity .
however , to our best knowledge , characterizations of the ergodic sum capacities as well as the optimality conditions for d - tdma over the fading c - mac and c - bc under various mixed transmit - power and interference - power constraints have not been addressed yet in the literature . in this paper
, we will provide the solutions to these problems .
the main results of this paper are summarized below for a brief overview : * for the fading cognitive siso - mac , we show that d - tdma is optimal for achieving the ergodic sum capacity when the lt - tpc is applied jointly with the lt - ipc .
this result is an extension of that obtained earlier in @xcite for the traditional fading siso - mac without the lt - ipc . for the other three cases of mixed power constraints ,
i.e. , lt - tpc with st - ipc , st - tpc with lt - ipc , and st - tpc with st - ipc , we show that although d - tdma is in general a suboptimal scheme and thus does not achieve the ergodic sum capacity , it can be optimal under some special conditions .
we formally derive these conditions from the karush - kuhn - tucker ( kkt ) conditions @xcite associated with the capacity maximization problems . in particular , for the case of lt - tpc with st - ipc , we show that the optimal number of secondary users that transmit at the same time should be no greater than @xmath2 .
therefore , for small values of @xmath1 , e.g. , @xmath3 corresponding to a single primary receiver , d - tdma is close to being optimal .
furthermore , for all cases considered , we derive the optimal transmit power - control policy for the secondary users to achieve the ergodic sum capacity .
for the two cases of lt - tpc with lt - ipc and st - tpc with lt - ipc , we provide the closed - form solutions for the optimal power allocation at each fading state .
particularly , in the case of st - tpc with lt - ipc , we show that for the active secondary users at one particular fading state , there is at most one user that transmits with power lower than its st power constraint , while all the other active users transmit with their maximum powers . * for the fading cognitive siso - bc , we show that for all considered cases of mixed power constraints , d - tdma is optimal for achieving the ergodic sum capacity .
the optimal transmit power allocations at the bs in these cases have closed - form solutions , which resemble the single - user `` water - filling ( wf ) '' solutions for the well - known fading ( parallel ) gaussian channels @xcite , @xcite . the rest of this paper is organized as follows .
section [ sec : system model ] provides the system model for the fading c - mac and c - bc .
section [ sec : mac ] and section [ sec : bc ] then present the results on the ergodic sum - capacity , the associated optimal power - control policy , and the optimality conditions for d - tdma , for the fading c - mac and c - bc , respectively , under different mixed lt / st transmit - power and interference - power constraints . section [ sec : numerical results ] provides the numerical results on the ergodic sum capacities of the fading c - mac and c - bc under different mixed power constraints , the capacities with vs. without the tdma constraint , and those with vs. without the optimal power control , and draws some insightful observations pertinent to the optimal dra in cr networks .
finally , section [ sec : conclusions ] concludes this paper .
consider a fading c - mac as shown in fig .
[ fig : c - mac model ] , where @xmath0 crs or secondary users ( sus ) transmit to the secondary bs by sharing the same narrow band with @xmath1 primary receivers ( prs ) , and all terminals are assumed to be equipped with a single antenna each .
a _ block - fading _
( bf ) channel model is assumed for all the channels involved .
furthermore , since this paper considers coherent communications , only the fading channel power gains ( amplitude squares ) are of interest . during each transmission block , the power gain of the fading channel from the @xmath4-th su to the secondary bs is denoted by @xmath5 , while that of the fading channel from the @xmath4-th su to the @xmath6-th pr is denoted by @xmath7 , @xmath8 .
these channel power gains are assumed to be drawn from a vector random process , which we assume to be ergodic over transmission blocks and have a continuous , differentiable joint cumulative distribution function ( cdf ) , denoted by @xmath9 , where @xmath10 $ ] denotes the power gain vector for all the channels of interest .
we further assume that @xmath5 s and @xmath7 s are independent .
in addition , it is assumed that the additive noises ( including any additional interferences from the outside of the secondary network , e.g. , the primary transmitters ) at the secondary bs are independent circular symmetric complex gaussian ( cscg ) random variables , each having zero mean and unit variance , denoted as @xmath11 .
since in this paper we are interested in the information - theoretic limits of the c - mac , it is assumed that the optimal gaussian codebook is used by each su transmitter .
it is assumed that the secondary bs knows _ a priori _ the channel distribution information @xmath9 and furthermore the channel realization @xmath12 at each transmission block .
thereby , the secondary bs is able to schedule transmissions of sus and allocate their transmit power levels and rate values at each transmission block , so as to optimize the performance of the secondary network and yet provide a necessary protection to each of the prs .
we denote the transmit power - control policy for sus as @xmath13 , which specifies a mapping from the fading channel realization @xmath12 to @xmath14 $ ] , where @xmath15 denotes the transmit power assigned to the @xmath4-th su . the long - term ( lt ) transmit - power constraint ( tpc ) for the @xmath4-th su , @xmath16 ,
can then be described as @xmath17\leq p^{\rm lt}_k\ ] ] where the expectation is taken over @xmath12 with respect to ( w.r.t . ) its cdf , @xmath9 , and the short - term ( st ) transmit - power constraint ( tpc ) for the @xmath4-th su is given as @xmath18 similarly , we consider both the lt and st interference - power constraints ( ipcs ) at the @xmath6-th pr , @xmath19 , described as @xmath20\leq \gamma^{\rm lt}_m\ ] ] @xmath21 respectively . for a given @xmath13 ,
the maximum achievable sum - rate ( in nats / complex dimension ) of sus averaged over all the fading states can be expressed as ( see , e.g. , @xcite ) @xmath22.\ ] ] the ergodic sum capacity of the fading c - mac can then be defined as @xmath23 where @xmath24 is the feasible set specified by a particular combination of the lt - tpc , st - tpc , lt - ipc and st - ipc .
note that all of these power constraints are affine and thus specify convex sets of @xmath15 s , so does any of their arbitrary combinations .
therefore , the capacity maximization in ( [ eq : sum capacity mac ] ) is in general a _ convex _ optimization problem , and thus efficient numerical algorithms are available to obtain its solutions . in this paper , we consider @xmath24 to be generated by one of the following four possible combinations of power constraints , which are lt - tpc with lt - ipc , lt - tpc with st - ipc , st - tpc with lt - ipc , and st - tpc with st - ipc , for the purpose of exposition .
next , we consider the siso fading c - bc as shown in fig .
[ fig : c - bc model ] , where the secondary bs transmits to @xmath0 sus while possibly interfering with each of the @xmath1 prs . without loss of generality , we use the same notation , @xmath5 , to denote the channel power gain from the bs to the @xmath4-th su , @xmath16 , as for the c - mac .
the interference channel power gains from the bs to prs are denoted as @xmath25 , @xmath19 , which are assumed to be mutually independent and also independent of @xmath5 s .
similar to the c - mac case , let @xmath26 $ ] denote the power gain vector for all the channels involved in the c - bc , which we assume to be drawn from an ergodic vector random process with a continuous , differentiable joint cdf , denoted by @xmath27 .
it is assumed that the additive noises at all su receivers are independent cscg random variables each distributed as @xmath11 ; and the optimal gaussian codebook is used by the transmitter of the bs . with the available channel distribution information @xmath27 as well as the csi on @xmath5 s and @xmath25 s at each transmission block , the secondary bs designs its downlink transmissions to the sus by dynamically allocating its transmit power levels and rate values .
let @xmath28 denote the transmit power - control policy for the secondary bs , which specifies a mapping from the fading channel realization @xmath29 to its transmit power @xmath30 .
similarly as for c - mac , we define the lt - tpc and st - tpc for the secondary bs as @xmath31\leq q^{\rm lt}\ ] ] where the expectation is taken over @xmath29 w.r.t .
its cdf , @xmath27 , and @xmath32 respectively ; and the lt - ipc and st - ipc at the @xmath6-th pr , @xmath19 , as @xmath33\leq \gamma^{\rm lt}_m\ ] ] and @xmath34 respectively .
now , consider an auxiliary siso fading c - mac for the siso fading c - bc of interest , where @xmath35 s remain the same as in the c - bc while @xmath36 .
thus , the channel realization @xmath12 in this auxiliary c - mac can be concisely represented by @xmath29 in the c - bc . by applying the mac - bc duality result @xcite at each fading state , for a given @xmath30 , the maximum sum - rate of the c - bc
can be obtained from its auxiliary c - mac as @xmath37 therefore , the ergodic sum capacity of the fading c - bc can be equivalently obtained from its auxiliary fading c - mac as @xmath38 where @xmath39 is specified by a particular combination of ( [ eq : lt tpc bc])-([eq : st ipc bc ] ) , with @xmath30 being replaced by @xmath40 .
note that we can obtain the optimal power - control policy @xmath28 to achieve the ergodic sum capacity of the c - bc from the corresponding optimal @xmath13 by solving the maximization problem in ( [ eq : sum capacity bc new ] ) .
similarly as for @xmath41 in ( [ eq : sum capacity mac ] ) , it can be shown that the optimization problem for obtaining @xmath42 in ( [ eq : sum capacity bc new ] ) is convex .
in this section , we consider the siso fading c - mac under different mixed transmit - power and interference - power constraints . for each case , we derive the optimal power - control policy for achieving the ergodic sum capacity , as well as the conditions for the optimality of d - tdma .
@xmath43-\sum_{k=1}^k\lambda_k\ { \mathbb{e}[p_k({\mbox{\boldmath{$ \alpha $ } } } ) ] - p^{\rm lt}_k\ } -\sum_{m=1}^m\mu_m\left\{\mathbb{e}\left[\sum_{k=1}^kg_{km}p_k({\mbox{\boldmath{$ \alpha $ } } } ) \right]- \gamma^{\rm lt}_m\right\}\end{aligned}\ ] ] from ( [ eq : sum rate mac ] ) and ( [ eq : sum capacity mac ] ) , the ergodic sum capacity under the lt - tpc and the lt - ipc can be obtained by solving the following optimization problem : [ prob : lt lt mac ] @xmath44 \\ \mathtt { subject \ to \
( s.t . ) } & & ( \ref{eq : lt tpc mac } ) , ( \ref{eq : lt ipc mac}).\end{aligned}\ ] ] the proposed solution to the above problem is based on the lagrange duality method .
first , we write the lagrangian of this problem as in ( [ eq : lagrangian lt lt ] ) ( shown on the next page ) , where @xmath45 and @xmath46 are the nonnegative dual variables associated with each corresponding power constraint in ( [ eq : lt tpc mac ] ) and ( [ eq : lt ipc mac ] ) , respectively , @xmath16 , @xmath19 .
then , the lagrange dual function , @xmath47 , is defined as @xmath48 the dual function serves as an upper bound on the optimal value of the original ( primal ) problem , denoted by @xmath49 , i.e. , @xmath50 for any nonnegative @xmath45 s and @xmath46 s .
the dual problem is then defined as @xmath51 let the optimal value of the dual problem be denoted by @xmath52 , which is achievable by the optimal dual solutions @xmath53 and @xmath54 , i.e. , @xmath55 . for a convex optimization problem with a strictly feasible point as in our problem , the slater s condition @xcite is satisfied and thus the duality gap , @xmath56 , is indeed zero .
this result ensures that problem [ prob : lt lt mac ] can be equivalently solved from its dual problem , i.e. , by first maximizing its lagrangian to obtain the dual function for some given dual variables , and then minimizing the dual function over the dual variables .
consider first the problem for obtaining @xmath47 with some given @xmath45 s and @xmath46 s .
it is interesting to observe that this dual function can also be written as @xmath57 + \sum_{k=1}^k\lambda_kp^{\rm lt}_k + \sum_{m=1}^m \mu_m\gamma^{\rm lt}_m\end{aligned}\ ] ] where @xmath58 thus , the dual function can be obtained via solving for sub - dual - function @xmath59 s , each for one fading state with channel realization , @xmath12 .
notice that the maximization problems in ( [ eq : dual function per state lt lt ] ) with different @xmath12 s all have the same structure and thus can be solved using the same computational routine . for conciseness
, we drop the @xmath12 in @xmath60@xmath61 s for the maximization problem at each fading state and express it as [ prob : lt lt mac per state ] @xmath62 this problem is convex since its objective function is concave and its constraints are all linear . by introducing nonnegative dual variables @xmath63 , for the corresponding constraints on the nonnegativity of @xmath64 s , we can write the following kkt conditions @xcite that need to be satisfied by the optimal primal and dual solutions of problem [ prob : lt lt mac per state ] , denoted as @xmath65 and @xmath66 , respectively . @xmath67 with @xmath68 and @xmath69 .
the following lemma can then be obtained from these kkt optimality conditions : [ lemma : tdma opt lt lt ] the optimal solution of problem [ prob : lt lt mac per state ] has at most one user indexed by @xmath70 , @xmath71 , with @xmath72 , i.e. , the solution follows a d - tdma structure
. please refer to appendix [ appendix : proof tdma opt lt lt ] .
given lemma [ lemma : tdma opt lt lt ] , the remaining tasks for solving problem [ prob : lt lt mac per state ] are to find the user that transmits at each fading state as well as the optimal transmit power , which are given by the following lemma : [ lemma : opt power user lt lt ] in the optimal solution of problem [ prob : lt lt mac per state ] , let @xmath70 denote the user that has @xmath73 , and @xmath74 be any of the other users that has @xmath75 , @xmath76
. then user @xmath70 must satisfy @xmath77 the optimal power allocation of user @xmath70 is @xmath78 where @xmath79
. please refer to appendix [ appendix : proof opt user power lt lt ]
. solutions of problem [ prob : lt lt mac per state ] across all the fading states are basically an optimal mapping between an arbitrary channel realization and the transmit power allocation for any given @xmath45 s and @xmath46 s , which can then be used to obtain the dual function @xmath47 .
next , the dual function needs to be minimized over @xmath45 s and @xmath46 s to obtain the optimal dual solutions @xmath80 s and @xmath81 s with which the duality gap is zero .
one method to iteratively update @xmath45 s and @xmath46 s toward their optimal values is the ellipsoid method @xcite , of which we omit the details here for brevity .
lemma [ lemma : tdma opt lt lt ] suggests that at each fading state , at most one su can transmit , i.e. , d - tdma is optimal . since this result holds for any given @xmath45 s and @xmath46 s , it must be true for the optimal dual solutions @xmath80 s and @xmath81 s under which the optimal value of the original problem or the ergodic sum capacity is achieved .
therefore , we have the following theorem : [ theorem : tdma opt lt lt ] d - tdma is optimal across all the fading states for achieving the ergodic sum capacity of the fading c - mac under the lt - tpc jointly with the lt - ipc . the optimal rules to select the su for transmission at a particular fading state and to determine its transmit power are given by lemma [ lemma : opt power user lt lt ] with all @xmath45 s and @xmath46 s replaced by their optimal dual solutions for problem [ prob : lt lt mac ]
. notice that if the lt - ipc given by ( [ eq : lt ipc mac ] ) is not present in problem [ prob : lt lt mac ] , or equivalently , the lt - ipc values @xmath82 s are sufficiently large such that these constraints are inactive with the optimal power solutions of problem [ prob : lt lt mac ] , it is then easy to verify from its kkt conditions that the optimal dual solutions for all @xmath46 s must be equal to zero .
from ( [ eq : user ordering lt lt ] ) , it then follows that only user @xmath70 with the largest @xmath83 among all the users can probably transmit at a given fading state .
this result is consistent with that obtained earlier in @xcite for the traditional fading siso - mac without the lt - ipc . however , under the additional lt - ipc , from ( [ eq : user ordering lt lt ] ) and ( [ eq : opt power lt lt ] ) it is observed that the selected su for transmission and its transmit power depend on the interference - power `` prices '' @xmath46 s for different prs and the instantaneous interference channel power gains @xmath7 s .
the ergodic sum capacity under the lt - tpc but with the st - ipc can be obtained as the optimal value of the following problem : [ prob : lt st mac ] @xmath84 \\ \mathtt { s.t .
} & & ( \ref{eq : lt tpc mac } ) , ( \ref{eq : st ipc mac}).\end{aligned}\ ] ] similar to problem [ prob : lt lt mac ] , we apply the lagrange duality method to solve the above problem .
however , different from problem [ prob : lt lt mac ] that has both the long - term transmit - power and interference - power constraints , it is noted that in problem [ prob : lt st mac ] , only the transmit - power constraints are long - term while the interference - power constraints are short - term .
therefore , the dual variables associated with the long - term constraints should be introduced first , in order to decompose the problem into individual subproblems over different fading states , to each of which the corresponding short - term constraints can then be applied .
let @xmath45 be the nonnegative dual variable associated with the corresponding lt - tpc in ( [ eq : lt tpc mac ] ) , @xmath16 .
the lagrangian of this problem can then be written as @xmath85 \nonumber \\ & ~ -\sum_{k=1}^k\lambda_k\left\ { \mathbb{e}\left[p_k({\mbox{\boldmath{$ \alpha $ } } } ) \right]- p^{\rm lt}_k\right\}.\end{aligned}\ ] ] let @xmath86 denote the set of @xmath87 specified by the remaining st - ipc in ( [ eq : st ipc mac ] ) . the lagrange dual function
is then expressed as @xmath88 the dual problem is accordingly defined as @xmath89 .
similar to problem [ prob : lt lt mac ] , it can be verified that the duality gap is zero for the convex optimization problem addressed here ; and thus solving its dual problem is equivalent to solving the original problem .
consider first the problem for obtaining @xmath90 with some given @xmath45 s .
similar to problem [ prob : lt lt mac ] , this dual function can be decomposed into individual sub - dual - functions , each for one fading state , i.e. , @xmath91 + \sum_{k=1}^k\lambda_kp^{\rm lt}_k\end{aligned}\ ] ] where @xmath92 with @xmath93 denoting the subset of @xmath86 corresponding to the fading state with channel realization @xmath12 . after dropping the @xmath12 in the corresponding maximization problem in ( [ eq : dual function per state lt st ] ) for a particular fading state
, we can express this problem as [ prob : lt st mac per state ] @xmath94 the above problem is convex , but in general does not have a closed - form solution .
nevertheless , it can be efficiently solved by standard convex optimization techniques , e.g. , the interior point method @xcite , or alternatively , via solving its dual problem ; and for brevity , we omit the details here . after solving problem [ prob : lt st mac per state ] for all the fading states ,
we can obtain the dual function @xmath90 .
next , the minimization of @xmath90 over @xmath45 s can be resolved via the ellipsoid method , similarly like that for problem [ prob : lt lt mac ] .
for this case , we next focus on studying the conditions under which d - tdma is optimal across the fading states .
this can be done by investigating the kkt optimality conditions for problem [ prob : lt st mac per state ] .
first , we introduce nonnegative dual variables @xmath46 , @xmath19 , and @xmath63 , for their associated constraints in ( [ eq : constraint 1 lt st ] ) and ( [ eq : constraint 2 lt st ] ) , respectively .
the kkt conditions for the optimal primal and dual solutions of this problem , denoted as @xmath65 , @xmath54 , and @xmath66 , can then be expressed as @xmath95 with @xmath96 , @xmath69 , and @xmath97 .
notice that in this case @xmath46 s are local variables for each fading state instead of being fixed as in ( [ eq : kkt1 lt lt ] ) for problem [ prob : lt lt mac per state ] . from these kkt conditions , the following lemma
can then be obtained : [ lemma : tdma opt lt st ] the optimal solution of problem [ prob : lt st mac per state ] has at most @xmath2 secondary users that transmit with strictly positive power levels
. please refer to appendix [ appendix : proof tdma opt lt st ] .
lemma [ lemma : tdma opt lt st ] suggests that the optimal number of sus that can transmit at each fading state may depend on the number of prs or interference - power constraints . for small values of @xmath1 ,
e.g. , @xmath3 corresponding to a single pr , the number of active sus at each fading state can be at most two , suggesting that d - tdma may be very close to being optimal in this case . in the theorem
below , we present the general conditions , for any @xmath0 and @xmath1 , under which d - tdma is both necessary and sufficient to be optimal at a particular fading state . again , without loss of generality , here we use @xmath45 s instead of their optimal dual solutions obtained by the ellipsoid method . [
theorem : tdma opt conditions lt st ] d - tdma is optimal at an arbitrary fading state for achieving the ergodic sum capacity of the fading c - mac under the lt - tpc jointly with the st - ipc if and only if there exists one user @xmath70 ( the user that transmits ) that satisfies either one of the following two sets of conditions .
let @xmath74 be any of the other users , @xmath98 ; and @xmath99 .
* @xmath100 and @xmath101 . in this case ,
@xmath102 ; * @xmath103 and @xmath104 . in this case , @xmath105
. please refer to appendix [ appendix : proof tdma opt conditions lt st ] .
notice that in theorem [ theorem : tdma opt conditions lt st ] , the first set of conditions holds when the optimal transmit power of the user with the largest @xmath83 among all the users satisfies the st - ipc at all the prs ; the second set of conditions holds when the first set fails to be true , and in this case any of @xmath0 sus can be the selected user for transmission provided that it satisfies the given @xmath106 inequalities .
in the special case where only the st - ipc given by ( [ eq : st ipc mac ] ) is present or active in problem [ prob : lt st mac ] , all @xmath45 s in theorem [ theorem : tdma opt conditions lt st ] can be taken as zeros .
as a result , the first set of conditions can never be true , while the second set of conditions are simplified as @xmath107 , and the optimal power of user @xmath70 that transmits is still @xmath108 .
we thus have the following corollary if it is further assumed that there is only a single pr . for conciseness
, the index @xmath6 for this pr is dropped below .
[ corol : lt st ] in the case that only the st - ipc given by ( [ eq : st ipc mac ] ) is present in problem [ prob : lt st mac ] and , furthermore , @xmath3 , d - tdma is optimal ; and the selected user @xmath70 for transmission satisfies that @xmath109 , with transmit power @xmath110 . in the case of st - tpc combined with lt - ipc , the ergodic sum capacity is the optimal value of the following optimization problem : [ prob : st lt mac ] @xmath84 \\ \mathtt { s.t . } & & ( \ref{eq : st tpc mac } ) , ( \ref{eq : lt ipc mac}).\end{aligned}\ ] ] again , we apply the lagrange duality method for the above problem .
let @xmath46 s be the nonnegative dual variables associated with the lt - ipc in ( [ eq : lt ipc mac ] ) , @xmath19 .
the lagrangian of problem [ prob : st lt mac ] can then be written as @xmath111\nonumber \\ - \sum_{m=1}^m\mu_m\left\{\mathbb{e}\left[\sum_{k=1}^kg_{km}p_k({\mbox{\boldmath{$ \alpha $ } } } ) \right]- \gamma^{\rm lt}_m\right\}.\end{aligned}\ ] ] let @xmath112 denote the set of @xmath87 specified by the remaining st - tpc in ( [ eq : st tpc mac ] ) .
the lagrange dual function is expressed as @xmath113 the dual problem is accordingly defined as @xmath114 .
similar to the previous two cases , this dual function can be equivalently written as @xmath115 + \sum_{m=1}^m\mu_k\gamma^{\rm lt}_m\end{aligned}\ ] ] where @xmath116 with @xmath117 denoting the subset of @xmath112 corresponding to the fading state with channel realization @xmath12 .
after dropping @xmath12 in the maximization problem in ( [ eq : dual function per state st lt ] ) , for each particular fading state we can express this problem as [ prob : st lt mac per state ] @xmath118 after solving problem [ prob : st lt mac per state ] for all the fading states , we obtain the dual function @xmath119 .
the dual problem that minimizes @xmath119 over @xmath46 s can then be solved again via the ellipsoid method .
next , we present the closed - form solution of problem [ prob : st lt mac per state ] based on its kkt optimality conditions .
let @xmath45 and @xmath120 , @xmath16 , be the dual variables for the corresponding user power constraints in ( [ eq : constraint 1 st lt ] ) and ( [ eq : constraint 2 st lt ] ) , respectively .
the kkt conditions for the optimal primal and dual solutions of this problem , denoted as @xmath65 , @xmath53 , and @xmath66 , can then be expressed as @xmath121 with @xmath68 , @xmath122 , and @xmath123 . from these kkt conditions ,
the following lemma can be first obtained : [ lemma : opt 1 st lt ] let @xmath70 and @xmath74 be any two arbitrary users , @xmath124 , with @xmath125 and @xmath75 in the optimal solution of problem [ prob : st lt mac per state ] .
then , it must be true that @xmath126
. please refer to appendix [ appendix : proof lemma 1 st lt ] .
let @xmath127 be a permutation over @xmath128 such that @xmath129 if @xmath130 .
supposing that there are @xmath131 users that can transmit with @xmath132 denoting this set of users , from lemma [ lemma : opt 1 st lt ] it is easy to verify that @xmath133 .
the following lemma then provides the closed - form solution to problem [ prob : st lt mac per state ] : [ lemma : opt 2 st lt ] the optimal solution of problem [ prob : st lt mac per state ] is @xmath134 where @xmath131 is the largest value of @xmath135 such that @xmath136
. please refer to appendix [ appendix : proof lemma 2 st lt ] . from lemma [ lemma : opt 2 st lt ]
, it follows that in the case of st - tpc along with lt - ipc , for the active secondary users at one fading state , there is at most one user that transmits with power lower than its st power constraint , while all the other active users transmit with their maximum powers .
furthermore , from lemma [ lemma : opt 2 st lt ] , we can derive the conditions for the optimality of d - tdma at any fading state , which are stated in the following theorem .
again , without loss of generality , we use @xmath46 s instead of their optimal dual solutions for problem [ prob : st lt mac ] in expressing these conditions . [
theorem : tdma opt conditions st lt ] d - tdma is optimal at an arbitrary fading state for achieving the ergodic sum capacity of the fading c - mac under the st - tpc jointly with the lt - ipc if and only if user @xmath137 satisfies @xmath138 user @xmath137 is then selected for transmission and its optimal transmit power is @xmath139 from lemma [ lemma : opt 2 st lt ] , it follows that d - tdma is optimal , i.e. , @xmath140 , occurs if and only if ( [ eq : tdma opt st lt ] ) holds . then
, ( [ eq : tdma opt power st lt ] ) is obtained from lemma [ lemma : opt 2 st lt ] by combining the cases of @xmath141 and @xmath142 . in the case of the traditional fading siso - mac with the user st - tpc
given in ( [ eq : st tpc mac ] ) , but without the lt - ipc given in ( [ eq : lt ipc mac ] ) , it can be easily verified that the ergodic sum capacity is achieved when all users transmit with their maximum available power values given by @xmath143 s at each fading state .
this is consistent with the results obtained in ( [ eq : opt power st lt ] ) by having all @xmath46 s associated with the lt - ipc take zero values . with zero @xmath46 s
, it can be easily verified that the condition given in theorem [ theorem : tdma opt conditions st lt ] is never satisfied , and thus d - tdma can not be optimal in this special case . the ergodic sum capacity under both the st - tpc and st - ipc can be obtained by solving the following optimization problem : [ prob : st st mac ] @xmath84 \\ \mathtt { s.t . } & & ( \ref{eq : st tpc mac } ) , ( \ref{eq : st ipc mac}).\end{aligned}\ ] ] notice that this case differs from all three previous cases in that all of its power constraints are short - term constraints and thus separable over fading states .
therefore , we can decompose the original problem into individual subproblems each for one fading state .
for conciseness , we drop again the @xmath12 and express the rate maximization problem at a particular fading state as [ prob : st st mac per state ] @xmath144 the above problem is convex , but in general does not have a closed - form solution .
similar to problem [ prob : lt st mac per state ] , the interior point method @xcite or the lagrange duality method can be used to solve this problem and thus we omit the details here . for this case ,
we next present in the following theorem the conditions for d - tdma to be optimal at an arbitrary fading state : [ theorem : tdma opt conditions st st ] d - tdma is optimal at an arbitrary fading state for achieving the ergodic sum capacity of the fading c - mac under the st - tpc jointly with the st - ipc if and only if there exists one user @xmath70 ( the user that transmits ) that satisfies both of the following two conditions .
let @xmath74 be any of the other users , @xmath98 , and @xmath145 .
* @xmath146 ; * @xmath147 .
the optimal transmit power of user @xmath70 is @xmath148
. please refer to appendix [ appendix : proof tdma opt conditions st st ] .
from ( [ eq : sum capacity bc new ] ) , the ergodic sum capacities for the siso fading c - bc under different mixed tpc and ipc constraints can be obtained as the optimal values of the following optimization problems : [ prob : bc ] @xmath149 \nonumber \\ \mathtt { s.t . } & ~ ( \ref{eq : lt tpc bc } ) , ( \ref{eq : lt ipc bc } ) \ ( { \rm case \ i : \ lt - tpc \ and \
lt - ipc } ) \nonumber \\ & ~ { \rm or } ~ ( \ref{eq : lt tpc bc } ) , ( \ref{eq : st ipc bc } ) \
( { \rm case \ ii : lt - tpc \ and \ st - ipc } ) \nonumber \\ & ~ { \rm or } ~ ( \ref{eq : st tpc bc } ) , ( \ref{eq : lt ipc bc } ) \ ( { \rm case \ iii : \ st - tpc \ and \ lt - ipc } ) \nonumber \\ & ~ { \rm or } ~ ( \ref{eq : st tpc bc } ) , ( \ref{eq : st ipc bc } ) \ ( { \rm case \ iv : \ st - tpc \ and \ st - ipc } ) . \nonumber\end{aligned}\ ] ] notice that in ( [ eq : lt tpc bc])-([eq : st ipc bc ] ) , the transmit power of the secondary bs at a given fading state , @xmath30 , needs to be replaced by the user sum - power in the dual c - mac , @xmath40 . compared with the problems addressed in section [ sec : mac ] for the c - mac ,
it is easy to see that the corresponding problems in the c - bc case are very similar , e.g. , both have the same objective function , and similar affine constraints in terms of @xmath15 s or @xmath150 s .
thus , we skip the details of derivations and present the results directly in the following theorem : [ theorem : tdma opt conditions bc ] in each of cases i - iv , d - tdma is optimal across all the fading states for achieving the ergodic sum capacity of the fading c - bc . in each case , the user @xmath70 with the largest @xmath151 among all the users should be selected for transmission at a particular fading state . the optimal rule for assigning the transmit power of the bs at each fading state ( for conciseness @xmath29
is dropped in the following expressions ) in each case is given below .
let @xmath74 be any of the users other than @xmath70 , @xmath98 ; @xmath152 ; and @xmath153 and @xmath46 s are the optimal dual variables associated with the lt - tpc in ( [ eq : lt tpc bc ] ) and the lt - ipc in ( [ eq : lt ipc bc ] ) , respectively , if they appear in any of the following cases .
* case i : @xmath154 * case ii : @xmath155 * case iii : @xmath156 * case iv : @xmath157 in the case of the traditional fading siso - bc without the lt- or st - ipc , by combining the results in @xcite for the fading siso - mac and the mac - bc duality results in @xcite , it can be inferred that it is optimal to deploy d - tdma by transmitting to the user with the largest @xmath151 at each time in terms of maximizing the ergodic sum capacity , regardless of the lt- or st - tpc at the bs .
theorem [ theorem : tdma opt conditions bc ] can thus be considered as the extensions of such result to the siso fading c - bc under the additional lt- or st - ipc .
also notice that the optimal power allocation strategies in ( [ eq : bc power opt 1])-([eq : bc power opt 3 ] ) resemble the well - known `` water - filling ( wf ) '' solutions for the single - user fading channels @xcite , @xcite .
in this section , we present numerical results on the performances of the proposed multiuser dra schemes for some example fading cr networks under different mixed transmit - power and interference - power constraints , namely : case i : lt - tpc with ( w/ ) lt - ipc ; case ii : lt - tpc w/ st - ipc ; case iii : st - tpc w/ lt - ipc ; and case iv : st - tpc w/ st - ipc . for simplicity
, we consider symmetric multiuser channels where all channel complex coefficients are independent cscg random variables distributed as @xmath11 . in total , @xmath158 randomly generated channel power gain vectors for @xmath12 or @xmath29 are used to approximate the actual ergodic sum - rate of the secondary network in each simulation result .
furthermore , we assume that the tpc ( lt or st ) values are identical for all sus , and the ipc ( lt or st ) values are identically equal to one , the same as the additive gaussian noise variance , at all prs . for convenience ,
we use @xmath159 to stand for all @xmath160 s and @xmath161 s , @xmath162 for both @xmath163 and @xmath164 , and @xmath165 for all @xmath166 s and @xmath167 s .
the simulation results are presented in the following subsections .
first , we compare the achievable ergodic sum capacities for the fading cr network under four different cases of mixed tpc and ipc .
[ fig : mac ] shows the results for the fading c - mac with @xmath168 and @xmath3 , and fig .
[ fig : bc ] for the fading c - bc with @xmath169 and @xmath170 . for the c - mac case
, it is observed in fig .
[ fig : mac ] that the ergodic sum capacity @xmath41 in case i is always the largest while that in case iv is the smallest for any given su transmit power constraint @xmath159 .
this is as expected since both the st - tpc and st - ipc are less favorable from the su s perspective as compared to their lt counterparts : the former one imposes more stringent power constraints than the latter one over the dra in the su network .
it is also observed that as @xmath159 increases , eventually @xmath41 becomes saturated as the ipc ( lt or st ) becomes more dominant than the tpc . on the other hand , for small values of @xmath159 where the tpc is more dominant than the ipc , it is observed that the lt - tpc ( where d - tdma is optimal in case i and close to being optimal in case ii ) leads to a capacity gain over the st - tpc ( where d - tdma is non - optimal in case iii or iv ) due to the well - known _ multiuser diversity _ effect exploited by d - tdma @xcite .
furthermore , @xmath41 in case ii is observed to be initially larger than that in case iii for small values of @xmath159 , but becomes equal to and eventually smaller than that in case iii as @xmath159 increases .
this is due to the facts that for small values of @xmath159 , tpc dominates ipc and furthermore lt - tpc is more flexible over st - tpc ; while for large values of @xmath159 , ipc becomes more dominant over tpc and lt - ipc is more flexible over st - ipc . for the c - bc case ,
similar results like those in the c - mac are observed .
however , there exists one quite different phenomenon for the c - bc .
as the secondary bs transmit power @xmath162 becomes large , the achievable ergodic sum capacity @xmath42 shown in fig .
[ fig : bc ] under the lt - ipc is much larger than that under the st - ipc , regardless of the lt- or st - tpc , as compared with @xmath171 shown in fig .
[ fig : mac ] .
this is due to the fact that for the c - bc with @xmath170 and a single bs transmitter , the st - ipc can limit the transmit power of the secondary bs more stringently than the case of c - mac shown in fig .
[ fig : mac ] , where there are two su transmitters but only a single pr .
since it is not always the case that both channels from the two sus to the pr have very large gains at a given time , in the c - mac case the su with the smallest instantaneous channel gain to the pr can be selected for transmission , i.e. , there exists an interesting _ new form of multiuser diversity effect _ in the fading c - mac .
in contrast , for the c - bc , the bs is likely to transmit with large power only if both channel gains from the bs to the two prs are reasonably low .
next , we consider the fading c - mac and examine the effect of the tdma constraint on its achievable ergodic sum capacity .
notice that for the fading c - bc , it has been shown in theorem [ theorem : tdma opt conditions bc ] that d - tdma is optimal for all cases of mixed tpc and ipc ; and for the fading c - mac , it has also been shown in theorem [ theorem : tdma opt lt lt ] that d - tdma is optimal in case i. therefore , in this subsection , we only consider the fading c - mac in cases ii , iii , and iv .
we compare the ergodic sum capacity @xmath171 achievable in each of these cases via the optimal dra rule proposed in this paper w / o the tdma constraint against that with an explicit tdma constraint , i.e. , at most one su is selected for transmission at any time .
however , for the cases with the explicit tdma constraint , we still allow dra over the su network to optimally select the su ( i.e. , using d - tdma ) and set its power level for transmission at each fading state , so as to maximize the long - term average sum - rate . for conciseness
, we discuss the optimal dra schemes for the fading c - mac under the explicit tdma constraint in appendix [ appendix : tdma ] . in figs .
[ fig : mac tdma 2 ] and [ fig : mac tdma ] , we compare the achievable @xmath41 w/ vs. w / o the tdma constraint for cases ii - iv with @xmath168 , @xmath3 , and @xmath172 , respectively .
it is observed in both figures that the achievable @xmath41 in each case of mixed tpc and ipc is larger without the tdma constraint .
this is as expected since tdma is an additional constraint that limits the flexibility of dra in the su network . in fig .
[ fig : mac tdma 2 ] , it is observed that the gap between the achievable @xmath41 s w/ and w / o the tdma constraint in each of cases ii - iv diminishes as the su transmit power constraint @xmath159 becomes sufficiently large . this phenomenon can be explained as follows .
first , note that as @xmath159 increases , eventually the tpc will become inactive and the ipc becomes the only active power constraint in each case . as a result ,
case ii and case iv only have the ( same ) st - ipc and case iii only has the lt - ipc as active constraints .
thus , the observed phenomenon is justified since d - tdma has been shown to be optimal for the above two cases , according to corollary [ corol : lt st ] ( notice that @xmath3 for fig .
[ fig : mac tdma 2 ] ) and theorem [ theorem : tdma opt lt lt ] ( with all @xmath45 s taking a zero value ) , respectively .
however , in fig .
[ fig : mac tdma ] with @xmath173 , only case iii has the same converged @xmath41 w/ and w / o the tdma constraint as @xmath159 becomes large , according to theorem [ theorem : tdma opt lt lt ] .
in general , the capacity gap between cases w/ and w / o the tdma constraint becomes larger as @xmath0 or @xmath1 increases , as observed by comparing figs .
[ fig : mac tdma 2 ] and [ fig : mac tdma ] . for example ,
for case ii , in fig . [
fig : mac tdma 2 ] with @xmath3 , the capacity gap is negligible for all values of @xmath159 , which is consistent with lemma [ lemma : tdma opt lt st ] ; but it becomes notably large in fig .
[ fig : mac tdma ] with @xmath170 . at last , we compare the ergodic sum capacity achievable with the optimal dra against the achievable average sum - rate of users via some heuristic fixed resource allocation ( fra ) schemes for the same fading cr network .
for dra , we select the most flexible power allocation scheme for the su network under the lt - tpc and the lt - ipc ( i.e. , case i ) , which is d - tdma based and gives the largest @xmath171 and @xmath42 among all cases of mixed power constraints under the same power - constraint values @xmath159 ( @xmath162 ) and @xmath165 for the fading c - mac ( c - bc ) . for fra ,
we also consider tdma , which uses the simple `` round - robin '' user scheduling rule , under the st - tpc and the st - ipc .
more specifically , for the fading c - mac , at each time the su , say user @xmath70 , which is scheduled for transmission , will transmit a power equal to @xmath174 , while for the fading c - bc , the bs transmits with the power equal to @xmath175 .
notice that the considered fra can be much more easily implemented as compared to the proposed optimal dra .
therefore , we need to examine the capacity gains by the optimal dra over the fra . in fig .
[ fig : mac comp ] , capacity comparisons between dra and fra are shown for the fading c - mac with @xmath168 or @xmath176 , and @xmath170 . notice that for the dra case we have normalized the su lt - tpc for @xmath177 by a factor of @xmath178 such that the sum of user transmit power constraints for both @xmath168 and @xmath177 are identical .
furthermore , for fair comparison between dra and fra , the su st - tpc values in the fra case are @xmath176 and @xmath178 times the lt - tpc value in the dra for @xmath177 and @xmath168 , respectively .
it is observed that dra achieves substantial throughput gains over fra for both @xmath168 and @xmath177 .
notice that for fra , it can be easily shown that with the user power normalization , the average sum - rate is statistically independent of @xmath0 .
furthermore , multiuser diversity gains in the achievable ergodic sum - rate for the dra are also observed by comparing @xmath177 against @xmath168 , given the same sum of user power constraints . in fig .
[ fig : bc comp ] , we show the capacity comparisons between the fading c - bc with dra and that with fra , for a fixed secondary bs transmit power constraint @xmath179db , @xmath3 or @xmath176 , and different values of @xmath0 .
since there is only one transmitter at the bs for the c - bc , there is no user power normalization required as in the c - mac case .
the capacity gains by dra over fra are observed to become more significant for both @xmath3 and @xmath180 cases , as @xmath0 increases , due to the multiuser diversity effect . as an example , at @xmath181 , the capacities with dra are @xmath182 and @xmath183 times of that with fra , for @xmath3 and @xmath180 , respectively .
this suggests that in contrast to the conventional fading bc without any ipc , the multiuser diversity gains obtained by the optimal dra become more crucial to the fading c - bc as the number of prs , @xmath1 , becomes larger .
in this paper , we have studied the information - theoretic limits of the cr network under wireless spectrum sharing with an existing primary radio network . by applying the interference - power constraint as a practical means to protect each primary link ,
we characterize the achievable ergodic sum capacity of the fading c - mac and c - bc under different mixed lt-/st - tpc and lt-/st - ipc .
optimal dra schemes for both cases w/ and w / o a tdma constraint are presented .
interestingly , except the cases where the optimality of d - tdma can be analytically proved , it is verified by simulation that there are also many circumstances where d - tdma with the optimal user scheduling and power control performs very closely to the optimal non - tdma - based schemes in the fading c - mac .
furthermore , an interesting new form of multiuser diversity is observed for the fading c - mac by exploiting the additional csi of channels between secondary transmitters and primary receivers , which differs from that in the conventional fading mac by exploiting only the csi of channels between secondary users and bs .
finally , it is worth pointing out that with the techniques introduced in this paper , it is possible to derive the optimal resource allocation for the more general cases where all lt / st tpc and ipc are present , and/or secondary users have different priorities for rate allocation ( i.e. , characterization of the capacity region instead of the sum capacity ) .
moreover , the results in this paper are also applicable to the general channel models consisting of parallel gaussian channels over which the average and instantaneous ( transmit or interference ) power constraints can be applied , e.g. , the frequency - selective fading broadband channel which is decomposable into parallel narrow - band channels at each fading state via the well - known orthogonal - frequency - division - multiplexing ( ofdm ) modulation / demodulation .
suppose that there are two arbitrary users @xmath70 and @xmath74 with @xmath125 and @xmath184 . from ( [ eq : kkt2 lt lt ] )
, it follows that @xmath185 and @xmath186 .
applying this fact to ( [ eq : kkt1 lt lt ] ) , the following equality must hold : @xmath187 since @xmath151 and @xmath188 s are independent of @xmath189 and @xmath190 s , and furthermore @xmath191 , @xmath192 , and @xmath46 s are all constants in problem [ prob : lt lt mac per state ] , it can be inferred that the above equality is satisfied with a zero probability .
thus , it is concluded that there is at most one user with a strictly positive power value .
let user @xmath70 be the user that can transmit , i.e , @xmath73 , while for the other users @xmath193 , @xmath75 .
problem [ prob : lt lt mac per state ] then becomes the maximization of @xmath194 subject to @xmath195 , for which @xmath196 given in ( [ eq : opt power lt lt ] ) can be easily shown to be the optimal solution .
next , we need to show that for the selected user @xmath70 for transmission , if @xmath125 , it must satisfy ( [ eq : user ordering lt lt ] ) . since @xmath125 , from ( [ eq : kkt2 lt lt ] ) it follows that @xmath185 . since @xmath197 , from ( [ eq : kkt1 lt lt ] ) , it follows that @xmath198 from which ( [ eq : user ordering lt lt ] ) can be obtained .
suppose that there are @xmath199 users with @xmath184 , where @xmath200 and @xmath201 .
then from ( [ eq : kkt3 lt st ] ) , it follows that @xmath186 , if @xmath200 .
. from ( [ eq : kkt1 lt st ] ) , the following equalities must hold : @xmath203 removing @xmath204 in the above equations yields @xmath205 where @xmath70 is an arbitrary user index in @xmath206 . notice that in ( [ eq : lt st equalities ] ) there are @xmath1 variables @xmath207 , , @xmath208 , but @xmath209 independent equations ( with probability one ) .
therefore , @xmath210 must hold in order for the above equations to have at least one set of solutions .
it then concludes that @xmath199 must be no greater than @xmath2 . since @xmath75 , @xmath125 , from ( [ eq : kkt2 st lt ] ) and ( [ eq : kkt3 st lt ] ) it follows that @xmath234 and @xmath185 , respectively .
then , from ( [ eq : kkt1 st lt ] ) it follows that @xmath235 from the above two inequalities , lemma [ lemma : opt 1 st lt ] can be easily shown .
the following lemma is required for the proof of lemma [ lemma : opt 2 st lt ] : [ lemma : opt 3 st lt ] the optimal solution of problem [ prob : st lt mac per state ] has at most one user , indexed by @xmath70 , which satisfies @xmath236 , where @xmath237 ; and the optimal sum - power of transmitting users must satisfy @xmath238 .
suppose that there are two users @xmath70 and @xmath74 with @xmath236 and @xmath239 . from ( [ eq : kkt2 st lt ] ) and ( [ eq : kkt3 st lt ] )
, it follows that @xmath240 and @xmath241 , respectively . using these facts , from ( [ eq : kkt1 st lt ] )
, it follows that the following two equalities must hold at the same time : @xmath242 thus , we have @xmath243 since @xmath151 and @xmath188 s are independent of @xmath189 and @xmath190 s , and @xmath46 s are constants , it is inferred that the above equality is satisfied with a zero probability .
thus , we conclude that there is at most one user @xmath70 with @xmath244 . from ( [ eq : sum power st lt ] )
, we have @xmath245 using ( [ eq : kkt1 st lt ] ) and ( [ eq : equality st lt ] ) , it is easy to see that for any user @xmath246 with @xmath247 , it must satisfy @xmath248 thus , we conclude that @xmath237 .
lemma [ lemma : opt 3 st lt ] suggests that only one of the following two sets of solutions for @xmath249 can be true , which are * case i : @xmath250 ; * case ii : @xmath251 , and @xmath252 . since @xmath253
, it then follows that @xmath254 the remaining part to be shown for lemma [ lemma : opt 2 st lt ] is that the optimal number of active users @xmath131 is the largest value of @xmath135 such that @xmath255 first , we show that in both case i and case ii , for any user @xmath256 , the above inequality holds .
since for ( [ eq : inequality st lt ] ) , from lemma [ lemma : opt 1 st lt ] it follows that its left - hand side decreases as @xmath135 increases , while its right - hand side increases with @xmath135 , it is sufficient to show that ( [ eq : inequality st lt ] ) holds for @xmath257 .
this is the case since from ( [ eq : kkt1 st lt ] ) with @xmath258 and @xmath259 , we have @xmath260 next , we show that for any user @xmath261 , ( [ eq : inequality st lt ] ) does not hold .
again , it is sufficient to consider user @xmath262 since if it does not satisfy ( [ eq : inequality st lt ] ) , neither does any of the other users @xmath263 . for user @xmath262 , from ( [ eq : kkt1 st lt ] ) with @xmath264 and @xmath265 , it follows that @xmath266 therefore , it is concluded that ( [ eq : inequality st lt ] ) can be used to determine @xmath131 .
suppose that user @xmath70 transmits with @xmath125 , while for the other users @xmath98 , @xmath75 .
we will consider the following two cases : i ) all @xmath81 s are equal to zero ; ii ) there is one and only one @xmath211 , denoted as @xmath212 , which is strictly positive .
notice that it is impossible for more than one @xmath81 s to be strictly positive at the same time , which can be shown as follows . for user @xmath70 , from ( [ eq : kkt2 lt st ] ) , @xmath213 suggests that @xmath214 .
supposing that there is @xmath215 such that @xmath216 and thus @xmath217 , a contradiction then occurs as @xmath218 holds with a zero probability .
first , we will prove the `` only if '' part of theorem [ theorem : tdma opt conditions lt st ] .
consider initially the case where all @xmath81 s are equal to zero .
suppose that @xmath125 , from ( [ eq : kkt3 lt st ] ) it follows that @xmath185 .
since @xmath219 , from ( [ eq : kkt1 lt st ] ) the followings must be true : @xmath220 thus , user @xmath70 must satisfy @xmath221 .
from ( [ eq : equality lt st ] ) , it follows that @xmath102 in this case .
also notice that from ( [ eq : kkt4 lt st ] ) @xmath222 must hold for @xmath223 .
therefore , we conclude that @xmath224 , where @xmath99 , and thus @xmath225 . therefore , the first set of conditions in theorem [ theorem : tdma opt conditions lt st ] is obtained . in the second case where there is one and only one @xmath213
, it follows from ( [ eq : kkt2 lt st ] ) that @xmath214 .
since from ( [ eq : kkt4 lt st ] ) we have @xmath226 , it follows that @xmath227 , and thus , again , @xmath99 , and @xmath108 in this case . from ( [ eq : kkt1 lt st ] ) , we have @xmath228 since @xmath213 , from ( [ eq : opt mu lt st ] ) it follows that @xmath229 . furthermore ,
from ( [ eq : kkt1 lt st ] ) , the followings must be true : @xmath230 thus , we have @xmath231 substituting @xmath212 in ( [ eq : opt mu lt st ] ) into the above inequalities yields @xmath232 @xmath233 .
the second set of conditions in theorem [ theorem : tdma opt conditions lt st ] is thus obtained .
next , the `` if '' part of theorem [ theorem : tdma opt conditions lt st ] can be shown easily by the fact that for a strictly - convex optimization problem , the kkt conditions are not only necessary but also sufficient to be satisfied by the unique set of primal and dual optimal solutions @xcite .
the proof of theorem [ theorem : tdma opt conditions st st ] is also based on the kkt optimality conditions for problem [ prob : st st mac per state ] .
let @xmath80 , @xmath81 , and @xmath267 , @xmath268 be the optimal dual variables associated with the constraints in ( [ eq : constraint 1 st st ] ) , ( [ eq : constraint 2 st st ] ) , and ( [ eq : constraint 3 st st ] ) , respectively .
the kkt conditions can then be expressed as @xmath269 with @xmath270 , and @xmath271 .
first , we will prove the `` only if '' part of theorem [ theorem : tdma opt conditions st st ] .
suppose that user @xmath70 should transmit with @xmath125 , while for the other users @xmath272 , @xmath75 . from ( [ eq : kkt2 st st ] ) and ( [ eq : kkt4 st st ] )
, it follows that @xmath273 and @xmath185 , respectively .
we will show that there is one and only one @xmath211 , denoted as @xmath212 , which is strictly positive .
notice that it is impossible for more than one @xmath81 s to be strictly positive at the same time . for user @xmath70 , from ( [ eq : kkt3
st st ] ) , @xmath213 suggests that @xmath214 .
supposing that there is @xmath215 such that @xmath216 and thus @xmath217 , a contradiction then occurs as @xmath218 holds with a zero probability .
second , we will show that it is also impossible for all @xmath81 s to be zero .
if this is the case , ( [ eq : kkt1 st st ] ) for any user @xmath193 , becomes @xmath274 .
this can be true only when @xmath275 , which occurs with a zero probability .
therefore , we conclude that there is one and only one @xmath213 . since @xmath214 and from ( [ eq : kkt6 st st ] ) we have @xmath226 , it follows that @xmath227 , and thus @xmath99 and @xmath108 . also notice from ( [ eq : kkt5 st st ] ) that in this case @xmath276 must hold . at last , considering ( [ eq : kkt1
st st ] ) for user @xmath70 and any other user @xmath74 , we have @xmath277 thus , we conclude that @xmath278 , must hold .
next , the `` if '' part of theorem [ theorem : tdma opt conditions st st ] follows due to the fact that for a strictly - convex optimization problem , the kkt conditions are both necessary and sufficient for the unique set of primal and dual optimal solutions @xcite .
in this appendix , we formally derive the optimal rule of user selection and power control to achieve the ergodic sum capacity for the siso fading c - mac _ under an explicit tdma constraint _
, in addition to any combination of transmit - power and interference - power constraints .
the tdma constraint implies that at each fading state there is only one su that can transmit .
let @xmath279 be a mapping function that gives the index of the su selected for transmission at a fading state with channel realization @xmath12 .
note that for this particular fading state , @xmath280 , while for the other sus @xmath281 , @xmath282 , @xmath283 .
the ergodic sum capacity of the fading c - mac under tdma constraint can be obtained as @xmath284\ ] ] where @xmath24 is specified by a particular combination of power constraints described in ( [ eq : lt tpc mac])-([eq : st ipc mac ] ) .
clearly , for any given function @xmath279 , the capacity maximization in ( [ eq : sum capacity mac tdma ] ) over @xmath24 is a convex optimization problem .
however , the maximization over the function @xmath279 may not be necessarily convex , and thus standard convex optimization techniques may not apply directly .
fortunately , it will be shown next that the optimization problem in ( [ eq : sum capacity mac tdma ] ) can be efficiently solved for all considered cases of mixed lt-/st - tpc and lt-/st - ipc . from ( [ eq : sum capacity mac tdma ] ) , the ergodic sum capacity under the tdma constraint , as well as the lt - tpc in ( [ eq : lt tpc mac ] ) and the lt - ipc in ( [ eq : lt ipc mac ] ) can be obtained by solving the following optimization problem : [ prob : lt lt mac tdma ] @xmath285 \nonumber \\ \mathtt { s.t . } & ~~ \mathbb{e}\left[p_k({\mbox{\boldmath{$ \alpha $ } } } ) \cdot{\mbox{\boldmath{$ 1 $ } } } ( \pi({\mbox{\boldmath{$ \alpha $ } } } ) = k)\right]\leq p^{\rm lt}_k , \ \forall k \label{eq : lt tpc mac tdma } \\ & ~~ \mathbb{e}\left[g_{\pi({\mbox{\boldmath{$ \alpha $ } } } ) m}p_{\pi({\mbox{\boldmath{$ \alpha $ } } } ) } ( { \mbox{\boldmath{$ \alpha $ } } } ) \right]\leq \gamma^{\rm lt}_m , \
\forall m \label{eq : lt ipc mac tdma}\end{aligned}\ ] ] where @xmath286 is the indicator function taking the values of 1 or 0 depending on the trueness or falseness of event @xmath287 , respectively .
@xmath288 -\sum_{k=1}^k\lambda_k\left\ { \mathbb{e}\left[p_k({\mbox{\boldmath{$ \alpha $ } } } ) \cdot{\mbox{\boldmath{$ 1 $ } } } ( \pi({\mbox{\boldmath{$ \alpha $ } } } ) = k)\right]- p^{\rm lt}_k\right\ } -\sum_{m=1}^m\mu_m\left\{\mathbb{e}\left[g_{\pi({\mbox{\boldmath{$ \alpha $ } } } ) m}p_{\pi({\mbox{\boldmath{$ \alpha $ } } } ) } ( { \mbox{\boldmath{$ \alpha $ } } } ) \right]- \gamma^{\rm lt}_m\right\}\end{aligned}\ ] ] first , we write the lagrangian of this problem , @xmath289 , as in ( [ eq : lagrangian lt lt tdma ] ) ( shown on the next page ) , where @xmath45 and @xmath46 are the nonnegative dual variables associated with the corresponding constraints in ( [ eq : lt tpc mac tdma ] ) and ( [ eq : lt ipc mac tdma ] ) , respectively , @xmath16 , @xmath19 .
then , the lagrange dual function , @xmath47 , is defined as @xmath290 the dual problem is accordingly defined as @xmath291 . since the problem at hand may not be convex , the duality gap between the optimal values of the original and the dual problems may not be zero .
however , it will be shown in the later part of this subsection that the duality gap for problem [ prob : lt lt mac tdma ] is indeed zero .
we consider only the maximization problem in ( [ eq : lagrange dual lt lt tdma ] ) for obtaining @xmath47 with some given @xmath45 s and @xmath46 s , while the minimization of @xmath47 over @xmath45 s and @xmath46 s can be obtained by the ellipsoid method , since it is always a convex optimization problem . for each fading state ,
the maximization problem in ( [ eq : lagrange dual lt lt tdma ] ) can be expressed as ( with @xmath12 dropped for brevity ) for any given user @xmath293 , the optimal power solution for the above problem can be obtained as @xmath294 substituting this solution into the objective function of problem [ prob : lt lt mac per state tdma ] yields @xmath295 it is easy to verify that the maximization of the above function over @xmath293 is attained with user @xmath70 that satisfies @xmath296 from ( [ eq : opt power lt lt tdma ] ) and ( [ eq : user ordering lt lt tdma ] ) , it follows that the same set of solutions for problem [ prob : lt lt mac per state ] without the tdma constraint , which is given in lemma [ lemma : opt power user lt lt ] , also holds for problem [ prob : lt lt mac per state tdma ] with the tdma constraint .
note that the optimal solutions of problem [ prob : lt lt mac ] without the tdma constraint are also tdma - based , and thus they are also feasible solutions to problem [ prob : lt lt mac tdma ] with the tdma constraint .
since these solutions have also been shown in the above to be optimal for the dual problem of problem [ prob : lt lt mac tdma ] , we conclude that the duality gap is zero for problem [ prob : lt lt mac tdma ] ; and both problem [ prob : lt lt mac ] and problem [ prob : lt lt mac tdma ] have the same set of solutions .
[ [ long - term - transmit - power - and - short - term - interference - power - constraints-1 ] ] long - term transmit - power and short - term interference - power constraints ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ [ prob : lt st mac tdma ] @xmath297 \nonumber \\ \mathtt { s.t . } & & ( \ref{eq : lt tpc mac tdma } ) \nonumber \\ & & g_{\pi({\mbox{\boldmath{$ \alpha $ } } } ) m}p_{\pi({\mbox{\boldmath{$ \alpha $ } } } ) } ( { \mbox{\boldmath{$ \alpha $ } } } ) \leq \gamma^{\rm st}_m , \
\forall { \mbox{\boldmath{$ \alpha $ } } } , m. \label{eq : st ipc mac tdma}\end{aligned}\ ] ] similarly as for problem [ prob :
lt lt mac tdma ] , we apply the lagrange duality method for solving the above problem by introducing the nonnegative dual variables @xmath298 , associated with the lt - tpc given in ( [ eq : lt tpc mac tdma ] ) .
however , since problem [ prob : lt st mac tdma ] is not necessarily convex , the duality gap for this problem may not be zero .
nevertheless , it can be verified that problem [ prob : lt st mac tdma ] satisfies the so - called `` time - sharing '' conditions @xcite and thus has a zero duality gap . for brevity , we skip the details of derivations here and present the optimal power - control policy in this case as follows : in the optimal solution of problem [ prob : lt st mac tdma ] , the user @xmath279 that transmits at a fading state with channel realization @xmath12 maximizes the following expression among all the users ( with @xmath12 dropped for brevity ) : @xmath299 where @xmath300 and @xmath298 , are the optimal dual solutions obtained by the ellipsoid method .
[ [ short - term - transmit - power - and - long - term - interference - power - constraints-1 ] ] short - term transmit - power and long - term interference - power constraints ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ [ prob : st lt mac tdma ] @xmath297 \nonumber \\ \mathtt { s.t . } & & p_{\pi({\mbox{\boldmath{$ \alpha $ } } } ) } ( { \mbox{\boldmath{$ \alpha $ } } } ) \leq p^{\rm st}_{\pi({\mbox{\boldmath{$ \alpha $ } } } ) } , \
\forall { \mbox{\boldmath{$ \alpha $ } } } \label{eq : st tpc mac tdma } \\ & & ( \ref{eq : lt ipc mac tdma } ) .
\nonumber\end{aligned}\ ] ] by introducing the nonnegative dual variables @xmath301 , associated with the lt - ipc given in ( [ eq : lt ipc mac tdma ] ) , problem [ prob : st lt mac tdma ] can be solved similarly as for problem [ prob : lt st mac tdma ] by the lagrange duality method .
for brevity , we present the optimal power - control policy in this case directly as follows : in the optimal solution of problem [ prob : st lt mac tdma ] , the user @xmath279 that transmits at a fading state with channel realization @xmath12 maximizes the following expression among all the users ( with @xmath12 dropped for brevity ) : @xmath302 where @xmath303 and @xmath301 , are the optimal dual solutions obtained by the ellipsoid method . in this case
, all the constraints are separable over the fading states and , thus , this problem is decomposable into independent subproblems each for one fading state . for brevity
, we present the optimal power - control policy in this case directly as follows : in the optimal solution of problem [ prob : st st mac tdma ] , the user @xmath279 that transmits at a fading state with channel realization @xmath12 maximizes the following expression among all the users ( with @xmath12 dropped for brevity ) : @xmath304 where @xmath305 c. cordeiro , k. challapali , d. birru , and n. s. shankar , `` ieee 802.22 : the first worldwide wireless standard based on cognitive radios , '' in _ proc .
ieee symp .
new frontiers in dynamic spectrum access networks ( dyspan ) _ , pp .
328 - 337 , nov . 2005 .
x. kang , y. c. liang , a. nallanathan , h. garg , and r. zhang , `` optimal power allocation for fading channels in cognitive radio networks : ergodic capacity and outage capacity , '' _ ieee trans .
wireless commun .
_ , vol . 8 , no .
940 - 950 , feb . 2009 .
y. xing , c. n. mathur , m. a. haleem , r. chandramouli , and k. p. subbalakshmi , `` dynamic spectrum access with qos and interference temperature constraints , '' _ ieee trans .
mobile computing _ , vol
423 - 433 , apr . 2007 .
d. tse and s. hanly , `` multi - access fading channels - part i : polymatroid structure , optimal resource allocation and throughput capacities , '' _ ieee trans .
inf . theory _ ,
44 , no . 7 , pp . 2796 - 2815 , nov .
1998 .
s. shamai and a. d. wyner , `` information theoretic considerations for symmetric , cellular , multiple access fading channels - part i , '' _ ieee trans .
inf . theory _ ,
43 , no . 6 , pp .
1877 - 1894 , nov . 1997 | this paper studies the information - theoretic limits of a secondary or cognitive radio ( cr ) network under spectrum sharing with an existing primary radio network . in particular , the fading cognitive multiple - access channel ( c - mac ) is first studied , where multiple secondary users transmit to the secondary base station ( bs ) under both individual transmit - power constraints and a set of interference - power constraints each applied at one of the primary receivers .
this paper considers the long - term ( lt ) or the short - term ( st ) transmit - power constraint over the fading states at each secondary transmitter , combined with the lt or st interference - power constraint at each primary receiver . in each case , the optimal power allocation scheme is derived for the secondary users to achieve the ergodic sum capacity of the fading c - mac , as well as the conditions for the optimality of the dynamic time - division - multiple - access ( d - tdma ) scheme in the secondary network . the fading cognitive broadcast channel ( c - bc ) that models the downlink transmission in the secondary network is then studied under the lt / st transmit - power constraint at the secondary bs jointly with the lt / st interference - power constraint at each of the primary receivers .
it is shown that d - tdma is indeed optimal for achieving the ergodic sum capacity of the fading c - bc for all combinations of transmit - power and interference - power constraints .
broadcast channel , cognitive radio , convex optimization , dynamic resource allocation , ergodic capacity , fading channel , interference temperature , multiple - access channel , spectrum sharing , time - division - multiple - access .
[ section ] [ section ] [ section ] [ section ] [ section ] [ section ] [ section ] |
it has been argued that strange quark matter , consisting of @xmath4- , @xmath5- and @xmath6-quarks , is energetically the most favorable state of quark matter @xcite .
@xcite suggested that there are two ways to form strange quark matter : the quark - hadron phase transition in the early universe and conversion of neutron stars into strange ones at ultrahigh densities .
if this strange matter hypothesis is correct , then it has profound impact on physics and astrophysics .
pulsars could be strange stars and not neutron stars as previously thought , and there could even be many strange dwarfs and strange planets in the universe @xcite .
several mechanisms have been proposed for the formation of strange quark stars .
for example , strange stars are expected to form during the collapse of the core of a massive star after a supernova explosion @xcite .
another possibility is that some rapidly spinning neutron stars in low - mass x - ray binaries ( lmxbs ) can accrete sufficient mass to initiate a phase transition to become strange stars @xcite .
some of the millisecond pulsars may be strange stars , because lmxbs are believed to be the progenitors of millisecond pulsars .
strange stars have also been proposed as sources of unusual astrophysical phenomena , such as soft @xmath7-ray repeaters @xcite , pulsating x - ray bursters @xcite , cosmological @xmath7-ray bursts @xcite , sax j1808.4 - 3658 @xcite etc .
the discovery of khz quasi - periodic oscillations in lmxbs @xcite implies that the compact stellar object must have a very soft equation of state ( eos ) , which is consistent with that of strange stars @xcite .
recently , hong et al .
@xcitehave argued that the first order phase transition of the color superconductivity occurring at the end point of some massive stars can release sufficient energy to produce hypernovae , which are considered as progenitors of cosmological @xmath0-ray bursts .
however , important as this strange - matter hypothesis maybe , it remains notoriously difficult to be proven or refuted on either observational or theoretical grounds .
the binding energy per nucleon for strange matter is estimated to be close to that of fe@xmath8 , but it has not been possible to calculate it from qcd to within the few percents level required to address the question .
model calculations of the binding energy are clearly inadequate and unreliable @xcite . lacking an accurate and reliable method to calculate the strange matter eos , the structure and stability of strange stars have been addressed @xcite typically in simplified models such as the mit bag model @xcite .
the quarks are treated as relativistic free particles confined in an impenetrable bag .
approximately then , the pressure comprises of the fermi gas pressure of the quarks minus a bag pressure , which mimics the strong interaction that holds the quarks together .
perturbative corrections to such a simple eos can be made based on one - gluon exchange calculations @xcite . with this eos
, it has been shown that self - bound stars exist with virtually no lower limit on the stellar mass @xcite ; these stars are basically giant hadrons , held together by strong interaction rather than gravity . it is arguable whether meaningful refinements of theoretical studies of strange stars can be made at this stage
. uncertainties in the strong interaction calculations easily make or kill strange stars , and arguments over details in model - dependent eos s are probably futile .
nevertheless , one could ask what kinds of effects commonly accepted ingredients of strong interaction have on the stability and structure of strange stars . whereas such studies are not meant to be quantitatively accurate , one may possibly gain insight into the physics of compact stars . with this in mind , we study the implications of chiral symmetry on the structure of strange stars .
chiral symmetry is widely recognized as an important driver of the strong interaction dynamics , witness the celebrated experimental success of partially conserved axial current ( pcac ) , which signifies the importance of chiral su(2)xsu(2 ) symmetry @xcite .
phenomenologically , the partial breaking of chiral symmetry is associated with the emergence of pion as a pseudo - goldstone boson and hence related to the small pion mass @xcite ; the coupling of pions to quarks / hadrons is thus a natural result of chiral dynamics and an essential ingredient of strong interaction physics .
there are many hadron models that incorporate various features of qcd @xcite .
whereas the mit bag model has the asymptotic freedom and the confinement of quarks built in , it is well known to violate chiral symmetry badly . by coupling the pions to the quarks , thus savaging the axial current at the bag , one can restore chiral symmetry .
there are many variants of such chiral bag models @xcite , and one can imagine these as either modifying the original bag model to include a meson cloud at the bag , or extending the sigma model to a nucleon with quark structure . for our purpose ,
a simple , representative chiral bag model suffices , and we have thus chosen the cloudy bag model @xcite .
the idea to model a hadron by putting a few relativistic quarks in a cavity has led to much phenomenological success @xcite . in the simplest version
, one considers free dirac fermions of mass @xmath9 confined in a static spherical bag of radius @xmath10 , and the ground state spatial wavefunction is simply @xcite @xmath11 \ \ , \ ] ] where @xmath12 , @xmath13 , @xmath14 , @xmath15r^3j_0 ^ 2(x)}}$ ] and @xmath16 satisfies @xmath17 } } } $ ] .
however , it was soon realized that the original mit bag boundary condition leads to severe violation of chiral symmetry @xcite .
the physical picture is that when a quark with a certain helicity collides with the impenetrable wall , its momentum is reversed but not its spin , leading to a change of the helicity .
mathematically , this is expressed by the non - conservation of the quark axial current @xmath18 , where the @xmath0 s are the dirac gamma matrices and @xmath19 is the isospin operator .
one resolution of the problem is to introduce a meson field , so that the total ( quark plus meson ) axial current is conserved @xcite .
physically , when a quark hits the bag , it produces mesons that carry away some of the axial current . in this line of thought ,
the hadron bag is always surrounded by a pion cloud , hence the name cloudy bag model .
for example , for a ground state nucleon , the pion field that will restore the chiral symmetry of a bare mit bag is @xmath20 where @xmath21 is the pauli matrix , and the spatial part of the wavefunction @xmath22 is given by @xmath23 here the pion - quark coupling constant @xmath24 is determined by the continunity of axial current at @xmath25 , which yields @xmath26 with @xmath27 is the total number of @xmath4 and @xmath5 quarks in the bag , @xmath28 is the meson mass , @xmath29 is the pion decay constant and @xmath30 .
the cloudy bag model lagrangian is a simple extension of that of the mit bag model .
here we use the linearized version : @xmath31 \theta _ v \nonumber \\ & & - { 1 \over 2 } { \bar q } \left ( 1 + i{\bf \tau } \cdot { \bf \phi } \gamma _ 5 /f \right ) q \delta _ s + { 1\over 2 } \left [ \partial _ { \mu } { \bf \phi } ( x ) \cdot \partial ^ { \mu } { \bf \phi } ( x ) - \mu ^2 { \bf \phi}^2 \right]\bar \theta_v \ \ , \end{aligned}\ ] ] where @xmath32 inside ( outside ) the bag , @xmath33 , @xmath34 is the surface delta function , and @xmath35 is the bag constant .
note that the quarks are completely confined in the bag , and the mesons only exist outside the bag .
they are free particles andare coupled to the quarks only at the bag surface , through the term proportional to @xmath36 , and are free outside or inside the bag .
the chiral bag model has essentially the same langrangian but differs from this one in that it excludes the mesons from the inside of the bag @xcite .
> from the lagrangian , one can obtain the stress - energy tensor : + @xmath37 $ ] , which gives the pressure and density of the system immediately , @xmath38 , @xmath39 . with some straightforward algebra ,
we get : @xmath40 -b \right\ } \theta _ v \ \ , \label{quarkp}\ ] ] @xmath41 + b \right\ } \theta _ v \label{quarkrho}\ ] ] inside the bag , note that if one ignores the quark mass , the eos becomes @xmath42 which is exactly the one used in previous studies of strange stars @xcite .
we assume that for a macroscopic object like a strange star , the pressure outside the bag due to the meson cloud is dominated by real and free pions , given by @xmath43 at the boundary , continunity of pressure is required .
therefore , when the quark pressure equals to the pion pressure just outside the bag , the eos is truncated to the pion one eq .
[ peos ] . in order to deduce the eos and apply it to study the properties of strange star , we need to determine the values of the bag constant ( @xmath35 ) and the strange quark mass ( @xmath44 ) . for simplicity
we have taken the masses of up and down quarks to be zero .
first , the hadron mass can be expressed as @xmath45 where @xmath10 is the radius of the hadron , @xmath46 is the eigenvalue of quarks with @xmath47 u , d and s. @xmath10 is fixed by minimizing the mass , i.e. , @xmath48 . then the hadron mass can be determined . since @xmath35 and @xmath44 are independent parameters , we first determine @xmath35 by fitting mass of the baryons which consist of only up and down quarks . in this simple model , hadrons with spin 1/2 and 3/2 are degenerate , and we minimizing the averaged root - mean - square error of @xmath49 and @xmath1 .
table [ table : b ] is the result of our fitting .
we obtain @xmath50mev .
after @xmath35 is fixed , @xmath44 can be determined by fitting baryons which consist of one or more strange quarks . > from table .
[ table : ms ] , we obtain @xmath44 = 220 mev . fig .
[ fig : eos ] shows the eos from the cloudy bag model .
we have extrapolated it into high - density region . in low - density region , since the pion field decay exponentially outside the bag , the pion ` tail ' in the eos is not important in calculation .
hence , it is not shown in the graph .
the eos s for different parameters were also shown in the figure as for comparison .
note that pion does not come into account in this part of the eos , therefore the one with @xmath51mev and @xmath52mev reduced to the mit bag case , which was used in previous studies of strange star .
assuming the pion cloud decay exponentially when leaving the stellar surface , it only exists within an effective range of scale @xmath53 ( @xmath54fm ) . the pion ` shell ' is so thin that has negligible effect on the stellar radius . using typical parameters : @xmath55 cm , @xmath56fm , , mass of the pion cloud surrounding the star can be estimated : @xmath57 which is completely negligible as compared to the stellar mass .
the pion cloud surrounding the star is a consequence of chiral symmetry consideration , it has no significance on global stellar structure such as radius and mass . using the eos of this model and the tolman - oppenheimer - volkoff equations , the properties of strange stars
can be obtained .
the global stellar structure is very similar to that of mit models @xcite . > from fig .
[ fig : mrhoc ] , maximum mass of cloudy bag strange stars using the best - fit parameters is about@xmath58 . as a comparison ,
the maximum mass of mit strange stars is 2.0@xmath59 .
these two values only differ slightly , @xmath60 .
similar to the case of mit strange stars , there is no minimum mass for the cloudy bag strange stars .
the mass - radius relation is shown in fig .
[ fig : mr ] . from the figure , radius of the cloudy bag strange star is in general larger , this is due to a softer eos obtained by a smaller value of @xmath35 .
the stellar density profile of @xmath61 strange stars is shown in fig .
[ fig : rhor14 ] , the density is quite uniform inside the star .
[ fig : z ] is the gravitational redshift ( @xmath62 ) versus central density ( @xmath63 ) for different parameters , where @xmath62 is given by @xmath64 . from the figure
, redshift depends sensitively on strange quark mass @xmath44 .
this result can be understood qualitatively from @xcite : since both the mass and radius of the star depend on the same factor of @xmath65 , their ratio , @xmath66 will be independent of @xmath35 .
then the only parameter remaining is @xmath44 .
this may provide a possible way to estimate the strange quark mass @xmath44 .
if the gravitational redshift of strange stars is observed , @xmath44 may be inferred from the graph .
rotation properties of strange star is important in observation , since the spin rate of pulsars can be measured very accurately . due to different visocity of strange matter and neutron matter , the maximum rotation frequency of strange stars
could be much higher than neutron stars @xcite .
@xcite derived the approximate mass and radius formulae for static and rotating strange stars using the general relativistic conditions of thermodynamic equilibrium and fitting with values obtained from numerical integration of structure equations . from the formulae , the maximum mass and radius of strange stars
can be predicted by fitting only one set of static configuration data . in our calculation
, we will adopt the formulae .
the equatorial radius of a rotating strange star can be expressed as a function of static mass ( @xmath67 ) and static radius ( @xmath68 ) : @xmath69^{-1/2},\ ] ] where @xmath70 is the constant of gravitation and @xmath71 is the angular velocity of rotation .
gravitational mass of a rotating strange is assumed to be in the form @xmath72 where @xmath73 is the eccentricity , defined by the ratio between polar and equatorial radii , @xmath74 .
we used the approximation as given by @xcite : @xmath75 now the gravitational mass is a function of rotation frequency .
the maximum frequency is given by the stability condition @xmath76 fig .
[ fig : mw ] shows the gravitational mass ( @xmath77 ) versus the rotation frequency ( @xmath71 ) for different strange stars with maximum masses . according to the graph ,
the maximum rotation frequency of the cloudy bag strange stars with best - fit parameters is 6300s@xmath78 , which corresponds to the maximum mass of @xmath79 . for mit strange stars ,
the kepler limit is about 9400s@xmath78 and the maximum mass is @xmath80 .
the maximum rotation frequency of a mit strange star is higher than the cloudy bag one by @xmath8130% .
this is due to different mass - radius properties .
a mit strange star in general more compact , having a smaller radius and larger mass .
hence the rotation can be faster .
in order to have a better understanding on the evolution of gravitational mass and equatorial radius as the spin rate increases , fig . [
fig : rotat ] is the plot of @xmath77 against @xmath82 and @xmath71 for the best - fit cloudy bag strange stars with different static masses .
the thick solid line on the @xmath77-@xmath82 plane is the mass - radius relation of the static stars , i.e. fig .
[ fig : mr ]
. then stars with static masses of @xmath83 @xmath84 , @xmath61 , @xmath85 and @xmath58 were investigated as rotation frequency increases .
> from the graph , @xmath82 increases monotonically with @xmath71 , while the mass attends a maximum value and then drops with @xmath71 .
the change in mass is not so obvious for a low mass strange star , e.g. @xmath83 . for the case of a @xmath61 star ,
the maximum mass is @xmath86 which corresponds to the kepler frequency of 5000s@xmath78 and the equatorial radius of @xmath87 cm .
the change in mass is about 20% and in radius is about 40% . for the star with maximum static mass ,
the kepler limit is about 6300s@xmath78 , with @xmath88 and @xmath89 cm .
the evolution of mass and radius can be investigated by the projection of the curves on the @xmath77-@xmath82 plane .
they are shown by the dotted lines .
the gravitational mass as well as equatorial radius increase with rotation frequency , until the maximum mass is reached .
projection of the curves on the @xmath71-@xmath77 plane is also shown by the dotted lines .
they demonstrate clearly the relationship between maximum rotation frequency and static stellar mass . from the graph , the kepler frequency increases with static stellar mass . hence the maximum possible
mass of a stable star is the one with largest static mass and rotating at kepler frequency .
the graph indicates that if a strange star with period less than 1ms is found , it probably has a mass greater than 1.4@xmath59 .
table [ table : summary ] gives a summary of the results for different parameters . for a rotating strange star ,
the mass and equatorial radius lie in different region from the static case . using the formulae , for given
any two stellar parameters from the gravitational mass , equatorial radius or rotation frequency , the remaining one can be determined from the graph .
this may help to determine the size or mass of a strange star since rotating period of a pulsar can be measure very accurately .
since the proposal of strange stars , much effort has been devoted to find observational properties that distinguish strange stars from neutron stars .
although many stellar properties are similar in both kinds of stars , their cooling behaviour is very different .
energy lost rate of strange matter is much higher than that of neutron matter , therefore the temperature of a young strange star is sufficiently lower than a neutron star of the same age @xcite . by chiral symmetry consideration , we suggested that a strange star should be surrounded by a pion field .
the thin pion cloud has negligible effect on global stellar structure such as mass and radius .
however , it plays an important part in strange star cooling .
the decay of pion is a rapid cooling mechanism in our model . as a result ,
the surface temperature of a cloudy bag strange star is lower than ordinary mit strange star significantly .
this may provide a hint to distinguish cloudy bag strange stars and verify the theory .
in addition , when the star is cooled by emission of pions , the huge amount of energy released within a short period may be a possible energy source of @xmath0-ray bursts . in this section
, we considered the cooling of a non - rotating bare strange star with typical mass 1.4@xmath59 , which is the typical mass of neutron star .
indeed , there is no reason why a strange star and a neutron star should have the same masses .
also it is unnecessary that there are identical twin between strange star and neutron star .
however , the radius of the strange star is insensitive to the stellar mass .
for example , the radius variation of strange star with masses between 1.4 to 2 solar masses is only 3@xmath90 .
furthermore , if the internal properties of the strange stars are the same , the cooling curves are almost identical .
of course , the density profiles of strange stars with different mass are different .
as we will point out in next section , strange stars with different masses will have different electron density due to the color - flavor locked ( cfl ) phase @xcite .
this has a very important effect in cooling calculations . since our focus in this paper is on the role of pion cloud surrounding a strange quark star , which is a natural consequence of chiral symmetry , and in particular , its effect on the cooling properties on strange stars .
we are not primarily interested in the detail of other effects , i.e. cfl of color superconductivity , in this paper .
therefore , in this paper we treat electron density as a parameter but the mass of the strange star is fixed as @xmath91 in the cooling calculations . for simplicity
we assume that all compact star models are non - rotating .
as argued by @xcite , when a strange star is formed , it is very hot and most of the normal matter in the curst is ejected .
therefore , it is expected that new born strange stars do not have crusts .
besides , we assumed that all pions will decay once they leave the stellar surface beyond the length scale of @xmath53 . for simplicity
, we assumed the star of uniform density and isothermal .
> from figure [ fig : rhor14 ] , the uniform density profile is a good approximation . also , the isothermal approximation is generally used in simplified models . as it is a bare strange star in our case ,
the core and surface temperature are equal .
note that the effects of magnetic field are neglected in this simple model .
this will be carried out in further works .
superfluidity ( superconductivity ) of quark matter is considered in the model .
it is generally believed that quarks can form colored cooper pairs near the fermi surface and become superfluid ( superconducting ) .
an early estimation of the magnitude of the gap @xmath1 was @xmath920.1 - 1 mev @xcite . on the other hand , some recent studies considered instanton - induced interaction between quarks and estimated @xmath93mev @xcite .
furthermore , recent works ( @xcite ) demonstrate the possibilities of diquark condensates characterized by large pairing gaps in quark cores of some neutron stars and in quark stars and discuss different possible phases of quark matter , for examples , the color - phase - locked u - d - s phase ( @xcite ) and two flavor u - d color - superconducting phase ( @xcite ) . in this paper ,
our main focus is on the role of the pion cloud surrounding a quark star and its effect on the cooling properties of strange stars .
we are not primarily interested in the detailed effects of color superconductivity , which has been discussed in some details in ref .
( @xcite ) . furthermore , there are large uncertainties as to the phase as well as the parameter values of color superconductivity in quark stars
. therefore we will assume that there are three flavors of u - d - s quarks in a strange star for both small and large color superconductivity gaps .
another main difference of these phases is the electron concentration , which plays a very important role in cooling calculations and will be discussed later . from bcs theory ,
the critical temperature @xmath94 is related to the pairing gap by @xmath95 where @xmath96 is the boltzmann constant .
when the temperature is below @xmath94 , it is in the superfluid state and there are some modifications to the specific heat and emissivity . in the cooling of cloudy bag
strange stars , the main cooling processes include quark urca process , pion emission and blackbody radiation .
the thermal evolution of the system is determined by the equation : @xmath97 where @xmath98 is the specific heat , @xmath99 , @xmath100 , @xmath101 are neutrino emissivity , pion emissivity and blackbody radiation respectively . in normal state
, we adopted the results of @xcite , where the quark fermi momentum was approximated by : @xmath102 the specific heat of quark matter is then given by : @xmath103 in superfluid state , the specific heat is modified , we adapted the formula by @xcite and @xcite : @xmath104 \\ & & \mathrm{\;for } \quad 0.2\,t_c\leq t\leq t_c \\ & = & 0 \quad \mathrm{\;for } \quad t < 0.2\,t_c . \end{array } \right\ } \label{eq : csf}\ ] ] it has been shown that the goldstone excitations in the quark - gluon plasma contribute to the specific heat of the strange star ( @xcite , @xcite ) .
@xmath105 where @xmath106 is the number of available massless gluon - photon states , which are present even in the color superconducting phase .
when temperature is low , the specific heat of quark matter vanishes .
however , electrons are not affected .
therefore , the specific heat of electrons becomes important .
we also considered this factor @xmath107 where @xmath108 is the electron fraction .
recently , rajagopal and wilczek @xcite have shown that the quark matter in the color - flavor locked phase of qcd , which occurs for large gaps @xmath109 ) , is rigorously electrically neutral , despite the unequal quark masses , and even in the presence of electron chemical potential . in other words ,
@xmath108 = 0 , and hence @xmath110 = 0 .
however , in order to have this electrical neutral situation the following condition must be satisfied @xmath111 where @xmath112 is the baryon chemical potential and @xmath113 is the strange quark mass .
realistically , a strange star must have a density profile .
if the above condition is satisfied in the core , it may not be satisfied near the surface , where the density is equal to 4b .
taking the typical values of baryon density near the strange star surface @xmath114@xmath115 , the baryon chemical potential is @xmath116mev and the right hand side of the equation gives @xmath117 with @xmath44 = 220 mev .
therefore we believe that @xmath108 will not be zero throughout a strange star even if its core is in color - flavor locked phase . for simplicity ,
we treat @xmath108 as a parameter .
neutrino is emitted from quark urca process : @xmath118 using the assumption of eq .
( [ eq : pf ] ) , @xcite obtained the neutrino emissivity : @xmath119 where @xmath120 is the strong coupling constant . in superfluid state ,
i.e. @xmath121 , it is suppressed by a factor of @xmath122 .
note that eq .
( [ eq : ev ] ) is an approximation to the exact form @xcite : + @xmath123 + where @xmath124 is the fermi weak coupling constant , @xmath125 is the cabibbo angle , @xmath96 is the boltzmann constant , @xmath126 , @xmath127 , @xmath128 , @xmath129 are fermi momenta of @xmath4 , @xmath5 quarks and electrons respectively , @xmath130 is the angle between the @xmath4 and @xmath73 momenta , which satisfies : @xmath131 the condition for reactions ( [ eq : dbeta1 ] ) and ( [ eq : dbeta2 ] ) to occur is @xmath132 .
we solved the equations numerically and found that in the density range we are interested , the reactions can always occur . on the other hand , we also considered the similar reactions for @xmath6 quarks , @xmath133 numerical results showed that the value of @xmath134 , which is defined by @xmath135 is smaller than 1 only if the density is as high as @xmath136 .
therefore , reactions ( [ eq : sbeta1 ] ) and ( [ eq : sbeta2 ] ) do not exist in the star . here , we have also ignored the neutrino emissivity due to the photon - gluon mixing , which can generate the a massive photon - gluon excitation @xcite .
it is because this excitation is only important for the temperature higher than 70 mev and our cooling calculations of quark stars start with temperature about 10mev .
the most important feature of cloudy bag strange stars is the emission of pions . due to the collision between quarks and the bag ,
pions are created at the stellar surface , which represents a conversion of the quark kinetic energy to the mass of the pion cloud .
some of the more energetic pions may even escape from the star carrying away substantial energy .
the pions are assumed to be decayed once left the stellar surface for a distance larger than @xmath53 .
the decay channel is @xmath137 as the decay rate is nearly 100% , it is an efficient cooling mechanism and the final product is a fireball of @xmath138 pairs and photons .
we estimated the escape velocity for pions to leave the region of @xmath53 from stellar surface .
numerical results showed that this value is very small such that almost every pion can escape and decay .
hence , we can assume all pions produced will decay .
the pion emissivity is given by : @xmath139 where @xmath140 is the energy density of pion field at stellar surface , which is fixed by the axial current conservation in cloudy bag model .
speed of pions emitted , @xmath141 , is estimated by @xmath142 , where @xmath28 is the pion mass .
therefore , @xmath143 put in @xmath144 , @xmath145 cm , we get : @xmath146 in superfluid state , since the collision between quarks and the bag is suppressed . by a similar argument ,
pion emissivity is also reduced by the factor of @xmath122 . in eq.(34 ) , we show that pions can carry away as much as @xmath147 erg / s .
however , the boundary condition requires the density of pion is only @xmath148 , from which we have estimated that the entire strange star will be covered by a pion cloudy with mass @xmath149 . the key reasons why pions with such low masses can carry away that much energy are : ( 1)the thermal conductivity of compact object like a strange star is extremely high ,
therefore it is a good approximation to assume that the entire strange star is isothermal .
in fact the characteristic time scale for the thermal equilibrium is @xmath150 where @xmath10 is the radius of the star and @xmath151 is the speed of sound , which is @xmath152 .
this time scale is @xmath153 .
therefore the thermal energy can quickly transport from core to the surface .
( 2)in order to carry away @xmath147erg / s , the pion production rate @xmath154 must be larger than @xmath155 .
since pions are produced when quarks collide the boundary of the bag , we can estimate that the collision rate between quarks and the boundary is @xmath156 , where @xmath157 is the quark number density and @xmath158 is the quark velocity .
we have assumed that the quark mass density is 4b and quarks are moving relativistically .
we can see that most pions will be re - absorbed by the boundary in order to maintain the pion density at the boundary .
only a small fraction of pions can escape .
the last cooling process we considered in cloudy bag strange stars is blackbody radiation .
it is given by the well - known formula : @xmath159 where @xmath160 is the stefan - boltzmann constant .
however , it has been pointed out that bare strange stars are very poor radiators of thermal photons with energy less than 20mev .
this is because the plasma frequency ( @xmath161 ) that is related to particle density of quarks is very high even at the surface of strange stars ( alcock et al .
on the other hand , chmaj et al .
( 1991 ) argue that low - energy photons with energy less than the plasma frequency can still leave the strange star surface due to the non - equilibrium quark - quark bremstralung radiation in the surface layer with the thickness of @xmath162fm .
nevertheless the efficiency of this non - equilibrium blackbody radiation is only @xmath163 of @xmath101 .
as it is a bare strange star in our case , the surface temperature is high and the energy of the blackbody radiation is large . besides , in superfluid state , other cooling processes are suppressed , so this non - equilibrium blackbody radiation will be a very important cooling mechanism at the later stage . here , we want to remark that eq.34 and eq.35 represent two completely different physical processes .
the production of pion is the requirement of chiral symmetry , which is believed to be a good symmetry for strong interaction , and they are produced outside the star due to the collisions between quarks and the bag boundary .
on the other hand , the reduced blackbody radiation ( chmaj et al .
1991 ) results from the collisions among quarks .
they are electromagnetic waves produced very near the stellar surface but yet inside the star and these em waves leak out the star through the bag boundary . in order to have a more realistic calculation , the effects of thermal equilibrium radiation and emission of @xmath138 pairs , as suggested by @xcite , were also discussed . in this theory , hot strange matter is filled with electromagnetic waves in thermodynamic equilibrium with quarks . the energy flux of thermal equilibrium photons radiated from a bare strange star surface is given by @xcite : @xmath164 where @xmath165mev , @xmath166 is the surface temperature , @xmath167 and @xmath168 is the coefficient of radiation transmission from strange matter to vacuum . @xmath169
@xmath170.\ ] ] it was pointed out by @xcite that the strong electric field at the surface of a strange star may be a powerful source of @xmath138 pairs .
the production rate is @xcite @xmath171j(\xi ) \,\mathrm{s^{-1}},\ ] ] where @xmath172 @xmath173 where @xmath174 is the fine - structure constant and @xmath17518mev .
the energy flux of @xmath138 pairs from unit surface is given by @xmath176 where @xmath177 is the mean energy of created particles .
the cooling mechanisms of mit strange stars are the same as those of cloudy bag strange stars except without the pion emission . as for comparison ,
the cooling of neutron stars is also discussed . following the work by @xcite
, we considered a simple case which only has neutrino emission and blackbody radiation for cooling .
@xmath94 is taken to be @xmath178k for the transition to superfluid state @xcite .
the detail relation between the core temperature @xmath179 and the surface temperature @xmath166 can be obtained by solving the energy transport equations numerically ( e.g. tsuruta 1979 ) and is found to be approximately related by a simple power law @xmath180(see shapiro and teukolsky 1983 , p.330 ) .
obviously , this simple relation can not take into account the gravitational effect .
@xcite have carried out a more detail analysis and found that : @xmath181 where @xmath182 is the surface gravity in the unit of @xmath183 .
specific heat of neutron matter is given by @xcite : @xmath184 where @xmath185 is the stellar mass in the unit of solar mass , @xmath186 .
in superfluid state , it is modified to be : + @xmath187 \ \mathrm{for } \ t\leq t_c,\ ] ] + which is exactly the same formula as eq .
( [ eq : csf ] ) . specific heat of electrons is @xmath188 neutrino luminosity consists of two parts : the first part is electron - proton coulomb scattering @xmath189 in the crust @xcite : @xmath190 where @xmath191 is the fractional mass of the crust @xmath192 .
the second part is neutrino bremsstrahlung process @xmath193 , which is given by : @xmath194 another case of neutron star is similar , but with the addition of a rapid cooling mechanism .
direct urca process @xmath195 is known to be the most efficient cooling way in neutron stars . at high temperature
, the neutrino emissivity can be higher than that from pion condensation and quark urca process by one to two orders of magnitude .
@xcite suggested that this process may occur in neutron stars if the proton concentration exceeds a certain value .
they estimated the neutrino emissivity to be @xmath196 where @xmath108 is the electron fraction , it is taken to be 0.1 in our calculation and @xmath197 is the nuclear saturation density .
we also include the pion condensation ( see shapiro and teukolsky 1983 p.323 ) as anther possible rapid cooling mechanism for neutron stars .
the cooling behaviors and the surface luminosities of 1.4@xmath59 strange stars and neutron stars are shown from fig.8 to fig.14 .
the cloudy bag strange star is the one obtained using best - fit parameters , i.e. @xmath198mev and @xmath199mev ; the mit one is using the linear eos and the neutron star is the one with typical mass of 1.4@xmath59 and radius of 10 km .
strange stars are bare while the neutron star has a crust of mass about 5% of the total stellar mass .
the initial temperature was chosen to be @xmath200k , which is a typical temperature when the star is born . due to gravitational redshift ,
the surface temperature and luminosity observed at infinity ( @xmath201 and @xmath202 ) are related to their value at stellar surface ( @xmath166 and @xmath203 ) by @xcite : @xmath204 where the surface redshift factor is @xmath205 @xmath77 is the gravitational mass of the star , @xmath10 is the stellar radius and @xmath70 is the gravitational constant .
figure [ fig:8 ] is the luminosities of various emission mechanisms observed at infinity as a function of temperature .
the calculations are based on cloudy bag strange star with @xmath206mev , @xmath207 . as a comparison , two cooling mechanisms for neutron stars are also shown .
we considered the direct urca process with @xmath207 and @xmath208k .
for pion condensation model , @xmath207 and @xmath209k .
according to the graph , energy released in direct urca process ( short dashed line ) is highest at high temperature .
then the luminosity by pion condensation ( solid line ) follows .
quark urca emission follows .
when temperature drops below @xmath210k , pion emission ( dash - dot - dotted line ) becomes the most efficient cooling mechanism and dominates over a wide range of temperature .
when temperature drops further , the star is in superfluid state .
most of the emissions are suppressed and only non - equilibrium blackbody radiation dominates . although thermal equilibrium radiation and @xmath138 pair emission remain unaffected , their values are small and negligible in low temperature .
the surface temperature of different kinds of stange stars observed at infinity is plotted in figure [ fig:9 ] .
we first considered a cloudy bag strange star with @xmath206mev and @xmath207 ( the dash - dot - dotted line ) .
as pion emission is an efficient cooling mechanism , temperature of the cloudy bag strange star drops rapidly at the beginning . from the figure , temperature drops from @xmath200k to @xmath211k within 1ms . a rough estimation of energy released : @xmath212 where @xmath213 is the total baryon number of the star .
it has been suggested that the conversion of neutron stars to strange stars could be the origin of @xmath0-ray bursts @xcite and 40% of the deconfinment energy ( @xmath214erg ) will be converted into @xmath138 pairs and photons fireball via neutrino - antineutrino annihilation . however , in our model , the evaporation of pions is a more efficient heat dissipation mechanism .
the pion decay products form a fireball of @xmath138 and photons directly instead of via neutrino - antineutrino annihilation . therefore
the fireball can carry away 100% of the deconfinment energy .
since both the time scale and order of magnitude of energy released is similar to typical @xmath0-ray bursts @xcite , we suggested that the rapid cooling of cloudy bag strange stars may be a possible energy source of @xmath0-ray bursts .
when the core temperature drops below @xmath215k , the cooling slows down . in this stage , the strange matter becomes superfluid .
therefore , emissivity of neutrino and pion are reduced by a factor of @xmath122 .
finally , the star is cooled by blackbody radiation only . for a mit strange star with @xmath1=1mev and @xmath207 ( the dash - dotted line )
, it is cooled by neutrino emission , thermal equilibrium radiation , @xmath138 emission and blackbody radiation .
thus , the cooling rate is much slower than that of a cloudy bag strange star , which emitts neutrinos as well as pions .
similarly , while temperature is lower than @xmath215k , all cooling mechanisms become negligible except blackbody radiation .
this is why the long term thermal evolution is the same in both mit and cloudy bag strange stars . in the case of @xmath1=100mev ,
temperature of the stars can never exceed the critical temperature @xmath94 .
in other words , they are already in superfluid state since born .
hence pion and neutrino emissivity are very small .
the cooling behaviour of cloudy bag and mit strange stars are similar , since they are mainly cooled by the smae mechanisms .
note that for @xmath1=100mev , stars with @xmath216 cool faster than those with @xmath207 . it is because in superfluid state
, specific heat mainly comes from the contribution of electrons .
therefore , if @xmath108 is smaller , the star can cool faster .
it is interesting to see that for mit strange stars , although the one with @xmath1=100mev cools slower , its surface temperature is lower than that of @xmath1=1mev after 1s .
this can be explained by the specific heat , since it depends on @xmath1 . under the same temperature ,
the specific heat is smaller for a larger gap .
hence , the temperature drops faster . for the long - term behavior , since it is dominated by the non - equilibrium blackbody radiation .
they only depend on @xmath108 , which is related to the specific heat .
figure [ fig:10 ] shows the surface temperature of strange stars and neutron stars observed at infinity . since neutron stars are covered by crusts , which act as thermal blankets to reduce the energy lost rate . at the beginning ,
although the core temperature of neutron stars is higher , their surface temperature is lower than strange stars by a lot .
this is due to eq .
( [ eq : tstc ] ) , surface temperature is different from core temperature by several orders of magnitude .
since neutron stars cool slowly , the surface temperature is eventually higher than that of strange stars at @xmath217s for @xmath1 = 100mev and at @xmath218s for @xmath1 = 1mev respectively . to investigate the observational properties , the luminosities of electromagnetic radiation from various emission mechanisms for cloudy bag strange stars and mit strange stars were plotted in figure [ fig:11 ] and [ fig:12 ] respectively .
photon luminosity plotted in the graphs consists of thermal equilibrium photons and non - equilibrium blackbody radiation .
note that in both graphs , photon luminosity and @xmath138 pair production start at the same point .
since they are independent of superfluidity , they are the same for all kinds strange stars .
they become different in later time because of different evolution of the star temperature . for cloudy bag strange stars with @xmath206mev and @xmath207 ( fig .
[ fig:11 ] ) , pion emission ( short dashed line ) dominates over a wide range of time until @xmath219s .
hence electromagnetic radiation mainly comes from the decay of pion in this case . in the case of cloudy bag
strange stars with @xmath220mev and @xmath216 , pion emission ( dash - dotted line ) is suppressed by superfluidity and becomes negligible .
therefore electromagnetic radiation mainly comes from blackbody radiation , equilibrium radiation and annihilation of @xmath138 pairs . for mit strange stars with @xmath206mev and @xmath207 ( fig .
[ fig:12 ] ) , electromagnetic radiation mainly comes from @xmath138 pairs ( dashed line ) from @xmath221s to @xmath219s , while photon emission ( dash - dot - dotted line ) dominates after @xmath219s .
similar for the case of @xmath220mev and @xmath216 , but the photon luminosity ( solid line ) dominates after @xmath222s . the total electromagnetic radiation from different strange stars and neutron stars are plotted in figure [ fig:13 ] and [ fig:14 ] .
strange stars in figure [ fig:13 ] were calculated using @xmath206mev and @xmath207 . from the figure ,
the cloudy bag strange star ( dotted line ) is most luminous at the beginning .
then after 1ms , it transits into superfluid state and the pion luminosity drops drastically . at this time
, the luminosity of mit strange star with @xmath206mev , @xmath207 ( solid line ) is highest . at @xmath223s
, it also drops due to superfluidity .
around @xmath224s , the luminosity of strange stars drop below neutron stars . for strange stars with @xmath220mev and @xmath216 ( figure [ fig:14 ] ) ,
both mit ( solid line ) and cloudy bag ( dotted line ) strange stars behave similarly , as pion emission is highly suppressed . in this case , the luminosity drops below that of neutron stars at @xmath225s . as compared to previous graph , @xmath108 is smaller for strange stars , therefore specific heat is smaller and the stars cools faster
. hence the luminosity drops below that of neutron stars at an earlier time .
we have used the cloudy bag model to describe the properties of strange stars .
the model parameters , i.e. , the bag constant @xmath226 and the strange quark mass @xmath227 , are fixed by comparing with hadron masses .
we found that the star is surrounded by a thin pion cloud , which is not significant in the stellar mass and radius .
hence , stellar properties are similar in the cloudy bag and mit case , except that the former has slightly larger radius and smaller mass due to a larger @xmath35 .
the maximum mass of cloudy bag strange stars is about @xmath228 . for the rotation properties of cloudy bag
strange stars , the approximate formulae were employed .
we found that the kepler limit for cloudy bag strange stars with the best - fit parameters is about 6500s@xmath78 , which is lower than the mit one by @xmath229 .
the maximum mass of a stable rotating star is about @xmath79 , similar to the value @xmath80 of a mit strange star .
the cooling properties of bare strange stars sensitively depend on the gap energy . for small gap energy ( @xmath1 = 1mev ) , the cooling properties of mit strange stars and cloudy strange stars are very much different because the pion emission can cool the strange star rapidly . for the large gap energy ( @xmath1 = 100mev ) , the quarks become superconducting immediately after the star is formed , and the pion emission is totally suppressed .
mit strange stars and cloudy strange stars become no difference . furthermore
, large gap energy also creates the effect of color - flavor locked , which significantly reduces the electron concentration and its capacity .
this makes the cooling curves lower than those with small gap energy . from fig13 and fig14
, we can see that the surface luminosity of a bare strange star is higher than that of a neutron star until @xmath2 and @xmath3 for ( @xmath1 = 100mev ) and ( @xmath1 = 1mev ) respectively .
after this period , the surface luminosity of a bare strange star becomes lower than that of a neutron star even rapidly cooling mechanisms , e.g. direct urca process or pion condensation , exist in the core of neutron stars .
hence , the cooling behavior may provide a possible way to distinguish a compact object between a neutron star , mit strange star and cloudy bag strange star in observations .
it is interesting to note that although the mit and cloudy strange stars are hotter than neutron stars before @xmath217s for @xmath1 = 100mev and @xmath218s for @xmath1 = 1mev respectively(cf .
fig10 ) , the thermal radiation of neutron stars are still stronger than those of strange stars until until @xmath2 and @xmath3 for ( @xmath1 = 100mev ) and ( @xmath1 = 1mev ) respectively . it is because the cooling mechanism at the later stage is dominated by the black body radiation , which is strongly suppressed for strange stars as pointed in section ivb .
> from fig13 and fig14 , we can see that the cooling curves of a mit strange star and a cloudy strange star are the same when the cooling mechanism is dominated by blackbody radiation
. however , the radiation properties of a mit strange star and a cloudy strange star are very much different in the early stage ( t @xmath230 ) for small gap energy case .
the cloudy strange star is first cooling very rapidly by emitting pions , which decay to 100mev photons and its temperature drops quickly below that of mit strange star . therefore the mit strange star is cooling mainly by emitting electrons / positrons .
the total energy of a cloudy strange star carried away by 100mev photons is @xmath231 , which makes the evaporation of pions a very efficient @xmath0-ray burst mechanism .
finally , we want to remark that the cooling curves of strange stars are lower than those cooling curves of neutron stars , which are consistent with the current observed data @xcite .
therefore we actually suggest that the cooling curves may be one possible way to differentiate strange stars from neutron stars .
for example , strange stars can behave as pulsars if they have strong magnetic field @xcite @xcite .
therefore their ages can be estimated by their period and period derivative .
if these pulsars are found to have thermal emission lower than that of the standard cooling neutron stars , then we can further to see if there are any e@xmath232 annihilation lines , which are red shifted larger than those of neutron stars ( cf .
fig . 5 ) .
the features of thermal x - ray emitted from these pulsars associated with bare strange stars are also very different from those of neutron stars @xcite . | as the chiral symmetry is widely recognized as an important driver of the strong interaction dynamics , current strange stars models based on mit bag models do not obey such symmetry .
we investigate properties of bare strange stars using the cloudy bag model , in which a pion cloud coupled to the quark - confining bag is introduced such that chiral symmetry is conserved .
the parameters in the model , namely the bag constant and strange quark mass are determined self - consistently by fitting the mass spectrum of baryons .
then the equation of state is obtained by evaluating the energy - momentum tensor of the system .
we find that the stellar properties of the cloudy bag strange stars are similar to those of mit bag models .
however , the decay of pions is a very efficient cooling way .
in fact it can carry out most the thermal energy in a few milliseconds and directly convert them into 100mev photons via pion decay .
this may be a very efficient @xmath0-ray burst mechanism .
numerical results indicate that temperature of a cloudy bag strange star is sufficiently lower than a mit one for the small gap energy of color superconductivity(@xmath1 = 1mev ) . on the other hand
, large gap energy ( @xmath1 = 100mev ) can suppress the pion emissivity and hence the cooling curves of cloudy model and mit model are almost identical .
the long term cooling behaviors of both mit model and cloudy model are determined by the color - flavor locked phase .
the surface luminosity of a bare strange star is higher than that of a neutron star until @xmath2 and @xmath3 for ( @xmath1 = 100mev ) and ( @xmath1 = 1mev ) respectively .
after this period , the surface luminosity of a bare strange star becomes lower than that of a neutron star even rapidly cooling mechanisms , e.g. direct urca process or pion condensation , exist in the neutron stars .
hence , the cooling behavior may provide a possible way to distinguish a compact object between a neutron star , mit strange star and cloudy bag strange star in observations . |
prediction of the critical behavior of modified spin models in the presence of site or bond dilution , random bonds , random fields has been the subject of many studies in the last decades ; this modification brings about considerable changes in the critical behavior of these systems , such as replacement of a first - order phase transition ( fopt ) by a second - order phase transition ( sopt ) , depression of tricritical points and critical end points , new critical points and universality classes , etc @xcite .
the study of the aforementioned disordered systems is based on the standard models , such as ising , blume - capel , baxter - wu , heisenberg , etc , modified accordingly to meet the requirements under consideration .
furthermore , extensions and versions of these models can be applied to describe many other different situations , such as multicomponent fluids , ternary alloys , @xmath15he -@xmath16he mixtures , in addition to the magnetic systems for which these were initially conceived . the most extensively investigated model in statistical and condensed matter physics
is the spin-1/2 ising model ( i m ) , since its two dimensional version , without an external magnetic field , was analytically solved by onsager ; it is a simple one relative to other models of cooperative phenomena and has a wide range of applicability to real physical systems .
its three - dimensional version has not yet been solved exactly , for which the only existing results are either from the renormalization group calculations or series expansions and are thus considered to be the `` exact '' ones . in its modified versions , it exhibits a variety of multicritical phenomena , such as a phase diagram with ordered ferromagnetic and disordered paramagnetic phases separated by a transition line that changes from an sopt to an fopt joined at a tricritical point ( tcp ) ; besides these , critical points , critical end points , ordered critical points of various orders , re - entrance can appear as in the presence of random fields .
the multicritical phenomena appear in systems presenting competition among distinct types of ordering and there are numerous circumstances in which this kind of phenomenon can arise . in ferromagnetic systems in the presence of random fields , the competition is between the parallel and random ordering , causing , occasionally , the conversion of a continuous transition into an fopt and the subsequent appearance of tcp as well as re - entrance in some cases @xcite .
random field effects on magnetic systems have been systematically studied not only for their own theoretical study but for their experimental importance , as well .
the methods used for the study of the i m are the mean field approximation ( mfa ) , high temperature expansions , series expansion , renormalization group , monte carlo , effective field theory ( eft ) .
the eldest one is the mfa , which is very popular because of its simplicity and has played an important role for the description of cooperative phenomena for many years ; it gives qualitative agreement with experimental results for many of the physical quantities involved in a phase transition .
however , mfa has some unsatisfactory results in describing correctly the critical region because of the omission of correlations when its results are compared with those of experiments . mfa is still implemented because it is not complicated and can give a general view of the expected behavior for the system under consideration .
many efforts were attempted towards improving mfa , such as the effective field theory by honmura and kaneyoshi @xcite .
eft relies on introducing a differential operator into the exact spin correlation function identities obtained by callen @xcite and using the van der waerden spin identities , which improves substantially on the standard mfa @xcite .
the eft approach is general and may be applied to systems with any spin value , adapting the van der waerden identities accordingly .
this procedure presents a great versatility and has been applied to several occasions such as pure , site- and bond - random ising model , although this procedure shall not yield accurate values for the physical quantities in the critical region due to the absence of long range fluctuations .
eft has already been used in numerous physical problems as a tool to study the magnetic behavior of complex spin systems , such as diluted ferromagnets @xcite , pure anisotropic systems @xcite , disordered systems @xcite , cylindrical nanowires @xcite .
the hamiltonian we shall adopt is that of ising model with nearest neighbor interactions , in which the bond @xmath12 between two neighboring spins varies from pair to pair randomly ; the system is also under the influence of a random field @xmath4 , varying from site to site ; both random variables are drawn from a suitable probability distribution function ( pdf ) .
the hamiltonian describing the above system is @xmath17 the summation in the first term extends over all nearest neighbors and is denoted by @xmath18 ; in the second term @xmath4 represents a random field that couples to the one dimensional spin variable @xmath19 .
the hamiltonian describes the competition between the long - range order ( expressed by the first summation ) and the random ordering fields .
the presence of random fields requires two averaging procedures , the usual thermal average , denoted by angular brackets @xmath20 and disorder average over the random fields denoted by @xmath21 for the respective pdf s , which are usually a version of the bimodal or gaussian distributions .
the random exchange integral @xmath12 between the sites @xmath22 and @xmath23 is drawn from the symmetric and anisotropic bimodal pdf , @xmath24 where @xmath10 and @xmath11 represent the average value and standard deviation of @xmath12 , respectively , implying that the exchange integral is fluctuating .
this choice of the pdf implies that both values ( @xmath25 ) of the random exchange integral are equally probable .
the pdf for the random fields @xmath4 is also of the bimodal type @xmath26 where @xmath27 , @xmath6 is the competition parameter and is considered to be positive so that the random fields are competitive .
the site probability @xmath28 is the fraction of lattice sites having a random magnetic field with strength @xmath5 , while the rest sites have a field with strength @xmath29 with site probability@xmath30 such that @xmath3 and the usual choice was @xmath31 , symmetric case , @xcite . as far as the pdf ( [ bimodal ] ) is non symmetric , the mean value for @xmath4 does not vanish a priori , but , instead , is @xmath32 ;
an immediate result is that the system is under the influence of a magnetic field in case @xmath33 and @xmath34 .
one of the main issues was the experimental realization of random fields .
fishman and aharony @xcite showed that the randomly quenched exchange interactions ising antiferromagnet in a uniform field @xmath35 is equivalent to a ferromagnet in a random field with the strength of the random field linearly proportional to the induced magnetization .
this identification gave new impetus to the study of the rfim , the investigation gained further interest and was intensified resulting in a large number of publications ( theoretical , numerical , monte carlo simulations and experimental ) in the last thirty years . although much effort had been invested towards this direction , the only well - established conclusion drawn was the existence of a phase transition for @xmath36 ( d space dimension ) , that is , the critical lower dimension @xmath37 is 2 after a long controversial discussion @xcite , while many other issues are still unanswered ; among them is the order of the phase transition ( first or second order ) , the universality class and the dependence of these points on the form of the random field pdf.the study of rfim has also highlighted another feature of the model , that of tricriticality and its dependence on the assumed distribution function of the random fields . according to the mean field approximation , the choice of the random field distribution can lead to a continuous fm / pm boundary as in the single gaussian probability distribution , whereas for the bimodal one this boundary is divided into two parts , an sopt branch for high temperatures and an fopt branch for low temperatures separated by a tcp at @xmath38 and @xmath39 @xcite , where @xmath40 is the coordination number and @xmath41 the boltzmann constant , such that for @xmath42 and @xmath43 the transition to the fm phase is first order for the symmetric case @xmath44 .
however , this behavior is not fully elucidated since in the case of the three - dimensional rfim , the high temperature series expansions by gofman et al @xcite yielded only continuous transitions for both probability distributions , whereas according to houghton et al @xcite both distributions predicted the existence of a tricritical point , with @xmath45 and @xmath46 for the bimodal and @xmath47 and @xmath48 for the single gaussian . in the monte carlo studies for @xmath49 , machta et al @xcite , using single gaussian distribution , could not reach a definite conclusion concerning the nature of the transition , since for some realizations of randomness the magnetization histogram was two - peaked ( implying an sopt ) whereas for other ones three - peaked , implying an fopt . in this investigation , we study the eft applied to the random field spin-1/2 ising model on a square lattice with random bond nearest neighbor interactions , calculating the relevant thermodynamic quantities , such as critical temperatures , susceptibility , magnetization profiles with respect to the temperature @xmath14 and random field @xmath5 . the paper is organized as follows : in the next section , the eft formalism is introduced and applied to the ising model in the presence of a random field and random bonds interactions ; the equation of state for the magnetization is derived . in section @xmath50 , the phase diagrams and magnetization profiles are calculated as functions of system s parameters ; we close with the conclusions in section @xmath51 .
considering the hamiltonian ( [ rham ] ) , each spin is under the influence of a molecular field @xmath52 @xmath53 where @xmath40 is the coordination number . for the current model
the starting point is the callen exact spin correlation function identity @xcite @xmath54 \rangle \rangle_{r } \label{spin1}\ ] ] where the summation takes all the nearest neighboring spin sites of @xmath22 and @xmath55 indicates the thermal and random configurational averages .
following honmura and kaneyoshi @xcite , introducing the differential operator @xmath56 into ( [ spin1 ] ) and using also the van der waerden identity @xmath57 eq .
( [ spin1 ] ) transforms into @xmath58\rangle \rangle_{r } \langle\tanh \left [ x + \beta h_{i}\right ] \rangle_{x=0 } \nonumber \\ & = & \langle \langle \prod_{j=1}^{z } \left[\cosh(\beta j_{ij}d ) + s_{j } \sinh(\beta j_{ij}d ) \right]\rangle \rangle_{r } f(x,\beta h_{i})_{x=0 } \label{spin3}\end{aligned}\ ] ] where @xmath59 before proceeding to the calculations , we shall consider the following approximations to make the problem mathematically tractable , @xmath60 the configurational average of spins and exchange integral are taken independently , @xmath61 , ( b ) the exchange integrals @xmath12 for different @xmath62 are also independent of each other , @xmath63 . under these assumptions the relation ( [ spin3 ] ) is written as @xmath64 where all the spins @xmath65 are nearest neighbors of @xmath19 and the coefficient @xmath66 is @xmath67 as far as the @xmath68 is an odd function of its argument , only odd @xmath69 appear in @xmath66 for the square ( @xmath70 ) lattice . considering the pdf ( [ bimodalj ] ) for @xmath12 ,
the averages in ( [ spin6a ] ) are @xmath71 thus ( [ spin6a ] ) becomes @xmath72 which , for even @xmath40 , takes the form @xmath73 using the operator relation @xmath74 , we obtain for the summand in ( [ spin7 ] ) @xmath75 + \tanh [ \beta ( 2\ell j - 2\nu \delta + h_{1 } ) ] + \nonumber \\
\tanh [ \beta ( 2\ell j - 2\nu \delta -h_{1 } ) ] + \tanh [ \beta ( 2\ell j + 2\nu \delta -h_{1 } ) ] \mbox{\large ) } + \nonumber \\ q \ , \mbox{\large ( } \tanh [ \beta ( 2\ell j + 2\nu \delta -h_{2 } ) ] \tanh [ \beta ( 2\ell j - 2\nu \delta -h_{2 } ) ] + \nonumber \\ \tanh [ \beta ( 2\ell j - 2\nu \delta + h_{2 } ) ] + \tanh [ \beta ( 2\ell j + 2\nu \delta + h_{2})]\mbox{\large ) } \mbox{\large \ } } \nonumber \\ \equiv \frac{1}{4}g_{\ell \nu } ( \beta , j,\delta , h_{1},h_{2 } ) \label{spin8}\end{aligned}\ ] ] thus eq . ( [ spin7 ] )
is written as @xmath76 with @xmath77|a_{\ell}^{(z)}| \nonumber \\
b_{\ell}^{(5 ) } & = & \frac{2\ell}{\prod _ { i=0}^{4}(z - i ) } [ ( 2\ell)^{4}-10(z-2)(2\ell)^{2}+ 15z^{2}-50z+24]|a_{\ell}^{(z)}| \label{spin10}\end{aligned}\ ] ]
applying this procedure to the linear chain ( @xmath78 ) without both types of randomness , namely , @xmath79 , ( [ spin5 ] ) implies @xmath80 where @xmath81 from relations ( [ spin7 ] ) , ( [ spin8 ] ) , ( [ spin9 ] ) , ( [ spin10 ] ) , so that for the respective critical point we get @xmath82 or @xmath83 which implies that @xmath84 in accordance with the known exact result for the one - dimensional i m , thus improving on the mfa result that is @xmath85 . for the disordered square lattice ( @xmath70 ) ,
the respective @xmath66 functions are @xmath86 where @xmath87 and the g s functions are defined in ( [ spin8 ] ) .
the resulting equation for the equilibrium magnetization from ( [ spin5 ] ) is @xmath88 which admits two solutions , the paramagnetic one , @xmath89 , and the ferromagnetic one given by @xmath90 whereas the critical boundary characterizing the ferromagnetic / paramagnetic phases is determined by the condition @xmath89 and results as a solution to the equation @xmath91 , which , in case both the random field strength @xmath5 and exchange integral deviation @xmath11 vanish , then it converts into @xmath92 as was also found in @xcite , and the critical temperature results as @xmath93 , which is closer to the exact one @xmath94 than the mfa one @xmath95 . at this point , it has to be noted that the decoupling scheme recalled in eq .
( [ magnz4 ] ) behaves as a better approximation than the one in the mfa , since within the eft framework the kinematics relations are treated exactly ( @xmath96 ) through the van der waerden identity ; in the present case , eft neglects correlations only between different spin variables , whereas the mfa neglects any kind of correlation , namely , the self- and multi - spin ones .
( color online ) random field @xmath5 variation of the critical temperature @xmath14 - phase diagram ( @xmath97 ) . in panel ( a ) , for the upper group of graphs @xmath98 , for the lower group @xmath99 . in each group from up to down @xmath100 . in panel
( b ) , d is fixed as @xmath101 , @xmath102 ( upper solid graphs ) , and @xmath103 ( lower dashed graphs ) , labelled by the respective @xmath104values .
the solid curves are arranged in increasing @xmath28 from up to down , whereas the dashed ones are arranged in increasing @xmath28 from down to up . in panel
( c ) , c is fixed as @xmath102 , @xmath105 upper solid graphs , and @xmath101 lower dashed graphs .
the plots in both groups are labelled by the respective @xmath104values .
@xmath14 in units of @xmath106 . ]
an important feature of the magnetic or fluid systems is their phase diagram , the temperature against a suitably chosen parameter , which , in the present case , allows the investigation more clearly of the effects of randomness itself on the transition temperature of the random system . in the current study , there are various parameters , such as @xmath107 , so that we can introduce a variety of phase diagrams by plotting the temperature with respect to one of these parameters keeping the other ones fixed .
however , in the numerical calculations to follow the main quantities @xmath108 are written with respect to the average value @xmath10 of the exchange integral , that is , @xmath109 , @xmath110 , scaled quantities ; this choice influences eqs .
( [ bimodalj ] ) and ( [ bimodal ] ) : the pdf in ( [ bimodalj ] ) is written as @xmath111 so that by choosing @xmath112 the former pdf changes into @xmath113 implying that some of the bonds are missing , diluted bonds . if we also set @xmath114 in eq .
( [ bimodal ] ) , then some of the sites are occupied either by non magnetic particles or are empty , since the respective pdf ( [ bimodal ] ) converts into @xmath115 . these cases are shown in fig [ figa](a ) with each graph labelled by the respective site probability @xmath28 with the graphs forming two groups : the upper one corresponding to @xmath98 ( site diluted system with vanishing exchange integral deviation ) and the lower group corresponding to @xmath116 , site and bond diluted system .
the choice @xmath117 implies that the deviation is of the order of the exchange integral ( @xmath118 ) so that some bonds may vanish according to the eq .
( [ bimodalj ] ) whereas the remaining ones are strengthened .
also , other phase diagrams are presented in figs .
[ figa](b , c ) with each line labelled by a p - value . in fig .
[ figa](b ) , with d fixed as @xmath101 , the five upper solid lines correspond to @xmath102 , whereas the five lower dashed ones to @xmath103 .
the topmost solid curve @xmath119 is for @xmath120 and as @xmath28 increases towards @xmath121 , ( curve @xmath122 ) the respective curve lies below that corresponding to a lower value for @xmath28 , the higher the p - value the lower the respective curve lies ; this causes a gradual reduction of the ferromagnetic region .
as far as the second group is concerned in fig .
[ figa](b ) ( dashed ones ) , an inversion of the order of the curves occurs in comparison to the upper solid ones with respect to @xmath28 , that is , now the lowermost one @xmath119 corresponds to @xmath120 , whereas as @xmath28 increases towards to @xmath121 @xmath122 , the respective line lies above that with a lower value for @xmath28 , so that , the higher the @xmath28-value the higher the curve lies , thus an enlargement of the ferromagnetic region takes place .
however , before the inversion of the order of the lines takes place in panel ( b ) , one observes coincidence of the phase - diagram boundaries for the all the @xmath28-values for @xmath123 ; in this case @xmath124 .
however , the order of the curves is unaltered if the temperature @xmath14 is drawn with respect to the random field @xmath5 for a fixed value of c ( @xmath102 ) , whereas @xmath125 varies as @xmath105 ( solid curves ) and @xmath101 ( dashed ones ) , fig .
[ figa](c ) , with the graphs for the same @xmath126value forming a separate group , whereas in fig .
[ figa](b ) they form a single group . a distinctive feature resulting from the plots in fig .
[ figa ] is that all the lines of the phase diagrams coincide for small values of @xmath5 , beginning from the same point on the t - axis for @xmath127 , implying that the critical temperature is the same regardless of @xmath6 and @xmath28 but depending only on @xmath125 .
also , in this figure the critical temperature for @xmath127 is greater the smaller the @xmath126value is .
all the critical lines , separating the fm and pm phases , are of second order phase transitions .
( color online ) the critical temperature @xmath14 as a function of the exchange - integral deviation @xmath125 . in both panels ,
the upper solid graph ( i ) corresponds to @xmath127 irrespectively of @xmath28 ( panel ( a ) ) or @xmath6 ( panel ( b ) . in panel
( a ) , the lower graphs are for fixed @xmath6 and @xmath5 as @xmath128 , @xmath129 and labelled by the selected @xmath104values . in panel ( b ) , the lower graphs are for fixed @xmath28 and @xmath5 as @xmath130 , @xmath129 and labelled by the selected @xmath131values . in both panels
the higher the labeling value the smaller the extent of the ferromagnetic phase . ]
the dependence of the critical temperature @xmath132 on the exchange - integral deviation @xmath133 appears in fig .
[ figb ] for fixed random field @xmath129 as well as for the case @xmath127 for comparison .
an overall feature of both panels is that each line starts from a different point on the @xmath134axis for @xmath135 as compared to figs .
[ figa ] . in fig .
[ figb](a ) , @xmath6 was fixed as @xmath128 with the respective curves labelled by a @xmath104value , whereas in fig .
[ figb](b ) , the constant quantity was @xmath28 as @xmath130 with the individual curves labelled by a @xmath131value . in both panels the boundary line for @xmath127 lies above the ones for non zero values of @xmath5 , thus defining the widest ferromagnetic phase space in comparison to that for non zero @xmath136 implying that the effect of including the random field is to reduce the ferromagnetic region ; also the boundaries are arranged one below the other in ascending order either for @xmath28 or @xmath6 so that each time @xmath28 or @xmath6 is increased the respective ferromagnetic region is reduced in comparison to that with a smaller @xmath28 or @xmath6 .
.[table1 ] the critical field @xmath137 values for constant values @xmath138
( second column ) and @xmath139 values for @xmath140 ( third column ) with respect the parameter @xmath28 .
@xmath141 refers to the parameter @xmath125 as a variable with constant values @xmath142
( fifth column ) , whereas @xmath143 refers to the parameter @xmath6 as a variable with constant values @xmath144
( seventh column ) . [ cols="^,^,^,^,^,^,^,^,^",options="header " , ] the critical field @xmath145 , the field for which the critical temperature vanishes , depends on the thermodynamic route followed , since the current random system contains various parameters , namely , @xmath146 ; consequently , in order to estimate @xmath145 , two of these parameters are fixed , only the remaining one varies .
initially , @xmath6 and @xmath125 are kept fixed as @xmath138 and later @xmath140 ; the values of the respective critical field @xmath145 appear in table [ table1 ] for selected @xmath104values for both choices . in the former case ,
these values are fitted by the sixth - degree polynomial @xmath147 , whereas in the latter case by the fifth - degree polynomial @xmath148 , with @xmath149 in both cases .
also , in the same table , there appear the values for the critical field @xmath141 ( fifth column ) corresponding to the constant quantities @xmath142 with parameter @xmath125 as a variable , for which the best fit polynomial is @xmath150 ( @xmath151 ) ; moreover , in the seventh column there appear the values for the critical field @xmath143 corresponding to the constants @xmath144 and parameter @xmath6 as a variable , with best fit polynomial @xmath152 ( @xmath153 ) .
( color online ) the thermal dependence of magnetization @xmath154 , order parameter
. in panel @xmath60 , the fixed parameters are @xmath155 ; each graph is labelled by the random field ratio @xmath156 : @xmath128 graph @xmath119 , @xmath157 graph @xmath158 and @xmath159 graph @xmath160 . in panel @xmath161 ,
the fixed parameters are @xmath162 ; each graph is labelled by @xmath28 and @xmath5 as : innermost graph @xmath163 , middle one @xmath164 and outermost @xmath165 . ]
in addition to the phase diagram , another important quantity is the magnetization ; its thermal behavior appears in fig .
[ figc ] for various values of the system s parameters @xmath166 , resulting by solving numerically the eq .
( [ magnz4 ] ) . in panel ( a ) the parameters @xmath167 are fixed as @xmath155 ; the individual graphs are labelled by the random field ratio @xmath156 , the outermost graph ( @xmath22 ) corresponds to @xmath128 , the middle one ( @xmath168 ) to @xmath157 and the innermost one ( @xmath169 ) to @xmath159 .
the magnetization exhibits its normal behavior but it is stronger for the smaller value of this parameter @xmath128 ( for @xmath114 the respective graph is very close to that for @xmath128 due to the finiteness of the calculations ) , which is in accordance with the respective plot in ref .
@xcite although in the latter publication the random field is not included . in the latter case ( @xmath159 ) the stronger negative random field @xmath27 yields smaller
magnetization because this field competes strongly the first term in the hamiltonian ( [ rham ] ) , which favors the parallel orientation of the spins . in panel
( b ) the parameters @xmath170 are fixed as @xmath171 , the plots are labelled by the site probability @xmath120 ( innermost graph ) and @xmath172 for the middle graph ; for comparison the case for the lack of the random field was also included , which is the outermost graph corresponding to @xmath173 . for small values of the temperature , the magnetization values are constant , independent of the relative parameters .
a susceptibility - like quantity @xmath174 can also be calculated as the first order derivative of the magnetization with respect to the applied external random magnetic field , namely @xmath175 so that @xmath176 or , for calculational purposes , @xmath177 since @xmath178 , although the product ( @xmath179 ) was earlier set as @xmath5 , now we divert for calculational purposes .
( color online ) the inverse magnetic susceptibility - like vs temperature @xmath14 , with constant @xmath180 ; graph ( i ) corresponds to @xmath181 and graph ( ii ) to @xmath117 . ]
the temperature dependence of the inverse susceptibility - like is shown in fig .
[ figd ] for selected values of the parameters @xmath166 . in both graphs ,
the susceptibility diverges twice , once at the zero temperature ( @xmath182 , finite clusters contribution ) and once again at the respective critical point @xmath132 , infinite cluster contribution .
although both graphs increase smoothly for small @xmath134values , they decrease steeply for larger ones .
a similar behavior for the susceptibility was also found by kaneyoshi et al @xcite .
in the current investigation we have determined the phase diagrams , critical temperatures , magnetization profiles and susceptibility of the nearest - neighbor spin-1/2 ising model on a square lattice , in the presence of competing random fields as well as fluctuating random bonds via eft , based on the differential operator technique and callen identity .
the topology of the system is taken into account through the coordination number .
the eft method is an improvement on the mfa , since it provides vanishing critical temperature for one - dimensional systems , also for two - dimensional systems the respective critical temperature is closer to the exact one than that of the mfa .
a variety of phase diagrams were obtained with respect to the various system s parameters , namely , @xmath183 .
the extent of the ferromagnetic region does not show systematic variation on varying the system s parameters ; it varies non monotonically for constant @xmath125 and varying @xmath6 , whereas in other circumstances it varies monotonically .
the magnetization profile was determined with respect to temperature as well as the random field @xmath5 .
an immediate result of the presence of the random field is to reduce the extent of the ferromagnetic region , in general , as well as numerical value of the magnetization as @xmath6 increases because of the strengthening of the random field @xmath27 and the subsequent increased tendency of this field to prevent spins to attain the parallel configuration and the competition with the other random field @xmath5 .
although the eft technique improves on the mfa , it presents the same shortcomings in the critical region ( non - classical region ) concerning the critical exponents like the mfa and landau theory , because the fluctuations occurring in the critical region , as the transition temperature is approached , become important and the non - classical behavior is observed ; these fluctuations are not taken under consideration properly by eft or mfa , thus a relative criterion , called ginzburg criterion , determines how close to the transition temperature the true critical behavior is revealed @xcite .
this criterion relies on any thermodynamic quantity but the specific heat is usually considered for determining the critical region around @xmath132 where the mean field solution can not describe correctly the phase transition .
the mfa is valid for lattice dimensionality greater than the upper critical dimension @xmath184 in case of presence of only thermal fluctuations .
however , in the current case the presence of random fields enhances fluctuations causing the critical region to be wider than the one due only to the thermal fluctuations and the upper critical dimension is increased by @xmath185 to @xmath186 @xcite . occasionally the non classical region is extremely narrow so that the respective critical behavior expected by landau or mfa is observed because the fluctuation region is very narrow and hardly accessible for experimental observation ; such a system is the weak - coupling superconductor in three dimensions for which the respective non classical region is @xmath187 ( @xmath188 is the reduced temperature ) .
however , on reducing the space dimension as in the case of the weak - coupling superconductor in two dimensions , the critical exponents have their classical values up to @xmath189 , thus the reduction of the space dimensionality causes serious repercussions on the behavior of the physical system ; on the contrary , for the superfluid helium transition the classical region extends up to @xmath190 so that fluctuations are detectable @xcite .
in addition to superconductivity , the extent of the non classical region for the ferroelectric system triglycine sulfate ( tgs ) is relatively small and its critical exponents have the respective classical values up to @xmath191 @xcite .
the results obtained in the current investigation by using the eft provide a basis for a comprehensive analysis by more sophisticated methods . however , they are of no less importance , since they show , nevertheless , the expected phenomena to be observed . | the spin-1/2 ising model on a square lattice , with fluctuating bond interactions between nearest neighbors and in the presence of a random magnetic field , is investigated within the framework of the effective field theory based on the use of the differential operator relation .
the random field is drawn from the asymmetric and anisotropic bimodal probability distribution @xmath0 , where the site probabilities @xmath1 take on values within the interval @xmath2 $ ] with the constraint @xmath3 ; @xmath4 is the random field variable with strength @xmath5 and @xmath6 the competition parameter , which is the ratio of the strength of the random magnetic field in the two principal directions @xmath7 and @xmath8 ; @xmath6 is considered to be positive resulting in competing random fields .
the fluctuating bond is drawn from the symmetric but anisotropic bimodal probability distribution @xmath9 , where @xmath10 and @xmath11 represent the average value and standard deviation of @xmath12 , respectively .
we estimate the transition temperatures , phase diagrams ( for various values of system s parameters @xmath13 ) , susceptibility , equilibrium equation for magnetization , which is solved in order to determine the magnetization profile with respect to @xmath14 and @xmath5 .
ising model , fluctuating pair interactions , asymmetric bimodal random field , effective field theory , phase diagram , phase transitions , magnetization 05.50.+q , 75.10.hk , 75.10.nr , 75.50.lk |
the double layered cmr manganite materials , la@xmath4sr@xmath5mn@xmath0o@xmath1 @xcite , demonstrate magnetic behaviour that is distinct from the perovskite manganites , la@xmath6sr@xmath7mno@xmath8 , including : ( i ) strong anisotropy of electrical ( magneto- ) transport @xcite ; ( ii ) nearly 2d character of its magnetism and strong afm magnetic correlations above t@xmath9 @xcite ; and ( iii ) anomalous magnetoelastic properties @xcite .
most experimental reports on the layered manganites have concentrated on the la@xmath4sr@xmath5mn@xmath0o@xmath1 compounds with @xmath10 , which are metallic and demonstrate strong magnetoresistive effects .
stochiometric lasr@xmath0mn@xmath0o@xmath1 ( x=0.5 ) exhibits interesting spin , charge and orbital ordering , and its resistivity is of the order of @xmath11 with a rather flat temperature dependence @xcite .
this can be compared with the situation in perovskite manganites where undoped lamno@xmath8 is an antiferromagnetic insulator with strong orbital ordering and la@xmath6sr@xmath7mno@xmath8 with @xmath10 is a metal with strong magnetoresistive effects @xcite . in comparison with the pseudocubic perovskites with three - dimensional networks of mno@xmath12 octahedra ( i.e. , ( la , sr)mno@xmath8 ) , these layered structures + ( la , sr)@xmath8mn@xmath0o@xmath1 have a reduced exchange coupling between the mn ions along the _ c _ direction .
indeed , the pseudo - cubic perovskites show ferromagnetism and metallic conductivity over a wide range of hole doping , suggesting that the double - exchange mechanism is dominant among itinerant _ e@xmath13 _ electrons . in the double - layered case , with the two - dimensional mn - o network , consisting of two perovskite blocks separated by an intervening insulating layer of ( la , sr)o ions along the _ c _ axis , the balance between antiferromagnetism and ferromagnetism is very sensitive to @xmath2 band filling @xcite .
bilayer la@xmath4sr@xmath5mn@xmath0o@xmath1 demonstrates ferromagnetism in a doping range @xmath14 , a canted antiferromagnetic structure for a hole concentration @xmath15 , and layered antiferromagnetic states for @xmath16 @xcite . from experimental data @xcite
, it is known that the ground state spin structure of lasr@xmath0mn@xmath0o@xmath1 phase is a - type layered antiferromagnet ( afm ) , where the magnetic moments lie in the @xmath17 plane and couple ferromagnetically within the single mno@xmath0 layer , but show afm order between the respective mno@xmath0 layers within the bilayer unit .
the interlayer coupling in the bilayer stack is ferromagnetic .
one of the most striking properties of the bilayered lasr@xmath0mn@xmath0o@xmath1 is the competition between the ( charge - disordered ) a - type afm spin ordering with a nel temperature t@xmath18170 k and the ce - type @xcite charge / orbital ordering which exists between @xmath19 and t@xmath20=210 k @xcite .
with decreasing temperature , the development of the charge / orbital ordering phase is disrupted by the onset of the a - type afm ordering at t@xmath21 which may be qualitatively explained by the onset of ferromagnetic order in each @xmath17 plane .
an important issue in the theory of cmr oxides is the role of coulomb correlation .
an accurate treatment of correlation may significantly affect the balance between ferromagnetic and antiferromagnetic interactions and hence the magnetic ground state .
lda band structure calculations for cmr perovskites demonstrated the possibility of lda theory predicting the correct magnetic ground state @xcite .
for the double layered perovskites , an investigation of the reliability of this theory has not yet been performed . in their report on the electronic structure calculations of lasr@xmath0mn@xmath0o@xmath1 within the general gradient approximation ( gga ) @xcite , boer and
de groot suggested to take into account some additional correlation corrections to obtain a truly half - metallic electronic structure in the layered manganites .
there is one report of an lda+u calculation @xcite performed for the doping level of 0.4 holes per mn site ( la@xmath22sr@xmath23mn@xmath0o@xmath1 ) , but details of their calculation have not been presented and the important question of how correlations affect the magnetic ground state has not been investigated .
the purpose of our work is to study the electronic structure which governs the microscopic origin of the magnetic phenomena in the bilayered manganite , and to determine its dependence on coulomb correlations and its relation with magnetic properties .
we consider the difference in the total energies , exchange interaction parameters , the population of mn - d and o - p states near the fermi level and d sub - band widths for ferromagnetic ( fm ) and a - type antiferromagnetic ( afm ) configurations as a function of the coulomb correlation parameter u and give a simple interpretation of exchange coupling between mn layers in lasr@xmath0mn@xmath0o@xmath1 within the double exchange model .
in contrast to the case of lamno@xmath8 , we find that the inclusion of u in lasr@xmath0mn@xmath0o@xmath1 results in important differences in the @xmath2-band characteristics that are responsible for the observed fm instability and for strong anisotropy in the electrical transport .
the lsda and lsda+u calculations were realized in the frame - work of the linear - muffin - tin - orbital method in the atomic sphere approximation ( lmto - asa ) @xcite .
we used the von barth - hedin - janak form @xcite for exchange - correlation potential .
the crystal parameters were taken from ref .
the atomic sphere radii were : r(la(sr))=3.4 a.u .
, r(mn)=2.8 a.u . , and r(o)=2.2 a.u . from the constrained lsda supercell calculations
@xcite , we obtained values of the coulomb and exchange parameters to be u=7.2 ev and j=0.78 ev .
these values are typical for the transition - metal oxides @xcite .
as a first step , we calculated the electronic structure of fm ( all atoms in every layer and between layers are ordered ferromagnetically ) and a - type afm ( ferromagnetic layers stacked antiferromagnetically ) lasr@xmath0mn@xmath0o@xmath1 by the standard lsda method ( u=0 ) .
the total and projected densities of states ( dos ) for these cases are shown in fig .
1 . for the majority - spin channel ( solid line ) , mn 3d states
form the bands between 2.2 ev below e@xmath24 and 2.5 ev above e@xmath24 .
as seen from the figure , the @xmath2 ( @xmath25 and @xmath26 ) bands are partially filled , cross e@xmath24 , and are rather broad compared to the @xmath27 ( @xmath28 and degenerate @xmath29 , @xmath30 ) bands which are about 1.2 ev wide and lie 1 ev below e@xmath24 .
the exchange interaction splits the mn 3d states such that the @xmath27 and @xmath2 minority - spin bands are located 2.6 ev higher in energy than the majority - spin states .
we obtained a band gap of 2.5 ev in the minority - spin @xmath31 bands and a band structure that is in a very good agreement with that obtained by the full - potential lmto method @xcite ( but their band gap within the minority spin channel is about 1.7 - 1.9 ev ) .
our results for the ferromagnetic state with u=0 differ from the results of dessau _ et al _ @xcite where a u of 2 ev was used to obtain a gap at e@xmath24 in the minority - spin bands for la@xmath22sr@xmath23mn@xmath0o@xmath1
. it should be noted , that the fermi level in our calculations was found to lie at the bottom of the @xmath27 minority conduction band ( as in the flmto calculation @xcite ) , so the electronic structure just misses to being a half - metallic one . in order to investigate the role of on - site coulomb correlation
, we calculated the electronic structures of lasr@xmath0mn@xmath0o@xmath1 for both fm and a - type afm orderings using a number of values of u , namely 2 , 4 , 6 , 7.2 , 8 , 9 and 10 ev .
the total and projected dos s of the fm and afm cases with u=7.2 ev are shown in fig .
2 . only majority @xmath2 states make a significant contribution to the dos at e@xmath24 . for u=0 , one can expect three - dimensional conduction in both spin channels ( fig .
1 ) , while for u@xmath320 we obtain a half - metallic state electron conduction is possible only within majority spin sublattice .
we want to emphasize here analogues in the main band characteristics near e@xmath24 in the calculated electronic spectra of double layered and doped perovskite manganites @xcite . and
since doped perovskite manganites are found to be half - metallic @xcite , this may also be the case for the double - layered manganite . to investigate the variation ( behavior ) of the electrically active states as a function of on - site coulomb correlation , we plot in fig .
3 the dependence on u of the @xmath26 and @xmath25 states contribution to the dos at e@xmath24 ( fig .
3(a ) ) and to the number of states integrated in a small energy window of 0.3 ev just above e@xmath24 ( denoted as n@xmath33 and n@xmath34 in fig .
the difference between fm and afm phases for a particular u value is clearly seen from fig .
3(a ) . for the fm phase ( dotted lines ) , both contributions of the @xmath26 ( circles ) and @xmath25 ( crosses ) states decrease with u in almost the same way , and only large u values ( u@xmath327 ev ) result in a splitting of the @xmath2 states .
the most significant changes occur for afm spin alignment ( solid lines ) .
the @xmath2 states are already split in the lsda calculation ( u=0 ) and the difference between the electron population of @xmath26 and @xmath25 orbitals is clearly seen ( fig .
3(a ) ) . further , while the contribution of the @xmath25 state to the dos at e@xmath24 does not change with an increase of u , the @xmath35 contribution , which is almost two times bigger than that of @xmath36 for u=0 , drastically decreases with u , taking its minimum for u@xmath327 ev . as can be seen from fig .
3(b ) , the situation does not change qualitativly in the energy window just above e@xmath24 , encouraging us to draw the following conclusions : for small u values we have 3d conduction , while for u@xmath326 the electron conduction ( hopping ) along the @xmath37 axis becomes very small .
thus , on - site correlation treated within lsda+u promotes 2d type ( in - plane ) electronic behavior in the system .
we need to add here that the significant difference between @xmath26 and @xmath25 states ( due to an decrease of the band width as well as a shift to higher energy of the @xmath26 state with increase of u ) is an essential feature of lasr@xmath0mn@xmath0o@xmath1 .
it sets the double layered manganite apart from perovskite manganites ( i.e. , ( la , sr)mno@xmath8 ) in which the main energy characteristics such as widths and centers of both @xmath2 bands are very close . the anisotropy in the population of the two @xmath2 states established above can be illustrated more clearly in the plot of the angular distribution of the electron density in the 0.3 ev energy window above e@xmath24 for u=0 and for the calculated value , u=7 ev ( c.f . ,
a comparison of the lowest spin configurations ( fm for u=0 ( fig .
4(a ) ) and afm for u=7 ev ( fig .
4(d ) ) as obtained from total energy differences ) shows that both on - site correlation and change in spin ordering significantly decrease the number of electrons that contribute to electrical transport along the @xmath37 axis .
this is contrary to the in - plane state which remain less sensitive to the spin alignment change and influence of u. ( note , however , that only on - site correlation supresses the minority @xmath27 ( @xmath28 ) state contribution to the electron density above e@xmath24 , resulting in a half - metallic state , as mentioned above . )
the decisive role of the on - site coulomb correlation in lowering the dimensionality of the conduction band is clearly seen from a comparison of two afm cases with u=0 and u=7 ev ( fig .
4(b ) and 4(d ) ) : for u=7 ev , only the @xmath25 orbital has some contribution just above e@xmath24 and hopping along the @xmath37 axis is negligible ( the @xmath26 state is responsible for it ) .
this is consistent with the point that the main issue for realizing the layered afm state is the strong anisotropy in the population of the two - dimensional @xmath38 state and the one - dimensional @xmath39 state . for the fm state
, the orbital configurations have more electron population freedom , because three crystallographic axes are equivalent .
qualitatively this situation does not change as u increases . according to the qualitative picture of the simple double exchange model , an electrically active electron has a finite hopping probability ( t ) between ferromagnetically ordered mn ions , but this hopping vanishes in the case of antiferromagnetic spin alignment due to strong hund coupling ( t@xmath40j ) .
figure 4(b ) clearly shows the failure of the lsda sheme .
thus , taking into account on - site coloumb correlation gives us the correct picture of 2d - like charge transport in the layered manganite : the increase of u suppresses the @xmath37 axis transport , keeping electron conduction in the @xmath17 plane .
however , it should be noted that the transition from fm to afm ordering ( fig .
4(c ) and 4(d ) ) also contributes to this tendency , but to a lesser degree than on - site correlation as can be seen from a comparison of fig .
4(a ) and 4(d ) . to investigate how the magnetic interaction and ground state depend on u , we calculated total energy differences for u=0 , 2 , 4 , 6 , 7.2 , 8 , 9 and 10 ev .
the results are shown in fig .
5(a ) , where the total energy of the fm spin ordering for each value of u is taken to be zero .
although the difference of the two calculated spin alignments is @xmath41 0.1 ev for all cases ( and so we have almost degenerate magnetic states ) , for u=0 , u=2 and u=4 ev the fm ordering is preferred , while for u=6 , 7.2 , 8 , 9 and 10 ev the afm phase has lower energy .
thus , the on - site mn - d electron coulomb correlations modify the magnetic ordering from fm to a - type afm as experimentally observed . note , that the total energy difference between fm and afm states does not change for u @xmath3 8 ev .
to determine the character of the exchange coupling in the layered manganite and to understand why the experimentally observed a - type afm spin ordering is simulated only with u@xmath327 ev , we also calculated the effective exchange interaction parameters ( for computational details see @xcite ) . in fig .
5(b ) , the values of the d - d effective exchange parameters , @xmath42 , between nearest mn neighbours which belong to the different layers of one bilayer are shown as a function of u. they are seen to follow the behavior of the total energy differences shown in fig .
usually , the exchange coupling is described within the superexchange ( sex ) and double exchange ( dex ) models and corresponding solutions of the simple hamiltonian problem with the following parameters : hopping between mn ions ( t ) , j ( dex model ) ; and t , u ( sex model ) . to provide interpretation to this model
, we plot the @xmath2-sub - band widths of the fm phase for a number of u values in fig .
5(c ) ( denoted as w@xmath33 and w@xmath34 below ) .
as can be seen , the @xmath25 band width almost does not change its value with u , and the main differences in the electronic spectra due to an increase of u are connected with the @xmath26 state .
the decrease of the difference between the w@xmath33 and w@xmath34 band widths from u=0 up to u@xmath435 ev in fig .
5(c ) , coming from the increase of w@xmath33 , promotes the dex interactions mediated by itinerant electrons , fig .
further , the positive ( fm ) contribution to the exchange interaction changes its sign at about u=7 ev , where a sharp decrease of the w@xmath33 value for the fm phase occurs .
in addition , due to a redistribution of the electron density between states , w@xmath34 increases slightly , and so the splitting of these two @xmath2-sub - bands increases . thus , coulomb correlations supress the dex contributions to the interlayer exchange interaction energy , and we believe that the sharp @xmath26 band width narrowing in the range of 6@xmath44u@xmath447 ev is responsible for the fm instability , and hence , is the mechanism that produces the change in magnetic ordering . in addition , this splitting of the @xmath2 states promotes strong anisotropy in the electrical transport , as described above .
in summary , we have calculated the electronic structure of double layered + lasr@xmath0mn@xmath0o@xmath1 using both the lsda and lsda+u methods and have investigated the influence of the coulomb interaction parameter on both the electronic structure and the magnetic ordering .
the main factors governing band formation in lasr@xmath0mn@xmath0o@xmath1 are : ( i ) the exchange splitting with almost unoccupied minority spin mn-3d states ; ( ii ) the ligand field splitting of the @xmath27 and @xmath2 states ; and ( iii ) futher splitting of the @xmath2 states with increase of u. we consider ( iii ) an essential factor in the layered manganites .
we have examined fm and a - type afm ground states of this compound by comparing their total energy differences for a number of u values ( from 0 to 10 ev ) and found that the experimentally observed magnetic ordering is reproduced within lsda+u only for u@xmath37 ev - in contrast with the 3d manganites ( lamno@xmath8 ) where lsda gives the true magnetic ground state .
it was shown that the correlations reduce significantly the effective hopping parameter for the electrically active @xmath2 electrons along the @xmath37 axis , so that taking into account the coloumb correlation gives the correct 2d - picture in the layered manganite .
the strong anisotropy in electrical transport which we obtained is consistent with observations of anisotropic transport - a property which sets this manganite apart from conventional 3d systems .
99 y. moritomo , a. asamitsu , h. kuwahara and y. tokura , nature ( london ) 380 ( 1996 ) 141 . t. kimura _
et al _ , science 274 ( 1996 ) 1698 . t. g. perring , g. aeppli , y. moritomo and y. tokura , phys .
78 ( 1997 ) 3197 .
t. kimura , y. tomioka , a. asamitsu and y. tokura , phys .
81 ( 1998 ) 5920 .
s. rosenkranz _
et al _ , j. appl .
83 ( 1998 ) 7348 .
m. kubota , h. yoshizawa , y. moritomo , h. fujioka , k.hirota and y. endoh , j. phys .
japan 68 ( 1999 ) 2202 .
a. urushibara , y. moritomo , t. arima , a. asamitsu , g. kido and y. tokura , phys .
b 51 ( 1995 ) 14103 . p. d. battle _
et al _ , chem .
( 1997 ) 1042 .
m. kubota _ et al _ , j. phys . and
solids 60 ( 1999 ) 1161 .
y. moritomo , y. maruyama , t. akimoto and a. nakamura , j. phys . soc .
japan 67 ( 1998 ) 405 .
t. hayashi , n. miura , m. tokunaga , t. kimura and y. tokura , j. phys .
: condens .
matter 10 ( 1998 ) 11525 .
d. n. argyriou _
et al _ , phys .
b 61 ( 2000 ) 15269 . tapan chatterji _
et al _ , phys .
b 61 ( 2000 ) 570 . e. o. wollan and w. c. koehler , phys .
100 ( 1955 ) 545 .
w. e. pickett and d. j. singh , phys .
b. 53 ( 1996 ) 1146 .
p. k. boer and r. a. de groot , phys .
b 60 ( 1999 ) 10758 . d. s. dessau _
et al _ , phys .
81 ( 1998 ) 192 .
o. k. andersen , phys .
b 12 ( 1975 ) 3060 .
u. von barth , l. hedin and b. janak , phys .
b 12 ( 1975 ) 1257 . o. gunnarsson , o. k. andersen , o. jepsen and j. zaanen , phys . rev .
b 39 ( 1989 )
1708 . v. i. anisimov and o. gunnarsson , phys .
b 43 ( 1991 ) 7570
. v. i. anisimov , j. zaanen and o. k. andersen , phys .
b 44 ( 1991 ) 943 ; v. i. anisimov , f. aryasetiawan and a. i. lichtenstein , j.phys . :
condens . matter 9 ( 1997 ) 767 .
s. satpathy , z. s. popovic and f. r. vukajlovic , phys .
76 ( 1996 ) 960 . i. solovyev , n. hamada and k. terakura , phys . rev .
b 53 ( 1996 ) 7158 .
x. huang , o. mryasov , d. novikov and a.j .
freeman , phys .
b , 62 ( 2000 ) 13318 .
z. fang , i. v. solovyev and k. terakura , phys .
84 ( 2000 ) 3169 .
j. y. t. wei , n .- c .
yeh , and r. p. vasquez , phys .
79 ( 1997 ) 5150 .
j .- h . park _
et al _ , nature ( london ) 392 ( 1998 ) 794 .
a. i. lichtenstein , j. zaanen and v. i. anisimov , phys .
b 52 ( 1995 ) r5467 . | a detailed study of the electronic structure and magnetic configurations of the 50 % hole - doped double layered manganite lasr@xmath0mn@xmath0o@xmath1 is presented .
we demonstrate that the on - site coulomb correlation ( u ) of mn d electrons _
( i ) _ significantly modifies the electronic structure , magnetic ordering ( from fm to afm ) , and interlayer exchange interactions , and _
( ii ) _ promotes strong anisotropy in electrical transport , reducing the effective hopping parameter along the _ c _ axis for electrically active @xmath2 electrons .
this findng is consistent with observations of anisotropic transport a property which sets this manganite apart from conventional 3d systems .
a half - metallic band structure is predicted with both the lsda and lsda+u methods .
the experimentally observed a - type afm ordering in lasr@xmath0mn@xmath0o@xmath1 is found to be energetically more favorable with u @xmath3 7 ev .
a simple interpretation of interlayer exchange coupling is given within double and super - exchange mechanisms based on the dependencies on u of the effective exchange parameters and @xmath2 state sub - band widths .
_ keywords : _ double layered manganite ; anisotropy in electrical transport ; half - metallic |
at present the top quark has not yet been conclusively discovered at the fermilab tevatron , even though there are events which look similar to those expected from top quark decays .
the experimental situation is presently confusing because both the cdf @xcite and d0 @xcite collaborations only have limited statistics .
as more events are collected one expects that the situation will be clarified . at the tevatron
, the top quark should be mainly produced through @xmath2 pair production from the light mass quarks and gluons in the colliding proton and antiproton .
both the top quark and the top antiquark then decay to @xmath3 pairs , and each @xmath4 boson can decay either hadronically or leptonically .
the @xmath5-quark becomes an on - mass - shell @xmath6-hadron which subsequently decays into leptons and ( charmed ) hadrons .
a large effort is being made to reconstruct the top quark mass from the measured particles in the decay , which is complicated by the fact that the neutrinos are never detected .
also there are additional jets so it is not clear which ones to choose to recombine .
the best channel for this mass reconstruction is where both @xmath4 bosons decay leptonically , one to a @xmath7 pair , the other to a @xmath8 pair ( a dilepton event ) because the backgrounds in this channel are small .
when only a single lepton is detected then it is necessary to identify the @xmath5 quark in the decay to remove large backgrounds from the production of @xmath9 jets @xcite . in all cases
the reconstruction of the particles in the final state involves both the details of the production of the top quark - antiquark pair as well as the knowledge of their fragmentation and decay products .
the cdf collaboration @xcite have reported two events with dilepton final states , six events with a single lepton and a @xmath5-quark identified by a secondary vertex , and seven single lepton events with the @xmath5-quark identified by a semileptonic decay .
the cdf collaboration then constructed a likelihood function for the invariant mass @xcite and quoted the value @xmath10 gev@xmath11 .
the top quark cross section quoted by the cdf collaboration is @xmath12 pb . the d0 collaboration @xcite have reported on nine events with an expected background of @xmath13 . if the excess is due to @xmath14 production and if the top quark mass is 180 gev@xmath15 , then the top quark cross section is @xmath16 pb . the values for the top quark production cross section as a function of the top quark mass used by the cdf and d0 collaborations contain both the nlo qcd corrections and an extension to include the resummation of initial state soft partons to all orders in perturbation theory @xcite .
a recent summary of the theoretical predictions has been presented in @xcite . in the laenen
paper @xcite the dis factorization scheme was used with the mrsd@xmath17 parton distributions @xcite , the two - loop running coupling constant with five active flavors , and @xmath18 gev . in the analysis of the decay distributions one
needs knowledge of the inclusive differential distributions of the heavy quarks in transverse momentum @xmath19 and rapidity @xmath20 .
these distributions are known in nlo .
what we would like to discuss in this paper is an updating of the resummation effects on the inclusive transverse momentum distribution of the top quark . in the original paper @xcite it
was not known which mass to choose whereas now we can assume that the mass is @xmath21 gev@xmath15 .
also we discuss here how the resummation effects modify the rapidity distribution of the top quark .
since there have been suggestions of using the mass and angular distributions in top quark production to look for physics beyond the standard model @xcite it is very important to know the normal qcd predictions for these quantities .
we first summarize what is known on the top quark cross section .
if the top quark mass is 175 gev@xmath15 then the dominant production channel is @xmath22 .
in lowest order qcd perturbation theory it contributes about 90 % of the total cross section with the reaction @xmath23 making up the remaining 10% .
one notes that the nlo corrections in the @xmath24 channel are small , whereas those in the @xmath25 channel are more than 80% . at this large top quark mass
the @xmath26 and @xmath27 channels give negligible contributions so we do not consider them . even though the @xmath25 channel contribution is small in born approximation it can be significant in nlo due to multiple soft parton radiation .
these large corrections are predominantly from the threshold region for heavy quark production .
it was shown previously @xcite that initial state gluon bremsstrahlung ( isgb ) is responsible for the large corrections at nlo near threshold .
in @xcite the dominant logarithms from isgb , which are the cause of the large corrections near threshold , were carefully examined .
such logarithms have been studied previously in drell - yan ( dy ) @xcite production at fixed target energies ( again near threshold ) where they are responsible for correspondingly large corrections .
the analogy between dy and heavy quark production cross sections was exploited in @xcite and a formula to resum the leading and next - to - leading logarithms in pqcd to all orders was proposed . since the contributions due to these logarithms are positive ( when all scales @xmath28 are set equal to the heavy quark mass @xmath29 ) , the effect of summing the higher order corrections increases the top quark production cross section over that predicted in @xmath1 .
this sum , which will be identified as @xmath30 , depends on a nonperturbative parameter @xmath31 .
the reason that a new parameter has to be introduced is that the resummation is sensitive to the scale at which pqcd breaks down .
as we approach the threshold region other , nonperturbative , physics plays a role ( higher twist , bound states , etc . )
indicated by a dramatic increase in @xmath32 and in the resummed cross section .
this is commonly called the effect of the infrared renormalon or landau pole @xcite .
we chose to simply cut off the resummation at a specific scale @xmath31 where @xmath33 since it is not obvious how to incorporate the nonperturbative effects .
note that our resummed corrections diverge for small @xmath31 but this is _ not _ physical since they should be joined smoothly onto some nonperturbative prescription and the total cross section will be finite .
another way to make it finite would be to avoid the infrared renormalon by a specific continuation around it , i.e. the principal value resummation method .
however , at the moment our total resummed corrections depend on the parameter @xmath31 for which we can only make a rough estimate .
see @xcite for more details .
to make this paper self - contained we list some relevant formulae .
the partonic processes under discussion will be denoted by @xmath34 where @xmath35 . the kinematical variables @xmath36 are introduced in the calculation of the corrections to the single particle inclusive differential distributions of the heavy ( anti)quark .
we do not distinguish in the text between the heavy quark and heavy antiquark since the distributions are essentially identical in our calculations . here
@xmath37 is the square of the parton - parton c.m .
energy and the heavy quark transverse momentum is given by @xmath38 .
the rapidity variable is defined by @xmath39 .
the born approximation differential cross sections can be expressed by @xmath40 with @xmath41\,,\ ] ] and @xmath42 \nonumber \\ & & \times\big [ \frac{t_1}{u_1 } + \frac{u_1}{t_1 } + \frac{4m^2s}{t_1u_1 } \big(1 - \frac{m^2s}{t_1u_1}\big ) \big ] \,.\end{aligned}\ ] ] here the color factors are @xmath43 and @xmath44 .
the color average factors are @xmath45 and @xmath46 .
the parameter @xmath28 denotes the renormalization scale . in @xcite
the inclusive cross section was examined near threshold ( @xmath47 ) where the contributions from the radiation of soft and collinear gluons are large .
a variable @xmath48 was defined , where @xmath49 and @xmath50 are inelastic variables in the channel @xmath51 .
the variable @xmath52 now depends on the four momentum of the extra parton(s ) emitted in the reaction . in the born approximation
there are no additional partons so @xmath53 . in @xcite
the nlo contributions were examined in the soft region ( where @xmath54 ) and it was found that the dominant contribution to the nlo cross section in this region had a similar form to the nlo correction in the drell - yan process .
as the latter correction is known exactly in nnlo @xcite this correspondence was used to write the differential cross section in order @xmath55 as follows @xmath56 \sigma^b_{ij}(s , t_1,u_1 ) \,.\end{aligned}\ ] ] here a small parameter @xmath57 has been introduced to allow us to distinguish between the soft ( @xmath58 ) and the hard ( @xmath59 ) regions in phase space .
the quantities @xmath60 contain terms involving the qcd @xmath61-functions and color factors .
the variables @xmath62 and @xmath63 were then mapped onto the variables @xmath64 and @xmath65 , where @xmath66 is the parton - parton c.m . scattering angle .
after explicit integration over the angle @xmath66 , the resulting series was exponentiated by the introduction of the @xmath64 variable into the argument of the running coupling constant . as noted in the previous paper @xcite in addition to the total cross section we can also derive the resummed heavy ( anti)quark inclusive @xmath19 ( and below the @xmath20 ) distributions .
the transverse momentum @xmath19 of the heavy quark is related to our previous variables by @xmath67^{1/2}\big\}\,,\ ] ] @xmath68^{1/2}\big\}\,,\ ] ] with @xmath69 the double differential cross section is therefore @xmath70^{1/2 } % \frac{d^2\sigma_{ij}(s , s_4 , p_t^2)}{dp_t^2ds_4 } \ , , \ ] ] with the boundaries @xmath71 the @xmath72 contribution to the inclusive transverse momentum distribution @xmath73 is given by @xmath74^{1/2 } } \sigma^b_{ij}(s , s_4,p_t^2 ) % \ , , \end{aligned}\ ] ] where we have inserted an extra factor of two so that @xmath75 .
after some algebra we can rewrite this result as @xmath76\,,\end{aligned}\ ] ] with the definition @xmath77^{1/2 } } \sigma_{ij}^b(s , s_4,p_t^2)\,,\ ] ] where @xmath78 again represents the born differential @xmath19 distribution . for the @xmath79 and @xmath25 subprocesses we have the explicit results @xmath80^{1/2 } }
% \nonumber \\ & & % \times \big[\frac{(s - s_4)^2 - 2sp_t^2}{s^2}\big ] \,,\end{aligned}\ ] ] and @xmath81^{1/2 } } % \nonumber \\ & & % \times\big[c_f - c_a \frac{m_t^2}{s}\big ] % \nonumber \\ & & % \times \big[\frac{(s - s_4)^2 - 2sm_t^2}{sm_t^2 } + % \frac{4m^2}{m_t^2 } \big ( 1 -\frac{m^2}{m_t^2}\big)\big ] \,.\end{aligned}\ ] ] since the above formulae are symmetric under the interchange @xmath82 the heavy quark and heavy antiquark inclusive @xmath19 distributions are identical .
note that ( 2.12 ) is basically the integral of a plus distribution together with a surface term . the corresponding formula to ( 2.12 ) for the rapidity @xmath20 of the heavy quark
is obtained by using @xmath83 @xmath84 the double differential cross section is therefore @xmath85 with the boundaries @xmath86 where @xmath87 .
the @xmath72 contribution to the inclusive rapidity distribution @xmath88 is given by @xmath89 after some algebra we can rewrite this result as @xmath90\,,\end{aligned}\ ] ] with the definition @xmath91 where @xmath92 again represents the born differential @xmath20 distribution . for the @xmath79 and @xmath25 subprocesses we have the explicit formulae @xmath93 \,,\end{aligned}\ ] ] and @xmath94 % \times \big[\cosh^2 y + \sinh^2 y % \nonumber \\ & & % + \frac{8m^2s \cosh^2 y}{(s - s_4)^2 } \big ( 1 -\frac{4m^2s\cosh^2 y}{(s - s_4)^2 } % \big)\big ] \,.\end{aligned}\ ] ] since the above formulae are symmetric under the interchange @xmath82 the heavy quark and heavy antiquark inclusive @xmath20-distributions are identical .
also ( 2.21 ) is again of the form of a plus distribution together with a surface term .
finally , we note that the terms in ( 2.12 ) and ( 2.21 ) are all finite .
the resummed contribution to the top quark cross section can be written as @xcite @xmath95 \sigma_{ij}^b(s , t_1,u_1),\ ] ] where @xmath96^{\eta}}{\gamma(1+\eta ) } % \exp(-\eta\gamma_e).\end{aligned}\ ] ] the straightforward expansion of the exponential plus the change of the argument in @xmath97 via the renormalization group equations generates the corresponding leading logarithmic terms written expicitly in @xcite .
the scheme dependent @xmath98 and @xmath97 in the above expression are given by @xmath99 in the @xmath100 scheme , and @xmath101 in the dis scheme , where @xmath102 is the lowest order coefficient of the qcd @xmath61-function .
the color factors @xmath103 are defined by @xmath104 and @xmath105 , and @xmath106 is the euler constant .
the quantity @xmath107 is given by @xmath108 following the procedure in @xcite for the resummation of the order @xmath109 contributions to the @xmath19 distribution we have @xmath110 note that we now have cut off the lower limit of the @xmath64 integration at @xmath111 because @xmath97 in ( 3.2 ) diverges as @xmath54 .
this parameter @xmath112 must satisfy the conditions @xmath113 and @xmath114 .
it is convenient to rewrite @xmath112 in terms of the scale @xmath28 as @xmath115 @xmath116 here @xmath31 is a nonperturbative parameter @xcite satisfying @xmath117 .
the derivative of @xmath118 is obtained from ( 3.2 ) .
it is equal to @xmath119^\eta}{\gamma(1+\eta ) } % \nonumber \\ & & \times \exp(-\eta\gamma_e)\ , , \end{aligned}\ ] ] where we have neglected terms which are higher order in @xmath97 .
the analogous formula for the rapidity distribution is @xmath120 with the conditions @xmath121 and @xmath114 .
following the notation in @xcite the total hadron - hadron cross section in order @xmath122 is @xmath123 where @xmath124 is the square of the hadron - hadron c.m . energy and
@xmath125 run over @xmath126 and @xmath127 .
the parton flux @xmath128 is defined via @xmath129 and @xmath130 is a product of the scale - dependent parton distribution functions @xmath131 , where @xmath132 stands for the hadron which is the source of the parton @xmath133 @xmath134 the mass factorization scale @xmath28 is chosen to be identical with the renormalization scale in the running coupling constant . since the @xmath19 distribution in hadron - hadron collisions is not altered by the lorentz transformation along the collision axis from the parton - parton c.m .
frame , we can write an analogous formula to ( 4.1 ) for the heavy - quark inclusive differential distribution in @xmath135 @xmath136 with @xmath137 . in the case of the all - order
resummed expressions the lower boundaries in ( 4.1 ) and ( 4.4 ) have to be modified according to the conditions @xmath138 or @xmath139 ( see above ) . resumming the soft gluon contributions to all orders
we obtain @xmath140 where @xmath141 is given in ( 3.24 ) of @xcite and @xmath142 with @xmath143 ( @xmath100 scheme ) or @xmath144 ( dis scheme ) [ see ( 3.7 ) ] .
the all - order resummed differential distribution in @xmath135 is @xmath145 with @xmath73 given in ( 3.6 ) and @xmath146 the corresponding formula to ( 4.4 ) for the heavy quark inclusive differential distribution in @xmath20 is @xmath147 order by order in perturbation theory the heavy quark rapidity plots in the parton - parton c.m .
frame show peaks away from @xmath148 ( see fig . 7 in @xcite ) . however , upon folding with the partonic densities the heavy quark rapidities in the hadron - hadron c.m .
frame peak near @xmath149 .
therefore we will assume that the plots for the resummed rapidity distribution show a similar feature .
the all - order resummed differential distribution in @xmath150 is therefore taken to be @xmath151 with @xmath88 given in ( 3.9 ) and @xmath152 the hadronic heavy quark rapidity @xmath150 is related to the partonic heavy quark rapidity @xmath20 by @xmath153 we now specialize to top quark production at the fermilab tevatron where @xmath154 tev and choose the top quark mass to be @xmath155 gev@xmath15 . in the presentation of our results for the exact , approximate and resummed hadronic cross sections we use the mrsd@xmath156 parametrization for the parton distributions @xcite .
note that the hadronic results only involve partonic distribution functions at moderate and large @xmath157 , where there is little difference between the various sets of parton densities .
we have used the mrsd@xmath156 set 34 as given in pdflib @xcite in the dis scheme with the number of active light flavors @xmath158 and the qcd scale @xmath159 gev .
we have used the two - loop corrected running coupling constant as given by pdflib . since we know the exact @xmath1 result
, we can make an even better estimate of the differential distributions by calculating the perturbation theory improved @xmath19 and @xmath20 distributions .
we define the improved @xmath19 distribution by @xmath160 and the improved @xmath150 distribution by @xmath161 to exploit the fact that @xmath162 and @xmath163 are known and @xmath164 and @xmath165 are included in @xmath166 and @xmath167 respectively .
we note that here @xmath168 denotes the @xmath169 contribution to the differential cross section .
moreover , @xmath170 denotes the exact calculated differential cross section , and @xmath171 the approximate one where only the leading soft gluon corrections are taken into account .
first we present the differential @xmath19 distributions at @xmath154 tev for a top quark mass @xmath172 gev@xmath15 . for these plots
the mass factorization scale is not everywhere equal to @xmath29 .
we chose @xmath173 in @xmath112 , @xmath174 and @xmath175 , but @xmath176 in the mrsd@xmath17 parton distribution functions and the running coupling constant @xmath177 .
we begin with the results for the @xmath24 channel in the dis scheme . in fig .
1 we show the born term @xmath178 , the first order exact result @xmath162 , the first order approximation @xmath164 , the second order approximation @xmath179 , and the resummed result @xmath166 for @xmath180 and for @xmath181 .
these are the same values for @xmath31 that were used in @xcite . as we decrease @xmath31 the differential cross sections increase . in fig .
2 we show the exact @xmath1 result @xmath182 , and , for comparison , @xmath183 for @xmath180 and @xmath181 .
the resummed differential cross sections were calculated with the cut @xmath184 while no such cut was imposed on the phase space for the individual terms in the perturbation series .
the improved differential distributions are uniformly above the exact @xmath1 result .
it is also evident from fig .
2 that the resummation of the soft gluon contributions to the @xmath19 distribution modifies the exact @xmath1 result only slightly for the values of @xmath31 that have been chosen .
we continue with the results for the @xmath25 channel in the @xmath100 scheme .
the corresponding plots are given in figures 3 and 4 .
in this case the values of @xmath31 have been chosen to be @xmath185 and @xmath186 to correspond to those in @xcite .
note that @xmath31 need not be the same in the @xmath24 and @xmath25 reactions because the convergence properties of the qcd perturbation series could be different in these channels and moreover depend on the factorization scheme .
the first and second order corrections in the @xmath25 channel in the @xmath100 scheme are larger than the respective ones in the @xmath24 channel in the dis scheme .
in fact , for the range of @xmath19 values shown the second - order approximate correction is larger than the first - order approximation .
hence , the relative difference in magnitude between the improved @xmath183 and the exact @xmath1 results is significantly larger than that in the @xmath24 channel in the dis scheme .
we finish our discussion of the differential @xmath19 distributions with the results of adding the @xmath24 and @xmath25 channels .
the plots appear in figures 5 and 6 .
it is evident that resummation produces an enhancement of the exact @xmath1 result , with very little change in shape .
now we turn to a discussion of the differential @xmath150 distributions at @xmath154 tev for a top quark mass @xmath172 gev@xmath15 . in this case
we set the factorization mass scale equal to @xmath29 everywhere .
we begin with the results for the @xmath24 channel in the dis scheme . in fig .
7 we show the born term @xmath187 , the first order exact result @xmath188 , the first order approximation @xmath165 , the second order approximation @xmath189 , and the resummed result @xmath167 for @xmath180 and @xmath181 . in fig .
8 we show the exact @xmath1 result @xmath190 , and , for comparison , @xmath191 for @xmath180 and @xmath181 . again , the resummed differential cross sections were calculated with the cut @xmath184 while no such cut was imposed on the phase space for the individual terms in the perturbation series .
it is also evident from fig . 8
that the resummation of the soft gluon contributions to the @xmath150 distribution modifies the exact @xmath1 result only slightly for the values of @xmath31 that have been chosen .
we continue with the results for the @xmath25 channel in the @xmath100 scheme .
the corresponding plots are given in figures 9 and 10 . here
, the values of @xmath31 are @xmath185 and @xmath186 . as in the case of the @xmath19 distributions ,
the first and second order corrections in this channel are larger than the respective ones in the @xmath24 channel in the dis scheme . for the range of @xmath150 values
shown the second - order approximate correction is larger than the first - order approximation .
again , as in the @xmath19 distributions , the relative difference in magnitude between the improved @xmath191 and the exact @xmath1 results is significantly larger than that in the @xmath24 channel in the dis scheme . finally , we conclude our discussion of the differential @xmath150 distributions by showing the results of adding the @xmath24 and @xmath25 channels .
the plots appear in figures 11 and 12 .
again , it is evident that resummation produces a non - negligible modification of the exact @xmath1 result .
however , the shape of the distribution is unchanged .
we have shown that the resummation of soft gluon radiation produces a small difference between the perturbation improved distributions in @xmath19 and @xmath150 and the exact @xmath1 distributions in @xmath19 and @xmath150 for the @xmath24 reaction in the dis scheme for the values of @xmath31 chosen . however , for the @xmath25 channel in the @xmath100 scheme the resummation produces a large difference .
the difference between the perturbation improved and the exact @xmath1 distributions depends on the mass factorization scheme ( dis or @xmath100 ) , the factorization scale @xmath28 , as well as the specific reaction under consideration ( @xmath24 or @xmath25 ) . for a mass @xmath155 gev@xmath15 the @xmath25 channel is not as important numerically as the @xmath24 channel .
however , since the corrections for the @xmath25 channel are quite large , resummation produces a non - negligible difference between the perturbation improved and the exact @xmath1 distributions when adding the two channels .
however , the shapes of the distributions are essentially unchanged . 99 ( cdf collaboration ) f. abe et al .
d * 50 * , 2966 ( 1994 ) .
( d0 collaboration ) s. abachi et al .
, fermilab - pub-94 - 354-e , 1994 . l. h. orr and w. j. stirling
, dtp-94 - 60 , 1994 . l. h. orr , t. stelzer , and w. j. stirling , dtp-94 - 112 , 1994 .
f. a. berends , j. b. tausk , and w. t. giele , phys .
d * 47 * , 2746 ( 1993 ) .
r. h. dalitz and g. r. goldstein , phys .
d * 45 * , 1531 ( 1992 ) .
p. nason , s. dawson , and r. k. ellis , nucl . phys .
* b303 * , 607 ( 1988 ) . w. beenakker , h. kuijf , w. l. van neerven , and j. smith , phys . rev .
d * 40 * , 54 ( 1989 ) .
e. laenen , j. smith , and w. l. van neerven , phys .
b * 321 * , 254 ( 1994 ) .
r. k. ellis , talk presented at the glasgow conference , july 1994 .
a. d. martin , r. g. roberts , and w. j. stirling , phys .
b * 306 * , 145 ( 1993 ) .
p. nason , s. dawson , and r. k. ellis , nucl .
* b327 * , 49 ( 1989 ) ; * b335 * , 260(e ) ( 1990 ) .
w. beenakker , w. l. van neerven , r. meng , g. a. schuler , and j. smith , nucl . phys . *
b351 * , 507 ( 1991 ) .
e. laenen , j. smith , and w. l. van neerven , nucl . phys . *
b369 * , 543 ( 1992 ) .
k. lane , buhep-95 - 2 , 1995 .
r. meng , g. a. schuler , j. smith , and w. l. van neerven , nucl .
b339 * , 325 ( 1990 ) .
g. sterman , nucl .
b281 * , 310 ( 1987 ) ; d. appel , p. mackenzie , and g. sterman , _ ibid .
_ * b309 * , 259 ( 1988 ) ; s. catani and l. trentadue , _ ibid .
_ * b327 * , 323 ( 1989 ) ; * b353 * , 183 ( 1991 ) ; h. contopanagos and g. sterman , _ ibid . _ * b400 * , 211 ( 1993 ) . a. h. mueller , nucl . phys . *
b250 * , 327 ( 1985 ) ; m. beneke , phys . lett .
b * 307 * , 154 ( 1993 ) ; g. grunberg , cpth - a328 - 1094 , 1994 ; yu . a. simonov , pisma zh .
. fiz . * 57 * , 513 ( 1993 ) [ jetp lett . * 57 * , 525 ( 1993 ) ] .
h. contopanagos and g. sterman , nucl . phys . * b419 * , 77 ( 1994 ) .
l. alvero and h. contopanagos , itp - sb-94 - 33 , 1994 ; h. contopanagos , l. alvero , and g. sterman , itp - sb-94 - 40 , 1994 .
r. hamberg , w. l. van neerven , and t. matsuura , nucl .
* b359 * , 343 ( 1991 ) ; w. l. van neerven and e. b. zijlstra , nucl . phys .
* b382 * , 11 ( 1992 ) ; p. j. rijken and w. l. van neerven , inlo - pub-14 - 94 , 1994 .
h. plothow - besch , _ pdflib : nucleon , pion and photon parton density functions and @xmath32 calculations , _ users s manual - version 4.16 , w5051 pdflib , 1994.01.11 , cern - ppe . | the inclusive transverse momentum and rapidity distributions for top quark production at the fermilab tevatron are presented both in order @xmath0 in qcd and using the resummation of the leading soft gluon corrections in all orders of qcd perturbation theory .
the resummed results are uniformly larger than the @xmath1 results for both distributions . |
the virgo cluster , at a distance of approximately @xmath1 mpc , is the nearest richly - populated cluster of galaxies and , consequently , one of the best studied .
a number of authors have pointed out that virgo s brightest elliptical galaxies have a remarkably linear arrangement , along a projected position angle of roughly @xmath2 ( measured north through east ) .
@xcite , for example , noted that `` all the e galaxies in the northern half of the virgo cluster fall on a line going through m87 . ''
similarly , @xcite suggested that `` the line connecting m87 and m84 appears as a fundamental axis of the cluster . ''
this can be seen in figure 1 , which plots the distribution of probable member galaxies in the northern portion of the virgo cluster , as seen on the plane of the sky .
the distribution of virgo dwarf elliptical galaxies also appears somewhat elongated in this direction @xcite , as does the distribution of hot x - ray emitting intracluster gas @xcite .
however , without accurate distances to individual galaxies , it is impossible to say for certain whether virgo s apparent principal axis is a genuine three - dimensional structure , or merely an illusory chance alignment of galaxies
. furthermore , three - dimensional information would allow one to measure the true shape and spatial orientation of this axis . here
we present evidence that virgo s bright elliptical galaxies trace a highly elongated , three - dimensional structure that is actually a small segment of a much larger filament passing through the heart of the virgo cluster .
a variety of methods exist for measuring galaxy distances ( see jacoby et al .
1992 for a review ) .
one of the most powerful is the technique of surface brightness fluctuations ( hereafter sbf ) , whereby distances to galaxies are estimated from the ratio of the second and first moments of their stellar luminosity functions ( tonry & schneider 1988 ; see blakeslee , ajar & tonry 1999 for a recent review ) .
this method works best for early - type galaxies , although it has also been applied to the bulges of some spirals . @xcite
have concluded that sbf is the most accurate early - type galaxy distance indicator reaching cosmologically interesting distances .
@xcite have recently published sbf distances for 300 nearby galaxies that were observed as part of their @xmath3-band sbf survey . based on distances for 31 probable virgo cluster members , they derived a mean cluster distance modulus @xmath4 , corresponding to a distance of @xmath5 mpc .
uncertainties in individual virgo galaxy distances are @xmath6 mpc , which suggests that it should be possible to resolve the virgo cluster along the line of sight , at least partially . indeed , tonry , ajhar , luppino ( 1990 ) attempted to do this a decade ago , but they lacked a suitable calibration .
they assumed that the @xmath3-band sbf magnitude @xmath7 was insensitive to stellar population variations among ellipticals , as implied by the best population models at the time . as a result
, they essentially found that bluer virgo ellipticals were systematically in front of redder ones .
following a decade of work , the stellar population dependence of @xmath7 has been empirically well characterized and calibrated out @xcite , so that there is no longer any correlation of the distances with stellar population parameters . moreover
, the empirical calibration now has strong support from theoretical modeling @xcite .
table 1 lists data for all elliptical galaxies in the northern portion of the virgo cluster with sbf distances from the @xcite survey .
column ( 1 ) gives the galaxy name , column ( 2 ) lists the sbf distance of each galaxy from @xcite , and column ( 3 ) gives the orientation of the galaxy major axis , taken from the lyon - meudon extragalactic database ( leda ) .
the positions and distance moduli of these galaxies are also indicated in figure 2 .
table 1 and figure 2 suggest some tendency for those galaxies located in the western region of the cluster to be more distant than those on the eastern side .
this trend can be seen more clearly in figure 3 , which shows a clear correlation between galaxy right ascension and distance , with a systematic trend of increasing galaxy distance as one moves in the westerly direction .
pearson ( parametric ) and spearman ( non - parametric ) rank correlation tests both confirm that this trend is statistically significant , at the 97% and 99% confidence levels , respectively ( the correlation becomes even stronger if ngc 4168 is also included ) .
we note that neilsen & tsvetanov ( 2000 ) have recently measured independent sbf distances with the _ hubble space telescope _ for 15 virgo galaxies , including 10 ellipticals along this central ridge line , and have observed the same trend in their data . to measure the three - dimensional shape and orientation of this axis , we computed the moments of inertia of the system of galaxies in table 1 after converting their right ascensions , declinations and distances to supergalactic cartesian coordinates .
diagonalization of the inertia tensor yields the three eigenvalues corresponding to the principal moments of inertia , and the associated eigenvectors provide information on the orientation of the principal axis .
we find that the virgo ellipticals have a remarkably collinear arrangement in three - dimensions , with an _ rms _ scatter of only @xmath8 kpc about the principal inertial axis along its @xmath9 mpc length between ngc 4387 and ngc 4660 .
this axis is inclined at an angle of approximately @xmath10 with respect to the plane of the sky , i.e. , close to the line of sight . presumably with smaller distance uncertainities , virgo s principal axis
would be found to be even narrower .
additional evidence of the special nature of the axis defined by virgo s brightest ellipticals comes from the orientations of the galaxies themselves .
as table 1 shows , the majority have projected major axis position angles between @xmath11 and @xmath12 , quite similar to the @xmath2 projected orientation of the cluster principal axis . a kolmogorov - smirnov ( ks )
test indicates a probability of only @xmath13 that the galaxy position angles in table 1 are randomly - oriented between 0 and 180 degrees . while one must be cautious not to overinterpret the statistical significance of results based on a small sample like this , nevertheless it is suggestive that the orientations of virgo s large ellipticals may be somehow related to the direction of the cluster principal axis .
a similar alignment effect is seen for the brightest elliptical galaxies in the coma cluster @xcite , as well as other clusters @xcite and references therein . to test whether fainter virgo ellipticals might also have preferred orientations with respect to the cluster major axis , we used the virgo photometry catalogue ( vpc ) of @xcite , which provides data , including orientations , for 1180 galaxies in the core region of the cluster . because the vpc includes both virgo members and unrelated galaxies along the line of sight , we cross - correlated it with the list of probable virgo members from @xcite to produce a sample of 108 virgo ellipticals with measured major axis position angles .
application of the ks test to this sample shows no statistically significant tendency for the fainter ellipticals to have preferred orientations .
however , when the roundest galaxies @xmath14 whose position angles are most uncertain @xmath14 are eliminated by restricting the sample to only those with ellipticities greater than 0.2 , then the ks test indicates that this subset of 69 galaxies has only a @xmath15 probability of being consistent with a randomly orientated population , with a median galaxy position angle of @xmath16 .
one of the most striking features of the large - scale distribution of galaxies is its filamentary appearance , with long , quasi - linear arrangements of galaxies that extend tens or perhaps even hundreds of mpc in length . given the linear arrangement of galaxies along virgo s principal axis , it is natural to ask whether this might be related to filamentary features on larger scales . as viewed on the plane of the sky , virgo s principal axis points in the direction of abell 1367 , a rich cluster located some @xmath17 mpc away along a projected position angle of @xmath18 .
a1367 itself forms part of a well - known supercluster with the coma cluster @xcite .
this raises the intriguing possibility that the virgo , a1367 and coma clusters may all be members of a common filamentary network , an idea suggested two decades ago by @xcite .
figure 4 plots the distribution of nearby poor galaxy clusters from the catalog of @xcite .
a narrow bridge of material is clearly seen connecting virgo and a1367 .
furthermore , the chain of giant elliptical galaxies that defines virgo s principal axis appears to be a segment of this filament .
hence , the virgo cluster points towards a1367 , not only in two dimensions , but in three . additional indirect evidence of a virgo - a1367 connection comes from a1367 s orientation ; x - ray observations show that this cluster is very elongated along a projected position angle of @xmath19 @xcite , and thus in the general direction of the virgo cluster . of course , the virgo cluster is often considered part of the larger supercluster which includes hydra - centaurus and pavo - indus ( e.g. , tully 1986 ) .
@xcite viewed this extended , planar supercluster complex as centered on the `` great attractor , '' with virgo near the outskirts .
thus , we can now trace an apparent link between two of the most massive structures in the local universe : from the `` great wall '' encompassing the coma cluster to the great attractor , through a1367 and the virgo cluster .
we have shown that the brightest elliptical galaxies in the virgo cluster have a remarkably collinear arrangement in three dimensions .
this axis appears to be part of a larger filament connecting the virgo cluster to abell 1367 .
virgo s elliptical galaxies also exhibit a tendency for their major axes to share this same orientation .
cosmological n - body simulations show that clusters of galaxies often form at the intersection of filaments , with material flowing into the cluster along one or more axes @xcite .
built by a series of subcluster mergers that occur along preferred directions , clusters naturally develop major axis orientations that reflect the orientation of the dominant filament feeding them @xcite . furthermore , if large elliptical galaxies are products of galaxy mergers , then the highly anisotropic nature of the merger process will tend to produce ellipticals whose major axis orientations are also aligned with the surrounding filamentary structure @xcite .
the results presented here are consistent with a picture in which the formation of the virgo cluster and its elliptical galaxy population has been driven by anisotropic inflow of material along the virgo - a1367 filament . although virgo may be fed by more than one filament @xcite , the one joining the cluster to a1367 appears to dominate .
additional evidence in support of this interpretation comes from x - ray observations of m86 , which show a plume of hot gas being stripped from this galaxy along a projected position angle of @xmath20 @xcite , presumably as a result of ram pressure as the galaxy travels along the cluster principal axis at over 1200 km@xmath21s . finally , it is intriguing that m87 s famous jet also emanates along the same direction as the virgo cluster major axis .
this is true not only in two dimensions , where both are oriented along projected position angles of @xmath22 , but also in three dimensions .
detailed models of the jet s observed properties indicate that it is most likely oriented within @xmath23 degrees of the line of sight @xcite , very close to the @xmath24 to @xmath25 line of sight inclination angle of the cluster principal axis found in 2 . while this might be purely coincidental , alternatively
, it might be an indication that virgo s principal axis has influenced not only the orientations of its member elliptical galaxies , but perhaps even the massive blackhole at the center of m87 that is believed to power its nuclear activity @xcite .
arp , h. 1968 , , 80 , 129 binggeli , b. 1982 , , 107 , 338 binggeli , b. , sandage , a. , & tammann , g.a .
1985 , , 90 , 1681 binggeli , b. , tammann , g.a . , &
sandage , a. 1987 , , 94 , 251 binggeli , b. , popescu , c.c . , & tammann , g.a .
1993 , , 98 , 275 binggeli , b. 1999 , in the radio galaxy messier 87 , ringberg workshop , eds .
rser & k. meisenheimer , springer , berlin , 9 biretta , j.a . , sparks , w.b . , &
macchetto , f. 1999 , , 520 , 621 blakeslee , j.p . ,
ajhar , e.a . , & tonry , j.l .
1999 , in post - hipparcos cosmic candles , eds .
a. heck & f. capulo ( boston : kluwer ) , 181 blakeslee , j.p . ,
vazdekis , a. , & ajhar , e.a .
2000 , , in press bhringer , h. , briel , u.g . , schwarz , r.a . , voges , w. , hartner , g. & trmper , j. 1994 , , 368 , 828 bond , j.r . ,
kofman , l. & pogosyan , d. 1996 , , 380 , 603 carollo , c.m . ,
franx , m. , illingworth , g.d . , & forbes , d.a .
1997 , , 481 , 710 de lapparent , v. , geller , m.j .
, & huchra , j.p .
1986 , , 302 , l1 dubinski , j. 1998 , , 502 , 141 ferrarese , l. , et al .
2000 , , 529 , 745 forman , w. , schwarz , j. , jones , c. , liller , w. , & fabian , a.c .
1979 , , 234 , l27 heinz , s. & begelman , m.c .
1997 , , 490 , 653 jacoby , g.h . , et al .
1992 , , 104 , 599 jones .
c. & forman , w. 1999 , 511 , 65 king , i.r .
1978 , , 222 , 1 liu , m. c. , charlot , s. , & graham , j. r. 2000 , , in press lynden - bell , d. , faber , s. m. , burstein , d. , davies , r. l. , dressler , a. , terlevich , r. j. , & wegner , g. 1988 , , 326 , 19 neilsen , e.h .
tsvetanov , z.i .
2000 , , 536 , 255 porter , a.c . ,
schneider , d.p . , & hoessel , j.g .
1991 , , 101 , 1561 schindler , s. , binggeli , b. , & bhringer , h. 1999 , , 343 , 420 gregory , s.a . , & thompson , l.a .
1978 , , 222 , 784 tonry , j.l . , ajhar , e.a . , & luppino , g.a .
1990 , , 100 , 1416 tonry , j.l . ,
blakeslee , j.p . ,
ajhar , e.a . , & dressler , a. 1997 , , 475 , 399 tonry , j.l . ,
dressler , a. , blakeslee , j.p . , ajhar , e.a . , fletcher , a.b . ,
luppino , g.a . ,
metzger , m.r . , & moore , c.p .
2000 , , in press tonry , j. , & schneider , d.p .
1988 , , 96 , 807 tully , r.b .
1982 , , 257 , 389 tully , r.b . 2986 , , 303 , 25 van haarlem , m. & van de weygaert , r. 1993 , , 418 , 544 west , m.j .
1994 , , 268 , 79 west , m.j .
1998 , a new vision of an old cluster : untangling coma berenices , a. mazure et al . ,
singapore : world scientific , 36 west , m.j . ,
jones , c. , & forman , w. 1995 , , 451 , l5 weil , m. , bland - hawthorn , j. , & malin , d.f .
1997 , , 490 , 664 white , r.a . ,
et al . 1999 , , 118 , 2014 worthey , g. 1994 , , 95 , 107 young , c.k . ,
& currie , m.j .
1998 , , 127 , 367 zeldovich , ya.b . ,
einasto , j. , & shandarin , s.f .
1982 , , 300 , 407 lcr ngc 4374 ( m84 ) & @xmath26 & 135 + ngc 4387 & @xmath27 & 140 + ngc 4406 ( m86 ) & @xmath28 & 130 + ngc 4458 & @xmath29 & 46 + ngc 4473 & @xmath30 & 100 + ngc 4478 & @xmath31 & 140 + ngc 4486 ( m87 ) & @xmath32 & 160 + ngc 4551 & @xmath33 & 70 + ngc 4552 ( m89 ) & @xmath34 & 125 + ngc 4564 & @xmath35 & 47 + ngc 4621 ( m59 ) & @xmath36 & 165 + ngc 4649 ( m60 ) & @xmath37 & 105 + ngc 4660 & @xmath38 & 100 + ngc 4733 & @xmath39 & 115 + | using accurate distances to individual virgo cluster galaxies obtained by the method of surface brightness fluctuations , we show that virgo s brightest ellipticals have a remarkably collinear arrangement in three dimensions .
this axis , which is inclined by @xmath0 from the line of sight , can be traced to even larger scales where it appears to join a filamentary bridge of galaxies connecting virgo to the rich cluster abell 1367 .
the orientations of individual virgo ellipticals also show some tendency to be aligned with the cluster axis , as does the jet of the supergiant elliptical m87 .
these results suggest that the formation of the virgo cluster , and its brightest member galaxies , have been driven by infall of material along the virgo - a1367 filament . |
the solar neutrino problem , the fact that the detected neutrino flux from the sun is less than the predicted flux , has been known for decades since the pioneering work of r. davis in the homestake mine @xcite . since then , substantial progress has been made .
the kamiokande collaboration confirmed that the neutrinos are indeed coming from the sun in a real - time experiment with directional capability @xcite .
both the kamiokande and super - kamiokande collaborations @xcite also reported a depletion of the predicted flux .
the gallex and sage experiments , which are sensitive to the ( dominant ) @xmath5 component of the solar neutrino flux @xcite , directly related to the solar luminosity , also found a depletion of the predicted flux . without relying on the standard solar model calculations
, one can conclude from the data that the electron neutrino flux from the @xmath6 is almost totally depleted ( see , _
e.g. _ , @xcite ) .
furthermore , the credibility of the standard solar model calculations has been verified by their agreement with the helioseismology data at better than one percent level @xcite .
these facts amount to strong evidence of new physics in the neutrino sector , in particular neutrino oscillations . even though the evidence for a `` real '' ( solar model independent ) solar neutrino problem is very strong ,
it is not yet completely established .
first , one needs to rely on ( at least ) two experiments to conclude that there is , model independently , a problem
. it would be far more convincing if one could see a signal of neutrino oscillations in a single experiment .
second , all of the experiments have been of the disappearance type , where one sees a depletion of the predicted flux . given the difficulty of neutrino experiments and of theoretical calculations of nuclear cross sections , an appearance experiment would be much more convincing evidence of neutrino oscillations .
the sno experiment @xcite will go a long way towards resolving the issues raised above .
it is designed to measure the solar @xmath7b neutrino flux via the charged - current ( cc ) reaction ( @xmath8 ) and the neutral - current ( nc ) reaction ( @xmath9 , @xmath10 ) . assuming both of these processes can be well understood
, a difference between the two measured fluxes would imply that there are neutrinos in the solar neutrino flux which are not of the electron type ; one may even call this an appearance experiment of @xmath3 .
there is also an additional oscillation signature in the possible distortion of the neutrino energy spectrum .
however , if for some astrophysical and/or nuclear - physics reason the @xmath11b neutrino flux is lower than currently predicted , the sno experiment may be unable to see an oscillation signature .
another possible concern is that the measurement of the cc / nc ratio involves the separate calibration of the efficiencies in the cc and nc processes . on the other hand ,
if the current data are correct and the solar neutrinos indeed oscillate ( even with an arbitrary @xmath7b flux ) , there must be neutrinos other than @xmath2 in the @xmath12be neutrino flux , and their detection would be an unambiguous signal of neutrino oscillations . the @xmath12be neutrinos will be studied using @xmath0-@xmath1 elastic scattering at borexino @xcite , and possibly also at kamland @xcite , if the background from natural radioactivity can be sufficiently suppressed . in this letter , we study the prospect of establishing the @xmath3 component of the solar neutrino flux in a completely solar model - independent analysis .
we point out that the recoil electron kinetic energy spectrum is different for @xmath2 and @xmath3 . by fitting the shape of the electron energy spectrum , one can determine the fraction of @xmath3 in the solar neutrino flux , without relying on the predicted neutrino flux from the standard solar model .
we discuss both the @xmath12be neutrinos at borexino or kamland and the @xmath11b neutrinos at super - kamiokande or sno .
this type of model - independent study seems to be difficult with the @xmath11b neutrinos at super - kamiokande or sno , but borexino or kamland should have enough statistics to do the analysis with the @xmath12be neutrinos . the sensitivity to the @xmath3 component is a strong function of the @xmath2 survival probability . in the parameter range of the small - angle msw solution , one can see the @xmath3 component of the @xmath12be neutrino flux at more than 95% confidence level with two years of borexino running , if the background is sufficiently small . under the same conditions
, the sensitivity at kamland would be even greater .
the differential cross - section for elastic @xmath13-@xmath1 scattering ( @xmath10 ) is known @xcite : @xmath14,\ ] ] where @xmath15 , @xmath16 and @xmath17 .
@xmath18 is the recoil electron kinetic energy and @xmath19 is the incoming neutrino energy in the lab frame . from the kinematics ,
@xmath20 is related to the recoil electron scattering angle @xmath21 by @xmath22 and ranges from @xmath23 to @xmath24 .
the sign in the definition of @xmath25 depends on the flavor of the incoming neutrino : it is plus for @xmath26 and minus for @xmath27 ( @xmath28 for active ) . in the presence of oscillations ,
the @xmath20 distribution is @xmath29 where @xmath30 is the oscillation probability for @xmath31 .
note that @xmath32 for @xmath33 . to illustrate the difference in the recoil electron kinetic energy spectra for different incoming neutrinos , we plot in fig .
[ spectra ] spectra for two neutrino energies , @xmath34 mev ( for @xmath11b neutrinos ) and @xmath35 mev ( for @xmath12be neutrinos ) .
the curves are all normalized to unit area such that their shapes can be compared .
the @xmath36 vs @xmath37 difference is more prominent at higher energies , but is not negligible even for the @xmath12be energy .
the central idea of this letter is the following .
one should fit the recoil electron kinetic energy spectrum with an _ arbitrary normalization , _ both for @xmath36 and @xmath37 .
the presence of a non - zero component of @xmath38-@xmath1 scattering is the evidence of neutrino oscillations .
this test does not depend on the theoretical prediction of the neutrino flux , and hence is independent of solar model and nuclear physics calculations .
it can be regarded as an `` appearance '' experiment of @xmath3 , similar to the sno experiment .
the rest of this letter is devoted to discussing under what conditions such a test can be performed .
we will analyze the recoil electron kinetic energy distributions for the case of solar neutrinos produced by the electron capture reaction @xmath4be @xmath39li @xmath40 . because of the @xmath41 kinematics the neutrinos are mono - energetic , which greatly simplifies our analyses.li nucleus , namely @xmath42 mev ( branching fractions @xmath43 vs @xmath44 ) .
we focus only on the higher energy value because the lower energy one does not produce recoil electron energies above the borexino threshold of 250 kev . ]
we mostly focus on borexino , because it is the only approved experiment which will specifically study the @xmath12be neutrinos .
we do comment on a possible statistically superior sample from kamland .
we find that borexino can in principle show the existence of a @xmath38 component in the @xmath12be solar neutrino flux at the two - sigma level after two years of running , if the background is negligible .
following the idea presented in the previous section , we will not rely on the overall rate of the scattering process , which depends on the theoretical prediction of the flux . to be completely model - independent ,
we use only the shape of the recoil electron kinetic energy spectrum by allowing the normalization to float in the fit . when discussing the sensitivity of borexino or kamland , however
, we do need to use some expected neutrino flux ; for this purpose , we use the standard solar model ( bp95 ) @xcite , plus the effect of neutrino oscillations .
the simulated `` data '' sample will consist of ten @xmath20 bins , with the number of events in the @xmath45-th bin given by @xmath46 where @xmath47 and @xmath48 , for @xmath49 .
we take the detection threshold energy to be 250 kev ( _ i.e. _ , @xmath50 ) for borexino , which is limited by the @xmath51c radioactivity background .
note that for the borexino @xmath20 range , @xmath52 .
@xmath53 is the number of events predicted by the standard solar model for borexino . in the upcoming analysis
, we will only consider statistical uncertainties , and no background .
a two - parameter @xmath54 fit of the `` data '' events was performed , by varying both @xmath55 and @xmath30 ( two parameters ) .
this is equivalent to fitting the data to a linear combination of the @xmath2-@xmath1 and @xmath38-@xmath1 differential scattering cross sections with arbitrary normalizations ( two parameters ) .
[ fit](a ) shows the extracted @xmath56 as a function of @xmath57 for two years of borexino running .
a nonzero value of @xmath56 implies the presence of @xmath31 oscillations .
the analysis indicates that , for two years of borexino running , the active neutrino component can be seen at the one - sigma level if @xmath58 .
for @xmath59 active neutrino oscillations would yield more than a two - sigma effect .
that is the case for the so - called small angle msw solution , which predicts @xmath60 . on the other hand ,
the so - called large angle msw solution predicts @xmath61 , and the vacuum oscillations ( the `` just - so '' solution ) @xmath62 @xcite .
a different type of analysis can be performed , with very similar results .
this different analysis might prove to be useful in order to deal with the background , if it is not negligible .
the integrated observable @xmath63 is defined by @xmath64 @xmath65 is the number of observed events @xmath66 , and the sub(super)script @xmath67 refers to the degree of the polynomial multiplying the data . in the absence of active neutrinos in the solar flux , @xmath68 .
note that @xmath63 is defined in such a way that the contribution of any background with a flat @xmath20 distribution cancels . in fig .
[ deltaa](a ) we plot @xmath63 as a function of @xmath57 , for the same conditions considered in the two - parameter fit .
the results are very similar to the ones obtained earlier , as expected .
even though the borexino experiment should have enough statistics for a model - independent test of the @xmath38 component in the solar @xmath12be flux , the experimental effort will still be very challenging .
the main concern is radioactive background from rn , u and th .
an accurate energy calibration is also crucial .
our simple analysis is valid only when the background is sufficiently small in the signal range .
if the background turns out to be significant , one can still use the recoil electron kinetic energy spectrum if ( 1 ) the background can be reliably subtracted and ( 2 ) the statistical significance can be kept after the background subtraction .
the first assumption is rather difficult to justify .
the counting test facility at borexino demonstrated that the background can be suppressed down to an extremely low level @xcite , but it was not possible to prove that it can be suppressed to the required level because the background was so low that it could not be studied ! even if the required level is achieved with the full - scale detector , understanding the energy spectrum of the background would require a challenging calibration procedure .
the validity of the second assumption , of course , depends on the level of the background .
it would be extremely valuable if kamland could also achieve the radio - purity planned for borexino , so that it can also study the recoil electron energy spectrum from the @xmath12be solar neutrinos , but with a larger fiducial volume . for comparison , the same plots as figs .
[ fit](a ) and [ deltaa](a ) are shown in figs .
[ fit](b ) and [ deltaa](b ) , for two years of kamland running .
we assume the bp95 estimate of 466 kamland events per day for a 1 kt fiducial volume .
the difference in the recoil electron kinetic energy spectra between incident @xmath2 and @xmath38 is more prominent for @xmath11b neutrinos than for lower energy neutrinos such as the @xmath12be neutrinos ( see fig .
[ spectra ] ) .
the main complication with the @xmath11b neutrinos is that , unlike the @xmath12be neutrinos , they have a continuous spectrum .
the super - kamiokande experiment has measured the recoil electron energy spectrum from @xmath69-@xmath1 elastic scattering @xcite , which is a convolution of the neutrino energy spectrum and the @xmath20 distribution discussed in section [ sec : spectrum ] .
if the spectrum is not consistent with expectations , it indicates either that ( 1 ) the neutrino energy spectrum is not the expected one , possibly due to unknown nuclear - physics uncertainties in the @xmath11b beta spectrum ( see , however , @xcite ) , ( 2 ) the neutrino energy spectrum is distorted due to an energy dependent neutrino oscillation , ( 3 ) there is some fraction of @xmath3 in the flux , which yields a different @xmath20 distribution , or ( 4 ) a combination of them .
the aim of this letter is to identify the possibility ( 3 ) .
the identification of ( 3 ) is , in principle , possible .
if one measures both the electron recoil energy _ and
_ the recoil angle ( which is an observable because we know the direction of the sun at the time of the event in a real - time experiment ) it is easy to solve the kinematics and calculate both the incident neutrino energy @xmath70 and @xmath20
. then one can select events with some specific value of @xmath70 and study the @xmath20 distribution .
this program , unfortunately , can not be done at super - kamiokande .
the main reason is that the recoil angle distribution to too forward - peaked , @xmath71 from eq .
[ angle ] , while the angular resolution is 25@xmath72 to 35@xmath72 in the relevant energy range @xcite .
the strong forward peak happens because of the high energy threshold for the recoil electron .
large time projection chamber ( tpc ) experiments , such as icarus @xcite or hellaz @xcite might have enough angular and recoil energy resolution to attempt such a program ; indeed , hellaz quotes a 35mrad ( @xmath73 ) angular resolution and a @xmath74 @xmath75 resolution , which is enough for our purposes . however , their statistics is very limited ( @xmath76 events per day ) and a positive result would require too long a running time .
sno studies the recoil electron energy from the charged current reaction @xmath77 , where the energy of the electron is approximately @xmath78 , where @xmath79 mev is the deuteron binding energy , when the kinetic energy of the recoil protons is neglected .
the measurement of this recoil electron energy spectrum does not reflect the @xmath20 distribution discussed in section [ sec : spectrum ] , but rather the neutrino energy spectrum .
this is , of course , a very valuable information in order to study the distortion of the neutrino energy spectrum due to oscillations .
this is , however , not the effect we wished to study in this letter . in principle
, one can also try to deconvolute the recoil electron energy spectrum at super - kamiokande using the measured neutrino energy spectrum from sno and then determine the presence of a @xmath38 component in the @xmath7b flux via the methods presented in the previous sections . as a matter of fact
, the sno experiment itself could also use the elastic scattering part of its signal to do this analysis . in principle
, sno could establish active neutrino oscillations even without its neutron capture capabilities .
this would , however , require a large elastic scattering sample and hence a very long running time .
it seems promising to try to establish neutrino oscillations by analyzing the recoil electron kinetic energy spectrum in the case of @xmath4be neutrinos .
in particular we have shown that , in the case of negligible background , two years of borexino running should be enough to determine the presence of a @xmath3 component in the solar neutrino flux model - independently if @xmath80 . under the same conditions , kamland is capable of obtaining even more significant results .
we emphasize that this effect is unrelated to the distortion of the incident _ neutrino _ energy spectrum , which has been thoroughly discussed in the literature .
it is certainly not clear that the background will be negligible .
unfortunately we can not simulate its effects clearly .
instead , we chose to define two different methods of establishing active neutrino oscillations .
we believe that the background will behave differently under the two methods , and therefore be more readily extracted .
another crucial issue is , of course , the energy calibration .
it is clear that a more thorough analysis can only be performed by detailed simulations of the detectors in questions ( and by the experiments themselves ! ) , which is beyond the scope of our letter . finally , the situation with the @xmath7b neutrinos is much less clear , in part due to their continuous energy spectrum .
it is hard to disentangle distortions in the neutrino energy spectrum , possibly due to oscillations , from changes in the recoil electron energy spectrum due to a @xmath3 component in the solar flux .
the tpc appears to be the right technology for this purpose , even though the currently proposed tpc - based experiments , icarus and hellaz , will not have enough statistics .
hm thanks lawrence hall , kevin lesko , and stuart freedman for useful conversations .
we also thank stuart freedman for comments on the manuscript . this work was supported in part by the u.s .
department of energy under contracts de - ac03 - 76sf00098 and in part by the national science foundation under grant phy-95 - 14797 .
hm was also supported by alfred p. sloan foundation , and adg by cnpq ( brazil ) .
gallex collaboration , _ phys .
* b388 * , 384 ( 1996 ) ; + t. kirsten , talk presented at neutrino 98 conference , june 49 , 1998 , takayama , japan ( http://www-sk.icrr.u-tokyo.ac.jp/nu98/scan/023 ) ; + v.n .
gavrin , talk presented at neutrino 98 conference , june 49 , 1998 , takayama , japan ( http://www-sk.icrr.u-tokyo.ac.jp/nu98/scan/022 ) . | we point out that the recoil electron kinetic energy spectra in the @xmath0-@xmath1 elastic scattering are different for incident @xmath2 or @xmath3 , and hence one can in principle establish the existence of the @xmath3 component in the solar neutrino flux by fitting the shape of the spectrum .
this would be a new model - independent test of the solar neutrino oscillation in a single experiment , free from astrophysical and nuclear physics uncertainties . for the @xmath4be neutrinos , it is possible to determine the @xmath3 component at borexino or kamland , if the background is sufficiently low .
note that this effect is different from the distortion in the incident _ neutrino _ energy spectrum , which has been discussed in the literature .
lbnl-42593 + ucb - pth-98/61 + hep - ph/9812307 + .1 in * establishing a @xmath3 component + in the solar neutrino flux * 0.3 in andr de gouva and hitoshi murayama 0.1 in _ department of physics , university of california + berkeley , ca 94720 _ + and 0.05 in _ theoretical physics group , lawrence berkeley national laboratory + berkeley , ca 94720 _ 0.05 in .1 in |
principal component analysis ( pca ) is a well - known technique initially designed to reduce the dimensionality of a typically huge data set while keeping most of its variance @xcite .
pca is intimately related to the singular value decomposition ( svd ) since the principal components of a data set , whose arithmetic mean is zero , will be equal to the eigenvectors of the covariance matrix sorted by their corresponding eigenvalue ; or equivalently by the variance they account for .
the principal coefficients are the linear coefficients allowing us to reconstruct the initial data set based on the principal components .
further details about pca will be given in section [ sec : pca ] of this paper .
interested readers are also invited to read @xcite for an accessible tutorial on this technique or @xcite for a deeper analysis .
pca has many applications in a wide variety of astronomical domains from the classification of the sloan digital sky survey ( sdss ) quasar spectra and their redshift determination @xcite to the study of the point spread function variation in lensing surveys @xcite .
the method described hereafter was originally developed in the framework of the _ gaia _ astrophysical parameters inference system @xcite where it is used to provide learning data sets of spectrophotometric data based on sdss quasar catalog spectra @xcite .
the latter cover the observed wavelength range 400010 000 and are extrapolated by our algorithm to the wavelength range 300011 000 covered by _
gaia_. even if developed for an astronomical purpose , it can be used in any problems requiring pca decomposition of weighted data .
the case of missing data being simply the limiting case of weights equal to zero .
classical pca is a mature tool whose performance in dimensionality reduction and pattern recognition has been assessed for a long time .
nevertheless , its main limitation comes from the fact that it is not adapted to the case of samples having weighted and/or missing data .
the inherent consequence is that the classical pca implementations made no difference between variance coming from a genuine underlying signal and variance coming from measurement noise .
most of the previous works cope with these limitations mainly by focusing on bypasses to the problem of noisy and/or missing data ; or deal explicitly with particular cases .
these encompass , for example , the interpolation of missing data @xcite or cases where the weight matrix can be factorized into per - observation and per - variable weight matrices @xcite . @xcite in sections 13.6 and 14.2 makes the point about these proposed solutions . at the present time , some methods are still able to deal with weight matrices having the same size as the corresponding data set @xcite .
nevertheless , none of these are able to provide the orthogonal principal components ordered by the data set variance it accounts for .
rather , they provide an unsorted set of not - necessary orthogonal vectors whose linear combination is optimized to describe the underlying variance but whose goal is not compatible with the explanation of the variance given a minimal number of vectors .
based on the idea of @xcite , @xcite and @xcite have recently proposed methods based upon the expectation - maximization ( em ) algorithm @xcite in order for the pca to include weights associated with each variable within each observation .
the objective of these methods is globally similar to the one of this paper .
differences mainly come from the fact that @xcite aim at finding an orthogonal decomposition of the original matrix such that the whole data set variance is the best accounted for .
instead , our implementation focuses on finding the orthogonal vectors that are the best at individually describing the data set variance at the expense of a lower explained global variance .
this trade - off comes from the fact that in a weighted case , the solution to the problem of finding the set of @xmath0 components explaining most of the variance of a data set is not guaranteed to contain the eigenvectors that are the best at individually describing this variance .
the implementation of @xcite takes benefits of the flexibility of the em algorithm in order to interpolate between these two solutions . in section [ sec : notation ] , we explain the notation used in this paper .
we summarize the properties of the classical pca in section [ sec : pca ] .
we see in details two current alternative implementations of weighted pcas in section [ sec : wempca ] . in section [ sec : wpca_new ] , we describe our new algorithm while in section [ sec : comparison ] , we see its application on simulated data and real cases and compare it against other algorithms . finally , some properties and extensions are discussed in section [ sec : discussion ] and we conclude in section [ sec : conclusion ] .
this paper uses the following notations : vectors are in bold italic , @xmath1 ; @xmath2 being the element @xmath3 of the vector @xmath1 .
matrices are in uppercase boldface or are explicitly stated ; i.e. @xmath4 from which the @xmath3th row will be denoted @xmath5 and the @xmath6th column by @xmath7 , element at row @xmath3 , column @xmath6 will then be @xmath8 . amongst matrix operators , @xmath9 denotes the element - wise product ( hadamard product ) of @xmath10 and @xmath11 and @xmath12 denotes the euclidian matrix norm of @xmath10 .
consider a problem where we have @xmath13 observations each containing @xmath14 variables , from which we want to retrieve @xmath15 principal components .
for reference , here are the often used matrices along with their corresponding sizes : @xmath4 the data set matrix ( @xmath16 ) from which we have subtracted the mean observation @xmath17 ( @xmath18 ) ; @xmath19 the weight of each variable within each observation ( @xmath16 ) ; @xmath20 the orthogonal matrix of principal components ( @xmath21 ) ; @xmath22 being the @xmath3th principal component ; @xmath23 the principal coefficient matrix ( @xmath24 ) ; @xmath25 the symmetric matrix of variance covariance ( @xmath26 ) associated with @xmath4 .
finally , @xmath27 means that @xmath28 is the nearest matrix from @xmath29 in a potentially weighted
least - squares sense .
mathematically , this is equivalent to have @xmath28 and @xmath29 such that @xmath30 is minimized .
regarding classical pca and assuming without any loss of generality that we would like to retrieve as many principal components as the number of variables ( ie .
@xmath31 ) , then the goal of the pca will be to find a decomposition @xmath32 such that @xmath33 is diagonal and for which @xmath34 note that based on equation ( [ eq : pca_d_pvp ] ) and according to the spectral theorem , @xmath20 will be orthogonal .
intuitively , the matrix @xmath20 can be seen as a change of basis allowing us to maximize the variance within @xmath35 and thus minimizing the off - diagonal elements corresponding to the covariance .
differently stated , each @xmath22 defines a privileged direction along which the data set variance is the best explained .
the fact that @xmath35 is ordered implies that for @xmath36 , the principal component @xmath22 accounts for more or equal variance than @xmath37 . for the sake of clarity , a comprehensive pca example is given in fig .
[ fig : pca_example ] . a common solution to such a classical pca is based on the svd of @xmath4 : @xmath38 where @xmath39 , @xmath40 are orthogonals , @xmath41 is diagonal and for which @xmath42 . by setting @xmath43 and @xmath44
, we find that equation ( [ eq : pca_d_pvp ] ) becomes @xmath45 that fulfils the conditions of equations ( [ eq : pca_d_pvp ] ) and ( [ eq : pca_d_ordered ] ) .
note that in equation ( [ eq : pca_d_pvp ] ) the exact variance covariance matrix should be normalized by @xmath13 but since we are solely interested in the diagonalization of @xmath25 , we drop it . can be seen as an orthogonal coordinate transformation from @xmath46 to @xmath47 such that the data set variance is maximized along the vector @xmath48 and @xmath49 .
note that , for a didactical purpose , we chose @xmath20 to be a rotation matrix but practically it can be any orthogonal matrix . ]
as already mentioned in section [ sec : intro ] , the current methods efficiently dealing with the problem of weighting pca @xcite aim at best explaining the whole data set variance according to a given number of principal components , which is equivalent to minimize @xmath50_{ij } \right)^2,\ ] ] where the weighted mean observation we subtracted ( cf . section [ sec : notation ] ) is given by @xmath51 we notice that equation ( [ eq : wpca_x_pc ] ) has latent variables such that it has to rely on an iterative procedure to be solved .
the em algorithm is a statistical tool specifically designed to optimize a likelihood function for models having latent or unknown / hidden variables @xcite .
this iterative procedure is composed of two steps . 1 . e - step : find the expectation value of the latent variables given the parameters of the current model .
m - step : find the parameters of the model such that the likelihood function is optimized .
based on the latter , @xcite has developed a fast and straightforward classical pca algorithm for which the conditions in equations ( [ eq : pca_x_pc ] ) , ( [ eq : pca_d_pvp ] ) and ( [ eq : pca_d_ordered ] ) are all fulfilled . regarding weighted pca and more specifically the @xmath52 described by equation ( [ eq : wpca_x_pc ] ) , we will have the following weighted expected maximization pca ( wempca ) algorithm : @xmath53 random orthogonal matrix while @xmath20 and @xmath23 have not converged _
( e - step ) _
find @xmath23 that minimizes @xmath52 given @xmath20 . _
( m - step ) _
find @xmath20 that minimizes @xmath52 given @xmath23 .
note that the convergence criterion is still relative .
it can be based on the @xmath52 or the change in the principal components @xmath54 falling under a given threshold , the fact that the algorithm has reached a given number of iterations or whatever criterion we consider as relevant .
@xcite designed a general approach to the modelling and dimensionality reduction of the sdss spectra called ` heteroskedastic matrix factorization ' .
more specifically , it attempts to minimize @xmath55 ^ 2,\ ] ] subject to @xmath56 we recognize the first part of equation ( [ eq : tslamantza_chi2 ] ) as being equation ( [ eq : wpca_x_pc ] ) while the second part is a smoothing regularization term whose scalar @xmath57 defines the strength .
non - negativity constraints reflect a particular need to have a meaningful physical interpretation of the resulting spectra . regarding the fact that we would like to model the widest variety of data sets , we will drop the non - negativity constraints that otherwise would have restricted our search space . concerning the smoothing regularization factor
, we have to note that it will be highly problem - dependent and that it can be tricky to optimize , this will result in a potential unfair comparison with other methods
. we will then consider the case @xmath58 .
moreover , as we will see in section [ sec : discussion_smoothing ] , our method can deal with principal components smoothing as well , consequently ignoring it will not constitute a major drawback to our implementation . the resulting function to optimize
will then be reduced to the sole equation ( [ eq : wpca_x_pc ] ) . nevertheless , and before going further
, we have to note that minimizing equation ( [ eq : wpca_x_pc ] ) will provide us a lower - rank matrix approximation of @xmath4 but it is not a sufficient condition for the resulting matrices @xmath20 and @xmath23 to be considered as a pca decomposition . according to @xcite , the solution to this problem can be solved in two steps .
first consider a lower - rank matrix decomposition of @xmath4 , similar to the one produced by the solution of equation ( [ eq : wpca_x_pc ] ) , @xmath59 where for clarity , the sizes of these matrices are @xmath60 and @xmath61 .
now suppose an orthogonal basis @xmath62 of @xmath28 ; such an orthogonal basis always exists for full - rank matrices and can be retrieved through a straightforward gram
schmidt process for example .
the associated coefficients matrix @xmath63 is then directly retrieved by @xmath64 secondly , in a way similar to equation ( [ eq : pca_cc_d ] ) , we will take the classical pca decomposition of @xmath63 , @xmath65 such that @xmath66 is diagonal .
the resulting principal coefficient matrix will then be given by @xmath67 that will be orthogonal and that will provide us with the final decomposition @xmath68 the above - mentioned steps have thus to be performed after the em algorithm minimizing equation ( [ eq : wpca_x_pc ] ) in order for @xmath20 to be orthogonal and for the covariance matrix @xmath69 to be diagonal . sections [ sec : wempca_tsalmantza_estep ] and [ sec : wempca_tsalmantza_mstep ] will now focus on details of the em algorithm .
as stated at the beginning of this section , the expectation step regarding wempca will be given by the retrieval of the coefficient matrix @xmath23 that minimizes equation ( [ eq : wpca_x_pc ] ) or equivalently that optimizes equation ( [ eq : wempca_x_pc ] ) while considering the principal component matrix @xmath20 being held fixed . since each observation
column of @xmath4 is a linear combination of the principal components , finding the solution of equation ( [ eq : wempca_x_pc ] ) is equivalent to solving @xmath70 whose solutions are given by the ` normal equations ' : ) . ]
@xmath71 with @xmath72 . similarly to the e - step ,
solution to the m - step that is the retrieval of @xmath20 that optimizes equation ( [ eq : wempca_x_pc ] ) given @xmath23 can be decomposed by noting that within each observation , a given variable is the linear combination of the corresponding principal components variables .
that is , @xmath73 whose solutions are @xmath74 with @xmath75 .
as we have seen in section [ sec : tsalmantza ] , the implementation of @xcite focuses on the solution of equation ( [ eq : wpca_x_pc ] ) while the pca decomposition can be seen as a supplemental step that is external to the em algorithm .
moreover , a single iteration of the algorithm requires the solution of @xmath76 systems of linear equations , each of size ( @xmath77 ) , at each iteration of the em algorithm .
this becomes quickly unmanageable regarding huge data sets .
@xcite takes the pragmatic approach that the sole solution of equation ( [ eq : wpca_x_pc ] ) can lead to good insights about the principal components if the latter were fitted individually .
this hypothesis is reasonable since this will allow each individual principal component to maximize the variance it accounts for .
the resulting implementation will be similar to the one of @xcite apart from the optimization function of the m - step .
indeed , in order for the principal components to be fitted separately , we have to consider the cross - product decomposition of equation ( [ eq : wempca_x_pc ] ) , that is @xmath78 from which each @xmath37 has to be individually fitted .
suppose that we already retrieved the @xmath79 first principal components .
let us also assume that the data projection along these @xmath79 principal components was already subtracted from the data set , that is @xmath80 then the retrieval of @xmath37 based on equation ( [ eq : wempca_bailey_mstep ] ) can be decomposed in a way similar to equation ( [ eq : wempca_tsalmantza_mstep_objective ] ) as @xmath81 whose solution is straightly given by @xmath82 with @xmath75 .
equation ( [ eq : wempca_bailey_proj_removal ] ) theoretically ensures that the last - retrieved component , @xmath37 will be orthogonal to any previous one .
nevertheless , due to machine round - off errors , this has to be manually checked .
finally , we have to note that solving equation ( [ eq : wempca_bailey_mstep ] ) will not minimize the global @xmath52 as defined by equation ( [ eq : wpca_x_pc ] ) such that the algorithm has to rely on a last e - step at the end of the main em algorithm .
though both mentioned algorithms @xcite correctly find lower - rank orthogonal decompositions that are suitable to explain the whole data set variance at best , none of them assures us that the retrieved principal components will be those that maximize the individual variance described by each of them .
these principal components are then efficient at reconstructing the initial data set but are not the best at individually describing the underlying data structure .
the basic idea of this new algorithm is to focus on the maximization of the weighted variance explained by each principal component through the diagonalization of the associated weighted covariance matrix .
the resulting principal components will then be those that are the most significant under the assumption that the definition of the used weighted variance is relevant in identifying pattern within the data set even if their linear combination is not necessarily the best at explaining the total data set variance as described by equation ( [ eq : wpca_x_pc ] ) . in the following
, we will consider that the weighted variance of a given discrete variable @xmath1 having weights @xmath83 is given by @xmath84 where @xmath85 and with the convention that @xmath86 .
the latter can be straightforwardly extended to the definition of the weighted covariance between two discrete variables , @xmath1 and @xmath87 , for which the weights are given by @xmath88 and @xmath89 respectively , that is @xmath90 based on these definitions , we can write the weighted covariance matrix of a data set @xmath4 with associated weights @xmath19 as @xmath91 we know , from the spectral theorem , that there exists an orthogonal matrix @xmath20 such that @xmath25 is diagonalized , and consequently that equation ( [ eq : pca_d_pvp ] ) is fulfilled ( as well as equation [ eq : pca_d_ordered ] if @xmath20 is ordered accordingly ) .
this matrix @xmath20 will then constitute the principal components of our implementation .
and classical principal components @xmath92 in presence of underweighted observations corresponding to the brighter points .
@xmath93 and @xmath94 correspond to the variances respectively explained by @xmath48 and @xmath49 . ]
[ fig : wpca_diff ] shows a two - dimensional example of classical principal components @xmath92 that are unable to individually describe the underlying data variance . in this example , we have supposed for didactical purpose that the observations corresponding to the bright points have far lower weights associated with the @xmath95 variable .
weighted principal components , @xmath20 , diagonalizing @xmath25 as described by equation ( [ eq : covar_mat_new ] ) are given along with the variance it explains ( that is the diagonal elements of @xmath96 ) .
note that such principal components maximize the variance explained by each variable as described by equation ( [ eq : var_new ] ) and consequently set to zero the associated covariance as described by equation ( [ eq : covar_new ] ) .
the goal of the algorithm is then to retrieve the dominant eigenvector @xmath97 from the covariance matrix @xmath25 along with its associated eigenvalue @xmath98 , that is the dominant eigenpair @xmath99 .
@xmath97 will then be the principal component explaining most of the data set variance , @xmath98 .
that is equivalent to find @xmath97 in @xmath100 such that @xmath98 is maximized is also solution of equation ( [ eq : eigenvector ] ) but for the sake of clarity , we will only consider the case of positive eigenvalues ]
. equation ( [ eq : eigenvector ] ) , corresponding to the eigenvector definition , is a well - studied problem for which many methods already exist .
the reference in the domain is @xcite where the interested reader may find a rather exhaustive list of such methods as well as proofs of the algorithms described hereafter .
unsurprisingly , in the context or our implementation , we choose the fastest and simplest algorithm called the power iteration method .
the idea behind this algorithm is to recognize that given a diagonalizable square matrix @xmath28 and a vector @xmath101 having a nonzero component in the direction of @xmath97 , the iterative relation @xmath102 will converge to a vector that is proportional to the dominant eigenvector @xmath97 as @xmath103 .
note that in practice , each vector @xmath104 is normalized to unity in order to avoid numerical round - off errors inherent to the computer representation of large numbers .
the final eigenvector will then be given by @xmath105 and the associated eigenvalue by the rayleigh quotient : @xmath106 convergence and assumptions made about this algorithm will be discussed in section [ sec : discussion_convergence ] .
further principal components can be retrieved by considering application of the above - mentioned algorithm to @xmath107 that is the matrix obtained by subtracting the data variance along the found principal components .
as we will see in section [ sec : discussion_convergence ] , the power iteration method may have a slow convergence rate under some particular conditions .
consequently it may be that some vectors did not effectively converge to an eigenvector that would have diagonalized the covariance matrix .
nevertheless , @xcite proposed an algorithm called rayleigh quotient iteration designed to tackle this kind of problem . even if the proof of this algorithm is beyond the scope of this paper , we still mention two basic facts to enable the reader to have a minimal understanding of how it works
first , as we have seen in equation ( [ eq : rayleigh_quotient ] ) , the rayleigh quotient of a matrix with one of its eigenvector is equal to its associated eigenvalue . secondly , given a matrix @xmath28 with eigenvalues @xmath108
, we find that the eigenvalues of the matrix @xmath109 will be @xmath110 .
based upon these facts , we will have that the sequence @xmath111 where each @xmath104 is normalized to the unit length , will converge cubically to the eigenpair @xmath99 that is the nearest regarding the absolute value of their rayleigh quotient from a starting point @xmath112 as @xmath103 .
note that equation ( [ eq : rayleigh_iter_power ] ) is a power iteration towards the eigenvector for which @xmath113 is minimized .
finally , the principal coefficients are retrieved by solving @xmath114 whose ` normal equation ' solution is given by equation ( [ eq : wempca_e_step ] ) .
real - world data often have sparse and unevenly distributed weights such that it may happen for some variables to have their corresponding variances to be based only on a small number of observations .
such situations may become problematic since these few variables will have a strong impact on the resulting first principal components in a way that is irrespective to their total weight .
such ` overweighted ' variables can be damped by using a regularization factor within the expression of the weighted covariance , @xmath115 , as defined by equation ( [ eq : covar_new ] ) .
the resulting regularized weighted covariance will take the form @xmath116^{\xi } { \sigma_{{\bmath{x}},{\bmath{y}}}^2},\ ] ] where the regularization parameter @xmath117 allows us to control the damping strength .
the typical value of the regularization parameter , @xmath117 , goes from zero , where we get back to the classical behaviour of the algorithm , to two for a strong damping of these rare variables .
conversely one might want to highlight such underrepresented variables by setting the regularization parameter to a negative value .
the performance of the method described in this paper was assessed regarding the two previously described algorithms , namely the one of @xcite and the one of @xcite .
the choice of these algorithms comes from the fact that they are fairly competitive and have goals that are comparable to those of the new algorithm .
all methods were tested on both simulated data as well as on real observational ones .
simulated data consist in random linear combinations of 10 orthogonal basis functions .
these bases are produced by taking 10 shifted sine functions having periods between @xmath118 and @xmath119 and by applying a gram
schmidt orthogonalization process to the latter .
resulting observations are then sampled over 100 evenly spaced points in the interval @xmath120 $ ] .
to each variable , @xmath95 , within each observation , @xmath1 , we also add a gaussian noise having a standard deviation given by @xmath121 where @xmath122 is a user - provided parameter corresponding to the desired noise amplitude , @xmath123 is an observation - specific noise ponderation uniformly drawn from @xmath124 $ ] and @xmath125 is a uniform random variable corresponding to the noise dispersion within observations and taking values in the range @xmath124 $ ] . the weight associated with the variable @xmath95
will then be set to @xmath126 .
finally , we discard @xmath127 contiguous and randomly positioned variables from each observation .
the latter will be used to assess performances of the various algorithms on data extrapolation while having their weights equal to zero during the pca retrieval phase .
examples of such simulated data are illustrated in fig .
[ fig : comparison_example ] . and @xmath127 .
] in order to perform the comparison with other algorithms , we built a given number of data sets , @xmath128 , each containing 1000 observations .
each of these sets was duplicated and altered through the realization of various values of the simulation parameters @xmath122 and @xmath127 .
we then retrieved , for each of the mentioned algorithms , the five first principal components out of the resulting altered data sets and computed estimators based on the following @xmath52 definition : @xmath129_{ij}^2}{\sum_{ij } { \bmath{\mathsf{w}}}_{ij}^ 2}}.\ ] ] the following estimators were computed : @xmath130 , the chi - square of the data set for which weights associated with the discarded variables are set to zero and @xmath131 where only rejected variables are considered and for which weights associated with the unrejected variables are set to zero .
let us note that @xmath130 will typically account for the quality of the fit while @xmath131 will account for the quality of the extrapolation .
if these estimators are to be computed based on the number of data set , @xmath128 , we use a @xmath132-clipped mean over all the @xmath130 and @xmath131 associated with each data set , namely @xmath133 and @xmath134 . for completeness
, only data sets having @xmath135 will be discussed here .
this decision comes from the fact that efficiently estimating the principal components of such sparse data sets while having ` only ' 1000 observations is a really tricky task strongly depending on the design of these data sets .
furthermore , as we will see in section [ sec : scaling_performance ] , the bailey and tsalmantza algorithms are big time consumers such that dealing with bigger data sets quickly becomes unmanageable .
since all the studied algorithms are based on iterative procedures , we have to take into account the convergence criterion for each of them .
to this aim , we performed a preliminary study whose goal is to determine the minimal number of iterations needed by each algorithm in order to reach convergence .
this was assessed by running 100 times each algorithm on data sets similar to those previously described and by setting the initial eigenvectors estimates to random values within each run . in order to make sure we can model unseen and potentially more complex data sets , the number of iterations was set to twice the found number of iterations needed to converge , giving respectively @xmath136 , @xmath136 and @xmath137 iterations for bailey , tsalmantza and our algorithm ( without refinement ) . for increasing noise without missing data ( top ) .
subtraction of bailey and tsalmantza @xmath130 from our algorithm @xmath130 ( middle ) . averaged @xmath130 with moderate noise and an increasing number of missing data ( bottom ) . ] regarding the quality of the fit , @xmath130 , differences between the various algorithms are fairly low even if as expected our algorithm is proven to have somewhat larger @xmath130 with a higher dispersion over all the data sets .
[ fig : comparison_chi2_fit ] shows the behaviour of the mean @xmath130 regarding two common cases , namely the case of increasing noise and no missing data and the case of moderate noise ( @xmath138 ) with increasing number of missing data . each point on these graphs is averaged over @xmath139 data sets .
practically , one can not distinguish the various algorithms if all data are present since differences are in @xmath140 .
in presence of missing data , differences start to be noticeable but still reasonable with differences in @xmath141 . for a didactical purpose ,
let us note that if only the first component out of the five retrieved was considered then our algorithm would have had a better or equal @xmath130 than bailey and tsalmantza implementations over all the values of the simulation parameters @xmath122 and @xmath127 .
maps of averaged @xmath131 regarding @xmath122 and @xmath127 simulation parameters . ] for moderate noise with an increasing number of missing data . ] regarding now the quality of the extrapolation , differences are more noticeable .
[ fig : comparison_chi2_test_map ] illustrates these discrepancies according to the averaged @xmath131 over @xmath142 data sets having reasonable @xmath130 in each of the @xmath122 and @xmath127 simulation parameters . at first glance ,
our new algorithm shows a globally better @xmath134 while suffering less from missing data : the latter being dominated by noise within the data . the tsalmantza s algorithm shows good performances but has a stronger dependence on missing data , reaching a maximum @xmath143 while bailey s algorithm shows strong numerical instabilities for @xmath144 that makes it unable to converge and makes it reach a maximum @xmath145 for @xmath146 and @xmath147 .
[ fig : comparison_chi2_test ] illustrates in more detail the behaviour of @xmath134 averaged over @xmath139 data sets in the common case of moderate noise ( @xmath138 ) with an increasing number of missing data . for clarity , the plot of bailey having @xmath144 has been removed from the graph .
finally , let us note that data sets having @xmath148 and @xmath149 can be solved using a classical pca algorithm .
consequently , both the variance explained by each individual component and the total variance can be simultaneously optimized . we will thus find in this particular case that all algorithms will provide us with identical results as it was already suggested in fig .
[ fig : comparison_chi2_fit ] .
similarly , if we choose to retrieve a single component , even with noisy and/or missing data , the algorithms of bailey and tsalmantza will maximize the variance explained by this component , the latter will thus match the first component that would have been retrieved by our algorithm .
comparisons against a concrete case were performed using the sdss dr10q quasar catalog from @xcite . out of the 166 583 qso spectra present in the initial data release ,
18 533 were rejected either due to spectra bad quality , strong uncertainties in redshift determination , presence of bal or insufficient number of high - s / n points .
the remaining spectra were set to the rest frame ; the continuum was then subtracted and the spectra were normalized such as to have a zero mean and a variance of 1 .
finally , visual inspection showed that in some cases the continuum was badly fitted such that the variance within these spectra can mainly be attributed to this error ; these regions were removed using a _ k_-sigma clipping algorithm for each variable among all observations .
there remain 148 050 spectra having observed wavelengths between 4000 and 10 000 and for which the variance within each spectrum is thought to be mainly caused by genuine signals . in the following tests , the 10 first components will be retrieved from each algorithm .
we also suppose , as previously , that @xmath150 .
the number of iterations associated with each algorithm was assessed by using a subset of the above - described data set ( 1000 @xmath151 2000 , @xmath152 ) and by running 10 times the various algorithms on it with random initial principal components .
convergence was assessed by minimizing the variance amongst the final principal components within these 10 runs .
the fact that only a subset of the above - described data set was used can be explained by the large amount of time needed by each algorithm to run as well as by the fact that as we will see tsalmantza and bailey algorithms often fail to converge in presence of a large amount of missing data .
the results show that the number of iterations chosen in section [ sec : comparison_simulated ] also match the complexity of this problem .
the initial test to be performed is similar to tests performed in section [ sec : comparison_simulated ] in the sense that pca were retrieved for all algorithms using the region 50009000 in observed wavelength while regions 40005000 and 900010 000 are rejected and kept to assess the quality of the extrapolation .
[ fig : comparison_sdss_spec ] shows an example of such a spectrum along with the fits by the various algorithms .
the resulting sdss dr10q data set @xmath130 and @xmath131 , as defined by equation ( [ eq : data set_chisq ] ) , are given in table [ tbl : sdss_data set_chi2 ] . .sdss
dr10q data set fit and extrapolation chi - squares followed by per - observation median chi - squares of extrapolation and associated ratio of outliers . [ cols="<,^,^,^",options="header " , ] .
right - hand part represents the number of qsos ( @xmath153 ) having @xmath154 .
] we see that @xmath130 is quite stable for all the algorithms while being a bit higher for our algorithm as expected . now watching at the extreme differences of @xmath131 for the various algorithms and given the fact that we know it to be an estimator that is quite sensitive to outliers , we found relevant to see how these discrepancies are distributed among the observations . to this aim , we computed a @xmath131 per - observation whose distribution is summarized in fig . [
fig : comparison_sdss_detail ] . given the significant number of observations used in this test , we can already draw some general trends .
beside the fact that the other algorithms have a better @xmath130 , they often fail to satisfactorily extrapolate the spectra .
indeed , for the bailey algorithm , @xmath155 of observations have @xmath154 with a peak up to @xmath156 ( median=@xmath157 ) ; the tsalmantza algorithm has respectively @xmath158 of observations having @xmath154 and a peak up to @xmath159 ( median=@xmath160 ) while our new algorithm has only @xmath161 of ` outliers ' with a maximal peak of @xmath162 ( median=@xmath163 ) .
if the observations are individually compared , our algorithm has a lower @xmath131 in @xmath164 of the time regarding bailey algorithm and @xmath165 of the time regarding tsalmantza algorithm while in other cases differences remain quite moderate with a median @xmath166 . for completeness
, we have to note that these huge discrepancies can be mainly attributed to the large amount of missing data ( @xmath167 ) within the resulting rest - frame dr10q data set .
another quality one would often desire is the ability to have the most general and representative set of principal components able to model unseen observations . in this optic
, two tests were performed .
first , we split the initial dr10q data set into two subsamples each of 74 025 spectra spanning from 4000 to 10 000 ( hereafter ls1 and ls2 ) . from each of these subsamples , we extract @xmath1688600 spectra for which the rest - frame region from 1000 to 2000 is entirely covered and consequently ` without ' missing data and @xmath16862 000 spectra for which the mentioned region is only partially covered and thus ` with ' missing data .
note that according to the selected observed wavelengths and the previous definitions , spectra without missing data will correspond to those having @xmath152 and spectra with missing data to those having @xmath169 .
the consistency and convergence of each algorithm were then tested by retrieving 10 principal components out of ls1 and ls2 data sets taken with and without missing data . prior to discussing the results
, we have to mention that the components may be swapped from data sets with missing data to the corresponding ones without missing data .
this is easily explained by the fact that some patterns may be highlighted due to the uneven coverage of the wavelength range while being damped in the total variance if a full coverage is considered
[ fig : comparison_ls1_ls2 ] illustrates the results of this test regarding the first component of the data set without missing data and associated component with missing data .
we see that in absence of missing data , all algorithms are consistent and succeed in converging towards the dominant eigenvector while in the case of missing data only our algorithm shows both a good consistency and a good convergence .
more precisely , regarding our algorithm , the mean differences per - point between ls1 and ls2 are in @xmath140 for both subsets with and without missing data .
differences noticed between subsets , that is mainly the general larger equivalent width and the stronger n@xmath170@xmath981240 emission line of the subset with missing data , are consistent with the underlying data set and make it able to model the larger variety of spectra coming from the larger redshift range .
let us mention , that no significant inconsistencies have been noticed up to the sixth component for ls1 and ls2 on both subsets for our algorithm .
concerning the bailey s algorithm , we see , as already suggested in section [ sec : comparison_simulated ] , that it fails to converge if a large amount of data is missing .
tsalmantza s implementation shows a better convergence but fails to correctly reproduce some emission lines ( o@xmath170@xmath981033 , n@xmath170@xmath981240 , c@xmath170@xmath981549 and c@xmath170@xmath981908 ) as well as the region @xmath1711200 in subsets with missing data . similarly to bailey s algorithm , analysis show that these are due to a convergence problem occurring on sparse data sets and arising from numerical instabilities .
in the second test performed , 10 principal components were retrieved from a set of 98 700 spectra spanning from 4000 to 10 000 ( hereafter ls3 ) .
we tested the fact that the principal components are the best at individually describing the underlying data variance by fitting the remaining 49 350 spectra ( hereafter vs ) using a subset of five components retrieved from ls3 .
that is , if the variance explained by each individual component is maximized then the fit of a smaller number of components to a similar data set should be minimal . in this case , if we consider computing @xmath52 of vs data set as defined by equation ( [ eq : data set_chisq ] ) for each of the algorithms , we have respectively @xmath172 for the bailey algorithm , @xmath173 for the tsalmantza algorithm and @xmath174 for our algorithm , thus supporting our last quality criterion .
note again that if all components were considered , bailey and tsalmantza algorithms would have had a better resulting @xmath52 .
at first glance it is surprising that the power iteration algorithm works at all , but in practice , it is easily demonstrated . consider a diagonalizable square matrix @xmath28 of size ( @xmath175 ) having eigenvectors @xmath176 and associated eigenvalues @xmath108 where @xmath177 . since the eigenvectors are orthogonal between each other , we can write the starting vector of the power iteration algorithm as @xmath178 then we will have that the vector at iteration @xmath179 given by @xmath180.\end{aligned}\ ] ] we see that @xmath181 will converge to @xmath182 as @xmath103 under the following conditions : @xmath183 , i.e. the starting vector has a nonzero component in the direction of the dominant eigenvector and @xmath184 , i.e. the data set has only one dominant eigenvalue .
also note that , in this case , the rate of convergence to the dominant eigenvector will be principally given by @xmath185 and that in the case of @xmath186 , the uniqueness of the solution is not guaranteed as it depends on the starting vector @xmath187 .
refinement seen in section [ sec : wpca_new_refinement ] obeys to the same conditions .
additionally , the dominant eigenvalue of @xmath188 will tend to @xmath189 as @xmath190 , where @xmath28 is the covariance matrix , @xmath191 the current ` eigenvalue ' ( i.e. the rayleigh quotient of @xmath28 and @xmath192 , the current ` eigenvector ' ) and @xmath98 a real eigenvalue of @xmath28 .
this may thus lead to numerical instabilities that can strongly deteriorate the vector used in the next step of the rayleigh quotient iteration .
that being said , failures against convergence can be easily checked .
for example , a satisfactory solution should have @xmath193 ; and if not , the power iteration algorithm can be resumed with another starting vector @xmath187 . under the condition that we have @xmath186 , we will have an infinite number of eigenvectors that are equally good at describing the data set variance , thus choosing one over another is irrelevant . concretely , during the @xmath168100 000 tests performed in the context of this paper , no such problems arose .
checks performed on eigenvectors from section [ sec : comparison_simulated ] and [ sec : comparison_observation ] show that the latter are orthogonal to a machine precision of @xmath194 and that the weighted covariance matrix ( as described by equation [ eq : covar_mat_new ] ) is diagonalized with the same precision .
nevertheless , people wanting extreme reproducibility and/or secure convergence may still use the svd in order to extract all eigenvectors from equation ( [ eq : covar_mat_new ] ) at the expense of a lower flexibility ( see sections [ sec : discussion_apriori ] and [ sec : discussion_smoothing ] ) .
in some situations , one may already have an approximation of the wanted principal components corresponding to a given data set .
typical examples include principal components update according to new observations added to the data set or in a real case we encountered , the use of sdss dr9q eigenvectors to the dr10q data set .
these will constitute a priori eigenvectors that it would be regrettable not to use .
the design of our algorithm allows us to easily take benefits of these a priori vectors . instead of using a random vector , @xmath101 , in equation ( [ eq : power_iteration ] ) one can straightforwardly substitute the known vectors to the random starting vector
usually used .
doing so will typically decrease the number of iterations needed to converge towards the new vectors . beware that vectors may be swapped between a priori eigenvectors and effective data set eigenvectors as encountered in section [ sec : comparison_observation ] and consequently iterations needed to converge will be the same as if we had used a random starting vector . in this case
, one can perform no power iteration and use only the refinement seen in section [ sec : wpca_new_refinement ] in order to converge to the nearest eigenvector .
it sometimes happens for some variables within eigenvectors to exhibit some sharp features that we know to be artefacts .
these occur mainly in cases where we have noise that is comparable in amplitude to the data variance , data regions covered only by a few numbers of observations or data sets containing corrupted observations . again
the flexibility of our algorithm allows us to efficiently deal with these drawbacks .
suppose that we retrieved such a ` noisy ' eigenpair @xmath99 from a given data set whose covariance matrix is given by @xmath28 , then , before removing the variance in the direction of the found component through equation ( [ eq : rayleigh_subtract ] ) , one can filter @xmath97 thanks to any existing smoothing function .
assuming that the resulting vector , @xmath195 , is near @xmath97 regarding the norm of their difference , we can suppose that the variance accounted for by @xmath195 ( i.e. its ` eigenvalue ' ) is similar to the one of @xmath97 and then subtract it along the direction of @xmath195 by @xmath196 note that the kind of smoothing function to use is highly data - dependent and no rule of thumb exists .
nevertheless , a quite general and commonly used filter producing efficient results in the field of qsos can be found in @xcite . finally , let us note that , when applying a filter , the orthogonality of the components has to be manually checked and that they will obviously not diagonalize the covariance matrix anymore .
pca is often used in cases where we have a lot of observations and a reasonable number of variables , typically we have @xmath197 . we know , for example , the classical algorithm to require @xmath198 basic operations in order to solve for the eigenvectors of the covariance matrix and @xmath199 operations to build this matrix . in the following
, we will compute the algorithmic complexity of the various explored algorithms in a similar way .
the tsalmantza algorithm requires for each iteration ( hereafter @xmath200 ) the solution of linear systems of equations for each of the e and m - step , respectively , in each observation and in each variable .
we will then have that its algorithmic complexity is given by @xmath201 .
the bailey algorithm will be identical except for the m - step that will be in @xmath202 and can thus be discarded , giving @xmath203 .
finally , our algorithm mainly requires the building of the covariance matrix , @xmath200 matrix multiplications , potentially @xmath204 matrix inversions in order to refine the eigenvectors and a final single e - step similar to the one for the previous algorithm ( discarded here ) , thus giving @xmath205 .
we see that in case @xmath197 , our algorithm is much faster than the other ones as the computing time is mainly spent in the covariance matrix building . as an illustration , if we take data similar to those described in section [ sec : comparison_simulated ] with @xmath206 10 000 , pca retrieval takes @xmath207s for the bailey and tsalmantza algorithms ( with @xmath208 ) and @xmath209s for the new one ( with @xmath21010 000 ) on a 2.4ghz cpu .
we presented a new method for computing principal components based on data sets having noisy and/or missing data .
the underlying ideas are intuitive and lead to a fast and flexible algorithm that is a generalization of the classical pca algorithm .
unlike existing methods , based on lower - rank matrix approximations , the resulting principal components are not those that aim to explain at best the whole variance of a given data set but rather those that are the most suitable in identifying the most significant patterns out of the data set while explaining most of its variance .
the main benefits of the current implementation are a better behaviour in presence of missing data as well as faster run times on data set having a large amount of observations .
privileged problems encompass data set extrapolation , patterns analysis and principal component usage over similar data sets .
we assessed the algorithm performance on simulated data as well as on qso spectra to which many applications are already foreseen in the field of the _ gaia _ mission .
the author acknowledges support from the esa prodex programme ` gaia - dpac qsos ' and from the belgian federal science policy office .
funding for sdss - iii has been provided by the alfred p. sloan foundation , the participating institutions , the national science foundation and the u.s .
department of energy office of science .
the sdss - iii web site is http://www.sdss3.org/. sdss - iii is managed by the astrophysical research consortium for the participating institutions of the sdss - iii collaboration including the university of arizona , the brazilian participation group , brookhaven national laboratory , carnegie mellon university , university of florida , the french participation group , the german participation group , harvard university , the instituto de astrofisica de canarias , the michigan state / notre dame / jina participation group , johns hopkins university , lawrence berkeley national laboratory , max planck institute for astrophysics , max planck institute for extraterrestrial physics , new mexico state university , new york university , ohio state university , pennsylvania state university , university of portsmouth , princeton university , the spanish participation group , university of tokyo , university of utah , vanderbilt university , university of virginia , university of washington and yale university .
bailer - jones c.a.l .
et al . , 2013 ,
a&a , 559 , a74 bailey s. , 2012 , pasp , 124 , 1015 beale e. m. l. , little r. j. a. , 1975 , j. r. stat .
b , 37 , 129 dempster a. p. , laird , n. m. , rubin d. b. , 1977 , j. r. statist .
b , 39 , 1 gabriel k.r . ,
zamir s. , 1979 , technometrics , 21 , 489 golub g.h . , van loan c.f . , 1996 , matrix computations , 3rd edn .
the johns hopkins univ . press ,
london greenacre m.j . , 1984 ,
theory and applications of correspondence analysis . academic press , london hotelling h. , 1933 , j. educ .
psychol . , 24 , 417 jarvis m. , jain b. , 2004 , preprint(astro - ph/0412234 ) jolliffe i.t . , 2002 , principal component analysis , 2nd edn .
springer , new york pris i. et al .
, 2014 , a&a , 563 , a54 parlett b.n . , 1974 , math .
comput . , 28 , 679 pearson k. , 1901 , phil . mag . , 2
, 559 press w.h . , tuekolsky s.a .
, vetterling w.t . ,
flannery b.p . , 2002 , numerical recipes in c++ : the art of scientific computing , 2nd edn .
cambridge univ .
press , new york roweis s. , 1997 , cns technical report cns - tr-97 - 02 , em algorithm for pca and spca .
available at : http://cs.nyu.edu/~roweis/papers/empca.pdf savitzky a. , golay m.j.e .
, 1964 , anal .
, 36 , 1627 schlens j. , 2009 , a tutorial on principal component analysis .
available at : http://www.snl.salk.edu/~shlens/pca.pdf srebro n. , jaakkola t. , 2003 , proc .
20th icml , weighted low rank approximation , aaai press , menlo park , p. 720
tipping m.e . ,
bishop c.m . , 1999 ,
j. r. stat .
b , 21 , 611 tsalmantza p. , hogg d.w .
, 2012 , apj , 753 , 122 wentzell p.d . ,
andrews d.t . ,
hamilton d.c . , faber k. , kowalski b.r . , 1997 , j. chemometr . , 11 , 339 yip c.w .
et al . , 2004 ,
aj , 128 , 2603
.... % function designed to find the weighted pca of a % given data set such that the resulting principal % components diagonalize the associated weighted % covariance matrix and that the associated principal % coefficients are the linear coefficients of these % principal components that best match the data set % in a least - squares sense .
% % input : % x : data set matrix from which the weighted % mean was already subtracted .
rows % correspond to variables , columns to % observations .
% w : weight matrix ( same size as x ) .
% ncomp : number of principal components to % retrieve .
% niter : number of iterations ( allows % convergence to the dominant principal % component ) % nrefine : number of refinments ( allows refinement % of the dominant principal component ) % xi : regularization factor controlling the % influence of rare variables % output : % p : principal components arranged in columns % c : principal coefficients such that x = p*c % in a least - squares sense % author : % ludovic delchambre % extragalactic astrophysics and space observations , % institute of astrophysics and geophysics , % university of liege , belgium function [ p , c ] = wpca(x , w , ncomp , niter , nrefine , xi ) nvar = size(x,1 ) ; nobs = size(x,2 ) ; p = zeros(nvar , ncomp);c = zeros(ncomp , nobs ) ; ws = sum(w,2 ) ; covar = ( ( ws * ws').^xi).*((x.*w)*(x.*w)')./(w*w ' ) ; covar(isnan(covar ) ) = 0 ; for i=1:ncomp u = ones(nvar,1 ) / nvar ; for j=1:niter u = covar*u ; d = sqrt(u'*u ) ; u = u / d ; end d = u'*covar*u ; for j=1:nrefine u = inv(covar - d * eye(nvar ) ) * u ; u = u / sqrt(u'*u ) ; d = u'*covar*u ; end covar = covar - u*d*u ' ; p(:,i ) = u ; end for i=1:nobs w = diag(w(:,i)).^2 ; c(:,i ) = inv(p'*w*p ) * ( p'*w*x(:,i ) ) ; end end .... | we present a new straightforward principal component analysis ( pca ) method based on the diagonalization of the weighted variance covariance matrix through two spectral decomposition methods : power iteration and rayleigh quotient iteration .
this method allows one to retrieve a given number of orthogonal principal components amongst the most meaningful ones for the case of problems with weighted and/or missing data .
principal coefficients are then retrieved by fitting principal components to the data while providing the final decomposition .
tests performed on real and simulated cases show that our method is optimal in the identification of the most significant patterns within data sets .
we illustrate the usefulness of this method by assessing its quality on the extrapolation of sloan digital sky survey quasar spectra from measured wavelengths to shorter and longer wavelengths .
our new algorithm also benefits from a fast and flexible implementation .
[ firstpage ] methods : data analysis quasars : general . |
considerable theoretical and observational interest has in recent years been focused on asymptotic giant branch ( agb ) stars but much remains mysterious about them .
the agb phase is enjoyed by low and intermediate mass stars ( approximate mass range 0.8 to 8m@xmath3 ) .
it is in this phase that a star experiences extensive internal nucleosynthesis whose fruits are dredged to the stellar surface * ( see herwig 2005)*. furthermore , mass - loss ensures that the products of nucleosynthesis are dispersed into the circumstellar and subsequently the interstellar environment .
thus , agb stars are likely major contributors of li , c , n , f and @xmath4-process elements among others to galactic chemical evolution * ( romano et al
. 2010)*. observational validation of theoretical investigations of how agb stars achieve internal nucleosynthesis , dredge - up and mass - loss are generally hampered by the fact that the more evolved and more interesting of these stars have low temperature atmospheres and therefore spectra replete with dense and complex molecular * lines*. although these spectra have been analysed by a few stellar spectroscopists , the abundance studies are generally limited to very few elements and isotopic ratios .
the agb phase of stellar evolution concludes with a rapid phase of evolution to the tip of the white dwarf cooling track .
this post - agb ( pagb ) phase takes a star in a few thousand years or so from a cool agb star to a very hot central star of a planetary nebula and onto a white dwarf cooling track . along this track ,
the star s spectrum is amenable to straightforward abundance analysis and thus offers apparently a way to infer the abundance changes brought by the pagb star s agb progenitor .
this inference is , of course , dependent on the assumption that the composition of an agb star is exactly preserved by the pagb star .
observations of certain pagb stars show that this is a false assumption .
for example , several pagb stars show abundance anomalies correlated with the condensation temperature ( t@xmath5 ) is the temperature at which half of a particular element in a gaseous environment condenses into dust grains .
condensable elements refer to the refractories like ti , ca , sc and the s - process elements which readily condense into dust grains because of their high t@xmath5 . * ] for dust grains to condense out of gas of normal composition ( see van winckel 2003 for a review ) a process we refer to as ` dust - gas winnowing ' .
other pagb candidate stars show abundance anomalies correlated with the ionization potential of the neutral atoms ( rao & reddy 2005 ) a process we refer to as ` the fip effect ' .
given that the pagb phase is rapid , the number of identified pagb stars is relatively small and the number subjected to an abundance analysis is , of course , even smaller . in this paper , we report abundance analyses for 11 candidate pagb stars and have compiled abundance data of previously analysed pagb stars in an attempt to seek explanations for the diverse compositions of these objects . habing ( 1988 ) and area enclosed in the dashed line which according to szczerba et al .
2007 is richly populated by pagb stars with colours like planetary nebulae .
the sample stars are numbered as follows : 1- iras 01259 + 6823 2- iras 04535 + 3747 3- iras 05208 - 2035 4- iras 07140 - 2321 5- iras 07331 + 0021 6- iras 08187 - 1905 8- iras
12538 - 2611 9- iras
17279 - 1119 10- iras 22223 + 4327 , width=377,height=377 ]
our sample stars are presented in table 1 and displayed in the iras two colour diagram figure [ loci1 ] which has proven to be a powerful tool for identifying candidate pagb stars ( van der veen & habing 1988 ; surez et al .
2006 ; szczerba et al .
nine of our eleven stars have measured iras fluxes and are identified in figure [ loci1 ] ; the stars , hd 107369 and bd + 39@xmath0 4926 were not detected by iras .
it is known that zone 1 of iras two colour diagram by van der veen & habing 1988 corresponds to fluxes from stellar photospheres warmer than 2000k .
the iras colours of zones ii - iii signify the emergence and evolution of circumstellar shell ( cs ) produced by increasing large mass - loss at agb .
sources at zone iv are at the super - wind phases or slightly beyond .
zone v contains objects with only signature of cold dust shell as the mass - loss has stopped hence warm dust is not being added .
the objects with detached cold cs would be seen at region via , while objects in zone vib contains objects with warm as well as cold cs .
the zone vii and dashed region is dominated by pagbs with pn like colours .
our sample stars iras 05208 - 2035 , iras 07140 - 2321 , iras 12538 - 2611 and iras
01259 + 6823 belong to zone iiia , iiib and iv where signatures of cs formed by increased mass - loss at late agb are evident .
iras 17279 - 1119 ( zone vib ) seems to possess the hot as well as cold dust shell while iras 07331 + 0021 , iras 22223 + 4327 , iras 08187 - 1905 have pn like colours . [ cols= " < , < , < ,
< , < " , ] @xmath6other refers to other names , @xmath7bin refers to binarity .
@xmath8giridhar et al .
2005 , @xmath9giridhar et al .
2000 , @xmath10guillermo gonzalez et al .
1997a , @xmath11giridhar et al .
1998 @xmath12giridhar et al .
2010 , @xmath13maas et al .
2005 , @xmath14van winckel ( 1996 ) , @xmath15giridhar & arellano ferro ( 2005 ) @xmath16arellano ferro et al .
2001 , @xmath17maas et al .
2007 , @xmath18present work in our study of extended sample for rv tau stars ( giridhar et al . 2005 ) we had proposed two scenarios ( based upon the single star and binary configuration ) and also discussed the parameters strongly affecting the depletion caused by dust - gas separation .
@xmath19 * the surface temperature : * the first important requirement was that the star must be hotter than @xmath20 5000k .
the cooler rv tau stars ( rva ) did not show the abundance anomalies , possibly their massive convective envelopes diluted the effect of accreted gas which was cleaned of refractory elements .
the masses of convective envelopes for pagb stars increases with decreased temperature ( see frankowski 2003 for more details ) varies from 0.016 m@xmath3 at 4000k to 0.0001 m@xmath3 at 6000k . at 6000k
the mass of convective envelope is about 1000 times that of observable photosphere . at hot temperature
the abundance deficiency of readily condensable elements ( those with high t@xmath21 ) may approach a thousandfold and in such cases the surface and convective envelope would be composed of almost undiluted accreted gas . at lower temperatures , in addition to larger convective mass
the stronger stellar wind may oppose the accretion of the clean gas .
@xmath19 * the metallicity : * the dust - gas separation process has not been seen in stars with intrinsic metallicity lower than @xmath20 @xmath11.0 irrespective of their temperatures .
it is likely consequence of the inability of radiation pressure on dust grains to force a separation of dust from gas when the mass fraction of the dust is very low as seen in metal - poor environments .
a dust - gas reservoir with metallicity in excess of @xmath11.0 is mandatory for efficient separation of dust from the gas .
these conditions can be easily realized in binary scenario with circumbinary disc since it may contain gas from either binary companion or the main star before the onset of accretion .
although a revised minimum temperature 4800k for discernible depletion has been identified , but stars in temperature range 6000k to 7500k tend to show large depletions . at hotter side _
t@xmath22_@xmath23 8500k di is smaller ; a shift in ionization structure causing the observable features to become too scarce for s and zn may be responsible . many stars such as cc lyr , hr 4049
, bd+39@xmath04926 show large differences between s and zn despite these elements having nearly the same t@xmath24 . for thick disc and halo stars relative enrichment of s by @xmath250.3 dex
can be anticipated since the s is an @xmath2 element but for larger differences found in the above mentioned objects there is no ready explanation .
another intriguing observation is that di are not correlated with iras colours nor do they show any correlation with ir flux at a given wavelength . since the sed characteristics can be used to develop a model for circumstellar material , de ruyter et al .
( 2006 ) have done optical photometry and have also used the ir and sub - millimeter data for a large sample of known binary pagb stars and other objects ( like rv tau stars ) of similar ir characteristics . using the optically thin dust model they found the dust at or near sublimation temperature ( 1200k ) and also very close to the star ( @xmath20 10 au from the central source ) irrespective of the effective temperature of the star .
these authors argued that at least a part of the dust must be gravitationally bound since any typical agb outflow velocity would bring the dust to cooler region within years .
although their sample of known pagb binary and rv tau stars of similar characteristics had a large range in the size of ir excess , but the shape of ir excess indicated that cs was not freely expanding but was stored in the system .
a study of orbital parameters for the well - studied objects point to orbits which are too compact for sizes of pagb stars .
these authors further suggest that it is highly improbable that these stars have evolved as single stars ; a strong interaction phase while the star was in giant stage is required to explain the present configuration .
the high spatial resolution interferometry in mid ir has been used to resolve compact dusty discs around the depleted binary post - agb stars . for red rectangle ,
hr 4049 and recently for hd 52961 the dusty discs have been resolved ( deroo et al .
it is comforting to see the additional information provided by high spatial resolution interferometry towards better understanding of these systems but the accuracies of disc dimensions are constrained by the radiative transfer models . a good number of pagb candidates displaying basic pagb characteristics like iras colours , high luminosity , small amplitude light and radial velocity variations do not belong to the above mentioned classes .
regarding s - process enhancements , the earlier samples of pagb belonged to only two classes `` have '' and `` have not '' ; the `` have '' showing strong s - process enhancements ( [ s / fe ] @xmath23 1 ] while many of the `` have nots '' had subsolar [ s / fe ] ratios . with increased sample
the moderately enhanced members are making an appearance . iras
17279 - 1119 has been known for some time but iras 01259 + 6823 , iras 05208 - 2035 and iras 08187 - 1905 are new additions .
then there are other pagb stars including rv tauris without significant s - process enhancement nor showing depletions .
we have tabulated 42 such objects ( more than half are rv tauris and semi - regular variable of type d srds ) in table 11
. a large fraction of them ( 30 ) are iras sources , one fourth ( 13 ) are binaries all showing iras fluxes . in terms of temperature range and iras
colours they are indistinguishable from those in table 10 .
pagb with s - process enhancements ( table 9 ) appear to be systematically warmer .
it should be noted that larger di pagb also seem to favour the same temperature range .
then it would appear that condition of low surface temperature are not conducive for the survival of abundance peculiarities although unglamorous objects are found even in among objects in temperatures range of 6000k to 7500k .
in our exploration of post - agb candidates we find mild enhancements of s - process elements in iras 01259 + 6823 , iras 05208 - 2035 and iras 08187 - 1905 , while significant enhancement is found for iras 17279 - 1119 and iras 22223 + 4327 . mild s - process enhancement and low temperatures of iras 05208 - 2035 and iras 01259 + 6823 are not of common occurrence in post - agb stars although the theory does not prohibit such objects since the degree of s - process enhancement gradually builds with number of thermal pulses .
however further evidence like intrinsic luminosity estimates are awaited .
a survey of well established s - process enhanced post - agb stars shows majority of this group to be most likely thick disc objects of intermediate temperatures .
it does not appear that their evolution is strongly affected by the presence of binary companion .
most of these heavily s - processed objects have temperatures between 6000k to 7500k , although a spread in [ s / fe ] for similar atmospheric parameters is seen indicating the influence of additional processes on third dredge - up efficiency .
a contemporary analysis of bd+39@xmath0 4926 has been conducted and initial metallicity of @xmath10.7 dex has been estimated .
hd 107369 appear to be a metal - poor , high galactic latitude supergiant with strong influence of cn processing .
the observed n enhancement is in excess of that predicted by the fdu but can not be ascribed to large rotation at main sequence as the stellar lines are very sharp and narrow @xmath1 not much broader than the instrumental profile .
our compilation of the pagb stars showing the abundance peculiarities caused by selective depletion of refractory elements reiterates the temperature and metallicity limits observed earlier for their detection . the observed spectral energy distribution with dust excess of these objects
have been interpreted by de ruyter et al .
2006 via a circumstellar dust shell contained in keplerian rotating discs .
but the actual resolution of dusty discs and measurement of their sizes has been possible only for heavily depleted objects like hr 4049 , red rectangle and more recently hd 52961 although hd 52961 does not exhibit large ir excess .
search of such depleted objects and their follow - up using high spatial resolution mid ir interferometry and ir spectrometry can provide valuable insight into the circumstellar geometries of these objects .
our compilation of pagb stars showing neither of these two peculiarities is indistinguishable from these two samples in terms of basic stellar parameters .
hence the cause separating these subgroups among pagb stars is not understood .
the extension of abundance analysis to pagb candidates located in uniform stellar system such as globular clusters , or in neighbouring galaxies would help since each system would have small range in mass and metallicity .
dll thanks the robert a. welch foundation of houston , texas for support through grant f-634 . 99 aldenius , m. , lundberg , h. , blackwell - whitehead , r. , 2009 , a&a , 502 , 989 allen , d. m. , porto de mello , g. f. , 2011 , a&a , 525 , 63 arellano ferro , a. , giridhar , s. , mathias , p. , 2001
, a&a , 368 , 250 arkhipova , v. p. , noskova , r. i. , ikonnikova , n. p. , komissarova , g. v. , 2003 , astronomy letters , 29 , 480 asplund , m. , grevesse , n. , sauval , a. j. , 2005 , asp conf.ser.336 , 25 bensby , t. , feltzing , s. , lundstrm , i. , ilyin , i. , 2005 , a&a , 433 , 185 bergemann , m. , cescutti , g. , 2010 , a&a , 522 , a9 bergemann , m. , 2011 , mnras , 413 , 2184 bimont , e. , grevesse , n. , hannaford , p. , lowe , r. m. , 1981 , apj , 248 , 867 bogaert , e. , ph.d .
thesis , k.u .
leuven , 1994 boyarchuk , a. a. , lyubimkov , l. s. , sakhibullin , n. a. , 1985 , astrophysics , 22 , 203 blcker , t. , schnberner , d. , 1991 , a&a , 244 , l43 curry , j.j .
, 2004 , j. phy .
data , 33 , 725 den hartog , e. a. , lawler , j. e. , sneden , c. , cowan , j. j , 2003 , apjs , 148 , 543 deroo , p. , ache , b. , verhoelst , t. , dominik , c. , tatulli , e. , van winckel , h. , 2007 , a&a , 474 , 45 de ruyter , s. , van winckel , h. , maas , t. , llyod evans , t. , waters , l. b. f. m. , dejonghe , h. , 2006 , a&a , 448 , 641 el eid , m. f. , champagne , a. e. , 1995 , apj , 451 , 298 frankowski , a. , 2003 , a&a , 406 , 265 fuhr , j. r. , martin , g. a. , wiese , w. l. , 1988 , j. phys . chem . ref .
data , 4 , 493 fuhr , j. r. , wiese , w. l. , 2005 , crc handbook of chemistry and physics , 86@xmath26 edition fuhr , j. r. , wiese , w. l. , 2006 , j. phys .
chem . ref .
data , 35 , 1669 gehren , t. , liang , y. c. , shi , j. r. , zhang , h. w. , zhao , g. , 2004 , a&a , 413 , 1045 gielen , c. , van winckel , h. , min , m. , waters , l. b. f. m. , lloyd evans , t. , 2008 , a&a , 490 , 725 giridhar , s. , rao , n. k. , lambert , d. l. , 1994 , apj , 437 , 476 giridhar , s. , arellano ferro , a. , parrao , l. , 1997 , pasp , 109 , 1077 giridhar , s. , lambert , d. l. , guillermo gonzalez , 1998 , 509 , 366 giridhar , s. , lambert , d. l. , guillermo gonzalez , 2000 , apj , 531 , 521 giridhar , s. , arellano ferro , a. , 2005 , a&a , 443 , 297 giridhar , s. , lambert , d. l. , reddy , b. e. , guillermo gonzalez . , david yong . , 2005 , apj , 627 , 432 giridhar , s. , molina , r. , arellano ferro , a. , selvakumar , g. , 2010 , mnras , 406 , 290 guillermo gonzalez , lambert , d. l. , giridhar , s. , 1997a , apj , 481 , 452 guillermo gonzalez , lambert , d. l. , giridhar , s. , 1997b , apj , 479 , 427 hannaford , p. , lowe , r. m. , grevesse , n. , bimont , e. , whaling , w. , 1982 , apj , 261 , 736 herwig , f. , 2005 , ara&a , 43 , 435 hrivnak , j. , 1995 , apj , 438 , 341 hrivnak , b. j. , van winckel , h. , reyniers , m. , bohlender , d. , waelkens , c. , wenxian lu . , 2008 ,
aj , 136 , 1557 kelleher , d. e. , podobedova , l. i. , 2008a , j. phy .
chem . ref .
data , 37 , 267 kelleher , d. e. , podobedova , l. i. , 2008b , j. phy .
chem . ref .
data , 37 , 709 kelleher , d. e. , podobedova , l. i. , 2008c , j. phy .
chem . ref .
data , 37 , 1285 kiss , l. l. , derekas , a. , szab , m. , bedding , t. r. , szabados , l. , 2007 , mnras , 375 , 1338 klochkova , v. g. , panchuk , v. e. , 1996 , bull.spec.astrophys.obs , 41 , 5 klochkova , v. g. , panchuk , v. e. , 1998 , astronomy letters , 24 , 650 klochkova , v. g. , panchuk , v. e. , tavolzhanskaya , n. s. , 2010 , astronomy reports , 54 , 234 kodaira , k. , greenstein , j. l. , oke , j. b. , 1970 , apj , 159 , 485 . korotin . s. a. , 2009 , astronomy reports , 53 , 651 kwok , s. , 1993 , ara&a , 31 , 63 kurucz , r.l . , furenlid , i. , brault , j. , testerman , l.,1984 , national solar observatory atlas , sunspot , new mexico : national solar observatory lawler , j. e. , bonvallet , g. , sneden , c. , 2001a , apj , 556 , 452 lawler , j. e. , wickliffe , m. e. , den hartog , e. a. , sneden , c. , 2001b , apj , 563 , 1075 lawler , j. e. , den hartog , e. a. , sneden , c. , cowan , j. j. , 2006 , apjs , 162 , 227 lawler , j. e. , sneden , c. , cowan , j. j. , ivans , i. i. , den hartog , e. a. , 2009 , apjs , 182 , 51 lind , k. , asplund , m. , barklem , p. s. , belyaev , a. k. , 2011 , a & a , 528 , 103 luck , r. e. , lambert , d. l. , bond , h. e. , 1983 , pasp , 98 , 413 luck , r.e . , bond , h.e .
, 1989 , apj , 342 , 476 lyubimkov , l. s. , lambert , d. l. , korotin , s. a. , poklad , d. b. , rachkovskaya , t. m. , rostopchin , s. i. , 2011 , mnras , 410 , 1774 maas , t. , van winckel , h. , waelkens , c. , 2002 , a&a , 386 , 504 maas , t. , van winckel , h. , llyod evans , t. , nyman , l. . , kilkenny , d. , martinez , p. , marang , f. , van wyk , f. , 2003 , a&a , 405 , 271 maas , t. , van winckel , h. , evans , l. t. , 2005 , a&a , 429 , 297 maas , t. , giridhar , s. , lambert , d. l. , 2007 , apj , 666 , 378 martin , g. a. , fuhr , j. r. , wiese , w. l. , 1988 , j. phys .
chem . ref .
data , 3 , 512 mashonkina , l. , korn , a. j. , przybilla , n. , 2007 , a&a , 461 , 261 mashonkina , l. , 2011 , arxiv:1104.4403 mcwilliam , a. , 1998 , aj , 115 , 1640 melndez , j. , barbuy , b. , 2009 , a&a , 497 , 611 mowlavi , n. , 1999 , a&a , 350 , 73 mucciarelli , a. , caffau , e. , freytag , b. , hans - gnter ludwig , bonifacio , p. , 2008
, a&a , 484 , 841 podobedova , l. i. , kelleher , d. e. , wiese , w. l. , 2009 , j. phys .
chem . ref .
data , 38 , 171 prochaska , j. x. , mcwilliam , a. , 2000 , apj , 537 , 57 rao , n. k. , goswami , a. , lambert , d. l. , 2002 , basi , 30 , 671 rao , n. k. , reddy , b. e. , 2005 , mnras , 357 , 235 reddy , b. e. , parthasarathy , m. , 1996 , aj , 112 , 2053 reddy , b. e. , bakker , e. j. , hrivnak , b. j. , 1999 , apj , 524 , 831 reddy , b. e. , guillermo gonzalez . , david yong , 2002 , apj , 564 , 482 reddy , b. e. , lambert , d. l. , allende prieto , c. , 2006 , mnras , 367 , 1329 reyniers , m. , 2002 , ph.d thesis , k. u. leuven reyniers , m. , van winckel , h. , gallino , r. , straniero , o. , 2004 , a&a , 417 , 269 reyniers , m. , van de steene , g. c. , van hoof , p. a. m. , van winckel , h. , 2007 , a&a , 471 , 247 romano , d. , karakas , a. i. , tosi , m. , matteucci , f. , 2010 , a&a , 522 , a32 sansonetti , j. e. , 2008 , j. phys .
data , 37 , 1659 schaller , g. , schaerer , d. , mainer , g. , maeder , a. , 1992 , a&as , 96 , 269 schiller , f. , pryzbilla , n. , 2008 , a&a , 479 , 849 sidmiak , n. , meixner , m. , ueta , t. , sugerman , b. e. k. , van de steene , g. c. , szczerba , r. , 2008 , apj , 677 , 382 sneden , c. , 1973 , ph.d thesis , univ . of texas at austin , usa sneden , c. , lawler , j. e. , cowan , j. j. , ivans , i. i. , den hartog , e. a. , 2009 , apjs , 182 , 80 sobeck , j. s. , lawler , j.e . , sneden , c. , 2007 , apj , 667 , 1267 surez , o. , garca - lario , p. , manchado , a. , manteiga , m. , ulla , a. , pottasch , s. r. , 2006 , a&a , 458 , 173 szczerba , r. , sidmiak , n. , stasiska , g. , borkowski , j. , 2007 , a&a , 469 , 799 takeda , y. , takada - hidai , m. , 1998 , pasj , 50 , 629 takeda , y. , sato , b. , murata , d. , 2008 , pasj , 60 , 781 thvenin , f. , idiart , t. p. , 1999
, apj , 521 , 753 tull , r. g. , macqueen , p. j. , sneden , c. , lambert , d. l. , 1995 , pasp , 107 , 251 van der veen , w. e. c. j. , habing , h. j. , 1988 , a&a , 194 , 125 van winckel , h. , 1995 , ph.d thesis , k. u. leuven van winckel , h. , 1996 , a&a , 319 , 561 van winckel , waelkens , c. , waters , l. b. f. m. , 1996 , a&a , 306 , 37 van winckel , h. , 1997 , a&a , 319 , 561 van winckel , h. , reyniers , m. , 2000 , a&a , 354 , 135 van winckel , h. , waelkens , c. , waters , l. b. f. m. , 2000 , iaus , 177 , 285 van winckel , h. , 2003 , ara&a , 41 , 391 velichko , a. b. , mashonkina , l. i. , nilsson , h. , 2010 , astronomy letters , 36 , 664 venn , k.a . , 1995 , apj , 449 , 839 .
ventura , p. , dantona , f. , 2009 , a&a , 499 , 835 wedemeyer , s. , 2001 , a&a , 373 , 998 wiese , w. l. , fuhr , j. r. , deters , t. m. , 1996 , j. phy .
data , monograph no . 7
wiese , w. l. , fuhr , j. r. , 2007 , j. phy .
chem . ref .
data , 36 , 1287 | we have derived elemental abundances for a sample of nine iras sources with colours similar to those of post - agb stars . for iras 01259 + 6823 , iras 05208 - 2035 ,
iras 04535 + 3747 and iras 08187 - 1905 this is the first detailed abundance analysis based upon high resolution spectra .
mild indication of s - processing for iras 01259 + 6823 , iras 05208 - 2035 and iras 08187 - 1905 have been found and a more comprehensive study of s - process enhanced objects iras 17279 - 1119 and iras 22223 + 4327 have been carried out .
we have also made a contemporary abundance analysis of the high galactic latitude supergiants bd+39@xmath0 4926 and hd 107369 .
the former is heavily depleted in refractories and estimated [ zn / h ] of @xmath10.7 dex most likely gives initial metallicity of the star . for hd 107369
the abundances of @xmath2 and fe - peak elements are similar to those of halo objects and moderate deficiency of s - process elements is seen . iras 07140 - 2321 despite being a short period binary with circumstellar shell does not exhibit selective depletion of refractory elements .
we have compiled the stellar parameters and abundances for post - agb stars with s - process enhancement , those showing significant depletion of condensable elements and those showing neither .
the compilation shows that the s - process enhanced group contains very small number of binaries , and observed [ @xmath2/fe ] are generally similar to thick disc values .
it is likely that they represent agb evolution of single stars .
the compilation of depleted group contains larger fraction of binaries and generally supports the hypothesis of dusty discs surrounding binary post - agb stars inferred via the shape of their sed and mid ir interferometry .
iras 07140 - 2321 and bd+39@xmath0 4926 are difficult to explain with this scenario and indicate the existence of additional parameter / condition needed to explain the depletion phenomenon .
however the conditions for discernible depletion , minimum temperature of 5000k and initial metallicity larger than @xmath11.0 dex found from our earlier work still serves as useful criteria . _
subject headings : stars : abundances stars : agb and post - agb stars : variables : other ( rvtauri ) _ [ firstpage ] post - agb stars , abundances , circumstellar matter : stars . |
the discovery @xcite that many galaxy clusters are also strong sources of x - rays opened a new window on cosmology which has proved fruitful for nearly three decades .
subsequent investigations ( e.g. , * ? ? ?
* ; * ? ? ?
* ) determined that this emission originates in a hot ( @xmath7 ) thermal plasma with electron number densities of @xmath8 .
since the sound - crossing time for pressure waves in this plasma is less than the hubble time , the plasma can be assumed to accurately trace the cluster gravitational potential .
cluster virial masses obtained in this manner , in conjunction with intracluster medium ( icm ) models derived from einstein and rosat observations , have shown an overdensity of baryons relative to the expectation for an @xmath9 , big bang nucleosynthesis universe @xcite , thus providing a powerful challenge to the cosmological orthodoxy . the inverse compton scattering of cosmic microwave background ( cmb ) photons known as the sunyaev - zeldovich effect ( sze) provides another direct probe of the icm .
as first indicated by @xcite , the combination of x - ray and sze measurements on a given cluster yield a _
direct _ measurement of the distance to the cluster .
if the thermal sze decrement predicted by the x - ray data under the assumption @xmath10 is @xmath11 , and the observed decrement is @xmath12 , then @xmath13 is given by @xmath14 such a calculation requires knowledge of the structure of the cluster along the line of sight ; since the x - ray data do not directly provide such information , the clusters are typically assumed to be spherically symmetric .
it then becomes important to select clusters from an orientation - unbiased sample .
since the error in @xmath6 is twice the error in @xmath11 , it is also important to accurately understand the statistical and systematic uncertainties inherent in the x - ray models .
the objective of the penn / ovro sze survey is to determine @xmath6 from observations of the sze in an x - ray flux - limited sample of nearby clusters .
the ovro 5.5-meter telescope is an ideal instrument for this purpose . at 32 ghz
, this telescope has a primary beam of @xmath15 ( fwhm ) and a dual - horn switching angle of @xmath16 . at the mean redshift of our sample , these correspond to @xmath17 and @xmath18 , respectively .
since the gas in clusters is distributed on a scale @xmath19 , and most of the gas is contained within the inner @xmath20 , the 5.5-meter main beam samples an astrophysically relevant scale while the switching only removes @xmath21 of the signal .
the first results from this survey were reported in @xcite : with sze measurements of four clusters ( coma , abell 478 , abell 2142 , and abell 2256 ) and x - ray models from the literature , myers et al .
find @xmath22 .
the accuracy of these results is limited primarily by the x - ray models . with this in mind
, we have undertaken to expand the sample of myers et al . and , using public x - ray data , construct high - quality x - ray models and rigorously evaluate their reliability , taking into account the specific observing strategy employed by our instrument .
the spatial resolution and large field of view of the rosat pspc make it ideal for our purposes . while the pspc does not have the energy or spatial resolution of chandra or xmm
, it has a larger field of view than any instrument on either of these observatories and so imaging analyses of extended objects based on rosat data will continue to be relevant for the foreseeable future
. the density models which result from our analysis will also be useful for near - future experiments capable of measuring the sz distortion resulting from the bulk motion of the cluster gas .
to this end we present predictions for the beam - averaged optical depth for the near - future acbar experiment , a sensitive bolometric receiver which will begin taking observations in antarctica early in 2001 .
this instrument has four frequency channels between 150 ghz and 345 ghz , with matched @xmath23 beams , and will be sensitive to both the thermal and kinematic sz effects . in the following section ,
we first present a brief summary of icm models ( [ sec : icmmodels ] ) , along with the formalism associated with these models to describe bremmstrahlung emission and the sze .
[ sec : sample ] describes our expanded cluster sample , and [ sec : dataanalysis ] the details of the x - ray data analysis , our monte carlo error analysis , and the error budget . in this section and the following one ,
we pay special attention to the effects of central cooling flow emission on our profile models .
we present our results in [ sec : results ] , including mass models and a quantitative assessment of the uncertainties in the sze predictions , as well as total masses and baryon mass fractions derived under the assumption of hydrostatic equilibrium .
we discuss our results and conclude in [ sec : conclusion ] . throughout
we use @xmath24 and @xmath25 unless otherwise specified ; we will comment on the impact that assuming other cosmologies has on our results in
[ sec : conclusion ] . we will report an improved measurement of @xmath6 using the observations of myers et al . and @xcite plus recent observations of abell 399 in a second paper ( paper ii ) .
this measurement relies upon the density models we present in this paper , as well as improved electron temperatures from the literature .
the consequences of this x - ray analysis in the context of our measurement of @xmath6 will also be discussed in paper ii .
one model for the cluster gas which has enjoyed great phenomenological success is the _ isothermal beta model _ @xcite . in this case
, the gas is assumed to be isothermal , and the electrons distributed according to @xmath26 here @xmath27 is the central electron number density , @xmath28 is the spherical metric radius , and @xmath29 is a characteristic scale .
the properties of this model are well - known and extensively tabulated in the literature .
we have assumed that the icm is spherically symmetric . while this is not generally the case for individual clusters , it should be a good description on average for an orientation - unbiased sample ( [ sec : sample ] ) .
one generalization of the isothermal beta model which has some support both observationally ( e.g. * ? ? ?
* ) and theoretically ( e.g. * ? ? ?
* ) is the hybrid isothermal / adiabatic model . in paper
ii we will study the effects of these models on the sze , but since the rosat data are not sensitive to temperature gradients in the outer parts of the cluster , we will assume an isothermal icm for the remainder of this analysis . some authors ( e.g. *
* ; * ? ? ?
* ; * ? ? ?
* ) have studied the possibility of a multi - phase icm . in
[ sec : spectralanalysis ] we will present some preliminary indications of cool icm phases we have found in our analysis .
it is , however , difficult to disentangle the effects of these phases from rosat calibration uncertainties
. we will assess the impact of the observed effect in
[ sec : spectralanalysis ] , but until better data from future missions is capable of accurately constraining icm phase models , we will adhere to the single - phase model .
the bolometric luminosity due to bremsstrahlung emission from an ionized thermal plasma of electrons and protons is @xcite @xmath30 where @xmath31 @xmath32 is the electron number density , @xmath33 is the proton number density , @xmath34 is boltzmann s constant , @xmath35 is the local temperature of the plasma , and @xmath36 is the thermally averaged gaunt factor . for a plasma with the cosmic helium mass fraction @xmath37 and metal abundances of 30% solar , @xmath38 ; this is the value we adopt .
this corresponds to a baryonic mass per electron of @xmath39 and an overall mean molecular mass ( in units of the proton mass ) @xmath40 ; these calculations assume the solar abundances of @xcite . for an isothermal plasma , the bolometric flux observed at a redshift @xmath41 is @xmath42 where @xmath43 is the luminosity distance @xcite @xmath44 { \rm mpc}\ ] ] ( for @xmath45 ) .
the integral on the right hand side is the emission measure ( em ) : @xmath46 for the beta model , the em over all space reduces to @xmath47 if we know the form of the electron density profile @xmath48 , the cluster temperature @xmath35 , the redshift , and the bolometric flux , we may determine the normalization of the density profile : @xmath49 here we have used the fact that @xmath50 . in practice the x - ray spectrum is observed over a finite bandpass with a finite aperture , and is modified by photoelectric absorption due to the intervening intra - galactic medium .
we account for this by modeling the spectrum with xspec , nasa gsfc s standard x - ray spectral analysis program .
xspec reports the normalization of the spectrum ( a quantity analogous to @xmath51 in the foregoing discussion ) through the parameter @xmath52 , defined as @xmath53 here all distances are in cm .
the em in the annulus between @xmath54 and @xmath55 is @xmath56 where @xmath57^{-3 \beta + 3/2}.\ ] ] the central density is then @xmath58^{\frac{1}{2 } } \nonumber.\end{aligned}\ ] ] note that eq .
[ eq : luminosity_distance ] used in deriving this equation assumed @xmath59 ; for @xmath60 this causes at most a @xmath61 error in @xmath27 for @xmath62 .
this is the expression we use to determine the normalization of the cluster density profiles .
@xcite show that the fractional change in intensity due to the inverse compton scattering of cmb photons by a thermal plasma with a velocity @xmath63 along the line of sight is @xmath64 where @xmath35 is the temperature of the plasma , @xmath65 is the optical depth for inverse compton scattering @xmath66 @xmath67 is the dimensionless frequency @xmath68 and @xmath69 @xmath70 is a relativistic correction @xcite , which generally has a magnitude of a few percent that of the leading term for @xmath71 . in eq .
[ eq : tau ] , @xmath72 is the thompson scattering cross section ( @xmath73 ) and the integral is along the line of sight .
since the rosat pspc data are not sensitive to temperature gradients , they are capable of directly constraining the cluster density profiles . for this reason we express our sze analysis in terms of the inverse compton optical depth .
note that this also renders the results we present here independent of the relativistic correction @xmath70 , although this will have to be taken into account when relating the @xmath65 values we present to observed intensity decrements .
a radio telescope will measure the intensity decrement convolved with the instrument beam pattern .
it is then convenient to consider the quantity @xmath74 where @xmath75 is the telescope beam normalized to unity at the maximum .
@xmath76 is typically well described by a gaussian : @xmath77 for the ovro 5.5-meter telescope we have @xmath78 @xcite , and for acbar , @xmath79 .
the 5.5-meter telescope also switches by @xmath16 in azimuth to remove atmosphere and ground emission .
we then define @xmath80 where @xmath81 is the switching vector which , at a given hour angle , corresponding to a @xmath16 offset in azimuth .
since our analysis assumes spherical symmetry for the clusters , we will always compute this quantity at transit . the equivalent quantity to @xmath82
then is @xmath83 in terms of these objects , the equivalent temperature decrement which a single - dish telescope will see for an isothermal cluster is @xmath84 for more details on this formalism , see @xcite and @xcite .
due to the assumption of spherical symmetry necessary to employ the sunyaev - zeldovich effect as a distance measure , it is vital to measure the sze in an unbiased sample . an x - ray selected , x
- ray flux limited sample satisfies this criterion .
the first steps in this direction were taken by @xcite who used the sample of @xcite to define an x - ray flux limited sample of 11 clusters with @xmath0 , @xmath85 ( 2 - 10 kev ) , galactic latitudes @xmath86 , and declinations @xmath87 .
we improve upon this by defining a larger sample .
this has two advantages .
first , it will allow the ( random ) noise caused by departures from spherical symmetry to be reduced in the sample average .
second , it will allow many possible systematic affects to be identified by dividing the sample into several sub - samples by , _
e.g. _ , morphology , optical richness , or x - ray spectral characteristics . as our parent sample
we choose the xbac catalog of @xcite .
ebeling et al .
have analyzed the data of the rosat all sky survey ( rass ) to obtain a sample of 242 abell clusters complete at the @xmath88 level down to a flux limit of @xmath89 in the 0.12.4 kev band . if we conservatively adopt a flux cutoff of twice this ( @xmath90 ) and impose the luminosity cutoff ( @xmath91 for @xmath92 ) corresponding to a volume complete sample out to @xmath93 , we find 31 clusters in the xbac catalog that meet these criteria .
these are listed in table [ tbl : clustersample ] .
since the xbac catalog ultimately uses the list of abell clusters as a parent sample , we must employ some caution in assessing the completeness of our sample . by comparing the overlap regions of the abell @xcite and aco @xcite catalogs ,
@xcite find that abell missed 29% of the richness class 1 or greater clusters which were in aco ; most of these are @xmath94 objects .
these results are consistent with the results of @xcite , who find that @xmath95 of @xmath96 objects are missed in monte carlo simulations with an optical selection algorithm designed to mimic that of abell and aco . in most cases ,
these clusters are missed due to fluctuations in the background galaxy counts , resulting in the misclassification of actual clusters as poorer objects which are then not included in the catalog ; since this misclassification preferentially affects the poorer clusters , most of the missed clusters are again likely to be richness class 1 objects .
our stringent luminosity criterion strongly selects against these poorer objects : only @xmath97 of our 31 cluster sample are @xmath98 , whereas @xmath99 of the 242 clusters in the xbac catalog have @xmath98 .
van haarlem et al .
find that a luminosity cutoff comparable to the one we have imposed , applied after the optical selection , results in a catalog which is free from false detections ( although this does not directly address the issue of orientation bias ) .
based on these considerations , we estimate our sample to be 90% complete for @xmath100 , @xmath101 .
this corresponds to 3 missed clusters at most .
this level of incompleteness will not significantly affect @xmath6 measurements derived from the sample as a whole .
near - future , fully x - ray selected surveys such as the eso reflex survey will provide an important cross - check on the completeness of this sample .
.sze cluster xbac subsample .
columns are : right ascension and declination ( j2000 ) , @xmath102 , @xmath103 , @xmath104 ; @xmath105 , @xmath106 .
fluxes and luminosities are as reported by ebeling et al . and
assume @xmath107 .
[ cols=">,^,^,^,>,>",options="header " , ]
we have defined an x - ray flux - limited sample of 31 nearby galaxy clusters and analyzed rosat pspc data on 22 of these .
the primary focus of our analysis is the quantification of sze modeling uncertainties . using a suite of monte carlo simulations
, we find that on average we predict the inverse compton optical depth with an accuracy of @xmath108 for the ovro 5.5-meter telescope .
these predictions are robust with respect to our cooling - flow modeling strategy and unaffected by realistic pspc systematics .
we have also presented similarly accurate predictions of the inverse compton optical depth for the near - future acbar experiment . while somewhat less robust than the 5.5-meter predictions due to the small angular scales this telescope samples , these predictions should be more than sufficient for real - world applications .
we have also confirmed the @xcite report of an excess column density of @xmath109 towards a478 , but do not find evidence for similar anomalies in any of 21 other clusters .
there appears to be an excess of soft counts in the rosat pspc spectra , similar that reported by @xcite . the mean baryon fraction within @xmath110 is found to be @xmath111 ; within @xmath112 , it is @xmath113 . for @xmath114 , these are @xmath115 and @xmath116 respectively .
these are consistent with the @xmath117 baryon fractions observed ( inside @xmath118 with @xmath107 ) by @xcite in analyses of einstein ipc data on galaxy clusters .
our results are also in good agreement with those of @xcite , who find baryon fractions of @xmath119 within @xmath112 for an analysis of the 44 clusters in the edge sample with rosat pspc data .
our mass fractions tend to be slightly lower on average . in part
this is due to the fact that the x - ray temperatures used by others in the literature are often biased low by cooling flow emission ; the remainder is due to the difference in cluster modeling strategies . for @xmath120 our findings also agree with gas mass fractions determined by @xcite on the basis of interferometric observations of the sze in distant clusters . in a sample of 14 clusters ,
they find a gas mass fraction of @xmath121 within @xmath112 ( at 68% confidence ) . while a number of other systematic x - ray cluster analyses exist in the literature ( e.g. , * ? ? ?
* ; * ? ? ?
* ) , ours is the first to attempt to realistically quantify the effect of the uncertainties in the x - ray modeling on the predicted sze decrement for a given instrument . since the analysis of mohr et al .
is closest to ours in spirit and technique , we have conducted a detailed comparison of their results to ours .
figure [ fig4 ] shows the mohr et al .
determination of the baryonic mass internal to @xmath122 versus our determination of this quanitity .
the overall agreement is excellent : the mean mass ratio ( @xmath123 ) is @xmath124 ( @xmath125 ) , strongly arguing against any systematic differences in our analyses .
one of the most significant outliers is a478 , for which mohr et al .
have used the anomalously low galactic value for @xmath126 ; the effect of correcting for this is shown by an arrow in figure [ fig4 ] .
the scatter between our results , however , is somewhat larger than the @xmath127 scatter which is expected
. this may be in part due to different strategies for profile modeling , spectral extraction , and dealing with cooling flows .
we have attempted to directly assess the robustness of our results , but there are several considerations beyond the scope of this analysis which could affect our conclusions .
the most significant is the possibility of substructure in the icm . @xcite
have studied this in simulations of rosat pspc cluster observations and find a mean overestimate of the cluster density of @xmath128 . since these simulations do not include astrophysically important mechanisms such as cooling and conduction , it will be important to address this issue with the current and upcoming x - ray missions chandra and xmm , as well as more powerful simulations .
also of some concern is the possibility of large - scale temperature gradients in the cluster . while this will not significantly affect the baryonic mass models we present here , it will affect thermal sze predictions and the inferred total masses . for consistency with myers et al . and many other authors we have assumed @xmath129
there is increasing evidence , however , that this may be wrong ( for a summary of the evidence see * ? ? ?
due to the @xmath130 dependence of the baryonic mass on the distance scale , these results will be most affected by any errors in the cosmology we assume ; the inverse compton optical depths and central densities will be least affected .
we have recomputed the sample average of the baryonic mass , baryonic mass fraction , and inverse compton optical depth for two currently viable cosmologies : an open ( @xmath131 ) model and a closed @xmath132 ( @xmath133 ) model . for the closed @xmath132 model , the sample average baryonic mass is increased by 8.8% , the baryonic mass fraction by 5.1% , and the inverse compton optical depth by 1.7% .
for the open model , the baryonic mass is increased by 3.0% , the baryonic mass fraction by 1.8% , and the inverse compton optical depth by only 0.6% .
clearly this translation must be done on a cluster - by - cluster basis for a comparison of our @xmath65 predictions with sz data . in paper ii we will use the models presented in this work , together with improved asca temperatures , to obtain a measurement of @xmath6 from sze measurements conducted with the ovro 5.5-meter telescope .
the wider implications for cosmology of this work will also be discussed therein .
we acknowledge helpful discussions with maxim markevitch , joe mohr , steve snowden , and alexey vikhlinin ; we also thank alexey vikhlinin for the use of his software .
we thank patricia udomprasert for help in some of the data reduction .
bsm was supported for part of the duration of this work by the zacheus daniels fund at the university of pennsylvania ; stm was supported by an alfred r. sloan fellowship at the university of pennsylvania .
in this appendix we present the profiles resulting from our analysis ( figures [ fig5 ] , [ fig6 ] , [ fig7 ] , and [ fig8 ] ) . the details of this analysis are described in [ sec : dataanalysis ] .
although it is conventional in the literature to present such fits on a log - log scale , we have chosen to employ a log - linear scale since this makes the goodness - of - fit at large radius more readily apparent . | we define a 90% complete , volume - limited sample of 31 @xmath0 x - ray clusters and present a systematic analysis of public rosat pspc data on 22 of these objects .
our efforts are undertaken in support of the penn / ovro sze survey , and to this end we present predictions for the inverse compton optical depth towards all 22 of these clusters .
we have performed detailed monte carlo simulations to understand the effects of the cluster profile uncertainties on the sze predictions given the ovro 5.5-meter telescope beam and switching patterns .
we also present a similar analysis for the near - future acbar experiment . for most of the clusters in the sample
we find less than a 5% uncertainty in the sze predictions due to an imperfect knowledge of the profile .
a comparison of different cooling flow modeling strategies shows that our results are robust with respect to this .
the profile uncertainties are then one of the least significant components of our error budget for sze - based distance measurements . the density models which result from this analysis also yield baryonic masses and , under the assumption of hydrostatic equilibrium , total masses and baryon mass fractions .
our monte carlo profile analysis indicates that the baryon masses within @xmath1 mpc for these clusters are accurate to better than @xmath2 and unaffected by realistic pspc systematics . in the sample as a whole
, we find a mean gas mass fraction of @xmath3 internal to @xmath4 mpc .
this is in agreement with previous x - ray cluster analyses , which indicate an overabundance of baryons relative to the prediction of big bang nucleosynthesis for an @xmath5 universe .
our analysis of the x - ray spectra confirms a previous claim of an excess absorbing column density towards a478 , but we do not find evidence for anomalous column densities in the other 21 clusters .
we also find some indications of an excess of soft counts in the rosat pspc data .
a measurement of @xmath6 using these models and ovro sze determinations will be presented in a second paper . |
the emerging field of cold molecules is a very vibrant topic in physics and physical chemistry .
the considerable interest in this topic is related to the properties of cold molecules and their many potential applications .
cold molecules have been identified as attractive systems for ultrahigh - resolution spectroscopy @xcite , quantum information processing @xcite , for developing new time standards and testing fundamental physical theories such as the time variation of physical constants @xcite , the existence of a dipole moment of the electron @xcite , and for the measurement of parity violation @xcite . in chemistry ,
cold molecules are essential tools to explore quantum - mechanical effects in chemical reactions .
in contrast to atoms , molecules have a very complicated level structure that consists of vibrational and rotational states as well as electronic levels .
this abundance of states is the main obstacle for direct laser cooling of molecules .
usually , molecules do not provide the closed transitions required for cooling and non - destructive state - selective detection .
this makes it impossible to perform direct spectroscopic measurements on _ single _ molecules a standard technique in atomic physics .
additional complications result from the small energy differences between the rotational levels , leading to a thermal distribution of the population over the molecules ro - vibrational states . despite important achievements @xcite , the control of molecular states never caught up with that of atomic systems
. however , there has been remarkable progress in the synthesis of ultra - cold alkali dimers from samples of ultra - cold atoms ; see e.g. , refs . @xcite .
furthermore , methods which enable the preparation of more diverse ( e.g. polyatomic ) cold molecular species in their vibrational ground - states have been successfully demonstrated .
these methods include : supersonic beam expansion followed by stark deceleration @xcite , optical stark deceleration @xcite , electrostatic velocity selection @xcite , collisional cooling in crossed molecular beams @xcite and buffer gas cooling @xcite .
the vibrationally cold ( but rotationally hot ) states that result will be taken to be the starting point for the schemes described in this paper .
the experimental advances which have enabled the production of these cold molecular states have inspired theoretical investigations of the cooling of molecules by laser pulses @xcite or even by coupling molecules to an optical cavity @xcite .
bartana et al .
@xcite used the electronic excited - state as a heat reservoir in order to cool the vibrational states of the electronic ground - state by means of short , shaped , laser pulses . in later work @xcite they employed state selective optical pumping , hiding
the target state in a dynamically trapped state . through this ,
bartana et al . achieved a vibrational ground state population of 97% after only 25 vibration periods .
a related scheme , investigated by schirmer @xcite , increased the vibrational ground - state population to a similar level .
these efforts to cool the internal degrees of freedom focus on the widely spaced vibrational states of molecules . in ref .
@xcite bartana et al .
investigated the possibility of cooling the _ rotational _ degrees of freedom in a simplified model employing the same techniques .
however , even though the results of their calculations are very promising , the model has limited application as it neglects the vibrational degree of freedom of the molecule .
in contrast to neutral molecules , ionized molecules can be sympathetically cooled by trapping them alongside atomic ions in a paul trap . under these conditions ,
state sensitive ultra - cold chemical reactions have been measured @xcite and high resolution spectroscopy has been demonstrated on small ensembles @xcite . despite these achievements , the rotational degrees of freedom
could not be controlled , but they led to new laser cooling schemes for the internal states of molecules @xcite exploiting the unique properties of these systems . in the present paper , we show that by means of purely coherent manipulations of internal states ( i.e. rotational states ) , and by using sympathetic cooling of motional states , single molecular ions can be cooled close to their motional and rotational ground - state .
the internal vibration of the molecule needs to be initially cold , in order for the final state to be cold in _ all _ its degrees of freedom .
this initial state can be achieved using the existing methods described above .
then the cooling of the internal molecular state is achieved in three steps : first a laser cooled atomic ion is trapped alongside the molecule and a common mode of vibration is used to prepare the molecule in its motional ground - state .
next , using adiabatic passage methods @xcite , the thermal internal state of the molecule can be mapped onto the molecule s motion in an external trapping field with high fidelity . during this mapping process the internal ( rotational ) states are transferred to the ground - state whereas the molecule s motion is excited .
finally , the molecule s motion is sympathetically cooled back into its ground - state using the common mode with a cooled atomic ion . by doing this without exciting other degrees of freedom , a molecule which is vibrationally , rotationally , and translationally cold
can be obtained .
the overall process is a kind of molecular `` heat pump '' : the heat energy in the rotational degree of freedom is transferred by the coherent processes to the motional degree of freedom .
this heat energy is in turn transferred to the environment by means of conventional cooling techniques , e.g. side - band cooling : this ensures that the process is uni - directional .
vogelius et al .
@xcite investigated the cooling of molecular ions by coupling a single rotational state to the motion of the ion . by means of black - body radiation
the population is pumped into the rotational ground - state resulting in a ground - state population of about 80% after a cooling time of the order of minutes . in this paper
we focus on the cooling process by employing techniques from coherent control providing much faster and more efficient cooling .
the result is a robust and highly efficient cooling process which enables the deterministic manipulation of the internal states of molecules .
we will examine adiabatic passage schemes which exhibit high fidelity in conjunction with relaxed requirements on the experimental parameters compared to direct raman transitions .
with cooling times of the order of milliseconds , and final ground - state populations of more than 92% , the proposed scheme provides a fast and efficient method for preparing molecules in their ro - vibrational ground - state .
even though we focus here on the cooling of molecular rotational levels , the technique is also applicable to other multi - level systems .
so the technique can also be employed in atoms with complicated level schemes , or more general ro - vibrational states of molecules .
the paper is organized as follows : in section [ sec:2 ] , we present the model used for our calculations . to find the best adiabatic passage process for our application we compare the results of numerical simulations for stimulated raman adiabatic passage ( stirap ) , stark chirped raman adiabatic passage ( scrap ) and chirped adiabatic rapid passage ( carp ) in a @xmath0-type level system in section [ sec:3 ] .
we also choose parameter ranges which are relevant for a possible experimental implementation .
section [ sec:4 ] contains the results for an extended level scheme , and in section [ sec : conclusion ] we present our conclusions .
the model we employ in this paper is based on the states of a quantum mechanical rigid rotator which is a good approximation for the rotational states of small diatomic molecules .
however , the techniques described here are applicable to most other level structures with the sole requirement that allowed raman transitions between the states involved exist . in general , the method can be applied to ro - vibrational states of molecules as well as zeeman and hyperfine levels . in order to simplify the discussion , and to avoid specializing to molecules with specific symmetries
, we will not apply specific selection rules to the raman transitions involved .
nevertheless , the results presented in this paper can be applied to a particular system by imposing the specific selection rules for that case with an appropriate relabelling of states .
for example , we could utilize @xmath1 for linear molecules .
we take the energy @xmath2 of the rotational levels of the electronic ground state to be @xmath3 , with the rotational quantum number @xmath4 and a rotational constant @xmath5 . in order to limit the number of levels in our calculations , only rotational levels up to a cut - off
are considered , i.e.@xmath6 . for typical , light , diatomic molecules at room temperature only rotational states with @xmath7
are significantly populated .
this decreases to below 10 states for rotational temperatures lower than about 50k .
this kind of temperature can be easily achieved by supersonic beam expansion @xcite . in our scheme the levels @xmath4
are coupled by laser pulses to an electronically excited state @xmath8 ( see fig.[fig : levelsystem ] ) . and
the rotational states @xmath9 .
the excited state decays towards the states @xmath10 at a rate @xmath11 and towards the uncoupled state @xmath12 at a rate @xmath13 .
the laser pulses have time - dependent rabi frequencies @xmath14 and are detuned from the relevant transition frequency by @xmath15 . ] from the excited state the molecule can spontaneously decay back into the electronic ground - state with a rate @xmath16 . to represent the decay of the excited state into levels outside the system , e.g. rotational states with @xmath17 or vibrationally excited states , an additional level @xmath12 is also included in the model .
the excited state can decay into this uncoupled state with a rate @xmath18 .
the single molecule is trapped in a harmonic potential as provided by rf - traps for molecular ions .
the quantized motion of the molecule in the trapping potential gives rise to an equally spaced ladder of motional states in addition to the internal states of the molecule .
to take the molecule s motion into account , we use the standard notation @xmath19 with @xmath20 representing the states @xmath10 , @xmath21 and @xmath12 .
the quantum number describing the motional state is @xmath22 . as discussed in the introduction , in order to prepare the molecule s motion in its ground - state , it is trapped alongside one or more atomic ions which can be directly laser cooled .
the two types of ion form a crystal - like structure due to their mutual coulomb repulsion , which enables sympathetic cooling of the molecule . in experiments with two types of atomic ion , ground - state populations of better than 95% have been achieved @xcite . in the molecular case
, the minimal system would be one trapped molecular ion and one trapped atomic ion . in this case
, we find that due to the frequency splitting of the motional modes , one mode ( com or stretch mode ) can be singled out and used for the proposed scheme .
the other mode only imposes an additional limit on the laser pulse length as we will discuss later in this paper . the laser - molecule interaction , in the rotating wave approximation @xcite ,
is described by the following hamiltonian @xcite : @xmath23\left\vert j-1\right\rangle\left\langle j-1\right\vert \\ & + \sum_{k = s , p}\left[\hat{h}^k_{c}(t)+\hat{h}^k_{r}(t ) + \hat{h}^k_{b}(t)\right ] , \end{split}\ ] ] where @xmath24 represents the carrier resonance @xcite for either the pump @xmath25 , or the stokes - pulse @xmath26 , with @xmath27 the corresponding red sideband transitions in eq .
are the @xmath28 , which are given by @xmath29 and for the first blue sideband transitions the @xmath30 are given by @xmath31 in eqs .
( [ eq:1]-[eq:4 ] ) , the secular frequency of the molecule in the external trapping potential is @xmath32 and @xmath33 are the corresponding lamb - dicke parameters @xcite .
the detuning between the @xmath34-th laser frequency @xmath35 and the transition frequency @xmath36 of the @xmath37 transition is @xmath38 and @xmath39 .
the raising operator for the internal states is @xmath40 and the lowering operator is @xmath41 .
the creation and annihilation operators for the motional number states @xmath42 are @xmath43 , and @xmath44 respectively .
note that we work in an interaction representation with explicit time - dependence removed , and keep both the resonant and non - resonant couplings .
the reason for this is that in pursuing the adiabatic limit in section [ sec:3 ] we will consider `` strong '' rabi frequencies ( @xmath45 ) which do not allow us to make a second rwa on the hamiltonians for the sideband transitions .
we start our calculations with a molecule in an internal thermal state such that it is already cooled in its internal vibrational mode ( e.g. by the methods mentioned in the introduction ) and such that the excited state @xmath8 is not populated .
we also assume that the molecule s vibrational motion in the trap has been cooled ( e.g. by sympathetic sideband cooling @xcite ) so that only the manifold of rotational states @xmath46 are populated .
the density matrix of this initial state is given by : @xmath47 where @xmath48 is the cut - off introduced for the numerical calculations .
the normalization factor @xmath49 is given by @xmath50 with @xmath51 , and with @xmath52 as the internal rotational temperature of the molecule .
starting from such an initial distribution , we will apply coherent control techniques to transfer population between the different states .
we aim to have a state mapping of the form @xmath53 where @xmath54 are the populations of the initial states @xmath55 and the final target states @xmath56 respectively .
a sequence of pulses will be used to map population in each @xmath4-state to a corresponding @xmath22-state with @xmath57 . throughout this work , we assume that the system is initially prepared in a state given by eq .
( [ eq:5 ] ) , and derive the requirements for achieving the state mapping in eq . ,
i.e. after completion of a number of the coherent pulse sequences .
thus we obtain a superposition of just the motional states which can then be cooled to the motional ground - state @xmath58 by applying the sympathetic cooling @xcite .
[ the state mapping can also be employed for non - destructive state detection : e.g. by measuring the initial thermal distribution ( [ eq:5 ] ) . by coupling the electronic state of an atom trapped alongside the molecules to its motion , the mapped state can be read out . ] for the coherent mapping the key idea is to use pairs of pulses , @xmath59 and @xmath60 , to induce population transfer between the states @xmath61 and @xmath62 , see fig.[fig : levelsystem ] . for this step
it is important to have a resonance , so that additional states do not get strongly involved and disturb the mapping . here , the resonance is arranged so that the quantity @xmath63 is conserved at each step
( although this is the simplest way to make the mapping , it would not be the only way , as we only require the transfer of population between unique pairs of states . ) repeating the @xmath64 step @xmath48 times , where in each step @xmath4 is decreased by one , the distribution of population can be moved to the @xmath57 motional states as in eq .
( [ eq:6 ] ) .
since states with @xmath65 are involved in the intermediate steps , it is clear that if we do not start in the motional ground - state the final state need not be entirely @xmath57 .
however , when we start in the motional ground - state ( i.e. @xmath66 ) , the population transfer is uni - directional .
hence , the population is transferred solely to the lower lying rotational states . in our analysis
we numerically integrate the master equation for the density matrix @xmath67 : @xmath68+\hat{\mathcal{l}}(\rho(t))+\hat{\mathcal{l}}_{{\scriptscriptstyle}u}(\rho(t)).\ ] ] the hamiltonian @xmath69 is given in eq . .
the two liouville terms @xmath70 and @xmath71 describe the irreversible decay of the excited electronic state @xmath72 to the rotational levels @xmath73 of the electronic and vibrational ground - state @xmath74 and the decay out of the system , i.e. to the uncoupled state @xmath12 , @xmath75 in eq .
the raising and lowering operators @xmath76 have the same form as @xmath77 , with the substitution @xmath78 .
we will assume that the laser pulses are gaussian , with a fixed width @xmath52 and rabi frequencies @xmath79 where @xmath80 .
the delay between the two pulses @xmath81 and @xmath82 that drive the transition @xmath83 is @xmath84 , whereas the delay between the @xmath4 and @xmath85 pulse pair is @xmath86 .
the corresponding detunings will be either constants or have the form of time - dependent frequency chirps .
we take all the lamb - dicke parameters to be the same , i.e. @xmath87 , and for simplicity we assume that the decay rates are the same , i.e.@xmath88 . in order to compare the three methods ( stirap , scrap and carp ) in section [ sec:3 ] we characterize the population transfer efficiency by a parameter @xmath89 representing the total population of the rotational ground - state after the transfer : @xmath90 where @xmath91 is the population of the @xmath56 state after the transfer . for an efficient state mapping
, the ground - state population ( @xmath57 ) will have increased . however
, if the initial state also has some population in the @xmath92 state , the efficiency @xmath89 may also decrease due to laser - induced transfer out of @xmath93 .
thus the definition is not only a measure for the transfer efficiency , but also for the uni - directionality of the mapping process . for ideal state mapping
the efficiency measure reaches the limit @xmath94 . in order to test the various passage methods in the next section
, the initial state is taken to be a mixture of 70% rotationally excited states ( @xmath95 ) and 30% ground - state ( @xmath57 ) population for each adiabatic method .
this approach will test how uni - directional the scheme is .
to transfer the population from the state @xmath9 to @xmath96 , various coherent processes can be employed . here
we focus on stimulated raman adiabatic passage ( stirap ) @xcite , stark chirped raman adiabatic passage ( scrap or siarp ) @xcite , and chirped adiabatic rapid passage ( carp ) @xcite which offer highly efficient population transfer in combination with robustness against variations of the pulse parameters . in order to simplify the numerical simulations we first investigate these processes in a system with just two rotational levels .
that is , we examine in detail a single step in our multi - pulse coherent transfer scheme . in this case the most important states are the two lowest motional states for @xmath57 and the lowest ( @xmath66 ) motional state for @xmath97 ( see fig . [
fig : lambdasystem ] ) .
-configuration used to investigate the transfer efficiency of various adiabatic passage schemes in section [ sec:3 ] . the pump and stokes pulses , @xmath98 and @xmath99 respectively ,
will induce a two - photon raman transition ( which may be chirped ) from state @xmath100 to state @xmath101 .
the best results are obtained if the initial population in @xmath102 remains there .
the off - resonant intermediate states @xmath103 decays towards the two rotational states and towards the uncoupled state @xmath12 at a rate @xmath18 .
the decay rate @xmath16 represents decay from the levels @xmath104 to the levels @xmath105 .
other off - resonant states are included as shown . ] however , to include the off - resonant effects all nine of the states shown in fig .
[ fig : lambdasystem ] are included in the numerical calculation .
we choose parameter ranges which are relevant for an experimental implementation .
the electronically excited state can decay to both rotational states as well as into the uncoupled state @xmath12 .
stirap is widely used in the optical control of molecules @xcite where a stokes pulse and a pump pulse are used in a `` counter - intuitive '' order to transfer the population between two states .
the main requirement for stirap is the two - photon resonance condition which corresponds to @xmath106 for a transition on a motional sideband ( see fig . [
fig : levelsystem ] ) .
in addition to this the pulse area must be large , @xmath107 , and the delay between the stokes and pump pulse must be of the order of the pulse width , @xmath108 , to ensure the adiabatic evolution of the system .
another constraint arises from the necessity to address particular motional sidebands .
the narrow splitting of the motional states imposes the use of laser pulses with narrow bandwidth @xmath109 @xcite .
fast pulses will inevitably result in the coupling of the target state to other , close - lying states .
this in turn will reduce the transfer efficiency .
furthermore , the resolved sideband condition @xcite requires that the rabi frequencies are small compared to the trap frequency , i.e.@xmath110 .
thus , only slow and weak pulses can be used which , as we will see , substantially reduces the efficiency of stirap .
however , this can be overcome by employing other adiabatic passage schemes as described below . + in fig .
[ fig : stirapa ] the transfer efficiency @xmath89 is plotted for different pulse widths @xmath52 and delay times @xmath111 . as mentioned above the initial mixed state
is described by the populations @xmath112 and @xmath113 .
the chosen decay rate of @xmath114 is in the typical range of values for the decay rate of an electronically excited molecule , when compared to a typical trap frequency @xmath32 ( of the order of several mhz ) .
the results in fig .
[ fig : stirapa ] show that the efficiency @xmath89 is below 76% .
the performance of stirap for short pulses is relatively poor due to the violation of the adiabaticity requirement ( @xmath115 ) and the limitation on the rabi frequency @xmath116 posed by the resolved sideband condition ( @xmath117 ) . for long pulses
the efficiency of stirap is compromised by the spontaneous decay @xmath118 , so the compensation of small rabi frequencies by long laser pulses is not an option .
therefore , the efficiency of stirap in this parameter range is low and the population transfer is governed by optical pumping rather than coherent evolution .
this is particularly visible in the plateau region for @xmath119 . in this regime
the pulse delay is too large to sustain the adiabatic evolution of the system , which results in a net loss of the ground state population . highly efficient
, fast stirap between motional states requires large motional frequencies which are beyond current ion - trap technology .
we can try to suppress the excited - state population by detuning the raman transition from the excited state .
however , it has been known for some time that detuning adversely affects the stirap process in the absence of decay @xcite . with decay present
one has to consider the balance of the adverse effect of detuning against a possible reduction in spontaneous emission from the excited state of a model @xmath0 system .
studies with such systems support the suggestion that the minimal losses ( for moderate decay rates ) are found by remaining on resonance @xcite .
figure [ fig : stirap - detuning ] shows how the efficiency of stirap drops , for our model system , as we detune from resonance .
the parameters are those for fig .
[ fig : stirapb ] , with the rabi frequency @xmath116 chosen to be at the peak of efficiency in fig .
[ fig : stirapb ] .
we see that both with , and without , decay processes it is best to be resonant . in the case
@xmath120 the resonance is much sharper , however .
( @xmath121 ) .
we show both the case with decay , @xmath122 ( solid line ) , and without decay , @xmath123 ( dashed line ) .
the other parameters are as in fig .
[ fig : stirap ] with the best values taken for @xmath124 , @xmath125 , and @xmath126 . ] to exploit the adiabatic evolution of the system and to suppress the excited - state population by far detuning the raman transition leads us to stark chirped raman adiabatic passage ( scrap ) and chirped adiabatic rapid passage ( carp ) , which will be discussed in sections [ sec:32 ] and [ sec:33 ] . in the limit of far detuning ,
the excited state population is strongly suppressed and the system s dynamics are effectively that of a two - level system @xcite .
the effective raman coupling between the states @xmath9 and @xmath127 is @xmath128 and the effective splitting of the coupled rotational levels is @xmath129 with the effective two - photon raman detuning @xmath130 , and the two stark shifts @xmath131 ^ 2/4\delta^s_{{\scriptscriptstyle}j}(t)$ ] and @xmath132 ^ 2/4\delta^p_{{\scriptscriptstyle}j}(t)$ ] induced by the stokes and pump pulse respectively . within this effective two - level model , adiabatic rapid passage techniques ( arp ) @xcite can be applied . the main idea behind arp
is to drive the system through the resonance ( @xmath133 ) adiabatically , to achieve a complete population transfer .
the technique of stark chirped raman adiabatic passage ( scrap ) @xcite takes advantage of the stark shifts whereas the chirped adiabatic rapid passage ( carp ) @xcite uses overlapping laser pulses along with frequency chirps to transfer the population .
we turn to these methods in the next sections . for the case of scrap (
earlier known as self - induced adiabatic passage , or siarp @xcite ) the laser pulses are engineered so that the system undergoes an avoided level crossing ( @xmath134 ) near a maximum of the effective coupling @xmath135 induced by the stark shifts due to the delay of the pump and stokes pulses . in order to obtain an efficient population transfer the system needs to evolve adiabatically in the crossing region @xcite . for gaussian pulses
this leads to the condition : @xmath136 + figure [ fig : siarp_a ] shows the efficiency @xmath89 for various rabi frequencies and pulse delays for a decay rate of @xmath114 and a fixed pulse length of @xmath137 .
the behavior described by the adiabaticity requirement is clearly visible . for a fixed rabi frequency
the efficiency decreases with increasing pulse delay @xmath111 as predicted by eq .
( [ eq:14 ] ) .
the improvement due to increased rabi frequency is also evident .
however , for large rabi frequencies the pulse violates the resolved sideband condition @xmath138 , leading to a sudden deterioration of the efficiency @xmath89 at large intensities , see fig .
[ fig : siarp_b ] . for small rabi frequencies
the efficiency @xmath89 strongly depends on the delay times . for @xmath139
the method is not robust .
as the effective decay rate increases with increasing rabi frequencies , the efficiency slowly degrades for greater laser intensity .
this leads to a ridge in the @xmath111-@xmath140 diagram . because the efficiency depends on the pulse delay and the rabi frequency , accurate knowledge of these pulse parameters
is required .
this can be moderated by increasing the detuning of the laser pulses . because the constraints on the pulse length are less severe than for stirap the population transfer can be faster with scrap . together with the far detuning this leads to a improved robustness against the detrimental effect of spontaneous decay @xcite .
another way of achieving an adiabatic population transfer is the application of simultaneous pump and stokes pulses ( @xmath141 ) with one laser having a frequency chirp .
this is a raman chirped adiabatic passage , sometimes called rcap @xcite , though here we refer to it as chirped adiabatic rapid passage ( in a @xmath0-system ) or carp . with this system
the stark shifts are eliminated , and the detuning @xmath142 reads @xmath143 in order to ensure the adiabatic evolution the chirp rate @xmath144 needs to fulfill @xmath145 and @xmath146 [ with the two - photon rabi frequency @xmath147 given by eq . ] .
these conditions arise from a landau - zener adiabaticity and from requiring completion of a landau - zener transfer within the time - scale of the pulse .
because there is no limit on the pulse duration arising directly from the adiabaticity requirements the transition can be fast : it is only limited by the narrow bandwidth condition @xmath148 @xcite .
this in turn reduces the susceptibility to spontaneous emission .
the resolved sideband condition @xcite requires that @xmath149 , which is easy to satisfy since the system is in the far - detuned limit @xmath150 . under these conditions the population transferred to the target state can be estimated with the landau - zener formula @xcite @xmath151 + where @xmath152 .
this behavior is confirmed by our numerical simulation [ see fig.[fig : carp_a ] ] .
it shows the simulated efficiency for different rabi frequencies and chirp rates for a system initially in a state with @xmath153 and @xmath113 .
the efficiency increases rapidly with increasing rabi frequency @xmath116 until it reaches a plateau . in this region
the transfer efficiency is well above 98% . for small rabi frequencies
the efficiency @xmath89 deteriorates with increasing chirp rate .
however , this effect diminishes for large pump intensities . similarly to scrap
, the increase in the effective decay rate for high rabi frequencies leads to a slow degradation of the transfer efficiency for large pulse intensities .
for very high laser intensities the efficiency rapidly drops due to the violation of the resolved sideband condition @xmath154 . in a different parameter regime , that of small chirp rates @xmath155
, the efficiency oscillates with changing rabi frequency , see fig .
[ fig : carp_b ] . here
the evolution is governed by rabi oscillations between the two rotational states which together with the finite pulse length leads to large fluctuations in the efficiency . in this regime a very precise control of the laser pulses
is required , so carp is not robust for very small chirp rates . for large rabi frequencies ,
the efficiency is essentially independent of the chirp rate and carp provides the best robustness against uncertainties in the pulse parameters .
both of the adiabatic rapid passage methods , scrap and carp , provide fast population transfer . together with the large detuning of the raman transition from the excited state they provide robustness against the adverse effect of spontaneous emission .
this is clearly evident in fig.[fig : comparison ] where the efficiency of the three methods is plotted against the decay rate .
-configuration of fig .
[ fig : lambdasystem ] , for different decay rates @xmath156 and the three methods we consider .
the parameters are optimised at @xmath122 for each of the three methods and are listed separately in the following .
for stirap ( solid line ) : @xmath157 , @xmath124 , @xmath125 , and @xmath158 . for scrap ( dashed line ) :
@xmath159 , @xmath137 , @xmath160 , and @xmath161 . for carp ( dotted line ) : @xmath162 , @xmath137 , @xmath163 , @xmath164 , @xmath165 with @xmath166 .
other parameters which are fixed for all three cases are : @xmath167 , @xmath168 , @xmath169 , @xmath112 , and @xmath113 . ] in this figure , the values of the parameters @xmath116 , @xmath52 and @xmath111 are optimised for @xmath170 and the detunings given .
( in the case of carp , the value of @xmath144 is also optimal . )
the same optimal values have been used for fixed parameters in figs .
[ fig : stirap][fig : carp ] .
the parameter @xmath171 has been fixed to a reasonable value for diatomic molecules like n@xmath172 or co@xmath173 .
this optimisation gives a fairly wide range of parameters in fig.[fig : comparison ] .
[ fig : comparison ] also shows that , even though stirap is efficient for small decay rates , it decreases rapidly for larger decay rates .
scrap and carp are far more efficient and their suppression of the excited - state population exceeds that of stirap within the limits imposed by the experimental requirements .
even though scrap and carp exhibit similar efficiencies carp is more robust against uncertainties in the laser pulse parameters .
hence , we use carp for our investigation of the population transfer in a multilevel system with @xmath174 in the next section .
+ having derived the conditions for efficient population transfer with chirped two - photon raman transitions , we turn now to systems where we include a larger number of rotational states in the calculation ( @xmath176 ) .
consequently the mapping process involves multiple ( @xmath48 ) pairs of laser pulses .
we investigate the state mapping given in eq.([eq:6 ] ) , for a thermal initial state distribution with @xmath177 .
this corresponds to a system where approximately six rotational levels are significantly populated and consequently we will take @xmath174 .
in our simulation we use a lamb - dicke parameter of @xmath169 and a decay rate of @xmath170 which for realistic trap frequencies ( of the order of a few mhz ) corresponds to typical decay rates of electronically excited states of diatomic molecules .
the pump laser is detuned by @xmath178 , and the stokes laser is chirped with the rate @xmath179 .
both lasers have a peak rabi frequency of @xmath180 , a pulse length of @xmath137 and a delay between successive pulse pairs of @xmath181 .
these parameters were chosen on the basis of the simulations with two rotational levels ( section [ sec:3 ] ) and realistic experimental parameters . in fig .
[ fig : multilevel1 ] the initial population distribution over all states @xmath9 is plotted , along with the final distribution in fig .
[ fig : multilevel2 ] . apart from some small population loss , see fig .
[ fig : losses ] , and some weak scattering of population into states other than the states @xmath182 , the two distributions agree very well .
the total population transferred into the rotational ground - state is @xmath183 , while the total population loss into the uncoupled states is only @xmath184 .
, see fig . [
fig : multilevel ] .
the detuning for the pump pulse is equal to @xmath164 and the rabi frequency is @xmath185 .
other parameters are as in fig .
[ fig : multilevel ] . ] the remaining 6.5% of the total population is mostly left in the initial states @xmath186 [ see fig . [
fig : multilevel2 ] ] .
losses due to spontaneous emission can be further reduced by increasing the detuning .
the population which remained in the higher , coupled rotational states @xmath187 can be transferred into the ground - state by sympathetically cooling the molecule s motion and reapplying the cooling pulse sequence .
using this approach the population of the ground - state can be even further increased from 92% to above 98.4% with a total loss into the uncoupled state of only 1.6% .
additional simulations with @xmath188 and @xmath189 , resulted in higher efficiencies , @xmath190 and @xmath191 , respectively .
the corresponding losses are less than @xmath192 for @xmath188 and @xmath193 for @xmath194 . for large numbers of populated states
the transfer efficiency can be estimated with the following equation : @xmath195 where @xmath89 is the transfer efficiency for the simple @xmath0-system , and @xmath196 is the initial population distribution ( [ eq:5 ] ) .
this agrees well with our numerical simulation of six rotational levels .
starting from a vibrationally cold system , these values of the efficiency @xmath89 show that we can reach the motional and rotational ground - state of molecules @xcite . for a co@xmath173 ion with a rotational temperature of @xmath197k , the first 15 rotational levels are significantly occupied initially with a ground - state population of only 3% . by applying the carp state mapping with a detuning of @xmath178 and taking the selection rule @xmath198 into account the population of the two lowest lying states can be
can be increased to 85% . using a detuning of @xmath194 this
can be improved to 97.8% for a single cooling cycle . for a trap frequency of 4 mhz
this cooling cycle will be completed within 10 ms .
in this work we presented an efficient method to cool the internal states of molecules by means of coherent processes ( and sympathetic cooling ) thus suppressing the problematic spontaneous decay into uncoupled states . by coupling the internal molecular state to the motion of the molecule , that internal state can be mapped onto a motional state . utilizing this , the _ internal _ state is cooled close to its ground - state if the molecule s motion was initially reduced to the motional ground - state through sympathetic cooling .
ultimately all the degrees of freedom of the molecule can be cooled by the application of sympathetic cooling to the final motional excitation . due to its high efficiency
the method presented here is not only useful to cool the internal state , but can also be employed to detect the internal state of the molecule by measuring its motional state with an atom which is trapped alongside .
we have studied various adiabatic methods for a range of laser pulse parameters which are relevant for an experimental implementation of this cooling scheme .
the motion of the ion imposes restrictions on the dynamics of the population transfer process which severely limit the possible parameter range for the laser pulses . for the near - resonant method ( stirap ) ,
population transfer efficiency is very low accompanied with a large population of the excited state .
population losses can be suppressed , if far - detuned chirped adiabatic two - photon raman passage methods are employed .
schemes that use chirped laser pulses ( carp ) , or self - induced adiabatic passage ( scrap / siarp ) by stark shifting the transition frequencies , turned out to be very efficient .
when it comes to the comparison of carp and scrap the former method has the advantage of easy optimization since it has no dependence with respect to pulse shape .
furthermore , for both methods , and unlike stirap , the resolved sideband condition imposes less severe constraints on the useful parameter space .
the requirements for all three methods were derived with simulations for a @xmath0-system . using the results from this simple model
, we were able to demonstrate the applicability of carp in systems with more than two rotational states .
for far - detuned transitions , a high - fidelity population mapping from the internal to the motional degrees of freedom is possible .
losses were very low and our simulations indicate that the fidelity can be further improved by detuning the laser pulses further from the transition . in the scheme we propose here , each rotational level
is coupled to the excited state by a laser .
due to the large rotational level splitting of light molecules this means in turn that multiple lasers are required .
the number of levels @xmath199 which have a population larger than the cut - off population @xmath200 , and therefore the number of required lasers , can be estimated as @xmath201 for small @xmath200 . even though the number of levels @xmath199 for molecules at room temperature can be of the order of 25 , for temperatures of a few kelvin this reduces to well below 10 states . in many experiments
molecular ions can be prepared in low lying rotational states by employing photo - association or state selective photo - ionisation in conjunction with supersonic beam expansion or buffer - gas cooling .
however , due to the interaction with black - body radiation and collisions the internal temperature quickly thermalises . by applying the scheme proposed here , this thermalisation can be suppressed to maintain the ground - state population .
additionally , the state mapping can be employed to detect the internal states of the molecule in a non - destructive manner which is beneficial for high - resolution spectroscopy of molecules . in conclusion , we have developed a fast scheme for cooling the internal states of single molecules by employing adiabatic passage methods which provide a high efficiency in conjunction with robustness against variations in the parameters of the involved laser pulses .
see n. v. vitanov , m. fleischhauer , b. w. shore , and k. bergmann , in _ advances in atomic , molecular , and optical physics _ , edited by h. walther and b. bederson ( academic press , 2001 ) , vol .
46 , pp . 55190 , and references therein .
m. d. barrett , b. demarco , t. schaetz , v. meyer , d. leibfried , j. britton , j. chiaverini , w. m. itano , b. jelenkovi , j. d. jost , c. langer , t. rosenband , and d. j. wineland , phys . rev . a * 68 * , 042302 ( 2003 ) . in the conext of ion traps , see , e.g. , d. j. wineland , j. j. bollinger , w. m. itano , and d. j. heinzen , phys .
a * 50 * , 67 ( 1994 ) , see section viii ; j. steinbach , j. twamley , and p. l. knight , phys .
a * 56 * , 4815 ( 1997 ) ; d. j. wineland , c. monroe , w. m. itano , d. leibfried , w. e. king , and d. m. meekhof , j. res .
inst . stand .
technol . * 103 * , 259 ( 1998 ) . as
explained in ref .
@xcite , the far off - resonant stirap process can be seen as a limit of scrap / siarp in the absence of decay . in the presence of decay
the stirap process is affected more due to the involvement of the excited state in the dynamics taking place over a long time . due to the generally different rotational constants ,
any population in vibrationally excited states is basically unaltered by the mapping process we have presented here .
thus only the vibrational ground - state of the molecule is transferred into its rotational ground - state . | in this work we investigate the theory for three different uni - directional population transfer schemes in trapped multilevel systems which can be utilized to cool molecular ions .
the approach we use exploits the laser - induced coupling between the internal and motional degrees of freedom so that the internal state of a molecule can be mapped onto the motion of that molecule in an external trapping potential . by sympathetically cooling the translational motion back into its ground state
the mapping process can be employed as part of a cooling scheme for molecular rotational levels .
this step is achieved through a common mode involving a laser - cooled atom trapped alongside the molecule . for the coherent mapping
we will focus on adiabatic passage techniques which may be expected to provide robust and efficient population transfers . by applying far - detuned chirped adiabatic rapid passage pulses we are able to achieve an efficiency of better than 98% for realistic parameters and including spontaneous emission . even though our main focus is on cooling molecular states ,
the analysis of the different adiabatic methods has general features which can be applied to atomic systems . |
methods for preparation and manipulation of nonclassical states of light have become an important research area in quantum optics @xcite , especially in relation to possible optical implementations of quantum computers and systems for quantum communication and quantum cryptography @xcite . among various schemes for optical - qubit generation , the so - called _ quantum scissors _ device of pegg _ et al . _
@xcite produces a superposition of vacuum and single - photon states , @xmath0 , by optical - state truncation of an input single - mode coherent light .
quantum scissors device was studied in numerous papers ( see , e.g. , @xcite ) , and tested experimentally by babichev _
_ @xcite and resch _ et al .
this simple scheme and its generalizations for truncation of an input optical state to a superposition of @xmath1 fock states ( the so - called qudits ) @xcite are based on linear optical elements , and thus referred to as the _ linear quantum scissors _ devices .
optical - state ` truncation ' can also be achieved in systems comprising nonlinear elements ( e.g. , kerr media ) @xcite , and thus will be referred to as the _ nonlinear quantum scissors _ devices .
the above - mentioned schemes are restricted to the single - mode optical truncation . here , by generalizing our former scheme @xcite , we present a realization of nonlinear quantum scissors for optical - state truncation of two cavity modes by means of a pumped nonlinear coupler .
two - mode nonlinear couplers have become , shortly after their introduction by jensen @xcite and maier @xcite , one of the important topics of photonics due to their wide potential applications and relative simplicity ( see , e.g. , reviews @xcite ) . among various types of the nonlinear optical couplers ,
those based on kerr effect have attracted especial interest both in classical @xcite and quantum @xcite regimes .
the kerr nonlinear couplers can exhibit variations of self - trapping , self - modulation and self - switching effects .
in quantum regime , they can also generate sub - poissonian and squeezed light @xcite .
possibilities of entanglement generation were also studied in nonlinear couplers operating by means of the kerr effect @xcite or degenerate parametric down - conversion @xcite . here ,
we analyze kerr nonlinear couplers , which can be modelled by systems composed of two quantum nonlinear oscillators linearly coupled to each other and placed inside a double - ring cavity .
we discuss two schemes based on the coupler with an external excitation of a single mode and the coupler with two modes pumped .
we show that the states generated in the excited nonlinear couplers under suitable conditions can be limited to a superposition of only vacuum and single - photon states , @xmath2 .
we compare the possibilities of generation of maximally entangled states by the couplers excited in single and two modes .
we also discuss effects of dissipation on the fidelity of truncation and suggest a method to achieve strong kerr interactions at low intensities in our system .
we consider a system , referred to as the pumped kerr nonlinear coupler , which consists of two nonlinear oscillators , with kerr nonlinearities @xmath3 and @xmath4 , linearly coupled to each other and additionally linearly coupled to an external classical field . in this section , we assume that the field is coupled to one of the oscillators only . thus , the system can be described by the following hamiltonian @xcite ( @xmath5 ) : @xmath6 where @xmath7 @xmath8 and @xmath9 ( @xmath10 ) is the bosonic annihilation operator corresponding to the mode @xmath11 ( @xmath12 ) of frequency @xmath13 ( @xmath14 ) .
hamiltonian ( [ n01 ] ) for @xmath15 describes the standard nonlinear coupler @xcite composed of two kerr nonlinear oscillators linearly coupled to each other , where the parameter @xmath16 is the strength of this coupling .
however , our model differs from the standard by inclusion of the extra term @xmath17 , which describes linear coupling between the driving single - mode classical field ( say , with frequency @xmath18 ) and the cavity mode @xmath11 .
the parameter @xmath19 describes the strength of this coupling and is proportional to the classical field amplitude .
a possible physical realization of the model is presented in figure 1 , where an external pump , described by @xmath20 , is off in the present analysis .
kerr media , marked by @xmath21 and @xmath22 , are linearly coupled as described by grey central region corresponding to @xmath23 .
the evolution of our system in the interaction picture can be described by the schrdinger equation @xmath24 where @xmath25 thus , the complex probability amplitudes @xmath26 satisfy the set of equations of motion : @xmath27c_{mn } \notag \\ & & + \epsilon c_{m-1,n+1 } \sqrt{m(n+1 ) } + \epsilon ^{\ast } c_{m+1,n-1 } \sqrt{(m+1)n } \notag \\ & & + \alpha c_{m-1,n } \sqrt{m } + \alpha ^{\ast } c_{m+1,n } \sqrt{m+1}. \label{n07}\end{aligned}\ ] ] a superficial analysis of ( [ n07 ] ) could lead to a conclusion that the evolution of the system pumped by classical external field can not be restricted to two lowest photon - numer states , but will also include states with a greater number of photons . however , by generalizing the method of single - mode nonlinear quantum scissors proposed in @xcite ( for a review see @xcite ) , we have observed in @xcite that evolution can be restricted to the only four states : @xmath28 , @xmath29 , @xmath30 and @xmath31 as a result of degeneracy of hamiltonians @xmath21 and @xmath22 . by assuming that the couplings @xmath32 , and @xmath33 are much smaller than the kerr nonlinearities @xmath3 and @xmath4 , we can interpret the evolution between the four states as resonant transitions , while the negligible evolution to other states as out of resonance , analogously to the single - mode case @xcite .
this phenomenon can be shown explicitly as follows : under the assumption of @xmath34 and short evolution times , equation ( [ n07 ] ) for @xmath35 can be approximated by @xmath36 c_{mn } \label{n08}\ ] ] which has the simple solution @xmath37 t\}c_{mn}(0 ) .
\label{n09}\ ] ] by setting the initial condition @xmath38 for @xmath35 , one gets @xmath39 .
by contrast , for @xmath40 , the terms proportional to @xmath3 and @xmath4 are vanishing due to degeneracy of the kerr hamiltonian and so the remaining terms proportional to @xmath16 and @xmath19 are significant .
thus , the ideally ` truncated ' two - mode state generated in the system has the following simple form @xmath41 where the evolution of @xmath42 , precisely given by ( [ n07 ] ) , can approximately be described by the following equations @xmath43 hereafter , in equations for the probability amplitudes under the discussed assumptions , sign ` = ' should be understood as ` @xmath44 ' .
although approximate equations ( [ n11 ] ) are independent of @xmath3 , our derivation clearly shows that the kerr nonlinearity plays a crucial role in the truncation process . by assuming that both oscillators are initially in vacuum states , @xmath45 , and parameters @xmath19 and @xmath16 are real , we find the following solutions of ( [ n11 ] ) for the time - dependent probability amplitudes : @xmath46 , \notag \\ c_{01}&=&\frac{\alpha}{\gamma}\left(\cos\tau _
{ 1}-\cos\tau _ { 2}\right ) , \notag \\
c_{10}&=&-\frac{i(\gamma+\epsilon)\omega_2}{4 \alpha \gamma}(\sin\tau _ { 1}+\sin\tau _ { 2 } ) , \notag \\ c_{11}&=&-\frac{i}{2\gamma}\left(\omega_2\sin\tau _ { 1}-\omega_1\sin\tau _ { 2}\right ) , \label{n12}\end{aligned}\ ] ] where @xmath47}$ ] , @xmath48 , and @xmath49 for @xmath50 .
note that the solution for @xmath51 can be written in a more symmetric form since the properties hold : @xmath52 . in a special case of equal couplings @xmath19 and @xmath16 , ( [ n12 ] ) simplifies to our former solution @xcite : @xmath53 where @xmath54 . to estimate the quality of the optical - state truncation ( up to single - photon states ) of the generated light , we apply fidelity as a measure of discrepancy between the ideally truncated two - qubit state @xmath55 , given by ( [ n10 ] ) , and the actually generated output state @xmath56 calculated numerically from @xmath57 for a large ( practically infinite - dimensional ) two - mode hilbert space
specifically , in our numerical analysis , we have chosen the dimension equal to @xmath58 for each subspace associated with single mode of the field .
the fidelity , also referred to as the uhlmann s transition probability for mixed states , is defined by ( see , e.g. , @xcite ) : = 4.4 cm = 4.4 cm = 4.4 cm = 4.4 cm = 4.4 cm = 4.4 cm @xmath59\right\}^2 .
\label{n15}\end{aligned}\ ] ] by assuming that one of the states is pure ( say @xmath60 ) , then ( [ n15 ] ) simplifies to @xmath61 .
instead fidelity , the bures distance @xmath62 is often applied as a measure of discrepancy between the states .
note that the bures distance satisfies the usual metric properties including symmetry @xmath63 , contrary to the quantum kullback - leibler ` distance ' ( quantum relative entropy ) often used in quantum - information context @xcite .
general expression ( [ n15 ] ) will be applied in our description of the effect of damping on the optical - state truncation in section 4 . in this section ,
we are focused on pure - state truncation , for which equation ( [ n15 ] ) simplifies to the familiar expression @xmath64 , where the ideally truncated state @xmath65 is given by ( [ n10 ] ) and the actually generated state @xmath66 is calculated numerically from ( [ n14 ] ) .
the fidelity for perfect truncation is equal to one .
figure 2 clearly indicates that the fidelity of truncation using the pumped coupler is close to one for short times and the coupling strengths much smaller than the nonlinearity parameters ( @xmath67 ) .
the numerical results shown in figure 2 confirm the validity of our analytical approach at least for short evolution times .
thus , we can refer to the system as a kind of nonlinear ( as operating by means of kerr nonlinearity ) quantum scissors device .
solutions for probability amplitudes of the truncated states enable a simple calculation of quantum entanglement , which is one of the most fundamental resources of quantum information theory @xcite .
it is well known that the entanglement of a bipartite pure state , described by a density matrix @xmath68 or @xmath69 or , equivalently , by the shannon entropy of the squared schmidt coefficients @xmath70 @xcite : @xmath71 for bipartite pure states this measure is often referred to as the _ entropy of entanglement_. in a special case of two qubits in a pure state , the entropy of entanglement @xmath72 ranges from zero for a separable state to 1 ebit for a maximally entangled state , and it is simply given in terms of the binary entropy @xmath73 .
in fact , for a general two - qubit pure state , given by ( [ n10 ] ) with arbitrary amplitudes @xmath42 ( @xmath74 ) , the entropy of entanglement given by ( [ n16 ] ) can simply be calculated as @xmath75 where @xmath76 and @xmath77 is the binary entropy . if the probability amplitudes @xmath26 evolve according to ( [ n12 ] ) then the evolution of the entropy of entanglement is given by @xmath78 \nonumber \\
\times \cos({{\textstyle{\frac{1}{2 } } } } \omega_{j}t)\big \ } \sin({{\textstyle{\frac{1}{2 } } } } \omega_{3-j}t)\big ) .
\label{n19}\end{aligned}\ ] ] solution ( [ n19 ] ) is further simplified by assuming real @xmath79 then the amplitudes @xmath42 are given by ( [ n13 ] ) .
thus , we obtain @xmath80 \sin ( \alpha t ) \nonumber \\ -\sqrt{5}\cos ( \alpha t)\sin ( \sqrt{5}\alpha t)\}\big ) .
\label{n20}\end{aligned}\ ] ] figures 3(a ) and 4(a ) show the plots of the entropy of entanglement for the case discussed here , i.e. , for the single - mode pumped coupler with the coupling parameters @xmath32 and @xmath33 much smaller than the nonlinearities @xmath81 .
we see in figure 3(a ) that the rapid oscillations in time ( with a period @xmath82 ) are modulated by oscillations of low frequency ( with period @xmath83 ) . as a consequence ,
their maxima are of various values but some of them approach 1 ebit corresponding to the formation of bell states . to show this explicitly , we represent the generated state in the basis @xmath84 spanned by the bell - like states : = 4.4 cm = 4.4 cm = 4.4 cm = 4.4 cm = 4.4 cm = 4.4 cm @xmath85 which differ from the standard bell states only by the phase factor @xmath86 .
clearly , @xmath87 and @xmath88 for @xmath89 , so states ( [ n21 ] ) will shortly be referred to as the bell states .
figures 5 and 6 show the probabilities for the generation of the bell states as a function of time for the single - mode driven couplers , when the initial state is the two - mode vacuum state .
it is seen that states @xmath90 and @xmath91 , being superpositions of @xmath28 and @xmath31 , are generated with a high accuracy . in detail , the maxima of ( [ n20 ] ) occur approximately at times @xmath92 $ ] if @xmath79 .
since the frequencies of oscillations in ( [ n20 ] ) are incommensurate , by waiting long enough ( assuming no dissipation ) , we can achieve 1 ebit with high precision .
for example , in the first four periods of @xmath93 , one could observe generation of entangled states approaching bell state @xmath90 with @xmath94=0.994 $ ] and @xmath95=0.997 $ ] ebits , and approaching @xmath91 with @xmath96=0.992 $ ] and @xmath97=0.998 $ ] .
note that only the first period is shown in figures 3(a ) and 5(a ) , for which the highest entanglement of 0.994 ebits occurs approximately at the time @xmath98 .
thus , we can effectively treat our system as a source of maximally entangled states . on the other hand ,
the bell states @xmath99 and @xmath100 , being superpositions of @xmath29 and @xmath30 , are not generated if the initial state is vacuum and @xmath101 .
this conclusion can be drawn by observing that the probabilities for @xmath99 and @xmath100 can reach only @xmath102 in figure 5(b ) . however , by relaxing the condition of equal couplings @xmath16 and @xmath19 , as shown in figure 6(b ) , bell states @xmath99 and @xmath100 can be generated from vacuum with high precision in the dissipation - free system .
hitherto , we have assumed that both cavity modes were initially in vacuum states .
now , we analyze a more general evolution when the cavity modes are initially not only in vacuum but also in single - photon fock states , i.e. , @xmath103 , where @xmath104 .
thus , by assuming as usual that @xmath105 , the evolutions of the initial states are found to be @xmath106 in terms of the time - dependent amplitudes @xmath107 given by ( [ n13 ] ) and @xmath108 where , as usual , @xmath109 . please note that @xmath110 and @xmath111 differ in sign .
we find that the generalized expression for the entropies of entanglement for the initial states @xmath112 with @xmath104 reads as @xmath113 \sin ( \alpha t)\hspace{1.5 cm } \nonumber \\
-(-1)^{k - l}\sqrt{5}\cos ( \alpha t)\sin ( \sqrt{5}\alpha t)\big ) , \label{n24}\end{aligned}\ ] ] which implies that @xmath114 it is worth noting that our system for the initial states @xmath30 or @xmath29 quasi - periodically evolves into the bell states @xmath99 and
@xmath100 with high precision but does not evolve into @xmath90 or @xmath91 assuming @xmath79 .
this is by contrast to the evolutions of the initial states @xmath28 ( see figure 5 ) or @xmath31 also for @xmath101 .
for brevity , we will not present any graphs of the evolutions of the initial states @xmath30 , @xmath29 , and @xmath31 , which would correspond to figures 25 plotted for the initial vacua .
the above solutions for the initial single - photon states are included for completeness of our mathematical approach to show that , in principle , all bell states can be generated in our system even for @xmath101 .
but it should be stressed that the system with the initial fock states is much more experimentally challenging than that assuming initially the vacuum states only . despite of experimental difficulty ,
our system enables generation of the one - photon fock states from vacuum assuming no coupling @xmath16 between the modes , which corresponds to having two independent pumped cavities with nonlinear kerr media .
a possibility of producing single - photon states in such systems was demonstrated in @xcite ( for a review see @xcite and references therein ) .
this section is devoted to the most general scheme presented in figure 1 , namely that involving two external excitations .
we assume here that both modes of the coupler are excited by external fields , whereas for the case discussed previously we assumed that only one of the modes was coupled to the external field .
the hamiltonian describing such a system is of the form @xmath115 which is the same as that defined by ( [ n01])([n04 ] ) , except for the extra term given by @xmath116 corresponding to the coupling of the cavity mode @xmath12 with an external driving single - mode classical field ( say , with frequency @xmath117 ) , where the parameter @xmath118 describes the strength of this interaction being proportional to the classical field amplitude .
the evolution of our system , described by hamiltonian ( [ n26 ] ) , can be given by the schrdinger equation from which we find the following set of equations for the amplitudes @xmath26 of the wavefunction ( [ n06 ] ) in the interaction picture : @xmath119c_{mn } \notag \\ & & + \epsilon c_{m-1,n+1 } \sqrt{m(n+1 ) } + \epsilon ^{\ast } c_{m+1,n-1 } \sqrt{(m+1)n } \notag \\ & & + \alpha c_{m-1,n } \sqrt{m } + \alpha ^{\ast } c_{m+1,n } \sqrt{m+1}\notag \\ & & + \beta c_{m , n-1 } \sqrt{n } + \beta ^{\ast } c_{m , n+1 } \sqrt{n+1}. \label{n28}\end{aligned}\ ] ] analogously to the analysis in the former section , we assume short evolution times as well as the couplings @xmath32 , @xmath120 , and @xmath33 to be much smaller than the kerr nonlinearities @xmath3 and @xmath4 .
then , equation ( [ n28 ] ) for @xmath121 can be approximated by ( [ n08 ] ) with the solution ( [ n09 ] ) , which vanishes for the initial condition @xmath38 .
thus , under the above assumptions , the infinite set of equations ( [ n28 ] ) reduces to the following four equations : @xmath122 assuming that at the time @xmath123 , both oscillator modes are in the vacuum states , i.e. , @xmath124 and @xmath125 , we can obtain analytical solutions of ( [ n29 ] ) . to solve ( [ n29 ] ) we need to find zeros of a fourth - order polynomial and , hence , the solutions in their general form are rather complicated and unreadable . however ,
if we assume that all coupling constants are real and the couplings with two external fields are of the same strength ( @xmath126 ) then the solutions become much simpler and easier to interpret .
thus , under these assumptions , the solutions are found to be : @xmath127e^{-i\epsilon t/2}\right\ } , \notag \\
c_{01 } & = & c_{10}=-i2\frac{\alpha } { \lambda } \sin ( \frac{\lambda t}{2})\,e^{-i\epsilon t/2 } , \notag \\ c_{11 } & = & c_{00}-1 , \label{n30}\end{aligned}\ ] ] where we have introduced the effective coupling constant @xmath128 . as in the case discussed in the previous section ,
the system s dynamics is closed within the finite set of @xmath129-photon states .
figure 2 shows that for the assumed parameters and evolution times shorter than @xmath130s , the fidelity between the ideally truncated states and the actually generated states by means of the coupler pumped in two modes deviates from 1 by the values less than @xmath131 in figure 2(a ) or even @xmath132 in figure 2(b ) .
this again confirms the validity of our analysis and justifies referring to this system as a kind of the quantum scissors device . for the parameters assumed in figure 2 , truncation with higher fidelity
is usually observed for the coupler pumped in a single mode rather than in two modes . in the latter case , the truncation fidelity depends on the relative phase between the pumping - field couplings @xmath19 and @xmath118 .
note that state @xmath133 for @xmath134 , contrary to @xmath135 , can be directly calculated from solution ( [ n12 ] ) .
for brevity , we have not presented here an analogous solution for @xmath135 , but we have used it for plotting the corresponding curves in figures 2 and 3(d ) .
it is seen , by comparing figures 2(a ) and 2(b ) , that by decreasing @xmath16 in comparison to @xmath19 , the fidelity of truncation can be improved for @xmath136 .
thus , pumping the coupler in two modes can lead sometimes to truncation better than that for the system driven in single mode as presented in figure 2(b ) by dot - dashed curve corresponding to @xmath135 and @xmath137 .
the evolution of the pure - state entanglement , given by ( [ n16 ] ) , generated in the coupler pumped in two modes can be calculated from ( [ n17 ] ) with the probability amplitudes given by ( [ n30 ] ) .
thus , we obtain @xmath138\big|\big).\label{n31}\ ] ] figures 3(b - d ) and 4(b ) show the evolution of the entropy of entanglement measured in ebits as a function of time for the system pumped in two modes in comparison to the results for the single - mode driven coupler shown in figures 3(a ) and 4(a ) .
it is seen that the first maximum in figures 3(b - d ) is the highest , contrary to the case shown in figures 3(a ) for the single - mode pumping .
nevertheless , the most important fact is that the value of the entropy of entanglement @xmath72 can approach unity to a high precision .
so , as in the case of the single - mode pumping , the two - mode driven system effectively generates bell states .
to find which bell states are generated , we can also transform the resulting wavefunction into the bell basis .
thus , figure 7 depicts probabilities for the four bell states .
as expected from the form of our analytical solutions for @xmath42 ( @xmath74 ) , the entanglement occurs for the states @xmath28 and @xmath31 , and leads to the generation of the states @xmath90 and @xmath91 ( with some unimportant global phase factor ) . clearly
, the highest peaks of the entropy @xmath139 and those of the probabilities @xmath140 of the bell state generation occur at the same evolution times , as seen by comparing figures : 3(a ) with 5(a ) , 4(a ) with 6(a ) , and 4(b ) with 7 .
analysis of the evolutions of the probabilities @xmath140 and the entropy of entanglement @xmath139 for relatively longer times reveals that the oscillations are modulated and some long - time oscillations occur in the system , which can be interpreted as a result of quantum beats .
it is seen from ( [ n31 ] ) that two various frequencies appear in our solution and one of them is considerably greater than the other , e.g. , the effective coupling constant @xmath141 is @xmath142 times greater than the internal coupling constant of the coupler @xmath101 .
we are not presenting here dissipation - free evolutions exhibiting modulated oscillations at times longer than those in figure 4
. it would be meaningless since the entanglement is lost at such evolution times due to dissipation , which inevitably occurs in real physical implementations of the coupler as will be discussed in the next section .
although our analysis , including all figures , is focused on evolution of the initial vacuum states , we present shortly some results for other states too .
we find that the evolution of the initial fock state @xmath143 with @xmath104 is of the form ( [ n22 ] ) but with the probability amplitudes @xmath107 given by ( [ n30 ] ) and @xmath144 equal to @xmath145 \big\ } , \notag \\
\widetilde{c}_{11}(\tau ) & = & \widetilde{c}_{00}(\tau ) -e^{i\epsilon t}. \label{n32}\end{aligned}\ ] ] with the help of these formulae , we can calculate the entropies of entanglement explicitly as @xmath146 \hspace{0.5 cm } \nonumber \\
+ \big[16 \alpha^2 + \epsilon^2 \cos(\lambda t)(-1)^{k - l}i \epsilon \lambda \sin(\lambda t)\big]\big|\big ) \label{n33}\end{aligned}\ ] ] implying the same properties as those given by ( [ n25 ] ) for the single - mode excited system .
we point out that both single- and two - mode pumped couplers for @xmath79 and the initial states @xmath30 or @xmath29 evolve into the bell states @xmath99 and @xmath100 but do not evolve into @xmath90 or @xmath91 .
this is contrary to the evolutions of the initial vacuum states ( or @xmath31 ) as , for example , presented in figures 5 and 7 .
in a more realistic description both fields @xmath11 and @xmath12 lose their photons from the cavities . according to the standard techniques in theoretical quantum optics , dissipation of our system
can be modelled by its coupling to reservoirs ( heat baths ) as described by the interaction hamiltonian @xmath147 where @xmath148 are the reservoir operators ; @xmath149 are the boson annihilation operators of the reservoir oscillators coupled with mode @xmath11 or @xmath12 , respectively ; @xmath150 are the coupling constants of the interaction with the reservoirs ; @xmath151 is given either by ( [ n01 ] ) or ( [ n26 ] ) dependent on the system analyzed .
we assume two kinds of functions @xmath152 and @xmath153 to describe standard damping and dephasing .
the standard description of a damped system is obtained for ( [ n35 ] ) with @xmath154 and @xmath155 , or explicitly @xmath156 corresponding to energy transfer between the system and reservoirs .
it should be stressed that the process described by ( [ n37 ] ) leads to combined effect of amplitude and phase damping .
the evolution in the markov approximation of the reduced density operator @xmath157 of the two cavity modes after tracing out over the reservoirs can be described in the interaction picture by the following master equation ( see , e.g. , @xcite ) @xmath158 + \hat { \cal l}_{\rm loss}\hat \rho , \label{n38}\end{aligned}\ ] ] where the liouvillian @xmath159+[\hat a,\hat \rho\hat a^\dagger ] ) + \frac{\gamma_b}{2 } ( [ \hat b\hat \rho,\hat b^\dagger ] \nonumber \\ \quad + [ \hat b,\hat \rho\hat b^\dagger ] ) + \gamma_a \bar
n_a [ [ \hat a,\hat \rho],\hat a^\dagger ] + \gamma_b \bar n_b [ [ \hat b,\hat \rho],\hat b^\dagger ] \label{n39}\end{aligned}\ ] ] is the usual loss term corresponding to @xmath160 , given by ( [ n37 ] ) ; @xmath161 is the damping rate of the @xmath162th ( @xmath163 ) ( ring ) cavity , and @xmath164^{-1}$ ] is the mean number of thermal photons at the reservoir temperature @xmath165 .
we will analyze both noisyreservoirs ( at @xmath166 implying @xmath167 ) and ` quiet ' reservoirs ( at @xmath168 , so @xmath169 ) .
the latter assumption implies that diffusion of fluctuations from the reservoirs into the system modes is negligible .
but still this simplified master equation describes the loss of photons from the system modes to the reservoirs .
one can raise some doubts @xcite against using liouvillian ( [ n39 ] ) in a description of lossy anharmonic oscillator models given by ( [ n02 ] ) .
also the approximation of @xmath170 in ( [ n39 ] ) is problematic .
nevertheless , master equation ( [ n38 ] ) with liouvillian ( [ n39 ] ) and hamiltonian @xmath171 set to @xmath21 was used in a number of works both for @xmath166 ( see , e.g. , @xcite ) but also for @xmath170 ( see , e.g. , @xcite ) . moreover , the same master equation , given by ( [ n38 ] ) for @xmath151 set to ( [ n02 ] ) , for coupled anharmonic oscillators was applied in , e.g. , @xcite . liouvillian ( [ n39 ] ) for @xmath170 was also used in ( @xcite p. 210 ) to describe a model essentially similar to ours comprising a nonlinear system , in which two quantized field modes in a cavity interact with a classical pump field .
the standard liouvillian for @xmath170 was also used in @xcite to describe a system of kerr nonlinearity , given by @xmath21 , and a parametric amplifier driven by a pulsed classical field .
nevertheless , it should be noted that a realistic master equation for kerr medium @xcite is more complicated .
phase damping ( also referred to as dephasing ) can be described by ( [ n35 ] ) assuming @xmath172 and @xmath173 , which gives the following loss hamiltonian @xcite : @xmath174 this interaction can be interpreted as a scattering process , where the number of photons remains unchanged contrary to the interaction described by ( [ n37 ] ) .
phase damping is essential in a fully quantum picture of dissipation of our system . as a simple generalization of the gardiner - zoller master equation for a single harmonic oscillator ( @xcite , equation ( 6.1.15 ) )
, we describe phase damping of our two - mode nonlinear system by master equation ( [ n38 ] ) for the liouvillian @xmath175 \notag \\
\quad + \frac{\gamma_b}{2 } ( 2 \bar n_b+1 ) [ 2\hat b^\dag \hat b \hat \rho \hat b^\dag \hat b - ( \hat b^\dag \hat b)^2
\hat \rho-\hat \rho ( \hat b^\dag \hat b)^2 ] , \label{n41}\end{aligned}\ ] ] which is significantly different from ( [ n39 ] ) .
to analyze dissipative evolution , governed by master equations ( [ n38 ] ) for liouvillians ( [ n39 ] ) and ( [ n41 ] ) , we apply standard numerical procedures for solving ordinary differential equations with constant coefficients as an exponential series .
the entropy of entanglement @xmath72 , given by ( [ n16 ] ) , is valid for qudits of arbitrary dimension in a pure state , but it fails to determine entanglement of a system in a mixed state .
thus , for a two - qubit mixed state @xmath176 , we have to apply more general measure , e.g. , the wootters measure of entanglement of formation given by @xcite : @xmath177 where @xmath178 is given by ( [ n18 ] ) with the argument being the concurrence @xmath179 defined as @xmath180 and @xmath181 are the square roots of the eigenvalues of @xmath182 , while @xmath183 is the pauli spin matrix of the @xmath162th qubit ( @xmath163 ) .
it is well - known that the entanglement of formation goes into the entropy of entanglement for any two - qubit pure states .
examples of evolution of the entanglement of formation and fidelity for the dissipative systems are shown in figures 811 both for quiet and noisy reservoirs . by analyzing the figures , the most important observation is that the generated two - qubit entangled states are very fragile to the leakage of photons from the cavities in the analyzed couplers .
our system is more fragile to losses described by the standard master equation , given by ( [ n38 ] ) and ( [ n39 ] ) , rather than the losses due to only phase damping as described by the gardiner - zoller master equation , given by ( [ n38 ] ) and ( [ n41 ] ) . inclusion of reservoir noise , at least for the assumed low mean numbers @xmath184 of thermal photons , does not cause a dramatic deterioration of fidelity and entanglement in comparison to the losses caused by coupling the system to the zero - temperature reservoirs .
note that we have chosen @xmath185 in the dephasing model shown in figure 11 and much smaller value of @xmath186 in the standard dissipation model in figure 9 .
the reason is that for the latter dissipation model , the number of photons in the system can be increased by absorbing thermal photons from the reservoirs .
if this absorption of thermal photons exceeded the loss of the system photons to the reservoirs then the generated fields could not be approximated as two - qubit states and the wootters function @xmath187 , given by ( [ n42 ] ) , would fail to be a good entanglement measure . for the parameters chosen in figure 9 ,
the generated states are well approximated by two - qubit density matrices and thus its entanglement of formation is well described by ( [ n42 ] ) . by contrast , the number of system photons is not affected by the reservoirs in the dephasing model ; thus ( [ n42 ] ) can be used for any number @xmath186 of thermal photons .
finally , it should be noted that fragility of our system to dissipation seems to be a serious drawback from an experimental point of view .
nevertheless , a method which enables a significant improvement of the entanglement robustness of the generated states , has recently been suggested for a similar system @xcite .
one of the crucial conditions for the successful truncation and generation of the bell states in our scheme concerns the couplings @xmath32 , @xmath120 and @xmath33 to be much smaller than the kerr nonlinearities @xmath3 and @xmath4 .
this implies that the kerr interaction should be strong at very low light intensities .
thus , it is desirable to discuss an implementation , in which such stringent conditions can experimentally be satisfied .
a possible realization can be based on the effect of the electromagnetically - induced transparency ( eit ) or atomic dark resonances as proposed by schmidt and imamolu @xcite ( see also @xcite ) and observed experimentally @xcite .
the schmidt - imamolu eit scheme can be realized in a low density system of four - level atoms , which level structure is shown in figure 12 , exhibiting giant resonantly - enhanced kerr nonlinearity at very low intensities .
the atoms are placed in cavity @xmath11 ( and analogously in cavity @xmath12 ) tuned to frequency @xmath13 of the mode @xmath11 resonant with the transition @xmath188 and detuned by @xmath189 of the transition @xmath190 .
the eit effect is created by a classical pumping field of frequency @xmath18 resonant with the transition @xmath191 ( see @xcite ) , all the atomic levels can adiabatically be eliminated , which results in the following formula for the kerr nonlinearity @xcite : @xmath192 where @xmath193 is third - order nonlinear susceptibility , @xmath194 are the coupling coefficients , @xmath195 is the rabi frequency of the classical driving ( and coupling ) field , @xmath196 is the electric dipole matrix element between the states @xmath197 and @xmath198 , @xmath199 is the total number of atoms contained in the cavity of volume @xmath200 , and @xmath201 is the permittivity of free space . by replacing subscript @xmath11 by @xmath12 in ( [ n44 ] ) ,
an analogous expression for @xmath4 can be obtained . by putting the stringent limit on the required cavity parameters @xcite , imamolu _ et al . _
estimated @xmath202 rad / sec @xcite .
note that some quantum information applications of these giant kerr nonlinearities have already been studied @xcite .
more details about application of the schmidt - imamolu scheme of resonantly - enhanced kerr nonlinearity in the analysis of dissipation effects on entanglement generation are given by one of us in @xcite .
it is worth stressing the differences between the present paper and our former works .
( i ) in @xcite , only the single - mode pumped systems were studied . here
, we analyze also two - mode pumped systems , which is basically a different model .
( ii ) the model described in @xcite is crucially different from the one used here as was based on the two - mode _ nonlinear _ interaction term @xmath203 . in the present work , as well as in @xcite , we apply the two - mode _ linear _ interaction hamiltonian @xmath204 , which is the same ( by neglecting the pump terms @xmath205 and @xmath206 ) as that used in @xcite .
the other novelties of the present work in comparison to @xcite can be summarized as follows : ( 1 ) for single - mode pumped system , we found a new generalized solution ( [ n12 ] ) , which , in a special case of equal couplings @xmath19 and @xmath16 , simplifies to our former solution obtained in @xcite .
( 2 ) analytical approximate solutions were found here for various initial fock states @xmath207 . in @xcite , solutions were given for initial vacuum states only .
( 3 ) here , we describe a possible realization of the model based on the effect of the electromagnetically - induced transparency .
( 4 ) in present numerical analysis based on the eit scheme , in comparison to @xcite , more realistic parameters were chosen for the coupling constants , kerr nonlinearities , and damping constants .
( 5 ) the deterioration of the fidelity of the generated bell states due to the standard dissipation and phase damping was analyzed here within two types of master equations both for quiet and noisy reservoirs . by contrast
, analysis of losses in @xcite was limited to the standard master equation and for the quiet reservoir only .
no effect of dissipation was studied in @xcite .
( 6 ) in the present manuscript , entanglement of formation was calculated analytically for dissipation - free systems and calculated numerically for dissipative systems .
no analytical formulae were given in @xcite , while the entanglement of formation was not at all studied in @xcite .
( 7 ) the quality of truncation was described here , but it was not studied in @xcite .
discrepancy between the really generated states and the exact truncated states was measured here by the fidelity . in conclusion ,
we have described a realization of the generalized two - mode optical - state truncation of two coherent modes via a nonlinear process .
our system is a two - mode generalization of the single - mode nonlinear quantum scissors device described in @xcite .
we have described an implementation of the kerr nonlinear couplers , where resonantly enhanced nonlinearities can be achieved in the schmidt - imamolu eit scheme .
we have compared kerr nonlinear couplers linearly excited in one or two modes by external classical fields .
we have shown under assumption of the coupling strengths to be much smaller than the nonlinearity parameters @xmath3 and @xmath4 that the optical states generated by the couplers are the two - qubit truncated states spanned by vacuum and single - photon states .
although our approximate solutions of the schrdinger equation are independent of @xmath208-constants , the kerr nonlinearity plays a crucial role in the physics as we have derived from the complete @xmath208-dependent hamiltonian .
in fact , the kerr interaction is the mechanism in our model , which enables truncation of the generated state at some energy level .
by contrast , the system without the kerr nonlinearities and pumped by an external field would gain more and more energy . to confirm our predictions ,
we have compared ` exact ' ( accurate up to double - precision ) direct numerical solutions of the schrdinger equation and compared with our approximate analytical solutions .
the discrepancies between the exact and approximate solutions are relatively small as shown by fidelities in figures 24 .
we have demonstrated that our system initially in vacuum state or single - photon fock states @xmath209 ( @xmath74 ) evolves into bell states .
we have discussed the fragility of the entanglement of formation of the generated states due to the standard dissipation and dephasing in the two distinct master equation approaches .
we thank professor ryszard tana , professor ryszard horodecki and professor robert alicki for discussions .
this work was supported by the polish state committee for scientific research under grant no
. 1 p03b 064 28 . | schemes for optical - state truncation of two cavity modes are analyzed .
the systems , referred to as the nonlinear quantum scissors devices , comprise two coupled nonlinear oscillators ( kerr nonlinear coupler ) with one or two of them pumped by external classical fields .
it is shown that the quantum evolution of the pumped couplers can be closed in a two - qubit hilbert space spanned by vacuum and single - photon states only .
thus , the pumped couplers can behave as a two - qubit system .
analysis of time evolution of the quantum entanglement shows that bell states can be generated .
a possible implementation of the couplers is suggested in a pumped double - ring cavity with resonantly enhanced kerr nonlinearities in an electromagnetically - induced transparency scheme .
the fragility of the generated states and their entanglement due to the standard dissipation and phase damping are discussed by numerically solving two types of master equations . |
recently , the physical meaning of the two parameters in static vacuum hypercylindrical solutions in five dimensions @xcite was correctly interpreted in ref . @xcite . in the analogy of weak field solutions for a cylindrical matter source distributed along the fifth direction
uniformly , the author identified the two parameters with mass " and tension " densities .
it is also pointed out that the well - known schwarzschild black string solution corresponds to the case that the tension - to - mass ratio is exactly one half .
this means that the schwarzschild black string , which was believed to be a family of solutions characterized by the mass density only , is indeed a special case of a wider class of solutions characterized by tension density as well .
note also that in @xmath3-dimensional spacetime , there is another class of stationary solutions which is characterized by the mass " and momentum " densities along the extra - directions @xcite .
the schwarzschild black string background is known to be unstable under small gravitational perturbations along the fifth direction - the so - called gregory - laflamme ( gl ) instability @xcite .
therefore , studying the physical role of tension might give a better understanding about what really causes the gl instability . in ref .
@xcite the geometric properties of this class of spacetimes with arbitrary tension in five dimensions were investigated in detail .
some of main properties are as follows . *
the solutions are classified by the tension - to - mass ratio @xmath4 .
the event horizon exists only when the tension density is a half of the mass density , _
i.e. _ @xmath5 . only in this case
the spacetime can be called a black string .
all spacetimes having other values of the tension - to - mass ratio than @xmath6 and @xmath7 have a naked singularity at ` center ' . *
even though there is a naked singularity instead of an event horizon , light radiated from it is infinitely red - shifted provided that @xmath8 .
* the geometry of some subspaces behaves interestingly . for @xmath9 or
@xmath10 the area of @xmath11 sphere monotonically decreases down to zero as one approaches to the naked singularity from infinity as usual . for @xmath12
, however , it bounces up and increases again up to infinity at the naked singularity as in the geometry of a wormhole spacetime .
* on the other hand , the proper length of a segment along the fifth direction shrinks down to zero for @xmath13 , but expands to infinity for @xmath14 as one approaches the naked singularity . *
although the @xmath11 area and the segment length compete each other , the total area of the segment @xmath15 turns out to decrease monotonically down to zero at the singularity except for @xmath16 . for the case of @xmath5
the area of @xmath11 becomes finite at the horizon and the scale factor of the fifth direction is constant .
the geodesic motions in this spacetime were also studied in ref .
@xcite . in this paper
we investigate how much these features change if the spacetime dimensionality becomes higher than five .
for the case of @xmath17 , for instance , since the area of @xmath11 sphere shrinks to zero whereas the proper length along the fifth direction diverges , one may expect that the total area might not shrink down to zero if one increases the number of extra dimensions large enough .
in addition , since there will be more number of tension parameters as the number of extra dimensions increases , one may wonder if there are more black brane solutions in addition to higher dimensional schwarzschild black branes .
higher dimensional spacetime solutions having translational symmetry along extra dimensions and spherical symmetry on slices perpendicular to them have been discovered by many authors in the literature in different contexts @xcite .
for instance , much more general solutions even in the presence of dilatonic scalar fields and antisymmetric forms were found in refs .
( see also references therein . )
however , the full and detailed analysis of the geometry and correct interpretation of the ` tension ' parameter have not been done so far as long as we know . in section
ii , we study the geometric properties of the trans - spherical vacuum solutions in @xmath1 dimensions in detail . in section iii
, we analyze the causal structure of the solution .
finally in section iv , we summarize our results and discuss physical implications of them .
we consider static solutions for the vacuum einstein equations in @xmath1 dimensions , which are invariant under translations along the extra @xmath0-dimensions and are spherically symmetric on the @xmath18-dimensions transverse to the @xmath0-dimensions .
the most general form of the metric with the symmetries may be written as @xmath19 here @xmath20 , @xmath21 , and @xmath22 are functions of the isotropic coordinate @xmath23 only .
this class of spacetime solutions will be characterized by the @xmath0-number of tension densities @xmath24 along @xmath25 and the adm mass density @xmath26 associated with the spatial and time translation symmetries , respectively @xcite .
interestingly , it turns out that the exponents in the metric component functions depend only on the dimensionless tension - to - mass ratio @xmath27 defined as @xmath28 for convenience , we define the sums of @xmath27 and @xmath29 by @xmath30 where @xmath31 runs over spatial extra - dimensions .
note that these definitions for @xmath4 and @xmath32 restrict values of @xmath32 such that @xmath33 as shown by the shaded region in fig . 1 below .
the solutions in this general higher dimensions were found by several authors @xcite and can be expressed as @xmath34 by using the adm mass and adm tensions . here
the integration constant @xmath35 is related to the adm mass @xmath26 by @xmath36^{\frac1{2}}\frac{4\pi g_d m}{\omega_{(n+2 ) } } , \label{m : ma}\end{aligned}\ ] ] where @xmath37 , and @xmath38 is the @xmath39-dimensional newton constant .
note that the quantity inside the square root in eq .
( [ m : ma ] ) is positive definite for all real values of @xmath27 .
the solutions ( [ sol2 ] ) are described by @xmath40 independent parameters , @xmath35 and @xmath27 .
it is interesting to see that the functions @xmath20 , @xmath21 , and @xmath41 depend only on the sums of @xmath27 and @xmath29 ( _ i.e. _ , @xmath4 and @xmath42 ) , not on the individual values of @xmath27 .
the double wick rotations @xmath43 and @xmath44 of the metric ( [ metric : ansatz ] ) also satisfies the vacuum einstein equation .
let us analyze geometrical properties of the spacetime solutions described above and see how they depend on the spacetime dimensionality .
first of all , one sees that these spacetimes become flat as either @xmath45 or @xmath46 .
the same happens in 4d schwarzschild and reissner - nordstrom solutions written in isotropic coordinates . from the form of solutions , we find that the component functions of the metric , @xmath20 , @xmath21 , and @xmath22 show different behaviors as @xmath47 , depending on the sign of their exponents . here
@xmath48 note that the region with @xmath49 is another copy of the spacetime region covered by @xmath50 as in the case of the @xmath3-dimensional one @xcite . ] .
although arbitrary values of @xmath35 ( _ e.g. _ , @xmath26 ) are allowed , we assume non - negative values of adm mass density for discussions below .
thus , @xmath51 . note also that the case of uniform tensions with @xmath52 for all @xmath31 corresponds to the well - known schwarzschild black @xmath0-brane in higher dimensions , @xmath53 where the event horizon is located at @xmath54 .
the other well known special cases are given by ( @xmath55 ) , the kaluza - klein bubble solution , @xmath56 which are related with the schwarzschild black @xmath0-brane solution mentioned above through the double wick rotation .
notice also that the metric becomes singular at @xmath54 .
it turns out that this is a coordinates singularity for the case of the schwarzschild black @xmath0-brane and the kaluza - klein bubble coordinate , _
e.g. _ , @xmath57 . ]
otherwise , it is a curvature singularity as will be shown below in detail . by investigating the spatial geometries of several spacelike hypersurfaces ,
we classify the solutions in the parameter space of @xmath27 .
note first that , in @xmath58 dimensions with @xmath59 , the area of @xmath11 diverges if @xmath60 , finite if @xmath5 or @xmath7 , whereas it vanishes if @xmath61 or @xmath62 @xcite .
in general , the area of the @xmath2 sphere at the @xmath63 surface is given by @xmath64 here @xmath65 is defined as @xmath66 as @xmath67 , this area increases as usual .
however , as @xmath47 , one can see that the property is crucially dependent on the signature of the exponent @xmath68 .
the area of @xmath54 surface has a non - vanishing finite value if @xmath69 . as shown by the thick curve in fig . 1 this equation gives the hyperbola defined by the following equation on @xmath4 and @xmath32 space @xmath70 with a restriction of @xmath71 .
note that this hyperbola intersects with the line @xmath72 one time if @xmath73 and two times if @xmath74 .
-brane solutions in @xmath75 .
the horizontal and vertical axes stand for @xmath4 and @xmath32 , respectively .
the definitions of @xmath76 and @xmath32 allow their values on the shaded region satisfying @xmath77 .
the dark shaded region is for physical parameters based on the strong energy condition .
as @xmath23 goes to @xmath78 , the area of the @xmath18-dimensional sphere @xmath79 becomes a non - vanishing finite value on the thick curve , the volume @xmath80 for extra @xmath0-dimensions on the vertical black line , and the total volume @xmath81 on the dashed curve . ] on the hyperbola , the area of the @xmath54 surface becomes @xmath82 therefore we see that , while only the cases of @xmath5 and @xmath7 in @xmath3-dimensions give a non - vanishing finite area @xmath83 , all solutions whose gravitational tensions are properly balanced in accordance to eq .
( [ an : fin ] ) have non - vanishing finite area @xmath79 .
namely , in the parameter space of @xmath0-dimensional @xmath27 , this region is given as the intersection of the @xmath84-dimensional sphere @xmath85 and a @xmath84-dimensional plane @xmath86 for each @xmath75 on the hyperbola ( [ an : fin ] ) in the shaded region in fig . 1 .
if @xmath87 , the behavior of the area is basically the same as the case of @xmath88 in @xmath3-dimensions .
namely , as @xmath23 decreases to @xmath78 from infinity , this area decreases only up to its minimum point @xmath89 given by @xmath90 and then starts to increase up to infinity instead of decreasing down monotonically .
therefore , this spatial geometry looks like a wormhole .
however , we point out that the @xmath91 surface can be reached from the throat ( _ i.e. _ , @xmath92 ) in a finite affine time . for @xmath73 , the value of @xmath93 does not have a maximum . on the other hand , for @xmath94
, it has a maximum . in fig . 1 ,
the corresponding parameter space is designated by the right - down side of the thick curve in the shaded region . on the other hand , if @xmath95 this area monotonically decreases down and shrinks to zero at @xmath54 as in the cases of @xmath61 or @xmath62 in @xmath3-dimensions . in fig . 1 ,
the corresponding parameter space is designated by the left - up side of the thick curve in the shaded region . in the special case of string configurations ( _ i.e.
_ , @xmath96 ) , @xmath4 and @xmath32 become same , and @xmath79 diverges as @xmath23 goes to @xmath78 in the region given by @xmath97 at @xmath98 , @xmath79 takes a non - vanishing finite value for the cases of @xmath99 or @xmath100 . for the cases of @xmath101 or @xmath102 , it vanishes .
for @xmath103 we recover the @xmath3-dimensional results .
now let us consider the spatial geometry in extra @xmath0-dimensions .
note first that , if @xmath104 for all @xmath31 , all scale factors of the extra space are constant since all @xmath105 .
this case corresponds to the schwarzschild black @xmath0-brane . for given values of tension - to - mass ratios @xmath106 ,
the @xmath25 direction becomes flat if @xmath107 equivalently , @xmath108 .
for other values of tension ratios , we see that @xmath22 in eq .
( [ sol2 ] ) becomes singular as @xmath23 goes to @xmath78 .
namely , as @xmath23 decreases to @xmath78 , the proper length of a unit segment with @xmath109 at @xmath110 constant surface shrinks to zero if @xmath111 , but it infinitely expands if @xmath112 .
however , we point out that all these cases do not necessarily correspond to singular spatial geometry in extra @xmath0-dimensions .
consider a kaluza - klein bubble solution along the @xmath113 direction , in which @xmath114 vanishes at @xmath54 with @xmath115 .
we can easily check that there exist no curvature singularity in this case .
indeed , the @xmath54 surface turns out to be a fixed point whose conical singularity can be removed by compactifying the @xmath25 coordinate suitably as mentioned before .
hence , the @xmath113 direction becomes a circle .
in general , if @xmath116 , one can find many solutions where some extra directions become circles in this way . however , all those solutions contain curvature singularity at @xmath54 except for the case of the kaluza - klein bubble solutions .
to summarize , we find that the regular spatial geometry in extra @xmath0-dimensions happens only for the schwarzschild black @xmath0-brane and kaluza - klein bubble solutions .
it is interesting to see how the total spatial volume @xmath81 for a given @xmath23 behaves as @xmath47 .
the total volume for @xmath117 constant surface per unit lengths of @xmath25 with respect to an asymptotic observer is given by @xmath118 where @xmath80 is the spatial volume of extra @xmath0-dimensions @xmath119 we see that the spatial volume @xmath80 becomes unit when @xmath120 which is designated by the vertical line in fig .
it is interesting to see that the total volume @xmath80 could be finite if the divergence at a certain direction can be precisely canceled out by the shrinking of some other directions .
as @xmath23 decreases to @xmath78 , @xmath80 monotonically decreases to zero for @xmath121 and monotonically increases to infinity for @xmath122 .
the total spatial volume of @xmath110constant section is given by @xmath123 thus the behavior of the total volume is determined by the exponent of @xmath124 $ ] , @xmath125 -(n+1)(a+1)^2 } { n(p+n+1)\sqrt{\bar a^2 + 1 -\frac{(a+1)^2}{p+n+1 } } \left(\sqrt{\bar a^2 + 1 -\frac{(a+1)^2}{p+n+1 } } + \sqrt{\frac{n}{n+1}}\right)}\end{aligned}\ ] ] which vanishes on a hyperbola ( the dashed curve given in fig.1 ) . from the definitions of @xmath4 and @xmath32 , the numerator of eq .
( [ a - exp ] ) can be reexpressed in the sum of squared terms given by @xmath126 ^ 2.\end{aligned}\ ] ] the eq .
( [ positive ] ) is non - negative for all real values of @xmath27 .
it vanishes only when @xmath127 the total volume with this set of @xmath27 becomes @xmath128 this case of uniform tensions is exactly the same as the case in which all extra @xmath0-directions are regular as mentioned above .
in addition , we point out that the geometry of the sphere is regular since this case satisfies eq .
( [ an : fin ] ) as well .
in fact , the solution with parameters in eq .
( [ pta ] ) corresponds to the schwarzschild black @xmath0-brane ( [ schwarz ] ) in higher dimensions .
since either @xmath79 or @xmath129 diverges depending on the values of tensions one may expect the total volume @xmath81 also diverges .
however , we claim that it never happens for any values of tension parameters .
this can be seen because the exponent of @xmath124 $ ] takes a positive value for all cases except for the case of uniform tensions so that the total volume @xmath81 vanishes as @xmath47 .
this happens because , even if the volume of the sphere diverges for certain values of parameters @xmath27 , @xmath80 shrinks more strongly for those values of parameters .
similarly , for the case in which @xmath80 diverges , @xmath79 shrinks more strongly .
the idea of making divergent total volume @xmath81 by adding expanding extra - dimensions arbitrarily does not work .
this is because as @xmath0 increases , both of the expanding rate in @xmath80 and shrinking in @xmath79 changes as well so that the net effect always gives vanishing total volume .
so far , we have not restricted the values of @xmath27 .
however , their physical ranges may be given by some energy conditions .
note that the strong energy condition in @xmath3-dimensional spacetime restricts the value of tension @xcite . in @xmath130-dimensional spacetimes , this extends as follows@xmath131 although we do not know whether the gravitational tensions satisfy the same condition , we assume that the same restriction holds as in the case of matter fields .
therefore , the physical range of @xmath4 may be given as , @xmath132 where the first inequality comes from the positivity theorem for gravitational tension , @xmath133 @xcite .
note that there exists a coordinate singularity at @xmath54 .
let us see whether this singularity is genuine one or not .
the kretschmann invariant for the metric ( [ sol2 ] ) is @xmath134 { k}^{2n}{\rho } ^{2n } { \nonumber}\right.\\ & & \quad\quad \left.- 4(1+n)\left ( \beta_1 ( 1+\beta_1 ^ 2 ) + n\beta_1(1 + 2\,\beta_2)+ n^2 \beta_3 \right ) ( { k}^{3n}{\rho}^n + { k^{n}}{\rho}^{3n } ) \right .
{ \nonumber}\\ & & \quad\quad \left .
+ ( 1 + n ) ( 2 + n ) ( \beta_1 ^ 2 + n\beta_2)({k}^{4n } + { \rho}^{4n } ) \right \}\end{aligned}\ ] ] where @xmath135 with @xmath136 .
here @xmath137 with @xmath138 is the half of the exponent of @xmath139 in @xmath22 .
we see that the right hand side of the first line becomes singular as @xmath47 since @xmath140 .
thus , there occurs a curvature singularity at @xmath54 unless the term inside the curly - bracket cancels this divergence .
we find that such cancelation happens only for the cases satisfying all the following conditions : @xmath141 by solving these algebraic equations above , we obtain either @xmath142 or , ( @xmath143 ) @xmath144 the first case ( [ calpha ] ) corresponds to the schwarzschild black @xmath0-brane .
the second case corresponds to the @xmath0 number of different kaluza - klein bubble solutions , which are related to the schwarzschild black @xmath0-brane by the double - wick rotations in eq .
( [ metric : ansatz ] ) . for the nonsingular cases ,
the value of the kretschmann invariant becomes @xmath145
let us consider the causal structure of these spacetime solutions .
it is enough to consider the @xmath25=constant surface with fixed angles .
then , the metric becomes @xmath146 here a tortoise coordinate @xmath147 is defined as @xmath148 the ingoing and outgoing null coordinates are defined by @xmath149 respectively . in order to see the causal properties of the @xmath54 surface ,
let us consider whether the null rays can escape from the surface or not .
the geodesic motion of an outgoing light in @xmath150 coordinates becomes @xmath151 in the vicinity of the @xmath54 surface . here
@xmath152 is given by @xmath153 where the equality holds only for @xmath154 for all positive @xmath31 .
the elapsed value @xmath155 for a light traveling from @xmath156 to @xmath157 outside becomes @xmath158 for the case of @xmath154 , which corresponds to the schwarzschild black brane , @xmath159 as @xmath160 .
this implies that the light can not escape from the @xmath54 surface and consequently the @xmath54 surface is indeed an event horizon . on the other hand
, @xmath161 takes a finite value for other cases .
therefore , the light can actually escape from the @xmath54 surface in a finite time . in this sense ,
the surface is not an event horizon and , in fact , is interpreted as a naked singularity because of the curvature singularity there .
the penrose diagram for the spacetime at @xmath25=constant is given in fig .
( a ) is for the case that all tension - to - mass ratio is @xmath162 , which corresponds to the schwarzschild black brane .
the surface of @xmath91 is indeed an event horizon .
the region of @xmath163 is not the inside of @xmath91 surface , but it is another copy of the spacetime as well known in the isotropic coordinate system .
the inside region can be obtained by analytic continuation of the spacetime covered by @xmath164 , revealing a space - like singularity at the center .
the case of kaluza - klein bubble is not shown here .
( b ) is for the case other than these two . here
the @xmath91 surface becomes a time - like singularity ( _ i.e. _ , a naked singularity ) , and the spacetime can not be analytically continued beyond this surface like the spacetime around the space - like singularity at the center in fig . ( a ) . [
cols="^,^ " , ]
we have investigated the geometrical properties of static vacuum @xmath0-brane solutions of einstein gravity in @xmath165 dimensions , which have spherical symmetry of @xmath2 orthogonal to the @xmath0-directions and are invariant under the translation along them .
the solutions are characterized by @xmath166 parameters , that are mass density @xmath26 and @xmath0-number of tension densities .
interestingly , some important geometric properties such as the area of @xmath167 and the total spatial volume ( [ r ] ) are characterized only by the three parameters ( @xmath168 ) , not by the individual value of tension ratios @xmath27 , where @xmath4 and @xmath169 are the summations of tension to mass ratio and of their squares , respectively , as defined in eq .
( [ a ] ) .
the surface area of @xmath2 sphere shows interesting behavior depending on the values of @xmath4 and @xmath32 .
there are three classes .
namely , i ) @xmath170 ; as @xmath23 decreases , this area does not decreases monotonically , but increases again as @xmath23 approaches to @xmath78 . in other words ,
the spatial geometry of the @xmath171 constant surface behaves like a wormhole geometry .
ii ) @xmath172 ; this case is designated by the thick curve ( @xmath79 ) in fig .
the schwarzschild black @xmath0-brane and the kaluza - klein bubble solutions belong to this class .
the area monotonically decreases to a non - vanishing finite value at @xmath54 .
iii ) @xmath173 ; the area monotonically decreases to zero at @xmath91 .
the spatial volume @xmath80 of unit extra @xmath0-directions takes a non - vanishing finite value only when @xmath174 .
as @xmath23 decreases to @xmath78 , it monotonically increases to infinity for @xmath122 and monotonically decreases to zero for @xmath121 .
although each @xmath79 and @xmath80 behaves variously as @xmath23 goes to @xmath78 , the total volume @xmath175 always shrinks to zero except for the case of @xmath154 for all @xmath31 .
it is interesting to see that this value is independent of the number of extra dimensions @xmath0 . for such exceptional case
, @xmath81 takes a finite value and it corresponds to the schwarzschild @xmath0-brane solution .
the curvature square diverges at @xmath54 except for the cases of the schwarzschild black @xmath0-brane ( [ schwarz ] ) and kaluza - klein bubble ( [ bubble1 ] ) .
this curvature singularity at the surface @xmath54 turned out to be naked .
some geometrical properties explicitly depend on each value of tensions .
in particular , the spatial geometry for each @xmath25 directions are dependent on @xmath27 in addition to @xmath4 and @xmath32 .
the proper length along @xmath25 becomes neither shrinking nor expanding as @xmath23 approaches to @xmath78 if the @xmath31-th tension takes a specific value which is determined by the values of other tensions as in eq .
( [ ai : regular ] ) .
the regular spatial geometry in all extra @xmath0-dimensions happens only for the schwarzschild black @xmath0-brane and kaluza - klein bubble solutions .
we have seen that the causal structure of the higher dimensional solutions are essentially the same as those of the five dimensional solutions @xcite .
namely , only the solutions where the value of tensions are equal to the mass divided by the number of angular coordinates _
i.e. _ @xmath176 have an event horizon located at @xmath54 .
as well known , for this schwarzschild black @xmath0-brane solutions , the spacetime can be continued beyond the @xmath54 surface and a spacelike curvature singularity appears inside the event horizon as shown in fig . 2 .
the kalua - klein bubble solution , where one of its tensions is given by mass times the number of angular coordinates and all other tensions are equal to the mass , do not possess an event horizon and the signature of this spacetime changes across the @xmath54 surface . beyond the @xmath177 surface .
note that , in the isotropic coordinate system , @xmath178 does not describe the inside of the @xmath91 surface .
] other than these two cases , the curvature singularity appears at @xmath54 surface and light can escape from this surface toward the spatial infinity .
consequently , this singularity is naked in nature .
investigation of such singular spacetimes might be meaningless physically .
however , studying the geometrical structure of such singular spacetimes in detail may be very important for the following reason .
it is interesting to see that the schwarzschild black brane is so special for geometrical properties in the whole solution space .
this fact was already pointed out in refs .
in particular , when one considers spherically symmetric perturbations with translational symmetries along extra - directions untouched , the schwarzschild black brane solution is singled out because it is stable under such perturbations whereas all other solutions possessing naked singularities are catastrophically unstable @xcite . on the other hand ,
the spacetimes we considered may be formed from the higher dimensional trans - spherical distribution of normal matter .
our study shows that a slight deviation of tension and mass densities from those evolving to the schwarzschild black brane will end up with a spacetime configuration possessing a naked singularity . assuming that the cosmic censorship conjecture is valid in higher dimensions as well
, we do not expect the development of a naked singularity .
this observation may indicate that some quantum effect becomes important right before forming a naked singularity during the gravitational collapse .
in fact , yoshimura and tanaka @xcite showed numerically that the naked singularity existing at some static spacetimes near the schwarzschild one in parameter space becomes null if the gauss - bonnet gravity term is added in 5-dimensions .
it is interesting to see how much various quantum effects change the formation of naked singularity in trans - spherical gravitational collapse . finally , it is also interesting to study how much the geometrical properties are modified if angular momentum or charge are added in the consideration .
these interesting issues deserve future work .
this work was supported by the korea research foundation grant funded by the korean government(moehrd , basic research promotion fund)(krf-2008 - 314-c00063 ) and the topical research program of apctp and in part by the e - science project of kisti .
gk was supported in part by the korea science and engineering foundation ( kosef ) grant funded by the korea government ( mest ) ( no .
r01 - 2008 - 000 - 20866 - 0 ) .
d. kramer , acta phys .
b2 * , 807 ( 1971 ) ; a. chodos and s. detweiler , gen .
rel . grav .
* 14 * , 879 ( 1982 ) ; j. gross and m. m. perry , nucl . phys . *
b226 * , 29 ( 1983 ) ; a. davidson and d. a. owen , phys . lett . * b155 * , 247 ( 1985 )
. r. gregory and r. laflamme , phys .
lett . * 70 * , 2837 ( 1993 ) ; s. s. gubser and i. mitra , jhep * 0108 * , 018 ( 2001 ) ; g. kang and j. lee , jhep * 0403 * , 039 ( 2004 ) ; t. harmark and n. obers , [ hep - th/0503020 ] ; h. kudoh , phys . rev .
* d73 * , 104034 ( 2006 ) ; y. brihaye , t. delsate , and e. radu , ( 2007 ) [ arxiv:0710.4034 ] .
i. cho , g. kang , s. p. kim and c. h. lee ; j. korean phys .
soc . * 53 * , 1089 ( 2008 ) , [ arxiv:0709.1021 ] ; geometrical properties of hypercylindrical vacuum solutions with arbitrary tension in five dimensions , " _ in preparation_. k. a. bronnikov , v. d. ivaschchuk , and v. n. melnikov , grav . cosmol . *
3 * , 205 ( 1997 ) , [ arxiv : gr - qc/9710054 ] ; k. a. bronnikov , grav .
cosmol . * 4 * , 49 ( 1998 ) , [ arxiv : hep - th/9710207 ] ; v. d. ivashchuk , on exact solutions in multidimensional gravity with antisymmetric forms " , in proceedings of the 18th course of the school on gravitational theories and experiments ( 2003 , erice ) , ed . by g.
t. gillies , v. n. melnikov , and v. de sabbata , ( kluwer academic publishers , dordrecht , 2004 ) , 39 [ arxiv : gr - qc/0310114 ] .
u. bleyer , k. a. bronnikov , s. b. fadeev , and v. n. melnikov , on black hole stability in multidimensional gravity , " [ gr - qc/9405021 ] ; k. a. bronnikov and v. n. melnikov , ann .
( n.y . ) , * 239 * , 40 ( 1995 ) ; nucl .
* b 584 * , 434 ( 2000 ) . | we investigate the geometrical properties of static vacuum @xmath0-brane solutions of einstein gravity in @xmath1 dimensions , which have spherical symmetry of @xmath2 orthogonal to the @xmath0-directions and are invariant under the translation along them .
the solutions are characterized by mass density and @xmath0 number of tension densities .
the causal structure of the higher dimensional solutions is essentially the same as that of the five dimensional ones .
namely , a naked singularity appears for most solutions except for the schwarzschild black @xmath0-brane and the kaluza - klein bubble .
we show that some important geometric properties such as the area of @xmath2 and the total spatial volume are characterized only by the three parameters such as the mass density , the sum of tension densities and the sum of tension density squares rather than individual tension densities .
these geometric properties are analyzed in detail in this parameter space and are compared with those of 5-dimensional case . |
low - dimensional time - discrete maps are among the most important models for exploring different aspects of chaos .
these systems display a very rich dynamical behavior but are still very amenable to straightforward computer simulations .
even more , in some cases rigorous analytical solutions are possible . after it was realized that diffusion processes can be generated by microscopic deterministic chaos in the equations of motion , time - discrete maps became useful tools in deterministic transport theory .
the analysis of these simple models required to suitably combine nonequilibrium statistical mechanics with dynamical systems theory leading to a more profound understanding of transport in nonequilibrium situations @xcite . however , time - discrete maps provide not only a suitable starting point for studying normal diffusion but also for investigating the anomalous case @xcite .
moreover , there are certain classes of more realistic models which share specific properties of maps such as being low - dimensional and exhibiting certain periodicities .
indeed , theoretical investigations of chaotic billiards subject to external fields @xcite , of periodic lorentz gases @xcite , and of pendulum - like differential equations @xcite showed that many properties of deterministic transport in maps carry over to these more complex chaotic dynamical systems . in this framework , recently a new feature of deterministic diffusion was discovered . for simple one - dimensional hyperbolic maps
it was shown that the diffusion coefficient is typically a fractal function of control parameters @xcite .
subsequently an analogous behavior was detected for other transport coefficients @xcite , and in more complicated models @xcite .
however , up to now the fractality of transport coefficients could be assessed for hyperbolic systems only , whereas , to our knowledge , the fractal nature of classical transport coefficients in the broad class of nonhyperbolic systems was not discussed .
on the other hand , studying nonhyperbolic dynamics appears to be more relevant in order to connect fractal transport coefficients to some known experiments .
here we think particularly of dissipative systems driven by periodic forces such as josephson junctions in the presence of microwave radiation @xcite , superionic conductors @xcite , and systems exhibiting charge - density waves @xcite in which certain features of deterministic diffusion were already observed experimentally . for these systems
the equations of motion are typically of the form of some nonlinear pendulum equation . in the limiting case of strong dissipation
they can be reduced to nonhyperbolic one - dimensional time - discrete maps sharing certain symmetries @xcite .
the so - called climbing sine map is a well - known example of this class of maps @xcite . in this paper
we pursue a detailed analysis of the diffusive and dynamical properties of the climbing sine map . particularly , we show that the nonhyperbolicity of this map does not destroy the fractal characteristics of deterministic diffusive transport as they were found in hyperbolic systems . on the contrary ,
fractal structures appear for normal diffusive parameters as well as for anomalous diffusive regions .
we argue that higher - order memory effects are crucial to understand the origin of these fractal hierarchies in this nonhyperbolic system . by using a green - kubo formula for diffusion ,
the dynamical correlations are recovered in terms of fractal takagi - like functions .
these functions appear as solutions of a generalized integro - differential de rham - type equation .
we furthermore show that the distribution of periodic windows exhibiting anomalous diffusion forms devil s staircase like structures as a function of the parameter and that the complementary sets of chaotic dynamics have a positive measure in parameter space that increases by increasing the parameter value . our paper is organized as follows . in sec .
ii we introduce the model . in sec .
iii we explore the coarse functional form of the parameter - dependent diffusion coefficient and discuss it in relation to previous results on hyperbolic maps . in sec .
iv our analysis is refined revealing complex scenarios of anomalous diffusion , which are explained in terms of correlated random walk approximations . in sec .
v generalized fractal takagi functions are constructed for the climbing sine map and the connection to the diffusion coefficient is worked out .
periodic windows exhibiting anomalous diffusion are studied in detail in sec .
we then draw conclusions and discuss our results in the final section .
the one - dimensional climbing sine map is defined as .
the dashed line indicates the orbit of a moving particle starting from the initial position @xmath0.,title="fig:",scaledwidth=50.0% ] + @xmath1 where @xmath2 is a control parameter and @xmath3 is the position of a point particle at discrete time @xmath4 .
obviously , @xmath5 possesses translation and reflection symmetry , @xmath6 the periodicity of the map naturally splits the phase space into different boxes , @xmath7 $ ] , @xmath8 , as shown in fig.[fig1 ] .
( [ map ] ) as restricted to one box , i.e. , on a circle , we call the reduced map , @xmath9 the probability @xmath10 to find a particle at a position between @xmath11 and @xmath12 at time @xmath4 then evolves according to the continuity equation for the probability density @xmath13 , which is the frobenius - perron equation @xcite @xmath14 the stationary solution of this equation is called the invariant density , which we denote by @xmath15 . due to its nonhyperbolicity
, the climbing sine map possesses a rich dynamics consisting of chaotic diffusive motion , ballistic dynamics , and localized orbits . under parameter variation these different types of dynamics are highly intertwined resulting in complicated scenarios related to the appearance of periodic windows @xcite . in order to study diffusion we will be interested in parameters that are greater than @xmath16 for which the extrema of the map exceed the boundaries of each box for the first time indicating the onset of diffusive motion .
in this section we explore the relationship between nonlinear maps like the climbing sine map and simple piecewise linear maps for which , in contrast to the climbing sine map , the diffusion coefficient can be calculated exactly .
it is well - known that in special cases such different types of maps are linked to each other via the concept of conjugacy .
indeed , we show that maps which are approximately conjugate to each other exhibit a very similar oscillatory behavior in the diffusion coefficient on coarse scales .
our argument refers to some existing methods for calculating the diffusion coefficient of piecewise linear maps , which we briefly review .
we then describe how we numerically calculated the complete parameter dependence of the diffusion coefficient for the climbing sine map and discuss a first result .
one speaks of normal deterministic diffusion if the mean square displacement of an ensemble of moving particles grows linearly in time . the diffusion coefficient is then given by the einstein relation @xmath17 where the brackets denote an ensemble average over the moving particles .
there exist various efficient numerical as well as , for some system parameters , analytical methods to exactly compute diffusion coefficients for piecewise linear hyperbolic maps , such as transition matrix methods based on markov partitions @xcite , cycle expansion methods@xcite , and more recently a very powerful method related to kneading sequences @xcite .
we first restrict our analysis of diffusion in the climbing sine map to parameters for which there are simple markov partitions . for one - dimensional maps ,
a partition is a markov partition if and only if parts of the partition get mapped again onto parts of the partition , or onto unions of parts of the partition , see ref .
@xcite and further references therein .
an example of a markov partition consisting of five parts is shown in the inset of fig.[d_mark ] . in case of the climbing sine map markov partitions
can be constructed simply by forward iteration of one of the critical points @xmath18 defined by the condition that @xmath19 in the reduced map . if higher iterations of this point fall onto a periodic orbit a markov partition exits .
indeed , if a markov partition is known , for piecewise linear maps the diffusion coefficient can often be calculated analytically via calculating the invariant measure of the map or via computing the second largest eigenvalue of the frobenius - perron operator written in form of a transition matrix .
one can now identify an infinite series of parameter values corresponding to a certain type of markov partition @xcite .
for parameter values which belong to such a markov partition series the corresponding invariant densities @xmath20 have a very similar functional form . note that , in case of nonlinear maps , singularities in the invariant density exactly correspond to the iteration of the critical point @xmath18 @xcite .
an example of @xmath20 for one series of parameter values ( marked as filled circles ) is shown in the inset of fig .
[ d_mark ] . by using respective series of markov partitions
piecewise linear maps can be related to nonlinear maps . for this purpose
let us consider , along with the climbing sine map , ( i ) the piecewise linear zig - zag map @xcite
@xmath21 with @xmath22 , @xmath23 and @xmath24 , and ( ii ) the nonlinear cubic map @xcite @xmath25 the definitions of both maps are given on the unit interval . in order to compare the diffusion coefficient of these different maps ,
the parameters @xmath26 were chosen such that the maps all display the same height @xmath27 , defined as the distance between the first iteration of the leftmost critical point @xmath28 and the zero bound in the first box @xmath29 $ ] .
thus , @xmath30 corresponds to the onset of diffusion for all three maps .
for the two simple markov partition series @xmath31 and @xmath32 , corresponding to integer and half - integer values of @xmath27 , respectively , the diffusion coefficient of the zig - zag map can be calculated analytically . for integer values of @xmath27
the result reads @xcite @xmath33 for half - integer values of @xmath27 the diffusion coefficient is easily calculated analytically , e.g. , by transition matrix methods @xcite , to @xmath34 in case of the climbing sine map and of the cubic map the diffusion coefficient was obtained from computer simulations by evaluating the mean square displacement eq .
( [ d_a ] ) for the same series of markov partitions .
results are shown in fig .
[ d_mark ] . for
the climbing sine map some more markov partition series points ( alltogether five different series ) were included .
for all three maps there is a very analogous oscillatory behavior of the parameter - dependent diffusion coefficient .
these oscillations can be explained in terms of the changes of the microscopic dynamics under parameter variation , that is , whenever there is a local maximum there is an onset of strong backscattering in the dynamics yielding a local decrease of the diffusion coefficient in the parameter , and vice versa at local minima @xcite .
however , the five markov partition series for the climbing sine diffusion coefficient already indicate that there are more irregularities on finer scales . for piecewise linear maps , the origin of these irregularities
was identified to be the topological instability of the dynamics under parameter variation @xcite .
that is , a small deviation of the parameter changes the markov partition and the corresponding invariant density which , in turn , is reflected in a change of the value of the diffusion coefficient .
note that the dependence of the diffusion coefficient for a single markov partition series appears to be a monotonously increasing function of the parameter @xcite .
nevertheless , computing @xmath35 for more and more markov partitions series will reveal more and more irregularities in @xmath35 thus forming a fractal structure @xcite .
since the climbing sine map shares the same topological features as piecewise linear maps in terms of these series of markov partitions , one may wonder whether it is not possible to straightforwardly calculate the diffusion coefficient for nonlinear maps from the one of piecewise linear maps by using the concept of conjugacy @xcite , see also the definition in appendix a. in fact , it was stated by grossmann and thomae @xcite that the diffusion coefficient is invariant under conjugacy , however , without giving a proof . in appendix a such a proof is provided .
unfortunately , conjugacies are explicitly known only in very specific cases and for maps acting on the unit interval @xcite .
as soon as the map extrema exceed the unit interval , which is reminiscent of the onset of diffusive behavior , only some approximate , piecewise conjugacies could be constructed in a straightforward way , see ref .
@xcite for an example .
we now apply this reasoning along the lines of conjugacy in order to understand the similarities between the diffusion coefficient of the three maps as displayed in fig
. 2 . the functional form of the cubic map can be obtained from a taylor series expansion of @xmath36 by keeping terms up to third order thus representing a low - order approximation of the climbing sine map .
this seems to be reflected in the fact that at any odd integer parameter value of @xmath27 the climbing sine map has an invariant density whose functional form is very close to the one of the cubic map at parameter value @xmath30 , @xmath37 .
hence , one may expect that both diffusion coefficients are possibly trivially related to each other , however , note the increasing deviations between the respective results at larger @xmath27 . for @xmath30 , the cubic map and the piecewise linear zig - zag map are now in turn conjugate to each other @xcite .
however , for @xmath38 we are not aware of the existence of any exact conjugacy between zig - zag and cubic map .
still , along the lines of ref .
@xcite one can at least approximately relate both maps to each other via using piecewise conjugacies .
this explains why the zig - zag map and the climbing sine map display qualitatively the very same oscillatory behavior in the diffusion coefficient , somewhat linked by diffusion in the cubic map . in summary , by using markov partitions and by arguing with the concept of conjugacy we have shown that the structure of the diffusion coefficient for the nonlinear climbing sine map has much in common with the one of respective piecewise linear maps , in the sense of displaying a non - trivial oscillatory parameter dependence .
however , to use conjugacies in order to exactly calculate the diffusion coefficient for nonlinear maps does not appear to be straightforward @xcite , hence in the following we restrict ourselves to alternative methods as discussed in the next subsection . ) ( squares ) , the climbing sine map ( circles ) , and the cubic map eq .
( [ cub_map ] ) ( triangles ) .
the values for the zig - zag map represent analytical results , see eqs.([cub_d1],[cub_d2 ] ) , the remaining values are from computer simulations .
the lines are guides for the eyes .
the inset shows an example of a markov partition for the climbing sine map on the unit interval and the corresponding invariant density.,title="fig:",scaledwidth=70.0% ] + in order to obtain the full parameter dependence for the diffusion coefficient of the climbing sine map we numerically evaluated the green - kubo formula for diffusion in maps @xcite reading @xmath39 here the angular brackets denote an average over the invariant density of the reduced map , @xmath40 .
the jump velocity @xmath41 is defined by @xmath42-[x_n]\equiv [ m_a(x_n ) ] \ : , \label{jumps_vel}\ ] ] where the square brackets denote the largest integer less than the argument .
points.,title="fig:",scaledwidth=70.0% ] + the sum @xmath43 gives the integer value of the displacement of a particle after @xmath4 time steps that started at some initial position @xmath11 , and we call it jump velocity function .
( [ gk ] ) defines a time - dependent diffusion coefficient which , in case of normal diffusion , converges to @xmath44 in our simulations for fig .
[ d_5box ] we truncated @xmath45 after @xmath46 time steps .
the invariant density was obtained by solving the continuity equation for @xmath47 eq .
( [ fp_eq ] ) with the histogram method of ref .
note that both the integer displacement and the density are coupled via eq .
( [ gk ] ) .
results for @xmath35 are shown in fig .
[ d_5box ] for a large range of parameters demonstrating a highly non - trivial behavior of the diffusion coefficient .
the large - scale oscillations as predicted from the simple markov partition series ( see fig.[d_mark ] ) are still clearly seen , however , on top of this there exist further oscillations on finer scales .
these are regions of anomalous diffusion that manifest themselves in form of abrupt divergences , @xmath48 , or by a vanishing diffusion coefficient , @xmath49 .
in this section we study the parameter - dependent diffusion coefficient in more detail . based on the green - kubo formula
we derive a systematic hierarchy of approximations for the diffusion coefficient and show how they can be used to understand the complex behavior of this curve in more detail .
we are first interested in understanding the coarse functional form of the parameter - dependent diffusion coefficient in the limit of very small and very large parameter values . for this purpose
we use simple random walk approximations that are based on the assumption of a complete loss of memory between the single jumps .
such an analysis was already performed for hyperbolic piecewise linear maps @xcite . here
we apply the same reasoning to the nonlinear case of the climbing sine map .
we start in the limit of very small parameter values , i.e. , near the onset of diffusion . here
we assume that particles make either a step of length one to the left or to the right , or just remain in the box . the transition probability is then given by integrating over the respective invariant density in the escape region .
putting all this information into eq .
( 5 ) yields @xcite @xmath50 where @xmath51 . making the additional approximation that @xmath52 we get @xmath53 the other limiting case concerns values of @xmath54 . here
the precise value of the width of the escape region is much less important than the precise value of the step length which is very large , hence by again assuming that @xmath52 eq .
( 5 ) can be approximated to @xcite @xmath55 , which give an indication of the exact functional form .
the magnification ( inset in normal scale ) of a small region around @xmath56 shows more irregularities on a finer scale pointing towards a fractal structure of the diffusion coefficient.,title="fig:",scaledwidth=70.0% ] + these two asymptotic random walk approximations are shown in fig.[crisis ] as bold lines .
one can clearly see that there is a dynamical crossover between the different functional forms of these two asymptotic regimes .
this crossover was first observed in piecewise linear maps and appears to be typical for diffusive systems exhibiting some spatial periodicity @xcite .
it was lateron also verified for diffusion in the periodic lorentz gas @xcite .
the coarse functional form of the random walk approximations should be compared to a certain series of higher - order approximations based on the green - kubo formula eq .
( [ gk ] ) , which is also shown in the figure .
these refined approximations are getting closer and closer to the exact functional form , as explained below , and thus give a good indication for these exact values . in the previous subsection
we have outlined two simple random walk approximations for diffusion that do not include any memory effects .
however , one can do better by systematically evaluating the single terms as contained in the series expansion of the green - kubo formula eqs.([gk],[vel_func ] ) . for a simple piecewise linear map and for the periodic lorentz gas this was done in ref .
@xcite and provided a simple approach to understand the origin of complex structures in the diffusion coefficient on fine scales .
the basic idea of this approach is as follows : the green - kubo formula eq.([gk ] ) splits the dynamics into an inter - box dynamics , in terms of integer jumps , and into an intra - box dynamics , as represented by the invariant density .
we first approximate the invariant density in eq .
( [ gk ] ) to @xmath57 irrespective of the fact that it is typically a very complicated function of @xmath11 and @xmath26 @xcite .
the resulting approximate diffusion coefficient we label with a superscript in eq .
( [ gk ] ) , @xmath58 .
the term for @xmath59 obviously excludes any higher - order correlations and was already worked out in form of the simple random walk approximation eqs .
( [ rw])-([rw_2 ] ) .
the generalization @xmath60 , which systematically includes more and more dynamical correlations , may consequently be denoted as _ correlated random walk approximation _ @xcite .
we now use this expansion to analyze the parameter dependence of the diffusion coefficient of the climbing sine map in terms of such higher - order correlations . , see eq .
( [ gk ] ) with uniform invariant density @xmath61 , for @xmath62 .
note the quick convergence for normal diffusive parameters .
the dashed lines define periodic windows , which are the same as in fig.[dif ] .
the insets ( b ) and ( c ) contain two magnifications of ( a ) in the region close to the onset of diffusion for @xmath63 only .
they show self - similar behavior on smaller and smaller scales.,title="fig:",scaledwidth=70.0% ] + fig .
[ crisis ] depicts results for @xmath64 at @xmath65 over a large range of parameters , whereas fig .
[ frw ] ( a ) presents a respective detailed analysis for the region close to the onset of diffusion , i.e. , for parameters @xmath66 $ ] , showing results for @xmath64 at @xmath67 .
the series of approximations in fig.[crisis ] clearly reveals finer and finer sequences of oscillations that eventually converge to a fractal structure , as is particularly shown in the inset of this figure .
however , here the order of the expansion is not large enough to identify parameter regions of anomalous diffusion .
these regions can be better seen in fig .
[ frw ] , where three different cases of parameter regions can be distinguished : ( i ) regions with quick convergence of this approximation corresponding to normal diffusion ( ii ) divergence of @xmath58 corresponding to ballistic motion , in agreement with @xmath68 , and ( iii ) localized dynamics where @xmath58 alternates in @xmath4 between two solutions , with @xmath69 for the exact diffusion coefficient .
this oscillation points to the dynamical origin of localization in terms of certain period - two orbits . that these approximate solutions are non - zero is due to the fact that the invariant density was set equal to one .
the dashed lines in fig.[frw ] indicate the largest regions of anomalous diffusion .
the approximate diffusion coefficient @xmath63 of this figure is compared to the `` numerically exact '' one in fig .
here `` numerically exact '' we wish to be understood in the sense that no further ad hoc - approximations are involved , i.e. , we evaluated the green - kubo formula according to the numerical method described in sec .
iii . by truncating it after @xmath70 time steps .
this comparison shows that in case of normal diffusion our approximation nicely reproduces the irregularities in the non - approximated diffusion coefficient . like the inset of fig .
[ crisis ] , the magnifications in fig.[frw ] give clear evidence for a self - similar structure of the diffusion coefficient .
these results thus show that dealing with correlated jumps only yields a qualitative and to quite some extent even quantitative understanding of the regions of normal and anomalous diffusion in the climbing sine map .
the impact of specific features of the microscopic dynamics on the diffusion coefficient is nicely elucidated by comparing the bifurcation diagram of the reduced climbing sine map eq .
( [ reduced_map ] ) with the numerically exact diffusion coefficient , see fig .
[ dif ] . as one can see in the upper panel of fig .
[ dif ] , the bifurcation diagram consists of ( infinitely ) many periodic windows . whenever there is a window the dynamics of eq .
( [ map ] ) is either ballistic or localized @xcite .
fig.[dif ] demonstrates the strong impact of this bifurcation scenario on the diffusion coefficient .
for localized dynamics , orbits are confined within some finite interval in phase space implying subdiffusive behavior for which the diffusion coefficient vanishes , whereas for ballistic motion particles propagate superdiffusively with a diverging diffusion coefficient being proportional to @xmath71 . only for normal diffusion @xmath35
is nonzero , finite and the limit in eq .
( [ dlim ] ) exists . at the boundaries of each periodic window the diffusion coefficient is related to intermittent - like transient behavior eventually resulting in normal diffusion with @xmath72 @xcite . in comparison with the correlated random walk approximation @xmath63 ( dots ) .
the dashed vertical lines connect regions of anomalous diffusion , @xmath68 or @xmath73 , with ballistic and localized windows , respectively , of the bifurcation diagram.,title="fig:",scaledwidth=70.0% ] +
in this section we further analyze the dynamical origin of the different structures in the parameter - dependent diffusion coefficient by constructing objects called fractal generalized takagi functions .
these functions somewhat resemble usual takagi functions but , as will be shown , they fulfill a more complicated type of functional recursion relations than standard de rham - type equations .
interestingly , takagi functions were known to mathematicians since about a hundred years @xcite , however , in the field of chaotic transport they were appreciated by physicists only very recently @xcite .
we first show how to construct fractal generalized takagi functions and study their properties with respect to the three different types of dynamics in the climbing sine map .
we then relate these objects directly to the diffusion coefficient . from the definition of the time - dependent jump velocity function there
follows the recursion relation @xcite @xmath74 since the time - dependent jump velocity function @xmath75 is getting extremely complicated after some time steps , we introduce the more well - behaved function @xmath76 integration of eq .
( [ rec ] ) yields the recursive functional equation @xmath77 with the integral term @xmath78 ( upper curve ) and @xmath79 ( lower curve ) .
( b ) , ( c ) : an example of nonconverging iterations of the generalized fractal takagi functions for the climbing sine map with parameters corresponding to ( b ) ballistic dynamics at @xmath80 and to ( c ) localized dynamics at @xmath81 , both for the time steps @xmath82 .
note the divergence of the iterations in ( b ) and the alternation between two states in ( c).,title="fig:",scaledwidth=55.0% ] + where @xmath83 is the second derivative of the inverse function of @xmath84 .
the function @xmath85 is given by @xmath86 where @xmath87 is defined to be constant on each subinterval where the jump velocity @xmath41 has a given value .
this constant is fixed by the condition for @xmath88 to be continuous on the unit interval supplemented by @xmath89 the generalized takagi function is now defined in the long - time limit of eq.([takeq ] ) , @xmath90 for piecewise linear hyperbolic maps @xmath91 in eq .
( [ takeq2 ] ) simply disappears , and the derivative in front of the second term of eq .
( [ takeq ] ) reduces to the local slope of the map thus recovering ordinary de rham - type equations @xcite .
it should be noted that for smooth nonlinear maps like the climbing sine map the reduced map @xmath84 is generally not invertible . in order to define a local inverse of @xmath84 , we split the unit interval into subintervals on which this function is piecewise invertible .
thus eq .
( [ takeq ] ) should be understood as a series of equations where each part is defined for a respective piecewise invertible part of @xmath84 .
the detailed derivation of eq .
( [ takeq ] ) is given in appendix b. it is not known to us how to directly solve this generalized de rham - type integro - differential equation for the climbing sine map , however , solutions can alternatively be constructed from eqs .
( [ takagi ] ) , ( [ t_a ] ) , ( [ t_0 ] ) starting from simulations .
results are shown in fig.[takfig ] . for normal diffusive parameters
the limit of eq.([takfun ] ) exists and the respective curves are fractal on the whole unit interval somewhat resembling standard takagi functions @xcite .
however , in case of periodic windows @xmath92 either diverges due to ballistic flights , or it oscillates indicating localization .
interestingly , in these functions the corresponding attracting sets appear in form of smooth , non - fractal regions on fine scales as marked by the dashed lines in fig .
[ takfig ] . the diffusion coefficient can now be formulated in terms of these fractal functions by carrying out the integrations contained in eq .
( [ gk ] ) . for simplicity we restrict ourselves to the parameter region of @xmath93 $ ]
in which the respective solution reads @xmath94 - d_0^{\rho}(a ) , \label{dtak}\ ] ] where @xmath95 , is defined by @xmath96:=1 $ ] , and @xmath97 .
our previous approximation @xmath58 with @xmath98 is recovered from this equation in form of @xmath99 - d_0 ^ 1(a)\;. \label{dtak2}\ ] ] hence , eqs .
( [ dtak ] ) , ( [ dtak2 ] ) explicitly relate the generalized fractal takagi functions shown in fig .
[ takfig ] to the fractals of fig.[dif ] .
one of the most important problems regarding periodic windows in maps remains the question of their total measure .
much understanding has been achieved for one - dimensional unimodal maps @xcite .
based on the sharkovskii theorem about the ordering of periodic orbits @xcite , metropolis , stein and stein organized periodic windows in universal symbolic sequences ( u - sequences ) such that the sequence of next order is uniquely determined by the previous one @xcite .
later jacobson came up with the proof that chaotic parameter values in one - dimensional unimodal maps with a single maximum do have positive measure @xcite .
related numerical studies were made by farmer @xcite .
furthermore , it was shown that periodic windows in such a map form so - called fat fractal cantor - like sets with positive measure .
however , for diffusive maps on a line , apart from the preliminary studies of refs.@xcite , nothing appears to be known . on the other hand , as was exemplified in sec .
iv there is an intimate relation between the irregular structures of the diffusion coefficient and the occurrence of periodic windows .
hence , in this section we investigate the periodic windows for the climbing sine map in full detail . due to the spatial extension of our model a new type of periodic motion , which is not present in unimodal maps , exists , which is that particles move on average in one direction .
ballistic solutions _ are born through tangent bifurcations , further undergo a feigenbaum - type scenario and die at crises points @xcite . _
localized solutions _ occurr at even periods only and start with tangent bifurcations followed by a symmetry breaking at slope - type bifurcation points @xcite . in this case
the bifurcation scenario is much more complex .
obviously , periodic windows are related to certain periodic orbits , thus there are infinitely many of them , and they are believed to be dense in the parameter space @xcite . of all period @xmath100-windows ( lines with symbols ) for the first three parameter intervals ( from top to bottom ) as defined in the text .
the dotted lines represent exponential fits ; for the parameters see table [ fit ] .
the measures corresponding to windows with localized orbits are shown as pluses .
lower panel : the partial sum @xmath101 for all periodic windows at a certain period @xmath100 .
the dashed curves represent approximations as calculated from eq .
( [ sp ] ) , the straight lines are their limiting values at @xmath102.,title="fig:",scaledwidth=60.0% ] + of all period @xmath100-windows ( lines with symbols ) for the first three parameter intervals ( from top to bottom ) as defined in the text .
the dotted lines represent exponential fits ; for the parameters see table [ fit ] .
the measures corresponding to windows with localized orbits are shown as pluses .
lower panel : the partial sum @xmath101 for all periodic windows at a certain period @xmath100 .
the dashed curves represent approximations as calculated from eq .
( [ sp ] ) , the straight lines are their limiting values at @xmath102.,title="fig:",scaledwidth=60.0% ] + by dividing the parameter line into subsets labeled by the integer value of the map maximum on the unit interval , @xmath103 = j , \ ; j \in\mathbb{z}$ ] , we computed all windows of a certain period @xmath100 in a certain subset .
the numerical procedure which was used for these computations is outlined in appendix c. let @xmath104 denote the total measure of all period @xmath100-windows in a subset @xmath105 and let @xmath101 be the partial sum of @xmath104 defined by @xmath106 . in fig .
[ fig5 ] @xmath107 is plotted as a function of the period for the three first subsets @xmath108 .
the measures corresponding to windows with localized orbits are shown in fig .
[ fig5 ] as pluses .
is it clear that they make the major contribution to the total measure for even periods hence explaining the origin of the pronounced oscillatory behavior of @xmath109 .
.fit parameters for the exponential decrease of the measure at even and odd periods for the first three subsets of the map control parameter labeled by @xmath105 .
[ fit ] [ cols="<,<,<,<,<",options="header " , ] is the integrated number of period six - windows .
the inset shows a blowup of the initial region.,title="fig:",scaledwidth=70.0% ] + the main message of table [ mes ] is that the measure of periodic windows is different for a different subset @xmath105 and obviously decreases by increasing the control parameter .
correspondingly , the measure of chaotic solutions increases as the parameter of the system is getting larger . in order to make this more quantitative we consider the dependence of the measure of period @xmath100-windows as a function of the box index @xmath105 .
first we check period one - windows .
since in each box there exists only one window of this period , we are able to go to up to @xmath110 . in the upper panel of fig .
[ fig6 ] the logarithm of @xmath111 is plotted against @xmath112 .
we find that the slope of this function almost exactly equal to @xmath113 .
the behavior of @xmath114 for different @xmath100 is shown in the lower panel of fig .
the slope of the line for period two is also @xmath113 , for period three and four it is @xmath115 , and for period five it is @xmath116 .
[ fig6 ] shows that even and odd periods decrease with respectively different laws , where the decay rate appears to be precisely given by the periodicity of the windows according to @xmath117 for even periods and @xmath118 for odd periods . , where @xmath119 is the total measure of windows that are smaller than @xmath120 .
results are plotted for the first three subsets @xmath121 of the map control parameter , from top to bottom . in all cases ,
the slope of the solid lines is approximately @xmath122 .
the dashed lines in each graph represent the corresponding coarse - grained window distribution functions . in the inset of fig .
( c ) the measure of any single period five - window in the subset @xmath123 is shown with respect to an integer label that accounts for the ordering according to the map control parameter.,title="fig:",scaledwidth=50.0% ] + , where @xmath119 is the total measure of windows that are smaller than @xmath120 .
results are plotted for the first three subsets @xmath121 of the map control parameter , from top to bottom . in all cases ,
the slope of the solid lines is approximately @xmath122 .
the dashed lines in each graph represent the corresponding coarse - grained window distribution functions . in the inset of fig .
( c ) the measure of any single period five - window in the subset @xmath123 is shown with respect to an integer label that accounts for the ordering according to the map control parameter.,title="fig:",scaledwidth=50.0% ] + , where @xmath119 is the total measure of windows that are smaller than @xmath120 .
results are plotted for the first three subsets @xmath121 of the map control parameter , from top to bottom . in all cases ,
the slope of the solid lines is approximately @xmath122 .
the dashed lines in each graph represent the corresponding coarse - grained window distribution functions . in the inset of fig .
( c ) the measure of any single period five - window in the subset @xmath123 is shown with respect to an integer label that accounts for the ordering according to the map control parameter.,title="fig:",scaledwidth=50.0% ] + in order to analyze the structure of the regions of anomalous diffusion in the parameter space , we sum up the number of period six - windows as a function of the parameter , that is , the total number is increased by one for any parameter value at which a new period six - window appears .
this sum forms a devil s staircase like structure in parameter space indicating an underlying cantor set like distribution for the corresponding anomalous diffusive region , see fig .
since the lebesque measure of periodic windows is positive , this set must be a fat fractal @xcite .
its self - similar structure can quantitatively be assessed by computing the so - called fatness exponent . following @xcite , let @xmath124 be the total measure of all periodic windows whose width is greater than or equal to @xmath120 . define the coarse - grained measure as @xmath125 , where @xmath126 is the total measure of a box related to the control parameter .
for quadratic maps on the interval , it was conjectured and confirmed numerically that @xmath119 asymptotically scales as a power law in the limit of @xmath127 , @xmath128 where @xmath129 is the measure of chaotic parameters .
@xmath130 was called the fatness exponent . for quadratic maps
it was found to be @xmath131 .
since our map belongs to the same universality class as considered in farmer s case , namely the map has a single quadratic maximum , one may expect that @xmath130 will have the same value .
for the climbing sine map a double - logarithmic plot of @xmath132 is shown in fig .
[ fig8 ] for the first three boxes , @xmath108 . in all cases
the bold lines have the slope @xmath122 with errors of @xmath133 , respectively , which seems to be in agreement with farmer s conjecture about the universality of @xmath130 .
however , apart from this coarse linear behavior one can see an interesting oscillatory behavior in @xmath134 with respect to @xmath120 .
this fine structure can be explained with respect to the histogram distribution functions @xmath135 of the window sizes @xmath136 , which are plotted in fig .
[ fig8 ] in form of dashed lines .
somewhat surprisingly , the periodic windows are not distributed uniformly or smoothly with respect to their size but form certain peaks , in which preferably windows of certain periods are grouped together . this non - uniformity is clearly reflected in the oscillations of @xmath137 .
moreover , we find that the size distributions of periodic windows have a fine structure that appears to resemble a fractal function . some evidence for this property
is given in the inset of fig .
[ fig8 ] ( c ) , which shows the size of every window of period five in the subset @xmath123 as a function of its appearance with respect to the map control parameter @xmath26 , i.e. , not the parameter itself is plotted but just an integer running index @xmath138 is given instead .
particularly the height of the peaks is important , and one can clearly see a complicated hierarchy of different peaks which are reminiscent of the fine structure in the corresponding distribution function shown in fig .
[ fig8 ] ( c ) .
in this paper we have performed a detailed analysis of the parameter dependence of the diffusion coefficient in a nonhyperbolic dynamical system .
the climbing sine map has been chosen as a paradigmatic example of such a system .
we have shown that , on a coarse scale , there are certain analogies between the parameter - dependent diffusion coefficient of this map and the ones in simple hyperbolic piecewise linear maps , such as the existence of an oscillatory structure , and the existence of asymptotic functional forms as derived from simple random walk models .
however , in contrast to hyperbolic maps showing normal diffusion only , in the nonhyperbolic climbing sine map fractal structures appear for both normal and anomalous diffusive regions of the diffusion coefficient .
an understanding of the origin of these fractal structures was given in terms of dynamical correlations starting from the green - kubo formula for diffusion .
we furthermore related these irregularities in the diffusion coefficient more microscopically to different characteristics in corresponding fractal generalized takagi functions .
for this purpose we derived a new functional recursion relation that defines these fractal forms and generalizes ordinary de rham - type equations .
our analysis was completed by extensive numerical studies of the periodic windows of the climbing sine map showing that both the periodic and the chaotic parameter regions have positive measures in the parameter space
. however , these measures are themselves parameter - dependent , and by increasing the parameter we found that the chaotic regions occupy larger and larger measures .
we finally provided evidence that these different sets form fat fractals on the parameter axis . in conclusion
, we wish to remark that the climbing sine map is of the same functional form as the respective nonlinear equation in the two - dimensional standard map , which is considered to be a standard model for many physical hamiltonian dynamical systems .
indeed , both models are motivated by the driven nonlinear pendulum , both are strongly nonhyperbolic , and though the standard map is area - preserving it too exhibits a highly irregular parameter - dependent diffusion coefficient .
understanding the origin of these irregularities was the subject of intensive research @xcite , however , so far the complexity of this system did not enable to reveal its possibly fractal nature . a suitably adapted version of our approach to nonhyperbolic diffusive dynamics
as presented in this paper may enable to make some progress in this direction .
another interesting problem is to possibly further exploit the concept of conjugacy between nonlinear and piecewise linear maps , as explained in sec.iii , in order to exactly calculate diffusion coefficients for nonlinear maps .
a very promising approach in this direction was presented in ref.@xcite .
based on these techniques we are planning to perform a spectral analysis of the frobenius - perron operator governing the probability density of the diffusive climbing sine map . combining such an analysis with the takagi function approach outlined here may lead to a general theory of nonhyperbolic transport .
it would furthermore be important to check out the applicability of periodic orbit theory for computing the parameter - dependent diffusion coefficient of the climbing sine map , which may provide an alternative method @xcite .
another promising direction of future research concerns establishing crosslinks between our work and the realm of strange kinetics and stochastic modeling as described in refs.@xcite , e.g. , by trying to apply continuous time random walk techniques to more complicated chaotic models exhibiting fractal diffusion coefficients such as the climbing sine map .
we finally emphasize the importance to look for possibly fractal transport coefficients in experiments .
a very promising candidate appears to be the phase dynamics in squid s , which was very recently analyzed theoretically @xcite and studied experimentally @xcite .
in this appendix we give a proof of the statement of grossmann and thomae @xcite that two diffusive maps which are conjugate to each other have the same diffusion coefficient .
two diffusive maps @xmath139 and @xmath140 are called conjugate @xcite if there exists a map @xmath141 such that @xmath142 .
let us assume in the following that the conjugation function @xmath143 is sufficiently smooth .
let the invariant densities of the corresponding reduced ( @xmath144 ) maps be @xmath145 for @xmath146 and @xmath147 for @xmath148 ; then it is , according to conservation of probability , @xmath149 .
the diffusion coefficients of the maps @xmath150 and @xmath151 we denote by @xmath152 and @xmath153 , respectively . without loss of generality let us furthermore assume that the maxima of both maps are in the interval @xmath154 $ ] .
we now start with the green - kubo formula written in _ correlated random walk _
terms as @xmath155 ^ 2 \tilde{\rho } ( x ) dx + \int\limits^1_0 \left [ f(x ) \right ] \cdot b(x ) \ ; \tilde{\rho } ( x ) dx , \label{ap1}\ ] ] where @xmath156 + ... + \left [ f\left ( \ { f \left ( \{ ... \left ( \ { f(x ) \ } \right) ... \ } \right ) \ } \right)\right ] + ... \ ; , \label{ap2}\ ] ] or shortly @xmath157 where @xmath158 ^ 2 \ ; \tilde{\rho } ( x ) \ ; dx , \label{ap3}\ ] ] @xmath159 \ ; \left[f\{f(x)\}\right ] \ ; \tilde{\rho } ( x ) \ ; dx , \label{ap4}\ ] ] and so on . focusing on the first term , one can rewrite this expression using the symmetry of the map to @xmath158 ^ 2 \ ; \tilde{\rho}(x ) \ ; dx = \int\limits^{x2}_{x1 } \tilde{\rho}(x ) \ ; dx , \label{ap5}\ ] ] where @xmath160 defines an escape region .
for the conjugate map @xmath151 the respective term reads @xmath161 ^ 2 \ ; \rho(y ) \ ; dy = \int\limits^{y2}_{y1 } \rho(y ) \ ; dy , \label{ap6}\ ] ] where @xmath162 is the corresponding escape region for @xmath163 .
note that the escape regions @xmath164 and @xmath165 are not the same , however , it is straightforward to show that @xmath166 , that is , the topology of both maps is conserved such that the two escape regions are mapped onto each other under conjugacy .
taking into account the conservation of probability mentioned before one immediately gets @xmath167 all other terms @xmath168 and @xmath169 have the form @xmath170 where @xmath171 is a constant , @xmath172 is the respective escape region and @xmath173 is the corresponding invariant measure .
thus , the same argument can be applied to show that @xmath174 , ( @xmath175 ) . combining all results we arrive at
we start with the recursion relation for the jump velocity function eq.([rec ] ) , @xmath177 by recalling the definition of the generalized takagi function eq.([takagi ] ) , @xmath178 or differently @xmath179 .
piecewise invertible branches are labeled by the integer numbers @xmath180.,title="fig:",scaledwidth=60.0% ] + we have to integrate eq .
( [ appb_1 ] ) , @xmath181 by using of eq .
( [ appb_2 ] ) we get @xmath182 , \label{appb_4}\ ] ] where @xmath183 . \label{appb_5}\ ] ] without loss of generality let us assume that the maximum of the map is in the interval @xmath154 $ ] . depending on @xmath11 the integral in eq .
( [ appb_5 ] ) can be decomposed into @xmath184 ; \label{appb_6}\ ] ] @xmath185 ; \end{aligned}\ ] ] @xmath186 ; \ ; ... \ ; i_6(x ) = ... , \ ; x \in ( x_5,x_6];\ ] ] @xmath187 .
\end{aligned}\ ] ] each integral in eq .
( [ appb_6 ] ) now contains only one piecewise invertible branch of the reduced map @xmath188 as shown in fig.[app_l ] . here
, the piecewise invertible branches of the reduced map are labeled by integers , and the corresponding branches of the inverse function @xmath189 have the same indices @xmath180 .
since all integrals in eq.([appb_6 ] ) have the same form ( only the inverse parts of the reduced map are different ) , we restrict ourselves to the integral @xmath190 making the change of variables @xmath191 and using the definition of the generalized takagi function eq .
( [ appb_2_a ] ) we get @xmath192 using integration by parts we arrive at @xmath193 now recall that according to eq .
( [ appb_2 ] ) it is @xmath194 , where the @xmath195 define the boundaries of the piecewise invertible parts of @xmath84 , see fig .
[ app_l ] , and that @xmath196 .
thus , by formally defining the inverse function @xmath197 as consisting of all branches @xmath198 , we can finally write eq .
( [ appb_4 ] ) in the form @xmath199 with the integral term @xmath200
the parameter values @xmath201 which correspond to the tangent bifurcations of the @xmath100-periodic windows were found by solving the two coupled transcendental equations @xmath202 where @xmath203 denotes the @xmath100-times iterated reduced map .
this corresponds to the situation where @xmath203 touches the bisector .
somewhat after a tangent bifurcation one will unavoidably find a situation where a critical point @xmath204 , which corresponds to an extremum of @xmath203 , crosses the bisector .
when this critical point is exactly located on the diagonal , the reduced map or its higher iterations have a fixed point and there exists a specific markov partition on the interval @xcite .
the periodic orbit generated by the corresponding parameter value @xmath205 is superstable , @xmath206 by further increasing the parameter value up to @xmath207 a crisis takes place , and this again corresponds to the existence of a certain markov partition .
based on this scenario , the full numerical procedure which was used for calculating the measure of periodic windows is as follows : the values of @xmath205 corresponding to superstable solutions were first calculated by a combination of bisection with the newton method .
the parameters for the tangent bifurcations could then usually be found by the modified two - dimensional newton method @xcite .
however , the highly discontinuous nature of @xmath203 made its implementation very inefficient .
instead , starting in the vicinity of each @xmath205 we again combined the one - dimensional newton and bisection methods .
this ensured that no windows were missed .
finally , the parameter values corresponding to crisis points @xmath207 , which are also defined by markov partitions , can be found by solving respective equations that are formally analogous to eq .
( [ sstable ] ) . | the nonlinear climbing sine map is a nonhyperbolic dynamical system exhibiting both normal and anomalous diffusion under variation of a control parameter .
we show that on a suitable coarse scale this map generates an oscillating parameter - dependent diffusion coefficient , similarly to hyperbolic maps , whose asymptotic functional form can be understood in terms of simple random walk approximations . on finer scales
we find fractal hierarchies of normal and anomalous diffusive regions as functions of the control parameter . by using a green - kubo formula for diffusion the origin of these different regions
is systematically traced back to strong dynamical correlations .
starting from the equations of motion of the map these correlations are formulated in terms of fractal generalized takagi functions obeying generalized de rham - type functional recursion relations .
we finally analyze the measure of the normal and anomalous diffusive regions in the parameter space showing that in both cases it is positive , and that for normal diffusion it increases by increasing the parameter value . deterministic diffusion , nonhyperbolic maps , fractal diffusion coefficient , anomalous diffusion , periodic windows 05.45.df , 05.10.-a , 05.45.ac , 05.60.-k |
quantum entanglement can be affected by a range of noise sources , both quantum and classical in nature .
noise can give rise to a loss of entanglement over a broad range of quantum states .
recently , the relationship between dephasing - decoherence and bipartite entanglement reduction under basis - specific classical noise has been studied in two - qubit systems @xcite and in pairs of qutrit systems @xcite ; this relationship has been studied in two - qubit systems under quantum dissipative vacuum noise as well @xcite .
for example , it has been noted that initially entangled two - qubit systems can suffer `` entanglement sudden death , '' in which a two - qubit system may suddenly lose entanglement in a finite time , even though each qubit itself maintains its quantum - coherence@xcite . here , we present the first general analysis of the disentanglement and dephasing of two qutrits as realized in atoms with `` v''-type energy - level configuration under a classical pure dephasing noise model , at the multi - local and collective level .
we compare the timescales of disentanglement and dephasing - decoherence , the latter timescale specifically in the basis on which this noise acts as in previous studies . in sec .
2 , we introduce our dephasing model . in sec . 3
, we examine the effects of the multi - local and collective dephasing noise on a general state . finally , in sec . 4 , we examine two specific classes of states and compare their disentanglement and decoherence rates explicitly .
the bipartite - entanglement decay rate , is found to be greater than or equal to the dephasing - decoherence rates , characterized by the decay of off - diagonal elements in the corresponding density matrix describing the full system or the reduced density matrix describing either qutrit , when dephasing occurs .
this result is the most general yet to be obtained in the comparative study of decoherence and disentanglement in two - qutrit systems .
our model consists of two - three level systems subjected to the time - dependent hamiltonian @xmath0,\end{aligned}\ ] ] where @xmath1 is the dephasing operator for three - level systems with subscripts denoting qutrits a , b , or both , @xmath2 , the time dependent noise terms @xmath3 refer to statistically independent classical markov processes satisfying @xmath4 with @xmath5 ; @xmath6 ; @xmath7 is the ensemble time average , @xmath8 and @xmath9 denote the phase - damping rates associated with @xmath10 ( x = a , b ) and @xmath11 , respectively .
the one - qutrit standard - basis eigenstates are @xmath12 , for example , representing the ground state , first - excited state , and second - excited state of the atom , respectively .
we assume that the states @xmath13 couple to the ground state but not to each other . here , we notate the standard two - qutrit basis states via the obvious one - to - one correspondence @xmath14 , for simplicity .
the time - dependent density matrix for the two - qutrit system is obtained by taking ensemble averages over the three noise fields , @xmath15 , @xmath16 , @xmath17 , that is , @xmath18 where the statistical density operator @xmath19 and the unitary operator @xmath20 associated with @xmath21 are @xmath22\ , $ ] respectively .
it is helpful to consider the dynamical evolution of @xmath23 as a completely positive trace preserving ( cptp ) linear map @xmath24 , that is , a combination of local and collective quantum channels , any of which can be turned off in particular cases , taking an input state @xmath25 to the output state @xmath26 given by the operator sum @xmath27 where @xmath28 are decomposition operators that satisfy the completeness relation @xmath29 in each of the various cases considered here , the internal structure of the @xmath28 in accordance with the hamiltonian ; various terms may or may not nontrivially contribute in a given case .
the most general solution of @xmath26 , assuming that the system is not initially correlated with any of the three environments , is @xmath30 , where the terms describing the interaction with the local magnetic fields @xmath15 and @xmath16 involve the decomposition operators @xmath31 and the terms associated with the global magnetic field @xmath32 involve the operators @xmath33 @xmath34 @xmath35 the time - dependent parameters appearing in the matrices above are given by : @xmath36 where @xmath37 are the phase - relaxation times associated with the the pertinent qubits @xmath38 and @xmath39 , respectively , as introduced in @xcite , with @xmath40 being the rate parameters . from here on , for tractability of notation , time does not explicitly appear as an argument for these quantities but is implied .
dephasing - decoherence rates are characterized by the decay of off - diagonal elements in the corresponding full density matrix describing the system or the reduced density matrix describing either qutrit given via characteristic decay times .
entanglement is bounded from above by the negativity @xmath41 where @xmath42 is the partial transpose of the density matrix with respect to qutrit a and @xmath43 denotes the trace norm @xcite .
in the standard - basis representation of eq . [ 2qutritbasis ] , the generic pure state of the two - qutrit system is @xmath44 a normalized state - vector with @xmath45 and @xmath46 .
our analysis proceeds as follows .
we find the explicit time evolution of the general state subjected to noise from the multi - local dephasing channel @xmath47 and the collective dephasing channel @xmath48 .
the decoherence timescales are then determined , as characterized by the decay of the off - diagonal elements at the level of the full density matrix of two - qutrits , as well as the reduced density matrix of each individual qutrit , for each of the dephasing cases , multi - local and collective .
we then analyze disentanglement timescales , using the monotone of negativity , when there is decoherence for the general state in each of these cases . in order to compare decoherence and disentanglement behavior explicitly
, we then consider the the behavior of two particular subclasses of states , the _ robust _ class and the _ fragile _ class .
first , consider the most general initial two - qutrit pure state , @xmath49 where @xmath50 is the projector corresponding to the generic state - vector argument @xmath51 explicitly given above , under multi - local and collective dephasing noise .
@xmath52 is the time - evolved full density matrix for multi - local dephasing .
our density matrices are labeled from here on as @xmath53 , with x denoting the pertinent qubits ( a , b , ab ) , c is the class examined ( g , f , r ) as described in the next section , and @xmath54 denoting the pertinent dephasing channel ( @xmath55 ) as noted above , for tractability of notation , time does not explicitly appear as an argument for the @xmath56 and @xmath57 . note that one can recover the local dephasing channel @xmath58 for qutrit a(b ) by effectively freezing the time parameter in the exponential of the @xmath57(@xmath56 ) factors of the opposite channel .
the effect of two local dephasing noise sources , one for each qutrit , is simply the combination of independent effects of the individual local dephasing channel just described .
thus , the combined matrix is just the component - wise multiplication of exponential decay factors arising from each of the local dephasing channels . due to their appearance in exponents , the decay rates add .
let us designate the different timescales as @xmath59 , representing the decoherence and disentanglement timescales , respectively ; @xmath60 denotes the number of qutrits affected ( @xmath61 ) , @xmath54 denotes the pertinent dephasing channel ( @xmath62 ) , @xmath63 is used as an index further to discriminate the timescales , and c denotes the class examined ( g , f , r ) .
the differing timescales of reduction of off - diagonal elements consist of various combinations of @xmath56 and @xmath57 . here
, we have the four timescales of decay as @xmath64 the reduced density matrices of the individual qutrit subsystems are @xmath65 again , one finds differing dephasing decoherence timescales . in particular , the off - diagonal elements decay in two different timescales before the density matrix becomes fully diagonal in the basis under consideration , @xmath66 the negativity , which bounds the ( non - bound ) entanglement , @xmath67 & = & \frac{1}{2}\bigg(-1 + \sqrt { \left [ \left|\bar{a}_{1}\right|^{2 } + ( \left|\bar{a}_{4}\right|^{2 } + \left|\bar{a}_{7}\right|^{2})\gamma_{\rm a}^{2 } \right ] \left [ \left|\bar{a}_{1}\right|^{2 } + ( \left|\bar{a}_{2}\right|^{2 } + \left|\bar{a}_{3}\right|^{2})\gamma_{\rm b}^{2 } \right ] } \nonumber \\ & + & \sqrt { \left [ \left|\bar{a}_{7}\right|^{2 } + \left|\bar{a}_{1}\right|^{2}\gamma_{\rm a}^{2 } + \left|\bar{a}_{4}\right|^{2}\gamma_{\rm a}^{4 } \right ] \left [ \left|\bar{a}_{7}\right|^{2 } + ( \left|\bar{a}_{8}\right|^{2 } + \left|\bar{a}_{9}\right|^{2})\gamma_{\rm b}^{2 } \right ] } \nonumber \\ & + & \sqrt { \left [ \left|\bar{a}_{4}\right|^{2 } + \left|\bar{a}_{1}\right|^{2}\gamma_{\rm a}^{2 } + \left|\bar{a}_{7}\right|^{2}\gamma_{\rm a}^{4 } \right ] \left [ \left|\bar{a}_{4}\right|^{2 } + ( \left|\bar{a}_{5}\right|^{2 } + \left|\bar{a}_{6}\right|^{2})\gamma_{\rm b}^{2 } \right ] } \nonumber \\ & + & \sqrt { \left [ \left|\bar{a}_{3}\right|^{2 } + ( \left|\bar{a}_{6}\right|^{2 } + \left|\bar{a}_{9}\right|^{2})\gamma_{\rm a}^{2 } \right ] \left [ \left|\bar{a}_{3}\right|^{2 } + \left|\bar{a}_{1}\right|^{2}\gamma_{\rm b}^{2 } + \left|\bar{a}_{2}\right|^{2}\gamma_{\rm b}^{4 } \right ] } \nonumber \\ & + & \sqrt { \left [ \left|\bar{a}_{6}\right|^{2 } + \left|\bar{a}_{3}\right|^{2}\gamma_{\rm a}^{2 } + \left|\bar{a}_{9}\right|^{2}\gamma_{\rm a}^{4 } \right ] \left [ \left|\bar{a}_{6}\right|^{2 } + \left|\bar{a}_{4}\right|^{2}\gamma_{\rm b}^{2 } + \left|\bar{a}_{5}\right|^{2}\gamma_{\rm b}^{4 } \right ] } \nonumber \\ & + & \sqrt { \left [ \left|\bar{a}_{9}\right|^{2 } + \left|\bar{a}_{3}\right|^{2}\gamma_{\rm a}^{2 } + \left|\bar{a}_{6}\right|^{2}\gamma_{\rm a}^{4 } \right ] \left [ \left|\bar{a}_{9}\right|^{2 } + \left|\bar{a}_{7}\right|^{2}\gamma_{\rm b}^{2 } + \left|\bar{a}_{8}\right|^{2}\gamma_{\rm b}^{4 } \right ] } \nonumber \\ & + & \sqrt { \left [ \left|\bar{a}_{2}\right|^{2 } + ( \left|\bar{a}_{5}\right|^{2 } + \left|\bar{a}_{8}\right|^{2})\gamma_{\rm a}^{2 } \right ] \left [ \left|\bar{a}_{2}\right|^{2 } + \left|\bar{a}_{1}\right|^{2}\gamma_{\rm b}^{2 } + \left|\bar{a}_{3}\right|^{2}\gamma_{\rm b}^{4 } \right ] } \nonumber \\ & + & \sqrt { \left [ \left|\bar{a}_{8}\right|^{2 } + \left|\bar{a}_{2}\right|^{2}\gamma_{\rm a}^{2 } + \left|\bar{a}_{5}\right|^{2}\gamma_{\rm a}^{4 } \right ] \left [ \left|\bar{a}_{8}\right|^{2 } + \left|\bar{a}_{7}\right|^{2}\gamma_{\rm b}^{2 } + \left|\bar{a}_{9}\right|^{2}\gamma_{\rm b}^{4 } \right ] } \nonumber \\ & + & \sqrt { \left [ \left|\bar{a}_{5}\right|^{2 } + \left|\bar{a}_{2}\right|^{2}\gamma_{\rm a}^{2 } + \left|\bar{a}_{8}\right|^{2}\gamma_{\rm a}^{4 } \right ] \left [ \left|\bar{a}_{5}\right|^{2 } + \left|\bar{a}_{4}\right|^{2}\gamma_{\rm b}^{2 } + \left|\bar{a}_{6}\right|^{2}\gamma_{\rm b}^{4 } \right]}\ \bigg ) \ , \ ] ] is rather unwieldy .
it is therefore illuminating to characterize its behavior for each of the dephasing cases : no dephasing , one - qutrit local dephasing , and two - qutrit multi - local dephasing . in the expression for negativity
, we can isolate the behavior of one local dephasing channel with respect to the other by setting the decay factors corresponding to the other channel to one , effectively freezing that factor to time zero .
\(1 ) in cases where there is no dephasing , any decay factors multiplying probability amplitudes do not fall off exponentially but instead their value remains at unity .
@xmath68 , which corresponds to a maximally entangled state for all time .
thus the ( trivial ) dephasing and decoherence rates are equal .
\(2 ) in the case of a single local dephasing channel , either channel @xmath69 on qutrit a only or channel @xmath70 on qutrit b only , the negativity @xmath71 in the large - time limit .
\(3 ) finally , in the case in which there is dephasing on both local channels @xmath69 and @xmath70 , @xmath72 in the large - time limit .
one then notes that in all these cases decoherence never proceeds faster than disentanglement , as we previously found to be the case for qubit pairs @xcite .
in later sections , when examining specific classes of states , we compare decoherence and disentanglement more explicitly .
when subjected to collective dephasing noise , the time - evolved density matrix of the two - qutrit system in general case is given by @xmath73 note the form of the matrix , in which there are regions unaffected by dephasing , with decoherence effects along the edges .
we show in the next section that this gives rise to a class of states that can exist in these decoherence free subspaces , where entanglement is also be explicitly characterized .
we see in the above general matrix that there exist two timescales of off - diagonal element decay , @xmath74 the reduced density matrices are given by the following matrices , @xmath75 note that the single - qutrit reduced density matrices do not become entirely diagonal .
the reason , as we stated above , is the existence of decoherence free subspaces . in this case , there exists only one timescale in which a subset of off - diagonal elements decay , @xmath76 the expression for negativity for the general class under the collective dephasing channel , is similar in character to the negativity found in the previous subsection for the multi - local dephasing channel given by eq .
5 . therefore ,
for clarity , we defer our discussion of the behavior of the negativity in the case of the collective dephasing channel until after treating specific states to the next section .
in order better to distinguish decoherence and disentanglement behavior , let us now turn our attention to two specific classes of states . in the standard - basis representation , the generic pure state of a two - qutrit system is @xmath77 , a normalized state vector with @xmath78 .
the generic class of two - qutrit pure states represented by @xmath79 contains two subclasses of interest , distinguished by their coherence behavior in large timescales under collective dephasing noise , that is , dephasing in which each qutrit interacts with the same collective noise @xmath80 .
one class is seen to be _ fragile _ , whereas the other is _ robust_. + ( i ) the _ fragile _ class @xmath81 , in which @xmath82 may be non - zero and all other terms @xmath83 , has the forms @xmath84 ( ii ) the _ robust _ class @xmath85 , in which all @xmath86 listed may be non - zero and @xmath87 , has the forms @xmath88 the specific forms of the two classes above constitute bell - like states in the hilbert space of our two - qutrit system , similarly to the qubit - pair - state classification given in @xcite .
the initial density matrix representing the two - qutrit fragile class is given by @xmath89 where the dots denote the remainder of the density matrix , which is filled out by zero entries .
let us turn our attention to the behavior of this state when subjected to the multi - local dephasing @xmath47 acting on qutrits a and b. recall that the multi - local dephasing channel captures the features of each of the local dephasing channels fully , so for this analysis , we consider the multi - local dephasing channel only , which provides information about each of the individual dephasing channels if one simply turns off one of the channels .
the density matrix subjected to noise is @xmath90 which has two different decoherence times of @xmath91 by comparison , the disentanglement timescales deriving from the negativity , @xmath92 are @xmath93 .
the reduced density matrices of qutrit a and b are @xmath94 we see that they are both initially fully mixed , so that no further decoherence is possible . the two - qutrit density matrix , as given by @xmath96 decays according to a single timescale , @xmath97 .
the disentanglement , characterized via the negativity , @xmath98 disentanglement proceeds at only a single timescale , @xmath99 .
the single qutrit matrices are always @xmath100 again , we see that the single qutrit states are both fully mixed , so that no further decoherence is possible . comparing timescales
, we see that decoherence never proceeds faster than disentanglement : @xmath101 the two - qutrit density matrix for the robust class , under the multi - local dephasing channel , is given by @xmath102 the rows and columns of zeros in this matrix are exactly the rows and columns that are affected by the collective dephasing channel , giving rise to the robust character of this density matrix .
for analysis in this section , we consider , without loss of generality , the first form of this class @xmath103 , which allows the decoherence and disentanglement timescales to be seen clearly .
if we were to include the entire subclass , we would have the same behavior as our forthcoming analysis .
the time - evolved density matrix for the multi - local dephasing channel acting on @xmath104 is given by @xmath105 we see that the off - diagonal elements decay according to the timescale @xmath106 .
the disentanglement , characterized via the negativity , is given by , @xmath107 .
disentanglement proceeds at the same single timescale , @xmath108 .
the reduced density matrices are given by @xmath109 always fully decohered as in the case of the fragile states .
again , we see that decoherence never proceeds faster than disentanglement , so that clearly the inequality @xmath110 is satisfied . here , @xmath111 as defined above in eq .
[ robusttimezero ] . and
the state neither decoheres nor disentangles : the off - diagonal elements remain unreduced and the negativity remains at the maximum value of 1 for all times .
these states are robust against collective dephasing noise .
we have shown for this model of two - qutrit systems under the chosen basis - specific dephasing noise , whether local , multi - local or collective in nature , that disentanglement proceeds at least as fast as decoherence .
this result accords with previous studies of similar nature for qubit systems .
because the case of two - dimensions in quantum mechanics is often a special one in quantum mechanics , for example , in the cases of the foundational theorems of gleason and kochen and specker , it is always important to examine cases beyond that of qubits for the possibility of different behavior from that exhibited in the special case of qubits .
our results suggest that the relation between basis - dependent dephasing decoherence and disentanglement found here can be expected to hold for pairs of initially entangled qu-@xmath112-it systems under analogous dephasing noise for general values of @xmath112 .
it remains an open question whether or not this phenomenon can be generalized to an arbitrary _ multi - partite _ quantum systems , each of arbitrary dimension , however , not least of all because good entanglement measures for mixed states of such systems is lacking . | we relate disentanglement and decoherence rates in a pair of three - level atoms subjected to multi - local and collective pure dephasing noise acting in a preferred basis . the bipartite entanglement decay rate , as bounded from above by the negativity ,
is found to be greater than or equal to the dephasing - decoherence rates characterized by the decay of off - diagonal elements in the corresponding full density matrix describing the system or the reduced density matrix describing either qutrit , extending previous results for qubit pairs subject to such noise . |
polymers are repetitive chains of elementary blocks called monomers ( or monomer units ) .
copolymers are inhomogeneous polymers , in the sense that each monomer unit carries a charge , and the charge is distributed along the chain in a _ disordered _ way .
it is well known that when the medium surrounding the copolymer is made of two solvents , separated for example by an interface , and the solvents interact with the monomers according to the value of the charge , the typical behavior of the copolymer may differ substantially from the case in which the medium is homogeneous . on copolymers
there is an extremely extended literature , given above all their practical relevance , see for example @xcite and @xcite and references therein .
moreover , for a realistic model of the interface , one should consider the possibility of the presence of impurities or fluctuations in the interface layer , as in @xcite . as an extreme , but very important example , one could consider also the case in which the interactions at the interface are essentially the only relevant ones , see @xcite . in order to be more concrete ,
let us introduce a specific model , which is just a particular example of the general class we consider .
it is based on the process @xmath4 , a simple random walk on @xmath5 , with @xmath6 and @xmath7 , @xmath8 , a sequence of iid random variables with @xmath9 .
the process @xmath10 has to be interpreted as a directed polymer in @xmath11dimension , and @xmath12 as its law in absence of any interaction with the environment ( _ free polymer _ ) .
the polymer
environment interaction depends on the _ charges _ @xmath13 and on four real parameters @xmath14 : without loss of generality we will assume @xmath15 , @xmath16 and @xmath17 to be non negative . let us set @xmath18 and let us introduce the family of boltzmann measures indexed by @xmath19 @xmath20 with the convention that @xmath21 for any @xmath22 such that @xmath23 .
the superscript @xmath24 will be often omitted .
the model is completely defined once @xmath25 is given : being interested in the disordered case , we choose for instance @xmath25 an realization of an iid family , of law @xmath26 , of symmetric variables taking value @xmath27 .
we invite the reader to look at figure [ fig : modello ] in order to get an intuitive idea on this model .
= 7 cm [ c][l]@xmath22 [ c][l]@xmath28 [ c][l]@xmath29 [ c][l]@xmath30 [ c][l]@xmath31 [ c][l]@xmath32 [ c][l]@xmath33 [ c][l]@xmath34 [ c][l]@xmath35 [ c][l]@xmath36 [ c][l]*water * [ c][l]*oil * [ c][l]@xmath37 [ c][l]@xmath38 [ c][l]@xmath39 [ c][l]@xmath40 [ c][l]@xmath41 an important observation on the model we have introduced is that its hamiltonian may be easily rewritten in terms of the sequence @xmath42 , defined by setting @xmath43 and , for @xmath44 , @xmath45 ( @xmath46 with probability one for every @xmath47 since @xmath10 is recurrent ) and in terms of the sign of the excursions , that , conditionally on @xmath48 , are just an independent sequence of iid symmetric variables taking values @xmath27 .
of course the ( strong ) markov property of @xmath10 immediately yields also that @xmath49 is an iid sequence .
we are now going to introduce the general class of models that we consider .
these models are based on a real valued free process @xmath50 , with law @xmath12 that satisfies the following properties : 1 .
the sequence @xmath51 of successive returns to @xmath29 is an infinite sequence with @xmath52 , so @xmath36 is a process starting from @xmath29 and for which @xmath29 is a recurrent state .
moreover @xmath48 is a renewal sequence , that is @xmath53 is a sequence of iid random variables , and we set @xmath54 .
2 . for some @xmath55 and for some function @xmath56 which is slowly varying at infinity ( see below for the definition and properties of slowly varying functions ) @xmath57 in particular @xmath58 for every @xmath59 .
3 . for @xmath44
such that @xmath60 we set @xmath61
. then conditionally on @xmath48 , @xmath62 is an iid sequence of symmetric random variables taking values @xmath27 .
conventionally we complete the sequence @xmath63 ( i.e. , we choose the @xmath64 for @xmath47 such that @xmath65 ) by tossing independent fair coins and , always by convention , we stipulate that @xmath66 .
of course , conditionally on @xmath48 , @xmath63 is just independent fair coin tossing .
4 . conditionally on @xmath48 , @xmath67 is an independent sequence of random vectors . moreover the law of @xmath68 ,
conditionally on @xmath48 , depends on @xmath48 only via the value of @xmath69 .
note that this property implies that , conditionally on @xmath70 , the process @xmath71 has the same distribution as the original process @xmath72 . with some abuse of language
, we will call this property the _ renewal property of @xmath36_. the free process @xmath36 is put in interaction with an environment via the _ charges _ @xmath13 .
the definition of @xmath73 is still as in , provided one suppresses the superscript rw in the left
hand side .
the process @xmath36 constructed in section [ sec : premodel ] starting from @xmath10 corresponds to the case of @xmath74 and @xmath75 .
we recall that a function @xmath76 is slowly varying ( at infinity ) if @xmath56 is a measurable function from @xmath77 to @xmath78 such that @xmath79 for every @xmath80 .
one of the properties of slowly varying functions is that both @xmath81 and @xmath82 are @xmath83 for every @xmath84 .
an example of slowly varying function is @xmath85 , for @xmath86 , but also @xmath87 , for @xmath88 , as well as any positive function for which @xmath89 .
a complete treatment of slowly varying functions is found in @xcite , but these functions are needed for us to in order to work in a reasonably general and well defined set up : we will use no fine property of slowly varying function since in most of the cases rough bounds on @xmath90 will suffice .
[ rem : s ] in short , one may think of building @xmath36 by first assigning the return times to zero according to a renewal process .
the excursions are then _ glued _ to the renewal points , but essentially the only relevant aspect of the excursions for us is the sign , that is chosen by repeated tossing of a fair coin . note again that the energy of the model depends only on @xmath48 and on @xmath91 , and not on the details of the excursions in the upper or lower half plane .
the last property in the list above , the renewal property of @xmath36 , makes a bit more precise what the excursions of the process really are : this property is very useful to have a nice pictorial vision of the process , but it is rather inessential for us .
we will use it only in stating theorem [ th : correlazioni ] since this way it turns out to be somewhat nicer and more intuitive , but the essence of our analysis lies in @xmath48 .
note also that if @xmath92 the energy does not depend on @xmath91 . in this case
we are dealing with a pure pinning model and insisting on @xmath36 taking values in @xmath93 is a useless limitation .
[ rem : s1 ] we have chosen @xmath29 to be a recurrent state of @xmath36 because it simplifies a little the notations , but everything carries over to the case in which @xmath29 is transient .
the sequence @xmath94 is chosen as a typical realization of an iid sequence of random variables , still denoted by @xmath95 .
we make the very same assumptions on @xmath96 and , in addition , @xmath97 and @xmath96 are independent .
the law of @xmath25 will be denoted by @xmath26 , and the corresponding expectation by @xmath98 .
a further assumption on @xmath25 is the following _ concentration inequality _ : there exists a positive constant @xmath99 such that for every @xmath28 , for every lipschitz and convex function @xmath100 with @xmath101 and for @xmath102 @xmath103\big\vert \ge t\right ) \ , \le \ , c \exp \left ( -\frac{t^2}{c \vert g \vert ^2 _ { \rm lip}}\right),\ ] ] where @xmath104 is the lipschitz constant of @xmath105 with respect to the euclidean distance on @xmath106 . of course such
an inequality implies that @xmath107 and @xmath108 are exponentially integrable : without loss of generality we assume @xmath107 and @xmath108 to be centered and of unit variance . in practice , it is sufficient to check that the concentration inequality holds for @xmath94 and for @xmath96 separately .
in fact , suppose that @xmath109\big\vert \ge t\right ) \ , \le \
, c \exp \left ( -\frac{t^2}{c \vert g \vert ^2 _ { \rm lip}}\right),\ ] ] for every @xmath110 and @xmath111 , and similarly for @xmath96
. then , @xmath112\big\vert \ge u\right)\le \\ { { \ensuremath{\mathbb p } } } \left(\big\vert { { \ensuremath{\mathbb e } } } [ g({{\underline{{\omega}}}})\vert { \omega}]-{{\ensuremath{\mathbb e } } } [ g({{\underline{{\omega}}}})]\big\vert \ge \frac u2\right ) + { { \ensuremath{\mathbb p } } } \left(\big\vert g({{\underline{{\omega}}}})-{{\ensuremath{\mathbb e } } } [ g({{\underline{{\omega}}}})\vert { \omega}]\big\vert \ge \frac u2\right).\end{gathered}\ ] ] using twice , once for @xmath113 $ ] and once for @xmath114 , noting that @xmath115 uniformly in @xmath94 , one obtains immediately .
the concentration inequality is known to hold with a certain generality : its validity for the gaussian case and for the case of bounded random variables is by now a classical result ( @xcite ) .
while of course such an inequality requires @xmath116 < \infty$ ] for some @xmath117 , a complete characterization of the distributions for which holds is , to our knowledge , still lacking , but among these distributions there are , for example , the cases in which the laws of @xmath107 and @xmath118 satisfy the log sobolev inequality , see @xcite , @xcite and @xcite , therefore , in particular , whenever @xmath107 and @xmath118 are continuous variables with positive densities of the form @xmath119 , with @xmath120 and @xmath121 when restricted to @xmath122 , for some @xmath123 .
we have chosen to work assuming concentration because it provides a unified rather general framework in which proofs are at several instances much shorter ( and , possibly , more transparent ) . however , as it will be clear , several results hold under weaker assumptions . on the other hand
we deal only with the polymer pinned at the endpoint , that is , constrained to @xmath124 : we have chosen this case for the sake of conciseness , but we could have decided for example to leave the endpoint free . under the above assumptions on the disorder the _ quenched free energy _ of the system exists ,
namely the limit @xmath125 exists @xmath126almost surely and in the @xmath127 sense .
the existence of this limit can be proven via standard super additivity arguments ( we refer for example to @xcite for details ) .
we stress that the concentration inequality implies immediately that @xmath128 is not random , but such a result may be proven under much weaker assumptions , see e.g. @xcite .
a simple but fundamental observation is that @xmath129 the proof of this is elementary : if we set @xmath130 for @xmath131 and @xmath132 we have @xmath133 \\ = \ , \frac { \lambda}n { \sum_{n=1}^n \left({{\ensuremath{\mathbb e } } } \left[{\omega}_1 \right]+h\right ) } + \frac { { \widetilde}{\lambda}}n \left ( { { \ensuremath{\mathbb e } } } \left[{\widetilde}{\omega}_n \right ] + { \widetilde}h \right ) + \frac 1n \log { { \ensuremath{\mathbf p } } } \left ( { \omega}_n^+\right)\ , \stackrel{n \to \infty}{\longrightarrow}\ , { \lambda}h,\end{gathered}\ ] ] where we have applied the fact that , by , @xmath134 , for @xmath135 .
the observation , above all if viewed in the light of its proof , suggests the definition @xmath136 and the following partition of the parameter space ( or _ phase diagram _ ) : * the localized region : @xmath137 ; * the delocalized region : @xmath138 . along with this definition
we observe that @xmath139 where @xmath140 of course the normalization constant @xmath141,\ ] ] changes , but @xmath142 so that the @xmath126a.s . and @xmath143 asymptotic behavior of @xmath144 are given by @xmath145 .
we will always work with @xmath146 , in order to conform with most of the previous mathematical literature .
for the model we introduced there is a very vast literature , mostly in chemistry , physics and bio physics .
it often goes under the name of _ copolymer with adsorption _ , see e.g. @xcite and references therein , and such a name clearly reflects the superposition of two distinct polymer environment interactions : * the monomer solvent interaction , associated to the charges @xmath94 .
some monomers prefer one solvent and some prefer the other one . since the charges are placed in an inhomogeneous way along the chain , energetically favored trajectories need to stick close to the interface . whether pinning actually takes place or not depends on the interplay between energetic gain and entropic loss associated to localization ( trajectories that stay close to the interface have a much smaller entropy than those which wander away ) .
+ if @xmath147 and @xmath148 only this interaction is present and we will call the model simply _ copolymer_. * the monomer interface interaction , associated to @xmath96 .
this interaction leads to a pinning ( or depinning ) phenomenon with a more direct mechanism : trajectories are energetically favored if they touch the interface _ as often as possible _ at points where @xmath149 , avoiding at the same time the points in which @xmath150 .
also in this case , a non trivial energy / entropy competition is responsible for the localization / delocalization transition .
+ when @xmath92 and @xmath151 we will refer to the model as _ pinning _ model .
the copolymer model has received a lot of attention : we mention in particular @xcite , in which it was first introduced and the replica method was applied in order to investigate the transition .
rigorous work started with @xcite , followed by @xcite : these works deal with the case @xmath152 and @xmath107 symmetric and taking only the values @xmath27 ( _ binary charges _ ) .
it turns out that in such a case there is no transition and the model is localized for every @xmath147 . in general
, one may distinguish between results concerning the free energy and results on pathwise behavior of the polymer .
about the first point , we mention that in @xcite the model with @xmath153 has been considered , still with the choice of binary charges , and the existence of a transition has been established , along with estimates on the critical curve and remarkable limiting properties of the free energy of the model in the limit of weak coupling ( @xmath15 small ) . improved estimates on free energy and critical curve
may be found in @xcite . in the physical literature one can find a number of conjectures , mostly on the free energy behavior , that are far from clarifying the phase diagram . in this respect , it is interesting to mention that recent numerical simulations ( @xcite ) show that the critical line is different from that predicted in the theoretical physics literature , which means that the localization mechanism is still poorly understood .
disordered pinning has been extensively studied in the physical literature , see e.g. @xcite and @xcite ( see also @xcite for more recent references ) , but much less in the mathematical one . however
the model has started attracting attention lately , see @xcite , @xcite and @xcite .
we should stress that there is no agreement in the physical literature on several important issues for disordered pinning .
for instance , it is still unclear whether the critical curve coincides with the so called annealed curve . about the study of path behavior
, there is a basic difference between the localized and the delocalized phase : in the first case , since @xmath154 , the interaction produces an exponential modification of the free polymer measure , and therefore _ large deviation
_ techniques apply very naturally .
the path behavior of the copolymer model in the localized phase has been considered in @xcite , for @xmath152 and binary charges , while in @xcite also the case @xmath155 is taken into account .
in @xcite the focus is on the gibbsian characterization of the infinite volume polymer measure ( in the localized phase ) .
the delocalized phase is more subtle , due to the fact that @xmath156 , and path delocalization estimates involve estimates on _ moderate deviations _ of the free energy .
results on the path behavior in the delocalized phase have been obtained only recently in @xcite , both for the copolymer and for the disordered pinning model .
in the present work , we consider the localized phase of the general model defined in section [ sec : model ] and formula . in addition to giving new results , our approach provides also a setting to reinterpret in a simpler way known results for copolymer and pinning models .
the free energy is everywhere continuous and almost everywhere differentiable , by convexity , so in particular @xmath157 is an open set .
however , one can go much beyond that , as our first theorem shows : [ th : cinf ] @xmath158 is infinitely differentiable in @xmath157 .
an interesting problem is to study the regularity properties of @xmath159 at the boundary between @xmath157 and @xmath160 , where it is non analytic .
this corresponds to investigating the order of the localization / delocalization transition .
recently , an important step in this direction was performed in @xcite and @xcite where it was proved , in particular , that the first derivatives of @xmath159 are continuous on the boundary . in other words ,
the ( de)localization transition is at least of second order .
as it will become clear in section [ sec : reg ] , the smoothness of the free energy in @xmath161 boils down to a property of exponential decay of ( average ) correlations . for this implication
we essentially rely on @xcite , where a similar result has been proven in the context of disordered ising models .
let us therefore state the decay of correlation property .
we say that @xmath162 is a bounded local observable if @xmath162 is a real bounded measurable local function of the path configurations
. in the sequel @xmath163 will denote the _ support of _
@xmath162 , that is the intersection of all the subsets @xmath164 of the form @xmath165 , with @xmath166 , such that @xmath162 is measurable with respect to the @xmath167algebra @xmath168 .
we have [ th : correlazioni ] for every @xmath169 there exist finite constants @xmath170 such that the following holds for every @xmath171 : * ( exponential decay of correlations . ) for every couple of bounded local observables @xmath162 and @xmath172 we have @xmath173 \\ \le\ , c_1 \vert a\vert_\infty\vert b\vert_\infty \exp(-c_2 d(\mathcal s(a),\mathcal s(b ) ) ) \end{gathered}\ ] ] where @xmath174 if @xmath175 . *
( influence of the boundary . ) for every bounded local observable @xmath162 and @xmath176 , such that @xmath177 , we have @xmath178\le c_1 \vert a\vert_\infty \exp(-c_2d(\mathcal s(a),\{k\})).\ ] ] * for every bounded local observable @xmath162 the following limit exists @xmath126almost surely : @xmath179 [ rem : a.s . ] from one may easily extract an almost sure statement .
choose two bounded local observables @xmath162 and @xmath172 and set @xmath180 , i.e. @xmath181 .
take the limit @xmath182 in to obtain @xmath183\le \exp(-c_2 k/4),\ ] ] for @xmath184 sufficiently large .
therefore the fubini tonelli theorem and yield @xmath185\ , < \ , \infty.\ ] ] the series appearing in the left
hand side is therefore @xmath126a.s .
this implies that there exists a random variable @xmath186 , @xmath187 @xmath126a.s .
, such that @xmath188 for every @xmath184 .
we consider now the question of whether or not knowing that @xmath169 does mean that the path of the polymer is really tight to the interface .
even if this question has not been treated for the general model we are considering here , the techniques used in @xcite , @xcite and @xcite , see also @xcite , may be applied directly and one obtains for example that , in the case @xmath189 , for every @xmath169 there exist finite constants @xmath190 such that for every @xmath28 , @xmath191 , and @xmath192 @xmath193 or one can obtain an analogous @xmath126a.s .
result , which is a bit more involved to state @xcite .
the reason for revisiting this type of results , besides generalizing them to our case , is that they are only bounds and we would like to find estimates that are sharp to leading order .
a notable exception is the case of some of the results in ( * ? ? ?
5.3 and th . 6.1 ) where the precise asymptotic size of the largest excursion ( and of the maximum displacement of the chain from the interface ) is obtained .
this result is a bit surprising since it depends on a certain annealed decay exponent .
this exponent turns out to be different from the decay exponent one finds for @xmath126a.s .
estimates , see discussion after our theorem [ th : zu ] .
the argument of the crucial point of the proof of ( * ? ? ?
5.3 ) looks obscure to us and we propose here a different one , based on decay of correlations . of course we present our results in terms of _ excursion lengths_. recall the definition of the return times @xmath48 in section [ sec : model ] . for every @xmath194 let us set @xmath195 , with @xmath196 equal to the value @xmath47 such that @xmath197 .
so @xmath198 is the left margin of the excursion to which @xmath184 belongs and @xmath199 is the length of such an excursion .
two distinct questions can be posed concerning polymer excursions : one may be interested about the typical length of a given excursion or , more globally , about the typical length of the _ longest _ excursion , @xmath200 .
denote by @xmath201 the left shift on @xmath94 , like for @xmath36 : @xmath202 .
about the first problem , we can prove : [ th : as ] take @xmath203 and let @xmath25 be the two sided sequence of iid random variables , @xmath204 , with law @xmath26 .
for every @xmath117 there exist random variables @xmath205 , @xmath206 , such that @xmath207 and @xmath208 for every @xmath28 , every @xmath209 and every @xmath191 : for the first inequality , the lower bound , we require also @xmath210 .
note that , in the definition of @xmath211 , the fact that @xmath25 is a doubly infinite sequence is completely irrelevant , since the polymer measure depends only on @xmath212 .
the introduction of two sided disorder sequences , which might seem a bit unnatural , is needed here to have the almost sure result uniformly in @xmath213 and @xmath214 .
about the maximal excursion , we have : [ th : zu ] the following holds : 1 . the limit @xmath215 \ , : = \ , \mu ( { { \underline{v } } } ) , \ ] ] exists and satisfies the bounds @xmath216 .
moreover , for every @xmath217 there exists @xmath218 such that @xmath219 , for @xmath220 and @xmath221 .
2 . fix @xmath169 . for every @xmath222 we have that @xmath223 and @xmath224 in the localized region , under rather general conditions on the law @xmath26
, we can prove that @xmath225 , see appendix [ sec : mu ] .
notice therefore the gap between and the result in proposition [ th : as ] , so that the largest excursion appears to be achieved in atypical regions . when the law of @xmath34 is symmetric , one can also prove ( see appendix [ sec : mumu ] ) that @xmath226 is equivalently given by @xmath227.\ ] ] this coincides with the expression given in @xcite , where only the case @xmath228 , @xmath152 and @xmath107 taking values in @xmath229 was considered . in this section ,
we investigate finite size corrections to the infinite volume limit of the quenched average of the free energy , and the behavior of its disorder fluctuations . about the first point
, it is quite easy to prove ( see also ( * ? ? ?
* proposition 2.6 ) ) [ th : nfin1 ] there exists @xmath230 such that , for every @xmath24 and @xmath171 , one has @xmath231
this bound is somehow optimal in general , in the sense that it is possible to prove a lower bound of order @xmath232 if @xmath24 is in the _ annealed region _ , i.e. , the sub region of @xmath160 where @xmath233 .
however , in the localized region we can go much farther : [ th : nfin2 ] assume that @xmath169 .
then , there exists @xmath234 such that , for every @xmath171 , one has @xmath235 about fluctuations , in @xcite it was proven that , for @xmath228 , @xmath152 and @xmath107 taking values in @xmath229 , the free energy satisfies in the large volume limit a central limit theorem on the scale @xmath236 .
here , we generalize this result to the entire localized region .
our proof employs basically the same idea as in @xcite ; however , the use of concentration of measure ideas , plus a more direct way to show that the limit variance is not degenerate , allow for remarkable simplifications .
the precise result is the following : [ th : clt ] if @xmath169 the following limit in law holds : @xmath237 with @xmath238 . for compactness
we introduce a notation for the hamiltonian @xmath239 and for @xmath240 we set @xmath241.\ ] ] moreover , we set @xmath242
the proof of theorem [ th : correlazioni ] is based on the following lemma , which is somehow similar in spirit to lemmas 4 and 5 in @xcite .
[ lemma : ritorni ] for every @xmath169 , there exist constants @xmath243 such that , for every @xmath244 , @xmath245 , @xmath246 and @xmath247 , @xmath248 moreover , let @xmath249 be two independent copies of the copolymer , distributed according to the product measure @xmath250 .
then , @xmath251 it will be clear from the proof that the constant @xmath252 in may be chosen smaller , but arbitrarily close to @xmath226 .
the quantitative estimate on the constant @xmath252 in that one can extract from the proof given below is instead substantially worse and certainly not optimal .
_ proof of theorem [ th : correlazioni]_. let @xmath253 and @xmath254 .
we assume that @xmath255 @xmath256 , otherwise holds trivially with @xmath257 . without loss of generality , we take @xmath258 . then , letting @xmath259 be the event @xmath260 one can write @xmath261 since , by a simple symmetry argument based on the renewal property of @xmath36 , one can show that the above average vanishes , if conditioned to the complementary of the event @xmath259 . at this point , using one obtains @xmath262 \le 2 c_1 \vert a\vert_\infty\vert b\vert_\infty e^{-c_2(b_1-a_2)},\end{aligned}\ ] ] which is statement of the theorem . as for we observe that , since we are assuming that @xmath263 , one has the identity @xmath264 where we recall that @xmath265 .
therefore , there exist positive constants @xmath266 and @xmath267 such that @xmath268 \ , = \ , { { \ensuremath{\mathbb e } } } \left [ \left \vert \frac { { { \ensuremath{\mathbf e } } } _ { n , { { \underline{{\omega}}}}}\left ( a\right){{\ensuremath{\mathbf e } } } _ { n , { { \underline{{\omega}}}}}\left ( \delta_k\right)- { { \ensuremath{\mathbf e } } } _ { n , { { \underline{{\omega}}}}}\left ( a\,\delta_k\right ) } { { { \ensuremath{\mathbf e } } } _ { n , { { \underline{{\omega}}}}}\left ( \delta_k\right ) } \right\vert \right ] \\ \le \ , c k^c { { \ensuremath{\mathbb e } } } \left[\zeta({{{\widetilde}\omega}}_k ) \left\vert { { \ensuremath{\mathbf e } } } _ { n , { { \underline{{\omega}}}}}\left ( a\right){{\ensuremath{\mathbf e } } } _ { n , { { \underline{{\omega}}}}}\left ( \delta_k\right)- { { \ensuremath{\mathbf e } } } _ { n , { { \underline{{\omega}}}}}\left ( a\,\delta_k\right)\right\vert \right ] \\
\le \ , c k^c \left({{\ensuremath{\mathbb e } } } \left [ \left\vert { { \ensuremath{\mathbf e } } } _ { n , { { \underline{{\omega}}}}}\left ( a\right){{\ensuremath{\mathbf e } } } _ { n , { { \underline{{\omega}}}}}\left ( \delta_k\right)- { { \ensuremath{\mathbf e } } } _ { n , { { \underline{{\omega}}}}}\left ( a\,\delta_k\right)\right\vert ^2 \right ] { { \ensuremath{\mathbb e } } } \left[\zeta({{{\widetilde}\omega}}_k)^2 \right]\right)^{1/2 } \\ \le c^\prime k^{c } \vert a \vert_\infty e^{-c_2 d(\mathcal s(a),\{k\})},\end{gathered}\ ] ] where in the first inequality we have applied lemma [ th : lbpk ] , in the second the cauchy
schwarz inequality and in the third theorem [ th : correlazioni ] , formula .
since @xmath162 is a local observable , @xmath269 as @xmath270 and therefore the proof of is complete . finally , is a consequence of the decay of the influence of boundary conditions expressed by .
note in fact that states that @xmath271 is a cauchy sequence in @xmath272 .
therefore it convergence in @xmath273 toward a limit random variable that we denote @xmath274 .
therefore holds if we set @xmath275 and , by the fubini tonelli theorem , this clearly implies that @xmath276<\infty,\ ] ] so the series in the expectation is @xmath126a.s .
convergent , which implies the almost sure convergence of @xmath271 , that is .
@xmath277 _ proof of lemma [ lemma : ritorni ] , equation _ .
it is immediate to realize that , letting @xmath278 and @xmath279 , @xmath280,\end{gathered}\ ] ] where @xmath90 was defined in section [ subs : fe ] and @xmath281 is the left shift . indeed
, it suffices to apply lemma [ prop : spezz ] with @xmath282 and @xmath283 .
thanks to part 1 of theorem [ th : zu ] and to the exponential integrability of @xmath118 , one has @xmath284
-_n1n@xmath285 =( ) = : > 0 , @xmath284 so that one obtains from @xmath286 @xmath287 _ proof of lemma [ lemma : ritorni ] , equation _ . in this proof
the positive constants , typically dependent on @xmath24 , will be denoted by @xmath288 .
the constants @xmath289 and @xmath252 are taken from , but since @xmath252 is repeated several times we set @xmath290 .
let us first define , for @xmath291 , @xmath292 , @xmath293 for @xmath294 and @xmath295 .
then , for @xmath296 we let @xmath297 . we will refer to the interval @xmath298 as to the @xmath299 excursion from zero of the walk @xmath300 , and to @xmath301 as to its length , see figure [ fig : figura ] .
( note that the @xmath302 and the @xmath303 excursions may have an endpoint outside @xmath304 . )
= 12 cm [ c][l]@xmath305 [ c][l]@xmath306 [ c][l]@xmath307 [ c][l]@xmath308 [ c][l]@xmath309 [ c][l]@xmath310 [ c][l]@xmath311 [ c][l]@xmath312 [ c][l]@xmath313 [ c][l]@xmath314 [ c][l]@xmath315 [ c][l]@xmath316 [ c][l]@xmath317 [ c][l]@xmath318 [ c][l]@xmath319 [ c][l]@xmath320 [ c][l]@xmath321 [ c][l]@xmath322 [ c][l]@xmath323 [ c][l]@xmath324 [ c][l]@xmath325 [ c][l]@xmath326 [ c][l]@xmath327 [ c][l]@xmath328 [ c][l]@xmath329 [ c][l]@xmath330 [ c][l]@xmath331 [ c][l]@xmath332 [ c][l]@xmath333 [ c][l]@xmath334 [ c][l ] @xmath22 [ c][c ] @xmath335 [ c][l ] @xmath336
[ c][l]@xmath337 [ c][l]@xmath22 the basic observation is that , as we will prove in a moment , there exists @xmath338 independent of @xmath24 such that@xmath339 uniformly in @xmath171 , @xmath340 , for some finite constant @xmath341 . in words , this means that with high probability at least @xmath342 of the interval @xmath304 is covered by excursions of @xmath36 whose length is smaller than @xmath343 ( we will call them _ short excursions _ ) . to prove , set @xmath344 , @xmath345 and @xmath346 so that the condition in the probability in the left hand side of reads @xmath347 .
obviously , the number @xmath348 of excursions entirely contained in @xmath349 and of length at least @xmath343 ( _ long excursions _ ) is at most @xmath350 , and one can ( very roughly ) bound above the number of possible ways one can place them in the stretch @xmath304 by @xmath351 these facts , together with a simple application of lemma [ prop : spezz ] and eq .
, allow to bound above the left hand side of by @xmath352 where the last inequality follows , if @xmath353 is sufficiently large , from the stirling formula .
we stress that the constant @xmath289 is the one appearing in .
the factor @xmath354 in just takes care of the possible location of @xmath315 and @xmath355 in @xmath356 .
for ease of notation , let @xmath357 be the event @xmath358 that the two walks do not touch zero at the same time between @xmath305 and @xmath306 , and for @xmath291 let @xmath359 be the event @xmath360 then , from eq .
, if @xmath361 , one has @xmath362 to estimate the last term in let us notice that if the event @xmath363 occurs then , denoting by @xmath364 the union of the short excursions of @xmath300 , the set @xmath365 contains at least @xmath366 sites .
as a consequence , recalling that short excursions do not exceed @xmath343 in length , @xmath367 contains at least @xmath368 sites . in words , if @xmath369 then either @xmath370 and @xmath371 belongs to a short excursion of @xmath336 , or the same holds interchanging the roles of @xmath337 and @xmath336 .
one can rewrite @xmath372 as the disjoint union @xmath373 , where @xmath374 and , of course , at least one among @xmath375 and @xmath376 contains @xmath377 points .
therefore , using also the symmetry between @xmath337 and @xmath336 , one has @xmath378 where @xmath379 is the subset of @xmath304 which satisfies the following properties : 1 .
@xmath370 for every @xmath380 ; 2
. for every @xmath380 there exist @xmath381 such that * @xmath382 * @xmath383 * @xmath384 if @xmath385 and @xmath386 .
note that , since we are working on @xmath363 , @xmath384 when @xmath387 : this prescription is not contained in the definition of @xmath379 .
one can now write ( see also fig .
[ fig : trem ] ) @xmath388 in the second step we have decomposed the probability by summing over all _ a priori _ admissible configurations @xmath389 of the set @xmath379 , i.e. , all possible subsets of @xmath304 containing at least @xmath390 sites .
we are now going to relax the constraint given by @xmath391 , but estimating the corresponding ratio of probabilities .
we claim in fact that , given @xmath392 and @xmath393 , we have @xmath394 where @xmath395 where @xmath396 is a positive constant that depends on @xmath90 and on the value of @xmath397 .
the bound , which will be applied with @xmath398 , is proven as follows .
first observe that @xmath399 and in the ratio in the right hand side we may factor the expression containing the copolymer energy term , that is we can set @xmath92 and we can restrict ourselves to considering @xmath400 . we write : @xmath401 where @xmath402 may be chosen , with a very rough estimate , equal to @xmath403 .
since @xmath404 , we obtain .
therefore from , using , we can extract @xmath405 = 12 cm [ c][l]@xmath406 [ c][l]@xmath407 [ c][l]@xmath408 [ c][l]@xmath409 [ c][l]@xmath336 [ c][l]@xmath337 [ c][l]@xmath305 [ c][l]@xmath410 [ c][l]@xmath411 [ c][l]@xmath412 [ c][l]@xmath413 [ c][l]@xmath414 [ c][l]@xmath306 [ c][l]@xmath22 the remaining problem now is that @xmath415 is random and can get arbitrarily close to @xmath37 .
this however does not happen too often : in fact there exists @xmath416 such that , for every @xmath169 , @xmath417 for every value of @xmath418 .
the bound follows from a direct ( large deviation ) estimate on the binomial random variable with parameters @xmath419 , which can be made arbitrarily small , and @xmath420 .
it suffices for example that @xmath421 .
thanks to @xmath422 , implies that , on the complementary of @xmath423 , @xmath424 together with and , this completes the proof of .
_ proof of theorem [ th : cinf]_. thanks to the ascoli - arzel theorem , it is sufficient to show that , for every integer @xmath184 , the @xmath426 derivative of @xmath427 with respect to any of the parameters @xmath428 is bounded above uniformly in @xmath28 .
for definiteness , let us show this property for @xmath429 the above derivative is given by @xmath430 which is expressed through truncated correlation functions ( ursell functions ) defined as @xmath431 where the sum runs over all partitions @xmath432 of @xmath433 into subsets @xmath434 .
starting from the property of decay of correlations for every pair of bounded local observables , one can prove by induction over @xmath435 that @xmath436 \\ \!\!\!\!\!\!\!\!\!\!\!\!\!\ ! & & \qquad \le c_1^{(k)}\vert a_1\vert_\infty\ldots \vert a_k\vert_\infty \exp(-c_2^{(k)}d(\mathcal s(a_1),\ldots,\mathcal s(a_k ) ) ) , \nonumber\end{aligned}\ ] ] for some finite positive constants @xmath437 where , if @xmath438 is the smallest interval including the supports of @xmath439 , then @xmath440 .
a proof that the exponential decay of average two points correlations implies exponential decay of the average @xmath22point truncated correlations can be found for instance in @xcite , in the context of disordered @xmath441dimensional spin systems in the high temperature or large magnetic field regime ( see remark [ rem : g ] below for a sketch of the proof ) . in @xcite ,
explicit bounds for the constants @xmath437 are also given , which is not needed in our case .
the proof in @xcite , which is basically an application of hlder s inequality , can be transposed to the present context almost without changes , to yield .
then , after an application of the cauchy - schwarz inequality , it is immediate to realize that the sum in converges uniformly in @xmath28 , for every @xmath184 .
@xmath442 [ rem : g]a sketch of how is deduced from goes as follows . consider the case @xmath443 and assume that @xmath444 ( which is just the case we need in view of ) and , without loss of generality , let @xmath445 .
then consider the simple identities @xmath446 from the first identity and we obtain @xmath447 \le 3c_1 e^{-c_2 ( n_3-n_2)},\end{aligned}\ ] ] while from the second identity we obtain the bound @xmath448 on the same quantity . therefore @xmath449\le
3c_1e^{-c_2((n_3-n_2)+(n_2-n_1))/2},\end{aligned}\ ] ] which is just the statement of in this specific case .
_ part 1 . _
the existence of the limit follows from the subadditivity of @xmath450\right\}_n$ ] , which is an immediate consequence of the renewal property ( of @xmath12 ) and of the iid property ( of @xmath25 ) : for every @xmath451 , @xmath452 , we have in fact @xmath453 \ , & \le \ , { { \ensuremath{\mathbb e } } } \left [ \frac{(1+e^{-2\lambda\sum_{n=1}^m({\omega}_n+h)})(1+e^{-2\lambda\sum_{n = m+1}^n({\omega}_n+h ) } ) } { z_{m , { { \underline{{\omega}}}}}^{{\underline{v}}}z_{n - m , \theta^m { { \underline{{\omega}}}}}^{{\underline{v } } } } \right]\\ & = { { \ensuremath{\mathbb e } } } \left [ \frac{1+e^{-2\lambda\sum_{n=1}^m({\omega}_n+h ) } } { z_{m , { { \underline{{\omega}}}}}^{{\underline{v } } } } \right ] { { \ensuremath{\mathbb e } } } \left [ \frac{1+e^{-2\lambda\sum_{n=1}^{n - m}({\omega}_n+h ) } } { z_{n - m , { { \underline{{\omega}}}}}^{{\underline{v } } } } \right ] .
\end{split}\ ] ] the inequality @xmath454 is an immediate consequence of the elementary lower bound @xmath455 a more refined lower bound on @xmath456 , valid in the localized region , follows from the concentration inequality : call @xmath457 the event @xmath458 since for @xmath28 sufficiently large @xmath459 with @xmath460 and @xmath353 suitable positive constants , one has @xmath461 \
, \le \ , 2\exp\left(-n{\textsc{f}}({{\underline{v}}})/4\right)+ \frac { { { \ensuremath{\mathbb e } } } \left[\zeta({{{\widetilde}\omega}}_n)^{-1}{\mathbf{1}}_{\ { \ , e_n^\complement\ } } \right]}{{{\ensuremath{\mathbf p } } } \left ( { \omega}_n^+\right ) } \\ \!\!\!\!\!\!\!\!\!\!\!\!\!\!\ ! & & \qquad\qquad\qquad\qquad\qquad \le 2\exp\left(-n{\textsc{f}}({{\underline{v}}})/4\right)+ \kappa'_1\exp \left(-\kappa'_2 n{\textsc{f}}({{\underline{v}}})^2 /\max({\lambda } , { \widetilde}{\lambda})^2\right),\end{aligned}\ ] ] which immediately implies @xmath462 and , for @xmath145 sufficiently small , @xmath463 .
finally , @xmath464 is an immediate consequence of jensen s inequality .
@xmath465 _ part 2 .
_ throughout this proof @xmath169 , so that @xmath462 , and we set @xmath466 .
we start with the proof of @xmath467 which clearly implies .
for @xmath468 we set @xmath469 so that @xmath470 , @xmath471 ranging up to @xmath472 , and we have @xmath473 let us estimate the terms in the sum by writing first @xmath474 as the disjoint union of the sets @xmath475 with @xmath476 and @xmath477 .
recalling the notations of section [ sec : conv ] , we have the bound @xmath478 this is an immediate consequence of the renewal property of @xmath36 .
notice that once we take the expectation of both sides of we may set @xmath479 in the right hand side and therefore , by , for every @xmath22 and every @xmath480 , @xmath481 sufficiently large , we have that @xmath482 indeed , the factor @xmath483 is negligible for @xmath348 large since , with probability at least of order @xmath484 , @xmath485 does not exceed @xmath486 , see , so that it does not modify the exponential behavior .
therefore , for @xmath28 sufficiently large and @xmath487 sufficiently small @xmath488 going back to we see that @xmath489 and is proven .
[ rem : cutoff ] it is immediate to realize by looking at the proof that @xmath490 with @xmath491 as @xmath492 .
let us then turn to proving @xmath493 let us set @xmath494 and @xmath495 for @xmath496 .
we now consider the family of events @xmath497 , recall the definition of @xmath498 in , and we observe that @xmath499 . in words , we are simply going to compute the probability that the walk makes at least one excursion of length exactly equal to @xmath500 at @xmath501 prescribed locations : the excursions have to start at @xmath502 for some @xmath371 , see figure [ fig : lb ] .
therefore @xmath503 where the last step defines @xmath504 ( we anticipate that , by theorem [ th : correlazioni ] , formula , @xmath504 is negligible , details are postponed to lemma [ th : rn ] below ) .
we need therefore a lower bound on @xmath505 : we will find a lower bound on this quantity that depends on @xmath25 only via @xmath0 and @xmath506 with @xmath507 and this will allow the direct use of independence when taking the expectation with respect to @xmath26 .
we use the explicit formula @xmath508 = 11.5 cm [ c][l]@xmath29 [ c][l]@xmath28 [ c][l]@xmath509 [ c][l]@xmath510 [ c][l]@xmath511 [ c][l]@xmath512 [ c][l]@xmath513 [ c][l]@xmath500 we proceed by finding an upper bound on the denominator and we first observe that , by , the event @xmath514 has @xmath126probability tending to @xmath37 for @xmath515 and @xmath516 : this is simply due to the fact that @xmath517 .
we will actually choose @xmath99 larger , so that to guarantee that @xmath518 \ge 1-n^{-2}$ ] for @xmath519 large , cf .
remark [ rem : cutoff ] . of the requirements defining the event
@xmath520 we now take advantage only of the fact that there exists a return to zero at distance at most @xmath521 from both sites @xmath502 and @xmath522 , for every @xmath371 . by lemma
[ th : lbpk ] we have that there exists @xmath523 such that for @xmath524 @xmath525 so that from there exist @xmath526 such that @xmath527 where the inequality holds for @xmath519 sufficiently large .
notice that the random variables @xmath528 are independent , since @xmath529 depends on @xmath530 with @xmath22 only in @xmath531 .
going back to , we have @xmath532 we take now the @xmath126expectation of both sides of : by independence , by the choice of @xmath99 in the definition of @xmath533 and by lemma [ th : rn ] @xmath534\right ) \ ,
+ \ , { { \ensuremath{\mathbb p } } } \left[a_{n}^\complement\right ] \ , + \ , { { \ensuremath{\mathbb e } } } \left[\left\vert r_n({{\underline{{\omega}}}})\right\vert\right ] \\ & = \ , \prod_{j=1}^{j_n } \left ( 1- { { \ensuremath{\mathbb e } } } \left[q_{n , j}({{\underline{{\omega } } } } ) \right]\right)\ , + \ , o(1/n^2 ) .
\end{split}\ ] ] we are reduced to estimating @xmath535 $ ] . by , for @xmath487 sufficiently small and @xmath28 sufficiently large , we have @xmath536\,\ge \ , \exp\left(- \mu ( { { \underline{v } } } ) ( 1+{\varepsilon}/2 ) a_n^-\right)\,\ge\ , n^{-1 + { \varepsilon}/4}.\ ] ] from this one gets @xmath537 by applying the markov inequality and borel cantelli lemma we conclude the proof of .
@xmath538 [ th : rn ] for every @xmath539 we have @xmath540 \ , = \ , 0,\ ] ] with @xmath504 defined in .
_ observe that a direct application of theorem , formula , yields that for @xmath541 we have @xmath542 \\ \le\ , c_1 \exp\left(-c_2 \frac{(\log n)^2}2\right ) .\end{gathered}\ ] ] one now applies iteratively this inequality starting from @xmath543 , down to @xmath544 , obtaining that @xmath545 $ ] is bounded above by @xmath546 .
@xmath547 it makes use of the identity @xmath548 _ upper bound .
_ we first observe that , reasoning as in section [ sec : maximal ] , we have @xmath549 where @xmath550 ( recall that we are working here with two sided disorder sequences , so that the @xmath25 variables may have negative indices ) .
now we claim that for every @xmath117 there exists @xmath551 , @xmath126a.s .
finite , such that @xmath552 for every @xmath553 and @xmath554 such that @xmath555 and every @xmath556 .
of course implies that for every @xmath117 @xmath557 by combining and one directly obtains the upper bound in . we are therefore left with the proof of .
this follows by observing first that @xmath558 where @xmath559 is the disorder sequence reflected around the origin : @xmath560 .
therefore , with @xmath561 , we have @xmath562 the leading terms are the last two : by definition of @xmath145 , for every @xmath117 there exists @xmath563 , @xmath126almost surely finite , such that for @xmath553 and @xmath554 larger than @xmath563 both @xmath564 and @xmath565 are larger than @xmath566 and this easily yields the existence of @xmath567 such that @xmath568 for every @xmath569 .
the remaining term in the last line of is treated by an analogous splitting of the sum and by applying the law of large numbers one sees that it gives a negligible contribution for @xmath214 sufficiently large .
finally the first two terms in the right hand side of are both vanishing as @xmath570 diverges by the properties of @xmath90 , cf . , and by the fact that @xmath108 is integrable , so that @xmath571 , @xmath126a.s .. @xmath572 _ lower bound .
_ by selecting in the sum in only the _ excursion _ from @xmath184 to @xmath573 we have that we can write for @xmath574 @xmath575 and we will use also @xmath576 where @xmath577 is the event @xmath578 ( with the conventions of section [ sec : conv ] for @xmath579 ) . thanks to lemma [ th : lbpk ] , one can write for some @xmath523 @xmath580 which , together with and , implies @xmath581^{-1 } -{{\ensuremath{\mathbf p } } } _ { n,{{\underline{{\omega}}}}}(o^{\complement}_{k , s}).\end{aligned}\ ] ] proceeding in analogy with the proof of the upper bound , and using in addition lemma [ th : lbpk ] to bound @xmath582 _ above _ , one can show that @xmath583 diverges @xmath126a.s . as @xmath584 , for every @xmath117 .
therefore , keeping in mind that @xmath585 , the first term on the r.h.s .
of is bounded _ below _ by @xmath586 if @xmath214 is larger than @xmath587 , where @xmath588 is a suitable @xmath126a.s .
finite number , and the same quantity may then be bounded below by @xmath589 where @xmath590 is a @xmath126a.s .
positive random variable . on the other hand , from the upper bound in proposition [ th : as ] one easily obtains @xmath591 taking @xmath487 small enough , this immediately implies the lower bound in proposition [ th : as ] for @xmath592 .
the restrictions on @xmath184 is of course due to having chosen in the first step of the proof the @xmath36_excursion _ from @xmath184 to @xmath573 .
but we may as well choose the @xmath36_excursion _ from @xmath593 to @xmath594 , and the argument may be repeated yielding the same bound , except for different @xmath25 dependent constants , for every @xmath184 ranging from @xmath595 to @xmath596 .
therefore the proof is complete for @xmath184 in @xmath597 and this is the whole set of sites smaller than @xmath28 if @xmath210 .
_ proof of proposition [ th : nfin1]_. just note that @xmath599 where the first inequality is a consequence of the renewal property of @xmath36 and the last one of lemma [ th : lbpk ] .
inequality then immediately follows .
@xmath600 _ proof of theorem [ th : nfin2]_. it is sufficient to prove that there exists @xmath601 such that , for every @xmath602 , @xmath603 it is convenient to define , for @xmath604 , @xmath605 which , in view of , just corresponds to @xmath606 if @xmath607 . note that , since @xmath608 , one has @xmath609 ( cf .
lemma [ lemma : derivne0 ] below ) and that @xmath610 provided that @xmath611 .
indeed , since @xmath612 is convex in @xmath613 one has @xmath614 from which immediately follows , since @xmath615 .
essentially the same argument shows that there esists a smooth ( e.g. differentiable ) path @xmath616 , with @xmath617 , such that @xmath618 , @xmath619 , and such that @xmath620 for every @xmath621 .
note that at @xmath622 the desorder dependence disappears and we have simply a homogenous pinning model which , thanks to , is in the localized phase .
for this model it is easy to prove ( this can be extracted , for instance , from appendix a of @xcite ) that @xmath623 on the other hand , we will prove in a moment that @xmath624 uniformly for @xmath625 .
of course , eqs .
- immediately give .
to prove , let us compute for instance the derivative with respect to @xmath613 : with obvious notations , @xmath626 which , thanks to eq . , is bounded above by @xmath627 @xmath628 being bounded above uniformly for @xmath629 belonging to the path .
similar estimates hold for the derivatives with respect to @xmath630 and therefore follows . @xmath631
the first step in the proof of theorem [ th : clt ] is to show that the variance of @xmath632 is not trivial in the infinite volume limit : [ lemma : nontriv ] if @xmath169 , then @xmath633 _ proof of lemma [ lemma : nontriv ] . _
the upper bound is an immediate consequence of the deviation inequality applied to @xmath634 .
indeed , it is immediate to verify that in this case @xmath635 , for some constant @xmath266 . to obtain the lower bound
, we employ a martingale method analogous to that developed in @xcite .
suppose first that @xmath607 .
observe that , if @xmath636 $ ] and @xmath637-{{\ensuremath{\mathbb e } } } \left[x_{n,{{{\widetilde}\omega}}}|{{{\widetilde}\omega}}_1,\ldots,{{{\widetilde}\omega}}_{k-1}\right],\ ] ] then @xmath638\ge \sum_{k=1}^n { { \ensuremath{\mathbb e } } } \left[\left({{\ensuremath{\mathbb e } } } \left(y^{(n)}_k|{{{\widetilde}\omega}}_k\right)\right)^2\right],\end{aligned}\ ] ] with the convention that @xmath639={{\ensuremath{\mathbb e } } } \left [ x_{n,{{{\widetilde}\omega}}}\right]$ ] if @xmath640 .
next , observe that @xmath641= { \widetilde}\lambda \,{{\ensuremath{\mathbb e } } } \left[{{\ensuremath{\mathbf e } } } _ { n,{{\underline{{\omega}}}}}(\delta_k)|{{{\widetilde}\omega}}_k\right]\ge0 , \end{aligned}\ ] ] and that @xmath642\right|\le { \widetilde}\lambda^2.\end{aligned}\ ] ] at this point , one can employ the following ( @xcite ) let @xmath643 , @xmath644 and let @xmath645 be the set of functions @xmath646 and , for every probability law @xmath647 on @xmath648 , @xmath649 then , @xmath650 is convex in @xmath651 and @xmath652 for @xmath653 , provided that @xmath647 is not concentrated on a single point .
identifying @xmath654 , @xmath655 $ ] and @xmath647 with the law of @xmath118 , using the convexity of @xmath656 and recalling , one obtains immediately @xmath657 to conclude the proof of lemma [ lemma : nontriv ] , it suffices therefore to show that [ lemma : derivne0 ] if @xmath169 , @xmath658 _ proof of lemma [ lemma : derivne0 ] .
_ it is enough to prove that there exists @xmath659 sufficiently small such that @xmath660 indeed , we have @xmath661 where in the first inequality we employed . using stirling s approximation , it is clear that , if @xmath659 is small enough , the right hand side of decays exponentially in @xmath28 .
@xmath662 in the case @xmath148 , the condition @xmath663 implies @xmath664 and the proof of lemma [ lemma : nontriv ] proceeds analogously , with the role of @xmath96 being played by @xmath94 and replaced by @xmath665 @xmath666 the second step in the proof of theorem [ th : clt ] is the observation that ( this follows for instance from lemma [ th : lbpk ] ) @xmath667 for some @xmath668 independent of @xmath25 .
therefore , keeping also into account the upper bound in , one obtains the following approximate subadditivity property : @xmath669 for some constant @xmath670 and @xmath28 large enough .
this , together with , implies that the following limit exists : @xmath671 next , employing repeatedly , one obtains the decomposition @xmath672 for every @xmath25 ( we assumed for simplicity that @xmath28 is multiple of @xmath673 ) .
therefore , if @xmath674 , @xmath675 where @xmath676 tends to zero in law and @xmath126a.s .. the summands in are iid random variables and lyapunov s condition , which in this case reads simply @xmath677 = 0\ ] ] for some @xmath678 , follows immediately since the deviation inequality implies that @xmath679 $ ] is bounded uniformly in @xmath28 . the central limit theorem is proven .
recall the notations in section [ sec : conv ] .
[ th : lbpk ] there exists @xmath523 such that for every @xmath681 , @xmath682 and every @xmath25 we have that @xmath683 _ proof .
_ we write @xmath684 and @xmath685 using and the definition of slow variation for @xmath56 , one finds that @xmath686 uniformly in @xmath687 , so that the expression can be bounded above by @xmath688 for some constant @xmath670 independent of @xmath25 , which directly yields .
@xmath689 [ prop : spezz ] let @xmath690 , @xmath691 with @xmath692 , and let @xmath693 @xmath694 be events depending only on @xmath695 , i.e. , @xmath696 .
then , @xmath697 _ proof .
_ this is elementary : just rewrite the probability in as @xmath698 and notice that one obtains an upper bound for it if one constrains the walk to touch zero at @xmath699 in the denominator @xmath700 .
at that point , the probability factorizes thanks to the renewal property of the random walk and one obtains just the right hand side of .
[ [ sec : mu ] ] in this section we sketch a proof of the fact that , under some reasonable conditions on the law @xmath26 , the strict inequality @xmath225 holds in the localized region .
since @xmath702 in @xmath157 , let us assume for definiteness that @xmath151 ( in the alternative case , the role of @xmath96 in the following is played by @xmath94 ) . in analogy with @xcite , we assume that one of the following two conditions holds for the iid sequence @xmath96 : * c * : _ continuous random variables_. the law of @xmath118 has a density @xmath703 with respect to the lebesgue measure on @xmath93 , and the function @xmath704 exists and is at least twice differentiable in a neighborhood of @xmath705 .
this is true in great generality whenever @xmath703 is positive , for instance in the case of @xmath706 , with @xmath707 a polynomial bounded below .
_ bounded random variables_. the random variable @xmath118 is bounded , @xmath708 assume first that condition * c1 * holds .
given @xmath709 , let @xmath710 be the law obtained from @xmath26 shifting the law of @xmath711 so that @xmath712=-\varepsilon$ ] . if @xmath659 is small enough , thanks to the assumed regularity of the function @xmath713 in one has @xmath714 for some finite constant @xmath715 .
then , applying the jensen inequality , @xmath716\,= \\ \frac1n\log{\widetilde}{{\ensuremath{\mathbb e } } } _ n \left [ \frac { \left(1+e^{-2\lambda\sum_{n=1}^n({\omega}_n+h)}\right ) } { z_{n , { { \underline{{\omega}}}}}^{{\underline{v } } } } e^{\log\left({{\,\text{\rm d}}{{\ensuremath{\mathbb p } } } } /{{\,\text{\rm d}}{\widetilde}{{\ensuremath{\mathbb p } } } _
n}\right ) } \right ] \\
\ge \ , \frac 1n{\widetilde}{{\ensuremath{\mathbb e } } } _ n \left ( \log \frac{{\,\text{\rm d}}{{\ensuremath{\mathbb p } } } } { { \,\text{\rm d}}{\widetilde}{{\ensuremath{\mathbb p } } } _ n}\right)-\frac1n{{\ensuremath{\mathbb e } } } \,\log z^{{{\underline{v}}}'}_{n,{{\underline{{\omega}}}}},\end{gathered}\ ] ] so that @xmath717 where @xmath718 is obtained from @xmath24 replacing @xmath719 with @xmath720 . since @xmath159 is smooth in the localized region and the derivative @xmath721 is not zero ( cf . ) , for @xmath659 sufficiently small one has @xmath722 in the case of condition * c2 * , the proof of is slightly more complicated . instead of shifting the law of @xmath711 , @xmath710 is obtained by tilting it : @xmath723 the estimate on the relative entropy is still valid , with a different constant @xmath715 , while the proof that @xmath724 for some positive @xmath266 , although rather intuitive in view of the previous case ( note in fact that , for @xmath659 small , the tilting produces a shift of @xmath725 of the average of @xmath711 ) , still requires some care .
we do not report here other details , which the interested reader can easily reconstruct from the proof of lemma 3.3 in @xcite .
@xmath726 here we prove that , if the law of @xmath107 is symmetric , then the definition of @xmath226 is equivalent to .
in fact , @xmath728}\\ \stackrel{s\stackrel{{{\ensuremath{\mathbf p } } } } { \sim}-s}= { { \ensuremath{\mathbb e } } } \frac 1{{{\ensuremath{\mathbf e } } } \left[\exp\left(2{\lambda}\sum_{n=1}^n({\omega}_n+h)\delta_n+{\widetilde}{\lambda}\sum_{n=1}^n({{{\widetilde}\omega}}_n+{\widetilde}h)\delta_n\right ) \delta_n\right]}\\ \le { { \ensuremath{\mathbb e } } } \frac 1{{{\ensuremath{\mathbf e } } } \left[\exp\left(2{\lambda}\sum_{n=1}^n({\omega}_n - h)\delta_n+{\widetilde}{\lambda}\sum_{n=1}^n({{{\widetilde}\omega}}_n+{\widetilde}h)\delta_n\right ) \delta_n\right]}\stackrel{\omega\stackrel{{{\ensuremath{\mathbb p } } } } { \sim}-{\omega}}={{\ensuremath{\mathbb e } } } \frac 1{z^{{\underline{v}}}_{n,{{\underline{{\omega}}}}}},\end{gathered}\ ] ] where in the inequality we simply used the fact that @xmath729 , see section [ sec : model ] .
therefore , @xmath730\ , & \le\ , \frac 1n \log { { \ensuremath{\mathbb e } } } \left [ \frac { 1+e^{-2\lambda\sum_{n=1}^n({\omega}_n+h ) } } { z_{n , { { \underline{{\omega}}}}}^{{\underline{v } } } } \right ] \\ & \le \ , \frac 1n \log { { \ensuremath{\mathbb e } } } \left [ \frac 1 { z_{n , { { \underline{{\omega}}}}}^{{\underline{v } } } } \right]+\frac{\log 2}n , \end{split}\ ] ] and the claim follows in the limit @xmath731
we are very grateful to an anonymous referee for his observations and , in particular , for having pointed out that our decay of correlation estimates yield the almost sure result of th .
[ th : correlazioni ] ( formula ) .
we would like to thank also francesco caravenna for interesting discussions on the content of section 5 .
this research has been conducted in the framework of the gip
anr project jc05_42461 ( _ polintbio _ ) .
m. ledoux , _ measure concentration , transportation cost , and functional inequalities _ , summer school on singular phenomena and scaling in mathematical models , bonn , 1013 june 2003 .
available online : http://www.lsp.ups-tlse.fr/ledoux/ | we analyze the localized phase of a general model of a directed polymer in the proximity of an interface that separates two solvents .
each monomer unit carries a charge , @xmath0 , that determines the type ( attractive or repulsive ) and the strength of its interaction with the solvents .
in addition , there is a polymer interface interaction and we want to model the case in which there are impurities @xmath1 , called again charges , at the interface .
the charges are distributed in an inhomogeneous fashion along the chain and at the interface : more precisely the model we consider is of quenched disorder type . + it is well known that such a model undergoes a localization / delocalization transition .
we focus on the localized phase , where the polymer sticks to the interface .
our new results include estimates on the exponential decay of correlations , the proof that the free energy is infinitely differentiable away from the transition and estimates on finite size corrections to the thermodynamic limit of the free energy per unit site .
other results we prove , instead , generalize earlier works that typically deal either with the case of copolymers near an homogeneous interface ( @xmath2 ) or with the case of disordered pinning , where the only polymer
environment interaction is at the interface ( @xmath3 ) .
moreover , with respect to most of the previous literature , we work with rather general distributions of charges ( we will assume only a suitable concentration inequality ) and we allow more freedom on the law of the underlying random walk .
+ + 2000 _ mathematics subject classification : 60k35 , 82b41 , 82b44 _
+ + _ keywords : directed polymers , copolymers , copolymers with adsorption , disordered pinning , localized phase _ |
observational effects caused by primordial molecules seem to be most promising in investigating dark ages epoch of the universe .
the basic properties of the molecules are discrete narrow energy levels and high efficiency of their interaction with cmbr .
this leads to forming ssf if proto - objects , containing these molecules , move with peculiar velocities v@xmath9 relative to cmbr ( dubrovich,1977,1982 , puy,1993 ) .
we may consider ssf as manifestation of the proto - objects at high redshifts 300 @xmath10 10 as scattering of cmb photons by primordial molecules must leave imprints on the cmbr temperature distribution .
the most abundant chemical elements predicted by the pure big bang model are h , he , d , @xmath11he , li and their ions .
they give some molecules in the primordial matter at z=100 - 200 such as h@xmath12,h@xmath13,hd , hd@xmath2 , heh@xmath2 , lih , lih@xmath2 , h@xmath12d@xmath2 .
there are other molecules which should be considered if non - standard nuclear synthesis at the early times ( z=10@xmath14 ) or star formation at z = 100 - 200 took place .
for the high efficiency of interaction between molecules and cmbr photons two values are important - the cross - section for scattering and the concentration ( or relative abundance ) of this molecule .
the first parameter depends on the specific quantum structure of the molecule - its symmetry and dipole moment .
the second one depends on the abundance of the chemical elements of which it is composed and on the rate of the appropriate chemical reactions . according to these constraints
, we should take into consideration only those molecules which have a large enough dipole moment and relatively high abundance .
these are hd@xmath2 , heh@xmath2 , lih , h@xmath12 d@xmath2 .
we consider them as the basic molecules interacting with cmbr at high z ( dubrovich,1977,1982,puy et al.,1993,stancil et al.,1996 ) .
let us remind physical conditions in dark ages epoch when the universe was younger than 300 mil .
the universe was in @xmath15 times smaller and in @xmath16 times more dense than now , temperature of cmbr and molecular gas followed @xmath17 law till z @xmath18150 .
then , temperature of matter should be lower than cmbr temperature due to different adiabatic indexes of radiation and matter ( varshalovich , hhersonskii , sunyaev , 1981 ) and it must follow @xmath19 law till first star formation .
calculated temperatures of cmbr and molecular gas in dark ages epoch are given in table 1 , we used @xmath20 = 2.726 k ( mather et al . , 1994 ) . in epoch before secondary ionization molecules are formed in accordance with kinetic rate of specific chemical reactions when matter and radiation are cooling .
after reionization molecular formation follows a more complicated law due to growth of ionization degree of hydrogen and existence of heavy chemical elements produced by first stars ( basu , hernandez - monteagudo , sunyaev , 2004 ) .
.temperatures of cmbr and molecular gas in dark ages epoch z=300 - 10 [ cols= " < , < , < , < , < " , ] if we realize a channel bandwidth of @xmath21=2% in 3 mm band preferable or optimal linear scales for ssf search are l=1.5 mpc for heh@xmath22(i=0 ) at z=20 and l=0.75 mpc for lih(i=6 ) at z=30 , correspondent angular sizes according to ( 7 ) are 10.7@xmath7 and 7.5@xmath7 .
recent wmap polarization measurements have shown strong probability of an early reionization scenario with @xmath23 ( kogut et al , 2003 , page et al , 2006 ) due to pop iii stars ( wyithe , loeb,2003 ) in galaxy halos .
ssf detection at z=20 - 30 may provide us with important information about ionization history of the universe .
the question is which maximum proto - object scales are reasonable , could exist at z=20 - 30 or are connected with some known scales in the universe .
let us compare them with spatial scales of primary cmba .
the angular scale of the first peak in primary cmbr power spectrum @xmath24 50@xmath7 , the following cmba peaks are @xmath25 1/2@xmath26,1/3@xmath26 , and 1/5@xmath26 ( see fig.2 down ) .
then possible spatial scales l of 1 - 4 peaks in cmbr power spectrum for different @xmath27 are given in the table 5 with ( 7 ) and ( 6 ) .
the above consideration shows that at @xmath4t / t=2@xmath2810 @xmath6 level and with @xmath21=2% we can search for ssf which are the result of large scale proto - object motion with peculiar velocity relative to cmbr at z=20 - 30 .
these possible proto - objects are still less than spatial scale of the fourth peak in cmbr power spectrum ( table 5 ) which might be related to a proto - galaxy scale . with less @xmath21 , higher sensitivity and multipole resolution we can approach to smaller spatial scales of proto - objects up to proto - star scales .
ssf firstly proposed as a new phenomena in ( dubrovich , 1977 ) is one of most promising secondary cmba .
ssf is non - isotropic spectral distortions of cmbr which are to be present at z=300 - 10 .
homogeneous and isotropic spectral distortions of cmbr may be present at z=6000 - 1000 due to primordial helium and hydrogen recombination ( dubrovich , 1975 , dubrovich , stolyarov , 1995 , dubrovich , shahvorostova , 2004 , dubrovich , grachev , 2004 ) .
non - isotropic distortions of cmbr similar to ssf may be present in this epoch at lower @xmath0 level as well(dubrovich , 1977/2 ) .
calculation of this effect in more details at z=1000 - 1100 is done in ( rubino - martin , mernandez - monteagudo and sunyaev , 2005 ) .
let us remind that cmbr temperature fluctuations are expanded in spherical harmonics ( peebles , 1993,peacock , 1999 ) : @xmath29 the angular power spectrum @xmath30 of @xmath0 can be related to the autocorrelation function @xmath31 as : @xmath32 where @xmath33 are legendre polynomials , l - multipole moment and @xmath34 .
theoretical predictions of @xmath35 are usually quoted as @xmath36 @xmath37 as we consider small - angle patches of the sky cmbr fluctuations @xmath38 may be generated by evaluating the simple fourier series ( bond et al , 1987 ) : @xmath39\ ] ] where @xmath40 denotes the size of the simulated region in radians ; @xmath41 are cartesian coordinates on the sky ( spatial domain ) ; @xmath42 are coordinates of fourier components @xmath43 in spatial frequency domain , @xmath44 , @xmath45 . for the search of ssf cmbr maps obtained in several spectral channels
are to be studied and then processed by the special frequency - differential or the difference method ( dubrovich , bajkova,2004 ) .
the difference method can be considered as an alternative method to the correlation function analysis . in this case
we are interested in the first derivative of spatial distortions with respect to the frequency .
the method used is based on the analysis of a difference of two cmbr temperature maps observed at different frequencies and reduced to one beam shape .
mainly , such a difference map contains information only on the secondary fluctuations because the primary cmbr fluctuations present in both maps will be eliminated due to their black - body spectrum nature .
the procedure of ssf simulation is the following .
firstly cmb map is generated with a standard cmb power spectrum .
then amplitudes of fft spectrum of cmb map are kept while phases are modified as random values uniformly distributed in the interval 0 - 2@xmath46 .
a backward fft gives us an approximation of cmb map which contains primary and secondary fluctuations in conditions of unknown theoretical power spectrum of ssf .
the actual theoretical power spectrum of ssf is to be a subject of nearest energetic efforts , it is strongly needed for following ssf investigation .
let the difference of the cmbr observation frequencies be equal to @xmath47 .
let the limiting angular fluctuation size @xmath26 corresponds to this frequency difference in accordance with equation ( 5 ) .
obviously , the fluctuations of size larger than @xmath26 will be seen at both maps .
after subtraction of one map from the other the fluctuations of size larger than @xmath26 will be mutually suppressed and the remainder map will contain only the fluctuations smaller than @xmath26 . evidently , this fact can be seen also from angular power spectrum of this map which is close to zero at frequencies with multipoles @xmath48 and different from zero at @xmath49 , where @xmath50 is the break of the spectrum .
now let the difference of two frequencies be @xmath51 , where @xmath52 .
the angular size @xmath53 corresponds to this frequency interval .
obviously , @xmath54 in accordance with equation ( 5 ) .
in this case we will see at the difference map the fluctuations with size smaller than @xmath53 and the corresponding angular power spectrum will have the break at @xmath55 .
our cmba model also includes convolution of expected cmba maps with a simulated radio telescope beam and adding a pixel noise .
differential maps are preliminary processed to decrease the noise in given power spectrum region .
let convolved cmbr maps have spatial scales l@xmath56 @xmath57 ( l@xmath58-m@xmath59 , then the spatial scale of differential maps is ( l@xmath58-n@xmath60l@xmath61)@xmath57l@xmath58 , where m=0,1,2 ... , @xmath4l@xmath61/l corresponds to @xmath62/@xmath632% in our case ) .
let us limit l@xmath64 in the simulation by l=1056 which is near to the lowest reasonable limit of l@xmath58 in the ground based mm - wave ssf experiment .
we will use in simulation a standard @xmath65cdm cmbr model ( spergel et al .
, 2003 , spergel et al , 2006 ) with parameters : @xmath66=0.2(@xmath67/100 km / smpc , @xmath68 km / smpc ) , @xmath69=0.65 , @xmath70 , n=1 , h=0.65 .
initial single frequency cmbr maps ( fig.2 up ) 15@xmath71x15@xmath71 ( 128x128 pixels ) have l=3 - 1056(1 ) , l=3 - 1032 ( 2 ) , l=3 - 1008 ( 3 ) , l=3 - 984(4 ) , l=3 - 960(5 ) , l=3 - 936(6 ) , l=3 - 912(7 ) , all @xmath72(quadratic sum)=25902 .
difference maps ( fig.2 down ) have l=1032 - 1056 ( 1 - 2 ) , @xmath72=14.4 , l=1008 - 1056(1 - 3 ) , @xmath73=21.3 , l=984 - 1056(1 - 4 ) , @xmath72=32.4 , l=960 - 1056(l-5 ) , @xmath73=42.9 , l=936 - 1056 ( 1 - 6 ) , @xmath72=60.8 , l=912 - 1056 , @xmath73=87.6 with the angular power spectrum corresponding to @xmath74cdm model for initial ( solid line ) and differential ( dashed line ) cmbr maps ( fig.2 middle ) .
x15@xmath71 ( 128x128 pixels ) ( up ) , difference maps ( middle ) , power spectrum of @xmath4 t corresponding to @xmath65cdm model for initial ( solid line ) and differential ( dashed line ) cmbr maps ( down ) ] it was shown in ( dubrovich , bajkova,2004 ) that the difference method may be used when s / n ratio for each map is 1.44 . in case of higher level of noise a more complicated preliminary signal processing directed to decreasing the noise in the given cmbr spectrum region would be required .
fig.3 demonstrates predicted difference ssf map without an instrumental noise(up ) and with white instrumental noise ( down ) added in single frequency maps with s / n=1.17 .
power equalization method of wiener filtration ( gorski , 1997 ) was applied in this case as signal preliminary processing to decrease the noise in given power spectrum region . in spite of a correlation coefficient of two difference maps of fig.3
is less than 0.15 the power spectra of the difference signal after wiener filtration are similar ( fig.3 right ) that is the most important to recognize ssf in a real ssf search experiment . but this is true for white instrumental noise and we concentrate our efforts in section 4 to avoid or minimize abnormal instrumental pixel noise in the ssf search experiment . 9.5 cm it was assumed in above simulation that the observed cmbr maps consist only of the primary and secondary cmbr fluctuations with some additive white instrumental noise .
however , in practice the cmbr detected maps in addition contain a number of astrophysical foregrounds such as extragalactic unresolved sources , free - free and synchrotron radiation , sz effect , galactic dust and so on .
astrophysical foregrounds are characterized by considerably low value of spectral index as compared with the analyzed secondary cmbr fluctuations and are successfully eliminated in difference maps .
up to now practically all cmbr ground based and space experiments study primary cmba which has black body spectrum .
primary cmba was 30 years under investigation at ratan-600 radio telescope since 1977 ( parijskij , korolkov , 1986 , parijskij et al , 2005 ) .
first attempts of ssf observations were done in 1992 at iram mm radio telescope with the upper @xmath4t / t limit 2@xmath2810@xmath75 at 1.3 mm ( de bernardis , dubrovich et al , 1993 ) and in 2001 at ratan-600 ( the near pole survey ) with the upper @xmath4t / t limit 10@xmath76 at 6 cm ( gosachinskii , dubrovich et al.,2001 ) .
surplus frequency resolution of the existing spectrum - analyzers ( 30 mhz and 20 khz ) increased @xmath4t / t limit in both cases . the far infrared absolute spectrometer(firas ) instrument has shown that cmb has ideal black body spectrum from 60 ghz to 600 ghz and cmb temperature is 2.725 + -0.001 k(fixsen , mather,2002 ) .
then fixsen et all reported about an effort to measure distortions of cmb black body spectrum due to of various processes in early universe .
this attempt was done in arcade balloon - born experiment where cmb temperatures were measured at 10 ghz and 30 ghz with accuracy + -0.010 k and + -0.032 k ( fixsen et all , 2004 ) . in this paper
we suggest a concept and strategy of the ssf experiment with @xmath60t / t limit @xmath77 .
high absolute measurement accuracy is not necessary in this experiment as we need to measure differential temperatures in nearby receiver channels without measuring absolute cmb temperatures .
simulation of possible ssf experiment is done for msrt ( khaikin et al , 2002,2003 ) , tuorla observatory , finland ( fig.4 ) equipped with a 7x4 beam cryo - microblometer array at 3 mm with rf multiplexer providing up to 7 spectral channels in each beam .
the scheme of the experiment is given in fig.5 .
firstly idea of such an experiment is suggested in ( dubrovich , bajkova , khaikin , 2004 , khaikin , dubrovich , 2004 , khaikin , luukanen , dubrovich , 2005 ) . in this paper
we develop the concept and present simulation results of a possible experiment .
the challenge is to split rf band in front of bolometers .
if superconducting nb microstrip resonant @xmath78/4 stub filters are used in rf multiplexer ( rfm ) the in - band transmission can approach unity due to negligible loss of the superconducter . to simplify a multiplexer design
the number of filters and spectral channels may be reduced to 5 or even 3 .
a wideband microstrip superconducting frequency multiplexer is applied in front of tes bolometers in samba(superconducting antenna - coupled multifrequency bolometric array ) , where low loss and excellent out - of - band rejection is demonstrated at 3 frequencies of mm band ( goldin et al , 2003 ) .
antenna - coupled tes bolometer with an integrated microstrip bp filter at 217 ghz is applied for a bolometer array in a ground based cmb polarization experiment ( myers et al , 2005 ) .
another way to split radio frequency band in a bolometric receiver is to build a quasi - optical filter based on metal mesh filters(mmf ) or frequency - selective surfaces ( fss ) .
a review of quasi - optical mmfs is given in ( ade et al , 2006 ) .
a few possible designs of cryogenic mmfs with bandwidth up to 6@xmath79 at 1 - 3 mm are given in ( page,1994 ) .
3 cryogenic filter wheels with several bandpass filters , are under development now in columbia university for millimeter bolometric array camera ( mbac ) to provide 5 - 10% frequency resolution in one octave mm band from 150 ghz to 300 ghz at atacama cosmology telescope(act ) .
fss may be also used to separate in space radio telescope foci in which several feeds or arrays at different frequencies may be placed ( kasyanov , khaikin , 2004 ) .
but it is difficult to achieve frequency resolution with mmf or fss better than 5@xmath79 and superconducting @xmath80 microstrip resonant filters are more effective in our case .
we consider several variants of an array feed . as
a narrow flare corrugated horn is difficult to build in 3 mm band a trimode conical horn may be used to provide a small flare angle with an acceptable sidelobe level ( less than 20 db in e / h plane ) and low frequency cut - off near 87 ghz .
a smooth - walled spline - profile horn(granet et al , 2004 ) is an alternative to both corrugated and trimode horns and it allows us to achieve an identical radiation pattern in e / h plane in the band , low sidelobe level(less than -25 db ) , more tight array packaging and lower manufacture cost .
we are in final stage now to build and test such a horn for array applications al lower frequencies ( 30 ghz-38 ghz ) in cooperation with c.granet(csiro)(khaikin et al,2007 ) .
a 2 m precise carbon fiber msrt dish ( khaikin et al,2002 ) allows us to work in 100 ghz and 150 ghz bands but the second one is much more difficult due to higher atmospheric attenuation and antenna scattering background .
angular resolution of msrt is 7@xmath7 - 6@xmath7 at 88 ghz-100 ghz that is enough for a large scale ssf experiment and corresponds to l@xmath58=@xmath46/@xmath81=1542 - 1800 ( for full sky coverage ) the same as at plank ( 7@xmath7 , l=1542 ) and twice better than at wmap(l@xmath58=750 ) .
nevertheless for a ground based or balloon experiment @xmath82 , where @xmath83 ( size of the symmetrical map ) and actual spatial resolution of msrt is not higher than l=1200 - 1500 for partial sky coverage . in boomerang balloon experiment
actual resolution was less than l=1025 at 150 ghz(de benardis et al.,2002 ) .
1000@xmath84 seems to be optimal for ssf search at current stage . at higher l with the low frequency resolution s
/ n ratio falls . besides , in ground based experiments with higher angular resolution such as cosmic microwave imager ( cbi ) ( readhead et al,2004 ) with l @xmath58= 3500 point sources were a serious problem at 26 - 36 ghz as the power spectrum of primary cmba falls rapidly with l while the point source contribution grows as l@xmath85 in the band .
point source contamination is still dangerous at 3 mm at higher l if secondary cmba has a falling spectrum .
some secondary cmba may have inverted @xmath30 spectra when the magnitude of @xmath30 grows at higher l ( doroshkevich , dubrovich , 2001 ) .
however ssf search at l@xmath861500 needs both higher sensitivity and higher frequency resolution which contradicts each other .
a multifeed or receiver array is needed in a ssf experiment to get a cmbr map in an unmovable radio telescope mode ( preferable ) or to accelerate mapping in a scanning mode ( less preferable because of variable instrumental effects ) .
we plan to use a chopping mode to reduce an abnormal instrumental noise and atmospheric effects .
a chopper with beam deviation @xmath87 will distort power spectrum of ssf due to filtration of scales @xmath88 and changing the amplitudes at scales @xmath89 but this is not so dangerous in the task of ssf search .
the main role of the chopper is to suppress uncorrelated residual 1/f noise of different bolometers which can not be eliminated in difference maps .
identical and stable calibration of all receiver channels is very important in this experiment .
it may be produced quasi - optically by a lens , forming a lateral plane wave for a whole array .
a high stable noise generator with a direct coupler or a matched black body load is placed in the lens focus .
a calibration source must be put in a reverse two level thermostat realizing 0.1 k accuracy at the first and not less than 0.01 k accuracy at the second level . according to our concept msrt
central beam is fixed near the north celestial pole ( ncp ) at @xmath90=86@xmath71 to increase cosmic source passing time via the beam up to 348s -402s and to observe the 1@xmath71 x 8@xmath71 strip preferably in low culmination from @xmath91=8@xmath9200@xmath93 to 16@xmath9200@xmath93 at high galactic latitudes .
actual spatial resolution of a combined cmbr map in the scheme of fig.5 is 6@xmath7 - 7@xmath7(hpbw ) .
the long - length focus cassegrain scheme with m=(e+1)/(e-1)=10 - 11 ( eccentricity e@xmath861 ) gives us negligible aberrations of off - axis beams but requires more complicated array feeds .
a short - length focus cassegrain radio telescope scheme gives us unacceptable level of off - axis aberrations(khaikin , luukanen , 2003 ) .
we simulated some antenna characteristics for the scheme of the experiment given at fig.5 .
antenna parameters are d=2 m , d=0.3 m , f / d=0.435 , interfocal distance f=150 cm , feed flare angle is 12 deg .
simulated characteristics are given in table 6 .
multibeam focal plane array ( fpa ) configuration presented in fig.5 allows us to avoid undersampling in a survey mode .
for simulation of the multibeam radiation pattern , calculation of some other antenna characteristics ( illumination , aperture , spillover , sidelobe efficiencies , beam overlapping level etc . )
focal plane array simulation ( fopas ) program package has been used ( khaikin , 2005 ) .
fopas takes into account geometry of a single or dual reflector radio telescope , expected aperture distribution or the beam pattern of an ideal broadband feed , the feed removal from the focus and off - axis aberrations .
fopas allows us to optimize fpa geometry with the beam spacing and the overlapping level which are optimal for the given astrophysical task .
the task of multibeam fpa optimization is also considered in ( majorova , khaikin , 2005 ) .
multibeam radiation pattern simulated with fopas is given in fig.6 . & & & & + & & & & + & & & & + & & & & + we consider transition edge sensors ( tes ) as cryo - bolometers for this experiment ( khaikin , luukanen , dubrovich , 2005 ) .
the lateral dimension of tes is less than 10 microns ( microbolometer ) and standard lithographic techniques may be used to build arrays of 100 - 1000 sensitive elements .
photon and phonon noise equivalent power ( nep ) and noise equivalent temperature difference ( netd ) of tes for the ground based ssf experiment are considered in ( khaikin , luukanen , dubrovich , 2005 ) .
an expected phonon nep of a cryo - microbolometer with the physical temperature of 0.3 k is @xmath941@xmath9510@xmath96 w / hz@xmath97 .
we do need the physical temperature near 0.3 k ( fig.5 ) as nep of tes dramatically grows at higher physical temperatures ( luukanen , pekola,2003 ) .
photon nep from radio telescope antenna and atmosphere is not less than 2@xmath9510@xmath96 in good atmospheric conditions .
this will give us @xmath72=1.0 - 2.0 mk s@xmath97 sensitivity per beam per channel ( 2 ghz bandwidth ) in the best atmospheric conditions with t@xmath98@xmath9420 k ( the lowest atmospheric opacity ) . in 8x8 element planar bolometer array at 90 ghz for gbt
operating temperature is 450 mk , which limits nep of tes bolometers to 2@xmath9910@xmath96 w / hz@xmath100 ( benford et.al,2004 ) .
the physical temperature @xmath101 of an antenna coupled with a superconducting bolometer must be chosen to provide a photon nep@xmath9410@xmath96 w / hz@xmath100 .
for rather high radiative antenna efficiency @xmath102=0.9 and t@xmath9=300 k an additional photon noise from an antenna is 10@xmath103 wt / hz@xmath97 and cooling of antenna ( horn)up to at least 10 k is necessary to provide its own photon nep at less than 10@xmath104 wt / hz @xmath97 level ( khaikin , luukanen , dubrovich,2004)(fig.5 ) . for deep cryogenic cooling sorption @xmath105 coolers at 0.3 k of applied physics institute of ras in nizhnij novgorodmay
be used(vdovin , 2005 ) . some ground based cmbr projects ( lawrence et al,2004 ) chose an alternative way and are being built cryogenic mmic mm - wave cameras with expected sensitivity of 500 @xmath106 in continuum in 3 mm band .
we have considered such alternative as well .
inp mmic lna at 15 k with a broadband subharmonic mixer and if microstrip 10 channel multiplexer may give us 4 - 5 mk@xmath107 per beam per channel(1% bandwidth ) at 85 ghz -105 ghz ( khaikin et al , 2004 ) .
the actual reachable integration time of a bolometer depends on the knee frequency f@xmath108 in its noise power spectrum . in a bolometer case f@xmath108
may be reduced from tens of hz to tens of mhz by biasing the bolometer with an ac current and synchronous demodulation in readout electronics ( crill et al , 2002 ) .
f@xmath108=10 mhz was reached by ac biasing of bolometers in a 144 pixel cryo - bolometer array bolocam at 140 - 270 ghz ( mauskopf et al . , 2000 ) . with ac biasing of the tes
bolometers we expect f@xmath108=20 mhz . in the effective chopping mode with chopper frequency f@xmath109=5 hz f@xmath108=5 mhz may be reached then 3 - 4 minutes integration time becomes available . using a 100 ms sampling rate for f@xmath109=5 hz and 170 s synchronous integration time we can reach @xmath72 of thermal noise not worse than 0.1 - 0.2 mk per beam per pixel .
four 1@xmath71x8@xmath71 maps ( 128x7 pixels ) are to be recorded during about 8 hours in clear cold winter nights ( after sunset and before sunrise ) and then combined into one map ( 64x14 pixels ) with minimum angular resolution @xmath110 and a noise factor of 1/2@xmath97 in comparison with initial maps .
20 - 25 identical maps must be summed for 20 - 25 clear cold nights with low atmospheric opacity in order to reach 30@xmath111k @xmath72 of thermal noise in the averaged single frequency maps .
we expect contribution of non thermal instrumental noise , atmosphere and foregrounds in differential maps up to 30 @xmath111 k as well .
so we may expect total @xmath72 = 50 - 55 @xmath111k in differential maps .
big efforts have to be made to reduce the contribution of instrumental and foreground effects at @xmath4t / t=10@xmath6 level that corresponds to 10@xmath112 of ambient temperature .
some instrumental effects such as 1/f noise , ground spillover , radiometric offset may be comparable with an atmospheric noise but their contribution may be significantly reduced in the ground based experiment .
atmospheric fluctuations are to be effectively suppressed with a chopper while high atmospheric attenuation is the most dangerous for a ground based mm - wave ssf experiment with cryo - bolometers .
atmosphere gives the main limitation in sensitivity in mm wave ground based ssf experiment with cryo - bolometers as it will contribute up to 80% in a system temperature and in some cases may saturate ( overheat ) a cryo - bolometer .
the bolometer must be able to absorb up to 1 pw in order to avoid saturation with photon nep @xmath113 wt / hz@xmath97 .
this requires a rather high dynamic range d which is the ratio of the maximum power detected to the minimal detectable power . for expected sensitivity 1
mk s@xmath97 d must be not worse than 45 db .
in fact only days with lowest atmospheric opacity ( lowest water vapour content ) may be used for such an experiment at sea level .
it was shown by a `` tipping curve '' method that typical atmospheric opacity at msrt site in zenith @xmath114 is 0.2 db at 36.5 ghz for dry clear autumn nights ( khaikin et al , 2003 ) that gives predicted @xmath115 at 94 ghz 0.3 db - 0.5 db in clear cold winter nights with lowest integrated water wapour content ( @xmath116 ) .
an effective beam wobbling made it possible to reduce residual atmospheric noise @xmath4t@xmath98 and spillover @xmath4t@xmath117 up to 10 @xmath111k at 40 ghz in sk cmbr experiment in canada ( wollack et al , 1994 ) , where the radiometric offset was less than 7 @xmath111k per day for all observations .
a special ground screen ( height up to 3 m ) is to be installed near msrt ( fig.5 ) to avoid the contribution of the hot atmosphere and ground via sidelobes , spillover and scattering background and provide t@xmath118= 33 k at high elevation angles including own antenna temperature , atmosphere ( 20 k ) and cmbr(khaikin , luukanen , dubrovich,2005 ) .
the residual non - thermal instrumental noise must be significantly reduced up to 30 @xmath111k in differential maps while thermal noise grows by a factor 2@xmath119 in comparison with single frequency maps .
non - thermal noise can not be reduced lower in difference maps due to residual abnormal ( 1/f ) instrumental noise , some difference in radiation patterns , in amplitude - frequency characteristics of the channels and some difference in atmospheric effects within 12% bandwidth . in spite of atmospheric limitations a ground based ssf mm wave experiment has advantages in the cost , flexibility and possible duration . in space conditions
tes bolometers may give one - two orders better sensitivity . in ( ali et al , 2003 ) netd=25 @xmath111 k / s@xmath119 per pixel is expected in space for cmbr at 90 ghz for 100 tes detectors with the physical temperature less than 0.1 k and nep=@xmath120 wt / hz@xmath97 . yet , most of new generation mm wave cmbr projects chose multi - element bolometer arrays to reach highest sensitivity in the ground based experiment . @xmath121 per detector
must be reached in 1000 element array of tes bolometers at 10 m south pole telescope(spt ) in 20% bandwidth at 95 ghz and 150 ghz(ruhl et al , 2004 ) .
millimeter bolometric array camera(mbac ) with 3 1000-element tes bolometer arrays at 150 ghz/220 ghz/270 ghz is under development now for 6 m atacama cosmology telescope(act ) project ( fowler , 2004 ) .
expected mbac sensitivity to cmbr is 300 @xmath111k/500@xmath111k/700@xmath111k@xmath122 per detector & & & & + & & & & + @xmath100 & & & & + & & & @xmath123 & + in a ground based experiment at sea level atmospheric effects contribute much more than astrophysical foregrounds nevertheless we should take them into account to better prepare the experiment .
galactic and extragalactic dust emission will significantly dominate at 3 mm over other foreground effects such as discrete and diffuse sources , galactic synchrotron and free - free emission , hot clusters of galaxies .
dust fluctuations are seen on all angular scales from 2@xmath7 to tens of degrees on iras data ( low et al.,1984 ) .
antenna temperature is @xmath124 , where @xmath125 is the spectral index which depends upon the emissivity of the dust and lies in the range of 1.5 - 2.2 according to cobe , iras and wmap data .
after 60 ghz galactic and extragalactic dust dominate and we should carefully chose cmbr map coordinates in dust free areas of the universe at high galactic latitudes . the full sky thermal dust map based on data from iras and cobe using mem procedure
shows a wide range of temperatures from 400 @xmath111k to 15 @xmath111k in w band of wmap ( 94 ghz ) ( bennet at el , 2004 ) .
contribution of dust emission is to be significantly eliminated in differential maps as this component must be well enough correlated in single frequency maps due to its black - body spectrum .
as it was shown in section 2 the expected signal ( @xmath0 of ssf ) depends on many cosmological parameters and conditions in early universe such as @xmath126,@xmath127 , @xmath128 , @xmath129,@xmath130,@xmath131 and others while the best s / n ratio at given z also depends on @xmath21 .
it is extremely difficult to predict the most probable combination of these parameters at different z in early universe and ssf experiments are urgently needed to set ssf an uppper limit at @xmath0=@xmath132 level and specify theoretical predictions of ssf .
we want to emphasize in final remarks the significance of ssf detection which leads to the discovery of the cosmological molecules .
cosmological molecules can provide an unique test to non - standard nuclear synthesis at high z. ssf can only help us in investigating dark ages epoch in the early universe ( @xmath133 ) .
ssf detection at z=20 - 30 may give an important information about ionization history of the universe .
finally , they will allow us to explore physical properties of dark matter and dark energy .
observational effects caused by interaction of primordial molecules with cmbr seem to be most promising in investigating of dark ages epoch of the early universe and ionization history of the universe . strategy and concept of the experiment
are given .
simulation of the experiment in an atmospheric transparency window at 3 mm with msrt equipped with a 7x4 beam cryo - microbolometer array gives us the expected sensitivity of 50 - 55 @xmath111k in the differential cmbr maps near ncp with angular resolution 6@xmath7 - 7@xmath7 ( l=1200 - 1500 ) and frequency resolution * * @xmath4**@xmath134/@xmath135=2% .
the expected integral sensitivity in the experiment is @xmath4t / t=2@xmath2810 @xmath136 which is near to optimistic estimates of ssf caused by molecules heh@xmath2 at z=20 - 30 .
simulation shows that ssf may be recognized in the angular power spectrum when s / n in single frequency cmbr maps is as small as 1.17 for a white noise .
such an experiment will give us a possibility to set upper limits in ssf observations in mm band and prepare future ssf experiments .
khaikin , v.b .
simulation of antenna characteristics of mm - wave radio telescopes with multibeam fpa . in proceedings of 28
th esa antenna workshop on space antenna systems and technologies , nordwijk , the netherlands , may , 2005 .
khaikin , v.b . ,
radzikhovsky , v.n . ,
kuzmin s.e . ,
shlenzin , s.v .
, popenko , chernobrovkin , r. granet , c. tightly packed multibeam fpa at 30 - 38 ghz . in proceed .
of 29 th esa antenna conference , noordwijk , the netherlands , april , 2007 .
mauskopf , p.d .
rownd , s.f .
edgington , v.v .
hristov , a. k. mainzer , j.glenn , a.e .
lange , j.j .
bock , and p. a. r ade .
imaging at radio through submillimeter wavelengths .
2000 , asp conference seriess , 217 , p.115 .
parijskij , yu.n .
bursov , n.n .
berlin , a.b .
balanovskij , a.a .
khaikin , v.b .
majorova , e.k .
mingaliev , m.g .
nizhelskij , n.a .
pylypenko , o.m .
tsibulev , p.a .
verkhodanov , o.v .
zhekanis , g.v .
zverev , yu.k .
ratan-600 new zenith field survey and cmb problems .
gravitation&cosmology , 10(2004 ) , pp.1 - 10 . | spectral spatial fluctuations ( ssf ) of the cosmic microwave background radiation ( cmbr ) temperature are considered as a result of an interaction of primordial atoms and molecules with cmbr in proto - objects moving with peculiar velocities relative to the cmbr .
expected optimistic values of @xmath0 are @xmath1 for ssf caused by heh@xmath2 at z = 20 - 30 which are possible redshifts of early reionization scenario .
the bandwidth of the lines is 0.1 - 2% depending on the scale of proto - objects and redshifts . for the ssf search cmbr maps in different spectral channels
are to be observed and then processed by the difference method .
simulation of the experiment is made for msrt ( tuorla observatory , finland ) equipped with a @xmath3 beam cryo - microbolometer array with a chopping flat and frequency multiplexer providing up to 7 spectral channels in each beam ( 88 - 100 ghz ) .
expected @xmath4t / t limit in the experiment is 2@xmath510 @xmath6 with 6@xmath7 - 7@xmath7 angular and 2% frequency resolution .
simulation shows that ssf may be recognized in the angular power spectrum when s / n in single frequency cmbr maps is as small as 1.17 or even something less for white noise .
such an experiment gives us a possibility to set upper limit of ssf in mm band and prepare future ssf observations .
cosmic microwave background ( cmb),cosmology , molecules , radio telescope 98.80.@xmath8k,34.50.@xmath8s,95.85.bh |
lattice solitons are a topic of great interest to ongoing studies of nonlinear dynamics in photonic media and becs ( bose - einstein condensates ) @xcite .
these localized modes are produced by the interplay of the intrinsic nonlinearity of the medium with an effective periodic potential induced in it by permanent or virtual lattice patterns .
in fact , the lattice may itself be a nonlinear structure if it is induced by a spatially periodic modulation of the local nonlinearity @xcite . in the limit of a deep periodic potential
, the fundamental models of lattice media reduce to various versions of the discrete nonlinear schrdinger ( dnls ) equation @xcite .
the realization of the one - dimensional ( 1d ) dnls model in arrayed optical waveguides was originally proposed in ref .
the same model was later applied to becs loaded into deep optical - lattice potentials @xcite ( see ref .
@xcite for a brief review ) .
a physical realization of the dnls model is also possible in the form of lattices of microcavities which serve as traps for polaritons @xcite .
lattice solitons take the form of discrete solitons in terms of the dnls equations , which correspond to quasi - discrete solitons in the respective experimental settings .
such solitons were created in a set of semiconductor waveguides built on top of a slab substrate @xcite , and also in arrays of optical fibers @xcite .
in addition to using permanent photonic structures , quasi - discrete solitons were also made in virtual waveguiding arrays , using the versatile technique of inducing interference lattices in photorefractive crystals @xcite .
the latter method was used to create the first examples of 2d quasi - discrete fundamental solitons moti2d , which was followed by the making of vortex solitons @xcite ,
i.e. , localized lattice excitations with embedded vorticity , that were predicted in ref .
another significant contribution to this area was the creation of 2d solitons in a bundle of fiber - like waveguides written in bulk silica @xcite .
such arrays and bundles are created by means of tightly focused femtosecond laser pulses @xcite . following the analysis of the fundamental localized discrete vortices with topological charge @xmath1 @xcite , their higher - order counterparts , with @xmath3 , and
multipole discrete solitons , such as quadrupoles , were predicted in refs . .
many other objects were studied in this area , including _ supervortices _ ( circular chains of compact vortices with an imprinted overall topological charge , which is independent of the vorticity of the individual eddies @xcite ) , necklace - shaped patterns @xcite , discrete solitons in hexagonal and honeycomb lattices @xcite , composite _ semidiscrete _ spatial solitons in arrays of waveguides with quadratic and cubic nonlinearities @xcite , quasi - discrete topological solitons in photonic - crystal fibers @xcite , etc .
nonstationary soliton effects were studied too .
these include the mobility of discrete solitons boulder , papa , collisions between traveling ones @xcite , and the onset of the spatiotemporal collapse in self - focusing arrayed waveguides @xcite .
most works on lattice solitons dealt with the spatial - domain settings . in particular , optically - induced lattices in photorefractive crystals do not makes it possible to observe the evolution in the temporal domain because of a very large response time in these materials .
however , the spatiotemporal dynamics can be realized in waveguiding arrays written in bulk silica jenawriting , where the spatially localized quasi - discrete patterns in the transverse plane can be combined with the temporal self - trapping in the longitudinal direction .
recently , the creation of the corresponding quasi - discrete light bullets " was reported in this system @xcite ( for a review of spatiotemporal solitons in nonlinear optics and bec , see ref .
@xcite ) . previously ,
a number of manifestations of the spatiotemporal self - trapping in similar settings were studied theoretically , including the related modulational instability @xcite , formation of bullets " in fiber arrays @xcite and photonic wires @xcite , and self - compression @xcite and steering @xcite of pulsed beams . continuing the work in this direction ,
semi - discrete spatiotemporal surface solitons were introduced , as _ surface modes _ , in models of semi - infinite waveguide arrays mihkiv , and in a system with an interface between different arrays staggunstagg .
also analyzed were spatiotemporal solitons in waveguide arrays with the quadratic nonlinearity @xcite .
once stable fundamental spatiotemporal soliton complexes in bundled arrays of waveguides are available , it is natural to seek for vortex solitons in the same setting .
a systematic analysis of spatiotemporal vortices and quadrupoles in the model based on the square lattice of discrete waveguides was reported in ref .
a vast stability area was found for the solitary vortices with @xmath4 and quadrupoles , which are built as rhombuses , alias on - site - centered modes , with respect to the underlying lattice ( the rhombus is built as a set of four bright " cores , with a nearly dark " one at the center ) .
the stability region is much smaller for the off - site - centered modes of the square " type , without an empty pivotal site in the middle ( the reduced stability domain of square - shaped vortices and quadrupoles , in comparison with their rhombic counterparts , in a generic feature of topological solitons in lattice media @xcite ) .
all the spatiotemporal vortex solitons with @xmath5 were found to be unstable unstable in the same model .
further , collisions between stable vortices and quadrupoles ( with identical or opposite topological charges ) , propagating along the bundle in opposite directions , were analyzed in ref .
four different outcomes of the collisions were identified : rebound of slowly moving solitary vortices , fusion , splitting , and quasi - elastic interactions between fast ones .
hexagonal lattices may be created by means of the same techniques which were used for the building the square - shaped structures . on the other hand
, the change of the underlying geometry may essentially alter fundamental properties of topological lattice solitons @xcite .
in particular , it was predicted theoretically and conformed in an experiment that spatial solitons in the form of double vortices ( with @xmath2 ) in hexagonal lattices may be stable , while their unitary counterparts ( with @xmath1 ) are unstable .
the objective of the present work is to study spatiotemporal vortex solitons in hexagonal lattices of discrete waveguides .
the model is formulated in section 2 , and at the end of it we also briefly consider fundamental solitons , driven by a temporally self - trapped pulse in a single waveguiding core . in section 3 ,
we demonstrate , also in a brief form , that a straightforward input in the form of a hexagon - shaped spatiotemporal vortex with @xmath1 always leads to a decay .
nevertheless , three different species of _ stable _ spatiotemporal complexes with the embedded vorticity are revealed by a systematic numerical analysis . in section 4 , we demonstrate that a spatiotemporal input of a triangular shape generates self - trapped vortices in the form of triangles with an empty core in the middle .
further , in section 5 it is shown that a hexagonally shaped input with @xmath2 produces stable spatiotemporal hexagons with the same ( _ double _ ) topological charge .
finally , a modified ( shifted ) input ansatz gives rise to stable densely packed triangular vortices with @xmath1 , without an empty central core , as shown in section 6 .
in addition to the study of these species of spatiotemporal vortex solitons , in section 7 collisions between counterpropagating triangular ones are studied .
the paper is concluded by section 8 .
we consider the hexagonal array of nonlinear waveguides , with cells in the transverse lattice numbered as shown in fig .
[ framec ] .
the transmission of waves in the array is described by the following system of coupled nls equations , written in the scaled form , similar to that used in many earlier works @xcite-@xcite , @xcite:@xmath6 \nonumber \\ & + & ( 1/2)\partial _ { t}^{2}u_{m , n}+u_{m , n}\left\vert u_{m , n}\right\vert ^{2}=0 .
\label{eq1}\end{aligned}\]]in terms of the optical setting , @xmath7 and @xmath8 are , respectively , the propagation distance and reduced time , assuming that each guiding core features the anomalous chromatic dispersion and cubic self - focusing , while @xmath9 is the propagation constant of the localized solution to be sought for . in terms of the corresponding bec model , eqs .
( [ eq1 ] ) is a system of coupled discrete gross - pitaevskii equations @xcite , with @xmath7 and @xmath8 playing the roles of the scaled time and axial coordinate , respectively , while @xmath10 is the chemical potential .
, with discrete coordinates @xmath11 defined as shown in the figure .
the map of integers @xmath12 into the cartesian coordinates in the transverse plane , @xmath13 , is performed as per eq .
( [ xmn ] ) , @xmath14 being the width of the hexagonal cell.,width=302 ] simulations of eq .
( [ eq1 ] ) were carried out in the fourier domain , with the help of the standard fourth - order runge - kutta scheme , the nonlinear term being evaluated by means of the combination of inverse and direct fast fourier transforms at each sub - step of the scheme .
we used an @xmath15 matrix in the plane of @xmath11 , 512 points for variable @xmath8 in the computation window of width @xmath16 , and the stepsize in the propagation direction @xmath17 .
the use of the fourier transform implies periodic boundary conditions in @xmath8 , which make sense if a characteristic temporal size of the localized objects will be essentially smaller than @xmath16 . as concerns the boundary conditions for
the discrete coordinates @xmath18 and @xmath19 , the values of @xmath20 corresponding to the coordinates which fall outside of the computation box are replaced by zeros . before proceeding to the search for complex spatiotemporal vortical patterns , it makes sense to test the propagation of fundamental solitons , which are carried , essentially , by a temporal pulse in a single core . for this purpose ,
the simulations were initiated with obvious initial conditions , @xmath21@xmath22setting @xmath23 .
the simulations were run in interval @xmath24 of values of the propagation constant .
it has been concluded that input ( [ fond ] ) decays , under the action of the lattice diffraction , at @xmath25 , and a stable fundamental soliton , concentrated in the central core , is formed at @xmath26 .
the temporal pulse which lies at the core of the so created fundamental soliton is not quite stationary , but rather features regular pulsations , as shown in fig .
[ osc_fond ] .
one may surmise that the oscillations of the fundamental soliton could be a result of its interaction with the radiation background , which was generated by the input field in the course of self - trapping into the fundamental soliton . to check this possibility ,
the background around the soliton was explicitly removed , at a particular step of the simulations .
nevertheless , the oscillations remain virtually unaffected by the cleaning " , i.e. , they seem to be a genuine feature of the dynamics of the soliton , possibly representing its intrinsic mode .
first , we attempted to create hexagonal vortical modes with @xmath1 , which seems a natural approach to the system based on a hexagonal lattice . to this end , we used the following input , based on an ansatz factorized in the longitudinal ( temporal ) and transverse ( spatial ) directions , cf . ref .
japan:@xmath27 \nonumber \\ & & \times \eta ~\mathrm{sech}\left ( \eta t\right ) , \label{hexag}\end{aligned}\]]with @xmath28 taken as per eq .
( [ eta ] ) , and@xmath29@xmath30@xmath14 being the width of the hexagonal cell , see fig .
[ framec ] .
the model s scale is fixed by setting @xmath31 .
factor @xmath32 in eq .
( [ hexag ] ) obviously corresponds to vorticity @xmath1 , and the exponential factor with @xmath33 taken as per eq .
( [ a ] ) is determined as in the 2d spatial soliton with propagation constant @xmath34 .
the choice of @xmath28 as per eq .
( [ eta ] ) implies that , simultaneously , the wave field in the factorized ansatz is localized in the longitudinal direction , in each core , as in the temporal soliton corresponding to the same propagation constant , cf .
the structure of the fundamental solitons considered above .
the phase and energy structure of ansatz ( [ hexag ] ) is illustrated , in a schematic form , by fig .
hexag_trans . )
patterns corresponding to input ansatz ( [ hexag ] ) which generates an ( unstable ) hexagonal vortex with @xmath35 .
, width=302 ] direct simulations of eq .
( [ eq1 ] ) with this input have been run in a broad range of values of the propagation constant , @xmath36 .
nonetheless , stable spatiotemporal vortices with the hexagonal structure and topological charge @xmath1 have never emerged .
in fact , the evolution of the input organized as the ansatz of this type never leads to formation of any stable pattern . in the interval of @xmath37 ,
the system makes an attempt to generate a robust pattern of a triangular shape , as shown in fig .
[ hexag_neuf ] : at three sites belonging to the original hexagon , the field quickly decays , while at three others it survives , for a while . however , the largest amplitude of the temporarily emerging triangular set is @xmath38 ( it is attained at @xmath39 ) , while triangular vortices may be stable for amplitudes above a threshold value of the amplitude which is @xmath40 ( see below ) , therefore the triangles developing from the unstable hexagons are also subject to an instability , eventually splitting into uncorrelated single - core excitations which separate in the longitudinal direction , see fig .
[ hexag_neuf ] . .
in this case , the simulations demonstrate the decay of the hexagon into a transient triangle , which is followed by a longitudinal instability ( splitting ) .
, width=302 ] further , in the interval of @xmath41 , the instability splits the original hexagon into a set of separating single - core excitations , the number of which varies randomly between @xmath42 and @xmath43 ( not shown here in detail ) . in an adjacent interval , @xmath44 ,
the instability - development scenario is similar but faster , so that the formation of a transient triangular structure can not be identified . at largest values of the propagation constant , @xmath45
, the instability - development scenario is different .
the six sites forming the hexagon keep their positions and amplitudes for a while , but loose the mutual phase coherence .
then , instabilities of amplitudes and positions set in , but they manifest themselves on a much longer scale of the propagation distance , with @xmath7 ranging from @xmath46 to a few hundreds , instead of @xmath47 at small @xmath10 , cf . fig .
[ hexag_neuf ] . the separation between excitations in individual cores grows very slowly too , in comparison with the quick split of the transient triangle observed in fig .
[ hexag_neuf ] .
the next step is an attempt to generate a triangular vortical structure , which is suggested by the emergence of a transient one in the course of the evolution of the unstable hexagon ( fig .
[ hexag_neuf ] ) . for this purpose
, we used the same input as defined by eqs .
( [ hexag])-([a ] ) , but with three main peaks suppressed , which was done by replacing the fields at the corresponding sites by those from adjacent sites in the outer layer : @xmath48 , @xmath49 , @xmath50 , as shown in fig . [ triangle ] .
the so constructed triangular ansatz keeps the vorticity of the original hexagon , @xmath51 .
the evolution of this input was simulated in a broad range of values of the propagation constant , @xmath52 . at small values , @xmath53 the three main peaks forming the triangle merge into a single - core fundamental soliton , which may be localized at the central site , or at any one belonging to the original triangle .
thus , stable vortices do not emerge in this case . for @xmath54 ,
the triangle is destroyed by an instability which splits it into separating uncorrelated single - core excitations ( the instability develops faster with the increase of @xmath10 ) .
finally , the same input generates _
stable _ triangular vortices at @xmath55 , an example of which is shown in fig .
[ triangle_200 ] .
the stability was verified by direct simulations for long propagation distances , e.g. , @xmath56 for @xmath57 .
the temporal pulses in the cores which represent vertices of the triangle remain well phase - locked , keeping the phase circulation of @xmath58 , which corresponds to vorticity @xmath1 .
the excitations at secondary sites ( between the vertices ) are phase - locked to the primary ones , but featuring some oscillations .
the oscillations enhance with @xmath10 , but the overall vortical phase pattern always persists .
on the other hand , the amplitudes of excitations at the secondary sites feature fast irregular oscillations , which also become stronger at larger @xmath59 ( variations of these amplitudes by a factor @xmath60 are observed already at the stability threshold , @xmath57 ) .
these amplitude oscillations are coupled to small variations of amplitudes at the primary sites , as shown in fig .
[ triangle_mumax ] .
it has been checked that the oscillations were not induced by reflection of small - amplitude radiation waves from edges of the integration domain ( absorbers installed at the edges do not suppress the oscillations ) .
.,title="fig:",width=151 ] .,title="fig:",width=151 ] figure [ triangle_curve ] displays the total energy , @xmath61 , and amplitude of the triangular vortices , both stable and unstable ones , as functions of the effective propagation constant , @xmath62 where the contribution @xmath63 from oscillations of the fields is computed as follows . for each vertex of the triangle , @xmath64 , peak time @xmath65 is defined , such that @xmath66 and the corresponding phase , @xmath67 , is identified .
next , we compute @xmath68where the average is taken over the three vertices of the triangle ( or six ones for stable hexagonal vortices with @xmath2 , see the next section ) , and , for the stable triangular modes , over @xmath7 .
for unstable triangles , the latter average was taken over a short interval @xmath69 , within which the pattern was not disturbed by the instability .
as concerns the amplitude shown in fig .
[ triangle_curve ] , it was defined as @xmath70 , averaged over @xmath7 and over the three vertices , to smooth effects of small persistent oscillations of the local amplitudes . a small gap between the unstable and stable portions of the amplitude plot in fig . ( [ deltamu ] ) is due to the difference in wavenumber shift ( [ deltamu ] ) , as computed for the stable and unstable solutions . ,
( [ deltamu ] ) .
blue ( dashed ) and red ( solid ) segments designate unstable and stable solution families , respectively.,title="fig:",width=147 ] , see eq .
( [ deltamu ] ) .
blue ( dashed ) and red ( solid ) segments designate unstable and stable solution families , respectively.,title="fig:",width=147 ]
as said above , the hexagonal input based on eqs .
( [ hexag])-([a ] ) could not produce any stable pattern .
however , the same initial ansatz , but with inverse signs of three of its main peaks
say @xmath71 , @xmath72 , @xmath73can give rise to _ stable _ hexagonal spatiotemporal patterns carrying vorticity @xmath74 ( if the topological charge of the original ansatz is defined as @xmath75 ) , see an example of the stable mode in fig .
[ double_223 ] .
note that no changes were made to the hexagonal input at sites in outer layers .
simulations with this input were also run in a broad range of values of the propagation constant , @xmath76 . at @xmath77 ,
the seeded pattern is subject to various instabilities : spread out due to the lattice diffraction and temporal dispersion , or splitting into separating excitations , or , sometimes , merger into one or two single - core fundamental solitons . in particular , in the interval of @xmath78 , the pattern forms a transient triangular structure , which eventually splits , and in a broad interval of @xmath79 the initial hexagon fissions into two triangles , which also turn out to be unstable essentially , because the amplitudes of the triangular patterns fall below the stability threshold . _
stable _ double ( @xmath2 ) hexagonal vortices emerge at @xmath80 , whose shape is illustrated by fig .
[ double_223 ] .
the stability of these vortices was verified in very long simulations . with the further increase of @xmath10 ,
an instability island was revealed around @xmath81 . in that case , the six temporal pulses remain locked to their positions , but the phase structure is lost at @xmath82 .
nevertheless , the hexagonal vortices recover their integrity at still larger @xmath10 .
it is possible that other narrow intervals of the instability may be found inside the stability region .
figure [ double_curve ] presents the energy and amplitude of the double vortices as functions of the effective wavenumber , similar to fig .
triangle_curve for the triangular vortices with @xmath1 .
however , only the stable family of the hexagonal vortices is shown here , as we were not able to measure characteristics of unstable ones at @xmath77 .
in fact , the simulations produce no evidence that such unstable modes exist . , but for the hexagon - shaped vortices with the double topological charge .
only stable modes are presented in these plots.,title="fig:",width=147 ] , but for the hexagon - shaped vortices with the double topological charge .
only stable modes are presented in these plots.,title="fig:",width=147 ]
still another type of stable spatiotemporal patterns can be produced by the input taken as per eqs .
( [ hexag])-([a ] ) , but centered at an edge of the original hexagon , i.e. , with @xmath18 and @xmath19 replaced by @xmath83 and @xmath84 , respectively . in this case , the results were collected for @xmath85 , and , as shown in fig .
[ ptttri_15 ] , the stable structure takes the form of a vortex with @xmath1 , shaped as a _ densely packed _ triangle , without an empty site in the center , cf .
[ triangle_200 ] .
this structure is stable for @xmath86 .
it is relevant to stress that this stabilization threshold is more than an order of magnitude lower than its counterparts for the triangular and hexagonal spatiotemporal vortices reported in the previous sections ( recall those thresholds were @xmath87 and @xmath88 , respectively ) .
in fact , direct simulations initiated by the above - mentioned shifted input ansatz generate the compact triangle which seems noisy " .
the noise can be removed by means of the temporal filtering " , setting the field equal to zero outside of the main pulse in each core , and running the additional propagation over @xmath89 .
furthermore , for @xmath90 , direct simulations starting from the shifted input ansatz lead to a phase instability . for instance , at @xmath91 and @xmath92 , the phases of the three vertices would take values @xmath93 ,
@xmath94 and @xmath95 , instead of those displayed in fig .
[ triangle_200 ] .
actually , this instability is caused by the fact that the input is far from the shape of the stable mode , giving rise to several temporal peaks in each core . if the initial data are cleaned up " by nullifying the field outside of the main temporal pulse , the simulations converge to stable compact triangular vortices .
the amplitude of stable compact triangular vortices evolves slowly and almost linearly versus the effective propagation constant @xmath96 : the computed values of the latter range from @xmath97 to 78 , then the amplitude goes from 40.3 to 41.2 .
the availability of stable solutions for the vortex spatiotemporal solitons , and the obvious galilean invariance of eq .
( [ eq1 ] ) suggest to study collisions between moving vortices .
in particular , it is interesting to simulate collisions between stable triangular modes shown in fig .
triangle_200 , rotated by angle @xmath98 relative to each other , to test a possibility of their fusion into a full hexagonal vortex of the type displayed in fig .
[ double_223 ] .
this was done taking a pair of the triangles separated by a relatively large temporal interval , @xmath99 , for values of @xmath100 , at which the triangular vortices are stable by themselves , as shown above .
they were set in motion , multiplying them by @xmath101 , which , obviously , lends the solitons velocities @xmath102 ( in terms of the optical waveguides , these are shifts of the inverse velocities ) .
in fact , the fusion of colliding triangles into a hexagon was never observed . instead ,
slowly moving triangles demonstrate a long - range repulsion and stop at finite distance ( but do not bounce back ) , as shown in fig .
[ inter11 ] .
relative to each other , for @xmath103 and velocities @xmath104.,width=302 ] at intermediate velocities , the colliding triangular vortices do bounce back , and eventually they get destroyed by the longitudinal instability ( splitting into uncorrelated temporal pulses in different cores ) , as shown in fig .
[ inter12 ] . and @xmath105 .
trajectories of individual pulses forming the vortices are displayed.,width=302 ] at high velocities , the solitons , quite naturally , pass through each other , loosing some kinetic energy .
there is a sharp threshold between the rebound regime and the passage . just above this threshold ,
the passing vortices get destroyed by the longitudinal instability shortly after the collision .
domains corresponding to different outcomes of the collisions in the @xmath106 plane are shown in fig .
[ li_inter ] .
.,width=302 ] while the collisions seem elastic with the increase of the velocity , it was not possible to conclude if the vortices remain stable indefinitely long after such quasi - elastic collisions .
indeed , since the numerical box has a finite length , and periodic boundary conditions in @xmath8 are used , the moving vortices undergo repeated collisions , loosing some velocity each time .
eventually , they would be destabilized by a collision occurring at a lower speed .
we have introduced a system of parallel waveguides with the linear coupling between nearest neighbors , based on the hexagonal lattice in the transverse plane .
each guiding core features the cubic self - attractive nonlinearity .
the system can find straightforward realizations in nonlinear optics , and in bec trapped in the corresponding optical lattice .
systematic simulations , starting with a natural input ansatz for vortical hexagons , reveal three distinct species of stable semi - discrete spatiotemporal complexes , which are discrete in the transverse plane and continuous in the longitudinal direction .
these are triangular modes with vorticity @xmath1 and hexagonal ones with @xmath2 , both built with an empty core at the center , and compact triangles carrying @xmath1 , without the central empty core .
collisions between stable triangular vortices were also studied by means of simulations , demonstrating the stoppage of the slowly moving vortex solitons , destabilizing rebounds , and quasi - elastic passage , depending on the collision velocity .
more complex structure of the arrayed waveguides can be considered as a generalization of this work ( in particular , quasi - periodic lattices ) .
it may also be interesting to study vortex complexes in two - component models of the same type .
this work was supported , in a part , by a grant from the high council for scientific and technological cooperation between france and israel .
the work of dm at laboratoire de photonique dangers , has been supported by a senior chair grant from the rgion pays de loire , france .
99 f. lederer , g. i. stegeman , d. n. christodoulides , g. assanto , m. segev , and y. silberberg , phys .
rep . * 463 * , 1 ( 2008 ) ; s. flach and a. v. gorbach , _
ibid_. * 467 * , 1 ( 2008 ) ; y. v. kartashov , v. a. vysloukh , and l. torner , in : progr .
optics * 52 * , 63 ( e. wolf , editor : north holland , amsterdam , 2009 ) ; eur .
j. special topics * 173 * , 87 ( 2009 ) .
a. trombettoni and a. smerzi , phys .
lett . * 86 * , 2353 ( 2001 ) ; g. l. alfimov , p. g. kevrekidis , v. v. konotop , m.salerno , phys .
e * 66 * , 046608 ( 2002 ) ; r. carretero - gonzlez and k. promislow , phys .
a * 66 * , * * * * 033610 ( 2002 ) ; f. s. cataliotti , s. burger , c. fort , p. maddaloni , f. minardi , a.trombettoni , a. smerzi , m. inguscio , science * 293 * , 843 ( 2001 ) ; m.greiner , o. mandel , t. esslinger , t. w. hnsch , i. bloch , nature * 415 * , 39 ( 2002 ) ; n. k. efremidis and d. n. christodoulides , phys . rev . a * 67
* , 063608 ( 2003 ) .
j. e. heebner and r. w. boyd , j. mod
. opt . * 49 * , 2629 ( 2002 ) ; p. chak , j. e. sipe , and s. pereira , opt . lett . * 28 * , 1966 ( 2003 ) ; j. j. baumberg , p. g. savvidis , r. m. stevenson , a. i. tartakovskii , m. s. skolnick , d. m. whittaker , and j. s. roberts , phys .
b * 62 * , r16247 ( 2000 ) ; p. g. savvidis and p. g. lagoudakis , semicond . sci .
technol . * 18 * , s311 ( 2003 ) .
n. k. efremidis , s. sears , d. n. christodoulides , j. w. fleischer , and m. segev , phys .
e * 66 * , 046602 ( 2002 ) ; n. k. efremidis , j. hudock , d. n. christodoulides , j. w. fleischer , o. cohen , and m. segev , phys .
lett . * 91 * , 213906 ( 2003 ) ; j. w. fleischer , g. bartal , o. cohen , t. schwartz , o. manela , b. freedman , m. segev , h. buljan , n. k. efremidis , opt
. exp . * 13 * , 1780 ( 2005 ) .
j. w. fleischer , m. segev , n. k. efremidis , and d. n. christodoulides , nature * 422 * , 147 ( 2003 ) ; j. w. fleischer , g. bartal , o. cohen , o. manela , m. segev , j. hudock , and d. n. christodoulides , phys .
* 92 * , 123904 ( 2004 ) .
d n. neshev , j. alexander , e. a. ostrovskaya , y. s. kivshar , h. martin , i. makasyuk , and z. chen , phys .
lett . * 92 * , 123903 ( 2004 ) ; j. w. fleischer , g. bartal , o. cohen , o. manela , m. segev , j. hudock , and d. n. christodoulides , phys .
lett . * 92 * , 123904 ( 2004 ) .
a. szameit , j. burghoff , t. pertsch , s. nolte , a. tnnermann , and f. lederer , opt . exp . * 14 * , 6055 ( 2006 ) ; a. szameit , t. pertsch , f. dreisow , s. nolte , a. tnnermann , u. peschel , and f. lederer , phys . rev .
a * 75 * , 053814 ( 2007 ) .
t. pertsch , u. peschel , f. lederer , j. burghoff , m. will , s. nolte , and a. tnnermann , opt .
* 29 * , 468 ( 2004 ) ; a. szameit , d. blmer , j. burghoff , t. schreiber , t. pertsch , s. nolte , a. tnnermann , and f. lederer , opt . exp . * 13 * , 10552 ( 2005 ) .
p. g. kevrekidis , b. a. malomed , and y. b. gaididei , phys .
e * 66 * , 016609 ( 2002 ) ; c. r. rosberg , d. n. neshev , a. a. sukhorukov , w. krlikowski , and y. s. kivshar , opt . lett .
* 32 * , 397 ( 2007 ) ; t. j. alexander , a. s. desyatnikov , and y. s. kivshar , _ ibid_. * 32 * , 1293 ( 2007 ) ; b. terhalle , t. richter , a. s. desyatnikov , d. n. neshev , w. krlikowski , f. kaiser , c. denz , and y. s. kivshar , phys . rev . lett . * 101 * , 013903 ( 2008 ) ; k. j. h. law , p. g. kevrekidis , v. koukouloyannis , i. kourakis , d. j. frantzeskakis , and a. r. bishop , phys . rev .
e * 78 * , 066610 ( 2008 ) ; k. j. h. law , h. susanto , and p. g. kevrekidis , phys .
a * 78 * , 033802 ( 2008 ) ; k. j. h. law , a. saxena , p. g. kevrekidis , and a. r. bishop , _ ibid_. * 79 * , p. xie , z .- q .
zhang , and x. zhang , phys .
e * 67 * , 026607 ( 2003 ) ; a. ferrando , m. zacars , p. f. de cordoba , d. binosi , and j. a. monsoriu , opt .
exp . * 11 * , 452 ( 2003 ) ; * 12 * , 817 ( 2004 ) ; a. ferrando , m. zacars , and m. a. garca - march , phys .
* 95 * , 043901 ( 2005 ) .
j. meier , g. i. stegeman , y. silberberg , r. morandotti , and j. s. aitchison , phys . rev . lett . * 93 * , 093903 ( 2004 ) ; j. meier , g. i. stegeman , d. n. christodoulides , r. morandotti , m. sorel , h. yang , g. salamo , j. s. aitchison , and y. silberberg , opt . exp . * 13 * , 1797 ( 2005 ) ; y. linzon , y. sivan , b. malomed , m. zaezjev , r. morandotti , and s. bar - ad , phys .
lett . * 97 * , 193801 ( 2006 ) .
s. minardi , f. eilenberger , y. v. kartashov , a. szameit , u. rpke , j. kobelke , k. schuster , h. bartelt , s. nolte , l. torner , f. lederer , a. tnnermann , and t. pertsch , phys .
* 105 * , 263901 ( 2010 ) .
a. b. aceves , c. de angelis , g. g. luther , and a. m. rubenchik , opt .
19 , 1186 ( 1994 ) ; e. w. laedke , k. h. spatschek , s. k. turitsyn , and v. k. mezentsev , phys
e * 52 * , 5549 ( 1995 ) ; s. darmanyan , i. relke , and f. lederer , _ ibid_. * 55 * , 7662 ( 1997 ) ; a. b. aceves , m. santagiustina , and c. de angelis , j. opt . soc . am .
b * 14 , * 1807 ( 1997 ) ; i. relke , phys .
e * 57 * , 6105 ( 1998 ) ; m. stepi , l. hadievski , and m. m. skori , _
ibid_. * 65 * , 026604 ( 2002 ) ; a. v. yulin , d. v. skryabin , and a. g. vladimirov , opt . exp . * 14 * , 12347 ( 2006 ) . a. b. aceves , c. de angelis , a. m. rubenchik , and s. k. turitsyn , opt
* 19 * , 329 ( 1994 ) ; a. v. buryak and n. n. akhmediev , ieee j. quant . electr . * 31 * , 682 ( 1995 ) ; b. b. baizakov , b. a. malomed , and m. salerno , phys . rev .
a * 70 * , 053613 ( 2004 ) ; d. mihalache , d. mazilu , f. lederer , y. v. kartashov , l .- c .
crasovan , and l. torner , phys .
e * 70 * , 055603 ( 2004 ) .
d. mihalache , d. mazilu , f. lederer , and y. s. kivshar , opt .
express * 15 * , 589 ( 2007 ) ; d. mihalache , d. mazilu , y. s. kivshar , and f. lederer , opt .
express * 15 * , 10718 ( 2007 ) ; d. mihalache , d. mazilu , f. lederer , and y. s. kivshar , opt . lett . * 32 * , 3173 ( 2007 ) ; d. mihalache , d. mazilu , f. lederer , and y. s. kivshar , phys .
a * 79 * , 013811 ( 2009 ) ; d. mihalache , d. mazilu , f. lederer , and y. s. kivshar , opt .
commun . * 282 * , 3000 ( 2009 ) .
a. trombettoni and a. smerzi , phys .
* 86 * , 2353 ( 2001 ) ; f. kh .
abdullaev , b. b. baizakov , s. a. darmanyan , v. v. konotop , and m. salerno , phys .
a * 64 * , 043606 ( 2001 ) ; g. l. alfimov , p. g. kevrekidis , v. v. konotop , and m. salerno , phys .
e * 66 * , 046608 ( 2002 ) ; r. carretero - gonzlez , and k. promislow , phys .
a * 66 * , 033610 ( 2002 ) ; n. k. efremidis and d. n. christodoulides , _ ibid_. * 67 * , 063608 ( 2003 ) . | by means of a systematic numerical analysis , we demonstrate that hexagonal lattices of parallel linearly - coupled waveguides , with the intrinsic cubic self - focusing nonlinearity , give rise to three species of stable semi - discrete complexes ( which are continuous in the longitudinal direction ) , with embedded vorticity @xmath0 : triangular modes with @xmath1 , hexagonal ones with @xmath2 , both centered around an empty central core , and compact triangles with @xmath1 , which do not not include the empty site .
collisions between stable triangular vortices are studied too .
these waveguiding lattices can be realized in optics and bec . |
the object catalogued as sdss j123813.73 - 033933.0 ( hereafter sdss1238 ) was identified with a faint ( @xmath2 mag ) short - period cataclysmic variable ( cv ) by @xcite .
the optical spectrum of sdss1238 shows a blue continuum with broad absorption features originating in the photosphere of a white dwarf surrounding double - peaked balmer emission lines , formed in a high inclination accretion disk .
the orbital period of the system is @xmath3 , based on spectroscopic data . the orbital period and the spectral features match those of wz - sge - type systems , but with this
, similarities practically end .
a number of observed aspects of the system differ from the majority of short period cvs .
the most intriguing characteristic , which we found in this system , is a sudden and fast rise in brightness up to @xmath4 mag during a short time , of about half of the orbital period . after reaching its peak , the brightness slowly decreases , lasting @xmath5 hours , down to the quiescence level .
we call these events _ brightenings _ in order to distinguish them from the more common outbursts , humps , flickering and other types of variability documented in short period cvs
. these _ brightenings _ seemed to happen cyclically about every 8 - 12hours .
in addition to _ brightenings _ , a nearly permanent sinusoidal variability was detected in the light curve of sdss1238 with a period half that of the spectroscopic orbital period @xmath6 ( hereafter , the double - humped light curve ) .
the amplitude of the double - hump variability depends on the phase of the _ brightenings_. it increases with a total rise in brightness of up to @xmath7 mag and decreases until almost disappears during the quiescence .
a similar behavior was found later by @xcite in another short period cv , sdssj080434.20 + 510349.2 ( hereafter sdss0804 ) which has an identical spectral appearance to sdss1238 in quiescence . advanced the hypothesis that the double - humped light curve is a signature of 2:1 resonance in the accretion disks of these systems . in order for the accretion disk to reach permanently the 2:1 resonance radius
, the mass ratio of the binary component must be extreme ( @xmath8 ) , and as such these objects could qualify as bounce - back systems , e.g. cvs , which are old enough to reach the period minimum and leap toward slightly longer orbital periods , as predicted by @xcite .
it is supposed that accretion disks of wz sge systems reach 2:1 resonance radius during super - outburst , when some of them have been noted to show double - humped light curves @xcite .
the super - outbursts of wz sge type systems are infrequent and happen every two dozen or more years .
oddly enough , sdss0804 went into the super - outburst in 2006 @xcite and exhibited all necessary attributes of a classical wz sge - type object . regrettably , the _ brightenings _ disappeared from the light curves of sdss0804 after the super - outburst , although the double - hump light curve persists .
thus , the sdss1238 remains the only object that still shows _
brightenings_. intrigued by the new photometric phenomenon observed in these two systems , we conducted a new time - resolved photometric study of sdss1238 to establish the reasons behind their common nature , understand the origin of the cyclic _ brightenings _ and its relation to the amplitude of the double - humped light curve .
meanwhile , we discovered that sdss1238 was marginally detected as an infrared source by 2mass survey , and secured accurate near - ir photometry of the object . in sect.2
we describe our observations and data reductions .
the data analyses and the results are presented in sects.3,4,5 , while a general discussion and conclusions are given in sect.6 .
the object is listed in the 2mass point source catalogue with @xmath9 , @xmath10 , @xmath11 magnitudes .
these magnitudes are close to the detection limits of 2mass , particularly in the @xmath12-band . in order have more accurate photometry , we obtained new observations of sdssj123813.73 - 033933.0 in @xmath13 on 17 june 2009 with the near - infrared camera panic ( martini et al .
2004 ) attached to the 6.5 m baade / magellan telescope at las campanas observatory ( lco ) .
panic provides an image scale of 0.125@xmath14 pixel@xmath15 on a hawaii hgcdte @xmath16 array detector .
the fwhm of the point - spread function was between @xmath17 and @xmath18 during our observations .
for each filter , 9 dithered frames spaced by @xmath19 were taken , with total on - source integration times of 540 s in each of the three filters .
the nine frames were shifted and averaged to produce the final images .
standard sky - subtraction and flat - field correction procedures were applied .
aperture ( @xmath20 ) photometry was performed with daophot within iraf in the standard way .
flux calibration was performed using standard stars sj 9146 and sj 9157 from the list of @xcite and the total errors are estimated to be less than 0.05 magnitudes .
the resulting magnitudes of sdss1238 are @xmath21 , @xmath22 , @xmath23 and corresponding colors @xmath24 , @xmath25 and @xmath26 . in order to investigate whether flux variations in the near - ir occur in this system in timescales of a few minutes , we measured the @xmath27 , @xmath28 and @xmath29 fluxes from each of the nine short - exposure ( 60 s ) frames in each filter . in the @xmath30min that lasted each series , we did not detect any variability within the @xmath31 mag photometric uncertainty associated with each single frame .
note , that in longer timescales ( comparable to the orbital period ) , the system is expected to show some variability in the near - ir , mostly due to the elliptical shape of roche - lobe filling secondary , but these could be missed in a 15 min time series .
the ellipsoidal variability of the secondary can be calculated and it has been taken into account in further considerations of ir magnitudes . the objective of our optical photometry of sdss1238 was to study of phenomenon of _
brightenings_. taking into account the long duration of the _ brightenings _ and the uncertainty of the cycle period , we planned and executed a multi - longitude observational campaign of this object .
time - resolved ccd photometry was obtained at several facilities : the 1.5 m telescope at the observatorio astronmico nacional at san pedro mrtir in mexico ; the 0.8 m iac80 telescope at the observatorio del teide in the canary islands , spain ; the 2.1 m telescope at the bohyunsan optical astronomy observatory ( boao ) in south korea ; the 2 m telescope at the terskol astrophysical observatory in the northern caucasus , russia , the 2 m himalayan chandra telescope of the indian astronomical observatory ( iao ) , hanle , india .
the data reduction was performed using both eso - midas and iraf software .
the images were bias - corrected and flat - fielded before aperture photometry was carried out .
the log of photometric observations is presented in table.1 .
the long - slit observations have been obtained with the boller & chivens spectrograph on the 2.1-m telescope at the spm site with a resolution of 3.03 pixel@xmath15 . the spectra span the wavelength range 4000 - 7100 . in order to improve the signal - to - noise ratio
, we obtained a series of phase - locked spectra : 10 spectra were taken at equal phase intervals over a single orbital period @xmath32min with an exposure time of 486 sec per spectrum .
this sequence of spectra was repeated at exactly the same phase intervals for subsequent periods and subsequent nights .
this allows us to calculate the phase - averaged spectra , summarizing the spectra of the same orbital phase obtained during one night and the whole set of observations without further decreasing the time resolution .
the log of spectroscopic observations is presented in table.2 .
the spectral energy distribution ( sed ) of the object in the range of 4000 - 25000 is shown in fig.[fig:1 ] .
the detailed description of the sdss1238 optical spectrum was given in our previous paper .
the overall appearance of the spectrum has not changed , but we detect a significant variability of the equivalent widths of balmer emission lines from epoch to epoch .
the balmer lines are about two times weaker in the 2009 observations compared to the 2004 spectra ( fig.[fig:2 ] ) .
the continuum , however , has not changed during the last five years , as we compare v - band magnitudes in quiescence between _ brightenings_. the average quiescence magnitude between _ brightenings _ remains constant at around @xmath33 .
the present near - ir measurements demonstrated that there is significant ir excess emission to that expected from the rayleigh
jeans tail of the optical spectrum for a white dwarf ( wd ) .
in fact , neither a wd nor a power - law flux from the accretion disk , nor their combination , can explain the observed ir excess .
the most probable source of ir excess is the radiation from the secondary star . to determine the spectral type of the secondary and the distance to the system we fitted the observed optical - infrared spectral energy distribution of the object with a simple model : the total flux @xmath34 is the sum of contributions from a wd with a hydrogen atmosphere , @xmath35 ( da type wds ) , an accretion disk with @xmath36 @xcite , and a red / brown dwarf with @xmath37 : @xmath38 brown dwarf fluxes were taken from the literature @xcite and on - line sources
the white dwarf spectra with a mass range of m@xmath39m@xmath40 were used with a 0.1m@xmath40 step and the radii were calculated using the white dwarf radius - mass relation of @xmath41 from @xcite and @xcite .
spectra of wds in the 4000 - 25000 range with pure hydrogen atmosphere were obtained using atlas9 @xcite and synth @xcite codes for an appropriate range of temperatures .
although our previous temperature estimate was @xmath42k based on fits to the absorption portion of balmer lines , in the present fitting procedure we allowed a wider temperature range from @xmath4311000 to 18000k , because of the larger number of free parameters .
the calculations were performed with a 1000k step and with the surface gravity @xmath44 .
the spectra are normalized to @xmath45 and the contribution of the wd is @xmath46 where @xmath47 , v = 17.8 is the object s magnitude in quiescence , and @xmath48 is a parameter , determining the contribution of the flux from the wd in the v band .
finally , the @xmath49 - is the constant to convert magnitudes into flux ( in @xmath50 ) in the v band .
the spectra of the accretion disk was assumed to be a simple power law @xmath51 where @xmath52 determines the contribution from the accretion disk in the v band , assuming that the wd and the accretion disk are the only contributors in that wavelength as the only other contributor is the brown dwarf , which has a negligible flux in v. the distance to the object is estimated to be @xmath53 where @xmath54 is the black body flux at @xmath55 with effective temperature @xmath56 .
observed seds of red / brown dwarfs of spectral types between m6 to l5 normalized to fit the observed flux in j were used .
the bolometric correction to the j magnitude for each spectral type was taken from @xcite .
the free parameters of the three - component model are : the white dwarf effective temperature , @xmath56 , the mass of the white dwarf , @xmath57 , the spectral type of the secondary star , spt , and the parameter @xmath48 ( in magnitudes ) .
the best fit model to the observed sdss1238 spectrum in the 0.4 - 2.5@xmath58 m range achieved for the following set of parameters ( fig.[fig:1 ] ) is : @xmath59 , spt = l4 , @xmath60 and @xmath61 .
the deduced distance to the object is 110pc .
we studied the behavior of @xmath62 vs a single fit parameter , when the other three are fixed to their corresponding best values .
fig.[fig:3 ] presents @xmath62 plots for various parameters with marked confidence levels corresponding to 95 , 80 and 60% .
we can state that @xmath62 tends to the minimum value always , when @xmath63k and @xmath64 @xmath65 regardless of the value of the other parameters . at the same time
, there is a dependence between the mass of the wd and the spectral type of the secondary : the lower the mass , the larger the radius of wd , and thus , the larger the distance to the system , resulting in an earlier spectral type of the secondary .
the best fit to the optical part of the spectrum is reached with the mass of the wd of @xmath66 , leading to the cited distance of , 110pc , and spectral type of the secondary , l4 . the entire range of secondary from @xmath67 at @xmath68 and @xmath69pc to @xmath70 at @xmath71 ,
@xmath72pc was considered .
however , a pronounced minimum in @xmath62 for rather massive wd strongly suggests the presence of brown dwarf in this system .
the indication of a massive wd is not accidental , a cv with a brown dwarf secondary , supposed to have evolved beyond minimum orbital period limit , has an age @xmath73 years and a long history of accretion . a priori ,
the bounce - back cvs are assumed to harbor a massive wd , although the distribution of known masses of all wds in eclipsing systems does not show any trend @xcite .
but when considering only the short - period end of that distribution as @xcite did , then it becomes apparent that evolved systems systematically have more massive wds .
the upper limit for the mass of the secondary is @xmath74 for @xmath75 or @xmath76 for @xmath77 @xcite .
but it is also a well known fact that the secondaries of cvs tend to show systematically an earlier spectral class and a larger radius with secondaries of lower mass than the corresponding single stars on the main sequence @xcite .
so we expect that the mass of the secondary in sdss1238 hovers around the lower edge of the above mentioned range of masses .
note , that due the high galactic latitude of this object ( @xmath78 ) and the inferred small distance , the interstellar extinction is negligible .
we also would like to emphasize that in the above calculations we took into account the fact that the state of the system was unknown at the moment of acquiring the ir data .
therefore , we conducted the fitting for both cases , considering that the system might have been at the maximum of the _ brightenings _ during the ir observations or at the bottom . that introduced only a minor change , basically decreasing the distance to the system by @xmath30pc and not affecting our conclusions regarding the spectral type of the secondary .
finally , we would like to comment on the discrepancy in the temperature determination of the wd , which is larger than we would like as compared to that of . in the latter the temperature and gravity
could not be determined simultaneously , and one parameter had to be fixed in order to calculate the other , which always introduces ambiguity , as none of these two parameters could be estimated independently . in this paper
we take a more complex approach : not only profiles of the lines are being fitted , but the continuum is taken into consideration as well . also , the presence of the secondary , adding additional restraints on the distance and thus , on the size of the wd .
the sed of the accretion disk , which may not necessarily obey the canonical power law index is the source of the largest uncertainty in our analysis .
since the contribution of the disk is limited to only @xmath79 % , so is the accuracy of our estimates . with all that in mind
, we still end up with a range of spectral classes for the secondary that implies that sdss1238 is a bounce - back system .
the h@xmath80 emission line originating in the accretion disk is the least affected by the absorption from the underlying white dwarf .
therefore , we constructed doppler maps @xcite of h@xmath80 using all our data obtained in 2004 and 2009 ( fig.[fig : dop ] , top ) .
the individual maps of separate nights resemble each other so much , that combining all available data did not smooth out details , but made them much more convincing . in order to over - plot contours of the secondary star , the location of the wd , the trajectory of the stream and resonance radius of the disk
, we used the best fit parameters to the sed ( see previous section ) .
thus , the white dwarf mass of @xmath81 was adopted . due to the ambiguity in determining precise masses of brown dwarfs
, we adopted a mass ratio of @xmath82 .
this value is typical for systems considered as bounce - back @xcite and it is the same mass ratio which was obtained for sdss0804 , a twin of sdss1238 , from the super - hump period observed during the super - outburst in 2006 . before sdss0804 underwent a wz - sge type super - outburst , it showed similar peculiar photometric variability to sdss1238 .
in addition , it shares every other characteristic of a wz sge - type object , and was also proposed as a candidate to the bounce back system .
as already noted , the structure of the accretion disk did not change between the two epochs of our observations .
there is a bright spot at the expected place where the stream of matter from the secondary collides with the accretion disk , but it overlaps with a much larger and prolonged structure , too extended to be a part of the spot .
another extended bright region of similar size can be seen at velocity coordinates ( @xmath83 km s@xmath15 , @xmath84 km s@xmath15 ) as well as a less bright structure at ( @xmath85 km s@xmath15 , @xmath86 km s@xmath15 ) .
similar doppler maps were obtained for wz sge during super - outburst in 2001 and in quiescence for bounce back candidates sdss1035 @xcite and sdss0804 @xcite .
such brightness distribution in the doppler map can be interpreted as evidence of spiral waves in the disk ( see for example , @xcite and @xcite and reference therein ) .
the formation of a spiral structure in an accretion disk of a close binary system was predicted by @xcite and explored by various authors ( and reference therein ) .
@xcite demonstrated from high resolution numerical calculations that spirals will always formed in accretion disks under tidal forces from the secondary .
they actually used @xmath87 in their models , but observationally , such spirals were detected in a number of systems only during outbursts of dwarf novae .
the careful examination of quiescent disks of the same systems did not reveal any spiral structures in longer period dne .
@xcite argued that little evidence of spiral arms in the emission lines is expected in systems with low values of viscosity .
on the other hand , spiral arms related to 2:1 resonance can be found in systems with extremely low mass ratio @xmath88 as originally was predicted by @xcite .
the bounce - back systems and related to them , wzsge stars , are examples of such objects . the long outburst recurrence time in wzsge systems
is probably explained by a very low viscosity in their accretion disks , yet spiral arms can be observed permanently in quiescent bounce - back systems in which , on one side there is a massive wd , which gained mass during a long accretion history , and on another side there is a late - type brown dwarf , providing a mass ratio of @xmath89 .
fig.[fig : dop ] ( bottom , left ) depicts a synthetic doppler map constructed from a model accretion disk that is shown on the bottom right panel .
the latter was calculated with the binary system parameters derived above by using smooth particle hydrodynamics according to @xcite .
the artificial doppler map reproduces the observed map in a case when there is a brightness excess within spiral arms .
most of the disk particles are on periodic orbits , which are most favorable from the point - of - view of viscosity .
however , the resonance dispatches some particles onto aperiodic orbits creating viscosity perturbations , which will create excess of heat .
a slightly different interpretation of spiral arm brightness is offered by @xcite .
the mechanism is not very well established , but it is natural to assume that in these regions there will be excess emission .
figure [ fig : phot ] displays the light curve of sdss1238 obtained in 2007 . in general , the object shows an identical behavior to previous years , as described by .
there are two distinct types of variability : a long - term variability ( ltv ) , lasting more than 8h , and a short - term variability with a period corresponding to half the orbital period . follow - up observations during 2008 and 2009 confirm a steady presence of both types of variability in the light curve ( fig.[fig : photnew ] ) .
continuous observations during about 15 hours obtained on 22/03/2007 ( hjd 2454182.15 - 2544182.60 , fig.[fig : phocont ] ) allowed us to observe two consecutive _ brightenings _ and , thus to determine their recurrence time directly ( @xmath90h ) . demonstrated that _
brightenings _ occurred cyclically with periods in between 8 and 12 hours , however it was not possible to establish if the phenomenon was strictly periodic or not .
armed with more data , and the advantage of detecting consecutive events by employing multi - longitude observations , we performed a simple period search .
the period analysis of all data obtained during 2007 , based on a discrete fourier transform method , results in a strong peak at a frequency @xmath91 day@xmath15 with fwhm of 0.1 .
this conforms with the period of @xmath920.36h @xmath93 ( fig.[fig : ltv1 ] ) .
however , the data folded with this period looks messy because , apparently , some re - brightenings happen with a different cycle period .
it shows that the large uncertainty in the period value is not just a result of a scarce amount of data , or its uneven distribution , but that the _ brightenings _ are not strictly periodic in long timescales .
the peak @xmath94 day@xmath15 , corresponding to the half of the orbital period , @xmath95 min , is also present in the power spectrum .
there are some additional peaks which are one - day aliases or a combination of high harmonics of @xmath96 with @xmath97 .
we have repeated the same analysis for the data obtained only within hjd 2454177.1 - 2544185.6 , when we observed the object with small observational gaps .
the result of the analysis of periodicity for this selection is presented in fig.[fig : ltv2 ] .
the strongest peak in the power spectrum appears at a frequency corresponding to @xmath98h .
the light curve folded with this period is coherent during the considered time ( fig.[fig : ltv2 ] ) , i.e macro - profile of the _ brightenings _ ( the rise , the peak , the declining slope and minimum duration ) repeats itself with high accuracy and , if not for the high - frequency modulation also present in the light curve , the folded curves would concur .
combining the entire available data obtained during the 2004 2009 observational runs does not permit a determination of a single period with which the entire data set can be folded . clearly , the ltv modulation is not strictly periodic , but probably maintains coherence in a time scale of several weeks and certainly has a cyclic nature . the long - term variability ( _ the brightenings _ ) presents a fast rise in the object brightness of up to 0.4 - 0.6 mag , lasting about half the orbital period ( @xmath99h ) with the brightness falling back to the quiescence level during the next 4.4 - 4.6h ( fig . [
fig : phocont ] ) .
after that , the object remains in quiescence during about another 4 hours .
the brightness increases at a rate of @xmath100 mag / h , and the falling rate is about @xmath101 mag / h .
sometimes , the rise in the brightness happens more slowly and lasts until almost one full orbital period ( fig.[fig : phot ] nights : 156 , 179 ) .
a short - term variability is clearly present during _
the brightenings_. the amplitude of the short - term variability depends strongly on the phase of _ brightenings_. the amplitude is larger then the total brightness increases and falls , sometimes practically disappearing , between the _ brightenings _ ( figs .
[ fig : phot ] , [ fig : photnew ] , [ fig : phocont ] )
. curiously , when the ltv is coherent , the signal at double orbital frequency decreases and a beat frequency between the ltv and @xmath97 becomes dominant in the range of frequencies around 36 cycles / day .
removing ltv has little effect , because this does affect the frequency modulation but not the amplitude modulation .
the strong relation between ltv cycle coherence and the strength of @xmath97 modulation attests that these two phenomena are physically linked in more than a simple amplitude correlation .
we suspect that the _ brightenings _ and formation of spiral arms in the disk are the result of the same process , but its cyclic nature remains unclear .
in recent years , a number of objects were discovered , mostly thanks to the sloan digital sky survey , which can be characterized as wz sge type cvs in quiescence , judging by their @xmath102 min periods , spectral appearance and lack of su uma style outburst activity . currently , there are about 20 objects with similar characteristics which were selected from the sdss lists .
three such objects are gw lib , v455 and and sdss0804 , which underwent a typical wzsge type super - outburst in 2006 - 2007 , confirming the correct assessment of their nature from their behavior in quiescence . but unlike classical wz sge stars , some of them differ by the presence of permanent double - humps per orbital period in the light curve .
the significance of the double - humped light curve lies in the assumption that it is produced by the spiral structure in the accretion disk .
the spiral arms in accretion disks have been detected before by means of doppler tomography in long period systems only during dn outburst .
we are not aware of any reports of double - humped light curves during these detections .
double - humped light curves have previously came to attention , as they were detected during super - outbursts of wz sge .
it was soon suggested that they are result of 2:1 resonance , as one of the possibilities .
also spirals arms are formed in the accretion disk of the wz sge undergoing super - outburst @xcite , because , according to our hypothesis wz sge accretion disk reaches the resonance radius only during super - outburst .
nevertheless , the spiral structure has been observed persistently along with the double - humped light curve in two similar short - period systems , sdss0804 and sdss1238 in quiescence .
we assume that the spiral structure formed in low viscosity , quiescent disks is a result of distortion of the disk by the 2:1 resonance .
the 2:1 resonance may happen only in the accretion disk of a system with extreme mass ratio of @xmath103 .
such mass ratio is achieved only in systems known as bounce - backed or , in other words , systems which have reached a minimum period of @xmath104 min and have turned to slightly longer periods according to @xcite .
it is expected that bounce - back systems are numerous @xcite , but until recently very few candidates have been found .
the sdss helped to uncover a large number of new cvs , with new interesting features @xcite . among them , there is a number of short period systems , some of which turned out to belong to the long - sought bounce - back systems @xcite .
we have unveiled a brown dwarf secondary in sdss1238 , probably as late as l4 , thus providing strong evidence that this object is a real bounce - back system .
we also estimated parameters of the wd .
the best fit to the sed converges if the primary is a massive @xmath105 white dwarf with a @xmath106k temperature .
both numbers seem plausible since the system is very old and the wd is expected to be relatively massive and cool .
these findings provide further support to our claim that by simply observing a permanent double humped light curves one can identify an evolved , bounce - back systems instead of using other complicated methods .
it is important to note that sdss1238 has another peculiarity , which termed as _
brightenings_. there was only one other system known to exhibit _ brightenings _ , namely sdss0804 , but since it underwent a super - outburst in 2006 , its photometric behavior has drastically changed .
the _ brightenings _ shortly detected by @xcite in sdss0804 before the super - outburst have been replaced by : a ) _ the mini - outburst _ activity with permanent presence of the double - humped light curve of constant amplitude and by : b ) a 12.6 min period , probably corresponding to pulsation activity of the wd @xcite .
therefore , at present , sdss1238 is the only object known to show _ brightenings_. it greatly complicates the study of this phenomenon .
based on new multi - longitude continuous monitoring , we demonstrated here that the _ brightenings _ are of cyclic nature with a recurrence time of @xmath107 hours and they are probably coherent over several cycles . there seem to be strong correlation between _ brightenings _ cycles and the amplitude of double - hump periodic variability .
this is believed to be the result of spiral arms in the accretion disk of bounce - back systems with an extreme mass ratio , in which accretion disk extends beyond a 2:1 resonance radius .
the tomogram of a simulated accretion disk in the regime of resonance , closely resembles the observed one and supports this hypothesis .
the formation of a spiral structure in the disk can be accounted by the appearance of double humps in the light curve , but it can not be the reason of increased brightness of the disk .
the disk brightness directly depends on the mass transfer rate and its change should probably reflect change in mass transfer .
no readily explanation is available as to why that rate can be variable and cyclical , but possible speculations include counteraction : a ) to the heating of the secondary by _ brightenings _ , or b ) to tidal interaction between the secondary with the resonance attaining accretion disk .
howell , s. b. , adamson , a. , & steeghs , d. 2003 , , 399 , 219 knigge , c. 2006 , , 373 , 484 kolb , u. , & baraffe , i. 1999 , , 309 , 1034 kunze , s. , speith , r. , & riffert , h. 1997 , , 289 , 889 kunze , s. , & speith , r. 2005 , the astrophysics of cataclysmic variables and related objects , 330 , 389 kurucz , r. 1993 , atlas9 stellar atmosphere programs and 2 km / s grid .
kurucz cd - rom no .
cambridge , mass . : smithsonian astrophysical observatory , 1993 . , 13 ogilvie , g. i. 2002 , , 330 , 937 paczynski , b. 1981 , acta astronomica , 31 , 1 patterson , j. , et al .
2002 , , 114 , 721 patterson , j. , et al .
2002 , , 114 , 1364 pavlenko , e. , et al . 2007 , 15th european workshop on white dwarfs , 372 , 511 llllccc date & hjd start+ & telescope & exp.time & duration + & 2454000 & & num .
of integrations & & + photometry & & & & & & + 25 feb .
2007 & 156.832 & 1.5m / spm / mxico & 120s@xmath108160 & 5.3h + 26 feb .
2007 & 157.845 & 1.5m / spm / mxico & 120s@xmath108144 & 4.8h + 15 mar .
2007 & 174.756 & 1.5m / spm / mxico & 220s@xmath10895 & 5.8h + 18 mar .
2007 & 177.767 & 1.5m / spm / mxico & 170s@xmath10891 & 4.3h + 19 mar .
2007 & 178.716 & 1.5m / spm / mxico & 240s@xmath108108 & 7.2h + 20 mar .
2007 & 179.414 & 2m / terksol / russia & 120s@xmath108118 & 3.9h + 20 mar .
2007 & 179.746 & 1.5m / spm / mxico & 160s@xmath10877 & 3.4h + 20 mar .
2007 & 180.290 & 2m / terksol / russia & 120s@xmath108152 & 5.1h + 21 mar .
2007 & 181.169 & 2m / iao / india & 120s@xmath108135 & 7.5h + 21 mar .
2007 & 181.286 & 2m / terksol / russia & 120s@xmath108210 & 7h + 22 mar .
2007 & 182.164 & 2m / iao / india & 120s@xmath108198 & 6.9h + 22 mar . 2007 & 182.302 & 2m / terksol / russia & 120s@xmath108195 & 6.5h + 22 mar .
2007 & 182.435 & 0.8m.iac80/spain & 270s@xmath10898 & 7.4h + 25 mar .
2007 & 185.032 & 2.1m / boao / korea & 135s@xmath108107 & 4.0h + 25 apr . 2007 & 216.696 & 0.8m / iac80/spain & 200s@xmath108101 & 5.6h + 27 apr . 2007 & 218.648 & 0.8m / iac80/spain & 200s@xmath108116 & 6.4h + 09 mar . 2008 & 535.904
& 1.5m / spm / mxico & 60s@xmath108150 & 2.5h + 10 mar .
2008 & 536.882 & 1.5m / spm / mxico & 60s@xmath108200 & 3.3h + 11 mar .
2008 & 537.889 & 1.5m / spm / mxico & 60s@xmath108170 & 2.8h + 27 feb .
2009 & 890.826 & 1.5m / spm / mxico & 60s@xmath108184 & 3.1h + 28 feb .
2009 & 891.814 & 1.5m / spm / mxico & 60s@xmath108227 & 3.8h + 01 mar .
2009 & 892.851 & 1.5m / spm / mxico & 60s@xmath108218 & 3.6h + spectroscopy & & & & + 24 jan .
2009 & 855.99 & 2.1m / spm / mxico & 486s@xmath1087 & 0.52h + 25 jan .
2009 & 856.96 & 2.1m / spm / mxico & 486s@xmath10817 & 2.43h + 26 jan .
2009 & 857.90 & 2.1m / spm / mxico & 486s@xmath10820 & 2.69h + 27 jan .
2009 & 856.96 & 2.1m / spm / mxico & 486s@xmath10810 & 1.2h + | we present infrared jhk photometry of the cataclysmic variable sdssj123813.73 - 033933.0 and analyze it along with optical spectroscopy , demonstrating that the binary system is most probably comprised of a massive white dwarf with @xmath0k and a brown dwarf of spectral type l4 .
the inferred system parameters suggest that this system may have evolved beyond the orbital period minimum and is a bounce - back system .
sdssj123813.73 - 033933.0 stands out among cvs by exhibiting the cyclical variability that called _
brightenings_. these are not related to specific orbital phases of the binary system and are fainter than dwarf novae outbursts , that usually occur on longer timescales .
this phenomenon has not been observed extensively and , thus , is poorly understood .
the new time - resolved , multi - longitude photometric observations of sdssj123813.73 - 033933.0 allowed us to observe two consecutive _ brightenings _ and to determine their recurrence time .
the period analysis of all observed _ brightenings _ during 2007 suggests a typical timescale that is close to a period of @xmath1 hours .
however , the _ brightenings _ modulation is not strictly periodic , possibly maintaining coherence only on timescales of several weeks . the characteristic variability with double orbital frequency that clearly shows up during _ brightenings
_ is also analyzed .
the doppler mapping of the system shows the permanent presence of a spiral arm pattern in the accretion disk .
a simple model is presented to demonstrate that spiral arms in the velocity map appear at the location and phase corresponding to the 2:1 resonance radius and constitute themselves as a double - humped light curves .
the long - term and short - term variability of this cv is discussed together with the spiral arm structure of an accretion disk in the context of observational effects taking place in bounce - back systems . |
optical microcavities such as photonic crystal nanocavities @xcite , microring resonators @xcite , toroidal microcavities @xcite and crystalline whispering gallery mode ( wgm ) cavities @xcite are attractive platforms for studying the optical linear and nonlinear properties of light @xcite .
these cavities have been employed in various studies including work on slow light generation @xcite , an ultra - narrow linewidth laser @xcite , an optoelectronic oscillator @xcite , nonlinear switching @xcite , frequency - comb generation @xcite , opto - mechanics @xcite , frequency conversion @xcite , and nanoparticle sensing @xcite .
wgm cavities , which use internal reflection for light confinement , are particularly attractive because they can provide a very high quality factor ( @xmath0 ) .
recent progress on the @xmath0 factor in such cavities has been noteworthy . in particular , cavities made of @xmath1 such as toroidal @xcite and bottle @xcite cavities exhibit a very high-@xmath0 exceeding @xmath2 because @xmath1 is a low - loss material and the laser reflow process is well established . on the other hand ,
wgm cavities made of crystal materials also exhibit an ultrahigh-@xmath0 , because their absorption loss is even lower than that of @xmath3 @xcite .
crystalline materials are also attractive because they have various unique properties .
the large @xmath4 nonlinearity is essential for making building blocks for optoelectronic modulators @xcite .
it is transparent at infrared wavelengths , which is an interesting regime for gas sensing @xcite .
it also has a large young s modulus , which is attractive for opto - mechanics research @xcite . with the development of ultra - precise machining ,
the properties of micro - resonators made of crystal materials have improved , and they are now good candidates especially as a platform on which to study nonlinear optics .
however , a high-@xmath0 crystalline wgm cavity is still difficult to fabricate because the material is hard and frangible . in particular
, the fabrication of a small cavity remains a challenge .
previously , we reported on the fabrication of sapphire ( @xmath5 ) wgm cavities by employing a modified laser heated pedestal growth ( lhpg ) method to overcome this problem @xcite .
the lhpg method was originally developed for manufacturing fiber lasers @xcite .
it has been applied to various materials such as @xmath6 , @xmath7 and yag @xcite .
in contrast to earlier work , in which researchers tried to fabricate crystal rods with a thinner and uniform radius , we used this method to fabricate a rod with a locally modulated radius to enable light confinement in the longitudinal direction and excite the wgm .
changing the growth rate enabled us to fabricate a swollen region , in which we obtained a wgm with a @xmath0 of @xmath8 . in our work
, we demonstrate the fabrication of sapphire wgm cavities that have a hexagonal cross - section but with round edges .
the controllability of the cross - section ( the sharpness of the edges ) is the key to controlling the @xmath0 factor of such a cavity . in this paper , we analyze in detail various wg - like modes excited in hexagonal cavities with different corner curvatures . in particular , we study a quasi - wgm and a newly discovered perturbed - wgm . to the best of our knowledge ,
this is the first time a perturbed - wgm has been characterized . by understanding the property of the perturbed - wgm , and investigating the coupling between these two modes , we reveal the limit of the @xmath0 of the quasi - wgm and study a possible way of improving the value .
we also demonstrate the fabrication and optical measurement of a sapphire hexagonal wgm cavity that supports such a mode .
we show that a high @xmath0 cavity is obtained with a cavity with round corners , whose growth is carefully controlled to obtain good crystal quality .
the polygonal cavity analysis described in this paper is useful for cavities fabricated with lhpg , moreover it provides information needed to understand the optical properties of a polygonal toroidal microcavity @xcite , and microcavity lasers made of zno @xcite , @xmath9 @xcite , @xmath10 @xcite , @xmath11 @xcite , and @xmath12 @xcite , that are fabricated with the vapor phase transport method .
these microlasers have recently attracted great interest .
it is known that these microcavities have a polygonal cross - section as a result of their crystal structure , and various experimental and numerical studies has been conducted to understand the mode property of such cavities , especially in zno .
for example , wiersing @xcite , nobis and grundmann @xcite , and huang @xcite have analyzed the features of the modes in this unique cavity with triangular , hexagonal and circular cross - sections .
crystals were grown experimentally on zno with different deposition temperatures .
this enabled the growth of zno micro wires with different cross - sections , including hexagonal , dodecagonal , and circular @xcite .
al _ tried to improve the @xmath0 value by rounding the corners of polygons @xcite .
however , the reported @xmath0 of these microresonators remains on the order of several thousand , even though the excited modes , which they call a `` quasi - wgm '' , seem to satisfy the total internal reflection in a ray - optics framework @xcite .
since we have a same problem in our sapphire hexagonal cavity , the purpose of the first half of this paper is to provide a clear understanding of the limiting @xmath0 of this mode in a polygonal cavity .
this paper is organized as follows . in sec .
[ sec2 ] we introduce the structure of the cavity we fabricated for investigation .
we investigate the properties of the perturbed - wgm and quasi - wgm , paying particular attention to the influence of the coupling between these two modes .
then in sec .
iii , we describe the fabrication and optical measurement of this cavity .
we conclude the paper in sec .
before analyzing a hexagonal cavity with round corners , we show side and cross - sectional views of our hexagonal sapphire cavity fabricated by lhpg in fig .
[ fig:16 ] .
we describe the fabrication and optical measurement of this cavity in detail in sec .
[ sec3 ] . as shown in fig .
[ fig:16](a ) the radius of the grown rod is locally modulated .
the light is confined in an area where the circumference is the largest , as shown in fig .
[ fig:16](b ) .
the typical mode volume is @xmath13 .
figure [ fig:16](c ) shows an optical microscope image of the cross - section of this cavity .
the shape is hexagonal as a result of the crystal structure of the sapphire .
however , the corners of the cavity are not sharp but round , and their curvature radius can be controlled by changing the growth condition .
this cavity is on interest of mode analysis .
figure [ fig:0 ] shows the ray - optics description of the possible modes in a hexagonal cavity with slightly rounded corners .
fabry - perot ( fp ) modes ( fig .
[ fig:0](a ) and ( b ) ) are excited by reflection at the opposing disk facets or at the hexagonal facets @xcite .
figure [ fig:0](c ) and ( d ) show `` quasi - wgms '' in which light is reflected at the side of the polygon .
since these modes satisfy the total internal reflection in a ray - optics framework , the quasi - wgm appears to exhibit an ultrahigh-@xmath0 .
although , the quasi - wgm has been studied in detail in previous work @xcite , we will show below that the discovery of a different wgm , which is shown in fig .
[ fig:0](e ) , is essential if we are to understand the low-@xmath0 of the quasi - wgm .
we call this mode a `` perturbed - wgm '' @xcite , because this is an ultrahigh-@xmath0 wgm when the cavity is round . a complete understanding of these modes is the main aim of this paper .
in addition , we show that a different mode , shown in fig .
[ fig:0](f ) , which has usually been categorized as a quasi - wgm , originates from a higher order wgm . to investigate these modes in detail
we perform a two - dimensional finite - difference time - domain ( fdtd ) calculation .
figure [ fig:1](a ) shows the structure that we used for the calculation .
it is a hexagonal cavity with round corners that have curvature radiuses @xmath14 .
when the microcavity is a perfect circle , an ideal wgm is excited as shown in fig .
[ fig:1](b ) .
next we change @xmath14 and make the cavity slightly polygonal .
then we obtain an optical mode as shown in fig .
[ fig:1](c ) .
this is the quasi - wgm shown in fig .
[ fig:0](d ) .
the light propagates while being reflected at the side of the cavity .
this simple picture leads us to expect that the resonant wavelength and the @xmath0 of the quasi - wgm do not depend strongly on the edge radius @xmath14 .
in contrast , we obtained a different mode in a hexagonal cavity for the same geometry as shown in fig .
[ fig:1](d ) .
this is the `` perturbed - wgm '' .
the original circular wgm is perturbed as a result of the modulation of the structure from circular to hexagonal .
but it retains the wgm characteristics , where the light propagates close to the surface .
note that if we look very carefully , we find that the light of this mode propagates close to the surface at the corner , but it propagates slightly inwards at the side @xcite .
m and the edge radius is @xmath14 .
( b ) mode profile of a wgm when the shape is a perfect circle ( @xmath15 ) .
( c ) mode profile of quasi - wgm when @xmath16 .
( d ) mode profile of a perturbed - wgm when @xmath16.,width=240 ] in contrast to the quasi - wgm , with the perturbed - wgm we can straightforwardly predict that the resonant wavelength and @xmath0 value change sensitively according to the corner radius @xmath14 . since this mode exhibits an ultrahigh-@xmath0 when the corner is round , we expect the @xmath0 of the perturbed - wgm to be higher than that of the quasi - wgm when @xmath14 is large .
however , the @xmath14 dependence of the @xmath0 is larger for the perturbed - wgm , therefore , the @xmath0 of the quasi - wgm should become higher when the corner is sharp and the cavity is hexagonal .
but , this simple description is insufficient as we discuss in the following sections . .
( b ) spectrum of the excited modes of a hexagonal cavity when @xmath17.,width=288 ] figure [ fig:2](a ) shows the resonance wavelength of the perturbed and quasi wgms with respect to the @xmath14 calculated using the fdtd method . as expected , the resonant wavelength changes much sensitive to the @xmath14 for the perturbed - wgm than for the quasi - wgm . but most importantly , we know from fig .
[ fig:2](a ) that the wgm of the circular cavity shown in fig . [ fig:1](b ) is on the same slope as the perturbed wgm in fig .
[ fig:1](d ) .
this indicates that the mode shown in fig . [
fig:1](d ) is indeed a `` perturbed '' wgm , and the quasi - wgm originates from a different mode from an ultrahigh-@xmath0 wgm when the cavity is circular . .
( b - e ) mode profiles of the cavity .
the corresponding resonance points are shown in ( a).,width=288 ] next , we investigate the resonance in more detail .
figure [ fig:2](b ) shows the resonant spectrum when @xmath18 .
this spectrum shows that not only two modes , but also various other different modes are excited in this hexagonal cavity . to explain the property of the hexagonal cavity
, we need to understand these modes .
since it is difficult to characterize the property of this number of modes from this single spectrum , we need to perform calculations with different @xmath14 .
the resonant peaks for cavities with different @xmath14 values are shown in fig .
[ fig:3](a ) .
the plots for the quasi and perturbed wgms shown in figs .
[ fig:1](c ) and ( d ) are indicated in the panel .
however , a number of different peaks appear in addition to these two modes .
although the behavior of these modes appears very complicated , they can be classified into two different groups .
we clearly obtained two different slopes .
figure [ fig:3](b ) shows a quasi - wgm but with a different longitudinal mode number .
similar resonance also appears at an 11.6 nm longer wavelength , where the interval is the free - spectral range ( fsr ) of this mode .
indeed , we obtained a similar fsr value from a simple ray - optics estimation .
similarly , fig .
[ fig:3](c ) shows a perturbed - wgm with a different longitudinal mode number .
the fsr for the perturbed - wgm is 11.3 nm
. however , fig .
[ fig:3](c ) shows that there are a number of different modes with a similar @xmath14 dependence to that of the perturbed - wgm in the hexagonal cavity .
one of the modes is shown in fig .
[ fig:3](d ) . in this mode
, the light propagates not only at the surface of the cavity , but also at the inner part of the structure .
this propagation occurs because the light reflects at the cavity wall with a large incident angle .
the light propagation in this mode is described in terms of ray - optics in fig .
[ fig:0](f ) , where the mode was originally classified as a quasi - wgm . however , the @xmath14 dependence of this mode in fig .
[ fig:3](b ) suggests that its behavior is similar to that of the perturbed - wgm .
indeed , when we examine the profile at @xmath19 for this mode , namely when the cavity is circular , we observe a higher order wgm as shown in fig .
[ fig:3](e ) .
hence , we can conclude that the mode shown in fig.5(d ) is the perturbed mode of the higher order wgm .
it is important to understand this picture , because it indicates the difficulty of obtaining a high-@xmath0 for the mode shown in fig .
[ fig:3](e ) , even though the ray - optics image in fig . [ fig:0](f ) gives the impression that this mode may exhibit an ultrahigh @xmath0 even when the corner is sharp . since the quasi and perturbed wgms ( fig .
[ fig:1](c ) and ( d ) ) are the fundamental modes of this hexagonal cavity , we now focus on these two modes and analyze the @xmath0 .
figure [ fig:4 ] shows the @xmath0s of these two modes with different @xmath14 values .
contrary to our initial expectation , the quasi - wgm exhibits a lower @xmath0 when the cavity is polygonal , although fig .
[ fig:0](d ) suggests that the @xmath0 is only slightly dependent on @xmath14 . of perturbed and quasi wgms for hexagonal cavity with different curvature radius @xmath14 values.,width=288 ] to understand this behavior , we pay close attention to the wavelength crossing between the quasi - wgm and the perturbed - wgm .
figure [ fig:6](a ) and ( b ) show enhanced views of the crossing points between the perturbed and quasi - wgms in fig .
[ fig:3](a ) .
the figures show that these two modes exhibit anti - crossing behavior , which indicates the existence of strong mode coupling between those two modes .
we can derive the coupling coefficient @xmath20 from the spectrum splitting width , and we obtained @xmath21 for fig .
[ fig:6](a ) and @xmath22 for fig .
[ fig:6](b ) .
this explains the low @xmath0 property of the quasi - wgm when the corners are sharp . according to the ray - optics description shown in fig .
[ fig:0](d ) , the @xmath0 of the quasi - wgm does not depend on the edge radius @xmath14 of the hexagonal cavity and should exhibit an ultrahigh @xmath0 . however , the coupling between the quasi- and perturbed - wgms needs to be taken into account .
this coupling is particularly important when @xmath14 is small .
coupling between modes , for example @xmath23 , which corresponds to a @xmath0 of about @xmath24 , means that the light energy of the high-@xmath0 quasi - wgm transfers to a low-@xmath0 perturbed - wgm and then easily couples to the out - of- cavity radiation .
therefore , the @xmath0 of the quasi - wgm is not dependent on the @xmath0 of the perturbed - wgm and the values of those two modes become very close . since the @xmath0 value of the perturbed - wgm is low when the corner is sharp , the @xmath0 value of the quasi - wgm also decreases .
the very low @xmath0 plots ( @xmath25 ) of the quasi - wgm in fig .
[ fig:4 ] result from coupling with higher order perturbed - wgms that have a low @xmath0 .
a numerical analysis revealed that the @xmath0 of quasi - wgm is fundamentally dependent on the @xmath0 of the perturbed - wgm .
since the @xmath0 of the perturbed - wgm is low when the corner radius @xmath14 is small , it is not easy to obtain a high-@xmath0 cavity even when we excite the quasi - wgm .
therefore , the only method with the potential to provide a high @xmath0 in such a crystalline cavity is to make the shape as circular as possible . in sec .
[ sec3 ] , we show methods for fabricating circular sapphire wgm cavities by using lhpg to achieve a high @xmath0 . in this section , we discuss the optimal radius of the hexagonal cavity .
we focus on the perturbed - wgm in a hexagonal cavity with a radius @xmath26 and a curvature radius @xmath27 .
it is known that the @xmath0 of a circular wgm cavity increases linearly as the radius increases @xcite . however
, this characteristic has not been studied for a polygonal wgm cavity .
factor of perturbed - wgm as a function of the radius @xmath26 of hexagonal cavities with a curvature radius @xmath28.,width=240 ] figure [ fig : diameter](a ) shows the @xmath0 factor for the perturbed - wgm for hexagonal cavities with different cavity radius @xmath14 .
it clearly shows that there is an optimal point at @xmath29 .
the @xmath0 value decreases when @xmath30 is small , because it is difficult for the light in the cavity to satisfy the total internal reflection condition when the cavity is small .
moreover , unlike a circular cavity , the @xmath0 value decreases when @xmath31 , which is difficult to explain solely in terms of total internal reflection .
a careful investigation revealed that mode coupling plays an important role . in a large cavity , the mode spacing of the lowest order perturbed - wgm and that of the higher order perturbed - wgm are both very small , which results in an easy overlap between these modes .
since higher order - wgms have a much lower @xmath0 , the @xmath0 of the lowest - order perturbed - wgm also decreases for a large cavity . therefore , there is an optimum radius as regards obtaining a high-@xmath0 in a hexagonal cavity .
figure [ fig:10 ] shows the experimental setup for the lhpg method .
the input @xmath32 laser beam ( coherent diamond c-70 ) is formed into a doughnut shape by using a pair of axicon lenses .
this minimizes the beam that hits the feed rod before it is focused with a concave mirror , and it makes the beam focusing more efficient . the @xmath32 laser beam is focused on the feed rod by using a concave mirror with a radius of 50 mm to heat the rod and form a molten zone .
the laser beam spot size is about @xmath33 , and the power density is @xmath34 when the @xmath32 laser power is 4.0 w. both the seed and feed rods are fixed to a linear translation stage ( oriental motor asm46ma ) .
the radius of the sapphire single crystal rods is @xmath35 for both the feed and seed rods .
laser with diameter @xmath36 and beam parameter product @xmath37 is used to heat the crystal rods.,width=240 ] then a crystal fiber is grown by pulling the feed rod upwards at a speed of @xmath38 , while the seed rod is moved in the same direction with a speed of @xmath39 ( @xmath40 ) . in lhpg , we can control the radius of the grown crystal rods by changing the velocity rate whose relationship is given by @xmath41 where @xmath42 and @xmath43 are the radiuses of the grown and feed rods , respectively .
lhpg was originally developed to create crystal rods with a uniform radius .
in contrast , by changing the velocity rate we can create a bulge with which we can confine light in the longitudinal direction and excite the wgm . in the experiment , we changed the velocity rate from 1/6 to 1/3 .
this corresponds to calculated @xmath44 values of @xmath45 and @xmath46 , when the feed rod radius @xmath47 is @xmath35 .
first , we started with standard lhpg .
@xmath48 was set at @xmath49 and @xmath50 was set at @xmath51 . in brief , the @xmath52 was 1/6 and the grown rod was fabricated with a radius of @xmath53 .
next , we reduced @xmath39 from @xmath54 to @xmath55 and set the @xmath52 ratio at 1/3 to fabricate a wgm cavity . the grown radius of the wgm cavity was then larger at @xmath56 . finally , we stopped the seed rod and pulled the feed rod downward . a side view and cross - section of the fabricated cavity
are shown in figs .
[ fig:16](a ) and ( c ) .
the cross - section of this cavity is hexagonal as a result of the crystalline structure of the sapphire .
as discussed earlier in this paper , it is essential to develop a method for controlling the curvature of the edge radius of this cavity in order to increase the @xmath0 .
we introduce a pre - heating method to make the cavity round .
the seed rod is heated with a @xmath57 laser at a power of 8 w prior to the crystal growth .
figure [ fig:22](a ) shows the @xmath35 radius pre - heating seed rod , where @xmath58 is the heating length .
we changed @xmath58 , and observed the cross - sectional shape of the fabricated wgm cavity .
figure [ fig:22](b)(e ) show the cross - section of the fabricated sapphire cavity with different @xmath58 values .
when @xmath58 is larger , the cross - sectional shape becomes close to a circle , and we obtained an optimal @xmath58 value of 0.78 mm .
this experimental result clearly shows that the preheating method is a promising way to control the cross - sectional shape of the cavity .
however , when @xmath58 is 0.80 mm or larger , the cross - section deforms as shown in fig .
[ fig:22](e ) .
in particular , the side view of the cavity ( fig . [ fig:22](f ) ) shows that the reduction of the crystal quality might be significant .
therefore , we need to employ the x - ray diffraction ( xrd ) technique to investigate the crystal quality of the fabricated cavity .
is the length of the preheated region .
( b)-(e ) optical microscope image of a cross - section of the fabricated sapphire cavity with different @xmath58 values .
( b ) @xmath59 , ( c ) @xmath60 , ( d ) @xmath61 , and ( e ) @xmath62 .
( f ) side view of the microcavity fabricated when @xmath58= 0.80 mm.,width=288 ] since we expected a tradeoff between the shape and the crystalline quality of the cavity , we conducted xrd spectrum measurements for cavities with different shapes . because we are interested in the crystal quality at the surface of the cavity where light propagates , we use the @xmath63@xmath64 method , which has an x - ray penetration depth of a few @xmath65 from the sample surface .
the experimental setup is shown in fig .
[ fig : xrd ] .
figures [ fig:20](a ) and ( b ) show 2d images of xrd measurements of cavities with hexagonal ( fig .
[ fig:16](c ) ) and circular cross - sections .
( note : the circular cavity used for this experiment is not the same as that shown in fig .
[ fig:22](d ) , but it has an almost identical shape . ) the x - ray was incident perpendicular to the c - axis of the crystal .
most importantly , in these images we observe bright spots , which indicate that the crystal quality is maintained .
note that debye - scherrer rings should appear if the crystallinity is low .
range in this image is from @xmath66 to @xmath67 .
the corresponding miller indices of @xmath6 are shown in the panel .
( b ) same as ( a ) but with a circular cavity .
( c ) xrd spectrum of the hexagonal cavity .
( d ) same as ( c ) but with a circular cavity.,width=307 ] figures [ fig:20](c ) and ( d ) are the xrd spectra of these cavities .
we observed diffraction peaks that originated from @xmath5 , but there is no clear evidence of different alumina compounds .
this also shows that relatively good crystal quality is maintained even for a circular cavity .
however , we found that the spectrum peaks diffracted owing to the high - order planes as seen in fig .
[ fig:20](b ) .
this indicates that in some part of the cavity , the c - axis of the crystal is tilted
. we must be careful if we want to use the nonlinear susceptibility of this material , which is a tensor .
however , for other applications , this effect is limited , although some randomness is inherent in the crystal structure .
we measured the transmittance spectrum of the sapphire microcavity in fig .
[ fig:14](a ) by coupling light with an optical tapered fiber setup .
we can control the coupling between the resonator and the waveguide in order to measure the cavity @xmath0 .
the transmittance spectrum of a circular sapphire wgm cavity is shown in fig .
[ fig:14](b ) .
we observe clear resonance with an equal free - spectral range .
figure [ fig:14](c ) is an enlarged view of fig .
[ fig:14](b ) . by fitting the spectrum with a lorentzian shape
, we obtained a @xmath0 of @xmath68 .
this value is higher than that obtained from the resonant spectrum of a hexagonal cavity shown in fig .
[ fig:14](d ) , which has a @xmath0 of @xmath69 .
as expected , with a numerical analysis , we obtained a higher @xmath0 for a circular cavity than for a hexagonal cavity . of @xmath70 .
( d ) transmission spectrum of the hexagonal cavity without preheating .
it has a @xmath0 of @xmath69.,width=288 ] although theory and experiment both predict a higher @xmath0 for a circular cavity , and are in a good agreement , the absolute @xmath0 values are far apart . to improve the experimental @xmath0
, we discuss the effect of surface roughness and perform some experiments to improve the smoothness .
one reason for the low-@xmath0 in the experiment is the relatively large radius of the wgm cavity , which results in a lower @xmath0 as described in sec .
[ sec2 ] . however , different factors can influence the cavity @xmath0 . of these
, surface roughness is clearly the most important factor in terms of impact on the cavity @xmath0 .
it is known that the @xmath0 value is proportional to the square of the surface roughness @xcite .
here we briefly describe an experiment designed to improve the smoothness of the surface by changing the crystal growth velocity .
we used a scanning white - light interferometer ( zygo new view tm6200 ) to examine the surface roughness of the side wall of the cavity .
figure [ fig:15](a ) is a 3d surface plot for a circular cavity with a @xmath0 of @xmath68 .
we fabricated this cavity with a crystal growth speed of @xmath51 .
the average surface roughness of the cavity surface was 65-nm rms . according to ref .
@xcite , we may be able to obtain a @xmath0 on the order of @xmath71 if we can reduce the surface roughness to a few nm . to fabricate a crystal rod with a smoother surface
, we reduced the growth velocity to @xmath72 .
then we obtained a surface roughness of about 6 nm rms , as shown in fig .
[ fig:15](b ) .
the result shows us that the growth velocity is the key to the surface roughness .
although optimization is still being studied , we should be able to fabricate a crystalline wgm cavity with a smoother surface , which may support an ultrahigh @xmath0 .
a different method of optimizing the surface might involve growing the crystal in a he gas environment @xcite .
since various studies have been conducted using the lhpg method , there is still plenty of room left for the optimization of the surface smoothness .
the roughness is 65 nm .
( b ) surface plot of a crystal rod grown with a velocity of @xmath49 .
the surface roughness is 6 nm.,width=288 ]
we investigated excited modes in polygonal cavities in detail and found that a perturbed - wgm exhibits an ultrahigh @xmath0 when the corner radius of the cavity is large .
we found that a number of lower and higher order perturbed - wgms are present in a hexagonal cavity coupled with a quasi - wgm .
although the quasi - wgm appears to exhibit an ultrahigh-@xmath0 with ray - optics , mode coupling with the perturbed - wgm limits the @xmath0 of this mode .
we revealed that we need to understand the coupling between the modes to explain the low @xmath0 of the quasi - wgm when the cavity is hexagonal . as a result ,
the only way to improve the @xmath0 of the crystalline cavity ( with both sapphire and zno cavities ) is to make the corner round . with this in mind , we showed pre - heating method that can control the cross - section of the fabricated cavity . although some degeneracy of the crystal was observed , this effect was limited . despite the experimental @xmath0 of the current device being limited by the surface roughness
, we showed that further optimization is possible .
research using wgm cavities is active in various fields including micro - laser fabrication , quantum - optics , sensing , and signal processing .
however , the fabrication of a crystalline wgm remains difficult because of its fragility .
this study will contribute to the provision of an alternative method for fabricating very small crystalline wgms by lhpg for use in these fields .
part of this work was supported by the strategic information and communications r&d promotion programme ( scope ) , a grant - in - aid for scientific research from the ministry of education , culture , sports , science and technology , japan ( kaken # 25600118 ) , and the keio university next generation research project promotion program .
the author would also like to thank dr .
atsushi yokoo from ntt basic research laboratories , prof .
yasuhiro kakinuma from keio university , and prof .
yoichi kamihara from keio university , for valuable discussions on lhpg , precise machining and xrd analysis , respectively .
99 t. tanabe , m. notomi , e. kuramochi , a. shinya , and h. taniyama , nat .
photonics * 1 , * 49 ( 2007 ) . y. akahane , t. asano , b. s. song , and s. noda , nature * 425 , * 944 ( 2003 ) .
v. r. almeida , c. a. barrios , r. r. panepucci , and m. lipson , nature * 431 , * 1081 ( 2004 ) .
p. delhaye , a. schliesser , o. arcizet , t. wilken , r. holzwarth , and t. j. kippenberg , nature * 450 , * 1214 ( 2007 ) .
i. s. grudinin , a. b. matsko , a. a. savchenkov , d. strekalov , v. s. ilchenko , and l. maleki , opt . commun .
* 265 , * 33 ( 2006 ) .
t. j. kippenberg , r. holzwarth , and s. a. diddams , science * 332 , * 555 ( 2011 ) .
w. yoshiki and t. tanabe , j. opt .
b , * 29 , * 3335 ( 2012 ) .
m. notomi , e. kuramochi , and t. tanabe , nat .
photonics * 2 , * 741 ( 2008 )
. y. a. vlasov , m. oboyle , and h. f. hamann , and s. j. mcnab , nature * 438 , * 65 ( 2005 ) .
a. a. savchenkov , v. s. ilchenko , a. b. matsko , and l. maleki , phys .
a * 40 , * 051804(r ) ( 2004 ) .
l. maleki , nat .
photonics * 5 , * 728 2011 .
t. tanabe , k. nishiguchi , e. kuramochi , and m. notomi , opt .
express * 17 , * 22505 ( 2009 ) .
t. tanabe , m. notomi , s. mitsugi , a. shinya , and e. kuramochi , opt .
* 30 , * 2575 ( 2005 ) .
a. a. savchenkov , a. b. matsko , v. s. ilchenko , i. solomatine , d. seidel , and l. maleki , phys .
* 101 , * 093902 ( 2008 ) .
s. weis , r. rivire , s. delglise , e. gavartin , o. arcizet , a. schliesser , t. j. kippenberg , science * 330 , * 1520 ( 2010 ) . s. inoue and y. aoyagi , phys . rev .
* 94 , * 103904 ( 2005 ) .
t. tanabe , m. notomi , h. taniyama , and e. kuramochi , phys .
* 102 , * 043907 ( 2009 ) .
t. lu , h. lee , t. chen , s. herchak , j .- h .
kim , s. e. fraser , r. c. flagan , and k. vahala , proc .
5976 ( 2011 ) .
f. vollmer , s. arnold , and d. keng , proc .
* 105 , * 20701 ( 2008 ) .
d. k. armani , t. j. kippenberg , s. m. spillane , and k. j. vahala , nature * 421 , * 925 ( 2003 ) .
m. pllinger and a. rauschenbeutel , opt .
express * 18 , * 17764 ( 2010 ) . c. y. wang , t. herr , p. delhaye , a. schliesser , j. hofer , r. holzwarth , t. w. hnsch , n. picqu , and t. j. kippenberg , nat .
* 4 , * 1345 ( 2013 ) .
j. hofer , a. schliesser , and t. j. kippenberg , phys .
rev . a * 82 , * 031804(r ) ( 2010 ) .
h. kudo , y. ogawa , t. kato , a. yokoo , and t. tanabe , `` fabrication of whispering gallery mode cavity using crystal growth , '' ( submitted , preprint is available in arxiv:1302.5488 ) .
a. yokoo , s. tomaru , i. yokohama , h. itoh , and t. kaino , j. cryst .
growth * 156 * , 279 ( 1995 ) .
m. m. fejer , l. j. nightingale , g. a. magel , and r. l. byer , rev .
. instrum .
* 55 , * 1791 ( 1984 ) .
r. feigelson , mater .
, b * 1 * , 67 ( 1988 ) .
t. kato , w. yoshiki , r. suzuki , and t. tanabe , appl .
* 101 , * 121101 ( 2012 ) .
u. vietze , o. krau , and f. laeri , phys .
* 81 , * 4628 ( 1998 ) .
t. nobis , e. m. kaidashev , a. rahm , m. lorenz , and m. grundmann , phys .
* 93 , * 103903 ( 2004 ) .
r. chen , t .-
d. tran , k. w. ng , w. s. ko , l. c. chuang , f. g. sedgwick and c. c .- hasnain , nat .
photonics * 5 , * 170 ( 2011 ) . j. h. paek , t. nishiwaki , m. yamaguchi , and n. sawaki , physica e * 42 , * 2722 ( 2010 ) .
peng , c .- y .
lu , w .- h .
wu , and s .- l .
wang , appl .
* 87 , * 161902 ( 2005 ) .
h. dong , z. chen , l. sun , j. lu , w. xie , h. h. tan , c. jagadish , and x. shen , appl .
* 94 , * 173115 ( 2009 ) .
j. wiersig , phys .
a * 67 , * 023807 ( 2003 ) .
t. nobis and m. grundmann , phys .
a * 72 , * 063806 ( 2005 ) .
y. z. huang , w .- h .
guo , and q .- m .
wang , ieee j. quantum electron .
* 37 , * 100 ( 2001 ) .
g. dai , y. zhang , r. liu , q. wan , q. zhang , a. pan , and b. zou , j. appl .
* 110 , * 033101 ( 2011 ) . c. p. dietrich , m. lange , t. bntgen , and m. grudmann , appl .
* 101 , * 141116 ( 2012 ) . c. czekall , t. nobis , a. rahm , b. cao , j. z- .
perz , c. sturm , r. s- .
grund , m. lorenz , and m. grundmann , phys .
status solidi ( b ) * 247 , * 1282 ( 2010 ) .
r. chen , bo ling , x. w. sun , and h. d. sun , adv .
mater , * 23 * 2199 ( 2011 )
. b. hua , j. motohisa , y. ding , s. hara , and t. fukui , appl .
* 91 , * 131112 ( 2007 ) .
b. wang , x. jin , h. y. wu , and z. q. zheng , j. appl .
* 113 , * 034313 ( 2013 ) .
v. phomsakha , r. s. f. chang , and n. djeu , rev .
* 65 , * 3860 ( 1994 ) . | we investigated the whispering gallery modes of cavities with a hexagonal cross - section .
we found two different modes , namely perturbed and quasi - wgms , of which the former exhibits the higher @xmath0 when the corner radius is large .
we studied the dependence of @xmath0 on the curvature radius of the polygonal cavities and found that the coupling between the two modes determines the @xmath0 of the cavity .
in addition we fabricated a cavity by employing laser heated pedestal growth and demonstrated a high @xmath0 . |
galois groups are not only symmetry groups of field extensions , they are also symmetry groups in enumerative geometry . this second point was made by jordan in 1870 @xcite who studied some classical problems in enumerative geometry , showing that several had galois groups which were not the full symmetric group , reflecting the intrinsic structure of these problems .
earlier , hermite gave a different connection to geometry , showing that the algebraic galois group coincided with a geometric monodromy group @xcite .
harris used this to study the galois group of several problems in enumerative geometry @xcite . for each he showed that their monodromy groups were the full symmetric groups on their sets of solutions and therefore the problem had no intrinsic structure .
the schubert calculus is a well - understood family of problems in enumerative geometry that involve linear subspaces having prescribed positions with respect to other , fixed linear spaces .
it provides a laboratory for studying galois groups in enumerative geometry .
for example , leykin and sottile @xcite directly computed monodromy for many schubert problems on small grassmannians involving simple ( codimension one ) schubert conditions and found that each problem had monodromy the full symmetric group . in @xcite , vakil gave a general combinatorial method based on the principle of specialization and group theory for showing that a problem in enumerative geometry has at least alternating galois group in that its galois / monodromy group contains the alternating group . with this method and his geometric littlewood - richardson rule @xcite
, vakil showed that many schubert problems on small grassmannians had at least alternating galois groups .
he also found schubert problems whose galois groups are not the full symmetric group .
brooks , et al .
@xcite used vakil s combinatorial criterion and some delicate estimates of integrals to show that all schubert problems on grassmannians of 2-planes have at least alternating galois groups .
vakil gave another , stronger , criterion for showing that a galois group was at least alternating which requires knowing that it is a doubly transitive permutation group .
we study double transitivity of galois groups of schubert problems on grassmannians .
vakil s stronger criterion is not our only motivation .
there appears to be a siginificant gap in transitivity : every galois group that we have studied is either at least alternating and therefore highly transitive , or else it fails to be doubly transitive and is imprimitive in that it preserves a partition of the solutions .
we conjecture that a galois group of a schubert problem is either the full symmetric group on its solutions , or it is imprimitive .
one purpose of this paper is to give theoretical and computational evidence supporting this conjecture .
a schubert condition on @xmath1-planes in @xmath2-space is if the condition is that the @xmath1-plane meets an @xmath3-plane nontrivially with @xmath4 .
a schubert problem is if it only has special conditions .
we state our main results .
* theorem .
* _ we have the following : _ 1 .
every special schubert problem has a doubly transitive galois group .
every schubert problem in @xmath5 has at least alternating galois group .
every schubert problem in @xmath6 has a doubly transitive galois group .
there are exactly fourteen schubert problems in @xmath0 whose galois groups are not at least alternating .
for each , the galois group is imprimitive .
part ( 2 ) was proven in @xcite . using part ( 1 ) and vakil s stronger criterion
, we give a significantly simpler proof of that result .
this approach does not generalize to show that all schubert problems involving 3-planes have at least alternating monodromy , for that requires a significantly more delicate analysis of geometric problems involving vakil s checkerboard varieties @xcite . by part
( 4 ) , our results on double transitivity can not be extended to grassmannians of 4-planes .
one of the fourteen schubert problems with imprimitive galois group was first described in 3.13 of @xcite , and is due to derksen .
we determine exactly the schubert problems on this grassmannian whose galois group is not at least alternating . in each case
, we compute the galois group , and demonstrate that the resulting groups are imprimitive .
in addition to derksen s example , there are essentially two others which generalize to give two families of schubert problems with imprimitive monodromy .
one family contains a problem in which all but two of its schubert conditions are special and the remaining two conditions are dual special schubert conditions .
hence our results on double transitivity can not be further extended to simple general statements .
this paper is organized as follows . in section [ s :
mg ] , we provide some background on galois / monodromy groups , and explain vakil s work , including his criteria . in section [ s : sc ]
we give basic definitions in the schubert calculus on grassmannians . in section [ s : dt ]
we discuss how to show double transitivity of galois groups using geometry and prove some geometric lemmas . in section [
s : mr ] we establish parts ( 1 ) and ( 3 ) of our main theorem about special schubert problems and schubert problems involving 3-planes . in section [ s : g2n ] , we use the double transitivity of special schubert problems to prove part ( 2 ) .
we close with section [ s : im ] , in which we prove part ( 4 ) and study the galois group of every schubert problem on @xmath0 .
we provide some background on galois / monodromy groups in enumerative geometry .
more information can be found in @xcite .
we work over an algebraically closed field @xmath7 of arbitrary characteristic .
suppose that @xmath8 is a proper , generically finite and separable ( i.e.generically tale ) morphism of @xmath7-schemes of degree @xmath9 , and @xmath10 is irreducible .
let @xmath11 be the zariski closure in the @xmath9-fold fiber product of the scheme @xmath12 where @xmath13 is the big diagonal .
let @xmath14 be a closed point in the open set over which @xmath15 is finite and separable such that the fiber @xmath16 consists of @xmath9 reduced points .
( such a point @xmath17 is a of @xmath15 . )
choose an ordering @xmath18 of the points of @xmath16 .
then the of @xmath8 is the subgroup of the symmetric group of permutations of @xmath19 $ ] consisting of those permutations @xmath20 such that the points @xmath21 and @xmath22 in the fiber above @xmath17 are in the same irreducible component of @xmath23 .
write , where @xmath24 , for the component of @xmath23 containing the point @xmath25 .
the group @xmath26 is well - defined as a subgroup of @xmath27 , up to conjugacy .
let @xmath28 be an integer .
a permutation group @xmath29 is if any two ordered sets of @xmath30 distinct numbers are mapped onto each other by an element of @xmath31 .
we interpret this geometrically for galois / monodromy groups .
define @xmath32 to be the zariski closure of the scheme @xmath33 where @xmath13 is the big diagonal as before .
let be the union of the irreducible components of @xmath34 that map dominantly onto @xmath10 .
each component of @xmath35 has the same dimension as @xmath10 and its projection to @xmath10 has finite fibers over the dense subset of regular values of @xmath15 .
the fiber of @xmath35 over the point @xmath17 consists of all @xmath30-tuples @xmath36 where @xmath37 are distinct elements of @xmath18 , taken in every possible order .
[ l : t - transitive ] the galois / monodromy group @xmath26 of @xmath8 is @xmath30-transitive if and only if @xmath35 is irreducible .
in particular , @xmath26 is transitive if and only if @xmath38 is irreducible .
let @xmath14 be a regular value of @xmath15 , order the points of @xmath16 , and set @xmath24 .
let @xmath39 be the component of @xmath23 containing @xmath25 and consider its image in @xmath34 under the projection map @xmath40 given by forgetting the last @xmath41 factors in the fiber product . as @xmath17 is a regular value , we have @xmath42 , and so @xmath43 is an irreducible component of @xmath35 .
the fiber of @xmath43 over @xmath17 consists of all @xmath30-tuples @xmath44 for @xmath45 .
if @xmath26 is @xmath30-transitive , then @xmath46 , which implies that @xmath35 is irreducible .
conversely , if @xmath35 is irreducible , then @xmath46 , which implies that @xmath26 is @xmath30-transitive .
suppose that we have @xmath8 as above with @xmath38 irreducible and @xmath10 smooth .
then the galois / monodromy group @xmath26 is a transitive subgroup of the symmetric group @xmath27 .
a subgroup @xmath31 of @xmath27 is if it is either the alternating subgroup of @xmath27 or the full symmetric group @xmath27 .
vakil gave several criteria which may be used to show that @xmath26 is at least alternating .
we follow the discussion of @xcite .
suppose that we have a fiber diagram @xmath47 where @xmath48 is the closed embedding of a cartier divisor , @xmath10 is smooth in codimension one along @xmath49 , and @xmath50 is a generically finite and separable morphism of degree @xmath9
. when @xmath51 has at most two components , we have the following . 1 .
if @xmath51 is irreducible , then there is an inclusion @xmath52 into @xmath26 .
if @xmath51 has two components , @xmath53 and @xmath54 , each of which maps dominantly to @xmath49 of respective degrees @xmath55 and @xmath56 , then the monodromy group of @xmath57 is a subgroup of @xmath58 that maps surjectively onto each factor @xmath59 and which includes into @xmath26 ( via @xmath60 ) . in @xcite , vakil gave criteria for deducing that @xmath26 is at least alternating .
this follows by purely group - theoretic arguments including goursat s lemma .
suppose that we have a fiber diagram @xmath61 .
then @xmath26 is at least alternating if one of the following holds . 1
. if we are in case @xmath62 and @xmath52 is at least alternating .
if we are in case @xmath63 , @xmath64 and @xmath65 are at least alternating , and either @xmath66 or @xmath67 .
if we are in case @xmath63 , @xmath64 and @xmath65 are at least alternating , one of @xmath55 or @xmath56 is not @xmath68 , and @xmath26 is @xmath69-transitive . in the introduction
, we referred to criterion ( b ) as vakil s combinatorial criterion , and ( c ) as his stronger criterion .
[ r : vakil ] these criteria apply to more general inclusions @xmath48 of an irreducible variety into @xmath10 .
all that is needed is that @xmath10 is generically smooth along @xmath49 , for then we may replace @xmath10 by an affine open set meeting @xmath49 and there are subvarieties @xmath70 with each inclusion @xmath71 that of a cartier divisor where @xmath72 is smooth in codimension one along @xmath73 , and then apply induction .
vakil observed that as the alternating group is least @xmath74-transitive , a consequence of criterion ( c ) is that to show that a galois / monodromy group is @xmath74-transitive , it often suffices to show that it is merely doubly transitive .
we develop the schubert calculus for the grassmannian in a form that we will use .
more material , including proofs and references , may be found in @xcite .
write for the linear span of a subset @xmath75 of a vector space . for
a positive integer @xmath2 , write for the set @xmath76 . for a finite - dimensional @xmath7-vector space @xmath77 ,
let be the projective space of 1-dimensional linear subspaces of @xmath77 .
for @xmath78 , the of @xmath1-dimensional linear subspaces of @xmath77 is a smooth irreducible algebraic variety of dimension @xmath79 .
then @xmath80 .
we will also write @xmath81 for @xmath82 . the schubert calculus concerns @xmath1-planes in @xmath77 having prescribed positions with respect to other linear subspaces .
the prescribed positions are encoded by , which are weakly decreasing sequences of nonnegative integers , @xmath83 for example @xmath84 is a partition for @xmath85 and @xmath86 is a partition for @xmath87 .
we typically omit trailing 0s and represent a partition by a left - justified array of boxes with @xmath88 boxes in row @xmath89 .
thus these two partitions are @xmath90 a partition prescribes the position of a @xmath1-plane with respect to a ( complete ) of subspaces , which is a sequence @xmath91 of linear subspaces with @xmath92 . given a partition @xmath93 and a flag @xmath94 , the schubert condition of type @xmath93 imposed by the flag @xmath94 on @xmath1-planes @xmath95 is @xmath96 for the partition @xmath97 this is @xmath98 and the other conditions given by the trailing 0s are that @xmath99 for @xmath100 , which always hold .
the set of all @xmath95 which satisfy the schubert condition @xmath93 on the flag @xmath94 is a @xmath101 this is an irreducible subvariety of @xmath82 of codimension @xmath102 and thus of dimension @xmath103 .
the schubert variety has a distinguished open subset @xmath104 , which is the locus where the inequalities are all equalities .
we will also need schubert varieties in manifolds of partial flags .
a partial flag @xmath94 is an increasing sequence of subspaces @xmath105 where @xmath106 .
this sequence of dimensions @xmath107 is the of the flag @xmath94 .
the set of all flags of type @xmath108 forms a smooth variety of dimension @xmath109 where @xmath110 is taken to be @xmath111 .
this flag variety has distinguished schubert varieties , which we describe as follows .
flags @xmath94 of type @xmath108 and @xmath112 of type @xmath113 have a , which is encoded by the .
this is the @xmath114 array where @xmath115 the set of flags @xmath116 whose position with respect to @xmath112 is at least that of @xmath94 , @xmath117 is a .
this irreducible subvariety of @xmath118 has a distinguished dense open subset @xmath119 consisting of @xmath120 for which @xmath121 for all @xmath122 .
only subspaces @xmath123 with @xmath124 in a flag @xmath94 are used to define the schubert variety @xmath125 .
thus we will often replace complete flags with partial flags of the type occurring in , that is @xmath126 and @xmath127 ( @xmath128 is the last nonzero part of @xmath93 . ) write for this space of partial flags .
it has dimension @xmath129 for example , when @xmath130 and @xmath131 , @xmath132 .
when @xmath133 has only one nonzero part , so that it consists of a single row , we will call it a , and simply write it as @xmath134 .
dually , when @xmath93 consists of a single column , so that @xmath135 with @xmath136 1s , then it is a , and we write it as @xmath137 .
we may omit the subscripts giving the dimensions of subspaces in a flag in @xmath138 .
thus for @xmath130 and @xmath131 , @xmath139 means that the partition is @xmath140 and @xmath141 has dimension @xmath142 , and @xmath143 means that the flag is @xmath144 .
associating a @xmath1-plane @xmath145 to its annihilator @xmath146 in the dual space @xmath147 of @xmath77 gives an isomorphism between @xmath82 and @xmath148 , where @xmath149 . under this isomorphism
, @xmath125 is sent to @xmath150 , where @xmath151 is the transpose of @xmath93 , the partition obtained by interchanging rows and columns , and @xmath152 is the flag whose @xmath89-plane is the annihilator of @xmath153 . in particular
, special schubert varieties are sent to dual special schubert varieties .
we close this subsection with a dimension calculation .
let @xmath154 and @xmath155 be types of @xmath156-step flags .
( note that we allow @xmath157 , so that a flag of type @xmath158 may have repeated subspaces . ) for any @xmath159 , consider the schubert subvariety of @xmath118 defined by @xmath160 [ lem : xeldot ] the schubert variety @xmath161 is nonempty if and only if @xmath162 where @xmath163 .
when @xmath164 is satisfied , @xmath161 has dimension @xmath165 summing for @xmath166 gives the obvious necessary condition for nonemptiness .
we prove this by induction on the length @xmath156 of the flags .
suppose first that @xmath167 .
then condition is @xmath168 , and @xmath169 , which is simply @xmath170 and has dimension @xmath171 .
this is nonempty if and only if @xmath168 .
now suppose that the statement of the lemma holds for flags of length @xmath172 .
given a flag @xmath112 of length @xmath156 , let @xmath173 be its truncation to length @xmath172 and @xmath174 , its last subspace .
set @xmath175 to be those @xmath176 in @xmath177 where @xmath178 , which is an open subset of @xmath177 .
this set is nonempty ( and therefore dense ) only if there exists a flag @xmath94 in @xmath177 with @xmath179 , which is equivalent to the inequality @xmath180 .
but this holds by , as @xmath181 . for @xmath182
we consider the fiber @xmath183 .
this is isomorphic to @xmath184 , which is @xmath185 where @xmath186 , which is @xmath187 , as @xmath182 .
we must have that @xmath188 for this to be nonempty , and when this is satisfied , we have @xmath189\,,\ ] ] the dimension of this grassmannian .
this completes the proof .
a on @xmath81 is a list @xmath190 of partitions with @xmath191 . given a schubert problem @xmath192 and a list of general flags @xmath193 with @xmath194 , the intersection @xmath195 is transverse @xcite and consists of finitely many points .
this number of points is independent of the choice of general flags and the algebraically closed field @xmath7 .
the schubert problem is if @xmath196 . by transversality ,
when the flags are general , all points of the intersection lie in the intersection of the open schubert varieties @xmath197 where @xmath198 is the subset of @xmath125 where the inequalities in are replaced by equalities .
for a schubert problem @xmath199 , consider the family of all intersections , @xmath200 set @xmath201 , the space of @xmath202-tuples of partial flags for schubert varieties indexed by the partitions of @xmath192 .
this family @xmath203 is equipped with two projections @xmath204 and @xmath205 .
fibers of the projection to @xmath206 are intersections , which , when transverse and nonempty , are zero - dimensional .
thus we expect that @xmath207 .
in fact , for a schubert problem @xmath192 we will always have this equality of dimension . to see this ,
define the subvariety @xmath208 for @xmath209 to be @xmath210 this is a schubert subvariety of @xmath211 of codimension @xmath212 , and it is irreducible .
now consider fibers of the projection @xmath213 . for @xmath95 , we have @xmath214 ( here , as elsewhere , we identify a fiber @xmath215 with its image under the second projection @xmath15 to @xmath206 . ) since any two fibers are isomorphic , @xmath204 realizes @xmath203 as a fiber bundle .
the dimension of a fiber @xmath215 is the sum of dimensions of the schubert varieties in the product , which is @xmath216 from which it follows that @xmath207 .
since the base and fibers of the projection @xmath204 are irreducible , we deduce that @xmath203 is irreducible .
let @xmath192 be a schubert problem . by the transversality of a general intersection , the map @xmath217 a proper , generically finite and separable ( i.e. generically tale ) morphism of @xmath7-schemes of degree @xmath218 .
furthermore @xmath206 is irreducible and smooth as it is a product of flag manifolds .
the of the schubert problem @xmath192 is the galois group of this family of all instances of the schubert problem @xmath192 , that is , @xmath219 which acts transitively , as @xmath203 is irreducible . a schubert problem
is reduced , if , roughly , it is not equivalent to a schubert problem on a smaller grassmannian .
every schubert problem @xmath192 is equivalent to a reduced schubert problem @xmath220 , and we have @xmath221 .
thus , when proving that a galois group has a given property , it suffices to assume that the schubert problem is reduced .
we give a definition of reduced schubert problems and sketch a proof of these facts .
[ d : reduced ] a schubert problem @xmath192 on @xmath81 is if for every pair of partitions @xmath222 from @xmath192 , none of the following hold .
1 . @xmath223 .
2 . @xmath224 .
3 . there is some @xmath225 with @xmath226 .
there is some @xmath227 with @xmath228 .
if the schubert problem is nontrivial , then ( c ) can hold at most with equality , for if @xmath229 and the flags @xmath230 are in general position , then @xmath231 .
[ prop : reduced ] any nontrivial schubert problem @xmath192 is equivalent to a reduced schubert problem @xmath220 having an isomorphic galois group , @xmath232 .
we indicate how a nonreduced schubert problem @xmath192 is equivalent to a schubert problem on a smaller grassmannian .
this process may be iterated to obtain an equivalent reduced schubert problem .
the arguments given in remark 4 of @xcite generalize to show that galois groups are preserved by this reduction process .
suppose that @xmath233 in @xmath234 with @xmath235 .
then @xmath236 thus if condition ( a ) of definition [ d : reduced ] holds , then @xmath237 and if condition ( b ) holds , then @xmath238 . if ( a ) holds , then @xmath239 , which is a schubert variety in @xmath240 , where @xmath241 is the partition @xmath242 and @xmath243 is the image of the flag @xmath112 in @xmath244 , specifically @xmath245 .
if @xmath246 where @xmath94 is a flag in general position with respect to @xmath112 , then @xmath247 , where @xmath248 is the image of the flag @xmath94 in @xmath244 , specifically , @xmath249 .
if ( b ) holds , then @xmath250 , which is a schubert variety in @xmath251 , where @xmath252 is the partition @xmath253 and @xmath254 is the flag in @xmath255 induced by @xmath112 , specifically , @xmath256 . if @xmath246 where @xmath94 is a flag in general position with respect to @xmath112 , then @xmath257 , where @xmath258 .
the point of these reductions is that the original schubert problem @xmath192 is equivalent to a schubert problem @xmath220 in a smaller grassmannian .
all partitions of @xmath220 are from @xmath192 , except that the partition @xmath259 has been changed in that either its first row of length @xmath260 has been removed or all of its columns of height @xmath1 have been removed . for condition ( c ) in definition [ d : reduced ] , note that if @xmath261 , then for every @xmath225 , @xmath262 since @xmath145 has dimension @xmath1 , these conditions imply that @xmath263 if @xmath264 are in general position and @xmath265 , then is impossible as @xmath266 . thus for the schubert problem @xmath192 to have solutions
, we must have @xmath267 .
suppose that @xmath268 .
then @xmath269 is a 1-dimensional linear space contained in @xmath145 that we will call and @xmath270 .
furthermore , if @xmath271 and @xmath272 are obtained from @xmath259 and @xmath273 by omitting the parts @xmath274 and @xmath275 and @xmath276 , @xmath277 are the images of @xmath112 and @xmath94 in @xmath278 , then @xmath279 . as before , if @xmath280 with @xmath281 in general position with respect to @xmath112 and @xmath94 , then @xmath282 . finally , note that if @xmath261 , we have as well as @xmath283 . if @xmath112 and @xmath94 are in general position , this implies that @xmath284 if we have @xmath228 , then this sum @xmath51 of subspaces of the flags has codimension @xmath285 in @xmath77 .
then @xmath286 , where @xmath271 is obtained from @xmath259 by subtracting @xmath285 from each of the first @xmath89 parts of @xmath259 and @xmath272 is obtained from @xmath273 by subtracting @xmath285 from each of the first @xmath287 parts of @xmath273 .
we give an example of this , where @xmath288 and @xmath289 with @xmath290 and @xmath291 .
then if @xmath292 , @xmath293 and @xmath294 . drawing @xmath259 together with @xmath273 ( rotated by @xmath295 ) in the @xmath296 box shows that there are two ( @xmath297 ) columns of height three covered .
removing those columns from @xmath259 and @xmath273 gives @xmath298 and @xmath299 .
@xmath300 as before , if @xmath280 in @xmath82 with @xmath281 in general position , then @xmath301 in @xmath302 .
we also note that any schubert problem @xmath192 on @xmath81 is equivalent to a dual schubert problem on the dual grassmannian @xmath303 which has an isomorphic galois group .
we develop tools for showing that the galois / monodromy group @xmath304 of a schubert problem @xmath192 on the grassmannian @xmath81 is doubly transitive .
we will assume that @xmath305 , for otherwise double transitivity is an empty condition .
by proposition [ prop : reduced ] , we may assume that @xmath192 is reduced .
by lemma [ l : t - transitive ] , it suffices to show that @xmath306 is irreducible . for this
, we investigate the projection @xmath307 . here , @xmath308 is a subset of the fiber product @xmath309 , but @xmath310 is simply the ordinary product of grassmannians . unlike the projection @xmath311 ,
this is not a fiber bundle as @xmath312 depends upon the relative position of @xmath313 and @xmath314 .
however , over the locus where @xmath315 it is a fiber bundle . since @xmath306 is a union of components of @xmath316 that project dominantly to @xmath206 , we will study @xmath317 and determine the components @xmath318 with @xmath319 . from our description of such components , we prove lemma [ lem : mainlemma ] , which gives a numerical condition for @xmath318 that is equivalent to @xmath320 . in section [
s : mr ] we use lemma [ lem : mainlemma ] to show that @xmath306 is irreducible when @xmath192 is a special schubert problem , or a schubert problem in @xmath6 .
we first determine the set - theoretic fibers @xmath312 of @xmath307 .
[ l:2-fiber ] let @xmath199 be a schubert problem on @xmath81 and @xmath321 .
if @xmath322 is the projection , then @xmath323 as before , we are identifying @xmath312 with its projection to @xmath206 in . by the definition of fiber product , @xmath324 since @xmath246 if and only if @xmath325 , we see that @xmath326 which implies .
[ r : dominant_components ] a component @xmath318 of @xmath317 that projects dominantly to @xmath206 must contain points @xmath327 where the flags @xmath328 are in general position so that @xmath329 .
this implies that @xmath330 .
consequently , in our arguments we may replace @xmath331 by any subset containing @xmath332 . by lemma [ l:2-fiber ]
, we must first understand the set - theoretic intersection @xmath333 when @xmath334 .
we begin with the case when the partition is a special schubert condition , @xmath335 , abbreviated .
flags of type @xmath134 are elements @xmath141 of @xmath336 . for @xmath145 of dimension @xmath1 and @xmath141 of dimension @xmath337 write @xmath338 and @xmath339 for the corresponding special schubert varieties , @xmath340 and note that @xmath341 .
set @xmath342 , so that @xmath343 .
we expect that @xmath344 .
[ lem : speccomponents ] let @xmath334 with @xmath345 . if @xmath346 or @xmath347 then the intersection @xmath348 has dimension @xmath349 and it is irreducible if @xmath350 or @xmath347 . when @xmath351 and @xmath352 , it consists of two components , one of dimension @xmath349 , and the other of dimension @xmath353 .
this second , possibly excess , component consists of those @xmath141 with @xmath354 .
set @xmath355 , which is @xmath356-dimensional .
since @xmath357 , we have that @xmath358 .
the intersection is a disjoint union @xmath359 where consists of those @xmath360 with @xmath361 and consists of those @xmath141 with @xmath362 .
we compute the dimensions of @xmath363 and @xmath364 by studying incidence varieties which map birationally to each .
set @xmath365 .
this projects birationally to @xmath363 , and its image in the last two factors @xmath366 is the open subset @xmath367 with fiber over @xmath368 those @xmath141 which contain the 2-dimensional linear span @xmath369 but do not meet @xmath370 .
this is a zariski open subset of the grassmannian @xmath371 thus the dimension of @xmath372 ( and thus of @xmath363 ) is @xmath373 for @xmath364 , define @xmath374 .
this projects birationally to @xmath364 .
the second projection @xmath375 is surjective with fiber over @xmath376 the set of those @xmath141 which contain @xmath377 , which is @xmath378 .
thus the dimension of @xmath379 ( and thus of @xmath364 ) is @xmath380 note that @xmath381 , the expected dimension of @xmath382 . when @xmath347 , @xmath383 , and this intersection is irreducible . if @xmath384 , then @xmath385 and so @xmath364 lies in the closure of @xmath363 . when @xmath386 , @xmath387 , and so its closure is a component of the intersection , which has the expected dimension @xmath388 only when @xmath389 .
we introduce notation for the sets @xmath363 and @xmath364 in the decomposition .
write for the set @xmath363 and for the set @xmath364 .
we now consider intersections @xmath390 , where @xmath93 is any schubert condition .
let @xmath391 with @xmath345 and set @xmath392 , so that @xmath393 .
since we assume that our schubert problem is reduced , we will have @xmath394 and @xmath395 for some @xmath396 .
we decompose @xmath397 into @xmath398 disjoint subsets and then compute the dimension of those that are nonempty .
[ def : u_s ] for @xmath399 $ ] , define @xmath400 to be @xmath401 \cap s|\,,\ j=1,\dotsc , m \}\,.\ ] ] recall that @xmath402:=\{1,\dotsc , j\}$ ] .
observe that we must have @xmath403 , and that @xmath404 indeed , for each @xmath405 and @xmath406 , we have @xmath407 , so that @xmath408 .
furthermore , if @xmath409 , then @xmath410 .
all possibilities for the dimensions of @xmath411 are given by @xmath412\cap s|$ ] for subsets @xmath413 of @xmath414 $ ] of cardinality at most @xmath356 .
[ lem : components ] when @xmath415 , the set @xmath416 is irreducible of dimension @xmath417 \smallsetminus s|\right)\,.\ ] ] the intersections of a flag @xmath418 with @xmath313 , @xmath314 , and @xmath370 are flags @xmath419 in these spaces with prescribed incidences .
specifically , for each @xmath406 and @xmath405 , we have @xmath420 \cap s|}\ : = \ f_{n - k+j-\lambda_j } \cap l\,.\ ] ] then we have @xmath421 the flags @xmath422 instantiate the conditions for @xmath94 to belong to @xmath423 and the flag @xmath424 measures how the flags @xmath425 , @xmath426 , and @xmath94 meet @xmath427 .
let consist of triples @xmath428 of flags , where @xmath429 \cap s|}\,.\ ] ] mapping @xmath418 to its intersections @xmath430 with @xmath313 , @xmath314 , and @xmath370 gives a projection @xmath431 .
we compute the dimension of the fiber over @xmath432 .
set @xmath433 , which has dimension @xmath434\cap s|$ ] .
then the fiber @xmath435 is identified with its projection to @xmath138 which is an open subset of the schubert variety @xmath436 of lemma [ lem : xeldot ] . in the notation of that lemma , @xmath437 and @xmath438\cap s|$ ]
. then the fiber @xmath439 has dimension @xmath440 as @xmath441 if @xmath442 and @xmath443 if @xmath444 .
we compute the dimension of @xmath445 .
let @xmath446 be the projection forgetting @xmath425 and @xmath426 . for @xmath447 ,
the fiber @xmath448 is an open dense subset of @xmath449 where @xmath161 is the schubert variety of lemma [ lem : xeldot ] . since @xmath450 has dimension @xmath451 and in the notation of lemma [ lem : xeldot ] , @xmath452 and @xmath453\cap s|$ ] , this fiber @xmath448 has dimension @xmath454 finally , as @xmath455 has dimension @xmath456 \cap s|)\ = \
\sum_{j\in s } ( d - j+|[j]\smallsetminus s|)\,,\ ] ] we have @xmath457 \smallsetminus s|)\,.\ ] ] since all sets and fibers that we considered are irreducible , @xmath416 is irreducible .
the decomposition of @xmath458 in definition [ def : u_s ] as a disjoint union of sets @xmath459 gives a decomposition of the product of the intersections @xmath460 . for each @xmath461 ,
let @xmath462 be the index of the last nonzero part of @xmath463 .
for each @xmath464 and each sequence @xmath465 where @xmath466 is a subset of @xmath467 $ ] of cardinality at most @xmath356 , define @xmath468 we write the product of the intersections @xmath460 as a disjoint union @xmath469 where @xmath470 ranges over all such sequences of subsets . for each choice of @xmath470 ,
the sets @xmath471 for @xmath334 form a fiber bundle over @xmath472 , which we write as .
the following proposition is clear .
[ p : xtwocomponents ] @xmath306 is a subset of the closure of @xmath473 . only sets
@xmath474 of dimension equal to that of @xmath206 could map dominantly to @xmath206 with finite fibers over an open subset , and so it is a necessary condition that @xmath475 for the closure of @xmath474 to be a component of @xmath306 .
we extract a numerical condition which is equivalent to this dimension condition .
[ lem : mainlemma ] let @xmath192 be a schubert problem on @xmath81 with at least two solutions , @xmath356 an integer between @xmath111 and @xmath476 , and @xmath465 with @xmath477 $ ] and @xmath478
. then @xmath479 if and only if @xmath480 \smallsetminus s^i| ) \ =
\ d(n-2k+d ) \,.\ ] ] by lemma [ lem : components ] , @xmath481 has dimension @xmath482 \smallsetminus s^i|)\bigr)\\ \ = \
n({{\boldsymbol{\lambda } } } ) - 2k(n - k)\ + \ \sum_{i=1}^s\sum_{j \in s^i } ( \lambda^i_j - k + d + |[j ] \smallsetminus s^i|)\ , . \qquad
\end{gathered}\ ] ] since @xmath483 , and @xmath484 , we have @xmath485 \smallsetminus s^i|)\,.\ ] ] since @xmath486 , the result follows .
we use lemma [ lem : mainlemma ] to show that if a schubert problem @xmath192 on @xmath81 is either special or if @xmath487 , then its galois group @xmath304 is doubly transitive .
a is a schubert problem all of whose conditions are special , that is , a list @xmath488 in which each partition consists of a single row . writing @xmath489 for @xmath461 , the non - negative integers @xmath490 satisfy the numerical condition @xmath491 we identify @xmath192 with the sequence @xmath492 .
thus , @xmath490 is always understood to be @xmath493 .
[ th : gspec_double ] every special schubert problem has doubly transitive galois group . let @xmath134 be a special schubert condition on @xmath81 and @xmath494 with @xmath495 . by lemma [ lem : speccomponents ] , the intersection @xmath382 has possibly two components @xmath496 and @xmath497 , which are the closures of the sets @xmath498 and @xmath499 of lemma [ lem : components ] .
we adapt the notation of proposition [ p : xtwocomponents ] to that of lemma [ lem : speccomponents ] .
the fiber @xmath312 has a decomposition into sets of the form @xmath500 where @xmath501 and @xmath502 $ ] .
this is a mild change in notation from lemma [ lem : mainlemma ] , where @xmath503 was indexed by a sequence @xmath504 , with @xmath505 when @xmath506 , and @xmath507 otherwise .
let @xmath508 be the subset of @xmath317 whose fiber over @xmath509 is @xmath510 .
then the condition that @xmath511 becomes @xmath512 let @xmath513 so that @xmath514 is zariski dense in @xmath310 .
we show that the only set @xmath508 which maps dominantly onto @xmath206 is the unique component of @xmath515 .
we do this by first showing that @xmath515 has dimension @xmath516 and then that no component of @xmath317 for @xmath517 which has dimension @xmath516 maps dominantly to @xmath206 .
this implies that @xmath306 is irreducible and thus @xmath304 is doubly transitive .
first suppose that @xmath518 .
by lemma [ lem : speccomponents ] , the @xmath89th factor @xmath519 in has a unique component which has dimension @xmath520 . indeed , if @xmath521 so that @xmath522 , then by lemma [ lem : speccomponents ] , @xmath519 is irreducible of dimension @xmath520 . if @xmath523 , then @xmath524 . since @xmath192 is a reduced special schubert problem , we have @xmath525 and so @xmath526 and again lemma [ lem : speccomponents ] implies that @xmath519 is irreducible of dimension @xmath520 .
hence @xmath312 is irreducible of dimension @xmath527 since @xmath528 , we have that @xmath529 .
we now show that if @xmath517 , then no set of the form @xmath508 of dimension @xmath516 maps dominantly to @xmath206 .
suppose that @xmath517 and @xmath502 $ ] satisfies so that @xmath530 .
we claim that the image @xmath531 of @xmath508 in @xmath206 is not dense .
suppose not .
then @xmath532 meets every open subset of @xmath206 . for @xmath334 , if @xmath533 , then the definition of @xmath510 ( via @xmath534 for @xmath535 ) , implies that if @xmath535 , then @xmath536 meets the @xmath356-plane @xmath392 nontrivially . since @xmath537
, we see that @xmath370 lies in the intersection of schubert varieties in @xmath538 , @xmath539 and so this intersection is nonempty .
since @xmath532 meets every open subset of @xmath206 , it meets the subset consisting of @xmath540 which are in sufficiently general position in that the intersection has the expected codimension @xmath541 .
suppose that @xmath509 is a point of @xmath472 with @xmath533 , which is in this general position .
since is nonempty , its codimension is at most the dimension of @xmath538 .
that is , @xmath542 rewrite the sum as @xmath543 and then use to obtain @xmath544 thus @xmath545 .
if we set @xmath546 , then we have @xmath547 whence @xmath548 .
thus there is some @xmath535 for which @xmath549 , contradicting our assumption that every @xmath550 and @xmath192 is reduced .
[ th : gr3n_double ] every schubert problem in @xmath6 has a doubly transitive galois group .
let @xmath192 be a schubert problem in @xmath6 .
we may assume that @xmath192 is reduced .
if not , then by proposition [ prop : reduced ] it is equivalent to a reduced schubert problem on either a @xmath551 , a @xmath552 , or a @xmath553 with @xmath554 .
all reduced schubert problems on @xmath555 have a single solution .
theorem [ th : gr3n_double ] covers the first case of @xmath551 and theorem [ th : gspec_double ] covers the second case of @xmath552 , for every reduced schubert problem on @xmath552 is special @xcite . since @xmath556 is reduced , for every two partitions
@xmath557 , we have @xmath558 in particular , these imply that at most one partition @xmath259 in @xmath192 has @xmath559 .
indices @xmath470 of sets @xmath474 in the decomposition of @xmath317 are lists @xmath465 of subsets @xmath560 where @xmath561 if @xmath562 and @xmath478 . for @xmath563 define @xmath564\mid s^i =
. then we can rewrite as @xmath565 we prove theorem [ th : gr3n_double ] by showing that there is at most one set @xmath474 with @xmath470 and @xmath356 satisfying , ( and thus having dimension @xmath516 ) which maps dominantly to @xmath206 . since @xmath566 , this set maps dominantly to @xmath567 , and so it is dense in @xmath568 .
this implies that @xmath569 is irreducible , and thus @xmath304 is doubly transitive .
we consider each case @xmath570 separately in the following three lemmas .
the set @xmath206 consists of @xmath202-tuples of partial flags @xmath571 where , if the @xmath89th partition is special , @xmath562 and so @xmath572 , then @xmath573 is omitted as it is not needed to define the schubert variety indexed by @xmath463 .
however , to simplify the arguments that follow , we will write these flags as if they are all two - step flags . the resulting ambiguity may be remedied by setting @xmath574 when @xmath562 .
[ lem : gr3n_part1 ] if there is a partition @xmath259 in @xmath192 with @xmath559 , then @xmath575 . in all other cases
, @xmath576 is irreducible and has dimension @xmath516 .
first suppose that there is a partition @xmath259 in @xmath192 with @xmath559 .
if @xmath577 , then @xmath578 and @xmath579 .
in particular this means that if @xmath580 , then @xmath581 and so @xmath582 .
thus @xmath575 .
suppose now that every partition @xmath259 in @xmath192 has @xmath583 .
the only index @xmath470 for @xmath347 is when @xmath584 .
then @xmath585 satisfies and so @xmath586 has dimension @xmath516 . as the index @xmath470 is unique
, @xmath576 is irreducible and has dimension @xmath516 .
[ lem : gr3n_part2 ] there is a set @xmath587 of dimension @xmath516 if and only if there is a partition @xmath259 in @xmath192 with @xmath559 , and in that case the index @xmath470 is unique . first note that if there is a partition @xmath259 in @xmath192 with @xmath559 , then it is unique , say @xmath588 .
consider the index @xmath465 , where @xmath589 , and @xmath590 for all @xmath591 .
then @xmath470 satisfies with @xmath592 , and so @xmath587 has dimension @xmath516 .
conversely , suppose that the index @xmath470 satisfies for @xmath592 .
one of the sets @xmath593 , @xmath594 , or @xmath595 must be nonempty .
since @xmath592 , @xmath596 , because @xmath597 .
let @xmath598 and set @xmath599 , which has dimension 1 .
suppose that @xmath600 and that the flags are in general position .
then by the definition of @xmath601 and @xmath602 , we have @xmath603 thus @xmath604 and so @xmath605 subtracting from this and dividing by 2 gives @xmath606 .
if @xmath607 then @xmath608 and implies that @xmath609 , and thus @xmath192 is not a reduced schubert problem , which is a contradiction .
thus @xmath610 and we have @xmath611 .
if @xmath612 with @xmath613 , then implies that @xmath614 , which implies that @xmath192 is not reduced , a contradiction .
we are left with the case of @xmath615 .
then implies that @xmath616 , which completes the proof .
[ lem : gr3n_part3 ] no set @xmath617 of dimension @xmath516 maps dominantly to @xmath206 .
we first state an auxiliary lemma which will be used in the proof of lemma [ lem : gr3n_part3 ] .
[ lem : codim ] let @xmath93 be a schubert condition for @xmath6 , @xmath618 , and @xmath619 . setting @xmath620 and @xmath621 ,
then @xmath370 and @xmath622 satisfy schubert conditions @xmath259 and @xmath273 , respectively , with respect to the flag @xmath623 depending upon which subset @xmath459 they belong to according to the following table .
@xmath624 here , @xmath625 .
suppose that the index @xmath470 satisfies for @xmath626 .
we will show that the projection of @xmath617 to @xmath627 is not dense .
let @xmath618 and suppose that @xmath628 .
setting @xmath629 and @xmath630 then these lie in intersections of schubert varieties given by flags in @xmath631 as detailed in lemma [ lem : codim ] .
if we assume that the flags in @xmath632 are in general position so that those schubert varieties intersect in the expected dimensions , then we show that the nonemptiness of those intersections gives an inconsistent system of linear equations and inequalities , which proves the lemma .
the condition that the intersection involving @xmath370 is nonempty is @xmath633 subtracting from this and dividing by 2 gives @xmath634 the condition that the intersection involving @xmath622 is nonempty is @xmath635 adding to this and dividing by two gives @xmath636 since @xmath192 is a schubert problem on @xmath6 , @xmath637 and so we obtain @xmath638 , which contradicts .
thus the projection of the set @xmath617 to @xmath206 can not meet the subset consisting of @xmath202-tuples of flags that are in general position for these two intersections of schubert varieties .
we determine the dimensions of @xmath639 , @xmath640 , @xmath641 , and @xmath642 and then apply the definition to obtain the partitions encoding those conditions .
recall that @xmath643 and @xmath644 , where @xmath645 .
we use the notation of the proof of lemma [ lem : components ] , for @xmath406 , setting @xmath646 and @xmath647 .
then the conditions on @xmath648 are that @xmath649 and @xmath650 , for each @xmath406 . if we set @xmath433 for @xmath651 , then @xmath652\cap s|$ ] and similarly @xmath653\cap s|$ ] .
these are lower bounds on the dimension of @xmath641 and @xmath642 .
suppose first that @xmath654 so that @xmath655 , and so @xmath656 and @xmath657
. then the flags @xmath623 impose no conditions on @xmath370 so that @xmath658 . if @xmath659 , so that @xmath660 , then @xmath661 and @xmath273 is @xmath662 .
however , if @xmath663 so that @xmath664 then @xmath665 , and we have @xmath666 .
suppose that @xmath667 so that @xmath668 and this has dimension 1 , so that @xmath669
. then @xmath670 and @xmath671 .
as in the previous case , the dimension of @xmath641 depends upon whether or not @xmath663 , and so we get @xmath672 .
suppose that @xmath673 so that @xmath674 , but @xmath675 .
then @xmath676 .
we have that @xmath677 and @xmath678 , so that @xmath679 .
finally , suppose that @xmath680 so that @xmath681 and @xmath682
. then @xmath683 and @xmath684 .
the main result of @xcite is that every schubert problem on @xmath5 has at least alternating galois group .
the proof relied on vakil s criterion ( b ) and used elementary , but very involved , estimates of trigonometric integrals to establish an inequality between two kostka numbers .
we use the stronger and simpler vakil s criterion ( c ) ( which applies , by theorem [ th : gspec_double ] ) to give a much simpler and purely combinatorial proof of that result .
[ thm : g2nalternating ] every schubert problem on @xmath5 has at least alternating galois group .
it suffices to prove this for reduced schubert problems , by proposition [ prop : reduced ] . by definition
[ d : reduced ] , a reduced schubert problem on @xmath5 has the form @xmath685 with @xmath686 and @xmath687 for each @xmath122 , in particular , a reduced schubert problem is special .
let @xmath688 be the number of solutions to a special schubert problem @xmath685 on @xmath5 .
this is a kostka number , which counts the number of young tableaux of shape @xmath689 and content @xmath685 .
these are fillings of the young diagram of shape @xmath689 with @xmath690 ones , @xmath691 twos , through @xmath692 s s , such that the filling is weakly increasing along the rows and strictly increasing down the columns .
section 2.1 of @xcite shows that there is a fiber diagram , such that @xmath693 and @xmath694 are isomorphic to the galois groups of the schubert problems @xmath695 and @xmath696 , respectively .
this gives , @xmath697 the following proposition is proven in @xcite using a combinatorial injection .
[ prop : injection ] let @xmath698 be a reduced schubert problem on @xmath5 such that @xmath699 with @xmath700 .
then @xmath701 we compute some kostka numbers .
[ lem : multparts ] let @xmath702 be positive integers with @xmath703
. then 1 .
@xmath704 , 2 .
@xmath705 , 3 .
@xmath706 and @xmath707 , where @xmath708 is even , 4 .
@xmath709 , 5 .
@xmath710 when @xmath711 .
statements ( 1)(3 ) are from lemmas 8 and 9 of @xcite .
statement ( 4 ) is simple , as @xmath712 is the number of triples @xmath713 of nonnegative integers whose sum is @xmath134 .
a young tableau of shape @xmath689 and content @xmath714 is determined by its second row which consists of @xmath715 fives and @xmath134 remaining numbers which are some twos , threes , and fours . for ( 5 ) ,
we construct tableaux of shape @xmath689 and content @xmath716 from tableaux of shape @xmath717 and content @xmath718 and of shape @xmath719 and content @xmath720 .
let @xmath711 and consider a tableau @xmath721 of shape @xmath717 and content @xmath722 .
for each @xmath723 , construct a tableau @xmath724 of @xmath689 and content @xmath716 as follows .
the second row of @xmath721 ends in @xmath715 entries of @xmath725 .
remove these to get a tableau @xmath726 of shape @xmath727 and content @xmath720 , which we extend to get @xmath724 as follows .
place @xmath89 entries of @xmath725 and @xmath728 entries of @xmath729 at the end of the first row of @xmath726 and then @xmath728 entries of @xmath725 and @xmath89 entries of @xmath729 in the second row , followed by @xmath715 entries of @xmath730 to obtain @xmath724 .
@xmath731 this gives @xmath732 tableaux of shape @xmath689 and content @xmath716 .
given a tableau @xmath721 of shape @xmath719 and content @xmath720 , add @xmath134 entries of @xmath725 and @xmath134 of @xmath729 to the first row and @xmath715 entries of @xmath730 to get a tableau of shape @xmath689 and content @xmath716 .
@xmath733 this gives @xmath734 tableaux of shape @xmath689 and content @xmath716 , all of which are different from the others we constructed , which proves statement ( 5 ) .
we prove the result by induction on @xmath202 and @xmath2 .
let @xmath735 be a reduced schubert problem .
we may assume that @xmath736 , for otherwise @xmath737 and there is nothing to prove .
since @xmath304 is doubly transitive by theorem [ th : gspec_double ] , vakil s criterion ( c ) and schubert s recursion imply that @xmath304 is at least alternating if one of @xmath738 or @xmath739 is not six , for some reordering of the list @xmath740 .
suppose first that not all the @xmath490 are equal and that they are ordered so that @xmath699 with @xmath700 . by , @xmath688 is equal to either the sum , @xmath741 or the sum @xmath742 by proposition [ prop : injection ] , @xmath743 , so they can not both be six , which implies @xmath304 is at least alternating .
assume now @xmath744 for some positive integer @xmath134 .
there are only two values of @xmath745 for which @xmath746 , and for both of these @xmath747 , which implies that @xmath748 is at least alternating . indeed , by lemma [ lem : multparts ] , if @xmath749 , then @xmath750 . if @xmath751 , then @xmath752 only for @xmath753 , and then @xmath754 . if @xmath755 , then @xmath756 only for @xmath757 , and then @xmath758 .
we show that if @xmath759 , then @xmath760 . by lemma [ lem : multparts ]
( 5 ) , @xmath761 so it suffices to show @xmath760 for @xmath762 . when @xmath763 , this becomes @xmath764 and when @xmath765 , @xmath766 is even , and we have @xmath767 this completes the proof .
using vakil s criteria and the frobenius algorithm of @xcite ( a symbolic method to prove that a galois group is full symmetric by computing cycle types of elements ) , we study the galois group of every reduced schubert problem on @xmath0 . a nonreduced problem in @xmath0 is equivalent to a problem in some @xmath81 with @xmath768 and @xmath769 , and vakil earlier showed that these problems have at least alternating galois groups .
there are 2987 reduced schubert problems on @xmath0 having two or more solutions .
all except fourteen have at least alternating galois group with many known to be the full symmetric group .
each of these fourteen have imprimitive galois group , which we determine , and they fall into three families according to their geometry .
much of this computation , except for the determination of the galois groups of the last thirteen schubert problems , was done by vakil in @xcite .
vakil wrote a maple script based on his geometric littlewood - richardson rule and his criterion ( b ) to test which schubert problems had galois groups that are at least alternating @xcite .
we altered it to only test reduced schubert problems on @xmath0 and found 28 problems for which it was inconclusive .
we list them , using a compact product notation in which the schubert problem @xmath770 with 32 solutions is written @xmath771 . in subsection [ s : largesp ]
we show that the two schubert problems in the first row have at least alternating galois groups .
we used the frobenius algorithm @xcite to show that the next twelve have full symmetric galois group .
the remaining fourteen on the last four lines have imprimitive galois groups .
they are grouped by similar geometry , which we indicate by strings of equalities .
we describe one problem from each family in the remaining three subsections .
the schubert problem @xmath789 is special and therefore has doubly transitive galois group by theorem [ th : gspec_double ] .
it asks for the 4-planes @xmath145 in @xmath790 that meet 16 general 4-planes @xmath791 nontrivially .
if @xmath792 so that @xmath793 is a 3-plane and @xmath794 is a 5-plane , then lemma [
lem : speccomponents ] shows that @xmath795 thus if @xmath796 and @xmath797 is the restriction of @xmath203 to @xmath798 ( as in ) , then @xmath799 where @xmath800 and @xmath801 are ( essentially ) total spaces of the schubert problems @xmath802 these are equivalent dual problems and each was found to have at least alternating galois group by vakil s maple script .
as the original problem @xmath772 is doubly transitive , vakil s criterion ( c ) and remark [ r : vakil ] implies that its galois group is at least alternating .
the schubert problem @xmath803 also has doubly transitive galois group . to see this note that by lemma [ lem : components ] the subset @xmath804 of @xmath805 has dimension at most @xmath806 unless @xmath807 and @xmath808 , and that @xmath602 has dimension 9 .
since @xmath809 has dimension @xmath810 for @xmath495 , we see that the only set @xmath474 having dimension @xmath516 is when @xmath347 and each component of @xmath470 is @xmath811 .
the special position gives a subset @xmath812 with the restriction of @xmath203 to @xmath798 having two components , each essentially the total space of one of the schubert problems @xmath813 these equivalent dual schubert problems were found to have at least alternating galois group by vakil s maple script .
as the original problem @xmath773 is doubly transitive , vakil s criterion ( c ) implies that its galois group is at least alternating .
derksen discovered that this schubert problem has galois group @xmath814 and it was described by vakil @xcite .
an instance is given by four @xmath815-planes @xmath816 in general position in @xmath790 .
its solutions are those @xmath817 for which @xmath818 for @xmath819 .
consider the _ auxiliary _
problem @xmath820 in @xmath821 given by @xmath816 .
this asks for those @xmath822 with @xmath823 for @xmath819 .
there are four solutions @xmath824 to this problem , and its galois group is the full symmetric group @xmath814 .
each of the 4-planes @xmath825 will meet each @xmath536 in a 2-plane , and so they are solutions to the original problem .
in fact , they are the only solutions .
it follows that the galois group of @xmath785 is @xmath814 acting on the pairs @xmath826 .
this is an imprimitive permutation group as it preserves the partition @xmath827 of the solutions .
we also see that @xmath569 has two components .
exactly two sets @xmath828 and @xmath829 have dimension @xmath516 .
the structure of this problem shows that if the @xmath536 are real , then either two or all six of the solutions will be real .
indeed , if all four solutions @xmath830 to the auxiliary problem are real , than all six solutions @xmath831 will also be real .
if however , two or four of the @xmath830 occur in complex conjugate pairs , then exactly two of the @xmath831 will be real .
let @xmath833 , @xmath834 , and @xmath835 be general .
these give an instance of this schubert problem , @xmath836 any solution @xmath145 meets each of the 6-planes @xmath837 in a 3-plane and therefore their four - dimensional intersection @xmath838 in a 2-plane , @xmath839 .
then @xmath839 must meet the 2-planes @xmath377 , @xmath840 , @xmath841 , and @xmath842 , so it is a solution to the problem @xmath843 in @xmath844 . as the subspaces are in general position , this has two solutions @xmath845 and @xmath846 .
any solution @xmath145 to our original problem also meets each of the 2-planes @xmath847 and @xmath848 in a 1-plane , and therefore meets their span , @xmath849 , in a 2-plane , @xmath156 . as @xmath850 , they are in direct sum and @xmath145 is the span of @xmath156 and one of the @xmath830 .
fix @xmath851 and suppose that @xmath852 .
then @xmath145 meets each of @xmath853 for @xmath651 in a 3-plane and therefore @xmath145 meets each of the 2-planes @xmath854 and @xmath855 .
thus @xmath856 is a solution to the problem @xmath857 in @xmath858 .
this has two solutions , @xmath859 and @xmath860 .
the four 4-planes @xmath861 for @xmath862 are the solutions to our schubert problem .
since each element of the galois group either fixes @xmath845 and @xmath846 or it interchanges them , it preserves the partition @xmath863 and so it is imprimitive .
in fact , it is a subgroup of the dihedral group @xmath864 of symmetries of the square whose diagonals are the partition .
we verified that it was the dihedral group @xmath864 by reducing modulo primes @xmath865 and computing the cycles type of the resulting frobenius elements ( this method is described in @xcite ) and found elements of the galois group with cycle types @xmath866 as the only subgroup of @xmath867 having elements of these cycle types is @xmath864 itself , we conclude that the original schubert problem had galois group @xmath864 .
it is an exercise to verify that the three problems @xmath868 , @xmath869 , and @xmath870 , have nearly the same geometry behind their solutions and also have galois group @xmath864 .
this last problem was mentioned by billey and vakil @xcite , who asked if its galois group was indeed @xmath864 .
an instance of this problem is given by the choice of two 6-planes @xmath872 , two 2-planes @xmath873 and four 4-planes @xmath816 , all in general position .
solutions will be those @xmath817 such that @xmath874 for @xmath406 and @xmath875 .
consider the first four conditions in .
let @xmath876 , the linear span of @xmath877 and @xmath878 , which is isomorphic to @xmath879 .
then @xmath880 is two - dimensional .
if we set @xmath881 and @xmath882 , then @xmath883 , and so @xmath884 meets each of the four two - planes @xmath885 , so @xmath839 is a solution to the problem @xmath886 in @xmath887 given by @xmath885 , and therefore there are two solutions , @xmath845 and @xmath846 .
now set @xmath888 , which is four - dimensional , and fix one of the solutions to the problem of the previous paragraph . for each @xmath875 ,
set @xmath889 , which is two - dimensional .
these four two - planes are in general position and therefore give an instance of @xmath886 in @xmath890 .
let and be the two solutions to this problem , so that @xmath891 for each @xmath892 .
then the four subspaces @xmath893 are solutions to the original schubert problem . indeed , since @xmath894 for @xmath875 and @xmath895
, we have @xmath896 , and so @xmath897 satisfies the first four conditions of . since @xmath898 , @xmath899 does not meet @xmath900 for @xmath875 , so @xmath901 , which implies that @xmath902 , and shows that @xmath897 is a solution .
the galois group preserves the partition @xmath903 and so it is imprimitive .
the same arguments as before show that @xmath864 is its galois group .
it is also an exercise to show that each of the following eight schubert problems has a solution whose description is nearly identical to @xmath871 , and therefore has galois group @xmath864 . | we investigate double transitivity of galois groups in the classical schubert calculus on grassmannians .
we show that all schubert problems on grassmannians of 2- and 3-planes have doubly transitive galois groups , as do all schubert problems involving only special schubert conditions .
we use these results to give a new proof that schubert problems on grassmannians of 2-planes have galois groups that contain the alternating group .
we also investigate the galois group of every schubert problem on @xmath0 , finding that each galois group either contains the alternating group or is an imprimitive permutation group and therefore fails to be doubly transitive .
these imprimitive examples show that our results are the best possible general results on double transitivity of schubert problems . |
numerous processes in meson physics are related to the adler - bell - jackiw ( abj ) axial anomaly @xcite appearing in the fermion loops connecting certain number of axial ( a ) and vector ( v ) vertices . concretely , in this paper we will deal with the processes related to the avv ( triangle " , fig . [
fig : triangle ] ) and vaaa ( box " , fig .
[ fig : box ] ) anomaly , exemplified by the famous @xmath1 and @xmath2 transitions . with two photons of momenta @xmath3 and @xmath4 .
the quark - photon coupling is in general given by dressed vector vertices @xmath5 , which in the free limit reduce to @xmath6.,height=151 ] , studied by the pseudoscalar coupling method in , e.g. , refs .
there are six different contributing graphs , obtained from the above graph by the permutations of the vertices of the three different pions .
the position of the @xmath7 and @xmath8 quark flavors on the internal lines , as well as @xmath9 or @xmath10 quark charges in the quark - photon vertex , varies from graph to graph , depending on the position of the quark - pion vertices .
the physical pion fields are @xmath11 and @xmath12 .
the momenta flowing through the four sections of the quark loop are conveniently given by various combinations of the symbols @xmath13 in @xmath14.,height=151 ] suppose one wants to describe such processes using qcd - related effective chiral meson lagrangians @xcite without adding ad hoc interactions of mesons with external gauge fields to reproduce empirical results .
for example , one can add by hand @xmath15 and this would reproduce the observed @xmath1 width for the favorable value of the @xmath16 coupling @xmath17 . however ,
if one does not want to add such ad hoc terms in the effective meson lagrangians , one must describe such `` anomalous '' processes through the term derived by wess and zumino @xcite . on the other hand , if one wants to utilize and explicitly take into account the fact that mesons are composed of quarks , another way of describing these processes is optimal in our opinion , and the main purpose of this paper is to stress and elucidate this .
axial vertices in the anomalous graphs such as the avv and vaaa ones , couple the quarks with pseudoscalar mesons . instead of anomalous graphs ,
another way to study the related amplitudes involving pseudoscalar mesons , is to calculate the corresponding graphs where axial vertices ( a ) are replaced by pseudoscalar ( p ) ones .
thereby , for example , the @xmath1 decay amplitude due to the avv triangle anomaly " , @xmath18 is reproduced by the calculation of the pvv triangle graph .
( [ pi2gammaamp ] ) pertains to the chiral limit , where the pion mass @xmath19 .
also , @xmath20 is the pion decay constant , @xmath21 is the proton charge , and @xmath22 is the number of quark colors . ]
the pvv triangle graph calculation of eq .
( [ pi2gammaamp ] ) can most simply be done essentially _ ` a la _
steinberger @xcite , that is , with a loop of `` free '' constituent quarks with the point pseudoscalar coupling ( i.e. , @xmath23 , where @xmath24 ) to quasi - elementary pion fields .
however , since the development of the dyson - schwinger ( ds ) approach to quark - hadron physics @xcite , the presently advocated method becomes even more convincing .
namely , the ds approach clearly shows how the light pseudoscalar mesons simultaneously appear both as quark - antiquark ( @xmath25 ) bound states and as goldstone bosons of the dynamical chiral symmetry breaking ( d@xmath26sb ) of nonperturbative qcd .
the solutions of bethe - salpeter ( bs ) equations for the bound - state vertices of pseudoscalar mesons then enter in the pvv triangle graph instead of the point @xmath23 coupling , and the current algebra result ( [ pi2gammaamp ] ) is again reproduced exactly and analytically , which is unique among the bound - state approaches . that the ( almost massless ) pseudoscalars are ( quasi-)goldstone bosons , is also a unique feature among the bound - state approaches to mesons .
a reason why the p - coupling method is simpler both technically and conceptually is that the pvv triangle graph amplitude is _ finite _ , unlike the avv one , which is divergent and therefore also ambiguous with respect to the momentum routing .
reference gives a more detailed discussion of the p - coupling method and why is that it is equivalent to the anomaly calculations , and illustrates this on the examples of the famous decay @xmath27 and processes of the type @xmath2 .
also , the pvv quark triangle amplitude leads to many ( over 15 ) decay amplitudes in agreement with data to within 3% and not involving free parameters @xcite .
this will be elaborated in more detail in sec .
[ manyproc ] .
additional advantages of this method is that its treatment of the @xmath28-@xmath29 complex and resolution of the @xmath30 problem , goes well with the absence of axions ( which were predicted to solve the strong cp problem but have _ not _ yet been observed @xcite ) and with the arguments of ref . , that there is really no strong cp problem .
all this will be discussed in sec .
[ sec : commreltoglounanom ] .
we state our conclusions in sec . [
sec : summary ] .
in this section we calculate the amplitudes for a number of processes using the quark triangle graphs .
figures [ fig : triangle ] and [ fig : pvvdiagrams ] show three such pvv processes .
first we consider @xmath31 decay via the @xmath7 and @xmath8 quark triangle graph for @xmath32 , @xmath22 and goldberger - trieman relation leading to the pion decay constant : @xmath33 .
this amplitude is finite and for the experimental value of the pion decay constant , @xmath34 @xcite , gives @xcite the chiral - limit amplitude ( [ pi2gammaamp ] ) of magnitude [ cols="^,^ " , ] @xmath35 very close to experimental data @xcite @xmath36^{1/2 } = ( 0.0252\pm 0.0009)~{\mbox{\rm gev}}^{-1}~.\ ] ] likewise , the @xmath7 , @xmath8 quark triangles for @xmath37 decay give @xcite @xmath38 for @xmath39 found from @xmath40 decay @xcite : @xmath41 the calculated @xmath42 is also near data @xcite , @xmath43^{1/2 } = ( 0.225\pm 0.011)~{\mbox{\rm gev}}^{-1}~,\ ] ] where @xmath44 is the photon momentum .
[ actually , the above value is a weighted average of @xmath45 and @xmath46 amplitudes . ]
next we predict the @xmath7 , @xmath8 quark triangle amplitude for @xmath47 taking @xmath48 as 99% nonstrange @xcite ( @xmath49 ) @xmath50 for @xmath51 found from @xmath52 decay .
the mixing angle is @xmath53 again this theory in eq .
( [ fomegatopi0gamma ] ) is near data @xmath54 @xcite
. other pvv photon decays involve the @xmath28 and @xmath55 mixed non strange and @xmath56 pseudoscalar mesons .
again the quark triangle amplitudes are a close match with data @xcite .
the quark
triangle ( qt ) calculation gives reliable predictions also for the @xmath28 and @xmath55 two photon decays : @xmath57 this should be compared with the experimental data : @xmath58^{1/2 } = ( 0.02498\pm 0.00064)~{\mbox{\rm gev}}^{-1}~ , \\
= \left[\frac{64\pi\gamma(\eta^\prime\to\gamma\gamma)}{m_{\eta^\prime}^3}\right]^{1/2 } = ( 0.03133\pm 0.00055)~{\mbox{\rm gev}}^{-1}~,\end{aligned}\ ] ] where @xmath59 and @xmath60 .
the ratio of the constituent quark masses is @xmath61 for @xmath62 and @xmath63 @xcite .
the mixing angle is @xcite @xmath64 next , we can calculate the @xmath65 amplitude employing the quark triangle diagram , @xmath66 again , this is close to the experimental data , @xmath67^{1/2 } = ( 0.48\pm 0.03)~{\mbox{\rm gev}}^{-1}~,\ ] ] where @xmath68 is the photon momentum and @xmath69 .
a similar situation is with the @xmath70 amplitude , for which the quark triangle calculation gives @xmath71 the corresponding experimental value is @xmath72^{1/2 } = ( 0.411\pm 0.017)~{\mbox{\rm gev}}^{-1}~,\ ] ] where @xmath73 is the photon momentum and @xmath74 is the experimental decay width @xcite .
the @xmath75 amplitude is @xmath76 where @xmath77 .
the @xmath75 decay width is @xmath78 where @xmath79 @xcite .
this is in a good agreement with the experimental value @xmath80 revealing that the vector meson dominance is the main effect , while the coupling through vppp quark box loop ( contact term " ) contributes little .
it is known that @xmath81 decay is dominated by @xmath82meson poles .
the required @xmath83 amplitude can be estimated as @xmath84 but can not be measured because there is no phase space for this process .
the @xmath83 amplitude is more precisely defined with ql , additionally enhanced with a meson loop associated with sigma exchange @xcite , @xmath85 the scalar amplitude @xmath86 is dominated by the @xmath82 meson in each of the three possible channels @xcite , @xmath87 \nonumber \\ & \approx & 1480~{\mbox{\rm gev}}^{-3}~.\end{aligned}\ ] ] following thew s phase space analysis @xcite , we get @xmath88 where @xmath89 is used .
the predicted value is close to the experimental value @xcite @xmath90 here we have taken @xmath48 as pure ns , although it is about 99% ns , since @xmath91 from our eq .
( [ phiv ] ) . in the quark
level @xmath92model a quark box diagram contributes to the @xmath81 decay .
this box diagram can be interpreted as a contact term .
it is shown that the contact contribution is small by itself , but can be enlarged through the interference effect @xcite .
using @xmath93 from our eq .
( [ phip ] ) , we predict the tensor @xmath94 branching ratios for @xmath95 : @xmath96 for center of mass momenta @xmath97 , @xmath98 , @xmath99 . the above data branching ratios follow from @xmath95 recent fractions @xcite : @xmath100 , @xmath101 and @xmath102 .
the approach using the pseudoscalar coupling is , in our opinion , also relevant for the effects related to the non - abelian , gluon " abj axial anomaly . here , we comment on this only briefly , and direct the reader to the original references for details .
the first point concerns the @xmath28-@xmath29 complex and the @xmath0 problem related to it . in the chiral limit @xmath103 , since all members of the flavor - su(3 ) pseudoscalar meson octet are massless in this theoretical , but very useful limit .
the only non - vanishing ground - state pseudoscalar meson mass in this limit is the mass of the su(3)-singlet pseudoscalar meson @xmath104 .
this is thanks to the non - abelian , gluon abj axial anomaly , i.e. , to the fact that the divergence of the su(3)-singlet axial current @xmath105 receives the contributions from both the finite quark masses and gluon fields @xmath106 : @xmath107 this removes the @xmath0 symmetry and explains why only eight pseudoscalar mesons are light , and not nine ; i.e. , why there is an octet of ( almost-)goldstone bosons , but not a nonet .
the physically observed @xmath28 and @xmath29 are then the mixtures of the anomalously heavy @xmath104 and ( almost-)goldstone @xmath108 in such a way that @xmath29 is predominantly @xmath104 and @xmath28 is predominantly @xmath108 .
this is how the gluon anomaly can save us from the @xmath0 problem in principle , and the details of how we achieve a successful description of the @xmath28-@xmath29 complex , are given in refs . .
here we just sketch some important points .
the mass matrix squared @xmath109 in the quark basis @xmath110 , @xmath111 , @xmath112 is @xmath113 + \beta \left [ \begin{array}{ccl } 1 & 1 & x \\ 1 & 1 & x \\ x & x & x^2 \end{array } \right]~,\ ] ] where @xmath114 is the non - anomalous part of the matrix , since @xmath115 and @xmath116 would be the masses of the respective non - strange " ( ns ) and strange " ( s ) pseudoscalar @xmath117 mesons if there were no gluon anomaly . in the ns sector , in the isospin symmetry limit ( which is very close to reality ) , the relevant combinations are @xmath118 as the neutral partner of the charged pions @xmath119 in the isospin 1 triplet , and the isospin 0 combination @xmath120 . in the absence of gluon anomaly , but with an @xmath121-quark mass heavier than the isosymmetric @xmath7 and @xmath8 ones , @xmath28 would reduce to @xmath122 with the mass @xmath123 , and @xmath29 to @xmath124 with the mass @xmath125 .
both of these assignments are in conflict with experiment .
the realistic contributions of various flavors to @xmath28 and @xmath29 and their masses ( i.e. , the realistic @xmath28-@xmath29 mixing ) are obtained only thanks to @xmath126 , the anomalous contribution to the mass matrix . in @xmath126
, the quantity @xmath127 describes transitions @xmath128 ( @xmath129 ) due to the gluon anomaly and @xmath130 describes the effects of the su(3 ) flavor symmetry breaking on these transitions . in refs .
, as the first step in solving the @xmath0 problem , we extract @xmath108 , @xmath104 masses from the @xmath28 , @xmath55 via @xmath131 where @xmath132 . the mesons @xmath108 and @xmath104 are defined as latexmath:[\[\begin{aligned } the @xmath108 meson mass ( [ meta8 ] ) @xmath134 is 4.56% greater than the observed @xcite @xmath135 .
the singlet @xmath104 mass ( [ meta1 ] ) @xmath136 is only 1.56% below the observed @xmath137 and close to the nonstrange@xmath56 mixing @xmath0 mass dictated by phenomenology @xcite @xmath138^{1/2 } = 915.31~{\mbox{\rm mev}}~ , \label{secondu1amass}\ ] ] ( this is also close to @xmath139 , which is the mass found in the analogous ds approach @xcite . ) we call the quantity ( [ secondu1amass ] ) the mixing @xmath0 mass " since the mass matrix ( which is especially clear in the nonstrange - strange quark basis ) reveals that @xmath140 induces the mixing between the nonstrange isoscalar @xmath141 and @xmath56 quark - antiquark states .
equivalently , @xmath140 can be viewed as being generated by the transitions among the @xmath142 , @xmath143 and @xmath56 pseudoscalar states ; via quark loops , these pseudoscalar @xmath144 bound states can annihilate into gluons which in turn via another quark loop can again recombine into another pseudoscalar @xmath145 bound state of the same or different flavor .
the quantity @xmath127 appearing in eq .
( [ secondu1amass ] ) is then the annihilation strength of such transitions , in the limit of an exact su(3 ) flavor symmetry .
( the realistic breaking of this symmetry is easily introduced and improves our description of the @xmath28-@xmath29 complex considerably . )
the `` diamond '' graph in fig .
[ fig : diamond ] gives just the simplest example of such an annihilation / recombination transition . since these annihilations occur in the nonperturbative regime of qcd , all graphs with any even number of gluons instead of just those two in fig .
[ fig : diamond ] , can be just as significant in annihilating and forming a @xmath146 pseudoscalar @xmath144 meson .
indeed , this nonperturbative @xmath0 mass scale , eq .
( [ secondu1amass ] ) , is still 3 times higher than the gluon `` diamond '' graph evaluated perturbatively @xcite .
thus , we can not calculate @xmath147 and the situation is much more complicated and less clear than in the abelian case , as explained in @xcite .
can it then be founded to think that the annihilation graphs with the pseudoscalar meson - quark coupling , such as the `` diamond '' graph in fig .
[ fig : diamond ] , give rise to the anomalous term @xmath148 in the divergence ( [ divganom ] ) of the su(3)-singlet current @xmath149 , and thus ultimately to the large mass of @xmath150 ( and of the observed @xmath29 ) ?
well , this conjecture may remain a speculation since we can not calculate @xmath127 due to the nonperturbative nature of the problem . nevertheless , when we use it in our approach as a parameter with the value given by eq .
( [ secondu1amass ] ) , we obtain a very good description of the @xmath28-@xmath29 complex phenomenology @xcite .
this includes not only the masses of @xmath28 and @xmath29 , but also their @xmath151 decay widths , and the mixing angle @xmath152 consistently following from the masses and @xmath151 widths .
this gives a strong motivation for the above conjecture .
pseudoscalar @xmath153 is coupled to a quark loop , whereby it can annihilate into gluons which in turn recombine into the pseudoscalar @xmath154 having the flavor content @xmath155.,height=113 ] we should also note that our conjecture in the previous subsection goes well with the arguments of banerjee _ et al . _
@xcite , that there is really no strong cp problem .
they find that one does _ not _ need vanishing @xmath156 ( where @xmath157 is the quark mass matrix ) .
thus , one does not need any fine - tuning , and all cp violation in the qcd lagrangian can be avoided by having @xmath158 in its cp - violating term @xmath159 this term in the qcd lagrangian breaks the @xmath0 symmetry and corresponds to the anomalous term @xmath160 in the divergence ( [ divganom ] ) of the singlet current .
the term ( [ thetaqcd ] ) is allowed by gauge invariance and renormalizability , but apparent nonexistence of the strong cp violation , and also of axions , is the solid reason to have it vanishing . our conjecture , that p - coupled annihilation graphs reproduce the effect of the gluon abj anomaly , naturally agrees with the vanishing of this term and with putting the case of the strong cp problem to rest _ la _ banerjee _ et al .
we have presented and surveyed in detail the method of pseudoscalar coupling of pseudoscalar mesons to the triangle " and box " quark loops .
we have reviewed how this method gives the equivalent results to the anomaly calculations .
the p - coupling method has also been illustrated on the example of many decay amplitudes .
the avv anomaly @xcite involves 10 invariant amplitudes ( reduced to 1 or 2 amplitudes for @xmath31 decay using additional ward identities ) .
if instead one considers the pvv transition with a pseudoscalar coupling , then the pvv quark triangle amplitude is finite and leads to many decay amplitudes ( over 15 ) then in agreement with data to within 3% and not involving free parameters @xcite . to solve instead the former avv decay problem , very light axion bosons
have been predicted but have _ not _ yet been observed @xcite . also , there is the @xmath0 and @xmath161 problem involving gluons whereby strong interaction qcd leads to cp violation , definitely a `` strong cp problem '' because cp violation is known to occur at the @xmath162 weak interaction amplitude level @xcite .
physicists have tried to circumvent this `` @xmath0 strong cp problem '' either via the topology of gauge fields or by investigating the @xmath161vacuum for this strong cp problem @xcite . in this paper
we have circumvented the need to deal directly with the above photon or gluon avv anomalies by studying instead ( finite ) pvv quark triangle graphs .
then we have given our phenomenological results which always are in approximate agreement with the data .
next we return to the @xmath0 problem and again use quark triangle diagrams coupled to 2 gluons . invoking nonstrange
strange particle mixing , the predicted @xmath0 mass is within 3% of data @xcite .
thus we circumvent both photon and , admittedly on a much more speculative level , also the gluon abj anomaly without resorting either to unmeasured axions or to a strong cp violating term in the qcd lagrangian .
r. delbourgo , dongsheng liu and m. d. scadron , int .
a * 14 * , 4331 ( 1999 ) [ arxiv : hep - ph/9905501 ] ; m. d. scadron , f. kleefeld , g. rupp and e. van beveren , nucl .
a * 724 * , 391 ( 2003 ) [ hep - ph/0211275 ] . | many meson processes are related to the @xmath0 axial anomaly , present in the feynman graphs where fermion loops connect axial vertices with vector vertices .
however , the coupling of pseudoscalar mesons to quarks does not have to be formulated via axial vertices .
the pseudoscalar coupling is also possible , and this approach is especially natural on the level of the quark substructure of hadrons . in this paper
we point out the advantages of calculating these processes using ( instead of the anomalous graphs ) the feynman graphs where axial vertices are replaced by pseudoscalar vertices .
we elaborate especially the case of the processes related to the abelian axial anomaly of qed , but we speculate that it seems possible that effects of the non - abelian axial anomaly of qcd can be accounted for in an analogous way . |
the magnetocaloric effect ( mce ) , which is characterized by an isothermal change of the entropy or an adiabatic change of the temperature upon magnetic - field variation , enjoys a great scientific interest mainly because of its immense application potential @xcite . besides a conventional mce observed in regular ferromagnets or paramagnets ,
there may also be detected an inverse mce , namely , in ferrimagnetic or antiferromagnetic materials . in the former case
the system cools down when the magnetic field is removed adiabatically , while in the latter case it heats up .
however , the mce has been so far rigorously studied only in one - dimensional spin systems ( see e.g. ref .
@xcite and references therein ) due to a lack of exactly solved spin models in higher dimensions accounting for a non - zero magnetic field .
theoretical description of the conventional and inverse mce in two- and three - dimensional magnetic systems is thus usually based on some approximative method @xcite .
the main goal of this work is to investigate the mce in the exactly solved spin-1/2 fisher s super - exchange antiferromagnet @xcite , which allows an exact theoretical description of this phenomenon also in a two - dimensional spin model .
the spin-1/2 fisher s super - exchange antiferromagnet represents the spin-1/2 ising model on a decorated square lattice , in which the antiferromagnetic ( ferromagnetic ) coupling @xmath1 on horizontal ( vertical ) bonds are supposed together with the external magnetic field @xmath2 acting on decorating spins .
the hamiltonian of the model reads @xmath3 in above , @xmath4 labels the decorated ising spin at @xmath5th horizontal ( @xmath6th vertical ) bond and @xmath7 denotes the nodal ising spin at @xmath8th site of the original square lattice . the first ( second ) summation in the hamiltonian ( [ eq : h ] )
is thus carried out over nearest - neighbour lattice sites on the horizontal ( vertical ) bonds , while the third term represents the zeeman s energy of the decorating spins @xmath9 . at zero temperature ,
the system passes from the antiferromagnetically ordered ground state to the paramagnetic one when the magnetic field applied on decorating spins exceeds the critical value @xmath10 .
the former ground state is characterized by a perfect antiferromagnetic arrangement of decorating spins placed on horizontal and vertical bonds , while in the latter ground state the system is broken into a set of @xmath11 spins polarized towards the magnetic - field direction due to a frustration of the nodal spins @xmath12 . at finite temperatures , the existence of the antiferromagnetic long - range order terminates at the critical temperature @xmath13 of the second - order phase transition , which monotonously decreases with increasing the magnetic field until it entirely tends to zero at @xmath10 .
since the fisher s super - exchange model ( [ eq : h ] ) is exactly solvable within the decoration - iteration mapping transformation ( for more computational details see works ) , it provides an excellent paradigmatic example of an exactly soluble two - dimensional spin system , which allows an examination of the mce in a vicinity of the continuous ( second - order ) phase transition at non - zero magnetic fields .
actually , the magnetocaloric quantities , such as the isothermal entropy change @xmath14 and the adiabatic temperature change @xmath15 upon the magnetic - field variation @xmath16 can be rigorously calculated by using the formulas : @xmath17 \label{eq : dt } \delta t_{ad}(s , \delta h)&= & t(s , h\!\neq0 ) - t(s , h\!=0).\end{aligned}\ ] ] recall that the former relation ( [ eq : ds ] ) is valid if the temperature @xmath18 of the model is constant , while the latter one ( [ eq : dt ] ) satisfies the adiabatic condition .
figure [ fig1 ] illustrates temperature dependencies of the isothermal entropy change ( @xmath19 ) normalized per site of the original square lattice for various values of the magnetic - field change @xmath16 .
crosses on relevant curves mark weak singularities of the zero - field entropy found at the critical temperature @xmath20 , while open circles denote weak singularities of the entropy located at critical points of second - order phase transitions at finite magnetic fields @xmath21 and @xmath22 . as one can see
, the magnetocaloric potential @xmath19 may be either positive or negative depending on the temperature , which points to both conventional ( @xmath23 ) and inverse ( @xmath24 ) mce for any value of @xmath25 . in the high - temperature region @xmath26 , where only short - range ordering occurs
, @xmath19 slowly increases to the broad maximum with decreasing temperature due to suppression of a spin disorder by the applied magnetic field . at certain temperature , @xmath19 starts to decrease and changes sign from positive to negative as @xmath18 further decreases .
the zero - temperature limits of @xmath19 are consistent with the entropy values @xmath27 , @xmath28 and @xmath29 of the antiferromagnetically ordered ground state ,
the paramagnetic ground state and the coexistence of both phases at a first - order phase transition , respectively ( for more details see ref .
minima in low - temperature parts of @xmath30 curves observed around the temperature interval for magnetic - field changes @xmath31 clearly indicate a large inverse mce slightly above the second - order phase transition ( see fig .
[ fig1]a ) .
the origin of this phenomenon can be attributed to strong thermal fluctuations of spins leading to an unusual thermally - induced increase of total magnetization in this region ( compare @xmath30 curves plotted in fig . [ fig1]a with thermal variations of total magnetization shown in fig .
9 of ref .
@xcite ) . in accordance with this statement ,
the inverse mce gradually increases and shifts to lower temperatures upon the increase of the field change @xmath25 .
it is obvious from fig .
[ fig1 ] that @xmath32 is always satisfied .
thus , one may conclude that the most pronounced inverse mce can be found for @xmath33 , which exactly coincides with the critical field @xmath34 of the first - order phase transition between the magnetically ordered and paramagnetic ground states . if @xmath35 , the inverse mce ( minimum in @xmath30 curves ) is gradually reduced with the increasing @xmath25 due to weakening of thermal excitations from paramagnetic ground state towards the antiferromagnetically ordered excited state ( see fig .
[ fig1]b ) . to discuss the mce
, one may alternatively investigate the adiabatic temperature change @xmath15 of the system at various magnetic - field changes @xmath16 .
temperature variations of this magnetocaloric potential for the considered spin model are displayed in fig .
all curves plotted in fig .
[ fig2 ] were calculated using eq .
( [ eq : dt ] ) by keeping the entropy constant .
crosses and open circles on relevant curves determine positions of critical temperatures for @xmath36 and @xmath37 , respectively .
obviously , the adiabatic temperature change @xmath15 clearly allows to distinguish the conventional mce ( @xmath38 ) from the inverse mce ( @xmath39 ) . in accordance to the previous discussion
, the investigated model heats up faster in a vicinity of the first - order phase boundary between antiferromagnetically ordered ground state and the paramagnetic ground state achieved upon the adiabatic reduction of the magnetic field .
indeed , the magnitude of the negative peak in @xmath40 curves gradually increases with magnetic - field change and shifts towards the zero - field critical temperature @xmath20 as the applied magnetic field approaches the critical value @xmath34 ( see fig .
[ fig2]a ) .
in addition , @xmath15 versus temperature plot ends at zero value in the asymptotic limit of zero temperature for any @xmath31 , which can be attributed to a perfect antiferromagnetic order of decorating spins placed on horizontal and vertical bonds of the square lattice at zero temperature .
by contrast , the adiabatic temperature change rapidly drops to the finite values @xmath41 and @xmath42 at the temperatures @xmath43 and @xmath44 for the magnetic - field changes @xmath45 ( see fig .
[ fig2]b ) . in this particular case
, the magnetocaloric potential @xmath15 can not be defined below aforementioned temperatures , because there is no temperature end point in the adiabatic process if @xmath45 .
this intriguing behaviour is evidently caused by residual entropies @xmath46 and @xmath29 detected within the paramagnetic ground state and the coexistence point at the first - order phase transition @xmath34 , respectively .
in this paper , we have investigated magnetocaloric properties of the exactly solved spin-1/2 fisher s super - exchange antiferromagnet by using the known exact solution for the magnetic entropy of the model @xcite .
the temperature dependencies of the isothermal entropy change and the adiabatic temperature change have been particularly examined for various values of the magnetic - field change . this study has enabled us to clarify the magnetic refrigeration efficiency of the model in a vicinity of the critical temperature of the second - order phase transition , which completely destroys the antiferromagnetic long - range order .
the obtained results for both magnetocaloric potentials clearly indicate on the fast heating of investigated spin system during the adiabatic demagnetization process ( on a presence of the enhanced inverse mce ) in this region due to strong thermal spin fluctuations leading to the thermally - induced increase of the total magnetization .
the maximal heating efficiency of the system has been observed for the magnetic - field change @xmath33 , which coincides with the critical field @xmath34 of the first - order phase transition between the antiferromagnetically ordered and paramagnetic phases . + * acknowledgments * : this work was financially supported by ministry of education , science , research and sport of the slovak republic provided under the vega grant no .
1/0043/16 and by the grant slovak research and development agency provided under the contract no .
apvv-0097 - 12
. 6 a. m. tishin , y. i. spichkin , _ the magnetocaloric effect and its applications _
, doi : 10.1016/s1369 - 7021(03)01134 - 9 iop pub . ,
bristol , philadelphia , 2003 . l. glisov , j. streka , _ phys . lett .
a _ * 379 * , 2474 ( 2015 ) .
doi : 10.1016/j.physleta.2015.07.007 .
y. hu , a. du , _ j. phys . : condens
. matter _ * 20 * , 125225 ( 2008 ) .
doi : http://dx.doi.org/10.1088/0953-8984/20/12/125225 m. e. fisher , _ proc .
a _ * 254 * , 66 ( 1960 ) .
10.1098/rspa.1960.0005 m. e. fisher , _ proc .
a _ * 256 * , 502 ( 1960 ) .
doi : 10.1098/rspa.1960.0122 | the isothermal entropy change and the adiabatic temperature change are rigorously calculated for the exactly solved spin-1/2 fisher s super - exchange antiferromagnet in order to examine magnetocaloric properties of the model in a vicinity of the second - order phase transition .
it is shown that the large inverse magnetocaloric effect occurs around the temperature interval for any magnetic - field change @xmath0 .
the most pronounced inverse magnetocaloric effect can be found for the magnetic - field change , which coincides with the critical field of a zero - temperature phase transition from the antiferromagnetically ordered ground state to the paramagnetic one . |
gas giant planets form in gaseous disks surrounding newly born stars .
as planets form in disks , their gravity perturbs the surrounding material and produces structures such as spiral density waves and gaps ( e.g. , * ? ? ?
these planet - induced structures are detectable in high angular resolution , direct imaging observations of disks at near - infrared ( nir ) wavelengths ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , which are sensitive to scattered light from the disk surface .
recently , spiral arms and gaps that are expected to result from planets have been found in disks surrounding ab aur @xcite , mwc 758 @xcite , sao 206462 @xcite , hd 100453 @xcite , hd 100546 @xcite , ak sco @xcite , and tw hya @xcite .
high angular resolution , high contrast scattered light images of disks are made possible by instruments such as vlt / naco @xcite , subaru / hiciao @xcite , vlt / sphere @xcite , gemini / gpi @xcite , and subaru / scexao @xcite . to anticipate how companion - induced disk structures may appear in nir images ,
we need synthetic observations . combining hydrodynamic simulations of disk - planet interactions with monte carlo radiative transfer ( mcrt ) calculations
, @xcite explored the appearance of gaps in scattered light images , and showed that the cavities in transitional disks @xcite can be explained as wide common gaps opened by multiple giant planets @xcite . by comparing the synthetic images of gapped disks in that study with the new gpi observations of tw hya
, @xcite concluded that the gap at 21 au in the system may be explained by a @xmath2 planet . using a similar method
, @xcite examined the morphology of the spiral shocks excited by a giant companion in the outer disk , finding two prominent arms in a near @xmath3 rotational symmetry in scattered light images ( see also * ? ? ? * ; * ? ? ? * ) , that resemble the arms discovered in mwc 758 @xcite and sao 206462 @xcite .
this giant - companion double - arm scenario was verified by @xcite in sphere images of the hd 100453 system @xcite , in which an m dwarf companion is seen together with the two arms it drives .
past models have focused on producing synthetic observations for disks viewed face - on ( but see e.g. , @xcite ) . in this work
we expand the parameter space to arbitrary inclinations and position angles .
we carry out three - dimensional ( 3d ) hydro simulations using the code ` penguin ` @xcite to calculate the density structure of spiral arms and gaps produced by a giant planet or a stellar companion on a circular orbit in the outer disk ( section [ sec : simulations ] ) .
the numerical disk models are subsequently read into the @xcite 3d monte carlo radiative transfer ( mcrt ) code to produce synthetic images at nir wavelengths . in section [ sec : results ] , we examine the morphology of the disk over a large grid of viewing geometries . in section [ sec : discussions ] , after criticizing some common reconstruction techniques for deprojecting inclined disks , we produce synthetic observations for ab aur and ak sco , comparing them with their respective subaru / hiciao @xcite and vlt / sphere observations @xcite .
a summary is given in section [ sec : summary ] .
we carry out 3d hydrodynamics simulations to calculate disk structures induced by a companion .
the resulting disk structures are fed into 3d mcrt simulations to produce synthetic images at several nir bands .
our numerical methods are described below .
we use the 3d lagrangian - remap hydrodynamics code ` penguin ` @xcite to simulate spiral arms in disks for two different models . in the first model
, we have a companion - to - star mass ratio , @xmath4 , of @xmath5 .
we call this the `` @xmath1 model '' ; the companion s mass is @xmath1 if the star is @xmath6 . for our `` @xmath7 model , ''
we have @xmath8 .
these two choices of @xmath4 are motivated by the three best studied double - arm systems to date : mwc 758 @xcite and sao 206462 @xcite which most likely harbor planetary mass companions with @xmath9 @xcite , and hd 100453 @xcite whose arms are almost certainly driven by the @xmath10 m dwarf companion @xcite .
both simulations are run for 100 orbits of the companion , at which point the disk structures have long settled into quasi - steady states . except for figure
[ fig : image_100mj_orbits ] , all figures in this paper are produced using the models after 100 companion orbits have elapsed .
` penguin ` solves the same equations as in @xcite , except in this paper all simulations are done in 3d , and with different input parameters .
we provide some details here .
we use @xmath11 to denote the usual radial , azimuthal , and polar coordinates , and @xmath12 for the cylindrical radius .
the lagrangian continuity and momentum equations solved by ` penguin ` are : @xmath13 where @xmath14 is the gas density , @xmath15 the velocity field , @xmath16 the gas pressure , @xmath17 the newtonian stress tensor , and @xmath18 the combined gravitational potential of the star and the companion .
we adopt a locally isothermal equation of state , such that @xmath19 , where @xmath20 is the sound speed of the gas .
@xmath17 is proportional to the kinematic viscosity @xmath21 , which we parameterize using the @xmath22-prescription by @xcite , such that @xmath23 , where @xmath24 is the orbital frequency of the disk . in all our simulations ,
we choose @xmath25 .
this relatively high viscosity is chosen to isolate the spirals excited by the companion from additional disk features caused by hydrodynamical instabilities , such as the rossby wave instability @xcite triggered by the formation of a sharp gap edge .
a higher viscosity also shortens the system s viscous timescale , allowing the simulation to converge faster with time .
the simulations are performed in the frame of the central star , so for a companion on a fixed , circular orbit , @xmath18 is : @xmath26 \,,\ ] ] where @xmath27 is the gravitational constant , @xmath28 the total mass of the star and the companion , @xmath29 the semi - major axis of the companion s orbit , @xmath30 the softening length of the companion s potential , and @xmath31 denotes the azimuthal separation from the companion . for convenience
, we also denote the keplerian velocity and frequency as @xmath32 and @xmath33 .
the third term in the bracket is the indirect potential due to the acceleration of the frame .
@xmath30 is chosen to be @xmath34 of the companion s hill radius , so for the @xmath35 model , we have @xmath36 , and for the @xmath7 model , @xmath37 .
we note that the companion is introduced into the simulation gradually @xmath4 increases from zero to the desired value over 5 orbits . because we adopt different disk sound speed profiles for the two models ( see the following section ) , we adjust our simulation domains and resolutions accordingly . in the @xmath35 model ,
our simulation box spans @xmath38 to @xmath39 in radius , the full @xmath40 in azimuth , and @xmath41 to @xmath42 in the polar angle measured from the midplane ( symmetry is enforced across the midplane ) .
the simulation grid contains @xmath43 cells , with the radial cells spaced logarithmically , while the spacing is uniform for the other two dimensions . for the @xmath7 model
, our simulation box spans @xmath44 to @xmath45 in radius , the full @xmath40 in azimuth , and @xmath41 to @xmath46 in the polar direction .
the resolution is higher in this model with @xmath47 cells , because of the choice of a lower sound speed .
the initial disk profile assumes a @xmath48 surface density profile .
additionally , we impose an initial axisymmetric gap that mimics the gap we expect the companion to open .
the inclusion of this initial gap allows for a significantly faster time - convergence , and avoids some numerical artifacts that would be caused by the sudden introduction of a companion into an unperturbed disk .
the initial density profile reads : @xmath49\right ) \left(\frac{1}{2 } - \frac{1}{\pi}\tan^{-1}\left[\frac{r_{\rm gap}-|r-{r_{\rm p}}|}{\delta r_{\rm gap}}\right]\right)\,,\ ] ] where @xmath50 , has no impact on our results .
] @xmath51 describes the disk sound speed profile , @xmath52 is the distance from @xmath29 to the gap edge , and @xmath53 describes the sharpness of the gap edge .
we choose @xmath52 to be approximately two times the companion s hill radius @xcite , which makes @xmath54 for the @xmath35 model , and @xmath55 for @xmath7 .
we choose @xmath56 in both models , which is sufficiently smooth to not trigger any hydrodynamical instabilities .
the sound speed profile reads : @xmath57 where @xmath58 is the sound speed at the companion s radial position . for the @xmath35 model ,
we choose @xmath59 , where @xmath60 is the keplerian speed at @xmath61 , and @xmath62 .
this constitutes a constant disk aspect ratio of @xmath63 . for the @xmath7 model ,
we choose @xmath64 and @xmath65 , which makes a flaring disk with @xmath66 .
the somewhat larger scale height for the @xmath1 model is motivated by the recent fits to hd 100453 @xcite , and also provides a contrast to the @xmath0 model .
the initial velocity field assumes hydrostatic equilibrium taking into account gas pressure , but neglecting viscosity . as a result ,
the initial disk has zero radial and polar velocities , and an orbital frequency of : @xmath67 both the inner and outer radial boundaries are fixed at their initial values . for the polar boundaries , we enforce symmetry at the midplane since we only simulate the upper half of the disk , and use a reflecting condition at the top to prevent mass from entering the simulation domain .
additionally , a wave killing zone is imposed next to the top boundary to remove reflections .
the prescription for this wave killing zone is : @xmath68 where @xmath69 represents the fluid variable @xmath14 , @xmath16 , or @xmath15 , @xmath70 is the polar angle at the top boundary , and @xmath71 is the width of the killing zone , which is @xmath72 for the @xmath35 model and @xmath73 for @xmath7 .
technically , this killing zone results in a continuous loss of disk mass , but since the top boundary is located about 5 scale heights above the midplane , the loss is negligible over the timescale of our simulations .
the density distributions obtained in our hydro simulations are post - processed via 3d mcrt calculations using the @xcite code to produce synthetic nir images .
this process largely follows the procedures described in @xcite and @xcite .
the central source is assumed to be a protostar with a temperature of 5200 k and a radius of @xmath74 ( mcrt simulations have only one illumination source , i.e. , the central star ) .
the temperature in each grid cell is calculated using the radiative equilibrium algorithm described in @xcite .
the systems are assumed to be at 140 pc from the earth , the distance to the taurus star forming region .
the companion is assumed to be at @xmath75 au in the @xmath0 case and 200 au in the @xmath1 case , chosen so that the spiral arms in both cases are most prominent around 50 to 100 au , as is suitable for nir imaging observations with currently achievable inner working angles and angular resolutions for systems at 140 pc .
these separations are also in the same range as the distances to the ( predicted ) companions in sao 206462 , mwc 758 , and hd 100453 .
the model grid is a 3d spherical grid in @xmath11 . in the @xmath76 and @xmath77 directions the grid is identical to the one in the corresponding hydro simulation . in the @xmath78 direction
, it extends from the dust sublimation radius @xmath79 , where the dust temperature reaches the sublimation temperature of 1600 k , to @xmath80 . from @xmath79
to @xmath81 the grid has 50 cells in @xmath78 with logarithmic spacing , while from @xmath81 to @xmath80 the grid is identical to the corresponding hydro simulation .
since hydro simulations produce disk density structures in gas , while nir images are determined by the distribution of dust in disks , we need to convert gas density structures into dust density structures .
the dust grains are assumed to be interstellar medium ( ism ) grains @xcite made of silicate , graphite , and amorphous carbon .
their size distribution obeys a smooth power law up to a grain radius of 0.25 @xmath82 followed by an exponential cut off at larger sizes .
the optical properties of the dust grains can be found in ( * ? ? ?
* figure 2 ) . from @xmath81 to @xmath80 , the dust volume density , @xmath83 , is assumed to be linearly proportional to the gas volume density , @xmath84 .
this is appropriate as these sub-@xmath82-sized particles are dynamically well coupled to the gas . between @xmath79 and @xmath81 , @xmath83 at @xmath85
is assumed to be the same as @xmath83 at @xmath86 .
we note that this inner disk is introduced in the mcrt simulations simply to fill in the central cavity of the hydro simulations . removing this inner disk leads to a @xmath87 increase in the surface brightness of the resultant image , but does not alter our reported outer disk morphologies .
the total ism dust disk mass within 100 au is normalized to @xmath88 in both models .
this corresponds to , for example , a total gas disk mass of 0.01 @xmath89 , a 100:1 gas - to - dust mass ratio , a @xmath34 dust mass fraction in the small ism grains , and a remaining 90% mass fraction in the large grains that are presumed to have settled to the disk midplane and do not affect nir scattering .
tests show that our results are insensitive to the total dust mass within one order of magnitude around this fiducial value . our mcrt simulations output all 4 stokes components , @xmath90 , from which total intensity ( ti=@xmath91 ) and polarized intensity ( pi=@xmath92 ) images ] , or [ 10@xmath93 ergs s@xmath94 @xmath95 hz@xmath94 arcsec@xmath96 .
] are produced at @xmath97 , @xmath98 , @xmath99 , and @xmath100-band using 1 billion photon packets over a range of viewing angles .
polarization fraction ( pf ) is defined as pf@xmath101pi / ti .
in addition , the following polar - coordinate stokes parameters are defined @xcite : @xmath102 @xmath103 where @xmath76 is the position angle of a point ( north to east ) in a disk image . by construction
, @xmath104 contains the polarization components in the tangential ( azimuthal ) and radial directions , while @xmath105 contains the components at @xmath106 with respect to that .
if the disk scattering surface is perpendicular to the line of sight ( this is approximately true for geometrically thin disks viewed face - on ) and observed photons are mostly produced by single - scattering events , only the tangential polarization component exists , and @xmath104 is equivalent to pi ( except unlike pi , @xmath104 can be both positive and negative ) .
in such cases , @xmath105 is expected to contain little signal and is often used to estimate the noise level @xcite . for inclined disks ( for which the scattering angle is often not @xmath107 ) and/or images built up of multiply scattered photons ( as is the case for disks surrounded by an envelope ) , polarization may not be purely tangential , and @xmath105 can contain non - trivial signal as well @xcite .
this issue has been recently investigated by @xcite , and will be discussed in section [ sec : results_arms ] .
the orientation of the disk inclination is indicated by a pair of parameters ( pa,@xmath108 ) , where pa is the position angle of the disk major axis ( north - to - east ; pa@xmath109 when the west side of the disk is the nearside ) and @xmath108 is the inclination ( @xmath41 is face - on and @xmath107 is edge - on ) .
the images emerging from the mcrt pipeline are convolved by a gaussian point spread function ( psf ) with a full width half maximum ( fwhm ) of @xmath110 at @xmath97-band , @xmath111 at @xmath98-band , @xmath112 at @xmath99-band , and @xmath113 at @xmath100-band .
these convolving psfs are chosen to achieve the diffraction limited angular resolution at each wavelength for an 8-meter telescope ( e.g. subaru , vlt , or gemini ) .
in this section , we present the morphology of the spiral arms in both the @xmath0 and @xmath1 models , and the morphology of the gap in the @xmath0 model ( in the @xmath1 model the disk is effectively truncated to a circumprimary disk ) , over the full range of viewing angles .
our hydro simulations settle into quasi - steady states after about 10 orbits , and the morphology of the arms remains practically the same afterwards , as shown in figure [ fig : image_100mj_orbits ] .
the definitions of various geometrical terms used in this paper , including the disk s pa ; its `` top '' ( illuminated ) and `` bottom '' ( obscured ) halves ; and its `` nearside '' and `` farside , '' are illustrated in figure [ fig : image_geometry ] .
we use a `` red - hot '' color scheme for the scattered light images and a greyscale scheme for other plots .
unless stated otherwise , all mcrt images are convolved polarized intensity images at @xmath99-band ( angular resolution @xmath112 ; we also examined the arm morphologies at @xmath97 , @xmath98 , @xmath99 , and @xmath100-bands , and found them indistinguishable ) ; the central @xmath114 in radius is masked out to mimic a coronagraph ; the star is marked by a white @xmath115 ; the nearside of the disk is indicated by a white asterisk on the minor axis ; and the projected location of the companion is indicated by an open green circle ( the absence of the green circle implies that the companion is outside the field of view ) .
we note that the noise level ( i.e. the detection limit ) in current nir pi imaging observations is @xmath1160.1 mjy arcsec@xmath117 or lower ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
is the detection limit for ao188+hiciao onboard subaru .
the detection limits for the newer generation of instruments , such as gemini / gpi and vlt / sphere , are expected to be better . ]
our color scheme is chosen to bring out all features above this noise floor .
we also experimented with varying the total disk dust mass @xmath118 by 1 order of magnitude around our fiducial value , and did not find any noticeable change in our results . since different frequency bands and @xmath118 values probe scattering surfaces of different altitudes , this insensitivity suggests that the positions of the arms are largely independent of altitude .
figure [ fig : basic ] shows the surface density maps and face - on images ( both full resolution and convolved ) for both models .
the companion in both cases excites two spiral shocks inside its orbit , which manifest themselves as arms in scattered light images , as seen in previous works @xcite .
we will label as primary ( `` p '' ) the arm pointing towards the companion ; the other arm is secondary ( `` s '' ) .
the two arms in the stellar mass companion case are in a near @xmath3 rotational symmetry . in the planetary companion case ,
the arms are @xmath116@xmath119 apart , consistent with the relation found by @xcite .
also , because of the smaller scale height in the @xmath0 model , the spiral arms in that model are less open than in the @xmath1 model .
the loci of the arms are marked in figure [ fig : image_locus ] .
the @xmath0 planet opens a gap from 75 to 130 au .
both the inner disk , which harbors the prominent inner arms , and the outer disk ring are visible in scattered light .
the planet also excites two arms in the outer disk ring ; however they are faint because of their large distance from the star , and they are difficult to distinguish because they are tightly wound @xcite . in this paper
we focus on the inner arms as they are most likely to be observable . in the @xmath1 model ,
the powerful tidal force of the companion truncates the disk to about half its orbital radius ( 100 au ) , leaving behind two separate components a circumprimary disk and a circumsecondary disk .
we note that the circumsecondary disk also has a pair of mini - arms excited by the primary star ; however , since the companion is not an illumination source , this mini - disk is not visible in scattered light ( it is shadowed from the primary star by the circumprimary disk ) .
figures [ fig : image_100mj_geometry ] , [ fig : image_100mj_geometry_big ] , and [ fig : image_3mj_geometry ] show synthetic @xmath99-band images for the two models at various viewing angles ( figure [ fig : image_100mj_geometry_big ] is a zoomed - out version of figure [ fig : image_100mj_geometry ] that shows the location of the companion ) . for the @xmath0 model
, we view the disk at 7 inclination angles from @xmath120 ( at @xmath121 disk images are very similar to face - on ; see figure [ fig : basic ] ) to @xmath122 ( top to bottom ) ; at each inclination the disk is oriented at 12 pas covering the entire @xmath40 . in total , @xmath123 images of different viewing angles are shown . for the @xmath1 model , because the two arms are nearly rotation - symmetric , at each inclination we show the disk at 6 pas covering @xmath124 in the azimuthal direction ( the other @xmath124 is redundant ) , and thus , in total , @xmath125 images of different viewing angles are shown
. the distortion of the arms introduced by a non - face - on viewing angle can be dramatic .
in general , the portion of the arms on the nearside of the disk gradually moves towards the major axis as the inclination increases .
this can be seen at all position angles , and is best illustrated when an entire arm is on the nearside ( e.g. , from top to bottom in the pa@xmath109 column in figure [ fig : image_100mj_geometry ] , and from top to bottom in the pa@xmath126 and @xmath127 columns in figure [ fig : image_3mj_geometry ] ) .
this behavior is expected when viewing a bowl - shaped surface at an angle .
the distortions as seen on the farside of the disk are less dramatic , until the disk reaches very high inclination ( @xmath128 ) and the entire top half becomes a bowl viewed edge - on . because of the distortions generated by finite inclinations , what appear as two trailing arms on two sides of the disk in face - on images may appear to be ( 1 ) just one arm ; ( 2 ) two trailing arms on one side of the disk ; or ( 3 ) two arms on one side winding in opposite directions .
these possibilities are summarized in figure [ fig : image_100mj_pianalysis ] , which shows the @xmath1 model at four viewing angles . at inclinations
lower than @xmath129 , the disk s bottom half is too faint to detect ( its emission is less than three times the noise level in current observations ) . as the disk becomes more inclined , the nearside of its bottom half , separated from the top half by the dark lane at the disk midplane , brightens . in the @xmath1 model , because there is no outer disk to block the spiral arms , one or both arms can be seen on the disk s bottom half depending on the position angle .
this is best illustrated by comparing the first and fourth columns of figure [ fig : image_100mj_geometry_big ] . in the @xmath0 model , only the outer disk ring can be seen on the bottom half , as the outer disk blocks the inner disk .
therefore the bottom half remains symmetric about the minor axis , regardless of position angle .
@xmath104 , @xmath105 , and polarization fractions for various viewing angles are shown in figure [ fig : image_100mj_pianalysis ] , along with ti and pi images .
pi and @xmath104 are nearly identical in all cases , and aside from a difference in normalization , both resemble the ti images . while the face - on @xmath105 contains little signal and represents noise in our mcrt simulations , at high inclinations there is some signal in @xmath105 ( particularly in the bright inner disk close to the model coronagraph edge ) , which shows patterns symmetric about the minor axis .
these findings are consistent with those by @xcite for disks viewed at modest - to - high inclination .
the @xmath1 disk is truncated by the companion and thus has no outer disk , while in the @xmath0 case the planet opens a gap that separates the well - defined inner and outer disks .
figure [ fig : image_3mj_geometry_gap ] shows pi , ti , and pf images of the entire 200-au - wide disk for the @xmath0 model at pa@xmath126 for various inclinations ( the pf images are the same as in the first column of figure [ fig : image_3mj_geometry ] ) .
the outer disk ring on the top half is visible in both pi and ti at all inclinations . in ti images ,
scattered light is brightest on the nearside of the ring , and decreases towards the farside because of preferential forward scattering by ism dust . in pi images ,
both the nearside and farside of the top half are fainter than in the ti images because of the inefficiency of polarized scattering at both small and large scattering angles ( note the two breaks on the ring along the minor axis at @xmath130 ) .
the two `` wings '' along the major axis on the top half of the disk are prominent as the scattering angle is close to @xmath107 . at high inclinations
, the two wings may look like two spiral arms winding in opposite directions ( as further discussed in section [ sec : discussions_aksco ] ) .
this can be further demonstrated in the right column : the pf in the two wings ( indicated by the red arrows ) at @xmath131 can reach @xmath1160.5 , while the farside of the top half and the nearside of the bottom half have pf close to 0 .
the gap is traceable in both ti and pi images at @xmath132 ( at larger inclinations only the ti images show the gap ) , but leaves little imprint in the pf maps .
lastly , the dark lane of the disk midplane is prominent at @xmath131 in both ti and pi , but is practically 0 in pf .
figure [ fig : image_3mj_edgeon ] shows the ti and pi images at 3 nearly edge - on viewing angles ( @xmath133 ) in a linear stretch to emphasize the structures in the outer disk ring . in all cases , clumps can be identified on the `` wings '' in both ti and pi .
they are spiral arms `` collapsed '' along the line of sight .
these structures will be discussed in section [ sec : discussions_aksco ] .
in resolved observations of disks , the original images are sometimes deprojected based on an estimated inclination of the system in an attempt to recover the face - on morphology of the disk , and also scaled by @xmath134 to enhance the visibility of features at large distance ( sometimes the two are combined as in , e.g. , @xcite ) .
clearly , deprojection assumes the disk surface is thin and flat ( i.e. , not curved ) , and @xmath134-scaling assumes a face - on viewing geometry .
neither of these conditions is typically satisfied . to assess the degree to which conventional deprojection and scaling procedures introduce unphysical artifacts , we show in figures [ fig : image_3mj_deproj ] and [ fig : image_100mj_deproj ] the results of deprojections and scalings .
@xmath134-scaling significantly enhances the visibility of features at large projected distances , as expected .
in particular , the originally faint bottom half of the disk becomes prominent , as does the ring in the outer disk in the @xmath0 model .
however we note that the compensation at large radius is not entirely physical .
for example , different parts of the outer disk ring in the @xmath0 model are not compensated by the same factor ( which they should be because the ring is a circle centered on the star ) because of their different projected distance @xmath78 from the center . as a result ,
the ring is disproportionally enhanced along the major axis and less so along the minor axis .
likewise , deprojecting an inclined image can sometimes lead to serious misinterpretation of the true morphology of the system . in the @xmath1 model , even at a modest @xmath135 inclination ( the second and third rows in figure [ fig : image_100mj_deproj ] ) , deprojection can not restore the morphology of the two arms at a face - on viewing angle ; at @xmath136 inclination ( the last two rows ) , deprojection severely distorts the arms it can not even recover the total number of arms .
in the @xmath0 model , deprojection also fails to restore the face - on arm morphologies , although the distortion is less severe than for the @xmath1 model because the disk is flatter in the @xmath0 model ( @xmath137 ) than in the @xmath1 model ( @xmath138 ) , and because the height variations of the arms are smaller for the @xmath0 model .
deprojection also can not restore the outer disk ring and the gap back to concentric circular structures they appear to be off - centered ellipses broken along the minor axis ( the eccentricity of the ellipses is somewhat reduced compared with the original inclined images ) .
lastly , it is obvious that the dark lane along the disk midplane can not be removed by deprojecting the images .
it follows from these tests that the method developed by @xcite to infer the mass of a planet based on the azimuthal separation between the two arms it excites can only be applied to nearly face - on systems ( @xmath139 ) . in this section ,
we connect our models to observed disk systems .
as shown in the bottom rows of figures [ fig : image_100mj_geometry ] , [ fig : image_100mj_geometry_big ] , and [ fig : image_3mj_geometry ] , nearly edge - on disks ( @xmath140 ) with spiral arms may appear to be asymmetric about their minor axis at certain pas , as the axisymmetry of the disk is broken by the non - axisymmetric spiral arms excited by companions . in the case of a stellar mass companion , the spiral
arms are big and wide , and not blocked by an outer disk ring when viewed edge - on ( since the disk is severely truncated ) ; the arms introduce global asymmetries in disks .
this kind of asymmetry has been found in directly imaged nearly edge - on disks , such as the hh 30 disk inclined by @xmath141 @xcite . in the case of planet - induced spiral arms ( which are thinner ) , when viewed at certain pas they may `` collapse '' along the line of sight to produce intensity inhomogeneities in the form of clumps in edge - on disks ( figure [ fig : image_3mj_edgeon ] ) .
these clumps can be seen in ti and especially pi images . in direct imaging observations
, they may mimic and be confused with point source detections . in recent gpi and sphere observations of the edge - on debris disk au mic , several clumps of unknown origin have been identified at tens of au on one side of the disk @xcite .
the morphology of these clumps are reminiscent of the ones in figure [ fig : image_3mj_edgeon ] .
though our results can not be directly applied to the au mic observations as gas - poor debris disks differ in their dynamics from gas - rich protoplanetary disks , figure [ fig : image_3mj_edgeon ] hints at a potential solution to the mystery , and warrants more specific investigation in the future .
as shown in figure [ fig : image_3mj_geometry_gap ] a planet - induced gap may be detectable in both ti and pi images at inclinations between @xmath116@xmath120 and @xmath116@xmath136 . in figure [ fig : image_abaur ]
we compare our @xmath0 model at @xmath142 with the subaru / hiciao polarized observation of ab aur @xcite .
the full resolution model image is rescaled to match the size of the object ( the @xmath0 planet is now at 67 au ) , and convolved by an appropriate psf to achieve the same angular resolution as the subaru observation .
the model disk has pa@xmath143 , @xmath142 ; rotates around the star in the counterclockwise direction ; and has a nearside to the southeast ; all these parameters are consistent with observations @xcite . in this paper
is equivalent to pa=@xmath144 in @xcite ; there is a difference of @xmath124 between our definitions . ]
the qualitative morphologies of the inner disk , the gap , and the outer disk ring resemble features seen in ab aur ; in particular , the gap at @xmath145 au appears to be deeper on the farside and shallower on the nearside because of the variation of dust scattering efficiency with scattering angle .
we note that although the comparison between our model and the observations is only qualitative , the model is `` unique '' in the sense that the viewing geometry has been fixed by the observations .
there are even hints of fine spiral arms in the inner disk in the subaru image , at the same locations as in the model images .
we note that in the model image the morphologies of the outer ring and the gap are more - or - less independent of the position angle of the planet as they are intrinsically axisymmetric , while the details of the inner disk structures are obviously sensitive to planet location . as shown in figures [ fig : image_100mj_geometry][fig : image_100mj_pianalysis ] , depending on the orientation of the disk , the trailing - double - arm pattern for a face - on viewing geometry may appear at other viewing angles to be one trailing arm , two trailing arms to one side , or two arms winding in opposite directions .
such features have been reported in real disks , as in the single trailing arm at @xmath116130 au in the inner disk of hd 141569 a ( viewed at @xmath116@xmath146 inclination ; * ? ? ?
* ) .5-myr - old system @xcite .
] companion - induced spiral arms viewed at a large inclination may explain such an observation .
intriguingly , `` spiral arms '' can be mimicked by an outer disk ring separated from the inner disk by a planetary gap when viewed at non - zero inclinations , as illustrated in figure [ fig : image_3mj_geometry_gap ] . in this case ,
the outer disk manifests itself as two `` pseudo - arms '' along the major axis , approximately symmetric about the minor axis , winding in opposite directions .
the two arms found in recent sphere observations of ak sco @xcite may provide a real - life example . in figure
[ fig : image_aksco ] , we compare a model image ( rescaled to match the size of ak sco ) with the sphere observations .
the model image adopts the geometry of ak sco : the position angle of the disk in this paper is equivalent to pa=@xmath136 in @xcite ; there is a difference of @xmath124 between our definitions . ]
is @xmath147 and the disk is inclined by @xmath148 @xcite .
the two pseudo - arms in the model image qualitatively resemble the arms in ak sco . as was the case above for ab aur ,
the model has no free parameter regarding viewing geometry .
moreover , the only remaining free parameter , the location of the planet , does not much affect the arm morphologies .
experiments show that the opening angle of the pseudo - arms is determined by the inclination ( figure [ fig : image_3mj_geometry_gap ] ) and the disk scale height ( both the absolute value and the extent of flaring ) . for a given distance to the object , the length of the arms is set by the size of the outer disk ring ; for a fixed ring size , the width of the arms is set by the gap size determined by @xmath149 , @xmath150 , and disk viscosity @xcite , as a wider gap narrows the outer disk ring .
the @xmath0 planet in our model of ak sco is located at 25 au from the star , while its azimuthal location can not be determined as the outer disk ring is approximately axisymmetric .
our model makes a unique prediction for such pseudo - arms .
as shown in figure [ fig : image_3mj_geometry_gap ] , they have high polarization fractions because they scatter light at @xmath107 .
given that the sphere observations in @xcite are in ti , future pi observations of ak sco can test our interpretation . also , deeper exposures can attempt to reveal the faint nearside of the bottom half of the disk , parallel to the two arms and separated by the dark midplane .
the two major arms identified in the inner disk of hd 100546 at @xmath151 au in recent gpi @xcite and sphere imaging ( * ? ? ?
* see the wing structure in their fig .
3 ) might also be pseudo - arms .
the geometry of the disk supports this interpretation : the disk is suitably inclined by @xmath116@xmath144 , and both arms are on the farside of the major axis .
the above tests outlined for ak sco also apply to hd 100546 .
in this paper , we carried out 3d hydrodynamics simulations of companion - induced spiral arms and gaps in protoplanetary disks .
we examined the observational signatures of these features by synthesizing direct imaging observations at near - infrared wavelengths using 3d monte carlo radiative transfer simulations .
we studied two models , one with a @xmath0 planet at 100 au and another with a @xmath1 companion at 200 au ( assuming the central star is 1 @xmath89 ) , surveying the entire parameter space in viewing inclination and position angle ( figures [ fig : image_100mj_geometry ] and [ fig : image_3mj_geometry ] ) . in the @xmath0 model
the planet opens a gap and excites two prominent spiral arms in the inner disk ; the inner disk arms , the outer disk ring , and the gap are readily detected when viewed face - on ( figure [ fig : basic ] ) .
in the @xmath1 model , the companion truncates the circumprimary disk to about half its orbital radius , and produces a prominent pair of arms with azimuthal wavenumber @xmath3 in face - on images ( figure [ fig : basic ] ) .
our results for face - on systems confirm previous findings in @xcite and @xcite based on _
athena_++ ( stone et al .
, in preparation ) hydro models , and can be compared with ( and provide predictions for ) current and future nir imaging observations using gemini / gpi , vlt / naco / sphere , and subaru / hiciao / scexao .
below , we summarize the arm and gap morphologies presented in this paper : 1 .
companion - induced spiral structures quickly settle into a quasi - steady state after @xmath11610 orbits ( figure [ fig : image_100mj_orbits ] ) . for face - on views ,
the morphologies are insensitive to observing wavelength and to the total ism dust mass assumed , suggesting the positions of the arms are largely independent of altitude .
total intensity and polarized intensity images of face - on disks exhibit the same morphologies ( figures [ fig : image_100mj_pianalysis ] and [ fig : image_3mj_geometry_gap ] ) .
2 . depending on the viewing inclination and position angle , the trailing - double - arm pattern for face - on views may appear as one trailing arm ; two trailing arms to one side of the star ; or two arms on the same side but winding in opposite directions ( e.g. , figure [ fig : image_100mj_pianalysis ] ) .
3 . at inclinations
@xmath152 , the bottom ( obscured ) half of the disk emerges . in the @xmath0 model
the nearside of the bottom half of the outer disk ring can be seen , and is always symmetric about the minor axis independent of the viewing angle ( figure [ fig : image_3mj_geometry_gap ] ) . in the @xmath1 model ,
the two spirals on the bottom half are directly visible as they are not blocked by an outer disk ; the bottom half in this case is asymmetric to varying degrees depending on the position angle ( figures [ fig : image_100mj_geometry ] and [ fig : image_100mj_pianalysis ] ) .
we confirmed the findings of @xcite that @xmath104 ( equation [ eq : qr ] ) and pi are generally similar . by comparison
, @xmath105 ( equation [ eq : ur ] ) represents observational noise in relatively face - on systems , but does contain signal for systems with @xmath153 ( figure [ fig : image_100mj_pianalysis ] ) .
we also highlight the shortcomings of two standard image reconstruction procedures : 1 . the conventional practice of deprojecting inclined disk images based on the known inclination can not restore the face - on morphology ( figures [ fig : image_3mj_deproj ] and [ fig : image_100mj_deproj ] ) , particularly for high inclinations .
the procedure ( 1 ) can not recover the total number of arms ; ( 2 ) can not recover the arm locations ; ( 3 ) can not restore gaps and outer disk rings back to concentric circular structures ; and ( 4 ) can not remove the dark lane of the disk midplane .
the dynamical planet - mass measurement technique based on arm morphology @xcite applies only to disks with inclinations @xmath154 .
2 . the rescaling of images by @xmath134 introduces artificial azimuthal variations , for the obvious reason that in such systems
, the physical distance to the star can differ from the projected distance . finally , we make the following connections between our models and observations : 1 .
asymmetric structures in nearly edge - on disks may be caused by spiral arms .
big and wide arms excited by massive ( e.g. , stellar mass ) companions can generate global scale asymmetries , similar to those seen in hst observations of the hh 30 disk which is inclined by @xmath141 @xcite .
narrower arms excited by planet - mass companions viewed from certain angles can collapse along the line of sight to appear as clumps in nearly edge - on disks ( figure [ fig : image_3mj_edgeon ] ) .
2 . the partial gap at @xmath145 au in ab aur observed by subaru / hiciao @xcite may be a gap opened by a @xmath0 planet at 67 au . in this interpretation ,
the disk viewed at @xmath120 inclination and @xmath155 position angle rotates counterclockwise on the sky , and has a nearside to the southeast , consistent with observations ( figure [ fig : image_abaur ] ) .
the single arm observed at @xmath116130 au in the hd 141569 a disk ( * ? ? ?
* hst ) may be a trailing - double - arm structure viewed at @xmath116@xmath144 inclination .
4 . at @xmath152 , the disk ring outside the gap opened by a giant planet manifests itself as two pseudo - arms placed symmetrically about the minor axis winding in opposite directions ( figure [ fig : image_3mj_geometry_gap ] ) .
our model image of a disk perturbed by a @xmath0 planet at 25 au , when viewed at @xmath148 inclination , resembles the ak sco system as observed by sphere ( @xcite ; figure [ fig : image_aksco ] ) .
a similar interpretation may explain the pair of minor - axis - symmetric arms at @xmath151 au in the inner disk of hd 100546 , as viewed at @xmath116@xmath144 inclination ( @xcite , gpi ; @xcite , sphere ) .
our models make a prediction that can be tested by observations in the near future : these arms should have a much higher polarization fraction than the rest of the disk ( see the two pf `` wings '' indicated by the arrows in figure [ fig : image_3mj_geometry_gap ] ) . a deeper exposure may also reveal the dark lane of the disk midplane parallel to these pseudo - arms .
we thank markus janson and jun hashimoto for kindly sharing with us the sphere image of ak sco and the hiciao image of ab aur , respectively .
an anonymous referee provided a helpful and encouraging report .
this project is supported by nasa through hubble fellowship grant hst - hf-51320.01-a ( r.d . ) 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 . e.c . acknowledges support from nasa and the nsf .
j.f . is grateful for the support from the center for integrative planetary science at the university of california , berkeley .
numerical calculations were performed on the savio cluster provided by the berkeley research computing program , supported by the uc berkeley vice chancellor for research and the berkeley center for integrative planetary science . | direct imaging observations of protoplanetary disks at near - infrared ( nir ) wavelengths have revealed structures of potentially planetary origin .
investigations of observational signatures from planet - induced features have so far focused on disks viewed face - on . combining 3d hydrodynamics and radiative transfer simulations , we study how the appearance of the spiral arms and the gap produced in a disk by a companion varies with inclination and
position angle in nir scattered light .
we compare the cases of a @xmath0 and a @xmath1 companion , and make predictions suitable for testing with gemini / gpi , vlt / naco / sphere , and subaru / hiciao / scexao .
we find that the two trailing arms produced by an external perturber can have a variety of morphologies in inclined systems they may appear as one trailing arm ; two trailing arms on the same side of the disk ; or two arms winding in opposite directions .
the disk ring outside a planetary gap may also mimic spiral arms when viewed at high inclinations .
we suggest potential explanations for the features observed in hh 30 , hd 141569 a , ak sco , hd 100546 , and ab aur .
we emphasize that inclined views of companion - induced features can not be converted into face - on views using simple and commonly practiced image deprojections . |
quenched disorder is obiquitous in condensed - matter systems , sometimes determining entirely new physical properties and phenomena . for the theoretical explanation of many experimental findings ,
models of classical and quantum spin systems are employed and - typically - disorder is introduced through a spatial variation of the exchange couplings , of the direction of the axial anisotropies , and of the sign of the couplings .
theoretical and experimental studies of model and real physical systems with these kinds of disorder have lead to much new physics .
much discussed have been the glassy phases in magnetic @xcite , and superconducting materials @xcite . in superconductors , specially with the advent of high-@xmath0 ceramic superconductors , the role of disorder has become central in the discussion of the physical properties of real materials @xcite .
many materials are naturally microstructured and granularity characterises the mesoscopic structures of most systems , leading to a phase diagram often displaying a superconductor - insulator transition at zero temperature due to the charging energy of the grains .
as first pointed out by abeles @xcite , when the grain charging energy arising from the charge @xmath1 and capacitance @xmath2 of the grain , @xmath3 ( @xmath4 is the grain diameter and @xmath5 the electronic charge ) is larger than the josephson - coupling energy @xmath6 between nearest neighbors grains , phase coherence is destroyed due to zero - point quantum fluctuations of the local phase of the superconducting order parameter @xcite .
granularity can also be realized in a controlled manner in artificially fabricated josephson - junction arrays of coupled superconducting grains @xcite , with a space dimensionality @xmath4 less than @xmath7 .
such granular systems can be theoretically modelled by pseudo - spin systems , the spins representing with their states the few relevant quantum charge states of the grains at relatively low temperatures . in such models
the pseudo - spin size is the same for all grains corresponding to the assumption that the grain - size distribution is very narrow , or alternatively , of negligible charging - energy disorder .
a well known model of this type is the pseudo - spin - one model introduced by de gennes and studied in different works @xcite where only charge states @xmath8 are allowed , corresponding to @xmath9 .
however , realistic systems may contain different kinds of disorder , such as a spatial distribution of josephson couplings between grains or / and a distribution of grain sizes , which leads to disorder in the grain electrical capacitances and charging energies .
studies of the effects of disorder in the electrical capacitance or charging energy of the grains have appeared recently @xcite . within a mean - field approximation @xcite , charging energy disorder widens the extent of the superconducting phase at the expense of the insulating one .
this behavior is consistent with results for the superconductor - insulator transition from quantum monte carlo simulations in two - dimensional models of josephson - junction arrays @xcite with disorder in the diagonal capacitance matrix .
earlier calculations for the related boson hubbard model with disorder in the onsite coulomb repulsion @xcite are also consistent with a decrease in the phase coherence threshold and moreover suggest a different universality class from the non - disordered case @xcite . in this work , we consider the effects of charging energy disorder in granular superconducting materials within the pseudo - spin approach . by considering a spatial distribution of the grain sizes , which leads to local charging energy disorder , a quantum pseudo - spin model with random on - site spin sizes can be constructed .
a mean - field theory is developed to obtain the phase diagram as a function of temperature , average charging energy and disorder .
spin - size disorder models have seldom been considered in the literature . in the context of classical spin models ,
spin - size disorder can be readily turned into exchange - coupling disorder ; the physics of quantum systems with spin - size disorder , however , appears to have not been investigated in such depths .
it has been considered mostly within the one - dimensional systems : dilution of a quantum spin-@xmath10 2-ladder was studied by sigrist and furusaki @xcite whilst the general problem of a quantum spin chain with random @xmath11 as well as random @xmath12 was considered by westerberg _ et .
@xcite within a real - space renormalization - group method showing that these systems belong to a different universality class of disordered spin systems .
the hamiltonian for a set of superconducting grains coupled by the josephson energy can be written as the sum of the coulomb charging energy and the josephson - coupling energy @xcite @xmath13 where @xmath14 is the net charge on the superconducting grain at site @xmath15 and @xmath16 is the phase of the local superconducting order parameter @xmath17 .
@xmath18 is the electrical capacitance matrix and @xmath19 is the josephson coupling between nearest - neighbor grains .
the charge @xmath14 in each grain can be expressed in terms of the excess number of cooper pairs @xmath20 as @xmath21 . considering a diagonal capacitance matrix @xmath22 and uniform josephson - coupling energies @xmath23 ,
leads to the self - charging model @xmath24 where @xmath25 are the charging energies of the grains .
the number operator @xmath20 is conjugate to the phase @xmath16 , satisfying the commutation relation @xmath26=-i\delta_{ij } , \label{jjacomm}\ ] ] and can be written as @xmath27 , having integer eigenvalues @xmath28 .
the model of eq .
( [ jja ] ) can also be regarded as a boson hubbard model @xcite where the charging energy represents the onsite coulomb repulsion of bosons and the josephson coupling represents the hopping term . when the charging energies are uniform @xmath29 , a pseudo - spin one model for the hamiltonian of eq .
( [ jja ] ) can be constructed by truncating the basis vectors of the number operators @xmath30 to @xmath31 and @xmath32 , corresponding to the lowest charging energy states . identifying the charging states , @xmath33 , as the eigenstates of @xmath34 for spin @xmath9 , writting the second term of eq .
( [ jja ] ) in terms of @xmath35 and making the correspondence @xmath36 one obtains the de gennes mapping to a @xmath9 pseudo - spin model with single - ion spin - anisotropy @xcite @xmath37 with @xmath38 .
the factor @xmath39 in front of @xmath40 in eq .
( [ pseudomap ] ) takes into account the length of the pseudo spin @xmath41 . in terms of this spin model ,
the superconductor to insulator transition in the original granular superconductor corresponds to a zero temperature transition where the ferromagnetic order of the spins in the @xmath42 plane is destroyed by quantum fluctuations when the @xmath34 component is confined to zero for an increasing ratio of @xmath43 .
although this mapping is not exact @xcite , the @xmath44 critical behavior observed in the original phase model of eq .
( [ jja ] ) without disorder and the spin model is the same @xcite as found in numerical calculations in one dimension . for higher dimensions
, results from the same mean - field approximation applied to both models also agree @xcite .
to consider the main effects of disorder in the charging energies @xmath45 , we generalize the above approximate mapping to a pseudo - spin model with randomness in the spin - size values @xmath46 .
since for a given charging energy ( or temperature fluctuation ) , low values of @xmath45 correspond to charging states with higher @xmath20 , it seems reasonable to use a truncation scheme which identifies the charging states @xmath47 and @xmath48 with the @xmath34-eigenstates of the spins of different sizes @xmath46 and @xmath49 , such that the corresponding maximum charging energies @xmath50 and @xmath51 of the grains are both comparable to the same truncation energy ( here @xmath52 is the length of the spin @xmath53 ) .
this ( approximate ) mapping leads to the effective spin hamiltonian @xmath54 with randomness in the spin - size values @xmath46 . here , @xmath55 , @xmath56 where @xmath57 is the average value of @xmath45 and @xmath58 the average value of @xmath59
the spin sizes @xmath46 are restricted to take only integer values @xmath60 in this mapping . in eq .
( [ haniso ] ) , randomness in the spin - size also leads to randomness in the local single - ion anisotropy parameter , but these random variables are only correlated at the same site . a typical spin - size distribution we shall be considering is @xmath61 where three spin - size values @xmath62 and @xmath63 are mixed with average value @xmath64 and a concentration @xmath65 , which is a measure of the disorder .
it is also convenient to rewrite the spin model of eq .
( [ haniso ] ) in terms of normalized spins @xmath66 @xmath67 in this latter form , a bimodal distribution of spin values can be considered : @xmath68 which in the special case @xmath69 corresponds to the dilution of a spin-@xmath70 system or to percolating granular superconductors @xcite in the presence of charging effects .
for high enough space dimensions @xmath4 , reasonable results can be obtained from simple mean - field approximations ( mfa ) .
the mfa replaces all variables around a given small cluster with their average value , or order parameter , and then use the resulting approximate hamiltonian to evaluate the order parameter itself . since the spin sizes are random , one must carry out the averaging procedure with respect to @xmath71 as well and we denote this by @xmath72_{ave}$ ] .
the simplest cluster is a single spin on a site . with @xmath73_{ave}$ ] the mean field , this yields a self - consistent equation for the order parameter @xmath74 : @xmath75_{ave}=\left [ \frac{1}{z_{mfa } } tr_{{\bf s}_i } { \bf s}_i e^{-\beta{\bar h}_{mfa } } \right]_{ave } , \label{mfa1}\ ] ] where @xmath76 is the mean - field hamiltonian and @xmath77 . for the anisotropic hamiltonian of eq .
( [ haniso ] ) , we assume that the ferromagnetic ordering takes place in the @xmath78 plane of the anisotropy , say @xmath79_{ave}=m$ ] . then , in the mfa we replace @xcite @xmath80 by virtue of the fact that @xmath81 . near the critical curve ,
the mfa hamiltonian then becomes : @xmath82 where @xmath83 is a dimensionless expansion parameter .
since we are interested in the phase transition line only , we can assume @xmath84 small and use a first - order perturbation expansion to solve eq .
( [ mfa1 ] ) for arbitrary spin @xmath11 . in a now single - site problem ,
we reabsorb a factor @xmath85 into @xmath86 and @xmath12 .
the unperturbed states @xmath87 are eigenstates of @xmath34 and thus also of the unperturbed hamiltonian @xmath88 , with @xmath89 .
we have @xmath90 where to first - order @xmath91 and @xmath92 here , the prime on @xmath93 means @xmath94 .
these can then be inserted in eq .
( [ def ] ) to get , keeping only first order terms in @xmath95 : @xmath96 where @xmath97 . using @xmath98
we get @xmath99 which , using eq .
( [ mfa1 ] ) , gives implicitly the critical temperature in the mfa @xmath100_{ave } .
\label{amfa7}\end{aligned}\ ] ] one can check that for @xmath9 and in absence of disorder the mfa transition line from eq .
( [ amfa7 ] ) agrees with the result obtained for the de gennes model of eq .
( [ gennes ] ) in the same approximation @xcite
1 shows the phase boundaries between the superconducting and normal phases obtained in absence of disorder for different values of uniform spin sizes @xmath9 , @xmath101 and @xmath102 . the critical value @xmath103 for the superconductor - insulator transition at @xmath44 remains unchanged and only at much higher temperatures there is a deviation in the transition lines when considering larger spin approximations .
this is in agreement with mf calculations using truncated basis vectors in the phase model of eq .
( [ jja ] ) for increasing number of states @xcite .
our present calculations generalize the mfa to arbitrary spin sizes and allow to obtain the transition line in the case of a random spin size distribution by simply averaging with respect to @xmath104 in eq .
( [ amfa7 ] ) .
the consequences of spin - size disorder depend very much now on the chosen distribution , but it is expected that in general within a simple mf approximation , the phase transition of the homogeneous system is preserved with at most a modification of the phase boundary line .
2a shows the transition lines for the spin - size distribution of eq .
( [ distr ] ) with @xmath105 and different values of the disorder parameter @xmath65 . for increasing disorder the critical value for phase coherence
@xmath106 decreases , increasing the extent of the superconducting phase at low enough temperatures .
this behavior is in qualitative agreement with mf calculations @xcite in the phase model of eq .
( [ jja ] ) .
on the other hand , a different behavior is expected when there is dilution of superconducting grains .
2b shows the transition lines for the spin - size distribution of eq .
( [ distrd ] ) with @xmath69 , @xmath107 , corresponding to a dilution of grains for different concentrations @xmath65 . in this case
, disorder decreases the extent of the superconducting phase . for large values of @xmath65 ,
corresponding to the system below the percolation threshold , the superconducing phase should disappear .
however , an improved mf approximation is required to describe this behavior .
one possible approach is the mf renormalizaton - group method @xcite used previously to study the de gennes model of eq .
( [ gennes ] ) in absence of disorder @xcite .
we have introduced a quantum pseudo - spin model with random spin sizes to model the effects of charging - energy disorder in granular superconducting materials .
randomness in the spin size is argued to arise from the inhomogeneous grain - size distribution . for a particular bimodal spin - size distribution ,
the model describes percolating granular superconductors .
a mean - field theory has been developed to obtain the phase diagram as a function of temperature , average charging energy and disorder .
the results are qualitatively consistent with previous mean - field calculations in the phase - number representation .
the pseudo - spin model should provide a useful framework to study the critical behavior and universality classes in presence of strong charging - energy disorder .
e.g. was suported by by fundao de amparo pesquisa do estado de so paulo ( fapesp ) under grant no .
07/08492 - 9 . g.j . acknowledges fapesp ( grant no . 07/ 04006 - 2 ) for supporting a visit to the instituto nacional de pesquisas espaciais , where part of this work was done . | a quantum pseudo - spin model with random spin sizes is introduced to study the effects of charging - energy disorder on the superconducting transition in granular superconducting materials .
charging - energy effects result from the small electrical capacitance of the grains when the coulomb charging energy is comparable to the josephson coupling energy . in the pseudo - spin model ,
randomness in the spin size is argued to arise from the inhomogeneous grain - size distribution . for a particular bimodal spin - size distribution ,
the model describes percolating granular superconductors .
a mean - field theory is developed to obtain the phase diagram as a function of temperature , average charging energy and disorder . and
granular superconductors , josephson - junction arrays , superconductor - insulator transition 74.81.-g , 74.25.dw , 74.40.+k , 64.70.tg |
an important question on complex quantum systems , which remains open to a large extent , addresses the conditions to have robust efficient transport of excitations across a disordered finite network @xcite . whereas it is well known how to define a hamiltonian system where perfect transport is obtained @xcite , the fact that such system requires the precise specification of a large number of parameters makes it difficult to achieve in practice .
this is the sense of robustness above : statistical changes in the parameters should lead to small fluctuations that preserve good efficiencies , instead of large transmission fluctuations .
clearly , the number of control parameters should be as small as possible .
inspired by ref .
@xcite , we study here the distribution of the transport efficiencies of an initial localized excitation in a disordered network of @xmath1 sites , which is modeled by random hamiltonian that includes many - body interactions , considering both fermions and bosons ; to the best of our knowledge , quantum efficiencies of this type of disordered networks have not been considered . in general , this question is of interest in a variety of fields , including understanding photosynthetic light - harvesting complexes @xcite , such as the fenna - matthews - olson ( fmo ) complex @xcite , or in quantum communication protocols across quantum spin chains @xcite . in either case
, a model with few - body interactions seems realistic . the transport efficiency from an input state @xmath2 to an output state @xmath3 ,
is quantified as the maximum transition probability achieved among theses states within a time interval @xmath4 $ ] .
the transport efficiency is defined as @xcite @xmath5 } |\langle { \rm out}| u(t)|{\rm in}\rangle|^2 . \label{eq1}\ ] ] here , @xmath6 is the unitary quantum evolution associated with the hamiltonian of the system and @xmath7 is a reasonable time scale ( @xmath8 ) .
the system is said to have perfect state transfer ( pst ) when @xmath9 @xcite .
the nodes of the network are the basis states of the hilbert space where the initial excitation is localized .
we define the hilbert space by distributing @xmath10 spinless particles ( bosons or fermions ) in @xmath1 single - particle states .
the @xmath10-body hamiltonian that we consider consists of a random @xmath0-body interaction among the @xmath10-particles states , with @xmath11 ; the non - zero hamiltonian matrix elements are related to the links of the network .
this matrix model is known as the ( bosonic or fermionic ) @xmath0-body embedded gaussian ensemble of random matrices @xcite . the motivation for choosing this matrix model comes from the fact that this ensemble displays correlations among the matrix elements @xcite , in the sense that the number of independent random variables is usually smaller than the number of independent links of the network .
in addition , the bosonic ensemble for the specific case where the bosons are distributed in two single - particle levels displays a systematic appearance of doublets in the spectrum @xcite .
thus , one of the design principles required in ref .
@xcite , the existence of a dominant doublet , is automatic fulfilled .
the extension to fermions arises naturally , and is also motivated by the transport processes within the fmo system @xcite . as in previous studies @xcite
, centrosymmetry @xcite is considered and we find that it increases dramatically the best efficiencies among certain pairs of distinct sites , in comparison to a non - centrosymmetric interaction ; we also study in this paper the case where time - reversal symmetry is broken , and the corresponding case including centrohermiticity @xcite . here , as usual , the benchmark for good efficiencies is @xmath12 .
in particular , for @xmath13 and few bosons distributed in two single - particle states , we find that pst is obtained for a large fraction of the best efficiencies of the ensemble when centrosymmetry or centrohermiticity are included ; for larger values of @xmath0 the best efficiencies are reduced . in the case of fermions and including centrosymmetry or centrohermiticity , we show that below half - filling ( @xmath14 , where @xmath10 is the number of fermions and @xmath1 the number of single - particle states ) the best efficiencies are obtained for @xmath13 ; above half - filling , our numerics indicate that the best efficiencies correspond to @xmath15 for @xmath16 . in both cases ,
the benchmark is achieved with non - zero probability .
the paper is organized as follows : in sect .
[ sec2 ] we define the matrix model considered and centrosymmetry , which is introduced at the one - particle level and then extended to the @xmath0- and @xmath10-particle spaces . in sect .
[ sec3 ] we describe the results for the distribution of the best efficiencies of each realization of the ensemble for bosons occupying two - single particle states , with or without the addition of centrosymmetry or centrohermiticity ; sect .
[ sec4 ] analyzes the fermionic case along the same lines . finally , in sect .
[ sec5 ] we present the conclusions of this work .
we introduce the @xmath0-body embedded gaussian ensembles of random matrices for bosons and fermions , for the cases of orthogonal ( @xmath17 ) and unitary ( @xmath18 ) symmetries , following ref .
we consider a set of @xmath1 degenerate single - particle states @xmath19 , with @xmath20 . the associated creation and
annihilation operators for fermions are @xmath21 and @xmath22 , and @xmath23 and @xmath24 for bosons , with @xmath25 .
these operators obey the usual ( anti)commutation relations that characterize the corresponding particles .
we define the operators that create a normalized state with @xmath26 fermions from the vacuum as @xmath27 , with the convention that the indexes are ordered increasingly @xmath28 ( @xmath29 simplifies the notation for these indexes ) ; the corresponding annihilation operators are @xmath30 .
likewise , the @xmath0-boson states are given by @xmath31 , where again @xmath32 .
here , @xmath33 is a factor that guarantees the normalization to unity of @xmath34 : if the index @xmath35 repeats @xmath36 times , @xmath33 contains a factor @xmath37 .
the random @xmath0-body hamiltonian for fermions reads @xmath38 a similar equation holds for bosons replacing @xmath39 by @xmath40 . in eq .
( [ eqham ] ) the coefficients @xmath41 are random distributed independent gaussian variables with zero mean and constant variance @xmath42 here , @xmath43 is dyson s parameter that accounts for the presence ( @xmath17 ) or absence ( @xmath18 ) of time - reversal symmetry @xcite , the bar denotes ensemble average , and we set @xmath44 without loss of generality .
the hamiltonian @xmath45 acts on a hilbert space spanned by distributing @xmath46 particles on the @xmath1 single - particle states .
then , a complete set of basis states is given by the set @xmath47 for fermions ( with @xmath48 ) , and @xmath49 for bosons .
the dimension of the hilbert spaces are , respectively , @xmath50 and @xmath51 .
this defines the @xmath0-body embedded gaussian ensembles of random matrices @xcite . by construction ,
the case @xmath52 is identical to the canonical ensembles of random matrix theory @xcite , i.e. , to the gaussian orthogonal ensemble ( goe ) for @xmath17 or the gaussian unitary ensemble ( gue ) for @xmath18 .
for @xmath53 , the matrix elements of @xmath45 may be identical to zero and display correlations .
the former property appears whenever there are no @xmath0-body operators that link together the @xmath10-body states , e.g. , for @xmath54 .
correlations arise because matrix elements of @xmath45 not related by symmetry may be identical . one of the notorious consequences of this is , for the bosonic ensemble in the dense limit ( @xmath55 , for @xmath0 and @xmath1 fixed ) that the ensemble is non - ergodic @xcite . in particular , for @xmath56 , @xmath57 and @xmath17 , the spectrum displays a significant number of quasi - degenerate states @xcite . as mentioned above , centrosymmetry is an important concept for an optimal efficiency @xcite .
a symmetric @xmath58 matrix @xmath59 is centrosymmetric if @xmath60=0 $ ] , where @xmath61 is the _ exchange matrix _
@xmath62 @xcite ; for complex hermitian matrices , centrohermiticity is defined when @xmath63 @xcite .
imposing centrosymmetry to the @xmath0-body embedded ensembles is subtle .
it can be introduced either at the one - particle level , which is the core for the definition of the @xmath0- and @xmath10-particle hilbert spaces , or at the @xmath0-body level , where the actual ( random ) parameters of the embedded ensembles are set , or at the @xmath10-body level , where the system evolves . considering a more realistic description which includes a one - body ( mean - field ) term and a two - body ( residual ) interaction , @xmath64
, it seems unnatural to define a specific transformation for each term separately .
hence , we shall introduce it in the one - particle space , and compute how is it transferred to the @xmath0-body and @xmath10-body space , which depends on @xmath43 as well .
then , in the one - particle space we define @xmath65 for @xmath25 , whose matrix representation in the one - body basis is precisely the exchange matrix . for two fermions ,
we define @xmath66 . in the last equality we keep the convention that the indexes are arranged in increasing order ; then , the fermionic anticommutation relations impose a global minus sign , which can be safely ignored .
this is generalized for @xmath0 particles as @xmath67 where we have dropped any global minus sign .
we note that in general the matrix @xmath68 , as defined by eq .
( [ cs ] ) , is not an exchange matrix , i.e. , the matrix with ones in the counterdiagonal and zeros elsewhere .
this follows from the possible existence of more than one state that is mapped by @xmath68 onto itself ; in this case , we shall say that @xmath68 is a _ partial _ exchange matrix . as an example , consider the case of fermions distributed in @xmath69 single - particle states ; for @xmath70 the @xmath0-particle space has dimension @xmath71 .
then , @xmath72 and @xmath73 , ignoring the minus signs mentioned above .
then , the entries in the @xmath74 matrix elements for these basis states are 1 in the diagonal .
in contrast , for the case @xmath69 and @xmath75 the resulting matrices @xmath76 and @xmath77 are exchange matrices . for bosons @xmath68 is defined as for fermions using eq .
( [ cs ] ) with the corresponding change of the creation operators . in this case , again , @xmath68 may not be an exchange matrix . as an example consider @xmath70 and @xmath78 ; then , we have @xmath79 , and likewise @xmath80 , which shows that there are more than one basis states mapped onto themselves under @xmath74 . yet , for the special case @xmath56 that we study below , @xmath68 is an exchange matrix for all @xmath0 , and @xmath81 ( in the @xmath10-particle space ) inherits the centrosymmetry ( or centrohermiticity ) of @xmath82 , i.e. , @xmath83=0 $ ] , with @xmath84 the exchange matrix of appropriate dimensions .
we analyze here the transport efficiency @xmath85 for bosons distributed in @xmath56 single - particle states , considering an arbitrary initial state @xmath86 and an arbitrary final one @xmath87 .
we address separately the cases where the @xmath0-body random interaction is invariant or not with respect to time - reversal ( @xmath17 or @xmath18 ) , and also when centrosymmetry ( or centrohermiticity ) is or not additionally imposed . in the simulations described below , we considered for concreteness the case with @xmath88 bosons , the corresponding hilbert space dimension is @xmath89 , the number of realizations of the ensemble is @xmath90 , and @xmath91 in eq .
( [ eq1 ] ) .
we discuss first the case where centrosymmetry or centrohermiticity , respectively for @xmath17 and @xmath18 , is absent .
the distributions of the efficiencies @xmath85 of the ensemble for any combination of the input and output states is in general rather broad , with negligible contributions to efficiencies above @xmath92 .
more interesting is to focus on the distribution of best efficiencies @xmath93 of each realization of the ensemble .
the results are presented in fig .
[ fig : fig1 ] .
the empty histograms correspond to @xmath17 and the shaded ones to @xmath18 .
with respect to time - reversal invariance , the case @xmath18 seems to yield marginally better efficiencies for a given value of @xmath0 , in the sense that the mean value of the distribution attains larger values and the distribution is somewhat narrower .
yet , we notice that the largest values of @xmath93 are slightly dominated by the @xmath17 case , except for @xmath13 . regarding the dependence on @xmath0 , the case
@xmath13 notably distinguishes itself as a special one .
first , the distribution of @xmath93 is dominated by a peak at @xmath94 , is wider than for other values of @xmath0 , and displays the largest number of realizations with efficiencies larger than the benchmark value @xmath92 ( vertical dashed line ) .
apart from this case , other values of @xmath0 display rather poor efficiencies , with smaller probabilities of reaching the benchmark value if they do it at all .
actually , the distributions for intermediate values of @xmath0 ( @xmath95 ) seem to be marginally better . in general , the states @xmath86 and @xmath87 that yield the best efficiencies differ by one boson in the occupation of each single - particle state , except for @xmath96 where the best efficiency is uniformly distributed for all pairs of states .
this statement is consistent with the fact that @xmath52 corresponds to a goe or gue . clearly , these unconstrained embedded gaussian ensembles yield poor transfer efficiencies , except maybe for @xmath13 .
we turn now to the best efficiencies of each realization of the ensemble when centrosymmetry or centrohermiticity is imposed .
as mentioned above , for the specific bosonic ensemble with @xmath56 , imposing centrosymmetry or centrohermiticity at the one - particle level carries over to the @xmath0- and @xmath10-particle spaces ; hence , @xmath97=0 $ ] , with @xmath84 the @xmath98 exchange matrix . in fig .
[ fig : fig2 ] we present the best efficiencies @xmath93 of each realization of the ensemble for this case . in comparison to the results presented in fig .
[ fig : fig1 ] , centrosymmetry ( centrohermiticity ) enhances dramatically the best efficiencies .
in particular , the case @xmath13 is remarkable since it displays perfect state transfer for _ all _ realizations of the ensemble for @xmath18 , while for @xmath17 there is a small remnant of a peak around @xmath94 .
quantitatively , for @xmath17 about @xmath99 of the realizations exhibit efficiencies that are larger than the benchmark value , while for @xmath18 this number is well above @xmath100 ; the mean of the distributions is @xmath101 and @xmath102 , respectively .
increasing @xmath0 shifts the mean value of the distribution towards smaller values and widens the distributions , being the effect larger for @xmath18 than for @xmath17 .
whereas for @xmath70 some realizations still display almost pst ( last bin of the histograms ) , this decreases for increasing values of @xmath0 . for @xmath17 , all values of @xmath0 exhibit realizations
whose best efficiencies are above the benchmark value , with better efficiencies appearing for smaller values of @xmath0 .
the states @xmath86 and @xmath87 that display the best efficiencies are those linked by centrosymmetry , i.e. , @xmath103 . while for @xmath52 all centrosymmetry - related pairs participate uniformly , for other values of @xmath0 there is a significant dominance of the pair of states where all bosons are initially in either of the single - particle states . from the perspective of the network ,
these states are located at the edges . as mentioned above and manifested in fig .
[ fig : fig2](a ) , the case @xmath13 is quite special .
indeed , for @xmath13 there are only three ( @xmath17 ) and four ( @xmath18 ) independent random matrix elements ; centrosymmetry ( centrohermiticity ) imposes that they become only two ( @xmath17 ) and three ( @xmath18 ) .
the consequence of this is that the random many - body centrosymmetric / centrohermitian hamiltonian has a constant main diagonal , which is proportional to the total number of bosons , and a second - diagonal whose matrix elements are proportional to @xmath104 , where @xmath105 is the boson occupation number of one of the single - particle levels for the many - body state in question . except for the constant diagonal and the random weights involved , for @xmath17
this matrix model has been shown to exhibit pst @xcite .
it can be shown that the eigenvectors of both models are identical , and the eigenvalues are related by a linear function .
this ensures the occurrence of pst as long as the off - diagonal matrix element of the random one - body interaction matrix is different from zero .
yet , the time scale where the pst occurs may be quite long , since it is inversely proportional to the absolute value of the off - diagonal matrix element of the random one - body matrix ; this explains the occurrence of other values for the efficiency , as shown in fig .
[ fig : fig2](a ) .
our results extend the validity of these statement to the broken time - reversal symmetric case and , more important , illustrate that they are robust under certain random centrosymmetric / centrohermitian perturbations that preserve the graph structure .
we emphasize the important role of centrosymmetry or centrohermiticity here : it guarantees that the diagonal matrix elements have a constant value , which results in pst among the edge states of the network .
we consider now the case of the fermionic embedded gaussian ensemble . for concreteness
we analyze the case @xmath106 , varying the number of fermions @xmath10 from @xmath107 to @xmath108 , and rank of interaction @xmath0 also from @xmath107 to @xmath108 . in general
, the distribution of efficiencies of the ensemble is rather broad with negligible contribution to efficiencies above @xmath109% and thus , we will focus on distribution of best efficiencies @xmath93 for each member of the ensemble .
we begin describing first the case where no centrosymmetry or centrohermiticity is considered .
the distribution of the best efficiencies @xmath110 of each realization are illustrated in fig [ fig : fig3 ] . in this figure , all frames in a row have the same number of fermions @xmath10 , and all frames in a column the same value of @xmath0 .
as before , the empty histograms correspond to @xmath17 and the filled ones to @xmath18 . in general , the lack of centrosymmetry or centrohermiticity yields poor efficiencies , with a marginal improvement for @xmath18 compared to the @xmath17 case .
interestingly , below and at half - filling ( first three rows ) only the case @xmath111 marginally reaches the benchmark value , while above half - filling ( last two rows of the figure ) the distributions of @xmath110 yield better results . in particular , the case @xmath112 is the only one where some realizations achieve the benchmark value . yet , the probability of such events is very small .
the distributions of the best efficiencies of each realization including centrosymmetry / centrohermiticity are displayed in fig .
[ fig : fig4 ] . as in the bosonic case
, @xmath17 seems to yield better efficiencies than the @xmath18 case . for @xmath13 , the distributions of @xmath93 for all @xmath10 surpass the benchmark value , and have a non - zero probability of displaying pst , specially for odd fermion number .
yet , the @xmath13 case does not correspond to the overall optimal situation : the probability that @xmath93 is above the benchmark value is clearly larger for @xmath112 and @xmath113 . indeed , while we find that nearly @xmath114 of the best efficiencies are larger than the benchmark for @xmath111 and @xmath17 ( similar values are obtained for @xmath13 and @xmath115 for @xmath116 ) , for @xmath117 and @xmath116 this percentage is close to @xmath118 .
for @xmath18 similar results are found with smaller percentages . while centrosymmetry clearly enhances the efficiencies , its role is more involved .
as we mentioned above , centrosymmetry / centrohermiticity is introduced at the one - particle level and then extended onto the @xmath0- and @xmath10-particle spaces using eq .
the corresponding @xmath68 and @xmath84 matrices are exchange matrices or partial exchange matrices depending on the parameters @xmath0 and @xmath10 , and @xmath1 . for the present case , @xmath106
, it can be shown that for odd particle - number space ( @xmath0 or @xmath10 ) the corresponding matrix ( @xmath68 or @xmath84 ) is a _ bona fide _ exchange matrix , whereas for even - values there are more than one states that are mapped onto itself , i.e. , the corresponding @xmath68 or @xmath84 matrix is a partial exchange matrix .
therefore , independent of @xmath0 , including the case where @xmath68 is a partial exchange matrix , the @xmath119 and @xmath116 many - body hamiltonians are fully centrosymmetric / centrohermitian . as shown in fig .
[ fig : fig4 ] , full centrosymmetry or centrohermiticity , in comparison with the partial cases , yield the best efficiency distributions .
when the many - body hamiltonian is partial centrosymmetric / centrohermitian , the best distributions correspond to the @xmath13 case .
in addition to centrosymmetry / centrohermiticity , the filling factor appears to play an important role : above half - filling the distributions of @xmath93 display the largest probability to yield efficiencies above the benchmark value ; in terms of @xmath0 , the optimal case corresponds to @xmath120 . regarding the pair of states that display the best efficiencies when centrosymmetry / centrohermiticity is imposed , as in the bosonic case , these occur among centrosymmetry related pairs , @xmath121 , excluding those which are mapped onto themselves by @xmath84 whenever @xmath84 is partial centrosymmetric .
these pairs of states appear uniformly distributed , that is , there is no special pair of states that display better transport properties .
this may be related to the fact that the fermionic graph is regular @xcite .
we have confirmed these results for @xmath122 , @xmath123 , @xmath13 , @xmath124 , @xmath71 and @xmath125 , @xmath126 , @xmath13 , @xmath124 , @xmath127 .
the primary aim of the present paper is to study transport of excitations in disordered networks with random @xmath0-body interactions .
this is important and certainly of interest because of possible applications in quantum communication protocols @xcite and artificial solar cells @xcite . towards this end
, we have studied the distribution of quantum efficiencies in disordered networks with many - body interactions , whose structure is modeled by the embedded gaussian ensemble , considering bosons and fermions , with and without time - reversal symmetry . in particular , we studied the role played by centrosymmetry / centrohermiticity , which is defined at the one - particle space , and then extended to the @xmath0- and @xmath10-particle spaces .
we have shown that ( full ) centrosymmetry enhances the efficiencies dramatically , being a requirement to have non - zero probability for pst ; the lack of centrosymmetry / centrohermiticity yields rather poor efficiencies . for bosons distributed in two single - particle levels ,
centrosymmetry is inherited in the @xmath0- and @xmath10-body spaces . in this case , pst is obtained for @xmath13 for almost all realizations ; the fact that our computation of the efficiencies involves an upper bound for the time , @xmath7 , constrains the relevant time scale for the achievement of the pst . in terms of @xmath0 ,
the probability of having pst decays with increasing @xmath0 .
however , we stress that for @xmath128 , the best efficient scenario is when @xmath129 , @xmath10 is the total number of bosons . with respect to the value of @xmath43 , the results are marginally better when time - reversal symmetry is preserved .
the pairs of states showing the best efficiencies are those at the edges of the network , i.e. , where all bosons are in one of the two single - particle levels
. then , in this case , state transfer corresponds to the physical transport of all bosons to the other single - particle state . for fermions
, we found that full centrosymmetry / centrohermiticity of the @xmath10-particle hamiltonian enhances considerably the best efficiencies , especially when the filling - factor is larger than @xmath130 .
we note that centrosymmetry / centrohermiticity is indroduced at the one - particle level , and then extended to the @xmath10-body space . for fermions , the rank of the interaction which displays the highest probability that the best efficiency @xmath93 is above the benchmark value corresponds to @xmath120
. a clear explanation of this is left open .
the pairs of states that yield the best efficiencies appear uniformly among those states linked by centrosymmetry .
previous results have shown that random perturbations on networks with pst destroy or affect significantly this property ; for details see @xcite and also @xcite and references therein .
our results show that , despite of the random character of the @xmath0-body interactions that we have considered , certain @xmath10-body networks display good efficiencies and may attain near perfect state transfer with non - zero probability .
our results could be exploited as new design principles of networks with good efficiency , which is preserved under certain many - body random perturbations .
for instance , considering the implementation of efficient quantum wires , it may be interesting to consider the case of filling factors that are smaller but close to one , where many - body interactions yield robustly very good efficiencies . finally , the results in our paper open the possibility to understand the good efficiency properties experimentally observed in exciton transport in biological systems , such as the fenna - matthews - olson complex @xcite .
it is a pleasure to thank thomas gorin , mattia walschaers and andreas buchleitner for enlightening discussions and correspondence .
we thank the financial support provided by the dgapa - unam project ig-101113 .
mv is dgapa - unam postdoctoral fellow .
lb acknowledges support through a ctedra marcos moshinsky ( 2012 ) . | the quantum efficiency in the transfer of an initial excitation in disordered finite networks , modeled by the @xmath0-body embedded gaussian ensembles of random matrices , is studied for bosons and fermions .
the influence of the presence or absence of time - reversal symmetry and centrosymmetry / centrohermiticity are addressed . for bosons and fermions ,
the best efficiencies of the realizations of the ensemble are dramatically enhanced when centrosymmetry ( centrohermiticity ) is imposed . for few bosons distributed in two single - particle levels
this permits perfect state transfer for almost all realizations when one - particle interactions are considered . for fermionic systems
the enhancement is found to be maximal for cases when all but one single particle levels are occupied . |
quantum monte carlo simulation , an undeniably useful tool in addressing a number of issues in quantum many - body physics , can not be used to simulate real - time dynamics . by means of analytic continuation to euclidean time @xmath3 ( @xmath5 ) , however , the schrdinger equation turns into a diffusion equation , which can be simulated using random walks which explore the potential landscape by spending proportionately more `` time '' near the valleys of the potential and less `` time '' near the potential heights .
the same transformation into imaginary time turns the path integral representation of the evolution operator from an integral over paths of computationally `` nasty '' phase factors , a problem almost impossible to treat stochastically , into an integral over paths in imaginary time weighted by a real and positive `` boltzmann - like '' weight ; this weight is interpreted as a well - behaved probability for a particular path to contribute to the sum and this interpretation allows a straightforward stochastic treatment@xcite .
this transformation , by itself , is useful because it yields the interacting ground state and physical quantities related to the equilibrium statistical mechanical description of a quantum many - body system .
however , if we are interested in obtaining information about the real - time dynamics and information about the excitations of the system , an `` inversion '' of this ill - defined transformation for the results of correlation functions @xmath0 of an operator @xmath6 representing a physical observable , namely , @xmath7 ( where @xmath8 is the imaginary - time ordering operator ) from the imaginary time @xmath3 back to real time @xmath9 is required .
the analytic continuation to real - time proceeds by means of a `` numerical inversion '' of the qmc data on @xmath0 to find the spectral function @xmath2 .
these inversion techniques , such as the so - called maximum entropy method@xcite or its generalization , the so - called stochastic analytical inference ( sai ) method@xcite , require very accurate qmc data on @xmath0 .
since collective phenomena emerge at energy scales significantly smaller than the typical short - range interaction energy scale , we are mainly interested in sampling the long imaginary part of such response or correlation functions .
such response functions obtained by qmc are noisy data and in a limited @xmath3 range .
if we are interested in extracting the low energy excitations , this information hides more clearly in the long - imaginary - time evolution of the correlation functions which is typically obscured by statistical errors .
flat histogram methods have been very useful in classical systems@xcite to overcome problems in simulations of first order phase transitions , systems with rough energy landscapes , etc .
the flat histogram idea has been extended in quantum many - body systems and , in particular , in stochastic series expansion@xcite to overcome the tunneling problem in first order phase transitions , in the continuous - time quantum monte carlo approach to the impurity solver problem used in dynamical mean - field - theory dmft@xcite and the diagrammatic monte carlo method@xcite . the main idea presented in this paper , in simple terms ,
is the following .
we show that the flat - histogram method can be applied to the qmc method itself to make the histogram of @xmath0 flat for all @xmath3 by sampling the variable @xmath3 , and keeping track of the factors in each imaginary - time interval needed to achieve this result . in this way
, we are able to compute @xmath0 in a greater @xmath3 range with significantly smaller stochastic error .
this approach allows us to achieve greater degree of accuracy when inverting the information contained in @xmath0 to find its corresponding spectral function @xmath2 in the entire range of values of @xmath10 of our interest . in order to demonstrate the idea in the present paper , we need to make specific choices of i ) a non - trivial quantum many - body problem , ii ) a specific qmc method , iii ) a specific correlation function @xmath0 , and iv ) a method to carry out the analytic continuation . using the same methods and techniques we will calculate the spectral function @xmath2 using qmc with and without the application of the flat histogram method during the qmc runs which produce the data on @xmath0 .
we will show that the flat histogram qmc method is superior to a simple qmc approach in which no flat histogram ideas have been implemented for such important observables . using our past specific experience with models and qmc techniques ,
we choose the problem of the green s function @xmath11 of a single hole in the @xmath4 model@xcite with the diagrammatic monte carlo ( dmc ) method@xcite using the method of the stochastic analytical inference@xcite for the analytic continuation to obtain the spectral function @xmath12 from the qmc data obtained for @xmath11 .
we apply the flat histogram idea with the combination of the dmc method and the wang - landau method , which we will refer to as the flat histogram diagrammatic monte carlo ( fhdmc ) method@xcite .
we will also use the standard implementation of the dmc@xcite where a guidance function is used when sampling @xmath11 . in the latter case , as we will see , the use of a parameter @xmath13 effectively makes the histogram of @xmath11 flat . on the other hand , we will also carry out dmc simulations using @xmath14 without the application of the flat histogram idea , which we will refer to as dmc0 which yields a histogram of @xmath11 very far from being flat , and we compare these results with those obtained with above two qmc methods , i.e. , with standard - dmc and fhdmc . this comparison is made in order to show the main point of our paper that if the flat histogram idea is incorporated in any qmc method , it will provide a more accurate analytic continuation to real time .
the paper is organized as follows . in sec .
[ method ] we describe the problem , the model , and the general approach which we will follow .
[ details ] describes the computational details and our implementation of the analytic continuation technique which we adopted . in sec .
[ results ] we present the results of the spectral function obtained with the dmc0 , the standard dmc , and fhdmc method , which are compared to what we believe to be the `` exact '' results of a soluble but restricted version of the @xmath4 model .
lastly , in sec . [ conclusion ]
we present the main conclusions of the paper .
in order to be precise we take the example where the operator @xmath15 is the single particle creation operator @xmath16 , in which case @xmath0 becomes the single particle green s function @xmath11 in imaginary time . in this case
the spectral function @xmath12 is related to @xmath17 as follows @xmath18 where the so - called kernel @xmath19 is simply @xmath20 .
the spectral function is a non - negative quantity normalized to unity .
we consider a finite imaginary time range , @xmath21 and we divide it into @xmath22 equal intervals . by integrating eq .
[ eq.1 ] in each time interval @xmath23 we obtain @xmath24 and @xmath25 is the average value of the kernel @xmath26 in the @xmath23 interval , i.e. , @xmath27 $ ] . in ref .
, we have shown that we can apply the flat histogram technique on the so - called diagrammatic monte carlo method@xcite to make the histogram @xmath28 of @xmath11 flat .
the idea was demonstrated on the frhlich polaron problem .
the results of the flat histogram dmc ( fhdmc ) on the estimate for the polaron ground state were significantly better than the result of the dmc0 .
however , as argued in the paper@xcite the polaron problem spectrum was characterized by a gap , and the full advantage of the flat histogram method over dmc0 could not fully demonstrated . in the present paper
we use a simplified version of the @xmath4 model , in which the heisenberg and the hole - hopping terms are linearized within the spin - wave approximation to obtain a polaron - like hamiltonian@xcite , i.e. , @xmath29 where the operator @xmath30 is the bogoliubov spin - wave creation operator , @xmath31 is the spin - wave dispersion of the square lattice quantum antiferromagnet@xcite and @xmath16 is the hole creation operator . here
@xmath32 is the coupling of the hole to spin waves and @xmath33 and @xmath34 are the coefficients of the the bogoliubov transformation as given in ref .
the single - hole spectral function of the above hamiltonian can be approximated by the non - crossing approximation(nca)@xcite .
the diagrammatic monte carlo ( dmc ) method with or without@xcite the incorporation of the flat histogram technique can be applied to the problem of a single - hole . in this paper in order to demonstrate the method , we wish to restrict our dmc to the sampling of only the diagrams contributing to the nca .
this restriction is actually a more difficult approach to numerically implement than the one in which the markov process samples all the connected diagrams contributing to @xmath11 .
the reason for restricting ourselves within the nca diagrammatic space , is because in this case there is an `` exact '' solution to the problem which we can use to measure the success of the approach discussed here . in order to obtain the `` exact '' solution ,
we have recalculated the single - hole green s function within the nca approach described in ref . for @xmath35 for a @xmath36 size square lattice and a finite imaginary part
was used in the free - hole propagator with @xmath37 ( in units of the hopping parameter @xmath9 ) which smears the lowest energy quasiparticle peak . note that as discussed in ref . on this size lattice
the finite - size effects were found to be small .
the results of these calculations will be used as default models and we will refer to them as `` exact '' solutions .
we have applied the dmc0 , the standard dmc , and fhdmc methods to obtain @xmath11 for @xmath35 and @xmath38 in the nca restricted diagrammatic space .
we have carried out qmc simulations for a range of @xmath39 to compare with the exact results in order to make sure that our computer programs are correct .
now imagine that we obtain a set of data @xmath40 @xmath41 , @xmath42 on @xmath28 obtained by any of the previously discussed three different qmc methods , where the @xmath43 denotes the data bin . for the analytic continuation we will use a generalization of the maximum entropy method , the so - called stochastic analytical inference ( sai ) technique . in the later approach , @xmath12 is obtained as the average over all its possible forms in a monte carlo integration where the particular form of @xmath12 is selected from a distribution determined by the so - called _ default model _ @xmath44 and the probability of any proposed form of @xmath12 to be the true form given the input data @xmath40 is given by @xmath45 & \sim & e^{-{{\chi^2[a ] } \over { 2 \alpha } } } , \label{eq.4}\end{aligned}\ ] ] where @xmath46 $ ] is the @xmath47 determined from the data @xmath40 on @xmath28 and the result which is obtained from eq .
[ eq.1 ] using the proposed form of @xmath12 . in the evaluation of @xmath46 $ ] the covariance matrix of the data is used .
the technical details of how one _ selects _ the particular proposed @xmath12 according to a given default model @xmath48 and other important details of this calculation are discussed in sec .
[ details ] . in the process of the analytic continuation
we will analyze the qmc data obtained with the three different qmc methods which we discussed above .
namely , first we will use dmc0 approach in which no guidance function has been implemented ( i.e. , @xmath14 ) .
second , we will use data obtained by means of standard - dmc ( sdmc ) where the histogram of @xmath49 is made approximately flat by multiplying @xmath49 with an exponential `` guidance '' function ( we use the same terminology used in ref .
@xcite ) of the form @xmath50 . in fig .
[ fig:1](a ) we compare the histograms of @xmath49 obtained with both dmc0 and standard - dmc .
it is clear that the standard dmc makes the histogram of @xmath49 approximately flat .
therefore , we will consider this method as an approximately flat histogram approach .
last , we will also use data obtained by the fhdmc .
notice that the histogram of @xmath49 which we obtain using the wang - landau algorithm is flat to better than @xmath51 .
furthermore , we will use three different default models in each of the above cases , i.e. , the total number of results presented will correspond to nine combinations of qmc data and default models . in one series of results presented in this paper , we will use as default model the `` exact '' solution to the problem . as a second choice for default model we will use a flat distribution , and as a third choice we will use a default model which shares some aspects of the `` exact '' solution but it differs significantly from the `` exact '' solution and we call it the `` incorrect '' default model .
these three choices are discussed in sec .
[ defaults ] in detail .
as discussed in the previous section , we will restrict our dmc approach to sample the diagrammatic subspace spanned only by the diagrams included in the non - crossing approximation@xcite .
the only reason for this restriction is that the problem of the single - hole spectral function can be solved `` exactly '' ( `` exact '' with the limitations discussed in the previous section ) and , this , can be used to test the accuracy of the results of qmc method .
furthermore , we will work with three sets of qmc data obtained for the histogram of @xmath11 for @xmath35 and @xmath38 .
one set is obtained with dmc0 , the second set is obtained with standard dmc , and the third set is obtained with fhdmc ; all three sets of data are obtained with approximately the same amount of cpu time in order to compare .
the first set of data for the histogram of @xmath11 is obtained for @xmath52 ( in units of the inverse hopping matrix element @xmath9 ) while the other two sets of data are obtained for @xmath53 and in all three sets of data we have used 600 @xmath3-intervals ( @xmath54 ) and the number of data bins @xmath55 for each time - interval . in fig .
[ fig:1](a ) we show that the histogram of @xmath49 obtained by the standard dmc is fairly flat over a very wide range of @xmath3 while that of the dmc0 is approximately an exponentially increasing function of @xmath3 .
therefore , we will consider that the standard dmc method is a method which produces an approximately flat histogram .
0.4 in 0.2 in 0.2 in the relative statistical error of the fhdmc results is more - or - less independent of @xmath3 ( see also fig . [ fig:1](b ) ) and it was approximately @xmath56 while for the results obtained using standard - dmc , @xmath57 ranges from @xmath58 for small @xmath3 up to @xmath59 for large @xmath3 . the relative error for the dmc0 results is significantly larger and of the order of @xmath60 ( see fig .
[ fig:1](b ) ) .
these relative errors were obtained by using approximately the same amount of cpu time such that the comparison between the three dmc methods to be meaningful . since our data on @xmath11 obtained by either dmc0 , standard
dmc or fhdmc are for a specific fixed value of @xmath61 , they will be simply denoted as @xmath62 and the histogram of @xmath11 will be denoted as @xmath63 , and @xmath2 is an abbreviation for @xmath12 for fixed @xmath61 .
there are @xmath64 data bins @xmath65 , @xmath66 obtained for @xmath67 in each imaginary time slice @xmath23 and @xmath68 . because as we produce the data table @xmath65 the imaginary time @xmath3 is sampled during any of the three dmc simulations , there is correlation between the data for the different imaginary time intervals .
therefore , in order to compute @xmath47 we need the data covariance matrix @xmath69 defined from its matrix elements as follows : @xmath70 \nonumber \\ & \times & [ \bar { g}^{(d)}(l)- g^{(d)}(l , j ) ] , \hskip 0.1 in k , l=1,2, ... ,l,\end{aligned}\ ] ] where @xmath71 is the average of the data for the @xmath72 time - slice , i.e , @xmath73 in terms of @xmath69 , @xmath74 and the proposed @xmath2 which gives a @xmath63 via eq .
[ eq.1 ] , the @xmath47 is given as @xmath75 where @xmath76 and @xmath77 .
we determine the orthogonal matrix @xmath78 and the diagonal matrix @xmath79 such that : @xmath80 then , we can simply write @xmath47 as follows : @xmath81 in the sai method @xmath2 is obtained as the average over all its possible forms by a metropolis monte carlo sampling , where the particular form of @xmath2 is selected from a distribution determined by the so - called _ default model _ @xmath48 and the probability of any proposed form of @xmath2 to be the true one , given the input data @xmath65 , is given by eq .
the @xmath46 $ ] is obtained using eq .
[ chis2 ] determined from the data @xmath65 on @xmath63 and @xmath63 which are determined from the assumed values of @xmath2 .
the optimum choice of the `` temperature '' @xmath82 is made according to the discussion in refs . .
we apply the metropolis algorithm using the expression given by eq .
[ eq.4 ] as the acceptance probability and we calculate the average spectral function @xmath2 . in the application of the metropolis algorithm we use as selection probability of a particular @xmath2 ,
a distribution which is related to the so - called default model @xmath48 , which contains our prior knowledge about the actual @xmath2 .
we consider the frequency interval @xmath83 $ ] and we assume that @xmath2 is zero outside this interval .
first , we slice this frequency interval into @xmath84 intervals @xmath85 around a middle - frequency @xmath86 , @xmath87 . we define a normalized histogram based on the default model @xmath48 as @xmath88 and based on it we define a new variable @xmath89 which takes values in the interval @xmath90 $ ] , which is sliced in intervals @xmath91 $ ] of width @xmath92 such that we can define the following @xmath93 we select normalized `` configurations '' of @xmath94 ( @xmath95 , @xmath96 ) from the uniform distribution@xcite .
the height @xmath97 of the histogram of @xmath2 in the interval @xmath85 of @xmath10 is obtained as @xmath98 in order to invert these data using the sai method discussed in the previous subsection , we need a default model .
we are going to use the following three default models .
\(a ) first , we will use the `` exact '' default model , i.e. , the solution obtained within nca by solving the dyson s equation self - consistently as was done in ref .
@xcite . for this purpose
we have re - calculated the spectral function for @xmath35 for @xmath38 on a @xmath36 size lattice and @xmath99 .
this result is close to the `` exact '' nca solution , but still has some small finite - size effects and finite-@xmath100 effects .
\(b ) a a second choice , we will use as a default model a flat distribution , i.e. , @xmath101 where @xmath102 and @xmath103 and @xmath104 .
\(c ) a third default model , which we will use is the following : using the approach described in ref .
we also recalculated the spectral function solution for @xmath12 for the same value of @xmath38 within nca but for @xmath105 .
this spectral function will be used as an option for a default model in the sai method .
the reason we would like to use this default model is to investigate the extent to which this method is capable of finding the correct solution starting from a default model which shares some features with the correct solution , such the location of the peaks but not others , such as the `` strength '' of the peaks
. this could be somewhat similar with a practical case scenario , where , for example , let us assume that we have performed a gw density - functional - theory based calculation for the quasiparticle spectral function of a real material , and , we have obtained more - or - less correct energy eigenvalues , but the quasiparticle wavefunctions are not close to the correct ones . in this case
the peaks will be close to the correct positions but the spectral weights should be different from the actual ones . in principle
, we can imagine that one starts a qmc simulation of a real material using as a default model the spectral function obtained in such a gw calculation .
0.2 in 0.2 in + 0.1 in 0.2 in + 0.1 in 0.2 in 0.3 in 0.3 in + 0.1 in + 0.1 in 0.3 in 0.4 in 0.2 in 0.2 in 0.5 in in fig .
[ fig:2 ] we present the results of the analytic continuation of the data obtained with dmc0 , the standard - dmc , and the fhdmc methods and the sai technique , and as default model the `` exact '' solution ( discussed in the previous section ) .
notice that the results using any sets of data are very close to the `` exact '' solution and independent of the `` temperature '' parameter @xmath82 which we used .
in general at higher value of @xmath82 the role of the default model is influencing more the result of the analytic continuation . since in practice
an exact solution is not available to use as the default model , we wish to examine the extent to which this method can be used for the analytic continuation .
therefore , we would like to test the method when we use a default model that is different from the exact answer . towards this goal ,
we have used the two different default models discussed in the previous section .
first we use the flat distribution given by eq . [ flat ] .
[ fig:3 ] presents the results of the application of the sai method using @xmath106 .
the results for lower values of @xmath82 are very similar .
the dmc0 method yields the results illustrated in fig .
[ fig:3](a ) ) .
the standard dmc method yields the results presented in fig .
[ fig:3](b ) ) , while the fhdmc approach yields the results illustrated in fig .
[ fig:3](c ) ) .
we least - squared fitted the dmc0 , the standard - dmc , and the fhdmc data on @xmath49 to the form @xmath107 and the values of the parameters @xmath108 and @xmath109 found by the fit are listed in table [ table1 ] .
the up - triangles in each of the subfigures show the values of the three frequencies found by the fits .
notice that the lowest two agree with the lowest two peaks found by the sai method .
in addition , the third frequency @xmath110 approximately accounts for the broad feature of the @xmath2 at high frequencies with approximately the correct spectral weight which is taken into account by the value of @xmath111 .
the value of @xmath112 , which corresponds to the residue of the lowest frequency quasiparticle peak , agrees rather well with the integral of @xmath2 around the quasiparticle peak ; thus , this agrees with the residue of the quasiparticle peak found in ref .
@xcite , which is approximately 0.2 .
.the results of fitting the dmc0 , the standard - dmc ( sdmc ) , and the fhdmc data on @xmath49 to the formula given by eq .
[ fit ] . [ cols="<,<,<,<,<,<,<",options="header " , ] we notice that the results of the analytic continuation can not reproduce the details of the exact solution even by changing the value of @xmath82 to a significantly lower value .
the overall performance of the standard dmc and the fhdmc methods are approximately the same and significantly better than the results obtained with the dmc0 method .
notice , however , that the width of the lowest energy peak is a little closer to the exact when we used the fhdmc data .
the default model used in the previous subsection assumes no a priori knowledge about the features of @xmath12 .
therefore , it heavily relies on minimizing @xmath47 . in many cases in practice , however , we have some partial information from either approximate analytic or semi - analytic ( for example , perturbation expansion ) or numerical techniques , such as from the density functional theory in electronic structure calculations .
it is typical to know the approximate location of the peaks , however , it is much harder to know the spectral weight of these peaks .
next we will use the `` exact '' spectral function @xmath113 obtained for @xmath114 as the default model in the sai approach to invert the dmc0 , the standard - dmc , and the fhdmc data on @xmath115 . as can be seen from fig .
[ fig:4 ] , most of the peaks of the default model are close to those of the `` exact '' solution .
the lowest energy peak is in the wrong place and the relative spectral weights ( i.e. , the heights of the peaks ) are very different than those in the `` exact '' solution . in fig.[fig:4
] we present the results of applying the sai method to data obtained with dmc0 , the standard dmc , and the fhdmc method using @xmath106 and as default model the above discussed `` incorrect '' solution which corresponds to @xmath116 instead of the correct value of @xmath117 .
fig.[fig:4](a ) presents the results obtained using the dmc0 data . in fig.[fig:4](b ) and fig .
[ fig:4](c ) we present the results of applying the sai method to data obtained with the standard dmc and the fhdmc method respectively .
the default model is shown by the green open circles .
notice that the overall performance of the standard dmc and the fhdmc methods is approximately the same and significantly better than that of the dmc0 data .
notice , however , that the width of the lowest energy quasiparticle peak when using the fhdmc data is somewhat smaller and closer to the `` exact '' , as compared to that obtained with the standard dmc data . also note that the location of the second peak is closer to the `` exact '' location when using the fhdmc data .
these differences in the degree of approximating the low - energy features of the spectral function can be explained when we compare the relative error in @xmath49 as obtained from the two methods as was done in fig .
[ fig:1](b ) . fig .
[ fig:1](b ) compares our results for the relative error for the @xmath11 as a function of @xmath3 as obtained with the dmc0 , fhdmc and standard - dmc for the approximately the same amounts of cpu time . for the case of dmc0
we have chosen @xmath118 in order to obtain a comparable error in @xmath49 with the other two methods .
first notice that for a fixed value of @xmath119 and for a fixed number of monte carlo steps in the dmc0 method , the relative error is weakly dependent of @xmath3 yielding an average value @xmath120 for the relative error .
however , for fixed number of monte carlo steps , this average relative error @xmath120 in the dmc0 method grows exponentially with @xmath119 . by choosing @xmath121
we have optimized the quality of the analytic continuation . while the greater range of @xmath3 allows us to obtain better resolution for small values of @xmath10 , if we increase the value of @xmath122 beyond the above chosen value , the errors grow to the point that the quality of the analytic continuation becomes much worse .
this is the reason for the superiority of the standard dmc and fhdmc methods .
we further note that the error in the standard - dmc method at short time @xmath3 is much better than the error in the fhdmc data .
however , the situation reverses at long imaginary time .
this helps the extraction of the low - energy physics of the problem . in the case of the present example , the crossing point of the relative error between the two methods occurs at @xmath123 , which implies that the features of the spectral function obtained with the two methods will differ for @xmath124 , ( where @xmath125 is measured from the lowest energy state ) . for the present problem this seems like a narrow range , however , in general , for other problems where we are interested in the low energy excitations it can be more important
we have shown that by combining flat histogram techniques@xcite and the qmc method in such a way to make the histogram of an imaginary - time correlation function @xmath49 flat , we are able to carry out accurate analytic continuation to real time with better degree of accuracy as compared to any quantum monte carlo method which does not use the flat histogram idea . this has been demonstrated using the dmc method which can be modified to simulate an exactly soluble problem .
this is within the flexibility of the dmc method , because we can select the set of feynman diagrams to sample .
thus , by selecting the diagrammatic series to be the one which takes into account all the diagrams which contribute to the single - hole spectral function of the 2d @xmath4 model within the nca approximation , we can compare the results of the analytic continuation directly to the `` exact '' solution@xcite . to carry out the analytic continuation to real time
, we have used the sai@xcite approach , a generalization of the maximum entropy method . from this paper
we draw the following main conclusions : \i ) the main point of the present paper is the following .
we find that if the qmc data are obtained by a technique in which the histogram of @xmath49 is more or less flat , such as the fhdmc or the standard dmc , the results of the analytic continuation improve significantly as compared to the results obtained by using a qmc method ( such as the dmc0 method ) in which the flat histogram idea has not been applied .
that is because we can approach large imaginary time with significantly small statistical error .
\ii ) this paper provides benchmarking on the power of the fhdmc and standard dmc methods to provide data for analytic continuation to real time using an exactly soluble model . in particular comparing these two methods
, we find that : * the standard dmc method requires a value of @xmath13 in order to approximate the long - time behavior of @xmath49 and this can be used to effectively make the histogram of @xmath49 approximately flat .
thus , the standard dmc method requires some amount of prior work in order to determine the optimum value of @xmath13 .
the fhdmc method makes the histogram of @xmath49 flat automatically without requiring any such a priori knowledge or prior work . *
the error on the short - time behavior of @xmath49 is significantly smaller in the case of standard dmc . *
the error on the long - time behavior of @xmath49 is smaller in the case of fhdmc .
this allows us to obtain somewhat better quality data on the spectrum and spectral weight of the very low - energy excitations .
this aspect of the fhdmc method seems to give some advantage over the standard dmc , when we are interested in the very low - energy excitations of a given system .
\iii ) lastly , we exploit the fact that we know the `` exact '' solution to this problem to draw the following additional conclusions from the present calculations : * when we use as default model , a distribution which is close to the exact , the result of the analytic continuation is very close to the exact result . *
if the correct answer is completely unknown and we use as default model the flat distribution , the result for the location for and the residue of the lowest energy quasiparticle peak is very close to the exact value , and the location and width of the second lowest peak is also in agreement with the `` exact '' answer .
the details of the spectral function can not be correctly captured by the analytic continuation using a flat default model .
* if we approximately know some aspects of the solution , i.e. , we use a default model which shares some features of the correct solution ( such as the location of some of the peaks , but not their relative strength in the spectral function ) the analytic continuation tends to correct the strength of the peaks and tends to move the incorrectly located peaks of the default model towards the correct positions .
the main point of the present paper is that , in general , if we incorporate the flat histogram idea with any qmc technique , in such a way to make the histogram of @xmath49 flat , the results of the analytic continuation to real time should be of better statistical quality . 0.2 in
this work was supported in part by the u.s
. national high magnetic field laboratory , which is partially funded by the nsf dmr-1157490 and the state of florida . | the quantum monte carlo ( qmc ) method can yield the imaginary - time dependence of a correlation function @xmath0 of an operator @xmath1 .
the analytic continuation to real - time proceeds by means of a `` numerical inversion '' of these data to find the response function or spectral density @xmath2 corresponding to @xmath1 .
such a technique is very sensitive to the statistical errors in @xmath0 especially for large values of @xmath3 , when we are interested in the low - energy excitations . in this paper
, we find that if we use the flat histogram technique in the qmc method , in such a way to make the _ histogram of _ @xmath0
_ flat _ , the results of the analytic continuation for low - energy excitations improve using the same amount of computational time . to demonstrate the idea we select an exactly soluble version of the single - hole motion in the @xmath4 model and the diagrammatic monte carlo technique . |
in the last 20 years , great progress has been made in the study of a new branch of nuclear physics , i.e , the isospin nuclear physics @xcite .
theoretical studies have shown that , within the parabolic approximation , the energy per particle in asymmetric nuclear matter can be approximately expressed as @xmath1 , where @xmath2 is the isospin asymmetry parameter and @xmath3 is the density - dependent nuclear symmetry energy .
the latter has been studied for decades due to its great importance in both nuclear physics and astrophysics @xcite .
although significant progress has been made , the symmetry energy is still subject to uncertainties especially at high - density region @xcite .
nowadays , many sensitive observables have been identified as promising probes of the symmetry energy , such as the @xmath4 ratio @xcite , energetic photon as well as @xmath5 @xcite , the neutron to proton ratio @xmath6 @xcite , @xmath7 @xcite , the isospin fractionation @xcite and the neutron - proton differential flow @xcite .
however , one only knows these observables are in general sensitive to the high - density or low - density behaviors of the symmetry energy at certain beam energy whereas none knows the decomposition of sensitivity of the symmetry energy observables in the whole density region .
such knowledge surely affects obtaining information of the density - dependent symmetry energy from comparisons of theoretical simulation and experimental data . in this study , based on the isospin - dependent transport model ,
we address the above question .
( color online ) evolution of the central baryon density in the reaction of @xmath8 at beam energies of 400 and 200 mev / nucleon with an impact parameter of b=1 fm .
@xmath0 denotes the nuclear saturation density.,scaledwidth=50.0% ] ( color online ) evolution of the distribution percentage of baryon number in different density regions in the central reaction of @xmath8 at beam energies of 400 and 200 mev / nucleon , respectively.,scaledwidth=50.0% ] in this study , for simplicity , we use the isospin and density - dependent single particle potential @xmath9 where the isoscalar potential reads @xmath10 this soft nuclear isoscalar potential ( sbkd ) with @xmath11 = 200 mev , was firstly introduced by bertsch , kruse and das gupta @xcite . for the isovector potential @xmath12 , we use the form @xcite @xmath13 where @xmath14 is the reduced bayron density and @xmath15 = 1/2 ( -1/2 ) for neutrons ( protons ) .
this symmetry potential roughly stands for one of the frequently used symmetry potential in nuclear transport models .
to probe the density - dependent symmetry energy , it is instructive to know the time evolution of central maximal baryon density reached in heavy - ion collisions . figure [ density ] shows the evolution of the central maximal baryon density reached in @xmath8 reaction at beam energies of 400 and 200 mev / nucleon with an impact parameter of b=1 fm .
one can see that the central maximal baryon density reached is about 2.5@xmath16 for the incident beam energy of 400 mev / nucleon and about 2.0@xmath16 for 200 mev / nucleon case .
and one can also see that the supra - density nuclear matter exists longer for 200 mev / nucleon case than that for 400 mev / nucleon case . in order to better understand our target of studying the sensitivity of symmetry - energy - sensitive observables in different density regions ,
as shown in figure [ percentage ] , we plot the evolution of the distribution percentage of baryon number in different density regions .
we can see that the distribution percentage of baryon number in different density regions changes with reaction time .
more baryons lie in the density region around [email protected]@xmath16 in the whole reaction process . at the beginning of the reaction ( t @xmath18 1 fm / c ) nucleons in the two colliding nuclei
are slightly compressed , thus more nucleons lie in the density region @xmath19 . immediately after this
, the central density of the reaction increases rapidly due to compression as shown in figure [ density ] .
since baryons lie in different density regions , it is thus necessary to study in which density region the symmetry - energy - sensitive observables show maximal sensitivity . in order to know in which density region the frequently used symmetry energy
sensitive observables show maximal sensitivity to the symmetry energy , similar with the study in ref .
@xcite , in the whole density region ( @xmath20 ) we use the @xmath21 as the standard calculation , which gives a value of one observable @xmath22 , i.e. , @xmath23 to see the relative sensitivity of this observable in different density regions ( i.e. , @xmath24 ) , we turn off the symmetry energy in one density region but keep the symmetry energy in the residual density region .
we thus get the other value of this observable @xmath25 , i.e. , @xmath26 through comparing these new computational results with the standard calculation @xmath22 , one can obtain the relative sensitivity in a certain density region , which reads @xmath27 in the following , we demonstrate decomposition of the sensitivity of the frequently mentioned observables in the literature . relative sensitivity ( @xmath28 ) of the symmetry - energy - sensitive observable @xmath4 as a function of density.,scaledwidth=50.0% ] relative sensitivity ( @xmath28 ) of the symmetry - energy - sensitive observable pre - equilibrium free n / p as a function of density.,scaledwidth=50.0% ] the @xmath4 ratio was first proposed as a probe of nuclear symmetry energy in 2002 by _ li _ @xcite .
it is generally considered to be sensitive to the high - density behavior of the symmetry energy . shown in figure [ rpion400 ]
is the decomposition of the sensitivity of the symmetry energy observable @xmath4 in different density regions . from figure [ rpion400 ]
we can clearly see that the maximal sensitivity of @xmath4 to the density - dependent symmetry energy lies in the density region around @xmath29 . above @xmath30 , sensitivity of @xmath4 to the symmetry energy
is roughly equal to that below @xmath16 .
with increase of density , collision effect becomes larger .
thus one can see that the maximal sensitivity does not lie in the maximal density region reached .
due to rescatterings of pion meson in the low - density region , charged pion ratio @xmath4 is also affected by the low - density behavior of the symmetry energy in some degree .
in general , the sensitivity of charged pion ratio @xmath4 to the nuclear symmetry energy is larger in the high density region @xmath31 than that in low density region @xmath32 . shown in figure [ rnp400 ] is decomposition of the sensitivity of the symmetry energy observable pre - equilibrium free neutron to proton ratio
n / p in different density regions . from figure [ rnp400 ] , we can clearly see that the maximal sensitivity of the free neutron to proton ratio
n / p at the incident beam energy of 400 mev / nucleon lies in the density region around @xmath30 .
because nucleon emission at pre - equilibrium of the reaction does not suffer scatterings from the low - density matter , the symmetry energy in the low density region thus has minor effect .
same as figure [ rpion400 ] , but at the beam energy of 200 mev / nucleon.,scaledwidth=50.0% ] same as figure [ rnp400 ] , but at the beam energy of 200 mev / nucleon.,scaledwidth=50.0% ] it is of interest to make similar study at incident beam energy below 400 mev / nucleon since such experiments of probing the symmetry energy and related are being planned or performed at riken in japan @xcite . shown in figure [ rpion200 ] is decomposition of the sensitivity of the symmetry energy observable @xmath4 in different density regions in au+au reactions at the beam energy of 200 mev / nucleon .
we can see that the maximal sensitivity of @xmath4 to the density - dependent symmetry energy lies also in the density region around @xmath29 . and
it is in general still sensitive to the high density ( @xmath33 ) behavior of the symmetry energy .
more interestingly , we can see that the sensitivity of charged pion ratio @xmath4 to the symmetry energy is about two times larger than that at 400 mev / nucleon beam energy .
therefore , one uses charged pion ratio @xmath4 to probe the high - density behavior of the symmetry energy , it is better to do experiments at relatively lower incident beam energy @xcite . as for the observable pre - equilibrium free neutron to proton ratio at incident beam energy of 200 mev / nucleon , it is shown in figure [ rnp200 ] .
the maximal sensitivity of the pre - equilibrium free neutron to proton ratio
n / p at the incident beam energy of 200 mev / nucleon also lies in the density region around @xmath30 .
in general , the symmetry energy in the low density region has minor effect . as shown in figure [ percentage ] , because only a small percentage of baryon lies in the maximum density region reached and also in the maximum density there are larger collision effects , symmetry - energy - sensitive observables are not sensitive to the symmetry energy in the maximal baryon - density region . to probe the symmetry energy at higher densities ,
one thus needs heavy - ion collisions at even higher beam energy .
however , the larger collision effects at higher beam energy increase the difficulty of studying nuclear symmetry energy .
in summary , within the isospin dependent ibuu transport model , we studied the au+au reaction at incident beam energies of 200 and 400 mev / nucleon .
it is found that the symmetry - energy - sensitive observables including charged pion ratio @xmath4 , pre - equilibrium free neutron - proton ratio @xmath6 mainly probe the density - dependent symmetry energy around 1.25@xmath0 ( for pion emission ) or 1.5@xmath0 ( for pre - equilibrium nucleon emission ) .
effects of the symmetry energy in the low - density region is in general small but observable . since the symmetry - energy - sensitive observables are not sensitive to the symmetry energy in the maximal baryon - density region
, it is therefore a challenge to probe the symmetry energy at higher densities .
this work is supported in part by the national natural science foundation of china under grant nos .
11375239 , 11435014 , 11275073 and the fundamental research funds for the central university of china under grant no .
2014zg0036 .
li , w. udo schrder(eds ) , isospin physics in heavy - ion collisions at intermediate energies , nova science , new york , 2001 .
steiner , m. prakash , j.m .
lattimer , et al .
, phys . rep . *
411 * , 325 ( 2005 ) .
v. baran , m. colonna , v. greco , m. di toro , phys .
rep . * 410 * , 335 ( 2005 ) .
chen , and c.m .
ko , phys .
rep . * 464 * , 113 ( 2008 ) .
k. sumiyoshi , h. toki , astrophys .
j. * 422 * , 700 ( 1994 ) .
p. danielewicz , r. lacey , and w.g .
lynch , science * 298 * , 1592 ( 2002 ) .
lattimer and m. prakash , science * 304 * , 536 ( 2004 ) ; phys . rep . *
442 * , 109 ( 2007 ) .
tsang , yingxun zhang , p. danielewicz , m. famiano , zhuxia li , w.g .
lynch , a.w .
steiner , phys .
lett . * 102 * , 122701 ( 2009 ) .
tsang , z. chajecki , d. coupland , p. danielewicz , f. famiano , r. hodges , m.kilburn , f. lu , w.g .
lynch , j. winkelbauer , m. youngs , yingxun zhang , prog .
. phys . * 66 * , 400 ( 2011 ) .
m. b. tsang , j. r. stone , f. camera , p. danielewicz , s. gandolfi , k. hebeler , c. j. horowitz , jenny lee , w. g. lynch , z. kohley , r. lemmon , p. moller , t. murakami , s. riordan , x. roca - maza , f. sammarruca , a. w. steiner , i. vida@xmath34a , s. j. yennello , phys .
c * 86 * , 015803 ( 2012 ) .
li , phys .
lett * 88 * , 192701 ( 2002 ) .
t. gaitanos , m.di toro , s. typei , v. baran , c. fuchs , v. greco , and h.h .
wolter , nucl .
* a732 * , 24 ( 2004 ) .
li et al . , phys .
c * 72 * , 034613 ( 2005 ) .
f. gulminelli and p. chomaz , phys .
c * 71 * , 054607 ( 2005 ) .
yong , b.a .
chen , w. zuo , phys .
c * 73 * , 051601 ( 2006 ) .
xiao et al , nucl .
phys * 36 * , 064040 ( 2009 ) .
y. gao , l. zhang , w. zuo , j.q .
li , phys .
c * 86 * , 034611 ( 2012 ) .
yong , b.a .
li , phys .
b * 723 * , 388 ( 2013 ) .
xiao , g.c .
yong , l.w .
chen , b.a .
li , m. zhang , g.q .
xiao , n. xu , eur .
j. * a50*,37 ( 2014 ) .
yong , b.a .
chen , phys .
b * 661 * , 82 ( 2008 ) .
li , phys .
lett * 78 * , 1644 ( 1997 ) .
w.p . tan et al .
c * 64 * , 051901 ( 2001 ) . v. baran et al .
* a703 * , 603 ( 2002 ) .
l.w . chen , c.m .
ko , and b.a .
li , phys , rev .
c * 68 * , 017601 ( 2003 ) ; nucl . phys . * a729 * , 809 ( 2003 ) .
y. zhang and z. li , phys .
c * 71 * , 024604 ( 2005 ) .
h. muller and b. serot , phys .
c * 52 * , 2072 ( 1995 ) .
li and c.m .
ko , nucl .
a * 618 * , 498 ( 1997 ) .
lett . * 85 * , 716 ( 2000 ) .
li , phys .
lett * 85 * , 4221 ( 2000 ) . v. greco et al .
c * 67 * , 015203 ( 2003 ) .
bertsch et al .
c * 29 * , 673 ( 1984 ) .
li et al . , phys .
c * 69 * , 011603 ( 2004 ) .
liu , g.c .
yong , d.h .
wen , phys .
c * 91 * , 024604 ( 2015 ) .
f. zhang , y. liu , g.c .
yong , w. zuo , chin .
* 29 * , 052502 ( 2012 ) . | to exactly answer which density region that some frequently used symmetry - energy - sensitive observables probe , for the first time , we make a study of decomposition of the sensitivity of some symmetry - energy - sensitive observables .
it is found that for the au+au reaction at incident beam energies of 200 and 400 mev / nucleon , frequently used symmetry - energy - sensitive observables mainly probe the density - dependent symmetry energy around 1.25@xmath0 ( for pionic observables ) or 1.5@xmath0 ( for nucleonic observables ) .
effects of the symmetry energy in the low - density region is in general small but observable .
the fact that the symmetry - energy - sensitive observables are not sensitive to the symmetry energy in the maximal baryon - density region increases the difficulty of studying nuclear symmetry energy at super - density . |
decades after the discovery of dark matter around spiral galaxies little is known about dark matter in early - type galaxies .
mass tracers at large radii ( such as stellar kinematics , kinematics of globular clusters and of planetary nebulae , and x - ray halos ) generally indicate that a constant mass - to - light ratio can not reproduce the observations , although there are typically large uncertainties and a wide variety of behavior is seen .
the main source of uncertainty in interpreting kinematic measures is that the derived mass profile depends on the assumed orbital structure .
this is commonly referred to as the mass - anisotropy degeneracy .
the situation is even more uncertain outside the local universe , because traditional kinematic tracers at large radii are generally not feasible at cosmological distances .
however , additional constraints can be gathered by looking at early - type galaxies that are gravitational lenses .
the configuration of multiple images provides information on the mass distribution of the lens ; first and foremost a robust measurement of the mass enclosed within the einstein radius ( typically larger than the effective radius ) , _ independent _ of its kinematic status . the lenses structure and dynamics ( lsd ) survey ( koopmans & treu 2002 , 2003 ; treu & koopmans 2002a,2003 ; hereafter kt ) is obtaining internal kinematics for a sample of of 11 relatively isolated e / s0 lens galaxies at @xmath1 . a joint lensing and dynamical analysis
is used to break the mass - anisotropy degeneracy and determine the amount and distribution of luminous and dark matter .
the first results from this project are briefly summarized and discussed in section 2 . at the cluster scale
, the presence of dark matter has been known for a long time ( zwicky 1937 ) .
the distribution of dark matter has been probed over a wide range of scales using x - ray analysis , dynamical studies and gravitational lensing , showing that dark matter dominates in mass .
the wealth of mass tracers in clusters in particular the existence of giant radial arcs makes them an ideal laboratory to test the existence of universal mass density profiles predicted by cold dark matter ( cdm ) simulations . combining a lensing analysis of clusters with giant arcs with a dynamical analysis of the kinematics of the brightest cluster galaxy ( bcg )
, we can measure very accurately the mass density profile within @xmath2 kpc , disentangle the stellar and dark matter , and pinpoint the inner logarithmic slope of the dark matter halo .
this has motivated a search of suitable systems with a central dominant bcg and giant arcs through the entire hst - wfpc2 archive ( sand et al .
2004 , in preparation ) , and the spectroscopic follow - up at keck ( sand , treu & ellis 2002 ; sand et al .
the first results from this project are briefly summarized and discussed in section 3 .
figure 1 illustrates via the lens 0047 - 281 ( @xmath3 ; warren et al .
1996 ) the data we are collecting as part of the lsd survey . archival hst imaging data ( mostly from the castles collaboration ) are combined with stellar kinematic data obtained using esi at the keck - ii telescope . in excellent conditions ( seeing @xmath4 )
it is possible to obtain a spatially resolved stellar velocity dispersion profile ( hatched boxes in the right panel ) out to the einstein radius ( typically 1 - 2 effective radii , 1 - 2 arcseconds ) .
a joint lensing and dynamical analysis is used to constrain the free parameters of a family of two - component mass models .
one component represents the stellar mass and follows the hst luminosity profile scaled by a stellar mass - to - light ratio , the other component represents the dark matter halo and is modeled as a generalized navarro , frenk & white ( 1997 ) profile ( see tk for details ) .
the main results from analysis of the first two objects ( 0047 - 281 and mg2016 ) are : \1 .
extended dark matter halos are detected in high redshift e / s0 galaxies at high significance ( @xmath5 @xmath6 ) ; dark matter contributes 5075 % of the total mass within the einstein radius .
the evolution of the stellar mass - to - light ratio with redshift , obtained from a two - component dynamical model , agrees with the independent estimate obtained by studying the evolution of the fundamental plane with redshift ( treu et al .
2002 ; van dokkum & ellis , gebhardt et al .
this is consistent with no structural or dynamical evolution of e / s0s between @xmath7 and today .
\3 . the effective slope ( i.e. the average slope measured between roughly the effective and einstein radius ) of the _ total _
mass distribution is very close to isothermal , i. e. @xmath8 with @xmath9 ( to within 5% ) , which we interpret as a possible indication of ( incomplete ) violent relaxation .
this result , if confirmed by a larger sample , indicates that already at @xmath0 there is a mechanism by which dark and luminous component conspire with each other to produce almost perfectly flat rotation curves , while still preserving their spatial segregation . clearly ,
analysis of the rest of the sample is essential to confirm and extend these findings .
for example , an open question which we intend to address and has profound cosmological implications
is understanding what is the distribution of the effective slope of total mass density profile .
an intrinsic scatter is expected on the basis of local studies ( e.g. gerhard et al .
it is necessary to quantify this scatter in order to estimate the uncertainties in the determination of the cosmological parameters from lens statistics , of velocity dispersion of early - type galaxies from image separation , and of the hubble constant from gravitational time delays .
the hubble constant is highly sensitive to the slope of the mass density profile , and some external information is needed if one wants to measure it with precision better than a few tens of per cent . in an effort parallel to the lsd survey , an enlarged collaboration is targeting lenses with known time - delays for spectroscopic follow - up with keck and vlt .
although this is generally complicated by the presence of bright qso images outshining the lens galaxy , analysis of the first two systems ( treu & koopmans 2002a ; koopmans et al .
2003 ) confirms that with suitable kinematic data it is possible to measure h@xmath10 at @xmath1115% accuracy from each individual system , _ including the uncertainty on the mass density profile_. we are collecting data for more systems , with the goal of measuring h@xmath10 from time - delays with precision comparable to the hst key project ( freedman et al .
as part of our study of the inner regions of clusters of galaxies we have measured arc redshifts and bcg stellar velocity dispersions for 6 clusters of galaxies ( figure 2 ; sand et al .
2002 , 2003 ) .
three systems ( upper panels ) are selected to have both a radial and a tangential gravitational arc , while three systems ( lower panels ) are selected to have only tangential arcs .
the location of the symetry break in a radial arc constrains the derivative of the enclosed mass and therefore the radial arc systems provide very tight constraints on the shape of the mass density profile .
however , the radial critical line disappears when the total mass density profile becomes steeper then @xmath12 .
therefore a radial arc selected sample may possibly be biased toward flatter mass profiles .
for this reason we introduced the tangential arc only systems as a control sample .
a joint lensing and dynamical analysis is performed , fitting a family of two component mass models to the data ( see sand et al .
2003 for details ) .
specifically , the dark matter mass component is parametrized as a generalized nfw profile ( navarro , frenk & white 1996,1997 ; moore et al .
1998 ) , i.e. the dark matter density goes as @xmath13 at small radii and @xmath14 at large radii .
numerical simulations predict @xmath15 in the range 1 - 1.5 ( hereafter , respectively , nfw and moore slopes ) , although numerical issues are still being debated ( see sand et al . 2003 and references therein ) .
figure 3 shows the posterior probability density for @xmath15 for the radial ( left ) and tangential ( right ) arc samples .
the radial arc sample indicates generally flat inner slopes , the average slope being inconsistent with the cdm `` cusps '' at more than 99.99% cl .
however , intrinsic scatter is detected and individual systems can be as steep as nfw ( rxj1133 )
. the tangential arc systems also confirm flatter slopes than @xmath16 , indicating that the sample with radial arcs is unbiased . the implications of this discrepancy could be profound . either the cdm - only simulations are insufficiently accurate and/or contain insufficient physics to describe faithfully the inner regions of clusters .
for example , baryons could play an important dynamical role driving energy and angular momentum out of the cluster core through mechanisms such as dynamical friction ( e.g. el - zant et al .
alternatively , the dm could be partially self - interacting , or manifest some interaction with the baryons ( spergel & steinhardt 2000 ) .
either outcome would be very exciting .
on the one hand , numerical simulations are being developed which will permit the realistic inclusion of baryons .
finding mechanisms that satisfy our observational constraints will be invaluable in understanding the complex physics of star formation and dissipative processes .
and of course , there is the possibility that there is some fundamental flaw in cdm which will necessitate considering more exotic scenarios , like self - interacting dm .
we are currently looking to expand our sample , drawing from the candidate radial arc systems identified from the hst - wfpc2 archive ( sand et al .
this would improve our determination of the mean of @xmath15 and also provide a first measure of the intrinsic scatter , which is essential for any meaningful comparison with simulations , especially those that include baryons .
tt thanks the organizers for this pleasant and exciting meeting .
tt thanks eric agol , stefano borgani , amr el - zant , raul jimenez , masataka fukugita , jean - paul kneib , ben moore , joel primack , paul schechter , jerry sellwood and simon white for stimulating conversations , during and around the sydney meeting .
tt gratefully acknowledges support from an aas travel grant and from hst proposals ar-09222 , ar-09527 , ar-09960 , provided by nasa through a grant from stsci , which is operated by aura , incorporated , under nasa contract nas5 - 26555 .
freedman et al .
2001 , apj , 553 , 47 el - zant , a. et al .
2003 , apj , submitted , astro - ph/0309412 gerhard , o. , kronawitter , a. , saglia , r. p. , & bender , r. 2001 , , 121 , 1936 gebhardt , k. et al .
2003 , apj , in press , astro - ph/0307242 koopmans , l.v.e . & treu , t. 2002 , apj , 568 , l5 koopmans , l.v.e . & treu , t. 2003 , apj , 583 , 606 koopmans , l.v.e . ,
treu , t. , fassnacht , c.d . , blandford , r.d . & surpi g. 2003 , apj , 599 , in press moore , b. , governato , f. , quinn , t. , stadel , j. & lake , g. 1998 , , 499 , l5 navarro , j. , frenk , c. s. , & white s. d. m 1997 , , 490 , 493 sand , d.j . ,
treu , t. & ellis r.s . 2002 , apj , 574 , l129 sand , d.j . ,
treu , t. , smith , g.p . & ellis , r.s .
2003 , apj , submitted , astro - ph/0309465 treu , t. & koopmans , l.v.e .
2002a , apj , 575 , 87 treu , t. & koopmans , l.v.e .
2002b , mnras , 337 , l6 treu , t. & koopmans , l.v.e .
2003 , mnras , 343 , l29 treu , t. , stiavelli , m. , casertano , s. , mller , p. & bertin 2002 , apj , 564 , l13 van dokkum , p.g . &
ellis , r.s .
2003 , apj , 592 , l53 warren , s.j . ,
hewett , p.c . ,
lewis , g.f . , mller , p. , iovino , a. , shaver , p.a . 1996 , mnras , 278 , 139 zwicky , f. 1937 , , 82 , 217 | we describe the first results from two observational projects aimed at measuring the amount and spatial distribution of dark matter in distant early - type galaxies ( e / s0s ) and clusters of galaxies . at the galaxy scale
, the lenses structure and dynamics ( lsd ) survey is gathering kinematic data for distant ( up to @xmath0 ) e / s0s that are gravitational lenses . a joint lensing and dynamical analysis
constrains the fraction of dark matter within the einstein radius , the mass - to - light ratio of the stellar component , and the total slope of the mass density profile .
these properties and their evolution with redshift are briefly discussed in terms of the formation and evolution of e / s0 galaxies and measurement of the hubble constant from gravitational time delay systems . at the cluster scale after careful removal of the stellar component with a joint lensing and dynamical analysis
systems with giant radial arcs can be used to measure precisely the inner slope of the dark matter halo .
an hst search for radial arcs and the analysis of a first sample are briefly discussed in terms of the universal dark matter halos predicted by cdm simulations .
# 1_#1 _ # 1_#1 _ = # 1 1.25 in .125 in .25 in |
recently , several restricted models of quantum computation have been investigated that seem to lie between classical and quantum , in terms of their computational complexity . while unlikely to be universal for quantum computation , these models seem to have some nontrivial computational power , as they were shown to perform tasks considered hard for classical computers . unlike factoring
, the tasks performed by these restricted models do not seem to be in np nor to be related to `` useful '' applications , making them less appealing from a practical point of view than the more standard applications of quantum computers , such as shor s factoring algorithm @xcite . on the other hand ,
it is conceivable that these models are easier to implement in experimental settings , since they do not require the full power of quantum computation , and thus provide good candidates for intermediate milestones for the field .
examples of these models include the one clean qubit model and generalizations @xcite , quantum circuits of constant depth three or greater @xcite , circuits of commuting 2-local quantum gates , also known as iqp @xcite , non - adaptive measurement - based quantum computation , known as mbcc ( measurement - based classical computation ) @xcite , and non - adaptive linear optics , also known as bosonsampling @xcite .
however , even if these restricted models propose demonstrations closer to the reach of current technologies than universal quantum computers , there are still important technological challenges to overcome before we can provide compelling evidence that quantum systems can outperform classical computers .
bosonsampling , for example , is in a sense more robust than the other models mentioned since its complexity claims hold even in an approximate setting , thus accommodating ( some level of ) experimental imperfections , and an experiment demonstrating interference of a few dozen photons in an interferometer of a few hundred modes already would provide evidence of the computational power of quantum systems long before any reasonable - sized implementation of shor s algorithm . even in this arguably more `` experimental - friendly '' scenario ,
the best experiments reported so far all lie in the limited range of 3 - 4 photons @xcite .
this motivates the study of further simplifications of these models , both from the practical point of view of bringing them even closer to experimental reality , as well as from a more conceptual point of view of understanding the origins of their complexity . in this paper ,
our main focus is on a restricted version of the bosonsampling model where the linear - optical circuits have only a constant number of layers of beam splitters , which we call constant - depth bosonsampling . using techniques similar to those used for constant - depth quantum circuits @xcite and iqp @xcite
we will prove that , if it was possible to efficiently produce a sample from the _ exact _ output distribution of a linear optical experiment of depth four ( i.e. with four sequential layers of beam splitters ) by only classical means , this would imply a large collapse in the structure of complexity classes ( namely , the polynomial hierarchy would collapse to its third level ) , which is considered a highly unlikely result .
along the way we will construct a simple example of a circuit that is both in iqp and has depth four , thus at the intersection of two of the aforementioned restricted models while this has been noticed before , we believe the presentation provided here may be of independent pedagogical interest . optimizing the bosonsampling model in terms of depth is of special interest for recent experimental implementations , as several use an integrated approach to linear optics where interferometers consist of waveguides etched into e.g. glass chips @xcite .
while having the benefit of miniaturization , allowing for implementation of networks of several dozen beam splitters within a few centimeters of material , these approaches also often suffer from certain scalability issues in particular , one can model losses by attributing a certain loss probability to each beam splitter a photon must transverse inside the chip , which leads to a decay of the signal that is exponential in the circuit depth . while further technological advances may mitigate this issue , this clearly motivates an investigation of how small the depth of an optical circuit needs to be to allow for some nontrivial complexity claim . in ref .
@xcite , the authors show than an @xmath0-photon @xmath1-mode bosonsampling device can be parallelized to a circuit of o(@xmath0 log @xmath1 ) depth even in the approximate setting .
our result shows that a depth of four is in fact already sufficient in the _ exact _ setting , and this result may be a first step in reducing the depth requirements of the original model .
this paper is organized as follows . in
we give some basic definitions and theoretical background . in
we discuss some relations between postselection and complexity classes relevant for this work , outlining the main complexity argument leading to our result . in
we give a brief overview of the klm scheme for quantum computing with linear optics . in
we give a simplified proof of the ( previously known ) fact that an efficient classical algorithm for sampling from the exact output distribution of a depth - d quantum circuit would imply the collapse of the polynomial hierarchy to its third level , for any @xmath2 . in we
show how the result of can be adapted to the setting of linear - optical circuits of depth four .
finally , contains concluding remarks and remaining open questions .
let us begin with a brief outline of the main complexity - theoretical argument behind our result .
this reasoning is a simplified version of that used , e.g. , in refs .
@xcite to prove similar results for different models .
let a be some model of probabilistic computation .
it may correspond to probabilistic classical computation ( bpp ) , quantum computation ( bqp ) , or some other restricted model such as circuits of commuting quantum gates ( iqp ) of ref .
@xcite , quantum circuits of constant depth @xcite , or linear optical circuits ( lo ) such as those used for bosonsampling @xcite . for all of these cases
, we can define a corresponding complexity class posta , or `` a with postselection '' .
informally , consider a family of circuits from the restricted model a which have two registers , the output register ( @xmath3 ) which encodes the answer to some desired problem , and the postselection register ( @xmath4 ) which flags whether the computation succeeded or not .
the class posta then corresponds to the set of problems for which there exists a family of circuits in a such that ( i ) the probability that the computation succeeds ( i.e. the probability of observing a specific predetermined outcome on register @xmath4 ) is nonzero and ( ii ) conditioned on this success , the circuit solves the desired problem with probability @xmath5 for some arbitrary positive constant @xmath6 . for our purposes ,
this informal definition will suffice , but a more formal definition of these classes can be found in refs .
@xcite .
two subtleties in the definition of posta should be pointed out .
first , we were intentionally vague about what the output and postselection registers actually consist of , since these depend on the details of model a. if a corresponds to classical ( resp .
quantum ) circuits , then @xmath3 and @xmath4 represent subsets of the output bits ( resp .
qubits ) , and the success condition corresponds to obtaining some particular bit string on @xmath4 . however , if a corresponds to linear optics , then @xmath3 and @xmath4 correspond to a subset of the output modes , and the success condition will correspond to observing some configuration of photons in @xmath4 .
the changes in the formal definitions necessary for each of these cases , however , are very straightforward .
the second subtlety is in the size of the register @xmath4 . in some cases , such as when defining postbqp ( see , e.g. , ref .
@xcite ) , @xmath4 is considered to be a single qubit without loss of generality , as it is trivial to encode , on a single qubit , whether poly@xmath7 output qubits are in a particular bit string using the operations available in bqp .
however , the same is not true for other models , such as iqp , lo or constant - depth quantum circuits , due to the limited set of operations allowed , and there we must consider the most general case where @xmath4 is a polynomial - sized register .
in fact , for these three cases it is possible to show that the probabilities associated with logarithmic - sized subsets of the output register can be computed efficiently classically @xcite , and only by looking at sufficiently large outputs can we obtain the models `` full '' computational power .
now we invoke a set of well - known results from complexity theory which ultimately provide evidence that , even with the unrealistic power of postselection , classical computers still seem to be much less powerful than quantum computers ( for formal definitions of the complexity classes mentioned here , see refs .
first , it is known that p with a postbpp oracle is contained within the third level of the polynomial hierarchy @xcite , which is a recursively - defined tower of complexity classes strongly conjectured to be infinite .
second , a result by aaronson @xcite shows that postbqp is in fact equal to a complexity class known as pp . finally , a theorem by toda shows that p with a pp oracle contains the whole polynomial hierarchy @xcite .
concatenating these results one can see that , if postbpp were equal to postbqp , the polynomial hierarchy would collapse to its third level , a surprising result that would contradict widely believed conjectures .
suppose now that our restricted model a , mentioned above , is such that posta = postbqp .
that is , suppose that a with postselection is as strong as bqp ( with postselection ) , even if a is not expected to be equal to bqp itself .
then the previous paragraph proves that an efficient classical weak simulation of a ( by which we mean the ability to sample exactly from the output distribution produced by a device from a with only a classical computer and polynomial resources @xcite ) is unlikely to be possible . to see this , note that if we could program a classical computer to efficiently sample from the output distribution of a , then this would mean that posta @xmath8 postbpp , since we could just apply the same postselection on the output of the classical simulation that would be applied on the output of a. but , as argued before , this would lead to postbpp = postbqp and the collapse of the polynomial hierarchy .
this relation , posta = postbqp , is in fact obtained for constant - depth circuits of depth greater than three @xcite , for iqp @xcite , for mbcc @xcite and for lo @xcite ( as we will review on the next section ) , and is the crucial reason why all of these restricted models are expected to be hard to simulate classically . in this work
, we will show that the exact same relation holds for constant - depth linear optics , thus adding it to this set of restricted models .
one important aspect of this postselection argument is that it only provides evidence that it is hard to sample from the _ exact _ output distribution of the corresponding device ( or at most from a distribution close within multiplicative error to the exact one , which is a very strong requirement given that most probabilities will be exponentially small @xcite ) . in this sense , bosonsampling seems to be more robust than the other models : the main contribution of ref .
@xcite was to prove that , under some reasonable technical conjectures , this sampling should be classically hard even in an _ approximate _ sense ( more specifically , even if we change the task to sampling from any distribution close to the ideal one in total variation distance ) .
this result , however , relies on different techniques than the postselection argument described and used here . in 2001 a seminal work by knill , laflamme and milburn @xcite showed how to implement any quantum circuit in a scalable manner using only linear - optical elements and adaptive measurements .
several variants have been developed since then , with the purpose of reducing the large ( but polynomial ) overheads in resources induced by the original construction . here
we will limit our discussion to that variant which is more suitable for our purposes , but for a full review see e.g. ref .
@xcite .
the klm scheme uses what is known as dual - rail encoding , where each qubit is encoded using one photon and two modes .
each computational state corresponds to the photon being found in one of the modes , as such : @xmath9 it is easy to see that , in this encoding , any single - qubit gate can be implemented deterministically using an arbitrary two - mode transformation , as represented in .
+ to obtain a universal gate set , we also need an entangling logical two - qubit gate @xcite .
the simplest candidate is the @xmath10 gate , since it only acts nontrivially on the @xmath11 state , thereby only requiring two of the four modes involved in the two - qubit state to interact .
it suffices to implement some two - mode transformation that adds a @xmath12 phase only if both modes are occupied by a single photon , and does nothing otherwise .
this seems to require an interaction between the photons : informally , the physical transformation would affect one photon conditioned on the other photon `` being there '' .
the crucial point of the klm scheme is that this interaction can be simulated by a measurement , at least probabilistically . in
we depict an optical circuit due to knill @xcite which does precisely that . by adding two extra ancilla modes , each carrying a single photon , and upon measuring again one photon in each of these modes , as depicted in
, the circuit is successful and implements a @xmath10 gate .
this happens with probability 2/27 , and any other measurement outcomes results in a failure of the gate . if this was the only way to implement such transformations then scalable linear - optical quantum computing would be hopeless , as the overall success probability of any circuit would decrease exponentially with the number of two - qubit gates .
the original klm paper also included an error - correcting step to deal with this issue , but we do not need to address it here our interest is in the capabilities of linear optics with postselection , so we can implement any quantum circuit just by postselecting on the success of every gate and never have to worry about what happens when they fail .
replacing error - correction with postselection on the klm scheme already shows that postlo = postbqp and thus , by the discussion in the previous section , provides evidence against a possible classical simulation of linear optics even in the non - adaptive setting .
this was pointed out in ref .
@xcite , but the authors go much further .
they define a random instance of bosonsampling as * preparation of @xmath1 photonic modes , the first @xmath0 containing one photon each ( with @xmath1 suitably larger than @xmath0 ) ; * evolution of these modes according to an @xmath13 random interferometer @xmath14 sampled from the uniform distribution ; and * measurement of the output in the coincidence basis ( i.e. only considering outputs with a single photon per mode ) . by relating the outcome probabilities to the permanents of certain sub - matrices of @xmath14 , together with very plausible conjectures about the hardness of computing the permanent for this particular ensemble of random matrices , the authors are able to provide the more robust result mentioned in the previous section without resorting to the postselection argument , as well as provide a worst - case / average - case equivalence conjecture for random bosonsampling instances . in the next section
we will consider a variation of bosonsampling where condition ( ii ) is replaced by * evolution of these modes according to an @xmath13 interferometer composed of four layers of arbitrary two - mode transformations ; remark that , in our model , the interferometers are not random , but rather they are constructed in a prescribed manner .
also , from this point on we will only consider linear optical circuits where the inputs consist of each photon in a different mode [ i.e. condition ( i ) ] .
this is justified by the fact that both the klm scheme and the original bosonsampling model satisfy this condition , but also because initializing several of the @xmath0 photons in the same mode tends to make the simulation easier @xcite .
let us begin with an alternative proof of the results found in ref . @xcite . in that paper , the authors show that quantum circuits with depth three can simulate any arbitrary quantum circuit if postselection is allowed from the discussion of this immediately means that an efficient classical weak simulation of such circuits would imply a collapse of the polynomial hierarchy .
they also show that this depth is the smallest for which this holds by giving an explicit simulation for circuits of depth @xmath15 . for the present purposes ,
the depth of a quantum circuit is defined as the number of layers of arbitrary two - qubit gates ( the authors in ref .
@xcite use a different convention , including the final round of measurements in the depth count , but this only changes the definition of depth by one ) .
we consider that gates acting on qubit pairs that share a common qubit ( e.g. , @xmath16 and @xmath17 ) can not be done simultaneously even if they commute .
single - qubit gates do not contribute to the depth count since we can consider them as part of the nearest two - qubit gate which means , in particular , that we can choose initialization and measurements in any single - qubit basis without loss of generality .
our starting point is an universal construction in the measurement - based model of quantum computing ( mbqc ) @xcite .
we will restrict ourselves to a description of how the computation is performed , with no discussion on how its universality is proved that would lie beyond the scope of this work , and we refer to @xcite and references therein .
let @xmath18 be the family of brickwork graphs , such as in , parameterized by some size parameter @xmath0 ( i.e. the graph has poly@xmath7 vertices ) .
consider , then , the corresponding graph state @xmath19 , built as follows : ( i ) for each vertex in @xmath18 prepare a qubit in the @xmath20 state and ( ii ) for each edge in @xmath18 apply a @xmath10 gate between the two corresponding qubits .
this generates a highly - entangled multi - qubit state , and the computation then proceeds by a sequence of single - qubit measurements on the qubits of @xmath19 . each measurement is done in one of a discrete set of bases , and their outcomes determine the bases of future measurements .
the complete set of measurements , including the order in which they are performed and the dependence of some measurements on the results of others , is known as a _ measurement pattern_. we now restate the following theorem , which is taken from @xcite ( we follow the result of @xcite for convenience , but the brickwork state was known to be universal for mbqc before that , see e.g. @xcite ) : [ thm : mbqc ] let @xmath18 be the brickwork graph shown in .
the corresponding graph state is universal for measurement - based quantum computation .
the computation is performed as follows : * initialize a qubit in the @xmath21 state for every vertex of @xmath18 ; * apply @xmath10 gates according to the edges of @xmath18 ; * sequentially measure each qubit of @xmath19 in one the bases @xmath22 any poly - sized quantum circuit on @xmath0 qubits can be simulated in this way using some graph @xmath18 of poly@xmath7 size , and by a suitable choice of the measurement pattern which , furthermore , can be computed efficiently classically . a formal proof of this result , found in ref .
@xcite and references therein , gives explicit measurement patterns corresponding to gates in the universal set @xmath23 , and thus an explicit procedure to simulate any quantum computation .
omitting further details from this proof , we would just like to point out that this protocol , described as a quantum circuit ( cf . ) , inherits a temporal structure almost exclusively from the `` classical '' adaptation of measurements .
in other words , preparation of the graph state only takes a few rounds of two - qubit gates , and the depth of the circuit stems from the fact that some qubits _ must _ be measured prior to others .
now consider what happens if we replace adaptive measurements by postselection .
that is , rather than making a measurement and conditioning future measurements on its outcome , we just _ postselect _ each measurement to a given outcome such that adaptation is unnecessary . as an example , suppose we measure one qubit in basis @xmath24 , with two possible outcomes labeled @xmath25 and @xmath26 , and the measurement pattern dictates that a second qubit must be measured either on basis @xmath27 or basis @xmath28 depending on the outcome of @xmath24 . this can be replaced simply by postselecting @xmath24 to the outcome @xmath25 and simultaneously measuring the second qubit in basis @xmath27 .
doing this for every measurement effectively flattens out the temporal structure of the protocol , allowing us to perform all measurements in a single round
if this sounds too good to be true , recall that postselection is indeed an extremely unrealistic `` power '' .
our goal is now basically reached , as the procedure described already constitutes a quantum circuit of depth three . to see this ,
consider the circuit description of the mbqc protocol , as shown in . in this circuit
, the depth count goes as follows : ( i ) a first round of @xmath10 gates ( green edges of ) ; ( ii ) a second round of @xmath10 gates ( red ) ; ( iii ) a third round of @xmath10 gates ( blue ) , followed by a round of single - qubit gates to prepare the measurement bases ( recall that neither state preparation in the @xmath20 state nor preparation for single - qubit measurements count for the depth ) .
only three rounds of @xmath10 gates suffice because vertices of the brickwork state have degree at most three , as shown in
. it should be clear by our arguments and that circuits of this form , when imbued with the power of postselection , can implement any computation in postbqp .
we then conclude , by the previous discussion , that an efficient classical simulation of its output would imply collapse of the polynomial hierarchy .
it is interesting that the mbqc approach is very well suited for this proof , since almost all the temporal structure lies in the adaptive measurements , which is precisely what we replace by postselection .
curiously , all information about the computation _ also _ lies in the measurements , since the brickwork state does not depend at all on the underlying quantum computation ( except , of course , in its size ) .
thus , it seems that all `` computational power '' ( to abuse the terminology ) of the constant - depth quantum circuit resides on the combinations of possible choices of measurement bases .
an interesting open question is whether there is a natural way to randomly choose the measurement bases such that the final simulation is a particularly hard instance , similar to the worst - case / average - case equivalence conjectured for approximate bosonsampling @xcite .
another curious aspect of this proof is that it encompasses several similar proofs for different models .
the resulting circuit obtained by flattening out the adaptivity of mbqc has depth three , is an mbcc protocol ( by definition ) , and is in iqp , since it only uses gates diagonal in the @xmath29 basis .
this is interesting , but maybe not specially surprising the concept of gate teleportation is present , in one form or another , in all of these results , and is closely related to the historical origin of mbqc .
we should also point out that , while this proof unifies several results , it does not provide concrete relations between these models the resulting circuit lies at their intersection , but each model may have circuits that perform tasks outside of this intersection . besides proving that a depth of three is sufficient for the hardness result , the authors in ref .
@xcite also prove that it is necessary .
this is done by giving an explicit simulation of any depth-2 circuit , which we briefly reproduce here .
consider an arbitrary depth-2 circuit .
that is , the qubits are initialized in the computational basis , undergo a first round of arbitrary two - qubit gates , followed by a second round of arbitrary two - qubit gates , and a final round of measurements .
the simulation becomes straightforward if we reinterpret this as a two - round computation : first the qubits are prepared in arbitrary two - qubit states , and then they are measured in arbitrary two - qubit bases .
the classical simulation then proceeds as follows : * choose one particular measurement @xmath30 .
this is a two - qubit measurement where each qubit may be entangled with some other qubit .
thus , its outcome probabilities only depend on a four - qubit state and are trivially easy to compute in constant time . *
compute the outcome probabilities and simulate @xmath30 by classically sampling a two - bit string from the corresponding distribution and fixing the output of the measurement accordingly .
* project the two measured qubits onto the sampled bit string , and update the description of the two qubits entangled to them accordingly . these two qubits are now in an arbitrary , but known , two - qubit state . * the next measurement involving one of the qubits updated in ( iii ) is again a two - qubit measurement that depends only on an arbitrary four - qubit state . * repeat this procedure until all measurements have been fixed .
this simulation samples from the same distribution as the corresponding quantum circuit .
it is also an example of weak simulation since at no point we compute the probabilities associated with the full output of the circuit , only small two - bit strings at a time . as a consistency check
, one could attempt to generalize this simulation for circuits of depth three .
this would naturally fail , because the first measurement in step ( i ) would depend on an 8-qubit state , hence step ( iii ) would collapse the unmeasured qubits to an arbitrary 6-qubit state .
but then , the next time we returned to step ( i ) the measurement might depend on the state of 14 qubits and so on , and the iteration is clearly disrupted .
we now combine the previous results into one the computational complexity of exactly simulating constant - depth bosonsampling .
the first step is a rather straightforward concatenation of the results mentioned so far .
consider a constant - depth circuit obtained from some universal mbqc protocol by replacing adaptive measurements with postselection , as in .
now map the resulting quantum circuit to a linear - optical circuit using the constructions of . finally , replace the adaptive measurements of the klm protocol by postselection .
the resulting circuit is naturally as strong as postbqp , and thus it can not have an efficient classical simulation unless the polynomial hierarchy collapses .
let us now count the depth of the resulting optical circuit .
the single - qubit gates can not be `` absorbed '' into the two - qubit gates anymore , as was done in , since in the linear - optical setting we have a dual - rail encoding where both single- and two - qubit gates correspond to two - mode transformations ( i.e. the single qubit gates involve the two modes of a single qubit , while two - qubit gates involve two modes of different qubits ) .
we will , however , absorb phase shifters into the closest beam splitter , and count layers of arbitrary two - mode transformations .
the total count then goes as follows : one layer of balanced beam splitters for the preparation of the @xmath31 states , followed by six layers for the preparation of the entangled state ( i.e. two for each layer of @xmath10 gates of the brickwork state , using the @xmath10 of ) . finally , one last layer of beam splitters and phase shifters for the preparation of measurement bases , and a round of measurements .
this amounts to a total depth of eight .
we can now use two tricks to reduce this depth count .
first note that , according to , when we perform a @xmath10 gate , only one mode of each encoded qubit is involved , while the other remains idle . by a simple adaptation of the circuit of , we can use the second mode of each qubit to simultaneously implement the second round of @xmath10s . to do this , we include two @xmath32 phase shifters before the circuit of , resulting in a gate which , up to a global phase , only adds a minus sign to the photonic @xmath33 state .
it is clear that , if we act with this gate on the modes encoding the @xmath34 state of two qubits , we obtain precisely a logical @xmath10 gate . by a clever alternation of which mode we use for each @xmath10 gate
, we can build the long paths of the brickwork state ( i.e. the red and green edges in ) using one round of @xmath10s , corresponding to two rounds of beam splitters .
this reduces the overall depth to six .
the second trick is to use a procedure similar to the usual gate teleportation @xcite , which can be found in ref .
@xcite , to parallelize the third layer of @xmath10s .
say we want to implement some faulty ( i.e. probabilistic ) gate @xmath14 on a computational qubit , but without endangering the information stored in that qubit .
we can then do the following : we prepare an auxiliary two - qubit state @xmath35 , implement @xmath14 on the second qubit of this auxiliary state , and , if @xmath14 succeeds , project the first qubit of the auxiliary state together with our original qubit on the @xmath36 state ( see ) . of course
, in standard quantum computation ( i.e. bqp ) we can not guarantee that a two - qubit measurement will project our qubits on the @xmath36 state , but for our purposes we can just postselect on observing this outcome .
now notice that , if rather than being a two - qubit state , @xmath37 actually represents a state of one photon in two modes , the mathematical structure is exactly the same , and so it allows us to use an equivalent scheme to teleport the states of the _ modes _ in the klm scheme .
since each @xmath10 gate only involves one mode of each qubit , as in , this allows us to implement the gate teleportation by teleporting only one mode , rather than a complete qubit , as shown in .
thus , we can perform all the @xmath10 gates for the brickwork state in parallel , and just teleport the computational states around , in a very similar spirit to the original result of @xcite .
this scheme , and how it can used to reduce the depth , is illustrated in .
this reduces the total depth count to four .
+ note that , of the two tricks mentioned , the second would already suffice to reduce the depth count to four .
we presented the first as it may be of independent interest for the parallelization of linear - optical circuits even in future implementations of the full klm scheme .
also , since in this setting we do not have to worry about the gates failing , it would be wasteful to do all the @xmath10 s in fresh ancilla modes rather than directly on the qubit themselves whenever possible .
we can now also ask whether , besides being sufficient , a depth of four is also necessary for the preceding result .
alas , unfortunately so far the answer seems to be no .
a direct adaptation of the simulation of only proves that depth-2 linear optical circuits are classically simulable .
our simulation follows the same basic principle as that described in , namely that we can re - interpret a depth-2 linear optical circuit as a two - mode preparation followed by a two - mode measurement .
the main difference between the simulation of and the analogous for linear optics is the following : assuming that each mode is initialized with either zero or one photons ( i.e. @xmath38 and @xmath39 ) , the state after the first layer of beam splitters can have some modes occupied by two photons ( i.e. @xmath40 ) .
thus one could argue that , when we do the iterative step of the simulation , the sample space changes from one step to the next , possibly disrupting the iteration .
but , if we observe that no mode can ever have more than four photons , this can easily be accommodated into the simulation scheme as follows : * choose a measurement @xmath30 .
this is a two - mode measurement , where each mode has up to two photons and may be `` entangled '' with some other mode ( by which we mean `` mode entanglement '' , where two modes are written in a nonseparable state such as @xmath41 , not to be confused with `` real entanglement '' between two photons due to some interaction ) .
thus , this measurement only depends on the states of four modes , each with up to two photons , and its probabilities can be computed in constant time . *
compute the outcome probabilities and simulate @xmath30 by sampling from the corresponding distribution and fixing the output of the measurement accordingly .
note that the space of outcomes can have from zero to four photons distributed in any way among the two measured modes .
* project the two measured modes onto the sampled state , and update the description of the two unmeasured modes accordingly .
they are now in an arbitrary , but known , two - mode state . by photon - number preservation ,
the outcome fixed in ( ii ) determines the number of photons in the collapsed state , but it is easy to check that it can have at most two photons per mode . * the next measurement involving one of the modes collapsed in ( iii ) is again a two - mode measurement that depends only on a four - mode state .
one caveat is that these four modes may now contain up to eight photons overall ( up to two per mode ) depending on previously fixed outcomes .
nonetheless , any subsequent two - mode measurement can only detect four of these photons , so step ( ii ) can always be applied . *
iterate this procedure until all measurements have been fixed .
as should be clear , since we have only two layers of beam splitters , no mode can ever have more than two photons , except immediately prior to a measurement , in which it can have up to four photons .
this guarantees that , even though the first measurement differ from subsequent measurements in the set of allowed states , this does not disrupt the iterative procedure .
also note that this simulation seems to break down for more layers .
if the optical circuit has depth three , say , then a first measurement could depend on the state of eight modes , and collapse the unmeasured modes into an arbitrary state of six modes .
but then the next measurement involving these modes could depend on 14 modes , and so forth .
since each new choice of measurement may depend on a larger number of modes , the iteration step fails .
in this paper , we showed that the exact version of the bosonsampling result in section 4 of ref .
@xcite already holds if the linear optical circuit has depth four , and also that it can not hold for depth two or lower .
this seems to be a consequence of the fact that a probabilistic linear - optical @xmath10 gate must be constructed with at least two layers of beam splitters .
this is natural , since it is known that ancilla modes and photons are necessary @xcite for the construction of the gate ( even probabilistically ) , and so a single layer of beam splitters would clearly not suffice to interact both the two computational modes and the ancilla modes in question .
it can not be ruled out , of course , that an ingenious use of postselection might provide a construction for a hard - to - simulate depth-3 exact bosonsampling instance via completely different means . besides the corresponding result for depth three
, we leave another open question : is it possible to somehow close the depth gap between this result and the main result of ref .
@xcite for approximate bosonsampling ? in the latter case ,
the minimum necessary depth is o(@xmath0 log @xmath1 ) , for @xmath0 photons in @xmath1 modes , as shown in ref .
@xcite , and for exact bosonsampling the minimal known depth , as we showed , is four .
there are two natural routes to close this gap : either prove that the approximate bosonsampling result can be simplified , using a further parallelized scheme or possibly a matrix ensemble other than the uniform one , or give an efficient classical algorithm for approximate constant - depth bosonsampling ( thus showing that constant - depth and regular bosonsampling are indeed fundamentally different ) .
the latter would not be too surprising : there exist problems for which the exact solution is hard ( e.g. in @xmath42p ) , while an approximate solution is easy ( i.e. in p ) , such as the permanent of positive matrices @xcite
. another possibility would be to explicitly investigate the permanents of constant - depth interferometers .
however , the unitary of a depth-4 interferometer is at most @xmath43-sparse , simply because a photon that enters one mode can not leave in more than @xmath43 different modes , and similarly a photon that exits in a particular mode can not have entered in more than @xmath43 different input modes .
thus , these matrices obey a very special structure , very different from the uniform ensemble , and a worst - case / average - case equivalence allowing us to condition the hardness of their permanents on some plausible conjecture seems less likely .
an approximate result for constant - depth bosonsampling would also have important consequences for the corresponding complexity classes . for example , it can contribute to the more general program of relating the different restricted models described so far .
it is known that an arbitrary @xmath0-photon , @xmath1-mode linear - optical circuit can be simulated in bqp simply by treating each mode as an @xmath0-level system , and using a standard mapping from @xmath1 qudits to o(@xmath1 log @xmath0 ) qubits @xcite . as discussed in ref .
@xcite , this mapping induces an o(log @xmath0 ) depth increase , simply because an arbitrary single - qudit gate will map to an arbitrary @xmath44log @xmath45-qubit circuit .
this logarithmic depth increase suggests that arbitrary linear - optics might not be simulable by constant - depth quantum circuits .
however , the same is not true for constant - depth linear optics . in an optical circuit consisting of a no - collision fock state input
( e.g. @xmath46 ) followed by a constant number of beam splitter layers , it is easy to see that no mode can have more than a constant number of photons .
for example , if the optical circuit has depth four , no mode ever has more than 16 photons , which means that this system can be viewed simply as a collection of @xmath1 16-level systems .
consequentially , simulating it on a quantum computer does not induce a logarithmic depth increase and it can , in fact , be simulated in constant depth .
so we see that constant - depth linear optics is provably contained within constant - depth quantum computing ( although the second constant might not be three , as in ) , and so showing that a classical simulation of the former is unlikely even in the approximate scenario , along the lines of @xcite , might suggest an equivalent result for the latter .
this would be the first evidence that other recent `` collapse - the - polynomial - hierarchy '' results might in fact be as robust as bosonsampling .
finally , we would like to discuss possible implications of this result for current experimental efforts .
to our knowledge , all bosonsampling experiments so far have been performed using integrated photonic circuits @xcite . in these devices ,
the depth of the circuit is one of the leading causes of photon loss and exponential attenuation of the signal , so any optimization of the model in terms of depth should be beneficial for the experimental efforts .
one drawback of the proof given here , however , is that it requires beam splitters acting between arbitrarily distant modes , a requirement that is not as well - suited for usual two - dimensional layouts of these devices .
we believe our result may pave the way for practical simplifications of current experimental efforts , but we leave for future work to determine the exact trade - off between losses induced by the depth count and losses induced by having to engineer crossovers between the waveguides .
the author would like to thank andrew childs and ernesto galvo for helpful and insightful discussions .
this research was supported in part by brazilian funding agency cnpq ( conselho nacional de desenvolvimento cientfico e tecnolgico ) , and in part by the perimeter institute for theoretical physics .
research at perimeter institute is supported by the government of canada through industry canada and by the province of ontario through the ministry of research and innovation .
28 ifxundefined [ 1 ] ifx#1 ifnum [ 1 ] # 1firstoftwo secondoftwo ifx [ 1 ] # 1firstoftwo secondoftwo `` `` # 1''''@noop [ 0]secondoftwosanitize@url [ 0 ] + 12$12 & 12#1212_12%12@startlink[1]@endlink[0]@bib@innerbibempty @noop * * , ( ) link:\doibase 10.1103/physrevlett.81.5672 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.112.130502 [ * * , ( ) ] @noop `` , '' ( ) , @noop * * , ( ) link:\doibase 10.1098/rspa.2010.0301 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.112.140505 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1126/science.1231440 [ * * , ( ) ] link:\doibase 10.1038/nphoton.2013.112 [ * * , ( ) ] link:\doibase 10.1126/science.1231692 [ * * , ( ) ] link:\doibase 10.1038/nphoton.2013.102 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.111.130503 [ * * , ( ) ] link:\doibase 10.1038/nphoton.2014.135 [ * * , ( ) ] link:\doibase 10.1038/nphoton.2014.152 [ * * , ( ) ] link:\doibase 10.1098/rspa.2005.1546 [ * * , ( ) ] link:\doibase 10.1137/s0097539792240467 [ * * , ( ) ] link:\doibase 10.1137/0220053 [ * * , ( ) ] link:\doibase 10.1038/35051009 [ * * , ( ) ] link:\doibase 10.1103/revmodphys.79.135 [ * * , ( ) ] @noop _ _ ( , ) link:\doibase 10.1103/physreva.66.052306 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.86.5188 [ * * , ( ) ] link:\doibase 10.1146/annurev - conmatphys-020911 - 125041 [ * * , ( ) ] in link:\doibase 10.1109/focs.2009.36 [ _ _ ] ( ) pp .
link:\doibase 10.1103/physreva.71.032318 [ * * , ( ) ] link:\doibase 10.1038/46503 [ * * , ( ) ] link:\doibase 10.1145/1008731.1008738 [ * * , ( ) ] | bosonsampling is a restricted model of quantum computation proposed recently , where a non - adaptive linear - optical network is used to solve a sampling problem that seems to be hard for classical computers .
here we show that , even if the linear - optical network has a constant number ( greater than four ) of beam splitter layers , the exact version of the bosonsampling problem is still classically hard , unless the polynomial hierarchy collapses to its third level .
this is based on similar result known for constant - depth quantum circuits and circuits of 2-local commuting gates ( iqp ) . |
the statistical properties of quantum system attract considerable attention in the broad field of modern physics ranging from nuclear , atomic and molecular physics to condensed matter and quantum computing . of a special interest
is the properties specifying the quantum chaos border for a system of many - body quantum states .
extensive investigations of many - body interacting systems such as nuclei , many - electron atoms , quantum dots , quantum spin glasses and quantum computer models @xcite have shown that the border is associated with a crossover of level spacing statistics from the poisson distribution to wigner - dyson one and with a transition to ergodic distribution for the eigenstate wavefunction over the large number of basis levels as in random matrix theory ( rmt ) @xcite . in a sense
the interaction leads to dynamic thermalization without coupling to an external thermal bath .
according to @xcite this crossover takes place when the coupling matrix elements of the hamiltonian become comparable to the energy spacing between directly coupled states .
we address this problem to a generic system of coupled quantum states , the hamiltonian matrix @xmath3 of that includes the leading diagonal with disordered random values and the random off - diagonal elements inside the band of size @xmath4 .
this band random matrix with the disordered diagonal ( brmdd ) is a reasonable model describing the systems with strong imperfection in the basis state energies .
models based on brmdd were applied to study the electron transport problem @xcite and the problem of interacting particles in a random potential @xcite .
the results obtained for these models are of an obvious interest also in analyzing such few - freedoms physical objects as the vibrational quasicontinuum of polyatomic molecule @xcite and a quantum computer @xcite .
the statistical properties of brmdd - based systems were studied in some details @xcite .
the shape , the localization length and the inverse participation ratio ( ipr ) for eigenfunction have been investigated with the help of numerical simulations @xcite and the supersymmetry approach @xcite .
these investigations have exhibited the lorentzian shape for local spectral density ( lsd ) in circumstances when a non - perturbative localization regime is realized .
the state - state interaction strength , at which the eigenstates are extended over the whole matrix size @xmath5 and the eigenenergy level spacing statistics has the wigner - dyson form , have been revealed for brmdd with sufficiently large band ( when @xmath6 ) @xcite .
unfortunately the ergodic properties of lsd have not been investigated in details .
an open question is also the quantum chaos border for brmdd with arbitrary small band . in the paper we study localization and ergodicity properties of lsd to be obtained from brmdd with arbitrary small band .
the lsd was introduced in 1955 by wigner @xcite and successfully employed in rmt to describe statistically the localization effects for complex quantum systems @xcite ( including the systems , represented by band random matrices with the reordered leading diagonal @xcite ) .
this quantity is the fourier transform for the correlation function @xmath7 and specifies spreading of the energy , initially concentrated in a specific unperturbed basis state @xmath8 , between the eigenstates due to state - state interaction .
generally the lsd may be characterized by the width @xmath9 measuring the energy scale , in that the individual state @xmath10 is localized . the number of eigenstates populating this scale is given in terms of the product @xmath11 designated here as _
an ergodic localization length _ of lsd ( @xmath12 is the eigenstate density ) .
hence the quantity @xmath13 specifies the greatest possible number of eigenstates , where the basis state @xmath14 can be effectively admixed .
the ergodic properties for the system can be identified with the structure of lsd .
the non - ergodic lsd is a strongly fluctuating spiked function and the ipr @xmath15 , which gives the actual number of eigenstates involving the state @xmath8 , is low in comparison with @xmath13 . in the ergodicity case
the lsd is monotonic and the number @xmath15 approaches the value of @xmath13 .
the main attention of our study is paid to the perturbative and localized regimes for lsd .
therefore we restrict the study to a moderate strength of state - state interaction .
this restriction implies that the width @xmath9 is essentially small as compared to the energy scale bounding the location of the eigenstate levels ( @xmath16 ) .
the finite - size effects are ignored .
hence the quantities @xmath13 and @xmath15 can be considered as functions of the interaction strength and the relative width of band .
we investigate the lsd shape and determine the ergodic localization length .
then we analyze the behavior of ipr for ergodic and non - ergodic lsd . from this analysis
we obtain criterions specifying the transition to localized and ergodic regimes for lsd .
we consider a brmdd , that represents the hamiltonian matrix @xmath3 in the basis of unperturbed states @xmath17 ( @xmath18 ) for isolated system of @xmath19 quantum states.the brmdd is a real symmetric matrix with statistically independent random elements @xmath20 where off - diagonal elements @xmath21 specify the state - state interaction .
the values of @xmath22 are distributed uniformly in the interval @xmath23 $ ] with @xmath24 and @xmath25 if @xmath26 or are zero otherwise .
the diagonal elements @xmath27 corresponding to energy levels for the states @xmath17 are uniformly distributed according to the poisson statistics with the mean spacing @xmath28 between adjacent energy levels : @xmath29 .
the energy level @xmath30 for a probing state @xmath8 is located in the midpoint of interval @xmath31 $ ] : @xmath32 .
the perturbation of an individual basis state depends on the relative level position for unperturbed states , those are coupled directly with the state by interaction .
therefore the strength of state - state interaction for brmdd can be specified in terms of the ratio @xmath0 , where @xmath33 is the energy spacing between directly coupled unperturbed states .
here the relative band width @xmath34 gives the relative number of states coupled directly with the individual state .
depending on the relation between @xmath0 and @xmath4 , three important regimes can be distinguished @xcite . when the coupling matrix elements are weak and @xmath35 the perturbative regime is realized . at the regime
the hamiltonian ( [ hamilt ] ) can be treated within the framework of perturbation theory .
the condition @xmath36 features the regime , at that the eigenstates are localized .
the strong interaction at @xmath37 is responsible for the delocalized regime for eigenfunction . in the regime
the eigenstates are extended over the whole matrix size @xmath5 .
the eigenfunctions @xmath38 and the eigenenergies @xmath39 are obtained from diagonalization of ( [ hamilt ] ) .
the quantity @xmath40 gives a probability to find the probing state @xmath8 in the eigenstate @xmath38 .
the numerical simulation is performed for the wide range of matrix parameters : @xmath41 , @xmath42 and @xmath43 .
the number of disorder realizations lies in the range from 100 to 1000 .
the shape of lsd is studied for the eigenenergy scale , which is no larger than @xmath44 . for this scale
the interaction - induced variations in level density are negligible and the eigenenergies are homogeneously distributed with the density @xmath45 . then the lsd can be defined as @xmath46 where @xmath47 means the averaging over disorder ( that is , over many random matrices ) .
the numerical investigation confirms that the lsd is described by the well - known breit - wigner distribution of the lorentzian shape @xmath48 ( see the inset in fig.[figl(q ) ] ) with the width @xmath9 .
this shape is observed for any @xmath1 , @xmath2 , @xmath4 and @xmath5 satisfying @xmath49 ( including the parameter region where @xmath50 ) .
we find the ergodic localization length @xmath13 from ( [ erglength ] ) where the width @xmath51 is estimated by fitting of the lorentzian ( [ bwdistr ] ) to an averaged lsd .
our analysis of numerical data obtained for different values of @xmath1 , @xmath2 and @xmath52 shows that the quantity @xmath13 can be considered as a function of the single parameter @xmath53 which defines essentially the localization of lsd in the eigenenergy scale . when the coupling of the basis states by perturbation is sufficiently strong and @xmath54 the quantity @xmath13 satisfies the fermi golden rule : @xmath55 . at low values of @xmath56 ( @xmath57 )
the length @xmath13 can be described by a linear function of @xmath56 : @xmath58 .
the width @xmath9 is then much smaller as compared to the spacing @xmath28 .
fig.[figl(q ) ] demonstrates the behavior of @xmath13 in the wide range of @xmath56 at three fixed magnitudes of @xmath52 .
[ h ! ] = 8 cm the linear dependence of @xmath13 and @xmath9 on the strength of state - state coupling for small @xmath56 can be explained in terms of a simple perturbation approach to the eigenenergy levels @xmath59 located in the vicinity of @xmath30 : @xmath60 . in the approach
the eigenstate @xmath38 is considered as a superposition of the nearest basis states @xmath10 and @xmath61 .
the probability @xmath62 to find the basis state @xmath8 in the eigenstate @xmath63 can be then estimated from a two - state ( the states @xmath64 and @xmath61 ) approximation . according to this approximation
the dependence of @xmath62 on @xmath59 for an individual disorder realization is lorentzian - like @xmath65 with the width of @xmath66 .
if the states @xmath10 and @xmath61 are not coupled by interaction the dependence ( [ wn0 ] ) is considered as a function , the width of that is negligible in comparison to @xmath1 .
the probability that the states @xmath10 and @xmath61 are coupled by interaction is equal to @xmath52 . as a result
the shape of disorder - averaged lsd is described by the lorentzian - like contour with the width @xmath67 .
it seems that this perturbation approach can be applied also for the eigenstates with @xmath68 , when the probability @xmath62 is essentially small as compared to the contributions of others basis states . in this limit
( [ wn0 ] ) gives @xmath69 , as predicted in the framework of a standard perturbation approach for non - degenerate states ( see , for instance , in @xcite ) . in order to estimate the localization properties of lsd at intermediate values of @xmath56 we approximate the behavior of ergodic length for the wide range of @xmath56 ( @xmath70 )
the analysis of numerical data shows that the length @xmath13 can be well approximated by @xmath71 with the fitted coefficients @xmath72 and @xmath73 . according to ( [ l(q ) ] ) an accurate boundary between the linear and quadratic dependencies of @xmath13 on @xmath56
is obtained at @xmath74 where @xmath75 . notice that this boundary separates the ranges where the value of @xmath13 is low or high as compared to @xmath76 .
the ipr @xmath77 is associated with the parameters @xmath56 and @xmath52 , which specify correspondingly the ergodic localization length @xmath13 for lsd and the band width for brmdd .
this association depends on whether or not the lsd is ergodic . in the ergodic case the probing state is monotonically spread over eigenstates in the energy scale @xmath9 .
then the ipr is a function of the only parameter @xmath56 and approaches the ergodic length @xmath13 .
the ergodic lsd can be realized from brmdd with large band .
our analysis of calculated data for the matrices with @xmath78 shows that the best approximation for ergodic ipr is @xmath79 the fitted coefficients @xmath80 and @xmath81 are obtained for full hamiltonian matrices ( [ hamilt ] ) with @xmath82 where all @xmath83 basis states are directly linked with the state @xmath8 .
notice that the approximation ( [ iprerg ] ) gives the maximum allowable magnitude of ipr for brmdd at fixed @xmath56 .
an essentially non - ergodic lsd can be obtained from brmdd with @xmath84 .
then the ipr is significantly low in comparison to @xmath13 and should be considered as a function depending both on the parameter @xmath56 and on @xmath52 .
the analysis of numerical data for matrices with low magnitudes of @xmath56 and @xmath52 shows that the lack of ergodicity for lsd can be clearly identified with an exponential increase in ipr with the quantity @xmath85@xmath86 at fixed @xmath52 the ipr is shown in fig.[figipronq ] to increase exponentially with the rate @xmath87 ( @xmath88 ) as the parameter @xmath56 rises if @xmath89 .
the increase is observed up to the point where @xmath90 .
then the ipr approaches asymptotically the length @xmath13 from below with increasing @xmath56 . at high magnitudes of @xmath56 and @xmath52 , when the exponential ( [ iprnonerg ] ) gives higher values of @xmath15 in comparison to @xmath13 ,
the lsd is ergodic and the ipr can be determined from ( [ iprerg ] ) .
[ h ! ] = 8 cm [ h ! ] = 7.5 cm for the broad range of @xmath56 and @xmath52 satisfying to @xmath91 the mean square fitting gives the coefficient @xmath92 .
fig.[figipronqb ] demonstrates a good accordance between the calculated and fitted values of ipr for non - ergodic lsd .
unfortunately we can give no heuristic explanation of the exponential rise in ipr with @xmath93 .
notice only that this rise is observed under conditions when the eigenstates are essentially localized and @xmath94 ( @xmath95 is the localization length for eigenstate ) .
we associate the localized regime for lsd with a high magnitude of ipr . for brmdd with arbitrary small band
the requirement @xmath96 is seen from ( [ iprerg ] ) and ( [ iprnonerg ] ) to result in @xmath97 no matter what the ergodic properties of lsd and the magnitude of @xmath52 . the parameters @xmath98 and @xmath99 defines then a localization border on the plane ( @xmath100 ) , that is a boundary between regions where the perturbative ( @xmath101 ) or localized ( @xmath96 ) regimes are realized .
notice that as for lsd , the criterion ( [ qcborder ] ) specifies a transition to the localized regime for eigenstate @xcite .
the attainment of ergodic lsd for brmdd is defined by the relative width of band . for the matrices with @xmath78
the localized lsd is nearly ergodic and the ergodicity criterion can be represented by the condition ( [ qcborder ] ) .
this condition is in good agreement with a quantum chaos border for many - body systems and gives the crossover of level spacing statistics from poisson to wigner - dyson distribution for a nucleus model ( @xmath102 @xcite ) and for a generic model of quantum computer ( @xmath103 @xcite ) . in ( [ qcborder ] )
the factor @xmath104 is assumed to be equal to @xmath105 . for brmdd with @xmath84 we should take into account the exponential growth ( [ iprnonerg ] ) of ipr with @xmath0 .
the criterion of lsd ergodicity can be determined then from the requirement @xmath106 , that results in : @xmath107 where the values of @xmath108 and @xmath109 specify an ergodicity border on the plane ( @xmath100 ) .
notice that the condition ( [ qeborder ] ) is a more severe restriction imposed on the parameters @xmath56 and @xmath52 than the criterion ( [ qcborder ] ) .
our study reveals some distinctive features in behavior of lsd for brmdd with arbitrary small band .
these features are outlined on the plane ( @xmath110 ) in fig.[figregimes ] to depend on the relation between @xmath111 and @xmath52 . as for eigenstate
, the perturbative regime for lsd can be associated with the requirement @xmath112 .
in contrast to eigenstate , the localization properties of lsd is essentially defined by the magnitude of @xmath56 . in the limit @xmath113 ( that is , @xmath114 )
the ergodic localization length is a linear function of @xmath56 and @xmath115 . due to weak state - state interaction
the lsd is a delta - like function confined in an eigenstate : @xmath116 . at higher values of @xmath56 ( @xmath117 or @xmath118 )
the quantity @xmath13 satisfies the fermi golden rule : @xmath119 . despite the wide spreading of basis states over the eigenenergy scale ,
the actual number of eigenstates associated with an individual basis state for this parameter region remains small : @xmath120 .
[ h ! ] = 8 cm the criterion ( [ qcborder ] ) specifies a transition from the perturbative to localized regime both for eigenstate and for lsd .
depending on the ergodicity properties of localized lsd , one can distinguish two important parameter ranges .
an area , located between the borders ( [ qcborder ] ) and ( [ qeborder ] ) on the plane ( @xmath100 ) , corresponds to the parameter region @xmath121 where @xmath122 . for this region the neighboring eigenstates
are weakly coupled by interaction since the density @xmath123 of eigenstates , those involve the probing state , is negligible in comparison to the total state density @xmath124 .
it means that the level spacing statistics @xmath125 should exhibit the poisson - like distribution as it shown in fig.[figspacing ] .
we identify this region with a non - ergodic localized regime for lsd .
the regime of ergodic localization for lsd at @xmath126 is represented by a domain above the ergodicity border ( [ qeborder ] ) where @xmath127 . for this domain
the level spacing statistics is close to the wigner - dyson distribution ( see fig.[figspacing ] ) to be associated usually with quantum chaos . in a sense
the condition ( [ qeborder ] ) can be considered as a quantum chaos criterion .
[ h ! ] = 7.5 cm
we have analyzed the properties of lsd for a generic conservative system of coupled states , the hamiltonian of which is represented in terms of a brmdd . as for eigenstate , the transition from the perturbative to localized regime for lsd can be associated with the same requirement to be imposed on the strength of state - state interaction .
nevertheless , our study has revealed some distinctive features in the properties of lsd .
we have determined localization and ergodicity borders for lsd obtained from brmdd with arbitrary small band . for the matrices with large band ( @xmath78 )
the borders coincide with each other and are in good agreement with the quantum chaos border for some generic many - body systems . for the matrices with small band ( @xmath84 )
the criterion of lsd ergodicity is a more severe requirement imposed on the band size and the interaction strength than the localization criterion . in the parameter range between the borders the ipr
is found to increase exponentially with the ratio @xmath0 , which specifies the strength of state - state interaction .
we suppose that such an effect can be observed for any isolated system of coupled states with strong imperfection in non - perturbed energies .
avouris p , gelbart w m and el - sayed m a 1977 _ chem .
rev . _ * 77 * 793 ; khundkar l r and zewail a h 1990 _ annu .
chem . _ * 41 * 15 ; crim f f 1996 _ j. phys .
chem . _ * 100 * 12725 ; nesbitt d j and field r w 1996 _ j. phys .
* 100 * 12735 | the shape and the inverse participation ratio ( ipr ) of local spectral density ( lsd ) are studied for a generic isolated system of coupled quantum states , the hamiltonian of which is represented by a band random matrix with the disordered leading diagonal .
we find for the matrices with arbitrary small band that the lack of ergodicity for lsd can be associated with an exponential increase in ipr with the ratio @xmath0 ( @xmath1 - the root of mean square for off - diagonal matrix elements , @xmath2 - the energy spacing between directly coupled basis states ) .
criterions specifying transition to localization and ergodicity for lsd are considered . |
estimating accurate masses of self - gravitating systems , such as star clusters , is of central importance in astronomy .
most estimates of such masses have been based on radial velocities from a few tracers of the system .
however , the advent of the next generation telescopes , proper motion data will become more accurate and reliable within an observation time of a few years .
proper motions at the milky way s center has already proven a boon ( e.g. @xcite ) .
combined with radial velocities , proper motions can provide us with 5 components of the 6-dimensional phase space of a large number of tracers .
such a wealth of information will help to set accurate limits to masses of objects in the local group .
apart from the masses , this new kinematic data can uncover more subtle dynamic processes in star clusters , helping us to verify some of our present understanding and models of star clusters .
the idea of using proper motion data for cluster dynamics is not new .
kudworth was the first to implement this idea in the 70 s , but he was limited by the data and telescopes of the time .
@xcite showed that proper motion data was sufficient to predict the dynamic mass of the cluster .
however , there was no complete proper motion data sample of a star cluster at the time . at the present ,
data from hst have reached sufficient quality to study internal cluster kinematics astrometrically @xcite . also , diffraction limited imaging at the vlt
, keck and other 8 m class telescopes is now a reality , and with the next generation of telescopes coming online ( e.g. the lbt with 23 m resolution imaging , @xcite ) , proper motions are becoming a broadly competitive dynamical tool , and are not restricted to areas with long time - baseline data , such as @xmath0 centauri ( e.g. @xcite ) .
hence , it is crucial to understand how best to use proper motion data for setting mass limits and to learn to deal with the systematic and random errors inherent to this technique . in section [ sec : estimators ] , we undertake a review of present mass estimators and develop tools geared towards proper motion data , in the hope of extracting maximum information about the same .
we then apply these tools to a proper motion data set of @xmath0 centauri ( figure [ fig : problem ] ) in section [ sec : omega ] to predict the mass of the cluster .
the virial theorem has been traditionally and widely used to estimate the masses of self - gravitational systems from the motions of individual tracers or constituents . however
, a major drawback of estimators that are based on the virial theorem is that they tend to be biased and formally have an infinite variance .
@xcite compared the virial mass estimator to an alternative method called the projected mass estimator first proposed by @xcite for binary galaxies .
@xcite later expanded their work to develop the projected mass estimator for self - gravitating systems . while for many self - gravitating systems ( e.g. extra - galactic ones ) radial velocities are the only measurable quantities , it has become increasingly clear in the last decade that proper motions studies could be used to estimate the masses of nearby clusters .
@xcite extended this approach to include proper motions in the mass estimate of the open cluster m35 .
projected mass estimators try to exploit the directly observable , projected properties of a self - gravitating system .
in addition , they are efficient , unbiased and have a finite variance .
the estimators are derived by taking moments of the jeans equation for a spherical system .
the projected mass estimator for proper motion velocities can be rewritten as ( @xcite eq .
@xmath4 where @xmath5 and @xmath6 are the radial and transverse proper motion velocities , @xmath7 is the projected distance and @xmath8 is the total mass of the cluster .
since the projected mass estimator involves a line - of - sight projection , care must be taken to have a proper radial sampling of the tracers of the system .
note that in the projected mass estimator , particles at larger projected distances are given more weight than particles at smaller projected distances , because of the inability to reconstruct the true radial distance for particles at small projected distance .
the projected mass estimator becomes biased when the velocity measurements have significant error bars .
the estimated mass bias is a quadratic effect w.r.t the error in velocity .
however , we can modify the projected mass estimator to take into account the velocity errors . since @xmath9 ,
a reasonable unbiased estimator is @xmath10 where @xmath11 .
the projected mass estimator does not provide a `` formula '' for mass error estimates .
however bootstrapping methods can be used to set limits to the mass .
correcting for the errors in the velocities as in the above estimator is a statistical process , and may fail with a limited number of particles .
overestimation or underestimation of the errors in the proper motion velocities may bias the estimator . here
we develop a maximum likelihood method mass estimator that is geared towards proper motion measurements and accounts for significant , and individually varying , error bars .
it uses high - precision measurements effectively and extracts maximum information from low - precision velocity measurements .
the key to this method is a suitable model to predict the proper motion velocity dispersions ( both radial and transverse ) as a function of the projected radius .
for example , the velocity dispersion can be predicted by the family of king models or by jeans modelling of the cluster .
the data then simply set the mass normalization of the model .
let us consider that we can model the proper motion velocity distribution locally as a maxwellian distribution .
thus the probability of a star in the cluster to have a certain proper motion velocity @xmath12 given that the total mass of the cluster is @xmath8 is : @xmath13 where the velocity dispersions vary with radius and scale with mass as @xmath14 and @xmath7 is the projected radial distance . on the other hand ,
every observation of proper motion velocity @xmath15 has a certain error @xmath16 associated with it .
the probability of actually measuring the velocity @xmath15 at a projected distance @xmath17 is a gaussian distribution given by : @xmath18 thus the probability that the cluster has a mass @xmath8 given an observation of proper motion velocity @xmath15 , at a projected distance @xmath17 with an observational error @xmath16 is , @xmath19 the log - likelihood of the cluster to have a mass @xmath8 given @xmath20 observations is : @xmath21 where one sums over both radial and transverse proper motion velocities .
the most probable mass of the cluster corresponds to the parameter @xmath8 of the maximum of the likelihood function .
limits of uncertainty of @xmath22 can also be set to the mass of the cluster by searching for the parameter @xmath8 such that @xmath23 ( in the astrophysical context see @xcite , whose statistic s @xmath24 ; @xcite ) .
the southern globular cluster @xmath0 centauri ( ngc 5139 ) is approximately at a distance of 5.1 kpc @xcite .
its large mass and size sets it apart from the bulk of globular clusters in the galaxy . in recent years , it has evoked interest because of the recent discovery of its multiple populations which implies a complex formation history @xcite .
the cluster is also isotropic @xcite .
@xcite reported a definite rotation in the cluster .
@xcite reports that there is very little mass segregation in the cluster , which means that the cluster is not in equipartition .
there has also been a considerable debate on the mass of @xmath0 centauri .
previous estimates of the mass of @xmath0 centauri have ranged from @xmath25 @xcite to @xmath26 @xcite .
setting accurate limits to the mass of @xmath0 centauri is essential to differentiate between the various theories of its formation process .
@xcite have completed a project providing the largest available set of proper motion data for a star cluster .
this large data set on @xmath0 centauri made available offers a fore - taste of the proper motion data which may be soon made available for more clusters . for our mass estimates
, we use the proper motion studies of @xmath0 centauri @xcite , which are based on 100 plates obtained with the yale - columbia 66 cm refractor , and ranging in epoch from 1931 till 1935 and 1978 till 1983 , differential proper motions are obtained for 9847 stars . the limiting magnitude ( photographic )
is 16.0 for the centre of the cluster , 16.5 for the outer part .
the precisions of the proper motions range from an average of 0.1 mas / yr for the brightest to an average of 0.65 mas / yr for the faintest stars .
@xcite uses the proper motions to study the internal kinematics of the cluster .
the centre of the cluster was taken to be @xmath27 .
we assume that the distance of the cluster is 5.1 kpc away @xcite .
we project the absolute celestial coordinates @xmath28 and @xmath29 into cartesian coordinates ( relative to the cluster centre ) onto the plane of the sky along the line - of - sight vector through the cluster centre : @xmath30 where @xmath31 is a scaling factor ( @xmath32 to have @xmath33 and @xmath34 in radians and @xmath35 to have @xmath33 and @xmath34 in units of degrees ) , @xmath36 and @xmath37 are the cluster centre coordinates , and @xmath38 and @xmath39 .
@xcite divide their proper motion observations into 4 classes based on the errors in the proper motion determination which depend upon crowding and plate quality .
they also calculate the membership probability of each star to belong to the cluster .
we only use stars taken from high - quality classes 0 and 1 , since the errors appear to be relatively well behaved ( see fig . 11 of @xcite ) .
the mass estimators are very sensitive to outliers and field stars in the cluster .
thus the mass estimates show a direct correlation to our membership probability cut - off criterion .
we use a conservative cut - off criterion : we use stars having a membership probability higher than or equal to @xmath40% ( @xmath41 ) to belong to the cluster .
this gives us a limited sample of 4056 stars , which have a @xmath42 probability to belong to the cluster , due to the integral nature of the membership probability assigned by @xcite .
these we term as class a stars .
there are 772 stars having a membership probability between @xmath43 ( @xmath44 ) and @xmath45 to belong to the cluster .
these are class b stars . to diminish the effect of any non - members on the mass estimators ,
we construct a monte - carlo test by using any subset of class b stars along with class a stars . in choosing our subset ,
we ensure that the number of class b stars does not exceed @xmath46 of the class a stars .
as shown in figure [ fig : problem ] , the errors in the proper motion data becomes comparable to the velocity dispersion profile at a magnitude of 16 , especially for stars in the outer parts of the cluster .
thus reasonable and robust estimates of the mass can be obtained at magnitude of 15 .
recent proper motions observations of @xmath0 centauri by @xcite with the wfpc2 aboard the hst have shown that the data from @xcite may be problematic at the centre of the cluster .
@xcite obtain proper motions dispersions much lower than reported in the proper motion study of @xcite ; however , the two groups agree with each other farther out .
this will not affect our mass estimates from the pme , as it gives a lower precedence to data at the centre of the cluster .
for the mle , we neglect stars from the inner region of @xmath47 ; this is indicated by the shaded region of figure [ fig : problem ] .
the radial stellar profile was taken from @xcite . in this section
, we attempt to estimate the mass of @xmath0 centauri from the proper motion data set of @xcite using the tools we have developed in the previous section .
first , we apply the pme ( eq . [ eqn : proj2 ] ) to the data set ; then we use the mle from section [ sec : mle ] with various models of the projected velocity dispersion profile .
one must be wary of selection effects in applying the pme .
crowding effects and lack of colour information at the centre of the cluster leads to a bias in the radial sampling of stars with proper motions .
a simple check can be made by calculating the average projected radius as a function of the limiting magnitude for our restricted sample : we find that fainter sub - samples lie at larger radii .
thus we have a larger number of faint stars away from the centre of the cluster in the sample . ) as a function of the class 0 and 1 stars from @xcite .
the @xmath48 error bars are calculated using bootstrapping .
the closed symbols use class a stars only .
the open symbols use both class a and class b stars in a monte - carlo method.,width=317 ] we can correct for this effect by using a monte - carlo algorithm to choose our tracers for the pme such that we always maintain the surface density profile of the cluster . for this purpose ,
we fit a kings density profile to surface density data ( see figure [ fig : problem ] ) published by @xcite .
the 1-@xmath49 error bars ( see figure [ fig : estimator ] ) are calculated using bootstrapping . at very bright magnitudes
, we have high precision measurements , but only few stars .
the large error bars demonstrate the lack of information due to the limited amount of tracers of the system . in trying to extract the maximum information in an economic way with the pme , one has to trade off between the above two effects . at magnitude of 15 using only class a stars , the pme estimates the mass of the cluster as @xmath50 , while the combination of class a as well as class b stars in a monte - carlo method predicts @xmath51 .
to apply the mle to @xmath0 centauri , we require a suitable model for the velocity dispersion .
we start with the classic king model @xcite framework for this task .
given a suitable dynamical model of the cluster , the tracers of the system effectively scale the projected velocity dispersion curve with the help of a maximum likelihood method to estimate the independent parameter : the mass of the cluster . as with the pme , we also correct for `` proper radial '' sampling .
we also neglect all stars in the central region of @xmath52 due to the uncertainty in the central velocity dispersion from crowding effects . at most @xmath53 of all the tracers of our limited sample
are present in this central region . in figure
[ fig : single - mass - kings - model ] , a single mass king model is fit to the surface density profile of @xmath0 centauri .
the central scaled potential used is @xmath54 , to match the surface radial profile @xcite .
since the search for anisotropy in @xmath0 centauri has yielded very limited results @xcite , we are justified in using one integral models .
even though the surface density profile provides a relatively good fit , we are unable to reproduce the projected velocity dispersion profile in the inner parts of the cluster ( even well outside @xmath52 ) .
the resulting mass estimate from the mle as a function of limiting magnitude is also shown . at magnitude of 15 using only class a stars ,
the mle with the king model estimates the mass of the cluster as @xmath55 , while the combination of class a as well as class b stars in a monte - carlo method predicts @xmath56 . without assuming any restrictive form of a distribution function
, we can still reconstruct the velocity dispersion using jeans equation . under the hypothesis of isotropy and neglecting the rotation in the cluster ,
the isotropic jeans equation for a spherical system is : @xmath57 where @xmath49 is the radial velocity dispersion , @xmath58 is the observable number density distribution , and @xmath59 is the potential generated by the matter distribution at the considered point .
note that @xmath0 centauri is one of the few systems known to be isotropic from direct observations .
we use the three dimensional number density distribution by @xcite using non - parametric analysis .
assuming that mass does not follow light , we try to disentangle the mass profile from the luminosity profile , by giving it an independent king profile @xcite .
note that the king mass profile only defines the potential well of the cluster , and does not assume the particular form of the kings distribution function .
the mass profile was adjusted until the form of the velocity dispersion curve predicted by jeans equation matched the kinematic data of @xmath0 centauri as derived by @xcite .
this is shown in figure [ fig : jeans - model ] .
the velocity dispersion curve is then used to predict the mass of the cluster using the mle .
the shape of the mass profile is consistent with the fact that @xmath0 centauri exhibits very little equipartition , as shown by the results of @xcite , where it was found that stars of a considerable range of mass down the main sequence of @xmath0 centauri differed very little in their radial distribution ; the differences found indicate only a small evolution towards equipartition .
the lack of equipartition is to be expected with a cluster of such large mass , and hence a very long relaxation time . at lower magnitude
the error - bars are large due to lack of information in the limited number of tracers . at magnitude of 15 using only class a stars ,
the mle with the jeans model estimates the mass of the cluster as @xmath50 , while the combination of class a as well as class b stars in a monte - carlo method predicts @xmath60 .
the resulting mle mass estimates for the cluster from the two models differ by @xmath3 because of the difference in the velocity dispersion profiles at the central region of the cluster ( @xmath61 ) . on the other hand ,
the results of the mle with jeans modelling are consistent with the results of the pme .
in this paper , we implement two independent methods to determine the mass of a cluster from the proper motion data .
when using the pme , one must be careful to correct for the errors in the velocities ( eq . [ eqn : proj2 ] ) and to use a `` proper radial sampling '' to correct for the incompleteness at the centre of the cluster .
the pme provides us an unbiased and robust estimate of the mass of the cluster , so long as the errors in the velocities are as large as the velocity dispersion itself .
using stars above this limiting magnitude leads us to a systematic error .
we independently estimated the mass of the cluster using the mle in conjunction with two dynamical models for the velocity dispersion profile . in this paper
, we use the king model and jeans modelling .
disentangling the mass profile from the density profile in the jeans modelling , gives us a better prediction of the shape of the velocity dispersion profile .
both models give us very similar mass estimates of the cluster within a @xmath3 systematic error .
the mass estimators are susceptible to a systematic bias due to contamination by outliers and field stars in the cluster .
the largest uncertainty in the absolute mass estimate is the distance to @xmath0 centauri which enters as the third power .
a comparison of the radial velocity dispersion @xcite with the proper motion velocity dispersions indicates a distance of the cluster of @xmath62 kpc @xcite
. however we follow @xcite in interpreting these results with caution , and assume that the cluster is at a distance of @xmath63 kpc away . thus , when implemented correctly , the two independent methods predict the mass of the cluster within @xmath64 on the basis of the @xcite data .
the mass of @xmath0 centauri is @xmath65^{3}$ ] with a @xmath2 statistical error and a @xmath3 systematic error which arises from the various models used .
the choice of the limiting magnitude introduces a further systematic error of @xmath3 . | a projected mass estimator and a maximum likelihood estimator are developed to best determine the mass of stellar clusters from proper motion data .
both methods effectively account and correct for errors in the velocities , and provide an unbiased and robust estimate of the cluster mass . using an extensive proper motion study of @xmath0 centauri by van leeuwen et al .
( 2000 ) , we estimate the mass of @xmath0 centauri using these two different methods to be @xmath1^{3}$ ] . within this modeling context , the statistical error is @xmath2 and the systematic error is @xmath3 .
[ firstpage ] stellar dynamics - celestial mechanics - galaxy : globular clusters : individual : ngc 5139 |
the main injector particle production experiment ( fnal e-907 , mipp ) @xcite is situated in the meson center beamline at fermilab .
it received its approval @xcite in november 2001 and has installed both the experiment and a newly designed secondary beamline in the interim .
it received its first beams in march 2004 and concluded its engineering run in august 2004 .
the physics data taking run of mipp is scheduled for the running period december 2004-july 2005 .
mipp is designed primarily as an experiment to measure and study in detail the dynamics associated with non - perturbative strong interactions .
it has nearly 100% acceptance for charged particles and excellent momentum resolution . using particle identification techniques that encompass @xmath0
, time - of - flight @xcite , multi - cell cerenkov @xcite and a ring imaging cerenkov ( rich)detector @xcite , mipp is able to identify charged particles at the 3@xmath1 or better level in nearly all of its final state phase space .
as currently envisaged , mipp will provide events of unparalleled quality and statistics for beam momenta ranging from 5 gev / c to 90 gev / c for 6 beam species ( @xmath2and@xmath3 . )
the primary physics motivation behind mipp is to restart the study of non - perturbative qcd interactions , which constitute over 99% of the strong interaction cross section .
the available data are of poor quality and old and are not in easily accessible form . the time projection chamber ( tpc ) @xcite that is at the heart of the mipp experiment represents the electronic equivalent of the bubble chamber with vastly superior data acquisition rates .
it also digitizes the charged tracks in three dimensions , obviating the need for track matching across stereo views .
coupled with the particle identification capability of mipp , the data from mipp would add significantly to our knowledge base of non - perturbative qcd .
one of the primary goals of the present run of mipp is to verify a general scaling law of inclusive particle production that states that the ratio of a semi - inclusive cross section to an inclusive cross section involving the same particles is a function only of the missing mass squared ( @xmath4 ) of the system and not of the other two mandelstam variables @xmath5 and @xmath6 , the center of mass energy squared and the momentum transfer squared , respectively . stated mathematically , the ratio @xmath7 _ i.e. _ , a ratio of two functions of three variables is only a function of one of them .
this scaling relation has been shown to hold very well in a limited number of reactions @xcite .
mipp will test this scaling for 36 reactions as a function of both @xmath5 and @xmath6 with great accuracy .
in addition to this , mipp will acquire high quality data in liquid hydrogen with excellent particle i d and statistics over a range of beam momenta , which should make possible a systematic study of exclusive reactions that is essential for testing any future theory of non - perturbative qcd .
we can also make forays into searches for exotic resonances such as glueballs and pentaquarks .
the existence of beams of differing flavor and energies will be a great advantage in sorting out the flavor content of any new states seen .
the other physics clients for mipp data are nuclear and heavy ion physics groups who are interested in data from several nuclear targets .
an important service measurement mipp hopes to perform is the measurement of particle production off the numi target in order to minimize the systematics in the near / far detector ratio in the minos @xcite experiment .
numi target measurements by mipp will also benefit the minerva @xcite and the nova @xcite experiments planned in the numi beamline in the future .
mipp will also make measurements with proton beams off various nuclei for the needs of proton radiography @xcite .
another measurement mipp hopes to make is that of pion and proton cross - sections off liquid nitrogen targets for the better prediction of atmospheric neutrino fluxes .
mipp is approved for @xmath8 spill seconds of beam to be partitioned according to the running modes shown in table [ tab1 ] .
c c c program & targets & number of events ( @xmath9 ) + atm .
neutrinos & @xmath10 & 3.00 + numi & thin c , numi target & 10.01 + proton - nucleus & be , cu , pb & 17.20 + qcd scaling & @xmath11,@xmath12 & 24.40 + proton radiography & be , c , cu , pb , u & 20.79 +
we designed a secondary beam @xcite specific to our needs .
the 120 gev / c primary protons extracted by slow resonant extraction from the fermilab main injector are transported down the meson center line at fermilab .
they impinge on a 20 cm long copper target producing secondary beam particles .
this target is imaged on to an adjustable momentum selection collimator which controls the momentum spread of the beam .
this collimator is re - imaged on to our interaction target placed next to the tpc .
the beam is tracked using three beam chambers and identified using two differential cerenkovs @xcite filled with gas , the composition and the pressure of which can be varied within limits depending onthe beam momentum and charge .
the study of non - perturbative qcd is of fundamental importance to our field .
however , we place too little priority on re - embarking on this quest , largely because of the lack of a fundamental theory .
mipp hopes to change this state of affairs by obtaining high quality data with large statistics , which it will publish as dst s on dvd s for public use .
however , it was important that the cost of the detector be low in order for the experiment to be approved and this was achieved by refurbishing several existing pieces of apparatus . figure [ mipp ] shows the layout of the apparatus .
the tpc sits in a wide aperture magnet ( the jolly green giant ) which has a peak field of 0.7 tesla .
downstream of the tpc are a 96 mirror multi - cell cerenkov detector filled with @xmath13 gas , and a time of flight system .
this is followed by a large aperture magnet ( rosie ) which runs in opposite polarity ( at -0.6 tesla)to the jolly green giant to bend the particles back into the ring imaging cerenkov counter .
the rich has @xmath14 as the radiator and an array of phototubes of 32 rows and 89 columns @xcite .
downstream of the rich we have an electromagnetic calorimeter @xcite and a hadron calorimeter @xcite to measure forward going photons and neutrons .
the electromagnetic calorimeter will also serve as a device to measure the electron content of our beam at lower energies , which will be useful for measuring cross sections .
the experimental setup .
the picture is a rendition in geant3 , which is used to simulate the detector ] mipp uses @xmath15 in the tpc to separate pions , kaons and protons for momenta less than @xmath16 , the time of flight array of counters to do the particle i d for momenta less than 2 gev / c , the multi - cell cerenkov detector @xcite contributes to particle i d in the momentum range 2.5 gev / c-7.5 gev / c and the rich @xcite for momenta higher than this . by combining information from all counters ,
we get the expected particle i d separation for @xmath17 and @xmath18 as shown in figure [ pid ] .
it can be seen that excellent separation at the @xmath19 or higher level exists for both @xmath17 and @xmath20 over almost all of phase space .
tracking of the beam particles and secondary beam particles is accomplished by a set of drift chambers @xcite and proportional chambers @xcite each of which have 4 stereo layers .
particle i d plots for pion / kaon separation and for kaon / proton separation as a function of the longitudinal and transverse momentum of the outgoing final state particle .
black indicates separation at the @xmath19 level or better and grey indicates separation at the @xmath21 level . ]
particle i d plots for pion / kaon separation and for kaon / proton separation as a function of the longitudinal and transverse momentum of the outgoing final state particle .
black indicates separation at the @xmath19 level or better and grey indicates separation at the @xmath21 level . ]
figure [ tpc ] shows the pictures of reconstructed tracks in the tpc obtained during the engineering run .
the tracks are digitized and fitted as helices in three dimensions . extrapolating three dimensional tracks to the other chambers makes the pattern recognition particularly easy .
pictures of tpc events .
the first set of four pictures on the top left hand side show the raw digitizations of an event in 4 views and the set of four pictures on the right hand side show the tracks pattern recognized on the same event .
tracks in different views have the same color .
the last picture is that of an upstream interaction .
the offline algorithm has found all the tracks ] pictures of tpc events .
the first set of four pictures on the top left hand side show the raw digitizations of an event in 4 views and the set of four pictures on the right hand side show the tracks pattern recognized on the same event .
tracks in different views have the same color .
the last picture is that of an upstream interaction .
the offline algorithm has found all the tracks ] pictures of tpc events .
the first set of four pictures on the top left hand side show the raw digitizations of an event in 4 views and the set of four pictures on the right hand side show the tracks pattern recognized on the same event .
tracks in different views have the same color .
the last picture is that of an upstream interaction .
the offline algorithm has found all the tracks ] pictures of tpc events .
the first set of four pictures on the top left hand side show the raw digitizations of an event in 4 views and the set of four pictures on the right hand side show the tracks pattern recognized on the same event .
tracks in different views have the same color .
the last picture is that of an upstream interaction .
the offline algorithm has found all the tracks ] figure [ rings ] shows events with rings in the rich counter .
some are due to single beam tracks and others are due to tracks from interactions .
examples of events with rings in the rich counter for a 40 gev / c beam . ] examples of events with rings in the rich counter for a 40 gev / c beam . ]
examples of events with rings in the rich counter for a 40 gev / c beam . ]
figure [ rbck ] shows the histogram of ring radii for a + 40gev secondary beam .
there is clean separation between pions and kaons and protons and their relative abundances @xcite match expectations . applying the particle
i d trigger from the beam cerenkovs enables us to separate the three particle species cleanly .
the kaons which form 4% of the beam are cleanly picked out by the beam cerenkov with very simple selection criteria .
these can be made much more stringent with offline cuts to produce a very clean kaon beam .
an example of a 40 gev / c primary beam trigger .
the rich identifies protons , kaons and pions by the ring radii .
the beam cerenkov detectors can be used to do the same .
when the beam cerenkov i d is used , one gets a very clean separation of pions , kaons and protons in the rich ] an example of a 40 gev / c primary beam trigger .
the rich identifies protons , kaons and pions by the ring radii .
the beam cerenkov detectors can be used to do the same .
when the beam cerenkov i d is used , one gets a very clean separation of pions , kaons and protons in the rich ] an example of a 40 gev / c primary beam trigger .
the rich identifies protons , kaons and pions by the ring radii .
the beam cerenkov detectors can be used to do the same .
when the beam cerenkov i d is used , one gets a very clean separation of pions , kaons and protons in the rich ] an example of a 40 gev / c primary beam trigger .
the rich identifies protons , kaons and pions by the ring radii .
the beam cerenkov detectors can be used to do the same .
when the beam cerenkov i d is used , one gets a very clean separation of pions , kaons and protons in the rich ]
at present the data taking rate in mipp is limited by the 60hz rate imposed on us by the tpc electronics .
these can be made considerably faster ( by at least a factor of 20 ) @xcite with more modern electronics .
upgrading the tpc electronics this way accompanied by modest upgrades to the rest of the system is expected to cost @xmath22 .
if approved in 2005 , such an upgrade can be implemented in time to continue the mipp run in 2006 .
this will enable mipp to expand its present scope to include pentaquark searches , searches for missing baryon resonances as well as partial wave analyses using low energy beams .
9 the main mipp web page is at http://ppd.fnal.gov / experiments / e907/. the mipp collaboration list may be found at http://ppd.fnal.gov/experiments/e907/contacts/ the mipp proposal and addendum to the proposal may be found at http://ppd.fnal.gov/experiments/e907/proposal/e907_proposal.html the time of flight detector was fabricated by mipp and consists of an array of 10@xmath2310 cm scintillators and 5@xmath235 cm scintillators .
see http://ppd.fnal.gov/experiments/e907/tof/tof.html for a detailed description of the detector .
the multi - cell cerenkov detector was initially built for brookhaven experiment e766 and later in fermilab experiment e690 and then used in several other brookhaven experiments . in mipp , we fill the detector with the gas @xmath13 which has the appropriate refractive index at atmospheric pressure .
the details of the selex rich construction and performance may be found at j. engelfried _ et al .
instr . and meth .
_ * a43*,53(1999 ) .
we have replaced the front end electronics , and done extensive work on the safety systems .
mipp uses co@xmath24 gas as the radiator for the rich .
the tpc was built by the bevalac group at berkeley in the 1990 s and used effectively at several brookhaven experiments ( e.g. e910 ) and then donated to fermilab by lbnl for use in mipp .
see , g. rai _ et al . , ieee trans.nucl.sci._*37*,56(1990);lbl-28141 .
r.raja , _ phys .
d _ * 18 * , 204 ( 1978 ) .
r.raja , _ phys .
d _ * 16 * , 142 ( 1977 ) .
r.raja , y.fisyak in proceedings of the dpf92 meeting , fermilab .
minos proposal may be found at `` p-875 : a long baseline neutrino oscillation experiment at fermilab '' , e. ables_et al . _ ; fermilab - proposal - p875,(1995 ) .
see also their website at http://www-numi.fnal.gov/ the minerva website is at http://www.pas.rochester.edu/ ksmcf / minerva/ the nova experiment is currently still a proposal and the website is at http://www-nova.fnal.gov/ `` beamline design for particle production experiment , e907 at fnal '' , c.johnstone _ et al .
_ , proceedings of the pac03 conference . for more details on the beam cerenkov system
, see + http://ppd.fnal.gov/experiments/e907/beam/beamcerenkov/beamcerenkov.html during the engineering run , we lost 20% of the phototubes in the rich due to a fire in one of the bases .
this does not impact adversely on our pattern recognition , since the cerenkov angle is large and there is plenty of light over most of our momentum range the electromagnetic calorimeter was fabricated by mipp and uses lead as the radiator and an array of proportional tubes with 2.54 cm wire spacing as the readout .
it has 10 radiation lengths and has 10 longitudinal segments .
the hadron calorimeter is recycled from the hypercp collaboration and uses scintillator fibers embedded in iron as readout .
it has 9.7 interaction lengths and has four longitudinal segmentations each of which is segmented in two transversely .
we have reused beam and drift chambers from the e690 collaboration .
instr . and meth .
_ * a345,62 * ( 1994 ) .
the large proportional chambers straddling the rich find their use again after having been used by numerous previous experiments .
m. de palma _ et al .
instr . and meth .
_ * 216 * ( 1983 ) 393 - 397 .
a. malensek , fermilab technical memo fn-341 , 1981 .
( unpublished ) .
the tpc daq upgrade scheme may be found at + http://ppd.fnal.gov/experiments/e907/notes/mippnotes/public/ps/mipp0042/mipp0042.ps.gz | we describe the physics capabilities and status of the mipp experiment which is scheduled to enter its physics data taking period during december 2004-july 2005 .
we show some of the results obtained from the engineering run that concluded in august 2004 and point out the unique features of mipp that make it an ideal apparatus to study non - perturbative qcd properties . . |
bekenstein - hawking entropy @xmath0 has been derived from different points of view @xcite .
but if it is considered thermal with a microscopic description , according to the statistical foundations of entropy , perhaps the most promising and appropriate formalism to explain @xmath0 is the entanglement entropy approach @xcite .
in particular , entanglement entropy of black shells is required , because the usual thermal entropies for black holes are divergent and geometric in nature @xcite . in the first part of this paper
we complete the concept of black shell presented in @xcite .
thus , by simplicity we effectively model the significant features of the gravitational collapse , in terms of a massive thin spherical collapsing shell with respect to an external observer . in the second part of the text we consider euclidean approach for entropy of black shells , thinking in its complementary description of entanglement entropy .
thermal entropy or entanglement entropy modeled by a black shell is a real physical model for @xmath0 , because it corresponds to the thermodynamics of hot quantum fields confined near the outside of the shell @xcite . unlike black holes that do nt have structure and require a picture of quantum gravity to describe its geometrical entropy as some unknown physics , black shells have structure and thermal physical entropy .
we show below that entropy calculated from gibbons - hawking euclidean approach for spherical black shells , retrieving the horizon integral and discarding boundary at infinity , preserves the same expression for bekenstein - hawking entropy : @xmath1 so entanglement entropy for black shells according to an external observer do not need to consider quantum gravity criteria to explain @xmath0 . in the last part of the paper we complete the model by describing complementary quantum details . in sec .
2 we present the black shell model and review the darmois - israel formalism @xcite that is needed in order to obtain the motion equations of this model that we solve in sec .
3 . in sec .
4 we reproduce the well known bekenstein - hawking entropy from euclidean approach for a schwarzschild black hole @xcite . in sec . 5
we retrieve the horizon integral and discard boundary at infinity in order to calculate the entropy of a black shell . there
we introduce a mathematical identity to support our two complementary descriptions of physics near an event horizon . in sec .
6 , for completeness , we return to entanglement entropy of black shells .
so , we reproduce some results from @xcite and complete details of the corresponding hamiltonian formulation .
7 is devoted to discuss some conclusions about this paper .
we present in this section a model where a black hole is formed by a thin contracting shell of dust of mass @xmath2 that contracts beginning from infinity and at last tends to the schwarzschild radius from the point of view of an external observer far from the shell in a gravitational collapse evolution of oppenheimersnyder kind @xcite , as we see in section 3 .
+ in this context we employ the darmois - israel formalism @xcite for a spherical thin shell @xmath3 which divides the spacetime in two regions : the interior region @xmath4 , described by flat minkowskian geometry and @xmath5 , the exterior geometry described by schwarzschild spacetime .
both regions are described by spherical coordinates : @xmath6 , and we use geometric units in which : @xmath7 , and signature : @xmath8 .
then the line element is : @xmath9 where : @xmath10 for @xmath5 and @xmath11 for @xmath4 .
the shell hypersurface @xmath12 is represented in the chosen frame by the following parametric equation : @xmath13 where @xmath14 is the thin shell radius .
the following intrinsic coordinates on the shell are used : + @xmath15 , and the 3-metric elements induced in the hypersurface are : @xmath16 we note that the relation between coordinates of @xmath17 and intrinsic coordinates on @xmath3 are : @xmath18 and the parametric equation for @xmath12 adopts the following general form : @xmath19 we assume that @xmath12 is non - null and the unit 4-normals to @xmath12 in @xmath20 are : @xmath21 the extrinsic curvature ( seccond fundamental form ) is defined by : @xmath22 using ( [ bs1 ] ) and ( [ f2 ] ) the normal ( [ f6 ] ) is : @xmath23 with the usual formula for the covariant derivative equations ( [ f4 ] ) and ( [ f7 ] ) yields the simple relation : @xmath24 raising indexes and considering that schwarzschild metric is diagonal , we arrive to the simple results : @xmath25=\frac{1}{2}f^{-\frac{1}{2}}\partial_{r}f,\ ] ] @xmath26=\left[k_{\varphi}^{\varphi}\right]=\frac{1}{r}f^{\frac{1}{2}}-\frac{1}{r},\ ] ] @xmath27=f^{-\frac{1}{2}}(\frac{1}{2}\partial_{r}f+\frac{2}{r}f - f^{\frac{1}{2}}\frac{2}{r}).\ ] ] where square brackets denotes a discontinuity across the layer , i.e. , @xmath28=f^{+}-f^{-}$ ] .
our following task is to perform the calculations to obtain the motion equations of the shell and the bekenstein - hawking entropy of the shell .
this is developed in sections 3 and 5 .
in this section we review the thin shells junction formalism @xcite , in order to study the motion of a spherical shell of dust that contracts beginning at rest from infinity as seen by a distant observer whose proper time is the schwarzschild time @xmath29 . for this purpose
we consider the surface energy tensor of a shell of dust : @xmath30 with the condition : @xmath31 and where @xmath32 is the rest mass surface density of the dust .
if we consider a trajectory @xmath33 of an element of rest mass on the shell , then the 3-vector associated @xmath34 is tangent to @xmath3 . for a spherical shell , the region @xmath5exterior to the shell
have the metric given by equation ( 2 ) . and the region @xmath35interior to the shell is minkowskian .
the equation of the shell is : @xmath36 where @xmath37 is the proper time measured by the dust particles .
the equation of motion of the shell could be obtained from the lanczos equation @xcite : @xmath38-g_{ij}\left[k_{ab}g^{ab}\right]\right).\ ] ] from equations ( [ bs1 ] ) , ( [ fc1 ] ) , ( [ fc2 ] ) and ( [ fc3 ] ) we obtain rearranging therms : @xmath39 we are interested in the velocity of the shell as measured by an external observer at rest at infinity , then we must replace : @xmath40 if the shell is initially at rest : @xmath41 because gravity is attractive it produces a flux entering to the thin gaussian region that encloses the shell and by this reason we must choose the negative sign . with these consederations
we finally obtain : @xmath42 this equation is of difficult integration but fortunately we can do the following good approximation : + + @xmath43 @xmath44 where @xmath45 must be chosen in order to match these two expressions for @xmath46 , where @xmath47 from this equation we obtain : @xmath48 and @xmath49
. then the motion equation reduces to : @xmath50 this equation could be integrated analytically to obtain : @xmath51\right)}e^{\left(-\frac{t}{\tau}\right)}.\ ] ] where @xmath52 , @xmath53 is the initial radius and @xmath54 and @xmath53 are measured in units of schwarzschild radius . taking little variations of @xmath55 respect to @xmath53 we can obtain : @xmath56 this important result sais that from the point of view of an external observer at rest and far from the horizon , the shell approaches to schwarzschild radius asymptotically .
this is that we call a black shell model of a black hole . for an observer comoving with the shell it is easy to see from ( [ fc5 ] ) , ( [ fc7 ] ) , ( [ fc8a ] ) and ( [ fc9 ] ) that
: @xmath57 integrating this equation we obtain : @xmath58 then we see that the shell reaches the radius @xmath59 in the finite time : @xmath60 + this is that we call a black shell free gravitational collapse .
in this section we review the statistical derivation of bekenstein - hawking entropy ( @xmath0 ) for stationary black holes , using analytic continuation to euclidean sector and imposing a period on euclidean time . according to gibbons - hawking derivation @xcite , and in order to obtain the bekenstein - hawking entropy
, we calculate the action @xmath61 for the schwarzschild metric ( [ bs1 ] ) : @xmath62 where @xmath20 is the schwarzschild spacetime with ( in natural units ) : @xmath10 , @xmath59 and @xmath45 is the second fundamental form . using the equations ( [ f11 ] ) , ( [ f12 ] ) and ( [ f13 ] ) for @xmath63 and @xmath64 we obtain : + + @xmath65 @xmath66 where @xmath67 is the euclidean period .
then for the black hole mass , @xmath68 , we obtain : @xmath69 this is the bekenstein - hawking entropy of a black hole as was derived by gibbons - hawking @xcite .
for a black shell , according to section 3 above , the radius approaches to the schwarzschild one and by this reason we retrieve the horizon integral obtaining the euclidean action : + + @xmath65 @xmath70 where @xmath67 is the euclidean period . for a black shell
the inner space is the empty space and by this reason : @xmath71 obtaining : @xmath72 both results : ( [ fo3 ] ) and ( [ euc13e ] ) , are mathematically equivalent derivations of @xmath0 . in order to relate both procedures , consider the following mathematical identity : @xmath73 @xmath74 this identity corresponds to a physical model that resolves one of the questions raised in mukohyama - israel @xcite , in the sense that @xmath0 is not a one - loop correction to the zero - loop gibbons - hawking contribution . indeed
this entropy is a zero - loop black shell contribution .
according to entanglement entropy model of black shells , thermal energy strongly concentrated near the exterior of a starlike object is clearly evident .
this object has a reflecting surface , compressed to nearly ( but not quite ) its gravitational radius . following this model
we may approximate the total stress - energy ( ground state and thermal excitations ) @xmath75 , near the wall , to the hartle - hawking stress - energy @xmath76 @xcite : @xmath77 where @xmath78 is boulware stress tensor and @xmath79 are thermal excitations .
thermofield dynamics encodes a reflexive symmetry between twin subsystems in such a way that each of them becomes macroscopically indistinguishable from a hot body at a definite temperature .
the reflection symmetry of these subsystems corresponds to the right and left regions of an eternal black hole , i.e. , kruskal sectors r and l see ( fig .
[ mg10i ] ) .
thus , the twin subsystems correspond to field modes propagating in these ( causally disjoint ) kruskal sectors , and the thermally entangled state of the total system corresponds to the ground state on the full kruskal manifold . in that sense
, this section defines modes and ground states appropriate for the subsystems @xmath80 and for the global system ( full kruskal manifold ) , i.e. , for stationary observers in a static spacetime ( killing - boulware modes ) and for free falling observers at the horizon of a black hole spacetime ( kruskal - hartle - hawking modes ) , and establishes the relationship between them .
+ from the line element ( [ bs1 ] ) consider the generic situation of a real scalar field @xmath81 propagating on a geometrical background with static metric given by @xmath82 on the other hand , @xmath81 may be expanded in terms of kruskal - hartle - hawking ( kh@xmath83)-modes @xmath84 and killing - boulware ( kb)-modes @xmath85 , which are connected by the bogoliubov transformation @xcite @xmath86 with @xmath87 . the physical sense of ( [ ete30 ] ) is based in the invariant action @xmath88 $ ] and invariant hamiltonian @xmath89 under this transformation , according to the action @xmath90 = \int \mathcal{l}[\phi ] d^{4}x = \int l[\phi ] dt_{+ } = \int_{-\infty}^{\infty}dt \left(\sum_{\epsilon}\epsilon l^{(\epsilon)}(\phi)\right),$}\ ] ] where the integration regions for the first and second integrals corresponds in fig .
[ mg10ii ] to the second and first graphs , respectively .
there @xmath91 takes same value at two mirror points in @xmath92 and @xmath55 sectors .
these two points are distinguished by different imaginary parts of @xmath93 : @xmath94 with @xmath95 and @xmath96 , the surface gravity .
@xmath97 is analytic over the full kruskal manifold , with the property that @xmath98 is positive frequency in kruskal time .
@xmath92 sector contribution to @xmath99 enters with negative sign , because @xmath29 and @xmath93 run backwards in this sector .
modes in the @xmath92 sector and beneath the horizon have negative energies .
we are talking about the continuous extension of the killing vector which is future timelike in the @xmath55 sector , and therefore becomes past - directed in the @xmath92 sector .
physically , this definition of energy includes gravitational potential energy , which is negative and becomes large below the horizon . in the hartle - hawking state , positive energies in the @xmath55 sector are evenly balanced with negative energies in the @xmath92 sector .
the state of the black hole is represented globally by the hartle - hawking state @xmath100 @xcite .
the total energy of a quantum field in this state is zero , because of the balance between positive - energy modes in the r sector and negative - energy modes in the l sector .
formally , computing the expectation value of the global energy operator @xmath101 in the state @xmath100 , @xmath102 where @xmath103 on the other hand , the ground state @xmath104 for a parochial observer in @xmath55 is energetically depressed below the hartle - hawking `` vacuum '' .
if we form the expectation value of @xmath101 in the state @xmath104 , only the l - sector part ( @xmath105 in eq .
( [ ete54 ] ) ) of @xmath101 contributes , since there are no r - sector modes in this state .
we thus find @xmath106 correspondingly , this same parochial observer , using his parochial energy operator @xmath107 , will perceive the hartle - hawking vacuum as energetically excited .
if we form the expectation value of @xmath107 in the state @xmath100 , the operator will pick out only the positive - energy modes in @xmath55 , so @xmath108 where we have dropped zero - point energies .
entanglement entropy actually allows us to think that entropy arises physically located near the horizon , given by @xcite @xmath109 where @xmath110 the expressions above are similars to brick wall model @xcite , where @xmath111 is defined by the external observer , which it is the main difference with respect to brick wall model .
in addition @xmath112 is finite according to thermo field dynamics @xcite . according to this model , the integral ( [ ete191 ] )
is dominated by two contributions , for large @xmath113 and for small @xmath114 .
the former corresponds to a volume term , proportional to @xmath115 , which represents the entropy and energy of a homogeneous quantum gas in a flat space at a uniform temperature @xmath116 .
the latter is the contribution of gas near the inner wall @xmath117 .
then , for this last contribution is required to introduce the ultrarelativistic approximations @xmath118 substituting eq .
( [ bw8 ] ) into eq .
( [ ete191 ] ) , the wall contribution to the total entropy is obtained @xmath119 where @xmath120 accounts for helicities and the number of particle species , @xmath121 is the wall area and @xmath122 is the proper radial distance from horizon to the shell .
now , depending on @xmath122 , we can obtain the bekenstein - hawking entropy from eq .
( [ bw10 ] ) @xmath123 where @xmath122 can be adjusted by resorting to times determined by an external observer in the context of the black shell model .
+ unlike black holes , equation ( [ bw10 ] ) corresponds to a microscopic description in terms of quantum field modes .
we have described in sections 2 and 3 the motion of a spherical shell of dust that contracts beginning at rest from infinity by using darmois - israel thin shells formalism @xcite , then we obtain from the point of view of an external observer far from the horizon , that shell approaches to schwarzschild radius .
this result is one of the main characteristic of the black shell model , introduced above .
4 we reproduce geometrical bekenstein - hawking entropy of a schwarzschild black hole using the well known gibbons - hawking euclidean approach @xcite and derive in sec
. 5 physical entropy of a black shell from gibbons - hawking euclidean approach retrieving horizon integral and discarding boundary at infinity .
this important result was obtained considering that for a black shell the mass is outside the horizon . + for the sake of completeness , in sec .
6 we return to entanglement entropy of black shells , because it is a complementary description of gibbons - hawking euclidean approach for a black shell .
so we have completed and extended the idea of black shell presented in @xcite . in particular , we reproduce some results about it and complete details of the corresponding hamiltonian formulation . + we agree with s. mukohyama and w. israel that entropy contributed by thermal excitations or entanglement is not a one loop correction to the zero - loop gibbons - hawking contribution
. actually we may consider these two entropy sources as equivalent but mutually exclusive descriptions of what is externally the same physical situation @xcite . in equation ( [ euc13f ] ) we observe that in the euclidean sector of the black shell space - time there is an inner boundary , the black shell itself .
thus inner boundary contribution to the euclidean action cancels that of the outer boundary at infinity .
so the gibbons - hawking zero - loop contribution is zero in this sense . + with the right identification of the ground state , the back - reaction problem in t hooft brick wall model is resolved @xcite . in that sense for black shells ,
we show in ( [ ete88 ] ) that the divergent parts cancel each other .
the boulware ground state contribution ( which is energetically depressed below the vacuum ) and thermal excitations cancel .
+ in summary we propose an effective model consisting on a massive thin spherical shell contracting toward its gravitational radius with respect to an external observer in order to describe significant features of a gravitational collapsing mass .
the collapsing massive shell is compressed near its gravitational radius defining an natural cut - off between horizon and shell depending on external observer . from this model
we can obtain a thermal and no divergent entanglement entropy that could explain @xmath0 .
+ + + * acknowledgements * + + we are indebted to werner israel , our collaborator of much of the work presented here , for many contributions , stimulating discussions and helpful comments on the manuscript .
specific lagrangian @xmath92 is given by @xmath124 where @xmath125 @xmath135 = \sum_{{\omega } , { \epsilon}}{\epsilon}\ , { \frac{1}{2}}\ , |{\omega}|\left({b_{\omega}}^{({\epsilon})\dag}\ , { b_{\omega}}^{({\epsilon } ) } + { b_{\omega}}^{({\epsilon})}\ , { b_{\omega}}^{({\epsilon})\dag}\right)\ , , \ ] ] @xmath137 = \sum_{{\omega } , { \epsilon}}{\epsilon}\ , { \frac{1}{2}}\ , |{\omega}|\left({a_{\omega}}^{({\epsilon})\dag}\ , { a_{\omega}}^{({\epsilon } ) } + { a_{\omega}}^{({\epsilon})}\ , { a_{\omega}}^{({\epsilon})\dag}\right).\ ] ] w. israel , phys .
164 * , ( 1967 ) 1776 .
b. carter , phys .
* 26 * , ( 1971 ) 331 .
d. c. robinson , phys .
lett . * 34 * , ( 1975 ) 905 . s. w. hawking , commun .
phys . * 25 * , ( 1972 ) 152 . s. w. hawking , commun .
phys . * 43 * , ( 1975 ) 199 .
g. t hooft , nucl .
b * 256 * , ( 1985 ) 727 .
w. israel , _ _ phys .
lett . a__**57 * * , 107 ( 1976 ) l. bombelli , r. k. koul , j. lee , and r. d. sorkin , _ _ phys .
rev . d__**34 * * , 373 ( 1986 ) s. mukohyama , w. israel , phys .
d * 58 * , ( 1998 ) 104005 .
g. w. gibbons and s. w. hawking , phys .
d * 15 * , ( 1977 ) 2752 .
v. p. frolov , d. v. fursaev , _ class .
quantum grav . _ * 15 * , 2041 ( 1998 ) d. v. fursaev , phys
nucl . , * 36 * , ( 2005 ) 81 g. t. horowitz , _ black holes and relativistic stars , _ part ii , 12 , 241 , edited by robert m. wald .
the univesity of chicago press , ( 1998 ) s. w. hawking , _ black holes and relativistic stars , _ part ii , 11 , 221 , edited by robert m. wald .
the univesity of chicago press , ( 1998 ) v. frolov and i. novikov , phys .
d * 48 * , 4545 ( 1993 ) hep - th/9802085 .
arenas , j.m .
tejeiro , _
nuovo cimento _ * 125b n.10 * , 1223 ( 2010 ) j.r .
oppenheimer and h. snyder , phys . rev . * 56 * , ( 1939 ) 455 .
w. israel , _
nuovo cimento vol xliv b _ * n.1 * , ( 1966 ) p. musgrave and k. lake , junctions and thin shells in general relativity using computer algebra , gr - qc/9510052v3 ( 1995 ) e. winstanley : phys .
d63 , 084013(2001 ) hep - th/0011176 f pretorius , d. vollick , w. israel , phys . rev . d**57 * * , 6311 ( 1998 ) . j. d. bekenstein , in _ proceedings of the seventh marcel grossmann meeting on general relativity , stanford , usa , 1994 _ , edited by r. t. jantzen and g. m. keiser ( world scientific , singapore , 1996 ) , p. 39 , gr - qc/9409015 . y. takahashi and h. umezawa , collective phenomena * 2 * , 55 ( 1975 ) .
h. umezawa , _ advanced field theory _
( aip press , new york , 1993 ) .
r. m. wald , phys .
d * 17 * , 1477 ( 1978 ) .
s. w. hawking , in _ general relativity : an einstein centenary survey _ , edited by s. w. hawking and w. israel ( cambridge university press , cambridge , 1979 ) , p. 746 .
w. israel , in _ current trends in relativistic astrophysics _ , edited by l. fernndez ,
l. m. gonzlez ( springer lecture notes in physics , 2003 ) , lnp 617 , 15 ( 2003 ) . | we introduce the concept of black shell , consisting on a massive thin spherical shell contracting toward its gravitational radius from the point of view of an external observer far from the shell , in order to effectively model the gravitational collapse . considering complementary description of entanglement entropy of a black shell and according to gibbons - hawking euclidean approach
, we calculate the bekenstein - hawking entropy retrieving horizon integral and discarding boundary at infinity .
+ key words : black hole entropy , entanglement entropy , black hole thermodynamics .
+ pacs numbers : 04.70.dy |
the cross sections for producing lepton pairs by drell - yan ( dy ) process have been proven to still fulfill the factorization property and are finite to first orders in perturbation theory at sufficiently large transverse momenta , @xmath1 . on the other hand , there is an extensive program of research treating the low @xmath1 region as the conventional factorization approaches give divergent results at @xmath2 ( see ref . @xcite and references therein ) .
in particular , in ref . @xcite it was shown that differential cross section in the region @xmath3 is driven by subprocesses initiated by incident gluons and therefore massive lepton - pair differential cross sections are useful sources of constraints on the gluon density .
that study using next - to - leading order qcd and @xmath1-resummation was generalized in ref .
@xcite to polarized scattering and in ref .
@xcite to electroweak gauge boson production .
the study of dy cross section with dileptons carrying large values of transverse momentum has long history and it is related to deep - inelastic - lepton scattering ( dis ) , prompt photon production and jet production as an important probe of short - distance hadron dynamics . besides helping to constraint the parton distribution functions ( pdfs ) in the nucleons ,
the great appeal of the production of dileptons in dy process in nuclear targets is that they are colorless probes of the dynamics of quarks and gluons @xcite .
namely , they escape unscathed through the colored medium of the high - energy collision .
thus , the dileptons can be a powerful probe of the initial state of matter created in heavy ion collisions , since they interact with the medium only electromagnetically and therefore provide a baseline for the interpretation of jet - quenching models .
along these lines , it has been shown @xcite that those electromagnetic probes are crucial to determine the dominant physics in the forward region at rhic and at the lhc .
it has been shown @xcite that both direct ( prompt ) photon production and drell - yan dilepton pair production processes can be described within the same color dipole approach without any free parameters .
such a formalism , developed in @xcite for the case of the total and diffractive cross sections , can be also applied to radiation @xcite . in the rest frame of a target
, the dy process looks like a bremsstrahlung @xcite of a massive photon from an incoming quark .
the photons can be emitted before or after a quark to be scattered on a proton .
although in the process of electromagnetic bremsstrahlung by a quark no real quark dipole participates , the cross section can be expressed through the more elementary cross section @xmath4 of interaction of a @xmath5 dipole @xcite .
the relation between this formalism and the usual collinear pqcd factorization has been studied in details in ref .
the dipole formalism offers an easy way to calculate the transverse momentum distribution in dy process even in the low-@xmath1 region .
the corresponding phenomenology investigating the role played by high energy approaches in the dy @xmath1 distribution has been investigated long before in ref .
results from order @xmath6 in the parton model can not be directly compared to such approach since it is not an expansion in any parameter .
all contribution from higher orders graphs enhanced by a factor @xmath7 and even non - perturbative corrections are included .
for instance , in ref .
@xcite a twist expansion in powers of @xmath8 ( @xmath9 is the saturation scale and @xmath10 is the invariant dilepton mass ) was developed .
it was shown that the leading twist is a good approximation to the full result for masses @xmath11 gev .
recently , the diffractive dy cross section has been investigated in detail within the dipole framework @xcite .
the corresponding factorization breaking effects in diffractive dy lead to very distinct properties of the observables compared to qcd factorisation - based calculations .
our goal is to investigate in detail the low mass dy cross section at the lhc energies using color dipole approach , discussing several phenomenological aspects .
we focus mainly on forward rapidities and at the energy available at the lhc .
the paper is organized as follows . in next section
we summarize the main formula for dy differential cross section within the dipole framework , which is suitable for forward rapidities and allows to incorporate parton saturation effects . in last section
we show our numerical results and predictions and summarize our main conclusions .
this section resumes the theoretical treatment for drell - yan production in @xmath12 collisions in high energies considered in our study
. we will use the dipole approach , which is well suited for high - energy processes , i.e. small parton momentum fraction in the target @xmath13 , and its range of validity is expected to be near @xmath14 .
the low mass dy production surely probes the small-@xmath15 physics , specially the forward rapidity case .
one advantage in such an approach is to describe simultaneously the direct photon and dilepton production in the same theoretical framework and it provides finite cross sections in the limit @xmath2 .
the transverse momentum @xmath16 distribution of virtual photon bremsstrahlung in quark - nucleon interactions , integrated over the final quark transverse momentum , was derived in @xcite in terms of the dipole formalism , @xmath17 , \label{di}\end{aligned}\ ] ] where @xmath18 and @xmath19 are the quark - photon transverse separations in the two radiation amplitudes contributing to the cross section , @xmath4 .
the parameter @xmath20 is the relative fraction of the quark momentum carried by the photon , and is the same in both amplitudes , since the interaction does not change the sharing of longitudinal momentum . in the equation above
, @xmath21 stands for transverse and @xmath22 for longitudinal photons .
the energy dependence of the dipole cross section , which comes through the variable @xmath23 ( with @xmath24 ) , where @xmath25 is the projectile four - momentum and @xmath26 is the four - momentum of the dilepton , is generated by additional radiation of gluons which can be resummed in the leading @xmath27 approximation . in eq .
( [ di ] ) the light - cone wavefunction of the projectile quark @xmath28 fluctuation has been decomposed into transverse @xmath29 and longitudinal @xmath30 components , and an average over the initial quark polarization and sum over all final polarization states of quark and photon is performed .
the expressions for @xmath21 and @xmath22 wavefunction components are well known at the lowest order @xcite .
the hadron cross section can be obtained from the elementary partonic cross section eq .
( [ di ] ) summing up the contributions from quarks and antiquarks weighted with the corresponding parton distribution functions ( pdfs ) , @xmath31 where the pdfs of the projectile have entered in a combination which can be written in terms of proton structure function @xmath32 , with @xmath33 .
for the hard scale @xmath34 entering into the proton structure function in eq .
( [ con ] ) , we take @xmath35 $ ] and the energy scale of the dipole cross section in eq .
( [ di ] ) is given by the variable @xmath36 .
the dependence of the cross section on the choice for the hard scale can be obtained by varying the @xmath37 value ( the default value here is @xmath38 ) .
the quantity @xmath39 takes into account the effective higher - order dy contributions . for simplicity
we take the expression @xcite : @xmath40 where the running coupling constant is computed at the scale @xmath41 .
an important piece in the color dipole calculations is the dipole cross section .
it is theoretically unknown , although several parameterizations have been proposed . here , we consider some analytical parameterizations which rely on the geometric scaling property . in this case , they are a function of a scaling variable @xmath42 where @xmath43 is the so called saturation scale .
it defines the transverse momentum scale where parton recombination physics is relevant and in general is modeled as @xmath44 .
a common feature on these models is that for decreasing @xmath15 , the dipole cross section saturates for smaller dipole sizes .
in addition , at small @xmath45 , as perturbative qcd implies @xmath46 , they vanish , i.e. the color transparency phenomenon occurs . in a general form
, they can be parameterized as @xcite , @xmath47,\label{gbw}\ ] ] where the quantity @xmath48 is the effective anomalous dimension .
the gbw parameterization @xcite uses @xmath49 and the remaining parameters are fitted to dis hera data at small @xmath15 .
the saturation scale is defined as @xmath50 .
this parameterization gives a quite good description of dis data at @xmath51 .
the main difference among the distinct phenomenological models using parameterizations as eq .
( [ gbw ] ) comes from the predicted behavior for the anomalous dimension , which determines the transition from the non - linear to the extended geometric scaling regime , as well as from the extended geometric scaling to the dglap regime .
it is the behavior of @xmath52 that determines the diminishing of the hadronic cross section as @xmath1 increases .
several models in the literature consider the general form @xmath53 , where @xmath54 is the anomalous dimension at the saturation scale and @xmath55 mimics the onset of the geometric scaling region and dglap regime . in order to take into account this possibility , here we also will consider the phenomenological saturation model proposed in ref .
@xcite which encodes the main properties of the saturation approaches , with the dipole cross section parameterized as follows @xmath56\ , , & \mbox{for $ \bar{\tau } > 2$}\ , , \end{array } \right .
\label{cgcfit}\end{aligned}\ ] ] where @xmath57 and the expression for @xmath58 ( saturation region ) has the correct functional form , as obtained from the theory of the color glass condensate ( cgc ) @xcite .
for the color transparency region near saturation border ( @xmath59 ) , the behavior is driven by the effective anomalous dimension @xmath60 , where @xmath61 is the lo bfkl anomalous dimension at saturation limit . before computing numerically the cross section given by eq .
( [ con ] ) , we discuss about the semi - analytical calculation which is allowed in color dipole picture in the color transparency region . in that case ,
an expression for @xmath1 distribution can be written using eq .
( [ con ] ) and the expressions for the transverse momentum @xmath16 distribution of photon bremsstrahlung in quark - nucleon interactions , eq .
( [ di ] ) . the explicit equation for the dy differential cross section , eq .
( [ con ] ) , reads as : @xmath62\left[\frac{{\cal i}_1}{(p_t^2+\varepsilon^2)}-\frac { { \cal i}_2}{4\varepsilon } \right ] \right .
\nonumber \\ & + & \left .
[ 1+(1-\alpha)^2]\left [ \frac{\varepsilon p_t \ , { \cal i}_3}{(p_t^2+\varepsilon^2 ) } -\frac{{\cal i}_1}{2}+\frac{\varepsilon \,{\cal i}_2}{4}\right ] \right \},\end{aligned}\ ] ] where @xmath63 .
the quantities @xmath64 are hankel s integral transforms of order @xmath65 ( @xmath66 ) and order @xmath67 ( @xmath68 ) given by : @xmath69 here , @xmath70 denotes the modified bessel function of the second kind and @xmath71 is an effective quark mass . considering a kinematic interval where the dipole cross section is dominated by the color transparency region
, one can use gbw parameterization and take its small-@xmath45 limit to compute analytically the integrals in eq .
( [ ints ] ) . in this case
, we can take the approximation @xmath72 in the region where @xmath73 . at the lhc energies ,
the typical saturation scale is of units of gev .
for instance , at mid - rapidity @xmath74 , which gives for @xmath75 gev a value @xmath76 at 7 tev .
then , the saturation scale is of order @xmath77 gev ( with @xmath78 and @xmath79 ) .
the final results for the hadron cross section is then , @xmath80\left [ \frac{p_t^2}{(p_t^2+\varepsilon^2)^4}\right ] \right.\nonumber \\ & + & \left .
[ 1+(1-\alpha)^2]\left [ \frac{p_t^4+\varepsilon^4}{2(p_t^2+\varepsilon^2)^4}\right ] \right \ } \label{analitic}\end{aligned}\ ] ] in what follows we compute the differential cross sections of low mass dy production in hadron - hadron collisions focusing in the recent measurements done by atlas @xcite and lhcb @xcite collaborations at the lhc .
, as a function of dilepton transverse momentum , @xmath1 , at energy of @xmath0 tev .
the plots are shown for fixed dilepton mass ( @xmath81 and @xmath82 gev ) and distinct lepton pair rapidities ( @xmath83 ) .
the results are presented using the gbw dipole cross section ( bold curves ) and the cgc dipole cross section ( thin curves ) . ]
let us now present some numerical calculations concerning the low mass dilepton production in the lhc energy regime of @xmath0 tev . for the proton structure function in eqs .
( [ con ] ) we have taken the allm parameterization @xcite , which is valid in the kinematic range we are interested in . the sensitivity to a different choice for @xmath84 is very small .
moreover , in order to account for the threshold region @xmath85 , we have corrected the dipole cross section by multiplying it by a threshold factor @xmath86 .
we consider @xmath87 gev for the effective quark mass . in fig .
[ fig:1 ] , we show the results for the differential cross section , @xmath88 ( in units of pb ) , as a function of the dilepton transverse momentum @xmath1 . here , the predictions are obtained using the gbw dipole cross section ( bold curves ) and the cgc dipole cross section ( thin curves ) and using the hard scale @xmath89 .
notice that the @xmath1-spectrum is quite sensitive to the particular model of dipole cross section ( specially at large transverse momentum ) as it depends on the behavior of effective anomalous dimension as discussed in the previous section . in the left panel
is shown the case for fixed invariant mass @xmath81 gev and for sample values of dilepton rapidity including central and forward rapidities , i.e. @xmath90 and 4 , respectively .
the same notation holds for the right panel , where now the invariant mass is @xmath91 gev .
as expected , the large rapidity cases give smaller cross sections and the peak on the distributions is shifted to larger values of transverse momentum . in the kinematical situation investigated here the peak
is located at momentum around @xmath92 gev . the shift and location of the peak
can be understood looking at the semi - analytical expression , eq .
( [ analitic ] ) . a clear problem in the numerical calculations using the full @xmath1 spectrum is due to the strongly oscillating integrand appearing in eqs .
( [ ints ] ) , which turn out the @xmath1 integrations a delicate task mainly for high mass values .
the imposed cuts at energy of @xmath0 tev are lepton pair rapidities @xmath93 and dilepton transverse momentum @xmath94 gev .
preliminary atlas data @xcite are shown for sake of comparison . ] in fig .
[ fig:2 ] we show the invariant mass distribution at midrapidities obtained from two different implementations of the dipole cross section taken from recent phenomenological works .
we considered the gbw model ( dot - dashed line ) and the phenomenological saturation model , labeled here cgc ( solid line ) , which involves a running anomalous dimension .
the main deviation between these two models occurs at large @xmath1 , which gives distinct overall normalizations for the dilepton invariant mass distribution .
the considered cuts are along the lines presented by the atlas analysis @xcite for low mass drell - yan di - muon process .
the selection cuts on that analysis at energy of @xmath0 tev and integrated luminosity of 36 pb@xmath95 were low muon transverse momentum , @xmath96 gev and low di - muon mass region @xmath97 gev . here , we consider the integration over the boson rapidity in the range @xmath93 and dilepton transverse momentum @xmath94 gev .
distinct @xmath1 cuts will lead to a different overall normalization for the invariant mass distribution . at this stage
we did not impose the selected cuts on individual muons as done by atlas analysis .
two different studies were done by atlas @xcite : selection of muons of different minimum @xmath98 and no requirement on boson selection except for the invariant mass constraint ( asymmetric analysis ) and selection of minimum @xmath98 for muons and a requeriment on the rapidity of boson is applied ( symmetric analysis ) .
the results presented here are somewhat consistent with the extrapolated born level differential cross section using the symmetric analysis . for sake of comparison
, we include the preliminary data @xcite in fig .
[ fig:2 ] ( filled circles ) .
the main @xmath99 dependence of dy cross section can be quantitatively understood in the color dipole framework . in the case of integrated cross section on @xmath1
, it was shown in ref .
@xcite that a twist expansion in positive powers of the ratio @xmath8 can be performed .
the leading twist contribution reads as @xcite : @xmath100\right\ } , \nonumber \label{golec}\end{aligned}\ ] ] modulo contributions of @xmath101 .
the exact leading twist formula is quite close to the all - twist result at the lhc energies in the region of @xmath102 gev @xcite . in central rapidities considered above , @xmath103 , the saturation scale changes from @xmath104 gev@xmath105 down to @xmath106 gev@xmath105 in the invariant mass range @xmath97 gev , respectively .
in this situation , the analytical result given in eq .
( [ analitic ] ) should be quite reliable
. it would be timely in the future to investigate the effects of qcd dglap evolution in the dipole cross section and also the impact parameter dependence in the cgc model as done recently in refs . @xcite and @xcite .
tev imposing the cut on dimuon transverse momentum @xmath107 gev and two invariant mass regions : ( upper plot ) @xmath108 gev and ( lower plot ) @xmath109 gev . ]
now , we focus on the forward rapidity region . in fig .
[ fig:3 ] , the rapidity distribution , @xmath110 , is computed for the interval @xmath111 considering the two phenomenological models ( same notation as the previous plot ) and using the hard scale @xmath112 $ ] .
it was found that the rapidity distribution is sensitive to the chosen hard scale as also occurs for the lo pqcd approach . here
, the deviations comparing the models are stronger than for the mass distribution .
the main point is that the rapidity distribution is driven by the effective anomalous dimension , which is distinct in the models .
we have imposed a cut for the dilepton transverse momentum of @xmath107 gev and two distinct intervals of invariant mass .
namely , in the upper plot one has @xmath108 gev whereas in the lower plot one has @xmath109 gev .
the motivation for such a cut is due to the recent lhcb collaboration @xcite measurement of low mass dy cross section .
the measurement collected with an integrated luminosity of 37 pb@xmath95 are for the di - muon final state having muons within pseudorapidities of 2 to 4.5 , muon transverse momentum @xmath113 gev ( @xmath114 gev for higher masses ) in two distinct mass regions . in the forward rapidities considered here ,
the saturation scale is in the interval @xmath115 gev@xmath105 for @xmath116 gev .
slightly lower values are found also for higher mass @xmath117 gev .
finally , we check the energy dependence of the dy differential cross section within the color dipole picture . in fig .
[ fig:4 ] ( upper panel ) we show the invariant cross section as a function of @xmath1 at energy @xmath118 gev .
the experimental results from the e866 collaboration @xcite are also presented ( @xmath119 and @xmath120 gev ) . in the bottom panel ,
the differential cross section @xmath121 ( for @xmath122 ) is shown for the energy @xmath123 gev as a function of dilepton invariant mass . the data from cdf collaboration @xcite are included in the plot , considering also the large invariant mass data points .
the solid curves refer to cgc and dot - dashed curves to gbw dipole cross section , respectively .
the color dipole picture reasonably describes the cross section from low to high energies in the kinematical regions where it is expected to be valid .
the approach is also somewhat consistent with calculations carried out in next - to - leading order qcd at both fixed target and collider energies @xcite .
gev as a function of @xmath1 compared to the e866 collaboration data @xcite . in bottom panel ,
the differential cross section @xmath124 at @xmath123 gev as a function of invariant mass is presented and compared to the cdf collaboration data @xcite.,title="fig : " ] gev as a function of @xmath1 compared to the e866 collaboration data @xcite . in bottom panel ,
the differential cross section @xmath124 at @xmath123 gev as a function of invariant mass is presented and compared to the cdf collaboration data @xcite.,title="fig : " ] as a summary , we showed that low mass dy production can be addressed in the color dipole picture without any free parameters by using dipole cross section determined from current phenomenology in dis .
it has been shown before @xcite that in central rapidities at rhic and tevatron , saturation effects do not play a significant role for the measured range of @xmath16 .
this situation can be changed at the lhc even at midrapidities as the saturation scale is enhanced by a sizable factor .
we write down also analytical results relevant for the calculation of @xmath1 spectrum of dy dileptons .
this work was partially financed by the brazilian funding agencies cnpq and fapergs and by the french - brazilian scientific cooperation project capes - cofecub 744/12 .
mvtm thanks murilo rangel , ronan mcnulty and michael klasen for helpful comments and mbgd thanks m.a .
betemps for useful discussions .
kopeliovich , proc . of the workshop hirschegg 95 : dynamical properties of hadrons in nuclear matter , hirschegg january 16 - 21 , 1995 , ed . by h. feldmeyer and w. nrenberg , darmstadt , 1995 , p. 102
( hep - ph/9609385 ) .
d. gabbert and l. de nardo , proceedings of 15th international workshop on deep - inelastic scattering and related subjects .
vol 1 and 2 , pages 373 - 376 : munich , germany , april 16 - 20 ( 2007 ) .
g. grindhammer and k. sachs . arxiv:0708.3196
[ hep - ph ] ; h. abramowicz , e.m .
levin , a. levy and u. maor , phys .
b * 269 * , 465 ( 1991 ) . | we present results for the low mass drell - yan production in proton - proton collisions at the lhc in the color dipole formalism .
the dy differential cross sections at @xmath0 tev as a function of dilepton rapidity , transverse momentum and invariant mass are discussed .
we have imposed kinematical cuts related to the low mass dy production investigated by atlas and lhcb collaborations . |
the extragalactic background light ( ebl ) supplies opacity for propagating high energy gev - tev photons via an electron - positron pair production ( @xmath7 ) @xcite . determining the transparency of the universe is of fundamental importance for a wide variety of current observatories such as the space - borne _ fermi_/lat instrument operating at energies @xmath8 gev to ground - based @xmath0-ray telescopes probing energies @xmath91 tev .
the distance at which the optical depth due to this interaction is @xmath10 defines a horizon of the observable universe at @xmath0-ray energies , and has been a subject of extensive efforts designed to model the build - up of ebl with time from the posited emission history of galaxy populations . in this letter
we show that , with observed galaxy populations over a wide range of wavelengths , one can uniquely reconstruct the optical depth of the universe at these energies out to redshifts @xmath11 .
this empirical reconstruction relies exclusively on _ data _ from an extensive library of galaxy luminosity functions ( lfs ) encompassing 18 finely sampled wavelengths from uv to mid - ir ( 0.15 - 24 ) relevant for the pair - production opacity .
this methodology enables robust calculation of the @xmath0-ray opacity in the _
fermi_/lat energy range using galaxy surveys probing @xmath12 out to @xmath13 . extending to tev energies , probed by the ground - based cherenkov observations , we use measurements out to 24
; this extrapolation is robust for the redshifts currently probed these observations .
this heuristic reconstruction using the _ observed _ galaxy populations defines the absolute floor of the photon - photon optical depth due to known galaxy populations and deviations from it would allow the characterization of any emissions inaccessible to direct telescopic studies @xcite .
we use the methodology developed in @xcite of reconstructing the ebl from observed galaxy populations in a compilation of measured lfs covering the uv , optical and near - ir ; that compilation is slightly updated compared to table 1 of @xcite .
the wealth of galaxy survey data has recently reached adequate redshift coverage to make such empirical estimation of the evolving ebl feasible and the reconstruction was shown to reproduce independent data from galaxy counts and the cosmic infrared background ( cib , * ? ? ? * ) .
our approach is completely independent of theoretical modeling describing the evolution of galaxy populations in that we use the lf data directly observed at all wavelengths out to @xmath14 in this heuristic reconstruction from which we derive the optical depth due to pair production ( see also * ? ? ?
standard cosmological parameters are used below : @xmath15 , @xmath16 , @xmath17 .
the solid curves are fits through the data points according to our fitting function ( equation [ eqn : ldfit ] ) with best - fit parameters displayed in each panel as ( @xmath18,@xmath19,@xmath20 ) where @xmath21 is also in @xmath22 .
all data points have been converted to units with @xmath23 .
the references for the data points can be found in table 1 of @xcite with the addition of uv data from @xcite .
, title="fig:",scaledwidth=45.0% ] quantifying the optical depth of the universe to high energy photons requires knowledge of the properties of the intervening ebl .
evolving galaxy populations compose the bulk of the ebl which is dominated by starlight in the uv / optical and thermally radiating dust at longer ir wavelengths .
we use multiwavelength survey data to fit the evolution of a single derived quantity , the luminosity density , in this otherwise assumption - free approach .
we cover the whole range from uv to mid - ir ( 0.15 - 25 ) , providing for the first time an empirical derivation of the @xmath2 optical depth out to several tev . @xcite
have used a similar approach to reconstruct the ebl at @xmath24 .
the survey data used in this letter extends the collection of lfs presented in @xcite with expanded coverage in the uv @xcite and out to mid - ir wavelengths @xcite .
this library now contains measured lfs and allows us to reconstruct the evolving ebl and its spectrum in a finely sampled wavelength grid encompassing 0.15 - 24 out to @xmath25 .
@xcite used this data to accurately recover the observed galaxy number counts in the 0.45 - 4.5 range .
this therefore assumes only the existence of populations now observed out to @xmath25 and magnitudes as faint as @xmath26 ; additional populations , such as the hypothetical galaxies with first stars would then exist at still earlier times and have much fainter fluxes . for infrared wavelengths , 1.6 - 24 .
the references for the data points can be found in table 1 of @xcite with the addition of mid - ir data from @xcite .
, scaledwidth=45.0% ] in the rest - frame uv to near - ir , the lf is well described by the conventional schechter parameterization whereas at mid - ir wavelengths , the lf seems to be better described by a broken power - law or a double - exponential profile .
regardless of the functional form , the lfs can be integrated to give the comoving volume emissivity ( we refer to this as the luminosity density ) in the given rest - frame band @xmath27 where @xmath28 is the measured lf .
each data point in figures [ fig_ld1 ] and [ fig_ld2 ] represents the luminosity density given by the authors along with 1@xmath29 error bars ( for references , see table 1 in @xcite ) . in the cases where
this value is not given explicitly in the original papers , we have integrated the best - fit parameterized lf to obtain @xmath30 and have estimated the errors from the distribution of all of the values of @xmath30 allowed within the 1@xmath29 solutions of the individual fit parameters . however , mutual comparison of uncertainties among the many different studies is not very meaningful since some authors include various effects in addition to the statistical errors from the method of lf estimation , such as cosmic variance , @xmath31-corrections , incompleteness and photometric system . here
, we have chosen to maximize our wavelength and redshift coverage by letting all available measurements contribute to our fitted evolution regardless of the error treatment . in the cases where
the median redshift of the sample is not explicitly given , we have placed the measurements at the midpoint of the redshift bin .
our wavelength interval is sampled at the rest - frame bands shown in the panels of figure [ fig_ld1 ] and [ fig_ld2 ] where most of the lfs have been measured .
the offsets from these defined wavelengths due to filter variations ( e.g. sdss @xmath32 and johnson @xmath33 ) are small enough to be neglected . motivated by the fitting formulae in @xcite , which we found to be reliable over a wide range of wavelengths , we consider the following three parameter fits for the evolution of the luminosity density @xmath34 where we fix @xmath35 . although not restricted to schechter lfs , this functional form for @xmath36 is nevertheless equivalent to the underlying schechter parameters evolving as @xmath37 and @xmath38 with a constant faint - end slope .
interpolating the rest - frame data between our 18 reference bands defines the rest - frame emissivity spectrum at any given epoch .
we apply a cutoff to the spectrum above the lyman limit , @xmath39 , corresponding to efficient absorption of ionizing photons by hydrogen in the local environments .
at all lower energies , the universe is assumed to be completely transparent to background photons .
the integrated light from galaxies seen today is ( e.g. peebles 1971 ) @xmath40 where @xmath41 is the rest - frame frequency and @xmath42 in @xcite we considered two limiting cases for the evolution of the faint - end of the lf and showed that the distribution of galaxies from lf data , when projected onto the sky , accurately recovered the observed galaxy counts across the optical and near - infrared .
the flux from the integrated counts is displayed as shaded regions in the lower panel of figure [ fig_nebl ] along with our empirically determined ebl ( solid line ) which is in good agreement with integrated counts data in the literature apart from wavelengths @xmath43 where the steep evolution of the mid - ir lfs ( 8 - 24 ) causes our ebl to be a factor of @xmath442 - 3 higher than integrated counts from @xcite , @xcite , @xcite and @xcite .
although these authors do not claim to fully resolve the cib at these wavelengths , the discrepancy is large enough to indicate a mismatch between number counts and mid - ir lf measurements at @xmath45 .
this issue is apparently also encountered in galaxy evolution models ; @xcite are not able to simultaneously account for the integrated counts and the bright - end of the observed lf in the 8 - 24 range using different dust templates .
in fact , this is the wavelength regime where varying degrees of dust contribution and pah emission make the spectrum less predictable .
recent upper limits derived from a tev source spectra also favor low levels of cib at these wavelengths @xcite . at this stage , one must therefore question the robustness of ebl reconstructed from lfs at @xmath46 .
the uv / blue end of the ebl turns out to be sensitive to the abundance of photons with energies just below the lyman limit , 13.6ev .
the redshifted far - uv contribution dominates the ebl below 0.5 due to the steep increase of the star formation rate at earlier times . for all galaxy types ,
there is considerable absorption in the lyman series which we do not account for and we illustrate this dependence by considering the case where lyman - series absorption completely suppresses the spectrum above @xmath47 ( instead of @xmath48 ) ; shown as gray lines in figure [ fig_nebl ] ( lower panel ) .
we subsequently display our optical depths for the both cases which bracket the true behavior .
the relevant quantity for computing the optical depth due to photon - photon interaction is the rest - frame number density of photons as a function of time and energy , @xmath49 .
we shall refer to the energy of a photon belonging to the ebl as @xmath50 and we use @xmath51 for the propagating @xmath0-rays .
rest - frame quantities are denoted with a prime . at any given epoch , the photon number density ( in proper coordinates ) is composed both of sources emitting in the rest - frame as well as the accumulated emission from earlier times @xmath52 where @xmath53 is the planck constant ( the extra @xmath53 is to convert @xmath30 to per unit energy , @xmath54 ) and @xmath55 .
the condition for pair production is that the total energy in the center - of - mass frame must satisfy @xmath56 where @xmath57 is the angle of incidence .
this means that in order to interact with a @xmath0-ray of energy @xmath58 , background photons must have wavelengths of @xmath59 .
the cross section for this interaction is @xmath60,\ ] ] where @xmath61 for the most likely angle of incidence , @xmath62 ( side - on ) , the probability for interaction is maximized at roughly four times the minimum threshold energy , @xmath63 .
the optical depth encountered by a high energy photon originating at @xmath64 ( @xmath65 ) can be expressed in terms of its observed energy , @xmath51 , as @xmath66 where @xmath49 comes from equation [ eq_n ] . in figure [ fig_tau ] we display the calculated optical depths as a function of observed energy for @xmath0-rays originating at several redshifts .
the optical depth roughly traces the shape of the number density of ebl photons with a sharp drop in the optical depth at the lowest energies due to the cutoff at the lyman limit .
we also show the regions encompassed by the two scenarios of the faint - end evolution from the reconstruction of @xcite .
optical depth contributed by known galaxy populations assuming @xmath67 ( gray curve in the lower panel of figure [ fig_nebl ] ) .
the curves are not drawn beyond the energy of @xmath68 as we do not consider data at @xmath69 .
the shaded regions show the boundaries of the upper / lower scenarios in the empirical reconstruction of @xcite out to 4.5 .
the dotted vertical line shows roughly the highest energy probed by _
fermi_/lat and the dashed line shows where @xmath70 for reference . , scaledwidth=45.0% ]
what do these reconstructed opacities imply for observations of high - energy sources with the current instruments ?
blazars and gamma - ray bursts ( grbs ) are examples of high energy extragalactic sources whose spectra is affected by attenuation of photons in excess of @xmath71 .
two types of datasets are relevant for this discussion : space - borne _ fermi_/lat measurements at @xmath72gev and ground - based telescopic measurements extending out to @xmath44tev energies . because extragalactic @xmath0-ray absorption increases with both with redshift and energy , the ebl can be constrained based on the highest energy photons observed from a source provided the redshift
is known @xcite .
grbs have the advantages of being observable across great distances and typically displaying harder spectra than most blazars at sub - gev energies . figure [ fig_nebl ] shows that for _ fermi_-observed sources it is sufficient to use data out to @xmath73 ( red symbols ) , whereas for tev range observations , survey data are needed out to the longer wavelengths ( green symbols ) . figure [ fig_tau ] shows the reconstructed optical depth explicitly confirming this .
we now briefly discuss the implications : * fermi / lat * detects blazars and grbs out to energies @xmath74 gev . in figure [ fig_fermi ] ( upper panel )
we show curves of constant @xmath2 optical depth in @xmath75 space and compare with the most constraining high energy _
fermi_/lat sources with known redshifts ( taken from * ? ? ? * ) .
the contours of @xmath76 correspond to probabilities of photon being absorbed by the ebl of 63% , 86% , 95% , 99.3% , respectively . in the absence of new populations , the universe remains fairly transparent at the _
fermi_/lat energies out to @xmath77 .
our reconstructed ebl is fully consistent with all the available lat data and , in fact , allows for non - negligible extra levels of the cib from new populations such as possibly have existed at higher @xmath64 . as the fermi mission progresses and figure [ fig_fermi ] ( upper ) becomes more populated at the highest energies , sources at high-@xmath64 will provide better constraints for the optical / nir ebl .
originating at @xmath64 , encounters exactly @xmath78 ( solid , dotted , dashed , dash - dotted respectively ) .
black and gray correspond to the cases of the rest - frame spectrum cutoff energy is at 10.2 and 13.6ev respectively .
the symbols show the highest energy photon observed in grbs ( blue squares ) and a selection of the most constraining of _ fermi_/lat blazars ( orange circles ) @xcite .
lower : the observed energy spectrum of the sources 1es 1101 - 232 ( @xmath79 ; _ left _ ) and 3c 279 ( @xmath80 ; _ right _ ) shown with black diamonds , and the corresponding deabsorbed data as blue squares .
the deabsorbed spectra have best - fit photon indices @xmath81 and @xmath82 respectively , but deviate substantially from a power - law at the highest energy bins . ,
scaledwidth=45.0% ] * ground - based cherenkov telescopes * have produced good quality spectra for tev - blazars , although for sources at significantly lower redshifts than _ fermi_/lat . if the evolving ebl is known to a good accuracy ,
one can deabsorb observed blazar spectra to reveal the intrinsic component , which is expected to have a power - law form , @xmath83 .
the lower panels in figure [ fig_fermi ] demonstrate how our reconstructed ebl affects the spectrum of two known blazars at relatively high redshifts , both of which have been used to place upper limits on the optical / nir ebl .
good quality spectrum of the bl lac object 1es 1101 - 232 ( @xmath79 ) has been obtained by hess in the energy range @xmath84 which at @xmath79 interacts most strongly with optical and near - ir background photons @xcite .
the observed spectrum is relatively hard ( @xmath85 ) and results in a best - fit intrinsic photon index of @xmath86 after deabsorption . for this particular source ,
the upturn at tev energies is largely driven by the ebl photons at @xmath87 and would be less pronounced if our ebl reflected the integrated counts data in the lower panel of figure [ fig_nebl ] . because the ebl changes with time in both shape and amplitude , the effects on @xmath0-ray absorption become even more prominent for more distant sources .
the spectrum of the distant radio quasar 3c 279 has been captured by magic during different flaring events @xcite .
the deabsorbed spectrum of 3c 279 shown in figure [ fig_fermi ] ( lower ) also deviates substantially from a simple power - law in the highest energy bins which is unlikely to be due to our near - ir background being overestimated .
this behavior of the deabsorbed 3c 279 spectrum has been pointed out by @xcite who suggest either improved emission models or instrumental systematic uncertainties as potential solutions .
another possibility is that some fundamental effects are missing such as secondary @xmath0-rays produced along the line of sight by cosmic rays accelerated by the blazar jet @xcite . in absence of such secondary effects
however , the universe should be completely opaque for tev sources at @xmath88 .
upper limits for the ebl derived from tev spectra rely on assumptions of the hardness of the intrinsic blazar spectrum . @xcite
derive limits for the whole range of optical to far - ir ebl using an extensive source sample from both _ fermi_/lat and ground - based cherenkov telescopes .
their results allow a total integrated nir flux ( 1 - 10 ) of @xmath4420 in excess of known galaxy populations whereas there is , at most , little room for extra contribution in the mid - ir ( @xmath89 ) .
in fact , our lf - derived ebl is inconsistent with the lowest mid - ir limits of @xcite .
we have shown that it is possible to robustly reconstruct the evolving ebl in the universe using our earlier library of multiwavelength survey data now updated to extend from the uv out to the mid - ir @xcite .
this reconstruction uniquely defines the @xmath0-ray opacity out to tev energies for sources at @xmath13 and shows that at the energy bands probed by _
fermi_/lat , the universe is fairly transparent out to @xmath77 , unless unknown sources at high redshifts contribute non - negligible amounts of cib . at tev energies , probed by ground based telescopes ,
the universe becomes optically thick at @xmath90 so any such photons associated with the sources at higher redshifts would have to be of secondary origin .
our reconstructed ebl and optical depths are available upon request .
this work was supported by nasa headquarters under the nasa earth and space sciences fellowship program - grant nnx11ao05h .
kh is also grateful to _ the leifur eiriksson foundation _ for its support .
we thank w. mcconville , b. magnelli and m. ricotti for useful communications . | we reconstruct @xmath0-ray opacity of the universe out to @xmath1 using an extensive library of observed galaxy luminosity function ( lf ) surveys extending to high redshifts .
we cover the whole range from uv to mid - ir ( 0.15 - 25 ) providing for the first time a robust empirical calculation of the @xmath2 optical depth out to several tev . here
, we use the same database as @xcite where the extragalactic background light was reconstructed from lfs out to 4.5 and was shown to recover observed galaxy counts to high accuracy .
we extend our earlier library of lfs to 25 such that it covers the energy range of pair production with @xmath0-rays ( 1 ) in the entire _
fermi_/lat energy range , and ( 2 ) at higher tev energies probed by ground - based cherenkov telescopes . in the absence of significant contributions to the cosmic diffuse background from unknown populations , such as the putative population iii era sources
, the universe appears to be largely transparent to @xmath0-rays at all _ fermi_/lat energies out to @xmath3 whereas becoming opaque to tev photons already at @xmath4 and reaching @xmath5 at @xmath6 .
comparison of the currently available _
fermi_/lat gamma - ray burst and blazar data shows that there is room for significant emissions originating in the first stars era . |
highly relativistic motion , essential to overcome the compactness problem @xcite , is a basic ingredient of all grb models .
the value of the bulk lorentz factor , @xmath1 , of the relativistic outflow is of outmost interest .
it is essential for understanding the nature of the inner engine , the outflow and its acceleration and collimation mechanisms , the conditions at the emitting regions and the radiation mechanism .
so far the most robust method to estimate @xmath1 was using the compactness .
the high energy photons set an upper limit on the optical depth for pair production .
the observations of gev photons from several bursts enabled the fermi team to set very high ( @xmath2 lower limits on @xmath1 for those grbs @xcite .
while other methods to estimate or limit @xmath1 depend on various assumptions ( see e.g. @xcite ) , the compactness limit seems to be independent of any model assumptions and hence it is considered to be the most robust one .
however , recent fermi observations have also shown that the onset of higher energy ( @xmath3 mev to a few gev - denoted hereafter gev ) emission lags after the onset of the lower energy ( @xmath3 kev to a few mev - denoted hereafter mev ) prompt emission @xcite .
fermi confirmed earlier egret results that the gev emission also lasts longer than the mev emission @xcite .
these two facts suggest the possibility of a different origin for the mev and the gev emission
. this would be the case , for example , if the gev emission arises from an external shock afterglow ( kumar & barniol duran 2009 , 2010 ; ghisellini et al .
2010 ; piran & nakar 2010 ; gao et al . 2009 ) , from multi - zone internal shocks @xcite or if the mev emission is the quasi - thermal radiation of the baryonic outflow while the gev emission is mainly from the subsequent internal shocks .
the compactness limits on the lorentz factor are based , however , on an implicit assumption that the mev and the gev photon arise from the same region .
we show here that the relaxation of this assumption reduces significantly the estimated lower limit on @xmath1 ( see also * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
the existence of two regimes leads to a rich variety of possibilities .
we consider in the following the most natural configuration , which is also consistent with the temporal delay of the gev emission , the mev emission is produced at lower radii ( say via internal shocks ) and the gev emission is produced at a larger radius ( say via external shocks ) - see fig .
[ fig.geometry.1 ] .
other geometrical options that we do nt consider are that the gev emission is emitted at a lower radius than the mev emission or that the mev and the gev emission are produced at different angular regions .
our estimates do nt depend on the origin of the emission ( e.g. internal or external shock ) but just on the overall geometry of the system .
we calculate the optical depth of a gev photon passing through the mev photons shell and we obtain new compactness limits on @xmath1 .
the mev photons beamed into an angle @xmath4 .
the gev photons can reach us only if they are directed in the line of sight , which is along the x - axis . ]
consider two emitting regions denoted by @xmath5 , for mev and @xmath6 , for gev respectively ( see fig .
[ fig.geometry.1 ] ) .
the mev ( gev ) emission region has a radius @xmath7 ( @xmath8 ) ( for simplicity we consider emission from thin shells ) with @xmath9 , a lorentz factor @xmath10 ( @xmath11 ) and an angular width @xmath12 ( @xmath13 ) .
we assume that the mev and gev jets are aligned and both are pointing toward us . in this configuration
a gev photon passes through a shell of mev photons on its way to the observer ( see fig .
[ fig.geometry.2 ] ) .
our goal is to estimate the optical depth for pair production of a gev and an mev photons shell .
the width of the mev photons shell is ( see their eq .
( 10 ) ) . ] : @xmath14 where @xmath15 is the observed duration of the mev pulse and @xmath16 is the redshift of the burst . a gev photon , emitted along the @xmath17 axis at @xmath18 and @xmath19 is immersed in mev photons until it leaves the mev shell at @xmath20 ( see fig .
[ fig.geometry.2 ] ) : @xmath21 } , r_{\max}(r_g,\theta_g ) = \frac{r_g}{1 -{2 c t_{90}}/[{(1+z)\theta_g^2r_g}]},\ ] ] for @xmath22 $ ] , while @xmath23 for @xmath24 $ ] . ) at @xmath8 and it leaves the shell at @xmath25 . ] assume , for simplicity , that the emitted flux of the mev photons is constant over time then the number density of the mev photons is : @xmath26 where @xmath27 is the isotropic equivalent mev luminosity , @xmath28 is the observed peak energy and @xmath29 is the radius .
the spectrum of the mev photons is described by the band function : @xmath30 where , in view of the two region model , the mev component has an upper limit to the energy : @xmath31 .
fortunately @xmath32 is insensitive to the exact value of @xmath31 . at @xmath7
the mev photons are beamed with an angular width @xmath4 along the radial direction outwards . since @xmath33 the mev photons travel almost radially outwards .
relative to the gev photons , that move along the @xmath17 axis , the angular width of the mev photons is of order @xmath34 and it decreases with @xmath35 .
this leads to two angular regimes : along the @xmath17 axis ( for @xmath36 ) , the mev photons have very small angles @xmath37 relative to the gev photons .
this leads to a very small optical depth along the axis .
for @xmath38 the angular spread of the mev photons can be neglected and the typical angle between the mev and the gev photons is simply @xmath39 . more generally , the angle between an mev photon emitted radially outwards at @xmath40 and a gev photon emitted parallel to the @xmath17 axis at @xmath41 is : @xmath42 where @xmath43 is the distance the gev photon travels before it collides with the mev photon and @xmath44 is the angle of a given mev photon relative to the axis parallel to the gev photon axis .
a gev photon , with an ( observed ) energy @xmath45 , and an mev photon , with an ( observed ) energy @xmath46 , can produce a pair if @xmath47 , where : @xmath48 the cross section behaves like ( e.g. * ? ? ? * ) : @xmath49 , \label{eq : sigma}\ ] ] where @xmath50 @xmath51 and @xmath52 is the thompson cross section .
we can estimate now the overall optical depth for the gev photon : @xmath53 the effective optical depth for an observed photon is averaged over all angles : @xmath54 = \frac{\int_{0}^{\theta_{g , j } } \mathcal{d}^{(3+\beta_g ) } e^{-\tau(\theta_g , e_g ) } \theta_g d\theta_g } { \int_{0}^{\theta_{g , j } } \mathcal{d } ^{(3+\beta_g ) } \theta_g d\theta_g } , \label{eq : e_tau_complete}\ ] ] where @xmath55 is the doppler factor , @xmath56 is the bulk velocity and @xmath57 is the photon index of the gev emission .
it is interesting to note that eq .
( [ eq : tau_theta ] ) is independent of @xmath11 .
the overall dependence of @xmath58 on @xmath11 arises from eq .
( [ eq : e_tau_complete ] ) where @xmath11 determines ( via @xmath59 ) the effective width of the integration .
we can simplify the above model and obtain an almost analytic formula by making a few approximations .
we show later that the full numerical results indeed agree with these formulae .
first , we approximate the mev spectrum in the relevant energy range using a single power law : @xmath60 .
second , we approximate the cross section ( eq .
( [ eq : sigma ] ) ) as @xmath61 ( this form allows us for an analytic integration . the numerical factor is chosen by comparison of the analytic approximate results with the full numerical solution . ) .
third , we divide the analysis to two regimes : for @xmath62 all the mev photons move radially and the collision angle is simply @xmath63 . for @xmath64
the scatter in the directions of the mev photons is important and we approximate the collision angle as @xmath65 .
define @xmath66 and @xmath67 $ ] .
for large angles the collision angle between the mev and gev photons is @xmath68 .
the minimal energy of the mev photon for pair production is : @xmath69 the number density of the mev photons : @xmath70 where @xmath71 . collecting the above expressions : @xmath72 , \label{eq : tau_theta_g_sim}\end{aligned}\ ] ] where @xmath73 and @xmath74 .
we use @xmath75 to express @xmath8 in terms of @xmath76 and @xmath11 , where @xmath76 is the time when the gev photon is observed .
generally @xmath77 with @xmath78 for an external shock and @xmath79(@xmath80 ) for an internal shock . as we see
later the result is not very sensitive to @xmath81 . realizing that the gev photons are mainly coming from @xmath82
, we can invert now this equation to obtain a rough estimate of the minimal lorentz factor .
the solution will be consistent if @xmath83 namely if @xmath84 .
to do so we assume that the factor in the square brackets of eq .
( [ eq : tau_theta_g_sim ] ) is order of unity .
we obtain : @xmath85^{\frac{1}{2\alpha_{_m}+2 } } \nonumber \\ & \simeq & 34 \left ( \frac{30 \times 10^{4\alpha_{_m}}}{\alpha_{_m } 34^{2\alpha_{_m}+2}}\right)^{\frac{1}{2\alpha_{_m}+2 } } \left [ \left(\frac{1+z}{2}\right)^{2\alpha_{_m}-3 } l_{m,51 } \epsilon_{g,4}^{\alpha_{_m}-1 } \epsilon_{p,0}^{\alpha_{_m}-2 } \eta^{-1 } t_{g,2}^{-1 } \right]^{\frac{1}{2\alpha_{_m}+2 } } , \label{eq : gamma_g_min1}\end{aligned}\ ] ] where the notation @xmath86 is used and @xmath87 . for @xmath88 , the overall coefficient is 34 . note the weak ( with a power @xmath89 to @xmath90 ) dependence of @xmath91 on @xmath81 . for @xmath92
the typical collision angle of the mev photons is @xmath93 .
the optical depth is similar to eq .
( [ eq : tau_theta_g_sim ] ) , with @xmath94 replacing @xmath19 and @xmath95 : @xmath96 .
\label{eq : tau_theta_m}\end{aligned}\ ] ] interestingly , this formula depends on @xmath10 and it is independent of @xmath11 . for typical values , @xmath97 , @xmath88 , @xmath98gev , @xmath99mev , @xmath100erg , @xmath101 cm , @xmath102 cm , we obtain @xmath103 , which provides only a very week constraint on the bulk lorentz factor of the mev region .
if we set @xmath104 , this expression reduces to the one zone case , where the relevant collisions occur at @xmath105 in the observer s frame . taking @xmath106 , where @xmath107s is the typical duration of @xmath108-ray pulse , @xmath109 , consistent with @xcite .
fig . [ fig.tau-theta ] depicts the dependence of @xmath58 on @xmath19 for a set of typical parameters : @xmath97 , @xmath110s and @xmath111 cm , @xmath112gev , @xmath99mev , @xmath100erg , and @xmath113s .
@xmath57 is always chosen to be 2 .
@xmath114 for these parameters ( with @xmath115 , @xmath116 ) for @xmath117 .
note that numerical experiments reveal that to obtain the average effective optical depth using eq .
( [ eq : e_tau_complete ] ) we have to integrate up to @xmath118 .
the results in this figure depict both the full ( eq .
( [ eq : tau_theta ] ) ) and the simplified ( eq . ( [ eq : tau_theta_g_sim ] ) ) calculations .
the later are depicted by the two thicker lines which are almost superposed on the corresponding numerical results , showing consistency of the analytical solution with the full numerical one .
as the angle increases , the small angle approximation ( @xmath119 ) breaks down and the approximate solution slightly diverges from the numerical one
. however , the deviation is small and it usually takes place in a regime that is not critical for the overall optical depth . one can clearly see different segments of power law dependence of the optical depth on @xmath19 . for a given gev photon
@xmath19 determines the colliding angle and hence @xmath120 . as
most photons are in the lowest energy and the cross section is largest near @xmath120 , the optical depth is dominated by the low energy photons at @xmath120 and the photon index can be simply taken the one at @xmath120 . for small angle , @xmath121 ( see eq .
( [ eq : e_min ] ) ) and the effective spectral slope is @xmath122 . for large angles @xmath123 and the effective spectral slope is @xmath124 .
correspondingly , lines with the same @xmath122 coincide at small angles while lines with the same @xmath124 conicide at large angles .
these results suggest that the single power law approximation for the spectrum is useful .
@xmath120 determines the relevant spectral , @xmath124 or @xmath122 .
a second transition takes place when @xmath125 approaches unity , namely the width of the interaction region is small compared to @xmath8 . according to eq .
( [ eq : tau_theta_g_sim ] ) , @xmath126 for @xmath127 and @xmath128 for @xmath129 .
the transition between the two takes place as @xmath130 .
for some parameters the two transitions may coincide to one and we have chosen @xmath27 so that the two transitions are clearly seen .
gev as a function of @xmath19 .
different lines correspond to different spectral indices .
the parameters of the lines are encoded using a combination of colors and line types .
@xmath131 are represented as blue , black and red respectively .
@xmath132 are represented as dashed , solid and dotted lines respectively .
the two thick lines are the analytical solution from eq .
( [ eq : tau_theta_g_sim ] ) for @xmath133 . ] fig .
[ fig.gamma_g-l ] depicts @xmath91 as a function of the mev luminosity @xmath27 . as one can expect ( see eq .
( [ eq : gamma_g_min1 ] ) ) @xmath91 increases with @xmath27 .
the optical depth for @xmath134s ( thin solid line , @xmath135 ) is much larger than the other ones for which @xmath136s , again in agreement with eq.([eq : gamma_g_min1 ] ) . if @xmath137 , @xmath91 increases with @xmath27 as a single power law
this is consistent with eq.([eq : gamma_g_min1 ] ) . for @xmath138 ,
the relationship breaks into two power law segments that are dominated by different parts of band spectrum of the mev photons .
as the luminosity increases the lorentz factor increases , and the dominant contribution to the optical depth arises from @xmath139 , with a lower effective @xmath19 , @xmath120 increases and hence for large values of @xmath27 the opacity is dominated by the high energy spectral slope , @xmath122 .
conversely , for low values of @xmath27 the opacity is dominated by @xmath124 .
the transition takes place at @xmath140 . as a function of the mev luminosity , @xmath27 .
different curves correspond to different photon indices .
the parameters of the lines are encoded using a combination of colors and line types .
@xmath131 are represented as blue , black and red respectively .
@xmath132 are represented as dashed , solid and dotted lines respectively .
the thick grey line has @xmath141s , corresponding to a short grb . ]
lccccccccccc grb & z & @xmath15(s ) & @xmath124 & @xmath122 & @xmath142(mev ) & @xmath143 ( erg / s ) & @xmath45(gev ) & @xmath76 & @xmath144 & @xmath145 & ref + 080916c & 4.35 & 66 & 1.02 & 2.21 & 1.17 & @xmath146 & @xmath147 & 40 & * 193*(414 ) & 880 & 1 + 090510 & 0.903 & 0.5 & 0.48 & 3.09 & 5.1 & @xmath148 & 3.4 & 0.5 & * 150 * ( 277 ) & 1200 & 2 + 090902b & 1.822 & 30 & 0.61 & 3.87 & 0.8 & @xmath149 & 33.4 & 82 & * 120 * ( 218 ) & 1000 & 3,4 + 090926a & 2.1062 & 20 & 0.693 & 2.34 & 0.27 & @xmath150 & 19.6 & 26 & * 150 * ( 318 ) & 1200 & 5 + + + + + + [ tab : gamma ] the two zone limits , @xmath91 , for four fermi bursts are shown in table [ tab : gamma ] together with the single zone limits , @xmath151 , and some parameters of the bursts .
while the single zone limits , @xmath151 , are of order 1000 and even larger , the two zone limits are around 200 ( 400 ) , for @xmath152 .
it should be stressed that in two of these bursts , grb 080916c and grb 090510 , the highest energy gev photon used to determine the single zone limit is coincident with a large mev flux , while an 11 gev photon is contemporaneous in grb 090902b with an mev spike .
still in both grb 090902b and grb 090926a gev photons are observed after the end of the prompt mev ( @xmath153 ) .
thus , it is not clear whether the single zone of the two zone limit should be used .
if the gev photons are from an external shock , another direct constraint on @xmath11 arises from the dynamics of external shock : @xmath154 @xcite . with observed gev peak emission time @xmath155 of @xmath156s @xcite , @xmath157erg and a circum - burst density @xmath158
, we obtain @xmath159 .
this value is larger than the lower limits obtained from the two - zone compactness estimate .
however , it is quite uncertain , in view of the uncertainty in the determination of @xmath155 .
following various indications that the high energy ( gev ) emission in grbs is produced in a different region than the lower energy ( mev ) emission we derived here revised compactness limits on the lorentz factor of grb outflows within a two - zone model .
we considered the natural " model in which the gev emission is produced at larger radii than the mev emission .
this would arise , for example , if the mev emission is produced by internal shocks and the gev emission by the afterglow , as has been suggested recently by several authors @xcite .
we calculated the optical depth for pair production by a gev photon passing through the mev photons shell .
our results reduce to the one - zone model when the emission region of the mev and gev coincide .
collisions between the gev and mev photons occur in the two - zone model at larger radii than the prompt emission radius , the density of the mev emission is smaller ( than in the prompt emission regime ) and the mev photons are more collimated along the line of sight .
consequently , the optical depth is smaller compared to the one - zone case and the compactness constraint on the lorentz factor becomes weaker .
the new constraint that we find is only for the lorentz factor of the gev region .
the constraint on the lorentz factor of the mev emitting region , arising from the optical depth of the gev photon , is rather weak .
the weak limit does not contradict to the neutrino driven jets@xcite nor to the magnetic driven jets@xcite .
for a canonical set of parameters , like @xmath97 , @xmath110s , @xmath112gev , @xmath99mev , @xmath160erg , and @xmath113s , the constraint on the gev region is @xmath161 .
when we apply the two zone constraint to four fermi bursts , we find minimal lorentz factors of about 200 - 400 , about one fifth to one half ( depending on @xmath81 ) of the the one - zone limit which is in the order of 1000 .
we conclude that one should proceed with care when applying the one - zone limits to the fermi data and unless we can verify that the gev emission is indeed produced in the same region as the lower energy prompt emission we should consider the more relaxed two zone limits .
we thank ehud nakar and uri vool for helpful discussions and an anonymous referee for helpful comments .
the research was supported by an erc grant , the israel center of excellence for high energy astrophysics , a special grant of chinese academy of sciences , national basic research programme of china grant 2009cb824800 and the national natural science foundation of china under the grant 10703002 and 11073057 .
tp thanks the purple mountain observatory of nanjing and huazhong university of science and technology for hospitality while some of this research was done .
abdo a. a. , et al . , 2009a , science , 323 , 1688 abdo a. a. , et al . ,
2009b , apj , 706 , l138 ackermann m. , et al .
, 2010 , apj , 716 , 1178 aloy m. a. , janka h .-
t . , & muller e. , 2005 , a&a , 436 , 273 aoi j. , et al . , 2009 , arxiv:0904.4878v1 de palma f. , et al . , 2009 , gcn , 9872 fenimore e. e. , epstein r. i. & ho , c. , 1993 , a&as , 97 , 59 fishman g. j. , & meegan c. a. , 1995 , ara&a , 33 , 415 gao w. h. , mao j. r. , xu d. , & fan y. z. , 2009 , apj , 706 , l33 ghisellini g. , ghirlanda g. , nava l. & celotti a. , 2010 , mnras , 403 , 926 gonzlez m. m. , dingus b. l. , kaneko y. , preece r. d. , dermer c. d. & briggs m. s. , 2003 , nature , 424 , 749 hurley k. , et al . , 1994 , nature , 372 , 652 jauch , j. m. , & rohrlich , f. , `` the theory of photons and electrons '' , springer - verlag , 1980 .
komissarov s. s. , vlahakis n. , konigl a. , & barkov m. v. , 2009 , mnras , 394 , 1182 kumar p. , & barniol duran r. , 2009 , mnras , 400 , l75 kumar p. , & barniol duran r. , 2010 , arxiv:0910.5726 li , z. , 2010 , apj , 709 , 525 lithwick y. , & sari r. , 2001 , apj , 555 , 540 piran t. , 1995 , in `` some open questions in astrophysics '' eds . j. n. bahcall and j. ostriker , princeton university press .
arxiv : astro - ph/9507114 piran t. , & nakar e. , 2010 , apj , 718 , l63 ruderman m. , 1975 , ann . n. y. acad .
, 262 , 164 sari r. , & piran t. , 1999 , apj , 520 , 641 swenson c. a. , et al . , 2010 ,
apj , 718 , l14 tchekhovskoy a. , mckinney j. c. , & narayan r. , 2009 , apj , 699 , 1789 woods e. , & loeb a. , 1995 , apj , 453 , 583 xue r. r. , fan y. z. , & wei d. m. , 2008 , mnras , 389 , 321 zhao x. h. , li z. , & bai j. m. , 2010 , arxiv:1005.5229 zou y. c. , piran t. , & sari r. , 2009 , apj , 692 , l92 zou y. c. , & piran t. , 2010 , mnras , 402 , 1854 | fermi observations of gev emission from grbs have suggested that the lorentz factor of some grbs is around a thousand or even higher . at the same time
the same fermi observations have shown an extended gev emission indicating that this higher energy emission might be a part of the afterglow and it does not come from the same region as the lower energy prompt emission .
if this interpretation is correct then we may have to reconsider the opacity limits on the lorentz factor which are based on a one - zone model .
we describe here a two - zone model in which the gev photons are emitted in a larger radius than the mev photons and we calculate the optical depth for pair creation of a gev photon passing the lower energy photons shell . we find that , as expected , the new two - zone limits on the lorentz factor are significantly lower . when applied to fermi bursts
the corresponding limits are lower by a factor of five compared to the one - zone model .
it is possible that both the mev and gev regions have relatively modest lorentz factors ( @xmath0 ) , which is significantly softer then one zone limit . |
the astrophysical rapid neutron - capture process ( r - process ) has been recognized for a long time as the scenario responsible for the synthesis of approximately half of the nuclear species in nature , which are more massive than fe @xcite .
it requires environments with a high neutron density , where neutron captures are faster than @xmath0-decays , even for neutron - rich unstable nuclei up to 15 - 30 units from stability . only under such conditions it is possible that highly unstable nuclei are produced near the neutron drip - line via neutron captures , @xmath1-photodisintegrations , @xmath2-decays and @xmath0-delayed processes , leading also to the formation of the heaviest elements in nature like th , u , and pu .
far from stability , magic neutron numbers are encountered for smaller mass numbers a than in the valley of stability , which shifts the r - process abundance peaks in comparison to the s - process peaks ( which occur at neutron shell closures at the stability line due to small neutron capture cross sections ) . besides this basic understanding , the history of r - process research has been quite diverse in suggested scenarios ( for reviews see @xcite ) . starting with a seed distribution somewhere around a=50 - 80 before massive neutron - capture sets in
, the operation of an r - process requires 10 to 150 neutrons per seed nucleus to form all heavier r - nuclei .
the question is which kind of environment can provide such a supply of neutrons to act before decaying with a 10 min half - life .
the logical conclusion is that only explosive environments , producing or releasing these neutrons suddenly , can account for such conditions . without having discussed any stellar models yet ,
we want to mention some recent literature on these issues .
the r - process site is not a settled one ; however , two astrophysical settings are suggested most frequently : ( i ) type ii supernovae ( snii ) with postulated high - entropy ejecta @xcite , where the delayed emergence of r - process matter in galactic evolution @xcite indicates that these can probably only be sn with small progenitor mass ; and ( ii ) neutron star mergers or other events with low - entropy ejecta @xcite . after this ad hoc introduction of sites ,
we want to return first to the conditions in high - temperature explosive plasmas in order to test what kind of nuclear - physics input is required , and also what type of thermodynamic environment properties can lead to the neutron to seed ratios mentioned above .
the main aspect is that a whole variety of burning processes responsible for the abundances of intermediate and heavy nuclei , like e.g. explosive o and si - burning , are leading to partial ( quasi ) or full equilibria of reactions ( qse or nse ) .
these are described by the chemical potentials of nuclei which form a boltzmann gas .
the abundance ratios are ( besides thermodynamic environment properties ) determined by mass differences
. therefore , mass uncertainties matter , while uncertainties in cross sections do not enter abundance determinations .
non - equilbrium regions are identified by small reaction cross sections , either due to small q - values for reactions out of the magic numbers or due to small level or resonance densities for light nuclei .
for sufficiently high temperatures all qse groups merge to a full nse .
the chemical equilibrium for neutron or proton captures leads to abundance maxima at specific neutron or proton separation energies @xmath3^{3/2 } \bigl[{{2\pi \hbar^2}\over { m_ukt}}\bigr]^{3/2 } { \rm exp}(s_{p}(a_c)/kt ) \\ { s_{p}(a_c ) \over kt}&= & 24 \ln\bigl [ ( { a_ta_p \over { a_c}})^{3/2 } ( { g(a_t)g_p \over 2g(a_c)})^{3/2 } ( { t \over { 10 ^ 9{\rm k}}})^{3/2 } { n_a \over { n_{p}/{\rm cm}^3}}\bigr].\end{aligned}\ ] ] these equations are valid for neutron , proton and/or @xmath4 captures , as indicated by the subscript @xmath5 ( projectile ) ; subscript @xmath6 stands for target and @xmath7 for the compound nucleus .
the @xmath8 s are abundances related to number densities @xmath9 via @xmath10 , where @xmath11 denotes the mass density and @xmath12 the avogadro number . at the maximum in an isotopic or isotonic
line we have @xmath13 , if the partition functions g are neglected .
this is slightly modified by logarithmic dependences on projectile ( neutron or proton ) density @xmath14 and temperature @xmath15 . in case
an equilibrium with neutrons and protons exists , the abundance maximum is found in the nuclear chart at the intersection of the relevant neutron and proton separation energies .
the free neutron and proton densities reflect the total neutron / proton ratio in matter , which is determined by slow ( not in equilibrium ) , weak interactions via changing the total proton / nucleon ratio @xmath16@xmath17@xmath18@xmath19 .
the understanding that explosive burning stages are governed by qse or even nse is growing @xcite .
this has been shown recently in calculations of snii nucleosynthesis @xcite with two different libraries of nuclear reaction rates .
different types of qse - groups can emerge in explosive burning .
the high temperature phase of the rp - process ( rapid proton capture ) in x - ray bursts , where no neutrons are available , witnesses isotonic lines in a ( @xmath20 equilibrium @xcite .
another application is the r - process , the focus of the present paper . here
the qse - groups are isotopic lines in an ( @xmath21 equilibrium .
the connecting weak interactions for both processes are @xmath22- or @xmath2-decays . during the final freeze - out from equilibria ,
when temperatures decline below equilibrium conditions , reaction rates and capture cross sections can count again .
in general , it should be pointed out that equilibria simplify the understanding of explosive nucleosynthesis processes and individual cross sections play a much less important role than reaction q - values . since the rp- and r - process explore exotic nuclei close to the proton or neutron drip - lines , we need to focus on nuclear masses and @xmath0-decay properties .
an understanding of the underlying nuclear structure is essential for both types of nuclear parameters .
such information usually comes from a close interplay between experiment and theory .
this will be discussed in more detail in sections 3 and 4 . in order to underline why ( within an understanding of the - nuclear - functioning of the r - process ) the two scenarios mentioned in the beginning of this section
seem to be the most promising ones , we want to present here two major results found recently by freiburghaus et al .
they performed an analysis of neutron to seed ratios in different types of adiabatic expansions with a given entropy s , proton / nucleon ratio @xmath23 , and expansion time scale . in order to obtain the required 10 to 150 neutrons per r - process seed nucleus ( in the fe - peak or somewhat beyond ) permitting to produce nuclei with masses a@xmath19200 , this translates into a @xmath16@xmath17@xmath18@[email protected] for a composition of fe - group nuclei and free neutrons .
such a high neutron excess is only available for high densities in neutron stars under @xmath0-equilibrium ( @xmath25 , @xmath26 ) , based on the high electron fermi energies which are comparable to the neutron - proton mass difference @xcite .
another option is a so - called extremely @xmath4-rich freeze - out in complete si - burning with moderate @[email protected] .
this corresponds to a freeze - out from qse with a weak connection between the light ( @xmath27 ) and heavy ( mg kr ) qse groups due to low densities .
the links accross the particle - unstable a=5 and 8 gaps are only possible via the three - body reactions @xmath28 and @xmath29 to @xmath30c and @xmath31be , whose reaction rates show a quadratic density dependence . the entropy can be used as a measure of the ratio between the remaining he mass - fraction and heavy nuclei .
similarly , the ratio of neutrons to heavy nuclei ( i.e. the neutron to seed ratio ) is a function of entropy and permits for high entropies , with large remaining he and neutron abundances compared to small heavy - seed abundances , neutron captures which proceed to form the heaviest r - process nuclei @xcite . after the freeze - out of charged - particle reactions in matter , which expands from high temperatures but relatively low densities , as much as 90% of all matter can be locked into @xmath32he with @xmath33=@xmath34 , which leaves even for moderate @xmath23 s a large neutron / seed ratio for the few existing heavier nuclei .
the behaviour of these two environments , representing a normal ( low - entropy ) and an @xmath4-rich ( high - entropy ) freeze - out , is summarized in fig . 1 .
the available number of neutrons per heavy nucleus @xmath35 after charged - particle freeze - out , when the large qse - groups break up into isotopic lines , is shown as a function of entropy and initial @xmath23 . at low entropies
the transition to a normal freeze - out occurs , indicated by the negligible entropy dependence .
this introduction centered ( in a cursory way ) on the questions : ( i ) how can one obtain - after an expansion and cooling to temperatures of @xmath36k , when the charged - particle freeze - out occurs - the neutron to seed ratios required for a successful r - process ; and ( ii ) what type of nuclear - physics input is of dominant importance . in the following section 2 , we will discuss in more detail the methods to follow abundance changes during the neutron - capture phase , and will show how in simplified r - process models the observed abundance features and nuclear properties far from stability are related . in section 3 , we present the nuclear - data input employed , and show the strong effects which the choice of nuclear data can have . section 4
summarizes the experimental information on r - process nuclei , to a large extent obtained at cern / isolde .
finally , in section 5 we discuss how r - process calculations evolved over the past decade with continuously improved nuclear - structure knowledge from initially classical equilibrium calculations to recent results from specific astrophysical sites .
the system of differential equations for an r - process network includes the terms for neutron captures , neutron - induced fission , photodisintegrations , @xmath0-decays , @xmath0-delayed neutron emission and fission .
this leads to a set of n differential equations for the individual nuclear abundances as in eq.(3 ) @xmath37 where the first term on the right - hand side includes @xmath0-decays ( with all possible emission channels , including delayed fission ) , spontaneous fission , and photodisintegrations .
the second term includes all neutron - induced reactions .
the terms are positive or negative , depending upon whether they produce or destroy the nucleus ( @xmath38 ) . for the heavy ( i.e. high-@xmath34 ) and neutron - rich nuclei in question , due to the high coulomb barriers between the particles only neutron - induced reactions are of importance for the temperatures of interest .
thus , only reactions involving neutrons as projectiles or emitted particles , such as ( @xmath39,n)-photodisintegrations and @xmath0-delayed neutron emission , need to be considered .
we can replace @xmath40 in eq.(3 ) by the neutron number density @xmath41 , underlining the fact that @xmath41 is the important quantity for the r - process rather than the neutron abundance . for nuclei with @xmath34@xmath1780
, fission does not play any role , and neutron captures , photodisintegrations and @xmath0-decays ( including delayed - neutron branchings ) dominate .
if we include @xmath0-decays leading to the emission of up to three delayed neutrons , the abundance change of nucleus ( z , a ) is given by @xmath42 where @xmath43 is the thermally averaged @xmath44-reaction rate @xmath45 of nucleus ( @xmath381 ) , @xmath46 is the photodisintegration rate @xmath47 for nucleus ( @xmath38 + 1 ) , @xmath48 is the @xmath0-decay rate of nucleus @xmath49 , and @xmath50 , @xmath51 , and @xmath52 are the rates of @xmath0-decay followed by the emission of one , two or three neutrons , respectively . depending upon the specific conditions , either @xmath0-decays can be faster than neutron captures and photodisintegrations like in an s - process , or vice versa , when an ( @xmath53)@xmath54(@xmath55 ) equilibrium exists [ see sect . 1 and eq(1 ) ] . if the @xmath0-flow ( i.e. the @xmath0-decays of the nuclei ) from each @xmath34-chain to ( @xmath34 + 1 ) is equal to the flow from ( @xmath34 + 1 ) to ( @xmath34 + 2 ) , then a steady - flow or @xmath0-flow equilibrium will exist .
when an ( @xmath53)@xmath54(@xmath55 ) equilibrium or a steady - flow condition is established , eq.(4 ) simplifies . with these simplifications
, a solution of the general set of equations , which would involve a linear system of the size of a nuclear network , can be avoided .
as long as there is a high neutron density and a large high - energy photon density ( i.e. a high temperature ) , we can expect that neutron captures and photodisintegrations occur on a much faster time scale than @xmath0-decays .
this leads after charged - particle freeze - out ( see sect . 1 ) to qse groups in form of isotopic chains ( denoted ( @xmath53)@xmath54(@xmath55 ) equilibrium or _ `
waiting - point ' _ approximation ) , where the nucleus with maximum abundance in each isotopic chain must wait for the longer @xmath0-decay time scale .
the chemical equilibrium @xmath56 of particles in a boltzmann gas causes abundance ratios which are dependent only on @xmath41 , @xmath15 and @xmath57 @xmath58^{3/2 } \bigl[{{2\pi \hbar^2}\over { m_ukt}}\bigr]^{3/2 } { \rm exp}(s_n(a+1)/kt).\ ] ] this equation is identical to eq.(1 ) with @xmath5(projectile)=@xmath9 , @xmath59=1 , @xmath60=@xmath61 , @xmath62=@xmath63 , and @xmath64=2 .
the neutron - separation energy @xmath57 introduces the dependence on nuclear masses in an ( @xmath53)@xmath54(@xmath55 ) equilibrium , where the abundance maxima in each isotopic chain are located at the @xmath57 given by eq.(2 ) , which is the same in all isotopic chains .
2 shows such a contour line of @xmath57@xmath652mev for the finite range droplet model , frdm @xcite .
in addition it displays the line of stability and the solar r - abundances as a function of a. the abundance flow from each isotopic chain to the next is governed by @xmath0-decays .
we can introduce a total abundance in each isotopic chain @xmath66=@xmath67 , and each @xmath68 can be expressed as @xmath68=@xmath69 .
the individual population coefficients @xmath70 are obtained from the equilibrium condition [ eq.(5 ) ] .
the set of differential equations which replaces eq.(4 ) in this case is @xmath71 where @xmath72 is the @xmath0-decay rate of nucleus @xmath49 .
this is a system of as many equations as the number of @xmath34-chains .
the individual abundances are obtained from eq.(5 ) .
the waiting - point approximation is only valid for high temperatures and neutron number densities of the gas before the freeze - out from chemical equilibrium with neutrons occurs @xcite ) for neutron number densities and temperatures as low as @xmath41@xmath6510@xmath73 @xmath74 and @xmath15@xmath6510@xmath31 k. process timescales in excess of @xmath0-decay half - lives lead to a steady - flow equilibrium , in addition to an ( @xmath53)@xmath54(@xmath55 ) equilibrium , i.e. @xmath75=0 in eq.(6 ) or @xmath76 in such a steady - flow equilibrium , the assumption of an abundance for @xmath77 at a minimum @xmath34-value is sufficient to predict the r - process curve .
the calculation of @xmath70 from eq.(5 ) requires the knowledge of nuclear masses ( or @xmath57-values ) , a neutron number density @xmath41 and a temperature @xmath15 .
the @xmath0-decay rates in eqs.(4 ) , ( 6 ) and ( 7 ) are related to the half - lives of very neutron - rich nuclei via @xmath78 .
thus , in the simplest of all cases , i.e. provided that an ( @xmath53)@xmath54(@xmath55 ) equilibrium and a steady - flow has been reached , the knowledge of @xmath41 , @xmath15 , @xmath57 , and @xmath79 alone would be sufficient to predict the whole set of abundances as a function of @xmath61 .
results by kratz et al .
@xcite indicated that this be the case for conditions which reproduce the low - mass wings of the @xmath61@xmath6580 and @xmath61@xmath65130 peak regions of the solar - system r - process abundances @xmath80 @xcite . in the most general case of ( astrophysical ) environment conditions
, we have thus to solve the set of eq.(4 ) for all nuclei from stability to the neutron drip - line ( 1000 - 2000 ) .
the type of conditions after charged - particle freeze - out discussed in section 1.2 permits , however , the use of eq.(6 ) , i.e. the assumption of an ( @xmath53)@xmath54(@xmath55 ) equilibrium . for small @xmath0-decay half - lives , encountered in between magic numbers and for small z s at magic numbers , even the steady - flow approximation of eq.(7 ) seems applicable , until the final neutron and temperature freeze - out no longer permits the use of any equilibrium approach .
here , one can imagine two extreme options : ( i ) the occurance of an instantaneous freeze - out ; then the results of eq.(6 ) just have to be followed by the final decay back to stability , where also @xmath0-delayed properties ( neutron emission and fission ) are needed .
( ii ) in the more general case of a slow freeze - out , the set of eq.(4 ) has to be solved .
then also neutron captures can affect the initial waiting - point abundance results .
for testing nuclear - structure properties it is advisable to make use of a simple astrophysical environment model , with @xmath41=const , @xmath15=const and a duration time @xmath81 , but chosen for conditions where an equilibrium in isotopic chains is obtained and eq.(6 ) can be applied ( however , without assuming the steady flow of eq.(7 ) ) .
3 shows three such calculations for @xmath41=@xmath82 , @xmath83 , and @xmath84 g @xmath74 at @xmath15=1.35@xmath85 k for duration times @xmath81 of 1.2 , 1.7 , and 2.1 s , respectively .
when testing the resulting abundances , one notices that a steady flow [ eq.(7 ) ] is actually obtained for the low - mass wings of the peaks and the connecting lighter - mass regions between magic neutron numbers . when entering these conditions , which reproduce the three r - process peaks , into eq.(2 ) , one notices that they correspond to three different r - process paths with different neutron - separation energies @xmath57 between 4 and 2 mev .
this indicates that a full fit to the solar r - process abundances requires a superposition of different stellar conditions ( but not necessarily different astrophysical sites ) . as outlined in the previous section , for a given neutron density ( or radiation entropy ) , the r - process proceeds along a contour line of constant @xmath57 to heavier nuclei ( see , e.g. fig .
4 in @xcite , where the abundance flow from each isotopic chain to the next is dominated by the @xmath0-decays .
the flow equilibrium then implies approximate equality of progenitor abundance @xmath86 ( i.e. the initial abundance of a waiting - point isotope lying in the r - process path ) times @xmath0-decay rate @xmath87 .
thus , with @xmath88=const .
the half - lives along the r - process flow path would directly define the @xmath86 , and when taking into account @xmath89 branching during freeze - out also the final @xmath80 .
still some 15 years ago , there were severe doubts among astrophysicists that the validity of this simple and elegant approximation would ever be proven , because it requires experimental information on far - unstable waiting - point isotopes believed to be inaccessible in terrestrial laboratories .
however , only a few years later in 1986 , a new area in nuclear astrophysics started with the identification of the first two classical neutron - magic waiting - point isotopes @xmath90cd@xmath91 at cern / isolde @xcite , and @xmath92zn@xmath93 at the fission - product mass separators osiris @xcite and tristan @xcite . as was shown by kratz et al .
@xcite , indeed , first evidence for the existence of at least a local steady - flow equilibrium could be achieved , which presented a stimulating challenge to both theoreticians and experimentalists in the nuclear - physics community in the following years . nevertheless , even today the vast majority of very neutron - rich r - process nuclei is experimentally not accessible .
therefore , a general understanding of their nuclear properties remains to be obtained only through theoretical means . since
a number of different quantities are needed in r - process calculations , as outlined above , in the past it was often not possible to derive them all from one source . taking them from different sources , however , may raise the question of consistency . in order to avoid an eventual vanishing of real signatures or the creation of artificial r - abundance effects from the use of nuclear - physics parameters from different models sometimes even of largely different sophistication we have performed our r - process calculations in a unified macroscopic - microscopic approach in which all nuclear properties can be studied in an internally consistent way .
this approach is , for example , discussed in detail in kratz et al .
@xcite and mller et al .
@xcite for the combination of nuclear masses from the finite range droplet model , frdm and @xmath0-decay properties from a quasi - particle random phase approximation , qrpa @xcite , in its present version using folded - yukawa single - particle levels and the lipkin - nogami pairing model ( mller and randrup @xcite ) .
analoguously , when adopting other mass formulae , such as the extended thomas - fermi plus strutinsky integral models , etfsi-1 ( aboussir et al .
@xcite ) or etfsi - q ( pearson et al .
@xcite ) , or the spherical hartree - fock - bogolyubov method with a specific skyrme force , hfb / skp ( dobaczewski et al .
@xcite ) , we normally use theoretical @xmath0-decay quantities deduced from qrpa calculations with masses and deformation parameters given by that particular model . with the above macroscopic - microscopic approaches , several rather sophisticated and internally consistent , global nuclear - data sets are now available for astrophysical calculations which are expected to yield more reliable predictions than earlier models , like the old droplet - type formulae of myers @xcite and von groote et al .
@xcite or the gross theory of @xmath0-decay ( takahashi et al .
@xcite ) . nevertheless , being aware that even the more microscopic approaches do have their deficiencies , we have continuously tried to improve the basic data sets by short - range extrapolations of known nuclear - structure properties , either model - inherently not contained in or not properly described by the above global methods
. for details , see e.g. kratz et al .
@xcite . and ,
in contrast to other authors who prefer to use exclusively model predictions in their r - process calculations ( see , e.g. @xcite ) , we have steadily updated our files by including all new experimental masses , @xmath79 and @xmath89 values ( @xcite ) .
these nuclear - data sets have been used extensively during the last years in our r - process calculations ( @xcite ) to reproduce the @xmath80 pattern @xcite and to test nuclear structure close to the neutron drip - line . as has been outlined in sect . 2 , calculated r - abundances are strongly dependent on model predictions of masses and @xmath0-decay properties .
therefore , it seems worthwhile to test the influence of the choice of nuclear theories in selected mass regions before performing global , multi - component r - process calculations .
such tests may well be able to shed some light on the development of nuclear structures towards the neutron drip - line , and with this may define certain key properties and key experiments .
of particular importance in this context are the n@xmath94 peak areas , since the related n=50 , 82 and 126 shell closures act as bottle necks for the r - process matter flow .
already burbidge et al .
@xcite pointed out that the r - process very likely passes through neutron - magic isotopes around a@xmath6580 , 130 and 195 which have longer - than - average @xmath0-decay half - lives . at these points
the r - abundances build up and form the observed n@xmath94 peaks , as nuclei like @xmath92zn@xmath93 , @xmath90cd@xmath91 and @xmath95tm@xmath96 wait to @xmath0-decay .
it is therefore immediately evident that these clasical waiting points would not only give important constraints on r - process conditions , but would also yield much needed information on shell struture far from @xmath0-stability .
in fact both , the long t@xmath97 of the above three classical waiting - point nuclei , as well as the low s@xmath98 values of the respective ( n@xmath99 + 1 ) isotopes will considerably slow down the nuclear flow through the a@xmath6580 , 130 and 195 mass regions , and with this will to a large extent determine the total duration of a canonical r - process @xcite . to first illustrate this for the @xmath0-decay half - lives , fig .
4 shows four calculations of an r - process component mainly responsible for the build - up of r - abundances in the a@xmath65130 peak region , where all stellar ( t@xmath100 , n@xmath98 , @xmath81 ) and nuclear - physics ( q@xmath101 , s@xmath98 , @xmath102 , p@xmath98 ) input parameters ( taken from @xcite ) , except the @xmath0-decay half - lives of @xmath103ag@xmath91 and @xmath90cd@xmath91 , are kept constant .
the t@xmath97 of these two waiting - point isotopes are varied within the range of earlier model predictions .
it is quite evident from this figure , that the t@xmath97 not only determine the height ( and shape ) of the a@xmath65130 n@xmath94 peak , but also the r - matter flow to higher masses . with ` short ' half - lives ( t@xmath97=20 ms
; see upper left part of fig . 4 )
, the r - process would pass the n=82 region more or less unhindered ; this would result in a short overall process duration . on the other hand , ` long ' half - lives ( t@xmath97=750 ms ;
see lower right part of fig .
4 ) would slow down the nuclear flow through the a@xmath104130 region considerably , and consequently would require an r - process duration of 510 s when starting from z=26 .
such time scales are , however , too long for the presently favoured r - process scenarios . with this , also the reliability of gross @xmath0-decay theories which yield model - inherently ` longer ' t@xmath97 far from stability @xcite , may be questioned .
we now want to test the behaviour of different mass models ; i.e. the old hilf et al .
formula @xcite , the two global macroscopic - microscopic theories frdm @xcite and etfsi-1 @xcite , and the recent microscopic spherical hfb+skp model @xcite . within a very simple _ static _ approximation ( see ,
e.g. @xcite ) , one can obtain a first rough estimate under which neutron - density conditions an r - process will start to break through the classical waiting points at n=50 , 82 and 126 . as can be seen from fig . 5 , depending on the magic shell there are considerable differences for the four mass models chosen .
the behaviour is clearly correlated with the respective shell strengths .
a strong shell closure is reflected by a low neutron separation energy ( in fig . 5 given as @xmath105@xmath98=[s@xmath98(a)+s@xmath98(a+1)]/2 ) and a high neutron density required to overcome the neutron - capture ` bottle - necks ' at @xmath92zn , @xmath90cd and @xmath95tm .
when focussing for example on n=82 , the strongest shell closure is predicted by the etfsi-1 model , whereas the recent hfb+skp indicates a much weaker magicity .
already from this simple picture it is evident , that for the production of the whole n@xmath94 distribution quite different neutron - density regimes will be required for different mass models . in a next step , we will check the above indication by more realistic time - dependent calculations . in fig .
6 , the results of four one - component r - process calculations are shown , which use exactly the same astrophysical parameters as in our t@xmath97 test to produce the a@xmath65130 r - abundance region ( see fig .
4 ) . in all calculations ,
the @xmath0-decay properties were now taken from our recent evaluation @xcite . in this picture , the decisive nuclear - physics quantity are the s@xmath98 values , in particular those of the waiting - point isotopes before and at n=82 . apart from a certain memory effect from the n=50 shell closure , it is quite obvious that the predicted n=82 shell strength ( see fig .
5 ) determines the r - matter flow in the a@xmath65130 peak region and beyond . for mass models with a pronounced n=82 shell closure , i.e. frdm and etfsi-1
, a significant r - abundance trough occurs around a@xmath65115 , and the low s@xmath98 s of the n=83 isotones of @xmath106nb to @xmath107cd prevent a rapid matter flow to heavier masses under the given astrophysical conditions . on the other hand , a weaker n=82 shell closure , as e.g. predicted by the hilf et al .
formula and the recent hfb / skp theory , allows a more rapid build - up and transit of the a@xmath65130 region . with these mass models ,
the n=82 magic shell is reached at a higher z ( i.e. not before @xmath108tc ) , thus avoiding the a@xmath65115 abundance deficiency . and , at the same time , due to their
somewhat higher s@xmath98(n=83 ) values the break - out from n=82 can occur already at a lower neutron density ( i.e. around n@xmath98@xmath6515@xmath10910@xmath110 @xmath74 , according to fig .
we see from fig . 6 , that in the ` extreme ' of the hfb / skp theory with bogolyubov - enhanced shell - quenching towards the neutron drip - line , a neutron density of n@xmath98@xmath6510@xmath111 @xmath74 and a time scale of @[email protected] s are already sufficient to drive the nuclear flow from fe up to the a@xmath65195 r - abundance peak , whereas with the frdm and etfsi-1 mass models just the a@xmath65130 peak is built up .
this result may be of particular interest for the recently favoured high - entropy bubble scenario of a type - ii supernova @xcite , which still encounters some problems in obtaining high enough neutron densities ( respectively radiation entropies ) to completely build up the a@xmath65195 r - abundance peak .
the experimental study of neutron - rich nuclides lying in and near the projected r - process path serves two purposes , provision of direct data for use in nucleosynthesis calculations , and testing the theories from which nuclear properties of far - unstable isotopes are derived when no data are available . as is infered from the nucleosynthesis snapshots shown in the previous section , the predictions of the presently existing global nuclear models quite obviously differ considerably when approaching the limits of particle binding .
the reason may well be that the model parameters used so far , which were mainly determined to reproduce known properties near @xmath0-stability , need not always be proper to be used at the drip - lines .
therefore , experiments very far from stability will be essential to verify possible nuclear - structure changes with isospin , and to motivate improvements in nuclear theories . as already mentioned in sect .
3 , only about a decade ago the area of experiments _ in the r - process path _ started with the identification of the first two classical , neutron - magic ` waiting - point ' isotopes : @xmath92zn@xmath93 ( situated 10 mass - units away from stability ) and @xmath90cd@xmath91 ( even 16 units beyond stable @xmath112cd ) @xcite . in the following , we will give some technical details on the identification of @xmath90cd , in order to convince also outsiders from our particular field that such , in principle quite simple experiments are not at all easy. neutron - rich cd isotopes were produced at that time at the old cern / isolde mass separator connected to the 600-mev proton synchro - cyclotron . in this experiment
, a uc@xmath113 target in a graphite - cloth matrix was connected to a plasma - discharge ion source via a heated quartz tansfer line .
this tube was assumed to serve as a kind of thermochromatography column allowing preferential extraction of volatile species , whereas non - volatile elements should be retained . in this way ,
some chemical selectivity was intruduced to the otherwise non - selective plasma ionization .
nevertheless , at a=130 strong isobaric contaminations of @xmath90 in and @xmath90cs were observed which made @xmath114- and @xmath39-spectroscopy of the orders of magnitude weaker produced @xmath90cd impossible . moreover , even with the selective detection method of @xmath0-delayed neutron ( @xmath0dn ) counting , the occurrence of a priori unexpected
[ @xmath115cabr]@xmath116 molecular ions , at a=130 containing the 1.9-s @xmath117br @xmath0dn - precursor , severely complicated the experimental conditions . to optimize the detection conditions for the searched @xmath90cd , data from about 36,000 growth ( 300 ms collections ) and decay ( 900 ms ) cycles were accumulated .
a careful analysis of the resulting complex @xmath0dn - multiscaling curves @xcite yielded after subtraction of the @xmath117br @xmath0dn - component an estimate of the @xmath90cd half - life between maximum 230 ms ( presumably representing a t@xmath97 mixture of cd - mother and in - daughter decays ) and 160 ms ( derived from the growth curve during activity collection , where the cd signals should be enhanced compared to the in - daughter activity ) .
after all these experimental difficulties , it was quite satisfying to see that our experimental half - life value of ( 195@xmath11835 ) ms turned out to be in reasonable agreement with the n@xmath94(z)@xmath109@xmath119@xmath101(z)@xmath120const . waiting - point expectation of ( 180@xmath11820 ) ms , derived independently at that time by hillebrandt ( see , kratz , thielemann , hillebrandt et al .
the correlation found between the @xmath121 in and @xmath90cd half - lives and the observed solar abundances of their stable isobars , which was for the first time exclusively based on experimental numbers , immediately became very important to constrain the equilibrium conditions of an r - process .
and this success strongly motivated further experimental and theoretical nuclear - structure investigations , as well as astrophysical r - process studies @xcite .
for example , at isolde the second n=50 waiting - point isotope @xmath122cu could be identified again using a chemically non - selective plasma - ion source @xcite ; experimentally known strong p@xmath98 branches were shown to be the nuclear - structure origin of the odd - even staggering in the a@xmath6580 n@xmath94 peak @xcite ; and from the interpretation of the @xmath92zn decay scheme in terms of j@xmath123=1@xmath116 , two - quasi - particle ( 2qp ) configurations , evidence for a vanishing of the spherical n=50 shell closure far from stability was obtained @xcite . since the late 1980 s , considerable progress has been achieved in the study of neutron - rich medium- to heavy - mass nuclei at various laboraties , with isolde always playing a leading role in this field . however , due to the generally very low production yields , the occurrance of isobar , multi - charged ion and molecular - ion contaminations from chemically non - selective ionization modes , or because of the application of non - selective detection methods , _ direct _ information on isotopes lying in the r - process path(s ) could be obtained but in a few exceptional cases . from the majority of investigations , originally dedicated to study nuclear - structure developments as a function of isospin , only _ indirect _ but nevertheless also important information for r - process calculations can be deduced ( see , e.g. @xcite ) . in this context , we would like to remind the reader of the fascinating phenomenon of the _ sudden onset and saturation of ground - state deformation at n=60 _
( see , e.g. @xcite ) , which had in fact led to the first astrophysical request of n=82 shell quenching @xcite .
because of the severe experimental problems to identify r - process nuclei , recent progress has undoubtedly benefitted from the * selectivity * in their production and detection , by applying e.g. z - selective laser ion - source ( lis ) systems , isobar separation or multifold - coincidence techniques . today , there are mainly three mass regions , where nuclear - structure information is of particular astrophysical interest .
the first is the seed region of the classical r - process which involves very neutron - rich fe - group isotopes up to the double - magic nucleus @xmath124ni .
we will discuss recent studies in that mass - region in subsect .
the second region of interest is that of far - unstable isotopes of refractory elements ( zr to pd ) around a@xmath65115 .
here , most r - process calculations show a pronounced r - abundance _ trough _ , which is believed to be due to deficiencies at and beyond n=72 mid - shell ( see , e.g. @xcite .
recent experimental information using chemical separation procedures , or ion - guide and projectile - fragmentation techniques combined with mass separation can , for example be found in @xcite .
the third region of strong nuclear - structure and astrophysics interest is that around the double - magic nucleus @xmath125sn , which we will discussed in subsect .
recent spectroscopic studies around the double ( semi- ) magic @xmath126ni ( z=28 , n=40 ) at lisol in louvain - la - neuve , at ganil / lise in caen and also at cern / isolde ( see , e.g. @xcite , and references therein ) , have revealed new structure features which are not easy to reconcile with established shell - model calculations in the neutron - rich region between @xmath127ca and @xmath124ni .
most ( but not all ) of the models find that n=40 is a good closed subshell and predict near - zero deformation for neutron numbers in its vicinity .
of course , these ideas are supported by existing data for @xmath128ni , @xmath129zn , @xmath130ge and @xmath131se isotopes for which there is a slight peak in the 2@xmath116 energies for n=38 and / or the supposed subshell at n=40 , and then a gentle drop in e(2@xmath116 ) beyond n=42 as the @xmath132g@xmath133 orbitals fill and some collectivity is observed ( see fig
. 7 ) .
however , recent experiments at the ps - booster isolde using the high z - selectivity of laser ionization have made possible decay - studies of very neutron - rich @xmath134mn isotopes , and the determination of detailed nuclear - structure information for their @xmath135fe - daughters .
in particular , it was found that
contrary to expectations
the n=40 isotone @xmath136fe , situated only one proton - pair below double ( semi- ) magic @xmath126ni , is already relatively deformed . from the well - known relationship between e(2@xmath116 ) , b(e2 ) values and collectivity @xcite , we deduce a deformation value of @xmath0@[email protected] for @xmath136fe .
as can be seen from fig . 7 , this trend toward collectivity below @xmath126ni
is already observed for @xmath137fe@xmath138 with @xmath0@[email protected] .
needless to say , these data serve as an indication that ` theoretical consensus ' for sphericity in n=40 isotones with z@xmath1728 does not provide a good description of the structure , and consequently also of nuclear masses and decay properties of these nuclides .
subsequent data from the ganil / lise spectrometer @xcite in this mass region are in support of our conclusion that the n=40 and adjacent nuclides below @xmath126ni are relatively deformed , with the maximum collectivity presumably centered around @xmath137cr@xmath139 .
hence , the use of theoretical results that do not account for deformation of neutron - rich isotopes in this mass region as input data for astrophysical calculations must be approached with caution . already with the use of z - selective laser ionization at the general - purpose separator ( gps ) of cern / isolde
, it has been possible to study decay properties of 14-ms @xmath140mn , which is one of the most neutron - rich isotopes with the shortest @xmath0-decay half - life beyond the sd - shell .
the prospects are good for an extension of such measurements to even heavier isotopes with lis and the high - resolution separator ( hrs ) , eventually including the study of detailed level structures up to @xmath141fe . combined with ongoing work on neutron - rich @xmath142cu and @xmath129zn at isolde and on @xmath143co and @xmath128ni at other facilities ,
nuclear - structure information towards the next major neutron - shell closure n=50 will accumulate , to answer e.g. the question about shell - quenching at double - magic @xmath124ni and its ` bottle - neck ' behaviour for the r - process matter flow through n=50 .
moreover , the nuclear properties in this region have gained additional importance in the context of the lately discussion of _ two _ distinct r - process components , a ` weak ' r - process below a@xmath65130 , and the ` main ' r - process beyond a@xmath65130 ( see sect
. 5.2 ) .
the mass region around the far - unstable double - magic nucleus @xmath125sn has long been and still is of particular interest for experimental and theoretical investigations .
apart from astrophysical importance ( formation of the a@xmath65130 peak of the n@xmath94 distribution @xcite ) , this region is of considerable shell - structure interest .
the neutron - rich isotope @xmath144sn@xmath91 with its pronounced magicity ( only comparable to stable @xmath145pb@xmath96 ) , together with the properties of the nearest - neighbour single - particle ( @xmath146sb@xmath91 and @xmath147sn@xmath148 ) and single - hole ( @xmath149sn@xmath150 and @xmath151in@xmath91 ) nuclides are essential for tests of the shell model , and as input for any reliable future microscopic nuclear - structure calculations towards the neutron drip - line .
the bulk of the data so far known in this region have been obtained from @xmath0-decay spectroscopy at the mass - separator facilities osiris ( sweden ) and isolde ( see , e.g. @xcite , and references therein ) .
the structures of @xmath152sn ( @xmath132-hole ) and @xmath153sb ( @xmath154-particle ) are fairly well known since more than a decade .
more recently , the @xmath132-particle states 2f@xmath155 ( g.s . ) , 3p@xmath156 ( 854 kev ) , 1h@xmath133 ( 1561 kev ) and 2f@xmath157 ( 2005 kev ) , as well as tentatively also the 3p@xmath97-particle ( 1655 kev ) and the 1h@xmath158-hole ( @xmath1203700 kev ) states in @xmath153sn@xmath159 have been identified @xcite at the general purpose separator ( gps ) of the new ps - booster isolde facility . from these data , valuable information on the spin - orbit splitting of
the 2f - orbital and tentatively the 3p - state splitting was obtained .
these results were compared to mean - field and hfb predictions , and it was found that none of the potentials currently used in _ ab initio _ shell - structure calculations was capable of properly reproducing the ordering and spacing of these states ( see , e.g. @xcite , where also possible astrophysical consequences are given ) . of particular interest in this context
are the surprisingly low - lying @xmath132p@xmath156- and @xmath132p@xmath97-states in @xmath153sn .
according to the standard nilsson model @xcite , for example , they are expected at 2.89 mev and 4.36 mev , respectively .
such lowering of the energies of low - j orbitals ( here , by as much as 2.0 mev and 2.7 mev , respectively ) seems to occur in different mass regions for very neutron - rich nuclides , and has been interpreted , for example , as monopole shifts of single - particle ( sp ) states@xcite .
the occurrence of such energy shifts of sp levels has recently also been predicted as a _ neutron - skin _ phenomenon to occur only near the neutron drip - line ( see , e.g. @xcite ) , but not yet in ( n@xmath99 + 1 ) @xmath153sn which is still neutron - bound by s@[email protected] mev .
nevertheless , following the suggestion of dobaczewski @xcite , we have modified the nilsson potential by reducing the strength of the @xmath160-term ( i.e. the spin - orbit interaction ) , in order to study its effect on different orbitals . and
indeed , this procedure has led to the desired change in the position and even the ordering of the @xmath132-particle states beyond n=82 , and has thus allowed at least a qualitative reproduction of the experimental observation of low - lying low - j and high - lying high - j orbitals in @xmath153sn .
recently , zhang et al .
@xcite have proposed a new set of nilsson parameters that correctly reproduce the sp levels in @xmath152sn and @xmath153sn .
they compare their new results with the experimental levels and older nilsson parameters , as well as sp levels obtained from relativistic mean - field calculations .
the danger with this approach is that the change in parameters may be compensating for underlying deficiencies in the model .
consequently , while the results do fit the levels of @xmath153sn , they may not be useful for astrophysical uses where the goal is to determine the structure of nuclides significantly further from stability . with these new data ,
the @xmath125sn valence - nucleon region is nearly complete .
the only missing information are the @xmath154-hole states in @xmath151 in , which can be studied through @xmath0-decay of the very exotic nucleus @xmath152cd@xmath159 .
recent lis developments at mainz @xcite and cern / isolde , using a novel frequency - tripling technique for the z - selective resonance ionization of cd , have made possible first test measurements on cd isotopes in the a@xmath65130 region at the gps of the ps - booster isolde .
although the data are not yet fully analyzed , we can already present some interesting new results .
for example , the complex @xmath0dn - curve at a=130 indicates a half - life component of ( 162@xmath1187 ) ms on top of the @xmath90 in @xmath0dn - activity .
we assign this component to the r - process waiting - point isotope @xmath90cd , which we had first identified using non - selective ionization with a somewhat longer half - life @xcite . as weighted average of the two determinations ,
we now favour a t@xmath97=(168@xmath11812 ) ms . in this experiment , it was also possible to measure , for the first time , a half - life for @xmath152cd@xmath159 .
the value of ( 68@xmath1183 ) ms @xcite is considerably shorter , and the estimated @xmath0dn - branch is much lower than might have been expected .
it is possible to estimate a forbidden g.s .- to - g.s .
@xmath0-branch ( @xmath132f@xmath155 @xmath161 @xmath154g@xmath133 ) with a partial half - life of about 120 ms .
this means that the gamow - teller ( gt ) allowed decay must also have a half - life of at least 150 ms , as compared to the earlier qrpa prediction of 943 ms @xcite .
as it is unlikely that the log(ft ) value is smaller , this result indcates a significantly larger @xmath0-energy driving the gt decay than predicted by the qrpa . in turn , this means either a significantly larger q@xmath162 ( meaning a less bound 83@xmath163 neutron ) or a significantly lower energy for the daughter states to be populated in @xmath152 in .
either of these possibilities is an indication that full understanding of the structure of nuclides along the n=82 closed neutron shell is not complete . in any case
, the present results from both detection methods , @xmath0dn - counting and @xmath39-spectroscopy , suggest the need of isobar separation , i.e. the use of the hrs at isolde , in order to discriminate the laser - ionized @xmath107cd from the unavoidable surface - ionized @xmath164 in isobar .
another recent study of nuclear - structure development towards n=82 concerns the @xmath0dn- and @xmath39-spectroscopic measurements of neutron - rich ag nuclides at the ps - booster isolde facility ( @xcite ) , using an improved version of the lis system described in @xcite together with new microgating procedures @xcite .
this approach was of considerable assistance in minimizing the activities from surface - ionized in and cs isobars . in this context , the additional ` selectivity ' of the spin- and moment - dependent hyperfine ( hf ) splitting was used to enhance the ionization of either the @xmath154p@xmath97-isomer or the @xmath154g@xmath133 g.s .- decay of the ag isotopes .
the laser response to hf - splitting was calibrated on the stable j@xmath123=1/2@xmath165 @xmath166ag whose magnetic moment is small ( @xmath167=-0.11 n.m . ) and on the radioactive j@xmath123=7/2@xmath116 isomer that has a large moment of @xmath167=4.4 n.m .
( see upper left part of fig . 8 , and sebastian et al .
subsequently , the @xmath0dn - activities were measured as a function of laser frequency for @xmath168ag up to @xmath169ag ( see fig .
all ag isotopes show
apart from the expected small isotope shift the same global pattern , where the moment of the odd g@xmath133-proton is the dominant source of hf - splitting . for @xmath168ag ,
previous decay - studies had already indicated the presence of both a low- and a high - spin isomer @xcite . and indeed , apart from the highly split peak that can be associated with a ( @xmath154g@xmath133@xmath132h@xmath158)@xmath170 configuration , another narrow peak shows up in the center that could arise from the ( @xmath154p@xmath97@xmath132d@xmath156)@xmath171 configuration ( see fig .
subsequently , both @xmath0dn- and @xmath39-decay from @xmath168ag was studied as a function of laser frequency . as is evident from the partial @xmath39-spectra shown in the lower part of fig .
9 , when the laser is set at the center of the frequency scale , @xmath39-rays from both isomers can be observed .
however , when the laser frequency is shifted away from the center ( to 35 units ) , ionization of the low - spin isomer is suppressed , and the @xmath39-spectrum indicates that only the decay of the high - spin isomer is oberved .
similarly , when following the @xmath0dn - decay curves as a function of laser frequency , different half - lives are obtained which represent varying admixtures of the two isomers . from a preliminary analysis
, we get a half - life of the low - spin isomer of t@xmath97@xmath65550(50 ) ms and a value of t@xmath97@xmath65200(50 ) ms for the high - spin isomer of @xmath168ag @xcite .
this experiment clearly demonstrates the sensitivity of ionization to laser frequency and marks the first spectroscopic application of this technique to short - lived isomers . in the meantime
, a gross analysis of @xmath39-spectra taken at different laser - frequency settings confirms isomerism also for the heavier ag isotopes up to ( at least ) @xmath172ag . an important nuclear - physics quantity for r - process nucleosynthesis calculations ( in particular for the r - matter flow through the a@xmath65130 region and the n@xmath94 peak shape )
is the @xmath0-decay half - life of the n=82 waiting - point isotope @xmath103ag , situated just below @xmath90cd .
qrpa calculations using different q@xmath101-values and sp wave functions resulted in decay rates between 15 ms and 170 ms @xcite .
this had left large uncertainties in the reproduction of the a@xmath65130 peak ( see fig . 1 in @xcite , and also fig . 4 in @xcite ) .
initial attempts @xcite to observe the @xmath0dn - decay of @xmath103ag , performed with the ( broad - band ) lasers centered with respect to the mean frequency for the stable @xmath154p@xmath97 @xmath166ag , had failed .
after the observation of enhanced ionization of the @xmath154g@xmath133 level at an off - center frequency , a new experiment was performed with the laser setting at the left @xmath154g@xmath133 peak value observed for @xmath169ag ( see lowest right part of fig .
this approach together with the microgating procedure @xcite mentioned above , finally permitted the unambiguous identification of the @xmath0dn - decay from @xmath154g@xmath133 @xmath173ag ( see fig .
the half - life of 46@xmath174 ms is in very good agreement with the recent qrpa prediction of 47 ms @xcite , but is lower than our old waiting - point requirement of about 130 ms @xcite .
however , that estimate was based on the earlier half - life measurement of t@xmath97@xmath65195 ms for @xmath90cd , and on the older r - process residuals n@xmath94 of @xcite
. however , for a non - equilibrium r - process , or after the breakdown of the @xmath0-flow equilibrium , also the t@xmath97 contribution from the @xmath154p@xmath97 @xmath103ag isomer would be of importance for the ` stellar ' half - life . with the qrpa model of @xcite
, it is possible to calculate the gt decay for such an isomer to t@xmath97@xmath65320 ms . when including an estimate for the expected first - forbidden strength ( extrapolated from the decay of the j@xmath123=1/2@xmath165 isomer in isotonic @xmath152 in ) , a minimum value of t@xmath97@xmath65125 ms is suggested . and ,
indeed , a careful re - examination and comparison of the a=129 @xmath0dn - decay curve taken with the laser at central frequency with the pure @xmath103 in curve from a laser - off run , gave a first indication of a weak , longer - lived @xmath103ag @xmath0dn - component with a half - life of roughly 160 ms .
now , it will be important to use isomer - specific laser ionization in combination with isobar separation at isolde - hrs to ascertain the existence of this @xmath154p@xmath97 isomer . the astrophysical half - life of @xmath103ag for non - equilibrium conditions , expected to be a mixture of the g.s and isomer values ,
might then lead to a better understanding of the bottle - neck behaviour and shape of the a@xmath65130 n@xmath94 peak . based on our present knowledge on level systematics in the @xmath125sn region , neutron - capture @xmath39-decay from s@[email protected] mev in @xmath103ag
would populate the @xmath154p@xmath97 isomer to roughly 35@xmath175 and the @xmath154g@xmath133 ground state to about 65 @xmath175 @xcite .
this would result in an average stellar t@xmath97@xmath6580 ms .
however , already within the waiting - point approximation requesting only the t@xmath97 of @xmath173ag , an excellent reproduction of the shape of the n@xmath94 peak from a=126 to a=133 is obtained . from this result , we conclude that the effect of neutrino - processing of the n@xmath176 during freeze - out is considerably smaller for the a@xmath65130 peak region than recently postulated by qian et al .
@xcite , and that the main effect is rather due to @xmath0dn - branching during the first 100 ms of the freeze - out .
in addition to the study of gross @xmath0-decay properties of neutron - rich ag isotopes , also the level systematics of the cd - daughters has been extended up to the r - process path @xcite .
first , an alternate source of nuclear structure for neutron - rich isotopes is the investigation of correlated @xmath39-emission of complementary fragments of spontaneous - fission systems .
recent data from such studies have revealed surprisingly low 2@xmath116 to 6@xmath116 energies in @xmath177sn ( only two neutrons beyond double - magic @xmath125sn ) , that suggest a reduction of the effects of pairing in very neutron - rich sn nuclides @xcite .
for the cd isotopes , however , such fission data extend only up to @xmath168cd .
our studies at isolde have now permitted even - even cd structures to be extended six neutrons further out , up to n=80 @xmath178cd .
the new results are shown in fig .
11 , together with known e(2@xmath116 ) and e(4@xmath116)/e(2@xmath116 ) level systematics of neighbouring even - z elements .
the e(4@xmath116)/e(2@xmath116 ) ratio for the 4n - hole nuclide @xmath172cd is 2.25 , a value almost unchanged relative to that for the 6n - hole isotope @xmath179cd .
already this static ratio is in contrast to the larger reduction observed for the z=52 isotones @xmath178te and @xmath90te . for the new 2n - hole nuclide @xmath178cd
, the energy of the first 2@xmath116 state at 645 kev is even * lower * than the e(2@xmath116 ) of 657 kev in @xmath172cd , and also the e(4@xmath116)/e(2@xmath116 ) ratio of 2.22 is only insignificantly smaller than those in @xmath179cd and @xmath172cd . as can be seen from the right side of fig . 11 , this trend for the z=48 isotopes clearly deviates from that of the z=50 ( sn ) , 52 ( te ) and 54 ( xe ) isotones .
this situation is similar to the hg isotopes below n=126 , where the e(2@xmath116 ) in the 2n - hole nuclide @xmath180hg@xmath181 is with 436.6 kev also slightly lower than the e(2@xmath116)=439.6 kev in the 4n - hole isotope @xmath182hg@xmath183 . for the ee - cd nuclides
this indicates an extension of the well - known vibrational character of the lighter isotopes up to n=82 , indicating a weakening of the spherical shell strength below @xmath125sn . for a=130
, we do have an indication for two weak @xmath39-lines at 957 kev and 1395 kev with a tentative assignment to the 4@xmath116 and 2@xmath116 levels in @xmath90cd .
if confirmed , this would result in an e(4@xmath116)/e(2@xmath116 ) ratio of 1.69 . as the above results were accomplished with lis and gps - isolde , again there is promise of additional data with hrs . in particular , a study of the decay of isobaric ` clean ' @xmath90ag should permit the identification of excited states of n=82 @xmath90cd up to the 8@xmath116 level in which two aligned g@xmath133 proton holes are coupled to double - magic @xmath125sn . as the structure of @xmath184cd , which has the same two g@xmath133 aligned protons coupled to the double - magic @xmath185sn core ,
is known , a direct comparison is already available for such a measurment at n=82 .
there also exists the possibilty of the presence of a low - spin @xmath154p@xmath97 isomer in @xmath152ag that should populate the @xmath132p@xmath156 state in the @xmath152cd - daughter . as has been mentioned previously , one possible effect of large neutron excess in nuclei
is the lowering of the low - j orbitals .
hence , an hrs study of @xmath152ag decay could demonstrate if the 854-kev gap between the @xmath132p@xmath156 first excited state and the @xmath132f@xmath155 g.s
. in @xmath153sn@xmath159 has narrowed with the removal of two protons in the n=83 ag isotone . under development and testing at isolde
is a scheme for laser ionization of sn nuclides @xcite .
if past experience is any guide , again , it should be possible with gps and hrs to produce and study the decay of at least six sn isotopes beyond @xmath177sn , the heaviest sn nuclide for which some nuclear properties are known .
@xmath186sn is a nucleus with 50 protons and 90 neutrons , a neutron / proton ratio of 1.8
. it would be the most neutron - rich nuclide studied beyond the sd shell ; and if there are to be strong divergences from expected nuclear properties because of the high neutron / proton ratio , the properties of this nuclide should reveal such signatures .
moreover , these heavy sn nuclides lie directly in the path of the r - process , and the degree to which they capture neutrons prior to undergoing @xmath0-decay is a measure of how quickly the r - process flow can be established beyond the waiting - point nuclides at n=82 . at this point , all nucleosynthesis calculations use predicted t@xmath97 values which are presumably too long .
given the recent , highly interesting results in the @xmath125sn region , one is tempted to re - inspect existing data for additional hidden or so far unrecognized signatures of shell quenching at n=82 . and
indeed , first qualitative evidence for such a phenomenon came already from a comparison of the old t@xmath97 measurement of @xmath90cd@xmath91 @xcite with predictions from the original qrpa model of mller and randrup @xcite .
when applying q@xmath162-values from the unquenched mass models frdm @xcite and etfsi-1 @xcite , theoretical t@xmath97 values for gt - decay of 1.12 s and 674 ms are obtained , respectively . however , with the quenched mass formulae hfb / skp @xcite and etfsi - q @xcite , shorter values of 246 ms and 364 ms , respectively , are derived which are in better agreement with experiment .
another , so far unrecognized indication is given by the measured masses in the @xmath125sn region @xcite .
as can be seen from fig .
12 , between the sequence of @xmath93sn and @xmath107cd isotopes there is a significant change in the trend of the experimental and theoretical mass differences ( normalized to the frdm predictions , m@xmath187-m@xmath188 ) . clearly , for the cd isotopic chain
the best agreement with the measured masses is obtained with the quenched etfsi - q .
all these signatures can , of course , only be taken as first evidence for a quenching of the spherical n=82 shell below @xmath125sn .
certainly , much more experimental data , including nuclear masses ( and , as e.g. shown in fig . 4 of @xcite , in particular s@xmath98 values of n=81 and 83 isotones ) , single - neutron levels , as well as spectroscopic factors of transfer and pickup reactions on radioactive isotopes around @xmath125sn , are necessary to quantify the effect of shell quenching .
as will be shown in the next section , a vanishing of the classical , spherical n=82 and 126 shell closures would not only be an interesting new nuclear - structure phenomenon close to the neutron drip - line , but may also have important consequences for r - process nucleosynthesis up to the th , u , pu cosmochronometers , and maybe even beyond ( see , e.g. @xcite ) .
mainly as a result of the improved nuclear - physics input , our r - process parameter studies have revealed that the entire isotopic n@xmath94 pattern can not be reproduced by assuming a global steady flow .
instead , even in a single sn ii event it requires a superposition of a multitude of components ( minimum three ) with different neutron densities or equivalenty different s@xmath98 s , characterizing different r - process paths and time scales . in our approach , the weighting of the individual r - components
is naturally given by the @xmath0-decay half - lives of the three waiting - point nuclei @xmath92zn , @xmath90cd and @xmath95tm which represent the ( main ) progenitors of the stable isobars @xmath92se , @xmath90te , @xmath95pt situated at the top of the respective n@xmath94 peaks .
that this approach has , indeed , a nearly one - to - one mapping to ` realistic ' astrophysical environments with time variations has been discussed in some detail in @xcite . in fig . 13 , we show global n@xmath189 distributions from a superpostion of sixteen n@xmath190 components for two versions of the etfsi nuclear mass model ( eftsi-1 @xcite and etfsi - q @xcite ) . in both cases ,
identical conditions for the stellar parameters were used . within this framework ,
the successful reproduction of the position and relative height of the n@xmath94 peaks as well as the remaining defiencies have been interpreted as signatures of nuclear structure near the neutron drip - line @xcite .
it should be mentioned at this point , that this interpretation is in contradiction with the conclusions of other authors ( see , e.g. @xcite ) who prefer a mathematical ` fourier - type ' analysis , assuming that the n@xmath94 pattern represents a ` multi - sn - event ' composition . as can be seen from fig . 13 ( dotted fit - curve ) , with the etfsi-1 mass model , apart from pronounced a@xmath65115 and a@xmath65175 abundance troughs , too little r - process material is observed in the whole region beyond the a@xmath65130 peak .
the solution to the a@xmath65115 r - abundance deficiency clearly lies in the prediction of nuclear properties along the r - process path in this mass range , which have their origin in an overestimation of the n=82 shell strength below @xmath125sn ( for a more detailed discussion see @xcite ) .
analogous deficiencies in describing shape - transitions around n@xmath65104 midshell and the n=126 shell strength seem to be the origin of the deep a@xmath65175 r - abundance trough .
it has therefore been suggested @xcite , that the trough problem might be resolved if for very neutron - rich nuclei the n=82 and 126 shell gaps are less pronounced than predicted by the global mass models @xcite used so far in r - process calculations .
a weakening of sperical shells near stability with increasing isospin , resulting in a gradual setting in of collectivity , is well established for the n=20 and 28 shells ( see , e.g. @xcite ) , and is also indicated at n=50 @xcite .
quenching of spherical shells at n=82 and 126 has recently been predicted by hfb calculations @xcite , and the infinite nuclear matter model @xcite . in order to check our predictions , in a first step we have replaced the frdm masses by the s@xmath98 s from the spherical hfb / skp model @xcite locally in the vicinity of the n=82 shell closure . with this
, we could indeed observe a considerable improvement of the calculated r - abundances in the ` pathological ' a@xmath65115 region ( for a more detailed discussion , see @xcite and figs . 2 and 3 therein ) .
motivated by this success , in a next step we have replaced the spherical hfb / skp masses by the global etfsi - q model which takes into account both deformation and shell quenching @xcite . the full curve in fig .
13 shows the result of a multicomponent fit with the latter masses . when compared to the n@xmath191 curve obtained with the etfsi-1 formula , a considerable improvement of the overall fit is observed .
in particular the prominent abundance troughs in the a@xmath65115 and 175 regions are eliminated to a large extent .
hence , a ` quenched ' mass model for r - abundance calculations beyond a@xmath65110 seems to be highly recommended , especially if predictions for the a@xmath192200 mass region are required .
of special interest in this context is the good agreement for the region between @xmath193tl and @xmath194bi with the recent n@xmath94 evaluation of @xcite . for @xmath194bi , for example , it indicates a nearly pure r - process origin of this isotope , which confirms that ( at least in the frame of the double - pulse s - process model ) only a minor contribution to this isotope is of s - origin . as a consequence ,
there is no further need for the so - called ` strong ' s - process component , which had been introduced originally to account for the pb and bi abundances .
this improvement with the etfsi - q masses should also make extrapolations up to the progenitors of the long - lived actinides @xmath195th and @xmath196u more reliable , in particular since they are for the first time based on an internally consistent nuclear - physics input . for galactic - age determinations , commonly production ratios of cosmochronometers , in particular @xmath195th/@xmath197u ,
are applied . in @xcite , we have compared the results derived from different mass models to literature values . only when using nuclear masses which correctly describe the entire n@xmath94 pattern , consistent values for a galactic age of 10 - 14 gyrs
could be obtained when following the procedure described in @xcite . recently , sneden et al .
@xcite have determined elemental abundances of 17 heavy elements between ba and th for the ultra - metal - poor , neutron - capture - rich halo star cs 22892052 . in fig .
14 , these values are compared to the solar r - element distribution and to our etfsi - q predictions , after adjustment to the solar metallicity .
the very good agreement clearly indicates , that the heavy elements in this star are of pure r - process origin .
with the above measured th abundance and our calculated initial abundance , a lower limit for the age of this star can be deduced when normalizing to a stable r - only element , e.g. eu . from the observed ratio th / eu=0.219 in this halo star , combined with our etfsi - q prediction of th / eu=0.480 for the initial ratio , a lower limit of its ` decay age ' of 14.5 gyrs is obtained @xcite , which overlaps with recent galactic ages on the young side ( see , e.g. @xcite ) .
this implies that the heavy elements in cs 22892052 must have been synthesized already very early in the galactic evolution , prior to the onset of s - process nucleosynthesis @xcite .
finally , the excellent agreement between the r - element abundances of this first - generation star and the global n@xmath94 pattern , which represents a summation over the whole galactic evolution , indicates that there is probably a _ unique _
scenario for the r - process beyond a@xmath65130 . in the meantime confirmed by observations of other metal - poor stars @xcite , this result seems to validate our deductive approach of deducing unique astrophysical constraints on r - process conditions from fitting the global n@xmath94 abundances .
only a few months ago , cowan et al .
@xcite have presented new element - abundance data for the halo - star cs 22892 - 052 obtained at the keck i high - resolution spectrograph .
they partly supercede the earlier analysis of the same group @xcite , and in addition give first results on the abundances between @xmath106nb and @xmath107cd . an interesting feature of the latter element pattern is the pronounced odd - even - z staggering .
all observed odd - z elements below z=50 ( @xmath198y , @xmath106nb , @xmath199rh and @xmath200ag ) are clearly under - abundant compared to the ( metallicity - scaled ) solar pattern , whereas the even - z elements ( @xmath139zr , @xmath201ru , @xmath202pd and @xmath107cd ) are within the given errors only slightly under - abundant ( see fig .
these new ` low - z ' data seem to support the recent evidence for a two - source nature of the solar r - process , originally concluded from extinct radioactivities ( 1.57@xmath20310@xmath204-y @xmath103i and 9@xmath20310@xmath205-y @xmath206hf ) in meteorites @xcite . taking advantage of our ` site - independent ' , exponential multicomponent approach to fit the n@xmath94 pattern , we now can test within the waiting - point assumption under which stellar conditions ( mainly neutron density and process duration ) the possible two r - processes have to run , in order to reproduce the ` low - z ' ( below z@xmath6550 ) and ` high - z ' ( beyond z@xmath6550 ) element - abundances of cs 22892 - 052 .
when assuming that the abundances of this extremely old metal - poor halo star are a living record of the first ( few ) generation(s ) of galactic nucleosynthesis @xcite , the observed pattern beyond zr up to th ( the trend of the lighter elements up to zr may be explainable in terms of the weak s - process ) should most likely be produced by only * one * r - process site .
this scenario ( the _ ` main ' _ r - process ) then produces the ` low - z ' elements under - abundant compared to solar , and reaches the solar r - pattern presumably beyond @xmath107cd .
14 shows a fit to the recent cs 22892 - 052 data , using our standard multicomponent approach @xcite with the etfsi - q nuclear masses . the trend for the ` low - z ' elements pronounced odd - even staggering , and approaching the solar values with increasing z
is nicely reproduced , and the good overall reproduction of the ` high - z ' elements ( beyond @xmath207ba ) is maintained .
when starting from an fe - group seed , this requires a * minimum * neutron density of n@xmath98@xmath19210@xmath111 @xmath74 in order to simulate the observed partial ` drying - out ' of the r - matter flow at low z. in terms of radiation entropy within the realistic scenario of the high - entropy wind in snii ( see sect .
5.3 ) , this would translate to s@xmath192200 k@xmath208 for y@xmath209=0.45 @xcite . as a consequence of the present observations for
cs 22892 - 052 ( but also for other halo stars ) and our calculations , the ` residuals ' at low z relative to the solar distribution , i.e. n@xmath94 - n@xmath210 , will require a separate _ ` weak ' _ r - process component of yet unknown stellar site .
again , when assuming an fe - group seed , our calculations can reproduce also the resulting n@xmath211 pattern up to @xmath200ag for neutron densities of 10@xmath73@xmath212n@xmath98@xmath21210@xmath110 at t@xmath100=1.35 and a process duration of @[email protected] s. a choice of @xmath139zr as seed ( in order to avoid the delay in the a@xmath6580 bottle - neck region ) , would speed up the low - z production to @[email protected] s. these are conditions that might be reached in explosive shell - burning scenarios ( see e.g. @xcite ) .
but , before being able to draw more definite conclusions , considerably more work has to be done in both astrophysics ( site , seed , n@xmath98,@xmath213 ) and nuclear physics ( masses , @xmath0-decay properties , shell - quenching at n=50 ) . after having tested nuclear properties with a site - independent approach in the previous subsection ,
we want to return briefly to the possible astrophysical r - process sites introduced already in section 1.2 , i.e. snii and neutron - star ( ns ) mergers . in the first case ,
a wind of matter from the neutron star surface ( within seconds after a successful supernova explosion ) is driven via neutrinos streaming out from the still hot neutron star @xcite .
this high entropy ` neutrino wind ' ( corresponding to conditions plotted in the left part of fig .
1 ) is leading to a superposition of ejecta with varying entropies .
an abundance pattern resulting from a superposition of such high - entropy environments , representing @xmath4-rich freeze - outs of various degrees , is shown in fig .
15 @xcite .
the high entropies ( up to 400 @xmath214/nucleon , responsible for the mass region a@xmath19110 ) reproduce nicely the n@xmath94 pattern , with the exception of the mass region 110120 , where the same conclusions can be drawn as in the site - independent studies ( see sect .
the trough is the result of the nuclear - structure deficiencies of the etfsi-1 mass model below n=82 .
the lower entropies are responsible for the abundances in the mass region 80110 .
we see that the predictions for that region do not reproduce the solar r - abundances .
the reason is that essentially no neutrons are left after an @xmath4-rich freeze - out with such entropies , and the abundance pattern is dictated by @xmath4-separation rather than neutron - separation energies .
thus , explaining the r - process by ejecta of snii faces two difficulties : ( i ) whether the required high entropies for reproducing heavy r - process nuclei can really be attained in supernova explosions has still to be verified , ( ii ) the mass region 80110 can not be reproduced adequately .
it has to be seen , for example , whether the inclusion of non - standard neutrino properties @xcite can cure both difficulties .
given the galactic occurence frequency , snii would need to eject about 10@xmath215 m@xmath216 per event .
an alternative site are neutron - star mergers .
interest in a scenario where a binary system , consisting of two neutron stars ( ns - ns binary ) , looses energy and angular momentum through the emission of gravitational waves comes from various sides .
such systems are known to exist ; five ns - ns binaries have been detected by now @xcite .
the measured orbital decay gave the first evidence for the existence of gravitational radiation @xcite .
further interest in the inspiral of a ns - ns binary arises from the fact that it is the prime candidate for a detection by the gravitational wave detector facilities that will come into operation in the very near future .
a merger of two ns may also lead to the ejection of neutron - rich material , and could be a promising site for the production of r - process elements .
it is even possible that such mergers account for _ all _ heavy r - process matter in the galaxy @xcite .
the decompression of cold neutron - star matter has been studied @xcite ; however , a hydrodynamical calculation coupled with a complete r - process calculation has not been undertaken , yet .
16 shows the composition of ejecta from a ns merger with an assumed superposition of components with @xmath23=0.08 - 0.14 , as expected when an average of spin orientation in merger events is taken @xcite .
it is seen that the large amount of free neutrons ( up to n@xmath98@xmath6510@xmath217 @xmath74 ) available in such a scenario leads to the build - up of the heaviest elements and also to fission cycling within very short timescales , while the flow from the fe - group to heavier elements ` dries out ' .
this leads to a complete composition void of abundances below the a@xmath65130 peak .
if this tendency is confirmed , e.g. by further observations of z@xmath1750 elements in very low metallicity stars , it would provide strong support for that r - process site , but would definitely require an additional astrophysical source which produces the bulk of the lighter r - abundances up to a@xmath65125 . the rate of ns mergers has been estimated to be of the order @xmath218y@xmath219 per galaxy @xcite ; more recent estimates @xcite tend towards the higher end of this range ( @xmath220y@xmath219 per galaxy ) .
hydrodynamic simulations of ns mergers are a formidable task . beyond 3d hydrodynamics
, it should include general relativitistic effects , employ a realistic equation of state , contain neutrino transport and neutrino cooling , as well as all possible nuclear reactions @xcite .
we @xcite have performed extensive ns - ns merger calculations and find typically ejecta of the order @xmath221m@xmath216 .
this is sufficient to reproduce the solar system r - process abundances , provided that this matter is converted into an r - process abundance pattern ( which still needs to be shown ) .
we have shown in this review which nuclear properties are of prime importance in the astrophysical r - process . due to the ( partial ) equilibrium nature at high temperatures and neutron densities , the major aspects to explore are nuclear masses and @xmath0-decay properties ( half - lives , delayed neutron emission and fission ) .
the application of recent experimental findings to ` site - independent ' r - process calculations has shown that the n@xmath94 abundances actually bear the possibility to test the evolution of nuclear structure and shell effects as a function of the distance from @xmath0-stability . in order to put these indications on a more solid footing , the experiments have to be extended further towards the neutron drip line , which sets strong requirements for future beam intensities and thus target techniques as well as experimental methods .
innovation has played a strong part in establishing a premier role for isolde in the study of properties of nuclides very far from stability .
the development of the techniques of laser ionization is one example of this kind of innovation .
a strong lesson from lis work is that _ selectivity _ is the key to extending the current results to even more exotic nuclides .
ultimately , the combination of z - selective laser ionization , isomer separation via hf - splitting and the use of the hrs for isobar separation promises enhanced selectivity .
astrophysical knowledge is not at all in a more advanced stage .
we have presented two ` realistic ' sites for the r - process , ( i ) supernovae of type ii and ( ii ) neutron star mergers , which represent the two options to obtain a high neutron to seed - nuclei ratio : high entropies with moderate @xmath23 or neutron to proton ratios , or low entropies with small @xmath23 or high neutron to proton ratios .
the first scenario is ( as is the sn explosion mechanism ) still riddled with the question whether the required high entropies of more than 200 @xmath222 per nucleon can be attained at all and that the lower entropy components give rise to the overproduction of non - solar r - patterns in the mass range a@xmath6580110 .
possibly only exoctic neutrino properties ( like sterile neutrinos ) might save this environment .
the ns mergers could overcome these problems , if the amount of ejected matter stays the same in models with general relativity @xcite , and a statistical average over neutron - star spins permits that matter in the range of @xmath23=0.1 is predominantly ejected @xcite .
an interesting feature of the second environment is that nuclei with masses below the a@xmath65130 peak are essentially unproduced ( see fig .
the fact that both scenarios are unable to reproduce the n@xmath94-`residuals ' in the a@xmath6580120 region , might be related to the increasing evidence of a two - source nature of the solar r - process , which was recently concluded from extinct radioactivities in meteorites @xcite and the low abundances of the odd - z elements @xmath198y to @xmath200ag measured in cs22892 - 052 @xcite , and in other low - metallicity stars ( which show for all elements heavier than ba ( z=56 ) a solar r - process pattern @xcite ) .
further detailed observations of r - process abundances in low - metallicity stars , especially for a@xmath17130 are highly needed for uncovering the nature and stellar site of this second ( _ ` weak ' _ ) r - process and fixing the time in galactic evolution when this nucleosynthesis process sets in .
the authors wish to acknowledge collaboration with many colleagues , in particular p. mller , j. dobaczewski , j.m .
pearson , b.a .
brown , w. hillebrandt , j.w .
truran , ch .
freiburghaus , t. rauscher , s. rosswog , j.j .
cowan , v.n .
fedoseyev , v.i .
mishin , h. gabelmann , g. lhersonneau , a. whr , w. bhmer , h.l .
ravn and the isolde collaboration , and all our present and former phd students and postdocs .
support for this work was provided by various grants from the german bmbf and dfg , the swiss nationalfonds and u.s .
burbidge et al . , rev . mod . phys . 29 ( 1957 ) , 547 .
cameron , atomic energy of canada , ltd . , crl-41 ( 1957 ) .
seeger et al . , ap .
97 ( 1965 ) , 121 .
cameron et al . ,
cern , 70 - 30 ( 1970 ) , 735 .
w. hillebrandt , space sci .
21 ( 1978 ) , 639 .
mathews and r.a .
ward , rep .
48 ( 1985 ) , 1371 .
cowan et al . , phys . rep .
208 ( 1991 ) , 267 .
k . thielemann et al . , in : _ nuclear and particle astrophysics _ , cambridge university press ( 1998 ) , 27 ; available via www at url : http://xxx.lanl.gov/astro-ph/9802077 g. wallerstein et al .
69 ( 1997 ) , 995 .
f. kppeler et al . , ann .
nucl . part .
48 ( 1998 ) , 175 .
woosley et al . , ap .
j. 433 ( 1994 ) , 229 .
k. takahashi et al . , a&a 286 ( 1994 ) , 857 .
c. freiburghaus et al . , ap .
j. 516 ( 1999 ) , 381 .
a. mcwilliam et al . , a. j. 109 ( 1995 ) , 2757 .
a. mcwilliam , ann .
35 ( 1997 ) , 503 .
lattimer et al . , ap .
j. 213 ( 1977 ) , 225 .
b. s. meyer , ap .
j. 343 ( 1989 ) , 254 .
d. eichler et al . ,
nature 340 ( 1989 ) , 126 .
s. rosswog et al .
, a&a 341 ( 1999 ) , 499 .
woosley et al . , ap .
j. suppl . 26
( 1973 ) , 231 .
hix and f .- k .
thielemann , ap .
j. 460 ( 1996 ) , 869 .
hix and f .- k .
thielemann , ap .
j. 511 ( 1999 ) , 862 .
hoffman et al .
, ap . j. ( august 1999 ) , in press .
thielemann et al .
, ap . j. 460 ( 1996 ) , 408 .
h. schatz et al . , phys . rep .
294 ( 1998 ) , 167 .
f. rembges et al .
j. 484 ( 1997 ) , 412 .
woosley and r.d .
hoffman , ap .
j. 395 ( 1992 ) , 202 .
hoffman et al . , ap .
j. 460 ( 1996 ) , 478 .
hoffman et al . , ap .
j. 482 ( 1997 ) , 951 .
p. mller et al . , at .
data nucl .
data tables 59 ( 1995 ) , 185 .
cameron et al . , astrophys .
space sci .
91 ( 1983 ) , 235 . v. bouquelle et al . ,
a&a 305 ( 1996 ) , 1005 .
kratz et al .
, j. phys .
g 14 ( 1988 ) , 331 .
kratz , rev .
astronomy 1 ( 1988 ) , 184 .
kratz et al . , ap .
j. 402 ( 1993 ) , 216 .
f. kppeler et al . , rep .
52 ( 1989 ) , 945 .
h. beer et al .
j. 474 ( 1997 ) , 843 .
thielemann et al .
a570 ( 1994 ) , 329c .
kratz et al . , z. phys .
a325 ( 1986 ) , 489 .
e. lund et al . , phys .
34 ( 1986 ) , 614 .
gill et al . , phys .
56 ( 1986 ) , 1874 .
p. mller et al . , at .
data nucl .
data tables 66 ( 1997 ) , 131 . j. krumlinde and p. mller , nucl .
phys . a417 ( 1984 ) , 419 .
p. mller and j. randrup , nucl .
phys . a514 ( 1990 ) , 1 .
y. aboussir et al , . at .
data nucl .
data tables , 61 ( 1995 ) , 127 .
pearson et al .
b387 ( 1996 ) , 455 .
j. dobaczewski et al . , phys .
( 1994 ) , 981 .
j. dobaczewski et al . , phys .
scr . t56 ( 1995 ) , 15 .
myers , at .
data nucl .
data tables 17 ( 1976 ) , 411 .
h. von groote et al . , at .
data nucl .
data tables 17 ( 1976 ) , 418 .
k. takahashi et al . , at .
data nucl .
data tables 12 ( 1973 ) , 101 .
g. audi et al .
a624 ( 1997 ) , 1 .
kratz et al .
, t@xmath97 and p@xmath98 evaluation , kch mainz report ( 1998 ) .
b. chen et al .
b355 ( 1995 ) , 37 .
kratz et al . , nucl .
a630 ( 1998 ) , 352c .
b. pfeiffer et al .
, z. phys .
a357 ( 1997 ) , 235 . c. freiburghaus et al .
a621 ( 1997 ) , 405c .
kratz et al .
, il nuovo cimento a111 ( 1998 ) , 1043 . t. tachibana et al .
, progr .
phys . 84 ( 1990 ) , 641 .
hilf et al . , suppl .
to proc .
. conf .
nffs-3 , cern 76 - 13 ( 1976 ) , 142 .
j. dobaczewski et al . , phys .
rev c53 ( 1996 ) , 2809 .
meyer et al . , ap . j. 399 ( 1992 ) , 656 .
kratz et al . , z. phys .
a340 ( 1991 ) 419 .
kratz et al . , z. phys .
a336 ( 1990 ) 357 .
kratz et al . , phys .
c38 ( 1988 ) , 278 .
rep . in phys . ,
springer , isbn 0 - 387 - 50120 - 7 ( 1988 ) .
, inst . of phys .
132 , iop , isbn 0 - 7503 - 0262 - 3 ( 1993 ) . , edition frontires , isbn 2 - 86332 - 186 - 2 ( 1995 ) .
s. schoedder et al . , z. phys .
a352 ( 1995 ) , 237 .
t. mehren et al .
77 ( 1996 ) , 458 .
g. lhersonneau et al .
phys . j. a1 ( 1998 ) , 285 . ,
bellaire , usa ; aip conf .
proc . 455 ( 1998 ) .
s. franchoo et al . , phys .
81 ( 1998 ) , 3100 .
o. sorlin et al . , nucl .
a632 ( 1998 ) , 205 .
r. grzywacz et al . , phys .
81 ( 1998 ) , 766 .
m. hannawald et al .
82 ( 1999 ) , 1391 .
s. raman et al .
c43 ( 1991 ) , 556 .
o. sorlin et al . , submitted to nucl .
phys . a f. azaiez and o. sorlin , nucl .
news 8 ( 1998 ) , 34 .
p. hoff et al .
77 ( 1996 ) , 1020 .
m. sanchez - vega et al .
80 ( 1998 ) , 5504 .
b. fogelberg et al . , in : _ exotic nuclei and atomic masses - enam98 _ , aip conf .
( 1998 ) , 502 , 552 , 785 .
stone et al . , phys .
78 ( 1997 ) , 820 .
zhang et al . , z. phys .
a358 ( 1997 ) , 9 .
fedoseyev et al . , z. phys .
a353 ( 1995 ) , 9 .
t. kautzsch et al . , phys .
c54 ( 1996 ) , 2811 .
kratz et al . , in : _ fission and properties of neutron - rich nuclei _ ,
world scientific , ( 1999 ) , in print .
t. rauscher et al . , phys .
c57 ( 1998 ) , 2031 .
i. ragnarsson and r.k .
sheline , phys .
29 ( 1984 ) , 385 .
walters , in : _ fission and properties of fission - product nuclides _ , aip conf .
( 1998 ) , 196 . jing - ye zhang et al . ,
c58 ( 1998 ) , 2663 .
n. erdmann et al . , appl .
b66 ( 1998 ) , 431 .
m. hannawald et al .
, to be published .
kratz , in : _ exotic nuclei and atomic masses - enam98 _ , aip conf .
( 1998 ) , 827 .
mishin et al . , nucl .
instr . meth .
b73 ( 1993 ) , 550 .
y. jading et al .
instr . meth .
b126 ( 1997 ) , 76 .
v. sebastian et al .
, in : _ exotic nuclei and atomic masses - enam98 _ , aip conf .
( 1998 ) , 126 .
zamfir et al . , phys .
c51 ( 1995 ) , 98 .
t. kautzsch , phd thesis , univ .
mainz ; and to be published .
kratz , in : _ nuclei in the cosmos iii _ , aip conf .
proc . 327 (
1995 ) , 113 .
g. walter et al . , a&a 167 ( 1986 ) , 186 .
t. rauscher ( 1998 ) , private communication .
qian et al . ,
j. 494 ( 1998 ) , 285 .
zhang et al . , z. phys .
a358 ( 1997 ) , 9 .
cowan et al . ,
j. 521 ( 1999 ) , in press ; available via www at url : http://xxx.lanl.gov/astro-ph/9808272 a.j .
westphal et al . ,
nature 396 ( 1998 ) , 50 .
orr et al . , phys .
b258 ( 1991 ) , 29 .
o. sorlin et al . , phys .
c47 ( 1993 ) , 2491 .
werner et al . , phys .
b333 ( 1994 ) , 303 .
j. dobaczewski , in : _ extremes of nuclear structure _
; gsi , issn 0720 - 8715 ( 1996 ) , 214 .
brown , phys .
c58 ( 1998 ) , 220 .
l. satpathy and r.c .
nayak , j. phys .
g : nucl . part .
( 1998),1527 . c. sneden et al . , ap .
j. 467 ( 1996 ) , 819 .
m. salaris and a. weiss , a&a 327 ( 1997 ) , 107 .
cowan et al . , 1999 ,
194th aas meeting , chicago , # 67.04 g.j .
wasserburg et al . , ap .
j. 466 ( 1996 ) , l109 .
thielemann et al .
, a&a 74 ( 1979 ) , 175 .
z . qian and s. e. woosley , ap .
j. 471 ( 1996 ) , 331 .
meyer et al . , phys .
c58 ( 1998 ) , 3696 .
mclaughlin et al . , phys .
c ( 1999 ) , in press .
thorsett , phys .
lett . 77 ( 1996 ) , 1432 .
j. taylor , rev .
66 ( 1994 ) , 711 .
rosswog et al . ,
a&a , submitted r. narayan et al . , ap . j. 395 ( 1992 ) , l83 .
e.p.j . van den heuvel and d.r .
lorimer , mnras 283 ( 1996 ) , l37 .
davies et al .
j. 431 ( 1994 ) , 742 .
janka and m. ruffert , a@xmath223a 307 ( 1996 ) , l33 .
m. ruffert and h .-
janka , a&a 338 ( 1998 ) , 535 .
baumgarte et al . , phys .
79 ( 1997 ) , 1182 . c. freiburghaus et al .
, in preparation j.j .
cowan et al . , ap .
j. 480 ( 1997 ) , 246 . | the focus of the present review is the production of the heaviest elements in nature via the r - process .
a correct understanding and modeling requires the knowledge of nuclear properties far from stability and a detailed prescription of the astrophysical environment .
experiments at cern / isolde have played a pioneering role in exploring the characteristics of nuclear structure in terms of masses and @xmath0-decay properties .
initial examinations paid attention to far unstable nuclei with magic neutron numbers related to r - process peaks , while present activities are centered on the evolution of shell effects with the distance from the valley of stability .
we first show in site - independent applications the effect of both types of nuclear properties on r - process abundances .
then , we explore the results of calculations related to two different ` realistic ' astrophysical sites , ( i ) the supernova neutrino wind and ( ii ) neutron star mergers .
we close with a list of remaining theoretical and experimental challenges needed to overcome for a full understanding of the nature of the r - process , and the role cern / isolde can play in this process .
nuclear astrophysics , r - process nucleosynthesis , isotope and isomer separation via hf splitting , laser ion source , nuclear structure , neutron - rich isotopes , ultra - metal - poor halo stars |
the motivation of this paper is to enrich the class of _ non - elementary amenable _ groups .
a group @xmath0 is amenable if there exists a finitely additive translation invariant probability measure on all subsets of @xmath0 .
this definition was given by john von neumann , @xcite , in a response to banach - tarski , and hausdorff paradoxes .
he singled out the property of a group which forbids paradoxical actions .
the class of _ elementary amenable groups _ , denoted by @xmath1 , was introduced by mahlon day in @xcite , as the smallest class of groups that contain finite and abelian groups and is closed under taking subgroups , quotients , extensions and directed unions .
the fact that the class of amenable groups is closed under these operations was already known to von neumann , @xcite , who noted at that at that time there was no known amenable group which did not belong to @xmath1 .
no substantial progress in understanding this class has been made until the 80s , when chou , @xcite , showed that all elementary amenable groups have either polynomial or exponential growth , and rostislav grigorchuk , @xcite gave an example of a group with intermediate growth .
grigorchuk s group served as a starting point in developing the theory of groups with intermediate growth , all of them being non - elementary amenable . in
the same paper chou showed that every simple finitely generated infinite group is not elementary amenable .
in @xcite it was shown that the topological full group of cantor minimal system is amenable . by the results of matui , @xcite , this group has a simple and finitely generated commutator subgroup , in particular , it is not elementary amenable .
this was the first example of infinite simple finitely generated amenable group .
currently there are only two sources of non - elementary amenable groups : groups acting on rooted trees and topological full groups of cantor minimal systems . in @xcite
, the author gives a unified approach to non - elementary amenability of groups acting on rooted trees .
here we give more examples of non - elementary amenable groups coming from topological full groups of cantor minimal systems .
[ main ] consider a minimal faithful action of @xmath2 on a cantor set conjugate to the action on a closed @xmath3-invariant subset of @xmath4 for some finite alphabet @xmath5 .
then the commutator subgroup of the topological full group @xmath6 $ ] is finitely generated .
we describe an explicit generating set , and give a simple direct proof .
subsequent paper @xcite gives a less direct proof of the finite generation of special subgroups @xmath7 of the topological full groups in the case of expansive action .
the groups @xmath7 are closely related to the derived subgroups of the full groups ( in particular , they coincide in the case of the actions of abelian groups ) , but precise relation is still not well understood .
it was proved in @xcite , that the commutator subgroup of @xmath6 $ ] is simple .
the topological full groups that correspond to interval exchange transformation group were studied in @xcite .
the authors prove that subgroups of _ rank _ equals to 2 are amenable .
these groups can be realized as topological full groups of minimal action of @xmath8 on the cantor set .
therefore , theorem [ main ] in the combination with @xcite , @xcite gives more examples of simple finitely generated infinite amenable groups , and thus by the result of chou non - elementary amenable groups .
matui , @xcite , showed that @xmath9 = [ [ t_1 , \ldots , t_n]]$ ] implies @xmath10 .
in particular , this implies that the groups from the corollary are different from the one previously obtained in @xcite . in the last section we associate a group @xmath11 to the penrose tiling .
the main result is the derived subgroup of @xmath11 is simple and finitely generated .
it is an open question to decide if the group @xmath11 is amenable .
[ lem : free ] let @xmath0 be an abelian group . if the action of @xmath0 on a cantor set @xmath13 is minimal and faithful , then it is free .
suppose that @xmath14 is a non - zero element . by faithfulness of the action
, there exists @xmath15 such that @xmath16 .
then there exists a neighborhood @xmath17 of @xmath18 such that @xmath19 .
let @xmath20 be an arbitrary point .
by minimality , there exists @xmath21 such that @xmath22 . since @xmath17 and @xmath23 are disjoint , the points @xmath24 and @xmath25 are different .
then @xmath26 .
it follows that @xmath27 has no fixed points in @xmath13 .
let us fix a minimal action of the free abelian group @xmath3 on a closed shift - invariant subset @xmath28 of the full shift over a finite alphabet @xmath5 .
then @xmath13 is homeomorphic to the cantor set .
we use the additive notation for the group @xmath3 .
if @xmath29 is a point of @xmath4 , then its image under the action of @xmath30 is defined by the rule @xmath31 in other words , elements of @xmath4 are labelings of the points of @xmath3 by elements of @xmath5 , and the elements @xmath30 act by shifting all the labels by @xmath27 .
alternatively , we may imagine the action of @xmath27 as the shift of the `` origin of coordinates '' in a given sequence @xmath32 by @xmath33 .
a _ patch _ @xmath34 is a finite subset @xmath35 together with a map @xmath36 .
the set @xmath37 is called the _ support _ of the patch .
we say that an element @xmath38 ( a _ @xmath3-sequence _ ) _ contains _ the patch @xmath39 if @xmath40 .
the set @xmath41 of all sequences containing a given patch @xmath42 is a clopen subset of @xmath4 called the _ cylindrical set _
defined by the patch , and the set of all such clopen subsets forms a basis of topology on @xmath4 , by definition .
we say that two patches @xmath43 and @xmath44 are _ compatible _ if there exists sequence @xmath45 containing @xmath46 and @xmath47 .
in other words , the patches are compatible if the intersection of the associated cylindrical sets is non - empty .
if the patches @xmath46 and @xmath47 are compatible then there _ union _
@xmath48 is the patch @xmath49 where @xmath50 and @xmath51 .
note that in terms of the cylindrical sets , we have @xmath52 . if @xmath34 is a patch , and @xmath53 is a finite set containing @xmath37 , then @xmath41 is equal to the disjoint union of the sets @xmath54 , where @xmath55 runs through the set of all maps @xmath56 such that @xmath57 .
note that some of these sets may be empty ( if the corresponding patch is not allowed for the elements of @xmath13 ) .
if @xmath34 is a patch and @xmath30 , then we have @xmath58 , where @xmath59 , where @xmath60 is given by @xmath61 , in accordance with the definition of the action of @xmath3 on @xmath4 .
[ lem : incompatible ] let @xmath62 be a finite set not containing zero
. then there exists @xmath63 such that for every @xmath64 and every @xmath65 the patches @xmath66 and @xmath67 are not compatible .
define the following metric on @xmath4 .
the distance @xmath68 is equal to @xmath69 , where @xmath70 is the biggest number such that restrictions of @xmath71 and @xmath72 to the ball of radius @xmath70 in @xmath3 with center in @xmath73 ( for example in the @xmath74 norm ) coincide .
the it is enough to prove that there exists @xmath75 such that @xmath76 for all @xmath65 and @xmath45 .
namely , for every @xmath75 there exists a finite set @xmath63 such that for every @xmath38 the set of all @xmath77 such that @xmath78 is contained in the @xmath75-neighborhood of @xmath32 .
suppose that it is not true , i.e. , that for every @xmath79 there exist @xmath32 and @xmath65 such that @xmath80 .
since @xmath81 is finite , this implies that there exists @xmath65 and a sequence of points @xmath82 such that @xmath83 as @xmath84 .
since @xmath13 is homeomorphic to the cantor set , this implies that @xmath27 has a fixed point , which is a contradiction lemma [ lem : free ] .
let @xmath85 be a clopen set , and @xmath86 elements such that @xmath87 are pairwise disjoint . denote by @xmath88 the element of @xmath12 $ ] given by @xmath89 in other words , @xmath88 cyclically permutes @xmath90 , @xmath91 , and @xmath92 in the natural way .
we will denote @xmath93 , for a patch @xmath42 .
let @xmath94 be subsets of a set @xmath95 such that only @xmath96 and @xmath97 have non - empty intersection , while all the other pairs of subsets are disjoint .
let @xmath98 be a permutation of @xmath95 acting trivially on @xmath99 , and satisfying @xmath100 , @xmath101 , @xmath102 , and @xmath103 .
similarly , let @xmath104 be a permutation acting trivially on @xmath105 and satisfying @xmath106 , @xmath107 , @xmath108 , and @xmath109
. then @xmath110 , [ b , a]]$ ] acts as @xmath98 on the set @xmath111 and identically outside of it .
note that we use the left action here , but the usual commutator @xmath112=g^{-1}h^{-1}gh$ ] .
let @xmath113 .
then the sets @xmath114 are pairwise disjoint .
the permutations @xmath98 and @xmath104 act as cycles of length three permuting @xmath115 and @xmath116 , respectively .
the element @xmath110 , [ b , a]]$ ] acts then on these fives sets in the same way as the similar expression involving commutators of the permutations @xmath117 and @xmath118 act on @xmath119 .
we have @xmath120=b^{-1}a^{-1}ba=(1 , 5 , 3)$ ] and @xmath121=bab^{-1}a^{-1}=(1 , 4 , 2)$ ] , and @xmath122=(1 , 4 , 2)^{-1}(1 , 5 , 3)^{-1}(1 , 4 , 2)(1 , 5 , 3)=(1 , 2 , 3)=a$ ] .
the set @xmath123 is @xmath98-invariant , and @xmath104 acts trivially on it .
it follows that the restriction of @xmath121 $ ] and @xmath120]$ ] to @xmath125 is equal to the restriction of @xmath126 $ ] and @xmath127 $ ] , which are trivial .
it follows that @xmath110 , [ b , a]]$ ] acts trivially on @xmath125 .
the same argument shows that @xmath110 , [ b , a]]$ ] acts trivially on @xmath128 . as a corollary , we get the following relation between the elements of the form @xmath129 .
[ cor : patchesunion ] let @xmath46 , @xmath47 be patches , @xmath130 be elements of @xmath3 such that @xmath131 are pairwise incompatible except for the pair @xmath46 and @xmath47
. then @xmath132 , [ t_{\pi_2 , ( 0 , h_1 , h_2 ) } , t_{\pi_1 , ( 0 , g_1 , g_2)}]]=t_{\pi_1\cup\pi_2 , ( 0 , g_1 , g_2)}.\ ] ] we will use the usual @xmath133 metric on @xmath3 , i.e. , the word metric associated with the standard generating set of @xmath3 .
denote by @xmath134 the ball of radius @xmath70 with the center in @xmath135 for this metric . by lemma [ lem : incompatible
] , there exists @xmath136 such that for every @xmath45 the patches @xmath137 and @xmath138 are incompatible for every @xmath30 of length @xmath139 .
let @xmath140 be the standard generating set of @xmath3 .
denote by @xmath141 the set of elements of @xmath12'$ ] of the form @xmath142 , where @xmath42 runs through the set of all patches of the form @xmath143 for @xmath45 . if @xmath144 , then the group generated by @xmath141 contains @xmath145 .
denote @xmath146 .
let @xmath63 be a finite subset containing @xmath147 .
define the patches @xmath148 , @xmath149 , for @xmath150 .
note that the patch @xmath151 contains the patch @xmath152 and that @xmath153 .
let us apply corollary [ cor : patchesunion ] for @xmath154 , @xmath155 , @xmath156 , @xmath157 , @xmath158 , @xmath159 , where @xmath160 are different elements of @xmath161 .
since @xmath162 and @xmath151 both contain the patch @xmath152 , the patches @xmath163 , @xmath164 , @xmath47 , @xmath165 , and @xmath166 are pairwise incompatible , except for @xmath46 and @xmath47 , which are both patches of @xmath32 .
it follows that we can apply corollary [ cor : patchesunion ] , hence @xmath167 , [ t_{\rho_h , ( 0 , h , 2h ) } , t_{\rho_0 , ( 0 , g , -g)}]]=t_{\rho_0\cup \rho_h , ( 0 , g , -g)}.\ ] ] note that @xmath168 .
let us apply now corollary [ cor : patchesunion ] to @xmath169 , @xmath170 , @xmath171 , @xmath157 , @xmath158 , @xmath172 .
the patches @xmath46 and @xmath47 are patches of @xmath173 , and contain @xmath174 .
it follows that , in the same way as above , we can apply corollary [ cor : patchesunion ] , and get @xmath175 , [ t_{\rho_g - g , ( 0 , h , -h ) } , t_{\rho_0-g , ( 0 , -2 g , -g)}]]=t_{(\rho_0-g)\cup(\rho_h - g ) , ( 0 , -2 g , -g)}.\ ] ] recall that @xmath176 and that we have @xmath177 .
we also have @xmath178 .
we have shown that the group generated by the set @xmath179 contains the set @xmath180 since @xmath181 , this finishes the proof of the proposition .
it follows that the group generated by @xmath182 contains @xmath141 for every @xmath144 . for every cylindrical set @xmath85
there exists @xmath70 such that @xmath17 is equal to the disjoint union of cylindrical sets @xmath41 such that @xmath42 is a patch with support @xmath134 .
it follows that every element of the form @xmath142 can be written as a product of elements of the same type such that @xmath42 is a patch with support @xmath134 for some @xmath70 big enough .
consequently , the group generated by @xmath182 contains all elements of the form @xmath142 .
the proof of theorem [ main ] is finished by the following , since the set @xmath182 is finite .
the derived subgroup of the full group of the action of @xmath3 on @xmath183 is generated by the set of elements of the form @xmath142 .
it is known , see @xcite , that the derived subgroup of @xmath12 $ ] is simple and is contained in every non - trivial normal subgroup of @xmath12 $ ] .
consider the set @xmath184 $ ] of all elements elements of order three permuting cyclically three disjoint clopen subsets @xmath185 of @xmath13 and acting identically outside their union .
the set @xmath186 is obviously invariant under conjugation by elements of @xmath12 $ ] , hence the group generated by @xmath186 is normal .
on the other hand , we have @xmath187'$ ] , as every element of @xmath186 is equal to the commutator of two transformations : one permuting @xmath188 with @xmath189 , and the other permuting @xmath189 with @xmath190 .
consequently , @xmath186 generates @xmath12'$ ] . for every element @xmath191 permuting cyclically clopen sets @xmath185
, there exists partitions of @xmath192 into cylindrical sets such that @xmath193 maps a piece of the partition to a piece of the partition , and restriction of @xmath193 to every piece of the partitions is equal to the restriction of an element of @xmath3 .
it follows that @xmath193 is a product of a finite set of elements of the form @xmath194 .
it remains to show that we can generate all elements of the form @xmath194 by elements of the form @xmath142 .
it is well known that the alternating group @xmath195 is generated by cycles @xmath196 .
it follows that the group generated by @xmath142 contains the set of elements of the form @xmath194 , where @xmath197 belong to one direct factor of @xmath3 .
let us prove the following technical lemma .
[ l1 ] let @xmath198 be the @xmath199-element set of @xmath200-letter words over the alphabet @xmath201 , and let @xmath202 be the symmetric group of permutations of @xmath203 . denote the alternating subgroup of even permutations of @xmath203 by @xmath204 .
consider the set @xmath205 of all elements of the type @xmath206 , where @xmath95 and @xmath207 are arbitrary ( possibly , empty ) words such that @xmath208 .
then @xmath204 is generated by the set @xmath205 .
the lemma can be proved by induction on @xmath200 . for @xmath209
, we use the well - known fact that @xmath210 is generated by 3-cycles @xmath211 . to apply this fact ,
we need to show that all 7 elements @xmath212 , @xmath213 , @xmath214 , @xmath215 , @xmath216 , @xmath217 , @xmath218 are generated by @xmath219 . this can be checked by hand : * @xmath220 ; * @xmath221 ; * @xmath222 ; * @xmath223 ; * @xmath224 ; * @xmath225 ; * @xmath226 .
suppose the statement holds for @xmath227 and consider the case @xmath228 .
since the alternating group is generated by 3-cycles , it s sufficient to show that every 3-cycle is generated by @xmath229 . assume we have a cycle @xmath230 , where @xmath231 are pairwise distinct , and @xmath232 , not necessarily distinct .
we know the following : * @xmath233 , @xmath234 , @xmath235 are generated by @xmath229 .
indeed , we can take the elements of @xmath236 generating @xmath237 and append the needed letter to each of them .
* @xmath238 , @xmath239 , @xmath240 are in @xmath229 by definition .
then , applying the induction base for the set @xmath241 , we conclude that @xmath230 is also generated by @xmath229 . in case
@xmath242 are not distinct , we can use a slightly modified version of the proof above . if , for example , @xmath243 ( which automatically implies @xmath244 ) , we can take an arbitrary word @xmath245 distinct from @xmath81 and @xmath246 in order to apply the induction base to @xmath247 .
clearly , the 3-cycle @xmath248 will belong to @xmath249 the induction step is complete .
lemma [ l1 ] implies that the group generated by all elements of the form @xmath194 , where @xmath197 belong to one factor of @xmath3 , contains all elements of the form @xmath194 , where @xmath250 are now arbitrary .
this finishes the proof of the proposition and theorem [ main ] .
let @xmath251 be irrational numbers such that the additive group @xmath252 generated by them modulo @xmath253 is isomorphic to @xmath2 ( this implies that @xmath254 is also isomorphic to @xmath2 ) .
the group @xmath255 is a subgroup of the circle @xmath256 , and hence acts on it in the natural way . by the classical kroneker s theorem ,
the action of each subgroup @xmath257 on @xmath256 is minimal , hence the action of @xmath255 on @xmath256 is also minimal .
let us lift @xmath255 as a set to @xmath258 $ ] by the natural quotient map @xmath258\to\mathbb{r}/\mathbb{z}$ ] , and let @xmath259 $ ] be the obtained set .
let us replace each number @xmath260 $ ] by two copies : @xmath261 and @xmath262 .
here we identify @xmath263 with @xmath264 and @xmath265 with @xmath73 , @xmath266 with 1 and @xmath267 with 0 , according to the natural cyclic order on @xmath256 ( seen also as the quotient of the interval @xmath258 $ ] ) .
denote by @xmath268 the obtained set ( equal to the disjoint union of @xmath258\setminus w$ ] and the set of doubled points @xmath269 ) .
the set @xmath268 is ordered in the natural way ( we assume that @xmath270 ) , and the order is linear ( total ) .
let us introduce the order topology on @xmath268 .
recall , that it is the topology generated by the open intervals @xmath271 .
the space @xmath272 is homeomorphic to the cantor set .
we use the following formulation of brouwer s theorem : a topological space is a cantor space if and only if it is non - empty , compact , totally disconnected , metrizable and has no isolated points . note that by classical metrization theorems , we can replace metrizability by hausdorffness and second countability .
the space @xmath268 is obviously non - empty , has no isolated points .
for any @xmath273 $ ] such that @xmath274 , we have @xmath275=(a_{-0 } , b_{+0})$ ] , hence the intervals @xmath276 are clopen .
the set of such intervals is a basis of topology , since the set @xmath269 is dense .
we also see that the space @xmath268 is second countable and hausdorff .
let @xmath277 be an arbitrary subset .
let us show that @xmath278 and @xmath279 exist , which will imply compactness .
let @xmath280 be the image of @xmath81 in @xmath258 $ ] .
we know that @xmath281 $ ] exist . if @xmath282 , then the corresponding element of @xmath268 is also a supremum of @xmath81 .
if @xmath283 , then @xmath284 , unless @xmath285 , in which case @xmath286 .
infima are treated in the same way .
the action of @xmath255 on @xmath256 naturally lifts to an action on @xmath268 : we just set @xmath287 and @xmath288 .
denote by @xmath289 the topological full group of the action @xmath290 .
for every element @xmath291 there exists a finite partition of @xmath268 into clopen subsets such that the action of @xmath27 on each of the subsets coincides with a translation by an element of @xmath255 .
clopen subsets of @xmath268 are finite unions of intervals of the form @xmath292 for @xmath293 .
it follows that @xmath27 is an _ interval exchange transformation _ : it splits the interval @xmath258 $ ] into a finite number of intervals and then rearranges them .
the endpoints of the intervals belong to @xmath269 .
conversely , every interval exchange transformation such that the endpoints of the subintervals belong to @xmath269 is lifted to an element of @xmath289 .
we have proved the following .
the group @xmath289 is naturally isomorphic to the group of all interval exchange transformations of @xmath258 $ ] such that the endpoints of the intervals into which @xmath258 $ ] is split belong to @xmath255 .
theorem [ main ] now implies the following .
the derived subgroup of @xmath289 is simple and finitely generated . a two - dimensional version of an interval exchange transformation group is considered in the next section .
there are several versions of the penrose tiling @xcite , let us describe one of them .
the tiles are two types of rhombi of equal side length 1 .
the angles of one rhombus are @xmath294 and @xmath295 .
the angles of the other are @xmath296 and @xmath297 .
we call these rhombuses `` thick '' and `` thin '' , respectively . mark a vertex of angle @xmath294 in the thick rhombus , and a vertex of angle @xmath297 of the thin rhombus .
mark the sides adjacent to the marked vertex by single arrows pointing towards the marked vertex .
mark the other edges by double arrows , so that in the thick rhombus they point away from the unmarked vertex of angle @xmath294 and in the thin rhombus they point towards the unmarked vertex of angle @xmath297 , see figure [ fig : penrosetiling ] .
penrose tiling _ is a tiling of the whole plane by such rhombi , where markings of the edges match ( adjacent tiles must have same number of arrows pointing in the same direction ) .
see figure [ fig : tiling ] for an example of a patch of a penrose tiling .
let us identify @xmath298 with @xmath299 , and consider all penrose tilings by rhombi such that their sides are parallel to the lines @xmath300 , @xmath301 .
a _ pointed _ penrose tiling is a penrose tiling with a marked vertex of a tile .
let @xmath186 be the set of all such pointed penrose tilings , up to translations ( two pointed tilings correspond to the same element of @xmath186 if and only if there exists a translation mapping one tiling to the other and the marked vertex of one tiling to the marked vertex of the other ) .
we sometimes identify a tiling with the set of vertices of its tiles .
let us introduce a topology on @xmath186 in the following way .
let @xmath302 be a finite set of vertices of a penrose tiling ( a _ patch _ ) , and let @xmath303 .
the corresponding open set @xmath304 is the set of all pointed tilings @xmath305 such that @xmath306 . in other words
, a pointed tiling @xmath305 belongs to @xmath304 if we can see the pointed patch @xmath307 around @xmath308 as a part of @xmath193 .
then the natural topology on @xmath186 is given by the basis of open sets of the form @xmath304 for all finite pointed patches @xmath309 of penrose tilings .
it follows from the properties of penrose tilings that the space @xmath186 is homeomorphic to the cantor set , and that for every penrose tiling @xmath193 the set of pointed tilings @xmath310 is dense in @xmath186 . in the literature , see ... , the space @xmath186 is called sometimes _
transversal_. consider a patch @xmath81 with two marked vertices @xmath311 .
then we have a natural homeomorphism @xmath312 mapping @xmath313 to @xmath314 .
the homeomorphism @xmath315 moves in every patch @xmath81 the marking from the vertex @xmath316 to the vertex @xmath317 .
it is easy to see that @xmath318 is a homeomorphism between clopen subsets of @xmath186 .
the set of all pointed tilings @xmath310 obtained from a given tiling @xmath193 is dense in @xmath186 and invariant under the action of the topological full group .
it follows that every element of the full group is uniquely determined by the permutation it induces on the set of vertices of the tiling . in terms of permutations of @xmath193
the full group can be defined in the following way .
we say that a map @xmath319 _ is defined by local rules _ if there exists @xmath70 such that for every @xmath320 the value of @xmath321 depends only on the set @xmath322 , where @xmath323 is the disc of radius @xmath70 around the origin @xmath324 .
a permutation @xmath319 is induced by the element of the full group if and only if @xmath325 is defined by a local rule .
consequently , the full group is isomorphic to the group of all permutations of @xmath193 defined by local rules .
denote @xmath326 , and let @xmath327\ ] ] be the group generated by the vectors on the sides of the regular pentagon @xmath328 .
note that @xmath329 . as an abelian group ,
@xmath37 is isomorphic to @xmath330 .
denote by @xmath331 the set of lines of the form @xmath332 , for @xmath333 and @xmath334 .
it is easy to see that for any two intersecting lines @xmath335 and any generator @xmath336 there exists @xmath337 such that @xmath338 is parallel to @xmath339 and @xmath340 .
it follows that for any pair of intersecting lines @xmath341 the intersection point @xmath342 belongs to @xmath37 .
consequently , a point @xmath343 belongs either to 0 , 1 , or to 5 lines from @xmath331 .
if @xmath343 does not belong to any line @xmath344 , then we call @xmath345 _ regular_. similarly to the case of interval exchange transformations , let us double each line @xmath344 .
let @xmath346 be the obtained space and let @xmath347 be the corresponding quotient map . if @xmath348 is regular , then @xmath349 consists of a single point . if @xmath350 belongs to a line @xmath344 , then @xmath349 consists of two points associated with each of the two half - planes into which @xmath351 separates @xmath299 .
every point @xmath352 has 10 preimages in @xmath346 associated with each of the ten sectors into which the lines from @xmath331 passing through @xmath345 separate the plane .
a sequence @xmath353 of points of @xmath346 converges to a point @xmath354 if and only if the sequence @xmath355 converges to @xmath356 and the sequence @xmath353 eventually belongs ( if @xmath356 is not regular ) to the associated closed half - plane or sector .
the space @xmath346 is locally compact and totally disconnected .
polygons with sides belonging to lines from @xmath331 form a basis of topology of @xmath346 .
the group @xmath37 acts on @xmath346 in the natural way , so that the action is projected by @xmath53 to the action of @xmath37 on @xmath299 by translations .
therefore , sums of the form @xmath357 , for @xmath358 and @xmath359 , are well defined .
let us describe , following @xcite , how a penrose tiling is associated with a point @xmath358 .
we will usually denote @xmath360 .
suppose that @xmath345 is regular .
the vertices of the corresponding tiling @xmath361 will be the points of the form @xmath362 , where @xmath363 are such that @xmath364 where @xmath365 is the pentagon with vertices @xmath366 , @xmath367 is the pentagon with vertices @xmath368 , @xmath369 , and @xmath370 .
( note that we have changed @xmath345 to @xmath371 comparing with @xcite . )
let @xmath375 and @xmath376 .
then @xmath377 if and only if @xmath378 .
it follows that action of @xmath37 on @xmath346 preserves the associated tilings up to translations .
in fact , it is not hard to show that two tilings @xmath379 and @xmath380 are translations of each other if and only if @xmath381 and @xmath382 belong to one @xmath37-orbit , see @xcite .... note that sides of the pentagons @xmath383 are contained in lines from the collection @xmath331 , hence they are naturally identified with compact open subsets of @xmath346 . see figure [ fig : vs ] for the pentagons @xmath384 .
denote by @xmath385 .
let @xmath390 , and let @xmath391 .
then the numbers @xmath392 and @xmath393 belong to @xmath37 , and the map @xmath394 is a bijection @xmath395 .
this maps moves @xmath18 to the marked vertex @xmath396 of the tiling corresponding to @xmath397 .
it follows that every pointed tiling , up to translation , corresponds to a point of @xmath398 .
it is easy to see that every pointed penrose tiling is represented by a unique point of @xmath398 , so that we get a bijection between @xmath398 and the space @xmath186 .
it follows from the results of @xcite that this bijection is a homeomorphism .
[ pr : fullvprime ] the group @xmath11 acts on @xmath399 locally by translations by elements of @xmath37 . in other words , for every @xmath400 there exists a partition of @xmath398 into disjoint clopen subsets @xmath401 such that @xmath325 acts on each of them by a translation @xmath402 for some @xmath403 and @xmath404 .
let us find some elements @xmath405 such that @xmath406 are pairwise disjoint , and denote by @xmath407 the union of the sets @xmath408 .
then it follows from proposition [ pr : fullvprime ] that @xmath11 is the group of all transformations @xmath409 that are locally equal to translations by elements of @xmath37 .
let us say that two clopen sets @xmath410 are _ equidecomposable _ if there exists a homeomorphism @xmath411 locally equal to translations by elements of @xmath37 . if @xmath17 is any clopen subset which is equidecomposable with @xmath412 , then @xmath11 is equal to the group of all transformations of @xmath17 that are locally translations by elements of @xmath37 .
the obtained triangles can be grouped into pairs of triangles @xmath416 such that @xmath417 is obtained from @xmath193 by a rotation by @xmath418 ( and translation ) .
such pairs can be put together to form parallelograms , as it is shown on figure [ fig : parallelograms ] .
figure [ fig : parallelograms ] also shows that each such parallelogram is equidecomposable with its rotation by @xmath419 .
it follows that each parallelogram is equidecomposable with its rotation by any angle of the form @xmath420 .
consequently , every parallelogram formed by the acute - angled triangles is equidecomposable with the parallelogram with the set of vertices @xmath421 , and each parallelogram formed by the obtuse - angled triangles is equidecomposable with the parallelogram with the set of vertices @xmath422 .
we get 5 parallelograms of each kind .
we can put all the obtained parallelograms together to form the parallelogram @xmath413 .
the parallelogram @xmath413 , seen as a subset of @xmath346 , is the fundamental domain of the group @xmath423 .
it is easy to check that @xmath424 is isomorphic to @xmath425 .
the space of orbits @xmath426 is naturally homeomorphic to the parallelogram @xmath413 ( and hence to the spaces @xmath398 and @xmath186 ) . | we give new examples of simple finitely generated groups arising from actions of free abelian groups on the cantor sets .
as particular examples , we discuss groups of interval exchange transformations , and a group naturally associated with the penrose tilings .
many groups in this class are amenable . |
in @xcite , @xcite and @xcite we carried out a program to generalize l.c .
siebenmann s famous manifold collaring theorem @xcite in ways applicable to manifolds with non - stable fundamental group at infinity .
motivated by some important examples of finite - dimensional manifolds and a seminal paper by t.a .
chapman and siebenmann @xcite on hilbert cube manifolds , we chose the following definitions .
a manifold @xmath2 with compact boundary is called a _
homotopy collar _
if @xmath3 is a homotopy equivalence .
if @xmath2 contains arbitrarily small homotopy collar neighborhoods of infinity , we call @xmath2 a _ pseudo - collar . _ clearly , an actual open collar @xmath2 , i.e. , @xmath4 , is a special case of a pseudo - collar .
fundamental to @xcite , @xcite , and our earlier work , is the notion of ` inward tameness ' .
a manifold @xmath5 is _ inward tame _ if each of its clean neighborhoods of infinity is finitely dominated ; it is _ absolutely inward tame _ if those neighborhoods all have finite homotopy type .
an alternative formulation of this definition ( see [ subsection : topology of ends of manifolds ] ) justifies the adjective ` inward'a term that helps distinguish this version of tameness from a similar , but inequivalent , version found elsewhere in the literature . in @xcite a classification of pseudo - collarable @xmath0-manifolds for @xmath6
was obtained . in simplified form
, it says : [ pseudo - collarability characterization
simple version][pct ] a 1-ended @xmath0-manifold @xmath5 ( @xmath1 ) with compact boundary is pseudo - collarable iff the following conditions hold : 1 .
@xmath5 is absolutely inward tame , and 2 .
the fundamental group at infinity is @xmath7-semistable . a `
@xmath7-semistable ( or perfectly semistable ) fundamental group at infinity ' indicates that an inverse sequence of fundamental groups of neighborhoods of infinity can be arranged so that bonding homomorphisms are surjective with perfect kernels . by way of comparison , the simple version of siebenmann s collaring theorem is obtained by replacing b ) with the stronger condition of @xmath8-stability , while chapman and siebenmann s pseudo - collarability characterization for hilbert cube manifolds is obtained by omitting b ) entirely .
thus , the differences among these three results lie entirely in the fundamental group at infinity . in this paper
we take a close look at @xmath0-manifolds satisfying only the inward tameness hypothesis . by necessity
, our attention turns to the group theory at the ends of those spaces .
unlike the case of infinite - dimensional manifolds , cw complexes , or even @xmath0-manifolds with noncompact boundary , inward tameness has major implications for the fundamental group at the ends of @xmath0-manifolds with compact boundary .
unfortunately , inward tameness ( ordinary or absolute ) does not imply @xmath7-semistability an example from @xcite attests to that but it comes remarkably close .
one of the main results of this paper is the following .
[ theorem : inward tame implies ap - semistability]let @xmath5 be an inward tame @xmath0-manifold with compact boundary .
then @xmath5 has an @xmath9-semistable ( almost perfectly semistable ) fundamental group at each of its finitely many ends . developing the appropriate group theory ( including the definition of @xmath9-semistable ) and proving the above theorem are the initial goals of this paper .
after that is accomplished , we apply those investigations by proving a structure theorem for manifolds that are inward tame , but not necessarily pseudo - collarable . [ near pseudo - collarability characterization
simple version][theorem : npc characterization
simple version]a 1-ended @xmath0-manifold @xmath5 ( @xmath1 ) with compact boundary is nearly pseudo - collarable iff the following conditions hold : 1 .
@xmath5 is absolutely inward tame , and 2 .
the fundamental group at infinity is @xmath10-semistable .
the notion of ` near pseudo - collarability ' will be defined and explored in [ section : generalizing 1-sided cobordisms , homotopy collars , and pseudocollars ] . for
now , we note that nearly pseudo - collarable manifolds admit arbitrarily small clean neighborhoods of infinity @xmath11 , containing codimension 0 submanifolds @xmath12 for which @xmath13 is a homotopy equivalence . obtaining a near pseudo - collar structure
requires a slight strengthening of @xmath9-semistability to @xmath10-semistability ( strong almost perfect semistability ) .
the essential nature of this stronger condition is verified by a final result , in which our group - theoretic explorations come together in a concrete set of examples . [
theorem : existence of counterexamples]for all @xmath1 , there exist 1-ended open @xmath0-manifolds that are absolutely inward tame but do not have @xmath10-semistable fundamental group at infinity , and thus , are not nearly pseudo - collarable . in [ section : remaining questions ] , we close with a discussion of some open questions .
_ throughout this paper attention is restricted to noncompact manifolds with compact boundaries . when a boundary is noncompact , its end topology gets entangled with that of the ambient manifold , leading to very different issues . in the study of noncompact manifolds ,
a focus on those with compact boundaries is analogous to a focus on closed manifolds in the study of compact manifolds .
an investigation of manifolds with noncompact boundaries is planned for @xcite . _
in this subsection we review the definition of ` perfect group ' and discuss some variations . given elements @xmath14 and @xmath15 of a group @xmath16 , the _ commutator _ @xmath17 will be denoted @xmath18 $ ] .
the _ commutator subgroup _ of @xmath16 , denoted @xmath19 $ ] is the subgroup generated by all commutators .
it is a standard fact that @xmath19 $ ] is normal in @xmath16 and is the smallest such subgroup with an abelian quotient .
we call @xmath16 _ perfect _ if @xmath20 $ ]
. now suppose @xmath16 and @xmath21 are normal subgroups of @xmath22 .
define @xmath23 $ ] to be the subgroup of @xmath22 generated by the set of commutators @xmath24 = \left\ { k^{-1}j^{-1}kj\mid k\in k\text { and } j\in j\right\ } .\ ] ] the following is standard and easy to verify .
[ lemma : generalized commutator subgroups]for normal subgroups @xmath16 and @xmath21 of a group @xmath22 , 1 .
@xmath25 \trianglelefteq g$ ] , 2 .
@xmath25 \trianglelefteq k$ ] and @xmath25 \trianglelefteq j$ ] , and 3 .
@xmath25 = \left [ j , k\right ] $ ] .
given the above setup , we say that @xmath16 is @xmath21-_perfect _ if @xmath26 $ ] , and that @xmath16 is _ strongly _
@xmath21-_perfect _ if @xmath27 $ ] _ _ .
_ _ by lemma [ lemma : generalized commutator subgroups ] , both of these conditions imply that @xmath28 ; so we customarily begin with that as an assumption .
the following two lemmas are immediate .
we state them explicitly for the purpose of comparison .
let @xmath28 be normal subgroups of @xmath22 . 1 .
@xmath16 is perfect if and only if each element of @xmath16 can be expressed as @xmath29 $ ] where @xmath30 for all @xmath31 .
@xmath16 is @xmath21-perfect if and only if each element of @xmath16 can be expressed as @xmath29 $ ] where @xmath32 for all @xmath31 .
@xmath16 is strongly @xmath21-perfect if and only if each element of @xmath16 can be expressed as @xmath29 $ ] where @xmath33 and @xmath34 for all @xmath31 .
[ basic implications]let @xmath35 be normal subgroups of @xmath22 . 1 .
if @xmath16 is [ strongly ] @xmath21-perfect , then @xmath16 is [ strongly ] @xmath36-perfect for every normal subgroup @xmath36 containing @xmath21 .
@xmath16 is [ strongly ] @xmath16-perfect if and only if @xmath16 is a perfect group .
_ lemma [ basic implications ] suggests a key theme : the smaller the group _
_ for which _
@xmath16 _ is [ strongly ] _
@xmath36-_perfect , the closer _
@xmath16 _ is to being a genuine perfect group .
_ the various levels of perfectness can be nicely characterized using _
group homology_. the @xmath37-homology of a group @xmath22 may be defined as the @xmath37-homology of a @xmath38 space @xmath39 .
if @xmath40 is a homomorphism , there is a map @xmath41 , unique up to basepoint preserving homotopy , inducing @xmath42 on fundamental groups .
define @xmath43 to be the homomorphisms induced by @xmath44 .
[ lemma : group homology]let @xmath28 , @xmath45 be inclusion , and @xmath46 be projection . 1 .
@xmath16 is perfect if and only if @xmath47 .
@xmath16 is @xmath21-perfect if and only if @xmath48 if and only if @xmath49 3 .
@xmath16 is strongly @xmath21-perfect if and only if @xmath16 is @xmath21-perfect and @xmath50 is surjective .
claim 1 ) is clear from the standard fact that @xmath51 $ ] .
claim 2 ) can be verified with elementary group theory .
claim 3 ) follows from a well - known 5-term exact sequence due to stallings @xcite and stammbach @xcite . due to its importance in this paper , we state it as a separate lemma .
[ 5-term exact sequence for group homology][lemma : 5 term exact ] given a normal subgroup @xmath16 of a group @xmath21 , there is a natural exact sequence:@xmath52 \rightarrow h_{1}\left ( j;\mathbb{z }
\right ) \rightarrow h_{1}\left ( j / k;\mathbb{z } \right ) \rightarrow0\text{.}\ ] ] the following elementary facts about group homology will be useful .
[ lemma : induced surjections on h_2]let @xmath53 be a map between connected cw complexes and @xmath54 the induced homomorphism
. then 1 .
@xmath55 , 2 .
@xmath56 realizes @xmath57 , and 3 . if @xmath58 is surjective , then @xmath59 is also surjective .
build a @xmath60 complex @xmath61 by attaching cells of dimension @xmath62 to @xmath63 and a @xmath64 complex @xmath65 by attaching cells of dimension @xmath62 to @xmath66 . both @xmath67 and @xmath68
induce isomorphisms on @xmath8 and @xmath69 , so ( 1 ) follows .
use the asphericity of @xmath65 to extend @xmath70 to @xmath71 , also inducing @xmath42 on @xmath8 .
clearly @xmath72 and @xmath73 are surjective .
this gives a commutative diagram@xmath74{ccc}h_{2}\left ( x;\mathbb{z } \right ) & \overset{f_{\ast}}{\twoheadrightarrow } & h_{2}\left ( y;\mathbb{z }
\right ) \\ i_{\ast}\downarrow & & \downarrow j_{\ast}\\ h_{2}\left ( \pi_{1}\left ( x\right ) ;
\mathbb{z } \right ) & \overset{f_{\ast}^{\prime}}{\longrightarrow } & h_{2}\left ( \pi _ { 1}\left ( y\right ) ; \mathbb{z } \right ) \end{array}\ ] ] since the other maps are all surjective , so is @xmath75 .
lastly we offer a topological characterization of the various levels of perfectness . for the purposes of this paper , these are possibly the most useful .
let @xmath76 denote a compact orientable surface of genus @xmath77 with a single boundary component .
a collection of oriented simple closed curves @xmath78 on @xmath76 with the property that each @xmath79 intersects @xmath80 transversely at a single point , and each of @xmath81 , @xmath82 , and @xmath83 is empty when @xmath84 , is called a _
complete set of handle curves _ for @xmath76 .
a complete set of handle curves on @xmath76 is not unique ; however given any such set , there exists a homeomorphism of @xmath76 to the ` disk with @xmath77 handles ' pictured in figure [ fig1-handle curves ] taking each @xmath79 and @xmath80 to the corresponding curves in the diagram .
[ ptb ] figure1-ends4.eps given a ( not necessarily embedded ) loop @xmath85 in a topological space @xmath63 , we say that @xmath85 _ bounds a compact orientable surface in _
@xmath63 if , for some @xmath77 , there exists a map @xmath86 such that @xmath87 . notice that we do not require that @xmath70 be an embedding .
we often abuse terminology slightly by saying that @xmath85 bounds the surface @xmath76 in @xmath63 .
similarly , we often do not distinguish between a set of handle curves on @xmath76 and their images in @xmath63 .
[ lemma : geometric perfect ] let @xmath63 be a space with @xmath88 and let @xmath28 be normal subgroups of @xmath22
. then 1 .
@xmath16 is perfect if and only if each loop @xmath85 in @xmath63 representing an element of @xmath16 bounds a surface @xmath76 in @xmath63 containing a complete set of handle curves @xmath89 with each @xmath79 and @xmath80 belonging to @xmath16 .
@xmath16 is @xmath21-perfect if and only if each loop @xmath85 in @xmath63 representing an element of @xmath16 bounds a surface @xmath76 in @xmath63 containing a complete set of handle curves @xmath89 with each @xmath79 and @xmath80 belonging to @xmath21 .
@xmath16 is strongly @xmath21-perfect if and only if each loop @xmath85 in @xmath63 representing an element of @xmath16 bounds a surface @xmath76 in @xmath63 containing a complete set of handle curves @xmath90 with each @xmath79 belonging to @xmath16 and each @xmath80 belonging to @xmath21 .
_ we are being informal in the statement of lemma [ lemma : geometric perfect ] . since the handle curves are not based , we should also choose , for each pair _ @xmath91 _ _ , an arc _ _ @xmath92 _ in _ @xmath76 _ from _ @xmath93 _ to _ @xmath94__. the element of _ _ @xmath95 _ represented by _
@xmath79 _ is then _ @xmath96 _ _ , and similarly for _ _
@xmath80__. notice that , by normality , the question of whether one of these loops belongs to _ _
@xmath16 _ or _
@xmath21 _ is independent of the choice of _ @xmath97__. _ _ understanding the ` fundamental group at infinity ' requires the language of inverse sequences .
we briefly review the necessary definitions and terminology . throughout this subsection
all arrows denote homomorphisms , while those of type @xmath98 or @xmath99 specify surjections .
the symbol @xmath100 denotes isomorphisms .
let @xmath101 be an inverse sequence of groups and homomorphisms .
subsequence _ is an inverse sequence of the form @xmath102 in the future we denote a composition @xmath103 ( @xmath104 ) by @xmath105 .
sequences @xmath106 and @xmath107 are _ pro - isomorphic _ if , after passing to subsequences , there exists a commuting diagram : @xmath74{ccccccc}g_{i_{0 } } & \overset{\lambda_{i_{0}+1,i_{1}}}{\longleftarrow } & g_{i_{1 } } & \overset{\lambda_{i_{1}+1,i_{2}}}{\longleftarrow } & g_{i_{2 } } & \overset { \lambda_{i_{2}+1,i_{3}}}{\longleftarrow } & \cdots\\ & \nwarrow\quad\swarrow & & \nwarrow\quad\swarrow & & \nwarrow\quad\swarrow & \\ & h_{j_{0 } } & \overset{\mu_{j_{0}+1,j_{1}}}{\longleftarrow } & h_{j_{1 } } & \overset{\mu_{j_{1}+1,j_{2}}}{\longleftarrow } & h_{j_{2 } } & \cdots \end{array } .\ ] ] clearly an inverse sequence is pro - isomorphic to each of its subsequences .
to avoid tedious notation , we often do not distinguish @xmath106 from its subsequences .
instead we assume @xmath108 has the properties of a preferred subsequence
prefaced by the words ` after passing to a subsequence and relabeling ' .
an inverse sequence @xmath106 is _ stable _ if it is pro - isomorphic to an @xmath109 for which each @xmath110 is an isomorphism .
a more usable formulation is that @xmath108 is stable if , after passing to a subsequence and relabeling , there is a commutative diagram of the form @xmath74{ccccccccc}g_{0 } & \overset{\lambda_{1}}{\longleftarrow } & g_{1 } & \overset{\lambda_{2}}{\longleftarrow } & g_{2 } & \overset{\lambda_{3}}{\longleftarrow } & g_{3 } & \overset{\lambda_{4}}{\longleftarrow } & \cdots\\ & \nwarrow\quad\swarrow & & \nwarrow\quad\swarrow & & \nwarrow\quad\swarrow & & & \\ & im(\lambda_{1 } ) & \overset{\cong}{\longleftarrow } & im(\lambda_{2 } ) & \overset{\cong}{\longleftarrow } & im(\lambda_{3 } ) & \overset{\cong}{\longleftarrow } & \cdots & \end{array } \tag{$\ast$}\ ] ] where all unlabeled maps are obtained by restriction . if @xmath109 can be chosen so that each @xmath110 is an epimorphism , we call our sequence
_ semistable _ ( or _ mittag - leffler , _ or
_ _ pro - epimorphic _ _ ) . in that case
, it can be arranged that the maps in the bottom row of ( @xmath111 ) are epimorphisms .
similarly , if @xmath109 can be chosen so that each @xmath110 is a monomorphism , we call our sequence _ pro - monomorphic _ ; it can then be arranged that the restriction maps in the bottom row of ( @xmath111 ) are monomorphisms .
it is easy to show that an inverse sequence that is semistable and pro - monomorphic is stable .
an inverse sequence is _ perfectly semistable _ if it is pro - isomorphic to an inverse sequence @xmath112 of finitely presentable groups and surjections where each @xmath113 is perfect .
a straightforward argument ( * ? ? ?
1 ) shows that sequences of this type behave well under passage to subsequences .
an _ augmentation _ of an inverse sequence @xmath114 is a sequence @xmath115 , where @xmath116 and @xmath117 for each @xmath31 .
the corresponding _ augmentation sequence _ is the sequence @xmath118 .
the _ minimal augmentation _ ( or the _ unaugmented case _ ) occurs when @xmath119 ; the _ maximal augmentation _ is the case where @xmath120 ; and the _ standard augmentation
_ occurs when @xmath121 for each @xmath31 .
any augmentation where @xmath122 for each @xmath31 is called a _ small augmentation .
_ for each subsequence @xmath123 of a sequence @xmath108 augmented by @xmath115 , there is a corresponding augmentation @xmath124 .
we say that @xmath106 satisfies the @xmath125__-perfectness property _ _ if , for each @xmath31 , @xmath126 is @xmath127-perfect ; it satisfies the _ strong _ @xmath115__-perfectness property _ _
if each @xmath128 is strongly @xmath129-perfect .
more concisely , if @xmath130 and @xmath131 , these conditions require that each @xmath132 be [ strongly ] @xmath133-perfect . employing the above terminology , we can restate the definition perfect semistability ( abbreviated @xmath7__-semistable _ _ ) by requiring that the sequence be pro - isomorphic an inverse sequence of finitely presented groups and surjections satisfying the @xmath115-perfectness property for the minimal augmentation @xmath134 .
more generally , we call an inverse sequence of groups : * @xmath9__-semistable _ _ ( for almost perfectly semistable ) if it is pro - isomorphic to an inverse sequence @xmath135 of finitely presentable groups and surjections , satisfying the @xmath125-perfectness property for some small augmentation @xmath125 , and * @xmath10__-semistable _ _ ( for strongly almost perfectly semistable ) if it is pro - isomorphic to an inverse sequence @xmath108 of finitely presentable groups and surjections satisfying the strong @xmath115-perfectness property for some small augmentation @xmath115 . _ note that an inverse sequence satisfies the [ strong ] _
@xmath125__-perfectness property for _ _ some _ small augmentation
_ @xmath115 _
if and only if it satisfies that property for the standard augmentation .
_ when applying sequences of the above types to geometric constructions , it is frequently desirable to pass to subsequences without losing the defining property of the sequence .
for that reason , the following observation is crucial .
[ prop : nearly sub]if an inverse sequence @xmath114 of surjections augmented by @xmath115 satisfies the [ strong ] @xmath115-perfectness property , then any subsequence @xmath123 satisfies the corresponding [ strong ] @xmath124-perfectness property .
since the proofs for perfectness and strong perfectness are similar , we prove only the latter .
assume @xmath106 augmented by @xmath115 satisfies strong @xmath136-perfectness .
simplifying notation , a portion of the given subsequence becomes@xmath137 where @xmath138 .
we must show that @xmath139 $ ] .
suppose the proposition holds for @xmath140 .
if @xmath141 , then @xmath142 , and the result follows by hypothesis .
now , assume @xmath143 and write @xmath144 let @xmath145 ; then @xmath146 .
by induction , @xmath147@xmath148 $ ] ; so , @xmath149 is a product of commutators @xmath150 $ ] where @xmath151 and @xmath152 .
since @xmath153 is surjective over @xmath154 we identify for each @xmath155 a pair of elements @xmath156 that map to @xmath157 and @xmath158 , respectively .
thus , @xmath159 , @xmath160 , and @xmath161\in\lbrack\ker\left ( \lambda_{b+1,c}\right ) , \lambda_{b+1,c}^{-1}\left ( l_{b}\right ) ] $ ] . now , let @xmath162 be the product of the commutators with @xmath163 $ ] replacing @xmath164 $ ] . by construction , @xmath165 and @xmath166 $ ] .
thus,@xmath167 \subseteq\left [ \ker\left ( \lambda_{b+1,c}\right ) , \lambda_{b+1,c}^{-1}\left ( l_{b}\right ) \right ] \text{.}\ ] ] consequently , @xmath168 $ ] as well , completing the proof of the proposition .
next we supply some topological definitions and background . throughout the paper , @xmath169 represents homeomorphism and @xmath170 indicates homotopic maps or homotopy equivalent spaces . the word _ manifold _ means _
manifold with ( possibly empty ) boundary_. a manifold is _ open _ if it is non - compact and has no boundary . as noted earlier , we restrict our attention to manifolds with compact boundaries . for convenience ,
all manifolds are assumed to be pl ; analogous results may be obtained for smooth or topological manifolds in the usual ways .
our standard resource for pl topology is @xcite .
some of the results presented here are valid in all dimensions .
others are valid in dimensions @xmath171 or @xmath172 , but require the purely topological @xmath173-dimensional techniques found in @xcite for the @xmath173 and/or @xmath174 dimensional cases ; there the conclusions are only topological .
the main focus of this paper is on dimensions @xmath175 .
let @xmath5 be a manifold with compact ( possibly empty ) boundary .
a set @xmath176 is a _ neighborhood of infinity _ if @xmath177 is compact .
a neighborhood of infinity @xmath11 is _ clean _
if * @xmath11 is a closed subset of @xmath5 , * @xmath178 , and * @xmath11 is a codimension 0 submanifold of @xmath5 with bicollared boundary .
it is easy to see that each neighborhood of infinity contains a clean neighborhood of infinity .
we say that @xmath5 _ has _
_ ends _ if it contains a compactum @xmath180 such that , for every compactum @xmath181 with @xmath182 , @xmath183 has exactly @xmath179 unbounded components , i.e. , @xmath179 components with noncompact closures .
when @xmath179 exists , it is uniquely determined ; if @xmath179 does not exist , we say @xmath5 _ has infinitely many ends_. if @xmath5 is @xmath179-ended , then it contains a clean neighborhood of infinity @xmath11 consisting of @xmath179 connected components , each of which is a 1-ended manifold with compact boundary .
thus , when studying manifolds with finitely many ends , it suffices to understand the @xmath184__-ended _ _ situation .
that is the case in this paper , where our standard hypotheses ensure finitely many ends .
( see theorem [ semistable ] . )
a connected clean neighborhood of infinity with connected boundary is called a _
0-neighborhood of infinity_. a @xmath185-neighborhood of infinity @xmath11 for which @xmath186 induces a @xmath8-isomorphism is called a _ generalized 1-neighborhood of infinity_. if , in addition , @xmath187 for @xmath188 , then @xmath11 is a _
generalized k - neighborhood of infinity_. a nested sequence @xmath189 of neighborhoods of infinity is _ cofinal _ if @xmath190 .
we will refer to any cofinal sequence @xmath191 of closed neighborhoods of infinity with @xmath192 , for all @xmath31 , as an _ end structure _ for @xmath5 .
descriptors will be added to indicate end structures with additional properties .
for example , if each @xmath193 is clean we call @xmath191 a _ clean end structure _ ; if each @xmath193 is clean and connected we call @xmath194 a _ clean connected end structure _ ; and if each @xmath193 is a generalized @xmath179-neighborhood of infinity , we call @xmath194 a _ generalized k - neighborhood end structure_. _ the word ` generalized ' in the above definitions is in deference to siebenmann s terminology in @xcite where the ambient manifold _ @xmath5 _ _ is assumed to have stable fundamental group at infinity . there a ( non - generalized ) k - neighborhood of infinity _ _ @xmath11 _ is also required to satisfy _ @xmath195__. _ _ building upon the above terminology , the primary goal of this paper can be described as : identify , construct , and detect the existence of various end structures for manifolds
. a central example the _ pseudo - collar _ can be described as an end structure @xmath194 where each @xmath193 is a homotopy collar .
we say @xmath196is _ inward tame _ if , for arbitrarily small neighborhoods of infinity @xmath11 , there exist homotopies @xmath197 \rightarrow n$ ] such that @xmath198 and @xmath199 is compact .
thus inward tameness means each neighborhood of infinity can be pulled into a compact subset of itself . in this case
we refer to @xmath200 as a _ taming homotopy_. in @xcite , the existence of generalized @xmath201-neighborhood end structures is shown for all inward tame @xmath5 ( @xmath202 ) .
recall that a space @xmath63 is _ finitely dominated _ if there exists a finite complex @xmath16 and maps @xmath203 and @xmath204 such that @xmath205 .
the following lemma uses this notion to offer equivalent formulations of inward tameness .
[ lemma : inward tame ] ( see ( * ? ? ?
* lemma 2.4 ) ) for a manifold @xmath5 , the following are equivalent . 1 .
@xmath5 is inward tame .
each clean neighborhood of infinity in @xmath5 is finitely dominated .
3 . for each clean end structure @xmath194
, the inverse sequence @xmath206 is pro - homotopy equivalent to an inverse sequence of finite polyhedra . given a clean connected end structure @xmath207 , base points @xmath208 , and paths @xmath209 connecting @xmath210 to @xmath211 , we obtain an inverse sequence : @xmath212 here , each @xmath213 is the homomorphism induced by inclusion followed by the change of base point isomorphism determined by @xmath79 . the singular ray obtained by piecing together the @xmath79 s
is called the _ base ray _ for the inverse sequence . provided
the sequence is semistable , its pro - isomorphism class does not depend on any of the choices made above ( see @xcite or @xcite ) . in the absence of semistability , the pro - isomorphism class of the inverse sequence depends on the base ray ; hence , the ray becomes part of the data .
the same procedure may be used to define @xmath214 for all @xmath215 . similarly , but without need for a base ray or connectedness , we may define @xmath216 .
in @xcite , wall showed that each finitely dominated connected space @xmath63 determines a well - defined @xmath217 \right ) $ ] ( the reduced projective class group ) that vanishes if and only if @xmath63 has the homotopy type of a finite complex .
given a clean connected end structure @xmath218 for an inward tame @xmath5 , we have a wall finiteness obstruction @xmath219 for each @xmath31 .
these may be combined into a single obstruction @xmath220 \right)\ ] ] that is well - defined and which vanishes if and only if each clean neighborhood of infinity in @xmath5 has finite homotopy type .
see @xcite or @xcite for details .
we may now state the full version of the main theorem of @xcite .
[ pseudo - collarability characterization
complete version][pct technical version]a 1-ended @xmath0-manifold @xmath5 ( @xmath1 ) with compact boundary is pseudo - collarable if and only if the following conditions hold : 1 .
@xmath5 is inward tame , 2 .
@xmath221 is @xmath7-semistable , and 3 .
in this section we show that , for manifolds with compact boundary , the inward tameness condition , by itself , has significant implications .
the main goal is a proof of theorem [ theorem : inward tame implies ap - semistability]that every inward tame manifold with compact boundary has @xmath9-semistable fundamental group at each of its finitely many ends .
results in this section are valid in all ( finite ) dimensions
. begin by recalling a theorem from @xcite . [ semistable]if an @xmath0-manifold with compact ( possibly empty ) boundary
is inward tame , then it has finitely many ends , each of which has semistable fundamental group and stable homology in all dimensions .
_ note that _ none _ of the above conclusions is valid for hilbert cube manifolds , polyhedra , or manifolds with noncompact boundary .
see , for example , @xcite .
_ as preparation for the proof of theorem [ theorem : inward tame implies ap - semistability ] , we look at an easier result that follows directly from theorem [ semistable ] .
let @xmath5 be an inward tame @xmath0-manifold with compact boundary . since @xmath5 is finite - ended
, there is no loss of generality in assuming that @xmath5 is 1-ended . by taking a product with @xmath223 ( @xmath224 ) if necessary , we may arrange that @xmath1 , without changing the fundamental group at infinity .
so , by the semistability conclusion of theorem [ semistable ] combined with the generalized @xmath225-neighborhood theorem ( * ? ? ?
* th.4 ) , we may choose a generalized @xmath225-neighborhood end structure @xmath194 for which each bonding map in the inverse sequence @xmath226 is surjective .
abelianization gives an inverse sequence@xmath227 which , by theorem [ semistable ] , is stable .
it follows that all but finitely many of the epimorphisms in ( [ sequence : h_1 sequence ] ) are isomorphisms , so by omitting finitely many terms ( then relabeling ) , we may assume all bonds in ( [ sequence : h_1 sequence ] ) are isomorphisms .
a term - by - term application of lemma [ lemma : group homology ] gives the following .
[ prop : near perfect 1]every 1-ended inward tame manifold @xmath5 with compact boundary admits a generalized @xmath225-neighborhood end structure @xmath191 for which all bonding maps in the sequence @xmath228 are surjective and each @xmath128 is @xmath229-perfect ; in other words , if @xmath230 is the maximal augmentation , then @xmath231 satisfies the @xmath115-perfectness property .
theorem [ theorem : inward tame implies ap - semistability ] is a stronger version of proposition [ prop : near perfect 1 ] . for clarity , we restate it in a similar form .
[ prop : near perfect 2]every 1-ended inward tame manifold @xmath5 with compact boundary admits a generalized @xmath225-neighborhood end structure @xmath191 for which all bonding maps in the sequence @xmath228 are surjective and , if we let @xmath130 for each @xmath232 ( the standard augmentation ) , then @xmath132 is @xmath233-perfect for all @xmath234 . in other words , @xmath231 satisfies the @xmath235-perfectness property ; so @xmath5 has @xmath9-semistable fundamental group at infinity .
assume the sequence @xmath194 was chosen so that , for each @xmath31 , @xmath236 is sufficiently small that a taming homotopy @xmath237 pulls @xmath193 into @xmath238 , i.e. , @xmath239 , and @xmath240 is sufficiently small that @xmath241 \right ) \cap n_{i+3}=\varnothing$ ] . by compactness of @xmath242 and @xmath243 \right ) $ ] those choices can be made .
now let @xmath234 be fixed and @xmath244 be the universal covering projection .
let @xmath245 and for @xmath246 , @xmath247 and @xmath248 then @xmath249 ; and @xmath250 lifts to a proper homotopy @xmath251 that pulls @xmath252 into @xmath253 and for which @xmath254 \right ) $ ] misses @xmath255 .
we may associate @xmath233 with @xmath256 and @xmath132 with @xmath257 .
thus , an arbitrary element of @xmath132 may be viewed as a loop @xmath258 in @xmath259 that bounds a disk @xmath181 in @xmath260 . to prove the proposition
, it suffices to show that @xmath258 bounds an orientable surface in @xmath261 . by @xmath8-surjectivity and the fact that the @xmath262 s are generalized @xmath225-neighborhoods
, @xmath258 may be homotoped within @xmath263 to a loop @xmath264 in @xmath265 .
let @xmath266 be the cylinder in @xmath263 between @xmath258 and @xmath264 traced out by that homotopy .
then the disk @xmath267 may be viewed as an element @xmath268\in h_{2}\left ( \widehat{a}_{i}\cup\widehat{a}_{i-1},\partial\widehat{n}_{i+1}\right ) $ ] .
let @xmath269\cup_{\partial\widehat { n}_{i}\times\{0\}}\widehat{a}_{i}\rightarrow\widetilde{a}_{i-2}\cup \widehat{a}_{i-1}\cup\widehat{a}_{i}\ ] ] be the identity on @xmath263 and @xmath270 on @xmath271 $ ] . by pl transversality theory ( see @xcite or @xcite ) , we may after a small proper adjustment that does not alter @xmath272 on @xmath273assume that @xmath274 is a manifold with boundary that is a homeomorphism over a collar neighborhood of @xmath265 .
let @xmath275 be the component of @xmath276 containing that neighborhood .
then , by local characterization of degree , @xmath277 is a proper degree @xmath225 map , and @xmath278 .
thus we have a surjection@xmath279 let @xmath280 $ ] be a preimage of @xmath281 $ ] .
we may assume that @xmath282 is an orientable surface with boundary in @xmath275 . since @xmath272 is the identity on @xmath283
, @xmath284 is homologous in @xmath283 to @xmath285 .
so , without loss of generality , we man assume that @xmath286 .
since @xmath275 lies in @xmath271\cup_{\partial\widehat{n}_{i}\times\{0\}}\widehat{a}_{i}$ ] , we may push @xmath282 , rel boundary , into @xmath263 .
this provides an orientable surface in @xmath287 with boundary @xmath264 . gluing the cylinder @xmath266 to that surface along
@xmath264 produces the bounding surface for @xmath258 that we desire .
early attempts to prove @xmath7-semistability ( hence pseudo - collarability ) with only an assumption of inward tameness , were brought to a halt by the discovery of a key example presented in @xcite .
ideas contained in that example play an important role here , so we provide a quick review . an easy way to denote normal subgroups will be helpful .
let @xmath22 be a group and @xmath288 .
the _ normal closure of s in g _ is the smallest normal subgroup of @xmath22 containing @xmath289 .
we denote it by @xmath290 .
[ main example from @xcite][example : 1 ] for all @xmath1 , there exist 1-ended absolutely inward tame open @xmath0-manifolds with fundamental group system@xmath112 where @xmath291,a_{2}=\left [ a_{2},a_{1}\right ] , \cdots , a_{i}=\left [ a_{i},a_{i-1}\right ] \right\rangle\ ] ] and @xmath292 sends @xmath293 to @xmath293 for @xmath294 and @xmath295 to @xmath225 . by a largely algebraic argument , it was shown that these examples do not have @xmath7-semistable fundamental group at infinity , and thus , are not pseudo - collarable .
notice , however , that each @xmath130 is the normal closure of @xmath295 and @xmath296 $ ] in @xmath297 ; so @xmath298 $ ] .
in other words , @xmath114 satisfies the strong @xmath235-perfectness property , and is therefore @xmath10-semistable .
in addition to the above algebra , these examples have nice topological properties .
although they do not contain small homotopy collar neighborhoods of infinity , they do contain arbitrarily small generalized @xmath225-neighborhoods of infinity @xmath11 for which @xmath186 is @xmath37-homology equivalence .
in fact , they contain a sequence @xmath191 of generalized @xmath225-neighborhoods of infinity with @xmath299 and @xmath300 a @xmath301$]-homology equivalence .
these observations provide much of the motivation for the remainder of this paper .
we begin developing ideas for placing example [ example : 1 ] into a general context .
we will see that end structures like those found in that example are possible only when kernels satisfy a strong relative perfectness condition .
conversely , we will show that whenever such a group theoretic condition is present , a corresponding ` near pseudo - collar ' structure is attainable .
we have already defined pseudo - collar structure on the end of a manifold @xmath5 to be an end structure @xmath194 for which each @xmath193 is a homotopy collar , i.e. , each @xmath302 is a homotopy equivalence .
the existence of such a structure allows us to express each @xmath193 as a union@xmath303 where @xmath304 , and each triple @xmath305 is a compact _ one - sided h - cobordism _ in the sense that @xmath306 is a homotopy equivalence ( and @xmath307 is probably not ) .
one - sided cobordisms play an important role in manifold topology in general , and the topology of ends in particular .
see @xcite for details . for later use , we review a few key properties of one - sided h - cobordisms .
see , for example , ( * ? ? ?
* theorem 2.5 ) [ one - sided]let @xmath308 be a compact cobordism between closed manifolds with @xmath309 a homotopy equivalence
. then 1 .
@xmath309 and @xmath310 are @xmath311 $ ] -homology equivalences , i.e. , @xmath312 \right ) \allowbreak=\allowbreak 0\allowbreak=\allowbreak h_{\ast}\left ( w , q;\mathbb{z } \left [ \pi_{1}\left ( w\right ) \right ] \right ) $ ] 2 .
@xmath313 is surjective , and @xmath314 3 .
@xmath315 is perfect moving forward , we require generalizations of the fundamental concepts : homotopy equivalence , homotopy collar , one - sided h - cobordism and pseudo - collar .
they are as follows : * let @xmath316 be a cw - pair for which @xmath317 induces a @xmath8-isomorphism , and let @xmath318 .
call @xmath31 a @xmath319__-homotopy equivalence _ _ if @xmath320\right ) = 0 $ ] for all @xmath111 .
extension to arbitrary maps is accomplished by use of mapping cylinders . *
a manifold @xmath11 with compact boundary is a @xmath319__-homotopy collar _ _ if @xmath321 and @xmath186 is a @xmath319-homotopy equivalence .
* let @xmath322 be a compact cobordism between closed manifolds and @xmath323 .
we call @xmath324 a @xmath319__-one - sided h - cobordism _ _ if @xmath325 is a @xmath319-homotopy equivalence and @xmath326 induces a surjection on fundamental groups .
* let @xmath194 be a generalized @xmath225-neighborhood end structure on a manifold @xmath5 , chosen so that the bonding maps in @xmath327 are surjective , and let @xmath115 be an augmentation of this sequence .
call @xmath194 a @xmath328 _ pseudo - collar structure _ if each @xmath302 is a @xmath329-homotopy equivalence . _
i ) _ * * _ each of the above definitions reduces to its traditional counterpart when the subgroup(s ) involved are trivial .
_ _ ii ) _ _ in the generalization of one - sided h - cobordism , we _ require @xmath330 _ to be surjective a condition that is automatic when _
@xmath331 _ _ , but not in general .
analogs of the other two assertions of theorem [ one - sided ] will be shown to follow . _
_ _ iii ) _ * * _ for the maximal augmentation , the generalization of pseudo - collar requires only that _ _ each _ @xmath300 _ be a _
@xmath37__-homology equivalence ; whereas , for the trivial augmentation , we have a genuine pseudo - collar .
the key dividing line between those extremes occurs when _
_ @xmath115 _ _ is a small augmentation ( _ _ @xmath332 _ for all _ @xmath31 _ _ ) . in those cases , we call _ _ @xmath194 _ a _ near pseudo - collar _ structure , and
say that a 1-ended _ @xmath5 _ with compact boundary is _ nearly pseudo - collarable _ if it admits such a structure .
the geometric significance of the ` small augmentation ' requirement will become clear in the proof of theorem [ npct technical version ] .
further discussion of that topic is contained in
[ section : remaining questions ] .
_ the following lemma adds topological meaning to the definition of @xmath333-homotopy equivalence .
[ lemma : ( modl ) h.e.]let @xmath316 be a cw - pair for which @xmath317 induces a @xmath8-isomorphism , @xmath334 , and @xmath335 for which @xmath336 . obtain @xmath337 from @xmath12 by attaching a @xmath338-disk @xmath339 along each @xmath340 ; let @xmath341 , and @xmath342 .
then @xmath31 is a @xmath319-homotopy equivalence if and only if @xmath343 is a homotopy equivalence .
let @xmath344 be the covering projection corresponding to @xmath36 .
then @xmath345 is the cover of @xmath12 corresponding to @xmath36 .
viewing @xmath289 as a collection of loops in @xmath12 and @xmath346 the set of all lifts of those loops , then attaching @xmath338-disks to @xmath347 ( and , simultaneously @xmath348 ) along @xmath346 produces universal covers @xmath349 and @xmath350 .
assume now that @xmath317 is a @xmath351-homotopy equivalence . then by shapiro s lemma @xcite , @xmath352 , so by excision @xmath353 .
since both spaces are simply connected , the relative hurewicz theorem implies that @xmath354 ; therefore @xmath355 . by whitehead s
theorem @xmath343 is a homotopy equivalence .
conversely , if @xmath343 is a homotopy equivalence , then its lift @xmath356 is a homotopy equivalence .
therefore @xmath357 , so by excision @xmath358 , and by shapiro s lemma @xmath320\right ) = 0 $ ] .
the following is a useful corollary .
[ lemma : ( modq ) h.e .
] let @xmath316 be a cw - pair for which @xmath317 induces a @xmath8-isomorphism .
suppose @xmath359 . if @xmath360\right ) = 0 $ ] , then @xmath361\right ) = 0 $ ] for any @xmath21 with @xmath362 .
in particular , @xmath363 the next observation is a direct analog of theorem [ one - sided ] .
[ thm : one - sided general ] let @xmath322 be a compact @xmath351-one - sided h - cobordism between closed manifolds with @xmath323 .
let @xmath326 and @xmath364 .
then 1 . both @xmath309 and @xmath310 are @xmath365$]-homology equivalences , i.e. , @xmath366\right ) = \allowbreak0=\allowbreak h_{\ast}\left ( w , q;\mathbb{z } \mathbb{[\pi}_{1}\left ( w\right ) /l]\right ) $ ] , and 2 .
@xmath367 @xmath368 is strongly @xmath369-perfect . first note that by the surjectivity of @xmath370 , there is a canonical isomorphism @xmath371 that is assumed throughout .
let @xmath372 be the covering projection corresponding to @xmath36 , @xmath373 and @xmath374
. then both @xmath375 and @xmath376 are connected , and their projections onto @xmath377 and @xmath378 are the coverings corresponding to @xmath36 and @xmath369 . the assertion that @xmath366\right ) = 0 $ ] is part of the hypothesis , and ( by shapiro s lemma @xcite ) equivalent to the assumption that @xmath379 . to show that @xmath380\right ) $ ] vanishes in all dimensions
, it suffices to show that @xmath381 .
this will follow from poincar duality for noncompact manifolds if we can verify : _ claim .
_ @xmath382 _ _ , where the _ _ @xmath70__ indicates cellular cohomology based on finite cochains .
( see ( * ? ? ?
_ applying lemma [ lemma : ( modl ) h.e . ] , attach @xmath338-cells to @xmath383 along a collection @xmath289 of loops in @xmath377 to kill @xmath36 , obtaining spaces @xmath384 and @xmath385 , and a homotopy equivalence @xmath386 .
since @xmath383 is compact , any strong deformation retraction of @xmath385 onto @xmath384 is proper , and hence , lifts to a proper strong deformation retraction of universal covers @xmath387 onto @xmath388 @xcite .
it follows that @xmath389 .
both universal covers are obtained by attaching disks along the collection @xmath346 of lifts to @xmath375 and @xmath390 of the loops in @xmath289 . by excising the interiors of those disks ,
we conclude that @xmath391 . to verify assertion ( 2 ) , consider the short exact sequence @xmath392 where @xmath393 may be identified with @xmath36 .
lemma [ lemma : 5 term exact ] provides the 5-term exact sequence @xmath394 \rightarrow h_{1}\left ( l^{\prime};\mathbb{z } \right ) \overset{q_{\ast1}}{\longrightarrow}h_{1}\left ( l^{\prime}/k;\mathbb{z } \right ) \rightarrow0\text{.}\ ] ] from which the @xmath369-perfectness of @xmath16 can be deduced by showing that @xmath395 is an epimorphism and @xmath396 an isomorphism .
since @xmath397 induces @xmath398 and since @xmath399 , the the long exact sequence for that pair ensures that @xmath400 .
in addition , the surjectivity of @xmath401 combines with lemma [ lemma : induced surjections on h_2 ] to imply surjectivity of @xmath402 .
with all necessary definitions in place , we are ready to prove the second main theorem described in the introduction .
we begin by stating a strong form of the theorem , written in the style of earlier characterization theorems from @xcite and @xcite .
[ near pseudo - collarability characterization][npct technical version]a 1-ended @xmath0-manifold @xmath5 ( @xmath1 ) with compact boundary is nearly pseudo - collarable iff each of the following conditions holds : 1 .
@xmath5 is inward tame , 2 .
the fundamental group at infinity is @xmath10-semistable , and 3 .
@xmath222 .
recall that condition ( 2 ) presumes the existence of a representation of @xmath403 of the form@xmath404 with a small augmentation @xmath115 ( @xmath405 , for all @xmath31 ) so that each @xmath132 is strongly @xmath133-perfect , where @xmath131 .
first we verify that a nearly pseudo - collarable 1-ended manifold with compact boundary must satisfy conditions ( 1)-(3 ) .
the hypothesis provides a generalized @xmath225-neighborhood end structure @xmath191 on @xmath5 with group data @xmath406 ( @xmath407 ) and a small augmentation @xmath125 ( @xmath408 ) such that each @xmath193 is a @xmath409-homotopy collar . to simultaneously verify ( 1 ) and ( 3 )
, it suffices to exhibit a cofinal sequence of clean neighborhoods of infinity , each having finite homotopy type .
lemma [ lemma : ( modq ) h.e . ] ensures that each @xmath193 is a @xmath410-homotopy collar , and since each @xmath292 is a surjection between finitely presented groups , each @xmath132 is finitely generated as a normal subgroup of @xmath297 .
let @xmath31 be fixed , and @xmath411 a finite collection of loops in @xmath412 that normally generates @xmath132 in @xmath297 . by lemma
[ lemma : ( modl ) h.e . ] , if we abstractly attach a @xmath338-disk @xmath413 along each @xmath414 , we obtain a homotopy equivalence @xmath415 in particular , @xmath416 has the homotopy type of a finite complex .
but , since each @xmath414 represents an element of @xmath128 , we may assume that each @xmath413 is properly embedded in @xmath417 . by thickening these @xmath338-disks to @xmath338-handles , we obtain a clean neighborhood of infinity @xmath418 with finite homotopy type , lying in @xmath419 .
this leaves only @xmath10-semistability to be checked .
we will show that ( [ inverse sequence in structure theorem ] ) satisfies the strong @xmath115-perfectness property ; in other words , each @xmath132 is strongly @xmath133-perfect , where @xmath420 . for each @xmath421 ,
let @xmath422 . _ claim .
_ @xmath423 _ is a _
@xmath424__-one - sided h - cobordism .
_ _ fix @xmath31 and let @xmath425 be the covering corresponding to @xmath426 ; let @xmath427 denote @xmath428 ; and let @xmath261 denote @xmath429 . then @xmath427 is the cover of @xmath430 corresponding to @xmath431 , and @xmath261 is the cover of @xmath193 corresponding to @xmath432 . by lemma
[ lemma : ( modq ) h.e . ] and shapiro s lemma@xmath433\right ) \cong h_{\ast}\left ( \widehat{n}_{i},\partial\widehat{n}_{i}\partial n_{i};\mathbb{z } \right ) \text{,}\ ] ] and from the long exact homology sequence for the triple @xmath434 , excision , and shapiro s lemma @xmath435\right ) = 0.\ ] ] the claim follows .
finally , since the bonding map @xmath436 is represented by the inclusion @xmath437 , @xmath132 is strongly @xmath133-perfect by theorem [ thm : one - sided general ] . for the converse , we must show that conditions ( 1)-(3 ) imply the existence of a near pseudo - collar structure on @xmath5 . though the proof is rather complicated
, it follows the same outline as @xcite , which followed the original proof in @xcite . for a full understanding , the reader should be familiar with @xcite .
the new argument presented here generalizes the the final portions of that proof .
a concise review of @xcite can be found in @xcite . in @xcite and
@xcite the goal was to improve arbitrarily small neighborhoods of infinity to homotopy collars .
that is impossible with our weaker hypotheses ; instead , the goal is to improve neighborhoods of infinity to homotopy collars modulo certain subgroups of their fundamental groups . by condition
( 2 ) the pro - isomorphism class of @xmath438 may be represented by a sequence@xmath112 of finitely presented groups , along with a small augmentation @xmath125 ( @xmath408 , for all @xmath31 ) so that each @xmath132 is strongly @xmath133-perfect , where @xmath439 . by ( * ? ? ?
* lemma 8) there is a sequence @xmath194 of generalized @xmath225-neighborhoods of infinity whose inverse sequence of fundamental groups is isomorphic to a subsequence of @xmath440
. @xmath74{ccccccccc}g_{i_{0 } } & \overset{\lambda_{i_{0}+1,i_{1}}}{\twoheadleftarrow } & g_{i_{1 } } & \overset{\lambda_{i_{1}+1,i_{2}}}{\twoheadleftarrow } & g_{i_{2 } } & \overset{\lambda_{i_{2}+1,i_{3}}}{\twoheadleftarrow } & g_{i_{3 } } & \overset{\lambda_{i_{3}+1,i_{4}}}{\twoheadleftarrow } & \cdots\\ \updownarrow\cong & & \updownarrow\cong & & \updownarrow\cong & & \updownarrow\cong & & \\ \pi_{1}\left ( n_{0},p_{0}\right ) & \overset{inc_{\#}}{\twoheadleftarrow } & \pi_{1}\left ( n_{1},p_{1}\right ) & \overset{inc_{\#}}{\twoheadleftarrow } & \pi_{1}\left ( n_{2},p_{2}\right ) & \overset{inc_{\#}}{\twoheadleftarrow } & \pi_{1}\left ( n_{3},p_{3}\right ) & \overset{inc_{\#}}{\twoheadleftarrow } & \cdots \end{array}\ ] ] this diagram and proposition [ prop : nearly sub ] ensure that , for each @xmath441 , @xmath442 is strongly@xmath443-perfect .
so by passing to this subsequence and relabeling , we may assume that sequence ( [ sequence : pi1-setup for main theorem ] ) and the corresponding subgroup data matches the fundamental group data of @xmath194 .
note here that the @xmath21-groups ( which are not viewed as part of the original data ) are not the same as the previous @xmath21-groups ; they are now preimages of _ compositions _ of the original bonding maps .
next we inductively improve the sequence @xmath444 to generalized @xmath179-neighborhoods of infinity for increasing values of @xmath179 , up to @xmath445 .
we must frequently pass to subsequences , however , each improvement of a given @xmath262 leaves its fundamental group and that of @xmath446 intact ; so at each stage , the ` new ' fundamental group data will be a subsequence of the original ( [ sequence : pi1-setup for main theorem ] ) , along with the subsequence augmentation .
the @xmath21-groups will change as per their definition , but , by proposition [ prop : nearly sub ] , we always maintain the appropriate strong relative perfectness condition .
this neighborhood improvement process uses only the hypothesis that @xmath5 is inward tame ; it is identical that used in ( * ? ? ?
* th . 5 ) and outlined in ( * ? ? ?
* theorem 3.2 ) . to save on notation we relabel the neighborhood sequences and their corresponding groups at each stage , designating the resulting cofinal sequence of generalized @xmath447-neighborhoods of infinity by @xmath191 , with @xmath407 , @xmath448 the corresponding homomorphism , @xmath408 , and @xmath449 . for each @xmath31 ,
let @xmath450 and consider the collection of cobordisms @xmath451 .
the following summary comprises the contents of lemmas 11 and 12 of @xcite , along with new hypotheses regarding kernels . 1 .
each @xmath193 is a generalized @xmath452-neighborhood of infinity .
each induced bonding map @xmath453 is surjective .
3 . each inclusion @xmath454 induces a @xmath8-isomorphism .
4 . each @xmath455 induces a @xmath8-epimorphism with kernel strongly @xmath133-perfect .
@xmath456 for all @xmath457 and all @xmath31
each @xmath458 admits a handle decomposition based on @xmath446 containing handles only of index @xmath447 and @xmath459 .
each @xmath193 admits an infinite handle decomposition with handles only of index @xmath447 and @xmath459 .
each @xmath460 has the homotopy type of a relative cw pair @xmath461 with@xmath462 .
the obvious next goal is attempting to improve the @xmath193 to generalized @xmath459-neighborhoods of infinity , which by item viii ) would necessarily be homotopy collars . in previous work @xcite , @xcite and @xcite , that is the final ( also the most difficult and interesting ) step .
the same is true here , where the weakened hypotheses create greater difficulties and the strategy and end goal must eventually be altered . for
now , we continue with the earlier strategies by turning attention to @xmath463 , which may be viewed as a @xmath464$]-module @xmath465)$ ] .
the content of ( * ? ? ? * lemma 13 ) is given by the next two items . 1 .
@xmath466 is a finitely generated projective @xmath464$]-module .
2 . as an element of @xmath467\right )
$ ] , @xmath468 = \left ( -1\right ) ^{n}\sigma\left ( n_{i}\right ) $ ] , where @xmath469 is the wall finiteness obstruction for @xmath193 . taken together , these elements of @xmath467\right ) $ ]
determine the obstruction @xmath470 found in condition ( 3 ) .
from now on we assume that @xmath471 vanishes .
this is equivalent to assuming that each @xmath469 is the trivial element of @xmath467\right ) $ ] , in other words , each @xmath466 is a stably free @xmath464$]-module .
therefore we have : 1 . by carving out finitely many trivial @xmath447-handles from each @xmath193
, we can arrange that @xmath466 is a finitely generated free @xmath464$]-module .
item ( xi ) can be done so that these sets remain a generalized @xmath472-neighborhood of infinity , and so that their fundamental groups and those of their boundaries are unchanged .
again , to save on notation , we denote the improved collection by @xmath194 .
see ( * ? ? ?
* lemma 14 ) for details .
the finite generation of @xmath466 allows us to , after again passing to a subsequence and relabeling , assume that 1 .
@xmath473 is surjective for each @xmath31 .
the long exact sequence for the triple @xmath474 from there shows that 1 .
@xmath475 is an isomorphism for each @xmath31 ( hence , @xmath476 is a finitely generated free @xmath477$]-module ) .
as above , we may choose handle decompositions for the @xmath478 based on @xmath446 having handles only of index @xmath479 and @xmath480 . from now on , let @xmath31 be fixed . after introducing some trivial @xmath481-handle pairs , an algebraic lemma and some handle slides allows us to obtain a handle decomposition of @xmath478 based on @xmath412 with @xmath459-handles @xmath482 and an integer @xmath483 , such that the subcollection @xmath484 is a free @xmath485 $ ] -basis for @xmath486 .
so we have : 1 .
@xmath485 $ ] -cellular chain complex for @xmath487 may be expressed as@xmath488 where * @xmath489 and @xmath490 represent free @xmath485 $ ] -submodules of @xmath491 generated by the corresponding handles ; * @xmath492 is the free @xmath485 $ ] -module generated by the @xmath447-handles in @xmath478 ; * @xmath493 ; and * @xmath494 takes @xmath495 injectively into @xmath496 .
item xiv ) and the preceding paragraph are the content ( * ? ? ?
* lemma 15 ) . to this point
, we have only used the hypotheses of inward tameness and triviality of the wall obstruction to build the structure described by items ( i)-(xiv ) .
all arguments used thus far appear in @xcite and @xcite , with simpler analogs in @xcite . under the @xmath8-stability hypothesis of @xcite
, @xmath497 can now be killed by sliding the offending @xmath459-handles @xmath498 off the @xmath447-handles and carving out their interiors . under the weaker @xmath7-semistability hypothesis of @xcite , a similar strategy works , but only after a significant preparatory step , made possible by perfect kernels . in @xcite
an alternate strategy was employed . instead of killing @xmath499 by removing its generating handles @xmath500 ,
the task was accomplished by introducing new @xmath447-handles , which became images of the @xmath498 under the resulting boundary map , thereby trivializing the kernel .
complete discussions of these approaches can be found in @xcite and @xcite ; the strategy employed here is based on the latter .
it is helpful to change our perspective by switching to the dual handle decomposition of @xmath478 .
let @xmath501 be a closed collar neighborhood of @xmath502 in @xmath478 , and for each @xmath459-handle @xmath503 identified earlier , let @xmath504 be its dual , attached to @xmath501 .
similarly , for each @xmath447-handle @xmath505 , let @xmath506 be its dual . as is standard , the attaching and belt spheres of a given handle switch roles in its dual .
let @xmath507 , @xmath508 , and @xmath509 be a closed collar on @xmath510 in @xmath511 .
observe that @xmath512 .
see figure [ fig2-ri ] .
[ ptb ] figure2-ends4.eps a simplified view of the next step is that we will find a collection of @xmath513-handles @xmath514 attached to the left hand boundary of @xmath478 and lying in @xmath515 so that the collection @xmath516 of attaching spheres of those @xmath513-handles is algebraically dual to the belt spheres of @xmath517 and has trivial algebraic intersection with the belt spheres of @xmath518 .
adding those @xmath513-handles to the mix , then inverting the handle decomposition again , results in a cobordism with chain complex @xmath519 in which @xmath520 as desired but with a caveat .
although addition of the @xmath513-handles does not change the fundamental group of the cobordism , the arranged algebraic intersections between the attaching spheres of @xmath521 and the belt spheres of the existing @xmath338-handles are @xmath522 $ ] -intersection numbers ; this is the best the hypotheses will allow .
then , to arrive at the desired conclusion that we have effectively killed the relative second homology , it is necessary to switch the coefficient ring to @xmath522 $ ] ( in other words , mod out by @xmath523 ) , and reinterpret ( [ chain complex - enhanced ] ) as a @xmath522 $ ] -complex .
then , letting @xmath524 , it follows that : @xmath525 ; @xmath526 induces a @xmath8-isomorphism ; and @xmath527 ) = 0 $ ] . in other words , @xmath528 is a @xmath529-homotopy collar . in order to carry out the above program , we first identify a collection @xmath516 of pairwise disjoint @xmath338-spheres in @xmath510 algebraically dual over @xmath522 $ ] to the collection @xmath530 of belt spheres of the @xmath338-handles @xmath531 and having trivial @xmath522 $ ] -intersections with the belt spheres @xmath532of the remaining @xmath338-handles @xmath533 . keeping in mind that @xmath534 is canonically isomorphic to @xmath535 , and using the hypothesis that @xmath536 is strongly @xmath537-perfect , such a collection @xmath516 exists , as is shown in @xcite . by general position
, the collection can be made disjoint from the attaching tubes of the @xmath513-handles @xmath538 , so they may be viewed as lying in @xmath446 .
if the collection @xmath539 bounds a pairwise disjoint collection of embedded @xmath513-disks in @xmath515 , regular neighborhoods of those disks would provide the desired @xmath513-handles , and the proof is complete .
( the argument from @xcite provides details . ) for @xmath540 , the issue is just whether the @xmath338-spheres @xmath541 contract in @xmath515 .
( in dimension @xmath542 , a special argument is needed to get pairwise disjoint embeddings . )
contractibility is not guaranteed ; but with additional work it can be arranged .
the additional work involves the _ spherical alteration _ of @xmath338-handles developed in @xcite .
the idea is to alter the @xmath338-handles @xmath543 in a preplanned manner so that the correspondingly altered @xmath516 contract in the new @xmath515 . along the way it will be necessary to reconstruct the @xmath513-handles @xmath538 as well ; for later use , let @xmath544 denote the attaching spheres of those handles .
all details were carefully laid out in @xcite , with this application in mind .
the tailor - made lemma , stated in the final section of that paper , is repeated here .
[ ( * ? ? ?
* lemma 6.1)][lemma from spherical alterations paper]let @xmath545 be a pair of n - manifolds @xmath546 with a common boundary component @xmath547 , and suppose there is a subgroup @xmath369 of @xmath548 for which @xmath549 is strongly @xmath369-perfect .
suppose further that there is a clean submanifold @xmath550 consisting of a finite collection @xmath551 of 2-handles in @xmath552 attached to a collar neighborhood @xmath289 of @xmath547 with @xmath553 inducing a @xmath554-isomorphism ( the 2-handles precisely kill the group @xmath16 ) and a finite collection @xmath555 of pairwise disjoint embedded 2-spheres in @xmath556 , each of which contracts in @xmath552 . then on any subcollection @xmath557 , one may perform spherical alterations to obtain 2-handles @xmath558 in @xmath552 so that in @xmath559 ( where @xmath560 is the correspondingly altered version of @xmath561 ) there is a collection of 2-spheres @xmath562 algebraically dual over @xmath563 $ ] to the belt spheres @xmath564 common to @xmath565 and @xmath558 with the property that each @xmath566 contracts in @xmath567
. furthermore , each correspondingly altered 2-sphere @xmath568 ( now lying in @xmath559 ) has the same @xmath569 $ ] -intersection number with those belt spheres and with any other oriented @xmath452-manifold lying in both @xmath556 and @xmath570 as did @xmath571 . whereas the 2-spheres @xmath572
each contracted in @xmath552 , the @xmath568 each contract in @xmath567 .
apply lemma [ lemma from spherical alterations paper ] to the current setup , with the following substitutions : @xmath74{llll}\underline{\mathbf{lemma}\text { \ref{lemma from spherical alterations paper } } } & & \hspace*{0.05 in } & \underline{\mathbf{current\ situation}\text { } } \\ r^{\prime } & \leftrightarrow & & r_{i}\\ r & \leftrightarrow & & r_{i}\bigcup r_{i-1}\\ b & \leftrightarrow & & \partial n_{i+1}\\ \mathcal{h}^{2 } & \leftrightarrow & & \left\ { \overline{h}_{1}^{2},\cdots,\overline{h}_{s}^{2},\overline{h}_{s+1}^{2},\cdots,\overline{h}_{r}^{2}\right\ } \\ l^{\prime } & \leftrightarrow & & j_{i+1}=\lambda_{i+1}^{-1}\left ( l_{i}\right ) \\ t & \leftrightarrow & & t_{i}=s_{i}\bigcup\left ( \overline{h}_{1}^{2}\cup\cdots\cup\overline{h}_{s}^{2}\cup\overline{h}_{s+1}^{2}\cup\cdots \cup\overline{h}_{r}^{2}\right ) \\ k\in\mathbb{z } & \leftrightarrow & & s\in\mathbb{z}\\ \left\ { h_{j}^{2}\right\ } _ { j=1}^{k } & \leftrightarrow & & \left\ { \overline{h}_{j}^{2}\right\ } _ { j=1}^{s}\\ \left\ { \gamma_{j}^{2}\right\ } _ { j=1}^{k } & \leftrightarrow & & \left\ { \gamma_{j}^{2}\right\ } _ { j=1}^{s}\\ \left\ { \theta_{t}^{2}\right\ } & \leftrightarrow & & \left\ { \theta _ { j}^{2}\right\ } _ { j=1}^{t}\end{array}\ ] ] after applying this lemma , the collection @xmath573 is replaced by altered versions @xmath574 and the original collection @xmath575 is retained .
let @xmath576 and @xmath577 . the collections @xmath516 and @xmath578 are replaced by altered versions @xmath579 and @xmath580 which lie in @xmath581 and contract in @xmath582 . the original @xmath513-handles @xmath583 must be discarded since their attaching tubes have been disrupted ; replacements will be constructed shortly .
when @xmath540 , use general position to choose pairwise disjoint collection of properly embedded @xmath513-disks in @xmath582 with boundaries corresponding to the @xmath338-spheres @xmath584 . those @xmath513-disks may be thickened to @xmath513-handles by taking regular neighborhoods . with all of these handles finally in place , the argument described earlier completes the proof .
when @xmath585 , the same is true , but the @xmath586-@xmath586 argument used in ( * ? ?
4.2 & 5.3 ) is needed in order to find pairwise disjoint embedded @xmath513-disks . _ in reality , we have shown a stronger result than what is stated in theorem [ npct technical version ] . specifically , the near pseudo - collar structures obtained are as close to actual pseudo - collars as the augmentation is to the trivial augmentation . for example , if _
@xmath136 _ is the trivial augmentation , the above argument contains an alternative proof of the main result of @xcite ( stated here as theorem [ pct technical version ] ) . more generally , if _ @xmath136 _ lies somewhere between the trivial augmentation and the standard augmentation , then a near pseudo - collar structure on _
@xmath5 _ can be chosen to reflect that augmentation . _
the main examples of @xcite , described here in example [ example : 1 ] , proved the existence of ( absolutely ) inward tame open manifolds that are not pseudo - collarable . in this section
we construct open manifolds that are absolutely inward tame but not nearly pseudo - collarable .
since the examples from @xcite are nearly pseudo - collarable , the new examples fill a gap in the spectrum of known end structures .
the examples of @xcite began with algebra .
the main theorems of that paper showed that all inward tame open manifolds have pro - finitely generated , semistable fundamental group , and stable @xmath37-homology , at infinity .
the missing ingredient for detecting a pseudo - collar structure was @xmath7-semistability . with that knowledge ,
an inverse sequence of groups satisfying the necessary properties , but failing @xmath7-semistability , became the blueprint for an example .
a nontrivial handle - theoretic strategy was needed to realize the examples , but the heart of the matter was the group theory .
a similar story plays out here
. we will begin with an inverse sequence of finitely presented groups with surjective bonding maps that become isomorphisms upon abelianization ; but this time , in light of theorems [ theorem : inward tame implies ap - semistability ] and [ theorem : npc characterization simple version ] , we want an @xmath9-semistable sequence that is not @xmath10-semistable .
the first step is to identify such a sequence .
let @xmath587 , the free group on the three generators ; @xmath588 $ ] , @xmath589 $ ] , and @xmath590 $ ] ; @xmath591 ; and @xmath592 .
notice that @xmath593 is precisely the commutator subgroup @xmath594 $ ] , so @xmath595 .
let @xmath596 $ ] , @xmath597 $ ] , and @xmath598 $ ] ; @xmath599 ; and @xmath600 .
since @xmath601 , there is an induced epimorphism @xmath602 which abelianizes to the identity map on @xmath603 .
continue inductively , letting @xmath604 $ ] , @xmath605 $ ] , and @xmath606 $ ] ; @xmath607 ; and @xmath608 .
the result is a nested sequence of normal subgroups of @xmath609 , @xmath610 , and a corresponding inverse sequence of quotient groups @xmath611 which abelianizes to the constant inverse sequence @xmath612 a more delicate motivation for our choices is the following : for each @xmath613 , @xmath614 ; similarly , for each @xmath615 , @xmath616 moreover , since the elements of @xmath617 are precisely the commutators of the elements of @xmath618 , @xmath619 .\ ] ] so , for the standard augmentation , @xmath121 , ( [ chosen inverse sequence ] ) is @xmath115-perfect , hence , @xmath9-semistable .
two tasks remain : * prove that ( [ chosen inverse sequence ] ) is not @xmath10-semistable , and * construct 1-ended absolutely inward tame open manifolds with fundamental groups at infinity representable by ( [ chosen inverse sequence ] ) .
since these tasks are independent , the ordering of the following two subsections is arbitrary .
let @xmath620 , the free group on @xmath0 generators .
we will exploit two standard constructions from group theory .
the _ derived series of _ @xmath621 is defined by @xmath622 and @xmath623 $ ] for @xmath624 .
the _ lower central series of _
@xmath625 is given by @xmath626 and then @xmath627 $ ] for @xmath624 . by inspection @xmath628 , @xmath629 , and @xmath630 for all @xmath179 .
a well - known fact , similar in spirit to our goal in this subsection , is that @xmath631 .
the following representation of @xmath621 was discovered by magnus ; our general reference is @xcite .
[ prop : representation ] ( * ? ? ?
* proposition 10.1 ) let @xmath632 be the non - commuting power series ring in indeterminates @xmath633 with @xmath634 for @xmath635 .
then the function @xmath636 @xmath637 induces a faithful representation of @xmath621 into @xmath638 , the multiplicative group of units of @xmath632 . in @xmath632 , the _ fundamental ideal _
@xmath639 is the kernel of the homomorphism @xmath640 that takes each @xmath641 to 0 .
the elements of @xmath639 are all sums of the form @xmath642 where each @xmath643 is a homogeneous polynomial of degree at least one .
consequently , for any positive integer @xmath179 the ideal @xmath644 is made of all sums of the form @xmath645 where each @xmath643 is a homogeneous polynomial of degree at least @xmath179 .
the next proposition and lemma are useful for monitoring the location of commutators in a group . [
prop : commutator ] ( * ? ? ?
* proposition 10.2 ) let @xmath646 be the representation given above .
if @xmath647 such that @xmath648 and @xmath649 , then @xmath650 \right ) -1\in\delta^{r+s}$ ] . by applying proposition [ prop : commutator ] inductively , we obtain the following useful facts .
[ lemma : commutator]for all integers @xmath651 , 1 .
@xmath652 2 .
@xmath653 3 .
@xmath654 , and 4 .
@xmath655 .
we now focus our attention on @xmath609 and its subgroups @xmath656 , as defined earlier .
[ lemma : topology prelim]for each @xmath215 and @xmath657 , 1 .
@xmath658 is a member of at least one free basis for @xmath659 , and 2 .
@xmath660 . assertion ( 1 ) can be obtained from an inductive argument using schreier systems .
a model argument can be found in ( * ? ? ?
* example 8.1 ) .
assertion ( 2 ) follows from ( 1 ) , since the quotient map @xmath661 is the abelianization of @xmath662 . since @xmath663 ,
the following is an easy consequence of lemmas [ lemma : commutator ] and [ lemma : topology prelim ] .
[ lemma : specialized commutators]for each @xmath232 and @xmath664 , 1 .
@xmath665 , and 2 .
@xmath666 .
the definitions of derived and lower central series are clearly applicable to arbitrary groups . to expand those notions further ,
the following definition is useful .
for @xmath667 , let @xmath668 and @xmath669 $ ] for @xmath670 . by normality , @xmath671 .
when @xmath200 is strongly @xmath22-perfect , @xmath672 for all @xmath179 .
[ prop : omega groups get small]for each @xmath232 , there exists @xmath673 and @xmath674 such that 1 . for each @xmath675 , @xmath676 , and 2
. @xmath677 .
let @xmath31 be fixed .
existence of @xmath210 follows from item ( 3 ) of lemma [ lemma : commutator ] .
existence of @xmath678 may be obtained from an inductive application of [ prop : commutator ] .
we shift focus one more time ; from @xmath609 and its subgroups to the quotient groups @xmath679 and their subgroups . in doing so
, we will allow a word in the generators of @xmath609 to represent both an element of @xmath609 and the corresponding element of a @xmath297 .
for example , recalling that @xmath680 , we say @xmath681 .
the following is simple but useful .
[ lemma : images of omega]suppose @xmath682 is a surjective homomorphism , @xmath667 , and @xmath683 . then @xmath684 .
lemma [ lemma : images of omega ] ensures that , for each @xmath685 and all @xmath683 , the quotient maps @xmath686 restrict to epimorphisms @xmath687 [ equation : key epimorphism ] [ prop : no omega ] for @xmath210 and @xmath678 as chosen in proposition [ prop : omega groups get small ] , and each @xmath675 , @xmath688 whenever @xmath689 .
suppose @xmath690 .
surjection ( [ equation : key epimorphism ] ) provides a @xmath691 with cosets @xmath692 .
consequently , there is an @xmath693 with @xmath694 in @xmath609 .
then@xmath74{rcl}\beta(r_{i , j})-1 & = & \beta(h)\beta\left ( w\right ) -1\\ & & \beta(h)\beta\left ( w\right ) -\beta(h)+\beta(h)-1\\ & & \beta(h)\left ( \beta\left ( w\right ) -1\right ) + \left ( \beta ( h)-1\right ) \end{array}\ ] ] since @xmath695 and @xmath696 , then @xmath697 , violating the choice of @xmath210 .
we are now ready for the main result of this subsection .. [ theorem : example ] the inverse sequence @xmath698 is not @xmath10-semistable .
in fact , @xmath699 is not pro - isomorphic to any inverse sequence @xmath109 of surjections that satisfies the strong @xmath700-perfectness property .
we proceed directly to the stronger assertion .
suppose @xmath106 is pro - isomorphic to an inverse sequence @xmath701 of surjections , that is strongly @xmath702-perfect ; in other words , @xmath703 $ ] for all @xmath31 . by proposition [ prop :
nearly sub ] , each subsequence of @xmath109 satisfies the same essential property , so by our assumption , @xmath106 contains a subsequence that fits into a commutative diagram of the following form : @xmath704 passing to a further subsequence if necessary , we may assume that @xmath705 for all @xmath0 . by lemma [ lemma : topology prelim ] , @xmath706 .
choose @xmath707 with @xmath708 .
then , @xmath709 , and consequently @xmath710 $ ] , since @xmath711 is strongly @xmath712-perfect ( again using proposition [ prop : nearly sub ] ) .
therefore @xmath713 for all @xmath714 .
moreover , since @xmath715 , @xmath716 for all @xmath714 , thereby contradicting proposition [ prop : no omega ] .
the goal of this subsection is to construct , for each @xmath1 , a 1-ended open manifold @xmath5 that is absolutely inward tame and has fundamental group at infinity represented by the inverse sequence ( [ chosen inverse sequence ] ) . by theorem [ theorem : npc characterization
simple version ] or [ npct technical version ] , such an example fails to be nearly pseudo - collarable , thus completing the proof of theorem [ theorem : existence of counterexamples ] .
we will construct @xmath5 as a countable union of codimension 0 submanifolds @xmath717 where @xmath718 is a compact core and @xmath719 is a sequence of compact cobordisms between closed connected @xmath720-manifolds where @xmath721 for each @xmath232 , and @xmath722 . letting @xmath723 gives a preferred end structure @xmath194 with @xmath724 for each @xmath31 .
see figure [ fig3-union of cobordisms ] .
[ ptb ] figure3-ends4.eps so that @xmath725-@xmath726 is represented by ( [ chosen inverse sequence ] ) , the @xmath727 will be constructed to satisfy : 1 . for all @xmath232 , @xmath728 and @xmath729
induces a @xmath8-isomorphism , and 2 .
the isomorphism between @xmath730 and @xmath297 may be chosen so that the following diagram commutes : @xmath74{ccccc}g_{i } & & \overset{\lambda_{i+1}}{\longleftarrow } & & g_{i+1}\\ \downarrow\cong & & & & \downarrow\cong\\ \pi_{1}(\gamma_{i},p_{i } ) & \overset{\cong}{\longleftarrow } & \pi_{1}\left ( a_{i},p_{i}\right ) & \overset{\psi_{i+1}}{\longleftarrow } & \pi_{1}(\gamma_{i+1},p_{i+1})\medskip \end{array}\ ] ] here @xmath731 is the composition@xmath732 where @xmath733 is induced by inclusion and @xmath734 is a change - of - basepoint isomorphism with respect to a path @xmath735 in @xmath727 between @xmath210 and @xmath211 .
from there it follows from van kampen s theorem that each @xmath736 induces a @xmath8-isomorphism , so by repeated application of ( a ) and ( b ) , the inverse sequence@xmath737 is isomorphic to ( [ chosen inverse sequence ] )
. it will be also be shown that each @xmath193 has finite homotopy type ; so @xmath5 is absolutely inward tame .
that argument requires specific details of the construction ; it will be presented later .
recall that a @xmath738-handle @xmath739 attached to an @xmath0-manifold @xmath740 and a @xmath741-handle @xmath742 attached to @xmath743 form a _
complementary pair _ if the attaching sphere of @xmath742 intersects the belt sphere of @xmath739 transversely in a single point . in that case @xmath744
; moreover , we may arrange ( by an isotopy of the attaching sphere of @xmath742 ) that @xmath745 is an @xmath746-ball in @xmath747 .
conversely , for any ball @xmath748 , one may introduce a pair of complementary handles @xmath749 so that @xmath750 .
we call @xmath751 a _ trivial handle pair .
_ note that the difference between a complementary pair and trivial pair is just a matter of perspective .
in general , we say that @xmath739 is _ attached trivially _ to @xmath740 if it is possible to attach an @xmath742 so that @xmath751 is a complementary pair .
after a preliminary step where we construct the core manifold @xmath718 , our proof proceeds inductively . at the @xmath752 stage
we construct the cobordism @xmath753 , along with a compact manifold @xmath754 with @xmath755 , to be used in the following stage . throughout the construction ,
we abuse notation slightly by letting @xmath756 $ ] denote a small regular neighborhood of @xmath757 in @xmath758 and @xmath759 $ ] to denote a small regular neighborhood of @xmath760 in @xmath727 * step 0 . *
( preliminaries ) let @xmath761 be the @xmath0-manifold obtained by attaching three orientable 1-handles @xmath762 to the @xmath0-ball @xmath763 . choose a basepoint @xmath764 and let @xmath765 , and @xmath766 be be embedded loops in @xmath767 intersecting only at @xmath768 .
abuse notation slightly by writing @xmath769 a convenient way to arrange that the 1-handles are orientable is by attaching three trivial @xmath770-handle pairs @xmath771 , then discarding the @xmath338-handles .
recall that @xmath772 where @xmath588 $ ] , @xmath589 $ ] , and @xmath590 $ ] .
attach a trio of @xmath338-handles @xmath773 to @xmath761 , where @xmath774 has attaching circle @xmath775 .
choose the framings of these handles so that , if the @xmath338-handles @xmath776 were added back in , then @xmath777 would be trivially attached ( to an @xmath0-ball ) .
let @xmath778 and note that @xmath779 .
* step 1 . *
( constructing @xmath780 and @xmath781 ) attach three trivial @xmath782-handle pairs to @xmath718 , disjoint from the existing handles , then perform handle slides on each of the trivial 2-handles ( over the handles @xmath783 ) so that the resulting 2-handles @xmath784 , @xmath785 and @xmath786 have attaching circles spelling out the words @xmath787 , @xmath788 and @xmath789 , respectively .
this is possible since each @xmath790 can be viewed as a product of the loops @xmath791 and their inverses , which are the attaching circles of @xmath783 .
sliding a 2-handle over @xmath774 inserts the loop @xmath792 into the new attaching circle of that 2-handle ( with @xmath793 depending on the orientation chosen ) . by keeping track of the attaching 2-spheres of the trivial 3-handles after the handle slides , it is possible to attach 3-handles @xmath794 , @xmath795 , and @xmath796 to @xmath797 that are complementary to @xmath784 , @xmath785 , and @xmath786 , respectively
. then @xmath798 for later purposes , it is useful to have a schematic image of the attaching circles of @xmath783 and the attaching @xmath338-spheres of the complementary handles @xmath783 . figure [ fig4-attachingcircle ] provides such an image for one complementary pair .
[ ptb ] figure4-ends4.eps the outer loop represents the attaching circle for an @xmath799 and the shaded region represents the ` lower hemisphere ' of the attaching 2-sphere of @xmath800 ; the ` upper hemisphere ' , which is not shown , is a parallel copy of the core of @xmath799 . within the lower hemisphere , the small central disk represents the lower hemisphere of the 2-sphere before handle slides .
the arms are narrow strips whose centerlines are the paths along which the handle slides were performed ; diametrically opposite paths lead to the same 2-handle , and are chosen to be parallel to a fixed path .
we have indicated this by labeling one pair of centerlines @xmath42 and the other @xmath801 .
the four outer disks are parallel to the cores of the 2-handles over which the slides were made . a twist in the strip leading to an outer disk
is used to reverse the orientation of the boundary of that disk .
thus , diametrically opposite outer disks are parallel to each other , but with opposite orientations .
center points of the outer disks represent transverse intersections with belt spheres of those handles ; thus , @xmath802 and @xmath803 are nearby intersections with the same belt sphere , and similarly for @xmath804 and @xmath805 . by rewriting @xmath806 as @xmath807
, we may reorder the handles so that @xmath784 , @xmath785 , and @xmath786 are attached first . define @xmath808 and note that @xmath809 .
furthermore , @xmath810 so , if we let @xmath811 ) \cup\left ( \cup_{j=1}^{3}h_{1,j}^{2}\right ) \cup\left ( \cup_{j=1}^{3}h_{2,j}^{3}\right ) \text{,}\ ] ] ( the result of excising the interior of a slightly shrunken copy of @xmath781 ) , then @xmath812 . by letting @xmath813 and @xmath814
we obtain the first cobordism of the construction @xmath815 . by avoiding the base point @xmath764 in all of the above handle additions
, we may let the arc @xmath816 be the product line @xmath817 $ ] , with @xmath818 and @xmath819 its end points .
conditions ( a ) and ( b ) are then clear
. * inductive step . *
( constructing @xmath727 and @xmath754 ) assume the existence of a cobordism @xmath820 satisfying ( a ) and ( b ) along with a compact manifold @xmath821 , with the attaching circle of each @xmath822 representing the relator @xmath823 in the presentation of @xmath297 , and @xmath824 .
attach three trivial @xmath782-handle pairs to @xmath758 , then perform handle slides on each of the trivial 2-handles ( over the handles @xmath825 ) so that the resulting 2-handles @xmath826 , @xmath827 and @xmath828 have attaching circles spelling out the words @xmath829 , @xmath830 and @xmath831 , respectively .
this is possible since each @xmath832 can be viewed as a product of the loops @xmath833 and their inverses , which are the attaching circles of @xmath825 . by keeping track of the attaching 2-spheres of the trivial 3-handles under the above handle slides , it is possible to attach 3-handles @xmath834 , @xmath835 , and @xmath836 to @xmath837 that are complementary to @xmath826 , @xmath827 , and @xmath828 , respectively . then @xmath838 a picture like figure [ fig4-attachingcircle ] , but with different indices , describes the current situation .
rewrite @xmath839 as @xmath840 , then reorder the handles so that @xmath826 , @xmath827 , and @xmath828 are attached first .
define @xmath841 and note that @xmath842 .
furthermore , @xmath843 excising the interior of a slightly shrunken copy of @xmath754 gives@xmath844 ) \cup\left ( \cup_{j=1}^{3}h_{i , j}^{2}\right ) \cup\left ( \cup_{j=1}^{3}h_{i+1,j}^{3}\right ) \text{,}\ ] ] then @xmath845 .
noting that @xmath846 and letting @xmath847 , we obtain @xmath753 . by avoiding @xmath848 in all of the handle additions , letting @xmath849 be the product line @xmath850 $ ] , and @xmath211 the new end point , conditions ( a ) and ( b ) are clear . assembling the pieces in the manner described in figure
[ fig3-union of cobordisms ] completes the construction .
in particular , we obtain a 1-ended open manifold@xmath851 whose fundamental group at infinity is represented by the inverse sequence ( [ chosen inverse sequence ] ) . [
remark : dual handle decompositions]_in the construction of _ @xmath753 _ _ , we have written _ _
@xmath760 _ on the left and _ @xmath852 _ on the right to match the blueprint laid out in figure [ fig3-union of cobordisms ] . in that case , the handle decomposition of _
_ implicit in the construction goes from right to left , with handles being attached to a collar neighborhood _
$ ] _ of _ @xmath852__. later , when our perspective becomes reversed , we will pass to the dual decomposition _ _
@xmath854 ) \cup\left ( \cup _ { j=1}^{3}\overline{h}_{1,j}^{n-3}\right ) \cup\left ( \cup_{j=1}^{3}\overline{h}_{2,j}^{n-2}\right)\ ] ] _ where each _ @xmath855 _ is the dual of an original _ @xmath739 _ and _
@xmath759 $ ] _ is a thin collar neighborhood of _ @xmath760__. _
_ the following proposition will complete the proof of theorem [ theorem : existence of counterexamples ] .
[ prop : absolute inward tameness of mn]for the manifolds @xmath5 constructed above , each clean neighborhood of infinity @xmath856 has finite homotopy type .
thus , @xmath5 is absolutely inward tame .
this will be accomplished by examining @xmath857 ( equivalently , @xmath858 viewed as a @xmath859-module ) where @xmath860 .
in particular , we will prove : [ claim : homology of n_i]for each @xmath31 , @xmath857 is trivial in all dimensions except for @xmath861 , where it is isomorphic to the free module @xmath862 .
once this claim is established , proposition [ prop : absolute inward tameness of mn ] follows from ( * ? ? ?
* lemma 6.2 ) . in remark [ siebenmann s lemma ] at the conclusion of this section ,
we explain why this final observation is elementary , requiring no discussion of finite dominations or finiteness obstructions . for proving the claim ,
it is useful to consider compact subsets of the form@xmath863 by repeated application of remark [ remark : dual handle decompositions ] , there is a handle decomposition of @xmath864 based on @xmath865 $ ] with handles only of indices @xmath479 and @xmath480 . by reordering the handles ,
@xmath866 is seen to be homotopy equivalent to a finite relative cw complex @xmath867 where @xmath868 consists of @xmath760 with an @xmath447-cell attached for each @xmath447-handle of @xmath864 followed by an @xmath459-cell for each @xmath869-handle .
in the usual way , the @xmath859-incidence number of an @xmath459-cell with an @xmath472-cell is equal to the @xmath859-intersection number between the belt sphere of the corresponding @xmath447-handle and the attaching sphere of the corresponding @xmath459-handle .
this process produces a sequence@xmath870 of relative cw complexes with direct limit a relative cw pair @xmath871 homotopy equivalent to @xmath872 .
so we can determine @xmath857 by calculating @xmath873 and taking the direct limit as @xmath874 the @xmath859-handle chain complex for @xmath866 ( equivalently , the @xmath859-cellular chain complex for @xmath867 ) looks like@xmath875 where @xmath876 and @xmath877 are finitely generated free @xmath859-modules generated by the handles of @xmath864 , and the boundary map is determined by @xmath859-intersection numbers between the belt spheres of @xmath201-handles and attaching spheres of the @xmath459-handles .
these intersection numbers will be determined by returning to the construction .
beginning with the compact manifold @xmath821 , attach three trivial @xmath782-handle pairs , then perform handle slides on the 2-handles ( over the handles @xmath825 ) to obtain @xmath826 , @xmath827 and @xmath828 with attaching circles @xmath829 , @xmath830 and @xmath878 , respectively .
having kept track of the attaching 2-spheres of the trivial 3-handles under the handle slides , attach 3-handles @xmath834 , @xmath835 , and @xmath836 to @xmath837 that are complementary to @xmath826 , @xmath827 , and @xmath828 , respectively ( all as described ` inductive step ' above ) .
this can all be done so that @xmath834 , @xmath835 , and @xmath836 do not touch the earlier @xmath338-handles @xmath879 , @xmath880 and @xmath881 .
next attach a second trio of trivial @xmath782-handle pairs , taking care that they are disjoint from the existing handles , and slide the trivial @xmath338-handles over the @xmath338-handles @xmath882 so that the resulting 2-handles @xmath883 have attaching circles @xmath884 , @xmath885 and @xmath886 .
again , having kept track of the attaching 2-spheres of the trivial 3-handles under the handle slides , attach 3-handles @xmath887 , @xmath888 , and @xmath889 to@xmath890 that are complementary to @xmath891 , @xmath892 , and @xmath893 , respectively , while taking care that these new 3-handles are completely disjoint from all 2- and 3-handles of lower index . continue this process @xmath894 times , at each stage : attaching three trivial @xmath782-handle pairs disjoint from the existing handles ; sliding the trivial @xmath338-handles over the @xmath338-handles created in the previous step , in the manner prescribed above ; then attaching 3-handles complementary to these new 2-handles ( and disjoint from earlier 2- and 3-handles ) along the images of the attaching 2-spheres of the trivial 3-handles after the handle slides .
since all of the 2- and 3-handles mentioned above , except for the original 2-handles @xmath879 , @xmath880 and @xmath881 , occur in complementary pairs , the manifold we just created is just a thickened copy of @xmath758 ; let us call it @xmath895 . by the standard reordering lemma , we may arrange that the 2-handles are pairwise disjoint , and all are attached before any of the 3-handles which are also are attached in a pairwise disjoint manner .
then @xmath896 where ,
going from the first to the second line , we apply the definition of @xmath758 ; going from the second to the third , we bring the last triple of 2-handles forward to the beginning ; and in going from the third to the fourth , we apply the definition of @xmath897 .
excising a slightly shrunken copy of the interior of @xmath897 from @xmath895 results in a cobordism between @xmath898 and @xmath899 , which has a handle decomposition@xmath900 ) \cup\left ( \cup_{j=1}^{3}h_{i , j}^{2}\right ) \cup\left ( \cup_{s=1}^{k-1}\left ( \cup_{j=1}^{3}h_{i+s , j}^{2}\right ) \right ) \cup\left ( \cup_{s=1}^{k}\left ( \cup_{j=1}^{3}h_{i+s , j}^{3}\right ) \right ) \text{.}\ ] ] comparing this handle decomposition to our earlier construction , reveals that this cobordism is precisely @xmath901 . in order to match the orientation of figure [ fig3-union of cobordisms ] , view @xmath902 as the right - hand boundary and @xmath760 as the left - hand boundary , with 2- and
3-handles being attached from right to left . before switching to the dual handle decomposition
, we analyze the @xmath859-intersection numbers between the attaching spheres of the 3-handles and the belt spheres of the 2-handles .
all should be viewed as submanifolds of the left - hand boundary of @xmath903 ) \cup\left ( \cup_{j=1}^{3}h_{i , j}^{2}\right ) \cup\left ( \cup_{s=1}^{k-1}\left ( \cup_{j=1}^{3}h_{i+s , j}^{2}\right ) \right ) $ ] , which has fundamental group @xmath297 . for each @xmath904 and @xmath675
let @xmath905 denote the attaching 2-sphere of @xmath906 ; and for each @xmath907 and @xmath908 let @xmath909 denote the belt @xmath201-sphere of @xmath910 there are three cases to consider .
* case 1 . * @xmath911 .
then for each @xmath441 , the pair @xmath912 is complementary ; in other words @xmath913 intersects @xmath914 transversely in a single point .
adjusting base paths , if necessary , and being indifferent to orientation ( since it will not affect our computations ) , we have @xmath915 .
if @xmath916 , then @xmath906 does not intersect @xmath917 , so @xmath918 . * case 2 . * @xmath919 . for each @xmath441 , @xmath913 can be split into a pair of disks .
the ` upper hemisphere ' lies in the the 2-handle @xmath920 and intersects @xmath921 transversely in a single point ; that point of intersection was accounted for in case 1 .
the ` lower hemisphere ' is analogous to the one pictured in figure [ fig4-attachingcircle ] . if @xmath922 , then one pair of the diametrically opposite disks has boundaries labelled @xmath923 and @xmath924 and the disks are parallel to the core of @xmath925 , so each intersects @xmath926 transversely in points @xmath927 and @xmath928 . due to the flipped orientation of one of the disks , these points of intersection , between @xmath913 and @xmath926 ,
have opposite sign .
connecting @xmath927 and @xmath928 by a path homotopic to @xmath929 in @xmath905 and a short path @xmath930 connecting @xmath927 and @xmath928 in @xmath926 yields a loop that is contractible in the left - hand boundary of @xmath903 ) \cup\left ( \cup_{j=1}^{3}h_{i , j}^{2}\right ) \cup\left ( \cup_{s=1}^{k-1}\left ( \cup_{j=1}^{3}h_{i+s , j}^{2}\right ) \right ) $ ] .
so together @xmath927 and @xmath928 contribute @xmath185 to to the @xmath859-intersection number of @xmath913 and @xmath926 ; hence , @xmath931 .
similarly @xmath932 .
finally , @xmath913 and @xmath933 do not intersect , so @xmath934 , as well .
* @xmath935 . in this case , the handles @xmath906 and @xmath936 are disjoint , so @xmath937 .
now invert the above handle decomposition , to obtain a handle decomposition of the cobordism @xmath938 , based on @xmath939 containing only @xmath447- and @xmath940-handles . specifically
, we have@xmath941 ) \cup\left ( \cup_{s=1}^{k}\left ( \cup_{j=1}^{3}\overline{h}_{i+s , j}^{n-3}\right ) \right ) \cup\left ( \cup_{j=1}^{3}\overline{h}_{i , j}^{n-2}\right ) \cup\left ( \cup_{s=1}^{k-1}\left ( \cup_{j=1}^{3}\overline{h}_{i+s , j}^{n-2}\right ) \right ) \text{.}\ ] ] since the belt sphere of each @xmath942 is the attaching sphere of its dual @xmath943 and the attaching sphere of of each @xmath944 is the belt sphere of its dual @xmath945 , the incidence numbers between these handles of this handle decomposition are determined ( up to sign ) by the earlier calculations .
so the cellular @xmath859-chain complex for the @xmath866 is isomorphic to @xmath946 where , the @xmath947 summands on the left are generated by the handles @xmath948 and those on the right by @xmath949 .
since @xmath915 for all @xmath950 and all other intersection numbers are @xmath185 , the boundary map is trivial on the @xmath951 copy of @xmath947 ; misses the @xmath952 copy of @xmath947 in the range ; and restricts to an isomorphism @xmath953 elsewhere .
thus @xmath954 where @xmath955 is generated by the @xmath956 summand , and @xmath957 is generated by the @xmath958 summand .
now consider the inclusion @xmath959 and the corresponding inclusion of @xmath859-chain complexes .
the chain complex of @xmath960 will contain an extra @xmath947 summand in each dimension , generated by @xmath961 and @xmath962 , respectively .
the boundary map takes the new summand in the domain onto the previous cokernel , thereby killing @xmath963 , and replacing it with a cokernel generated by @xmath964 .
said differently , the inclusion induced map@xmath965 is trivial . on the other hand , the expansion from @xmath868 to @xmath966 does not change @xmath967 , which is still generated by the handles @xmath968 .
in other words , the inclusion induced map @xmath969 is an isomorphism . taking direct limits ,
we have @xmath970{ccc}\left ( \mathbb{z } g_{i}\right ) ^{3 } & & \text{if } \ast\ = n-2\\ 0 & & \text{otherwise}\end{array } \right . \text{.}\ ] ] so the claim is proved . _
[ siebenmann s lemma]the appeal to ( * ? ? ?
* lemma 6.2 ) may give the impression that obtaining proposition [ prop : absolute inward tameness of mn ] from claim [ claim : homology of n_i ] is complicated that is not the case .
the conclusion can be obtained directly as follows : if _ @xmath971 _ represents the cores of the _ @xmath869__-handles _ _ @xmath972 _ _ , which generate _ _ @xmath857 _ _ , abstractly attach _ _
@xmath459__-disks _ _ @xmath973 _ to _ @xmath760 _ along their boundaries .
this does not affect fundamental groups , so by excision , the pair _
@xmath974 _ has the same _ @xmath859__-homology as _ _
@xmath872 _ _ , with the same generating set .
now attach an _
_ @xmath746__-cell _ _ @xmath975 _ along each sphere _ @xmath976 _ to obtain a pair _ @xmath977 _ with trivial _
@xmath859__-homology in all dimensions .
it follows that _ _
@xmath978 _ is a homotopy equivalence .
but notice that each _
@xmath979 _ has a free face _
_ _ @xmath981 _ collapses onto _
@xmath193__. therefore , _ _
@xmath193 _ is homotopy equivalent to _ @xmath982__. _ _
in the introduction we commented that nearly pseudo - collarable manifolds admit arbitrarily small clean neighborhoods of infinity @xmath11 , containing codimension 0 submanifolds @xmath12 for which @xmath13 is a homotopy equivalence .
call such a pair @xmath983 a _ wide homotopy collar_. the difference , of course , between a wide homotopy collar and a homotopy collar is that , in the latter , the subspace is required to be a codimension 0 submanifold .
the fact that nearly pseudo - collarable manifolds contain arbitrarily small wide homotopy collars is immediate from the following easy lemma .
suppose @xmath984 is a ( @xmath985)-homotopy collar neighborhood of infinity in a manifold @xmath5 ( @xmath986 , where @xmath21 is a normally finitely generated subgroup of @xmath987 .
then @xmath5 contains a wide homotopy collar neighborhood of infinity @xmath988 , where @xmath989 .
choose a finite collection of pairwise disjoint properly embedded 2-disks @xmath990 in @xmath991 with boundaries comprising a normal generating set for @xmath992 .
let @xmath983 be a regular neighborhood pair for @xmath993 and apply lemma [ lemma : ( modl ) h.e . ] .
examples constructed in this paper and in @xcite show that the existence of arbitrarily small wide homotopy collars in a manifold @xmath5 does not imply the existence of a pseudo - collar structure .
the following seems likely but , so far , we have been unable to find a proof . | we continue our study of ends of non - compact manifolds , with a focus on the inward tameness condition . for manifolds with compact boundary , inward tameness , by itself ,
has significant implications .
for example , such manifolds have stable homology at infinity in all dimensions .
we show that these manifolds have ` almost perfectly semistable ' fundamental group at each of their ends .
that observation leads to further analysis of the group theoretic conditions at infinity , and to the notion of a ` near pseudo - collar ' structure .
we obtain a complete characterization of @xmath0-manifolds ( @xmath1 ) admitting such a structure , thereby generalizing @xcite .
we also construct examples illustrating the necessity and usefulness of the new conditions introduced here .
variations on the notion of a perfect group , with corresponding versions of the quillen plus construction , form an underlying theme of this work . |
-reinforcing brakes are a subject of intensive investigation during last years .
working principle of such brakes is to use the wheelset s inertia momentum of a vehicle as the source of power for braking .
one major advantage of self - reinforcing brakes is the energy consumption decrease which makes this research direction perspective . at present
the development of a new brake concept of a self - energizing electro - hydraulic brake for a railway application is being carried out at the institute for fluid power drives and control ( ifas , rwth aachen university ) .
working principle of the braking system can be found in [ 69 ] .
the control task of the self - energizing electro - hydraulic brake is to track the reference signal of the pressure @xmath0 ( the output variable ) in the supply line of the brake system .
the control system with a pure proportional controller allows reaching the goal , as shown in fig .
1 , where the reference signal represents a step function , whose values are the sequence \{27 , 59 , 91 , 59 , 27 bar } , which corresponds to brake forces of \{5 , 10 , 15 , 10 , 5 kn}. however , as seen from the figure , the oscillations of the supply line pressure appear in the system , which cause the oscillations of the brake force . such a behavior of braking , of course , is inadmissible for passenger trains .
therefore , the main requirement for the closed - loop system behavior is to get an aperiodic process of supply line pressure changes independent of the desired brake force level and friction coefficient variations of brake pads .
to reach the required behavior of the controlled variable a controller of nonlinear structure is proposed to synthesize on basis of the exact feedback linearization method [ 3 ] , [ 5 ] .
mathematical description of the electro - hydraulic system can be represented by a nonlinear model of 10^th^ order [ 6 ] .
high order of the model result in the analytically complicated controller structure , which is a problem for simulation and experiments . due to simplifications based on the results of the brake system analysis [ 6 ] as well as the approach used in [ 1 ]
, the basic model can be reduced to a 4^th^ order nonlinear model . for positive ( * case a * ) and negative ( * case b * ) valve opening this model has the form * case
a : * @xmath1 * case b : * @xmath2 where @xmath3 the load pressure of the brake actuator , the pressure in the supply line , the velocity and the control valve spool movement , respectively ; @xmath4 the ratio between piston areas of the brake actuator ; @xmath5 the pressure in the low pressure line ; @xmath6 the control valve parameters ; @xmath7 the known constant parameters ; @xmath8 the parameters dependent of the brake pads friction coefficient ; @xmath9 the control valve input signal .
the output of the brake system is @xmath10 this model will be used for the controller synthesis in the following .
the nonlinear model ( 1 ) ( 3 ) can be represented as & = f(x)+g(x)u + y & = h(x ) the idea of the exact feedback linearization method is to find a nonlinear transformation ( linearization algorithm ) of a control signal , for which the model ( 4 ) , ( 5 ) is linear or equivalent to a linear model in new coordinates @xmath11 , where @xmath12 is the coordinates transformation . an important notion of an input - output model is its relative degree . for linear systems ,
as known , the relative degree is the difference between the number of poles and zeros of the transfer function of a system .
this is also the number of times an output needs to be differentiated in order that an input appears in the equation .
for nonlinear systems it is defined in the similar manner . by differentiating the output @xmath13 and substituting ( 4 )
we get ( see [ 2 ] ) @xmath14 where @xmath15 and @xmath16 are lie derivatives of the function @xmath17 along the vector field @xmath18 and @xmath19 , respectively . for the electro - hydraulic brake , for positive and negative valve opening , @xmath20 for all @xmath21 in the range of operating points .
therefore @xmath22 continuing in this way , we get @xmath23 where @xmath24 .
consequently , @xmath25 the time derivative @xmath26 yields : @xmath27 where the lie derivative @xmath28 for all @xmath21 in the range of operating points and for positive and negative valve opening has the form * case a : * @xmath29 * case b : * @xmath30 thus , the equation ( 8) with the nonzero factor for @xmath9 describes the relation between the input @xmath9 and the output @xmath13 . here , according to the definition in [ 3 ] , the relative degree of the 4th order nonlinear model of the electro - hydraulic brake is 3 . for linear state - space systems , derivatives of an output are chosen as state variables of a plant . using the similar approach
, we determine new coordinates as @xmath31 for positive and negative valve spool movement , the new coordinates @xmath32 can be expressed by the coordinates @xmath21 of the simplified brake model ( 1 ) ( 3 ) , i.e. * case a : * z_1 & = p_sup + z_2 & = t^a_sup x _
+ z_3 & = t^a_sup ( t^a_sup - t^a_l ) x^2 _ + t^a_sup v _ * case b : * z_1 & = p_sup + z_2 & = t^b_sup x _
+ z_3 & = t^b_sup ( t^b_sup - t^b_l ) x^2 _ + t^a_sup v _ in accordance with ( 11 ) , equations ( 5 ) ( 7 ) can be rewritten in the form @xmath33 where @xmath34 the function @xmath35 , for positive and negative valve spool movement , is determined by equations ( 9 ) , ( 10 ) , the function @xmath36 is expressed as * case a : * @xmath37 * case b : * @xmath38 if models ( 1 ) ( 3 ) , and ( 12 ) where equivalent , than the exact feedback linearization problem would be solvable .
this , in its turn , would mean that a control signal of the form @xmath39 where @xmath40 is a new control signal , due to overall compensation of the nonlinear functions @xmath36 and @xmath35 , would lead the plant model ( 1 ) ( 3 ) , on the assumption of parametric certainty of the model and measurability of the state variables @xmath21 , to a system , whose behaviour is exactly identical with the behaviour of the linear model = 0 & 1 & 0 + 0 & 0 & 1 + 0 & 0 & 0 + z_1 + z_2 + z_3 + & + 0 + 0 + 1 + , + y= 1 & 0 & 0 + & z_1 + z_2 + z_3 + , whose output @xmath41 would coincide with the output @xmath42 of the simplified nonlinear model .
however , since rank @xmath43 for all @xmath21 in the range of operating points , which is less then the dimension @xmath44 of the model ( 1 ) ( 3 ) , the mapping @xmath12 is not a diffeomorphism .
let us note that a mapping @xmath45 , where @xmath46 and @xmath47 are smooth manifolds of dimension @xmath48 , is a diffeomorphism if @xmath49 is bijective and both @xmath49 and @xmath50 are smooth mapping ( see appendix a in [ 3 ] ) .
considering aforesaid , the model ( 12 ) is not equivalent to the simplified nonlinear model ( 1 ) ( 3 ) .
this is caused by the internal dynamics , i.e. behaviour of @xmath51-dimensional part of the nonlinear model , which , however , has no influence on the output ( 15 ) of the linear model ( 14 ) . in that case
the problem of input - output linearization can be solved [ 4 ] .
this means that the control ( 13 ) transforms the nonlinear model ( 1 ) ( 3 ) of the electro - hydraulic brake into a model whose input - output behaviour can be represented in the form ( 14 ) , ( 15 ) .
additionally , one more variable @xmath52 has to be supplemented with the new coordinates @xmath32 . according to proposition 4.1.3 in [ 3 ]
it is always possible to find such a function that the jacobian matrix of the mapping @xmath53 is nonsingular at some point @xmath54 . for the plant in question the load pressure of the brake actuator @xmath55
was chosen as the additional variable , i.e. @xmath56 . the jacobian matrix of the mapping @xmath57 is nonsingular at each operating point of the brake .
the time derivative of @xmath58 for positive and negative valve opening has the form * case a : * @xmath59 * case b : * @xmath60 thus , the equivalent model in the new coordinates can be written as & = q(z ) , + = 0 & 1 & 0 + 0 & 0 & 1 + 0 & 0 & 0 + z_1 + z_2 + z_3 + & + 0 + 0 + 1 + ( ( , z)+(,z)u ) , + y&=z_1 , where the equation ( 18 ) for describes the zero - dynamics of the model ( 1 ) ( 3 ) . taking into consideration that @xmath61 , the functions @xmath62 and @xmath63 are expressed as @xmath64 the block diagram of the equivalent model is depicted in fig . 2 .
thus we can conclude that for the self - energizing electro - hydraulic brake the controller synthesis based on the exact feedback linearization method is conventionally divided into the following steps .
firstly , the control signal @xmath40 is determined for the linear model ( 14 ) , ( 15 ) , which describes the relationship between input and output variables of the nonlinear model ( 1 ) ( 3 ) .
secondly , we take into consideration the internal dynamics of the plant , described by the equation ( 18 ) , which has no influence on the output of the linear model , see fig . 3 .
finally , we generate the control signal @xmath9 of the 4^th^ order nonlinear model using the nonlinear transformation for the signal @xmath40 .
the control @xmath9 is expressed as @xmath65 the obtained signal @xmath9 is the input signal for the real brake system , described by the nonlinear model of 10^th^ order .
the linear model ( 14 ) gives the opportunity to use classical methods of linear systems theory to achieve the required control performance , i.e. to get an aperiodic behaviour of the output variable ( 15 ) of the model ( 14 ) .
one such approach is the pole placement method which allows reaching any prescribed placement of closed loop system poles [ 10 ] . according to the method the control input signal @xmath40 for the linear model ( 14 ) is chosen in the form @xmath66 where is the error , i.e. the difference between the reference state vector and the real state vector .
since the reference signal @xmath67 is assumed to be a piecewise constant function , the reference signal @xmath68 of the model ( 14 ) during each time period , where @xmath67 is constant , can be represented as @xmath69 the row - vector @xmath70 provides prescribed poles placement of the following closed loop system @xmath71 where @xmath72 since the electro - hydraulic brake system in question is a siso system , i.e. a single - input single - output system , only the output @xmath0 is measured .
consequently , not all state vector @xmath21 is known .
it means that both the state vector @xmath32 of the model ( 14 ) , ( 15 ) and the variable @xmath58 of the model ( 18 ) are unmeasured .
instead of these variables their estimates have to be used .
therefore , the control system is supplemented with a full - order state observer for ( 14 ) , ( 15 ) . by means of the estimate @xmath73 of the vector
@xmath32 the estimate @xmath74 of the variable @xmath58 can be calculated ( see ( 16 ) and ( 17 ) ) .
the full - order state observer model is determined on the basis of the model ( 14 ) , ( 15 ) and has the form : & = a + b+ k_obs(y- ) , + & = c^t , where @xmath75 is a column - vector of preset coefficients and .
the vector @xmath75 is chosen such that the matrix of closed loop estimation system @xmath76 be hurwitz stable [ 10 ] .
simulation results of the electro - hydraulic brake system of 10^th^ order with the controller of described nonlinear structure are shown in fig .
one can see from the figure , that the nonlinear control algorithm synthesized on the basis of the exact feedback linearization method allows to get the desired closed - loop system behavior .
the results of accomplished simulation demonstrated as well , that the control system is robust with respect to small variations of the friction coefficient .
expansion of the robust stability ranges stimulates further development of the controller for the self - energizing electro - hydraulic brake .
nonlinear control algorithm synthesis of the self - energizing electro - hydraulic brake on the basis of the exact feedback linearization method has been fulfilled . with the help of the obtained algorithm
the problem of the desired closed - loop system behavior has been solved .
the author would like to thank german academic exchange service ( daad ) and ministry of education and science of russian federation for supporting the research stay at the institute for fluid power drives and control in aachen ( project no .
7.375.c2007 ) . m. liermann and c. stammen .
development of a self - energizing electro - hydraulic brake ( sehb ) for rail vehicles . in : _
the tenth scandinavian international conference on fluid power : sicfp07 _ , tampere : university of technology , institute of hydraulics and automation , 2007 .
m. liermann , c. stammen and h. murrenhoff .
development of a self - energizing electro - hydraulic brake ( sehb ) . _ paper no . 2007 - 01 - 4236 , sae commercial vehicle congress and exhibition 2007 _ , chicago , usa , 2007 .
m. liermann , c. stammen and h. murrenhoff .
pressure tracking control for a self - energizing hydraulic brake . in _ the 20th bath symposium on power transmission and motion control _
, bath : hadleys ltd . , p. 315330 | nonlinear control algorithm for a self - energizing electro - hydraulic brake is analytically designed .
the desired closed - loop system behavior is reached via a synthesized nonlinear controller .
shell nonlinear system , nonlinear control , algorithm , electro - hydraulic brake . |
the presence of a supermassive compact object , presumably a black hole , at the center of the dwarf elliptical galaxy m32 has been suspected for some time ( @xcite ) .
the evidence consists of rapid rotation of the stars near the center of m32 and a central peak in the stellar velocity dispersions ( @xcite ; @xcite ; @xcite ; @xcite ; @xcite ) .
the most recent study ( @xcite , 1998 ) used data from the faint object spectrograph ( fos ) on the _ hubble space telescope _ ( hst ) to infer the rotation and dispersion velocities with a spatial resolution of @xmath4 in the inner @xmath5 of m32 .
the fos data revealed a sharper rise in the stellar velocity dispersions than had been observed from the ground ; however the velocity resolution of the fos is limited , making that instrument only marginally useful for the study of a low velocity dispersion system like m32 .
in fact , van der marel et al . (
1998 ) found large point - to - point variations in their velocity dispersion measurements , making the dynamical interpretation uncertain .
nevertheless , the case for a supermassive black hole in m32 , of mass @xmath6 , was considerably strengthened . here
we present observations of m32 carried out using stis , the space telescope imaging spectrograph , on hst .
our data were obtained as part of the stis investigation definition team s ( idt ) key program to observe the nuclei of a sample of @xmath7 nearby galaxies in the spectral region centered on the calcium triplet , @xmath8 .
this paper is the first in a series that will present stellar - kinematical evidence from stis for the presence ( or absence ) of supermassive black holes in galactic nuclei .
the stis data improve on earlier ground - based and fos data from m32 in several ways .
the spatial resolution of stis is @xmath4 , or @xmath9 pc at the distance of m32 , similar to that of the single fos aperture ; however stis provides continuous spatial sampling along a slit .
the spectral resolution of stis in the g750 m mode ( @xmath10 km / s ) is much greater than that of the fos making stis a more suitable instrument for observing m32 , whose velocity dispersion outside of the nucleus is only @xmath11 .
we were able to obtain from the stis spectra not only the lowest - order moments of the stellar velocity distribution the rotation velocity and velocity dispersion but also the full line - of - sight velocity distribution ( losvd ) as a function of position along the major axis . a number of modelling studies ( @xcite ; @xcite ; @xcite ) have made predictions about the observable signatures at hst resolution of a supermassive black hole in m32 .
the black hole is expected to be associated with three kinematical features .
( 1 ) the stellar rotation velocity should remain flat , or rise slightly , into the central resolution element .
( 2 ) the stellar velocity dispersion should exhibit a sudden upturn at a distance of @xmath12 from the center , reaching a central value of @xmath13 or greater depending on the mass of the black hole .
( 3 ) the distribution of line - of - sight stellar velocities in the resolution element centered on the black hole should be strongly non - gaussian , with extended , high - velocity wings .
part of the predicted rise in the velocity dispersion near the center would be due to these wings ; part to blending of the rotation curve from the two sides of the galaxy ; and part to an intrinsic rise in the random velocities .
we see all three signatures of the black hole in the stis data .
the velocity dispersion spike is most impressive ; in terms of the usual parametrization @xmath14 ( the dispersion of the gaussian core of the losvd ) , the central measured value is @xmath1 , while the value of @xmath15 corrected for the non - gaussian wings of the losvd is considerably greater , at least @xmath16 , or @xmath17 times the value in the main body of m32 .
a detailed description of the observations and the data reduction is given in 2 . in
3 we describe the methods used to extract the stellar losvds from the stis spectra .
we carried out independent analyses based on two spectral deconvolution routines : the fcq algorithm of r. bender ( 1990 ) , which is a fourier method ; and the mpl routine of d. merritt ( 1997 ) , which is based on nonparametric function - estimation techniques .
the two algorithms gave consistent results for the low - order velocity moments of m32 ( the mean and dispersion ) but there were systematic differences in the recovered values of @xmath18 , the parameter that measures symmetric deviations of the losvd from a gaussian .
we argue that the @xmath18 values are consistent after taking into account the biases in the two methods ; if our analysis is correct , the true values of @xmath18 in m32 are significantly greater than zero both within the region where the black hole s force is dominant and at larger radii .
a preliminary interpretation of the m32 kinematical data is presented in 4 , where we address two basic questions : have we resolved the central spike in the velocity dispersions ? and : is the shape of the central losvd consistent with what is expected for a nucleus containing a black hole ?
we answer `` yes '' to both questions , although the resolution of the central spike may be marginal if the black hole mass is near the lower limit of the allowed range . in any case
, m32 is the first galaxy for which the imprint of the black hole s gravitation on the stellar velocities has been observed with a resolution comparable to that of the gas - dynamical studies ( e.g. @xcite ; @xcite ) .
we show that both the velocity dispersion spike and the wings of the central losvd are consistent with the predictions of simple dynamical models containing black holes , with masses @xmath19 in the range @xmath20 .
more accurate estimates of @xmath19 will be presented in paper ii where the full kinematical data set will be compared to three - integral axisymmetric models .
we adopt a distance to m32 of @xmath21 mpc ; thus @xmath22 corresponds to 3.25 pc .
m32 was observed on 1998 september 04 with stis in the long - slit mode with wavelength centered on the ca ii triplet feature near 8561 .
stis is described by woodgate et al .
( 1998 ) and its on - orbit performance by kimble et al . (
the present data are part of a survey of the nuclei of nearby galaxies being conducted by the stis idt ( hst program i d : 7566 ) .
the goal of the survey is to place stellar kinematical constraints on the masses of nuclear black holes .
two orbits of data with a total integration time of 4,898 seconds were obtained .
the hst spacecraft tracking was operated in fine lock with a reported jitter of no more than 3 mas rms or 11 mas peak - to - peak .
the aperture was @xmath23 with a position angle of 163@xmath24 , coincident with the m32 isophotal major axis .
the ccd data were read out in the unbinned mode .
spatial sampling at the focal plane was at every @xmath25 , corresponding to a 2-pixel optical resolution of about @xmath26 fwhm .
outside of @xmath27 , the data were binned spatially to enhance the signal .
the spectral resolution was approximately 38 km s@xmath28 .
m32 spectra were obtained in the cr - split mode to assist with cosmic ray ( cr ) identification and rejection .
the location of the galaxy center in the image was moved by approximately 4.5 rows along the aperture between the two orbits to ensure that residual detector sensitivity variations , that may not be completely removed from the data during reduction , are not mistaken for weak features .
this form of dithering also assists with the identification of hot pixels in the ccd which do not rectify well .
a spectrum of the star hr 7615 was obtained with the same stis aperture earlier in the program ( figure [ fig_spectra ] ) .
this bright , k0 iii giant was used as a template when inferring the losvd of the stars in m32 from the stis spectra .
template spectra of a handful of other bright , cool stars of various spectral types were also obtained and used to test the sensitivity of the spectral deconvolution to template mismatch .
a set of spectral images of hr 7615 were taken centered at @xmath29 and @xmath30 with respect to the centerline of the aperture ( i.e. offset along the dispersion direction ) .
these data were added together using appropriate weights to match the aperture illumination profile of m32 .
each of these spectral images has a slightly different shift in its velocity centroid and the combined image provides a more accurate template for determining the kinematics of an extended object such as a galaxy .
internal wavelength calibration images ( `` wavecals '' ) as well as an internal continuum lamp image ( `` flat field '' ) were taken for calibration purposes .
an image that has had all of the instrumental response removed is said be a rectified image .
the contemporaneous flat field spectrum was obtained in the portion of the orbit where hr 7615 was behind the earth .
the flat spectrum was taken through the @xmath31 aperture rather than through the long aperture since the former is superior for removing the pixel - to - pixel detector response of a stellar point source .
a contemporaneous flat field was used to remove the internal fringing which is significant for wavelengths greater than 7500 and which changes over time ( @xcite ) .
we note that the fringing is far more serious for low dispersion spectra especially for high s / n than it is for our data taken in medium dispersion . cosmic rays ( crs ) account for approximately 20% of the total signal and contaminate approximately 5% of the pixels in a typical exposure .
crs were identified and removed using the following procedure .
the centroids in the cross - dispersion direction were determined for each rectified image and the images shifted so that the galaxy core appeared on the same row .
crs were identified and removed by comparing the flux in a given pixel to the flux in the corresponding pixel in subsequent images . for each pixel
, outlying values were rejected and excluded when the frames were averaged together .
our data set for m32 included 4 raw images .
most pixels were found to have 4 frames contributing to their average values , while fewer pixels had 3 or less frames .
only 2 pixels within the central @xmath32 had contributions from no frames .
those pixels were assigned values representing the average of their adjacent pixels .
the m32 data were reduced using two separate approaches : 1 ) `` shift and add , '' and 2 ) `` frame by frame . ''
the latter relies heavily on a standard software package called calstis at goddard space flight center .
the `` shift and add '' method starts by removing the detector response using contemporaneous flat , bias , and dark calibration files .
cr hits are then removed using the procedure described previously .
the frames are averaged together and the resulting frame is remapped to place the spectra from a single location along the aperture onto a single row . as with most spectrographs
, stis produces spectra with s - shaped and pincushion distortions as well as spectra that are not aligned exactly with a row .
a cubic interpolation was used to remap the spectra for later analysis .
the remapping is not perfect , with centroid errors of approximately 0.1 pixel rms .
this level of accuracy was deemed adequate , although we are working to improve it .
in addition , the remapping produces very minor ( @xmath33 ) residual moir ripples in the data , which can be minimized but not completely eliminated .
further work is also underway to measure and correct for this residual affect .
the strength of the `` shift and add '' technique is that it preserves photometric accuracy .
however it has the disadvantage of introducing a subpixel image smearing since each frame is registered to the nearest integer pixel .
the philosophy of the `` frame by frame '' method is to apply all calibration corrections ( including the remapping described above ) to a single frame before the resulting frames are added together .
this approach has the advantage of preserving the highest spatial resolution .
however , a substantial fraction of the pixels in each frame in this case are interpolated values , potentially sacrificing some photometric precision . as noted above , crs contaminate approximately 5% of the pixels in a 20 minute exposure .
after remapping , as many as 20% of the pixels must be assigned a reduced weight for the final averaging since a single pixel often gets remapped partially into several pixels in the new image .
+ fortunately , both approaches to the data reduction were found to give very similar results .
while the `` frame by frame '' method produced a higher apparent s / n ratio than the `` shift and add '' method , the former method is more prone to the introduction of systematic error .
we adopted the more conservative approach of accepting a somewhat higher variance rather than risk the introduction of a bias .
we therefore adopted the `` shift and add '' spectra for the analysis of the losvds in the present study .
figure 1 shows spectra at several positions along the aperture . as noted above ,
we are continuing in our efforts to refine the data reduction techniques still further .
an observed spectrum @xmath34 is the convolution of the line - of - sight velocity distribution @xmath35 of the stars within the aperture with the spectrum of a single star @xmath36 : @xmath37 the goal is to extract an estimate @xmath38 of the true stellar broadening function @xmath35 given @xmath34 and @xmath36 , both observed with the same instrument .
for @xmath36 we adopt the template spectrum of figure [ fig_spectra ] .
two independent deconvolution routines were used .
the first algorithm , the `` fourier correlation quotient '' ( fcq ) method ( @xcite ; @xcite ) , constructs an estimate of the broadening function using fourier techniques .
the fcq routine differs from earlier fourier algorithms ( e.g. @xcite ) in that the deconvolution is based on the template - galaxy correlation function rather than on the spectra themselves .
this approach is less sensitive to template mismatch ( @xcite ) .
the second algorithm , `` maximum penalized likelihood '' ( mpl ) , finds @xmath38 as the solution to a penalized likelihood problem .
the mpl estimate of @xmath35 is computed on a grid in @xmath39 in such a way as to optimize the fit of the convolved template to the galaxy spectrum , subject to a `` penalty '' that measures the lack of smoothness of @xmath38 ( @xcite ) .
both algorithms are nonparametric in the sense that no explicit constraints are placed on the functional form of @xmath38 .
however they differ in two ways that are important for the current study .
the fcq algorithm requires that the absorption lines in the template spectrum be narrow compared to the broadened lines of the galaxy spectrum , i.e. that the galaxy velocity dispersion be large compared to the instrumental resolution .
the mpl routine works well even when the galaxy velocity dispersion is small , as long as both template and galaxy spectra are observed at the same spectral resolution , at least in the case that the template star and galaxy have the same intrinsic absorption line properties ( an assumption that will not be tested here ) .
the two algorithms differ also in the way they deal with the amplification of noise that accompanies the deconvolution .
the fcq routine uses a wiener filter to suppress high - frequency components of the template - galaxy correlation function @xmath40 .
the degree of smoothing is determined by a factor , called here @xmath41 , which fixes the width of the gaussian function used to model the low - frequency , or signal , component of @xmath40 .
the choice @xmath42 corresponds to `` optimal '' filtering and larger values produce less smoothing ; the fcq algorithm adopts a default value of @xmath42 but automatically increases @xmath41 ( to a maximum of 1.3 ) if the recovered losvd shows evidence of significantly non - gaussian wings . in the mpl algorithm ,
the level of smoothing is determined by a factor @xmath43 that multiplies the smoothness penalty function .
this penalty function defines any @xmath35 that is gaussian as `` smooth , '' regardless of its mean or dispersion , via silverman s ( 1982 ) prescription ; in the limit @xmath44 , the mpl estimate of @xmath35 is the gaussian function which is most consistent , in a maximum - likelihood sense , with the galaxy spectrum .
there is no a priori way of computing the optimum value of @xmath43 in the mpl algorithm , a point that we return to below .
the different effects of smoothing on the form of @xmath38 are illustrated in figure [ fig_compare ] , which shows estimates of the losvd in the central resolution element of m32 as computed by the two routines .
both algorithms produce rapidly - fluctuating solutions when undersmoothed , a consequence of the amplification of noise that always accompanies deconvolution .
the only significant difference in this regime is the non - negativity of the mpl estimates , a consequence of the logarithmic form of the penalty function ( @xcite ) . as the smoothing
is increased , systematic differences begin to appear which are related to the different smoothing algorithms in the two codes .
solutions obtained via mpl tend to be more robust with respect to the degree of smoothing , producing in the limit of large @xmath43 a gaussian fit
. however figure [ fig_compare ] suggests that estimates of certain quantities , e.g. the wings of the losvd , might depend sensitively on the choice of smoothing level in either algorithm .
once an estimate of @xmath35 has been obtained , various quantites related to the line - of - sight velocity distribution can be derived .
the simplest of these are the mean and rms velocities , which we denote by @xmath45 and @xmath15 respectively .
as is well known , both quantities are difficult to estimate for numerically - recovered losvd s since they are sensitively dependent on the form of @xmath38 at large velocities where this function is most poorly determined .
a standard alternative is to describe @xmath38 by a gram - charlier or gauss - hermite ( gh ) series , the product of a normalizing gaussian with a sum of hermite polynomials @xmath46 , both expressed in terms of @xmath47 ( @xcite ) .
the parameters @xmath48 and @xmath14 take the place of @xmath45 and @xmath15 ; while their definitions are to an extent arbitrary , these parameters are typically determined by requiring the coefficients of @xmath49 and @xmath50 , called @xmath51 and @xmath52 , to be zero ( @xcite ; @xcite ) . because @xmath48 and @xmath14 describe the gaussian core of the losvd , they are relatively insensitive to deviations of @xmath38 from gaussianity at high velocities .
information about these deviations is contained in the higher - order coefficients @xmath53 , @xmath18 etc . ;
@xmath53 measures asymmetries in @xmath35 and @xmath18 measures the strength of symmetric , non - gaussian wings .
the fcq and mpl algorithms derive the gh parameters from @xmath38 in slightly different ways ; details are given in appendix a. when applied to the stis m32 spectra , the two algorithms were found to give consistent results for the lowest moments of @xmath35 , i.e. @xmath48 , @xmath14 and @xmath53 .
however the estimates of @xmath18 differed significantly at positions outside of the central @xmath4 .
the fcq algorithm gave @xmath54 at almost all positions ; negative values of @xmath18 imply an @xmath35 that falls off more sharply than a gaussian at large velocities .
the mpl algorithm gave values for @xmath18 in the range @xmath55 , almost all positive , corresponding to losvd s with super - gaussian wings .
positive values of @xmath18 are expected near a black hole ( @xcite ; @xcite ) and are also characteristic of models with radially - anisotropic velocity distributions .
we discuss the origin of this discrepancy in appendix b. we believe that the primary reason for the systematic difference in @xmath56 values is the low velocity dispersion of m32 .
when a galaxy s velocity dispersion is comparable to the dispersion of the template star spectrum ( @xmath57 in the case of hr7615 ) , the fcq algorithm has difficulty recovering the true losvd ( figures b2,3 ; @xcite ) . the @xmath38 s recovered by fcq in this regime are more sharply truncated than the true @xmath35 s , leading to systematically low estimates of @xmath56 . for values of @xmath14 and s
/ n comparable to those of m32 at @xmath58 , figure b2 shows that the estimates of @xmath18 generated by fcq depend only weakly on the true @xmath18 , with a bias that approaches @xmath59 for a true @xmath18 of 0.1 .
the mpl algorithm suffers much less from this bias ( figures b3 - 5 ) .
bias in nonparametric function estimates can always be reduced by reducing the degree of smoothing ( e.g. @xcite ) , which in the case of the fcq algorithm means increasing @xmath41 .
figure b5c suggests that increasing @xmath41 from its default value of @xmath60 to values of @xmath61 can reduce the bias in fcq estimates of @xmath18 by factors of @xmath62 or greater , even when @xmath14 is as large as @xmath63 .
we carried out this experiment with the stis data ; the results are shown in figure [ fig_h4 ] .
the average @xmath56 values recovered by fcq in m32 are indeed dependent on @xmath41 ; a change in @xmath41 from @xmath60 to @xmath64 has the effect of increasing @xmath65 from @xmath66 to @xmath67 .
the latter value is essentially identical to the mean value of @xmath56 recovered via mpl .
we note that the dependence of @xmath56 on @xmath41 could be due either to the suggested explanation , i.e. the need to include more frequency channels for dispersions close to the instrumental resolution , or alternatively to a mismatch between galaxy and template spectral properties in the wings of the lines which results in incorrect continuum subtraction
. we will not explore the second possibility here but note again that the monte carlo experiments in appendix b suggest that values of @xmath41 of order 2 are appropriate even when there is no template mismatch .
the values of @xmath56 recovered by mpl are also dependent on the value of the smoothing parameter @xmath43 but much less so ( cf . figure [ fig_compare ] ) , until @xmath43 is made so large that the losvd is forced into a gaussian shape .
the monte carlo tests summarized in figure b5b suggest that the bias in @xmath56 as recovered by mpl is likely to be only of order @xmath68 , several times smaller than with fcq .
we conclude that the values of @xmath56 recovered by the two algorithms are consistent once their different biases are taken into account and that the values returned by mpl are likely to be more accurate .
henceforth we adopt the mpl estimates .
the full set of gh parameters derived from the stis spectra and their @xmath69 confidence intervals are given in table 2 ( fcq ) and table 3 ( mpl ) .
for radii @xmath70 from the center the data were sampled at full resolution(@xmath71 ) while at larger radii they were binned spatially .
the sampling at small radii is fine enough that the data points are somewhat correlated ; this was done to ensure that no information concerning the steep radial gradients of the profiles was lost .
figure [ fig_broad ] presents losvds computed via the mpl algorithm at positions separated by about @xmath72 along the slit .
one expects these broadening functions to obey @xmath73 , since for a point - symmetric galaxy , the velocity distributions should reverse after passing from one side of the galaxy to the other . the losvds of fig .
[ fig_broad ] show approximately the expected symmetry .
the right - hand column of figure [ fig_broad ] plots mean broadening functions averaged over the two sides of the galaxy , @xmath74 $ ] ; the central losvd has been symmetrized about @xmath75 .
these broadening functions show clear and consistent deviations from gaussian form , in two respects .
first , the central losvd exhibits strong super - gaussian `` wings '' at high velocities .
these wings are possibly present also in some of the off - center losvds although with lower amplitude .
second , the off - center losvds are asymmetric , with tails extending toward velocities opposite in sign to the mean velocity at each radius .
these asymmetric tails are similar to those exhibited by a rotating system superimposed on a slowly - rotating bulge .
systematic problems in the spectral deconvolution , e.g. template mismatch or incorrect continuum subtraction , can easily produce features like the wings and tails seen in the broadening functions of figure [ fig_broad ] .
such errors in most cases would be expected to produce features located at the _ same _ velocity on both sides of the galaxy ( e.g. @xcite ) and are therefore an unlikely explanation for the asymmetric tails seen in the off - center losvds .
the strong wings seen in the central losvd might more plausibly be attributed to systematic errors .
however we found that the wings in the central losvd were robust ; they appeared in both the mpl and fcq estimates of @xmath35 ( though less clearly in the latter
see figure [ fig_compare ] ) and were relatively unaffected by changes in the assumed continuum level or slope .
we carried out mpl deconvolutions where the fit to the galaxy spectrum was restricted to the region around only one , or two , of the three calcium - triplet lines ; these losvds also exhibited strong wings .
we also tried using one of the other available stis stellar templates ; again the wings were only slightly affected .
( the adopted template , figure 1 , produced the best overall fit to the galaxy spectrum . ) finally , we show in 4.2 that the wings are consistent with those predicted by stars in the gravitational field of a supermassive black hole .
figure [ fig_rotate ] shows the gauss - hermite parameter @xmath48 , a measure of the stellar rotation , in the inner arc second of m32 . also plotted
are @xmath53 , the lowest , odd gh moment of the losvd , and the `` corrected '' rotation velocity , @xmath76 .
@xmath77 is a closer approximation than @xmath48 to the true mean line - of - sight velocity @xmath45 ( @xcite ) .
the corrected rotation velocity is lower in absolute magnitude than @xmath78 due to the asymmetric wings of the losvd noted above .
the stis rotation curve is consistent with earlier ground - based measurements ( figure [ fig_gb ] ) at radii @xmath79 but with a larger peak value , @xmath11 .
furthermore the rotation curve remains flat or slightly rising into smaller radii than seen heretofore , before falling at @xmath80 due to the blending of light from the two sides of the galaxy .
there is a suggestion of an east - west asymmetry in the rotation curve though the effect is probably not significant .
the @xmath53 profile is approximately antisymmetric about the center of m32 , as expected in a relaxed galaxy .
@xmath81 reaches a maximum value of @xmath82 at @xmath83 and appears to gradually decline at larger radii .
this behavior is similar to that predicted in axisymmetric models ( e.g. figure 8 of @xcite ) where @xmath53 remains essentially constant at radii outside the seeing disk .
the gauss - hermite parameter @xmath14 is shown in figure [ fig_dispersion ] .
also plotted is @xmath18 , the lowest , even moment of the losvd , and the `` corrected '' velocity dispersion , @xmath84 ; @xmath85 is a closer approximation than @xmath14 to the true rms velocity @xmath15 .
the velocity dispersion rises suddenly inside of @xmath0 , approximately the same radius at which the rotation curve begins to fall .
this coincidence suggests that at least part of the rise in @xmath14 is due to averaging of the rotation velocity over the two sides of the galaxy near the center , which has the effect of converting a rotation into an apparent dispersion ( @xcite ) .
the corrected velocity dispersion @xmath85 rises well above @xmath14 near the center due to the strong non - gaussian wings of the losvd .
the central value of @xmath85 is @xmath86 km s@xmath28 ; this should probably be interpreted as a lower limit since @xmath18 is only sensitive to the inner parts of the wings .
( we argue below , based on model fits , that the rms velocity in the central resolution element may be as high as @xmath87 . )
the ground - based data ( figure [ fig_gb ] ) are consistent with the stis dispersions at radii @xmath79 but fail to resolve the continued rise in @xmath14 inside of @xmath5 . dynamical models ( e.g. @xcite ; @xcite ) predict @xmath88 profiles similar to that in figure [ fig_dispersion ] when observed with hst resolution : a central maximum ; a rapid drop , to small or negative values , at @xmath89 ; and a nearly constant value at larger radii .
the predicted drop at @xmath4 is due to blending of the light from the two sides of the galaxy , which broadens the low - velocity part of the losvd and lowers the observed @xmath18 .
the predicted central value of @xmath18 depends strongly on the black hole mass and on the psf ; our value , @xmath90 , is larger than in the two studies just cited , but these studies were based on rather low assumed black hole masses , @xmath91 .
the true black hole mass is probably greater ( @xcite ) .
the behavior of @xmath56 at larger radii is surprising .
previous observational studies ( e.g. @xcite ; @xcite ) have returned smaller estimates for @xmath18 in m32
. however we believe that these earlier results are not inconsistent with ours giving the difficulties involved with estimating this parameter .
the van der marel ( 1994a ) study was based on wht observations with a much lower spatial resolution than the stis data . at radii @xmath92 ,
the value of @xmath56 recovered by those authors was strongly affected by the psf blending of the rotation curve discussed above , yielding negative values in the central aperture . outside of @xmath93 , van der marel et al .
found @xmath56 to increase sharply to @xmath94 on both sides of the galaxy ( their figure 12 ) .
van der marel ( private communication ) notes that the values of @xmath56 derived from the wht data depended sensitively on the choice of template spectrum and on the algorithm for continuum subtraction . using a single , best - fit template ,
@xmath56 was found to lie between @xmath95 and @xmath96 throughout the inner @xmath32 ; the lower values of @xmath56 in the published paper were derived using a spectral deconvolution routine that constructs an `` optimal '' template by linear superposition of a set of stellar spectra . in another ground - based study ,
bender , kormendy & dehnen ( 1996 ) applied the fcq algorithm to cfht data of higher spatial resolution and found @xmath97 inside of @xmath98 , gradually falling to @xmath99 at @xmath100 .
however the spectral resolution in this study was only @xmath101 and the derivation of @xmath53 and @xmath18 correspondingly difficult ; as noted above , we also found @xmath102 from the stis data using the fcq algorithm and argued that these values were significantly negatively biased .
although we believe that all of these studies are consistent with our conclusion that @xmath18 is significantly positive throughout the nucleus of m32 , we are less willing to strongly endorse the precise @xmath56 values shown in figure [ fig_dispersion ] , due to the sensitive dependence of this parameter on the details of the spectral deconvolution algorithm , continuum subtraction , smoothing level , etc . in paper ii we will construct dynamical models based on a range of assumed @xmath18 profiles in order to test the dependence of the inferred black hole mass on this parameter .
we may also compare our results to the van der marel et al .
( 1997 , 1998 ) hst / fos measurements of @xmath48 and @xmath14 ( figure [ fig_fos ] ) .
the fos measurements were taken through square apertures as small as @xmath4 on a side , hence their spatial resolution is comparable to that of the stis data .
however the fos is a low spectral resolution instrument and not well suited to objects like m32 with a relatively low velocity dispersion ; furthermore there are difficulties in positioning the fos and these were probably the cause of the large point - to - point variations seen by van der marel et al .
one advantage of stis over fos is the continuous spatial sampling which avoids potential errors in aperture placement .
we find a hint in the stis data of the asymmetry seen in the fos @xmath103 profile ( a more rapid falloff on the west side ) .
the central fos value of @xmath14 seems significantly bigger than found here , and the fos rotation velocities are systematically lower .
the stis data show what appear to be clear signatures of the gravitational influence of a massive compact object on the stellar velocity distribution within the central parsec of m32 . here
we use simple dynamical models to address the question of whether these observed features are consistent with the presence of a supermassive black hole .
our aim is not to derive the best possible estimate of the black hole mass or the stellar velocity distribution those are the goals of paper ii but rather to address two , more basic issues concerning the interpretation of the data .
\1 . the velocity dispersion profile ( figs .
[ fig_dispersion ] , [ fig_gb ] ) exhibits a sudden upturn at a distance of @xmath0 from the center , presumably due in part to the gravitational force from a massive compact object . at roughly the same radius ,
the rotation curve falls ( figs .
[ fig_rotate ] , [ fig_gb ] ) , presumably due to blending of light from opposite sides of the galaxy which are rotating in opposite directions .
the blending would also be expected to contribute to the rise in the observed velocity dispersion ( @xcite ) , consistent with the fact that the upturn in @xmath15 and the drop in @xmath45 occur at roughly the same radius .
we would like to estimate the degree to which the velocity dispersion spike is a product of this blending , and the degree to which it is due to a real upturn in the stellar random velocities .
in other words : do the stis data resolve the black hole s sphere of influence ?
the central losvd in m32 exhibits strong , super - gaussian wings ( fig .
[ fig_compare ] , [ fig_broad ] ) .
such wings are a generic prediction of the black hole model ( @xcite ; @xcite ; @xcite ) ; they result from stars on high - velocity orbits within the black hole s sphere of influence .
however systematic errors in the spectral deconvolution can also produce spurious features in the losvd s , particularly at large velocities where the form of the broadening function is only weakly constrained by the spectra .
we would like to verify that the inferred , non - gaussian shape of the central losvd is consistent with that expected from the black hole model .
we note that m32 is the only galaxy so far to exhibit either a resolved central spike in the stellar velocity dispersions , or strong non - gaussian wings in the nuclear losvd .
either feature , if observed with sufficient spatial resolution and s / n , could independently place strong constraints on the mass of a central black hole .
we will in fact generate estimates of @xmath19 from our analyses of both features but we stress that the best estimates of @xmath19 can only come from more complete modelling based on the entire kinematical data set .
one expects to see a rise in the stellar velocities in a hot stellar system at a distance @xmath104 from the black hole , where @xmath105 the black hole s `` radius of influence '' ( @xcite ) . here
@xmath19 is the black hole mass and @xmath106 is the stellar velocity dispersion at @xmath107 .
setting @xmath108 and @xmath109 km s@xmath28 ( figures [ fig_dispersion ] , [ fig_gb ] ) gives @xmath110 .
this value seems comfortably larger than the hst / stis resolution but it is based on an assumed value of @xmath19 and furthermore it refers to the true radius whereas we observe the galaxy in projection , which tends to hide otherwise sharp features . modelling the spike therefore requires us to predict the true , projected velocity field of the galaxy in two dimensions on the plane of the sky , including both random and rotational velocities , and then to convolve it with the stis psf .
we begin by constructing solutions to the stellar hydrodynamical equations .
we assume the simplest possible axisymmetric model for the stars , in which the velocity dispersions are isotropic in the meridional plane @xmath111 , i.e. @xmath112 ; the model is flattened by an inequality between @xmath113 and the mean square azimuthal velocity @xmath114 .
the second moments of the internal velocity distribution are given by @xmath115 where @xmath116 is the stellar number density , @xmath117 is the azimuthal velocity and @xmath118 is the combined gravitational potential from the stars and the central black hole , @xmath119 ( @xcite ) .
we evaluated these expressions assuming a stellar density @xmath120^{\beta}\left[1+(m / c)^2\right]^{\gamma},\ \ \ \ m^2=\varpi^2+(z / q)^2,\ ] ] a parametrized form proposed by van der marel et al .
( 1998 ) ; those authors found a good match between their model ( observed edge - on ) and m32 with the parameters @xmath121 and @xmath122pc@xmath123 .
the luminosity density @xmath124 is converted into a mass density @xmath125 by the factor @xmath126 , with @xmath127 the mass - to - light ratio of the stars in solar units .
given @xmath128 and @xmath129 as obtained from equations ( 3 ) and ( 4 ) , the projected , line - of - sight , mean square velocity @xmath130 is obtained by a density - weighted integration through the galaxy , assumed here to be edge - on : @xmath131{\varpi d\varpi\over\sqrt{\varpi^2-r^2}}\ ] ] ( @xcite ) , where @xmath132 are coordinates on the plane of the sky and @xmath133 is measured parallel to the long axis of the galaxy s figure ; @xmath134 is the stellar surface density .
values of @xmath130 were computed on a rectangular grid of @xmath135 locations with separations of @xmath136 in @xmath133 and @xmath137 in @xmath138 .
these values were then convolved with the stis psf and averaged over the pixel area and the aperture after weighting by the model surface brightness .
the stis psf at @xmath139 has a fwhm of @xmath140 .
the psf is also slightly asymmetric ( @xcite ) .
we ignored this slight asymmetry when carrying out the convolutions with our models .
the second velocity moments of models constructed in this way are uniquely determined by the two parameters @xmath141 that specify the potential . figure [ fig_contour ] shows the goodness of fit of the models to the observed , mean square velocities ; only data points within the inner @xmath142 were used in evaluating @xmath143 . as estimates of @xmath130 , we took @xmath144 , where @xmath77 and @xmath85 are the gauss - hermite parameters corrected by @xmath53 and @xmath18 respectively ( figures [ fig_rotate ] , [ fig_dispersion ] ) .
the best - fit model has @xmath145 and @xmath146 with @xmath147 ; a @xmath148 of unity includes models with @xmath19 as small as @xmath149 and as large as @xmath150 .
the degree of net rotation in our models may be adjusted by partitioning the azimuthal motions between streaming , @xmath151 , and dispersion , @xmath152 .
we followed the standard practice ( @xcite ) of making @xmath153 a weighted average of @xmath113 and @xmath114 , i.e. @xmath154 the parameter @xmath155 ( assumed independent of position ) may be varied between zero ( corresponding to no streaming motions ) and a maximum value , of order unity , at which @xmath156 is forced below zero at some point in the meridional plane ; @xmath157 yields an `` isotropic oblate rotator . ''
when we add @xmath155 as a free parameter and require the models to fit the observed rotation and velocity dispersion profiles separately , the best - fit values of @xmath141 were nearly unchanged but @xmath148 increased to @xmath158 since the model is now being asked to fit twice as many data points with only one extra parameter .
the best - fit value of @xmath155 was found always to be close to @xmath159 , implying slightly smaller @xmath152 ( i.e. greater rotation ) than in an isotropic oblate rotator .
figure [ fig_three ] compares the data with the predicted profiles for @xmath160 , close to the best - fit value , and for @xmath161 and @xmath162 .
we draw the following conclusions from these comparisons .
the lowest - order moments of the line - of - sight velocity distribution in m32 are reasonably well fit near the center by our simple axisymmetric model , with a black hole mass @xmath6 .
the rotation curve is best fit by a smaller mass ( @xmath163 ) and the velocity dispersions by a larger mass ( @xmath164 ) ; if we require the models only to fit the mean square velocities , the fit is essentially perfect within the inner arc second .
the stis observations probably come close to resolving the central upturn in the stellar velocity dispersions , which is predicted to occur at a projected radius of @xmath4 for @xmath165 and @xmath166 for @xmath167 .
smearing of the stellar rotation field probably accounts for only a small part of the observed upturn in the dispersions .
we note that dehnen ( 1995 ) was able to improve the fit of his axisymmetric models to the m32 data then available by varying the ratio of rotational to non - ordered azimuthal motions , which in our models would correspond to varying @xmath155 with position .
if our simple model for the internal dynamics of m32 is approximately correct , figures [ fig_contour ] and [ fig_three ] imply that the stis data can place strong constraints on the mass of the central black hole .
however such a conclusion must await the results of the more complete modelling of paper ii .
that study may yield tighter constraints on @xmath19 , due to the use of the full kinematical data set ; or weaker constraints , due to the increased flexibility of general , anisotropic models .
the losvd near the projected center of a galaxy containing a black hole is expected to be very non - gaussian due to high velocity stars orbiting near the central mass ( @xcite ) .
if the stellar density follows a power law into the center , @xmath168 , the losvd in an aperture containing the black hole has power - law wings , @xmath169 , @xmath170 ( @xcite ) .
the amplitude of these wings depends on the ratio of black hole mass to stellar mass within the aperture , and on the slope @xmath171 of the stellar density profile ( @xcite ) , among other factors .
the wings are most prominent in a galaxy , like m32 , for which the stellar cusp is steep , @xmath172 , since a large fraction of the light near the projected center comes from the region near the black hole . here
we ask whether the strong wings seen in the central m32 broadening function ( figures [ fig_compare ] , [ fig_broad ] ) are consistent with the black hole model .
to answer this question we must compute a stellar distribution function @xmath173 and integrate it over the two velocity components in the plane of the sky ( @xcite ) .
the result must then be smeared by the instrumental resolution .
we once again restrict ourselves to the simplest model which permits a meaningful comparison with the data ; in this case , a spherical galaxy with an isotropic distribution function , @xmath174 .
our model will ignore rotation , which acts to broaden the central losvd but leaves it symmetric .
in any case we do not know what the contribution of rotation to the stellar velocity distribution is very near to the black hole .
for the stellar density profile we assume the spherical version of equation ( 4 ) ; @xmath175 then follows from the standard formula ( @xcite ) and the projected velocity distribution is also straightforward to compute ( @xcite ) .
we chose to fix @xmath127 for the stars at @xmath176 as the black hole mass was varied ; this @xmath127 reproduces approximately the observed mean square velocities for @xmath177 .
figure [ fig_center ] compares the central losvd in m32 to the broadening functions predicted by the spherical model , for black hole masses @xmath178 .
the m32 losvd has been symmetrized about @xmath75 .
the fit is quite reasonable for @xmath179 , although the high - velocity wings are better fit by still larger masses .
the model losvd for @xmath179 has the parameters : @xmath180 the first two numbers are essentially identical to the values inferred from the m32 data ( figure [ fig_dispersion ] ) ; @xmath18 is lower than , but consistent with , the m32 estimate ( @xmath181 ) .
we conclude that the central losvd in m32 is consistent with that expected for a stellar nucleus containing a massive compact object , with a mass comparable to that found in the fit to the axisymmetric models .
we used hst and stis to obtain stellar absorption line spectra near the center of m32 in a wavelength region centered on the calcium triplet .
the spectra were analyzed using two independent spectral deconvolution routines ; these gave fully consistent results except in the case of the gauss - hermite @xmath18 parameter , but we argued that the differences could be reconciled after taking into account the different biases of the two algorithms .
the stellar rotation velocities in m32 are slightly higher than observed from the ground and remain constant into @xmath182 from the center .
the velocity dispersions exhibit a clear spike beginning at approximately the same radius .
these two kinematical profiles are consistent with those predicted by simple axisymmetric models containing central black holes with masses in the range @xmath3 .
the stellar losvds show significant deviations from gaussian form as measured by the gauss - hermite parameters @xmath53 and @xmath18 .
the central losvd is particularly non - gaussian , exhibiting strong , high - velocity wings .
we showed that the amplitude of these wings is consistent with that predicted by simple models containing black holes with masses of order @xmath183 .
we thank w. dehnen and r. van der marel for helpful discussions .
this work was supported by nasa grants nag 5 - 3158 and nag 5 - 6037 , by nsf grant ast 96 - 17088 , and by stis gto funding .
data presented here were based on observations with the nasa / esa _ hubble space telescope _
, obtained at the space telescope science institute , which is operated by the association of universities for research in astronomy , inc . ( aura ) , under nasa contract nas5 - 26555 . the two spectral deconvolution algorithms described above yield nonparametric estimates @xmath38 of the stellar losvd . here
we describe the methods used by the two algorithms to derive the gh moments from @xmath38 .
let @xmath184 be the distribution of line - of - sight stellar velocities in the aperture centered at @xmath185 .
define the gh moments of @xmath186 as h_i(x , y ) = 2_-^ n(x , y;v)g(w)h_i(w)dv , where @xmath46 are the hermite polynomials ( as defined by @xcite ) and the weight function g(w ) = 1_0 e^-w^2/2 , w=(v - v_0)/_0 has three free parameters @xmath187 . following van der marel & franx ( 1993 ) , we choose these parameters at every point @xmath185 such that h_0(x , y)=1 , h_1(x , y)=h_2(x , y)=0 .
these definitions impose the following implicit conditions on @xmath187 : _ 0 = _ 0_-^ n(v)e^-w^2/2dw , 0 = _ -^ n(v)e^-w^2/2 w dw , 0 = _ -^ n(v)e^-w^2/2(2w^2 - 1 ) dw . the relations ( a4 )
define a nonlinear minimization problem with solutions @xmath187 given @xmath35 .
the mpl algorithm ( @xcite ) derives the three parameters in just this way , using the nag routine e04fdf to minimize the sum @xmath188 as a function of @xmath187 .
the higher - order gh moments are then derived using equation ( a1 ) , by numerical integration over @xmath38 .
most spectral deconvolution algorithms of which we are aware derive the parameters @xmath187 in a different way .
the losvd is compared to the trial function ( v ) = _ 0_0e^-w^2/2 where @xmath189 is the index of the highest gh moment fitted to @xmath38 ; typically @xmath190 . the integrated square deviation between @xmath38 and @xmath191 is then minimized by varying the @xmath192 free parameters @xmath193 .
this is the technique used by the fcq algorithm .
a theorem ( @xcite ) guarantees the equivalence of the two approaches if @xmath194 in equation ( a5 ) ( @xcite ) . however
if @xmath195 , and if the input @xmath35 can not be precisely represented by a finite gh series with @xmath196 , the results given by the two algorithms will differ .
for instance , in attempting to represent an @xmath35 having @xmath197 using @xmath190 , the fcq algorithm will adjust @xmath14 and @xmath18 to incorrect values in order to better fit the high - velocity wings of the profile with the limited number of terms allowed to it .
this is illustrated in figure a1 , which shows the values of @xmath14 and @xmath18 generated by the second algorithm , @xmath198 and @xmath199 , compared to the true values for an input @xmath35 with @xmath200 and nonzero @xmath201 : n(v ) = 1e^-v^2/2\{1 + 0.15h_4(v ) + h_6h_6(v)}. for @xmath202 , the errors in @xmath14 and @xmath18 as derived from the second algorithm are @xmath203 and @xmath204 respectively .
here we compare the performance of the fcq and mpl algorithms given simulated data .
our primary goal is to understand the source of the systematic offset of @xmath18 values as derived from the m32 spectra by the two algorithms ( 3.1 ) .
two independent sets of tests were carried out , the first by r. bender and the second by d. merritt .
all tests were based on synthesized galaxy spectra generated from the stis template spectrum ( figure [ fig_spectra ] ) by convolving it with an assumed @xmath35 and adding noise .
the first set of tests addressed the accuracy of fcq estimates when the galaxy velocity dispersion is low .
it is well known that the accuracy of fcq begins to fall off when the galaxy velocity dispersion becomes comparable to the dispersion of the template spectrum ( e.g. @xcite ) .
figure b1 shows values of @xmath205 recovered by fcq given a gaussian - broadened template spectrum and thirty random noise realizations .
the default value ( @xmath206 ) of the smoothing parameter was used .
there is a positive bias in the estimated values beginning at @xmath207 ; the bias increases with decreasing @xmath14 becoming significant for @xmath208 .
the bias is only weakly dependent on s / n .
this bias in the estimation of @xmath14 is unlikely to be important for the nucleus of m32 where @xmath209
. figure b2 shows the performance of fcq at recovering @xmath18 .
the template spectrum was broadened using an @xmath35 of the form n_1(v ) = 1e^-v^2/2_0 ^ 2\{1 + h_4h_4(v/_0 ) } with various values of @xmath14 and @xmath18 .
figure b2 reveals significant biases in @xmath56 for @xmath210 , even when s / n is as great as 100 .
when @xmath208 and s / n @xmath211 , characteristic of m32 at @xmath58 , the bias in @xmath18 is @xmath212 for an input @xmath18 of @xmath213 . the second set of tests compared the performance of the fcq and mpl algorithms on galaxy spectra generated from the broadening function n_2(v ) = 111+(v/)^2 , a lorentzian function , with @xmath214 .
this losvd is qualitatively similar to what is expected in a black - hole cusp , with @xmath215 high - velocity wings .
the non - trivial gh parameters are _ 0=0.76986 _ 0=108.07 km s^-1 h_4=0.14546 h_6=0.01850 .
figure b3 shows mean estimates of @xmath35 obtained using the two algorithms for 100 random realizations of the noise .
the smoothing parameter in both algorithms was adjusted to minimize the mean square error of @xmath38 ( as defined below ) for each value of s / n .
there is a greater bias in the fcq estimates , as well as a persistent `` ringing '' at high velocities .
figure b4 plots the mean integrated square error ( mise ) and integrated square bias ( isb ) of the recovered broadening functions as functions of s / n ; in the case of the fcq algorithm , the integrated errors are shown both for the optimal choice of smoothing parameter @xmath216 that minimizes the mise , as well as for the value chosen by the algorithm ( @xmath217 ) .
the mise of an estimate @xmath218 is defined as = e\{(x ) - f(x)}^2dx = \{e(x ) - f(x)}^2dx + ( e\{^2(x ) } - e\{(x)}^2 ) dx = isb+ iv , the sum of the integrated square bias isb and the integrated variance iv ( @xcite ) ; here @xmath219 denotes the expectation value , i.e. the average over many random realizations of the noise .
the mise and isb displayed in fig .
b4 were divided by the normalizing factor @xmath220 ^ 2 dv$ ] .
the mise of the mpl estimates falls roughly as a power law , @xmath221\sim ( s / n)^{-1.3}$ ] , close to the asymptotic @xmath222 of parametric estimators .
approximately @xmath223 of the total square error comes from the bias and @xmath223 from the variance . in the case of the fcq algorithm
, the mise behaves in a more complicated way with s / n , at first falling with s / n then appearing to level off for s / n @xmath224 .
this levelling off is a consequence of the low - velocity - dispersion bias of fcq discussed above . for the fcq estimates ,
the bulk of the mise is due to the variance ; adjusting the smoothing parameter primarily affects the bias and has little effect on the mise . for s / n @xmath225 ,
the mean square error of the optimal fcq estimate is a factor @xmath17 greater than that of the mpl estimates .
the bias in @xmath38 is in the direction of wider and more steeply truncated functions , particularly in the case of the fcq estimates ( fig .
this bias in @xmath38 is consistent with the negative bias found above in estimates of @xmath18 .
figure b5 compares the ability of the two algorithms to recover @xmath18 from the lorentzian @xmath226 .
plotted there are the mean square error ( mse ) and bias in estimates of @xmath18 from 100 random noise realizations ; the mse is defined , for any estimated parameter @xmath227 , as ( ) = e\ { - p}^2 , which can also be decomposed into contributions from the squared bias sb and the variance v : mse ( ) = ( e - p)^2 + ( e\{^2 } - e\{}^2 ) = sb ( ) + v ( ) .
the mse of estimates obtained with the mpl algorithm again varies roughly as a power law , @xmath228 .
the bias in the mpl estimates is always negative , i.e. in the direction of more gaussian @xmath35 s ; for @xmath229 , this bias is a modest @xmath230 , dropping to @xmath231 for s / n @xmath232 .
the fcq estimates of @xmath18 show a considerably greater error , both in the bias and the variance .
two sets of fcq estimates were made : first using the default value of the smoothing parameter returned by the code , @xmath233 ; and second using the optimum value @xmath216 that minimized the mse of the @xmath18 estimates at each s / n .
for the default value of @xmath41 , the algorithm returns mean estimates of @xmath18 that lie in the range @xmath234 for all values of s / n @xmath235 , an average error of @xmath236 .
however the optimum smoothing parameter for the recovery of @xmath18 was found to vary strongly with s / n , from @xmath237 for s / n @xmath238 to @xmath239 for s / n @xmath240 ( fig .
nevertheless a substantial bias remains when @xmath216 is used , of order @xmath241 even for s / n @xmath242 .
these biases are larger than found above using a more gaussian @xmath35 with smaller @xmath18 and suggest that fcq estimates of @xmath18 may be substantially biased even for @xmath14 as large as @xmath243 when the true @xmath35 is sufficiently non - gaussian .
bahcall , j. & wolf , s. 1976 , apj , 209 , 214 bender , r. 1990 , a&a , 229 , 441 bender , r. , kormendy , j. & dehnen , w. 1996 , apj , 464 , l123 bender , r. , paquet , a. & nieto , j .- l . 1991 , , 246 , 349 bender , r. , saglia , r. & gerhard , o. 1994 , mnras , 269 , 785 bower , g. a. 1998 , apj , 492 , l111 bower , g. a. et al . 2000 , in preparation carter , d. & jenkins , c. r. 1993 , mnras , 263 , 1049 dehnen , w. 1995 , mnras , 274 , 919 dressler , a. & richstone , d. o. 1988 , apj , 324 , 701 eddington , a. s. 1916 , mnras , 76 , 572 ferrarese , l. & ford , h. c. 1999 , apj , 515 , 583 fillmore , j. a. 1986 , aj , 91 , 1096 gerhard , o. e. 1993 , mnras , 265 , 213 goudfrooij , p. , baum , s. a. & walsh , j. r. 1997 , in the 1997 hst calibration workshop with a new generation of instruments , 100 .
kimble , r. a. et al .
1998 , apj l , 492 , 283 merritt , d. 1987 , apj , 313 , 121 merritt , d. 1993 , apj , 413 , 79 merritt , d. 1997 , aj , 114 , 228 merritt , d. 1999 , pasp , 111 , 129 myller - lebedeff , w. 1908 , math .
64 , 388 peebles , p. j. e. 1972 , gen .
grav . , 3 , 63 qian , e. e. , de zeeuw , p. t. , van der marel , r. p. & hunter , c. 1995 , mnras , 274 , 602 sargent , w. l. w. , schechter , p. l. , boksenberg , a. & shortridge , k. 1977 , apj , 212 , 326 satoh , c. 1980 , pasj , 32 , 41 silverman , b. w. 1982 , ann .
10 , 795 silverman , b. w. 1986 , density estimation for statistics and data analysis ( chapman & hall : london ) thompson , j. r. & tapia , r. a. 1990 , nonparametric function estimation , modeling , and simulation ( siam , philadelphia ) , 37 tonry , j. l. 1987 , apj , 322 , 632 van der marel , r. 1994 , apj , 432 , 91 van der marel , r. & franx , m. 1993 , apj , 407 , 525 van der marel , r. p. , rix , h .- w . ,
carter , d. , franx , m. , white , s. d. m. & de zeeuw , t. 1994a , mnras , 268 , 521 van der marel , r. p. , wyn - evans , n. , rix , h .- w . , white , s. d. m. & de zeeuw , t. 1994b , mnras , 271 , 99 van der marel , r. p. , de zeeuw , p. t. & rix , h. w. 1997 , apj , 488 , 119 van der marel , r. p. , cretton , n. , de zeeuw , p. t. & rix , h. w. 1998 , apj , 493 , 613 woodgate , b. e. et al .
1998 , pasp , 110 , 1183 ll gain ( e@xmath244/adu ) & 1.0 + wavelength coverage & 8275 8847 + reciprocal dispersion ( pixel@xmath28 ) & 0.56 + aperture & @xmath245 + comparison line fwhm ( pixels ) & 2.0 + @xmath246 & 7644 + instrumental dispersion ( @xmath247 ) ( ) & 17.1 + spatial scale ( @xmath248 pixel@xmath28 ) & 0.05071 + 4.497 & 25.2 & 32.2 & 67.2 & 19.0 & 0.110 & 0.436 & -0.289 & 0.436 & 38.0 & 60.2 & 19.6 & 75.5 + 4.000 & 49.7 & 31.5 & 97.5 & 32.2 & 0.053 & 0.294 & -0.055 & 0.294 & 58.7 & 58.9 & 84.3 & 77.3 + 3.465 & 55.5 & 23.6 & 60.7 & 13.9 & -0.063 & 0.353 & -0.230 & 0.353 & 48.8 & 44.0 & 26.5 & 54.9 + 2.983 & 54.4 & 20.6 & 82.2 & 18.9 & 0.017 & 0.227 & -0.090 & 0.227 & 56.8 & 38.3 & 64.1 & 49.6 + 2.546 & 35.0 & 18.8 & 60.1 & 14.0 & 0.040 & 0.284 & -0.148 & 0.284 & 39.1 & 35.0 & 38.3 & 44.4 + 2.173 & 24.1 & 25.1 & 103.3 & 18.7 & -0.052 & 0.221 & -0.148 & 0.221 & 14.8 & 46.9 & 65.9 & 59.4 + 1.842 & 35.9 & 17.2 & 64.0 & 13.0 & -0.058 & 0.245 & -0.144 & 0.245 & 29.5 & 32.2 & 41.4 & 40.8 + 1.562 & 61.5 & 17.2 & 88.8 & 15.6 & -0.074 & 0.176 & -0.093 & 0.176 & 50.1 & 32.2 & 68.6 & 41.5 + 1.336 & 42.8 & 13.7 & 76.9 & 13.5 & -0.013 & 0.162 & -0.067 & 0.162 & 41.0 & 25.6 & 64.3 & 33.4 + 1.159 & 34.4 & 15.4 & 81.6 & 13.9 & 0.010 & 0.171 & -0.095 & 0.171 & 35.8 & 28.6 & 62.6 & 37.0 + 1.034 & 53.6 & 15.3 & 82.9 & 12.6 & 0.059 & 0.168 & -0.122 & 0.168 & 62.1 & 28.6 & 58.1 & 36.6 + 0.932 & 52.8 & 13.5 & 74.4 & 11.2 & -0.024 & 0.165 & -0.120 & 0.165 & 49.7 & 25.2 & 52.5 & 32.3 + 0.831 & 65.7 & 12.9 & 72.2 & 10.4 & -0.078 & 0.163 & -0.129 & 0.163 & 55.9 & 24.2 & 49.4 & 30.8 + 0.729 & 69.6 & 11.2 & 77.1 & 11.3 & -0.001 & 0.132 & -0.057 & 0.132 & 69.4 & 20.9 & 66.3 & 27.4 + 0.654 & 62.4 & 12.6 & 72.5 & 11.0 & 0.018 & 0.158 & -0.105 & 0.158 & 64.7 & 23.5 & 53.9 & 30.3 + 0.603 & 57.5 & 12.6 & 79.0 & 8.1 & -0.055 & 0.144 & -0.180 & 0.144 & 50.0 & 23.4 & 44.2 & 29.2 + 0.553 & 56.8 & 13.0 & 75.3 & 11.1 & -0.014 & 0.157 & -0.112 & 0.157 & 55.0 & 24.3 & 54.7 & 31.2 + 0.502 & 58.8 & 10.7 & 72.4 & 10.1 & 0.019 & 0.135 & -0.083 & 0.135 & 61.1 & 20.1 & 57.7 & 26.1 + 0.451 & 58.8 & 11.6 & 81.1 & 11.2 & 0.000 & 0.130 & -0.074 & 0.130 & 58.8 & 21.6 & 66.4 & 28.2 + 0.401 & 65.2 & 10.4 & 76.0 & 10.2 & -0.089 & 0.124 & -0.065 & 0.124 & 53.5 & 19.4 & 63.9 & 25.3 + 0.350 & 58.8 & 10.0 & 86.0 & 10.2 & -0.067 & 0.106 & -0.055 & 0.106 & 48.8 & 18.7 & 74.4 & 24.6 + 0.299 & 52.9 & 9.6 & 89.3 & 10.3 & -0.072 & 0.098 & -0.037 & 0.098 & 41.8 & 18.0 & 81.2 & 23.8 + 0.248 & 60.6 & 8.4 & 77.4 & 8.9 & -0.041 & 0.099 & -0.044 & 0.099 & 55.1 & 15.7 & 69.1 & 20.8 + 0.198 & 53.4 & 8.1 & 90.4 & 8.3 & -0.049 & 0.082 & -0.057 & 0.082 & 45.7 & 15.2 & 77.8 & 20.0 + 0.147 & 40.8 & 9.2 & 99.4 & 8.9 & 0.038 & 0.084 & -0.072 & 0.084 & 47.3 & 17.2 & 81.8 & 22.4 + 0.096 & 38.9 & 10.2 & 112.2 & 11.2 & 0.012 & 0.083 & -0.030 & 0.083 & 41.2 & 19.1 & 104.0 & 25.4 + 0.046 & 36.3 & 8.7 & 123.0 & 10.8 & -0.009 & 0.064 & 0.019 & 0.064 & 34.4 & 16.2 & 128.7 & 22.1 + -0.005 & -2.9 & 9.0 & 136.6 & 12.6 & -0.013 & 0.060 & 0.075 & 0.060 & -6.0 & 16.8 & 161.7 & 23.8 + -0.056 & -24.6 & 9.6 & 124.2 & 10.1 & 0.032 & 0.071 & -0.045 & 0.071 & -17.7 & 18.1 & 110.5 & 23.9 + -0.106 & -34.4 & 9.2 & 104.2 & 9.8 & 0.070 & 0.081 & -0.041 & 0.081 & -21.7 & 17.3 & 93.8 & 22.9 + -0.157 & -40.9 & 9.9 & 98.3 & 11.5 & 0.087 & 0.092 & -0.007 & 0.092 & -26.1 & 18.6 & 96.6 & 25.0 + -0.208 & -53.1 & 9.9 & 90.8 & 10.4 & 0.037 & 0.100 & -0.045 & 0.100 & -47.3 & 18.6 & 80.8 & 24.6 + -0.259 & -64.0 & 9.9 & 87.3 & 10.6 & 0.013 & 0.103 & -0.036 & 0.103 & -62.0 & 18.4 & 79.6 & 24.5 + -0.309 & -62.6 & 9.0 & 76.4 & 8.5 & 0.052 & 0.107 & -0.079 & 0.107 & -55.7 & 16.8 & 61.6 & 21.8 + -0.360 & -63.4 & 10.2 & 80.8 & 10.8 & 0.045 & 0.114 & -0.042 & 0.114 & -57.1 & 18.9 & 72.5 & 25.0 + -0.411 & -50.9 & 12.0 & 90.4 & 13.8 & 0.038 & 0.120 & -0.010 & 0.120 & -44.9 & 22.3 & 88.2 & 30.0 + -0.461 & -57.6 & 11.6 & 70.5 & 11.0 & 0.016 & 0.150 & -0.078 & 0.150 & -55.7 & 21.7 & 57.0 & 28.2 + -0.512 & -60.1 & 10.8 & 63.5 & 9.0 & 0.033 & 0.155 & -0.118 & 0.155 & -56.5 & 20.2 & 45.1 & 25.9 + -0.563 & -55.0 & 13.5 & 83.4 & 12.1 & 0.092 & 0.147 & -0.096 & 0.147 & -41.7 & 25.2 & 63.8 & 32.5 + -0.614 & -51.5 & 15.5 & 93.2 & 15.2 & 0.047 & 0.151 & -0.067 & 0.151 & -43.9 & 28.9 & 77.9 & 37.8 + -0.664 & -51.4 & 13.4 & 66.4 & 11.6 & 0.053 & 0.183 & -0.106 & 0.183 & -45.3 & 25.0 & 49.2 & 32.1 + -0.715 & -58.4 & 13.2 & 78.0 & 10.7 & 0.173 & 0.154 & -0.126 & 0.154 & -35.0 & 24.8 & 53.9 & 31.5 + -0.766 & -46.2 & 18.0 & 87.3 & 17.3 & 0.189 & 0.188 & -0.076 & 0.188 & -17.6 & 34.1 & 71.0 & 43.9 + -0.841 & -31.8 & 12.2 & 77.9 & 10.1 & -0.079 & 0.142 & -0.119 & 0.142 & -42.4 & 22.7 & 55.2 & 29.1 + -0.943 & -51.4 & 13.4 & 77.1 & 11.1 & -0.009 & 0.158 & -0.121 & 0.158 & -52.6 & 25.0 & 54.3 & 32.0 + -1.044 & -44.9 & 13.5 & 74.9 & 13.8 & 0.079 & 0.164 & -0.054 & 0.164 & -34.6 & 25.3 & 65.0 & 33.2 + -1.145 & -48.6 & 14.9 & 67.0 & 11.8 & 0.037 & 0.202 & -0.132 & 0.202 & -44.3 & 27.8 & 45.3 & 35.4 + -1.247 & -54.2 & 15.0 & 73.8 & 14.1 & 0.064 & 0.184 & -0.082 & 0.184 & -46.0 & 27.9 & 59.0 & 36.3 + -1.372 & -51.8 & 23.3 & 109.2 & 23.9 & -0.044 & 0.194 & -0.053 & 0.194 & -60.1 & 43.5 & 95.0 & 57.2 + -1.549 & -69.3 & 17.1 & 83.9 & 12.4 & 0.180 & 0.186 & -0.154 & 0.186 & -43.1 & 32.2 & 52.3 & 40.5 + -1.778 & -52.1 & 16.9 & 75.6 & 15.4 & -0.075 & 0.204 & -0.092 & 0.204 & -61.9 & 31.7 & 58.5 & 40.9 + -2.052 & -56.6 & 16.3 & 81.1 & 13.5 & 0.077 & 0.183 & -0.121 & 0.183 & -45.8 & 30.5 & 57.1 & 39.0 + -2.382 & -66.0 & 18.0 & 78.2 & 19.2 & 0.172 & 0.209 & -0.039 & 0.209 & -42.8 & 34.0 & 70.7 & 44.4 + -2.765 & -54.6 & 29.4 & 102.9 & 29.4 & 0.048 & 0.259 & -0.061 & 0.259 & -46.1 & 54.8 & 87.5 & 71.7 + -3.185 & -53.9 & 21.7 & 62.1 & 17.2 & 0.129 & 0.317 & -0.131 & 0.317 & -40.1 & 40.6 & 42.2 & 51.5 + -3.673 & -70.0 & 51.5 & 167.0 & 55.9 & -0.003 & 0.280 & -0.033 & 0.280 & -70.8 & 96.0 & 153.5 & 127.6 + -4.206 & -46.9 & 25.2 & 30.0 & 33.3 & 0.039 & 0.763 & 0.047 & 0.763 & -44.9 & 47.0 & 33.5 & 65.3 + -4.722 & -67.4 & 32.1 & 6.5 & 53.7 & -0.295 & 4.505 & 0.167 & 4.505 & -70.7 & 65.8 & 9.1 & 92.0 + ccccccccccccc 4.559 & 18.9 & 12.8 & 51.0 & 10.4 & 0.176 & 0.140 & 0.022 & 0.033 & 37.9 & 19.7 & 53.7 & 12.1 + 4.184 & 70.2 & 18.9 & 97.2 & 30.0 & -0.235 & 0.155 & 0.040 & 0.079 & 7.15 & 51.4 & 106.7 & 31.4 + 3.545 & 47.3 & 8.4 & 47.9 & 6.8 & 0.054 & 0.102 & 0.008 & 0.032 & 56.9 & 10.1 & 48.8 & 8.9 + 3.068 & 54.4 & 8.1 & 60.3 & 12.7 & -0.054 & 0.098 & 0.094 & 0.033 & 49.5 & 14.3 & 74.2 & 16.4 + 2.632 & 44.8 & 6.7 & 45.4 & 7.0 & -0.008 & 0.111 & 0.028 & 0.018 & 35.4 & 14.1 & 48.6 & 9.6 + 2.247 & 27.2 & 9.8 & 70.0 & 10.1 & 0.043 & 0.094 & 0.111 & 0.060 & 29.0 & 12.9 & 89.0 & 14.4 + 1.922 & 37.6 & 7.8 & 62.7 & 9.0 & -0.102 & 0.080 & 0.048 & 0.031 & 26.9 & 11.4 & 70.0 & 10.1 + 1.648 & 71.8 & 7.9 & 64.6 & 12.6 & -0.124 & 0.076 & 0.080 & 0.027 & 60.2 & 16.7 & 77.3 & 16.9 + 1.415 & 46.4 & 6.9 & 54.8 & 7.1 & -0.008 & 0.079 & 0.069 & 0.023 & 46.1 & 10.0 & 64.1 & 9.5 + 1.242 & 44.7 & 7.2 & 64.4 & 8.6 & -0.079 & 0.091 & 0.077 & 0.026 & 37.6 & 10.7 & 76.5 & 11.6 + 1.111 & 49.1 & 8.1 & 87.5 & 11.2 & 0.070 & 0.079 & 0.037 & 0.047 & 62.7 & 13.2 & 95.5 & 11.8 + 1.009 & 58.2 & 7.2 & 72.7 & 11.2 & -0.144 & 0.087 & 0.104 & 0.041 & 37.1 & 13.9 & 91.1 & 14.1 + 0.908 & 71.7 & 4.6 & 49.6 & 6.5 & -0.034 & 0.072 & 0.036 & 0.012 & 69.3 & 6.7 & 54.0 & 8.5 + 0.806 & 69.1 & 4.9 & 68.7 & 8.3 & 0.000 & 0.070 & 0.082 & 0.023 & 70.6 & 8.2 & 82.6 & 8.7 + 0.705 & 49.5 & 7.0 & 67.4 & 8.7 & 0.138 & 0.090 & 0.071 & 0.026 & 74.2 & 13.4 & 79.1 & 13.4 + 0.654 & 65.3 & 5.3 & 64.0 & 7.1 & -0.004 & 0.060 & 0.041 & 0.016 & 63.5 & 7.4 & 70.4 & 8.5 + 0.604 & 68.3 & 5.2 & 61.7 & 7.7 & -0.157 & 0.056 & 0.051 & 0.018 & 52.0 & 9.3 & 69.4 & 8.1 + 0.553 & 62.2 & 5.4 & 56.3 & 6.4 & -0.034 & 0.060 & 0.060 & 0.019 & 61.5 & 7.0 & 64.6 & 8.8 + 0.502 & 64.6 & 4.4 & 54.8 & 6.7 & -0.022 & 0.053 & 0.070 & 0.022 & 63.9 & 5.9 & 64.2 & 7.6 + 0.451 & 62.5 & 5.3 & 74.1 & 7.0 & -0.071 & 0.066 & 0.054 & 0.028 & 53.6 & 7.9 & 83.9 & 7.1 + 0.401 & 63.9 & 5.0 & 61.8 & 7.6 & -0.040 & 0.048 & 0.089 & 0.018 & 61.8 & 7.4 & 75.3 & 10.6 + 0.350 & 55.4 & 4.3 & 78.0 & 7.5 & -0.040 & 0.049 & 0.062 & 0.024 & 51.1 & 7.7 & 89.8 & 9.2 + 0.299 & 60.1 & 4.5 & 76.8 & 7.0 & -0.178 & 0.046 & 0.053 & 0.023 & 37.4 & 8.9 & 86.8 & 6.5 + 0.249 & 62.5 & 3.9 & 66.5 & 6.8 & -0.021 & 0.039 & 0.080 & 0.029 & 60.1 & 5.2 & 79.6 & 7.2 + 0.198 & 55.2 & 3.9 & 82.4 & 6.6 & -0.065 & 0.042 & 0.066 & 0.022 & 45.9 & 6.7 & 95.8 & 6.2 + 0.147 & 43.3 & 4.7 & 101.2 & 6.0 & -0.004 & 0.048 & -0.022 & 0.035 & 42.5 & 7.9 & 95.8 & 6.4 + 0.096 & 38.8 & 5.1 & 110.3 & 6.3 & -0.009 & 0.044 & 0.032 & 0.032 & 41.0 & 8.3 & 118.9 & 7.1 + 0.046 & 37.2 & 5.0 & 120.6 & 7.0 & -0.028 & 0.038 & 0.118 & 0.034 & 31.3 & 9.1 & 115.6 & 9.6 + -0.005 & -3.7 & 4.7 & 132.0 & 6.9 & -0.051 & 0.037 & 0.139 & 0.034 & -15.2 & 10.7 & 176.8 & 11.5 + -0.056 & -24.5 & 5.4 & 130.0 & 7.0 & -0.006 & 0.037 & 0.010 & 0.034 & -25.7 & 10.1 & 133.2 & 9.7 + -0.106 & -31.6 & 4.7 & 103.0 & 6.0 & 0.018 & 0.039 & 0.028 & 0.030 & -28.1 & 7.5 & 110.2 & 8.4 + -0.157 & -39.2 & 4.9 & 92.4 & 6.8 & 0.026 & 0.042 & 0.076 & 0.028 & -35.9 & 8.0 & 109.7 & 8.5 + -0.208 & -52.1 & 4.8 & 87.6 & 6.5 & 0.033 & 0.041 & 0.033 & 0.031 & -47.5 & 7.1 & 94.8 & 6.4 + -0.259 & -63.7 & 4.9 & 79.8 & 7.0 & 0.012 & 0.042 & 0.062 & 0.030 & -62.1 & 6.4 & 91.9 & 7.9 + -0.309 & -64.2 & 5.0 & 67.6 & 6.5 & 0.090 & 0.039 & 0.068 & 0.022 & -56.6 & 6.3 & 78.9 & 9.0 + -0.360 & -65.6 & 5.8 & 68.0 & 10.4 & 0.089 & 0.049 & 0.122 & 0.030 & -56.0 & 11.5 & 88.3 & 12.4 + -0.411 & -55.7 & 5.5 & 65.1 & 7.8 & 0.017 & 0.043 & 0.102 & 0.022 & -54.3 & 7.6 & 81.5 & 10.8 + -0.461 & -59.9 & 4.9 & 58.8 & 6.3 & 0.071 & 0.055 & 0.067 & 0.025 & -53.7 & 6.0 & 68.4 & 7.4 + -0.512 & -59.0 & 5.4 & 50.7 & 6.4 & 0.028 & 0.046 & 0.061 & 0.018 & -58.4 & 5.8 & 58.3 & 9.8 + -0.563 & -53.9 & 6.5 & 72.4 & 8.5 & 0.097 & 0.061 & 0.058 & 0.025 & -40.3 & 9.6 & 82.7 & 10.4 + -0.614 & -54.2 & 7.4 & 67.2 & 12.0 & 0.036 & 0.064 & 0.127 & 0.040 & -51.1 & 11.2 & 88.0 & 14.8 + -0.664 & -49.5 & 6.0 & 49.3 & 7.6 & 0.028 & 0.057 & 0.077 & 0.030 & -48.4 & 9.3 & 58.5 & 12.2 + -0.715 & -59.2 & 7.1 & 61.0 & 9.4 & 0.175 & 0.061 & 0.069 & 0.028 & -42.6 & 14.3 & 71.3 & 14.1 + -0.816 & -50.8 & 6.8 & 55.6 & 9.5 & 0.042 & 0.055 & 0.113 & 0.029 & -49.6 & 10.5 & 71.0 & 13.8 + -0.918 & -43.3 & 6.9 & 58.9 & 10.1 & 0.050 & 0.080 & 0.101 & 0.031 & -36.2 & 10.3 & 73.5 & 15.2 + -1.019 & -50.7 & 5.6 & 60.6 & 6.8 & -0.004 & 0.065 & 0.076 & 0.041 & -50.7 & 8.2 & 72.0 & 9.1 + -1.121 & -56.2 & 6.7 & 53.3 & 8.1 & 0.080 & 0.073 & 0.066 & 0.027 & -49.7 & 14.2 & 61.9 & 12.6 + -1.252 & -58.6 & 7.3 & 60.2 & 8.5 & 0.092 & 0.072 & 0.066 & 0.022 & -47.2 & 10.8 & 69.9 & 12.4 + -1.425 & -55.7 & 9.5 & 88.1 & 18.1 & 0.035 & 0.077 & 0.182 & 0.061 & -48.3 & 17.0 & 127.4 & 21.6 + -1.658 & -64.8 & 6.4 & 51.2 & 7.4 & 0.037 & 0.058 & 0.056 & 0.019 & -60.5 & 9.0 & 58.2 & 9.4 + -1.932 & -56.9 & 8.1 & 58.0 & 8.1 & 0.009 & 0.094 & 0.076 & 0.041 & -54.9 & 9.8 & 68.8 & 9.9 + -2.257 & -75.5 & 9.5 & 67.2 & 13.5 & 0.164 & 0.096 & 0.084 & 0.041 & -43.6 & 17.1 & 81.1 & 20.4 + -2.642 & -57.6 & 10.3 & 62.1 & 11.9 & 0.037 & 0.105 & 0.068 & 0.036 & -47.4 & 15.9 & 72.5 & 16.9 + -3.078 & -56.0 & 9.9 & 53.8 & 11.1 & 0.268 & 0.106 & 0.088 & 0.043 & -28.7 & 17.7 & 65.4 & 15.9 + -3.555 & -41.6 & 27.4 & 75.6 & 50.9 & 0.070 & 0.283 & 0.404 & 0.302 & -28.8 & 34.7 & 150.3 & 30.9 + -4.194 & -9.2 & 15.6 & 56.4 & 14.9 & -0.144 & 0.105 & 0.021 & 0.037 & -22.5 & 34.1 & 59.3 & 20.1 + -4.569 & -39.7 & 10.8 & 45.3 & 9.2 & 0.240 & 0.136 & 0.042 & 0.036 & -19.7 & 17.4 & 49.9 & 13.2 + top : stis spectra of m32 at five positions along the slit .
solid curves are convolutions of the mpl - derived broadening functions @xmath38 ( fig . [ fig_broad ] ) with the stellar template .
bottom : spectrum of hr7615 , a k0iii giant , the template spectrum .
the vertical scale of the template spectrum is compressed with respect to that of the m32 spectra . broadening functions recovered from the central spectrum of m32 using the two spectral deconvolution algorithms discussed in the text .
the degree of smoothing increases downward . left column : fcq .
( a ) @xmath249 ; ( b ) @xmath250 ; ( c ) @xmath251 ; ( d ) @xmath252 .
right column : mpl .
( e ) @xmath253 ; ( f ) @xmath254 ; ( g ) @xmath255 ; ( h ) @xmath256 .
the mpl estimates tend toward a gaussian for large @xmath43 while the fcq estimates become increasingly distorted as the smoothing is increased .
this is the source of the greater bias in the fcq estimates ( although in practice smoothing parameters as small as @xmath252 would never be used ) . the mean value of @xmath18 computed by the fcq algorithm between @xmath257 and @xmath258 , as a function of smoothing parameter @xmath41 .
large values of @xmath41 correspond to small degrees of smoothing and hence to less biased estimates .
these @xmath18 values are based on data that were heavily binned in radius in order to increase the s / n as much as possible . line - of - sight velocity distributions derived from the stis m32 spectra using the mpl deconvolution algorithm .
note the sudden increase in the width of the broadening functions inside of @xmath166 .
the losvds are roughly antisymmetric about the center of m32 , as expected for a relaxed system ; the right column shows @xmath38 averaged over the left and right sides , @xmath259 $ ] .
the central losvd exhibits strong non - gaussian wings , a likely consequence of high - velocity stars near the central black hole .
the broadening functions at larger radii exhibit asymmetries suggestive of a second kinematic subcomponent which rotates with a velocity closer to the systemic velocity of m32 .
stis rotation curve for m32 , derived from losvds obtained using the mpl spectral deconvolution algorithm .
upper panel : filled circles : @xmath48 , the parameter that measures the velocity shift of the gaussian function that multiplies the gauss - hermite series .
open circles : @xmath260 , an estimate of the true mean line - of - sight velocity .
lower panel : the gauss - hermite parameter @xmath53 that measures asymmetries in the losvds . the mean velocity is smaller than @xmath78 due to the nonzero value of @xmath53 , which in turn reflects asymmetries in the losvd s ( fig .
[ fig_broad ] ) .
stis velocity dispersion profile for m32 , derived from losvds obtained using the mpl spectral deconvolution algorithm .
upper panel : filled circles : @xmath14 , the parameter that measures the dispersion of the gaussian function that multiplies the gauss - hermite series .
open circles : @xmath261 , an estimate of the true rms line - of - sight velocity .
lower panel : the gauss - hermite parameter @xmath18 that measures the amplitude of symmetric non - gaussian distortions in the losvd .
the velocity dispersion is generally greater than @xmath14 due to the nonzero values of @xmath18 .
this difference is substantial in the inner @xmath166 due to the strongly non - gaussian wings of the central losvds ( fig .
[ fig_broad ] ) .
comparison of @xmath48 and @xmath14 derived from the m32 stis data ( filled circles ) with earlier ground - based determinations .
squares : wht measurements from van der marel et al .
( 1994a ) .
triangles : cfht measurements from bender et al .
( 1996 ) . reduced @xmath143 contours describing the fit of the axisymmetric models described in the text to the observed , mean square line - of - sight velocity at points within the inner arc second of m32 .
the contour spacing is 0.5 and the innermost contour is at 0.75 .
the plus symbol marks the best - fit model
. predicted kinematical profiles for three axisymmetric models with different black hole masses .
( a ) @xmath262 ; ( b ) @xmath160 ; ( c ) @xmath263 .
thin curves show the models as observed with infinite resolution ; heavy curves are the models after convolution with the stis psf ; open circles are the data points . for each @xmath19
, the mass to light ratio @xmath127 and rotational parameter @xmath155 have been adjusted to optimize the fit .
the central m32 losvd ( heavy line ) , symmetrized about @xmath75 , compared to the losvd s predicted by spherical nonrotating models with three black holes masses .
thin line : @xmath264 ; dashed line : @xmath179 ; thin line : @xmath265 . estimated gauss - hermite parameters @xmath266 and @xmath267 , derived by fitting the @xmath35 of equation ( a6 ) to the assumed form ( a5 ) , with @xmath190 .
filled circles indicate the input values of @xmath14 and @xmath18 ; these values are recovered only when the input @xmath35 has @xmath268 .
recovery of @xmath18 via fcq .
circles : @xmath269 .
triangles : @xmath270 .
diamonds : @xmath271 .
there is a significant negative bias in the recovered values of @xmath18 when the velocity dispersion is less than about 100 .
mise and isb of estimates of @xmath35 obtained from the two deconvolution algorithms .
the input losvd was a lorentzian ( equation b2 ) with @xmath273 and @xmath200 .
both mise and isb have been normalized as described in the text .
solid lines : mpl algorithm .
open circles : fcq algorithm , using a fixed smoothing parameter @xmath274 . filled circles : fcq algorithm , using the value @xmath216 that minimizes the mise of the estimated @xmath35 .
mse and bias in estimates of @xmath18 obtained from the two deconvolution algorithms .
the input @xmath35 was a lorentzian ( equation b2 ) with @xmath273 and @xmath200 .
solid lines : mpl algorithm .
open circles : fcq algorithm , with @xmath251 . filled circles : fcq algorithm , using the value @xmath216 that minimizes the mise of the estimate @xmath56 .
@xmath216 is plotted vs. s / n in the bottom panel . | we have obtained optical long - slit spectroscopy of the nucleus of m32 using the space telescope imaging spectrograph aboard the _ hubble space telescope_. the stellar rotation velocity and velocity dispersion , as well as the full line - of - sight velocity distribution ( losvd ) , were determined as a function of position along the slit using two independent spectral deconvolution algorithms .
we see three clear kinematical signatures of the nuclear black hole : a sudden upturn , at @xmath0 from the center , in the stellar velocity dispersions ; a flat or rising rotation curve into the center ; and strong , non - gaussian wings on the central losvd .
the central velocity dispersion is @xmath1 ( gaussian fit ) or @xmath2 ( corrected for the wings ) .
both the velocity dispersion spike and the shape of the central losvd are consistent with the presence of a supermassive compact object in m32 with a mass in the range @xmath3 .
these data are a significant improvement on previous stellar kinematical data , making m32 the first galaxy for which the imprint of the black hole s gravitation on the stellar velocities has been observed with a resolution comparable to that of gas - dynamical studies .
# 1#23.6pt keywords : galaxies : elliptical and lenticular galaxies : structure galaxies : nuclei stellar dynamics |
the mechanism called _ bulk - mediated surface diffusion _ typically arises at interfaces separating a liquid bulk phase and a second phase which may be either solid , liquid , or gaseous .
whenever the adsorbed species is soluble in the liquid bulk , adsorption - desorption processes occur continuously .
these processes generate surface displacement because desorbed molecules undergo fickian diffusion in the liquid s bulk , and are latter re - adsorbed elsewhere .
when this process is repeated many times , an effective diffusion results for the molecules on the surface .
the importance of bulk - surface exchange in relaxing homogeneous surface density perturbations is experimentally well established @xcite .
adsorption at the solid - liquid interfaces arises , for instance , in many biological contexts involving protein deposition @xcite , in solutions or melts of synthetic macromolecules @xcite , in colloidal dispersions @xcite , and in the manufacture of self - assembly mono- and multi - layers @xcite .
usually the studies performed in this type of systems are done within the framework of a master equation scheme @xcite , where the particle s motion through the bulk and the adsorption - desorption processes are markovian . in two recent papers @xcite
we have shown the most important features of these phenomena .
non markovian diffusion processes have been used in order to modelling a great diversity of phenomena some of the most relevant are : anomalous charge transport in amorphous materials @xcite , diffusion of particles with internal states @xcite , chromatography @xcite , dielectric relaxation due to defect diffusion @xcite .
we address one of the fundamental issues in surface physics : the dynamics of adsorbed molecules .
we study this problem when the particle s desorption is characterized by a non markovian process , while the particle s adsorption and its motion in the bulk are governed by a markovian dynamics .
we analyze the diffusion of particles in a semi - infinite cubic lattice , and focus on the effective diffusion process at the interface @xmath0 .
we calculate analytically the conditional probability to find the particle on the @xmath1 plane as well as the surface dispersion as functions of time , and test those results comparing with monte carlo simulations .
when non markovian processes are present generalize master equations are involved , these equations being characterized by a memory kernel " that is related univocally with a continuous time random walk ( ctrw ) scheme @xcite .
non markovian desorption has been introduced in many fields of natural science describing transport processes in chemical reactions @xcite , re - emission when surfaces contains deep traps " @xcite , capture and re - emission from surfaces that contain sites with many internal states such the ladder trapping model " @xcite , proteins with active sites deep inside its matrix @xcite , etc .
the main goal of this work is to get information about the influence of a non markovian desorption dynamics on the effective diffusion process at the interface @xmath1 , and in this way to develop some criteria for looking for non markovian desorption effects in experimental situations . for that purpose
we calculate analytically the temporal evolution of the variance ( @xmath2 ) and the conditional probability of being on the surface at time @xmath3 since the particle arrived there at @xmath4 ( @xmath5 ) , that we indicate by @xmath6 . when @xmath7 , the waiting time density for desorption defined in the ctrw scheme , has a * finite first moment * we have been able to establish the long time asymptotic behavior for @xmath2 and @xmath6 for * any * non - markovian desorption .
this behavior is the same as the markovian one and only depends on the first moment of @xmath7 .
we also analyzed the behavior associated to two different families of desorption waiting time densities .
for the first one we have observed two regions , called _ strong _ and _ weak adsorption _ regime ( to be defined later ) . in the strong adsorption limit
we observe a transient regime , where the temporal evolution is characterized by damped oscillations ( its frequency not depending on the degree of departure from markovian behavior ) , and a markovian asymptotic one .
the oscillatory effects disappear in the weak adsorption region .
for the second kind of desorption dynamics , a transient regime with a non - monotonic behavior for the slope of the variance emerges , the asymptotic behavior being again markovian like .
the organization of the paper is as follows . in the next section
we formally present the model , supported by a generalized master equation which describe the particle s dynamics through the bulk and the adsorption - desorption dynamics at the surface .
we devote section iii to present and discuss the time dependence of @xmath8 and @xmath6 for two families of specific memory kernels , and compare the analytical results with monte carlo simulations in the first case , and discuss the general characteristics in the second . in section
iv we study the asymptotic behavior of the aforementioned magnitudes , in the asymptotic long time limit . in the last section
we discuss the results and present some conclusions .
above equations for @xmath37 are similar to eq.(2 ) and eq.(5 ) in @xcite with @xmath38 replacing @xmath24 .
therefore , all results obtained in that paper remains valid for non - markovian dynamic when @xmath24 is replaced by @xmath38 in the laplace domain .
the expression for the laplace transform of the variance ( @xmath39 ) became @xmath40 = \langle r^2(s)\rangle _ { plane}=\frac{n(s)}{d(s)},\ ] ] with @xmath41 \\ \nonumber & & \left [ 2 \gamma^3-s^2(\sqrt{s ( 4 \gamma + s ) } - s ) - 3 \gamma^2 ( \sqrt{s ( 4 \gamma + s)}- 3 u)- 2 \gamma s ( 2 \sqrt{s ( 4 \gamma + s ) } - 3 s ) \right],\end{aligned}\ ] ] @xmath42 ^{2}.\end{aligned}\ ] ] the laplace transform of @xmath43 , the conditional probability of being in the surface , is @xmath44 = p(z=1,s)= & & \left[\gamma ( 2 \gamma + s - \sqrt{s
( 4 \gamma + s)}\right ] \\ \nonumber & & \left [ \gamma s ( 2 \gamma + s - \sqrt{s ( 4 \gamma + s ) } ) + k(s)(\gamma ( \sqrt{s ( 4 \gamma + s)}- 3 s ) \right . \\ \nonumber & & \left . + s ( \sqrt{s ( 4 \gamma + s)}-s))\right]^{-1}.\end{aligned}\ ] ] finally we want to point out that the relation between @xmath7 , the waiting time density for the desorption processes defined in the ctrw scheme , and the memory kernel of eq .
( [ c6mod1 ] ) @xcite in the laplace domain is given by @xmath45
here we show the results obtained analytically and compare them with monte carlo simulations . in all cases
we have fixed the parameters : @xmath46 .
the simulations results were obtained averaging over @xmath47 realizations . to describe the desorption dynamics from the surface we have chosen two families of waiting time densities ( @xmath7 ) .
the first of them was introduced by scher and lax @xcite to describe the frequency dependence of the electric conductivity in disordered solids when transport is due to impurity hopping .
it has been extensively exploited in modelling non markovian cases that emerge when an average over transition rates ( disorder ) is taking into account @xcite and @xcite .
the reason of its wide use are its versatile functional form and its simplicity which allows to take into account a controllable spread of transition rates @xcite .
when only one transition rate is present the markovian description is reobtained ( the memory kernel is a dirac delta function ) and when the spread is very wide , steps occur at fixed regular intervals of time ( see below ) .
the adopted function ( see also @xcite ) is @xmath48 where @xmath49 is a positive integer and @xmath50 is the gamma or factorial function .
it is worth remarking here two important facts about this family of functions .
first , as can be seen from the eq .
( [ pdf ] ) , there are two parameters which characterized the function .
the parameter @xmath49 , called _
markovianicity parameter _ , defines the degree of function s departure from the markovian behavior ( @xmath51 corresponds to the markovian case ; @xmath52 to the non - markovian case ) , while the parameter @xmath53 is the average desorption s rate " .
second , as shown in @xcite , the mean value of these waiting time densities is @xmath54 showing that the average desorption s time " does not depends on the parameter @xmath49 .
for the form of this family of functions , see fig . 1 in @xcite . in figs .
[ fig2 ] and [ fig3 ] we present the temporal evolution for the conditional probability ( @xmath6 ) for the markovian and two non markovian cases ( two different values of @xmath49 ) with the same value of @xmath53 . from these figures
we can observe two temporal regions : a transient one which ranges from @xmath55 to @xmath56 and an asymptotic region ( @xmath57 ) where the behavior approaches that of the markovian case .
it is important to remark the existence of damped oscillations in the transient region for the non markovian case .
this oscillatory behavior is due to the waiting time density functions used as will be explained later . .
[ fig2 ] . [ fig3 ] in figs .
[ fig4 ] and [ fig5 ] we depict the results for the variance @xmath58 , for both the markovian and the two different non markovian cases ( @xmath59 and @xmath60 ) .
we can observe two important features .
the first one is that the variance in the non markovian regime shows a delay in the beginning of the spreading .
the second feature is that the system presents an oscillatory - like behavior , with the oscillations attenuating as the time grows .
this fact is still more apparent in the insert of the figures where we show the temporal evolution of the _ spreading velocity _ ( @xmath61 ) .
it is worth remarking here that oscillations only appears in the non markovian case with @xmath62 and are due to the particular behavior of the family of waiting time density defined by eq .
( [ pdf ] ) .
as it is known , this function goes to a dirac delta function as the markovianicity parameter @xmath63 ( @xmath64 ) , implying a periodic - like behavior . this periodicity " explains the oscillatory behavior of @xmath6 and @xmath58 .
also the figures show an excellent agreement between the theoretical and monte carlo simulation results .
we remark here that we can distinguish two region defined by the ratio @xmath65 .
the _ strong adsorption region _ characterized by @xmath66 where the non markovian character shows important differences respect to the markovian case in the transient temporal regime and the _ weak adsorption region _ ( @xmath67 ) where these differences disappear .
another worth remarking point from these figures is that @xmath68 , the frequency in the oscillation , remains unperturbed due to the fact that all desorption waiting time densities used have the same average desorption s time " .
this aspect becomes apparent in fig .
[ fig6 ] .
the second class of desorption dynamics that we consider in this work corresponds to the so called direct - access multi - trapping process " @xcite .
such an approach was proposed in order to describe in a phenomenological manner the transport of excited charge carriers across amorphous material @xcite , or the motion of intersticial in metals @xcite , etc .
when the walker arrives at a surface site this model considers transitions to and from @xmath69 internal states , both with probability @xmath70 per unit time , before it desorbs to the bulk with probability @xmath24 .
the resulting desorption probability waiting time density in laplace domain is @xmath71 and the average desorption s rate " in this case is @xmath72 . for a detailed analysis of this model and derivation of eq .
( [ diacc ] ) see ref .
@xcite . in figs .
[ fig6b ] and [ fig7b ] we depict the results for the variance @xmath58 for two different non markovian cases ( @xmath73 and @xmath74 respectively ) .
we also show the markovian behavior with the same average time for each case ( @xmath75 for @xmath73 and @xmath76 for @xmath74 ) .
we can observe that the temporal evolution of the variance in the non markovian case shows a transient regime characterized by the slope s non - monotonic behavior , which increases with @xmath77 .
this fact is still more apparent in the insert of the figures where we show the temporal evolution of the _ spreading velocity _
( @xmath61 ) .
the asymptotic behavior for large @xmath3 of @xmath78 and @xmath6 can be obtained analyzing the limit @xmath79 .
we assume that @xmath7 , the waiting time density for desorption , has a * finite first moment * @xmath80 which means that @xmath81 when @xmath82 and consequently @xmath83 in this limit . from eqs .
( [ variance4 ] ) , ( [ num ] ) , ( [ den ] ) and ( [ laplacep ] ) we obtain @xmath84 @xmath85 exploiting tauberian theorems @xcite , the behavior for large @xmath3 can be obtained resulting in @xmath86 } } \,\,\,\,\ t^{\frac{1}{2}},\ ] ] @xmath87 } \,\,\,\,\ t^{-\frac{1}{2}}.\ ] ] equations ( [ asimprt ] ) and ( [ asimppt ] ) represent the behavior of @xmath58 and @xmath6 for * any * non - markovian desorption for large values of time .
this behavior is the same as the markovian one and only depends on the first moment of @xmath7 .
it is important to remark that the only assumption was that @xmath7 , the waiting time density for desorption , has a finite first moment @xmath80 .
we have studied here the evolution of particles diffusing on a surface .
the diffusion have been performed across the bulk surrounding the surface , this phenomenon being called _ bulk mediated surface diffusion_@xcite .
usually the proposed models are based on markovian desorption processes .
the main feature of this work was to present an analytical model for non markovian desorption from the surface .
the bulk that surrounds this surface was considered to be semi - infinite and that the particles undergo a markovian motion on them .
we observed the influence of the non markovian desorption dynamic on the effective diffusion process at the interface @xmath1 by calculating analytically , in the laplace domain , the temporal evolutions of the variance ( @xmath58 ) and the conditional probability of being on the surface at time @xmath3 since the particle arrived there at @xmath4 ( @xmath6 ) .
when the waiting time density for desorption has a * finite first moment * we have established the behavior of the above magnitudes for * any * non - markovian desorption for large values of time .
this behavior is the same as the markovian one and only depends on the first moment of @xmath7 .
we have chosen two families of non - markovian desorption waiting time densities and tested the analytical results for @xmath2 and @xmath6 by comparison with monte carlo simulations , obtaining an excellent agreement .
for the first desorption waiting time density we can establish two regions based on the ratio @xmath65 , calling them _ strong adsorption _
( this occurs when @xmath88 ) and _ weak adsorption _
( when @xmath89 ) . in the strong adsorption limit we can observe a transient regime and a stationary one .
the main feature in the transient regime is that the temporal evolution is characterized by damped oscillations whose frequencies are in direct relation to @xmath53 , the average desorption rate .
it is worth remarking here that the frequency does not depend on the markovianicity parameter .
the effect of this parameter appears on the amplitude of the oscillations which disappear in the weak adsorption region . for the second kind of desorption dynamics we found a transient regime with an emerging non - monotonic behavior for the slope s variance . in the asymptotic regime
, the conditional probability of the non markovian system tends to the markovian one as is expected from the analytical results obtained in section iv . in a recent paper @xcite we have shown that , when we consider * finite * or * infinite biased * systems , and desorption waiting time densities with finite first moment , the variance growths linearly with time in the asymptotic regime .
finally , it is worth here remarking an important aspect of the present approach .
we have shown through the above results that the behavior of @xmath58 and @xmath6 results to be strongly dependent on the desorption mechanism .
as the effective dispersion and the percentage of particles that remain on the plane @xmath1 are measurable magnitudes @xcite they could be used to investigate the characteristic and get information about the fundamental parameters of the desorption processes . | here we address a fundamental issue in surface physics : the dynamics of adsorbed molecules .
we study this problem when the particle s desorption is characterized by a non markovian process , while the particle s adsorption and its motion in the bulk are governed by a markovian dynamics .
we study the diffusion of particles in a semi - infinite cubic lattice , and focus on the effective diffusion process at the interface @xmath0 .
we calculate analytically the conditional probability to find the particle on the @xmath1 plane as well as the surface dispersion as functions of time .
the comparison of these results with monte carlo simulations show an excellent agreement . |
the kuiper belt has a complex orbital structure and can be divided in multiple sub - populations ( see gladman et al . , 2008 for a review ) . among them
are the cold and the hot populations , which are defined as the collections of objects inwards of the 1/2 resonance with neptune ( @xmath1 au ) with , respectively , inclinations smaller or larger than 4 degrees .
the cold and hot populations have also distinct physical properties ( see morbidelli and brown , 2004 for a review ) .
there is a quite general consensus that the hot population formed originally closer to the sun , was dynamically excited by the perturbations from the giant planets and finally was transported into the kuiper belt ( gomes , 2003 ; levison et al . , 2008 ) .
however , there is no consensus on the origin of the cold population .
some models argue that it also was transported into the kuiper belt from a region closer to the sun ( levison and morbidelli , 2003 ; levison et al . , 2008 ) , while others argue that the cold population formed locally ( e.g. parker et al . , 2011 ; batygin et al . , 2011 ) .
an important point in this debate was made by dawson and murray - clay ( 2012 ) .
first they observed that the usual partition of the cold and hot populations according to the 4-degree inclination boundary is simplistic ; in reality these populations have distinct , but partially overlapping inclination distributions ( see brown , 2001 ) .
thus , to limit the contamination of the cold population by the hot population , they restricted their analysis to objects with inclination @xmath2 , where the relative fraction of low - inclination `` hot '' objects is expected to be negligible .
then they showed that , inside of 43.5 au , this low - inclination population has also small eccentricities ( @xmath3 ) , even though orbits would be stable up to @xmath4 ( lykawka and mukai , 2005a ) .
the hot population , in fact , has eccentricities up to this limit .
the lack of moderate eccentricity orbits in the cold population obviously can not be explained by observational biases .
dawson and murray - clay therefore interpreted this result as evidence that the cold kuiper belt was only very moderately excited relative to its original quasi - circular and coplanar orbits .
this argues against models in which the cold population originates closer to the sun and is implanted into the kuiper belt , because such models predict a cold population with an eccentricity distributions covering the whole stability range .
given the importance of this argument , we have decided to revisit the problem of the eccentricity distribution of the cold kuiper belt . in section 2 we redo the same analysis as dawson and murray - clay , but using proper elements instead of osculating elements .
our results confirm theirs , but we notice that the transition between the inner part of the cold population , where eccentricities are all small , to the outer part , where the eccentricities cover a wider range , happens exactly at the 4/7 mean motion resonance with neptune .
this suggests that this resonance might have played a role in sculpting the inner cold belt during a phase of outward migration .
then , in section 3 we conduct migration experiments , testing different migration timescales and eccentricities of neptune .
section 4 analyzes more in details how moderate - eccentricity cold kuiper belt objects are removed by resonance sweeping and compares their evolution with that of high inclination bodies .
our conclusions are discussed in section 5 .
we have selected all tnos from the mpc catalog with semi major axis larger than 25 au and orbits determined from observations covering at least three oppositions . such procedure selected a set of 811 tnos .
for each of these objects we computed numerically the orbital proper elements using integrations covering 132 my .
the proper semi major axis was computed by numerical average of the values recorded during the simulation , with an output time - step of 1,000y . for the proper eccentricities and inclinations , the computational procedure was more elaborated , although standard ( knezevic and milani , 2000 ) .
we first computed the fourier series of @xmath5 and @xmath6 , where : @xmath7\cr k(t))&=&e(t)\sin[\varpi(t)]\cr p(t)&=&i(t)\cos[\omega(t)]\cr q(t))&=&i(t)\sin[\omega(t)]\end{aligned}\ ] ] and @xmath8 and @xmath9 are the values of eccentricity , inclination , longitude of perihelion and longitude of node recorded over time @xmath10 .
we then removed from the series expansions the terms with frequencies close ( i.e. within one arcsec / y ) to the proper frequencies of the planets . finally , we selected as proper eccentricity and inclination the coefficients of the largest remaining term in the each of the two fourier series . in order to evaluate the accuracy of proper eccentricity and inclination
, we performed the procedure described above for the first and the second halves of the whole integration , i.e. , 65.536my .
then , we adopted as an estimate of the error , the largest difference among the proper elements calculated for the whole integration time - span and those computed in each of the two half time - spans .
the relative accuracy in proper semi major axis was always better than @xmath11 .
the absolute accuracies in proper eccentricity and inclinations were better than 0.01 and 0.1 degrees throughout the region of interest ( i.e. inside of 44 au and not in the 4/7 mean motion resonance with neptune ) .
once in possession of this proper element catalog , following dawson and murray - clay we retained as members of an `` uncontaminated cold population '' the objects with proper inclination smaller than 2 degrees .
[ helton ] shows the distribution of the selected objects ( dots ) in proper semi major axis vs. eccentricity plane .
we also computed a new stability map .
in fact , the map used by dawson and murray - clay , from lykawka and mukai ( 2005a ) , was computed for a wide range of inclinations , whereas here we are interested to very low inclinations only .
we could have used the stability map in duncan et al .
( 1995 ) , which was computed for @xmath12 , but the latter was quite sparse , due to the computing limitations of the time
. moreover , both lykawka and mukai and duncan et al . reported their maps relative to the initial osculating elements . here ,
for a consistent comparison with the proper elements of the real objects , we needed a map computed in proper elements space . to compute the stability map
, we proceeded as follows .
we adopted a grid of particles initial conditions , with osculating elements in the following ranges : 42 au @xmath13 48 au , @xmath14 and @xmath15 deg , with resolutions of 0.2 au in @xmath16 , 0.01 in @xmath17 and 0.5 degrees in @xmath18 .
the secular angles @xmath19 and @xmath20 were set equal to 0 degrees .
each particle was integrated for 132 my .
we then computed their proper elements following the same procedure described above .
finally , we continued the simulations for 4 gy in order to asses the long - term survival of the test particles . to construct fig .
[ helton ] , for each given pair of initial @xmath16 and @xmath17 , we selected the particle with the smallest value of proper inclination . then , for the particles that survived in the 4 gy integration ( i.e. they did not encounter neptune within a hill radius within this time ) , we plotted on a white background a light - gray square of size 0.2 au@xmath21 centered on their values of proper semi major axis and eccentricity measured on the first 132 my .
moreover , particles that did not survive were denoted by dark - gray squares centered on the initial pair of osculating semi major axis and eccentricity . the stability map of fig . [ helton
] shows few surprises . in general , particles are unstable at large eccentricity and stable at low eccentricity , where the perihelion distance is larger than @xmath2240 au .
mean motion resonances represent the exception to this general rule .
the light - gray squares at large eccentricity are all associated to mean motion resonances , as well as the vertical white columns at low or moderate eccentricities .
some mean motion resonances , therefore , clearly stand out from the stability map , and they are labelled on the figure .
in general , as expected , the dots fall on light - gray squares .
those that do nt , are associated to mean motion resonances .
in fact , in a mean motion resonance there is a third dimension characterizing the orbit : the resonant amplitude .
it is well possible that none of the test particles that we used for the stability map sampled the orbit of a real particle because their libration amplitude is different . in this case ,
a dot is plotted over the white background . overall , fig .
[ helton ] confirms the results of dawson and murray - clay . inside of 43.5 au ,
all real objects have small eccentricities , barely exceeding 0.05 .
the stability map , however , ranges up to 0.1 in the 42.2 - 42.6 au region .
thus , there is clearly a stable region ( approximately in the range @xmath23 ) in the inner belt that is not inhabited by the cold population . at a closer inspection
, one sees that the transition in eccentricity distribution of the cold belt is sharply at the location of the 4/7 resonance ( see lykawka and mukai , 2005b for a description of this resonance ) .
this result suggests that the 4/7 resonance might have played a role in depleting the moderate - eccentricity cold objects inside of its current location , as it swept through the 42.543.5 au region during the putative radial migration of neptune .
thus , in the following section , we report on numerical experiments of radial migration that address whether this is indeed possible and at which conditions .
we set up simple numerical experiments , where jupiter , saturn and uranus were assumed to be on their current orbits , while neptune s mean semi major axis was forced to change from @xmath24 au to its current average value @xmath25au as @xmath26 .
the eccentricity of neptune was also damped from an initial value @xmath27 as @xmath28 with an appropriate coefficient @xmath29 that allowed us to match the current eccentricity of neptune at the end of the simulation .
each simulation was run for a time - span of @xmath30 .
the semi major axis drift and the eccentricity damping were implemented by applying synthetic forces to neptune s equations of motion . in particular , we employed the forces described in malhotra ( 1995 ) for evolving the semi - major axes and those in kominami et al .
( 2005 ) for controlling the eccentricity . my and @xmath31 .
the top panel depicts the time evolution of the eccentricity and the bottom panel that of the semi major axis .
the green area close to the @xmath32 axis is obtained by plotting the evolution of neptune in the current solar system , and represents the target @xmath16 and @xmath17 that a good migration simulation needs to hit . ]
[ neptune ] shows the evolution of neptune in a simulation with @xmath33 my and @xmath31 .
notice that , although the _ initial _ eccentricity is 0.1 , the _
mean _ eccentricity at the beginning of the simulation is only 0.075 .
the green area at the right - hand - side of each panel shows the current range of oscillations of the semi major axis and eccentricity of neptune in the real solar system .
this shows that our migration evolution reproduces the actual orbit of neptune with a good accuracy .
overall we did 9 simulations , corresponding to the values of @xmath34 and @xmath27 reported in table [ table ] .
these values cover a range of proposed migration models , with and without a putative initial eccentricity excitation of neptune s orbit .
we will discuss in sect .
[ conclusions ] what the successful parameters imply for the history of the solar system . in each simulation
we considered a population of 1,000 test particles , initially spread in osculating elements between 42.3 and 44.2 au ( in some cases only up to 43.6 au ) in semi major axis , from 0 to 0.15 in eccentricity and up to 2 degrees in inclination ( with a uniform distribution in @xmath35 ) .
particles got discarded when they had encounters with neptune . for the surviving particles we computed the proper elements over the final part of the simulation . only particles with final proper inclination smaller than 2 degrees were considered .
my and @xmath31 .
the red dots depict the simulated particles and the blue dots the real cold belt objects ( the same as shown in fig .
[ helton ] ) .
the column of objects at @xmath36 au marks the location of the 4/7 resonance with neptune .
the small black dots depict the initial conditions of the simulated particles .
horizontal lines at @xmath37 and @xmath38 are plotted for reference . ]
[ results ] shows the final particle distribution ( red dots ) obtained in the simulation with @xmath33 my and @xmath31 .
their final proper semi major axes have been multiplied by a factor 1.00092 , to compensate for the slight offset of the final semi major axis of neptune relative to the real one , visible in fig .
[ neptune ] . for comparison ,
the blue dots represent the real low - inclination cold population ( the same as in fig .
[ helton ] ) and the small black dots show the initial conditions . first ,
notice the cluster of simulated particles in the 4/7 resonance at @xmath39 .
this cluster is visible in the observations as well , in the right proportion .
in fact , the ratio between the number of resonant particles and that of particles with @xmath40 au is 26.3% at the end of the simulation ; among observed objects , this ratio is 25.6% .
obviously , one has to take into account that observational biases may act differently on resonant and non - resonant objects , so that this agreement may just be accidental .
however , we argue that in this case the differential bias effect is probably not a big issue .
in fact , both resonant and non - resonant objects considered here have small inclinations and eccentricities ; moreover objects in the 4/7 resonance can reach perihelion at 4 different position in the sky relative to neptune , so that it is unlikely that all these sweet - spots have been missed by the surveys .
second , notice that inside of the final location of the 4/7 resonance , most of the particles with moderate eccentricities have been removed .
thus , the final distribution is strongly skewed towards small eccentricities , more or less similar to the observed distribution .
.a summary of the results of the statistical tests for 9 simulations .
the first line reports the value of @xmath34 and the first column that of @xmath27 ( in parenthesis the initial mean value for @xmath17 ) .
each case of the matrix reports @xmath41 , where @xmath42 is the probability that criterion 1 is fulfilled , @xmath43 is the probability for criterion 2 and @xmath44 is the probability that both criteria are fulfilled simultaneously .
see text for definition of criteria . [
cols="^,^,^,^",options="header " , ] [ table ] notice also that , beyond 43.8 au , a similar truncation in the particle distribution occurs , but at proper @xmath37 .
remember that fig .
[ helton ] shows that , in the current solar system , orbits in this region would be stable up to @xmath45 .
the truncation at @xmath37 is also visible in the observed distribution and in our simulation is operated by the 5/9 resonance sweeping .
resonant sweeping , however , can not explain the `` wedge '' , namely the paucity of low eccentricity objects ( @xmath46 ) beyond 43.8 au ; therefore , some other explanation is needed ( e.g. batygin et al . , 2011 ) for this structure . in order to quantify
how well the simulated distribution reproduces the observed distribution we proceeded as follows .
we considered objects and test particles in the 42.443.6 range only .
the observed distribution in this range is made of 43 objects .
the object with the largest proper eccentricity has @xmath47 so it is included here ] . in total , there are only 7 objects with proper @xmath48 .
the simulated distribution has 194 test particles . in this count
, we discarded the two particles with proper eccentricity larger than 0.1 because , from the stability map in fig .
[ helton ] , we know that these particles are unstable on the long term .
we then did a monte carlo simulation . from the simulated distribution
, we generated 10,000 synthetic populations , each of which contained 43 particles ( the same number as that of the objects in the observed population ) .
we considered two criteria of `` success '' .
the first was that a synthetic population contained no particles with proper eccentricity larger than 0.0653 ( criterion 1 ) ; the second was that the population contained no more than 7 particles with proper @xmath48 ( criterion 2 ) .
we found that 18% of the synthetic populations fulfilled criterion 1 ; 80% of them fulfilled criterion 2 and 17% fulfilled simultaneously criterion 1 and 2 .
from this test , therefore , we conclude that the simulated population is statistically equivalent to the observed population ( in the sense that it can not be rejected even at a @xmath49 level ) .
thus , we conclude that the slow migration of the 4/7 resonance with an initial moderate eccentricity of neptune can explain the properties of the inner part of the cold population of the kuiper belt .
the results of the same statistical tests for the other simulations are reported in table [ table ] .
as one can appreciate , fast migrations with @xmath50 or 30 my do a poor job in reproducing the observations .
the simulation with @xmath51 and @xmath52 does a decent job if one considers criterion 2 , but it is consistent with the observed population only at the @xmath53 level if one considers criterion 1 . if one considers both criteria simultaneously , the simulated population has only a 1% chance to match the observed population .
so , both the migration speed and the initial eccentricity of neptune are important ( see sect . [ comparison ] for an explanation of this result )
. however , if neptune is too eccentric ( e.g. in the simulations with @xmath54 ) the entire inner kuiper belt is destabilized , as described in levison et al .
( 2008 ) and almost no particles survive with proper @xmath2 in the considered semi major axis range . all these results do not depend strictly on the adopted @xmath2 limit ; they would basically be the same for any reasonable boundary in proper inclination .
our proposed explanation for the peculiar orbital structure of the cold population would not be acceptable without showing that the hot population avoids being sculpted in a similar fashion by the same process .
in fact , as shown by dawson and murray - clay ( 2012 ) , the hot population does not exhibit any apparent deficit of objects with @xmath48 . for this purpose ,
we run again our best - case simulation , that with @xmath33 my and @xmath27=0.1 , with a population of test particles having the following initial osculating orbital elements : @xmath55 , 42.3 au @xmath13 43.6 au , and @xmath56 .
the initial conditions and the proper elements of the surviving population are shown in fig .
[ hot ] .
but for test particles initially with @xmath40 au and @xmath57 and observed kbos with current inclination larger than 4 degrees .
notice that the depletion of test particles with @xmath48 is much less pronounced than for the low - inclination population . ]
it can be immediately appreciated that the depletion of particles with eccentricity above 0.05 is much less effective than for the cold population .
thus , the hot population should have preserved its original ( i.e. post - capture ) eccentricity distribution .
for this reason , and given the simple initial distribution ( uniform ) of our particles , the simulated final distribution is not intended to match the observations , but just to demonstrate that a moderate - eccentricity hot population could have survived the same migration scenario that explains the removal of the moderate - eccentricity cold population . the differences between fig .
[ results ] and [ hot ] is so striking that it calls for an explanation . to understand what happens ,
we looked at the individual evolution of particles in both simulations . a representative evolution of particles removed at low inclination is shown in figs .
[ lowi - evolve ] and that of particles surviving at high inclination is shown in fig .
[ highi - evolve ] .
each figure shows the behavior of semi major axis ( bottom panel ) and eccentricity ( top panel ) over time , while depicting also the evolving location of the 4/7 and 5/9 resonances with neptune .
notice that both particles have initially comparable values of semi major axis and eccentricity .
the low inclination particle has @xmath58 and the high inclination particle has @xmath59 . and
@xmath60 , namely they depict the locations of the 5/9 and 4/7 resonances . ]
[ lowi - evolve ] shows that the low - inclination particle was captured in mean motion resonance with neptune twice .
first , the particle was captured in the 5/9 resonance , between 60 and 100 my .
this is clear from the drift in particle s semi major axis along the resonant track
. then the particle was released and it was captured in the 4/7 resonance at @xmath61 my .
this was a brief capture episode , but enough to raise the the eccentricity up to 0.2 .
thus , the particle was destabilized : it was scattered by neptune until it was dynamically removed .
this shows that the cold population with moderate eccentricities was not lifted to into the hot population , but rather it was removed into the scattered disk .
but for a high inclination particle .
] fig .
[ highi - evolve ] , shows a completely different behavior : the high - inclination particle crossed both resonances without being captured .
the semi major axis and the eccentricity show a distinctive jump each time that a resonance was crossed , but the eccentricity was not significantly affected overall .
thus , the particle remained stable till the end of the simulation .
why is the probability of capture into resonance so different for low inclination and high inclination particles ?
we argue that the explanation is in the resonant structure . at small inclination
, a mean motion resonance can be approximated by an integrable single - harmonic hamiltonian .
for the 4/7 resonance the harmonic term is @xmath62 where @xmath63 and @xmath64 are the mean longitudes of the particle and of neptune and @xmath20 is the perihelion longitude of the particle .
similarly , for the 9/5 resonance the harmonic term is @xmath65 .
but at high inclination there are additional major resonant harmonics . for the 4/7 resonance
the second major harmonic is @xmath66 , where @xmath19 is the particle s longitude of node . for the 5/9 resonance one
has two additional harmonics : @xmath67 and @xmath68 .
as explained in chapter 9 of morbidelli ( 2002 ) , a resonance described by multiple harmonics differing for the combinations of secular angles is analog to a _ modulated pendulum _ , which exhibits a wide chaotic band around the central resonant island .
the probability of capture into resonance is very different in the integrable approximation and in the modulated pendulum approximation ( henrard and henrard , 1991 ; morbidelli and henrard , 1993 ) . in the case of the modulated pendulum ,
if the drift of the resonance is fast relative to the diffusion timescale inside the chaotic band , the capture into resonance is still possible ( indeed we see in fig .
[ hot ] particles aligned at the 4/7 resonance at @xmath69 au ) , but much more unlikely than in the integrable approximation . from this analysis
we can also understand why the results for the cold population depend on the parameters @xmath34 and @xmath27 as reported in table 1 .
if temporary resonant capture is the key , it is obvious that a faster migration speed ( i.e. smaller @xmath34 ) makes less likely that particles are trapped in resonance . in the case of fast migration
, most low - i particles just jump across resonance , with an evolution similar to that shown in fig .
[ highi - evolve ] .
the dependence on eccentricity is more subtle .
a larger neptune s eccentricity makes the resonances more effective in exciting eccentricities through secular effects .
in fact , if the eccentricity of the planet were null and the planet were not migrating , mean motion resonances would only force an eccentricity oscillation coupled with the resonant libration ( see morbidelli , 2002 - chapter 9 ) . in the case of a migrating circular planet , particles -once trapped in resonance- would have their eccentricity increased monotonically as they move outward ( e.g. see malhotra , 1995 for an illustration concerning the 2/3 resonance ) .
but in our case , this outward motion is short - ranged ( less than @xmath70 au ) , so this effect would not be very dramatic .
the large and fast eccentricity increase observed in fig .
[ lowi - evolve ] at @xmath71 my is possible only because the eccentricity of neptune is not null .
it should be noticed in fig .
[ neptune ] that the eccentricity of the planet is not damped to zero in our simulations , so that the final orbit of the planet is similar to the current one .
this implies that , in simulations starting with different initial eccentricities @xmath27 , eventually the eccentricity of the planet is the same .
more precisely , in the simulations with @xmath52 and @xmath31 the eccentricities of neptune became basically indistinguishable after 150my .
thus , why are the resulting cold belts different , as suggested by the the data reported in table 1 ? .
but for the simulation with initial neptune s eccentricity @xmath52 ( and still @xmath72100 my ) .
notice that in this case the particle population initially extends only up to 43.6 au . ] in the time range up to 150 my , the 4/7 resonance has swept the belt up to 43 au .
the inspection of the final distribution of the particles in the simulation with @xmath52 ( fig .
[ ae005 ] ) shows that the difference with the @xmath31 case is indeed mostly for particles with @xmath73 au .
specifically , the @xmath52 simulations leaves several particles there with @xmath48 . whereas the @xmath31 simulation removed almost all of them .
this difference is enough to reduce drastically the probability to fit the observed distribution , as reported in table 1 .
this explanation of the importance of the eccentricity of neptune for the sculpting of the cold population has an interesting implication .
it argues that the ( already excellent ) fit to observations obtained in the simulation with @xmath33 my and @xmath74 could improve if the eccentricity of neptune were damped more slowly than we originally assumed .
in fact , the number of particles surviving with @xmath48 and @xmath16 in the 43.043.5 au range would presumably be reduced if neptune s eccentricity remained somewhat larger until the 4/7 resonance reached 43.5 au .
the origin of the cold population of the kuiper belt is still elusive .
a debate is open on whether this population formed in - situ or was transported into the kuiper belt from a location closer to the sun during the primordial `` wild '' evolution of the giant planet orbits .
dawson and murray - clay ( 2012 ) pointed out a property of the cold population that had passed previously unnoticed : the inner part of this population ( the one with @xmath0 au ) has eccentricities smaller than @xmath75 , despite orbits up to @xmath76 could be stable in this region .
they interpreted this as an evidence that the cold population was only moderately excited from its original circular orbits , which argues in favor of its in - situ formation . in this work
we have confirmed the analysis of dawson and murray clay using proper elements .
however , we showed that a slow migration of neptune ( on a timescale of 100 my ) , initially on a moderately eccentric orbit ( @xmath77 ) , can reproduce the observations starting from a population uniformly distributed in eccentricity .
therefore , we disagree that the eccentricity distribution of the inner cold belt can be used as an argument for minor excitation and in - situ formation . obviously , however , our results do not imply that the cold population formed elsewhere and was transported into the kuiper belt .
it s origin , therefore , remains elusive . on this subject , notice from fig .
[ helton ] that the proper eccentricity distribution of the cold population beyond the 4/7 resonance extends up to @xmath78 .
it would be strange if the outer population had been excited more than the inner population .
this suggests that the _ entire _ cold population got strongly excited , so to cover the entire stability region ( and probably going beyond it ) , but it was then depleted near the stability border by the last bit of resonance migration .
we also notice in fig .
[ helton ] that beyond the 4/7 resonance there is a deficit of low - eccentricity ( i.e. proper @xmath46 ) cold objects , known as the `` wedge '' ( batygin et al . , 2011 ) . if this deficit is not due to observational biases , and it can not be explained the sweeping of the 5/9 resonance .
thus , it is an important diagnostic feature for excitation / implantation models of the cold population .
our mechanism for the depletion of the inner cold population with moderate eccentricities by resonance sweeping requires a migration timescale of @xmath79 my .
this timescale is reasonable for a tail - end of the planet migration process , and it is typically observed in simulations where neptune , after a wild phase of evolution due to close encounters with the other planets , settles down in the planetesimal disk ( gomes et al . , 2005 ; nesvorny and morbidelli , 2012 ) . the fact that our mechanism requires that the orbit of neptune had some moderate eccentricity when its semi major axis was at 28.5 au argues that the planet underwent previously some form of dynamical instability .
the reason is that planetesimal - driven migration can only damp , not excite , the planet s eccentricity and therefore , without an instability phase , the eccentricity of neptune would have always been small . in principle , the eccentricity of neptune at 28.5 au could be the remnant of a much larger eccentricity acquired when the planet was closer to the sun and had close encounters with the other planets .
therefore , the results of this paper are not inconsistent with the scenario of levison et al .
( 2008 ) on the origin of the kuiper belt , provided that the tail - end of neptune s migration is slow enough ( which was not the case in levison et al . , who adopted @xmath80 my throughout their simulation ) .
however , we acknowledge that said scenario seems to be inconsistent with the existence of wide binaries in the cold population ( parker et al . , 2012 )
. more recently , nesvorny and morbidelli ( 2012 ) explored the possibility that the outer solar system contained initially an extra neptune - mass planet , which eventually was ejected during the giant planet instability . in the simulations that reproduced the best the current orbits of the planets , neptune had an evolution much less wild than that considered in levison et al .
neptune migrated out quite smoothly and had only a moderate eccentricity excitation when it encountered the lost planet .
it will be important to investigate whether this kind of evolution can transport the cold population into the kuiper belt via the mechanism of levison and morbidelli ( 2003 ) .
this transport mechanism would in principle preserve wide binaries , satisfying the constraint discussed in parker et al .
( 2012 ) and would also be consistent with the results of this paper .
a.m. thanks german holmholtz alliance for funding this research through their `` plan etary evolution and life '' programme .
d.n . thanks nasa s opr program for supporting his work .
* batygin , k. , brown , m. e. , fraser , w. c. 2011 .
retention of a primordial cold classical kuiper belt in an instability - driven model of solar system formation .
the astrophysical journal 738 , 13 .
* brown , m. e. 2001 .
the inclination distribution of the kuiper belt .
the astronomical journal 121 , 2804 - 2814 . * dawson , r. i. , murray - clay , r. 2012 .
neptune s wild days : constraints from the eccentricity distribution of the classical kuiper belt . the astrophysical journal 750 , 43 . * duncan , m. j. , levison , h. f. , budd , s. m. 1995 .
the dynamical structure of the kuiper belt .
the astronomical journal 110 , 3073 . * gladman , b. , marsden , b. g. , vanlaerhoven , c. 2008 .
nomenclature in the outer solar system . the solar system beyond neptune 43 - 57 . * gomes , r. s. 2003 .
the origin of the kuiper belt high - inclination population .
icarus 161 , 404 - 418 .
* gomes , r. , levison , h. f. , tsiganis , k. , morbidelli , a. 2005 .
origin of the cataclysmic late heavy bombardment period of the terrestrial planets .
nature 435 , 466 - 469 .
* henrard , j. and henrard , m. 1991 .
slow crossing of a stochastic layer .
physica d 54 , 135 - 146 . *
knev zevi , z. , milani , a. 2000 .
synthetic proper elements for outer main belt asteroids . celestial mechanics and dynamical astronomy 78 , 17 - 46 .
* kominami , j. , tanaka , h. , ida , s. 2005 .
orbital evolution and accretion of protoplanets tidally interacting with a gas disk .
i. effects of interaction with planetesimals and other protoplanets .
icarus 178 , 540 - 552 . *
levison , h. f. , morbidelli , a. 2003 .
the formation of the kuiper belt by the outward transport of bodies during neptune s migration .
nature 426 , 419 - 421 . * levison , h. f. , morbidelli , a. , van laerhoven , c. , gomes , r. , tsiganis , k. 2008 .
origin of the structure of the kuiper belt during a dynamical instability in the orbits of uranus and neptune .
icarus 196 , 258 - 273 . * lykawka , p. s. , mukai , t. 2005a . long term dynamical evolution and classification of classical tno@xmath81 . earth moon and planets 97 , 107 - 126 .
* lykawka , p. s. , mukai , t. 2005b . exploring the 7:4 mean motion resonance i : dynamical evolution of classical transneptunian objects .
planetary and space science 53 , 1175 - 1187 . * malhotra , r. 1995 .
the origin of pluto s orbit : implications for the solar system beyond neptune .
the astronomical journal 110 , 420 .
* morbidelli a. and henrard , j. , 1993 .
slow crossing of a stochastic layer .
physica d. 68 , 187 - 200 .
* morbidelli a. , 2002 .
modern celestial mechanics .
taylor and francis * morbidelli , a. , brown , m. e. 2004 .
the kuiper belt and the primordial evolution of the solar system . comets ii 175 - 191 .
* nesvorn , d. , morbidelli , a. 2012
. statistical study of the early solar system s instability with four , five , and six giant planets . the astronomical journal 144 , 117 . * parker , a. h. , kavelaars , j. j. , petit , j .-
m . , jones , l. , gladman , b. , parker , j. 2011 .
characterization of seven ultra - wide trans - neptunian binaries . the astrophysical journal 743 , 1 . | dawson and murray - clay ( 2012 ) pointed out that the inner part of the cold population in the kuiper belt ( that with semi major axis @xmath0 au ) has orbital eccentricities significantly smaller than the limit imposed by stability constraints . here , we confirm their result by looking at the orbital distribution and stability properties in proper element space . we show that the observed distribution could have been produced by the slow sweeping of the 4/7 mean motion resonance with neptune that accompanied the end of neptune s migration process .
the orbital distribution of the hot kuiper belt is not significantly affected in this process , for the reasons discussed in the main text .
therefore , the peculiar eccentricity distribution of the inner cold population can not be unequivocally interpreted as evidence that the cold population formed in - situ and was only moderately excited in eccentricity ; it can simply be the signature of neptune s radial motion , starting from a moderately eccentric orbit .
we discuss how this agrees with a scenario of giant planet evolution following a dynamical instability and , possibly , with the radial transport of the cold population . |
* time resolved photoluminescence of a - exciton . * ( a ) laser pulse ( blue line ) and pl emission ( black line ) intensity at t = 4 k detected at maximum of a - exciton
pl e@xmath1 ev as a function of time .
inset : chiral optical selection rules in 1ml mos@xmath0 ( b ) t=4k , e@xmath2 ev , e@xmath1 ev . laser polarization @xmath3 . left axis : @xmath3 ( @xmath4 ) polarized pl emission intensity presented in black ( red ) as a function of time .
right axis : circular polarisation degree during exciton emission ( blue hollow squares : excitation power p@xmath5 , green full squares:0.01 p@xmath6 ) , errors bars take into account uncertainty in time origin @xmath7 .
( c ) same as ( b ) , but for t=300k , e@xmath8 ev , e@xmath9 ev.,scaledwidth=45.0% ] transition metal dichalcogenides such as mos@xmath0 emerge as an exciting class of atomically flat , two - dimensional materials for electronics @xcite , optics @xcite and optoelectronics @xcite .
in contrast to graphene , monolayer ( ml ) mos@xmath0 has a direct bandgap @xcite in the visible region of the optical spectrum .
inversion symmetry breaking ( usually absent in graphene ) together with the spin - orbit interaction leads to a unique coupling of carrier spin and k - space valley physics .
the circular polarization ( @xmath3 or @xmath4 ) of the absorbed or emitted photon can be directly associated with selective carrier excitation in one of the two non - equivalent k valleys ( @xmath10 or @xmath11 , respectively ) @xcite , where the role of strong excitonic effects merits further investigation in this context @xcite .
the chiral optical selection rules open up very exciting possibilities of manipulating carriers in valleys with contrasting berry phase curvatures @xcite , aiming for experimental manifestations of the predicted valley hall effect @xcite .
also stable spin states have been predicted for valence and conduction states @xcite for this material .
+ up to now optical valley initialisation in ml mos@xmath0 is based on the analysis of the large circular polarization degree @xmath12 of the emitted light from the direct bandgap observed in continuous wave ( cw ) measurements following circularly polarized laser excitation @xcite .
an important drawback seemed to be the drastic decrease of @xmath12 as the temperature is raised to 300k @xcite . in a simple approach , the stationary ( time integrated ) polarisation is determined by the initially created polarization @xmath13 , the lifetime of the electron - hole pair @xmath14 and the polarization decay time @xmath15 through @xmath16 @xcite .
we emphasize that the polarization decay time does not correspond directly to the carrier spin flip time as in most semiconductors like gaas @xcite , but it includes the scattering time between the two non - equivalent k valleys ( k@xmath17 or k@xmath18 ) @xcite .
+ in this letter we present the first time resolved polarization measurements in mos@xmath0 monolayers , providing vital information on the valley dynamics from 4k to room temperature .
we determine the key parameters that govern the stationary polarization degree @xmath12 : using quasi - resonant excitation of the a - exciton transitions , we can infer that the photoluminescence ( pl ) decays within @xmath194ps . for pulsed laser excitation
, we observe a decrease of p@xmath20 with increasing laser power .
we show that the pl polarisation remains nearly constant in time for experiments from 4k up to 300k , a necessary condition for the success of future valley hall effect experiments based on optically initialized k - valley polarization @xcite .
in addition , @xmath14 does not vary significantly over this temperature range .
these results are surprising when considering the reported decrease of @xmath12 in cw experiments when going from 4k to 300k reported in the literature @xcite . by tuning the laser following
the shift of the a - exciton resonance with temperature we are able to recover at room temperature @xmath21 of the polarization observed at 4k in our sample .
* time resolved photoluminescence * ( a ) t=200k , e@xmath22 ev , e@xmath23 ev . * a - exciton * pl intensity as function of time , plotted in log - scale to emphasize weak , slow component that appears when raising t. the small peak at about 50ps is due to a laser reflection in the set - up .
inset : cw - pl spectrum at t=4k showing energy of intrinsic a - exciton emission and localised exciton emission .
( b ) t=4k , e@xmath22 ev , e@xmath24 ev .
* localised exciton * pl intensity as a function of time ( black ) .
blue line : fit with a single exponential decay ( @xmath25 ) .
inset : the polarization resolved emission of a - exciton and localised exciton which spectrally slightly overlap .
, scaledwidth=40.0% ]
mos@xmath0 flakes are obtained by micro - mechanical cleavage of a natural bulk mos@xmath0 crystal @xcite(from spi supplies , usa ) on a si/90 nm sio@xmath0 substrate .
the 1ml region is identified by optical contrast and very clearly in pl spectroscopy @xcite .
experiments between t=4 and 300k are carried out in a confocal microscope optimized for polarized pl experiments@xcite .
the mos@xmath0 flake is excited by picosecond pulses generated by a tunable frequency - doubled optical parametric oscillator ( opo ) synchronously pumped by a mode - locked ti : sa laser .
the typical pulse and spectral width are 1.6 ps and 3 mev respectively ; the repetition rate is 80 mhz . the laser wavelength can be tuned between 500 and 740 nm . the detection spot diameter is @xmath26 @xmath27 m . for time - integrated experiments ,
the pl emission is dispersed in a spectrometer and detected with a si - ccd camera . for time - resolved experiments ,
the pl signal is dispersed by an imaging spectrometer and detected by a synchro - scan hamamatsu streak camera with an overall time resolution of 4 ps . the pl polarization @xmath12 defined as @xmath28 is analyzed by a quarter - wave plate placed in front of a linear polarizer . here
@xmath29 denotes the intensity of the right ( @xmath3 ) and left @xmath30 circularly polarized emission .
figure [ fig : fig1]a displays the total photoluminescence intensity dynamics at t=4 k following a @xmath3 polarized picosecond excitation laser pulse with an energy e@xmath2 ev , which is within the broad a - exciton absorption line @xcite .
the detection energy corresponds to the pl peak energy e@xmath1 ev . the average laser power used for all experiments in figures [ fig : fig1 ] to [ fig : fig3 ] is below 1 @xmath31 , well below absorption saturation and in the absence of sample heating effects , as discussed in detail in the supplementary material in section[sec : supp ] .
we do not observe any variation of the dynamics when the detection energy is varied within the a - exciton spectrum @xcite .
though the mos@xmath0 pl dynamics is very fast we see in figure [ fig : fig1 ] that it occurs on a slightly longer time scale compared to the one defined by the temporal resolution of the set - up ( compare the mos@xmath0 pl and laser pulse detection in figure [ fig : fig1]a ) . using a deconvolution based on gaussian functions
we can infer that the mos@xmath0 emission time is about 4.5 ps .
we emphasize that this fast pl dynamics is obtained in excitation conditions where exclusively the a - exciton ( not b ) in the k@xmath17 valley is excited ( see inset of fig.[fig : fig1]a ) . here
the energy difference between excitation and detection @xmath32 is about 100 mev .
similarly fast dynamics were recorded in strongly non - resonant excitation conditions with @xmath33 ev @xcite .
+ in figure [ fig : fig1]b , the right ( @xmath34 ) and left ( @xmath35 ) circularly polarized luminescence components have been detected using a @xmath3 polarized laser ( measured @xmath36 ) .
remarkably the pl circular polarization degree is large and remains almost constant during the short exciton emission , around 50% for a laser excitation power p@xmath5 ( blue curve in figure [ fig : fig1]b ) .
lowering the excitation power by two orders of magnitude has a strong impact : the polarization still remains nearly constant , but at a higher value of 60% ( green curve in figure [ fig : fig1]b ) .
as @xmath37 either the polarization generation at this laser energy is not 100% efficient ( due to the optical selection rules ) or due to the sequential recording of @xmath3 and @xmath4 polarized kinetics there is an experimental uncertainty of @xmath7 when fixing the time origin of the @xmath3 emission with respect to @xmath4 .
this results in an experimental uncertainty when determining the circular pl polarization , as indicated by the error bars in figure [ fig : fig1]b . as a result of this time jitter and the short exciton emission time , our experiments do not allow an accurate determination of the spin / valley relaxation time @xmath15 .
the emission at room temperature is very similar to 4k with @xmath38 .
figure [ fig : fig1]c shows @xmath39 constant in time for a laser excitation energy e@xmath8 ev .
the dependence of @xmath12 as a function of the excitation laser energy will be discussed below .
remarkably the pl circular polarization degree which probes the valley initialization measured in time - resolved experiment is similar to the one obtained in cw experiments as a result of the very short exciton lifetime and the absence of measurable polarization decay within this short emission time .
+ in figure [ fig : fig2]a we show that when raising the temperature , in addition to the ps exciton decay a longer lived component is observed .
this is similar to the findings of @xcite under highly non - resonant excitation .
this measured long lived pl component is essentially unpolarized in figure [ fig : fig2]a and the intensity is very weak compared to the short , main pl emission ( log .
therefore even at room temperature the short component , see figure [ fig : fig1]c , determines p@xmath20 .
in addition to the main a - exciton also localised excitons emitting at lower energy ( e@xmath24 ev ) are observed @xcite , see inset of figure [ fig : fig2]a .
the localised exciton emission decays within about 125ps at t=4k ( see figure [ fig : fig2]b ) and is not detectable at higher temperature .
the polarization dynamics is shown in the inset of figure [ fig : fig2]b : as the a - exciton and the broad , localised exciton emission spectrally slightly overlap , we detect the remaining a - exciton polarization at short times , before detecting the essentially unpolarized emission of the localised excitons . it is important to note that the localised exciton polarization dynamics is expected to be more sensitive to the sample parameters ( substrate material , interface defects etc ) than the a - exciton emission . + * pl polarization as a function of laser excitation energy * ( a ) t = 4k , p@xmath5 detected on the a - exciton pl maximum e@xmath1 ev . the a - exciton emission ( blue ) is shown .
( b ) same as ( a ) but for t = 300 k , p@xmath40 , e@xmath9 ev , scaledwidth=40.0% ] the time resolved measurements in figure [ fig : fig1]c show a nearly constant polarization during the exciton pl emission at t=300k . yet in the literature the time - integrated polarization of the a - exciton pl at room temperature is reported to be considerably lower than at 4k @xcite .
we confirm these observations in our sample , see supplementary data in section[sec : supp ] . here
it is important to take into account the considerable red - shift of the direct bandgap as the temperature increases @xcite .
we have therefore performed time and polarization resolved photoluminescence excitation ( ple ) experiments , see figure [ fig : fig3 ] , to vary the initially generated polarization p@xmath41 .
the time - integrated pl is detected at its peak energy ( 1.867 ev and 1.828 ev at t=4 k and 300 k respectively ) . as already observed by different groups @xcite the pl circular polarization degree and thus the valley initialization decrease at t=4
k ( figure [ fig : fig3]a ) when the laser excitation energy increases : it varies for the sample investigated here from @xmath42 for @xmath43 ev down to a value close to zero for @xmath44 ev , similar to the findings by @xcite . here
we observe that @xmath12 slightly increases again up to @xmath45 for @xmath46 ev .
though the polarization minimum is observed roughly in the region where the b - exciton is photogenerated ( see the vertical dotted line ) , its origin needs further clarification since the b - exciton absorbs and emits the same light helicity as the a - exciton in a given k valley @xcite .
the energetically close lying indirect transition from the @xmath47 valley valence band to the conduction band , which is unpolarized , could play a role @xcite . when comparing results obtained on different samples in the literature @xcite , it is important to take into account the laser excitation ( power , pulsed or cw ) and the exact form of the emission spectrum , see section[sec : supp ] .
+ a key result is presented in figure [ fig : fig3]b .
we perform the same ple experiments as in figure [ fig : fig3]a , but at room temperature . when the laser is far from resonance , we observe close to zero polarization .
remarkably , as we lower the laser energy and become more and more resonant , the polarization drastically increases , in the same manner as at 4k .
for the closest energy to resonance that was achievable in practice with our set - up ( filter cut - off for stray laser light ) , we measure an emission polarization of @xmath48 , close to the maximum observed at 4k of @xmath49 @xcite .
this is very encouraging as optical initialization of valley polarization with a suitable excitation source can therefore be very efficient even at room temperature . here
it would be extremely useful to investigate how the strong coulomb interaction @xcite influences the polarization at 300k .
the few time resolved measurements reported in the literature for 1ml mos@xmath0 use highly non - resonant excitation and did not analyse the polarization of the emission @xcite . in @xcite and @xcite
the authors suggest that non - radiative recombination of the excitons could explain the first , short decay observed from 4k to 300k .
we can infer from our time and polarization resolved measurements that the detected polarization in time - integrated experiments is due to the emission during this short time window .
if this initial decay is indeed limited by non - radiative processes or an intrinsic exciton lifetime is still an open question .
the exciton binding energy is estimated to be in the hundreds of mev range @xcite .
systems with large exciton binding energies such as organic films and carbon nanotubes have intrinsic exciton radiative lifetimes on the order of a few ps @xcite . in our measurements
the pl decay time @xmath14 does not change with temperature within our time resolution , we do not observe any activation or any other typical signature of non - radiative processes .
for comparison , for high quality gaas quantum well structures the free wannier - exciton radiative recombination time increases with temperature @xcite .
an argument in favour of a radiative exciton decay within a few ps comes from the observation of the b - exciton emission , 150mev higher in energy than the a - exciton @xcite , as radiative recombination of the b - exciton is in competition with non - radiative decay and relaxation to the a - exciton . here a theoretical prediction of the intrinsic exciton recombination is needed to guide future experiments in this promising system . + in the future also the influence of the substrate on the polarization dynamics needs to be investigated .
here we used mos@xmath0 mls on sio@xmath0 , the most practical substrate for ( opto-)electronic devices .
although time resolved experiments on suspended mos@xmath0 mls gave similar time - resolved absorption results to experiments using substrates @xcite , a comparison with mos@xmath0 on bn where the localised exciton emission was suppressed @xcite would be important to clarify the nature of the observed pl emission .
+ the very fast exciton decay time measured here in mos@xmath0 mls could help explaining several key observations : ( i ) application of a transverse magnetic fields will only shown an influence on the polarization of the ml pl if the precession time is shorter than 4ps @xcite .
( ii ) coherence between valley excitons probed through the observation of stationary , linearly polarized luminescence in related wse@xmath0 devices , which are expected to have similar physical properties @xcite .
we acknowledge partial funding from erc starting grant no . 306719 , programme investissements davenir anr-11-idex-0002 - 02 , reference anr-10-labx-0037-next , and cas grant no .
2011t1j37 , and the national basic research program of china ( 2009cb930502 , 2009cb929301 ) and national science foundation of china [ grants no .
11174338 , no . 11225421 , no .
10934007 and no .
10911130356 ( spinman ) ] .
delphine lagarde and louis bouet contributed equally to this work .
in this supplementary material we present additional , power dependent results obtained under pulsed and a continuous wave excitation conditions at 4k and room temperature in a mos@xmath0 monolayer sample .
* power dependence of photoluminescence under pulsed laser excitation at 4k * using streak camera detection at e@xmath1 ev and e@xmath2 ev ( a ) pl intensity as a function of laser power ( b ) peak position of a - exciton pl emission ( black squares ) . the fwhm of the transitions is indicated ( blue lines ) .
( c ) pl circular polarization degree detected at central point of pl kinetics as a function of laser power ( blue circles).,scaledwidth=40.0% ] optical spectroscopy of mos@xmath0 monolayers is carried out either with standard continuous wave ( cw ) lasers such as the hene laser ( 1.96ev ) or with tunable pulsed lasers such as a ps frequency - doubled optical parametric oscillator synchronously pumped by a mode - locked ti : sa laser ( opo ) , see main text and the work by @xcite .
the laser power used in the experiments is dictated by practical considerations : at low power the signal level has to be sufficient to achieve a high signal - to - noise ratio , at high excitation power structural damage of the mos@xmath0 monolayer has to be avoided .
to the best of our knowledge their is no systematic study of the power dependence of the circular polarization degree @xmath12 of the photoluminescence of ml mos@xmath0 . in the experiments below
we demonstrate that the laser excitation power range used for the results presented in the main text is still within the linear absorption regime ( below saturation ) and does not induce any measurable sample heating or degradation .
+ the power of the opo used in the experiments in the main text was chosen to provide high signal - to - noise ratio time resolved photoluminescence ( trpl ) spectra without causing sample damage .
in general , time resolved measurements using a streak - camera demand a higher signal level than time - integrated measurements using a sensitive si - ccd camera . in figure
[ fig : figsi1](a ) we show the pl intensity as a function of laser power over 2 orders of magnitude detected with the streak camera system i.e. under the same conditions as the data in figure 1 of the main text .
the absorption is not saturated for the laser power values used in time resolved spectroscopy , see figure 1 of the main text , where p@xmath5 corresponds to the highest power used in figure s1 .
* power dependence of photoluminescence under pulsed laser excitation at 300k * detected at si - ccdfor and e@xmath50 ev ( a ) pl intensity as a function of laser power ( black squares ) ( b ) pl circular polarization degree detected at e@xmath9 ev as a function of laser power ( blue circles).,scaledwidth=40.0% ] we find as a general trend that the pl polarization decreases as the laser power increases , a shown in figure [ fig : figsi1](c ) . the transition energy is constant for the laser power range investigated , see figure [ fig : figsi1](b ) .
we therefore conclude that sample lattice heating is an unlikely cause of this polarization decrease , it is possible that depolarization due to a dyakonov - perel type mechanism plays a role .
+ at a temperature of 300k the optical emission of the mos@xmath0 monolayer is weaker than at 4k . in order to obtain an acceptable signal to noise ratio for the pl at low laser power
, we performed the power dependent studies at 300k using a sensitive si - ccd camera at a detection energy of 1.828 ev for a laser energy of 1.913 ev . in figure
[ fig : figsi2](a ) we confirm that also at 300k we are still working in the linear absorption regime . in figure
[ fig : figsi1](b ) we show a decrease of the circular pl polarization degree as the laser power increases , similar to the trend shown at low temperature .
+ as previously reported in the literature @xcite , we observe a decrease in our sample of the circular pl polarization as a function of temperature , see figure [ fig : figsi3 ] . here
the laser energy is kept constant and does not follow the bandgap shrinking with temperature .
+ * photoluminescence of monolayer mos@xmath0 using cw hene laser * ( a ) polarization of the pl emission at 300k : black squares from _ position 1 _ on the sample , red circles from _ position 2 _ ; at 4k : blue upwards triangles _ position 1 _ , pink downwards triangles _
position 2_. ( b ) spectrum of pl at 4k at _ position 2 _ , the energy of the maximum a - exciton emission is e@xmath52 ( c ) pl at 4k at _ position 1 _ ( d ) pl at 300k at _ position 2 _ ( e ) pl at 300k at _ position 1_,scaledwidth=45.0% ] very high pl polarization under cw hene laser excitation at 4k and at 300k has been reported before @xcite and is confirmed in figure [ fig : figsi4](a ) .
first , the laser emission is spectrally narrower than the pulsed opo .
second , although the average laser power monitored with a photodiode , which is used to measure the laser power plotted in figure s1 and s2 , might be comparable in the experiments , the power is several orders of magnitude lower than the peak power in pulsed operation . here
we use the same set - up as in our previous work @xcite , but a different sample . over the investigated power range the polarization does not change drastically , in contrast to the measurements with the opo excitation .
this seems to indicate that the generation rate with the opo at the lowest power is still much higher than the hene generation rate at its maximum .
+ our confocal microscope is also used for spectroscopy on single semiconductor quantum dots ( typical diameter 20 nm ) @xcite and due the high spatial resolution we are able to investigate _ different _ regions of the _ same _ mos@xmath0 monolayer as the sample is mounted on piezo - positioners with a step - size in nm - range in a variable temperature cryostat .
we find measurable variations of the polarization for different positions in the monolayer : the first type of spectra have a more pronounced a - exciton emission and only a very small contribution of the localized exciton . analysing the a - exciton emission
, we find @xmath12 on the order of 70% at 4k and 10% at 300k , see figure [ fig : figsi3 ] .
the second type of spectra is about 15 mev red - shifted with a more pronounced contribution of localized excitons . here
the a - exciton emission is more polarized , with @xmath12 on the order of 85% at 4k and 25% at 300k , close to the values reported in @xcite .
please note that we can not directly compare the pl polarization measured in figures [ fig : figsi1 ] and [ fig : figsi2 ] ( opo excitation ) with the results in figure [ fig : figsi4 ] ( hene excitation at 1.96ev ) as the laser energies are different and the polarization changes abruptly for different energies ( see main text ) . | in monolayer mos@xmath0 optical transitions across the direct bandgap are governed by chiral selection rules , allowing optical valley initialization . in time
resolved photoluminescence ( pl ) experiments we find that both the polarization and emission dynamics do not change from 4k to 300k within our time resolution .
we measure a high polarization and show that under pulsed excitation the emission polarization significantly decreases with increasing laser power .
we find a fast exciton emission decay time on the order of 4ps . by compensating the
temperature induced change in bandgap energy with the excitation laser energy an emission polarization of 40% is recovered at 300k , close to the maximum emission polarization for this sample at 4k . |
among many absorption features in spectra of cool stars , tio molecular absorptions are be recognized as an indicator of basic stellar parameters mainly in two type of stars , cool pulsating giants and super giants and chromospheric active stars which tio absorption can be used as a starspot indicator @xcite .
the temperature sensitivity of tio bands at near - infrared is so pronounced that these bands have always been used as a primary indicator to divide m stars into their subtypes .
another interesting observed fact is that the tio absorption strength is independent of luminosity class , which is approximately the same for dwarf , giants and supergiants @xcite .
@xcite defined an index by measuring the tio@xmath0 absorption band strength at @xmath1 = 886 nm in m giants , and calibrated it versus temperature by using the relation between spectral type and temperature given by ridgway @xcite .
as shown in figure 2 of ramsey , the depth of the tio band at 886 nm is a sensitive temperature indicator for giants in the range @xmath2 k. it is also somewhat useful in the range @xmath3 k. at @xmath4 k , the tio bands become too weak to be useful .
@xcite has invented a tio sensitive index which was proposed to measure the absorption strength of tio molecules at 719 nm .
he showed that this index can distinguish between m super giant sub - types .
@xcite used wing tio - index to measure longitudinal distribution of starspots on the surface of the g8iii star @xmath1 andromeda and have estimated a period of 10 - 15 years for chromospheric activity of this star . to provide a correct measure of absorption at 719 nm wing has estimated the continuum over this absorption band by extrapolating the continuum in 754 and 1024 which he claimed are mostly free of absorption .
this has made a three filter system which is known as the wing filter system .
one of the major advantages of wing tio - index has been its strong sensitivity to effective temperature for atmospheres cooler than 4000 k. the linear reduction of the index by increasing @xmath5 has made it a temperature index which has been used by several authors including himself .
the possible contamination of other molecular absorptions in continuum regions would make this linear relation inaccurate and misleading .
technological advancements in spectroscopic facilities which provide us with ultra high resolution spectra especially in near infrared have made it possible to better investigate the wing continuum bands and to check if they are contaminated or if it is possible to shift them to represent a more accurate continuum , free of molecular absorbtion .
the basic motivation of this paper is investigating the synthetic and also high resolution observed spectra of a sample of cool giants to find a more precise temperature calibration for cool stars .
+ in this paper , we present an updated tio - index which is proposed to measure absorption strength of tio bands at @xmath1 = 719 nm , by using the high - resolution spectroscopic data from the .
we first search in a grid of synthetic spectra to find a spectral band which closely resembles the continuum .
section 2 describes wing system properties for measuring the tio@xmath1719 nm absorption strength and the definition for tio - index . in section 3 ,
a modified continuum band is selected which we claim is less affected from absorption line contamination .
the stellar sample is described in section 4 .
section 5 presents the results , including wing tio - index and our index values for a sample of stars along with their calibration as a function of the effective temperature .
@cccc@ filter & region measured & central & band - pass + & & wavelength & fwhm + & & ( nm ) & ( nm ) + a ... & tio@xmath1719 nm band & 719 & 11 + b ... & continuum & 754 & 11 + c ... & continuum & 1024 & 42 + the wing three color system @xcite consists of three filters , designated a , b and c , that can be used to measure the four principal properties of cool stars : their magnitude in near - infrared region , color index in near infrared , temperature , and spectral type .
the central wavelengths and full width at half - maximum band - pass at these filters are listed in table [ tbl:1 ] .
the central wavelength of the a filter ( full width at half - maximum is 11 nm ) at 719 nm is one of the strongest absorption bands of tio @xmath0 molecules .
the b filter with central wavelength of 754 nm measures the continuum intensity in the right wing of tio absorption band and the c filter ( full width at half - maximum is 42 nm ) with central wavelength of 1024 nm is located in a continuum region far from a and b. the strength of the tio absorption at 719 nm is measured by dividing the stellar flux received through filter a to filter b. however , this ratio is slightly sensitive to the slope of the energy distribution of the star .
@xcite has removed this effect by linearly extrapolating the b and c band - passes to the center of a band and by introducing a measure of absorption at a band by a tio index : @xmath6 where a , b and c are the apparent magnitudes in a , b and c filters . the numerical coefficient of 0.13 is measured by the spacings of the filters in wavelength .
as mentioned earlier , tio - index is a measure of the depth of tio@xmath1719 nm molecular absorption band .
precise measurement of the continuum at a band - pass will give a more accurate absorption strength and by the way a more precise calibration .
we use a set of synthetic spectra to search for more appropriate b and c bands , free from contamination of other absorption features , using kurucz model atmosphere @xcite . with [ m / h ] = 0 ( i.e. , star abundance ) , @xmath7 g between 3 to 5 , @xmath8 g = 0.5 and effective temperature between 2900 and 4000 k , @xmath9 t = 100 k where tio molecules have significant absorbtion at near - infrared , a number of 60 synthetic spectra were obtained .
figure [ fig:1 ] shows a few sample synthetic spectra which were produced in this exercise .
the results of this investigation lead to a redefinition of two new filters in the cleaner continuum areas that are designated d and e. the central wavelength and their bandwidth along with the a filter features are given in table [ tbl:2 ] .
the cool red stars spectra are dominated by molecular absorption bands , especially tio and vanadium oxide ( vo ) @xcite .
careful investigation of the wing continuum shows that there are weaker molecular absorption features like [email protected] nm , [email protected] nm , [email protected] nm , [email protected] nm , and [email protected] nm in his selected continuum region @xcite . @cccc@ filter & region measured & central & band - pass + & & wavelength & fwhm + & & ( nm ) & ( nm ) + a ... & tio@xmath1719 nm band & 719 & 11 + d ... & continuum & 702 & 6 + e ... & continuum & 956 & 30 + @xcite carried out a survey of a sequence of m and k
stars spectra in the red and infrared ranges , have provided a reference list for the absorption features in these stars of all luminosity classes . in table [ tbl:3 ] , vo absorption wavelength regions are merely given for a quick review . also , it is remarkable that the dependence of vo absorption with spectral type in the dwarfs and giants is similar @xcite .
+ table [ tbl:3 ] shows that the new continuum band - passes d and e are essentially free of molecular absorption .
@l l l@ wavelength & absorption & notes + ( nm ) & feature & + 733.4 , 737.2 , 739.3 , 740.5 , 741.8 , & vo & seen only in late - m stars + 747.2 , 753.4 & & + + 785.1 , 786.5 , 789.7 , 790.0 , 791.9 , & vo & obvious in late - m dwarfs + 792.9 , 793.9 , 796.1 , 796.7 , 797.3 & + + 852.1 , 853.8 , 857.4 , 859.7 , 860.5 , & vo & found only in late - m stars + 862.4 , 864.9 , 866.8 & + + using the new filters , the tio - index is redefined as @xmath10 where a , d and e are the apparent magnitudes in a , d and e filters . according to the central wavelength of the new filters ,
the numerical coefficient of 0.016 is measured by the spacings of the filters in wavelength .
to check the feasibility of the new tio - index , we select a set of 23 giant k5-m8 observed spectra from uves paranal observatory project @xcite .
all target stars have been observed by the vlt ut2 ( very large telescope ) at european southern observatory using the ultraviolet - visual echelle spectrograph ( uves ) .
uves is a very efficient high resolution spectrograph designed to operate from about 300 to 1100 nm .
the spectra are accessible from the eso archives .
these sample stars are listed in the first column of table [ tbl:4 ] .
@l l l l l l l@ star & spectral & obs . & wing & updated & @xmath5 & ref . , @xmath5 + & type & date & tio - index & tio - index & & + hd11695 & m4 iii & 2003 - 02 - 08 & 0.728 & 0.749 & 3550 & 16 + hd18191 & m6 iii & 2001 - 12 - 05 & 1.833&1.092 & 3266 & 4,12,13 + hd18242 & m7 iii & 2001 - 07 - 24 & 1.886&1.397 & 3102 & 3,9 + hd18884 & m1.5 iii & 2003 - 01 - 02 & 0.174&0.362 & 3788 & 6,8,14,17 + hd29139 & k5 iii & 2002 - 03 - 04 & 0.024 & 0.279&3943 & 2 + hd29712 & m8 iii & 2002 - 12 - 28 & 1.535&1.967 & 2900 & 1,5 + hd39801 & m2 iab & 2002 - 09 - 13 & 0.307 & 0.403 & 3659 & 12 + hd102212 & m1 iii & 2004 - 03 - 10 & 0.050&0.311 & 3824 & 11,14,15 + hd118767 & m5 iii & 2001 - 07 - 09 & 1.382&1.036 & 3339 & 3,9 + hd119149 & m2 iii & 2001 - 07 - 12 & 0.243&0.406 & 3782 & 12 + hd120052 & m2 iii & 2001 - 07 - 13 & 0.207&0.397 & 3715 & 12,14 + hd123934 & m1 iii & 2002 - 01 - 24 & 0.090&0.320 & 3850 & 12 + hd126327 & m7.5iii & 2006 - 03 - 20 & 1.725&1.578 & 3050 & 11,13 + hd146051 & m0.5 iii & 2006 - 03 - 27 & 0.428 & 0.282&3879 & 6,18 + hd149447 & k5 iii & 2003 - 03 - 13 & 0.001&0.242 & 4000 & 12 + hd167006 & m3 iii & 2003 - 09 - 07 & 0.403&0.537 & 3640 & 12 + hd167818 & k5 iii&2001 - 06 - 23 & 0.092&0.080 & 4450 & 10 + hd175588 & m4 ii & 2003 - 05 - 01 & 0.815&0.746 & 3550 & 4 + hd189124 & m6 iii & 2001 - 06 - 20 & 1.396&0.992 & 3334 & 3,9 + hd189763 & m4 iii & 2001 - 07 - 28 & 0.676&0.668 & 3568 & 3,9 + hd213080 & m4.5 iiia & 2001 - 07 - 12 & 0.708&0.696 & 3514 & 3,9 + hd214952 & m5 iii & 2002 - 05 - 24 & 0.810&0.742&3480 & 7 + hd224935 & m3 iii & 2001 - 11 - 29 & 0.536&0.581&3634 & 6,12 + for k5-m8 stars ; new filters ] for k5-m8 stars ; wing filters ]
to determine spectral types , it is necessary to calibrate the tio - index as a function of effective temperature by our sample stars of known spectral types .
first , the tio - index of the sample stars is calculated by using our proposed index .
the results are given in the fifth column of table [ tbl:4 ] .
the effective temperatures of most of the sample stars were determined by averaging over the temperatures found from the existing literature . for stars where these were not found in the literature , we used the calibration of houdashelt and bell @xcite from their published v - k indices ( results are given in the sixth column of table [ tbl:4 ] ) .
figure [ fig:2 ] represents the variation of updated tio - index relative to effective temperature for our sample stars .
a quadratic least - squares fit was performed to the points as @xmath11 the dispersion of the residuals around the fit is 0.04532 . as it can be seen from the curve , the value of tio@xmath12-index is declining with increasing temperature and for temperatures above 4500 k , asymptotically approaching zero where the tio absorption feature is not present or is very weak . by decreasing temperature
, the tio@xmath12-index increases up to a value about 2.00 which corresponds to spectral type m8 . to compare our results with original wing calibration , we recalculate the wing tio - index for these stars .
the results of calculation are given in the fourth column of table [ tbl:4 ] .
figure [ fig:3 ] represents the variation of the original tio - index relative to the effective temperature for our sample stars .
a quadratic least - squares fit was performed to the points as @xmath13 the dispersion of the residuals around the fit is 0.2431 .
this curve displays the same trend in fig .
3 but with more scatter , especially at later spectral - types .
in this paper , we have presented a new continuum filters set for measuring the tio - index and an updated calibration of the tio - index with effective temperature . by comparing the dispersion of the residuals around the fit for the two curves ,
it is clear that the quality of the fit for the updated tio - index is better than for the original wing index .
moreover , we used the uves high resolution spectrum . the high scatter in fig .
3 might be due to the presence of some absorption lines in the original wing b and c continuum band - passes .
the new tio - index proposed here can provide more precise information about starspot activity in solar like stars . to test the proposed revised wing tio photometric system ,
actual photometry should be carried out .
another advantage of the new system is that the e filter at 956 nm is better placed at shorter wavelength than the original wing c filter at 1024 nm .
most ccd detectors rapidly decrease in sensitivity at wavelengths @xmath14 1000 nm .
bagnulo , s. , jehin , e. , ledoux , c. , cabanac , r. , melo , c. , & gilmozzi , r. 2003 , the eso messenger , 114 , 10 .
bedding , t.r . ,
zijlstra , a.a . , von der luhe , o. , robertson , j.g . , marson , r.g . ,
barton , j.r . , & carter , b.s .
bell r.a . , & gustafsson , b. 1989 , , 236 , 653 .
bell , r.a .
, & houdashelt , m.l .
1997 , american astronomical society , 191st aas meeting , bulletin of the american astronomical society , vol .
29,p . 1283 .
carroll , bradley w. , & ostlie , dale a. , 2007 , `` an introduction to modern astrophysics '' , pearson : san francisco , boston , new york , cape town , hong kong , london , madrid , mexico city , montreal , munich , paris , singapore , sydney , tokyo , toronto , vol .
8 , p. 205 .
castelli , f. , & kurucz , r.l . : in : iaus , vol .
210 , 2004 .
cenarro , a.j . ,
peletier , r.f .
, sanchez blazquez , selam , s.o.,toloba , e. , cardiel , n. , falcon barroso , j. , gorgas , j. , jimenez - vicente , j. , & vazdekis , a. 2007 , , 374 , 664 .
dumm , t. , & schild , h. 1998 , , 3 , 137 .
dyck , h.m .
, van belle , g.t . , & thompson , r.r .
, 1998 , , 116 , 981 .
engelke , c.w . , price , s.d . , &
kraemer , k.e .
, 2006 , , 132 , 1445 .
flower , p.j .
, 1996 , , 469 , 355 .
houdashelt , m.l . ,
bell r.a . , & sweigart , a.v .
2000 , , 119 , 1448 .
jehin , e. , bagnulo , s. , melo , c. , ledoux , c. , & cabanac , r. 2005 , iaus , 261 , 228 .
jones , k.l . , robinson , r.d .
, slee , o.b .
, & stewart , r.t . , 1992 , , 256 , 535 .
kirkpatrick j.d . ,
henry t.j .
, & mc carthy d.w .
1991 , , 77 , 417 .
koleva , m. , & vazdekis , a. , 2012 , , 538 , a143 .
kurucz , r.l .
1991 , in precision photometry : astrophysics of the galaxy , ed .
p. davis , a. upgren , and k. janes ( schenectady : l. davis press ) , 27 .
mcdonald . , zijlstra , a.a .
, & boyer , m.l .
2012 , , 1 , 16 .
mirtorabi , m.t . ,
wasatonic , r. , & guinan , e.f .
2003 , , 125 , 3265 .
neff , j.e .
, oneal d. , & saar s.h .
1995 , , 452 , 879 .
oneal , d. , neff , j.e . , & saar , s.h .
1998 , , 507 , 919 .
oneal , d. , saar , s.h . , & neff , j.e . , 1996 , , 463 , 766 .
oneal , d. , saar , s.h . ,
neff , j.e .
, & cuntz , m. , 13th cool stars workshop , hamburg , 2004 .
prugniel , ph . , vauglin , i. , & koleva , m. , 2011 , , 531 , 165 .
ramsey , l.w .
1981 , , 86 , 557 .
ridgway , s.t .
, joyce , r.r . , white , n.m . , & wing , r.f .
1980 , , 235 , 127 .
strassmeier , k.g . , & schordan , p. , 2000 ,
astronomische nachrichten , 321 , 277 .
park , s. , kang , w. , lee , j.e .
, & lee , s.g . , 2013 , , 146 , 73 .
valenti , j.a .
, piskunov , n. , & johns - krull , c.m . 1998 , , 498 , 851 .
wing , r.f .
, & white , n.m .
1978 , iaus , 80 , 451 .
wing , r.f .
1992 , javso , 21 , 42 .
wittkowski , m. , aufdenberg , j.p . , & kervella , p. , 2004
, , 413 , 711 .
wittkowski , m. , aufdenberg , j.p . ,
driebe , t. , roccatagliata , v. , szeifert , t. , & wolff , b. , 2006 , , 460 , 855 .
yue wu , harinder p. singh , philippe prugniel , ranjan gupta , & mina koleva , 2010 , , 525 , a71 . | by careful searching of synthetic and observed spectra in a sample of cool giant and supergiant stars , we have updated the continuum band - passes of near - infrared wing three filter system . this photometric system measures the strength of titanium oxide ( tio ) absorption in near - infrared ( nir ) at 719 nm .
we show that new reference continuum band - passes are essentially free from molecular absorptions and the updated tio - index defines the temperature variation in a sample of cool giants with less scatter .
a tio - index vs. effective temperature calibration is derived based on new continuum band - passes . |
start with four mutually tangent circles , one of them internally tangent to the other three as in fig .
[ circles ] .
one can inscribe into each of the curvilinear triangles in this picture a unique circle ( the uniqueness follows from an old theorem of apollonius of perga circa 200 bc ) .
if one continues inscribing the circles in this way the resulting picture is called an apollonian circle packing ( acp ) .
a key aspect of studying such packings is to consider the radii of the circles which come up in a given acp . however , since these radii become small very quickly , it is more convenient to study the _ curvatures _ of the circles , or the reciprocals of the radii
. studied in this way , acp s possess a beautiful number theoretic property that all of the circles in an acp have integer curvature if the initial four have integer curvature .
the number theory associated with these integral acp s has been investigated extensively in @xcite , @xcite , and @xcite .
a central theorem to any of the results in these papers is descartes theorem , which says that the curvatures @xmath1 of any four mutually tangent circles satisfy what is called the descartes equation , @xmath2 where a circle which is internally tangent to the other three is defined to have negative curvature ( see @xcite for a proof ) .
given this formula we may assign to every set of @xmath3 mutually tangent circles in an integral packing @xmath4 a vector @xmath5 of the circles curvatures .
we use descartes equation to express any integral acp as an orbit of a subgroup of the orthogonal group @xmath6 acting on @xmath7 .
this subgroup , called the apollonian group , is specified in @xcite , and we denote it by @xmath8 .
it is a group on the four generators @xmath9 @xmath10 derived by fixing three of the coordinates of @xmath7 and solving @xmath11 for the fourth .
note that each @xmath12 is of order @xmath13 and determinant @xmath14 .
also , @xmath12 fixes all but the @xmath15th coordinate of @xmath5 , producing a new curvature in the @xmath15th coordinate . in their paper @xcite the five authors graham , lagarias , mallows , wilks , and yan
ask several fundamental questions about the curvatures in a given integer acp , which have mostly been resolved in @xcite , @xcite , @xcite , and @xcite .
in particular , they make some observations about the congruence classes of curvatures which occur in any given acp , and suggest a strong density " conjecture , that every sufficiently large integer satisfying these congruence conditions should appear as a curvature in the packing . in @xcite ,
the first author and bourgain prove a weaker conjecture of graham et.al . of this flavor that the integers appearing as curvatures in a given acp make up a positive fraction of @xmath16 .
proving the strong density " conjecture would be significantly more difficult . in this paper
, we use the @xmath17-adic description of the apollonian orbit from @xcite to formulate this conjecture precisely and provide strong experimental evidence in section [ locglobal ] in support of it .
our conjecture is specified further in the case of two different acp s in section [ locglobal ] .
it is stated generally here .
* local to global principle for acp s : * [ lg ] let @xmath4 be an integral acp and let @xmath18 be the set of residue classes mod @xmath19 of curvatures in @xmath4 .
then there exists @xmath20 such that any integer @xmath21 whose residue mod @xmath19 lies in @xmath18 is in fact a curvature of a circle in @xmath4 .
we note that @xmath22 above depends on the packing @xmath4 under consideration . in this paper , we investigate two acp s which we call the _ bugeye _ packing @xmath23 and the _ coins _ packing @xmath24 .
these packings are represented by right action of the apollonian group on @xmath25 and @xmath26 , respectively ( see fig . [ bugcoin ] for a picture ) . in the case of @xmath23 , our data suggests that @xmath27 exists and is @xmath28 , as we find no integers @xmath29 which violate the above conjecture .
the data for @xmath24 , however , suggests that @xmath30 exists , but that it is @xmath31 .
namely , there are integers @xmath32 in certain residue classes in the set @xmath33 which do _ not _ appear as curvatures in the packing we consider .
we explain this further in section [ locglobal ] . another interesting problem regarding acp s is counting circles of prime curvature in a given packing .
sarnak proves in @xcite that there are infinitely many circles of prime curvature in any packing . in light of this
, we give a heuristic in section [ primeacp ] for the weighted prime count @xmath34 : @xmath35 where @xmath36 is a circle in the packing @xmath4 and @xmath37 is its curvature .
this count is closely related to the number @xmath38 of prime curvatures less than @xmath0 in a packing @xmath4 ( see remark [ katsremark ] ) .
we confirm experimentally that our heuristic holds for the packings @xmath23 and @xmath24 .
we note that our heuristic does not depend on the chosen packing @xmath4 in fact , it yields the correct count of prime curvatures for all of the packings we checked . we summarize it in the following conjecture : [ pconj ]
let @xmath39 be the number of circles in a packing @xmath4 of curvature less than @xmath0 , and let @xmath34 be as in ( [ intropsi ] ) .
then as @xmath40 , @xmath41 where @xmath42 is the value of the dirichlet @xmath43-series at @xmath13 with character @xmath44 for @xmath45 and @xmath46 for @xmath47 .
sarnak also shows in @xcite that there are infinitely many pairs of tangent circles of prime curvature ( we call these _ kissing primes _ ) .
we address the question of counting kissing primes in @xmath4 via the weighted sum @xmath48 : @xmath49 where @xmath50 is the set of unordered pairs of tangent circles @xmath51 of prime curvature in a packing @xmath4 , and @xmath37 and @xmath52 denote their respective curvatures . in this case , it is less obvious what the relation is between @xmath48 and the number @xmath53 of kissing prime circles in a packing @xmath4 both of whose curvatures are less than @xmath0 .
we therefore stick with @xmath48 in our computation : [ tpconj ] let @xmath48 be as in ( [ introsum2 ] ) , and let @xmath39 be the number of circles in a packing @xmath4 of curvature less than @xmath0 .
then @xmath54 where @xmath55 is as above and @xmath56 is given by @xmath57 these heuristics are computed by counting primes in orbits of the apollonian group , which is possible due to recent results of bourgain , gamburd , and sarnak in @xcite , as well as the recent asymptotic count of kontorovich and oh in @xcite for the number @xmath39 .
our computer experiments were conducted using java and matlab , and the programs are available at http://www.math.princeton.edu/~ksanden/elenakatcode.html .
a brief description of our algorithm and a discussion of its running time can be found in section [ algorithm ] . since all of the computations and claims in this paper concern the orbit @xmath58 of the apollonian group @xmath8 acting on a vector @xmath5 , we recall the description of the orbit modulo @xmath59 for any integer @xmath59 from @xcite .
we use this description throughout sections [ primeacp ] and [ locglobal ] .
( fuchs ) : [ padicorbit ] let @xmath58 be an orbit of @xmath8 acting on a root quadruple of a packing , and let @xmath60 be the reduction of this orbit modulo an integer @xmath61 .
let @xmath62 denote the cone without the origin , and let @xmath63 be @xmath36 over @xmath64 : @xmath65 write @xmath66 with @xmath67 and @xmath68 where @xmath69 . write @xmath70 where @xmath71
. then * the natural projection @xmath72 is surjective .
* let @xmath73 be the natural projection
. then @xmath74 . *
the natural projection @xmath75 is surjective and @xmath76 .
this result is obtained by analyzing the reduction modulo @xmath59 of the inverse image of the apollonian group @xmath8 in the spin double cover of @xmath77 .
we note that theorem [ padicorbit ] implies that the orbit @xmath58 of @xmath8 has multiplicative structure in reduction modulo @xmath78 and that it is completely characterized by its reduction mod @xmath19 , or by @xmath79 in our notation .
this explains the dependence on @xmath18 in conjecture [ lg ] .
* acknowledgements : * we thank peter sarnak , alex kontorovich , and kevin wayne for many insightful conversations and helpful suggestions .
in @xcite , bourgain et.al . construct an affine linear sieve that gives lower and upper bounds for prime and almost - prime points in the orbits of certain groups . in this section
, we use their analysis to predict precise asymptotics on the number of prime curvatures less than @xmath0 , as well as the number of pairs of tangent circles of prime curvature less than @xmath0 in a given primitive apollonian packing @xmath4 .
the conditions associated with the affine linear sieve for @xmath8 are verified in @xcite .
we recall the setup below .
let @xmath80 , and note that @xmath81 is finite since the number of circles of any given radius can be bounded in terms of the area of the outermost circle .
we consider @xmath82 and note that the sum @xmath83 where @xmath39 is the number of circles of curvature less than @xmath0 and is determined by the asymptotic formula in @xcite ( see lemma [ oh ] ) .
key to obtaining our asymptotics is computing the averages of progressions mod @xmath59 of curvatures less than @xmath0 , where @xmath61 ranges over positive square - free integers of suitable size . to this end , we define @xmath84 and introduce a multiplicative density function @xmath85 for which @xmath86 where the remainder @xmath87 is small according to the results in @xcite . in the case of acp s , we define @xmath85 as follows .
let @xmath58 be an integral orbit of @xmath8 , and let @xmath60 be the reduction of @xmath58 modulo @xmath59 for a square - free positive integer @xmath59 .
then @xmath88 where @xmath89 is the @xmath90th coordinate of @xmath7 .
we recall from theorem [ padicorbit ] that the orbit @xmath60 has a multiplicative structure which carries over to the function @xmath91 so that @xmath92 thus in order to evaluate @xmath93 for arbitrary square - free @xmath59 , we need only to determine @xmath94 for @xmath17 prime .
this is summarized in the following theorem .
[ betathm ] let @xmath95 be the prime factorization of a square - free integer @xmath61 .
then * @xmath96 for @xmath97 . * for @xmath98
, we have @xmath99 * for any orbit @xmath58 there exist two coordinates , @xmath15 and @xmath90 , such that @xmath100 @xmath101 we say that the @xmath15th and @xmath90th coordinates are even throughout the orbit , while the other two coordinates are odd throughout the orbit . * for @xmath102 , let @xmath103 for @xmath104 . then @xmath105 the statements in ( i ) and ( ii ) follow from theorem [ padicorbit ]
let @xmath7 be the root quadruple ( the quadruple of the smallest curvatures ) of the packing @xmath4 . to show ( iii ) , note that any quadruple in a primitive integral acp consists of two even and two odd curvatures ( see @xcite for a discussion ) . without loss of generality , assume @xmath106 mod @xmath13 , so @xmath107 and @xmath108 in this case .
since the apollonian group is trivial modulo @xmath13 , we have that every vector in the orbit is of the form @xmath109 mod @xmath13 , so we have what we want . to prove ( iv ) , we use results in @xcite and recall from theorem [ padicorbit ] that @xmath110 is the cone @xmath111 for @xmath112 . thus the numerator of @xmath113 is @xmath114 where @xmath115 is the descartes quadratic form and @xmath112 .
so the numerator counts the number of nontrivial solutions to the ternary quadratic form obtained by setting one of the @xmath116 in the descartes form @xmath117 to @xmath118 .
similarly we have that the denominator of @xmath113 is @xmath119 where @xmath112 .
so the denominator counts the number of nontrivial solutions to the descartes form .
the number of nontrivial solutions to ternary and quaternary quadratic forms over finite fields is well known ( see @xcite , for example ) .
namely , @xmath120 for @xmath112 , and @xmath121 combining ( [ orbit ] ) and ( [ orbitbeta ] ) , we obtain the expression in ( [ bvalue ] ) for @xmath112 . for @xmath122
, we compute @xmath110 explicitly and find that there are two possible orbits of @xmath8 modulo @xmath123 which are illustrated via finite graphs in fig .
[ mod31 ] and fig .
[ mod32 ] .
both of these orbits consist of @xmath124 vectors @xmath5 . in both orbits ,
@xmath3 of the vectors @xmath7 have @xmath125 for any @xmath126 .
thus @xmath127 as desired . in the following two sections ,
we use this setup to produce a precise heuristic for the number of circles of prime curvature as well as the number of pairs of tangent circles of prime curvature less than @xmath0 in a given acp . in order to compute the number of prime curvatures in an acp as proposed in conjecture [ pconj ]
, we use the setup above paired with properties of the moebius function to pick out primes in the orbit of @xmath8 ( see ( [ lam ] ) ) .
we use the asymptotic in @xcite for the number @xmath39 of curvatures less than @xmath0 in a given packing @xmath4 : ( kontorovich , oh ) : [ oh ] given a bounded apollonian circle packing @xmath4 , there exists a constant @xmath128 which depends on the packing , such that as @xmath129 , @xmath130 where @xmath131 is the haussdorf dimension of the limit set of @xmath8 acting on hyperbolic space . for the purpose of our computations
, we will need a slightly stronger statement of theorem [ oh ] , as we will sum over each coordinate of the points in the orbit of @xmath8 separately . namely , each circle in the packing is uniquely represented in the orbit @xmath58 as a maximal coordinate of a vector @xmath7 in @xmath132 .
we would like to know how many circles there are of curvature less than @xmath0 that are represented in this way in the @xmath15th coordinate of a vector in the orbit .
we denote this by @xmath133 : @xmath134 where @xmath135 denotes the @xmath15th coordinate of @xmath5 which is also a maximal coordinate of @xmath7 . to this end
we have [ sullivan ] let @xmath133 and @xmath39 be as above .
then @xmath136 as @xmath0 approaches infinity .
the computation in @xcite of the main term in the asymptotics in theorem [ oh ] relies on the patterson - sullivan measure on the limit set of the apollonian group @xmath8 . in order to prove lemma [ sullivan ] ,
we show that this measure is invariant under transformations on the coordinates of a vector @xmath7 in an orbit @xmath58 of @xmath8 . to this end , let @xmath137 be the group of permutations of the coordinates @xmath138 of a vector @xmath139 .
the group @xmath137 is finite and its elements can be realized as @xmath3 by @xmath3 integer matrices .
for example , the matrix @xmath140 switches the first and second coordinates by left action on @xmath141 .
let @xmath142 be the group of @xmath3 by @xmath3 matrices generated by the apollonian group @xmath8 together with @xmath137 , and note that each element of @xmath137 normalizes @xmath8 .
for example , if @xmath143 is as above , we have @xmath144 , where @xmath145 and @xmath146 are as in ( [ gens ] ) .
similarly , any element of @xmath137 switching the @xmath15th and @xmath90th coordinate of @xmath7 conjugates @xmath12 to @xmath147 in this way .
thus @xmath148 is finite , and so @xmath43 is a finite extension of @xmath8 . in particular
, this implies that the patterson - sullivan measure for @xmath43 is the same as for @xmath8 . since @xmath43 is precisely an extension of @xmath8 by the permutations of the coordinates of @xmath7 , we have that the patterson - sullivan measure is invariant under @xmath137 . together with theorem
[ oh ] and its proof in @xcite , this proves the lemma
. since we are interested in counting the points in @xmath58 for which @xmath135 is prime , we sum over points for which @xmath135 is @xmath118 modulo some square - free @xmath59 .
it is convenient to count primes in the orbit of @xmath8 with a logarithmic weight . to this end
, we consider the function @xmath149 for which it is well known that @xmath150 where @xmath151 is the moebius function . using this we write down a concrete expression for the number of prime curvatures less than @xmath0 in a packing @xmath4 counted with a logarithmic weight : [ primtm ] let @xmath135 be the @xmath15th coordinate of a vector @xmath7 in @xmath58 such that @xmath135 is the maximal coordinate of @xmath7 , and let @xmath34 be as before .
then @xmath152 the sum in ( [ lambdasum1 ] ) is a count of all circles whose curvatures are _ powers _ of primes . including powers of primes in our count will not affect the final answer significantly .
namely , let @xmath153 be the number of circles in a packing @xmath4 whose curvatures are less than @xmath0 and perfect squares .
note that @xmath154 this is insignificant compared to the count of all curvatures in theorem [ oh ] , so the sum in ( [ lambdasum1 ] ) is the correct one to consider .
denote by @xmath155 be the level distribution from the analysis in @xcite
i.e. our moduli @xmath59 are taken to be less than @xmath156 .
we combine lemma [ primtm ] , lemma [ sullivan ] , and the expression for @xmath157 in ( [ lam ] ) to get @xmath158 assuming that the moebius function @xmath151 above becomes random as @xmath59 grows , the sum over @xmath159 in ( [ sumprimes1 ] ) is negligible , and we omit it below .
we proceed by rewriting the sum over @xmath160 in ( [ sumprimes1 ] ) using the density function @xmath85 in ( [ betadef ] ) .
recall that the analysis in @xcite and @xcite gives us that @xmath161 where @xmath87 is small on average .
in particular , @xmath162 for some @xmath163 .
paired with the assumption that @xmath164 is random , this evaluation of the remainder term allows us to rewrite ( [ sumprimes1 ] ) as follows : @xmath165 to compute the innermost sum in the final expression above , note that @xmath166 assuming once again that the sum over @xmath159 is insignificant due to the conjectured randomness of the moebius function , we have that the sum over @xmath160 in ( [ infinity ] ) can be approximated by the sum over all @xmath59 . with this in mind , the following lemma yields the heuristic in conjecture [ pconj ] .
[ infinitesumcalc ] let @xmath167 be as before .
we have @xmath168 where @xmath169 is the value of the dirichlet @xmath43-function at @xmath13 with character @xmath170 we introduce a function @xmath171 and note that its derivative at @xmath118 is precisely what we want :
@xmath172 since the functions @xmath173 , @xmath164 , and @xmath174 are all multiplicative , we may rewrite @xmath175 as an euler product and obtain @xmath176 where @xmath177 is holomorphic in @xmath178 .
differentiating , we obtain @xmath179 since @xmath180 and @xmath181 .
thus it remains to compute @xmath182 part ( iii ) of lemma [ betathm ] says @xmath183 for two coordinates @xmath184 .
for these two coordinates @xmath185 and so @xmath186 otherwise @xmath187 and we have @xmath188 thus the sum we wish to compute is @xmath189 , as desired .
lemma [ infinitesumcalc ] implies that the contribution of the two of the coordinates that are even throughout the orbit to the sum in ( [ sumprimes ] ) is @xmath118 , and the contribution for the other two coordinates is @xmath190 yielding the predicted result in conjecture [ pconj ] .
[ katsremark ] it is well known that @xmath191 as @xmath40 .
thus conjecture [ pconj ] can also be stated in terms of @xmath38 : @xmath192 since our computations rely on the multiplicativity inherent to the reduction mod @xmath59 of the apollonian group @xmath8 , our heuristic is independent of the chosen packing @xmath4 in which we count prime curvatures .
this is confirmed by our data : fig .
[ pntfigure ] and fig .
[ pntfigureb ] above show the graphs of @xmath193 where @xmath194 and @xmath23 as in section [ intro ] and @xmath195 . in both cases , ( [ graphp ] )
converges to @xmath196 as predicted in conjecture [ pconj ] . in this section
, we use the analysis in @xcite , as well as the conjectured randomness of the moebius function to arrive at the heuristic in conjecture [ tpconj ] for the number of kissing primes , i.e. pairs of tangent circles both of prime curvature less than @xmath0 . with the same notation as in section [ theory ] , we would now like to count the points in @xmath58 for which @xmath135 and @xmath89 are prime for some @xmath197 , so we sum over points for which either @xmath135 or @xmath89 is @xmath118 modulo some square - free @xmath59 . to do this
we need the total number of pairs of mutually tangent circles of curvature less than @xmath0 in a packing @xmath4 . if @xmath39 is the number of circles of curvature up to @xmath0 as specified by theorem [ oh ] , it is not difficult to see that @xmath198 since each circle of curvature @xmath199 in the packing is tangent to a distinct triple of circles of curvature @xmath200
. this can be seen via a simple proof by induction .
we again employ the function @xmath157 in order to write down a concrete expression for the number of kissing primes less than @xmath0 in a packing @xmath4 .
[ twintm ] let @xmath135 and @xmath89 be two distinct coordinates of a vector @xmath7 in @xmath58 , where @xmath135 denotes the maximum coordinate of @xmath7 , and let @xmath48 be as before .
then @xmath201 again , the sum in ( [ lambdasumm ] ) is a count of all mutually tangent pairs of circles whose curvatures are _ powers _ of primes , but by a similar argument to that in section [ theory ] we have that including powers of primes in our count does not affect the final answer significantly .
note that in order to evaluate ( [ lambdasumm ] ) we introduce in ( [ referg ] ) a function which counts points in @xmath58 for which _ two _ of the coordinates are @xmath118 modulo @xmath17 .
denote by @xmath202 be the level distribution from the analysis in @xcite i.e. the moduli @xmath61 in the computations below may be taken to be less than @xmath156 .
we rewrite the expression in ( [ lambdasumm ] ) using ( [ lam ] ) and get @xmath203 where @xmath204 and @xmath205 as in section [ theory ] , we omit @xmath206 in ( [ sum11 ] ) under the assumption that @xmath164 behaves randomly for large values of @xmath59 . along with the results about the remainder term in the sieve in @xcite ,
the expression in ( [ sum11 ] ) becomes @xmath207\leq d'}\beta_i\left(\frac{d_i}{(d_i , d_j)}\right)\beta_j\left(\frac{d_j}{(d_i , d_j)}\right)g((d_i , d_j))\mu(d_i)\mu(d_j)\log d_i\log d_j+\textrm{o}(x^{\delta-\epsilon_0}),\ ] ] where @xmath167 is as before , @xmath208 $ ] is the least common multiple of @xmath209 and @xmath210 , and @xmath211 is their gcd .
the function @xmath212 above is the ratio @xmath213 where @xmath211 is square - free in our case .
note that @xmath214 is multiplicative outside of the primes @xmath13 and @xmath123 by theorem [ padicorbit ] , so @xmath215 and we need only to compute @xmath216 for @xmath17 prime in evaluating the sum above .
[ gthm ] let @xmath216 be as before where @xmath17 is a prime . then * @xmath217 * @xmath218 to prove ( ii ) , we note that theorem [ padicorbit ] implies that the numerator of @xmath216 is @xmath219 for @xmath112 .
thus it is the number of non - trivial solutions to a binary quadratic form with determinant @xmath118 ( the descartes form in ( [ descartes ] ) with two of the @xmath116 , @xmath89 set to @xmath118 ) , and so @xmath220 for all @xmath112 ( see @xcite , for example ) . in the case
@xmath122 , we observe that in both of the possible orbits of @xmath8 mod @xmath123 in figures [ mod31 ] and ( [ mod32 ] ) we have @xmath221 as desired .
part ( i ) follows from the structure of the orbit @xmath222 as observed in the proof of lemma [ betathm ] .
denote by @xmath223 the gcd of @xmath209 and @xmath210 ( we write just @xmath224 from now on and keep in mind that @xmath224 depends on @xmath209 and @xmath210 ) .
note that @xmath225}{\nu}\right)=\beta_i\left(\frac{d_i}{\nu}\right)\beta_j\left(\frac{d_j}{\nu}\right),\ ] ] where @xmath226 and @xmath227 are relatively prime , so we rewrite the expression in ( [ sum1 ] ) below : @xmath228\leq d'}\beta_i\left(\frac{d_i}{\nu}\right)\beta_j\left(\frac{d_j}{\nu}\right)g(\nu)\mu(d_i)\mu(d_j)\log d_i\log d_j.\ ] ] to compute the sum in ( [ desired ] ) , we use a similar argument to that in section [ theory ] .
we note that the inner sum in the expression above is equal to @xmath229 >
d'}\beta_i\left(\frac{d_i}{\nu}\right)\beta_j\left(\frac{d_j}{\nu}\right)g(\nu)\mu(d_i)\mu(d_j)\log d_i\log d_j\ ] ] where we assume the sum over @xmath208 > d'$ ] is insignificant by the conjectured randomness of the moebius function , and thus the sum over all @xmath209 and @xmath210 is a good heuristic for the sum in ( [ desired ] ) .
we compute the infinite sum in the following lemma and obtain the heuristic in conjecture [ tpconj ] .
[ infinitesum2calc ] let @xmath167 , @xmath93 , and @xmath230 be as before .
we have @xmath231 we introduce the function @xmath232 and note that @xmath233 which is a good heuristic for the sum in ( [ desired ] ) as @xmath0 tends to infinity .
the difficulty in computing this is the interaction of @xmath209 and @xmath210 in @xmath230 . to this end , write @xmath234 and @xmath235 where @xmath236 .
this gives us the following formula for @xmath237 .
@xmath238 where @xmath239 where @xmath240 @xmath241 @xmath242 we write @xmath243 @xmath244 which gives us @xmath245 note that the sums over @xmath224 and @xmath246 , as well as @xmath247 converge and are holomorphic .
we now compute the desired derivative .
write @xmath248 then we have @xmath249 since @xmath250 , we need not compute the partial derivative of @xmath137 . if @xmath251 in the expression for @xmath252 , we have that @xmath253 , since @xmath254 as in section [ theory ] .
this holds for the sums in ( [ desired ] ) over @xmath255 and @xmath256 where the @xmath15th and @xmath90th coordinates in our orbit are everywhere odd . the contribution to ( [ desired ] ) from the terms where @xmath135 or @xmath89 is
even is @xmath118 : [ zero ] let @xmath58 be an orbit of the apollonian group .
given that the @xmath15-th or @xmath90-th coordinate of each vector @xmath257 is even , we have @xmath258 recall that two of the coordinates of the vectors in @xmath58 are always even , and two are odd .
we write @xmath259 _ case 1 : only one of the @xmath15-th and @xmath90-th coordinates is odd throughout the orbit .
_ recall from lemma [ gthm ] that @xmath260 in the case that only one of the coordinates @xmath261 is even throughout the orbit , and @xmath262 if both coordinates are even throughout the orbit .
so , if only one of the coordinates @xmath261 is even throughout the orbit , we have @xmath263 recalling that @xmath264 for @xmath265 from lemma [ betathm ] , we have @xmath266 since lemma [ betathm ] implies either @xmath267 or @xmath268 in the first term , and either @xmath269 or @xmath270 in the second term in ( [ 2 m ] ) .
we now compute the expression in ( [ split ] ) in the case that both the @xmath15th and @xmath90th coordinate are even throughout the orbit .
_ case 2 : the @xmath15-th and @xmath90-th coordinates are both even throughout the orbit .
_ in this case , lemma [ gthm ] implies that @xmath271 and that @xmath272 for odd @xmath224 . also , we again have that @xmath264 for @xmath265 , so the first sum in ( [ split ] ) is @xmath273 to compute the second sum in ( [ split ] ) , we note that @xmath274 by lemma [ betathm ] and write @xmath275 where @xmath276 @xmath277 where @xmath278 note that , since @xmath279 , we have @xmath280 in this case , so @xmath281 on the other hand , @xmath282 combining ( [ first ] ) and ( [ second ] ) , we have that the expression in ( [ split ] ) is @xmath118 , as desired . therefore , we have that in the contributing terms of ( [ desired ] ) both @xmath135 and @xmath89 are odd .
this makes up two of the terms in ( [ desired ] ) , so we have that the sum we wish to compute is equal to @xmath283 it remains to compute @xmath284 in this case .
recall that @xmath251 in this case , and that @xmath264 for @xmath265 .
we have @xmath285 we write @xmath286 and get @xmath287 therefore our infinite sum is equal to @xmath288 as desired .
conjecture [ tpconj ] follows from our assumption that the moebius function is random and from lemma [ infinitesumcalc2 ] .
namely , we predict @xmath289 where @xmath290 as in conjecture [ tpconj ] . as with our heuristic for the number of prime curvatures less than @xmath0 in a packing
, this count does not depend on the packing @xmath4 .
[ kpnt ] and fig .
[ kpntc ] above show graphs of @xmath291 for @xmath292 and @xmath24 , and @xmath195 .
both of these graphs converge to @xmath293 as conjectured , even though they converge slower than in the case of counting circles of prime curvature in section [ theory ] .
in this section we present numerical evidence in support of conjecture [ lg ] , which predicts a local to global principle for the curvatures in a given integral acp .
since the apollonian group @xmath8 is small it is of infinite index in @xmath294 it is perhaps surprising that its orbit should eventually cover all of the integers outside of the local obstruction mod @xmath19 as specified in theorem [ padicorbit ] . proving this rigorously
, however , appears to be very difficult .
an analogous problem over @xmath295 would be to show that all large integers satisfying certain local conditions are represented by a general ternary quadratic form this analogy is realized by fixing one of the curvatures in descartes form and solving the problem for the resulting ternary form . while this problem has been recently resolved in general in @xcite and @xcite , even there
the local to global principle comes in a much more complicated form , relying on congruence obstructions specified in the spin double cover .
our conjecture , which therefore has the flavor of hilbert s 11th problem for an indefinite form , predicts a local to global principle of a more straightforward nature .
our computations suggest that this conjecture is true and we predict the value @xmath296 in the examples we check .
we consider the packings @xmath23 and @xmath24 introduced in section [ intro ] .
recall that @xmath23 corresponds to the orbit of @xmath8 acting on @xmath25 , and @xmath24 corresponds to the orbit of @xmath8 acting on @xmath26 . in order to explain the data we obtain in both cases , we use theorem [ padicorbit ] to determine the congruence classes ( mod @xmath19 ) in the given packing .
recall that the apollonian group @xmath8 is generated by the four generators @xmath12 in ( [ gens ] ) .
we can view an orbit of @xmath8 modulo @xmath19 as a finite graph @xmath297 in which each vertex corresponds to a distinct ( mod @xmath19 ) quadruple of curvatures , and two vertices @xmath7 and @xmath298 are joined by an edge iff @xmath299 for some @xmath126 .
recall from theorem [ padicorbit ] that for any orbit @xmath58 of the apollonian group , @xmath300 so the graph @xmath297 is completely determined by the structure of @xmath301 and @xmath302 .
there are only two possible orbits modulo @xmath123 , pictured in fig .
[ mod31 ] and fig .
[ mod32 ] .
there are many more possible orbits modulo @xmath303 , and we provide the graphs for these orbits in the case of @xmath23 and @xmath24 in fig .
[ mod8 ] .
note that each vertex in @xmath304 and @xmath305 is connected to its neighboring vertices via all of the generators @xmath12 .
therefore the curvatures of circles in a packing modulo @xmath19 are equally distributed among the coordinates of the vertices in @xmath297 .
combined with theorem [ padicorbit ] , this lets us compute the ratio of curvatures in a packing which fall into a specific congruence class modulo @xmath19 .
namely , let @xmath306 be the orbit mod @xmath19 corresponding to a given packing @xmath4 .
for @xmath307 let @xmath308 be the @xmath15th coordinate of @xmath309 .
we define @xmath310 as the proportion of coordinates in @xmath306 congruent to @xmath311 modulo @xmath19 .
that is , @xmath312 with this notation , a packing @xmath4 contains a circle of curvature congruent to @xmath311 modulo @xmath19 iff @xmath313 . given ( [ 2483 ] ) , we express @xmath314 as follows : @xmath315 the significance of @xmath314 in the case of any packing ( not only the two we consider ) is explained in the following lemma .
[ gammalemma ] let @xmath39 be as before , let @xmath36 be a circle in an integral appolonian packing p , and let @xmath37 be the curvature of @xmath36 .
then @xmath316 this follows from theorem [ padicorbit ] .
note that in general the orbits @xmath317 and @xmath318 have @xmath3 and , respectively , @xmath124 vertices in the corresponding finite graphs .
therefore @xmath297 always has @xmath319 vertices , and the ratio in ( [ multgamma ] ) is easily computed using this graph . with this in mind , we observe the following about the packing @xmath23 . [ bugob ] let @xmath320 denote the possible congruence classes of curvatures mod @xmath19 in the packing @xmath23 , and let @xmath321 be as in theorem [ oh ] .
then we have * @xmath322 , where @xmath323 * @xmath324 * @xmath325 * for @xmath326 , let @xmath327 denote @xmath0 mod @xmath19 . if @xmath328 then @xmath0 is a curvature in the packing @xmath23 .
part ( iv ) is an observation based solely on our computations using the algorithm described in section [ algorithm ]
these are illustrated in the histograms below .
the first three parts follow from computations combined with theorem [ oh ] and lemma [ gammalemma ] .
note that @xmath329 for this particular packing , one can hence express the local obstructions modulo @xmath330 rather than modulo @xmath19 .
whenever this is the case for an integral acp , we will find that there are eight congruence classes modulo @xmath19 in the curvatures of the circles .
graham et.al .
observe this in @xcite where they compute which integers less than @xmath331 are exceptions " for @xmath23 they find integers which satisfy these local conditions for @xmath23 modulo @xmath330 but do not occur as curvatures in the packing .
our data extends their findings ( we consider integers up to @xmath332 ) , and shows that all integers between @xmath331 and @xmath333 belonging to one of the congruence classes in part ( ii ) of observation [ bugob ] appear as curvatures in the packing @xmath23 .
the following histograms illustrate the distribution of the frequencies with which each integer in the given range satisfying the specified congruence condition occurs as a curvature in the packing @xmath23 .
the frequencies seem normally distributed , and the means of the distributions can be computed , as we do in ( [ meaneq ] ) .
the variance , however , is much more difficult to predict at this time explaining the behavior of the variance as we consider larger integers would shed more light on our local to global conjecture . note that there are no exceptions to the local to global principle in this range whenever @xmath118 is not a frequency represented in the histogram ( i.e. each integer occurs at least once ) .
there are several other frequencies not represented in the histograms for both @xmath23 and @xmath24 ( these show up as gaps in the graphs ) , and an explanation of this aspect would be interesting .
we do the same analysis for the packing @xmath24 . in this case
, we must consider much larger integers than in the case of @xmath23 in order to get comparable results .
this can be explained partially by the fact that the constant @xmath334 in kontorovich - oh s formula @xmath335 is much smaller for the packing @xmath24 than the packing @xmath23 since the initial four circles in @xmath24 are much larger than the initial four in @xmath23 ( see part ( i ) of lemmata [ bugob ] and [ coinob ] ) .
specifically , @xmath336 and @xmath323 .
however , our data suggests that the proposed local to global principle should hold for this packing as well .
[ coinob ] let @xmath337 denote the possible congruence classes mod @xmath19 in the packing @xmath24 , and let @xmath338 be as in theorem [ oh ] .
then we have * @xmath339 , where @xmath336 * @xmath340 * @xmath341 * for @xmath342 , let @xmath327 denote @xmath0 mod @xmath19 . if @xmath343 or @xmath344 , then @xmath0 is a curvature in the packing @xmath24 . again , note
that part ( iv ) is an observation based solely on our computations , while the first three parts in lemma [ coinob ] rely on lemma [ gammalemma ] and theorem [ oh ] .
the histograms below illustrate the distribution of the frequencies with which each integer in the given range satisfying the specified congruence condition occurs as a curvature in the packing @xmath24 .
note that , as with @xmath23 , the frequencies with which integers are represented in the packing seem to have a normal distribution .
however , since the mean of this distribution is much smaller for @xmath24 than for @xmath23 , we find that @xmath118 is often a frequency represented in the histograms , and so there are still some exceptions to the proposed local to global principle in the range we consider .
as we mentioned before , the mean in each of these histograms is easily computable : let @xmath36 denote a circle in an apollonian packing @xmath4 and let @xmath37 denote the curvature of @xmath36 .
let @xmath345 $ ] be an interval of length @xmath346 and let @xmath347 be an integer .
let @xmath348 be the number of times @xmath0 is a curvature of a circle in @xmath4
. for and integer @xmath349 , let @xmath350 then by lemma [ gammalemma ] , @xmath351 the equivalence of the two sums above is easy to observe one counts the same set of curvatures , but partitions them differently .
in particular , the expression in ( [ equivalentsums ] ) allows us to determine the mean of the distributions in the histograms above .
namely , denote by @xmath347 an integer in some interval @xmath345 $ ] of length @xmath346 .
let @xmath352 , and let @xmath353 denote the mean of the number of times @xmath354 mod @xmath19 is represented as a curvature in the packing @xmath4 .
note that there are precisely @xmath355 integers congruent to @xmath311 mod @xmath19 in the interval @xmath356 . combined with ( [ equivalentsums ] ) , this gives us @xmath357 this formula predicts the following values for the means in @xmath23 in the range @xmath358 ; and @xmath24 in the range @xmath359 : @xmath360 which coincides with the means observed in the histograms .
this clarifies why the mean is small for packings where the constant @xmath334 in the formula @xmath361 is small , and why one needs to consider very large integers to see that the local to global principle for such packings should hold .
this analysis can be carried out for any acp , and will likely yield similar results . in the direction of proving conjecture [ lg ]
, one might investigate how @xmath296 depends on the given packing
can it perhaps be expressed in terms of the constant @xmath334 in ( [ c ] ) ?
one might also ask how the variance of the distributions above depend on the packing , and how it changes with the size of the integers we consider answering this would give further insight into the local to global correspondence for curvatures in integer acp s .
we represent an acp by a tree of quadruples . fig .
[ tree ] shows the first two generations of the tree corresponding to @xmath24 . to generate all curvatures of magnitude less than @xmath0
, we use a lifo ( last - in - first - out ) stack to generate and prune this tree .
the algorithm is as follows : * push the root quadruple onto the stack . * until the stack is empty , perform an iterative process : * * pop a quadruple off of the stack and generate its children . * * for each child , if the new curvature created ( i.e. the maximum entry of the quadruple ) is less than @xmath0 , then push the child onto the stack . by pushing a quadruple onto the stack
only if its maximum entry is less than @xmath0 , we effectively prune the the tree .
since we know each quadruple has a larger maximum entry than its parent , we use step 2__b ) _ _ to avoid generating branches whose quadruples are known to have entries greater than @xmath0 .
although we use the concept of a tree to _ generate _ curvatures , we note that the entire tree structure is not necessary to _ store _ such curvatures .
instead , we store the curvatures in a one - dimensional array of @xmath0 elements , all initialized to zero . the @xmath15th element of the array contains the number of curvatures with magnitude @xmath15 . for instance , the @xmath19th element of the array for @xmath24 is equal to @xmath362 , while the @xmath363th element is equal to @xmath118 , since there are no curvatures equal to 25 in @xmath24 .
we use these arrays to generate the histograms in section [ locglobal ] . due to matlab s memory constraints
, we limit our matlab arrays to @xmath364 entries .
so , to check for exceptions in the entire range of @xmath365 , we check each of the intervals @xmath358 , @xmath366 individually .
we have chosen to display the interval @xmath358 in our figures in section [ locglobal ] . to count primes less than @xmath0
, we simply increment a sum whenever a prime curvature is produced . to count kissing primes less than @xmath0
, we increment a sum whenever a prime curvature is produced _ and _ some other member of the curvature s quadruple is prime . 1 j. bourgain , a. gamburd , p. sarnak , _ sieving , expanders , and sum product _ , http://www.math.princeton.edu/sarnak ( 2007 ) j. bourgain , e. fuchs , _ on the density of integers occurring as curvatures in integer apollonian circle packings
_ , in preparation ( 2010 ) j.w.s .
cassels , _ rational quadratic forms _ , dover publications , inc . , mineola , ny ( 1978 ) j. cogdell , _ on sums of three squares_. j. thor
. nombres bordeaux , v. 15 ( 2003 ) h.s.m .
coxeter , _ an absolute property of four mutually tangent circles _ , _ non - euclidean geometries , janos bolyai memorial volume _ ( eds .
a. prkopa and e. molnr ) , kluwer academic pub .
( 2005 ) w. duke , r. schulze - pillot , _ representation of integers by positive ternary quadratic forms and equidistribution of lattice points on ellipsoids , _ invent .
99 , pp.49 - 57 ( 1990 ) e. fuchs , _ arithmetic properties of apollonian circle packings _ , ph .
d. thesis , princeton university ( 2010 ) e. fuchs , _
strong approximation in the apollonian group _
, in preparation ( 2010 ) r.l .
graham , j.c .
lagarias , c.l .
mallows , a.r .
wilks , c.h .
yan , _ apollonian circle packings : number theory _ , journal of number theory , 100 , pp.1 - 45 ( 2003 ) k.e .
hirst , _ the apollonian packing of circles _ , proc .
usa , 29 , pp .
378 - 384 ( 1943 ) a. kontorovich , h. oh , _
apollonian circle packings and closed horospheres on hyperbolic @xmath123-manifolds _ , preprint , http://arxiv.org/pdf/0811.2236 ( 2008 ) k. sanden , _ integral apollonian circle packings : a prime number conjecture _ , senior thesis , princeton university .
ksanden / thesis.htm ( 2009 ) p. sarnak , _ letter to lagarias _ , http://www.math.princeton.edu/sarnak ( 2007 ) p. sarnak , _ maa lecture notes _ , http://www.math.princeton.edu/sarnak ( 2009 ) r. sedgewick , k. wayne , _ introduction to programming in java _ , http://www.cs.princeton.edu/introcs/home ( 2007 ) | bounded apollonian circle packings ( acp s ) are constructed by repeatedly inscribing circles into the triangular interstices of a configuration of four mutually tangent circles , one of which is internally tangent to the other three .
if the original four circles have integer curvature , all of the circles in the packing will have integer curvature as well . in @xcite
, sarnak proves that there are infinitely many circles of prime curvature and infinitely many pairs of tangent circles of prime curvature in a primitive integral acp . in this paper , we give a heuristic backed up by numerical data for the number of circles of prime curvature less than @xmath0 , and the number of kissing primes , " or _
pairs _ of circles of prime curvature less than @xmath0 in a primitive integral acp .
we also provide experimental evidence towards a local to global principle for the curvatures in a primitive integral acps . |
the dynamics of particles of finite inertia suspended in turbulent flows is a fundamental problem with applications ranging from industrial processes ( e.g. spray combustion engines ) to geophysical flows ( e.g. , atmospheric clouds ) .
the interaction between turbulence and particles has been studied to understand rain initiation in warm terrestrial clouds ( e.g. , pinsky & khain 1997 ; falkovich , fouxon , & stepanov 2002 ; shaw 2003 ) , cloud evolution in the atmospheres of planets , cool stars and brown dwarfs ( e.g. , rossow 1978 ; pruppacher & klett 1997 ; freytag et al .
2010 ; helling et al .
2011 ) , collisions and growth of dust particles in protoplanetory disks ( e.g. , dullemond & dominik 2005 ; zsom et al . 2010 , 2011 ;
birnstiel et al .
2011 ) and in the interstellar medium ( e.g. , ormel et al .
2009 ) .
the evolution of the particle size depends on the particle collision rate which may be significantly enhanced by turbulent motions in the carrier flow , as illustrated by recent numerical and theoretical advances in this field ( e.g. , wang et al . 2000 ;
zhou et al . 2001 ; falkovich et al . 2002 ; zaichik & alipchenkov 2003 , 2009 ; zaichik et al .
2003 , 2006 ; wilkinson et al .
2005 , 2006 ; falkovich & pumir 2007 ; gustavsson & mehlig 2011 ; gustavsson et al . 2012 ) .
an accurate evaluation of the collision rates requires understanding the effects of two interesting phenomena : the preferential concentration or clustering of inertial particles ( e.g. , maxey 1987 and squires & eaton 1991 ) and the turbulence - induced collision velocity . in this work
, we will focus on the statistics of turbulence - induced relative velocities , and briefly discuss the role of turbulent clustering on the collision rate ( see pan et al .
2011 for a detailed discussion of turbulent clustering in the context of planetesimal formation ) .
we restrict our discussion to the relative velocity between same - size particles , usually referred to as the monodisperse case , and will address the general bidisperse case ( collisions between particles of different sizes ) in a follow - up paper .
the main motivation of our study is to improve the modeling of the evolution of dust particles in protoplanetary disks , which sets the stage for the formation of planetesimals , the likely precursors to fully - fledged planets .
for example , the planetesimal formation model by johansen et al .
( 2007 , 2009 , 2011 ) requires particle growth up to decimeter to meter size , in order to achieve good frictional coupling to the disk rotation and hence the maximum clustering effect by the streaming instability .
cuzzi et al .
( 2008 , 2010 ) and chambers ( 2010 ) proposed an alternative model of planetesimal formation based on the strong turbulent clustering of chondrule - size particles .
other studies ( e.g. lee et al .
2010 ) focus on the possibility that small particles settle to the disk midplane , where gravitational instability can result in planetesimal formation ( e.g. goldreich & ward 1973 ; youdin 2011 ) , despite the turbulence stirring caused by the kelvin - helmholtz instability induced by the vertical settling of the particles ( e.g. weidenschilling 1980 ; chiang 2008 ) .
the evolution of the size distribution of dust particles is controlled by collisions .
small particles tend to stick together when colliding , and thus their size grows by coagulation . as the size increases , the particles become less sticky ( blum & wurm 2010 ) , and , depending on the collision velocity , the collisions may result in bouncing or fragmentation .
a detailed summary of experimental results for the dependence of the collision outcome on the particle properties ( such as the particle size and porosity ) and on the collision velocity can be found in guttler et al .
the coagulation , bouncing and fragmentation processes may lead to a quasi - equilibrium distribution of particle sizes ( e.g. , birnstiel et al .
2011 ; zsom et al .
2010 , 2011 ) . due to the dependence of the collision outcome on the collision velocity ,
an accurate evaluation of the turbulence - induced relative velocity is important for modeling the size distribution of dust particles .
saffman and turner ( 1956 ) studied the relative velocity in the limit of small particles with the particle friction or stopping time , @xmath0 , much smaller than the kolmogorov timescale , @xmath10 , of the turbulent flow .
this limit , known as the saffman - turner limit , is usually expressed as @xmath11 , where the stokes number is defined as @xmath12 .
saffman and turner ( 1956 ) predicted that , at a given distance , @xmath13 , the relative velocity of identical particles is independent of @xmath1 , and , at a given @xmath1 , it scales linearly with @xmath13 for small @xmath13 . in the opposite limit of large particles with @xmath0 larger than the largest timescale of the turbulent flow , abrahamson ( 1975 ) showed that the relative velocity scales with the friction time as @xmath8 .
a variety of models have been developed to bridge the two limits and to predict the relative velocity for particles of any size , i.e. , with @xmath0 covering the entire scale range of the carrier flow ( williams & crane 1983 , yuu 1984 , kruis & kusters 1997 , & alipchenkov 2003 , zaichik et al .
2006 , ayala et al .
2008 ) . among these models ,
the formulation of zaichik and collaborators is particularly impressive , as it examines turbulent clustering and turbulence - induced relative velocity simultaneously .
the model prediction for the relative velocity agrees well with simulation results at low resolutions .
however , the model lacks a transparent physical picture . pan & padoan ( 2010 ) developed a new model for the relative velocity of inertial particles of any size that provides an insightful physical picture of the problem .
their formulation illustrates that the relative velocity of identical particles is determined by the memory of the flow velocity difference along their trajectories in the past .
the model also shows that the separation of inertial particle pairs backward in time plays an important role in their relative velocity .
the model prediction can correctly reproduce the scaling behaviors of the relative speed in the extreme limits of small and large particles , and was found to successfully match the simulation data of wang et al .
falkovich et al .
( 2002 ) discovered an interesting effect , named the sling effect , which provides an important contribution to the collision rate .
the basic physical picture of this effect is that inertial particles may be shot out of fluid streamlines with high curvature , causing their trajectories to cross with those of other particles ( see fig . 1 of falkovich & pumir 2007 ) .
in particular , in flow regions with large negative velocity gradients , fast particles can catch up with the slower ones from behind .
the trajectory crossing causes the particle velocity to be multi - valued at a given point .
this gives rise to folds , usually referred to as caustics , in the momentum - position phase space of the particles ( wilkinson et al .
2006 ; see fig . 1 of gustavsson & mehlig 2011 for a clear illustration ) .
for small particles with @xmath11 , the sling events correspond to high - order statistics of the flow velocity gradient , and the effect is not reflected in the prediction of saffman and turner ( 1956 ) .
the formulations of falkovich et al .
( 2002 ) and gustavsson & mehlig ( 2011 ) for the collision kernel of @xmath14 particles consist of two contributions . following wilkinson et al .
( 2006 ) , we name them as continuous and caustic contributions , corresponding to two types of particle pairs with low and high relative velocities , respectively . in the continuous contribution ,
the relative speed follows the saffman - turner prediction and decreases linearly with the particle distance , @xmath13 .
the contribution is amplified by turbulent clustering . however , the scaling exponents of the relative speed and the degree of clustering suggest that the continuous contribution approaches zero in the limit @xmath15 , as pointed out by hubbard ( 2012 ) .
the caustic contribution to the collision kernel per unit cross section was predicted to be independent of the particle size or distance , @xmath13 , and is thus expected to dominate at sufficiently small @xmath13 .
the effect of slings or caustics causes a rapid rise in the collision rate as @xmath1 approaches 1 , which has been proposed to be responsible for the initiation of rain shower in terrestrial clouds ( wilkinson et al .
2006 ) . applying this effect to dust particle collisions in protoplanetary disks
, one may expect that the collision rate greatly accelerates as the particle grows past sub - mm to mm size , corresponding to @xmath16 for typical protoplanetary turbulence conditions .
the recent developments mentioned above have not been considered in coagulation models for dust particles in circumstellar disks .
we will show that the general formulation of the collision kernel commonly used in the astrophysical literature for dust coagulation is inaccurate .
in particular , the dust coagulation models usually adopt collision velocities from the work of volk et al .
( 1980 ) and its later extensions ( e.g. , markiewicz , mizuno & volk 1991 , cuzzi and hogan 2003 , and ormel & cuzzi 2007 ) , which have a number of limitations .
pan & padoan ( 2010 ) pointed out a weakness in the physical picture of these models . roughly speaking , these models assume that the velocities of two particles induced by turbulent eddies with turnover time significantly smaller ( larger ) than @xmath0 are independent ( correlated ) .
as shown by pan & padoan ( 2010 ) , whether the particle velocities contributed by turbulent eddies of a given size are correlated or not also depends on how the eddy size compares to the separation of the particles at the time the eddies were encountered .
therefore , the eddy turnover time is not the only factor that determines the degree of correlation .
the role of the particle separation relative to the eddy size is not captured by the approach of volk et al .
we also find that the model of volk et al .
overestimates the relative velocity by a factor of 2 for particles with @xmath0 on the order of the large eddy turnover time of the turbulent flow . in this paper , we conduct a @xmath17 numerical simulation to study inertial particle dynamics in a hydrodynamic turbulent flow . in the simulated flow ,
we evolve inertial particles in an extended size range , with @xmath0 covering the entire scale range of the turbulent flow . to our knowledge
, such a systematic simulation of a significant resolution has not been previously conducted in the astrophysical literature . using the simulation data , we first test the model prediction of pan & padoan ( 2010 ) for the rms relative velocity of inertial particles as a function of @xmath1 , and validate the physical picture revealed by the model .
we then apply the pan & padoan ( 2010 ) model to interpret the probability distribution function ( pdf ) of the relative velocity .
the pdf study is motivated by the importance of the pdf of the collision speed in modeling dust particle collisions ( windmark et al .
2012 , garaud et al . 2013 ) , which determines the fractions of collisions leading to sticking , bouncing or fragmentation .
the relative velocity pdf of inertial particles has been shown to be highly non - gaussian by numerical , experimental and theoretical studies ( e.g. , sundaram and collins 1997 , wang et al .
2000 , gustavsson et al . 2008 ,
bec et al .
2009 , de jong et al .
2010 , gustavsson & mehlig 2011 , hubbard 2012 ) .
our simulation further confirms high non - gaussianity , which should be incorporated into coagulation models for dust particles in protoplanetary disks .
we will also investigate the particle collision kernel as a function of @xmath1 .
due to the computational cost , the number of particles included in our simulation is limited and only allows to accurately measure the relative velocity statistics at significant particle distances .
the distance range explored is @xmath18 where @xmath19 is the kolmogorov scale of the simulated flow .
this raises the question concerning the direct applicability of our measured statistics to dust particle collisions .
the size of dust particles is many orders of magnitudes smaller than the kolmogorov scale ( @xmath20 km ) in protoplanetary turbulence .
therefore , dust particles should be viewed as nearly point particles , and one is required to examine the @xmath15 limit in order to model their collisions ( hubbard 2012 , 2013 ) .
this suggests that the relative velocity measured in our simulation at @xmath21 would be distinct from the collision speed of dust particles , unless the statistics have already converged at @xmath22 .
we find that the measured relative velocity statistics for particles with @xmath23 actually converge at @xmath24 , and are thus directly applicable for the collision velocity of dust particles . on the other hand , for small to intermediate particles with @xmath25 ,
the measured statistics show an @xmath13-dependence in the @xmath13 range explored in this study . for these particles , an appropriate extrapolation to the @xmath15 limit
is needed for applications to dust particle collisions . in the current paper , we focus on understanding the fundamental physics of turbulence - induced relative velocity at finite distances ( @xmath26 ) .
our theoretical and numerical results provide an important step toward the final goal of estimating the dust particle collision velocity at @xmath27 .
to underhand the @xmath15 limit , we make an initial and preliminary attempt to separate particle pairs into two types , i.e. , continuous and caustic types , which show different scalings with @xmath13 . in particular , we evaluate the contributions of two types of pairs to the collision kernel and examine their behaviors as @xmath28 . a systematical study for the @xmath28 limit
is deferred to a future work . in this work
, we will consider the particle dynamics only in statistically homogeneous and isotropic turbulence .
this is clearly an idealized situation , considering various complexities in protoplanetary disks .
for example , the disk rotation induces large - scale anisotropy , which may have significant effects on the prediction for particles with friction time close to the rotation period .
nevertheless , the idealized problem is a very useful tool to understand the fundamental physics .
we also neglect the vertical settling and radial drift .
these processes do not directly affect the relative velocity between identical particles , although they may provide important contributions for particles of different sizes that we address in a follow - up work .
the paper is organized as follows . in 2 , we present a simple model for the rms velocity of a single particle , which provides an illustration for our formulation of the particle relative velocity . in 3 , we introduce the model of pan & padoan ( 2010 ) for the relative velocity of nearby particles .
our simulation setup and the statistical properties of the simulated turbulent flow are described in 4 .
5 presents simulation results for the one - particle rms velocity . in 6
, we test the model prediction of pan & padoan ( 2010 ) for the rms relative velocity , and discuss in details the probability distribution of the relative velocity as a function of the particle inertia . in 7
, we evaluate the collision kernel .
the conclusions of our study are summarized in 8 .
the dynamics of inertial particles depends crucially on its friction or stopping timescale , @xmath0 . to evaluate of the friction timescale
, we first need to compare the particle size , @xmath29 , with the mean free path of the gas particles in the carrier flow .
if the particle size is larger than the mean free path , the friction timescale is given by the stokes law @xmath30 , where @xmath31 ( @xmath6 g @xmath32 ) is the density of the dust material , @xmath33 is the gas density , and @xmath34 is the kinematic viscosity of the flow . on the other hand , if @xmath29 is smaller than the gas mean free path , the particle is in the epstein regime and @xmath35 , where @xmath36 is the sound speed in the flow .
for example , for a typical gas density in protoplanetary discs , @xmath37 g @xmath32 , at 1 au , the mean free path of the gas particles is @xmath38 cm , and thus particles with @xmath29 larger ( smaller ) than 1 cm are in the stokes ( epstein ) regime . the velocity , @xmath39 , of an inertial particle suspended in a turbulent velocity field , @xmath40 , obeys the equation , @xmath41 where @xmath42 is the position of the particle at time @xmath43 , and @xmath44 corresponds to the flow velocity seen " by the particle .
( [ particlemomentum ] ) has a formal solution , @xmath45 where it is assumed that @xmath46 and the particle has already lost the memory of its initial velocity at @xmath47 .
the formal solution indicates that the velocity of an inertial particle is determined by the memory of the flow velocity along its trajectory within a timescale of @xmath48 in the past .
although the aim of the present work is the relative velocity of inertial particle pairs , we start with a discussion of the single - particle ( or `` 1-particle '' ) velocity induced by turbulent motions .
we provide a simple model for the 1-particle rms velocity as a function of @xmath0 .
the derivation of this model helps to illustrate our formulation for the relative velocity between two nearby particles .
the 1-particle rms velocity can be calculated using the formal solution of eq .
( [ formalsolution ] ) .
we assume the turbulent flow is statistically stationary , and the particle statistics eventually relax to a steady state .
we consider a time when the steady state is already reached and denote this time as time 0 .
using eq .
( [ formalsolution ] ) at @xmath49 , we have , @xmath50 where @xmath51 denotes the ensemble average and @xmath52 is the temporal correlation tensor of the flow velocity along the trajectory , @xmath53 , of the inertial particle . the subscript
t " stands for trajectory " .
we changed the lower integration limit ( @xmath54 ) in eq .
( [ formalsolution ] ) to @xmath55 , based on the assumption that the particle dynamics is fully relaxed at time 0 ( i.e. , @xmath56 ) . with statistical stationarity and isotropy ,
the trajectory correlation tensor can be written as @xmath57 , where @xmath58 is the 1d rms velocity of the turbulent flow and the correlation coefficient @xmath59 is a function of the time lag only .
the subscript 1 " is used to indicate that the correlation is along the trajectory of _ one _ particle .
the correlation coefficient , @xmath59 , is unknown , and a common assumption is to approximate it with the lagrangian correlation function , @xmath60 , of tracer particles ( or fluid elements ) , which has been extensively studied . the assumption is likely valid for small particles , but can not be justified for large particles on a theoretical basis .
we will validate the assumption _ a posteriori _ using simulation results .
the simplest choice for @xmath60 is an exponential function , @xmath61 , where @xmath62 is the time lag and @xmath4 the lagrangian correlation timescale . setting @xmath63 in eq .
( [ 1particlevelocity ] ) , we have @xmath64 , where the 1d rms particle velocity , @xmath65 , is given by , @xmath66 this result shows that the particle rms velocity approaches the flow velocity for @xmath67 and decreases as @xmath68 for @xmath9 ( e.g. , abrahamson 1975 ) . in the large particle limit , @xmath9 , the action of even the largest turbulent eddies on the particle would appear to be random kicks when viewed on a timescale of @xmath0 . in that case , eq .
( [ particlemomentum ] ) is essentially a langevin equation , and the particle motions are similar to brownian motions .
the @xmath8 scaling corresponds to an equilibrium " between the velocity of these particles and the turbulent motions of the flow .
numerical simulations have shown that the lagrangian correlation function , @xmath69 , is better fit by a bi - exponential form ( e.g. , sawford 1991 ) .
a single - exponential form does not reflect the smooth part of the correlation function for @xmath70 smaller than the taylor micro timescale , @xmath71 .
the taylor timescale is defined as @xmath72 , where @xmath73 is the rms acceleration of the turbulent velocity field .
the bi - exponential form for @xmath69 is , @xmath74 - \big(1-\sqrt{1 - 2z^2}\big ) \times \notag\\ \exp \left [ { - \frac{2 |\delta \tau| } { \big(1 - \sqrt{1 - 2z^2 } \big ) t_{\rm l } } } \right ] { \bbigg@{3}}\ } , \hspace{1.8 cm } \label{biexponential}\end{gathered}\ ] ] where the parameter @xmath75 is defined as @xmath76 . from the above equation , it is easy to show that @xmath77 , and the bi - exponential function is smooth , @xmath78 , at @xmath79 . in the limit @xmath80 , eq .
( [ biexponential ] ) reduces to the single exponential with a timescale of @xmath4 . adopting the bi - exponential form , eq .
( [ biexponential ] ) , for the trajectory correlation coefficient , @xmath81 , we find that the 1-particle rms velocity is given by , @xmath82 where @xmath83 is defined as @xmath84 . in the limits
@xmath85 and @xmath86 , eq . ( [ 1particlevelocitybiexp ] ) has the same behavior as eq .
( [ 1particlevelocityexp ] ) from the single exponential correlation .
in fact , the two predictions , eqs .
( [ 1particlevelocityexp ] ) and ( [ 1particlevelocitybiexp ] ) , are close to each other at all values of @xmath83 , differing only by a few percent at @xmath87 .
this suggests that , for a given correlation timescale , @xmath4@xmath88 ) , the integral in eq .
( [ 1particlevelocity ] ) is insensitive to the exact function form of @xmath89 .
we will measure @xmath75 and @xmath4 using lagrangian tracer particles in our simulated turbulent flow , and test the predictions , eqs .
( [ 1particlevelocityexp ] ) and ( [ 1particlevelocitybiexp ] ) , against the simulation data .
we briefly review the 2-point eulerian statistics of the velocity field in fully - developed turbulence , which is crucial to understand the relative velocity of two inertial particles .
we consider the structure tensor of a turbulent flow , defined as @xmath90 where @xmath91 is the velocity increment across a separation @xmath92 .
the statistics of @xmath93 is independent of @xmath94 and @xmath43 from the assumption of homogeneity and stationarity . with statistical isotropy
, the velocity structure tensor takes the form ( e.g. , monin and yaglom 1975 ) , @xmath95\frac{\ell_i \ell_j } { \ell^2 } \label{flowstructure}\ ] ] where the longitudinal and transverse structure functions , @xmath96 and @xmath97 , are functions of the amplitude , @xmath98 , but not the direction ( @xmath99 ) of @xmath92 . from eq .
( [ flowstructure ] ) , we see @xmath100 , where @xmath101 is the radial component of @xmath102 .
similarly , @xmath97 can be written as @xmath103 with @xmath104 being one of the two components of @xmath105 on the tangential / transverse plane perpendicular to @xmath92 .
the statistical isotropy indicates that the probability distribution of @xmath104 is invariant under any rotation about the direction @xmath99 . in incompressible turbulence , which is approximately the case for gas flows in protoplanetary disks
, we have the relation @xmath106 , which can be derived from the incompressibility condition : @xmath107 ( monin and yaglom 1975 ) .
the structure functions exhibit different scaling behaviors in different scale ranges .
there are three subranges divided by two length scales , the kolmogorov length scale , @xmath19 , and the integral length scale @xmath108 .
the kolmogorov scale , @xmath19 , is defined as @xmath109 , where @xmath34 and @xmath110 are , respectively , the kinematic viscosity and the average energy dissipation rate in the turbulent flow .
it essentially corresponds to the size of the smallest eddies .
scales below @xmath19 are called the viscous or dissipation range , where the velocity field is laminar and differentiable due to the smoothing effect of the viscosity . in the dissipation range ,
the velocity difference scales linearly with @xmath98 , and the longitudinal structure function is @xmath111 .
@xmath97 is twice larger , i.e. , @xmath112 , as required by the incompressibility constraint . in the inertial range , @xmath113
, @xmath96 follows the kolmogorov scaling , @xmath114 , where @xmath115 is the kolmogorov constant .
the typical value of @xmath115 is @xmath116 .
the incompressibility condition gives @xmath117 in the inertial range .
the integral scale , @xmath108 , is essentially the correlation length of the velocity field . at @xmath118
, the velocity field is uncorrelated , and both @xmath119 and @xmath120 are constant and equal to @xmath121 with @xmath58 the 1d rms velocity of the flow .
to bridge the scalings of @xmath96 in the three scale ranges , we adopt a connecting formula ( zaichik et al .
2006 ) , @xmath122^{4/3 } \times } \hspace{2 cm } \notag \\ \hspace{3 cm } { \displaystyle \left [ \frac{(\ell/ \eta)^4}{(\ell/ \eta)^4 + ( 2 u'^2 /c_{\rm k } u_\eta^2)^{6 } } \right]^{1/6 } } , \label{sll}\end{gathered}\ ] ] where @xmath123 is the kolmogorov velocity scale defined as @xmath124 . with eq .
( [ sll ] ) for @xmath96 , we can obtain @xmath97 using the incompressibility condition ( see above ) .
alternatively , one may adopt a separate connecting formula for @xmath97 ( see 4.3 ) .
the goal of this work is to understand the relative velocity of two nearby inertial particles . the relative velocity across a distance , @xmath13 , equal to the sum of the particle radii corresponds to the speed at which the two particles collide ( saffman and turner 1956 ) .
as mentioned earlier , dust particles in protoplanetary disks are nearly point - like , as their size is much smaller than the kolmogorov length scale , @xmath19 .
the collision speed of dust particles is therefore the relative velocity at @xmath27 . in this paper , we focus on the relative speed at finite distances , @xmath125 , and the @xmath27 limit will be examined systematically in a future work .
we label two particles coming together with superscripts @xmath126 and @xmath127 .
for example , we denote their positions as @xmath128 and @xmath129 , and their velocities as @xmath130 and @xmath131 ( see fig . 1 for illustration ) .
when the superscripts @xmath126 and @xmath127 are not used , the discussion is general and not referring to a specific particle . at a given time @xmath43 , we consider the relative velocity @xmath132 , of particle pairs at a given separation , @xmath133 , which corresponds to a constraint @xmath134 for the particle positions .
we first present a theoretical model for the second - order moment of @xmath135 , and then use simulations to explore its full statistics including the probability distribution function ( pdf ) .
similar to the structure tensor of the flow velocity , we characterize the second - order statistics of the particle relative velocity @xmath135 by a structure tensor , @xmath136 which was referred to as the particle velocity structure tensor by pan and padoan ( 2010 ) . here
@xmath137 denotes the average over all particle pairs at a separation of @xmath133 .
once the particle dynamics is fully relaxed , the particle velocity is expected to possess the same statistical symmetries as the flow , including stationarity , homogeneity and isotropy . with these symmetries
, @xmath138 can be written in a similar form as the structure tensor of the flow ( eq . [ flowstructure ] ) , @xmath139 where @xmath140 and @xmath141 are the variances of the radial / longitudinal component , @xmath142 ( @xmath143 ) , and a tangential / transverse component , @xmath144 , of the relative velocity , respectively . for particle collisions , we are interested in @xmath145 at small distances only , with @xmath13 in the dissipation range of the flow . under the assumption of isotropy , the tangential component , @xmath144 ,
is expected to be statistically invariant for any rotations about the axis @xmath133 .
we can thus measure the statistics of the tangential relative velocity by projecting @xmath135 into an arbitrary direction on the plane perpendicular to @xmath133 . in the rest of this section
, we consider theoretical models for the variances , @xmath140 and @xmath141 , of the relative velocity , which can be computed from the particle structure tensor @xmath146 .
for example , @xmath147 .
the 3d variance , @xmath148 , of the relative velocity , is given by the contraction of the tensor , i.e. , @xmath149 .
we point out that the relative velocity variances can not be directly applied to estimate the collision kernel , which depends on @xmath150 or @xmath151 ( see 7 ) .
one may use @xmath140 to approximately estimate the collision rate by a conversion to @xmath150 under an assumption for the pdf shape of @xmath142 ( e.g. , wang et al .
2000 ) .
furthermore , the 3d variance @xmath148 does not accurately reflect the average collisional energy for each collision .
as pointed out by hubbard ( 2012 ) , a collision - rate weighting is needed to evaluate the average collisional energy per collision .
in particular , @xmath148 is defined as the variance over all particle pairs at a distance , @xmath13 .
but not all the pairs may give a significant contribution to the collision rate in the @xmath27 limit ( see 7.2 ) , and in that case @xmath148 does not provide a reliable estimate for the average collision energy for those pairs that dominate the collision rate at @xmath15 . despite these limitations in the practical use of the overall rms ( or variance ) of the relative speed , its theoretical modeling is an important step toward understanding the fundamental physics . as mentioned earlier , we focus on the monodisperse case with equal - size particles .
we first consider small particles in the saffman - turner limit ( hereafter the s - t limit ) . in this limit
, the friction timescale , @xmath0 , is much smaller than the kolmogorov timescale , @xmath152 , of the carried flow , which is defined as @xmath153 .
the kolmogorov timescale is the smallest timescale in a turbulent flow , corresponding to the turnover time of the smallest eddies .
therefore , the velocity of particles with @xmath154 can be approximated by a taylor expansion of eq .
( [ particlemomentum ] ) , @xmath155 , where @xmath156 is the acceleration of the local fluid element . applying the approximation to both particles @xmath126 and @xmath127 , we have @xmath157 , where @xmath158 ( @xmath159 ) and @xmath160 ( @xmath161 ) are the flow velocity and acceleration at the positions of particles ( 1 ) and ( 2 ) , respectively .
saffman and turner ( 1956 ) assumed that the correlation coefficient of the flow accelerations , @xmath162 , and @xmath163 , across a small distance , @xmath13 , is unity , which is equivalent to assuming @xmath164 .
the acceleration terms then cancel out for identical particles , and the particle structure tensor , @xmath138 , is simply equal to the flow structure tensor , @xmath165 , defined in eq .
( [ flowstructure ] ) . using the flow structure functions @xmath96 and @xmath97 at @xmath166 in incompressible turbulence
, we have the saffman - turner formula , @xmath167 for identical particles with @xmath11 .
the equation shows that in the s - t limit the relative speed is caused by the flow velocity difference across the particle separation . the effect is usually referred to as the shear contribution . from eq .
( [ saffmanturner ] ) , the 3d variance of the relative velocity is given by @xmath168 .
the s - t formula predicts that the tangential variance of the relative velocity , @xmath141 , is twice larger than that in the radial direction , @xmath140 .
( [ saffmanturner ] ) also indicates a constant relative speed at a given separation , @xmath13 , and a linear scaling with @xmath13 at a given @xmath11 .
the accuracy of the saffman - turner formula for the small particle limit has been questioned , as it neglects the effect of slings and caustic formation ( e.g. , falkovich et al .
2002 , wilkinson et al .
we will test the s - t prediction against our simulation data . in the s - t limit ,
the particle memory is short and the relative speed is determined largely by the local flow velocity at small scales .
the memory effect becomes more important for larger particles with @xmath169 ( see 3.2 ) .
we next consider the other extreme limit , i.e. , large particles with @xmath0 much larger than the lagrangian correlation time , @xmath4 , of the flow .
as discussed in 2 , the motions of these particles are similar to brownian motions , and the velocities of any two particles are statistically independent .
this is because the velocity of a large particle has a significant contribution from its memory of the flow velocity long time ago , and the flow velocities seen " by the two particles at that time were uncorrelated because the particle separation was likely larger than the flow integral length scale , @xmath108 . with the independence of @xmath170 and @xmath171 , the particle structure tensor defined in eq .
( [ particlestructure ] ) can be written as @xmath172 \delta_{ij}$ ] , where @xmath173 and @xmath174 are the ( 1d ) rms velocities of particles ( 1 ) and ( 2 ) , respectively .
as shown in 2 , for particles with @xmath9 , the rms velocity is given by @xmath175 .
we therefore have ( e.g. , abrahamson 1975 ) , @xmath176 for identical particles with @xmath9 .
the equation suggests that the rms relative speed decreases with @xmath1 as @xmath177 .
the physical picture for the large particle limit is clear , and eq .
( [ largelimit ] ) is thus robust . in between the two extreme limits
are particles in the inertial range , i.e. , particles with friction timescale @xmath178 , corresponding to inertial - range scales in the turbulent flow . unlike the two extreme limits where the velocities of two nearby particles are either highly correlated ( small particles ) or essentially independent ( large particles ) , the velocity correlation of nearby inertial - range particles is at an intermediate level .
we will show that a key physics for the relative velocity of these particles is their memory of the flow velocity difference in the past and the separation of the particle pair backward in time . as mentioned in the introduction ,
a variety of models for the particle relative velocity covering the whole range of particle sizes have been developed ( e.g. , volk et al . 1980 ,
ormel & cuzzi 2007 , zaichik & alipchenkov 2003 , zaichik et al .
2006 , and pan & padoan 2010 ) . the models listed here all predict a @xmath179 scaling for inertial - range particles in turbulent flows with an extended inertial range .
the @xmath179 scaling may be obtained by a simple scale - invariant assumption for inertial - range particles ( e.g. , hubbard 2012 ) , which we argue , however , does not provide a sufficient physical picture to understand the full statistics , e.g. , the pdf shape , of the relative velocity .
the models of zaichik and collaborators and pan and padoan ( 2010 ) can reproduce both the s - t limit ( eq . ( [ saffmanturner ] ) ) and the large particle limit ( eq . ( [ largelimit ] ) ) .
we will focus on the model of pan and padoan ( 2010 ) , which provides a clearer physical picture than that of zaichik et al .
the physical differences between various models have been summarized in pan and padoan ( 2010 ) .
we review the formulation and the physical picture of the model by pan & padoan ( 2010 ; hereafter pp10 ) for the relative velocity of identical particles . the pp10 model aimed at predicting the variance or rms of the relative velocity . as mentioned earlier , although the rms relative velocity does not directly enter the collision kernel or the average collisional energy per collision , its theoretical modeling is essential for understanding the underlying physics .
for example , the physical picture revealed by the pp10 model is very successful in the interpretation of the probability distribution of the relative velocity ( 6.2 ) , which , in turn , is helpful for the evaluation of the collision kernel ( 7 ) .
the main idea of the pp10 model was to compute the particle velocity structure tensor , @xmath180 , using the formal solution ( eq . ( [ formalsolution ] ) ) for the particle velocity .
applying eq .
( [ formalsolution ] ) to the velocities of particles ( 1 ) and ( 2 ) at @xmath49 , we have , @xmath181 \exp \left ( \frac { \tau}{\tau_{\rm p } } \right ) d\tau , \label{velocitydifference}\ ] ] where @xmath182 ( @xmath183 ) are the flow velocities seen " by the two particles at time @xmath184 .
again we have changed the lower integration limit in the formal solution , eq .
( [ formalsolution ] ) , to @xmath55 .
inserting eq .
( [ velocitydifference ] ) into the definition ( eq .
( [ particlestructure ] ) ) of @xmath138 , it is straightforward to find that , @xmath185 where @xmath186 , named the trajectory structure tensor by pp10 , is defined as , @xmath187 \left [ u_j^{(2 ) } ( \tau ' ) - u_j^{(1 ) } ( \tau ' ) \right]\right \rangle.\ ] ] this tensor represents the correlation of the flow velocity differences on the trajectories of the two particles at two times @xmath184 and @xmath188 .
@xmath189 depends on the separation , @xmath133 , through the constraint that @xmath190 .
( [ particlestructureintegral ] ) is in close analogy with eq .
( [ 1particlevelocity ] ) for the 1-particle velocity . here
the trajectory structure tensor , @xmath189 , replaces the trajectory correlation tensor , @xmath191 , in eq .
( [ 1particlevelocity ] ) .
the physical meaning of eqs .
( [ velocitydifference ] ) and ( [ particlestructureintegral ] ) is clear : the relative velocity of two identical inertial particles is controlled by the particles memory of the flow velocity difference within a friction timescale , @xmath192 , in the past .
the physical picture is illustrated in fig . 1 .
the trajectory structure tensor , @xmath189 , is unknown , and we model it using the approach of pp10 .
since the flow velocity difference scales with the distance , @xmath189 has an indirect dependence on the particle separation at @xmath184 and @xmath188 .
we denote the particle separation at @xmath184 as @xmath193 ( @xmath194 ) .
the vector @xmath195 is stochastic because of the random dispersion of the particle pair by turbulent motions .
@xmath189 also has a dependence on the time lag @xmath196 .
this dependence is associated with the temporal correlation of turbulent structures or eddies encountered by the two particles between @xmath184 and @xmath188 , and the correlation time is essentially the turnover time of these eddies . to estimate @xmath189 , we consider the ( indirect ) spatial dependence on the particle separation and the temporal dependence on the time lag separately .
we use a typical particle separation @xmath197 between @xmath184 and @xmath188 to model the spatial dependence .
like @xmath198 and @xmath199 , @xmath197 is also a random vector .
we approximate the dependence on the separation by the eulerian structure tensor of the flow velocity , @xmath200 , defined in eq .
( [ flowstructure ] ) .
we denote as @xmath201 the temporal correlation of the flow structure at the scale @xmath202 .
@xmath203 is expected to be an even function of the time lag and is normalized to unity , @xmath204 , at zero time lag . to distinguish from the temporal correlation , @xmath205 , along the trajectory of a single particle ( see 2 ) ,
we have used a subscript 2 " here for the two - particle case .
the trajectory structure tensor is then modeled as the product of the two dependences ( pp10 ) , @xmath206 where @xmath207 denotes the average over the statistics of the random vector , @xmath208 .
this average is over the probability distributions of both the amplitude , @xmath202 , and the direction of @xmath208 .
( [ pp10assumption0 ] ) implicitly assumes the statistical independence of the velocity difference , @xmath209 , seen by the two particles from their separation , @xmath208 .
rigorously , the amplitudes of @xmath209 and @xmath208 may have a correlation .
if the particle pair encounters an eddy with a larger velocity , the particle separation tends to be larger .
for example , if @xmath202 is in the inertial range of the flow , @xmath210 from the refined similarity hypothesis ( kolmogorov 1962 ) , where @xmath211 is the average dissipation rate over the scale @xmath202 seen by the particle pair . a positive correlation is expected between the fluctuations in @xmath211 and @xmath202 .
( [ pp10assumption0 ] ) neglects this correlation and may underestimate @xmath189 and hence the particle relative velocity .
the @xmath203 term in eq .
( [ pp10assumption0 ] ) does not depend on the direction of @xmath208 , so one can first take the angular average of @xmath200 and then average the entire term over the pdf of the amplitude , @xmath202 .
the latter can not be exactly performed because the pdf of @xmath202 is unknown . with some simple estimates ,
pp10 argued that simply using the rms of @xmath202 to evaluate @xmath189 ( instead of averaging over the pdf of @xmath202 ) only causes a small difference ( @xmath212 ) in the model prediction .
following pp10 , we ignore the pdf of @xmath202 and insert the rms of @xmath202 to evaluate @xmath189 . for the simplicity of notation
, we use @xmath202 to denote the rms particle distance in between @xmath184 and @xmath188 in the rest of the paper .
a similar notation is adopted for @xmath213 and @xmath214 , which will denote the rms separations at @xmath184 and @xmath188 , respectively .
we approximate @xmath202 by the geometric average of @xmath213 and @xmath214 , @xmath215^{1/2}. \label{typicalseparation}\ ] ] the rms separation @xmath213 as a function of time @xmath184 will be discussed in 3.2.3 . with the above assumptions ,
the trajectory structure tensor is modeled as , @xmath216 the angular average of @xmath165 over the direction of @xmath208 will be carried out in 3.2.2 . in eq .
( [ pp10assumption ] ) , the dependence of @xmath189 on @xmath133 is through the dependence of @xmath213 , @xmath214 and @xmath202 on @xmath133 .
we refer to @xmath217 as the initial " separation , although it actually corresponds to the current or final separation of the two particles .
our formulation indicates that the separation of particle pairs backward in time is crucial for the prediction of the particle relative speed . inserting eq .
( [ pp10assumption ] ) into eq .
( [ particlestructureintegral ] ) gives the pp10 model for the particle structure tensor , @xmath218 we will numerically compute this double integral after evaluating or modeling the angular average , the temporal correlation and the particle separation backward in time .
a simplification of the pp10 model is to set @xmath208 to one of two distances , @xmath198 or @xmath199 , instead of their geometric average .
we find that replacing @xmath208 in eq .
( [ particlestructuremodel ] ) by either @xmath198 or @xmath199 leads to equivalent model prediction for the particle relative speed .
this is because @xmath203 in eq .
( [ particlestructuremodel ] ) is an even function of @xmath70(@xmath219 ) , and the product of the two exponential cutoffs are invariant under the exchange of @xmath184 and @xmath188 . if one sets @xmath220 in eq .
( [ particlestructuremodel ] ) , the integral over @xmath188 can be isolated , yielding , @xmath221 where the angular average is over the direction of @xmath198 and the function @xmath222 is defined as , @xmath223 the factor @xmath222 may be roughly viewed as a response function of the particle pair to turbulent eddies at the scale @xmath213 . although not indicated explicitly , the factor @xmath222 also depends on @xmath13 through its dependence on @xmath213
. we will refer to eqs .
( [ separatedintegral ] ) and ( [ ffactor ] ) as the simplified model . in the simplified model , @xmath222 can be integrated analytically using assumed function forms of @xmath203 in 3.2.1 , and one only needs to numerically solve a single integral in eq .
( [ separatedintegral ] ) . on the other hand , for the original pp10 model
, one must numerically evaluate the double integral in eq .
( [ particlestructuremodel ] ) . to estimate the temporal correlation , @xmath203 , in the trajectory structure tensor , @xmath189
, we first consider a special case where the particle separation , @xmath202 , is much larger than the integral length scale , @xmath108 , of the flow . in this case , the flow velocities , @xmath224 and @xmath225 , seen " by the two particles are independent , and @xmath186 can be written as @xmath226 ( see eq .
( [ particlestructure ] ) ) .
both terms correspond to the trajectory correlation tensor @xmath191 defined below eq .
( [ 1particlevelocity ] ) in 2 , and for identical particles the two terms are equal .
therefore , for @xmath227 , @xmath228 is the same as the temporal correlation coefficient , @xmath89 , along the trajectory of one particle . in 2
, we approximated @xmath81 by the lagrangian correlation function , @xmath60 . using the approximation again
, we have @xmath229 for @xmath118 .
two function forms , single- and bi - exponential , were adopted for @xmath60 in 2 . with the single - exponential form , we set @xmath230 for @xmath118 .
an extension of this function to smaller scales gives , @xmath231 where @xmath232 is essentially the correlation time or lifetime of turbulent eddies of size @xmath98 . for @xmath118 , we set @xmath233 . at
smaller @xmath98 , @xmath232 can be estimated using the velocity scalings in the turbulent flow . for @xmath98 in the inertial range ,
we obtain @xmath232 by dividing @xmath98 by the amplitude of the turbulent velocity fluctuations at this scale , which is @xmath234 . using the kolmogorov scaling for structure functions
, we have @xmath235 , where @xmath236 .
the factor , @xmath237 , is from the incompressibility relation @xmath238 in the inertial range . since the kolmogorov constant @xmath115 is @xmath116 , we set @xmath239 .
a similar value of @xmath240 was adopted by zaichik & alipchenkov ( 2003 ) . in the viscous range with @xmath241 ,
the flow velocity difference goes linearly with @xmath98 , and @xmath242 is expected to be constant .
lundgren ( 1981 ) predicted that @xmath243 for @xmath244 , which was later confirmed by numerical simulations of girimaji & pope ( 1990 ) .
we thus take @xmath245 for @xmath241 in our model .
we will use the bridging formula for @xmath232 from zaichik et al .
( 2006 ) , @xmath246^{-2/3 } \times \hspace{1 cm } \notag\\ \hspace{2.7 cm } \left [ \frac{(\ell/
\eta)^4}{(\ell/ \eta)^4 + ( t_{\rm l}/ ( c_{\rm t } \tau_\eta))^{6 } } \right]^{1/6 } , \label{tr}\end{gathered}\ ] ] which satisfies the scalings of @xmath232 in different scale ranges . one may also adopt a bi - exponential form for @xmath247 based on eq .
( [ biexponential ] ) for the lagrangian correlation function @xmath60 ( see 2 ) . replacing @xmath4 in eq .
( [ biexponential ] ) by @xmath248 gives , @xmath249 - \big(1-\sqrt{1 - 2z^2}\big )
\times\notag\\ \exp \left [ { - \frac{2 |\delta \tau| } { \big(1 - \sqrt{1 - 2z^2 } \big ) t(\ell ) } } \right ] { \bbigg@{3}}\}.\hspace{1.3 cm } \label{biexponential2}\end{gathered}\ ] ] this bi - exponential form for @xmath247 was used in all the calculations in pp10 .
we will compute the predictions of the pp10 model using both the single- and bi - exponential correlation functions .
we find the results from the two cases are close to each other , suggesting that the double integral in eq .
( [ particlestructuremodel ] ) is insensitive to the function form of @xmath250 . after the integration ,
the dependence on @xmath250 is essentially condensed to a dependence on the timescale @xmath232 .
this is similar to the case of the one - particle velocity , which is insensitive to the form of @xmath251 ( see 2 ) .
pp10 also considered the possible dependence of the parameter @xmath75 on the length scale @xmath98 .
it was found that including a reasonable length scale dependence of @xmath75 barely changes the model prediction .
we will set @xmath75 to be constant in this study .
we next consider the simplified model represented by eqs .
( [ separatedintegral ] ) and ( [ ffactor ] ) . with a single - exponential @xmath203 ( eq . ( [ singleexponential ] ) ) , the response factor @xmath222 defined in eq . ( [ ffactor ] )
can be integrated analytcally , @xmath252 since @xmath184 is negative , @xmath222 is dominated by the first term if @xmath253 , and it approaches @xmath254 in that limit .
on the other hand , for @xmath255 , the leading term is @xmath256 .
note that eq .
( [ response ] ) does not diverge at @xmath257 .
applying the lhospital s rule shows that it converges to @xmath258 , as @xmath259 .
therefore , when numerically integrating eq .
( [ separatedintegral ] ) , we set @xmath260 for @xmath261 around @xmath0 . with the bi - exponential temporal correlation , eq .
( [ biexponential2 ] ) , the response factor , @xmath222 , can also be integrated analytically .
the integration is straightforward , but the resulting function for @xmath222 is complicated and is thus omitted here .
the predictions of the simplified model with single- and bi - exponential @xmath250 are also found to be close to each other .
we evaluate the angular average of @xmath200 over the direction of @xmath208 .
it follows from eq .
( [ flowstructure ] ) that @xmath262 r_i r_j / r^2 $ ] .
the contraction of the tensor is @xmath263 , which does not have a direct dependence on the direction of @xmath208 .
therefore , to predict the 3d rms relative speed , we do not need to perform the angular average .
however , for the radial and tangential components , one must make an assumption for the direction of @xmath208 and compute the angular average for the term @xmath264 . in pp10
, we assumed that the direction of the separation change , @xmath265 , caused by turbulent dispersion is completely random or isotropic .
one can then insert @xmath266 into @xmath200 and take the average over the direction of @xmath267 . from the assumed isotropy of @xmath268 , we have @xmath269 and @xmath270 , and hence @xmath271 ( see pp10 ) . the angular average @xmath272 is then given by , of @xmath208 and hence @xmath273 and @xmath274 have a dependence on the direction of @xmath268 .
however , the average of these quantities over the direction of @xmath268 is complicated and can not be done analytically . for simplicity
, we kept @xmath202 , @xmath96 and @xmath97 fixed , and only accounted for the angular average of @xmath275 .
] , @xmath276 + \big[s_{\rm ll}(r ) - s_{\rm nn}(r)\big]\frac{r_i r_j}{r^2}. \label{randomdirection1}\end{gathered}\ ] ] the equation approaches @xmath277 in the limit @xmath278 .
pp10 showed that eq .
( [ randomdirection1 ] ) reproduces the s - t formula for the radial and tangential relative speeds . in the limit @xmath279
, we have @xmath280 $ ] . here
we make a simpler assumption than pp10 : we take the direction of @xmath208 ( rather than @xmath268 ) to be isotropic .
this means @xmath281 , and we have , @xmath282 \delta_{ij } , \label{randomdirection2}\ ] ] which suggests that the particle structure tensor @xmath283 ( see eq .
( [ particlestructuremodel ] ) ) , and hence @xmath284 ( see eq .
( [ particlestructure2 ] ) ) for particles of any size . a comparison of the two assumptions , eqs .
( [ randomdirection1 ] ) and ( [ randomdirection2 ] ) , shows that they differ only at @xmath285 .
as expected , the contraction @xmath286 of both eq .
( [ randomdirection1 ] ) and eq .
( [ randomdirection2 ] ) , is equal to @xmath287 $ ] , indicating that the two assumptions give the same prediction for the 3d rms relative velocity .
the only difference between the two assumptions is the prediction for the radial and tangential components at @xmath288 . in 3.2.4 , we will compare the model predictions by the two assumptions .
the angular average @xmath289 in the simplified model ( eq . ( [ separatedintegral ] ) ) can be evaluated similarly , and the resulting expressions are in the same form as eqs . ( [ randomdirection1 ] ) and ( [ randomdirection2 ] ) with @xmath213 replacing @xmath290 .
we finally specify the ( rms ) particle separation , @xmath213 , as a function of @xmath184 .
the separation of inertial particle pairs backward in time has not been explored in the literature .
fortunately , bec et al . (
2010 ) carried out a detailed numerical study of the forward - in - time pair dispersion of inertial particles .
following pp10 , we use their results to guide the assumption for the backward dispersion .
we first consider the separation behavior of inertial - range particles with @xmath178 .
bec et al . ( 2010 ) found that the separation of inertial particles shows different behaviors at early and late times . at early times , a clear ballistic phase is observed for particles with @xmath291 . in this phase ,
the separation increases linearly with time , and the phase lasts for about a friction timescale .
the ballistic behavior is easy to understand : the particle velocity tends to be roughly constant for a memory timescale , @xmath0 .
this also applies to the dispersion backward in time .
we thus assume that , for particle pairs at an initial " distance of @xmath13 , the separation @xmath213 in the time range @xmath292 is given by , @xmath293 where @xmath148 is the 3d variance of the particle relative velocity at time 0 .
the particle relative speed is actually what our model aims to predict .
therefore , the dependence of @xmath213 on @xmath148 in the ballistic phase leads to an implicit equation for @xmath148(see 3.2.4 ) .
bec et al . ( 2010 ) also showed that , after a friction timescale , the dispersion of inertial - range particles make a transition to a tracer - like phase , where the separation variance increases as time cubed , a behavior known as the richardson law .
the richardson law was first discovered for tracer pair dispersion at inertial - range scales .
the transition to the richardson phase at a friction timescale or so suggests the ballistic separation for a duration of @xmath0 already brings the average particle distance into the inertial range of the flow .
the richardson behavior was observed in the tracer pair dispersion both forward and backward in time ( berg et al .
2006 ; see appendix a ) .
it is thus likely to exist also in the backward separation of inertial particles .
we connect the richardson phase to the ballistic phase at @xmath294 , and use the richardson law @xmath295 at @xmath296 , where @xmath297 is called the richardson constant and @xmath110 is the average dissipation rate of the flow . as the backward separation is typically faster than the forward case , the transition to the richardson phase might occur slightly earlier than assumed here .
bec et al . (
2010 ) did not report the value of @xmath297 in the richardson phase of inertial particle pair dispersion . as in pp10
, we will take @xmath297 as a parameter . in our model , we use a combined separation behavior that connects a ballistic and a richardson phase at @xmath294 . the richardson behavior would end when the separation becomes larger than the integral length scale , @xmath108 , of the turbulent flow . at such a large distance , the flow velocities seen " by the two particles
is uncorrelated , and the particle separation is expected to be diffusive like in a random walk . it is thus appropriate to switch the richardson behavior to a diffusive phase with @xmath298 at @xmath299 .
however , we find that the exact separation behavior at @xmath300 ( or @xmath227 ) does not affect the prediction of our model .
this is because at these scales both the structure functions , @xmath301 and @xmath96 , and the timescale , @xmath261 ( or @xmath302 ) , become independent of @xmath303 ( or @xmath202 ) . therefore , eq . ( [ particlestructuremodel ] ) ( or eq . ( [ separatedintegral ] ) )
is insensitive to the behavior of the separation once it becomes much larger than @xmath108 .
this is confirmed by the numerical solutions of eqs .
( [ particlestructuremodel ] ) and ( [ separatedintegral ] ) . for convenience , we keep using the richardson s law even after @xmath303 exceeds @xmath108 . the separation behavior discussed above is based on the simulation results of bec et al .
( 2010 ) for particles in the inertial range . for simplicity , we will use the same behavior for all particles , although its validity is questionable for small ( @xmath304 ) and large ( @xmath305 ) particles . for small particles with @xmath306 , a ballistic phase
is not clearly observed in the @xmath307 vs. time plots in fig . 5 of bec et al .
we expect that a short ballistic phase is likely to exist if one plots @xmath308 ( instead of @xmath307 ) vs. time ( see fig . [ separation ] in appendix a for the @xmath308 vs. time plot for tracer particle pairs ) . however , for @xmath309 particles , the connection of the short ballistic phase to the richardson behavior is more complicated than in the case of larger particles ( fig . 5 of bec et al .
this is because the pair separation of these particles does not enter the inertial range of the flow in a friction timescale or so .
therefore , an intermediate phase exists in between the ballistic and richardson phases .
ideally , a three - phase behavior should be considered .
unfortunately , the separation behavior in the intermediate phase is completely unknown , and thus , to include it , one must adopt a pure parameterization . here
we take a simpler approximation : we still connect the richardson behavior directly to the ballistic phase for @xmath306 particles , although it can not be justified physically .
essentially , this parameterizes the later two phases by a single richardson law with a free parameter @xmath297 .
future numerical studies for the entire separation behavior of small particles is needed to improve the approximation . for the particle distance range , @xmath310 , considered in our data analysis ( see 6.1 ) , our model with the assumed behavior
does give acceptable prediction for @xmath311 particles .
however , in the @xmath15 limit , a careful study of the intermediate separation phase of @xmath311 particles is necesary to accurately model their relative velocity .
the problem of using the assumed behavior for particles with @xmath9 is that the richardson phase does not exit .
the velocities of these large particles are uncorrelated even at small distances ( 3.1 ) . therefore
, at timescales larger than @xmath0 , the separation is likely diffusive , i.e. , @xmath312 .
realistically , one needs to connect the ballistic phase to a diffusive behavior rather than a richardson law at @xmath313 .
however , it turns out that , at the end of the ballistic phase of these particles , the separation is already @xmath314 . as discussed earlier , once the separation exceeds @xmath108 , the exact separation behavior would not significantly affect the model prediction
this justifies using a combined separation behavior with a ballistic and a richardson phase also for @xmath9 particles .
so far , the initial distance , @xmath13 , just provides a floor value in our assumption for the particle separation @xmath303 ( eq .
( [ ballistic ] ) ) .
it is , however , possible that the value of @xmath13 has additional effects on the separation behavior .
bec et al . (
2010 ) only explored @xmath13 above the kolmogorov scale , and it is not clear whether the separation behavior has a qualitative difference if @xmath125 . to model the particle collision speed , we are interested in the backward separation with @xmath166 , and it would thus be helpful to systematically investigate whether and how the separation behavior changes as @xmath13 decreases below @xmath19 .
we defer such a study to a later work . due to the uncertainty in the separation behavior for @xmath166
, we will focus on testing the model prediction for the relative velocity at significant fractions of the kolmogorov scale ( @xmath310 ) .
we assume that the two - phase behavior discussed above applies for this range of @xmath13 .
considering the existence of various uncertainties , the assumed separation behavior should be viewed more or less as a parameterization . to constrain @xmath297 in the richardson phase , in appendix a we measure @xmath297 for the backward dispersion of tracer particle pairs in our simulated flow , which is used as a reference for inertial particles .
the measured @xmath297 for tracers in our flow at a limited resolution shows a dependence on @xmath13 , suggesting that the richardson constant for inertial particles may also depend on @xmath13 .
when comparing our model prediction to the simulation results at different @xmath13 , we will adjust @xmath297 to obtain best fits , and examine whether the best - fit values are consistent with the range of @xmath297 measured from tracer particles .
the richardson constant for inertial particles may also have a dependence on @xmath0 ( or @xmath1 ) , which will be ignored for simplicity .
finally , we point out that our model for the rms relative velocity does not directly account for the effect of the spatial clustering of the particles ( see 7 ) .
ideally , a theoretical model needs to consider the clustering and relative velocity statistics simultaneously . at a given time
, the relative velocity determines the evolution of the spatial distribution of the particles , while the particle distribution may affect how the particles see " the flow velocity and hence the evolution of the relative velocity statistics .
however , modeling clustering and the relative velocity together self - consistently is very challenging , and is out of the scope of the current work .
our model for the particle structure tensor , @xmath138 , is now complete .
here we discuss the qualitative behavior of our model prediction .
we start by considering the 3d variance , @xmath148 .
the contraction of eq .
( [ particlestructuremodel ] ) gives , @xmath315 \times } \hspace{1cm}\notag \\
\hspace{2 cm } { \displaystyle \phi_2 \big(\tau ' -\tau , r \big ) \exp \left(\frac{\tau}{\tau_{\rm p } } \right ) \exp \left(\frac{\tau'}{\tau_{\rm p}}\right ) } , \label{w2}\end{gathered}\ ] ] which is an implicit equation of @xmath316 because @xmath202 depends on @xmath148 in the ballistic separation phase . in 6.1
, we will solve the equation numerically using an iterative method .
the qualitative behavior of the model prediction for @xmath148 can be obtained by analyzing the integrand in eq .
( [ w2 ] ) . in the s - t limit ( @xmath317 ) , the exponential cutoff terms ,
@xmath318 and @xmath319 , in the integrand can be viewed as delta functions at @xmath320 and @xmath321 , respectively .
this suggests that @xmath148 is approximately given by @xmath322 at @xmath323 . since @xmath324
, we have @xmath325 for @xmath13 in the dissipation rate , which is consistent with the s - t prediction ( see eq .
( [ saffmanturner ] ) ) for the 3d variance of the relative velocity . the analysis of eq .
( [ w2 ] ) for larger particles is more complicated .
we first note that @xmath273 , @xmath326 , and the timescale @xmath302 in the correlation function @xmath203 are all increasing functions of @xmath202 . since @xmath202 increases backward in time , the first factor in the integrand of eq .
( [ w2 ] ) increases with increasing @xmath327 and @xmath328 .
a larger @xmath302 also tends to increase the integral because , with increasing @xmath302 , @xmath203 allows contributions from a broader range of time lag ( @xmath70 ) .
together with the exponential cutoffs , these suggest that the contribution to the integral peaks at @xmath329 .
we denote the particle separation at @xmath330 as @xmath331 ( @xmath332 ) , and refer to it as the primary distance . in the extreme limit of large particles with @xmath9
, @xmath331 is expected to be much larger than the integral scale , @xmath108 , of the flow . at @xmath333 , we have @xmath334 , and @xmath335 .
the exponential cutoff by @xmath203 indicates that only the time pairs ( @xmath184 and @xmath188 ) that satisfy the constraint @xmath336 give a significant contribution to the integral .
since @xmath337 , @xmath203 reduces the range of @xmath184 and @xmath188 that contributes to the double integral by a factor of @xmath338 . assuming the main contribution to the integral is from @xmath339 and accounting for the effect of @xmath203 , we find @xmath340 .
this is consistent with eq .
( [ largelimit ] ) , meaning that our model correctly reproduces the large particle limit . for inertial - range particles with @xmath178 ,
the primary distance , @xmath331 , corresponds to inertial - range scales of the turbulent flow , is in analogy with the critical wavenumber , @xmath341 , defined in the model of volk et al .
( 1980 ) . in the language of volk et al .
, the velocity structures at scales much larger than @xmath331 would be counted as class i eddies , while structures below @xmath331 would belong to class iii . however , pan and padoan ( 2010 ) pointed out a physical weakness in the evaluation of @xmath341 in the volk et al .
model for the two - particle relative velocity , as the role of the separation of the two particles is not properly accounted for .
the reader is referred to section 4.1 of pan and padoan ( 2010 ) for a detailed discussion on this issue . ] .
using the kolmogorov scaling gives @xmath342 and @xmath343 . from its definition , @xmath331 is roughly the particle distance at the time when the ballistic phase connects to the richardson phase ( see 3.2.3 ) .
we thus assume that @xmath331 is determined by a ballistic separation of duration @xmath0 , i.e. , @xmath344 .
the effect of @xmath203 depends on how @xmath345 compares to @xmath0 . if @xmath346 , @xmath347 for all time pairs in the range @xmath348 . on the other hand , if @xmath349 , @xmath203 provides a factor of @xmath350 , which follows from the same argument used above for the large particle limit .
we find that both cases lead to the same scaling of @xmath148 with @xmath0 . in the first case , eq .
( [ w2 ] ) is approximated by @xmath351 \propto r_{\rm p}^{2/3}$ ] . with @xmath344
, we obtain @xmath352 . in the second case with @xmath349 ,
we include a factor of @xmath350 and estimate @xmath148 as @xmath353t(r_{\rm p})/\tau_{\rm p } \propto r_{\rm p}^{4/3}/\tau_{\rm p}$ ] .
it is straightforward to see that setting @xmath344 in this estimate gives the same scaling , @xmath352 , as the first case .
therefore , whether @xmath345 is larger or smaller than @xmath0 , our model predicts a @xmath354 ( or @xmath179 ) scaling for inertial - range particles . using a similar argument , pp10 found that if the primary distance is determined by the richardson s law , @xmath355 , the model also predicts a @xmath179 scaling for inertial - range particles . since both the ballistic and richardson behaviors yield a @xmath179 scaling , a combination of a ballistic and a richardson phase produces the same scaling ( pp10 ) .
the @xmath179 scaling has been previously predicted by models of volk et al .
( 1980 ) , cuzzi and hogan ( 2003 ) , ormel and cuzzi ( 2007 ) , and zaichik & alipchenkov ( 2003 ) .
as mentioned earlier , the scaling may also be obtained from a dimensional analysis under a scale - invariant assumption .
if the dimensional analysis exactly holds , then the departure from the @xmath179 scaling for inertial - range particles in a simulation of limited resolution is caused completely by the effects from dissipation or driving scales .
the derivation of the @xmath179 scaling in all the models assumes a sufficiently broad inertial range .
the scaling would not exist if the reynolds number of the turbulent flow is low .
in fact , the predicted @xmath179 behavior has never been confirmed by simulations due to the low numerical resolution .
pp10 showed that , to see the @xmath179 scaling , the taylor reynolds number of the turbulent flow must be larger than @xmath356 .
this is higher than in the @xmath17 simulation used in the present study , and thus a clear @xmath179 scaling is not observed .
it appears likely that the existence of this scaling can be verified at a twice larger resolution .
we will conduct a @xmath357 simulation in a future work .
the above analysis for the scaling behavior of @xmath148 in different @xmath1 ranges can be similarly applied to the simplified model , eqs .
( [ separatedintegral ] ) and ( [ ffactor ] ) .
the prediction of the simplified model is qualitatively the same as the original pp10 model .
finally , we examine the model prediction for the radial and tangential components of the relative velocity . the prediction for @xmath140 and
@xmath141 depends on the angular average of @xmath189 in eq .
( [ particlestructuremodel ] ) ( or eq . [ separatedintegral ] ) . in 3.3.2 , we made two assumptions , eqs .
( [ randomdirection1 ] ) and ( [ randomdirection2 ] ) , for the angular average . inserting the first assumption , eq .
( [ randomdirection1 ] ) , into eq .
( [ particlestructuremodel ] ) and comparing it with eq .
( [ particlestructure2 ] ) , we find , @xmath358 \phi_2 \big(\tau ' -\tau , r \big ) \exp \left(\frac{\tau+\tau'}{\tau_{\rm p}}\right ) , \notag \ ] ] @xmath359 \phi_2 \big(\tau ' -\tau , r \big ) \exp \left(\frac{\tau+\tau'}{\tau_{\rm p } } \right ) .
\label{wrwt}\end{gathered}\ ] ] in order to integrate these two equations , one needs to first solve eq .
( [ w2 ] ) for @xmath148 due to the dependence of @xmath202 on @xmath148 in the ballistic phase .
it is easy to show that , in the limit @xmath317 , @xmath278 , and eq .
( [ wrwt ] ) reduces to @xmath360 and @xmath361 , reproducing the s - t formula , eq .
( [ saffmanturner ] ) . for larger particles with @xmath362
, we have @xmath363 , and thus eq . ( [ wrwt ] ) predicts @xmath364 , therefore , like @xmath148 , both @xmath140 and @xmath365 scale as @xmath179 for inertial - range particles and as @xmath177 in the large particle limit . as discussed in 3.3.2 , the second assumption , eq .
( [ randomdirection2 ] ) , for the angular average of @xmath189 predicts that @xmath364 for all particles . in the s - t limit , eq .
( [ w2 ] ) gives @xmath366 , and thus @xmath367 .
this means that the prediction by the second assumption for the radial and tangential relative speeds of @xmath368 particles differs from the s - t formula , although it reproduces the s - t prediction for the 3d rms .
we will test the model predictions for the relative velocity variances measured from our simulation data .
in this section , we describe the numerical method used in our simulation and discuss the statistical properties of the simulated flow .
our simulation was conducted in a periodic @xmath17 box with a length of @xmath369 on each side . using the pencil code ( brandenburg & dobler 2002 , johansen , andersen , & brandenburg 2004 ) , we evolved the hydrodynamic equations , @xmath370 \hspace{3.5 cm } \notag \\
-\frac{1}{\rho}\frac{\partial p}{\partial x_i } + { f_i } , \hspace{1.7cm}\end{gathered}\ ] ] with an isothermal equation of state , @xmath371 .
the sound speed is set to unity , i.e. , @xmath372 . the kinematic viscosity , @xmath34 , is taken to be constant , @xmath373 .
a large - scale force , @xmath374 , generated in fourier space using 20 modes in the wavenumber range of @xmath375 is applied to drive and maintain the turbulent flow .
the driving length scale , @xmath376 , is thus about 1/2 box size .
the balance between the energy input by the driving force and the dissipation by viscosity leads to a statistical steady state with a 1d rms velocity , @xmath58 , of 0.05 , or a 3d rms of @xmath377 .
this weakly compressible flow is suitable for the application to turbulence in protoplanetary disks . at an rms mach number of @xmath377 ,
the flow statistics is essentially the same as incompressible turbulence ( padoan et al .
2004 , pan & scannapieco 2011 ) . the integral length scale , @xmath108 , in our simulated flow is found to be @xmath6 , i.e. , about 1/6 box size .
it is about 3 times smaller than the driving scale , @xmath376 .
the integral scale , @xmath108 , represents the ( longitudinal ) correlation length of the velocity field , and we computed it from the energy spectrum , @xmath378 , of the flow , using the relation @xmath379 ( monin & yaglom 1975 ) . the energy spectrum , @xmath378 , is plot in the inset of fig . ( [ structureandpower ] ) .
with @xmath380 , the large - eddy turnover time is @xmath381 in units in which the sound crossing time is @xmath382 .
the average energy dissipation rate per unit volume by the viscosity term is given by @xmath383 , where @xmath384 is the average density . in our weakly compressible flow ,
the density fluctuations and the velocity divergence can be neglected , and the dissipation rate can be estimated by @xmath385 , where @xmath386 is the vorticity variance .
we find that @xmath387 , implying that @xmath388 .
we also evaluated the dissipation rate from the 3rd - order longitudinal structure function using kolmogorov s 4/5 law , @xmath389 , for @xmath98 in the inertial range .
this latter method gives a larger dissipation rate , @xmath390 , suggesting that a small fraction , @xmath391 , of kinetic energy is dissipated by numerical diffusion .
the effective viscosity is thus larger than the adopted value by the same amount .
we take the effective viscosity to be @xmath392 and use it in our estimates of the kolmogorov scales .
we compute the kolmogorov timescale from the vorticity variance as @xmath393 .
the kolmogorov length scale is estimated to be @xmath394 , which corresponds to @xmath395 cell size of the computation grid .
the kolmogorov velocity scale is @xmath396 in units of the sound speed .
the reynolds number of our simulated flow is @xmath397 .
a more commonly - used reynolds number in turbulence studies is the taylor reynolds number , @xmath398 , where the taylor micro length scale is defined as @xmath399 .
we find that @xmath400 in our simulated flow , and thus @xmath401 . from the definitions of @xmath123 and @xmath402 , we have @xmath403 .
to study the lagrangian statistics , we integrated the trajectories of @xmath404 million tracer particles with zero inertia in the simulated flow .
the total number of tracer particles corresponds to an average number density of 1 particle per 4 computational cells . to obtain the particle velocity inside a cell , we selected the triangular - shaped - cloud interpolation method already implemented in the pencil code ( johansen and youdin 2007 ) .
we output the particle positions to a data file in each @xmath405 . the lagrangian correlation function , @xmath69 , is computed as the average of the velocity correlation , @xmath406 , along the trajectories , @xmath407 , of all particles .
we considered both positive and negative @xmath70 , corresponding to lagrangian trajectories forward and backward in time , respectively .
our data confirmed that @xmath60 is an even function of @xmath70 , as expected from statistical stationarity ( see 2 ) .
we find that the lagrangian correlation timescale , @xmath4 ( @xmath408 ) , is @xmath409 , which is about 0.75 eddy turnover time , @xmath410 .
this is consistent with the simulation result of yeung et al .
( 2006 ) . since @xmath411 in our flow
, we have @xmath412 . ; circles ) and eulerian ( @xmath413 ; triangles ) temporal correlation functions in our simulated flow . the time lag , @xmath70 , is normalized to the kolmogorov timescale , @xmath10 and the lagrangian correlation time , @xmath4 , on the bottom and top x - axises , respectively .
the solid line shows the best fit for @xmath60 using the bi - exponential function , eq .
( [ biexponential ] ) . ]
the lagrangian correlation , @xmath60 , in our flow is plot as circles in fig .
[ correlation ] , where the time lag , @xmath414 , is normalized to the kolmogorov timescale , @xmath152 and the lagrangian correlation time , @xmath4 , on the bottom and top x - axises , respectively .
the solid line shows the bi - exponential function , eq .
( [ biexponential ] ) , given in 2 .
the parameter @xmath75 is set to 0.3 , which suggests that the taylor micro timescale , @xmath71 , is @xmath415 .
this value of @xmath71 corresponds to an acceleration variance , @xmath416 .
the bi - exponential function matches very well the simulation data . on the other hand
, we find that a single exponential function could not give a satisfactory fit to @xmath60 .
we also considered the eulerian temporal correlation function , @xmath417 .
it is computed as the average , @xmath418 , over all grid points @xmath94 .
the result is plot as triangles in fig .
[ correlation ] .
@xmath413 is smaller than the lagrangian correlation @xmath60 at small time lags , and then becomes larger at @xmath419 .
due to the slower decrease of @xmath413 at large time lags , the eulerian correlation time , @xmath420 , is slightly ( 10% ) larger than @xmath4 .
we find that @xmath421 .
the eulerian correlation function is of interest for large inertial particles with @xmath9 . due to their large inertia , these particles have small velocities and
thus may stay around as the flow sweeps by .
therefore , unlike small particles , the temporal series of the flow velocity seen " by the large particles may be better described by the eulerian velocity .
this suggests that , for @xmath305 , it may be appropriate to replace the lagrangian correlation used in our model by the eulerian correlation
. however , the eulerian correlation function and timescale are quite close to the lagrangian ones , and using the lagrangian correlation for all particles in our model gives satisfactory predictions for both the 1-particle velocity and the 2-particle relative velocity at any @xmath1 ( 5 and 6.1 ) .
we summarize the relevant timescales in the simulated flow and list them in an increasing order .
the smallest timescale is kolmogorov time @xmath152 , and we use it as a reference timescale .
the taylor micro scale , @xmath71 , was found to be @xmath422 from the bi - exponential fit to the lagrangian correlation function .
the next timescale is the lagrangian correlation time , @xmath4 , which is @xmath423 .
the eulerian correlation time is slightly larger , @xmath424 .
the large eddy turnover time , @xmath410 , was measured to be @xmath425 .
another commonly - used timescale is the dynamical time , @xmath426 , defined as the forcing length scale , @xmath376 , divided by the 3d rms velocity ( @xmath427 ) .
we find that @xmath428 . in this work
, we will express the particle friction time primarily by @xmath1 and @xmath83 .
they correspond to normalizations to @xmath152 and @xmath4 , which are convenient for small and large particles , respectively .
one may also normalize @xmath0 to the large eddy turnover time , and define @xmath429 , which may be more convenient for practical applications .
however , we prefer using @xmath83 than @xmath430 , because , according to our model , it is @xmath4 that directly enters the physics of turbulence - induced particle velocity . using the measured values of the timescales in our simulation , one may convert the normalizations by @xmath431 . in fig .
[ structureandpower ] , we show the longitudinal ( @xmath96 ; open circles ) and transverse ( @xmath97 ; filled circles ) structure functions in our simulated flow . the structure functions are measured from the velocity differences along the 3 directions , @xmath432 , @xmath433 and @xmath434 , of the simulation grid . for @xmath435 , we computed and averaged the variances of @xmath436(@xmath437 ) , @xmath438 and @xmath439 over all the points , @xmath94 .
similarly , @xmath440 is obtained by averaging the variances of @xmath441(@xmath442 ) , @xmath443 , @xmath444 , @xmath445 , @xmath446 and @xmath447 .
as discussed earlier , kolmogorov s similarity theory predicts that @xmath448 for @xmath98 in the inertial range .
we thus compensated the structure functions by @xmath449 in fig .
[ structureandpower ] .
a limited inertial range is seen in both @xmath96 and @xmath97 .
the kolmogorov constant @xmath115 is about 2 . in the inertial range ,
the scaling exponent for @xmath96 is found to be slightly larger than 2/3 , while the slope of @xmath97 is close to 2/3 .
the ratio of the two structure functions in the inertial range is about 1.25 , slightly smaller than the value , 4/3 , expected from the incompressibility condition ( see 3 ) .
this is perhaps because our flow is weakly compressible .
another possibililty is that the inertial range is too short to allow an accurate measurement of this ratio .
both structure functions become smooth , i.e. , @xmath450 , as @xmath98 decreases toward the kolmogorov scale , and approach @xmath451 in the limit @xmath118 ( 3 ) . ) and ( [ snn ] ) , for @xmath96 and @xmath97 , respectively .
the bottom and top x - axises normalize @xmath98 to the kolmogorov scale and the integral scale , respectively .
the inset shows the energy spectrum of the flow , compensated by @xmath452 . ] the solid line in fig .
[ structureandpower ] is the connecting formula , eq .
( [ sll ] ) , for @xmath96 ( 3 ) .
we set @xmath453 in the formula .
the line gives a fairly good fit to the data points . as discussed in 3 , with the connecting formula for @xmath96
, one may obtain a fitting function for @xmath97 using the incompressibility relation @xmath454 .
however , the fitting function obtained this way overestimates @xmath97 in the inertial range , perhaps because the incompressibility condition does not exactly hold in our flow ( see above ) . for a more accurate fit
, we adopted a separate connecting formula for @xmath97 , @xmath455\times \hspace{2 cm } \notag\\ \hspace{2cm}\left [ \frac{(\ell/ \eta)^4}{(\ell/ \eta)^4 + ( 2u'^2 /c_{\rm kn } u_\eta^2 ) ^{6 } } \right]^{1/6 } , \label{snn}\end{gathered}\ ] ] where @xmath456 is the scaling coefficient for @xmath97 in the inertial range .
this connecting formula correctly reproduces the scaling behaviors of @xmath97 in different scale ranges .
its form is slightly different from eq .
( [ sll ] ) for @xmath96 . the dotted line in fig .
[ structureandpower ] corresponds to eq .
( [ snn ] ) with @xmath457 .
we will use eqs .
( [ sll ] ) and ( [ snn ] ) in the computation of our model prediction for the particle relative velocity . the inset of fig
. [ structureandpower ] show the energy spectrum , @xmath378 , of our flow .
the kolmogorov theory predicts @xmath458 in the inertial range , and we compensated the spectrum by @xmath452 .
the power - law range ( @xmath459 ) in the spectrum appears to be shorter than in the structure functions .
the constant @xmath460 is measured to be @xmath461 , consistent with previous studies ( ishihara et al .
it is also consistent with the relation @xmath462 ( monin & yaglon 1975 ) , as the kolmogorov constant , @xmath115 , for @xmath96 was found to be @xmath116 .
in our simulation , we included 14 species of inertial particles of different sizes .
the friction timescale of the particles spans about four decades from @xmath463 to @xmath464 ( @xmath465 or @xmath466 ) , covering the entire scale range of the simulated flow .
the friction timescale is equally spaced , increasing by a factor of two in each successive species .
the number of particles contained in each species is 33.6 million , corresponding to an average particle density of one per 4 computational cells .
the same number of tracer particles was included to study the lagrangian statistics ( 4.1 ) .
the integration of the particle trajectories is computationally very expensive . using 4096 cores ( 512 harpertown nodes ) on the nasa / ames pleiades supercomputer
, the simulation was run for 14 days , corresponding to a total cpu cost of 1.4 million hours . to evolve the particle equation of motion ( eq . [ particlemomentum ] ) , we adopted the triangular - shaped - cloud ( tsc ) method to interpolate the flow velocity inside the computational cells .
the tsc interpolation is a well - established method ( hockney & eastwood 1981 , johansen & youdin 2007 ) that makes use of the nearest 27 grid points in a 3d simulation . in 1d ,
the weighting factor for the nearest 3 grid points is set to be quadratic with the distance to the points .
the velocity difference in our simulated flow is linear with @xmath98 around and below the cell size , @xmath467 , as seen from the @xmath468 scaling of the structure functions toward @xmath467 ( bottom data points ) in fig .
[ structureandpower ] .
this implies that the subgrid velocity field can be well approximated by a linear interpolation ( pan et al .
the linear scaling is also captured by the tsc method .
it is straightforward to show that , if the flow velocity is already linear around the resolution scale ( approximately the case in our simulated flow ) , the scaling of the interpolated velocity at subgrid scales by the tsc method would be exactly linear , as the quadratic terms in the weighting functions cancel out in this special case . in comparison to the linear interpolation ,
the tsc method is of higher order and has the advantage of smoother connections at cell boundaries .
initially , the 33.6 million particles in each species are distributed randomly in the simulation box .
each component of the initial particle velocity is also random , independently drawn from a uniform distribution in the range [ -0.01 , 0.01 ] .
therefore , the initial ( 1d ) rms , @xmath469 , of each velocity component of all the particles is @xmath470 , equivalent to a 3d rms of 0.01 .
the numerical values given here are in units of the gas sound speed , which was set to unity in the simulation .
the initial particle conditions for all the 14 species are the same .
we evolved the turbulent flow and the particle trajectories together right from the beginning of the simulation . at time zero , the gas velocity and density are set to zero and unity , respectively .
our simulation run lasted for about @xmath471 ( or @xmath472 ) , and we saved 52 snapshots with an equal separation of @xmath473 . the black dotted line in fig .
[ rms0 ] is the 3d rms of the flow velocity ( @xmath474 ) as a function of time , which shows that the flow is fully developed and reaches a ( quasi ) steady state at @xmath475 . from top to bottom , the colored lines in fig .
[ rms0 ] plot the 3d rms velocities of inertial particles with @xmath476 = 0.39(0.027)[0.02]$ ] , @xmath477 $ ] , @xmath478 $ ] , @xmath479 $ ] , @xmath480 $ ] , @xmath481 $ ] , @xmath482 $ ] , @xmath483 $ ] , and 795(54)[41.4 ] , respectively .
the numbers in the parentheses and square brackets correspond to @xmath83 and @xmath430 , respectively .
we find there are dips at earlier times in the curves for relatively large particles .
when our first snapshot was saved at @xmath473 , particles with @xmath484 partially lost the memory of the initial rms velocity , and meanwhile their velocity had some contribution from the flow velocity , @xmath58 , between @xmath49 and @xmath473 .
however , since @xmath58 was 0 at @xmath49 , this contribution turns out to be small and does not compensate the decrease due to the memory loss of the initial velocity .
this causes a decrease in @xmath485 and leads to dips at @xmath486 . due to their short memory time
, the small particles forgot their initial velocity , @xmath469 , at the first snapshot , and their velocity was close to the flow velocity at @xmath473 , which is already slightly larger than @xmath469 . therefore ,
no dips appear for the small particles with @xmath487 .
the top six color lines appear to reach a steady state at @xmath488 .
for the bottom three lines , @xmath489 keeps increasing gradually but almost monotonically .
this may imply that these largest particles need more time to relax .
it is also possible that the slow increase of @xmath65 at late times is simply caused by the slight rise of the flow velocity ( see the black dotted line ) . , of inertial particles of different sizes .
the thin dotted line shows the 3d rms flow velocity , which reaches a ( quasi ) steady state at @xmath490 .
the flow rms velocity is 0 at time zero , and the dotted line starts from @xmath473 when the first snapshot is saved . from top to bottom , the color lines correspond to the rms velocity for particles with @xmath491 , 6.12 , 12.4 , 24.9 , 49.7 , 99.4 , 199 , 398 and 795 , respectively .
the velocity is in units of the sound speed of the flow , which is set to unity in the simulation . ]
the relaxation timescale for inertial particles in a stationary turbulent flow is essentially the time for the particles to forget the initial condition , and is roughly given by the friction timescale , if the initial velocity is not much larger than the final steady - state value ( as is the case for our initial conditions ) .
the estimate for the relaxation time in our simulation is a little complicated because the particles are released to the flow before @xmath492 .
for particles with @xmath493 , the dynamical relaxation is expected once the flow is fully developed , i.e. , at @xmath494 .
this is the case for the top 5 - 6 color lines in fig .
[ rms0 ] . for the bottom 3 - 4 lines , @xmath495 , and
we expect these particles would be relaxed at some time in the range ( @xmath0 , @xmath496 ) .
the lower limit is the minimum relaxation time , and the upper limit is based on the consideration that , if the particle evolution started at @xmath492 instead of time 0 , the particles would relax at @xmath497 . from this estimate , the third largest particles ( @xmath498 )
are relaxed by @xmath499 , and the second largest ones ( @xmath500 ) are likely relaxed by the end of the simulation . on the other hand , the largest ( @xmath501 ) particles
may not have reached a relaxed state .
however , the quite flat @xmath502 of the @xmath501 particles indicates the possibility they are actually relaxed toward the end of the simulation . if that is the case , a likely reason for it is that the chosen initial condition ( e.g. , the rms velocity ) happens to be very similar to the expected relaxed state of these particles .
this similarity may reduce the relaxation time .
we assume that all particles in our simulation are relaxed in the last 5 - 6 @xmath410 . in our data analysis
, we average over three snapshots at @xmath503 , 24 and @xmath471 . for the uniformity of the data sample
, we use the same snapshots for all particle species .
since the largest particles become relaxed around the end of the simulation , we only select late snapshots at @xmath504 .
the purpose of averaging over a number of snapshots is to obtain better statistics by increasing the sample size .
it is thus helpful to use well - separated snapshots with independent statistics .
a temporal separation of @xmath505 guarantees the particle velocities at the selected snapshots are independent for the first 10 particle species .
the velocities of the largest four particles remain correlated for significantly longer than @xmath505 . therefore , unlike the case of smaller particles , using the selected snapshots may not effectively increase the independent sample size or the measurement accuracy . if the computation resources allow , it would be ideal to run the simulation much longer and collect snapshots separated by a few friction times of the largest particles . in fig .
[ rms ] , we show the simulation result for the 1d rms , @xmath65 , of the particle velocity as a function of @xmath1 .
we normalized @xmath65 by the rms velocity , @xmath58 , of the flow .
the top x - axis normalizes the friction time to the lagrangian correlation time .
one may convert @xmath1 ( or @xmath83 ) to @xmath430 by @xmath506 ( or @xmath507 .
the dotted and solid lines are predictions , eq . ( [ 1particlevelocityexp ] ) and eq .
( [ 1particlevelocitybiexp ] ) , of our model using single- and bi - exponential forms for the temporal correlation function , respectively .
the model approximates the trajectory correlation function , @xmath81 , by the lagrangian correlation function , @xmath60 ( see 2 ) . for the bi - exponential case
, we set the parameter @xmath508 , which best fits @xmath60 measured from the lagrangian trajectories ( see 4.1 ) .
the two lines almost coincide , indicating that the model prediction for @xmath65 is insensitive to the exact form of the correlation function , and depends only on the correlation timescale is crucial for predicting the relative velocity between different particles in the s - t limit ( pp10 ) . in the bidisperse case , adopting the bi - exponential form for @xmath60 and @xmath81 is needed to reproduce the acceleration contribution to the relative velocity , while a single exponential form can not correctly capture the effect of the flow acceleration on @xmath60 at small time lags . ] . in both curves ,
we set @xmath509 , generally consistent with the directly measured value of @xmath423 ( see 4.1 ) .
as expected , @xmath65 follows the @xmath177 scaling ( the dotted line segment ) at @xmath86 .
this supports our claim that the largest particles are dynamically relaxed at the end of the run .
, as a function of the particle friction time .
the bottom and top x - axises normalize @xmath0 to @xmath10 and @xmath4 , respectively .
the dotted and solid lines are the predictions , eq .
( [ 1particlevelocityexp ] ) and eq .
( [ 1particlevelocitybiexp ] ) , of our model using single- and bi- exponential forms for the temporal correlation function , respectively . in both cases , @xmath4 is set to @xmath510 , and the parameter @xmath75 in the bi - exponential case is set to 0.3 .
the thin dotted line segment denotes a @xmath177 scaling . ]
the flow velocity seen " by large particles with @xmath305 may be closer to eulerian than lagrangian ( 4.1 ) , and thus using @xmath60 for the trajectory correlation @xmath81 is not well justified ( 2 ) .
however , the assumption is validated by our simulation result for all particles .
this is because , first , based on our model prediction , @xmath65 is controlled mainly by the correlation time , not the form of the correlation function , and , second , the eulerian correlation timescale , @xmath511 , was found to be close to @xmath4 .
thus , whether @xmath81 is approximated by @xmath413 or @xmath60 , the predicted 1-particle velocity would be similar , justifying the use of @xmath60 for all particles .
the best - fit correlation timescale , @xmath510 , used in fig .
[ rms ] is in between the measured values of @xmath4 ( @xmath423 ) and @xmath511 ( @xmath512 ) .
this suggests that the temporal statistics of the flow velocity seen " by a large particle is in between that along a lagrangian trajectory and that at a fixed eulerian point .
it is interesting to note that , although a single - exponential function does not well fit either @xmath413 or @xmath60 ( 4.1 ) , our model prediction for the rms particle velocity with an exponential correlation is in good agreement with the simulation data .
we also computed the probability distribution of the 1-particle velocity .
the velocity of small particles is expected to be gaussian because it simply samples the 1-point pdf of the flow velocity , which is close to gaussian .
for large particles with @xmath9 , the velocity would also be gaussian because their equation of motion is essentially a langevin equation .
we find the 1-particle velocity pdf is indeed nearly gaussian at all @xmath1 .
the nearly gaussian pdf for the largest particles in the three selected snapshots also supports our assumption that these particles are relaxed at @xmath513 .
now we explore the statistics of 2-particle relative velocity in our simulation , focusing on the monodisperse case of equal - size particles . using the simulations data
, we can compute the joint probability distribution , @xmath514 , of the particle separation and the relative velocity as a function of @xmath1 ( see , e.g. , zaichik et al .
2003 , gustavsson and mehlig 2011 , hubbard 2013 ) .
the joint distribution is defined such that the number of particles located in a volume @xmath515 at a separation @xmath133 from a reference particle and moving at a relative velocity in the range @xmath516 $ ] is given by @xmath517 , where @xmath518 is the average number density .
once the particle statistics become isotropic , it is convenient to study the distributions , @xmath519 , @xmath520 , and @xmath521 , for the radial , tangential components and the 3d amplitude of the relative velocity .
the normalization of @xmath519 with respect to @xmath142 is given by @xmath522 , where @xmath523 is the so - called radial distribution function ( rdf ) .
the same normalization applies for @xmath520 and @xmath521 .
the rdf represents the overall probability of finding a neighbor with any relative speed , and is a measure of the spatial clustering of inertial particles ( see 7 ) . with the joint distributions and the rdf
, we define the relative speed pdfs as @xmath524 , @xmath525 , and @xmath526 . for the simplicity of notation ,
the dependence of the pdfs on @xmath13 is not explicitly indicated . a systematic study of these pdfs is given in 6.2 .
our simulation results for the variances of the pdfs are presented in 6.1 . for each species
( @xmath1 ) , we measure the relative velocity of particle pairs mainly at three distances , @xmath527 , @xmath528 and @xmath529 . these distances are below the resolution scale of the simulation . measuring the statistics at subgrid scales
is justified , as the subgrid flow velocity is reliably captured by the adopted tsc interpolation ( 5 ) . for each @xmath1 and @xmath13 , we search the simulation box for all particle pairs in a distance shell from @xmath530 to @xmath531 . for @xmath527 and @xmath528
, we set the shell thickness @xmath532 to @xmath533 . to increase the number of particle pairs hence the measurement accuracy at @xmath534
, we used a larger thickness , @xmath535 , which is likely the largest value one can reasonably adopt . if @xmath532 is increased further by a factor of 2 , it would be comparable to @xmath13 , and one may not safely attribute the measured statistics to a single particle distance . for @xmath534 and @xmath535 ,
the number of particle pairs in one snapshot is typically on the order of @xmath536 .
this number of pairs is about enough to provide sufficient statistics , although the measured relative speed pdf at @xmath534 already shows considerable noises at the tails ( see fig .
[ pmpdf ] in 6.2.3 ) .
a study of smaller values of @xmath13 is desirable .
however , at smaller @xmath13 , the number of particle pairs available becomes more limited and does not allow accurate statistical analysis . we take the statistical accuracy as priority , and restrict our data analysis to @xmath537 . to check the relaxation of the 2-particle statistics
, we examined the temporal evolution of the particle pair counts , or equivalently the rdf , and the rms relative velocity in our simulation .
the rdfs for all the 14 particle species reach a quasi steady state at @xmath488 , when the turbulent flow is fully developed ( fig . [ rms0 ] ) .
this further confirms that the dynamics of the particles in the first 10 species is well relaxed at @xmath538 . however , the steady state of the rdf is not a perfect indicator for the relaxation of the largest few particles .
these particles do not show significantly clustering ( see 7.1 ) , and their rdfs are close to unity at all times .
the rms of the relative velocity also reaches a quasi steady state at @xmath539 for essentially all particles .
as discussed in 5 , there is an uncertainty in the relaxation of the largest particles ( @xmath540 ) because the expected relaxation time is larger than the end time of the simulation .
our data shows that the rms relative velocity of the largest particles is about equal to @xmath541 times the 1-particle rms velocity , consistent with the expected relaxed state ( see 3.1 ) .
the relaxation of the largest particles is also supported by our later results that the rms relative speed obeys the expected @xmath177 scaling ( 6.1 ) and the relative velocity pdf approaches gaussian ( 6.2 ) . in summary , there is sufficient evidence that the 2-particle statistics is well relaxed toward the end of our simulation for all particles .
again we use the three snapshots at @xmath542 , @xmath543 , and @xmath471 in our analysis .
for the first 10 species with @xmath544 , the velocity of each particle is independent in the three snapshots .
averaging over these snapshots increases the measurement accuracy for both the 1-particle velocity and the 2-particle relative velocity .
we find that using the three snapshots also improves the relative velocity measurement for the four largest particles ( unlike the case of the 1-particle rms ; see 5 ) , even though the velocity of each individual particle is correlated for longer than @xmath545-@xmath546 . to effectively increase the independent sample size
, one should avoid the same two particles to appear as a pair in two successive snapshots selected ( see hubbard 2013 ) .
it turns out that , for the typical relative velocity of the four largest particles , a particle pair at @xmath310 separates to a significant distance and no longer makes a pair in @xmath547 . since the velocities of any two large particles with @xmath548 at any distance in a snapshot are independent ( 3.1 ) , the relative velocities of new pairs that appear in @xmath547 are typically independent of those in the earlier snapshot . therefore , including the three snapshots does increase the independent sample size and the measurement accuracy for the relative velocity of the largest particles . to decompose the relative velocity , @xmath135 , into radial and tangential components
, we set up a local coordinate system ( @xmath549 , @xmath550 and @xmath551 ) for each selected particle pair .
the direction @xmath549 is chosen to coincide with the particle separation , @xmath133 . in terms of the unit base vectors , @xmath432 , @xmath433 and @xmath434 , of the simulation grid , @xmath549 is expressed as @xmath552 , where @xmath553 , @xmath554 , @xmath555 , and @xmath556 .
the radial component is calculated as @xmath557 .
for the two tangential directions , we set @xmath558 and @xmath559 , which are obtained by two consecutive rotations of the original coordinates .
the first rotation is about @xmath434 by @xmath560 , which moves @xmath433 to @xmath550 , and the second one is about @xmath550 by @xmath561 , which further brings the original base vectors @xmath432 and @xmath434 to @xmath549 and @xmath551 .
we then calculate @xmath562 and @xmath563 .
the pdfs of @xmath564 and @xmath565 are found to be almost the same , as expected from the statistical isotropy by an arbitrary angle .
this changes @xmath550 and @xmath551
. however , from the statistical isotropy , @xmath566 and @xmath565 would be statistically invariant under this third rotation . ] .
we thus take the pdf of a tangential component , @xmath144 , to be the average pdf of @xmath564 and @xmath565 .
the variance of @xmath144 is calculated as @xmath567 .
we first study the root - mean - square of the relative velocity .
we test the prediction of the pp10 model and validate the physical picture revealed by the model .
[ 3drms ] shows the simulation result for the 3d rms , @xmath568 , of the relative velocity as a function of the particle inertia .
the data points correspond to the measured relative velocity at a distance of @xmath569 . on the bottom and top x - axises , we normalize the friction timescale to the kolmogorov timescale ( @xmath570 ) and the lagrangian correlation timescale(@xmath571 ) , respectively .
the left and right y - axises normalize the relative speed to the kolmogorov velocity and the 3d rms flow velocity , respectively . as a reminder , @xmath572 in our simulated flow .
similar normalizations are adopted in most figures in the rest of the paper .
the normalization to kolmogorov scales is a convention commonly adopted in the turbulence literature , which is convenient for the study of small particles with @xmath288 and the related phenomena at small length scales . on the other hand ,
the normalization to large - scale quantities is more useful for large particles , and may be more convenient for practical applications since observations constrain the large - scales properties of protoplanetary turbulence . at small @xmath1 , the 3d rms relative speed
is roughly constant , and its value is consistent with the s - t prediction , @xmath573 .
the relative speed starts to rise at @xmath16 , as the effect of the particle memory and the backward separation becomes important .
for the largest particles , we find that @xmath574 , with @xmath575 , in agreement with eq .
( [ largelimit ] ) for the large particle limit , @xmath9 . like the earlier result for the 1-particle rms velocity ,
this provides a validation for using the lagrangian correlation function for the trajectory correlation , @xmath81 , of large particles with @xmath9 , even though their trajectories may significantly deviate from lagrangian tracers .
the predicted @xmath179 scaling for inertial - range particles by various models is not observed due to the limited inertial range of the simulated flow .
, as a function of the stokes number , @xmath1 .
the data points show the simulation result .
the solid line is the prediction of the pp10 model ( eq . [ w2 ] ) using a bi - exponential temporal correlation @xmath203 and a two - phase separation with @xmath576 in the richardson phase .
the dotted line corresponds to the same model but with a single - exponential @xmath203 .
the line is barely visible because it almost coincides with the solid line .
the dot - dashed line assumes ballistic separation at all times .
the dashed line is the prediction of the simplified model , eqs .
( [ separatedintegral ] ) and ( [ ffactor ] ) , using single - exponential @xmath203 and a two - phase separation behavior with @xmath577 . the top x - axis and the right y - axis normalize the friction timescale and the relative speed to @xmath4 and @xmath474 , respectively .
the left and right dotted line segments show @xmath179 and @xmath177 scalings . ]
the solid curve in fig .
( [ 3drms ] ) is the prediction of the pp10 model , and it is obtained by numerically solving eq .
( [ w2 ] ) . in the computation , we used eq .
( [ sll ] ) for @xmath96 , eq .
( [ snn ] ) for @xmath97 , and eq . ( [ tr ] ) for @xmath232 , respectively .
the parameters in these equations were set to @xmath578 , @xmath579 , @xmath580 and @xmath581 . a bi - exponential form , eq .
( [ biexponential2 ] ) , is adopted for the temporal correlation @xmath203 .
we used a two - phase behavior for the particle separation backward in time ( 3.4.3 ) , connecting the ballistic and richardson phases at @xmath582 .
we set @xmath583 for @xmath584 , and then switch to the richardson s law , @xmath585 , for @xmath586 , where @xmath587 . to fit the simulation data , the richardson constant , @xmath297 , is set to 1.6 .
the solid line is in good agreement with the data points , confirming the validity of the physical picture of our model . adopting a larger @xmath297
could further improve the fitting quality at intermediate @xmath1 . in appendix
a , we investigate the backward separation of tracer particles in our simulated flow , and find that @xmath588 .
therefore , the @xmath297 value used in the solid line in fig .
[ 3drms ] is significantly larger than that of tracer particles .
there exist two possibilities .
first , the backward separation of inertial particles in the richardson phase is indeed faster than tracers .
second , the accuracy of our model for the trajectory structure tensor , @xmath189 , may be inadequate .
for example , approximating the correlation timescale , @xmath232 , in @xmath203 by the eddy turnover time at the scale of the particle separation , @xmath202 , is essentially a qualitative assumption .
also there is an order - of - unity uncertainty in the adopted parameter @xmath240 for the scaling of @xmath242 in the inertia range ( see eq .
( [ tr ] ) ) .
if a larger @xmath240 were adopted , we could obtain a good fit to the data with a smaller @xmath297 .
finally , in our model for @xmath189 , we neglected the correlation between the particle distance , @xmath202 , and the fluctuations in the flow velocity difference , @xmath589 , seen by the particle pair ( see 3.2 ) .
this tends to underestimate the relative velocity . the dotted line in fig .
( [ 3drms ] ) is the prediction of the same pp 10 model , but with a single - exponential @xmath203 ( eq . ( [ singleexponential ] ) ) .
the dotted line is actually not distinguishable from the solid line , confirming the earlier statement in 3.2.1 that our model prediction for the relative speed is insensitive to the function form of @xmath203 .
this leaves us some freedom for the choice of @xmath203 , as long as the correlation timescale is accurately esimated .
in particular , it provides a justification for approximating @xmath203 by the same function form , e.g. , a bi - exponential function , for all particles , although realistically the form of @xmath203 may have a dependence on the particle inertia .
the dot - dashed curve plots the prediction of the pp10 model assuming that the particle separation is ballistic with @xmath590 at all times .
the model is otherwise the same as the solid line .
a pure ballistic separation is not realistic , and we show it here just to illustrate whether the richardson phase provides important contribution to the relative velocity . at @xmath591 ,
the dot - dashed line significantly underestimates @xmath148 , and , from @xmath592 , it becomes close to both the data points and the solid line using a two - phase separation behavior .
a possible explanation for this is that , for @xmath591 , the relative velocity receives a significant contribution from the richardson phase , even though this phase occurs at times beyond the particle memory timescale , i.e. , at @xmath593 . in that case , accounting for this phase would be necessary for particles with intermediate @xmath1 .
however , the validity of the above interpretation is subject to future tests .
there is the possibility that the discrepancy between the dot - dashed line and the data points may be caused by various uncertainties in our model for the trajectory structure tensor , @xmath189 ( see above ) .
the dashed line is the prediction of the simplified model , eqs .
( [ separatedintegral ] ) and ( [ response ] ) , using a single - exponential @xmath203 . as before , the simplified model with a bi - exponential @xmath203 gives almost the same prediction .
the same two - phase separation as in the solid line for the original pp10 model is adopted .
for the simplified model , the best - fit @xmath297 is found to be @xmath6 , which is close to the @xmath297 values measured from tracer particles .
the simplified model also fits the data better for intermediate @xmath1 , although the assumption made in the model is physically not better than the original pp10 model .
the simplified model may be a preferred choice , as its prediction is easier to compute .
we find that the stokes number , @xmath594 , at which the rms relative velocity peaks is @xmath595 , corresponding to a friction timescale of @xmath596 ( or @xmath597 ) .
the peak value of the 3d rms relative velocity is @xmath598 , which is about half the 3d rms velocity ( @xmath474 ) of the flow .
we give an explanation for the behavior of the peak relative velocity using the qualitative analysis of our model prediction disucssed in 3.2.4 .
the analysis was based on the primary distance @xmath331 , estimated as @xmath599 .
@xmath331 generally increases with @xmath0 . around the relative velocity peak , @xmath600 and @xmath601 , and thus @xmath602 . from eq .
( [ tr ] ) , the correlation time , @xmath232 , at @xmath603 is about @xmath604 , which is close to @xmath4 . for @xmath605 , @xmath345 would be constant and @xmath606 .
consequently , the @xmath203 term in eq .
( [ w2 ] ) provides a factor of @xmath338 for all particles with @xmath605 ( see 3.2.4 ) . using the same analysis as in 3.2.4
, one can show that this factor causes the relative speed to decrease with @xmath0 , even though the structure functions @xmath607 and @xmath608 are still increasing with @xmath331 at @xmath609 ( see fig .
[ structureandpower ] ) . for particles with @xmath610 , both the structure functions and @xmath345 decrease with decreasing @xmath331 , and
thus the relative speed would decrease with decreasing @xmath0 .
therefore a peak forms at @xmath611 .
we find that , for particles with @xmath612 , the amplitude of the flow velocity difference at the primary distance ( @xmath602 ) is smaller than @xmath427 , and this is responsible for why the maximum relative velocity is significantly lower than the rms flow velocity .
the discussion here shows that the relative velocity is the largest for the particles whose primary distance @xmath331 corresponds to the size of turbulent eddies with lifetime @xmath606 .
clearly , the backward particle separation plays an important role in determining the peak stokes number , @xmath594 . the model of volk et al .
( 1980 ) and its later developments predict that the relative speed reaches the maximum when @xmath0 is equal to a large eddy time , @xmath613 ( e.g. , markiewicz , mizuno & volk 1991 , cuzzi and hogan 2003 , ormel & cuzzi 2007 ) . the definition of @xmath613 in these studies is different from the timescales we used , and it is not clear whether , using parameters appropriate for our simulated flow , these models may correctly produce a peak at @xmath614 .
another issue is that , in the volk et al .
model , the peak relative speed is predicted to be equal or close to the rms velocity of the flow .
this overestimates the relative speed around the peak by a factor of 2 . a physical problem of the volk et al .
model has been discussed in the introduction ( see pp10 for details ) . the performance of the volk et al .
( 1980 ) model may improve as the reynolds number of the flow increases , which will be tested by future higher - resolution simulations . in fig .
[ 3drmsscale ] , we plot the 3d rms relative velocity at different distances , @xmath13 .
the squares , circles and diamonds correspond to @xmath615 , @xmath616 and @xmath529 , respectively . in this distance range
, the relative velocity shows a @xmath13-dependence at @xmath617 , while it is independent of @xmath13 for @xmath618 particles . in the context of our physical picture ,
this is because the friction time , @xmath0 , of @xmath618 particles is long enough that the backward particle separation after a duration of @xmath48 is insensitive to the initial " value , @xmath13 .
on the other hand , the relative speed of smaller particles relies on the flow velocity difference they saw in the near past , when the particle separation was still dependent on @xmath13 . , at at @xmath615 ( squares ) , @xmath616 ( circles ) and @xmath529 ( diamonds ) .
solid and dashed lines are predictions of the pp10 model with bi - exponential @xmath203 and the simplified model with single - exponential @xmath203 , respectively . a two - phase separation connecting at @xmath619 is adopted in both models . in the pp10 model , the richardson constant @xmath297
is set to 1.6 , 1.3 and 1.0 for solid lines from top to bottom .
the corresponding @xmath297 values used in the simplified model are 1 , 0.7 and 0.5 , respectively .
the dotted line segments denote @xmath179 and @xmath177 scalings . ]
the solid and dotted lines are predictions of the pp10 model with bi - exponential correlation function @xmath203 and the simplified version with single - exponential @xmath203 , respectively .
the lines for @xmath620 have already been shown in fig .
( [ 3drms ] ) , and the richardson constant , @xmath297 , was set to 1.6 and 1.0 , respectively , in the two models . at smaller @xmath13 ,
the best - fit value of @xmath297 becomes smaller .
for the pp10 model , we adopted @xmath621 and @xmath622 for @xmath623 and @xmath624 , respectively .
the decrease of @xmath297 with decreasing @xmath13 is consistent with our result in appendix a for the tracer particle pair dispersion .
the backward separation of tracer pairs was found to be slower for smaller @xmath13 .
the @xmath297 value used in the simplified model also decreases with decreasing @xmath13 . in the dashed lines for @xmath615 , @xmath616 and @xmath624 ,
the value of @xmath297 is set to 1 , 0.7 and 0.5 , respectively .
the relative speed of the smallest particles ( @xmath625 ) in our simulation appears to be larger than the second smallest ones ( @xmath626 ) , especially at smaller @xmath13 .
slight dips are seen at @xmath626 ( fig .
[ 3drmsscale ] ) .
this is in contrast to the s - t formula , which predicts the relative speed at a given @xmath13 is constant at sufficiently small @xmath1 .
these dips are not expected from the physical picture of our model either , and their existence is thus questionable .
one possibility is that the rise of the relative speed toward @xmath625 is a numerical artifact .
this suspicion is based on the consideration that the trajectory integration in our simulation is likely less accurate for smaller particles .
the accuracy of the trajectory computation depends on the integration time step relative to the particle friction time .
since the time step is the same for all particles , the accuracy would be lower for smaller particles .
this suggests the rise of the relative speed toward @xmath625 may be caused by numerical errors in the trajectory integration .
we expect it to disappear as the computation accuracy increases .
this will be tested by future simulations with a better temporal resolution for the integration of small particles .
the s - t formula predicts that @xmath568 scales linearly with @xmath13 in the @xmath11 limit .
this linear scaling is not confirmed by our simulation result for the smallest particles . a rough power fit for the rms relative speed as a function of @xmath13 gives @xmath627 at @xmath628 .
this means that , for particles with @xmath629 , the s - t formula is already invalid at @xmath630 . at a given @xmath1 ,
a critical particle distance is expected , below which the linear scaling does not apply .
the physical reason is that , as @xmath13 decreases , the local flow velocity difference across @xmath13 becomes smaller , and it is easier for the particle memory of the flow velocity difference in the past to provide a significant contribution , which tends to invalidate the s - t prediction .
equivalently , at a given @xmath13 , the s - t formula is valid only below a critical @xmath1 . in fig .
[ 3drmsscale ] , the lines for @xmath630 show that , as @xmath1 decreases to @xmath631 , the relative speed predicted by our model is not flat yet , suggesting a significant contribution from the particle memory . at sufficiently small @xmath1 , the rms relative speed at a distance @xmath630 is expected to finally become constant . to verify this , a simulation of higher accuracy for small particles is needed to fix the problem of the artificial rise in the relative velocity toward @xmath625 . as mentioned earlier ,
the rise is expected to disappear as the computation accuracy for the smallest particles increases .
our model does not directly consider the sling effect ( falkovich et al .
2002 , falkovich and pumir 2007 ) or the related caustic formation ( wilkinson & mehlig 2005 ; wilkinson et al . 2006
; gustavsson & mehlig 2011 ) .
the effect of slings is usually explored for small particles with @xmath288 . as mentioned in the introduction
, it corresponds to crossing of particle trajectories that occurs at fluid streamlines with high curvature or local flow regions with large velocity gradient . in our physical picture ,
the effect of slings or caustics could be viewed as a contribution to the backward particle separation . in the sling events ,
the particle pairs come together from a farther distance than the average . based on the model of wilkinson et al .
( 2006 ) and gustavsson & mehlig ( 2011 ) , the frequency of slings or caustic formation increases with @xmath1 .
the effect also becomes more important as @xmath13 decreases , and would finally dominate over the s - t contribution at a sufficiently small @xmath13 .
falkovich and pumir ( 2007 ) showed that the sling effect is already significant at @xmath626 . in fig .
[ 3drmsscale ] , we see that our model prediction underestimates the relative velocity of @xmath632 particles at @xmath630 .
a likely reason is that the sling effect is not sufficiently reflected by the assumed backward separation behavior . in principle , the effect can be better incorporated into our model by directly and accurately evaluating the frequency of such events and their contribution to the backward separation .
we will discuss the effect of slings or caustics on the particle collision rate in details in 7 . for @xmath288 particles ,
the sling events are rare . at @xmath291 ,
these events become very frequent , and essentially all particle pairs at @xmath633 should be counted as sling or caustic pairs ( 7.2 ) . in our physical picture , this corresponds to the fact that the backward separation of @xmath291 particle pairs at a friction timescale ago is significantly larger the initial distance , @xmath13 . concerning the @xmath13-dependence of @xmath568 for @xmath634 particles seen in fig
. [ 3drmsscale ] , a fundamental question is whether the dependence disappears as @xmath13 further decreases below @xmath635 , or , equivalently , whether @xmath568 approaches a finite constant as @xmath15 .
based on the prediction of gustavsson & mehlig ( 2011 ) , it is possible that , for particles that exhibit significant clustering ( see 7.1 ) , the overall rms relative velocity may approach 0 as @xmath15 . in the case ,
@xmath568 for @xmath636 particles would not converge with decreasing @xmath13 , and thus in principle could not be resolved .
intuitively , the rms relative speed , @xmath568 , of intermediate particles with @xmath637 may converge at sufficiently small @xmath13 , while , for @xmath638 particles , @xmath568 may decrease to zero as @xmath15 .
the convergence of @xmath568 for @xmath634 particles needs to checked with larger simulations that allow accurate measurements at @xmath639 .
[ radialtangential ] shows the rms relative speeds in the radial ( circles ) and tangential ( diamonds ) directions for particle pairs at @xmath640 , @xmath616 and @xmath624 . at @xmath288 ,
the tangential rms speed , @xmath641 , is slightly larger ( by @xmath212 ) than the radial rms , @xmath642 .
this difference is considerably smaller than the prediction of the s - t formula , eq .
( [ saffmanturner ] ) , which indicates that for @xmath11 particles the tangential rms relative speed should be larger than the radial rms by a factor of @xmath541 .
our data implies that this prediction is not valid at least for particles with @xmath643 .
it remains to be checked whether the factor of @xmath541 difference between @xmath642 and @xmath641 would be recovered at smaller stokes numbers . in fig .
[ radialtangential ] , we see that @xmath642 and @xmath641 become exactly equal at @xmath644 . ; circles ) and tangential ( @xmath641 ; diamonds ) directions . from top to bottom , data points show simulation results at @xmath615 , @xmath616 , and @xmath529 , respectively .
lines are predictions of the pp10 model .
the solid lines adopt eq .
( [ randomdirection2 ] ) for the angular average of @xmath186 , which predicts that @xmath364 .
the particle separation behavior assumed here is exactly the same as in the solid lines in fig .
[ 3drmsscale ] for the 3d rms .
the richardson constant , @xmath297 , is set to 1.6 , 1.3 and 1.0 for @xmath615 , @xmath616 , and @xmath529 , respectively .
the dashed and dotted lines for @xmath620 are solutions of eq .
( [ wrwt ] ) for the radial and tangential relative speeds , respectively .
the two lines reproduce the s - t prediction for @xmath642 and @xmath641 at @xmath645 . ]
the solid lines in fig .
[ radialtangential ] correspond to the prediction of the pp10 model using eq .
( [ randomdirection2 ] ) for the angular average , @xmath646 , of the trajectory structure tensor .
the equation assumes that the direction of the particle separation @xmath208 at any time is completely random , and predicts that @xmath364 for all particles ( 3.2.2 ) .
this prediction is in good agreement with our simulation data .
the equality of the radial and tangential rms speeds for @xmath644 is expected , because the separation @xmath208 of these particles at a friction timescale ago is significantly larger than the initial distance , @xmath13 , and its direction is likely random with respect to @xmath133 . on the other hand ,
the near equality of @xmath642 and @xmath641 at @xmath647 is somewhat surprising . for @xmath620 , the backward separation of these particles does not contribute to make the 3d rms , @xmath648 , of the relative velocity significantly larger than the s - t prediction
this suggests that the near equality of @xmath642 and @xmath641 is due to a conversion of the relative velocity from the tangential to the radial direction .
the conversion is probably caused by the deviation of the particle trajectories from the fluid elements .
even though the deviation does not considerably change the 3d amplitude @xmath648 at @xmath649 , it could efficiently alter the direction of @xmath135 with respect to @xmath133 .
the trajectory deviation is stochastic , and thus tends to randomize the direction of @xmath135 and equalize its radial and tangential components .
this reduces the tangential - to - radial ratio .
the randomization effect is expected to be more efficient in the slings events , where the particles are shot out of the flow streamlines , and encounter the trajectories of other particles . at smaller @xmath13
, the contribution from the backward separation to the 3d rms of the relative velocity is larger , and the random direction of the particle separation in the past also tends to equalize the radial and tangential components . in the @xmath15 limit
, we would expect that @xmath642 and @xmath641 are exactly equal at all @xmath1 .
when computing the solid lines , we used a bi - exponential form for @xmath203 , and the separation behavior adopted here is exactly the same as for the solid lines in fig .
[ 3drmsscale ] for the 3d rms .
the solid lines shown here correspond to those in fig .
[ 3drmsscale ] divided by @xmath650 .
the richardson constant is set to 1.6 , 1.3 and 1.0 in the three lines for @xmath651 , and @xmath529 , respectively .
the dashed and dotted lines for @xmath620 are the solutions of eq .
( [ wrwt ] ) for the radial and tangential rms relative speeds , respectively .
( [ wrwt ] ) was derived from eq .
( [ randomdirection1 ] ) for @xmath652 , which assumes that the direction of the separation change @xmath267 ( rather than @xmath208 itself ) is random . when solving eq .
( [ wrwt ] ) , we used the same two - phase separation behavior ( with @xmath653 ) as in the corresponding solid line . at small @xmath1 ,
the dashed and solid lines reproduce the s - t prediction that @xmath654 . the discrepancy between the simulation data and the s - t formula implies that , for particles with @xmath655 , the direction of @xmath208 is more random than assumed in eq .
( [ randomdirection1 ] ) .
the dependence of the radial and tangential rms relative speeds on @xmath13 is similar to that of the 3d rms ( see fig .
[ 3drmsscale ] ) . in an attempt to roughly fit them as power - law functions of @xmath13
, we find that @xmath642and @xmath641 scale with @xmath13 as @xmath656 at @xmath657 .
similar to the 3d rms , the slight dips at @xmath626 for both @xmath642and @xmath641 may be due to a numerical artifact .
the simulations of wang et al .
( 2000 ) found that the tangential - to - radial variance ratio , @xmath658 , is @xmath659 at @xmath629 .
this is closer to the s - t prediction and larger than the corresponding value ( @xmath660 ) in our simulation .
this is probably because our flow has a much larger reynolds number .
although wang et al . ( 2000 ) claimed that the ratio is independent of @xmath402 based on several simulations with @xmath661 , it is not clear if this is also true at @xmath662 .
as speculated above , it is the deviation of the particle trajectories from the flow elements that tends to equalize the radial and tangential relative speeds of small particles .
clearly , the trajectory deviation would be larger in flow regions with larger velocity gradients , where the flow experiences a faster velocity change .
the probability of finding large flow velocity gradients hence large trajectory deviations increases with @xmath402 .
therefore , the tangential - to - radial ratio is likely smaller at higher @xmath402 . for @xmath288 particles ,
the sling events occur in regions with extreme flow velocity gradients , and the frequency of slings would increase with @xmath402 ( falkovich and pumir 2007 ) .
this also tends to reduce the tangential - to - radial ratio .
however , we can not rule out the possibility that the trajectory integration of the smallest particles in our simulation is not sufficiently accurate to allow an accurate measurement of @xmath658 at small @xmath1 .
so far , our analysis for the particle relative velocity included all particle pairs at given distances . however , not all pairs at a small distance lead to collisions . particles with a negative radial relative velocity , @xmath663 , approach each other and may collide , while particle pairs with @xmath664 move away from each other . since the final goal of our study is to examine the particle collisions ,
it is appropriate to split particle pairs at a given distance into two groups with @xmath663 and @xmath665 , respectively .
we refer to them as the minus and plus groups .
although only the first group is relevant for particle collisions , it is theoretically interesting to compare the two groups . for the radial component , @xmath142 , of the relative velocity ,
we denote the variances in the minus and plus groups as @xmath666 and @xmath667 , respectively . in terms of the pdf , @xmath668 , of @xmath142 ,
the variances are written as @xmath669 and @xmath670 .
we denote the pdfs of the tangential component in the minus and plus groups as conditional pdfs , @xmath671 and @xmath672 .
the tangential variances in the two groups are then given by @xmath673 .
similarly , for the 3d amplitude , @xmath674 , the minus and plus variances are expressed as @xmath675 , where @xmath676 are the pdfs of @xmath674 for approaching and separating pairs .
the pdfs , @xmath668 , @xmath677 , and @xmath678 , will be studied in 6.2 .
the data points in the left panel of fig .
( [ plusminus ] ) show the radial rms relative speeds of particle pairs at @xmath620 ( squares ) and @xmath624 ( circles ) .
the right panel plots the tangential ( squares ) and 3d ( circles ) rms speeds at @xmath569 . in both panels ,
the filled and open symbols correspond to particle pairs in the minus and plus groups , respectively , and the lines plot the overall rms relative velocities counting all particle pairs . if the velocity of @xmath11 particles closely follow the flow velocity , we expect that @xmath679 are determined by the variances of the longitudinal flow velocity increments , @xmath680 , for negative and positive @xmath681 , respectively .
the definition of @xmath680 is given in appendix b , and they correspond to the fluctuation amplitudes in the left and right wings of the pdf of @xmath681 . in appendix b , we find the ratio @xmath682 is 1.47 at the size , @xmath467 , of the computation cell .
this suggests that , in the @xmath683 limit , @xmath684 would be larger than @xmath685 by @xmath686 .
the simulation result confirms this expectation . at @xmath625 , @xmath684 is larger than @xmath685 by @xmath687 for both @xmath615 and @xmath529 ( see the left panel of fig .
[ plusminus ] ) . using a similar analysis to the tangential component
gives @xmath688 for @xmath11 , where @xmath689 are variances of the transverse flow velocity increment , @xmath104 , conditioned on the sign of the longitudinal increment @xmath681 ( see appendix b ) .
the ratio @xmath690 is found to be 1.16 at @xmath628 , consistent with the ratio of @xmath691 to @xmath692 at the grid cell size , @xmath467 ( see appendix b ) . for the the 3d amplitude ( @xmath693 ) ,
the rms ratio between the minus and plus groups is 1.18 at @xmath11 . as @xmath1 increases , the relative speed for the plus group first decreases slightly and reaches a minimum at @xmath694 in all cases with @xmath620 .
this can be explained by considering the effects of the particle memory and the particle separation backward in time .
particle pairs in the plus group with @xmath695 are coming from smaller distances , meaning that the separation of the particles was smaller in the near past .
as @xmath1 increases from 0.1 to 0.4 , the contribution from the particle memory of the flow velocity difference becomes more important , and this contribution tends to reduce the relative speed since the particle distance was smaller in the immediate past .
however , if we look back further into the past ( i.e. , at larger @xmath327 ) , the two particles may pass each other , and their distance would make a transition from decreasing to increasing .
this explains the increase of @xmath667 , @xmath696 , and @xmath697 at @xmath698 .
the minimum of @xmath667 for @xmath699 appears at @xmath700 instead of @xmath701 , because , for smaller @xmath13 , it takes a shorter time for the particle distance in the past to change from decreasing to increasing . for approaching particles in the minus group , the particle distance would increase monotonically toward the past
therefore , the relative speed for this group is expected to increase monotonically as @xmath702 increases from 0 to @xmath703 .
this is confirmed by the filled data points in fig .
[ plusminus ] , except the slight dips at @xmath626 .
these dips are not expected , and again may be caused by insufficient numerical accuracy in the trajectory integration of the smallest particles ( 6.1.1 ) .
[ plusminus ] shows that approaching particles tend to have a larger relative speed than separating ones .
the difference between the two groups first increases with @xmath1 and then decreases at @xmath704 . at @xmath705 , the rms relative speeds in the two groups are close and
coincide with the overall rms .
the reason is that , for these larger particles , the separation of the particle pairs in the two groups at a friction timescale ago becomes insensitive to the initial " condition around @xmath706 .
the asymmetry in the relative velocity of @xmath634 particles is related to the spatial clustering of these particles .
the fact that approaching pairs move faster than separating ones may imply that particles tend to cluster at small distances to a reference particle ( see also 7.1 ) .
an interesting question is whether the asymmetry found at @xmath310 would exist also in the @xmath27 limit .
we expect the asymmetry to decrease with decreasing @xmath13 because the difference caused by the different separation behaviors in the near past for approaching and separating pairs would be smaller at smaller @xmath13 . however , it is not clear whether it completely vanishes as @xmath27 .
for example , the asymmetry in the flow velocity difference , @xmath209 , persists at any tiny but finite @xmath13 , and it may leave an imprint on the relative velocity of small particles ( say , with @xmath707 ) . the actual behavior of the asymmetry as @xmath27 will be checked by future simulations that allow to resolve smaller @xmath13 .
the prediction of the pp10 model was made only for the overall rms , and it could be modified to give separate predictions for approaching and separating pairs if the different separation behaviors of the two groups are properly specified .
we also find that @xmath684 and @xmath708 almost coincide for all @xmath1 , suggesting that each relative velocity component would provide equal amount of collision energy . on average ,
the radial component contributes 1/3 collision energy , while the rest 2/3 is from the two tangential components .
an accurate estimate of the pdf of the particle collision velocity is important for modeling the growth and evolution of dust particles in protoplanetary disks .
as mentioned in the introduction , the outcome of particle collisions depends on the collision velocity , and due to the random nature of the turbulent - induced collision velocity , collisions of particles with exactly the same properties may have different outcomes , and thus using a single value , e.g. , the rms , for the collision speed of particles of a given size is insufficient .
the probability distribution of the collision velocity is needed to calculate the fractions of collisions resulting in sticking , bouncing or fragmentation . in this section ,
we explore the probability distribution of the particle relative velocity .
we will primarily show the pdfs at @xmath709 , where the statistical measurements are most sufficient and accurate .
the measured pdfs for @xmath23 particles already converge at @xmath24 . on the other hand , for smaller particles
the pdfs have an @xmath13-dependence at @xmath710 , and an appropriate extrapolation to @xmath15 using larger simulations will be needed for the application to dust particle collisions .
the physical picture of pp10 shows that the relative velocity of inertial particles depends on the flow velocity difference the particle pairs saw within a memory timescale or so .
this suggests that the statistics of the velocity difference in the carrier flow is crucial for the understanding of the relative velocity pdf of inertial particles .
therefore , we analyzed the pdfs , @xmath711 and @xmath712 , of the longitudinal and transverse velocity increments , @xmath681 and @xmath104 , as functions of the length scale , @xmath98 , in our simulated flow .
the results are discussed in details in appendix b. the flow velocity pdfs are used in the physical explanation for the pdf of the particle relative velocity as a function of particle inertia in 6.2.2 . in fig .
[ rpdfs ] , we show the pdf , @xmath668 , of the radial component of the relative velocity as a function of the stokes number . all the pdfs
are measured at a particle distance of @xmath569 .
the relative speed is normalized to the kolmogorov velocity , @xmath713 , and the 1d flow rms velocity , @xmath58 , on the bottom and top x - axises , respectively .
each pdf is normalized to its value at the central peak . in the left panel
, the thin dashed line corresponds to the pdf of tracer particles ( @xmath714 ) at @xmath569 .
the shape of this line is found to be identical to the pdf , @xmath715 , of the longitudinal flow velocity increment , @xmath681 , at the computational cell size ( @xmath716 ; see the top line in the left panel of fig .
[ flowpdf ] in appendix b ) .
this is expected as tracer particles exactly follow the flow velocity , and the shape of @xmath715 is independent of @xmath98 in the dissipation range ( appendix b ) .
the solid color lines of increasing width show the pdfs of larger particles .
this corresponds to the increase of the rms relative speed with @xmath1 for @xmath717 ( see figs .
[ radialtangential ] and [ plusminus ] ) . for @xmath718 ,
the pdf of @xmath142 has a negative skewness , which is inherited from the flow velocity pdf @xmath715 .
the pdf becomes symmetric at @xmath719 .
it is interesting to note that , as @xmath1 increases , the tails of the pdfs become broader , while the innermost part remains unaffected and the same as the pdf of the tracer particles .
as to be shown in 6.2.2 , the amplification of the pdf tails correspond to the effect of slings or caustic formation . due to the tail amplification ,
the overall pdf shape becomes fatter as @xmath1 increases to @xmath720 . with increasing @xmath1 , the amplification effect proceeds towards the inner parts of the pdf , leading to a sharp cusp - like shape at the center , especially for @xmath721 .
for particles with @xmath721 , the slope of the outer parts of the pdf tends to steepen when extending to higher tails , i.e. , the pdf shape becomes thinner at larger @xmath722 .
this thinning trend toward the high tails causes a decrease in the overall fatness of the pdf for @xmath1 above @xmath723 . in the right panel ,
the pdf becomes narrower as @xmath1 increases above @xmath724 , corresponding to the decrease of the rms relative velocity with @xmath1 in the large particle limit ( figs .
[ radialtangential ] and [ plusminus ] ) . for @xmath725 ( @xmath726 ) ,
the friction timescale is larger than the correlation timescale ( @xmath4 ) of the flow velocity at largest scales , meaning that the memory time of the flow velocity is shorter than the memory of the particles .
this induces a factor of @xmath338 in the relative velocity variance ( 3.2.4 ) , causing a decrease of the pdf width at larger @xmath1 .
the dotted line in the right panel is the gaussian fit to the pdf of the largest particles ( @xmath727 ; or @xmath728 ) in our simulation .
for these particles , @xmath0 is 54 times larger than @xmath4 , suggesting that the assumption of a gaussian relative velocity pdf applies only in the extreme limit @xmath9 .
[ tpdfs ] shows the pdf , @xmath729 , of a tangential component of the relative velocity conditioned on @xmath730 .
the measurement of @xmath731 only counts particle pairs approaching each other .
the figure is plot in the same way as fig .
[ rpdfs ] for the radial component .
again , the thin dashed line in the left panel is for tracer particles ( @xmath732 ) .
it corresponds to the pdf of the transverse difference , @xmath104 , of the flow velocity conditioned on @xmath733 .
the shape of @xmath734 for tracer particles at @xmath735 is close to the dashed line in the right panel of fig .
[ flowpdf ] for @xmath736 at @xmath737 ( appendix b ) .
the qualitative behavior of @xmath738 as a functions of @xmath1 is similar to that of @xmath668 . for @xmath739 particles , the radial pdf tails
are significantly amplified with respect to tracers ( fig .
[ rpdfs ] ) , while the conditional pdf of the tangential component almost coincides with the tracer pdf ( fig .
[ tpdfs ] ) . from the physical picture for the pdf behavior given in 6.2.2
, the effects of the particle memory and the backward separation tend to amplify the pdf tails of @xmath739 particles .
for the tangential pdf , this amplification effect is counteracted by the conversion of the relative velocity from the tangential to the radial direction , which reduces the pdf width of the tangential component .
as discussed in 6.1.2 , the conversion is caused by the deviation of inertial particle trajectories from the flow elements and the particle memory of the flow velocity in the past , which randomize the direction of @xmath740 relative to the particle separation , @xmath133 .
the conversion is expected to be more efficient at the pdf tails ( corresponding to the sling events ) .
it appears that the two opposite effects cancel out for the tangential pdf of @xmath741 particles , as it almost coincides with the dotted line for tracers . on the other hand ,
both effects broaden the pdf of the radial component , leading to significantly amplified tails with respect to tracers .
unlike @xmath742 , which has a negative skewness for @xmath718 particles , @xmath738 is symmetric at all @xmath1 . as mentioned earlier , the symmetry of the tangential pdf is expected from the statistical isotropy .
we find that the left wing of the radial pdf @xmath668 almost coincides with that of @xmath743 at all @xmath1 .
this is apparently due to the randomization of the relative velocity direction discussed above . on the other hand ,
the right wing of @xmath668 is narrower than that of @xmath744 , until it becomes symmetric at @xmath721 . to study particle collisions
, we are mainly interested in the pdfs for approaching pairs , i.e. , the left wing of @xmath668 and the entire tangential pdf , @xmath738 , conditioned on @xmath730 .
we give a more quantitative description for the shape of @xmath738 .
the description also applies to the left wing of @xmath668 , as it coincides with the left wing of @xmath745 .
we first quantify the fatness of @xmath729 by computing the kurtosis , defined as @xmath746 . at @xmath620 ,
the kurtosis for @xmath747 and @xmath748 particles is @xmath749 , which is already much larger than 3 for a gaussian pdf . with increasing @xmath1 , the kurtosis first increases due to the tail amplification .
it reaches a maximum value of @xmath750 at @xmath751 , indicating extremely high non - gaussianity .
the kurtosis decreases slightly to 32 at @xmath752 , and then drops rapidly and approaches @xmath753 for the largest particles ( @xmath501 ) .
this decrease corresponds to the thinning trend of the high pdf tails for the large particles .
we also measured the kurtosis for the pdfs at smaller @xmath13 , and found that , for @xmath634 particles , it keeps increasing as @xmath13 decreases to @xmath635 .
the pdfs of these particles are fatter at smaller @xmath13 because the effect of the tail amplification on the pdf shape is relatively stronger ( see more detailed discussions in 6.2.3 ) .
following sundaram & collions ( 1997 ) and wang et al .
( 2000 ) , we attempted to fit @xmath729 with stretched exponential pdf .
the generic stretched exponential function is given by , @xmath754 , \label{se}\ ] ] where @xmath755 is the gamma function .
the variance of @xmath756 is given by @xmath757 .
thus , to fit a given pdf by eq .
( [ se ] ) with a chosen @xmath758 , one can fix @xmath759 by the variance of the pdf .
the index @xmath758 controls the pdf shape , and smaller @xmath758 corresponds to fatter tails .
the pdfs for @xmath747 and @xmath748 at @xmath620 almost have the same shape , and both can be fit by a stretched exponential with @xmath760 .
this value of @xmath758 is consistent with that ( 0.7 ) used to fit the pdf tails of the flow velocity difference at @xmath761 ( see appendix b and fig . [ flowpdf ] ) . at @xmath762 , @xmath763 and @xmath764 ,
the best - fit @xmath758 is @xmath765 , @xmath766 , and @xmath767 , respectively .
the decrease of @xmath758 in the @xmath1 range from 0.1 to 0.78 indicates increasing fatness of the pdf .
the pdf shape at @xmath752 is very close to that at @xmath751 . for particles with @xmath768 ,
the pdfs are more complicated , due to the existence of sharp cusps at the center and the steepening trend of the pdf slope toward to the far tails .
these features can not be captured simultaneously by a single stretched exponential function .
it is , however , possible to fit these pdfs with a combination of two different stretched exponential functions for the cusp and the tails respectively .
we postpone a detailed study of fitting functions for these intermediate particles to a future work . to give a quantitative idea for the pdf shape of these particles ,
here we simply list the best - fit @xmath758 for the far tails without accounting for the central cusp . at @xmath769 , 6.21 , 12.4 , 24.9 and @xmath724 , the best - fit @xmath758 for the pdf tails
is found to be 1 , 1.1 , 1.3 , 1.33 , and 1.45 , respectively .
the stretched exponential fits for the pdf tails of @xmath770 ( @xmath771 ) particles are shown as dotted lines in figs .
[ rpdfs ] and [ tpdfs ] , where the index @xmath758 is set to @xmath772 .
such a 4/3 stretched exponential pdf was predicted by gustavsson et al .
( 2008 ) assuming an exact gaussian flow velocity field with kolmogorov scaling and a rapid temporal decorrelation .
an alternative derivation for the 4/3 stretched exponential is given in 6.2.2 using the physical picture of the pp10 model .
our derivation does not assume a short temporal correlation for the flow velocity , and is thus more general than that of gustavsson et al .
( 2008 ) .
starting from @xmath773 ( @xmath774 ) , the central cusp becomes sufficiently small , leading to simpler pdf shapes .
this allows the entire pdf to be satisfactorily fit by a single stretched exponential again .
the measured @xmath758 values for @xmath775 , @xmath776 , @xmath777 and @xmath778 are @xmath779 , @xmath780 , @xmath781 and @xmath782 , respectively . note that the pdf at @xmath540 is close to gaussian , but the best - fit value for @xmath758 is actually 1.9 instead of 2 .
a similar trend of the pdf fatness and the best - fit @xmath758 as a function of @xmath1 was found in previous studies with low - resolution simulations ( sundaram & collions 1997 ; wang et al .
2000 ) .
we give an explanation for the behavior of the relative velocity pdf using the physical picture of pp10 , which was illustrated in fig .
[ cartoon ] .
we first consider particles with @xmath783 . in
3.2.4 , we showed that the 3d relative velocity variance of these particles may be roughly estimated by @xmath784 , where @xmath785 is the flow structure tensor and @xmath331 is the primary distance .
for simplicity , we have neglected the effect of the temporal correlation function , @xmath203 , which may provide a factor , @xmath786 , of order of unity for particles with @xmath783 .
@xmath331 was estimated by @xmath787 , assuming a ballistic backward separation within a friction timescale .
this picture for the rms relative velocity can be generalized to understand the behavior of the full pdf as a function of @xmath1 .
consider a pair of particles at a distance @xmath13 at time 0 , and suppose their relative velocity is @xmath788 . applying the above physical picture to this particular pair
, the relative speed , @xmath788 , is estimated as @xmath789 , where @xmath790 is the flow velocity difference the two particles saw " at @xmath619 and @xmath791 is the primary distance of this pair .
we have used @xmath788 and @xmath790 to represent either the radial or the tangential component .
the generalized picture suggests that the particle relative velocity samples the pdf of the flow velocity difference in a certain way .
an immediate implication is that the particle relative velocity would inherit intermittency of the turbulent flow . assuming a ballistic separation again , @xmath791 is estimated by @xmath792 , where @xmath793 corresponds to the difference between the 3d separation speed of the particle pair and the 1d speed in the radial or tangential direction .
we point out that , for particles with @xmath794 , it may not be valid to assume the contribution to the particle relative speed is dominated by the ballistic separation phase .
as discussed in 6.1 , the richardson phase may provide a crucial contribution for these particles ( see fig . [ 3drms ] ) .
however , using the ballistic assumption to estimate @xmath791 would be sufficient for a qualitative understanding of the relative velocity pdf .
the above argument provides a satisfactory explanation for our simulation results for the relative speed pdf , @xmath795 , of particles with @xmath796 . at @xmath368 ,
the primary distance @xmath791 ( @xmath797 ) for particle pairs in the inner part of the pdf ( i.e. , at @xmath798 ) would be close to @xmath13 . as a result ,
the central pdf follows the pdf , @xmath799 , of the flow velocity difference at @xmath800 , as observed in the left panels of figs .
[ rpdfs ] and [ tpdfs ] . at the tails of @xmath795
, @xmath791 is larger , and @xmath788 samples the flow velocity pdf @xmath801 at larger @xmath98 .
this implies that higher tails broaden faster because @xmath799 is wider at larger @xmath98 .
the effect may be viewed as a self - amplification " of the pdf tails .
the tail amplification makes the overall shape of @xmath802 at @xmath288 considerably fatter than the pdf of tracer particles . as @xmath1 increases , @xmath791 becomes larger at the same value of @xmath788 , and the amplification " proceeds deeper into the inner part of the pdf , as seen in the left panels of figs .
[ rpdfs ] and [ tpdfs ] .
the overall pdf broadening appears to be driven by the tail amplification .
the amplification in the far pdf tails of @xmath288 particles actually corresponds to the effects of slings or caustic formation .
this is because the tail of @xmath795 is associated with local flow regions with large velocity gradients , which are indeed where the slings or caustics are expected to occur .
the tail amplification of @xmath288 particles thus corresponds to the caustic contribution to the collision kernel in the model of wilkinson et al .
( 2006 ) . as @xmath1 increases above 1
, the range of the central pdf that follows @xmath799 becomes narrower , and the outer parts continue to get more extended . as discussed in 6.2.1 , for @xmath803 ,
the pdf tails show slope changes as they extend to high values of @xmath804 .
this is because different parts of the relative velocity pdf samples the flow velocity difference pdf , @xmath799 , at different length scales . as the fatness of @xmath799 decreases with increasing @xmath98
( see appendix b and fig .
[ flowpdf ] ) , the shape of @xmath795 at higher tails becomes thinner .
this thinning trend occurs at smaller values of @xmath804 for particles with larger @xmath0 .
the trend explains why the overall shape of the pdf becomes less fat as @xmath1 increases above @xmath6 .
note that the central cusp for @xmath805 keeps a sharp shape , corresponding to @xmath806 at small @xmath98 .
the fact that the broadening of the pdf starts from the tail amplification is not captured by the pp10 model for the rms relative velocity ( 3.2 ) .
the model only considers the 2nd - order moments of the flow velocity increment , the particle separation and the particle relative velocity .
this essentially assumes that the pdf shape does not considerably change with @xmath1 , or the shape change of the pdf at the outer or tail parts does not have significant effects on the variance of the pdf .
this gives rise to uncertainties in the prediction for the rms relative velocity because the pdf @xmath807 is found to be very fat especially for @xmath16 .
even the far tails give considerable contribution to the variance .
the tail amplification also provides evidence for a positive correlation between the fluctuations in the flow velocity increment seen " by the particle pair and the particle separation .
the tails of @xmath795 correspond to the pdf tails of both the flow velocity difference , @xmath790 , and the the primary distance , @xmath791 .
in other words , in flow regions with @xmath790 larger than its rms value , the backward separation of particles is also faster than the rms separation rate .
as discussed in 3.2 and 6.1 , the pp10 model neglects this correlation , and thus tends to underestimate the rms relative speed .
the effect of the pdf tail amplification on the variance of the relative speed may be incorporated in the pp10 model if the correlation between @xmath790 and the particle separation is properly accounted for . as mentioned in 6.1 , accounting for this correlation , our model may fit the rms relative velocity with a smaller richardson constant . in principle , if the pdf , @xmath799 , of the flow velocity increment as a function of the scale @xmath98 is provided , one can derive the particle relative velocity pdf as a function of @xmath0 .
for illustration , we consider a simplified example .
we assume the flow velocity is exactly gaussian , i.e. , @xmath808 where @xmath809 is the flow structure function or the variance of @xmath790 at @xmath98 . to estimate the pdf , @xmath810 , of the particle relative speed @xmath788
, we ask the question what the probability is for two nearby particles to see a flow velocity difference of @xmath788 at a friction timescale ago , i.e. , at @xmath294 .
the probability is roughly estimated by @xmath811 .
using eq .
( [ gaussianu ] ) for @xmath812 and setting @xmath813 , we have , @xmath814 , \label{pdfmodel}\ ] ] where it is assumed all uncertainties in the rough estimate can be absorbed in a parameter @xmath815 .
if @xmath791 is in the inertial range of the flow , we may apply the kolmogorov scaling @xmath816 . further assuming that @xmath815 is independent of @xmath788
, we find that eq .
( [ pdfmodel ] ) corresponds to a stretched exponential with @xmath817 ( see eq .
( [ se ] ) ) at @xmath818 .
this suggests that the relative speed of inertial - range particles would be non - gaussian even if the flow statistics were exactly gaussian
. this non - gaussianity originates purely from the particle dynamics , and is thus distinct from that inherited from the intermittency of the turbulent flow . in other words , we identified two sources , namely , the turbulent intermittency and the particle dynamics , that contribute to the non - gaussianity of the particle relative velocity .
the predicted stretched exponential with @xmath819 was found to satisfactorily fit the pdf tails of @xmath820 ( @xmath821 ) particles ( see dashed lines in the left panels of figs .
[ rpdfs ] and [ tpdfs ] ) . for these particles ,
the two assumptions made in the derivation of the stretched exponential , i.e. , gaussianity and kolmogorov scaling of the flow velocity , are both satisfied .
we point out that these assumptions are strong , and thus the validity of the 4/3 stretched exponential is quite limited .
in fact , the prediction applies only to particles around the peak stokes number , @xmath611 .
as discussed in 6.1 , for particles with @xmath822 in our simulation , the typical primary distance is around @xmath823 . from fig .
[ flowpdf ] ( appendix b ) , we see that , above this length scale , the pdf of the flow velocity increment is close to gaussian . therefore , the gaussian assumption made in eq .
( [ pdfmodel ] ) is valid for @xmath824 .
also , fig .
[ structureandpower ] shows that @xmath823 is toward the end of but still within the inertial range of the flow , meaning that , only for particles with @xmath825 , can one apply the kolmogorov scaling around the primary distance , @xmath791 .
these suggest that the two assumptions are simultaneously met only at @xmath822 .
our finding that the 4/3 stretched exponential fits the pdf tails of @xmath820 particles confirms the validity of our physical picture .
the 4/3 stretched exponential can also acceptably fit the pdf tails of @xmath826 particles , but not for other particles . at the central part of the pdf of @xmath820 ( or 12.4 ) particles , both assumptions beak down , and the 4/3 stretched exponential does not apply .
we next consider large particles with @xmath824 .
the friction time of these particles is much larger than @xmath4 , and , accounting for the effect of the memory time of the flow velocity , the relative velocity of a given particle pair is roughly estimated by @xmath827 ( see 3.2.4 ) . due to the large friction time
, @xmath791 is typically comparable to or even larger than the integral scale , @xmath108 , of the turbulent flow , and thus @xmath801 at @xmath828 is close to gaussian .
since the flow velocity seen " by @xmath824 particles is typically gaussian , the pdf shape for their relative velocity is simpler than particles with intermediate @xmath0 ( see figs .
[ rpdfs ] an [ tpdfs ] ) . using eq .
( [ gaussianu ] ) and the same analysis that led to eq .
( [ pdfmodel ] ) , we find @xmath829 for @xmath830 .
the structure function , @xmath809 , starts to become constant at @xmath831 .
therefore , as @xmath702 increases , the typical @xmath791 increases , and @xmath832 becomes less dependent on @xmath791 or @xmath788 . as a consequence
, the shape of the relative velocity pdf becomes less fat . in the limit @xmath833 , @xmath834 , and @xmath795
finally approaches a gaussian pdf with a variance @xmath835 . as observed in figs .
[ rpdfs ] and [ tpdfs ] , a nearly gaussian pdf is indeed observed for the largest particles in our simulation .
the physical interpretation of the relative velocity pdf at the beginning of this subsection suggests that the pdf can be split into two parts .
we use the pdf of @xmath142 as an example , and , in particular , we consider the left wing of @xmath668 , corresponding to approaching particle pairs .
we divide the pdf into two parts using a critical value , @xmath836 .
we choose @xmath836 such that the inner pdf at @xmath837 follows the flow velocity difference pdf , while in the outer part ( @xmath838 ) the effects of the particle memory and the backward separation dominate . at a given @xmath13 ,
@xmath836 is roughly estimated by @xmath839 .
this critical value is consistent with the physical picture of falkovich et al .
( 2002 ) for the sling effect .
falkovich et al .
( 2002 ) showed that the velocity gradient of small particles ( @xmath288 ) blows up in a finite time once it exceeds @xmath840 , leading to the sling events . at a given scale , @xmath13 , this gradient criterion can be written as @xmath841 .
the model of gustavsson & mehlig ( 2011 ) suggests the same criterion for caustic formation .
a similar critical value was proposed by hubbard ( 2013 ) for inertial - range particles . following wilkinson et al .
( 2006 ) , we name the central part ( @xmath842 ) and the tail part ( @xmath843 ) of the pdf as the continuous ( or s - t ) part and the caustic ( sling ) part , respectively . the terminology is based on the geometry in the position - momentum phase diagram of inertial particles ( see fig . 1 of gustavsson and mehlig 2011 ) . because the two parts have different scaling behaviors with @xmath13 , the division is especially useful for the prediction of particle collisions in the @xmath15 limit , appropriate for applications to dust particles . in 7 , we evaluate the collision kernel and show the contribution from the continuous part vanishes as @xmath15 .
only the caustic particle pairs contribute to the collisions of nearly point - like particles .
therefore , to obtain the collision velocity pdf for the application to dust particles , one may exclude the contribution of the continuous part , and meanwhile push @xmath13 to as small values as possible ( e.g. , hubbard 2013 ) .
a detailed study of this topic will be conducted in a future work .
our study for the rms relative velocity in 6.1 did not split the particle pairs into the two types , and the main purpose was to to understand the general physics of turbulence - induced relative velocity and to validate the physical picture of pp10 . to see the shape of the pdf more clearly
, we show in fig .
[ rpdfnormalized ] the pdf of the radial component normalized to have unit variance .
the radial relative speed is normalized to its rms value , @xmath642 .
the solid , dashed and dotted lines lines are the pdfs at @xmath620 , @xmath616 and @xmath624 , respectively , and the curves of different colors correspond to different @xmath1 . the bottom curves in the left and right panels plot the actual pdf values for @xmath844 and @xmath540(@xmath728 ) , respectively . for clarify ,
the pdf curves are shifted upward by a factor of 16 for each larger @xmath1 in the left panel or each smaller @xmath1 in the right panel .
the asymmetry of the pdfs at @xmath845 is clearly seen .
as @xmath1 increases from 0.1 to 1.55 , the central part of the normalized pdf @xmath846 becomes sharper . before normalization ,
the innermost part of the pdf follows the pdf of the flow velocity difference , and is thus essentially the same for particles in the range @xmath847 ( see fig .
[ rpdfs ] ) . since the rms of @xmath142 increases with @xmath1 due to the tail amplification , normalizing @xmath142 by its rms tends to make the central part of the pdf sharper . at @xmath769 and @xmath848 , the central cusp in the normalized pdf
is very sharp . for @xmath1 above 3.11 ,
the outer pdf parts become less fat with increasing @xmath1 .
the shape of the normalized pdf is helpful to understand the behavior of the ratio @xmath849 with @xmath1 discussed in 7.1 .
it is interesting to note that , as @xmath1 increases from 0.1 to 0.39 , the pdf , @xmath846 , at small to intermediate @xmath850 in the right wing decreases .
this corresponds to the decrease of @xmath685 in the @xmath1 range from 0.1 to 0.39 ( see the left panel of fig .
[ plusminus ] ) .
the physical reason is that the right wing corresponds to separating particle pairs , and the particle distance decreases toward the near past .
this leads to a decrease in the primary distance , @xmath791 , for separating pairs with small to intermediate @xmath850 , as @xmath1 increases from 0.1 to 0.39 . for the far right tail with large @xmath142
, the particle pairs may quickly move past each other , and their distance starts to increase within a friction time in the past .
this explains why the far right tail of @xmath851 particles becomes slightly broader than that for @xmath844 .
as @xmath1 increases to @xmath763 , the pdf at immediate to large @xmath850 in the right wing is significantly amplified , while the effect of the initial " decrease of the particle distance is still visible at small positive @xmath850 . for @xmath852 ,
the particle memory is longer , and the initial " separation phase does not cause a significant difference in the primary distances @xmath791 of separating and approaching particle pairs .
the two wings become almost symmetric at @xmath721 . for particles with @xmath853 ,
the normalized pdf has a dependence on the distance @xmath13 for the range of @xmath13 shown here .
as @xmath13 decreases , the central part of the normalized pdf becomes sharper , and the outer parts become slightly broader , leading to a fatter pdf shape at smaller @xmath13 . before the normalization ,
the central part of the pdf follows the flow velocity difference at @xmath13 , and its width thus decreases linearly with @xmath13 . on the other hand ,
the dependence of the tails on @xmath13 is weaker because the primary distance , @xmath791 , for particle pairs in the outer parts of the pdf has a larger contribution from backward separation . also , as @xmath13 decreases , the contribution from the outer parts of the pdf to the variance increases .
consequently , normalizing the pdf to unit variance gives a sharper central part and broader tails at smaller @xmath13 .
the fattening of the pdf with decreasing @xmath13 can also be viewed as due to the relatively larger contribution of caustics at the tails . the critical value @xmath854 decreases with decreasing @xmath13 , suggesting that , for a given @xmath1 , there are more caustic pairs at smaller @xmath13 .
we also observe that the asymmetry in the two wings decreases with decreasing @xmath13 .
however , it is remains to be verified whether it exactly disappears as @xmath15 . at @xmath855 ,
the pdf is essentially independent of @xmath13 , and thus directly applicable for dust particle collisions . for these larger particles , @xmath791
is mainly contributed by the backward separation even for pairs lying around the central part of the pdf , and thus the pdf is independent of @xmath13 for @xmath142 in any range . the left panel of fig .
[ pmpdf ] shows the pdfs of the tangential relative speed at @xmath615 , 0.5 , and @xmath529 for three values of @xmath1 . at smaller @xmath13 , the number of particle pairs available is smaller , leading to a decrease in the sample size . as a consequence ,
the pdfs become quite noisy at @xmath856 , especially for small particles .
for the @xmath13 range shown here , the pdf width decreases with decreasing @xmath13 for @xmath857 particles .
the difference between the pdf tails at @xmath635 and @xmath858 appears to be smaller than that between @xmath858 and @xmath19 , indicating relatively stronger contribution from caustic formation at smaller @xmath13 . like the radial pdf ( fig .
[ rpdfnormalized ] ) , the pdf becomes independent of @xmath13 for @xmath855 . in the right panel of fig .
[ pmpdf ] , we compare the pdfs of the tangential relative velocity for approaching ( @xmath738 ) and separating ( @xmath859 ) particle pairs at @xmath620 . for @xmath860 ,
the pdf of approaching particles is broader than the separating ones , consistent with our earlier result for the rms relative speeds , @xmath861 ( see the right panel of fig .
[ plusminus ] ) . again
, this is because , for a given initial " value @xmath13 , the distance of approaching particles was larger in the near past than the separating ones .
therefore , the pdf of the relative velocity for approaching pairs samples the pdf , @xmath799 , of the flow difference at larger @xmath98 . since the width of @xmath799 increases with @xmath98 , @xmath738
is expected to be broader than @xmath859 . at @xmath855 ,
the pdfs for approaching and separating pairs are almost equal . for these larger particles ,
the primary distance at @xmath294 is insensitive to the initial conditions around @xmath320 .
although @xmath859 for separating particle pairs is not relevant for particle collisions , a comparison of @xmath738 with @xmath859 provides an interesting illustration for the role of the backward separation of particle pairs in determining their relative velocity .
[ 3dpdf ] plots the pdf of the 3d amplitude , @xmath674 , of the relative velocity for approaching particle pairs at @xmath620 . for each @xmath1 ,
the amplitude @xmath674 is normalized to the rms value @xmath862 , i.e. , @xmath863 , so that all the normalized pdfs have unit variance .
the rms of the 3d amplitude , @xmath862 , for approaching pairs has been shown in fig .
[ plusminus ] .
the left and right panels of fig .
[ 3dpdf ] show simulation results for small ( @xmath864 ) and large ( @xmath865 ) particles , respectively .
the thin dashed line in the left panel corresponds to approaching tracer particles ( @xmath732 ) , while the dashed line in right panel is the normalized pdf for a gaussian vector with three independent components of equal variance .
the pdf for tracer particles in the left panel is already highly non - gaussian , as can be seen from a comparison with the dashed line in the right panel . in the left panel , the degree of non - gaussianity increases as @xmath1 increases from @xmath866 to @xmath6 . at larger @xmath1 , the pdf peaks at smaller @xmath867 .
the pdf around the rms value ( i.e. , @xmath868 ) decreases with increasing @xmath1 , and more probability is distributed toward smaller and larger values of @xmath867 .
this corresponds to the sharpening of @xmath668 and @xmath869 in the central part and the broadening of the tail parts in this @xmath1 range ( see figs .
[ rpdfs ] , [ tpdfs ] and [ rpdfnormalized ] ) .
the trend is reversed as @xmath1 increases further above @xmath870 .
the peak of the pdf moves back to around the rms value , @xmath868 , at @xmath871 ( @xmath872 ) .
the pdf eventually approaches gaussian in the limit @xmath873 .
however , note that , even for @xmath540 ( @xmath728 ) particles , the pdf shows a difference from the gaussian distribution at small relative speeds .
we find that , for @xmath1 in the range from 0.78 to 6.22 , the pdf shows an extended power - law range at intermediate values of @xmath867 .
for example , at @xmath874 , the pdf goes like @xmath875 in the range @xmath876 .
the slope of the pdf in the power - law range becomes shallower with increasing @xmath1 . for @xmath877 , @xmath878 and @xmath879 ,
the power - law exponent of the pdf in the intermediate @xmath867 range is @xmath880 , @xmath881 , and @xmath882 , respectively .
the pdf of @xmath674 can be easily computed from the pdfs of the radial and tangential components , if the three components are completely independent . in that case
, one may obtain fitting functions for @xmath883 using the fitting functions discussed earlier for the pdfs , @xmath668 and @xmath869 , of approaching particle pairs .
one could also directly fit the pdf of @xmath674 with simple function forms or tabulate it as a function of @xmath1 .
we will provide fitting functions for @xmath884 in a future paper .
the strong non - gaussianity in the amplitude of the relative velocity has interesting implications for the growth and evolution of dust particles in planetary disks .
we assume that the shape of the collision velocity pdf of dust particles at @xmath15 is more or less similar to the measured pdfs at @xmath21 in our simulation , even though the pdfs of @xmath885 particles have not converged at @xmath886 .
the shape of the pdfs shown in fig .
[ 3dpdf ] suggests that there would more collisions with extremely large ( @xmath887 ) or small ( @xmath888 ) relative speeds than estimated from a gaussian distribution . the higher probability for collisions with small relative speed would favor sticking , while there are also more collisions that would result in fragmentations .
the competition of the two opposite effects would determine whether the non - gaussian pdf of the collision velocity accelerates or slows down the particle growth .
a coagulation model incorporating the non - gaussian statistics of the collision speed would give a more realistic prediction for the evolution of the size distribution of dust particles .
the prediction of the collision rate is one of the main goals of our study of the particle relative velocity .
if the mean number density of inertial particles of a given size is @xmath889 , the collision rate per unit volume between these identical particles is estimated by @xmath890 , where @xmath755 is the collisional kernel and the factor 1/2 is used to avoid counting the same pair twice .
saffman and turner ( 1956 ) presented two formulations for the collision kernel .
the formulations were based on spherical and cylindrical geometries , respectively , and were thus named as the spherical and cylindrical formulations by wang et al .
( 1998 ) . for the spherical formulation ,
we make use of the joint distribution , @xmath519 , of the particle distance and the radial relative speed , defined at the beginning of 6 .
the collision kernel for identical particles is written as ( e.g. , saffman and turner 1956 ) , @xmath891 where @xmath892 is the diameter of the particle , and the integral limits include only particle pairs moving toward each other . using the definition of the radial pdf ,
@xmath668 as @xmath893 , eq . ( [ sphericalformal ] )
can be expressed in a simpler form , @xmath894 , where @xmath895 is the inward flux of particles toward a given reference particle defined as @xmath896 .
the particle statistics become stationary when the dynamics is fully relaxed . in the steady state , the inward flux , @xmath897 , is equal to the outward flux @xmath898(@xmath899 ) ( wang et al .
2000 ) , which is confirmed by our simulation data , of approaching and separating pairs , one can also define @xmath900 and @xmath901 . using the same reasoning in 6.1.3
, we expect @xmath902 for inertial particles , which is confirmed by the simulation data . together with the steady - state condition @xmath903
, we have @xmath904 .
this means that there tend to be more particles coming out from a reference particle , suggesting a larger particle density at small distances to the reference particle .
this provides an interesting physical perspective for the origin of particle clustering . ] .
this is because the average radial velocity @xmath905 , as expected from statistical isotropy , and @xmath906 by definition .
we thus have @xmath907 where @xmath908 is the ensemble average of the absolute value of the radial relative velocity .
the collision kernel can then be written as ( wang et al .
2000 ) , @xmath909 the cylindrical formulation assumes that all particles inside a cylinder of length @xmath910 located at a distance @xmath892 from a given particle will collide with the particle in a time interval @xmath911 .
similar to eq .
( [ spherical ] ) , the cylindrical collision kernel can be written as @xmath912 , with @xmath913 being the average of the 3d amplitude of the relative velocity . to evaluate the collision kernel
, one can start directly from eq .
( [ sphericalformal ] ) using the joint distribution , @xmath519 .
however , previous studies have shown that considering the rdf and @xmath914 separately provides interesting insights on the estimate of the particle collision rate ( e.g. , sundaram & collins 1997 , wang et al .
2000 ) . in 7.1 , we compute the rdf and the absolute average of the relative speed from our simulation data .
the rdf represents the probability of finding a neighboring particle at a distance of @xmath13 with any relative velocity , and is a measure for the spatial clustering of the particles . due to their finite inertia , inertial particles
do not exactly follow the flow velocity and have been found to exhibit inhomogeneous distribution even in incompressible turbulence .
turbulent clustering of inertial particles has been extensively investigated in the literature ( see e.g. , maxey 1987 , sundaram and collins 1997 , wang et al .
2000 , cuzzi et al .
2001 , hogan & cuzzi 2001 , balkovsky et al .
2001 , zaichik et al .
2003 , falkovich and pumir 2004 , cuzzi et al .
2008 , pan et al .
the general physical interpretation for turbulent clustering is that vortical structures in turbulent flows tend to expel inertial particles .
the particles are pushed out of high - vorticity regions by the centrifugal force , leading to the formation of clusters in strain - dominated regions .
, as a function of the stokes number at @xmath615 ( solid ) , @xmath616 ( dashed ) and @xmath529 ( dotted ) . ] fig .
[ rdf ] plots the rdf as a function of @xmath1 at @xmath615 , 0.5 and @xmath529 .
consistent with previous studies , the rdf is largest for @xmath16 particles , whose friction timescale couples with the smallest scale of the turbulent flow . for @xmath13 in the dissipation rage
, the rdf increases toward smaller @xmath13 as a power - law , i.e. , @xmath915 .
the scaling exponent @xmath916 peaks at @xmath16 , and approaches zero in the limits @xmath11 and @xmath917 .
we measured @xmath916 using the values of @xmath13 shown in fig .
[ rdf ] , and found that @xmath918 for @xmath751 , consistent with the result of pan et al .
the interested reader is referred to pan et al .
( 2011 ) for the scaling exponent , @xmath916 , as a function of @xmath1 . in the left panel of fig .
( [ absolute ] ) , we plot the simulation result for @xmath919 at @xmath651 , and @xmath624 . for comparison
, we also show the data ( circles ) for the rms of the radial relative speed at @xmath620 .
qualitatively , @xmath919 as a function of @xmath1 and @xmath13 is similar to the rms . it is smaller than the rms , as it corresponds to the 1st - order moment of the pdf , @xmath668 , of @xmath142 .
most theoretical models , including our own , for the particle relative velocity are based on the computation of the variance ( e.g. , @xmath920 ) , and can not be directly applied to predict @xmath908 .
the conversion between @xmath920 and @xmath908 relies on the shape of the pdf of @xmath142 , which is difficult to predict . we thus did not attempt to fit @xmath908 with a model prediction .
similar to the s - t formula ( eq . [ saffmanturner ] ) for the variances of the relative velocity , we would predict that @xmath921 in the @xmath11 limit , where @xmath922 is the absolute average of the longitudinal flow velocity increment . at @xmath923 , @xmath922
is expected to scale linearly with @xmath98 .
we find that , for @xmath11 , @xmath924 , which is slightly shallower than the linear scaling .
this is likely caused by the contribution from the effect of slings or caustics .
the scaling is steeper than @xmath925 for the radial rms velocity ( see 6.1.2 ) , suggesting that @xmath908 follows the flow velocity scaling better .
the behavior of @xmath922 at small @xmath1 also appears to be more regular than that of @xmath642 .
for all three values of @xmath13 , the @xmath908 curves become flat at @xmath926 .
the likely reason is that @xmath908 represents statistics at a lower order than the variance ( or rms ) , and is thus less affected by the rare and extreme sling events , or by the numerical uncertainty in the particle trajectory computation .
the inset of the left panel shows the ratio of the absolute average to the rms .
the ratio depends on the pdf shape of @xmath142 , and particularly on the central part of the pdf because both @xmath908 and the rms are lower - order moments . as a reference , if the pdf @xmath668 is gaussian , we have @xmath927 ( e.g. , wang et al . 2000 ) , and for an exponential pdf it is equal to @xmath928 . the ratio from a gaussian pdf was usually used to convert the model predictions for the rms to @xmath908 ( wang et al . 2000 , zaichik et al . 2003 , 2006 ) .
generally , the ratio is smaller if the central pdf is sharper and the tails are fatter . as seen in the inset , the ratio reaches a minimum at @xmath929 , corresponding to a maximum fatness of the pdf at @xmath929 ( see 6.2.3 ) . at @xmath930 ,
the minimum is smaller than 0.45 , corresponding to highly non - gaussian pdf . at @xmath931 ,
@xmath932 reaches 0.78 , close to the expected value for a gaussian pdf . for the range of @xmath13 shown here
, the ratio also decreases with decreasing @xmath13 for @xmath617 , which is expected from the trend of the pdf shape with @xmath13 for these particles ( 6.2.3 ) .
our simulation result for @xmath913 is shown in the right panel of fig .
( [ absolute ] ) , which is very similar to the left panel for @xmath914 .
at @xmath933 , we find @xmath913 also scales as @xmath934 with @xmath13
. the ratio of @xmath913 to @xmath568 also shows a dip at @xmath16 .
the ratio approaches 0.9 at the largest @xmath1 , as expected from a 3d gaussian distribution . .
the solid lines plot @xmath935 for the spherical formulation , and the dashed lines show @xmath936 for the cylindrical formulation .
the lines from bottom to top correspond to normalized kernels measured at @xmath937 , @xmath616 and @xmath569 , respectively .
the dotted line segment corresponds to a @xmath177 scaling . ] in fig .
[ product ] , we plot @xmath938 and @xmath939 normalized to @xmath940 .
the normalized kernels correspond to the product @xmath941 ( solid lines ) for the spherical formulation and @xmath942 ( dashed lines ) for the cylindrical formulation .
we will also refer to these products as the collision kernel per unit cross section .
at each @xmath13 , the solid and dashed lines almost coincide , meaning that @xmath938 and @xmath939 are nearly equal at all @xmath1 and @xmath13 .
this suggests that @xmath943 since @xmath944 .
the two collision kernels have a noticeable difference only at @xmath945 , where @xmath938 is smaller than @xmath939 by @xmath946 .
this is consistent with the result of wang et al .
wang et al . ( 2000 ) also showed that the spherical formulation provides an almost exact description for the particle collision rate .
the near equality of @xmath938 and @xmath939 at all @xmath1 and @xmath13 suggests that one can apply either formulation to evaluate the collision rate .
[ product ] shows that @xmath947 and @xmath948 are independent of @xmath13 for @xmath704 .
apparently , this is because , at @xmath644 , the @xmath13-dependences of the rdf @xmath949 and the absolute average @xmath950 ( or @xmath913 ) almost cancel out ( see figs . [ rdf ] and [ absolute ] ) .
a more interesting perspective is that the inverse scaling of @xmath949 and @xmath914 is expected from the intuition that the normalized kernel approaches a finite constant , corresponding to a finite collision rate , at sufficiently small @xmath13 .
therefore , the scalings of @xmath949 and @xmath914 must cancel out once the kernel converges ( see more discussions in 7.2 ) . at @xmath951 ,
@xmath949 increases with decreasing @xmath13 as @xmath952 , while both @xmath950 and @xmath913 scale with @xmath13 as @xmath953 .
note that , if the @xmath954 scaling of the relative speed persists in the @xmath27 limit , it does mean the collision energy is zero for nearly - point particles .
this is because @xmath950 or @xmath913 does not represent the collision energy .
the average collision energy per collision is expected to be finite even if @xmath950 or @xmath913 approaches zero at @xmath28 ( see 7.2 ) .
the cancellation between the overall @xmath949 and @xmath914 for @xmath644 particles needs to be interpreted with care . as discussed in 6.2 , at a given @xmath13 , there are two types of particle pairs , i.e. , continuous and caustic pairs .
the two types have different properties , and the inverse scalings of the overall rdf and relative velocity could be an artifact of not properly splitting the two types of particle pairs particles one type of pairs dominates the contribution to the rdf , while the other provides a dominant and @xmath13-independent contribution to the collision kernel , then the cancellation of the scaling exponents of the overall rdf and relative velocity would not be meaningful physically . ] . in 7.2 , we will evaluate the contribution from each type of pairs .
we find that the collision kernel per unit cross section shows an abrupt increase as @xmath1 increases toward 1 ( see , e.g. , sundaram & collins 1997 ) , increases only slightly for @xmath1 between @xmath955 and @xmath611 , and starts to decrease at @xmath612 .
it finally scales with @xmath1 as @xmath177 at @xmath9 .
a physical discussion for the behaviors is given in 7.2 .
the @xmath956independence of the kernel of @xmath704 particles at @xmath21 implies that one may apply our result to estimate the collision kernel of @xmath644 dust particles in protoplanetary turbulence , even though the dust particle diameter @xmath892 is smaller than @xmath19 by orders of magnitude . on the other hand , at @xmath288 ,
the measured collision kernel depends on @xmath13 in the range of @xmath13 considered , and is thus not directly applicable for dust particles with @xmath288 . to achieve a general understanding of the @xmath15 limit
, it would be useful if one could isolate an @xmath13-independent contribution .
for this purpose , we make a preliminary attempt to separate the contributions from the continuous and caustic pairs .
as discussed in 6.2 , at a given distance , @xmath13 , there are two types of particle pairs , corresponding to the continuous ( inner ) and caustic ( tail ) parts of the relative velocity pdf . for the radial relative speed pdf of approaching particle pairs
, the two parts can be roughly divided by a critical value , @xmath957 . in this section
, we separate the contributions of the two types of pairs to the collision kernel in the spherical formulation . using the relation @xmath958 , and splitting the left wing of @xmath668 into two parts , the collision kernel , eq .
( [ sphericalformal ] ) , is written as , @xmath959 where @xmath960 and @xmath961 .
the continuous and caustic contributions can be expressed in more convenient forms , @xmath962 where @xmath963 and @xmath964 are the rdfs of continuous and caustic pairs , corresponding to the probabilities of finding an approaching neighbor that makes a continuous and caustic pair , respectively .
the sum of @xmath965 and @xmath966 is given by @xmath967 . because there are more separating pairs than approaching ones , @xmath968 is smaller ( by 10 - 20% ) than @xmath969 at @xmath970 .
it becomes equal to @xmath971 at @xmath23 .
the average radial relative speeds for the two types of pairs are defined as @xmath972 , and @xmath973 .
the four quantities , @xmath965 , @xmath966 , @xmath974 , and @xmath975 , depend on @xmath836 , whose exact value is uncertain .
we will set @xmath976 here , and defer a study on the effect of the choice of @xmath836 to a later work . in the formulation of falkovich
et al . ( 2002 ) for the collision kernel of @xmath288 particles , the clustering effect enters only for the s - t ( continuous ) contribution .
the rdf was theoretically modeled and evaluated by considering the continuous - type particles only . for the continuous pairs
, @xmath974 scales linearly with @xmath13 , as in the s - t prediction . the continuous contribution to the kernel
is amplified by the clustering effect . on the other hand , the falkovich et al .
formulation suggests that clustering plays no role for the sling ( caustic ) contribution . both @xmath975 , and the normalized kernel @xmath977 due to the slings are independent of the particle distance ( falkovich et al .
these also imply that @xmath966 is @xmath13-independent .
the model of gustavsson & mehlig ( 2011 ) provides a similar formulation for the collision kernel .
the continuous contribution is essentially the same as that in falkovich et al .
they also predicted an @xmath13-independent caustic contribution to the normalized kernel .
however , unlike falkovich et al . (
2002 ) , their derivation accounts for the possibility that both @xmath966 and @xmath975 may depend on @xmath13 . but their scaling exponents exactly cancel out , giving an @xmath13-independent contribution to the kernel . in the left panel of fig .
[ splitrdfandv ] , we plot @xmath965 ( blue ) and @xmath966 ( red ) computed from our simulation data at three distances . at @xmath288 , the number of caustic pairs is small , and most particle pairs belong to the continuous type .
the continuous - type particles show significant clustering at @xmath306 , and the rdf @xmath965 increases toward smaller @xmath13 with similar scaling exponents as the overall rdf .
these suggest that the origin of strong clustering at @xmath306 corresponds to the continuous - type particles ( falkovich et al .
similar to the overall rdf ( fig .
[ rdf ] ) , @xmath965 peaks at @xmath16 . at larger @xmath1
, @xmath965 decreases rapidly with @xmath1 , and becomes comparable to @xmath966 at @xmath978 .
the fast drop of @xmath965 at @xmath291 is because it becomes sensitive to the critical value , @xmath836 , as the range @xmath979 narrows with increasing @xmath1 .
this range also becomes narrower with decreasing @xmath13 , and this causes @xmath965 to slightly decease with decreasing @xmath13 at @xmath291 . at @xmath980 , @xmath965 is significantly smaller than @xmath966 , and essentially all particle pairs are of the caustic type .
the caustic rdf @xmath966 increases with decreasing @xmath13 for particles with @xmath25 , and is @xmath13-independent at larger @xmath1 .
concerning the @xmath13-dependence of @xmath966 at @xmath25 , there are several possibilities .
the first is that this dependence is generic , and @xmath981 as @xmath27 , as considered by gustavsson & mehlig ( 2011 ) .
second , the @xmath13-dependence of @xmath966 may disappear at sufficiently small @xmath13 , and @xmath966 finally converges to a finite constant , as assumed in falkovich et al .
( 2002 ) . in this case
, one may in principle achieve a converged @xmath966 by using a larger number of @xmath982 particles in the simulation to resolve smaller scales .
we checked @xmath983 in our simulation and found that the convergence is not reached yet .
third , the @xmath13-dependence of @xmath966 may be an artifact of our selected value for @xmath836 .
we attempted to vary @xmath836 around @xmath839 by a factor 4 , but were not able to find a considerable decrease in the @xmath13-dependence of @xmath966 .
it is also possible that the transition from the continuous to the caustic type is gradual , and thus exactly splitting two types by a single critical value ( @xmath836 ) is impossible at least at the scales explored here .
we note that @xmath966 is larger than 0.5 at @xmath870 , suggesting some degree of clustering for caustic - type particles .
this , however , could be an artifact that the two types of pairs are not split precisely .
the right panel of fig .
[ splitrdfandv ] shows the average relative speeds , @xmath974 and @xmath975 , for continuous ( blue ) and caustic ( red ) pairs . at @xmath11 ,
@xmath974 is constant with @xmath1 and decreases linearly with @xmath13 , as in the s - t prediction . at @xmath644 ,
@xmath974 decreases with @xmath1 .
this is because the number of continuous pairs becomes small and the average relative speed starts to have a sensitive dependence on the lower limit @xmath836 , which becomes closer to zero as @xmath1 increases . by definition ,
the caustic pairs have larger relative velocity than continuous pairs . at @xmath638 ,
the sling events or caustic formation occur only in regions with very large velocity gradient , and this is responsible for the large relative velocity for caustic pairs at small @xmath1 . as @xmath1 increases , the criterion for the sling events to occur becomes weaker , as reflected by the decrease of @xmath854 .
this results in a decrease in @xmath975 toward @xmath16 . at @xmath291 , most particles belong to the caustic type , and thus the trend of @xmath975 as a function @xmath1 becomes similar to the overall relative speed @xmath150 ( see fig .
[ absolute ] ) .
it first increases with @xmath1 for @xmath984 , and then decreases at @xmath824 . at the three distances shown here
, @xmath975 decreases with decreasing @xmath13 for @xmath25 particles , as the condition to have a caustic pair is weaker at smaller @xmath13 .
again , the @xmath13-dependence of @xmath975 may disappear at much smaller @xmath13 and/or with a better method to split the two types of pairs . at @xmath23 , both @xmath966 and @xmath975
are independent of @xmath13 , indicating full resolution of their statistics in our simulation . in fig .
[ splitproduct ] , we show the continuous ( blue ) and caustic ( red ) contributions to the collision kernel per unit cross section . in the @xmath13 range shown here ,
the continuous and caustic contributions dominate for @xmath288 and @xmath644 , respectively . the continuous contribution decreases with @xmath13 , corresponding to the @xmath13-dependence of the overall kernel , @xmath938 , in fig .
[ product ] .
the decrease of the continuous contribution with @xmath13 is expected from the scaling exponents of @xmath965 and @xmath985 . with decreasing @xmath13
, @xmath985 decreases linearly , which is faster than the increase of @xmath965 . in the limit @xmath15 , the continuous contribution
is expected to vanish for all particles .
this means that , although amplified by clustering , the continuous - type particles do not significantly contribute to the collision rate of dust particles , as pointed out by hubbard ( 2012 ) .
the contribution of the caustic - type particles to the normalized kernel is found to be approximately @xmath13-independent .
the differences in three red lines are small , and one may claim achieving numerical convergence at @xmath986 for all particles .
the @xmath13-independence of @xmath987 for @xmath25 particles at the distances explored here is related to the cancellation of the scalings of @xmath966 and @xmath975 .
intuitively , one would expect that the collision rate of inertial particles converges to a finite constant in the @xmath27 limit .
therefore , @xmath966 and @xmath975 must have inverse scalings with @xmath13 , once the caustic kernel converges . on the other hand
, one may prove the intuition of a finite constant caustic kernel at sufficiently small @xmath13 by showing that @xmath966 and @xmath975 scale inversely with @xmath13 , which was actually predicted by the theoretical model of gustavsson & mehlig ( 2011 ) . as discussed earlier
, both @xmath966 and @xmath975 may become independent of @xmath13 at much smaller @xmath13 .
however , it is also possible that , as @xmath27 , @xmath981 and @xmath988 , especially for @xmath288 particles .
note that @xmath975 does not represent the collisional energy at all . even if @xmath988 , it does not suggest the average collisional energy is zero .
in fact , a computation of the average collision energy per collision ( i.e. , the relative velocity variance weighted by the collision rate ; see , e.g. , hubbard 2012 ) by caustic pairs for particles with @xmath989 in our simulation shows that it already converges to a finite constant at @xmath986 .
the collision energy ( velocity ) of @xmath25 particles by caustic pairs in the @xmath27 limit will be examined systematically in a future study .
the @xmath13-independence of the caustic contribution is responsible for that of the overall normalized kernel at @xmath644 observed in fig .
[ product ] . because the continuous contribution vanishes as @xmath27 , and due to the convergence of the caustic contribution
, one may use our result for @xmath990 to estimate the collision rate of nearly point - like dust particles .
one caveat , however , is that @xmath990 has a dependence on the critical value , @xmath836 , for @xmath288 particles , an issue that will be refined in a future work . ;
blue ) and caustic ( @xmath990 ; red ) contributions to the kernel normalized to @xmath940 or @xmath991 on the left and right y - axis , respectively .
the solid , dashed and dotted lines correspond to normalized kernels measured at @xmath615 , @xmath616 and @xmath529 , respectively .
the dotted line segment shows a @xmath177 scaling . ] from the behaviors of @xmath992 and @xmath990 with @xmath13 , there exists a distance , @xmath993 , below which the caustic contribution dominates .
the critical distance @xmath993 is generally a function of @xmath1 . as seen in fig .
[ splitproduct ] , @xmath994 for @xmath644 . at @xmath874 ,
the two contributions become about equal at @xmath995 . below @xmath874
, @xmath993 decreases very rapidly .
for example , @xmath996 for @xmath762 . achieving sufficient statistics at such a small distance
is computationally challenging , as it requires a huge number of particles .
directly resolving @xmath938 for nearly point particles with @xmath997 would be extremely difficult , and the best approach is thus to properly isolate the caustic and continuous contributions .
the normalized caustic kernel experiences a sharp rise as @xmath1 increase from 0.1 to 1 , suggesting that the collision frequency greatly accelerates as the particle size grows toward @xmath16 .
the rise corresponds to the fast increase of @xmath966 ( see the left panel of fig .
[ splitrdfandv ] ) , which has been interpreted as due to the rapid caustic formation ( wilkinson et al .
it has been claimed that the effect of caustic formation may be modeled as an activation process .
if we fit the normalized caustic kernel by @xmath998 ( wilkinson et al .
2006 , falkovich & pumir 2007 ) , the activation value , @xmath999 , is found to be around @xmath1000 .
the sling events or caustic formation occur at places where the flow velocity gradient is larger than @xmath840 ( falkovich et al .
2002 ) . at these locations , the flow velocity changes faster than the response of the particle . for small particles with @xmath154
, the probability of the sling events corresponds to the high pdf tail of the flow velocity gradient . with increasing @xmath0 ,
the probability becomes larger as it samples toward the inner parts of the flow velocity gradient pdf .
since the rms flow gradient is on the order of @xmath1001 , the probability would increase rapidly as @xmath840 approaches @xmath1001 , leading to a sharp rise in the sling frequency and the caustic kernel at @xmath1002 . for @xmath1 between @xmath955 and the peak
stokes number , @xmath611 , the normalized caustic kernel is almost constant , increasing only by 50% .
this is apparently due to the opposite trends of @xmath966 and @xmath975 with @xmath1 in this range .
the kernel starts to decrease at @xmath612 due to the decrease of @xmath975 .
it finally scales with @xmath1 as @xmath177 ( dotted line segment in fig .
[ splitproduct ] ) for particles with @xmath9 .
we finally comment on the collision kernel commonly used in coagulation models for the dust particle growth in protoplanetary disks . in these models ,
the kernel is typically set to @xmath1003 ( e.g. , dullemond and dominik 2005 ) .
the models usually adopt the rms collision velocities based on the model of volk et al .
( 1980 ) or its later developments .
@xmath1004 is of the cylindrical type discussed in 7.1 .
however , it would be more appropriate to use @xmath913 , rather than the 3d rms , @xmath1005 , in the cylindrical formulation ( wang et al . 2000 ; see the inset in the right panel of fig .
[ absolute ] for the difference between @xmath913 and @xmath1005 ) .
we argue that @xmath1004 can not sufficiently capture the wealth of physics for the turbulence - induced collision rate , especially the interesting phenomena that occur at small @xmath1 .
the prediction of volk et al . for @xmath1005 should be viewed as for the @xmath15 limit , as it does not consider the @xmath13-dependence of the relative speed .
thus , we compare @xmath1004 with our simulation result for @xmath990 , which provides the dominant contribution in the @xmath15 limit .
we point out several problems in the qualitative behavior of the commonly - used collision kernel .
first , volk et al.s model predicts that @xmath1005 drops exactly to zero for small particles with @xmath0 below the turnover time of the smallest eddies in the assumed flow energy spectrum ( e.g. , cuzzi & hoggan 2003 ) .
however , our simulation result shows that , for any finite @xmath1006 , @xmath990 is nonzero , and it can be estimated from the activation formula , @xmath1007 . clearly , the volk et al .
model does not capture the sling effect , and can not accurately estimate the kernel at @xmath14 .
second , based on fig . 5 of cuzzi & hoggan ( 2003 ) , @xmath1004 would increase quite rapidly as @xmath1 increases from 1 to 2 - 3 , and then connect to a @xmath179 scaling at @xmath291 .
this is in contrast to @xmath990 found in our simulation , which changes quite slowly in the range @xmath1008 .
this suggests that the commonly - used @xmath1004 is inaccurate for @xmath25 .
this range of @xmath1 may include particle sizes up to chondrules for typical disk parameters at a radius of 1 au .
it remains to be checked by simulations at higher resolutions whether and how @xmath1004 switches to a @xmath179 scaling for inertial - range particles .
finally , the volk et al . model overestimates the peak relative speed at @xmath594 ( 6.1.1 ) , and would thus overestimate the collision rate by a factor of 2 or so for inertial - range particles and/or large particles with @xmath1009 .
we investigated the turbulence - induced relative velocity of inertial particles using both theoretical modeling and numerical simulations .
we conducted a 512@xmath1010 simulation of a weakly compressible turbulent flow with an rms mach number of @xmath1011 , and evolved 14 species of inertial particles with friction timescales , @xmath0 , covering the entire scale range of the flow .
the stokes number , @xmath1 , of the particles spans about 4 order of magnitudes , ranging from 0.1 to 800 .
the friction time of the largest particles is about 41 times the large eddy turnover time , or 54 times the lagrangian correlation time .
we used the simulation to test the theoretical model for the rms relative velocity of inertial particles by pan & padoan ( 2010 ) , in the case of identical particles ( equal friction times ) .
we explored the probability distribution function ( pdf ) of the relative velocity .
consistent with previous studies , our simulation result for the pdf indicates two different types of particle pairs , named as the continuous and caustic pairs , respectively . in this work
, the relative velocity statistics is measured at distances , @xmath13 , in the range from @xmath635 to @xmath19 .
an accurate measurement at smaller scales would require a larger simulation with a larger number of particles .
our study for the relative velocity helps reveal the fundamental physics , and provides a theoretical preparation toward modeling the collision velocity of nearly - point dust particles at @xmath15 . using the simulation data , we computed the particle collision kernel , and , in particular , we evaluated the contributions from the continuous and caustic pairs .
distinguishing the two types of contributions is crucial for the estimate of the kernel at @xmath15 .
we summarize the main results of this work . 1 .
we introduced the formulation for the particle relative velocity by pan & padoan ( 2010 ) , which reveals an insightful physical picture .
the relative velocity of two nearby identical particles is determined by the memory of the flow velocity difference along their trajectories in the past , and hence depends on the separation behavior of particle pairs backward in time .
we adopted a two - phase separation behavior consisting of a ballistic and a richardson phase , and showed that the model predicts a @xmath179 scaling for inertial - range particles in turbulent flows with an extended inertial range .
the model can correctly reproduce the expected behaviors in the extreme limits of small and large particles .
the model prediction for the rms relative velocity is in good agreement with the simulation results for @xmath1012 .
the physical picture also provides a successful explanation for the the relative velocity pdf as a function of @xmath1 .
2 . to improve the understanding of the inertial particle statistics , we analyzed both lagrangian and eulerian temporal correlation functions in the simulated flow .
while the flow velocity along the trajectory of a small particle is close to lagrangian , the velocity seen by a very large particle may be better approximated by eulerian .
the eulerian and lagrangian correlation timescales , @xmath511 and @xmath4 , were found to be close to each other , with @xmath511 slightly larger ( by 10% ) .
our model predictions for both 1-particle rms velocity and the rms relative speed between two identical particles depend mainly on the correlation timescale and are insensitive to the function form of the temporal correlation .
these provides a validation for using the lagrangian correlation function form for all particles .
our simulation data shows that , in the small particle limit ( @xmath11 ) , the 3d rms relative velocity of particle pairs at a distance @xmath1013 is constant , @xmath1014 , consistent with the saffman - turner prediction .
it starts to rise at @xmath644 , and peaks for particles with @xmath7 , corresponding to @xmath1015 in our simulated flow .
as expected , the relative velocity scales with @xmath1 as @xmath177 in the limit @xmath9 .
the pp10 model with reasonable parameters provides an excellent fit to the simulation result for the 3d rms .
the maximum relative speed at @xmath594 is twice smaller than the rms velocity of the turbulent flow , indicating a factor of @xmath116 overestimate by the commonly - used model of volk et al .
( 1980 ) and its later developments ( e.g. , markiewicz et al . 1991 ; cuzzi & hogan 2003 ; ormel & cuzzi 2007 ) .
the rms relative speed of particles with @xmath617 shows a @xmath13-dependence in the range @xmath310 .
the dependence for the smallest particles ( @xmath657 ) in our simulated flow at @xmath1016 was found to be slower than the linear scaling predicted by the saffman - turner formula , suggesting considerable contributions from the sling events or caustic formation . for larger particles with @xmath618 , the backward separation at a friction
timescale ago becomes insensitive to the initial distance @xmath13 , and the rms relative speed is independent of @xmath13 . it remains to be examined whether and at which scale the rms relative velocity of @xmath1017 particles would converge and become @xmath13-independent as @xmath27 .
+ the rms relative speeds in the radial and tangential directions are nearly equal for all @xmath643 particles . for @xmath11 particles ,
this is in contrast to the saffman - turner formula , which predicts the tangential rms is larger than the radial rms by @xmath541 .
this near equality is due to the randomization of the relative velocity direction with respect to the particle separation @xmath217 , which is caused by the deviation of particle trajectories from the fluid elements and/or the stochastic backward separation of particle pairs .
+ in the distance range explored , we find an asymmetry in the relative speed of @xmath617 particles : approaching pairs that may lead to collisions have a larger relative speed than separating ones .
the asymmetry is related to the phenomenon of turbulent clustering .
the asymmetry is expected to decrease with decreasing @xmath13 , but it remains to be tested whether it completely vanishes as @xmath15 .
the probability distribution function for the particle relative velocity is highly non - gaussian , exhibiting extremely fat tails . for small particles with @xmath288 , the effects of the particle memory and the backward separation lead to a self - amplification starting at the far tails , corresponding to the effect of slings or caustic formation .
as @xmath1 increases , the amplification becomes stronger and proceeds toward the inner part of the pdf , causing an increase in the fatness of the overall pdf shape . on the other hand , as @xmath1 increases above @xmath6 ,
the pdf shape becomes less fat .
for the larger particles , the relative velocity samples the pdf , @xmath799 , of the flow velocity increment , @xmath790 , at larger scales , @xmath98 . as the fatness of @xmath812 decreases with increasing @xmath98 , the pdf of @xmath1018 particles keeps thinning with increasing @xmath1 . at a particle distance of @xmath1019 ,
the pdf shape is fattest at @xmath16 , with a kurtosis of @xmath595 .
we identified two sources of non - gaussianity for the particle relative velocity : the imprint of intermittency of the turbulent flow and an intrinsic contribution from the particle dynamics .
we predicted a 4/3 stretch exponential pdf , @xmath1020 , for inertial - range particles in an exactly gaussian velocity field with kolmogorov scaling .
this 4/3 stretched exponential is observed in the pdf tails of particles with @xmath822 ( or @xmath1021 ) , confirming the validity of the physical picture of pp10 .
+ based on the pp10 picture and our simulation result for the pdf , the particle pairs at a give distance can be split into two types , namely , the continuous type and the caustic type ( wilkinson et al .
the two types correspond to the inner part of the pdf that follows the flow velocity difference , and the outer pdf part that is affected by the particle memory and the backward separation , respectively .
+ the pdf of the particle collision velocity is expected to play a crucial role in the growth of dust particles in protoplanetary turbulence , as it determines the fractions of collisions resulting in sticking , bouncing and fragmentation .
the relative velocity pdfs of @xmath1022 particles already converge at @xmath886 . on the other hand ,
the pdfs of smaller particles show @xmath13-dependence at @xmath1023 , and an appropriate extrapolation is needed for the application to dust particle collisions at @xmath15 .
the shape of the measured pdfs provides interesting clues for the collisional energy .
for example , the pdf of the 3d amplitude of the relative velocity suggests much higher probabilities of extremely small and large collision speeds than estimated from a gaussian pdf . only for very large particles with @xmath1024
does the pdf approach gaussian .
the highly non - gaussian nature of the relative / collision velocity needs to be incorporated into dust coagulation models for protoplanetary disks . pushing the pdf toward smaller @xmath13 and finally to @xmath15
will be pursued in a future work .
we computed the particle collision kernel from the simulation data using both spherical and cylindrical formulations .
it was found that the two formulations give nearly equal predictions for all particles . adopting the formulation of falkovich et al .
( 2002 ) and gustavsson and mehlig ( 2011 ) , we calculated the contributions to the collision kernel from continuous and caustic particle pairs separately .
we showed that , although amplified by the effect clustering , the collision kernel due to the continuous - type pairs decreases with @xmath13 , and would vanish in the @xmath28 limit .
consistent with the theoretical prediction of gustavsson and mehlig ( 2011 ) , the contribution to the normalized kernel by caustic pairs is found to be @xmath13-independent , and the convergence is reached at @xmath1025 . this caustic contribution is dominant at sufficiently small @xmath13 , and can be used to estimate the collision rate of nearly point - like dust particles .
+ the caustic contribution to the collision kernel per unit cross section shows an abrupt rise as @xmath1 increases toward 1 , which may be viewed as an activation process corresponding to the rapid formation of caustics .
as @xmath1 increases from @xmath6 to @xmath594 , the normalized caustic kernel is roughly constant , increasing only slightly by 50% .
it finally decreases as @xmath1026 for large particles with @xmath9 .
coagulation models for dust particle growth need to incorporate these important features
. we will provide fitting functions for the caustic collision kernel and the collisional energy as a function of @xmath1 in a separate study . in this work
, we have focused on the monodisperse case with identical particles .
a systematic analysis for the relative velocity of different particles will be conducted in a follow - up paper .
future simulations at higher resolutions will further improve our understanding of the problem .
for example , a @xmath357 simulation would help verify the existence of the predicted @xmath179 scaling for inertial - range particles by various models , which is not yet confirmed numerically .
we have only partially addressed the @xmath28 limit , necessary for the application to dust particle collisions .
a computationally more demanding simulation with a larger number of inertial particles would allow us to examine the particle statistics at smaller scales . with such simulations and by isolating the caustic pairs from the continuous ones , we can systematically evaluate the collision rate and the pdf of the collisional energy at @xmath15 .
the results of these future studies will significantly improve the formulation of coagulation models to compute the dust particle evolution in protoplanetary disks .
we thank the referee , alexander hubbard , for an extensive report that helped us improve the paper , and anders johansen for useful discussions and for support with the pencil code .
pp is supported by the fp7-people-2010-rg grant pirg07-ga-2010 - 261359 .
the simulations were carried out on the nasa / ames pleiades supercomputer .
our model for the relative velocity of inertial particles depends on the particle pair dispersion backward in time ( 3.2.3 ) .
we adopted a two - phase behavior , consisting of a ballistic and a richardson phase . to constrain the richardson constant , @xmath297 , in the latter phase , we study the dispersion of tracer particles in our simulated flow , and take the measured value of @xmath297 as a reference for inertial particles . in fig .
[ separation ] , the three solid lines from bottom to top show the backward - in - time separation of tracer particle pairs with initial " distance @xmath1027 , @xmath1028 and @xmath1029 , respectively .
we subtracted @xmath1030 from the separation variance , @xmath1031 .
the initial " time is set to 0 , and @xmath184 is negative for the backward separation . as seen in fig .
[ separation ] , the particle separation at small @xmath327 shows a ballistic behavior with @xmath1032 increasing as @xmath1033 .
the physical origin of the ballistic behavior is that the velocity at which two tracer particles separate is determined by the flow velocity difference across the initial distance , @xmath13 , and remains roughly constant before the particle distance becomes significantly larger than @xmath13 .
this ballistic phase of the tracer particles is physically different from that of inertial particles discussed in 3.2.3 . for inertial particles , the duration and the separation speed of the ballistic phase
depend on the particles memory timescale .
but for tracer particles , the ballistic phase is determined purely by the initial distance . for @xmath1034 ,
the ballistic phase lasts for a few kolmogorov timescales . , @xmath545 and @xmath1035 , respectively .
the dashed line plots the forward separation with @xmath1036 .
the initial time is set to 0 , and @xmath184 is negative ( positive ) for the backward ( forward ) separation .
the separation shows a ballistic phase and a richardson behavior at small and large @xmath327 , respectively . ] the richardson separation behavior ( the @xmath1037 scaling ) is observed at large @xmath327 after the particle separation enters the inertial range of the flow . in the bottom solid curve for @xmath620 , a @xmath1037
scaling exists in a very limited range .
a rough estimate of the richardson constant in that range gives @xmath1038 .
consistent with previous studies ( e.g. , sawford et al .
2008 ) , the time range that exhibits the richardson scaling becomes broader as @xmath13 increases to @xmath1029 .
this allows a more accurate estimate of @xmath297 , and we find @xmath1039 for @xmath1036 , consistent with the experimental results of berg et al .
similar to sawford et al .
( 2008 ) , the measured @xmath297 has a dependence on @xmath13 .
if the inertial range of the flow is considerably broader and the richardson behavior exists in a larger time range , the curves for different initial distances are expected to converge at sufficiently large time lags , with @xmath297 eventually approaching a universal value . the dashed line in fig . [ separation ] plots the forward - in - time pair dispersion of tracer particles at @xmath1036 .
a comparison with the top solid line shows that the forward separation is slower than the backward separation . for @xmath1036 , @xmath297
is estimated to be 0.5 in the forward separation , about twice smaller than the value ( 1.2 ) for the corresponding backward separation .
this is consistent with the result of berg et al .
( 2006 ) . a physical explanation for the faster backward separation will be given in appendix b. in our model for the relative velocity of inertial particles , it is the backward separation that is relevant , and the purpose of showing an example for the forward separation in fig . [ separation ] is to illustrate the difference between the forward and backward separations .
the richardson constant , @xmath297 , in the tracer - like phase of inertial particle separation may be different from tracers .
however , it is reasonable to assume that @xmath297 for the backward separation of inertial particles lies in a similar range . like tracers , the value of @xmath297 for inertial particles may also depend on the initial distance , @xmath13 . in 6.1 , we adjusted the value of @xmath297 in our model to obtain best fits to the simulation results for the rms relative velocity at different @xmath13 .
in this appendix , we analyze the probability distribution functions ( pdf ) of the flow velocity increments .
we measure the pdfs , @xmath715 and @xmath712 , of the longitudinal and transverse increments as a function of the length scale , @xmath98 .
similar to the computation of the structure functions in 4.2 , the pdf measurement also uses the velocity differences along the three directions of the simulation grid .
the variances of @xmath715 and @xmath1040 correspond to the longitudinal ( @xmath435 ) and transverse ( @xmath440 ) structure functions , which have been shown in fig .
[ structureandpower ] ( see 4.2 )
. clearly , the pdfs are wider at larger scales .
also , @xmath1040 is wider than @xmath715 because @xmath1041 at all @xmath98 ( see fig . [ structureandpower ] ) . .
the normalized velocity increments are defined as @xmath1042 and @xmath1043 .
the top line in each panel plots the exact pdf at @xmath1044 ( the cell size ) , and , for clarity , the pdf at each larger @xmath98 is shifted downward by a factor of 4 . except at the largest scales ,
the pdf , @xmath1045 , of the radial increment has a negative skewness , whereas @xmath1046 for the transverse increment is symmetric at all scales .
the pdf tails are highly non - gaussian at small scales .
with increasing @xmath98 , the pdfs become less fat and finally approach gaussian . in the right panel ,
the dashed and dotted lines for @xmath1044 are the normalized pdfs of the transverse increment conditioned on @xmath1047 and @xmath1048 , respectively .
one may change the normalization of @xmath98 to the integral scale , @xmath108 , using @xmath1049.,title="fig:",height=249 ] .
the normalized velocity increments are defined as @xmath1042 and @xmath1043 .
the top line in each panel plots the exact pdf at @xmath1044 ( the cell size ) , and , for clarity , the pdf at each larger @xmath98 is shifted downward by a factor of 4 . except at the largest scales , the pdf , @xmath1045 , of the radial increment has a negative skewness , whereas @xmath1046 for the transverse increment is symmetric at all scales .
the pdf tails are highly non - gaussian at small scales .
with increasing @xmath98 , the pdfs become less fat and finally approach gaussian . in the right panel ,
the dashed and dotted lines for @xmath1044 are the normalized pdfs of the transverse increment conditioned on @xmath1047 and @xmath1048 , respectively
. one may change the normalization of @xmath98 to the integral scale , @xmath108 , using @xmath1049.,title="fig:",height=249 ] to better see the pdf _ shape _ as a function of @xmath98 , we normalized the pdfs at each scale to have unit variance .
the radial and transverse velocity increments are normalized to their rms values , i.e. , @xmath1050 and @xmath1043 .
the left panel of fig .
[ flowpdf ] shows the normalized pdf , @xmath1045 , of the radial increment . except at the largest scales ,
the pdf is negatively skewed . for inertial - range scales , this can be understood from kolmogorov s 4/5 law , @xmath1051 , which indicates a negative skewness for the pdf of @xmath681 .
the connection of the 3rd order moment of @xmath681 to the energy dissipation rate suggests that the skewness originates from the dissipative nature of turbulence .
the skewness of the pdf of @xmath681 also provides an explanation for the faster backward separation found in appendix a. the left and right tails of @xmath1045 correspond to tracer pairs receding from each other backward and forward in time , respectively .
the broader left tail of the pdf thus suggests that the backward separation of tracer particles is faster than the forward case . unlike @xmath1045 , the pdf , @xmath1046 , of the transverse increment in the right panel is symmetric at all @xmath98 , as expected from statistical isotropy .
both @xmath1045 and @xmath1046 are close to gaussian at the largest scales , @xmath1052 ( 1/4 box size , or 1.6 times the integral scale , @xmath108 ) and @xmath1053 ( 1/2 box size , or @xmath1054 ) , of the simulated flow .
this is consistent with the gaussian 1-point statistics in fully developed turbulence . at smaller @xmath98 ,
the pdfs become non - gaussian , and the tails keep getting fatter with decreasing @xmath98 , a phenomenon known as intermittency in turbulence theory ( frisch 1995 ) .
as mentioned in the text , we use the word fat " ( or thin " ) specifically for the shape of the pdf , while broad " ( or narrow " ) refers to the extension or width of the pdf .
the smallest scale , @xmath1055 , in the figure corresponds to the size , @xmath467 , of the computational cell .
the shape of the normalized pdf is expected to remain unchanged once @xmath98 becomes smaller than @xmath1056 .
physically , the viscosity acts to smooth the velocity field and makes it differentiable in the dissipation range , and thus the velocity increment at any scale @xmath923 is proportional to the local velocity gradient , whose pdf is fixed . in our simulation ,
the velocity field inside a computation cell is obtained by interpolation , and thus the pdf of the velocity difference below the cell size is controlled by the velocity gradient pdf at @xmath467 . in
6.2 , we showed that the trend of the pdf shape of the flow velocity difference with @xmath98 has interesting effects on the pdf of the relative velocity of inertial particles as a function of the particle inertia .
the tails of @xmath1046 for the transverse increment can be approximately described by stretched exponentials , @xmath756 ( see eq .
( [ se ] ) in 6.2 ) . at largest scales , @xmath1046 are nearly gaussian , and @xmath1057 . with decreasing @xmath98 , @xmath758 decreases , corresponding to fatter tails .
for example , the best - fit @xmath758 for @xmath1058 at @xmath1059 is @xmath6 , and it further decreases to @xmath1060 at @xmath1061 . due to the asymmetry of the radial pdf , @xmath1045
, one needs to obtain the fits separately for the left and right wings . comparing the left wing of @xmath1062 with @xmath1046
, we see that their shape has a similar level of fatness at the same scale @xmath98 .
in fact , the best - fit @xmath758 for the left tail of @xmath1045 is very close to that for @xmath1046 .
we also find that the best - fit values of @xmath758 for the left and right tails of @xmath1045 are close , indicating that the two tails have a similar shape and differ only in the fluctuation amplitude . to quantify the fluctuation amplitudes in the left and right wings of @xmath715
, we define the variances in the two wings as @xmath1063 and @xmath1064 .
the definition of @xmath1065 is similar to @xmath679 for the relative velocity of inertial particles ( 6.1.3 ) .
we find that the ratio of @xmath1066 to @xmath1067 is @xmath1068 at @xmath1069 .
this ratio decreases with increasing @xmath98 , and reaches unity at the largest scales .
the variances of the left and right wings of @xmath715 was used in the discussion on the relative velocity of approaching and separating particle pairs in the @xmath11 limit ( see 6.1.3 ) .
we also considered the pdf of @xmath104 conditioned on the sign of @xmath681 .
we denote two conditional pdfs as @xmath1070 and their variances as @xmath1071 . at @xmath1069 , @xmath1072
is found to be larger than @xmath1073 by 28% . in the right panel of fig .
[ flowpdf ] , the dashed and dotted lines show the normalized conditional pdfs , @xmath1074 and @xmath1075 , at @xmath1069 .
the conditional variances and pdfs of @xmath104 are useful to understand the tangential relative velocity of approaching and separating particle pairs with @xmath11 ( 6.1.3 and 6.2.4 ) . | we study the relative velocity of inertial particles suspended in turbulent flows and discuss implications for dust particle collisions in protoplanetary disks .
we simulate a weakly compressible turbulent flow , evolving 14 particle species with friction timescale , @xmath0 , covering the entire range of scales in the flow .
the particle stokes numbers , @xmath1 , measuring the ratio of @xmath0 to the kolmogorov timescale , are in the range @xmath2 . using simulation results ,
we show that the model by pan & padoan ( pp10 ) gives satisfactory predictions for the rms relative velocity between identical particles .
the probability distribution function ( pdf ) of the relative velocity is found to be highly non - gaussian .
the pdf tails are well described by a 4/3 stretched exponential function for particles with @xmath3 , where @xmath4 is the lagrangian correlation timescale , consistent with a prediction based on pp10 .
the pdf approaches gaussian only for very large particles with @xmath5 .
we split particle pairs at given distances into two types with low and high relative speeds , referred to as continuous and caustic types , respectively , and compute their contributions to the collision kernel .
although amplified by the effect of clustering , the continuous contribution vanishes in the limit of infinitesimal particle distance , where the caustic contribution dominates .
the caustic kernel per unit cross section rises rapidly as @xmath1 increases toward @xmath6 , reaches a maximum at @xmath7 , and decreases as @xmath8 for @xmath9 . |
effects of ferromagnetism have been known since antiquity . but a mathematical understanding of the microscopic origin of ferromagnetism has remained somewhat elusive , until today ! pauli paramagnetism , ferro- , ferri- and antiferromagnetism are quantum phenomena connected to the spin of electrons and to pauli s exclusion principle .
the theory of _ paramagnetism _ in ( free ) electron gases is quite straightforward , @xcite .
_ antiferromagnetism _ is relatively well understood : a mechanism for the generation of antiferromagnetic exchange interactions has been proposed by anderson @xcite , who discovered a close relationship between the half - filled hubbard model and the heisenberg antiferromagnet using perturbative methods ; ( see also @xcite for mathematically more compelling and more general variants of anderson s key observation ) .
it has been proven rigorously by dyson , lieb and simon @xcite , using the method of infrared bounds previously discovered in @xcite , that the _ quantum heisenberg antiferromagnet _ with nearest - neighbour exchange couplings exhibits a phase transition accompanied by spontaneous symmetry breaking and the emergence of gapless spin waves , as the temperature is lowered , in _ three _ or _ more _ dimensions .
( the mermin - wagner theorem says that , in ( one and ) two dimensions , continuous symmetries can not be broken spontaneously in models with short - range interactions , @xcite . )
our mathematical understanding of _ ferromagnetism _ is far less advanced .
some kind of heuristic theory of ferromagnetism emerged , long ago , in the classic works of heisenberg , bloch , stoner , dyson , landau and lifshitz , and others @xcite .
various insights have been gained on the basis of some form of mean - field theory , with small fluctuations around mean - field theory taken into account within a linear approximation .
this approximation , however , is known to break down in the vicinity of the critical point of a ferromagnetic material , where nonlinear fluctuations play a crucial role , @xcite . in a variety of tight - binding models of itinerant electrons ,
ferromagnetic order has been exhibited in the _ ground state _
( i.e. , at zero temperature ) ; see @xcite,@xcite,@xcite .
one of these models is a fairly natural two - band model in which ferromagnetism arises from a competition between electron hopping , coulomb repulsion and an on - site hund s rule , @xcite .
( hund s rule says that the spin - tripled state of two electrons occupying the same site is energetically favoured over the spin - singlet state .
it should be emphasized , however , that a mathematically rigorous derivation of hund s rule in atomic physics from first principles has _ not _ been accomplished , so far . )
none of the results in @xcite,@xcite,@xcite comes close to providing some understanding of ferromagnetic order and of an order - disorder phase transition at _ positive _ temperature .
it is not known how to derive , with mathematical precision , an effective hamiltonian with explicit ferromagnetic exchange couplings from the microscopic schrdinger equation , or a tight - binding approximation thereof , of ferromagnetic materials .
but even if we resort to a phenomenological description of such materials in terms of models where ferromagnetic exchange couplings have been put in _ by hand _ we face the problem that we are unable to exhibit ferromagnetic order at low enough temperature and to establish an order - disorder phase transition in three or more dimensions . _
no mathematically rigorous proof _ of the phase transition in the _ quantum heisenberg ferromagnet _ is known , to date ! ( such a result has , however , been established for _ classical _ heisenberg models in @xcite . )
ab - initio quantum monte carlo simulations of models of quantum ferromagnets are plagued by the well known `` sign- ( or complex - phase ) problem '' .
thus , until now , there are neither substantial mathematically rigorous results on , nor are there reliable ab - initio numerical simulations of , realistic models of ferromagnetic metals , such as ni , co or fe ! given that ferromagnetism is among the most striking _ macroscopic _ manifestations - apparent , e.g. , in the needle of a compass - of the _ quantum - mechanical _ nature of matter , this is clearly a desolate state of affairs .
in the present paper we shall not remedy this unsatisfactory situation .
however , first , we attempt to draw renewed attention to it , and , second , we outline a formalism and some fairly elementary analytical observations of which we hope that they will ultimately lead to a better , mathematically rather precise understanding of ferromagnetism . were it not known already , our analysis and the one in @xcite would make clear that ferromagnetism is a non - perturbative phenomenon involving strong correlations and gapless modes . to understand it mathematically will most probably necessitate a full - fledged _ multi - scale _ ( renormalization group ) _ analysis_. the formalism presented in this paper and our calculations are intended to provide a convenient starting point for such an analysis . analytical work on ferromagnetism may seem to be rather unfashionable . however , there are recent developments , such as spintronics , fast magnetic devices , etc . that may make work like ours appear worthwhile . our own motivation for the work that led to this paper actually originated in studying recent experiments with beams of spin - polarized , hot electrons shot through ferromagnetically ordered films consisting of ni , co or fe that were carried out in the group of h.c .
siegmann at eth ; see @xcite,@xcite . back in 1998
, it became clear to one of us that the concept of the `` _ weiss exchange field _ ''
( see @xcite,@xcite ) would play a useful role in a theoretical interpretation of the experimental results reported in @xcite,@xcite .
more generally , the weiss exchange field appears to offer a key to a systematic study of phase transitions in magnetic materials , magnetic order , spin precession and magnon dynamics . in this paper , we focus on elucidating the microscopic origin of the weiss exchange field , the role it plays in the theory of magnetism , and its dynamics .
our paper is organized as follows . in section 2
, we describe the experiments reported in @xcite,@xcite and sketch a phenomenological interpretation , based on scattering theory , of the results found in these experiments , merely adding some conceptual remarks to the discussion of our experimental colleagues and describing the role played by the weiss exchange field .
we also propose some further experiments , in particular a stern - gerlach experiment for electrons traversing an inhomogeneous exchange field . a mathematically precise analysis of the scattering of electrons ( or neutrons , photons , ... ) at dynamical targets , such as magnetic films , metallic solids , liquid droplets , ... , will appear elsewhere ; ( some first results appear in @xcite ) . in section 3 , we reformulate one - band @xmath1 and hubbard models in terms of a dynamical weiss exchange field . for this purpose ,
the standard imaginary - time functional integral formalism for the analysis of thermal equilibrium states of quantum many - body systems is recalled .
the _ weiss exchange field _ is seen to be a lagrange - multiplier field in a hubbard - stratonovich transformation of the original functional integral that renders the action functional quadratic in the grassmann variables describing the electronic degrees of freedom ; see e.g. @xcite,@xcite . the effective field theory of the weiss exchange field
is obtained after integrating over those grassmann variables .
the effective ( imaginary - time ) action functional of the exchange field and identities for green functions of spin operators are derived . in section 4 ,
we determine the leading terms of the effective action of the weiss exchange field , @xmath2 , in the approximation where fluctuations of the _ length _ of the exchange field are neglected . for this purpose
, we derive the ferro- and antiferromagnetic mean - field equations from the exact effective action of the exchange field . by solving these equations
we determine the most likely length , @xmath3 , of the exchange field .
from that point on , the length of the exchange field is frozen to be @xmath4 .
we then consider a one - band hubbard model with a half - filled band and find that , in this situation , the effective action of @xmath5 is the one of a nonlinear @xmath6-model with a minimum that favours nel order .
this result is found on the basis of controlled perturbative calculations and goes beyond linear stability analysis of the antiferromagnetic mean - field solution , ( which has been presented , e.g. , in @xcite ) .
it represents a functional - integral version of anderson s basic observations @xcite .
we then turn to ferromagnetically ordered mean field solutions and show that @xmath7-independent fluctuations are not a source of instability of such a solution .
then we consider a one - band hubbard model with a weakly filled , fairly flat band .
in this situation one expects that ferromagnetism prevails .
indeed , we find that , at low temperatures , the ferromagnetic mean - field equation has a non - trivial solution , and that this solution belongs to a _ quadratically stable _ critical point of the effective action of w. this conclusion is the result of somewhat subtle calculations involving processes close to the fermi surface , which make the dominant contribution ( but would lead to small - energy denominators in a purely perturbative analysis ) .
details will appear in @xcite .
our calculations support the idea that the one - band hubbard model with a weakly filled , fairly flat band describes coexistence of metallic behaviour with ferromagnetic order , at sufficiently low temperatures . a similar conclusion was reached , tentatively , in @xcite for some two - band hund - hubbard models .
the methods of the present paper also apply to the model discussed in @xcite ; see @xcite . in the next to last subsection of section 4
, we exhibit a universal wess - zumino term in the effective action of the weiss exchange field @xmath5 and calculate its coefficient , which is purely imaginary .
the wess - zumino term is `` irrelevant '' for antiferromagnets , but plays a _
crucial _ role in the dynamics of magnons in _ ferromagnetically _ ordered systems . repeating arguments in @xcite , we derive the landau - lifshitz equations for magnons in a ferromagnet . finally , we draw attention to two well known arguments explaining why there is no magnetic ordering at positive temperature , in one and two dimensions ; ( but see @xcite ) . in section 5 , we sketch novel rigorous proofs , based on analyzing the effective field theory of the exchange field , @xmath5 , of the existence of phase transitions and magnetic order at low temperatures in a class of @xmath0-models , heisenberg antiferromagnets and ferromagnets of localized @xmath8- spins , for @xmath9 .
see @xcite for the original results .
our proof is based on establishing reflection positivity of the effective field theory of the exchange field @xmath5 and then using the original techniques developed in @xcite ; ( see also @xcite , @xcite ) .
it should be emphasized that the concept of the weiss exchange field has a number of further , quite exciting applications .
we hope to return to these matters in future papers .
some of the material in this paper has a review character ; but some of it is new .
we hope it is fairly easy to read .
if it draws renewed attention to some of the deep technical problems in the quantum theory of magnetism it has fulfilled its purpose .
we gratefully dedicate this paper , belatedly , to two great colleagues and friends of the senior author ( j.f . ) : g. jona - lasinio , on the occasion of his seventieth birthday , and h .- c .
siegmann , on the occasion of his retirement from eth .
_ acknowledgements .
_ j. frhlich thanks s. riesen and h .- c .
siegmann for very stimulating discussions of the experiments in @xcite , p. wiegmann for some crucial advice with the calculations in section 4 , and the ihs for hospitality during much of the work on this paper .
we all thank m. azam for very useful and pleasant discussions on the uses of the exchange field .
we start this section with a brief description of recent experiments carried out by oberli , burgermeister , riesen , weber and siegmann at eth - zrich @xcite , @xcite . in these experiments , a beam of hot , spin
polarized electrons is shot through a thin ferromagnetic film ( ni , co , or fe ) and the polarization of the outgoing beam is observed .
their experimental setup is as described in fig .
[ fig : exp1 ] .
* the thickness , @xmath10 , of the film ; @xmath10 is a few nanometers . * the average energy , @xmath11 , of an incident electron ; if @xmath12 denotes the fermi energy of the magnetic film then @xmath13 varies between 4@xmath14 and 16@xmath14 .
the group velocity of the electrons inside the film is denoted by @xmath15 ; it is not directly measurable , but it is comparable to @xmath16 , where @xmath14 is the average potential energy of an electron and @xmath17 its effective mass inside the film . * the degree , @xmath18 , and the direction , @xmath19 , of the spin polarization of the incident electron beam ; ( in fig .
[ fig : exp1 ] , @xmath19 is parallel to the @xmath7-direction , @xmath20 to the @xmath21-direction ) ; the same quantities , @xmath22 and @xmath23 , for the outgoing beam . *
the direction of the magnetization , @xmath24 , of the film ( in fig .
[ fig : exp1 ] chosen to be parallel to the @xmath25-axis ) ; the angles , @xmath26 and @xmath27 , between @xmath28 and @xmath24 and between @xmath23 and @xmath29 , respectively ( @xmath30 , in fig .
[ fig : exp1 ] ) .
experimentally , the angle @xmath27 is found to be considerably smaller than @xmath26 , i.e. , the spins of the transmitted electrons rotate into the direction of the spontaneous magnetization @xmath24 of the film .
this is interpreted as being mainly due to an enhanced absorption of minority - spin electrons , as compared to majority - spin electrons ; see ( vi ) .
( it appears that the contribution of spin flip processes accompanied by magnon emission into the film
`` stoner excitations '' to the total spin rotation is only around 5% , @xcite ) . *
the spin precession angle , @xmath31 , between the projections of @xmath32 and of @xmath33 onto the plane perpendicular to @xmath24 ( the @xmath34-plane of fig .
[ fig : exp1 ] ) ; @xmath31 is found to be `` large '' ( tens of degrees ) .
* let @xmath35 be the intensity of the incident beam , and let @xmath36 and @xmath37 be the intensities of the outgoing beam of electrons with spin parallel or antiparallel to @xmath24 , respectively , assuming the incident beam has intensity @xmath35 and the spins of its electrons are parallel to @xmath24 ( @xmath28 parallel to @xmath24 , @xmath38 ) , or antiparallel to @xmath24 ( @xmath28 anti - parallel to @xmath24 , @xmath39 ) , respectively .
then @xmath40 , or @xmath41 , respectively , and @xmath42 .
@xmath43 and @xmath44 can be measured and yield the spin - transmission asymmetry @xmath45 @xmath46 is found to be positive and large .
this is interpreted in terms of rates of transitions of electrons into unoccupied 3d states ( holes ) : there are more unoccupied 3d states in the film with spin antiparallel to @xmath24 ( minority spin ) than with spin parallel to @xmath24 ( majority spin ) .
this explains qualitatively the experimental results found for @xmath46 and for @xmath47 ( see ( iv ) ) ; @xcite . * the orbital deflection angle , @xmath48 , between the directions of the incident and the transmitted beam ( not indicated in fig .
[ fig : exp1 ] ) . experimentally @xmath48
is found to be negligibly small .
this tells us that the integrated lorentz force on the electrons transmitted through the film is tiny .
the precession of the spins of the electrons when they traverse the film can therefore _ not _ be explained by zeeman coupling of the spins to the magnetic field inside the layer .
it is mainly due to zeeman coupling of the spins to what will be called the _
weiss exchange field_. in iron , the weiss exchange field causing the observed spin precession would correspond to a magnetic field of roughly 8000 tesla ( which is gigantic ) .
a theoretical interpretation of the experimental results reported in @xcite can be attempted within the formalism of _ scattering theory_. if the luminosity of the incident beam is low we can consider a single incoming electron .
the incoming state is described as a tensor product of a pauli spinor , @xmath49 , describing the incident electron and a state , @xmath50 , of the film .
typically , @xmath50 is the ground state ( temperature @xmath51 ) or a thermal equilibrium state ( @xmath52 ) of the film .
the outcoming state , long after the interactions between an outgoing electron and the film have taken place , is more complicated and will , in general , exhibit _ entanglement _ between the electron and the degrees of freedom of the film .
if only measurements far away from the film are performed , as in @xcite , the outgoing state can be described as a density matrix @xmath53 where @xmath54 is a non - negative , trace - class operator on the hilbert space of @xmath55 _ outgoing electrons _ ( the incident electron has knocked @xmath56 electrons out of the film ) , @xmath57 ; @xmath58 is a non - negative number , the _ absorption probability _ , @xmath59 is a non - negative trace - class operator on the hilbert space @xmath60 of square - integrable pauli spinors and describes the ( generally _ mixed _ ) state of _ one _ outgoing electron . ``
conservation of probability '' implies that @xmath61 the state @xmath62 is obtained by taking a partial trace of the outgoing state of the _ total system , including _ the film , over the degrees of freedom of the film .
this is justified , because the degrees of freedom of the film are _ not _ observed in the experiment .
if the energy , @xmath11 , of the incident electron is below ( or comparable to ) the threshold , @xmath63 , for emission of two or more electrons from the film then @xmath64 in the interpretation of the experimental data provided in @xcite , this is tacitly assumed . experimentally , the absorption probability @xmath58 and the spin polarizations @xmath65 and @xmath66 of the incoming and the outgoing electron , respectively , are measured .
the vectors @xmath67 and @xmath66 are given by @xmath68 and @xmath69 where @xmath49 is the wave function of the incident electron , @xmath70 is the vector of pauli matrices , and @xmath71^{-1 } \rho_1 $ ] is the _
conditional _ state of the outgoing electron , given that it has _ not _ been absorbed in the film . if ( [ eq : rhon ] ) is assumed to hold then @xmath72 it would be highly interesting to estimate , experimentally , the amount of entanglement with the film ( or _ decoherence _ ) in the state of the outgoing electron by measuring the quotient @xmath76 .
if @xmath77 then @xmath78^{-1 } \rho_1 $ ] is _ not _ a pure state , anymore , meaning there is entanglement with the film .
apparently , @xmath76 has not been measured accurately , yet . a moment
s reflection shows that spin flip processes accompanied by magnon emission in the film ( `` stoner excitations '' ) lead to entanglement ; while absorption of electrons into unoccupied 3d states need not be correlated with entanglement of the states of _ those _ electrons that _ do _ traverse the film .
in fact , it is implicitly assumed in @xcite that if stoner excitations are neglected then @xmath74 is close to a pure state ( at least in spin - space ) .
the experimental techniques of @xcite could be used to test this hypothesis .
next , we express the spin transmission asymmetry @xmath46 ( see ( vi ) ) in terms of outgoing states .
let @xmath79 and @xmath80 be incoming states with spin polarization @xmath81 parallel to @xmath24 ( majority spin ) and @xmath82 antiparallel to @xmath83 ( minority spin ) .
let @xmath84 and @xmath85 be outgoing states corresponding to @xmath86 , @xmath87 , respectively .
then @xmath88 as follows from ( [ eq : cons ] ) .
more interesting would be measurements of @xmath89 clearly @xmath90 are parallel or anti - parallel to @xmath83 ; but their lengths @xmath91 ought to be measured . in @xcite , it is tacitly assumed that @xmath92 , and that the states @xmath93 are close to _ pure _ states ; but serious experimental data backing up this hypothesis appear to be lacking .
it is clear that it would be invalidated if `` stoner excitations '' played an important role . in the following ,
we outline a phenomenological description of the experiments in @xcite , assuming that ( [ eq : rhon ] ) and the hypothesis just discussed ( purity of @xmath93 ) are valid .
( a more detailed discussion of the scattering approach to electron transmission and reflection experiments will be presented elsewhere . ) when an incoming electron enters the film it occupies an empty state of the film .
if the film is crystalline this state belongs to a band of states ; let @xmath48 be the corresponding band index .
the state of an electron in band @xmath48 is described by a pauli spinor @xmath94 where @xmath95 and @xmath96 are the components of @xmath97 with spin parallel or anti - parallel to the magnetization @xmath24 , respectively . adopting the approximation of the peierls substitution , the pauli equation for @xmath97 in configuration space has the form @xmath98 where @xmath99 is the band function of band @xmath48 , @xmath100 here
@xmath101 denotes a ( complex ) electrostatic potential , @xmath102 is a ( complex ) _ weiss exchange field _
, @xmath103 is the electromagnetic vector potential , and @xmath104 is an @xmath105-vector potential responsible for _ spin - orbit interactions_. as discussed in @xcite , ( [ eq : cd0 ] ) displays electromagnetic @xmath106-_gauge invariance and _ @xmath105-_gauge invariance _ ,
i.e. , covariance with respect to local @xmath105-rotations in spin space .
a number of important consequences of these gauge symmetries have been pointed out in @xcite . according to ( vii ) above
, effects of the electromagnetic vector potential @xmath107 are apparently negligible ; so @xmath107 is set to 0 .
the electrostatic vector potential @xmath101 is given , approximately , by @xmath108 where @xmath14 is the surface exit work , and the imaginary part , @xmath109 , of @xmath101 provides a phenomenological description of spin - independent inelastic absorption processes inside the film .
velocity - dependent spin - flip processes due to _ spin - orbit interactions _ appear to play a very minor role in the experiments reported in @xcite ; so we may set @xmath110 to 0 , @xmath111 . the weiss exchange field @xmath112 is given by @xmath113 where the real part @xmath114 describes _ exchange interactions _ between the incoming electron and the electron density of the film , and the imaginary part , @xmath115 , yields a phenomenological description of spin - dependent absorption processes .
let @xmath116 denote the pauli spinor describing the state of an electron when it enters the film at some time @xmath117 .
( [ eq : cd0 ] ) can be solved for @xmath118 , @xmath119 , with @xmath120 .
the solution is explicit if @xmath121 , @xmath122 , @xmath111 .
let us suppose that the real part @xmath114 and the imaginary part @xmath115 of the exchange field @xmath112 are both anti - parallel to the magnetization @xmath24 of the film .
then ( [ eq : cd0 ] ) leads us to consider two simple , quasi - one - dimensional scattering problems in a complex potential well of depth @xmath123 with @xmath124 and @xmath125 , for electrons with spin parallel to @xmath126 , or anti - parallel to @xmath127 , respectively .
the solution to these scattering problems can be found in every book on elementary quantum mechanics .
for the purposes of interpreting the results in @xcite , a semi - classical treatment appears to be adequate .
the group velocity , @xmath128 , of an incoming electron wave with energy peaked at @xmath11 and spin up @xmath129 , or down @xmath130 , inside the film can be found by solving the equation @xmath131 for @xmath132 and then setting @xmath133 where @xmath134 is a solution of ( [ eq : xip ] ) chosen such that @xmath135 points in the positive @xmath21-direction .
the sojourn time @xmath136 of the wave inside the film is then given by latexmath:[\[\tau_{\pm } = \frac{d}{v_{\pm } } , \quad \text{with } \quad v_{\pm } = state of the electron with spin up @xmath129 , or spin down @xmath130 , respectively , when it enters the film its state @xmath138 upon leaving the film is then given , approximately , by . ]
@xmath139\right ) \phi_{\text{in}}^{\pm } .\ ] ] the presence of inelastic absorption processes implies that @xmath140 . setting @xmath141\right )
= c \sqrt{1 \pm a},\ ] ] we find that @xmath142 where @xmath46 is the spin transmission asymmetry ( see point ( iv ) and ( [ eq : asy ] ) ) , @xmath31 is the spin precession angle , and @xmath27 is an ( unimportant ) spin - independent phase .
if @xmath143 then @xmath144 , and ( [ eq : phiout ] ) yields @xmath145 thus , measuring @xmath31 and @xmath10 and estimating @xmath146 yields an approximate value for the size of the spin precession angular velocity @xmath147 and hence of the size of the weiss exchange field .
( [ eq : phiout ] ) and ( [ eq : phiout2 ] ) could be mistaken for equations describing _ kaon_- or _ neutrino oscillations _ and are analogous to the equations describing the _ faraday rotation _ of light traversing a magnetized medium .
the origin of the weiss exchange field @xmath114 is hardly a mystery : the spin of an electron traversing the film in a band @xmath48 apparently experiences ( exchange ) interactions with the spins of the occupied states of the film .
since states with spin up and with spin down are occupied asymmetrically ( corresponding to the fact that @xmath148 ) , the net spin density , @xmath149 , at a point @xmath7 in the film is different from zero .
the weiss exchange field @xmath150 is given by @xmath151 where @xmath152 is the strength of the _ exchange coupling _ between spins in the @xmath153 band and those in the occupied band .
the theoretical discussion above is based on a _
mean field ansatz _ : the exchange field @xmath114 in ( [ eq : cd-0 ] ) , ( [ eq : w0c ] ) is chosen to be @xmath154 ( our conventions are such that @xmath155 is parallel to the magnetization @xmath24 . )
one of the surprising implications of the experimental results of @xcite is that , apparently , @xmath152 is quite large , even for rather high - lying bands ( @xmath48 ) , implying that the orbitals of states in such bands must have substantial overlap with those of states in the partially occupied , spin - polarized band .
( [ eq : wx ] ) makes it clear that @xmath156 is a _ dynamical _ field .
one of the main purposes of this paper is to derive the effective quantum dynamics of @xmath156 within a lagrangian functional integral formalism and to sketch what can be accomplished with this formalism .
* by applying an external magnetic field to the film rotating in the @xmath34-plane with angular velocity @xmath157 , the exchange field @xmath114 could be made to rotate around the @xmath25-axis : @xmath158 a polarized beam of electrons shot through this film must exhibit _ bloch spin resonance _ ; but , in this experiment , it would be due to the rotation of the exchange field .
* one may envisage a _
stern gerlach experiment _ for electrons .
one would start by constructing a sandwich of two ferromagnetic metals , i ( e.g. fe ) and ii ( e.g. ni ) , with exchange fields @xmath159 and @xmath160 of different strength , joined by a transition region , a mixture of i and ii , of width @xmath161 .
the transition region would be parallel to the @xmath34-plane ( see fig .
[ fig : exp2 ] ) .
one shoots an unpolarized beam of ( not very hot ) electrons through the sandwich along the transition region between i and ii , as shown in fig .
[ fig : exp2 ] .
one would expect to detect two beams emerging on the other side of the film in slightly different directions that are spin - polarized in opposite directions .
+ the force , @xmath162 , in the @xmath25-direction on an electron with spin up / down inside the film is given , approximately , by @xmath163 it yields a change in the @xmath25-component of the momentum of the electron during its passage through the film given by @xmath164 where @xmath146 is the average group velocity .
the deflection angle @xmath48 is found from @xmath165 of course , as discussed above , the intensity , @xmath166 of the upper beam can be expected to be much larger than the intensity , @xmath167 , of the lower beam , due to spin - asymmetric absorption inside the film . in order to achieve adequate focussing , the incident beam could come from the tip of a scanning tunnelling microscope . * in an experimental set - up similar to the one above , one would force an electron current , spin - polarized in the @xmath25-direction , through the film .
then a hall tension in the @xmath25-direction should be observed ( hall effect for spin currents ; see @xcite ) .
e. brezin , j. c. le guillou and j. zinn - justin , _ field theoretical approach to critical phenomena , _ in _ phase transitions and critical phenomena , _ c. domb and m.s .
green ( eds . )
volume 6 , 125 - 247 , london : academic press 1976 .
d. oberli , _ the ferromagnetic spin filter , _ phd thesis eth zrich , nr .
12933 , ( 1998 ) .
j. frhlich , u. m. studer and e. thiran , _ quantum theory of large systems of nonrelativistic matter _ , in _ fluctuating geometries in statistical mechanics and field
( les houches 1994 ) , f. david , p. ginsparg , j. zinn - justin ( eds . ) ; elsevier 1996 , [ arxiv : cond - mat/9508062 ] .
t. chen , j. frhlich and m. seifert , _ renormalization group methods : landau - fermi liquid and bcs superconductor , _ in _ fluctuating geometries in statistical mechanics and field theory _
, f. david , p. ginsparg , j. zinn - justin ( eds . ) ; elsevier 1996 , [ arxiv : cond - mat/9508063 ] .
j. frhlich and t. spencer , _ phase transitions in statistical mechanics and quantum field theory , _ in _ new developments in quantum field theory and statistical mechanics _ ; m. lvy , p. mitter ( eds . ) , new york , london : plenum 1977 . | the huge spin precession frequency observed in recent experiments with spin - polarized beams of hot electrons shot through magnetized films is interpreted as being caused by zeeman coupling of the electron spins to the so - called _ weiss exchange field _ in the film .
a `` stern - gerlach experiment '' for electrons moving through an inhomogeneous exchange field is proposed .
the microscopic origin of exchange interactions and of large mean exchange fields , leading to different types of magnetic order , is elucidated .
a microscopic derivation of the equations of motion of the weiss exchange field is presented .
novel proofs of the existence of phase transitions in quantum @xmath0-models and antiferromagnets , based on an analysis of the statistical distribution of the exchange field , are outlined . |
decision - making is an important issue in many fields such as signal detection , psychology , radiology , and medicine .
for example , preoperative diagnostic tests are medically necessary and implemented in clinical preventive medicine to determine those patients for whom surgery is beneficial . for the sake of cost - saving or performance improvement , new diagnostic tests are often introduced and the classification accuracies of them are evaluated and compared with the existing ones .
the roc curve , a plot of the true positive rate ( tpr ) versus the false positive rate ( fpr ) for each possible cut point , has been widely used for this purpose when the considered diagnostic tests are continuous .
one advantage of the roc curve is that it describes the inherent classification capability of a biomarker without specifying a specific threshold .
moreover , the invariance characteristic of roc curve in measurement scale provides a suitable base to compare different biomarkers . generally , the more the curve moves toward the point @xmath0 , the better a biomarker performs . in many applications , the area under the roc curve ( auc ) ,
one of the most popular summary measures of the roc curve , is used to evaluate the classification ability of a biomarker .
it has the probability meaning that the considered biomarker of a randomly selected diseased case is greater than that of a non - diseased one .
generally , a perfect biomarker will have the auc of one while a poor one takes a value close to 0.5 . since the auc is the whole area under the roc curve
, relevant information might not be entirely captured in some cases .
for example , two crossed roc curves can have the same auc but totally different performances .
furthermore , there might be limited or no data in the region of high fpr . in view of these drawbacks , it is more useful to see the pauc within a certain range of tpr or fpr . to evaluate the performance of several biomarkers , mcclish ( 1989 )
adopted the summary measure pauc for the fpr over a practically relevant interval . on the other hand ,
jian , metz , nishikawa ( 1996 ) showed that women with false - negative findings at mammography can not be benefited from timely treatment of the cancer .
thus , these authors suggested using the pauc with a portion of the true positive range in their applied data .
although their perspectives are different , the main frame is the same : only the practically acceptable area under the roc curve is assessed . as mentioned by dwyer ( 1997 )
, the pauc is a regional analysis of the roc curve intermediate between the auc and individual points on the roc curve .
the pauc becomes a good measure of classification accuracy because it is easier for a practitioner to determine a range of tpr or / and fpr that are relevant .
several estimation and inference procedures have been proposed by emir , wieand , jung , and ying ( 2000 ) , zhang , zhou , freeman , and freeman ( 2002 ) , dodd and pepe ( 2003 ) , among others .
a more thorough understanding of the roc , auc , and pauc can be also found in zhou , mcclish , and obuchowski ( 2002 ) .
recent research in roc methodology has extended the binary disease status to the time - dependent setting .
let @xmath1 denote the time to a specific disease or death and @xmath2 represent the continuous diagnostic marker measured before or onset of the study with joint survivor function @xmath3 .
for each fixed time point @xmath4 , the disease status can be defined as a case if @xmath5 and a control otherwise . to evaluate the ability of @xmath2 in classifying subjects who is diseased before time @xmath4 or not ,
heagerty , lumley , and pepe ( 2000 ) generalized the traditional tpr and fpr to the time - dependent tpr and fpr as @xmath6 and @xmath7 , which can be further derived to be @xmath8 and @xmath9 with @xmath10 . for the time - dependent auc , chambless and diao ( 2006 ) , chiang , wang , and hung ( 2008 ) , and chiang and hung ( 2008 ) proposed different nonparametric estimators and developed the corresponding inference procedures .
as for the time - dependent pauc , there is still far too little research on this topic . we propose nonparametric estimators , which are shown to converge weakly to gaussian processes , and the estimators for the corresponding variance - covariance functions .
the established properties facilitate us to make inference on the time - dependent pauc and can be reasonably applied to the time - dependent auc because it is a special case of this summary measure .
the rest of this paper is organized as follows . in section 2 ,
the nonparametric estimation and inference procedures are proposed for the time - dependent pauc .
the finite sample properties of the estimators and the performance of the constructed confidence bands are studied through monte carlo simulations in section 3 .
section 4 presents an application of our method to the actg 175 study . in this section ,
an extended inference procedure is further provided for the comparison of the time - dependent paucs .
some conclusions and future works are addressed in section 5 .
finally , the proof of main results is followed in the appendix .
in this section , we estimate the time - dependent pauc and develop the corresponding inference procedures . without loss of generality , the time - dependent pauc is discussed for restricted @xmath11 because that for restricted @xmath12 can be derived in the same way by reversing the roles of case and control subjects .
let @xmath13 be the minimum of @xmath1 and censoring time @xmath14 , @xmath15 represent the censoring status , and @xmath16 , @xmath17 $ ] , denote the @xmath18 quantile of @xmath2 conditioning on @xmath19 at the fixed time point @xmath4 .
following the expression @xmath20 for the time - dependent auc , the time - dependent pauc @xmath21 with the @xmath11 less than @xmath22 is derived to be functional of @xmath23 : @xmath24 note that the value of @xmath21 for a perfect biomarker should be @xmath22 while a useless one is @xmath25 .
same with the interpretation of cai and dodd ( 2008 ) , the rescaled time - dependent pauc @xmath26 can be explained as the probability that the test result of a case @xmath27 is higher than that of a control @xmath28 with its value exceeding @xmath29 for @xmath30 , i.e. , @xmath31 . from the formulation in ( [ 1.1 ] ) , an estimator of @xmath21 can be obtained if @xmath23 is estimable . under marker
dependent censoring ( @xmath1 and @xmath14 are independent conditioning on @xmath2 ) , akritas ( 1994 ) suggested estimating @xmath23 by @xmath32 , where @xmath33 is an estimator of @xmath34 with @xmath35 and @xmath36 being estimators of @xmath37 and @xmath38 . here , @xmath39 and @xmath40 is a nonnegative smoothing parameter . substituting @xmath41 for @xmath23 in ( [ 1.1 ] ) ,
@xmath21 is proposed to be estimated by @xmath42 where @xmath43 , @xmath44 , @xmath45 , and @xmath46 . in the appendix
, we show that @xmath47 is uniformly approximated by @xmath48 and converges weakly to a mean zero gaussian process with variance - covariance function @xmath49 $ ] for @xmath50 $ ] and @xmath51 .
the application of kernel function @xmath52 provides the nearest neighbor estimator of @xmath23 .
an alternative choice of kernel function is possible and will lead to a different estimator of @xmath21 . as mentioned in akritas ( 1994 ) ,
the asymptotic properties of @xmath41 is irrelevant to the choice of kernel function under some regularity conditions and so is @xmath53 .
the author further showed that any other estimator for @xmath23 is at least as dispersed as @xmath41 and the choice of @xmath40 is irrelevant to the measurement scale of @xmath2 .
it is not difficult to see that the estimation problem of @xmath21 becomes that of @xmath23 . from this perspective
, the proposed estimation procedure can be extended to any censoring or truncation mechanisms provided that @xmath23 is estimable .
alternatively , one might be interested in making inference on the time - dependent pauc @xmath54 over the range @xmath55 $ ] , @xmath56 , of @xmath57 .
the estimator @xmath58 is suggested and the limiting gaussian process of @xmath59 is a direct consequence of the large sample property of @xmath53 . when the complete failure time data @xmath60 are available , @xmath23 can be estimated by an empirical estimator @xmath61 .
a natural estimator for @xmath21 is obtained as @xmath62 where @xmath63 , @xmath64 , and @xmath65 . by substituting the disease and disease - free groups for the time - varying case and control ones , @xmath66 and the estimator @xmath67 will reduce to the time - invariant pauc and the nonparametric estimator of dodd and pepe ( 2003 ) . by the similar argument as in the proof of the asymptotic gaussian process of @xmath53 , it is straightforward to derive that @xmath68 converges weekly to a gaussian process with mean zero and variance - covariance function @xmath69 $ ] , where @xmath70 @xmath71 - 2h(t , y)$ ] , @xmath72 , and @xmath73 .
the confidence intervals for @xmath21 and @xmath74 can be constructed by the asymptotic gaussian processes and the estimated variance - covariance functions . replacing the parameters with their sample analogues ,
@xmath75 is proposed to be estimated by @xmath76 where @xmath77 , @xmath78 , and @xmath79 with @xmath80 , @xmath81 , and @xmath82 .
thus , it is straightforward to have an estimated variance - covariance function @xmath83 and a @xmath84 , @xmath85 , pointwise confidence interval for @xmath21 : @xmath86 where @xmath87 is the @xmath88 quantile value of the standard normal distribution . with the independent and identically distributed representation @xmath48 , the re - sampling technique of lin , wei , yang , and ying ( 2000 ) is applied to determine a critical point @xmath89 so that @xmath90}|\frac{\sqrt{n}(\widehat{\theta}_t(\widehat{q}_{\alpha t})-\theta_t(q_{\alpha t}))}{\widehat{\sigma}_{\alpha}^{1/2}(t , t)}|<l_{\varsigma})\doteq 1-\varsigma \label{2.5}\end{aligned}\ ] ] for a subinterval @xmath91 $ ] of interest within the time period @xmath92 $ ] .
the validity of ( [ 2.5 ] ) enables us to construct a @xmath84 simultaneous confidence band for @xmath93\}$ ] via @xmath94\}.\label{2.6}\end{aligned}\ ] ] note that both pointwise and simultaneous confidence bands for @xmath74 can be constructed as the above ones .
when @xmath67 is applicable , the confidence bands are easily obtained by substituting @xmath95 for @xmath96 in ( [ 2.4 ] ) and ( [ 2.6 ] ) , where @xmath97 @xmath98 , @xmath99 , and @xmath100 .
in this section , monte carlo simulations are conducted to investigate the finite sample properties of the proposed estimators and the performance of the inference procedures .
the continuous biomarker @xmath2 is designed to follow a standard normal distribution .
conditioning on @xmath101 , the failure time @xmath1 and the censoring time @xmath14 are independently generated from a lognormal distribution with parameters @xmath102 and @xmath103 , and an exponential distribution with scale parameter @xmath104 , where the constant @xmath105 is set to produce the censoring rates of @xmath106 , @xmath107 and @xmath108 . in our numerical studies ,
500 data sets of 500 and 1000 observations are simulated .
the estimators and the pointwise confidence intervals of @xmath66 are evaluated at the selected time points @xmath109 , @xmath110 , and @xmath111 with @xmath1120.1 , 0.2 , and 0.3 , where @xmath113 is the @xmath114th quantile of the distribution of @xmath1 . moreover , the simultaneous confidence bands for @xmath66 over the subintervals @xmath115 $ ] and @xmath116 $ ] are considered . since
a small portion of cases or controls occur outside @xmath116 $ ] under the above design , the simulation results are presented within this time period .
when survival times are subject to censoring , an appropriate smoothing parameter urgently becomes necessary in the estimation of @xmath66 .
it usually attempts to select a bandwidth that minimizes the asymptotic mean squared error of an estimator , which is obtained by using the plug - in method for unknown parameters .
this approach , however , would lead to further bandwidth selection problems and is infeasible in our current setting . for the bandwidth selection ,
we propose a simple and easily implemented data - driven method .
this procedure is to find a bandwidth , say , @xmath117 which minimizes the following integrated squared error @xmath118 where @xmath119 is the kaplan - meier estimator computed based on the data @xmath120 , @xmath121 , @xmath122 is computed as @xmath123 with the @xmath124th observation @xmath125 deleted , and @xmath126 . the rationale behind ( [ 3.1 ] )
is that @xmath127 can be shown to be an independent censored sample from a uniform distribution @xmath128 under the validity of conditionally independent censoring . for each generated sample ,
@xmath129 and @xmath130 s are computed by using @xmath117 and the subjective bandwidths of 0.01 and 0.2 . among the 500 simulated samples ,
the bandwidths obtained from minimizing @xmath131 in ( [ 3.1 ] ) have a range between 0.01 and 0.2 , and medians of about 0.09 , 0.1 , 0.1 , 0.07 , 0.08 , and 0.08 for @xmath132 of @xmath133 , @xmath134 , @xmath135 , @xmath136 , @xmath137 , and @xmath138 , where @xmath139 and @xmath140 represent the sample size and the censoring rate .
tables 1 - 3 summarize the averages and standard deviations of estimates , the standard errors , and the empirical coverage probabilities of 0.95 pointwise confidence intervals for @xmath66 . for the complete failure time data ( i.e. , @xmath141 ) , @xmath142 and @xmath129 , computed using @xmath117 , give separately a slight overestimate and underestimate of @xmath66 .
moreover , the variance @xmath143 tends to be underestimated , which leads to a lower coverage probability .
as expected , the bias and standard deviation of @xmath129 will separately increase and decrease as the bandwidth becomes larger .
it is also detected from these tables that the poor estimates of @xmath144 s appear at extremely small or large bandwidths . in the numerical studies , the larger bias of @xmath145
is found and the main reason for this is because it is computed via comparing only ( at most ) the top ten percent of subjects in the control group with those in the case group . for @xmath146 , the availability of data used for statistical analysis is expanded and , hence , the performance becomes better .
the biases of the proposed estimators is indistinguishable in the presence of heavy censoring , whereas the standard deviations and the standard errors will become larger . at each simulated sample
, we can see that the estimators using @xmath117 provide generally satisfactory results .
the empirical coverage probabilities of pointwise confidence intervals ( tables 1 - 3 ) show the good performance of bandwidths selected from the automatic selection procedure for interval estimation , except for @xmath147 of @xmath148 and @xmath149 at the time point @xmath111 . however , most of the coverage probabilities are lower than 0.95 for the bandwidths of 0.01 and 0.2 . for samples without censoring , it is revealed from tables 1 - 3 that the empirical coverage probabilities of the pointwise confidence intervals computed based on @xmath129 with @xmath117 are more close to the nominal level of 0.95 than those based on @xmath150 .
similar conclusions can be also drawn from table 4 for the simultaneous coverage probabilities .
the empirical coverage probabilities of the simultaneous confidence band ( [ 2.6 ] ) with @xmath117 roughly stay around 0.95 except for the wider interval @xmath116 $ ] .
in the actg 175 study ( hammer et al . ( 1996 ) ) , the classification accuracy of cd4 cell counts on the time in weeks from entry to aids diagnosis or death might depend on whether they received the prior antiretroviral therapy .
a total of 2467 hiv-1-infected patients , which were recruited between december 1991 and october 1992 , are considered . of these patients
, 1395 received the prior therapy while the rest 1072 did not receive the therapy . during the study period ,
308 patients died of all causes or were diagnosed with aids . for a negative association between cd4 counts and the time to aids and death , we let @xmath2 be a strictly decreasing function of the cd4 marker and @xmath1 be the minimum of time - to - aids and time - to - death . currently , there is still no standard of clinically meaningful values of fpr for the pauc in aids research . in this data analysis ,
we restrict our attention to the pauc of @xmath2 with the fpr less than 0.1 or 0.2 or 0.3 . to simplify the presentation ,
the marker and the time - dependent pauc for non - therapy and therapy patients are denoted separately by @xmath151 and @xmath152 .
based on two independent data sets @xmath153 and @xmath154 , @xmath155 and @xmath156 are computed as @xmath129 in ( [ 2.1 ] ) using the bandwidths of 0.042 and 0.069 , which are the minimizers of @xmath131 in ( [ 3.1 ] ) .
the confidence intervals for @xmath157 s are further constructed from ( [ 2.4 ] ) and ( [ 2.6 ] ) . due to the large variation of estimators before week 98
, we only provide the estimated time - dependent paucs and 0.95 pointwise and simultaneous confidence bands from week 98 to the end of study in figures 1 ( a)-(f ) .
based on the summary measures @xmath158 s and compared with 0.005 , one can conclude from the simultaneous confidence bands that the cd4 count is a useless biomarker in classifying patient s survival time within the considered time period for both therapy and non - therapy patients .
however , a different conclusion will be drawn at each time point based on the pointwise confidence intervals .
the time - dependent paucs @xmath159 and @xmath160 are detected to be significantly higher than @xmath161 for @xmath22=0.2 and 0.3 after week 110 and week 98 , respectively .
figures 1 ( a ) , ( c ) , and ( e ) give a clear indication that the paucs decrease slightly over time for patients without prior therapy .
however , the paucs stay very close to a constant throughout the study period for those with prior therapy ( figures 1 ( b ) , ( d ) , and ( f ) ) .
the difference in the classification accuracies of @xmath162 and @xmath163 can be measured by the summary index @xmath164 .
when @xmath165 , @xmath166 is the usual comparison of aucs in the time - dependent setting .
it is natural to estimate @xmath166 by @xmath167 . along the same lines as the proof in the appendix
, we can derive that @xmath168 converges weakly to a mean zero gaussian process with variance - covariance function @xmath169+(1-\kappa)^{-1}e[\psi_{\alpha i}^{(2)}(s)\psi_{\alpha i}^{(2)}(t)]$ ] provided that @xmath170 ( @xmath171 ) as @xmath172 , where @xmath173 is a counterpart of @xmath75 , @xmath174 . to make inference on @xmath166 , @xmath175 is first estimated by @xmath176 a @xmath84 pointwise confidence interval for @xmath166 and a @xmath84 simultaneous confidence band for @xmath177\}$ ] are separately given via @xmath178\}\label{4.6}\end{aligned}\ ] ] with @xmath179 being obtained as ( [ 2.5 ] ) .
it is revealed in figures 2 ( a)-(c ) that @xmath166 , @xmath22=0.1 , 0.2 , and 0.3 , tend to be positive within the study period and the difference becomes negligible as @xmath22 increases .
in other words , with small values of @xmath11 , a prior antiretroviral therapy might lower the discrimination ability of cd4 counts in classifying subject s t - week survival .
one possible explanation for this conclusion is that the prior therapy makes patients more homogeneous in survival time and cd4 counts .
the estimates are further found to be around zero after about week 160 .
it means that for long term survival classification the performance of cd4 counts is irrelevant to whether patients receive prior therapy or not . due to the large variability in the data
, we could not detect any significant difference between @xmath159 and @xmath160 .
it would necessitate extremely large sample sizes to enable demonstration of significant differences between the paucs .
for the time - dependent pauc , it was traditionally estimated by the trapezoidal numerical integration method . the derivation for its
sampling distribution becomes complicated and the computation load is prohibitively expensive .
although the inferences can be developed through a bootstrap technique , there is still no rigorous theoretical justification for this procedure .
we can see in this article that the proposed estimators are simple and have explicit mathematical expressions .
the confidence bands are built based on the asymptotic gaussian process of the estimators as well as the corresponding estimates of the asymptotic variances .
the estimation and inference procedures are further shown to be useful through simulation studies and an application to the actg 175 data .
it is detected from our numerical studies that the performance of the proposed estimator @xmath53 is very sensitive to small value of @xmath22 . to obtain a more stable estimate of @xmath21 , a large sample size relative to @xmath22
is usually suggested especially in the presence of censoring .
moreover , the price for the assumption of marker dependent censoring is to find an appropriate bandwidth in estimation .
to this problem , we propose a simple and easily implemented selection procedure and show its good performance through simulations . as for the estimation of @xmath180
, it can be also derived via using the bivariate estimation methods of campbell ( 1981 ) or burke ( 1988 ) for @xmath23 .
however , these estimators are only valid under independent censorship which appears to be very limited and may not always be met in applications .
one advantage of totally independent censoring assumption is that no smoothing technique is required
. in some empirical examples , censored survival data of the form @xmath181 are often occurred in a paired design with @xmath182 being the different biomarkers of the @xmath124th subject .
the scientific interest usually focuses on comparing the discrimination abilities of @xmath162 and @xmath163 on subject s survival status at each time point within the study period .
obviously , the assumptions of marker dependent censoring made separately on @xmath183 and @xmath184 are often unreasonable in practice . under a more flexible assumption of conditionally independent censorship ( @xmath1 and @xmath14 are independent conditioning on @xmath185 ) , the estimated joint survivor function of @xmath1 and @xmath162 and that of @xmath1 and @xmath163 in this article are quite inadequate in the estimation of @xmath166 without modification .
it is worthwhile to investigate the associated comparison procedure in our future study . for the proof of main results , the assumptions in akritas ( 1994 ) and the conditions ( * a1 * : @xmath186 exists with @xmath187 ) and ( * a2 * : @xmath188 as @xmath189 ) are made throughout the rest of this paper . from theorem 3.1 of akritas ( 1994 ) , one has @xmath190 where @xmath191 with @xmath192 @xmath193 , @xmath194 , and @xmath195 .
let @xmath196 and @xmath197 $ ] .
the uniform consistency of @xmath123 ( cf . dabrowska ( 1987 ) ) ensures that @xmath198 with @xmath199 . by a direct calculation and
( [ a.5 ] ) , a simplified form of the second term in the righthand side of ( [ a.1 ] ) is obtained as follows : @xmath200 where @xmath201 .
similarly , the third term can be expressed as @xmath202 with @xmath203 .
it follows from ( [ a.1])-([a.3 ] ) , the decomposition of a u - statistic into a sum of degenerate u - statistics ( serfling ( 1980 ) ) , and corollary 4 of sherman ( 1994 ) that @xmath204 where @xmath205 - 2h(t , y ) + ( s_y(y_i)-s(t , y))\xi_i(t)i(y_i > y)$ ] . by the taylor expansion of @xmath206 at @xmath207 , ( [ a.5 ] ) , and ( [ a.4 ] )
, one has @xmath208 where @xmath209 .
thus , @xmath210 can be shown to converge to a mean zero gaussian process by an application of the functional cental limit theorem .
for the asymptotic gaussian process of @xmath53 , it is established through the equality @xmath211 .
let @xmath212 with @xmath213 . by assumptions ( a1)-(a2 ) ,
the hadamard differentiability of @xmath214 is a direct result of lemma a.1 in daouia , florens , and simar ( 2008 ) . together with the functional delta method ( cf .
van der vaart ( 2000 ) ) , we have @xmath215 and the weak convergence of @xmath216 . it is further ensured by ( a version of ) lemma 19.24 of van der vaart ( 2000 ) and ( [ a.6 ] ) that @xmath217 moreover , the first order taylor expansion of @xmath218 at @xmath219 , the continuity of @xmath220 , @xmath221 , and the continuous mapping theorem imply that @xmath222 it follows from ( [ a.6])-([a.10 ] ) that @xmath223 @xmath224 finally , the proof is completed by applying the functional central limit theorem to the approximated term @xmath48 in ( [ a.11 ] ) .
: : akritas , m. g. ( 1994 ) . nearest neighbor estimation of a bivariate distribution under random censoring . _
annals of statistics _ * 22 * , 1299 - 1327 .
: : burke , m. d. ( 1988 ) .
estimation of a bivariate distribution function under random censorship
. _ biometrika _ * 75 * , 379 - 382 .
: : cai , t. and dodd , l. e. ( 2008 ) .
regression analysis for the partial area under the roc curve .
_ statistica sinica _ * 18 * , 817 - 836 .
: : campbell , g. ( 1981 ) .
nonparametric bivariate estimation with randomly censored data
. _ biometrika _ * 68 * , 417 - 423 .
: : chambless , l. e. and diao , g. ( 2006 ) .
estimation of time - dependent area under the roc curve for long - term risk prediction .
_ statistics in medicine _ * 25 * , 3474 - 3486 . : : chiang , c. t. , wang , s. h. , and hung , h. ( 2008 ) . random weighting and edgeworth expansion for the nonparametric time - dependent auc estimator . _ statistic sinica_. accepted .
: : chiang , c. t. and hung , h. ( 2008 ) .
nonparametric estimation for time - dependent auc
. technical report , national taiwan university .
: : dabrowska , d. m. ( 1987 ) .
uniform consistency of nearest neighbor and kernel conditional kaplan - meier estimates .
technical report no .
86 , univ .
california , berkeley .
: : daouia , a. , florens , j. p. , and simar , l. ( 2008 )
. functional convergence of quantile - type frontiers with application to parametric approximations . _ journal of statistical planning and inference _ * 138 * , 708 - 725 . : : dodd , l. and pepe , m. s. ( 2003 ) . partial auc estimation and regression . _
biometrics _ * 59 * , 614 - 623 .
: : dwyer , a. j. ( 1996 ) .
in pursuit of a piece of the roc . _ radiology _ * 201 * , 621 - 625 . : : emir , b. , wieand , s. , jung , s. h. , and ying , z. ( 2000 ) .
comparison of diagnostic markers with repeated measurements : a non - parametric roc curve approach .
_ statistics in medicine _ * 19 * , 511 - 523 . : : hammer , s. m. , katzenstein , d. a. , hughes , m. d. , gundacker , h. , schooley , r. t. , haubrich , r. h. , henry , w. k. , lederman , m. m. , phair , j. p. , niu , m. , hirsch , m. s. , and merigan , t. c. ( 1996 ) . a trial comparing nucleoside monotherapy with combination therapy in hiv - infected adults with cd4 cell counts from 200 to 500 per cubic millimeter .
_ the new england journal of medicine _ * 335 * , 1081 - 1090 .
: : heagerty , p. j. , lumley , t. , and pepe , m. s. ( 2000 ) . time - dependent roc curves for censored survival data and a diagnostic marker
. _ biometrics _ * 54 * , 124 - 135 .
: : jiang , y. , metz , c. e. , and nishikawa , r. m. ( 1996 ) .
a receiver operating characteristic partial area index for highly sensitive diagnostic tests .
_ radiology _ * 201 * , 745 - 750 . : : lin , d. y. , wei , l. j. , yang , i. , and ying , z. ( 2000 ) .
semiparametric regression for the mean and rate functions of recurrent events .
_ journal of the royal statistical society _ * b62 * , 711 - 730 .
: : mcclish , d. k. ( 1989 ) . analyzing a portion of the roc curve
_ medical decision making _ * 9 * , 190 - 195 . : : serfling , r. j. ( 1980 ) .
approximation theorems of mathematical statistics .
new york : wiley .
: : sherman , r. p. ( 1994 ) .
maximal inequalities for degenerate u - processes with applications to optimization estimators . _
annals of statistics _
* 22 * , 439 - 459 .
: : van der vaart , a. w. ( 2000 ) .
asymptotic statistics .
cambridge : cambridge university press .
: : zhang , d. d. , zhou , x. h. , freeman , daniel h. j. , and freeman , j. l. ( 2002 ) . a non - parametric method for the comparison of partial areas under roc curves and its application to large health care data set .
_ statistics in medicine _ * 21 * , 701 - 715 .
: : zhou , x. h. , mcclish , d. k. , and obuchowski , n. a. ( 2002 ) . statistical methods in diagnostic medicine .
new york : wiley . | to assess the classification accuracy of a continuous diagnostic result , the receiver operating characteristic ( roc ) curve is commonly used in applications .
the partial area under the roc curve ( pauc ) is one of widely accepted summary measures due to its generality and ease of probability interpretation . in the field of life science , a direct extension of the pauc into the time - to - event setting
can be used to measure the usefulness of a biomarker for disease detection over time . without using a trapezoidal rule ,
we propose nonparametric estimators , which are easily computed and have closed - form expressions , for the time - dependent pauc .
the asymptotic gaussian processes of the estimators are established and the estimated variance - covariance functions are provided , which are essential in the construction of confidence intervals .
the finite sample performance of the proposed inference procedures are investigated through a series of simulations .
our method is further applied to evaluate the classification ability of cd4 cell counts on patient s survival time in the aids clinical trials group ( actg ) 175 study .
in addition , the inferences can be generalized to compare the time - dependent paucs between patients received the prior antiretroviral therapy and those without it .
= 0.3 in |
the groundbreaking achievements in experiments with ultracold fermi gases @xcite have opened up unprecedented possibilities to study new regimes of strongly interacting quantum matter .
ultracold gases represent well - controllable model systems for the exploration of many - body regimes in a way not possible in conventional condensed - matter systems @xcite .
a new frontier in the field is currently being explored in experiments on ultracold fermi - fermi mixtures of @xmath0li and @xmath1k atoms @xcite . because of the mass imbalance and the possibility to apply species - specific optical potentials , such systems promise manifold intriguing applications both in many - body physics @xcite and few - body physics @xcite . to prepare degenerate fermi gases ,
all - optical approaches have proven to be simple and robust and they facilitate highly efficient evaporative cooling
. therefore they are routinely applied in many present experiments ; see ref .
@xcite for a review of earlier work and refs .
@xcite for more recent examples .
spin mixtures of @xmath0li atoms near a broad feshbach resonance are particularly well suited for this cooling approach because of their exceptional collision properties , which offer extremely large cross sections for elastic collisions in combination with very weak inelastic decay .
this favorable situation motivates the general idea of using the strongly interacting @xmath0li gas as a cooling agent for sympathetic cooling of another species .
following this idea in ref .
@xcite , we recently demonstrated the sympathetic cooling of @xmath1k atoms by an evaporatively cooled , optically trapped spin mixture of @xmath0li , reaching the double - degenerate regime . in this article , we first present more details on our all - optical approach of preparing a double - degenerate fermi - fermi mixture of @xmath0li and @xmath1k .
we then show new results related to interactions and molecule formation near interspecies feshbach resonances . in sec .
[ sec : trapping ] , we discuss our dual - species cooling and trapping setup and the special loading procedures used for the optical traps . in sec .
[ sec : spinrelax ] , we present an important preparation step where spin relaxation in the mixture brings the k atoms into their lowest internal state . in sec . [
sec : evaporation ] , we describe the combined evaporative and sympathetic cooling process . in sec .
[ sec : stateprep ] , we show how the mixture can be prepared near interspecies feshbach resonances . in sec .
[ sec : molecules ] , we finally demonstrate the creation of ultracold heteronuclear fermi - fermi molecules by feshbach association methods .
li and @xmath1k versus magnetic field . ] here , we outline the basic concept of our dual - species setup ( sec . [ sec : setup ] ) , and we present the special procedures applied to prepare the optically trapped mixture . in sec .
[ sec : mot ] we describe how we operate a dual - species magneto - optical trap ( mot ) . in sec .
[ sec : odt ] we present the optical dipole traps ( odt ) used in the experiments . in sec . [
sec : transfer ] we discuss the special loading procedure for the odt .
the whole scheme is optimized for a large number of @xmath0li atoms , as this species is used as the cooling agent for sympathetic cooling of @xmath1k into degeneracy @xcite .
figure [ fig : levelscheme ] shows the atomic ground state energy structure of @xmath0li and @xmath1k .
we label the energy levels li@xmath2 and k@xmath3 , counting the states with rising energy .
the hyperfine splitting of @xmath0li is 228.2mhz . for @xmath1k ,
the hyperfine structure is inverted and the splitting amounts to 1285.8mhz @xcite . for the low - lying states with @xmath4 and @xmath5 , the projection quantum numbers are given by @xmath6 and @xmath7 .
for the cooling and trapping of li and k we apply standard laser cooling and trapping techniques @xcite combining a zeeman - slowed atomic beam and a dual - species mot for initial collection of atoms in the vacuum chamber .
a detailed description of the experimental setup and the laser systems can be found in ref .
a dual - species oven , which is connected to the main chamber via a differential pumping section , delivers a well - collimated atomic beam .
we operate the oven with isotopically enriched samples containing 80% of @xmath0li and 7% of @xmath1k .
the zeeman slower can cool both species individually with the respective settings of the magnetic field gradients .
the central element of our vacuum chamber is a glass cell that allows for very good optical access .
we achieve excellent vacuum conditions with a pressure on the order of 10@xmath8 mbar . for both species we use diode laser systems with
one grating stabilized master oscillator in combination with tapered amplifiers .
the li ( k ) laser system provides 11mw ( 12mw ) per mot beam and 80mw ( 100mw ) for the zeeman slower beam .
figure [ fig : optpump ] shows a schematic drawing of the atomic energy levels and optical transitions used for cooling and trapping of li and k. the li mot laser beams contain two frequency parts tuned to the cooling ( @xmath9 ) and repumping ( @xmath10 ) transitions and having equal power . for k the cooling ( @xmath11 ) to repumper ( @xmath12 ) ratio is three to two .
the initial collection and cooling of @xmath0li and @xmath1k is achieved by conventional laser - cooling and trapping techniques .
as loading of both species requires different settings of the zeeman slower magnetic field , we use a sequential mot - loading scheme .
the basic idea is to first load li and then add k in the presence of the li mot .
the li mot is operated with a field gradient of 26g / cm along the symmetry axis of the field coils and a laser detuning of @xmath13mhz , for both , cooling and repumping transition . after a loading time of about 15s
, a few @xmath14 li atoms are accumulated in the mot . at this point
we increase the magnetic field gradient to 38g / cm , where the k mot works optimally . in 5s , about @xmath15 k atoms are added to the trap .
the k mot is operated with a laser detuning of @xmath16mhz . during the k loading phase we operate the li mot with a relatively large detuning of @xmath17mhz in order to compensate for the effect of the higher magnetic field gradient on volume and density .
this avoids increased losses by inelastic interspecies collisions , enabling the efficient sequential loading of both species . in order to reduce shot - to - shot fluctuations of the number of atoms in the trap , we control the li and k mot loading by monitoring their fluorescence independently .
when the fluorescence of the li mot reaches its threshold value , li loading is stopped and loading of the k mot is initiated .
once the k mot fluorescence reaches its threshold , the k loading is turned off . at this point
the odt is ramped on in 100ms , and the magnetic fields of the zeeman slower are ramped to zero in 10ms . for further storage and cooling of the atomic cloud ,
we use a trapping scheme that employs the optical dipole force of an intense infrared laser beam @xcite and the magnetic force in the curvature of the magnetic field @xcite .
the latter becomes important when the optical trap is operated with low laser power .
a schematic drawing of this hybrid trapping scheme is shown in fig .
[ fig : traps ] .
the principal odt is formed by a single beam ( beam 1 ) delivered by a 200-w fiber laser that emits light at a central wavelength of 1070 nm ( ipg ylr-200sm - lp ) .
the beam is focused down to a waist of 38@xmath18 m at the center of the mot . during loading
, the trap is operated with an optical power of 96w , which results in a depth of 2.6mk for li . for k
the trap depth is larger by a factor of 2.1 and thus amounts to about 5.5mk .
two sets of magnetic field coils are used in our setup to control the bias and curvature field independently ; the coils setup is described in more detail in appendix [ apx : magnets ] . for
small bias fields the magnetic confinement is very small in the axial direction ; here additional axial confinement can be obtained from another infrared beam ( beam 2 ) delivered by a 5-w fiber laser with a central wavelength of 1065 nm ( ipg ylp-5-lp ) .
the single beam is focused to a waist of 97@xmath18 m and intersects beam 1 at an angle of @xmath19 .
loading cold atoms of a single species from a mot into a dipole trap is a standard procedure in many experiments .
sub - doppler cooling and mot compression are common methods to enhance transfer into the optical trap .
the optimized loading of two species , however , needs special procedures .
here we describe the sequential loading scheme that gives us excellent starting conditions for the evaporation of li and k in a common optical trap . a schematic illustration of the loading and transfer sequence is shown in fig .
[ fig : sequence ] .
we found that an optimum is achieved by first transferring k into the optical trap while keeping li stored in the large - volume , low - density mot and then performing the li transfer .
after switching on the odt , in a first step the k mot is compressed by ramping up the magnetic field gradient within 50ms to 96g / cm and bringing the frequencies of the k lasers closer to resonance to a detuning of a few mhz with an intensity lowered to 70% . at the same time
the detuning of the li mot is increased to @xmath20mhz to avoid compression of the li mot . at this point
the k mot light is extinguished and the k atoms are confined in the dipole trap while li is stored practically undisturbed in a mot at a reduced magnetic field gradient of 64g / cm . with the k mot
beams off , untrapped atoms are allowed to escape for 50ms .
potassium has ten zeeman sublevels in the lowest hyperfine ground state ; see fig .
[ fig : levelscheme ] . in order to produce a spin - polarized sample of k in its lowest internal state
, we apply an optical pumping scheme that not only transfers the atoms to the lower hyperfine state , but also pumps the atoms to the lowest @xmath21 state . for optical pumping the quadrupole field
is switched off for 2ms and only a small guiding field is kept on .
parallel to the field we shine in a @xmath22-polarized laser beam for 10@xmath18s , which optically pumps the k atoms into state @xmath23 .
the optical pumping beam contains two frequency components , one for zeeman pumping tuned to the ( @xmath24 ) transition and another one for hyperfine pumping tuned to the ( @xmath12 ) transition as shown in fig .
[ fig : optpump ] .
each frequency component has about 50 times the saturation intensity . during the optical pumping stage ,
the cloud of li atoms remains trapped in an optical molasses and can be recaptured without significant losses .
the high - power trapping laser induces a large light - shift on the optical cooling and pumping transitions .
potassium has two optical transitions between the excited 4@xmath25p@xmath26 state and the 5@xmath25s@xmath27 and 3@xmath25d@xmath28 states with wavelengths of 1252 nm and 1177 nm , respectively . at high intensities of the odt
these two transitions shift the 4@xmath25p@xmath26 level by several 100mhz .
therefore optical pumping can not be performed in the trap and we have to switch it off for a short time . after the 10@xmath18s off - time of the odt needed for the optical pumping , essentially all k atoms are recaptured into the odt . at this point
we have a sample of a few 10@xmath29 polarized @xmath1k atoms in the odt , surrounded by a magneto - optically trapped cloud of @xmath0li atoms .
we finally apply a compressed mot stage for li in order to efficiently load this species into the dipole trap . for this
, the quadrupole field is ramped back up to 64g / cm and the mot lasers are operated at a very small detuning of @xmath30mhz from resonance while the power is lowered to 180@xmath18w per beam for a duration of 15ms .
hyperfine pumping of li to the lower state is performed by switching the repumping laser off during the last 50@xmath18s of the pulse . with this sequence
we obtain a few 10@xmath31 li atoms in the lowest two spin states in the odt at a temperature of about 300@xmath18k .
a li@xmath2k@xmath3 mixture can undergo rapid decay via spin relaxation if exoergic two - body collisions can take place that preserve the total projection quantum number @xmath32 .
in such a process , @xmath33 the energy @xmath34 is released .
whenever one of the species is in the absolute ground state , and the other one is in a low - lying state ( @xmath35 and @xmath5 or @xmath36 and @xmath4 ) , spin relaxation is strongly suppressed @xcite .
since optical zeeman pumping does not lead to a perfect transfer of all k atoms into the lowest spin state , we exploit spin relaxation to fully spin polarize k into state @xmath23 .
we investigated the conditions under which spin relaxation can be used for this purpose . in these measurements
we apply only hyperfine pumping , but no zeeman pumping .
we start with an almost equal mixture of li in its lowest two hyperfine states , @xmath23 and @xmath37 , trapped together with a population of k in all zeeman substates @xmath38 .
( @xmath23 ) atom number , the filled circles are k@xmath23 .
the two pronounced features that are visible at 40 g and 207 g are fitted by lorentzians to determine their positions and widths . ]
we investigate the magnetic field dependence of the spin relaxation by holding the sample for 500ms at a variable magnetic field .
the trap is operated under the same conditions as during trap loading , i.e. with a trap depth of 2.6mk for li and 5.5mk for k. the atom numbers are measured using spin - selective absorption images , which are always taken at a bias field of 1190 g @xcite .
figure [ fig : spinrelaxbfield ] shows the resulting atom numbers of li in states @xmath23 ( open squares ) and @xmath37 ( filled squares ) as well as k in state @xmath23 ( filled circles ) .
two distinct peaks in the k@xmath23 atom number are visible at 40 g and 207 g and coincide with dips in the li@xmath37 atom number .
these features are fitted with lorentzians to determine their positions and widths .
the release energy @xmath34 at 40 g ( 207 g ) corresponds to 2.1mk ( 5.8mk ) . for an inelastic collision between two atoms with different masses , the resulting kinetic energy contributions are inversely proportional to the mass . for the @xmath0li-@xmath1k combination , 87% of the released energy
is transferred to the lighter li atoms ( mass @xmath39 ) and only 13% to the heavier k ( mass @xmath40 ) .
a necessary condition for the trap depth @xmath41 under which a k atom stays confined is @xmath42 and analogously for @xmath43 .
the mass factor along with the about two times larger trap depth for k explains why we observe loss of li atoms during the spin relaxation while k stays trapped .
furthermore , a k atom in a zeeman level higher than @xmath37 will need multiple collisions with li@xmath37 in order to fully polarize .
that explains why much more li@xmath37 is lost than k@xmath23 is gained during this process . and k@xmath44 @xcite .
also plotted is the lorentzian fit to the li@xmath37 loss feature from fig .
[ fig : spinrelaxbfield ] for comparison . ]
we interpret our data by comparing the position of the two spin relaxation features with the location of known interspecies feshbach resonances , since we expect enhanced inelastic loss close to a feshbach resonance @xcite . as fig .
[ fig : li2kswave ] shows , there is a series of @xmath45-wave feshbach resonances between li@xmath37 and k@xmath44 near the 207 g feature .
the distribution of feshbach resonances coincides with the width of the observed spin relaxation feature .
note that the experiment is performed at relatively high temperature causing considerable broadening of the feshbach resonances
. therefore individual resonances can not be resolved . for the feature of enhanced spin relaxation at 40 g
, there are no interspecies @xmath45- or @xmath46-wave feshbach resonances and thus we can not explain it by means of scattering resonances .
however , at low magnetic fields @xmath47 the release energy @xmath34 increases @xmath48 , which leads to a corresponding increase in the density of continuum states in the decay channel .
we speculate that the corresponding threshold behavior may explain the increase at lower fields .
then , already at a few ten gauss the nuclear spin of li decouples from the electron spin , which may lead to a reduction of loss . in a second set of experiments
we investigate the time scale on which spin relaxation occurs at the two relevant magnetic fields 40 g and 207 g . for both fields
we find that the time scale for the process is 150ms and a steady state is essentially reached after 500ms .
spin relaxation is a very efficient process that allows us to fully polarize our k sample without loss of k atoms . since initially much more li atoms are present in the trap , the li loss is a minor problem .
the resulting imbalance of the two li spin states can be removed by driving radio - frequency ( rf ) transitions between the two states .
a spin - mixture of li@xmath23 and li@xmath37 near the broad 834-g feshbach resonance facilitates highly efficient evaporative cooling , as it is well known in the field of strongly interacting fermi gases @xcite .
the efficiency of the cooling process is due to the very favorable combination of a large elastic scattering cross section with very low losses . in ref .
@xcite we have already demonstrated the possibility of using the @xmath0li spin - mixture as an efficient cooling agent to sympathetically cool another species . in this way we have demonstrated the attainment of a double - degenerate fermi - fermi mixture of @xmath0li and @xmath1k .
here we present additional information on the experimental procedures , and the combined evaporative and sympathetic cooling process .
let us first summarize our main findings of ref .
@xcite on the collisional stability of the three - component fermi gas of li in the lowest two spin states together with k in the lowest spin state .
interspecies collisional loss is negligible on the bcs side of the broad li resonance ( @xmath49 g ) and quite weak even exactly on resonance ( 834 g ) . substantial loss , however , occurs on the bec side of the resonance in a range between about 650 and 800 g .
the latter is a result of inelastic collisions between k atoms with weakly bound li dimers .
consequently , the combined evaporation and sympathetic cooling process needs to be performed on the bcs side of the li resonance . for our experiments
we choose a field of 1190 g . here
the li scattering length is @xmath50 and the interspecies scattering length is about @xmath51 . before starting the evaporation process , we carefully balance the population of the two spin states li@xmath23 and li@xmath37 .
this is particularly important in cases when the spin relaxation stage has caused considerable losses in @xmath37 .
the spin balance is accomplished by driving the rf transition @xmath52 using a series of 20 ramps over 10khz with a duration of 50ms each @xcite .
this procedure is performed at 1190 g , where the li spin mixture is outside of the strongly interacting regime and interaction - induced rf line shifts are relatively small .
note that the evaporation process is much more sensitive to a spin imbalance on the bcs side of the resonance than on the bec side of the resonance .
the reason is that in the latter case the molecule formation can lead to a self - balancing of the spin population during evaporation @xcite .
evaporation of the mixture is performed in the principal odt , beam 1 .
the evaporation ramp consists of two stages , technically implemented in different ways . in the first stage ,
we use a digital input of the laser control unit to reduce the odt power to about 15w .
this ramp is linear and takes 1.5s . in a second stage , an acousto - optical modulator ( aom )
is used to decrease the power in a nearly exponential ramp .
the evaporation ramp is typically stopped after 6s when the laser power is 60mw . at this point
the trap frequencies in the radial directions are 394hz for li and 219hz for k. in the axial direction the trap frequency is dominated by the magnetic confinement and is 27hz for li and 11hz for k. the experimental data in fig .
[ fig : evapbfield ] show the number of atoms remaining after the complete evaporation ramp for a magnetic field varied across the full tuning range offered by the li feshbach resonance .
the data correspond to the observations of ref .
@xcite , showing pronounced loss on the bec side of the li feshbach resonance and large stability on its bcs side . in the high - field region between 950 g and 1250 g , where the li scattering length varies between @xmath53 and @xmath54
, the magnetic field has no significant influence . in order to analyze the cooling process
, we stop the evaporation ramp at a variable endpoint and measure the number of li and k atoms .
the measurements are performed by recapture into the mot and subsequent detection of the fluorescence intensity or , at lower power , by absorption imaging after release from the odt into free space .
figure [ fig : evaporationpower ] shows that the li atom number steadily decreases while the k cooling proceeds essentially without losses .
note that the trap depth for k is a factor of 2.1 larger than for li .
this changes at about 100mw , as the gravitational sag of k reduces the trap depth and we begin to see significant loss of k when further lowering the power of the odt .
figure [ fig : temp ] shows the temperature evolution of li and k in the last part of the evaporation ramp .
we extract the li temperature by fitting a thomas - fermi profile to absorption images of the atomic cloud after time - of - flight .
the k temperature is determined using a simple gaussian fit , as the sample here remains in the non - degenerate regime . throughout the whole evaporation the temperature of k lags behind the temperature of li . at the end of an extended evaporation ramp , at a trap power of 40mw ,
the radial trap frequencies for li ( k ) amounts to 320hz ( 160hz ) . in the axial direction the trap frequency is dominated by the magnetic confinement and is 27hz for li and 11hz for k. the fermi temperatures are calculated to be @xmath55nk for li and @xmath56nk for k. following the scheme we have presented in ref .
@xcite , we continue cooling of the mixture by holding it in this shallow trap for 5s . this way we achieve a final k temperature of 50nk corresponding to a degeneracy of @xmath57 while li is deeply degenerate with @xmath58 .
since li is the coolant of our evaporation scheme , we adjust the amount of k with which evaporation starts such that at the end of evaporation we have about ten times more atoms of li in each spin state than atoms in k@xmath23 .
in this situation k can be used as a probe for the li mixture .
one example of this idea has already been presented in ref .
the measurement of the k temperature was used in order to get a firm upper bound for the temperature of the li bath even in the strongly interacting regime .
this method was recently adopted to using @xmath59li as a probe in ref .
the @xmath0li-@xmath1k mixture offers several @xmath45-wave feshbach resonances in the range between 150 g and 200 g @xcite .
all of them tend to be quite narrow , which is a common situation in cases of moderate values of the background scattering length @xcite .
the broadest resonances were found for the channels li@xmath23k@xmath60 with widths between 1 g and 2 g @xcite .
the energetically lowest channel li@xmath23k@xmath23 , which is of particular interest because of the energetic suppression of any two - body decay , features two resonances with calculated widths around 100 mg . in this work ,
we focus on the resonance near 168 g .
we show how a degenerate two - component li - k mixture can be prepared near this resonance after sympathetic cooling at high magnetic field and present measurements on inelastic and elastic properties of the mixture .
-wave ( @xmath46-wave ) interspecies resonances .
the li@xmath23 and li@xmath37 intraspecies @xmath46-wave resonances ( not shown ) are located at 160 g and 215 g respectively @xcite and nearly coincide with interspecies resonances . for the @xmath45-wave resonances ,
the relative widths are indicated by the size of the symbols .
state transfer , indicated by vertical dashed lines , is achieved by rf transitions .
after the evaporation at 1190 g , the li@xmath37 population is removed by a resonant laser pulse , as indicated by the * x*. ] when ramping down the magnetic field from its evaporation setting ( 1190 g ) to the interspecies resonances , one has to cross the region of the broad 834-g li@xmath23@xmath37 feshbach resonance .
if this spin channel is populated , the formation of @xmath0li dimers inevitably leads to strong losses from the atomic sample . to avoid this problem
, we remove one of the li spin components by the light pressure of a resonant laser pulse @xcite before starting the ramp .
note that already the momentum kick from one photon is sufficient to push a li atom out of the shallow trap after evaporation .
the pulse is applied for 10@xmath18s with a few times the saturation intensity .
we find that at 1190 g the interaction between the two spin components is weak enough to avoid any significant effect on the population in the remaining spin state . in this way , we reduce the three - component fermi - fermi gas to a two - component mixture . to approach a specific interspecies resonance ,
it is also important to avoid the effect of other inter- and intraspecies resonances .
we find that our ramps are fast enough ( ramp speed up to 20g / ms ) to cross all the @xmath46-wave resonances without any problem .
however , we find substantial losses on the interspecies @xmath45-wave resonances , even on the weaker ones .
this already points to efficient molecule association @xcite as we will discuss in sec .
[ sec : molecules ] . figure [ fig : ramp2fbr ] illustrates the procedures applied to reach specific interspecies feshbach resonances . while it is straightforward to reach the 168-g resonance in the li@xmath23k@xmath23 channel by a fast ramp after removal of the state li@xmath37 , other resonances require more elaborate methods . as an example , we discuss the 155-g resonance in the li@xmath23k@xmath61 channel , which is of interest as one of the broadest resonances ( width between 0.5 and 1 g ) in the low - lying spin channels . here a possible way is to transfer the li atoms from @xmath23 to @xmath37 after ramping down the field to @xmath62200 g , thus converting the sample into a li@xmath37k@xmath23 mixture .
this can be done with very high efficiency using rf transfer methods .
then the ramp is continued down to a value close to 155 g and three subsequent rf transfers are applied to convert the population from li@xmath37k@xmath23 to li@xmath23k@xmath61 .
this procedure avoids all detrimental resonances .
analogous schemes can be applied to reach any other desired resonance . in a set of experiments performed at the 168-g interspecies feshbach resonance in the li@xmath23k@xmath23 channel
, we investigate aspects of inelastic and elastic collisions .
initially , we prepare about @xmath63 li atoms together with about @xmath64 k atoms at a temperature of about 300nk
. the power of the trapping beam ( beam 1 ) is 170mw , corresponding to a radial ( axial ) trap frequency of 660hz ( 14hz ) for li . for
k the trap frequencies are 375hz and 6hz respectively .
the peak densities of the clouds are @xmath65@xmath66 and @xmath67@xmath66 and the degeneracies are @xmath68 and @xmath69 . note that these conditions are deliberately prepared with an incomplete evaporation ramp , stopped at 170mw instead of the usual final power of 60mw ( sec .
[ sec : evaporation ] ) . in the first series of measurements ,
we ramp the magnetic field to a variable value and study the loss of atoms after a hold time of 5s . for detection
, the remaining atoms are recaptured into the two - species mot and their fluorescence is recorded .
the k atom number , normalized to the background value away from resonance , is plotted in fig .
[ fig : likfbrevap ] ( open squares ) .
we observe a loss feature centered at 168.22 g .
ramping across the feshbach resonance does not lead to loss , indicating that the phase - space density used in these experiments is insufficient for adiabatic molecule creation during the magnetic field ramp . in the second series of measurements , we investigate whether an effect of enhanced elastic collisions can be observed in evaporative cooling near the interspecies resonance .
here we lower the power of beam 1 to 55mw within 3s , which results in a radial ( axial ) trap frequency of 375hz ( 14hz ) for li and 210hz ( 5hz ) for k. as before , the number of remaining atoms is determined by recapture into a mot and fluorescence detection .
the corresponding data , plotted in fig . [ fig : likfbrevap ] ( filled circles ) , show a pronounced asymmetry and thus a different behavior on the two sides of the feshbach resonance . on its high - field side ,
corresponding to large negative scattering length , we observe a maximum in the recaptured atom number at 168.26 g .
this signifies evaporative cooling with an enhanced elastic scattering cross section as compared to the background value . at lower fields
, however , loss enhanced by the resonance dominates and leads to a minimum in atom number at 168.18 g .
the loss properties on the two sides of the feshbach resonance are thus found to be strikingly different with more favorable conditions on the side of negative scattering length , where no weakly bound molecular state exists .
-width of the loss feature ( indicated by the dotted lines ) is determined by fitting a gaussian to the experimental data and amounts to 33 mg . ] to determine the resonance position more precisely , we prepare a sample with higher phase - space density than used for the previous sets of experiments . now both beams of the odt are used .
beam 2 is held at a constant power of 250mw , corresponding to a trap depth of 1@xmath18k for li and 2.1@xmath18k for k. its purpose is to add confinement along the weak direction of beam 1 at the very end of evaporation .
evaporation at 1190 g and the ramp to low magnetic field proceed as described above . for further cooling
, we create a balanced mixture of li@xmath23 and @xmath37 at 170.5 g , using a sequence of rf sweeps .
afterwards , beam 1 is ramped from a li trap depth of 1.9@xmath18k to 1.5@xmath18k during one second and the sample is left to thermalize for another second .
then li@xmath37 is removed by a short pulse of resonant light . at this point
the sample has a temperature of @xmath62200nk and contains about @xmath70 k atoms and @xmath71 li atoms . with the trap oscillation frequencies of axially 90hz ( 50hz ) , and radially 390hz
( 210hz ) for li ( k ) , we calculate fermi temperatures of about 900nk for li and 270nk for k , corresponding to @xmath72 and @xmath73 .
the k cloud has less than half the size of the li cloud . for both components the density in the center of the trap is about @xmath74@xmath66 . under these deep cooling conditions ,
we detect the fast atom loss as a function of the magnetic field . in order to approach the magnetic field value of interest without forming molecules ,
k is transferred into state @xmath37 by an rf @xmath75-pulse prior to the magnetic field ramp . at the final field
, k is transferred back to state @xmath23 by another @xmath75-pulse .
after a hold time of 10ms , the remaining k atom number is measured .
figure [ fig : fastloss ] shows the corresponding data .
we observe maximum loss of atoms centered at 168.217 g , with an estimated calibration uncertainty of 10 mg .
here , we describe our basic methodology for molecule creation and detection ( sec . [ sec : creationanddetectionschemes ] ) , present experimental results ( sec . [ sec : experimentalresults ] ) and discuss our findings ( sec .
[ sec : discussion ] ) .
, across a feshbach resonance .
transitions to higher atomic spin states are driven by rf pulses .
atoms bound in a molecule are not affected because of the binding energy @xmath76 . ]
the creation of the molecules starts with a li@xmath23k@xmath23 mixture under the same conditions as prepared for fig .
[ fig : fastloss ] .
the molecules are associated by a magnetic field ramp from 170.5 g to 168.19 g within 10ms , crossing the li@xmath23k@xmath23 168-g feshbach resonance ( ramp speed 0.23g / ms ) .
instantly after the ramp , the sample is released from the odt .
selective imaging of molecules and remaining unpaired atoms is possible after transfer of the unpaired atoms to the states li@xmath37 and k@xmath37 .
an rf @xmath75-pulse , tuned to the atomic @xmath23@xmath77@xmath37 transition , is used for this purpose @xcite ; see fig .
[ fig : imagerf ] . atoms bound in lik molecules are not transferred to state @xmath37 if the molecular binding energy detunes the transition far enough from the free atom transition to be outside of the fourier spectrum of the rf pulse .
this condition requires a detuning of 23khz for k , which is reached 9 mg below resonance according to the relative magnetic moment of the molecular state @xcite .
the rf pulses are applied one after the other during the 0.4ms free expansion of the sample .
state - selective absorption images are taken simultaneously on cycling transitions starting from the li@xmath23 and k@xmath23 states . this way , molecules are imaged directly .
the resulting pictures are shown in the left - hand column of fig .
[ fig : likmolecules ] @xcite . a second pair of images , this time of the unpaired atoms , which have been transferred to the @xmath37 states , is taken 1ms later and shown in the right - hand column of fig .
[ fig : likmolecules ] @xcite .
absorption imaging of the molecules gives lower boundaries @xmath78 and @xmath79 for the real molecule numbers @xmath80 and @xmath81 since the absorption cross section of atoms bound in lik molecules is somewhat smaller than the one of unpaired atoms .
close to the feshbach resonance the cross section is similar to the one of free atoms and decreases for increasing binding energy . the number of remaining unpaired atoms @xmath82 and @xmath83 can be obtained from the second pair of absorption images . from k images ( top row of fig . [
fig : likmolecules ] ) we obtain @xmath84 and @xmath85 and from li images ( bottom row of fig .
[ fig : likmolecules ] ) @xmath86 and @xmath87 .
the small cloud of k is immersed in a much larger degenerate li bath .
the molecule conversion efficiency is therefore best characterized by the k conversion efficiency .
a lower bound for the molecule fraction can be determined from k absorption images as @xmath88 ) . from the images shown in fig .
[ fig : likmolecules ] we obtain @xmath89 . in dependence of the final magnetic field value of the molecule association magnetic field ramp .
molecules are detected for fields below 168.218 g .
this field corresponds to the center of the loss feature shown in fig .
[ fig : fastloss ] , which is marked by the vertical solid line here .
the dashed vertical lines mark the @xmath90-width of the loss feature and the horizontal dashed line marks a systematic offset . ]
we now examine the molecule creation process and properties of the molecules in more detail .
first , we determine the magnetic field value of the onset of molecule creation . for this , we perform experiments as the one just described , but we vary the endpoint of the magnetic field ramp , keeping the ramp duration fixed .
the frequency of the rf pulse for the separation of free k atoms and lik molecules and the probe beam frequencies are adapted accordingly .
the li rf pulse was not used in these experiments .
figure [ fig : conveff ] shows the lower bound for the molecule fraction @xmath91 .
imperfect rf pulses lead to a 3% systematic offset in the data , indicated by the horizontal dashed line @xcite .
it is found that the detected molecule fraction depends strongly on the endpoint of the magnetic field ramp .
no molecules are detected down to a final field of 168.217 g .
only 13 mg lower the maximum molecule fraction is observed .
this magnetic field range corresponds well to the required detuning from resonance for our atom - molecule separation method to work , as discussed above . for lower fields the molecular signal drops again , first steeply down to about 168.19 g and then much slower . at about 167.5 g ( outside the range of the plotted data ) it becomes indiscernible from the background noise .
the dependence of the detected molecule fraction on the field may have several reasons .
it might be caused by the change in absorption cross section of the molecules with the magnetic field .
the slow decrease away from resonance comes from loss of molecules as more time is spent between molecule association and detection . in dependence on the hold time after the fast magnetic field ramp .
the solid line is an exponential fit to the data , yielding a lifetime of 1.7ms . ] within the measurement precision of a few mg the onset of molecule detection coincides with the center of the loss feature from fig . [ fig : fastloss ] , marked by the solid vertical line in fig .
[ fig : conveff ] .
this observation is in accordance with the standard picture of molecule formation close to a feshbach resonance @xcite .
the maximum k molecule conversion efficiency extracted from this data is reached at 168.204 g and amounts to about 40% . a different method to determine
the k molecule conversion efficiency is to examine the number of free k atoms at a magnetic field just above the onset of molecule production and just below 168.19 g . assuming all missing atoms have formed molecules , the molecule conversion efficiency is also determined to be 40% .
the assumption that no molecules are lost is well justified since the time spent in the feshbach resonance region during the magnetic field ramp to 168.19 g ( 120@xmath18s ) is short compared to the lifetime of the molecules .
the lifetime of the lik molecules is determined by holding the sample after molecule creation for a varying time in the odt at a constant magnetic field of 168.204 g and measuring the molecule number afterwards .
a fit to the decay of the molecule number gives a lifetime of 1.7ms ; see fig .
[ fig : likdecay ] . this lifetime does not change if the remaining free li atoms are removed just after molecule creation by a resonant flash of light , indicating that the dominant loss mechanism does not involve free li atoms .
we did not investigate the effect of unpaired k atoms on the molecule lifetime .
-width of gaussian fits to integrated density profiles . ] a striking manifestation of molecule formation can be observed by comparing the expansion behavior of clouds of lik molecules with the one of clouds of unpaired li atoms in imaging with li light . for this comparison
, we record the expansion of the molecules and the remaining unpaired li atoms after a molecule association magnetic field ramp to 168.196 g ; see fig .
[ fig : likmoltof ] .
we find the average expansion velocity of molecules to be slower by a factor of 3.3 , as determined by fits to the expansion .
we interpret this difference mainly as a result of the higher mass of the molecules compared to unpaired li atoms .
it corresponds well to the expected velocity ratio of @xmath92 in the approximation of thermal clouds of equal temperature .
this observation tells us that li atoms that remain in state @xmath23 after the @xmath75-pulse are bound to k atoms . in our experiment ,
molecule association is achieved in basically the same way as demonstrated in many other cold atom experiments before @xcite and our results agree well with the standard picture of molecule formation close to a feshbach resonance @xcite .
we observe that molecule association is most efficient in samples of high phase - space density and obtain a maximum molecular conversion efficiency for k of 40% .
this conversion efficiency is typical for experiments employing the feshbach ramp technique . a monte carlo simulation based on the method presented in ref .
@xcite agrees with our results , giving a conversion efficiency of about 50% for k.
the lifetime of our molecules is quite short , only 1.7ms .
because of this , it would be technically challenging to observe the standard signature of molecule association , which is the reduction of the absorption imaging signal when ramping to the molecular side of the feshbach resonance and the recovery of the signal after ramping back .
our rf state separation detection technique , which allows to obtain images of molecules less than 0.1ms after molecule association , overcomes this detection problem .
the li@xmath23k@xmath23 molecule lifetime that we measure is much shorter than typical lifetimes of li@xmath23k@xmath61 molecules that were measured by the munich group @xcite .
presently , we do not know whether the different spin channels and therefore the different feshbach resonances used for molecule association can explain the different lifetimes .
there are also other possible , more technical reasons , which we presently can not rule out .
one possibility , which needs further investigation , is the loss of molecules because of the absorption of photons from the broad spectrum of the multi - mode fiber laser used for the odt @xcite .
we have presented an all - optical evaporative and sympathetic cooling scheme for the preparation of a double degenerate @xmath0li-@xmath1k fermi - fermi mixture .
we have also shown the general methodology to prepare the sample close to specific interspecies feshbach resonances . as a first application
, we have demonstrated the formation of fermi - fermi heteronuclear molecules and we have examined the molecule association process and some properties of the molecules . with the basic tools at hand , we are now in the position for the next steps towards our main goal of realizing strongly interacting regimes in the fermi - fermi mixture .
since the available feshbach resonances are quite narrow this requires precise knowledge of the exact resonance position and the magnetic - field dependent elastic and inelastic interaction properties . we are currently inspecting the relatively broad 155-g resonance in the li|1>-k|3 > channel as a promising candidate , for which we experimentally find a width of about 800 mg @xcite .
strongly interacting conditions generally require a scattering length exceeding the interparticle spacing . under our typical experimental conditions
this would be realized with a magnetic detuning below 10 mg , which is experimentally feasible .
we want to thank paul julienne for fruitful discussions and support on the theoretical understanding of the interspecies scattering properties , tobias tiecke for the calculated feshbach resonance positions displayed in fig .
[ fig : li2kswave ] , christoph kohstall for useful comments on the manuscript , and clarice aiello for contributions to the experimental set - up .
this work is supported by the austrian science fund ( fwf ) and the european science foundation ( esf ) within the euroquam / fermix project and by the fwf through the sfb foqus .
three pairs of magnetic field coils are present in the setup : a pair of high - current , large - diameter coils , which we call feshbach coils , a smaller - diameter pair of high - current coils , which we call curvature coils , and a third , low - current , low inductance pair of coils , which we call fast coils . normally , the currents in all coils circulate in the same direction . to achieve a quadrupole field configuration for mot operation
, the direction of current in one coil of the feshbach coil pair and one coil of the curvature coil pair can be reversed using mechanical relays . in the normal configuration ,
the feshbach coils are in helmholtz configuration and give a very homogeneous bias field near the trap center of up to 3000 g . the curvature coils exhibit a magnetic field curvature , which gives rise to an additional contribution to the trapping potential @xcite . with the current used during evaporation , the curvature coils give a homogeneous bias field of 600 g and a magnetic field curvature of 27g / cm@xmath25 along the axial direction of the dipole trap beams ( perpendicular to the symmetry axis of the coils ) .
this curvature gives rise to a magnetic confinement corresponding to trap frequencies of 27hz for li and 10hz for k. when working at bias fields between 150 g and 170 g , where the interspecies feshbach resonances are , the curvature coils provide a magnetic confinement corresponding to 13hz for li and 5hz for k. when high magnetic field - stability is needed , we make use of a battery - powered current supply . since the interspecies feshbach resonances are very narrow , it is necessary to control the magnetic field with very high precision . passive stabilization methods ,
not employing any shielding , lead to a stability of about 10 mg peak - to - peak over a 50hz cycle . by synchronizing the experimental sequence to line , we achieve a magnetic field stability of a few mg for times on the order of one ms , which is much larger than the typical duration of rf @xmath75-pulses we use for internal state transfer .
magnetic field values are calibrated using rf transitions . for probing the interspecies resonances we make use of the fast coils , which have helmholtz configuration .
using these coils we make precise magnetic field ramps of up to 3 g in about 0.1ms .
this response time of the magnetic field was characterized by measuring the change in frequency of an atomic rf transition with time after a step change of the current . the response time is not limited by the speed of change of the current through the coil , but by eddy currents . | we present the essential experimental steps of our all - optical approach to prepare a double - degenerate fermi - fermi mixture of @xmath0li and @xmath1k atoms , which then serves as a starting point for molecule formation .
we first describe the optimized trap loading procedures , the internal - state preparation of the sample , and the combined evaporative and sympathetic cooling process .
we then discuss the preparation of the sample near an interspecies feshbach resonance , and we demonstrate the formation of heteronuclear molecules by a magnetic field ramp across the resonance . |
celestial bodies and dust particles migrate to near - earth space from different regions of the solar system ( from the main asteroid and edgeworth - kuiper belts , the oort and hills clouds , etc . ) . some scientists ( farinella et al .
1993 ; bottke et al .
2002 ) considered that most near - earth objects ( neos ) are asteroidal fragments , others ( wetherill 1988 ) supposed that half of neos are former short - period comets .
we studied migration of small bodies and dust particles based on extensive integrations , considering more jupiter - crossing objects than before .
the interval of considered diameters of dust particles was wider than in previous publications .
our reviews of papers on these problems can be found in ( ipatov 2001 , ipatov & mather 2004a - b , ipatov et al .
as migration of trans - neptunian objects ( tnos ) to jupiter s orbit was investigated by several authors ( e.g. , levison & duncan , 1997 ) , ipatov & mather ( 2003 , 2004a - b ) integrated the orbital evolution of 30,000 jupiter - family comets ( jfcs ) under the gravitational influence of planets , and based on the results of these integrations they investigated the migration of tnos to the near - earth space . also the orbital evolution of 1300 resonant main - belt asteroids was studied .
we omitted the influence of mercury ( except for comet 2p / encke ) and pluto and used the bulirsch - stoer and symplectic methods ( bulsto and rmvs3 codes ) from the integration package of levison and duncan ( 1994 ) . in the first series of runs ( denoted as @xmath6 ) , we calculated the evolution of 3100 jupiter - crossing objects ( jcos ) moving in initial orbits close to those of 20 real comets with period 5@xmath1@xmath7@xmath19 yr , and in the second series of runs ( denoted as @xmath8 ) we considered 13,500 jcos moving in initial orbits close to those of 10 real comets with period 5@xmath1@xmath7@xmath115 yr . in other series of runs , initial orbits were close to those of a single comet ( 2p , 9p , 10p , 22p , 28p , 39p , or 44p ) .
we investigated the orbital evolution during the dynamical lifetimes of objects ( at least until all the objects reached perihelion distance @xmath9@xmath106 au ) . in our runs
, planets were considered as material points , so literal collisions did not occur . however , based on the orbital elements sampled with a 500 yr step , we calculated the mean probability @xmath11 of collisions during lifetime of a jco .
the results can differ considerably depending on the initial orbits of comets .
the values of @xmath11 for the earth were about ( 1 - 4)@xmath12@xmath13 for comets 9p , 22p , 28p , and 39p .
they were ( 6 - 10)@xmath12@xmath13 for comet 10p and exceeded @xmath14 for comet 2p .
the ratio of the mean probability of a jco with semi - major axis @xmath0@xmath101 au with a planet to the mass of the planet was greater for mars by a factor of several than that for earth and venus . using @xmath11=@xmath15 ( this value is smaller than the mean value of @xmath11 obtained in our runs ) and assuming that the total mass of planetesimals that ever crossed jupiter s orbit is about 100@xmath16 ( ipatov 1987 , 1993 ) , where @xmath16 is the mass of the earth , we obtained that the total mass of water delivered from the feeding zone of the giant planets to the earth could be about the mass of water in earth s oceans .
ancient venus and mars could have large oceans . for series @xmath6 with rmvs3
, the probability of a collision with the earth for one object with initial orbit close to that of comet 44p was 88.3% of the total probability for 1200 objects from this series , and the total probability for 1198 objects was only 4% .
the probabilities of collisions of one object with earth and venus could be greater than for thousands other objects combined .
a few jfcs got earth - crossing orbits with aphelion distance @[email protected] au ( and sometimes with @xmath0@xmath12 au ) and moved in such orbits for more than 1 myr ( up to tens or even hundreds of myrs ) .
the probability of a collision of one of such objects , which move for millions of years inside jupiter s orbit , with a terrestrial planet can be greater than analogous total probability for thousands other objects . with bulsto , six and nine objects , respectively from 10p and 2p series , moved into apollo orbits with @xmath0@xmath12 au for at least 0.5 myr each , and five of them remained in such orbits for more than 5 myr each .
only two objects in series @xmath8 got apollo orbits with @xmath0@xmath12 au during more than 1 myr .
four considered former jfcs even got inner - earth object ( ieo ) orbits ( with @[email protected] au ) or aten orbits for myrs .
note that ipatov ( 1995 ) obtained migration of jcos into ieos using the method of spheres to consider the gravitational influence of planets .
now let s return to our recent results of integration .
one former jco got aten orbits ( @xmath0@xmath11 au , @[email protected] au ) during @xmath103 myr , but a probability of its collision with the earth was greater than that for @xmath17 other considered former jcos .
this object also moved during about 10 myr ( before its collision with venus ) in ieo orbits , and during this time interval a probability of its collision with venus was even greater than the total probability of collisions of @xmath17 considered jcos with earth .
after 40 myr one considered object with initial orbit close to that of comet 88p got @[email protected] au , and it moved in orbits with @xmath0=2.60 - 2.61 au , 1.7@xmath1@[email protected] au , 3.1@xmath1@[email protected] au , eccentricity @xmath18=0.2 - 0.3 , and inclination @xmath19=5 - 10@xmath20 for 650 myr .
another object ( with initial orbit close to that of comet 94p ) moved in orbits with @xmath0=1.95 - 2.1 au , @[email protected] au , @[email protected] au , @xmath18=0.2 - 0.3 , and @xmath19=9 - 33@xmath20 for 8 myr
( and it had @xmath2@xmath13 au for 100 myr ) .
so jfcs can rarely get typical asteroid orbits and move in them for myrs . in our opinion , it can be possible that comet 133p ( elst - pizarro ) moving in a typical asteroidal orbit was earlier a jfc and it circulated its orbit also due to non - gravitational forces .
jfcs got typical asteroidal orbits less often than neo orbits .
results of the orbital evolution of jcos show that many earth - crossing objects can move in highly eccentrical ( @[email protected] ) orbits and , probably , most of 1-km objects in such orbits have not yet been discovered .
the obtained results show that during the accumulation of the giant planets the total mass of icy bodies delivered to the earth could be about the mass of water in earth s oceans .
the results obtained by the bulirsh - stoer method ( bulsto code ) with an integration step error @xmath21 @xmath410@xmath22 - 10@xmath23 , with 10@xmath24 , and by a symplectic method ( rmvs3 code ) with an integration step @xmath25@xmath2610 days were usually similar , and the difference at these three series was about the difference at small variation of @xmath21 or @xmath25 . in the case of close encounters with the sun ,
the values of collision probability @xmath27 with the sun obtained by bulsto and rmvs3 and at different @xmath21 or @xmath25 could be different , but all other results were usually similar . for comet 2p ,
comet 96p , and the 3:1 resonance with jupiter , close encounters with the sun took place , and with a symplectic method we got greater mean probabilities of collisions with the sun than with bulsto .
the fraction of asteroids migrated from the 3:1 resonance with jupiter that collided with the earth was greater by a factor of several than that for the 5:2 resonance .
using our results of the orbital evolution of jcos and the results of migration of tnos obtained by duncan et al .
( 1995 ) and considering the total of @xmath28 1-km tnos with 30@xmath1@xmath0@xmath150 au , ipatov & mather ( 2003 , 2004a - b ) estimated the number of 1-km former tnos in neo orbits .
results of our runs testify in favor of at least one of these conclusions : 1 ) the portion of 1-km former tnos among neos can exceed several tens of percents , 2 ) the number of tnos migrating inside the solar system could be smaller by a factor of several than it was earlier considered , 3 ) most of 1-km former tnos that had got neo orbits disintegrated into mini - comets and dust during a smaller part of their dynamical lifetimes if these lifetimes are not small .
our runs showed that if one observes former jfcs in neo orbits , then most of them could have already moved in such orbits for millions ( or at least hundreds of thousands ) of years ( if they did nt disintegrate ) .
some former comets that have moved in typical neo orbits for millions or even hundreds of millions of years , and might have had multiple close encounters with the sun , could have lost their typically dark surface material , thus brightening their low albedo and assuming the aspect typical of an asteroid ( for most observed neos , the albedo is greater than that for comets ) .
the number of extinct comets can exceed the number of active comets by several orders of magnitude if most of comets do not disintegrate into mini - comets and dust during their dynamical lifetimes .
using the bulirsh - stoer method of integration , ipatov et al .
( 2004a ) integrated the orbital evolution of 7000 asteroidal , trans - neptunian , and cometary dust particles , under the gravitational influence of planets , the poynting - robertson drag , radiation pressure , and solar wind drag .
the values of the ratio @xmath29 between the radiation pressure force and the gravitational force were varied from 0.0004 to 0.4 . for silicates ,
such values correspond to particle diameters @xmath3 between 1000 and 1 microns ; for water ice , the diameters are greater by a factor of 3 than those for silicates . after the above publication
, we have made more calculations for small values of @xmath29 and more analysis of the runs .
considered asteroidal and trans - neptunian ( kuiperoidal ) dust particles started with zero relative velocity from the numbered main - belt asteroids and known trans - neptunian objects , respectively .
we also studied particles started from comet encke .
several hundred of particles were usually considered at each @xmath29 .
the simulations continued until all of the particles either collided with the sun or reached 2000 au from the sun . in our runs , planets were considered as material points , but using orbital elements obtained with a step @xmath30 of @xmath2620 yr , we calculated the mean probability @xmath11 of a collision of a particle with a planet during the lifetime of the particle .
we also made similar runs without planets to investigate the role of planets in interplanetary dust migration .
the mean probabilities @xmath11 of collisions of asteroidal and cometary dust particles with earth and venus during lifetimes of particles were maximum at @[email protected] - 0.004 ( i.e. , at @xmath3@xmath4100 - 200 microns for silicate particles ) . at @[email protected]
the values of @xmath11 and the mean times @xmath32 spent by particles in orbits with @xmath9@xmath11 au quickly decrease with an increase of @xmath29 ( usually @xmath11 and @xmath32 are proportional to 1/@xmath29 ) . for asteroidal dust particles at @[email protected] - 0.001 the values of @xmath11 and @xmath32 were smaller than those at @[email protected] - 0.004 , though maximum times elapsed until collisions of particles with the sun were greater for smaller @xmath29 ( greater times were needed for larger particles to migrate to the orbits of the terrestrial planets ) .
the probability of a collision of a migrating dust particle with the earth for @[email protected] is greater by a factor @xmath10200 than that for @xmath29=0.4 .
cratering records in lunar material and on the panels of the long duration exposure facility showed that the mass distribution of dust particles encountering earth peaks at @xmath3=200 @xmath5 m .
the probability of collisions of cometary particles with the earth is smaller than that for asteroidal particles at the same @xmath29 , and this difference is greater for larger particles .
almost all asteroidal particles with @xmath3@xmath264 @xmath5 m collided with the sun . at @[email protected] and @[email protected]
some asteroidal particles migrated beyond jupiter s orbit .
a few asteroidal particles can collide with the sun after moving outside jupiter s orbit for a long time . at @[email protected] (
i.e. , at @xmath3@xmath110 @xmath5 m ) the probability of a collision of a trans - neptunian particle with the earth and the mean time @xmath32 spent in orbits with @xmath9@xmath11 au were usually less than those for a typical asteroidal particle by less than an order of magnitude ( only at @xmath29=0.05 values of @xmath32 differed by a factor of 20 ) .
the total mass of the edgeworth - kuiper belt is at least two orders of magnitude greater than that of the main asteroid belt .
so the fraction of trans - neptunian particles with @xmath3@xmath110 @xmath5 m ( runs with greater @xmath3 have not yet been finished ) among all particles near the orbit of the earth can be considerable . at @[email protected] - 0.4 ,
the fraction @xmath27 of trans - neptunian particles collided with the sun was 0.1 - 0.2 , i.e. , it was smaller by a factor of 5 than that of asteroidal particles for the same @xmath29 .
several plots of the distribution of migrating asteroidal particles in their orbital elements and the distribution of particles with their distance @xmath33 from the sun and their height @xmath34 above the initial plane of the earth s orbit were presented by ipatov et al .
( 2004a ) . for @xmath29=0.01 the local maxima of the mean time @xmath35 ( the total time divided by the number @xmath36 of particles ) during which an asteroidal dust particle had a semi - major axis @xmath0 in an interval of fixed width corresponding to the 6:7 , 5:6 , 3:4 , and 2:3 resonances with the earth are greater than the maximum at 2.4 au .
there are several other local maxima corresponding to the n:(n+1 ) resonances with earth and venus ( e.g. , the 7:8 and 4:5 resonances with venus ) .
the trapping of dust particles in the n:(n+1 ) resonances causes earth s dust ring .
the greater the @xmath29 , the smaller the local maxima corresponding to these resonances . at @[email protected]
there are gaps with @xmath0 a little smaller than the semi - major axes of venus and earth that correspond to the 1:1 resonance for each ; the greater the @xmath29 , the smaller the corresponding values of @xmath0 .
a small gap for mars is seen only at @[email protected] .
there are also gaps corresponding to the 3:1 , 5:2 , and 2:1 resonances with jupiter .
spatial density @xmath37 of considered trans - neptunian particles near ecliptic at distance from the sun @xmath33=1 au was greater than at @xmath33@xmath101 au . at 0.1@xmath26@[email protected] and 2@xmath1@xmath33@xmath145
au ( at @xmath29=0.05 for @xmath38@xmath1@xmath33@xmath150 au ) , @xmath37 varied with @xmath33 by less than a factor of 4 . for asteroidal and cometary particles , @xmath37 quickly decrease with an increase of @xmath33 , e.g. , for @xmath29=0.2 at @xmath33=5 au @xmath37 was smaller than at @xmath33=1 au by a factor of 70 and 50 for asteroidal and cometary particles , respectively .
in contrast to the asteroidal dust particles , for cometary dust particles the values of a mean time in a planet - crossing orbit did not differ much between venus , earth , and mars .
collision probabilities with earth were greater by a factor of 10 - 20 than those with mars and greater for particles starting at perihelion than aphelion .
for the same values of @xmath29 , the probability of cometary dust particles colliding with a terrestrial planet was several times smaller than for asteroidal dust particles , mainly due to the greater eccentricities and inclinations of the cometary particles . this difference is greater for larger particles . in almost all cases , inclinations
@xmath19@xmath1@xmath39 , and at @xmath0@xmath1010 au the maximum @xmath19 was less for larger asteroidal particles . based on the obtained distributions of migrating dust particles , ipatov et al .
( 2004b ) investigated how the solar spectrum is changed by scattering by the zodiacal cloud grains .
as positions of particles were stored with a step of 10 yr , for each such stored position we calculated many ( @xmath1010@xmath40 ) different positions of a particle and the earth considering that orbital elements do not vary during a period of revolution of the particle around the sun .
one of the obtained results is that the results of modeling are relatively insensitive to the scattering function considered . in some models , the scattering function depended on a scattering angle @xmath41 in such a way : 1/@xmath41 for @xmath41@xmath1c , 1+(@xmath41-c)@xmath42 for @xmath41@xmath10c , where c=2@xmath43/3 radian .
in other runs , we added the same dependence on elongation @xmath44 ( considered westward from the sun ) . in the third
runs , the scattering function did nt depend on these angles at all .
the plots of the obtained spectrum are in general agreement with the observations made by reynolds et al .
they measured the profile of the scattered solar mg i@xmath455184 absorption line in the zodiacal light .
unlike results by clarke et al .
( 1996 ) , our modeled spectra do not exhibit strong asymmetry .
as these authors , we obtained that minima in the plots of dependencies of the intensity of light on its wavelength near 5184 angstrom are not so deep as those for the initial solar spectrum .
the details of plots depend on diameters , inclinations , and a source of particles .
for example , at inclination @xmath19=0 and @xmath44=90@xmath20 , for kuiperoidal and cometary particles the shift of the plot to the blue and its minimum were greater than those for asteroidal particles .
our preliminary models and comparison with observational data indicate that for more precise observations it will be possible to distinguish well the sources of the dust and impose constrains on the particle size . for asteroidal and cometary particles at @xmath29=0.2 ,
about 65 - 80% of brightness is due to particles at distance from the earth @xmath46@xmath11 au ( 90 - 95% for @[email protected] au ) , and at most values of @xmath44 ( not close to 180@xmath20 ) 75 - 90% of brightness is from particles with @[email protected] au .
for trans - neptunian particles , these fractions were smaller . at @xmath29=0.2 for @xmath44=270@xmath20
about half of brightness was from particles with @xmath33@xmath104 au and @xmath46@xmath104 au ; for @xmath44=90@xmath20 30 - 40 % ( depending on the considered scattering function ) of brightness was from particles with such values of @xmath33 and @xmath46 .
velocities of dust particles relative to the earth that mainly contributed to brightness were different for different @xmath44 . at @xmath19=0
they were between -25 and 25 km / s .
our results show that some former jfcs got earth - crossing orbits with aphelion distance @[email protected] au and even inner - earth object orbits ( with @[email protected] au ) , aten orbits , or typical asteroidal orbits for myrs
. the probability of a collision of one former jfc moving inside jupiter s orbit for millions of years with a terrestrial planet can be greater than analogous total probability for thousands other jfcs .
results obtained by the bulirsch - stoer method were mainly similar ( except for probabilities of close encounters with the sun when they were high ) to those obtained by a symplectic method .
the total mass of icy bodies delivered to the earth during the accumulation of the giant planets could be about the mass of water in earth s oceans .
we concluded that the trans - neptunian belt can provide a significant portion of near - earth objects , or the number of trans - neptunian objects migrating inside solar system could be smaller than it was earlier considered , or most of 1-km former trans - neptunian objects that had got near - earth object orbits disintegrated into mini - comets and dust during a smaller part of their dynamical lifetimes if these lifetimes are not small .
the probability of a collision of an asteroidal or cometary particle with the earth was maximum at diameter @xmath3@xmath4100 @xmath5 m . at @xmath3@xmath110 @xmath5 m
such probability for a trans - neptunian particle was less than that for an asteroidal particle by less than an order of magnitude , and the fraction of trans - neptunian particles with such diameter near earth can be considerable .
the peaks in the distribution of migrating asteroidal dust particles with semi - major axis corresponding to the n:(n+1 ) resonances with earth and venus and the gaps associated with the 1:1 resonances with these planets are more pronounced for larger particles . 2001 , _
advances in space research _ 28 , 1107 2003 , _ earth , moon , and planets _ 92 , 89 in : e. belbruno , d. folta , & p. gurfil ( ed . ) 2004a , _ astrodynamics , space missions , and chaos _ , annals of the new york academy of sciences , vol . 1017 , 46 ( http://arxiv.org/format/astro-ph/0308448 ) . 2004b , _ advances in space research _ 33 , 1524 in : e. belbruno , d. folta , & p. gurfil ( ed .
) 2004a , _ astrodynamics , space missions , and chaos _ , annals of the new york academy of sciences , vol .
1017 , 66 ( http://arxiv.org/format/astro-ph/0308450 ) . | we integrated the orbital evolution of 30,000 jupiter - family comets , 1300 resonant asteroids , and 7000 asteroidal , trans - neptunian , and cometary dust particles . for initial orbital elements of bodies close to those of comets 2p , 10p , 44p , and 113p , a few objects got earth - crossing orbits with semi - major axes @xmath0@xmath12 au and moved in such orbits for more than 1 myr ( up to tens or even hundreds of myrs ) .
four objects ( from 2p and 10p runs ) even got inner - earth orbits ( with aphelion distance @[email protected] au ) and aten orbits for myrs .
our results show that the trans - neptunian belt can provide a significant portion of near - earth objects , or the number of trans - neptunian objects migrating inside the solar system can be smaller than it was earlier considered , or most of 1-km former trans - neptunian objects that had got near - earth object orbits for millions of years disintegrated into mini - comets and dust during a smaller part of their dynamical lifetimes .
the probability of a collision of an asteroidal or cometary particle during its lifetime with the earth was maximum at diameter @xmath3@xmath4100 @xmath5 m . at @xmath3@xmath110 @xmath5 m
such probability for trans - neptunian particles was less than that for asteroidal particles by less than an order of magnitude , so the fraction of trans - neptunian particles with such diameter near earth can be considerable . |
the idea of _ internet - of - things _ ( iot ) originated from an internet - connected coke machine at carnegie mellon university in the 1980 s and took over in the late 1990 s with the vision of networking everyday objects so as to automate our societies and daily lives . in the past decades
, the iot has seen technological innovations in a wide range of applications such as smart city , smart home , and autonomous robots , vehicles and _ unmanned aerial vehicles _ ( uavs ) .
the iot is expected to comprise tens of billions of sensors in the near future among other types of nodes .
the big - data collected via the sensors over iot is centralized and analyzed in the cloud to measure and regulate phenomena of common interest in the real world such as pollution , traffic congestion , and parking availability . empowered by cloud computing
, the iot technologies can penetrate many sectors of our society ranging from traffic control to business and foster breakthroughs in diversified fields ranging from physics to biology .
however , keeping the massive number of energy - constrained iot sensors alive poses a key design challenge for iot .
this is especially challenging given a large popular of the sensors are hidden ( e.g. , in the walls or appliances ) or deployed in remote or hazardous environments ( e.g. , in radioactive areas or pressurized pipes ) , making battery recharging or replacement difficult if not impossible .
thus , it is highly desirable to power iot nodes by ambient energy harvesting @xcite or wireless _ power transfer _ ( pt ) @xcite .
one particular promising solution is _ backscatter communications _ ( backcom ) that allows an iot node to transmit data by reflecting and modulating an incident rf wave @xcite .
the conventional radio architecture comprises power - hungry rf chains having oscillators , mixers and digital - to - analog converters , which results in non - compact form factors and limiting the battery lives of iot devices .
in contrast , a backscatter node has no active rf components and as a result can be made to have miniature hardware with extremely low power consumption , facilitating large - scale deployment at flexible location or even in - body implantation . in the past two decades
, point - to - point backcom has been widely deployed in the application of _ radio - frequency identification _ ( rfid ) for a passive rfid tag to report an i d to a enquiring reader over the near field ( typically several centimeters ) . in its early stage ,
iot comprised of primarily rfid devices for logistics and inventory management . however , iot is expected to connect tens of billions of devices and accomplish much more sophisticated and versatile tasks with city - wise or even global - scale influences .
this demands the communication capabilities and ranges ( tens of meters ) between iot nodes to be way beyond the primitive rfid operations supporting bursty and low - rate ( several - bytes pre - written i d sequence ) uni - directional transmission over several meters .
this can be achieved via a full - fledged backcom theory leveraging the existing well - developed communication technologies such as small - cell networks , full - duplexing , multi - antenna communications and wireless pt , as well as advancements in electronics such as miniature radios ( e.g. , button - size radios ) and low - power electronics .
furthermore , instead of supporting a single rfid link , a backcom networking theory has to be developed for allowing multiple access by a massive number of simultaneous iot nodes . from the computing perspective , it is desirable for iot sensors to pre - process sensing data to reduce its redundancy and latency while rfid tags have no such capabilities . despite the constraint on their form factors
, iot sensors can acquire powerful computing capabilities leveraging the advancements in miniature computers e.g. , computer on a stick .
these iot design challenges present many promising research opportunities , resulting in a recent surge in research interests in backcom .
this article aims at introducing the active area of backcom .
first , we introduce the basic principles , system and network architectures for backcom .
second , we discuss how specific communication techniques have been redesigned for backcom .
the last part focuses on the recent applications of backcom for iot .
a basic backcom system consists of two devices : a mobile backscatter node , i.e. , a tag , and a reader @xcite .
the tag is a passive node that harvests energy from an incident single - tone sinusoidal _ continuous wave _ ( cw ) radiated by the reader , and also modulates and reflects a fraction of the wave back to the reader .
specifically , the wave reflection is due to an intentional mismatch between the antenna and load impedance .
varying the load impedance makes the reflection coefficient to vary following a random sequence that modulates the reflected wave with tag s information bits .
such modulation scheme is named as the backscatter modulation .
therefore , the passive tag is powered by rf energy harvesting and does not require any active rf component . on the contrary
, the reader has its own power supply and a full set of conventional rf components for emitting cw and information transmission / reception .
the mobile architecture of the tag consists of an rf energy harvesting block , a battery , a modulation block and an information decoder , as illustrated in fig .
[ fig : architecture ] .
the incident cw is connected to the rf energy harvesting block and either the modulation block or the information decoder depending on the transmission or reception mode of the tag .
the rf energy harvesting block converts the cw into a direct current ( dc ) signal to charge the battery . powered by the battery , the key functional blocks , i.e.
, the modulation block and the information decoder , are able to perform backscatter modulation - based information transmission ( it ) and energy detection - based information detection , respectively . in general , the communication between the reader and the tag has two modes : the forward it , i.e. , the reader - to - tag transmission , and the backward it , i.e. , the tag - to - reader transmission . for the forward it , as illustrated in fig .
[ fig : forwardit](a ) , the reader transmits a binary intensity modulated signal to the tag .
the tag connects its information decoder and utilizes the received rf signal for rf energy harvesting and energy - detection based demodulation .
for example , the decoded bit is ` @xmath0 ' or ` @xmath1 ' when high or low signal energy is detected , respectively .
the use of this primitive on / off modulation and energy detection is due to constraint that a typical tag is provisioned with an energy detector instead of a power hungry rf chain needed for coherent demodulation . for the backward it , the reader sends a cw signal to the tag , and the tag connects its modulation block and utilizes the received rf signal for rf energy harvesting and backscatter modulation . generally , a tag can modulate the reflected signal by switching over a given set of impedances , generating a set of reflection coefficients forming a constellation .
for example , assuming _ binary phase - shift keying _
( bpsk ) modulation , as illustrated in fig .
[ fig : forwardit](b ) , the tag has a pair of impedances corresponding to two symbols , and it chooses either one of them for backscattering depending on the value of the transmitted bit . for backward - it modulation , one unique design issue is the consideration of energy - harvesting efficiency . specifically , it is desirable to design a modulation scheme for backcom that reduces the reflected energy and thereby increases the harvested energy . in general , the objective can be achieved by shifting the constellation points on the complex plane towards the original @xcite .
this , however , may increase the detection error - rate , introducing an energy - rate tradeoff .
the backward it is the dominant mode for most of the conventional rfid applications , which have asymmetric data traffics , e.g. , a low - rate command through the forward it and a high - rate information - bearing data through the backward it .
however , the forward and backward it are equally important for future iot applications , which require peer - to - peer network architecture and symmetric communication links between the massive number of devices .
besides the said tradeoff arising from modulation design , there exists another one due to bursty transmission by iot devices . for instance , a sensor for crowd - sensing reports data only upon receiving a query and spends the remaining time on other activities e.g. , sensing and computing . leveraging this characteristic , a backscatter tag can be designed to periodically switch between two modes , namely the _
silent _ ( or energy harvesting ) and _ active _ modes .
then the _ duty cycle _ is defined as the percentage of time for the active mode . in the silent mode ,
the energy of the incident wave is fully harvested without reflection , by matching the variable impedance to that of the antenna , and the circuit may be turned off for energy conservation . in the active mode ,
the tag circuit is activated to receive or transmit data by backscattering where only the fraction of harvested energy is much smaller ( see fig .
[ fig : architecture ] ) .
consequently , the duty - cycle is a key design parameter for regulating an energy - rate tradeoff .
the conventional backcom system discussed in sec . [ sec : basic ] is the simplest backcom system , as illustrated in fig . [
fig : networks](a ) .
the backcom systems for iot are much more complex and can be divided into the following categories .
many real - life applications like these can be modeled as a _ multiple - access _ ( mac ) backcom system where a single reader serves multiple tags , as illustrated in fig .
[ fig : networks](b ) . in a warehouse ,
an administrator can use a single reader to collect information simultaneously from hundreds to thousands of items equipped with rfid tags . or in a smart city , a data aggregator can receive sensing data from a large number of backscatter sensors at the same time
. the efficiency of the mac backcom system can be improved by avoiding collisions between multi - tag transmissions using the traditional mac schemes including _ space / code / time - division multiple - access _ ( sdma / cdma / tdma ) .
the other scheme , _ orthogonal frequency - division multiple - access _ ( ofdma ) , may not be suitable as the required fft operation is too complex for backscatter nodes . due to its simplicity ,
tdma is perhaps the most practical scheme for mac backcom systems where tags transmits in separate pre - assigned time slots in each frame . the inherent close - loop signaling for backcom facilitates the needed reader - tag synchronization for tdma .
researchers have also attempted to design new mac schemes exploiting the characteristics of backcom . for instance , an interesting new method for collision avoidance was proposed in @xcite for mac backcom that treats bursty backscatter transmissions by tags as a sparse code and decodes multi - tag data at the reader using a compressive - sensing algorithm . to avoid unnecessarily overloading the core network and reduce latency , distributed _ device - to - device _ ( d2d ) communications is common in iot , creating backcom interference channels . compared with conventional interference channels ,
a backscatter node reflects all incident interference signals , resulting in _ interference regeneration_. the phenomenon is illustrated in fig .
[ fig : networks](c ) showing a two - link backcom interference channel where readers 1 and 2 each transmit a cw to backscatter tags 1 and 2 , respectively . in total , each reader , reader 1 , is exposed to two interference components due to tag 2 s modulation and reflection of cws from readers 1 and 2 . in theory
, the number of interference components received by each reader can increase as the square of the number of coexisting links instead of linearly with the number in the conventional case . as a result ,
the interference issue is particular severe in backcom interference networks due to interference regeneration .
one effective way for coping with this issue is to adopt _ spread spectrum _
techniques in backcom exploiting its low data rates @xcite .
ultra - dense iot nodes are expected to be deployed in the environment in the near future and communicate by d2d transmissions .
most transmitting nodes are energy constrained and can not act as readers to emit high - power cw and power their receivers nor to enable d2d transmissions . for iot , there are two practical solutions for the power - source problem : one solution is to deploy dense low - complexity and low - cost _ power beacons _ ( pbs ) , stations dedicated for microwave power transfer ( mpt ) @xcite , to enable d2d backcom links in their proximity by beaming towards them strong cws , as illustrated in fig . [
fig : networks](d ) . then the tag is able to perform rf energy harvesting and backcom to the reader .
such backcom system is usually named as _
wirelessly powered backcom_. the other solution is to harvest the energy from ambient rf signals , such as the signals from cellular , tv broadcasting and wifi networks .
leveraging the ambient rf signals can allow direct tag - to - reader ( or d2d ) communication without readers supplying power , which motivated various relevant designs recently @xcite .
for instance , a backscatter tag can transmit data to a peer by reflecting an incident base - station signal , as illustrated in fig . [
fig : networks](e ) .
such an energy harvesting backcom system differs from one with a reader or pb in two aspects .
first , the cw is replaced with a modulated signal and thus the reflected is _
double modulated _ with superimposed unintended and intended data .
this problem can be tackled by exploiting the asymmetry in the high - rate ambient and low - rate backscatter signals . as result
, the latter can be detected after suppressing the former by time averaging over each symbol duration .
the second issue is the incident ambient signal is much weaker than the cw from an intended reader or pb due to a much long propagation distance .
consequently , backcom with ambient rf signal is suitable only for either short - distance or infrequent d2d communications . to communicate with passive tags embedded with different types of available commercialized devices , backcom systems with
technology conversion come into being .
one technology conversion is to leverage a bluetooth signal for backcom between a tag and a wifi device @xcite , as illustrated in fig .
[ fig : networks](f ) . consider a wearable - device network , which consists of an implanted backcom sensor ( i.e. , a tag ) , a bluetooth watch , and a smart phone , which is a wifi device .
the bluetooth watch adopts a _ gaussian frequency shift keying _ ( gfsk ) method that encodes bits using two frequency tones .
hence , the watch emits a cw at either of these frequency tones . by leveraging such cw - like bluetooth signal ,
the backcom tag would like to get its sensed e - health information collected by the smart phone through backscatter modulation .
although the bluetooth cw frequency is usually different from wifi , by adopting a fsk modulation through properly switching between different impedances , the backcom tag is able to shift the bluetooth cw to the wifi channel , hence enables the backcom with the smart phone .
next , consider a backcom system for smart home applications , which consists of a wifi _ access point _ ( ap ) and a backscatter iot sensor .
the ap transmits packets to the wifi clients , such as laptops and smart phones , which is also received by the backscatter iot sensor .
then the sensor is able to modulate its data on the unintended signal and backscatters the signal to the ap .
the double modulated backscattered signal is used by the ap , which obtains the tag s information by removing its own transmitted information . in this way
, sensors can be powered by wifi aps and also rely on them to access internet even in the presence of access by wifi devices , thereby providing inter - connectivity of everything for smart homes . to quantify the benefits for the iot nodes with backscatter circuit , we simulate a iot _ wireless sensor network _ ( wsn ) , where each iot sensor node adopts either a backscatter circuit or a traditional rf circuit ( including a mixer , dac and amplifier ) .
both types of the nodes are powered by a pb .
the iot nodes with a density of @xmath2 nodes / m@xmath3 are randomly distributed in a disk region with the radius of @xmath4 m and served by the pb at the center . each node aims to perform sensing and transmission to its intended receiver at a fixed distance of @xmath5 m. the iot nodes adopt a harvesting - then - sensing - and - transmission task sequence per @xmath6-ms time slot , where energy harvesting occupies @xmath7 ms and the other operations @xmath8 ms if there is sufficient energy . the antenna effective apertures of all nodes are @xmath9 m@xmath3 .
we consider friis free - space channels for wireless power transfer and it .
the carrier frequency is @xmath10 ghz , the modulation scheme bpsk , and the noise power at the receiver @xmath11 dbm , and the rf energy harvesting efficiency @xmath12 . each sensing task consumes @xmath13 @xmath14j of energy ( e.g. , ambient light sensor tsl2550d ) , and the digital circuit power consumption during the active mode is @xmath15 @xmath14w @xcite .
each backcom iot node does not perform rf energy harvesting while backscattering , i.e. , with @xmath16 reflection . for the traditional iot node ,
the power consumptions of the mixer and dac during active mode are @xmath17 @xmath14w and @xmath13 mw , respectively ( e.g. , mixer ad831 and dac8830 ) , and the power amplifier ( class ab ) works at an efficiency of @xmath12 . based on these practical settings , we investigate how much performance improvement of the iot network by adopting the backscatter circuit .
[ fig : ber ] plots the average ber at the receiver versus the pb s transmit power .
we see that the bers with the two kinds of iot nodes decrease inversely with the pb s transmit power and converge to the same value due to interference .
also we see that the ber achieved by the backscatter nodes is significantly reduced compared with the traditional nodes within a practical range of pb s transmit power , e.g. , the ber is reduced as @xmath6 times when the pb has a @xmath18 dbm transmission power .
[ fig : active ] plots the percentage of active nodes .
we see that the percentage of active iot nodes increases with the pb s transmit power , and the percentage of the active backscatter nodes is much larger than that of the traditional nodes .
for example , the improvement of the percentage of active nodes is @xmath19 when the transmit power is @xmath20 dbm , and the improvement is @xmath21 when the transmit power is @xmath18 dbm . therefore , simulation results clearly show that the backscatter nodes can significantly improve the performance of the iot network .
the future iot applications require the backcom systems to enable far - distance , low - latency , and high - rate communications between dense iot devices and also enable their ad hoc communication .
a few advanced backcom techniques aiming at tackling these challenges are discussed as follows .
different from conventional rfid tags which only need to report its i d information , the iot backcom tags are also required to perform sensing and computing which consume much more energy .
how to wirelessly power backcom using pbs ( see earlier discussions ) and maximize the power - transfer efficiency is a significant design issue .
the efficiency can be increased by energy beamforming from a multi - antenna pb and a tag . to this end ,
the pb needs to know the forward _ channel state information _ ( csi ) . for a free - space channel
, the csi reduces to the tag direction with respect to the pb .
the pb can form a beam using the well - known retrodirective beamforming technique that automatically generates a beamformer by conjugating the observation of the pilot signal sent ( i.e. , reflected ) by the tag .
on the other hand , for a scattering channel , the beamformer designs are more complex but can exploit the so called key - hole " channel structure due to backscattering ( see e.g. @xcite ) . another technique for improving the power - transfer efficiency is to optimize the cw waveform , e.g. , as a weighted sum of multiple sinusoidal waves ( see e.g. , @xcite ) .
the design aims at increasing the peak - to - average power ratio of the pb signal that yields a higher energy - harvesting efficiency due to its non - linearity as a function of incident power . for future iot
, there would be billions of d2d links exist at the same time .
though information flow in rfid applications is usually uni - directional , message exchange between nodes is common in iot .
thus , full - duplex communication can substantially reduce latency and interference , and improve the efficiency of spectrum utilization of the iot - d2d links .
there are two methods for implementing full - duplex backcom .
both requires a reader to have a full - duplex antenna allowing simultaneous transmission and reception @xcite .
consider simultaneously forward and backward it in a backcom system having one pair of reader and tag .
leverage prior knowledge of forward information , the reader can cancel it in the backward it signal and thereby retrieve the backward information . while this method supports symmetric bidirectional data rates , the other exploits the rate asymmetry in data transfer ( tag to reader ) and control signaling ( reader to tag ) .
specifically , the signals have different frequencies and can be decoupled by filtering or averaging @xcite .
as mentioned , interference in dense d2d iot poses a key design challenge that is exacerbated by backscattering .
fortunately , a large population of iot devices are sensors for smart cities and homes , and the burstiness in their transmissions can be exploited for tackling interference .
a suitable transmission technique is _ time - hopping spread spectrum _ ( th - ss ) , where each tag randomly selects one of @xmath22 time slots for transmitting a single symbol and the choices of different tags are independent @xcite . as a result
, the number of simultaneous links is reduced by the factor @xmath22 , called the processing gain , thereby suppressing interference .
an extreme form of th - ss can be realized by _ ultra - wideband _ ( uwb ) transmission , where an extremely large processing gain is achieved using ultra - narrow pulses whose durations are in the order of nano - second @xcite .
a key characteristic of backcom is the _
double path - loss _ due to the fact that the backscatter signal received at a reader propagates through the close - loop channel cascading the downlink and uplink channels .
the resultant path loss is especially server as the propagation distances in iot are much longer than those for rfid applications . to enhance link reliability , one solution is to deploy antenna arrays at readers and tags and apply spatial - diversity techniques
furthermore , multi - antennas can enhance the efficiency of wireless power transfer by enabling transmit energy beamforming and increasing receive antenna apertures .
backscattering introduces a special channel structure for the backward it in a _ multiple - input - multiple - output _ ( mimo ) backcom system , called a _
dyadic mimo channel _ , which captures the composite fading in the forward and backward channels @xcite .
to be specific , the cw signals sent by the transmit antennas of the reader propagate through the forward mimo channel , and are first combined at each antenna of the tag and then backscattered , and lastly propagate through the backward mimo channel to arrive at the receive antennas of the reader .
the resultant dyadic mimo channel has a similar structure as the classic _ keyhole mimo channel_. the space - time coding is a simple but suitable technique for achieving the diversity gain of such a channel . by adopting space - time coding ,
it has been proved that the achievable maximum diversity order is equal to the number of the tag s antennas @xcite . in other words , in contrary with the conventional mimo channel , increasing the number of receive antenna at the reader can not continuously enhance the reliability of the backward it .
low - power or passive backcom devices with energy harvesting capabilities can be densely deployed to provide pervasive and uninterrupted sensing and computing services that provide a platform for implementing applications for smart homes / cities . in a smart home
, a large number of passive backcom sensors can be placed at flexible locations ( e.g , embedded in walls , ceilings and furnitures ) .
they are freed from the constraints due to recharging or battery replacements as one or multiple in - house pbs can be deployed to simultaneously power all the sensors or otherwise they can operate on ambient energy harvesting .
the tasks performed by the sensors have a wide range such as detection of gas leak , smoke and co , monitoring movements , indoor positioning , and surveillance [ see fig . [
fig : app](a ) ] . as an example , backcom - based smart dustbins are able to monitor their trash levels and communicate the information to passing - by garbage trucks by backscattering , streamlining the trash - collection process . in another example ,
household robots are able to use the backscattered signals from the tags located on doors and furnitures for indoor navigation @xcite . in a smart city
, ubiquitous backcom sensor nodes can be placed in every city conner such as buildings , bridges , trees , street lamps , and parking areas .
they can streamline the city operations and improve our life quality via e.g. , monitoring of air / noise pollution and traffic and parking - availability indicating . the efficient sensing - data fusion and wireless power for backcom sensors
can be realized by the deployment of integrated pbs and aps at fixed locations or mounted on autonomous ground vehicles or uavs , providing full - city coverage without costly backhaul networks @xcite .
iot biomedical applications , such as plant / animal monitoring , wearable and implantable human health monitoring , require tiny and low heat - radiation communication devices .
backcom devices , which do not rely on any active rf component , can meet such requirements and thereby avoid causing any significant effect on the under - monitoring plants , animals , tissues or organs .
the advantages make backcom a promising solution for iot biomedical applications .
one example is the backcom - based smart google contact lens , as illustrated in fig .
[ fig : app](b ) .
the lens was invented in google in 2014 for the purpose of assisting people with diabetes by constantly measuring the glucose levels in their tears ( once per second ) .
the device consists of a miniaturized glucose sensor and a tiny backcom tag .
the tag is able to provide energy to the sensor by rf energy harvesting from a wireless controller , and also backscatter the measured blood sugar level to the wireless controller for diagnose purpose . looking into the future , backcom will find a wide range of biomedical applications .
in particular , implantable tiny backcom neural devices with ultra - low power consumption and heat radiation , will be placed on the surface of the patient s brain to help the study , diagnosis and treatment of diseases such as epilepsy and parkinson s disease , where the backcom implants will act as the brain - computer interface . until now
the most successful commercialization of the backcom technique is the application in logistics due to the ultra - low manufacturing cost of simple and passive backcom tags , as illustrated in fig .
[ fig : app](c ) .
the low - cost backcom tags make the commercialization easy and quickly . for example
, as early as 2007 , the biggest @xmath6 suppliers of the global renowned chain commercial group wal - mart have used the backcom technology for logistics tracking .
the technology have been helping the companies to substantially reduce operational cost , guarantee product quality , and accelerate the processing speed . in the past decade
, the popularization and the application of backcom have brought revolutionary changes to the logistics industry , due to its advantages compared with the conventional bar code technology , such as reduced manual control , long service lives , long reading distances , and encrypt - able and rewritable data . in the future ,
apart from the existing backcom techniques for logistics tracking and management , backcom - based three - dimensional orientation tracking is an emerging technique . by attaching an array of low - cost passive backcom tags as orientation sensors on the surface of the target objects ,
three - dimensional orientation information is available at the reader by analyzing the relative phase offset between different tags . in this way
, human workers can be warned when the angle of a cargo is larger than a threshold .
iot that aims to enable both the activity and the connectivity of billions of energy - consuming smart nodes , is challenged from the energy perspective .
the backcom systems and techniques provide promising solutions for tackling the challenge . in this article
, we have introduced the basic principles for backcom , summarising existing backcom system and network architectures and discussed several emerging backcom techniques .
moreover , we have described various applications of backcom in iot applications . with rapid advancements in both theory and applications ,
the technology is expected to play a key role in future iot by enabling truly ubiquitous network connectivity , and pervasive sensing and computing .
w. liu , k. huang , x. zhou , and s. durrani , `` full - duplex backscatter interference networks based on time - hopping spreading spectrum , '' _ arxiv e - prints _ , 2016 .
[ online ] .
available : https://arxiv.org/abs/1609.00062 v. liu , a. parks , v. talla , s. gollakota , d. wetherall , and j. r. smith , `` ambient backscatter : wireless communication out of thin air , '' _ acm sigcomm computer communication review _ , vol .
43 , no . 4 , pp .
3950 , 2013 . | the internet - of - things ( iot ) is an emerging concept of network connectivity at anytime and anywhere for billions of everyday objects , which has recently attracted tremendous attentions from both the industry and academia .
the rapid growth of iot has been driven by recent advancements in consumer electronics , wireless network densification , 5 g communication technologies [ e.g. , millimeter wave and massive multiple - input and multiple - output ( mimo ) ] , and cloud - computing enabled big - data analytics .
one of the remaining key challenges for iot is the limited network lifetime due to massive iot devices being powered by batteries with finite capacities .
the low - power and low - complexity backscatter communications ( backcom ) has emerged to be a promising technology for tackling the challenge . in this article
, we present an overview of the active area by discussing basic principles , system and network architectures and relevant techniques .
last , we describe the iot applications for backcom and how the technology can solve the energy challenge for iot . |
dealing with quantum vacuum whose ` microscopic ' physics is still unknown the high - energy , general - relativity and condensed - matter communities use different experience developed in working in each of those fields @xcite . in condensed matter there is a rather general class of fermionic systems , where the relativistic quantum field theory gradually emerges at low energy and where the momentum - space topology is responsible for the mass protection for fermions , so that masses of all the fermions are much smaller than the natural energy scale provided by the microscopic ( trans - planckian ) physics @xcite .
since the vacuum of the standard model belongs to the same universality class of quantum vacua , this condensed - matter example provides us with some criteria for selection of the particle physics theories : the theory which incorporates the standard model must be consistent with its condensed - matter analog .
here we apply such criteria to the multiple point principle ( mpp ) @xcite . according to mpp
, nature chooses the parameters of the standard model such that two or several phases of the quantum vacua have the same energy density .
these phases coexist in our universe in the same manner as different phases of quantum liquids , such as superfluid phases a and b of @xmath1he or mixtures of @xmath1he and @xmath2he liquids . using mpp nielsen and
co - workers arrived at some prediction for the correlation between the fine structure constants in their extension of the standard model .
the fine tuning of the coupling constants is similar to the fine tuning of the chemical potentials of the coexisting quantum liquids in equilibrium . the problem of the standard model parameters is thus related to the problem of the vacuum energy , and correspondingly to the cosmological constant problem .
it was suggested @xcite that mpp can serve as a basic principle to explain the present value of the cosmological constant . from the condensed - matter point of view
such connection is rather natural .
according to observations the cosmological constant is ( approximately ) zero in our phase of the quantum vacuum , that is why it must be ( almost ) zero in all the coexisting vacua as well . in condensed matter
such nullification of the vacuum energy occurs for the arbitrary phase of the quantum vacuum .
this happens due to the thermodynamic gibbs - duhem relation , according to which the microscopic ( trans - planckian ) degrees of freedom exactly cancel the contribution to the vacuum energy from the low - energy ( sub - planckian ) degrees of freedom @xcite .
the phenomenon of nullification is so general that it must be applicable to any macroscopic system including the quantum vacuum of relativistic quantum fields , irrespective of whether the vacuum is true or false , and even if we do not know the microscopic physics .
since the mpp is justified by condensed - matter analog , we can apply it to different problems related to quantum vacuum . here
we discuss the scenario of the baryonic asymmetry of the universe , which follows from mpp .
the phase transitions between different quantum vacua does not influence the phenomenon of nullification of the vacuum energy : the energies of the vacuum is zero both above the transition and also after some transient period below the phase transition . in quantum liquids ,
the microscopic degrees of freedom which adjust themselves to nullify the energy density in a global equilibrium are the underlying bare particles atoms of the liquid .
the number density @xmath3 of atoms changes after the phase transition and this compensates the change of the vacuum energy .
for example , after the phase transition between superfluid @xmath1he - a and superfluid @xmath1he - b the relative change in the particle density is @xmath4 here the superfluid transition temperature @xmath5 characterizes the energy scale of the superfluid phase transition , and also the transition between @xmath1he - a and @xmath1he - b ; the fermi energy @xmath6 characterizes the atomic ( planck ) energy scale of the liquid .
it is important that the correction to the microscopic parameter @xmath3 ( and also to @xmath7 : @xmath8 ) is very small , and thus it does not influence the parameters of the effective theory of superfluidity in the low - energy corner .
the translation to the language of the standard model is almost straightforward .
let us consider the relative change of the microscopic ( trans - planckian ) parameters needed to nullify the vacuum energy of the standard model after , say , the electroweak phase transition .
in this case one must identify @xmath9 and @xmath10 ( the planck energy ) .
the relative change of the microscopic parameters in quantum liquids corresponds to the relative change of the planck physics parameters , and thus one can identify @xmath11 . however , the equation ( [ particledensitychange ] ) is not applicable for the standard model .
the reason is that the fermionic density of states ( dos ) in standard model above the electroweak phase transition differs from dos in liquid @xmath1he above the superfluid phase transition .
the vacuum in non - superfluid normal @xmath1he above the superfluid transition belongs to the fermi - surface universality class , while the vacuum of the standard model above the electroweak transition belongs to the universality class with fermi points .
that is why they have different density of fermionic states : @xmath12 in the vicinity of the fermi surface and @xmath13 in the vicinity of the fermi point .
thus the energy density related to superfluidity is @xmath14 , while the energy denstity involved in the electroweak transition is @xmath15 .
this gives an additional factor @xmath16 , as a result the relative correction to the planck energy needed to compensate the energy change of the vacuum after the electroweak transition is @xmath17 such response of the vacuum is so extremely small that it can not influence the parameters of the effective low - energy theory the standard model .
this demostrates that the adjustment of the deep vacuum does not lead to any sizable correlation between the parameters of the standard model , and thus the cosmological constant problem has nothing to do with the parameters of the effective theory .
however , the mpp contains a more strong assumption than the statement that each vacuum always acquires zero energy .
it assumes that several essentially different vacua have zero energy simultaneously , i.e. these vacua coexist in the same universe ( though the phase boundaries between different vacua can be well behind the cosmological horizon ) .
the coexistence , though it does not influence the parameters of the effective theories , leads to other physical consequences , such as baryonic asymmetry of universe .
let us discuss the principles of the coexistence of quantum vacua using as an example the coexisting quantum liquids , where the coexistence can be regulated both by microscopic and macroscopic parameters ( analogs of microscopic or macroscopic fermionic charges ) .
let us first consider the quantum liquid formed by the mixture of @xmath18 sorts of atoms .
an example of the mixture of @xmath19 components is provided by the liquid solution of @xmath1he atoms in @xmath2he liquid .
the number of atoms @xmath20 of each species @xmath21 is conserved , and it serves as the conserved microscopic fermionic charge of the vacuum ( the ground state of the mixture ) . the relevant vacuum energy whose gradient expansion gives rise to the effective quantum field theory for quasiparticles at low energy is @xcite @xmath22 where @xmath23 is the hamiltonian of the system ; @xmath24 is the particle number operator for atoms of sort @xmath21 in the mixture ; and @xmath25 is their chemical potential . if the liquid is in equilibrium it obeys the gibbs - duhem relation which expresses the energy @xmath26 through the other thermodynamic varaibles including the temperature @xmath27 , the entropy @xmath28 , the particle number @xmath29 and the pressure
@xmath30 : @xmath31 from this thermodynamic relation and eq.([vacuumenergy ] ) one obtains at @xmath32 the familiar equation of state for the vacuum , which is valid both for relativistic and non - relativistic systems : @xmath33 where we also introduced the energy density @xmath34 and particle number density @xmath35 . if the system is isolated from the environment , its pressure is zero and thus the energy density is zero too : @xmath36 for such condensed - matter systems , in which the effective gravity emerges in the low - energy corner , this equation means that the effective cosmological constant is zero .
this nullification occurs without fine - tuning for any vacuum since the microscopic degrees of freedom the particle number densities @xmath37 and chemical potentails @xmath25 automatically adjust themselves in equilibrium in such a way that the gibbs - duhem relation ( [ gibbs - duhemrelation ] ) is satisfied .
the more components the liquid has , the more flexible is the vacuum state , and as the result the number @xmath38 of different vacua which can coexist is bigger . in such flexible system the multiple point principle naturally emerges . for the system with @xmath18 components ,
the maximal number of different vacua which can coexist being separated by the phase boundaries is @xmath39 ( see fig .
[ coexistingvacuafig ] for @xmath40 ) , and all of these vacua have zero energy density : @xmath41 ( @xmath42 ) .
this results from the following consideration .
the coexisting vacua must have the same chemical potentials @xmath25 because of the exchange of particles between the vacua .
they also have the same pressure @xmath43 ( and the same temperature @xmath32 ) .
thus for each vacuum @xmath44 the pressure as a function of the chemical potentials must be zero : @xmath45 .
all these @xmath38 equations can be satisfied simultaneously if @xmath46 .
coexisting vacua a , b and c in a droplet of a substance with @xmath47 conserved charges , @xmath48 , @xmath49 , and @xmath50 .
the droplet is isolated from environment , so that in all three vacua the pressure @xmath43 if the curvature of the boundary of the droplet and the curvature of interfaces are neglected .
that is why the energy density is also zero in all three coexisting vacua : @xmath51 .
volumes @xmath52 , @xmath53 and @xmath54 occupied by the three vacua are determined by the total microscopic fermionic charges of the droplet : the particle numbers @xmath55 , @xmath56 and @xmath57 . ]
does the coexistence of quantum vacua lead to observable consequences for the effective field theories emerging in these vacua ?
the answer is yes , if some of the variables @xmath37 are soft variables belonging to the low - energy world , such as the density of the baryonic charge stored in the vacuum .
an example is provided by the superfluid phases of @xmath1he , a and b , which can coexist at @xmath32 and @xmath43 in an applied magnetic field @xmath58 @xcite .
the corresponding vacuum energy density is @xmath59 where @xmath3 is the number density of @xmath1he atoms ; @xmath60 is the density of the angular momentum which comes from the spins of the atoms ( each atom has spin @xmath61 ) ; @xmath62 ; and @xmath63 is the gyromagnetic ratio of the @xmath1he atom . for a given direction of the magnetic field ,
say @xmath64 , the liquid can be represented as the mixture of the @xmath19 components , with spin up and spin down : @xmath65 where @xmath66 since we have effectively @xmath19 components , the @xmath67 vacua can coexist in the absence of the environment , i.e. at @xmath43 .
this is the reason why a and b phases can coexist at @xmath32 and @xmath43 . as distinct from the variable @xmath3 , the fermionic charge
@xmath68 is a soft variable since in typical situations it is zero ( in the absence of external magnetic field ) . when the two vacua coexist , both variables @xmath3 and @xmath68 adjust themselves to nullify the pressure ( and thus to nullify the ` cosmological constant ' @xmath69 ) in each of the two phases .
however , while the change of the particle density is negligibly small , @xmath70 , the change of the variable @xmath68 is essential since it changes from zero .
the energy density related to non - zero @xmath68 is on the order of @xmath71 .
comparing this with the superfluid energy @xmath72 one obtains that the characteristic spin density of the vacuum which emerges to compensate the ab phase transition is @xmath73 this is much bigger than the relative change in the particle density @xmath3 after adjustment in eq.([particledensitychange ] ) . as a result
the parameters of the effective theory , which describe superfluidity , also change considerably .
in particular , the originally isotropic b - phase becomes highly anisotropic in the applied magnetic field ( or at non - zero spin density of the liquid ) needed for coexistence of a and b phases .
note that @xmath68 is the fermionic charge of the vacuum , and in principle it is not related to the fermionic charge of matter , since in our example the matter ( quasiparticles ) is absent .
however , the charge asymmetry in the vacuum sector can cause the charge asymmetry in the matter sector due to exchange between the vacuum and matter . the resulting excess of the fermionic charge in the matter sector can induce the non - zero matter density even at @xmath32 .
this is similar to the non - zero matter density in our universe caused by the baryonic asymmetry of matter .
let us consider how this baryonic charge can be induced .
let us start with the baryonic charge of the vacuum of the relativistic quantum field exploiting an analogy between the macroscopic global charges : the spin @xmath68 of the quantum liquid in its ground state and the global charges in our quantum vacuum , such as the baryonic charge @xmath74 ( or the family charge @xmath75 @xcite ) .
the spin @xmath68 of the liquid must be nonzero to provide the coexistence of the a and b vacua in superfluid @xmath1he at @xmath32 . in the same manner the baryonic charge @xmath74 or family charge @xmath75 could naturally arise in the quantum vacuum to establish the equilibrium between the coexisting phases of the vacuum .
let us consider the coexistence of several vacua whose physics differ below the energy scale @xmath76 .
such vacua can result from the broken symmetry phase transition , which occurs at @xmath77 , and we assume that the ordered phases differ by their residual symmetries @xmath78 . the energy densities involved in the coexistence of the vacua are on the order of @xmath79 .
let us assume that the coexistence is regulated by the exchange of the baryonic charge between the vacua
. then one can estimate the density of this baryonic charge in the vacua by equating the energy density difference @xmath80 and the energy density of the vacuum due to the nonzero charge density @xmath74 .
if the baryonic charge is stored in the microscopic degrees of freedom of the quantum vacuum , the energy density related to this charge must be on the order of @xmath81 . as a result
the baryonic charge density of the vacuum needed for the coexistence of different vacua is @xmath82 thus the coexistence results in the baryonic asymmetry in the vacuum sector . in turn , the non - zero baryonic charge of the vacuum could be in the origin of the baryonic asymmetry of the matter in our universe an excess of the baryons over antibaryons , @xmath83 .
let us consider this mechanism of baryogenesis .
if an exchange of the baryonic charge between the vacuum and matter is possible , the chemical potential for the baryons in matter must be equal to the chemical potential for the baryonic charge in the vacuum .
the latter is non - zero due to the non - zero baryonic charge in the vacuum sector in eq.([baryonicchargdensity ] ) : @xmath84 at temperature @xmath85 one obtains the following estimation for the baryonic charge stored in the matter sector ( in the gas of relativistic fermions ) : @xmath86 however , the exchange between the vacuum and matter occurs ( due to axial anomaly ) only at @xmath27 above the electroweak transition , @xmath87 . below the transition , at @xmath88 the exchange with the quantum vacuum
is highly suppressed : the transition rate due to the sphaleron mechanism becomes exponentially weak @xcite . at the moment of the phase transition ,
i.e. at @xmath89 , the baryonic asymmetry of matter ( primordial baryon - to - entropy ratio ) is : @xmath90 below the transition , the baryonic charge in the matter sector is completely separated from the vacuum and evolves together with matter , while the density of the baryionic charge in the vacuum sector remains constant . in the matter sector
the baryonic density evolves in the same way as the entropy , and thus the baryon - to - entropy ratio @xmath91 remains the same as at the moment of transition . to obtain the value @xmath92 , which follows from the cosmological observations @xcite , the characteristic energy @xmath93 , related to the baryonic charge of the vacuum ,
must be @xmath94 in analogy with a and b phases of @xmath1he , this corresponds to the transition temperature @xmath5 at which the coexisting vacua were formed due to symmetry breaking .
the main point in this scenario of the baryogenesis is that the vacuum and matter are two subsystems , whose properties related to the fermionic ( baryonic ) charge are different . in condensed matter the analogous exchange of spin charge between the superfluid vacuum and quasiparticles ( matter ) plays an important role in the spin dynamics of the system ( see @xcite and sec.8.6 in @xcite ) .
in conclusion , the gravitating part of the vacuum energy is always zero in equilibrium vacuum , @xmath95 , even if the cosmological phase transition occurs . the nullification after the phase transition is supported by automatic adjustment of the microscopic ultraviolet degrees of freedom .
however , because of the huge energy stored in the microscopic degrees , the relative change in the microscopic parameters is extremely small , and this adjustment practically does not influence the parameters of the effective infrared theories . as a result ,
the multiple point principle , which implies the coexistence of two or several different ( i.e. not connected by symmetry ) naturally occurs , and all the coexisting vacua automatically acquire zero energy without any fine - tuning .
if the universe is on the coexistence curve , this may lead to the observable physical consequences related to the fermionic charges of the vacuum and matter .
in particular , if the coexistence is regulated by the exchange of the baryonic charge , all the coexisting vacua acquire the baryonic asymmetry .
the latter in turn gives rise to the baryonic asymetry in the matter sector . according to eqs.([energyscale ] ) and ( [ baryonicchargdensity ] ) the density of the baryonic charge in the vacuum sector is rather high , @xmath96 .
what are the consequences of such @xmath97 violation in the quantum vacuum is the subject of further investigations . | according to the multiple point principle our universe is on the coexistence curve of two or more phases of the quantum vacuum .
the coexistence of different quantum vacua can be regulated by the exchange of the global fermionic charges between the vacua , such as baryonic , leptonic or family charge . if the coexistence is regulated by the baryonic charge , all the coexisting vacua exhibit the baryonic asymmetry . due to the exchange of the baryonic charge between the vacuum and matter which occurs above the electroweak transition , the baryonic asymmetry of the vacuum induces the baryonic asymmetry of matter in our standard - model phase of the quantum vacuum .
the present baryonic asymmetry of the universe indicates that the characteristic energy scale which regulates the equilibrium coexistence of different phases of quantum vacua is about @xmath0 gev . |
, title="fig:",width=192 ] , title="fig:",width=192 ] during recent years many ob stars have been discovered to be binary systems .
non - thermal radio emission is observed to be produced by some of these binary stars .
the non - thermal emissivity is thought to be due to synchrotron emission radiated by relativistic electrons .
the electrons are accelerated up to high energies by strong shocks produced by the collision between the two radiatively driven stellar winds ( ( * ? ? ?
* eichler & usov 1993 ) ) .
several parameters of the system can be constrained by the synchrotron emission , among them the mass loss rates from the primary and the secondary .
investigating the synchrotron radiation is thus necessary .
we model the non - thermal emission for two colliding wind systems , cyg ob2 no .
8a and cyg ob2 no . 9 , and compare the obtained results with the observations
the two colliding winds are separated by a contact discontinuity .
its position ( which , in our model , we assume to be coincident with the two shocks ) is defined as in ( * ? ? ?
* antokhin et al .
( 2004 ) ) .
the electrons are accelerated at the shock .
we follow them as they advect away and cool down due to adiabatic and inverse compton losses along the post - shock streamlines .
the momenta follow a modified power - law distribution .
the synchrotron emissivity from the relativistic electrons is calculated along the post - shock streamlines in the orbital plane .
the razin effect is included .
the third dimension is recovered by rotating the orbital plane along the line which connects the two stars
. we also include free - free emission and then calculate the fluxes and spectral indices at different orbital phases using adam s method ( @xcite ) . for cyg
8a the parameters are provided by ( * ? ? ?
* de becker et al . ( 2006 ) ) , for cyg ob2 no .
9 the orbital parameters are provided by ( * ? ? ?
* naz et al . ( 2010 ) ) , stellar parameters by ( * ? ? ?
* martins et al .
( 2005 ) ) and wind parameters by ( * ? ? ?
* vink et al .
( 2001 ) ) .
[ [ cygob2-no.8a - results . ] ] * cyg ob2 no . 8a results .
* + + + + + + + + + + + + + + + + + + + + + + + + + for cyg ob2 no .
8a ( see @xcite ) the radio data at 3.6 and 6 cm are obtained with vla , the x - ray data with xmm and rosat .
variability that is locked with the orbital phase is observed in both radio and x - rays . the x - ray and
the radio light curves are anti - correlated due to different formation regions .
the radio formation region is far out in the wind , along the contact discontinuity , while the x - rays are formed much closer to the apex of the contact discontinuity .
the model predicts phase - locked radio variability which is consistent with the observations ( see fig .
[ fig1 ] ) , even if the phases of the flux maximum and minimum do not agree .
[ [ cygob2-no.9-results . ] ] * cyg ob2 no . 9 results .
* + + + + + + + + + + + + + + + + + + + + + + + + @xcite studied the observed vla radio fluxes of cyg ob2 no . 9 at 3.6 , 6 , and 20 cm .
they found a 2.35 yr period from the data . a preliminary 6 cm light curve from our modelling shows variability in the radio flux linked to the orbital period that is the fingerprint of non - thermal radiation .
compared to the observations , the theoretical fluxes are much too high and the maximum occurs too early .
to improve the current results for cyg ob2 no . 8a and 9 , we need to include the orbital motion and solve the hydrodynamics equations .
a more detailed study is also necessary to better determine the star and wind parameters of the binary components and to investigate the porosity / clumping problem . | some ob stars show variable non - thermal radio emission .
the non - thermal emission is due to synchrotron radiation that is emitted by electrons accelerated to high energies .
the electron acceleration occurs at strong shocks created by the collision of radiatively - driven stellar winds in binary systems . here
we present results of our modelling of two colliding wind systems : cyg ob2 no .
8a and cyg ob2 no .
9 . |
capillarity is one of the most interesting subjects to teach in condensed matter physics , as its detailed understanding involves three otherwise separated domains : macroscopic thermodynamics@xcite , fluid mechanics and statistical physics@xcite .
the microscopic origin of surface tension lies in the intermolecular interactions and thermal effects@xcite , while macroscopically it can be seen as a force acting along the interface or an energy per unit surface . in the present article
we discuss the link between these three aspects of capillarity , on the basis of simple academic examples .
we first discuss the standard problems faced by students and many researchers in the understanding of surface tension .
we will see that the difficulty of understanding surface tension forces is often caused by the improper or incomplete definition of a system on which the forces act .
we bring up four basic questions , such as the one raised in the title , which are answered in the rest of the article .
contrary to many textbooks on the subject , this provides a picture that reconciles the microscopic , thermodynamic and mechanical aspects of capillarity . following the pioneering work of gibbs@xcite , we introduce surface tension as the excess free energy due to the presence of an interface between bulk phases .
let us consider a molecule in the vicinity of an interface , for example near a liquid - vapor interface .
the environment of this molecule is manifestly different from the molecules in the bulk .
this is usually represented schematically by drawing the attractive bonds around molecules , as shown in fig .
[ missingbounds ] .
one clearly sees from this picture that approximately half of the bonds are missing for a molecule at the interface , leading to an increase of the free energy .
one thus defines the surface tension from the free energy @xmath0 per unit area : @xmath1 for a system of volume @xmath2 containing @xmath3 molecules at temperature @xmath4 .
hence , @xmath5 is the energy needed to increase the interfacial area by one unit .
its dimension is @xmath6=m.t^{-2}$ ] ( mass per time squared ) and it is usually expressed in @xmath7 ( force per unit length ) or @xmath8 ( energy per unit area ) .
it is instructive to estimate the magnitude of surface tension , which must be of the order of the bond energy @xmath9 divided by the cross section area @xmath10 of a molecule
@xmath11 is a fraction of nanometer . for oils ,
interaction through van der waals interactions leads to @xmath12 and thus @xmath13
n / m . for water ,
hydrogen bonds lead to a higher value @xmath14 n / m . for mercury , the high energy bonds ( @xmath15 )
lead to a high surface tension @xmath16 n / m .
sketch showing the ` missing ' intermolecular bonds close to the liquid - vapor interface .
this gives rise to an increase in free energy per unit area , i.e. the surface tension . ] in fluid mechanics , the surface tension is not defined in terms of a surface energy but rather as a force per unit length . in the bulk of a fluid at rest ,
two sub - parts of a fluid exert a repulsive interaction on one another , which is called the pressure .
if the surface separating these two subsystems crosses the liquid - vapor interface , an additional force needs to be taken into account : surface tension . as shown in fig .
[ forcepicture ] , the surface tension is a force tangent to the surface and normal to the contour separating the two subsystems .
the total force is proportional to the width of the contour , which we will call @xmath17 throughout the paper .
contrarily to pressure , surface tension is an attractive force . sketch showing surface tension as a force per unit length exerted by one subsystem on the other .
the system on which the forces act is the dotted region .
the force is parallel to the interface and perpendicular to the dividing line .
] the link between mechanics and thermodynamics is provided by the virtual work principle .
if one moves a contour of width @xmath17 by a length @xmath18 , the area of the interface of the subsystem considered increases by @xmath19 .
consequently , the free energy is increased by @xmath20 .
the free energy should equal the work done by the surface tension force , which means that this force must be parallel to the interface , normal to the contour and have a magnitude @xmath21 . per unit length ,
the tension force is thus @xmath5 . on the other hand ,
the link between mechanics and statistical physics is much less obvious for students .
one clearly sees in fig .
[ missingbounds ] that the molecule at the interface is submitted to a net force ( which would be represented by the sum of the vectors ) along the direction perpendicular to the interface .
however , we just argued from the mechanical point of view , that the force is _ parallel _ to the interface .
this leads to the first key question of this article : + _ @xmath22 q1 : why is surface tension a force parallel to the interface while it is so obvious that it must be perpendicular to it ? _ a standard experimental method for determining the liquid - vapor surface tension is to measure the force required to pull a metallic plate ( usually made of platinum ) out of a liquid bath .
this force is related to the liquid - vapor surface tension @xmath5 , as is usually explained by the diagram of fig .
[ young]a .
imagine that the plate is moved vertically by a distance @xmath18 .
the area of the liquid - vapor interface does not change by this motion , so the corresponding interfacial energy is unaffected .
however , the motion does lead to a decrease of the immersed solid - liquid interface area , by @xmath23 , while the solid - vapor interface increases by the same amount .
in other words , part of the wetted surface is exchanged for dry surface .
this leads to a variation of the free energy @xmath24 , where @xmath25 and @xmath26 are the solid - vapor and solid - liquid surface tensions respectively .
this energy is provided by the work done by the experimentalist , due to the force required to displace the plate by @xmath18 .
hence , this force must be equal to @xmath27 .
( a ) experimental method to determine the liquid - vapor surface tension .
the force per unit length needed to pull a plate from a bath of liquid is equal to @xmath28 , where @xmath29 is the equilibrium contact angle .
( b ) a classical way to interpret young s law as a force balance of surface tensions . question 2 : why is there no force balance in the normal direction ?
question 3 : why do we draw a single surface tension force in the left panel ( @xmath5 ) while there are three in the right panel ( @xmath30 ) ? ] to relate this force to the value of the liquid - vapor surface tension @xmath5 , one has to invoke young s law for the contact angle @xmath29 .
when the three interfaces ( between solid , liquid and vapor ) join at the contact line , the liquid makes contact with the substrate at an angle @xmath29 given by@xcite : @xmath31 with this , the force exerted on the plate can be expressed as @xmath32 , and can thus be used to design a tensiometer . from the mechanical point of view , we can interpret this force as the surface tension acting parallel to the liquid - vapor interface . by symmetry ,
the total force exerted on the solid is vertical ( the horizontal components add up to zero ) . projecting the surface tension force on the vertical direction and multiplying by the length @xmath17 of the contact line , one indeed gets @xmath33 . by
a similar argument one usually interprets young s law for the contact angle as the balance of forces at the contact line ( fig .
[ young]b ) . by projection along the direction parallel to the solid substrate , one obtains @xmath34 , which is the same as equation ( [ young ] ) .
this force interpretation is a common source of confusion for students : + _ @xmath22 q2 : from fig .
[ young]b , there seems to be an unbalanced force component in the vertical direction @xmath35 .
what force is missing to achieve equilibrium ? _ + _ @xmath22 q3 : why does one draw a single force acting on the contact line in the case of the plate ( fig .
[ young]a ) , while for young s law we need to balance all three forces ( fig .
[ young]b ) ? _
actually , when measuring a surface tension using the plate technique , one often uses a platinum plate to be sure that the liquid completely wets the solid . in that case , however , @xmath36 and young s law does not apply ! then , the thermodynamic and mechanical approaches give conflicting answers : + _
@xmath22 q4 : in the situation of complete wetting , is the force on the plate given by @xmath5 or by @xmath37 ? _ before addressing these points in detail , we start with a short overview of the answers to the questions raised above .
we emphasize that the thermodynamic result ( i.e. from the virtual work principle ) always gives the correct total force . if one wants to know the _ local _ force distribution , which can not be extracted from thermodynamics , it is imperative that the system on which the forces act is properly defined .
confusion regarding the forces is often caused by an improper or incomplete definition of such a system .
@xmath22 a1 : the schematic of fig . [ missingbounds ] represents only the attractive intermolecular forces .
the real force balance requires both repulsive and attractive interactions between liquid molecules .
_ to answer the questions related to the contact line it is crucial to specify the system of molecules on which the forces are acting : + _ @xmath22 a2 : in young s law , the system _ on which the forces act _ is a corner of liquid bounded by the contact line .
@xmath5 is indeed the force exerted on this system inside the liquid - vapor interface , but the forces exerted by the solid on the corner are incomplete in fig .
[ young]b .
an extra vertical force on the liquid , caused by the attraction of the solid , exactly balances the upward force @xmath38 . + _ _
a3 : to obtain the force on the plate , the system to consider is the solid plate . in this case , the force exerted by the liquid on the solid is in fact equal to @xmath39 per unit length . + _ _
a4 : the correct vertical force on the plate is @xmath33 . in the case of complete wetting ( @xmath40 )
, the virtual work principle can be applied , but only when taking into account the prewetting film . _
lennard - jones intermolecular potential @xmath41 .
the interaction is strongly repulsive for intermolecular distances @xmath42 , where @xmath11 reflects the ` hard - core ' of the molecules . at large distances the molecules
are attracted to one another .
the gray arrow points to the presence of thermal fluctuations , which , in a liquid , lead to substantial variations of the intermolecular distance . ] to address the origin of capillarity , we first have to understand how a liquid phase and a vapor phase can coexist .
this goes back to the work of van der waals who derived an equation of state for fluids that can account for a liquid - gas phase transition@xcite : @xmath43 here @xmath44 is the pressure , @xmath45 the volume per molecule , and @xmath4 the temperature .
this equation of state corrects the ideal gas law to incorporate the effect of intermolecular forces .
the constant @xmath46 introduces repulsion between molecules as an excluded volume effect : the pressure diverges when the total volume per molecule reaches a minimal size @xmath46 . in this limit the molecules are densely packed and constitute a liquid phase . in this phase , the volume per molecule no longer depends on pressure , which means that the liquid phase is incompressible .
ultimately , this effect comes from the repulsion of the electron clouds of the molecules , due to pauli exclusion principle in quantum mechanics .
the constant @xmath47 represents the long - range attraction between molecules which finds its origin in the dipole - dipole interaction ( van der waals attraction ) .
van der waals s equation of state ( [ vdw ] ) explains how a low density gas phase can coexist with a high density liquid phase .
this coexistence requires the pressures to be identical on both sides of the interface , despite the striking difference in density . in a gas , where @xmath48 is large ,
most of the energy is of kinetic origin ( @xmath49 ) : the pressure is @xmath50 . in the liquid phase the volume per unit molecule is almost in the incompressible limit : @xmath51
this strong repulsive effect ( @xmath52 ) , however , is counterbalanced by the presence of attractions ( @xmath53 ) so that , for the same temperature , the pressure in the liquid phase can be in equilibrium with the pressure in the gas phase .
this gives rise to a stable liquid - vapor interface .
how can a liquid at the same time be repulsive and attractive ?
a single pair of atoms can of course only attract or repel each other , depending on the distance separating their two nuclei .
this interaction is shown schematically in fig .
the steep potential reflects the short - range repulsion , while the negative tail of the potential represents the long - range attraction .
this means that the balance of attractive and repulsive interactions in ( [ vdw ] ) only has a statistical meaning : some particles are in an attractive state while others are in a repulsive state .
this property of ` simultaneous ' repulsion and attraction is what makes a liquid very different from a solid : in an elastic solid , molecules in the same region are either all compressed ( if the solid is submitted to an external compression ) , or all attracted to each other ( if the solid is submitted to an external tension ) . this point is briefly discussed in the appendix . the difference between a solid and a liquid can be traced back to the importance of thermal fluctuations , i.e. the kinetic energy of the molecules . in the solid phase ,
these fluctuations are relatively small with respect to the potential energy , i.e. @xmath54 , where @xmath9 is the energy scale for the intermolecular forces . as a consequence the system only explores a small region of the potential . hence the solid is either in the compressed or in the tensile state . on the other hand ,
the liquid phase is characterized by large fluctuations , for which @xmath55 . a broad range of the potential is therefore sampled by molecules in the same region of space ( fig . [ lj ] ) .
finally , the case @xmath56 corresponds to a gas phase of weakly interacting particles that is dominated by kinetic energy . the liquid - vapor interface .
the vertical axis is in units of the molecular scale @xmath11 .
( a ) snapshot of lennard - jones simulation of a liquid - vapor interface .
( b ) time - averaged normalized density profile @xmath57 across the interface .
( c ) tangential force per unit area exerted by the left part on the right part of the system . technically speaking ,
the plot shows the difference @xmath58 between the normal and tangential components of the stress tensor . ]
let us now consider the liquid - vapor interface in more detail .
figure [ liquidvapor]a shows a snapshot of the interface obtained in molecular dynamics simulations of molecules that interact through a lennard - jones potential @xcite .
the corresponding time - averaged density profile is plotted in fig .
[ liquidvapor]b .
the transition from the high density liquid to the low density gas takes place in a very narrow region that is a few molecules wide . in order to determine the capillary forces we need to divide the system along the direction normal to the interface into two subsystems ( see fig . [ liquidvapor]a ) .
one considers the force per unit area , called the stress , exerted by the left subsystem on the right subsystem as a function of the vertical position @xmath59 .
this stress can be decomposed into two contributions : the pressure @xmath44 , which we recall to be the same in the vapor and the liquid bulk , plus an extra stress @xmath60 acting along the direction parallel to the interface , plotted in fig .
[ liquidvapor]c .
the profile of this stress anisotropy shows that there is a force localized at the interface , acting in the direction parallel to the interface .
this force spreads over a few molecular scales , which is also the typical thickness of the density jump across the interface .
the integrated contribution of this force is indeed equal to @xmath5 per unit length , the surface tension .
this shows that surface tension really is a mechanical force .
having seen that in our simulations there is a parallel force localized at the interface , let us turn to question 1 .
why is the tension force parallel and not normal to the interface ?
first , we note that fig .
[ missingbounds ] only depicts the attraction between molecules .
a more complete picture also incorporates the repulsive forces in the internal pressure , as sketched as dashed arrows in fig .
[ missingboundscorrect ] . away from the surface
there is perfect force balance due to the symmetry around the molecule . near the interface ,
however , the up - down symmetry is broken . to restore the force balance in the vertical direction , the upward repulsive arrow ( dashed ) has to balance with the downward attractive arrow ( solid ) . in the direction parallel to the interface
the symmetry is still intact , thus automatically ensuring a force balance parallel to the interface .
this means that along the direction parallel to the interface , there is no reason why the attractive forces should have the same magnitude as the repulsive forces .
as described above , we find that in practice , the attractive forces are stronger , which indeed give rise to a positive surface tension force .
sketch showing repulsive ( dashed black arrows ) and attractive ( gray arrows ) forces in the bulk and at the surface . ]
however , we still need to explain why these intermolecular forces give rise to such a strong tension along the surface .
this question was previously addressed by berry@xcite , who noted the separate roles of attraction and repulsion .
the key observation is that , to a good approximation , the repulsive contribution to the pressure is isotropic while attraction is strongly anisotropic .
this is because the repulsion is very short ranged due to the hard core of the molecules , and can therefore be thought of as a contact force " . as such
, repulsion is not very sensitive to the changes in molecular structure near the interface and is equally strong in all directions@xcite .
by contrast , the long - range nature of the attractive forces make them very susceptible for the molecular structure .
this is the origin of the observed pressure anisotropy near the interface that generates the surface tension force . to see how this works out in detail it is useful to divide the liquid into two subsystems using control surfaces parallel to the liquid - vapor interface , as shown in fig .
[ interfacelv]a .
the force exerted on the dotted subsystem by the rest of the liquid results from the superposition of attractive ( vertical gray arrows ) and repulsive ( dashed black arrows ) interactions ( fig .
[ interfacelv]a ) .
as the subsystem is at equilibrium , these attractive and repulsive components must balance each other .
the magnitude of the attractive force increases with the size of attracting region this is because the density increases as one moves from the vapor towards the liquid phase .
the magnitude of the attraction saturates to the bulk value when the control surface is a few molecular sizes from the interface .
we then divide the liquid into two subsystems using a control surface perpendicular to the liquid vapor interface ( fig .
[ interfacelv]b ) .
one can now use that the repulsive short - range forces are isotropic .
this means that the magnitude of repulsion ( dashed black arrows ) exerted by the left side on the subsystem ( dotted region ) increases with depth in a way analogous to that in fig .
[ interfacelv]a .
by contrast , the strength of attraction has a much weaker dependence on depth for the sake of simplicity we draw it at a constant magnitude that equals the attraction in the bulk . as a result
, there indeed is a net attraction of the subsystem by the rest of the liquid ( dark gray arrow in fig .
[ interfacelv]c ) .
forces exerted on a subsystem of liquid ( dotted region ) by the rest of the liquid ( gray region without dots ) ( a ) the subsystem considered is the lower part of the liquid and is separated from the interfacial zone by a line parallel to the liquid - vapor interface .
the subsystem ( dotted region ) is submitted to an attractive force ( gray arrows ) and a repulsive force ( dashed black arrows ) exerted by the rest of the liquid ( gray region without dots ) .
they must balance each other .
( b ) the liquid is now divided along a line perpendicular to the interface .
the subsystem considered ( dotted region on the right ) is submitted to an attractive force ( gray arrows ) and to a repulsive force ( dashed arrows ) exerted by the rest of the liquid ( gray region without dots).as the repulsive force is isotropic , it has the same magnitude as in ( a ) and therefore decays close to the surface . on the contrary ,
the attractive force is nearly constant and remains almost unchanged close to the surface .
( c ) this leads to a net attractive force from one side on the other . ]
we now consider the liquid - solid interface ( see fig . [ interfacels ] ) . here , two effects superimpose .
first , due to the presence of the solid , there is a lack of liquid in the lower half - space ( hatched region in fig .
[ interfacels ] ) .
this missing liquid induces an anisotropy of the attractive liquid - liquid force in the same way as it does in the case of the liquid - vapor interface .
therefore , as in fig .
[ interfacelv]c , the left hand side of the liquid exerts a net attractive force @xmath5 per unit length on the right hand side subsystem .
the second effect is due to the liquid - solid interaction . in the same way as with the liquid - vapor interface
, we divide the liquid into two subsystems using a control surface parallel to the interface , as shown in fig .
[ interfacels]a .
the attraction by the solid ( gray arrow ) decreases with the distance and is perfectly balanced by a short range liquid - liquid repulsive force ( dashed arrows ) .
now , we divide the liquid into two subsystems using a control surface _
perpendicular _ to the liquid solid interface ( fig .
[ interfacels]b ) .
assuming once more that the liquid - liquid repulsion is isotropic , the left part of the liquid exerts a net repulsive force on the right subsystem ( dotted region ) .
this force is induced by the influence of the solid and , perhaps surprisingly , it is not equal to @xmath26 . instead
, it has been shown@xcite that this force is equal to @xmath61 , and this will be motivated in more detail below .
to combine these two effects , we subtract the unbalanced attractive force due to the absent liquid ( @xmath5 ) from the repulsive force due to the solid ( @xmath61 ) to find the net repulsive force between the subsystems : @xmath37 .
forces exerted by the solid ( dashed line ) on a subsystem of liquid ( dotted region ) . on these schematics , the attractive liquid - liquid interactions already treated in fig .
[ interfacelv ] , are not considered .
( a ) the liquid subsystem is semi - infinite .
it is delimited by a line parallel to the liquid - solid interface , at different distances above it . the subsystem ( dotted region )
is submitted to an attractive force ( gray arrows ) exerted by the solid and to a repulsive force ( dashed arrows ) exerted by the rest of the liquid ( gray region without dots ) .
as the subsystem is in equilibrium , they must balance each other .
( b ) the liquid is now divided along a line perpendicular to the interface .
only the horizontal force components are shown .
the solid exerts no horizontal attraction .
as the repulsive interactions are isotropic , this results into an horizontal repulsive force exerted by one side on the other . ]
relation between adhesion work and surface tensions .
( a ) to split a liquid volume into two semi - infinite volume , one has to create two liquid - vacuum interfaces .
it costs an energy @xmath62 .
( b ) to create a solid - liquid interface one first needs to create a liquid - vacuum and solid - vacuum interface , which costs an energy @xmath63 .
joining the liquid - vacuum and solid - vacuum interfaces yields an energy reduction @xmath64 due to the solid - liquid attraction . ]
let us motivate why the strength of the solid - liquid interaction does not couple directly to the solid - liquid surface tension @xmath26 , but to the combination @xmath61@xcite .
note that this feature will also be crucial for understanding the wetting phenomena discussed in the following section .
the solid - liquid surface tension represents the free energy needed to create a solid - liquid interface . to make such an interface ,
one first has to break " a bulk solid and a bulk liquid into two separate parts , and then join these solid and liquid parts together .
this breaking of liquid is depicted schematically in fig .
[ sl]a ( it works similarly for the solid ) .
the corresponding energy is the work of adhesion " @xmath65 due to liquid - liquid attractions ( @xmath66 for the solid ) .
this gives rise to a surface tension @xmath67 ( 2@xmath68 ) , since the liquid ( solid ) is connected to a vacuum at this intermediate stage .
when joining the solid - vacuum and liquid - vacuum interfaces , the attractive solid - liquid interaction will reduce the surface energy by the solid - liquid work of adhesion @xmath69 ( fig .
hence , the resulting solid - liquid surface tension becomes @xmath70 from this one finds that , indeed , the strength of the solid - liquid adhesion reads @xmath71 in the last step we used young s law for the equilibrium contact angle . as a consequence
, the capillary forces induced by solid - liquid attractions will have a magnitude @xmath72 and not @xmath26 .
the question of the force balance is even more intricate in the vicinity of the contact line , where the liquid - vapor interface meets a solid .
it is crucial to note that _ the contact line itself does not represent any material _ : it is a mathematical line that marks the separation between wetted and dry parts of the solid .
the question what is the force on the contact line ? " is thus ill - posed , since there are no molecules on which such a force would act . as a matter of fact ,
only a material system ( a collection of matter ) can be submitted to a force .
therefore , care should be taken to properly define the systems that play a role near the contact line : the liquid near the contact line and the solid underneath it . in the following sections we will show how a careful consideration of all the forces on the appropriate material systems will lead to proper force balances , consistent with the thermodynamic predictions .
all results and sketches provided in this section , some of which may appear counterintuitive , are backed up by a density functional theory model for microscopic interactions @xcite and molecular dynamics simulations.@xcite solid and liquid forces acting on a liquid subsystem ( dotted regions ) near the contact line . ( a ) sketch of a wedge of liquid near the contact line with the three forces exerted on the system .
( b ) each of the three corners of this system must be treated differently .
the upper right corner is at the liquid vapor interface .
following fig .
[ interfacelv ] , the rest of the liquid exert a net attractive force parallel to the interface equal to @xmath5 per unit length .
the lower right corner is at the liquid solid interface .
following figs .
[ interfacelv ] and [ interfacels ] , the rest of the liquid exerts a repulsive force @xmath37 .
the liquid near the solid - liquid interface is attracted by the solid , this force is balanced everywhere by repulsion at the solid - liquid interface , except in the vicinity of the contact line . ]
let us consider the forces on the wedge - shaped liquid corner in the vicinity of the contact line , as shown in fig .
[ youngcorrect ] . choosing this subsystem
, we will now explain young s force construction from fig .
[ young]b and answer question 2 : what happens to the force balance normal to the solid - liquid interface ?
there are two types of forces acting on the liquid molecules inside the subsystem : interactions with the solid and interactions with other liquid molecules outside the subsystem .
we first consider the solid - on - liquid forces .
one can immediately see that , since the solid spans an infinite half space , every liquid molecule experiences a resultant force that is oriented purely normal to the solid - liquid interface : the left - right symmetry of the solid ensures that there is no force component parallel to the interface .
far from the contact line at the solid - liquid interface , this attractive force is balanced by a repulsive force , as discussed in fig .
[ interfacels ] .
however , since the repulsive force is continuous and zero outside the droplet , the repulsive force must decay close to the contact line .
this means that there is an unbalanced attractive force that is strongly localized in the vicinity of the contact line .
it has been shown@xcite that this force per unit length is equal to @xmath73 .
the existence of this force has recently been challenged@xcite . to show that this force indeed must exist to achieve equilibrium , we consider the droplet shown in fig . [ laplace ] .
choosing the drop as the system , and recognizing that the force in the interior of the droplet at the liquid - solid interface ( small arrows ) is due to the laplace pressure @xmath74 ( with @xmath75 the curvature @xmath76 ) , the attractive force at the contact line must be exactly @xmath35 to achieve a force balance@xcite .
this provides the answer to question 2 : the downward solid - on - liquid force is not drawn in fig .
[ young]b . this missing force has often been interpreted as a reaction from the solid@xcite , whose existence is demonstrated experimentally by the elastic deformation of soft solids below the contact line@xcite . here , we clarify the molecular origin of this normal force@xcite . forces acting on a system , in this case a liquid drop ( dotted area ) .
the system is in equilibrium so the sum of all ( external ) forces must be zero . due to laplace pressure , there is a repulsive force exerted by the solid on the liquid across the liquid - solid interface ( upward black arrows ) . in the vicinity of the contact line ,
repulsion and attraction of the liquid by the solid do not balance each other .
therefore , the solid attracts the liquid with a vertical force equal to @xmath73 per unit length ( downward dark gray arrows ) . ] to finalize the force construction near the contact line we return to the wedge shown in fig .
[ youngcorrect]b .
since the solid can only exert a normal force on the liquid , all parallel force components drawn in young s construction are purely due to the liquid molecules outside the corner .
the force drawn along the liquid - vapor interface can be understood directly from the tension @xmath5 inside the liquid - vapor interface ( cf .
the discussion of fig .
[ interfacelv ] ) .
a similar force arises at the solid - liquid interface ( fig .
[ interfacels ] ) , which , however , is repulsive and has a magnitude @xmath37 , as explained in the previous section . including these forces gives a perfect force balance on the liquid corner , as seen in fig .
[ youngcorrect]a .
in fact , one can easily verify that even the resultant torque ( or force moment ) is zero for this force construction .
as such , it provides a more physical alternative to the classical picture of young s law . the measurement of surface tension in fig .
[ young]a directly relies on the force exerted by the liquid on the solid plate .
again , we emphasize the importance of a proper definition of `` the system '' on which the forces act , and in this case this is the solid on which the liquid rests .
the situation is thus very different from the forces on the liquid corner , which are in equilibrium and experiences a zero resultant force .
this _ difference between systems _ provides the key to question 3 . in fig .
[ young]a , the total force exerted by the liquid on the solid is represented by the resultant @xmath77 , while fig .
[ young]b represents the balance of the forces acting on the liquid wedge .
forces acting on the solid subsystem ( hatched areas ) by the liquid ( gray areas ) near the contact line .
( a ) distribution of forces acting on the solid near the contact line . due to the attraction of the liquid , the solid
is attracted towards the liquid ( solid gray arrows ) .
the absence of liquid on the left part of the contact line ensures the tangential force is always towards the liquid , even for @xmath78 .
the repulsion ( dotted arrows ) arises from the contact force at the solid - liquid interface .
far from the contact line , repulsion and attraction balance each other .
( b ) resultant force acting on the solid near the contact line .
the net normal force is @xmath79 , and the parallel force @xmath80 . ]
once again , we turn to the microscopic description of the forces in the vicinity of the contact line .
it turns out that the normal component of the force exerted on the solid is equal to @xmath79 , consistent with the macroscopic picture of a tension along the liquid - vapor interface .
however , the parallel component of the liquid - on - solid force does not have the expected magnitude @xmath81 , but @xmath82 .
this can be understood as follows .
[ forcesolid]a illustrates that the tangential force component originates from the long - range attraction between solid and liquid molecules . in the previous section we already demonstrated the strength of this solid - liquid adhesion to be @xmath83 .
hence , there is no reason why the total force on the solid should be @xmath81 .
indeed , the dft calculation confirms a tangential liquid - on - solid force of magnitude @xmath83@xcite . the physics of this surprising result is nicely illustrated by fig . [ forcesolid ] .
the macroscopic intuition that the resultant surface tension force is pulling along the liquid - vapor interface would predict a force to the left whenever the contact angle @xmath84 .
however , it is clear from the sketch of the attractive forces that the sum of all parallel components must be oriented towards the liquid ( right side in the figure ) .
this stems from the asymmetry between the amount of liquid attracting the solid molecules on both sides of the contact line : there are always more liquid molecules on the right side of the contact line in this figure .
this is consistent with a parallel force @xmath85 , but not with a force @xmath28 ( which changes sign at @xmath86 ) .
note that when considering the force exerted by the solid _ on the liquid _
, this asymmetry does not occur since the solid is left - right symmetric , therefore there is no tangential component .
this once more illustrates that a detailed force interpretation crucially relies on the definition of the system .
forces acting on a solid at the solid - liquid interface .
( a ) without liquid , there is neither repulsion nor attraction .
( b ) when liquid is present there is repulsion and attraction .
however , the repulsion is not completely balanced in this curved interface , as there is more liquid than in a plane geometry .
( c ) the resulting force per unit surface is @xmath87 where @xmath75 is the curvature of the liquid ( positive in this case ) , equivalently to the laplace pressure .
it only shows a dependency on the liquid - liquid interactions as it is only the absence of liquid that is relevant to the curvature . ]
to solve this apparent discrepancy with the thermodynamic result one has to consider all the forces exerted by the liquid on the solid , i.e. not just the forces near the contact line .
the key point is that the submerged solid bodies can not be flat everywhere and must have curved pieces of liquid - solid interface .
if the interface separating the solid from the liquid is flat , the net normal force is locally zero as repulsion balances attraction ( far away from the contact line ) .
however , when this interface is curved , a repulsive force inside the liquid is enhanced due to the curvature of the solid - liquid interface , in a way similar to the laplace pressure . as shown in fig .
[ laplacesolid ] , the presence of a curved half - space of liquid acts on the solid , and creates an unbalanced liquid - on - solid force @xmath87 per unit area . here
@xmath75 is the local curvature ( inverse radius of curvature ) of the solid .
interestingly , dft calculations@xcite show that the resultant pressure couples only to @xmath5 and not to @xmath26 . as we show below , this is exactly what is needed to restore consistency between microscopic and thermodynamic forces .
capillary forces on immersed objects .
( a ) schematic of a pin floating at the surface of a liquid under partial wetting conditions and in zero - gravity .
the downward thin gray arrows are the forces exerted by the liquid on the pin located in the vicinity of the contact line .
the small light gray arrows show the laplace pressure @xmath88 acting on the solid , due to the curvature @xmath75 of the solid - liquid interface .
( b ) @xmath29 denotes the contact angle and @xmath89 and @xmath90 are two unit vectors tangential to the pin , pointing upwards , at the two contact lines .
( c ) the thick gray arrows show the resultants of the capillary forces detailed in ( a ) , which apply on each half of the pin .
they reduce to forces tangential to the liquid - vapor interface at the contact lines .
note that this schematic does not show the distribution of capillary forces .
( d ) distribution of the capillary forces in the case of an irregular shape . since the integral over the curvature is equal to the sum of the two tangential vectors at the contact lines , the resultant is independent of the shape of the body .
it is thus the same as in ( c ) .
( e ) pin floating at the surface of a liquid under gravity .
the upward thick gray arrows are the resultants of capillary forces .
they balance the effect of gravity ( corrected by the archimedes force ) , shown as the downward black arrow . ]
an excellent demonstration of this effect is the long debated case of a ` floating - pin ' under zero - gravity , as shown in fig [ pin ] . whereas a floating pin in a system with gravity leaves a visible depression in the liquid - vapor interface near the contact line ( cf .
[ pin]e ) , the zero - gravity condition ensures that the interface has a constant curvature , i.e. it is straight everywhere .
this means that the vertical position depends on the equilibrium contact - angle alone , and not on the density ratio of the involved materials . as discussed in fig .
[ forcesolid ] , the liquid - on - solid force near the contact line is not oriented along the liquid - vapor interface , but points towards the interior of the liquid .
this creates a resultant downward force that has to be compensated to restore equilibrium .
additionally , the curvature of the solid - liquid interface creates a normal force distributed over the whole immersed surface of the solid of magnitude @xmath87 per unit surface ( cf .
[ pin]a ) . integrating over the curvature of the submerged surface from one contact line to the other
then gives the resultant of the laplace pressure : @xmath91 where @xmath89 and @xmath90 are the unit vectors tangential to the pin , pointing upwards .
therefore , the resultant is orientated upwards and is equal to @xmath92 per unit depth ( see fig .
[ pin]b ) .
it balances exactly the downward forces induced close to the contact lines ( fig .
[ forcesolid]b ) , hence the pin is in equilibrium .
importantly , this result is independent of the shape of the body ( see fig . [ pin]d ) .
the same principle applies to the partially wetted plate of fig .
[ young]a : the force exerted by the fluid on the plate results from two contributions , as shown schematically in fig .
[ platepartialtotal]c .
first , there is the vertical force component ( per unit length ) due to the vicinity of the contact line : @xmath93 ( cf .
[ forcesolid ] ) .
second , there are submerged surfaces of the plate where a localized curvature exists at the corners .
this curvature induces a laplace force on the pin ( see fig .
[ pin]d ) which results into a net upward force @xmath5 per unit length of contact line which means the total force ( per unit length of contact line ) on the plate is @xmath28 , in agreement with the thermodynamic result .
( a - b ) force per unit length of the contact line needed to keep a plate in equilibrium in a bath in partial ( a ) and complete wetting ( b ) .
( c ) partial wetting .
the vertical force at the contact line , equal to @xmath94 , is balanced by the laplace pressure induced by the curvature of the plate .
note that any plate shape would lead to the same resultant force as the integral of the curvature over the surface reduces to the local tangents at the contact line ( see fig .
[ pin]d ) .
( d ) complete wetting case . due to the mesoscopic pre - wetting film , of which
the thickness is exaggerated in the picture , there is no contact line , thus there is no force located near the apparent macroscopic contact line .
the forces are related only to the laplace pressure : the curvature of the solid gives a zero resultant force , as it is completely immersed in the liquid , and the curvature of the liquid is acting on the solid only in the pre - wetting zone as it is compensated by gravity in the meniscus .
the resultant force per unit length is then equal to @xmath5 . ] in the case of complete wetting , young s law for the contact angle is no longer applicable .
instead , the apparent contact angle @xmath29 vanishes because the three surface tensions do not balance each other : @xmath95 physically , there is no real contact line in this configuration ( fig .
[ platepartialtotal]b ) , but there is a meniscus where the liquid - vapor interface approaches the solid . beyond the meniscus
, there exists a mesoscopic liquid film called a pre - wetting film , which covers the solid completely ( fig .
[ platepartialtotal]d ) . the existence of an _ apparent _ contact line is only due to the effect of gravity : on a flat surface , the liquid would simply spread . the interface between the liquid and vapor phases consequently presents two regions . in the lower region ,
the meniscus can be described by the balance between the laplace pressure and the gravity potential @xmath96 where @xmath59 is the height above the bath ( thus , no additional constant needed ) and @xmath75 the curvature of the interface . introducing the capillary length @xmath97 this equation
can be written as : @xmath98 . in the upper region
, there is the pre - wetting film whose thickness @xmath99 is determined by the balance between the gravity potential and the disjoining pressure @xmath100 .
the disjoining pressure is by definition the potential energy per unit volume at the surface of a liquid layer of thickness @xmath101 , therefore the balance reads@xcite : @xmath102 as the pre - wetting film is very flat , the contribution of laplace pressure can be neglected ( @xmath103 ) in this regime .
the disjoining pressure scales as : @xmath104 for films where @xmath105 , where @xmath11 is a length on the order of the molecular size .
the disjoining pressure vanishes in the limit case @xmath106 , which can be interpreted as the situation for which the interaction is the same with the liquid and with the solid .
then , one indeed does not expect any influence of the thickness @xmath101 on the energy .
equating gravity and disjoining pressure ( eq . [ prewetpressurebalance ] ) , we obtain the thickness profile in the pre - wetting region : @xmath107^{1/3}\ ] ] in the vicinity of the apparent contact line , where the two zones must match , the thickness is thus of order @xmath108 . as @xmath109 is millimetric and @xmath11 nanometric , @xmath101 is mesoscopic ( @xmath110 nm ) . from the microscopic point of view , the solid is then completely surrounded by a semi - infinite layer of liquid ( @xmath111 )
. therefore , the only forces acting on a solid in complete wetting are normal contact forces , such as laplace pressures .
there are no contact line forces such as those described in fig .
[ forcesolid]b .
the forces exerted by the liquid on the solid are then related , as previously , to the curvature of the liquid - solid interface but also , inside the pre - wetting film , to the curvature of the liquid - vapor interface ( fig [ platepartialtotal]d ) .
integrated over the whole submerged solid , the curvature of the solid gives a zero resultant force , whereas the curvature of the liquid is only integrated where the pre - wetting film exists . as a result ,
the resultant force is vertical and has an amplitude @xmath5 per unit length of the apparent contact line .
this result is indeed consistent with the thermodynamic perspective .
since the solid is covered by a liquid layer much thicker than the molecular size the surface tension above the apparent contact line is not @xmath25 but @xmath112 : the plate is always completely submerged .
in essence , this means that the plate never leaves the liquid bath when the plate is pulled upwards .
when moving , there is no change of the solid - vapor interfacial area ( it remains zero ) or of the solid - liquid interfacial area ( which is simply the total area of the plate ) .
the only change occurs at the liquid - vapor interfacial area , which is increased , and the required pulling strength is thus @xmath5 per unit length of the apparent contact line .
in this article we have raised simple questions about capillarity that many students face . by studying the interfaces from a microscopic perspective
, we were able to provide answers to these questions , and also reconcile thermodynamics and statistical physics .
we have provided a mechanical perspective about why there exists an attractive force parallel to interfaces : surface tension .
the absence of liquid creates an attractive anisotropic force within a few molecular lengths from the interface whereas the repulsion remains isotropic and scales with the local density of the fluid .
the attractive anisotropy leads to a strong localized stress parallel to the interface called surface tension .
this occurs at liquid - solid interfaces as well , where there is also a half - space of liquid missing .
the problems when constructing force pictures in capillarity often arise from the improper definition of a _ system _ on which the forces act . considering a corner of liquid near the contact line as a system
, we have proposed here an alternative to young s construction ( fig .
[ young]b ) .
the analysis allows to locate and understand the different forces , in particular the attractive force exerted by the solid .
this new force constructions leads to a perfect mechanical equilibrium : vectorial force balance and torque balance .
when looking at the force that is exerted by the liquid on the solid near the contact line we find that , surprisingly , this force is _ not _
@xmath81 , but @xmath113
. moreover , a normal stress is exerted in all the regions of any curved solid - liquid interface : the liquid pulls the solid when the latter is convex .
this force is equivalent to the usual laplace pressure .
one has to take _ both _ these forces into account , to obtain the net force from the thermodynamic result .
the advantage of this microscopic force description is that it provides a simple answer to an academic problem which has become controversial in the last decade : the floating pin paradox@xcite .
the drawings and several relations presented in this article are based on results obtained using the density functional theory in the sharp - kink approximation@xcite , which allows to make quantitative predictions of the force distributions in the liquid and in the solid . .
we realize that the detailed picture of microscopic forces is not necessarily the most accessible for teaching purposes . in particular when introducing the basic concepts of capillarity , it is much simpler to work from the thermodynamic perspective : energy minimization naturally yields the equilibrium conditions , while the resultant forces can be computed from the virtual work principle .
nevertheless , our analysis provides a number of insights that are useful when teaching capillarity : * to determine capillary forces " it is crucial to explicitly specify the system (= collection of matter ) to which the forces are applied .
* the surface tensions @xmath26 and @xmath25 do not pull on the solid .
* the global force exerted on the solid by the liquid can be computed by adding the contributions of ( laplace ) pressure inside the liquid and a localized tension force @xmath5 parallel to the liquid - vapor interface .
while this gives the correct answer , it does not reflect the true microscopic distribution of liquid - on - solid forces .
* by contrast , the resultant force on the liquid near the contact line _ does _ involve the three surface tensions @xmath114 and @xmath25 . * the classical construction of fig .
[ young]b to explain young s law does not accurately represent the force balance .
a complete picture is provided in fig .
[ youngcorrect]a .
we hope in particular that the force construction of fig .
[ youngcorrect]a will find its way to the classroom to explain young s law ( [ young ] ) .
it is conceptually simple , it clarifies the system " to which forces are applied , and represents a perfect mechanical equilibrium .
namely , besides a balance of normal and tangential components , the forces also exert a zero torque .
finally , we note that the virtual work principle yields the correct resultant force on a solid , but that it can not recover the true microscopic force distribution .
the knowledge of such a force distribution is crucial when one wants to take into account how a solid is elastically deformed by the contact line@xcite .
even though these deformations can be as small as a few nanometers , these can be measured due to the advent of modern experimental techniques@xcite .
this experimental access renews the fundamental interest in the microscopic details of capillarity@xcite .
* acknowledgements : * we thank j. van honschoten and k. winkels for critically reading the manuscript . b.a .
thanks the students of the university paris - diderot s master of physics for both their impertinent and pertinent questions .
the example developed in the appendix was brought to us by y. forterre .
unlike liquids , solids are locally either in a repulsive state or an attractive state .
this is illustrated here by the contact between two adhesive solid elastic spheres @xcite .
( a ) geometry of the contact in the presence ( solid line ) or absence ( dashed line ) of adhesion .
( b ) radial distribution of normal force in the contact zone ] from the mechanical point of view , the surface tensions of liquids and solids are fundamentally different . to illustrate this difference , let us analyse the example of fig .
[ elasticspring ] , which shows two small adhesive solid spheres such as grains in a powder of flour .
one might imagine that the particles are attracted all over the contact , but in reality , there is also repulsion to give a zero resultant force .
although there is both attraction and repulsion in this example , the attractive and repulsive parts are spatially separated : attractive near the edge , repulsive in the central part of the contact . in a liquid , by contrast , the attraction and repulsion occur at the same location . | a paperclip can float on water .
drops of mercury refuse to spread on a surface .
these capillary phenomena are macroscopic manifestations of molecular interactions , and can be explained in terms of surface tension . for students ,
the concept of surface tension is quite challenging since the microscopic intuition is often in conflict with the common macroscopic interpretations . in this paper
we address a number of conceptual questions that are often encountered when teaching capillarity . by answering these questions
we provide a perspective that reconciles the macroscopic viewpoints , from thermodynamics or fluid mechanics , and the microscopic perspective from statistical physics . |
the phenomenon of electric conduction in metals and semiconductors has been a prominent research topic ever since the early days of solid - state physics .
the idea that electric currents flowing inside solid materials are effected by the transport of `` small '' charged particles ( `` charge carriers '' ) was first conceived by weber.@xcite following the discovery of the electron by thomson,@xcite weber s idea quickly found its concrete expression in attempts to understand electric ( as well as thermal ) conduction in solids as a manifestation of electron transport .
the basic concept for a theoretical treatment of conduction in terms of the motion of individual carriers was outlined by riecke.@xcite relying to some extent on this concept , drude@xcite formulated his celebrated transport model , which subsequently was refined by lorentz.@xcite in drude s model , the atomistic picture of matter and the kinetic theory of gases are combined to describe conduction in terms of a homogeneous gas of non - interacting , mobile charge carriers in thermodynamic equilibrium , which are assumed to move against a background of spatially fixed , heavy atoms . when an external electric field is applied , the interplay of field - induced acceleration and subsequent thermalizing collisions with the heavy atoms gives rise to a `` drift current '' of the carriers .
the magnitude of this current is determined by the `` mean free path '' , i.e. , the average distance the carriers travel between two collisions . for the drift
current to be a valid concept , the mean free path must be very small as compared with typical dimensions of the sample .
while originally conceived to describe carrier transport in metals , drude s model later on has been frequently used in qualitative , or semi - quantitative , analyses of transport properties of semiconductors as well . to overcome the shortcomings of drude s model in the quantitative description of carrier transport in semiconductors , particularly in inhomogeneously doped systems , the drift current was supplemented with a diffusion current,@xcite whereby
drude s model has been extended to the `` drift - diffusion model '' of transport .
the latter model , while representing a substantial improvement over drude s model and serving as a benchmark of semiclassical transport theories even in modern times , is again valid in the range of very small carrier mean free paths only . in the opposite case of very large mean free paths , carrier transport in semiconductors can be described in terms of the ballistic ( `` thermionic '' ) model,@xcite in which carriers thermally emitted at the ends of a sample are assumed to traverse it without collisions with the background atoms . until recently
, no systematic attempts had been made to bridge the gap between the limiting cases of the drift - diffusion and ballistic descriptions within a unified approach . in view of this situation
, we set out to develop the `` thermoballistic approach''@xcite to carrier transport in semiconductors .
apart from perceiving the challenge to fill a long - standing gap in the theory of carrier transport in semiconductors , we found that recent progress in device physics , spintronics , and photovoltaics called for an extension of the theoretical framework hitherto available for analyzing experimental data in these fields .
the physical concept underlying the thermoballistic approach rests on a random partitioning of the length of a semiconducting sample into `` ballistic transport intervals '' . here , `` random '' implies ( i ) an arbitrary number of intervals and ( ii ) arbitrary positions of the end - points , and thus arbitrary lengths , of the intervals .
[ the drift - diffusion model may be viewed as implying a partitioning of the sample length into infinitesimally short ballistic intervals ; in the ballistic transport model , on the other hand , the sample length constitutes a single ballistic interval . ]
an individual partition defines what we call a `` ballistic configuration '' .
the points linking adjacent intervals in a ballistic configuration are assumed to be points of local thermodynamic equilibrium characterized by a local chemical potential .
any such point acts both as a source of , and a sink for , carriers .
that is , on the one hand , carriers are thermally emitted there , with a velocity distribution determined by the local temperature ; subsequent to their emission , the carriers are ballistically transmitted across either interval to the left and right , facing , in general , potential energy barriers arising from the internal and external potentials inside the sample . on the other hand , carriers emitted at the two equilibrium points neighboring the point under consideration on either side and transmitted to it are `` absorbed '' there , i.e. , they return to thermodynamic equilibrium instantaneously .
this equilibration is assumed to result from collisions of the carriers with spatially fixed scattering centers ( `` impurities '' ) randomly distributed over the sample , a view adopted from drude s transport model . in the manner by which the ballistic intervals have been introduced ,
the lengths of these intervals are stochastic variables , with associated probabilities given by the probabilities for carriers to traverse an interval without impurity scattering .
these probabilities are governed by the carrier mean free path , which is allowed to have arbitrary magnitude . by averaging the ( in general , spin - dependent ) carrier currents in the individual ballistic intervals over all ballistic configurations , a position - dependent total ( i.e.
, spin - summed ) thermoballistic current as well as a thermoballistic spin - polarized current are derived .
these currents , in conjunction with the associated ballistic densities , represent the key elements of the thermoballistic approach .
they form the point of departure for the calculation of the spin - resolved equilibrium chemical potentials . from the latter functions , in turn ,
transport quantities , like the current - voltage characteristic , the magnetoresistance , and ( position - dependent ) current and density spin polarizations , are obtained in terms of the potential energy profile , the mean free path , the ballistic spin relaxation length , and other parameters characterizing the sample . the present article is devoted to a comprehensive exposition of the physical concept underlying the thermoballistic approach and its detailed implementation .
we begin by describing a prototype model of thermoballistic transport ( the `` prototype model '' , for short ) .
this model introduces averaging over ballistic configurations as the constitutive element of the thermoballistic concept .
it is based on the simplifying assumption that the carrier current is conserved across the points of local thermodynamic equilibrium linking the ballistic transport intervals .
this assumption allows the derivation of an explicit , transparent expression for the current - voltage characteristic , as well as the construction of a local chemical potential in a heuristic form . after describing the prototype model
, we present a general formulation of the thermoballistic concept proper , which makes use of a local chemical potential as the essential dynamical quantity , and takes into account arbitrarily shaped , spin - split potential energy profiles as well as spin relaxation during ballistic carrier motion . throughout this article ,
the formulation deviates in many respects from that in our original publications .
apart from correcting flaws and inconsistencies , we introduce modifications and extensions of the formalism that make it more transparent and more comprehensible . aiming at a self - contained , unified presentation , we will develop the elements of the thermoballistic concept in considerable detail .
this includes a recollection of early attempts to describe charge carrier transport in semiconductors and , at one place or another , the coherent recapitulation of background material which otherwise can be found only scattered over textbooks .
while the thermoballistic approach , in its most general form , would allow a consistent treatment of three - dimensional , bipolar carrier transport in semiconducting systems , we confine ourselves here to the case of one - dimensional , unipolar transport throughout . to be specific , we consider electron transport in the conduction band of n - doped systems , with the understanding that all results for hole transport in the valence band of p - doped systems can be obtained by transcribing the results for electron transport in an obvious way .
in implementing the thermoballistic approach , we aim at carrying the development to the point where we obtain explicit equations from which the relevant physical quantities can be calculated .
we do not enter here into applications requiring numerical calculations .
pertinent results presented in our previous publications will be briefly mentioned at various places in this article .
the organization of this article is as follows . in the next section ,
we present an account of drude s transport model and the standard drift - diffusion and ballistic models . in sec .
[ sec : gendrift ] , we begin with an outline of the probabilistic approach to carrier transport , which is prerequisite to the formulation of the thermoballistic model .
this is followed by the description of the prototype model . in sec .
[ sec : therconc ] , the thermoballistic concept proper is developed at length , where particular emphasis is placed on the inclusion of spin degrees of freedom . the spin - resolved ballistic currents and associated densities
are introduced in terms of an `` average chemical potential '' and a `` spin accumulation function '' connected with the splitting of the spin - resolved chemical potentials .
total and spin - polarized thermoballistic currents and densities are constructed by averaging the corresponding ballistic quantities over the ballistic configurations , and are evaluated in the drift - diffusion regime and in the ballistic limit .
energy dissipation ( `` heat production '' ) is analyzed within the thermoballistic approach . in sec .
[ sec : thersyno ] , in particular , a synopsis of the thermoballistic concept is presented , with emphasis on its physical content as well as its merits and weaknesses . the procedures devised for the implementation of the thermoballistic concept , i.e. , for the explicit calculation of the average chemical potential and of the spin accumulation function as functions of intrinsic and external physical parameters ,
are described in sec .
[ sec : therimpl ] .
expressions are derived for the current and density spin polarizations and the magnetoresistance in terms of the values of the spin accumulation function at the boundaries of a semiconducting sample . as specific examples of current interest in semiconductor and spintronics
research , we treat spin - polarized transport in heterostructures formed of a nonmagnetic semiconductor and two ferromagnetic contacts , and spin - polarized transport in heterostructures involving diluted magnetic semiconductors in their paramagnetic phase . finally , in sec . [
sec : summconc ] , we summarize the contents of this article and give an outlook towards future developments .
as the thermoballistic approach is devised to bridge the gap between the drift - diffusion and ballistic descriptions of transport , we survey , in this section , the standard formulations of these two limiting cases .
the drift - diffusion model is an extension of the transport model of drude , so that we begin here with an account of the latter .
drude s model is far from being able to describe transport properties of semiconductors quantitatively ; nevertheless , its exposition provides us with the opportunity to introduce and discuss the basic notions on which classical and semiclassical transport theories rely , and which will appear ubiquitously throughout this article . in the transport model of drude,@xcite
one considers a ( three - dimensional ) homogeneous classical gas of non - interacting electrons in thermodynamic equilibrium at temperature @xmath0 , which is subjected to an externally applied , constant electric field @xmath1 .
the electrons collide with randomly distributed , spatially fixed scattering centers ( `` impurities '' ) .
the collisions are assumed to cause the electrons to return instantaneously to complete thermodynamic equilibrium .
the average vectorial velocity of the electrons emerging from this equilibration is equal to zero . their mean speed @xmath2 , i.e.
, the thermal average of the magnitude of the electron velocity as derived from the three - dimensional maxwell - boltzmann velocity distribution function , is given by @xmath3 where @xmath4 is the effective electron mass , and @xmath5 , with @xmath6 the boltzmann constant .
[ note that in drude s original paper,@xcite the mean electron speed was derived from boltzmann s equipartition theorem ; the maxwell - boltzmann velocity distribution was introduced into the description of electron transport in metals by lorentz.@xcite ] defining the _ mean free path _ ( or _ momentum relaxation length _ ) @xmath7 as the average distance ( measured along the transport direction ) travelled by the electrons between two collisions , one finds for the _ collision time _ ( or _ relaxation time _ ) @xmath8 , defined as the average time - of - flight between successive collisions , @xmath9 the mean free path @xmath7 characterizes a set of collision - free _ ballistic intervals _ of average length @xmath7 , across which electrons move ballistically under the influence of the external field @xmath1 . on average ,
they acquire a velocity , the _ drift velocity _ @xmath10 , given by the acceleration @xmath11 times their average time - of - flight across the ballistic intervals , i.e. , times the collision time @xmath8 , @xmath12 where @xmath13 is the electron mobility . in the picture of drude ,
the charge current density @xmath14 ( current , for short ) is a _ drift current _ , @xmath15 , driven by the external electric field @xmath1 , @xmath16 where @xmath17 is the electron density . since , in drude s model , @xmath17 as well as the field @xmath1 are independent of position , @xmath18 is constant . in terms of the electron conductivity @xmath19 the current @xmath14
is expressed as @xmath20 i.e. , in the form of ohm s law . the transport mechanism in drude s model
can be elucidated@xcite by noting that eq .
( [ eq : drude003 ] ) , when written as @xmath21 states that the effect of the external force @xmath22 is balanced , on average , by that of a _ friction force _ @xmath23 , with the friction coefficient @xmath24 given by @xmath25 the friction force reflects the reset to zero of their individual velocities when , subsequent to their acceleration through the external field , the electrons are thermally equilibrated due to impurity scattering .
limits on the range of validity of drude s model are set by the requirement that the perturbation of the electron gas due to the external field is sufficiently small so that the gas stays close to thermodynamic equilibrium .
this condition can be met by requiring the magnitude of the drift velocity to be very small as compared with the mean speed of the electrons , @xmath26 using eqs .
( [ eq : drude001 ] ) , ( [ eq : drude003 ] ) , and ( [ eq : drude004 ] ) , condition ( [ eq : drude004c ] ) can be re - expressed in a form that restricts , for given mean free path @xmath7 , the magnitude of the electric - field vector @xmath1 , @xmath27 for a homogeneous sample of length @xmath28 subjected to an external voltage bias @xmath29 , when @xmath30 , one has @xmath31 reversely , for given voltage bias @xmath29 , the ratio of mean free path to sample length is restricted by the condition @xmath32 for room temperature , i.e. , @xmath33 ev , and values of @xmath34 of a few tens of mev , which are typical for semiconducting devices , the right - hand side of condition ( [ eq : drude004e ] ) is of order unity .
hence , one may use the condition @xmath35 as a rough criterion for delimiting the range of validity of the drude model .
the value of @xmath7 remains small if the density of impurities is sufficiently high .
electron transport in inhomogeneous semiconductors is outside of the scope of drude s model .
the nonuniformity of the donor density in inhomogeneous samples entails a position dependence of the electron density , @xmath36 , and gives rise to an @xmath37-dependent internal ( `` built - in '' ) electrostatic potential.@xcite [ throughout this article , we consider _ one - dimensional _ transport in _ three - dimensional _ , `` plane - parallel '' semiconducting samples , i.e. , samples whose parameters do not vary in the directions perpendicular to the transport direction , which is taken as the @xmath37-direction ( for a discussion of this assumption , see sec .
[ sec : thersyno ] ) .
further , the temperature @xmath0 is assumed to be constant across the sample . ]
transport in inhomogeneous systems near thermodynamic equilibrium can be described in terms of a _ local _ thermodynamic equilibrium@xcite characterized by a _ local chemical potential _
@xmath38.@xcite disregarding spin degrees of freedom and adopting the effective - mass approximation,@xcite we have for the equilibrium electron density @xmath39 in a nondegenerate system @xmath40 here , @xmath41 is planck s constant , and we have introduced cylindrical coordinates in three - dimensional velocity space ( with cartesian coordinates @xmath42 ) such that @xmath43 with @xmath44 .
the function @xmath45 is the maxwell - boltzmann energy distribution function , and @xmath46 is the total electronic energy at the equilibrium point @xmath37 , with the kinetic energy @xmath47 the potential energy profile @xmath48 comprises the conduction band edge potential ( which includes the position - dependent internal potential ) and the external electrostatic potential . in general
, @xmath48 exhibits a `` multiple - barrier '' shape associated with local maxima of the profile .
evaluation of the integrals in eq .
( [ eq : drude102 ] ) results in the standard form for the density , @xmath49 } , \label{eq : drude014}\ ] ] where @xmath50 is the effective density of states at the conduction band edge.@xcite the effects of the spatial variation of the electron density and of the associated occurrence of an internal potential in inhomogeneous semiconductors can be described within an extension of drude s model , the _ drift - diffusion model_. in this model , a generalized , @xmath37-dependent drift current @xmath51 is supplemented@xcite by a _ diffusion current _ @xmath52 proportional to the density gradient along the @xmath37-axis , so that the total ( conserved ) current @xmath53 is given by @xmath54 the generalized drift current is obtained from expression ( [ eq : drude005 ] ) by replacing ( i ) the conductivity @xmath55 with the _ local conductivity _
@xmath56 given by eq .
( [ eq : drude006 ] ) , with @xmath39 in lieu of @xmath17 , and ( ii ) the constant external electric field @xmath1 with the field @xmath57/e$ ] associated with the potential energy profile @xmath48 , so that one obtains @xmath58 the diffusion current is written as @xmath59 where @xmath60 is the diffusion coefficient , which is related to the electron mobility @xmath61 via the einstein relation@xcite @xmath62 and @xmath61 is now assumed to have the form corresponding to one - dimensional transport , @xmath63 [ see eq .
( [ eq : drude004 ] ) for the three - dimensional form of @xmath61 ] . then , generalizing eq .
( [ eq : drude006 ] ) , one has @xmath65 for the total current @xmath53 , one now finds from eqs .
( [ eq : drude008f])([eq : drude007 ] ) , using eq .
( [ eq : drude006f ] ) , @xmath66 , \label{eq : drude008}\ ] ] which is the standard drift - diffusion expression for the total charge current . for the derivation of this expression from boltzmann s transport equation and for its application in device simulation , see , e.g. , ref . . in ref
. , the expression is derived within a time - dependent tutorial treatment of diffusion in the presence of an external electric field ( `` biased - random - walk model '' ) . substituting expression ( [ eq : drude014 ] ) for the equilibrium density @xmath39 in eq .
( [ eq : drude008 ] ) , we obtain the current @xmath53 in the form @xmath67 which shows that , in the drift - diffusion model , it is the local chemical potential which provides the driving force for electron transport .
using eqs .
( [ eq : drude014 ] ) and ( [ eq : drude006f ] ) , we can rewrite eq .
( [ eq : diff001 ] ) as @xmath68 considering a sample extending from @xmath69 to @xmath70 and integrating eq .
( [ eq : diff002 ] ) over the interval @xmath71 $ ] , we can express the local chemical potential @xmath38 in the form @xmath72 an equivalent representation of @xmath38 is obtained by integrating eq .
( [ eq : diff002 ] ) over the interval @xmath73 $ ] . now , setting @xmath74 in eq .
( [ eq : diff003 ] ) , identifying the chemical potentials at the end - points @xmath75 with the potentials @xmath76 in the contacts connected to the semiconducting sample , @xmath77 and solving the resulting equation for the current @xmath53 , we then obtain , using eq . ( [ eq : drude014 ] ) , the current - voltage characteristic of the drift - diffusion model in the form @xmath78 here , @xmath79 is the voltage bias , and @xmath80 is the maximum barrier height of the potential energy profile relative to its value @xmath81 at the left end of the sample , where @xmath82 is the overall maximum of @xmath48 in the interval @xmath83 $ ] .
finally , the quantity @xmath84 } \label{eq : diff008}\ ] ] is the `` effective sample length '' .
it has the appealing property of becoming equal to the sample length @xmath85 for a flat profile , i.e. , if @xmath86 , and otherwise satisfies @xmath87 . writing the current - voltage characteristic in the particular form ( [ eq : diff006 ] ) facilitates comparison with analogous expressions given below . expression ( [ eq : diff006 ] ) shows that the current - voltage characteristic of the drift - diffusion model is controlled ( i ) by the `` barrier factor '' @xmath88 , which involves the overall maximum of the profile @xmath48 , and ( ii ) by the ratio @xmath89 or , owing to eq .
( [ eq : drude004hg ] ) , by the ratio @xmath90 , in which the effective sample length reflects , in an integral way , the shape of @xmath48 .
[ in the ballistic description of transport , the barrier factor @xmath91 re - appears as the thermally averaged probability for electron transmission from @xmath69 to @xmath70 ; see eq .
( [ eq : ball005 ] ) below . ]
the drift - diffusion model is based on the assumption of a continuously varying equilibrium chemical potential @xmath38 , which implies that the points of local thermodynamic equilibrium lie arbitrarily dense .
then , strictly speaking , the mean free path @xmath7 must be confined to arbitrarily small values . on the other hand , @xmath7 ( or the mobility @xmath61 ) must be finite and large enough to give rise to a non - vanishing conductivity of a magnitude in the range of typical experimental values , which calls for a relaxation of the former condition . to obtain a practical criterion for
the range of validity of the drift - diffusion model , one may require @xmath7 to be so small that the effective number of points of local thermodynamic equilibrium along the length of the sample is so large that the spatial variations in the potential energy profile @xmath48 , and hence in the electron density @xmath39 , are `` resolved '' with sufficient accuracy . in terms of a local
, @xmath37-dependent mean free path @xmath92 , this requirement may be expressed as @xmath93 where @xmath94 is the length of an interval , centered at the point @xmath37 , over which the relative variation of @xmath48 is very small compared to unity , i.e. , @xmath95 for constant mean free path @xmath7 , one may fulfil the above requirement in an overall way by adopting the condition @xmath96 the effective sample length @xmath97 defined by the integral ( [ eq : diff008 ] ) tends to decrease exponentially when strong variations in @xmath48 are `` switched on '' . condition ( [ eq : diff009 ] ) tightens condition ( [ eq : drude004f ] ) so as to permit the `` resolution '' of the details of the profile @xmath48 .
in contrast to the drift - diffusion model , in which the points of local thermodynamic equilibrium are assumed to lie arbitrarily dense , the ballistic ( thermionic ) transport model presupposes the complete absence of such points inside the sample .
then , the electrons in the sample perform a collision - free _ ballistic motion _ in the field associated with the potential energy @xmath48 . without thermal equilibration , it is meaningless to speak of a local chemical potential . only at the sample
ends at @xmath75 , the electrons are forced into equilibrium , with densities @xmath98 } \label{eq : ball000}\ ] ] determined by the boundary values of the potential energy profile , @xmath99 , and by the chemical potentials @xmath76 in the contacts . here
, the mean free path @xmath7 , which has been of central importance in drude s model and in the drift - diffusion model , is effectively of infinite length and does not appear in the formalism of the ballistic model .
roughly speaking , one may use the condition @xmath100 to delimit the range of validity of the ballistic model . in the ballistic model ,
the end - points @xmath75 of the sample are fixed points of local thermodynamic equilibrium with chemical potentials @xmath101 , out of which thermal electron currents are symmetrically emitted towards the left and right , so that only
_ one half _ of each of these currents are emitted towards the inner region of the sample . for a nondegenerate system , the classical _ electron current _
@xmath102 emitted at the left end - point @xmath69 towards the right , say , is expressed , in extension of the electron density @xmath39 given by eq .
( [ eq : drude102 ] ) , in the form @xmath103 from which we find , using eq .
( [ eq : ball000 ] ) , @xmath104 } = v_{e } n(x_{1 } ) .
\label{eq : ball001}\ ] ] here , @xmath105 is the _ emission velocity _ , which is actually equal to one half of the one - dimensional mean electron speed @xmath106 the three - dimensional analogue of which is given by eq .
( [ eq : drude001 ] ) .
the electrons emitted at @xmath69 with velocity component @xmath107 move along ballistic trajectories , thereby conserving their total energy , @xmath108 where @xmath37 is any point inside the interval @xmath83 $ ] ( and is not a point of local thermodynamic equilibrium ) .
the energy distribution at @xmath37 is therefore equal to that at @xmath69 .
however , only electrons with total energy larger than the overall maximum @xmath82 of @xmath48 [ see eq .
( [ eq : diff007 ] ) ] are classically able to reach the right end - point of @xmath83 $ ] at @xmath70 .
thus , part of the current @xmath102 emitted at @xmath69 will be reflected , and the electrons forming it are absorbed when they return to their origin at @xmath69 .
the other , transmitted part @xmath109 , called the ( left ) `` ballistic current '' , is absorbed into the contact connected to the sample at @xmath70 . modifying expression ( [ eq : ball001h ] ) , we have for this part @xmath110 in the integration , the potential energy profile @xmath48 contained in the function @xmath111 drops out , and we obtain for the ( left ) ballistic current the @xmath37-independent expression @xmath112 } , \label{eq : ball004f}\ ] ] i.e. , the ballistic current is conserved ( independent of @xmath37 ) , as expected .
the ( right ) ballistic current @xmath113 transmitted from the _ right _ end - point @xmath70 of the interval @xmath83 $ ] is given by @xmath114 } , \label{eq : ball006}\ ] ] in analogy to eq .
( [ eq : ball004f ] ) .
the ballistic currents ( [ eq : ball004f ] ) and ( [ eq : ball006 ] ) bear a close analogy to the `` thermionic emission current '' associated with the evaporation of electrons from a heated metal ( `` richardson effect'').@xcite in semiconductor physics , `` thermionic emission '' was introduced as a mechanism of carrier transport by bethe@xcite in his treatment of electron transport across a schottky barrier .
associated with the ballistic currents @xmath115 are the `` ballistic densities '' @xmath116 of the electrons making up the currents inside the ballistic interval @xmath83 $ ] .
these densities will turn out to be instrumental in establishing the spin - dependent thermoballistic scheme ( see sec .
[ sec : balltrans ] ) . the density @xmath117 associated with the current @xmath109 of eq .
( [ eq : ball004 ] ) is given by @xmath118 which is evaluated to yield @xmath119 } .
\label{eq : ball004ka}\end{aligned}\ ] ] here , @xmath120 } \nonumber \\ & \ & \times { \rm erfc } \bm { ( } \ { \beta [ e_{c}^{m}(x_{1 } , x_{2 } ) - e_{c}(x ) ] \}^{1/2 } \bm { ) } , \nonumber \\ \label{eq : ball004kaa}\end{aligned}\ ] ] where the function @xmath121 is the complementary error function.@xcite the ballistic density is position - dependent via the @xmath37-dependence of the function @xmath122 , i.e. , of the potential energy profile @xmath48 .
the ballistic density @xmath123 associated with the current @xmath113 is obtained by replacing @xmath124 with @xmath125 in expression ( [ eq : ball004ka ] ) .
the function @xmath122 determines the `` ballistic velocities '' @xmath126 which have the same magnitude for the currents transmitted from the left and right . for constant potential energy profile
, one has @xmath127 , and the electrons move with speed @xmath128 , i.e , with the mean electron speed @xmath2 given by eq .
( [ eq : ball002hvl ] ) . for position - dependent profiles ,
when @xmath129 , the magnitude of the ballistic velocities is larger than @xmath2 . the _ net ballistic current _
@xmath130 in the interval @xmath131 $ ] , @xmath132 equals the ( conserved ) total current @xmath133 , which we can express , using eqs .
( [ eq : ball004f ] ) and ( [ eq : ball006 ] ) , as @xmath134 this can be rewritten , using eqs .
( [ eq : diff004 ] ) and ( [ eq : ball000 ] ) , in the form @xmath135 here , @xmath136 with @xmath137 given by eq .
( [ eq : diff007 ] ) , is the thermal average of the classical transmission probability @xmath138 for electrons emitted at @xmath69 with total energy @xmath139 to be transmitted to the point @xmath70 .
if , in particular , the potential energy profile is constant across the interval @xmath83 $ ] , or if its maximum lies at the emission point @xmath69 itself , then @xmath140 in eq . (
[ eq : diff007 ] ) , and hence @xmath141 .
relation ( [ eq : ball009 ] ) is the current - voltage characteristic of the ( classical ) ballistic transport model .
in contrast to the characteristic ( [ eq : diff006 ] ) of the drift - diffusion model , which involves the mean free path @xmath7 and the potential energy profile @xmath48 [ via the effective sample length @xmath97 ] , the characteristic ( [ eq : ball009 ] ) is controlled by one `` material parameter '' only , _ viz .
_ , the thermally averaged transmission probability @xmath143 . for the _ joint ballistic density _
@xmath144 , @xmath145 we have @xmath146 in analogy to expression ( [ eq : ball008 ] ) for the net ballistic current . the ballistic transport model is straightforwardly extended so as to include _ electron tunneling _ by replacing the classical transmission probability @xmath147 of eq .
( [ eq : ball005a ] ) with the corresponding _ quantal _
probability @xmath148 .
the thermally averaged quantal transmission probability is then , in generalization of expression ( [ eq : ball005 ] ) , given by @xmath149 where @xmath150 the probability @xmath148 is obtained by solving the stationary schrdinger equation with the potential energy function @xmath48 . [ owing to time reversal invariance , the probability for transmission from the left equals that for transmission from the right . ] the integration in eq .
( [ eq : ball010 ] ) starts at the total energy @xmath151 , so that scattering boundary conditions can be imposed on the wavefunction both in the ranges @xmath152 and @xmath153 .
in wkb approximation,@xcite the transmission probability @xmath154 to be used in eq .
( [ eq : ball010 ] ) is composed of the classical ( over - barrier " ) part @xmath147 given by eq .
( [ eq : ball005a ] ) and the remaining quantal ( `` sub - barrier '' ) part @xmath155 , @xmath156 the sub - barrier contribution has the form @xmath157 where @xmath158 with @xmath159 \}^{1/2 } \theta \bm { ( } e_{c}(x ) - e \bm { ) } , \label{eq : ball012g}\ ] ] is the barrier penetration factor . in writing @xmath160 in the form ( [ eq : ball012 ] ) , we disregard resonance effects that may occur when @xmath48 exhibits two or more local maxima in the interval @xmath83 $ ] , with a corresponding number of one or more minima in between .
then , when the energy @xmath161 is located below the second - highest maximum and above the lowest minimum , there is at least one `` valley '' in @xmath48 , across which the eletron motion is classically allowed , so that resonance formation due to quantum coherence becomes possible . in semiconductor physics ,
a concrete realization of this situation occurs in resonant tunneling in multiple - barrier quantum - well structures.@xcite for this case , the full wkb tunneling probability for double - barrier and triple - barrier structures , respectively , has been presented in ref . . from eq .
( [ eq : ball011 ] ) , we now find for the wkb form of the thermally averaged transmission probability @xmath162 of eq .
( [ eq : ball010 ] ) @xmath163 here , we have @xmath164 with @xmath165 for the over - barrier contribution , and @xmath166 for the sub - barrier contribution .
the thermally averaged quantal probability @xmath167 for transmission from the left end - point at @xmath69 can be expressed in terms of @xmath168 , using eqs .
( [ eq : diff007 ] ) and ( [ eq : ball003a ] ) , as @xmath169 } .
\label{eq : ball006b}\ ] ] in analogy to eq .
( [ eq : ball009 ] ) for the classical case , we have @xmath170 for the current - voltage characteristic of tunneling - enhanced ballistic transport . in the degenerate case ,
when the electron system obeys fermi - dirac statistics , we write the ballistic current @xmath171 in the form [ see eqs .
( [ eq : ball001h ] ) and ( [ eq : ball004 ] ) for the nondegenerate case ] @xmath172 where @xmath173 is the fermi - dirac energy distribution function , @xmath174 expression ( [ eq : degen040 ] ) for the current @xmath175 formally agrees with the expression for the current of evaporated electrons encountered in the degenerate treatment of the richardson effect.@xcite then , following the procedure of ref .
, we can reduce the threefold integration in eq .
( [ eq : degen040 ] ) to a single integration over the kinetic energy @xmath176 , obtaining @xmath177 the ballistic current @xmath113 transmitted from the right end - point of the sample at @xmath70 is the negative of expression ( [ eq : degen010 ] ) , with @xmath125 substituted for @xmath124 .
the total current @xmath133 [ see eq .
( [ eq : ball007 ] ) ] is thus obtained as @xmath178 \nonumber \\ & \ & \hspace{3.0 cm } \times \theta \bm { ( } \epsilon - e_{c}^{m } ( x_{1 } , x_{2 } ) \bm { ) } .
\label{eq : degen011e}\end{aligned}\ ] ] here , we have expressed the factor preceding the integral in eq .
( [ eq : degen010 ] ) in terms of the emission velocity , @xmath179 , and the effective density of states , @xmath180 , which are nondegenerate quantities given by eqs .
( [ eq : ball002 ] ) and ( [ eq : diff001a ] ) , respectively . expression ( [ eq : degen011e ] ) for the total currrent @xmath133 is the degenerate analogue to the nondegenerate current - voltage characteristic ( [ eq : ball008 ] ) . for zero bias , the chemical potentials at the sample ends , @xmath124 and @xmath125 , differ only by an infinitesimal @xmath181 , @xmath182 then , expanding the right - hand side of eq .
( [ eq : degen011e ] ) to first order in @xmath181 , we find @xmath183_{\mu = \mu_{1 } } \nonumber \\ & = & v_{e } n_{c } \beta \delta \ln \bm { ( } 1 + e^{-\beta[e_{c}^{m}(x_{1 } , x_{2 } ) - \mu_{1 } ] } \bm { ) } .
\label{eq : degen001}\end{aligned}\ ] ] setting @xmath184 , we now have for the zero - bias conductance per unit area in the ballistic transport model @xmath185 } \bm { ) } .
\label{eq : degen009}\end{aligned}\ ] ] in highly doped , degenerate semiconductors , we may have @xmath186 ( see , e.g. , ref . , where grain - boundary - limited transport in polycrystalline materials is considered ) .
then , if @xmath187 \gg 1 , \label{eq : degen009ofd}\ ] ] we find from eq .
( [ eq : degen009 ] ) @xmath188 , \label{eq : degen009ze}\ ] ] which equals the conductance of a ballistic point contact.@xcite in closing this subsection , we note that the ballistic transport model does not provide information on where the resistance causing the voltage drop is located along the sample . evidently , it can not be inside the collision - free sample . in the quantal description of ballistic electron transport in mesoscopic systems as formulated by landauer,@xcite the resistance is made up solely of the _ interface resistances _ arising from the abrupt change in the density of states ( `` transverse modes '' ) that the electrons encounter when they move across the interfaces separating the contacts ( with infinitely many modes ) from the sample ( with a few modes only ) .
the voltage drop is located , therefore , in the immediate vicinity of the interfaces , so that , when a chemical potential is introduced _ ad hoc _ , this must be constant inside the sample and discontinuous at the interfaces . prior to the work of landauer ,
the importance of interface resistances in ballistic transport had been emphasized by sharvin.@xcite anticipating the later development , we remark at this point that in the thermoballistic approach , i.e. , for _ finite _ magnitude of the mean free path , the ( local ) equilibrium chemical potential is a constitutive element of the transport mechanism ; it is defined , and can be explicitly calculated , all along a semiconducting sample .
this potential has discontinuities at the contact - sample interfaces , whose magnitude increases from near - zero in the small-@xmath7 , drift - diffusion regime to the sharvin value in the large-@xmath7 , ballistic regime . for more details , see sec .
[ sec : sharvin ] .
in the drift - diffusion and ballistic transport models , the parameter of central importance is the electron mean free path @xmath7 or , equivalently , the collision time @xmath8 originally introduced in drude s model .
the collision time represents the average time - of - flight that elapses between successive electron collisions with the randomly distributed scattering centers in the sample .
the drift - diffusion and ballistic transport mechanisms are limiting cases associated with very small and very large ( effectively , infinite ) collision times , respectively . in the present section , we proceed to the `` prototype thermoballistic model'',@xcite which represents an attempt to unify the drift - diffusion and ballistic models within a stationary ( time - independent ) description in which the mean free path is not limited in magnitude and enters as a parameter that controls collision _
probabilities_. [ in our original paper,@xcite
we have called this model the `` generalized drude model '' ; the new name appears to give a better description of its features . ]
we begin this section by introducing the probabilistic definition of @xmath7 ( or @xmath8 ) . in the time - dependent probabilistic approach ,
one introduces the probability @xmath189 that an electron undergoes a collision , i.e. , is equilibrized with its surroundings , within unit time .
that is , @xmath190 is the probability for the collision to occur within an infinitesimally short time interval @xmath191 .
then , assuming the collision time @xmath8 to be constant , the probability @xmath192 that an electron moves without collision over a finite time interval @xmath193 is given@xcite by @xmath194 and the conditional probability @xmath195 that an electron undergoes a collision within the time interval @xmath191 after a collision - free flight over a time interval @xmath193 , by @xmath196 the mean time - of - flight between successive collisions now becomes @xmath197 identifying @xmath198 with the collision time @xmath8 of the drude model , one sees that the inverse of this quantity is just equal to the collision probability per unit time , @xmath199 , of the probabilistic approach .
we recall that the term `` collision '' used here is that proffered by drude , meaning a collision leading to instantaneous , complete equilibration of the momentum of an electron with its surroundings . in that picture
, @xmath8 is the average time of collision - free flight between points of equilibration , hence the appellation .
however , @xmath189 is , more generally , the probability for a _ complete equilibration _ to occur in unit time ; this may happen as a consequence of a `` complete collision '' ( in the drude sense ) after the average time @xmath8 , or by `` incomplete collisions '' ( leading to incomplete equilibration ) in a sequence of shorter collision times , which add up , on average , to the time @xmath8 . within the time - dependent probabilistic picture
, one can set up@xcite an expression for the net electron current @xmath200 at position @xmath37 and time @xmath193 , which embodies features that are relevant for devising the thermoballistic concept ( see sec .
[ sec : thermcurr ] ) . assuming a sample with position - dependent electron density @xmath39
, one writes @xmath201 considering for the moment electrons with _ arbitrary , constant _
velocity @xmath202 , the currents @xmath203 are expressed , using eqs .
( [ eq : prob001 ] ) and ( [ eq : prob002 ] ) , as @xmath204 the contributions to these currents for fixed time @xmath193 arise from electrons that are transmitted , subsequent to their emission at the points @xmath205 [ lying to the left ( right ) of the point @xmath37 ] , without collision from @xmath207 @xmath208 to the point @xmath37 .
the intervals @xmath209 $ ] and @xmath210 $ ] define `` ballistic intervals '' ; the currents transmitted across these intervals contribute to the total current with weight @xmath211 .
note that , in contrast to the situation considered above , where ballistic motion occurs _ before _ equilibration in the time interval @xmath191 ( with probability @xmath212 ) , here the ballistic motion occurs _
after _ the electrons have been equilibrated in the time interval @xmath213 ( with probability @xmath214 ) . when the collision time @xmath8 is sufficiently small , such that @xmath215 is much smaller than the average length over which the density @xmath39 changes appreciably , one has , by expanding in expressions ( [ eq : prob004 ] ) @xmath39 to first order , @xmath216 taking into account that only electrons with @xmath217 ( @xmath218 ) contribute to the current arriving at the point @xmath37 from the left ( right ) , _ thermal averaging _ of @xmath219 now yields @xmath220 so that one obtains the local expression @xmath221 for the thermally averaged electron current .
the charge current @xmath222 is then found , in view of expression ( [ eq : drude011 ] ) for the diffusion coefficient @xmath60 , to agree with relation ( [ eq : drude007 ] ) for the diffusion ( charge ) current @xmath52 of the drift - diffusion model .
this result may thus be regarded as a proof of the einstein relation ( [ eq : drude011 ] ) .
when @xmath223 , only electrons emitted at the left and right end , respectively , of the sample contribute to the currents @xmath203 . then , the ( thermally averaged ) net electron current @xmath200 essentially reduces to the ballistic current given by expression ( [ eq : ball008 ] ) . in the intermediate regime ,
i.e. , for nonzero , finite values of @xmath8 , @xmath200 combines elements of both drift - diffusion and ballistic transport . in analogy
to the time - dependent probability @xmath192 given by eq .
( [ eq : prob001 ] ) , one can introduce the probability @xmath224 for an electron to travel without collision over a finite distance @xmath37 , @xmath225 with a constant electron mean free path @xmath7 . the conditional probability for an electron to undergo a collision in the infinitesimal collision interval @xmath226 after a collision - free flight over the distance @xmath37 reads @xmath227 then one finds @xmath228 for the mean distance @xmath229 between successive electron collisions . in the following ,
the mean free path @xmath7 will be assumed to take on any finite value , while , as before , the collisions are assumed to lead instantaneously to a complete equilibration of the electron momenta , and are to be interpreted as a simulation of the real situation where the distances between successive collisions are short ( of atomic dimensions ) , while the equilibration during a collision is ( usually ) far less than complete . the quantity @xmath7 , in parallel to the collision time @xmath8 , is simply to be regarded as a parameter which determines the probability ( [ eq : prob001a ] ) of collision - free ( complete or not ) traversal by an electron of the distance @xmath37 .
it is noted here that when treating inhomogeneous systems , one should , strictly speaking , allow for a position dependence of the collision time , @xmath230 , and of the mean free path , @xmath231 . in that case ,
expressions ( [ eq : prob001 ] ) and ( [ eq : prob001a ] ) for the probabilities @xmath192 and @xmath224 , respectively , must be replaced with the more general forms @xmath232 and @xmath233 respectively .
then , of course , the relation @xmath234 [ see eq .
( [ eq : drude002 ] ) ] ceases to be valid .
the unification of the drift - diffusion and ballistic transport models within the prototype thermoballistic model@xcite is based on the introduction of configurations of ballistic transport intervals , which cover the length of the sample .
the individual ballistic intervals in a configuration are linked by points of local thermodynamic equilibrium , and their lengths are stochastic variables occurring with probabilities determined by the probabilities for collision - free electron motion across the intervals , where the electron mean free path is allowed to have _ arbitrary _ magnitude .
the electron current is , of course , conserved in each ballistic interval .
in addition , it is assumed that _ the current is conserved also from one interval to the next_. averaging over the ballistic configurations then leads to a transparent and intuitively appealing expression for the current voltage - characteristic in terms of an effective transport length , in which the detailed shape of the potential energy profile is taken into account .
moreover , a local chemical potential can be constructed in a heuristic way . here , as well as in sec .
[ sec : averball ] , we confine ourselves to ( one - dimensional ) classical transport in nondegenerate systems .
the effects of tunneling and degeneracy will be considered in secs .
[ sec : quantum ] and [ sec : degener1 ] , respectively . exhibiting a single barrier with maximum at @xmath235 .
the partitioning of the sample length @xmath28 into @xmath236 ballistic intervals @xmath237 $ ] ( @xmath238 ) is indicated , where @xmath239 is the maximum of @xmath48 in interval @xmath240 . for the definition of the mirror point
@xmath241 associated with a point @xmath37 , see eq .
( [ eq : gendr009uvy ] ) below .
, scaledwidth=45.0% ] for given @xmath236 ( @xmath242 ) , the length @xmath28 of the sample is randomly partitioned into @xmath236 `` ballistic intervals '' labeled @xmath240 ( @xmath243 ) , in which the electrons move ballistically across the potential energy profile @xmath48 [ see fig .
[ fig:1 ] ] .
these intervals are linked by points of local thermodynamic equilibrium @xmath244 and @xmath245 ( @xmath246 ) which , in addition , include the two ( true ) equilibrium points at the interfaces with the left and right contacts at @xmath247 and @xmath248 , respectively . [ in this section , in order to avoid confusion with the labeling of the equilibrium points inside , we denote the end - point coordinates of the sample by @xmath249 and @xmath28 , instead of @xmath69 and @xmath70 , as used elsewhere . ] at @xmath244 and @xmath245 , electrons are thermally emitted towards the right and left , respectively , into the interval @xmath237 $ ] .
any set @xmath250 of @xmath251 equilibrium points linking @xmath236 ballistic intervals characterizes a _
ballistic configuration_. introducing an equilibrium chemical potential @xmath38 all across the sample , in terms of which the values @xmath252 and @xmath253 are defined , the net ballistic current @xmath254 transmitted across the interval @xmath240 in a partition with @xmath236 intervals is given , in analogy to expression ( [ eq : ball008 ] ) for the current in the ballistic transport model , by @xmath255 , \label{eq : gendr001}\ ] ] where @xmath256 denotes the absolute maximum of the potential energy profile @xmath48 in the interval @xmath240 . in terms of the (
classical ) thermally averaged transmission probability @xmath257 } \label{eq : gendr001x}\ ] ] defined in analogy to expression ( [ eq : ball005d ] ) , we can write @xmath258 in the form @xmath259 , \nonumber \\ \label{eq : gendr001y}\end{aligned}\ ] ] where @xmath260 the current @xmath258 is , of course , conserved across each interval @xmath240 . in its general form , expression ( [ eq : gendr001y ] ) obviously does not lend itself for the calculation of transport properties in terms of external physical quantities .
therefore , we make the simplifying assumption that the electron current is conserved also across the points of local thermodynamic equilibrium linking one ballistic interval to its adjacent intervals ( and hence is constant all along the length of the sample ) , so that in eq .
( [ eq : gendr001y ] ) @xmath261 for all @xmath240 and @xmath236 , where @xmath133 is the _ physical current_. we can then iterate eq . (
[ eq : gendr001y ] ) with respect to @xmath240 ( for fixed @xmath236 ) , with the result @xmath262 here , @xmath263 denotes the overall maximum of the profile @xmath48 in the interval @xmath264 $ ] , and the function @xmath265 , defined as @xmath266 } \nonumber \\ & \equiv & \sum_{i=1}^{n } e^{-\beta [ e_{c}^{m}(0,s ) - e_{c}^{m}(i ) ] } , \label{eq : gendr002a}\end{aligned}\ ] ] can be formally viewed as the partition function corresponding to the `` energy spectrum '' @xmath267 .
it depends on the equilibrium points @xmath268 in the ballistic configuration @xmath269 via eq .
( [ eq : gendr001 m ] ) . in order to arrive at results in terms of physical quantities
, we must average eq .
( [ eq : gendr002 ] ) over the complete set of ballistic configurations @xmath270 @xmath271 . in doing so
, we identify the average of the left - hand side of eq .
( [ eq : gendr002 ] ) with the `` bias term '' @xmath272 , where @xmath273 and @xmath274 are the values of the chemical potential at the contact side of the left and right contact - semiconductor interface , respectively , and obtain @xmath275 here , @xmath276 denotes the average of the partition functions @xmath265 over all ballistic configurations @xmath270 . from eq .
( [ eq : gendr004 ] ) , the current - voltage characteristic of the prototype thermoballistic model is now obtained in the general form @xmath277 } \frac{1}{\cal r } ( 1 - e^{-\beta e v } ) , \label{eq : gendr004a}\ ] ] with the voltage bias @xmath29 given by @xmath278 aside from the barrier factor involving the absolute maximum , @xmath263 , of the potential energy profile , the current @xmath133 is controlled here by the parameter @xmath279 , which alone comprises the detailed dependence of the current - voltage characteristic on the shape of the profile and on the material parameters characterizing the sample . as it appears in eq .
( [ eq : gendr004a ] ) , @xmath279 can be viewed as a ( dimensionless ) _ reduced resistance_. in the averaging process implied in eq .
( [ eq : gendr003 ] ) , the contribution of each ballistic interval @xmath240 in a partition with @xmath236 intervals is to be weighted with the conditional probability @xmath280 for an electron to make a collision in the infinitesimal interval @xmath281 after having traveled freely across the interval @xmath240 of length @xmath282 , into which it was emitted after a collision in the interval @xmath283 .
the collision probability is connected with the _ two _ ends of each ballistic interval , and the collision interval @xmath281 is the _ equilibration interval _ for the ballistic interval @xmath240 , but it is simultaneously the _ emission interval _ for the ballistic interval @xmath284 .
this `` sharing '' of collision intervals @xmath281 between the ballistic intervals @xmath240 and @xmath284 implies that effectively the density of points of equilibration is only _ one - half _ of that given originally .
therefore , the conditional probability @xmath280 is , in conformance with eqs .
( [ eq : prob001a ] ) and ( [ eq : prob002a ] ) , expressed as @xmath285 with the _ effective mean free path _
@xmath286 given by @xmath287 here , the meaning of the mean free path @xmath7 is the original one introduced within the drude model .
the conditional probability @xmath288 for an electron emitted at @xmath289 to undergo @xmath290 collisions at the equilibrium points @xmath245 @xmath291 , and finally be absorbed with unit probability at @xmath292 , is then found as @xmath293 \nonumber \\ & \ & \times e^{-(s - x^{n}_{n-1})/\ell } \label{eq : gendr007}\end{aligned}\ ] ] for @xmath294 , while @xmath295 in the product in eq .
( [ eq : gendr007 ] ) , the exponentials cancel out except for the factor @xmath296 , so that @xmath297 and @xmath298 [ the symbol @xmath299 is meant to imply @xmath300-fold integration over @xmath301 from @xmath249 to @xmath28 ] , i.e. , the total probability is unity .
the reduced resistance @xmath279 defined by eq .
( [ eq : gendr003 ] ) can now be expressed as @xmath303 where the operators @xmath304 acting on the partition functions @xmath265 are given by @xmath305 and @xmath306 for @xmath294 .
it can be shown that , owing to the separable form of @xmath307 [ see eqs .
( [ eq : gendr002a ] ) and ( [ eq : gendr002c ] ) ] , the multi - dimensional integrals in eq .
( [ eq : gendr009 ] ) can always be reduced to one - dimensional and two - dimensional integrals .
the remaining infinite series can be summed up in terms of exponentials . for the classical partition function @xmath265 of eq .
( [ eq : gendr002a ] ) , the reduced resistance @xmath279 is thus obtained in the form @xmath308 \nonumber \\ & + & \hspace{-0.1 cm } \int_{0}^{s } \frac{dx'}{\ell } \int_{x'}^{s } \frac{dx''}{\ell } e^{-(x''-x')/\ell } \bar{r}(x',x '' ) , \label{eq : gendr009a}\end{aligned}\ ] ] where the ( dimensionless ) function @xmath309 is given by @xmath310 } \nonumber \\ & = & e^{- \beta [ e_{c}^{m}(0,s ) - e_{c}^{m}(x',x '' ) ] } .
\label{eq : gendr009b}\end{aligned}\ ] ] here , the thermally averaged classical transmission probability @xmath311 is defined in analogy to eqs .
( [ eq : ball005d ] ) and ( [ eq : gendr001x ] ) , @xmath312 is the absolute maximum of the potential energy profile @xmath48 in the interval @xmath313 $ ] , and @xmath314 in general , we have @xmath315 and , in particular , @xmath316 . for a flat potential energy profile , @xmath317 , when @xmath318 , expression ( [ eq : gendr009a ] ) can be immediately evaluated with the result @xmath319 so that in the general case , owing to eq .
( [ eq : gendr009uhd ] ) , @xmath320 i.e. , @xmath279 _ decreases _ when a position - dependent potential energy profile is introduced . in the current - voltage characteristic ( [ eq : gendr004a ] ) , this decrease of @xmath279 is overcompensated by the effect of the barrier factor @xmath321}$ ] , so that the current @xmath133 decreases as well . by reducing expression ( [ eq : gendr009 ] ) for @xmath279 to the form ( [ eq : gendr009a ] ) , we have transcribed the prototype thermoballistic model , which , according to its original concept , relies on a discrete partitioning of the sample length into ballistic intervals , into a pure continuum model .
expression ( [ eq : gendr009a ] ) is composed of three contributions .
the first corresponds to ballistic transport all across the sample length @xmath28 , with associated probability @xmath322 .
the second reflects the integrated effect of ballistic transport across intervals @xmath323 $ ] and @xmath324 $ ] , respectively , with probabilities @xmath325 and @xmath326 .
the third , finally , corresponds to the integrated effect of ballistic transport across intervals @xmath327 $ ] , with probabilities @xmath328 .
the central task in implementing the prototype thermoballistic model is to evaluate the reduced resistance @xmath279 from eq .
( [ eq : gendr009a ] ) [ or from the analogous equation derived from the quantal partition function ( [ eq : gendr002c ] ) ] for given potential energy profile @xmath48 . in general , the computation of @xmath48 is accomplished ( see , e.g. , ref . ) by solving , for given distribution of the space - fixed charges in the sample , a nonlinear poisson equation . in the ballistic transport model ( see sec .
[ sec : balltunn ] ) , we have taken into account the effect of electron tunneling by introducing quantal probabilities to describe electron transmission across the whole sample length . in the prototype model , we can include tunneling effects by using quantal transmission probabilities in expression ( [ eq : gendr001y ] ) for the current @xmath258 in the individual ballistic intervals @xmath329 $ ] .
replacing in eq .
( [ eq : gendr001y ] ) the probability @xmath330 with the thermally averaged quantal transmission probability @xmath331 defined in analogy to expression ( [ eq : ball010 ] ) , with @xmath75 replaced with @xmath332 , the partition function @xmath265 in eq .
( [ eq : gendr002 ] ) acquires the form @xmath333 } \nonumber \\ & \equiv & \sum_{i=1}^{n } \frac{\bar{t}(i)}{\bar{\cal t}(i ) } e^{-\beta [ e_{c}^{m}(0,s ) - e_{c}^{m}(i ) ] } , \label{eq : gendr002c}\end{aligned}\ ] ] where eq .
( [ eq : gendr001x ] ) for @xmath330 has been used .. in generalization of expression ( [ eq : gendr009a ] ) , the reduced resistance @xmath334 including tunneling effects is obtained from the quantal partition function @xmath335 as @xmath336 \nonumber \\ & + & \int_{0}^{s } \frac{dx'}{\ell } \int_{x'}^{s } \frac{dx''}{\ell } e^{-(x''-x')/\ell } \bar{\cal r}(x',x '' ) , \label{eq : gendr009e}\end{aligned}\ ] ] where @xmath337 } \nonumber \\ & \equiv & \frac{\bar{t}(x',x'')}{\bar{\cal t}(x',x '' ) } \bar{r}(x ' , x '' ) .
\label{eq : ball003y}\end{aligned}\ ] ] here , the thermally averaged quantal transmission probability @xmath338 is defined in analogy to expression ( [ eq : ball010 ] ) , and @xmath339 is given by eq .
( [ eq : gendr009b ] ) . in view of eq .
( [ eq : ball003y ] ) , the inequality ( [ eq : gendr009y ] ) providing an upper limit on the values of the classical function @xmath279 holds also for the quantal function @xmath334 .
we now adopt the wkb approximation , for which we have , using eq .
( [ eq : ball010xa ] ) , @xmath340 with @xmath341 inserting expression ( [ eq : ball003za ] ) in eq .
( [ eq : gendr009e ] ) , we find that @xmath334 separates in the form @xmath342 where the classical contribution @xmath343 is given by eq .
( [ eq : gendr009a ] ) , and the tunneling ( sub - barrier ) contribution @xmath344 by eq .
( [ eq : gendr009e ] ) , with @xmath345 replaced with @xmath346 .
evidently , @xmath347 and @xmath348 . according to its definition in analogy to eq .
( [ eq : ball010x ] ) , the thermally averaged sub - barrier transmission probability @xmath349 is nonzero only if the potential energy profile @xmath48 exhibits at least one local maximum in the range @xmath350 such that @xmath351 . in the evaluation of the sub - barrier function @xmath344 from eq .
( [ eq : gendr009e ] ) , ballistic intervals for which @xmath48 does not meet this requirement can , therefore , be excluded from the outset . recalling the procedure developed in the nondegenerate case ( see secs .
[ sec : ballconf ] and [ sec : averball ] ) , in which expression ( [ eq : gendr001 ] ) for the net ballistic current @xmath258 forms the starting point of the derivation of the current - voltage characteristic ( [ eq : gendr004a ] ) , we find that an analogous procedure can not be set up , in general , in the degenerate case .
the reason is that for the argument leading to eq .
( [ eq : gendr004a ] ) , @xmath258 must factorize into a term depending on the potential energy profile @xmath48 , and terms depending on the chemical potential @xmath352 . in the degenerate case ,
the current analogous to @xmath258 is the current @xmath353 given by eqs .
( [ eq : ball007 ] ) and ( [ eq : degen011e ] ) , with @xmath69 and @xmath70 , respectively , replaced with @xmath354 and @xmath355 .
this current does not factorize if the bias is nonzero .
in the zero - bias case , however , the chemical potentials at the sample ends , @xmath356 and @xmath357 , differ only infinitesimally , and the increment of @xmath38 across a ballistic interval @xmath240 , @xmath358 is infinitesimally small as well .
then , we find in analogy to eq .
( [ eq : degen001 ] ) @xmath359_{\mu = \mu(x_{i-1}^{n } ) } \nonumber \\ & = & v_{e } n_{c } \beta \delta_{i } \ln \bm { ( } 1 + e^{-\beta[e_{c}^{m}(i ) - \mu(0 ) ] } \bm { ) } , \label{eq : degen001y}\end{aligned}\ ] ] where we have replaced , in the second equation , @xmath360 with @xmath361 .
expression ( [ eq : degen001y ] ) has the factorization property alluded to above , so that , following the procedure leading to eq .
( [ eq : gendr002 ] ) in the nondegenerate case , we here find @xmath362 = \frac{j}{v_{e } n_{c } } \frac{{\cal z}^{n}_{d } } { \ln \bm { ( } 1 + e^{- \beta [ e_{c}^{m}(0,s ) - \mu(0 ) ] } \bm { ) } } , \label{eq : degen003}\ ] ] where the function @xmath363 } \bm{)}}{\ln \bm { ( } 1 + e^{- \beta [ e_{c}^{m}(i ) - \mu(0 ) ] } \bm { ) } } \nonumber \\ \label{eq : degen003a}\ ] ] generalizes the partition function @xmath265 given by eq .
( [ eq : gendr002a ] ) to the degenerate case .
averaging eq .
( [ eq : degen003 ] ) over the ballistic configurations @xmath364 yields , in analogy to eq .
( [ eq : gendr004 ] ) , @xmath365 } \bm { ) } } , \label{eq : degen005}\ ] ] where @xmath366 is the reduced resistance for the degenerate case . within the prototype model ,
we can construct , in a heuristic way , a unique chemical potential @xmath38 all along the sample length by modifying relation ( [ eq : gendr004 ] ) that connects the values of the chemical potential at the contact - semiconductor interfaces , @xmath273 and @xmath274 .
we proceed as follows . on the one hand , by replacing the fixed position @xmath28 with a variable position @xmath37 inside the sample , we turn @xmath274 into a chemical - potential function @xmath367 and , simultaneously , reduce the range of averaging over the ballistic configurations in eq .
( [ eq : gendr003 ] ) for the reduced resistance @xmath279 from @xmath368 $ ] to @xmath369 $ ] .
we then obtain @xmath370 @xmath371 , where we have introduced the `` resistance function '' @xmath372 as @xmath373 } , \label{eq : prochem03}\ ] ] which , in turn , can be expressed as @xmath374 .
\label{eq : prochem03gcx}\ ] ] on the other hand , by replacing in eq .
( [ eq : gendr004 ] ) the fixed position @xmath249 with the variable @xmath37 , so that @xmath273 turns into a chemical - potential function @xmath375 and the range of averaging is reduced to @xmath376 $ ] , we have @xmath377 @xmath371 , with the resistance function @xmath378 defined by @xmath379 } , \label{eq : prochem04}\ ] ] so that @xmath380 .
\label{eq : prochem04ldp}\ ] ] from the definitions ( [ eq : gendr003 ] ) , ( [ eq : prochem03 ] ) , and ( [ eq : prochem04 ] ) , it follows that @xmath381 the resistance functions @xmath372 and @xmath378 are obtained in explicit form from expression ( [ eq : gendr009a ] ) [ or from expression ( [ eq : gendr009e ] ) , which includes tunneling effects ] by transcribing it so as to correspond to samples with end - point coordinates @xmath382 and @xmath383 , respectively . from the potentials
@xmath384 , we now construct a unique chemical potential @xmath352 by setting @xmath385 \label{eq : prochem05}\ ] ] for @xmath386 , while @xmath387 using eqs .
( [ eq : prochem01 ] ) and ( [ eq : prochem02 ] ) , along with eq .
( [ eq : gendr004 ] ) in eq .
( [ eq : prochem05 ] ) , we obtain @xmath388 where the function @xmath389 is defined as @xmath390 .
\nonumber \\ \label{eq : prochem06gcl}\end{aligned}\ ] ] with the resistance functions @xmath391 and the reduced resistance @xmath279 calculated from eq .
( [ eq : gendr009a ] ) [ or eq . ( [ eq : gendr009e ] ) ] , the equilibrium chemical potential in the prototype model can now be explicitly evaluated in terms of the potential energy profile @xmath48 and the momentum relaxation length @xmath7 .
a noteworthy feature of the chemical potential @xmath38 is the occurrence of discontinuities at @xmath392 and @xmath393 . using eqs .
( [ eq : gendr004 ] ) , ( [ eq : gendr009a ] ) , ( [ eq : gendr009b ] ) , and ( [ eq : prochem04zwe ] ) , we find from eq .
( [ eq : prochem06 ] ) @xmath394 } - 1 = - \frac{j}{2 v_{e } n_{c } } e^{\beta [ e_{c}(0 ) - \mu_{0 } ] } \label{eq : prochem06amf}\ ] ] for @xmath392 , and @xmath395 } - 1 = \frac{j}{2 v_{e } n_{c } } e^{\beta [ e_{c}(s ) - \mu_{s } ] } \label{eq : prochem06pky}\ ] ] for @xmath393 . in the fully thermoballistic approach ( see sec . [
sec : equichem ] below ) , we will derive an expression for the average chemical potential which closely resembles expression ( [ eq : prochem06 ] ) and which exhibits discontinuities analogous to those expressed by eqs .
( [ eq : prochem06amf ] ) and ( [ eq : prochem06pky ] ) , but relies on different assumptions .
the present subsection is devoted to the application of the prototype thermoballistic model to specific examples . with a view to inhomogeneous semiconductors and heterostructures ,
in which barriers in the potential energy profile arise , e.g. , at heterojunctions , grain boundaries , and metal - semiconductor contacts ( `` schottky barriers '' ) , we deal with the evaluation of the reduced resistance @xmath279 for profiles @xmath48 exhibiting an arbitrary number of barriers and interjacent valleys . as a prelude to the general case , we treat the cases of a single barrier and of two barriers enclosing a valley separately .
another example considered here is that of field - driven transport in homogeneous semoconductors , for which we evaluate the position dependence of the chemical potential . here
, we consider the case of a single barrier in the profile @xmath48 , with its maximum located at some position @xmath396 $ ] ( see fig .
[ fig:1 ] ) .
then , we have @xmath397 .
if , in particular , @xmath398 or @xmath399 , the profile is monotonic . in the calculation of the corresponding reduced resistance @xmath279 from expression ( [ eq : gendr009a ] ) , one must distinguish three cases .
if the ballistic interval @xmath313 $ ] contains @xmath400 , we have @xmath401 ; if it lies to the left or right of @xmath400 , we have @xmath402 or @xmath403 , respectively .
we then find @xmath404 where @xmath97 is the effective sample length , @xmath405 } .
\label{eq : gendr012}\ ] ] this expression is formally equal to the effective sample length ( [ eq : diff008 ] ) introduced in the context of the drift - diffusion model , but it differs in physical origin . while expression ( [ eq : gendr012 ] ) results from averaging over ballistic configurations involving intervals of arbitrary , finite length , expression ( [ eq : diff008 ] ) arises from the assumption of arbitrarily short ballistic intervals , on which the drift - diffusion model is implicitly based . in expression ( [ eq : gendr011 ] ) for @xmath279 , the unit term corresponds to ballistic transmission all across the sample , governed by the barrier maximum at @xmath400 , while the effective sample length @xmath97 represents the effect of the potential energy profile in an integral way . the ( classical ) current - voltage characteristic ( [ eq : gendr004a ] ) for the case of a single potential energy barrier now reads @xmath406 } \frac{\ell}{\ell + \tilde{s } } ( 1 - e^{-\beta e v } ) .
\label{eq : gendr004c}\ ] ] if the effective mean free path is much longer than the effective sample length , @xmath407 , the characteristic ( [ eq : gendr004c ] ) reduces to that of the ballistic transport model , eq .
( [ eq : ball009 ] ) . in the opposite case , @xmath408 ,
using the relation @xmath409 for the electron mobility @xmath61 [ see eqs .
( [ eq : drude004hg ] ) and ( [ eq : ball002 ] ) ] , the characteristic of the drift - diffusion model , eq .
( [ eq : diff006 ] ) , is retrieved .
expression ( [ eq : gendr004c ] ) exemplifies the unification of the ballistic and drift - diffusion transport mechanisms in the prototype thermoballistic model .
we note that a heuristic attempt to unify drift - diffusion and ballistic transport has been made , for the special case of transport across a schottky barrier , by crowell and sze@xcite ( see also ref . ) , who assumed ballistic transport to prevail in the vicinity of the barrier maximum , and drift - diffusion transport elsewhere .
they obtained an expression for the current - voltage characteristic equivalent to ( [ eq : gendr004c ] ) , @xmath410 } \frac{v_{e } v_{di}}{v_{e } + v_{di } } ( 1 - e^{-\beta e v } ) , \label{eq : gendr004f}\ ] ] where @xmath411 is an effective diffusion velocity , with the electron mobility @xmath61 given by eq .
( [ eq : ballim20a ] ) . a picture similar to that of ref . has been developed and applied by evans and nelson@xcite in a study of transport across a single grain boundary barrier .
other studies elucidating the transition region between drift - diffusion and ballistic transport were presented by de jong@xcite and prins _
et al._@xcite considering now , for the single barrier , the effect of electron tunneling , we obtain , following sec .
[ sec : quantum ] , @xmath412 for the sub - barrier part @xmath344 of the reduced resistance @xmath334 , while the classical part @xmath343 is given by eq .
( [ eq : gendr011 ] ) . since @xmath413 , we have from eq .
( [ eq : gendr009s ] ) @xmath414 since @xmath415 , the transmission probabilities @xmath311 and @xmath416 , and consequently the function @xmath346 , depend on @xmath417 and @xmath418 only via the function @xmath419 [ see eqs .
( [ eq : ball005d ] ) and ( [ eq : ball010x ] ) ] .
then , @xmath346 is a function of one coordinate only , determined by the potential energy profile @xmath48 , @xmath420 as @xmath418 increases from @xmath400 to @xmath28 , for fixed @xmath421 $ ] , @xmath422 changes to @xmath423 when @xmath424 or , equivalently , @xmath425 ; here , the `` mirror point '' @xmath241 is defined as that position to the right ( left ) of a maximum or minimum of the potential energy profile where it has the same value as at the point @xmath37 to the left ( right ) , i.e. , @xmath426 ( see figs . [ fig:1 ] and [ fig:2 ] ) . the double integral in expression ( [ eq :
gendr009s ] ) then reduces to two single integrals which , together with the single integrals preceding it , combine to just one single integral , so that @xmath344 attains the simple form @xmath427 this expression bears a close formal similarity to the shape term appearing in the classical treatment of the double - barrier case [ see eq . ( [ eq : vall002a ] ) below ] .
next , we consider a potential energy profile of the type shown in fig . [ fig:2 ] , exhibiting two barriers with maxima at @xmath428 and @xmath429 , respectively ( without loss of generality , the maximum at @xmath430 is assumed to be the higher one ) , and a valley with minimum at @xmath431 in between . in the evaluation of the reduced resistance @xmath279 ,
a number of cases are to be distinguished when calculating the function @xmath432 in dependence on the location of the ballistic interval @xmath313 $ ] with respect to the different ranges delimited by the points @xmath392 , @xmath430 , @xmath433 , @xmath434 , @xmath435 , and @xmath28 . from expression ( [ eq : gendr009a ] ) , we obtain @xmath436 where the effective sample length @xmath97 is given by eq .
( [ eq : gendr012 ] ) with @xmath437 .
the `` shape term '' @xmath438 has the form @xmath439 .
\label{eq : vall002a}\end{aligned}\ ] ] in writing @xmath438 in this compact form , we have used the identity @xmath440 where @xmath441 is the width of the valley ( see fig . [ fig:2 ] ) .
the position of the minimum of the valley , @xmath434 , enters @xmath279 implicitly via the shape of @xmath48 in the interval @xmath442 $ ] . exhibiting two barriers with maxima at @xmath428 and @xmath443 , respectively , which enclose a valley with minimum at @xmath431 and width @xmath444 . for the definition of the mirror point @xmath241 associated with a point @xmath37 , see eq .
( [ eq : gendr009uvy ] ) .
, scaledwidth=45.0% ] comparing in eq .
( [ eq : vall002a ] ) the magnitude of the integral involving the term @xmath445 in the brackets to that of the integral involving @xmath446 , it is seen that for @xmath447 , where @xmath448 are the widths of the two barriers centered about @xmath449 , the second integral can be neglected .
hence , @xmath450 } , \label{eq : vall004}\ ] ] i.e. , the two barriers contribute independently to the reduced resistance ( first and third term ) . when @xmath451 ( ballistic limit ) , we have @xmath452 } < s , \label{eq : vall005}\ ] ] so that both the terms @xmath453 and @xmath454 in eq .
( [ eq : vall002 ] ) can be neglected .
then , @xmath279 reduces to the unit term , which reflects ballistic transmission across the higher barrier maximum at @xmath430 ; this maximum `` eclipses '' the lower maximum at @xmath435 . in the foregoing cases of a single barrier and of a valley in - between two barriers
, we have obtained explicit expressions for the reduced resistance @xmath279 in terms of the potential energy profile @xmath48 and the effective momentum relaxation length @xmath286 .
if the profile contains an arbitrary combination of barriers and valleys , @xmath279 must be evaluated , in general , for each case anew , starting from the basic formula ( [ eq : gendr009a ] ) .
however , an explicit expression for @xmath279 can still be obtained in the special case where the _ height _ of the barriers decreases or increases _ monotonically _ along the sample length . exhibiting @xmath455 barriers with maxima at @xmath456 ( @xmath457 ) , and heights decreasing monotonically with increasing @xmath37 .
the valleys enclosed by two adjacent barriers have minima located at @xmath458 ( @xmath459 ) and widths @xmath460 .
, scaledwidth=45.0% ] assuming , without loss of generality , the barrier height to decrease monotonically when @xmath37 increases from 0 to @xmath28 ( which includes the case of equal height of all barriers ) , we consider @xmath455 barriers , with maxima at @xmath456 @xmath461 , enclosing @xmath462 valleys , with minima at @xmath458 ( @xmath459 ) and widths @xmath460 ( see fig . [ fig:3 ] ) . then , using eq .
( [ eq : gendr009a ] ) and proceeding as in the case of a single valley , we obtain @xmath279 in the general form given by eq .
( [ eq : vall002 ] ) , with @xmath97 again given by eq .
( [ eq : gendr012 ] ) with @xmath437 , where @xmath430 is now the position of the maximum of the heighest ( left - most ) barrier of the @xmath455 barriers considered . the shape term @xmath438 entering expression ( [ eq : vall002 ] ) , @xmath463 , \nonumber \\ \label{eq : vall006a}\end{aligned}\ ] ] generalizes expression ( [ eq : vall002a ] ) to the case of @xmath462 valleys ; it appears as a sum over separate contributions from the different valleys . the contribution of valley @xmath464 , with the maximum of the adjoining lower barrier located at @xmath465 , consists of an integral which extends over the width of the valley from @xmath466 to @xmath465 . when the barrier heights in the potential energy profile do not behave monotonically , inspection of formula ( [ eq : gendr009a ] )
shows that the reduced resistance @xmath279 can always be expressed as a sum of a term formally identical to expression ( [ eq : vall002 ] ) [ with @xmath467 in @xmath97 and @xmath438 replaced with @xmath468 , where @xmath469 is the position of the maximum of the heighest barrier ] plus additional terms arising from the combined effect of barriers and valleys on the electron transport in the ballistic intervals .
introducing now the effective transport length as @xmath470 we have from eqs .
( [ eq : gendr009x ] ) , ( [ eq : gendr009y ] ) , and ( [ eq : vall002 ] ) @xmath471 from eqs .
( [ eq : gendr004a ] ) and ( [ eq : vall007a ] ) , we then obtain the current - voltage characteristic for a potential energy profile of the type depicted in fig .
[ fig:3 ] in the form @xmath472 } \frac{\ell}{l } ( 1- e^{-\beta ev } ) .
\label{eq : vall008}\ ] ] this expression is the principal result of the prototype thermoballistic model . it has been derived here in the framework of classical transport in nondegenerate systems , where its interpretation is most transparent .
the properties of the characteristic ( [ eq : vall008 ] ) are determined by the barrier factor @xmath473}$ ] and by the ratio @xmath474 . in the effective transport length @xmath475 ,
the effective mean free path @xmath286 represents the ballistic contribution to the current , which is associated with the maximum @xmath467 of the highest barrier in the potential energy profile .
the remaining terms give a quantitative measure of the influence of that part of the electron motion which is not purely ballistic .
their contribution amounts to at most the sample length @xmath28 .
the effective sample length @xmath97 given by eq .
( [ eq : gendr012 ] ) with @xmath437 represents a contribution that characterizes the potential energy profile @xmath476 in an integral way ; it does not manifestly depend on @xmath286 , only implicitly so via the profile ( an indirect relationship between profile and mean free path arises from their common dependence on the donor density ) .
the shape term @xmath438 given by eq .
( [ eq : vall006a ] ) , on the other hand , depends on the detailed structure of the profile as well as explicitly on the mean free path , and thus represents the interplay of ballistic and drift - diffusion transport .
this term is a distinctive feature of expression ( [ eq : vall008 ] ) , and therefore of the prototype thermoballistic model .
tunneling effects in the current - voltage characteristic ( [ eq : vall008 ] ) can be taken into account by generalizing the effective transport length @xmath475 of eq .
( [ eq : vall007a ] ) so as to include the sub - barrier contribution @xmath344 to the reduced resistance . using eq .
( [ eq : ball003p ] ) , we have @xmath477 here , as exemplified by expression ( [ eq : gendr009st ] ) for the single - barrier case , @xmath344 depends explicitly on the effective mean free path @xmath286 . from expression ( [ eq : gendr009r ] ) , the inclusion of tunneling is seen to _ lower _ the _ ballistic _ contribution to the effective transport length [ see eq .
( [ eq : vall007a ] ) ] .
this behavior is in line with the role of @xmath334 as a resistance , which ought to decrease when the barriers become `` transparent '' to ballistic electron transport .
the ( classical ) reduced resistance in the degenerate case , @xmath478 [ see eqs .
( [ eq : degen003a ] ) and ( [ eq : degen004 ] ) ] , is expressed , in analogy to eqs .
( [ eq : vall007a ] ) and ( [ eq : vall007 ] ) for the nondegenerate case , in terms of a ( classical ) effective transport length @xmath479 given by @xmath480 where , in generalization of eq .
( [ eq : gendr012 ] ) , @xmath481 } \bm{)}}{\ln \bm { ( } 1 + e^{- \beta [ e_{c}(x ) - \mu_{0 } ] } \bm { ) } } \label{eq : degen007}\ ] ] is the effective sample length for the degenerate case , and @xmath482 } \bm { ) } } { \ln \bm { ( } 1 + e^{- \beta [ e_{c}(x_{v } ) - \mu_{0 } ] } \bm { ) } } - \frac { \ln \bm { ( } 1 + e^{- \beta [ e_{c}(x_{0 } ) - \mu_{0 } ] } \bm { ) } } { \ln \bm { ( } 1 + e^ { -\beta [ e_{c}(x ) - \mu_{0}]}\bm { ) } } \right\ } \nonumber \\ \label{eq : degen008}\end{aligned}\ ] ] generalizes the shape term ( [ eq : vall006a ] ) .
the inclusion of tunneling effects is , in the degenerate case , a highly intricate task and will not be considered here . setting @xmath483 in eq .
( [ eq : degen005 ] ) , we now have for the zero - bias conductance per unit area in the prototype thermoballistic model @xmath484 in generalization of expression ( [ eq : degen009 ] ) for the conductance in the ballistic transport model . previously , we have applied@xcite the prototype model in calculations of transport properties of poly- and microcrystalline semiconducting materials , in particular , of materials relevant to photovoltaics .
the occurence of grain boundaries in this kind of materials gives rise to a multi - barrier structure of the band edge profile .
adopting the trapping model@xcite to describe the grain boundaries , we have solved@xcite the corresponding nonlinear poisson equation to obtain zero - bias potential energy profiles for chains of grains . then , taking into account tunneling corrections and using a phenomenological relation@xcite to express the momentum relaxation length @xmath7 in terms of the donor density , we have calculated zero - bias conductivities and electron mobilities as a function of @xmath7 , and of the number and lengths of the grains .
it turns out that neither the ballistic ( thermionic ) model nor the drift - diffusion model can provide an adequate description of electron transport in poly- and microcrystalline materials .
the application of the prototype model in the analysis of experimental data@xcite has led to promising results .
the chemical potential @xmath352 given by eq .
( [ eq : prochem06 ] ) can be expressed in closed form for the case of electron transport in a homogeneous semiconductor ( no space charges ) , driven by a constant external electric field of magnitude @xmath485 . with the field assumed to be directed antiparallel to the @xmath37-axis ,
the corresponding potential energy profile reads @xmath486 where @xmath487 . using this in the properly transcribed expression ( [ eq : gendr009a ] ) [ or , alternatively , in eq .
( [ eq : gendr011 ] ) , considering the profile ( [ eq : potfi01 ] ) a particular case of a barrier ] , we obtain the resistance functions @xmath391 in the form @xmath488 and @xmath489 \equiv { \cal r}_{1}(s - x ) .
\label{eq : potfi03}\ ] ] from eq .
( [ eq : prochem04zwe ] ) , we then have @xmath490 and from eq .
( [ eq : prochem06gcl ] ) , @xmath491 inserting expressions ( [ eq : potfi04a ] ) and ( [ eq : potfi04 ] ) in eq .
( [ eq : prochem06 ] ) gives @xmath38 in closed form . for zero bias ,
when @xmath492 , we have @xmath493 and @xmath494 i.e. , @xmath495 varies linearly with position .
the prototype thermoballistic model developed in the preceding section has been based on the random partitioning of the length of a semiconducting sample into ballistic transport intervals linked by points of local thermodynamic equilibrium , which make up a ballistic configuration .
electrons thermally emitted at either end - point of a ballistic interval , and subsequently transmitted across the potential energy profile in the sample , form the net ballistic electron current in the interval .
this current is conserved across an individual ballistic interval .
a distinctive assumption of the prototype model has been that the current is conserved also across the points of local thermodynamic equilibrium linking the ballistic intervals , and equals the physical current . by averaging over all ballistic configurations , the current - voltage characteristic
can then be expressed , without requiring knowledge of the equilibrium chemical potential inside the sample , essentially in terms of a reduced resistance that comprises the effect of the sample parameters .
the position dependence of the chemical potential has been constructed in a heuristic way only . while adhering to the idea of introducing ballistic configurations and averaging thereover , the thermoballistic concept proper refines the prototype model in that it abandons the assumption of current conservation across the points of local thermodynamic equilibrium .
[ a simple example contradicting this assumption is provided by the case of the ballistic currents in a homogeneous semiconductor at zero bias considered in sec .
[ sec : homsem ] below . ]
position - dependent total and spin - polarized thermoballistic currents as well as the associated densities are defined in terms of an average chemical potential and a spin accumulation function related to the splitting of the spin - resolved chemical potentials . by imposing appropriately chosen physical conditions on these dynamical functions , procedures for their explicit determination are devised .
we have developed the thermoballistic approach to semiclassical carrier transport@xcite in a series of papers . in ref . , the concept of a thermoballistic current was introduced and implemented without regard to spin degrees of freedom .
the extension of this concept to spin - polarized electron transport across a spin - degenerate potential energy profile was presented in ref . ,
in which spin injection out of ferromagnetic contacts into a nonmagnetic semiconducting sample was treated in detail . in ref .
, we have generalized the thermoballistic concept to the case of arbitrary spin splitting of the profile , thereby covering , in particular , spin - polarized transport in diluted magnetic semiconductors in their paramagnetic phase . in the present section ,
we formulate the thermoballistic concept within the frame set by ref . .
classical transport in nondegenerate systems is considered throughout .
effects of electron tunneling and degeneracy can be included , in principle , by resorting to the corresponding developments in secs .
[ sec : ballmech ] and [ sec : gendrude ] . for comprehensive surveys of the fundamentals of spin physics in semiconductors and their application in the field of spintronics
, we refer the reader to refs . and .
enclosed between two metal contacts .
the spin splitting @xmath496 of the electrostatic potential energy profile @xmath48 gives rise to the spin - dependent potential energy profiles @xmath497 .
the coordinates @xmath417 and @xmath418 denote points of local thermodynamic equilibrium which delimit a ballistic transport interval @xmath327 $ ] .
electrons thermally emitted at @xmath498 toward the right ( left ) move ballistically across the profiles @xmath497 to reach the end - point @xmath499 of the ballistic interval , where they are absorbed ( equilibrated ) .
two net electron currents are indicated : the ( conserved ) net total ballistic current @xmath500 determined by the values @xmath501 and @xmath502 of the average chemical potential @xmath503 and the values @xmath504 and @xmath505 of the spin accumulation function @xmath506 according to eq .
( [ eq : ball004jvx ] ) [ with the mean spin function @xmath507 expressed in terms of @xmath503 via eq .
( [ eq : eqden03zk ] ) ] , and the net relaxing ballistic spin - polarized current @xmath508 determined by @xmath504 and @xmath505 according to eq .
( [ eq : netbal02 ] ) .
the quantities @xmath76 and @xmath509 are the values of the equilibrium chemical potential and the spin accumulation function , respectively , at the contact side of the contact - semiconductor interfaces [ see eqs .
( [ eq : ballim05 ] ) and ( [ eq : ballim05s ] ) ] . , scaledwidth=45.0% ] in sec .
[ sec : gendrude ] , we have introduced ballistic transport intervals with end - point coordinates @xmath510 characterized by discrete labels @xmath511 . when the ballistic configurations made up of these intervals are averaged over , the description in terms of discrete coordinates turns into one in terms of the _ continuous _ coordinates @xmath512 in expressions ( [ eq : gendr009a ] ) and ( [ eq : gendr009e ] ) for the reduced resistance @xmath279 . transferring this feature into the formulation of the thermoballistic concept , we work here with ballistic intervals @xmath313 $ ] with continuous end - point coordinates @xmath417 and @xmath418 representing points of local thermodynamic equilibrium ( see also sec .
[ sec : probappr ] ) .
for notational convenience , we will henceforth label the end - points of a semiconducting sample by @xmath69 and @xmath70 , respectively , so that we have @xmath513 and the sample length @xmath28 is given by @xmath85 . in fig .
[ fig:4 ] , a paramagnetic semiconducting sample enclosed between two metal contacts is depicted in a schematic diagram .
various physical quantities appearing in the thermoballistic description are indicated .
we write the spin - dependent potential energy profiles @xmath514 corresponding to spin - up @xmath515 and spin - down @xmath516 conduction band electrons , respectively , in the form @xmath517 here , the spin - independent part @xmath518 is assumed to comprise the conduction band edge potential and the external electrostatic potential , and @xmath519 is the spin splitting of the conduction band . having in mind electron transport in diluted magnetic semiconductors in their paramagnetic phase , we identify this splitting with the ( giant ) zeeman splitting due to an external magnetic field @xcite [ we restrict ourselves to considering a single landau level whose energy is assumed to be included in @xmath518 ] . in presenting the general formalism , we assume both @xmath518 and @xmath519 , and hence @xmath521 , to be continuous functions of @xmath417 in the interval @xmath83 $ ] .
the case of abrupt changes in one or the other of these functions , which occur at the interfaces in heterostructures , may be described , in a simplified picture , in terms of discontinuous functions ( see sec .
[ sec : dmsnmshe ] below ) . in terms of the boltzmann factors
@xmath522 , the ( static ) local spin polarization of the conduction band electrons , @xmath523 , is given by @xmath524 where @xmath525 so that we have @xmath526 the function @xmath527 defined by @xmath528^{2 } \right\}^{1/2 } \nonumber \\ & = & \frac{1}{\cosh \bm { ( } \beta \delta(x')/2 \bm { ) } } \label{eq : spfra17c}\end{aligned}\ ] ] will be frequently used below .
for the spin - resolved equilibrium electron densities @xmath529 , we have , in generalization of expression ( [ eq : drude014 ] ) , @xmath530 } .
\label{eq : eqden02}\ ] ] here , @xmath531 are the spin - resolved chemical potentials associated with the local thermodynamic equilibrium at @xmath417 , and @xmath532 , with @xmath180 given by eq .
( [ eq : diff001a ] ) , is the effective density of states of either spin at the conduction band edge ( for simplicity , the effective electron mass @xmath4 entering @xmath180 is assumed here to be independent of position and of the external magnetic field ) .
the spin - resolved chemical potentials @xmath533 and @xmath534 are independent dynamical functions whose position dependence is to be determined within the thermoballistic approach and from which , subsequently , all transport properties are to be derived .
however , in implementing the thermoballistic approach , we will work not with @xmath533 and @xmath534 , but with suitably defined combinations of these functions : ( i ) an `` average chemical potential '' , and ( ii ) a `` spin accumulation function '' related to the splitting of the potentials @xmath533 and @xmath534 .
we define the `` spin functions '' @xmath535 as @xmath536 and the `` mean spin function '' @xmath507 as @xmath537 where @xmath538 we can express @xmath539 in the form @xmath540 where @xmath541 \label{eq : spfra15xxx}\ ] ] is the mean value of @xmath533 and @xmath534 [ mean chemical potential ] , and @xmath542 their splitting .
[ in the following , the @xmath543 notation introduced in eqs .
( [ eq : eqden03oh ] ) and([eq : spfra15x ] ) , respectively , will be frequently used _
mutatis mutandis _ to denote spin - summed and spin - polarized quantities .
] further , writing @xmath544 we introduce the _ `` average chemical potential '' _ @xmath545 .
the _ `` spin accumulation function '' _
@xmath546 is defined as @xmath547 [ this function agrees with the `` spin transport function '' @xmath504 of ref . , but differs by a factor of two from the identically named function introduced in ref . .
] the name here chosen for @xmath546 derives from the fact that in the limit @xmath548 , this function becomes proportional to the splitting @xmath549 , which is the dynamical quantity used in the drift - diffusion approach to spin - polarized transport , and which is commonly@xcite referred to as the spin accumulation there . using eq .
( [ eq : eqden03nn ] ) , we can rewrite @xmath546 in the form @xmath550 introducing the `` reduced '' spin accumulation function @xmath551 as @xmath552 we then obtain , using eqs .
( [ eq : eqden03zk])([eq : spfra01fv ] ) , the relations @xmath553 expressing the difference between average and mean chemical potential in terms of the splitting @xmath549 and of @xmath551 , respectively .
the spin functions @xmath535 are proportional to the spin - resolved electron densities @xmath554 [ see eqs .
( [ eq : eqden02 ] ) ] , so that their use results in a formulation in terms of _ linear _ equations , instead of the nonlinear description ensuing from using the spin - resolved chemical potentials @xmath531 themselves .
[ this aspect has been emphasized previously @xcite within a study , based on the standard drift - diffusion approach , of electric - field effects on spin - polarized transport in nondegenerate semiconductors . ] from eqs .
( [ eq : eqden02 ] ) , the total ( i.e. , spin - summed ) equilibrium density , @xmath555 , and the spin - polarized equilibrium density , @xmath556 , are now obtained in terms of the functions @xmath557 as @xmath558 , \label{eq : ball004jux}\ ] ] where eqs .
( [ eq : eqden60 ] ) and ( [ eq : eqden61 ] ) have been used . in this subsection , the spin - resolved electron currents transmitted across a ballistic transport interval as well as the associated densities ( called `` ballistic currents '' and `` ballistic densities '' , for short ) are constructed by closely following the development in the ballistic ( see sec . [
sec : ballmech ] ) and the prototype thermoballistic ( see sec . [
sec : ballconf ] ) transport models , in which spin degrees of freedom have been disregarded . introducing spin relaxation during the ballistic electron motion
, we obtain the ballistic spin - polarized currents and densities , whose dynamics are determined from a balance equation .
the left end - point , @xmath417 , as well as the right end - point , @xmath418 , of the ballistic interval @xmath313 $ ] are points of local thermodynamic equilibrium but are , in general , _ not _ points of spin equilibrium .
we first assume that the electrons thermally emitted at @xmath417 towards the right are _ not _ affected by spin relaxation during their motion across the interval @xmath313 $ ] . the spin - resolved densities
@xmath554 [ see eq .
( [ eq : eqden02 ] ) ] then give rise to conserved ballistic spin - resolved currents ( i.e. , currents independent of the position @xmath559 $ ] ) in that interval , which have the form @xmath560 ( see sec .
[ sec : ballnonc ] ) , where the thermally averaged ( classical ) transmission probabilities @xmath561 are given [ see eqs .
( [ eq : ball005 ] ) and([eq : ball005a ] ) ] by @xmath562 } , \label{eq : ball004iii}\ ] ] with @xmath563 the overall maxima of the potential energy profiles @xmath564 in @xmath313 $ ] . in terms of the spin functions @xmath535 given by eqs .
( [ eq : eqden03 mm ] ) , we can rewrite the currents @xmath565 in the form @xmath566 we then have for the total ( i.e. , spin - summed ) ballistic current , @xmath567 , and the ballistic spin - polarized current , @xmath568 , @xmath569 , \nonumber \\ \label{eq : ball004ju}\end{aligned}\ ] ] where @xmath570 and @xmath571 which are nonlocal extensions of expressions ( [ eq : eqden60 ] ) and ( [ eq : eqden06 ] ) for the functions @xmath572 and the local polarization @xmath523 , respectively .
we now consider the effect of spin relaxation on the ballistic currents inside the interval @xmath313 $ ] .
we introduce functions @xmath573 that depend on the position @xmath574 $ ] and are required to satisfy the initial conditions @xmath575 using these functions , we generalize the definitions of the currents @xmath576 given by eqs .
( [ eq : ball004ju ] ) to @xmath577 .
\label{eq : spirel08}\end{aligned}\ ] ] now , also in the presence of spin relaxation , the _ total _ ballistic current must be conserved , @xmath578 where , from eq .
( [ eq : ball004ju ] ) , @xmath579 .
\label{eq : ball004juhu}\end{aligned}\ ] ] equating this to expression ( [ eq : spirel08 ] ) for @xmath580 , we obtain the relation @xmath581 .
\label{eq : balden09}\end{aligned}\ ] ] therefore , the spin - polarized current can be written as @xmath582 where @xmath583^{2 } , \label{eq : spirel15}\ ] ] and @xmath584^{2 } \right\}^{1/2 } \label{eq : emtra14}\ ] ] is the nonlocal extension of expression ( [ eq : spfra17c ] ) for the function @xmath527 .
the ballistic spin - polarized current of eq .
( [ eq : spirel11 ] ) has the form @xmath585 where the first ( @xmath37-independent , nonrelaxing ) term is the `` persistent '' ballistic spin - polarized current , @xmath586 while the second ( @xmath37-dependent ) term is the `` relaxing '' ballistic spin - polarized current , @xmath587 the latter describes the spin dynamics in the current @xmath588 via the `` spin relaxation function '' @xmath589 . the @xmath37-dependence of @xmath589 will be determined explicitly in sec .
[ sec : balance ] , while the procedure for calculating the spin accumulation function @xmath590 will be described in sec .
[ sec : accufunc ] .
the persistent ballistic spin - polarized current @xmath591 depends , via the total ballistic current @xmath567 , on the spin accumulation function @xmath546 .
when @xmath546 is calculated within the thermoballistic approach , spin relaxation in all ballistic intervals ( including the interval @xmath313 $ ] under consideration ) is taken into account .
therefore , while being not _ directly _ affected by spin relaxation _ inside _ the interval @xmath313 $ ] , the persistent current @xmath591 depends , via the dependence of the factor @xmath592 on @xmath546 [ see eqs .
( [ eq : ball004juhu ] ) and ( [ eq : spirel12 ] ) ] , in an _ indirect _ way on spin relaxation .
the ballistic densities @xmath593 associated with the ballistic currents @xmath594 of eqs .
( [ eq : spirel08 ] ) are given by @xmath595 \nonumber \\ \label{eq : ball004m}\end{aligned}\ ] ] [ see eq .
( [ eq : ball004ka ] ) ] . here
, we have taken into account that the ballistic densities correspond to _ one half _ of the thermal currents emitted symmetrically at a point of local thermodynamic equilibrium [ see , e.g. , eq . ( [ eq : ball001h ] ) ] .
the functions @xmath596 and @xmath597 are defined as @xmath598 where @xmath599 } \nonumber \\ & \ & \times { \rm erfc } \bm { ( } \ { \beta [ e_{\uparrow , \downarrow}^{m}(x ' , x '' ) - e_{\uparrow , \downarrow}(x ) ] \}^{1/2 } \bm { ) } \nonumber \\ \label{eq : ball004n}\end{aligned}\ ] ] [ see eq .
( [ eq : ball004kaa ] ) ] , and @xmath600 in generalization of expression ( [ eq : emtra09 ] ) for the function @xmath601 . now , inserting expression ( [ eq : balden09 ] ) for @xmath602 in eqs .
( [ eq : ball004 m ] ) , we find that the total ballistic density , @xmath603 , and the ballistic spin - polarized density , @xmath604 , have the form @xmath605 here , the first ( @xmath37-dependent , but nonrelaxing ) term is the persistent part , @xmath606 \nonumber \\ \label{eq : balden13}\end{aligned}\ ] ] and the second term is the relaxing part , @xmath607 where @xmath608 in contrast to the @xmath37-independent currents @xmath567 and @xmath591 , the persistent parts of the ballistic densities , @xmath609 , depend on position via the @xmath37-dependence of the potential energy profiles @xmath564 . for zero spin splitting of the conduction band , @xmath610 , when @xmath611 , the persistent part of the spin - polarized density as well as the relaxing part of the total density vanish , @xmath612 . in a time - dependent formulation , spin relaxation in a system described by spin - resolved electron densities @xmath613 and currents @xmath614
is governed by the local coupled spin balance equations@xcite @xmath615 and @xmath616 where @xmath617 @xmath618 is the rate for spin - flip scattering from spin - up ( spin - down ) to spin - down ( spin - up ) states . in the stationary case , when @xmath619 , this leads to the balance equation @xmath620 connecting the relaxing part of the spin - polarized current , @xmath621 , to the relaxing parts of the spin - polarized , @xmath622 , and of the total density , @xmath623 . here ,
@xmath624 , defined as @xmath625 is the spin relaxation time .
we apply eq .
( [ eq : spirel01ac ] ) to spin relaxation during ballistic transport . in doing this , we will disregard the term involving @xmath626 for two reasons : ( i ) as it is preceded by the _ difference _ of the two relaxation rates @xmath617 and @xmath627 , which are estimated to be of comparable magnitude , the term may generally be considered small in comparison with the term involving @xmath628 .
( ii ) since both @xmath628 and @xmath626 are proportional to the spin relaxation function @xmath629 [ see eq .
( [ eq : balden1 ] ) ] , we can account for @xmath626 by combining its prefactors with those of @xmath628 in an _ effective _ spin relaxation time ( still denoted by @xmath624 ) depending , in general , on the potential energy profiles .
[ note that we must not assume @xmath630 from the outset by adopting the arguments leading to eq .
( 2.7 ) of ref . .
the density @xmath631 appearing in that equation is the deviation of the total density from its spin equilibrium value .
by contrast , the relaxing total density @xmath626 , when used in the thermoballistic description , gives the deviation of the total density from the persistent total density .
in the latter , a spin non - equilibrium part enters via the spin accumulation function @xmath632 [ see eq .
( [ eq : balden13 ] ) ] . ] in the thermoballistic approach , it is assumed that the thermally emitted electrons spend only an infinitesimally short time span at the emission point , and it is only during their motion across the ballistic interval that they can undergo spin relaxation .
spin relaxation in ballistic transport is commonly described@xcite in terms of a ( ballistic ) spin relaxation length @xmath633 given by @xmath634 [ as in the case of the effective electron mass @xmath4 , we assume @xmath624 , and hence @xmath633 , to be independent of position and of the external magnetic field ; we consider @xmath633 here an _ effective _ quantity , in line with the interpretation of @xmath624 . ] in our description of spin - polarized electron transport , spin relaxation is thus completely separated from momentum relaxation at the points of local thermodynamic equilibrium and , in this respect , is similar to the dyakonov - perel relaxation mechanism.@xcite in terms of the relaxing ballistic spin - polarized current @xmath635 and the corresponding density @xmath636 , the balance equation governing spin relaxation during ballistic transport now reads @xmath637 where the spin relaxation length @xmath633 is given by eq .
( [ eq : spirel01 ] ) . inserting in eq .
( [ eq : baleq01 ] ) the expressions for @xmath635 and @xmath636 from eqs .
( [ eq : spirel14 ] ) and ( [ eq : balden1 ] ) , respectively , we obtain a first - order differential equation for the spin relaxation function @xmath629 , @xmath638 where @xmath639 the solution of eq .
( [ eq : baleq02 ] ) obeying the initial condition ( [ eq : spirel05 ] ) is @xmath640 where @xmath641 with @xmath642 .
now , inserting expression ( [ eq : baleq05 ] ) in eqs .
( [ eq : spirel14 ] ) and ( [ eq : balden1 ] ) , respectively , we obtain the relaxing ballistic spin - polarized current and density explicitly in terms of the spin accumulation function @xmath546 , @xmath643 and @xmath644 the @xmath37-dependence of the relaxing spin - polarized current and density in the ballistic interval @xmath327 $ ] is hence governed by the factor @xmath645 .
it departs from a purely exponential behavior unless the potential energy profiles @xmath497 are constant over the interval , in which case @xmath646 .
so far , we have only considered thermal emission at the _ left _ end - point , @xmath417 , of the ballistic interval @xmath327 $ ] , obtaining a variety of ballistic currents and densities summarily denoted here by @xmath647 and @xmath648 , respectively .
the analogous ballistic currents and densities @xmath649 and @xmath650 corresponding to emission at the _
right _ end - point @xmath418 can be expressed in terms of those emitted at @xmath417 as @xmath651 and @xmath652 [ note that , owing to the symmetry of the functions @xmath653 , the functions @xmath654 , @xmath655 , @xmath656 , and @xmath657 entering the expressions for @xmath647 and @xmath648 are symmetric under the exchange of @xmath417 and @xmath418 . ]
we then have @xmath658 and @xmath659 for the _ net _ ballistic currents and _ joint _ ballistic densities summarily denoted by @xmath660 and @xmath661 , respectively . for the ( conserved ) net _ total _ ballistic current @xmath500 inside the ballistic interval , we now find , using eq .
( [ eq : ball004juhu ] ) , @xmath662 \nonumber \\ & \ & + { \mbox{$\frac{1}{2}$ } } p^{m}(x ' , x'')[a(x ' ) -a(x '' ) ] \ } .
\label{eq : ball004jvx}\end{aligned}\ ] ] here , the function @xmath663 has been replaced with @xmath664 [ see eq .
( [ eq : eqden03 ] ) ] , and the subscript attached to the spin accumulation function @xmath546 has been omitted , i.e. , we have set @xmath665 the current @xmath500 is seen to be dynamically determined , in general , both by the average chemical potential @xmath545 [ via the mean spin function@xmath539 ] and the spin accumulation function @xmath504 .
the same then holds for the net _ persistent _ ballistic _ spin - polarized _
current @xmath667 , for which we have , using eq .
( [ eq : spirel12 ] ) , @xmath668 for zero spin splitting , when @xmath669 and hence @xmath670 , the dependence of @xmath500 on @xmath506 drops out and , further , @xmath671 . for the net
_ relaxing _ ballistic _ spin - polarized _ current , we have , using eq .
( [ eq : baleq07 ] ) , @xmath672 , \label{eq : netbal02}\end{aligned}\ ] ] which is dynamically determined by the spin accumulation function @xmath506 alone . for the joint _ total _ ballistic density , @xmath673 , and the joint ballistic _ spin - polarized _ density , @xmath674 , respectively , we have from eq .
( [ eq : spirel11kx ] ) @xmath675 with _ persistent _ parts @xmath676 and _ relaxing _ parts @xmath677 given by @xmath678 \nonumber \\ & \ & + { \mbox{$\frac{1}{2}$ } } p^{m}(x ' , x'')[a(x ' ) + a(x '' ) ] \ } \nonumber \\ \label{eq : ball004juy}\end{aligned}\ ] ] [ see eq .
( [ eq : balden13 ] ) ] and @xmath679 \nonumber \\ \label{eq : netbal02xy}\end{aligned}\ ] ] [ see eqs .
( [ eq : balden1 ] ) and ( [ eq : baleq08 ] ) ] , in analogy to eqs .
( [ eq : ball004jvx ] ) , ( [ eq : netbal06 ] ) , and ( [ eq : netbal02 ] ) for the net currents . from expressions ( [ eq : netbal02 ] ) and
( [ eq : netbal02xy ] ) , we now derive , using eqs .
( [ eq : baleq03 ] ) and ( [ eq : baleq06 ] ) , the balance equation @xmath680 connecting the net relaxing ballistic spin - polarized current @xmath508 and the associated joint density @xmath681 .
this equation can be obtained , of course , simply by adding the balance equation eq .
( [ eq : baleq01 ] ) and the corresponding equation for emission at @xmath418 . in a significant advance over its prototype presented in sec .
[ sec : gendrift ] , where overall current conservation was introduced via the condition ( [ eq : gendr001qal ] ) , the thermoballistic concept proper rests on the introduction of a `` reference coordinate '' @xmath37 that characterizes an arbitrary point inside the semiconducting sample extending from @xmath69 to @xmath70 , as shown in fig .
[ fig:4 ] . singling out this coordinate
, we consider the net ballistic currents and joint ballistic densities within the ensemble of all ballistic intervals @xmath327 $ ] enclosing @xmath37 .
these currents and densities form the building blocks for establishing , at @xmath37 , the corresponding thermoballistic quantities . the point labeled by the reference
coordinate @xmath37 is _ not _ a point of local thermodynamic equilibrium ( we may call it a `` ballistic point '' ) .
however , the `` equilibrium points '' @xmath512 may come infinitesimally close to @xmath37 . the thermoballistic currents and densities at the point @xmath37 are constructed by performing weighted summations of the corresponding net ballistic currents and joint ballistic densities over all ballistic intervals @xmath327 $ ] subjected to the condition @xmath682 just as in the prototype thermoballistic model ( see sec . [
sec : averball ] ) , we adopt here the probabilistic picture outlined in sec . [ sec : probappr ] : the contributions from the interval @xmath327 $ ] are weighted with the probability @xmath683 ( one - dimensional transport is assumed here ; see the remark at the beginning of sec . [
sec : diffmech ] ) that the electrons traverse the interval without collisions , multiplied by the probability @xmath684 ( @xmath685 ) that they are absorbed or emitted in an interval @xmath686 ( @xmath687 ) around the end - point @xmath417 ( @xmath418 ) . at the ends of the semiconducting sample at @xmath75 , absorption and emission occur with unit probability . like the effective electron mass @xmath4 and the spin relaxation length @xmath633
, the momentum relaxation length @xmath7 is assumed here to be independent of position and of the external magnetic field . representing the net ballistic currents and joint ballistic densities of eqs .
( [ eq : ball004jvx])([eq : netbal02xy ] ) summarily by a function @xmath688 , and the corresponding thermoballistic currents and densities by @xmath689 , we write @xmath690 in the form @xmath691 where @xmath692 , and the infinitesimal @xmath181 has been introduced in accordance with condition ( [ eq : thebal00 ] ) .
further , we set @xmath693 in parallel to eq .
( [ eq : gendr008x ] ) for the total probability in the prototype model , we have @xmath694 for @xmath695 . according to eq .
( [ eq : thebal01 ] ) .
, scaledwidth=45.0% ] specifically , the total thermoballistic current @xmath696 is given by expression ( [ eq : thebal01 ] ) with @xmath500 of eq .
( [ eq : ball004jvx ] ) substituted for @xmath697 . the persistent thermoballistic spin - polarized current @xmath698 follows by identifying @xmath688 with @xmath667 of eq .
( [ eq : netbal06 ] ) .
the relaxing thermoballistic spin - polarized current @xmath699 is obtained by replacing @xmath688 with @xmath700 of eq .
( [ eq : netbal02 ] ) .
further , the thermoballistic densities @xmath701 , @xmath702 , and @xmath703 corresponding to the currents @xmath696 , @xmath698 , and @xmath699 follow by substituting the respective joint ballistic densities [ see eqs .
( [ eq : netbal03ax])([eq : netbal02xy ] ) ] for @xmath688 in eq .
( [ eq : thebal01 ] ) . in expression ( [ eq : thebal01 ] ) when specialized to the total thermoballistic current @xmath696 , the first term on the right - hand side represents the net electron current passing through the point @xmath37 while being ballistically transmitted between @xmath69 and @xmath70 [ which occurs with probability @xmath704 , the second term refers to ballistic electron motion between any point @xmath417 @xmath705 and @xmath70 [ probability @xmath706 , the third term to ballistic electron motion between @xmath69 and any point @xmath418 @xmath707 [ probability @xmath708 , etc .
the four types of ballistic current appearing in expression ( [ eq : thebal01 ] ) are illustrated by the double arrows labeled i to iv in fig .
[ fig:5 ] .
the derivatives with respect to @xmath37 of the various thermoballistic quantities can be written in the general form @xmath709 where the first term on the right - hand side comprises the contributions arising from differentiating the functions @xmath710 in the integrands of expression ( [ eq : thebal01 ] ) , and the second , those from differentiating the limits of integration:@xmath711 ) - \mathbb{f}_{2}(x ; [ f ] ) \ } .
\label{eq : thebal06jb}\end{aligned}\ ] ] here , the functionals @xmath712)$ ] and @xmath713)$ ] , defined by @xmath714 ) & = & e^{- ( x - x_{1})/l } f(x_{1 } , x^{+ } ; x ) \nonumber \\ & \ & + \int_{x_{1}}^{x^{- } } \frac{dx'}{l } e^{- ( x - x')/l } f(x ' , x^{+ } ; x ) \nonumber \\ \label{eq : thec00ohx}\end{aligned}\ ] ] and @xmath715 ) & = & e^{- ( x_{2 } - x)/l } f(x^{- } , x_{2 } ; x ) \nonumber \\ & \ & + \int_{x^{+}}^{x_{2 } } \frac{dx''}{l } e^{- ( x '' - x)/l } f(x^{- } , x '' ; x ) , \nonumber \\ \label{eq : thec00gma}\end{aligned}\ ] ] respectively , represent the contributions of the function @xmath710 arising from the ranges to the left and right of the point @xmath37 . for the total thermoballistic current @xmath696 constructed from the @xmath37-independent ballistic current @xmath500 , the first term on the right - hand side of eq .
( [ eq : thebal01dz ] ) vanishes , and we have @xmath716 with @xmath717 given by expression ( [ eq : thebal06jb ] ) for @xmath718 .
we note that @xmath719 is not , in general , equal to zero , so that the total thermoballistic current @xmath696 is not conserved .
the quantity @xmath720 is the increment of @xmath696 across the infinitesimal interval @xmath226 at position @xmath37 of the sample .
the four types of ballistic current appearing in expression ( [ eq : thebal06jb ] ) , which contribute to this increment , are depicted by the double arrows labeled i to iv in fig .
[ fig:6 ] , with the arrows i and ii representing the term @xmath712)$ ] , and iii and iv the term @xmath713)$ ] . of the total thermoballistic current @xmath696
according to eq .
( [ eq : thebal06jb ] ) .
, scaledwidth=45.0% ] if we set @xmath721 in expressions ( [ eq : thebal01 ] ) and ( [ eq : thebal06jb ] ) , we find from eq .
( [ eq : thebal01dz ] ) for the relaxing thermoballistic spin - polarized current @xmath722 in the first term on the right - hand side of this equation , we have introduced the relaxing thermoballistic spin - polarized density @xmath703 by using the balance equation ( [ eq : baleq01zs ] ) to replace the derivatives @xmath723 which we encounter when differentiating the integrals in expression ( [ eq : thebal01 ] ) . the thermoballistic currents and densities , eq . ( [ eq : thebal01 ] ) , are expressed , via the corresponding ballistic currents and densities , eqs .
( [ eq : ball004jvx])([eq : netbal02xy ] ) , in terms of two dynamical functions , _ viz . _ , the mean spin function @xmath507 and the spin accumulation function @xmath506 , or , equivalently , the average chemical potential @xmath503 and the splitting of the spin - resolved chemical potentials , @xmath724 [ see eqs . ( [ eq : eqden03zk ] ) , ( [ eq : spfra01fv ] ) , and ( [ eq : ball004jhm ] ) ] . to implement the thermoballistic concept
, we must establish algorithms for calculating @xmath507 and @xmath506 in terms of the intrinsic parameters of the semiconducting system , like momentum and spin relaxation lengths , as well as of the external parameters , like applied voltage and spin polarization in the external leads . before establishing these algorithms within the thermoballistic approach ,
i.e. , for arbitrary momentum relaxation length @xmath7 , we directly evaluate the thermoballistic currents and densities in the drift - diffusion regime ( small @xmath7 ) and for the ballistic case ( @xmath725 ) .
the results will turn out to be equal to the respective standard physical expressions summarized in sec .
[ sec : dibatran ] , which demonstrates that the thermoballistic description indeed bridges the gap between the drift - diffusion and ballistic descriptions of carrier transport . in the _ drift - diffusion regime _
, when @xmath726 and @xmath727 , nonzero contributions to the thermoballistic currents and densities defined by eq .
( [ eq : thebal01 ] ) arise only from very short ballistic intervals @xmath313 $ ] enclosing the point @xmath37 , so that @xmath728 and @xmath729 are infinitesimals and only the double integral over @xmath417 and @xmath418 contributes .
then , from eq .
( [ eq : ball004juy ] ) , the persistent part of the total thermoballistic density , @xmath730 , is seen to reduce to the total equilibrium density @xmath731 [ see eq .
( [ eq : ball004jux ] ) ] , @xmath732 with @xmath731 expressed in terms of the functions @xmath507 and @xmath506 , @xmath733 .
\label{eq : ball004jum}\ ] ] further , since @xmath734 in the drift - diffusion regime , the relaxing part of the joint total ballistic density , @xmath735 [ see eq .
( [ eq : netbal02xy ] ) ] , vanishes .
consequently , we have @xmath736 for the relaxing part of the total thermoballistic density , @xmath737 , and hence @xmath738 for the total thermoballistic density @xmath701 in the drift - diffusion limit .
the persistent part of the thermoballistic spin - polarized density , @xmath702 , is obtained from eq .
( [ eq : ball004juy ] ) as @xmath739 , \label{eq : ball004juz}\ ] ] and the relaxing part of the thermoballistic spin - polarized density , @xmath703 , from eq .
( [ eq : netbal02xy ] ) as @xmath740 a(x ) \nonumber \\ & = & \frac{n_{c}}{4 } b_{+}(x ) q^{2}(x ) a(x ) , \label{eq : ball004jua}\end{aligned}\ ] ] with @xmath741 given by eq . ( [ eq : spfra17c ] ) .
then , re - expressing the spin - polarized equilibrium density @xmath742 [ see eq .
( [ eq : ball004jux ] ) ] in the form @xmath743 , \label{eq : ball004jvz}\ ] ] we have @xmath744 for the thermoballistic spin - polarized density @xmath745 in the drift - diffusion limit .
the net total ballistic current @xmath746 is found from eq .
( [ eq : ball004jvx ] ) , by expanding to first order in @xmath747 and @xmath748 , in the form @xmath749 where @xmath750 now , evaluating the double integral over @xmath417 and @xmath418 in eq .
( [ eq : thebal01 ] ) for @xmath751 and taking the limit @xmath752 at fixed @xmath753 , we are left with a factor @xmath754 , so that the drift - diffusion limit of the total thermoballistic current @xmath696 is given by @xmath755 here , we have used relation ( [ eq : ballim20a ] ) to introduce the electron mobility @xmath61 , and we have identified the constant , total thermoballistic current in the drift - diffusion regime with the ( conserved ) total physical current @xmath133 . then , using eqs .
( [ eq : ball004jum ] ) and ( [ eq : ball004jvz ] ) , we can express @xmath133 in terms of the equilibrium densities @xmath756 [ see eqs .
( [ eq : ball004jup ] ) and ( [ eq : ball004jut ] ) ] in the form @xmath757 .
\label{eq : ballim20cc}\end{aligned}\ ] ] for zero spin splitting , @xmath758 , this expression becomes equivalent to expression ( [ eq : drude008 ] ) for the total current in the standard drift - diffusion model .
however , by contrast with the latter model , we have obtained eq .
( [ eq : ballim20cc ] ) without invoking the einstein relation ( [ eq : drude011 ] ) . this feature can be traced back to the probabilistic description underlying the thermoballistic approach ( see sec . [
sec : probappr ] ) , which allows the diffusion current to be directly expressed in terms of the collision time ( and hence of the momentum relaxation length @xmath7 ) [ see eq .
( [ eq : prob006 ] ) ] . to obtain an explicit expression for the mean spin function @xmath507 , and hence for the average chemical potential @xmath503
, we observe that in the drift - diffusion regime the total thermoballistic density @xmath701 is equal to the total equilibrium density @xmath731 [ see eq .
( [ eq : ball004jup ] ) ] , which , in turn , is related to @xmath507 via eq .
( [ eq : ball004jum ] ) . using eq .
( [ eq : ballim30aha ] ) to solve eq .
( [ eq : ballim20 ] ) for @xmath759 , integrating over the interval @xmath71 $ ] , and using eqs . ( [ eq : eqden60])([eq : spfra17c ] ) to simplify the integrals , we can express @xmath503 in the form @xmath760 \nonumber \\ & - & \frac{\beta}{4 } \int_{x_{1}}^{x } dx ' \frac{d \delta ( x')}{dx ' } q^{2}(x ' ) a(x ' ) , \nonumber \\ \label{eq : ballim20kf}\end{aligned}\ ] ] which generalizes expression ( [ eq : diff003 ] ) .
integration over the interval @xmath73 $ ] leads to another expression for @xmath503 , which is different in form , but numerically equal to expression ( [ eq : ballim20kf ] ) .
the drift - diffusion form of the spin accumulation function @xmath506 in expression ( [ eq : ballim20kf ] ) is determined by a differential equation [ see eq .
( [ eq : ballim26xx ] ) below ] .
we now set @xmath761 in eq .
( [ eq : ballim20kf ] ) and identify the boundary values of @xmath503 at the interface positions @xmath75 with the values , @xmath76 , of the equilibrium chemical potential at the contact side of the contact - semiconductor interfaces , @xmath762 [ see eq .
( [ eq : diff005 ] ) ] , so that from eq .
( [ eq : eqden03zk ] ) @xmath763 similarly , we identify the boundary values of @xmath506 with external values @xmath509 , @xmath764 we then obtain the drift - diffusion form of the current - voltage characteristic , which , for simplicity , is written down here for the case of constant spin splitting , when @xmath765 , @xmath766 , and @xmath767 , @xmath768 } \nonumber \\ & \ & \times [ 1 - e^{- \beta e v } + { \mbox{$\frac{1}{2}$ } } e^{- \beta \mu_{1 } } p ( a_{1 } - a_{2 } ) ] .
\label{eq : diff006hsa}\end{aligned}\ ] ] here , we have used eqs .
( [ eq : diff004 ] ) and ( [ eq : diff008 ] ) , respectively , to introduce the voltage bias @xmath29 and the effective sample length @xmath97 . for zero spin splitting , when @xmath769 and @xmath770 , expressions ( [ eq : ballim20kf ] ) and ( [ eq : diff006hsa ] ) become equivalent to the expressions ( [ eq : diff003 ] ) and ( [ eq : diff006 ] ) , respectively , in the standard drift - diffusion model . the drift - diffusion limit of the persistent thermoballistic spin - polarized current @xmath698 immediately follows from eqs .
( [ eq : netbal06 ] ) and ( [ eq : ballim20 ] ) as @xmath771 so that @xmath772 for zero spin splitting .
proceeding in analogy to the derivation of @xmath696 , we find the drift - diffusion limit of the relaxing thermoballistic spin - polarized current @xmath699 from eqs .
( [ eq : netbal02 ] ) and ( [ eq : thebal01 ] ) in the form @xmath773 using eq .
( [ eq : ball004jua ] ) , we can express @xmath699 in terms of the analogous thermoballistic density @xmath703 , @xmath774 , \label{eq : ballim2010}\ ] ] where @xmath775 so that @xmath776 for zero spin splitting .
inserting expressions ( [ eq : ballim21 ] ) and ( [ eq : ball004jua ] ) for @xmath699 and @xmath703 , respectively , in the general balance equation ( [ eq : ballim24new ] ) , we obtain a homogeneous second - order differential equation for the spin accumulation function @xmath506 , @xmath777 where @xmath778 is the spin diffusion length .
equation ( [ eq : ballim26xx ] ) can be converted into an analogous equation for the density @xmath703 , @xmath779 - \frac{1}{l^{2}_{s } } \check{\mathfrak{n}}_{-}(x ) = 0 , \nonumber \\ \label{eq : ballim25}\end{aligned}\ ] ] where eq .
( [ eq : ball004jua ] ) has been used . for zero spin splitting
, this equation generalizes , by including arbitrarily shaped potential energy profiles , the drift - diffusion equation commonly used to describe electric - field effects in spin - polarized transport in semiconductors [ see eq . ( 2.8 ) of ref . ; in that equation , the spin diffusion length appears as @xmath780 , with @xmath60 an effective diffusion coefficient and @xmath624 the spin relaxation time ] .
note that , in contrast to eq .
( [ eq : ballim26xx ] ) for @xmath506 , eq .
( [ eq : ballim25 ] ) for @xmath781 contains terms proportional to the first and second derivatives of @xmath48 [ see eq .
( [ eq : ballim24xy ] ) ] . in the strictly _ ballistic limit _
, when @xmath725 , there are no points of local thermodynamic equilibrium in the interval @xmath83 $ ] , and the average chemical potential @xmath782 is not defined inside this interval . expression ( [ eq : thebal01 ] ) reduces to @xmath783 i.e. , the thermoballistic currents and densities are given by the corresponding expressions for the net ballistic currents and joint ballistic densities ( see sec . [
sec : netball ] ) , evaluated at @xmath784 and @xmath785 . in expressions ( [ eq : ball004jvx ] ) , ( [ eq : netbal02 ] ) , ( [ eq : ball004juy ] ) , and ( [ eq : netbal02xy ] ) , the boundary values of the mean spin function , @xmath786 , and those of the spin accumulation function , @xmath787 , are to be identified with the corresponding values in the contacts [ see eqs .
( [ eq : ballim05ups ] ) and ( [ eq : ballim05s ] ) , respectively ] . from eq .
( [ eq : ball004jvx ] ) , we then obtain the total thermoballistic current @xmath696 in the form @xmath788 , \nonumber \\ \label{eq : ballim02}\end{aligned}\ ] ] where @xmath789 with @xmath790 defined by eq .
( [ eq : ballim05ups ] ) , and @xmath791 for zero spin splitting , expression ( [ eq : ballim02 ] ) becomes equivalent to expression ( [ eq : ball008 ] ) for the total current in the ballistic transport model .
the ballistic limit of the persistent thermoballistic spin - polarized current @xmath698 is obtained from @xmath792 by replacing in expression ( [ eq : ballim02 ] ) the quantity @xmath793 with @xmath794 [ see eqs .
( [ eq : emtra09 ] ) and ( [ eq : netbal06 ] ) ] . for the total thermoballistic density @xmath701
, we have , defining @xmath795 and using eqs .
( [ eq : netbal03ax])([eq : netbal02xy ] ) , @xmath796 \nonumber \\ & \ & + { \mbox{$\frac{1}{2}$ } } { \cal d}^{m}_{+}(x_{1},x_{2};x ) { \cal a}^{+}_{12}(x_{1 } , x_{2 } ; x)\ } .
\label{eq : ballim03}\end{aligned}\ ] ] for zero spin splitting , this becomes equivalent to expression ( [ eq : ball007za ] ) .
further , using eq .
( [ eq : netbal02 ] ) and ( [ eq : netbal02xy ] ) , we express the relaxing thermoballistic spin - polarized current @xmath699 and the corresponding density @xmath703 as @xmath797 and @xmath798 respectively . in view of eq .
( [ eq : baleq01zs ] ) , the current @xmath699 and density @xmath703 in the ballistic limit are trivially connected by the general balance equation ( [ eq : ballim24new ] ) .
the stochastic equilibration of the electrons that occurs during their motion across the sample is associated with the dissipation of energy , i.e. , the net transfer of energy out of the ensemble of conduction band electrons into a reservoir ( `` heat bath '' ) of electrons in thermodynamic equilibrium . to describe this transfer within the thermoballistic concept
, we introduce the ( conserved ) ballistic _ energy _
currents @xmath799 generated by thermal electron emission at the end - points @xmath417 and @xmath418 , respectively , of the ballistic interval @xmath313 $ ] . neglecting spin degrees of freedom , we have for the current @xmath800 , by an obvious modification of expression ( [ eq : ball004 ] ) for the electron current @xmath801 , writing @xmath802 , @xmath803 } \nonumber \\ & \ & \times \theta \bm { ( } \epsilon - e_{b}^{l}(x ' , x '' ) \bm { ) } , \label{eq : thec01}\end{aligned}\ ] ] where , in analogy to eq .
( [ eq : diff007 ] ) , @xmath804 is the maximum barrier height of the potential energy profile @xmath48 relative to its value at @xmath417 . evaluating the integral in expression ( [ eq : thec01 ] ) , we obtain , in extension of eq .
( [ eq : ball004f ] ) , @xmath805 e^{- \beta [ e_{c}^{m}(x ' , x '' ) - \tilde{\mu}(x ' ) ] } .
\label{eq : thec03}\end{aligned}\ ] ] for the ballistic energy current @xmath806 , we have @xmath807 in line with eq .
( [ eq : ball004rfd ] ) for the corresponding ballistic electron current , and for the net ballistic energy current in the interval @xmath313 $ ] , @xmath808 in line with eq .
( [ eq : ball004zgx ] ) for the net ballistic electron current .
now , we again introduce a `` reference coordinate '' @xmath37 inside the sample , whose meaning , however , differs from that of the coordinate @xmath37 in the definition ( [ eq : thebal01 ] ) of the thermoballistic currents and densities .
while in that definition , @xmath37 is a coordinate _ inside ballistic intervals _ , we here consider @xmath37 a coordinate characterizing the position of a _ point of local thermodynamic equilibrium_. more precisely , we consider a collection of such points , with density @xmath809 , in an interval @xmath226 centered about @xmath37 , in which ballistic energy currents are absorbed from , and emitted towards , either side .
the equilibrium point @xmath37 , where incoming electron currents are completely equilibrated and the outgoing currents are solely determined by the parameters of the reservoir ( i.e. , by the chemical potential ) , dynamically separates the two sample partitions to the left and right of @xmath37 .
we denote by @xmath810 the ( kinetic ) energy transferred per unit volume and unit time _ into _ the reservoir by absorption of electrons at the point @xmath37 , and correspondingly by @xmath811 the change in energy of the reservoir due to thermal electron emission _ out of _ it .
the _ net _ energy @xmath812 ,
i.e. , the energy dissipated locally at the point @xmath37 , is then given by @xmath813 the energies @xmath810 and @xmath811 are each composed of two parts ( see also fig .
[ fig:6 ] ) .
one part of @xmath810 is given by the weighted sum of the energy currents @xmath814 over all ballistic intervals @xmath313 $ ] lying to the left of @xmath37 and having their right end at @xmath815 , where the currents are absorbed .
the other part is expressed analogously in terms of the currents @xmath816 in ballistic intervals to the right of the absorption point @xmath37 .
for the energy @xmath817 dissipated in an interval @xmath226 centered around @xmath37 , we then have @xmath818 ) + \mathbb{f}_{2}(x ; [ \mathbb{e}^{r } ] ) \ } .
\label{eq : thec00kcg}\ ] ] here , the functionals @xmath819)$ ] and @xmath820)$ ] are given , for an arbitrary ballistic current @xmath821 , by eqs .
( [ eq : thec00ohx ] ) and ( [ eq : thec00gma ] ) , respectively .
the energy current @xmath822 is obtained from eq .
( [ eq : thec00kcg ] ) by interchanging the role of @xmath814 and @xmath823 , @xmath824 ) + \mathbb{f}_{2}(x ; [ \mathbb{e}^{l } ] ) \ } , \label{eq : thec00tuz}\ ] ] which comprises all ballistic energy currents emitted at the point @xmath37 towards either side . for the dissipated energy @xmath812
, we then have @xmath825 ) + \mathbb{f}_{2}(x ; [ \mathbb{e } ] ) \ } , \label{eq : thec00phy}\ ] ] where the net ballistic energy current @xmath826 is given by eq .
( [ eq : thec00gjb ] ) .
we note that in obtaining expression ( [ eq : thec00phy ] ) for @xmath812 we have _ not _ drawn on the thermoballistic energy current @xmath827 that results from identifying in eq .
( [ eq : thebal01 ] ) the function @xmath710 with net ballistic energy current @xmath826 .
naively , one might expect that the dissipated energy @xmath812 can be represented by the derivative @xmath828 given by eq .
( [ eq : thebal06jb ] ) .
however , comparing expressions ( [ eq : thec00phy ] ) and ( [ eq : thebal06jb ] ) , one observes that the contributions of the ballistic energy currents from the two sides of @xmath37 add up in the former , and are subtracted from one another in the latter .
the result ( [ eq : thec00phy ] ) can be obtained from the thermoballistic energy current @xmath827 in the form @xmath829 at variance with expression ( [ eq : thebal06jb ] ) .
we do not write down here the general expression for @xmath812 in terms of the average chemical potential @xmath782 obtained by using expressions ( [ eq : thec03])([eq : thec00gjb ] ) in eq .
( [ eq : thec00phy ] ) , and confine ourselves to considering the ballistic limit and the drift - diffusion regime . owing to the overall factor @xmath830 in the right - hand side of eq .
( [ eq : thec00phy ] ) , @xmath812 vanishes in the _ ballistic limit _ @xmath725 , which reflects the complete absence of equilibration in this limit . in the _ drift - diffusion regime _ , on the other hand , when @xmath831 , only the integral terms contribute .
we evaluate these terms for the special case of a homogeneous sample subjected to a constant external electric field of magnitude @xmath485 directed antiparallel to the @xmath37-axis , so that the potential energy profile has the form @xmath832 where @xmath833 for this profile , the average chemical potential @xmath503 is found from eqs .
( [ eq : diff003 ] ) and ( [ eq : diff006 ] ) , using the relation @xmath834 , to run parallel to the profile , @xmath835 + e_{c}(x ) .
\label{eq : the03b}\ ] ] therefore , the total equilibrium electron density is constant , @xmath836 .
the net ballistic energy current @xmath826 can now be expressed as @xmath837 } \left [ 1 + \epsilon ( x '' - x ' ) \right ] \ } \nonumber \\ & \approx & \frac{v_{e}}{2 \beta } n(x_{1 } ) \epsilon^{2 } ( x''- x')^{2 } , \label{eq : the03c}\end{aligned}\ ] ] where the second , approximate equation holds in the zero - bias limit @xmath838 . inserting the approximate representation of @xmath826 in the integral terms of expression ( [ eq : thec00phy ] ) , we find for the locally dissipated energy @xmath812 in the drift - diffusion regime @xmath839 where we have introduced the conductivity @xmath55 via eqs .
( [ eq : drude006 ] ) and ( [ eq : ballim20a ] ) . for
the total energy @xmath840 dissipated per unit time ( `` heat production '' ) in a sample with cross - sectional area @xmath841 and resistance @xmath842 , we then recover the ohmic expression @xmath843 .
we conclude this section with a synopsis of the concept underlying the thermoballistic description of charge carrier transport in semiconductors , in which we briefly comment on its basic ingredients and elucidate the physical content of its formal structure .
this will be followed by an assessment of its merits as well as its weaknesses .
the basic ingredients of the ( semiclassical ) thermoballistic concept are the ballistic carrier currents and densities which are constructed within the following framework .
( i ) thermal emission of carriers occurs at points of local thermodynamic equilibrium randomly distributed over the sample .
( ii ) the equilibrium points link `` ballistic transport intervals '' across which the emitted carriers move ballistically under the influence of potential energy profiles arising from internal and external electrostatic potentials . during their ballistic motion ,
the carriers undergo spin relaxation controlled by a ballistic spin relaxation length @xmath633 .
( iii ) at the end - points of the ballistic intervals , instantaneous `` point - like '' thermalization ( `` absorption '' ) of the carriers takes place . here
, we invoke the picture of `` reflectionless contacts''@xcite , according to which the ballistic carriers that enter a contact ( representing a `` bath '' with an effectively infinite number of transverse modes ) are completely absorbed there , after having been emitted from a similar contact at the opposite end of the sample . in the thermoballistic concept ,
the emission and absorption of carriers are treated in this way at _ all _ points of local thermodynamic equilibrium inside the sample . at these points ,
the absorption of carriers into the bath is complete ( `` reflectionless '' ) , but at the same instant carriers are emitted out of the bath into the ballistic intervals on either side.the opposing collision - free currents emitted at either end into the ballistic interval combine to form a _ net ballistic current _ inside the interval , with an associated _ joint ballistic density_. the random distribution of points of local thermodynamic equilibrium is mirrored in a random partitioning of the length of the sample into ballistic intervals , where each partition defines a `` ballistic configuration '' , a central notion of the concept .
the ballistic carrier currents and densities in these configurations are assembled to form the corresponding _ thermoballistic currents and densities_. these are constructed , at a reference position @xmath37 located arbitrarily inside the sample , by performing weighted summations , with weights controlled by a momentum relaxation length @xmath7 , over the net ballistic currents and joint ballistic densities in all ballistic intervals containing the point @xmath37 .
the thermoballistic currents and densities constitute the key element of the thermoballistic concept .
the physical content of the thermoballistic transport mechanism can be exhibited by analyzing the underlying formalism with regard to the intertwined effects of thermal electron emission and ballistic motion .
let us consider expressions ( [ eq : ball004jvx])([eq : netbal02xy ] ) for the net ballistic currents and joint ballistic densities , which are essentially composed of two factors each .
on the one hand , they contain the nonlocal barrier factors @xmath844 and @xmath845 in the expressions for the total and persistent spin - polarized currents and the relaxing spin - polarized current , respectively , and the factors @xmath846 and @xmath847 in the analogous expressions for the densities .
these factors describe the collision - free motion of the electrons across the ballistic interval @xmath313 $ ] , which is essentially determined by the potential energy profiles @xmath848 inside the interval .
they represent the _ ballistic attribute _ of thermoballistic transport . the factors in brackets , on the other hand , contain terms depending on the average chemical potentials @xmath545 , @xmath849 and the spin accumulation functions @xmath504 , @xmath505 , which are directly related to the spin - resolved chemical potentials @xmath850 , @xmath851 at the end - points @xmath512 of the ballistic interval . these factors , which describe the thermal emission ( `` thermal activation '' ) of the ballistic currents at the points of local thermodynamic equilibrium at @xmath417 and @xmath418 , respectively , represent the _ thermal attribute _ of thermoballistic transport . the term in brackets is the `` activation term '' .
[ the joint appearance of ballistic and thermal attributes shows that the term `` thermoballistic '' indeed provides an appropriate characterization of our approach . ] the contributions of the ballistic currents and densities ( [ eq : ball004jvx])([eq : netbal02xy ] ) , summarily denoted by @xmath710 , to the corresponding thermoballistic currents and densities @xmath689 are to be read from eq .
( [ eq : thebal01 ] ) : they are given by the current ( or density ) @xmath710 in the interval @xmath313 $ ] , multiplied by the probability @xmath852 that the electrons traverse this interval ballistically .
the momentum relaxation length @xmath7 controls the magnitude of the ballistic contribution to the entire transport process . at the same time
, it determines the average number of collisions , @xmath853 , in a sample of length @xmath28 . in the _ ballistic limit _
, when @xmath725 and there is no point of local thermodynamic equilibrium inside the sample , the transport is purely ballistic between the end - points @xmath69 and @xmath70 , and only the first term on the right - hand side of eq .
( [ eq : thebal01 ] ) contributes . in the opposite limit ,
when @xmath854 , the points of local thermodynamic equilibrium at which the electrons are equilibrated , lie infinitesimally close to one another .
then only the double integral in eq .
( [ eq : thebal01 ] ) survives , and we arrive at @xmath855 for the total thermoballistic current @xmath856 [ see eq .
( [ eq : ballim20 ] ) ] .
this current is essentially given in terms of equilibrium quantities , a property that characterizes the _ drift - diffusion limit_. expression ( [ eq : syn02 ] ) has the form of the current in the standard drift - diffusion approach [ see eqs .
( [ eq : drude006f ] ) and ( [ eq : diff001 ] ) ] , @xmath857 where @xmath61 is the electron mobility given by eq .
( [ eq : ballim20a ] ) , and @xmath731 the equilibrium electron density given by eq .
( [ eq : ball004jum ] ) . in that approach ,
the momentum relaxation length @xmath7 is nonzero ( so that @xmath61 remains nonzero ) , but small compared with the length scales over which the other parameters vary appreciably .
the activation term reduces to a derivative , and the relation between the total thermoballistic current and the average chemical potential becomes a local one .
the principal merit of the thermoballistic concept is that it allows to establish a consistent and transparent formalism for bridging , within the semiclassical approximation , the gap between the standard drift - diffusion and ballistic descriptions of charge carrier transport in semiconductors . while incorporating basic features of these descriptions , the thermoballistic concept consistently unifies and generalizes them by introducing random partitionings of the sample length into ballistic configurations .
the concept is transparent in a twofold way .
first , as shown above , a lucid interpretation of its physical content can be given in terms of ballistic and thermal attributes .
second , owing to its semiclassical character,@xcite the concept allows the effects of the different parameters describing a semiconducting system to be clearly distinguished . in the implementation of this concept , explicit equations for various transport quantities
can be derived , and simple solutions can be obtained in important special cases . the merit of transparency of the thermoballistic concept carries with it some simplifications and weak points which , however , in many cases can be remedied , albeit at the cost of increased complexity : they are not detrimental to the concept as a whole .
while the formulation of the full thermoballistic concept given here describes semiclassical transport in nondegenerate semiconducting systems , we have demonstrated within the prototype model how effects of electron tunneling and degeneracy can be taken into account .
quantum interference effects in the electron motion are not treated explicitly , but they may be assumed to be implicitly incorporated via an extended interpretation of the mean free path ( or momentum relaxation length ) @xmath7 , which from the outset has been taken as a phenomenological parameter . by treating it formally as the average distance that the carriers travel without collision between points of _ complete _ thermodynamic equilibrium , as in the relaxation time approximation,@xcite it simulates the effects of incomplete equilibration due to elastic or inelastic impurity scattering , of dimensionality , and , in the extreme , of quantal phase correlations .
indeed , in this work the momentum relaxation length @xmath7 is the determining parameter in which a great diversity of detail is subsumed .
it is introduced as a constant , so it must include in an average way the effect of spatial variations in the internal and external parameters characterizing the semiconducting system;in particular , this constant is chosen to be independent of the potential energy profile , to which , however , it should be related in a self - consistent way . moreover , the choice of the momentum relaxation length @xmath7 , rather than the relaxation time @xmath8 , is also merely one of convenience for the stationary treatment in this work .
we may work with position - dependent momentum relaxation lengths , but this would increase the complexity of the formalism and obscure the general line of argument .
the spin relaxation mechanism in terms of the ballistic spin relaxation length @xmath633 is again a phenomenological one , having certain similarities with the dyakonov - perel mechanism.@xcite we assume spin relaxation to occur only during the ballistic electron motion ; however , simultaneous spin and momentum relaxation could be taken into account by introducing additional terms in the spin balace equation . in principle , the thermoballistic concept allows a fully three - dimensional treatment of bipolar carrier transport to be implemented .
however , in the present paper , in order to keep the formalism manageable , we work within a narrowed framework .
first , we confine ourselves to unipolar transport , dealing specifically with electron transport in a spin - split conduction band .
second , we consider three - dimensional `` plane - parallel '' samples whose parameters ( in particular , the average density of the scattering centers associated with impurities ) do not vary in the directions perpendicular to the transport direction ( the @xmath37-direction ) .
nevertheless , electron transport in this kind of sample depends on the number of dimensions , @xmath17 , via `` no - scattering probabilities '' @xmath858 [ see sec .
ii of ref . ] . here , in order to be able to write down physically transparent formulae , we use one - dimensional no - scattering probabilities of the form @xmath859 ( see sec .
[ sec : probappr ] ) .
having presented , in the preceding section , the concept underlying the thermoballistic approach , we now turn to the implementation of this concept .
we begin by establishing the physical conditions from which the algorithms for calculating the dynamical functions , _ viz . _ , the average chemical potential @xmath503 [ via the mean spin function @xmath507 ] and the spin accumulation function @xmath506 , are developed .
thereafter , these algorithms will be described in detail .
we call _ physical conditions _
( i ) a relation introduced to connect the total thermoballistic current @xmath696 with the cognate total physical current @xmath133 , which allows us to calculate the function @xmath507 , and ( ii ) an assumption concerning the detailed spin relaxation mechanism , which leads to the determination of the function @xmath506 .
the current @xmath861 , owing to its construction in terms of ballistic currents averaged over random ballistic configurations ( with weights controlled by the momentum relaxation length @xmath7 ) , is to be interpreted as an `` ensemble average '' of the electron current at the point @xmath37 .
this average is spatially varying , and , therefore , it is the _ spatial average _ of the ensemble average over the length of the sample which is to be identified with the ( constant ) total physical current @xmath862 inside the sample . by current conservation at @xmath69 and @xmath70 , this current is equal to the total current @xmath133 in the left and right leads , so that we have @xmath863 in this condition , the current @xmath864 is defined at a ballistic point @xmath37 located inside an ensemble of ballistic intervals @xmath313 $ ] , where @xmath417 and @xmath418 , unlike @xmath37 , are points of local thermodynamic equilibrium [ see eqs .
( [ eq : thebal00 ] ) and ( [ eq : thebal01 ] ) ] .
the integration over @xmath37 starts at the fixed equilibrium point @xmath69 and ends at the fixed equilibrium point @xmath70 .
now , just as we have introduced the _
ballistic _ reference point @xmath37 , we consider here an _
point of reference , i.e. , a point of local thermodynamic equilibrium anywhere inside the sample , which , again , is labeled by the coordinate @xmath37 @xmath865 .
such a point acts in the same way as the fixed equilibrium points @xmath69 and @xmath70 .
the ballistic current entering @xmath37 , say , from the left , is completely absorbed , whereupon a thermal current is instantaneously emitted to either side of @xmath37 .
the same happens to the current entering the equilibrium point @xmath866 from the right .
however , in contrast to the `` true '' , externally controlled equilibrium points @xmath75 , which are located at the contact side of the contact - semiconductor interfaces , no sharvin - type interface resistance ( see sec .
[ sec : sharvin ] below ) appears at @xmath37 . within this scheme
, we introduce the thermoballistic current @xmath867 , with @xmath868 , where @xmath37 is an equilibrium point , and @xmath869 a ballistic point . in conformance with eq .
( [ eq : thebal04abc ] ) , the spatial average of this current over the range @xmath71 $ ] is again , by current conservation at @xmath69 , equal to the physical current in the left lead , so that @xmath870 analogously , we have for the thermoballistic current @xmath871 in the range @xmath73 $ ] , using similar arguments as above , @xmath872 ( @xmath873 ) . equations ( [ eq : thebal04ohg ] ) and ( [ eq : thebal04zul ] ) , when expressed in terms of the function @xmath507 via eqs .
( [ eq : ball004jvx ] ) and ( [ eq : thebal01 ] ) , lead to two different volterra - type integral equations with solutions @xmath874 and @xmath875 , respectively .
trivially , by satisfying eqs .
( [ eq : thebal04ohg ] ) and ( [ eq : thebal04zul ] ) , these solutions also satisfy condition ( [ eq : thebal04abc ] ) .
they define , for given spin accumulation function @xmath506 , two different total equilibrium densities @xmath876 and @xmath877 , respectively , via eq .
( [ eq : ball004jum ] ) .
the total equilibrium density at @xmath37 is obtained as the mean value of these , @xmath878 , \label{eq : thebal05lg}\ ] ] so that , using eq .
( [ eq : ball004jum ] ) again , we have for the unique ( thermoballistic ) mean spin function @xmath879 .
\label{eq : thebal05pk}\ ] ] the physical conditions ( [ eq : thebal04ohg ] ) and ( [ eq : thebal04zul ] ) , together with eq .
( [ eq : thebal05lg ] ) , determine the procedure for calculating the average chemical potential @xmath503 . using the mean value ( [ eq : thebal05lg ] ) as the point of departure for constructing @xmath503 reflects the fact that this function expresses an intrinsic property of the semiconducting sample , with no preference for one or the other of the sample ends at @xmath69 and @xmath70 . with @xmath507 given by eq .
( [ eq : thebal05pk ] ) , it follows from eqs .
( [ eq : ball004jvx ] ) and ( [ eq : thebal01 ] ) that the unique thermoballistic current @xmath864 is given by @xmath880 , \nonumber \\ \label{eq : thebal04lfs}\end{aligned}\ ] ] a symmetric combination as in eq .
( [ eq : thebal05pk ] ) .
we note that in ref .
we have applied a different procedure for constructing the average chemical potential @xmath503 . there
, the combination of the functions @xmath874 and @xmath875 was not chosen to be symmetric as in eq .
( [ eq : thebal05pk ] ) , but was determined by the requirement that the values of the unique total thermoballistic current at the two sample ends be equal , @xmath881 this condition derives from postulating that the position dependence of the total thermoballistic current is compensated by that of a `` background current'',@xcite such that these currents add up to the ( conserved ) total physical current .
the background current is assumed to be fed by sources and sinks whose effect averages out to zero when the current is integrated over the length of the sample . in the present work
, this hypothesis has been abandoned for being unphysical . in the thermoballistic transport mechanism
, we assume that spin relaxation takes place only _ inside _ the ballistic intervals ( see sec . [ sec : balance ] ) , and that an infinitesimal shift of the end - points of the ballistic intervals does not affect the current @xmath699 .
accordingly , we set the term arising from differentiating the limits of integration in expression ( [ eq : thebal01 ] ) for @xmath699 equal to zero , @xmath882 from eq .
( [ eq : thebal06 ] ) , the relaxing thermoballistic spin - polarized current and density are then seen to be connected by the balance equation @xmath883 which is of the same form as the balance equation ( [ eq : baleq01zs ] ) connecting the relaxing spin - polarized current and density in the individual ballistic intervals .
the _ spin accumulation function _
@xmath506 , and hence the relaxing thermoballistic spin - polarized current @xmath699 , are determined by condition ( [ eq : thebal06rt ] ) . when written in terms of @xmath506 by using expression ( [ eq : netbal02 ] ) in eq .
( [ eq : thebal06jb ] ) , this condition turns into a linear , inhomogeneous fredholm - type integral equation of the second kind@xcite for @xmath506
. the explicit form of this equation and its conversion , in a specific case of particular importance , into a differential equation will be the subject of sec .
[ sec : accufunc ] .
the average chemical potential @xmath782 is determined , via expression ( [ eq : thebal05pk ] ) for the mean spin function @xmath507 , by the solutions @xmath884 of the integral equations ( [ eq : thebal04ohg ] ) and ( [ eq : thebal04zul ] ) , respectively , which are conveniently solved in terms of `` resistance functions '' .
we begin by considering eq .
( [ eq : thebal04ohg ] ) . to solve it for the function @xmath874 ,
we define functions @xmath885 and @xmath886 via @xmath887 and @xmath888 where the spin accumulation function @xmath506 is assumed to be given .
furthermore , we introduce the `` resistance function '' @xmath889 [ the choice of this name will be substantiated in sec .
[ sec : currvolt ] below ] as @xmath890 along with the function @xmath891 \label{eq : inteq02ax}\ ] ] [ for given @xmath506 , @xmath892 is a given function as well ] . now , expressing the net total ballistic current @xmath746 given by eq .
( [ eq : ball004jvx ] ) in terms of the functions @xmath889 and @xmath892 , we obtain from eq .
( [ eq : thebal01 ] ) the total thermoballistic current @xmath893 to be inserted in eq .
( [ eq : thebal04ohg ] ) in the form @xmath894 here , @xmath895 \nonumber \\ & \ & \hspace{-2.2 cm } + \int_{\xi^{+}}^{x } \frac{dx''}{l } w_{+}(x_{1 } , x '' ; l ) [ \tilde{\mathfrak{r}}_{1}(x '' ) - \tilde{\mathfrak{r}}_{1}(x_{1 } ) ] \nonumber \\ & \ & \hspace{-2.2 cm } + \left .
\int_{x_{1}}^{\xi^{- } } \frac{dx'}{l } \int_{\xi^{+}}^{x } \frac{dx''}{l } w_{+}(x ' , x '' ; l ) [ \tilde{\mathfrak{r}}_{1}(x '' ) - \tilde{\mathfrak{r}}_{1}(x ' ) ] \right\ } , \nonumber \\ \label{eq : thebal01tzg}\end{aligned}\ ] ] while the term @xmath896 is obtained from expression ( [ eq : thebal01tzg ] ) by replacing @xmath889 with @xmath892 throughout , and @xmath897 with @xmath898 , where @xmath899 to evaluate condition ( [ eq : thebal04ohg ] ) , we use the relation @xmath900 to carry out the integration over @xmath869 , so that the condition can be expressed in the explicit form @xmath901 here , the inhomogeneity @xmath902 is given by @xmath903 and the integral kernels @xmath904 are defined as @xmath905 with @xmath906 equation ( [ eq : inteq04ftx ] ) is a linear , inhomogeneous , volterra - type integral equation of the second kind@xcite in the range @xmath907 . the kernels @xmath908 depend on the potential energy profiles @xmath909 solely via the quantities @xmath654 , and are independent of the total physical current @xmath133 .
nonetheless , for nonzero spin splitting , a @xmath133-dependence of the resistance function @xmath889 can arise , via the function @xmath892 , from a nonlinear @xmath133-dependence of the spin accumulation function @xmath506 [ see secs .
[ sec : accufunc ] and [ sec : currden ] below ] . in the zero - bias limit ,
when @xmath506 is proportional to @xmath133 , the function @xmath892 , and hence @xmath889 , become independent of @xmath133 . for zero spin splitting ,
when @xmath910 , we have @xmath911 and @xmath889 does not depend on @xmath506 .
the resistance function @xmath889 is discontinuous at @xmath912 : we have @xmath913 from eq .
( [ eq : inteq02 ] ) , whereas we obtain @xmath914 by expanding eq .
( [ eq : inteq04ftx ] ) to first order in @xmath915 ( we assume the potential energy profiles @xmath916 to be continuous in the interval @xmath83 $ ] ) . for arbitrary functions @xmath48 and @xmath917 , the calculation of the resistance function @xmath889 for a chosen parameter set consists of _ three , consecutive _ steps : ( i ) calculation of the spin accumulation function @xmath506 as solution of the integral equation ( [ eq : spiacc02 ] ) [
see sec .
[ sec : accufunc ] below ] , using the boundary conditions ( [ eq : ballim05s ] ) with given values @xmath509 at the contact - semiconductor interfaces ; ( ii ) with @xmath506 as input in expression ( [ eq : inteq02ax ] ) , calculation of the inhomogeneity @xmath918 from eq .
( [ eq : inteq03 ] ) ; ( iii ) using @xmath902 in the volterra equation ( [ eq : inteq04ftx ] ) , calculation of @xmath889 by , in general , numerical methods . turning now to the function @xmath919
, we proceed as for @xmath885 . in analogy to eq .
( [ eq : inteq04ftx ] ) for the resistance function @xmath889 , we introduce a resistance function @xmath920 as @xmath921 for which we obtain from eq .
( [ eq : thebal04zul ] ) a volterra - type integral equation in the range @xmath922 , @xmath923 with the inhomogeneity
@xmath924 expressed in terms of the function @xmath925 \label{eq : inteq02im}\ ] ] as @xmath926 the kernels @xmath927 are obtained by replacing in expression ( [ eq : inteq06opa ] ) for @xmath928 the pair of arguments @xmath929 with @xmath930 .
the resistance function @xmath931 is discontinuous at @xmath932 , @xmath933 @xmath934 in analogy to the discontinuity of the function @xmath889 at @xmath935 [ see eqs .
( [ eq : inteq08 ] ) and ( [ eq : inteq09 ] ) ] . as to the calculation of @xmath931 from eq .
( [ eq : inteq04iz ] ) , the foregoing discussion regarding the calculation of @xmath889 from eq .
( [ eq : inteq04ftx ] ) applies _ mutatis mutandis_. alternatively , one may calculate @xmath931 by using the relation @xmath936 where @xmath937 is the solution of eq .
( [ eq : inteq04ftx ] ) corresponding to spatially reversed potential energy profiles , @xmath938 using as input a spin accumulation function @xmath506 calculated from eq .
( [ eq : spiacc02 ] ) below with the reversed profiles and with the boundary values @xmath939 and @xmath940 interchanged . the calculation of the resistance functions @xmath889 and @xmath931 simplifies considerably if the spin splitting is constant over the sample , @xmath941 [ while @xmath48 still may be arbitrary ] . with @xmath942
we have @xmath943 and hence for the functions @xmath654 from eqs .
( [ eq : ball004u ] ) and ( [ eq : emtra09 ] ) , using eqs .
( [ eq : eqden06])([eq : spfra17c ] ) , @xmath944 and @xmath945 here , we have defined @xmath946 @xmath947 and @xmath948 is the static spin polarization
. considering , for instance , the calculation of @xmath889 , we then find from eqs .
( [ eq : inteq06opa ] ) for the kernels @xmath904 , using expressions ( [ eq:66x ] ) and ( [ eq:66y ] ) in eq .
( [ eq : thebal02 ] ) , @xmath949 and @xmath950 where the reduced kernel @xmath951 corresponds to @xmath952 .
now , inserting expressions ( [ eq : inteq06wuf ] ) and ( [ eq : inteq06waf ] ) in eqs .
( [ eq : inteq04ftx ] ) and ( [ eq : inteq03 ] ) , respectively , we can rewrite the integral equation ( [ eq : inteq04ftx ] ) in the form @xmath953 @xmath954 , where @xmath955 , \label{eq : inteq09wu}\ ] ] and the function @xmath892 is independent of @xmath956 ( see sec . [
sec : intequat ] below ) . as a solution of eq .
( [ eq : inteq04aha ] ) , the function @xmath957 is a universal function that describes an intrinsic property of the semiconducting sample ; it is determined by the potential energy profile @xmath48 and the momentum relaxation length @xmath7 , and does not depend on the spin splitting @xmath956 and the spin accumulation function @xmath506 .
the resistance function @xmath889 for constant spin splitting can now be expressed as @xmath958 the calculation here separates into _ two , independent _ steps : ( i ) solution of eq .
( [ eq : inteq04aha ] ) for @xmath957 ; ( ii ) calculation of @xmath506 as solution of the integral equation ( [ eq : spiacc02 ] ) and determination of @xmath892 via eqs . ( [ eq : inteq04lil ] ) and ( [ eq : inteq02ax ] ) . in particular , for zero spin splitting , we have @xmath959 introducing the mean value @xmath960 of the functions @xmath961 , which both satisfy condition ( [ eq : thebal04abc ] ) , @xmath962 \label{eq : inteq09plb}\ ] ]
[ see eq .
( [ eq : thebal05pk ] ) ] , we observe that @xmath960 satisfies condition ( [ eq : thebal04abc ] ) as well .
using eqs .
( [ eq : inteq02 ] ) and ( [ eq : spiche02 ] ) , we can express @xmath960 in the form @xmath963 - { \mbox{$\frac{1}{2}$ } } [ \tilde{\mathfrak{r}}_{1}(x ) - \tilde{\mathfrak{r}}_{2}(x ) ] .
\label{eq : spiche08gs}\end{aligned}\ ] ] now , writing down this expression for @xmath935 and @xmath761 , respectively , and using eqs .
( [ eq : inteq04scp ] ) , ( [ eq : inteq08 ] ) , and ( [ eq : spiche05 ] ) together with the boundary conditions ( [ eq : ballim05ups ] ) , we add the resulting two expressions to obtain for the quantity @xmath964 defined by eq .
( [ eq : ballim04 ] ) , @xmath965 \ } .
\label{eq : spiche16}\end{aligned}\ ] ] on the other hand , subtracting the two expressions , we have @xmath966 with the parameter @xmath967 defined as @xmath968 .
\label{eq : spiche15}\ ] ] then , using eqs .
( [ eq : spiche16 ] ) and ( [ eq : spiche14 ] ) , we can eliminate from expression ( [ eq : spiche08gs ] ) the dependence on the boundary values @xmath969 and @xmath970 of the functions @xmath961 . introducing the function @xmath971 , \nonumber \\
\label{eq : spiche18}\end{aligned}\ ] ] so that , in view of eqs .
( [ eq : inteq08 ] ) and ( [ eq : spiche05 ] ) , @xmath972 we find for the mean spin function @xmath507 , using eq .
( [ eq : inteq04scp ] ) , @xmath973 for @xmath974 , and @xmath975 equivalently , we write @xmath976 for @xmath974 , and @xmath977 expression ( [ eq : spiche17zg ] ) represents the final , general expression for the ( local ) thermoballistic average chemical potential @xmath503 inside the sample [ henceforth , we drop the attribute `` thermoballistic '' when referring to @xmath978 .
it is seen to be uniquely determined by the values of the _ external _ parameters at the contact sides of the contact - semiconductor interfaces , _ viz . , _ the values @xmath76 of the equilibrium chemical potential and the values @xmath509 of the spin accumulation function [ which enter via the resistance functions @xmath889 and @xmath931 ] . comparing expression ( [ eq : spiche17zg ] ) to expression ( [ eq : prochem06 ] ) for the chemical potential @xmath38 in the prototype thermoballistic model
, we observe a formally identical structure , but differences in the explicit forms of the functions @xmath389 and @xmath980 and of the reduced resistances @xmath279 and @xmath981 .
an explicit expression for @xmath503 can be obtained if @xmath982 ( `` ballistic regime '' ) , in which case the probability for an electron to suffer a collision when traversing the sample of length @xmath28 , @xmath983 , is vanishingly small .
then , keeping terms of order @xmath853 , we have from eq .
( [ eq : inteq04ftx ] ) @xmath984 and similarly from eq .
( [ eq : inteq04iz ] ) @xmath985 since from eqs .
( [ eq : ballim05s ] ) , ( [ eq : inteq04scp ] ) , ( [ eq : inteq02ax ] ) , and ( [ eq : inteq02im ] ) @xmath986 we now find from eq .
( [ eq : spiche15 ] ) @xmath987 for the function @xmath988 , we have from eq .
( [ eq : spiche18 ] ) , using eq .
( [ eq : inteq53cb ] ) , @xmath989 with expressions ( [ eq : inteq53pl ] ) and ( [ eq : spiche18fv ] ) inserted in eq .
( [ eq : spiche17zg ] ) , we obtain the average chemical potential @xmath503 in explicit form . in the particular case of zero spin
splitting , when @xmath990 and @xmath991 , we have @xmath992 .
\label{eq : spiche17ld}\ ] ] this result , here obtained in the ballistic regime , @xmath982 , contrasts with the ( strict ) ballistic limit , @xmath725 , considered in sec .
[ sec : ballim ] , when there is absolutely no point of local thermodynamic equilibrium inside the sample . in this extreme case
, no meaning can be attached to an average chemical potential , and there is nothing a calculation of such a quantity can be based upon . in ref . , we have presented and discussed numerical results for @xmath503 , calculated with the potential energy profile ( [ eq : difeq01 ] ) for values of the ratio @xmath993 ranging between @xmath994 to @xmath995 . if calculated with the procedure adopted in the present work ( see the remarks at the end of sec .
[ sec : detmsf ] ) , the results for @xmath503 would differ quantitatively from the former ones , but would agree with those in all qualitative respects .
the discontinuities in the resistance functions @xmath889 and @xmath931 at the points of local thermodynamic equilibrium @xmath935 and @xmath761 , respectively , give rise to discontinuities in the function @xmath988 both at @xmath996 and @xmath761 . from eq .
( [ eq : spiche18 ] ) , we find , using eqs .
( [ eq : inteq08 ] ) and ( [ eq : inteq09 ] ) , @xmath997 and similarly , using eqs .
( [ eq : spiche05 ] ) and ( [ eq : spiche06 ] ) , @xmath998 therefore , according to eq .
( [ eq : spiche17zg ] ) , the average chemical potential @xmath503 exhibits discontinuities at these points as well ( `` sharvin effect '' ; see ref . ) , @xmath999 } - 1 & = & - \frac{\eta_{12}^{- } e^{- \beta \mu_{1 } } } { 2 } \frac{\tilde{\mathfrak{r}}_{1 } ( x_{1}^{+})}{\tilde{\mathfrak{r } } } \nonumber \\ & = & - { \mbox{$\frac{1}{2}$ } } \beta e^{2 } j \tilde{\rho}_{1 } , \label{eq : sharv03}\end{aligned}\ ] ] @xmath1000 } - 1 & = & \frac{\eta_{12}^{- } e^{- \beta \mu_{2 } } } { 2 } \frac{\tilde{\mathfrak{r}}_{2 } ( x_{2}^{-})}{\tilde{\mathfrak{r } } } \nonumber \\ & = & { \mbox{$\frac{1}{2}$ } } \beta e^{2 } j \tilde{\rho}_{2 } , \label{eq : sharv04}\end{aligned}\ ] ] where we have used eqs .
( [ eq : spiche14 ] ) and ( [ eq : spiche15 ] ) to introduce the total physical current @xmath133 .
the quantities @xmath1001 are the _ sharvin interface resistances_,@xcite @xmath1002 @xmath1003 for zero spin splitting , when @xmath1004 , so that @xmath1005 and @xmath1006 , we can express @xmath1001 in the form @xmath1007 where @xmath1008 } \label{eq : sharv05tf}\end{aligned}\ ] ] are the total equilibrium electron densities at the semiconductor side of the contact - semiconductor interfaces [ see eqs .
( [ eq : ball004jux ] ) and ( [ eq : ball004jum ] ) ] . equations ( [ eq : sharv03 ] ) and ( [ eq : sharv04 ] ) , respectively , are analogous to eqs .
( [ eq : prochem06amf ] ) and ( [ eq : prochem06pky ] ) for the discontinuities of the chemical potential @xmath352 in the prototype thermoballistic model .
the _ current - voltage characteristic _ of the thermoballistic transport model is obtained from eq .
( [ eq : spiche14 ] ) in the form @xmath1009 where @xmath1010 is the voltage bias between the metal contacts ( we assume the voltage drop across the contacts to be negligibly small ) . in the zero - bias limit , when @xmath1011 , we have @xmath1012 for the resistance times cross - sectional area of the sample . comparing the current - voltage characteristic ( [ eq : sharv06 ] ) to that of the ( spinless ) prototype thermoballistic model , eq .
( [ eq : gendr004a ] ) , one sees that the ( dimensionless ) parameter @xmath967 directly corresponds to the reduced resistance @xmath279 of the prototype model , eq .
( [ eq : gendr003 ] ) .
therefore , @xmath967 is here also called the _ reduced resistance _ of the sample , and the functions @xmath889 and @xmath931 from which it is derived , the _ resistance functions_. the reduced resistance @xmath967 is a central element of the thermoballistic description of electron transport in semiconductors .
it comprises the effect of the detailed shape of the potential energy profiles as well as that of the momentum relaxation length .
moreover , via its dependence on the spin accumulation function @xmath506 , it takes into account the effect of spin relaxation . as exemplified by expression ( [ eq : inteq53pl ] ) for @xmath967 , which holds in the ballistic regime ,
the reduced resistance depends on the total current @xmath133 unless @xmath506 is proportional to @xmath133 .
the explicit form of @xmath967 in the drift - diffusion regime can be obtained by rewriting eq .
( [ eq : ballim20kf ] ) for @xmath761 in the form of a current - voltage characteristic and comparing it to the characteristic ( [ eq : sharv06 ] ) .
defining now the _ relative magnetoresistance _ @xmath1013 of the semiconducting sample@xcite as @xmath1014 where @xmath1015 is the resistance at zero external magnetic field , i.e. , at zero spin splitting , we have from eq . ( [ eq : sharv07tf ] ) @xmath1016 the reduced resistance at zero spin splitting , @xmath1017 , is obtained by solving eq .
( [ eq : inteq04aha ] ) for the function @xmath957 and the analogous equation for @xmath1018 and using eq .
( [ eq : spiche15 ] ) , @xmath1019 .
\label{eq : spiche15axa}\ ] ] for _ constant _ splitting , we find from eq .
( [ eq : spiche15 ] ) , using eq .
( [ eq : inteq09wq ] ) and the analogous equation for @xmath931 as well as eq .
( [ eq : inteq53cb ] ) , @xmath1020 and hence @xmath1021 for the magnetoresistance .
the net relaxing ballistic spin - polarized current @xmath508 [ see eq .
( [ eq : netbal02 ] ) ] and the joint relaxing ballistic spin - polarized density @xmath681 [ see eq .
( [ eq : netbal02xy ] ) ] , and thus also the corresponding thermoballistic current @xmath699 and density @xmath703 evaluated from eq .
( [ eq : thebal01 ] ) , are dynamically determined solely by the spin accumulation function @xmath506 .
the calculation of this function is , in general , prerequisite to a complete determination of the average chemical potential @xmath782 [ see sec .
[ sec : resisfunc ] ] .
following sec .
[ sec : thermcurr ] , we invoke condition ( [ eq : thebal06rt ] ) , which is equivalent to the spin balance equation ( [ eq : ballim24new ] ) , and insert expression ( [ eq : netbal02 ] ) for the quantity @xmath710 in eq .
( [ eq : thebal06jb ] ) to obtain for the spin accumulation function @xmath506 an integral equation of the form @xmath1022 where @xmath1023 and @xmath1024 here , the values @xmath787 of the spin accumulation function at the interface positions @xmath75 have been set equal to the external values , @xmath509 , at the contact side of the contact - semiconductor interfaces [ see eq .
( [ eq : ballim05s ] ) ] . the function @xmath1025 is defined as @xmath1026 [ see eq .
( [ eq : baleq06 ] ) ] , and @xmath1027 with @xmath655 given by eq .
( [ eq : spirel15 ] ) . with the first two terms in eq .
( [ eq : spiacc02 ] ) acting as an inhomogeneity , this equation is a linear , inhomogeneous , fredholm - type integral equation of the second kind@xcite for the spin accumulation function @xmath506 .
the corresponding homogeneous equation is solved by @xmath1028 only , so that the solution @xmath506 of eq .
( [ eq : spiacc02 ] ) for @xmath1029 is linear and homogeneous in @xmath939 and @xmath940 . just like the average chemical potential @xmath503 , the spin accumulation function @xmath506 exhibits the sharvin effect ,
i.e. , it is not , in general , continuous at the interfaces with the contacts , @xmath1030 .
this will be demonstrated in sec .
[ sec : diffequa ] below by way of a specific example .
the solution of eq .
( [ eq : spiacc02 ] ) is found , in general , numerically by applying matrix methods after discretization . for constant spin splitting , @xmath1031 , when the functions @xmath1032 are given by eq .
( [ eq:66ixy ] ) , the function @xmath655 defined by eq .
( [ eq : spirel15 ] ) reduces to @xmath1033 with @xmath1034 given by eq .
( [ eq:66asa ] ) , and @xmath1035 by eq .
( [ eq:66blg ] ) .
further , using eqs .
( [ eq:66hjc ] ) and ( [ eq:66ixy ] ) in expression ( [ eq : ball004n ] ) , we see that @xmath956 cancels out in the functions @xmath1036 , @xmath1037 inserting this in eq .
( [ eq : ball004p ] ) , we then find , via eqs .
( [ eq : balden15 ] ) , ( [ eq : baleq03 ] ) , and ( [ eq : baleq06 ] ) , the function @xmath1025 defined by eq .
( [ eq : spiacc06 ] ) to be independent of @xmath956 , @xmath1038 with this result , in conjunction with eqs . ( [ eq : thebal03 ] ) and ( [ eq : thebal03ihi ] ) , used in eqs .
( [ eq : spiacc03 ] ) and ( [ eq : spiacc04 ] ) , the spin splitting @xmath956 is seen to drop out from the integral equation ( [ eq : spiacc02 ] ) , i.e. , its solution @xmath506 does not depend on @xmath956 . in _ homogeneous _ semiconductors without space charge ,
the conduction band edge potential is constant . in this case , the integral equation for @xmath506 can be converted into a differential equation . assuming the electrons to be driven by an external electric field of magnitude @xmath485 directed antiparallel to the @xmath37-axis , the potential energy profile ( [ eq : eqden01 ] ) for zero spin splitting , @xmath1039 , is given by eq .
( [ eq : difeq01 ] ) .
[ according to the above discussion , the inclusion of a nonzero , constant spin splitting would not alter the results . ] for the profile ( [ eq : difeq01 ] ) , the function @xmath1040 defined by eq .
( [ eq : baleq03 ] ) takes the form @xmath1041^{1/2}\bm { ) } \nonumber \\ \label{eq : difeq02}\end{aligned}\ ] ] @xmath1042 , as can be shown by using eqs .
( [ eq : ball004p ] ) , ( [ eq : ball004n ] ) , and ( [ eq : balden15 ] ) . using expression ( [ eq : difeq02 ] ) , via eq .
( [ eq : baleq06 ] ) , in eq .
( [ eq : spiacc06 ] ) , the integral equation ( [ eq : spiacc02 ] ) can now be reduced to @xmath1043 where @xmath1044 x } , \label{eq : difeq05}\ ] ] @xmath1045 x } , \label{eq : difeq06}\ ] ] and @xmath1046 \ } .
\label{eq : difeq07}\ ] ] the function @xmath1047 defined by @xmath1048 obeys the relations @xmath1049 , @xmath1050 for @xmath1051 , and @xmath1052 for @xmath1053 .
we can convert the inhomogeneous integral equation ( [ eq : difeq04 ] ) into a homogeneous integrodifferential equation for @xmath506 by supplementing eq .
( [ eq : difeq04 ] ) with the equations obtained by forming its first and second derivative with respect to @xmath37 , and subsequently eliminating from this set of three equations the boundary values @xmath939 and @xmath940 .
[ in principle , a similar procedure could also be applied to the general equation ( [ eq : spiacc02 ] ) , but this does not seem to lead to any advantage . ] if we replace in the integral equation ( [ eq : difeq04 ] ) the function @xmath1047 with a position - independent average value @xmath1054 , the functions @xmath1055 and @xmath1056 become pure exponentials , and the corresponding integrodifferential equation reduces to a homogeneous second - order differential equation of the form @xmath1057 here , @xmath1058 , \label{eq : difeq10}\ ] ] @xmath1059 , \label{eq : difeq11}\ ] ] and @xmath1060 \ } , \label{eq : difeq12}\ ] ] where @xmath1061 and @xmath1062 with the ballistic spin relaxation length @xmath633 given by eq .
( [ eq : spirel01 ] ) .
[ note that expression ( [ eq : difeq12 ] ) for the function @xmath1063 differs from the corresponding expression ( 3.48 ) of ref . , which was derived by introducing the average value @xmath1054 in the integrodifferential equation , rather than in the original integral equation .
] since , owing to the presence of the factor @xmath1064 in the functions @xmath1055 and @xmath1056 , only the values of @xmath1065 in the range @xmath1066 contribute appreciably , we may choose @xmath1054 as the average of @xmath1065 over an @xmath37-interval of length equal to the momentum relaxation length @xmath7 , @xmath1067 in the right - hand integral in this equation , the range of small @xmath1068 , when @xmath1069 , is emphasized because of the weight factor @xmath1070 . for large @xmath1071 ( i.e. , in the ballistic regime and/or for strong fields )
, the variation of @xmath1065 with @xmath37 becomes essential .
however , for the purpose of demonstrating the principal effects of the transport mechanism , the approximation in terms of a constant average value @xmath1054 of the function @xmath1072 , in conjunction with the choice ( [ eq : difeq15 ] ) for @xmath1054 , appears to be sufficiently accurate . in the drift - diffusion regime , when @xmath726 , @xmath727 , and @xmath1073 , we have @xmath1074 and hence @xmath1075 , where @xmath1076 equation ( [ eq : difeq09 ] ) then reduces to @xmath1077 where @xmath1078 is the spin diffusion length given by eq .
( [ eq : ballim26 ] ) . setting @xmath1079 in eq .
( [ eq : ballim26xx ] ) , we find eq .
( [ eq : difeq16 ] ) to agree with the former equation , which was obtained by directly evaluating the thermoballistic current @xmath699 and density @xmath703 in the drift - diffusion limit , and which leads to the standard form ( [ eq : ballim25 ] ) of the drift - diffusion equation for spin - dependent transport . in the zero - bias limit
@xmath1080 , the integral equation ( [ eq : difeq04 ] ) reduces to @xmath1081 from which we obtain the differential equation @xmath1082 here , @xmath1083 is the generalization of the spin diffusion length @xmath1078 given by eq .
( [ eq : ballim26 ] ) .
the length @xmath475 becomes equal to the latter length , @xmath1084 , in the drift - diffusion regime , when @xmath727 , and to the ballistic spin relaxation length , @xmath1085 , in the ballistic case , when @xmath1086 and hence @xmath1087 . for @xmath974 , the general solution of eq .
( [ eq : difeq18 ] ) reads @xmath1088 inserting this expression for @xmath506 in eq .
( [ eq : difeq17 ] ) and setting @xmath935 and @xmath761 , we obtain , using eq .
( [ eq : ballim05s ] ) , two linear equations for the coefficients , with the solution @xmath1089 , \label{eq : difeq21}\ ] ] @xmath1090 .
\label{eq : difeq22}\ ] ] here , @xmath1091 , \label{eq : difeq23}\ ] ] with @xmath1092 it then follows from eqs .
( [ eq : difeq20])([eq : difeq24 ] ) that @xmath506 is discontinuous at @xmath935 and @xmath761 , @xmath1093 @xmath1094 where @xmath1095 \leq 1 , \label{eq : difeq27}\ ] ] with @xmath1096 in the drift - diffusion regime , when @xmath1084 and @xmath1097 , we have @xmath1098 and in the ballistic case , when @xmath1085 and @xmath1099 , @xmath1100 .
\label{eq : difeq30}\ ] ] at @xmath935 , for example , the discontinuity of @xmath506 is @xmath1101 in the drift - diffusion regime , and @xmath1102 $ ] in the ballistic case .
the position dependence of the current and density spin polarizations as well as the magnetoresistance are the physical quantities of principal interest in the study of spin - polarized electron transport in paramagnetic semiconducting systems . in the thermoballistic approach ,
we define the _ persistent _ current spin polarization in the semiconducting sample , @xmath1103 , in terms of the persistent thermoballistic spin - polarized current @xmath698 and the total physical current @xmath133 as @xmath1104 @xmath1105 .
analogously , the _ relaxing _ current spin polarization @xmath1106 is defined in terms of the relaxing thermoballistic spin - polarized current @xmath699 and the total current @xmath133 as @xmath1107 for the _ total _ current spin polarization @xmath1108 , we then have @xmath1109 [ note that in refs . and , we have defined the current spin polarization in terms of the total _ thermoballistic _
current @xmath696 , rather than in terms of the total _ physical _ current @xmath133 . ] the _ persistent _ density spin polarization @xmath1110 and the _ relaxing _ density spin polarization @xmath1111 are introduced by replacing in expressions ( [ eq:65 ] ) and ( [ eq:65ax ] ) the thermoballistic currents @xmath698 and @xmath699 with the respective densities @xmath702 and @xmath703 , and the current @xmath133 in the denominator of those expressions with the thermoballistic joint total density @xmath701 .
hence , we obtain @xmath1112 for the _ total _ density spin polarization @xmath1113 . in expression ( [ eq:91xx ] ) for the total current spin polarization inside the semiconductor , the current @xmath698 depends on the reduced spin accumulation function @xmath1114 , and hence linearly on the boundary values @xmath1115 , via the dependence of the total ballistic current @xmath500 on @xmath1114 , @xmath1116 [ see eqs .
( [ eq : ball004jvx ] ) and ( [ eq : netbal06 ] ) ; here , we have assumed @xmath500 to be expressed in terms of @xmath1114 , rather than @xmath506 , using eq .
( [ eq : spfra01fv ] ) ] .
the relaxing thermoballistic spin - polarized current @xmath699 is determined by the ballistic current @xmath508 given by eq .
( [ eq : netbal02 ] ) , which is a linear - homogeneous functional of @xmath1114 . hence , @xmath699 is linear and _ homogeneous _ in the boundary values @xmath1115 , so that we can write @xmath1117 , \nonumber \\
\label{eq:91x}\end{aligned}\ ] ] with ( dimensionless ) `` formfactors '' @xmath1118 and @xmath1119 .
then , we can express the values of the total current spin polarization at the semiconductor side of the interfaces at @xmath75 as @xmath1120 .
\nonumber \\ \label{eq:66c}\end{aligned}\ ] ] the boundary values @xmath1115 of the reduced spin accumulation function , which are undetermined as yet , are to be expressed in terms of the current @xmath133 and the parameters of the system by matching the current spin polarization in the semiconductor to the corresponding polarization in the contacts ( see sec .
[ sec : match ] below ) . in order to keep the formulation sufficiently general , we assume the semiconducting sample to be connected to ( semi - infinite ) _ ferromagnetic _ metal contacts , which are treated as fully degenerate fermi systems . in the left contact located in the range @xmath1121 ,
the spin - resolved chemical potentials @xmath1122 have the form@xcite @xmath1123 here , @xmath1124 is the spin diffusion length , and @xmath1125 are the ( @xmath37-independent ) spin - up and spin - down bulk conductivities , respectively . setting @xmath1126 in eqs .
( [ eq:50 ] ) , we obtain @xmath1127 with the spin - resolved currents @xmath1128 given by @xmath1129 we then find for the current spin polarization in the left contact , using @xmath1130 , @xmath1131 here , @xmath1132 is the bulk ( current or density ) spin polarization , and the parameter @xmath1133 which has the dimension of interface conductance , characterizes the spin - dependent transport in the ferromagnet . for the current spin polarization in the right contact located in the range @xmath1134
, we have @xmath1135 which follows by replacing in eq .
( [ eq:50 ] ) the coordinate @xmath37 with @xmath1136 , and the labels `` l '' attached to the parameters in eqs .
( [ eq:50])([eq:57xyz ] ) with `` r '' . when spin - selective interface resistances are absent , the chemical - potential splitting @xmath724 is continuous at the interfaces , @xmath1137 @xmath1138 where @xmath1139 are its values at the contact side of the interfaces .
the latter values are to be identified with the corresponding values @xmath1140 at the semiconductor side of the interface , @xmath1141 where the right - hand equations have been obtained by solving eq .
( [ eq : spfra01hl ] ) for @xmath549 , setting @xmath1142 , and defining @xmath1143 for the boundary values of the reduced spin accumulation function @xmath551 . with eqs .
( [ eq:57zxl])([eq:68 ] ) used in eqs .
( [ eq:57lala ] ) and ( [ eq:57zzz ] ) , respectively , we now have for the current spin polarizations at the contact side of the interfaces @xmath1144 @xmath1145 which are to be identified with the corresponding polarizations @xmath1146 at the semiconductor side of the interfaces [ see eq .
( [ eq:66zara ] ) below ] .
spin - selective interface resistances @xmath1147 give rise to discontinuities of the spin - resolved chemical potentials on the contact sides of the interfaces.@xcite at @xmath935 , for example , the discontinuity has the form @xmath1148 the interface resistances @xmath1149 are located between @xmath1126 and @xmath69 in the ferromagnetic contact , adjacent to the sharvin interface resistance @xmath1150 [ see eq .
( [ eq : sharv05 ] ) ] between @xmath935 and @xmath1151 in the semiconductor .
the quantity @xmath1152 to be substituted in eq .
( [ eq:57lala ] ) is obtained from eqs .
( [ eq:50 ] ) , ( [ eq : zeropol01 ] ) , and ( [ eq:57xyz ] ) as @xmath1153 \right\ } .
\label{eq : karol}\end{aligned}\ ] ] the same procedure applies _ mutatis mutandis _ to the interface at @xmath74 .
the connection of @xmath1139 with the interface values of the reduced spin accumulation function , @xmath1115 , is again given by eqs .
( [ eq:68 ] ) .
to evaluate the current and density spin polarizations all across the contact - semiconductor system , we have to relate the boundary values of the reduced spin accumulation function , @xmath1115 , to the total physical current @xmath133 and the internal parameters characterizing the system . to this end , we exploit the continuity of the total current spin polarization @xmath1108 across the contact - semiconductor interfaces at @xmath75 , @xmath1154 using expressions ( [ eq:57angi ] ) and ( [ eq:57guido ] ) for @xmath1155 and expressions ( [ eq:66c ] ) for @xmath1146 , we obtain a pair of coupled nonlinear equations for @xmath1115 , which we will not write down here in their general form . in the zero - bias limit , when @xmath1156 , we have @xmath1157 [ see eq .
( [ eq : spfra01hl ] ) ] , so that @xmath1158 and @xmath1159 become linear in @xmath1160 and @xmath1161 , respectively .
further , to first order , the term involving the reduced spin accumulation function @xmath1114 in the persistent ballistic spin - polarized current @xmath667 [ see eq .
( [ eq : netbal06 ] ) , with @xmath746 expressed in terms of @xmath1114 via eqs.([eq : spfra01hl ] ) and ( [ eq : ball004jhm ] ) ] can be neglected , so that this current becomes independent of @xmath1115 .
the same then holds for the corresponding thermoballistic current , @xmath698 , as well as for the persistent current spin polarization @xmath1103 derived therefrom .
the polarizations @xmath1146 are linear in @xmath1115 , and the coupled equations for @xmath1115 become linear . since @xmath1162 in the zero - bias limit
, they can be expressed , using eq .
( [ eq : ball004jux ] ) , in the form @xmath1163 \breve{a}_{1 } + v_{e } n_{+}(x_{1 } ) { \cal f}_{2}(x_{1 } ) \breve{a}_{2 } \nonumber \\ & \ & \hspace{-4.0 cm } = p_{l } j - \overstar{\mathfrak{j}}_{-}(x_{1 } ) , \label{eq:66e}\end{aligned}\ ] ] @xmath1164 \breve{a}_{2 } \nonumber \\ & \ & \hspace{-4.0 cm } = p_{r } j - \overstar{\mathfrak{j}}_{-}(x_{2 } ) .
\label{eq:66ezg}\end{aligned}\ ] ] the current @xmath698 is proportional to @xmath133 , so that the quantities @xmath1115 are , as solutions of eqs .
( [ eq:66e ] ) and ( [ eq:66ezg ] ) , proportional to @xmath133 as well .
hence , the relaxing thermoballistic spin - polarized current @xmath699 is also proportional to @xmath133 , so that the total current spin polarization @xmath1108 calculated from eq .
( [ eq:91xx ] ) is , in the zero - bias limit , independent of the total physical current @xmath133 .
summing up , we obtain the total current spin polarization along the entire heterostructure , @xmath1108 , as follows . in the metal contacts ,
it is given by expressions ( [ eq:57lala ] ) and ( [ eq:57zzz ] ) , respectively , with @xmath1152 and @xmath1165 expressed in terms of @xmath1115 via eqs .
( [ eq:57zxl])([eq:68 ] ) . in the semiconductor , it is given by expression ( [ eq:91xx ] ) in terms of the persistent , @xmath698 , and relaxing , @xmath699 , thermoballistic spin - polarized currents , where the formfactors @xmath1166 entering the latter current are determined by the solution of the integral equation ( [ eq : spiacc02 ] ) for the spin accumulation function @xmath506 .
the total density spin polarization @xmath1113 is obtained in an analogous way .
this section deals with the application of the thermoballistic approach to various specific examples .
the emphasis is on spin - polarized transport in heterostructures including nonmagnetic and magnetic semiconductors . in considering the case of a homogeneous semiconductor at zero bias
, we illustrate the thermoballistic formalism for a particularly simple example and , moreover , collect results needed for treating heterostructures involving homogeneous layers of nonmagnetic and magnetic semiconductors ( see secs .
[ sec : fmnmshet ] and [ sec : dmsnmshe ] below ) . for a homogeneous semiconducting layer at zero bias , the potential energy profiles @xmath497 are given by @xmath1167 @xmath1105 . specializing the development presented in sec .
[ sec : resf ] for constant spin splitting to the case @xmath1168 , we find from eq .
( [ eq : inteq06opa ] ) the reduced @xmath1169 kernel @xmath1170 in the explicit form @xmath1171 .
\label{eq : inteq06woft}\ ] ] from the corresponding solution @xmath957 of eq .
( [ eq : inteq04aha ] ) and the solution of the analogous equation for the function @xmath1018 , we then obtain for @xmath1172 and @xmath1173 , respectively , using eq .
( [ eq : inteq09wq ] ) , @xmath1174 for the resistance functions @xmath1175 of the homogeneous semiconductor ; at @xmath935 and @xmath761 , on the other hand , we have @xmath1176 ; see eqs .
( [ eq : inteq08 ] ) and ( [ eq : spiche05 ] ) . for the reduced resistance @xmath967
, we then find from eq .
( [ eq : spiche15 ] ) , using eq .
( [ eq : inteq53cb ] ) , @xmath1177 and for the function @xmath988 from eq .
( [ eq : spiche18 ] ) , using eqs .
( [ eq : ballim04st ] ) , ( [ eq : inteq04lil ] ) , ( [ eq : inteq02ax ] ) , and ( [ eq : inteq02im ] ) , @xmath1178 \label{eq:66zora}\end{aligned}\ ] ] for @xmath1029 , whereas from eq .
( [ eq : spiche15ug ] ) , @xmath1179 .
the spin accumulation function @xmath506 to be inserted in this expression is given by eqs .
( [ eq : ballim05s ] ) and ( [ eq : difeq20 ] ) . inserting expressions ( [ eq:66zick ] ) and ( [ eq:66zora ] ) in eq .
( [ eq : spiche17zg ] ) and using eq .
( [ eq : spiche14 ] ) , we find from eq .
( [ eq : ball004jvx ] ) for the net total ballistic current in a homogeneous semiconductor at zero bias @xmath1180 @xmath1181 for @xmath1182 , @xmath1183 for @xmath1184 , and @xmath1185 for @xmath1186 , in which the dependence on @xmath506 present in @xmath967 and @xmath988 has cancelled out completely . at this point , it is appropriate to note that the dependence of the ballistic current @xmath746 on the end - point coordinates @xmath512 exhibited here contradicts the basic assumption of the prototype thermoballistic model as expressed by eq .
( [ eq : gendr001qal ] ) .
now , using expressions ( [ eq:66jafz])([eq:66juhu ] ) in eq .
( [ eq : thebal01 ] ) , we find @xmath1187 i.e. , the total thermoballistic current is equal to the ( constant ) total physical current .
further , using eq .
( [ eq : netbal06 ] ) , we have @xmath1188 for the persistent thermoballistic spin - polarized current , and hence from eq .
( [ eq:65 ] ) for the persistent current spin polarization @xmath1189 thereby retrieving the static spin polarization .
turning now to the determination of the relaxing thermoballistic spin - polarized current @xmath699 in a homogeneous semiconductor at zero bias , we have from eq .
( [ eq : netbal02 ] ) for the corresponding ballistic current @xmath1190 \nonumber \\ \label{eq : netbal02ihk}\end{aligned}\ ] ] @xmath1191 . inserting this expression in eq .
( [ eq : thebal01 ] ) and subsequently using the derivative of eq .
( [ eq : difeq17 ] ) with respect to @xmath37 , we obtain @xmath699 in the form @xmath1192 [ see eq .
( 4.11 ) of ref . ;
the factor 2 appearing in the right - hand side of the latter equation again reflects the fact that the normalization of @xmath506 used there differs from that used in the present article ] . from eqs .
( [ eq:91xx ] ) , ( [ eq:66zyxv ] ) , and ( [ eq : netbal02fln ] ) , we now have for
the total current spin polarization @xmath1193 [ note that owing to its normalization to the total thermoballistic current , the expression for @xmath1108 given by eq .
( 136 ) of ref . differs from expression ( [ eq:66ze ] ) ; see the remark following eq .
( [ eq:91xx ] ) above . ] as to the density spin polarization in a homogeneous semiconductor , we have from eq .
( [ eq : ball004juy ] ) for the persistent part , @xmath1194 , of the joint total ballistic density @xmath1195 @xmath1196 for @xmath1182 , @xmath1197 for @xmath1184 , and @xmath1198 for @xmath1186 . for the relaxing part of the joint total ballistic density , @xmath1199
, we find from eq .
( [ eq : netbal02xy ] ) @xmath1200 so that @xmath1201 observing eq .
( [ eq:66trfc ] ) , we now use expressions ( [ eq:66kohj])([eq:66fsnb ] ) in eq .
( [ eq : thebal01 ] ) to obtain for the thermoballistic joint total density @xmath1202 with @xmath1203 given explicitly by eq .
( [ eq:66kohj ] ) . since @xmath1204 from eq .
( [ eq : ball004juy ] ) , we have for the persistent part of the joint ballistic spin - polarized density @xmath1205 and hence for the persistent density spin polarization @xmath1206 in agreement with the corresponding current spin polarization . by inserting expression ( [ eq : netbal02fln ] ) for the corresponding current @xmath699 in the balance equation ( [ eq : ballim24new ] ) and using eq .
( [ eq : difeq18 ] ) for the spin accumulation function @xmath506 , the relaxing thermoballistic spin - polarized density @xmath1207 for a homogeneous semiconductor is readily obtained in the form @xmath1208 so that we have for the total density spin polarization @xmath1209 with @xmath701 given by eq .
( [ eq:66gvya ] ) .
now , omitting in expression ( [ eq:66gvya ] ) for @xmath701 the term proportional to @xmath133 , so that @xmath1210 , we can use expression ( [ eq:66gxky ] ) for @xmath1113 in eq .
( [ eq:66ze ] ) to express the current spin polarization @xmath1108 in terms of @xmath1113 in the form @xmath1211 differentiating this equation and using eqs .
( [ eq : difeq18 ] ) and ( [ eq:66gxky ] ) , we then obtain @xmath1212 for the density spin polarization @xmath1113 in terms of @xmath1108 . for the magnetoresistance of a homogeneous semiconductor at zero bias , we find @xmath1213 by inserting the zero - field limit of expression ( [ eq:66zick ] ) , @xmath1214 for @xmath1017 in eq .
( [ eq : sharv07lz ] ) .
we now consider heterostructures formed of a _ homogeneous , nonmagnetic _ semiconducting ( nms ) layer and two _ ferromagnetic _ metal ( fm ) contacts .
the study of spin - polarized transport in this kind of structure , both experimentally and theoretically , marked the beginning of semiconductor spintronics@xcite ( for recent surveys of the physics of semiconductor - based spintronic devices , see refs . ) . for zero bias
, we evaluate the position dependence of the current and density spin polarizations using the results of secs .
[ sec : pollead ] , [ sec : match ] , and [ sec : homsem ] .
for nonzero bias , we obtain the spin polarizations `` injected '' from the contacts into the semiconductor , i.e. , the polarizations generated inside the semiconductor in the vicinity of either fm / nms interface regardless of the influence of the opposite interface . inside the semiconducting layer
, we have from eq .
( [ eq:66ze ] ) with @xmath1215 and @xmath1216 , using expression ( [ eq : difeq20 ] ) for the spin accumulation function @xmath506 , @xmath1217 , \nonumber \\ \label{eq:40c}\end{aligned}\ ] ] where @xmath24 is given by eq .
( [ eq : difeq24 ] ) .
the coefficients @xmath1218 can be expressed via eqs .
( [ eq : difeq21 ] ) and ( [ eq : difeq22 ] ) , and using eq .
( [ eq : spfra01hl ] ) , in terms of the values @xmath1115 of the reduced spin accumulation function @xmath1114 on the contact sides of the interfaces .
in order to determine the quantities @xmath1115 , we evaluate expression ( [ eq:40c ] ) on the semiconductor sides of the interfaces , obtaining @xmath1219 @xmath1220 here , we have introduced the total equilibrium density for zero spin splitting and zero bias , @xmath1221 which is constant inside the semiconducting layer .
the coefficients @xmath1222 and @xmath41 in expressions ( [ eq:61a ] ) and ( [ eq:61abc ] ) are given by eqs .
( [ eq : difeq27 ] ) and ( [ eq : difeq28 ] ) , respectively . using the continuity of @xmath1108 at the interfaces , eq .
( [ eq:66zara ] ) , we now equate expressions ( [ eq:57angi ] ) and ( [ eq:61a ] ) , and expressions ( [ eq:57guido ] ) and ( [ eq:61abc ] ) .
[ when spin - selective interface resistances are included , expression ( [ eq:57angi ] ) for @xmath1158 is to be replaced with the more general expression obtained from using expression ( [ eq : karol ] ) for @xmath1152 in eq .
( [ eq:57lala ] ) , and analogously for @xmath1159 .
] for zero bias , when @xmath1157 , this results in a system of coupled linear equations for @xmath1115 [ see eqs .
( [ eq:66e ] ) and ( [ eq:66ezg ] ) ] , @xmath1223 \breve{a}_{1 } - h \breve{a}_{2 } = \frac{2 j}{v_{e } n_{+}^{(0 ) } } p_{l } , \label{eq:69}\ ] ] @xmath1224 \breve{a}_{2 } = \frac{2 j}{v_{e } n_{+}^{(0 ) } } p_{r } , \label{eq:69abc}\ ] ] where @xmath1225 the solutions of eqs .
( [ eq:69 ] ) and ( [ eq:69abc ] ) are found to be @xmath1226 p_{l } - h p_{r } \ } , \label{eq:72abc}\ ] ] @xmath1227 p_{r } \ } , \label{eq:72}\ ] ] where @xmath1228 [ g + \tilde{g}_{r}^{(0 ) } ] - h^{2 } .
\label{eq:72ijk}\ ] ] expressions ( [ eq:72abc ] ) and ( [ eq:72 ] ) determine the values of the reduced spin accumulation function , @xmath1160 and @xmath1161 in terms of the current @xmath133 , of the polarizations @xmath1229 and @xmath1230 in the left and right ferromagnetic contact , respectively , and of material parameters , such as the conductivities @xmath1231 and the spin diffusion lengths @xmath1232 of the contacts [ via @xmath1233 $ ] and the momentum relaxation length @xmath7 and the spin relaxation length @xmath633 in the semiconducting sample as well as the sample length @xmath28 ( via @xmath1222 and @xmath41 ) and the ( constant ) total equilibrium electron density @xmath1234 .
since the quantities @xmath1115 are proportional to the total current @xmath133 , the current spin polarization @xmath1108 is independent of @xmath133 , while the density spin polarization @xmath1113 is proportional to @xmath133 .
the spin polarizations along the entire heterostructure are now obtained as follows . inside the semiconductor , the current spin polarization @xmath1108
is given by expression ( [ eq:40c ] ) , with @xmath1218 calculated from the solutions @xmath1115 of eqs .
( [ eq:72abc ] ) and ( [ eq:72 ] ) .
the corresponding density spin polarization @xmath1113 is then readily obtained from eq .
( [ eq:66pgmq ] ) .
the expressions for @xmath1108 in the ferromagnetic contacts are provided by eqs .
( [ eq:57lala ] ) and ( [ eq:57zzz ] ) , respectively , where the quantities @xmath1152 and @xmath1165 are calculated from eq .
( [ eq : karol ] ) and from its analogue for @xmath1165 , respectively .
we do not write down the density spin polarizations in the ferromagnets , but only mention that they do not , in general , match the polarizations @xmath1235 and @xmath1236 on the semiconductor sides of the interfaces . in refs .
and , we have presented results of detailed calculations for @xmath1108 at zero bias for typical parameter values of fm / nms / fm heterostructures , emphasizing the dependence of the polarization on the momentum and spin relaxation lengths , @xmath7 and @xmath633 , and on the spin - selective interface resistances @xmath1237 ( see sec . [
sec : pollead ] ) .
we define the `` injected spin polarization '' as the polarization inside the semiconductor in the vicinity of the interface , e.g. , at @xmath912 , generated by the bulk polarization @xmath1229 of the left ferromagnet regardless of the influence of the right ferromagnet .
more explicitly , we define the injected current and density spin polarizations as the polarizations @xmath1238 and @xmath1235 , respectively , in the limit of infinite sample length , @xmath1239 .
the injected spin polarizations at @xmath1240 provide the initial values of the left - generated polarizations in the semiconductor , which propagate into the region @xmath1172 while being degraded by the effect of spin relaxation .
we now consider the injected spin polarizations for electron transport in a homogenenous semiconductor , driven by an external electric field , i.e. , for a ( spin - degenerate ) potential energy profile of the form ( [ eq : difeq01 ] ) , with the parameter @xmath1241 defined by eq .
( [ eq : difeq01erd ] ) . in order to obtain the spin accumulation function @xmath506 for this case
, one has to solve eq .
( [ eq : difeq09 ] ) [ in general , numerically ] under the asymptotic condition @xmath1242 to determine the decay length @xmath1243 , we solve eq .
( [ eq : difeq09 ] ) in the range @xmath1244 in which the @xmath37-dependence of the coefficient functions @xmath1245 , @xmath1246 , and @xmath1063 arising from the function @xmath1247 can be disregarded .
this yields @xmath1248^{1/2 } .
\label{eq:80}\ ] ] since @xmath1249 , @xmath1243 is a real number .
the injected current spin polarization is obtained from eq .
( [ eq:91xx ] ) with @xmath772 .
we calculate the relaxing thermoballistic spin - polarized current at the interface , @xmath1250 , from the relation @xmath1251 , \label{eq:80ijhb}\ ] ] where the function @xmath1252 is given by eq .
( [ eq : spiacc04 ] ) .
this relation can be shown to hold , for arbitrary sample length @xmath28 and potential energy profiles @xmath564 , by ( i ) evaluating @xmath1250 from eq .
( [ eq : thebal01 ] ) with @xmath1253 [ see eq .
( [ eq : netbal02 ] ) ] and ( ii ) eliminating from the resulting expression the boundary value @xmath940 by using the integral equation ( [ eq : spiacc02 ] ) for @xmath935 .
for the case considered here , we write down the integral equation ( [ eq : difeq04 ] ) [ with @xmath1254 for @xmath935 and take the limit @xmath1255 to obtain the relation @xmath1256 where @xmath1257 relation ( [ eq:80lcvs ] ) fixes the normalization of the spin accumulation function @xmath506 in terms of the boundary value @xmath939 . then , using this relation in eq .
( [ eq:80ijhb ] ) with @xmath1258 , we can express the current @xmath1250 in the form @xmath1259 where the quantity @xmath1260 is independent of the normalization of @xmath506 , and @xmath1261 } .
\label{eq:82jpfc}\ ] ] in obtaining eq .
( [ eq:81 ] ) , we have used eq .
( [ eq : ball004jum ] ) , with @xmath769 , as well as eq .
( [ eq : spfra01hl ] ) . for the injected current spin polarization , we now have from eq .
( [ eq:91xx ] ) @xmath1262 which , by continuity , is equal to the polarization @xmath1263 at the contact side of the interface . then , equating the right - hand sides of eqs .
( [ eq:57angi ] ) and ( [ eq:81ijk ] ) , we obtain @xmath1264 this nonlinear equation for @xmath1160 is to be solved for given values of the parameters @xmath1241 , @xmath1229 , @xmath1265 , @xmath1266 , @xmath7 , and @xmath633 .
we now turn to the calculation of the injected density spin polarization from eq .
( [ eq:91olsb ] ) , where @xmath1267 and @xmath1268 in the present case . proceeding similarly as for the current @xmath1250
, we obtain the relaxing thermoballistic spin - polarized density at the interface , @xmath1269 , in the form @xmath1270 where @xmath1271 for the persistent total thermoballistic density at the interface , @xmath1272 , we find by evaluating eq .
( [ eq : thebal01 ] ) with @xmath1273 [ see eq .
( [ eq : ball004juy ] ) ] and taking the limit @xmath1239 @xmath1274 where the quantity @xmath1275 } \label{eq:82erkg}\ ] ] has been expressed in terms of the average chemical potential @xmath503 using eq .
( [ eq : eqden03zk ] ) .
the injected density spin polarization now follows as @xmath1276 where the boundary value @xmath1160 of the reduced spin accumulation function is again to be determined by solving eq .
( [ eq:69rst ] ) . the dependence of the injected current spin polarization , @xmath1238 , on the electric - field parameter @xmath1241 , for @xmath7-values ranging from the drift - diffusion to the ballistic regime and for zero as well as nonzero interface resistances @xmath1277 , has been studied numerically in ref . .
in the drift - diffusion regime , when @xmath727 and @xmath1278 , the spin accumulation function @xmath506 is given by the exponentially decreasing solution of eq .
( [ eq : difeq16 ] ) , @xmath1279 @xmath1280 , with the field - dependent spin diffusion length @xmath1281 given by @xmath1282 where @xmath1078 is the spin diffusion length defined by eq .
( [ eq : ballim26 ] ) . from eqs .
( [ eq:82 ] ) and ( [ eq:82pqr ] ) , respectively , we then have @xmath1283 and @xmath1284 and hence from eqs .
( [ eq:81ijk ] ) and ( [ eq:81zyx ] ) , respectively , @xmath1285 and @xmath1286 for the injected current and density spin polarizations in the drift - diffusion regime . in deriving the right - hand equation ( [ eq:81jslc ] ) , we have used eq .
( [ eq : diff006hsa ] ) , with @xmath1287 , @xmath1288 , and @xmath1289 , as well as eq .
( [ eq:82jpfc ] ) , to express the total current @xmath133 in the form @xmath1290 where @xmath1291 is the ( spin - summed ) conductivity of the semiconductor [ see eq .
( [ eq : drude006f ] ) ] .
the right - hand equation ( [ eq:81pfqn ] ) has been obtained by using eq .
( [ eq : ballim20kf ] ) in eq .
( [ eq:82erkg ] ) . the boundary value @xmath1160 is determined by the equation @xmath1292 which follows from eq .
( [ eq:69rst ] ) by replacing its right - hand side with the right - hand side of eq .
( [ eq:81jslc ] ) and inserting expression ( [ eq:81kalo ] ) for @xmath133 .
the effect of external electric fields on the current spin polarization injected at fm / nms interfaces has been studied within the standard drift - diffusion approach by yu and flatt@xcite .
comparing our description to that of these authors , we find that the field - dependent spin diffusion length @xmath1281 given by eq .
( [ eq:80hui ] ) agrees with the `` up - stream '' spin diffusion length @xmath1293 given by eq .
( 2.23b ) of ref . , provided the `` intrinsic '' spin diffusion length @xmath475 of that reference is identified with the spin diffusion length @xmath1294 of the present work .
then , identifying in eq .
( 3.5 ) of ref .
( with the interface resistances set equal to zero ) the spin - summed conductivity of the semiconductor , @xmath1295 , with @xmath1296 , and the spin injection efficiency at the interface , @xmath1297 , with @xmath1298 , and using eq .
( [ eq:57xyz ] ) , we observe the equivalence of the former equation with eq .
( [ eq:69hlbv ] ) .
consequently , the injected current and density spin polarizations of either work are actually identical . for arbitrary @xmath7 , we now consider the injected current spin polarization in the zero - bias limit , when @xmath1299 . here , the spin accumulation function @xmath506 is given by the exponentially decreasing solution of eq .
( [ eq : difeq18 ] ) , @xmath1300 @xmath1280 , where the generalized spin diffusion length @xmath475 is given by eq .
( [ eq : difeq19 ] ) . from eq .
( [ eq:82 ] ) , we then have @xmath1301 with @xmath24 given by eq .
( [ eq : difeq24 ] ) , and hence from the zero - bias limit of eq .
( [ eq:69rst ] ) @xmath1302 where @xmath1303 is the sharvin interface conductance [ see eq .
( [ eq : sharv05kl ] ) ] . inserting this in eq .
( [ eq:81ijk ] ) , we find @xmath1304 i.e. , for zero bias , the injected current spin polarization is independent of the total current @xmath133 . in the drift - diffusion regime , when @xmath727 , we have @xmath1305 so that @xmath1306 reduces to @xmath1307 where @xmath1308 with @xmath1296 defined by eq .
( [ eq:75zqsp ] ) , and @xmath1078 by eq .
( [ eq : ballim26 ] ) .
the quantity @xmath1309 is seen to be the semiconductor analogue of the `` interface conductance '' @xmath1265 characterizing the ferromagnet [ see eq .
( [ eq:57xyz ] ) ] . for typical values of the parameters of the fm / nms / fm system ( see , e.g. ,
ref . ) , the values of @xmath1265 exceed those of @xmath1309 by several orders of magnitude .
this indicates a `` conductance mismatch '' between ferromagnet and semiconductor,@xcite which gives rise to very low values of the injected spin polarizations at fm / nms interfaces .
[ to remedy this mismatch , one may introduce spin - selective interface resistances , for example , in the form of tunneling barriers.@xcite ] in the ballistic regime , when @xmath1310 , we have @xmath1311 so that @xmath1312 here , the sharvin interface conductance @xmath1313 takes the place of the quantity @xmath1309 in eq .
( [ eq:75 ] ) . as @xmath1313 is proportional to the density @xmath1266 , eq .
( [ eq:79y ] ) yields values for @xmath1238 close to the bulk spin polarization of the feromagnet , @xmath1229 , if @xmath1266 is sufficiently high . for typical parameter values,@xcite
however , the high donor densities needed to obtain the required high electron densities implies very small values of the momentum relaxation length @xmath7 , such that ballistic transport is excluded as the dominating transport mechanism . on the other hand , for densities
so low that ballistic transport prevails , the injected spin polarization is confined to very small values .
this result corroborates previous estimates@xcite according to which spin injection is suppressed even in the ballistic regime unless spin - selective interface resistances are introduced . as an example of particular interest from the point of view of applications , we now consider spin - polarized electron transport in heterostructures involving diluted magnetic semiconductors ( dms ) in their paramagnetic phase ( see , e.g. , refs . and ) . in structures of this kind , a nonmagnetic semiconducting
( nms ) layer is sandwiched between two dms layers which , in turn , are enclosed between nonmagnetic metal contacts ( see fig . [ fig:7 ] ) . in a complete thermoballistic description ,
the semiconductor part of a dms / nms / dms structure should be treated as a _ single _ sample .
then , the ballistic intervals @xmath327 $ ] covering the sample may contain one or both of the dms / nms interfaces , at which the potential energy profiles must be expected to change abruptly . the same holds for the material parameters , in particular the momentum and spin relaxation lengths , so that a description in terms of position - dependent parameters would become necessary . here , we adopt a simplified treatment by assuming the different layers in a dms / nms / dms heterostructure to be _ homogeneous _ and requiring the interfaces to act as _ fixed points _ of local thermodynamic equilibrium .
this allows us to apply the thermoballistic description separately to the different layers ( the potential energy profiles are then , in general , _ discontinuous _ at the interfaces ) . for each layer
, we evaluate the spin accumulation function and the zero - bias current spin polarization for a homogeneous semiconductor ( see sec .
[ sec : homsem ] ) , and subsequently match the current spin polarization at the interfaces to obtain its full position dependence as well as the magnetoresistance .
heterostructures of the kind depicted schematically in fig .
[ fig:7 ] are parametrized here by attaching labels @xmath1314 to quantities referring to the left dms layer , the nms layer , and the right dms layer , respectively .
the positions of the interfaces ( including the interfaces between the dms layers and the contacts ) are denoted by @xmath1315 @xmath1316 ; this notation deviates from the one used in the preceding sections . inside the left and right metal contacts , the position dependence of the current spin polarization
is given by eqs .
( [ eq:57lala ] ) and ( [ eq:57zzz ] ) , respectively [ with @xmath70 in the latter equation replaced with @xmath1317 , with @xmath1318 [ see eq .
( [ eq : spfra01hl ] ) for @xmath1319 and @xmath1320 , @xmath1321 for @xmath1121 and @xmath1322 . inside the semiconducting part of the heterostructure
, @xmath1108 is found in terms of the static spin polarizations @xmath1323 of the three layers , and of the boundary values of the spin accumulation function @xmath506 at the interface positions @xmath1315 , @xmath1324 , which are all points of local thermodynamic equilibrium .
[ the interface positions @xmath1325 are points of local thermodynamic equilibrium in the same sense as are the interface positions @xmath1326 , except that there the interface is between metal contact and dms , while here it is between dms and nms . ] in each layer @xmath53 , the function @xmath506 has the sharvin discontinuities @xmath1327 and
@xmath1328 at the interface positions @xmath1329 and @xmath1330 , respectively [ see eqs .
( [ eq : difeq25 ] ) and ( [ eq : difeq26 ] ) ] .
using now , for zero bias , eq .
( [ eq:66ze ] ) together with eq .
( [ eq : difeq20 ] ) separately in each layer , we have @xmath1331 \nonumber \\ \label{eq : a2}\end{aligned}\ ] ] for @xmath1332 @xmath1333 . here , @xmath1334 with @xmath1335 and @xmath1336 defined by eqs .
( [ eq : difeq14c ] ) and ( [ eq : difeq19 ] ) in terms of the momentum relaxation length , @xmath1337 , and spin relaxation length , @xmath1338 , of layer @xmath53 .
further , @xmath1339 \label{eq : a4}\ ] ] and @xmath1340 , \label{eq : a5}\ ] ] where @xmath1341 , \label{eq : a7}\ ] ] and @xmath1342 is the thickness of layer @xmath53 .
[ owing to a difference in normalization , expression ( [ eq : a2 ] ) for @xmath1108 differs from the corresponding expression given by eq .
( 141 ) of ref .
; see the remark following eq .
( [ eq:66ze ] ) above . ] in order to relate the quantities @xmath1343 to the total physical current @xmath133 and the parameters of the contact - semiconductor system , we invoke the continuity of the current spin polarization @xmath1108 at all interfaces , setting @xmath1344 ( @xmath1345 ) in eq .
( [ eq : a2 ] ) . from eqs .
( [ eq : a1 ] ) and ( [ eq : a2 ] ) , we then obtain a set of four coupled linear equations for @xmath1343 ,
@xmath1346 @xmath1347 @xmath1348 @xmath1349 here , @xmath1350 where @xmath1351 } \label{eq : a853}\ ] ] is the ( constant ) total equilibrium electron density in layer @xmath53 , and @xmath1352 are the band offsets .
further , @xmath1353 , \label{eq : a12}\ ] ] with @xmath1354 the system of equations ( [ eq : a8])([eq : a11 ] ) can be easily solved , so that the complete position dependence of the zero - bias current spin polarization is obtained in explicit form .
we now specialize to the case of a _ symmetric _ heterostructure , for which the parameters of the right dms layer and metal contact are identical to those of the left dms layer and metal contact . assuming arbitrarily high conductivities of the contacts , @xmath1355 , so that @xmath1356 , we have from eqs . ( [ eq : a1 ] ) and ( [ eq : a987 ] ) @xmath1357 such that @xmath1358 remains nonzero and finite . then , from eqs .
( [ eq : a9 ] ) and ( [ eq : a10 ] ) with @xmath1359 and @xmath1360 , @xmath1361 where @xmath1362 hence , from eqs .
( [ eq : a8 ] ) and ( [ eq : a11 ] ) , @xmath1363 here , the dms parameters have been labeled by
d " , and the nms parameters by
thus , in the symmetric case , the zero - bias current spin polarization @xmath1108 is completely determined by the quantity @xmath1161 .
explicitly , we obtain , setting @xmath1364 ( where the superscript `` @xmath1365 '' refers to the metal contacts ) in eq .
( [ eq : a1 ] ) and using eqs .
( [ eq : a987 ] ) and ( [ eq : a16 ] ) , @xmath1366 if @xmath1121 and @xmath1322 , respectively .
further , from eq .
( [ eq : a2 ] ) , we have , using eqs .
( [ eq : a6])([eq : a7 ] ) along with eqs .
( [ eq : spfra01hl ] ) , ( [ eq:61ujo ] ) , and ( [ eq : a15 ] ) , @xmath1367 \rule{0mm}{5 mm } \right\ } , \label{eq : aa2}\end{aligned}\ ] ] if @xmath1368 and @xmath1369 , respectively , and @xmath1370 \nonumber \\ & \ & \times \cosh \bm { ( } [ x - ( x_{2 } + x_{3})/2]/l_{n } \bm { ) } , \label{eq : aa3}\end{aligned}\ ] ] if @xmath1371 . in refs . and , we have presented numerical results for @xmath1108 for values of the momentum relaxation length @xmath7 ranging from the drift - diffusion regime to the ballistic regime . the relative magnetoresistance @xmath1013 of the dms / nms / dms heterostructure follows by inserting in eq .
( [ eq : sharv07xbs ] ) the reduced resistance @xmath967 and the corresponding zero - field resistance @xmath1017 obtained by summing up the respective contributions @xmath1372 and @xmath1373 of the ( homogeneous ) layers @xmath53 .
these contributions are obtained from the ( appropriately labeled ) expressions ( [ eq:66zick ] ) and ( [ eq:66tfk ] ) by multiplication with a factor @xmath1374 , such that @xmath967 and @xmath1017 are normalized in conformance with the definition ( [ eq : inteq04scp ] ) for a single sample extending from @xmath69 to @xmath1375 .
specializing immediately to the case of the symmetric structure considered above , we have , defining @xmath1376 @xmath1377 and using eqs .
( [ eq : a14 ] ) , @xmath1378 ( we have omitted here the overall factor @xmath1379 which drops out when @xmath1013is formed ) , from which @xmath1017 follows by setting @xmath1215 and @xmath1216 . then , using eq .
( [ eq : a15 ] ) , we obtain @xmath1380 here , we note that in ref .
, in contrast to its definition in terms of the zero - field resistance @xmath1017 via eq .
( [ eq : sharv07xbs ] ) , the relative magnetoresistance @xmath1013 has been defined with respect to the _ spin - equilibrium _ resistance obtained by setting @xmath1381 in expression ( [ eq:66fdlv ] ) .
the latter definition does not seem to correspond to a genuine magnetoresistance , and so one should not attach quantitative significance to the numerical results for @xmath1013 shown in fig . 4 of ref . . in the limit of low external magnetic field , when the static spin polarization @xmath1382 depends linearly on the field strength , we have from eq .
( [ eq : a17 ] ) , keeping terms of order @xmath1383 , @xmath1384 - { \cal s}_{d}/2 \ } p^{2}}{{\cal s}_{d } + e^{- \beta \delta_{dn } } { \cal s}_{n}/2 } .
\label{eq : a17pdfg}\ ] ] while based on assumptions that differ from , and are more general , than those underlying the two - band model for the ( transverse ) magnetoresistance,@xcite expression ( [ eq : a17pdfg ] ) exhibits the quadratic dependence on the field strength characterizing the latter model in the low - field limit .
considering first the thermoballistic description of spin - polarized transport in dms / nms / dms heterostructures in the drift - diffusion regime , we can compare our results to those of ref .
obtained within the standard drift - diffusion approach . in the drift - diffusion regime ,
when @xmath1385 and @xmath1386 , we have @xmath1387 and @xmath1388 [ see eqs .
( [ eq : ballim26 ] ) , ( [ eq : difeq14c ] ) , ( [ eq : difeq19 ] ) , and ( [ eq : a6 ] ) ] . hence , from eqs .
( [ eq : a12 ] ) and ( [ eq : a13 ] ) , @xmath1389 and @xmath1390 respectively . then , specializing to the symmetric heterostructure , we find for the quantity @xmath1391 defined by eq .
( [ eq : a4711 ] ) @xmath1392 .
\label{eq : a23}\end{aligned}\ ] ] introducing the ( spin - summed ) conductivity @xmath1393 for layer @xmath53 @xmath1394 , @xmath1395 [ see eq .
( [ eq : drude006f ] ) ] , with @xmath1396 @xmath1397 , we can rewrite expression ( [ eq : a23 ] ) in the form @xmath1398 .
\label{eq : a23cayk}\end{aligned}\ ] ] now , inserting expression ( [ eq : a23cayk ] ) in eq .
( [ eq : a15 ] ) and using the resulting expression for @xmath1161 in eq .
( [ eq : a2 ] ) , we find that the values of the zero - bias current spin polarization @xmath1108 at the positions @xmath1326 , @xmath1325 , and @xmath1399 , respectively , agree with those given by eqs .
( 17 ) , ( 16 ) , and(18 ) of ref . , if we identify in the latter equations the spin - flip lengths @xmath1400 and @xmath1401 with @xmath1402 and @xmath1403 , respectively , and the layer thicknesses @xmath1404 and @xmath1405 with @xmath1406 and @xmath1407 , respectively . for the magnetoresistance in the drift - diffusion regime
, we obtain from eq .
( [ eq : a17 ] ) , setting @xmath1408 and inserting expression ( [ eq : a23cayk ] ) for @xmath1391 , @xmath1409 + p^{2}/\mathfrak{q}_{dn } } { s_{d}/ \sigma_{+}^{(0;d ) } + s_{n}/2 \sigma_{+}^{(0;n ) } } , \label{eq : a17zglv}\ ] ] where @xmath1410 is the conductivity for zero spin splitting , which is given by eq .
( [ eq : a16a ] ) with @xmath1411 , and @xmath1412 then , identifying the parameters as above , we find that @xmath1013 agrees with expression ( 13 ) of ref . . turning now to the ballistic regime , when @xmath1413 and @xmath1414 , we have @xmath1415 and @xmath1416 [ see eqs .
( [ eq : ballim26 ] ) , ( [ eq : difeq14c ] ) , ( [ eq : difeq19 ] ) , and ( [ eq : a6 ] ) ] , so that @xmath1417 and @xmath1418 for the symmetric heterostructure , we then find from eq .
( [ eq : a4711 ] ) @xmath1419 e^{\beta \delta_{dn } } .
\label{eq : a21a}\ ] ] with this used in eq .
( [ eq : a15 ] ) for @xmath1161 as well as in eq .
( [ eq : a16 ] ) for @xmath1420 , we obtain the position dependence of the zero - bias current spin polarization from eqs .
( [ eq : a1 ] ) and ( [ eq : a2 ] ) . for the magnetoresistance in the ballistic regime , we have from eqs .
( [ eq : a17 ] ) , setting @xmath1421 and inserting expression ( [ eq : a21a ] ) for @xmath1391 , @xmath1422 e^{\beta \delta_{dn } } \ } } { 1 + e^{- \beta \delta_{dn}}/2 } . \label{eq:22}\ ] ] depending on the values assigned to the different parameters in this expression , the ballistic ( relative ) magnetoresistance is positive or negative .
in this article , we have presented a comprehensive survey of the thermoballistic approach to charge carrier transport in semiconductors .
the principal aim has been to develop the basic physical concept underlying this approach in detail , and to give a coherent exposition of the ensuing formalism , which unifies , and partly modifies , generalizes , and corrects , the formal developments presented in our previous publications . to make the presentation self - contained and easy to follow , we have proceeded step by step , starting with an account of drude s model as the origin of all semiclassical transport models .
we then reviewed the standard drift - diffusion and ballistic transport models , basic features of which have been adopted to shape the thermoballistic description of carrier transport in terms of averages over random configurations of ballistic transport intervals .
the contributions of the individual ballistic intervals to the total carrier current are governed by collision probabilities involving the carrier mean free path , or momentum relaxation length , as the determining parameter of the thermoballistic concept .
this concept finds a first concrete expression in the prototype thermoballistic model , which is based on the simplifying assumption of current conservation across the points of local thermodynamic equilibrium linking ballistic intervals .
the implementation of the prototype model results in a current - voltage characteristic containing a reduced resistance , which can be expressed explicitly in terms of the parameters of the semiconducting system . in the full thermoballistic concept , current conservation across the equilibrium points is abandoned , and position - dependent , total and spin - polarized thermoballistic currents and densities are introduced in terms of an average chemical - potential function and a spin accumulation function related to the spin splitting of the chemical potential .
the algorithms for determining these dynamical functions follow from two physical conditions .
first , the average of the total thermoballistic current over the length of the semiconducting sample ( as well as over the range between either end and an arbitrary point of local thermodynamic equilibrium ) is required to equal the conserved physical current , whereby one is able to set up a scheme for obtaining the average chemical potential .
the explicit calculation of this function is implemented in terms of resistance functions , for which volterra - type integral equations are derived .
second , spin relaxation is assumed to act only inside the ballistic transport intervals . as a result , the thermoballistic spin - polarized current and density
are connected by a spin balance equation , from which one obtains an inhomogeneous fredholm - type integral equation for the spin accumulation function . in the general case of
arbitrarily shaped potential energy profiles , considerable numerical effort is needed for solving the integral equations for the resistance functions and the spin accumulation function . in a number of important special cases , however , these equations reduce to a form which greatly facilitates their solution .
examples are the equations for the resistance functions when the spin splitting of the band edge profile is independent of position , or the equation for the spin accumulation function for a homogeneous sample in an external electric field . in the latter case
, the integral equation can be converted into an easily tractable second - order differential equation . for homogeneous semiconductors at zero bias , the solutions of the equations for the dynamical functions can be obtained in closed form throughout . for the purpose of demonstrating the potentialities of the thermoballistic approach in present - day semiconductor and spintronics research , we have summarized in this article the treatment of a number of specific examples .
the prototype model is employed to describe electron transport across potential energy profiles exhibiting an arbitrary number of barriers , where effects of tunneling and degeneracy are included .
this approach proves to be of particular relevance to the description of grain boundary effects in electron transport in polycrystalline semiconductors and has already been applied with promising results in the analysis of experimental data . the full thermoballistic approach is used in the treatment of spin - polarized transport in heterostructures . for the prototype problem of semiconductor spintronics , _ viz .
_ , the injection of spin polarization from ferromagnetic contacts into a semiconducting sample , the thermoballistic description extends the standard drift - diffusion description so as to allow for arbitrary values of the momentum relaxation length .
the same holds for transport in heterostructures composed of layers of diluted magnetic and nonmagnetic semiconductors at zero bias , where the position dependence of the current spin polarization as well as the magnetoresistance are obtained in closed form . in our previous publications
, we have presented results of explicit calculations for a variety of specific cases .
these calculations were mostly of exploratory character , with the aim to reveal qualitative trends in the parameter dependence of the relevant transport properties . in future
work , emphasis should be placed on the application of the thermoballistic approach in quantitative studies , as demanded for the analysis of specific experimental results . for these studies to become successful ,
the careful evaluation of the potential energy profiles is prerequisite , and advanced numerical techniques for solving the integral equations for the dynamical functions are to be employed . with these goals
achieved , the thermoballistic approach will certainly prove useful as a practical tool , beyond its basic theoretical relevance as the bridge between the drift - diffusion and ballistic descriptions of charge carrier transport in semiconductors .
we are indebted to the board of directors and the staff of the helmholtz - zentrum berlin fr materialien und energie for generously granting access to its premises and facilities beyond the date of our retirement , thereby allowing us to continue and complete our research work on charge carrier transport in semiconductors .
p. drude , ann .
( leipzig ) * 306 * , 566 ( 1900 ) [ reprinted in _
ostwalds klassiker der exakten wissenschaften _ , vol . 298 , edited by h. t. grahn and d. hoffmann ( verlag harri deutsch , frankfurt / m . , 2006 ) ] ; * 308 * , 369 ( 1900 ) .
a. sommerfeld and h. bethe , in _ handbuch der physik _
, vol . 24/2 , second edition , edited by h. geiger and k. scheel ( springer , berlin , 1933 ) [ reprinted in _ heidelberger taschenbcher _
, vol . 19 ( springer , berlin , 1967 ) ] .
t. weis , r. lipperheide , and u. wille , in _ proc .
2nd world conf .
photovoltaic solar energy conversion _ , edited by j. schmid , h. ossenbrink , p. helm , h. ehmann , and e. d. dunlop ( european comm .
joint res .
centre , ispra , 1998 ) , p. 1438 .
we use the term `` semiclassical '' essentially as defined in chap .
12 of ref . , where the band structure of the unperturbed system is assumed to be generated from quantum - mechanical calculations , whereas the effect of an externally applied electric field is treated classically ( the effect of a static external magnetic field is assumed here to be included in the band edge potential ) .
we go beyond this scope when considering corrections to the ballistic currents due to tunneling and degeneracy . | a comprehensive survey is given of the thermoballistic approach to charge carrier transport in semiconductors .
this semiclassical approach bridges the gap between the drift - diffusion and ballistic ( `` thermionic '' ) models of carrier transport , whose validity is limited to the range of very small and very large values , respectively , of the carrier mean free path .
the physical concept underlying the thermoballistic approach , while incorporating basic features of the drift - diffusion and ballistic descriptions , constitutes a novel , unifying scheme .
it is based on the introduction of `` ballistic configurations '' defined by a random partitioning of the length of a semiconducting sample into ballistic transport intervals .
the points linking adjacent ballistic intervals are assumed to be points of local thermodynamic equilibrium characterized by a local chemical potential .
carriers thermally emitted at any such point are ballistically transmitted across either interval , while at the same time carriers transmitted from the equilibrium points next to it are `` absorbed '' at that point , i.e. , they are assumed to be instantaneously equilibrated there . during their transmission ,
the carriers face , in general , potential energy barriers arising from internal and external electrostatic potentials in the sample .
the lengths of the ballistic intervals are stochastic variables , with associated probabilities given by the probabilities for carriers to traverse an interval without collisions with the scattering centers randomly distributed over the sample .
these probabilities are controlled by the carrier mean free path , whose magnitude is arbitrary . by averaging the ballistic carrier currents over all ballistic configurations ,
a position - dependent thermoballistic current is derived , which is the key element of the thermoballistic concept and forms the point of departure for the calculation of various transport properties .
the present article starts out with a preparatory account of the standard drift - diffusion and ballistic transport models which form the cornerstones of the thermoballistic concept , and of a prototype model which paves the way for the fully developed form of that concept . in the main body of the article , a coherent exposition of the thermoballistic approach
is given within a general formulation that takes into account arbitrarily shaped , spin - split potential energy profiles and spin relaxation during the carrier motion across ballistic intervals .
the calculational procedures devised for implementing the thermoballistic concept are described .
specific examples relevant to present - day semiconductor and spintronics research are considered . |
the exploration of underlying mechanism for anomalous and normal heat conduction in low dimensional systems represents a huge challenge in the area of statistical physics @xcite . after enormous efforts for more than one decade from numerical simulations@xcite , theoretical predictions@xcite and experimental observations@xcite
, there is still no consensus for the exact physical mechanism causing anomalous heat conduction .
it is believed that momentum conservation plays an important role in determining the actual heat conduction behavior . in general ,
1d non - integrable lattices with momentum conserving property should have anomalous heat conduction where the thermal conductivity @xmath1 diverges with the lattice size @xmath2 as @xmath3 where @xmath4 @xcite .
however , the 1d coupled rotator lattice is a well known exception .
it exhibits normal heat conduction behavior despite its momentum conserving nature @xcite . the normal heat conduction in 1d coupled rotator lattice was discovered via numerical simulations by two groups independently @xcite .
in order to understand the underlying mechanism , the temperature dependence of thermal conductivity has been studied in detail in both works . in ref .
@xcite , the temperature dependence of thermal conductivity of coupled rotator lattice has been found to be like @xmath5 where @xmath6 is a positive constant .
however , in ref .
@xcite , a temperature dependence of thermal conductivity of @xmath7 was claimed where @xmath8 is a positive constant .
there it was also argued that a possible phase transition at temperature around @xmath9 where heat conduction is normal above this temperature while anomalous below this temperature exists @xcite . although this phase transition statement was challenged as a finite size effect @xcite , a later work supported this phase transition conjecture after deriving a similar temperature dependent thermal conductivity as @xmath7 @xcite .
most recently , the 1d coupled rotator lattice re - attracts much attention due to the new finding of simultaneously existing normal diffusion of momentum as well as heat energy @xcite .
it is argued that the normal behavior might be due to the reduced number of conserved quantities which is unique for periodic interaction potentials @xcite .
but the new debate is that whether the stretch is conserved or not in 1d coupled rotator lattice .
the investigation of 1d coupled rotator lattice will be the key to unravel the true mechanism behind the connection between momentum conservation and normal or anomalous heat conduction .
therefore , it is the right time to revisit the temperature dependence of thermal conductivity of 1d coupled rotator lattice as a first step . in this work , we will revisit the temperature dependence of thermal conductivities for the 1d coupled rotator lattice .
we find that the temperature dependence is neither @xmath5 nor @xmath7 as previously claimed @xcite .
the actual temperature dependence is a power - law dependence as @xmath10 .
the possible connection with the momentum diffusion of single kicked rotator or the chirikov standard map has also been discussed . in order to determine whether there is a phase transition
, we also present the temperature dependent thermal conductance at different system sizes .
all the thermal conductances for different sizes collapse to the same value at low temperatures while approach to the power - law behavior as @xmath11 at high temperatures
. however , the crossover temperature decreases as the system size increases .
this fact indicates that there is no phase transition between normal and anomalous heat conduction . in thermodynamical limit ,
the heat conduction is normal for all temperatures except the trivial zero temperature point . in sec .
ii the lattice model will be introduced .
numerical results and discussions will be presented in sec .
iii and the results will be summarized in sec .
the hamiltonian for the 1d coupled rotator lattice is defined as the following @xcite : @xmath12\right]\ ] ] where @xmath13 and @xmath14 denote the displacement from equilibrium and momentum for @xmath15-th atom , respectively .
the mass of the atom @xmath16 and the boltzmann constant @xmath17 has been set into unity .
the parameter @xmath18 with energy dimension represents the coupling strength of the inter - atom potential .
therefore , the system temperature @xmath19 can be rescaled by @xmath20 and we can also set @xmath21 for simplicity @xcite .
the equations of motion for @xmath15-th atom are @xmath22 at low temperature limit , the displacements are small values so that @xmath23 .
the hamiltonian can be expanded into the taylor series as : @xmath24\end{aligned}\ ] ] this hamiltonian will approach to the integrable harmonic lattice only at zero temperature @xmath25
. the first stable nonlinear potential term will be the sextic potential .
on the other hand , at infinite high temperature limit @xmath26 , the rotator lattice will approach to another integrable system consisting of @xmath2 independent and isolated free particles as @xmath27 this is because the kinetic energy is proportional to the temperature as @xmath28 and the potential energy in eq .
( [ ham ] ) is confined by the cosine function . without getting into numerics ,
we can get a qualitative picture of the thermal conductivity of 1d coupled rotator lattice as the function of temperature .
the thermal conductivity @xmath29 will diverge in the low temperature limit approaching to the harmonic limit and will decay to zero in the high temperature limit as approaching to the unconnected @xmath2 free particles . as a general picture
, the thermal conductivity @xmath29 will decrease as the temperature increases .
( color online ) .
temperature profiles for the 1d coupled rotator lattice at ( a ) @xmath30 and ( b ) @xmath31 .
the lattice sizes are @xmath32 and @xmath33 and the color rule of the lines are the same for both ( a ) and ( b ) . the first and last atom are put into contact with a langevin heat bath with temperature set as @xmath34 where @xmath35 here . ] in non - equilibrium numerical simulations , we put the first and last atom into contact with langevin heat bath . the temperatures for left and right heat bath
are set as @xmath36 where @xmath19 denotes the average temperature and @xmath37 gives rise to the temperature gradient along the lattice .
the fixed boundary conditions with @xmath38 are also been applied .
before we discuss the results of thermal conductivities , we first show the temperature profiles . in fig . [ fig : t - profile ] , the temperature profiles at two different temperatures @xmath30 and @xmath31 are plotted .
the lattice sizes are chosen as @xmath32 and @xmath33 .
for relative low temperature at @xmath30 , the temperature jumps at two boundaries are obvious .
however , the temperature jumps are reduced with the increase of lattice size @xmath2 as can be seen in fig .
[ fig : t - profile](a ) . in fig .
[ fig : t - profile](b ) where the temperature is relatively high at @xmath31 , all the temperature profiles collapse to the same straight line as the temperature jumps are very small for all lattice sizes . in order to obtain the temperature dependence of thermal conductivities , we need first define the way how @xmath29 can be calculated numerically .
we notice that the temperature profiles all collapse to the same straight line if the temperature jumps can be ignored at high temperatures or long lattice sizes .
it is appropriate to define the thermal conductivity @xmath29 as : @xmath39 where @xmath40 is the average heat flux along the lattice in the stationary state and the local heat flux @xmath41 is defined as @xmath42 via the energy continuity equation .
( color online ) . thermal conductivities @xmath29 as the function of temperature for different lattice sizes @xmath32 and @xmath33 .
the straight line is proportional to @xmath43 which describes the temperature behavior of @xmath29 at high temperature region very well . ] in fig .
[ fig : kappa ] , the thermal conductivities @xmath29 as the function of temperature are plotted for different lattice sizes @xmath32 and @xmath33 . at high temperature region ,
all the thermal conductivities @xmath29 for different lattice sizes collapse together indicating the saturation of thermal conductivities as the increase of lattice size .
this is characteristic for lattices with normal heat conduction .
as the temperature decreases , the thermal conductivities first increases and then becomes flat .
this is because the phonon mean free paths are getting longer as the temperature reduces and the ballistic regime will be approached if the phonon mean free paths are longer than the lattice size .
the boundary jumps will dominate the temperature profiles as in fig .
[ fig : t - profile](a ) and the definition of @xmath29 of eq .
( [ kappa - def ] ) will no longer reflect the actual thermal conductivities . at high temperatures
, it is clearly seen that the thermal conductivities follows a power - law dependence as @xmath10 . for the longest size we considered here as @xmath44
, this power - law behavior can be best fitted for more than two orders of magnitudes for the @xmath29 value as shown in fig .
[ fig : kappa - t_1600 ] .
this also explains the poor fitting of the @xmath5 dependence in ref .
@xcite and the @xmath7 dependence in ref .
@xcite .
( color online ) . thermal conductivities @xmath29 as the function of temperature for lattice size @xmath44 .
the data are taken from fig .
[ fig : kappa ] and can be best fitted to be a power - law dependence as @xmath10 . ] this power - law dependence of @xmath10 can not be explained by the effective phonon theory which is able to predict the temperature dependent thermal conductivities for other typical 1d nonlinear lattices such as fpu-@xmath0 lattice and generalized nonlinear klein - gordon lattices @xcite . according to the effective phonon theory , the thermal conductivity for low temperature rotator lattice with hamiltonian of eq .
( [ rotator - low - t ] ) can be derived as @xmath45 where @xmath46 is the nonlinearity strength with the following temperature dependence @xmath47 at low temperature region .
here we consider the sextic potential term as the lowest nonlinear term because the dynamics governed by the negative quartic potential term is unstable .
therefore , the actual temperature behavior of @xmath10 is steeper than the prediction of @xmath48 from effective phonon theory . although it is difficult to unravel the exact physical mechanism behind the power - law dependence of thermal conductivities for coupled rotator lattice , it is very helpful to look into the transport properties of the single kicked rotator ( the chirikov standard map ) @xcite .
as the name indicates , the coupled rotator lattice is a kind of connected kicked rotators .
the equations of motion for the single kicked rotator are @xmath49 where @xmath50 and @xmath51 denote the coordinate and momentum after @xmath52-th kick .
( color online ) .
momentum diffusion constant @xmath53 as the function of temperature for 1d coupled rotator lattice .
the numerical data are obtained via equilibrium md simulations as in ref .
the solid line of @xmath43 is guided for the eye . ]
the variation of momentum @xmath54 is unbounded and can be characterized by a normal diffusion @xcite @xmath55 where @xmath56 is the diffusion constant .
this is similar to the normal momentum diffusion for 1d coupled rotator lattice . according to ref .
@xcite , the diffusion constant @xmath56 depends on the coupling strength @xmath18 as @xmath57 where @xmath58 is the chaos threshold for the kicked rotator . if one assume the transport properties of coupled rotator lattice are the same as that of single kicked rotator , one would expect that the energy diffusion constant @xmath59 as well as the momentum diffusion constant @xmath53 of coupled rotator lattice should also follow the same dependence as that of single kicked rotator as in eq .
( [ dif - single ] ) . to translate the @xmath18 dependence into @xmath19 dependence
, we notice that the parameter @xmath18 in single kicked rotator of eq .
( [ equation - single ] ) plays the same role as that in the coupled rotator lattice as in eq .
( [ equation - lattice ] ) .
as we have discussed above , the dynamics of coupled rotator lattice can be rescaled with @xmath20 .
the @xmath18 dependence should be inversely proportional to the @xmath19 dependence .
low @xmath18 value region corresponds to the high @xmath19 region and vice verse . and
for normal heat conduction , the thermal conductivity @xmath1 is proportional to the energy diffusion constant @xmath59 .
this will finally give rise to the prediction of temperature dependent thermal conductivity @xmath29 of coupled rotator lattice from the analogy with single kicked rotator : @xmath60 as a consistence check , we plot the momentum diffusion constant @xmath53 for the coupled rotator lattice as the function of temperature in fig .
[ fig : dp ] . the same temperature dependence of @xmath61 has been obtained .
therefore , the prediction of @xmath62 with analogy to the single kicked rotator at high temperature or low @xmath18 region is close to the numerical observation of @xmath63 .
this indicates there might be some deeper connection between the diffusion behavior of single kicked rotator and 1d coupled rotator lattice .
in addition , the analogy analysis also predicts a @xmath64 behavior at low temperature region which is the same as the prediction from the effective phonon theory of eq .
( [ kappa - ept ] ) .
the reason why this temperature behavior can not be observed in numerical simulations might be due to the severe finite size effect at low temperature region as can be seen in fig .
[ fig : t - profile](a ) .
( color online ) .
heat conductance @xmath65 as the function of temperature @xmath19 for different lattice size @xmath32 and @xmath33 .
all the other parameters are the same as in fig .
[ fig : kappa ] .
the left - pointing arrow represents the trend of decreasing crossover temperature as the lattice size increases . ]
in the final part we will discuss the issue about the possible phase transition between normal and anomalous heat conduction for 1d coupled rotator lattice . as we have shown in fig .
[ fig : kappa - t_1600 ] , the actual temperature dependence of @xmath29 is a power - law dependence as @xmath63 .
this indicates the analytic derivation of @xmath7 in ref .
@xcite originally for a lattice with pinned on - site potential can not be extended to 1d coupled rotator lattice . on the other hand , the claim that the transition temperature is around @xmath66 in ref .
@xcite has been challenged in ref .
@xcite with numerical simulations of longer sizes .
the possible phase transition could be a finite size effect .
here we reconfirm the finite size effect by giving a more illustrative picture in fig .
[ fig : conductance ] . the heat conductance @xmath67 as the function of temperature has been plotted for different lattice sizes . at high temperature regions , all the heat conductances
follow the same temperature behavior of @xmath68 .
as the temperature decreases , the phonon mean free path will overcome the lattice size and the heat conductance will saturate to a value determined by the harmonic limit of coupled rotator lattice . as can be seen from fig .
[ fig : conductance ] , the heat conductance for short lattice will first bend and become flat as the temperature decreases .
this crossover temperature decreases as the lattice size increases smoothly and no sign of phase transition can be observed .
it is also noticed that the crossover temperature happens to be around @xmath66 for lattice sizes of a few thousands .
we have systematically investigated the temperature dependence of thermal conductivities of 1d coupled rotator lattice .
the actual temperature dependence is a power - law dependence of @xmath10 which is different with the observations of previous studies . the possible connection with the single kicked rotator or
the chirikov standard map has been discussed where a @xmath69 dependence can be implied .
our results also reconfirm that the previously claimed possible phase transition should be a finite size effect .
the numerical calculations were carried out at shanghai supercomputer center . this work has been supported by the nsf china with grant no .
11334007 ( y.l .
, b.l . ) , the nsf china with grant no .
11347216 ( y.l ) , tongji university under grant no .
2013kj025 ( y.l ) , the nsf china with grant no .
11205114 ( n.l . ) , the program for new century excellent talents of the ministry of education of china with grant no .
ncet-12 - 0409 ( n.l . ) and the shanghai rising - star program with grant no .
13qa1403600 ( n.l . ) . | in contrary to other 1d momentum - conserving lattices such as the fermi - pasta - ulam @xmath0 ( fpu-@xmath0 ) lattice , the 1d coupled rotator lattice is a notable exception which conserves total momentum while exhibits normal heat conduction behavior .
the temperature behavior of the thermal conductivities of 1d coupled rotator lattice had been studied in previous works trying to reveal the underlying physical mechanism for normal heat conduction .
however , two different temperature behaviors of thermal conductivities have been claimed for the same coupled rotator lattice .
these different temperature behaviors also intrigue the debate whether there is a phase transition of thermal conductivities as the function of temperature . in this work
, we will revisit the temperature dependent thermal conductivities for the 1d coupled rotator lattice .
we find that the temperature dependence follows a power law behavior which is different with the previously found temperature behaviors .
our results also support the claim that there is no phase transition for 1d coupled rotator lattice .
we also give some discussion about the similarity of diffusion behaviors between the 1d coupled rotator lattice and the single kicked rotator also called the chirikov standard map . |
blue compact dwarf galaxies ( hereafter , bcds ) have emerged as a particularly intriguing subset of dwarf galaxies that have been studied in the local universe ( [email protected] ) . since the early studies and characterizations by @xcite and @xcite , many investigations have classified bcds as systems that harbor spatially and temporally concentrated ongoing star formation in a low - mass galaxy potential well .
while definitions in the literature vary , bcds are instrinsically faint systems ( m@xmath617 ) that exhibit ongoing star formation over the bulk of an otherwise physically compact stellar population .
detailed studies of individual , nearby bcds have provided important insights into their physical properties and into the mechanisms that initiate and govern their star formation .
@xcite found that five compact dwarf galaxies that are currently experiencing strong bursts of star formation exhibit central h i surface densities that are significantly higher than those of dwarf irregular ( dirr ) galaxies .
similarly , some well - studied bcds harbor very high column density neutral hydrogen gas in the regions of active star formation ( e.g. , n@xmath7 @xmath4 2.4@xmath810@xmath9 @xmath10 in izw18 ; see brown etal .
2002 and van zee etal .
other recent studies have verified that many nearby and well - resolved systems have high neutral gas surface densities in the regions of ongoing star formation ( see , e.g. , thuan etal .
2004 , elson etal .
2010 , and references therein ) . the physical mechanism(s ) that create this central concentration of neutral gas remain(s ) ambiguous .
one commonly proposed solution is that the current burst events in bcds are the result of a gravitational interaction with a nearby companion ( e.g. , * ? ? ?
* ) . while there are examples of apparently interaction - driven burst events ( e.g. , the sbs1129@xmath11576 system studied by ekta etal .
2006 ) , there are also systems that have been shown to be quite isolated ( see , for example , van zee etal . 1998 , bureau etal . 1999 , and further discussion below ) .
while undiscovered , very low - mass companions may reside near some of these isolated systems ( consider , for example , the discovery of an interaction with a low surface brightness companion to the magellanic irregular ngc4449 by martnez - delgado etal .
2012 ) , the gravitational interaction trigger does not appear to be universal in bcd galaxies .
the origin of the concentrated star formation in at least some of the isolated bcds appears to require a different initiation mechanism . while the properties of isolated bcds have been studied in some detail ( e.g. , the studies of ngc2915 by meurer etal .
1994 , 1996 ) , a particularly intriguing system was discovered in the _ arecibo dual beam survey _ ( adbs ) by @xcite : adbs113845 + 2008 .
this galaxy was subsequently cataloged as a uv - excess source in the 2nd kiso survey ( miyauchi - isobe & maehara , 2000 ) .
its extremely compact nature was first recognized during _ wiyn _ 0.9-meter broadband imaging of selected adbs galaxies .
the optical component was found to be extremely compact ( b - band exponential scale length @xmath20.57 kpc ) ; h@xmath1 narrowband imaging confirmed a high ongoing star formation rate ( 0.06 m@xmath12yr@xmath3 ) within this optically compact stellar component .
subsequent _ very large array _
( _ vla _ ) h i spectral line imaging of the source revealed that the neutral gas disk was extremely extended compared to the compact optical structure .
as discussed in detail in @xcite , h i gas was detected to a maximum radial distance of @xmath1325 kpc at the 10@xmath14 @xmath10 level ; it thus possesses one of the largest h i - to - optical size ratios of any known dwarf galaxy .
in contrast to some detailed h i studies of other bcd systems that have found complex velocity structure ( e.g. , * ? ? ?
* ; * ? ? ?
* ) , adbs113845 + 2008 is undergoing well - ordered rotation throughout most of the neutral gas disk .
perhaps the most mysterious quality of adbs113845 + 2008 is that its neutral gas morphology is dominated by a broken ring structure that is roughly 15 kpc in diameter .
this ring contains the highest column density neutral gas in the galaxy and is not coincident with the optical component .
rather , the star - forming component resides in a relative h i under - density .
@xcite interpret this result to be a signature of an extremely inefficient mode of star formation in a very isolated galaxy .
we hereafter refer to this striking collection of physical properties as the `` giant disk '' phenomenon in dwarf galaxies .
the remarkable characteristics of adbs113845 + 2008 are the origin of the present investigation .
stated most simply , we seek to understand if this `` giant disk '' scenario is common or rare in star - forming dwarf galaxies that have stellar components that are as strongly centrally concentrated as the one in adbs113845 + 2008 . to this end
, we present new optical , infrared , and h i imaging of a sample of extremely optically compact dwarf galaxies . the new observations presented here confirm the rarity of the `` giant disk '' found in the adbs113845 + 2008 system .
the sample of optically compact galaxies studied in this paper was produced using specific selection criteria designed to identify star - forming galaxies with compact stellar components ; adbs113845 + 2008 and ic2271 were already known to have such characteristics from previous imaging . starting with the full _ kpno international spectroscopic survey _
( _ kiss _ , an objective - prism survey for extragalactic emission - line objects ; salzer etal . 2000 , 2001 , 2002 ) database , we searched for emission - line objects that were spectroscopically - confirmed . to maximize physical resolution
an upper velocity constraint was set at v@xmath15 @xmath16 5,000 kms@xmath3 .
photometric parameters were obtained for these objects from the _ sdss _ using the _ skytools _ interface .
objects with optical radii less than 3 were selected ; these were then individually inspected to ensure a compact optical structure . finally , we searched for archival single - dish h i observations of these remaining systems in order to determine their suitability for _ vla _ observing .
this resulted in a sample of six galaxies selected from the _ kiss _ catalogs .
we stress that this sample is meant to be representative of the compact dwarf class , but it is not complete in the sense of being a volume - limited sample of such sources .
all of these galaxies share the following characteristics : 1 ) optical compactness ( optical radii at 50% peak brightness of @xmath51 kpc ) ; 2 ) faint blue absolute magnitudes ( m@xmath1717 ) ; 3 ) ongoing star formation ( based on emission - line selection by the h@xmath1 or [ ] lines ) ; and 4 ) proximity for spatially resolved work with the _ vla _ c - array ( 15 beam @xmath55 kpc for all systems ; mean velocity @xmath2 3315 kms@xmath3 @xmath4 45 mpc ) .
note that in the work of @xcite , adbs113845 + 2008 was selected using these same criteria .
it is important to emphasize that the six _ kiss _ galaxies in the present study ( see table [ t1 ] for salient properties ) were selected based on their optical properties alone ; the h i flux integrals were used only to determine observational feasibility .
h i spectral line imaging of each of the sample galaxies was obtained with the _ vla _ in the c configuration via observing programs ac 841 and ac 919 ( pi : cannon ) .
the _ vla _
correlator was used in a standard single polarization mode , providing 64 or 128 channels over a 1.56 or 3.12 mhz bandwidth . the velocity resolution ( 2.56 , 5.15 , or 10.3 kms@xmath3ch@xmath3 )
was selected based on the ( single dish ) line width of each source .
combinations of bandwidth and frequency resolution were selected to guarantee sufficient line - free channels to allow an accurate removal of the continuum emission .
table [ t2 ] provides details about each observing session from which data are presented in this work .
all reductions were performed using the astronomical image processing ( aips ) package .
due to the array being in a `` hybrid '' state , containing both _
vla _ and _ evla _ receivers , standard reduction procedures were followed with necessary modifications to account for the `` aliasing '' effect that arose due to the digital - to - analog conversion required to correlate _ evla _ signals using the old _ vla _ correlator .
this effect was most egregious for narrow bandwidth observations .
we thus followed the nrao instructions of flagging _
baselines before and unflagging them after the tasks bpass and calib .
the resulting continuum subtraction was performed using a first - order fit in uvlsf to channels where the bandpass was demonstrably flat .
we found no measurable differences between this calibration approach and one where we disregarded the _ evla _ antennas completely and followed standard reduction procedures .
we note that there were only 5 _ evla _ antennas in the array during the c configuration observation of adbs113845 + 2008 ( which would contain the most measurable aliasing effect , given the bandwidths listed in table [ t2 ] ) .
two data cubes were produced for each galaxy using the task imagr and each one was cleaned to 2 times the rms noise ( hgbom 1974 ; clark 1980 ) . first , `` low - resolution '' cubes were created using natural weighting ( robust = 5 in the aips imagr task ) .
these cubes were convolved to a circular beam size of 20 x 20 .
second , `` high - resolution '' cubes were created using robust weighting ( robust = 0.5 ; see briggs 1995 ) .
these cubes were convolved to a circular beam size of 15 x 15 . to create blanked cubes and moment maps ,
the naturally weighted cubes were convolved to a beam size of 30 .
these cubes were blanked at the 2 @xmath18 level found in line - free channels .
the cubes were then inspected by hand , and emission in three or more consecutive channels was considered to be real .
these cubes were then used to blank both the `` low- '' ( 20 ) and `` high - resolution '' ( 15 ) cubes . through this procedure , we assure that the same pixels contribute to both the `` low- '' and the `` high - resolution '' cubes .
moment maps showing integrated h i emission and velocity structures were created using the techniques discussed in detail in @xcite . as seen in table [ t3 ] ,
the total h i flux integrals derived for the sample galaxies are each lower than the _ arecibo observatory _ single dish observations @xcite .
the cause of these differences is at least partly due to the missing short _ uv _ spacings of the _ vla _ compared to a single dish measurement .
for example , the observed h i flux integral of ic2271 of 1.22 jy kms@xmath3 is significantly lower than the value found by @xcite using the _ vla _ d array ( 4.7 jy kms@xmath3 ) .
similarly , the total h i flux integral derived for adbs113845 + 2008 ( 1.58 jy kms@xmath3 ) is lower than the 2.58 jy kms@xmath3 found by @xcite using combined c and d configuration data .
a re - reduction of the d configuration data sets as part of the present analysis confirms a higher flux integral using the more compact configuration ( s@xmath7 = [email protected] jykms@xmath3 for adbs 113845 + 2008 and s@xmath7 = [email protected] jykms@xmath3 for ic 2271 , both with 55 circular beam sizes ) .
since the kissr galaxies ( which constitute the bulk of the present sample ) have only c configuration observations , we chose to intentionally omit the d configuration data for ic2271 and adbs113845 + 2008 from the present analysis ; this assures that all eight galaxies have similar thresholds for surface brightness sensitivity at different spatial scales . in an effort to recover as much flux from each of the target galaxies as possible , we performed various tests with the _ vla _ _ uv _ datasets .
long baselines were downweighted in imagr via tapering in the _ uv _ plane .
circular beam size cubes with resolution elements of 15 , 30 , and 60 were created for each galaxy using appropriate uvtaper values .
the results of these tests are summarized in table [ t4 ] .
there are only modest gains in the recovered flux integrals for most sources .
kissr278 shows the most pronounced increase in s@xmath7 when moving from 15 to 30 resolution .
note that ic2271 and adbs113845 + 2008 , which have the largest single dish flux integrals ( see table [ t3 ] ) , show only modest increases in s@xmath7 when downweighting long baselines .
this is strong evidence that the missing shortest spacing baselines are important for the recovery of all of the h i flux from these systems
. we also experimented with calibrating our _ uv _ data using only certain baselines in the array .
using kissr561 as a template , we independently calibrated the _
uv _ data in five different ways : by removing the _ vla_-_evla _ baselines ; by removing the _ vla_-_vla _ baselines ; by retaining only the _ evla_- _ evla _ baselines ; by retaining only the _ vla_-_vla _ baselines ; and by the previously discussed method in which the _ evla_-_evla _ baselines were flagged and unflagged around specific tasks .
we find no significant differences in the galaxy s morphology amongst these tests .
the last method ( i.e. , the flagging and unflagging of _ evla_-_evla _ baselines ) proved to recover the most flux , and so was used for all datasets in the sample . as we discuss in detail in
[ s3.2 ] below , five of the eight sample galaxies have nearby companions , some of which are very h i - bright .
a third possible explanation for the lower flux integrals derived for some of the systems studied here is that these high - surface - brightness companions introduce sidelobes , the imperfect cleaning of which will prevent a full recovery of the flux integral from the dim target galaxies .
in fact , this effect can lead to the single dish flux measurements being erroneously high .
the system that is most egregiously affected in this way is kissr396 ; as we discuss below , this source is a companion to the massive spiral ngc5375 .
it is possible that the single dish measurement of kissr396 is an over - estimate , since ngc5375 is in the first sidelobe of the single - dish primary beam . in an attempt to image this complex field with high fidelity
, we implemented cleaning algorithms using both multi - faceted cleans around the detected galaxies and multi - scale clean algorithms @xcite . neither approach produced significant increases in the flux integral recovered from kissr396 .
we thus use the _ vla _ s@xmath7 values quoted in table [ t3 ] throughout the remainder of this work .
optical and infrared imaging of the sample galaxies was obtained from public archival databases to facilitate detailed comparisons with the h i imaging presented here .
this was necessary because the imaging data obtained as part of _ kiss _
@xcite had an image scale of 2.0pixel@xmath3 ; none of the _ kiss _ galaxies studied here are resolved in the survey data .
optical r - band images were obtained from the _ sdss _ data release eight @xcite . 3.6 @xmath20 m and 4.5 @xmath20 m
_ spitzer _ images of the six kissr galaxies were obtained from the _ spitzer heritage archive _ ( _ spitzer _ imaging of ic2271 and adbs113845 + 2008 is not available ) .
no further treatments of the pipeline - processed archival data were required .
these images are discussed in further detail in the sections below . throughout this paper
we will discuss the physical sizes of the galaxies in this sample .
we thus draw specific attention to the angular resolution of the optical and infrared imaging used in our analysis .
the fwhm of the _
spitzer_/irac images is 1.66 at 3.6 @xmath20 m and 1.72 at 4.5 @xmath20 m .
similarly , the median psf fwhm of _ sdss _ r - band images is quoted as 1.3 , and we adopt this value here ; using the individual images we analyze in this work , we measure empirical psf fwhm values ranging from 0.94 ( ic2271 ) to 1.22 ( kissr245 ) .
using the blanked datacubes discussed in
[ s2.2 ] , real emission in each channel was summed in order to create the global h i profiles shown in figures [ figcap1 ] and [ figcap2 ] .
it is immediately evident that the neutral gas disks of the sample galaxies span a wide range of physical characteristics .
two systems show broad h i profiles ( @xmath21v @xmath22100 kms@xmath3 ) that are indicative of bulk rotation ( e.g , adbs113845 + 2008 , kissr561 ) .
three systems show what may be double - horned or broad profiles that are measured with poor fidelity ( e.g. , kissr245 , kissr396 , kissr1021 ) .
three systems show simple single - peak profiles indicative of either low rotation velocities or nearly face - on inclination angles ( e.g , ic2271 , kissr278 , kissr572 ) .
the profiles shown in figures [ figcap1 ] and [ figcap2 ] should be compared with the recovered flux integrals presented in table [ t3 ] and discussed in
the three systems with the smallest recovered s@xmath7 values in table [ t3 ] ( kissr278 , kissr396 , kissr572 ) are those systems with the narrowest h i profiles . note specifically that kissr278 , kissr396 , and kissr572 are each only detected in 4 , 7 , and 3 consecutive channels of their datacubes , respectively .
it is important to keep the low signal to noise ratio of some of these galaxies in mind during the discussion that follows .
single dish profiles of the six kissr galaxies in figures [ figcap1 ] and [ figcap2 ] were presented in @xcite .
a comparison of these two sets of profiles shows similar morphologies in some cases ; the profiles of kissr278 , kissr396 , and kissr561 shown in figures [ figcap1 ] and [ figcap2 ] are similar in shape as those shown in @xcite , although with slightly weaker peak flux densities in the interferometric global profiles .
conversely , the other three kissr systems ( kissr245 , kissr572 , and kissr1021 ) show significant differences in the shapes of their profiles ; the single - dish profiles not only recover more flux ( see detailed discussion in [ s2.2 ] above ) , but they also recover different morphologies ( note , for example , that the single dish profile of kissr1021 shows a pronounced double - horn morphology that is only poorly recovered in the interferometric global profile ) .
figures [ figcap3 ] through [ figcap10 ] compare the h i , optical , and infrared ( available for six of the eight systems , excluding ic2271 and adbs113845 + 2008 ) images of the eight sample galaxies . using the 2 kpc physical length denoted in panel ( b ) of each figure
, it is immediately evident that the high surface brightness stellar component of each of these eight galaxies is extremely compact .
in fact , some of these galaxies are only slightly more extended than foreground stars found in the fields [ note the close proximity of foreground stars to adbs113845 + 2008 ( figure [ figcap4 ] ) and kissr561 ( figure [ figcap8 ] ) ] ; we discuss the effects of resolution on our results in more detail below . while the h i surface brightness varies from one system to the next , each of these eight galaxies is detected at high significance in both the _ sdss _ and in the _ spitzer _ images .
an examination of the h i column density images and contours in figures [ figcap3 ] through [ figcap10 ] reveals that all of the galaxies in this sample , with the exception of adbs113845 + 2008 , harbor centrally concentrated neutral gas distributions at these spatial resolutions . in most cases
the optical and infrared surface brightness maxima are spatially coincident with the highest detected column density region ( e.g. , ic2271 in figure [ figcap3 ] , kissr572 in figure [ figcap9 ] ) ; in some systems there is a small offset between these two maxima , but most of these offsets are smaller than the synthesized beam sizes and are thus not significant .
the most significant discrepancy between the stellar and gaseous components is found in adbs113845 + 2008 : as discussed in detail in @xcite , the h i morphology is dominated by a broken ring structure that surrounds the optical body of the system .
the stellar component is many kpc away from the highest column density neutral gas .
similarly , an examination of the optical and infrared images of these sources reveals that most members of the sample are extremely compact .
however , it is important to stress that each of the systems is spatially resolved in the _ sdss _ images , the _ spitzer _ images , or both ; we discuss this in more detail in [ s4.1 ] below .
a few of the galaxies do have low surface brightness emission in the optical and infrared that may be indicative of the compact burst being only the highest surface brightness component of a larger stellar disk .
the most obvious systems with these properties are kissr278 ( figure [ figcap6 ] ) , kissr561 ( figure [ figcap8 ] , but note that the compact and high surface brightness object in the northern part of the stellar disk is a foreground star ) , and kissr572 ( figure [ figcap9 ] ) .
the images shown in figures [ figcap3 ] through [ figcap10 ] reveal that five of the eight neutral gas disks extend well beyond the beam size at this sensitivity level ; the h i disks of kissr278 , kissr396 , and kissr572 are the same size as the beam .
there are low surface brightness extensions of h i gas in some systems : ic2271 ( figure [ figcap3 ] ) and kissr1021 ( figure [ figcap10 ] ) harbor h i disks that are roughly twice the size of the stellar body ; kissr245 ( figure [ figcap5 ] ) and kissr561 ( figure [ figcap8 ] ) show tenuous h i filaments extending toward the southwest . however , the low signal to noise ratio of most of these datasets ( see discussion in [ s2 ] ) precludes a robust kinematic study of seven of the eight systems in this work .
the detailed dynamical analysis of adbs113845 + 2008 is already presented in @xcite , to which we refer the interested reader for details and results .
while dynamical analyses are not possible with most of the data presented here , the derived surface densities are robustly measured ( at this modest sensitivity level )
. stated differently , an analysis of the ( projected ) surface densities in each system does not require a detailed knowledge of the gas dynamics .
we thus limit our interpretation to those qualities and quantities of the neutral gas that do not depend on dynamical properties .
we focus on quantifying the relative sizes of the neutral gas and stellar disks , as a measure of the frequency of the `` giant disk '' phenomenon ; most of [ s4 ] is devoted to addressing this science goal .
one such well - studied metric is the empirical relationship between h i column density and ongoing star formation . while we do not have h@xmath1 imaging of all of these sources ,
each is a spectroscopically confirmed emission line source from kiss @xcite .
we make the implicit assumption that the majority of the ongoing star formation in these systems is spatially coincident with the high surface brightness stellar components shown in the optical images in figures [ figcap3 ] through [ figcap10 ] .
we note explicitly that the most extreme member of this sample , adbs113845 + 2008 , has an h@xmath1 morphology that is essentially indistinguishable from that of the r - band image shown in figure [ figcap4 ] ( see , e.g. , figure 5 of cannon etal .
2009 ) .
a gas column density of 10@xmath23 @xmath10 has been shown to trace locations of active star formation ( e.g. , skillman 1987 ; kennicutt 1989 ; kennicutt 1998 ; and references therein ) .
interestingly , in all of the galaxies , except ic 2271 and kissr561 , the peak observed h i column densities are below this 10@xmath23 @xmath10 level , even when the h i column density and stellar surface brightness maxima are spatially coincident .
this is consistent with the results of other studies of dirr systems ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
it is important to note that beam smearing may be removing the signatures of high column density gas at our current resolutions of 15 and 20 , so we are cautious not to over - interpret this signature at the present time .
higher resolution and sensitivity observations of these galaxies , while requiring extensive amounts of additional observing time , would be very insightful in revealing the small - scale structure of the neutral gas disks .
the enigmatic properties of adbs113845 + 2008 led @xcite to argue for an evolutionary scenario where isolation allows for a very inefficient mode of star formation or , alternatively , very inefficient gas consumption .
one might naively expect that systems with similar optical properties might also be similarly isolated .
we were thus somewhat surprised to discover that the majority of the sample members ( five of eight systems ) are in relative physical proximity to other galaxies ; we hereafter refer to these systems as loose associations .
h i and optical images of these loose associations are shown in figure [ figcap11 ] ( ic 2271 ) , figure [ figcap12 ] ( kissr245 ) , figure [ figcap13 ] ( kissr396 ) , figure [ figcap14 ] ( kissr561 ) , and figure [ figcap15 ] ( kissr572 ) . using the 10 kpc physical length denoted in panel ( b ) of each of these figures , one can discern the ( projected ) separation of the galaxies in each loose association .
we draw attention to the ( projected ) separation of each association member from the compact dwarf system studied here in the caption of each figure and in table [ t5 ] ; the minimum and maximum separations are 26 kpc ( ic2271 and sdssj081808.31 + 243006.4 ) and 230 kpc ( kissr572 and sdss j144754.68 + 291929.4 ) .
a commonly posited mechanism to initiate a burst of star formation ( i.e. , a starburst ) is a tidal interaction .
dynamical instabilities can drive gas deep into the potential well , initiating spatially and temporally concentrated star formation ( e.g. , * ? ? ?
* ; * ? ? ?
while a direct signature of a tidally triggered star formation event has been difficult to produce observationally , one might suspect that extended neutral gas in the form of tidal tails might be considered to be an indirect signature of interaction - driven evolution ( e.g. , the extended tidal structure in the m81 group of galaxies ; * ? ? ?
* ) . to search for such signs of interactions amongst the loose associations studied here , we examined the moment zero ( integrated h i intensity ) maps and full datacubes at a variety of spatial resolutions ( from 15 to 60 ) for signatures of tidal material between the compact dwarf systems and other association members .
some systems have asymmetric , low surface brightness extensions of h i gas away from the central disk ( see discussion of ic2271 , kissr1021 , kissr245 , and kissr561 above ) that can be interpreted as weak evidence for gravitational interactions playing a role in their recent evolution . none of the loose associations shown in figures [ figcap11 ] through [ figcap15 ] show direct evidence of strong tidal interaction in the form of tails or diffuse gas ( at the n@xmath7 @xmath22 10@xmath14 @xmath10 level ) connecting one system to another .
the most promising candidate for a strong interaction scenario might be the loose association containing ic2271 ; there is some low surface brightness h i gas surrounding multiple galaxies in this field . given the lack of very short _ uv _ spacings in the data presented here ,
the surface brightness sensitivity to large - scale structure in these images is only modest .
we conclude that the membership of the majority of these star - forming systems in loose associations is intriguing but weak evidence for tidally triggered star formation events ; very deep observations of these systems with the compact _ vla _ d configuration could be very fruitful in further quantifying the role of interactions in initiating starburst events .
the primary goal of the present investigation is to quantify the relative sizes of the neutral gas and stellar disks of each of the optically compact dwarf galaxies in our sample , thus estimating the frequency of the `` giant disk '' phenomenon exemplified by adbs113845 + 2008 and discussed in detail in @xcite . to make this comparison ,
we create radially averaged surface brightness profiles of each system in the optical , infrared , and h i spectral line .
we measure the exponential scale lengths at optical and infrared wavelengths , and compare these to both the exponential scale length of the neutral gas disks ( when possible ) and to the sizes of the gaseous disks at the 10@xmath14 @xmath10 level .
it is important to emphasize that while this comparison does not include dynamical parameters for each system ( e.g. , an inclination derived from dynamical modeling of an individual galaxy s velocity field ) , the comparisons discussed below do use the same corrections ( when applicable ) for galaxy inclination at each wavelength .
similarly , our analysis uses matched apertures between wavelengths to assure self - similar measurements .
we define a galaxy as being `` resolved '' in a certain image if the derived scale length or size exceeds the fwhm or beam size .
therefore , we classify systems as `` resolved '' in the _ sdss _ r - band , the _
3.6 @xmath20 m band , or the _ spitzer _ 4.5 @xmath20 m band , if the derived scale lengths exceed 1.3 , 1.66 , or 1.72 , respectively .
similarly , we classify systems as `` resolved '' in h i if the size of the h i disk at the 10@xmath14 @xmath10 column density level exceeds the 15 beam size .
we discuss the details of each of these metrics in the subsections that follow .
optical and infrared surface brightness profiles for each of the galaxies were produced by fitting elliptical isophotes to the _ sdss _ r - band , _ spitzer _ 3.6 @xmath20 m , and _ spitzer _
4.5 @xmath20 m images using the iraf task ellipse .
as noted above , some of these systems have easily identifiable foreground stars that can contaminate the surface brightness profiles [ e.g. , adbs113845 + 2008 ( figure [ figcap4 ] ) and kissr561 ( figure [ figcap8 ] ) ] .
such foreground objects were masked by hand using the iraf task fixpix .
for each system , the position angle and inclination were determined from the _ sdss _ r - band image .
when a system was extended in one dimension , the position angle was defined as that angle east of north that aligns with the larger projected optical axis .
the inclination was defined by @xmath24 where _ a _ and _ b _ are the major and minor axes , respectively .
the ellipse parameters used to derive the isophotal photometry are given in table [ t6 ] .
surface photometry was measured in elliptical isophotes incremented in one - pixel steps along the major axis ( 0.4 per ellipse for the _ sdss _ images and 0.6 per ellipse for the _ spitzer _ images ) .
the r - band ellipses are shown in red in figures [ figcap3 ] though [ figcap10 ] , while the infrared ellipses are shown in green in figures [ figcap5 ] though [ figcap10 ] .
figures [ figcap16 ] and [ figcap17 ] show the resulting surface brightness plots for all eight of the sample galaxies . as expected
based on the selection criteria used to create the sample ( two of which were an optical radius less than 3 and an upper limit to the recessional velocity or distance ; see discussion in [ s2 ] above ) , the profiles are all centrally peaked and show little structure throughout the optical disks of the galaxies . while there is some variation in the steepness of the profile in the inner regions of the systems ( i.e. , within the inner few arcseconds )
, it is immediately obvious that each system harbors an exceptionally compact stellar component . to quantify the sizes of these compact stellar populations , we next derived the scale lengths of each system in the _ sdss _ r - band and in the _ spitzer _ 3.6 @xmath20 m and 4.5 @xmath20 m bands . in deriving these profiles
, we follow the methodology in @xcite . using logarithmic units of mag arcsec@xmath25
, a very good approximation for the exponential disk component of the optical radial profile can be made using : @xmath26 where @xmath27 is the central surface brightness in units of mag arcsec@xmath25 , _ r _ is the radius , @xmath28 is the surface brightness in units of mag arcsec@xmath25 as a function of radius , and _ h _ is the scale length .
table [ t7 ] provides the calculated scale lengths ( in both angular and in physical units ) at each wavelength , derived from the profiles shown in figures [ figcap16 ] and [ figcap17 ] , with upper limit errors of 25% at each ellipse location .
note that we use the nomenclature of _ h_@xmath29 , _
h_@xmath30 , and _
h_@xmath31 to denote the scale lengths ( in angular units and in physical units of kpc ) of each galaxy in the _ sdss _ r - band , the _
3.6 @xmath20 m band , and the _ spitzer _ 4.5 @xmath20 m band , respectively .
the scale lengths presented in table [ t7 ] show that all of the galaxies are resolved in at least one optical or infrared image , and that most are resolved in both .
recall that the fwhm of the _ sdss _ r - band , _ spitzer _ 3.6 @xmath20 m , and _
4.5 @xmath20 m images are 1.3 , 1.66 , and 1.72 , respectively .
values in table [ t7 ] show that all systems except for kissr245 are resolved in the _ sdss _ images ( that is , _ h_@xmath29 exceeds the 1.3 fwhm of the _ sdss _ r - band images ) .
h_@xmath29 value for kissr245 is formally smaller than the fwhm , although it is consistent within the errorbar .
similarly , _ spitzer _ resolves each galaxy ; note , however , that the scale lengths of kissr245 at 3.6 @xmath20 m and of kissr1021 at 4.5 @xmath20 m are consistent with the fwhm , given the errorbars .
the optical and infrared scale lengths in table [ t7 ] quantify the sizes of the compact stellar components of the sample galaxies . at the adopted distances , the stellar component of each system
has an r - band scale length much less than 1 kpc .
all systems have _
@xmath320.5 kpc , except for kissr561 .
moving to the infrared , the scale lengths increase .
this is exactly the behavior expected of an older and non - star forming stellar population .
while each of the infrared scale lengths exceeds the r - band value , it is remarkable to note that each system is still exceptionally compact even when considering the older stellar populations .
for example , at 4.5 @xmath20 m , some systems have scale lengths slightly larger than 1 kpc ( kissr278 , kissr561 , kissr572 ) , but others show very little increase in scale length with increasing wavelength ( consider kissr1021 , whose scale length does not change between the optical and the infrared , within the measurement errors ) . having quantitatively established the compact nature of the stellar population of each of the sample galaxies , we now seek to quantify the sizes of the h i neutral gas disks via two methods .
first , we derive radially averaged surface brightness profiles for each system ; when possible , we use these profiles to derive an exponential scale length .
second , we quantify the size of the disks at the 10@xmath14 @xmath10 levels . we discuss each method in detail below .
we begin by creating radial h i surface brightness profiles for each of the galaxies using the 15 beam images and the gipsy task ellint . as an inspection of figures [ figcap3 ] through [ figcap10 ] shows ,
the neutral gas disks studied here are larger than the beam size for five of the eight galaxies ( excluding kissr278 , kissr396 , and kissr572 ) .
a detailed dynamical analysis is only possible for adbs113845 + 2008 ; this precludes the derivation of kinematic position angles and inclinations for most galaxies . in order to assure an accurate comparison of stellar and gaseous properties , we thus use the same position angles and inclinations as derived using the _ sdss _ r - band images ( see table [ t6 ] and the detailed discussion in [ s4.1 ] above ) .
these ellipses are shown in blue in figures [ figcap3 ] through [ figcap10 ] .
a semi - major axis and step size of 5 was used to create each of the h i profiles ; technically this means that the profiles we discuss below are oversampled ( compared to the 15 beam size ) by a factor of 3 .
however , this oversampling does allow us to extract meaningful radially averaged surface brightness profiles .
figures 18 and 19 show these surface brightness profiles for each of the eight sample galaxies . as was evident from the surface density images presented in [ s3.1 ] ,
most of the sample galaxies have h i surface brightness profiles that are centrally peaked at the resolution of these observations and that fall off with radius in a more or less smooth way .
note that low surface brightness features in some of the galaxies moment zero ( integrated h i intensity ) maps do cause departures from smooth profiles in the outer disks ; for example , the low surface brightness feature extending to the southwest of kissr245 in figure [ figcap5 ] is easily identified as the slight increase in h i surface density between @xmath1325 and @xmath1350 in figure [ figcap18 ] . the obvious exception to this trend is adbs113845 + 2008 , whose radially averaged h i surface density profile ( created with the optically derived inclination and position angle discussed above , and shown in red in figure [ figcap18 ] ) rises with increasing galactocentric radius , peaks , and then falls off in the outer disk .
the peak surface densities are located in the broken ring structure @xmath137.5 kpc from the stellar component .
a cursory comparison of the profiles in figures [ figcap18 ] and [ figcap19 ] already reveals the striking differences between adbs113845 + 2008 and the other compact dwarf galaxies in the present sample ; we discuss these differences in more detail below , and refer the interested reader to @xcite for a more complete discussion of the remarkable characteristics of the adbs113845 + 2008 system .
the radial profiles shown in figures [ figcap18 ] and [ figcap19 ] ( specifically , the red profile for adbs 113845 + 2008 ) were created using exactly the same inclinations and position angles as used to derive the optical and infrared surface brightness profiles in figures [ figcap16 ] and [ figcap17 ] . using these neutral gas radial profiles , exponential fits were made using a version of equation [ eq1 ] converted to linear units : @xmath33 where @xmath34 is the central surface brightness of the disk in m@xmath12 pc@xmath25 , _ r _ is the radius , @xmath35 is the surface brightness of the disk in m@xmath12 pc@xmath25 as a function of radius , and _
h _ is the scale length @xcite .
we hereafter refer to the h i scale length derived using this equation as _ h_@xmath7 .
we tabulate the derived _
values in table [ t7 ] . for all of the systems except adbs113845 + 2008 ,
the calculated values of _
h_@xmath7 are smaller than the beam size .
formally , this metric to measure the size of the h i disks only provides meaningful results when it can be fit over multiple independent beams .
as noted above , our radially averaged profiles required over - sampling by a factor of three . while the sizes of the h i disks based on the _ h_@xmath7 metric are interesting , and we discuss these briefly below , we seek an alternative measure of the size of the neutral gas disks of these galaxies .
it is obvious from the neutral gas morphology ( figure [ figcap3 ] ) and surface brightness profile ( figure [ figcap18 ] ) of adbs113845 + 2008 that an exponential scale length is undefined for a disk whose morphology is not centrally concentrated . while we considered fitting the outer disk of adbs113845 + 2008 ( i.e. , that region at radii larger than that of the broken h i ring ) with an exponential function , the striking disparity between the neutral gas and stellar disks of this system warrants caution against such an approach . due to the remarkable qualities of this galaxy s neutral gas disk
, we thus do not calculate an exponential h i scale length .
the properties of adbs113845 + 2008 naturally require a second metric for the size of the neutral gas disks of these systems . to address these issues
, we thus use the radially averaged h i surface density profiles in figures [ figcap18 ] and [ figcap19 ] to determine the ( radially averaged ) size of the h i disk at the 10@xmath14 @xmath10 column density level .
this column density corresponds to a mass surface density of 0.08 m@xmath12 pc@xmath25 ; recall from the discussion in [ s2.2 ] that this emission is above the 2@xmath18 level in at least 3 consecutive channels .
we extract the radii of the sample galaxies at this level ( hereafter referred to as r@xmath7 ) from figures [ figcap18 ] and [ figcap19 ] and tabulate the values in table [ t7 ] ; note that @xcite quantified the physical size of the adbs113845 + 2008 disk in the same manner . as expected , r@xmath7
h_@xmath7 in all systems where both quantities are defined .
it is important to stress that for kissr245 , kissr278 , kissr396 , and kissr572 , r@xmath7 is equal ( within errors ) to the beam size ; this means that the r@xmath7 metric for these galaxies is formally an upper limit ( but note that kissr245 does have h i gas extending to the southwest , well beyond a single beam size ) .
these four systems are `` unresolved '' by the present observations ( that is , r@xmath7 @xmath5 beam size , thus the r@xmath7 and related values being listed as upper limits in tables [ t7 ] and [ t8 ] ) ; they may harbor smaller neutral gas components than calculated here .
we discuss the ratios of both _
h_@xmath7/_h_@xmath29 and r@xmath7/_h_@xmath29 for all eight sample galaxies in [ s4.3 ] .
we again note that no kinematic constraints are used to derive the scale lengths or disk sizes discussed here ; the profiles shown in figures [ figcap16 ] through [ figcap19 ] use only optically - derived position angles and inclinations . since seven of the eight galaxies
are not sufficiently resolved to undertake a robust dynamical analysis , we can only test these dependences for the one system that does have a full kinematic model , adbs113845 + 2008 . to test how using the position angles and inclinations derived from the optical images affect our calculations ( as opposed to using those derived from the h i dynamics ) , we show three radial mass surface density profiles for adbs113845 + 2008 in figure [ figcap18 ] .
the red profile is derived using the optical inclination ( 0@xmath37 ) and position angle ( 0@xmath37 ) and is discussed above .
the blue profile uses the dynamically - derived position angle ( 155@xmath37 ) and inclination ( 30@xmath37 ) from @xcite .
the green profile uses the same position angle ( 155@xmath37 ) as in @xcite , but assumes a larger value for the galaxy s inclination angle ( 60@xmath37 ) . since _
h_@xmath7 is undefined for adbs113845 + 2008 , it is difficult to use this test to assess the dependence of h i scale length on inclination or position angle .
the r@xmath7 values , on the other hand , do change slightly for these different fits ( r@xmath7 @xmath2 17.3 @xmath19 3.5 kpc , 18.1 @xmath19 3.6 kpc , and 22.3 @xmath19 4.5 kpc , for the red , blue , and green profiles in figure [ figcap18 ] , respectively ) , but the resulting h i disk radii are equal within the ( substantial ) errorbars . while differences from one system to another are expected
, this simple test gives a baseline for how to interpret the measurements made using the position angles and inclinations derived from the _ sdss _ r - band images within the rest of the sample .
we consider the r@xmath7 values for ic2271 , adbs113845 + 2008 , kissr561 , and kissr1021 to be accurate representations of their ( average ) physical size at the 10@xmath14 @xmath10 level .
the r@xmath7 values in table [ t7 ] for the other four systems should be considered upper limits on the physical size of each neutral gas disk .
when available , we also tested how the disk size measurements change when using c configuration versus d configuration data .
this comparison can only be made for adbs113845 + 2008 and ic2271 .
the scale length measurement of ic2271 with the lower resolution d configuration data is _
kpc ( [email protected] ) ; this can be compared with the _ h_@[email protected] kpc ( [email protected] ) value derived from the c configuration data only ( see table [ t7 ] ) .
the radii of the galaxy disks , when using the c versus d configuration datasets , change from r@[email protected] kpc to [email protected] kpc ( ic2271 ) and from [email protected] kpc to [email protected] kpc ( adbs113845 + 2008 ) .
while these increases are slightly larger than the sizes of the errorbars , we conclude that there is only a modest dependency of the size estimates of these galaxy disks on the use of only the c configuration data .
given that our analysis uses the same imaging strategy for all systems , the surface brightness sensitivity of each dataset is very comparable .
we can now compare , in a quantitative sense , the relative sizes of the stellar and gaseous disks of the eight galaxies in this sample .
table [ t8 ] contains the ( dimensionless ) ratios of _
h_@xmath7 and r@xmath7 versus
_ h_@xmath29 , _
h_@xmath30 , and _
h_@xmath31 , for all galaxies with available data ( recall that adbs113845 + 2008 has an undefined _ h_@xmath7 , and that no _ spitzer _ imaging is available for adbs113845 + 2008 or for ic2271 ) . from the contents of this table , we note the following trends and behaviors . as discussed in [ s4.2 ] above , four of the eight systems have h i disks that exceed the sizes of the stellar bodies based on the r@xmath7 metric ( and are thus considered `` resolved '' in h i ) . hence , the dimensionless ratios in table [ t8 ] for kissr245 , kissr278 , kissr396 , and kissr572 should be considered upper limits ; the actual size ratios of neutral to stellar disks may in fact be smaller for these systems . for the entire sample , the ratios of _
h_@xmath7 to r - band or infrared scale lengths ( spanning a range of @xmath132 to @xmath137 ) are systematically smaller than the ratios of r@xmath7 to the same r - band or infrared values ( which span a range of @xmath134 to @xmath1358 ) . similarly , for each individual galaxy , the relative size ratios measured using either r@xmath7 or _
h_@xmath7 decrease with increasing wavelength used to measure the stellar disk size .
these general trends hold across the entire sample , regardless of which tracer is used to measure the size of either the stellar or the gaseous disk . the r@xmath7/_h_@xmath29 ratio for adbs113845 + 2008 ( 58@xmath1921 ) is more than a factor of two larger than the same ratio for any of the other systems ; if the ratios in table [ t7 ] for the other sample members are smaller , then this reinforces the enormity of the h i disk of adbs113845 + 2008 . using the other seven sample galaxies as benchmarks , it is physically intuitive that the infrared scale length of adbs113845 + 2008 would be larger than the measured _
h_@xmath29 value .
however , the enormous r@xmath7/_h_@xmath29 ratio of this system , and the relative size of the broken h i ring ( @xmath137.5 kpc in radius ) versus the optical scale length ( [email protected] kpc ) , are challenging parameters to explain unless the optical body is only a high surface brightness knot in an otherwise extremely low surface brightness underlying stellar disk .
we again refer the interested reader to @xcite for more detailed discussion ; for the purposes of this work , we interpret the remarkable properties of adbs113845 + 2008 as the signature of a `` giant '' h i disk .
the present sample of compact dwarf galaxies was created using consistent selection criteria ; the observations and analysis have produced a uniform set of results . while the sample is small , it appears that the frequency of `` giant disk '' dwarf galaxies such as adbs113845 + 2008 is low .
a few other examples of such systems have been identified ( e.g. , ddo154 , carignan & beaulieu 1989 , carignan & purton 1998 ; ngc2915 , meurer etal .
1996 ; ugc5288 , van zee 2004 ; ngc3741 , bremnes etal .
2000 , begum etal .
this small collection of galaxies suggests that systems with qualities similar to adbs113845 + 2008 are comparatively rare in the local volume
. a larger sample of h i observations of local compact dwarf systems could be fruitful in improving the statistics of the trends identified in table [ t8 ] .
we have presented _ vla _ c configuration imaging of eight optically compact dwarf galaxies .
these sources were selected to share the following common characteristics : 1 ) compact stellar populations ; 2 ) faint blue absolute magnitudes ( i.e. , low masses ) ; 3 ) emission lines indicative of recent star formation ; and 4 ) proximity for spatially resolved work .
the sample includes the `` giant disk '' dwarf galaxy adbs113845 + 2008 studied in @xcite .
the _ vla _
data are supplemented by archival optical ( _ sdss _ r - band ) and infrared ( _ spitzer _ 3.6 @xmath20 m and 4.5 @xmath20 m ) imaging .
we use this data suite to quantitatively address the relative sizes of the stellar and gaseous components of these galaxies .
a comparison of the morphology of the stellar and gaseous disks of each galaxy in the sample reveals that all eight systems have extremely compact ( though in almost all cases resolved ) stellar components ; this is of course expected given the selection criteria .
seven of the eight systems in the sample harbor centrally concentrated neutral gas distributions .
adbs113845 + 2008 differs from the other systems in that its neutral gas disk is structured in a broken ring that is well outside of the stellar component . in most of the other cases , the stellar and gaseous disks have surface brightness maxima that are in close physical proximity to one another .
five of the eight systems are located in loose associations , with neighboring systems with projected separations of less than a few hundred kpc and with recessional velocities within a few hundred kms@xmath3 .
some of the loose association galaxies have low surface brightness neutral gas in the outer disks that might be construed as evidence of tidal interactions .
however , given the lack of very short _ uv _ spacings in our data , we conservatively conclude that the membership of these star - forming dwarfs in such loose associations is only weak evidence for a tidal burst trigger . while each of the eight sample galaxies is detected in h i at meaningful significance , these _ vla _ data do not allow a robust dynamical analysis in any of the systems except for adbs113845 + 2008 .
thus , we concentrate on the morphological properties of the galaxies .
we use the optical , infrared , and h i data to derive radially averaged surface brightness profiles that can be used to measure the sizes of both the stellar and the gaseous components .
we quantify the sizes of the h i disks via both an exponential scale length ( available for all systems except for adbs113845 + 2008 ) and a ( radially averaged ) limit of the gaseous disk at the 10@xmath14 @xmath10 column density ( 0.08 m@xmath12 pc@xmath25 ) level .
we find that four of the eight systems ( specifically , ic2271 , adbs113845 + 2008 , kissr561 , and kissr1021 ) have larger gaseous disks than stellar disks ; the remaining four ( kissr245 , kissr278 , kissr396 , and kissr572 ) have only upper limits on the physical size of their h i disks ( but note that kissr245 has low - surface brightness h i gas extending well beyond a single beam size ; see figure [ figcap5 ] ) .
when parameterized by an h i disk scale length , the neutral gas disks of the seven available galaxies ( excluding adbs113845 + 2008 ) are larger than the stellar disks by factors of @xmath1327 . when parameterized by the size of the h i disk at the 10@xmath14 @xmath10 column density level , the neutral gas disks of the eight sample galaxies ( including adbs113845 + 2008 ) are larger than the stellar disks by factors of @xmath13458 .
for each galaxy , the relative size ratios measured using either r@xmath7 or _ h_@xmath7 decrease with increasing wavelength used to measure the stellar disk size ( i.e. , the optical scale lengths are slightly shorter than the infrared scale lengths ) . if the values of r@xmath7 and _ h_@xmath7 are upper limits , then these ratios decrease .
adbs113845 + 2008 stands out among the members of this sample as harboring a `` giant '' h i disk .
while the h i - to - optical size ratios of some other members of the sample are substantial ( e.g. , ic2271 , kissr245 ; see table [ t8 ] ) , this ratio for adbs113845 + 2008 is more than twice as large as any other measured here .
as discussed in @xcite , the physical properties of this galaxy are extreme and enigmatic ; adbs113845 + 2008 has one of the largest known h i - to - optical size ratios in the local universe . given that the present sample was selected _ only _ based on optically - derived properties , and that the observations of the present sample were acquired and analyzed in a uniform way , suggests that this `` giant disk '' dwarf galaxy is rare in the local volume .
the present sample is small and is limited in number by available ( single dish h i ) data ; the statistics behind the present results could be significantly strengthened by a larger study of additional extremely compact , star - forming dwarf galaxies in the local volume .
this research was supported by a cottrell college science award from research corporation .
jlr acknowledges support from nsf ast-000167932 .
the authors thank the national radio astronomy observatory and macalester college for partial support of this work , and dr .
gustaaf van moorsel for helpful discussions .
this work is based in part on archival data obtained with the _
spitzer space telescope _ , which is operated by the jet propulsion laboratory , california institute of technology under a contract with nasa .
support for this work was provided by an award issued by jpl / caltech .
this work has made use of data from the _ sloan digital sky survey_. funding for the _ sdss _ and _ sdss - ii _ has been provided by the alfred p. sloan foundation , the participating institutions , the national science foundation , the u.s .
department of energy , the national aeronautics and space administration , the japanese monbukagakusho , the max planck society , and the higher education funding council for england .
the _ sdss _
web site is http://www.sdss.org/. the _ sdss _ is managed by the astrophysical research consortium for the participating institutions .
the participating institutions are the american museum of natural history , astrophysical institute potsdam , university of basel , university of cambridge , case western reserve university , university of chicago , drexel university , fermilab , the institute for advanced study , the japan participation group , johns hopkins university , the joint institute for nuclear astrophysics , the kavli institute for particle astrophysics and cosmology , the korean scientist group , the chinese academy of sciences ( lamost ) , los alamos national laboratory , the max - planck - institute for astronomy ( mpia ) , the max - planck - institute for astrophysics ( mpa ) , new mexico state university , ohio state university , university of pittsburgh , university of portsmouth , princeton university , the united states naval observatory , and the university of washington .
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 , and nasa s astrophysics data system .
lcccccc ic 2271 & 08:18:19.7 & 24:31:37 & 2184 @xmath19 10 & 29.9 & 25.1 @xmath19 6.3 & 101 + adbs 113845 + 2008 & 11:38:45.2 & 20:08:24 & 3054 @xmath19 3 & 41.8 & 40.2 @xmath19 10.1 & 84.6 + kissr245 & 13:16:27.9 & 29:25:11 & 4762 @xmath19 5 & 65.2 & 13.0 @xmath19 3.3 & 28.4 + kissr278 & 13:23:37.7 & 29:17:17 & 4022 @xmath19 5 & 55.1 & 4.4 @xmath19 1.1 & 12.0 + kissr396 & 13:57:10.0 & 29:13:10 & 2238 @xmath19 5 & 30.7 & 1.0
@xmath19 0.3 & 11.6 + kissr561 & 14:29:53.6 & 29:20:11 & 3750 @xmath19 5 & 51.4 & 34.3 @xmath19 8.6 & 56.5 + kissr572 & 14:46:48.2 & 29:25:18 & 3693 @xmath19 5 & 50.6 & 2.3 @xmath19 0.6 & 16.6 + kissr1021 & 16:19:02.5 & 29:10:22 & 2519 @xmath19 5 & 34.5 & 7.0 @xmath19 1.8 & 14.5 + [ t1 ] lccccc ic2271 & 2006 oct .
31 & 241 & 3.12 & 64 & 10.30 + adbs113845 + 2008 & 2006 oct .
31 & 245 & 1.56 & 128 & 2.58 + kissr245 & 2008 apr .
23 & 252 & 3.12 & 128 & 5.15 + kissr278 & 2008 apr .
29 & 253 & 3.12 & 128 & 5.15 + kissr396 & 2008 may 1 & 254 & 3.12 & 128 & 5.15 + kissr561 & 2008 may 13 & 252 & 3.12 & 128 & 5.15 + kissr572 & 2008 may 8 & 251 & 3.12 & 128 & 5.15 + kissr1021 & 2008 may 15 & 255 & 3.12 & 128 & 5.15 + [ t2 ] lcccc ic2271 & 2121 - 2246 & 0.39 & 1.22 @xmath19 0.12 & 4.702 + adbs113845 + 2008 & 2987 - 3117 & 0.73 & 0.98 @xmath19 0.10 & 1.982 + kissr245 & 4744 - 4779 & 0.52 & 0.13 @xmath19 0.01 & 0.28 + kissr278 & 4014 - 4030 & 0.48 & 0.062 @xmath19 0.006 & 0.17 + kissr396 & 2221 - 2253 & 0.53 & 0.043
@xmath19 0.004 & 0.51 + kissr561 & 3695 - 3809 & 0.50 & 0.55 @xmath19 0.06 & 0.89 + kissr572 & 3688 - 3699 & 0.48 & 0.038 @xmath19 0.004 & 0.27 + kissr1021 & 2473 - 2562 & 0.46 & 0.25 @xmath19 0.03 & 0.51 + [ t3 ] lccc ic2271 & 1.2 & 1.5 & 1.5 + adbs113845 + 2008 & 0.98 & 0.90 & 0.90 + kissr245 & 0.13 & 0.14 & 0.15 + kissr278 & 0.062 & 0.30 & - + kissr396 & 0.043 & 0.038 & 0.039 + kissr561 & 0.55 & 0.65 & 0.66 + kissr572 & 0.038 & 0.028 & - + kissr1021 & 0.25 & 0.35 & 0.23 + [ t4 ] lcccc ic2271 & adbs j081707 + 2433 & 995 & 121 & 144 + & ic 2267 & 790 & 146 & 115 + & sdss j081808.31 + 243006.4 & 180 & 105 & 26 + kissr245 & maps - ngp_o_323_0594797 & 477 & 12 & 151 + kissr396 & ngc 5375 & 270 & 137 & 40 + & ngc 5375a & 633 & 18 & 94 + kissr561 & ngc 5657 & 861 & 110 & 215 + kissr572 & sdss j144754.68 + 291929.4 & 936 & 33 & 230 + [ t5 ] lcc ic2271 & 143@xmath37 & 45@xmath37 + adbs113845 + 2008 & 0@xmath37 & 0@xmath37 + kissr245 & 0@xmath37 & 0@xmath37 + kissr278 & 40@xmath37 & 73@xmath37 + kissr396 & 175@xmath37 & 43@xmath37 + kissr561 & 0@xmath37 & 57@xmath37 + kissr572 & 60@xmath37 & 70@xmath37 + kissr1021 & 130@xmath37 & 48@xmath37 + [ t6 ] lccccc ic2271 & 0.33 @xmath19 0.08 & n / a & n / a & 1.72 @xmath19 0.43 & 8.74 @xmath19 2.19 + adbs113845 + 2008 & 0.30 @xmath19 0.08 & n / a & n / a & - & 17.3 @xmath19 4.3 + kissr245 & 0.38 @xmath19 0.10 & 0.55 @xmath19 0.14 & 0.70 @xmath19 0.18 & 2.48 @xmath19 0.62 & @xmath168.28 @xmath19 2.07 + kissr278 & 0.41 @xmath19 0.10 & 0.80 @xmath19 0.20 & 1.36 @xmath19 0.34 & 2.64 @xmath19 0.66 & @xmath166.19 @xmath19 1.55 + kissr396 & 0.19 @xmath19 0.05 & 0.31 @xmath19 0.08 & 0.36 @xmath19 0.09 & 0.78 @xmath19 0.20 & @xmath162.65 @xmath19 0.66 + kissr561 & 0.72 @xmath19 0.18 & 1.30 @xmath19 0.33 & 1.20 @xmath19 0.30 & 2.59 @xmath19 0.65 & 11.8 @xmath19 3.0 + kissr572 & 0.50 @xmath19 0.13 & 0.77 @xmath19 0.19 & 1.18 @xmath19 0.30 & 2.19 @xmath19 0.55 & @xmath164.64 @xmath19 1.16 + kissr1021 & 0.29 @xmath19 0.07 & 0.37 @xmath19 0.09 & 0.31 @xmath19 0.08 & 1.18 @xmath19 0.30 & 5.39 @xmath19 1.35 + & & & & & + & & & & & + ic2271 & 2.29 @xmath19 0.57 & n / a & n / a & 11.9 @xmath19 3.0 & 60.3 @xmath19 15.1 + adbs113845 + 2008 & 1.44 @xmath19 0.36 &
n / a & n / a & - & 85.3 @xmath19 21.3 + kissr245 & 1.18 @xmath19 0.30 & 1.71 @xmath19 0.43 & 2.21 @xmath19 0.55 & 7.84 @xmath19 1.96 & @xmath1626.2 @xmath19 6.6 + kissr278 & 1.51 @xmath19 0.38 & 2.99 @xmath19 0.75 & 5.12 @xmath19 1.28 & 9.88 @xmath19 2.47 & @xmath1623.2 @xmath19 5.8 + kissr396 & 1.33 @xmath19 0.33 & 2.13 @xmath19 0.53 & 2.50 @xmath19 0.63 & 5.25 @xmath19 1.31 & @xmath1617.8 @xmath19 4.5 + kissr561 & 2.93 @xmath19 0.73 & 5.27 @xmath19 1.32 & 4.86 @xmath19 1.22 & 10.4 @xmath19 2.6 & 47.4 @xmath19 11.9 + kissr572 & 2.07 @xmath19 0.52 & 3.18 @xmath19 0.80 & 4.87 @xmath19 1.22 & 8.93 @xmath19 2.23 & @xmath1618.9 @xmath19 4.7 + kissr1021 & 1.74 @xmath19 0.44 & 2.23 @xmath19 0.56 & 1.86 @xmath19 0.47 & 7.05 @xmath19 1.76 & 32.2 @xmath19 8.1 + [ t7 ] lcccccc ic2271 & 5.2 @xmath19 1.8 & 26 @xmath19 9 &
n / a & n / a & n / a & n / a + adbs113845 + 2008 & - & 58 @xmath19 21 & n / a & n / a & n / a & n / a + kissr245 & 6.5 @xmath19 2.4 & @xmath1622 @xmath19 8 & 4.5 @xmath19 1.6 & @xmath1615 @xmath19 5 & 3.5 @xmath19 1.3 & @xmath1612 @xmath19 4.2 + kissr278 & 6.4 @xmath19 2.2 & @xmath1615 @xmath19 5 & 3.3 @xmath19 1.2 & @xmath167.7 @xmath19 2.7 & 1.9 @xmath19 0.7 & @xmath164.6 @xmath19 1.6 + kissr396 & 4.1 @xmath19 1.5 & @xmath1614 @xmath19 5 & 2.5 @xmath19 0.9 & @xmath168.5 @xmath19 3.0 & 2.2 @xmath19 0.8 & @xmath167.4 @xmath19 2.6 + kissr561 & 3.6 @xmath19 1.3 & 16 @xmath19 6 & 2.0 @xmath19 0.7 & 9.1 @xmath19 3.2 & 2.2 @xmath19 0.8 & 9.8 @xmath19 3.5 + kissr572 & 4.4 @xmath19 1.6 & @xmath169.3 @xmath19 3.3 & 2.8 @xmath19 1.0 &
@xmath166.0 @xmath19 2.1 & 1.9 @xmath19 0.7 & @xmath163.9 @xmath19 1.4 + kissr1021 & 4.1 @xmath19 1.4 & 19 @xmath19 6 & 3.2 @xmath19 1.1 & 14.6 @xmath19 5.2 & 3.8 @xmath19 1.4 & 17 @xmath19 6.2 + [ t8 ] | we present new _ very large array _
h i spectral line , archival _ sloan digital sky survey _ , and archival _ spitzer space telescope _
imaging of eight star - forming blue compact dwarf galaxies that were selected to be optically compact ( optical radii < 1 kpc ) .
these systems have faint blue absolute magnitudes ( m@xmath017 ) , ongoing star formation ( based on emission - line selection by the h@xmath1 or [ ] lines ) , and are nearby ( mean velocity @xmath2 3315 kms@xmath3 @xmath4 45 mpc ) .
one galaxy in the sample , adbs 113845 + 2008 , is found to have an h i halo that extends 58 r - band scale lengths from its stellar body .
in contrast , the rest of the sample galaxies have h i radii to optical - scale - length ratios ranging from 9.3 to 26 .
the size of the h i disk in the `` giant disk '' dwarf galaxy adbs 113845 + 2008 appears to be unusual as compared to similarly compact stellar populations . |
we consider a tournament among players @xmath0 , in which each match involves two players . the tournament is assumed to be of knockout type in that the losers of matches are eliminated and do not move on to the next round , and the tournament continues until all but one player is eliminated , with that player being declared the winner of the tournament .
the match format is specified by the set of positive integers @xmath9 with the interpretation that there are a total of @xmath10 rounds , with round @xmath2 consisting of @xmath11 matches , @xmath12 because @xmath13 players have been eliminated by the end of round @xmath14 , we must have that @xmath15 we suppose that the constitution of the matches in a round is totally random . that is , for instance ,
the @xmath16 players that play in round @xmath5 are randomly chosen from all @xmath17 players and then randomly arranged into @xmath18 match pairs .
the winners of these @xmath18 matches , along with the @xmath19 players that did not play a match in round @xmath5 then move to round @xmath20 , and so on .
we suppose that the players have respective values @xmath1 , and that a match involving players @xmath2 and @xmath3 is won by player @xmath2 with probability @xmath21 we let @xmath22 be the probability that player @xmath2 wins the tournament , @xmath23 in section 3 we derive a lower bound on the probability that the strongest player ( e.g. , the one with the largest @xmath24 ) wins the tournament , and an upper bound on the probability that the weakest player wins the tournament . in section 4 we derive upper and lower bounds on the win probabilities @xmath25 and also show that if @xmath26 then @xmath27 in section 5 , we consider the special case where @xmath28 and show that when @xmath29 the best format for the strongest player is the so - called _ balanced format _ that has @xmath30 matches in the first round and then has all remaining players competing in each subsequent round .
we also show that whenever the number of remaining players , say t , is even there is an optimal ( from the point of view of the best remaining player ) format that calls for @xmath31 matches in the next round .
we also show , for the sections special case , that the worst format for the best player is to have exactly one match each round .
analogous results for the worst player are also shown .
although we do not have a proof we conjecture that , among all possible formats , the balanced format maximizes and the one match per round minimizes the best players probability of winning the tournament even in the case of general @xmath32 ( we show by a counterexample that the format that calls for @xmath31 matches when an even number @xmath33 of players remain is not optimal for the best player in the general case . )
most papers in the literature on random knockout tournaments consider more structured formats than the ones we are considering which suppose that the number of matches in each round is fixed and that the game participants in each round are randomly chosen from those that remain . in these papers
the structure of the tournament is fixed and players are randomly assigned to positions of the structure .
for example , a structure with 6 players having a balanced format ( meaning 2 matches in round 1 , 2 matches in round 2 and 1 match in round 3 ) might in round 1 have a match between those in positions 1 and 2 and one between those in positions 3 and 4 , with the winners then playing each other in round 2 and 5 playing 6 in round 2 .
another possibility would be to have the winners of round 1 each play in round 2 one of those that did not play in round 1 .
figures [ fig_bal_str1 ] and [ fig_bal_str2 ] indicate these two structures .
.5 .5 an example of a non - balanced structure with 6 players is one where there is a single match in each round , with the winner of a match playing in the following round someone who has not yet played .
maurer @xcite proved for random structured formats , when @xmath34 , that the win probability of player @xmath5 is maximized under the balanced structures . because , as noted by maurer , a random format can be expressed as a mixture of random structured formats , this also establishes the result for random formats . in section 5
we give another proof of this result , which is both quite elementary and also shows , under the same condition , that when there are @xmath35 remaining players the optimal random format can schedule either @xmath30 or @xmath36 games if @xmath17 is even , and @xmath30 games if @xmath17 is odd .
we also prove , when @xmath34 , that the win probability of player @xmath5 is uniquely minimized by the random format that has exactly one match in each round .
chung and hwang @xcite proved that for any structured format , if @xmath37 is decreasing in @xmath2 then so is the winning probability , which also proves the result for any random unstructured format . in corollary 1 of section 4
we give a proof of this result that also proves , when @xmath37 is decreasing in @xmath38 that the probability player @xmath2 reaches round @xmath39 is decreasing in @xmath2 for all @xmath39 .
there are other papers studying knockout tournaments , for example , david @xcite , glenn @xcite , hennessy and glickman @xcite , horen and riezman @xcite , hwang @xcite , marchand @xcite .
the paper ross and ghamami @xcite discusses efficient ways of simulating win probabilities for such tournaments .
in this section we establish a lower bound on the probability that the strongest player wins the tournament , and an upper bound on the probability that the weakest player wins .
@xmath8 if @xmath40 is decreasing in @xmath2 , then @xmath41 @xmath8 preliminary to proving the theorem , we define the _ weight _ of a player as follows .
say that a player is alive if that player has not been eliminated .
if @xmath42 is the current set of alive players then , for @xmath43 , the weight of player @xmath2 is defined to equal @xmath44 if @xmath2 is no longer alive , its weight is defined to be @xmath45 @xmath8 the following lemma will be needed to prove the theorem .
+ suppose at the start of a round that players @xmath46 and @xmath47 are alive and have respective weights @xmath48 and @xmath49 , where @xmath50 if @xmath51 and @xmath52 are random variables denoting the respective weights of @xmath46 and @xmath47 after the round , then @xmath53 \geq x e[y ] .
$ ] * proof : * the result is immediate if @xmath54 so suppose that @xmath55 define a random variable @xmath56 whose value is determined by the results of the following round , as follows : * @xmath57 if @xmath46 and @xmath47 both lose * @xmath58 if @xmath46 and @xmath47 both advance * @xmath59 if @xmath46 and @xmath47 play each other * @xmath60 if one of the players @xmath46 or @xmath47 defeats @xmath2 whereas the other loses to @xmath3 * @xmath61 if one of the players @xmath46 or @xmath47 advances by not being selected to play whereas the other loses to @xmath3 by considering the possible values of @xmath56 , we now show that @xmath62 \geq x e[y|z ] .$ ] 1 . @xmath63 = e[y|z=0 ] = 0 $ ] 2 .
conditional on @xmath64 , let @xmath65 represent the sum of the weights of those aside from @xmath46 and @xmath47 that advance .
then , @xmath66 = e [ \frac{x}{x+y+r } ] , \quad e[y|z=1 ] = e [ \frac{y}{x+y+r}]\ ] ] showing that @xmath67 = x e[y|z=1].$ ] 3 .
conditional on @xmath68 , let @xmath65 represent the sum of the weights of those aside from @xmath46 and @xmath47 that advance ( and note that @xmath65 is independent of who wins the game between @xmath46 and @xmath47 ) .
then , @xmath69 = \frac{x}{x+y } e[\frac{x}{x+r } ] , \quad e[y|z=2 ] = \frac{y}{x+y } e[\frac{y}{y+r}]\ ] ] hence , @xmath70 - x e[y|z=2 ] = \frac{xy}{x+y } e[\frac{x}{x+r } - \frac{y}{y+r } ] \geq 0\ ] ] where the inequality follows because @xmath71 implies that @xmath72 for any @xmath73 4 . with @xmath74 and @xmath24 representing , respectively , the weights of @xmath2 and @xmath3 before the round , we have @xmath75 where @xmath76
let @xmath65 represent the sum of the weights of those , aside from @xmath77 and @xmath3 , that advance , and note that @xmath65 is independent of which of @xmath46 or @xmath47 loses .
then @xmath78 = c(\frac{x}{x+w } ) ( \frac{v}{v+y } ) e [ \frac{x}{r+x+ v } ] \ ] ] and , similarly , + @xmath79 = c(\frac{y}{y+w } ) ( \frac{v}{v+x } ) e [ \frac{y}{r+y+ v } ] \ ] ] hence , letting @xmath80 - x e[y|z= ( i , j ) ] $ ] , we see that + @xmath81\ ] ] as it is easy to check that for every @xmath82 and @xmath71 @xmath83 it follows that @xmath84 - x e[y|z= ( i , j ) ] \geq 0 $ ] for @xmath50 5 . that @xmath85 - x e[y|z= ( 0,j ) ] \geq 0 $ ] follows from the preceding result by setting @xmath86 hence , we have shown that @xmath87 \geq x e[y|z],$ ] and the result follows upon taking expectations of both sides of this inequality .
@xmath88 @xmath8 * proof of theorem 1 : * we give the proof that @xmath89 the proof that @xmath90 is similar .
let @xmath91 be the weight of player @xmath3 after @xmath30 rounds have been played .
also , let @xmath92 be the history of all results through the first @xmath30 rounds .
we claim that @xmath93 \geq w_1(k)\ ] ] because the claim is true when @xmath94 assume that @xmath95 let @xmath96 denote the set of alive players after round @xmath10 .
now , from lemma 1 , we have , for @xmath97 that @xmath98 \geq w_1(k ) e[w_j(k+1)| h_k]\ ] ] hence , if @xmath99 < w_1(k)$ ] , then for any @xmath100 @xmath101\ ] ] which is a contradiction since @xmath102 .$ ] hence , @xmath99 \geq w_1(k)$ ] , and taking expectations of both sides gives @xmath103 \geq e[w_1(k ) ] , \ ; r \geq 0.\ ] ] if the tournament has @xmath10 rounds , then the preceding yields that @xmath104 \geq e[w_1(0)]$ ] , which gives the result since @xmath104 = p_1 $ ] whereas @xmath105 @xmath8 _ remark : _ the preceding argument shows that @xmath106 is a submartingale , and that @xmath107 is a supermartingale . @xmath8
let @xmath108 be the probability that @xmath2 would win a match against a randomly chosen opponent .
in this section we prove that @xmath109 is smaller than it would be if it were the case that @xmath2 would win each game it plays with probability @xmath110 that is , we will prove the following . if the tournament format is @xmath111 then @xmath112 where @xmath113 is the number of players that advance to round @xmath39 . to prove the preceding theorem we will need a couple of lemmas . before giving these lemmas ,
we introduce the following notation .
we let @xmath114 be the event that player @xmath2 reaches round @xmath115 , and let @xmath116 be the event that @xmath2 wins the tournament .
if a player receives a bye in a round ( that is , if it reaches that round but is not chosen to play a match ) say that it plays player @xmath117 in that round .
also , let @xmath118 be the probability that @xmath2 beats @xmath3 in a game .
lemma 2 is easily proven by a coupling argument .
[ strdeclem ] for all @xmath119 is an increasing function of @xmath120 ( when @xmath121 , this states that @xmath122 is an increasing function of @xmath32 ) assume @xmath123 . for fixed @xmath2
, @xmath124 decreases in @xmath3 for @xmath125 .
* proof : * fix @xmath126 , where @xmath127 and @xmath128 the proof is by induction on @xmath17 . for @xmath129 ,
the only format is to play one game each round .
the result holds in this case because @xmath130 implies that @xmath131 assume the results holds for up to @xmath132 players and for all formats .
we now consider the @xmath17 player case .
let @xmath133 be an arbitrary format .
define a random vector @xmath56 whose value is determined by the results of the first round of the tournament under @xmath133 , as follows : * @xmath134 if @xmath2 plays against @xmath3 and @xmath30 plays against @xmath135 , or @xmath2 plays against @xmath30 and @xmath3 plays against @xmath135 ; * @xmath136 if @xmath2 plays against @xmath137 and @xmath3 plays against @xmath30 ; * @xmath138 if @xmath2 plays against @xmath135 , @xmath3 beats @xmath139 and @xmath30 beats @xmath74 ; * @xmath140 if @xmath2 plays against @xmath135 , @xmath3 loses to @xmath139 and @xmath30 loses to @xmath74 ; * @xmath141 if either ( a ) : @xmath2 plays against @xmath135 , @xmath3 beats @xmath139 and @xmath30 loses to @xmath74 , or ( b ) : @xmath2 plays against @xmath135 , @xmath30 beats @xmath139 and @xmath3 loses to @xmath74 . by considering the possible values of z , we now show that @xmath142 . 1 .
let @xmath143 .
because @xmath144 and @xmath145 , it follows that @xmath146 let @xmath147 denote the set of players other than @xmath148 that reach round 2 .
given that @xmath2 and @xmath3 have reached round 2 , the probability of @xmath2 and @xmath3 reaching round @xmath149 equals to the probability that @xmath2 and @xmath3 reaches round @xmath150 in a new tournament which begins with players @xmath2 , @xmath3 and @xmath147 , and follows the same format as @xmath133 after round 1 .
let @xmath151 denote the probability that @xmath2 and @xmath3 reaches round @xmath150 in the new tournament . similarly , let @xmath152 denote the probability that @xmath2 and @xmath30 reaches round @xmath150 in a tournament which begins with players @xmath2 , @xmath30 and @xmath147 , and follows the same format as @xmath133 after round 1
. then by lemma [ strdeclem ] , @xmath153 .
therefore for @xmath154 , @xmath155 \\ & \geq & p(r_{i,2}r_{k,2}|z = ( 1,u))e[p_{i , k , t}(s-1 ) ] \\ & = & p(r_{i , s}r_{k , s}|z = ( 1,u)).\end{aligned}\ ] ] where the preceding used that @xmath147 is independent of the event @xmath156 and of the event @xmath157 2 . since @xmath158
, we have that @xmath159 let @xmath147 denote the set of players other than @xmath160 and @xmath30 that reaches round 2 . with the same definition and argument as when @xmath161 , it can be shown that for @xmath149 , @xmath162 which implies that @xmath163 3 . in this case
, @xmath164 now , given that @xmath160 and @xmath30 have reached round 2 , the induction hypothesis implies that the probability of @xmath2 and @xmath3 reaching round @xmath39 is greater than or equal to the probability of @xmath2 and @xmath30 reaching round @xmath39 for @xmath149 .
therefore for @xmath149 , @xmath165 4 . in this case ,
@xmath166 . 5 .
in this case , @xmath167 @xmath168 where @xmath169 . + since @xmath170 and @xmath171 , we have that @xmath172 let @xmath147 denote the set of players other than @xmath160 and @xmath30 that reach round 2 , and note that @xmath147 has the same distribution whether ( a ) or ( b ) resulted . with the same definition and argument as when
@xmath161 , it can be shown that for @xmath149 , @xmath162 which implies that @xmath173 hence , we have shown that @xmath142 , and the result follows upon taking expectations of both sides of this inequality .
@xmath88 + + * proof of theorem 2 : * assume that @xmath174 is decreasing in @xmath3 .
now , given that @xmath2 reaches round @xmath39 , the conditional probability that @xmath3 also reaches round @xmath39 is @xmath175 hence , from lemma 3 it follows that @xmath176 is a decreasing function of @xmath177 .
now , @xmath178 let @xmath179 be the event that @xmath2 competes in round @xmath39 ( that is , it is the event that @xmath2 reaches round @xmath39 and then plays a match in that round ) . with @xmath180
we have that @xmath181 where @xmath182 hence , @xmath183 is a decreasing function of @xmath3 . letting @xmath52
be a random variable such that @xmath184 , it thus follows that @xmath52 is stochastically smaller than the random variable @xmath51 having @xmath185 therefore , @xmath186 \\ & \leq & 1 - \frac{2m_s}{r_s } + \frac{2m_s}{r_s } e[\frac{v_i}{v_i + v_x } ] \\ & = & 1 - \frac{2m_s}{r_s } + \frac{2m_s}{r_s } \frac{1}{n-1 } \sum_{j \neq i } \frac{v_i}{v_i + v_j } \end{aligned}\ ] ] where the preceding used that @xmath187 is an increasing function of @xmath3 and @xmath51 is stochastically larger than @xmath52 to conclude that @xmath188 \leq e[\frac{v_i}{v_i + v_x } ] .
\qquad \rule{2mm}{2mm}$ ] @xmath8 _ remark : _ it follows from lemma 2 that for all random formats , @xmath189 is an increasing function of @xmath37 . on the other hand , it seems intuitive that @xmath109 is a decreasing function of @xmath174 for @xmath190 .
however , while this is true for @xmath191 , it is not true for @xmath192 .
the argument , when @xmath129 , uses that @xmath193 this gives @xmath194 for @xmath195 a counter example can be constructed as follows .
consider the balanced format and let @xmath196 denote the probability that @xmath2 wins the tournament when player @xmath3 has value @xmath197 . conditioning on
whether or not player 1 first plays against player 4 , we have @xmath198 and thus @xmath199 @xmath8 using lemma 3 , it is easy to show that if @xmath37 is decreasing in @xmath2 then so is @xmath122 . + if @xmath200 then @xmath201 .
* proof : * suppose @xmath202 . because @xmath203 is the number of players that are still alive at the start of round @xmath39 , it follows that given @xmath2 reaches round @xmath39 , the expected number of others that also reach round @xmath39 is @xmath204 consequently , @xmath205 hence , @xmath206 @xmath207 which by lemma 3 shows the desired result for @xmath208 .
in addition , we have , @xmath209 @xmath210 which by lemma 3 , and by the assumption that @xmath211 , completes the proof
. @xmath88 + + we now give lower bounds on the win probabilities .
suppose @xmath212 let @xmath213 be independent with @xmath214 and let @xmath215 .
then @xmath216 * proof : * define the surrogate of @xmath2 as follows : the initial surrogate of @xmath2 is @xmath2 itself ; anyone who beats a surrogate of @xmath2 becomes the current surrogate of @xmath2 .
note that at any time there is exactly one player who is the current surrogate of @xmath2 .
if we let @xmath217 be the indicator of whether the surrogate of @xmath2 plays a match in round @xmath3 , it follows that @xmath213 are independent with @xmath218 also , let @xmath215 be be the number of games played by surrogates of @xmath2 ( while they are the current surrogate ) , and let @xmath219 be their set of opponents in these games . because @xmath220 it follows that @xmath221 \\ & \geq & e [ \prod_{k=1}^{\min(n , i-1 ) } \frac{v_i}{v_i + v_k } \prod_{k = i+1}^{n+1 } \frac{v_i}{v_i + v_k } ] \end{aligned}\ ] ] where @xmath222 is equal to @xmath5 when @xmath223 hence , @xmath224 _ remarks : _ 1 .
the probability mass function of @xmath225 , which does not depend on @xmath2 , is easily obtained by solving recursive equations ( see example 3.24 of ross @xcite ) . in the special case
where @xmath226 and where there are @xmath30 matches in round @xmath5 and afterwards all remaining players have matches in each subsequent round , we have @xmath227 2 . a weaker bound than the one provided in
is obtained by noting that , for @xmath228 , @xmath229 and so for @xmath228 @xmath230 \\ & = & e [ ( \frac{v_i}{v_i+ v_1 } ) ^{x_1 + \ldots x_r } ] \\ & = & \prod_{j=1}^r e [ ( \frac{v_i}{v_i+ v_1 } ) ^{x_j } ] \\ & = & \prod_{j=1}^r ( \frac{2m_j}{r_j } \frac{v_i}{v_i+ v_1 } + 1- \frac{2m_j}{r_j } ) \end{aligned}\ ] ]
suppose @xmath231 if @xmath232 then the balanced format that schedules @xmath30 matches in round @xmath5 and then has all remaining players competing in each subsequent round leads to the maximal possible value of @xmath233 * proof : * let @xmath234 be the probability that player @xmath5 wins in a match against another player .
the proof is by induction .
as there is nothing to prove when @xmath235 , let @xmath236 and suppose that the result is true for all smaller values of @xmath17 .
consider a format that calls for @xmath3 matches in the first round , where @xmath237 then , by the induction hypothesis , the format of this type that is best for player @xmath5 will call for @xmath238 matches in the second round .
the probability that player @xmath5 is among the final @xmath239 players under these conditions is @xmath240 [ \frac{2(k - j ) } { n - j } \,p + 1 - \frac{2(k - j)}{n - j } ] \ ] ] however , the probability that player @xmath5 is among the final @xmath239 players if the first round has @xmath30 matches is @xmath241 now , @xmath242 when @xmath243 is either @xmath244 or @xmath5 . because @xmath245 is easily seen to be strictly convex this implies that @xmath246 when @xmath247 thus , in searching for the best format for player @xmath5 we need not consider any format that calls for less than @xmath30 matches in round @xmath5 .
now consider any format that calls for @xmath248 matches in round @xmath5 where @xmath249 . by the induction hypothesis
it will call for @xmath250 matches in round @xmath20 .
this will result in player @xmath5 being among the final @xmath251 players with probability @xmath252 [ \frac { 2^s - 2i } { 2^s - i } \,p + 1 - \frac { 2^s - 2 i } { 2^s - i } ] \ ] ] however , the format that calls for @xmath30 matches in round @xmath5 and @xmath251 matches in round @xmath20 leads to player @xmath5 being among the final @xmath251 players with probability @xmath253 now , @xmath254 when @xmath243 is either @xmath244 or @xmath5 . because @xmath255 it follows that @xmath256 is equivalent to @xmath257 which is easily seen to be equivalent to @xmath258 which holds because having @xmath248 two person matches implies that @xmath259 hence , @xmath260 is convex , which shows that there is an optimal format that initially has @xmath30 matches in round 1 .
the induction hypothesis then proves the result .
@xmath88 @xmath8 _ remarks _ + 1 .
because @xmath260 is strictly convex unless @xmath261 , it follows that the format specified in theorem 2 is uniquely optimal ( in the sense of maximizing @xmath262 ) when @xmath17 is odd , whereas when @xmath17 is even there is also an optimal format that schedules @xmath36 matches in round 1 .
+ + 2 . in the case of general @xmath37 ,
the format that calls for @xmath36 matches when an even number @xmath17 of players remain is not optimal for the best player . for a counter example , consider a knockout tournament with 6 players having values @xmath263 . under the balanced format ,
the probability of player 1 winning the tournament is @xmath264 . under a format that plays 3 games in round 1 and then one game in each round ,
the probability of player 1 winning the tournament is @xmath265 . @xmath8
* proof : * the proof is by induction .
suppose it is true for all tournaments with fewer than @xmath17 players and now suppose there are @xmath17 players .
consider any format that calls for @xmath39 matches in the first round , where @xmath266 .
the probability that player @xmath5 is still alive when there are only @xmath267 alive players is @xmath268 on the other hand , the probability that player @xmath5 is still alive when there are only @xmath267 alive players when the format is one match per round is @xmath269 because @xmath270 is a polynomial whose coefficients are all positive it follows that @xmath271 as @xmath272 this implies that @xmath273 is strictly concave , which since @xmath274 when @xmath275 or @xmath276 , enables us to conclude that @xmath277 when @xmath247 hence , by the induction hypothesis , the format that calls for one match in each round results in a win probability for player 1 that is strictly smaller than what it is under any format that calls for @xmath278 matches in round 1 .
@xmath88 @xmath8 suppose @xmath279 the format that results in the highest win probability for player @xmath17 is the one that has one match in each round . if @xmath280 then the format that minimizes @xmath281 is the one that schedules @xmath30 matches in round @xmath5 and then has all remaining players competing in each subsequent round .
( i ) : : if @xmath285 , then @xmath286 is maximized by the balanced format , and its value is @xmath287 .
( ii ) : : if @xmath288 , then @xmath289 for all eligible formats .
( iii ) : : if @xmath290 , @xmath291 is maximized by the format where at each round there is exactly one match , and its value is @xmath292 .
* note that for a tournament where the players have respective values @xmath293 , @xmath294 .
now , ( i ) and ( iii ) are direct corollaries of theorems 2 and 4 respectively , while ( ii ) is true because if @xmath295 then all players have equal probability of winning the tournament regardless of the format . | we consider a random knockout tournament among players @xmath0 , in which each match involves two players .
the match format is specified by the number of matches played in each round , where the constitution of the matches in a round is random .
supposing that there are numbers @xmath1 such that a match between @xmath2 and @xmath3 will be won by @xmath2 with probability @xmath4 , we obtain a lower bound on the tournament win probability for the best player , as well as upper and lower bounds for all the players .
we also obtain additional bounds by considering the best and worst formats for player @xmath5 in the special case @xmath6 * random knockout tournaments * * ilan adler * + department of industrial engineering and operations research + university of california , berkeley + @xmath7 + * yang cao * + department of industrial and systems engineering + university of southern california + @xmath7 + * richard karp * + computer science department + university of california , berkeley + @xmath7 + * erol pekoz * + school of business + boston university + @xmath7 + * sheldon m. ross * + department of industrial and systems engineering @xmath7 + university of southern california @xmath8 [email protected] + [email protected] + [email protected] + [email protected] + [email protected] @xmath8 |
the philosophy derived from the principle o__f _ _ relativity , according to which the laws o__f _ _ physical phenomena must be the same for a `` stationary '' inertial observer as for one that is in uniform relative motion with the `` stationary '' inertial observer , states that there exists no means by which any inertial observer can determine whether or not they are in motion .
this philosophy introduces some uncomfortable inconsistencies that have made some critics of the str to spend a considerable amount of their time ( such as professor herbert dingle who spent about thirty years , see e.g. mccausland 2008 ) arguing that these inconsistencies rendered the str obsolete .
the str has never failed any experimental test to which has been subjected to , and this has lead to the mainstream scientific community to ignore any such criticism .
this philosophy that there exists no means mechanical or optical by which any inertial observer can determine whether or not they are in motion rests its weight on the michelson - mosley experiment ( mm - exp ) ( michelson 1881 , 1887 ) .
the mm - exp is an experiment that was designed to measure the speed of the earth in the hypothetical aether .
this aether was thought to exist since james clerk maxwell had shown that light was a wave and this light wave travelled at a constant speed denoted by the symbol @xmath1 . since a wave needs a medium which to travel in ; it was supposed that the aether filled the whole universe and was an absolute stationary frame of reference which was rigid to electromagnetic waves but completely permeable to ponderable matter .
the mm - exp was then designed and conducted .
much to suprise of the then present scientific community , the experiment showed no proof of the existence aether or lack thereof . without the knowledge of the mm - exp , albert einstein reasoned that it was not necessary to invoke this hypothetical medium .
he reasoned , that , if the laws o__f _ _ physical phenomena where to be the same for all inertial observers , and speed of light where an absolute constant as predicted by maxwell s theory , then the speed of light ought to be a universal and absolute constant for every observer in the universe everywhere and everytime .
we argue here in favor of the aether and we here cajole the reader to not stop but proceed . normally , well seasoned physics readers could nt give much time to a reading that tries to revive the aether hypothesis hence the cajoling effort to draw the reader .
as aforesaid , this reading seeks to revive the aether hypothesis
we call upon the dead to arise and shine .
we re - examine closely the long held underlaying philosophy of the str emanating from the principle o__f _ _ relativity via the twin paradox gedanken .
first we give an exposition of the well known twin paradox whereafter a modified version of is propounded .
this modified version is unlike the the original version , symmetric in every respect .
the symmetric nature of the new version , brings about an inconsistency that the str is unable to solve . even if the general theory o__f _ _ relativity ( gtr ) where to be brought to the str s rescue , this inconsistency is insoluble unless we critically and carefully revise the underpinning philosophy emanating from the principle o__f _ _ relativity namely that : _ * `` no inertial observer can determine their state of motion by any means be it via mechanical means , optical means or otherwise ; it is simple impossible . ''
* _ we persuade and ask the reader to approach this reading with an open mind - that is , devoid of the now seemingly entrenched philosophy that einstein s str trashed once and forever the aether to the `` science museum _ of _ great _ but _ failed ideas '' .
the idea of the possibility that absolute motion may be real has support from other some authors see for example cahill ( 2007 , 2008 ) .
cahill approaches this topic from a different vantage point known as process physics which seems to have some experiment basis .
the detection of absolute motion directly leads to the aether hypothesis and the works cahill ( 2007 , 2008 ) surely point to this .
when it comes to the str , a natural source of confusion for those encountering the str for the first in their endeavor to comprehend the time - dilation effect is summed up in the so - called twin paradox , which is not really a paradox in the true sense of a paradox .
this so - called paradox goes as follows : suppose we have a set of twins , taurai and taurwi , and taurwi decides to celebrate his @xmath2 birthday in style by rocketing at a constant relativistic speed ( i.e. speeds comparable to the speed of light , for which the effects predicted by the str become important and easily measurable ) to the nearest star to planet earth
alpha - centauri which is @xmath3 lightyears away and this is incordance with a `` stationery '' earth bound observer . taurai and taurwi are recent physics graduates who understand very well albert einstein s ( 1905 ) str .
taurwi makes a round - trip , that is , he travels to alpha - centauri at a constant speed and upon arriving he returns to mother earth .
the other twin taurai , decides to stay at home and not join his adventurous twin brother . according the str ,
taurai sees taurwi as moving away from the earth and at the sametime , taurwi has equal claim in his own frame of reference that he is not moving but tauri is , that is he [ tauri ] is moving away from him at the same speed as that taurai sees him move albeit in the opposite direction .
the paradox arises because according to the str , the one that is moving will experience time dilation so the question is ; since each sees the other as moving , then , who amongst them will experience this time dilation ? and
thus seem younger to the other .
this apparent paradox arises from an incorrect application of the principle o__f _ _ relativity to the description of the story by the traveling twin s point of view and the widely accepted resolution of this apparent paradox goes as follows . from his [ taurwi ] point of view , the argument goes ; his non - adventurous stay - at - home brother is the one who travels backward on the receding earth , and then returns as the earth approaches the spaceship again , while in the frame of reference fixed to the spaceship , the astronaut twin is not moving at all .
it would then seem that the twin on earth is the one whose biological clock should tick more slowly , not the one on the spaceship .
the `` flaw '' in the reasoning is that the principle o__f _ _ relativity only applies to frames that are in motion at constant velocity relative to one another , i.e. , inertial frames of reference .
the astronaut twin s frame of reference , is a noninertial system , because his spaceship must accelerate when it leaves , decelerate when it reaches its destination , and then repeat the whole process again on the way back home .
their experiences are not equivalent , because the astronaut twin feels accelerations and decelerations thus leading to the conclusion that the traveling twin will be younger when they are reunited .
given this solution and that this typically presented in books that deal with the str at length , it is suprising that some notable authors ( see e.g. kark 2007 ) still regard the twin paradox as a paradox .
the real trick is the accelerations and decelerations experienced by the traveling twin ; these bring about the asymmetry which leads to taurwi being the one that experiences the time dilation . from the purely idealized point of view
, we can neglected these accelerations and decelerations .
if we did this will the scenario be symmetric ?
since it is these accelerations and decelerations that bring in the asymmetry , it must follow that we must have a paradox because symmetry ought in this case to be restored thus leading to a real paradox .
we wish to point - out here that our treatment of the twin paradox since it was conceived has been erroneous because the idealized twin paradox without the accelerations and decelerations , is not symmetric at all .
why do we say this ? if two people where to give a succinct description of their experiences and these experiences where truly symmetric , one would naturally not be able to differentiate the difference in their statements because their experiences would appear exactly the same as well as their world - view via - - vis the description of their experience .
for example , taking the most condensed and succinct description of their experience and the swapping or interchanged some key words in their statements , the resultant statement ought to be the same as the description by the other twin .
this is not the case with the present scenario . *
:* he is stationery and taurwi is moving toward alpha - centauri and alpha - centauri is not moving .
* :* he is stationery while both the taurai and alpha - centauri are moving as a whole unit like a rigid body .
( taurai and alpha - centauri are stationery relative to each other . )
the description of events by the taurai and taurwi is not the same hence not symmetric . in order to understand what we mean here by : _ * `` the description of events by each of the observers must be the same or symmetric . ''
* _ the reader will have to wait until the end of the next section .
the asymmetry seen in the description of events here is all one needs in order to come to the conclusion that the astronaut twin taurwi , is older when the twins reunite .
what we need if we are to have a real paradox is to bring about a perfect symmetry into the whole situation .
we shall set forth a new version of the twin paradox which is truly symmetric and this will introduce a true paradox and we shall provide a solution .
suppose taurai unlike in the previous version , decided to be adventurous too .
he decides to rocket into space and travels not with his twin brother but all by himself and instead of alpha - centauri he travels at the same constant relativistic speed as taurwi [ this speed is measured by the earth bound observers ] to an imaginary constellation ( call it constellation alpha - christina ) which is equidistant and directly opposite to alpha centauri along the line of sight joining the earth and alpha centauri .
= 9.8 cm on their day of departure , their family and friends bid them farewell and wish them a safe travel . without much say , on the day of reunion ,
the family and friends [ who all have studied physics at university and understand very well the str ] have no doubt that they [ the twins ] will all have aged the same .
the big question is , will the twins agree with their family and friends that they have aged the same ?
the truth is that , each of the twins will see the other as having aged less than they so they would not agree with their family and friends that they must be the same age .
herein we have a paradox ! who is older than who here ?
if @xmath4 is the speed with which the earth bound observers ( family and friends ) see the twins travel at , then , according to the twins in their own respective frames of references , the earth is receding at a speed @xmath4 and the other twin is receding from them at a speed @xmath5 .
this scenario is perfectly symmetric and each of the twins has every right according to the str to say the other twin is the one that is younger and they will not agree that their ages are equal upon reuniting .
we are here presented with a true paradox which the str in its presently understood form [ as is found in most if not all the textbooks of physics that deal with the subject of the str ] , is unable to provide an answer .
it will not help to call the gtr to our rescue because what we shall do with it for one twin , we shall have to do exactly the same with the other twin we have a catch here .
the situation exhibits a perfect symmetry that runders us imobile as along as we stick with the provinces of einstein s philosophy _ of _ relativity .
now , what we meant in the previous section by `` the description of event by both observers must be the same if their experience are symmetric '' is as follows : * :* he is stationery and taurwi is receding from him at a speed @xmath5 and the earth is receding from him at a speed @xmath4 .
alpha - centauri is receding at a speed @xmath4 while alpha - christina is approaching him at a speed @xmath4 . * :* he is stationery and taurai is receding from him at a speed @xmath5 and the earth is receding from him at a speed @xmath4 .
alpha - christina is receding at a speed @xmath4 while alpha - centauri is approaching him at a speed @xmath4 .
the above description is congruent .
we just have to swap the alpha - christina with alpha - centauri and taurai with taurwi , that is where there is alpha - centauri @xmath6 alpha - christina where there is alpha - christina we make the replacement alpha - christina @xmath6 alpha - centauri and where there is taurai @xmath6 taurwi .
it is not possible to do the same in the case of the asymmetric twin paradox of the previous section . once again before leaving this section , we shall re - emphasis that unlike the asymmetric twin paradox , where one can seek refuge by invoking the gtr to deal with the accelerations of one of the twins , here , this clearly wo nt work since both twins will all undergo the same experience .
their ages will be less than that recorded by the earth observers and these observers will measure these ages to be exactly the same but according to the twins , their ages can not be the same , hence the dilemma !
how do we solve this ?
the solution to the symmetric twin paradox will require us to rethink the very nimbus of the str s central philosophy , namely that it is impossible for an inertial observer to detect their state of motion .
this revision , will not alter the mathematical content of the str , but will bring us back to the long obsoloted idea of the existence of the all pervading and permeating medium , the aether .
suppose we have an inertial observer @xmath7 stationed at point @xmath7 in a closed rectangular cabin oabcde as shown in figure 2 . the axis @xmath8 and @xmath9 are orthogonal and the corners of the cabin @xmath10 , @xmath11 , @xmath12 and @xmath13 are right angles . at point
a , observer @xmath7 places a photon emitter that emits a single photon at a time in the vertical direction parallel to eo and bc .
point d is vertically and directly above point a. since the point d is directly above point a and the cabin oabcde is an inertial system , according to our current understanding of inertial systems , it goes without saying that the photon emitted in the vertical direction at point a will reach point d since light travels in straight lines . at this point d ,
observer @xmath7 places a photon detector that is linked to the photon emitter at point a such that observer @xmath7 is able to determine the time taken by this photon to travel from point a to point d. if @xmath14 , the time of travel ( @xmath15 ) according to observer @xmath7 of the photon will be @xmath16 where @xmath1 is the speed of light .
so far so good and no problem .
lets proceed !
let us introduce another inertial observer @xmath17 stationed at point @xmath18 in a closed rectangular cabin @xmath19 as shown in figure 3 . as is the case with the @xmath20 axis , the axis @xmath21 and @xmath22 are orthogonal and the corners of the cabin @xmath23 , @xmath24 , @xmath25 and @xmath26 are right angles . at point @xmath27 ,
observer @xmath17 bores a large enough hole so much that for a photon entering via this hole , diffraction effects can be neglected and the photon can be treated as a particle .
point @xmath28 is vertically and directly above point @xmath27 .
the roof of the cabin @xmath29 is photo - sensitive .
let this cabin move along the positive @xmath30 at speed @xmath4 such that when the lines @xmath31 and @xmath32 are coincident , the photon realized at point @xmath33 by observer @xmath7 will be at the opening of the cabin @xmath19 at point @xmath27 .
so far every this looks good , lets proceed .
we have agreed that the photon can be treated here as a particle because the opening at point @xmath27 is large enough for us to neglect completely any diffraction effects .
this photon entering at this opening will have its direction of motion being parallel to the walls , @xmath34 & @xmath35 and @xmath36 & @xmath37 of the both cabins @xmath38 and @xmath19 respectively .
now our trouble begins !
= 8.0 cm [ a ] = 8.0 cm [ b ] = 8.0 cm [ c ] = 8.0 cm [ d ] since @xmath17 is a inertial observer and s / he has knowledge that the particle that just entered is a photon and the direction of motion of this photon is as aforedescribed .
the question is ; will s / he see the photon continue to travel parallel to the walls of her / his cabin ? if it does , then , s / he will expect at some finite time in the future that this photon will be detected at point @xmath28 .
if it so happens that at this point @xmath28 , we have an opening , the photon will travel outside the cabin of observer @xmath17 upon arriving at point @xmath28 and this photon will be detected on the roof of observer @xmath7 s cabin albeit off - set from point @xmath39 ( to the right - side of ) .
the reason the photon will be detected off - set the point @xmath39 is because at the time of exit of the photon at point @xmath28 , this point is no - longer directly above point @xmath39 because this cabin is moving relative to the cabin of observer @xmath7 and the photon will have to continue its journey in a straight line parallel to wall of both cabins .
let us re - state or rephrase what we have just said in the previous paragraph .
if the photon travels the same path as that in figure ( @xmath40 ) , then , according to observer @xmath17 , its path will be inclined at an angle @xmath41 to her / his walls and this photon will traverse path @xmath42 and not @xmath31 as would expected for a photon traveling in the vertical direction from point @xmath27 in the cabin @xmath43 .
this angle @xmath41 is such that : @xmath44 if the photon traversed along a straight path according @xmath17 just as in the case figure ( @xmath40 ) , then the photon will have to exit the cabin @xmath43 at point @xmath28 which will be offset thus the photon will have to be detected by observer @xmath7 off - set from the point @xmath39 to the right - side .
if this is the case , observer @xmath7 is forced to draw the conclusion that the cabin @xmath43 at point @xmath28 affected the motion of the photon . at this point
we leave the reader to decide whether they think the photon will exhit from @xmath17 at point @xmath28 or not .
we believe the reader will reach the conclusion that the photon will not exit via point @xmath28 but would exit at point @xmath45 is a hole was bored there and that observer @xmath7 will then be able to detect the photon at point @xmath39 . accepting this ,
amounts to accepting that observer @xmath17 will be able to determine his speed in - accordance with equation ( [ v ] ) . from the foregoing
, the solution to the symmetric twin - paradox is clear .
taurai and taurwi can determine their state of motion and even measure their velocity which they will each find to be @xmath4 and this velocity is their velocity relative to some absolute and universal medium that is absolute rest and this medium clearly must be the one in which light has this constant speed @xmath1 . if the laws _ of _ nature are to be the same everywhere in space and time , then , it follows that this medium must fill all of space .
the length contraction and time dilation occur relative to this medium and these properties are exactly as those of the lorentz aether .
we are thus are brought back to the old ideas that now `` safely '' belongs to the science museum _ of _ great _ but _ failed ideas .
the idea of a universal all - pervading and permeating medium , is a superseded idea .
einstein s 1905 str rendered it obsolete and ever since then , research on this idea is not taken seriously hence the reason for the lack of citation of recent reseach on this field .
the more than century old mm - exp is said to be enough proof against this idea and it is said / thought / and or supposed , that this experiment alone closed down the curtain once and for all on this subject . if the arguments presented in this reading are correct , then , we are called back to the drawing board to rethink our long held belief that a universal all - pervading and permeating medium , is a superseded idea .
this believe stems from the fact the str proclaimed that it is impossible for an inertial observer to detect their state of motion .
we have shown here that not only is an observer able to determine where or not they are moving , but that they are able to deduce their velocity .
this velocity will have to measured relative to the same `` medium '' in which light is the speed of light has this same speed @xmath1 .
we realize , that if our arguments are correct , then , not only is the speed of light the same for all inertial observers , but the direction of motion of this light .
this would mean we have to re - write the second postulate of the str which in most physics texts book reads ( see e.g cutnell & johnson 2003 ; halliday & resnick walker 1997 ) : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ the speed of light in vacuum has the same value @xmath1 in all directions and in all inertial reference frames " _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ to read : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ the velocity of light in a gravity free vacuum is the same for all observers . " _
+ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ the term `` velocity '' is different from `` speed '' as this term [ velocity ] includes the speed and the direction of propagation of the beam of light .
what this means is that all inertial observers will agree not just on the speed but on the direction of propagation of the beam . in the case
as presented in figures ( 2)-(5 ) , the photon will not change its direction of motion relative to observer @xmath7 thus the meaning of will be that , @xmath17 will see the photon traverse at an inclined angle @xmath41 to her / his walls .
we have already argued , this angle is enough to deduce the speed of the cabin . once again , if these ideas are correct , what this really means is that time dilation and length contraction are real physical phenomena in much the same way as lorentz ( 1892 , 1904 ) and fitz - gerald ( 1889 ) envisaged and the aether is also real in the sense envisaged by maxwell ( 1973 ) ( and many advocated it the aether theory ) when he propounded his electromagnetic theory which einstein mused until he arrived at the str .
the shift @xmath46 measured by observer @xmath17 in his / her cabin as shown in figure ( 5 ) is given : @xmath47 where @xmath48 is the height of the cabin .
this can be generalized to any given inertial observer .
thus if one is in an inertial frame of reference and they projected a light beam vertically up - wards and this light beam strikes the roof not on a point directly above the point when the beam of light was realized , the conclusion they have to make is that their cabin of system of reference is in motion and the shift is related to the speed of their frame of reference and the height of this system by equation ( [ shift ] ) . given that the gravitational pull between the earth and sun causes the earth to travel around , or orbit , the sun at a speed of @xmath49 this would mean a laboratory that is say @xmath50 in height will be expected to register a shift in accordance with equation ( [ shift ] ) , of about @xmath51 . given modern day precision ,
it should be possible to detect such a shift .
therefore , we propose that this experiment be carried out . should the results provide a negative result , the present ideas are immediately rendered null and void and einstein s philosophy about the obsoleteness of absolute motion is holds .
if the experiments prove this shift , then , nothing of the mathematical structure of the str will change , expect its philosophy and this philosophy will exactly be that championed by lorentz in his works ( lorentz 1892 , 1904 ) .
cahill t. r. , 2007 , _
dynamical 3-space : a review _ ; arxiv:0705.4146 + cahill t. r. , 2008 , _ unravelling lorentz covariance and the spacetime formalism_;arxiv:0807.1767 + cutnell j. & johnson k. 2003 , _ physics extended _ , _ john wiley & sons inc . _ : isbn 0 - 471 - 39219 - 7 , fithy edition , page 865 .
+ einstein a. , 1905 , _ zur elektrodynamik bewegter krper _ ,
_ annalen der physik
* 17 * , page 891 .
english translation : http://www.fourmilab.ch/etexts/einstein/specrel/www/ + fitz - gerald , g. f. , 1889 , _ the ether and the earth s atmosphere _ ,
_ science _ , vol . * 13 * , 390 .
+ halliday d. , resnick r. & wallker j. 1997 , _ fundamentals of physics extended _ , _ john wiley & sons inc . _ : isbn 0 - 471 - 105597 , fithy edition , page 990 .
+ kak1 s. , 2007 , _ moving observers in an isotropic universe _ , _ international journal of theoretical physics _
46 * , pages 1424 - 1430 . + lorentz h. a. , 1892 , _ the relative motion of the earth and the ether _ ,
. wetensch . * 1 * , 74 . + lorentz h. a. , ( 1904 ) , _ electromagnetic phenomena in a system moving with any velocity smaller than that of light _ , proc .
science amsterdam , vol .
* 6 * , pages , 809 - 831 .
+ maxwell j. c. , 1873 , _ oxford : clarendon press_. ( reprint of 3th ed .
, 1998 , oxford classic series ) , trease on electricity and magnetism , * 1 * , ix . + mccausland i. , 2008 , _ a question of relativity _ , _ apeiron journal _ ,
15 * , no 2 , pages 545 - 557 . :
+ michelson a. a. 1881 , _ the relative motion of the earth and the luminiferous aether . " _ , _ amer .
_ , * 22 * , 120 - 129 .
+ michelson a. a. & morley , e. w. , 1887 , _ on the relative motion of the earth and the luminiferous ether .
_ , _ amer .
_ , * 34 * , 333 - 345 ; also see _ philos . mag .
_ 24 , 449 - 463 , 1887 . | ( 1,0)400 + + @xmath0 the famous twin paradox of the special theory o__f _ _ relativity by einstein ( 1905 ) is revisited and revised .
this paradox is not a paradox in the true sense of a paradox but a reflection of a misunderstanding of the problem and the principle o__f _ _ relativity .
the currently accepted solution to this takes into account the accelerations and deceleration of the traveling twin thus introducing an asymmetry that solves the paradox .
we argue here that , with the acceleration and deceleration neglected , the problem is asymmetric hence leading to the same conclusion that the traveling twin will age less than the stay at home .
we introduce a symmetric twin paradox whose solution can not be found within the currently accepted provinces of the str if one adopts the currently accepted philosophy of the str namely that it is impossible for an inertial observer to determine their state of motion . to resolve this
, we present ( in our modest view ) a simple and convincing argument that leads us to conclude that it must be possible for an inertial observer to determine their own state of motion . with this
, we are able to solve the symmetric twin paradox .
the fact that it is possible for an inertial observer to determine their state of motion brings us back to the long rejected idea of an all pervading and permeating medium the aether , namely the lorentz luminiferous aether .
an experiment capable of validating or invalidating this claim is suggested .
+ + * keywords : * absolute motion , aether , asymmetry , symmetry , principle _ of _ relativity , relative motion . + ( 1,0)400 * jules henri poincar * ( 1854 - 1912 ) |
two non - orthogonal quantum states can not be discriminated with certainty , while the discrimination error can be made smaller as their copies are provided .
this leads to one of the well - known no - go theorems , that quantum states can not be copied with certainty @xcite , although approximate quantum cloning is possible with the use of quantum operations and ancillary quantum systems @xcite .
interestingly , the impossibility of perfect quantum cloning can be connected to the no - signalling principle in physics , which dictates that information can not be sent faster than light . as a consequence , quantum communication that makes use of ( non - local ) quantum correlations
can not be performed faster than light .
in fact , the relation between the no - cloning theorem and the no - signalling constraint has been established in both qualitative and quantitative terms .
for instance , it has been shown in ref .
@xcite that any no - signalling theory predicting non - locality , i.e. violation of bell inequalities , has a no - cloning theorem .
this implies that quantum theory necessarily has the no - cloning theorem .
in addition , the optimal fidelity of approximate quantum cloning has been derived by applying the no - signalling condition to the cloning process of quantum states @xcite , provided that corresponding quantum operations are positive @xcite . on the other hand , the impossibility of perfect state discrimination can also be connected to the no - signalling condition in a qualitative way , at least through the existing relation that the no - cloning theorem is implied by the no - signalling constraint .
now the question naturally arises whether there is a quantitative connection as well .
in fact , recent progress along this line has shown that the no - signalling condition would imply the optimality of state discrimination , constrained to those figures of merit such as minimum - error discrimination @xcite , unambiguous state discrimination @xcite , and maximum confidence measurement @xcite .
it is remarkable that the optimality of state discrimination in different figures of merit can be derived from the single condition , no - signalling constraint , despite the fact that quantum theory is not a maximally non - local theory @xcite . in this work ,
we relate the no - signalling condition to minimum - error state discrimination of ensembles of quantum states i.e. mixed states .
the novelty of this work is that two quantum states to be discriminated between are not purely quantum but mixtures of pure quantum states .
we shall derive the minimum error for discriminating between two ensembles of quantum states from the no - signalling constraint . the proof is built on a communication scenario between two parties , alice and bob .
before starting the main proof , for clarity we first fix notations to be used in the communication scenario and then briefly review some known facts about quantum states , minimum - error discrimination and ensemble decompositions .
we also restrict our consideration to qubit states , so one can make use of the representation in which a single qubit state is fully characterised by its bloch vector @xmath0 : @xmath1 where @xmath2 .
pure states have unit bloch vectors @xmath3 .
the communication scenario to be considered assumes that , sharing entangled states with bob , alice measures her qubits to prepare the ensemble decomposition of bob s state , either one of two ensembles , @xmath4 and @xmath5 , as follows , _
b^(0 ) & = & p_0 + ( 1-p)|| , + _
b^(1 ) & = & p_1 + ( 1-p)|--| , [ bobs ] where @xmath6 and @xmath7 are two qubit states for which we are to derive the discrimination bound using the no - signalling constraint , and @xmath8 are two orthogonal quantum states such that it holds @xmath9 , see fig [ fig ] .
it is convenient to express those quantum states in ( [ bobs ] ) in terms of bloch vectors as follows , _
b^(i ) = _ b^(i ) ( _ b^(i ) ) , _ i = _ i ( _ i ) , || = ( _ ) , [ parameter ] where @xmath10 .
then , the relation between quantum states in ( [ bobs ] ) can be translated to the following relations between bloch vectors _
b^(0 ) & = & p _ 0 + ( 1-p ) _ , + _
b^(1 ) & = & p _ 1 + ( 1-p ) _ - . [ vec ]
recall that @xmath11 and @xmath12 in ( [ bobs ] ) are two quantum states that we wish to discriminate between .
the minimum - error discrimination between two quantum states has been completely analyzed in ref .
@xcite optimizing over all possible measurement bases .
the minimum error , known as the helstrom bound , is p_e= - _ 0 - _
1,[hel ] assuming that the _ a priori _ probability that the quantum state @xmath13 occurs is equally @xmath14 . here
the trace norm for an operator @xmath15 , denoted by @xmath16 , is defined as @xmath17 $ ] . as is shown in ( [ hel ] ) ,
the optimal discrimination between two quantum states only depends on their ( trace - norm ) distance separation , and in particular it does not depend on the dimensions of the states . , and
here we restrict to two decompositions that can be seen in a cross - section of the bloch sphere spanned by @xmath18 and @xmath19 or @xmath20 . as is shown in ( [ vec ] )
, @xmath21 can be a mixture of @xmath22 and @xmath23 , or @xmath24 and @xmath25 , both of which point to the same point in the bloch sphere , meaning that they have the same ensemble average.,scaledwidth=32.0% ] we now review properties of _ ensembles _ of quantum states , known as mixed states .
these are classical mixtures of some pure quantum states .
for instance , each mixed state @xmath26 ( for @xmath10 ) in ( [ parameter ] ) can be expressed as a convex combination of some other pure states @xmath27 , i.e. _ i = _ j _ j |v_j^i v_j^i| .
[ pure ] note that such a decomposition is not unique @xcite .
this means that , in reality , once an ensemble of quantum states is given , its actual decomposition can not be known without knowing how the state was prepared .
one can only deduce ensemble - averaged properties of the state , which does not depend on any particular decomposition . now let us assume that two quantum states in ( [ bobs ] ) are the same . using the expression in ( [ vec ] ) , one can see that the vector of each ensemble of quantum states points to the same point in the bloch sphere , see fig .
that is , the two quantum states have different decompositions but have the same ensemble average .
in addition , each @xmath13 , @xmath28 , is again a mixture of pure quantum states , so finally quantum states of bob in @xmath29 can be expressed as an ensemble of pure quantum states .
recall that for any mixed state , there exists higher - dimensional pure quantum state such that partial reduction of the pure state would be the mixed state itself . in this case ,
two - dimensional ancillary systems suffice to purify those quantum states in ( [ bobs ] ) . for two different decompositions of the same state in ( [ bobs ] ) ,
the corresponding purifications are equivalent up to local unitary transformations .
this simply depends on choice of the measurement basis on the ancillary systems .
if we assume that alice holds the purification , her measurement will decide the decomposition of bob s quantum state .
however , unless alice announces what basis her measurement was made in , bob will never know which decomposition he has .
performing quantum state tomography will only allow bob to understand his quantum state as an ensemble average of other quantum states .
this is in fact what prevents the two parties communicating faster than light .
we are ready to prove that the no - signalling constraint implies that the helstrom bound in ( [ hel ] ) is optimal .
consider that alice and bob are separated in space so that local actions performed by one can not affect the other .
suppose that they share copies of entangled states of @xmath30 |_ab = _
j|a_j_a |b_j_b,[ab ] which is not written in orthonormal basis of alice and bob , so @xmath31 may exceed the rank of systems belonging to two parties . as we have mentioned above , alice chooses measurement basis so that after the measurement on her particle , the ensemble decomposition of bob s state is determined
. in other words , alice performs a measurement @xmath32 or @xmath33(which are general positive - operator - valued - measures ) , after which decomposition of bob s state will be either @xmath4 or @xmath5 in ( [ bobs ] ) respectively .
alice actually can make use of the measurement @xmath32 and @xmath33 as an encoding such that a value @xmath31 is encoded by applying @xmath34 .
bob then decodes the value by discriminating between two ensembles @xmath4 and @xmath5 .
two ensembles are the same , i.e. _
b^(0 ) = _ b^(1 ) and so their bloch vectors are also the same , i.e. @xmath35 from ( [ vec ] ) , and therefore the value @xmath36 in ( [ vec ] ) is uniquely determined , p=. [ p ] let us now consider a detector that bob applies , that works as follows .
once a qubit in a state @xmath37 being measured , the detector gives an outcome @xmath38 with some probability @xmath39 where @xmath10 , i.e. the detector can be thought of being in a black box from which we only know the input / output list .
therefore , it fulfills that @xmath40 note that we do not know all the properties of the detector , e.g. the internal structure , etc . , but the probabilities @xmath39 for @xmath10 . for two states @xmath6 and @xmath7
, we can assume that in general @xmath41 , by which it also follows that @xmath42 .
while alice not announcing which measurement is applied , bob can make guess of it by designing his detector to discriminate the two states @xmath43 of @xmath4 and @xmath44 of @xmath5 .
if the detector discriminates between @xmath6 and @xmath7 and gives an output @xmath38 ( either @xmath45 or @xmath46 ) , bob will then conclude that the ensemble is @xmath47 .
note that alice applies measurement @xmath48 or @xmath49 with equal probabilities , and then the error rate is @xmath50/2.\ ] ] two ensembles in ( [ bobs ] ) consist of not only @xmath26 but also @xmath51 .
although the detector being designed to discriminate between the two states @xmath43 and @xmath44 in the ensemble , the detector produces some outcomes for @xmath52 and @xmath53 with some probabilities @xmath54 and @xmath55 respectively .
assume first that alice applies the measurement @xmath32 so that @xmath56 is prepared on the bob s side .
bob is then with the state @xmath43 with probability @xmath36 , from which the value @xmath45 is found with probability @xmath57 .
bob is also with the state @xmath53 with probability @xmath58 , from which he can guess the value @xmath45 with probability @xmath59 , a non - negative number .
finally , the overall probability that detector s output says the value @xmath45 when alice prepares the ensemble @xmath4 , denoted by @xmath60 denote the probability that the detector says @xmath38 when alice prepares @xmath61 for @xmath10 and @xmath28 . ] and @xmath62 , is greater than or equal to @xmath63 , i.e. , @xmath64 we also have @xmath65 , the overall probability that detector s output says the value @xmath46 when the same ensemble @xmath4 is prepared .
clearly it holds that @xmath66 next , assume that alice applies the measurement @xmath33 so that @xmath5 is prepared . in the same way we have done for the ensemble @xmath4 , the overall probability that the detector says the value @xmath46 , denoted by @xmath67 , is greater than or equal to @xmath68 , i.e. , @xmath69 it holds again that @xmath70 however , if the following holds @xmath71 bob can discriminate the two ensembles since equations in ( [ b ] ) , ( [ d ] ) , and ( [ e ] ) tell that @xmath72 the above ( [ f ] ) means that , once @xmath34 is applied so that @xmath61 is prepared , the measurement data show more frequency on the value @xmath31 than @xmath73 , through which the two ensemble can be discriminated .
this implies that the no - signalling principle excludes ( [ e ] ) and therefore the following holds , @xmath74 combining ( [ a ] ) , ( [ c ] ) , and ( [ g ] ) , we have @xmath75 as well , from ( [ y]),([z ] ) , and ( [ h ] ) we obtain @xmath76 finally from ( [ p ] ) and ( [ i ] ) , the following is the discrimination error limited by the non - signalling principle @xmath77 which is already saturated by the helstrom bound in ( [ hel ] ) .
in conclusion , we relate two no - go theorems - the no - perfect state discrimination and the no - signalling - by showing that the optimal discrimination between two quantum states is derived from the no - signalling condition .
it can thus be said that the impossibility of perfect state discrimination is a consequence of the no - signalling condition , not only qualitatively , but also quantitatively as is the case with the no - cloning theorem @xcite .
furthermore , it was shown in ref .
@xcite that the fidelity of optimal quantum cloning converges asymptotically to the fidelity obtained through optimal state estimation .
therefore , the quantitative relationship between three no - go theorems- no - signalling , no - cloning , and no - perfect state estimation- has been established at the most basic level i.e. two non - orthogonal quantum states
. it would be interesting to investigate whether optimal values of all figures of merit in state discrimination , more generally state estimation , of quantum states could be derived solely from the no - signalling condition .
recently , it is shown that the optimal measurement for the maximum confidence is also implied by the no - signalling condition @xcite .
we thank j. hur , d. song , and h. nha for discussions . this work is supported by the it r@xmath78d program of mke / iita ( 2008-f-035 - 01 ) , the korean government ( moehrd ) ( krf-2005 - 003-c00047 ) , and the korea science and engineering foundation ( r01 - 2006 - 000 - 10354 - 0 ) . | in this work we relate the well - known no - go theorem that two non - orthogonal ( mixed ) quantum states can not be perfectly discriminated , to the general principle in physics , the no - signalling condition .
in fact , the minimum error in discrimination between two quantum states is derived from the no - signalling condition . |
optimization and ( feedback ) control of dynamical systems is often computationally infeasible for high dimensional plant models .
therefore , one tries to reduce the order of the system , so that the input - output mapping is still computable with sufficient accuracy , but at considerably smaller cost than for the original system , @xcite . to guarantee the desired accuracy , computable error bounds are required .
moreover , system properties which are relevant in the context of control system design like asymptotic stability need to be preserved .
it has long been known that for linear time - invariant ( lti ) systems the method of balanced truncation preserves asymptotic stability and provides an error bound for the @xmath1-induced input - output norm , that is the @xmath0-norm of the associated transfer function , see @xcite .
when considering model order reduction of more general system classes , it is natural to try to extend this approach .
this has been worked out for descriptor systems in @xcite , for time - varying systems in @xcite , for bilinear systems in @xcite and general nonlinear systems e.g. in @xcite .
yet another generaliztion of lti systems is obtained considering dynamics driven by noise processes .
this leads to the class of stochastic systems , which have been considered in a system theoretic context e.g. in @xcite . quite recently
, balanced truncation has also been described for linear stochastic systems of it type in @xcite .
already the formulation of the method leads to two different variants that are equivalent in the deterministic case , but not so for stochastic systems .
it is natural to ask which of the above mentioned properties of balanced truncation also hold for these variants . the aim of this paper is to answer this question .
let us first recapitulate balanced truncation for linear deterministic control systems of the form @xmath2 here @xmath3 , @xmath4 , @xmath5 , and @xmath6 , @xmath7 and @xmath8 are the state , output , and input of the system , respectively .
moreover @xmath9 denotes the spectrum of @xmath10 and @xmath11 the open left half complex plane .
let @xmath12 denote the lyapunov operator and @xmath13 its adjoint with respect to the frobenius inner product .
then @xmath14 if and only if there exists a positive definite solution @xmath15 of the lyapunov inequality @xmath16 , by lyapunov s classical stability theorem , see e.g. @xcite .
balanced truncation means truncating a balanced realization .
this realization is obtained by a state space transformation computed from the gramians @xmath17 and @xmath18 , which solve the dual pair of _ lyapunov equations _ [ eq : lyap ] @xmath19 or more generally the _ inequalities _ @xmath20 these ( in)equalities are essential in the characterization of stability , controllability and observability of system .
if @xmath21 , the inequalities can be written as [ eq : lyapequiv ] @xmath22 in the present paper we discuss extensions of and for stochastic linear systems .
as indicated above , the equivalent formulations and lead to different generalizations , if we consider it - type stochastic systems of the form @xmath23 where @xmath24 are as in ( [ eq : lti ] ) and @xmath25 .
system is asymptotically mean - square stable ( e.g. @xcite ) , if and only if there exists a positive definite solution @xmath15 of the generalized lyapunov inequality @xmath26 here @xmath27 and @xmath28 .
this stability criterion indicates that in the stochastic context the generalized lyapunov operator @xmath29 takes over the role of @xmath30 . substituting @xmath30 by @xmath29 in and , we obtain two different dual pairs of generalized lyapunov inequalities .
we call them _ type i _ : [ eq : gram1 ] @xmath31 and _ type ii _ : [ eq : gram2 ] @xmath32 note that corresponds to in the sense that @xmath33 has been replaced by @xmath34 , while corresponds to , where @xmath35 has been replaced by @xmath36 . in general (
if @xmath37 and @xmath17 do not commute ) , the inequalities and are not equivalent . at first glance it is not clear which generalization is more appropriate .
if the system is asymptotically mean - square stable and certain observability and reachability conditions are fulfilled , then for both types there are solutions @xmath38 . by a suitable state space - transformation ,
it is possible to balance the system such that @xmath39 is diagonal .
consequently , the usual procedure of balanced truncation can be applied to reduce the order of . for simplicity ,
let us refer to this as _ type i _ or _ type ii balanced truncation_. under natural assumptions , this reduction preserves mean - square asymptotic stability . for type
i , this nontrivial fact has been proven in @xcite .
moreover , in @xcite , an @xmath40-error bound has been provided .
however , different from the deterministic case , there is no @xmath0-type error bound in terms of the truncated entries in @xmath41 .
this will be shown in example [ ex : noerrbound ] .
in contrast , for type ii , an @xmath0-type error bound has been obtained in @xcite . in the present paper , as one of our main contributions , we show in theorem [ satz : trunc_stab2 ] that type ii balanced truncation also preserves mean - square asymptotic stability .
the proof differs significantly from the one given for type i. using this result , we are able to give a more compact proof of the error bound , theorem [ thm : newgramianbound ] , which exploits the stochastic bounded real lemma @xcite . we illustrate our results by analytical and numerical examples in section iv .
consider a stochastic linear control system of it - type @xmath42 where @xmath43 are uncorrelated zero mean real wiener processes on a probability space @xmath44 with respect to an increasing family @xmath45 of @xmath46-algebras @xmath47 ( e.g. @xcite ) .
+ to simplify the notation , we only consider the case @xmath48 and set @xmath49 , @xmath50 .
but all results can immediately be generalized for @xmath51 .
+ let @xmath52 denote the corresponding space of non - anticipating stochastic processes @xmath53 with values in @xmath54 and norm @xmath55 where @xmath56 denotes expectation . let the homogeneous equation @xmath57 be asymptotically mean - square - stable , i.e. @xmath58 , for all solutions @xmath59 . then , by theorem [ thm : hans ] the equations @xmath60 have unique solutions @xmath61 and @xmath62 . under suitable observability and controllability
conditions , @xmath18 and @xmath17 are nonsingular .
a similarity transformation @xmath63 of the system implies the contragredient transformation as @xmath64 choosing e.g. @xmath65 , with cholesky factorizations @xmath66 , @xmath67 and a singular value decomposition @xmath68 , we obtain @xmath69 and @xmath70 after suitable partitioning @xmath71,\ ; s=\left [ \begin{array}{cc } s_1&s_2 \end{array } \right],\ ; s^{-1}=\left [ \begin{array}{c } t_1\\t_2 \end{array } \right]\end{aligned}\ ] ] a truncated system is given in the form @xmath72 the following result has been proven in @xcite .
[ satz : trunc_stab1 ] let @xmath73 satisfy @xmath74 for a block - diagonal matrix @xmath75 with @xmath76 , assume that @xmath77 then , with the usual partitioning of @xmath10 and @xmath37 , we have @xmath78 its implication for mean - square stability of the truncated system is immediate . consider an asymptotically mean square stable stochastic linear system @xmath79 assume that a matrix @xmath80 is given as in theorem [ satz : trunc_stab1 ] and @xmath10 and @xmath37 are partitioned accordingly . + then the truncated system @xmath81 is also asymptotically mean square stable .
if the diagonal entries of @xmath82 are small , it is expected that the truncation error is small .
in fact this is supported by an @xmath40-error bound obtained in @xcite .
additionally , however , from the deterministic situation ( see @xcite ) , one would also hope for an @xmath0-type error bound of the form @xmath83 with some number @xmath84 .
the following example shows that no such general @xmath85 exists .
[ ex : noerrbound ] let @xmath86 $ ] with @xmath87 , @xmath88 $ ] , @xmath89 $ ] , @xmath90 $ ] .
+ solving with equality , we get @xmath91 $ ] , @xmath92 $ ] with @xmath93 so that @xmath94 , where @xmath95 and @xmath96 .
the system is balanced by the transformation @xmath97^{1/4}$ ] . + then @xmath98 $ ] so that @xmath99 for the truncated system of order @xmath100
. thus the output of the reduced system is @xmath101 , and the truncation error @xmath102 is equal to the stochastic @xmath0-norm ( see @xcite ) of the original system , @xmath103 we show now that this norm is equal to @xmath104 .
thus , depending on @xmath105 , the ratio of the truncation error and @xmath106 can be arbitrarily large . according to the stochastic bounded real lemma , theorem [ thm : sbrl ]
, @xmath107 is the infimum over all @xmath108 so that the riccati inequality @xmath109\nonumber\end{aligned}\ ] ] possesses a solution @xmath110<0 $ ] .
if a given matrix @xmath15 satisfies this condition , then so does the same matrix with @xmath111 replaced by @xmath112 .
hence we can assume that @xmath113 , and end up with the two conditions @xmath114 and ( after multiplying the upper left entry with @xmath115 ) @xmath116 thus necessarily @xmath117 , i.e. @xmath118 .
this already proves that @xmath119 , which suffices to disprove the existence of a general bound @xmath85 in .
taking infima , it is easy to show that indeed @xmath120 .
we now consider the inequalities .
[ lemma : gramexists ] assume that @xmath57 is asymptotically mean - square - stable .
then inequality is solvable with @xmath121 .
* proof : * by theorem [ thm : hans ] , for a given @xmath122 , there exists a @xmath123 , so that @xmath124
. then @xmath125 , for sufficiently small @xmath126 , satisfies @xmath127 so that holds even in the strict form .
@xmath128 + it is easy to see that like in the previous section a state space transformation @xmath129 leads to a contragredient transformation @xmath130 , @xmath131 of the solutions . that is , @xmath18 and @xmath17 satisfy and , if and only if @xmath132 and @xmath133 do so for the transformed data .
as before , we can assume the system to be balanced with @xmath134\;,\label{eq : defsigma}\end{aligned}\ ] ] where @xmath135 and @xmath136 , @xmath137 .
hence , we will now assume ( after balancing ) that a diagonal matrix @xmath41 as in is given which satisfies @xmath138 [ eq : pq_ineq ] partitioning @xmath10 , @xmath37 , @xmath139 , @xmath140 like @xmath41 , we write the system as @xmath141 the reduced system obtained by truncation is @xmath142 the index @xmath143 is the number of different singular values @xmath144 that have been kept in the reduced system . in the following subsections , we consider matrices @xmath145,\quad n=\left [ \begin{array}{cc } n_{11}&n_{12}\\n_{21}&n_{22 } \end{array } \right]\ ; , \end{aligned}\ ] ] @xmath80 as in , and equations of the form [ eq : cbtilde ] @xmath146 with arbitrary right - hand sides @xmath147 and @xmath148 . for convenience , we write out the blocks of these equations explicitly : @xmath149\nonumber a_{12}^t\sigma_1+\sigma_2a_{21}+n_{12}^t&\sigma_1n_{11}\\&=-n_{22}^t\sigma_2 n_{21}-\tilde c_2^t\tilde c_1\label{eq : allcomps2}\\[2 mm ] \nonumber a_{22}^t\sigma_2+\sigma_2a_{22}+n_{22}^t&\sigma_2n_{22}\\&=-n_{12}^t\sigma_1 n_{12}-\tilde c_2^t\tilde c_2\label{eq : allcomps3}\\[2 mm ] \nonumber
a_{11}^t\sigma^{-1}_1+\sigma^{-1}_1a_{11}+n_{11}^t&\sigma^{-1}_1n_{11}\\&=-n_{21}^t\sigma^{-1}_2 n_{21}-\tilde b_1\tilde b_1^t\label{eq : allcomps4 } \\[2mm]\nonumber a_{12}^t\sigma^{-1}_1+\sigma^{-1}_2a_{21}+n_{12}^t&\sigma^{-1}_1n_{11}\\&=-n_{22}^t\sigma^{-1}_2 n_{21}-\tilde b_2\tilde b_1^t\label{eq : allcomps5}\\[2 mm ] \nonumber
a_{22}^t\sigma^{-1}_2+\sigma^{-1}_2a_{22}+n_{22}^t&\sigma^{-1}_2n_{22}\\&=-n_{12}^t\sigma^{-1}_1 n_{12}-\tilde b_2\tilde b_2^t\label{eq : allcomps6}\end{aligned}\ ] ]
the following theorem is the main new result of this paper .
[ satz : trunc_stab2 ] let @xmath10 and @xmath37 be given such that @xmath150 assume further that for a block - diagonal matrix @xmath75 with @xmath76 , we have [ eq : sigma_ineq ] @xmath151 then , with the usual partitioning of @xmath10 and @xmath37 , we have @xmath152 again we have an immediate interpretation in terms of mean - square stability of the truncated system .
consider an asymptotically mean square stable stochastic linear system @xmath79 assume that a matrix @xmath80 is given as in theorem [ satz : trunc_stab2 ] and @xmath10 and @xmath37 are partitioned accordingly . +
then the truncated system @xmath81 is also asymptotically mean square stable .
* proof of theorem [ satz : trunc_stab2 ] : * note that the inequalities are equivalent to the equations with appropriate right - hand sides @xmath153 and @xmath154 .
+ by way of contradiction , we assume that does not hold .
then by theorem [ thm : kreinrutman ] , there exist @xmath155 , @xmath156 , @xmath157 such that @xmath158 taking the scalar product of the equation with @xmath159 , we obtain @xmath160 whence @xmath161 and @xmath162 , @xmath163 by corollary [ cor : semidefy ] .
hence @xmath164 analogously , we have @xmath165 by .
+ in particular , from @xmath163 , we get @xmath166\right)\\ & = \left [ \begin{array}{cc } a_{11}v+va_{11}^t+n_{11}vn_{11}^t & va_{21}^t+n_{11}vn_{21}^t \\ a_{21}v+n_{21}vn_{11}^t&n_{21}vn_{21}^t \end{array } \right]\\&=\left [ \begin{array}{cc } 0 & va_{21}^t \\
\end{array } \right]\;.\end{aligned}\ ] ] we will show that @xmath167 , which implies @xmath168 in contradiction to , and thus finishes the proof . we first show that @xmath169 is invariant under @xmath170 and @xmath171 . to this end
let @xmath172 .
then by , @xmath173 whence also @xmath174 , i.e. @xmath175 . from this
, we have @xmath176 implying @xmath177 .
thus @xmath178 and @xmath179 .
since @xmath180 , it follows further that @xmath169 is invariant under @xmath170 and @xmath171 .
let @xmath181 , where @xmath182 has full column rank , i.e. @xmath183 .
then by the invariance , there exist square matrices @xmath15 and @xmath184 , such that @xmath185 it follows that @xmath186 whence @xmath187 .
moreover , from , we get @xmath188 using this substitution in the following computation , we obtain @xmath189 we will show that the right hand side has nonnegative trace .
this then implies that the whole term vanishes . note
that @xmath190 taking the trace in , we have @xmath191^t m \left [ \begin{array}{c } v_1y\\v_1 \end{array } \right ] \;.\end{aligned}\ ] ] where @xmath192\;.\end{aligned}\ ] ] the matrix @xmath193 is positive semidefinite , because the upper left block is positive definite , and the corresponding schur complement @xmath194 vanishes .
hence @xmath195\left [ \begin{array}{c } v_1y\\v_1 \end{array } \right]&=0\end{aligned}\ ] ] implying via the first block row that @xmath196 . from
, using also again , we thus have @xmath197 i.e. @xmath198 .
it follows that for arbitrary @xmath199 , the eigenvector @xmath159 in can be replaced by @xmath200 because @xmath201 induction leads to @xmath202 as above , we conclude that @xmath203 , @xmath204 , and @xmath205 . multiplying with @xmath206 and with @xmath207 ,
we get @xmath208 hence ( after multiplication with @xmath82 ) , for all @xmath209 , we have @xmath210 applying this identity repeatedly , we get @xmath211 if @xmath212 is the minimal polynomial of @xmath213 , then @xmath76 implies @xmath214 and @xmath215 whence @xmath216 and also @xmath167 . hence we obtain the contradiction .
@xmath128 + the following theorem has been proven in @xcite using lmi - techniques . exploiting the stability result in the previous subsection
, we can give a slightly more compact proof based on the stochastic bounded real lemma , theorem [ thm : sbrl0 ] .
[ thm : newgramianbound ] let @xmath10 and @xmath37 satisfy @xmath217 assume furthermore that for @xmath75 with @xmath218 and @xmath76 , the following lyapunov inequalities hold ,
@xmath219 if @xmath220 and @xmath221 , then for all @xmath222 , it holds that @xmath223)}\le2 ( \sigma_{r+1}+\ldots+\sigma_\nu)\|u\|_{l^2_w([0,t])}\;.\ ] ] * proof : * we adapt a proof for deterministic systems e.g. ( * ? ? ?
* theorem 7.9 ) . in the central argument
we treat the case where @xmath224 and show that @xmath225}\le 2\sigma_\nu \|u\|_{l^2_w[0,t]}\;.\end{aligned}\ ] ] from and , we can see that also @xmath226 hence we can repeat the above argument to remove @xmath227 successively . by the triangle inequality we find that @xmath228}&\le \sum_{j = r}^{\nu-1}\|y_{j+1}-y_j\|_{l^2_w[0,t]}\\ & \le 2(\sigma_\nu+\ldots+\sigma_{r+1 } ) \|u\|_{l^2_w[0,t]}\;.\end{aligned}\ ] ] which then concludes the proof .
+ to prove , we make use of the stochastic bounded real lemma . in the following let @xmath229 and consider the error system defined by @xmath230 where @xmath231,\quad a_e=\left [ \begin{array}{ccc } a_{11}&a_{12}&0\\a_{21}&a_{22}&0\\0&0&a_{11 } \end{array } \right],\\ n_e&=\left [ \begin{array}{ccc } n_{11}&n_{12}&0\\n_{21}&n_{22}&0\\0&0&n_{11 } \end{array } \right],\quad b_e=\left [ \begin{array}{c } b_1\\b_2\\b_1 \end{array } \right],\\ c_e&=\left [ \begin{array}{ccc } c_1&c_2&-c_1 \end{array } \right]\;.\end{aligned}\ ] ] applying the state space transformation @xmath232&=\left [ \begin{array}{c } x_1-x_r\\x_2\\x_1+x_r \end{array } \right]=\underbrace{\left [ \begin{array}{ccc } i_r&0&-i_r\\0&i_{n - r}&0\\i_r&0&i_r \end{array } \right]}_{=s^{-1}}\left [ \begin{array}{c } x_1\\x_2\\x_r \end{array } \right ] , \end{aligned}\ ] ] we obtain the transformed system @xmath233\;,\\\tilde n_e&=s^{-1}n_es=\left [ \begin{array}{ccc } n_{11}&n_{12}&0\\\tfrac12n_{21}&n_{22}&\tfrac12n_{21}\\0&n_{12}&n_{11 } \end{array } \right]\;,\\ \tilde b_e&=s^{-1}b\left [ \begin{array}{c } 0\\b_2\\2b_1 \end{array } \right]\;,\quad\tilde c_e = c_es=\left [ \begin{array}{ccc } c_1&c_2&0 \end{array } \right]\;.\ ] ] by theorem [ thm : sbrl0 ] , we have @xmath234 , if the riccati inequality @xmath235 possesses a solution @xmath236 .
we will show now that the block - diagonal matrix @xmath237 satisfies . partitioning @xmath238 $ ]
, we have @xmath239 with the permutation matrix @xmath240 $ ] we define @xmath241 where @xmath242 by . using , we have @xmath243\\ & -\frac{\sigma_\nu}2\left [ \begin{array}{c } n_{21}^t\\0\\-n_{21}^t \end{array } \right]\left [ \begin{array}{c } n_{21}^t\\0\\-n_{21}^t \end{array } \right]^t+\sigma_\nu^2\left [ \begin{array}{c|c } 0&0\\\hline0 & m \end{array } \right]\le 0\;,\end{aligned}\ ] ] which is inequality .
@xmath128 + [ ex : noerror_cont ] let the system @xmath244 and @xmath18 be as in example [ ex : noerrbound ] .
the matrix @xmath245^{-1}>0\;,\text { where } 0<p\le 1\;,\end{aligned}\ ] ] satisfies inequality . as in example
[ ex : noerrbound ] , we have @xmath246 for the corresponding reduced system of order @xmath100 , so that the truncation error again is @xmath247 , independently of @xmath2480,1]$ ] . on the other hand we have @xmath249 with equality for @xmath250 .
theorem [ thm : newgramianbound ] thus gives the sharp error bound @xmath251 .
note , that there is no @xmath121 satisfying the _ equation _ .
the previous example illustrates the problem of optimizing over all solutions of inequality .
to compare the reduction methods we need to compute @xmath252 from or . instead of the inequalities , , we can consider the corresponding equations , for which quite efficient algorithms have been developed recently , e.g. @xcite .
these also allow for a low - rank approximation of the solutions .
in contrast we can not replace by the corresponding equation , because this may not be solvable ( see example [ ex : noerror_cont ] ) .
even worse , we do not have any solvability or uniqueness criteria nor reliable algorithms
. therefore , in general , we have to work with the inequality , which is solvable according to lemma [ lemma : gramexists ] , but of course not uniquely solvable . in view of our application
, we aim at a solution @xmath17 of , so that ( some of ) the eigenvalues of @xmath253 are particularly small , since they provide the error bound . choosing a matrix @xmath122 and a very small @xmath254 along the lines of the proof of lemma [ lemma : gramexists ] can be contrary to this aim .
hence some optimization over all solutions of is required .
note also that a matrix @xmath121 satisfies , if and only if it satisfies the linear matrix inequality ( lmi ) @xmath255&\le 0\;. \end{aligned}\ ] ] thus , lmi optimal solution techniques are applicable .
however , their complexity will be prohibitive for large - scale problems .
therefore further research for alternative methods to solve adequately is required . by @xmath256 and @xmath257
, we always denote the original and the @xmath143-th order approximated system .
the stochastic @xmath0-type norm @xmath102 is computed by a binary search of the infimum of all @xmath108 such that the riccati inequality is solvable .
the latter is solved via a newton iteration as in @xcite .
finally , the lyapunov equations are solved by preconditioned krylov subspace methods described in @xcite .
+ unfortunately , for small @xmath108 , i.e. for small approximation errors , this method of computing the error runs into numerical problems , because contains the term @xmath258 .
this apparently leads to cancellation phenomena in the newton iteration , if e.g. @xmath259 .
therefore we mainly concentrate on cases where the error is larger , that is we make @xmath143 sufficiently small . in many examples we observe
that type ii reduction gives a valid error bound , but the approximation error still is better with type i. this , however , is not always true , as the example @xmath260,\left [ \begin{array}{rr } 0&0\\1&0 \end{array } \right ] , \left [ \begin{array}{r } 0\\3 \end{array } \right ] , \left [ \begin{array}{r } 3\\0 \end{array } \right]\right)\end{aligned}\ ] ] shows .
it can easily be verified that the type i lyapunov equations are solved by @xmath261\text { and } p=\left [ \begin{array}{rr } 3&3\\3&6 \end{array } \right]\;.\end{aligned}\ ] ] the type ii inequalities are e.g. solved by @xmath261\text { and } p=\left [ \begin{array}{rr } 8&0\\0&12 \end{array } \right]\;.\end{aligned}\ ] ] reduction to order @xmath262 gives the following error bounds and approximation errors for both types : [ cols="^,^,^",options="header " , ] the main contributions of this paper are the preservation of asymptotic stability for type ii balanced truncation proved in theorem [ satz : trunc_stab2 ] and the new proof of the @xmath0 error bound in theorem [ thm : newgramianbound ] .
the efficient solution of is an open issue and requires further research .
the same is true for the computation of the stochastic @xmath0-norm .
consider the stochastic linear system of it - type @xmath263 where @xmath264 is a zero mean real wiener process on a probability space @xmath44 with respect to an increasing family @xmath45 of @xmath46-algebras @xmath47 ( e.g. @xcite ) .
+ let @xmath52 denote the corresponding space of non - anticipating stochastic processes @xmath53 with values in @xmath54 and norm @xmath55 where @xmath56 denotes expectation . by definition , system is asymptotically mean - square - stable , if @xmath58 , for all initial conditions @xmath265 .
we have the following version of lyapunov s matrix theorem , see @xcite . here
@xmath266 denotes the kronecker product .
[ thm : hans ] the following are equivalent .
( i ) : : system is asymptotically mean - square stable . (
ii ) : : @xmath267 ( iii ) : : @xmath268 : @xmath269 ( iv ) : : @xmath270 : @xmath269 ( v ) : : @xmath271 : @xmath269 the theorem ( like all other results in this paper ) carries over to systems @xmath272 with more than one noise term , and many more equivalent criteria can be provided , see e.g. @xcite or ( * ? ? ?
* theorem 3.6.1 ) .
+ the following theorem does not require any stability assumptions ( see ( * ? ? ? * theorem 3.2.3 ) ) .
it is central in the analysis of mean - square stability .
[ thm : kreinrutman ] let @xmath273 then there exists a nonnegative definite matrix @xmath156 , such that @xmath274 we also note a simple consequence of this theorem ( * ? ? ?
* corollary 3.2 ) . here
@xmath275 is the frobenius inner product for symmetric matrices .
[ cor : semidefy ] let @xmath276 as in the theorem .
for given @xmath277 assume that @xmath278 then @xmath279 .
moreover , if @xmath161 then @xmath280 .
now let us consider system with input @xmath281 and output @xmath282 .
if system is asymptotically mean - square stable , then defines an input output operator @xmath283 from @xmath284 to @xmath285 , see @xcite . by @xmath107
we denote the induced operator norm , which is an analogue of the deterministic @xmath0-norm .
it can be characterized by the stochastic bounded real lemma .
* system is asymptotically mean - square stable and @xmath287 .
* there exists a negative definite solution @xmath288 to the riccati inequality @xmath289 * there exists a positive definite solution @xmath290 to the riccati inequality @xmath291 we have stated the obviously equivalent formulations ( ii ) and ( iii ) to avoid confusion arising from different formulations in the literature . under additional assumptions also non - strict versions can be formulated .
the following sufficient criterion is given in ( * ? ? ?
* corollary 2.2.3 ) ( where also the signs are changed ) .
unlike in the previous theorem , here asymptotic mean - square stability is assumed at the outset .
w. s. gray , j. mesko , energy functions and algebraic gramians for bilinear systems , in : preprints of the 4th ifac nonlinear control systems design symposium , enschede , the netherlands , 1998 , pp .
103108 .
p. benner , t. damm , m. redmann , y. rocio rodriguez cruz , positive operators and stable truncation , linear algebra appl .
doi:10.1016/j.laa.2014.12.005 , in press . published electronically , dec . 30 , 2014 . | we consider two approaches to balanced truncation of stochastic linear systems , which follow from different generalizations of the reachability gramian of deterministic systems . both preserve mean - square asymptotic stability , but
only the second leads to a stochastic @xmath0-type bound for the approximation error of the truncated system .
generalized lyapunov equation , model order reduction , balanced truncation , stochastic linear system , asymptotic mean square stability 15a24 , 93a15 , 93b36 , 93b40 , 93d05 , 93e15 , |
the structure and success of general relativity have lead us to the believe that the spin-2 graviton field mediates quantum gravitational interactions ; but to date this believe has eluded experimental detection . furthermore , the quantum limit of general relativity fails to renormalize properly forcing us to further believe that general relativity comes from a classical projection of a 10 dimensional string .
on the other hand , we know that the geometrical object that describes the geodesics in general relativity is the connection and not the metric which enters that construction only after arbitrarily imposing the metric compatibility constraint
. then , we can in practice construct a quantum theory of gravity where the fundamental field is the connection as in@xcite or a similar object as we do here .
nothing but experimental observation constrains these possibilities .
therefore , we should not force the graviton upon ourselves and instead focus on matching the proposed model s differential cross sections to experimental evidence ; more so , when we must yet detect the particle mediating gravitational interactions . here
we present quantum gravitodynamics , qgd , the only renormalizable 4-dimensional theory that reproduces the interactions in general relativity to @xmath2 and couples to the standard model through a covariant derivative like the other three interactions in nature . as we show below , all diagrams in qgd
match those of it counterpart in general relativity to @xmath2 and the potential that determines a test particle s classical trajectories , are identical for both theories to that same order .
as opposed to general relativity which does not renormalize , qgd has a metric void of any dynamics , explicitly lacks a graviton field or any spin-2 state , and instead uses so(4 ) yang mills fields coupled to spinor singlets to mediate the gravitational interactions .
furthermore , qgd reproduces all diagrams in general relativity to @xmath2 and therefore all experimental evidence to that same order .
however , as opposed to general relativity , the terms in the qgd lagrangian involve products of fields and their derivatives with total dimensionality of at most four and therefore renormalization follows@xcite .
thus , qgd is renormalizable , and , like general relativity , constructed in 4 dimensions .
because qgd is defined in terms of yang mills fields , it also couples naturally to the standard model to produce a unified model with @xmath3 gauge symmetry that reproduces all the experimentally observed differential cross sections in nature , including those produced by gravitational interactions to @xmath2 .
qgd also has the ability to fit those diagrams contributions from gravity that may become available in the future such as higgs - gluon gravitational scattering .
in addition , qgd has gravitational su(2 ) yang mills bps states that match the entropy of black holes and type - ii strings .
pure yang s gravity@xcite has similar classical equations to those of einstein s relativity and includes schwarzschild - like solutions , whose singularity at the horizon , as opposed to black hole solutions , can not be removed through a coordinate transformation@xcite .
qgd and yang s gravity are both constructed in terms of the same connection defined over a fiber bundle , however , yang s gravity is not renormalizable because it couples its yang - mills field directly to the energy - momentum tensor . instead , in qgd , the yang mills fields couple to matter through a @xmath4 covariant derivative , in the same manner as the other forces in nature and which preserves renormalization .
therefore , the couplings in qgd ensure that all terms in the lagrangian have dimensionality four or less which guarantees renormalization@xcite .
string theory is the only other theory , besides qgd , to incorporate gravitational interactions while preserving renormalization .
it has been a successful gravitational theory , because like qgd , scattering processes associated with the @xmath5-function constrain the metric to be a solution of einstein s equation .
therefore , like qgd it reproduces all general relativity tests to @xmath2 . like strings , qgd has the necessary bps states to resolve the black hole entropy problem .
however , unlike qgd , string theory can not be defined without ambiguity in 4 dimensions and neither can it couple to the standard model which instead must be reformulated along side gravity .
the qgd connection defined on the so(4 ) fiber bundle can be projected to the connection defined on the so(1,3 ) tangent bundle .
this projection allows the generators of the so(4 ) gauge fields to be defined in terms of dirac gamma matrices @xmath6,\label{so4g}\ ] ] and therefore these generators do not commute with the spacetime dirac @xmath7 matrices .
this non - commutativity contrasts qgd from the other gauge theories in the standard model because it introduces terms proportional to @xmath8 $ ] which couple to fermionic matter fields with arbitrary couplings .
the map from the so(4 ) gravitational gauge field , @xmath9 , to the spin - connection defined in@xcite motivate calling @xmath9 the connecton . in order to test qgd
, we must reproduce the experimental evidence supporting general relativity .
as discussed at the end of section 4 , this also suffices to show that qgd will reproduce all effects of general relativity to @xmath2 .
the procedure used to fit qed to experimental evidence , see for example@xcite , transfers well to qgd and permits the fitting of all the qgd parameters to experimental observation .
the t - matrix calculation in@xcite associated with the tree - level @xmath10 scattering process produces a differential cross section ; comparison of that differential cross section with the one obtained from the schroedinger equation with a coulomb potential allows the fitting of the electron coupling found in qed to the experimentally measured charge used in the coulomb potential .
this procedure extends to the quantized but non - renormalizable general relativity theory .
in@xcite the tree - level scattering process for two bosons , @xmath11 , produces a differential scattering matrix which fits the schroedinger equation with a newton potential while the 1-loop scattering correction fits the differential cross section produced by the schroedinger equation with the leading classical post newtonian correction to the potential .
the strategy implemented in@xcite and@xcite was also successfully implemented to the gravitational sector in string theory @xcite-@xcite .
thus , matching the model s differential cross sections to experimental / expected differential cross sections allows us to test if a model correctly describes nature . here
we follow the same strategy as that used in@xcite-@xcite to show that qgd coupled to the standard model fits all the known experimental differential cross sections , including those gravitational in nature .
section 2 details the symmetries and action derived for a single matter field and the u(1 ) electromagnetic field , both coupled to gravity . in section 3 we calculate the propagators and vertices of the model presented in section 2 . in section 4
we calculate t - matrix elements to produce differential cross sections which are then fitted to the differential cross section of the schroedinger equations with the coulomb , the newtonian and post newtonian potentials as well as the differential cross section for the deflection of light by a point particle in the small angle approximation found in general relativity .
section 5 presents the full qgd coupling to the standard model while section 6 fits the differential cross sections to experimentally observed differential cross sections .
section 7 moves outside the realm of experimental observation and shows that qgd can include the expected gravitational interactions of the su(2 ) and su(3 ) gauge bosons as well as the higgs doublet . in section 8
we consider the isomorphic @xmath12 yang mills representation of qgd , which when @xmath13 supersymmetrized has bps states whose entropy corresponds to that of a black hole , and that can bridge to type - ii strings@xcite .
the conclusions appear in section 9 .
the fundamental geometrical object of general relativity , the connection on the tangent bundle , does not require a metric to describe a particle s geodesics .
consistent with general relativity and in the spirit of@xcite , qgd chooses a connection over the metric as the fundamental field .
qgd does not use a yet to be detected graviton field or spin-2 particle and it uses instead a constant metric void of any dynamics .
instead , connections defined on the so(4 ) fiber bundle mediate all gravitational interactions ; these so(4 ) gauge fields alone mediates all the gravitational interactions between particles .
these fields suffice to obtain all the differential cross sections expected from general relativity for a single matter - photon system .
more importantly , this construction does not have any experimental impediments because we have yet to detect the graviton or any other spin-2 particle .
we present the lagrangian @xmath14\psi.\label{lagf}\end{aligned}\ ] ] we use the following conventions .
the diagonal and flat space - time metric , @xmath15 , does not carry any dynamics and has signature @xmath16 .
@xmath17 represents the levi - civita tensor in 4 dimensions .
the vielbeins @xmath18 connect space - time coordinates ( greek indices ) with the so(4 ) fiber bundle coordinates ( latin indices ) .
in particular @xmath19 and @xmath20 .
the metric @xmath21 , defined on the fiber bundle , has signature @xmath22 .
we define the field strengths appearing in ( [ lagb ] ) @xmath23 @xmath24 represents the u(1 ) electromagnetic gauge field and the so(4 ) gauge field @xmath25 , the connecton , mediates of gravitational interactions .
the couplings @xmath26 , @xmath27 , @xmath28 and @xmath29 denote coupling constants and @xmath30 . with @xmath31
defined as a copy of a textbook so(n=4 ) gauge field necessarily antisymmetric in indices @xmath32 and @xmath33 , then @xmath34 so that the indispensable property @xmath35 holds .
the covariant derivative operator @xmath36 defined in ( [ lagf ] ) acts on a 4 dimensional dirac spinor field @xmath37 @xmath38 and depends on the so(4 ) generators @xmath39 defined in ( [ so4 g ] ) .
the u(1 ) local gauge transformation @xmath40 , with @xmath41 a scalar function , acts on the following fields @xmath42 which also transforms the covariant derivative in the standard way @xmath43 but leaves ( [ lagtodo ] ) unchanged .
the so(4 ) local gauge transformations with @xmath44 and @xmath45 a 2-tensor , act on the following fields @xmath46 which also transforms the covariant derivative in the standard way @xmath47,\ ] ] but also leaves ( [ lagtodo ] ) unchanged .
the so(1,3 ) local gauge transformations acting on the fermions have the same effect as the so(4 ) transformation because @xmath48 however , the so(1,3 ) gauge transformations , can not act on the pure yang - mills lagrangian because the cartan - killing metric for that gauge group , given by the trace over two group elements , is not positive definite which yields states not bounded from below@xcite .
this property therefore distinguishes theories defined over the tangent bundle with local so(1,3 ) symmetry from those defined over the fiber bundle with local so(4 ) symmetry .
the former can not be used to define yang - mills theories even when the fiber bundles can be projected to the tangent bundle . for the avoidance of doubt , gravitational theories like spin - gravity can be defined over the tangent bundle only because , as a result of the different construction from yang - mills , their spectrum does not become unbounded from below when the cartan - killing metric is not positive definite .
the lorenz gauge fixings @xmath49 and property ( [ prop1 ] ) simplify the quadratic component in ( [ lagb ] ) to @xmath50 the structure constant defined in@xcite @xmath51 simplifies the relevant terms of the cubic terms in ( [ lagb ] ) to @xmath52 the terms in ( [ lag13 ] ) take the form @xmath53 and @xmath54 . in the low energy limit the photon momentum , @xmath55 , and the connecton momentum , @xmath56 , satisfy @xmath57 or equivalently @xmath58 this condition accounts for the omission of all other terms .
for example , for matter - photon scattering the vertex generated by @xmath59 would only participate at 1-loop level , would necessarily appear twice and thus takes values proportional to the square of the connecton momentum . instead , the vertices produced by the second term in ( [ lag13 ] ) would appear first at tree - level and then at 1-loop level where it would contribute terms proportional to the photon momentum and its square respectively .
dimensional analysis of general relativity shows that it does not renormalize properly .
however , it follows from ( [ eq1a ] ) that the connecton has the same dimensionality as the u(1 ) gauge field .
therefore , the terms in the lagrangian ( [ lag12 ] ) and ( [ lag13 ] ) involve products of fields and their derivatives with total dimensionality four or less .
renormalization then follows as shown in@xcite .
since @xmath60\ne0 $ ] , we use the relation @xmath61 to express @xmath62\psi&=&g \bar\psi\bar\omega_{\;\;\;\;\mu}^{\mu a}\gamma_a\psi.\label{lf2}\end{aligned}\ ] ] note that ( [ lf2 ] ) does not depend on @xmath63 .
this means that gauge invariance constrains ( [ lf1 ] ) but not ( [ lf2 ] ) .
thus the rescaling of ( [ lf1 ] ) by a constant redefines the coupling .
however , the redefinition guarantees gauge invariance only after a subsequent redefinition of the spinor @xmath37 . on the other hand ,
a recalling by a constant , even a complex one , of ( [ lf2 ] ) does not affect gauge invariance .
this contrasts with su(2 ) of su(3 ) couplings where the dirac matrices commute with their generators and therefore ( [ lf2 ] ) necessarily vanishes . the relations ( [ lf1 ] ) and ( [ lf2 ] ) simplify the lagrangian ( [ lagf ] ) to @xmath64 here
the real coupling constants @xmath26 , @xmath27 , @xmath29 and @xmath28 adjust the theory to observation . in particular
@xmath26 represents the electric charge and gauge invariance constrains @xmath27 and coincides with the coupling constant found in the purely bosonic sector ( [ lag13 ] ) . instead ,
gauge invariance does not constrain @xmath28 which can take any value , real , purely imaginary or complex .
again , the terms in the lagrangian ( [ lag2 ] ) involve products of fields and their derivatives with total dimensionality four or less . therefore , the terms in the lagrangians ( [ lag12 ] ) , ( [ lag13 ] ) and ( [ lag2 ] ) , and thus lagrangian ( [ lagtodo ] ) , involve products of fields and their derivatives with total dimensionality four or less .
renormalization then follows as shown in@xcite .
the lagrangian @xmath65 has an complex coupling constant @xmath28 while @xmath66 has a real coupling constant @xmath27 .
then , self energy corrections can have imaginary contributions and therefore it would be tempting to assume that the probability density determined by the wavefunction decreases over time and unitarity is lost .
this is not the case .
we choose a pure real @xmath27 and a pure imaginary @xmath28 and will show when fitting theory to experiment in the subsequent sections that @xmath67 where @xmath68 represents the gravitational constant .
the self - energy contributions to @xmath69 for both bosons @xmath24 and @xmath9 come from purely bosonic diagrams and those with fermionic loops .
we construct the purely bosonic diagrams with vertex ( c ) and vertex ( d ) defined in section [ pv1 ] .
therefore , the diagrams depend only on the real valued coupling constant @xmath27 @xmath70 @xmath71 thus , those diagrams always contribute real amounts to the self energies @xmath72 , @xmath73 , @xmath74 and @xmath75 . as we show below , real contributions to the self energy
do not affect unitarity .
the diagrams involving fermionic loops always involve an even number of vertices ( a ) or an even number of vertices ( b ) defined in section [ pv1 ] .
therefore , corrections to @xmath69 to the self energies of both @xmath24 and @xmath9 always contribute even powers of @xmath28 @xmath76 @xmath77 for some real constants @xmath78 and @xmath79 . even though the coupling @xmath28 is imaginary , all corrections to the self energies for both bosons @xmath24 and @xmath9 contribute real amounts to @xmath69 , even more , because fermions always appear in loops regardless of the order of perturbation , the self energy corrections to the bosons @xmath24 and @xmath9 always remain real .
on the other hand , the contributions to the self energies of the fermion contribute real amounts to @xmath80 but complex amounts to @xmath69 .
the contribution to the fermion self energy to @xmath80 comes from diagrams constructed from a pair of vertices ( a ) or a pair of vertices ( b ) .
therefore @xmath81 for some real constants @xmath82 and @xmath83 . the first imaginary contribution comes from the diagram involving one vertex ( c ) and three vertices ( a ) .
therefore @xmath84 therefore , the leading imaginary correction to the fermion self energy contributes at most of @xmath69 .
the time evolution of the wavefunction for the fermion @xmath85 where @xmath86 describes the wavefunction at the emitter and @xmath87 represents the wavefunction at the detector , implies that the probability density at time @xmath88 , @xmath89 , decays as @xmath90 when @xmath91 particles are lost as they travel between emitter and detector .
this experimentally unobserved reduction in flux prevents large imaginary values of @xmath92 .
however , when @xmath93 satisfies @xmath94 then , the self energy @xmath95 can take on imaginary values without loss of unitarity .
since the leading imaginary correction to the self energy of the fermion contributes at most @xmath69 , then , @xmath96 whenever @xmath97 years .
bearing in mind that the universe has existed less than @xmath98 years and that estimates put the the proton s half life at @xmath99 years , qgd can take imaginary couplings without affecting unitarity .
this contrasts with other theories where the coupling constants are much larger than @xmath28 leading to a loss of unitarity if not of gauge invariance as well since @xmath100 .
this conclusion is further supported by the absence of diagrams contributing to @xmath69 to the self energies constructed with vertex ( a ) and vertices involving the electromagnetic field .
the lagrangian densities ( [ lag12 ] ) and ( [ lag2 ] ) yield the following propagators gluon ( 70,15 ) propagator ( a ) & @xmath101 + fermion ( 70,15 ) propagator ( b ) & @xmath102 + photon ( 70,15 ) propagator ( c ) & @xmath103 + for the avoidance of doubt , in the limit considered here @xmath104 or equivalently @xmath105 . the lagrangian densities ( [ lag13 ] ) and ( [ lag2 ] ) generate the following vertices : |p4.5cm| p8.5cm| vertex1 ( 100,65 ) vertex ( a ) & @xmath106 + vertex2 ( 100,65 ) vertex ( b ) & @xmath107 + vertex3 ( 100,70 ) vertex ( c ) & @xmath108 @xmath109 + vertex4 ( 100,65 ) vertex ( d ) & @xmath110 + in the following section we relate the tree and 1-loop level diagrams to physical processes in order to determine the four parameters in qgd .
in this section we calculate 4 scattering amplitudes : the tree - level photon exchange between two particles with equal mass , the tree - level connecton exchange and the 1-loop connecton exchange between two particles with equal mass and the tree - level connecton exchange between a photon and a particle with mass .
these scattering amplitudes yield a total of 4 differential cross sections ; we equate the first 3 to the differential cross section for a schroedinger equation with 3 different potentials : the coulomb potential , the newton potential and the post - newtonian potential while the last cross section matches the cross section for gravitational lensing of a photon by a massive point particle .
the lagrangian provides 4 coupling constants : @xmath26 , @xmath27 , @xmath28 and @xmath29 .
therefore by the end of this section , the 4 coupling constants appearing in the differential cross sections calculated from qgd match those of the schroedinger equation as well as that obtained from gravitational lensing in general relativity .
|p5cm| p5cm| dia2 ( 120,80 ) diagram ( a ) & dia1 ( 120,80 ) diagram ( b ) + the contribution from the photon exchange between two mass particles in diagram ( a ) produces the t - matrix @xmath111 with @xmath112 and @xmath113 .
the property @xmath114 suppresses the second term in ( [ tree - a ] ) .
a further limiting to the spinless schroedinger limit requires summation over the spin of all in - states and out - states so that @xmath115 .
thus , in the non - relativistic quantum limit , the third term in ( [ tree - a ] ) vanish exactly reducing ( [ tree - a ] ) to @xmath116 this t - matrix yields the differential cross section @xmath117 while the schroedinger equation with a coulomb potential , @xmath118 with @xmath119 , the measured electric charge , and @xmath120 , the distance between the two particles , produces a differential cross section @xmath121 comparing the differential cross section ( [ cross - a ] ) and ( [ coulomb - a ] ) yields @xmath122 diagram ( b ) contributes to the differential scattering cross section of the connecton exchange between two mass particles .
the t - matrix element @xmath123 produces the quantum gravitodynamic differential cross section @xmath124 the schroedinger equation with a newtonian potential , @xmath125 with @xmath68 representing the gravitational constant , @xmath126 representing the measured mass and @xmath120 representing the distance between the two particles , produces a differential cross section @xmath127 comparing the differential cross section ( [ cross - b ] ) and ( [ newton - b ] ) yields @xmath128 and @xmath129 dia3 ( 160,160 ) diagram ( c ) & dia32 ( 160,160 ) diagram ( d ) + 1.0 cm the post - newtonian correction comes from diagrams ( c ) and ( d ) .
all other 1-loop diagrams contribute to either analytic terms , or quantum corrections of order @xmath130 and do not contribute to the low energy limit considered here@xcite .
in addition , diagrams involving photons and connectons simultaneously do not contribute because the transformation @xmath131 leaves the contribution of the above diagram unchanged while all terms but the first two in ( [ lag13 ] ) remain present in the action .
these two diagrams contribute to the quantum gravitodynamic post newtonian t - matrix @xmath132 this result as those below require the following steps .
first , we focus only on the electric " form factor which means that @xmath133 equivalently , we disregard terms proportional to @xmath134 $ ] that contribute to the magnetic " form factor .
second , the on - shell external momenta imply the relations @xmath135 third , we suppress terms of @xmath136 using the low energy limit property @xmath137 .
fourth , we use the approximation @xmath138 and @xmath139 .
finally , we use the appendix in@xcite with the expressions for the several feynman integrals .
the t - matrix ( [ t - cd ] ) produces the differential scattering cross section @xmath140 the schroedinger equation with a post newtonian correction to the gravitational potential ( see equation ( 39 ) of@xcite ) , @xmath141 where the constant @xmath32 depends on the post - newtonian expansion , produces , after using ( [ tree - b ] ) and ( [ tree - b2 ] ) , the differential scattering cross section @xmath142 comparing the differential cross section ( [ cross - cd ] ) and ( [ einstein - cd ] ) yields @xmath143 therefore , by fixing @xmath26 , @xmath27 and @xmath28 through equations ( [ eq - a ] ) , ( [ tree - b ] ) and ( [ eq - cd ] ) , qgd reproduces the low energy limit of general relativity for the matter sector .
therefore , the so(4 ) yang mills theory coupled to matter and an inert metric is a valid renormalizable relativistic quantum field theory to describe gravity s experimental observations .
|p5cm| dia6 ( 120,80 ) diagram ( e ) + diagram ( e ) describes the scattering of a fermion @xmath144 with a photon @xmath145 that produces the t - matrix @xmath146 which in turn , in the small angle approximation where @xmath147 , yields the differential cross section @xmath148 where the photon energy , @xmath149 , satisfies @xmath150 derivation of ( [ saa1 ] ) requires the photon on - shell conditions , @xmath151 along with the steps used for the matter - matter scattering above . substitution of ( [ tree - b ] ) and ( [ eq - cd ] ) into ( [ saa1 ] ) yields @xmath152 in the low energy limit , equation ( [ saa2 ] ) describes the rutherford scattering of a mass @xmath153 projectile off of a mass @xmath149 target . in the small angle deflection approximation and after setting the speed of light to 1 , the gravitational lensing of a photon by a point particle of mass @xmath153 in general relativity relates the impact parameter @xmath154 with the deflection angle @xmath155 @xmath156 the relation between the impact parameter , @xmath157 , and the differential cross section , @xmath158 simplifies the latter to @xmath159 or equivalently @xmath160 equation ( [ saa5 ] ) describes the rutherford scattering of mass @xmath149 projectile off of a mass @xmath153 target .
the exchange @xmath161 modifies this differential cross section to @xmath162 which then describes the scattering of a mass @xmath153 particle of a photon target with energy @xmath149 . equating ( [ saa2 ] ) with ( [ saa3 ] ) requires that @xmath163 which determines @xmath29 as a function of the known constant @xmath32 . with ( [ tree - light ] )
satisfied , fourier transform in the born approximation of ( [ saa2 ] ) yields the low energy potential @xmath164 which describes the motion of a photon as it had mass equal to its energy in the newtonian approximation .
diagram ( e ) also describes the gravitational redshift .
when the impact parameter vanishes , a photon starting at @xmath165 approaches the mass target and experiences the potential ( [ potphoton ] ) which accelerates the photon and blueshifts its frequency .
thus the energy of the photon at point @xmath166 reads @xmath167 and when @xmath168 @xmath169 as required .
diagram ( e ) has the same value as its counterpart diagram in general relativity obtained following the methods of @xcite which produces the exact same potential .
renormalization aside , it does so in the frame of reference where the photon scatters off the fermion , and again confirms that we can model photons as particles with mass @xmath170 and ( [ potphoton ] ) . like the gravitational redshift , the shapiro delay also stems from diagram ( e ) and it is just the length of the path traveled by the photon when bending about a mass particle and after reflection from some other body retraces its path back to the source .
this feature transpires the quantum limit were diagrams in general relativity match the diagrams for qgd to @xmath2 . in the classical limit , the trajectories of general relativity on a schwarzschild background also follow from consideration of a particle moving about a potential .
therefore , since we have shown that both theories have the same potential to @xmath2 , any classical or quantum test that general relativity passes will also be passed by qgd to that same order .
the qgd lagrangian ( [ lagtodo ] ) has 4 free parameters @xmath26 , @xmath27 , @xmath28 and @xmath29 .
equations ( [ eq - a ] ) , ( [ tree - b ] ) , ( [ eq - cd ] ) and ( [ tree - light ] ) , repeated here : @xmath171 fix these parameters to experimentally observed quantities from qed , and general relativity : the electric charge , the newtonian mass and gravitational constant , the precession of a mass particle about another mass particle and the gravitational lensing of light by mass particles . therefore , qgd suffices to describe the gravitational effects observed in nature without resorting to a spin-2 particle and using instead a quantum connection defined over an so(4 ) fiber bundle .
the dimensionality of the connecton is the same as that of the u(1 ) gauge field .
therefore , the terms in the lagrangians ( [ lag12 ] ) , ( [ lag13 ] ) and ( [ lag2 ] ) involve products of fields and their derivatives with total dimensionality four or less
. then all scattering amplitudes considered in this section renormalize by finite amounts@xcite .
the matching of the above 4 differential cross sections with those of general relativity coupled to matter and light are more than simple tests . by matching the differential cross sections we have , at least for two particle interactions , shown that the equations of motion for matter and light coupled to general relativity exactly match those for qgd to @xmath2 .
this is manifested by the fact that the schroedinger equation for both theories are the same .
therefore , any test that follows from general relativity coupled to matter and light will be reproduced by qgd to order @xmath172 . while it may very well be the case that general relativity and qgd diverge beyond @xmath2 , the reader is reminded that qgd needs to replicate nature and not general relativity . we know
that because qgd and general relativity are indistinguishable to @xmath2 , it replicates nature to that order .
but general relativity , safe for @xcite , has not been experimentally proven beyond that order , and therefore it may very well be that if qgd and general relativity diverge beyond @xmath2 , it will be up to nature to judge which most closely resembles it , always taking into account that general relativity is at best and effective theory because it is not renormalizable like qgd . it should be further noted that while the differential cross sections pertain two particle interactions and omit single particle and three- , four - particle interactions , these do not impose further constraints .
the single particle diagrams involve loops which general relativity can not handle due to lack of renormalization .
furthermore , three and four particle interactions are constrained by two particle scattering ; see @xcite for the diagrams used in general relativity .
therefore , by considering only the equivalency of two particle scattering processes , we fix the equations of motions for all other scattering processes in both theories .
lack of evidence supporting the existence of a spin-2 particle along with the existence of a mapping between an so(4 ) fiber bundle to an so(1,3 ) tangent bundle motivate the formulation of a gravitational theory based on the so(4 ) connecton instead of one based on the graviton . without a graviton field and with a constant diagonal metric void of any dynamics , qgd again incorporates all the necessary gravitational interactions to the standard model using instead @xmath173 so(4 ) gauge fields @xmath174 .
the inclusion of the fields @xmath174 suffices to obtain all the experimentally verified differential cross sections , and those expected from general relativity . the experimental evidence for general relativity exists only for the matter and u(1 ) gauge sector . at present
, no experimental evidence exists supporting that the su(2 ) , su(3 ) and the higgs sectors of the standard model couple to gravity ; however , here we show that qgd can incorporate such couplings to satisfy the intuitive expectation of general relativity .
the lagrangian @xmath175\psi^{(j)}\label{lagfsw}\\ l_{h}&=&\frac{1}{2\beta_{higgs}^2}d_\mu \phi\cdot d^\mu \phi -v(\phi\cdot \phi)\label{laghsw}\end{aligned}\ ] ] remains invariant under so(4 ) transformations as well as @xmath176 transformations .
there are @xmath173 covariant derivatives @xmath177 in ( [ lagfsw ] ) .
this new artifice allows us to introduce enough parameters to fit all the gravitational scattering amplitudes .
the u(1 ) gauge field @xmath178 , the su(2 ) gauge field @xmath179 and the su(3 ) gauge field @xmath180 along with the so(4 ) gauge fields @xmath181 define the strength fields in ( [ lagbsw ] ) @xmath182,\\ w_{\mu\nu}^a&=&\frac{1}{\beta_{su(2)}}\bigg[\partial_\mu w^a_\nu-\partial_\nu w^a_\mu -\frac{g_{su(2)}}{\beta_{su(2)}}c^{abc}w_\mu^b w_\nu^c\bigg ] , \\
g_{\mu\nu}^a&=&\frac{1}{\beta_{su(3 ) } } \bigg[\partial_\mu g^a_\nu-\partial_\nu g^a_\mu -\frac{g_{su(3)}}{\beta_{su(3 ) } } c^{abc}g_\mu^b g_\nu^c\bigg ] , \\ \omega_{\mu\nu}^{(i , m)ab}&=&(\partial_\mu\omega_\nu^{(i , m)ab}+\frac{\alpha_{im } } { \beta_{im}}\epsilon_\nu^{ab\rho}\partial_\mu a^m_\rho)- ( \partial_\nu\omega_\mu^{(i , m)ab}+\frac{\alpha_{im } } { \beta_{im}}\epsilon_\mu^{ab\rho}\partial_\nu a^m_\rho)\nonumber\\ & & -g_{i\,m } c^{ab}_{\;\;\;cdef}(\omega_\mu^{(i , m)cd}+\frac{\alpha_{im } } { \beta_{im}}\epsilon_\mu^{cd\rho } a^m_\rho)(\omega_\nu^{(i , m)ef}+ \frac{\alpha_{im } } { \beta_{im}}\epsilon_\nu^{ef\sigma } a^m_\sigma)\nonumber\\ & & \end{aligned}\ ] ] we define each @xmath183 as a copy of a textbook so(4 ) gauge field necessarily having antisymmetry in indices @xmath32 and @xmath33 , then , @xmath184 so that @xmath185 since @xmath186 where @xmath187 and @xmath188 describes the higgs in the fundamental of so(4 ) , the constant couplings @xmath189 must take the form @xmath190 so that @xmath191 and @xmath192 when representing a quark doublet , the spinors @xmath193 , @xmath194 transform in the irreducible representation of @xmath176 while when representing a lepton doublet , the spinors @xmath193 , @xmath195 transform in the irreducible representation of @xmath196
. then , each of the covariant derivatives @xmath197 , @xmath198,\label{cdsw}\ ] ] acting on a quark doublet @xmath193 transforms covariantly under @xmath3 .
the u(1 ) transformation generated by @xmath199 acts on the following fields @xmath200 and leaves ( [ lagtodosw ] ) invariant while each @xmath201 transforms covariantly .
since @xmath202=[\tau^a , t^{ab}]=[t^a,\tau^a]=0,\ ] ] su(3 ) transformations generated by @xmath203 act on the following fields @xmath204 where @xmath205 represents the su(3 ) structure constant , to leave ( [ lagtodosw ] ) invariant while each @xmath201 transforms covariantly .
similarly , su(2 ) transformations generated by @xmath206 act on the following fields @xmath207 where @xmath205 represents the su(2 ) structure constant , and also leave ( [ lagtodosw ] ) invariant while each @xmath201 transforms covariantly .
so(4 ) transformations generated by @xmath208 act on the following fields @xmath209 and also leave ( [ lagtodosw ] ) invariant while each @xmath201 transforms covariantly .
under @xmath210 transformations , ( [ cdsw ] ) acting on a lepton doublet @xmath193 transforms covariantly while ( [ lagtodosw ] ) remains invariant .
the bosonic component of the lagrangian , ( [ lagbsw ] ) , has quadratic terms which are included in ( [ nov6 ] ) instead . ]
@xmath211 the dimensionality of the connecton is the same as that of the u(1 ) gauge field .
therefore , the terms in the lagrangian ( [ nov4 ] ) involve products of fields and their derivatives with total dimensionality four . the pure standard model su(2 ) and su(3 ) cubic and quartic bosonic interactions present in ( [ lagbsw ] ) do not mix with the so(4 ) gauge fields and do not merit further treatment here . for the same reasons presented for ( [ lag13 ] ) ,
the relevant cubic terms in lagrangian ( [ lagbsw ] ) include @xmath212 as in ( [ lag13 ] ) , the terms in ( [ nov5 ] ) take the form @xmath53 and @xmath54 .
in the low energy limit the gauge boson momentum , @xmath55 , and the connecton momentum , @xmath56 , satisfy @xmath213 .
the terms in the lagrangian ( [ nov5 ] ) also involve products of fields and their derivatives with total dimensionality four
. therefore any diagrams constructed from the terms in ( [ nov5 ] ) will renormalize by finite amounts@xcite . after adding the contributions from ( [ lagbsw ] ) dependent on @xmath214 to ( [ laghsw ] ) we obtain the standard model higgs lagrangian along with qgd interactions @xmath215 , \label{nov6}\end{aligned}\ ] ] where @xmath216 describes the higgs in the su(2 ) doublet representation instead of the so(4 ) vector representation and @xmath217 the dual representation of the higgs
further strengthens the proposal of the so(4 ) connecton as the carrier of quantum gravitational interactions .
the terms in the lagrangian ( [ nov6 ] ) again involve products of fields and their derivatives with total dimensionality four .
thus , any corrections involving only these terms will renormalize by finite amounts@xcite .
when @xmath193 represents a quark doublet , the lagrangian ( [ lagfsw ] ) simplifies to @xmath218 \psi^{(j ) } , \label{nov8}\end{aligned}\ ] ] where @xmath219 and @xmath220 are the electric charges of each respective quark and @xmath221 @xmath222 similarly , when @xmath193 represents a lepton doublet , the lagrangian ( [ lagfsw ] ) simplifies to @xmath223 \psi^{(j ) } , \label{nov8b}\end{aligned}\ ] ] where @xmath224 and @xmath225 are the electric charges of each respective lepton .
in addition to the transformation of the terms in the pure standard model , the electroweak rotation which produces the u(1 ) gauge field @xmath24 only transforms the last term in ( [ nov8 ] ) and ( [ nov8b ] ) @xmath226 where @xmath227 describes the weak mixing angle and takes the same values as in the pure standard model .
thus , for a quark doublet @xmath228 \psi^{(j ) } , \nonumber\\ \label{nov8 - 2}\end{aligned}\ ] ] with @xmath229 representing the fermionic sector of the pure standard model after the rotation between @xmath178 and @xmath230 .
a similar expression follows for the lepton doublet .
the dimensionality of the connecton is the same as that of the u(1 ) gauge field .
therefore , the terms in the lagrangian ( [ nov8 - 2 ] ) involve products of fields and their derivatives with total dimensionality of at most four .
renormalization then follows@xcite .
more importantly , the whole lagrangian , the sum of ( [ nov4 ] ) , ( [ nov5 ] ) , ( [ nov6 ] ) and ( [ nov8 - 2 ] ) , involve products of fields and their derivatives also with total dimensionality at most four and therefore any corrections involving their terms renormalize by finite amounts@xcite .
the electroweak rotation transforms @xmath231 into @xmath232 and it transforms @xmath233 into @xmath234 after choosing @xmath235 , @xmath236 and @xmath237 . at this point
we can apply the higgs mechanism by selecting the usual vev for the higgs field which will produce the same effects that take place in the standard model .
the rotation between the fields does not affect the dimensionality of the fields .
therefore , the rotated terms in the lagrangian involve products of fields and their derivatives with total dimensionality at most four .
renormalization then follows@xcite .
the experimental evidence for gravitational interactions only includes observations for the u(1 ) gauge fields and the fermionic matter fields .
thus , while the model can include gravitational interactions with the su(2 ) , su(3 ) and higgs sectors , we limit ourselves in this section to the study of the u(1 ) gauge fields and the fermionic matter fields .
this removes all terms in ( [ nov4 ] ) , ( [ nov8 - 2 ] ) , and ( [ nov11 ] ) which carry index @xmath153 as well and allows us to drop altogether the higgs sector ( [ nov6 ] ) .
then , lagrangian ( [ nov8 - 2 ] ) expressed in terms of true fermions reduces to @xmath238 \psi^{(j ) } , \nonumber\\ & & \label{nov12}\end{aligned}\ ] ] with @xmath239 running over the number of singlet spinors , @xmath240 and @xmath241 .
the lagrangian ( [ nov47 ] ) reduces to @xmath242 and the lagrangian ( [ nov11 ] ) reduces to @xmath243 the dimensionality of the connecton is the same as that of the u(1 ) gauge field .
therefore , the terms in the lagrangian ( [ nov12 ] ) , ( [ nov13 ] ) and ( [ nov14 ] ) involve products of fields and their derivatives with total dimensionality four .
renormalization then follows as shown in@xcite . in sections 2 to 4 we considered a single particle , like the electron , and successfully reproduced the necessary post - newtonian corrections to fit theory to experiment .
expanding the particle spectrum increases the number of scattering cross sections well beyond 4 ; this leads to a lack of solution when a single connecton describes the gravitational interactions .
instead we introduced @xmath244 connectons to increase the number of couplings from @xmath245 to @xmath246 and from @xmath247 to @xmath248 . with this increase in the number of connectons
the number of parameters in the model also increases sufficiently to fit all these differential cross sections . the lagrangian densities ( [ nov12])-([nov14 ] ) yield the following propagators gluonsm ( 70,15 ) propagator ( d ) & @xmath101 + fermionsm ( 70,15 ) propagator ( e ) & @xmath102 + photon ( 70,15 ) propagator ( f ) & @xmath103 + for the avoidance of doubt , in the low energy limit @xmath104 such that @xmath105 .
these same lagrangian densities generate the following vertices : |p5.5cm| p7.5cm| vertex1sm ( 100,70 ) vertex ( e ) & @xmath249 + [ 7ex ] |p5.5cm| p8.5cm|
vertex2sm ( 100,70 ) vertex ( f ) & @xmath250 + vertex3sm ( 100,70 ) vertex ( g ) & [ cols="^ " , ] + vertex4sm ( 100,70 ) vertex ( h ) & @xmath251 + [ 7ex ] |p5cm| p5cm| dia2bis ( 120,80 ) diagram ( f ) & dia1bis ( 120,80 ) diagram ( g ) + after assuming as before that the axial current vanishes , the scattering process in diagram ( f ) describing the photon exchange between particle @xmath252 and @xmath56 produces a t - matrix element , @xmath253 which in turn yields a differential cross section contribution , @xmath254 with @xmath255 defined as the reduce mass of particles @xmath252 and @xmath56 . on the other hand ,
the schrodinger equation for a coulomb potential @xmath256 with @xmath257 and @xmath258 describing the particles measured electric charge and @xmath120 describing the distance between the two particles , yields the differential cross section @xmath259 comparing the differential cross section ( [ sqed ] ) and ( [ coulomb - abis ] ) constrains the parameters @xmath260 and @xmath261 to @xmath262 diagram ( g ) , exchanging a connecton between particle @xmath252 and @xmath56 produces the t - matrix , @xmath263 which in turn produces the differential cross section , @xmath264 on the other side , the differential cross section obtained from the schrodinger equation for a newton potential , @xmath265 where @xmath266 and @xmath267 describe the particles measured mass and @xmath120 the distance between them , produces the differential cross section , @xmath268 comparing the differential cross sections ( [ sqgd ] ) and ( [ newton - abis ] ) imposes the constraint @xmath269
dia3bis ( 160,160 ) diagram ( h ) & dia32bis ( 160,160 ) diagram ( i ) + 1.0 cm diagrams ( h ) and ( i ) produce the 1-loop correction to the connecton exchange between particles @xmath252 and @xmath56 to the t - matrix , @xmath270 after using the same steps used in section 4 , and in turn it produces the differential cross section
, @xmath271 the schroedinger equation with the einstein correction to the gravitational potential between two particles , @xmath272 yields the differential scattering cross section @xmath273 comparing the differential cross section ( [ einstein - cdbis ] ) and ( [ solve22 ] ) yields @xmath274 |p5cm| dia6bis ( 120,80 ) diagram ( j ) + the t - matrix contribution from diagram ( j ) scattering a mass particle off of a photon simplifies to @xmath275 after using the small angle approximation , the differential cross section obtained from this t - matrix equates to @xmath276 in the low energy limit , equation ( [ sab1 ] ) describes the rutherford scattering of a mass @xmath277 projectile off of a mass @xmath149 target . for a small angle deflection in general relativity the gravitational lensing of a photon by a point particle of mass @xmath277
, the differential cross section calculation follows in the same manner as in section 4.3 and resulting in @xmath278 the exchange @xmath279 transforms the differential cross section to describe the scattering of a particle of mass @xmath277 from a photon target with energy @xmath149 and as a result takes the form @xmath280 equating ( [ sab1 ] ) with ( [ sab5 ] ) requires that @xmath281 renormalization follows by construction because the structure added to the standard model is just @xmath244 standard yang - mills gauge fields with @xmath1 symmetry .
dimensional analysis of the standard model coupled to qgd shows that the connectons @xmath282 have the same dimension as the u(1 ) , su(2 ) and su(3 ) gauge fields .
therefore , all integrands in the standard model coupled to qgd , including those involving the connectons @xmath283 have the dimension four or less which guarantees renormalization of all corrections to physical quantities@xcite .
if a solution also exists for the system of equations ( [ eqpn11 ] ) , ( [ eqpn12 ] ) and ( [ eqpn13 ] ) , then the renormalizable lagrangian ( [ nov12 ] ) , ( [ nov13 ] ) and ( [ nov14 ] ) correctly describes experimental observation without resorting to the graviton and instead using @xmath1 gauge fields .
if a solution exists to the system of equations @xmath284 then , a solution also exists for the system of equations ( [ eqpn11 ] ) , ( [ eqpn12 ] ) and ( [ eqpn13 ] ) . equation ( [ eqqn11 ] ) is a stricter for of ( [ eqpn11 ] ) , ( [ eqqn12 ] ) is a stricter for of ( [ eqpn12 ] ) and ( [ eqqn13 ] ) is a stricter for of ( [ eqpn13 ] ) . while ( [ eqpn11 ] ) , ( [ eqpn12 ] ) and ( [ eqpn13 ] ) are all real , the complex system of equations ( [ eqqn11 ] ) , ( [ eqqn12 ] ) and ( [ eqqn13 ] ) require their imaginary components to vanish .
then , the solution to the real components of the system of equations ( [ eqqn11 ] ) , ( [ eqqn12 ] ) and ( [ eqqn13 ] ) also solves the system of equations ( [ eqpn11 ] ) , ( [ eqpn12 ] ) and ( [ eqpn13 ] ) .
we require the complex system of equations ( [ eqqn11 ] ) with @xmath285 , @xmath286 instead of ( [ eqpn11 ] ) because solutions to ( [ eqqn11 ] ) ensure that the @xmath80 correction to the @xmath287fermion self energy , @xmath288 does not have an imaginary component , so that @xmath289 does not affect unitarity to @xmath80 . to show existence of a solution to the system ( [ eqqn11 ] ) , ( [ eqqn12 ] ) and ( [ eqqn13 ] ) first note that ( [ eqqn12 ] ) and ( [ eqqn13 ] ) is a complex linear equation system in @xmath248 and then assume that @xmath290 .
our model contains 12 fermions with @xmath291 , so that the linear system ( [ eqqn12 ] ) has @xmath292 real equations and @xmath292 imaginary equations .
the linear system ( [ eqqn13 ] ) has 12 real and 12 imaginary equations .
we can also represent the complex linear system ( [ eqqn12 ] ) and ( [ eqqn13 ] ) instead as a real system of 180 linear equations @xmath293 where @xmath294 and @xmath295 represents a square matrix with dimension n=180 .
then , a solution to ( [ lsys1 ] ) exists whenever the @xmath296 or equivalently when the vector rows in @xmath295 are not parallel among themselves .
a simple inspection of the coefficients of @xmath297 produced by ( [ eqqn12 ] ) and ( [ eqqn13 ] ) shows that when the complex vectors @xmath298 are not parallel among themselves , the different vector rows in @xmath297 are not parallel .
a solution to the complex system of equations ( [ eqqn11 ] ) can be obtained for the set @xmath299 following these steps : 1 ) assign a real random number between 0 and 1 to each @xmath300 and @xmath301 , @xmath302 , 2 ) use the fletcher - reeves - polak - ribiere algorithm to minimize the equation @xmath303 this method produces a minimum of zero for this equation which amounts to solving ( [ eqqn11 ] ) .
note that the first squared term in ( [ lsys2 ] ) is the real part of ( [ eqqn11 ] ) while the second squared term is the imaginary part of ( [ eqqn11 ] ) .
it is not surprising that a minimum of zero exists for ( [ lsys2 ] ) ; there are @xmath304 variables and only 88 equations .
furthermore , the vectors produced by this method are not parallel and explicit calculation of @xmath297 shows that its determinant does not vanish .
then , we can choose @xmath305 effectively reducing the number of couplings to matter and the number of connectons by half . ] .
therefore , a solution can always be found to the system of 12 fermions and one u(1 ) gauge boson provided there are 180 connectons .
this does not preclude the existence of solutions whenever @xmath306 , specially after abandoning the assumption that @xmath290 .
the reader should be concerned by the number of parameters necessary to fit qgd coupled to the standard model when compared with general relativity . however
, two points should be kept in mind .
first , the intention of this section was to show the existence of a solution rather than to find a solution with the smallest possible number of couplings . given the non - linear nature of the system of equations , it is very likely that non - linear solutions with a significantly smaller number of couplings may very well exist ; for example solutions obtained through the gradient method or simulated annealing .
second , qgd should only be compared with theories that successfully incorporate gravity and more importantly renormalize .
thus , when compared with heterotic strings , which have 496 gauge fields before even considering compactification , the number of gauge fields required by qgd is similar . or , alternatively
, like string theory , the gauge fields may represent some compactification of m - theory .
nevertheless , we can cast qgd in a more symmetric manner where each of the 25 particles , whether a fermion or a gauge boson , couples to 25 different connectons ; each particle carries charge with respect to 25 non - abelian gauge fields ; each particle connects to a different particle through a unique connecton , including one which self couples .
then in this construction a picture emerges where each of the 325 connectons mediates the interaction between two particular particles only ( for example between a higgs and a @xmath307 ) .
then , the number , @xmath308 , of differential cross sections to @xmath2 is smaller than the number of couplings which , when omitting the @xmath29 s , counts to @xmath309 : the first term counts the covariant derivative couplings and the second term counts the yang mills couplings .
experimental data evidencing the gravitational interactions does not yet exist for the @xmath310 , the @xmath311 and the higgs particles or the gluons .
however , we expect for them to become available in the future .
nevertheless , qgd can incorporate gravitational interactions for all these particles through the second term in ( [ nov5 ] ) .
we present the propagators for the remainder of the bosonic sector gluonb ( 70,15 ) propagator ( g ) & , @xmath312 + wb ( 70,15 ) propagator ( h ) & , @xmath313 + higgsb ( 70,15 ) propagator ( i ) & + where the first propagator describes the gluons , the second describes the electroweak vector bosons @xmath314 and @xmath311 , and the third describes the higgs in the so(4 ) representation .
lagrangian ( [ nov5 ] ) produces the following vertices : |p7cm| p7.5cm| vertex5 ( 100,70 ) vertex ( i ) & @xmath315 + [ 7ex ] vertex6 ( 100,70 ) vertex ( j ) & , @xmath316 + [ 7ex ] vertex7 ( 100,70 ) vertex ( k ) & @xmath317 + [ 7ex ] in general we expect that gluon couplings satisfy @xmath318 ; but , due to the mass difference between the w and z particles , we expect @xmath319 and @xmath320 .
these additional vertices produce the following tree - level and @xmath80 diagrams : |p6.5cm|p6.5cm| dia6 g ( 120,80 ) diagram ( k ) + dia6w ( 120,80 ) diagram ( l ) + dia6h ( 120,80 ) diagram ( m ) + these diagrams produce the t - matrix elements , @xmath321 which in turn produce the differential cross sections , @xmath322 and @xmath323 after using the low energy limit @xmath324 .
although these differential cross sections have little difference among themselves they represent different scattering processes : the gluon does not carry mass while the @xmath325 and higgs do carry mass .
therefore , ( [ sg1 ] ) must match the small angle deflection calculated in general relativity which requires @xmath326 and whose derivation follows exactly along the lines in section 6 above . on the other hand
the w , z and higgs bosons are all massive and therefore we must compare their differential cross sections to those obtained from the schroedinger equation with a standard newtonian potential .
these constraints yield the following conditions @xmath327 |p7cm| p7cm| dia7 ( 160,160 ) diagram ( o ) & dia8 ( 160,160 ) diagram ( p ) + dia7r ( 160,160 ) diagram ( r ) & dia8q ( 160,160 ) diagram ( q ) + finally , diagrams ( o)-(q ) above give the desired correction to @xmath69 for the w , z , and higgs particles interacting with fermions : in the low energy limit , these diagrams produce the t - matrix elements , @xmath328,\\\end{aligned}\ ] ] with @xmath329 . these t - matrices in turn produce the following differential cross sections , @xmath330 comparison with the differential cross sections derived for a schrodinger with a post newtonian potential ( [ einstein - cdbis ] ) imposes @xmath331 inclusion of these differential cross sections further increases the number of differential cross section by @xmath332 for a total of 252 differential cross sections with a consequent increase in the total number of equations to be fitted .
however , a solution can be shown to exist following the method in the previous section .
we could go on and construct the @xmath80 amplitude for the scattering described by the exchange of a connecton between different bosons as well as the @xmath69 amplitude for the corresponding 1-loop corrections .
this requires modifying @xmath333 in ( [ lagtodosw ] ) to read @xmath334 the requirement that the quadratic lagrangian ( [ nov4 ] ) remain invariant requires that the constant real symmetric matrices @xmath335 satisfy @xmath336 the vertices ( h)-(k ) for bosons emitting a connecton will then have a quadratic dependence on the matrices @xmath337 .
then , diagrams for fermion - fermion scattering remain unchanged ; diagrams for fermion - boson scattering will produce tree - level t - matrices with quadratic dependence on the matrices @xmath337 ; their 1-loop corrections will produce t - matrices with quadratic and quartic dependence on the matrices @xmath337 ; diagrams for boson - boson scattering will produce tree - level t - matrices with quartic dependence on the matrices @xmath337 ; their 1-loop corrections will produce t - matrices with sextic dependence on the matrices @xmath337 .
the coefficients of these matrices will be determined in the same manner as in the previous section : equating the differential cross sections of qgd to those obtained from the schroedinger equation and gravitational lensing .
given the lack of experimental evidence , the task of writing each equation which reproduce general relativity expectations is left as a curiosity for the reader .
however we note that it may very well turn out that some processes not described by general relativity , like those analogous to s - channel bhabha scattering produced by the term @xmath338 , are confirmed through experiment or that some processes described by general relativity , like those produced by the term @xmath339 , are confirmed to be absent through experiment .
we stress that the matrices @xmath337 allow all these possible experimental scenarios .
qgd also explains the black hole entropy problem in the same way that strings do . to construct the map between qgd and strings we need first break down so(4 ) into its isomorphic representation @xmath12 .
we n=2 supersymmetrize at least one of qgd s su(2 ) gauge groups and consider for example a k3 spacetime manifold .
each su(2 ) yang mills theory then has bps states which have the same moduli space as that of a supersymmetric sigma model which in turn can be used to calculate the degeneracy of bps states for large energy states@xcite .
the degeneracy thus obtained determines the entropy which coincides with that of certain black hole solutions in string theory @xcite .
thus qgd has states whose entropy is that of certain black hole solutions , and therefore , those states represent the quantum description of the black hole solutions in the exact same way they do in string theory . the solution for @xmath340 yang - mills found in@xcite differs from the schwarzschild black hole solution because the singularity of its horizon can not be removed through a coordinate transformation .
however , both singularities can be located at @xmath341 . therefore in su(2 ) yang mills , the horizon area and the black hole mass are related in the same way as for the schwarzschild black hole : @xmath342 .
this result implies that the su(2 ) yang mills black hole has the same entropy as the schwarzschild black hole which in turn can be expressed in terms of bps solutions .
therefore one has a dual representation of the same physical objects in the @xmath13 supersymmetric limit : the macroscopic system is described by black hole solution found in@xcite , while the microscopic system is described by bps states .
qgd , as presented here , reproduces the expected newtonian and post newtonian interaction evidenced by experimental measurements for both matter fields and photon fields .
in addition to matching the differential cross sections from qgd to those of the schroedinger equation with newtonian and post newtonian corrections as well as the differential cross section for the deflection of light by a point mass particle in general relativity , qgd couples to the standard model in a straight forward and natural manner . thus
, qgd does not require us to re - invent the standard model to have a unified theory .
this contrasts other attempts at describing quantum gravity which require a recasting of the already successful standard model . the fundamental theory , a straight forward @xmath1 yang - mills , couples to matter in the same manner as su(3 ) and su(2 ) theories describe the other fundamental forces and thus bears a high degree of symmetry with the non - gravitational forces found in nature .
dimensional analysis of qgd and qgd coupled to the standard model show that all terms in the lagrangians have dimensionality four or less and therefore renormalization follows@xcite .
qgd accommodates the full particle spectrum currently known and yields in the low energy limit the expected relativistic forces provided a sufficiently large number of @xmath1 symmetries exists .
the absence in qgd or qgd coupled to the standard model of the graviton should not concern the reader ; after all , we have yet to detect such a particle .
the theory here proposed not only reproduces the observed experimental results of general relativity , exhibits renormalization and couples in a straight forward manner to the standard model , it also incorporates yet to be observed , but highly expected , gravitational interactions between matter and the su(3 ) , su(2 ) and higgs bosons as well as those between these bosons .
these features make qgd coupled to the standard model with @xmath3 local symmetry the most attractive theory to describes all interactions in nature .
i am most grateful to j. lee for providing _ ricci.m _ which was indispensable for all calculations . also indispensable was the encouragement and funding by w. israel during the early attempts to construct this model as well as funding by e. braun in its latter stages .
v. n. gribov , j. nyiri , _ quantum electrodynamics : gribov lectures on theoretical physics _ ,
volume 13 of cambridge monographs on particle physics , nuclear physics and cosmology ( cambridge university press , 2001 ) . | we show that the renormalizable @xmath0 yang mills coupled to matter and the higgs field fits all the experimentally observed differential cross sections known in nature .
this extended standard model reproduces the experimental gravitational differential cross sections without resorting to the graviton field and instead by exchanging @xmath1 gauge fields . by construction ,
each so(4 ) generator in quantum gravitodynamics does not commute with the dirac gamma matrices .
this produces additional interactions absent to non - abelian gauge fields in the standard model .
the contributions from these new terms yield differential cross sections consistent with the newtonian and post newtonian interactions derived from general relativity .
dimensional analysis of the lagrangian shows that all its terms have total dimensionality four or less and therefore that all physical quantities in the theory renormalize by finite amounts .
these properties make qgd the only renormalizable 4-dimensional theory describing gravitational interactions . |
a single polymer molecule can transition between elongated coil and compact globule states upon changing the temperature , ph of salts or addition of other polymer molecules . in general , the extended ( coil ) state of a long polymer
is stabilized when the interaction between the polymer segment and solvent is favorable , typically occurring in good solvents .
@xcite in contrast , the contracted ( globule ) state of a long polymer is stabilized when the interaction between the polymer segments is favorable , which occurs in poor solvents . by varying the solvent quality
, the coil - globule transition of a polymer can be induced .
transitions between coil and globule states can also be induced by adding solutes that can be regarded as another solvent .
@xcite when a good solvent is added to another good solvent , a coil - to - globule transition may be observed as if the long polymer is in a homogeneous poor solvent .
similarly , an unintuitive transition from globule to coil can be obtained by mixing a polymer with poor solvents .
the added solute can be short polymers .
when short polymers are added , both coil - to - globule and globule - to - coil transitions have been observed .
@xcite inspired by the complex transitions of polymers induced by addition of polymer solvents , here we study theoretically the structural transformations of a long polymer induced by polymer solvents . from another point of view , addition of polymer solvents puts a polymer in crowded environment ( macromolecular crowding ) .
coil - globule transition can be regarded as the structural transition of a long polymer caused by
macromolecular crowding " . a qualitative description of the coil
globule transition was initially formulated using mean field theory .
@xcite later , the nature of this phase transition , that is , whether it was continuous or discontinuous one , was extensively investigated using simulations , self - consistent field theory and sophisticated models that considered polymer stiffness .
@xcite however , the additional effect of short polymers on the coil globule transition of long polymers has not yet been fully studied .
part of the observed concentration dependences has been reproduced by phenomenological mean field theories , but the results were not further justified using statistical theories .
@xcite although the coil globule transition of polymers has been widely studied both theoretically and experimentally , it remains difficult to construct theories when a single molecular transition is induced by a change in the concentration of other polymers .
here we study the transition between coil and globule states of long polymers induced by addition of short polymers by a statistical theory using the partition function .
the excluded - volume effect of short polymers is taken into account as a factor that decreases the emptiness in the partition function in a mean field sense .
we also include flory s @xmath0-parameters to take into account the interactions between the long polymer , short polymers and solvents . from the extremum of the partition function
, we obtain an optimal expansion factor for a long polymer . among many possible conformations ,
the optimal ones can be determined from both the entropy in the long polymer and its mixing entropy with other polymer molecules as well as their interaction energy .
there is some ambiguity in combining the mixing entropy of polymers and elastic energy originating from entropy in a single polymer molecule . using the partition function ,
such ambiguity is removed .
we show that the same equation for the expansion factor can be derived by combining the mixing entropy of polymers and the elastic energy originating from the entropy in a single polymer molecule . in our approach
, the optimal expansion factors are obtained from the extrema of the partition function , whereas the steady - state expansion factor was obtained by imposing the osmotic pressure balance at the boundary of the region occupied by a long polymer in previous approaches .
@xcite in these theories , the region occupied by a long polymer is tacitly assumed to be in a phase different from that of the outer region for short polymers and the solvent .
furthermore , in principle , the pressure balance should include the hydrostatic pressure . in these theories ,
osmotic pressure was taken into account , but the hydrostatic pressure was not fully described . as in the previous theories ,
we ignore the hydrostatic pressure .
our results , nevertheless , reproduce the classical results obtained by virial expansion in the absence of short polymers .
various types of shape transformations of the long polymer can be obtained when the solvent acts as a good solvent for the long polymer and a poor solvent for short polymers and vice versa . these competitive interactions and
the excluded - volume interaction cause the shape transformation of the long polymer . by systematically changing flory s @xmath0-parameter between the long polymer , solvent and short polymers
, we investigate the shape transformation of the long polymer induced by adding short polymers .
this paper is organized as follows . in sec .
[ sec : partition function ] , the shape equation determining the optimal values of the expansion factor is derived from the partition function .
we also present an alternative derivation of the shape equation from the mixing entropy using the effective volume of the long polymer . in sec .
[ sec : shapeeq ] , we qualitatively examine the shape equation using an approximation .
numerical solutions of the shape equation are described in sec .
[ sec : shape transitions ] . in sec .
[ sec : equilibrium shape ] , the chemical equilibrium between inside and outside of the region occupied by a long polymer is considered .
[ sec : discussion ] presents conclusion .
in this section , we derive the shape equation of a long polymer from the partition function .
we study the coil globule transition of a long polymer with a degree of polymerization of @xmath1 in the presence of other polymers with a degree of polymerization of @xmath2 . for both types of polymers ,
we introduce the kuhn length of the monomer denoted by @xmath3 .
@xcite we consider the case where @xmath4 .
we focus on the coil
globule transition of the long polymer and do not study the shape transition of short polymers here .
the excluded volume of monomers of the long polymer is taken into account in the partition function .
the short polymers are regarded as ideal chains .
denotes the gyration radius of the long polymer , and dashed circles indicate the region occupied by the long polymer .
small red zigzags represent short polymers inside the region occupied by the long polymer .
small gray zigzags represent short polymers outside the region occupied by the long polymer . ] the region occupied by a long polymer is characterized by the gyration radius @xmath5 ( fig .
[ fig : inside ] ) .
we also introduce the natural end - to - end distance of the long polymer denoted as @xmath6 .
@xmath7 can be obtained from the extremum of @xmath8 , where @xmath9 is given by @xcite @xmath10 for an ideal gaussian chain .
the expansion factor of the chain is given by @xmath11 .
the volume fraction occupied by the long polymer inside the region within @xmath5 of the long polymer is expressed as , @xmath12 @xmath13 is expressed using the expansion ratio as @xmath14 where @xmath15 is a numerical factor independent of @xmath16 and @xmath1 .
the short polymers can move into the region occupied by the long polymer , as depicted in fig .
[ fig : inside ] .
the number of short polymers inside the region occupied by a long polymer is denoted as @xmath17 .
the volume fraction of short polymers inside the region occupied by the long polymer is denoted as @xmath18 and can be obtained from @xmath19 . an expression for the partition function in the absence of short polymers was described by di marzio .
@xcite in the presence of short polymers , the partition function can be generalized to ( see appendix [ sec : appa ] for details ) @xmath20 \left [ \prod_{j=0}^{m_c-1 } \left ( 1- m_p p_i - j \frac{\ell^3}{r^3 } \right ) \right ] \times \\
\left [ -m_c \chi_{\phi , s } \left(1-m_p
p_i -\phi \right ) - m_c \chi_{\phi , p } m_p p_i - m_p n_i\chi_{p , s } \left(1-m_p p_i - \phi \right ) \right ] , \label{eq : p_2}\end{gathered}\ ] ] where @xmath21 , @xmath22 , and @xmath23 represent flory s @xmath24parameter between the long and short polymers , the long polymer and solvents , and the short polymer and solvents , respectively .
@xcite flory s @xmath24parameters are given by the interaction energies divided by @xmath25 , where @xmath26 is the boltzmann constant and @xmath27 is temperature .
the occupancy probability is assumed to be proportional to the volume fraction of emptiness .
@xcite the factor given by @xmath28 represents the volume fraction of emptiness to place the @xmath29th segment of the long polymer when short polymers and @xmath30 segments of the long polymer are already placed .
similarly , the factor given by @xmath31 represents the emptiness to place the @xmath32th short polymer in the region occupied by the long polymer . in eq .
( [ eq : p_2 ] ) , we rewrite the term relating the occupancy probability as @xmath33 \\ & = \left(\frac{\ell}{r}\right)^{3 m_c } \frac{\gamma\left[\frac{r^3}{\ell^3}\left ( 1- m_p p_i\right)+1\right]}{\gamma\left[\frac{r^3}{\ell^3}\left ( 1- m_p p_i\right)-m_c+1\right ] } , \label{eq : p_3}\end{aligned}\ ] ] where @xmath34 is the gamma function . using stirling s formula expressed by @xmath35 , @xcite
we obtain @xcite @xmath36 - m_c .
\label{eq : p_5}\end{aligned}\ ] ] in the same way , we rewrite @xmath37 \\ & = \left(\frac{m_p \ell^3}{r^3}\right)^{n_i } \frac{\gamma\left[\frac{r^3}{m_p \ell^3}+1\right]}{\gamma\left[\frac{r^3}{m_p \ell^3}-n_i+1\right ] } .
\label{eq : p_5_1}\end{aligned}\ ] ] using stirling s formula , we obtain @xmath38 from eq .
( [ eq : p_5_2 ] ) , we find @xmath39 we also applied stirling s formula to the rest of the factors in eq .
( [ eq : p_2 ] ) to obtain @xmath40 \approx - \left(\frac{r}{\ell}\right)^{3 } p_i\ln \left ( p_i \right ) + n_i . \label{eq : p_5_4}\end{aligned}\ ] ] the optimal shape of the long polymer can be obtained by maximizing @xmath41 with respect to @xmath16 .
we note that the volume fraction occupied by short polymers can be varied even if @xmath17 is not altered when @xmath5 is changed . in this sense
, @xmath18 depends on the volume fraction of the long polymer , @xmath42 .
because @xmath43 depends on @xmath16 through @xmath42 , it is convenient to define a new variable , @xmath44 here , @xmath17 can be regarded as a constant for the infinitesimal small variation of @xmath5 . in this work
, we consider the small variation of @xmath5 and regard @xmath45 as being independent of @xmath16 . by taking into account eqs .
( [ eq : p_5])-([eq : r ] ) , we obtain @xmath46 , \label{eq : p_6}\end{gathered}\ ] ] where the effective @xmath0-parameter is defined by @xmath47 and @xmath48 is defined by @xmath49 the optimal shape of the long polymer can be obtained by maximizing eq .
( [ eq : p_6 ] ) with respect to @xmath16 . by calculating @xmath50(d \ln w(r))/(d \alpha)=0 $ ] using eq .
( [ eq : p_6 ] ) , we find @xmath51=0 .
\label{eq : p_7}\end{aligned}\ ] ] equation ( [ eq : p_7 ] ) was derived previously when @xmath52 using the partition function except for a minor difference associated with the definition of @xmath16 .
@xcite by further expanding eq .
( [ eq : p_7 ] ) in terms of @xmath53 , we obtain @xmath54- \frac{s^2}{2\alpha^3}\left(1+r\right)^3=0 .
\label{eq : p_8}\end{aligned}\ ] ] for simplicity , we have expressed the volume of the polymer segment as @xmath55 .
in fact , the volume of a polymer segment can be given by @xmath56 using the diameter of the polymer @xmath57 .
in this case , @xmath58 is proportional to @xmath59 .
below , we regard @xmath58 as a parameter that characterizes the anisotropy of a polymer segment .
we further show that the partition function given by eq .
( [ eq : p_6 ] ) can be derived by combining the mixing entropy and entropic elastic energy of the long polymer when the mixing entropy is multiplied by the properly defined volume occupied by the long polymer . using the expansion factor @xmath16 , the entropic elastic energy of the long polymer
is given by , @xcite @xmath60 we consider configurations of polymers using a lattice model . in the mean field approximation ,
the mixing entropy at a site inside the long polymer is given by , @xcite @xmath61 .
\label{eq : s}\end{aligned}\ ] ] in the absence of the elastic energy given by eq , ( [ eq : el ] ) , the free energy at a site inside the long polymer can be written as , @xcite @xmath62 the entropic elastic energy for a long polymer is given in eq .
( [ eq : el ] ) and the free energy per site is given by eq .
( [ eq : fin ] ) . to add these terms
, we should multiply the free energy of a site by a certain volume .
the real polymer volume can be approximated by @xmath63 .
we assumed that the effective volume occupied by a long polymer is @xmath64 . by comparing the site free energy given by eq .
( [ eq : fin ] ) and the probability distribution given by eq .
( [ eq : p_6 ] ) using the polymer volume given by @xmath64 , @xcite the total free energy @xmath65 including the elastic energy given by eq .
( [ eq : el ] ) can be expressed as , @xmath66 + \frac{e_{\rm el}}{k_{\rm b } t}. \label{eq : tenergy}\end{gathered}\ ] ] equation ( [ eq : tenergy ] ) is equal to @xmath67 obtained from eq .
( [ eq : p_6 ] ) except for the unimportant term @xmath68 . in the limit of @xmath69 ,
( [ eq : tenergy ] ) is simplified to the equation obtained before . @xcite the shape equation given by eq .
( [ eq : p_7 ] ) can be obtained by minimizing eq .
( [ eq : tenergy ] ) with respect to @xmath16 .
previously , the equation representing the osmotic pressure balance was derived from eq .
( [ eq : tenergy ] ) and the osmotic pressure balance was imposed by regarding the region occupied by the long polymer as being in a phase different from that of the other region .
@xcite we did not impose the osmotic pressure balance at the boundary of the region occupied by the long polymer .
instead , we derive the optimal shape by minimization of eq .
( [ eq : tenergy ] ) with respect to the expansion factor . in the absence of short polymers
, the resultant shape equation simplifies to the conventional form .
first , we study the shape equation derived from the partition function analytically . for convenience ,
we express eq .
( [ eq : p_8 ] ) as @xmath70 where parameters are given by , @xmath71 and @xmath72 from eq .
( [ eq : r ] ) . when @xmath52 , eq .
( [ eq : minimaa ] ) is simplified to the same form of equation as that derived using the virial expansion .
the term proportional to @xmath73 and @xmath74 in eq .
( [ eq : minimaa ] ) represents the results of excluded - volume interactions , and that proportional to @xmath75 represents the results of third - body interactions .
@xcite below , we summarize known results for eq .
( [ eq : minimaa ] ) when @xmath76 and generalize them for @xmath77 .
the results can be used when we study the coil
globule transition by solving eq .
( [ eq : p_7 ] ) numerically .
extended states can be characterized by @xmath78 . if @xmath79 is positive , extended states are stabilized according to eq .
( [ eq : minimaa ] ) .
when @xmath80 , the extended states obey the scaling form @xmath81 . by substituting @xmath82 and @xmath83 into the scaling form , we obtain the scaling relation for @xmath5 in good solvents , @xcite @xmath84 when @xmath85 . if @xmath86 , @xmath87 can be derived , which represents the same scaling form of the free polymers without taking into account the excluded - volume interactions .
compact states are characterized by @xmath88 .
when @xmath89 , the compact states obey scaling form , @xmath90 . the scaling relation for @xmath5
is obtained as @xcite @xmath91 when @xmath92 .
the scaling of eq .
( [ eq : gsolvents ] ) represents the reasonable assumption that the molecular weight of the long polymer is proportional to its volume given by @xmath93 in the compact state . for the values other than @xmath94 and @xmath88 , we studied eq .
( [ eq : minimaa ] ) analytically by assuming that @xmath79 can take any value and vary independently of @xmath95 .
the results can be used as a guide for numerical solutions of eq .
( [ eq : p_7 ] ) , where we take into account proper constraints on @xmath96 and @xmath95 , in particular the fact that they are not independent functions of the concentration of short polymers @xmath43 .
it is convenient to define @xmath97 equation ( [ eq : minimaa ] ) can be rewritten using @xmath98 as @xmath99 if @xmath98 has a maximum and a minimum in the region of @xmath100 , the coexistence of coil and globule states is possible depending on the value of @xmath101
. the above condition can be restated as @xmath102 has two solutions for @xmath100 .
the explicit condition can be obtained by defining @xmath103 , which satisfies both @xmath104 and @xmath105 for @xmath106 .
using @xmath107 , @xcite the condition is given by @xmath108 , where @xmath109 therefore , the coexistence of coil and globule states may be possible when @xmath108 depending on the right - hand side of eq .
( [ eq : gtau ] ) .
we should also note that @xmath110 can be transformed into @xmath111 .
if @xmath112 , _ i.e. _ , @xmath113 , @xmath110 is always satisfied because @xmath95 is positive according to eq .
( [ eq : y ] ) .
in addition , @xmath92 is required if the solution of @xmath114 satisfies @xmath112 because @xmath115 and @xmath116 .
the coexistence of coil and globule states implies that we have two solutions and one of them should satisfy @xmath117 .
therefore , coexisting states can be obtained when @xmath92 .
we next present numerical results of the shape transitions of the long polymer as a function of @xmath45 .
@xmath45 is proportional to the number of short polymers inside the region occupied by the long polymer @xmath17 . as a function of @xmath45 .
@xmath45 is proportional to the number of short polymers inside the region occupied by the long polymer .
( a ) @xmath118 , @xmath119 , and @xmath120 ( thick solid line ) .
( b ) @xmath121 , @xmath122 , and @xmath123 ( thin solid line ) .
( c ) @xmath124 , @xmath125 , and @xmath126 ( dashed line ) . ]
we first estimate appropriate parameters to discuss the shape transition by considering @xmath101 as a function of @xmath45 ( fig .
[ fig : tau_cg ] ) .
in general , the long polymer is in the elongated state if @xmath127 and in the compact state if @xmath92 , and the sign of @xmath101 can be changed by changing @xmath45 . by using eq .
( [ eq : chieff ] ) , eq .
( [ eq : tau ] ) can be rewritten as @xmath128 at @xmath52 , the elongated coil state can be obtained when @xmath129 , and the compact globule state can be obtained when @xmath130 . as for @xmath23 , when @xmath131 , @xmath101 becomes negative upon increasing @xmath45 , as can be understood from eq .
( [ eq : tau1 ] ) .
in contrast , when @xmath132 , @xmath101 becomes positive upon increasing @xmath45 . based on the above reasoning , if @xmath129 and @xmath131 ( fig . [
fig : tau_cg ] ( a ) ) , the coil - to - globule transition can be obtained , and if @xmath130 and @xmath132 ( fig .
[ fig : tau_cg ] ( b ) ) , the globule - to - coil transition occurs .
figure [ fig : tau_cg ] ( c ) presents the results when @xmath133 and @xmath134 . in this case
, @xmath101 slowly increases as @xmath45 is increased .
below , we assume that @xmath135 , @xmath136 , and @xmath137 . as a function of @xmath45
when @xmath118 , @xmath119 , and @xmath120 .
here , @xmath45 is proportional to the number of short polymers inside the region occupied by the long polymer .
the red dashed lines were obtained using eq .
( [ eq : p_7 ] ) and the solid lines were obtained using the approximate expression given by eq .
( [ eq : minimaa ] ) using @xmath138 , @xmath136 , and @xmath137 . ] figure [ fig : cg_1 ] shows the expansion factor @xmath16 as a function of @xmath45 with @xmath131 and @xmath139 where the coil - to - globule transition can be expected . in the absence of short polymers ( @xmath52 ) ,
the long polymer extends in the good solvent characterized by @xmath139 .
the short polymers tend to aggregate in the poor solvent characterized by @xmath131 . when there are excluded - volume interactions between the long and short polymers , the long polymer may shrink if the short polymers tend to aggregate and the motion of short polymers is hindered by the presence of the long polymer .
therefore , the globule state can be obtained when @xmath45 exceeds a certain value .
figure [ fig : cg_1 ] shows the coil - to - globule transition .
@xmath16 has two different values that correspond to the coil and globule states when @xmath45 exceeds a certain value , and a single value ( @xmath140 ) when @xmath45 is less than a certain value .
such a discontinuous transition suggests that it is a first - order transition and the compact globule state can be found when @xmath45 exceeds a certain value . incidentally , in fig .
[ fig : cg_1 ] , the dashed line was obtained using the original equation given by eq .
( [ eq : p_7 ] ) and the solid line was obtained using the approximate expression given by eq .
( [ eq : minimaa ] ) .
we note that the difference between these lines is small and the qualitative @xmath45 dependence is well captured by eq .
( [ eq : minimaa ] ) . as a function of @xmath45
when @xmath121 , @xmath122 , and @xmath123 .
@xmath45 is proportional to the number of short polymers inside the region occupied by the long polymer .
the red dashed lines were obtained using eq .
( [ eq : p_7 ] ) and the solid lines were obtained using the approximate expression given by eq .
( [ eq : minimaa ] ) using @xmath138 , @xmath136 , and @xmath137 . ]
figure [ fig : gc_1 ] shows the results for @xmath132 and @xmath141 where the globule - to - coil transition can be expected . in the absence of short polymers ( @xmath52 ) , the long polymer is in a compact state in a poor solvent characterized by @xmath130 .
in contrast , the short polymers are in good solvents .
if a sufficient number of short polymers tend to be dispersed and involved in excluded - volume interactions with the long polymer , the extended coil state is expected .
figure [ fig : gc_1 ] clearly shows that the long polymer changes from the compact globule state to the extended coil one as @xmath45 increases .
the transition is discontinuous .
it should be noted that the extended coil state can be expected if a sufficient number of short polymers are involved in excluded - volume interactions with the long polymer regardless of the values of @xmath23 and @xmath21 .
indeed , we obtain a similar figure even when the short polymers can be regarded as a poor solvent for the long polymer characterized by @xmath142 if the other values in fig .
[ fig : gc_1 ] are fixed . the discontinuous transition to the extended coil state
is also obtained when @xmath23 is changed to @xmath143 and the other values in fig .
[ fig : gc_1 ] are fixed . in the latter two cases ,
the transition to the extended state is mainly caused by the excluded - volume interactions between the long and short polymers . as a function of @xmath45 when @xmath124 , @xmath125 , and @xmath126 .
@xmath45 is proportional to the number of short polymers inside the region occupied by the long polymer .
the red dashed line was obtained using eq .
( [ eq : p_7 ] ) and the solid line was obtained using the approximate expression given by eq .
( [ eq : minimaa ] ) using @xmath138 , @xmath136 , and @xmath137 . ]
figure [ fig : cg_2nd ] shows the results for @xmath124 , @xmath144 and @xmath126 .
the extended coil state expands gradually and the discontinuous transition is not obtained . in experiments , coil - to - globule transitions of dna
were observed when the solvent acted as a good solvent for both long dna and short polyethylene glycol ( peg ) .
@xcite although a coil - to - globule transition can be obtained according to our theoretical approach when the solvent acts as a good solvent for the long dna , the condition for short polymers is different . by including the translational entropy of small ions ,
experimental results have been reproduced using mean field theory .
@xcite although such an ion effect is ignored in our theory , we systematically varied flory s @xmath0-parameters and obtained the general conditions for transitions between coil and globule states for neutral polymers . in some experiments , a coil
coil transition was observed .
@xcite further study is needed to understand theoretically the globule - to - coil reentrance transition .
here we discuss a way to determine the number of short polymers @xmath17 inside the region occupied by the long polymer .
we consider the situation where short polymers can enter the region occupied by the long polymer and the concentration of short polymers inside the region occupied by the long polymer is different from that in the region outside the long polymer .
the concentrations of short polymers outside and inside the region occupied by the long polymer can be expressed as @xmath145 and @xmath146 , respectively .
the relationship between the concentrations of short polymers in these regions can be determined by imposing the equality of chemical potentials .
the free energy at a site outside the region occupied by the long polymer can be given by , @xcite @xmath147 in the mean field approximation , where @xmath148 denotes the concentration of the long polymers outside the region occupied by a single long polymer .
when short polymers can move across the boundary of the region occupied by a long polymer , the chemical potential outside the region containing the long polymer , @xmath149 , should be equal to that inside the area with the long polymer , @xmath150 , which was determined using eqs .
( [ eq : s])-([eq : fin ] ) .
the result can be written as , @xmath151 = \delta \chi m_p\left ( \phi - m_c c \right)+ 2 \chi_{p , s } m_p^2 \left(p - p_i \right ) , \label{eq : chemeq } \end{aligned}\ ] ] where @xmath48 is defined in eq .
( [ eq : dchi ] ) . equation ( [ eq : chemeq ] ) can be expressed as @xmath152 = \\
\mbox { } \frac{m_p p_i } { \left ( 1 - \phi - m_p p_i \right)^{m_p } } \exp \left [ \delta \chi m_p \phi - 2 \chi_{p , s } m_p^2 p_i \right ] .
\label{eq : chemeqphi } \end{gathered}\ ] ] @xmath148 and @xmath153 for the regions outside and @xmath42 and @xmath43 for those inside the area occupied by a specific long polymer are related by eq .
( [ eq : chemeqphi ] ) .
it is more convenient to rewrite eq .
( [ eq : chemeqphi ] ) in terms of @xmath45 and @xmath16 , where @xmath43 is replaced by @xmath45 using eq .
( [ eq : r ] ) , @xmath154 . equation ( [ eq : chemeqphi ] ) can therefore be expressed as @xmath155}{\left(1-(1+r)s/\left(\alpha^3 \sqrt{m_c } \right ) \right)^{m_p } } = \frac{m_p p\exp\left(-2 \chi_{p , s } m_p^2 p \right)}{\left(1- m_p p -m_c c \right)^{m_p } } .
\label{eq : chemeq2_1}\end{aligned}\ ] ] the optimal expansion factor , @xmath16 , can be determined from eq .
( [ eq : p_7 ] ) and eq .
( [ eq : chemeq2_1 ] ) . in fig .
[ fig : gc_p_s ] , we show @xmath148 as a function of @xmath153 for the parameters used to draw fig . [
fig : cg_1 ] .
the results indicate that the same values of @xmath16 and @xmath45 can be obtained if @xmath148 and @xmath153 satisfy the relationship represented by the line in fig .
[ fig : gc_p_s ] . as a function of @xmath153 obtained using eq .
( [ eq : chemeq2_1 ] ) when @xmath156 and @xmath157 , with @xmath118 , @xmath119 , @xmath120 , @xmath138 , @xmath136 , and @xmath137 . ]
we theoretically investigated the excluded - volume and interaction effects of added polymer solvents on the transition between the coil and globule states of a long polymer .
the excluded - volume effect was taken into account by generalizing the partition function of the long polymer to include the excluded - volume effect of short polymers .
the interactions between the segments of short polymers , those of long polymer and solvents were described using flory s @xmath0-parameters as in other theories . following this approach
, we obtained the optimal expansion factors denoted by @xmath16 from the maxima of the partition function . to be more precise ,
the partition function was maximized by assuming that the number of short polymers @xmath17 inside the region occupied by the long polymer was conserved against the infinitesimal change in the gyration radius @xmath5 of the long polymer .
the shape equation was given in terms of the effective flory s
@xmath0-parameter denoted by @xmath158 .
@xmath158 was expressed by @xmath17 as well as flory s intrinsic @xmath0-parameters .
the shape equation was simplified using asymptotic expansion in terms of @xmath16 . in the absence of short polymers
, the simplified equation reduces to the classical expression derived by the virial expansion .
the simplified shape equation captured the main features of the nonlinear shape equation and thus is very useful to analytically investigate the shape transition of polymers .
previously , the shape transition had been extensively studied using the simplified equation obtained by the virial expansion in the absence of short polymers .
@xcite guided by this simplified equation , we systematically studied the shape transition of a long polymer induced by the presence of short polymers . using the simplified shape equation and the effective @xmath0-parameter @xmath158
, we showed that coil - to - globule transitions can be induced by increasing @xmath17 when the solvent acts as a good solvent for the long polymer and a poor solvent for the short polymers .
the long polymer tends to extend in a good solvent and short polymers tend to aggregate in a poor solvent . under the competitive effect induced by such a solvent , the long polymer shrinks if the long polymer segments are involved in excluded - volume interactions with a sufficient number of short polymers .
these transitions were confirmed by numerical calculations using the complete shape equation .
in contrast , globule - to - coil transitions can be induced by increasing @xmath17 when the solvent acts as a poor solvent for the long polymer .
this is because long polymer shrinks in the poor solvent , but expands through the excluded - volume interaction between the long polymer segments and short polymers .
the globule - to - coil transition occurs at a lower concentration of short polymers when the solvent acts as a good solvent for the short polymers compared to that in the case where the solvent acts as a poor solvent for the short polymers .
the globule - to - coil transition of a long polymer results from excluded - volume interaction with a sufficient number of short polymers and the threshold value for the number of short polymers can be low when the short polymers tend to be well dispersed in the good solvent . the excluded - volume effects can be also taken into account by considering the site - mixing entropy of the lattice model .
however , there was some ambiguity in defining the effective volume of a long polymer to combine the site - mixing entropy and entropic elastic energy of a long polymer .
we show that the minimization of the total free energy leads to the equation for the optimal expansion factor derived using the partition function if the mean volume is defined by @xmath159 , where @xmath3 is the kuhn length and @xmath5 denotes the radius of gyration of the long polymer . in a related topic ,
effects of macromolecular crowding on protein folding has been studied both experimentally and theoretically @xcite .
protein folding stability was studied using a reduction factor for chain conformations by crowding objects .
@xcite the reduction factor for folding states was obtained from a scaled - particle theory.@xcite the reduction factor for unfolding states was expressed using the unoccupied volume to place the protein and the fraction of the reduced protein conformations due to prohibiting overlapping with crowding objects . by calculating the folding free energy using a reduction factor , the folding stability was discussed . although the method of calculation is different , entropy reduction by excluded volume effect was taken into account both in the present theory and in the theory of protein folding stability . in this sense
, this study gives another insight into the effect of macromolecular crowding using a single polymer chain .
in this paper , we have implicitly assumed that the long polymer is flexible .
flexibility of conjugated polymers can be increased by formation of defects .
although conjugated polymers are stiff , various compact shapes of conjugated polymers were simulated by introducing defects .
@xcite it has been pointed out that the folding characteristics of a long semiflexible chain differ from those of a short stiff chain .
@xcite recently , it was demonstrated that a polymer of intermediate length formed various types of collapsed states , while the spherical collapsed state was obtained for long polymers .
@xcite spherical symmetric globules can be also classified into several states .
@xcite it was also shown that combined effect of rigidity and hydrophobicity is correlated with the folding kinetics and structures.@xcite to identify the shape of a long polymer , its stiffness , chemical interactions and finite length should be explicitly considered .
although such complexity was ignored in this paper , the competitive and cooperative effects of excluded volume and the interaction between monomers and solvents were taken into account in this work using the partition function in the mean field approximation . by maximizing the partition function
, we showed that various kinds of transition between the coil and globule states are possible by adding two different solvents , even if one of the solvents is short polymers .
in particular , we qualitatively show that the shape transition can be induced by the excluded volume interactions if solvent acts as good ( poor ) solvent for a long polymer and poor ( good ) solvent for short macromolecules .
the mechanism will apply for real polymers having stiffness and other chemical interactions .
ko was financially supported by a kenkyu - josei grant from senshu university .
if the excluded volume interactions are ignored , the total number of configurations of a long chain is proportional to @xmath160 . in the presence of excluded volume interactions ,
the number of configurations is influenced by the presence of short polymers .
we first consider the possible ways to place short polymers under excluded volume effect .
when the first segment of the @xmath32th short polymer is placed inside the region occupied by the long polymer , the number of possible ways to place the first segment is proportional to the free volume @xmath161 , where @xmath162 denotes the total segments previously placed in the region .
the second segment of the @xmath32th short polymer can be placed to the nearest void if the space is allowed .
when @xmath163 short polymers have been previously placed randomly inside the area occupied by the long polymer , the probability of finding a sufficient size of nearest void is given by @xmath164 in the mean field approximation .
the probability of adding @xmath165 segments of the @xmath32th short polymer after placing the first segment can be given by @xmath166^{m_p-1}$ ] . in this way , the number of distinct configurations to place all @xmath17 short polymers inside the region occupied by a long polymer is proportional to the factor @xmath167 , \label{eq : app_1}\end{aligned}\ ] ] where @xmath168 is multiplied to account for @xmath169 ways of ordering indistinguishable short polymers .
now , we consider the excluded volume effect of the long polymer segments under the presence of short polymers .
for each configuration whose incidence is proportional to @xmath160 , it requires that all @xmath1 successive segments of the long polymer can be placed in a void .
the probability of appearance of such a void can be estimated as follows .
when short polymers and @xmath30 segments of the long polymer are already placed , the next segment of the long polymer can be placed in the nearest void and the probability of finding the void is given by the volume fraction expressed by @xmath170 in the mean field approximation .
the probability of placing @xmath1 segments of the long polymer in a void can be given by @xmath171 $ ] . by taking into account all the above factors , we obtain @xmath167 \left [ \prod_{j=0}^{m_c-1 } \left ( 1- m_p p_i - j \frac{\ell^3}{r^3 } \right ) \right ] .\end{aligned}\ ] ] by also taking into account the interaction energies and expansion potential of the long polymer , we obtain eq .
( [ eq : p_2 ] ) .
v. v. vasilevskaya , a. r. khokhlov , y. matsuzawa and k. yoshikawa , j. chem . phys . * 102 * , 6595 ( 1995 ) .
y. t. maeda , t. tlusty , and a. libchaber , proc .
usa , * 109 * , 17972 ( 2012 ) .
m. kojima , k. kubo , and k. yoshikawa , j. chem .
* 124 * , 024902 ( 2006 ) .
p. j. flory , _ principles of polymer chemistry _ ( cornell university press , ithaca , 1953 ) .
i. c. sanchez , macromolecules * 12 * , 980 ( 1979 ) . g. ziv , d. thirumalai , and g. haran , phys .
* 11 * 83 ( 2009 ) .
f.tanaka , _ polymer physics - applications to molecular association and thermoreversible gelation _ ( cambridge university press , uk , 2011 ) .
zhou , g. rivas , and a. p. minton , annu .
biophys . * 37 * , 375 ( 2008 ) .
zhou , j. mol .
* 17 * , 368 ( 2004 ) .
zhou , arch .
biophys . * 469 * , 76 ( 2008 ) .
b. van den berg , r. j. ellis , and c. m. dobson , embo .
j. * 10 * , 3870 ( 2000 ) .
r. j. ellis , trends biochem .
* 26 * , 597 ( 2001 ) .
f. h. stillinger , j. solution chem .
* 2 * , 141 ( 1973 ) .
lebowitz , j. s. rowlinson , j. chem .
phys . * 41 * , 133 ( 1964 ) .
b. bagchi , _ water in biological and chemical processes : from structure and dynamics to function _
( cambridge university press , cambridge , 2014 ) .
d. hu , j. yu , k. wong , b. bagchi , p. j. roskky , and p. f. barbara , nature * 405 * , 1030 ( 2000 ) . t. sakaue and k. yoshikawa , j. chem .
phys . * 117 * , 6323 ( 2002 ) .
s. chakrabarty and b. bagchi , j. phys .
b * 113 * , 84 ( 2009 ) . | polymers adopt extended coil and compact globule states according to the balance between entropy and interaction energies .
the transition of a polymer between an extended coil state and compact globule state can be induced by changing thermodynamic force such as temperature to alter the energy / entropy balance .
previously , this transition was theoretically studied by taking into account the excluded - volume interaction between monomers of a polymer chain using the partition function . for binary mixtures of a long polymer and short polymers ,
the coil - globule transition can be induced by changing the concentration of the shorter polymers . here
we investigate the transition caused by short polymers by generalizing the partition function of the long polymer to include the excluded - volume effect of short polymers .
the coil - globule transition is studied as a function of the concentration of mixed polymers by systematically varying flory s @xmath0-parameters .
we show that the transition is caused by the interplay between the excluded - volume interaction and the dispersion state of short polymers in the solvent .
we also reveal that the same results can be obtained by combining the mixing entropy and elastic energy if the volume of a long polymer is properly defined . |
developing a deeper understanding of active matter @xcite has been the major focus of a considerable amount of theoretical work over the last decades @xcite . in recent years
, kinetic theory has gained considerable popularity to assess the ordering behavior in systems of actively propelled particles @xcite . in this context , the boltzmann equation provides a particularly compelling approach to active matter systems .
apart from its inherent limitations , which are largely due the assumptions of binary particle interactions and molecular chaos ( cf . refs .
@xcite ) , the following advantages of this framework are manifest : ( i ) the structure of the boltzmann equation , relating convection and collision processes on the level of the one - particle distribution function , is ideally suited to explicitly implement a microscopic picture of particle dynamics .
( ii ) due to its mesoscopic character , the boltzmann equation provides an immediate connection to the system s hydrodynamic variables , which naturally arise in the form of the various moments of the one - particle distribution function .
therefore , the boltzmann equation sets up a direct link between the microscopic dynamics of the system s constituent particles and the corresponding physics emerging on hydrodynamic length and time scales .
moreover , since the boltzmann equation keeps track of all parameters used to formulate a particle - level description of the system under study , the resulting hydrodynamics can be studied explicitly in terms of those microscopic parameters .
the boltzmann equation and related kinetic frameworks have been employed in a number of previous studies on two - dimensional systems of self - propelled particles @xcite , mainly to derive a hydrodynamic description . to this end
, an expansion technique is used to derive an infinite hierarchy of coupled equations of motion for the fourier modes of the angular degrees of freedom of the one - particle distribution function . since the zeroth , first , second , and higher order fourier modes are directly related to the local average particle number , polar , nematic , and higher symmetry order parameters , the fourier space description lends itself as a starting point to extract a hydrodynamic description . in general , this is achieved by truncating the hierarchy of fourier space equations by means of a perturbation ansatz for the fourier modes , which itself typically depends on system s symmetries . following a similar strategy , a numerical approach based on an enskog - like kinetic theory has recently been proposed in ref .
@xcite .
while such fourier space approaches have been successfully applied to investigate the onset of collective motion , the validity of the underlying truncation schemes remains largely elusive for parameters deeper inside the ordered region .
in this work , we propose a boltzmann equation based approach which we will refer to as _ snake _ ( olving umerically ctive inetic quations ) . _
_ focuses on a direct numerical solution of the boltzmann equation in real space , without any explicit recourse to the aforementioned coupled hierarchy of equations in fourier space .
such a direct numerical approach offers a number of advantages in the study of active systems , which can be summarized as follows .
first of all , the real space boltzmann equation constitutes a closed description for the dynamics of the one - particle distribution function .
consequently , the range of validity of this direct numerical strategy is restricted only by inherent limitations of the boltzmann equation itself but is not principally confined to a parameter region around the onset of collective motion .
secondly , _ snake
_ provides full access to the flexibility of the boltzmann equation in the implementation of specific model systems which are based on a concrete microscopic picture of particle dynamics .
this , thirdly , equips this approach with particularly powerful capabilities to incorporate physical boundary conditions , which can be formulated on the basis of actual particle - wall interactions rather than on the level of macroscopic , hydrodynamic field variables .
finally , since all pertinent hydrodynamic fields are easily computed from the one - particle distribution function by taking appropriate averages , this numerical approach allows to fully exploit the boltzmann equation s potential to mediate between the physics on microscopic and macroscopic scales . to demonstrate the benefits of _ snake
_ , we will present a detailed study of an archetypical model for active particles subject to binary , polar particle interactions , which has been proposed originally by bertin et al .
@xcite and which we will henceforth refer to as `` bdg model '' . as has been established previously @xcite , the bdg model transitions from a spatially homogeneous , fully isotropic state at low densities / high noise levels to a collectively moving state at high densities / low noise levels . in the vicinity of this phase transition ,
the polar order parameter field was shown to be susceptible to a longitudinal instability @xcite , which in turn has been linked to the emergence of stable solitary wave patterns propagating on an otherwise isotropic low density background @xcite .
well beyond the transition line to collective motion , this longitudinal instability disappears from the boltzmann equation @xcite and a spatially homogeneous state of collective motion has been shown to emerge @xcite .
our work is structured as follows . in section [ sec : discretization ] , we discuss a discretization scheme for the boltzmann equation of self - propelled particles in two spatial dimensions . the central result of this section is an update rule , eq . , which lends itself to direct numerical implementation
this section also contains a discussion of the general strategy to implement virtually arbitrary boundary conditions and geometries .
readers not interested in the details of the actual discretization procedure underlying the _ snake _ algorithm may skip sections [ sec : dimlessbe ] [ sec : bcs ] and proceed to section [ sec : applications ] . there
, we will exploit _ snake _
s capabilities to address the stationary states of the bdg model over a broad range in parameter space ( section [ sec : phasediagram ] ) .
in particular , we will show that our implementation of _ snake _ consistently reproduces previously established analytical results on this model @xcite .
in section [ sec : ld_transition ] we focus on the onset of collective motion , which is accompanied by the formation of density segregated patterns and hysteresis effects , as previously reported in numerical studies of the vicsek model @xcite .
additionally , in section [ sec : dsp ] , we extend previous results on the flocking transition in polar systems @xcite well beyond the onset of order .
specifically , we demonstrate the presence of an additional first order phase transition from a spatially inhomogeneous , density segregated phase to a spatially homogeneous polar phase , taking place well inside the ordered , broken - symmetry parameter regime ( section [ sec : hd_transition ] ) .
overall , we find striking similarities between the flocking transition to spatially homogeneous polar order , on the one hand , and the features of liquid - gas phase transitions in equilibrium systems , on the other hand .
analogous observations have been made previously in the context of a lattice model of active ising spins , where a discrete up - down symmetry is spontaneously broken in two spatial dimensions @xcite . in section [ sec : cluster ] , we shift our focus to a deeper discussion of density segregated patterns and report on the emergence of previously unseen `` cluster - lane patterns '' which seem to occupy the same parameter region as the familiar solitary wave patterns and to constitute a stable limit - cycle solution of the underlying boltzmann equation .
these patterns consist of parallel lanes of polar clusters moving in opposite directions ; cf . figs .
[ fig : cluster_particle_currents](b , c ) and figs .
[ fig : cluster_lane_geometry](b , c ) for snapshots .
intriguingly , the stability of such patterns provides a mechanism to establish states of nematic macroscopic order , arising from perfectly polar particle interactions on microscopic scales .
finally , in section [ sec : coasening ] , we investigate the coarsening dynamics for both , wave and cluster lane patterns within the density segregated phase .
section [ sec : outlook ] , then summarizes our numerical approach , discusses our main findings , and gives an outlook of potential applications of _ snake_. in particular , this section gives a comprehensive and self - contained discussion of the most pertinent results established in the more technical sections [ sec : discretization ] and [ sec : applications ] , and can be read independently of these technical parts .
section [ sec : outlook ] has been written to provide the reader with a direct and non - technical accessibility to the core results of this work .
here and in the following , we focus on a kinetic description for two - dimensional systems of actively propelled particles in an over - damped environment . as a simplifying assumption , we consider particles moving at constant speed @xmath0 but variable orientation @xmath1 , such that each particle s velocity vector @xmath2 can be written @xmath3 ( @xmath4 : unit vector ) . for particles with nearly constant propelling forces , this is a reasonable approximation in the `` aristotlean regime '' where the dynamics is governed by a balance between propelling and dissipative forces and inertial effects can be neglected .
moreover , for sufficiently dilute systems , binary particle interactions are believed to dominate over higher order interactions and the spatio - temporal dynamics of the one - particle distribution function @xmath5 can be captured by means of the boltzmann equation .
the most general form of the boltzmann equation reads [ eq : boltzmann ] @xmath6.\ ] ] the density of particles in the phase space element @xmath7\times[\theta,\,\theta+d\theta]$ ] is being convected due the particles active propulsion ( left hand side ) , and is , simultaneously , subject to sudden changes due to particle collisions , chemical reactions , etc . which is captured on the right hand side of eq .
in the form of source terms @xmath8 . to make contact to previous works on the boltzmann equation for active systems @xcite , we consider two separate contributions to the source term , @xmath9 = i_{\text{sd}}\left[f;\,\theta\right ] + i_{\text{c}}\left[f;\,\theta\right],\ ] ] which take into account rotational diffusion due to a fluctuating background and / or noisy propelling mechanism , @xmath10 , and binary particle collisions , @xmath11 , respectively . to be specific , rotational diffusion is implemented using a constant `` tumbling '' rate @xmath12 , and drawing the corresponding angular shifts from a given probability distribution @xmath13 .
the self - diffusion integral @xmath10 then takes the following form ( omitting any explicit reference to space and time coordinates for brevity ) : @xmath14 = \\ &
-\lambda\,f(\theta ) + \lambda\int_{-\pi}^{\pi}[d\phi]_{(m)}\,f(\phi ) \ ,
p_0(\theta-\phi+2m\pi ) , \end{split}\ ] ] where we introduced the short hand notation @xmath15_{(m)}\equiv\sum_{m=-\infty}^{\infty } \int_{-\pi}^{\pi}\,d\phi.\ ] ] to take into account the @xmath16-periodicity of @xmath17 . in eq .
, the two terms on the right hand side quantify the diffusion - related loss and gain of particles in the phase space element @xmath18 , respectively .
to asses the effect of binary particle collisions , we assume that the one - particle distribution function @xmath5 varies on time and length scales which are large compared to the typical time between particle collisions and typical inter - particle separations , respectively . therefore , on the scale of the boltzmann equation , the impact of collisions on the distribution of angles @xmath1 can be formulated locally in space . to capture the effect of binary collisions , we resort to a collision rule , i.e. a model - specific mapping @xmath19 between pre- and post - collisional angles : @xmath20 further
, we assume that the collision process is subject to noise , which causes a rotation of the post - collisional angles @xmath21 by some random amount drawn from a probability distribution @xmath22 .
we then arrive at the following form of the collision integrals @xmath11 ( again , omitting any explicit reference to space and time coordinates ) : @xmath23 = \\ & -f(\theta)\int_{\theta-\psi}^{\theta+\psi}d\phi\,\gamma(\theta-\phi)f(\phi)\\ & + \int_{-\pi}^{\pi}[d\phi_1]_{(m)}\int_{\phi_1-\psi}^{\phi_1+\psi}d\phi_2\,\gamma(|\phi_2-\phi_1|)\,f(\phi_1)f(\phi_2)\\ & \times\frac{1}{2}\left[p\left(\theta-\theta_w^{(1)}+2m\pi\right ) + p\left(\theta-\theta_w^{(2)}+2m\pi\right)\right ] , \end{split}\ ] ] where @xmath24 is a phase space factor ( `` boltzmann stozylinder '' ) , quantifying the rate of collisions between particles with orientations @xmath25 and @xmath26 . the first term on the right hand side of eq .
gives the rate at which the phase space element @xmath18 loses particles due to binary collisions .
similarly , the second term quantifies the rate at which particles are scattered into @xmath18 . for the sake of greater generality , we further introduced the parameter @xmath27 $ ] in eq . , to account for systems with
limited angular interaction ranges @xcite .
. have been investigated in a number of previous studies @xcite by mapping the mesoscopic boltzmann equation to a set of hydrodynamic equations of motion for the first few moments of the one - particle distribution function @xmath5 . as discussed in the introduction ,
this mapping is based on a truncation of terms in fourier space representation of the boltzmann equation which itself rests on a perturbation ansatz for the fourier modes around an isotropic base state .
to which extent such an approach yields reliable results in the context of ordered systems remains an open question ; cf .
@xcite . in the following sections ,
we take a different route and focus on a kinetic description for active systems by a direct numerical solution of the boltzmann equation .
before we derive a discretized version of eqs . , one last remark is in order : since eqs . imply @xmath28=0 $ ] , eq .
gives rise to a conservation law for the local particle density @xmath29 : @xmath30 where @xmath31 denotes the local momentum density . in what follows
, we will restrict ourselves to the discussion of closed systems in which the particle flux across its boundaries vanishes .
eq . then entails conservation of the overall particle density @xmath32 where @xmath33 denotes the system s two - dimensional spatial domain , and @xmath34 its volume .
we start by rewriting eqs . in dimensionless form , in order to reduce the number of parameters . to this end , we measure time in units of the self - diffusion time scale @xmath35 , distances in units of the `` ballistic flight length '' @xmath36 , and the one - particle distribution function @xmath37 in units of the ( constant ) overall particle density @xmath38 [ cf .
finally , the phase space factor @xmath39 quantifies the rate particle collisions and is , therefore , proportional to the particle speed @xmath0 and the particle extension @xmath40 .
dimensional analysis then dictates @xmath41 .
we thus arrive at the following convention of units [ eq : rescalings ] @xmath42 which will be used throughout this work .
the boltzmann equation then takes the dimensionless form @xmath43 + \rho_0\,i_{\text{c}}\left[f;\,\theta\right].\ ] ] here the expressions for @xmath10 and @xmath11 are given in eqs .
( with @xmath44 ) and , respectively , with @xmath37 and @xmath39 in their dimensionless forms . in the present system of units , the dimensionless density parameter @xmath38 measures the rate of collisions [ @xmath45 relative to the rate of self - diffusion [ @xmath12 ] .
as an aside we note that in a regime of high densities and / or large ballistic flight lengths with @xmath47 , the characteristic scales introduced above are no longer meaningful . in such cases ,
the characteristic time scale would be set by the typical time between subsequent collisions , @xmath48 , the characteristic length scale by the typical inter - particle distance @xmath49 . in this limit ,
self diffusion can be neglected and the problem becomes independent of the overall particle density .
having established a dimensionless form for the boltzmann equation for active , two - dimensional systems , eq . , this section is devoted to the derivation of a discretized version of eq . .
in order to keep the notation compact , we use `` co- and contravariant indices '' such that summation over repeated co- and contravariant index pairs is implied . throughout this work ,
we reserve greek letters for spatial indices , and latin letters for angular indices .
moreover , we use the symbol @xmath50 to denote the measure of spatial domains and angular intervals . to discretize the spatial domain @xmath33
, we consider a general tessellation by means of an arbitrary set of connected domains @xmath51 such that they are a partition of @xmath33 , i.e. [ @xmath52 : kronecker delta ] [ eq : spatialtessellation ] @xmath53 for all @xmath54 .
similarly , to partition the angular range @xmath55 $ ] , we consider tessellations by means of a set of simply connected intervals @xmath56 , such that [ eq : angulartessellation ] @xmath57 for all @xmath58 ; cf fig .
[ fig : tessellations ] .
the particular choices for the spatial tessellations are of no importance for the general development of our discretization scheme .
however , some care must be taken in the construction of the angular tessellations . to avoid creating an artificial angular bias ,
the angles @xmath59 , used to represent the partition @xmath56 , should sum up to zero .
in addition , the angular partition should be constructed such that any special lattice found in the spatial tessellations ( e.g. when space - filling , regular partitions are being used ) should be reflected in any particular choice of angular tessellations . in what follows
, we will use @xmath60 , \qquad\theta_n=\frac{2\pi n}{k},\ ] ] and choose @xmath61 such that ( i ) the angular partition @xmath62 and ( ii ) for each angular domain @xmath63 there exists a domain @xmath64 such that @xmath65 ( modulo @xmath16 ) .
this then implies @xmath66 is discretized using an arbitrary tessellation @xmath67 obeying eqs .
a set of angular channels according to the tessellation is attached to each spatial cell @xmath68 .
_ spatial transport : _ convection in the bulk of the system along the direction @xmath69 ( indicated by the dashed arrow ) amounts to translating each spatial cell @xmath68 by the convection length @xmath70 .
the _ convective transport operator _ @xmath71 , eq .
, then redistributes the current value of @xmath72 among all neighboring cells according to their intersection with the convected cell . ]
we proceed by defining indicator functions for the spatial partition of @xmath33 @xmath73 and the angular partition of @xmath74 $ ] , @xmath75 which can be used to write @xmath76 here , the time dependent _ density matrix _
@xmath77 constitutes a discrete representation of the continuous one - particle distribution function @xmath5 . to assess the temporal evolution of the _ density matrix _ @xmath78
, it suffices to establish a transformation law of this matrix under small time increments .
this transformation law then provides us with an `` update rule '' to compute the temporal evolution @xmath79 iteratively , starting from some initial distribution @xmath80 , and hence to numerically solve the boltzmann equation . in the following ,
we restrict ourselves to considering the time evolution in the bulk of the system , postponing a detailed discussion of boundary conditions to section [ sec : bcs ] . consider a small translation in the ( rescaled ) time coordinate , @xmath81 , where @xmath82 . to assess the convective contribution to the time transformation law for the _ density matrix _ @xmath78
, we replace the convective derivative on the left hand side of the boltzmann equation by finite differences . using a forward difference to substitute for the partial time derivative , we get [ eq : convectivederivative ] @xmath83 further , noting that @xmath84 is the directional derivative of @xmath37 along @xmath4 , we write @xmath85 where now a backward difference is used to replace the spatial derivative in order to achieve cancellation of the terms @xmath86 .
we chose equal step lengths in eqs . and to account for the fact that the particles move at unit speed @xmath87 according to our units convention . to derive the time transformation law for the _ density matrix _
@xmath78 due to particle convection , we add eqs . and , substitute the discretization ansatz for the one - particle distribution function @xmath5 , and project the resulting expression onto the discretized phase space volume @xmath88 .
introducing the projection operator @xmath89 we get [ eq : convection ] @xmath90 f({\mathbf{x}},\theta , t)\right\}\equiv\\ & \int d\bm\tilde\omega_n^{\alpha}\,\mathcal{f}^k(\theta)\bigl[f_k^{\gamma}(t+\tau)\mathcal{f}_{\gamma}({\mathbf{x}})-f_k^{\gamma}(t)\mathcal{f}_{\gamma}({\mathbf{x}}-\tau\,\hat{{\mathbf{e}}}_{\theta})\bigr]/\tau\\ & = \bigl[f_n^{\alpha}(t+\tau)-\mathcal{t}_{n\gamma}^{\alpha k}\,f_k^{\gamma}(t)\bigr]/\tau , \end{split}\ ] ] where we introduced the _ convective transport operator _
@xmath91 which quantifies the time propagation of the _ density matrix _ @xmath92 , due to the convective transport of particles . from eq .
we can extract a direct geometrical interpretation for the implementation of convective transport processes via @xmath71 ; cf .
[ fig : tessellations ] : the _ convective transport operator _ is computed by `` convecting '' the indicator function @xmath93 for the spatial domain @xmath94 along @xmath69 , and integrating the resulting function over the domain @xmath95 .
geometrically , this amounts to computing the area of intersection between @xmath95 and a convected copy of @xmath94 divided by the volume of the phase space element @xmath96 .
the operator element @xmath71 thus measures the number particles being transported from @xmath97 to @xmath98 during @xmath99 .
more precisely , @xmath71 gives the increase in particle density in the phase space element @xmath98 relative to the particle density in the phase space element @xmath97 at time @xmath100 due to convective coupling between these elements in the time interval @xmath99 .
a few remarks on the properties of the _ convective transport operator _ are in order .
first of all , since , in the absence of self - diffusion and collisions , the particles maintain their current orientation , the _ convective transport operator _ is diagonal with respect to the angular indices for convective processes in the bulk of the system .
this is , however , no longer true for convection in the vicinity of a ( non - periodic ) boundary , where particles generally reorient due to collisions with boundary walls . a corresponding redefinition of the _ convective transport operator _
will be given in section [ sec : bcs ] , where we discuss the implementation of non - periodic boundary conditions .
secondly , since @xmath101 reduces to the identity operator in the limit @xmath102 , the _ convective transport operator _ becomes `` quasi diagonal '' in the spatial indices for sufficiently small time steps @xmath70 .
more precisely , while the diagonal elements @xmath103 are @xmath104 , off - diagonal elements are @xmath105 , i.e. convective coupling between neighboring elements becomes increasingly small as @xmath102 .
this , in turn , has important consequences for the computation of the angular self - diffusion and collision contributions to the time evolution of the _ density matrix _ @xmath92 , which themselves are @xmath105 as will be shown below .
thus , up to linear order in @xmath70 , convective processes decouple from angular diffusion and collision , such that the latter processes can be computed independently inside each spatial domain @xmath95 .
to discretize the self - diffusion integral @xmath10 , we proceed along the same lines as in the preceding section and substitute the discrete representation of @xmath5 , eq . , and project the resulting expression onto the phase space volume @xmath106 . using the analytical form of the self - diffusion matrix , eq .
, we find @xmath107_{(m)}\\ & \times\mathcal{f}^{k}(\phi)\,f_k^{\gamma}(t)\,\mathcal{f}_{\gamma}({\mathbf{x } } ) \ , p_0(\theta-\phi+2m\pi)\\[.2 cm ] & = \mathcal{d}_n^k\,f_k^{\alpha}(t ) , \end{split}\ ] ] where the matrix elements of the _ self - diffusion operator _ @xmath108_{(m)}\,p_0(\theta-\phi+2m\pi)-\delta_n^k\ ] ] quantify the net gain of particles ( per unit time ) in the angular element @xmath109 , by diffusive scattering from the angular element @xmath110 . finally , the discretized version of the collision integral @xmath11 defines the _ collision operator _ @xmath111 , which can be calculated using the standard procedure outlined in the previous two subsections . starting loss term in eq .
, we obtain : [ eq : collisionloss ] @xmath112 & = -f_n^{\alpha}(t)\,\mathcal{s}_n^l\,f_l^{\alpha}(t ) , \end{split}\ ] ] where the matrix elements [ eq : smatrix1 ] @xmath113 measure the loss of particles ( per unit time ) in the angular element @xmath109 due to collisions with particles in the element @xmath110 . similarly , considering the gain term of eq .
, we find : [ eq : collisiongain ] @xmath114_{(m)}\int_{\phi_1-\psi}^{\phi_1+\psi}d\phi_2\,\gamma(|\phi_2-\phi_1|)\\ & \times \mathcal{f}_{\gamma}({\mathbf{x}})\mathcal{f}_{\delta}({\mathbf{x}})f_k^{\gamma}(t)f_l^{\delta}(t)\mathcal{f}^{k}(\phi_1)\mathcal{f}^{l}(\phi_2)\\ & \times \frac{1}{2}\left[p\left(\theta-\theta_w^{(1)}+2\pi m\right)+p\left(\theta-\theta_w^{(2)}+2\pi m\right)\right]\\[.2 cm ] & = f_k^{\alpha}(t)\mathcal{s}_n^{k l}f_l^{\alpha}(t ) , \end{split}\ ] ] where the matrix elements @xmath115_{(m)}\int_{\phi_1-\psi}^{\phi_1+\psi}d\phi_2\,\gamma(|\phi_2-\phi_1|)\mathcal{f}^{l}(\phi_2)\\ & \times \frac{1}{2}\left[p\left(\theta-\theta_w^{(1)}+2\pi m\right)+p\left(\theta-\theta_w^{(2)}+2\pi m\right)\right ] \end{split}\ ] ] count the number of particles ( per unit time ) being scattered into the angular element @xmath109 as a result of binary collisions between particles with angles @xmath116 and @xmath117 .
combining eqs . and , we find [ eq : collisionterms ] @xmath118 with the _ collision operator _
@xmath111 given by : @xmath119 having discretized the various terms in the boltzmann equation , eq .
, we can now combine our findings from the previous three subsections , to arrive at the following transformation law , capturing the evolution of the _ density matrix _ @xmath120 under a small transformation in time , @xmath81 , @xmath121\,f_k^{\gamma}(t),\ ] ] with the various operators , acting on the _ density matrix _ @xmath120 , given in eqs .
. defines a recursive relation , which allows us to numerically compute the solutions of the boltzmann equation , eq . , given initial conditions @xmath122 .
it , therefore , constitutes a discrete version of the boltzmann equation for active systems , as introduced in section [ sec : continuousboltzmanneq ] .
we conclude the current section by demonstrating that the particle conservation property , inherent in the original boltzmann equation , eq . ,
is being preserved by the transformation law , eq . .
to this end , we use the following operator summation rules , which themselves follow directly from the respective definitions of the corresponding operators : [ eq : summationrules ] @xmath123 .
eqs . and capture the particle conserving properties of the _ self diffusion _ and _ collision operators _ , by asserting that the total gain of particles in each angular element is balanced by the total loss of particles in all remaining angular elements , and vice versa .
now , using eqs .
, we find @xmath124\,f_k^{\gamma}(t)\\ & = ||\mathfrak s_{\gamma}||\,||\mathfrak a^{k}||\,f_k^{\gamma}(t)=n(t ) , \end{split}\ ] ] i.e. conservation of the total number of particles . finally , we note that , without further simplification , the computation time for an algorithm based on eq .
is of order @xmath125 , operations to update each one of @xmath126 angular channels of the local _ density matrix _ @xmath92 . ] and is thus effectively set by the number of angular slices @xmath126 . in the absence of collision noise , on the other hand , the transition probabilities in the definition of the _ collision operator _ [ cf .
reduce to delta peaks about @xmath21 , rendering the implementation of eq . of order @xmath127 .
unless stated otherwise , we will use @xmath128 in what follows . to test whether this is an appropriate choice for the angular resolution , we computed the spatially homogeneous solution of the _ density matrix _ @xmath92 for a range of values @xmath129 and a number of different density levels , using a @xmath130 spatial grid .
in fact , we observe no significant quantitative improvement upon increasing the number of angular slices in the presence of collision noise ; cf .
[ fig : phase_diagram_cuts](a ) . the stationary , spatially homogeneous solution for the one - particle distribution function @xmath17 as depicted in fig .
[ fig : phase_diagram_cuts](a ) has previously been observed numerically @xcite and predicted analytically @xcite . the analytical form of eq
. applies equally well in the bulk and at the boundaries of the system . for non - periodic boundaries , however , the _ convective transport operator _
@xmath101 differs from its bulk form , eq . , and has to be modified . in this section
, we start by discussing these modifications in abstract terms , and will later apply the generalized form of the _ convective transport operator _ to the specific case of reflective boundaries . in the following , we shall confine ourselves to the discussion of fixed boundaries with a model specific reflection rule .
specifically , we assume particles to undergo a spontaneous reorientation upon collisions with the boundary , @xmath131 where @xmath1 and @xmath132 denote the particle s orientation before and after the boundary contact , and where @xmath133 can be either a deterministic or a random variable . for the sake of clarity , we will discuss deterministic particle - boundary interactions .
the treatment of stochastic boundary conditions follows along similar lines . to the considered order ,
self - diffusion processes and particle - particle collisions are evaluated at a fixed point in space and time in eq . .
to account for the presence of boundaries , we are , therefore , led to reconsider the derivation of the _ convective transport operator _
@xmath101 ; cf .
section [ sec : convectiveoperator ] . for small but finite values of the time shift @xmath70
, we have to redefine the convective derivative , eqs .
, for all @xmath134 here @xmath135 denotes the set of all points in space which are within `` convective reach '' of the boundary @xmath136 of the system s domain @xmath33 within the next time interval @xmath137 $ ] .
an appropriate redefinition of the _ convective transport operator _ can then be given using _
`` inverse propagation functions '' _ : [ eq : inverse_propagation_functions ] @xmath138 using these _ inverse propagation functions _ , we redefine the finite difference approximation of the convective derivative as follows : @xmath139 \ , f({\mathbf{x}},\theta , t ) \rightarrow\\[.2 cm ] & \frac{f({\mathbf{x}},\theta , t+\tau)-f({\mathbf{c}}_{s}^{-1}({\mathbf{x}},\theta ; \tau),c_{a}^{-1}({\mathbf{x}},\theta ; \tau),t)}{\tau } , \end{split}\ ] ] which reduces to the bulk form , eq .
, upon substituting the inverse bulk propagation functions @xmath140 , and @xmath141 .
using this generalized finite difference representation to replace the convective derivative , we can proceed along the same lines of section [ sec : convectiveoperator ] to arrive at the following , generalized form of the _ convective transport operator _ @xmath142 since @xmath143 for particles undergoing a collision with the system s boundary during @xmath137 $ ] , @xmath71 is no longer diagonal in the angular indices for @xmath144 . the generalized _ convective transport operator _ , eq .
lends itself to implement virtually arbitrary boundary conditions , which enter the eq . by an appropriate definition of the _ inverse propagation functions _ . to illustrate the general concept
, we conclude this section with a brief discussion of the implementation of reflective boundary conditions . after the collision of a particle with a reflective boundary , the particle s orientation changes according to @xmath145 where @xmath146 denotes the local orientation of the system s boundary , and where @xmath147 and @xmath1 denote the particle s orientation before and after the collision with the reflective boundary , respectively . moreover ,
if a particle s state is described by @xmath148 at time @xmath149 and is known to have undergone a collision with the system s boundary in the time interval @xmath150 $ ] , its spatial position @xmath151 at time @xmath152 is given by @xmath153 where @xmath154 denotes a reflection with respect to the boundary s local tangent of orientation @xmath155 . in summary ,
the _ inverse propagation functions _ in the case of reflective boundary conditions read : [ eq : inverse_propagation_reflective_bc ] @xmath156 and @xmath157 where @xmath158 denotes the line segment joining the points @xmath159 and @xmath160 . as a final remark we mention that eqs .
can be implemented straightforwardly using arbitrary boundary geometries .
although a detailed study on the impact of confining geometries on the behavior of active systems lies beyond the scope of our present work , _ snake _ lends itself as a convenient starting point to address this important question . for sample simulations of active polar systems in circular and ring - like confinements
, we refer the interested reader to the supplemental material @xcite .
having established a general framework to discretize the boltzmann equation for active particles , eq . , we now discuss actual applications of _ snake _ in the context of active polar systems .
we will focus our considerations on a simple model system with a polar collision rule @xmath161 and full angular interaction range . for the self - diffusion and collision noise probability functions @xmath22 and @xmath13 , we choose zero - mean gaussian probability distributions with variances @xmath162 and @xmath163 , respectively . this model has previously been proposed and studied by bertin et al .
@xcite using mainly analytical methods .
our purpose here is twofold .
first , we can resort to a number of well established analytical results @xcite for this model , as well as a number of previous studies of closely related systems @xcite , to test for the accuracy and reliability of our computational scheme .
secondly , we can use _ snake _ to go beyond the hydrodynamic picture of the onset of collective motion , as discussed in refs .
@xcite , and study the nature of the various phases , as well as the ensuing transitions between them . in the following sections
, we will use the system of units introduced section [ sec : dimlessbe ] ; cf .
moreover , we will use spatial tessellations in the form of regular two - dimensional arrays of @xmath164 rectangular grid sites with linear extensions @xmath165 , and periodic boundary conditions .
some care must be taken for the appropriate choice of the size of one `` boltzmann element '' , @xmath166 : the spatial patterns typically formed by these systems at the onset of polar order , are known to develop rather steep wave fronts , with density profiles varying on length scales comparable to the collision length ( i.e. the typical length a particle travels between successive collisions ) @xcite . to be able to resolve such patterns with sufficient accuracy , we need @xmath167 . in all subsequent sections , we will use densities @xmath168 , such that @xmath169 is an appropriate choice .
unless stated otherwise , @xmath169 and @xmath170 will be used throughout this work . [ cols="^,^,^ " , ] and the _ length parameter _ @xmath171 ; cf .
each plot indicates the projection of the cluster lane s midline onto the torus equatorial plane ( the system s extent in the equatorial plane is indicated by a pair of dashed , concentric circles ) .
the _ winding number _
@xmath172 ( rows ) gives the number of full revolutions about the torus midpoint in the equatorial plane .
the special case @xmath173 is indicated by the straight solid lines in the top row .
the _ length parameter _ @xmath171 ( columns ) gives the total length of the closed contour representing the cluster lane s midline . ] here , we will give a brief account of possible cluster lane orientations in systems with periodic boundary conditions .
in particular , we will illustrate that the observation rates for cluster lane patterns do strongly depend on cluster lane orientations , and , therefore , explain why only a small number of typical cluster lane orientations is expected to be observed in simulations with finite system sizes . for the sake of greater clarity , we will consider systems with a @xmath174 aspect ratio .
further , we will use the one - dimensional `` midline '' of each cluster lane [ i.e. the zero of the time averaged particle current inside the _ cluster zone _ ; cf . fig .
[ fig : cluster_particle_currents](a ) ] to represent the entire ( two - dimensional ) lane . since clusters moving on different lanes ( and antiparallel directions ) must not collide , each cluster lane on the system s periodic domain @xmath33 must be closed .
this can expressed by requiring @xmath175 where @xmath176 denote the distance along the @xmath177-axis between subsequent intersections of the cluster lane s midline with the lines @xmath178 and @xmath179 , and where @xmath172 and @xmath171 are coprime integers ; cf .
[ fig : cluster_lane_geometry](a )
. since all values of @xmath176 with @xmath180 are equivalent upon interchanging @xmath177- and @xmath181-axis , we can assert @xmath182 without loss of generality . in eq .
the integers @xmath172 and @xmath171 characterize the topological and geometrical properties of the cluster lane , as is shown in fig .
[ fig : cluster_lane_geometry](a ) .
specifically , @xmath172 gives the cluster lane s _ winding number _ in the equatorial plane of the torus representing the periodic domain @xmath33 ; cf . fig .
[ fig : winding_number ] .
the _ length parameter _
@xmath171 measures the total length of the cluster lane in units of the ( orientation dependent ) length parameter @xmath183 , where @xmath184 denotes the cluster lane s orientation with respect to the @xmath181-axis . according to eq .
, the ratio @xmath185 can take on arbitrary rational values . at first sight
, this does not seem to imply a severe restriction , since the set of rational numbers is a dense subset of the real numbers .
in other words , arbitrary cluster lane orientations @xmath186 are `` infinitely close '' to meeting the requirement eq . . while this is true in the hydrodynamic limit , @xmath187
, finite size effects considerably restrict the number of possible cluster lane orientations @xmath188 . to see this , note that the area @xmath189 each cluster lane of orientation @xmath186 covers on @xmath33
is given by ( @xmath190 : cluster lane s transversal extension ) @xmath191 since the total system size is finite , @xmath192 , and since the minimum number of cluster lanes is @xmath193 ( one cluster lane alone corresponds to a wave pattern ) , we have @xmath194 therefore , noting that the cluster lane width @xmath190 is proportional to the transverse correlation length @xmath195 of the polar order parameter , and that @xmath196 , we get @xmath197 for the system sizes used in our numerical computations , we found @xmath198 .
we , therefore , observe cluster lane patterns characterized by @xmath199 , i.e. @xmath200 , amounting to cluster lane orientations @xmath201 ; cf .
[ fig : cluster_lane_geometry](b , c ) .
this corresponds to the cluster lane geometries indicated in the top left plot in fig .
[ fig : winding_number ] . visual inspection of fig .
[ fig : winding_number ] illustrates the rapidly increasing space requirements for cluster lane patterns with increasing values of @xmath172 and @xmath171 .
these space requirements , in turn , directly affect the competition between cluster lane and wave patterns during the pattern selection process : although eq .
has to be fulfilled for any line , representing the lateral extension of a wave pattern , as well , the resulting space requirements [ as captured by eqs . and ]
are far less restrictive for waves .
more precisely , a single wave band can attain any orientation @xmath188 such that @xmath202 where @xmath203 denotes the longitudinal extension of the wave .
since @xmath204 , eq . is therefore compatible with a much larger set of orientations @xmath188 than eqs .
and ( for finite system sizes @xmath205 ) .
finite system sizes @xmath205 , therefore , imply a bias in favor of the formation of wave patterns .
in this appendix , we briefly explain the numerical protocol used to obtain the data points shown in fig . [
fig : hd_hysteresis ] ( open symbols ) .
special care must be taken to properly assess the location of the density scale @xmath206 , where spatially homogeneous polar ordered states become linearly unstable toward the emergence of density segregated patterns . at the point where the homogeneous system turns linearly unstable ,
the development of spatial heterogeneities out of a virtually homogeneous state is extremely slow . as a sensitive means to detect the buildup of spatial heterogeneities
, we introduce the maximum absolute deviation , @xmath207 for the _ density matrix _ @xmath92 , where @xmath208 this maximum absolute deviation approaches zero for spatially homogeneous systems and attains finite values [ typically of order @xmath209 for states inside the dsp regime .
to estimate the value of @xmath206 , we proceed as follows .
we start from a spatially homogeneous base state , equilibrated at @xmath210 with @xmath211 .
we then quench the overall density @xmath38 to successively lower values and let the system evolve for a time span of at least @xmath212 . had to be run for even longer time
spans up to @xmath213 in order to observe the formation of wave patterns . for @xmath214
, we observed a monotonic but extremely slow increase in @xmath215 . to keep computation times at a tolerable level
, we therefore perturbed the system by setting @xmath216 at @xmath212 ( random variable @xmath217 uniformly distributed on @xmath218 $ ] ) ] during this time , we record the evolution of the quantity @xmath215 . for all data points in fig .
[ fig : hd_hysteresis ] with @xmath219 , the maximum absolute deviation @xmath215 is a monotonically decreasing function of time , and the system approaches a spatially homogeneous state .
conversely , in systems being represented by data points with @xmath220 , @xmath215 increases such that the spatially homogeneous initial states eventually give rise to density segregated patterns .
as mentioned earlier , for density quenches down to final densities @xmath221 in the vicinity of @xmath206 , the actual buildup of spatial heterogeneities happens on a very large time scale . in actual computations ,
this time scale turns out to be much slower than the local `` equilibration '' of the density matrix @xmath92 toward the new density level @xmath38 .
we , therefore , do not expect that the results shown in fig .
[ fig : hd_hysteresis ] should depend on whether they are obtained from a `` quasi - static '' protocol , or a density quench protocol as discussed here .
55ifxundefined [ 1 ] ifx#1 ifnum [ 1 ] # 1firstoftwo secondoftwo ifx [ 1 ] # 1firstoftwo secondoftwo `` `` # 1''''@noop [ 0]secondoftwosanitize@url [ 0 ]
+ 12$12 & 12#1212_12%12@startlink[1]@endlink[0]@bib@innerbibempty link:\doibase 10.1016/j.aop.2005.04.011 [ * * , ( ) ] link:\doibase 10.1146/annurev - conmatphys-070909 - 104101 [ * * , ( ) ] @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1103/physrevlett.75.1226 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.75.4326 [ * * , ( ) ] link:\doibase 10.1103/physreve.58.4828 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.92.025702 [ * * , ( ) ] link:\doibase 10.1103/physreve.74.030904 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1103/physreve.77.011920 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.101.268101 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1103/physreve.77.046113 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.104.184502 [ * * , ( ) ] link:\doibase 10.1103/physreve.81.061916 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.106.128101 [ * * , ( ) ] @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1103/physreve.71.050901 [ * * , ( ) ] link:\doibase 10.1103/physreve.74.022101 [ * * , ( ) ] link:\doibase 10.1103/physreve.73.031109 [ * * , ( ) ] @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1103/physreve.75.040901 [ * * , ( ) ] link:\doibase 10.1103/physreve.83.030901 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1103/physrevlett.109.098101 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.109.268701 [ * * , ( ) ] \doibase http://dx.doi.org/10.1063/1.4812822 [ * * , ( ) ] http://stacks.iop.org/1367-2630/15/i=4/a=045014 [ * * , ( ) ] @noop * * , ( ) http://stacks.iop.org/1367-2630/15/i=8/a=085032 [ * * , ( ) ] http://stacks.iop.org/1367-2630/16/i=3/a=035003 [ * * , ( ) ] @noop * * , ( ) @noop *
* , ( ) @noop @noop * * , ( ) link:\doibase 10.1103/physrevlett.112.148102 [ * * , ( ) ] link:\doibase 10.1080/00018739400101505 [ * * , ( ) ] \doibase http://dx.doi.org/10.1016/0022-3697(61)90054-3 [ * * , ( ) ] link:\doibase 10.1080/00018737600101402 [ * * , ( ) ] http://stacks.iop.org/1742-5468/2011/i=02/a=p02029 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.110.055701 [ * * , ( ) ] @noop * * ( ) link:\doibase 10.1103/physrevlett.111.145702 [ * * , ( ) ] | kinetic theories constitute one of the most promising tools to decipher the characteristic spatio - temporal dynamics in systems of actively propelled particles . in this context
, the boltzmann equation plays a pivotal role , since it provides a natural translation between a particle - level description of the system s dynamics and the corresponding hydrodynamic fields . yet
, the intricate mathematical structure of the boltzmann equation substantially limits the progress toward a full understanding of this equation by solely analytical means . here , we propose a general framework to numerically solve the boltzmann equation for self - propelled particle systems in two spatial dimensions and with arbitrary boundary conditions .
we discuss potential applications of this numerical framework to active matter systems , and use the algorithm to give a detailed analysis to a model system of self - propelled particles with polar interactions . in accordance with previous studies , we find that spatially homogeneous isotropic and broken symmetry states populate two distinct regions in parameter space , which are separated by a narrow region of spatially inhomogeneous , density - segregated moving patterns .
we find clear evidence that these three regions in parameter space are connected by first order phase transitions , and that the transition between the spatially homogeneous isotropic and polar ordered phases bears striking similarities to liquid - gas phase transitions in equilibrium systems . within the density segregated parameter regime
, we find a novel stable limit - cycle solution of the boltzmann equation , which consists of parallel lanes of polar clusters moving in opposite directions , so as to render the overall symmetry of the system s ordered state nematic , despite purely polar interactions on the level of single particles . |
quantum entanglement has been extensively studied , both due to its fundamental significance in quantum theory @xcite and its utility as a resource for quantum communication and quantum information processing @xcite .
atomic physics offers a domain with sufficient control of the system and isolation from noise that it has been possible to perform precision experiments on quantum entanglement @xcite , and , therefore , it is also a productive target for theoretical study of the issue as well .
one of the simplest scenarios for theoretical studies of the dynamics of entanglement between atoms is that of two atoms which are isolated from one another and interact with different electromagnetic fields . studying this sort of model led to the discovery of the entanglement sudden death ( sd ) phenomenon @xcite , in which entanglement decays to zero in a finite time rather than asymptotically .
sd has garnered significant interest because it is both unintuitive and potentially undesirable .
an alternative simple model for atom - field interaction in which to study entanglement dynamics is the model studied by dicke @xcite and tavis and cummings @xcite , where one assumes all atoms are grouped in a sample whose size is small compared to the resonant wavelength ( resulting in identical coupling to every atom ) .
however , such a simplistic model may miss the variety of behavior that can result when the atoms are not confined to such a small sample .
entanglement dynamics have been shown to have a significant distance dependence when two atoms are interacting with a common field @xcite . for atoms weakly interacting with a continuum of field modes in the born - markov approximation , it has been shown @xcite that changing the atomic separation of two atoms can affect whether there is sd and whether there is revival of entanglement , as well as modify the dynamical generation of entanglement ; in short , the qualitative features are sensitive to the atomic spacing . at short interatomic distance
non - markovian effects associated with induced interactions between the atoms due to the quantum field become more pronounced whose qualitative behavior varies greatly with different classes of initial states , as detailed in the study of @xcite .
so it is clear that distance dependence gives rise to a significant variety of behaviors , and some seemingly innocuous approximations can qualitatively alter the entanglement dynamics in unintuitive ways @xcite . in view of this , we seek to study entanglement dynamics in a model that is sufficiently complex to manifest some of this variety of behavior yet simple enough that the dynamics may be understood in considerable detail and obtained with fewer approximations .
the simplest model one may pursue along this line would be two two - level atoms coupled to a single mode with couplings that reflect the distance dependence ; however , as discussed in sec . [ sec : sysham ] , one needs to include at least two field modes before any non - trivial distance dependence can arise in the problem .
so we will study such a model with two field modes here .
we choose to have traveling wave modes so as to eliminate the effect of atomic center of mass position and focus on the relative separation between the atoms , this can be experimentally realized in the situation where one has two counter - propagating modes in a toroidal resonator cavity @xcite .
the motivation for working with as few modes as as possible rather than a continuum is to explore the variety of dynamical behavior that can arise in a non - dissipative system and its sensitivity to initial state . in order to better qualify the entanglement dynamics
, we must specify a way of quantifying entanglement . for two qubits
one may compute the entanglement of formation efficiently in terms of the wootters concurrence , and the concurrence itself may be used as an entanglement monotone @xcite .
this is the quantity we will use most often to quantify the atom - atom entanglement in this paper .
alternatively , to quantify entanglement one can also use the negativity @xcite ( which quantifies the degree to which the state violates the peres - horodecki positive partial transpose condition ) .
this is useful on larger hilbert spaces , and we use this to quantify entanglement when considering aspects of the the entanglement dynamics involving field states . in section [ sec : sysham ] we describe a useful way to parameterize the class of hamiltonians that can give rise to different entanglement dynamics due to atomic separation . in section [ sec : sp ] we focus on two special cases in terms of separation with the aim of illustrating the variety of different behaviors that can arise in different initial atomic and field states .
specifically , we will show how sd , dynamical entanglement generation , and other phenomena differ between these two cases by comparing with analogous situations of the two well - studied types of models - where each system is a single two - level atom interacting with a single field mode @xcite and two two - level atoms interacting with a single common field mode @xcite .
we will show that the entanglement dynamics in these models differs significantly to illustrate the effect of atomic separation .
finally , in section [ sec : gen ] we examine the entanglement dynamics for other atomic separations , using analytic results for a few cases and numerical results more generally , to examine to what degree the behaviors of the special cases examined generalize .
to look at the entanglement evolution for intermediate values of separation analytically , we use the resolvent expansion method of @xcite restricted to initial states with a single excitation among the atoms and field modes .
also , for the most general case we can look at the dynamics by numerical diagonalization of the hamiltonian for all the other initial states and arbitrary separations .
in this way we explore the implications of interatomic distance on entanglement dynamics in this simplest non - trivial context .
consider a pair of identical two - level atoms coupled via the minimal coupling interaction hamiltonian to a collection of electromagnetic ( em ) field modes under the dipole and rotating - wave approximation .
we will assume that atoms can be effectively considered to have fixed center of mass positions @xmath0 .
the interaction hamiltonian of the system in the schrdinger picture has the form @xmath1 where @xmath2 and @xmath3 are the raising and lowering operators for the two - level system representing the @xmath4 atom , and @xmath5 and @xmath6 are the creation and annihilation operators corresponding to the @xmath7 normal mode with frequency @xmath8 and polarization indexed by @xmath9 in some mode decomposition of the em field with ( complex ) classical electric field mode function @xmath10 .
let us assume that the atomic transition in question does not alter the angular momentum of the atom . under these assumptions @xmath11 where the field is quantized in a region of volume v
, @xmath12 is the complex dipole matrix element for the transition in the @xmath4 atom , and the dipole moment has a direction @xmath13 @xcite . in general , the mode functions @xmath14 can be quite complicated , and even for a single atom there can be position dependence in the dynamics arising from the boundary conditions that the mode functions obey .
in order to distinguish these effects from the effect of atomic separation , we will consider mode functions of the form @xmath15 in terms of some set of coordinates @xmath16 , with @xmath17 indexing the polarization .
this form describes traveling - wave mode functions that are solutions to an em boundary value problem which is translation invariant under translations in the coordinate @xmath18 , and it appears in a number of interesting physical systems , including spherical resonators @xcite and toroidal resonators @xcite .
we will consider situations in which all atoms share the same coordinates @xmath19 and @xmath20 , differing only in the coordinate @xmath21 , so that the variation in behavior due to the atomic separation will manifest without additional position dependence arising from the boundary conditions . let us further assume that the atoms are arranged such that the atomic transition dipole vectors are aligned with one of the mode polarizations , so @xmath22 .
in this situation only one polarization is relevant , and all polarization labels will be suppressed . with these further assumptions
we may express the coupling constants as @xmath23 however , it turns out that without loss of generality one may study a much smaller set of hamiltonians . given a particular set of complex phases for the coupling constants @xmath24 , one may make a trivial basis transformation @xmath25 which simply amounts to redefining the reference for the phases of the atoms by @xmath26 and for the fields by @xmath27 . after this transformation ,
the interaction hamiltonian becomes @xmath28 with @xmath29 , and we may obtain the solution to the original problem by using the transformed hamiltonian : @xmath30 in terms of the entanglement dynamics , we are not concerned with the basis transformation as it only acts locally on each of the atoms and field modes . if we consider two identical atoms , with frequencies @xmath31 and dipole strengths @xmath32 coupled to a single em field mode , then the previous paragraph implies that it suffices to consider only a hamiltonian where both couplings are real , and the atomic separation does not enter ; thus , there can be no non - trivial distance dependence . however , if we consider the two identical atoms coupled to two field modes , then without loss of generality we can write the total hamiltonian of the system as @xmath33,\nonumber \end{aligned}\ ] ] where all distance dependence arises from @xmath34 , and `` h.c . ''
denotes the hermitian conjugate terms . for a pair of counter - propagating modes
, we have @xmath35 and hence the parameter @xmath36 . for simplicity
, we will further assume that @xmath37 , which implies @xmath38 .
our aim is to get a sense of the variety of different entanglement dynamics that can result from different separations and initial conditions . to that end
, we will first consider two special cases , which are arguably extreme cases of the general model , @xmath39 ( where @xmath40 ) which we will call the case of symmetric coupling , and @xmath41 ( where @xmath42 ) , which we will call the case of antisymmetric coupling ( of the second atom to the second mode ) . next we will discuss how the qualitative features of entanglement dynamics can differ significantly between these two special cases of atomic separation .
in order to quantitatively study the entanglement dynamics , we must specify a way of quantifying entanglement . for two qubits
one may compute the entanglement measure known as the entanglement of formation efficiently in terms of the wootters concurrence @xcite . in order to compute the concurrence
, one must choose an arbitrary basis and construct the spin - flipped operator @xmath43 where the star denotes the complex conjugation in that basis and @xmath44 is the pauli matrix in the same basis .
the concurrence is @xmath45 where the @xmath46s are the eigenvalues of the matrix @xmath47 in decreasing order .
the entanglement of formation can then be computed as @xmath48 where @xmath49 alternatively , to quantify entanglement one can also use the negativity @xcite ( which quantifies the degree to which the state violates the peres - horodecki positive partial transpose condition ) , defined @xmath50 where @xmath51 is the partial transpose of the density matrix , defined by @xmath52 in an arbitrary basis b , and @xmath53 is the trace norm . either of these approaches can be used to quantify entanglement .
the negativity is arguably less precise , because in some systems it can be zero even for an entangled state , but we will not consider such situation .
the advantage of the negativity is that it can be efficiently computed even in a large hilbert space .
we will most often use the concurrence in this article ; however , we do quantify the entanglement of field states using the negativity .
as a first step toward evaluating the influence of interatomic separation on entanglement dynamics , we now examine two particularly simple cases that one might expect to represent two distinct extremes .
since we have assumed the two field modes have the same frequency , rather than using the original modes @xmath54 and @xmath55 one could equally well choose a different mode decomposition of the field where the two modes are replaced by linear combinations @xmath56 and @xmath57 with annihilation operators @xmath58 these will be two orthogonal standing wave modes .
the assumed translation invariance of the mode structure in the direction of the coordinate @xmath18 allows us to choose the phase of the standing wave modes arbitrarily ( subject to the constraint of orthogonality ) , and the mathematical definitions we have given here specify that @xmath56 will have an anti - node at the position of atom 1 and @xmath57 will have a node at that location .
the coupling to atom 2 then depends on the separation . with this mode transformation
we can now turn to the special cases of interest . when @xmath59 we have the case of two modes with symmetrical coupling ( sc ) to the second atom .
the interaction hamiltonian can then be written @xmath60\label{eq : tmsctranshaq}\ ] ] in terms of these transformed modes , so that the atoms only couple to @xmath61 and not @xmath62 .
this reflects the fact that the atoms are separated by @xmath40 , so the second atom also lies at a node of @xmath57 and anti - node of @xmath56 .
thus , the hamiltonian is equivalent to the model of a single mode symmetrically coupled to two atoms , a tavis - cummings ( tc ) model , and all work done on that model can be applied here .
the evolution of the reduced density matrix of the atom @xmath63 in the sc model is be the same as in the tc model with a properly transformed initial state .
namely , if the total system ( atoms and modes ) is in an initial state described by the density matrix @xmath64 , then the appropriate initial density matrix for the tc model is obtained by making the mode transformation of eqs . and and tracing out the second transformed mode @xmath57 . as
this mapping of initial states involves a partial trace , it is a many - to - one mapping from the sc problem to the tc problem ( and this mapping does not preserve purity ) . in order to solve the dynamics in the tc model , and by extension the sc model ,
we simply compute the time evolution operator expressed in the atomic basis @xmath65 directly by exponentiation ( as in , e.g. @xcite ) : @xmath66 where @xmath67 we observe that there is generally oscillatory behavior for the state dynamics coming from the periodic time dependence and the frequency of these oscillations depends on the number of excitations in the initial field state .
also , a fock state has a well defined frequency .
when @xmath68 we have a two - mode model with asymmetric coupling ( ac ) to the second atom . in terms of the transformed modes of eqs . and the hamiltonian can be written as @xmath69.\label{eq : tmactranshaq}\ ] ] in this case , rather than both atoms coupling to one mode we see that atom 1 couples only to transformed mode @xmath56 while atom 2 couples only to @xmath57 . because in this case @xmath42 , atom 2 now lies at a node of @xmath56 and anti - node of @xmath57
this situation is equivalent to a model comprised of two subsystems that are totally isolated from one another , each composed of a single atom coupled to a single mode .
we will call this the double jaynes - cummings ( djc ) model .
this sort of model with isolated subsystems is common to the study of entanglement sudden death @xcite , and the djc model specifically has been studied @xcite , so the mapping allows those investigations to bear on this model . as in the previous case ,
the evolution of @xmath63 for the ac model should be the same as given by the djc model with the proper mapping of initial states .
this implies that there can , for example , be no dynamical generation of atomic entanglement in the ac model unless the djc initial field state obtained by the mapping is entangled . in the djc model
we can write the unitary time - evolution operators for the two separate non - interacting atom - field subsystems as @xmath70 and @xmath71 , and then the total time evolution operator is @xmath72 .
we again compute the two subsystem unitary time evolution operators by direct exponentiation to obtain @xmath73 with @xmath74 in order to illustrate the variety of behavior that can arise among these models , we examine the entanglement dynamics of a selection of initial states in which the atoms are uncorrelated with the fields . in each case
we select the atomic state from the set of pure states @xmath75 , where @xmath76 , @xmath77 .
the initial field state will be either a product of fock states @xmath78 , a product of glauber coherent states @xmath79 , a product of squeezed vacuum states @xmath80 , a two - mode squeezed vacuum state ( tmss ) @xmath81 , a thermal state @xmath82 ( with both modes having equal temperature ) , the pure state @xmath83 , or the mixed state @xmath84 .
the two - mode squeezed state is defined as the state resulting from the action of the two - mode squeezing operator @xmath85 on vacuum .
the state @xmath83 is the result of mapping the state @xmath86 in the original modes of the ac problem to the transformed modes equivalent to the djc problem , with @xmath87 and @xmath88 $ ] is the state in the tc model that gives equivalent evolution to the state @xmath86 in the sc model . to find the atom - atom density matrix we trace out the field modes from the time evolved state of the closed atom - field system .
[ cols="<,<,^,^,^,^,^ " , ] we summarize our findings for the entanglement behavior given the various initial states considered in the four models in table [ phi0 ] , listing the equivalent sc - tc cases and ac - djc cases .
when discussing entanglement sudden death , we adopt the usage of the term as in @xcite in applying it only to instances where the entanglement goes to zero for some finite ( non - zero ) duration . in the case where entanglement goes to zero only for an instant during the time evolution we refer to it as death for an instant ( di ) .
if there is a non - zero entanglement at all times once it is generated in the system then we label it as being `` always living '' ( al ) . in the context of that general summary
, we now turn to commenting on some specific features . by reducing the sc and ac models to the tc and djc models , respectively
, we see that the change in atomic separation leads to very different entanglement dynamics . in the case of the ac model
, we have a mapping to the dynamics of the djc model , where each atom interacts only with a separate field mode .
so in the djc model , the dynamics can not increase entanglement between the two atom - field subsystems ; therefore , if the atomic state is not entangled initially then it will remain separable , unless there is an initial entanglement between the field modes that can be transferred to the atoms by the dynamics . knowing this , we can see that any initial field state for the ac model that maps to a separable djc field state fails to generate entanglement .
the nature of the mapping means that even some entangled field states will fail to generate atomic entanglement in the ac model , while some separable states will map to an entangled djc state and will generate entanglement .
for example , when considering the ac model with an initial field state that is a product of squeezed states @xmath80 we find that entanglement is dynamically generated as shown in fig .
[ djctmss ]
. one can understand this from the fact that the state @xmath80 maps to the djc model with an initial tmss , so the generation of atomic entanglement occurs simply because the dynamics transfers the entanglement between the field modes to the atoms .
further , we find that there is an optimal squeezing parameter value for entanglement generation when the field state is close to being a maximally entangled qubit state @xmath89 , with @xmath90 and @xmath91 being comparable .
on increasing the squeezing parameter further there are contributions from higher fock states which decrease the transfer of entanglement .
transfer of field - field ( @xmath56-@xmath57 ) correlations can be explicitly seen in a small squeezing approximation when a two mode squeezed state interacts with an initially separable atomic state @xmath92 . since the probability of higher photon numbers is small , by restricting to the 4-dimensional subspace of lowest energy states in the fock basis for @xmath56-@xmath57 we obtain the negativities for the two subsystems as , @xmath93 where @xmath94 , @xmath95 and @xmath96 .
[ lowtemp ] shows how the dynamics of the system transfer the entanglement between field - field subsystem and atom - atom repeatedly .
turning to the sc model , the correspondence drawn to the tc model leads to the essential feature that the map for the initial field states from sc to tc is many to one ; the set of initial states that have the same reduced density matrix for the first transformed mode @xmath56 have identical entanglement dynamics in terms of atom - atom entanglement .
as a counterintuitive example of this feature we note that a squeezed state of the form @xmath80 and a thermal field in the sc model give the same entanglement dynamics provided the average number of photons in the thermal field corresponds to that in the squeezed state ( @xmath97 ) .
we can also turn to previous work on the tc model to expose interesting features of the sc model .
as had been shown in @xcite , in the tc model an atomic state @xmath92 interacting with any fock state @xmath98 in the field mode never gets entangled .
this can be generalized to any initial field state with a density matrix that is diagonal in the fock basis as follows : if one defines the time evolved state @xmath99 , then the atom - atom density matrix given as @xmath100 $ ] remains separable . extending to a general field density matrix diagonal in the fock basis , the time evolved atom - atom density matrix given as @xmath101 $ ] is clearly a convex sum of separable density matrices , and hence there is no entanglement generation .
looking at the initial state @xmath102 in the tc case , we observe no sd in fig .
this can be explained by considering the state as a superposition @xmath103 , where due to the symmetry of the coupling constants the maximally entangled dark state @xmath104 does not interact with the field .
it is only momentarily during the evolution that the state of the system returns to being the original separable superposition @xmath105 . as a result
we always observe some entanglement between the two atoms for an initial field density matrix diagonal in the fock basis .
the dynamics of the @xmath106 state in the tc model , shown in fig .
[ ggn ] , exhibit sd in general except for the special case of @xmath107 where because of symmetry reasons the state oscillates between the states @xmath108 and @xmath109 , going from being separable to maximally entangled .
hence , any density matrix of the field modes diagonal in the fock basis with a high component of @xmath110 would generate more entanglement in general .
+ + + apart from the fock state in the tc model and a thermal field , another example of having a initial field density matrix diagonal in the fock basis is to have a fock state @xmath111 in the sc which corresponds to the tc density matrix @xmath112 where @xmath113 . for this state
again we see no entanglement generation for @xmath92 , di for @xmath105 and maximal entanglement in @xmath114 , as shown in figs .
[ egnm ] and [ ggnm ] .
another point we observe is that for an initial state @xmath115 there is no entanglement generation , which is a common feature between sc and ac . a fock state @xmath111 in ac transforms into an entangled state in the djc model
, so we expect to have entanglement generation in the system for an initially separable atomic state . as an exception we find that for @xmath116
, if there is no atom - atom entanglement to begin with then the atoms remain separable evolving to a diagonal density matrix .
this is counterintuitive in the sense that there is no entanglement transfer to the atomic subsystem from the entangled field modes .
this feature can be explained by considering the initial field state @xmath117 .
if the atoms are initially in the state @xmath92 then time evolution will lead to an entanglement of atomic and field states such that detecting whether the number of photons in each mode is even or odd tells us the state of the two atoms .
on tracing out the field this gives us a diagonal atom - atom density matrix with no atom - atom entanglement .
the same is true for the initial atomic states @xmath105 and @xmath118 .
+ + while table [ phi0 ] encapsulates the qualitative features of entanglement dynamics for many states , it may be useful to point to a few interesting examples .
note that many familiar initial field states that are separable in the djc model fail to dynamically generate entanglement in the ac model .
in contrast , entanglement generation is a common feature in the sc model for the same selection of field states .
for example , starting with a two - mode squeezed state in the field modes does not generate any entanglement in ac , because the corresponding field state in djc possesses no correlations amongst @xmath56 and @xmath57 . on the other hand , a two - mode squeezed state in the symmetrically coupled case generates entanglement in the initially separable atomic states as can be seen in fig .
furthermore , for the separable initial atomic states @xmath92 and @xmath118 , if the field is sufficiently squeezed , entanglement is dynamically generated and once generated sustains forever , protected from sudden death . this behavior is shown in figs .
[ ssqee ] and [ ssqgg ] . if the atoms are initially in the entangled atomic state @xmath119 , then fig .
[ ssqeegg ] shows that the entanglement stays protected as the state becomes almost a dark state for large squeezing parameter values .
one can also see entanglement protection for a thermal field in the sc model so long as the thermal average photon number is below a threshold value @xmath120 ( fig .
[ singlethermeegg ] ) with the state being al ( above this critical temperature @xmath119 experiences sd as well ) . in considering a selection of familiar states for these two special cases of symmetric and anti - symmetric coupling
, we find that the same sort of initial states can have very different entanglement dynamics when the spacing of the atoms is altered . for the selection of initial states considered , the symmetric coupling , where @xmath40 , is better for generating entanglement between the atoms , and for protecting entanglement from sudden death .
however , these remain only special cases that reveal some facets of the differences in entanglement dynamics arising from a change in distance .
it remains to be seen what happens for a more general selection of distances .
to extend the results found in the previous section to arbitrary atomic separations we use two different approaches to solve the problem the resolvent approach as described in @xcite and numerical diagonalization of the hamiltonian .
the first approach gives an analytic solution but only yields sufficiently simple expressions to be of use for initial states that have at most one excitation among the atoms and field modes , meaning superpositions and mixtures of states belonging to the set @xmath121 .
using the resolvent expansion of the hamiltonian @xmath122 and expanding @xmath123 we can find an exact expression for the matrix elements @xmath124 by resumming the series expansion in ( [ pr ] ) .
it can be seen that the action of the operator @xmath125 on the subsets @xmath126 and @xmath127 gives a state in the same subspace .
say , @xmath128 and @xmath129 , such that @xmath130 then , @xmath131 and @xmath132 .
now consider the matrix element @xmath133 , by using the expansion in ( [ pr ] ) this can be summed up to obtain @xmath134 similarly , @xmath135 from those matrix elements we obtain the matrix elements of the propagator as in ( [ res ] ) : @xmath136 where @xmath137 and @xmath138 .
it can be seen as a general observation that for any initial state that lies in the single excitation subspace the time evolved atom - atom density matrix has the form @xmath139 @xcite . for this general form of the density matrix
the atom - atom concurrence simplifies to @xmath140 , meaning for a state in the single excitation subspace the entanglement of the two atoms has a dynamics that is either al or di ( since @xmath141 is analytic and can never vanish over a finite non - zero length of time ) . using the propagator for the one - excitation subspace , we can examine the analytic expressions for entanglement dynamics of a few initial states .
we begin with the state @xmath142 .
the time evolved reduced density matrix of the two atoms after tracing out the field modes is obtained as @xmath143 this gives a concurrence for the two qubits as @xmath144 it can be seen from the above expression that , unless @xmath68 , the atoms are entangled except at a discrete , periodic set of instants given by @xmath145 , @xmath146 where the state is separable ( i.e. , `` dead '' ) . as the separation parameter @xmath147 approaches @xmath148 , @xmath149 approaches @xmath150 and the death time approaches infinity .
thus for antisymmetric coupling , there is no entanglement generated at all , as the situation corresponds to having the two atoms interacting with two separate transformed modes .
thus , this special case is uncharacteristic of the qualitative features of the entanglement dynamics for general separations , and for any other separation the atoms will become entangled , if perhaps on a long timescale .
+ for the initial state @xmath151 the atom - atom entanglement is @xmath152 in this situation it can be seen that the entanglement remains al except for separations where @xmath153 is rational , where the entanglement dynamics is di .
this latter case includes symmetric and antisymmeric couplings , where the concurrence oscillates as @xmath154 and @xmath155 respectively .
+ here we see that the special cases describing states with high symmetry are not at all representative of the generic qualitative features for arbitrary separation . for the initially entangled atomic state @xmath156 ,
the concurrence is given as @xmath157 , again we can see that entanglement is al unless @xmath153 is a rational number
. + restricting to the single excitation subspace , one might also seek to find the approximate concurrence for thermal and coherent states with low average photon numbers ( keeping only one excitation in the field modes ) interacting with @xmath118 ; however , one can find that the resulting values are of the same order as two - excitation terms that have been neglected , so this approach is not useful in that case .
to look at the entanglement dynamics for other initial states we numerically diagonalize the hamiltonian , restricting to a small photon number approximation .
we find that sd is generally observed for most distances with the initial states we have discussed in the preceding section . in many cases , however , the special cases of symmetric and antisymmetric coupling show qualitatively different dynamical behavior . for example ,
if we look at the entanglement dynamics for the maximally entangled state @xmath119 interacting with a thermal state in the sc model , we had seen that the entanglement was protected upto some critical temperature , while there was sd for the ac model . upon extending this to arbitrary separation values
we find that for all separations where @xmath158 the entanglement always shows sd for all temperatures as shown in fig .
[ sdgen ] .
similarly , for @xmath119 interacting with a two - mode squeezed state one has sd except for symmetric coupling or @xmath59 .
+ as we had discussed in the previous section for the case of an initial em field state diagonal in the fock basis in the tc model , if both the atoms are excited there is no entanglement generation and for the atoms in @xmath105 the entanglement dies only for an instant . considering
the fock state @xmath159 in the sc model translates to the state @xmath160 in the tc model which is diagonal in the fock basis , hence we see no entanglement generation for @xmath92 and di for @xmath105 . on the other hand , we had also shown that for such a state there is no entanglement generation in the ac situation .
we use numerical approach to consider the entanglement behavior for intermediate separations .
it can be seen from fig.[fockgen ] that the intermediate distances have sudden death for the case of @xmath161 , even if one deviates slightly from the symmetric coupling @xmath59 .
the state @xmath162 shows entanglement generation for asymmetric couplings .
+ while examining the special cases with high symmetry in the setup it is useful to show the over - riding effect of symmetry on the entanglement dynamics of this system , computing the results for arbitrary distances suggests that these special cases are not representative in terms of the general qualitative behavior of entanglement .
as we have discussed above , in many cases the dynamics falls in one qualitative class almost everywhere on the interval of asymmetrically coupled cases ( @xmath163 or @xmath148 ) but in a different class on one or both boundaries .
moreover , of the three qualitative classes we have discussed here , all are possible at every distance , given the correct initial state .
an analysis of linear oscillators coupled to a common vacuum field ( consisting of a continuum of modes ) by lin and hu found that there exists an entanglement distance , so that oscillators with a separation less than that distance will be asymptotically entangled , while initially entangled oscillators spaced further will undergo sd @xcite .
in contrast , an analysis of two two - level atoms interacting with a common em thermal field ( with a continuum of modes ) found @xcite that all initial states will eventually undergo sd for any non - zero separation .
in this two - mode model we find no evidence of an `` entanglement distance '' which determines a change in the qualitative features of entanglement dynamics .
for each of initial states considered , the qualitative class into which the entanglement dynamics falls is the same for almost all distances , with departures only on a set of measure zero . for a thermal field , which is the environment considered in @xcite ( the former considers specifically the vacuum ) ,
sd occurs at almost all distances .
this bears some resemblance to the result of @xcite , though the non - dissipative nature of this model ensures revival and , in this regard , is qualitatively quite different .
we have analyzed the entanglement dynamics in a model consisting of two two - level atoms and two electromagnetic field modes for a variety of familiar field states and classified the various cases in terms of different dynamical behavior .
one aim of this analysis is to get a sense of the variety of entanglement dynamics that can arise from different atomic separations in a case where two atoms interact with a shared em field .
it is useful to examine this question in a very simple model that can be solved with a minimum of approximations ( using which can lead to unphysical effects ) and understood in detail .
we have argued that there is no non - trivial distance dependence in a single mode model , and , therefore , a two - mode model represents the simplest case for the study of distance dependence in the entanglement dynamics . on the other hand ,
including a continuum of modes leads to dissipation and misses the variety of dynamical behavior that can be seen with a few number of modes .
in particular , the effect of initial state is largely washed out .
we have first studied two arguably extreme cases out of the class of two - mode hamiltonians that can arise , one in which the two modes are symmetrically coupled ( sc ) to the second atom , and one where the two modes have asymmetric coupling ( ac ) .
a useful insight in understanding these models is that the atomic dynamics in the sc model correspond exactly to the dynamics for a tavis - cummings model with a single field mode symmetrically coupled ( tc ) to both atoms with a suitable mapping of the field state , while the atomic dynamics for the ac model correspond exactly to the dynamics for a double jaynes - cummings ( djc ) model , made up of two isolated subsystems each with one atom coupled to one mode , under a proper transformation of the field state .
these mappings help one understand the significant differences in behavior in the two seemingly similar models , giving a window into how symmetry introduced by specific atomic separation can affect entanglement dynamics .
another significant implication of the mapping between the sc and tc models comes from the fact that the mapping of initial field states between those models is many - to - one , this shows us entire classes of field states for the sc model that will give exactly the same atomic dynamics . in examining the dynamical generation of entanglement from an initially separable atomic state , we find quite a marked contrast between the sc and ac models . while entanglement generation is a relatively common feature , present for a variety of field states , in the sc model , it is comparatively much more rare in the ac model .
this difference is not so surprising however , if one views it in terms of the mappings we have introduced to the other models . when considering that the ac maps to the djc model , where the two subsystems are isolated , one would expect entanglement generation to be relatively rare ; it can only exist in cases where the field state is mapped to an entangled state whose entanglement can then be transferred to the atoms . with the sc ,
by contrast , we have a mapping to the tc model where a single shared field mode can readily introduce entanglement between the two atoms .
these models ( tc and djc ) have previously been studied for some of the cases that we have looked at as well .
for example , tessier et al in @xcite had also analyzed the the initial state @xmath164 and coherent state in the tc model .
also , interaction of atomic states @xmath92 , @xmath105 and @xmath118 with a thermal field was studied in the context of thermal generation of entanglement in @xcite .
we extend the conclusion of these results more generally to the set of initial field states that are diagonal in the fock basis .
we find that for a field state diagonal in the fock basis there is no entanglement generation in @xmath92 , di for @xmath105 and for @xmath118 an initial field state with a high component of @xmath110 would generate more entanglement in general .
other than that we study the case of an initially entangled atomic state interacting with a thermal field and find that the entanglement is preserved for low temperatures and the incidence of sudden death occurs only after a certain critical temperature .
the djc model was previously studied from the viewpoint of transfer of entanglement phenomenon in @xcite and more generally by @xcite for the case of an entangled atomic state interacting with a vacuum field and transferring correlations via the atom - field entanglement .
we look at the phenomenon for the reversed situation , where one starts with a maximally correlated field state ( tmss ) and looks at the entanglement generated in the atomic subsystem .
since the field state does not remain gaussian as the system evolves , we resort to a low temperature approximation in order to quantify the entanglement .
we also see that the transfer is maximum for an optimal squeezing parameter value such that the field state approximates a maximally entangled qubit state . as a potentially useful phenomenon
we observe cases where we find that the entanglement always stays alive once generated or stays protected if already present . in particular , for the case of a squeezed state interacting with an initially entangled atomic state @xmath119 we see that the qubits end up in an almost dark state and stay almost maximally entangled for all times . as a natural question one may ask what would be the extension of these two extreme cases to intermediate separations and how does the dynamics on one extreme change into a different behavior on the other .
we look at the general separations using the resolvent method for states in the single excitation subspace to find the concurrence as a function of separation and time .
we see that these states can only exhibit al / di behavior based on the general form of the atom - atom density matrix as noted in @xcite , in particular it can be seen explicitly how the behavior changes for the symmetric and anti - symmetric couplings .
we can look at all the other cases using numerical analysis and we find that for all the cases considered there is no critical separation behavior as seen in the continuum cases @xcite .
for example , situations where entanglement stays protected in the symmetric coupling case ( thermal state and squeezed state ) shows sd as soon as there is any asymmetry in the coupling .
we have chosen in this analysis to try to isolate the effect of atomic separation on entanglement dynamics from the position - dependent effects arising from boundary conditions . for this purpose ,
we have assumed in sec .
[ sec : sysham ] an atom - field coupling that depends on the coordinate separating the atoms only by a phase factor . for clarity
we have supposed our two field modes are traveling - wave modes in free space , however a more experimentally relevant situation would be a toroidal resonator , where there is rotational invariance in the azimuthal direction that satisfies our requirements . in this case
the two modes of interest would be two resonant , counter - propagating whispering - gallery modes .
because strong coupling between an atom and a whispering - gallery modes of a microtoroidal resonator has observed experimentally @xcite , there is the possibility of experimentally probing quite directly the model we have considered .
however , in the experimental system there will be dissipative dynamics arising from emission into other modes outside the resonator , so a detailed comparison would require either including the dissipative effects in the theoretical model or a restriction to the case of sufficiently strong coupling to the resonator modes and early times that the dissipation could be neglected .
our analysis of special cases of the entanglement dynamics arising in two atoms interacting with two modes suggests a wide variety of different behaviors can arise , with qualitative features of the dynamics changing entirely between the two special cases considered , even with the same initial field state .
this suggests that understanding the distance dependence of entanglement dynamics for multiple atoms interacting with a common field will be quite important for predicting even the qualitative features that may arise .
furthermore , if one has the practical goal of dynamically generating entanglement or protecting entanglement once generated , the special cases we have considered suggest that the ability to achieve these goals will be greatly impacted by the separation of the atoms .
having gotten some sense of the variation in behavior that can arise , we quantify the entanglement dynamics over a range of atomic separations rather than just special or extreme cases that highlight the importance of symmetry factors in entanglement dynamics . on this point it would be interesting to compare the exact results of the present model with the general statements on classifying the entanglement behavior @xcite based on geometrical and topological properties of quantum state spaces @xcite .
this research is supported in part by grants from the nsa - laboratory for physical sciences , the darpa - quest program ( darpahr0011 - 09 - 1 - 0008 ) and the nsf - itr program ( phy-0426696 ) .
results here were reported by blh at the program on quantum open systems at kitp - china ( beijing ) supported by the project of knowledge innovation program ( pkip ) of the chinese academy of sciences , grant no . | we analyze the time evolution of quantum entanglement in a model consisting of two two - level atoms interacting with a two - mode electromagnetic field for a variety of initial states and interatomic separations .
we study two specific atomic separations which give rise to symmetric atom - field couplings . for general atomic distances
we consider a subset of initial states analytically , and then treat the more general situation numerically .
we examine a variety of qualitative features such as entanglement sudden death , dynamical generation , protection , and transfer between subsystems .
our analysis shows a stark contrast in features of entanglement between the two special coupling schemes often considered ; further , these features are uncharacteristic of those arising for general distances , due to the high degree of symmetry present in the special cases .
the variety of behaviors in these two - mode cases suggest the importance of considering atomic separation carefully for any model where two atoms interact with a common field .
# 1|#1 # 1#1| # 1 # 1 # 1#2^#1#2 |
[ horolimdefn ] let @xmath0 be a locally symmetric space of noncompact type ( with universal cover @xmath1 ) , and let @xmath5 .
a point @xmath2 on the visual boundary of @xmath1 is a _ horospherical limit point _ for @xmath3 if every horoball based at @xmath2 intersects the orbit @xmath6 .
( see [ cgammahorolimdefn ] for an alternate characterization which makes it clear that this notion is independent of the choice of the basepoint @xmath7 . )
our main theorem characterizes the horospherical limit points for any finite - volume locally symmetric space @xmath0 of noncompact type .
the result is slightly easier to state if we assume that the lattice @xmath3 is arithmetic .
( see [ nonarithsect ] for the general case . )
let @xmath8 be the real points of a connected , semisimple algebraic group over @xmath4 , and let @xmath9 be the corresponding symmetric space of noncompact type .
nobreaktrue 1 .
it is well known that if @xmath10 is any @xmath11-split torus in @xmath8 , then there exists @xmath5 , such that @xmath12 is a flat in @xmath1 ( cf .
6.1 , pp . 245 ) ) .
we say the flat @xmath12 is _
@xmath4-split _ if the torus @xmath10 is ( defined over @xmath4 and ) @xmath4-split .
the killing form on @xmath8 induces a metric on @xmath13 that gives it the structure of a symmetric space ( * ? ? ?
we call this the _ killing - form metric_. see [ notkilling ] for a formulation of our results that applies to the other symmetric metrics on @xmath13 . the direction ( @xmath14 ) of the following result
has already been proved by t.hattori ( * ? ? ?
a or prop .
4.4 ) , but we provide a proof of both directions because our methods are quite different . [ horolimiffgrat ] let @xmath15 be an arithmetic locally symmetric space of noncompact type with the killing - form metric .
a point @xmath16 is a horospherical limit point for @xmath3 if and only if @xmath2 is not on the boundary of any @xmath4-split flat .
since @xmath17 acts transitively on the set of maximal @xmath4-split tori , we have the following reformulation : [ ratbldg ] let nobreaktrue * @xmath8 be the real points of a connected , semisimple algebraic group over @xmath4 , * @xmath9 be the corresponding symmetric space of noncompact type with the killing - form metric , and * @xmath18 be the boundary of some maximal @xmath4-split flat in @xmath1 .
then the set of horospherical limit points for @xmath19 is the complement of @xmath20 .
for the special case of @xmath4-split groups , we can state this another way : [ qsplithorolimits ] let @xmath21 be a connected , @xmath4-split , semisimple algebraic group over @xmath4 .
a point @xmath2 on the visual boundary of the corresponding symmetric space @xmath22 is * not * a horospherical limit point for @xmath23 if and only if @xmath2 is fixed by some parabolic @xmath4-subgroup of @xmath21 .
[ genrems ] nobreaktrue 1 .
the set @xmath20 in the statement of [ ratbldg ] is known as the `` rational tits building '' of @xmath21 @xcite .
thus , the result states that the set of horospherical limit points of @xmath23 is equal to the complement of the rational tits building of @xmath21 .
this was conjectured by w.h.rehn @xcite ( in somewhat less generality ) , but the inclusion ( @xmath24 ) has remained open even for the case where @xmath25 with @xmath26 .
a geodesic ray @xmath27 is _ divergent _ if the function @xmath28 is a proper map .
it is easy to see that if the endpoint of @xmath27 is not a horospherical limit point , then @xmath27 must be divergent .
the converse is not true , because s.g.dani @xcite has shown that if @xmath29 , then there are many geodesic rays that diverge for `` non - obvious '' reasons , and [ ratbldg ] shows that the endpoints of such rays are horospherical limit points .
3 . in [ qsplithorolimits ] , the assumption that @xmath21 is @xmath4-split can be weakened to the assumption that @xmath30 .
we also note that this does not assume @xmath1 has the killing - form metric it is valid for every symmetric metric on @xmath31 ( if @xmath30 ) .
4 . given a locally symmetric space @xmath0 and a finitely generated @xmath3-module @xmath32 , a corresponding set @xmath33 of _ horospherical limit points _ has been defined by r.bieri and r.geoghegan @xcite .
it reduces to [ horolimdefn ] when @xmath34 is the trivial @xmath3-module , but it would be interesting to extend [ horolimiffgrat ] by calculating @xmath33 for other @xmath3-modules .
the proof of [ horolimiffgrat ] is short ( about a page for each direction ) , but relies on definitions and other background material from the theory of algebraic groups , lie groups , and unipotent dynamics .
these preliminaries are presented in [ prelimsect ] .
onflatnothorosect proves that the boundary points of a @xmath4-split flat are not horospherical .
the other direction of [ horolimiffgrat ] is proved in [ mainsect ] .
( see [ horolimiff ] for a summary that provides several alternative formulations of [ horolimiffgrat ] . )
the presents a generalization of [ horolimiffgrat ] that allows @xmath3 to be non - arithmetic .
see @xcite for a generalization of [ horolimiffgrat ] that allows @xmath3 to be an @xmath35-arithmetic group .
we thank ross geoghegan for explaining the conjecture of rehn that motivated this line of research , and we thank the park city mathematics institute for bringing the two of us together and providing an opportunity to start work on this problem .
we also thank tam nguyen phan and kevin wortman for helpful conversations about the structure of horospheres in symmetric spaces of higher rank .
in addition , we thank the latter for calling @xcite to our attention , and pointing out that it proves one direction of [ horolimiffgrat ] .
for any lie group @xmath36 , we let @xmath37 be the identity component of @xmath36 . @xmath38 .
we record a few well - known , elementary observations .
[ cgammahorolimdefn ] @xmath2 is a horospherical limit point for @xmath3 iff there is a compact subset @xmath39 of @xmath1 , such that @xmath40 intersects every horoball based at @xmath2 .
( @xmath14 ) let @xmath41 , where @xmath7 is the basepoint chosen in [ horolimdefn ] . ( @xmath42 )
choose @xmath43 , such that @xmath44 for all @xmath45 .
any horoball @xmath46 based at @xmath2 contains a smaller horoball @xmath47 , such that the distance from @xmath47 to the complement of @xmath48 is greater than @xmath49 . by assumption , there exist @xmath45 and @xmath50 , such that @xmath51 . since @xmath52 , this implies @xmath53 .
cgammahorolimdefn implies that the set of horospherical limit points is independent of the choice of the basepoint @xmath5 , and also does not change if we replace @xmath3 by any finite - index subgroup .
therefore , we have the following consequence : [ invtundercommgamma ] the set of horospherical limit points for @xmath3 is invariant under the action of the commensurator group @xmath54 on @xmath55 .
in particular , if @xmath56 ( and @xmath8 is defined over @xmath4 ) , then the set of horospherical limit points for @xmath19 is invariant under the action of @xmath57 on @xmath55 .
[ horosubgrp ] let nobreaktrue * @xmath32 be a maximal @xmath11-split torus of @xmath8 , * @xmath58 , such that @xmath59 is a flat in @xmath1 , * @xmath60 be a nontrivial one - parameter subgroup of @xmath32 , * @xmath16 be the endpoint of the ray @xmath61 , * @xmath62 be the codimension - one subgroup of @xmath32 that is orthogonal to @xmath60 ( with respect to the killing form ) , * @xmath63 be a weyl chamber of @xmath32 that contains the ray @xmath64 , and * @xmath65 be the maximal unipotent subgroup of @xmath8 , such that @xmath66 as @xmath67 for all @xmath68 and all @xmath69 in the interior of @xmath63 . then nobreaktrue 1 . [ horosubgrp - subgrp ] @xmath70 is a horosphere based at @xmath2 , for each @xmath71 .
[ horosubgrp - limit ] @xmath2 is not a horospherical limit point for @xmath3 iff @xmath72 where @xmath73 is the identity element of @xmath8 .
( [ horosubgrp - subgrp ] ) let @xmath74 . for each @xmath75 ,
the geodesic ray @xmath76 is at bounded distance from @xmath77 ( because @xmath78 is a bounded set ) .
therefore , @xmath79 fixes the point @xmath2 , so it acts ( continuously ) on the set of horospheres based at @xmath2 .
since these horospheres are parametrized by @xmath11 , and every continuous homomorphism @xmath80 is trivial on @xmath65 , we conclude that @xmath65 fixes every horosphere based at @xmath2 .
therefore , @xmath81 is contained in the horosphere through @xmath82 .
since the iwasawa decomposition @xmath83 tells us that @xmath8 is the disjoint union of these sets ( and every point of @xmath1 is on a unique horosphere ) , the set must be the entire horosphere .
( [ horosubgrp - limit ] ) from ( [ horosubgrp - subgrp ] ) , we know that each horoball based at @xmath2 is of the form @xmath84 ( for some @xmath85 ) . therefore , the equivalence in ( [ horosubgrp - limit ] ) is a restatement of [ cgammahorolimdefn ] ( by using ( * ? ? ?
* thm . 1.12 , p. 22 ) ) . [ inspanmustlarge ]
suppose nobreaktrue 1 .
@xmath86 , with @xmath87 , 2 .
[ inspanmustlarge - span ] @xmath88 is in the span of @xmath89 , 3 .
[ inspanmustlarge - acute ] @xmath90 for all @xmath91 , 4 .
[ inspanmustlarge - obtuse ] @xmath92 for @xmath93 , and 5 .
@xmath94 .
then , for all sufficiently large @xmath95 and all @xmath96 , there is some @xmath91 , such that @xmath97 .
this is a standard argument . from ( [ inspanmustlarge - span ] )
, we may write @xmath98 with @xmath99 .
also , by passing to a subset , we may assume @xmath89 is linearly independent , and that @xmath100 for every @xmath91 . then , by replacing @xmath101 with the span of @xmath89 , we may assume that @xmath89 is a basis .
permute the elements of @xmath89 so that the negative values of @xmath102 come first .
that is , there is some @xmath103 with @xmath104 for @xmath105 and @xmath106 for @xmath107 .
let @xmath108 .
then @xmath109 and @xmath110
so we must have equality throughout , which implies @xmath111 .
therefore @xmath112 , so we must have @xmath113 ( since @xmath114 is linearly independent ) .
this means @xmath106 for all @xmath91 .
we claim there is some @xmath115 , such that , for every @xmath96 , there exists @xmath91 , such that @xmath116 .
suppose not .
then there must be some nonzero @xmath96 , such that @xmath117 for all @xmath91 .
so @xmath118 hence , we must have @xmath119 for all @xmath91 .
since @xmath114 is a basis , this implies @xmath120 , which is a contradiction .
since @xmath121 is a basis ( and @xmath88 is nonzero ) , we must have @xmath122 for some @xmath123 .
then @xmath124 is large whenever @xmath125 is large .
thus , if the conclusion of the fails to hold , then @xmath126 must be large ( and negative ) , so @xmath127 must be large . by making it so large that @xmath128 , and applying the claim of the preceding paragraph , we have @xmath129 , as desired .
if @xmath79 is a parabolic subgroup of a connected , semisimple lie group @xmath8 with finite center , then we may write @xmath130 , where nobreaktrue * @xmath10 is an @xmath11-split torus , * @xmath131 is a connected , reductive subgroup that centralizes @xmath10 and has compact center , and * @xmath132 is the unipotent radical of @xmath79 .
[ splitperpnochar ]
let @xmath133 be a field of characteristic @xmath134 .
if @xmath36 is a reductive @xmath133-subgroup of an algebraic @xmath133-group @xmath8 , and @xmath36 has no nontrivial @xmath133-characters , then @xmath36 is orthogonal to every @xmath133-split torus @xmath10 that centralizes it .
let @xmath135 , @xmath136 , and @xmath137 be the lie algebra of @xmath8 , @xmath36 , and @xmath10 , respectively .
consider any minimal @xmath138-invariant @xmath133-subspace @xmath139 of @xmath135 .
since @xmath36 has no @xmath133-characters , it must act on @xmath139 via @xmath140 , so @xmath141 for every @xmath142 .
on the other hand , since @xmath10 centralizes @xmath36 ( and is @xmath133-split ) , schur s lemma tells us that any @xmath143 acts by a scalar @xmath144 on @xmath139 .
therefore @xmath145 since @xmath36 is reductive , we know that @xmath135 is the direct sum of such submodules @xmath139 , so the trace of @xmath146 is @xmath134 .
this means @xmath147 ( with respect to the killing form ) .
[ mperpa ] if @xmath130 is a parabolic subgroup of @xmath8 , then @xmath10 is orthogonal to @xmath131 .
[ aninp ] let @xmath83 be an iwasawa decomposition of @xmath8 . if @xmath79 is a parabolic subgroup of @xmath8 , and @xmath148 , then @xmath149 .
let @xmath150 be the normalizer of @xmath65 , so @xmath133 is a ( minimal ) parabolic subgroup of @xmath8 , such that @xmath151 and @xmath152 .
since @xmath148 and @xmath153 is normal in @xmath79 , we know that @xmath154 is a unipotent subgroup . since @xmath65 is a maximal unipotent subgroup of @xmath8 , this implies @xmath155 . in other words , @xmath156 .
since @xmath79 and @xmath133 are parabolic subgroups , this implies @xmath157 ( cf .
so @xmath158 .
[ afixesxi ] let @xmath32 be a maximal @xmath11-split torus of @xmath8 , @xmath2 be a point on the visual boundary of @xmath13 , and @xmath159 .
if @xmath32 fixes @xmath2 , and @xmath160 is a ( maximal ) flat in @xmath13 , then @xmath2 is on the boundary of @xmath160 .
let @xmath161 , and choose a maximal flat @xmath162 , such that @xmath2 is on the boundary of @xmath162 .
since @xmath163 is abelian , it is clear that @xmath163 fixes @xmath2 , so @xmath164 . also , since @xmath79 is a parabolic subgroup ( cf .
the start of the proof of [ horosubgrp]([horosubgrp - subgrp ] ) ) , it is zariski closed , so any two maximal @xmath11-split tori in @xmath79 are conjugate . hence , there is some @xmath75 , such that @xmath165 .
then @xmath166 is on the boundary of the flat @xmath167 .
the uniqueness of the flat fixed by @xmath32 implies this flat is @xmath160 .
[ closureishomog ] if nobreaktrue * @xmath65 is a maximal unipotent subgroup of a connected , semisimple lie group @xmath8 , and * @xmath3 is a lattice in @xmath8 , then there is a closed , connected subgroup @xmath36 of @xmath8 , such that nobreaktrue 1 .
@xmath168 , 2 .
@xmath169 is a lattice in @xmath36 , 3 .
@xmath170 , and 4 .
@xmath65 acts ergodically on @xmath171 , with respect to the @xmath36-invariant probability measure .
we can describe the subgroup @xmath36 quite explicitly if the lattice @xmath3 is arithmetic : [ closureisporbit ] suppose nobreaktrue * @xmath172 , where @xmath21 is a connected , semisimple algebraic group over @xmath4 , * @xmath3 is a subgroup of finite index in @xmath173 , and * @xmath65 is a maximal unipotent subgroup of @xmath8 .
then there is a parabolic @xmath4-subgroup @xmath79 of @xmath8 , with real langlands decomposition @xmath130 , and a connected , closed , normal subgroup @xmath174 of @xmath131 , such that nobreaktrue * @xmath175 , and * @xmath176 . since @xmath177 contains the maximal unipotent subgroup @xmath65 , we know that @xmath174 contains all of the noncompact , simple factors of @xmath131 .
however , it may be missing some of the compact factors .
closureishomog has been vastly generalized by m.ratner ( * ? ? ?
a and cor .
[ onflatnothoro ] let nobreaktrue * @xmath178 , where @xmath21 is a connected , semisimple @xmath4-group , * @xmath9 be the corresponding symmetric space of noncompact type , with the killing - form metric , * @xmath179 , where @xmath180 is a maximal @xmath4-split torus of @xmath21 , * @xmath5 , such that @xmath181 is a ( @xmath4-split ) flat in @xmath1 , and * @xmath60 be a one - parameter subgroup of @xmath35 .
then the endpoint of the geodesic ray @xmath61 is not a horospherical limit point for @xmath23 .
let : nobreaktrue * @xmath182 be the system of roots of @xmath21 with respect to @xmath180 , * @xmath183 be a base of @xmath182 , such that @xmath184 for all @xmath185 and all @xmath186 , * @xmath187 be a @xmath4-torus in @xmath8 that contains some maximal @xmath11-split torus @xmath32 , and also contains @xmath35 ( such a torus can be constructed by applying ( * ? ? ?
* cor . 3 of 7.1 , p. 405 ) to @xmath188 ) , and
* @xmath62 be the orthogonal complement of @xmath60 in @xmath32 . for each @xmath185
, let : nobreaktrue * @xmath189 , such that @xmath190 for all @xmath191 , and * @xmath192 be the parabolic @xmath4-subgroup of @xmath8 corresponding to @xmath193 , where nobreaktrue * * @xmath194 is the one - dimensional subtorus of @xmath35 on which all roots in @xmath195 are trivial , * * @xmath196 is reductive with @xmath4-anisotropic center , and * * the unipotent radical @xmath197 is generated by the roots in @xmath198 that are _ not _ trivial on @xmath194 .
let @xmath65 be a maximal unipotent subgroup of @xmath8 that is normalized by @xmath32 and is contained in the minimal parabolic @xmath4-subgroup @xmath199 .
( in other words , let @xmath65 be the unipotent radical of a minimal parabolic @xmath11-subgroup of @xmath8 that contains @xmath32 and is contained in @xmath199 . )
note that : nobreaktrue * since @xmath183 is a basis for the dual of @xmath35 ( viewed as a vector space ) , we know that @xmath200 spans @xmath35 .
hence , @xmath60 is contained in the span of @xmath200 . * for @xmath185 and @xmath95
, we have @xmath201 . * for @xmath202 with @xmath203 ,
it is a basic property of root systems that @xmath204 . therefore @xmath205 .
so [ inspanmustlarge ] tells us that if @xmath95 is sufficiently large , then , for all @xmath206 , there exists @xmath185 , such that @xmath207 is large .
note that @xmath193 extends uniquely to a @xmath4-character @xmath208 of @xmath187 .
namely , @xmath208 must be trivial on the @xmath4-anisotropic part of @xmath187 , which is complementary to @xmath35 .
then , since [ splitperpnochar ] tells us that the anisotropic part is orthogonal to @xmath35 , we have @xmath209 for all @xmath210 ( not only for @xmath191 ) .
hence , the conclusion of the preceding paragraph tells us that @xmath211 is large . since conjugation by the inverse of @xmath212 contracts the haar measure on @xmath197 by a factor of @xmath213 for some @xmath214 , and
the action of @xmath65 on @xmath197 is volume - preserving , this implies that , for any @xmath215 , conjugation by the inverse of @xmath216 contracts the haar measure on @xmath197 by a large factor . since @xmath217 is a cocompact lattice in @xmath197 ( * ? ? ?
* thm . 2.12 )
, this implies there is some nontrivial @xmath218 , such that @xmath219 is small .
therefore , [ horosubgrp]([horosubgrp - limit ] ) implies that @xmath2 is not a horospherical limit point for @xmath23 .
suppose @xmath0 is a locally symmetric space of noncompact type , and @xmath2 is a point on the visual boundary of @xmath1 .
we say the horospheres based at @xmath2 are _ uniformly coarsely dense _ in @xmath0 if there exists @xmath220 , such that , for every horosphere @xmath221 based at @xmath2 , every point of @xmath0 is at distance @xmath222 from some point in @xmath223 , where @xmath224 is the natural covering map .
suppose @xmath225 .
it is obvious that if the horospheres based at @xmath2 are uniformly coarsely dense in @xmath226 , then they are uniformly coarsely dense in @xmath227 .
horolimiff implies that the converse is true if @xmath226 has finite volume .
[ horotoinfty ] let nobreaktrue * @xmath178 , where @xmath21 is a connected , semisimple @xmath4-group , * @xmath13 be the corresponding symmetric space of noncompact type with the killing - form metric , * @xmath3 be a subgroup of finite index in @xmath19 , and * @xmath2 be a point on the visual boundary of @xmath13 .
if the horospheres based at @xmath2 are not uniformly coarsely dense in @xmath228 , then there is a parabolic @xmath4-subgroup @xmath229 of @xmath21 , such that nobreaktrue 1 .
[ horotoinfty - fixxi ] @xmath230 fixes @xmath2 , and 2 .
[ horotoinfty - fixhoro ] @xmath231 fixes some ( or , equivalently , every ) horosphere based at @xmath2 .
fix any @xmath232 .
choose nobreaktrue * a maximal ( connected ) @xmath11-split torus @xmath32 of @xmath8 , and * a one - parameter subgroup @xmath60 of @xmath32 , such that nobreaktrue * @xmath160 is a ( maximal ) flat in @xmath13 , and * @xmath2 is the endpoint of the geodesic ray @xmath233 .
let nobreaktrue * @xmath63 be a weyl chamber of @xmath32 that contains @xmath234 , and * @xmath235 .
note that @xmath83 is an iwasawa decomposition of @xmath8 .
let @xmath130 and @xmath174 be as in [ closureisporbit ] .
denote by @xmath62 the orthogonal complement of @xmath60 in @xmath32 ( with respect to the killing form ) , so @xmath62 is a ( codimension - one ) connected subgroup of @xmath32 . since @xmath236 ( and @xmath79 is parabolic ) , we have @xmath149 ( see [ aninp ] ) .
therefore , since all maximal @xmath11-split tori of @xmath79 are conjugate ( * ? ? ?
20.9(ii ) , p. 228 ) , and @xmath237 contains a maximal @xmath11-split torus , there is no harm in assuming @xmath238 , by replacing @xmath237 with a conjugate .
horosubgrp([horosubgrp - subgrp ] ) tells us that the horosphere based at @xmath2 through the point @xmath82 is @xmath239 ( note that @xmath65 preserves the horosphere and thus also the point @xmath2 , so the proof of ( [ horotoinfty - fixxi ] ) will be complete when we show that the levi subgroup @xmath240 also preserves @xmath2 .
) we have @xmath241 by assumption , there is some @xmath125 , such that @xmath223 is not dense .
assuming , as we may , that @xmath242 is the stabilizer of @xmath7 , this implies @xmath243 . since @xmath244 and @xmath245 , we conclude that @xmath246 .
( note that this implies @xmath247 . )
let @xmath248 , so @xmath249 . then , since @xmath246 , we must have @xmath250 .
since @xmath62 has codimension one in @xmath32 , this implies @xmath251 , which means @xmath252 . on the other hand , [ splitperpnochar ] tells us @xmath253 , which implies that @xmath10 is the orthogonal complement of @xmath254 in @xmath32 .
therefore @xmath255 , so @xmath256 .
hence @xmath257 since @xmath258 and @xmath65 each preserve the point @xmath2 at infinity , we conclude that the parabolic @xmath4-subgroup @xmath79 preserves the point @xmath2 at infinity .
this completes the proof of ( [ horotoinfty - fixxi ] ) .
now , we turn to the proof of ( [ horotoinfty - fixhoro ] ) . fixing a basepoint in @xmath13
yields a natural parametrization @xmath221 of the horospheres based at @xmath2 .
if @xmath216 is any isometry of @xmath13 that fixes @xmath2 , then there is some @xmath259 , such that @xmath260 for all @xmath125 .
thus , an isometry that fixes one of these horospheres must fix all of them .
suppose there is some element @xmath261 of @xmath231 with @xmath262 .
( this will lead to a contradiction . ) by replacing @xmath261 with a power of itself , we may assume @xmath50 ( since @xmath263 for all @xmath264 ) .
then , for any @xmath71 , we have @xmath265 since every point in @xmath13 is on some horosphere , this implies that every point is at distance less than @xmath266 from @xmath267 .
therefore , the horospheres based at @xmath2 are uniformly coarsely dense in @xmath268 ( since @xmath266 is a constant , independent of @xmath125 ) .
this is a contradiction .
[ onqsplit ] assume the notation of [ horotoinfty ] .
if there is a parabolic @xmath4-subgroup @xmath229 of @xmath21 , such that nobreaktrue * @xmath230 fixes @xmath2 , and * @xmath231 fixes every horosphere based at @xmath2 , then @xmath2 is on the boundary of a @xmath4-split flat .
let @xmath269 .
there exists a @xmath4-torus @xmath10 of @xmath79 , such that @xmath10 contains a maximal @xmath11-split torus @xmath32 ( * ? ? ?
* cor . 3 of 7.1 , p. 405 ) .
choose @xmath159 , such that @xmath59 is a ( maximal ) flat .
since @xmath149 fixes @xmath2 , afixesxi provides a geodesic @xmath270 in @xmath160 , such that @xmath271 ( and @xmath272 ) .
write @xmath273 , where @xmath35 is @xmath4-split and @xmath274 is @xmath4-anisotropic .
then @xmath275 is a cocompact lattice in @xmath274 ( * ? ? ?
* thm . 4.11 , p. 208 ) and , by assumption , @xmath275 fixes the horosphere through @xmath7 .
this implies that all of @xmath274 fixes this horosphere , so the flat @xmath276 is contained in the horosphere , and is therefore perpendicular to the geodesic @xmath261 . since [ splitperpnochar ] tells us that the orthogonal complement of @xmath276 is @xmath277 , we conclude that @xmath278 .
so @xmath2 is on the boundary of the @xmath4-split flat @xmath277 .
[ horolimiff ]
let @xmath0 be an arithmetic locally symmetric space of noncompact type with the killing - form metric , and let @xmath2 be a point on the visual boundary of @xmath1
. then the following are equivalent : nobreaktrue 1 .
[ horolimiff - horo ] @xmath2 is a horospherical limit point for @xmath3 .
[ horolimiff - notflat ] @xmath2 is not on the boundary of any @xmath4-split flat .
[ horolimiff - parab ] there does not exist a parabolic @xmath4-subgroup @xmath229 of @xmath21 , such that @xmath230 fixes @xmath2 , and @xmath231 fixes some ( or , equivalently , every ) horosphere based at @xmath2 .
[ horolimiff - spheresdense ] the horospheres based at @xmath2 are uniformly coarsely dense in @xmath0 . 5 .
[ horolimiff - ballsdense ] the horoballs based at @xmath2 are uniformly coarsely dense in @xmath0 .
[ horolimiff - onto ] @xmath279 for every horoball @xmath48 based at @xmath2 , where @xmath224 is the natural covering map .
( [ horolimiff - horo ] @xmath14 [ horolimiff - notflat ] ) is the contrapositive of [ onflatnothoro ] .
( [ horolimiff - notflat ] @xmath14 [ horolimiff - parab ] ) is the contrapositive of [ onqsplit ] .
( [ horolimiff - parab ] @xmath14 [ horolimiff - spheresdense ] ) is the contrapositive of [ horotoinfty ] .
( [ horolimiff - spheresdense ] @xmath14 [ horolimiff - ballsdense ] ) is obvious , because horoballs are bigger than horospheres .
( [ horolimiff - ballsdense ] @xmath14 [ horolimiff - horo ] ) is [ cgammahorolimdefn](@xmath42 ) .
( [ horolimiff - horo ] @xmath280 [ horolimiff - onto ] ) is a restatement of [ horolimdefn ] .
suppose @xmath21 is @xmath4-split ( or , more generally , suppose @xmath30 ) . under this assumption , it is easy to show that if @xmath2 is not on the boundary of a @xmath4-split flat , then every horosphere based at @xmath2 is dense in @xmath268 , not just coarsely dense . to see this
, we prove the contrapositive .
note that the proof of [ horotoinfty]([horotoinfty - fixxi ] ) only assumes there is a horosphere that is not dense .
now , if we let @xmath35 be any maximal @xmath4-split torus of @xmath79 , then @xmath35 is also a maximal @xmath11-split torus ( by our assumption that @xmath30 ) , so [ afixesxi ] tells us that @xmath2 is on the boundary of the corresponding ( @xmath4-split ) maximal flat @xmath277 ( since @xmath281 fixes @xmath2 ) .
to state a generalization of [ horolimiffgrat ] that does not require @xmath0 to be arithmetic , we need an appropriate generalization of the notion of a @xmath4-split flat . *
a parabolic subgroup @xmath79 of @xmath8 is _ @xmath3-rational _ if @xmath3 contains a lattice subgroup of @xmath153 . *
assume @xmath283 .
a torus @xmath35 in @xmath8 is _
@xmath3-split _ if @xmath35 is @xmath11-split and @xmath35 is contained in the intersection of two different @xmath3-rational , proper , parabolic subgroups of @xmath8 .
* from the margulis arithmeticity theorem ( * ? ? ?
* thm . 1 , p. 2 )
, we know that , after passing to a finite cover of @xmath0 ( in other words , after passing to a finite - index subgroup of @xmath3 ) , we can write @xmath284 where @xmath285 is arithmetic , @xmath286 is compact with all factors of real rank one , and each @xmath287 is noncompact with real rank one . a torus in @xmath8 is _ @xmath3-split _ if it is contained in some torus of the form @xmath288 , where @xmath289 is @xmath4-split , and each @xmath290 is @xmath291-split . * a flat @xmath12 in @xmath1 is _ @xmath3-split _ if the torus @xmath10 is @xmath3-split .
[ horolimiffnonarith ] let @xmath0 be a finite - volume locally symmetric space of noncompact type with the killing - form metric , and let @xmath2 be a point on the visual boundary of @xmath1
. then the following are equivalent : nobreaktrue 1 .
@xmath2 is a horospherical limit point for @xmath3 .
@xmath2 is not on the boundary of any @xmath3-split flat .
there does not exist a @xmath3-rational parabolic subgroup @xmath79 of @xmath8 , such that @xmath79 fixes @xmath2 , and @xmath292 fixes some ( or , equivalently , every ) horosphere based at @xmath2 .
4 . the horospheres based at @xmath2 are uniformly coarsely dense in @xmath0 .
the horoballs based at @xmath2 are uniformly coarsely dense in @xmath0 .
@xmath279 for every horoball @xmath48 based at @xmath2 , where @xmath224 is the natural covering map .
[ notkilling ] the above results apply only to the killing - form metric on @xmath13 , but it is well known that any other symmetric metric @xmath216 differs only by a scalar multiple on each irreducible factor of @xmath13 @xcite . if the endpoint of a particular geodesic ray @xmath61 is a horospherical limit point in the killing - form metric , and we let @xmath293 where @xmath294 is the scaling factor of @xmath216 on the irreducible factor @xmath295 , then the endpoint of @xmath296 is a horospherical limit point with respect to the metric @xmath216 . in fact
, it is easy to see that the two different geodesic rays ( in the two different metrics ) have exactly the same horospheres in @xmath13 .
this means that the above proofs apply in general if we replace the phrase `` @xmath4-split '' with `` @xmath4-good , '' where a torus @xmath35 is _ @xmath4-good _
if @xmath35 is contained in a maximal @xmath4-torus @xmath10 of @xmath8 , such that @xmath10 contains a maximal @xmath4-split torus of @xmath8 , and @xmath35 is orthogonal to the maximal @xmath4-anisotropic torus of @xmath10 ( cf .
[ splitperpnochar ] ) .
linear algebraic groups . _ springer , new york , _ 1991 . http://www.ams.org/mathscinet-getitem?mr=1102012[mr1102012 ] ( 92d:20001 ) , http://zbmath.org/?q=an:0726.20030[zbl 0726.20030 ] .
invariant measures and minimal sets of horospherical flows .
_ * 64 * ( 1981 ) 357385 .
( 83c:22009 ) , http://zbmath.org/?q=an:0498.58013[zbl 0498.58013 ] divergent trajectories of flows on homogeneous spaces and diophantine approximation , _ j. reine angew .
_ * 359 * ( 1985 ) 5589 .
( 87g:58110a ) , http://zbmath.org/?q=an:0578.22012[zbl 0578.22012 ] geometric limit sets of higher rank lattices
london math .
( 3 ) * 90 * ( 2005 ) 689710 .
( 2006e:22013 ) , http://zbmath.org/?q=an:1077.22015[zbl 1077.22015 ] lie groups and symmetric spaces . _
battelle rencontres : 1967 lectures in mathematics and physics , _ 171 .
_ benjamin , new york , _ 1968 .
http://www.ams.org/mathscinet-getitem?mr=0236325[mr0236325 ] ( 38 # 4622 ) , http://zbmath.org/?q=an:0177.50601[zbl 0177.50601 ] differential geometry , lie groups , and symmetric spaces . _ academic press , new york , _ 1978 .
( 80k:53081 ) , http://zbmath.org/?q=an:0451.53038[zbl 0451.53038 ] tits geometry , arithmetic groups , and the proof of a conjecture of siegel . _ j. lie theory _ * 14 * ( 2004 ) 317338 .
http://www.ams.org/mathscinet-getitem?mr=2066859[mr2066859 ] , http://zbmath.org/?q=an:1086.53073[zbl 1086.53073 ] discrete subgroups of semisimple lie groups .
_ springer , berlin , _ 1991 .
( 92h:22021 ) , http://zbmath.org/?q=an:0732.22008[zbl 0732.22008 ] horospherical limit points of @xmath35-arithmetic groups .
/ abs/1309.7113 algebraic groups and number theory . _ academic press , new york , _ 1994 .
( 95b:11039 ) , http://zbmath.org/?q=an:0841.20046[zbl 0841.20046 ] discrete subgroups of lie groups .
_ springer , new york , _ 1972 .
http://www.ams.org/mathscinet-getitem?mr=507234[mr507234 ] ( 58 # 22394a ) , http://zbmath.org/?q=an:0254.22005[zbl 0254.22005 ] raghunathan s topological conjecture and distributions of unipotent flows . _ duke math
. j. _ * 63 * ( 1991 ) 235280 .
( 93f:22012 ) , http://zbmath.org/?q=an:0733.22007[zbl 0733.22007 ] closed orbits for actions of maximal tori on homogeneous spaces . _ duke math .
j. _ * 119 * ( 2003 ) 367392 .
( 2004g:22006 ) , http://zbmath.org/?q=an:1040.22005[zbl 1040.22005 ] | suppose @xmath0 is an arithmetic locally symmetric space of noncompact type ( with the natural metric induced by the killing form of the isometry group of @xmath1 ) , and let @xmath2 be a point on the visual boundary of @xmath1 .
t.hattori showed that if each horoball based at @xmath2 intersects every @xmath3-orbit in @xmath1 , then @xmath2 is not on the boundary of any @xmath4-split flat in @xmath1 .
we prove the converse .
( this was conjectured by w.h.rehn in some special cases . )
furthermore , we prove an analogous result when @xmath3 is a nonarithmetic lattice . |
the correlation found between spiral arm pitch angle and rotation curve shear rate ( seigar , block & puerari 2004 ; seigar et al .
2005 ) suggests that there is a link between the tightness of spiral structure and the central mass concentration in spiral galaxies .
the shear rate , @xmath8 , a dimensionless quantity , can be measured directly from rotation curves and is defined as follows , @xmath9 where a is the first oort constant , @xmath10 is the angular velocity and @xmath11 is the velocity at a radius @xmath12 .
the shear rate depends upon the shape of the rotation curve . for a rotation curve that remains flat
@xmath13 , for a falling rotation curve @xmath14 and for a continually rising rotation curve @xmath15 .
as the shape of a rotation curve depends on the mass distribution , the shear rate at any given position depends upon the mass within that radius , or the central mass concentration . as a result ,
the spiral arm pitch angle is dependent upon the central mass concentration , and this is consistent with the expectations of most spiral density wave models ( e.g. bertin et al .
1989a , b ; bertin 1991 , 1993 , 1996 ; bertin & lin 1996 ; fuchs 1991 , 2000 ) , although density wave models predict that pitch angles also depend on stability ( i.e. the toomre @xmath16-parameter ) . the modal theory of spiral structure ( bertin et al .
1989a , b ; bertin 1991 , 1993 , 1996 ; bertin & lin 1996 ) predicts that the tightness of the arms comes from the central mass concentration .
galaxies with higher central mass concentrations , i.e. higher overall densities ( including dark matter ) , are predicted to have more tightly wound spiral arms .
the models of fuchs ( 1991 , 2000 ) result in disks with rigidly rotating spiral modes , wherein bulges act as inner reflectors of waves or modes induced by the swing amplification method , thus leading to modal spiral waves , which form as a result of toomre swing amplification ( toomre 1981 ) .
fuchs ( 2000 ) adopts a stellar - dynamical analogue ( julian & toomre 1966 ) of the goldreich & lynden - bell ( 1965 ) sheet , which describes the local dynamics of an unbounded patch of thin , differentially rotating stellar disk .
he then increases @xmath16 in the goldreich & lynden - bell ( 1965 ) sheet , which has the affect that the bulge acts as an inner boundary .
the result is that , instead of shearing density waves ( as in the unbounded sheet ) , spiral modes appear .
these models best show how central mass concentration correlates with spiral arm pitch angle .
if the disk is very light ( low @xmath17 where @xmath17 is the disk density ) the mode can be very tight , and we are in the domain of small epicycles .
formally , if the stability parameter @xmath18 is close to unity , the value of @xmath19 must also be small , where @xmath19 is the radial velocity dispersion and @xmath20 is the epicyclic frequency .
if we increase the relative mass of the disk , we find a trend towards more open structures .
recent observational studies of spiral structure have highlighted a difference in the morphologies seen in the optical compared with the near - infrared .
the classification of galaxies by hubble type ( hubble 1926 ) is performed in the optical regime , where dust extinction still has a large effect , and where the light is dominated by the young population i stars . in the near - infrared dust extinction
is minimized and the light is dominated by old and intermediate age stars ( rhoads 1998 ; worthey 1994 ) .
previous work has found that optically determined hubble type is not correlated with the near - infrared morphology .
there is not a good correlation between hubble type and near - infrared bulge - to - disk ratio or near - infrared spiral arm pitch angle ( de jong 1996 ; seigar & james 1998a , b ) .
however , even in the optical regime , the correlation between quantitative pitch angle and hubble type is weak ( kennicutt 1981 ) , even though the tightness of the spiral arm pattern , as judged by eye , is one of the defining hubble morphological criteria .
furthermore , it has been shown that near - infrared morphologies of some spiral galaxies can be significantly different from their optical morphologies ( block & wainscoat 1991 ; block et al .
1994 ; thornley 1996 ; seigar & james 1998a , b ; seigar , chorney & james 2003 ) .
galaxies with flocculent spiral structure in the optical may present grand - design spiral structure in the near - infrared ( thornley 1996 ; seigar et al . 2003 ) .
these results suggest that the optical morphology gives incomplete information about the nature of spiral structure in disk galaxies .
this is the basis of the near - infared studies of block et al .
( 1999 ) and seigar et al .
( 2005 ) . however , work by eskridge et al .
( 2002 ) showed that when assigning a hubble type to galaxies in several different wavebands , from the optical to the near - infrared , a correlation exists between the optically determined hubble type and that determined in the near - infrared .
this suggests that , while there may be large small - scale differences between optical and near - infrared morphologies of spiral structure , the large - scale differences are minimal . as a result
, it may be possible that there is no significant difference between the large - scale spiral arm pitch angle , regardless of whether it is determined in the optical or near - infrared for the regular grand - design spirals investigated here .
it should be noted that for flocculent spiral structure it is difficult to determine a pitch angle for the two - armed component in the optical , while a @xmath21 grand - design structure may be seen in the near - infrared ( e.g. thornley 1996 ; grosbol & patsis 1998 ; seigar , chorney & james 2003 ) . as a result
it may be possible to use optical imaging to expand the sample of galaxies having quantitative measurements of spiral arm pitch angle and therefore strengthen the correlation between spiral arm pitch angle and rotation curve shear rate reported by seigar et al .
( 2005 ) .
in addition to providing an important constraint on models of spiral arm formation , the correlation between spiral arm pitch angle and shear rate opens up a fundamentally new approach for probing mass distributuions in spiral galaxies .
this approach relies on imaging data alone without the need for full rotation curve information .
specifically , the tully - fisher relation ( tully & fisher 1977 ) for spiral galaxies coupled with the shear rate - pitch angle relation can be used to determine a rotation curve normalization _ and _ slope . in this paper
we explicitly demonstrate how , given a bulge - disk decomposition , a pitch angle determination , and a tully - fisher normalizaion , one can constrain galaxy mass distributions , dark matter halo concentrations , and other galaxy formation parameters .
we obtain our constraints within the context of the the standard framework of disk formation put forth by fall & efstathiou ( 1980 ) and blumenthal et al .
( 1986 ) . in principle
, the technique we demonstrate here can be applied generally to a large sample of galaxies and to galaxies at high redshift , when spiral arms are detected .
this is the first in a series of papers in which we use spiral arm pitch angles to determine the mass distribution in spiral galaxies . in this paper
we determine mass concentrations in a few disk galaxies from their spiral arm pitch angles ( or shear rates ) and their disk masses and scalelengths ( determined via a bulge - disk decomposition technique ) . in future papers
we will apply the outline methods in this paper to a large sample of nearby disk galaxies and to a large sample of galaxies at higher redshift in order to determine the evolution in the central mass concentrations of disk galaxies as a function of look - back time .
this paper is arranged as follows . in section 2
we describe the data used and the methods for determining spiral arm pitch angles and rotation curve shear rates . in section 3 we compare pitch angles in the near - infrared and optical . in section 4
we discuss the relationship between spiral arm pitch angle and rotation curve shear rate found by seigar et al .
( 2005 ) and we use more data to strengthen this correlation . in section 5
we use a bulge - disk decomposition routine and an adiabatic infall model for disk galaxy formation to investigate the link between shear rate and central mass concentration and we summarize our findings in section 6 .
we have made use of @xmath22 and @xmath2 band images of 57 galaxies from the ohio state university bright spiral galaxy survey ( osubsgs ; e.g. eskridge et al .
these galaxies were chosen to be as face - on as possible ( with a ratio of the minor- to major - axis @xmath23 ) so that a comparison of spiral arm morphology could be made in the optical and near - infrared .
we also include @xmath2 band images of 31 galaxies that also have measured rotation curves ( mathewson et al .
1992 ; persic & salucci 1995 ) .
we observed these galaxies as part of an ongoing survey of the 600 brightest southern hemisphere galaxies .
the observations make use of the ccd direct imaging camera at the 2.5-m du pont telescope at the las campanas observatory in chile .
integration times used were @xmath24 seconds . for this work , we selected from the full imaging dataset a sample of galaxies having a ratio of the minor- to major - axis @xmath25 , so that spiral arm morphology can be studied , and rotation curves are also available . of these galaxies , 6 also have @xmath26 band data .
the @xmath26 band images were observed using the wide - field infrared camera ( wirc ) , also at the 2.5-m du pont telescope .
a 5-point dither pattern was used with integration times of @xmath27 seconds in each position .
the total exposure was therefore 900 seconds .
we also include @xmath2 and @xmath26 band images of a further 3 galaxies from our survey , which do not have rotation curves .
the selection for these galaxies satisfies the same selection criteria that was used to select the osubsgs galaxies above .
llcccll [ tab1 ] ngc 150 & sbbc & 0.50 & 118 & 70126 ; 65131 ; 75121 & [email protected] & [email protected] + ngc 157 & sabbc & 0.64 & 30 & 65121 ; 60126 ; 70116 & [email protected] & [email protected] + ngc 289 & sabbc & 0.71 & 130 & 40150 ; 35155 ; 45145 & [email protected] & [email protected] + ngc 578 & sabc & 0.63 & 110 & 65125 ; 60130 ; 70120 & [email protected] & [email protected] + ngc 613 & sbbc & 0.76 & 120 & 73135 ; 68140 ; 78130 & [email protected] & [email protected] + ngc 864 & sabc & 0.74 & 20 & 55143 ; 50148 ; 60138 & [email protected] & [email protected] + ngc 1073 & sbc & 0.92 & 15 & 73183 ; 68188 ; 78178 & [email protected] & [email protected] + ngc 1087 & sabc & 0.59 & 5 & 57147 ; 52152 ; 62142 & [email protected] & [email protected] + ngc 1187 & sbc & 0.75 & 130 & 90132 ; 85137 ; 95127 & [email protected] & [email protected] + ngc 1241 & sbb & 0.61 & 140 & 2551 ; 2056 ; 3046 & [email protected] & [email protected] + ngc 1300 & sbbc & 0.66 & 106 & 80116 ; 75121 ; 85111 & [email protected] & [email protected] + ngc 1350 & sbab & 0.54 & 0 & 7587 ; 7290 ; 7884 & [email protected] & [email protected] + ngc 1792 & sabc & 0.50 & 137 & 45211 ; 40216 ; 50206 & [email protected] & [email protected] + ngc 2090 & sab & 0.50 & 13 & 64272 ; 59277 ; 69267 & [email protected] & [email protected] + ngc 2139 & sabcd & 0.73 & 140 & 50118 ; 45123 ; 55113 & [email protected] & [email protected] + ngc 2964 & sabbc & 0.55 & 97 & 70162 ; 65 - 167 ; 75157 & [email protected] & [email protected] + ngc 3223 & sabc & 0.61 & 135 & 3274 ; 2779 ; 3769 & [email protected] & [email protected] + ngc 3261 & sbbc & 0.76 & 85 & 3090 ; 2595 ; 3586 & [email protected] & [email protected] +
ngc 3338 & sac & 0.61 & 100 & 74164 ; 69169 ; 79159 & [email protected] & [email protected] + ngc 3507 & sbb & 0.85 & 110 & 120196 ; 115201 ; 125191 & [email protected] & [email protected] + ngc 3513 & sbc & 0.79 & 75 & 80130 ; 75135 ; 85125 & [email protected] & [email protected] + ngc 3583 & sbb & 0.64 & 125 & 6084 ; 5787 ; 6381 & [email protected] & [email protected] + ngc 3646 & sabc & 0.56 & 50 & 2052 ; 1557 ; 2547 & [email protected] & [email protected] + ngc 3686 & sbbc & 0.78 & 15 & 99169 ; 94 - 174 ; 104164 & [email protected] & [email protected] + ngc 3726 & sabc & 0.69 & 10 & 100258 ; 95263 ; 105253 & [email protected] & [email protected] + ngc 3810 & sac & 0.70 & 15 & 112200 ; 107205 ; 117195 & [email protected] & [email protected] + ngc 3887 & sbbc & 0.76 & 20 & 93163 ; 88168 ; 98158 & [email protected] & [email protected] + ngc 3893 & sabc & 0.62 & 165 & 102218 ; 97223 ; 107213 & [email protected] & [email protected] + ngc 4027 & sbdm & 0.75 & 167 & 60126 ; 55131 ; 65121 & [email protected] & [email protected] + ngc 4030 & sabc & 0.71 & 27 & 51161 ; 46166 ; 56156 & [email protected] & [email protected] + ngc 4051 & sabbc & 0.75 & 135 & 185257 ; 180262 ; 190267 & [email protected] & [email protected] + ngc 4145 & sabd & 0.73 & 100 & 93211 ; 88216 ; 98206 & [email protected] & [email protected] + ngc 4414 & sac & 0.56 & 155 & 142290 ; 137295 ; 147285 & [email protected] & [email protected] + ngc 4548 & sbb & 0.80 & 150 & 250386 ; 245391 ; 255381 & [email protected] & [email protected] + ngc 4580 & saba & 0.76 & 165 & 82218 ; 77223 ; 87213 & [email protected] & [email protected] + ngc 4654 & sabcd & 0.57 & 128 & 99199 ; 94204 ; 104 - 194 & [email protected] & [email protected] + ngc 4930 & sbbc & 0.82 & 40 & 4377 ; 3882 ; 4872 & [email protected] & [email protected] + ngc 4939 & sabc & 0.51 & 10 & 2476 ; 1981 ; 2971 & [email protected] & [email protected] + ngc 4995 & sabb & 0.64 & 95 & 31145 ; 26150 ; 36140 & [email protected] & [email protected] + ngc 5054 & sabc & 0.59 & 160 & 51127 ; 46132 ; 56122 & [email protected] & [email protected] + ngc 5085 & sac & 0.88 & 38 & 20138 ; 15143 ; 25133 & [email protected] & [email protected] +
ngc 5247 & sabc & 0.88 & 20 & 47181 ; 42186 ; 52176 & [email protected] & [email protected] + ngc 5371 & sabbc & 0.80 & 8 & 4278 ; 3783 ; 4773 & [email protected] & [email protected] + ngc 5483 & sbc & 0.92 & 25 & 50124 ; 55119 ; 45129 & [email protected] & [email protected] + ngc 5921 & sbbc & 0.82 & 130 & 70140 ; 65145 ; 75135 & [email protected] & [email protected] +
ngc 6215 & sac & 0.86 & 78 & 47141 ; 42146 ; 52136 & [email protected] & [email protected] + ngc 6221 & sbbc & 0.71 & 5 & 73135 ; 68140 ; 78130 & [email protected] & [email protected] + ngc 6300 & sbb & 0.67 & 118 & 99179 ; 94184 ; 104174 & [email protected] & [email protected] + ngc 6384 & sabbc & 0.66 & 30 & 60126 ; 55131 ; 65121 & [email protected] & [email protected] + ngc 6907 & sbbc & 0.82 & 46 & 4058 ; 3860 ; 4256 & [email protected] & [email protected] + ngc 7083 & sabc & 0.59 & 5 & 3278 ; 2981 ; 3575 & [email protected] & [email protected] + ngc 7412 & sabc & 0.74 & 65 & 5090 ; 4595 ; 55 - 85 & [email protected] & [email protected] + ngc 7418 & sabcd & 0.74 & 139 & 63151 ; 58156 ; 68146 & [email protected] & [email protected] + ngc 7479 & sbc & 0.76 & 25 & 4882 ; 4585 ; 5179 & [email protected] &
[email protected] + ngc 7552 & sbab & 0.79 & 1 & 63133 ; 58138 ; 68128 & [email protected] & [email protected] + ngc 7723 & sbb & 0.66 & 40 & 72104 ; 67109 ; 7799 & [email protected] & [email protected] + ngc 7741 & sbcd & 0.68 & 170 & 165247 ; 160252 ; 170242 & [email protected] & [email protected]
+ spiral arm pitch angles are measured using the same technique employed by seigar et al .
a two - dimensional fast - fourier decomposition technique is used , which employs a program described by schrder et al .
logarithmic spirals are assumed in the decomposition .
the resulting pitch angles are listed in tables 1 , 2 and 3 .
llcccll [ tab1a ] ic 2522 & sacd & 0.71 & 0 & 2577 ; 2082 ; 3072 & [email protected] & [email protected] + ngc 1964 & sabb & 0.38 & 32 & 21175 ; 16180 ; 26170 & [email protected] & [email protected] + ngc 2082 & sabc & 0.94 & 60 & 50130 ; 45135 ; 55 - 125 & [email protected] & [email protected] + ngc 2280 & sacd & 0.49 & 163 & 55117 ; 50122 ; 60112 & [email protected] & [email protected] + ngc 2417 & sabbc & 0.68 & 81 & 2571 ; 2076 ; 3066 & [email protected] & [email protected] + ngc 2935 & sabb & 0.78 & 0 & 4393 ; 3898 ; 4888 & [email protected] & [email protected] +
ngc 3318 & sabb & 0.54 & 78 & 2389 ; 1894 ; 2884 & [email protected] & [email protected]
+ ngc 3450 & sbb & 0.88 & 140 & 2848 ; 2551 ; 3145 & [email protected] & [email protected] + ngc 4050 & sbab & 0.68 & 85 & 7091 ; 6794 ; 7388 & [email protected] & [email protected] + the amplitude of each fourier component is given by @xmath29}}{\sum_{i=1}^{i}\sum_{j=1}^{j}i_{ij}(\ln{r},\theta)},\ ] ] where @xmath30 and @xmath31 are polar coordinates , @xmath32 is the intensity at position @xmath33 , @xmath34 represents the number of arms or modes , and @xmath35 is the variable associated with the pitch angle @xmath36 , defined by @xmath37 . throughout this work we measure the pitch angle @xmath36 of the @xmath21 component .
the resulting pitch angle measured using equation [ fft ] is in radians , and this is later converted to degrees for ease of perception .
the range of radii over which the fourier fits were applied were selected to exclude the bulge or bar ( where there is no information about the arms ) and to extend out to the outer limits of the arms in our images , in such a way that the 10 kpc radius fell approximately in the middle of this range .
the radial extent of the bar was measured manually ( see e.g. grosbol , patsis & pompei 2004 ) , and the inner radial limit applied to the fft was chosen to be outside this radius .
physical distances are calculated using a hubble constant @xmath38 km s@xmath39 mpc@xmath39 and recessional velocities from the nasa extragalactic database ( ned ) .
pitch angles are then determined from peaks in the fourier spectra , as this is the most powerful method to find periodicity in a distribution ( considre & athanassoula 1988 ; garcia - gomez & athanassoula 1993 ) .
the radial range over which the fourier analysis was performed was chosen by eye and is probably the dominant source of error in the calculation of pitch angles , as spiral arms are only approximately logarithmic and sometimes abrupst changes can be seen in spiral arm pitch angles ( e.g. seigar & james 1998b ) as a result , three radial ranges were chosen for each galaxy , and a mean pitch angle and standard error calculated for every object .
the images were first projected to face - on .
mean uncertainties of position angle and inclination as a function of inclination were discussed by considre & athanassoula ( 1988 ) . for a galaxy with low inclination , there are clearly greater uncertainties in assigning both a position angle and anaccurate inclination .
these uncertainties are discussed by block et al .
( 1999 ) and seigar et al .
( 2005 ) , who take a galaxy with low inclination ( @xmath40 ) and one with high inclination ( @xmath41 ) and varied the inclination angle used in the correction to face - on .
they found that for the galaxy with low inclination , the measured pitch angle remained the same .
however , the measured pitch angle for the galaxy with high inclination varied by @xmath42% . since inclination corrections are likely to be largest for galaxies with the highest inclinations cases where inclination is @xmath41 are taken as the worst case scenario .
for galaxies with inclination @xmath43 we take into account this uncertainty .
our deprojection method assumes that spiral galaxy disks are intrinsically circular and flat in nature .
31 of the galaxies observed here have h@xmath44 rotation curve data measured by mathewson et al .
( 1992 ) and persic & salucci ( 1995 ) .
these rotation curves are of good quality with an rms error @xmath45 km s@xmath39 , and an error associated with folding the two sides of the galaxy also @xmath45 km s@xmath39 .
these rotation curves have been used to estimate the shear rates in these galaxies , using the same method used by other authors ( e.g. block et al .
1999 ; seigar et al . 2005 ; seigar 2005 ) .
using equation [ shearrate ] , we have calculated the shear rates for these galaxies , over the same radial ranges for which the fourier analysis was performed and pitch angles calculated .
we have selected several different radial ranges , just as in the fourier analysis , and we present mean shear rates and standard errors .
the dominant sources of error on the shear rate are the rms error in the rotation curve and the error associated with folding the two sides of the galaxy .
this is typically @xmath45% . in order to calculate the shear rate ,
the mean value of @xmath46 measured in km s@xmath39 arcsec@xmath39 is calculated by fitting a line of constant gradient to the outer part of the rotation curve ( i.e. past the radius of turnover and any bar or bulge that may exist in the galaxy ) .
mean shear rates are then calculated from shear rates measured over three radial ranges , corresponding to the same radial ranges over which the fourier analysis was performed .
the resulting shear rates are listed in table 3 .
the choice of using 10 kpc as the radius at which to measure the shear and spiral arm pitch angle is somewhat arbitrary .
the correlation between these two quantities ( seigar et al .
2005 and figure 3 ) only exists when they are measured at a physical radius , chosen indepently of the disk scalelength .
if we measure pitch angles and shear at a radius chosen using the disk scalelength , the correlation no longer exists , and the range of shear is narrower ( between 0.4 and 0.6 ) . as a result , we chose to measure shear and pitch angles at a radius independent of the disk scalelength .
we chose 10 kpc ( although it could just as easily be 8 kpc or 12 kpc ) .
the fact that this correlation only exists for a physical radius , suggests that other quantities may be important in the determination of shear and pitch angle ( e.g. the disk scalelength itself or the stability of the disk as a function of radius ) .
indeed , when modeling galaxies we use a 1-dimensional bulge - disk decomposition and use the derived disk scalelength in our codes .
figure [ opt - ir - pitch ] shows of plot of @xmath2 band pitch angle versus near - infrared pitch angle for the 57 face - on and nearly face - on galaxies taken from the osubsgs and 9 face - on and nearly face - on galaxies taken from our survey .
this plot shows a very tight 1:1 correlation between the optical and near - infrared spiral arm pitch angles ( correlation coefficient=0.95 ; [email protected]% ) , showing that pitch angles are similar whether measured in the optical or near - infrared regimes . in this plot ,
only one galaxy has a pitch angle that shows a difference of @xmath48 when measured at @xmath2 compared with its near - infrared measurement .
the outlying galaxy is ngc 613 , which appears to have a strong 4-armed component in the optical , whereas this is a much weaker feature in the near - infrared .
this result is in contrast to the claims of block et al .
( 1999 ) and block & puerari ( 1999 ) , who argue that there is usually a large difference between the spiral arm pitch angle when measured in the optical and the near - infrared .
our result is consistent with the result that hubble types assigned to galaxies in the optical and the near - infrared correlate well with each other ( eskridge et al .
2002 ) .
we now analyze the absolute difference between the pitch angles measured the the @xmath2 band and those measured in the near - infrared .
the mean of this difference in pitch angles is @xmath1 , where the error is a standard deviation , and this is therefore consistent with the same pitch angle being measured , regardless of the waveband .
the mode of the distribution is in the range [email protected] and the median is 1.46 .
furthermore a linear fit to the data in fig [ opt - ir - pitch ] yields the following relationship , @xmath50 where @xmath51 is the pitch angle measured in the near - infrared and @xmath52 is the pitch angle measured in the optical . the point at which this line of best fit intersects the @xmath51-axis is consistent with zero , and the gradient of the line is consistent with unity , and so the line of best fit is consistent with a 1:1 correlation between spiral arm pitch angles in the optical and near - infrared .
it should be noted that although the difference between optical and near - infrared spiral arm pitch angles is consistent with zero from a statistical analysis , minor systematic differences can not be excluded .
small differences between optical and near - infrared pitch angles have been measured of @xmath53 ( e.g. grosbol & patsis 1998 ) .
furthermore , color gradients across arms may be caused by such small differences in pitch angles ( e.g. gonzalez & graham 1996 ) .
of course , it is still a well - known fact that spiral structure measured in the near - infrared appears much smoother than that measured in the optical ( e.g. block & wainscoat 1991 ; block et al .
1994 ; thornley 1996 ; seigar et al .
2003 ) . even in galaxies that appear grand - design in the optical , small spokes and bifurcations
are clearly seen along the length of the dominant arms .
these spokes and bifurcations all but disappear in the near - infrared .
however , our result suggests that the pitch angle of the large - scale spiral structure remains constant with wavelength in the range 0.4@[email protected]@xmath0 m for spirals with a relatively strong two - armed pattern in the optical . for weaker , more flocculent galaxies there
may well be significant differences due to dust .
this has serious implications for the physical meaning of the dust - penetrated classification scheme ( block & puerari 1999 ; block et al .
these implications are discussed in section [ next ] .
the correlation between spiral arm pitch angle and rotation curve shear rate presented by seigar et al .
( 2005 ) only had 17 data points , including 3 from an earlier study by block et al .
( 1999 ) . here
we include a further 31 galaxies with measure pitch angles and shear rates .
figure [ shear - pitch ] shows a plot of shear rate versus spiral arm pitch angle .
a very good correlation still exists ( correlation coefficient=0.89 ; significance=99.75% ) .
the shape of a rotation curve , beyond the turnover radius , is determined by the amount and distribution of matter contained in a galaxy .
the increase of shear rates from low to high dictates a change in mass distributions from small to large central mass concentrations .
the correlation with pitch angle is therefore interpreted as follows : galaxies with higher rates of shear present a larger central mass concentration and more tightly wound arms .
in contrast , open arm morphologies are associated with lower rates of galactic shear and lower central mass concentrations .
this correlation was alluded to by the pioneering work of lin & shu ( 1964 ) and by the later spectroscopic study of burstein & rubin ( 1985 ) .
it is in agreement with modal theories of spiral structure ( e.g. bertin et al .
1989a , b ; bertin & lin 1996 ) and other numerical models based on the modal theory ( e.g. fuchs 1991 , 2000 ) .
llcccll [ tab2 ] eso 9-g10 & sac & 0.76 & 171 & 27101 ; 22 - 106 ; 3296 & [email protected] & [email protected] + eso 121-g26 & sbc & 0.63 & 115 & 5185 ; 4690 ; 5680 & [email protected] & [email protected] + eso 582-g12 & sac & 0.64 & 48 & 32102 ; 27107 ; 3797 & [email protected] & [email protected] + ic 2522 & sacd & 0.71 & 0 & 2577 ; 2082 ; 3072 & [email protected] & [email protected] + ic 2537 & sabc & 0.65 & 26 & 3082 ; 2587 ; 3577 & [email protected] & [email protected] + ic 3253 & sac & 0.39 & 23 & 2292 ; 1797 ; 2787 & [email protected] & [email protected] + ic 4538 & sabc & 0.77 & 50 & 4266 ; 3771 ; 4761 & [email protected] & [email protected] + ic 4808 & sac & 0.42 & 45 & 1545 ; 1050 ; 2040 & [email protected] & [email protected] + ngc 150 & sbbc & 0.50 & 118 & 70126 ; 65131 ; 75121 & [email protected] & [email protected] + ngc 151 & sbbc & 0.42 & 75 & 3052 ; 2755 ; 3349 & [email protected] & [email protected] + ngc 578 & sabc & 0.63 & 110 & 65125 ; 60130 ; 70120 & [email protected] & [email protected] + ngc 908 & sac & 0.43 & 75 & 56150 ; 51155 ; 61145 & [email protected] & [email protected] + ngc 1232 & sabc & 0.88 & 108 & 53141 ; 48146 ; 58136 & [email protected] & [email protected] +
ngc 1292 & sac & 0.43 & 7 & 53173 ; 48178 ; 58168 & [email protected] & [email protected] + ngc 1300 & sbbc & 0.66 & 106 & 80116 ; 75121 ; 85111 & [email protected] & [email protected] + ngc 1353 & sabc & 0.41 & 138 & 63139 ; 58144 ; 68134 & [email protected] & [email protected] + ngc 1365 & sbb & 0.55 & 32 & 61129 ; 56134 ; 66124 & [email protected] & [email protected] + ngc 1559 & sbcd & 0.57 & 64 & 82156 ; 77161 ; 87151 & [email protected] & [email protected] + ngc 1566 & sabbc & 0.80 & 60 & 73133 ; 68138 ; 78128 & [email protected] & [email protected] + ngc 1964 & sabb & 0.38 & 32 & 21175 ; 16180 ; 26170 & [email protected] & [email protected] + ngc 2280 & sacd & 0.50 & 163 & 55117 ; 50122 ; 60112 & [email protected] & [email protected] + ngc 2417 & sabbc & 0.68 & 81 & 2571 ; 2076 ; 3066 & [email protected] & [email protected] + ngc 2835 & sabc & 0.67 & 8 & 110240 ; 105245 ; 115235 & [email protected] & [email protected] + ngc 2935 & sabb & 0.78 & 0 & 4393 ; 3898 ; 4888 & [email protected] & [email protected] + ngc 3052 & sabc & 0.65 & 102 & 2557 ; 2062 ; 3052 & [email protected] & [email protected] + ngc 3054 & sabbc & 0.61 & 118 & 4484 ; 3989 ; 4979 & [email protected] & [email protected] + ngc 3223 & sabc & 0.61 & 135 & 3274 ; 2779 ; 3769 & [email protected] & [email protected] +
ngc 3318 & sabb & 0.54 & 78 & 2389 ; 1894 ; 2884 & [email protected] & [email protected] + ngc 5967 & sabc & 0.59 & 90 & 4676 ; 4181 ; 5171 & [email protected] & [email protected] + ngc 7083 & sabc & 0.59 & 5 & 3278 ; 2981 ; 3575 & [email protected] & [email protected] + ngc 7392 & sbab & 0.62 & 123 & 3066 ; 2571 ; 3561 & [email protected] & [email protected] + the correlations shown in figures [ opt - ir - pitch ] and [ shear - pitch ] have significant implications for the dust - penetrated classification scheme introduced by block & puerari ( 1999 ) .
these authors argue that their dust - penetrated classification scheme is needed , because spiral arm pitch angles are different in the near - infrared when compared to the optical , and as a result they classify galaxies into three bins , based on their pitch angles when measured in the near - infrared .
however , we have shown ( in figure [ opt - ir - pitch ] ) that spiral galaxies with a relatively strong two - armed spiral pattern , have optical and near - infrared pitch angles that are nearly always very similar .
this , therefore , seems to lessen the need for the dust - penetrated classification scheme . in figure
[ shear - pitch ] , however , we show that a correlation still exists between shear rate and spiral arm pitch angle , no matter what waveband the pitch angle is measured in .
this suggests that classifying galaxies based on their pitch angles has some physical basis , because it tells us something about their central mass concentrations .
given the correlation between shear rate and pitch angle we derive the following expression relating these terms @xmath55 where @xmath36 is the pitch angle in degrees and @xmath8 is the shear rate .
our goal now is to demonstrate how spiral arm pitch angle measurements coupled with a bulge - disk decomposition can be used to constrain the overall mass distribution in spiral galaxies .
we outline the method for two example galaxies with pseudo - bulges ( see kormendy & kennicutt 2004 for a review about pseudo - bulges ) .
the two galaxies we chose are chosen such that one galaxy , eso 582-g12 , has a flat rotation curve , with a shear , @xmath56 in the middle of the range , and the other galaxy , ic 2522 , has a continually rising rotation curve , with a shear , @xmath57 , towards the low end of the range of shears . as a normalization for each galaxy we take its measured rotation speed at 2.2 disk scale lengths , @xmath58 ( listed in the last column of table 4 ) , and include a bulge - disk decomposition in order to estimate the baryonic contribution to each rotation curve . in future work paper
we plan to use the same methodology for a large sample of galaxies .
for each of our galaxies we produce surface brightness profiles using the iraf * ellipse * routine , which fits ellipses to isophotes in an image , using an iterative method described by jedrzejewski ( 1987 ) . from the surface brightness profile
we then determine the disk and bulge b - band luminosity using an exponential disk and an exponential ( pseudo ) bulge . in future papers
a more general srsic parameterization will be used for the bulge component .
we utilize a 1-d bulge - disk decompostion routine , which performs levenberg - marquardt least - squares minimization .
explicity we fit an exponential profile for each bulge via @xmath59 ) } , \label{bulge}\ ] ] where @xmath60 is the effective radius , containing 50% of the total light of the bulge and @xmath61 is the surface brightness at @xmath60 .
we fit the disk component using @xmath62 where @xmath63 is the central surface brightness and @xmath64 is the scalelength of the disk . the results of this bulge - disk decompostion can be seen in figure [ decompose ] and table 4 .
we assign masses to the disk and bulge components using a range of stellar mass - to - light ratios from bell & de jong ( 2001 ) .
specifically , in our rotation curve models we allow mass - to - light ratios of @xmath65 and @xmath66 ( measured in b band solar units ) , and use our photometrically - derived disk and bulge light profiles @xmath67 to determine the stellar mass contributuion to each rotation curve : @xmath68 .
we now explore a range of allowed dark matter halo masses and density profiles , adopting two extreme models for disk galaxy formation . in the first , we assume that the dark matter halos surrounding these galaxies have not responded significantly to the formation of the disks , i.e. adiabatic contraction ( ac ) does not occur .
we refer to this as our `` non - ac '' model . in this case , the dark matter contributuion to each galaxy rotation curve is described by a density profiles that mirrors those found in dissipationless dark matter simulations : @xmath69 where @xmath70 is a characteristic inner " radius , and @xmath71 a corresponding inner density . here
we have adopted the profile shape of navarro , frenk & white ( 1996 ; hereafter nfw ) .
the nfw profile is a two - parameter function and is completely specified by choosing two independent paramters , e.g. the virial mass @xmath72 ( or virial radius @xmath73 ) and concentration @xmath74 define the profile completely ( see bullock et al . 2001b , for a discussion ) .
similarly , given a virial mass @xmath72 and the dark matter circular velocity at any radius , the halo concentration @xmath75 is completely determined .
in the second class of models we adopt the scenario of `` adiabatic contraction '' ( ac ) discussed by blumenthal et al .
( 1986 ; see also bullock et al .
2001a , pizagno et al .
2005 ) . here
we assume that the baryons and dark matter initially follow an nfw profile and that the baryons cool and settle into the halo center slowly compared to a typical orbital time .
this slow infall provokes an adiabatic contraction in the halo density distributuion and gives rise to a more concentrated dark matter profile .
the idea of adiabatic contraction was originally discussed as to explain the `` conspiracy '' between dark halos and disk sizes which gives rise to a featureless rotation curve ( rubin et al .
1985 ) but has since proven to be remarkably accurate in describing the formation of disk galaxies in numerical simulations ( e.g. gnedin et al .
2004 , and references therein ) , although the degree to which this process operates in the real universe is currently uncertain . for example , dutton et al .
( 2005 ) , showed that adiabatic contraction models are inconsistent with the rotation curves measured , and the expected nfw concentrations , for a sample of 6 galaxies .
they suggest that mechanisms such as stellar feedback and stellar bars may result in less concentrated halos than predicted by adiabatic concentration . in our ac model , we take the contraction into account following the prescription of blumenthal et al .
note that gnedin et al .
( 2004 ) advocate a slightly modified prescription , but the differences between the two methods are small compared to the differences between our `` ac '' model and our non - ac model . in principle , any observational probe that can distinguish between ac and non - ac type scenarios provides an important constraint on the nature of gas infall into galaxies ( i.e. was it fast or was it slow ? ) . for each galaxy we iterate over the central and @xmath76 values found in the bulge - disk decompositions for @xmath64 and @xmath77 and explore the three values of mass - to - light ratio discussed above @xmath78 and @xmath66 .
in each case we assume that @xmath79 remains constant with radius . for each choice of bulge - disk model parameters and mass - to - light ratios
we allow the ( initial ) halo nfw concentration parameter to vary over the range of viable values : @xmath80 ( bullock et al .
we then determine the halo virial mass @xmath81 necessary to reproduce the value of @xmath58 for the galaxy and determine the implied fraction of the mass in the system in the form of stars compared to that `` expected '' from the universal baryon fraction : @xmath82 .
we make the ( rather loose ) demand that @xmath83 lies within the range of plausible values @xmath84 .
llllllc [ tab4 ] eso 582-g12 & [email protected] & [email protected] & [email protected] & 0.106 & [email protected] & 145 + ic 2522 & [email protected] & [email protected] & [email protected] & 0.164 & [email protected] & 216 + for each chosen value of @xmath85 and adopted disk formation scenario ( ac or non - ac ) the chosen @xmath58 constraint defines the rotation curve completely , and thus provides an implied shear rate at every radius .
the three panels of figures [ eso582_m ] and [ ic2522_m ] show the results of this exercise for eso 582-g12 and ic 2522 respectively . in each panel ,
open symbols are for the non - ac model and the solid symbols reflect the ac assumption .
each point represents a distinct input combination of @xmath64 , @xmath77 , and @xmath86 .
the measured shear rate is illustrated by a solid vertical line and the @xmath76 range in the observe shear rate for each galaxy is shown by the two vertical dashed lines in each panel .
consider first the left hand panel of figure [ eso582_m ] .
here we plot the dark matter halo concentration parameter versus the shear rate measured at @xmath87 kpc .
more concentrated halos generally produce higher shear rates , as expected .
it can be seen that for a given nfw concentration , @xmath88 , several values of shear are possible .
this is due to changes in the baryon contrbution ( i.e. the disk mass and disk scalelength ) to the rotation curve .
whether an increase in the baryon contribution causes the shear to increase or decrease depends on the size of the disk ( i.e. the disk scalelength ) .
the same is true for figure 6 . in the ac model ( solid )
, @xmath75 refers to the halo concentration _ before _ the halo is adiabatically contracted .
this is why the ac points tend to have higher shear values at fixed @xmath75 compared to the non - ac case .
every point ( or @xmath64 , @xmath77 , @xmath86 input combination ) has an associated stellar baryon fraction @xmath83 and dark halo virial mass @xmath72 .
these values are plotted versus shear rate in the middle and right - hand panels in figure [ eso582_m ] .
observe that with only the @xmath58 constraint imposed ( all points ) , then a wide range of dark matter halo properties are allowed .
once we constrain the models by forcing the predicted shear to be consistent the observed range @xmath89 , we favor models that include adiabatic contraction . the dark matter of halo of eso 582-g12 is constrained to be of relatively low mass , @xmath90m@xmath91 , and to have a fairly high nfw concentration , @xmath92 . this is consistent with the high - end of the the @xmath75 distribution predicted for lcdm @xcite .
the stellar mass in the galaxy is between @xmath93 and @xmath94% of the the baryonic mass associated with its dark matter halo .
figure [ ic2522_m ] shows analogous results for ic 2522 . unlike eso 582 g12
, this galaxy prefers a more massive , low - concentration halo : @xmath95m@xmath91 with @xmath96 , at the low end of the @xmath85 distribution @xcite .
the implied stellar mass is quite small : @xmath7% of the universal baryon budget has ended up on stars .
furthermore , the measured shear rate of this galaxy favors our non - ac model .
this is in qualitative agreement with results presented by dutton et al .
( 2005 ) .
model results for the central mass concentration in each galaxy are shown in figures [ eso582 ] and [ ic2522 ] . here
@xmath97 and @xmath98 correspond to the fraction of the total and dark matter mass contained within central @xmath87 kpc of each system .
we see that eso 582-g12 contains roughly @xmath99% of its mass within @xmath87 kpc , while ic 2552 is quite diffuse with @xmath100% . what drives these differences ?
take ic 2522 for example , which has a very low concentration .
this arises because of the combination of its ( small ) disk scalelength _ and _ the fact that the rotation curve is rising at 10 kpc ( implied by a shear of [email protected] ) .
the rotation curve of the stellar disk alone peaks at 2.2 scalelengths ( @xmath101 kpc ) , and will be falling at 10 kpc ( where the shear is measured ) .
the measured shear of [email protected] at 10 kpc implies that the total rotation curve is rising at 10 kpc , which means the dark matter must be very extended ( or of low concentration ) .
eso 582-g12 , on the other hand , has a falling rotation curve at 10 kpc , but a _ larger _ disk scalelength .
this requires a more concentrated halo to explain the observed shear rate .
the two galaxies chosen for this modeling are , in many ways , quite comparable . their hubble types , b / d ratios and disk luminosities are very similar .
the only significant difference is in their measured rates of shear .
the differences in their favored formation models ( ac for eso 582-g12 versus non - ac for ic 2522 ) , halo masses and concentrations , comes down to a different measurement for their shear and the sizes of their disks .
it is quite interesting that two galaxies , which seem so similar in size and hubble type , seem to inhabit very different types of dark matter halos .
it is tempting to speculate that the shear rate itself provides an underlying physical driver to push galaxies in drastically different dark matter halos towards similar luminosities .
however , the problem would then be what stops such a drive and makes the galaxy luminosities so similar , when the halo masses show large differences .
this clearly demonstrates that the shear rate adds an important constraint on galaxy formation models compared to what can be learned from standard tully - fisher constraints alone . indeed , the results for ic 2522 and eso 582-g12 are tantalizing .
the first , with a low shear rate , favors a non - adiabatically contracted halo , with a low halo concentration and a massive , extended dark matter halo .
the second , with a high shear rate , favors an adiabatically contracted halo with a high nfw concentration , and relatively low virial mass .
these two cases motivate the application of the shear - rate constraint to a larger sample of galaxies .
since shear is related to the spiral arm pitch angle and spiral arms are clearly detectable in disk galaxies at @xmath102 ( e.g. elmegreen , elmgreen & hirst 2004 ) , it is possible to constrain models of disk formation and dark matter halo structure back to a look - back time of @xmath103 gyrs , and evaluate how the mass distribution in disk galaxies changes with time .
an important test of our modeling is how well the modeled rotation curves fit the data .
figure 9 shows the observed rotation curves from persic & salucci ( 1995 ) overlaid with our modeled rotation curves . in both cases
the rotation curve is modeled well in the outer parts , where we are measuring the shear . in the inner regions , where the bulge is important ,
both rotation curve models overestimate the observed rotation curve , although the model for eso 582-g12 presents a much better case than that for ic 2522 . for ic 2522 ,
the solid line represents a model rotation curve , with a bulge - to - disk ratio of 0.164 ( see table 4 ) and a bulge effective radius of 0.2 kpc ( as measured in the 1-d decomposition ) .
however , infrared imaging of ic 2522 ( see figure 10 ) reveals that this galaxy has a bar with a major axis length @xmath1042 kpc ( i.e. the radius within which the model overestimates the observed rotational velocities ) and a major - to - minor axis ratio of @xmath1042 .
( it is common for near - ir imaging to reveal bars that may have been hidden due to dust extinction in an optical image , e.g. seigar & james 1998a ; eskridge et al .
2000 ; seigar 2002 ; seigar et al .
the spectroscopic data taken by mathewson et al .
( 1992 ) were observed with a slit aligned along the major axis of each galaxy , and in the case of ic 2522 , the major axis of the bar is well aligned with the galaxy major axis .
the stellar orbits within a bar are such that the dominant motion is parallel to its major - axis ( e.g. athanassoula 1992 ) , and so this will account ( to some extent ) for the low values of the measured rotation velocities ( compared to the modeled values ) within a few kpc . furthermore
, if we double the effective radius of the modeled bulge , the resulting rotation curve is the dashed line in figure 9 .
this still fits the outer part of the galaxy rotation curve extremely well , and the central velocities are not as badly overestimated .
modifying the mass models in this way does not affect the trends displayed in figures 6 and 8 , it just affects how well the model reproduces the inner part of the observed rotation curve .
this is probably because the main constraints we are using to model these galaxies are at a 10 kpc radius , well outside the region where the bulge or bar is dominating the rotational velocities . as a result ,
the concentrations we determine from our modeling do not depend on the mass distribution assumed for the bulge ( i.e. the mass distribution within the central few kpc ) .
this approach is , therefore , a powerful and robust method , because its constraints are relatively insensitive to the details of bars versus bulges in the central regions .
in this paper we have shown that near - infrared and optical spiral arm pitch angles are the same , on average , and as a result we have strengthened the correlation between spiral arm pitch angle and shear rate originally shown in seigar et al .
( 2005 ) and expanded here using optical data . using an infall model ,
we have shown that the use of rates of shear ( which can now be derived from spiral arm pitch angles ) allow us to put constraints on the total central mass concentration , the dark matter concentration and the initial nfw concentration . in some cases it may be possible to determine
if the infall has to occur adiabatically or non - adiabatically , and this is demonstrated by ic 2522 which has to undergo non - adiabatic infall .
this method can be used to determine the central concentrations of galaxies and to constrain galaxy formation models in any galaxy that has detectable spiral structure . in future papers
we will apply our technique to a large sample of galaxies .
furthermore , since spiral structure can clearly be seen in disk galaxies out to @xmath102 ( elmegreen et al .
2004 ) , the evolution of central mass concentrations in disk galaxies can be estimated as a function of look - back time , and we plan to investigate this as well .
support for this work was provided by nasa through grant number hst - ar-10685.01-a from the space telescope science institute , which is operated by the association of universities for research in astronomy , inc .
, under nasa contract nas5 - 26555 .
this research has made use of the nasa / ipac extragalactic database ( ned ) which is operated by the jet propulsion laboratory , california institute of technology , under contract with the national aeronautics and space administration .
this work also made use of data from the ohio state university bright spiral galaxy survey , which was funded by grants ast-9217716 and ast-9617006 from the united states national science foundation , with additional support from the ohio state university .
the authors wish to thank paolo salucci for supplying the rotation curve data from which shear and the @xmath58 velocities were derived .
we also wish to show our appreciation to the referee , dr . p. grosbol , for his many useful comments .
seigar , m. s. , block , d. l. , & puerari , i. 2004 , in penetrating bars through masks of cosmic dust : the hubble tuning fork strikes a new note , ed .
d. l. block , i. puerari , k. c. freeman , r. groess , & e. k. block ( dordrehct : springer ) , 155 | we investigate the use of spiral arm pitch angles as a probe of disk galaxy mass profiles .
we confirm our previous result that spiral arm pitch angles ( p ) are well - correlated with the rate of shear ( s ) in disk galaxy rotation curves , by using a much larger sample ( 51 galaxies ) than used previously ( 17 galaxies ) .
we use this correlation to argue that imaging data alone can provide a powerful probe of galactic mass distributions out to large lookback times .
in contrast to previous work , we show that observed spiral arm pitch angles are similar when measured in the optical ( at 0.4 @xmath0 m ) and the near - infrared ( at 2.1 @xmath0 m ) with a mean difference of @xmath1 .
this is then used to strengthen the known correlation between p and s using @xmath2 band images .
we then use two example galaxies to demonstrate how an inferred shear rate coupled with a bulge - disk decomposition model and a tully - fisher derived velocity normalization can be used to place constraints on a galaxy s baryon fraction and dark matter halo profile .
we show that eso 582-g12 , a galaxy with a high shear rate ( slightly declining rotation curve ) at @xmath3 kpc , favors an adiabatically contracted halo , with high initial nfw concentration ( @xmath4 ) and a high fraction of halo baryons in the form of stars ( @xmath5% ) .
in contrast , ic 2522 has a low shear rate ( rising rotation curve ) at @xmath3 kpc and favors non - adiabatically contracted models with low nfw concentrations ( @xmath6 ) and a low stellar baryon fraction @xmath7% . |
the last few years have settled the debate about the origin of long gamma - ray bursts ( grbs ) .
the hint provided by grb980425 and sn1998bw @xcite was finally taken when the spectroscopic follow - up of the afterglow of grb030329 revealed the unambiguous signatures of a very energetic supernova
sn 2003dh @xcite . soon thereafter , another clear - cut sn1998bw look - alike emerged in the afterglow of grb031203 @xcite . while the swift satellite @xcite has been very successful in finding grbs over a large redshift range ( e.g. , * ? ? ?
* ) , the wait for the next spectacular case of a nearby grb - supernova has lasted more than two years .
grb060218 was detected by the bat instrument on - board the swift satellite @xcite on february 18.149 2006 ut .
this burst had exceptional high - energy properties @xcite .
the peak energy of the event ( sect [ xrays ] ) was very low and we will hereafter refer to this burst as an x - ray flash .
xrf060218 is one of the longest bursts ever detected , and the unusual properties gave a very confused early impression .
several gcns indicated that this was probably not a proper grb , and our optical monitoring programme was therefore somewhat delayed .
however , eventually this turned out to be a very interesting low-@xmath4 event @xcite with a likely association to a supernova @xcite .
the transient has now been detected over a wide wavelength range , from x - rays @xcite to radio @xcite . in this paper
we focus on the optical transient , and the early spectral and photometric evolution of this supernova ( sn ) .
the paper is organized as follows .
[ observations ] outlines how the optical observations were obtained and reduced .
the results are presented in sect .
[ results ] , which includes @xmath0- , @xmath1- and @xmath2-band lightcurves as well as spectra of the sn , and an analysis of the host galaxy .
we end the paper with a discussion ( sect .
[ discussion ] ) where we compare the properties of this sn with other relevant sne .
the observations for xrf060218 were somewhat complicated and hampered by the celestial position of the burst .
being close to the sun it could only be observed for a short time right after sunset .
we have used the combined efforts of two telescopes , at a northern and a southern observatory , to follow the object until it faded into the glare of the sun , about 50 days past the burst . for the final observations we had to restrict ourselves to a single passband ( @xmath2 ) due to the limited time available for observations in the twillight .
we obtained imaging of the transient of xrf060218 with the eso / danish 1.5 m telescope ( d1.5 m ) on la silla equipped with the dfosc instrument , which offers a @xmath5 arcminute field - of - view ( fov ) at 0.395 arcsec per pixel .
we also used the 2.56 m nordic optical telescope ( not ) on la palma equipped with alfosc which offers a fov of @xmath6 arcminutes with a pixel scale of 0.189 arcsec , as well as stancam which has a pixel scale of 0.176 arcsec over @xmath7 arcminutes .
the journal of observations is given in table [ t : log ] .
the data were reduced using standard techniques for de - biasing and flat - fielding .
spectra of the source were obtained with alfosc at four epochs , february 21 , 22 and 24 and on march 2 .
these epochs correspond to 3.78 , 4.71 , 6.71 and 12.71 days past the burst .
each spectrum had an integration time of 2400 seconds using grism 4 and a 1.3 arcsec wide slit .
this set - up provided a dispersion of 3 per pixel .
the spectral range covered is from 3300 to 9200 .
there is some second order contamination above 6600 , and significant fringing above @xmath8 .
the spectra were taken at or close to the parallactic angle .
we note that apart from the first spectrum , taken at an airmass of 1.93 , all spectra were achieved at an airmass @xmath9 .
the not / alfosc has a high efficiency in the uv , so we put emphasis on obtaining the bluest part of the spectrum .
the spectra were reduced following standard procedures in midas and iraf .
wavelength calibration was achieved by comparison to images taken of helium and neon lamps .
the flux calibration was performed using the spectrophotometric standard star gd71 @xcite , which was observed every night close in time to the supernova observation .
finally , the absolute flux - calibration was achieved by comparison to the contemporary ( or interpolated ) @xmath1-band photometry .
when comparing to synthetic photometry obtained by integrating each spectrum under the filter profiles , we discovered that some of the spectra have suffered from differential slit losses .
this has been considered in the analysis below .
aperture photometry of the transient was carried out using a combination of daophot @xcite and sextractor @xcite .
we measured the magnitudes of the supernova as well as for 9 stars in the field in the @xmath1 and @xmath2 bands ( 7 local standards in the @xmath0 band ) .
the relative magnitudes were transformed to the standard system using observations of photometric standard stars @xcite .
we estimate an absolute photometric accuracy of 0.08 , 0.06 and 0.04 mags in the @xmath0 , @xmath1 and @xmath2 bands , respectively . in fig .
[ f : lightcurves ] we plot the @xmath0- , @xmath1- and @xmath2-band lightcurves .
this is the data from table [ t : log ] .
the dates are given with respect to the time of the burst in the observers frame .
we have not plotted the data with errors larger than 0.15 mag , if there are more accurate data from the same night . the @xmath2- and @xmath1-band lightcurves are followed from well before peak and are traced to way past maximum . in fig .
[ f : lightcurves ] we have applied no corrections to subtract the host galaxy ( estimated at @xmath10 , see sect . [ host ] )
. this can be an important contribution , @xmath11 mag at maximum light , and is considered in the following analysis .
also , we have made no k - corrections for the magnitudes given in table [ t : log ] and plotted in fig .
[ f : lightcurves ] . at the early epochs where we have spectra
, we estimate this correction to be @xmath12 mag in the @xmath1 band , and @xmath13 mag in the @xmath2 band .
the final spectrum is taken closest in time to the maximum light in these bands , and indicate k - corrections of @xmath14 mag in the @xmath1 band , and @xmath15 mag in @xmath2 . to estimate the time of maximum , peak brightness and the lightcurve shape as described by @xmath16m@xmath17 ( the number of magnitudes the supernova decayed in the 15 days following maximum brightness ) we have fitted the lightcurves with smooth functions ( see * ? ? ?
we estimate the rise times of @xmath18 days past burst and @xmath19 days .
we further estimate @xmath16m@xmath20 and @xmath16m@xmath21 mag from the observed data . when correcting for the underlying emission from the host galaxy ( sect .
[ other ] ) , as well as for time dilation , the corrected numbers are @xmath16m@xmath22 and @xmath16m@xmath23 mag . the peak magnitudes are estimated to be m(@xmath1)=@xmath24 and m(@xmath2)=@xmath25 mag . to determine the absolute magnitudes we need estimates of the distance and extinction .
the redshift of this burst is @xmath26 ( * ? ? ?
* see also sect .
[ spectrum ] ) , and assuming a cosmology where @xmath27kms@xmath28mpc@xmath28 , @xmath29 and @xmath30 , this corresponds to a luminosity distance of 145.4 mpc .
the reddening associated with galactic extinction is e(@xmath31)=0.14 mag according to the maps by @xcite .
high - resolution spectra @xcite can be used to check this . using the sodium lines to estimate the reddening
( see e.g. , * ? ? ?
* ) provides e(@xmath31)=0.127 for the milky way , and e(@xmath31)=0.042 mag for the grb host .
this is consistent with adopting a total reddening of e(@xmath31)=0.14 mag .
we note that @xcite required e(@xmath32 mag based on the assumption of thermal radiation detected by uvot .
this is higher than claimed by @xcite .
it is known that the na i d lines do not provide a robust measure of the extinction , and could be influenced by e.g. , the ionization state in the host ( see e.g. , * ? ? ? * ; * ? ? ?
* ) . however , the overall properties of the host galaxy based on the spectral energy distribution ( sed ) modeling ( sect . [ modeling ] ) , as well as the measured balmer line decrement ( sect .
[ other ] , see also * ? ? ?
* ) , also argue for a low host extinction . in the following
we will therefore adopt a total extinction of e@xmath33 mag .
the absolute magnitudes of the sn are then m@xmath34 and m@xmath35 mag . finally , correcting these estimates for host contamination ( @xmath36 ) and k - corrections ( @xmath37 ) our best estimates are m@xmath38 and m@xmath39 mag .
these are the magnitudes adopted for comparison to other sne , and are given in table [ t : summary ] .
lightcurves for type ic sne are relatively rare in the @xmath0 band . for sn2006aj , we started our @xmath0-band imaging campaign 5 days past the burst . at this epoch ,
the @xmath0 band was already close to maximum light .
we then followed the evolution of the @xmath0-band flux until 25 days past the burst , after which the supernova became to faint ( also compared to the host ) to allow further monitoring .
given the sparse pre - maximum coverage , the estimates are somewhat more uncertain in the @xmath0 band .
we estimate @xmath40 days .
the estimate of the corrected light curve shape is rather uncertain , due to the large correction for host contamination on the already steep lightcurve .
we estimate @xmath16m@xmath41 .
the peak brightness is m@xmath42 mag , which converts to an absolute @xmath0-band magnitude of m@xmath43 in the vega magnitude system .
the k - corrections are most uncertain in the @xmath0 band , and could also be significant in particular for the latest epochs where the spectrum falls very steeply in that region . at around @xmath0-band maximum light
, we estimate a k - correction of @xmath44 mag . applying this k - correction , and a correction for the host galaxy ( @xmath45 mag ) we therefore estimate the final absolute magnitude m(@xmath46 .
we summarize all corrected lightcurve parameters in table [ t : summary ] . in this table , the rise times are corrected for time dilation , as are the light curve shapes which are also corrected for the underlying host galaxy .
the absolute magnitudes are corrected for extinction , host galaxy contamination and are k - corrected .
the flux - calibrated spectra are shown in fig .
[ f : spectra ] .
the continuum - like spectrum with broad bumps renders a classification of this burst as a type ic supernova , based on the lack of conspicuous sn lines ( e.g. , * ? ? ?
the spectral evolution is well represented , the most obvious development being the depression of the uv flux with time .
some of the depression seen in our final spectrum may be attributed to differential slit losses , but the overall evolution of the spectra are correct , as can be seen from the comparison to the broad band light curves .
in fact , the uv depression is a common feature of sne and reflects the increased line blanketing due to low ionization iron group elements .
as seen from the lightcurve , the @xmath0 band actually peaked close to the date of our third spectrum , so we see the rise of the @xmath0 band up to that epoch in the spectral evolution , followed by a rapid decline to the final spectrum .
our latest spectrum , taken 12.3 days past burst in the rest frame , shows a dramatic evolution of the flux towards the red part of the spectrum .
the broad red bumps are common features of so - called hypernovae and signal huge expansion velocities of the ejecta . interpreting the inflection point at @xmath47 as the si ii 6355 feature seen in other grb - sne @xcite , we can estimate an expansion velocity of @xmath48 km s@xmath28 . at 12 days past burst ,
this is similar to the expansion velocities measured in sn1998bw and sn2003dh .
however , since this feature is quite loosely defined , this estimate can only be approximate .
from the multitude of narrow emission lines from the host galaxy we can also measure the redshift to the supernova .
the positions and fluxes of a number of detected narrow lines are given in table [ t : lines ] .
we derived the redshift by measuring the positions of the [ o ii ] line , the [ o iii ] lines as well as h@xmath49 and h@xmath50 at all 4 epochs , and conclude @xmath3 .
the fluxes of the lines were measured by gaussian fits , using both idl and iraf splot .
we use these below to estimate the star formation rate and the metallicity .
we note that the values given in table [ t : lines ] are averages for the four spectra corrected for e@xmath33 mag . apart from the stronger lines listed in table [ t : lines ] , we also detect [ ne iii ] @xmath51 , which signals the presence of ionizing radiation .
this line is about 3 times weaker than h@xmath50 , although the uncertainty in such a weak line is @xmath52 in our spectra .
the stronger lines have uncertainties of @xmath53 .
the @xmath54-band sdss pre - imaging of the field @xcite allowed us to construct the optical spectral energy distribution of the host galaxy .
however , it was noted @xcite that the absolute calibration of this field was not correct . to correct the sdss model magnitudes we used field star photometry @xcite , which was transformed from landolt to the sdss - system using @xcite . corrected host magnitudes
were then deduced from the offsets between the sdss model magnitudes and the transformed values , giving : @xmath55 .
we then used the best fit sed ( see below ) to transform these magnitudes back to landolt photometry : @xmath56mag .
these are the host galaxy magnitudes used to correct the light curve parameters in sect .
[ lightcurves ] .
the @xmath54 host galaxy photometric points were then de - reddened by the galactic extinction following @xcite and then fitted based on the sdss filter+ccd efficiency curves @xcite and using the synthetic sed templates constructed with the hyperz code @xcite .
for the construction of the synthetic templates three initial mass functions ( imfs ) were used @xcite . we also used four different extinction laws : mw @xcite , lmc @xcite , smc @xcite and one for starburst galaxies @xcite .
solar metallicity was assumed for all the templates .
the redshift of the templates was fixed at @xmath26 .
in addition , a wide range of star - formation histories were considered ( see more details on the @xmath57 parameter in * ? ? ?
* ) , creating different families of templates : elliptical , starburst , lenticular , irregular and spiral galaxies .
the @xmath54-band photometric points were satisfactorily fitted by the sed templates ( @xmath58 ; see fig .
[ f : sed ] ) .
our sed fits did not favour any imf , extinction law or galaxy type .
this means that the inferred host galaxy extinction is independent on the input model , and is stable at around @xmath59 mag .
this is why we were favoring a low host galaxy extinction in sect .
[ lightcurves ] .
the host magnitude of @xmath60 mag , and the galactic extinction of @xmath61 gives an absolute magnitude for the host of m@xmath62 mag at the measured redshift .
adopting @xmath63 this corresponds to @xmath64 . from the h@xmath49 and [ o ii ] lines
we can estimate the star formation rate ( sfr ) . from both these lines we get sfr @xmath65 yr@xmath28 , following @xcite .
this is of course only measured from the part of the galaxy that falls on the spectroscopic slit .
the specific star formation rate for the host galaxy of xrf060218 is thus @xmath666 m@xmath67 yr@xmath68 .
finally , we can estimate the metallicity of the galaxy using the r23 technique . from the results presented in table [ t : lines ] ,
we derive a log(r@xmath69 .
this indicates a somewhat sub - solar metallicity , although the exact value can not be determined from this ratio alone ( see e.g. , fig . 5 by *
the luminosity and star formation rate thus indicates a small but fairly normal dwarf galaxy , similar to other nearby grb host galaxies @xcite .
the low metallicity is also similar to that of other grb host galaxies .
the spectral evolution reveals a rapidly evolving type ic supernova with very broad lines .
very few type ic spectra exist for such early epochs .
our first spectrum was obtained 3.7 days past the burst . the first spectrum for sn1998bw
was not obtained until after a week , and for sn2003dh the emission was still dominated by the afterglow at this epoch .
therefore , it is difficult to make any one - to - one comparisons of the apparent bumps in these spectra with those of other similar sne ( see e.g. , fig . 1 in * ? ? ?
* ) . in fig .
[ f : comparison ] we have re - plotted our latest spectrum of sn2006aj from march 2 .
this is 12.3 days past burst in the sn rest frame .
we have also plotted spectra for sn1998bw @xcite and sn2003dh @xcite .
these spectra are very similar to the one for sn2006aj , but are taken at a later epoch .
sn2006aj thus displays a fast spectral evolution .
this agrees with the narrow lightcurve . from the lightcurves , as well as from the spectral evolution
, we can see that the emission is dominated by the supernova rather than by the afterglow from very early on .
this is similar to sn1998bw , where no optical afterglow was ever detected , but very different from sn2003dh which was dominated by the afterglow for more than a week before it emerged .
the peak magnitudes we have estimated show that sn2006aj was a fairly normal type ic supernova in that respect @xcite .
in particular , it was not as bright as sn1998bw or sn2003dh .
sn1998bw ejected @xmath70 of radioactive @xmath71ni ( see e.g. , * ? ? ?
* ; * ? ? ?
* ) . that sn2006aj was only @xmath52 as luminous as sn1998bw thus means that sn2006aj ejected @xmath72 of radioactive @xmath71ni
this is still more than seen in other broad - line supernovae , such as sne 1997ef and 2002ap .
we note that the assumption that the peak magnitude scales with the nickel mass may not be valid for very asymmetric explosions @xcite .
grbs are expected to be asymmetric , although they should all be pointed within a few degrees to our line of sight .
the shape of the lightcurve is also of interest . for sn2006aj
we have summarized the properties in table [ t : summary ] . for comparison ,
the type ic sn1994i displayed @xmath16m@xmath73 and @xmath16m@xmath74 mag .
the peak magnitude for sn1994i was reached after @xmath75 days in @xmath0 , and after 10 days in the @xmath1 band @xcite .
the rise time is , however , very uncertain for sn1994i ; since the exact epoch of the explosion was not observed . for sn1998bw , @xcite estimated @xmath16m@xmath76 and @xmath16m@xmath77 mag .
the peak magnitude was reached after 13.5 days in @xmath0 , and after 17 days in the @xmath1 band @xcite .
this is clearly slower than observed for sn2006aj .
finally , sn2002ap reached @xmath0-band maximum at about 6.2 days @xcite .
this lightcurve seems to be most similar to sn2006aj in this respect . in fig .
[ f : lightcurvescompare ] we compare the light curves of sn2006aj with those for sne 1994i , 1998bw and 2002ap .
note that in this figure we have corrected the light curves for time dilation ( for sne2006aj and 1998bw ) and also corrected sn2006aj for the underlying host galaxy .
this correction is quite substantial , in particular at late stages ( compare fig . [
f : lightcurves ] ) . the comparison in fig .
[ f : lightcurvescompare ] demonstrates that sn2006aj is in fact a fast version of sn2002ap .
another important aspect of grb - sne is clearly the possibility to relate the supernova shock - wave breakout with the exact time of the explosion .
@xcite used the uvot instrument onboard swift to follow the uv lightcurves from the early shock break - out to the following peak due to radioactive heating ( the latter being the optical peak we are probing in this paper ) .
such a shock break - out was also seen in sn1999ex @xcite and in sn1998bw @xcite . for sn1999ex
, the time of shock break - out could be estimated , and the rise time in the @xmath1 band was @xmath78 days @xcite .
sn2006aj has a substantially faster lightcurve , which is related to the faster expansion velocities , and possibly also to a lower ejecta mass .
we note that grb060218 was a very soft burst , and thus qualifies as an ( unusual ) xrf .
@xcite estimated e@xmath79 kev , at the very end of the observed distribution of peak energies . while the case for an association between long grbs and sne has been established ( see sect .
[ introduction ] ) , the case is more unclear for xrfs .
xrf030723 showed a very conspicuous light curve bump at @xmath80 days past burst , suggesting the presence of a fast rising supernova @xcite .
in fact , doubts were raised against this interpretation since the required supernova light curve was very fast and narrow .
the very fast @xmath0-band lightcurve of sn2006aj may be taken as support for the hypothesis of a sn in xrf030723 . at a cosmological redshift of @xmath81
the @xmath2-band light curve bump would correspond to rest frame @xmath0 , as also noted by @xcite .
more recently , xrf050824 showed a less conspicuous bump @xcite .
moreover , xrf020903 has a lightcurve and spectrum consistent with a supernova at @xmath82 @xcite .
these findings all argue for a common progenitor for grbs and xrfs .
the situation appeared less clear as other xrfs with late - time coverage did not show clear evidence for a bright supernova bump ( e.g. , * ? ? ? * ) .
however , with xrf060218 the case for a supernova origin for such bursts is obvious . among the many remaining questions are the lack of conspicuous afterglow emission .
compared to grb030329 , the supernova emerged much faster from the afterglow for xrf060218 .
it is interesting to note that there was also no conspicuous afterglow in sn1998bw .
for sn2003lw ( grb031203 ) there were claims of a very faint and fast decaying afterglow @xcite .
these bursts also had low values of e@xmath83 , and in fact @xcite considered 031203 to be an xrf .
@xcite considered an off - axis model for xrf031203 in which this was really a normal grb although viewed from an angle of about twice the opening angle . from fig .
[ f : xrfs ] we see that it may be difficult to reconcile these diverse observations by a simple geometric scenario . in this four - field diagram
we have divided bursts into xrfs and grbs .
we have also divided them according to the dominating component in the optical lightcurve ; supernova or afterglow .
we have indicated the spectroscopically confirmed sn - grbs .
the upper left corner is represented by grb030329 where sn2003dh was not apparent until after a week .
xrf020903 was dominated by an afterglow until the late supernova bump , and occupies the upper right field .
the same applies to xrf030723 .
the lower left box is represented by sn1998bw and grb980425 , although the peak energy was not very high .
finally , xrf060218 now fills in the lower right field in this diagram .
it has a very low e@xmath83 and shows supernova signatures already from the very early photometry @xcite and spectroscopy ( * ? ? ?
* ; * ? ? ?
* and this work fig .
[ f : spectra ] ) .
a one - parameter explanation such as an on- vs. off - axis picture would have problem to explain all the combinations in fig .
[ f : xrfs ] .
it seems that ( sn)-xrfs can come both with and without a conspicious afterglow , and the afterglow can moreover behave quite differently ( flat early lightcurve in xrf030723 vs. constant decay in xrf050824 ) .
a larger sample of sn - grbs will be needed to unveil whether we observe different classes of objects , or simply a continuum of burst properties .
@xcite actually predicted swift to detect a significant population of faint bursts and hence allow the study of core - collapse sne at much earlier times than had been previously possible , and indicated that this would have a substantial impact on sn research .
the discovery of the first nearby swift grb - sn substantiates this prediction .
this work was done at the dark cosmology centre funded by the danish national research foundation .
js also acknowledge support from danmarks nationalbank and from the anna - greta and holger crafoord fund .
aoj acknowledges support from the norwegian research council .
some of the data presented here have been taken using alfosc , which is owned by the instituto de astrofisica de andalucia ( iaa ) and operated at the nordic optical telescope under agreement between iaa and the nbi of the astronomical observatory of copenhagen .
+ date & @xmath16 t & pass band & exptime & magnitude & magnitude error & telescope + ( ut ) & ( days ) & & ( s ) & & ( @xmath84 & + feb 23.932 & 5.783 & u & 900 & 17.74 & 0.06 & not + feb 24.006 & 5.857 & u & 120 & 17.43 & 0.30 & d1.5 m + feb 24.009 & 5.860 & u & 200 & 17.33 & 0.16 & d1.5 m + feb 24.012 & 5.863 & u & 200 & 17.55 & 0.07 & d1.5 m + feb 24.016 & 5.867 & u & 200 & 17.61 & 0.06 & d1.5 m + feb 24.022 & 5.873 & u & 600 & 17.66 & 0.04 & d1.5 m + feb 24.030 & 5.881 & u & 600 & 17.60 & 0.04 & d1.5 m + feb 24.898 & 6.749 & u & 300 & 17.69 & 0.10 & not + feb 25.009 & 6.860 & u & 200 & 17.62 & 0.09 & d1.5 m + feb 25.014 & 6.865 & u & 400 & 17.67 & 0.06 & d1.5 m + feb 25.021 & 6.872 & u & 600 & 17.64 & 0.04 & d1.5 m + feb 25.029 & 6.880 & u & 600 & 17.63 & 0.05 & d1.5 m + feb 26.006 & 7.857 & u & 200 & 17.65 & 0.24 & d1.5 m + feb 26.011 & 7.862 & u & 200 & 17.51 & 0.12 & d1.5 m + feb 26.016 & 7.867 & u & 600 & 17.66 & 0.03 & d1.5 m + feb 26.024 & 7.875 & u & 600 & 17.71 & 0.03 & d1.5 m + feb 27.008 & 8.859 & u & 200 & 17.35 & 0.17 & d1.5 m + feb 27.011 & 8.862 & u & 200 & 17.61 & 0.10 & d1.5 m + feb 27.017 & 8.868 & u & 600 & 17.69 & 0.05 & d1.5 m + feb 27.025 & 8.876 & u & 600 & 17.72 & 0.05 & d1.5 m + feb 28.007 & 9.858 & u & 200 & 17.57 & 0.16 & d1.5 m + feb 28.010 & 9.861 & u & 200 & 17.79 & 0.10 & d1.5 m + feb 28.016 & 9.867 & u & 600 & 17.77 & 0.08 & d1.5 m + feb 28.024 & 9.875 & u & 600 & 17.87 & 0.04 & d1.5 m + mar 1.012 & 10.863 & u & 600 & 17.85 & 0.10 & d1.5 m + mar 2.003 & 11.854 & u & 200 & 18.48 & 0.24 & d1.5 m + mar 2.007 & 11.858 & u & 200 & 18.16 & 0.25 & d1.5 m + mar 2.012 & 11.863 & u & 600 & 18.04 & 0.08 & d1.5 m + mar 2.914 & 12.765 & u & 300 & 18.00 & 0.14 & not + mar 3.003 & 12.854 & u & 200 & 18.06 & 0.18 & d1.5 m + mar 3.009 & 12.860 & u & 600 & 18.19 & 0.10 & d1.5 m + mar 4.005 & 13.856 & u & 200 & 18.01 & 0.15 & d1.5 m + mar 4.010 & 13.861 & u & 600 & 18.45 & 0.08 & d1.5 m + mar 5.005 & 14.856 & u & 200 & 18.31 & 0.18 & d1.5 m + mar 5.011 & 14.862 & u & 600 & 18.46 & 0.12 & d1.5 m + mar 6.006 & 15.857 & u & 200 & 18.57 & 0.20 & d1.5 m + mar 6.012 & 15.863 & u & 600 & 18.48 & 0.13 & d1.5 m + mar 6.851 & 16.702 & u & 900 & 18.79 & 0.17 & not + mar 7.854 & 17.705 & u & 900 & 18.89 & 0.15 & not + mar 8.866 & 18.717 & u & 900 & 18.79 & 0.16 & not + mar 10.878 & 20.729 & u & 1500 & 19.16 & 0.20 & not + mar 14.864 & 24.715 & u & 2400 & 19.55 & 0.15 & not + feb 21.013 & 2.864 & v & 120 & 18.22 & 0.05 & d1.5 m + feb 21.015 & 2.866 & v & 120 & 18.16 & 0.08 & d1.5 m + feb 21.017 & 2.868 & v & 120 & 18.21 & 0.03 & d1.5 m + feb 21.034 & 2.885 & v & 300 & 18.21 & 0.03 & d1.5 m + feb 21.051 & 2.902 & v & 600 & 18.17 & 0.03 & d1.5 m + feb 22.020 & 3.871 & v & 120 & 18.01 & 0.03 & d1.5 m + feb 22.022 & 3.873 & v & 120 & 18.02 & 0.05 & d1.5 m + feb 22.040 & 3.891 & v & 300 & 18.01 & 0.03 & d1.5 m + feb 22.045 & 3.896 & v & 300 & 18.00 & 0.03 & d1.5 m + feb 22.049 & 3.900 & v & 300 & 18.03 & 0.03 & d1.5 m + feb 23.027 & 4.878 & v & 300 & 17.84 & 0.05 & d1.5 m + feb 23.031 & 4.882 & v & 300 & 17.87 & 0.02 & d1.5 m + feb 23.046 & 4.897 & v & 300 & 17.84 & 0.03 & d1.5 m + feb 23.050 & 4.901 & v & 300 & 17.85 & 0.04 & d1.5 m + feb 24.041 & 5.892 & v & 300 & 17.71 & 0.04 & d1.5 m + feb 25.041 & 6.892 & v & 300 & 17.61 & 0.03 & d1.5 m + feb 26.031 & 7.882 & v & 300 & 17.53 & 0.03 & d1.5 m + feb 27.032 & 8.883 & v & 300 & 17.50 & 0.04 & d1.5 m + feb 28.031 & 9.882 & v & 300 & 17.49 & 0.04 & d1.5 m + mar 1.018 & 10.869 & v & 200 & 17.50 & 0.08 & d1.5 m + mar 2.019 & 11.870 & v & 200 & 17.50 & 0.03 & d1.5 m + mar 2.924 & 12.775 & v & 300 & 17.52 & 0.04 & not + mar 3.016 & 12.867 & v & 200 & 17.54 & 0.05 & d1.5 m + mar 4.017 & 13.868 & v & 200 & 17.61 & 0.03 & d1.5 m + mar 5.018 & 14.869 & v & 200 & 17.62 & 0.15 & d1.5 m + mar 6.863 & 16.714 & v & 300 & 17.77 & 0.10 & not + mar 7.002 & 16.853 & v & 300 & 17.74 & 0.04 & d1.5 m + mar 7.012 & 16.863 & v & 300 & 17.81 & 0.02 & d1.5 m + mar 7.863 & 17.714 & v & 300 & 17.85 & 0.08 & not + mar 8.002 & 17.853 & v & 300 & 17.92 & 0.07 & d1.5 m + mar 8.006 & 17.857 & v & 300 & 17.84 & 0.07 & d1.5 m + mar 8.874 & 18.725 & v & 300 & 18.00 & 0.09 & not + mar 9.008 & 18.859 & v & 450 & 17.94 & 0.04 & d1.5 m + mar 9.837 & 19.688 & v & 300 & 18.04 & 0.11 & not + mar 11.004 & 20.855 & v & 300 & 18.12 & 0.07 & d1.5 m + mar 12.000 & 21.851 & v & 200 & 18.20 & 0.07 & d1.5 m + mar 12.999 & 22.850 & v & 100 & 18.24 & 0.17 & d1.5 m + mar 13.859 & 23.710 & v & 900 & 18.26 & 0.06 & not + mar 13.995 & 23.846 & v & 100 & 18.29 & 0.09 & d1.5 m + mar 15.992 & 25.843 & v & 180 & 18.34 & 0.12 & d1.5 m + mar 18.840 & 28.691 & v & 900 & 18.65 & 0.07 & not + mar 20.852 & 30.703 & v & 900 & 18.78 & 0.06 & not + feb 21.021 & 2.872 & r & 200 & 18.00 & 0.02 & d1.5 m + feb 21.025 & 2.876 & r & 300 & 17.97 & 0.09 & d1.5 m + feb 21.029 & 2.880 & r & 300 & 17.99 & 0.09 & d1.5 m + feb 21.041 & 2.892 & r & 600 & 17.99 & 0.07 & d1.5 m + feb 21.057 & 2.908 & r & 150 & 18.00 & 0.08 & d1.5 m + feb 21.915 & 3.766 & r & 900 & 17.79 & 0.03 & not + feb 22.026 & 3.877 & r & 300 & 17.80 & 0.04 & d1.5 m + feb 22.031 & 3.882
& r & 300 & 17.82 & 0.06 & d1.5 m + feb 22.054 & 3.905 & r & 300 & 17.81 & 0.08 & d1.5 m + feb 22.898 & 4.749 & r & 900 & 17.64 & 0.03 & not + feb 23.016 & 4.867 & r & 200 & 17.62 & 0.07 & d1.5 m + feb 23.019 & 4.870 & r & 200 & 17.64 & 0.08 & d1.5 m + feb 23.023 & 4.874 & r & 200 & 17.63 & 0.09 & d1.5 m + feb 23.036 & 4.887 & r & 300 & 17.61 & 0.08 & d1.5 m + feb 23.041 & 4.892 & r & 300 & 17.64 & 0.09 & d1.5 m + feb 23.879 & 5.730 & r & 250 & 17.51 & 0.04 & not + feb 24.036 & 5.887 & r & 300 & 17.49 & 0.10 & d1.5 m + feb 24.854 & 6.705
& r & 300 & 17.36 & 0.04 & not + feb 25.036 & 6.887 & r & 300 & 17.38 & 0.05 & d1.5 m + feb 26.036 & 7.887
& r & 300 & 17.29 & 0.03 & d1.5 m + feb 27.037 & 8.888 & r & 300 & 17.25 & 0.08 & d1.5 m + feb 28.035 & 9.886 & r & 300 & 17.21 & 0.02 & d1.5 m + mar 1.023 & 10.874 & r & 200 & 17.26 & 0.04 & d1.5 m + mar 2.022 & 11.873 & r & 200 & 17.21 & 0.03 & d1.5 m + mar 2.849 & 12.700 & r & 300 & 17.22 & 0.03 & not + mar 3.020 & 12.871 & r & 200 & 17.24 & 0.09 & d1.5 m + mar 4.020 & 13.871 & r & 200 & 17.26 & 0.08 & d1.5 m + mar 5.022 & 14.873 & r & 200 & 17.30 & 0.18 & d1.5 m + mar 6.843 & 16.694 & r & 300 & 17.43 & 0.07 & not + mar 7.016 & 16.867 & r & 300 & 17.34 & 0.06 & d1.5 m + mar 7.020 & 16.871 & r & 200 & 17.35 & 0.03 & d1.5 m + mar 7.867 & 17.718 & r & 300 & 17.43 & 0.06 & not + mar 8.012 & 17.863 & r & 300 & 17.40 & 0.02 & d1.5 m + mar 8.017 & 17.868 & r & 300 & 17.43 & 0.05 & d1.5 m + mar 8.848 & 18.699 & r & 1500 & 17.43 & 0.03 & not + mar 9.841 & 19.692 & r & 300 & 17.51 & 0.07 & not + mar 10.010 & 19.861 & r & 400 & 17.47 & 0.07 & d1.5 m + mar 11.009 & 20.860 & r & 300 & 17.60 & 0.03 & d1.5 m + mar 12.005 & 21.856 & r & 400 & 17.68 & 0.02 & d1.5 m + mar 13.001 & 22.852 & r & 150 & 17.76 & 0.09 & d1.5 m + mar 13.872 & 23.723 & r & 900 & 17.80 & 0.04 & not + mar 13.998 & 23.849 & r & 100 & 17.79 & 0.08 & d1.5 m + mar 14.000 & 23.851 & r & 100 & 17.81 & 0.06 & d1.5 m + mar 18.860 & 28.711 & r & 900 & 18.10 & 0.04 & not + mar 20.866 & 30.717 & r & 900 & 18.12 & 0.04 & not + mar 27.889 & 37.740 & r & 1800 & 18.43 & 0.04 & not + mar 28.874 & 38.725 & r & 1800 & 18.58 & 0.03 & not + mar 29.874 & 39.725 & r & 2100 & 18.60 & 0.03 & not + mar 31.884 & 41.735 & r & 1800 & 18.63 & 0.08 & not + apr 07.858 & 48.709 & r & 1800 & 18.89 & 0.05¬ + the @xmath0- , @xmath1- and @xmath2-band lightcurves of sn2006aj .
dates are given in days after the high - energy burst in the observers frame .
the filled circles are data from the d1.5 m , and the open circles from the not ( data from table [ t : log ] ) . for clarity ,
we have excluded points for which the errors are greater than 0.15 mag when more accurate data were available for the same night .
these magnitudes are not corrected for the host galaxy contribution , and have not been k - corrected .
corrections for the host galaxy is done in fig .
[ f : lightcurvescompare ] [ f : lightcurves ] the flux - calibrated and de - reddened spectra of the emerging supernova .
the spectra have been absolute flux - calibrated by comparison to @xmath1-band photometry . in the feb .
24 spectrum we note some components of intermediate widths @xmath85 km s@xmath28 in the blue part of the spectrum . [ f : spectra ] spectral comparison to other grb sne , sn1998bw at 16 days past burst and sn2003dh at 20 days past burst .
times are in the sn rest frames .
the spectra of sne 1998bw and 2003dh have been arbitrarily shifted in flux .
they have also been shifted to the redshift of sn2006aj .
note that sn2003dh has an afterglow that adds to the uv part .
[ f : comparison ] comparison of @xmath0- and @xmath2-band lightcurves for sn2006aj and three other type ic sne .
sn1998bw is seen to evolve quite a bit slower , while sn1994i is clearly faster .
the best match is with sn2002ap .
the lightcurves have been corrected for time dilation and matched at date of peak and at maximum brightness . in this plot we have also corrected sn2006aj for the host galaxy contribution .
note that the final @xmath0-band datapoint is uncertain due to a large relative host extinction correction .
[ f : lightcurvescompare ] four - field diagram representing different varieties of supernova - grbs .
there are differences in high - energy properties defining the peak energy of the burst , as well as in the optical afterglow appearance .
xrf060218 / sn2006aj fills in the lower right field of this diagram , as an xrf with supernova light dominating the early optical transient .
[ f : xrfs ] | we have studied the afterglow of the gamma - ray burst ( grb ) of february 18 , 2006 .
this is a nearby long grb , with a very low peak energy , and is therefore classified as an x - ray flash ( xrf ) .
xrf060218 is clearly associated with a supernova dubbed sn2006aj .
we present early spectra for sn2006aj as well as optical lightcurves reaching out to 50 days past explosion .
our optical lightcurves define the rise times , the lightcurve shapes and the absolute magnitudes in the @xmath0 , @xmath1 and @xmath2 bands , and we compare these data with data for other relevant supernovae .
sn2006aj evolved quite fast , somewhat similarly to sn2002ap , but not as fast as sn1994i .
our spectra show the evolution of the supernova over the peak , when the @xmath0-band portion of the spectrum rapidly fades due to extensive line blanketing .
we compare to similar spectra of very energetic type ic supernovae .
our first spectra are earlier than spectra for any other grb - sn .
the spectrum taken 12 days after burst in the rest frame is similar to somewhat later spectra of both sn1998bw and sn2003dh , implying a rapid early evolution .
this is consistent with the fast lightcurve . from the narrow emission lines from the host galaxy we derive a redshift of @xmath3 .
this makes xrf060218 the second closest gamma - ray burst detected .
the flux of these emission lines indicate a high - excitation state , and a modest metallicity and star formation rate of the host galaxy . |
a complementary astrophysical data from type ia supernova , large scale structure ( lss ) and cosmic microwave background ( cmb ) indicate that our universe is currently undergoing a phase of accelerating @xcite .
a component which is responsible for this accelerated expansion is usually dubbed `` dark energy '' ( de ) . the simplest candidate for de
is the cosmological constant @xcite which is located in the center both from theoretical and observational evidences .
however , there are different alternative theories for the dynamical de scenarios which have been proposed to interpret the accelerating universe
. one of these models , which has arisen a lot of enthusiasm recently , is hde .
it was shown by cohen et al .
@xcite that in quantum field theory a short distance cutoff could be related to a long distance cutoff ( ir cutoff l ) due to the limit sets by black hole formation . if the quantum zero - point energy density is due to a short distance cutoff , then the total energy in a region of size @xmath13 should not exceed the mass of a black hole of the same size , namely @xmath14 .
the largest @xmath13 is the one saturating this inequality , so that we obtain the hde density as @xcite @xmath15 where @xmath16 is the reduced planck mass . in general the factor
@xmath17 in holographic energy density can vary with time very slowly @xcite . by slowly varying we mean that @xmath18 is upper bounded by the hubble expansion rate , @xmath19 , i.e.
, @xcite @xmath20 it was also argued that @xmath17 depends on the ir length , @xmath13 @xcite . for the case of @xmath21 , one can take @xmath17 approximately constant in the late time where de dominates ( @xmath22 ) @xcite .
since in the present work we consider the late time cosmology where the de dominates , we shall assume the factor @xmath17 to be constant . depending on the ir cutoff , @xmath13 ,
many hde models have been proposed in the literatures .
a comprehensive , but not complete , list of ir cutoffs which have been used includes the particle horizon radius @xmath23 @xcite , the hubble horizon @xmath21 @xcite , the future event horizon radius @xmath24 @xcite , the apparent horizon radius @xmath25 @xcite , the ricci scalar curvature radius @xmath26 @xcite , the so - called granda - oliveros ( go ) cutoff , which is the formal generalization of @xmath27 , namely @xmath28 @xcite , the age of our universe @xmath29 @xcite , the conformal age of our universe @xmath30 @xcite , the radius of the cosmic null hypersurface @xcite , etc . on the other side , in recent years
, the theories of large extra dimensions in which the observed universe is realized as a brane embedded in a higher dimensional spacetime , have received a lot of interest . in these theories the cosmological evolution on the brane
is described by an effective friedmann equation that incorporates non - trivially with the effects of the bulk onto the brane .
one of the well - know picture in the braneworld scenarios was proposed by dvali - gabadadze - porrati ( dgp ) @xcite . in this model
our four - dimensional universe is a friedmann - robertson - walker ( frw ) brane embedded in a five - dimensional minkowskian bulk with infinite size . in this model
the recovery of the usual gravitational laws on the brane is obtained by adding an einstein - hilbert term to the action of the brane computed with the brane intrinsic curvature .
the self - accelerating branch of dgp model can explain the late time cosmic speed - up without recourse to de or other components of energy @xcite .
however , the self - accelerating dgp branch has ghost instabilities and it can not realize phantom divide crossing by itself . to realize phantom divide crossing it
is necessary to add at least a component of energy on the brane . on the other hand
, the normal dgp branch can not explain acceleration but it has the potential to realize a phantom - like phase by dynamical screening on the brane . in the present work we consider the hde model in the framework of dgp braneword with
go cutoff , @xmath31 , proposed in @xcite .
our work differs from @xcite in that , they studied hde model with go cutoff in standard cosmology , while we investigate this model in the framework of dgp braneword and incorporate the effect of the extra dimension on the evolution of the cosmological parameters on the brane .
the main difference between the hde with go cutoff in the framework of dgp braneword , with the one considered in standard cosmology @xcite , is that the equation of state parameter of the hde with go cutoff in standard cosmology is a constant @xcite , namely @xmath32 however , as we shall see in dgp braneworld , due to the bulk effects , @xmath33 becomes a time variable parameter .
clearly , a time variable de is more compatible with observations .
in particular , the analysis of data from wmap9 or planck-@xmath34 results on cmb anisotropy , bao distance ratios from recent galaxy surveys , magnitude - redshift relations for distant sne ia from snls3 and union2.1 samples indicate that the time varying de gives a better fit than a cosmological constant @xcite .
in addition , current data still slightly favor the quintom de scenario with eos across the cosmological constant boundary @xmath35 @xcite .
furthermore , we are able to derive explicitly , the cosmological parameters on the brane as functions of redshift parameter and provide a profile of the cosmic evolution on the brane . indeed , this is the advantages of the present model in compared to all other hde models .
as far as we know , this is the first model of hde which leads to analytical solution , @xmath36 and @xmath5 .
then , we establish the correspondence between our model and scalar field models of de and reconstruct the evolution of scalar fields and potentials . finally , we investigate the stability of this model against perturbation in different cases .
we find that for some range of the parameter spaces our model is stable which indicates the viability of this model for explanation of the late time acceleration .
we consider a homogeneous and isotropic frw universe on the brane which is described by the line element @xmath37 where @xmath38 represent a flat , closed and open maximally symmetric space on the brane , respectively . the modified friedmann equation in dgp braneworld model
is given by @xcite @xmath39 where @xmath40 corresponds to the two branches of solutions @xcite , and @xmath41 stands for crossover length scale between the small and large distances in dgp braneworld defined as @xcite @xmath42 we also assume that there is no energy exchange between the brane and the bulk and so the energy conservation equation holds on the brane , @xmath43 for @xmath44 , the friedmann equation in standard cosmology is recovered @xmath45 recent observations indicate that our universe is spatially flat . for a flat frw universe on the brane , eq .
( [ friedeq01 ] ) reduces to @xmath46 depending on the sign of @xmath47 , there are two different branches for the dgp model . for @xmath48 and in the absence of any kind of energy or matter field on the brane ( @xmath49 ) , there is a de - sitter solution for eq .
( [ friedeq02 ] ) with constant hubble parameter @xmath50 clearly , eq .
( [ de sitter ] ) leads to an accelerating universe with constant equation of state parameter @xmath51 , exactly like the cosmological constant . however , there are some unsatisfactory problems with this solution .
first of all , it suffers the well - known cosmological constant problems namely , the fine - tuning and the coincidence problems . besides , it leads to a constant @xmath52 , while many cosmological evidences , especially the analysis of the type ia supernova data indicates that the time varying de gives a better fit than a cosmological constant @xcite .
most of these data favor the evolution of the equation of state parameter and in particular it can have a transition from @xmath53 to @xmath54 at recent stage .
although some evidence such as the galaxy cluster gas mass fraction data do not support the time - varying @xmath52 @xcite , an overwhelming flood of papers has appeared to understand the @xmath51 crossing in the past decade @xcite .
in addition , to arrive at eq .
( [ de sitter ] ) one ignores all parts of energy on the brane including de , dark matter and byronic matter , which is not a reasonable assumption . on the other hand , for @xmath55 and @xmath56 , one can neglect the term @xmath57 in eq .
( [ friedeq02 ] ) to arrive at @xmath58 this is the friedmann equation in spatially flat rs ii braneworld @xcite . clearly , eq . ( [ rs ] )
does not have a self accelerating solution , so it implies the requirement of some kind of de on the brane . in the present paper
, we consider the hde model in flat dgp braneworld with go cutoff , which is defined as @xcite @xmath59 with this ir cutoff , the energy density ( [ hde ] ) can be written @xmath60 where @xmath61 and @xmath62 are constants which should be constrained by the recent observational data and we have absorbed the constant @xmath17 in @xmath61 and @xmath62 .
hereafter , we work in a unit in which @xmath63 .
we also restrict our study to the current cosmological epoch , and hence we are not considering the contributions from matter and radiation by assuming that the dark energy @xmath64 dominates , thus the friedman equation becomes simpler . substituting eq .
( [ rho1 ] ) into friedmann equation ( [ friedeq02 ] ) we can obtain the differential equation for the hubble parameter as + @xmath65 solving this equation , the hubble parameter is obtained as @xmath66 where @xmath67 is constant of integration . since for @xmath68
, the effects of the extra dimension should be disappeared and the result of @xcite , namely @xmath69 must be restored , thus the constant @xmath67 should be chosen as @xmath70 substituting @xmath67 in eq .
( [ hubble1 ] ) , we obtain @xmath71 taking the time derivative of eq .
( [ hubble2 ] ) yields @xmath72 we can also solve eq .
( [ hubble2 ] ) to obtain the scale factor .
we find @xmath73 using the fact that @xmath74 , where @xmath3 is the redshift parameter , and combining eqs .
( [ hubble2 ] ) and ( [ scale ] ) we find explicitly the hubble parameter as a function of @xmath3 , @xmath75}{(1+z)^{\frac{1-\alpha}{\beta}}}\nonumber \\ & = & \frac{\epsilon}{r_{\rm c}(1-\alpha)}\left[1+(1+z)^{\frac{\alpha-1}{\beta}}\right].\end{aligned}\ ] ] this equation is valid in two cases .
first , for @xmath48 and @xmath10 , and second for @xmath55 and @xmath8 , and has no solution for @xmath76 . inserting eqs .
( [ hubble2 ] ) and ( [ hubbledot ] ) in ( [ rho1 ] ) , we get @xmath77 substituting eqs .
( [ hubble2 ] ) and ( [ hubbledot ] ) in the time derivative of eq .
( [ rho1 ] ) , @xmath78 we arrive at @xmath79.\ ] ] combining eqs .
( [ hubble2 ] ) , ( [ rho ] ) and ( [ rhodot ] ) with ( [ conserveq ] ) , we find the equation of state parameter of hde on the brane as a function of time , @xmath80.\ ] ] in order to see the evolution of @xmath52 during the history of the universe , it is better to express it as a function of the redshift parameter @xmath3 , @xmath81.\ ] ] versus redshift parameter @xmath3 for different model parameters . left panel corresponds to @xmath10 and @xmath11 , while the right panel shows the case @xmath8 and @xmath9 . , title="fig:",width=302 ] versus redshift parameter @xmath3 for different model parameters
. left panel corresponds to @xmath10 and @xmath11 , while the right panel shows the case @xmath8 and @xmath9 . , title="fig:",width=302 ] to check the limit of ( [ wdz ] ) in standard cosmology where @xmath82 , we note that , from eq .
( [ scale ] ) and relation @xmath83 we have @xmath84 , as @xmath82 .
therefore , ( [ wdz ] ) reduces to @xmath85 which is exactly the result obtained in @xcite .
it is worth mentioning that unlike in standard cosmology , where the equation of state parameter of hde with go cutoff becomes a constant @xcite , in dgp braneworld scenario @xmath52 varies with time .
thus , it seems that the presence of the extra dimension brings rich physics .
for @xmath76 we have @xmath51 , similar to the cosmological constant .
let us consider the case with @xmath10 and @xmath8 separately .
in the first case where @xmath10 , we find that the equation of state parameter can explain the acceleration of the universe provided @xmath11 . in the second case where @xmath8 , the accelerated expansion can be achieved for @xmath9 . in both cases ,
our universe has a transition from deceleration to the acceleration phase around @xmath86 , compatible with the recent observations @xcite , and mimics the cosmological constant at the late time .
these results can be easily seen in fig .
[ w ] which shows the behaviour of @xmath52 versus @xmath3 for both cases .
the first and second derivatives of the distance can be combined to obtain the deceleration parameter @xmath2 .
it was shown that the zero redshift value of @xmath87 , is independent of space curvature , and can be obtained from the first and second derivatives of the coordinate distance @xcite .
it was argued that @xmath87 , which indicates whether the universe is accelerating at the current epoch , can be obtained directly from the supernova and radio galaxy data @xcite .
the deceleration parameter is given by @xmath88 using eqs .
( [ hubble2 ] ) and ( [ hubbledot ] ) , the deceleration parameter is obtained as @xmath89}.\ ] ] for @xmath76 we have @xmath90 .
although , the equation of state and the deceleration parameters do not depend explicitly on the crossover length scale @xmath41 which is the characterization of the dgp branworld , they depend on @xmath41 via the relation between the scale factor and the redshift parameter as in ( [ scale ] ) .
we have plotted the behavior of the deceleration parameter versus @xmath3 in figure [ q ] .
form this figures , we see that the transition from deceleration phase to the acceleration phase can be occurred around @xmath86 , which is consistent with the recent observations @xcite .
the zero - redshift value of the deceleration parameter is obtained as @xmath91 recent cosmological data from the combined sample of @xmath92 supernovae and @xmath93 radio galaxies , show that @xmath94 , for a window function of width @xmath95 in redshift and a transition from deceleration to the acceleration phase at redshift @xmath96 @xcite .
the zero - redshift value ( [ q2 ] ) , for @xmath97 and @xmath98 , leads to @xmath99 , while for @xmath100 and @xmath101 , we get @xmath102 , which is consistent with the observations @xcite . versus redshift parameter @xmath3 for different model parameters .
left panel corresponds to @xmath10 and @xmath11 , while the right panel shows the case @xmath8 and @xmath9 .
, title="fig:",width=302 ] versus redshift parameter @xmath3 for different model parameters . left panel corresponds to @xmath10 and @xmath11 , while the right panel shows the case @xmath8 and @xmath9 .
, title="fig:",width=302 ]
the dynamical de proposal is often realized by some scalar field mechanism which suggests that the energy form with negative pressure is provided by a scalar field evolving down a proper potential .
scalar fields naturally arise in particle physics including supersymmetric field theories and string / m theory .
therefore , scalar field is expected to reveal the dynamical mechanism and the nature of de . however , although fundamental theories such as string / m theory do provide a number of possible candidates for scalar fields , they do not predict its potential , @xmath103 , uniquely .
consequently , it is meaningful to reconstruct the potential @xmath103 from some de models possessing some significant features of the quantum gravity theory , such as hde models .
famous examples of scalar field de models include quintessence @xcite , k - essence @xcite , tachyon @xcite , phantom @xcite , ghost condensate @xcite , quintom @xcite , and so forth . for a comprehensive review on scalar filed models of dark energy ,
see @xcite .
generically , there are two points of view on the scalar field models of dynamical de .
one viewpoint regards the scalar field as a fundamental field of the nature .
the nature of de is , according to this viewpoint , completely attributed to some fundamental scalar field which is omnipresent in supersymmetric field theories and in string / m theory .
the other viewpoint supports that the scalar field model is an effective description of an underlying theory of de .
if we regard the scalar field model as an effective description of such a theory , we should be capable of using the scalar field model to mimic the evolving behavior of the hde and reconstructing the scalar field model according to the evolutionary behavior of hde . in this section
we implement a correspondence between hde in dgp braneworld with go cutoff and various scalar field models , by equating the equation of state parameters for these models with the obtained equation of state parameter of eq .
( [ wdz ] ) .
let us start with reconstructing the potential and dynamics of quintessence scaler field .
the energy density and pressure of the quintessence scalar field are given by @xmath104 thus the potential and the kinetic energy term can be written as @xmath105 where @xmath106 . in order to implement the correspondence between hde and quintessence scaler field , we identify @xmath107 and @xmath108 . using eqs .
( [ wdz ] ) , ( [ vphi ] ) and ( [ dotphi ] ) , we find @xmath109 @xmath110 + \frac{1-\alpha}{3\beta}\left((1+\alpha)(1+z)^{\frac{1-\alpha}{\beta}}+2\right)\big].\ ] ] eq .
( [ phidotq ] ) can also be rewritten @xmath111 using relation @xmath112 one can also obtain @xmath113 integrating , yields @xmath114 for different parameters . in the right panel the behavior of the quintessence scalar field @xmath115 for hde model in dgp braneworld is illustrated for different parameters.,title="fig:",width=302 ] for different parameters . in the right panel the behavior of the quintessence scalar field @xmath115 for hde model in dgp braneworld is illustrated for different parameters.,title="fig:",width=302 ] therefore , we have established holographic quintessence de model and reconstructed the potential and the dynamics of scalar field according to evolutionary behaviour of hde on the brane .
theoretically , one may expect to omit the redshift @xmath3 from ( [ vphiq ] ) and ( [ phizq ] ) to derive the potential as a function of the scalar field , namely , @xmath103 .
however , due to the complexity of the equations , the analytical form of the potential @xmath116 is hard to be derived .
nevertheless , we can plot the reconstructed potential as a function of @xmath117 numerically .
the reconstructed quintessence potential @xmath103 and the evolutionary form of the field are plotted in figure [ quint ] , where we have taken the zero redshift value of the scalar field equal to zero , namely , @xmath118 .
selected curves are plotted for @xmath10 and the different model parameter @xmath11 .
we can also plot the figures for the case @xmath8 and @xmath9 , however , the behaviour is the same as the former case and for the economic reason we have not plotted here the latter case . from these figures
we can see the dynamics of the potential as well as the scalar field explicitly .
figure [ quint ] shows that the reconstructed quintessence potential is steeper in the early epoch and becomes flat near the present time . in other words , it mimics a cosmological constant at the the present time .
the tachyon field is another candidate for de .
the equation of state parameter of the rolling tachyon smoothly interpolates between @xmath119 and @xmath120 @xcite .
thus , tachyon can be realized as a suitable candidate for the inflation at high energy @xcite as well as a source of de depending on the form of the tachyon potential @xcite .
choosing different self - interacting potentials in the tachyon field model lead to different consequences for the resulting de model . due to all the above reasons , the reconstruction of tachyon potential @xmath103 is of great importance .
the correspondence between tachyon field and various de scenarios such as hde @xcite and agegraphic de @xcite has been already established .
the study has also been generalized to the entropy corrected holographic and agegraphic de models @xcite .
the tachyon condensates in a class of string theories and can be described by an effective scalar field with a lagrangian of the form by@xcite @xmath121 where @xmath103 is the tachyon potential .
the corresponding energy momentum tensor for the tachyon field can be written in a perfect fluid form @xmath122 where @xmath123 and @xmath124 are , respectively , the energy density and pressure of the tachyon , and the velocity @xmath125 is @xmath126 the energy density and pressure of tachyon field are given by @xmath127 thus the equation of state parameter of tachyon field is given by @xmath128 to establish the correspondence between hde and tachyon field , we equate @xmath52 with @xmath129 . combining eqs .
( [ wdz ] ) , ( [ rhot ] ) and ( [ pt ] ) , we find @xmath130}\frac{(1+\alpha(1+z)^{\frac{1-\alpha}{\beta}})(1+(1+z)^{\frac{1-\alpha}{\beta}})}{(1+z)^{\frac{2(1-\alpha)}{\beta}}},\\ \dot{\phi}^2(z)&=&\frac{\alpha-1}{3\beta } \left[\frac{(1+\alpha)(1+z)^{\frac{1-\alpha}{\beta}}+2}{(1+(1+z)^{\frac{1-\alpha}{\beta}})(1+\alpha ( 1+z)^{\frac{1-\alpha}{\beta}})}\right].\label{phit1}\end{aligned}\ ] ] we can further rewrite ( [ phit1 ] ) as @xmath131^{\frac{1}{2}}.\ ] ] using the fact that @xmath132 we have @xmath133^{\frac{1}{2}}.\ ] ] integrating and setting the constant of integration equal to zero by assuming @xmath118 , we obtain @xmath134^{\frac{1}{2}}dz}.\ ] ] for different parameters . in the right panel the behavior of the tachyon field @xmath115 for hde model in dgp braneworld is illustrated for different parameters.,title="fig:",width=302 ] for different parameters . in the right panel the behavior of the tachyon field @xmath115 for hde model in dgp braneworld is illustrated for different parameters.,title="fig:",width=302 ] in this way we connect the hde in dgp braneworld with a tachyon field , and reconstruct the potential and the dynamics of the tachyon field which describe tachyon cosmology . the reconstructed tachyon field and the evolution of the tachyon potential are plotted in figure [ tachyon ] .
again , we see that tachyon potential is steeper in the early epoch and becomes flat near today . thus , the universe enters a de - sitter phase at the late time and the constant potential plays the role of cosmological constant .
in this section we would like to investigate the stability of the hde in dgp braneworld against perturbation .
it is expected that any viable de model should result a stable de dominated universe .
a simple , but not complete , approach to check the stability of a proposed de model is to study the behavior of the square sound speed ( @xmath135 ) @xcite .
it was argued that the sign of @xmath136 plays a crucial role in determining the stability of the background evolution .
if @xmath137 , it implies that the perturbation of the background energy density is not an oscillatory function and may grow or decay with time , and so we have the classical instability of a given perturbation . on the other hand , the positivity of @xmath138 indicates that the perturbation in the energy density , propagates in the environment and so we expect a stable universe against perturbations .
it is important to note that the positivity of @xmath136 is necessary but is not enough to conclude that the model is stable .
indeed , the negativity of it shows a sign of instability in the model .
the behavior of the squared sound speed for hde @xcite , agegraphic de @xcite and the ghost de model @xcite were investigated .
it was found that all these models @xcite are instable against background perturbations and so can not lead to a stable de dominated universe . in the linear perturbation regime
, the perturbed energy density of the background can be written as @xmath139 where @xmath140 is unperturbed background energy density .
the energy conservation equation ( @xmath141 ) yields @xcite @xmath142 for @xmath12 , eq .
( [ pert2 ] ) becomes a regular wave equation whose solution is given by @xmath143 , which indicates a propagation mode for the density perturbations . for @xmath144 ,
the perturbation becomes an irregular wave equation and in this case the frequency of the oscillations are pure imaginary and the density perturbation will grow with time as @xmath145 .
hence the negative squared speed shows an exponentially growing mode for a density perturbation .
thus the growing perturbation with time indicates a possible emergency of instabilities in the background . for hde model in dgp braneworld for different model parameters .
left panel corresponds to @xmath10 and @xmath11 , while the right panel shows the case @xmath8 and @xmath9 .
, title="fig:",width=302 ] for hde model in dgp braneworld for different model parameters . left panel corresponds to @xmath10 and @xmath11 , while the right panel shows the case @xmath8 and @xmath9 .
, title="fig:",width=302 ] the squared speed of sound can be written as @xmath146 where @xmath147 taking the time derivative of ( [ wdz ] ) and substituting the result in ( [ vs1 ] ) , after some calculation , we arrive at @xmath148 - 1.\ ] ] the evolution of @xmath7 versus redshift parameter are plotted in figure [ vs ] . from these figures
we see that @xmath7 can be positive provided the parameters of the model are chosen suitably . for example
, for @xmath10 , @xmath11 , and also for @xmath8 , @xmath9 the square of sound speed is always positive during the history of the universe , and so in these cases the stable de dominated universe can be achieved .
in this paper , we have investigated the holographic model of de in the framework of dgp braneworld .
we have chosen the go ir cutoff in the form , @xmath149 , with two free parameters @xmath61 and @xmath62 , which should be constrained by comparing with observations .
this cutoff is the generalization of the well - known ricci scalar cutoff in flat universe .
we have restricted our study to the current cosmological epoch , and so we have not considered the contributions from matter and radiation by assuming that the dark energy @xmath64 dominates , thus the friedman equation becomes simpler .
the main difference between the hde with go cutoff studied in this work in the framework of dgp braneword , with the one considered in standard cosmology @xcite , is that the equation of state parameter of the hde with go cutoff in standard cosmology is a constant @xcite , however , in dgp braneworld , due to the bulk effects , @xmath33 becomes a time variable parameter .
the application of the hde with go cutoff , in dgp braneworld allows us to solve the friedmann equation and derive the hubble parameter , @xmath150 , as well as the scale factor , @xmath151 , analytically .
we also obtained the equation of state and the deceleration parameters of hde as the functions of the redshift parameter @xmath3 and plotted the evolutionary behaviour of these parameters against @xmath3 .
the cosmological quantities depend on the two model parameter @xmath61 and @xmath62 .
to show the viability of the model , we studied the zero - redshift values of the cosmological parameters by chosen the suitable values for @xmath61 and @xmath62 .
for example , we found that for @xmath97 and @xmath152 , the transition from deceleration phase to the acceleration phase occurred around @xmath153 , which is consistent with recent observations @xcite .
the zero - redshift value of the deceleration parameter was obtained @xmath99 for @xmath97 , @xmath98 , and @xmath102 for @xmath100 , @xmath101 which is again compatible with the cosmological data @xcite .
besides , for @xmath76 this model mimics a cosmological constant with @xmath51 , and @xmath90 , independent of the redshift parameter @xmath3 . for @xmath8 , @xmath9 and @xmath10 , @xmath11 the equation of state parameter , however , is always larger than @xmath119 , and the universe enters a de - sitter phase at the late time .
we have also established a connection between the quintessence / tachyon scalar field and the hde in dgp braneworld . as a result
, we reconstructed the corresponding potentials of the scalar field , @xmath103 , and the dynamics of the scalar fields as a function of redshift parameter , @xmath154 , according to the evolutionary behavior of the hde model .
finally , we studied the stability of the presented model by studying the evolution of the squared sound speed @xmath7 whose sign determines the sound stability of the model .
interestingly enough , we found that @xmath12 provided the parameters of the model are chosen suitably .
for example , for @xmath97 , @xmath155 , and @xmath100 , @xmath156 the squared sound speed is always positive during the history of the universe , and so in this case the stable de dominated universe can be achieved .
this is in contrast to hde in standard cosmology , which is instable against background perturbations and so can not lead to a stable de dominated universe @xcite .
this implies that the presence of the extra dimension in hde model , can bring rich physics and in particular it has an important effect on the stability of the hde model .
this issue deserves further investigations
. 99 a.g .
riess , et al .
j. * 116 * , 1009 ( 1998 ) ; + s. perlmutter , et al .
, astrophys .
j. * 517 * , 565 ( 1999 ) ; + p. debernardis , et al . , nature * 404 * , 955 ( 2000 ) ; + s. perlmutter , et al .
, astrophys .
j. * 598 * , 102 ( 2003 ) .
s. weinberg , rev .
* 61 * , 1 ( 1989 ) .
a. g. cohen , d. b. kaplan and a. e. nelson , phys .
* 82 * , 4971 ( 1999 ) .
p. horava and d. minic , phys .
lett . * 85 * , 1610 ( 2000 ) ; + s. d. thomas , phys .
lett . * 89 * , 081301 ( 2002 ) .
s. d. h. hsu , phys .
b * 594 * , 13 ( 2004 ) .
d. pavon , w. zimdahl , phys .
b * 628 * , 206 ( 2005 ) ; + a. sheykhi , phys .
d * 84 * , 107302 ( 2011 ) .
a. sheykhi , phys .
b * 680 * , 113 ( 2009 ) ; + a. sheykhi , phys .
b * 682 * , 329 ( 2010 ) 329 ; + a. sheykhi , phys .
d * 81 * , 023525 ( 2010 ) ; + ahmad sheykhi , mubasher jamil , phys .
b * 694 * , 284 ( 2011 ) ; + a. sheykhi , m. r. setare , int . j. theor
. phys . * 49 * , 2777 ( 2010 ) . c. j. gao ,
nt . j. mod .
* 10 * , 95 ( 2012 ) .
xia , h. li , x. zhang , phys .
d * 88 * , 063501 ( 2013 ) .
w. yang , l. xu , phys .
d * 89 * , 083517 ( 2014 ) ; + b. novosyadlyj , o. sergijenko , r. durrer , v. pelykh , arxiv:1312.6579 u. alam , v. sahni , a.a .
starobinsky , jcap * 0406 * , 008 ( 2004 ) ; + d. huterer , a. cooray , phys .
d * 71 * , 023506 ( 2005 ) ; + y.g .
gong , int .
d * 14 * , 599 ( 2005 ) ; + y.g .
gong , class .
quantum grav .
* 22 * , 2121 ( 2005 ) .
b. wang , y. gong and e. abdalla , phys .
b * 624 * , 141 ( 2005 ) ; + b. wang , c. y. lin and e. abdalla , phys .
b * 637 * , 357 ( 2005 ) ; + m. r. setare , phys .
b * 642*,1 ( 2006 ) .
p. binetruy , c. deffayet , u. ellwanger , d. langlois , phys . lett .
b * 477 * , 285 ( 2000 ) .
p. j. e. peebles and b. ratra , astrophys .
j. * 325 * , l17 ( 1988 ) ; + b. ratra and p. j. e. peebles , phys .
d * 37 * , 3406 ( 1988 ) ; + c.wetterich , nucl .
b * 302 * , 668 ( 1988 ) ; + j. a. frieman , c. t. hill , a. stebbins and i. waga , phys .
lett . * 75 * , 2077 ( 1995 ) ; + x. zhang , mod .
a * 20 * , 2575 ( 2005 ) ; + x. zhang , phys .
b * 611 * , 1 ( 2005 ) .
a. mazumdar , s. panda , and a. perez - lorenzana , nucl .
b * 614 * , 101 ( 2001 ) ; + a. feinstein , phys .
d * 66 * , 063511 ( 2002 ) ; + y. s. piao , r. g. cai , x. m. zhang , and y. z. zhang , phys .
d * 66 * , 121301 ( 2002 ) .
t. padmanabhan , phys .
d * 66 * , 021301 ( 2002 ) ; + j.s .
bagla , h. k. jassal , and t. padmanabhan , phys .
d * 67 * , 063504 ( 2003 ) ; + e. j. copeland , m. r. garousi , m. sami , and s. tsujikawa , phys .
d * 71 * , 043003 ( 2005 ) .
m. r. setare , j. sadeghi , a.r .
amani , phys .
b * 673 * , 241 ( 2009 ) ; + m. r. setare , phys .
b 653 , 116 ( 2007 ) ; + k. karami , j. fehri , phys . lett .
b * 684 * , 61 ( 2010 ) .
j. cui , l. zhang , j. zhang , and x. zhang , chin .
b * 19 * , 019802 ( 2010 ) ; + a. sheykhi , phys .
b * 682 * , 329 ( 2010 ) .
m. jamil and a. sheykhi , int .
. phys . * 50 * , 625 ( 2011 ) ; + m. umar farooq , muneer a. rashid , mubasher jamil , int .
* 49 * , 2278 ( 2010 ) ; + e. ebrahimi and a. sheykhi , phys .
scr . * 84 * , 045016 ( 2011 ) ; + k. karami , m. s. khaledian , m. jamil , phys .
scr . * 83 * , 025901 ( 2011 ) . | we consider the holographic dark energy ( hde ) model in the framework of dgp braneworld with granda - oliveros infrared ( ir ) cutoff , @xmath0 . with this choice for ir cutoff
, we are able to derive evolution of the cosmological parameters such as the equation of state and the deceleration parameters , @xmath1 and @xmath2 , as the functions of the redshift parameter @xmath3 . as far as we know
, most previous models of hde presented in the literatures , do not gives analytically @xmath4 and @xmath5 .
we plot the evolution of these parameters versus @xmath3 and discuss that the results are compatible with the recent observations . with suitably choosing the parameters , this model can exhibit a transition from deceleration to the acceleration around @xmath6 .
then , we suggest a correspondence between the quintessence and tachyon scalar fields and hde in the framework of dgp braneworld .
this correspondence allows us to reconstruct the evolution of the scalar fields and the scalar potentials .
we also investigate stability of the presented model by calculating the squared sound speed , @xmath7 , whose sign determines the stability of the model .
our study shows that @xmath7 could be positive provided the parameters of the model are chosen suitably .
in particular , for @xmath8 , @xmath9 , and @xmath10 , @xmath11 , we have @xmath12 during the history of the universe , and so the stable dark energy dominated universe can be achieved .
this is in contrast to the hde in standard cosmology , which is unstable against background perturbations and so can not lead to a stable dark energy dominated universe . |
after the discovery @xcite of @xmath10 followed by the other experiments @xcite , identifying the properties of the particle is one of the central problems in hadron physics . while the isospin of @xmath11 is likely to be zero @xcite , the spin and the parity and the origin of its tiny width still remain open questions @xcite . in spite of many theoretical studies on @xmath11 @xcite ,
the nature of this exotic particle , including the very existence of the particle , is still controversial . among theoretical approaches ,
the lattice qcd calculation is considered as one of the most reliable _ ab initio _
methods for studying the properties of hadronic states , which has been very successful in reproducing the non - exotic hadron mass spectra @xcite . up to now , several lattice qcd studies have been reported , which aim to look for pentaquarks in various different ways . however , the conclusions are unfortunately contradictory with each other . on one hand , the authors in refs .
@xcite conclude that the parity of @xmath11 is likely to be negative , while in ref .
@xcite the state with the similar mass to @xmath11 in the positive parity channel is reported . in refs .
@xcite , the absence of @xmath11 is suggested .
one of the difficulties in the spectroscopy calculation with lattice qcd arises from the fact that the hadron masses suffer from systematic errors due to the discretization , the chiral extrapolation , the quenching effect , the finite volume effect and the contaminations from higher excited - states .
the difficulty specific to the present problem is that the signal of @xmath11 is embedded in the discrete spectrum of nk scattering states in finite volume . in order to verify the existence of a resonance state
, one needs to isolate the first few low energy states including the lowest nk scattering state , identify a resonance state and study its volume dependence which can distinguish itself from other scattering states .
therefore , ideally one should extract multistates from a high statistics unquenched calculation for several different physical volumes , where both the continuum and the chiral limits are taken .
however , due to enormous computational costs , so far there are no lattice qcd study which performs all these steps . in the present situation where even the very existence of the resonance state is theoretically in dispute ,
the primary task is to provide evidences which distinguish the candidate resonance state from a scattering state .
as long as other systematic errors only affect the numerical values of the masses but not the characteristic evidences of the resonance state , they may be neglected .
even so , the isolation of the first few low energy states and the study of the volume dependence is a minimum requisite .
therefore , at this stage as a first step towards a more complete analysis , we propose to focus only on analyses using rather heavy quarks on coarse quenched lattices but with a good statistics .
by such a strategy , we can afford taking several different lattice volumes with thousands of gauge configurations so that the careful separation of states and the studies of volume dependence are possible .
although giving well controlled continuum and chiral extrapolations may be important , we simply assume that the contents in spectra would not be drastically changed , although there are some cases where level crossings of resonance states occur as the quark masses decrease @xcite .
even with such a compromise , we can hopefully learn about the existence and much of the qualitative properties of @xmath11 . in this paper , we study @xmath8 channel in quenched lattice qcd to search for possible resonance states .
we adopt two independent operators with @xmath12 and @xmath13 and diagonalize the @xmath7 correlation matrices by the variational method for all the combinations of lattice sizes and quark masses to extract the 2nd - lowest state slightly above the nk threshold in this channel . after the careful separation of the states
, we investigate the volume dependence of the energy as well as the spectral weight @xcite of each state so that we can distinguish the resonance state from simple scattering states .
the paper is organized as follows .
we present the formalism used in the analysis in sec .
[ formalism ] and show the simulation conditions in sec .
[ setup ] .
the process of the analysis is shown in sec .
[ negative1]-[positive ] are devoted to the interpretation of the obtained results and the verification of the existence of a resonance state , as well as some checks on the consistency and the reliability of the obtained data . in sec .
[ discussion ] , we discuss the operator dependence of the results and compare our results with the previous works .
we finally summarize the paper in sec .
[ summary ] . in appendix
, we show the result of another trial to estimate the volume dependences of the spectral weights , which requires no multi - exponential fit .
as @xmath11 lies above the nk threshold , any hadron correlators , which have the @xmath11 signal , also contain the discrete - level nk scattering states in a finite volume lattice . in order to isolate the resonance state from the scattering states
, one needs to extract _ at least _ two states before anything else .
since a double - exponential fit of a single correlator becomes numerically ambiguous , we adopt the variational method using correlation matrices constructed from independent operators @xcite .
a set of independent operators , \{@xmath14 } for sinks and\{@xmath15 } for sources , is needed to construct correlation matrices @xmath16 , which can be decomposed into the sum over the energy eigenstates @xmath17 as @xmath18 with the general matrices which depend on the operators as @xmath19 and the diagonal matrix @xmath20 from the product @xmath21 we can extract the energies \{@xmath22 } as the logarithm of eigenvalues \{@xmath23 } of the matrix @xmath24 . while there are @xmath25 independent operators for the correlation matrix , the number of the intermediate states @xmath17 which effectively contribute to this matrix may differ from @xmath25 in general .
let us call this number as @xmath26 .
if @xmath26 is larger than @xmath25 , the higher excited - states are non negligible and their contaminations give rise to a @xmath27-dependence of eigenvalues as \{@xmath28}. if on the other hand @xmath26 is smaller than @xmath25 , @xmath29 becomes non - invertible so that the extracted energies become numerically fairly unstable and we can not extract all the @xmath25 eigenvalues . in order to have a reliable extraction of states , we therefore need to find an appropriate window of @xmath27 ( @xmath30 ) so that @xmath31 .
( of course , even in the case when @xmath32 , we can extract @xmath26 eigenvalues with the reduced @xmath33 correlation matrices . ) the stability of \{@xmath28 } against @xmath27 is expected in this @xmath27 range and we can obtain @xmath25 eigenenergies \{@xmath22 } ( @xmath34 ) by fitting the eigenvalues @xmath28 as @xmath35 in @xmath30 . since finding the stability of the energies against @xmath27 in noisy data may suffer from uncontrollable biases , the result could be quite subjective . in order to avoid such biases
, one should impose some concrete criteria to judge the stability as will be explained in later sections and select only those data which satisfy the criteria .
after the separation of the states , we need to distinguish a possible resonance state from nk scattering states by the volume dependence of each state .
it is expected that the energies of resonance states have small volume dependence , while the energies of nk scattering states are expected to scale as @xmath36 according to the relative momentum @xmath37 between n and k on a finite periodic lattice , provided that the nk interaction is weak and negligible which is indeed the case for the leading order in chiral perturbation theory .
although the variational method is powerful for extracting the energy spectrum , one can obtain only part of the information on the spectral weights @xmath38 . in order to extract the spectral weights , we also perform constrained double exponential fits using the energies from the variational method as inputs .
we carry out simulations on four different sizes of lattices , @xmath2 , @xmath3 , @xmath4 and @xmath5 with 2900 , 2900 , 1950 and 950 gauge configurations using the standard plaquette ( wilson ) gauge action at @xmath39 and the wilson quark action .
the hopping parameters for the quarks are @xmath40=@xmath41 , @xmath42 , @xmath43 , @xmath44 and @xmath45 , which correspond to the current quark masses @xmath46 , @xmath47 , @xmath48 , @xmath49 and @xmath50 , respectively in the unit of mev @xcite .
the lattice spacing @xmath51 from the sommer scale is set to be 0.17 fm , which implies the physical lattice sizes are @xmath52 @xmath53 , @xmath54 @xmath53 , @xmath55 @xmath53 and @xmath56 @xmath53 .
we adopt the following two operators used in ref .
@xcite for the interpolating operators at the sink @xmath57 ; @xmath58 \{u_e(x)[{\overline{s_e}(x)\gamma_5d_c(x ) } ] -(u\leftrightarrow d)\ } , \label{theta1}\end{aligned}\ ] ] which is expected to have a larger overlap with @xmath11 state , and @xmath59 \{u_c(x)[{\overline{s_e}(x)\gamma_5d_e(x ) } ] -(u\leftrightarrow d)\ } , \label{theta2}\end{aligned}\ ] ] which we expect to have larger overlaps with nk scattering states . here
, the dirac fields @xmath60 , @xmath61 and @xmath62 are up , down and strange quark fields , respectively and the roman alphabets \{a , b , c , e } denote color indices . for measuring the energy spectrum ,
the two operators at the source @xmath63 are chosen to be @xmath64 and @xmath65 defined using spatially spread quark fields @xmath66 with the coulomb gauge : @xmath67\{u_e(x_3)[{\overline{s_e}(x_4 ) \gamma_5 d_c(x_5 ) } ] -(u\leftrightarrow d)\},\end{aligned}\ ] ] and @xmath68\{u_c(x_3)[{\overline{s_e}(x_4 ) \gamma_5 d_e(x_5 ) } ] -(u\leftrightarrow d)\}.\end{aligned}\ ] ] the above operators give a @xmath7 correlation matrix in the channel with the quantum number of @xmath8 .
we note here that the baryonic correlators have the spinor indices , which we omit in the paper , and they contain the propagations of both the positive and negative parity particles . for the parity projection
, we simply multiply the correlators by @xmath69 and extract the contributions proportional to @xmath70 from the negative - parity and positive - parity particles , respectively .
we fix the source operator @xmath71 on @xmath72 plane to reduce the effect of the dirichlet boundary on @xmath73 plane @xcite .
we adopt the operators @xmath74 as sink operators , which is summed over all space to project out the zero - momentum states .
we finally calculate @xmath75 using two independent operators , we can extract the first _ two _ states , namely the lowest and the next - lowest states .
the lowest state is considered to be the `` lowest '' nk scattering state . in order to extract a possible resonance state with controlled systematic errors ,
we need to choose the physical volume of the lattice in an appropriate range .
if we choose @xmath76 to be too large , the resonance state becomes heavier than the 2nd - lowest nk scattering state whose energy is naively expected to scale as @xmath77 according to the spatial lattice extent @xmath76 . in this case
we need to extract the 3rd state using a @xmath78 correlation matrix , which requires more computational time .
the energy difference between the lowest and the next - lowest nk scattering states @xmath79 , for example , ranges from 180 mev to 860 mev in @xmath80 . taking into account that @xmath11 lies about 100 mev above the nk threshold , we take 3.5 fm as the upper limit of @xmath76 .
on the other hand , if we choose @xmath76 too small , unwanted finite - volume artifacts from the finite sizes of particles become non - negligible .
it is however difficult to estimate the lower limit of @xmath76 , because the finite - volume effect is rather uncontrollable .
we shall take the spatial extents @xmath81 at @xmath82 as a trial .
we take periodic boundary conditions in all directions for the gauge field , whereas we impose periodic boundary conditions on the spatial directions and the dirichlet boundary condition on the temporal direction for the quark field in order to avoid possible contaminations from those propagating beyond the boundary at @xmath73 in ( anti)periodic boundary conditions .
since the source of possible contaminations is peculiar to the pentaquark and has not been properly noticed in previous studies , it is worthwhile to dwell on this problem for a moment .
let us denote the correlators in the pentaquark channel with periodic / antiperiodic boundary conditions and dirichlet boundary conditions as @xmath83 and @xmath84 , respectively . inserting complete set of states
these correlators read @xmath85 where the states @xmath86 are the eigenstates with energies @xmath87 , @xmath88 , @xmath89 , @xmath90 respectively .
@xmath91 is the state which corresponds to the dirichlet boundary condition and @xmath92 is the factor which represents the @xmath93 sign with antiperiodic boundary condition .
the factor @xmath92 is equal to @xmath94 when @xmath95 contains an even(odd ) number of valence quarks . it should be noted that the up , down , and strange quark numbers @xmath96 for the state @xmath97 are restricted to zero since all the quark fields @xmath98 are set to zero in dirichlet boundary condition , while there is no restriction in the quark sector for the states @xmath95 which appear in periodic / antiperiodic boundary conditions .
this means that the following states can contribute in the correlators @xmath99 where @xmath100 and @xmath101 are the antiparticle states of @xmath102 and @xmath103 , respectively .
therefore , the correlators have the contributions in the long range limit , @xmath104 in fig . [ hadronic_corr ]
, we give a schematic picture of the contributions to the correlators .
the first two terms in eq.([cpap ] ) are the contributions from the five quark states as given in diagram ( a ) .
the third , the fourth and the fifth terms in eq.([cpap ] ) which correspond to diagrams ( c),(d ) and ( b ) are hadronic contributions which propagate beyond the boundary .
schematic figure for the explanation on the possible contaminations of the particles propagating over the temporal boundary . @xmath105 and @xmath106 are the interpolating operators and their arguments are the distances from the source point .
@xmath105 is a generic hadronic operator , which creates and annihilates the particles which can not decay in quenched qcd .
the wave lines represent the propagations of states .
resonance states like @xmath11 are represented by `` 5q '' in the figure , as well as nk scattering states .
_ the five - quark state can dissociate into forward - propagating nucleon ( kaon ) and backward - propagating kaon ( nucleon ) .
_ ] as a result , the correlation @xmath107 inevitably contains unwanted contributions such as @xmath108 in this case , the effective mass plot approaches @xmath109 below the nk threshold as @xmath27 is increased . on the other hand , the contributions corresponding to diagrams ( c),(d ) and ( b ) do not exist with dirichlet boundary conditions ( eq.([cd ] ) ) .
therefore , we find that it would be safest to impose the dirichlet boundary condition on the temporal direction , since no quark can go over the boundary on t=0 in the temporal direction . although the boundary is transparent for the particles composed only by gluons ; _ i.e. _ glueballs , due to the periodicity of the gauge action , it would be however safe to neglect these gluonic particles going beyond the boundary since these particles are rather heavy .
then , the correlation @xmath110 mainly contains only such terms ( ( a ) in fig .
[ hadronic_corr ] ) as @xmath111 with @xmath112 the weight factor and @xmath113 the eigenenergy of @xmath114-th state in positive / negative parity channel , respectively .
one sees that one can now apply the prescription mentioned in the last section .
one may wonder if these contaminations can be discarded with the parity projection of the correlators by taking linear combinations with periodic and antiperiodic boundary conditions .
this method indeed works for ordinary three quark states where one can single out one of the two contributions diagram ( x ) and ( y ) in fig .
[ hadronic_corr ] .
however even if one takes such linear combinations , one can not make the contributions from diagram ( c ) seen in eq.([cpap ] ) cancel out as opposed to the contributions from diagram ( b ) and ( d ) .
it is because of the fact that the factor @xmath92 for the contribution ( c ) is always equal to @xmath115 .
( we note here that we can avoid these contaminations using the `` averaged quark propagator '' @xcite . )
some of the previous lattice qcd studies on @xmath11 adopted a parity projection method using the combination with periodic and antiperiodic boundary conditions @xcite .
we stress that one should in principle be careful whether the result is free from the contamination owing to the boundary condition which is peculiar to the pentaquark and can mimic a fake plateau in the propagator . after obtaining the energy spectrum ,
we carry out a study of the spectral weight for @xmath40=@xmath44 . introducing two smeared operators @xmath116 , @xmath117 we compute the following correlators @xmath118 from which we extract the spectral weights using a constrained double exponential fit .
the details will be explained in sec .
[ negative2 - 3 ] .
the `` effective mass '' plot @xmath119 as the function of @xmath27 , the separation between source operators and sink operators , in @xmath120 channel with the hopping parameters @xmath121 on @xmath2 , @xmath3 , @xmath4,@xmath5 lattice at @xmath82 .
the stability of each @xmath119 against @xmath27 means the smallness of the unwanted higher excited - state contaminations .
the solid line and the dashed lines represent the central value and the error of the fitted masses @xmath122 and @xmath123 . , title="fig : " ] the `` effective mass '' plot @xmath119 as the function of @xmath27 , the separation between source operators and sink operators , in @xmath120 channel with the hopping parameters @xmath121 on @xmath2 , @xmath3 , @xmath4,@xmath5 lattice at @xmath82 .
the stability of each @xmath119 against @xmath27 means the smallness of the unwanted higher excited - state contaminations .
the solid line and the dashed lines represent the central value and the error of the fitted masses @xmath122 and @xmath123 . , title="fig : " ] + the `` effective mass '' plot @xmath119 as the function of @xmath27 , the separation between source operators and sink operators , in @xmath120 channel with the hopping parameters @xmath121 on @xmath2 , @xmath3 , @xmath4,@xmath5 lattice at @xmath82 .
the stability of each @xmath119 against @xmath27 means the smallness of the unwanted higher excited - state contaminations .
the solid line and the dashed lines represent the central value and the error of the fitted masses @xmath122 and @xmath123 . , title="fig : " ] the `` effective mass '' plot @xmath119 as the function of @xmath27 , the separation between source operators and sink operators , in @xmath120 channel with the hopping parameters @xmath121 on @xmath2 , @xmath3 , @xmath4,@xmath5 lattice at @xmath82 .
the stability of each @xmath119 against @xmath27 means the smallness of the unwanted higher excited - state contaminations .
the solid line and the dashed lines represent the central value and the error of the fitted masses @xmath122 and @xmath123 . , title="fig : " ] the `` effective mass '' plot @xmath119 as the function of @xmath27 , the separation between source operators and sink operators , in @xmath120 channel with the hopping parameters @xmath124 on @xmath2 , @xmath3 , @xmath4,@xmath5 lattice at @xmath82 .
, title="fig : " ] the `` effective mass '' plot @xmath119 as the function of @xmath27 , the separation between source operators and sink operators , in @xmath120 channel with the hopping parameters @xmath124 on @xmath2 , @xmath3 , @xmath4,@xmath5 lattice at @xmath82 .
, title="fig : " ] + the `` effective mass '' plot @xmath119 as the function of @xmath27 , the separation between source operators and sink operators , in @xmath120 channel with the hopping parameters @xmath124 on @xmath2 , @xmath3 , @xmath4,@xmath5 lattice at @xmath82 .
, title="fig : " ] the `` effective mass '' plot @xmath119 as the function of @xmath27 , the separation between source operators and sink operators , in @xmath120 channel with the hopping parameters @xmath124 on @xmath2 , @xmath3 , @xmath4,@xmath5 lattice at @xmath82 .
, title="fig : " ] the `` effective mass '' plot @xmath119 as the function of @xmath27 , the separation between source operators and sink operators , in @xmath120 channel with the hopping parameters @xmath124 on @xmath2 , @xmath3 , @xmath4,@xmath5 lattice at @xmath82 .
, title="fig : " ] the `` effective mass '' plot @xmath119 as the function of @xmath27 , the separation between source operators and sink operators , in @xmath120 channel with the hopping parameters @xmath124 on @xmath2 , @xmath3 , @xmath4,@xmath5 lattice at @xmath82 .
, title="fig : " ] + the `` effective mass '' plot @xmath119 as the function of @xmath27 , the separation between source operators and sink operators , in @xmath120 channel with the hopping parameters @xmath124 on @xmath2 , @xmath3 , @xmath4,@xmath5 lattice at @xmath82 .
, title="fig : " ] the `` effective mass '' plot @xmath119 as the function of @xmath27 , the separation between source operators and sink operators , in @xmath120 channel with the hopping parameters @xmath124 on @xmath2 , @xmath3 , @xmath4,@xmath5 lattice at @xmath82 .
, title="fig : " ] the `` effective mass '' plot @xmath119 as the function of @xmath27 , the separation between source operators and sink operators , in @xmath120 channel with the hopping parameters @xmath125 on @xmath2 , @xmath3 , @xmath4,@xmath5 lattice at @xmath82 .
, title="fig : " ] the `` effective mass '' plot @xmath119 as the function of @xmath27 , the separation between source operators and sink operators , in @xmath120 channel with the hopping parameters @xmath125 on @xmath2 , @xmath3 , @xmath4,@xmath5 lattice at @xmath82 .
, title="fig : " ] + the `` effective mass '' plot @xmath119 as the function of @xmath27 , the separation between source operators and sink operators , in @xmath120 channel with the hopping parameters @xmath125 on @xmath2 , @xmath3 , @xmath4,@xmath5 lattice at @xmath82 .
, title="fig : " ] the `` effective mass '' plot @xmath119 as the function of @xmath27 , the separation between source operators and sink operators , in @xmath120 channel with the hopping parameters @xmath125 on @xmath2 , @xmath3 , @xmath4,@xmath5 lattice at @xmath82 .
, title="fig : " ] the `` effective mass '' plot @xmath119 as the function of @xmath27 , the separation between source operators and sink operators , in @xmath120 channel with the hopping parameters @xmath125 on @xmath2 , @xmath3 , @xmath4,@xmath5 lattice at @xmath82 .
, title="fig : " ] the `` effective mass '' plot @xmath119 as the function of @xmath27 , the separation between source operators and sink operators , in @xmath120 channel with the hopping parameters @xmath125 on @xmath2 , @xmath3 , @xmath4,@xmath5 lattice at @xmath82 .
, title="fig : " ] + the `` effective mass '' plot @xmath119 as the function of @xmath27 , the separation between source operators and sink operators , in @xmath120 channel with the hopping parameters @xmath125 on @xmath2 , @xmath3 , @xmath4,@xmath5 lattice at @xmath82 .
, title="fig : " ] the `` effective mass '' plot @xmath119 as the function of @xmath27 , the separation between source operators and sink operators , in @xmath120 channel with the hopping parameters @xmath125 on @xmath2 , @xmath3 , @xmath4,@xmath5 lattice at @xmath82 .
, title="fig : " ] before obtaining the energies of the lowest state and the 2nd - lowest state , there are only a few simple steps .
first , we calculate the @xmath7 correlation matrix @xmath126 defined in eq.([correlation_matrix ] ) and obtain the `` energies '' \{@xmath119 } as the logarithm of eigenvalues \{@xmath28 } of the matrix product @xmath24 . after finding the @xmath27 range ( @xmath30 ) , where \{@xmath119 } are stable against @xmath27
, we can extract the energies @xmath22 by the least @xmath127-squared fit of the data as @xmath128 in @xmath30 .
since the volume dependence of the energy is crucial to judge whether the state is a resonance or not , a great care must be paid in extracting the energy .
therefore , the systematic error by the contaminations from higher excited - states should be avoided by a careful choice of fitting ranges .
for this reason , we impose the following criteria for the reliable extraction of the energies .
although this set of the criteria is nothing more than just one possible choice , we believe it is important to impose some concrete criteria for the fit so that we can reduce the human bias for the fit , though not completely . 1 .
the effective mass plot should have `` plateau '' for both the lowest and the 2nd - lowest states simultaneously in a fit range [ @xmath129 , @xmath130 , where the length @xmath131 should be larger than or equal to 3 ( @xmath132 ) .
2 . in the plateau region , the signal for the lowest and the 2nd - lowest states should be distinguishable , so that the gap between the central values of the lowest and the 2nd - lowest energies should be larger than their errors .
the fitted energies should be stable against the choice of the fit range ; i.e. the results of the fit with @xmath133 time slices and with @xmath134 time slices should be consistent within statistical errors for both the lowest and the 2nd - lowest states .
4 . the lowest state energy obtained by the diagonalization method using
the @xmath135 correlation matrix should be consistent with the value from a single exponential fit for a sufficiently large @xmath136 . if the fit does not satisfy the above conditions , we discard the result since either the data in the fit range may be contaminated by higher excited - states or the 2nd - lowest - state signal is too noisy for a reliable fit .
figs [ effective_mass ] show the `` effective mass '' plot @xmath119 for the heaviest combination of quarks @xmath121 . as is mentioned in sec .
[ formalism ] , we need to find the @xmath27 region ( @xmath30 ) where each @xmath119 shows a plateau . in the case of @xmath4 lattice in fig .
[ effective_mass ] , for example , we choose the fit range of @xmath137 and @xmath138 ( @xmath139 ) .
notice that the source operators are put on the time slice with @xmath140 .
the plateau in this region satisfies the above criteria so that we consider the fit @xmath122 and @xmath123 for the range @xmath141 as being reliable .
the situation is similar for the cases of @xmath3 and @xmath5 lattices . on the other hand , in the case of @xmath2 lattice we do not find a plateau region satisfying the above criteria .
[ meff221],[meff112],[meff331],[meff111 ] shows the `` effective mass '' plots for the combinations with smaller quark masses .
we find that the signal is noisier for the lighter quarks and the fit with the smaller volumes @xmath2 and @xmath3 lattices do not satisfy the criteria .
now we show the lattice qcd results of the lowest state in @xmath12 and @xmath142 channel .
the filled circles in fig . [ negativegses ] show the lowest - state energies @xmath122 in @xmath12 and @xmath142 channel on four different volumes . here
the horizontal axis denotes the lattice extent @xmath76 in the lattice unit and the vertical axis is the energy of the state .
the lower line denotes the simple sum @xmath143 of the nucleon mass @xmath144 and kaon mass @xmath145 obtained with the largest lattice .
though @xmath144 and @xmath145 are slightly affected by finite volume effects , the deviation of @xmath146 from @xmath147 is about a few % ( table [ results ] . ) .
we therefore simply use @xmath147 as a guideline . at a glance , we find that the energy of this state takes almost constant value against the volume variation and coincides with the simple sum @xmath143
. we can therefore conclude that the lowest state in @xmath12 and @xmath142 channel is the nk scattering state with the relative momentum @xmath148 . the good agreement with the sum
@xmath143 implies the weakness of the interaction between n and k. in fact , the scattering length in the @xmath12 channel is known to be tiny ( @xmath149 fm ) from compilations of hadron scattering experiments @xcite , whereas the current algebra prediction from pcac with su(3 ) symmetry predicts that the scattering length @xmath150 .
we here compare our data with the previous lattice qcd studies , which were performed with almost the same conditions as ours , in order to confirm the reliability of our data .
the well - known hadron masses @xmath151,@xmath152 and @xmath153 listed in table [ results ] can be compared with the values in ref .
our data are consistent with those in ref .
the lowest nk scattering state in @xmath9 channel is carefully investigated in ref .
@xcite with almost the same parameters as our present study .
it is worth comparing our data with them . for the complete check of our data
, we re - extract the lowest state by the ordinary single - exponential fit of the correlator as @xmath154 in the large - t region , and compare them with the present lattice data @xmath122 obtained by the multi - exponential method as well as the data in ref . @xcite . in table
[ results ] , we list the data of the lowest state @xmath155 obtained by the single - exponential fit
. they almost coincide with the present data @xmath122 extracted by the multi - exponential method with about 1% deviations , which may be considered as the slightly remaining contaminations of the higher excited - state . in ref .
@xcite , the authors extracted the energy difference @xmath156 with the hopping parameters @xmath157 using @xmath158 lattice .
we therefore compare our data @xmath156 obtained with the hopping parameters @xmath159 on @xmath4 lattice .
the energy difference @xmath160 in our study is found to be @xmath161 , which is consistent with the value of @xmath162 in ref .
@xcite taking into account that this error includes only statistical one .
it is now confirmed that the lowest state extracted using the multi - exponential method is consistent with the previous works and that our data and method are reliable enough to investigate the 2nd - lowest state in this channel .
the @xmath9 state is one of the candidates for @xmath10 . since @xmath11 is located above the nk threshold , it would appear as an excited state in this channel .
we show the lattice data of the 2nd - lowest state in this channel . in order to distinguish a possible resonance state from nk scattering states , we investigate the volume dependence of both the energy and the spectral weight of each state .
it is expected that the energies of resonance states have small volume dependence , while the energies of nk scattering states are expected to scale as @xmath36 according to the relative momentum @xmath37 between n and k on a finite periodic lattice , provided that the nk interaction is weak and negligible .
we can take advantage of the above difference for the discrimination .
a possible candidate for the volume dependence of the energies of nk scattering states is the simple formula as @xmath163 with the relative momentum @xmath37 between n and k in finite periodic lattices , which is justified on the assumption that nucleon and kaon are point particles and that the interaction between them is negligible . in practice
, there may be some corrections to the volume dependence of @xmath164 .
we therefore estimate here three possible corrections ; the existence of the nk interaction , the application of the momenta on a finite discretized lattice and the estimation of the implicit finite - size effects . there can be small hadronic interactions between nucleon and kaon , which may lead to correction to naively expected energy spectrum @xmath164 of the nk scattering states . using l " uscher formula @xcite
, one can relate the scattering phase shift to the energy shift from @xmath164 on finite lattices .
for example , in the case when a system belongs to the representation @xmath165 of cubic groups , which is relevant in the present case , the relation between the phase shift and the possible momentum spectra is @xmath166 here @xmath167 is the zeta function defined as @xmath168 with the eigenenergy @xmath169 on a finite lattice .
we have simply omitted the corrections from the partial waves with angular momenta higher than the next smallest one ( @xmath170 ) .
although our current quark masses are heavier than those of the real quarks , we use the empirical values of the phase shift in nk scattering in ref.@xcite , by simply neglecting the quark mass dependence .
the correction using the empirical values results in at most a few % larger energy than the simple formula @xmath164 within the volume range under consideration ; the energies are slightly increased by the weak repulsive force between nucleon and kaon .
one may claim that one has to adopt momenta on a finite discretized lattice : @xmath171 for kaon and @xmath172 for nucleon , respectively .
this correction turns out to be within only a few % lower energy than @xmath164 , although it is not certain whether this correction is meaningful or not for composite particles like nucleon or kaon .
we find that these corrections lead to at most a few % deviations from @xmath164 .
we then neglect these corrections for simplicity in the following discussion and use the simple form @xmath164 .
so far , we have neglected the implicit finite - size effects in @xmath173 , other than the explicit ones due to the lattice momenta @xmath174 .
some smart readers may suspect that the dispersions @xmath175 and @xmath176 may be affected by the uncontrollable finite - size effects due to the finite sizes of n and k , and no longer valid . in order to make sure of the small implicit artifacts especially with @xmath177 , which we are mainly interested in
, we also calculate the sum @xmath178 of energies of nucleon @xmath179 and kaon @xmath180 with the smallest non - zero lattice momentum @xmath181 .
we extract @xmath182 and @xmath183 from the correlators @xmath184 and @xmath185 .
these results are denoted by the open squares in fig .
[ negativegses ] as the sum @xmath178 .
the upper lines in fig . [ negativegses ] show @xmath186 .
the deviations of @xmath178 from @xmath187 are very small , which implies the smallness of the implicit finite - size artifacts .
therefore , provided that the interaction between nucleon and kaon is weak , which we assume throughout the present analysis , the naive expectation for the 2nd - lowest nk scattering states denoted by the upper line in fig .
[ negativegses ] would be able to follow the energies of the 2nd - lowest nk scattering states even on the @xmath188 lattices in our setup .
the black ( gray ) filled - squares denote the lattice qcd data of the 2nd - lowest state in @xmath9 channel extracted with @xmath134 ( @xmath133 ) data plotted against the lattice extent @xmath76 .
the filled circles represent the lattice qcd data @xmath122 of the lowest state in @xmath9 channel .
the open symbols are the sum @xmath178 of energies of nucleon @xmath179 and kaon @xmath180 with the smallest lattice momentum @xmath181 .
the upper line represents @xmath189 with @xmath190 the smallest relative momentum on the lattice .
the lower line represent the simple sum @xmath143 of the masses of nucleon @xmath144 and kaon @xmath145 .
we adopt the central values of @xmath144 and @xmath145 obtained on the largest lattice to draw the two lines .
, title="fig : " ] + the black ( gray ) filled - squares denote the lattice qcd data of the 2nd - lowest state in @xmath9 channel extracted with @xmath134 ( @xmath133 ) data plotted against the lattice extent @xmath76 .
the filled circles represent the lattice qcd data @xmath122 of the lowest state in @xmath9 channel .
the open symbols are the sum @xmath178 of energies of nucleon @xmath179 and kaon @xmath180 with the smallest lattice momentum @xmath181 .
the upper line represents @xmath189 with @xmath190 the smallest relative momentum on the lattice .
the lower line represent the simple sum @xmath143 of the masses of nucleon @xmath144 and kaon @xmath145 .
we adopt the central values of @xmath144 and @xmath145 obtained on the largest lattice to draw the two lines .
, title="fig : " ] the black ( gray ) filled - squares denote the lattice qcd data of the 2nd - lowest state in @xmath9 channel extracted with @xmath134 ( @xmath133 ) data plotted against the lattice extent @xmath76 .
the filled circles represent the lattice qcd data @xmath122 of the lowest state in @xmath9 channel .
the open symbols are the sum @xmath178 of energies of nucleon @xmath179 and kaon @xmath180 with the smallest lattice momentum @xmath181 .
the upper line represents @xmath189 with @xmath190 the smallest relative momentum on the lattice .
the lower line represent the simple sum @xmath143 of the masses of nucleon @xmath144 and kaon @xmath145 .
we adopt the central values of @xmath144 and @xmath145 obtained on the largest lattice to draw the two lines .
, title="fig : " ] + the black ( gray ) filled - squares denote the lattice qcd data of the 2nd - lowest state in @xmath9 channel extracted with @xmath134 ( @xmath133 ) data plotted against the lattice extent @xmath76 .
the filled circles represent the lattice qcd data @xmath122 of the lowest state in @xmath9 channel .
the open symbols are the sum @xmath178 of energies of nucleon @xmath179 and kaon @xmath180 with the smallest lattice momentum @xmath181 .
the upper line represents @xmath189 with @xmath190 the smallest relative momentum on the lattice .
the lower line represent the simple sum @xmath143 of the masses of nucleon @xmath144 and kaon @xmath145 .
we adopt the central values of @xmath144 and @xmath145 obtained on the largest lattice to draw the two lines .
, title="fig : " ] the black ( gray ) filled - squares denote the lattice qcd data of the 2nd - lowest state in @xmath9 channel extracted with @xmath134 ( @xmath133 ) data plotted against the lattice extent @xmath76 .
the filled circles represent the lattice qcd data @xmath122 of the lowest state in @xmath9 channel .
the open symbols are the sum @xmath178 of energies of nucleon @xmath179 and kaon @xmath180 with the smallest lattice momentum @xmath181 .
the upper line represents @xmath189 with @xmath190 the smallest relative momentum on the lattice .
the lower line represent the simple sum @xmath143 of the masses of nucleon @xmath144 and kaon @xmath145 .
we adopt the central values of @xmath144 and @xmath145 obtained on the largest lattice to draw the two lines .
, title="fig : " ] we compare the lattice data @xmath123 with the expected behaviors @xmath187 for the 2nd - lowest nk scattering states .
the filled squares in fig . [ negativegses ] denote @xmath123 , the 2nd - lowest - state energies in this channel .
the black and gray symbols are the lattice data obtained by the fits with @xmath134 and @xmath133 time slices , respectively ( see the criterion.3 in the sec . [ data ] ) .
the upper line shows the expected energy - dependence on @xmath191 of the 2nd - lowest nk scattering state @xmath186 estimated with the next - smallest relative momentum between n and k , and with the masses @xmath145 and @xmath144 extracted on the @xmath192 lattices . although the lattice qcd data @xmath123 and the expected lines @xmath193 almost coincide with each other on the @xmath192 lattices , which one may take as the characteristics of the 2nd - lowest scattering state , the data @xmath123 do not follow @xmath193 in the smaller lattices .
( at the smallest lattices with @xmath188 ( 1.4 fm ) in the physical unit , some results apparently coincide with each other again .
however we consider that the volume with @xmath194 fm is too small for the pentaquarks ; it is difficult to tell which is the origin of the coincidence , uncontrollable finite volume effects of the pentaquarks or expected volume dependence of the 2nd - lowest nk scattering state . )
especially when the quarks are heavy , composite particles will be rather compact and we expect smaller finite volume effects besides those arising from the lattice momenta @xmath195 .
moreover the statistical errors are also well controlled for the heavy quarks .
thus , the significant deviations in @xmath196 fm with the combination of the heavy quarks , such as @xmath197 , are reliable and the obtained states are difficult to explain as the nk scattering states
. therefore one can understand this behavior with the view that this state is a resonance state rather than a scattering state .
in fact , while the data with the lighter quarks have rather strong volume dependences which can be considered to arise due to the finite size of a resonance state , the lattice data exhibit almost no volume dependence with the combination of the heavy quarks especially in @xmath196 fm , which can be regarded as the characteristic of resonance states .
for further confirmation , we investigate the volume dependence of the spectral weight @xcite . as mentioned in sec .
[ formalism ] , the correlation function @xmath198 can be expanded as @xmath199 .
the spectral weight of the @xmath114-th state is defined as the coefficient @xmath200
corresponding to the overlap of the operator @xmath201 with the @xmath114-th excited - state . the normalization conditions of the field @xmath202 and the states @xmath203 give rise to the volume dependence of the weight factors @xmath200 in accordance with the types of the operators @xmath201 .
for example , in the case when a correlation function is constructed from a point - source and a zero - momentum point - sink , as @xmath204 , the weight factor @xmath200 takes an almost constant value if @xmath203 is the resonance state where the wave function is localized . if the state @xmath203 is a two - particle state , the situation is more complicated .
nevertheless if there is almost no interaction between the two particles , the weight factor is expected to be proportional to @xmath205 . in the case
when a source is a wall operator @xmath206 as taken in this work , a definite volume dependence of @xmath200 is not known .
therefore , we re - examine the lowest state and the 2nd - lowest state in @xmath9 channel using the locally - smeared source @xmath207 with @xmath208 , which we introduce to partially enhance the ground - state overlap .
since smeared operators , whose typical sizes are much smaller than the total volume , can be regarded as local operators , we can discriminate the states using the locally - smeared operators as in the case of point operators .
( we also investigated the weight factor using the point source .
the results are consistent with those obtained using the locally - smeared one , but are rather noisy . )
we adopt the hopping parameter @xmath197 and additionally employ @xmath209 lattice for this aim .
the spectral weight factors defined in sec .
[ negative2 - 3 ] are plotted against the lattice volume @xmath191 .
the left figure shows @xmath210 for the lowest state in @xmath9 channel and the right figure shows @xmath211 for the 2nd - lowest state . in the case
when the weight factor @xmath200 for the @xmath114-th state @xmath203 in a point - point correlator shows no volume dependence , @xmath203 is likely to be a resonance state . on the contrary ,
when the @xmath114-th state @xmath203 is a two - particle state , @xmath200 scales according to @xmath212 .
, title="fig : " ] the spectral weight factors defined in sec .
[ negative2 - 3 ] are plotted against the lattice volume @xmath191 .
the left figure shows @xmath210 for the lowest state in @xmath9 channel and the right figure shows @xmath211 for the 2nd - lowest state . in the case
when the weight factor @xmath200 for the @xmath114-th state @xmath203 in a point - point correlator shows no volume dependence , @xmath203 is likely to be a resonance state . on the contrary ,
when the @xmath114-th state @xmath203 is a two - particle state , @xmath200 scales according to @xmath212 .
, title="fig : " ] we extract @xmath210 and @xmath211 using the two - exponential fit as @xmath213 .
the fit with four free parameters @xmath210 , @xmath211 , @xmath214 and @xmath215 is however unstable and therefore we fix the exponents using the obtained values @xmath122 and @xmath123 .
the weight factors @xmath210 and @xmath211 are then obtained through the two - parameter fit as @xmath216 in as large @xmath136 range ( @xmath217 ) as possible in order to avoid the contaminations of the higher excited - states than the 2nd - excited state ( 3rd - lowest state ) , which will bring about the instability of the fitted parameters .
the fluctuations of @xmath122 and @xmath123 are taken into account through the jackknife error estimation .
[ weightfactor ] includes all the results with the various fit range as ( @xmath129,@xmath218)= ( 16,18),(17,19),(18,20 ) to see the fit - range dependence .
though the results have some fit - range dependences , the global behaviors are almost the same among the three .
the left figure in fig . [ weightfactor ] shows the weight factor @xmath210 of the lowest state in @xmath9 channel against the lattice volume @xmath191 .
we find that @xmath210 decreases as @xmath191 increases and that the dependence on @xmath191 is consistent with @xmath205 , which is expected in the case of two - particle states .
it is again confirmed that the lowest state in this channel is the nk scattering state with the relative momentum @xmath148 .
next , we plot the weight factor @xmath211 of the 2nd lowest state in the right figure . in this figure ,
almost no volume dependence against @xmath191 is found , which is the characteristic of the state in which the relative wave function is localized .
( in appendix , we try another prescription to estimate the volume dependences of the spectral weights , which requires no multi - exponential fit . )
this result can be considered as one of the evidences of a resonance state lying slightly above the nk threshold . to summarize this section ,
the volume dependence analysis of the eigenenergies and the weight factors of the 2nd - lowest state in @xmath9 channel suggests the existence of a resonance state .
although there remain the statistical errors and the possible finite - volume artifacts , the data can be consistently accounted for assuming the 2nd - lowest state to be different from ordinary scattering states .
if the 2nd - lowest state were an ordinary scattering state , one had to assume a large systematic errors for heavier quarks which is hard to understand consistently
the lattice qcd data in the @xmath219 channel are plotted against the lattice extent @xmath76 .
the solid line denotes the simple sum @xmath220 of the masses of the lowest - state negative - parity nucleon @xmath221 and kaon @xmath145 obtained with the largest lattice .
, title="fig : " ] the lattice qcd data in the @xmath219 channel are plotted against the lattice extent @xmath76 .
the solid line denotes the simple sum @xmath220 of the masses of the lowest - state negative - parity nucleon @xmath221 and kaon @xmath145 obtained with the largest lattice . ,
title="fig : " ] the lattice qcd data in the @xmath219 channel are plotted against the lattice extent @xmath76 .
the solid line denotes the simple sum @xmath220 of the masses of the lowest - state negative - parity nucleon @xmath221 and kaon @xmath145 obtained with the largest lattice .
, title="fig : " ] the lattice qcd data in the @xmath219 channel are plotted against the lattice extent @xmath76 .
the solid line denotes the simple sum @xmath220 of the masses of the lowest - state negative - parity nucleon @xmath221 and kaon @xmath145 obtained with the largest lattice .
, title="fig : " ] the lattice qcd data in the @xmath219 channel are plotted against the lattice extent @xmath76 .
the solid line denotes the simple sum @xmath220 of the masses of the lowest - state negative - parity nucleon @xmath221 and kaon @xmath145 obtained with the largest lattice .
, title="fig : " ] in the same way as @xmath9 channel , we have attempted to diagonalize the correlation matrix in @xmath219 channel using the wall - sources @xmath71 and the zero - momentum point - sinks @xmath222 . in this channel ,
the diagonalization is rather unstable and we find only one state except for tiny contributions of possible other states .
we plot the lattice data in fig .
[ positivess ] .
one finds that they have almost no volume dependence and that they coincide with the solid line which represents the simple sum @xmath220 of @xmath221 and @xmath145 , with @xmath221 the mass of the ground state of the _ negative - parity _ nucleon . from this fact
, the state we observe is concluded to be the @xmath223-@xmath224 scattering state with the relative momentum @xmath148 .
it may sound strange because the p - wave state of n and k with the relative momentum @xmath190 should be lighter than the @xmath223-@xmath224 scattering state with the relative momentum @xmath148 ; this lighter state is missing in our analysis
. this failure would be due to the wall - like operator @xmath225 .
the fact that the wall operator @xmath225 is constructed by the spatially spread quark fields @xmath226 with zero momentum may leads to the large overlaps with the nk scattering state with zero relative momentum .
the relation between operators and overlap coefficients is an interesting problem and is to be explored in detail for further studies .
anyway , the strong dependence on the choice of operators suggests that it is needed to try various types of operators before giving the final conclusion . before closing this section , we show the spectral weight @xmath227 obtained by the fit using the form @xmath228 in fig . [ weightfactor2 ] .
the spectral weight factor of the extracted state in @xmath219 channel with the hopping parameters @xmath197 is plotted against the lattice volume @xmath191 .
we note here that the @xmath229-like volume dependence _ is not _ always the characteristics of scattering states when we do not adopt point - point correlators as is the present case .
to conclude from this dependence , we need to determine the precise volume dependences of the weight factors in wall - point correlators . ]
although one sees the @xmath205-like volume dependence in fig .
[ weightfactor2 ] , one can conclude nothing only from this behavior unless the precise volume dependences of the weight factors in wall - point correlators are estimated .
we here mention the operator dependences in @xmath9 channel . as is seen in sec .
[ positive ] , the overlap factors with states strongly depend on the choice of operators .
we survey the effective masses of the five correlators ; @xmath230 , @xmath231 , @xmath232 , @xmath233 and @xmath234 .
the effective mass e(t ) is defined as a ratio between correlators with the temporal separation @xmath27 and @xmath235 , @xmath236 which can be expressed in terms of the eigenenergies and spectral weights as @xmath237 a plateau in e(t ) at @xmath238 implies the ground - state dominance in the correlator .
effective mass plots @xmath239 are often used to find the range where correlators show a single - exponential behavior ; the higher excited - state contributions @xmath240 are negligible in comparison with the ground - state component @xmath241 . here
, @xmath242 is an interpolation operator defined as @xmath243[u_g(x)c\gamma_5d_h(x)]c\bar s_c(x)\ ] ] which has a di - quark structure similar to that proposed by jaffe and wilczek @xcite , and is also used in refs .
@xcite .
the effective mass plots constructed from @xmath230 , @xmath231 , @xmath232 , @xmath233 and @xmath234 , in @xmath9 channel in @xmath5 lattice at @xmath82 employing the hopping parameters ( @xmath244)=(0.1600,0.1600 ) are plotted , along with the dashed line which denotes the lowest - state energy @xmath122 . ]
[ variouseffectivemass ] shows the effective mass plots constructed from @xmath230 , @xmath231 , @xmath232 , @xmath233 and @xmath234 .
one can see two typical behaviors in the figure .
one is the line damping from a large value to the energy @xmath122 of the lowest nk scattering state .
the other is the one arising upward to @xmath122 .
surprisingly , the differences of the spinor structure or the color structure among the operators are hardly reflected in the effective mass plots .
the difference is enough to perform the variational method but seems insufficient for a clear change of the effective mass plots .
instead , the effective mass plots seem sensitive to the spatial distribution of operators .
the upper three symbols are data using the point - point correlators and the lower two symbols are those from the wall - point correlators .
this means the overlap factor with each state is controlled mainly by the spatial distribution rather than the internal structure of operators , except for the overall constant .
the spatially smeared operators seem to have larger overlaps with the scattering state with the relative momentum @xmath148 .
( one can find that the overlap factor of the wall operator with the observed state in @xmath219 in fig .
[ weightfactor2 ] is 1000 times larger than those of point operators in fig .
[ weightfactor ] . )
one often expects that the overlap with a state could be enhanced using an operator whose spinor or color structures resembles the state .
we find however no such tendency in the present analysis .
the insensitivity to the spinor structures may come from the fact that the kn - type operator ( @xmath245 ) and the di - quark type operator ( @xmath242 ) are directly related by a factor of @xmath246 and a fierz rearrangement @xcite .
though we have no idea about the mechanism of the insensitivity to the color structure at present , this insensitivity would have some connection with the internal color - structure of @xmath11 .
the upper three data slowly damp and do not reach the lowest energy @xmath122 in this @xmath27 range , which can be explained in terms of the spectral weight . as is seen in fig .
[ weightfactor ] , @xmath210 is ten - times smaller than @xmath211 in the case of the point - point correlator
. then , the term @xmath247 in the effective mass survives at relatively large @xmath27 .
hence the effective mass needs larger @xmath27 to show a plateau at @xmath122 .
the insensitivity of the overlaps to the internal structure of operators could be helpful for us : we have adopted two operators whose color and spinor structures are different from each other . although the difference is enough in @xmath9 channel , it may be insufficient in @xmath219 channel , which leads to the failure in the diagonalization .
if we use operators with spatial distributions different from each other , it would be more effective in the diagonalization method . here
we comment on other works previously published , especially for the pioneering works by csikor _
et al . _
@xcite and sasaki @xcite .
the simulation condition for the former is rather similar to ours .
csikor _ et al .
_ first reported the possible pentaquark state slightly above the nk threshold in @xmath9 channel in @xcite . in ref .
@xcite , they tried chiral extrapolations and taking the continuum limit at the quenched level for the possible pentaquark state .
however they used the single - exponential fit analysis for the non - lowest state , namely the possible pentaquark state , for the main results .
it is difficult to justify their result unless the coupling of the operators to the lowest nk state is extremely small .
sasaki found a double - plateau in the effective mass plot and identified the 2nd - lowest plateau as the signal of @xmath11 .
unfortunately , we does not find a double - plateau in the present analysis .
the double - plateau - like behavior in effective mass plots can appear only under the extreme condition that @xmath211 is much larger than @xmath210 .
@xmath211 which is ten times larger than @xmath210 seen in fig .
[ weightfactor ] and the statistical fluctuations may cause the deviation of the effective mass plot from the single monotonous line .
in fact , the effective mass plot very slowly approaches @xmath122 as @xmath27 increases in fig .
[ variouseffectivemass ] .
he extracted the mass of the next - lowest state with a single and double exponential fits . the result
do not contradict with ours .
@xcite reports a lattice qcd study which adopted the overlap - fermions with the exact chiral symmetry .
the hybrid - boundary method was suggested in ref .
@xcite and the authors tried to single out the possible resonance state .
in these two studies , the absence of resonance states with a mass a few hundred mev above the nk threshold was concluded .
we have not found the resonance state which coincides just with the mass of @xmath11 in the chiral limit . in this sense , the results in refs .
@xcite are not inconsistent with ours .
we perform chiral extrapolations for kaon , nucleon , nk threshold ( a simple sum of a kaon mass and a nucleon mass ) and the 2nd - lowest state in the @xmath9 channel .
we adopt the lattice data with @xmath5 lattice , the largest lattice in our analysis .
one can find in fig .
[ negativegses ] that the 2nd - lowest state , which is expected to be a resonance state , is already affected by the finite volume effects for @xmath248 with the lightest combination of quarks , and we therefore adopt the largest - lattice data for safety .
we can expect from this fact that the typical diameter of this resonance is about 2 fm or longer and that it is desirable to use larger lattices than , @xmath249 for the analysis of @xmath11 . in fig .
[ chiralextrapolations ] , @xmath250 and @xmath123 obtained with each combination of quark masses for @xmath4 and @xmath5 lattices are plotted against @xmath251 .
we assume the linear function of quark masses , @xmath252 , for nucleon and the 2nd - lowest state with @xmath253 free parameters fitted using the five lattice data .
we determine the critical @xmath254 ( @xmath255 ) by @xmath256 and fix the @xmath257 so that the physical kaon mass is reproduced in the chiral limit , using the form for pseudo scalar mesons @xmath258 .
the chiral - extrapolated values of @xmath145 , @xmath144 , @xmath250 and @xmath123 for @xmath5 lattice are 0.4274(12 ) , 0.7996(60 ) , 1.227(6 ) and 1.500(52 ) in the lattice unit and 0.5001(14 ) , 0.9355(70 ) , 1.436(7 ) and 1.755(61 ) in the unit of gev , respectively .
we find that the results for @xmath4 lattice are consistent within errors as shown in fig .
[ chiralextrapolations ] .
the value of @xmath123=1755(61 ) mev in the chiral limit is significantly larger than the mass of @xmath10 in the real world .
how can we interpret this deviation ?
one possibility is the systematic errors from the discretization , the chiral extrapolation , or quenching .
another possibility is that the observed 2nd - lowest state might be a signal of a resonance state lying higher than @xmath11 .
unfortunately there is no clear explanation at this point . obviously more extensive studies on finer lattices with lighter quark masses in unquenched qcd are required .
however , we can at least conclude that _ our quenched lattice calculations suggest the existence of a resonance - like state slightly above the nk threshold for the parameter region we have investigated . _
a comparison of the chiral extrapolations in @xmath9 channel on @xmath4 ( left ) and @xmath5 ( right ) lattices for the nk threshold energy @xmath250 and the 2nd - lowest state energy @xmath123 . @xmath250 and @xmath123 are plotted against @xmath259 .
the filled circles ( triangles ) denote the energies of the nk threshold ( 2nd - lowest state ) in the chiral limit .
@xmath257 is fixed to be @xmath260 so that the physical kaon mass is reproduced in the chiral limit .
, title="fig : " ] a comparison of the chiral extrapolations in @xmath9 channel on @xmath4 ( left ) and @xmath5 ( right ) lattices for the nk threshold energy @xmath250 and the 2nd - lowest state energy @xmath123 . @xmath250 and
@xmath123 are plotted against @xmath259 .
the filled circles ( triangles ) denote the energies of the nk threshold ( 2nd - lowest state ) in the chiral limit .
@xmath257 is fixed to be @xmath260 so that the physical kaon mass is reproduced in the chiral limit .
, title="fig : " ]
we have performed the lattice qcd study of the @xmath261 states on @xmath2 , @xmath3 , @xmath4 and @xmath5 lattices at @xmath6=5.7 at the quenched level with the standard plaquette gauge action and wilson quark action . to avoid the possible contaminations originating from the ( anti)periodic boundary condition , which are peculiar to the pentaquark and
have not been properly noticed in previous studies , we have adopted the dirichlet boundary condition in the temporal direction for the quark field . with the aim to separate states clearly , we have adopted two independent operators with @xmath12 and @xmath262 so that we can construct a @xmath7 correlation matrix . from the correlation matrix of the operators ,
we have successfully obtained the energies of the lowest state and the 2nd - lowest state in the @xmath9 channel .
the volume dependence of the energies and spectral weight factors show that the 2nd - lowest state in this channel is likely to be a resonance state located slightly above the nk threshold and that the lowest state is the nk scattering state with the relative momentum @xmath148 . as for the @xmath219 channel ,
we have observed only one state in the present analysis , which is likely to be a @xmath263 scattering state of the ground state of the negative - parity nucleon @xmath223 and kaon with the relative momentum @xmath148 .
we have also investigated the overlaps using five independent operators . as a result
, we have found that the overlaps seem to be insensitive to the spinor and color structure of operators while the overlaps are mainly controlled by the spatial distributions of operators , at least for a few low - lying state in this analysis .
for the diagonalization method , it may be more effective to vary the spatial distributions rather than the internal structures .
the volume dependence of @xmath123 suggests that this resonance - like state in the @xmath9 channel is a rather spread object with the radius of about 1 fm or more .
the possibility of a resonance state lying in @xmath9 channel is desired to be confirmed by other theoretical studies , such as quark models , qcd sum rules , string models and so on @xcite .
unfortunately , four quarks _ uudd _ and one antiquark @xmath264 in @xmath265 state can hardly reproduce the unusually narrow width of @xmath11 so far , while the obtained mass in @xmath142 channel could be assigned to the observed resonance state @xcite .
hence , @xmath266 or @xmath267 states are favored to reproduce the width in terms of the quark model . however , there are many unknown problems left so far such as the internal structure of multi - quark hadrons @xcite or the dynamics of the string / flux - tubes @xcite .
the discovery of @xmath11 gives us many challenges in the hadron physics and more detailed theoretical study including the lattice qcd studies are awaited .
for further analyses , a variational analysis using the @xmath78 correlation matrix or larger matrices will be desirable .
the observation of wave functions will be also useful to distinguish a resonance state from scattering states and to investigate the internal structures of hadrons .
we can use the lattice qcd calculations in order to estimate the decay width @xcite and to study the flux - tube dynamics @xcite , which should give useful inputs for model calculations .
we acknowledge the yukawa institute for theoretical physics at kyoto university , where this work was initiated from the discussions during the yitp - w-03 - 21 workshop on `` multi - quark hadrons : four , five and more ? '' .
t. t. t. thanks dr .
f. x. lee for the useful advice .
t. u. and t. o. thank dr .
t. yamazaki for the fruitful discussion .
t. t. t. and t. u. were supported by the japan society for the promotion of science ( jsps ) for young scientists .
t. o. and t. k. are supported by grant - in - aid for scientific research from the ministry of education , culture , sports , science and technology of japan ( nos .
13135213,16028210 , 16540243 ) and ( nos .
14540263 ) , respectively .
this work is also partially supported by the 21st century for center of excellence program .
the lattice qcd monte carlo calculations have been performed on nec - sx5 at osaka university and on hitachi - sr8000 at kek .
after the completion of this paper , refs .
@xcite which also study the pentaquark state with lattice qcd have appeared on the preprint server .
0 t. nakano _ et al .
_ [ leps collaboration ] , phys . rev . lett . * 91 * , 012002 ( 2003 ) [ arxiv : hep - ex/0301020 ] .
v. v. barmin _ et al . _ [ diana collaboration ] , phys . atom .
nucl . * 66 * , 1715 ( 2003 ) [ yad .
fiz . * 66 * , 1763 ( 2003 ) ] [ arxiv : hep - ex/0304040 ] .
s. stepanyan _ et al . _ [ clas collaboration ] , phys .
lett . * 91 * , 252001 ( 2003 ) [ arxiv : hep - ex/0307018 ] .
j. barth _ et al .
_ [ saphir collaboration ] , phys .
b * 572 * , 127 ( 2003 ) .
s. chekanov _ et al . _
[ zeus collaboration ] , phys .
b * 591 * , 7 ( 2004 ) [ arxiv : hep - ex/0403051 ] .
a. e. asratyan , a. g. dolgolenko and m. a. kubantsev , phys .
nucl . * 67 * , 682 ( 2004 ) [ yad .
* 67 * , 704 ( 2004 ) ] [ arxiv : hep - ex/0309042 ] .
a. airapetian _ et al . _
[ hermes collaboration ] , phys .
b * 585 * , 213 ( 2004 ) [ arxiv : hep - ex/0312044 ] .
m. abdel - bary _ et al . _
[ cosy - tof collaboration ] , phys .
b * 595 * , 127 ( 2004 ) [ arxiv : hep - ex/0403011 ] . v. kubarovsky _ et al .
_ [ clas collaboration ] , phys .
* 92 * , 032001 ( 2004 ) [ erratum - ibid . *
92 * , 049902 ( 2004 ) ] [ arxiv : hep - ex/0311046 ] .
m. oka , prog .
. phys . * 112 * , 1 ( 2004 ) and references therein [ arxiv : hep - ph/0406211 ] .
r. l. jaffe , phys .
rept . * 409 * , 1 ( 2005 ) [ nucl .
suppl . * 142 * , 343 ( 2005 ) ] [ arxiv : hep - ph/0409065 ] .
d. diakonov , v. petrov and m. polyakov , z. phys .
* a 359 * , 305 ( 1997 ) .
m. praszalowicz , phys .
b * 575 * , 234 ( 2003 ) [ arxiv : hep - ph/0308114 ] .
stancu and d. o. riska , phys
* b575 * , 242 ( 2003 ) [ arxiv : hep - ph/0307010 ] ; + c. e. carlson , c.d .
carone , h.j .
kwee and v. nazaryan , phys . lett .
* b579 * , 52 ( 2004)[arxiv : hep - ph/0310038 ] ; + y. kanada - enyo , o. morimatsu and t. nishikawa , arxiv : hep - ph/0404144 .
+ s. takeuchi and k. shimizu , arxiv : hep - ph/0411016 .
a. hosaka , phys . lett . *
b571 * , 55 ( 2003 ) [ arxiv : hep - ph/0307232 ] .
j. sugiyama , t. doi and m. oka , phys .
b * 581 * , 167 ( 2004 ) [ arxiv : hep - ph/0309271 ] ; + t. nishikawa , y. kanada - enyo , o. morimatsu and y kondo , phys .
* d71 * , 016001 ( 2005)[arxiv : hep - ph/0410394 ] ; + r. d. matheus and s. narison , arxiv : hep - ph/0412063 .
for instance , a. ali khan _ et al . _
[ cp - pacs collaboration ] , phys .
d * 65 * , 054505 ( 2002 ) [ erratum - ibid .
d * 67 * , 059901 ( 2003 ) ] [ arxiv : hep - lat/0105015 ] .
f. csikor , z. fodor , s. d. katz and t. g. kovacs , jhep * 0311 * , 070 ( 2003 ) [ arxiv : hep - lat/0309090 ]
. s. sasaki , phys .
lett . *
93 * , 152001 ( 2004 ) [ arxiv : hep - lat/0310014 ] .
t. t. takahashi , t. umeda , t. onogi and t. kunihiro , arxiv : hep - lat/0410025 . t. w. chiu and t. h. hsieh , arxiv : hep - ph/0403020 . n. mathur _ et al .
_ , phys .
d * 70 * , 074508 ( 2004 ) [ arxiv : hep - ph/0406196 ] .
n. ishii , t. doi , h. iida , m. oka , f. okiharu and h. suganuma , phys .
d * 71 * , 034001 ( 2005 ) [ arxiv : hep - lat/0408030 ] .
n. mathur _ et al .
_ , phys .
b * 605 * , 137 ( 2005 ) [ arxiv : hep - ph/0306199 ] .
s. perantonis and c. michael , nucl .
b * 347 * , 854 ( 1990 ) .
m. luscher and u. wolff , nucl .
b * 339 * , 222 ( 1990 ) .
t. t. takahashi and h. suganuma , phys .
lett . * 90 * , 182001 ( 2003 ) [ arxiv : hep - lat/0210024 ] ; phys .
d * 70 * , 074506 ( 2004 ) and references therein [ arxiv : hep - lat/0409105 ] .
f. butler , h. chen , j. sexton , a. vaccarino and d. weingarten , nucl .
b * 430 * , 179 ( 1994 ) [ arxiv : hep - lat/9405003 ] .
t. yamazaki _ et al .
_ [ cp - pacs collaboration ] , phys .
d * 70 * , 074513 ( 2004 ) [ arxiv : hep - lat/0402025 ] .
m. fukugita , y. kuramashi , m. okawa , h. mino and a. ukawa , phys .
d * 52 * , 3003 ( 1995 ) [ arxiv : hep - lat/9501024 ] .
k. sasaki and s. sasaki , arxiv : hep - lat/0503026 .
o. dumbrajs , r. koch , h. pilkuhn , g. c. oades , h. behrens , j. j. de swart and p. kroll , nucl .
b * 216 * ( 1983 ) 277 .
m. m. nagels _ et al .
_ , nucl .
b * 147 * ( 1979 ) 189 .
m. l " uscher , nucl .
b * 354 * , 531 ( 1991 ) .
j. s. hyslop , r. a. arndt , l. d. roper and r. l. workman , phys .
d * 46 * , 961 ( 1992 ) .
r. l. jaffe and f. wilczek , phys .
lett . * 91 * , 232003 ( 2003 ) [ arxiv : hep - ph/0307341 ] . s. takeuchi and k. shimizu , arxiv : hep - ph/0411016 . t. hyodo and a. hosaka , arxiv : hep - ph/0502093 .
t. t. takahashi , h. matsufuru , y. nemoto and h. suganuma , phys .
lett . * 86 * , 18 ( 2001 ) [ arxiv : hep - lat/0006005 ] ; t. t. takahashi , h. suganuma , y. nemoto and h. matsufuru , phys .
d * 65 * , 114509 ( 2002 ) [ arxiv : hep - lat/0204011 ] . m.
bando , t. kugo , a. sugamoto and s. terunuma , prog .
phys . * 112 * , 325 ( 2004 ) [ arxiv : hep - ph/0405259 ] . c. alexandrou and g. koutsou , phys .
d * 71 * , 014504 ( 2005 ) [ arxiv : hep - lat/0407005 ] .
f. okiharu , h. suganuma and t. t. takahashi , arxiv : hep - lat/0407001 ; arxiv : hep - lat/0410021 ; arxiv : hep - lat/0412012 . g. m. de divitiis , l. del debbio , m. di pierro , j. m. flynn , c. michael and j. peisa [ ukqcd collaboration ] , jhep * 9810 * , 010 ( 1998 ) [ arxiv : hep - lat/9807032 ] .
j. m. cornwall , arxiv : hep - ph/0412201 .
b. g. lasscock _ et al .
_ , arxiv : hep - lat/0503008 . f. csikor , z. fodor , s. d. katz , t. g. kovacs and b. c. toth , arxiv : hep - lat/0503012 .
c. alexandrou and a. tsapalis , arxiv : hep - lat/0503013 .
( @xmath268,@xmath257)=(0.1650,0.1650 ) .[results ] the pion masses @xmath269 , kaon masses @xmath145 , nucleon masses @xmath144 , masses of the ground state of negative - parity nucleon @xmath221 , energies of the lowest state @xmath122 , energies of the 2nd - lowest state @xmath123 , energies of the lowest state @xmath155 ( obtained by single - exponential fit ) in the @xmath9 channel are listed .
the energies of the obtained state @xmath270 in the @xmath219 channel are also listed .
@xmath268 and @xmath257 are the hopping parameters for @xmath271 quarks and @xmath272 quark respectively . [
cols="^,^,^,^,^,^,^,^,^",options="header " , ]
in this appendix , we make another trial to estimate volume dependences of weight factors in @xmath9 channel . as seen in sec .
[ negative2 - 3 ] , we have extracted the weight factors using double - exponential fit , which is however rather unstable and we have therefore fixed the exponents .
we here discuss the possibility of methods without any multi - exponential fits .
let us again consider @xmath273 correlation matrices constructed by @xmath274-sink @xmath275-source and @xmath274-sink @xmath6-source correlators . here
@xmath276 denote the types of operators , such as `` point '' or `` wall '' or `` smear '' and so on .
the notations are the same as those in sec . [ formalism ] .
the @xmath277 and @xmath278 correlation matrices are described as @xmath279 with @xmath273 matrices ( @xmath280 ) and ( @xmath281 ) being possible higher excited - state contaminations .
we hereby consider two quantities ; @xmath282^t { \cal c}_{ij}^{\rm \gamma\alpha}(t)$ ] defined using one type of the correlation matrix and @xmath283^{-1}{\cal c}_{ij}^{\rm \gamma\beta}(t)$ ] , which with large @xmath27 lead to @xmath284^t { \cal c}_{ij}^{\rm \gamma\alpha}(t ) = c^{\gamma\dagger } c^\alpha + { \cal f}(d(t))+ ... .\ ] ] and @xmath285^{-1}{\cal c}_{ij}^{\rm \gamma\beta}(t ) = ( c^{\alpha})^{-1}c^\beta + { \cal f}'(d(t))+ ... , \ ] ] respectively . here
@xmath286 and @xmath287 are terms including @xmath273 diagonal matrix @xmath288 .
then , each component of @xmath282^t { \cal c}_{ij}^{\rm \gamma\alpha}(t)$ ] and @xmath283^{-1}{\cal c}_{ij}^{\rm \gamma\beta}(t)$ ] gets stable and shows a plateau in large @xmath27 region , where @xmath286 and @xmath287 are negligible . next , we relate these quantities to spectral weights . for this aim , we simply take the determinants . in the case when the correlation matrices are @xmath7 matrices , the determinant @xmath289 is explicitly written as @xmath290 , and the determinant @xmath291 is expressed as @xmath292 . the term @xmath293 ( @xmath294 )
denotes the product of the overlaps of the @xmath6(@xmath275)-type operator with the lowest state and the 2nd - lowest state . on the other hand
, @xmath295 ( @xmath296 ) corresponds to the spectral weight for the lowest ( 2nd - lowest ) state in the @xmath6-@xmath274 correlator in terms of a volume dependence .
let us consider the several cases when ( @xmath276)=\{w(wall ) , s(smeared ) , p(point)}. the term @xmath297 behaves showing the same volume dependence as the product of the spectral weights for the lowest and the 2nd - lowest state in the smeared - point correlator , which _ should be _ @xmath298 _
if _ the lowest state is a scattering state and the 2nd - lowest state is a resonance state .
the left panel in fig .
[ app1 ] represents @xmath299^t { \cal c}_{ij}^{\rm ps}(t)\right)$ ] on each volume .
however , @xmath299^t { \cal c}_{ij}^{\rm ps}(t)\right)$ ] on each volume , which approaches @xmath297 with large @xmath27 , has relatively large errors and fluctuations with no clear plateau and we fail to extract @xmath297 .
this would be due to the smallness of the signals in smeared - point correlators .
meanwhile , @xmath300^t { \cal c}_{ij}^{\rm pw}(t)\right)$ ] shown in the middle panel in fig . [ app1 ] and @xmath301^{-1}{\cal c}_{ij}^{\rm ps}(t)\right)$ ] shown in the right panel in fig .
[ app1 ] , which approach @xmath302 and @xmath303 respectively , show relatively clear plateaus .
therefore we extract @xmath302 and @xmath303 by the fits @xmath304^t
{ \cal c}_{ij}^{\rm pw}(t)\right)$ ] and @xmath305^{-1}{\cal c}_{ij}^{\rm ps}(t)\right)$ ] in the @xmath27 range where they show plateaus and we finally obtain @xmath297 as @xmath306 . in fig . [ app2 ] , we show @xmath297 obtained by the prescription shown above .
the solid line denotes the best - fit curve by @xmath307 and the dashed line does the best - fit curve by @xmath308 .
the best fit parameters are @xmath309 and @xmath310 , and @xmath311 is 1.72 and 7.13 respectively .
the volume dependence of @xmath297 seems not to be inconsistent with @xmath205 .
if we know the precise volume dependence of overlaps of wall operators , we may be discriminate the states using @xmath302 or @xmath303 . ^t { \cal c}_{ij}^{\rm ps}(t)\right)$ ] as the function of @xmath27 on each volume .
the middle figure is the plot of @xmath300^t { \cal c}_{ij}^{\rm pw}(t)\right)$ ] against @xmath27 .
@xmath312^{-1}{\cal c}_{ij}^{\rm ps}(t)\right)$ ] is plotted in the right figure . @xmath313^t { \cal c}_{ij}^{\rm ps}(t)\right)$ ] ,
@xmath300^t { \cal c}_{ij}^{\rm pw}(t)\right)$ ] and @xmath301^{-1}{\cal c}_{ij}^{\rm ps}(t)\right)$ ] show plateaus in the large @xmath27 region and coincide with @xmath297 , @xmath302 and @xmath303 respectively .
[ app1],title="fig : " ] ^t { \cal c}_{ij}^{\rm ps}(t)\right)$ ] as the function of @xmath27 on each volume .
the middle figure is the plot of @xmath300^t { \cal c}_{ij}^{\rm pw}(t)\right)$ ] against @xmath27 .
@xmath312^{-1}{\cal c}_{ij}^{\rm ps}(t)\right)$ ] is plotted in the right figure .
@xmath313^t { \cal c}_{ij}^{\rm ps}(t)\right)$ ] , @xmath300^t { \cal c}_{ij}^{\rm pw}(t)\right)$ ] and @xmath301^{-1}{\cal c}_{ij}^{\rm ps}(t)\right)$ ] show plateaus in the large @xmath27 region and coincide with @xmath297 , @xmath302 and @xmath303 respectively . [ app1],title="fig : " ]
^t { \cal c}_{ij}^{\rm ps}(t)\right)$ ] as the function of @xmath27 on each volume .
the middle figure is the plot of @xmath300^t { \cal c}_{ij}^{\rm pw}(t)\right)$ ] against @xmath27 .
@xmath312^{-1}{\cal c}_{ij}^{\rm ps}(t)\right)$ ] is plotted in the right figure . @xmath313^t { \cal c}_{ij}^{\rm ps}(t)\right)$ ] ,
@xmath300^t { \cal c}_{ij}^{\rm pw}(t)\right)$ ] and @xmath301^{-1}{\cal c}_{ij}^{\rm ps}(t)\right)$ ] show plateaus in the large @xmath27 region and coincide with @xmath297 , @xmath302 and @xmath303 respectively .
[ app1],title="fig : " ] is plotted as the function of the lattice volume @xmath191 .
the solid line denotes the best - fit curve by @xmath307 and the dashed line does the best - fit curve by @xmath308 .
the best fit parameters are @xmath309 and @xmath310 , and @xmath311 is 1.72 and 7.13 respectively .
these data behaves consistently in accordance with @xmath212 . | we study spin @xmath0 hadronic states in quenched lattice qcd to search for a possible @xmath1 pentaquark resonance .
simulations are carried out on @xmath2 , @xmath3 , @xmath4 and @xmath5 lattices at @xmath6=5.7 at the quenched level with the standard plaquette gauge action and the wilson quark action .
we adopt a dirichlet boundary condition in the time direction for the quark to circumvent the possible contaminations due to the ( anti)periodic boundary condition for the quark field , which are peculiar to the pentaquark . by diagonalizing the @xmath7 correlation matrices constructed from two independent operators with the quantum numbers @xmath8 ,
we successfully obtain the energies of the lowest state and the 2nd - lowest state in this channel .
the analysis of the volume dependence of the energies and spectral weight factors indicates that a resonance state is likely to exist slightly above the nk threshold in @xmath9 channel . |
kastan & alon @xcite identify and describe the configurations of relatively simple and small subnetworks that occur more frequently in biological networks than would be the case if the network were random .
they refer to these subnetworks as _
network motifs_. later , in his 2007 book on systems biology @xcite , alon makes the following comment _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ `` ideally , we would like to understand the dynamics of the entire network based on the dynamics of the individual building blocks . '' _
( alon @xcite . ) _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ the underlying premise behind this comment is that a modular , or engineering , approach to network dynamics is feasible .
identify building blocks , connect together to form networks and then describe dynamical properties of the resulting network in terms of the dynamics of its components .
this approach works well in linear systems theory , where a superposition principle holds , or in , for example , the study of synchronization in weakly coupled systems of nonlinear approximately identical oscillators where network dynamics can be closely related to the dynamics of individual nodes ( oscillators ) .
however , from the perspective of a mathematician or physicist , alon s comment seems unhelpful for the study of dynamics of heterogenous networks modelled by systems of nonlinear odes .
this is a well - known issue in complex systems @xcite . yet
engineers do couple gadgets together to make more complex systems and so it is natural to ask if it is possible to reconcile these viewpoints . in this note , we outline some of the basic ideas involved in an approach to network dynamics based on what we call _ asynchronous networks_. we allow for features seen in networks from technology , engineering , and biology ( for example , neuroscience or gene transcription networks ) .
network dynamics can involve a mix of distributed and decentralized control , adaptivity , event driven dynamics , switching , varying network topology and hybrid dynamics .
typically network dynamics will be piecewise smooth , time - irreversible , nodes may stop and later restart , and there may be no intrinsic global time .
significantly , many of these networks have a function : transportation networks bring people and goods from one point to another , neural networks may perform pattern recognition or computation , etc .
our goal is to address alon s comment in the context of functional asynchronous networks .
specifically , we describe a factorization of dynamics theorem where it is possible to describe the function of a large class of functional asynchronous networks in terms of the function of constituent subnetworks . as this article
is only intended to be an introduction , we work always with the simplest examples and models and omit technical details .
we refer the reader to @xcite for greater detail , generality and proofs .
we briefly describe some characteristic examples of asynchronous networks ( we refer to @xcite for more detail ) . in threaded or parallel computation , computation is broken into blocks or ` threads ' which are then computed _ independently _ of each other at a speed that depends on the clock rates of the individual processors .
as the computation proceeds , threads may need to exchange information with other threads .
this process involves stopping and synchronizing ( updating ) the thread states .
each thread may have to stop and wait for other threads to complete their computations before it can continue with its own computation .
each thread has its own clock ( determined by its associated processor ) .
if threads run on separate processors , threads will be unaware of the clock times of other threads except during the stopping and synchronization events .
we give a simple detailed example of a transport network in section [ sec : trans ] . in production networks , parts , compounds , etc .
are repeatedly built and combined as part of a production process leading to the desired item ( for example , a car or protein ) . in particular , there is variation in both connection structure , typically intermittent , and in the set of nodes ( nodes may be combined or decomposed ) .
a power grid consists of a network of various types of generators and loads connected by transmission lines .
a microgrid is a local network , capable of existing independently of the main power grid , and typically powered by renewable energy sources ( for example , solar or wind power ) .
critical questions involve the stability of the power grid in case of loss of a transmission line or generator ( variation in network structure ) , and control and stability issues related to combining and separation ( islanding ) of a large set of microgrids from the main power grid .
many mathematical models from engineering and biology incorporate thresholds . for networks ,
when a node attains a threshold , there are often changes ( addition , deletion , weights ) in connections to another nodes . for networks of neurons , reaching a threshold can result in a neuron firing ( spiking ) and short term connections to other neurons ( for transmission of the spike ) .
for learning mechanisms , such as spike - timing dependent plasticity ( stdp ) @xcite relative timings ( the order of firing ) are crucial @xcite and so each neuron , or connection between a pair of neurons , comes with a ` local clock ' that governs the adaptation in stdp . in general , networks with thresholds , spiking and adaptation provide characteristic examples of asynchronous networks where dynamics is piecewise smooth and hybrid a mix of continuous and discrete dynamics .
we summarize some of the characteristic features of asynchronous networks as revealed in the examples above . 1 . variable network structure and dependencies between nodes .
changes depend on the state of the system or are given by a stochastic process .
synchronization events associated with stopping or waiting states of nodes .
3 . order of events may depend on the initialization of the system or stochastic effects .
dynamics is only piecewise smooth and there may be a mix of continuous and discrete dynamics .
aspects involving function , adaptation and control . 6 .
evolution only defined for forward time
systems are not time reversible .
we use the notational conventions that @xmath0 and @xmath1 , @xmath2 . let @xmath3 .
assume given a network with @xmath4 nodes , @xmath5 .
suppose that @xmath6 has associated phase space @xmath7 , @xmath8 .
set @xmath9 the network phase space .
a vector field @xmath10 on @xmath11 is a _ network vector field_. stopping , waiting , and synchronization are a feature of asynchronous networks . if nodes are stopped or partially stopped , node dynamics will be constrained to subsets of node phase space .
we codify this situation by introducing a _ constraining node _
@xmath12 that , when connected to @xmath6 , implies that dynamics on @xmath6 is constrained .
set @xmath13 .
interactions between distinct nodes in the network are given by the network graph .
connections @xmath14 encode _ dependencies _ if @xmath15 , and _ constraints _ if @xmath16 , @xmath8 .
a _ connection structure _ @xmath17 is a directed network graph on the nodes @xmath18 such that for all @xmath8 , @xmath19 , @xmath20 , there is at most one directed connection @xmath14 .
we do not allow self - loops or connections to the constraining node @xmath12 .
@xmath17-admissible _ vector field @xmath21 is a network vector field with dependencies given by @xmath17 .
if @xmath22 , @xmath23 , then @xmath24 does not depend on @xmath25 ( and conversely , see @xcite ) . referring to figure [ fig:1 ] ,
suppose @xmath26 is @xmath17-admissible . for @xmath27
, we have @xmath28 @xmath29 where here we have assumed that the constraint @xmath30 results in @xmath31 being stopped .
a _ generalized connection structure _ @xmath32 is a ( nonempty ) set of connection structures on @xmath18 .
an @xmath32-structure @xmath33 is a set @xmath34 of network vector fields such that each @xmath35 is @xmath17-admissible .
suppose that @xmath32 is a generalized connection structure and @xmath33 is an @xmath32-structure .
interactions between nodes in asynchronous networks may be state dependent or change over time ( stochastically ) . here
we only consider state dependence .
we handle interactions and constraints using an _ event map _ @xmath36 .
the quadruple @xmath37 defines an _
asynchronous network_. dynamics on @xmath38 is given by the state dependent network vector field @xmath33 defined by @xmath39 we refer to @xcite for the definition of an integral curve for ( [ eq:1 ] ) and note that the definition we use is different from that used in filippov systems @xcite . under simple conditions on the event map @xcite
, it can be shown that if @xmath40 is compact then for each @xmath41 there is a unique piecewise smooth integral curve @xmath42 with initial condition @xmath43 and corresponding semiflow @xmath44 .
in general , @xmath45 will not be continuous as a function of @xmath46 .
we consider a single track railway line joining stations marked @xmath47 in figure [ fig:2 ] .
suppose there is a passing loop at @xmath48 .
trains , marked @xmath49 and @xmath50 in figure [ fig:2 ] , start from stations @xmath51 and @xmath52 respectively and proceed towards the opposite station .
there is no central control or communication between the train drivers except when both trains are in the passing loop .
we further assume that both drivers are running a nonlinear oscillator . when a train enters the passing loop it stops . when both trains are in the passing loop , the drivers cross couple their nonlinear oscillators .
in order for a train to leave the passing loop , two conditions must be satisfied . 1
. both trains must be in the passing loop .
2 . the two nonlinear oscillators must be phase synchronized to within @xmath53 where @xmath54 .
we model this setup using an asynchronous network with two nodes corresponding to the two trains .
the phase space for train @xmath55 will be @xmath56\times \mathbb{t}$ ] , @xmath57 ( @xmath58 ) , where the interval @xmath59 $ ] models the line joining @xmath51 to @xmath52 , @xmath51 has coordinate @xmath60 , @xmath52 has coordinate @xmath61 , the passing loop @xmath48 will be at @xmath62 , and @xmath63 will be the phase space for the nonlinear oscillator .
assume train motion given by @xmath64 where @xmath65 @xmath66 $ ] .
note that neither @xmath67 or @xmath68 can be zero anywhere on @xmath59 $ ] otherwise the trains will never both reach their destination stations in finite time .
we define four connection structures .
@xmath69 let @xmath70 be the associated generalized connection structure .
model the uncoupled oscillator dynamics for train @xmath55 by @xmath71 , where @xmath72 , and the coupled dynamics by the kuramoto phase oscillator system @xmath73 it remains to define the admissible vector fields and event map that give the required dynamics for this example . as admissible vector fields ( on @xmath74 \times \mathbb{t})^2 $ ] )
we take @xmath75 we define the event map by @xmath76 finally , dynamics are given by the network vector field @xmath77 we leave it as an easy exercise for the reader to check that if @xmath55 is initialized at @xmath78\times \mathbb{t}$ ] , @xmath79 , where @xmath80 , then ( [ eq:2 ] ) has a well defined integral curve @xmath81\times \mathbb{t})^2 $ ] , with specified initial condition , that gives the correct dynamics for the passing loop problem .
the passing loop gives an example of a simple _ functional _ asynchronous network .
the function is for the trains to go from one station to the opposite station in finite time .
observe that for this example there is the possibility of a _ dynamical deadlock _ : if the trains start at the same time and if @xmath82 , then the coupled phase oscillators will never phase synchronize
@xmath83 for all @xmath84 and so the trains will never exit the passing loop .
we refer to @xcite for more details on deadlocks in functional asynchronous networks .
we follow the notational conventions of section [ sec : async ] and let @xmath85 denote an asynchronous network .
we assume that @xmath38 has associated semiflow @xmath86 suppose that we are given _ initialization _ and _ termination _ sets @xmath87 where @xmath88 typically , @xmath89 will be closed disjoint hypersurfaces that separate @xmath7 into three connected components , @xmath8 .
that is , @xmath90 where @xmath91 and @xmath92 we call @xmath93 a _ functional asynchronous network_. the network function is getting from @xmath94 to @xmath95 and is expressed by the transition and timing functions @xmath96 that is , if @xmath97 , then for all @xmath8 there exists @xmath98 such that @xmath99 for the passing loop example discussed in the previous section , we take @xmath100 , @xmath101 . in this case
, there is the implicit assumption that trains stop when they reach their termination set .
equally well , we could take @xmath102 so that @xmath103\times\mathbb{t}$ ] etc .
( see also @xcite )
. finally , observe that @xmath104 .
more generally , we allow for general initialization times and define generalized transition and timing functions @xmath105 we refer to @xcite for details . for our main result , it is required that the network has a generalized transition and timing functions with @xmath106 . in figure [ fig:3 ]
we show a nine node functional asynchronous network that is built from the eight `` events '' @xmath107 .
the initialization and termination sets are indicated on the left and right sides of the figure respectively .
the events signify regions of phase space where there can be ( state dependent ) interaction between nodes .
for example , the event labelled @xmath108 involves interaction between nodes @xmath109 , @xmath110 , @xmath111 , and @xmath112 .
observe that there is only a partially ordered temporal structure on the events .
thus , the event @xmath108 must occur after @xmath113 but can occur before or after event @xmath114 . in figure [ fig:4 ]
we represent a basic building block with the same number of inputs and outputs .
the initialization sets are represented by the symbols @xmath115 , termination sets by @xmath116 .
interaction between nodes occurs only in the event region denoted by the rectangle . outside of the event region ,
nodes evolve independently .
more generally , we can allow for different number of inputs and outputs : nodes may merge or split .
our immediate aim is describe some basic operations that we can define on functional asynchronous networks that allow us enable us to find a ( maximal ) decomposition of a functional asynchronous network into the form shown in figure [ fig:3 ] .
if @xmath117 , @xmath118 , are functional asynchronous networks with distinct node sets ( @xmath119 , @xmath120 ) , define the _ product _ @xmath121 to be the functional asynchronous network @xmath93 , where @xmath122 and @xmath123 is defined in the obvious way to be the asynchronous network with node set @xmath124 ( we refer to @xcite for details ) .
we say the @xmath4-node functional asynchronous network @xmath93 is _ trivial _ if @xmath125 where each @xmath126 has exactly one node @xmath127 .
in particular , if @xmath128 is trivial there are no interactions between nodes and no constraints .
next let @xmath129 , @xmath118 , be a family of functional asynchronous networks with common initialization set , termination set and node set @xmath130 .
suppose that for each @xmath131 , there exists @xmath132 such that 1 .
@xmath133 where @xmath134 has node set @xmath135 and @xmath136 is trivial . 2 .
if @xmath137 , @xmath138 .
we define the _ amalgamation _ @xmath139 to be the functional asynchronous network @xmath140 , where @xmath141 is the trivial network defined as the product of the common trivial factors in @xmath136 , @xmath131 .
thus the node set of @xmath141 will be @xmath142 . referring to figure [ fig:5 ]
, we have @xmath143 and @xmath144 .
the amalgamation @xmath145 is trivial when restricted to nodes @xmath146 .
finally , we outline the operation of concatenation , referring the reader to @xcite for the details ( most ) we omit .
suppose that @xmath147 , @xmath148 , are functional asynchronous networks with common node set .
assume that @xmath149 .
the _ concatenation _
@xmath150 will be a temporal merging @xmath151 .
we define 1 .
@xmath152 , @xmath153 .
2 . @xmath154 , where @xmath155 denote the join of the graphs .
the definition of the set of admissible vector fields @xmath33 for @xmath38 is trickier and requires additional conditions on @xmath151 we refer to @xcite for details .
we define the event map by @xmath156 .
we refer to figure [ fig:6 ] for the operation of concatenation .
the concatenation @xmath157 has the important property that if @xmath126 has generalized transition and timing functions @xmath158 ,
@xmath159 , @xmath160 , then @xmath157 has generalized transition function @xmath161 given by @xmath162 ( * ? ? ?
* corollary 4.15 ) .
we have deliberately avoided listing the detailed properties required of functional asynchronous networks in order to define amalgamations and concatenations .
briefly , apart from requiring the existence of generalized transition and timing functions , we require ( 1 ) the uncoupled vector vectors defining intrinsic dynamics of a node @xmath6 to be transverse to @xmath163 , @xmath164 and ( 2 ) a local product structure on the network .
we refer to @xcite for the details .
a functional asynchronous network is _ primitive _ if it can not be written as a nontrivial amalgamation or concatenation .
[ fan ] under general conditions , a functional asynchronous network @xmath128 has a unique ( up to rearrangements ) decomposition @xmath165 where @xmath166 , @xmath167 , and and each @xmath168 is primitive .
the generalized transition function @xmath161 for @xmath128 can be expressed in terms of the generalized transition and timing functions @xmath169 of @xmath170 ( or @xmath171 for @xmath168 ) by : @xmath172 consider the network shown in figure [ fig:3 ] and assume that each event @xmath173 .
@xmath174 is primitive .
a decomposition satisfying the requirements of theorem [ fan ] is indicated in figure [ fig:7 ] the dashed lines indicate the initialization and termination sets for the subnetworks . the factorization for the network is @xmath175 this factorization corresponds to maximizing from the left hand side . however ,
if we maximize from the right we obtain the factorization @xmath176 in either case there is a concatenation of five networks that is the minimum number possible .
theorem [ fan ] allows us to write the function of a network explicitly in terms of the transition functions of the constituent subnetworks .
results of this type depend crucially on intermittent connection structure and nonsmooth dynamics .
for example , no such result is possible for a classical coupled network of phase oscillators .
the approach works because we have adopted an engineer s viewpoint : we emphasise function rather than dynamics .
indeed , we are indifferent to the specific dynamics occurring _ between _ the initialization and termination sets .
of course , both the timing and transition functions provide the key information about network function .
theorem [ fan ] is a prototypical theorem providing proof of concept .
the conditions for the theorem can be significantly weakened from those required in @xcite .
the theorem yields maximal feedforward structure on a functional asynchronous network ( note that individual events may have feedback loops ) .
3 . the result suggests the utility of starting with a small functional asynchronous network ; understanding the structure in depth and then then evolving to optimize function ( for example by adding feedback ) .
there are many as yet unexplored issues such as bifurcation , hidden deadlocks , and the effects of noise .
there is the problem of how far one can determine internal structure on the basis of input / output time series data
. 99 u alon . _ an introduction to systems biology .
design principles of biological circuits _
( chapman & hall / crc , boca raton , 2007 ) .
m di bernardo , c j budd , a r champneys , & p kowalczyk . _ piecewise - smooth dynamical systems _ ( springer , applied mathematical science 163 , london , 2008 ) .
c bick & m j field . `
asynchronous networks and event driven dynamics ' , ( 2016 ) , preprint .
c bick & m j field . ` asynchronous networks : modularization of dynamics theorem ' , ( 2016 ) , preprint .
n caporale & y dan . `
spike timing dependent plasticity : a hebbian learning rule ' _ annu .
neurosci . _ * 31 * ( 2008 ) , 2536 .
a f filippov . _ differential equations with discontinuous righthand sides _ ( kluwer academic publishers , 1988 ) .
w gerstner , r kempter , l j van hemmen & h wagner . ` a neuronal learning rule for sub - millisecond temporal coding ' , _ nature _ * 383 * ( 1996 ) , 7678 .
w gerstner & w kistler .
_ spiking neuron models _ ( cambridge university press , 2002 ) .
n kashtan & u alon . `
spontaneous evolution of modularity and network motifs ' , _ pnas _ * 102 * ( 39 ) ( 2005 ) , 1377313778 .
j ladyman , j lambert , and k wiesner .
` what is a complex system ? ' , _ eur .
j. for phil . of sc . _ * 3*(1 ) ( 2013 ) , 3367 .
a morrison , m diesmann & w gerstner . ` phenomenological models of synaptic plasticity based on spike timing ' , _ biol .
* 98 * ( 2008 ) , 459478 . | in this note we describe the theory of functional asynchronous networks and one of the main results , the modularization of dynamics theorem , which for a large class of functional asynchronous networks gives a factorization of dynamics in terms of constituent subnetworks . for these networks
we can give a complete description of the network function in terms of the function of the events comprising the network and thereby answer a question originally raised by alon in the context of biological networks . |
let @xmath0 be a real - valued lvy process , and let @xmath1 be the epochs of an independent poisson process with intensity @xmath2 ; add @xmath3 .
the probability law corresponding to @xmath4 started at @xmath5 will be denoted by @xmath6 ( with @xmath7 denoting the expectation ) . when @xmath5 is not mentioned
explicitly we assume that @xmath8 and write simply @xmath9 and @xmath10 .
define @xmath11 which we interpret as the first passage times under continuous and poisson observations , respectively .
observe that @xmath12 and , moreover , @xmath13 converges in probability to @xmath14 as @xmath15 ( the same is true for @xmath16 and @xmath17 ) .
thus exit theory under poisson observation can be regarded as a generalization of the classical exit theory . throughout this paper , however , we keep @xmath2 fixed .
observation at poisson epochs is both of theoretical and practical interest .
firstly , some exit problems with poisson observation yield transforms of certain occupation times , e.g. @xmath18.\ ] ] secondly , poisson observation is relevant in various applications such as queueing ( see e.g. @xcite ) , reliability and insurance risk theory ( see e.g. @xcite ) . in particular , in many applications discrete - time observation of stochastic processes would often be considered more natural , but for equidistant discrete time epochs the explicit and tractable analytical structure of continuous - time processes is typically destroyed , so that one is forced towards numerical techniques for the determination of exit probabilities and related quantities .
the poisson observation structure is a bridge between continuous - time and discrete - time observation that still leads to rather explicit , and as will be shown below , also somewhat elegant modifications of the continuous - time formulas . in order to stress the intuition behind the derivation of the identities , we will start with a simple case and
gradually generalize the setup .
most of the results are stated in terms of relations between transforms , but can also be understood as relations between the corresponding laws in an obvious way .
+ some of the wording throughout the manuscript will be in terms of the insurance application , where @xmath4 is the surplus process of a portfolio of insurance contracts , @xmath14 is the time of ruin of the portfolio , @xmath19 is the event of ( infinite - time ) survival , and @xmath13 is the time of observed ruin under poisson observation of the surplus process ( in the application the poisson epochs can for instance be interpreted as the observation times of the regulatory authority ) . + in section [ sec : exit1 ] we discuss survival probabilities corresponding to the two observation types , and then proceed to the general one - sided exit problems including the time of exit and the overshoot . in section [ sec : exit2 ]
we consider more complex problems .
firstly , the two - sided exit problem with one continously observed and one discretely observed boundary is related to the one where the observation types at the boundaries are interchanged .
secondly , we provide a link between poisson exit of a reflected process and continuous exit of the process reflected at poisson epochs .
we also show that a two - sided problem with poisson exit at both boundaries yields an identity as well , but with a non - standard first passage time . the latter quantity is then linked to a parisian ruin problem with erlang - distributed implementation delay .
finally , we establish a link between parisian ruin problems with continuous and poisson observations .
we conclude with section [ sec : spectral1 ] , where we specialize to the case of spectrally - one sided processes and demonstrate the use of our simple identities , providing simpler proofs and additional insight to some identities established in earlier literature .
the wiener - hopf factorization plays a crucial role in the derivations below .
define @xmath20\ } , & \underline g_t=\inf\{s\in[0,t]:x_s\wedge x_{s-}=\underline x_t\},\end{aligned}\ ] ] the infimum and its ( first ) time of occurrence up to horizon @xmath21 .
similarly , the supremum and its ( last ) time of occurrence are defined by @xmath22\ } , & \overline g_t=\sup\{s\in[0,t]:x_s\vee x_{s-}=\overline x_t\}.\end{aligned}\ ] ] finally , let the pairs @xmath23 and @xmath24 be distributed as @xmath25 and @xmath26 respectively ( under @xmath9 ) , and sampled independently of each other and of everything else ( @xmath27 and @xmath28 stand for ` down ' and ` up ' ) . recall that according to the wiener - hopf factorization we have @xmath29 see , e.g. , ( * ? ? ?
vi.5 ) , and @xcite for applications of factorization embeddings .
let us first consider @xmath30 which in the insurance application are the probabilities of survival with initial capital @xmath5 under continuous and poisson observation , respectively .
in fact , the two quantities are connected by two very simple relations : [ prop : survival ] for @xmath31 it holds that @xmath32 survival under poisson observation is determined by the sequence @xmath33 , whereas survival under continuous observation is determined by the sequence of infima in between the observation epochs ( black and grey dots in figure [ fig : wh_points ] , respectively ) .
let @xmath34 and define @xmath35 @xmath36 in the same way but for the shifted process @xmath37 and exponential time @xmath38 .
let @xmath39 and @xmath40 be the partial sum processes corresponding to @xmath41 respectively ; @xmath42 and @xmath43 are the heights of the black and grey dots in figure [ fig : wh_points ] .
observe that all @xmath44 and @xmath45 are independent , because of independence of increments and the wiener - hopf factorization .
since the @xmath45 s have the law of @xmath27 we obtain @xmath46 similarly , @xmath47 hence @xmath48 and , since @xmath31 , @xmath49 relation allows to interpret the transition from continuous to discrete poisson observation simply as a ( random ) increase of the starting value ( initial capital ) @xmath5 by @xmath28 , as far as the survival probability is concerned ; that is the structure of @xmath50 as a function of @xmath5 is otherwise completely preserved .
likewise , relation shows that moving from discrete poisson to continuous observation preserves the structure , reducing the initial capital by @xmath27 ( which has all its probability mass on the negative half - line ) .
suppose we modify the poisson observation model , so that there is no observation at time @xmath51 . then is still valid ( even for negative @xmath5 then ) , whereas does not hold any more . by the same token
one can connect the finite - time survival probabilities @xmath52 and @xmath53 for @xmath54 : @xmath55 that is , survival under continuous observation up to an independent erlang distributed time horizon is intimately related to survival under poisson observation up to a certain arrival epoch . [
prop : exit1 ] for @xmath31 and @xmath56 it holds that @xmath57{{\mathbb e}}e^{-{\alpha}t^d+{\beta}d},\nonumber\\ \label{eq : exit1b}&{{\mathbb e}}_{u}\left(e^{-{\alpha}\tau_0 ^ -+{\beta}x_{\tau_0 ^ -}};\tau_0 ^ -<\infty\right){{\mathbb e}}e^{-{\alpha}t^d+{\beta}d}\\&\qquad = { { \mathbb e}}\left [ e^{-{\alpha}t^d}{{\mathbb e}}_{u+d}\left(e^{-{\alpha}{\widehat \tau}_0 ^ -+{\beta}x_{{\widehat \tau}_0 ^
-}};{\widehat \tau}_0 ^ -<\infty\right)\right].\nonumber\end{aligned}\ ] ] observe that the left - hand side of gives the transform of the undershoot of the first grey point below 0 in figure [ fig : wh_points ] , according to the strong markov property applied at @xmath14 .
now one can establish the relation between black and grey points as in the proof of proposition [ prop : survival ] , additionally taking time into account .
essentially , we just shift the picture so that we start at the first grey point .
let @xmath58 and define @xmath59 in the same way but for the shifted process @xmath37 and exponential time @xmath38 .
as in the proof of proposition [ prop : survival ] we consider the sequences of black and grey dots in figure [ fig : wh_points ] , but now we also add the time component : @xmath60 and @xmath61 which are the partial sum processes corresponding to @xmath62 respectively . similarly to and we observe that @xmath63 letting @xmath64 and @xmath65 be the first passage epochs we can write @xmath66 where in the last line we applied the strong markov property at @xmath14 .
identity can be derived analogously .
in this section we consider two - sided exit problems with one continuous and one poisson exit at the boundaries .
it turns out that there is a simple relation between the problems when the roles of the continuous and the poisson exit are interchanged , i.e. problems corresponding to @xmath67 and @xmath68 .
here we extend the ideas of section [ sec : exit1 ] to their full potential .
[ prop : exit2 ] for @xmath69 and @xmath56 it holds that @xmath70{{\mathbb e}}e^{-{\alpha}t^d+{\beta}d},\nonumber\\ \label{eq : exit2b}&{{\mathbb e}}_{u}\left(e^{-{\alpha}\tau_0 ^ -+{\beta}x_{\tau_0 ^ -}};\tau_0 ^ -<{\widehat \tau}_a^+\right){{\mathbb e}}e^{-{\alpha}t^d+{\beta}d}\\&\qquad = { { \mathbb e}}\left[e^{-{\alpha}t^d}{{\mathbb e}}_{u+d}\left(e^{-{\alpha}{\widehat \tau}_0 ^ -+{\beta}x_{{\widehat \tau}_0 ^ -}};{\widehat \tau}_0 ^ -<\tau_a^+\right)\right].\nonumber\end{aligned}\ ] ] the proof is by inspection : for , consider the embeddings illustrated in figure [ fig : exit2 ] . in the left picture the grey dots correspond to the observations and the black to the suprema in between two observations . in the right picture
the black dots correspond to observations and the grey to the infima in between observations .
note that the position of a black point with respect to the previous grey point has the same distribution in both cases , namely @xmath24 .
the same is true for the position of the grey points with respect to their previous black points with @xmath23 .
so the patterns of points in each case have the same law up to a certain shifting ; we illustrate this by using the same patterns of points in both pictures in figure [ fig : exit2 ] and by drawing different sample paths .
now it follows that @xmath71 , see the left picture , must coincide with @xmath72 , see the right picture , because the interpretation of points was reversed. finally , we include the value of @xmath4 at first passage and its time using the strong markov property at @xmath14 as in the proof of proposition [ prop : exit1 ] . the same type of reasoning yields . by considering the negative of @xmath4
we immediately obtain the following result from proposition [ prop : exit2 ] .
[ cor : exit2 ] for @xmath69 and @xmath56 it holds that @xmath73{{\mathbb e}}e^{-{\alpha}t^u-{\beta}u},\nonumber\\ & { { \mathbb e}}_{u}\left(e^{-{\alpha}\tau_a^+-{\beta}(x_{\tau_a^+}-a)};\tau_a^+<{\widehat \tau}_0 ^ -\right){{\mathbb e}}e^{-{\alpha}t^u-{\beta}u}\\ & \qquad = { { \mathbb e}}\left[e^{-{\alpha}t^u}{{\mathbb e}}_{u+u}\left(e^{-{\alpha}{\widehat \tau}_a^+-{\beta}(x_{{\widehat \tau}_a^+}-a)};{\widehat \tau}_a^+<\tau_0 ^ -\right)\right].\nonumber\end{aligned}\ ] ] in this section we consider the process @xmath4 reflected at a barrier @xmath74 in a continuous and poisson manner , and study its first passage below 0 in poisson and continuous manner respectively ( with always opposite manners ) . again , these two problems are closely related .
note that in an insurance context reflection at @xmath75 results when paying out dividends according to a _ barrier strategy _ , either continuously or at poisson epochs ( see e.g. @xcite ) .
let @xmath76 be the law of @xmath4 started in @xmath5 and continuously reflected at @xmath75 and let @xmath77 be the corresponding regulator , i.e. @xmath78 under @xmath76 is @xmath79 similarly , let @xmath80 be the law of @xmath4 started at @xmath5 and reflected in poisson manner at @xmath75 , i.e. @xmath78 under @xmath80 is @xmath81 [ prop : reflected ] for @xmath69 and @xmath82 it holds that @xmath83{{\mathbb e}}e^{-{\alpha}t^d+\beta d},\nonumber\\ & { \widehat { { \mathbb e}}^a}_{u } \left ( e^{-{\alpha}\tau_0 ^ -+{\beta}x_{\tau_0 ^ -}-\gamma r_{\tau_0 ^ -}};\tau_0 ^ -<\infty\right){{\mathbb e}}e^{-{\alpha}t^d+\beta d}\\ & \qquad = { { \mathbb e}}\left[e^{-{\alpha}t^d}{{{\mathbb e}}^a}_{u+d } \left ( e^{-{\alpha}{\widehat \tau}_0 ^ -+{\beta}x_{{\widehat \tau}_0 ^ -}-\gamma r_{{\widehat \tau}_0 ^
-}};{\widehat \tau}_0 ^ -<\infty\right)\right].\nonumber\end{aligned}\ ] ] again , the proof follows merely by inspection in a similar way as for the previous results .
the first relation can be seen from figure [ fig : exit_ref ] , where the left picture depicts continuous reflection at @xmath75 and poisson observation at 0 , and the right picture depicts the corresponding ( shifted ) poisson reflection at @xmath75 and continuous observation at 0 . in the left picture
poisson observations yield the sequence : @xmath84 . in the right picture
the infima in between poisson reflection epochs are given by @xmath85 , where we choose @xmath86 to be distributed as @xmath87 .
these sequences can be complemented with the respective times as in the proof of proposition [ prop : exit1 ] .
finally , it is easy to see that @xmath77 enters the transforms without requiring any changes .
the second relation follows accordingly .
it is easy to see that proposition [ prop : reflected ] can be generalized from reflection to so - called refraction .
concretely , consider the processes @xmath88 and @xmath89 for @xmath90 $ ] . in the insurance context
such a refraction has the interpretation of taxation according to a loss - carry - forward scheme and tax rate @xmath91 , see e.g. @xcite .
the two - sided exit with poisson observation at both barriers can be related to a model with another type of exit time .
define the random time @xmath92 of the first observation such that the process has stayed above @xmath75 during the entire preceding inter - observation period , i.e. @xmath93 with @xmath94\}>a\}$ ] .
similarly , define @xmath95 with @xmath96\}<0\}$ ] as the first observation time such that the process has stayed below @xmath51 during the entire preceding inter - observation period .
then for @xmath97 $ ] it holds that @xmath98 to see this , one follows the same ideas as above : for the first equality consider infima in between two observations , see figure [ fig : exit2 ] , and for the second equality consider suprema in between two observations .
similarly , we also have the reverse identities : @xmath99 parisian ruin is defined as the first time when an excursion of @xmath4 below 0 is longer than some time @xmath100 ( sometimes referred to as _ implementation delay _ ) .
whereas the classical definition is in terms of a deterministic @xmath101 , for analytic tractability it is often assumed that @xmath101 is a random variable , and that an independent copy of @xmath101 is assigned to each excursion , see e.g. @xcite and @xcite .
firstly , from the memoryless property it follows that the time @xmath13 of poisson ruin is also the time of parisian ruin in the case where @xmath101 is an exponential random variable with rate @xmath102 . secondly , @xmath103 as defined in section [ 3.3 ] is the time of parisian ruin in the case where @xmath101 is erlang@xmath104 distributed ( since the latter is the sum of two independent exponential variables ) .
similarly to , equation [ eq : parisian1 ] can easily be extended to @xmath105{{\mathbb e}}e^{-{\alpha}t^d+{\beta}d},\nonumber\end{aligned}\ ] ] which under the present interpretation relates parisian ruin quantities with exponential and erlang(2)-distributed implementation delay ( here we took @xmath106 for simplicity ) . more generally ,
consider parisian ruin with erlang@xmath107 implementation delay and let @xmath108 denote the corresponding ruin time .
so in particular @xmath109 , @xmath110 and @xmath111 . on the other hand ,
define @xmath112 as the first epoch @xmath113 such that @xmath114 , i.e. the process has been observed negative at the last @xmath115 poisson epochs . then , along the same line of arguments , we can extend ( cf .
figure [ fig : paris ] ) : for @xmath116 and @xmath31 we have @xmath117{{\mathbb e}}e^{-{\alpha}t^d+{\beta}d}.\end{gathered}\ ] ] and @xmath118.\end{gathered}\ ] ]
if @xmath4 is a one - sided lvy process , some of the identities lead to more explicit forms , and this will allow to retrieve a number of results previously obtained in the literature , now with alternative proofs , revealing some more structure of the formulas . without loss of generality assume that @xmath4 is a spectrally - negative lvy process , i.e. it may only have negative jumps and it is not a non - increasing process . consider its laplace exponent @xmath119 and put @xmath120 for @xmath121 .
let us first recall some basic functions which play a fundamental role in exit theory , see e.g. ( * ? ? ?
let @xmath122 be the largest ( non - negative ) zero of @xmath123 , and let @xmath124 be the so - called scale function : a continuous non - negative function determined by its laplace transform @xmath125 .
in addition , we need a second scale function @xmath126 which can be rewritten as @xmath127 for @xmath128 , see also @xcite .
the two basic one - sided exit identities under continuous observation are @xmath129 and the wiener - hopf factors are given by @xmath130 see e.g. ( * ? ? ?
in order to apply formula of proposition [ prop : exit1 ] to , we first need the following identities : [ lem : wz ] for @xmath31 it holds that @xmath131 firstly , @xmath132 which can be checked by taking transforms and comparing to the wiener - hopf factor .
hence @xmath133 according to .
for the first identity it is left to note that @xmath134 .
using several times we can write for @xmath135 : @xmath136 plugging in @xmath137 and multiplying by @xmath138 we obtain the second identity . by analytic extension @xmath139 can be chosen arbitrarily .
assume that @xmath140 and consider the survival probabilities .
it is well known that @xmath141 , see also . according to proposition [ prop : survival ] and
lemma [ lem : wz ] we have @xmath142 which is corollary 1 of @xcite .
note that due to , for the spectrally negative lvy process the identity simplifies to the pleasant form @xmath143 where @xmath144 is an exponential random variable with parameter @xmath145 .
this for instance immediately explains why for a compound poisson process @xmath4 with exponential jump sizes the discrete poisson observation changes the classical ruin probability formula just by a multiplicative factor @xmath146 , where @xmath147 is the lundberg adjustment coefficient ( cf .
* eq.2.18 ) ) . using the standard identity , proposition [ prop : exit1 ] and lemma [ lem : wz ]
we obtain @xmath148 also , by considering proposition [ prop : exit1 ] for the negative of @xmath4 , see also corollary [ cor : exit2 ] , we arrive at @xmath149{{\mathbb e}}e^{-{\alpha}t^u-{\beta}u}\\ & = { { \mathbb e}}e^{-{\alpha}t^d-\phi_{\alpha}(a - u - d)}{{\mathbb e}}e^{-{\alpha}t^u-{\beta}u}=e^{-\phi_{\alpha}(a - u)}\frac{{\lambda}}{\phi_{\lambda}}\frac{\phi_{{\lambda}+{\alpha}}-\phi_{\alpha}}{{\lambda}-\psi_{\alpha}(\phi_{\alpha } ) } \frac{\phi_{\lambda}}{\phi_{{\lambda}+{\alpha}}+{\beta}}\\&=e^{-\phi_{\alpha}(a - u)}\frac{\phi_{{\lambda}+{\alpha}}-\phi_{\alpha}}{\phi_{{\lambda}+{\alpha}}+{\beta}},\end{aligned}\ ] ] because of and the fact that @xmath150 on @xmath16 .
these two identities were obtained in ( * ? ? ?
3.1 ) by virtue of a rather technical argument using the expression for the potential density of @xmath4 .
finally , we relate our results to previous literature on parisian ruin .
firstly , taking @xmath8 and @xmath151 in ( use @xmath152 ) one retrieves corollary 3.2 of @xcite , which is based on exponential implementation delay . furthermore , from the form of ( * ? ? ? *
eq.49 ) one can , after some lengthy calculations , obtain the following expression for erlang@xmath104 implementation delay : @xmath153 where the derivative is with respect to the subindex .
we can alternatively obtain directly using the results of this paper : and imply @xmath154{{\mathbb e}}e^{-{\alpha}t^d}\\ & = \frac{{\lambda}}{{\lambda}+{\alpha}}{{\mathbb e}}\left[e^{-{\alpha}t^u}\left(z_{\alpha}(u,0)-z_{\alpha}(u,\phi_{{\lambda}+{\alpha}})\frac{{\alpha}}{{\lambda}}\frac{\phi_{{\lambda}+{\alpha}}-\phi_{\alpha}}{\phi_{\alpha}}\right)\right]{{\mathbb e}}e^{-{\alpha}t^d}.\end{aligned}\ ] ] using lemma [ lem : wz ] and noting again that @xmath155 we get @xmath156 which together with the expression for the wiener - hopf factor @xmath157 readily yields .
the authors would like to thank ton dieker for stimulating discussions on the topic . | we consider exit problems for general lvy processes , where the first passage over a threshold is detected either immediately or at an epoch of an independent homogeneous poisson process .
it is shown that the two corresponding one - sided problems are related through a surprisingly simple identity .
moreover , we identify a simple link between two - sided exit problems with one continuous and one poisson exit . finally , poisson exit of a reflected process is connected to the continuous exit of a process reflected at poisson epochs , and a link between some parisian type exit problems is established . with the appropriate perspective , the proofs of all these relations turn out to be quite elementary . for
spectrally one - sided lvy processes this approach enables alternative proofs for a number of previously established identities , providing additional insight . |
with gpu hardware and the corresponding software environments becoming mature , compute clusters with gpu - accelerated nodes establish as a new , powerful platform for high - performance computing ( hpc ) .
mainly motivated by the expected boost for application performance ( i.e. reducing `` time to solution '' ) and also by energy - efficiency considerations ( i.e. reducing `` energy to solution '' ) , major research organizations and providers of hpc resources have already deployed an appreciable amount of gpu - accelerated resources worldwide @xcite .
moreover , gpu - like architectures are expected to play a major role in the upcoming exascale era @xcite .
it is well known , however , in the community , that the new hardware architecture together with the apparently disruptive programming models pose substantial challenges to scientific application developers ( e.g. @xcite ) .
while selected algorithms and applications have in fact been demonstrated to keep up with the shiny performance promises of gpus , in some cases even at the very large scale ( e.g. @xcite ) , it remains to be seen whether a broader class of scientific applications can take advantage of gpu - accelerated systems with reasonable programming effort and in a sustainable way .
often inappropriately termed `` legacy applications '' in this context , leading scientific hpc codes are typically being actively developed , comprise many tens or hundreds of thousands of lines of code achieved by a team effort of many dozens of person years , they provide state - of - the - art functionality , as well as high optimization , parallel scalability and portability .
the codes gene and vertex , which have been developed in the max - planck - society with continuous support from its high - performance computing centre ( rzg ) may serve as prototypical examples in this respect .
but as a matter of fact , such highly tuned codes are often reaching the limits of ( strong ) scalability .
for example , due to increasing inter - node communication times , or , as in the case of vertex , due to the lack of conventionally exploitable parallelism in the code structure , the time to solution for a given setup can no more be efficiently reduced by utilizing more cpu resources .
thus , a significantly increased node performance due to accelerators appears as a promising route towards further boosting application performances at scale .
although both , gene and vertex are written in fortran ( with mpi and hybrid mpi / openmp parallelization , respectively ) we decided to adopt the c - based cuda programming model because it is the performance reference for nvidia gpus . while we found the commercial cuda - fortran language to deliver competitive performance on the gpu
, the employed pgi compiler falls behind the intel compiler ( which is our reference for the cpu ) on the remaining cpu parts which marginalizes the overall application speedups of the heterogeneous code . for the same reason we did not yet make productive use of the openacc programming model @xcite .
the performance baseline for all comparisons is defined by a highly optimized , parallel cpu implementation of the respective algorithms . rather than quoting single - core speedups ( which , in our opinion is hardly meaningful in most cases )
our comparisons are always based on the same number of gpu cards and _ multicore _ cpus ( `` sockets '' ) .
specifically , we compare the run time obtained on a certain number of nodes , each equipped with two intel xeon e5 - 2670 8-core cpus and two nvidia k20x gpus , with the run time measured with the original , parallel cpu code on the same number of nodes ( without gpus ) .
gene @xcite is a massively parallel code for the simulation of plasma turbulence in fusion devices .
the code solves the time - dependent , five - dimensional vlasov - maxwell system of equations on a fixed phase - space grid . depending on the physical problem
, typical gene simulations take between a few days and many weeks , using thousands of cores on x86_64-based hpc systems .
gene is open - source @xcite and has a world - wide user base .
the gene algorithm employs coordinates aligned to the magnetic field lines in a fusion device like a tokamak . in this paper
we use the so called _ x - global _ version , where all physical quantities are handled in a spectral representation with respect to the @xmath0 coordinate , which is the second of the three space dimensions @xmath1 ( radial ) , @xmath0 ( binormal ) and @xmath2 ( along the field line ) .
the remaining phase - space coordinates are ( in this order ) the velocity along the field line @xmath3 and the magnetic moment @xmath4 ( see @xcite for details ) . although gene is able to handle any number of ion species and the electrons in the framework of gyrokinetics , we use for this paper only a single ion species , neutralized by electrons . for all performance comparisons a problem setup with a number of @xmath5 grid points
is used .
the starting point of this work was a profiling of the gene code ( svn revision 3440 ) , with the times given in table [ tab : profiling_snb ] .
[ cols= " < , < " , ] table [ vertex : tab1]a shows that the run time of vertex on the cpu is dominated by solving the radiative transfer equations ( item ` transport ` ) , and in particular for computing neutrino absorption and emission rates ( item ` rates ` ) .
[ vertex : fig1]a identifies the positions of the individual routines in the execution flow .
about 50 percent of the run time is spent in the computation of one particular interaction rate ( named ` rate kernel ` , ` c2 ` ) .
the different interaction rates are often termed `` local physics '' , which expresses the fact that the computations are to a high degree independent of each other and provide a data parallelism on the grid level .
different interaction processes ( ` rates ` ) can be computed independently of each other , which implies additional , coarse - grained parallelism on the function level ( see blowup in fig .
[ vertex : fig1]a ) . in the following the algorithm for offloading the ` rate kernel ` to the gpu is outlined . due to its dominance in the code ,
high data parallelism and arithmetic intensity the suitability for the gpu shall become immediately apparent . as input for the computations
a few one - dimensional arrays are needed , which represent the local thermodynamic conditions for which the interaction kernel is evaluated .
all operations are performed on a five - dimensional grid representing discretized phase space .
the size of this grid varies with the resolution , in a typical setup the total number of grid points is about @xmath6 .
for the major part of the kernel , computations on each grid zone can be done independently of the others , which leads to a high degree of data parallelism ( up to @xmath7 threads ) .
only after all grid zones are processed , a reduction ( corresponding to a phase space integral ) to a three dimensional grid is performed .
this can be still done in parallel , but with much less parallelism ( @xmath8 threads ) .
all computations are done twice for subsets of different input data , accounting for two possible reaction channels .
the actual implementation of the part ` c2 ` is straightforward : the data is copied asynchronously to the gpu and the five - fold nested loops of the cpu version are separated in kernel calls with about 100000 threads .
the kernels are scheduled in streams , in order to allow the cuda run time to overlap kernel executions corresponding to the twofold computation of the processes .
the problem is compute bound , as data transfer is negligible ( 0.9 ms ) compared to gpu computations ( 40 ms ) and at least 140 double - precision floating - point operations are executed per transferred byte . for
good performance results it turned out to be crucial to use shared memory for the input data and to use as much registers as possible on the device . after tuning our cuda code with the help of the nvidia profiler
, we achieve an occupancy of 93% of the theoretical upper limit for the most important kernels .
however , we still encounter about 10% of branch - divergence overhead and 25% of global memory replay overhead . work is still ongoing to improve on the latter performance metrics . as mentioned above , the different sub - steps ` c1 ` to ` cn ` ( see fig . [
vertex : fig1]b ) are independent of each other and can be computed in any order within one openmp thread , or ray. in the original code , however , the order across different openmp threads is always the same , e.g. when a thread computes sub - step ` c1 ` , also the other threads work on the same sub - step .
an overlap of computations on the cores and the gpu was thus achieved by : a ) individually shuffling the computations of the sub - steps ` c1 ` to ` cn ` on each ray , and b ) ensuring that the sub - step ` c2 ` from each ray to the gpu is offloaded in a queue ( see fig .
[ vertex : fig1]b ) . in an ideal situation
where all steps ` c1 ` to ` cn ` take the same amount of execution time , work on the cpus and the gpu would be perfectly overlapped . in reality , a balancing of about 80% could be reached .
the rate kernel ` c2 ` requires 2.16 s on one cpu thread ( cf .
tab [ vertex : tab1 ] ) and scales almost perfectly with openmp .
the same kernel can be computed on the gpu in 0.04 s. thus , with one gpu , speedups of 7 or 54 are achieved when comparing with one cpu socket or a single core , respectively .
this demonstrates that a significant speedup was achieved with respect to a sandy bridge cpu . as the coarse grained openmp parallelization of vertex ( which is crucial for achieving its excellent weak scalability )
does not allow to use the threaded rate kernel on the cpu , the acceleration factor of 54 applies for production applications which effectively eliminates the rate kernel from the computing time budget and in practice accounts for a twofold acceleration ( corresponding to the original 50% share of the rate kernel , cf .
tab [ vertex : tab1 ] ) of the entire application .
with the specific cases of gene and vertex we have shown that complex hpc applications can successfully be ported to heterogeneous cpu - gpu clusters .
besides writing fast gpu code , exploiting and balancing both the gpu _ and _ the cpu resources of the heterogeneous compute nodes turned out to be an essential prerequisite for achieving good overall `` speedups '' , which we define as the ratio of the run time obtained on a number of gpu - accelerated sockets and the run time measured with parallel code on the same number of cpu sockets .
in the case of vertex we have demonstrated twofold speedups which hold for production applications on gpu - clusters with many hundreds of nodes . in particular , the excellent weak scalability of vertex @xcite is not affected by the additional acceleration due to gpus . threefold speedups appear in reach but would require at least additional porting of a linear solver for a block - tridiagonal system .
limitations in the software environment ( lack of device - callable lapack functionality ) have so far impeded a successful port of this part of the algorithm .
importantly , due to the specific code structure of vertex , such speedups would not have been possible with comparable programming effort by simply using more cpu cores .
the performance of gene is currently limited by the data transfer between the host cpu and the gpu as we have shown by an elaborate performance - modeling analysis .
after this bottleneck will have relaxed by upcoming hardware improvements ( pcie 3 ) further optimization efforts on the gpu code will increase the overall speedups on a heterogeneous cluster .
the question whether the effort of several person - months , which we have invested for each code , and which we consider typical for such projects , is well justified can not be answered straightforwardly . for complex , and `` living '' scientific hpc codes , for which gene and vertex can serve as prototypical examples , achieving up to threefold speedups in overall application performance appears very competitive @xcite . also from the point of view of hardware investment
( buying gpus instead of cpus ) and operational costs ( `` energy to solution '' ) the migration of applications from pure cpu machines to gpu - accelerated clusters can be considered cost - effective if speedups of at least about two are achieved . on the other hand , while very valuable for increasing simulation throughput , twofold or threefold application speedups usually do not enable qualitatively new science objectives .
for this reason we sometimes observe reluctance in the scientific community to invest significant human resources for achieving gpu - performance improvements in this range .
this is further exacerbated by legitimate concerns about sustainability , maintainability and portability of gpu - kernel code .
these are no serious issues for gene and vertex , where the parts we have ported to the gpu are not under heavy algorithmic development and were also carefully encapsulated by us . in general , however , the _ need _ for kernel programming , which is considered as a pain by many , currently appears as the largest hurdle for a broader adoption of gpu programming in the scientific hpc community . moreover
, it may turn out necessary to port significant parts of the application code to the gpu , e.g. in cases like gene where the data transfers become a limiting factor , or even to completely reimplement the application .
these concerns could be mitigated by the establishment of a high - level , directive based programming model , based e.g. on the openacc standard @xcite or a future revision of openmp @xcite , together with appropriate compiler support .
also intel s xeon phi many - core coprocessor with its less disruptive programming model appears very prospective in this respect . despite serious efforts ,
however , we were not yet successful with gene or vertex to achieve performances on this platform which are competitive with the gpu .
we attribute this mostly to a comparably lower maturity of the xeon phi software stack and we expect improvements with upcoming versions of the compiler and the openmp run time
. most importantly , today s gpus ( and many - core coprocessors ) might provide a first glimpse on the architecture and the related programming challenges of future hpc architectures of the exascale era @xcite .
applications need to be prepared _ in time _ for the massive simt and simd parallelism which is expected to become prevalent in such systems . even on contemporary multicore cpus with comparably moderate thread - counts and simd width , the experience we have gained with porting gene and vertex has already led to appreciable performance improvements of the cpu codes .
we thank f. jenko and h .- th .
janka for encouraging the development of gpu versions for gene and vertex , respectively .
nvidia corp . and intel corp .
are acknowledged for providing hardware samples and technical consulting .
buras , r. , janka , h .- th . , rampp , m. , kifonidis , k. _ two - dimensional hydrodynamic core - collapse supernova simulations with spectral neutrino transport . i. numerical method and results for a 15 m@xmath9 star_. astronomy & astrophysics * 447
* , 1049 ( 2006 ) .
cardall , c. , endeve , e. , budiardja , r. d. , marronetti , p. , mezzacappa , a. _ towards the core - collapse supernova explosion mechanism_. advances in computational astrophysics : methods , tools , and outcome .
asp conference proceedings , * 453 * , 81 ( 2012 ) .
shimokawabe , t. , aoki , t. et al .
_ peta - scale phase - field simulation for dendritic solidification on the tsubame 2.0 supercomputer_. proceedings of 2011 international conference for high performance computing , networking , storage and analysis , sc 11 , acm new york , ny , us ( 2011 ) 3:13:11 | we have developed gpu versions for two major high - performance - computing ( hpc ) applications originating from two different scientific domains .
gene @xcite is a plasma microturbulence code which is employed for simulations of nuclear fusion plasmas .
vertex @xcite is a neutrino - radiation hydrodynamics code for `` first principles''-simulations of core - collapse supernova explosions @xcite .
the codes are considered state of the art in their respective scientific domains , both concerning their scientific scope and functionality as well as the achievable compute performance , in particular parallel scalability on all relevant hpc platforms .
gene and vertex were ported by us to hpc cluster architectures with two nvidia _ kepler _ gpus mounted in each node in addition to two intel xeon cpus of the _ sandy bridge _ family .
on such platforms we achieve up to twofold gains in the overall application performance in the sense of a reduction of the time to solution for a given setup with respect to a pure cpu cluster .
the paper describes our basic porting strategies and benchmarking methodology , and details the main algorithmic and technical challenges we faced on the new , heterogeneous architecture . , and gpu , hpc application , gene , vertex |
quantum entanglement is a key resource for quantum information processing ( qip ) protocols [ 1 - 4 ] .
information processing involving multi - particle states requires entangled channels which can process the information from one remote location to another with reliability .
experimental realization of multi - particle systems and the detection of all orthogonal basis states forming a complete set of entangled states remains a challenge [ 5 - 9 ] , nevertheless , efficient theoretical construction and characterization of different multi - particle entangled channels for analyzing different information protocols is an important precursor to successful design of experiments .
quantum teleportation involving many particles has been studied theoretically using different multi - particle entangled systems [ 10 - 22 ] .
many experiments have also been performed which provide partial experimental support to this concept [ 23 - 28 ] .
information processing protocols such as dense coding deal with sending classical information using an entangled quantum state as a shared resource [ 29 - 33 ] .
quantum information processing techniques through nuclear magnetic resonance have been considered in detail elsewhere [ 34 - 42 ] . in this article
, we propose generalized multi - particle entangled systems for improving the efficiency of information processing .
we do this by proposing particle correlations , as a direct measure of entanglement , using standard ursell - mayer terms which are firmly founded on the principles of many body statistical mechanics [ 43 - 47 ] .
the approach presented here can be expanded and is applicable to statistical ensembles , and therefore , to electrons and other spin-1/2 systems as well as photons [ 48 - 50 ] .
statistical correlation coefficients are shown to be useful in distinguishing entangled systems belonging to different families .
the properties of correlation coefficients are used to determine whether the states under study are related through local transformations or not . in section 3
, we propose and discuss the properties of a five - particle entangled channel and generalize the quantum channel for @xmath0 number of particles .
the quantum channel proposed in that section is used for various information processing protocols successfully .
this is followed by a conclusion .
in this section , we first review the entanglement properties of a few maximally entangled states used in the past by others and then propose multi - particle genuinely entangled states for use in information processing . a criterion is used to define the extent of correlation between particles and several examples of entangled states of many particles are considered .
the entanglement properties of bipartite states and a few multi - partite states have been studied extensively [ 51 - 60 ] . however , the same for the multi - particle states is not well established . here
, the extent of entanglement is assessed by the well established statistical mechanical formula for correlation coefficients [ 43 - 47 ] .
correlation measures for multi - particle systems defined using ursell - mayer type cluster coefficients are suggested by us as a means for generalizing the defining of degree of entanglement between many particles .
correlation coefficients for the three - particle systems are represented as @xmath24 they are components of a third rank tensor .
the non - zero correlation coefficients @xmath25 for the three - particle greenberger - horne - zeilinger ( ghz ) states [ 5 ] , given by @xmath26_{123 } & , & \left| \psi \right\rangle_{123}^{(3),(4 ) } = \frac{1}{\sqrt{2 } } \left [ \ , \left| 001 \right\rangle \pm \left| 110 \right\rangle \ , \right]_{123 } , \ \nonumber \\
\left| \psi \right\rangle_{123}^{(5),(6 ) } = \frac{1}{\sqrt{2 } } \left [ \ , \left| 010 \right\rangle \pm \left| 101 \right\rangle \ , \right]_{123 } & { \rm and } & \left| \psi \right\rangle_{234}^{(7),(8 ) } = \frac{1}{\sqrt{2 } } \left [ \ , \left|011 \right\rangle \pm \left| 100 \right\rangle \ , \right]_{123 } \end{aligned}\ ] ] are either @xmath27 or @xmath28 for the coefficients @xmath29 , @xmath30 , @xmath31 , @xmath32 .
the values suggest that the correlations between three particles are genuine and quantum .
the three - particle ghz states , though maximally entangled , are not robust with respect to disposal of any of the particles i.e. tracing of any of the particles results in the disappearance of quantum correlation between the rest of the particles .
the other popular three - particle entangled state is w state [ 62 ] , given by @xmath33_{123}\ ] ] have the value @xmath34 for the non - zero correlation coefficients @xmath35 , @xmath36 , @xmath37 , @xmath38 , @xmath39 , @xmath40 , @xmath41 which suggests that the correlation between three particles is less than the maximum .
the state is robust with respect to tracing of any of the particles .
a similar calculation of correlation coefficients for a set of states such as @xmath42 shows that these states are maximally entangled as well @xmath43 , @xmath30 , @xmath38 , @xmath44 @xmath45 .
in angular momentum algebraic parlance states represented in eq .
( 7 ) and ghz states refer to different coupling schemes and can be locally transformed into each other .
the entanglement properties of these states are similar to the ghz states if we consider the extent of correlation between three particles .
thus , if the value of correlation coefficients associated with a particular system is maximum then it indicates that the state in question possesses genuine multi - particle quantum correlations and is maximally entangled . however , if the value is not maximum but more than one non - zero correlation coefficients exists the state is non - maximally entangled . for a direct product state
all the correlation coefficients are zero suggesting no genuine multi - particle correlation between the particles . the criteria to measure the degree of entanglement using statistical correlations is compared with the existing criteria s such as concurrence [ 52,53 ] ( for two - particle systems ) and with 3-tangle for three - particle maximally entangled ghz states and average value of square of the concurrence for less than maximally entangled w state [ 54 , 62 ] .
concurrence for a two - particle system is defined as @xmath46 where @xmath47 and @xmath48 is complex conjugate of @xmath49 . above expression
shows that the value of concurrence is equal to one of the coefficient @xmath9 of second rank symmetric traceless tensor representing the correlation between the particles .
table 1 summarizes the comparison between the value of concurrence and correlation coefficients obtained for bell states .
table 1 [ cols="^,^,^,^,^ " , ] table 3 @xmath30 & @xmath31 & @xmath44 & @xmath36 & @xmath38 + [ 2.0 ex ] @xmath50_{123}$ ] & 1 & @xmath51 & @xmath52 & @xmath52 & @xmath52 & - & - + [ 2.0 ex ] @xmath53_{123}$ ] & 1 & @xmath51 & @xmath52 & @xmath51 & @xmath51 & - & - + [ 2.0 ex ] @xmath54_{123}$ ] & 1 & @xmath51 & @xmath51 & @xmath52 & @xmath51 & - & - + [ 2.0 ex ] @xmath55_{123}$ ] & 1 & @xmath51 & @xmath51 & @xmath51 & @xmath52 & - & - + [ 2.0 ex ] @xmath56 & 1 & - & -1 & - & -1 & -1 & + 1 + [ 2.0 ex ] @xmath57 & 1 & - & -1 & - & -1 & + 1 & -1 + [ 2.0 ex ] @xmath58 & 1 & - & + 1 & - & + 1 & + 1 & -1 + [ 2.0 ex ] @xmath59 & 1 & - & + 1 & - & + 1 & -1 & + 1 + [ 2.0 ex ] @xmath60 & 1 & - & + 1 & - & -1 & + 1 & + 1 + [ 2.0 ex ] @xmath61 & 1 & - & + 1 & - & -1 & -1 & -1 + [ 2.0 ex ] @xmath62 & 1 & - & -1 & - & + 1 & -1 & -1 + [ 2.0 ex ] @xmath63 & 1 & - & -1 & - & + 1 & + 1 & + 1 + [ 2.0 ex ] table 3 and table 2 show that the value of non - zero correlation coefficients for three - particle ghz state(s ) and three - particle @xmath64 are in excellent argument with the value of 3-tangle whereas average value of square of the concurrence for @xmath65 is also a match with the value of non - zero correlation coefficients obtained .
this suggests that the criterion using statistical correlation coefficients to measure the degree of entanglement include all possible type of entanglement in multi - particle systems and is a noble idea to study and analyze the properties of multi - particle systems .
this can thus be generalized for arbitrary number of particles .
the expression for four - particle correlation coefficients is given by @xmath66- \left\langle \sigma ^2_{\beta } \right\rangle\left[c_{\alpha\gamma\delta}^{134}\right]- \left\langle \sigma ^3_{\gamma } \right\rangle\left[c_{\alpha\beta\delta}^{124}\right ] \nonumber \\ & - & \left\langle \sigma ^4_{\delta } \right\rangle\left[c_{\alpha\beta\gamma}^{123}\right ] - \left \langle \sigma^1_{\alpha } \sigma^2_{\beta } \right \rangle \left \langle \sigma^3_{\gamma } \sigma^4_{\delta } \right \rangle - \left \langle \sigma^1_{\alpha } \sigma^3_{\gamma } \right \rangle \left \langle \sigma^2_{\beta } \sigma^4_{\delta } \right \rangle \nonumber \\ & - & \left \langle \sigma^1_{\alpha } \sigma^4_{\delta } \right \rangle \left \langle \sigma^2_{\beta } \sigma^3_{\gamma } \right \rangle + 2 \left \langle \sigma^1_{\alpha } \right \rangle \left \langle \sigma^2_{\beta } \right \rangle \left \langle \sigma^3_{\gamma } \right \rangle \left \langle \sigma^4_{\delta } \right \rangle \end{aligned}\ ] ] the non - zero correlation coefficients calculated for the four - particle ghz states , namely @xmath67\ ] ] where if @xmath68 then @xmath69 and _ vice versa _ are @xmath70 , @xmath71 , @xmath72 , @xmath73 , @xmath74 , @xmath75 , @xmath76 , @xmath77 , @xmath78 and indicate that four - particle ghz states possess maximum correlations .
similarly , the non - zero correlation coefficients calculated for the four - particle w state , @xmath79_{1234}$ ] , are @xmath80 , @xmath81 , @xmath82 , @xmath83 , @xmath84 , @xmath85 , @xmath86 , @xmath87 , @xmath88 , @xmath89 , @xmath90 , @xmath91 and @xmath78 and show the value as @xmath92 indicating less than maximum correlations between particles .
rigolin [ 17 ] proposed a generalized bell basis as a set of four - particle states to be used for information processing , however , all the 16 four - particle correlation coefficients associated with the generalized bell basis are zero suggesting that there is no genuine correlation between the four - particles .
yeo and chua [ 20 ] proposed a four - particle entangled system @xmath93 ; the maximum value of non - zero correlation coefficients @xmath73 , @xmath82 , @xmath89 and @xmath78 indicates that the state is maximally and genuinely entangled .
we consider here three sets of four - particle maximally entangled states , in addition to ghz states , given by @xmath94 , @xmath95 and @xmath96 where @xmath97 , \end{aligned}\ ] ] @xmath98 , \end{aligned}\ ] ] and @xmath99.\end{aligned}\ ] ] the non - zero correlation coefficients calculated for the above three sets are @xmath100 , @xmath73 , @xmath74 and @xmath101 , @xmath102 , @xmath103 , @xmath104 , @xmath105 and @xmath106 , @xmath107 , @xmath108 , @xmath109 , respectively and indicate maximum entanglement .
the set of states represented by eq .
( 12 ) and eq .
( 14 ) are cluster type of states [ 63 ] and can be transformed into each other through local transformations whereas eq .
( 13 ) represents @xmath110 type of states [ 20 ] .
the expression for the five - particle correlation coefficient is given in the appendix a. the generalized five - particle ghz states are represented as @xmath111\ ] ] and are maximally correlated as shown by non - zero correlation coefficients @xmath112 , @xmath113 , @xmath114 , @xmath115 , @xmath116 , @xmath117 , @xmath118 , @xmath119 , @xmath120 , @xmath121 , @xmath122 , @xmath123 , @xmath124 , @xmath125 , @xmath126 and @xmath127 . unlike the ghz state , the generalized five - particle w state , @xmath128_{12345}$ ] , is not maximally correlated as shown by non - zero correlation coefficients @xmath129 , @xmath130 , @xmath131 , @xmath132 , @xmath133 , @xmath134 , @xmath135 , @xmath136 , @xmath137 , @xmath138 , @xmath139 , @xmath140 , @xmath141 , @xmath142 , @xmath143 , @xmath144 , @xmath145 , @xmath146 , @xmath147 , @xmath148 and @xmath149 .
other five - particle entangled systems to be considered are two sets of basis states given as @xmath150 and @xmath151 , where @xmath152\end{aligned}\ ] ] and @xmath153 .
\end{aligned}\ ] ] @xmath154 are three - particle ghz states and are given by eq .
the non - zero correlation coefficients for the two sets are @xmath155 , @xmath156 , @xmath157 , @xmath158 and @xmath159 , @xmath160 , @xmath161 , @xmath162 , respectively and show maximum value .
the extent of correlation between five - particles remains the same even after interchanging the particle indices .
the general expression for the @xmath163-particle correlation coefficient can be obtained by solving the equations for cluster functions derived formally from the @xmath163-th quantum virial coefficient .
the following summarizes the relation between correlation coefficients and the degree of entanglement .
+ ( i ) existence of maximum values for more than one correlation coefficient for a system under study , indicates that the state of the system possesses genuine and maximum entanglement .
+ ( ii ) for non - maximally entangled states the value of correlation coefficients lies between @xmath164 and @xmath165 .
+ ( iii ) null results for all correlation coefficients of state suggest that it is a direct product of fewer particle states and there exists no genuine multi - particle entanglement .
+ ( iv ) the value of non - zero correlation coefficients remains the same for states connected to each other by local unitary transformations .
+ ( v ) the extent of correlation remains invariant to changing the particle indices .
the criterion to use cluster coefficients as a measure of entanglement of the state under study allows one to characterize the extent of correlation of multi - particle states on the same scale irrespective of number of particles involved .
a consistent description emerges for systems irrespective of the number of particles which are entangled . in this subsection
, we discuss some of the properties of correlation coefficients in addition to those described in previous section . 1 .
the relation between correlation coefficients of states which differ from each other only through permutation of particle indices can be seen immediately as follows : + ( 1 ) the state @xmath166_{1234}$ ] [ eq .
( 12 ) ] is obtained from @xmath167_{1234}$ ] by permuting particles 1 and 2 .
hence , the non - zero correlation coefficients associated with @xmath168 and @xmath169 are @xmath100 , @xmath73 , @xmath74 , @xmath101 and @xmath100 , @xmath72 , @xmath75 , @xmath101 , respectively .
+ ( 2 ) conversely , by examining two sets of equal number of correlation coefficients , we can also relate the states .
for example , the two sets @xmath170 , @xmath156 , @xmath157 , @xmath171 and @xmath172 , @xmath173 , @xmath174 , @xmath175 are related to each other through particle permutations @xmath176 and @xmath177 .
the first set is the only non - null set of coefficients for the state given by eq .
hence another can be obtained by particle permutations .
thus , a family of states can be quickly enumerated .
if the number of non - zero correlation coefficients corresponding to two entangled sets are not equal , then they belong to two different family of states .
+ ( 1 ) three - particle ghz state @xmath29 , @xmath30 , @xmath31 , @xmath32 and three - particle w state @xmath35 , @xmath36 , @xmath37 , @xmath38 , @xmath39 , @xmath40 , @xmath41 show four and seven non - zero correlation coefficients , respectively .
they belong to two different families of states .
+ ( 2 ) four - particle maximally entangled ghz states @xmath178 , @xmath71 , @xmath72 , @xmath73 , @xmath74 , @xmath75 , @xmath76 , @xmath77 , @xmath179 represented by eq . ( 11 ) and four - particle maximally entangled set represented by eq .
( 12 ) @xmath100 , @xmath74 , @xmath73 , @xmath101 show nine and four non - zero correlation coefficients , respectively which indicates that these two sets belong to different families of states .
+ ( 3 ) the set of five - particle states represented by eq . ( 17 ) @xmath180 , @xmath160 , @xmath161 , @xmath181 and five - particle brown state [ 16 ] @xmath182 , @xmath183 , @xmath184 , @xmath185 , + @xmath186 , @xmath187 possess four and six non - zero correlation coefficients , respectively and hence belong to two different families of entangled systems .
3 . even if the number of correlation coefficients associated with different entangled states are equal the states need not have to belong to the same family . for example , the three states @xmath166_{1234}$ ] , @xmath188_{1234}$ ] and @xmath189_{1234}$ ] belong to maximally entangled four - particle sets represented by eq .
( 12 ) , eq .
( 13 ) and eq .
( 14 ) , respectively and possess four non - zero correlation coefficients .
although @xmath168 and @xmath190 belong to same family , @xmath191 belongs to different family of states .
the extent of correlation between particles remains invariant under standard local unitary transformations .
the set of states @xmath192 represented by eq .
( 7 ) can be obtained by applying hadamard operation to the second particle of three - particle ghz states ( eq .
( 5 ) ) and possesses the same degree of correlation as that of ghz states .
doing a hadamard operation would not affect the y - component of the correlation coefficient but would convert x - component to z - component and _ vice versa_. thus , the non - zero correlation coefficients for the set of states represented by eq .
( 2 ) and eq . ( 3 ) ( which differ from each other by a hadamard transformation on particle 2 )
are @xmath8 , @xmath9 , @xmath10 and @xmath13 , @xmath9 , @xmath14 , respectively .
this is a trivial example .
however , as the number of particles in an entangled set increases , the number of ways of doing transformations also increases and hence this scheme is useful for nontrivial , multiple local transformations as shown below .
+ ( 1 ) the correlation coefficients , @xmath193 , @xmath82 , @xmath89 , @xmath179 and @xmath194 , @xmath195 , @xmath196 , @xmath197 , correspond to the entangled states @xmath198_{1234}$ ] and @xmath199_{1234}$ ] , respectively .
the set of coefficients can be transformed into each other by doing hadamard transformations on @xmath17 , @xmath200 and @xmath201 particles and permuting particles 1 and 2 .
+ ( 2 ) the five - particle maximally entangled states , namely @xmath202_{12345}$ ] and @xmath203_{12345}$ ] can be converted into one another by doing local transformations as revealed by their correlation coefficients , namely @xmath204 , @xmath114 , @xmath116 , @xmath119 , @xmath132 , @xmath121 , @xmath122 , @xmath124 , @xmath127 , @xmath205 and @xmath206 , @xmath207 , @xmath208 , @xmath119 , @xmath112 , @xmath209 , @xmath210 , @xmath184 , @xmath127 , @xmath211 , respectively .
thus by doing three hadamard operations on particle two , three and four @xmath212 can be locally transformed to @xmath213 .
in this section we propose a maximally and genuinely entangled five - particle state and describe different information processing protocols using the state . in the past , multi - particle entangled channels involving odd number of particles have been proposed _ with the use of a controller _ to assist the sender for successful and optimal information transfer [ 10 , 13 , 14 ] .
we show that one can _ eliminate the intermediate observer _ controlling the process such that information processing is successful in all the measurement outcomes performed by the sender .
the formation of the state proposed here ensures efficient information transfer between two or more users in the communication protocol .
the five - particle maximally entangled set proposed here is given by @xmath214 \nonumber \\ & & \end{aligned}\ ] ] where @xmath95 are given by @xmath215 the set represented above is same as four - particle entangled set given in eq .
( 13 ) , however , the order in which the states are represented is different .
the set proposed here shows values @xmath51 for the non - zero correlation coefficients @xmath216 , @xmath217 , @xmath218 , @xmath219 .
depending on the discussions of previous section and due to the absence of y in the @xmath220 s the five - particle entangled set proposed above belongs to a different family of states with respect to those represented by eq .
( 16 ) and eq .
( 17 ) . in order to communicate an arbitrary two - particle information to bob i.e. @xmath221_{12}$ ]
, alice must share any one of the five - particle entangled state @xmath222 given by eq . (
18 ) with bob such that particles 3 , 4 and 5 are with alice and particles 6 and 7 are with bob .
thus , using @xmath223 as the quantum channel shared between alice and bob where @xmath224_{34567}$ ] , alice can communicate her unknown message with bob by interacting her particles 1 and 2 with her share of entangled particles 3 , 4 and 5 so that @xmath225 eq .
( 20 ) can be re - expressed in form of alice s projection basis given by eq .
( 17 ) as @xmath226 where @xmath227 and @xmath228 . for alice s measurement outcomes @xmath229 and @xmath230 , bob s particles are instantaneously projected on to the state alice wanted to communicate , with total probability of 1/16 , however , for all other measurement outcomes of alice , bob require _ only _ single qubit transformations to recover the message successfully .
the preparation of above set of states and teleportation of an arbitrary two - particle state are represented in fig .
( 1 ) and fig . ( 2 ) , respectively . for controlled teleportation
, the quantum state @xmath223 is shared between alice , charlie and bob such that the particles 3 and 4 are with alice , particles 5 and 6 are with bob and particle 7 is with charlie .
alice projects her four particles on to the basis set given by eq .
( 13 ) so that eq .
( 20 ) becomes @xmath231 where @xmath232 and @xmath233 .
for example , if alice s measurement outcome is @xmath234 , the combined state of bob s and charlie s particles is given by @xmath235 \left|0\right\rangle_{7 } \nonumber \\ & + & \frac{1}{\sqrt{2}}\left [ -a\left|10\right\rangle_{56}-b\left|11\right\rangle_{56}+c\left|00\right\rangle_{56}+d\left|01\right\rangle_{56 } \right]\left|1\right\rangle_{7}.\end{aligned}\ ] ] for charlie s outcome of @xmath236 , bob s particles are in the state identical to the one communicated by alice , however , for his outcome @xmath237 , bob needs to do a @xmath238 and @xmath239 operation on the @xmath240-particle to complete the process successfully .
again , for all the outcomes of alice and charlie , bob can recover the message with single qubit unitary transformations , if needed .
the above processes can be generalized for the case of @xmath0 number of particles as follows .
the generalized entangled basis set corresponding to eq .
( 18 ) is @xmath241\end{aligned}\ ] ] where @xmath242 s are ordered in the same way as in eq .
( 19 ) and + @xmath242 s are @xmath243-particle generalization of eq .
( 13 ) , namely @xmath244.\end{aligned}\ ] ] this can be used for teleportation of @xmath163-particle arbitrary information from alice to bob .
the @xmath0-particle channel is shared between alice and bob such that the first @xmath245 particles are with alice and the rest @xmath163 particles are with bob .
it is important to choose the correct projection basis such that the teleportation becomes feasible in all outcomes with only single qubit operations on bob s end . for controlled teleportation
the same basis set is shared between alice , charlie and bob such that the first @xmath163 particles are with alice , @xmath0-th particle is with charlie and rest @xmath163 particle are with bob . to realize successful teleportation alice projects her particle s on the @xmath243-particle generalized basis set given by eq .
( 25 ) . to control
the process effectively charlie measures his @xmath0-th particle in computational basis .
dense coding is concerned with the transfer of two bits of classical message between two parties by sending only one qubit ( particles ) from sender to receiver provided they share a maximally entangled pair of two qubits . in this subsection
we discuss dense coding protocol using maximally entangled state proposed in this article .
alice and bob must share a maximally entangled quantum channel given by @xmath246 such that the qubits 1 , 2 , and 3 are with alice and qubits 4 and 5 are with bob .
alice can locally operate her qubits to encode the desired message , using operators from the set @xmath247 , @xmath248 and @xmath249 corresponding to her three qubits 1 , 2 and 3 , respectively .
alice sends her encoded qubits to bob who decodes the message by an appropriate measurement on the joint five qubit state . however , for the five - qubit quantum channel @xmath250 , alice will only produce 32 orthogonal states and hence can encode a 5-bit message .
the operator set that is used to prepare all the orthogonal states belonging to the entangled set is given by @xmath251 , @xmath252 , @xmath253 , @xmath254 , @xmath255 , @xmath256 , @xmath257 , @xmath258 , @xmath259 , @xmath260 , @xmath261 , @xmath262 , @xmath263 , @xmath264 , @xmath265 , @xmath266 , @xmath267 , @xmath268 , @xmath269 , @xmath270 , @xmath271 , @xmath272 , @xmath273 , @xmath274 , @xmath275 , @xmath276 , @xmath277 , @xmath278 , @xmath279 , @xmath280 , @xmath281 , @xmath282 $ ] .
the capacity of dense coding channel [ 64 , 65 ] is @xmath283 where @xmath284 is the dimension of alice s system , @xmath285 is reduced density matrix of bob s system with respect to alice s system and @xmath286 is the von neumann entropy . for the entangled channel @xmath246 , @xmath287 , @xmath288 , and @xmath289 which shows @xmath290 and thus maximizes the channel capacity .
it has been shown earlier that by using a maximally entangled five - qubit ghz state or a generalized five - qubit w state as a quantum channel , alice can only send 4-bit information .
quantum dense coding can also be realized involving a controller between alice and bob . by doing so ,
we compare the transfer efficiency of the channel proposed by us to that of others discussed in section 2 .
we do this by using three different states of maximally entangled five - qubits @xmath291_{12345}.\end{aligned}\ ] ] @xmath292_{12345}. \end{aligned}\ ] ] @xmath293_{12345}\end{aligned}\ ] ] given by eq .
( 16 ) , eq .
( 17 ) and eq .
( 18 ) , respectively .
the quantum channel @xmath294 is shared by three users alice ( 1 and 2 ) , charlie ( 3 ) and bob ( 4 and 5 ) .
charlie performs a von neumann measurement on his share of qubit in the new basis @xmath295 such that @xmath296 where @xmath297 if @xmath298 . for @xmath299 in the range @xmath300 ,
replace @xmath299 in eq .
[ 27 ] by @xmath301 . in this basis
, @xmath294 is given by @xmath302_{1245 } \nonumber \\ & + & \frac{\left|x_{2}\right\rangle_{3}}{2}\left [ \sin{\theta}\left|0000\right\rangle-\cos{\theta}\left|1001\right\rangle-\cos{\theta}\left|0110\right\rangle-\sin{\theta}\left|1111\right\rangle \right]_{1245}.\end{aligned}\ ] ] in the special case of @xmath303 , the four qubits are in maximally entangled four - qubit state and alice can send 4 bits of message to bob by first encoding the message and then sending 2 qubits of hers to him .
however , charlie can do measurement for any @xmath299 .
+ ( a ) if charlie s result is @xmath304 and he communicates his measurement result to alice then the dense coding process is given schematically in fig .
alice introduces an auxiliary qubit @xmath305 and does a joint unitary operation @xmath306 on her qubits 1 , 2 and auxiliary qubit ( in the computational basis for qubits 1 , 2 and 3 ) .
the unitary operation is @xmath307 if charlie s measurement outcome is @xmath308 , the unitary operation that alice will use is @xmath309 where @xmath310 if alice makes a measurement on auxiliary qubit and gets the measurement result @xmath305 ( with the probability @xmath311 ) , she knows that the joint state of four qubits is in a maximally entangled four - qubit state ; however , the measurement result @xmath312 ( with the probability @xmath313 ) , will confirm that the four - qubits are in ghz state .
in the first case , alice can encode her message using 16 binary operators from the set @xmath314 and @xmath315 corresponding to her two qubits 1 and 2 .
she sends her two qubits to bob who decodes the message by doing a joint measurement on four - qubits based on the entangled state alice has prepared and thus decodes the original message .
however , doing a joint measurement on four - qubits to discriminate 16 orthogonal states is experimentally challenging . hence the following : + ( i ) bob does a joint unitary operation on alice s @xmath316 qubit and his @xmath240 qubit given as @xmath317 which evolves the four - qubit state(s ) into a four - qubit ghz state(s ) , measures the joint state in the four - qubit ghz basis and recovers the encoded message .
+ ( ii ) bob applies two c - not operations on the state of four qubits keeping alice s qubits 1 and 2 as controls and his qubits 5 and 4 as targets , respectively which leads to a state of direct products of four qubits ; alice s two qubits are in an entangled state and bob s two qubits are in a computational basis state . by doing these operations bob ensures that he differentiates between the subsets of alice s operations , i.e. if he measures his two qubits in computational basis states and finds @xmath318 , he knows that the operation alice has used to encode the message belongs to the subset @xmath319 . similarly other measurements @xmath320 , @xmath321 and @xmath322 belong to the subsets @xmath323 , @xmath324 and @xmath325 , respectively .
bob uses bell basis to measure alice s qubits ( which are just a local transformation away from bell states ) and decodes the 4-bit classical message with relative ease . in the other case ,
where alice s measurement result yields @xmath312 , she can still send 3-bit classical message to bob .
therefore on the average @xmath326 -bit classical message is transferred where the suffix c - a denotes that charlie informs his measurement result to alice only .
+ ( b ) if charlie sends his measurement result to bob and not to alice , then the process for sending the information is pictorially represented in fig .
alice does local operations on her qubits and sends them to bob who , can not however , do a joint measurement to discriminate all the state of four - qubits as they may or may not be orthogonal .
we take alice s first operational subset @xmath319 , as an example , such that the four - qubit state immediately after alice s operation is @xmath327_{1245}$ ] and where -ve signs are always in odd numbers . after receiving her qubits
bob does two c - not operations as described in the case ( a ) .
alice s qubits are partially entangled , whereas bob s qubits are in computational basis state such that @xmath328_{12 } \otimes \left|00\right\rangle_{45}$ ] . by measuring his qubits in the computational basis
, bob gets alice s operational subset on her two qubits , thus he needs to differentiate between four states of two - qubit system related to each operational subset .
for this bob introduces an auxiliary qubit in state @xmath305 and performs a joint unitary operation given by eq .
( 29 ) on three qubits ( 1 , 2 and aux ) in the computational basis of @xmath329 , such that @xmath330_{12}\otimes \left|0\right\rangle_{aux } \nonumber \\ & + & \cos{\theta}.\sqrt{1-\frac{\sin^2{\theta}}{\cos^2{\theta}}}.\frac{1}{\sqrt{2}}\left [ \left|00\right\rangle \mp \left|11\right\rangle \right]_{12}\otimes \left|1\right\rangle_{aux}.\end{aligned}\ ] ] the average information transfer from alice to bob will then be @xmath331 where the suffix c - b denotes that charlie informs his measurement result to bob only .
it is clear that in both the cases [ ( a ) and ( b ) ] the amount of information transfer , on average , between alice and bob is the same .
however , in the case of a five - qubit ghz state the amount of information transferred in controlled manner is @xmath332 .
a similar calculation for information transfer between alice and bob using the quantum channel proposed , @xmath250 , shows that irrespective of the measurement basis used by charlie , alice is always able to send 4-bit information to bob using her 2-qubits . even if charlie does not inform alice about the measurement basis used , she is able to send maximum information to bob using her two qubits such that the information transfer between alice and bob is independent of the value of analyzer angle @xmath299 .
at the same time , the information transfer between alice and bob using @xmath333 is the same as one obtains in the case of @xmath334 .
it is seen that all the five - qubit states used for controlled dense coding possess maximum correlation between the particles , however , the amount of information transfer using different quantum channels is not the same .
the entangled set proposed here is advantageous in terms of average information transfer between the sender and the receiver as compared to other five - qubit entangled sets .
thus the representation of a quantum channel used in a communication network and distribution of qubits between different users is an important factor in information processing .
a graphical comparison is made in fig .
( 5 ) to compare the efficiency of genuinely entangled five - qubit states discussed here and the five - qubit ghz states in terms of average information transferred during the process and the analyzer angle @xmath299 . for generalized dense coding the @xmath0-qubit quantum channel
is shared between alice and bob such that @xmath245 qubits @xmath335 to @xmath336 are with alice and rest of the qubits are with bob . by locally manipulating her qubits , alice encodes her message and sends her qubits to bob who , in turn , does the required measurements involving all the qubits and decodes the message . in principle alice
can prepare @xmath337 orthogonal basis states and hence @xmath337 distinguishable messages for bob .
thus by using the generalized entangled channels alice can send @xmath0-bit information to bob .
alternatively , the @xmath0-qubit state is shared between the three users in communication protocol such that @xmath163 qubits @xmath335 to @xmath338 are with alice , one qubit is with charlie [ @xmath245-th ] and the rest @xmath163 qubits are with bob @xmath339 .
charlie measures his qubit in the basis given by eq .
( 27 ) and either sends his measurement results to alice or to bob .
there are two instances : + ( a ) if charlie sends his measurement results to alice , she introduces an auxiliary qubit and does a combined unitary transformation on her @xmath163 qubits and the auxiliary qubit .
after this she measures the state of auxiliary qubit , encodes her message and sends her qubits to bob who does a joint measurement on @xmath243 qubits and decodes the message . in practice
it is really difficult to discriminate multi - qubit states experimentally . as an alternate
, bob can do @xmath163 c - not operations keeping alice s qubits @xmath340 as controls and his qubits @xmath341 , respectively as targets and measure last @xmath342 qubits in a computational basis and first two qubits in bell basis ( which are generally a local transformation away from bell states ) and recover the message .
+ ( b ) if charlie sends his measurement results to bob , alice encodes her message and sends her qubits to bob who applies @xmath163 c - not operations as discussed above , measures last @xmath342 qubits in a computational basis and then introduces an auxiliary qubit so that he can discriminate alice s operation by doing joint unitary transformation on three qubits . after the transformation ,
bob measures the state of the auxiliary qubit and discriminates alice s operations to decode the message . although the average information transferred in both the cases is @xmath343 $ ] , the case where charlie sends his measurement results to bob is more appealing as the joint transformation that bob needs to do involves only 3 qubits , however , the one which alice does involves @xmath245 qubits . in the case of @xmath0-qubit generalization of @xmath246 ,
alice is able to send a @xmath243-bit information to bob independent of charlie s measurement basis .
we have given a criterion to assess the degree of entanglement between qubits / spin-1/2 particles using statistical correlation coefficients as a measure of entanglement .
ursell - mayer type correlation functions have been suggested to calculate the correlations between multiple particles which are extensions to the two - particle functions . the use of ursell - mayer type correlation functions to calculate the entanglement between multi - particles is an attempt to generalize the definition of degree of entanglement in multi - particle systems irrespective of the number of particles involved .
the criterion is shown to be unique in characterizing different entangled systems in different families .
it has been shown that the local transformations between two - states can be established by visual examination of the non - zero correlation coefficients associated with the systems under study .
the criterion developed here is compared with the existing entanglement norms for two and three particle maximally and non - maximally entangled systems and found to be in excellent match .
we have proposed a maximally and genuinely entangled five - particle quantum channel and described an efficient theoretical approach for direct quantum teleportation of multi - particle information .
the process discussed here overcomes the difficulty of getting null results in half of the projections when dealing with odd number of particles comprising a quantum channel .
teleportation using a controller has also been shown to be effective using appropriate projection basis .
physical realization of states and quantum teleportation protocols are analyzed using standard quantum gates and circuit diagrams .
quantum dense coding using the state proposed has been shown to be optimal . for
controlled dense coding process it has been observed that unlike cluster and @xmath110 type of states where amount of information transfer depends on the analyzer angle used by the controller , the information transfer using the state proposed is always maximum irrespective of the measurement basis used by the controller . a comparison between average information transferred in the case of the state proposed here and other five - qubit states including ghz state has been made through a plot at various analyzer angles to have more insight into the entanglement properties and representation of quantum channel used .
it has been observed that for higher number of qubits when a controller is involved , it is desirable that measurement results be transmitted to the receiver and not to the sender .
ak is grateful to iit madras for a graduate fellowship .
msk would like to acknowledge iit madras for research funds .
the correlation coefficients @xmath344 for five - particle systems are given by @xmath345 -\left \langle \sigma^{2}_{\beta}\right \rangle \left[c^{1345}_{\alpha\gamma\delta\kappa}\right ] -\left \langle \sigma^{3}_{\gamma}\right \rangle \left[c^{1245}_{\alpha\beta\delta\kappa}\right ] \nonumber \\ & - & \left \langle \sigma^{4}_{\delta}\right \rangle \left[c^{1235}_{\alpha\beta\gamma\kappa}\right ] -\left \langle \sigma^{5}_{\kappa}\right \rangle \left[c^{1234}_{\alpha\beta\gamma\delta}\right ] -\left \langle \sigma^1_{\alpha } \sigma^2_{\beta } \right \rangle \left \langle \sigma^3_{\gamma } \sigma^4_{\delta } \sigma^5_{\kappa}\right \rangle \nonumber \\ & - & \left \langle \sigma^1_{\alpha } \sigma^3_{\gamma } \right \rangle \left \langle \sigma^2_{\beta } \sigma^4_{\delta } \sigma^5_{\kappa}\right \rangle - \left \langle \sigma^1_{\alpha } \sigma^4_{\delta } \right \rangle \left \langle \sigma^2_{\beta } \sigma^3_{\gamma } \sigma^5_{\kappa}\right \rangle - \left \langle \sigma^1_{\alpha } \sigma^5_{\kappa } \right \rangle \left \langle \sigma^2_{\beta } \sigma^3_{\gamma } \sigma^4_{\delta}\right \rangle \nonumber \\ & - & \left \langle \sigma^2_{\beta } \sigma^3_{\gamma } \right \rangle \left \langle \sigma^1_{\alpha } \sigma^4_{\delta } \sigma^5_{\kappa}\right \rangle- \left \langle \sigma^2_{\beta }
\sigma^4_{\delta } \right \rangle \left \langle \sigma^1_{\alpha } \sigma^3_{\gamma } \sigma^5_{\kappa}\right \rangle - \left \langle \sigma^2_{\beta } \sigma^5_{\kappa } \right \rangle \left \langle \sigma^1_{\alpha } \sigma^3_{\gamma } \sigma^4_{\delta } \right \rangle \nonumber \\ & - & \left \langle \sigma^3_{\gamma } \sigma^4_{\delta } \right \rangle \left \langle \sigma^1_{\alpha } \sigma^2_{\beta } \sigma^5_{\kappa}\right \rangle - \left \langle \sigma^3_{\gamma } \sigma^5_{\kappa } \right \rangle \left \langle \sigma^1_{\alpha } \sigma^2_{\beta } \sigma^4_{\delta}\right \rangle - \left \langle \sigma^4_{\delta } \sigma^5_{\kappa } \right \rangle \left \langle \sigma^1_{\alpha } \sigma^2_{\beta } \sigma^3_{\gamma}\right \rangle \nonumber \\ & + & \left\langle \sigma^1_{\alpha } \right\rangle \left\langle \sigma^2_{\beta } \sigma^3_{\gamma}\right\rangle \left\langle \sigma^4_{\delta } \sigma^5_{\kappa}\right\rangle + \left\langle \sigma^1_{\alpha } \right\rangle \left\langle \sigma^2_{\beta } \sigma^4_{\delta}\right\rangle \left\langle \sigma^3_{\gamma } \sigma^5_{\kappa}\right\rangle + \left\langle \sigma^1_{\alpha } \right\rangle \left\langle \sigma^2_{\beta } \sigma^5_{\kappa}\right\rangle \left\langle \sigma^3_{\gamma } \sigma^4_{\delta}\right\rangle \nonumber \\ & + & \left\langle \sigma^2_{\beta } \right\rangle \left\langle \sigma^1_{\alpha } \sigma^3_{\gamma}\right\rangle \left\langle \sigma^4_{\delta } \sigma^5_{\kappa}\right\rangle + \left\langle \sigma^2_{\beta } \right\rangle \left\langle \sigma^1_{\alpha } \sigma^4_{\delta}\right\rangle \left\langle \sigma^3_{\gamma } \sigma^5_{\kappa}\right\rangle + \left\langle \sigma^2_{\beta } \right\rangle \left\langle \sigma^1_{\alpha } \sigma^5_{\kappa}\right\rangle \left\langle \sigma^3_{\gamma } \sigma^4_{\delta}\right\rangle \nonumber \\ & + & \left\langle \sigma^3_{\gamma } \right\rangle \left\langle \sigma^1_{\alpha } \sigma^2_{\beta}\right\rangle \left\langle \sigma^4_{\delta } \sigma^5_{\kappa}\right\rangle + \left\langle \sigma^3_{\gamma } \right\rangle \left\langle \sigma^1_{\alpha } \sigma^4_{\delta}\right\rangle \left\langle \sigma^2_{\beta } \sigma^5_{\kappa}\right\rangle+ \left\langle \sigma^3_{\gamma } \right\rangle \left\langle \sigma^1_{\alpha } \sigma^5_{\kappa}\right\rangle \left\langle \sigma^2_{\beta } \sigma^4_{\delta}\right\rangle \nonumber \\ & + & \left\langle \sigma^4_{\delta } \right\rangle \left\langle \sigma^1_{\alpha } \sigma^2_{\beta}\right\rangle \left\langle \sigma^3_{\gamma } \sigma^5_{\kappa}\right\rangle+ \left\langle \sigma^4_{\delta } \right\rangle \left\langle \sigma^1_{\alpha } \sigma^3_{\gamma}\right\rangle \left\langle \sigma^2_{\beta } \sigma^5_{\kappa}\right\rangle + \left\langle \sigma^4_{\delta } \right\rangle \left\langle \sigma^1_{\alpha } \sigma^5_{\kappa}\right\rangle \left\langle \sigma^2_{\beta } \sigma^3_{\gamma}\right\rangle \nonumber \\ & + & \left\langle \sigma^5_{\gamma } \right\rangle \left\langle \sigma^1_{\alpha } \sigma^2_{\beta}\right\rangle \left\langle \sigma^3_{\gamma } \sigma^4_{\delta}\right\rangle+ \left\langle \sigma^5_{\gamma } \right\rangle \left\langle \sigma^1_{\alpha } \sigma^3_{\gamma}\right\rangle \left\langle \sigma^2_{\beta } \sigma^4_{\delta}\right\rangle + \left\langle \sigma^5_{\gamma } \right\rangle \left\langle \sigma^1_{\alpha } \sigma^4_{\delta}\right\rangle \left\langle \sigma^2_{\beta } \sigma^3_{\gamma}\right\rangle \nonumber \\ & - & 6 \left\langle \sigma^1_{\alpha}\right\rangle \left\langle \sigma^2_{\beta}\right\rangle \left\langle \sigma^3_{\gamma}\right\rangle \left\langle \sigma^4_{\delta}\right\rangle \left\langle \sigma^5_{\kappa}\right\rangle.\end{aligned}\ ] ] list of figures 1 . quantum circuit to prepare the set of states @xmath246 represented by eq . ( 18 ) .
2 . quantum network to teleport an arbitrary two qubit state @xmath346 $ ] using @xmath223 as quantum channel .
3 . controlled dense coding of five - qubit state @xmath334 with controller - sender interface .
4 . controlled dense coding of five - qubit state @xmath334 with controller - receiver interface .
comparison of the efficiency of information transfer between states given by eq .
( 26 ) and the five qubit ghz state .
60 bennett c h , brassard g , crepeau c , jozsa r , peres a and wootters w k 1993 _ phys . rev . lett . _ * 70 * 1895 bennett c h and wiesner s j 1992 _ phys .
lett . _ * 69 * , 2881 zukowski m , zeilinger a , horne m a and eckert a 1993 _ phys .
lett . _ * 71 * 4287 schumacher b 1995 _ phys .
a _ * 51 * 2738 greenberger d m , horne m and zeilinger a 1989 _ bell s theorem , quantum theory and conceptions of the universe _
kluwer , dordrecht ; bouwmeester d , pan j w , daniell m , weinfurter h and zeilinger a 1999 _ phys .
* 82 * 1345 pan j w , daniell m , gasparoni s , weihs g and zeilinger a 2001 _ phys .
lett . _ * 86 * 4435 weinfurter h and zukowski m 2001 _ phys .
rev . a _ * 64 * 010102(r ) zhao z , chen y a , zhang a n , briegel h j and pan j w 2004 _ nature _ * 430 * 54 walther p , resch k j , rudolph t , schenck e , weinfurter h , vedral v , aspelmeyer m and zeilinger a 2005 _ nature _ * 434 * 169 karlsson a and bourennane m 1998 _ phys .
a _ * 58 * 4394 bernstein h j _ quant .
* 5 * 451 wang x w , su y h and yang g j 2009 _ quant .
* 8 * , 319 shi b s and tomita a 2002 _ phys .
a _ * 296 * 161 gorbachev v n and trubilko a i 2000 _ jetp _ * 91 * 894 cao h j. song h s 2007 _ quant .
proc . _ * 6 * , 1636 muralidharan s , panigrahi p k 2008 _ phys . rev .
a _ * 77 * 032321 rigolin g 2005 _ phys .
a _ * 71 * , 032303 fang j , lin y , zhu s and chen x 2003 _ phys .
a _ * 67 * 014305 deng f g , li c y , li y s , zhou h y and wang y 2005 _ phys . rev . a _ * 72 * 022338 yeo y and chua w k 2006 _ phys .
lett . _ * 96 * 060502 yang c p , chu s and han s 2004 _ phys .
a _ * 70 * 022329 man z x , xia y j and an n b 2007 _ phys .
rev . a _ * 75 * 052306 bouwmeester d , mattle k , pan j w , weinfurter h , zeilinger a and zukowski m 1997 _ nature _ * 390
* 575 braunstein s l and kimble h j 1998 _ nature _ * 394 * 840 furusawa a , sorensen j l , braunstein s l , fuchs c a , kimble h j and polzik e s 1998 _ science _ * 282 * 706 boschi d , branca s , martini f d , hardy l and popescu s 1998 _ phys .
* 80 * 1121 kim y h , kulik sp and shih y 2001 _ phys .
lett . _ * 86 * 1370 sherson j f , krauter h , olsson r k , julsgaard b , hammerer k , cirac i and polzik e s 2006 _ nature _ * 443 * 557 mattle k , weinfurter h , kwiat p g , zeilinger a 1996 _ phys .
lett . _ * 76 * 4656 hao j c , li c f and guo g c 2001 _ phys .
a _ * 63 * 054301 bose s , plenio m and vedral v 2000 _ j. mod .
_ * 47 * 291 lee h j , ahn d and hwang s w 2002 _ phys . rev .
a _ * 66 * 024304 schaetz t , barrett m d , leibfried d , chiaverini j , britton j , itano w m , jost j d , langer c and wineland d j 2004 _ phys .
lett . _ * 93 * 040505 cory d g , price m d and havel t f 1998 _ physica d _ * 120 * 82 nielsen m a , knill e and laflamme r 1998 _ nature _ * 396 * 52 boulant n , emerson j , havel t f , cory d g and furuta s 2004 _ j. chem .
phys . _ * 121 * 2955 cory d g , fahmy a m r f and havel t f 1997 _ proc .
* 94 * 1634 fang x , zhu x , feng m , mao x and du f 2000 _ phys . rev .
a _ * 61 * 022307 furman g b , meerovich v m and sokolovsky v l 2009 _ quant .
_ * 8 * 283 jones j a 2001 _ prog .
spectrosc . _ * 38 * 325 mahesh t s , sinha n , ramanathan k v and kumar a 2002 _ phys . rev . a _ * 65 * 022312 suter d and mahesh t s 2008 _ j. chem . phys . _ * 128 * 052206 schlienz j and mahler g 1995 _ phys .
a _ * 52 * 4396 osborn t a 1977 _ phys
a _ * 16 * 334 huang k 1987 _ statistical mechanics _
john wiley new york altafini c 2004 _ phys .
rev . a _ * 69 * 012311 watson k m 1956 _ phys .
* 103 * 489 fano u 1957 _ rev
* 29 * 74 fano u 1983 _ rev .
_ * 55 * 855 blum k 1981 _ density matrix theory and application _ plenum press new york bennett c h , bernstein h j , popescu s and schumacher b 1996 _ phys . rev . a _ * 53 * 2046 hill s and wootters w k 1997 _ phys . rev .
lett . _ * 78 * 5022 wootters w k 1998 _ phys .
lett . _ * 80 * 2245 coffman v , kundu j and wootters w k 2000 _ phys .
a _ * 61 * 052306 love p j , brink a m , smirnov a yu , amin m h s , grajcar m , ilichev e , izmalkov a and zagoskin a m 2007 _ quant .
proc . _ * 6 * 187 vedral v , plenio m b , rippin m a and knight p l 1997 _ phys
_ * 78 * 2275 vedral v 2002 _ rev .
phys . _ * 74 * 197 bruss d , datta n , ekert a , kwek l c and macchiavello c 2005 _ phys . rev .
a _ * 72 * 014301 wei t c , das d , mukhopadyay s , vishveshwara s and goldbart p m 2005 _ phys .
* 71 * 060305(r ) facchi p. florio g and pascazio s 2006 _ phys .
rev . a _ * 74 * 042331 audenaert k m r and plenio m b 2006 _ new .
j. phys . _
* 8 * 266 dur w , vidal d and cirac j i 2000 _ phys .
rev . a _ * 62 * 062314 briegel h j and raussendorf r 2001 _ phys .
_ * 86 * 910 hausladen p , jozsa r , schumacher b , westmorelan , m and wootters w k _ phys .
a _ * 54 * 1869 bowen g 2001 _ phys .
a _ * 63 * 022302 | we discuss and generalize multi - particle entanglement based on statistical correlations using ursell - mayer type of cluster coefficients .
cluster coefficients are used to distinguish different , independent entangled systems as well as those which are connected through local unitary transformations .
we propose a genuinely and maximally entangled five - particle state for efficient information processing .
the physical realization of entangled states and information processing protocols are analyzed using quantum gates and circuit diagrams .
we show that direct as well as controlled communication can be achieved using the state proposed here , with certainty in the case of teleportation and with a high degree of optimity in the case of dense coding . for controlled dense coding the amount of information transferred from the sender to the receiver is always a maximum irrespective of the measurement basis used by the controller . |
although the nature of the superconducting pair wave function in high -@xmath7 cuprates is not yet known strong evidences of a major @xmath8 symmetry exists @xcite .
experiments sensitive to the internal phase structure of the pair wave function reported a sign reversal of the order parameter supporting @xmath3 wave symmetry @xcite .
most recently from various experiments and theory it appears that the pairing symmetry of these family could be a mixed one like @xmath0 where @xmath4 could be something in the @xmath9 wave family or @xmath10 .
there were early questions from tunneling experiments regarding the pure d - wave symmetry@xcite as the data supports an admixture of d and s - wave components due to orthorhombicity in ybco@xcite .
possibility of a minor but finite @xmath11 symmetry alongwith the predominant @xmath12 has also been suggested@xcite in connection with magnetic defects or small fractions of a flux quantum @xmath13 in ybco powders .
similar proposals came from various other authors in the context of magnetic field , magnetic impurity , interface effect etc .
@xcite the experimental result by krishana _
_ , was interpreted as a signature of induction of a minor component _ eg _ , @xmath11 or @xmath14 in a @xmath3-wave superconducture with the application of magnetic field along the @xmath5 axis . in this work
, we study in details the effect of a weak zeeman magnetic field on superconductors with mixed order parameter symmetry like @xmath15 with @xmath16 , @xmath9 for arbitrary @xmath6 .
it is well known that such superconductors with @xmath17 and @xmath18 corresponds to broken time reversal states ( btrs ) .
these btrs states lift the directional degeneracy of charge currents by admixing a subdominant @xmath4-wave component to the d - wave pairing state and a spontaneous finite current appears@xcite .
an application of zeeman magnetic field can lift the spin degeneracy leading to suppression of btrs ; a pure @xmath3-wave occurs with increasing magnetic field .
for mixed symmetry as above with @xmath19 that preserves time reversal symmetry and are nodeful respond differently to the zeeman field as compared to @xmath17 states . for nodefull @xmath19 state , the local gap @xmath20 of small magnitude over the fermi surface may be destructed with the application of zeeman field leading to a paramagnet pocket .
this although true for @xmath17 states , but such states correspond to _ fully gapped _ situation all over the fermi surface which causes _ weak _ response to the magnetic field . a clear picture on the above will be demonstrated in this article .
it may be mentioned that the high temperature superconductors are quasi two dimensional in nature and therefore , a magnetic field parallel to the @xmath2 plane does not couple to the orbital motion of the electrons in the plane .
therefore , we shall not consider spin - orbit interaction in this work . in connection with the discussion of order parameter symmetry in cuprates
, we would further like to mention that the proposal of mixed order parameter symmetry got the correct momentum when experimental data on longitudinal thermal conductivity by krishana _
@xcite of @xmath21 compounds and that by movshovich _ et al , _
@xcite showed supportive indication to such proposals .
there are experimental results related to interface effects as well as in the bulk that indicates mixed pairing symmetry ( with dominant @xmath3-wave ) @xcite , thus providing a strong threat to the pure @xmath3 wave models .
there were early orginal works as regards to the modifaction of superconductivity due to application of zeeman magnetic @xcite and very recently , the zeeman suppression was discussed in mesoscopic systems @xcite .
the free energy of a two dimensional planar superconductor with arbritary pairing symmetry in presence of a magnetic field may be written as , @xmath22 where @xmath23 are the energy eigen values of a hamiltonian that describes superconductivity , @xmath24 is the magnetic moment of the electrons .
this includes the assumption that the zeeman field raises / lowers the energy of the spin up / down quasiparticle states .
we minimize the free energy , eq .
( [ free ] ) _ i.e _ , @xmath25 = 0 , to get the gap equation as , @xmath26 where @xmath27 is the dispersion relation taken from the arpes data @xcite and @xmath28 the chemical potential will control band filling through a number conserving equation given below . since the applied zeeman field modifies the sc quasiparticles of spin up and down differently ,
their occupation probablities are also modified .
the number conserving equation that controls the band filling through chemical potential , @xmath28 in presence of zeeman field is given by , @xmath29 \label{dens}\end{aligned}\ ] ] where @xmath30 .
let us consider that the overlap of orbitals in different unit cells is small compared to the diagonal overlap . then in the spirit of tight binding lattice description , the matrix element of the pair potential used in the sc gap equation eq .
( [ gapeq ] ) may be obtained as , @xmath31 where in the first result of the equation ( [ ppot ] ) @xmath32 locates nearest neighbour and further neighbours , @xmath33 labels and @xmath34 , @xmath35 represents strength of attraction between the respective neighbour interaction .
the first term in the above equation @xmath36 refers to the on - site interaction which has an effective attractive value giving rise isotropic @xmath9 wave .
the second and third terms are responsible for @xmath3 and extended @xmath9 wave symmetry superconductivity whereas the @xmath37 and the @xmath38 terms are responsible for @xmath10 and @xmath39 symmetries respectively .
we restrict only to singlet pairing states ( _ i.e _ , @xmath40 ) as applicable for high temperature superconductors .
the momentum form factors are obtained as , @xmath41 for two component order parameter symmetries as mentioned above , we substitute the required form of the potential and the corresponding gap structure into the either side of eq .
( [ gapeq ] ) which gives us an identity equation . then separating the real and imaginary parts together with comparing the momentum dependences on either side of it we get gap equations for the amplitudes in different channels as , @xmath42 \label{gapcomp}\end{aligned}\ ] ] where @xmath43 corresponding to two components @xmath3 and @xmath9 or @xmath3 and @xmath10 symmetries . considering mixed symmetry of the form
@xmath44 one identifies @xmath45 , @xmath46 , @xmath47 , @xmath48 and @xmath49 , @xmath50 in eq .
( 4 ) . similarly , for mixed symmetries of the form @xmath51 @xmath52 , @xmath53 and @xmath49 , @xmath54 of eq.(4 ) .
the potential required to get such pairing symmetries are discussed in eq .
( [ symm ] ) .
we solve self - consistently the above three equations ( eq.[gapcomp ] and eq.[dens ] ) in order to study the phase diagram of a mixed order parameter superconducting phase in presence of zeeman magnetic field .
the numerical results obtained for the gap amplitudes through eqs .
( [ gapcomp],[dens ] ) will be compared with free energy minimizations via eq .
( [ free ] ) to get the phase diagrams .
we present in this section our numerical results for a set of fixed parameters , _ e.g _ , a cut - off energy @xmath55= 500 k around the fermi level above which superconducting condensate does not exist , a fixed transition temperature of the minor component @xmath56 k and the bulk @xmath57 k determined by the @xmath3-wave order parameter . in figures 1 and 2 we present results for @xmath58 symmetries for @xmath59 and @xmath19 respectively .
such symmetries would arise from a combination of two component pair potentials ( @xmath60 ) , ( @xmath61 ) terms in eq .
[ ppot ] for @xmath62 respectively .
the amplitudes of extended @xmath9-wave states like @xmath63 , @xmath39 are found to be finite only towards very low band filling , @xmath64 for @xmath65 hence does not cause any mixing with the predominant @xmath3-wave .
therefore , we shall discuss only the results of @xmath19 and @xmath66 for @xmath62 .
these two phases of @xmath6 can cause important differences .
it is known that for any @xmath17 , time reversal symmetry is locally broken @xcite at lower temperatures with the onset of the secondary component which correspond to a phase transition to an fully gapped phase for @xmath67 , from a partially ungapped phase of @xmath12 symmetry . on the other hand ,
the @xmath68 phase still remains nodeful , although the nodal lines shifts a lot from the usual @xmath69 lines of the @xmath12 . in fig .
1(a ) and ( b ) we present the temperature dependencies of the order parameters in the complex mixed symmetry .
curves corresponding to the @xmath3-wave channels are represented with thinner joining lines of different styles whereas the minor @xmath9 component is denoted through that of same style but with thicker lines ( same strategy will be carried out in other figures as well ) .
for a zero zeeman field ( @xmath67 ) , the amplitude of the @xmath3-wave is suppressed with the onset of the minor @xmath9 component at @xmath70 k which leads to a kink like structure ( cf .
the thin solid curve in fig1 ( a ) ) .
= 4.5truein = 4.5truein with application of weak zeeman field the transition temperature of the minor @xmath9 state decreases while the zero temperature magnitude remains the same .
this causes a shift in the kink like structure in the thermal dependence of the @xmath3 wave channel towards lower temperature with increasing field and hence a small enhancement in the @xmath3-wave with field at lower temperature occurs .
this point will be clearer from fig .
5(b ) as discussed latter . at a field value
@xmath71 ev , the @xmath9 wave component is completely suppressed leading to a pure @xmath3 wave phase .
thus we have a magnetic field induced transition at lower temperature from _ fully gapped phase _ of the @xmath72 state to a _ partially _ gapped phase of the @xmath3-wave .
therefore , in absence or very low magnetic field , there is a transition from a _ partially gapped _
phase of the @xmath3-wave at higher temperature to a _ fully gapped _
@xmath72 phase at lower temperatures . with increasing field the fully gapped phase region with respect to temperature decreases and brings back the ungapped phase of the @xmath3-wave .
these phase transitions will have important bearings in the thermodynamic and transport properties .
= 3.25truein = 3.25truein in fig .
1(b ) we present the paramagnetic state of the @xmath3-wave .
this phase has also been qualitatively investigated recently by yang and sondhi @xcite with possibility of pairing with finite momentum .
we therefore restrict to present only the details thermal dependence of the @xmath3-wave superconductivity in presence of zeeman magnetic field that were not discussed .
we show that with increasing field ( at lower fields ) the zero temperature magnitude of the @xmath3-wave remains unchanged whereas the @xmath7 is reduced . at higher field _ e.g _ , @xmath73 one sees a first order transition from superconducting state to the normal state with respect to temperature .
this behaviour causes a magnetic field induced enhancement of the @xmath74 ratio .
this ratio is crucial for many physical properties like specific heat jump etc . and
hence expected to have drastic effect with magnetic field .
= 3.25truein = 3.25truein in fig.2 we describe the thermal behaviour of superconductors with @xmath75 symmetry in presence of magnetic field .
first of all , in absence of magnetic field , the thermal behaviors at lower temperatures is quite different ( as we have seen in fig .
1 ) , the @xmath9-wave gap opens very fast below @xmath70 k and also induces a growth to the @xmath3 at a faster rate than that above @xmath70 k. while the @xmath9-wave has about three times the zero temperature value compared to @xmath72 phase , the @xmath3 wave also have quite larger value .
with application of a very small magnetic field the @xmath9-component is suppressed largely ; both its @xmath7 and the zero temperature gap being suppressed .
the @xmath3-wave gap magnitude is also suppressed while its @xmath7 remained practically unchanged with such small values of the magnetic field .
more importantly , although both the @xmath3 and the @xmath9 channel has larger magnitude in the @xmath75 phase the critical field ( @xmath76 ev ) at which the @xmath9 component vanishes completely is _ smaller _ compared to that for the @xmath59 phase ( @xmath77 ev ) of the mixed symmetry .
distinctly , the response of the zeeman field to the @xmath75 superconductors is more pronounced than the @xmath72 superconductors .
this study therefore also revealed the importance of the phase of the minor component in @xmath3-wave superconductors with a pratical example of effect of zeeman magnetic field , for the first time . in order to establish the importance of the phase of the minor component we also study the effect of zeeman magnetic field in @xmath78 and @xmath79 symmetry superconductors respectively
the qualitative as well as quantitative behaviors remain almost same as that shown in figures 1 and 2 . thus establishing that the response of the superconducors with mixed order parameter symmetries with @xmath68 phase of the minor component is stronger .
= 3.25truein = 3.25truein in figures 5 and 6 we concentrate on studying behavior of such mixed order parameter symmetry superconductors in presence of magnetic field as a function of band filling ( @xmath80 ) for @xmath10 and @xmath9 as minor components respectively .
with application of a small field the minor component @xmath16 or @xmath9 is suppressed only at lower fillings and then with increasing field it is suppressed in the optimal doping regime suddenly .
therefore , with increasing field one finds mixed symmetry region as well as pure @xmath3-wave region with respect to filling a _ non - uniform _ superconductivity .
this behavior depends on the nature of the minor component . for @xmath81 ,
the mixed symmetry is possible around half - filling and at around @xmath82 in an intermediate field value . with increasing field mixed symmetry regions around both the band fillings shrinks and at a field value of @xmath83 the @xmath9-wave vanishes leading to a pure @xmath3-wave phase .
it may be noticed the @xmath3-wave channel is enhanced ( cf .
fig 5(b ) ) in the optimal doping region with intermediate field values as was also mentioned earlier while discussing fig1 ( a ) .
this however does not occur in case of @xmath10 .
for @xmath10 minor component , the minor component is suppressed strongly only from the lower filling with increasing field and at an intermidiate field the mixing is possible only near half filling .
the @xmath3-wave boundary towards lower fillings as well as around half - filling also shrinks with increasing field values in contrast to @xmath84 .
one always see a sharp transition from a mixed phase to a pure @xmath3-wave phase or otherway around with respect to band filling irrespective of the minor component in a given magnetic field . at larger fields ( @xmath85 ) when the minor component is suppressed completely _
i.e _ , one has a pure @xmath3-wave phase , the @xmath3-wave phase also show sharp transition from @xmath3-wave superconducting state to a normal state ( cf . fig .
= 4.5truein = 4.5truein in fig .
6 we study the same as that in fig . 5 for @xmath75 and @xmath86 symmetries .
in contrast to the @xmath66 phase of the minor component ( as in fig .
5 ) , the minor components have very large values in the @xmath19 phase although have the same @xmath87 as mentioned earlier , in absence of the magnetic field . with very small magnetic field such large amplitudes of the condensation in the minor channel is strongly suppressed ( see fig 6(b ) specially ) . the nature of suppression , as in fig.5 , is different for different minor component .
for example , @xmath10 is suppressed only from the lower filling whereas the @xmath9-wave is suppressed both from the lower filling as well as around the optimal doping .
= 4.5truein = 4.5truein overall , it is very distinct that the magentic field affects the @xmath19 mixed phase more stongly than the @xmath66 phase .
this is due to the fact that the response of applied zeeman field is paramagnetic with destruction of superconductivity over parts of the fermi surface where the zeeman field exceeds the local magnitude of the @xmath88-dependent gap resulting a spin polarization in the normal electrons . for @xmath59 phase
, the nodes are missing and the fermi surface is gapped all over , although the gap will have local minima .
thus response of the @xmath66 phase is weaker compared to the @xmath89 phase .
noticiably , the critical value at which the minor component vanishes completely in all band fillings for the @xmath19 phase is @xmath90 ev whereas that for @xmath66 is @xmath91 ev .
( the value of @xmath92 depends on @xmath80 , as well as @xmath4 and @xmath6 ) . above @xmath93 ,
the paramegnetic state of the @xmath3-wave superconductors are also presented .
the region of @xmath3-wave superconductivity shrinks with increasing field around the band filling @xmath94 .
the change from @xmath3-wave to spin polarized normal state is very sharp with respect to the band filling . in the @xmath75 state
a plateau has been observed in the @xmath3 channel in weak or zero field , similar to the behaviour known for the ybco systems @xcite .
it may be mentioned that krishana _
et al . _ , found the longitudinal thermal conductivity of @xmath95 at lower temperatures ( 5k to 20 k ) decreases with the increase in magnetic field applied along @xmath96-axis . above a critical value of the magnetic field @xmath97 , the thermal conductivity cease to change with the magnetic field and develops a plateau .
it was proposed @xcite that the @xmath8 pairing state is unstable against the formation of @xmath98 ( where @xmath99 ) in presence of @xmath100 such that the loss of quasiparticle transport in the thermal conductivity can be explained . in contrast , we started with a @xmath98 picture and application of zeeman field ( which may be mapped as application of magnetic field parallel to the 2d @xmath2 plane ) without considering the orbital effect did not find any enhancement in the condensation of the minor channel but suppression leading to a pure paramegnetic @xmath3 state .
in summary we have performed a detailed study on the effect of zeeman magnetic field on mixed pairing symmetry with predominant @xmath3-wave which seems to be very promising symmetry for the high @xmath7 systems .
thus we have described the paramagnetic state in the mixed symmetry superconductors and subsequently in the @xmath3-wave superconductors .
in particular we established that the phase of the minor component mixed with predominant @xmath3-wave is of immense importance . the @xmath19 phase minor component symmetry responds to zeeman field more profoundly than the @xmath66 of the minor component .
it will be very interesting to calculte the specific heat , magnetization , density of states as a function of magnetic field using this model .
we argued that the orbital effect is secondary when a magnetic field is applied parallel to the conducting plane .
this may indicate that the experimental observation by krishana _
et al . _ , involve strong coupling of spins to orbitals due to application of magnetic field perpendicular to the plane at lower temperatures .
however , the order parameter alone does not completely determine the thermodynamic property of a system . in ref .
@xcite , some estimations and scaling relations in the change of various physical properties due to zeeman field are given for a pure @xmath3-wave superconductor .
it turns out , a weak zeeman field does little to the order parameter but may profoundly affect the thermodynamic property of a pure @xmath12 supercondcutor .
such effects will presumably also remain for mixed superconductors with @xmath89 as the gapnodes similar to @xmath12 remains , its effect in the @xmath66 phase has to be studied more carefully .
a large part of this work was carried out at the instituto de fsica , universidade federal fluminense , brazil and was financially supported by the brazilian funding agency faperj , project no .
e-26/150.925/96-bolsa .
d. l. cox and m. b. maple , phys .
today , 32 , feb . 1995 .
d. j. van harlingen , rev .
phys , * 67 * , 515 ( 1995 ) .
d. j. scalapino , phys . rept . * 250 * , 329 ( 1995 ) .
d. a. wollman , d. j. van harlingen , w. c. lee , d. m. ginsberg and a. j. leggett , , 2134 ( 1993 ) ; d. a. brawner and h. r. ott , , 6530 ( 1994 ) ; c. c. tsuei , j. r. kirtley , c. c. chi , lock see yu - jahnes , a. gupta , t. shaw , j. z. sn and m. b. ketchen , , 593,(1994 ) ; w. n. hardy , d. a. bonn , d. c. morgan , r. liang and k. zhang , , 3999 ( 1990 ) ; l. a. de vaulchier , j. p. vieren , y. guldner , n. bontemps , r. combescot , y. lema @xmath101 and j. c. image , europhys . lett . * 33 * , 153 ( 1996 ) .
dong ho wu , j. mao , s. n. mao , j. l. pong , x. x. xi , t. venkatesan , r. l. greene and s. m. anlage , , 85 ( 1993 ) ; s. m. anlage , dong - ho wu , j. mao , s. n. mao , x. x. xi , t. venkatasan , j. l. peng and r. l. greene , , 523 ( 1994 ) ; a. f. annett , n. goldenfield and a. j. leggett , jr .
phys . * 105 * , 473 ( 1996 ) .
sun , d. a. gajensk , m. b. maple and r. c. dynes , phys .
* 72 * , 2267 , ( 1994 ) .
m. b. walker , , 5835 , ( 1996 ) . c. odonovan , d. branch , j. p. carbotte and j. s. preston , , 6588 , ( 1995 ) .
m. r. norman , m. randeria , h. ding and j. c. campuzano , , 615 , ( 1995 ) .
m. sigrist , d. b. bailey , and r. b. laughlin , phys .
* 74 * , 3249 , ( 1995 ) .
r. b. laughlin , phys .
lett . * 80 * , 5188 , ( 1998 ) .
k. krishana , n. p. ong , q. li , g. d. gu , and n. koshizuka , science , * 277 * , 83 , ( 1997 ) ; k. a. kouznetsov _
et al . _ , , 3050 ( 1997 ) ; a. v. balatsky , , 1972 ( 1998 ) ; r. movshvich , m. a. hubbard , m. b. salamon , a. v. balatsky , r. yoshizaki , j. l. sarrao and m. jaime , , 1968 ( 1998 ) ; r. b. laughlin , , 5158 ( 1998 ) ; t. v. ramakrishnan , cond - mat/9803069 ; haranath ghosh , europhys . lett . * 43 * , 707 ( 1998 ) .
m. covington , g. aprili , e. paraoanu , l. h. greene , f. xu , j. zhu and c. a. mirkin , , 277 ( 1997 ) ; m. forgelstrm , d. rainer and j. a. sauls , , 281 ( 1997 ) ; m. e. zhitomirsky and m. b. walker , 1734 ( 1997 ) ; r. j. kelley , c. quitmann , m. onellion , h. berger , p. almeras , g. margaritondo , science * 271 * , 1255 ( 1996 ) ; j. ma , c. quitmann , r. j. kelley , h. berger , g. margaritondo , m. onellion , science * 267 * , 862 ( 1995 ) . for
a review see m. sigrist , prog .
* 99 * , 899 ( 1998 ) .
a. m. clogston , , 266 ( 1962 ) ; b. s. chandrasekhar , appl .
* 1 * , 7 ( 1962 ) ; p. fulde and r. a. ferrell , phys . rev . * 135 * , a550 ( 1964 ) ; a , i. larkin and yu .
n. ovchinnikov , sov .
jetp , * 20 * , 762 ( 1965 ) ; w. wu and p. adams , , 1412 ( 1994 ) ; k. gloos , r. modler , h. schimanski , c. d. bredl , c. geibel , f. steglich , a. i. buzdin , n. sato and t. komatsubara , , 501 ( 1993 ) . f. braun , jan von delft , d. c. ralph and m. tinkham , , 921 ( 1997 ) .
kun yang and s. l. sondhi , , 8566 ( 1998 ) . for a mixed phase symmetry that breaks time reversal symmetry locally , like @xmath102 , where @xmath4 = iso-@xmath9 or @xmath10 see last ref .
@xcite haranath ghosh , europhys . lett . * 43 * , 707 ( 1998 ) ; haranath ghosh , 3357 ( 1999 ) .
r. j. cava , b. b. batlogg , c. h. chen , e. a. rietman , s. m. zahurak and d. werder , nature * 329 * 423 ( 1987 ) . | we study the effect of zeeman magnetic field in the superconducting phase with two component order parameter scenario , such as , @xmath0 , where @xmath1 .
this scenario is equivalent to applying magnetic field parrllel to @xmath2 planes . in a weak magnetic field , which does not cause much change to the predominant @xmath3-wave ,
supresses the minor @xmath4 component leading to pure @xmath3-wave phase .
this observation is in contrary to the effect of magnetic field applied in the @xmath5-direction to the @xmath2 planes which is believed to induce a minor component @xmath4 to @xmath3-wave superconductors .
we also show that the response of such superconductors to a weak zeeman magnetic field can be quite different depending on the phase @xmath6 of the minor component ( @xmath4 ) .
epsf 2 |
ventricular arrhythmia is the leading cause of cardiac arrest and sudden death .
clinical observations and implantable cardioverter defibrillators have accumulated a substantial amount of data on the occurrences of ventricular arrhythmia in patients [ 1 , 36 , 37 , 38 , 52 , 59 - 61 ] .
temporal patterns of initiations and terminations of ventricular arrhythmia tend to exhibit substantial variations across different time scales , and their occurrences do not correlate decisively with medication , exertion , stress , lifestyles and similar factors .
arrhythmia events are not random ; they show circadian patterns [ 36 , 60 ] and also tend to occur in clusters . however , the detection times between consecutive events and clusters are spread out over time [ 37 , 52 , 59 ] , making it difficult to understand their causes and make predictions about their occurrences . unlike the circadian patterns ,
these clusterings or their patterns of occurrences are not affected by the long - term administration of antiarrhythmic drugs [ 59 ] . in spite of the abundant data in existence ,
the basis for non - circadian patterns is not well - understood .
many factors , ranging from internal cardiac mechanisms to external chance events play significant roles in shaping the electrocardiogram ( ecg ) recordings and the implantable defibrillator data . separating
all of the possible contributions is a formidable task , but understanding the influences of various factors and the extent to which each plays a role may have important consequences for the diagnosis and treatment of tachyarrhythmias .
a number of prior studies using special preparations of animal cell cultures establish that complex , spontaneously generated patterns may occur without some of the features peculiar to the heart ( e.g. the 3-dimensional geometry or biological features such as valves , different tissue layers and types , etc ) [ 3 , 25 , 35 , 44 ] .
these studies also indicate ( both theoretically and experimentally ) that causes external to the heart , whether random or deterministic , are not always necessary for the occurrence of irregular behavior . a different class of studies that involve interactive self - oscillatory sources also establish the capability of basic cardiac mechanisms to generate complex rhythms without considering the whole heart [ 23 , 28 , 43 ] .
from these and similar studies it may be inferred that changes in the basic internal cardiac mechanisms can play a significant role in generating and sustaining arrhythmias . to clarify the role of these basic internal mechanisms at a fundamental level , in this paper we focus on the unidirectional reentrant circulation of action potentials in a loop of ventricular tissue .
because of their pervasiveness , relative simplicity and importance in clinical arrhythmogenesis reentrant loops have received a great deal of attention [ 4 , 8 , 11 , 14 - 16 , 20 , 26 , 30 , 41 , 44 , 48 - 50 , 55 ] .
studying spontaneous initiation and termination of reentry ( sitr ) patterns in the loop context is considerably simpler than in the whole heart and can offer potentially useful insights into the clinically observed patterns of arrhythmia . in the construction of a workable model , a certain level of abstraction bridges over physiological complexities that do not play a central role in the long - term evolution of temporal patterns . in this paper
we consider a model that combines the traditional discrete loop with a pacer to form an interactive system .
fundamental thresholds are dynamically integrated so they can either trigger reentry or inhibit it .
although a relatively simple mechanism , this composite system has two important features : it is an essential component of a well - known cardiac anomaly ; and it is capable of generating complex initiation and termination patterns . in a series of case - studies
we show that the loop - pacer mechanism may be responsible for irregularities in onset and termination of reentry beyond such pervasive factors as tissue heterogeneity , the geometry of the heart and other physiological and non - physiological factors .
even without such complexities , the sitr patterns generated by the loop - pacer system can be complex ; i.e. they contain many regular features yet their evolution over time is difficult to predict .
we trace this complexity to the interaction of a threshold with bistability of the reentrant circulation and in some cases , also to the inherent discontinuity of phase resetting interactions between the loop and the pacer .
bistability is an inherent physiological characteristic that may be attributed to non - uniformities in the restitution of action potential duration ( apd ) ; this feature emerges prominently when conduction velocity ( cv ) is properly matched with apd .
when this happens ( e.g. when the length of the loop is within a certain range ) the sitr patterns become considerably more complex and unpredictable .
in this section we describe the main features of the loop - pacer model , leaving a few additional details to section 4 .
the pacer is a self - oscillatory mechanism that is tied to the reentrant circuit ( the loop in this paper ) via a time lag or delay parameter and a special threshold .
consider a loop of cardiac tissue consisting of cells that conduct action potentials and assume that a single conducting pathway connects the loop to the rest of the heart .
let the physical length of the loop be denoted by @xmath4 measured in centimeters .
divide the loop into @xmath5 sets or aggregates of cells , each of which may be called a _ cell aggregate _ or for brevity , just a cell .
the cell aggregate that is connected to the only pathway out of the loop is also the gateway through which pulses enter the loop or exit it ; we label it cell 1 ; see figure 1 .
this cell is adjacent not only to cell 2 , but also to cell @xmath5 at its other end .
[ schematicloop ] for each integer @xmath6 between @xmath7 and @xmath8 label the length of the @xmath6-th cell @xmath9 these cell aggregates lengths are not necessarily equal but of course , they add up to @xmath10 since each cell aggregate must have at least one cardiac cell in it , @xmath11 has a natural lower bound , namely the nominal length of a single cardiac cell ( about 0.01 cm or 100 microns ) .
the closer all @xmath11 are to this lower bound , the greater is the number @xmath5 of cell aggregates in a loop of fixed length @xmath4 ; for a 12 cm loop with cells of length 0.01 cm , the number of cells is 1200 .
while for discrete modeling it is not necessary to pick @xmath5 that large , it needs to be large enough ( hence each cell aggregate small enough ) that the conduction velocity from one end of an aggregate to the other can be taken to be approximately constant .
this assumption is essential if the restitution of conduction time ( ct ) is defined in terms of the restitution of cv , as we do later on .
we assume that all the cardiac cells within a given aggregate are identical .
if all cells in the _ entire loop _ have identical physiological characteristics then we may also let cell aggregates all have equal lengths .
in such a case , the loop is said to be homogeneous .
otherwise the loop is heterogeneous . to initiate unidirectional circulation ,
it is necessary that a unidirectional block ( ub ) exist somewhere in the loop ; for simplicity , we place it in cell 1 , the gateway to the loop ; see figure 1 .
thus cell 1 is able to activate cell 2 but not cell @xmath12 for our simulations , we use a simplified version of the ub time window that is defined by a threshold value for the diastolic interval ( di ) .
thus conduction is blocked in any cell whose di is not greater than the threshold @xmath13 ; see section 4 below for more details .
such a minimum value is used in [ 26 ] as a termination mechanism for explaining the experimental results of [ 20 ] ; also see [ 19 ] .
we use the term restitution generally for relations that give a particular quantity , e.g. action potential duration or conduction time , as a function of the diastolic interval ( di ) .
the di ( usually measured in milliseconds , ms ) is the rest or recovery period for the cell which essentially starts with the end of the refractory period and ends when the cell is activated again . in its most basic form ,
the action potential duration or apd is the length of time ( usually measured in milliseconds , ms ) that a cell is active after excitation . for our purposes
, we may think of apd as a cell s effective refractory period ( erp ) during which no excitations , even strong ones , can elicit new action potentials .
the restitution of apd , which can be experimentally measured or derived from ionic models , is the most extensively studied of restitution relations [ 2 , 4 - 8 , 10 , 11 , 21 , 24 , 26 , 29 , 32 , 33 , 42 , 45 , 48 , 50 , 53 - 57 ] . any apd restitution function , whether experimentally measured or analytically derived can be used in this model . for the numerical simulations in this paper
, we use an apd restitution function that is fitted to the experimental data by koller , et al [ 33 ] where apd values were recorded in two types of patients , those with and those without structural heart disease ( shd ) .
the following is a possible fit to their averaged data displayed in figure 1 in [ 33 ] for shd patients : @xmath14 with parameter values [ c]|c|c|c|c|c|c|c|c|c|c|c|c|@xmath15 & @xmath16 & @xmath17 & @xmath18 & @xmath19 & @xmath20 & @xmath21 & @xmath22 & @xmath23 & @xmath24 & @xmath25 & @xmath26 + 350 & 157 & 8 & 20 & 1700 & 1200 & 0.0021 & 0.025 & 0.0004 & 80 & 136 & 82 + a graph of this function is shown in figure 2 . expressing the apd as a function of di in the above manner represents only a first order of approximation ( see section 4 below ) .
[ apdrc ] each cell conducts an action potential through it in a finite amount of time .
this time interval , usually measured in milliseconds , is referred to as the conduction time ( ct ) [ 4 , 26 , 50 ] .
experimentally measured ct restitution functions are not as readily available for human hearts as the apd restitution functions ; however , they may be readily derived from cv restitution functions through the relation@xmath27 where @xmath28 and @xmath29 are , respectively , the ct and the cv restitution functions .
this derivation is valid if as mentioned earlier , @xmath29 is essentially constant over the length of each cell ( or cell aggregate ) and thus no spatial dependence is required in @xmath30 for discussions of cv restitution functions see [ 2 , 7 , 8 , 11 , 12 , 14 - 16 , 20 , 22 , 53 , 54 , 56 ] . for our simulations we use the following expression @xmath31,\quad di>0 \label{ct}%\ ] ] with parameter values : [ c]|c|c|c|c|@xmath32 & @xmath33 & @xmath34 & @xmath35 + 0.1 cm & 0.07 cm / ms & 1 & 0.02 ms@xmath36 + these numbers are obtained using ( [ cv2ct ] ) and a reasonable fit to the human - rescaled , guinea pig data for cv from [ 22 , 54 ] .
the value of @xmath32 corresponds to 10 nominal cells .
the number @xmath33 in ( [ ct])is the maximum cv ; in more common ( though less uniform ) units , the fitted number in the above table would be 0.7 m / s . by changing the parameters in ( [ ct ] )
we may achieve conduction slow - downs in different ways , as shown in figure 3 .
[ ctrc ] and @xmath37 .
( a ) @xmath38 and @xmath39 ; ( b ) @xmath40 and @xmath39 ; and ( c ) @xmath41 and @xmath42.,width=288 ] in the absence of reentry , a pacer ( the sinus node or an ectopic focus ) drives the ventricular contraction cycle .
if reentry is initiated and interrupted then the pacer and the loop fall out of phase and it is necessary to carefully model their interaction . for this purpose
we define two rhythms or clocks : the pacer s rhythm which has a prescribed beat pattern ( fixed or variable ) and the system s rhythm which is reset when reentry is initiated or terminated .
these two rhythms are related via a pacer threshold ( seen the next section ) and a delay term that we now define .
the _ time of occurrence _ for the @xmath43-th beat ( following a fixed reference beat ) can be tracked and calculated using the following quantity , namely , the sum of usually variable cycle lengths ( after a di adjustment ) : @xmath44 let the time of occurrence of the @xmath45-th pacer beat be denoted by @xmath46 ( starting from the same reference point in time as that for @xmath47 ) .
the difference @xmath48 is the duration of each such beat . in our simulations @xmath49
is defined arbitrarily in order to facilitate our study of the sitr patterns .
if the pacer is the sinus node then @xmath49 is influenced by a variety of deterministic and stochastic factors , such as circadian , weekly and seasonal variations , lifestyles , drugs , the autonomic system and so on .
thus a single , all purpose formula for @xmath49 does not exist .
the two rhythms @xmath47 and @xmath46 fall out of phase if reentry occurs as noted earlier .
to model their interaction , we define the _ first _ pacer pulse that reaches cell 1 in beat @xmath43 as the _ least _ integer @xmath50 such that @xmath51 the variable delay term @xmath52 here represents a _ time lag _ that results from the blocking of the pacer by the retrograde reentrant wave propagating toward the pacer .
a number of things may affect @xmath52 ; these include the different types of tissue in which pulses propagate ( retrograde and antegrade ) , the electrotonic spread of current in conducting tissue between the pacer and the loop , and the effect of reentrant waves on the pacer s intrinsic beat rate .
for simplicity we do not consider the electrotonic currents and tissue heterogeneity .
we assume also that the pacer itself is protected in the sense that it s intrinsic beat rate is not affected by the reentrant activity .
therefore , we use a simple definition for @xmath52 as follows:@xmath53{ll}% \delta , & \text{if pulse } n-1\text { was reentrant}\\ di^{\ast } , & \text{otherwise } % \end{array } \right . , \ \delta
> di^{\ast}.\ ] ] if the loop is not in reentry , then @xmath52 may take the minimum value @xmath13 since cell 1 must have at least that much rest time after its erp before it can be reactivated .
high frequency reentrant waves typically disassociate a slow pacer such as the sinus node .
this situation is modeled here by choosing a sufficiently large value of the delay parameter @xmath54 so as to inhibit the pacer from interfering with the reentrant circulation .
small values for @xmath54 , which promote pacer interference , may be feasible in certain circumstances , e.g. when there are pathways from the pacer s site to the site of the loop that are protected from the fast reentrant waves by pathologies . to determine the index value @xmath50 we may generate the sequence @xmath46 independently of @xmath47 and for each @xmath43 mark the values @xmath50 that satisfy ( [ kn ] ) .
this process is carried out internally during each simulation run ; an analytical relationship is generally not easy to find .
however , if @xmath55 is constant then such a relationship is easy to find ; in this case we have @xmath56 with @xmath57
i.e. , the least integer that is greater than or equal to the ratio @xmath58 this simple relationship makes explicit the discontinuity that results from phase resetting in our model .
we distinguish between two primary dynamic modes for the loop - pacer system : the _ reentry mode _ where an action potential circulates in the loop by itself and the _ paced mode _ where the action potential in the loop comes from the pacer .
the system s mode in each beat changes on the basis of threshold relations to be defined below .
the modes are defined by systems of partial difference equations that track the circulating pulse in both space and time .
introductory material on ordinary and partial difference equations can be found in [ 9 , 17 , 31 , 40 , 51 ] . technically , the dynamical system consisting of the loop , the pacer and the associated thresholds is a polymodal structure in the sense of [ 51 ] .
let @xmath59 be the di of cell @xmath6 in beat @xmath60 define the _ firing indicator _ function as @xmath61{ll}% 1 , & \text{if } di_{i , n-1}>di^{\ast}\text { and } \phi_{i-1,n}=1\\ 0 , & \text{if } di_{i , n-1}\leq
di^{\ast}\text { or } \phi_{i-1,n}=0\text { } % \end{array } \right . , \ i=2,\ldots , m.\end{aligned}\ ] ] this quantity specifies conditions under which the @xmath6-th cell fires an action potential in beat @xmath43 ; these conditions simply require that ( a ) the preceding cell @xmath62 fired and ( b ) the di of cell @xmath6 in the preceding beat @xmath63 exceeded the minimum value @xmath13 as is required for propagation .
we assume that cell 1 fires in every beat whether by a reentrant pulse or by a pulse from the pacer .
let @xmath64 denote the action potential duration of cell @xmath6 in beat @xmath43 so that if the @xmath6-th cell does not fire in beat @xmath43 then @xmath65 ( otherwise , @xmath66 ) .
the dynamic _ activation duration _ is defined as @xmath67 where @xmath68 is the apd restitution function for cell @xmath6 .
note that @xmath69 for cell 1 in every beat . in the following sections ( except in case - study 4 )
we assume that the loop is apd homogeneous so that @xmath70 for all @xmath71 where @xmath72 is given by ( [ apd - shd ] ) . in this mode
the pacer determines the di values @xmath59 rather than a reentrant pulse in the primary loop ; the circulation is assumed unidirectional owing to an active ub .
we use the following system of @xmath5 equations : [ udc11]@xmath73 here @xmath74 is the ct restitution function for cell @xmath75 the above system of equations determines the di values using the first pacer beat after time instance @xmath47 ( beat @xmath43 in progress ) .
figure 4 illustrates equations ( [ udc4 ] ) and ( [ re2 ] ) .
the first five terms in ( [ udc4 ] ) simply express in mathematical terms what we might state less precisely using the english language .
they are the same for tissue fibers as they are for loops ; see , e.g. [ 68 ] .
the last two terms in the definition of @xmath59 in ( [ udc4 ] ) allow for proper adjustment of the di value ( without double counting ) when cell @xmath6 does not fire in beat @xmath43 ( see comments on the reentry mode below ) . in typical loop models in prior literature
all cells fire so these two terms would drop out . cells ; a is apd , d is di and ct is the conduction time .
the sizes of a , di and ct are selected arbitrarily for clarity of illustration.,width=336 ] in this mode the reentrant loop drives the ventricles ectopically .
thus @xmath76 is the difference between the apd of cell @xmath7 and the total conduction time for unidirectional circulation once around the ring , i.e. [ re0]@xmath77 it is not hard to show that the system of @xmath5 equations ( [ re0 ] ) is equivalent to the coupled map lattice equations in [ 51 ] when @xmath78 for all @xmath6 and @xmath60 equations ( [ re0 ] ) are not valid if some of the di values fall below the threshold @xmath13 or if a stray pacer pulse gets into the loop ( i.e. if the right hand side of ( [ re1 ] ) exceeds the right hand side of ( [ udc1 ] ) ) .
the following threshold relations ensure a consistent scenario .
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ tc : the _ circulation threshold _ ( or _ head - tail _ or _ conduction block _ _ threshold _ ) _ _ is based on the following set of inequalities : @xmath79 they ensure that the di values @xmath59 in equations ( [ re0 ] ) exceed @xmath80 as is necessary for the generation of action potentials .
if any of the above inequalities fails in a cell @xmath6 then a wavefront reaches its own tail in that particular cell and circulation stops until the arrival of the next pacer pulse . in this case
we say that tc fails in cell @xmath75 _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ tp : the _ pacer _ _ threshold _ ( or _ phase resetting _ _ threshold _ ) is defined by the following inequality:@xmath81 this inequality ensures that the reentrant pulse ( rather than a stray one from the pacer ) reactivates cell 1 .
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ * mode switching criteria*. _ equations ( [ re0 ] ) are used if tp holds and also tc holds in cell 1 .
if for some cell _ @xmath82 _ tc fails , then _ ( [ re2 ] ) _ can still be used to find the di values for beat n , but we switch to equations ( [ udc11 ] ) for beat _
@xmath83__. if tc fails for _ _ @xmath84 _ or if tp fails then we use ( [ udc11 ] ) for beat n. _ in the absence of unidirectional block the pacer sends two pulses in the loop with cell 1 activating both cell 2 and cell @xmath5 .
these pulses propagate in opposite directions to eventually annihilate each other in some cell in the loop .
transitions into and out of this bidirectional circulation mode are governed by tc , tp and a ub threshold ( tub ) ; see section 4 below .
we assume for simplicity that tub never fails so that the bdc mode need not be discussed in this paper .
in this section we present the results of numerical simulations of our threshold model along with some analytical observations . due to
the large number of different types of behavior that can occur by changing one or more of the many variables involved , providing a comprehensive list of essentially different patterns is not feasible here .
we therefore present our main results in the form of a few model case - studies and for additional results we refer the interested reader to our web site : _ www.people.vcu.edu / hsedagha / sitr_. each case - study below consists of one or more runs , i.e. sets of iterations of the propagation equations .
each run is based on a fixed set of parameter values and we call each iteration in a given run a beat . for each run
we need to specify @xmath5 initial di values @xmath85 .
this specification of di values is a technical representation of the occurrence of a premature stimulation in our model .
computational accuracy is not a major issue in the qualitative studies in this paper .
however , calculating the values of the various nonlinear functions ( e.g. the exponentials in the apd and cv restitution functions ) generate round - off errors that can accumulate in certain cases ( e.g. runs in case - study 4 ) .
therefore , results obtained by different programs , codes or levels of precision ( single or double ) may be different from some of those that we have presented , although the qualitative features are usually retained .
the following conventions facilitate the presentation of results in this section : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ for a loop consisting of @xmath5 cells if a reentrant pulse is blocked at cell @xmath86 in a particular beat , then we define the _ reentry value _ of that beat as the ratio @xmath87 this ratio will be abbreviated as @xmath88.$ ] if a reentrant pulse completes a turn around the loop then the reentry value is 1 .
if a reentrant pulse is blocked in cell @xmath86 after @xmath89 complete turns around the loop then the reentry value is @xmath90.$ ] if a beat occurs in the paced mode ( i.e. one of the thresholds tc or tp fails ) then we assign it the value @xmath91 thus @xmath89 consecutive paced beats appear as @xmath92 .
if the pattern is locked in the reentry mode , we use the symbol @xmath93 if it is locked in the paced mode , we use @xmath94 _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ adding up consecutive reentry mode beats and consecutive paced mode beats gives a sequence of numbers consisting of integers whose sign tell us the mode of system in various sets of beats
. examples of mode sequences and more details on them are found in the case - studies below .
we call the loop - pacer system _ bistable _ if there are two or more coexisting stable di configurations or state vectors , each of which can be reached from a particular region of the @xmath5-dimensional state space .
if there are more than two distinct , coexisting stable states then the system is _
multistable_. bistability may be attributed to the non - concavity of the apd restitution function [ 50 ] .
the apd function ( [ apd - shd ] ) whose graph has bumps and twists is also non - concave .
proper ct restitution parameters are required to realize bistability ; i.e. the ct and apd restitution parameters must be properly matched .
because the length @xmath4 of the loop affects the conduction time through it , bistability may emerge or fade as @xmath4 is changed with other ct parameters fixed [ 50 ] .
a bistable regime can affect sitr patterns by causing the violation of the circulation threshold tc without any changes in the apd or ct parameters , or any changes in the pacing rate .
this situation occurs if the bistable regime satisfies the following _ threshold bistability _ condition : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ t - bistability _ : there are two distinct stable states in the reentry mode : in one state the di values cross @xmath13 and cause the failure of the circulation threshold tc at some cell within the loop , but in the other state the di values are always greater than @xmath13 .
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ an example of t - bistability is shown in figure 7 below ; other examples of t - bistability occur in the next section ( case - study 1 ) .
the @xmath5-dimensional state space of a t - bistable system is partitioned into two regions or basins of attraction . as a sitr pattern evolves with the motion of the di state vector ,
mode changes occur if the state vector crosses over into the basin of attraction of a different stable regime . in this section
we establish that t - bistability lends a measure of unpredictability to sitr patterns that is not due to phase resetting disruptions by the pacer ( i.e. tp failures ) .
for instance , comparing the mode sequences in runs 1.2 , 1.4 and 1.6 below we see that the initial few bursts of fast , reentrant beats provide no obvious clues about the eventual modes that the patterns lock into .
consider a homogeneous loop with the apd and ct restitutions given by ( [ apd - shd ] ) and ( [ ct ] ) .
we also set the length @xmath4 and other parameters as : [ c]|c|c|c|c|c|@xmath4 & @xmath5 & @xmath13 & @xmath54 & @xmath95 + 12.5 cm & 125 & 15.3 ms & 120 ms & 320 ms + where @xmath95 is the fixed cycle length of the pacer ; it is chosen small in this case - study to facilitate re - initiations of reentry .
the value of @xmath54 is set high enough to inhibit pacer interference ( tp never fails ) .
the following table summarizes the results of simulations ; it is followed by a series of observations and elaborations .
[ c]|c|c|c|run no .
& @xmath96 & sitr pattern mode sequence + 1.1 & 200 ms & @xmath97 + 1.2 & 100 ms & @xmath98,-2,\infty\right\ } $ ] + 1.3 & 80 ms & @xmath99 + 1.4 & 70 ms & @xmath100,-\infty\right\ } $ ] + 1.5 & 60 ms & @xmath101 + 1.6 & 50 ms & @xmath102,-2,\infty\right\ } $ ] + \(i ) in run 1.1 the sitr pattern immediately locks into a regular form of 2 paced beats followed by 2 reentrant beats . in runs 1.2 , 1.3 , 1.5 and 1.6 the sitr pattern locks into the reentry mode . in run 1.4
the paced mode is locked into .
note that the only thing that is changing from one run to the next is the initial di values .
\(ii ) in runs 1.2 , 1.3 , 1.5 and 1.6 the di values in the sustained or locked reentry modes converge to a fixed number and the eventual values of the fundamental parameters are [ c]|c|c|c|di & apd & cycle length cl = apd+di + 59.4 ms & 173.6 ms & 233 ms + the fixed , limiting di value 59.4 represents a stable _ convergent state_. this equilibrium di is locally stable ( attracting ) because using the restitution functions ( [ apd - shd ] ) and ( [ ct ] ) we compute [ 4 , 26]@xmath103 \(iii ) in runs 1.2 , 1.4 and 1.6 all the spontaneously terminated reentry bursts have oscillating ( quasiperiodic ) di .
thus apd and cl also oscillate similarly .
the oscillatory state is stable and eventually crosses @xmath104 ms .
hence there is t - bistability ( the oscillatory one and the convergent one ) which is responsible for mode changes in each run .
for example , in run 1.2 the 2nd initiation of reentry ( the 32-beat burst ) starts with the di state - vector in the basin of attraction of an oscillatory state which expands enough for the di values to cross @xmath13 ( tc fails ) ; see figure 5 .
this terminates the second reentry burst at cell 73 in the 41st beat .
this is the same type of termination mechanism as that in [ 19 , 20 , 26 ] ; this type of oscillations in di are also seen in [ 56 ] where they are related to the t wave alternans in ecg recordings ; also see [ 35 ] in this regard . with the 3rd initiation of reentry ,
the di state - vector falls in the basin of attraction of the convergent state which then effectively traps the di trajectory and locks the sitr pattern into the reentry mode ( here then , simplicity of behavior is not a good thing ) .
[ run1 - 2 ] .
amplification of oscillations in di leads to termination of reentry in beat 41 , cell 73 because @xmath105 . with beat 44
reentry is re - initiated and the system locks into the stable convergent state ( after some transient oscillations).,width=432 ] \(iv ) examination of the simulation data for run 1.4 shows a recurrent failing of tc in cell 1 after the last burst of 23 reentrant beats .
for all sufficiently large values of @xmath43 , [ c]|c|c|c|@xmath76 & @xmath106 & @xmath107 + 96.6 & 223.4 & 320 + the number 96.6 is the equilibrium di for the paced mode with pacing period @xmath108 .
the paced - mode equilibrium happens to be stable because in the absence of reentry the paced mode is governed by the one - dimensional difference equation@xmath109 with @xmath110 i.e. @xmath111 why the transition to the stable paced equilibrium occurs in this run and not others is unclear .
figure 6 shows the changes in di values in this run .
[ run1 - 4 ] .
the four interruptions of reentry due to conduction block ( tc fails ) are seen as brief jumps to high di .
the last interruption terminates reentry and locks the system in the paced mode.,width=432 ] \(iv ) in run 1.5 , the given initial di values put the state vector in the basin of attraction of the convergent state of the system .
this is evidently not the case in run 1.6 with even shorter di . in run 1.6 , as in run 1.4 , the initial mode is oscillating - di reentry .
we now consider some longer loops than in case - study 1 for comparison . in these loops
bistability is present but does not satisfy the t - bistability condition .
a smaller variety of different sitr patterns are obtained ( with all other system parameters having the same values as in case - study 1 ) .
we consider two different lengths : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ @xmath112 cm ( @xmath113 ) where stable oscillating states exist all of which fail tc , but the convergent mode is absent since the di fixed point of about @xmath114 ms is unstable [ 4 , 26]@xmath115 @xmath116 cm ( @xmath117 ) where all of the stable oscillating states ( there is at least one ) satisfy tc .
the convergent mode is again absent for the same reason as above:@xmath118 _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ to save space let us abbreviate repetitions or locked mode patterns with bars:@xmath119 the following table summarizes the results of simulations ; it is followed by a series of observations and elaborations .
[ c]|c|c|c|c|run no .
& @xmath85 & @xmath112 cm & @xmath116 cm + 2.1 & 200 ms & @xmath120 & @xmath121 + 2.2 & 100 ms & @xmath122,\overline{-2,2}\right\ } $ ] & @xmath123 + 2.3 & 80 ms & @xmath124 & @xmath125 + 2.4 & 70 ms & @xmath126,-2,17[79],\overline{-2,5[79]}\right\ } $ ] & @xmath127 + 2.5 & 60 ms & @xmath128 & @xmath129 + 2.6 & 50 ms & @xmath130,\overline{-2,2}\right\ } $ ] & @xmath131 + \(i ) the sitr patterns for the 14 cm loop are expected to lock into the reentry mode since the stable states do not fail tc to cause terminations .
the exception is run 2.6 where possibly owing to the greater length of the loop , a pulse from the pacer has time to activate cell 1 and block the premature simulation from reentering the loop ( threshold tp fails ) .
it is not clear why reentry does not re - initiate as in run 2.5 .
\(ii ) in the case of the 13 cm loop the repeated terminations can be attributed to the fact that the stable states all fail tc ; thus the patterns can not lock into the reentry mode .
\(iii ) for both of these loops , the eventual form of the sitr pattern is more predictable than the t - bistable loop of case - study 1 . on the other hand ,
the transient bursts of fast beats still seem difficult to predict .
* run 3.1 * ( _ more t - bistability _ ) .
this run presents a complex sitr pattern that locks into a long cycle of several initiations and terminations .
make the following changes to parameters in run 1.2 : [ c]|c|c|@xmath34 & @xmath95 + 1.3 & 315 ms + note that the increase in @xmath34 from 1 to 1.3 in this run causes a slight elevation of the ct restitution curve ( i.e. , a conduction slow - down ) at low di values ; see figure 3 .
\(i ) in this run , after a transient period of spontaneous initiations and terminations , a long 62-beat pattern emerges that is repeated ; i.e. the sitr pattern locks into a 62-beat cycle containing several bursts of fast reentrant beats .
the mode sequence for this run is as follows , with the pattern between starred numbers repeating : @xmath132,-1,[35],-2,4,-2,7[24],-1,[18],-2,8[18],^{\ast}% -1,[12],-2,7[12],-1,[12],-2,\\ & 10[49],-2,5[49],-2,4,-2,7[22],-1,[19],-2,7[19],^{\ast}% -1,[12],-2,7[12],-1,[12],-2,\ldots\}\end{aligned}\ ] ] \(ii ) the di equilibrium of approximately @xmath133 ms is unstable in this run since@xmath134 this instability is caused by the greater value of @xmath34 and results in the absence of the convergent state .
nevertheless , there is t - bistability where one of the stable oscillating states does not fail tc but another oscillating state does . in the first oscillatory state , di values come very close to @xmath13 ; see figure 7 .
this proximity increases the sensitivity to transient effects of mode changes and makes locking into the reentry mode an improbable event .
[ run3 - 1 ] for all @xmath6 , the graph crosses the dashed horizontal line @xmath135 ; for the sake of illustrating the quasiperiodic nature of oscillations in reentry mode , the graph ignores tc since reentry would terminate with the first such crossing .
( b ) @xmath136 for all @xmath6 .
reentry is sustained with quasi - periodic variation of di.,width=528 ] * run 3.2 * ( _ pacer action _ ) . in case - studies 1 and 2
we see that the convergent reentry pattern is sticky i.e. once this pattern is attained , reentry does not self - terminate by the circulation mechanism in the loop .
however , pacer interaction with the loop can lead to termination in some cases by changing the convergent pattern to a terminating oscillatory one in a t - bistable case .
all parameters are as in run 1.5 ( hence there is t - bistability ) except that now : [ c]|c|c|@xmath54 & @xmath95 + 45 ms & 800 ms + we have set the pacing period at the nominal sinus length to eliminate the effects of fast pacing .
the smaller value of @xmath54 promotes interference by the pacer .
\(i ) the mode sequence for this run , which may be compared with run 1.5 , is:@xmath137
\(ii ) reentry is initiated with beat 1 by a premature stimulation , and the di values begin to approach the limiting value 59.4 ms as in run 1.5
. however , tp fails in beat 41 ; a stray pulse from the pacer enters the loop and changes the reentry di pattern from the convergent type to an oscillating one ; see figure 8 .
the quantities @xmath138 , @xmath139 in figure 8 are the right hand sides of equations ( [ udc1 ] ) and ( [ re1 ] ) , respectively .
[ run4 - 1 ] ( see text ) .
the negative value in beat 41 ( indicated by the arrow ) corresponds to termination of reentry due to failure of tp .
( b ) di values in beats 30 through 60 ; note the phase resetting that occurs in beat 41 and changes the reentrant circulation pattern from convergent to oscillatory in this t - bistable case.,width=528 ] the oscillating di pattern crosses the @xmath13 threshold ( tc fails ) in beat 55 and reentry terminates in this beat . when the pacing period is long ( say , within the nominal sinus range of 800@xmath140 ms ) and the loop is homogeneous , _ re - initiations _ of reentry
do not occur with the above ct and apd parameters in the absence of premature stimulations or ectopic sources .
the low ct ( or high cv ) in the loop blocks reentry when pacing is slow .
nevertheless , the rather common occurrences of ventricular tachycardia at low pacing rates in patients [ 58 ] motivate us to study this important special case in the context of a heterogeneous loop with a variable - rate pacer .
suppose that the loop consists of two patches of cells .
patch 1 consists of cells 1 through @xmath86 and patch 2 contains cells @xmath141 and beyond where of course , @xmath86 is a positive integer less than @xmath8 the total number of cells .
the cells in patch 1 may have different apd and/or ct restitution parameters than those in patch 2 . for simulations
, we also assume that the pacer has a variable - rate within the aforementioned nominal sinus range .
assume that the pacer s oscillation period varies in a sinusoidal fashion:@xmath142 this beat pattern may represent an idealization ( without random effects ) of a single cycle of a complex , multi - pattern stretch of pacer beats .
more complex variations may be considered in later studies if needed .
the number @xmath143 is the period of a full cycle of variable - rate oscillations , which we refer to as a _ full pacing cycle _ ( fpc ) .
the halfway point of the fpc , or the bottom of the fpc well occurs at @xmath144 at this point , the pacer oscillates with minimum period ( fastest beat rate ) . on the other hand ,
when @xmath145 or @xmath146 the pacer has its largest oscillation period ( slowest beat rate ) .
the number @xmath147 is the fpc amplitude since it modulates the pacer s oscillation period within the fpc . if @xmath148 then the period is a fixed @xmath95 as in the preceding case - studies .
* run 4.1*. we fix the apd and ct parameters for patch 2 to be the same as those used in case - study 1 .
patch 1 apd and ct have the same parameters as patch 2 except for the following : [ c]|c|c|c|@xmath15 & @xmath16 & @xmath33 + 300 ms & 170 ms & 0.05 cm / ms + thus apd is reduced ( by a shift in the restitution function ) in patch 1 ; also the cells in this patch conduct more slowly with a lower value of @xmath33 ( the maximum cv ) .
additional loop and pacer parameters are set as follows : [ c]|c|c|c|c|c|c|c|@xmath4 & @xmath95 & @xmath147 & @xmath143 & @xmath86 & @xmath54 & @xmath149 + 13.5 cm & 800 ms & 350 ms & 1000 & 50 & 140 ms & 600 ms + the integer @xmath86 gives the number of cells in patch 1 and the three pacing parameters @xmath150 and @xmath147 define a variable pacing protocol with fpc period of 1000 beats and amplitude 350 about a fixed , nominal sinus trend of 800 ms .
\(i ) the mode sequence of the sitr pattern obtained is@xmath151,-1,[51],-1,[51],-1,[51],-1,[51],-2,\infty\}.\ ] ] \(ii ) the mode sequence exhibits a pattern of 5 incomplete reentry initiations followed by a pacer beat ( or 2 in one case ) .
each incomplete reentry start which is blocked at the patch junction ( cell 51 ) corresponds to a short cycle length ( about 224 ms ) compared to the pacer s period of approximately 460 ms near the middle of the fpc .
hence the pattern of 5 alternating reentry / pacer beats resembles a series of premature ventricular contractions ( pvc ) in a bigeminal form before reentry finally takes hold for good .
\(iii ) we note that if @xmath152 or fewer , then no reentry is initiated in this run . on the other hand , larger @xmath86 than 50
sustain reentry more quickly by increasing the percentage of slow cells in the loop .
* run 4.2*. in the preceding run the pacer did not interfere with the reentrant circulation ( tp did not fail ) because @xmath153 ms was large enough to block the pacer .
we now show that decreasing @xmath54 to promote pacer interactions may cause the eventual termination of reentry in the preceding run .
we make only one change in this run relative to run 4.1 by setting @xmath154 ms .
\(i ) the mode sequence of the resulting sitr pattern is as follows:@xmath155,-1,\ldots,[51],-1}% } , [ 51],-2,1,-1,11,-3,1,-1,4,-1,6,-3,1,-1,\\ & 4,-1,6,-3,1,-1,11,-1,6,-3,\underset{\text{38 times}}{\underbrace { [ 51],-1,\ldots,[51],-1}},[51],-374,\ldots\}\end{aligned}\ ] ] after several pvc s and reentrant bursts of different durations , the pattern locks into the paced mode until reentry is initiated in the next fpc .
this mode sequence is sensitive to computer round - off errors ; however , the varied nature of mode changes persists .
figure 9 gives a graphical representation of mode changes for this run .
\(ii ) the first deviation from the mode sequence of run 4.1 occurs with a tp failure that interrupts reentry after just one full cycle .
some of the @xmath156 entries in the mode sequence above ( mostly the ones _ not _ associated with conduction blocks in cell 51 ) indicate failures of tp also .
in this section we present additional supporting details about the model .
experimental measurements and other studies reveal the difficulty of fitting a single curve to the data , particularly at the lower end of the di scale . to address this and related issues ,
refinements have been considered in the forms of memory and latency .
beat to beat memory has been studied in [ 13 , 19 , 21 , 23 , 28 , 46 , 56 ] and has been modeled in a number of ways . in many cases ,
modeling memory can be accomplished by using di values from several earlier stimulations .
latency ( roughly a brief period between the arrival of a pulse and the firing of the cell ) has also been expressed as a restitution function of di [ 12 , 13 ] .
if this effect is added to apd we obtain a composite restitution function that may no longer be monotonic for small di values . for simplicity , we ignore both memory and latency in this paper ; however , they can ( and need to ) be included in the model for greater refinement in future studies . the apd restitution function may be measured experimentally using two different pacing protocols : dynamic or standard ( s1-s2 ) [ 29 , 32 ] .
these protocols generally produce apd curves that are shifted or have different slopes . for a discussion of this issue regarding the apd restitution that we use in this paper see [ 33 ] .
the fit in ( [ apd - shd ] ) is to the dynamic restitution data in [ 33 ] .
the apd values ( and thus the restitution function ) may be affected by the steady fast pacing of the ventricles [ 34 ] or of the atria [ 39 ] .
the implied shift in the apd restitution can have a significant effect on the sitr patterns in certain borderline cases .
for simplicity we do not consider such effects in this paper ; however , a suitable modification of the apd restitution function that incorporates explicit time dependence can accommodate dynamically - induced shifts . experimental evidence is offered in [ 53 ] that the apd restitution curve may be non - monotonic for small di values .
working with a apd restitution curve having a single local minimum at low di , the same study demonstrates the occurrence of spatial bifurcations along the length of a periodically paced fiber of cardiac tissue .
although not involving a loop structure , the work in [ 53 ] uses restitution functions in a discrete model .
a localized form of non - monotonicity ( like a bump or dent ) in apd may occur at any di value and has interesting consequences for additional complexity in reentry mode . in [ 48 ] where a continuous model of the loop as a cable is used , this type of non - monotonicity is shown to be responsible for the appearance of spatio - temporal chaos in apd and other key variables over the length of the loop .
the occurrence of spatio - temporal chaos on the loop may raise the level of unpredictability in sitr if the amplitudes of oscillations in di are large enough to let the wavefront reach its tail ( thus causing a threshold failure ) .
various anatomical and functional mechanisms for ub in a loop are discussed in the model study [ 49 ] .
these mechanisms include a properly timed premature stimulus delivered to a suitable location in the loop , inhomogeneities in the degree of cellular uncoupling and gap junction resistance , in membrane excitability and in fiber cross sectional area .
the discussion in [ 49 ] points to time , space and voltage windows of vulnerability for the occurrence of unidirectional block and initiation of reentry . of these
, the time window is seen to be the easiest to measure and its value is related to the space and voltage windows via standard mathematical formulas .
each cell requires a certain amount of depolarizing current to fire an action potential .
this _ activation threshold _ can be represented by a restitution function [ 27 ] or equivalently , the strength - interval curve [ 13 ] shifted to the left so that the origin represents the end of the erp . in figure 10 this curve is shown as a decreasing restitution function @xmath157 . in principle , at any time following the end of the erp a sufficiently large current will be able to elicit an action potential .
however , in the absence of external electrical shocks , there is a limited amount of current @xmath158 available to depolarize the next cell .
if a restitution of @xmath158 is given then the intersection point of @xmath157 and @xmath158 is the @xmath159 threshold or cut - off value @xmath13 in the sense that if @xmath160 then action potential is fired and propagation occurs .
the time interval @xmath13 is similar in nature to the chronaxie [ 47 ] although the definition of @xmath13 requires another restitution curve in addition to @xmath161 a monotonically increasing restitution function for @xmath158 might be appropriate ; see figure 10 .
[ currents ] and @xmath162 . in this figure , @xmath163 ( see text).,width=288 ] for our purposes here we do not need explicit formulas for either @xmath157 or @xmath164 the value @xmath13 may be chosen arbitrarily or estimated experimentally without reference to its potential sources .
values as high as 160 ms were used in [ 26 ] . in human cases the value of @xmath13 in vivo is likely to be in a much smaller range .
we use the conservatively small value of about 15 ms in this study .
an analog of the time window in [ 49 ] can be defined here using the restitution functions @xmath157 and @xmath158 .
we define two sets of currents restitution functions for cell 1 corresponding to the two possible directions in the loop : @xmath165 , @xmath166 and @xmath167 , @xmath168 for the left and the right directions .
the asymmetry creates a _ ub window _ in cell 1 as follows : let @xmath169 and @xmath170 with at least one of these inequalities strict .
let @xmath13 be the ( unique ) intersection point of @xmath165 and @xmath167 and let @xmath162 to be the intersection point of @xmath166 and @xmath171 then @xmath172 and the interval between these two values is where the ub occurs . for
if @xmath76 represents the di of cell 1 in a cycle or beat @xmath173 then the conditions @xmath174 imply that propagation is possible in one direction because the threshold marked by @xmath13 is crossed but it is inhibited in the other direction where the @xmath162 threshold is not crossed .
we emphasize that the unidirectional block as defined here is functional and may not occur even if the ub window exists ( for anatomical or functional reasons ) .
if the number @xmath76 does not enter the the ub window ( i.e. the interval from @xmath13 to @xmath162 ) in a particular beat @xmath43 then the ub does not materialize .
this option adds another mode to the loop - pacer system , namely the bdc mode mentioned previously .
in addition to tc and tp , the following threshold is associated wtih the bdc mode : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ tub : @xmath175 with beat @xmath63 in the paced mode . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ this inequality defines the _ ub threshold _ to ensure that cell 1 conducts unidirectionally .
if beat @xmath63 is in reentry mode then conduction in beat @xmath43 is automatically unidirectional whether tub holds or not , because the pulse is transmitted from a still active cell @xmath12 to limit the variety of cases to consider , we assume in this paper that @xmath162 is so large that the unidirectional block is always present in cell 1 ( permanent ub ) so tub always holds .
in the preceding sections , we developed an interactive loop - pacer threshold model that is a part of a familiar cardiac anomaly . using experimental data on restitution parameters from the existing literature for action potential duration and conduction time
, we obtained various conditions and parameter values that would cause irregular patterns of spontaneous initiations and terminations of reentry .
a major cause of unpredictable behavior is threshold bistability where one stable regime causes a failure of the circulation threshold tc and another does not . in this situation
the initiation and termination of reentry depends on the location of the di state vector in the @xmath5-dimensional state space , where the nonlinearity in propagation equations makes it difficult to track the state vector .
thus , it is quite difficult to forecast the long - term behavior of an evolving sitr pattern for a t - bistable loop .
another potential source of unpredictability is the interference by the pacer .
if reentry occurs , then the pacer and the reentrant circuit are thrown out of phase so a phase resetting pacer pulse may unexpectedly end the reentrant circulation .
a number of prior studies of a different nature from ours also find complex sitr patterns in simple settings , such as in sheets [ 3 , 25 ] , loops [ 44 ] and aggregates [ 35 ] of animal cardiac tissue . in [ 3 ] and
[ 25 ] complex patterns in monolayers of growing cell cultures are linked to changing densities and to local inhomogeneities , respectively . in [ 35 ]
bursting beat patterns are studied in cell aggregates under external pacing .
closer to the subject matter of this paper , in [ 44 ] a ring - shaped lab preparation is considered with two localized pacemakers .
threshold relations for handling the pacers in this loop appear in the mathematical discussion . in a different class of studies , namely , the modulated parasystole , a permanent ectopic rhythm coexists with the sinus or another competing rhythm . in such relatively simple systems ,
complex beat patterns arise that are highly sensitive to parameter changes [ 23 , 28 , 43 ] . like the tp threshold in our model ,
a mathematical relation with a discontinuity is responsible for the complexity of temporal patterns in parasystole ; see [ 23 , 43 ] .
we omitted many features from our discussion here that one might include in an enhanced version of this model .
for example , ub may be treated functionally by refining the ub time window and considering bidirectional propagation in the loop , as noted above .
other enhancements to this model might include adding memory and latency to the apd restitution function .
the inclusion of memory is especially important in clarifying the dynamic significance of delayed responses ( not to mention improved fits to the apd restitution data ) .
additional directions for exploring the dynamics of the loop - pacer system include consideration of multiple exit / entry points into the loop that can affect reentry , a systematic approach to heterogeneity and also some theory to better model the time lag @xmath52 that influences the loop - pacer interactions
. a somewhat different direction that may be explored would be to incorporate a programmed shock routine to artificially end long bursts of fast , reentrant beats . with some of the above enhancements the sitr patterns will likely become more complex and less predictable , though it is also possible that some of these enhancements ( e.g. memory ) can have a moderating effect .
understanding the precise mechanisms that generate such complex sitr patterns may in turn offer new insights into the nature and causes of long - term temporal patterns of tachyarrhythmia occurrences . [ 1 ] anastasiou - nana , m.i .
menlove , j.n .
nanas and j.l .
anderson , _ changes in spontaneous variability of ventricular ectopic activity as a function of time in patients with chronic arrhythmias _ , circulation , 78 ( 1988 ) , 286 - 295 . [ 2 ] banville , i. and r.a .
gray , _ effects of action potential duration and conduction velocity restitution on alternans and the stability of arrhythmias _ , j. cardiovascular electrophysiol .
, 13 ( 2002 ) , 1141 - 1149 . [ 7 ] cao , j - m . , z. qu , y - h .
kim , t - j .
wu , a. garfinkel , j.n .
weiss , h.s .
karagueuzian and p - s .
chen , _ spatiotemporal heterogeneity in the induction of ventricular fibrillation by rapid pacing : importance of cardiac restitution properties _ , circ .
research , 84 ( 1999 ) , 1318 - 1331 .
[ 10 ] cherry , e.m . and f.h .
fenton , _ suppression of alternans and conduction blocks despite steep apd restitution : electrotonic , memory , and conduction velocity restitution effects _ , am .
j. physiol .
286 ( 2004 ) , h2332-h2341 .
[ 12 ] derksen , r. , h.v.m .
van rijen , r. wilders , s. tasseron , r.n.w .
hauer , w.l.c .
rutten and j.m.t .
de bakker , _ tissue discontinuities affect conduction velocity restitution : a mechanism by which structural barriers may promote wave break _ , circulation , 108 ( 2003 ) , 882 - 888 . [
16 ] courtemanche , m. and a. vinet , _ reentry in excitable media _ , in : a. beuter , l. glass , m.c .
mackey and m.s .
titcombe ( ed.s ) , _ nonlinear dynamics in physiology and medicine _ , chapter 7 , springer , new york , 2003 .
[ 20 ] frame , l.h . and m.b .
simson , _ oscillations of conduction , action potential duration and refractoriness : a mechanism for spontaneous termination of reentrant tachycardias _ , circulation , 78 ( 1988 ) , 2182 - 2185 . [ 21 ] gilmour , r.f . , m.a . watanabe and n.f .
otani , _ restitution properties and dynamics of reentry _ , in : _ cardiac electrophysiology : from cell to bedside _
, 3rd ed . ,
saunders , london , 1999 , 378 - 385 . [ 22 ] girouard , s.d .
pastore , k.r .
laurita , k.w .
gregory and d.s .
rosenbaum , _ optical mapping in a new guinea pig model of ventricular tachycardia reveals mechanisms for multiple wavelengths in a single reentrant circuit _ , circulation , 93 ( 1996 ) , 603 - 613 . [ 29 ] kalb , s.s .
dobrovolny , e.g. tolkacheva , s.f .
idriss , w. krassowska and d.j .
gauthier , _ the restitution portrait : a new method for investigating rate - dependent restitution _
, j. cardiovascular electrophysiol . ,
15 ( 2004 ) , 698 - 709 .
[ 33 ] koller , m.l .
maier , a.r .
gelzer , w.r .
bauer , m. meesman and r.f .
gilmour , _ altered dynamics of action potential restitution and alternans in humans with structural heart disease _ , circulation , 112 ( 2005 ) , 1542 - 1548 . [ 34 ] krebs , m.e .
szwed , t. shinn , w.m .
miles and d.p .
zipes , _ short - term rapid ventricular pacing prolongs ventricular refractoriness in patients _ , j. cardiovascular electrophysiol .
, 9 ( 1998 ) , 1036 - 1042 .
[ 36 ] lampert , r. , l. rosenfeld , w. batsford , f. lee and c. mcpherson , _ circadian variation of sustained ventricular tachycardia in patients with coronary artery disease and implantable cardioverter defibrillators _ , circulation , 90 ( 1994 ) , 241 - 247 .
[ 37 ] liebovitch , l.s .
todorov , m. zochowski , d. scheurle , l. colgin , m.a .
wood , k.a .
ellenbogen , j.m .
herre and r.c .
bernstein , _ nonlinear properties of cardiac rhythm abnormalities _ , phys . rev .
e , 59 ( 1999 ) , 3312 - 3319 . [ 39 ] manios , e.g. , e.m .
kallergis , e.m .
kanoupakis , h.e .
mavrakis , h.k .
mouloudi , n.k .
klapsinos and p.e .
vardas , _ effects of successful cardioversion of persistent atrial fibrillation on right ventricular refractoriness and repolarization _ , europace , 7 ( 2005 ) , 34 - 39 .
[ 54 ] ten tusscher , k.h.w.j .
, o. bernus , r. hren and a.v .
panfilov , _ comparison of electrophysiological models for human ventricular cells and tissues _ , prog .
molecular biology , 90 ( 2006 ) , 326 - 345 .
[ 60 ] wood , m.a . ,
p.m. simpson , w.b .
london , b.s .
stambler , j.m .
herre , r.c .
bernstein , k.a .
ellenbogen , _ circadian patterns of ventricular tachycardia in patients with implantable cardioverter defibrillators _ , j. am . coll .
, 25 ( 1995 ) , 901 - 907 . [ 61 ] wood , m.a . , p.m. simpson , l.s .
liebovitch , a.t .
todorov and k.a .
ellenbogen , _ temporal patterns of ventricular tachyarrhythmias : insights from the implantable cardioverter - defibrillator _ , in : s.b .
dunbar , k.a .
ellenbogen and a.e .
epstein ( ed.s ) , _ sudden cardiac death : past , present and future _ , futura pub .
co. , armonk , ny , 1997 . | a two - component model is developed consisting of a discrete loop of cardiac cells that circulates action potentials as well as a pacing mechanism .
physiological properties of cells such as restitutions of refractoriness and of conduction velocity are given via experimentally measured functions .
the dynamics of circulating pulses and their interactions with the pacer are regulated by two threshold relations .
patterns of spontaneous initiations and terminations of reentry ( sitr ) generated by this system are studied through numerical simulations and analytical observations .
these patterns can be regular or irregular ; causes of irregularities are identified as the threshold bistability of reentrant circulation ( t - bistability ) and in some cases , also phase - resetting interactions with the pacer . * complex patterns of spontaneous initiation and termination * * of reentrant circulation in a loop of cardiac tissue * h. sedaghat@xmath0 , m.a .
wood@xmath1 , j.w .
cain@xmath0 , c.k .
cheng@xmath2 c.m .
baumgarten@xmath3 , d.m .
chan@xmath0 @xmath0department of mathematics and the center for the study of biological complexity ; @xmath1department of internal medicine - cardiology and the pauley heart center ; @xmath2department of computer science and the center for the study of biological complexity ; @xmath3department of physiology and the pauley heart center ; virginia commonwealth university , richmond , virginia , 23284 - 2014 , usa * keywords * reentry ; loop ; pacer ; thresholds ; bistability ; difference equations |
Subsets and Splits